Merge tag 'perf-tools-for-v6.8-1-2024-01-09' of git://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools

Pull perf tools updates from Arnaldo Carvalho de Melo: "Add Namhyung Kim as tools/perf/ co-maintainer, we're taking turns processing patches, switching roles from perf-tools to perf-tools-next at each Linux release. Data profiling: - Associate samples that identify loads and stores with data structures. This uses events available on Intel, AMD and others and DWARF info: # To get memory access samples in kernel for 1 second (on Intel) $ perf mem record -a -K --ldlat=4 -- sleep 1 # Similar for the AMD (but it requires 6.3+ kernel for BPF filters) $ perf mem record -a --filter 'mem_op == load || mem_op == store, ip > 0x8000000000000000' -- sleep 1 Then, amongst several modes of post processing, one can do things like: $ perf report -s type,typeoff --hierarchy --group --stdio ... # # Samples: 10K of events 'cpu/mem-loads,ldlat=4/P, cpu/mem-stores/P, dummy:u' # Event count (approx.): 602758064 # # Overhead Data Type / Data Type Offset # ........................... ............................ # 26.09% 3.28% 0.00% long unsigned int 26.09% 3.28% 0.00% long unsigned int +0 (no field) 18.48% 0.73% 0.00% struct page 10.83% 0.02% 0.00% struct page +8 (lru.next) 3.90% 0.28% 0.00% struct page +0 (flags) 3.45% 0.06% 0.00% struct page +24 (mapping) 0.25% 0.28% 0.00% struct page +48 (_mapcount.counter) 0.02% 0.06% 0.00% struct page +32 (index) 0.02% 0.00% 0.00% struct page +52 (_refcount.counter) 0.02% 0.01% 0.00% struct page +56 (memcg_data) 0.00% 0.01% 0.00% struct page +16 (lru.prev) 15.37% 17.54% 0.00% (stack operation) 15.37% 17.54% 0.00% (stack operation) +0 (no field) 11.71% 50.27% 0.00% (unknown) 11.71% 50.27% 0.00% (unknown) +0 (no field) $ perf annotate --data-type ... Annotate type: 'struct cfs_rq' in [kernel.kallsyms] (13 samples): ============================================================================ samples offset size field 13 0 640 struct cfs_rq { 2 0 16 struct load_weight load { 2 0 8 unsigned long weight; 0 8 4 u32 inv_weight; }; 0 16 8 unsigned long runnable_weight; 0 24 4 unsigned int nr_running; 1 28 4 unsigned int h_nr_running; ... $ perf annotate --data-type=page --group Annotate type: 'struct page' in [kernel.kallsyms] (480 samples): event[0] = cpu/mem-loads,ldlat=4/P event[1] = cpu/mem-stores/P event[2] = dummy:u =================================================================================== samples offset size field 447 33 0 0 64 struct page { 108 8 0 0 8 long unsigned int flags; 319 13 0 8 40 union { 319 13 0 8 40 struct { 236 2 0 8 16 union { 236 2 0 8 16 struct list_head lru { 236 1 0 8 8 struct list_head* next; 0 1 0 16 8 struct list_head* prev; }; 236 2 0 8 16 struct { 236 1 0 8 8 void* __filler; 0 1 0 16 4 unsigned int mlock_count; }; 236 2 0 8 16 struct list_head buddy_list { 236 1 0 8 8 struct list_head* next; 0 1 0 16 8 struct list_head* prev; }; 236 2 0 8 16 struct list_head pcp_list { 236 1 0 8 8 struct list_head* next; 0 1 0 16 8 struct list_head* prev; }; }; 82 4 0 24 8 struct address_space* mapping; 1 7 0 32 8 union { 1 7 0 32 8 long unsigned int index; 1 7 0 32 8 long unsigned int share; }; 0 0 0 40 8 long unsigned int private; }; This uses the existing annotate code, calling objdump to do the disassembly, with improvements to avoid having this take too long, but longer term a switch to a disassembler library, possibly reusing code in the kernel will be pursued. This is the initial implementation, please use it and report impressions and bugs. Make sure the kernel-debuginfo packages match the running kernel. The 'perf report' phase for non short perf.data files may take a while. There is a great article about it on LWN: https://lwn.net/Articles/955709/ - "Data-type profiling for perf" One last test I did while writing this text, on a AMD Ryzen 5950X, using a distro kernel, while doing a simple 'find /' on an otherwise idle system resulted in: # uname -r 6.6.9-100.fc38.x86_64 # perf -vv | grep BPF_ bpf: [ on ] # HAVE_LIBBPF_SUPPORT bpf_skeletons: [ on ] # HAVE_BPF_SKEL # rpm -qa | grep kernel-debuginfo kernel-debuginfo-common-x86_64-6.6.9-100.fc38.x86_64 kernel-debuginfo-6.6.9-100.fc38.x86_64 # # perf mem record -a --filter 'mem_op == load || mem_op == store, ip > 0x8000000000000000' ^C[ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 2.199 MB perf.data (2913 samples) ] # # ls -la perf.data -rw-------. 1 root root 2346486 Jan 9 18:36 perf.data # perf evlist ibs_op// dummy:u # perf evlist -v ibs_op//: type: 11, size: 136, config: 0, { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|ADDR|CPU|PERIOD|IDENTIFIER|DATA_SRC|WEIGHT, read_format: ID, disabled: 1, inherit: 1, freq: 1, sample_id_all: 1 dummy:u: type: 1 (PERF_TYPE_SOFTWARE), size: 136, config: 0x9 (PERF_COUNT_SW_DUMMY), { sample_period, sample_freq }: 1, sample_type: IP|TID|TIME|ADDR|CPU|IDENTIFIER|DATA_SRC|WEIGHT, read_format: ID, inherit: 1, exclude_kernel: 1, exclude_hv: 1, mmap: 1, comm: 1, task: 1, mmap_data: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, ksymbol: 1, bpf_event: 1 # # perf report -s type,typeoff --hierarchy --group --stdio # Total Lost Samples: 0 # # Samples: 2K of events 'ibs_op//, dummy:u' # Event count (approx.): 1904553038 # # Overhead Data Type / Data Type Offset # ................... ............................ # 73.70% 0.00% (unknown) 73.70% 0.00% (unknown) +0 (no field) 3.01% 0.00% long unsigned int 3.00% 0.00% long unsigned int +0 (no field) 0.01% 0.00% long unsigned int +2 (no field) 2.73% 0.00% struct task_struct 1.71% 0.00% struct task_struct +52 (on_cpu) 0.38% 0.00% struct task_struct +2104 (rcu_read_unlock_special.b.blocked) 0.23% 0.00% struct task_struct +2100 (rcu_read_lock_nesting) 0.14% 0.00% struct task_struct +2384 () 0.06% 0.00% struct task_struct +3096 (signal) 0.05% 0.00% struct task_struct +3616 (cgroups) 0.05% 0.00% struct task_struct +2344 (active_mm) 0.02% 0.00% struct task_struct +46 (flags) 0.02% 0.00% struct task_struct +2096 (migration_disabled) 0.01% 0.00% struct task_struct +24 (__state) 0.01% 0.00% struct task_struct +3956 (mm_cid_active) 0.01% 0.00% struct task_struct +1048 (cpus_ptr) 0.01% 0.00% struct task_struct +184 (se.group_node.next) 0.01% 0.00% struct task_struct +20 (thread_info.cpu) 0.00% 0.00% struct task_struct +104 (on_rq) 0.00% 0.00% struct task_struct +2456 (pid) 1.36% 0.00% struct module 0.59% 0.00% struct module +952 (kallsyms) 0.42% 0.00% struct module +0 (state) 0.23% 0.00% struct module +8 (list.next) 0.12% 0.00% struct module +216 (syms) 0.95% 0.00% struct inode 0.41% 0.00% struct inode +40 (i_sb) 0.22% 0.00% struct inode +0 (i_mode) 0.06% 0.00% struct inode +76 (i_rdev) 0.06% 0.00% struct inode +56 (i_security) <SNIP> perf top/report: - Don't ignore job control, allowing control+Z + bg to work. - Add s390 raw data interpretation for PAI (Processor Activity Instrumentation) counters. perf archive: - Add new option '--all' to pack perf.data with DSOs. - Add new option '--unpack' to expand tarballs. Initialization speedups: - Lazily initialize zstd streams to save memory when not using it. - Lazily allocate/size mmap event copy. - Lazy load kernel symbols in 'perf record'. - Be lazier in allocating lost samples buffer in 'perf record'. - Don't synthesize BPF events when disabled via the command line (perf record --no-bpf-event). Assorted improvements: - Show note on AMD systems that the :p, :pp, :ppp and :P are all the same, as IBS (Instruction Based Sampling) is used and it is inherentely precise, not having levels of precision like in Intel systems. - When 'cycles' isn't available, fall back to the "task-clock" event when not system wide, not to 'cpu-clock'. - Add --debug-file option to redirect debug output, e.g.: $ perf --debug-file /tmp/perf.log record -v true - Shrink 'struct map' to under one cacheline by avoiding function pointers for selecting if addresses are identity or DSO relative, and using just a byte for some boolean struct members. - Resolve the arch specific strerrno just once to use in perf_env__arch_strerrno(). - Reduce memory for recording PERF_RECORD_LOST_SAMPLES event. Assorted fixes: - Fix the default 'perf top' usage on Intel hybrid systems, now it starts with a browser showing the number of samples for Efficiency (cpu_atom/cycles/P) and Performance (cpu_core/cycles/P). This behaviour is similar on ARM64, with its respective set of big.LITTLE processors. - Fix segfault on build_mem_topology() error path. - Fix 'perf mem' error on hybrid related to availability of mem event in a PMU. - Fix missing reference count gets (map, maps) in the db-export code. - Avoid recursively taking env->bpf_progs.lock in the 'perf_env' code. - Use the newly introduced maps__for_each_map() to add missing locking around iteration of 'struct map' entries. - Parse NOTE segments until the build id is found, don't stop on the first one, ELF files may have several such NOTE segments. - Remove 'egrep' usage, its deprecated, use 'grep -E' instead. - Warn first about missing libelf, not libbpf, that depends on libelf. - Use alternative to 'find ... -printf' as this isn't supported in busybox. - Address python 3.6 DeprecationWarning for string scapes. - Fix memory leak in uniq() in libsubcmd. - Fix man page formatting for 'perf lock' - Fix some spelling mistakes. perf tests: - Fail shell tests that needs some symbol in perf itself if it is stripped. These tests check if a symbol is resolved, if some hot function is indeed detected by profiling, etc. - The 'perf test sigtrap' test is currently failing on PREEMPT_RT, skip it if sleeping spinlocks are detected (using BTF) and point to the mailing list discussion about it. This test is also being skipped on several architectures (powerpc, s390x, arm and aarch64) due to other pending issues with intruction breakpoints. - Adjust test case perf record offcpu profiling tests for s390. - Fix 'Setup struct perf_event_attr' fails on s390 on z/VM guest, addressing issues caused by the fallback from cycles to task-clock done in this release. - Fix mask for VG register in the user-regs test. - Use shellcheck on 'perf test' shell scripts automatically to make sure changes don't introduce things it flags as problematic. - Add option to change objdump binary and allow it to be set via 'perf config'. - Add basic 'perf script', 'perf list --json" and 'perf diff' tests. - Basic branch counter support. - Make DSO tests a suite rather than individual. - Remove atomics from test_loop to avoid test failures. - Fix call chain match on powerpc for the record+probe_libc_inet_pton test. - Improve Intel hybrid tests. Vendor event files (JSON): powerpc: - Update datasource event name to fix duplicate events on IBM's Power10. - Add PVN for HX-C2000 CPU with Power8 Architecture. Intel: - Alderlake/rocketlake metric fixes. - Update emeraldrapids events to v1.02. - Update icelakex events to v1.23. - Update sapphirerapids events to v1.17. - Add skx, clx, icx and spr upi bandwidth metric. AMD: - Add Zen 4 memory controller events. RISC-V: - Add StarFive Dubhe-80 and Dubhe-90 JSON files. https://www.starfivetech.com/en/site/cpu-u - Add T-HEAD C9xx JSON file. https://github.com/riscv-software-src/opensbi/blob/master/docs/platform/thead-c9xx.md ARM64: - Remove UTF-8 characters from cmn.json, that were causing build failure in some distros. - Add core PMU events and metrics for Ampere One X. - Rename Ampere One's BPU_FLUSH_MEM_FAULT to GPC_FLUSH_MEM_FAULT libperf: - Rename several perf_cpu_map constructor names to clarify what they really do. - Ditto for some other methods, coping with some issues in their semantics, like perf_cpu_map__empty() -> perf_cpu_map__has_any_cpu_or_is_empty(). - Document perf_cpu_map__nr()'s behavior perf stat: - Exit if parse groups fails. - Combine the -A/--no-aggr and --no-merge options. - Fix help message for --metric-no-threshold option. Hardware tracing: ARM64 CoreSight: - Bump minimum OpenCSD version to ensure a bugfix is present. - Add 'T' itrace option for timestamp trace - Set start vm addr of exectable file to 0 and don't ignore first sample on the arm-cs-trace-disasm.py 'perf script'" * tag 'perf-tools-for-v6.8-1-2024-01-09' of git://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools: (179 commits) MAINTAINERS: Add Namhyung as tools/perf/ co-maintainer perf test: test case 'Setup struct perf_event_attr' fails on s390 on z/vm perf db-export: Fix missing reference count get in call_path_from_sample() perf tests: Add perf script test libsubcmd: Fix memory leak in uniq() perf TUI: Don't ignore job control perf vendor events intel: Update sapphirerapids events to v1.17 perf vendor events intel: Update icelakex events to v1.23 perf vendor events intel: Update emeraldrapids events to v1.02 perf vendor events intel: Alderlake/rocketlake metric fixes perf x86 test: Add hybrid test for conflicting legacy/sysfs event perf x86 test: Update hybrid expectations perf vendor events amd: Add Zen 4 memory controller events perf stat: Fix hard coded LL miss units perf record: Reduce memory for recording PERF_RECORD_LOST_SAMPLES event perf env: Avoid recursively taking env->bpf_progs.lock perf annotate: Add --insn-stat option for debugging perf annotate: Add --type-stat option for debugging perf annotate: Support event group display perf annotate: Add --data-type option ...

Linus Torvalds committed 2 years ago

Commit 9d64bf42 parents 57f22c8 d988c9f249 files · +7843 −2141Browse files

Unified · Split

249 files

MAINTAINERS +1 −1
tools/build/Makefile.feature +1 −0
tools/build/feature/Makefile +4 −0
tools/build/feature/test-dwarf_getcfi.c +9 −0
tools/build/feature/test-libopencsd.c +2 −2
tools/include/uapi/linux/perf_event.h +13 −0
tools/lib/api/fs/fs.c +29 −51
tools/lib/api/io.h +9 −3
tools/lib/perf/Documentation/examples/sampling.c +1 −1
tools/lib/perf/Documentation/libperf-sampling.txt +1 −1
tools/lib/perf/Documentation/libperf.txt +2 −2
tools/lib/perf/cpumap.c +38 −27
tools/lib/perf/evlist.c +3 −3
tools/lib/perf/evsel.c +1 −1
tools/lib/perf/include/internal/mmap.h +2 −1
tools/lib/perf/include/perf/cpumap.h +35 −5
tools/lib/perf/libperf.map +3 −3
tools/lib/perf/mmap.c +17 −3
tools/lib/perf/tests/test-cpumap.c +2 −2
tools/lib/perf/tests/test-evlist.c +3 −3
tools/lib/perf/tests/test-evsel.c +1 −1
tools/lib/subcmd/help.c +14 −4
tools/perf/.gitignore +4 −0
tools/perf/Documentation/itrace.txt +1 −0
tools/perf/Documentation/perf-annotate.txt +11 −0
tools/perf/Documentation/perf-config.txt +2 −2
tools/perf/Documentation/perf-list.txt +7 −5
tools/perf/Documentation/perf-lock.txt +1 −1
tools/perf/Documentation/perf-record.txt +4 −0
tools/perf/Documentation/perf-report.txt +3 −0
tools/perf/Documentation/perf-stat.txt +28 −24
tools/perf/Documentation/perf.txt +3 −0
tools/perf/Makefile.config +9 −4
tools/perf/Makefile.perf +11 −2
tools/perf/arch/arm/util/cs-etm.c +8 −8
tools/perf/arch/arm64/util/arm-spe.c +2 −2
tools/perf/arch/arm64/util/header.c +1 −1
tools/perf/arch/loongarch/annotate/instructions.c +3 −3
tools/perf/arch/x86/tests/hybrid.c +30 −7
tools/perf/arch/x86/util/dwarf-regs.c +38 −0
tools/perf/arch/x86/util/event.c +58 −45
tools/perf/arch/x86/util/intel-bts.c +2 −2
tools/perf/arch/x86/util/intel-pt.c +5 −5
tools/perf/bench/epoll-ctl.c +1 −1
tools/perf/bench/epoll-wait.c +1 −1
tools/perf/bench/futex-hash.c +1 −1
tools/perf/bench/futex-lock-pi.c +1 −1
tools/perf/bench/futex-requeue.c +1 −1
tools/perf/bench/futex-wake-parallel.c +1 −1
tools/perf/bench/futex-wake.c +1 −1
tools/perf/bench/sched-seccomp-notify.c +1 −1
tools/perf/builtin-annotate.c +272 −28
tools/perf/builtin-c2c.c +1 −1
tools/perf/builtin-ftrace.c +1 −1
tools/perf/builtin-inject.c +6 −0
tools/perf/builtin-lock.c +3 −1
tools/perf/builtin-record.c +45 −48
tools/perf/builtin-report.c +74 −39
tools/perf/builtin-stat.c +19 −14
tools/perf/builtin-top.c +26 −26
tools/perf/builtin-trace.c +2 −4
tools/perf/perf-archive.sh +80 −6
tools/perf/perf.c +30 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereone/core-imp-def.json +1 −1
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/branch.json +125 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/bus.json +20 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/cache.json +206 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/core-imp-def.json +464 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/exception.json +47 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/instruction.json +128 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/intrinsic.json +14 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/memory.json +41 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/metrics.json +442 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/mmu.json +170 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/pipeline.json +41 −0
tools/perf/pmu-events/arch/arm64/ampere/ampereonex/spe.json +14 −0
tools/perf/pmu-events/arch/arm64/arm/cmn/sys/cmn.json +1 −1
tools/perf/pmu-events/arch/arm64/mapfile.csv +1 −0
tools/perf/pmu-events/arch/powerpc/mapfile.csv +1 −2
tools/perf/pmu-events/arch/powerpc/power10/datasource.json +14 −4
tools/perf/pmu-events/arch/riscv/mapfile.csv +2 −0
tools/perf/pmu-events/arch/riscv/starfive/dubhe-80/common.json +172 −0
tools/perf/pmu-events/arch/riscv/starfive/dubhe-80/firmware.json +68 −0
tools/perf/pmu-events/arch/riscv/thead/c900-legacy/cache.json +67 −0
tools/perf/pmu-events/arch/riscv/thead/c900-legacy/firmware.json +68 −0
tools/perf/pmu-events/arch/riscv/thead/c900-legacy/instruction.json +72 −0
tools/perf/pmu-events/arch/riscv/thead/c900-legacy/microarch.json +80 −0
tools/perf/pmu-events/arch/x86/alderlake/adl-metrics.json +8 −7
tools/perf/pmu-events/arch/x86/amdzen4/memory-controller.json +101 −0
tools/perf/pmu-events/arch/x86/amdzen4/recommended.json +84 −0
tools/perf/pmu-events/arch/x86/cascadelakex/clx-metrics.json +6 −0
tools/perf/pmu-events/arch/x86/emeraldrapids/floating-point.json +24 −3
tools/perf/pmu-events/arch/x86/emeraldrapids/pipeline.json +1 −17
tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-interconnect.json +4 −4
tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-io.json +30 −0
tools/perf/pmu-events/arch/x86/icelakex/icx-metrics.json +6 −0
tools/perf/pmu-events/arch/x86/icelakex/other.json +1 −1
tools/perf/pmu-events/arch/x86/icelakex/pipeline.json +1 −1
tools/perf/pmu-events/arch/x86/icelakex/uncore-interconnect.json +3 −3
tools/perf/pmu-events/arch/x86/mapfile.csv +3 −3
tools/perf/pmu-events/arch/x86/rocketlake/rkl-metrics.json +1 −1
tools/perf/pmu-events/arch/x86/sapphirerapids/floating-point.json +24 −3
tools/perf/pmu-events/arch/x86/sapphirerapids/pipeline.json +1 −1
tools/perf/pmu-events/arch/x86/sapphirerapids/spr-metrics.json +6 −0
tools/perf/pmu-events/arch/x86/sapphirerapids/uncore-interconnect.json +4 −4
tools/perf/pmu-events/arch/x86/sapphirerapids/uncore-io.json +30 −0
tools/perf/pmu-events/arch/x86/skylakex/skx-metrics.json +6 −0
tools/perf/pmu-events/jevents.py +3 −1
tools/perf/scripts/python/arm-cs-trace-disasm.py +16 −14
tools/perf/scripts/python/compaction-times.py +1 −1
tools/perf/scripts/python/exported-sql-viewer.py +2 −2
tools/perf/tests/Build +14 −0
tools/perf/tests/attr.c +1 −1
tools/perf/tests/attr/base-record +1 −1
tools/perf/tests/attr/test-record-user-regs-no-sve-aarch64 +1 −1
tools/perf/tests/attr/test-record-user-regs-sve-aarch64 +1 −1
tools/perf/tests/builtin-test.c +15 −2
tools/perf/tests/code-reading.c +25 −55
tools/perf/tests/cpumap.c +1 −1
tools/perf/tests/dso-data.c +12 −3
tools/perf/tests/keep-tracking.c +1 −1
tools/perf/tests/make +1 −1
tools/perf/tests/maps.c +38 −23
tools/perf/tests/mmap-basic.c +1 −1
tools/perf/tests/openat-syscall-all-cpus.c +1 −1
tools/perf/tests/parse-events.c +187 −73
tools/perf/tests/perf-time-to-tsc.c +1 −1
tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c +0 −1
tools/perf/tests/shell/coresight/thread_loop/thread_loop.c +0 −1
tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c +0 −1
tools/perf/tests/shell/diff.sh +108 −0
tools/perf/tests/shell/lib/perf_has_symbol.sh +21 −0
tools/perf/tests/shell/lib/setup_python.sh +16 −0
tools/perf/tests/shell/list.sh +19 −0
tools/perf/tests/shell/pipe_test.sh +8 −1
tools/perf/tests/shell/record+probe_libc_inet_pton.sh +4 −1
tools/perf/tests/shell/record.sh +37 −1
tools/perf/tests/shell/record_offcpu.sh +2 −2
tools/perf/tests/shell/script.sh +66 −0
tools/perf/tests/shell/stat+json_output.sh +3 −13
tools/perf/tests/shell/stat_all_pmu.sh +1 −1
tools/perf/tests/shell/stat_metrics_values.sh +4 −10
tools/perf/tests/shell/test_arm_callgraph_fp.sh +6 −0
tools/perf/tests/shell/test_brstack.sh +6 −0
tools/perf/tests/shell/test_data_symbol.sh +14 −2
tools/perf/tests/shell/test_perf_data_converter_json.sh +3 −10
tools/perf/tests/sigtrap.c +84 −22
tools/perf/tests/sw-clock.c +1 −1
tools/perf/tests/switch-tracking.c +1 −1
tools/perf/tests/task-exit.c +1 −1
tools/perf/tests/tests.h +1 −0
tools/perf/tests/topology.c +1 −1
tools/perf/tests/vmlinux-kallsyms.c +98 −83
tools/perf/tests/workloads/thloop.c +1 −3
tools/perf/trace/beauty/arch_errno_names.sh +6 −4
tools/perf/trace/beauty/beauty.h +0 −2
tools/perf/trace/beauty/prctl_option.sh +2 −2
tools/perf/trace/beauty/socket.sh +2 −2
tools/perf/ui/browsers/annotate.c +48 −56
tools/perf/ui/browsers/hists.c +14 −22
tools/perf/ui/browsers/hists.h +0 −2
tools/perf/ui/browsers/scripts.c +1 −0
tools/perf/ui/gtk/annotate.c +2 −4
tools/perf/ui/gtk/gtk.h +0 −2
tools/perf/ui/tui/setup.c +22 −0
tools/perf/util/Build +2 −0
tools/perf/util/annotate-data.c +405 −0
tools/perf/util/annotate-data.h +143 −0
tools/perf/util/annotate.c +522 −165
tools/perf/util/annotate.h +100 −44
tools/perf/util/auxtrace.c +5 −2
tools/perf/util/auxtrace.h +3 −0
tools/perf/util/block-info.c +4 −6
tools/perf/util/block-info.h +1 −2
tools/perf/util/block-range.c +6 −1
tools/perf/util/bpf-event.c +7 −4
tools/perf/util/bpf-event.h +6 −6
tools/perf/util/bpf_counter.c +1 −1
tools/perf/util/bpf_lock_contention.c +1 −1
tools/perf/util/compress.h +5 −2
tools/perf/util/cpumap.c +1 −1
tools/perf/util/cputopo.c +1 −1
tools/perf/util/cs-etm.c +18 −3
tools/perf/util/db-export.c +2 −2
tools/perf/util/debug.c +15 −7
tools/perf/util/debug.h +1 −0
tools/perf/util/debuginfo.c +205 −0
tools/perf/util/debuginfo.h +64 −0
tools/perf/util/dso.c +4 −0
tools/perf/util/dso.h +2 −0
tools/perf/util/dwarf-aux.c +232 −17
tools/perf/util/dwarf-aux.h +49 −2
tools/perf/util/dwarf-regs.c +34 −0
tools/perf/util/env.c +64 −18
tools/perf/util/env.h +17 −0
tools/perf/util/event.c +2 −2
tools/perf/util/evlist.c +32 −2
tools/perf/util/evlist.h +1 −0
tools/perf/util/evsel.c +45 −10
tools/perf/util/evsel.h +3 −1
tools/perf/util/genelf.c +3 −3
tools/perf/util/header.c +33 −10
tools/perf/util/hisi-ptt.c +1 −0
tools/perf/util/hist.h +11 −17
tools/perf/util/include/dwarf-regs.h +19 −0
tools/perf/util/machine.c +41 −34
tools/perf/util/map.c +2 −18
tools/perf/util/map.h +43 −40
tools/perf/util/maps.c +445 −83
tools/perf/util/maps.h +19 −29
tools/perf/util/mem-events.c +14 −11
tools/perf/util/mmap.c +2 −3
tools/perf/util/mmap.h +0 −1
tools/perf/util/parse-branch-options.c +1 −0
tools/perf/util/parse-events.c +38 −14
tools/perf/util/perf_api_probe.c +2 −2
tools/perf/util/perf_event_attr_fprintf.c +1 −0
tools/perf/util/pmu.c +6 −2
tools/perf/util/pmu.h +2 −1
tools/perf/util/probe-event.c +26 −14
tools/perf/util/probe-finder.c +5 −196
tools/perf/util/probe-finder.h +1 −18
tools/perf/util/record.c +2 −2
tools/perf/util/s390-cpumcf-kernel.h +2 −0
tools/perf/util/s390-sample-raw.c +104 −10
tools/perf/util/sample.h +1 −0
tools/perf/util/scripting-engines/trace-event-perl.c +3 −0
tools/perf/util/scripting-engines/trace-event-python.c +15 −2
tools/perf/util/session.c +18 −2
tools/perf/util/sort.c +207 −9
tools/perf/util/sort.h +7 −0
tools/perf/util/stat-display.c +1 −1
tools/perf/util/stat-shadow.c +1 −1
tools/perf/util/stat.c +2 −2
tools/perf/util/stat.h +0 −1
tools/perf/util/symbol-elf.c +2 −4
tools/perf/util/symbol-minimal.c +6 −4
tools/perf/util/symbol.c +29 −277
tools/perf/util/symbol.h +0 −1
tools/perf/util/symbol_conf.h +5 −1
tools/perf/util/synthetic-events.c +67 −51
tools/perf/util/thread.c +20 −22
tools/perf/util/thread.h +14 −0
tools/perf/util/top.c +4 −5
tools/perf/util/top.h +0 −1
tools/perf/util/unwind-libdw.c +17 −4
tools/perf/util/unwind-libunwind-local.c +24 −10
tools/perf/util/vdso.c +23 −12
tools/perf/util/zstd.c +33 −30

MAINTAINERS+1−1

@@ -17140,10 +17140,10 @@ PERFORMANCE EVENTS SUBSYSTEM · 17139 unmodified lines
17140	17140	M: Peter Zijlstra <peterz@infradead.org>
17141	17141	M: Ingo Molnar <mingo@redhat.com>
17142	17142	M: Arnaldo Carvalho de Melo <acme@kernel.org>
	17143	+M: Namhyung Kim <namhyung@kernel.org>
17143	17144	R: Mark Rutland <mark.rutland@arm.com>
17144	17145	R: Alexander Shishkin <alexander.shishkin@linux.intel.com>
17145	17146	R: Jiri Olsa <jolsa@kernel.org>
17146		−R: Namhyung Kim <namhyung@kernel.org>
17147	17147	R: Ian Rogers <irogers@google.com>
17148	17148	R: Adrian Hunter <adrian.hunter@intel.com>
17149	17149	L: linux-perf-users@vger.kernel.org

tools/build/Makefile.feature+1−0

@@ -32,6 +32,7 @@ FEATURE_TESTS_BASIC := \ · 31 unmodified lines
32	32	backtrace \
33	33	dwarf \
34	34	dwarf_getlocations \
	35	+ dwarf_getcfi \
35	36	eventfd \
36	37	fortify-source \
37	38	get_current_dir_name \

tools/build/feature/Makefile+4−0

@@ -7,6 +7,7 @@ FILES= \ · 6 unmodified lines
7	7	test-bionic.bin \
8	8	test-dwarf.bin \
9	9	test-dwarf_getlocations.bin \
	10	+ test-dwarf_getcfi.bin \
10	11	test-eventfd.bin \
11	12	test-fortify-source.bin \
12	13	test-get_current_dir_name.bin \
@@ -154,6 +155,9 @@ $(OUTPUT)test-dwarf.bin: · 141 unmodified lines
154	155	$(OUTPUT)test-dwarf_getlocations.bin:
155	156	$(BUILD) $(DWARFLIBS)
156	157
	158	+$(OUTPUT)test-dwarf_getcfi.bin:
	159	+ $(BUILD) $(DWARFLIBS)
	160	+
157	161	$(OUTPUT)test-libelf-getphdrnum.bin:
158	162	$(BUILD) -lelf
159	163

tools/build/feature/test-dwarf_getcfi.c+9−0

@@ -0,0 +1,9 @@
	1	+// SPDX-License-Identifier: GPL-2.0
	2	+#include <stdio.h>
	3	+#include <elfutils/libdw.h>
	4	+
	5	+int main(void)
	6	+{
	7	+ Dwarf *dwarf = NULL;
	8	+ return dwarf_getcfi(dwarf) == NULL;
	9	+}

tools/build/feature/test-libopencsd.c+2−2

@@ -4,9 +4,9 @@ · 3 unmodified lines
4	4	/*
5	5	* Check OpenCSD library version is sufficient to provide required features
6	6	*/
7		−#define OCSD_MIN_VER ((1 << 16) \| (1 << 8) \| (1))
	7	+#define OCSD_MIN_VER ((1 << 16) \| (2 << 8) \| (1))
8	8	#if !defined(OCSD_VER_NUM) \|\| (OCSD_VER_NUM < OCSD_MIN_VER)
9		−#error "OpenCSD >= 1.1.1 is required"
	9	+#error "OpenCSD >= 1.2.1 is required"
10	10	#endif
11	11
12	12	int main(void)

tools/include/uapi/linux/perf_event.h+13−0

@@ -204,6 +204,8 @@ enum perf_branch_sample_type_shift { · 203 unmodified lines
204	204
205	205	PERF_SAMPLE_BRANCH_PRIV_SAVE_SHIFT = 18, /* save privilege mode */
206	206
	207	+ PERF_SAMPLE_BRANCH_COUNTERS_SHIFT = 19, /* save occurrences of events on a branch */
	208	+
207	209	PERF_SAMPLE_BRANCH_MAX_SHIFT /* non-ABI */
208	210	};
209	211
@@ -235,6 +237,8 @@ enum perf_branch_sample_type { · 25 unmodified lines
235	237
236	238	PERF_SAMPLE_BRANCH_PRIV_SAVE = 1U << PERF_SAMPLE_BRANCH_PRIV_SAVE_SHIFT,
237	239
	240	+ PERF_SAMPLE_BRANCH_COUNTERS = 1U << PERF_SAMPLE_BRANCH_COUNTERS_SHIFT,
	241	+
238	242	PERF_SAMPLE_BRANCH_MAX = 1U << PERF_SAMPLE_BRANCH_MAX_SHIFT,
239	243	};
240	244
@@ -982,6 +986,12 @@ enum perf_event_type { · 741 unmodified lines
982	986	* { u64 nr;
983	987	* { u64 hw_idx; } && PERF_SAMPLE_BRANCH_HW_INDEX
984	988	* { u64 from, to, flags } lbr[nr];
	989	+ * #
	990	+ * # The format of the counters is decided by the
	991	+ * # "branch_counter_nr" and "branch_counter_width",
	992	+ * # which are defined in the ABI.
	993	+ * #
	994	+ * { u64 counters; } cntr[nr] && PERF_SAMPLE_BRANCH_COUNTERS
985	995	* } && PERF_SAMPLE_BRANCH_STACK
986	996	*
987	997	* { u64 abi; # enum perf_sample_regs_abi
@@ -1427,6 +1437,9 @@ struct perf_branch_entry { · 439 unmodified lines
1427	1437	reserved:31;
1428	1438	};
1429	1439
	1440	+/* Size of used info bits in struct perf_branch_entry */
	1441	+#define PERF_BRANCH_ENTRY_INFO_BITS_MAX 33
	1442	+
1430	1443	union perf_sample_weight {
1431	1444	__u64 full;
1432	1445	#if defined(__LITTLE_ENDIAN_BITFIELD)

tools/lib/api/fs/fs.c+29−51

@@ -16,6 +16,7 @@ · 15 unmodified lines
16	16	#include <sys/mount.h>
17	17
18	18	#include "fs.h"
	19	+#include "../io.h"
19	20	#include "debug-internal.h"
20	21
21	22	#define _STR(x) #x
@@ -344,53 +345,24 @@ int filename__read_ull(const char filename, unsigned long long value) · 322 unmodified lines
344	345	return filename__read_ull_base(filename, value, 0);
345	346	}
346	347
347		−#define STRERR_BUFSIZE 128 /* For the buffer size of strerror_r */
348		−
349	348	int filename__read_str(const char filename, char buf, size_t sizep)
350	349	{
351		− size_t size = 0, alloc_size = 0;
352		− void bf = NULL, nbf;
353		− int fd, n, err = 0;
354		− char sbuf[STRERR_BUFSIZE];
	350	+ struct io io;
	351	+ char bf[128];
	352	+ int err;
355	353
356		− fd = open(filename, O_RDONLY);
357		− if (fd < 0)
	354	+ io.fd = open(filename, O_RDONLY);
	355	+ if (io.fd < 0)
358	356	return -errno;
359		−
360		− do {
361		− if (size == alloc_size) {
362		− alloc_size += BUFSIZ;
363		− nbf = realloc(bf, alloc_size);
364		− if (!nbf) {
365		− err = -ENOMEM;
366		− break;
367		− }
368		−
369		− bf = nbf;
370		− }
371		−
372		− n = read(fd, bf + size, alloc_size - size);
373		− if (n < 0) {
374		− if (size) {
375		− pr_warn("read failed %d: %s\n", errno,
376		− strerror_r(errno, sbuf, sizeof(sbuf)));
377		− err = 0;
378		− } else
379		− err = -errno;
380		−
381		− break;
382		− }
383		−
384		− size += n;
385		− } while (n > 0);
386		−
387		− if (!err) {
388		− *sizep = size;
389		− *buf = bf;
	357	+ io__init(&io, io.fd, bf, sizeof(bf));
	358	+ *buf = NULL;
	359	+ err = io__getdelim(&io, buf, sizep, /delim=/-1);
	360	+ if (err < 0) {
	361	+ free(*buf);
	362	+ *buf = NULL;
390	363	} else
391		− free(bf);
392		−
393		− close(fd);
	364	+ err = 0;
	365	+ close(io.fd);
394	366	return err;
395	367	}
396	368
@@ -475,15 +447,22 @@ int sysfs__read_str(const char entry, char buf, size_t sizep) · 78 unmodified lines
475	447
476	448	int sysfs__read_bool(const char entry, bool value)
477	449	{
478		− char *buf;
479		− size_t size;
480		− int ret;
	450	+ struct io io;
	451	+ char bf[16];
	452	+ int ret = 0;
	453	+ char path[PATH_MAX];
	454	+ const char *sysfs = sysfs__mountpoint();
	455	+
	456	+ if (!sysfs)
	457	+ return -1;
481	458
482		− ret = sysfs__read_str(entry, &buf, &size);
483		− if (ret < 0)
484		− return ret;
	459	+ snprintf(path, sizeof(path), "%s/%s", sysfs, entry);
	460	+ io.fd = open(path, O_RDONLY);
	461	+ if (io.fd < 0)
	462	+ return -errno;
485	463
486		− switch (buf[0]) {
	464	+ io__init(&io, io.fd, bf, sizeof(bf));
	465	+ switch (io__get_char(&io)) {
487	466	case '1':
488	467	case 'y':
489	468	case 'Y':
@@ -497,8 +476,7 @@ int sysfs__read_bool(const char entry, bool value) · 7 unmodified lines
497	476	default:
498	477	ret = -1;
499	478	}
500		−
501		− free(buf);
	479	+ close(io.fd);
502	480
503	481	return ret;
504	482	}

tools/lib/api/io.h+9−3

@@ -12,6 +12,7 @@ · 11 unmodified lines
12	12	#include <stdlib.h>
13	13	#include <string.h>
14	14	#include <unistd.h>
	15	+#include <linux/types.h>
15	16
16	17	struct io {
17	18	/* File descriptor being read/ */
@@ -140,8 +141,8 @@ static inline int io__get_dec(struct io io, __u64 dec) · 122 unmodified lines
140	141	}
141	142	}
142	143
143		−/* Read up to and including the first newline following the pattern of getline. */
144		−static inline ssize_t io__getline(struct io io, char line_out, size_t line_len_out)
	144	+/* Read up to and including the first delim. */
	145	+static inline ssize_t io__getdelim(struct io io, char line_out, size_t line_len_out, int delim)
145	146	{
146	147	char buf[128];
147	148	int buf_pos = 0;
@@ -151,7 +152,7 @@ static inline ssize_t io__getline(struct io io, char line_out, size_t line_l · 3 unmodified lines
151	152
152	153	/* TODO: reuse previously allocated memory. */
153	154	free(*line_out);
154		− while (ch != '\n') {
	155	+ while (ch != delim) {
155	156	ch = io__get_char(io);
156	157
157	158	if (ch < 0)
@@ -184,4 +185,9 @@ err_out: · 26 unmodified lines
184	185	return -ENOMEM;
185	186	}
186	187
	188	+static inline ssize_t io__getline(struct io io, char line_out, size_t line_len_out)
	189	+{
	190	+ return io__getdelim(io, line_out, line_len_out, /delim=/'\n');
	191	+}
	192	+
187	193	#endif /* __API_IO__ */

tools/lib/perf/Documentation/examples/sampling.c+1−1

@@ -39,7 +39,7 @@ int main(int argc, char **argv) · 38 unmodified lines
39	39
40	40	libperf_init(libperf_print);
41	41
42		− cpus = perf_cpu_map__new(NULL);
	42	+ cpus = perf_cpu_map__new_online_cpus();
43	43	if (!cpus) {
44	44	fprintf(stderr, "failed to create cpus\n");
45	45	return -1;

tools/lib/perf/Documentation/libperf-sampling.txt+1−1

@@ -97,7 +97,7 @@ In this case we will monitor all the available CPUs: · 96 unmodified lines
97	97
98	98	[source,c]
99	99	--
100		− 42 cpus = perf_cpu_map__new(NULL);
	100	+ 42 cpus = perf_cpu_map__new_online_cpus();
101	101	43 if (!cpus) {
102	102	44 fprintf(stderr, "failed to create cpus\n");
103	103	45 return -1;

tools/lib/perf/Documentation/libperf.txt+2−2

@@ -37,7 +37,7 @@ SYNOPSIS · 36 unmodified lines
37	37
38	38	struct perf_cpu_map;
39	39
40		− struct perf_cpu_map *perf_cpu_map__dummy_new(void);
	40	+ struct perf_cpu_map *perf_cpu_map__new_any_cpu(void);
41	41	struct perf_cpu_map perf_cpu_map__new(const char cpu_list);
42	42	struct perf_cpu_map perf_cpu_map__read(FILE file);
43	43	struct perf_cpu_map perf_cpu_map__get(struct perf_cpu_map map);
@@ -46,7 +46,7 @@ SYNOPSIS · 2 unmodified lines
46	46	void perf_cpu_map__put(struct perf_cpu_map *map);
47	47	int perf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx);
48	48	int perf_cpu_map__nr(const struct perf_cpu_map *cpus);
49		− bool perf_cpu_map__empty(const struct perf_cpu_map *map);
	49	+ bool perf_cpu_map__has_any_cpu_or_is_empty(const struct perf_cpu_map *map);
50	50	int perf_cpu_map__max(struct perf_cpu_map *map);
51	51	bool perf_cpu_map__has(const struct perf_cpu_map *map, int cpu);
52	52

tools/lib/perf/cpumap.c+38−27

@@ -9,6 +9,7 @@ · 8 unmodified lines
9	9	#include <unistd.h>
10	10	#include <ctype.h>
11	11	#include <limits.h>
	12	+#include "internal.h"
12	13
13	14	void perf_cpu_map__set_nr(struct perf_cpu_map *map, int nr_cpus)
14	15	{
@@ -27,7 +28,7 @@ struct perf_cpu_map *perf_cpu_map__alloc(int nr_cpus) · 12 unmodified lines
27	28	return result;
28	29	}
29	30
30		−struct perf_cpu_map *perf_cpu_map__dummy_new(void)
	31	+struct perf_cpu_map *perf_cpu_map__new_any_cpu(void)
31	32	{
32	33	struct perf_cpu_map *cpus = perf_cpu_map__alloc(1);
33	34
@@ -66,15 +67,21 @@ void perf_cpu_map__put(struct perf_cpu_map *map) · 32 unmodified lines
66	67	}
67	68	}
68	69
69		−static struct perf_cpu_map *cpu_map__default_new(void)
	70	+static struct perf_cpu_map *cpu_map__new_sysconf(void)
70	71	{
71	72	struct perf_cpu_map *cpus;
72		− int nr_cpus;
	73	+ int nr_cpus, nr_cpus_conf;
73	74
74	75	nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
75	76	if (nr_cpus < 0)
76	77	return NULL;
77	78
	79	+ nr_cpus_conf = sysconf(_SC_NPROCESSORS_CONF);
	80	+ if (nr_cpus != nr_cpus_conf) {
	81	+ pr_warning("Number of online CPUs (%d) differs from the number configured (%d) the CPU map will only cover the first %d CPUs.",
	82	+ nr_cpus, nr_cpus_conf, nr_cpus);
	83	+ }
	84	+
78	85	cpus = perf_cpu_map__alloc(nr_cpus);
79	86	if (cpus != NULL) {
80	87	int i;
@@ -86,9 +93,27 @@ static struct perf_cpu_map *cpu_map__default_new(void) · 5 unmodified lines
86	93	return cpus;
87	94	}
88	95
89		−struct perf_cpu_map *perf_cpu_map__default_new(void)
	96	+static struct perf_cpu_map *cpu_map__new_sysfs_online(void)
90	97	{
91		− return cpu_map__default_new();
	98	+ struct perf_cpu_map *cpus = NULL;
	99	+ FILE *onlnf;
	100	+
	101	+ onlnf = fopen("/sys/devices/system/cpu/online", "r");
	102	+ if (onlnf) {
	103	+ cpus = perf_cpu_map__read(onlnf);
	104	+ fclose(onlnf);
	105	+ }
	106	+ return cpus;
	107	+}
	108	+
	109	+struct perf_cpu_map *perf_cpu_map__new_online_cpus(void)
	110	+{
	111	+ struct perf_cpu_map *cpus = cpu_map__new_sysfs_online();
	112	+
	113	+ if (cpus)
	114	+ return cpus;
	115	+
	116	+ return cpu_map__new_sysconf();
92	117	}
93	118
94	119
@@ -180,27 +205,11 @@ struct perf_cpu_map perf_cpu_map__read(FILE file) · 85 unmodified lines
180	205
181	206	if (nr_cpus > 0)
182	207	cpus = cpu_map__trim_new(nr_cpus, tmp_cpus);
183		− else
184		− cpus = cpu_map__default_new();
185	208	out_free_tmp:
186	209	free(tmp_cpus);
187	210	return cpus;
188	211	}
189	212
190		−static struct perf_cpu_map *cpu_map__read_all_cpu_map(void)
191		−{
192		− struct perf_cpu_map *cpus = NULL;
193		− FILE *onlnf;
194		−
195		− onlnf = fopen("/sys/devices/system/cpu/online", "r");
196		− if (!onlnf)
197		− return cpu_map__default_new();
198		−
199		− cpus = perf_cpu_map__read(onlnf);
200		− fclose(onlnf);
201		− return cpus;
202		−}
203		−
204	213	struct perf_cpu_map perf_cpu_map__new(const char cpu_list)
205	214	{
206	215	struct perf_cpu_map *cpus = NULL;
@@ -211,7 +220,7 @@ struct perf_cpu_map perf_cpu_map__new(const char cpu_list) · 4 unmodified lines
211	220	int max_entries = 0;
212	221
213	222	if (!cpu_list)
214		− return cpu_map__read_all_cpu_map();
	223	+ return perf_cpu_map__new_online_cpus();
215	224
216	225	/*
217	226	* must handle the case of empty cpumap to cover
@@ -268,10 +277,12 @@ struct perf_cpu_map perf_cpu_map__new(const char cpu_list) · 50 unmodified lines
268	277
269	278	if (nr_cpus > 0)
270	279	cpus = cpu_map__trim_new(nr_cpus, tmp_cpus);
271		− else if (*cpu_list != '\0')
272		− cpus = cpu_map__default_new();
273		− else
274		− cpus = perf_cpu_map__dummy_new();
	280	+ else if (*cpu_list != '\0') {
	281	+ pr_warning("Unexpected characters at end of cpu list ('%s'), using online CPUs.",
	282	+ cpu_list);
	283	+ cpus = perf_cpu_map__new_online_cpus();
	284	+ } else
	285	+ cpus = perf_cpu_map__new_any_cpu();
275	286	invalid:
276	287	free(tmp_cpus);
277	288	out:
@@ -300,7 +311,7 @@ int perf_cpu_map__nr(const struct perf_cpu_map *cpus) · 22 unmodified lines
300	311	return cpus ? __perf_cpu_map__nr(cpus) : 1;
301	312	}
302	313
303		−bool perf_cpu_map__empty(const struct perf_cpu_map *map)
	314	+bool perf_cpu_map__has_any_cpu_or_is_empty(const struct perf_cpu_map *map)
304	315	{
305	316	return map ? __perf_cpu_map__cpu(map, 0).cpu == -1 : true;
306	317	}

tools/lib/perf/evlist.c+3−3

@@ -39,7 +39,7 @@ static void __perf_evlist__propagate_maps(struct perf_evlist *evlist, · 38 unmodified lines
39	39	if (evsel->system_wide) {
40	40	/* System wide: set the cpu map of the evsel to all online CPUs. */
41	41	perf_cpu_map__put(evsel->cpus);
42		− evsel->cpus = perf_cpu_map__new(NULL);
	42	+ evsel->cpus = perf_cpu_map__new_online_cpus();
43	43	} else if (evlist->has_user_cpus && evsel->is_pmu_core) {
44	44	/*
45	45	* User requested CPUs on a core PMU, ensure the requested CPUs
@@ -619,7 +619,7 @@ static int perf_evlist__nr_mmaps(struct perf_evlist *evlist) · 573 unmodified lines
619	619
620	620	/* One for each CPU */
621	621	nr_mmaps = perf_cpu_map__nr(evlist->all_cpus);
622		− if (perf_cpu_map__empty(evlist->all_cpus)) {
	622	+ if (perf_cpu_map__has_any_cpu_or_is_empty(evlist->all_cpus)) {
623	623	/* Plus one for each thread */
624	624	nr_mmaps += perf_thread_map__nr(evlist->threads);
625	625	/* Minus the per-thread CPU (-1) */
@@ -653,7 +653,7 @@ int perf_evlist__mmap_ops(struct perf_evlist *evlist, · 27 unmodified lines
653	653	if (evlist->pollfd.entries == NULL && perf_evlist__alloc_pollfd(evlist) < 0)
654	654	return -ENOMEM;
655	655
656		− if (perf_cpu_map__empty(cpus))
	656	+ if (perf_cpu_map__has_any_cpu_or_is_empty(cpus))
657	657	return mmap_per_thread(evlist, ops, mp);
658	658
659	659	return mmap_per_cpu(evlist, ops, mp);

tools/lib/perf/evsel.c+1−1

@@ -120,7 +120,7 @@ int perf_evsel__open(struct perf_evsel evsel, struct perf_cpu_map cpus, · 119 unmodified lines
120	120	static struct perf_cpu_map *empty_cpu_map;
121	121
122	122	if (empty_cpu_map == NULL) {
123		− empty_cpu_map = perf_cpu_map__dummy_new();
	123	+ empty_cpu_map = perf_cpu_map__new_any_cpu();
124	124	if (empty_cpu_map == NULL)
125	125	return -ENOMEM;
126	126	}

tools/lib/perf/include/internal/mmap.h+2−1

@@ -33,7 +33,8 @@ struct perf_mmap { · 32 unmodified lines
33	33	bool overwrite;
34	34	u64 flush;
35	35	libperf_unmap_cb_t unmap_cb;
36		− char event_copy[PERF_SAMPLE_MAX_SIZE] __aligned(8);
	36	+ void *event_copy;
	37	+ size_t event_copy_sz;
37	38	struct perf_mmap *next;
38	39	};
39	40

tools/lib/perf/include/perf/cpumap.h+35−5

@@ -19,10 +19,23 @@ struct perf_cache { · 18 unmodified lines
19	19	struct perf_cpu_map;
20	20
21	21	/**
22		− * perf_cpu_map__dummy_new - a map with a singular "any CPU"/dummy -1 value.
	22	+ * perf_cpu_map__new_any_cpu - a map with a singular "any CPU"/dummy -1 value.
	23	+ */
	24	+LIBPERF_API struct perf_cpu_map *perf_cpu_map__new_any_cpu(void);
	25	+/**
	26	+ * perf_cpu_map__new_online_cpus - a map read from
	27	+ * /sys/devices/system/cpu/online if
	28	+ * available. If reading wasn't possible a map
	29	+ * is created using the online processors
	30	+ * assuming the first 'n' processors are all
	31	+ * online.
	32	+ */
	33	+LIBPERF_API struct perf_cpu_map *perf_cpu_map__new_online_cpus(void);
	34	+/**
	35	+ * perf_cpu_map__new - create a map from the given cpu_list such as "0-7". If no
	36	+ * cpu_list argument is provided then
	37	+ * perf_cpu_map__new_online_cpus is returned.
23	38	*/
24		−LIBPERF_API struct perf_cpu_map *perf_cpu_map__dummy_new(void);
25		−LIBPERF_API struct perf_cpu_map *perf_cpu_map__default_new(void);
26	39	LIBPERF_API struct perf_cpu_map perf_cpu_map__new(const char cpu_list);
27	40	LIBPERF_API struct perf_cpu_map perf_cpu_map__read(FILE file);
28	41	LIBPERF_API struct perf_cpu_map perf_cpu_map__get(struct perf_cpu_map map);
@@ -31,12 +44,23 @@ LIBPERF_API struct perf_cpu_map perf_cpu_map__merge(struct perf_cpu_map orig, · 2 unmodified lines
31	44	LIBPERF_API struct perf_cpu_map perf_cpu_map__intersect(struct perf_cpu_map orig,
32	45	struct perf_cpu_map *other);
33	46	LIBPERF_API void perf_cpu_map__put(struct perf_cpu_map *map);
	47	+/**
	48	+ * perf_cpu_map__cpu - get the CPU value at the given index. Returns -1 if index
	49	+ * is invalid.
	50	+ */
34	51	LIBPERF_API struct perf_cpu perf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx);
	52	+/**
	53	+ * perf_cpu_map__nr - for an empty map returns 1, as perf_cpu_map__cpu returns a
	54	+ * cpu of -1 for an invalid index, this makes an empty map
	55	+ * look like it contains the "any CPU"/dummy value. Otherwise
	56	+ * the result is the number CPUs in the map plus one if the
	57	+ * "any CPU"/dummy value is present.
	58	+ */
35	59	LIBPERF_API int perf_cpu_map__nr(const struct perf_cpu_map *cpus);
36	60	/**
37		− * perf_cpu_map__empty - is map either empty or the "any CPU"/dummy value.
	61	+ * perf_cpu_map__has_any_cpu_or_is_empty - is map either empty or has the "any CPU"/dummy value.
38	62	*/
39		−LIBPERF_API bool perf_cpu_map__empty(const struct perf_cpu_map *map);
	63	+LIBPERF_API bool perf_cpu_map__has_any_cpu_or_is_empty(const struct perf_cpu_map *map);
40	64	LIBPERF_API struct perf_cpu perf_cpu_map__max(const struct perf_cpu_map *map);
41	65	LIBPERF_API bool perf_cpu_map__has(const struct perf_cpu_map *map, struct perf_cpu cpu);
42	66	LIBPERF_API bool perf_cpu_map__equal(const struct perf_cpu_map *lhs,
@@ -51,6 +75,12 @@ LIBPERF_API bool perf_cpu_map__has_any_cpu(const struct perf_cpu_map *map); · 8 unmodified lines
51	75	(idx) < perf_cpu_map__nr(cpus); \
52	76	(idx)++, (cpu) = perf_cpu_map__cpu(cpus, idx))
53	77
	78	+#define perf_cpu_map__for_each_cpu_skip_any(_cpu, idx, cpus) \
	79	+ for ((idx) = 0, (_cpu) = perf_cpu_map__cpu(cpus, idx); \
	80	+ (idx) < perf_cpu_map__nr(cpus); \
	81	+ (idx)++, (_cpu) = perf_cpu_map__cpu(cpus, idx)) \
	82	+ if ((_cpu).cpu != -1)
	83	+
54	84	#define perf_cpu_map__for_each_idx(idx, cpus) \
55	85	for ((idx) = 0; (idx) < perf_cpu_map__nr(cpus); (idx)++)
56	86

tools/lib/perf/libperf.map+3−3

@@ -1,15 +1,15 @@
1	1	LIBPERF_0.0.1 {
2	2	global:
3	3	libperf_init;
4		− perf_cpu_map__dummy_new;
5		− perf_cpu_map__default_new;
	4	+ perf_cpu_map__new_any_cpu;
	5	+ perf_cpu_map__new_online_cpus;
6	6	perf_cpu_map__get;
7	7	perf_cpu_map__put;
8	8	perf_cpu_map__new;
9	9	perf_cpu_map__read;
10	10	perf_cpu_map__nr;
11	11	perf_cpu_map__cpu;
12		− perf_cpu_map__empty;
	12	+ perf_cpu_map__has_any_cpu_or_is_empty;
13	13	perf_cpu_map__max;
14	14	perf_cpu_map__has;
15	15	perf_thread_map__new_array;

tools/lib/perf/mmap.c+17−3

@@ -19,6 +19,7 @@ · 18 unmodified lines
19	19	void perf_mmap__init(struct perf_mmap map, struct perf_mmap prev,
20	20	bool overwrite, libperf_unmap_cb_t unmap_cb)
21	21	{
	22	+ /* Assume fields were zero initialized. */
22	23	map->fd = -1;
23	24	map->overwrite = overwrite;
24	25	map->unmap_cb = unmap_cb;
@@ -51,13 +52,18 @@ int perf_mmap__mmap(struct perf_mmap map, struct perf_mmap_param mp, · 26 unmodified lines
51	52
52	53	void perf_mmap__munmap(struct perf_mmap *map)
53	54	{
54		− if (map && map->base != NULL) {
	55	+ if (!map)
	56	+ return;
	57	+
	58	+ zfree(&map->event_copy);
	59	+ map->event_copy_sz = 0;
	60	+ if (map->base) {
55	61	munmap(map->base, perf_mmap__mmap_len(map));
56	62	map->base = NULL;
57	63	map->fd = -1;
58	64	refcount_set(&map->refcnt, 0);
59	65	}
60		− if (map && map->unmap_cb)
	66	+ if (map->unmap_cb)
61	67	map->unmap_cb(map);
62	68	}
63	69
@@ -223,9 +229,17 @@ static union perf_event perf_mmap__read(struct perf_mmap map, · 159 unmodified lines
223	229	*/
224	230	if ((startp & map->mask) + size != ((startp + size) & map->mask)) {
225	231	unsigned int offset = *startp;
226		− unsigned int len = min(sizeof(*event), size), cpy;
	232	+ unsigned int len = size, cpy;
227	233	void *dst = map->event_copy;
228	234
	235	+ if (size > map->event_copy_sz) {
	236	+ dst = realloc(map->event_copy, size);
	237	+ if (!dst)
	238	+ return NULL;
	239	+ map->event_copy = dst;
	240	+ map->event_copy_sz = size;
	241	+ }
	242	+
229	243	do {
230	244	cpy = min(map->mask + 1 - (offset & map->mask), len);
231	245	memcpy(dst, &data[offset & map->mask], cpy);

tools/lib/perf/tests/test-cpumap.c+2−2

@@ -21,7 +21,7 @@ int test_cpumap(int argc, char **argv) · 20 unmodified lines
21	21
22	22	libperf_init(libperf_print);
23	23
24		− cpus = perf_cpu_map__dummy_new();
	24	+ cpus = perf_cpu_map__new_any_cpu();
25	25	if (!cpus)
26	26	return -1;
27	27
@@ -29,7 +29,7 @@ int test_cpumap(int argc, char **argv) · 1 unmodified line
29	29	perf_cpu_map__put(cpus);
30	30	perf_cpu_map__put(cpus);
31	31
32		− cpus = perf_cpu_map__default_new();
	32	+ cpus = perf_cpu_map__new_online_cpus();
33	33	if (!cpus)
34	34	return -1;
35	35

tools/lib/perf/tests/test-evlist.c+3−3

@@ -46,7 +46,7 @@ static int test_stat_cpu(void) · 45 unmodified lines
46	46	};
47	47	int err, idx;
48	48
49		− cpus = perf_cpu_map__new(NULL);
	49	+ cpus = perf_cpu_map__new_online_cpus();
50	50	__T("failed to create cpus", cpus);
51	51
52	52	evlist = perf_evlist__new();
@@ -261,7 +261,7 @@ static int test_mmap_thread(void) · 208 unmodified lines
261	261	threads = perf_thread_map__new_dummy();
262	262	__T("failed to create threads", threads);
263	263
264		− cpus = perf_cpu_map__dummy_new();
	264	+ cpus = perf_cpu_map__new_any_cpu();
265	265	__T("failed to create cpus", cpus);
266	266
267	267	perf_thread_map__set_pid(threads, 0, pid);
@@ -350,7 +350,7 @@ static int test_mmap_cpus(void) · 82 unmodified lines
350	350
351	351	attr.config = id;
352	352
353		− cpus = perf_cpu_map__new(NULL);
	353	+ cpus = perf_cpu_map__new_online_cpus();
354	354	__T("failed to create cpus", cpus);
355	355
356	356	evlist = perf_evlist__new();

tools/lib/perf/tests/test-evsel.c+1−1

@@ -27,7 +27,7 @@ static int test_stat_cpu(void) · 26 unmodified lines
27	27	};
28	28	int err, idx;
29	29
30		− cpus = perf_cpu_map__new(NULL);
	30	+ cpus = perf_cpu_map__new_online_cpus();
31	31	__T("failed to create cpus", cpus);
32	32
33	33	evsel = perf_evsel__new(&attr);

tools/lib/subcmd/help.c+14−4

@@ -52,11 +52,21 @@ void uniq(struct cmdnames *cmds) · 51 unmodified lines
52	52	if (!cmds->cnt)
53	53	return;
54	54
55		− for (i = j = 1; i < cmds->cnt; i++)
56		− if (strcmp(cmds->names[i]->name, cmds->names[i-1]->name))
57		− cmds->names[j++] = cmds->names[i];
58		−
	55	+ for (i = 1; i < cmds->cnt; i++) {
	56	+ if (!strcmp(cmds->names[i]->name, cmds->names[i-1]->name))
	57	+ zfree(&cmds->names[i - 1]);
	58	+ }
	59	+ for (i = 0, j = 0; i < cmds->cnt; i++) {
	60	+ if (cmds->names[i]) {
	61	+ if (i == j)
	62	+ j++;
	63	+ else
	64	+ cmds->names[j++] = cmds->names[i];
	65	+ }
	66	+ }
59	67	cmds->cnt = j;
	68	+ while (j < i)
	69	+ cmds->names[j++] = NULL;
60	70	}
61	71
62	72	void exclude_cmds(struct cmdnames cmds, struct cmdnames excludes)

tools/perf/.gitignore+4−0

@@ -39,6 +39,9 @@ trace/beauty/generated/ · 38 unmodified lines
39	39	pmu-events/pmu-events.c
40	40	pmu-events/jevents
41	41	pmu-events/metric_test.log
	42	+tests/shell/*.shellcheck_log
	43	+tests/shell/coresight/*.shellcheck_log
	44	+tests/shell/lib/*.shellcheck_log
42	45	feature/
43	46	libapi/
44	47	libbpf/
@@ -49,3 +52,4 @@ libtraceevent/ · 4 unmodified lines
49	52	libtraceevent_plugins/
50	53	fixdep
51	54	Documentation/doc.dep
	55	+python_ext_build/

tools/perf/Documentation/itrace.txt+1−0

@@ -25,6 +25,7 @@ · 24 unmodified lines
25	25	q quicker (less detailed) decoding
26	26	A approximate IPC
27	27	Z prefer to ignore timestamps (so-called "timeless" decoding)
	28	+ T use the timestamp trace as kernel time
28	29
29	30	The default is all events i.e. the same as --itrace=iybxwpe,
30	31	except for perf script where it is --itrace=ce

tools/perf/Documentation/perf-annotate.txt+11−0

@@ -155,6 +155,17 @@ include::itrace.txt[] · 154 unmodified lines
155	155	stdio or stdio2 (Default: 0). Note that this is about selection of
156	156	functions to display, not about lines within the function.
157	157
	158	+--data-type[=TYPE_NAME]::
	159	+ Display data type annotation instead of code. It infers data type of
	160	+ samples (if they are memory accessing instructions) using DWARF debug
	161	+ information. It can take an optional argument of data type name. In
	162	+ that case it'd show annotation for the type only, otherwise it'd show
	163	+ all data types it finds.
	164	+
	165	+--type-stat::
	166	+ Show stats for the data type annotation.
	167	+
	168	+
158	169	SEE ALSO
159	170	--------
160	171	linkperf:perf-record[1], linkperf:perf-report[1]

tools/perf/Documentation/perf-config.txt+2−2

@@ -251,7 +251,8 @@ annotate.*:: · 250 unmodified lines
251	251	addr2line binary to use for file names and line numbers.
252	252
253	253	annotate.objdump::
254		− objdump binary to use for disassembly and annotations.
	254	+ objdump binary to use for disassembly and annotations,
	255	+ including in the 'perf test' command.
255	256
256	257	annotate.disassembler_style::
257	258	Use this to change the default disassembler style to some other value
@@ -722,7 +723,6 @@ session-<NAME>.*:: · 464 unmodified lines
722	723	Defines new record session for daemon. The value is record's
723	724	command line without the 'record' keyword.
724	725
725		−
726	726	SEE ALSO
727	727	--------
728	728	linkperf:perf[1]

tools/perf/Documentation/perf-list.txt+7−5

@@ -81,11 +81,13 @@ For Intel systems precise event sampling is implemented with PEBS · 80 unmodified lines
81	81	which supports up to precise-level 2, and precise level 3 for
82	82	some special cases
83	83
84		−On AMD systems it is implemented using IBS (up to precise-level 2).
85		−The precise modifier works with event types 0x76 (cpu-cycles, CPU
86		−clocks not halted) and 0xC1 (micro-ops retired). Both events map to
87		−IBS execution sampling (IBS op) with the IBS Op Counter Control bit
88		−(IbsOpCntCtl) set respectively (see the
	84	+On AMD systems it is implemented using IBS OP (up to precise-level 2).
	85	+Unlike Intel PEBS which provides levels of precision, AMD core pmu is
	86	+inherently non-precise and IBS is inherently precise. (i.e. ibs_op//,
	87	+ibs_op//p, ibs_op//pp and ibs_op//ppp are all same). The precise modifier
	88	+works with event types 0x76 (cpu-cycles, CPU clocks not halted) and 0xC1
	89	+(micro-ops retired). Both events map to IBS execution sampling (IBS op)
	90	+with the IBS Op Counter Control bit (IbsOpCntCtl) set respectively (see the
89	91	Core Complex (CCX) -> Processor x86 Core -> Instruction Based Sampling (IBS)
90	92	section of the [AMD Processor Programming Reference (PPR)] relevant to the
91	93	family, model and stepping of the processor being used).

tools/perf/Documentation/perf-lock.txt+1−1

@@ -119,7 +119,7 @@ INFO OPTIONS · 118 unmodified lines
119	119
120	120
121	121	CONTENTION OPTIONS
122		−--------------
	122	+------------------
123	123
124	124	-k::
125	125	--key=<value>::

tools/perf/Documentation/perf-record.txt+4−0

@@ -445,6 +445,10 @@ following filters are defined: · 444 unmodified lines
445	445	4th-Gen Xeon+ server), the save branch type is unconditionally enabled
446	446	when the taken branch stack sampling is enabled.
447	447	- priv: save privilege state during sampling in case binary is not available later
	448	+ - counter: save occurrences of the event since the last branch entry. Currently, the
	449	+ feature is only supported by a newer CPU, e.g., Intel Sierra Forest and
	450	+ later platforms. An error out is expected if it's used on the unsupported
	451	+ kernel or CPUs.
448	452
449	453	+
450	454	The option requires at least one branch type among any, any_call, any_ret, ind_call, cond.

tools/perf/Documentation/perf-report.txt+3−0

@@ -118,6 +118,9 @@ OPTIONS · 117 unmodified lines
118	118	- retire_lat: On X86, this reports pipeline stall of this instruction compared
119	119	to the previous instruction in cycles. And currently supported only on X86
120	120	- simd: Flags describing a SIMD operation. "e" for empty Arm SVE predicate. "p" for partial Arm SVE predicate
	121	+ - type: Data type of sample memory access.
	122	+ - typeoff: Offset in the data type of sample memory access.
	123	+ - symoff: Offset in the symbol.
121	124
122	125	By default, comm, dso and symbol keys are used.
123	126	(i.e. --sort comm,dso,symbol)

tools/perf/Documentation/perf-stat.txt+28−24

@@ -422,7 +422,34 @@ See perf list output for the possible metrics and metricgroups. · 421 unmodified lines
422	422
423	423	-A::
424	424	--no-aggr::
425		−Do not aggregate counts across all monitored CPUs.
	425	+--no-merge::
	426	+Do not aggregate/merge counts across monitored CPUs or PMUs.
	427	+
	428	+When multiple events are created from a single event specification,
	429	+stat will, by default, aggregate the event counts and show the result
	430	+in a single row. This option disables that behavior and shows the
	431	+individual events and counts.
	432	+
	433	+Multiple events are created from a single event specification when:
	434	+
	435	+1. PID monitoring isn't requested and the system has more than one
	436	+ CPU. For example, a system with 8 SMT threads will have one event
	437	+ opened on each thread and aggregation is performed across them.
	438	+
	439	+2. Prefix or glob wildcard matching is used for the PMU name. For
	440	+ example, multiple memory controller PMUs may exist typically with a
	441	+ suffix of _0, _1, etc. By default the event counts will all be
	442	+ combined if the PMU is specified without the suffix such as
	443	+ uncore_imc rather than uncore_imc_0.
	444	+
	445	+3. Aliases, which are listed immediately after the Kernel PMU events
	446	+ by perf list, are used.
	447	+
	448	+--hybrid-merge::
	449	+Merge core event counts from all core PMUs. In hybrid or big.LITTLE
	450	+systems by default each core PMU will report its count
	451	+separately. This option forces core PMU counts to be combined to give
	452	+a behavior closer to having a single CPU type in the system.
426	453
427	454	--topdown::
428	455	Print top-down metrics supported by the CPU. This allows to determine
@@ -475,29 +502,6 @@ highlight 'tma_frontend_bound'. This metric may be drilled into with · 46 unmodified lines
475	502
476	503	Error out if the input is higher than the supported max level.
477	504
478		−--no-merge::
479		−Do not merge results from same PMUs.
480		−
481		−When multiple events are created from a single event specification,
482		−stat will, by default, aggregate the event counts and show the result
483		−in a single row. This option disables that behavior and shows
484		−the individual events and counts.
485		−
486		−Multiple events are created from a single event specification when:
487		−1. Prefix or glob matching is used for the PMU name.
488		−2. Aliases, which are listed immediately after the Kernel PMU events
489		− by perf list, are used.
490		−
491		−--hybrid-merge::
492		−Merge the hybrid event counts from all PMUs.
493		−
494		−For hybrid events, by default, the stat aggregates and reports the event
495		−counts per PMU. But sometimes, it's also useful to aggregate event counts
496		−from all PMUs. This option enables that behavior and reports the counts
497		−without PMUs.
498		−
499		−For non-hybrid events, it should be no effect.
500		−
501	505	--smi-cost::
502	506	Measure SMI cost if msr/aperf/ and msr/smi/ events are supported.
503	507

tools/perf/Documentation/perf.txt+3−0

@@ -64,6 +64,9 @@ OPTIONS · 63 unmodified lines
64	64	perf-event-open - Print perf_event_open() arguments and
65	65	return value
66	66
	67	+--debug-file::
	68	+ Write debug output to a specified file.
	69	+
67	70	DESCRIPTION
68	71	-----------
69	72	Performance counters for Linux are a new kernel-based subsystem

tools/perf/Makefile.config+9−4

@@ -476,6 +476,11 @@ else · 475 unmodified lines
476	476	else
477	477	CFLAGS += -DHAVE_DWARF_GETLOCATIONS_SUPPORT
478	478	endif # dwarf_getlocations
	479	+ ifneq ($(feature-dwarf_getcfi), 1)
	480	+ msg := $(warning Old libdw.h, finding variables at given 'perf probe' point will not work, install elfutils-devel/libdw-dev >= 0.142);
	481	+ else
	482	+ CFLAGS += -DHAVE_DWARF_CFI_SUPPORT
	483	+ endif # dwarf_getcfi
479	484	endif # Dwarf support
480	485	endif # libelf support
481	486	endif # NO_LIBELF
@@ -680,15 +685,15 @@ ifndef BUILD_BPF_SKEL · 198 unmodified lines
680	685	endif
681	686
682	687	ifeq ($(BUILD_BPF_SKEL),1)
683		− ifeq ($(filter -DHAVE_LIBBPF_SUPPORT, $(CFLAGS)),)
684		− dummy := $(warning Warning: Disabled BPF skeletons as libbpf is required)
685		− BUILD_BPF_SKEL := 0
686		− else ifeq ($(filter -DHAVE_LIBELF_SUPPORT, $(CFLAGS)),)
	688	+ ifeq ($(filter -DHAVE_LIBELF_SUPPORT, $(CFLAGS)),)
687	689	dummy := $(warning Warning: Disabled BPF skeletons as libelf is required by bpftool)
688	690	BUILD_BPF_SKEL := 0
689	691	else ifeq ($(filter -DHAVE_ZLIB_SUPPORT, $(CFLAGS)),)
690	692	dummy := $(warning Warning: Disabled BPF skeletons as zlib is required by bpftool)
691	693	BUILD_BPF_SKEL := 0
	694	+ else ifeq ($(filter -DHAVE_LIBBPF_SUPPORT, $(CFLAGS)),)
	695	+ dummy := $(warning Warning: Disabled BPF skeletons as libbpf is required)
	696	+ BUILD_BPF_SKEL := 0
692	697	else ifeq ($(call get-executable,$(CLANG)),)
693	698	dummy := $(warning Warning: Disabled BPF skeletons as clang ($(CLANG)) is missing)
694	699	BUILD_BPF_SKEL := 0

tools/perf/Makefile.perf+11−2

@@ -134,6 +134,8 @@ include ../scripts/utilities.mak · 133 unmodified lines
134	134	# x86 instruction decoder - new instructions test
135	135	#
136	136	# Define GEN_VMLINUX_H to generate vmlinux.h from the BTF.
	137	+#
	138	+# Define NO_SHELLCHECK if you do not want to run shellcheck during build
137	139
138	140	# As per kernel Makefile, avoid funny character set dependencies
139	141	unexport LC_ALL
@@ -227,8 +229,15 @@ else · 87 unmodified lines
227	229	force_fixdep := $(config)
228	230	endif
229	231
	232	+# Runs shellcheck on perf test shell scripts
	233	+ifeq ($(NO_SHELLCHECK),1)
	234	+ SHELLCHECK :=
	235	+else
	236	+ SHELLCHECK := $(shell which shellcheck 2> /dev/null)
	237	+endif
	238	+
230	239	export srctree OUTPUT RM CC CXX LD AR CFLAGS CXXFLAGS V BISON FLEX AWK
231		−export HOSTCC HOSTLD HOSTAR HOSTCFLAGS
	240	+export HOSTCC HOSTLD HOSTAR HOSTCFLAGS SHELLCHECK
232	241
233	242	include $(srctree)/tools/build/Makefile.include
234	243
@@ -1152,7 +1161,7 @@ bpf-skel-clean: · 917 unmodified lines
1152	1161
1153	1162	clean:: $(LIBAPI)-clean $(LIBBPF)-clean $(LIBSUBCMD)-clean $(LIBSYMBOL)-clean $(LIBPERF)-clean arm64-sysreg-defs-clean fixdep-clean python-clean bpf-skel-clean tests-coresight-targets-clean
1154	1163	$(call QUIET_CLEAN, core-objs) $(RM) $(LIBPERF_A) $(OUTPUT)perf-archive $(OUTPUT)perf-iostat $(LANG_BINDINGS)
1155		− $(Q)find $(or $(OUTPUT),.) -name '.o' -delete -o -name '\..cmd' -delete -o -name '\.*.d' -delete
	1164	+ $(Q)find $(or $(OUTPUT),.) -name '.o' -delete -o -name '\..cmd' -delete -o -name '\..d' -delete -o -name '.shellcheck_log' -delete
1156	1165	$(Q)$(RM) $(OUTPUT).config-detected
1157	1166	$(call QUIET_CLEAN, core-progs) $(RM) $(ALL_PROGRAMS) perf perf-read-vdso32 perf-read-vdsox32 $(OUTPUT)$(LIBJVMTI).so
1158	1167	$(call QUIET_CLEAN, core-gen) $(RM) .spec .pyc .pyo /.pyc /.pyo $(OUTPUT)common-cmds.h TAGS tags cscope $(OUTPUT)PERF-VERSION-FILE $(OUTPUT)FEATURE-DUMP $(OUTPUT)util/-bison $(OUTPUT)util/-flex \

tools/perf/arch/arm/util/cs-etm.c+8−8

@@ -199,7 +199,7 @@ static int cs_etm_validate_config(struct auxtrace_record *itr, · 198 unmodified lines
199	199	{
200	200	int i, err = -EINVAL;
201	201	struct perf_cpu_map *event_cpus = evsel->evlist->core.user_requested_cpus;
202		− struct perf_cpu_map *online_cpus = perf_cpu_map__new(NULL);
	202	+ struct perf_cpu_map *online_cpus = perf_cpu_map__new_online_cpus();
203	203
204	204	/* Set option of each CPU we have */
205	205	for (i = 0; i < cpu__max_cpu().cpu; i++) {
@@ -211,7 +211,7 @@ static int cs_etm_validate_config(struct auxtrace_record *itr, · 5 unmodified lines
211	211	* program can run on any CPUs in this case, thus don't skip
212	212	* validation.
213	213	*/
214		− if (!perf_cpu_map__empty(event_cpus) &&
	214	+ if (!perf_cpu_map__has_any_cpu_or_is_empty(event_cpus) &&
215	215	!perf_cpu_map__has(event_cpus, cpu))
216	216	continue;
217	217
@@ -435,7 +435,7 @@ static int cs_etm_recording_options(struct auxtrace_record *itr, · 217 unmodified lines
435	435	* Also the case of per-cpu mmaps, need the contextID in order to be notified
436	436	* when a context switch happened.
437	437	*/
438		− if (!perf_cpu_map__empty(cpus)) {
	438	+ if (!perf_cpu_map__has_any_cpu_or_is_empty(cpus)) {
439	439	evsel__set_config_if_unset(cs_etm_pmu, cs_etm_evsel,
440	440	"timestamp", 1);
441	441	evsel__set_config_if_unset(cs_etm_pmu, cs_etm_evsel,
@@ -461,7 +461,7 @@ static int cs_etm_recording_options(struct auxtrace_record *itr, · 19 unmodified lines
461	461	evsel->core.attr.sample_period = 1;
462	462
463	463	/* In per-cpu case, always need the time of mmap events etc */
464		− if (!perf_cpu_map__empty(cpus))
	464	+ if (!perf_cpu_map__has_any_cpu_or_is_empty(cpus))
465	465	evsel__set_sample_bit(evsel, TIME);
466	466
467	467	err = cs_etm_validate_config(itr, cs_etm_evsel);
@@ -536,10 +536,10 @@ cs_etm_info_priv_size(struct auxtrace_record *itr __maybe_unused, · 68 unmodified lines
536	536	int i;
537	537	int etmv3 = 0, etmv4 = 0, ete = 0;
538	538	struct perf_cpu_map *event_cpus = evlist->core.user_requested_cpus;
539		− struct perf_cpu_map *online_cpus = perf_cpu_map__new(NULL);
	539	+ struct perf_cpu_map *online_cpus = perf_cpu_map__new_online_cpus();
540	540
541	541	/* cpu map is not empty, we have specific CPUs to work with */
542		− if (!perf_cpu_map__empty(event_cpus)) {
	542	+ if (!perf_cpu_map__has_any_cpu_or_is_empty(event_cpus)) {
543	543	for (i = 0; i < cpu__max_cpu().cpu; i++) {
544	544	struct perf_cpu cpu = { .cpu = i, };
545	545
@@ -802,7 +802,7 @@ static int cs_etm_info_fill(struct auxtrace_record *itr, · 256 unmodified lines
802	802	u64 nr_cpu, type;
803	803	struct perf_cpu_map *cpu_map;
804	804	struct perf_cpu_map *event_cpus = session->evlist->core.user_requested_cpus;
805		− struct perf_cpu_map *online_cpus = perf_cpu_map__new(NULL);
	805	+ struct perf_cpu_map *online_cpus = perf_cpu_map__new_online_cpus();
806	806	struct cs_etm_recording *ptr =
807	807	container_of(itr, struct cs_etm_recording, itr);
808	808	struct perf_pmu *cs_etm_pmu = ptr->cs_etm_pmu;
@@ -814,7 +814,7 @@ static int cs_etm_info_fill(struct auxtrace_record *itr, · 5 unmodified lines
814	814	return -EINVAL;
815	815
816	816	/* If the cpu_map is empty all online CPUs are involved */
817		− if (perf_cpu_map__empty(event_cpus)) {
	817	+ if (perf_cpu_map__has_any_cpu_or_is_empty(event_cpus)) {
818	818	cpu_map = online_cpus;
819	819	} else {
820	820	/* Make sure all specified CPUs are online */

tools/perf/arch/arm64/util/arm-spe.c+2−2

@@ -232,7 +232,7 @@ static int arm_spe_recording_options(struct auxtrace_record *itr, · 231 unmodified lines
232	232	* In the case of per-cpu mmaps, sample CPU for AUX event;
233	233	* also enable the timestamp tracing for samples correlation.
234	234	*/
235		− if (!perf_cpu_map__empty(cpus)) {
	235	+ if (!perf_cpu_map__has_any_cpu_or_is_empty(cpus)) {
236	236	evsel__set_sample_bit(arm_spe_evsel, CPU);
237	237	evsel__set_config_if_unset(arm_spe_pmu, arm_spe_evsel,
238	238	"ts_enable", 1);
@@ -265,7 +265,7 @@ static int arm_spe_recording_options(struct auxtrace_record *itr, · 26 unmodified lines
265	265	tracking_evsel->core.attr.sample_period = 1;
266	266
267	267	/* In per-cpu case, always need the time of mmap events etc */
268		− if (!perf_cpu_map__empty(cpus)) {
	268	+ if (!perf_cpu_map__has_any_cpu_or_is_empty(cpus)) {
269	269	evsel__set_sample_bit(tracking_evsel, TIME);
270	270	evsel__set_sample_bit(tracking_evsel, CPU);
271	271

tools/perf/arch/arm64/util/header.c+1−1

@@ -57,7 +57,7 @@ static int _get_cpuid(char buf, size_t sz, struct perf_cpu_map cpus) · 56 unmodified lines
57	57
58	58	int get_cpuid(char *buf, size_t sz)
59	59	{
60		− struct perf_cpu_map *cpus = perf_cpu_map__new(NULL);
	60	+ struct perf_cpu_map *cpus = perf_cpu_map__new_online_cpus();
61	61	int ret;
62	62
63	63	if (!cpus)

tools/perf/arch/loongarch/annotate/instructions.c+3−3

@@ -61,10 +61,10 @@ static int loongarch_jump__parse(struct arch arch, struct ins_operands ops, st · 60 unmodified lines
61	61	const char *c = strchr(ops->raw, '#');
62	62	u64 start, end;
63	63
64		− ops->raw_comment = strchr(ops->raw, arch->objdump.comment_char);
65		− ops->raw_func_start = strchr(ops->raw, '<');
	64	+ ops->jump.raw_comment = strchr(ops->raw, arch->objdump.comment_char);
	65	+ ops->jump.raw_func_start = strchr(ops->raw, '<');
66	66
67		− if (ops->raw_func_start && c > ops->raw_func_start)
	67	+ if (ops->jump.raw_func_start && c > ops->jump.raw_func_start)
68	68	c = NULL;
69	69
70	70	if (c++ != NULL)

tools/perf/arch/x86/tests/hybrid.c+30−7

@@ -47,7 +47,7 @@ static int test__hybrid_hw_group_event(struct evlist *evlist) · 46 unmodified lines
47	47	evsel = evsel__next(evsel);
48	48	TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
49	49	TEST_ASSERT_VAL("wrong hybrid type", test_hybrid_type(evsel, PERF_TYPE_RAW));
50		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_INSTRUCTIONS));
	50	+ TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_BRANCH_INSTRUCTIONS));
51	51	TEST_ASSERT_VAL("wrong leader", evsel__has_leader(evsel, leader));
52	52	return TEST_OK;
53	53	}
@@ -102,7 +102,7 @@ static int test__hybrid_group_modifier1(struct evlist *evlist) · 48 unmodified lines
102	102	evsel = evsel__next(evsel);
103	103	TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
104	104	TEST_ASSERT_VAL("wrong hybrid type", test_hybrid_type(evsel, PERF_TYPE_RAW));
105		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_INSTRUCTIONS));
	105	+ TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_BRANCH_INSTRUCTIONS));
106	106	TEST_ASSERT_VAL("wrong leader", evsel__has_leader(evsel, leader));
107	107	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
108	108	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
@@ -163,6 +163,24 @@ static int test__checkevent_pmu(struct evlist *evlist) · 54 unmodified lines
163	163	return TEST_OK;
164	164	}
165	165
	166	+static int test__hybrid_hw_group_event_2(struct evlist *evlist)
	167	+{
	168	+ struct evsel evsel, leader;
	169	+
	170	+ evsel = leader = evlist__first(evlist);
	171	+ TEST_ASSERT_VAL("wrong number of entries", 2 == evlist->core.nr_entries);
	172	+ TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
	173	+ TEST_ASSERT_VAL("wrong hybrid type", test_hybrid_type(evsel, PERF_TYPE_RAW));
	174	+ TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	175	+ TEST_ASSERT_VAL("wrong leader", evsel__has_leader(evsel, leader));
	176	+
	177	+ evsel = evsel__next(evsel);
	178	+ TEST_ASSERT_VAL("wrong type", PERF_TYPE_RAW == evsel->core.attr.type);
	179	+ TEST_ASSERT_VAL("wrong config", evsel->core.attr.config == 0x3c);
	180	+ TEST_ASSERT_VAL("wrong leader", evsel__has_leader(evsel, leader));
	181	+ return TEST_OK;
	182	+}
	183	+
166	184	struct evlist_test {
167	185	const char *name;
168	186	bool (*valid)(void);
@@ -171,27 +189,27 @@ struct evlist_test { · 2 unmodified lines
171	189
172	190	static const struct evlist_test test__hybrid_events[] = {
173	191	{
174		− .name = "cpu_core/cpu-cycles/",
	192	+ .name = "cpu_core/cycles/",
175	193	.check = test__hybrid_hw_event_with_pmu,
176	194	/* 0 */
177	195	},
178	196	{
179		− .name = "{cpu_core/cpu-cycles/,cpu_core/instructions/}",
	197	+ .name = "{cpu_core/cycles/,cpu_core/branches/}",
180	198	.check = test__hybrid_hw_group_event,
181	199	/* 1 */
182	200	},
183	201	{
184		− .name = "{cpu-clock,cpu_core/cpu-cycles/}",
	202	+ .name = "{cpu-clock,cpu_core/cycles/}",
185	203	.check = test__hybrid_sw_hw_group_event,
186	204	/* 2 */
187	205	},
188	206	{
189		− .name = "{cpu_core/cpu-cycles/,cpu-clock}",
	207	+ .name = "{cpu_core/cycles/,cpu-clock}",
190	208	.check = test__hybrid_hw_sw_group_event,
191	209	/* 3 */
192	210	},
193	211	{
194		− .name = "{cpu_core/cpu-cycles/k,cpu_core/instructions/u}",
	212	+ .name = "{cpu_core/cycles/k,cpu_core/branches/u}",
195	213	.check = test__hybrid_group_modifier1,
196	214	/* 4 */
197	215	},
@@ -215,6 +233,11 @@ static const struct evlist_test test__hybrid_events[] = { · 17 unmodified lines
215	233	.check = test__hybrid_cache_event,
216	234	/* 8 */
217	235	},
	236	+ {
	237	+ .name = "{cpu_core/cycles/,cpu_core/cpu-cycles/}",
	238	+ .check = test__hybrid_hw_group_event_2,
	239	+ /* 9 */
	240	+ },
218	241	};
219	242
220	243	static int test_event(const struct evlist_test *e)

tools/perf/arch/x86/util/dwarf-regs.c+38−0

@@ -113,3 +113,41 @@ int regs_query_register_offset(const char *name) · 112 unmodified lines
113	113	return roff->offset;
114	114	return -EINVAL;
115	115	}
	116	+
	117	+struct dwarf_regs_idx {
	118	+ const char *name;
	119	+ int idx;
	120	+};
	121	+
	122	+static const struct dwarf_regs_idx x86_regidx_table[] = {
	123	+ { "rax", 0 }, { "eax", 0 }, { "ax", 0 }, { "al", 0 },
	124	+ { "rdx", 1 }, { "edx", 1 }, { "dx", 1 }, { "dl", 1 },
	125	+ { "rcx", 2 }, { "ecx", 2 }, { "cx", 2 }, { "cl", 2 },
	126	+ { "rbx", 3 }, { "edx", 3 }, { "bx", 3 }, { "bl", 3 },
	127	+ { "rsi", 4 }, { "esi", 4 }, { "si", 4 }, { "sil", 4 },
	128	+ { "rdi", 5 }, { "edi", 5 }, { "di", 5 }, { "dil", 5 },
	129	+ { "rbp", 6 }, { "ebp", 6 }, { "bp", 6 }, { "bpl", 6 },
	130	+ { "rsp", 7 }, { "esp", 7 }, { "sp", 7 }, { "spl", 7 },
	131	+ { "r8", 8 }, { "r8d", 8 }, { "r8w", 8 }, { "r8b", 8 },
	132	+ { "r9", 9 }, { "r9d", 9 }, { "r9w", 9 }, { "r9b", 9 },
	133	+ { "r10", 10 }, { "r10d", 10 }, { "r10w", 10 }, { "r10b", 10 },
	134	+ { "r11", 11 }, { "r11d", 11 }, { "r11w", 11 }, { "r11b", 11 },
	135	+ { "r12", 12 }, { "r12d", 12 }, { "r12w", 12 }, { "r12b", 12 },
	136	+ { "r13", 13 }, { "r13d", 13 }, { "r13w", 13 }, { "r13b", 13 },
	137	+ { "r14", 14 }, { "r14d", 14 }, { "r14w", 14 }, { "r14b", 14 },
	138	+ { "r15", 15 }, { "r15d", 15 }, { "r15w", 15 }, { "r15b", 15 },
	139	+ { "rip", DWARF_REG_PC },
	140	+};
	141	+
	142	+int get_arch_regnum(const char *name)
	143	+{
	144	+ unsigned int i;
	145	+
	146	+ if (*name != '%')
	147	+ return -EINVAL;
	148	+
	149	+ for (i = 0; i < ARRAY_SIZE(x86_regidx_table); i++)
	150	+ if (!strcmp(x86_regidx_table[i].name, name + 1))
	151	+ return x86_regidx_table[i].idx;
	152	+ return -ENOENT;
	153	+}

tools/perf/arch/x86/util/event.c+58−45

@@ -14,66 +14,79 @@ · 13 unmodified lines
14	14
15	15	#if defined(__x86_64__)
16	16
17		−int perf_event__synthesize_extra_kmaps(struct perf_tool *tool,
18		− perf_event__handler_t process,
19		− struct machine *machine)
	17	+struct perf_event__synthesize_extra_kmaps_cb_args {
	18	+ struct perf_tool *tool;
	19	+ perf_event__handler_t process;
	20	+ struct machine *machine;
	21	+ union perf_event *event;
	22	+};
	23	+
	24	+static int perf_event__synthesize_extra_kmaps_cb(struct map map, void data)
20	25	{
21		− int rc = 0;
22		− struct map_rb_node *pos;
23		− struct maps *kmaps = machine__kernel_maps(machine);
24		− union perf_event *event = zalloc(sizeof(event->mmap) +
25		− machine->id_hdr_size);
	26	+ struct perf_event__synthesize_extra_kmaps_cb_args *args = data;
	27	+ union perf_event *event = args->event;
	28	+ struct kmap *kmap;
	29	+ size_t size;
26	30
27		− if (!event) {
28		− pr_debug("Not enough memory synthesizing mmap event "
29		− "for extra kernel maps\n");
30		− return -1;
31		− }
	31	+ if (!__map__is_extra_kernel_map(map))
	32	+ return 0;
32	33
33		− maps__for_each_entry(kmaps, pos) {
34		− struct kmap *kmap;
35		− size_t size;
36		− struct map *map = pos->map;
	34	+ kmap = map__kmap(map);
37	35
38		− if (!__map__is_extra_kernel_map(map))
39		− continue;
	36	+ size = sizeof(event->mmap) - sizeof(event->mmap.filename) +
	37	+ PERF_ALIGN(strlen(kmap->name) + 1, sizeof(u64)) +
	38	+ args->machine->id_hdr_size;
40	39
41		− kmap = map__kmap(map);
	40	+ memset(event, 0, size);
42	41
43		− size = sizeof(event->mmap) - sizeof(event->mmap.filename) +
44		− PERF_ALIGN(strlen(kmap->name) + 1, sizeof(u64)) +
45		− machine->id_hdr_size;
	42	+ event->mmap.header.type = PERF_RECORD_MMAP;
46	43
47		− memset(event, 0, size);
	44	+ /*
	45	+ * kernel uses 0 for user space maps, see kernel/perf_event.c
	46	+ * __perf_event_mmap
	47	+ */
	48	+ if (machine__is_host(args->machine))
	49	+ event->header.misc = PERF_RECORD_MISC_KERNEL;
	50	+ else
	51	+ event->header.misc = PERF_RECORD_MISC_GUEST_KERNEL;
48	52
49		− event->mmap.header.type = PERF_RECORD_MMAP;
	53	+ event->mmap.header.size = size;
50	54
51		− /*
52		− * kernel uses 0 for user space maps, see kernel/perf_event.c
53		− * __perf_event_mmap
54		− */
55		− if (machine__is_host(machine))
56		− event->header.misc = PERF_RECORD_MISC_KERNEL;
57		− else
58		− event->header.misc = PERF_RECORD_MISC_GUEST_KERNEL;
	55	+ event->mmap.start = map__start(map);
	56	+ event->mmap.len = map__size(map);
	57	+ event->mmap.pgoff = map__pgoff(map);
	58	+ event->mmap.pid = args->machine->pid;
59	59
60		− event->mmap.header.size = size;
	60	+ strlcpy(event->mmap.filename, kmap->name, PATH_MAX);
61	61
62		− event->mmap.start = map__start(map);
63		− event->mmap.len = map__size(map);
64		− event->mmap.pgoff = map__pgoff(map);
65		− event->mmap.pid = machine->pid;
	62	+ if (perf_tool__process_synth_event(args->tool, event, args->machine, args->process) != 0)
	63	+ return -1;
66	64
67		− strlcpy(event->mmap.filename, kmap->name, PATH_MAX);
	65	+ return 0;
	66	+}
68	67
69		− if (perf_tool__process_synth_event(tool, event, machine,
70		− process) != 0) {
71		− rc = -1;
72		− break;
73		− }
	68	+int perf_event__synthesize_extra_kmaps(struct perf_tool *tool,
	69	+ perf_event__handler_t process,
	70	+ struct machine *machine)
	71	+{
	72	+ int rc;
	73	+ struct maps *kmaps = machine__kernel_maps(machine);
	74	+ struct perf_event__synthesize_extra_kmaps_cb_args args = {
	75	+ .tool = tool,
	76	+ .process = process,
	77	+ .machine = machine,
	78	+ .event = zalloc(sizeof(args.event->mmap) + machine->id_hdr_size),
	79	+ };
	80	+
	81	+ if (!args.event) {
	82	+ pr_debug("Not enough memory synthesizing mmap event "
	83	+ "for extra kernel maps\n");
	84	+ return -1;
74	85	}
75	86
76		− free(event);
	87	+ rc = maps__for_each_map(kmaps, perf_event__synthesize_extra_kmaps_cb, &args);
	88	+
	89	+ free(args.event);
77	90	return rc;
78	91	}
79	92

tools/perf/arch/x86/util/intel-bts.c+2−2

@@ -143,7 +143,7 @@ static int intel_bts_recording_options(struct auxtrace_record *itr, · 142 unmodified lines
143	143	if (!opts->full_auxtrace)
144	144	return 0;
145	145
146		− if (opts->full_auxtrace && !perf_cpu_map__empty(cpus)) {
	146	+ if (opts->full_auxtrace && !perf_cpu_map__has_any_cpu_or_is_empty(cpus)) {
147	147	pr_err(INTEL_BTS_PMU_NAME " does not support per-cpu recording\n");
148	148	return -EINVAL;
149	149	}
@@ -224,7 +224,7 @@ static int intel_bts_recording_options(struct auxtrace_record *itr, · 74 unmodified lines
224	224	* In the case of per-cpu mmaps, we need the CPU on the
225	225	* AUX event.
226	226	*/
227		− if (!perf_cpu_map__empty(cpus))
	227	+ if (!perf_cpu_map__has_any_cpu_or_is_empty(cpus))
228	228	evsel__set_sample_bit(intel_bts_evsel, CPU);
229	229	}
230	230

tools/perf/arch/x86/util/intel-pt.c+5−5

@@ -369,7 +369,7 @@ static int intel_pt_info_fill(struct auxtrace_record *itr, · 368 unmodified lines
369	369	ui__warning("Intel Processor Trace: TSC not available\n");
370	370	}
371	371
372		− per_cpu_mmaps = !perf_cpu_map__empty(session->evlist->core.user_requested_cpus);
	372	+ per_cpu_mmaps = !perf_cpu_map__has_any_cpu_or_is_empty(session->evlist->core.user_requested_cpus);
373	373
374	374	auxtrace_info->type = PERF_AUXTRACE_INTEL_PT;
375	375	auxtrace_info->priv[INTEL_PT_PMU_TYPE] = intel_pt_pmu->type;
@@ -774,7 +774,7 @@ static int intel_pt_recording_options(struct auxtrace_record *itr, · 398 unmodified lines
774	774	* Per-cpu recording needs sched_switch events to distinguish different
775	775	* threads.
776	776	*/
777		− if (have_timing_info && !perf_cpu_map__empty(cpus) &&
	777	+ if (have_timing_info && !perf_cpu_map__has_any_cpu_or_is_empty(cpus) &&
778	778	!record_opts__no_switch_events(opts)) {
779	779	if (perf_can_record_switch_events()) {
780	780	bool cpu_wide = !target__none(&opts->target) &&
@@ -832,7 +832,7 @@ static int intel_pt_recording_options(struct auxtrace_record *itr, · 51 unmodified lines
832	832	* In the case of per-cpu mmaps, we need the CPU on the
833	833	* AUX event.
834	834	*/
835		− if (!perf_cpu_map__empty(cpus))
	835	+ if (!perf_cpu_map__has_any_cpu_or_is_empty(cpus))
836	836	evsel__set_sample_bit(intel_pt_evsel, CPU);
837	837	}
838	838
@@ -858,7 +858,7 @@ static int intel_pt_recording_options(struct auxtrace_record *itr, · 19 unmodified lines
858	858	tracking_evsel->immediate = true;
859	859
860	860	/* In per-cpu case, always need the time of mmap events etc */
861		− if (!perf_cpu_map__empty(cpus)) {
	861	+ if (!perf_cpu_map__has_any_cpu_or_is_empty(cpus)) {
862	862	evsel__set_sample_bit(tracking_evsel, TIME);
863	863	/* And the CPU for switch events */
864	864	evsel__set_sample_bit(tracking_evsel, CPU);
@@ -870,7 +870,7 @@ static int intel_pt_recording_options(struct auxtrace_record *itr, · 5 unmodified lines
870	870	* Warn the user when we do not have enough information to decode i.e.
871	871	* per-cpu with no sched_switch (except workload-only).
872	872	*/
873		− if (!ptr->have_sched_switch && !perf_cpu_map__empty(cpus) &&
	873	+ if (!ptr->have_sched_switch && !perf_cpu_map__has_any_cpu_or_is_empty(cpus) &&
874	874	!target__none(&opts->target) &&
875	875	!intel_pt_evsel->core.attr.exclude_user)
876	876	ui__warning("Intel Processor Trace decoding will not be possible except for kernel tracing!\n");

tools/perf/bench/epoll-ctl.c+1−1

@@ -330,7 +330,7 @@ int bench_epoll_ctl(int argc, const char **argv) · 329 unmodified lines
330	330	act.sa_sigaction = toggle_done;
331	331	sigaction(SIGINT, &act, NULL);
332	332
333		− cpu = perf_cpu_map__new(NULL);
	333	+ cpu = perf_cpu_map__new_online_cpus();
334	334	if (!cpu)
335	335	goto errmem;
336	336

tools/perf/bench/epoll-wait.c+1−1

@@ -444,7 +444,7 @@ int bench_epoll_wait(int argc, const char **argv) · 443 unmodified lines
444	444	act.sa_sigaction = toggle_done;
445	445	sigaction(SIGINT, &act, NULL);
446	446
447		− cpu = perf_cpu_map__new(NULL);
	447	+ cpu = perf_cpu_map__new_online_cpus();
448	448	if (!cpu)
449	449	goto errmem;
450	450

tools/perf/bench/futex-hash.c+1−1

@@ -138,7 +138,7 @@ int bench_futex_hash(int argc, const char **argv) · 137 unmodified lines
138	138	exit(EXIT_FAILURE);
139	139	}
140	140
141		− cpu = perf_cpu_map__new(NULL);
	141	+ cpu = perf_cpu_map__new_online_cpus();
142	142	if (!cpu)
143	143	goto errmem;
144	144

tools/perf/bench/futex-lock-pi.c+1−1

@@ -172,7 +172,7 @@ int bench_futex_lock_pi(int argc, const char **argv) · 171 unmodified lines
172	172	if (argc)
173	173	goto err;
174	174
175		− cpu = perf_cpu_map__new(NULL);
	175	+ cpu = perf_cpu_map__new_online_cpus();
176	176	if (!cpu)
177	177	err(EXIT_FAILURE, "calloc");
178	178

tools/perf/bench/futex-requeue.c+1−1

@@ -174,7 +174,7 @@ int bench_futex_requeue(int argc, const char **argv) · 173 unmodified lines
174	174	if (argc)
175	175	goto err;
176	176
177		− cpu = perf_cpu_map__new(NULL);
	177	+ cpu = perf_cpu_map__new_online_cpus();
178	178	if (!cpu)
179	179	err(EXIT_FAILURE, "cpu_map__new");
180	180

tools/perf/bench/futex-wake-parallel.c+1−1

@@ -264,7 +264,7 @@ int bench_futex_wake_parallel(int argc, const char **argv) · 263 unmodified lines
264	264	err(EXIT_FAILURE, "mlockall");
265	265	}
266	266
267		− cpu = perf_cpu_map__new(NULL);
	267	+ cpu = perf_cpu_map__new_online_cpus();
268	268	if (!cpu)
269	269	err(EXIT_FAILURE, "calloc");
270	270

tools/perf/bench/futex-wake.c+1−1

@@ -149,7 +149,7 @@ int bench_futex_wake(int argc, const char **argv) · 148 unmodified lines
149	149	exit(EXIT_FAILURE);
150	150	}
151	151
152		− cpu = perf_cpu_map__new(NULL);
	152	+ cpu = perf_cpu_map__new_online_cpus();
153	153	if (!cpu)
154	154	err(EXIT_FAILURE, "calloc");
155	155

tools/perf/bench/sched-seccomp-notify.c+1−1

@@ -32,7 +32,7 @@ static bool sync_mode; · 31 unmodified lines
32	32	static const struct option options[] = {
33	33	OPT_U64('l', "loop", &loops, "Specify number of loops"),
34	34	OPT_BOOLEAN('s', "sync-mode", &sync_mode,
35		− "Enable the synchronious mode for seccomp notifications"),
	35	+ "Enable the synchronous mode for seccomp notifications"),
36	36	OPT_END()
37	37	};
38	38

tools/perf/builtin-annotate.c+272−28

@@ -20,6 +20,7 @@ · 19 unmodified lines
20	20	#include "util/evlist.h"
21	21	#include "util/evsel.h"
22	22	#include "util/annotate.h"
	23	+#include "util/annotate-data.h"
23	24	#include "util/event.h"
24	25	#include <subcmd/parse-options.h>
25	26	#include "util/parse-events.h"
@@ -45,7 +46,6 @@ · 19 unmodified lines
45	46	struct perf_annotate {
46	47	struct perf_tool tool;
47	48	struct perf_session *session;
48		− struct annotation_options opts;
49	49	#ifdef HAVE_SLANG_SUPPORT
50	50	bool use_tui;
51	51	#endif
@@ -56,9 +56,13 @@ struct perf_annotate { · 4 unmodified lines
56	56	bool skip_missing;
57	57	bool has_br_stack;
58	58	bool group_set;
	59	+ bool data_type;
	60	+ bool type_stat;
	61	+ bool insn_stat;
59	62	float min_percent;
60	63	const char *sym_hist_filter;
61	64	const char *cpu_list;
	65	+ const char *target_data_type;
62	66	DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
63	67	};
64	68
@@ -94,6 +98,7 @@ static void process_basic_block(struct addr_map_symbol *start, · 29 unmodified lines
94	98	struct annotation *notes = sym ? symbol__annotation(sym) : NULL;
95	99	struct block_range_iter iter;
96	100	struct block_range *entry;
	101	+ struct annotated_branch *branch;
97	102
98	103	/*
99	104	* Sanity; NULL isn't executable and the CPU cannot execute backwards
@@ -105,6 +110,8 @@ static void process_basic_block(struct addr_map_symbol *start, · 5 unmodified lines
105	110	if (!block_range_iter__valid(&iter))
106	111	return;
107	112
	113	+ branch = annotation__get_branch(notes);
	114	+
108	115	/*
109	116	* First block in range is a branch target.
110	117	*/
@@ -118,8 +125,8 @@ static void process_basic_block(struct addr_map_symbol *start, · 7 unmodified lines
118	125	entry->coverage++;
119	126	entry->sym = sym;
120	127
121		− if (notes)
122		− notes->max_coverage = max(notes->max_coverage, entry->coverage);
	128	+ if (branch)
	129	+ branch->max_coverage = max(branch->max_coverage, entry->coverage);
123	130
124	131	} while (block_range_iter__next(&iter));
125	132
@@ -315,9 +322,153 @@ static int hist_entry__tty_annotate(struct hist_entry *he, · 189 unmodified lines
315	322	struct perf_annotate *ann)
316	323	{
317	324	if (!ann->use_stdio2)
318		− return symbol__tty_annotate(&he->ms, evsel, &ann->opts);
	325	+ return symbol__tty_annotate(&he->ms, evsel);
	326	+
	327	+ return symbol__tty_annotate2(&he->ms, evsel);
	328	+}
	329	+
	330	+static void print_annotated_data_header(struct hist_entry he, struct evsel evsel)
	331	+{
	332	+ struct dso *dso = map__dso(he->ms.map);
	333	+ int nr_members = 1;
	334	+ int nr_samples = he->stat.nr_events;
	335	+
	336	+ if (evsel__is_group_event(evsel)) {
	337	+ struct hist_entry *pair;
	338	+
	339	+ list_for_each_entry(pair, &he->pairs.head, pairs.node)
	340	+ nr_samples += pair->stat.nr_events;
	341	+ }
	342	+
	343	+ printf("Annotate type: '%s' in %s (%d samples):\n",
	344	+ he->mem_type->self.type_name, dso->name, nr_samples);
	345	+
	346	+ if (evsel__is_group_event(evsel)) {
	347	+ struct evsel *pos;
	348	+ int i = 0;
	349	+
	350	+ for_each_group_evsel(pos, evsel)
	351	+ printf(" event[%d] = %s\n", i++, pos->name);
	352	+
	353	+ nr_members = evsel->core.nr_members;
	354	+ }
	355	+
	356	+ printf("============================================================================\n");
	357	+ printf("%s %10s %10s %s\n", 11 nr_members, "samples", "offset", "size", "field");
	358	+}
	359	+
	360	+static void print_annotated_data_type(struct annotated_data_type *mem_type,
	361	+ struct annotated_member *member,
	362	+ struct evsel *evsel, int indent)
	363	+{
	364	+ struct annotated_member *child;
	365	+ struct type_hist *h = mem_type->histograms[evsel->core.idx];
	366	+ int i, nr_events = 1, samples = 0;
	367	+
	368	+ for (i = 0; i < member->size; i++)
	369	+ samples += h->addr[member->offset + i].nr_samples;
	370	+ printf(" %10d", samples);
319	371
320		− return symbol__tty_annotate2(&he->ms, evsel, &ann->opts);
	372	+ if (evsel__is_group_event(evsel)) {
	373	+ struct evsel *pos;
	374	+
	375	+ for_each_group_member(pos, evsel) {
	376	+ h = mem_type->histograms[pos->core.idx];
	377	+
	378	+ samples = 0;
	379	+ for (i = 0; i < member->size; i++)
	380	+ samples += h->addr[member->offset + i].nr_samples;
	381	+ printf(" %10d", samples);
	382	+ }
	383	+ nr_events = evsel->core.nr_members;
	384	+ }
	385	+
	386	+ printf(" %10d %10d %*s%s\t%s",
	387	+ member->offset, member->size, indent, "", member->type_name,
	388	+ member->var_name ?: "");
	389	+
	390	+ if (!list_empty(&member->children))
	391	+ printf(" {\n");
	392	+
	393	+ list_for_each_entry(child, &member->children, node)
	394	+ print_annotated_data_type(mem_type, child, evsel, indent + 4);
	395	+
	396	+ if (!list_empty(&member->children))
	397	+ printf("%s}", 11 nr_events + 24 + indent, "");
	398	+ printf(";\n");
	399	+}
	400	+
	401	+static void print_annotate_data_stat(struct annotated_data_stat *s)
	402	+{
	403	+#define PRINT_STAT(fld) if (s->fld) printf("%10d : %s\n", s->fld, #fld)
	404	+
	405	+ int bad = s->no_sym +
	406	+ s->no_insn +
	407	+ s->no_insn_ops +
	408	+ s->no_mem_ops +
	409	+ s->no_reg +
	410	+ s->no_dbginfo +
	411	+ s->no_cuinfo +
	412	+ s->no_var +
	413	+ s->no_typeinfo +
	414	+ s->invalid_size +
	415	+ s->bad_offset;
	416	+ int ok = s->total - bad;
	417	+
	418	+ printf("Annotate data type stats:\n");
	419	+ printf("total %d, ok %d (%.1f%%), bad %d (%.1f%%)\n",
	420	+ s->total, ok, 100.0 * ok / (s->total ?: 1), bad, 100.0 * bad / (s->total ?: 1));
	421	+ printf("-----------------------------------------------------------\n");
	422	+ PRINT_STAT(no_sym);
	423	+ PRINT_STAT(no_insn);
	424	+ PRINT_STAT(no_insn_ops);
	425	+ PRINT_STAT(no_mem_ops);
	426	+ PRINT_STAT(no_reg);
	427	+ PRINT_STAT(no_dbginfo);
	428	+ PRINT_STAT(no_cuinfo);
	429	+ PRINT_STAT(no_var);
	430	+ PRINT_STAT(no_typeinfo);
	431	+ PRINT_STAT(invalid_size);
	432	+ PRINT_STAT(bad_offset);
	433	+ printf("\n");
	434	+
	435	+#undef PRINT_STAT
	436	+}
	437	+
	438	+static void print_annotate_item_stat(struct list_head head, const char title)
	439	+{
	440	+ struct annotated_item_stat istat, pos, *iter;
	441	+ int total_good, total_bad, total;
	442	+ int sum1, sum2;
	443	+ LIST_HEAD(tmp);
	444	+
	445	+ /* sort the list by count */
	446	+ list_splice_init(head, &tmp);
	447	+ total_good = total_bad = 0;
	448	+
	449	+ list_for_each_entry_safe(istat, pos, &tmp, list) {
	450	+ total_good += istat->good;
	451	+ total_bad += istat->bad;
	452	+ sum1 = istat->good + istat->bad;
	453	+
	454	+ list_for_each_entry(iter, head, list) {
	455	+ sum2 = iter->good + iter->bad;
	456	+ if (sum1 > sum2)
	457	+ break;
	458	+ }
	459	+ list_move_tail(&istat->list, &iter->list);
	460	+ }
	461	+ total = total_good + total_bad;
	462	+
	463	+ printf("Annotate %s stats\n", title);
	464	+ printf("total %d, ok %d (%.1f%%), bad %d (%.1f%%)\n\n", total,
	465	+ total_good, 100.0 * total_good / (total ?: 1),
	466	+ total_bad, 100.0 * total_bad / (total ?: 1));
	467	+ printf(" %-10s: %5s %5s\n", "Name", "Good", "Bad");
	468	+ printf("-----------------------------------------------------------\n");
	469	+ list_for_each_entry(istat, head, list)
	470	+ printf(" %-10s: %5d %5d\n", istat->name, istat->good, istat->bad);
	471	+ printf("\n");
321	472	}
322	473
323	474	static void hists__find_annotations(struct hists *hists,
@@ -327,6 +478,11 @@ static void hists__find_annotations(struct hists *hists, · 3 unmodified lines
327	478	struct rb_node nd = rb_first_cached(&hists->entries), next;
328	479	int key = K_RIGHT;
329	480
	481	+ if (ann->type_stat)
	482	+ print_annotate_data_stat(&ann_data_stat);
	483	+ if (ann->insn_stat)
	484	+ print_annotate_item_stat(&ann_insn_stat, "Instruction");
	485	+
330	486	while (nd) {
331	487	struct hist_entry *he = rb_entry(nd, struct hist_entry, rb_node);
332	488	struct annotation *notes;
@@ -359,11 +515,38 @@ find_next: · 26 unmodified lines
359	515	continue;
360	516	}
361	517
	518	+ if (ann->data_type) {
	519	+ /* skip unknown type */
	520	+ if (he->mem_type->histograms == NULL)
	521	+ goto find_next;
	522	+
	523	+ if (ann->target_data_type) {
	524	+ const char *type_name = he->mem_type->self.type_name;
	525	+
	526	+ /* skip 'struct ' prefix in the type name */
	527	+ if (strncmp(ann->target_data_type, "struct ", 7) &&
	528	+ !strncmp(type_name, "struct ", 7))
	529	+ type_name += 7;
	530	+
	531	+ /* skip 'union ' prefix in the type name */
	532	+ if (strncmp(ann->target_data_type, "union ", 6) &&
	533	+ !strncmp(type_name, "union ", 6))
	534	+ type_name += 6;
	535	+
	536	+ if (strcmp(ann->target_data_type, type_name))
	537	+ goto find_next;
	538	+ }
	539	+
	540	+ print_annotated_data_header(he, evsel);
	541	+ print_annotated_data_type(he->mem_type, &he->mem_type->self, evsel, 0);
	542	+ printf("\n");
	543	+ goto find_next;
	544	+ }
	545	+
362	546	if (use_browser == 2) {
363	547	int ret;
364	548	int (annotate)(struct hist_entry he,
365	549	struct evsel *evsel,
366		− struct annotation_options *options,
367	550	struct hist_browser_timer *hbt);
368	551
369	552	annotate = dlsym(perf_gtk_handle,
@@ -373,14 +556,14 @@ find_next: · 3 unmodified lines
373	556	return;
374	557	}
375	558
376		− ret = annotate(he, evsel, &ann->opts, NULL);
	559	+ ret = annotate(he, evsel, NULL);
377	560	if (!ret \|\| !ann->skip_missing)
378	561	return;
379	562
380	563	/* skip missing symbols */
381	564	nd = rb_next(nd);
382	565	} else if (use_browser == 1) {
383		− key = hist_entry__tui_annotate(he, evsel, NULL, &ann->opts);
	566	+ key = hist_entry__tui_annotate(he, evsel, NULL);
384	567
385	568	switch (key) {
386	569	case -1:
@@ -422,9 +605,9 @@ static int __cmd_annotate(struct perf_annotate *ann) · 35 unmodified lines
422	605	goto out;
423	606	}
424	607
425		− if (!ann->opts.objdump_path) {
	608	+ if (!annotate_opts.objdump_path) {
426	609	ret = perf_env__lookup_objdump(&session->header.env,
427		− &ann->opts.objdump_path);
	610	+ &annotate_opts.objdump_path);
428	611	if (ret)
429	612	goto out;
430	613	}
@@ -457,8 +640,20 @@ static int __cmd_annotate(struct perf_annotate *ann) · 26 unmodified lines
457	640	evsel__reset_sample_bit(pos, CALLCHAIN);
458	641	evsel__output_resort(pos, NULL);
459	642
460		− if (symbol_conf.event_group && !evsel__is_group_leader(pos))
	643	+ /*
	644	+ * An event group needs to display other events too.
	645	+ * Let's delay printing until other events are processed.
	646	+ */
	647	+ if (symbol_conf.event_group) {
	648	+ if (!evsel__is_group_leader(pos)) {
	649	+ struct hists *leader_hists;
	650	+
	651	+ leader_hists = evsel__hists(evsel__leader(pos));
	652	+ hists__match(leader_hists, hists);
	653	+ hists__link(leader_hists, hists);
	654	+ }
461	655	continue;
	656	+ }
462	657
463	658	hists__find_annotations(hists, pos, ann);
464	659	}
@@ -469,6 +664,20 @@ static int __cmd_annotate(struct perf_annotate *ann) · 4 unmodified lines
469	664	goto out;
470	665	}
471	666
	667	+ /* Display group events together */
	668	+ evlist__for_each_entry(session->evlist, pos) {
	669	+ struct hists *hists = evsel__hists(pos);
	670	+ u32 nr_samples = hists->stats.nr_samples;
	671	+
	672	+ if (nr_samples == 0)
	673	+ continue;
	674	+
	675	+ if (!symbol_conf.event_group \|\| !evsel__is_group_leader(pos))
	676	+ continue;
	677	+
	678	+ hists__find_annotations(hists, pos, ann);
	679	+ }
	680	+
472	681	if (use_browser == 2) {
473	682	void (*show_annotations)(void);
474	683
@@ -495,6 +704,17 @@ static int parse_percent_limit(const struct option opt, const char str, · 20 unmodified lines
495	704	return 0;
496	705	}
497	706
	707	+static int parse_data_type(const struct option opt, const char str, int unset)
	708	+{
	709	+ struct perf_annotate *ann = opt->value;
	710	+
	711	+ ann->data_type = !unset;
	712	+ if (str)
	713	+ ann->target_data_type = strdup(str);
	714	+
	715	+ return 0;
	716	+}
	717	+
498	718	static const char * const annotate_usage[] = {
499	719	"perf annotate [<options>]",
500	720	NULL
@@ -558,9 +778,9 @@ int cmd_annotate(int argc, const char **argv) · 57 unmodified lines
558	778	"file", "vmlinux pathname"),
559	779	OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules,
560	780	"load module symbols - WARNING: use only with -k and LIVE kernel"),
561		− OPT_BOOLEAN('l', "print-line", &annotate.opts.print_lines,
	781	+ OPT_BOOLEAN('l', "print-line", &annotate_opts.print_lines,
562	782	"print matching source lines (may be slow)"),
563		− OPT_BOOLEAN('P', "full-paths", &annotate.opts.full_path,
	783	+ OPT_BOOLEAN('P', "full-paths", &annotate_opts.full_path,
564	784	"Don't shorten the displayed pathnames"),
565	785	OPT_BOOLEAN(0, "skip-missing", &annotate.skip_missing,
566	786	"Skip symbols that cannot be annotated"),
@@ -571,15 +791,15 @@ int cmd_annotate(int argc, const char **argv) · 4 unmodified lines
571	791	OPT_CALLBACK(0, "symfs", NULL, "directory",
572	792	"Look for files with symbols relative to this directory",
573	793	symbol__config_symfs),
574		− OPT_BOOLEAN(0, "source", &annotate.opts.annotate_src,
	794	+ OPT_BOOLEAN(0, "source", &annotate_opts.annotate_src,
575	795	"Interleave source code with assembly code (default)"),
576		− OPT_BOOLEAN(0, "asm-raw", &annotate.opts.show_asm_raw,
	796	+ OPT_BOOLEAN(0, "asm-raw", &annotate_opts.show_asm_raw,
577	797	"Display raw encoding of assembly instructions (default)"),
578	798	OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style",
579	799	"Specify disassembler style (e.g. -M intel for intel syntax)"),
580		− OPT_STRING(0, "prefix", &annotate.opts.prefix, "prefix",
	800	+ OPT_STRING(0, "prefix", &annotate_opts.prefix, "prefix",
581	801	"Add prefix to source file path names in programs (with --prefix-strip)"),
582		− OPT_STRING(0, "prefix-strip", &annotate.opts.prefix_strip, "N",
	802	+ OPT_STRING(0, "prefix-strip", &annotate_opts.prefix_strip, "N",
583	803	"Strip first N entries of source file path name in programs (with --prefix)"),
584	804	OPT_STRING(0, "objdump", &objdump_path, "path",
585	805	"objdump binary to use for disassembly and annotations"),
@@ -598,7 +818,7 @@ int cmd_annotate(int argc, const char **argv) · 12 unmodified lines
598	818	OPT_CALLBACK_DEFAULT(0, "stdio-color", NULL, "mode",
599	819	"'always' (default), 'never' or 'auto' only applicable to --stdio mode",
600	820	stdio__config_color, "always"),
601		− OPT_CALLBACK(0, "percent-type", &annotate.opts, "local-period",
	821	+ OPT_CALLBACK(0, "percent-type", &annotate_opts, "local-period",
602	822	"Set percent type local/global-period/hits",
603	823	annotate_parse_percent_type),
604	824	OPT_CALLBACK(0, "percent-limit", &annotate, "percent",
@@ -606,7 +826,13 @@ int cmd_annotate(int argc, const char **argv) · 1 unmodified line
606	826	OPT_CALLBACK_OPTARG(0, "itrace", &itrace_synth_opts, NULL, "opts",
607	827	"Instruction Tracing options\n" ITRACE_HELP,
608	828	itrace_parse_synth_opts),
609		−
	829	+ OPT_CALLBACK_OPTARG(0, "data-type", &annotate, NULL, "name",
	830	+ "Show data type annotate for the memory accesses",
	831	+ parse_data_type),
	832	+ OPT_BOOLEAN(0, "type-stat", &annotate.type_stat,
	833	+ "Show stats for the data type annotation"),
	834	+ OPT_BOOLEAN(0, "insn-stat", &annotate.insn_stat,
	835	+ "Show instruction stats for the data type annotation"),
610	836	OPT_END()
611	837	};
612	838	int ret;
@@ -614,13 +840,13 @@ int cmd_annotate(int argc, const char **argv) · 1 unmodified line
614	840	set_option_flag(options, 0, "show-total-period", PARSE_OPT_EXCLUSIVE);
615	841	set_option_flag(options, 0, "show-nr-samples", PARSE_OPT_EXCLUSIVE);
616	842
617		− annotation_options__init(&annotate.opts);
	843	+ annotation_options__init();
618	844
619	845	ret = hists__init();
620	846	if (ret < 0)
621	847	return ret;
622	848
623		− annotation_config__init(&annotate.opts);
	849	+ annotation_config__init();
624	850
625	851	argc = parse_options(argc, argv, options, annotate_usage, 0);
626	852	if (argc) {
@@ -635,13 +861,13 @@ int cmd_annotate(int argc, const char **argv) · 8 unmodified lines
635	861	}
636	862
637	863	if (disassembler_style) {
638		− annotate.opts.disassembler_style = strdup(disassembler_style);
639		− if (!annotate.opts.disassembler_style)
	864	+ annotate_opts.disassembler_style = strdup(disassembler_style);
	865	+ if (!annotate_opts.disassembler_style)
640	866	return -ENOMEM;
641	867	}
642	868	if (objdump_path) {
643		− annotate.opts.objdump_path = strdup(objdump_path);
644		− if (!annotate.opts.objdump_path)
	869	+ annotate_opts.objdump_path = strdup(objdump_path);
	870	+ if (!annotate_opts.objdump_path)
645	871	return -ENOMEM;
646	872	}
647	873	if (addr2line_path) {
@@ -650,7 +876,7 @@ int cmd_annotate(int argc, const char **argv) · 2 unmodified lines
650	876	return -ENOMEM;
651	877	}
652	878
653		− if (annotate_check_args(&annotate.opts) < 0)
	879	+ if (annotate_check_args() < 0)
654	880	return -EINVAL;
655	881
656	882	#ifdef HAVE_GTK2_SUPPORT
@@ -660,6 +886,13 @@ int cmd_annotate(int argc, const char **argv) · 3 unmodified lines
660	886	}
661	887	#endif
662	888
	889	+#ifndef HAVE_DWARF_GETLOCATIONS_SUPPORT
	890	+ if (annotate.data_type) {
	891	+ pr_err("Error: Data type profiling is disabled due to missing DWARF support\n");
	892	+ return -ENOTSUP;
	893	+ }
	894	+#endif
	895	+
663	896	ret = symbol__validate_sym_arguments();
664	897	if (ret)
665	898	return ret;
@@ -702,6 +935,14 @@ int cmd_annotate(int argc, const char **argv) · 36 unmodified lines
702	935	use_browser = 2;
703	936	#endif
704	937
	938	+ /* FIXME: only support stdio for now */
	939	+ if (annotate.data_type) {
	940	+ use_browser = 0;
	941	+ annotate_opts.annotate_src = false;
	942	+ symbol_conf.annotate_data_member = true;
	943	+ symbol_conf.annotate_data_sample = true;
	944	+ }
	945	+
705	946	setup_browser(true);
706	947
707	948	/*
@@ -709,7 +950,10 @@ int cmd_annotate(int argc, const char **argv) · 1 unmodified line
709	950	* symbol, we do not care about the processes in annotate,
710	951	* set sort order to avoid repeated output.
711	952	*/
712		− sort_order = "dso,symbol";
	953	+ if (annotate.data_type)
	954	+ sort_order = "dso,type";
	955	+ else
	956	+ sort_order = "dso,symbol";
713	957
714	958	/*
715	959	* Set SORT_MODE__BRANCH so that annotate display IPC/Cycle
@@ -731,7 +975,7 @@ out_delete: · 15 unmodified lines
731	975	#ifndef NDEBUG
732	976	perf_session__delete(annotate.session);
733	977	#endif
734		− annotation_options__exit(&annotate.opts);
	978	+ annotation_options__exit();
735	979
736	980	return ret;
737	981	}

tools/perf/builtin-c2c.c+1−1

@@ -2320,7 +2320,7 @@ static int setup_nodes(struct perf_session *session) · 2319 unmodified lines
2320	2320	nodes[node] = set;
2321	2321
2322	2322	/* empty node, skip */
2323		− if (perf_cpu_map__empty(map))
	2323	+ if (perf_cpu_map__has_any_cpu_or_is_empty(map))
2324	2324	continue;
2325	2325
2326	2326	perf_cpu_map__for_each_cpu(cpu, idx, map) {

tools/perf/builtin-ftrace.c+1−1

@@ -333,7 +333,7 @@ static int set_tracing_func_irqinfo(struct perf_ftrace *ftrace) · 332 unmodified lines
333	333
334	334	static int reset_tracing_cpu(void)
335	335	{
336		− struct perf_cpu_map *cpumap = perf_cpu_map__new(NULL);
	336	+ struct perf_cpu_map *cpumap = perf_cpu_map__new_online_cpus();
337	337	int ret;
338	338
339	339	ret = set_tracing_cpumask(cpumap);

tools/perf/builtin-inject.c+6−0

@@ -2265,6 +2265,12 @@ int cmd_inject(int argc, const char **argv) · 2264 unmodified lines
2265	2265	"perf inject [<options>]",
2266	2266	NULL
2267	2267	};
	2268	+
	2269	+ if (!inject.itrace_synth_opts.set) {
	2270	+ /* Disable eager loading of kernel symbols that adds overhead to perf inject. */
	2271	+ symbol_conf.lazy_load_kernel_maps = true;
	2272	+ }
	2273	+
2268	2274	#ifndef HAVE_JITDUMP
2269	2275	set_option_nobuild(options, 'j', "jit", "NO_LIBELF=1", true);
2270	2276	#endif

tools/perf/builtin-lock.c+3−1

@@ -2285,8 +2285,10 @@ setup_args: · 2284 unmodified lines
2285	2285	else
2286	2286	ev_name = strdup(contention_tracepoints[j].name);
2287	2287
2288		− if (!ev_name)
	2288	+ if (!ev_name) {
	2289	+ free(rec_argv);
2289	2290	return -ENOMEM;
	2291	+ }
2290	2292
2291	2293	rec_argv[i++] = "-e";
2292	2294	rec_argv[i++] = ev_name;

tools/perf/builtin-record.c+45−48

@@ -270,7 +270,7 @@ static int record__write(struct record rec, struct mmap map __maybe_unused, · 269 unmodified lines
270	270
271	271	static int record__aio_enabled(struct record *rec);
272	272	static int record__comp_enabled(struct record *rec);
273		−static size_t zstd_compress(struct perf_session session, struct mmap map,
	273	+static ssize_t zstd_compress(struct perf_session session, struct mmap map,
274	274	void dst, size_t dst_size, void src, size_t src_size);
275	275
276	276	#ifdef HAVE_AIO_SUPPORT
@@ -405,9 +405,13 @@ static int record__aio_pushfn(struct mmap map, void to, void *buf, size_t size · 128 unmodified lines
405	405	*/
406	406
407	407	if (record__comp_enabled(aio->rec)) {
408		− size = zstd_compress(aio->rec->session, NULL, aio->data + aio->size,
409		− mmap__mmap_len(map) - aio->size,
410		− buf, size);
	408	+ ssize_t compressed = zstd_compress(aio->rec->session, NULL, aio->data + aio->size,
	409	+ mmap__mmap_len(map) - aio->size,
	410	+ buf, size);
	411	+ if (compressed < 0)
	412	+ return (int)compressed;
	413	+
	414	+ size = compressed;
411	415	} else {
412	416	memcpy(aio->data + aio->size, buf, size);
413	417	}
@@ -633,7 +637,13 @@ static int record__pushfn(struct mmap map, void to, void *bf, size_t size) · 219 unmodified lines
633	637	struct record *rec = to;
634	638
635	639	if (record__comp_enabled(rec)) {
636		− size = zstd_compress(rec->session, map, map->data, mmap__mmap_len(map), bf, size);
	640	+ ssize_t compressed = zstd_compress(rec->session, map, map->data,
	641	+ mmap__mmap_len(map), bf, size);
	642	+
	643	+ if (compressed < 0)
	644	+ return (int)compressed;
	645	+
	646	+ size = compressed;
637	647	bf = map->data;
638	648	}
639	649
@@ -1350,7 +1360,7 @@ static int record__open(struct record *rec) · 710 unmodified lines
1350	1360	evlist__for_each_entry(evlist, pos) {
1351	1361	try_again:
1352	1362	if (evsel__open(pos, pos->core.cpus, pos->core.threads) < 0) {
1353		− if (evsel__fallback(pos, errno, msg, sizeof(msg))) {
	1363	+ if (evsel__fallback(pos, &opts->target, errno, msg, sizeof(msg))) {
1354	1364	if (verbose > 0)
1355	1365	ui__warning("%s\n", msg);
1356	1366	goto try_again;
@@ -1527,10 +1537,10 @@ static size_t process_comp_header(void *record, size_t increment) · 170 unmodified lines
1527	1537	return size;
1528	1538	}
1529	1539
1530		−static size_t zstd_compress(struct perf_session session, struct mmap map,
	1540	+static ssize_t zstd_compress(struct perf_session session, struct mmap map,
1531	1541	void dst, size_t dst_size, void src, size_t src_size)
1532	1542	{
1533		− size_t compressed;
	1543	+ ssize_t compressed;
1534	1544	size_t max_record_size = PERF_SAMPLE_MAX_SIZE - sizeof(struct perf_record_compressed) - 1;
1535	1545	struct zstd_data *zstd_data = &session->zstd_data;
1536	1546
@@ -1539,6 +1549,8 @@ static size_t zstd_compress(struct perf_session session, struct mmap map, · 2 unmodified lines
1539	1549
1540	1550	compressed = zstd_compress_stream_to_records(zstd_data, dst, dst_size, src, src_size,
1541	1551	max_record_size, process_comp_header);
	1552	+ if (compressed < 0)
	1553	+ return compressed;
1542	1554
1543	1555	if (map && map->file) {
1544	1556	thread->bytes_transferred += src_size;
@@ -1912,21 +1924,13 @@ static void __record__save_lost_samples(struct record rec, struct evsel evsel, · 367 unmodified lines
1912	1924	static void record__read_lost_samples(struct record *rec)
1913	1925	{
1914	1926	struct perf_session *session = rec->session;
1915		− struct perf_record_lost_samples *lost;
	1927	+ struct perf_record_lost_samples *lost = NULL;
1916	1928	struct evsel *evsel;
1917	1929
1918	1930	/* there was an error during record__open */
1919	1931	if (session->evlist == NULL)
1920	1932	return;
1921	1933
1922		− lost = zalloc(PERF_SAMPLE_MAX_SIZE);
1923		− if (lost == NULL) {
1924		− pr_debug("Memory allocation failed\n");
1925		− return;
1926		− }
1927		−
1928		− lost->header.type = PERF_RECORD_LOST_SAMPLES;
1929		−
1930	1934	evlist__for_each_entry(session->evlist, evsel) {
1931	1935	struct xyarray *xy = evsel->core.sample_id;
1932	1936	u64 lost_count;
@@ -1949,6 +1953,15 @@ static void record__read_lost_samples(struct record *rec) · 16 unmodified lines
1949	1953	}
1950	1954
1951	1955	if (count.lost) {
	1956	+ if (!lost) {
	1957	+ lost = zalloc(sizeof(*lost) +
	1958	+ session->machines.host.id_hdr_size);
	1959	+ if (!lost) {
	1960	+ pr_debug("Memory allocation failed\n");
	1961	+ return;
	1962	+ }
	1963	+ lost->header.type = PERF_RECORD_LOST_SAMPLES;
	1964	+ }
1952	1965	__record__save_lost_samples(rec, evsel, lost,
1953	1966	x, y, count.lost, 0);
1954	1967	}
@@ -1956,9 +1969,19 @@ static void record__read_lost_samples(struct record *rec) · 1 unmodified line
1956	1969	}
1957	1970
1958	1971	lost_count = perf_bpf_filter__lost_count(evsel);
1959		− if (lost_count)
	1972	+ if (lost_count) {
	1973	+ if (!lost) {
	1974	+ lost = zalloc(sizeof(*lost) +
	1975	+ session->machines.host.id_hdr_size);
	1976	+ if (!lost) {
	1977	+ pr_debug("Memory allocation failed\n");
	1978	+ return;
	1979	+ }
	1980	+ lost->header.type = PERF_RECORD_LOST_SAMPLES;
	1981	+ }
1960	1982	__record__save_lost_samples(rec, evsel, lost, 0, 0, lost_count,
1961	1983	PERF_RECORD_MISC_LOST_SAMPLES_BPF);
	1984	+ }
1962	1985	}
1963	1986	out:
1964	1987	free(lost);
@@ -2216,32 +2239,6 @@ static void hit_auxtrace_snapshot_trigger(struct record *rec) · 251 unmodified lines
2216	2239	}
2217	2240	}
2218	2241
2219		−static void record__uniquify_name(struct record *rec)
2220		−{
2221		− struct evsel *pos;
2222		− struct evlist *evlist = rec->evlist;
2223		− char *new_name;
2224		− int ret;
2225		−
2226		− if (perf_pmus__num_core_pmus() == 1)
2227		− return;
2228		−
2229		− evlist__for_each_entry(evlist, pos) {
2230		− if (!evsel__is_hybrid(pos))
2231		− continue;
2232		−
2233		− if (strchr(pos->name, '/'))
2234		− continue;
2235		−
2236		− ret = asprintf(&new_name, "%s/%s/",
2237		− pos->pmu_name, pos->name);
2238		− if (ret) {
2239		− free(pos->name);
2240		− pos->name = new_name;
2241		− }
2242		− }
2243		−}
2244		−
2245	2242	static int record__terminate_thread(struct record_thread *thread_data)
2246	2243	{
2247	2244	int err;
@@ -2475,7 +2472,7 @@ static int __cmd_record(struct record rec, int argc, const char *argv) · 227 unmodified lines
2475	2472	if (data->is_pipe && rec->evlist->core.nr_entries == 1)
2476	2473	rec->opts.sample_id = true;
2477	2474
2478		− record__uniquify_name(rec);
	2475	+ evlist__uniquify_name(rec->evlist);
2479	2476
2480	2477	/* Debug message used by test scripts */
2481	2478	pr_debug3("perf record opening and mmapping events\n");
@@ -3580,9 +3577,7 @@ static int record__mmap_cpu_mask_init(struct mmap_cpu_mask *mask, struct perf_cp · 1098 unmodified lines
3580	3577	if (cpu_map__is_dummy(cpus))
3581	3578	return 0;
3582	3579
3583		− perf_cpu_map__for_each_cpu(cpu, idx, cpus) {
3584		− if (cpu.cpu == -1)
3585		− continue;
	3580	+ perf_cpu_map__for_each_cpu_skip_any(cpu, idx, cpus) {
3586	3581	/* Return ENODEV is input cpu is greater than max cpu */
3587	3582	if ((unsigned long)cpu.cpu > mask->nbits)
3588	3583	return -ENODEV;
@@ -3989,6 +3984,8 @@ int cmd_record(int argc, const char **argv) · 400 unmodified lines
3989	3984	# undef set_nobuild
3990	3985	#endif
3991	3986
	3987	+ /* Disable eager loading of kernel symbols that adds overhead to perf record. */
	3988	+ symbol_conf.lazy_load_kernel_maps = true;
3992	3989	rec->opts.affinity = PERF_AFFINITY_SYS;
3993	3990
3994	3991	rec->evlist = evlist__new();

tools/perf/builtin-report.c+74−39

@@ -96,9 +96,9 @@ struct report { · 95 unmodified lines
96	96	bool stitch_lbr;
97	97	bool disable_order;
98	98	bool skip_empty;
	99	+ bool data_type;
99	100	int max_stack;
100	101	struct perf_read_values show_threads_values;
101		− struct annotation_options annotation_opts;
102	102	const char *pretty_printing_style;
103	103	const char *cpu_list;
104	104	const char *symbol_filter_str;
@@ -171,7 +171,7 @@ static int hist_iter__report_callback(struct hist_entry_iter *iter, · 66 unmodified lines
171	171	struct mem_info *mi;
172	172	struct branch_info *bi;
173	173
174		− if (!ui__has_annotation() && !rep->symbol_ipc)
	174	+ if (!ui__has_annotation() && !rep->symbol_ipc && !rep->data_type)
175	175	return 0;
176	176
177	177	if (sort__mode == SORT_MODE__BRANCH) {
@@ -541,8 +541,7 @@ static int evlist__tui_block_hists_browse(struct evlist evlist, struct report · 363 unmodified lines
541	541	evlist__for_each_entry(evlist, pos) {
542	542	ret = report__browse_block_hists(&rep->block_reports[i++].hist,
543	543	rep->min_percent, pos,
544		− &rep->session->header.env,
545		− &rep->annotation_opts);
	544	+ &rep->session->header.env);
546	545	if (ret != 0)
547	546	return ret;
548	547	}
@@ -574,8 +573,7 @@ static int evlist__tty_browse_hists(struct evlist evlist, struct report rep, c · 25 unmodified lines
574	573
575	574	if (rep->total_cycles_mode) {
576	575	report__browse_block_hists(&rep->block_reports[i++].hist,
577		− rep->min_percent, pos,
578		− NULL, NULL);
	576	+ rep->min_percent, pos, NULL);
579	577	continue;
580	578	}
581	579
@@ -670,7 +668,7 @@ static int report__browse_hists(struct report *rep) · 88 unmodified lines
670	668	}
671	669
672	670	ret = evlist__tui_browse_hists(evlist, help, NULL, rep->min_percent,
673		− &session->header.env, true, &rep->annotation_opts);
	671	+ &session->header.env, true);
674	672	/*
675	673	* Usually "ret" is the last pressed key, and we only
676	674	* care if the key notifies us to switch data file.
@@ -745,7 +743,7 @@ static int hists__resort_cb(struct hist_entry he, void arg) · 68 unmodified lines
745	743	if (rep->symbol_ipc && sym && !sym->annotate2) {
746	744	struct evsel *evsel = hists_to_evsel(he->hists);
747	745
748		− symbol__annotate2(&he->ms, evsel, &rep->annotation_opts, NULL);
	746	+ symbol__annotate2(&he->ms, evsel, NULL);
749	747	}
750	748
751	749	return 0;
@@ -859,27 +857,47 @@ static struct task tasks_list(struct task task, struct machine *machine) · 107 unmodified lines
859	857	return tasks_list(parent_task, machine);
860	858	}
861	859
862		−static size_t maps__fprintf_task(struct maps maps, int indent, FILE fp)
	860	+struct maps__fprintf_task_args {
	861	+ int indent;
	862	+ FILE *fp;
	863	+ size_t printed;
	864	+};
	865	+
	866	+static int maps__fprintf_task_cb(struct map map, void data)
863	867	{
864		− size_t printed = 0;
865		− struct map_rb_node *rb_node;
	868	+ struct maps__fprintf_task_args *args = data;
	869	+ const struct dso *dso = map__dso(map);
	870	+ u32 prot = map__prot(map);
	871	+ int ret;
866	872
867		− maps__for_each_entry(maps, rb_node) {
868		− struct map *map = rb_node->map;
869		− const struct dso *dso = map__dso(map);
870		− u32 prot = map__prot(map);
	873	+ ret = fprintf(args->fp,
	874	+ "%*s %" PRIx64 "-%" PRIx64 " %c%c%c%c %08" PRIx64 " %" PRIu64 " %s\n",
	875	+ args->indent, "", map__start(map), map__end(map),
	876	+ prot & PROT_READ ? 'r' : '-',
	877	+ prot & PROT_WRITE ? 'w' : '-',
	878	+ prot & PROT_EXEC ? 'x' : '-',
	879	+ map__flags(map) ? 's' : 'p',
	880	+ map__pgoff(map),
	881	+ dso->id.ino, dso->name);
871	882
872		− printed += fprintf(fp, "%*s %" PRIx64 "-%" PRIx64 " %c%c%c%c %08" PRIx64 " %" PRIu64 " %s\n",
873		− indent, "", map__start(map), map__end(map),
874		− prot & PROT_READ ? 'r' : '-',
875		− prot & PROT_WRITE ? 'w' : '-',
876		− prot & PROT_EXEC ? 'x' : '-',
877		− map__flags(map) ? 's' : 'p',
878		− map__pgoff(map),
879		− dso->id.ino, dso->name);
880		− }
	883	+ if (ret < 0)
	884	+ return ret;
	885	+
	886	+ args->printed += ret;
	887	+ return 0;
	888	+}
	889	+
	890	+static size_t maps__fprintf_task(struct maps maps, int indent, FILE fp)
	891	+{
	892	+ struct maps__fprintf_task_args args = {
	893	+ .indent = indent,
	894	+ .fp = fp,
	895	+ .printed = 0,
	896	+ };
881	897
882		− return printed;
	898	+ maps__for_each_map(maps, maps__fprintf_task_cb, &args);
	899	+
	900	+ return args.printed;
883	901	}
884	902
885	903	static void task__print_level(struct task task, FILE fp, int level)
@@ -1341,15 +1359,15 @@ int cmd_report(int argc, const char **argv) · 455 unmodified lines
1341	1359	"list of cpus to profile"),
1342	1360	OPT_BOOLEAN('I', "show-info", &report.show_full_info,
1343	1361	"Display extended information about perf.data file"),
1344		− OPT_BOOLEAN(0, "source", &report.annotation_opts.annotate_src,
	1362	+ OPT_BOOLEAN(0, "source", &annotate_opts.annotate_src,
1345	1363	"Interleave source code with assembly code (default)"),
1346		− OPT_BOOLEAN(0, "asm-raw", &report.annotation_opts.show_asm_raw,
	1364	+ OPT_BOOLEAN(0, "asm-raw", &annotate_opts.show_asm_raw,
1347	1365	"Display raw encoding of assembly instructions (default)"),
1348	1366	OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style",
1349	1367	"Specify disassembler style (e.g. -M intel for intel syntax)"),
1350		− OPT_STRING(0, "prefix", &report.annotation_opts.prefix, "prefix",
	1368	+ OPT_STRING(0, "prefix", &annotate_opts.prefix, "prefix",
1351	1369	"Add prefix to source file path names in programs (with --prefix-strip)"),
1352		− OPT_STRING(0, "prefix-strip", &report.annotation_opts.prefix_strip, "N",
	1370	+ OPT_STRING(0, "prefix-strip", &annotate_opts.prefix_strip, "N",
1353	1371	"Strip first N entries of source file path name in programs (with --prefix)"),
1354	1372	OPT_BOOLEAN(0, "show-total-period", &symbol_conf.show_total_period,
1355	1373	"Show a column with the sum of periods"),
@@ -1401,7 +1419,7 @@ int cmd_report(int argc, const char **argv) · 45 unmodified lines
1401	1419	"Time span of interest (start,stop)"),
1402	1420	OPT_BOOLEAN(0, "inline", &symbol_conf.inline_name,
1403	1421	"Show inline function"),
1404		− OPT_CALLBACK(0, "percent-type", &report.annotation_opts, "local-period",
	1422	+ OPT_CALLBACK(0, "percent-type", &annotate_opts, "local-period",
1405	1423	"Set percent type local/global-period/hits",
1406	1424	annotate_parse_percent_type),
1407	1425	OPT_BOOLEAN(0, "ns", &symbol_conf.nanosecs, "Show times in nanosecs"),
@@ -1426,7 +1444,14 @@ int cmd_report(int argc, const char **argv) · 18 unmodified lines
1426	1444	if (ret < 0)
1427	1445	goto exit;
1428	1446
1429		− annotation_options__init(&report.annotation_opts);
	1447	+ /*
	1448	+ * tasks_mode require access to exited threads to list those that are in
	1449	+ * the data file. Off-cpu events are synthesized after other events and
	1450	+ * reference exited threads.
	1451	+ */
	1452	+ symbol_conf.keep_exited_threads = true;
	1453	+
	1454	+ annotation_options__init();
1430	1455
1431	1456	ret = perf_config(report__config, &report);
1432	1457	if (ret)
@@ -1445,13 +1470,13 @@ int cmd_report(int argc, const char **argv) · 12 unmodified lines
1445	1470	}
1446	1471
1447	1472	if (disassembler_style) {
1448		− report.annotation_opts.disassembler_style = strdup(disassembler_style);
1449		− if (!report.annotation_opts.disassembler_style)
	1473	+ annotate_opts.disassembler_style = strdup(disassembler_style);
	1474	+ if (!annotate_opts.disassembler_style)
1450	1475	return -ENOMEM;
1451	1476	}
1452	1477	if (objdump_path) {
1453		− report.annotation_opts.objdump_path = strdup(objdump_path);
1454		− if (!report.annotation_opts.objdump_path)
	1478	+ annotate_opts.objdump_path = strdup(objdump_path);
	1479	+ if (!annotate_opts.objdump_path)
1455	1480	return -ENOMEM;
1456	1481	}
1457	1482	if (addr2line_path) {
@@ -1460,7 +1485,7 @@ int cmd_report(int argc, const char **argv) · 2 unmodified lines
1460	1485	return -ENOMEM;
1461	1486	}
1462	1487
1463		− if (annotate_check_args(&report.annotation_opts) < 0) {
	1488	+ if (annotate_check_args() < 0) {
1464	1489	ret = -EINVAL;
1465	1490	goto exit;
1466	1491	}
@@ -1615,6 +1640,16 @@ repeat: · 148 unmodified lines
1615	1640	sort_order = NULL;
1616	1641	}
1617	1642
	1643	+ if (sort_order && strstr(sort_order, "type")) {
	1644	+ report.data_type = true;
	1645	+ annotate_opts.annotate_src = false;
	1646	+
	1647	+#ifndef HAVE_DWARF_GETLOCATIONS_SUPPORT
	1648	+ pr_err("Error: Data type profiling is disabled due to missing DWARF support\n");
	1649	+ goto error;
	1650	+#endif
	1651	+ }
	1652	+
1618	1653	if (strcmp(input_name, "-") != 0)
1619	1654	setup_browser(true);
1620	1655	else
@@ -1673,7 +1708,7 @@ repeat: · 52 unmodified lines
1673	1708	* so don't allocate extra space that won't be used in the stdio
1674	1709	* implementation.
1675	1710	*/
1676		− if (ui__has_annotation() \|\| report.symbol_ipc \|\|
	1711	+ if (ui__has_annotation() \|\| report.symbol_ipc \|\| report.data_type \|\|
1677	1712	report.total_cycles_mode) {
1678	1713	ret = symbol__annotation_init();
1679	1714	if (ret < 0)
@@ -1692,7 +1727,7 @@ repeat: · 12 unmodified lines
1692	1727	*/
1693	1728	symbol_conf.priv_size += sizeof(u32);
1694	1729	}
1695		− annotation_config__init(&report.annotation_opts);
	1730	+ annotation_config__init();
1696	1731	}
1697	1732
1698	1733	if (symbol__init(&session->header.env) < 0)
@@ -1746,7 +1781,7 @@ error: · 47 unmodified lines
1746	1781	zstd_fini(&(session->zstd_data));
1747	1782	perf_session__delete(session);
1748	1783	exit:
1749		− annotation_options__exit(&report.annotation_opts);
	1784	+ annotation_options__exit();
1750	1785	free(sort_order_help);
1751	1786	free(field_order_help);
1752	1787	return ret;

tools/perf/builtin-stat.c+19−14

@@ -653,7 +653,7 @@ static enum counter_recovery stat_handle_error(struct evsel *counter) · 652 unmodified lines
653	653	if ((evsel__leader(counter) != counter) \|\|
654	654	!(counter->core.leader->nr_members > 1))
655	655	return COUNTER_SKIP;
656		− } else if (evsel__fallback(counter, errno, msg, sizeof(msg))) {
	656	+ } else if (evsel__fallback(counter, &target, errno, msg, sizeof(msg))) {
657	657	if (verbose > 0)
658	658	ui__warning("%s\n", msg);
659	659	return COUNTER_RETRY;
@@ -1204,8 +1204,9 @@ static struct option stat_options[] = { · 544 unmodified lines
1204	1204	OPT_STRING('C', "cpu", &target.cpu_list, "cpu",
1205	1205	"list of cpus to monitor in system-wide"),
1206	1206	OPT_SET_UINT('A', "no-aggr", &stat_config.aggr_mode,
1207		− "disable CPU count aggregation", AGGR_NONE),
1208		− OPT_BOOLEAN(0, "no-merge", &stat_config.no_merge, "Do not merge identical named events"),
	1207	+ "disable aggregation across CPUs or PMUs", AGGR_NONE),
	1208	+ OPT_SET_UINT(0, "no-merge", &stat_config.aggr_mode,
	1209	+ "disable aggregation the same as -A or -no-aggr", AGGR_NONE),
1209	1210	OPT_BOOLEAN(0, "hybrid-merge", &stat_config.hybrid_merge,
1210	1211	"Merge identical named hybrid events"),
1211	1212	OPT_STRING('x', "field-separator", &stat_config.csv_sep, "separator",
@@ -1255,7 +1256,7 @@ static struct option stat_options[] = { · 43 unmodified lines
1255	1256	OPT_BOOLEAN(0, "metric-no-merge", &stat_config.metric_no_merge,
1256	1257	"don't try to share events between metrics in a group"),
1257	1258	OPT_BOOLEAN(0, "metric-no-threshold", &stat_config.metric_no_threshold,
1258		− "don't try to share events between metrics in a group "),
	1259	+ "disable adding events for the metric threshold calculation"),
1259	1260	OPT_BOOLEAN(0, "topdown", &topdown_run,
1260	1261	"measure top-down statistics"),
1261	1262	OPT_UINTEGER(0, "td-level", &stat_config.topdown_level,
@@ -1316,7 +1317,7 @@ static int cpu__get_cache_id_from_map(struct perf_cpu cpu, char *map) · 54 unmodified lines
1316	1317	* be the first online CPU in the cache domain else use the
1317	1318	* first online CPU of the cache domain as the ID.
1318	1319	*/
1319		− if (perf_cpu_map__empty(cpu_map))
	1320	+ if (perf_cpu_map__has_any_cpu_or_is_empty(cpu_map))
1320	1321	id = cpu.cpu;
1321	1322	else
1322	1323	id = perf_cpu_map__cpu(cpu_map, 0).cpu;
@@ -1622,7 +1623,7 @@ static int perf_stat_init_aggr_mode(void) · 299 unmodified lines
1622	1623	* taking the highest cpu number to be the size of
1623	1624	* the aggregation translate cpumap.
1624	1625	*/
1625		− if (!perf_cpu_map__empty(evsel_list->core.user_requested_cpus))
	1626	+ if (!perf_cpu_map__has_any_cpu_or_is_empty(evsel_list->core.user_requested_cpus))
1626	1627	nr = perf_cpu_map__max(evsel_list->core.user_requested_cpus).cpu;
1627	1628	else
1628	1629	nr = 0;
@@ -2289,7 +2290,7 @@ int process_stat_config_event(struct perf_session *session, · 660 unmodified lines
2289	2290
2290	2291	perf_event__read_stat_config(&stat_config, &event->stat_config);
2291	2292
2292		− if (perf_cpu_map__empty(st->cpus)) {
	2293	+ if (perf_cpu_map__has_any_cpu_or_is_empty(st->cpus)) {
2293	2294	if (st->aggr_mode != AGGR_UNSET)
2294	2295	pr_warning("warning: processing task data, aggregation mode not set\n");
2295	2296	} else if (st->aggr_mode != AGGR_UNSET) {
@@ -2695,15 +2696,19 @@ int cmd_stat(int argc, const char **argv) · 399 unmodified lines
2695	2696	*/
2696	2697	if (metrics) {
2697	2698	const char *pmu = parse_events_option_args.pmu_filter ?: "all";
	2699	+ int ret = metricgroup__parse_groups(evsel_list, pmu, metrics,
	2700	+ stat_config.metric_no_group,
	2701	+ stat_config.metric_no_merge,
	2702	+ stat_config.metric_no_threshold,
	2703	+ stat_config.user_requested_cpu_list,
	2704	+ stat_config.system_wide,
	2705	+ &stat_config.metric_events);
2698	2706
2699		− metricgroup__parse_groups(evsel_list, pmu, metrics,
2700		− stat_config.metric_no_group,
2701		− stat_config.metric_no_merge,
2702		− stat_config.metric_no_threshold,
2703		− stat_config.user_requested_cpu_list,
2704		− stat_config.system_wide,
2705		− &stat_config.metric_events);
2706	2707	zfree(&metrics);
	2708	+ if (ret) {
	2709	+ status = ret;
	2710	+ goto out;
	2711	+ }
2707	2712	}
2708	2713
2709	2714	if (add_default_attributes())

tools/perf/builtin-top.c+26−26

@@ -147,7 +147,7 @@ static int perf_top__parse_source(struct perf_top top, struct hist_entry he) · 146 unmodified lines
147	147	return err;
148	148	}
149	149
150		− err = symbol__annotate(&he->ms, evsel, &top->annotation_opts, NULL);
	150	+ err = symbol__annotate(&he->ms, evsel, NULL);
151	151	if (err == 0) {
152	152	top->sym_filter_entry = he;
153	153	} else {
@@ -261,9 +261,9 @@ static void perf_top__show_details(struct perf_top *top) · 107 unmodified lines
261	261	goto out_unlock;
262	262
263	263	printf("Showing %s for %s\n", evsel__name(top->sym_evsel), symbol->name);
264		− printf(" Events Pcnt (>=%d%%)\n", top->annotation_opts.min_pcnt);
	264	+ printf(" Events Pcnt (>=%d%%)\n", annotate_opts.min_pcnt);
265	265
266		− more = symbol__annotate_printf(&he->ms, top->sym_evsel, &top->annotation_opts);
	266	+ more = symbol__annotate_printf(&he->ms, top->sym_evsel);
267	267
268	268	if (top->evlist->enabled) {
269	269	if (top->zero)
@@ -450,7 +450,7 @@ static void perf_top__print_mapped_keys(struct perf_top *top) · 180 unmodified lines
450	450
451	451	fprintf(stdout, "\t[f] profile display filter (count). \t(%d)\n", top->count_filter);
452	452
453		− fprintf(stdout, "\t[F] annotate display filter (percent). \t(%d%%)\n", top->annotation_opts.min_pcnt);
	453	+ fprintf(stdout, "\t[F] annotate display filter (percent). \t(%d%%)\n", annotate_opts.min_pcnt);
454	454	fprintf(stdout, "\t[s] annotate symbol. \t(%s)\n", name?: "NULL");
455	455	fprintf(stdout, "\t[S] stop annotation.\n");
456	456
@@ -553,7 +553,7 @@ static bool perf_top__handle_keypress(struct perf_top *top, int c) · 96 unmodified lines
553	553	prompt_integer(&top->count_filter, "Enter display event count filter");
554	554	break;
555	555	case 'F':
556		− prompt_percent(&top->annotation_opts.min_pcnt,
	556	+ prompt_percent(&annotate_opts.min_pcnt,
557	557	"Enter details display event filter (percent)");
558	558	break;
559	559	case 'K':
@@ -646,8 +646,7 @@ repeat: · 86 unmodified lines
646	646	}
647	647
648	648	ret = evlist__tui_browse_hists(top->evlist, help, &hbt, top->min_percent,
649		− &top->session->header.env, !top->record_opts.overwrite,
650		− &top->annotation_opts);
	649	+ &top->session->header.env, !top->record_opts.overwrite);
651	650	if (ret == K_RELOAD) {
652	651	top->zero = true;
653	652	goto repeat;
@@ -1027,8 +1026,8 @@ static int perf_top__start_counters(struct perf_top *top) · 373 unmodified lines
1027	1026
1028	1027	evlist__for_each_entry(evlist, counter) {
1029	1028	try_again:
1030		− if (evsel__open(counter, top->evlist->core.user_requested_cpus,
1031		− top->evlist->core.threads) < 0) {
	1029	+ if (evsel__open(counter, counter->core.cpus,
	1030	+ counter->core.threads) < 0) {
1032	1031
1033	1032	/*
1034	1033	* Specially handle overwrite fall back.
@@ -1044,7 +1043,7 @@ try_again: · 9 unmodified lines
1044	1043	perf_top_overwrite_fallback(top, counter))
1045	1044	goto try_again;
1046	1045
1047		− if (evsel__fallback(counter, errno, msg, sizeof(msg))) {
	1046	+ if (evsel__fallback(counter, &opts->target, errno, msg, sizeof(msg))) {
1048	1047	if (verbose > 0)
1049	1048	ui__warning("%s\n", msg);
1050	1049	goto try_again;
@@ -1241,9 +1240,9 @@ static int __cmd_top(struct perf_top *top) · 190 unmodified lines
1241	1240	pthread_t thread, thread_process;
1242	1241	int ret;
1243	1242
1244		− if (!top->annotation_opts.objdump_path) {
	1243	+ if (!annotate_opts.objdump_path) {
1245	1244	ret = perf_env__lookup_objdump(&top->session->header.env,
1246		− &top->annotation_opts.objdump_path);
	1245	+ &annotate_opts.objdump_path);
1247	1246	if (ret)
1248	1247	return ret;
1249	1248	}
@@ -1299,6 +1298,7 @@ static int __cmd_top(struct perf_top *top) · 49 unmodified lines
1299	1298	}
1300	1299	}
1301	1300
	1301	+ evlist__uniquify_name(top->evlist);
1302	1302	ret = perf_top__start_counters(top);
1303	1303	if (ret)
1304	1304	return ret;
@@ -1536,9 +1536,9 @@ int cmd_top(int argc, const char **argv) · 231 unmodified lines
1536	1536	"only consider symbols in these comms"),
1537	1537	OPT_STRING(0, "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]",
1538	1538	"only consider these symbols"),
1539		− OPT_BOOLEAN(0, "source", &top.annotation_opts.annotate_src,
	1539	+ OPT_BOOLEAN(0, "source", &annotate_opts.annotate_src,
1540	1540	"Interleave source code with assembly code (default)"),
1541		− OPT_BOOLEAN(0, "asm-raw", &top.annotation_opts.show_asm_raw,
	1541	+ OPT_BOOLEAN(0, "asm-raw", &annotate_opts.show_asm_raw,
1542	1542	"Display raw encoding of assembly instructions (default)"),
1543	1543	OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
1544	1544	"Enable kernel symbol demangling"),
@@ -1549,9 +1549,9 @@ int cmd_top(int argc, const char **argv) · 4 unmodified lines
1549	1549	"addr2line binary to use for line numbers"),
1550	1550	OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style",
1551	1551	"Specify disassembler style (e.g. -M intel for intel syntax)"),
1552		− OPT_STRING(0, "prefix", &top.annotation_opts.prefix, "prefix",
	1552	+ OPT_STRING(0, "prefix", &annotate_opts.prefix, "prefix",
1553	1553	"Add prefix to source file path names in programs (with --prefix-strip)"),
1554		− OPT_STRING(0, "prefix-strip", &top.annotation_opts.prefix_strip, "N",
	1554	+ OPT_STRING(0, "prefix-strip", &annotate_opts.prefix_strip, "N",
1555	1555	"Strip first N entries of source file path name in programs (with --prefix)"),
1556	1556	OPT_STRING('u', "uid", &target->uid_str, "user", "user to profile"),
1557	1557	OPT_CALLBACK(0, "percent-limit", &top, "percent",
@@ -1609,10 +1609,10 @@ int cmd_top(int argc, const char **argv) · 51 unmodified lines
1609	1609	if (status < 0)
1610	1610	return status;
1611	1611
1612		− annotation_options__init(&top.annotation_opts);
	1612	+ annotation_options__init();
1613	1613
1614		− top.annotation_opts.min_pcnt = 5;
1615		− top.annotation_opts.context = 4;
	1614	+ annotate_opts.min_pcnt = 5;
	1615	+ annotate_opts.context = 4;
1616	1616
1617	1617	top.evlist = evlist__new();
1618	1618	if (top.evlist == NULL)
@@ -1642,13 +1642,13 @@ int cmd_top(int argc, const char **argv) · 23 unmodified lines
1642	1642	usage_with_options(top_usage, options);
1643	1643
1644	1644	if (disassembler_style) {
1645		− top.annotation_opts.disassembler_style = strdup(disassembler_style);
1646		− if (!top.annotation_opts.disassembler_style)
	1645	+ annotate_opts.disassembler_style = strdup(disassembler_style);
	1646	+ if (!annotate_opts.disassembler_style)
1647	1647	return -ENOMEM;
1648	1648	}
1649	1649	if (objdump_path) {
1650		− top.annotation_opts.objdump_path = strdup(objdump_path);
1651		− if (!top.annotation_opts.objdump_path)
	1650	+ annotate_opts.objdump_path = strdup(objdump_path);
	1651	+ if (!annotate_opts.objdump_path)
1652	1652	return -ENOMEM;
1653	1653	}
1654	1654	if (addr2line_path) {
@@ -1661,7 +1661,7 @@ int cmd_top(int argc, const char **argv) · 6 unmodified lines
1661	1661	if (status)
1662	1662	goto out_delete_evlist;
1663	1663
1664		− if (annotate_check_args(&top.annotation_opts) < 0)
	1664	+ if (annotate_check_args() < 0)
1665	1665	goto out_delete_evlist;
1666	1666
1667	1667	if (!top.evlist->core.nr_entries) {
@@ -1787,7 +1787,7 @@ int cmd_top(int argc, const char **argv) · 119 unmodified lines
1787	1787	if (status < 0)
1788	1788	goto out_delete_evlist;
1789	1789
1790		− annotation_config__init(&top.annotation_opts);
	1790	+ annotation_config__init();
1791	1791
1792	1792	symbol_conf.try_vmlinux_path = (symbol_conf.vmlinux_name == NULL);
1793	1793	status = symbol__init(NULL);
@@ -1840,7 +1840,7 @@ int cmd_top(int argc, const char **argv) · 46 unmodified lines
1840	1840	out_delete_evlist:
1841	1841	evlist__delete(top.evlist);
1842	1842	perf_session__delete(top.session);
1843		− annotation_options__exit(&top.annotation_opts);
	1843	+ annotation_options__exit();
1844	1844
1845	1845	return status;
1846	1846	}

tools/perf/builtin-trace.c+2−4

@@ -2470,9 +2470,8 @@ static int trace__fprintf_callchain(struct trace trace, struct perf_sample sam · 2469 unmodified lines
2470	2470	static const char errno_to_name(struct evsel evsel, int err)
2471	2471	{
2472	2472	struct perf_env *env = evsel__env(evsel);
2473		− const char *arch_name = perf_env__arch(env);
2474	2473
2475		− return arch_syscalls__strerrno(arch_name, err);
	2474	+ return perf_env__arch_strerrno(env, err);
2476	2475	}
2477	2476
2478	2477	static int trace__sys_exit(struct trace trace, struct evsel evsel,
@@ -4264,12 +4263,11 @@ static size_t thread__dump_stats(struct thread_trace *ttrace, · 1785 unmodified lines
4264	4263	printed += fprintf(fp, " %9.3f %9.2f%%\n", max, pct);
4265	4264
4266	4265	if (trace->errno_summary && stats->nr_failures) {
4267		− const char *arch_name = perf_env__arch(trace->host->env);
4268	4266	int e;
4269	4267
4270	4268	for (e = 0; e < stats->max_errno; ++e) {
4271	4269	if (stats->errnos[e] != 0)
4272		− fprintf(fp, "\t\t\t\t%s: %d\n", arch_syscalls__strerrno(arch_name, e + 1), stats->errnos[e]);
	4270	+ fprintf(fp, "\t\t\t\t%s: %d\n", perf_env__arch_strerrno(trace->host->env, e + 1), stats->errnos[e]);
4273	4271	}
4274	4272	}
4275	4273	}

tools/perf/perf-archive.sh+80−6

@@ -4,8 +4,73 @@ · 3 unmodified lines
4	4	# Arnaldo Carvalho de Melo <acme@redhat.com>
5	5
6	6	PERF_DATA=perf.data
7		−if [ $# -ne 0 ] ; then
8		− PERF_DATA=$1
	7	+PERF_SYMBOLS=perf.symbols
	8	+PERF_ALL=perf.all
	9	+ALL=0
	10	+UNPACK=0
	11	+
	12	+while [ $# -gt 0 ] ; do
	13	+ if [ $1 == "--all" ]; then
	14	+ ALL=1
	15	+ shift
	16	+ elif [ $1 == "--unpack" ]; then
	17	+ UNPACK=1
	18	+ shift
	19	+ else
	20	+ PERF_DATA=$1
	21	+ UNPACK_TAR=$1
	22	+ shift
	23	+ fi
	24	+done
	25	+
	26	+if [ $UNPACK -eq 1 ]; then
	27	+ if [ ! -z "$UNPACK_TAR" ]; then # tar given as an argument
	28	+ if [ ! -e "$UNPACK_TAR" ]; then
	29	+ echo "Provided file $UNPACK_TAR does not exist"
	30	+ exit 1
	31	+ fi
	32	+ TARGET="$UNPACK_TAR"
	33	+ else # search for perf tar in the current directory
	34	+ TARGET=`find . -regex "\./perf.*\.tar\.bz2"`
	35	+ TARGET_NUM=`echo -n "$TARGET" \| grep -c '^'`
	36	+
	37	+ if [ -z "$TARGET" -o $TARGET_NUM -gt 1 ]; then
	38	+ echo -e "Error: $TARGET_NUM files found for unpacking:\n$TARGET"
	39	+ echo "Provide the requested file as an argument"
	40	+ exit 1
	41	+ else
	42	+ echo "Found target file for unpacking: $TARGET"
	43	+ fi
	44	+ fi
	45	+
	46	+ if [[ "$TARGET" =~ (\./)?$PERF_ALL.*.tar.bz2 ]]; then # perf tar generated by --all option
	47	+ TAR_CONTENTS=`tar tvf "$TARGET" \| tr -s " " \| cut -d " " -f 6`
	48	+ VALID_TAR=`echo "$TAR_CONTENTS" \| grep "$PERF_SYMBOLS.tar.bz2" \| wc -l` # check if it contains a sub-tar perf.symbols
	49	+ if [ $VALID_TAR -ne 1 ]; then
	50	+ echo "Error: $TARGET file is not valid (contains zero or multiple sub-tar files with debug symbols)"
	51	+ exit 1
	52	+ fi
	53	+
	54	+ INTERSECT=`comm -12 <(ls) <(echo "$TAR_CONTENTS") \| tr "\n" " "` # check for overwriting
	55	+ if [ ! -z "$INTERSECT" ]; then # prompt if file(s) already exist in the current directory
	56	+ echo "File(s) ${INTERSECT::-1} already exist in the current directory."
	57	+ while true; do
	58	+ read -p 'Do you wish to overwrite them? ' yn
	59	+ case $yn in
	60	+ [Yy]* ) break;;
	61	+ [Nn]* ) exit 1;;
	62	+ * ) echo "Please answer yes or no.";;
	63	+ esac
	64	+ done
	65	+ fi
	66	+
	67	+ # unzip the perf.data file in the current working directory and debug symbols in ~/.debug directory
	68	+ tar xvf $TARGET && tar xvf $PERF_SYMBOLS.tar.bz2 -C ~/.debug
	69	+
	70	+ else # perf tar generated by perf archive (contains only debug symbols)
	71	+ tar xvf $TARGET -C ~/.debug
	72	+ fi
	73	+ exit 0
9	74	fi
10	75
11	76	#
@@ -39,9 +104,18 @@ while read build_id ; do · 27 unmodified lines
39	104	echo ${filename#$PERF_BUILDID_LINKDIR} >> $MANIFEST
40	105	done
41	106
42		−tar cjf $PERF_DATA.tar.bz2 -C $PERF_BUILDID_DIR -T $MANIFEST
43		−rm $MANIFEST $BUILDIDS \|\| true
	107	+if [ $ALL -eq 1 ]; then # pack perf.data file together with tar containing debug symbols
	108	+ HOSTNAME=$(hostname)
	109	+ DATE=$(date '+%Y%m%d-%H%M%S')
	110	+ tar cjf $PERF_SYMBOLS.tar.bz2 -C $PERF_BUILDID_DIR -T $MANIFEST
	111	+ tar cjf $PERF_ALL-$HOSTNAME-$DATE.tar.bz2 $PERF_DATA $PERF_SYMBOLS.tar.bz2
	112	+ rm $PERF_SYMBOLS.tar.bz2 $MANIFEST $BUILDIDS \|\| true
	113	+else # pack only the debug symbols
	114	+ tar cjf $PERF_DATA.tar.bz2 -C $PERF_BUILDID_DIR -T $MANIFEST
	115	+ rm $MANIFEST $BUILDIDS \|\| true
	116	+fi
	117	+
44	118	echo -e "Now please run:\n"
45		−echo -e "$ tar xvf $PERF_DATA.tar.bz2 -C ~/.debug\n"
46		−echo "wherever you need to run 'perf report' on."
	119	+echo -e "$ perf archive --unpack\n"
	120	+echo "or unpack the tar manually wherever you need to run 'perf report' on."
47	121	exit 0

tools/perf/perf.c+30−0

@@ -39,6 +39,7 @@ · 38 unmodified lines
39	39	#include <linux/zalloc.h>
40	40
41	41	static int use_pager = -1;
	42	+static FILE *debug_fp = NULL;
42	43
43	44	struct cmd_struct {
44	45	const char *cmd;
@@ -162,6 +163,19 @@ static void commit_pager_choice(void) · 117 unmodified lines
162	163	}
163	164	}
164	165
	166	+static int set_debug_file(const char *path)
	167	+{
	168	+ debug_fp = fopen(path, "w");
	169	+ if (!debug_fp) {
	170	+ fprintf(stderr, "Open debug file '%s' failed: %s\n",
	171	+ path, strerror(errno));
	172	+ return -1;
	173	+ }
	174	+
	175	+ debug_set_file(debug_fp);
	176	+ return 0;
	177	+}
	178	+
165	179	struct option options[] = {
166	180	OPT_ARGUMENT("help", "help"),
167	181	OPT_ARGUMENT("version", "version"),
@@ -174,6 +188,7 @@ struct option options[] = { · 6 unmodified lines
174	188	OPT_ARGUMENT("list-cmds", "list-cmds"),
175	189	OPT_ARGUMENT("list-opts", "list-opts"),
176	190	OPT_ARGUMENT("debug", "debug"),
	191	+ OPT_ARGUMENT("debug-file", "debug-file"),
177	192	OPT_END()
178	193	};
179	194
@@ -287,6 +302,18 @@ static int handle_options(const char **argv, int argc, int *envchanged) · 107 unmodified lines
287	302
288	303	(*argv)++;
289	304	(*argc)--;
	305	+ } else if (!strcmp(cmd, "--debug-file")) {
	306	+ if (*argc < 2) {
	307	+ fprintf(stderr, "No path given for --debug-file.\n");
	308	+ usage(perf_usage_string);
	309	+ }
	310	+
	311	+ if (set_debug_file((*argv)[1]))
	312	+ usage(perf_usage_string);
	313	+
	314	+ (*argv)++;
	315	+ (*argc)--;
	316	+
290	317	} else {
291	318	fprintf(stderr, "Unknown option: %s\n", cmd);
292	319	usage(perf_usage_string);
@@ -547,5 +574,8 @@ int main(int argc, const char **argv) · 254 unmodified lines
547	574	fprintf(stderr, "Failed to run command '%s': %s\n",
548	575	cmd, str_error_r(errno, sbuf, sizeof(sbuf)));
549	576	out:
	577	+ if (debug_fp)
	578	+ fclose(debug_fp);
	579	+
550	580	return 1;
551	581	}

tools/perf/pmu-events/arch/arm64/ampere/ampereone/core-imp-def.json+1−1

@@ -110,7 +110,7 @@ · 109 unmodified lines
110	110	{
111	111	"PublicDescription": "Flushes due to memory hazards",
112	112	"EventCode": "0x121",
113		− "EventName": "BPU_FLUSH_MEM_FAULT",
	113	+ "EventName": "GPC_FLUSH_MEM_FAULT",
114	114	"BriefDescription": "Flushes due to memory hazards"
115	115	},
116	116	{

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/branch.json+125−0

@@ -0,0 +1,125 @@
	1	+[
	2	+ {
	3	+ "ArchStdEvent": "BR_IMMED_SPEC"
	4	+ },
	5	+ {
	6	+ "ArchStdEvent": "BR_RETURN_SPEC"
	7	+ },
	8	+ {
	9	+ "ArchStdEvent": "BR_INDIRECT_SPEC"
	10	+ },
	11	+ {
	12	+ "ArchStdEvent": "BR_MIS_PRED"
	13	+ },
	14	+ {
	15	+ "ArchStdEvent": "BR_PRED"
	16	+ },
	17	+ {
	18	+ "PublicDescription": "Instruction architecturally executed, branch not taken",
	19	+ "EventCode": "0x8107",
	20	+ "EventName": "BR_SKIP_RETIRED",
	21	+ "BriefDescription": "Instruction architecturally executed, branch not taken"
	22	+ },
	23	+ {
	24	+ "PublicDescription": "Instruction architecturally executed, immediate branch taken",
	25	+ "EventCode": "0x8108",
	26	+ "EventName": "BR_IMMED_TAKEN_RETIRED",
	27	+ "BriefDescription": "Instruction architecturally executed, immediate branch taken"
	28	+ },
	29	+ {
	30	+ "PublicDescription": "Instruction architecturally executed, indirect branch excluding return retired",
	31	+ "EventCode": "0x810c",
	32	+ "EventName": "BR_INDNR_TAKEN_RETIRED",
	33	+ "BriefDescription": "Instruction architecturally executed, indirect branch excluding return retired"
	34	+ },
	35	+ {
	36	+ "PublicDescription": "Instruction architecturally executed, predicted immediate branch",
	37	+ "EventCode": "0x8110",
	38	+ "EventName": "BR_IMMED_PRED_RETIRED",
	39	+ "BriefDescription": "Instruction architecturally executed, predicted immediate branch"
	40	+ },
	41	+ {
	42	+ "PublicDescription": "Instruction architecturally executed, mispredicted immediate branch",
	43	+ "EventCode": "0x8111",
	44	+ "EventName": "BR_IMMED_MIS_PRED_RETIRED",
	45	+ "BriefDescription": "Instruction architecturally executed, mispredicted immediate branch"
	46	+ },
	47	+ {
	48	+ "PublicDescription": "Instruction architecturally executed, predicted indirect branch",
	49	+ "EventCode": "0x8112",
	50	+ "EventName": "BR_IND_PRED_RETIRED",
	51	+ "BriefDescription": "Instruction architecturally executed, predicted indirect branch"
	52	+ },
	53	+ {
	54	+ "PublicDescription": "Instruction architecturally executed, mispredicted indirect branch",
	55	+ "EventCode": "0x8113",
	56	+ "EventName": "BR_IND_MIS_PRED_RETIRED",
	57	+ "BriefDescription": "Instruction architecturally executed, mispredicted indirect branch"
	58	+ },
	59	+ {
	60	+ "PublicDescription": "Instruction architecturally executed, predicted procedure return",
	61	+ "EventCode": "0x8114",
	62	+ "EventName": "BR_RETURN_PRED_RETIRED",
	63	+ "BriefDescription": "Instruction architecturally executed, predicted procedure return"
	64	+ },
	65	+ {
	66	+ "PublicDescription": "Instruction architecturally executed, mispredicted procedure return",
	67	+ "EventCode": "0x8115",
	68	+ "EventName": "BR_RETURN_MIS_PRED_RETIRED",
	69	+ "BriefDescription": "Instruction architecturally executed, mispredicted procedure return"
	70	+ },
	71	+ {
	72	+ "PublicDescription": "Instruction architecturally executed, predicted indirect branch excluding return",
	73	+ "EventCode": "0x8116",
	74	+ "EventName": "BR_INDNR_PRED_RETIRED",
	75	+ "BriefDescription": "Instruction architecturally executed, predicted indirect branch excluding return"
	76	+ },
	77	+ {
	78	+ "PublicDescription": "Instruction architecturally executed, mispredicted indirect branch excluding return",
	79	+ "EventCode": "0x8117",
	80	+ "EventName": "BR_INDNR_MIS_PRED_RETIRED",
	81	+ "BriefDescription": "Instruction architecturally executed, mispredicted indirect branch excluding return"
	82	+ },
	83	+ {
	84	+ "PublicDescription": "Instruction architecturally executed, predicted branch, taken",
	85	+ "EventCode": "0x8118",
	86	+ "EventName": "BR_TAKEN_PRED_RETIRED",
	87	+ "BriefDescription": "Instruction architecturally executed, predicted branch, taken"
	88	+ },
	89	+ {
	90	+ "PublicDescription": "Instruction architecturally executed, mispredicted branch, taken",
	91	+ "EventCode": "0x8119",
	92	+ "EventName": "BR_TAKEN_MIS_PRED_RETIRED",
	93	+ "BriefDescription": "Instruction architecturally executed, mispredicted branch, taken"
	94	+ },
	95	+ {
	96	+ "PublicDescription": "Instruction architecturally executed, predicted branch, not taken",
	97	+ "EventCode": "0x811a",
	98	+ "EventName": "BR_SKIP_PRED_RETIRED",
	99	+ "BriefDescription": "Instruction architecturally executed, predicted branch, not taken"
	100	+ },
	101	+ {
	102	+ "PublicDescription": "Instruction architecturally executed, mispredicted branch, not taken",
	103	+ "EventCode": "0x811b",
	104	+ "EventName": "BR_SKIP_MIS_PRED_RETIRED",
	105	+ "BriefDescription": "Instruction architecturally executed, mispredicted branch, not taken"
	106	+ },
	107	+ {
	108	+ "PublicDescription": "Instruction architecturally executed, predicted branch",
	109	+ "EventCode": "0x811c",
	110	+ "EventName": "BR_PRED_RETIRED",
	111	+ "BriefDescription": "Instruction architecturally executed, predicted branch"
	112	+ },
	113	+ {
	114	+ "PublicDescription": "Instruction architecturally executed, indirect branch",
	115	+ "EventCode": "0x811d",
	116	+ "EventName": "BR_IND_RETIRED",
	117	+ "BriefDescription": "Instruction architecturally executed, indirect branch"
	118	+ },
	119	+ {
	120	+ "PublicDescription": "Branch Record captured.",
	121	+ "EventCode": "0x811f",
	122	+ "EventName": "BRB_FILTRATE",
	123	+ "BriefDescription": "Branch Record captured."
	124	+ }
	125	+]

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/bus.json+20−0

@@ -0,0 +1,20 @@
	1	+[
	2	+ {
	3	+ "ArchStdEvent": "CPU_CYCLES"
	4	+ },
	5	+ {
	6	+ "ArchStdEvent": "BUS_CYCLES"
	7	+ },
	8	+ {
	9	+ "ArchStdEvent": "BUS_ACCESS_RD"
	10	+ },
	11	+ {
	12	+ "ArchStdEvent": "BUS_ACCESS_WR"
	13	+ },
	14	+ {
	15	+ "ArchStdEvent": "BUS_ACCESS"
	16	+ },
	17	+ {
	18	+ "ArchStdEvent": "CNT_CYCLES"
	19	+ }
	20	+]

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/cache.json+206−0

@@ -0,0 +1,206 @@
	1	+[
	2	+ {
	3	+ "ArchStdEvent": "L1D_CACHE_RD"
	4	+ },
	5	+ {
	6	+ "ArchStdEvent": "L1D_CACHE_WR"
	7	+ },
	8	+ {
	9	+ "ArchStdEvent": "L1D_CACHE_REFILL_RD"
	10	+ },
	11	+ {
	12	+ "ArchStdEvent": "L1D_CACHE_INVAL"
	13	+ },
	14	+ {
	15	+ "ArchStdEvent": "L1D_TLB_REFILL_RD"
	16	+ },
	17	+ {
	18	+ "ArchStdEvent": "L1D_TLB_REFILL_WR"
	19	+ },
	20	+ {
	21	+ "ArchStdEvent": "L2D_CACHE_RD"
	22	+ },
	23	+ {
	24	+ "ArchStdEvent": "L2D_CACHE_WR"
	25	+ },
	26	+ {
	27	+ "ArchStdEvent": "L2D_CACHE_REFILL_RD"
	28	+ },
	29	+ {
	30	+ "ArchStdEvent": "L2D_CACHE_REFILL_WR"
	31	+ },
	32	+ {
	33	+ "ArchStdEvent": "L2D_CACHE_WB_VICTIM"
	34	+ },
	35	+ {
	36	+ "ArchStdEvent": "L2D_CACHE_WB_CLEAN"
	37	+ },
	38	+ {
	39	+ "ArchStdEvent": "L2D_CACHE_INVAL"
	40	+ },
	41	+ {
	42	+ "ArchStdEvent": "L1I_CACHE_REFILL"
	43	+ },
	44	+ {
	45	+ "ArchStdEvent": "L1I_TLB_REFILL"
	46	+ },
	47	+ {
	48	+ "ArchStdEvent": "L1D_CACHE_REFILL"
	49	+ },
	50	+ {
	51	+ "ArchStdEvent": "L1D_CACHE"
	52	+ },
	53	+ {
	54	+ "ArchStdEvent": "L1D_TLB_REFILL"
	55	+ },
	56	+ {
	57	+ "ArchStdEvent": "L1I_CACHE"
	58	+ },
	59	+ {
	60	+ "ArchStdEvent": "L2D_CACHE"
	61	+ },
	62	+ {
	63	+ "ArchStdEvent": "L2D_CACHE_REFILL"
	64	+ },
	65	+ {
	66	+ "ArchStdEvent": "L2D_CACHE_WB"
	67	+ },
	68	+ {
	69	+ "ArchStdEvent": "L1D_TLB"
	70	+ },
	71	+ {
	72	+ "ArchStdEvent": "L1I_TLB"
	73	+ },
	74	+ {
	75	+ "ArchStdEvent": "L2D_TLB_REFILL"
	76	+ },
	77	+ {
	78	+ "ArchStdEvent": "L2I_TLB_REFILL"
	79	+ },
	80	+ {
	81	+ "ArchStdEvent": "L2D_TLB"
	82	+ },
	83	+ {
	84	+ "ArchStdEvent": "L2I_TLB"
	85	+ },
	86	+ {
	87	+ "ArchStdEvent": "DTLB_WALK"
	88	+ },
	89	+ {
	90	+ "ArchStdEvent": "ITLB_WALK"
	91	+ },
	92	+ {
	93	+ "ArchStdEvent": "L1D_CACHE_REFILL_WR"
	94	+ },
	95	+ {
	96	+ "ArchStdEvent": "L1D_CACHE_LMISS_RD"
	97	+ },
	98	+ {
	99	+ "ArchStdEvent": "L1I_CACHE_LMISS"
	100	+ },
	101	+ {
	102	+ "ArchStdEvent": "L2D_CACHE_LMISS_RD"
	103	+ },
	104	+ {
	105	+ "PublicDescription": "Level 1 data or unified cache demand access",
	106	+ "EventCode": "0x8140",
	107	+ "EventName": "L1D_CACHE_RW",
	108	+ "BriefDescription": "Level 1 data or unified cache demand access"
	109	+ },
	110	+ {
	111	+ "PublicDescription": "Level 1 data or unified cache preload or prefetch",
	112	+ "EventCode": "0x8142",
	113	+ "EventName": "L1D_CACHE_PRFM",
	114	+ "BriefDescription": "Level 1 data or unified cache preload or prefetch"
	115	+ },
	116	+ {
	117	+ "PublicDescription": "Level 1 data or unified cache refill, preload or prefetch",
	118	+ "EventCode": "0x8146",
	119	+ "EventName": "L1D_CACHE_REFILL_PRFM",
	120	+ "BriefDescription": "Level 1 data or unified cache refill, preload or prefetch"
	121	+ },
	122	+ {
	123	+ "ArchStdEvent": "L1D_TLB_RD"
	124	+ },
	125	+ {
	126	+ "ArchStdEvent": "L1D_TLB_WR"
	127	+ },
	128	+ {
	129	+ "ArchStdEvent": "L2D_TLB_REFILL_RD"
	130	+ },
	131	+ {
	132	+ "ArchStdEvent": "L2D_TLB_REFILL_WR"
	133	+ },
	134	+ {
	135	+ "ArchStdEvent": "L2D_TLB_RD"
	136	+ },
	137	+ {
	138	+ "ArchStdEvent": "L2D_TLB_WR"
	139	+ },
	140	+ {
	141	+ "PublicDescription": "L1D TLB miss",
	142	+ "EventCode": "0xD600",
	143	+ "EventName": "L1D_TLB_MISS",
	144	+ "BriefDescription": "L1D TLB miss"
	145	+ },
	146	+ {
	147	+ "PublicDescription": "Level 1 prefetcher, load prefetch requests generated",
	148	+ "EventCode": "0xd606",
	149	+ "EventName": "L1_PREFETCH_LD_GEN",
	150	+ "BriefDescription": "Level 1 prefetcher, load prefetch requests generated"
	151	+ },
	152	+ {
	153	+ "PublicDescription": "Level 1 prefetcher, load prefetch fills into the level 1 cache",
	154	+ "EventCode": "0xd607",
	155	+ "EventName": "L1_PREFETCH_LD_FILL",
	156	+ "BriefDescription": "Level 1 prefetcher, load prefetch fills into the level 1 cache"
	157	+ },
	158	+ {
	159	+ "PublicDescription": "Level 1 prefetcher, load prefetch to level 2 generated",
	160	+ "EventCode": "0xd608",
	161	+ "EventName": "L1_PREFETCH_L2_REQ",
	162	+ "BriefDescription": "Level 1 prefetcher, load prefetch to level 2 generated"
	163	+ },
	164	+ {
	165	+ "PublicDescription": "L1 prefetcher, distance was reset",
	166	+ "EventCode": "0xd609",
	167	+ "EventName": "L1_PREFETCH_DIST_RST",
	168	+ "BriefDescription": "L1 prefetcher, distance was reset"
	169	+ },
	170	+ {
	171	+ "PublicDescription": "L1 prefetcher, distance was increased",
	172	+ "EventCode": "0xd60a",
	173	+ "EventName": "L1_PREFETCH_DIST_INC",
	174	+ "BriefDescription": "L1 prefetcher, distance was increased"
	175	+ },
	176	+ {
	177	+ "PublicDescription": "Level 1 prefetcher, table entry is trained",
	178	+ "EventCode": "0xd60b",
	179	+ "EventName": "L1_PREFETCH_ENTRY_TRAINED",
	180	+ "BriefDescription": "Level 1 prefetcher, table entry is trained"
	181	+ },
	182	+ {
	183	+ "PublicDescription": "L1 data cache refill - Read or Write",
	184	+ "EventCode": "0xd60e",
	185	+ "EventName": "L1D_CACHE_REFILL_RW",
	186	+ "BriefDescription": "L1 data cache refill - Read or Write"
	187	+ },
	188	+ {
	189	+ "PublicDescription": "Level 2 cache refill from instruction-side miss, including IMMU refills",
	190	+ "EventCode": "0xD701",
	191	+ "EventName": "L2C_INST_REFILL",
	192	+ "BriefDescription": "Level 2 cache refill from instruction-side miss, including IMMU refills"
	193	+ },
	194	+ {
	195	+ "PublicDescription": "Level 2 cache refill from data-side miss, including DMMU refills",
	196	+ "EventCode": "0xD702",
	197	+ "EventName": "L2C_DATA_REFILL",
	198	+ "BriefDescription": "Level 2 cache refill from data-side miss, including DMMU refills"
	199	+ },
	200	+ {
	201	+ "PublicDescription": "Level 2 cache prefetcher, load prefetch requests generated",
	202	+ "EventCode": "0xD703",
	203	+ "EventName": "L2_PREFETCH_REQ",
	204	+ "BriefDescription": "Level 2 cache prefetcher, load prefetch requests generated"
	205	+ }
	206	+]

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/core-imp-def.json+464−0

@@ -0,0 +1,464 @@
	1	+[
	2	+ {
	3	+ "PublicDescription": "Level 2 prefetch requests, refilled to L2 cache",
	4	+ "EventCode": "0x10A",
	5	+ "EventName": "L2_PREFETCH_REFILL",
	6	+ "BriefDescription": "Level 2 prefetch requests, refilled to L2 cache"
	7	+ },
	8	+ {
	9	+ "PublicDescription": "Level 2 prefetch requests, late",
	10	+ "EventCode": "0x10B",
	11	+ "EventName": "L2_PREFETCH_UPGRADE",
	12	+ "BriefDescription": "Level 2 prefetch requests, late"
	13	+ },
	14	+ {
	15	+ "PublicDescription": "Predictable branch speculatively executed that hit any level of BTB",
	16	+ "EventCode": "0x110",
	17	+ "EventName": "BPU_HIT_BTB",
	18	+ "BriefDescription": "Predictable branch speculatively executed that hit any level of BTB"
	19	+ },
	20	+ {
	21	+ "PublicDescription": "Predictable conditional branch speculatively executed that hit any level of BTB",
	22	+ "EventCode": "0x111",
	23	+ "EventName": "BPU_CONDITIONAL_BRANCH_HIT_BTB",
	24	+ "BriefDescription": "Predictable conditional branch speculatively executed that hit any level of BTB"
	25	+ },
	26	+ {
	27	+ "PublicDescription": "Predictable taken branch speculatively executed that hit any level of BTB that access the indirect predictor",
	28	+ "EventCode": "0x112",
	29	+ "EventName": "BPU_HIT_INDIRECT_PREDICTOR",
	30	+ "BriefDescription": "Predictable taken branch speculatively executed that hit any level of BTB that access the indirect predictor"
	31	+ },
	32	+ {
	33	+ "PublicDescription": "Predictable taken branch speculatively executed that hit any level of BTB that access the return predictor",
	34	+ "EventCode": "0x113",
	35	+ "EventName": "BPU_HIT_RSB",
	36	+ "BriefDescription": "Predictable taken branch speculatively executed that hit any level of BTB that access the return predictor"
	37	+ },
	38	+ {
	39	+ "PublicDescription": "Predictable unconditional branch speculatively executed that did not hit any level of BTB",
	40	+ "EventCode": "0x114",
	41	+ "EventName": "BPU_UNCONDITIONAL_BRANCH_MISS_BTB",
	42	+ "BriefDescription": "Predictable unconditional branch speculatively executed that did not hit any level of BTB"
	43	+ },
	44	+ {
	45	+ "PublicDescription": "Predictable branch speculatively executed, unpredicted",
	46	+ "EventCode": "0x115",
	47	+ "EventName": "BPU_BRANCH_NO_HIT",
	48	+ "BriefDescription": "Predictable branch speculatively executed, unpredicted"
	49	+ },
	50	+ {
	51	+ "PublicDescription": "Predictable branch speculatively executed that hit any level of BTB that mispredict",
	52	+ "EventCode": "0x116",
	53	+ "EventName": "BPU_HIT_BTB_AND_MISPREDICT",
	54	+ "BriefDescription": "Predictable branch speculatively executed that hit any level of BTB that mispredict"
	55	+ },
	56	+ {
	57	+ "PublicDescription": "Predictable conditional branch speculatively executed that hit any level of BTB that (direction) mispredict",
	58	+ "EventCode": "0x117",
	59	+ "EventName": "BPU_CONDITIONAL_BRANCH_HIT_BTB_AND_MISPREDICT",
	60	+ "BriefDescription": "Predictable conditional branch speculatively executed that hit any level of BTB that (direction) mispredict"
	61	+ },
	62	+ {
	63	+ "PublicDescription": "Predictable taken branch speculatively executed that hit any level of BTB that access the indirect predictor that mispredict",
	64	+ "EventCode": "0x118",
	65	+ "EventName": "BPU_INDIRECT_BRANCH_HIT_BTB_AND_MISPREDICT",
	66	+ "BriefDescription": "Predictable taken branch speculatively executed that hit any level of BTB that access the indirect predictor that mispredict"
	67	+ },
	68	+ {
	69	+ "PublicDescription": "Predictable taken branch speculatively executed that hit any level of BTB that access the return predictor that mispredict",
	70	+ "EventCode": "0x119",
	71	+ "EventName": "BPU_HIT_RSB_AND_MISPREDICT",
	72	+ "BriefDescription": "Predictable taken branch speculatively executed that hit any level of BTB that access the return predictor that mispredict"
	73	+ },
	74	+ {
	75	+ "PublicDescription": "Predictable taken branch speculatively executed that hit any level of BTB that access the overflow/underflow return predictor that mispredict",
	76	+ "EventCode": "0x11a",
	77	+ "EventName": "BPU_MISS_RSB_AND_MISPREDICT",
	78	+ "BriefDescription": "Predictable taken branch speculatively executed that hit any level of BTB that access the overflow/underflow return predictor that mispredict"
	79	+ },
	80	+ {
	81	+ "PublicDescription": "Predictable branch speculatively executed, unpredicted, that mispredict",
	82	+ "EventCode": "0x11b",
	83	+ "EventName": "BPU_NO_PREDICTION_MISPREDICT",
	84	+ "BriefDescription": "Predictable branch speculatively executed, unpredicted, that mispredict"
	85	+ },
	86	+ {
	87	+ "PublicDescription": "Preditable branch update the BTB region buffer entry",
	88	+ "EventCode": "0x11c",
	89	+ "EventName": "BPU_BTB_UPDATE",
	90	+ "BriefDescription": "Preditable branch update the BTB region buffer entry"
	91	+ },
	92	+ {
	93	+ "PublicDescription": "Count predict pipe stalls due to speculative return address predictor full",
	94	+ "EventCode": "0x11d",
	95	+ "EventName": "BPU_RSB_FULL_STALL",
	96	+ "BriefDescription": "Count predict pipe stalls due to speculative return address predictor full"
	97	+ },
	98	+ {
	99	+ "PublicDescription": "Macro-ops speculatively decoded",
	100	+ "EventCode": "0x11f",
	101	+ "EventName": "ICF_INST_SPEC_DECODE",
	102	+ "BriefDescription": "Macro-ops speculatively decoded"
	103	+ },
	104	+ {
	105	+ "PublicDescription": "Flushes",
	106	+ "EventCode": "0x120",
	107	+ "EventName": "GPC_FLUSH",
	108	+ "BriefDescription": "Flushes"
	109	+ },
	110	+ {
	111	+ "PublicDescription": "Flushes due to memory hazards",
	112	+ "EventCode": "0x121",
	113	+ "EventName": "GPC_FLUSH_MEM_FAULT",
	114	+ "BriefDescription": "Flushes due to memory hazards"
	115	+ },
	116	+ {
	117	+ "PublicDescription": "ETM extout bit 0",
	118	+ "EventCode": "0x141",
	119	+ "EventName": "MSC_ETM_EXTOUT0",
	120	+ "BriefDescription": "ETM extout bit 0"
	121	+ },
	122	+ {
	123	+ "PublicDescription": "ETM extout bit 1",
	124	+ "EventCode": "0x142",
	125	+ "EventName": "MSC_ETM_EXTOUT1",
	126	+ "BriefDescription": "ETM extout bit 1"
	127	+ },
	128	+ {
	129	+ "PublicDescription": "ETM extout bit 2",
	130	+ "EventCode": "0x143",
	131	+ "EventName": "MSC_ETM_EXTOUT2",
	132	+ "BriefDescription": "ETM extout bit 2"
	133	+ },
	134	+ {
	135	+ "PublicDescription": "ETM extout bit 3",
	136	+ "EventCode": "0x144",
	137	+ "EventName": "MSC_ETM_EXTOUT3",
	138	+ "BriefDescription": "ETM extout bit 3"
	139	+ },
	140	+ {
	141	+ "PublicDescription": "Bus request sn",
	142	+ "EventCode": "0x156",
	143	+ "EventName": "L2C_SNOOP",
	144	+ "BriefDescription": "Bus request sn"
	145	+ },
	146	+ {
	147	+ "PublicDescription": "L2 TXDAT LCRD blocked",
	148	+ "EventCode": "0x169",
	149	+ "EventName": "L2C_DAT_CRD_STALL",
	150	+ "BriefDescription": "L2 TXDAT LCRD blocked"
	151	+ },
	152	+ {
	153	+ "PublicDescription": "L2 TXRSP LCRD blocked",
	154	+ "EventCode": "0x16a",
	155	+ "EventName": "L2C_RSP_CRD_STALL",
	156	+ "BriefDescription": "L2 TXRSP LCRD blocked"
	157	+ },
	158	+ {
	159	+ "PublicDescription": "L2 TXREQ LCRD blocked",
	160	+ "EventCode": "0x16b",
	161	+ "EventName": "L2C_REQ_CRD_STALL",
	162	+ "BriefDescription": "L2 TXREQ LCRD blocked"
	163	+ },
	164	+ {
	165	+ "PublicDescription": "Early mispredict",
	166	+ "EventCode": "0xD100",
	167	+ "EventName": "ICF_EARLY_MIS_PRED",
	168	+ "BriefDescription": "Early mispredict"
	169	+ },
	170	+ {
	171	+ "PublicDescription": "FEQ full cycles",
	172	+ "EventCode": "0xD101",
	173	+ "EventName": "ICF_FEQ_FULL",
	174	+ "BriefDescription": "FEQ full cycles"
	175	+ },
	176	+ {
	177	+ "PublicDescription": "Instruction FIFO Full",
	178	+ "EventCode": "0xD102",
	179	+ "EventName": "ICF_INST_FIFO_FULL",
	180	+ "BriefDescription": "Instruction FIFO Full"
	181	+ },
	182	+ {
	183	+ "PublicDescription": "L1I TLB miss",
	184	+ "EventCode": "0xD103",
	185	+ "EventName": "L1I_TLB_MISS",
	186	+ "BriefDescription": "L1I TLB miss"
	187	+ },
	188	+ {
	189	+ "PublicDescription": "ICF sent 0 instructions to IDR this cycle",
	190	+ "EventCode": "0xD104",
	191	+ "EventName": "ICF_STALL",
	192	+ "BriefDescription": "ICF sent 0 instructions to IDR this cycle"
	193	+ },
	194	+ {
	195	+ "PublicDescription": "PC FIFO Full",
	196	+ "EventCode": "0xD105",
	197	+ "EventName": "ICF_PC_FIFO_FULL",
	198	+ "BriefDescription": "PC FIFO Full"
	199	+ },
	200	+ {
	201	+ "PublicDescription": "Stall due to BOB ID",
	202	+ "EventCode": "0xD200",
	203	+ "EventName": "IDR_STALL_BOB_ID",
	204	+ "BriefDescription": "Stall due to BOB ID"
	205	+ },
	206	+ {
	207	+ "PublicDescription": "Dispatch stall due to LOB entries",
	208	+ "EventCode": "0xD201",
	209	+ "EventName": "IDR_STALL_LOB_ID",
	210	+ "BriefDescription": "Dispatch stall due to LOB entries"
	211	+ },
	212	+ {
	213	+ "PublicDescription": "Dispatch stall due to SOB entries",
	214	+ "EventCode": "0xD202",
	215	+ "EventName": "IDR_STALL_SOB_ID",
	216	+ "BriefDescription": "Dispatch stall due to SOB entries"
	217	+ },
	218	+ {
	219	+ "PublicDescription": "Dispatch stall due to IXU scheduler entries",
	220	+ "EventCode": "0xD203",
	221	+ "EventName": "IDR_STALL_IXU_SCHED",
	222	+ "BriefDescription": "Dispatch stall due to IXU scheduler entries"
	223	+ },
	224	+ {
	225	+ "PublicDescription": "Dispatch stall due to FSU scheduler entries",
	226	+ "EventCode": "0xD204",
	227	+ "EventName": "IDR_STALL_FSU_SCHED",
	228	+ "BriefDescription": "Dispatch stall due to FSU scheduler entries"
	229	+ },
	230	+ {
	231	+ "PublicDescription": "Dispatch stall due to ROB entries",
	232	+ "EventCode": "0xD205",
	233	+ "EventName": "IDR_STALL_ROB_ID",
	234	+ "BriefDescription": "Dispatch stall due to ROB entries"
	235	+ },
	236	+ {
	237	+ "PublicDescription": "Dispatch stall due to flush",
	238	+ "EventCode": "0xD206",
	239	+ "EventName": "IDR_STALL_FLUSH",
	240	+ "BriefDescription": "Dispatch stall due to flush"
	241	+ },
	242	+ {
	243	+ "PublicDescription": "Dispatch stall due to WFI",
	244	+ "EventCode": "0xD207",
	245	+ "EventName": "IDR_STALL_WFI",
	246	+ "BriefDescription": "Dispatch stall due to WFI"
	247	+ },
	248	+ {
	249	+ "PublicDescription": "Number of SWOB drains triggered by timeout",
	250	+ "EventCode": "0xD208",
	251	+ "EventName": "IDR_STALL_SWOB_TIMEOUT",
	252	+ "BriefDescription": "Number of SWOB drains triggered by timeout"
	253	+ },
	254	+ {
	255	+ "PublicDescription": "Number of SWOB drains triggered by system register or special-purpose register read-after-write or specific special-purpose register writes that cause SWOB drain",
	256	+ "EventCode": "0xD209",
	257	+ "EventName": "IDR_STALL_SWOB_RAW",
	258	+ "BriefDescription": "Number of SWOB drains triggered by system register or special-purpose register read-after-write or specific special-purpose register writes that cause SWOB drain"
	259	+ },
	260	+ {
	261	+ "PublicDescription": "Number of SWOB drains triggered by system register write when SWOB full",
	262	+ "EventCode": "0xD20A",
	263	+ "EventName": "IDR_STALL_SWOB_FULL",
	264	+ "BriefDescription": "Number of SWOB drains triggered by system register write when SWOB full"
	265	+ },
	266	+ {
	267	+ "PublicDescription": "Dispatch stall due to L1 instruction cache miss",
	268	+ "EventCode": "0xD20B",
	269	+ "EventName": "STALL_FRONTEND_CACHE",
	270	+ "BriefDescription": "Dispatch stall due to L1 instruction cache miss"
	271	+ },
	272	+ {
	273	+ "PublicDescription": "Dispatch stall due to L1 data cache miss",
	274	+ "EventCode": "0xD20D",
	275	+ "EventName": "STALL_BACKEND_CACHE",
	276	+ "BriefDescription": "Dispatch stall due to L1 data cache miss"
	277	+ },
	278	+ {
	279	+ "PublicDescription": "Dispatch stall due to lack of any core resource",
	280	+ "EventCode": "0xD20F",
	281	+ "EventName": "STALL_BACKEND_RESOURCE",
	282	+ "BriefDescription": "Dispatch stall due to lack of any core resource"
	283	+ },
	284	+ {
	285	+ "PublicDescription": "Instructions issued by the scheduler",
	286	+ "EventCode": "0xD300",
	287	+ "EventName": "IXU_NUM_UOPS_ISSUED",
	288	+ "BriefDescription": "Instructions issued by the scheduler"
	289	+ },
	290	+ {
	291	+ "PublicDescription": "Any uop issued was canceled for any reason",
	292	+ "EventCode": "0xD301",
	293	+ "EventName": "IXU_ISSUE_CANCEL",
	294	+ "BriefDescription": "Any uop issued was canceled for any reason"
	295	+ },
	296	+ {
	297	+ "PublicDescription": "A load wakeup to the scheduler has been canceled",
	298	+ "EventCode": "0xD302",
	299	+ "EventName": "IXU_LOAD_CANCEL",
	300	+ "BriefDescription": "A load wakeup to the scheduler has been canceled"
	301	+ },
	302	+ {
	303	+ "PublicDescription": "The scheduler had to cancel one slow Uop due to resource conflict",
	304	+ "EventCode": "0xD303",
	305	+ "EventName": "IXU_SLOW_CANCEL",
	306	+ "BriefDescription": "The scheduler had to cancel one slow Uop due to resource conflict"
	307	+ },
	308	+ {
	309	+ "PublicDescription": "Uops issued by the scheduler on IXA",
	310	+ "EventCode": "0xD304",
	311	+ "EventName": "IXU_IXA_ISSUED",
	312	+ "BriefDescription": "Uops issued by the scheduler on IXA"
	313	+ },
	314	+ {
	315	+ "PublicDescription": "Uops issued by the scheduler on IXA Par 0",
	316	+ "EventCode": "0xD305",
	317	+ "EventName": "IXU_IXA_PAR0_ISSUED",
	318	+ "BriefDescription": "Uops issued by the scheduler on IXA Par 0"
	319	+ },
	320	+ {
	321	+ "PublicDescription": "Uops issued by the scheduler on IXA Par 1",
	322	+ "EventCode": "0xD306",
	323	+ "EventName": "IXU_IXA_PAR1_ISSUED",
	324	+ "BriefDescription": "Uops issued by the scheduler on IXA Par 1"
	325	+ },
	326	+ {
	327	+ "PublicDescription": "Uops issued by the scheduler on IXB",
	328	+ "EventCode": "0xD307",
	329	+ "EventName": "IXU_IXB_ISSUED",
	330	+ "BriefDescription": "Uops issued by the scheduler on IXB"
	331	+ },
	332	+ {
	333	+ "PublicDescription": "Uops issued by the scheduler on IXB Par 0",
	334	+ "EventCode": "0xD308",
	335	+ "EventName": "IXU_IXB_PAR0_ISSUED",
	336	+ "BriefDescription": "Uops issued by the scheduler on IXB Par 0"
	337	+ },
	338	+ {
	339	+ "PublicDescription": "Uops issued by the scheduler on IXB Par 1",
	340	+ "EventCode": "0xD309",
	341	+ "EventName": "IXU_IXB_PAR1_ISSUED",
	342	+ "BriefDescription": "Uops issued by the scheduler on IXB Par 1"
	343	+ },
	344	+ {
	345	+ "PublicDescription": "Uops issued by the scheduler on IXC",
	346	+ "EventCode": "0xD30A",
	347	+ "EventName": "IXU_IXC_ISSUED",
	348	+ "BriefDescription": "Uops issued by the scheduler on IXC"
	349	+ },
	350	+ {
	351	+ "PublicDescription": "Uops issued by the scheduler on IXC Par 0",
	352	+ "EventCode": "0xD30B",
	353	+ "EventName": "IXU_IXC_PAR0_ISSUED",
	354	+ "BriefDescription": "Uops issued by the scheduler on IXC Par 0"
	355	+ },
	356	+ {
	357	+ "PublicDescription": "Uops issued by the scheduler on IXC Par 1",
	358	+ "EventCode": "0xD30C",
	359	+ "EventName": "IXU_IXC_PAR1_ISSUED",
	360	+ "BriefDescription": "Uops issued by the scheduler on IXC Par 1"
	361	+ },
	362	+ {
	363	+ "PublicDescription": "Uops issued by the scheduler on IXD",
	364	+ "EventCode": "0xD30D",
	365	+ "EventName": "IXU_IXD_ISSUED",
	366	+ "BriefDescription": "Uops issued by the scheduler on IXD"
	367	+ },
	368	+ {
	369	+ "PublicDescription": "Uops issued by the scheduler on IXD Par 0",
	370	+ "EventCode": "0xD30E",
	371	+ "EventName": "IXU_IXD_PAR0_ISSUED",
	372	+ "BriefDescription": "Uops issued by the scheduler on IXD Par 0"
	373	+ },
	374	+ {
	375	+ "PublicDescription": "Uops issued by the scheduler on IXD Par 1",
	376	+ "EventCode": "0xD30F",
	377	+ "EventName": "IXU_IXD_PAR1_ISSUED",
	378	+ "BriefDescription": "Uops issued by the scheduler on IXD Par 1"
	379	+ },
	380	+ {
	381	+ "PublicDescription": "Uops issued by the FSU scheduler",
	382	+ "EventCode": "0xD400",
	383	+ "EventName": "FSU_ISSUED",
	384	+ "BriefDescription": "Uops issued by the FSU scheduler"
	385	+ },
	386	+ {
	387	+ "PublicDescription": "Uops issued by the scheduler on FSX",
	388	+ "EventCode": "0xD401",
	389	+ "EventName": "FSU_FSX_ISSUED",
	390	+ "BriefDescription": "Uops issued by the scheduler on FSX"
	391	+ },
	392	+ {
	393	+ "PublicDescription": "Uops issued by the scheduler on FSY",
	394	+ "EventCode": "0xD402",
	395	+ "EventName": "FSU_FSY_ISSUED",
	396	+ "BriefDescription": "Uops issued by the scheduler on FSY"
	397	+ },
	398	+ {
	399	+ "PublicDescription": "Uops issued by the scheduler on FSZ",
	400	+ "EventCode": "0xD403",
	401	+ "EventName": "FSU_FSZ_ISSUED",
	402	+ "BriefDescription": "Uops issued by the scheduler on FSZ"
	403	+ },
	404	+ {
	405	+ "PublicDescription": "Uops canceled (load cancels)",
	406	+ "EventCode": "0xD404",
	407	+ "EventName": "FSU_CANCEL",
	408	+ "BriefDescription": "Uops canceled (load cancels)"
	409	+ },
	410	+ {
	411	+ "PublicDescription": "Count scheduler stalls due to divide/sqrt",
	412	+ "EventCode": "0xD405",
	413	+ "EventName": "FSU_DIV_SQRT_STALL",
	414	+ "BriefDescription": "Count scheduler stalls due to divide/sqrt"
	415	+ },
	416	+ {
	417	+ "PublicDescription": "Number of SWOB drains",
	418	+ "EventCode": "0xD500",
	419	+ "EventName": "GPC_SWOB_DRAIN",
	420	+ "BriefDescription": "Number of SWOB drains"
	421	+ },
	422	+ {
	423	+ "PublicDescription": "GPC detected a Breakpoint instruction match",
	424	+ "EventCode": "0xD501",
	425	+ "EventName": "BREAKPOINT_MATCH",
	426	+ "BriefDescription": "GPC detected a Breakpoint instruction match"
	427	+ },
	428	+ {
	429	+ "PublicDescription": "Core progress monitor triggered",
	430	+ "EventCode": "0xd502",
	431	+ "EventName": "GPC_CPM_TRIGGER",
	432	+ "BriefDescription": "Core progress monitor triggered"
	433	+ },
	434	+ {
	435	+ "PublicDescription": "Fill buffer full",
	436	+ "EventCode": "0xD601",
	437	+ "EventName": "OFB_FULL",
	438	+ "BriefDescription": "Fill buffer full"
	439	+ },
	440	+ {
	441	+ "PublicDescription": "Load satisified from store forwarded data",
	442	+ "EventCode": "0xD605",
	443	+ "EventName": "LD_FROM_ST_FWD",
	444	+ "BriefDescription": "Load satisified from store forwarded data"
	445	+ },
	446	+ {
	447	+ "PublicDescription": "Store retirement pipe stall",
	448	+ "EventCode": "0xD60C",
	449	+ "EventName": "LSU_ST_RETIRE_STALL",
	450	+ "BriefDescription": "Store retirement pipe stall"
	451	+ },
	452	+ {
	453	+ "PublicDescription": "LSU detected a Watchpoint data match",
	454	+ "EventCode": "0xD60D",
	455	+ "EventName": "WATCHPOINT_MATCH",
	456	+ "BriefDescription": "LSU detected a Watchpoint data match"
	457	+ },
	458	+ {
	459	+ "PublicDescription": "Counts cycles that MSC is telling GPC to stall commit due to ETM ISTALL feature",
	460	+ "EventCode": "0xda00",
	461	+ "EventName": "MSC_ETM_COMMIT_STALL",
	462	+ "BriefDescription": "Counts cycles that MSC is telling GPC to stall commit due to ETM ISTALL feature"
	463	+ }
	464	+]

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/exception.json+47−0

@@ -0,0 +1,47 @@
	1	+[
	2	+ {
	3	+ "ArchStdEvent": "EXC_UNDEF"
	4	+ },
	5	+ {
	6	+ "ArchStdEvent": "EXC_SVC"
	7	+ },
	8	+ {
	9	+ "ArchStdEvent": "EXC_PABORT"
	10	+ },
	11	+ {
	12	+ "ArchStdEvent": "EXC_DABORT"
	13	+ },
	14	+ {
	15	+ "ArchStdEvent": "EXC_IRQ"
	16	+ },
	17	+ {
	18	+ "ArchStdEvent": "EXC_FIQ"
	19	+ },
	20	+ {
	21	+ "ArchStdEvent": "EXC_HVC"
	22	+ },
	23	+ {
	24	+ "ArchStdEvent": "EXC_TRAP_PABORT"
	25	+ },
	26	+ {
	27	+ "ArchStdEvent": "EXC_TRAP_DABORT"
	28	+ },
	29	+ {
	30	+ "ArchStdEvent": "EXC_TRAP_OTHER"
	31	+ },
	32	+ {
	33	+ "ArchStdEvent": "EXC_TRAP_IRQ"
	34	+ },
	35	+ {
	36	+ "ArchStdEvent": "EXC_TRAP_FIQ"
	37	+ },
	38	+ {
	39	+ "ArchStdEvent": "EXC_TAKEN"
	40	+ },
	41	+ {
	42	+ "ArchStdEvent": "EXC_RETURN"
	43	+ },
	44	+ {
	45	+ "ArchStdEvent": "EXC_SMC"
	46	+ }
	47	+]

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/instruction.json+128−0

@@ -0,0 +1,128 @@
	1	+[
	2	+ {
	3	+ "ArchStdEvent": "SW_INCR"
	4	+ },
	5	+ {
	6	+ "ArchStdEvent": "ST_RETIRED"
	7	+ },
	8	+ {
	9	+ "ArchStdEvent": "LD_SPEC"
	10	+ },
	11	+ {
	12	+ "ArchStdEvent": "ST_SPEC"
	13	+ },
	14	+ {
	15	+ "ArchStdEvent": "LDST_SPEC"
	16	+ },
	17	+ {
	18	+ "ArchStdEvent": "DP_SPEC"
	19	+ },
	20	+ {
	21	+ "ArchStdEvent": "ASE_SPEC"
	22	+ },
	23	+ {
	24	+ "ArchStdEvent": "VFP_SPEC"
	25	+ },
	26	+ {
	27	+ "ArchStdEvent": "PC_WRITE_SPEC"
	28	+ },
	29	+ {
	30	+ "ArchStdEvent": "BR_IMMED_RETIRED"
	31	+ },
	32	+ {
	33	+ "ArchStdEvent": "BR_RETURN_RETIRED"
	34	+ },
	35	+ {
	36	+ "ArchStdEvent": "CRYPTO_SPEC"
	37	+ },
	38	+ {
	39	+ "ArchStdEvent": "ISB_SPEC"
	40	+ },
	41	+ {
	42	+ "ArchStdEvent": "DSB_SPEC"
	43	+ },
	44	+ {
	45	+ "ArchStdEvent": "DMB_SPEC"
	46	+ },
	47	+ {
	48	+ "ArchStdEvent": "RC_LD_SPEC"
	49	+ },
	50	+ {
	51	+ "ArchStdEvent": "RC_ST_SPEC"
	52	+ },
	53	+ {
	54	+ "ArchStdEvent": "INST_RETIRED"
	55	+ },
	56	+ {
	57	+ "ArchStdEvent": "CID_WRITE_RETIRED"
	58	+ },
	59	+ {
	60	+ "ArchStdEvent": "PC_WRITE_RETIRED"
	61	+ },
	62	+ {
	63	+ "ArchStdEvent": "INST_SPEC"
	64	+ },
	65	+ {
	66	+ "ArchStdEvent": "TTBR_WRITE_RETIRED"
	67	+ },
	68	+ {
	69	+ "ArchStdEvent": "BR_RETIRED"
	70	+ },
	71	+ {
	72	+ "ArchStdEvent": "BR_MIS_PRED_RETIRED"
	73	+ },
	74	+ {
	75	+ "ArchStdEvent": "OP_RETIRED"
	76	+ },
	77	+ {
	78	+ "ArchStdEvent": "OP_SPEC"
	79	+ },
	80	+ {
	81	+ "PublicDescription": "Operation speculatively executed - ASE Scalar",
	82	+ "EventCode": "0xd210",
	83	+ "EventName": "ASE_SCALAR_SPEC",
	84	+ "BriefDescription": "Operation speculatively executed - ASE Scalar"
	85	+ },
	86	+ {
	87	+ "PublicDescription": "Operation speculatively executed - ASE Vector",
	88	+ "EventCode": "0xd211",
	89	+ "EventName": "ASE_VECTOR_SPEC",
	90	+ "BriefDescription": "Operation speculatively executed - ASE Vector"
	91	+ },
	92	+ {
	93	+ "PublicDescription": "Barrier speculatively executed, CSDB",
	94	+ "EventCode": "0x7f",
	95	+ "EventName": "CSDB_SPEC",
	96	+ "BriefDescription": "Barrier speculatively executed, CSDB"
	97	+ },
	98	+ {
	99	+ "PublicDescription": "Prefetch sent to L2.",
	100	+ "EventCode": "0xd106",
	101	+ "EventName": "ICF_PREFETCH_DISPATCH",
	102	+ "BriefDescription": "Prefetch sent to L2."
	103	+ },
	104	+ {
	105	+ "PublicDescription": "Prefetch response received but was dropped since we don't support inflight upgrades.",
	106	+ "EventCode": "0xd107",
	107	+ "EventName": "ICF_PREFETCH_DROPPED_NO_UPGRADE",
	108	+ "BriefDescription": "Prefetch response received but was dropped since we don't support inflight upgrades."
	109	+ },
	110	+ {
	111	+ "PublicDescription": "Prefetch request missed TLB.",
	112	+ "EventCode": "0xd108",
	113	+ "EventName": "ICF_PREFETCH_DROPPED_TLB_MISS",
	114	+ "BriefDescription": "Prefetch request missed TLB."
	115	+ },
	116	+ {
	117	+ "PublicDescription": "Prefetch request dropped since duplicate was found in TLB.",
	118	+ "EventCode": "0xd109",
	119	+ "EventName": "ICF_PREFETCH_DROPPED_DUPLICATE",
	120	+ "BriefDescription": "Prefetch request dropped since duplicate was found in TLB."
	121	+ },
	122	+ {
	123	+ "PublicDescription": "Prefetch request dropped since it was found in cache.",
	124	+ "EventCode": "0xd10a",
	125	+ "EventName": "ICF_PREFETCH_DROPPED_CACHE_HIT",
	126	+ "BriefDescription": "Prefetch request dropped since it was found in cache."
	127	+ }
	128	+]

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/intrinsic.json+14−0

@@ -0,0 +1,14 @@
	1	+[
	2	+ {
	3	+ "ArchStdEvent": "LDREX_SPEC"
	4	+ },
	5	+ {
	6	+ "ArchStdEvent": "STREX_PASS_SPEC"
	7	+ },
	8	+ {
	9	+ "ArchStdEvent": "STREX_FAIL_SPEC"
	10	+ },
	11	+ {
	12	+ "ArchStdEvent": "STREX_SPEC"
	13	+ }
	14	+]

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/memory.json+41−0

@@ -0,0 +1,41 @@
	1	+[
	2	+ {
	3	+ "ArchStdEvent": "LD_RETIRED"
	4	+ },
	5	+ {
	6	+ "ArchStdEvent": "MEM_ACCESS_RD"
	7	+ },
	8	+ {
	9	+ "ArchStdEvent": "MEM_ACCESS_WR"
	10	+ },
	11	+ {
	12	+ "ArchStdEvent": "LD_ALIGN_LAT"
	13	+ },
	14	+ {
	15	+ "ArchStdEvent": "ST_ALIGN_LAT"
	16	+ },
	17	+ {
	18	+ "ArchStdEvent": "MEM_ACCESS"
	19	+ },
	20	+ {
	21	+ "ArchStdEvent": "MEMORY_ERROR"
	22	+ },
	23	+ {
	24	+ "ArchStdEvent": "LDST_ALIGN_LAT"
	25	+ },
	26	+ {
	27	+ "ArchStdEvent": "MEM_ACCESS_CHECKED"
	28	+ },
	29	+ {
	30	+ "ArchStdEvent": "MEM_ACCESS_CHECKED_RD"
	31	+ },
	32	+ {
	33	+ "ArchStdEvent": "MEM_ACCESS_CHECKED_WR"
	34	+ },
	35	+ {
	36	+ "PublicDescription": "Flushes due to memory hazards",
	37	+ "EventCode": "0x121",
	38	+ "EventName": "BPU_FLUSH_MEM_FAULT",
	39	+ "BriefDescription": "Flushes due to memory hazards"
	40	+ }
	41	+]

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/metrics.json+442−0

@@ -0,0 +1,442 @@
	1	+[
	2	+ {
	3	+ "MetricName": "branch_miss_pred_rate",
	4	+ "MetricExpr": "BR_MIS_PRED / BR_PRED",
	5	+ "BriefDescription": "Branch predictor misprediction rate. May not count branches that are never resolved because they are in the misprediction shadow of an earlier branch",
	6	+ "MetricGroup": "branch",
	7	+ "ScaleUnit": "100%"
	8	+ },
	9	+ {
	10	+ "MetricName": "bus_utilization",
	11	+ "MetricExpr": "BUS_ACCESS / (BUS_CYCLES * 1)",
	12	+ "BriefDescription": "Core-to-uncore bus utilization",
	13	+ "MetricGroup": "Bus",
	14	+ "ScaleUnit": "100percent of bus cycles"
	15	+ },
	16	+ {
	17	+ "MetricName": "l1d_cache_miss_ratio",
	18	+ "MetricExpr": "L1D_CACHE_REFILL / L1D_CACHE",
	19	+ "BriefDescription": "This metric measures the ratio of level 1 data cache accesses missed to the total number of level 1 data cache accesses. This gives an indication of the effectiveness of the level 1 data cache.",
	20	+ "MetricGroup": "Miss_Ratio;L1D_Cache_Effectiveness",
	21	+ "ScaleUnit": "1per cache access"
	22	+ },
	23	+ {
	24	+ "MetricName": "l1i_cache_miss_ratio",
	25	+ "MetricExpr": "L1I_CACHE_REFILL / L1I_CACHE",
	26	+ "BriefDescription": "This metric measures the ratio of level 1 instruction cache accesses missed to the total number of level 1 instruction cache accesses. This gives an indication of the effectiveness of the level 1 instruction cache.",
	27	+ "MetricGroup": "Miss_Ratio;L1I_Cache_Effectiveness",
	28	+ "ScaleUnit": "1per cache access"
	29	+ },
	30	+ {
	31	+ "MetricName": "Miss_Ratio;l1d_cache_read_miss",
	32	+ "MetricExpr": "L1D_CACHE_LMISS_RD / L1D_CACHE_RD",
	33	+ "BriefDescription": "L1D cache read miss rate",
	34	+ "MetricGroup": "Cache",
	35	+ "ScaleUnit": "1per cache read access"
	36	+ },
	37	+ {
	38	+ "MetricName": "l2_cache_miss_ratio",
	39	+ "MetricExpr": "L2D_CACHE_REFILL / L2D_CACHE",
	40	+ "BriefDescription": "This metric measures the ratio of level 2 cache accesses missed to the total number of level 2 cache accesses. This gives an indication of the effectiveness of the level 2 cache, which is a unified cache that stores both data and instruction. Note that cache accesses in this cache are either data memory access or instruction fetch as this is a unified cache.",
	41	+ "MetricGroup": "Miss_Ratio;L2_Cache_Effectiveness",
	42	+ "ScaleUnit": "1per cache access"
	43	+ },
	44	+ {
	45	+ "MetricName": "l1i_cache_read_miss_rate",
	46	+ "MetricExpr": "L1I_CACHE_LMISS / L1I_CACHE",
	47	+ "BriefDescription": "L1I cache read miss rate",
	48	+ "MetricGroup": "Cache",
	49	+ "ScaleUnit": "1per cache access"
	50	+ },
	51	+ {
	52	+ "MetricName": "l2d_cache_read_miss_rate",
	53	+ "MetricExpr": "L2D_CACHE_LMISS_RD / L2D_CACHE_RD",
	54	+ "BriefDescription": "L2 cache read miss rate",
	55	+ "MetricGroup": "Cache",
	56	+ "ScaleUnit": "1per cache read access"
	57	+ },
	58	+ {
	59	+ "MetricName": "l1d_cache_miss_mpki",
	60	+ "MetricExpr": "(L1D_CACHE_LMISS_RD * 1e3) / INST_RETIRED",
	61	+ "BriefDescription": "Misses per thousand instructions (data)",
	62	+ "MetricGroup": "Cache",
	63	+ "ScaleUnit": "1MPKI"
	64	+ },
	65	+ {
	66	+ "MetricName": "l1i_cache_miss_mpki",
	67	+ "MetricExpr": "(L1I_CACHE_LMISS * 1e3) / INST_RETIRED",
	68	+ "BriefDescription": "Misses per thousand instructions (instruction)",
	69	+ "MetricGroup": "Cache",
	70	+ "ScaleUnit": "1MPKI"
	71	+ },
	72	+ {
	73	+ "MetricName": "simd_percentage",
	74	+ "MetricExpr": "ASE_SPEC / INST_SPEC",
	75	+ "BriefDescription": "This metric measures advanced SIMD operations as a percentage of total operations speculatively executed.",
	76	+ "MetricGroup": "Operation_Mix",
	77	+ "ScaleUnit": "100percent of operations"
	78	+ },
	79	+ {
	80	+ "MetricName": "crypto_percentage",
	81	+ "MetricExpr": "CRYPTO_SPEC / INST_SPEC",
	82	+ "BriefDescription": "This metric measures crypto operations as a percentage of operations speculatively executed.",
	83	+ "MetricGroup": "Operation_Mix",
	84	+ "ScaleUnit": "100percent of operations"
	85	+ },
	86	+ {
	87	+ "MetricName": "gflops",
	88	+ "MetricExpr": "VFP_SPEC / (duration_time * 1e9)",
	89	+ "BriefDescription": "Giga-floating point operations per second",
	90	+ "MetricGroup": "InstructionMix"
	91	+ },
	92	+ {
	93	+ "MetricName": "integer_dp_percentage",
	94	+ "MetricExpr": "DP_SPEC / INST_SPEC",
	95	+ "BriefDescription": "This metric measures scalar integer operations as a percentage of operations speculatively executed.",
	96	+ "MetricGroup": "Operation_Mix",
	97	+ "ScaleUnit": "100percent of operations"
	98	+ },
	99	+ {
	100	+ "MetricName": "ipc",
	101	+ "MetricExpr": "INST_RETIRED / CPU_CYCLES",
	102	+ "BriefDescription": "This metric measures the number of instructions retired per cycle.",
	103	+ "MetricGroup": "General",
	104	+ "ScaleUnit": "1per cycle"
	105	+ },
	106	+ {
	107	+ "MetricName": "load_percentage",
	108	+ "MetricExpr": "LD_SPEC / INST_SPEC",
	109	+ "BriefDescription": "This metric measures load operations as a percentage of operations speculatively executed.",
	110	+ "MetricGroup": "Operation_Mix",
	111	+ "ScaleUnit": "100percent of operations"
	112	+ },
	113	+ {
	114	+ "MetricName": "load_store_spec_rate",
	115	+ "MetricExpr": "LDST_SPEC / INST_SPEC",
	116	+ "BriefDescription": "The rate of load or store instructions speculatively executed to overall instructions speclatively executed",
	117	+ "MetricGroup": "Operation_Mix",
	118	+ "ScaleUnit": "100percent of operations"
	119	+ },
	120	+ {
	121	+ "MetricName": "retired_mips",
	122	+ "MetricExpr": "INST_RETIRED / (duration_time * 1e6)",
	123	+ "BriefDescription": "Millions of instructions per second",
	124	+ "MetricGroup": "InstructionMix"
	125	+ },
	126	+ {
	127	+ "MetricName": "spec_utilization_mips",
	128	+ "MetricExpr": "INST_SPEC / (duration_time * 1e6)",
	129	+ "BriefDescription": "Millions of instructions per second",
	130	+ "MetricGroup": "PEutilization"
	131	+ },
	132	+ {
	133	+ "MetricName": "pc_write_spec_rate",
	134	+ "MetricExpr": "PC_WRITE_SPEC / INST_SPEC",
	135	+ "BriefDescription": "The rate of software change of the PC speculatively executed to overall instructions speclatively executed",
	136	+ "MetricGroup": "Operation_Mix",
	137	+ "ScaleUnit": "100percent of operations"
	138	+ },
	139	+ {
	140	+ "MetricName": "store_percentage",
	141	+ "MetricExpr": "ST_SPEC / INST_SPEC",
	142	+ "BriefDescription": "This metric measures store operations as a percentage of operations speculatively executed.",
	143	+ "MetricGroup": "Operation_Mix",
	144	+ "ScaleUnit": "100percent of operations"
	145	+ },
	146	+ {
	147	+ "MetricName": "scalar_fp_percentage",
	148	+ "MetricExpr": "VFP_SPEC / INST_SPEC",
	149	+ "BriefDescription": "This metric measures scalar floating point operations as a percentage of operations speculatively executed.",
	150	+ "MetricGroup": "Operation_Mix",
	151	+ "ScaleUnit": "100percent of operations"
	152	+ },
	153	+ {
	154	+ "MetricName": "retired_rate",
	155	+ "MetricExpr": "OP_RETIRED / OP_SPEC",
	156	+ "BriefDescription": "Of all the micro-operations issued, what percentage are retired(committed)",
	157	+ "MetricGroup": "General",
	158	+ "ScaleUnit": "100%"
	159	+ },
	160	+ {
	161	+ "MetricName": "wasted",
	162	+ "MetricExpr": "1 - (OP_RETIRED / (CPU_CYCLES * #slots))",
	163	+ "BriefDescription": "Of all the micro-operations issued, what proportion are lost",
	164	+ "MetricGroup": "General",
	165	+ "ScaleUnit": "100%"
	166	+ },
	167	+ {
	168	+ "MetricName": "wasted_rate",
	169	+ "MetricExpr": "1 - OP_RETIRED / OP_SPEC",
	170	+ "BriefDescription": "Of all the micro-operations issued, what percentage are not retired(committed)",
	171	+ "MetricGroup": "General",
	172	+ "ScaleUnit": "100%"
	173	+ },
	174	+ {
	175	+ "MetricName": "stall_backend_cache_rate",
	176	+ "MetricExpr": "STALL_BACKEND_CACHE / CPU_CYCLES",
	177	+ "BriefDescription": "Proportion of cycles stalled and no operations issued to backend and cache miss",
	178	+ "MetricGroup": "Stall",
	179	+ "ScaleUnit": "100percent of cycles"
	180	+ },
	181	+ {
	182	+ "MetricName": "stall_backend_resource_rate",
	183	+ "MetricExpr": "STALL_BACKEND_RESOURCE / CPU_CYCLES",
	184	+ "BriefDescription": "Proportion of cycles stalled and no operations issued to backend and resource full",
	185	+ "MetricGroup": "Stall",
	186	+ "ScaleUnit": "100percent of cycles"
	187	+ },
	188	+ {
	189	+ "MetricName": "stall_backend_tlb_rate",
	190	+ "MetricExpr": "STALL_BACKEND_TLB / CPU_CYCLES",
	191	+ "BriefDescription": "Proportion of cycles stalled and no operations issued to backend and TLB miss",
	192	+ "MetricGroup": "Stall",
	193	+ "ScaleUnit": "100percent of cycles"
	194	+ },
	195	+ {
	196	+ "MetricName": "stall_frontend_cache_rate",
	197	+ "MetricExpr": "STALL_FRONTEND_CACHE / CPU_CYCLES",
	198	+ "BriefDescription": "Proportion of cycles stalled and no ops delivered from frontend and cache miss",
	199	+ "MetricGroup": "Stall",
	200	+ "ScaleUnit": "100percent of cycles"
	201	+ },
	202	+ {
	203	+ "MetricName": "stall_frontend_tlb_rate",
	204	+ "MetricExpr": "STALL_FRONTEND_TLB / CPU_CYCLES",
	205	+ "BriefDescription": "Proportion of cycles stalled and no ops delivered from frontend and TLB miss",
	206	+ "MetricGroup": "Stall",
	207	+ "ScaleUnit": "100percent of cycles"
	208	+ },
	209	+ {
	210	+ "MetricName": "dtlb_walk_ratio",
	211	+ "MetricExpr": "DTLB_WALK / L1D_TLB",
	212	+ "BriefDescription": "This metric measures the ratio of data TLB Walks to the total number of data TLB accesses. This gives an indication of the effectiveness of the data TLB accesses.",
	213	+ "MetricGroup": "Miss_Ratio;DTLB_Effectiveness",
	214	+ "ScaleUnit": "1per TLB access"
	215	+ },
	216	+ {
	217	+ "MetricName": "itlb_walk_ratio",
	218	+ "MetricExpr": "ITLB_WALK / L1I_TLB",
	219	+ "BriefDescription": "This metric measures the ratio of instruction TLB Walks to the total number of instruction TLB accesses. This gives an indication of the effectiveness of the instruction TLB accesses.",
	220	+ "MetricGroup": "Miss_Ratio;ITLB_Effectiveness",
	221	+ "ScaleUnit": "1per TLB access"
	222	+ },
	223	+ {
	224	+ "ArchStdEvent": "backend_bound"
	225	+ },
	226	+ {
	227	+ "ArchStdEvent": "frontend_bound",
	228	+ "MetricExpr": "100 - (retired_fraction + slots_lost_misspeculation_fraction + backend_bound)"
	229	+ },
	230	+ {
	231	+ "MetricName": "slots_lost_misspeculation_fraction",
	232	+ "MetricExpr": "(OP_SPEC - OP_RETIRED) / (CPU_CYCLES * #slots)",
	233	+ "BriefDescription": "Fraction of slots lost due to misspeculation",
	234	+ "DefaultMetricgroupName": "TopdownL1",
	235	+ "MetricGroup": "Default;TopdownL1",
	236	+ "ScaleUnit": "100percent of slots"
	237	+ },
	238	+ {
	239	+ "MetricName": "retired_fraction",
	240	+ "MetricExpr": "OP_RETIRED / (CPU_CYCLES * #slots)",
	241	+ "BriefDescription": "Fraction of slots retiring, useful work",
	242	+ "DefaultMetricgroupName": "TopdownL1",
	243	+ "MetricGroup": "Default;TopdownL1",
	244	+ "ScaleUnit": "100percent of slots"
	245	+ },
	246	+ {
	247	+ "MetricName": "backend_core",
	248	+ "MetricExpr": "(backend_bound / 100) - backend_memory",
	249	+ "BriefDescription": "Fraction of slots the CPU was stalled due to backend non-memory subsystem issues",
	250	+ "MetricGroup": "TopdownL2",
	251	+ "ScaleUnit": "100%"
	252	+ },
	253	+ {
	254	+ "MetricName": "backend_memory",
	255	+ "MetricExpr": "(STALL_BACKEND_TLB + STALL_BACKEND_CACHE) / CPU_CYCLES",
	256	+ "BriefDescription": "Fraction of slots the CPU was stalled due to backend memory subsystem issues (cache/tlb miss)",
	257	+ "MetricGroup": "TopdownL2",
	258	+ "ScaleUnit": "100%"
	259	+ },
	260	+ {
	261	+ "MetricName": "branch_mispredict",
	262	+ "MetricExpr": "(BR_MIS_PRED_RETIRED / GPC_FLUSH) * slots_lost_misspeculation_fraction",
	263	+ "BriefDescription": "Fraction of slots lost due to branch misprediciton",
	264	+ "MetricGroup": "TopdownL2",
	265	+ "ScaleUnit": "1percent of slots"
	266	+ },
	267	+ {
	268	+ "MetricName": "frontend_bandwidth",
	269	+ "MetricExpr": "frontend_bound - frontend_latency",
	270	+ "BriefDescription": "Fraction of slots the CPU did not dispatch at full bandwidth - able to dispatch partial slots only (1, 2, or 3 uops)",
	271	+ "MetricGroup": "TopdownL2",
	272	+ "ScaleUnit": "1percent of slots"
	273	+ },
	274	+ {
	275	+ "MetricName": "frontend_latency",
	276	+ "MetricExpr": "(STALL_FRONTEND - ((STALL_SLOT_FRONTEND - ((frontend_bound / 100) * CPU_CYCLES * #slots)) / #slots)) / CPU_CYCLES",
	277	+ "BriefDescription": "Fraction of slots the CPU was stalled due to frontend latency issues (cache/tlb miss); nothing to dispatch",
	278	+ "MetricGroup": "TopdownL2",
	279	+ "ScaleUnit": "100percent of slots"
	280	+ },
	281	+ {
	282	+ "MetricName": "other_miss_pred",
	283	+ "MetricExpr": "slots_lost_misspeculation_fraction - branch_mispredict",
	284	+ "BriefDescription": "Fraction of slots lost due to other/non-branch misprediction misspeculation",
	285	+ "MetricGroup": "TopdownL2",
	286	+ "ScaleUnit": "1percent of slots"
	287	+ },
	288	+ {
	289	+ "MetricName": "pipe_utilization",
	290	+ "MetricExpr": "100 * ((IXU_NUM_UOPS_ISSUED + FSU_ISSUED) / (CPU_CYCLES * 6))",
	291	+ "BriefDescription": "Fraction of execute slots utilized",
	292	+ "MetricGroup": "TopdownL2",
	293	+ "ScaleUnit": "1percent of slots"
	294	+ },
	295	+ {
	296	+ "MetricName": "d_cache_l2_miss_rate",
	297	+ "MetricExpr": "STALL_BACKEND_MEM / CPU_CYCLES",
	298	+ "BriefDescription": "Fraction of cycles the CPU was stalled due to data L2 cache miss",
	299	+ "MetricGroup": "TopdownL3",
	300	+ "ScaleUnit": "100percent of cycles"
	301	+ },
	302	+ {
	303	+ "MetricName": "d_cache_miss_rate",
	304	+ "MetricExpr": "STALL_BACKEND_CACHE / CPU_CYCLES",
	305	+ "BriefDescription": "Fraction of cycles the CPU was stalled due to data cache miss",
	306	+ "MetricGroup": "TopdownL3",
	307	+ "ScaleUnit": "100percent of cycles"
	308	+ },
	309	+ {
	310	+ "MetricName": "d_tlb_miss_rate",
	311	+ "MetricExpr": "STALL_BACKEND_TLB / CPU_CYCLES",
	312	+ "BriefDescription": "Fraction of cycles the CPU was stalled due to data TLB miss",
	313	+ "MetricGroup": "TopdownL3",
	314	+ "ScaleUnit": "100percent of cycles"
	315	+ },
	316	+ {
	317	+ "MetricName": "fsu_pipe_utilization",
	318	+ "MetricExpr": "FSU_ISSUED / (CPU_CYCLES * 2)",
	319	+ "BriefDescription": "Fraction of FSU execute slots utilized",
	320	+ "MetricGroup": "TopdownL3",
	321	+ "ScaleUnit": "100percent of slots"
	322	+ },
	323	+ {
	324	+ "MetricName": "i_cache_miss_rate",
	325	+ "MetricExpr": "STALL_FRONTEND_CACHE / CPU_CYCLES",
	326	+ "BriefDescription": "Fraction of cycles the CPU was stalled due to instruction cache miss",
	327	+ "MetricGroup": "TopdownL3",
	328	+ "ScaleUnit": "100percent of slots"
	329	+ },
	330	+ {
	331	+ "MetricName": "i_tlb_miss_rate",
	332	+ "MetricExpr": "STALL_FRONTEND_TLB / CPU_CYCLES",
	333	+ "BriefDescription": "Fraction of cycles the CPU was stalled due to instruction TLB miss",
	334	+ "MetricGroup": "TopdownL3",
	335	+ "ScaleUnit": "100percent of slots"
	336	+ },
	337	+ {
	338	+ "MetricName": "ixu_pipe_utilization",
	339	+ "MetricExpr": "IXU_NUM_UOPS_ISSUED / (CPU_CYCLES * #slots)",
	340	+ "BriefDescription": "Fraction of IXU execute slots utilized",
	341	+ "MetricGroup": "TopdownL3",
	342	+ "ScaleUnit": "100percent of slots"
	343	+ },
	344	+ {
	345	+ "MetricName": "stall_recovery_rate",
	346	+ "MetricExpr": "IDR_STALL_FLUSH / CPU_CYCLES",
	347	+ "BriefDescription": "Fraction of cycles the CPU was stalled due to flush recovery",
	348	+ "MetricGroup": "TopdownL3",
	349	+ "ScaleUnit": "100percent of slots"
	350	+ },
	351	+ {
	352	+ "MetricName": "stall_fsu_sched_rate",
	353	+ "MetricExpr": "IDR_STALL_FSU_SCHED / CPU_CYCLES",
	354	+ "BriefDescription": "Fraction of cycles the CPU was stalled and FSU was full",
	355	+ "MetricGroup": "TopdownL4",
	356	+ "ScaleUnit": "100percent of cycles"
	357	+ },
	358	+ {
	359	+ "MetricName": "stall_ixu_sched_rate",
	360	+ "MetricExpr": "IDR_STALL_IXU_SCHED / CPU_CYCLES",
	361	+ "BriefDescription": "Fraction of cycles the CPU was stalled and IXU was full",
	362	+ "MetricGroup": "TopdownL4",
	363	+ "ScaleUnit": "100percent of cycles"
	364	+ },
	365	+ {
	366	+ "MetricName": "stall_lob_id_rate",
	367	+ "MetricExpr": "IDR_STALL_LOB_ID / CPU_CYCLES",
	368	+ "BriefDescription": "Fraction of cycles the CPU was stalled and LOB was full",
	369	+ "MetricGroup": "TopdownL4",
	370	+ "ScaleUnit": "100percent of cycles"
	371	+ },
	372	+ {
	373	+ "MetricName": "stall_rob_id_rate",
	374	+ "MetricExpr": "IDR_STALL_ROB_ID / CPU_CYCLES",
	375	+ "BriefDescription": "Fraction of cycles the CPU was stalled and ROB was full",
	376	+ "MetricGroup": "TopdownL4",
	377	+ "ScaleUnit": "100percent of cycles"
	378	+ },
	379	+ {
	380	+ "MetricName": "stall_sob_id_rate",
	381	+ "MetricExpr": "IDR_STALL_SOB_ID / CPU_CYCLES",
	382	+ "BriefDescription": "Fraction of cycles the CPU was stalled and SOB was full",
	383	+ "MetricGroup": "TopdownL4",
	384	+ "ScaleUnit": "100percent of cycles"
	385	+ },
	386	+ {
	387	+ "MetricName": "l1d_cache_access_demand",
	388	+ "MetricExpr": "L1D_CACHE_RW / L1D_CACHE",
	389	+ "BriefDescription": "L1D cache access - demand",
	390	+ "MetricGroup": "Cache",
	391	+ "ScaleUnit": "100percent of cache acceses"
	392	+ },
	393	+ {
	394	+ "MetricName": "l1d_cache_access_prefetces",
	395	+ "MetricExpr": "L1D_CACHE_PRFM / L1D_CACHE",
	396	+ "BriefDescription": "L1D cache access - prefetch",
	397	+ "MetricGroup": "Cache",
	398	+ "ScaleUnit": "100percent of cache acceses"
	399	+ },
	400	+ {
	401	+ "MetricName": "l1d_cache_demand_misses",
	402	+ "MetricExpr": "L1D_CACHE_REFILL_RW / L1D_CACHE",
	403	+ "BriefDescription": "L1D cache demand misses",
	404	+ "MetricGroup": "Cache",
	405	+ "ScaleUnit": "100percent of cache acceses"
	406	+ },
	407	+ {
	408	+ "MetricName": "l1d_cache_demand_misses_read",
	409	+ "MetricExpr": "L1D_CACHE_REFILL_RD / L1D_CACHE",
	410	+ "BriefDescription": "L1D cache demand misses - read",
	411	+ "MetricGroup": "Cache",
	412	+ "ScaleUnit": "100percent of cache acceses"
	413	+ },
	414	+ {
	415	+ "MetricName": "l1d_cache_demand_misses_write",
	416	+ "MetricExpr": "L1D_CACHE_REFILL_WR / L1D_CACHE",
	417	+ "BriefDescription": "L1D cache demand misses - write",
	418	+ "MetricGroup": "Cache",
	419	+ "ScaleUnit": "100percent of cache acceses"
	420	+ },
	421	+ {
	422	+ "MetricName": "l1d_cache_prefetch_misses",
	423	+ "MetricExpr": "L1D_CACHE_REFILL_PRFM / L1D_CACHE",
	424	+ "BriefDescription": "L1D cache prefetch misses",
	425	+ "MetricGroup": "Cache",
	426	+ "ScaleUnit": "100percent of cache acceses"
	427	+ },
	428	+ {
	429	+ "MetricName": "ase_scalar_mix",
	430	+ "MetricExpr": "ASE_SCALAR_SPEC / OP_SPEC",
	431	+ "BriefDescription": "Proportion of advanced SIMD data processing operations (excluding DP_SPEC/LD_SPEC) scalar operations",
	432	+ "MetricGroup": "Instructions",
	433	+ "ScaleUnit": "100percent of cache acceses"
	434	+ },
	435	+ {
	436	+ "MetricName": "ase_vector_mix",
	437	+ "MetricExpr": "ASE_VECTOR_SPEC / OP_SPEC",
	438	+ "BriefDescription": "Proportion of advanced SIMD data processing operations (excluding DP_SPEC/LD_SPEC) vector operations",
	439	+ "MetricGroup": "Instructions",
	440	+ "ScaleUnit": "100percent of cache acceses"
	441	+ }
	442	+]

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/mmu.json+170−0

@@ -0,0 +1,170 @@
	1	+[
	2	+ {
	3	+ "PublicDescription": "Level 2 data translation buffer allocation",
	4	+ "EventCode": "0xD800",
	5	+ "EventName": "MMU_D_OTB_ALLOC",
	6	+ "BriefDescription": "Level 2 data translation buffer allocation"
	7	+ },
	8	+ {
	9	+ "PublicDescription": "Data TLB translation cache hit on S1L2 walk cache entry",
	10	+ "EventCode": "0xd801",
	11	+ "EventName": "MMU_D_TRANS_CACHE_HIT_S1L2_WALK",
	12	+ "BriefDescription": "Data TLB translation cache hit on S1L2 walk cache entry"
	13	+ },
	14	+ {
	15	+ "PublicDescription": "Data TLB translation cache hit on S1L1 walk cache entry",
	16	+ "EventCode": "0xd802",
	17	+ "EventName": "MMU_D_TRANS_CACHE_HIT_S1L1_WALK",
	18	+ "BriefDescription": "Data TLB translation cache hit on S1L1 walk cache entry"
	19	+ },
	20	+ {
	21	+ "PublicDescription": "Data TLB translation cache hit on S1L0 walk cache entry",
	22	+ "EventCode": "0xd803",
	23	+ "EventName": "MMU_D_TRANS_CACHE_HIT_S1L0_WALK",
	24	+ "BriefDescription": "Data TLB translation cache hit on S1L0 walk cache entry"
	25	+ },
	26	+ {
	27	+ "PublicDescription": "Data TLB translation cache hit on S2L2 walk cache entry",
	28	+ "EventCode": "0xd804",
	29	+ "EventName": "MMU_D_TRANS_CACHE_HIT_S2L2_WALK",
	30	+ "BriefDescription": "Data TLB translation cache hit on S2L2 walk cache entry"
	31	+ },
	32	+ {
	33	+ "PublicDescrition": "Data TLB translation cache hit on S2L1 walk cache entry",
	34	+ "EventCode": "0xd805",
	35	+ "EventName": "MMU_D_TRANS_CACHE_HIT_S2L1_WALK",
	36	+ "BriefDescription": "Data TLB translation cache hit on S2L1 walk cache entry"
	37	+ },
	38	+ {
	39	+ "PublicDescrition": "Data TLB translation cache hit on S2L0 walk cache entry",
	40	+ "EventCode": "0xd806",
	41	+ "EventName": "MMU_D_TRANS_CACHE_HIT_S2L0_WALK",
	42	+ "BriefDescription": "Data TLB translation cache hit on S2L0 walk cache entry"
	43	+ },
	44	+ {
	45	+ "PublicDescrition": "Data-side S1 page walk cache lookup",
	46	+ "EventCode": "0xd807",
	47	+ "EventName": "MMU_D_S1_WALK_CACHE_LOOKUP",
	48	+ "BriefDescription": "Data-side S1 page walk cache lookup"
	49	+ },
	50	+ {
	51	+ "PublicDescrition": "Data-side S1 page walk cache refill",
	52	+ "EventCode": "0xd808",
	53	+ "EventName": "MMU_D_S1_WALK_CACHE_REFILL",
	54	+ "BriefDescription": "Data-side S1 page walk cache refill"
	55	+ },
	56	+ {
	57	+ "PublicDescrition": "Data-side S2 page walk cache lookup",
	58	+ "EventCode": "0xd809",
	59	+ "EventName": "MMU_D_S2_WALK_CACHE_LOOKUP",
	60	+ "BriefDescription": "Data-side S2 page walk cache lookup"
	61	+ },
	62	+ {
	63	+ "PublicDescrition": "Data-side S2 page walk cache refill",
	64	+ "EventCode": "0xd80a",
	65	+ "EventName": "MMU_D_S2_WALK_CACHE_REFILL",
	66	+ "BriefDescription": "Data-side S2 page walk cache refill"
	67	+ },
	68	+ {
	69	+ "PublicDescription": "Data-side S1 table walk fault",
	70	+ "EventCode": "0xD80B",
	71	+ "EventName": "MMU_D_S1_WALK_FAULT",
	72	+ "BriefDescription": "Data-side S1 table walk fault"
	73	+ },
	74	+ {
	75	+ "PublicDescription": "Data-side S2 table walk fault",
	76	+ "EventCode": "0xD80C",
	77	+ "EventName": "MMU_D_S2_WALK_FAULT",
	78	+ "BriefDescription": "Data-side S2 table walk fault"
	79	+ },
	80	+ {
	81	+ "PublicDescription": "Data-side table walk steps or descriptor fetches",
	82	+ "EventCode": "0xD80D",
	83	+ "EventName": "MMU_D_WALK_STEPS",
	84	+ "BriefDescription": "Data-side table walk steps or descriptor fetches"
	85	+ },
	86	+ {
	87	+ "PublicDescription": "Level 2 instruction translation buffer allocation",
	88	+ "EventCode": "0xD900",
	89	+ "EventName": "MMU_I_OTB_ALLOC",
	90	+ "BriefDescription": "Level 2 instruction translation buffer allocation"
	91	+ },
	92	+ {
	93	+ "PublicDescrition": "Instruction TLB translation cache hit on S1L2 walk cache entry",
	94	+ "EventCode": "0xd901",
	95	+ "EventName": "MMU_I_TRANS_CACHE_HIT_S1L2_WALK",
	96	+ "BriefDescription": "Instruction TLB translation cache hit on S1L2 walk cache entry"
	97	+ },
	98	+ {
	99	+ "PublicDescrition": "Instruction TLB translation cache hit on S1L1 walk cache entry",
	100	+ "EventCode": "0xd902",
	101	+ "EventName": "MMU_I_TRANS_CACHE_HIT_S1L1_WALK",
	102	+ "BriefDescription": "Instruction TLB translation cache hit on S1L1 walk cache entry"
	103	+ },
	104	+ {
	105	+ "PublicDescrition": "Instruction TLB translation cache hit on S1L0 walk cache entry",
	106	+ "EventCode": "0xd903",
	107	+ "EventName": "MMU_I_TRANS_CACHE_HIT_S1L0_WALK",
	108	+ "BriefDescription": "Instruction TLB translation cache hit on S1L0 walk cache entry"
	109	+ },
	110	+ {
	111	+ "PublicDescrition": "Instruction TLB translation cache hit on S2L2 walk cache entry",
	112	+ "EventCode": "0xd904",
	113	+ "EventName": "MMU_I_TRANS_CACHE_HIT_S2L2_WALK",
	114	+ "BriefDescription": "Instruction TLB translation cache hit on S2L2 walk cache entry"
	115	+ },
	116	+ {
	117	+ "PublicDescrition": "Instruction TLB translation cache hit on S2L1 walk cache entry",
	118	+ "EventCode": "0xd905",
	119	+ "EventName": "MMU_I_TRANS_CACHE_HIT_S2L1_WALK",
	120	+ "BriefDescription": "Instruction TLB translation cache hit on S2L1 walk cache entry"
	121	+ },
	122	+ {
	123	+ "PublicDescrition": "Instruction TLB translation cache hit on S2L0 walk cache entry",
	124	+ "EventCode": "0xd906",
	125	+ "EventName": "MMU_I_TRANS_CACHE_HIT_S2L0_WALK",
	126	+ "BriefDescription": "Instruction TLB translation cache hit on S2L0 walk cache entry"
	127	+ },
	128	+ {
	129	+ "PublicDescrition": "Instruction-side S1 page walk cache lookup",
	130	+ "EventCode": "0xd907",
	131	+ "EventName": "MMU_I_S1_WALK_CACHE_LOOKUP",
	132	+ "BriefDescription": "Instruction-side S1 page walk cache lookup"
	133	+ },
	134	+ {
	135	+ "PublicDescrition": "Instruction-side S1 page walk cache refill",
	136	+ "EventCode": "0xd908",
	137	+ "EventName": "MMU_I_S1_WALK_CACHE_REFILL",
	138	+ "BriefDescription": "Instruction-side S1 page walk cache refill"
	139	+ },
	140	+ {
	141	+ "PublicDescrition": "Instruction-side S2 page walk cache lookup",
	142	+ "EventCode": "0xd909",
	143	+ "EventName": "MMU_I_S2_WALK_CACHE_LOOKUP",
	144	+ "BriefDescription": "Instruction-side S2 page walk cache lookup"
	145	+ },
	146	+ {
	147	+ "PublicDescrition": "Instruction-side S2 page walk cache refill",
	148	+ "EventCode": "0xd90a",
	149	+ "EventName": "MMU_I_S2_WALK_CACHE_REFILL",
	150	+ "BriefDescription": "Instruction-side S2 page walk cache refill"
	151	+ },
	152	+ {
	153	+ "PublicDescription": "Instruction-side S1 table walk fault",
	154	+ "EventCode": "0xD90B",
	155	+ "EventName": "MMU_I_S1_WALK_FAULT",
	156	+ "BriefDescription": "Instruction-side S1 table walk fault"
	157	+ },
	158	+ {
	159	+ "PublicDescription": "Instruction-side S2 table walk fault",
	160	+ "EventCode": "0xD90C",
	161	+ "EventName": "MMU_I_S2_WALK_FAULT",
	162	+ "BriefDescription": "Instruction-side S2 table walk fault"
	163	+ },
	164	+ {
	165	+ "PublicDescription": "Instruction-side table walk steps or descriptor fetches",
	166	+ "EventCode": "0xD90D",
	167	+ "EventName": "MMU_I_WALK_STEPS",
	168	+ "BriefDescription": "Instruction-side table walk steps or descriptor fetches"
	169	+ }
	170	+]

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/pipeline.json+41−0

@@ -0,0 +1,41 @@
	1	+[
	2	+ {
	3	+ "ArchStdEvent": "STALL_FRONTEND",
	4	+ "Errata": "Errata AC03_CPU_29",
	5	+ "BriefDescription": "Impacted by errata, use metrics instead -"
	6	+ },
	7	+ {
	8	+ "ArchStdEvent": "STALL_BACKEND"
	9	+ },
	10	+ {
	11	+ "ArchStdEvent": "STALL",
	12	+ "Errata": "Errata AC03_CPU_29",
	13	+ "BriefDescription": "Impacted by errata, use metrics instead -"
	14	+ },
	15	+ {
	16	+ "ArchStdEvent": "STALL_SLOT_BACKEND"
	17	+ },
	18	+ {
	19	+ "ArchStdEvent": "STALL_SLOT_FRONTEND",
	20	+ "Errata": "Errata AC03_CPU_29",
	21	+ "BriefDescription": "Impacted by errata, use metrics instead -"
	22	+ },
	23	+ {
	24	+ "ArchStdEvent": "STALL_SLOT"
	25	+ },
	26	+ {
	27	+ "ArchStdEvent": "STALL_BACKEND_MEM"
	28	+ },
	29	+ {
	30	+ "PublicDescription": "Frontend stall cycles, TLB",
	31	+ "EventCode": "0x815c",
	32	+ "EventName": "STALL_FRONTEND_TLB",
	33	+ "BriefDescription": "Frontend stall cycles, TLB"
	34	+ },
	35	+ {
	36	+ "PublicDescription": "Backend stall cycles, TLB",
	37	+ "EventCode": "0x8167",
	38	+ "EventName": "STALL_BACKEND_TLB",
	39	+ "BriefDescription": "Backend stall cycles, TLB"
	40	+ }
	41	+]

tools/perf/pmu-events/arch/arm64/ampere/ampereonex/spe.json+14−0

@@ -0,0 +1,14 @@
	1	+[
	2	+ {
	3	+ "ArchStdEvent": "SAMPLE_POP"
	4	+ },
	5	+ {
	6	+ "ArchStdEvent": "SAMPLE_FEED"
	7	+ },
	8	+ {
	9	+ "ArchStdEvent": "SAMPLE_FILTRATE"
	10	+ },
	11	+ {
	12	+ "ArchStdEvent": "SAMPLE_COLLISION"
	13	+ }
	14	+]

tools/perf/pmu-events/arch/arm64/arm/cmn/sys/cmn.json+1−1

@@ -107,7 +107,7 @@ · 106 unmodified lines
107	107	"EventName": "hnf_qos_hh_retry",
108	108	"EventidCode": "0xe",
109	109	"NodeType": "0x5",
110		− "BriefDescription": "Counts number of times a HighHigh priority request is protocolretried at the HN‑F.",
	110	+ "BriefDescription": "Counts number of times a HighHigh priority request is protocolretried at the HN-F.",
111	111	"Unit": "arm_cmn",
112	112	"Compat": "(434\|436\|43c\|43a).*"
113	113	},

tools/perf/pmu-events/arch/arm64/mapfile.csv+1−0

@@ -42,3 +42,4 @@ · 41 unmodified lines
42	42	0x00000000480fd010,v1,hisilicon/hip08,core
43	43	0x00000000500f0000,v1,ampere/emag,core
44	44	0x00000000c00fac30,v1,ampere/ampereone,core
	45	+0x00000000c00fac40,v1,ampere/ampereonex,core

tools/perf/pmu-events/arch/powerpc/mapfile.csv+1−2

@@ -11,8 +11,7 @@ · 10 unmodified lines
11	11	#
12	12	# Multiple PVRs could map to a single JSON file.
13	13	#
14		−
15		−# Power8 entries
16	14	0x004[bcd][[:xdigit:]]{4},1,power8,core
	15	+0x0066[[:xdigit:]]{4},1,power8,core
17	16	0x004e[[:xdigit:]]{4},1,power9,core
18	17	0x0080[[:xdigit:]]{4},1,power10,core

tools/perf/pmu-events/arch/powerpc/power10/datasource.json+14−4

@@ -99,6 +99,11 @@ · 98 unmodified lines
99	99	"EventName": "PM_INST_FROM_L2MISS",
100	100	"BriefDescription": "The processor's instruction cache was reloaded from a source beyond the local core's L2 due to a demand miss."
101	101	},
	102	+ {
	103	+ "EventCode": "0x0003C0000000C040",
	104	+ "EventName": "PM_DATA_FROM_L2MISS_DSRC",
	105	+ "BriefDescription": "The processor's L1 data cache was reloaded from a source beyond the local core's L2 due to a demand miss."
	106	+ },
102	107	{
103	108	"EventCode": "0x000380000010C040",
104	109	"EventName": "PM_INST_FROM_L2MISS_ALL",
@@ -161,9 +166,14 @@ · 56 unmodified lines
161	166	},
162	167	{
163	168	"EventCode": "0x000780000000C040",
164		− "EventName": "PM_INST_FROM_L3MISS",
	169	+ "EventName": "PM_INST_FROM_L3MISS_DSRC",
165	170	"BriefDescription": "The processor's instruction cache was reloaded from beyond the local core's L3 due to a demand miss."
166	171	},
	172	+ {
	173	+ "EventCode": "0x0007C0000000C040",
	174	+ "EventName": "PM_DATA_FROM_L3MISS_DSRC",
	175	+ "BriefDescription": "The processor's L1 data cache was reloaded from beyond the local core's L3 due to a demand miss."
	176	+ },
167	177	{
168	178	"EventCode": "0x000780000010C040",
169	179	"EventName": "PM_INST_FROM_L3MISS_ALL",
@@ -981,7 +991,7 @@ · 811 unmodified lines
981	991	},
982	992	{
983	993	"EventCode": "0x0003C0000000C142",
984		− "EventName": "PM_MRK_DATA_FROM_L2MISS",
	994	+ "EventName": "PM_MRK_DATA_FROM_L2MISS_DSRC",
985	995	"BriefDescription": "The processor's L1 data cache was reloaded from a source beyond the local core's L2 due to a demand miss for a marked instruction."
986	996	},
987	997	{
@@ -1046,12 +1056,12 @@ · 58 unmodified lines
1046	1056	},
1047	1057	{
1048	1058	"EventCode": "0x000780000000C142",
1049		− "EventName": "PM_MRK_INST_FROM_L3MISS",
	1059	+ "EventName": "PM_MRK_INST_FROM_L3MISS_DSRC",
1050	1060	"BriefDescription": "The processor's instruction cache was reloaded from beyond the local core's L3 due to a demand miss for a marked instruction."
1051	1061	},
1052	1062	{
1053	1063	"EventCode": "0x0007C0000000C142",
1054		− "EventName": "PM_MRK_DATA_FROM_L3MISS",
	1064	+ "EventName": "PM_MRK_DATA_FROM_L3MISS_DSRC",
1055	1065	"BriefDescription": "The processor's L1 data cache was reloaded from beyond the local core's L3 due to a demand miss for a marked instruction."
1056	1066	},
1057	1067	{

tools/perf/pmu-events/arch/riscv/mapfile.csv+2−0

@@ -15,3 +15,5 @@ · 14 unmodified lines
15	15	#
16	16	#MVENDORID-MARCHID-MIMPID,Version,Filename,EventType
17	17	0x489-0x8000000000000007-0x[[:xdigit:]]+,v1,sifive/u74,core
	18	+0x5b7-0x0-0x0,v1,thead/c900-legacy,core
	19	+0x67e-0x80000000db0000[89]0-0x[[:xdigit:]]+,v1,starfive/dubhe-80,core

tools/perf/pmu-events/arch/riscv/starfive/dubhe-80/common.json+172−0

@@ -0,0 +1,172 @@
	1	+[
	2	+ {
	3	+ "EventName": "ACCESS_MMU_STLB",
	4	+ "EventCode": "0x1",
	5	+ "BriefDescription": "access MMU STLB"
	6	+ },
	7	+ {
	8	+ "EventName": "MISS_MMU_STLB",
	9	+ "EventCode": "0x2",
	10	+ "BriefDescription": "miss MMU STLB"
	11	+ },
	12	+ {
	13	+ "EventName": "ACCESS_MMU_PTE_C",
	14	+ "EventCode": "0x3",
	15	+ "BriefDescription": "access MMU PTE-Cache"
	16	+ },
	17	+ {
	18	+ "EventName": "MISS_MMU_PTE_C",
	19	+ "EventCode": "0x4",
	20	+ "BriefDescription": "miss MMU PTE-Cache"
	21	+ },
	22	+ {
	23	+ "EventName": "ROB_FLUSH",
	24	+ "EventCode": "0x5",
	25	+ "BriefDescription": "ROB flush (all kinds of exceptions)"
	26	+ },
	27	+ {
	28	+ "EventName": "BTB_PREDICTION_MISS",
	29	+ "EventCode": "0x6",
	30	+ "BriefDescription": "BTB prediction miss"
	31	+ },
	32	+ {
	33	+ "EventName": "ITLB_MISS",
	34	+ "EventCode": "0x7",
	35	+ "BriefDescription": "ITLB miss"
	36	+ },
	37	+ {
	38	+ "EventName": "SYNC_DEL_FETCH_G",
	39	+ "EventCode": "0x8",
	40	+ "BriefDescription": "SYNC delivery a fetch-group"
	41	+ },
	42	+ {
	43	+ "EventName": "ICACHE_MISS",
	44	+ "EventCode": "0x9",
	45	+ "BriefDescription": "ICache miss"
	46	+ },
	47	+ {
	48	+ "EventName": "BPU_BR_RETIRE",
	49	+ "EventCode": "0xA",
	50	+ "BriefDescription": "condition branch instruction retire"
	51	+ },
	52	+ {
	53	+ "EventName": "BPU_BR_MISS",
	54	+ "EventCode": "0xB",
	55	+ "BriefDescription": "condition branch instruction miss"
	56	+ },
	57	+ {
	58	+ "EventName": "RET_INS_RETIRE",
	59	+ "EventCode": "0xC",
	60	+ "BriefDescription": "return instruction retire"
	61	+ },
	62	+ {
	63	+ "EventName": "RET_INS_MISS",
	64	+ "EventCode": "0xD",
	65	+ "BriefDescription": "return instruction miss"
	66	+ },
	67	+ {
	68	+ "EventName": "INDIRECT_JR_MISS",
	69	+ "EventCode": "0xE",
	70	+ "BriefDescription": "indirect JR instruction miss (inlcude without target)"
	71	+ },
	72	+ {
	73	+ "EventName": "IBUF_VAL_ID_NORDY",
	74	+ "EventCode": "0xF",
	75	+ "BriefDescription": "IBUF valid while ID not ready"
	76	+ },
	77	+ {
	78	+ "EventName": "IBUF_NOVAL_ID_RDY",
	79	+ "EventCode": "0x10",
	80	+ "BriefDescription": "IBUF not valid while ID ready"
	81	+ },
	82	+ {
	83	+ "EventName": "REN_INT_PHY_REG_NORDY",
	84	+ "EventCode": "0x11",
	85	+ "BriefDescription": "REN integer physical register file is not ready"
	86	+ },
	87	+ {
	88	+ "EventName": "REN_FP_PHY_REG_NORDY",
	89	+ "EventCode": "0x12",
	90	+ "BriefDescription": "REN floating point physical register file is not ready"
	91	+ },
	92	+ {
	93	+ "EventName": "REN_CP_NORDY",
	94	+ "EventCode": "0x13",
	95	+ "BriefDescription": "REN checkpoint is not ready"
	96	+ },
	97	+ {
	98	+ "EventName": "DEC_VAL_ROB_NORDY",
	99	+ "EventCode": "0x14",
	100	+ "BriefDescription": "DEC is valid and ROB is not ready"
	101	+ },
	102	+ {
	103	+ "EventName": "OOD_FLUSH_LS_DEP",
	104	+ "EventCode": "0x15",
	105	+ "BriefDescription": "out of order flush due to load/store dependency"
	106	+ },
	107	+ {
	108	+ "EventName": "BRU_RET_IJR_INS",
	109	+ "EventCode": "0x16",
	110	+ "BriefDescription": "BRU retire an IJR instruction"
	111	+ },
	112	+ {
	113	+ "EventName": "ACCESS_DTLB",
	114	+ "EventCode": "0x17",
	115	+ "BriefDescription": "access DTLB"
	116	+ },
	117	+ {
	118	+ "EventName": "MISS_DTLB",
	119	+ "EventCode": "0x18",
	120	+ "BriefDescription": "miss DTLB"
	121	+ },
	122	+ {
	123	+ "EventName": "LOAD_INS_DCACHE",
	124	+ "EventCode": "0x19",
	125	+ "BriefDescription": "load instruction access DCache"
	126	+ },
	127	+ {
	128	+ "EventName": "LOAD_INS_MISS_DCACHE",
	129	+ "EventCode": "0x1A",
	130	+ "BriefDescription": "load instruction miss DCache"
	131	+ },
	132	+ {
	133	+ "EventName": "STORE_INS_DCACHE",
	134	+ "EventCode": "0x1B",
	135	+ "BriefDescription": "store/amo instruction access DCache"
	136	+ },
	137	+ {
	138	+ "EventName": "STORE_INS_MISS_DCACHE",
	139	+ "EventCode": "0x1C",
	140	+ "BriefDescription": "store/amo instruction miss DCache"
	141	+ },
	142	+ {
	143	+ "EventName": "LOAD_SCACHE",
	144	+ "EventCode": "0x1D",
	145	+ "BriefDescription": "load access SCache"
	146	+ },
	147	+ {
	148	+ "EventName": "STORE_SCACHE",
	149	+ "EventCode": "0x1E",
	150	+ "BriefDescription": "store access SCache"
	151	+ },
	152	+ {
	153	+ "EventName": "LOAD_MISS_SCACHE",
	154	+ "EventCode": "0x1F",
	155	+ "BriefDescription": "load miss SCache"
	156	+ },
	157	+ {
	158	+ "EventName": "STORE_MISS_SCACHE",
	159	+ "EventCode": "0x20",
	160	+ "BriefDescription": "store miss SCache"
	161	+ },
	162	+ {
	163	+ "EventName": "L2C_PF_REQ",
	164	+ "EventCode": "0x21",
	165	+ "BriefDescription": "L2C data-prefetcher request"
	166	+ },
	167	+ {
	168	+ "EventName": "L2C_PF_HIT",
	169	+ "EventCode": "0x22",
	170	+ "BriefDescription": "L2C data-prefetcher hit"
	171	+ }
	172	+]

tools/perf/pmu-events/arch/riscv/starfive/dubhe-80/firmware.json+68−0

@@ -0,0 +1,68 @@
	1	+[
	2	+ {
	3	+ "ArchStdEvent": "FW_MISALIGNED_LOAD"
	4	+ },
	5	+ {
	6	+ "ArchStdEvent": "FW_MISALIGNED_STORE"
	7	+ },
	8	+ {
	9	+ "ArchStdEvent": "FW_ACCESS_LOAD"
	10	+ },
	11	+ {
	12	+ "ArchStdEvent": "FW_ACCESS_STORE"
	13	+ },
	14	+ {
	15	+ "ArchStdEvent": "FW_ILLEGAL_INSN"
	16	+ },
	17	+ {
	18	+ "ArchStdEvent": "FW_SET_TIMER"
	19	+ },
	20	+ {
	21	+ "ArchStdEvent": "FW_IPI_SENT"
	22	+ },
	23	+ {
	24	+ "ArchStdEvent": "FW_IPI_RECEIVED"
	25	+ },
	26	+ {
	27	+ "ArchStdEvent": "FW_FENCE_I_SENT"
	28	+ },
	29	+ {
	30	+ "ArchStdEvent": "FW_FENCE_I_RECEIVED"
	31	+ },
	32	+ {
	33	+ "ArchStdEvent": "FW_SFENCE_VMA_SENT"
	34	+ },
	35	+ {
	36	+ "ArchStdEvent": "FW_SFENCE_VMA_RECEIVED"
	37	+ },
	38	+ {
	39	+ "ArchStdEvent": "FW_SFENCE_VMA_RECEIVED"
	40	+ },
	41	+ {
	42	+ "ArchStdEvent": "FW_SFENCE_VMA_ASID_RECEIVED"
	43	+ },
	44	+ {
	45	+ "ArchStdEvent": "FW_HFENCE_GVMA_SENT"
	46	+ },
	47	+ {
	48	+ "ArchStdEvent": "FW_HFENCE_GVMA_RECEIVED"
	49	+ },
	50	+ {
	51	+ "ArchStdEvent": "FW_HFENCE_GVMA_VMID_SENT"
	52	+ },
	53	+ {
	54	+ "ArchStdEvent": "FW_HFENCE_GVMA_VMID_RECEIVED"
	55	+ },
	56	+ {
	57	+ "ArchStdEvent": "FW_HFENCE_VVMA_SENT"
	58	+ },
	59	+ {
	60	+ "ArchStdEvent": "FW_HFENCE_VVMA_RECEIVED"
	61	+ },
	62	+ {
	63	+ "ArchStdEvent": "FW_HFENCE_VVMA_ASID_SENT"
	64	+ },
	65	+ {
	66	+ "ArchStdEvent": "FW_HFENCE_VVMA_ASID_RECEIVED"
	67	+ }
	68	+]

tools/perf/pmu-events/arch/riscv/thead/c900-legacy/cache.json+67−0

@@ -0,0 +1,67 @@
	1	+[
	2	+ {
	3	+ "EventName": "L1_ICACHE_ACCESS",
	4	+ "EventCode": "0x00000001",
	5	+ "BriefDescription": "L1 instruction cache access"
	6	+ },
	7	+ {
	8	+ "EventName": "L1_ICACHE_MISS",
	9	+ "EventCode": "0x00000002",
	10	+ "BriefDescription": "L1 instruction cache miss"
	11	+ },
	12	+ {
	13	+ "EventName": "ITLB_MISS",
	14	+ "EventCode": "0x00000003",
	15	+ "BriefDescription": "I-UTLB miss"
	16	+ },
	17	+ {
	18	+ "EventName": "DTLB_MISS",
	19	+ "EventCode": "0x00000004",
	20	+ "BriefDescription": "D-UTLB miss"
	21	+ },
	22	+ {
	23	+ "EventName": "JTLB_MISS",
	24	+ "EventCode": "0x00000005",
	25	+ "BriefDescription": "JTLB miss"
	26	+ },
	27	+ {
	28	+ "EventName": "L1_DCACHE_READ_ACCESS",
	29	+ "EventCode": "0x0000000c",
	30	+ "BriefDescription": "L1 data cache read access"
	31	+ },
	32	+ {
	33	+ "EventName": "L1_DCACHE_READ_MISS",
	34	+ "EventCode": "0x0000000d",
	35	+ "BriefDescription": "L1 data cache read miss"
	36	+ },
	37	+ {
	38	+ "EventName": "L1_DCACHE_WRITE_ACCESS",
	39	+ "EventCode": "0x0000000e",
	40	+ "BriefDescription": "L1 data cache write access"
	41	+ },
	42	+ {
	43	+ "EventName": "L1_DCACHE_WRITE_MISS",
	44	+ "EventCode": "0x0000000f",
	45	+ "BriefDescription": "L1 data cache write miss"
	46	+ },
	47	+ {
	48	+ "EventName": "LL_CACHE_READ_ACCESS",
	49	+ "EventCode": "0x00000010",
	50	+ "BriefDescription": "LL Cache read access"
	51	+ },
	52	+ {
	53	+ "EventName": "LL_CACHE_READ_MISS",
	54	+ "EventCode": "0x00000011",
	55	+ "BriefDescription": "LL Cache read miss"
	56	+ },
	57	+ {
	58	+ "EventName": "LL_CACHE_WRITE_ACCESS",
	59	+ "EventCode": "0x00000012",
	60	+ "BriefDescription": "LL Cache write access"
	61	+ },
	62	+ {
	63	+ "EventName": "LL_CACHE_WRITE_MISS",
	64	+ "EventCode": "0x00000013",
	65	+ "BriefDescription": "LL Cache write miss"
	66	+ }
	67	+]

tools/perf/pmu-events/arch/riscv/thead/c900-legacy/firmware.json+68−0

@@ -0,0 +1,68 @@
	1	+[
	2	+ {
	3	+ "ArchStdEvent": "FW_MISALIGNED_LOAD"
	4	+ },
	5	+ {
	6	+ "ArchStdEvent": "FW_MISALIGNED_STORE"
	7	+ },
	8	+ {
	9	+ "ArchStdEvent": "FW_ACCESS_LOAD"
	10	+ },
	11	+ {
	12	+ "ArchStdEvent": "FW_ACCESS_STORE"
	13	+ },
	14	+ {
	15	+ "ArchStdEvent": "FW_ILLEGAL_INSN"
	16	+ },
	17	+ {
	18	+ "ArchStdEvent": "FW_SET_TIMER"
	19	+ },
	20	+ {
	21	+ "ArchStdEvent": "FW_IPI_SENT"
	22	+ },
	23	+ {
	24	+ "ArchStdEvent": "FW_IPI_RECEIVED"
	25	+ },
	26	+ {
	27	+ "ArchStdEvent": "FW_FENCE_I_SENT"
	28	+ },
	29	+ {
	30	+ "ArchStdEvent": "FW_FENCE_I_RECEIVED"
	31	+ },
	32	+ {
	33	+ "ArchStdEvent": "FW_SFENCE_VMA_SENT"
	34	+ },
	35	+ {
	36	+ "ArchStdEvent": "FW_SFENCE_VMA_RECEIVED"
	37	+ },
	38	+ {
	39	+ "ArchStdEvent": "FW_SFENCE_VMA_RECEIVED"
	40	+ },
	41	+ {
	42	+ "ArchStdEvent": "FW_SFENCE_VMA_ASID_RECEIVED"
	43	+ },
	44	+ {
	45	+ "ArchStdEvent": "FW_HFENCE_GVMA_SENT"
	46	+ },
	47	+ {
	48	+ "ArchStdEvent": "FW_HFENCE_GVMA_RECEIVED"
	49	+ },
	50	+ {
	51	+ "ArchStdEvent": "FW_HFENCE_GVMA_VMID_SENT"
	52	+ },
	53	+ {
	54	+ "ArchStdEvent": "FW_HFENCE_GVMA_VMID_RECEIVED"
	55	+ },
	56	+ {
	57	+ "ArchStdEvent": "FW_HFENCE_VVMA_SENT"
	58	+ },
	59	+ {
	60	+ "ArchStdEvent": "FW_HFENCE_VVMA_RECEIVED"
	61	+ },
	62	+ {
	63	+ "ArchStdEvent": "FW_HFENCE_VVMA_ASID_SENT"
	64	+ },
	65	+ {
	66	+ "ArchStdEvent": "FW_HFENCE_VVMA_ASID_RECEIVED"
	67	+ }
	68	+]

tools/perf/pmu-events/arch/riscv/thead/c900-legacy/instruction.json+72−0

@@ -0,0 +1,72 @@
	1	+[
	2	+ {
	3	+ "EventName": "INST_BRANCH_MISPREDICT",
	4	+ "EventCode": "0x00000006",
	5	+ "BriefDescription": "Mispredicted branch instructions"
	6	+ },
	7	+ {
	8	+ "EventName": "INST_BRANCH",
	9	+ "EventCode": "0x00000007",
	10	+ "BriefDescription": "Retired branch instructions"
	11	+ },
	12	+ {
	13	+ "EventName": "INST_JMP_MISPREDICT",
	14	+ "EventCode": "0x00000008",
	15	+ "BriefDescription": "Indirect branch mispredict"
	16	+ },
	17	+ {
	18	+ "EventName": "INST_JMP",
	19	+ "EventCode": "0x00000009",
	20	+ "BriefDescription": "Retired jmp instructions"
	21	+ },
	22	+ {
	23	+ "EventName": "INST_STORE",
	24	+ "EventCode": "0x0000000b",
	25	+ "BriefDescription": "Retired store instructions"
	26	+ },
	27	+ {
	28	+ "EventName": "INST_ALU",
	29	+ "EventCode": "0x0000001d",
	30	+ "BriefDescription": "Retired ALU instructions"
	31	+ },
	32	+ {
	33	+ "EventName": "INST_LDST",
	34	+ "EventCode": "0x0000001e",
	35	+ "BriefDescription": "Retired Load/Store instructions"
	36	+ },
	37	+ {
	38	+ "EventName": "INST_VECTOR",
	39	+ "EventCode": "0x0000001f",
	40	+ "BriefDescription": "Retired Vector instructions"
	41	+ },
	42	+ {
	43	+ "EventName": "INST_CSR",
	44	+ "EventCode": "0x00000020",
	45	+ "BriefDescription": "Retired CSR instructions"
	46	+ },
	47	+ {
	48	+ "EventName": "INST_SYNC",
	49	+ "EventCode": "0x00000021",
	50	+ "BriefDescription": "Retired sync instructions (AMO/LR/SC instructions)"
	51	+ },
	52	+ {
	53	+ "EventName": "INST_UNALIGNED_ACCESS",
	54	+ "EventCode": "0x00000022",
	55	+ "BriefDescription": "Retired Store/Load instructions with unaligned memory access"
	56	+ },
	57	+ {
	58	+ "EventName": "INST_ECALL",
	59	+ "EventCode": "0x00000025",
	60	+ "BriefDescription": "Retired ecall instructions"
	61	+ },
	62	+ {
	63	+ "EventName": "INST_LONG_JP",
	64	+ "EventCode": "0x00000026",
	65	+ "BriefDescription": "Retired long jump instructions"
	66	+ },
	67	+ {
	68	+ "EventName": "INST_FP",
	69	+ "EventCode": "0x0000002a",
	70	+ "BriefDescription": "Retired FPU instructions"
	71	+ }
	72	+]

tools/perf/pmu-events/arch/riscv/thead/c900-legacy/microarch.json+80−0

@@ -0,0 +1,80 @@
	1	+[
	2	+ {
	3	+ "EventName": "LSU_SPEC_FAIL",
	4	+ "EventCode": "0x0000000a",
	5	+ "BriefDescription": "LSU speculation fail"
	6	+ },
	7	+ {
	8	+ "EventName": "IDU_RF_PIPE_FAIL",
	9	+ "EventCode": "0x00000014",
	10	+ "BriefDescription": "Instruction decode unit launch pipeline failed in RF state"
	11	+ },
	12	+ {
	13	+ "EventName": "IDU_RF_REG_FAIL",
	14	+ "EventCode": "0x00000015",
	15	+ "BriefDescription": "Instruction decode unit launch register file fail in RF state"
	16	+ },
	17	+ {
	18	+ "EventName": "IDU_RF_INSTRUCTION",
	19	+ "EventCode": "0x00000016",
	20	+ "BriefDescription": "retired instruction count of Instruction decode unit in RF (Register File) stage"
	21	+ },
	22	+ {
	23	+ "EventName": "LSU_4K_STALL",
	24	+ "EventCode": "0x00000017",
	25	+ "BriefDescription": "LSU stall times for long distance data access (Over 4K)",
	26	+ "PublicDescription": "This stall occurs when translate virtual address with page offset over 4k"
	27	+ },
	28	+ {
	29	+ "EventName": "LSU_OTHER_STALL",
	30	+ "EventCode": "0x00000018",
	31	+ "BriefDescription": "LSU stall times for other reasons (except the 4k stall)"
	32	+ },
	33	+ {
	34	+ "EventName": "LSU_SQ_OTHER_DIS",
	35	+ "EventCode": "0x00000019",
	36	+ "BriefDescription": "LSU store queue discard others"
	37	+ },
	38	+ {
	39	+ "EventName": "LSU_SQ_DATA_DISCARD",
	40	+ "EventCode": "0x0000001a",
	41	+ "BriefDescription": "LSU store queue discard data (uops)"
	42	+ },
	43	+ {
	44	+ "EventName": "BRANCH_DIRECTION_MISPREDICTION",
	45	+ "EventCode": "0x0000001b",
	46	+ "BriefDescription": "Branch misprediction in BTB"
	47	+ },
	48	+ {
	49	+ "EventName": "BRANCH_DIRECTION_PREDICTION",
	50	+ "EventCode": "0x0000001c",
	51	+ "BriefDescription": "All branch prediction in BTB",
	52	+ "PublicDescription": "This event including both successful prediction and failed prediction in BTB"
	53	+ },
	54	+ {
	55	+ "EventName": "INTERRUPT_ACK_COUNT",
	56	+ "EventCode": "0x00000023",
	57	+ "BriefDescription": "acknowledged interrupt count"
	58	+ },
	59	+ {
	60	+ "EventName": "INTERRUPT_OFF_CYCLE",
	61	+ "EventCode": "0x00000024",
	62	+ "BriefDescription": "PLIC arbitration time when the interrupt is not responded",
	63	+ "PublicDescription": "The arbitration time is recorded while meeting any of the following:\n- CPU is M-mode and MIE == 0\n- CPU is S-mode and delegation and SIE == 0\n"
	64	+ },
	65	+ {
	66	+ "EventName": "IFU_STALLED_CYCLE",
	67	+ "EventCode": "0x00000027",
	68	+ "BriefDescription": "Number of stall cycles of the instruction fetch unit (IFU)."
	69	+ },
	70	+ {
	71	+ "EventName": "IDU_STALLED_CYCLE",
	72	+ "EventCode": "0x00000028",
	73	+ "BriefDescription": "hpcp_backend_stall Number of stall cycles of the instruction decoding unit (IDU) and next-level pipeline unit."
	74	+ },
	75	+ {
	76	+ "EventName": "SYNC_STALL",
	77	+ "EventCode": "0x00000029",
	78	+ "BriefDescription": "Sync instruction stall cycle fence/fence.i/sync/sfence"
	79	+ }
	80	+]

tools/perf/pmu-events/arch/x86/alderlake/adl-metrics.json+8−7

@@ -69,12 +69,6 @@ · 68 unmodified lines
69	69	"MetricName": "C9_Pkg_Residency",
70	70	"ScaleUnit": "100%"
71	71	},
72		− {
73		− "BriefDescription": "Uncore frequency per die [GHZ]",
74		− "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9",
75		− "MetricGroup": "SoC",
76		− "MetricName": "UNCORE_FREQ"
77		− },
78	72	{
79	73	"BriefDescription": "Percentage of cycles spent in System Management Interrupts.",
80	74	"MetricExpr": "((msr@aperf@ - cycles) / msr@aperf@ if msr@smi@ > 0 else 0)",
@@ -809,6 +803,13 @@ · 728 unmodified lines
809	803	"ScaleUnit": "100%",
810	804	"Unit": "cpu_atom"
811	805	},
	806	+ {
	807	+ "BriefDescription": "Uncore frequency per die [GHZ]",
	808	+ "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9",
	809	+ "MetricGroup": "SoC",
	810	+ "MetricName": "UNCORE_FREQ",
	811	+ "Unit": "cpu_core"
	812	+ },
812	813	{
813	814	"BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.",
814	815	"MetricExpr": "(cpu_core@UOPS_DISPATCHED.PORT_0@ + cpu_core@UOPS_DISPATCHED.PORT_1@ + cpu_core@UOPS_DISPATCHED.PORT_5_11@ + cpu_core@UOPS_DISPATCHED.PORT_6@) / (5 * tma_info_core_core_clks)",
@@ -1838,7 +1839,7 @@ · 1023 unmodified lines
1838	1839	},
1839	1840	{
1840	1841	"BriefDescription": "Average number of parallel data read requests to external memory",
1841		− "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / cpu_core@UNC_ARB_DAT_OCCUPANCY.RD\\,cmask\\=1@",
	1842	+ "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / UNC_ARB_DAT_OCCUPANCY.RD@cmask\\=1@",
1842	1843	"MetricGroup": "Mem;MemoryBW;SoC",
1843	1844	"MetricName": "tma_info_system_mem_parallel_reads",
1844	1845	"PublicDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches",

tools/perf/pmu-events/arch/x86/amdzen4/memory-controller.json+101−0

@@ -0,0 +1,101 @@
	1	+[
	2	+ {
	3	+ "EventName": "umc_mem_clk",
	4	+ "PublicDescription": "Number of memory clock cycles.",
	5	+ "EventCode": "0x00",
	6	+ "PerPkg": "1",
	7	+ "Unit": "UMCPMC"
	8	+ },
	9	+ {
	10	+ "EventName": "umc_act_cmd.all",
	11	+ "PublicDescription": "Number of ACTIVATE commands sent.",
	12	+ "EventCode": "0x05",
	13	+ "PerPkg": "1",
	14	+ "Unit": "UMCPMC"
	15	+ },
	16	+ {
	17	+ "EventName": "umc_act_cmd.rd",
	18	+ "PublicDescription": "Number of ACTIVATE commands sent for reads.",
	19	+ "EventCode": "0x05",
	20	+ "RdWrMask": "0x1",
	21	+ "PerPkg": "1",
	22	+ "Unit": "UMCPMC"
	23	+ },
	24	+ {
	25	+ "EventName": "umc_act_cmd.wr",
	26	+ "PublicDescription": "Number of ACTIVATE commands sent for writes.",
	27	+ "EventCode": "0x05",
	28	+ "RdWrMask": "0x2",
	29	+ "PerPkg": "1",
	30	+ "Unit": "UMCPMC"
	31	+ },
	32	+ {
	33	+ "EventName": "umc_pchg_cmd.all",
	34	+ "PublicDescription": "Number of PRECHARGE commands sent.",
	35	+ "EventCode": "0x06",
	36	+ "PerPkg": "1",
	37	+ "Unit": "UMCPMC"
	38	+ },
	39	+ {
	40	+ "EventName": "umc_pchg_cmd.rd",
	41	+ "PublicDescription": "Number of PRECHARGE commands sent for reads.",
	42	+ "EventCode": "0x06",
	43	+ "RdWrMask": "0x1",
	44	+ "PerPkg": "1",
	45	+ "Unit": "UMCPMC"
	46	+ },
	47	+ {
	48	+ "EventName": "umc_pchg_cmd.wr",
	49	+ "PublicDescription": "Number of PRECHARGE commands sent for writes.",
	50	+ "EventCode": "0x06",
	51	+ "RdWrMask": "0x2",
	52	+ "PerPkg": "1",
	53	+ "Unit": "UMCPMC"
	54	+ },
	55	+ {
	56	+ "EventName": "umc_cas_cmd.all",
	57	+ "PublicDescription": "Number of CAS commands sent.",
	58	+ "EventCode": "0x0a",
	59	+ "PerPkg": "1",
	60	+ "Unit": "UMCPMC"
	61	+ },
	62	+ {
	63	+ "EventName": "umc_cas_cmd.rd",
	64	+ "PublicDescription": "Number of CAS commands sent for reads.",
	65	+ "EventCode": "0x0a",
	66	+ "RdWrMask": "0x1",
	67	+ "PerPkg": "1",
	68	+ "Unit": "UMCPMC"
	69	+ },
	70	+ {
	71	+ "EventName": "umc_cas_cmd.wr",
	72	+ "PublicDescription": "Number of CAS commands sent for writes.",
	73	+ "EventCode": "0x0a",
	74	+ "RdWrMask": "0x2",
	75	+ "PerPkg": "1",
	76	+ "Unit": "UMCPMC"
	77	+ },
	78	+ {
	79	+ "EventName": "umc_data_slot_clks.all",
	80	+ "PublicDescription": "Number of clocks used by the data bus.",
	81	+ "EventCode": "0x14",
	82	+ "PerPkg": "1",
	83	+ "Unit": "UMCPMC"
	84	+ },
	85	+ {
	86	+ "EventName": "umc_data_slot_clks.rd",
	87	+ "PublicDescription": "Number of clocks used by the data bus for reads.",
	88	+ "EventCode": "0x14",
	89	+ "RdWrMask": "0x1",
	90	+ "PerPkg": "1",
	91	+ "Unit": "UMCPMC"
	92	+ },
	93	+ {
	94	+ "EventName": "umc_data_slot_clks.wr",
	95	+ "PublicDescription": "Number of clocks used by the data bus for writes.",
	96	+ "EventCode": "0x14",
	97	+ "RdWrMask": "0x2",
	98	+ "PerPkg": "1",
	99	+ "Unit": "UMCPMC"
	100	+ }
	101	+]

tools/perf/pmu-events/arch/x86/amdzen4/recommended.json+84−0

@@ -330,5 +330,89 @@ · 329 unmodified lines
330	330	"MetricGroup": "data_fabric",
331	331	"PerPkg": "1",
332	332	"ScaleUnit": "6.103515625e-5MiB"
	333	+ },
	334	+ {
	335	+ "MetricName": "umc_data_bus_utilization",
	336	+ "BriefDescription": "Memory controller data bus utilization.",
	337	+ "MetricExpr": "d_ratio(umc_data_slot_clks.all / 2, umc_mem_clk)",
	338	+ "MetricGroup": "memory_controller",
	339	+ "PerPkg": "1",
	340	+ "ScaleUnit": "100%"
	341	+ },
	342	+ {
	343	+ "MetricName": "umc_cas_cmd_rate",
	344	+ "BriefDescription": "Memory controller CAS command rate.",
	345	+ "MetricExpr": "d_ratio(umc_cas_cmd.all * 1000, umc_mem_clk)",
	346	+ "MetricGroup": "memory_controller",
	347	+ "PerPkg": "1"
	348	+ },
	349	+ {
	350	+ "MetricName": "umc_cas_cmd_read_ratio",
	351	+ "BriefDescription": "Ratio of memory controller CAS commands for reads.",
	352	+ "MetricExpr": "d_ratio(umc_cas_cmd.rd, umc_cas_cmd.all)",
	353	+ "MetricGroup": "memory_controller",
	354	+ "PerPkg": "1",
	355	+ "ScaleUnit": "100%"
	356	+ },
	357	+ {
	358	+ "MetricName": "umc_cas_cmd_write_ratio",
	359	+ "BriefDescription": "Ratio of memory controller CAS commands for writes.",
	360	+ "MetricExpr": "d_ratio(umc_cas_cmd.wr, umc_cas_cmd.all)",
	361	+ "MetricGroup": "memory_controller",
	362	+ "PerPkg": "1",
	363	+ "ScaleUnit": "100%"
	364	+ },
	365	+ {
	366	+ "MetricName": "umc_mem_read_bandwidth",
	367	+ "BriefDescription": "Estimated memory read bandwidth.",
	368	+ "MetricExpr": "(umc_cas_cmd.rd * 64) / 1e6 / duration_time",
	369	+ "MetricGroup": "memory_controller",
	370	+ "PerPkg": "1",
	371	+ "ScaleUnit": "1MB/s"
	372	+ },
	373	+ {
	374	+ "MetricName": "umc_mem_write_bandwidth",
	375	+ "BriefDescription": "Estimated memory write bandwidth.",
	376	+ "MetricExpr": "(umc_cas_cmd.wr * 64) / 1e6 / duration_time",
	377	+ "MetricGroup": "memory_controller",
	378	+ "PerPkg": "1",
	379	+ "ScaleUnit": "1MB/s"
	380	+ },
	381	+ {
	382	+ "MetricName": "umc_mem_bandwidth",
	383	+ "BriefDescription": "Estimated combined memory bandwidth.",
	384	+ "MetricExpr": "(umc_cas_cmd.all * 64) / 1e6 / duration_time",
	385	+ "MetricGroup": "memory_controller",
	386	+ "PerPkg": "1",
	387	+ "ScaleUnit": "1MB/s"
	388	+ },
	389	+ {
	390	+ "MetricName": "umc_cas_cmd_read_ratio",
	391	+ "BriefDescription": "Ratio of memory controller CAS commands for reads.",
	392	+ "MetricExpr": "d_ratio(umc_cas_cmd.rd, umc_cas_cmd.all)",
	393	+ "MetricGroup": "memory_controller",
	394	+ "PerPkg": "1",
	395	+ "ScaleUnit": "100%"
	396	+ },
	397	+ {
	398	+ "MetricName": "umc_cas_cmd_rate",
	399	+ "BriefDescription": "Memory controller CAS command rate.",
	400	+ "MetricExpr": "d_ratio(umc_cas_cmd.all * 1000, umc_mem_clk)",
	401	+ "MetricGroup": "memory_controller",
	402	+ "PerPkg": "1"
	403	+ },
	404	+ {
	405	+ "MetricName": "umc_activate_cmd_rate",
	406	+ "BriefDescription": "Memory controller ACTIVATE command rate.",
	407	+ "MetricExpr": "d_ratio(umc_act_cmd.all * 1000, umc_mem_clk)",
	408	+ "MetricGroup": "memory_controller",
	409	+ "PerPkg": "1"
	410	+ },
	411	+ {
	412	+ "MetricName": "umc_precharge_cmd_rate",
	413	+ "BriefDescription": "Memory controller PRECHARGE command rate.",
	414	+ "MetricExpr": "d_ratio(umc_pchg_cmd.all * 1000, umc_mem_clk)",
	415	+ "MetricGroup": "memory_controller",
	416	+ "PerPkg": "1"
333	417	}
334	418	]

tools/perf/pmu-events/arch/x86/cascadelakex/clx-metrics.json+6−0

@@ -1862,6 +1862,12 @@ · 1861 unmodified lines
1862	1862	"MetricName": "uncore_frequency",
1863	1863	"ScaleUnit": "1GHz"
1864	1864	},
	1865	+ {
	1866	+ "BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data receive bandwidth (MB/sec)",
	1867	+ "MetricExpr": "UNC_UPI_RxL_FLITS.ALL_DATA * 7.111111111111111 / 1e6 / duration_time",
	1868	+ "MetricName": "upi_data_receive_bw",
	1869	+ "ScaleUnit": "1MB/s"
	1870	+ },
1865	1871	{
1866	1872	"BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data transmit bandwidth (MB/sec)",
1867	1873	"MetricExpr": "UNC_UPI_TxL_FLITS.ALL_DATA * 7.111111111111111 / 1e6 / duration_time",

tools/perf/pmu-events/arch/x86/emeraldrapids/floating-point.json+24−3

@@ -23,26 +23,47 @@ · 22 unmodified lines
23	23	"UMask": "0x10"
24	24	},
25	25	{
26		− "BriefDescription": "FP_ARITH_DISPATCHED.PORT_0",
	26	+ "BriefDescription": "FP_ARITH_DISPATCHED.PORT_0 [This event is alias to FP_ARITH_DISPATCHED.V0]",
27	27	"EventCode": "0xb3",
28	28	"EventName": "FP_ARITH_DISPATCHED.PORT_0",
29	29	"SampleAfterValue": "2000003",
30	30	"UMask": "0x1"
31	31	},
32	32	{
33		− "BriefDescription": "FP_ARITH_DISPATCHED.PORT_1",
	33	+ "BriefDescription": "FP_ARITH_DISPATCHED.PORT_1 [This event is alias to FP_ARITH_DISPATCHED.V1]",
34	34	"EventCode": "0xb3",
35	35	"EventName": "FP_ARITH_DISPATCHED.PORT_1",
36	36	"SampleAfterValue": "2000003",
37	37	"UMask": "0x2"
38	38	},
39	39	{
40		− "BriefDescription": "FP_ARITH_DISPATCHED.PORT_5",
	40	+ "BriefDescription": "FP_ARITH_DISPATCHED.PORT_5 [This event is alias to FP_ARITH_DISPATCHED.V2]",
41	41	"EventCode": "0xb3",
42	42	"EventName": "FP_ARITH_DISPATCHED.PORT_5",
43	43	"SampleAfterValue": "2000003",
44	44	"UMask": "0x4"
45	45	},
	46	+ {
	47	+ "BriefDescription": "FP_ARITH_DISPATCHED.V0 [This event is alias to FP_ARITH_DISPATCHED.PORT_0]",
	48	+ "EventCode": "0xb3",
	49	+ "EventName": "FP_ARITH_DISPATCHED.V0",
	50	+ "SampleAfterValue": "2000003",
	51	+ "UMask": "0x1"
	52	+ },
	53	+ {
	54	+ "BriefDescription": "FP_ARITH_DISPATCHED.V1 [This event is alias to FP_ARITH_DISPATCHED.PORT_1]",
	55	+ "EventCode": "0xb3",
	56	+ "EventName": "FP_ARITH_DISPATCHED.V1",
	57	+ "SampleAfterValue": "2000003",
	58	+ "UMask": "0x2"
	59	+ },
	60	+ {
	61	+ "BriefDescription": "FP_ARITH_DISPATCHED.V2 [This event is alias to FP_ARITH_DISPATCHED.PORT_5]",
	62	+ "EventCode": "0xb3",
	63	+ "EventName": "FP_ARITH_DISPATCHED.V2",
	64	+ "SampleAfterValue": "2000003",
	65	+ "UMask": "0x4"
	66	+ },
46	67	{
47	68	"BriefDescription": "Counts number of SSE/AVX computational 128-bit packed double precision floating-point instructions retired; some instructions will count twice as noted below. Each count represents 2 computation operations, one for each element. Applies to SSE* and AVX* packed double precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT DPP FM(N)ADD/SUB. DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element.",
48	69	"EventCode": "0xc7",

tools/perf/pmu-events/arch/x86/emeraldrapids/pipeline.json+1−17

@@ -1,20 +1,4 @@
1	1	[
2		− {
3		− "BriefDescription": "AMX retired arithmetic BF16 operations.",
4		− "EventCode": "0xce",
5		− "EventName": "AMX_OPS_RETIRED.BF16",
6		− "PublicDescription": "Number of AMX-based retired arithmetic bfloat16 (BF16) floating-point operations. Counts TDPBF16PS FP instructions. SW to use operation multiplier of 4",
7		− "SampleAfterValue": "1000003",
8		− "UMask": "0x2"
9		− },
10		− {
11		− "BriefDescription": "AMX retired arithmetic integer 8-bit operations.",
12		− "EventCode": "0xce",
13		− "EventName": "AMX_OPS_RETIRED.INT8",
14		− "PublicDescription": "Number of AMX-based retired arithmetic integer operations of 8-bit width source operands. Counts TDPB[SS,UU,US,SU]D instructions. SW should use operation multiplier of 8.",
15		− "SampleAfterValue": "1000003",
16		− "UMask": "0x1"
17		− },
18	2	{
19	3	"BriefDescription": "This event is deprecated. Refer to new event ARITH.DIV_ACTIVE",
20	4	"CounterMask": "1",
@@ -505,7 +489,7 @@ · 484 unmodified lines
505	489	"UMask": "0x1"
506	490	},
507	491	{
508		− "BriefDescription": "INT_MISC.UNKNOWN_BRANCH_CYCLES",
	492	+ "BriefDescription": "Bubble cycles of BAClear (Unknown Branch).",
509	493	"EventCode": "0xad",
510	494	"EventName": "INT_MISC.UNKNOWN_BRANCH_CYCLES",
511	495	"MSRIndex": "0x3F7",

tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-interconnect.json+4−4

@@ -4825,11 +4825,11 @@ · 4824 unmodified lines
4825	4825	"Unit": "M3UPI"
4826	4826	},
4827	4827	{
4828		− "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (AD Bouncable)",
	4828	+ "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (AD Bounceable)",
4829	4829	"EventCode": "0x47",
4830	4830	"EventName": "UNC_MDF_CRS_TxR_INSERTS.AD_BNC",
4831	4831	"PerPkg": "1",
4832		− "PublicDescription": "AD Bouncable : Number of allocations into the CRS Egress",
	4832	+ "PublicDescription": "AD Bounceable : Number of allocations into the CRS Egress",
4833	4833	"UMask": "0x1",
4834	4834	"Unit": "MDF"
4835	4835	},
@@ -4861,11 +4861,11 @@ · 25 unmodified lines
4861	4861	"Unit": "MDF"
4862	4862	},
4863	4863	{
4864		− "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (BL Bouncable)",
	4864	+ "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (BL Bounceable)",
4865	4865	"EventCode": "0x47",
4866	4866	"EventName": "UNC_MDF_CRS_TxR_INSERTS.BL_BNC",
4867	4867	"PerPkg": "1",
4868		− "PublicDescription": "BL Bouncable : Number of allocations into the CRS Egress",
	4868	+ "PublicDescription": "BL Bounceable : Number of allocations into the CRS Egress",
4869	4869	"UMask": "0x4",
4870	4870	"Unit": "MDF"
4871	4871	},

tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-io.json+30−0

@@ -1185,6 +1185,36 @@ · 1184 unmodified lines
1185	1185	"UMask": "0x70ff010",
1186	1186	"Unit": "IIO"
1187	1187	},
	1188	+ {
	1189	+ "BriefDescription": ": IOTLB Hits to a 1G Page",
	1190	+ "EventCode": "0x40",
	1191	+ "EventName": "UNC_IIO_IOMMU0.1G_HITS",
	1192	+ "PerPkg": "1",
	1193	+ "PortMask": "0x0000",
	1194	+ "PublicDescription": ": IOTLB Hits to a 1G Page : Counts if a transaction to a 1G page, on its first lookup, hits the IOTLB.",
	1195	+ "UMask": "0x10",
	1196	+ "Unit": "IIO"
	1197	+ },
	1198	+ {
	1199	+ "BriefDescription": ": IOTLB Hits to a 2M Page",
	1200	+ "EventCode": "0x40",
	1201	+ "EventName": "UNC_IIO_IOMMU0.2M_HITS",
	1202	+ "PerPkg": "1",
	1203	+ "PortMask": "0x0000",
	1204	+ "PublicDescription": ": IOTLB Hits to a 2M Page : Counts if a transaction to a 2M page, on its first lookup, hits the IOTLB.",
	1205	+ "UMask": "0x8",
	1206	+ "Unit": "IIO"
	1207	+ },
	1208	+ {
	1209	+ "BriefDescription": ": IOTLB Hits to a 4K Page",
	1210	+ "EventCode": "0x40",
	1211	+ "EventName": "UNC_IIO_IOMMU0.4K_HITS",
	1212	+ "PerPkg": "1",
	1213	+ "PortMask": "0x0000",
	1214	+ "PublicDescription": ": IOTLB Hits to a 4K Page : Counts if a transaction to a 4K page, on its first lookup, hits the IOTLB.",
	1215	+ "UMask": "0x4",
	1216	+ "Unit": "IIO"
	1217	+ },
1188	1218	{
1189	1219	"BriefDescription": ": Context cache hits",
1190	1220	"EventCode": "0x40",

tools/perf/pmu-events/arch/x86/icelakex/icx-metrics.json+6−0

@@ -1846,6 +1846,12 @@ · 1845 unmodified lines
1846	1846	"MetricName": "uncore_frequency",
1847	1847	"ScaleUnit": "1GHz"
1848	1848	},
	1849	+ {
	1850	+ "BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data receive bandwidth (MB/sec)",
	1851	+ "MetricExpr": "UNC_UPI_RxL_FLITS.ALL_DATA * 7.111111111111111 / 1e6 / duration_time",
	1852	+ "MetricName": "upi_data_receive_bw",
	1853	+ "ScaleUnit": "1MB/s"
	1854	+ },
1849	1855	{
1850	1856	"BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data transmit bandwidth (MB/sec)",
1851	1857	"MetricExpr": "UNC_UPI_TxL_FLITS.ALL_DATA * 7.111111111111111 / 1e6 / duration_time",

tools/perf/pmu-events/arch/x86/icelakex/other.json+1−1

@@ -19,7 +19,7 @@ · 18 unmodified lines
19	19	"BriefDescription": "Core cycles where the core was running in a manner where Turbo may be clipped to the AVX512 turbo schedule.",
20	20	"EventCode": "0x28",
21	21	"EventName": "CORE_POWER.LVL2_TURBO_LICENSE",
22		− "PublicDescription": "Core cycles where the core was running with power-delivery for license level 2 (introduced in Skylake Server microarchtecture). This includes high current AVX 512-bit instructions.",
	22	+ "PublicDescription": "Core cycles where the core was running with power-delivery for license level 2 (introduced in Skylake Server microarchitecture). This includes high current AVX 512-bit instructions.",
23	23	"SampleAfterValue": "200003",
24	24	"UMask": "0x20"
25	25	},

tools/perf/pmu-events/arch/x86/icelakex/pipeline.json+1−1

@@ -519,7 +519,7 @@ · 518 unmodified lines
519	519	"BriefDescription": "Cycles when Reservation Station (RS) is empty for the thread",
520	520	"EventCode": "0x5e",
521	521	"EventName": "RS_EVENTS.EMPTY_CYCLES",
522		− "PublicDescription": "Counts cycles during which the reservation station (RS) is empty for this logical processor. This is usually caused when the front-end pipeline runs into stravation periods (e.g. branch mispredictions or i-cache misses)",
	522	+ "PublicDescription": "Counts cycles during which the reservation station (RS) is empty for this logical processor. This is usually caused when the front-end pipeline runs into starvation periods (e.g. branch mispredictions or i-cache misses)",
523	523	"SampleAfterValue": "1000003",
524	524	"UMask": "0x1"
525	525	},

tools/perf/pmu-events/arch/x86/icelakex/uncore-interconnect.json+3−3

@@ -38,7 +38,7 @@ · 37 unmodified lines
38	38	"EventCode": "0x10",
39	39	"EventName": "UNC_I_COHERENT_OPS.CLFLUSH",
40	40	"PerPkg": "1",
41		− "PublicDescription": "Coherent Ops : CLFlush : Counts the number of coherency related operations servied by the IRP",
	41	+ "PublicDescription": "Coherent Ops : CLFlush : Counts the number of coherency related operations serviced by the IRP",
42	42	"UMask": "0x80",
43	43	"Unit": "IRP"
44	44	},
@@ -65,7 +65,7 @@ · 20 unmodified lines
65	65	"EventCode": "0x10",
66	66	"EventName": "UNC_I_COHERENT_OPS.WBMTOI",
67	67	"PerPkg": "1",
68		− "PublicDescription": "Coherent Ops : WbMtoI : Counts the number of coherency related operations servied by the IRP",
	68	+ "PublicDescription": "Coherent Ops : WbMtoI : Counts the number of coherency related operations serviced by the IRP",
69	69	"UMask": "0x40",
70	70	"Unit": "IRP"
71	71	},
@@ -454,7 +454,7 @@ · 382 unmodified lines
454	454	"EventCode": "0x11",
455	455	"EventName": "UNC_I_TRANSACTIONS.WRITES",
456	456	"PerPkg": "1",
457		− "PublicDescription": "Inbound Transaction Count : Writes : Counts the number of Inbound transactions from the IRP to the Uncore. This can be filtered based on request type in addition to the source queue. Note the special filtering equation. We do OR-reduction on the request type. If the SOURCE bit is set, then we also do AND qualification based on the source portID. : Trackes only write requests. Each write request should have a prefetch, so there is no need to explicitly track these requests. For writes that are tickled and have to retry, the counter will be incremented for each retry.",
	457	+ "PublicDescription": "Inbound Transaction Count : Writes : Counts the number of Inbound transactions from the IRP to the Uncore. This can be filtered based on request type in addition to the source queue. Note the special filtering equation. We do OR-reduction on the request type. If the SOURCE bit is set, then we also do AND qualification based on the source portID. : Tracks only write requests. Each write request should have a prefetch, so there is no need to explicitly track these requests. For writes that are tickled and have to retry, the counter will be incremented for each retry.",
458	458	"UMask": "0x2",
459	459	"Unit": "IRP"
460	460	},

tools/perf/pmu-events/arch/x86/mapfile.csv+3−3

@@ -7,7 +7,7 @@ GenuineIntel-6-56,v11,broadwellde,core · 6 unmodified lines
7	7	GenuineIntel-6-4F,v22,broadwellx,core
8	8	GenuineIntel-6-55-[56789ABCDEF],v1.20,cascadelakex,core
9	9	GenuineIntel-6-9[6C],v1.04,elkhartlake,core
10		−GenuineIntel-6-CF,v1.01,emeraldrapids,core
	10	+GenuineIntel-6-CF,v1.02,emeraldrapids,core
11	11	GenuineIntel-6-5[CF],v13,goldmont,core
12	12	GenuineIntel-6-7A,v1.01,goldmontplus,core
13	13	GenuineIntel-6-B6,v1.00,grandridge,core
@@ -15,7 +15,7 @@ GenuineIntel-6-A[DE],v1.01,graniterapids,core · 1 unmodified line
15	15	GenuineIntel-6-(3C\|45\|46),v33,haswell,core
16	16	GenuineIntel-6-3F,v28,haswellx,core
17	17	GenuineIntel-6-7[DE],v1.19,icelake,core
18		−GenuineIntel-6-6[AC],v1.21,icelakex,core
	18	+GenuineIntel-6-6[AC],v1.23,icelakex,core
19	19	GenuineIntel-6-3A,v24,ivybridge,core
20	20	GenuineIntel-6-3E,v24,ivytown,core
21	21	GenuineIntel-6-2D,v24,jaketown,core
@@ -26,7 +26,7 @@ GenuineIntel-6-1[AEF],v4,nehalemep,core · 4 unmodified lines
26	26	GenuineIntel-6-2E,v4,nehalemex,core
27	27	GenuineIntel-6-A7,v1.01,rocketlake,core
28	28	GenuineIntel-6-2A,v19,sandybridge,core
29		−GenuineIntel-6-8F,v1.16,sapphirerapids,core
	29	+GenuineIntel-6-8F,v1.17,sapphirerapids,core
30	30	GenuineIntel-6-AF,v1.00,sierraforest,core
31	31	GenuineIntel-6-(37\|4A\|4C\|4D\|5A),v15,silvermont,core
32	32	GenuineIntel-6-(4E\|5E\|8E\|9E\|A5\|A6),v57,skylake,core

tools/perf/pmu-events/arch/x86/rocketlake/rkl-metrics.json+1−1

@@ -985,7 +985,7 @@ · 984 unmodified lines
985	985	},
986	986	{
987	987	"BriefDescription": "Average number of parallel data read requests to external memory",
988		− "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / cpu@UNC_ARB_DAT_OCCUPANCY.RD\\,cmask\\=1@",
	988	+ "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / UNC_ARB_DAT_OCCUPANCY.RD@cmask\\=1@",
989	989	"MetricGroup": "Mem;MemoryBW;SoC",
990	990	"MetricName": "tma_info_system_mem_parallel_reads",
991	991	"PublicDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches"

tools/perf/pmu-events/arch/x86/sapphirerapids/floating-point.json+24−3

@@ -23,26 +23,47 @@ · 22 unmodified lines
23	23	"UMask": "0x10"
24	24	},
25	25	{
26		− "BriefDescription": "FP_ARITH_DISPATCHED.PORT_0",
	26	+ "BriefDescription": "FP_ARITH_DISPATCHED.PORT_0 [This event is alias to FP_ARITH_DISPATCHED.V0]",
27	27	"EventCode": "0xb3",
28	28	"EventName": "FP_ARITH_DISPATCHED.PORT_0",
29	29	"SampleAfterValue": "2000003",
30	30	"UMask": "0x1"
31	31	},
32	32	{
33		− "BriefDescription": "FP_ARITH_DISPATCHED.PORT_1",
	33	+ "BriefDescription": "FP_ARITH_DISPATCHED.PORT_1 [This event is alias to FP_ARITH_DISPATCHED.V1]",
34	34	"EventCode": "0xb3",
35	35	"EventName": "FP_ARITH_DISPATCHED.PORT_1",
36	36	"SampleAfterValue": "2000003",
37	37	"UMask": "0x2"
38	38	},
39	39	{
40		− "BriefDescription": "FP_ARITH_DISPATCHED.PORT_5",
	40	+ "BriefDescription": "FP_ARITH_DISPATCHED.PORT_5 [This event is alias to FP_ARITH_DISPATCHED.V2]",
41	41	"EventCode": "0xb3",
42	42	"EventName": "FP_ARITH_DISPATCHED.PORT_5",
43	43	"SampleAfterValue": "2000003",
44	44	"UMask": "0x4"
45	45	},
	46	+ {
	47	+ "BriefDescription": "FP_ARITH_DISPATCHED.V0 [This event is alias to FP_ARITH_DISPATCHED.PORT_0]",
	48	+ "EventCode": "0xb3",
	49	+ "EventName": "FP_ARITH_DISPATCHED.V0",
	50	+ "SampleAfterValue": "2000003",
	51	+ "UMask": "0x1"
	52	+ },
	53	+ {
	54	+ "BriefDescription": "FP_ARITH_DISPATCHED.V1 [This event is alias to FP_ARITH_DISPATCHED.PORT_1]",
	55	+ "EventCode": "0xb3",
	56	+ "EventName": "FP_ARITH_DISPATCHED.V1",
	57	+ "SampleAfterValue": "2000003",
	58	+ "UMask": "0x2"
	59	+ },
	60	+ {
	61	+ "BriefDescription": "FP_ARITH_DISPATCHED.V2 [This event is alias to FP_ARITH_DISPATCHED.PORT_5]",
	62	+ "EventCode": "0xb3",
	63	+ "EventName": "FP_ARITH_DISPATCHED.V2",
	64	+ "SampleAfterValue": "2000003",
	65	+ "UMask": "0x4"
	66	+ },
46	67	{
47	68	"BriefDescription": "Counts number of SSE/AVX computational 128-bit packed double precision floating-point instructions retired; some instructions will count twice as noted below. Each count represents 2 computation operations, one for each element. Applies to SSE* and AVX* packed double precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT DPP FM(N)ADD/SUB. DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element.",
48	69	"EventCode": "0xc7",

tools/perf/pmu-events/arch/x86/sapphirerapids/pipeline.json+1−1

@@ -505,7 +505,7 @@ · 504 unmodified lines
505	505	"UMask": "0x1"
506	506	},
507	507	{
508		− "BriefDescription": "INT_MISC.UNKNOWN_BRANCH_CYCLES",
	508	+ "BriefDescription": "Bubble cycles of BAClear (Unknown Branch).",
509	509	"EventCode": "0xad",
510	510	"EventName": "INT_MISC.UNKNOWN_BRANCH_CYCLES",
511	511	"MSRIndex": "0x3F7",

tools/perf/pmu-events/arch/x86/sapphirerapids/spr-metrics.json+6−0

@@ -1964,6 +1964,12 @@ · 1963 unmodified lines
1964	1964	"MetricName": "uncore_frequency",
1965	1965	"ScaleUnit": "1GHz"
1966	1966	},
	1967	+ {
	1968	+ "BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data receive bandwidth (MB/sec)",
	1969	+ "MetricExpr": "UNC_UPI_RxL_FLITS.ALL_DATA * 7.111111111111111 / 1e6 / duration_time",
	1970	+ "MetricName": "upi_data_receive_bw",
	1971	+ "ScaleUnit": "1MB/s"
	1972	+ },
1967	1973	{
1968	1974	"BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data transmit bandwidth (MB/sec)",
1969	1975	"MetricExpr": "UNC_UPI_TxL_FLITS.ALL_DATA * 7.111111111111111 / 1e6 / duration_time",

tools/perf/pmu-events/arch/x86/sapphirerapids/uncore-interconnect.json+4−4

@@ -4825,11 +4825,11 @@ · 4824 unmodified lines
4825	4825	"Unit": "M3UPI"
4826	4826	},
4827	4827	{
4828		− "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (AD Bouncable)",
	4828	+ "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (AD Bounceable)",
4829	4829	"EventCode": "0x47",
4830	4830	"EventName": "UNC_MDF_CRS_TxR_INSERTS.AD_BNC",
4831	4831	"PerPkg": "1",
4832		− "PublicDescription": "AD Bouncable : Number of allocations into the CRS Egress",
	4832	+ "PublicDescription": "AD Bounceable : Number of allocations into the CRS Egress",
4833	4833	"UMask": "0x1",
4834	4834	"Unit": "MDF"
4835	4835	},
@@ -4861,11 +4861,11 @@ · 25 unmodified lines
4861	4861	"Unit": "MDF"
4862	4862	},
4863	4863	{
4864		− "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (BL Bouncable)",
	4864	+ "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (BL Bounceable)",
4865	4865	"EventCode": "0x47",
4866	4866	"EventName": "UNC_MDF_CRS_TxR_INSERTS.BL_BNC",
4867	4867	"PerPkg": "1",
4868		− "PublicDescription": "BL Bouncable : Number of allocations into the CRS Egress",
	4868	+ "PublicDescription": "BL Bounceable : Number of allocations into the CRS Egress",
4869	4869	"UMask": "0x4",
4870	4870	"Unit": "MDF"
4871	4871	},

tools/perf/pmu-events/arch/x86/sapphirerapids/uncore-io.json+30−0

@@ -1249,6 +1249,36 @@ · 1248 unmodified lines
1249	1249	"UMask": "0x70ff010",
1250	1250	"Unit": "IIO"
1251	1251	},
	1252	+ {
	1253	+ "BriefDescription": ": IOTLB Hits to a 1G Page",
	1254	+ "EventCode": "0x40",
	1255	+ "EventName": "UNC_IIO_IOMMU0.1G_HITS",
	1256	+ "PerPkg": "1",
	1257	+ "PortMask": "0x0000",
	1258	+ "PublicDescription": ": IOTLB Hits to a 1G Page : Counts if a transaction to a 1G page, on its first lookup, hits the IOTLB.",
	1259	+ "UMask": "0x10",
	1260	+ "Unit": "IIO"
	1261	+ },
	1262	+ {
	1263	+ "BriefDescription": ": IOTLB Hits to a 2M Page",
	1264	+ "EventCode": "0x40",
	1265	+ "EventName": "UNC_IIO_IOMMU0.2M_HITS",
	1266	+ "PerPkg": "1",
	1267	+ "PortMask": "0x0000",
	1268	+ "PublicDescription": ": IOTLB Hits to a 2M Page : Counts if a transaction to a 2M page, on its first lookup, hits the IOTLB.",
	1269	+ "UMask": "0x8",
	1270	+ "Unit": "IIO"
	1271	+ },
	1272	+ {
	1273	+ "BriefDescription": ": IOTLB Hits to a 4K Page",
	1274	+ "EventCode": "0x40",
	1275	+ "EventName": "UNC_IIO_IOMMU0.4K_HITS",
	1276	+ "PerPkg": "1",
	1277	+ "PortMask": "0x0000",
	1278	+ "PublicDescription": ": IOTLB Hits to a 4K Page : Counts if a transaction to a 4K page, on its first lookup, hits the IOTLB.",
	1279	+ "UMask": "0x4",
	1280	+ "Unit": "IIO"
	1281	+ },
1252	1282	{
1253	1283	"BriefDescription": ": Context cache hits",
1254	1284	"EventCode": "0x40",

tools/perf/pmu-events/arch/x86/skylakex/skx-metrics.json+6−0

@@ -1806,6 +1806,12 @@ · 1805 unmodified lines
1806	1806	"MetricName": "uncore_frequency",
1807	1807	"ScaleUnit": "1GHz"
1808	1808	},
	1809	+ {
	1810	+ "BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data receive bandwidth (MB/sec)",
	1811	+ "MetricExpr": "UNC_UPI_RxL_FLITS.ALL_DATA * 7.111111111111111 / 1e6 / duration_time",
	1812	+ "MetricName": "upi_data_receive_bw",
	1813	+ "ScaleUnit": "1MB/s"
	1814	+ },
1809	1815	{
1810	1816	"BriefDescription": "Intel(R) Ultra Path Interconnect (UPI) data transmit bandwidth (MB/sec)",
1811	1817	"MetricExpr": "UNC_UPI_TxL_FLITS.ALL_DATA * 7.111111111111111 / 1e6 / duration_time",

tools/perf/pmu-events/jevents.py+3−1

@@ -83,7 +83,7 @@ def c_len(s: str) -> int: · 82 unmodified lines
83	83	"""Return the length of s a C string
84	84
85	85	This doesn't handle all escape characters properly. It first assumes
86		− all \ are for escaping, it then adjusts as it will have over counted
	86	+ all \\ are for escaping, it then adjusts as it will have over counted
87	87	\\. The code uses \000 rather than \0 as a terminator as an adjacent
88	88	number would be folded into a string of \0 (ie. "\0" + "5" doesn't
89	89	equal a terminator followed by the number 5 but the escape of
@@ -286,6 +286,7 @@ class JsonEvent: · 196 unmodified lines
286	286	'imx8_ddr': 'imx8_ddr',
287	287	'L3PMC': 'amd_l3',
288	288	'DFPMC': 'amd_df',
	289	+ 'UMCPMC': 'amd_umc',
289	290	'cpu_core': 'cpu_core',
290	291	'cpu_atom': 'cpu_atom',
291	292	'ali_drw': 'ali_drw',
@@ -354,6 +355,7 @@ class JsonEvent: · 62 unmodified lines
354	355	('SampleAfterValue', 'period='),
355	356	('UMask', 'umask='),
356	357	('NodeType', 'type='),
	358	+ ('RdWrMask', 'rdwrmask='),
357	359	]
358	360	for key, value in event_fields:
359	361	if key in jd and jd[key] != '0':

tools/perf/scripts/python/arm-cs-trace-disasm.py+16−14

@@ -45,8 +45,8 @@ parser = OptionParser(option_list=option_list) · 44 unmodified lines
45	45	# Initialize global dicts and regular expression
46	46	disasm_cache = dict()
47	47	cpu_data = dict()
48		−disasm_re = re.compile("^\s*([0-9a-fA-F]+):")
49		−disasm_func_re = re.compile("^\s([0-9a-fA-F]+)\s.:")
	48	+disasm_re = re.compile(r"^\s*([0-9a-fA-F]+):")
	49	+disasm_func_re = re.compile(r"^\s([0-9a-fA-F]+)\s.:")
50	50	cache_size = 64*1024
51	51
52	52	glb_source_file_name = None
@@ -188,6 +188,17 @@ def process_event(param_dict): · 135 unmodified lines
188	188	dso_end = get_optional(param_dict, "dso_map_end")
189	189	symbol = get_optional(param_dict, "symbol")
190	190
	191	+ cpu = sample["cpu"]
	192	+ ip = sample["ip"]
	193	+ addr = sample["addr"]
	194	+
	195	+ # Initialize CPU data if it's empty, and directly return back
	196	+ # if this is the first tracing event for this CPU.
	197	+ if (cpu_data.get(str(cpu) + 'addr') == None):
	198	+ cpu_data[str(cpu) + 'addr'] = addr
	199	+ return
	200	+
	201	+
191	202	if (options.verbose == True):
192	203	print("Event type: %s" % name)
193	204	print_sample(sample)
@@ -209,16 +220,6 @@ def process_event(param_dict): · 15 unmodified lines
209	220	if (name[0:8] != "branches"):
210	221	return
211	222
212		− cpu = sample["cpu"]
213		− ip = sample["ip"]
214		− addr = sample["addr"]
215		−
216		− # Initialize CPU data if it's empty, and directly return back
217		− # if this is the first tracing event for this CPU.
218		− if (cpu_data.get(str(cpu) + 'addr') == None):
219		− cpu_data[str(cpu) + 'addr'] = addr
220		− return
221		−
222	223	# The format for packet is:
223	224	#
224	225	# +------------+------------+------------+
@@ -258,8 +259,9 @@ def process_event(param_dict): · 33 unmodified lines
258	259
259	260	if (options.objdump_name != None):
260	261	# It doesn't need to decrease virtual memory offset for disassembly
261		− # for kernel dso, so in this case we set vm_start to zero.
262		− if (dso == "[kernel.kallsyms]"):
	262	+ # for kernel dso and executable file dso, so in this case we set
	263	+ # vm_start to zero.
	264	+ if (dso == "[kernel.kallsyms]" or dso_start == 0x400000):
263	265	dso_vm_start = 0
264	266	else:
265	267	dso_vm_start = int(dso_start)

tools/perf/scripts/python/compaction-times.py+1−1

@@ -260,7 +260,7 @@ def pr_help(): · 259 unmodified lines
260	260
261	261	comm_re = None
262	262	pid_re = None
263		−pid_regex = "^(\d)-(\d)$\|^(\d*)$"
	263	+pid_regex = r"^(\d)-(\d)$\|^(\d*)$"
264	264
265	265	opt_proc = popt.DISP_DFL
266	266	opt_disp = topt.DISP_ALL

tools/perf/scripts/python/exported-sql-viewer.py+2−2

@@ -677,8 +677,8 @@ class CallGraphModelBase(TreeModel): · 676 unmodified lines
677	677	# sqlite supports GLOB (text only) which uses * and ? and is case sensitive
678	678	if not self.glb.dbref.is_sqlite3:
679	679	# Escape % and _
680		− s = value.replace("%", "\%")
681		− s = s.replace("_", "\_")
	680	+ s = value.replace("%", "\\%")
	681	+ s = s.replace("_", "\\_")
682	682	# Translate * and ? into SQL LIKE pattern characters % and _
683	683	trans = string.maketrans("*?", "%_")
684	684	match = " LIKE '" + str(s).translate(trans) + "'"

tools/perf/tests/Build+14−0

@@ -77,3 +77,17 @@ CFLAGS_python-use.o += -DPYTHONPATH="BUILD_STR($(OUTPUT)python)" -DPYTHON="BUI · 76 unmodified lines
77	77	CFLAGS_dwarf-unwind.o += -fno-optimize-sibling-calls
78	78
79	79	perf-y += workloads/
	80	+
	81	+ifdef SHELLCHECK
	82	+ SHELL_TESTS := $(shell find tests/shell -executable -type f -name '*.sh')
	83	+ TEST_LOGS := $(SHELL_TESTS:tests/shell/%=shell/%.shellcheck_log)
	84	+else
	85	+ SHELL_TESTS :=
	86	+ TEST_LOGS :=
	87	+endif
	88	+
	89	+$(OUTPUT)%.shellcheck_log: %
	90	+ $(call rule_mkdir)
	91	+ $(Q)$(call echo-cmd,test)shellcheck -a -S warning "$<" > $@ \|\| (cat $@ && rm $@ && false)
	92	+
	93	+perf-y += $(TEST_LOGS)

tools/perf/tests/attr.c+1−1

@@ -188,7 +188,7 @@ static int test__attr(struct test_suite *test __maybe_unused, int subtest __mayb · 187 unmodified lines
188	188	if (perf_pmus__num_core_pmus() > 1) {
189	189	/*
190	190	* TODO: Attribute tests hard code the PMU type. If there are >1
191		− * core PMU then each PMU will have a different type whic
	191	+ * core PMU then each PMU will have a different type which
192	192	* requires additional support.
193	193	*/
194	194	pr_debug("Skip test on hybrid systems");

tools/perf/tests/attr/base-record+1−1

@@ -6,7 +6,7 @@ flags=0\|8 · 5 unmodified lines
6	6	cpu=*
7	7	type=0\|1
8	8	size=136
9		−config=0
	9	+config=0\|1
10	10	sample_period=*
11	11	sample_type=263
12	12	read_format=0\|4\|20

tools/perf/tests/attr/test-record-user-regs-no-sve-aarch64+1−1

@@ -6,4 +6,4 @@ args = --no-bpf-event --user-regs=vg kill >/dev/null 2>&1 · 5 unmodified lines
6	6	ret = 129
7	7	test_ret = true
8	8	arch = aarch64
9		−auxv = auxv["AT_HWCAP"] & 0x200000 == 0
	9	+auxv = auxv["AT_HWCAP"] & 0x400000 == 0

tools/perf/tests/attr/test-record-user-regs-sve-aarch64+1−1

@@ -6,7 +6,7 @@ args = --no-bpf-event --user-regs=vg kill >/dev/null 2>&1 · 5 unmodified lines
6	6	ret = 1
7	7	test_ret = true
8	8	arch = aarch64
9		−auxv = auxv["AT_HWCAP"] & 0x200000 == 0x200000
	9	+auxv = auxv["AT_HWCAP"] & 0x400000 == 0x400000
10	10	kernel_since = 6.1
11	11
12	12	[event:base-record]

tools/perf/tests/builtin-test.c+15−2

@@ -14,6 +14,7 @@ · 13 unmodified lines
14	14	#include <sys/wait.h>
15	15	#include <sys/stat.h>
16	16	#include "builtin.h"
	17	+#include "config.h"
17	18	#include "hist.h"
18	19	#include "intlist.h"
19	20	#include "tests.h"
@@ -32,6 +33,7 @@ · 12 unmodified lines
32	33
33	34	static bool dont_fork;
34	35	const char *dso_to_test;
	36	+const char *test_objdump_path = "objdump";
35	37
36	38	/*
37	39	* List of architecture specific tests. Not a weak symbol as the array length is
@@ -60,8 +62,6 @@ static struct test_suite *generic_tests[] = { · 22 unmodified lines
60	62	&suite__pmu,
61	63	&suite__pmu_events,
62	64	&suite__dso_data,
63		− &suite__dso_data_cache,
64		− &suite__dso_data_reopen,
65	65	&suite__perf_evsel__roundtrip_name_test,
66	66	#ifdef HAVE_LIBTRACEEVENT
67	67	&suite__perf_evsel__tp_sched_test,
@@ -513,6 +513,15 @@ static int run_workload(const char work, int argc, const char *argv) · 445 unmodified lines
513	513	return -1;
514	514	}
515	515
	516	+static int perf_test__config(const char var, const char value,
	517	+ void *data __maybe_unused)
	518	+{
	519	+ if (!strcmp(var, "annotate.objdump"))
	520	+ test_objdump_path = value;
	521	+
	522	+ return 0;
	523	+}
	524	+
516	525	int cmd_test(int argc, const char **argv)
517	526	{
518	527	const char *test_usage[] = {
@@ -529,6 +538,8 @@ int cmd_test(int argc, const char **argv) · 10 unmodified lines
529	538	"Do not fork for testcase"),
530	539	OPT_STRING('w', "workload", &workload, "work", "workload to run for testing"),
531	540	OPT_STRING(0, "dso", &dso_to_test, "dso", "dso to test"),
	541	+ OPT_STRING(0, "objdump", &test_objdump_path, "path",
	542	+ "objdump binary to use for disassembly and annotations"),
532	543	OPT_END()
533	544	};
534	545	const char * const test_subcommands[] = { "list", NULL };
@@ -538,6 +549,8 @@ int cmd_test(int argc, const char **argv) · 3 unmodified lines
538	549	if (ret < 0)
539	550	return ret;
540	551
	552	+ perf_config(perf_test__config, NULL);
	553	+
541	554	/* Unbuffered output */
542	555	setvbuf(stdout, NULL, _IONBF, 0);
543	556

tools/perf/tests/code-reading.c+25−55

@@ -185,7 +185,7 @@ static int read_via_objdump(const char filename, u64 addr, void buf, · 184 unmodified lines
185	185	int ret;
186	186
187	187	fmt = "%s -z -d --start-address=0x%"PRIx64" --stop-address=0x%"PRIx64" %s";
188		− ret = snprintf(cmd, sizeof(cmd), fmt, "objdump", addr, addr + len,
	188	+ ret = snprintf(cmd, sizeof(cmd), fmt, test_objdump_path, addr, addr + len,
189	189	filename);
190	190	if (ret <= 0 \|\| (size_t)ret >= sizeof(cmd))
191	191	return -1;
@@ -511,38 +511,6 @@ static void fs_something(void) · 319 unmodified lines
511	511	}
512	512	}
513	513
514		−#ifdef __s390x__
515		−#include "header.h" // for get_cpuid()
516		−#endif
517		−
518		−static const char *do_determine_event(bool excl_kernel)
519		−{
520		− const char *event = excl_kernel ? "cycles:u" : "cycles";
521		−
522		−#ifdef __s390x__
523		− char cpuid[128], model[16], model_c[16], cpum_cf_v[16];
524		− unsigned int family;
525		− int ret, cpum_cf_a;
526		−
527		− if (get_cpuid(cpuid, sizeof(cpuid)))
528		− goto out_clocks;
529		− ret = sscanf(cpuid, "%*[^,],%u,%[^,],%[^,],%[^,],%x", &family, model_c,
530		− model, cpum_cf_v, &cpum_cf_a);
531		− if (ret != 5) /* Not available */
532		− goto out_clocks;
533		− if (excl_kernel && (cpum_cf_a & 4))
534		− return event;
535		− if (!excl_kernel && (cpum_cf_a & 2))
536		− return event;
537		−
538		− /* Fall through: missing authorization */
539		−out_clocks:
540		− event = excl_kernel ? "cpu-clock:u" : "cpu-clock";
541		−
542		−#endif
543		− return event;
544		−}
545		−
546	514	static void do_something(void)
547	515	{
548	516	fs_something();
@@ -583,8 +551,10 @@ static int do_test_code_reading(bool try_kcore) · 34 unmodified lines
583	551	int err = -1, ret;
584	552	pid_t pid;
585	553	struct map *map;
586		− bool have_vmlinux, have_kcore, excl_kernel = false;
	554	+ bool have_vmlinux, have_kcore;
587	555	struct dso *dso;
	556	+ const char *events[] = { "cycles", "cycles:u", "cpu-clock", "cpu-clock:u", NULL };
	557	+ int evidx = 0;
588	558
589	559	pid = getpid();
590	560
@@ -618,7 +588,7 @@ static int do_test_code_reading(bool try_kcore) · 27 unmodified lines
618	588
619	589	/* No point getting kernel events if there is no kernel object */
620	590	if (!have_vmlinux && !have_kcore)
621		− excl_kernel = true;
	591	+ evidx++;
622	592
623	593	threads = thread_map__new_by_tid(pid);
624	594	if (!threads) {
@@ -640,13 +610,13 @@ static int do_test_code_reading(bool try_kcore) · 15 unmodified lines
640	610	goto out_put;
641	611	}
642	612
643		− cpus = perf_cpu_map__new(NULL);
	613	+ cpus = perf_cpu_map__new_online_cpus();
644	614	if (!cpus) {
645	615	pr_debug("perf_cpu_map__new failed\n");
646	616	goto out_put;
647	617	}
648	618
649		− while (1) {
	619	+ while (events[evidx]) {
650	620	const char *str;
651	621
652	622	evlist = evlist__new();
@@ -657,7 +627,7 @@ static int do_test_code_reading(bool try_kcore) · 4 unmodified lines
657	627
658	628	perf_evlist__set_maps(&evlist->core, cpus, threads);
659	629
660		− str = do_determine_event(excl_kernel);
	630	+ str = events[evidx];
661	631	pr_debug("Parsing event '%s'\n", str);
662	632	ret = parse_event(evlist, str);
663	633	if (ret < 0) {
@@ -675,32 +645,32 @@ static int do_test_code_reading(bool try_kcore) · 11 unmodified lines
675	645
676	646	ret = evlist__open(evlist);
677	647	if (ret < 0) {
678		− if (!excl_kernel) {
679		− excl_kernel = true;
680		− /*
681		− * Both cpus and threads are now owned by evlist
682		− * and will be freed by following perf_evlist__set_maps
683		− * call. Getting reference to keep them alive.
684		− */
685		− perf_cpu_map__get(cpus);
686		− perf_thread_map__get(threads);
687		− perf_evlist__set_maps(&evlist->core, NULL, NULL);
688		− evlist__delete(evlist);
689		− evlist = NULL;
690		− continue;
691		− }
	648	+ evidx++;
692	649
693		− if (verbose > 0) {
	650	+ if (events[evidx] == NULL && verbose > 0) {
694	651	char errbuf[512];
695	652	evlist__strerror_open(evlist, errno, errbuf, sizeof(errbuf));
696	653	pr_debug("perf_evlist__open() failed!\n%s\n", errbuf);
697	654	}
698	655
699		− goto out_put;
	656	+ /*
	657	+ * Both cpus and threads are now owned by evlist
	658	+ * and will be freed by following perf_evlist__set_maps
	659	+ * call. Getting reference to keep them alive.
	660	+ */
	661	+ perf_cpu_map__get(cpus);
	662	+ perf_thread_map__get(threads);
	663	+ perf_evlist__set_maps(&evlist->core, NULL, NULL);
	664	+ evlist__delete(evlist);
	665	+ evlist = NULL;
	666	+ continue;
700	667	}
701	668	break;
702	669	}
703	670
	671	+ if (events[evidx] == NULL)
	672	+ goto out_put;
	673	+
704	674	ret = evlist__mmap(evlist, UINT_MAX);
705	675	if (ret < 0) {
706	676	pr_debug("evlist__mmap failed\n");
@@ -721,7 +691,7 @@ static int do_test_code_reading(bool try_kcore) · 14 unmodified lines
721	691	err = TEST_CODE_READING_NO_KERNEL_OBJ;
722	692	else if (!have_vmlinux && !try_kcore)
723	693	err = TEST_CODE_READING_NO_VMLINUX;
724		− else if (excl_kernel)
	694	+ else if (strstr(events[evidx], ":u"))
725	695	err = TEST_CODE_READING_NO_ACCESS;
726	696	else
727	697	err = TEST_CODE_READING_OK;

tools/perf/tests/cpumap.c+1−1

@@ -213,7 +213,7 @@ static int test__cpu_map_intersect(struct test_suite *test __maybe_unused, · 212 unmodified lines
213	213
214	214	static int test__cpu_map_equal(struct test_suite *test __maybe_unused, int subtest __maybe_unused)
215	215	{
216		− struct perf_cpu_map *any = perf_cpu_map__dummy_new();
	216	+ struct perf_cpu_map *any = perf_cpu_map__new_any_cpu();
217	217	struct perf_cpu_map *one = perf_cpu_map__new("1");
218	218	struct perf_cpu_map *two = perf_cpu_map__new("2");
219	219	struct perf_cpu_map *empty = perf_cpu_map__intersect(one, two);

tools/perf/tests/dso-data.c+12−3

@@ -394,6 +394,15 @@ static int test__dso_data_reopen(struct test_suite *test __maybe_unused, int sub · 393 unmodified lines
394	394	return 0;
395	395	}
396	396
397		−DEFINE_SUITE("DSO data read", dso_data);
398		−DEFINE_SUITE("DSO data cache", dso_data_cache);
399		−DEFINE_SUITE("DSO data reopen", dso_data_reopen);
	397	+
	398	+static struct test_case tests__dso_data[] = {
	399	+ TEST_CASE("read", dso_data),
	400	+ TEST_CASE("cache", dso_data_cache),
	401	+ TEST_CASE("reopen", dso_data_reopen),
	402	+ { .name = NULL, }
	403	+};
	404	+
	405	+struct test_suite suite__dso_data = {
	406	+ .desc = "DSO data tests",
	407	+ .test_cases = tests__dso_data,
	408	+};

tools/perf/tests/keep-tracking.c+1−1

@@ -81,7 +81,7 @@ static int test__keep_tracking(struct test_suite *test __maybe_unused, int subte · 80 unmodified lines
81	81	threads = thread_map__new(-1, getpid(), UINT_MAX);
82	82	CHECK_NOT_NULL__(threads);
83	83
84		− cpus = perf_cpu_map__new(NULL);
	84	+ cpus = perf_cpu_map__new_online_cpus();
85	85	CHECK_NOT_NULL__(cpus);
86	86
87	87	evlist = evlist__new();

tools/perf/tests/make+1−1

@@ -183,7 +183,7 @@ run += make_install_prefix_slash · 182 unmodified lines
183	183	# run += make_install_pdf
184	184	run += make_minimal
185	185
186		−old_libbpf := $(shell echo '\#include <bpf/libbpf.h>' \| $(CC) -E -dM -x c -\| egrep -q "define[[:space:]]+LIBBPF_MAJOR_VERSION[[:space:]]+0{1}")
	186	+old_libbpf := $(shell echo '\#include <bpf/libbpf.h>' \| $(CC) -E -dM -x c -\| grep -q -E "define[[:space:]]+LIBBPF_MAJOR_VERSION[[:space:]]+0{1}")
187	187
188	188	ifneq ($(old_libbpf),)
189	189	run += make_libbpf_dynamic

tools/perf/tests/maps.c+38−23

@@ -14,44 +14,59 @@ struct map_def { · 13 unmodified lines
14	14	u64 end;
15	15	};
16	16
	17	+struct check_maps_cb_args {
	18	+ struct map_def *merged;
	19	+ unsigned int i;
	20	+};
	21	+
	22	+static int check_maps_cb(struct map map, void data)
	23	+{
	24	+ struct check_maps_cb_args *args = data;
	25	+ struct map_def *merged = &args->merged[args->i];
	26	+
	27	+ if (map__start(map) != merged->start \|\|
	28	+ map__end(map) != merged->end \|\|
	29	+ strcmp(map__dso(map)->name, merged->name) \|\|
	30	+ refcount_read(map__refcnt(map)) != 1) {
	31	+ return 1;
	32	+ }
	33	+ args->i++;
	34	+ return 0;
	35	+}
	36	+
	37	+static int failed_cb(struct map map, void data __maybe_unused)
	38	+{
	39	+ pr_debug("\tstart: %" PRIu64 " end: %" PRIu64 " name: '%s' refcnt: %d\n",
	40	+ map__start(map),
	41	+ map__end(map),
	42	+ map__dso(map)->name,
	43	+ refcount_read(map__refcnt(map)));
	44	+
	45	+ return 0;
	46	+}
	47	+
17	48	static int check_maps(struct map_def merged, unsigned int size, struct maps maps)
18	49	{
19		− struct map_rb_node *rb_node;
20		− unsigned int i = 0;
21	50	bool failed = false;
22	51
23	52	if (maps__nr_maps(maps) != size) {
24	53	pr_debug("Expected %d maps, got %d", size, maps__nr_maps(maps));
25	54	failed = true;
26	55	} else {
27		− maps__for_each_entry(maps, rb_node) {
28		− struct map *map = rb_node->map;
29		−
30		− if (map__start(map) != merged[i].start \|\|
31		− map__end(map) != merged[i].end \|\|
32		− strcmp(map__dso(map)->name, merged[i].name) \|\|
33		− refcount_read(map__refcnt(map)) != 1) {
34		− failed = true;
35		− }
36		− i++;
37		− }
	56	+ struct check_maps_cb_args args = {
	57	+ .merged = merged,
	58	+ .i = 0,
	59	+ };
	60	+ failed = maps__for_each_map(maps, check_maps_cb, &args);
38	61	}
39	62	if (failed) {
40	63	pr_debug("Expected:\n");
41		− for (i = 0; i < size; i++) {
	64	+ for (unsigned int i = 0; i < size; i++) {
42	65	pr_debug("\tstart: %" PRIu64 " end: %" PRIu64 " name: '%s' refcnt: 1\n",
43	66	merged[i].start, merged[i].end, merged[i].name);
44	67	}
45	68	pr_debug("Got:\n");
46		− maps__for_each_entry(maps, rb_node) {
47		− struct map *map = rb_node->map;
48		−
49		− pr_debug("\tstart: %" PRIu64 " end: %" PRIu64 " name: '%s' refcnt: %d\n",
50		− map__start(map),
51		− map__end(map),
52		− map__dso(map)->name,
53		− refcount_read(map__refcnt(map)));
54		− }
	69	+ maps__for_each_map(maps, failed_cb, NULL);
55	70	}
56	71	return failed ? TEST_FAIL : TEST_OK;
57	72	}

tools/perf/tests/mmap-basic.c+1−1

@@ -52,7 +52,7 @@ static int test__basic_mmap(struct test_suite *test __maybe_unused, int subtest · 51 unmodified lines
52	52	return -1;
53	53	}
54	54
55		− cpus = perf_cpu_map__new(NULL);
	55	+ cpus = perf_cpu_map__new_online_cpus();
56	56	if (cpus == NULL) {
57	57	pr_debug("perf_cpu_map__new\n");
58	58	goto out_free_threads;

tools/perf/tests/openat-syscall-all-cpus.c+1−1

@@ -37,7 +37,7 @@ static int test__openat_syscall_event_on_all_cpus(struct test_suite *test __mayb · 36 unmodified lines
37	37	return -1;
38	38	}
39	39
40		− cpus = perf_cpu_map__new(NULL);
	40	+ cpus = perf_cpu_map__new_online_cpus();
41	41	if (cpus == NULL) {
42	42	pr_debug("perf_cpu_map__new\n");
43	43	goto out_thread_map_delete;

tools/perf/tests/parse-events.c+187−73

@@ -162,6 +162,22 @@ static int test__checkevent_numeric(struct evlist *evlist) · 161 unmodified lines
162	162	return TEST_OK;
163	163	}
164	164
	165	+
	166	+static int assert_hw(struct perf_evsel evsel, enum perf_hw_id id, const char name)
	167	+{
	168	+ struct perf_pmu *pmu;
	169	+
	170	+ if (evsel->attr.type == PERF_TYPE_HARDWARE) {
	171	+ TEST_ASSERT_VAL("wrong config", test_perf_config(evsel, id));
	172	+ return 0;
	173	+ }
	174	+ pmu = perf_pmus__find_by_type(evsel->attr.type);
	175	+
	176	+ TEST_ASSERT_VAL("unexpected PMU type", pmu);
	177	+ TEST_ASSERT_VAL("PMU missing event", perf_pmu__have_event(pmu, name));
	178	+ return 0;
	179	+}
	180	+
165	181	static int test__checkevent_symbolic_name(struct evlist *evlist)
166	182	{
167	183	struct perf_evsel *evsel;
@@ -169,10 +185,12 @@ static int test__checkevent_symbolic_name(struct evlist *evlist) · 1 unmodified line
169	185	TEST_ASSERT_VAL("wrong number of entries", 0 != evlist->core.nr_entries);
170	186
171	187	perf_evlist__for_each_evsel(&evlist->core, evsel) {
172		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->attr.type);
173		− TEST_ASSERT_VAL("wrong config",
174		− test_perf_config(evsel, PERF_COUNT_HW_INSTRUCTIONS));
	188	+ int ret = assert_hw(evsel, PERF_COUNT_HW_INSTRUCTIONS, "instructions");
	189	+
	190	+ if (ret)
	191	+ return ret;
175	192	}
	193	+
176	194	return TEST_OK;
177	195	}
178	196
@@ -183,8 +201,10 @@ static int test__checkevent_symbolic_name_config(struct evlist *evlist) · 4 unmodified lines
183	201	TEST_ASSERT_VAL("wrong number of entries", 0 != evlist->core.nr_entries);
184	202
185	203	perf_evlist__for_each_evsel(&evlist->core, evsel) {
186		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->attr.type);
187		− TEST_ASSERT_VAL("wrong config", test_perf_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	204	+ int ret = assert_hw(evsel, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	205	+
	206	+ if (ret)
	207	+ return ret;
188	208	/*
189	209	* The period value gets configured within evlist__config,
190	210	* while this test executes only parse events method.
@@ -861,10 +881,14 @@ static int test__group1(struct evlist *evlist) · 670 unmodified lines
861	881	evlist__nr_groups(evlist) == num_core_entries());
862	882
863	883	for (int i = 0; i < num_core_entries(); i++) {
	884	+ int ret;
	885	+
864	886	/* instructions:k */
865	887	evsel = leader = (i == 0 ? evlist__first(evlist) : evsel__next(evsel));
866		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
867		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_INSTRUCTIONS));
	888	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_INSTRUCTIONS, "instructions");
	889	+ if (ret)
	890	+ return ret;
	891	+
868	892	TEST_ASSERT_VAL("wrong exclude_user", evsel->core.attr.exclude_user);
869	893	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
870	894	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -878,8 +902,10 @@ static int test__group1(struct evlist *evlist) · 7 unmodified lines
878	902
879	903	/* cycles:upp */
880	904	evsel = evsel__next(evsel);
881		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
882		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	905	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	906	+ if (ret)
	907	+ return ret;
	908	+
883	909	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
884	910	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
885	911	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -907,6 +933,8 @@ static int test__group2(struct evlist *evlist) · 21 unmodified lines
907	933	TEST_ASSERT_VAL("wrong number of groups", 1 == evlist__nr_groups(evlist));
908	934
909	935	evlist__for_each_entry(evlist, evsel) {
	936	+ int ret;
	937	+
910	938	if (evsel->core.attr.type == PERF_TYPE_SOFTWARE) {
911	939	/* faults + :ku modifier */
912	940	leader = evsel;
@@ -939,8 +967,10 @@ static int test__group2(struct evlist *evlist) · 26 unmodified lines
939	967	continue;
940	968	}
941	969	/* cycles:k */
942		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
943		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	970	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	971	+ if (ret)
	972	+ return ret;
	973	+
944	974	TEST_ASSERT_VAL("wrong exclude_user", evsel->core.attr.exclude_user);
945	975	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
946	976	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -957,6 +987,7 @@ static int test__group2(struct evlist *evlist) · 10 unmodified lines
957	987	static int test__group3(struct evlist *evlist __maybe_unused)
958	988	{
959	989	struct evsel evsel, group1_leader = NULL, *group2_leader = NULL;
	990	+ int ret;
960	991
961	992	TEST_ASSERT_VAL("wrong number of entries",
962	993	evlist->core.nr_entries == (3 * perf_pmus__num_core_pmus() + 2));
@@ -1045,8 +1076,10 @@ static int test__group3(struct evlist *evlist __maybe_unused) · 82 unmodified lines
1045	1076	continue;
1046	1077	}
1047	1078	/* instructions:u */
1048		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1049		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_INSTRUCTIONS));
	1079	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_INSTRUCTIONS, "instructions");
	1080	+ if (ret)
	1081	+ return ret;
	1082	+
1050	1083	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1051	1084	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
1052	1085	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -1070,10 +1103,14 @@ static int test__group4(struct evlist *evlist __maybe_unused) · 17 unmodified lines
1070	1103	num_core_entries() == evlist__nr_groups(evlist));
1071	1104
1072	1105	for (int i = 0; i < num_core_entries(); i++) {
	1106	+ int ret;
	1107	+
1073	1108	/* cycles:u + p */
1074	1109	evsel = leader = (i == 0 ? evlist__first(evlist) : evsel__next(evsel));
1075		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1076		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	1110	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1111	+ if (ret)
	1112	+ return ret;
	1113	+
1077	1114	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1078	1115	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
1079	1116	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -1089,8 +1126,10 @@ static int test__group4(struct evlist *evlist __maybe_unused) · 9 unmodified lines
1089	1126
1090	1127	/* instructions:kp + p */
1091	1128	evsel = evsel__next(evsel);
1092		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1093		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_INSTRUCTIONS));
	1129	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_INSTRUCTIONS, "instructions");
	1130	+ if (ret)
	1131	+ return ret;
	1132	+
1094	1133	TEST_ASSERT_VAL("wrong exclude_user", evsel->core.attr.exclude_user);
1095	1134	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1096	1135	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -1108,6 +1147,7 @@ static int test__group4(struct evlist *evlist __maybe_unused) · 11 unmodified lines
1108	1147	static int test__group5(struct evlist *evlist __maybe_unused)
1109	1148	{
1110	1149	struct evsel evsel = NULL, leader;
	1150	+ int ret;
1111	1151
1112	1152	TEST_ASSERT_VAL("wrong number of entries",
1113	1153	evlist->core.nr_entries == (5 * num_core_entries()));
@@ -1117,8 +1157,10 @@ static int test__group5(struct evlist *evlist __maybe_unused) · 3 unmodified lines
1117	1157	for (int i = 0; i < num_core_entries(); i++) {
1118	1158	/* cycles + G */
1119	1159	evsel = leader = (i == 0 ? evlist__first(evlist) : evsel__next(evsel));
1120		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1121		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	1160	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1161	+ if (ret)
	1162	+ return ret;
	1163	+
1122	1164	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1123	1165	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1124	1166	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1133,8 +1175,10 @@ static int test__group5(struct evlist *evlist __maybe_unused) · 8 unmodified lines
1133	1175
1134	1176	/* instructions + G */
1135	1177	evsel = evsel__next(evsel);
1136		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1137		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_INSTRUCTIONS));
	1178	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_INSTRUCTIONS, "instructions");
	1179	+ if (ret)
	1180	+ return ret;
	1181	+
1138	1182	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1139	1183	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1140	1184	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1148,8 +1192,10 @@ static int test__group5(struct evlist *evlist __maybe_unused) · 7 unmodified lines
1148	1192	for (int i = 0; i < num_core_entries(); i++) {
1149	1193	/* cycles:G */
1150	1194	evsel = leader = evsel__next(evsel);
1151		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1152		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	1195	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1196	+ if (ret)
	1197	+ return ret;
	1198	+
1153	1199	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1154	1200	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1155	1201	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1164,8 +1210,10 @@ static int test__group5(struct evlist *evlist __maybe_unused) · 8 unmodified lines
1164	1210
1165	1211	/* instructions:G */
1166	1212	evsel = evsel__next(evsel);
1167		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1168		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_INSTRUCTIONS));
	1213	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_INSTRUCTIONS, "instructions");
	1214	+ if (ret)
	1215	+ return ret;
	1216	+
1169	1217	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1170	1218	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1171	1219	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1178,8 +1226,10 @@ static int test__group5(struct evlist *evlist __maybe_unused) · 6 unmodified lines
1178	1226	for (int i = 0; i < num_core_entries(); i++) {
1179	1227	/* cycles */
1180	1228	evsel = evsel__next(evsel);
1181		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1182		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	1229	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1230	+ if (ret)
	1231	+ return ret;
	1232	+
1183	1233	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1184	1234	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1185	1235	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1201,10 +1251,14 @@ static int test__group_gh1(struct evlist *evlist) · 15 unmodified lines
1201	1251	evlist__nr_groups(evlist) == num_core_entries());
1202	1252
1203	1253	for (int i = 0; i < num_core_entries(); i++) {
	1254	+ int ret;
	1255	+
1204	1256	/* cycles + :H group modifier */
1205	1257	evsel = leader = (i == 0 ? evlist__first(evlist) : evsel__next(evsel));
1206		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1207		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	1258	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1259	+ if (ret)
	1260	+ return ret;
	1261	+
1208	1262	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1209	1263	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1210	1264	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1218,8 +1272,10 @@ static int test__group_gh1(struct evlist *evlist) · 7 unmodified lines
1218	1272
1219	1273	/* cache-misses:G + :H group modifier */
1220	1274	evsel = evsel__next(evsel);
1221		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1222		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CACHE_MISSES));
	1275	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CACHE_MISSES, "cache-misses");
	1276	+ if (ret)
	1277	+ return ret;
	1278	+
1223	1279	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1224	1280	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1225	1281	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1242,10 +1298,14 @@ static int test__group_gh2(struct evlist *evlist) · 16 unmodified lines
1242	1298	evlist__nr_groups(evlist) == num_core_entries());
1243	1299
1244	1300	for (int i = 0; i < num_core_entries(); i++) {
	1301	+ int ret;
	1302	+
1245	1303	/* cycles + :G group modifier */
1246	1304	evsel = leader = (i == 0 ? evlist__first(evlist) : evsel__next(evsel));
1247		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1248		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	1305	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1306	+ if (ret)
	1307	+ return ret;
	1308	+
1249	1309	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1250	1310	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1251	1311	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1259,8 +1319,10 @@ static int test__group_gh2(struct evlist *evlist) · 7 unmodified lines
1259	1319
1260	1320	/* cache-misses:H + :G group modifier */
1261	1321	evsel = evsel__next(evsel);
1262		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1263		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CACHE_MISSES));
	1322	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CACHE_MISSES, "cache-misses");
	1323	+ if (ret)
	1324	+ return ret;
	1325	+
1264	1326	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1265	1327	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1266	1328	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1283,10 +1345,14 @@ static int test__group_gh3(struct evlist *evlist) · 16 unmodified lines
1283	1345	evlist__nr_groups(evlist) == num_core_entries());
1284	1346
1285	1347	for (int i = 0; i < num_core_entries(); i++) {
	1348	+ int ret;
	1349	+
1286	1350	/* cycles:G + :u group modifier */
1287	1351	evsel = leader = (i == 0 ? evlist__first(evlist) : evsel__next(evsel));
1288		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1289		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	1352	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1353	+ if (ret)
	1354	+ return ret;
	1355	+
1290	1356	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1291	1357	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
1292	1358	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -1300,8 +1366,10 @@ static int test__group_gh3(struct evlist *evlist) · 7 unmodified lines
1300	1366
1301	1367	/* cache-misses:H + :u group modifier */
1302	1368	evsel = evsel__next(evsel);
1303		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1304		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CACHE_MISSES));
	1369	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CACHE_MISSES, "cache-misses");
	1370	+ if (ret)
	1371	+ return ret;
	1372	+
1305	1373	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1306	1374	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
1307	1375	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -1324,10 +1392,14 @@ static int test__group_gh4(struct evlist *evlist) · 16 unmodified lines
1324	1392	evlist__nr_groups(evlist) == num_core_entries());
1325	1393
1326	1394	for (int i = 0; i < num_core_entries(); i++) {
	1395	+ int ret;
	1396	+
1327	1397	/* cycles:G + :uG group modifier */
1328	1398	evsel = leader = (i == 0 ? evlist__first(evlist) : evsel__next(evsel));
1329		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1330		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	1399	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1400	+ if (ret)
	1401	+ return ret;
	1402	+
1331	1403	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1332	1404	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
1333	1405	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -1341,8 +1413,10 @@ static int test__group_gh4(struct evlist *evlist) · 7 unmodified lines
1341	1413
1342	1414	/* cache-misses:H + :uG group modifier */
1343	1415	evsel = evsel__next(evsel);
1344		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1345		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CACHE_MISSES));
	1416	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CACHE_MISSES, "cache-misses");
	1417	+ if (ret)
	1418	+ return ret;
	1419	+
1346	1420	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1347	1421	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
1348	1422	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -1363,10 +1437,14 @@ static int test__leader_sample1(struct evlist *evlist) · 14 unmodified lines
1363	1437	evlist->core.nr_entries == (3 * num_core_entries()));
1364	1438
1365	1439	for (int i = 0; i < num_core_entries(); i++) {
	1440	+ int ret;
	1441	+
1366	1442	/* cycles - sampling group leader */
1367	1443	evsel = leader = (i == 0 ? evlist__first(evlist) : evsel__next(evsel));
1368		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1369		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	1444	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1445	+ if (ret)
	1446	+ return ret;
	1447	+
1370	1448	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1371	1449	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1372	1450	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1379,8 +1457,10 @@ static int test__leader_sample1(struct evlist *evlist) · 6 unmodified lines
1379	1457
1380	1458	/* cache-misses - not sampling */
1381	1459	evsel = evsel__next(evsel);
1382		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1383		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CACHE_MISSES));
	1460	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CACHE_MISSES, "cache-misses");
	1461	+ if (ret)
	1462	+ return ret;
	1463	+
1384	1464	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1385	1465	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1386	1466	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1392,8 +1472,10 @@ static int test__leader_sample1(struct evlist *evlist) · 5 unmodified lines
1392	1472
1393	1473	/* branch-misses - not sampling */
1394	1474	evsel = evsel__next(evsel);
1395		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1396		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_BRANCH_MISSES));
	1475	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_BRANCH_MISSES, "branch-misses");
	1476	+ if (ret)
	1477	+ return ret;
	1478	+
1397	1479	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1398	1480	TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->core.attr.exclude_kernel);
1399	1481	TEST_ASSERT_VAL("wrong exclude_hv", !evsel->core.attr.exclude_hv);
@@ -1415,10 +1497,14 @@ static int test__leader_sample2(struct evlist *evlist __maybe_unused) · 15 unmodified lines
1415	1497	evlist->core.nr_entries == (2 * num_core_entries()));
1416	1498
1417	1499	for (int i = 0; i < num_core_entries(); i++) {
	1500	+ int ret;
	1501	+
1418	1502	/* instructions - sampling group leader */
1419	1503	evsel = leader = (i == 0 ? evlist__first(evlist) : evsel__next(evsel));
1420		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1421		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_INSTRUCTIONS));
	1504	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_INSTRUCTIONS, "instructions");
	1505	+ if (ret)
	1506	+ return ret;
	1507	+
1422	1508	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1423	1509	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
1424	1510	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -1431,8 +1517,10 @@ static int test__leader_sample2(struct evlist *evlist __maybe_unused) · 6 unmodified lines
1431	1517
1432	1518	/* branch-misses - not sampling */
1433	1519	evsel = evsel__next(evsel);
1434		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1435		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_BRANCH_MISSES));
	1520	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_BRANCH_MISSES, "branch-misses");
	1521	+ if (ret)
	1522	+ return ret;
	1523	+
1436	1524	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
1437	1525	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
1438	1526	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
@@ -1472,10 +1560,14 @@ static int test__pinned_group(struct evlist *evlist) · 33 unmodified lines
1472	1560	evlist->core.nr_entries == (3 * num_core_entries()));
1473	1561
1474	1562	for (int i = 0; i < num_core_entries(); i++) {
	1563	+ int ret;
	1564	+
1475	1565	/* cycles - group leader */
1476	1566	evsel = leader = (i == 0 ? evlist__first(evlist) : evsel__next(evsel));
1477		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1478		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	1567	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1568	+ if (ret)
	1569	+ return ret;
	1570	+
1479	1571	TEST_ASSERT_VAL("wrong group name", !evsel->group_name);
1480	1572	TEST_ASSERT_VAL("wrong leader", evsel__has_leader(evsel, leader));
1481	1573	/* TODO: The group modifier is not copied to the split group leader. */
@@ -1484,13 +1576,18 @@ static int test__pinned_group(struct evlist *evlist) · 2 unmodified lines
1484	1576
1485	1577	/* cache-misses - can not be pinned, but will go on with the leader */
1486	1578	evsel = evsel__next(evsel);
1487		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1488		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CACHE_MISSES));
	1579	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CACHE_MISSES, "cache-misses");
	1580	+ if (ret)
	1581	+ return ret;
	1582	+
1489	1583	TEST_ASSERT_VAL("wrong pinned", !evsel->core.attr.pinned);
1490	1584
1491	1585	/* branch-misses - ditto */
1492	1586	evsel = evsel__next(evsel);
1493		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_BRANCH_MISSES));
	1587	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_BRANCH_MISSES, "branch-misses");
	1588	+ if (ret)
	1589	+ return ret;
	1590	+
1494	1591	TEST_ASSERT_VAL("wrong pinned", !evsel->core.attr.pinned);
1495	1592	}
1496	1593	return TEST_OK;
@@ -1517,10 +1614,14 @@ static int test__exclusive_group(struct evlist *evlist) · 20 unmodified lines
1517	1614	evlist->core.nr_entries == 3 * num_core_entries());
1518	1615
1519	1616	for (int i = 0; i < num_core_entries(); i++) {
	1617	+ int ret;
	1618	+
1520	1619	/* cycles - group leader */
1521	1620	evsel = leader = (i == 0 ? evlist__first(evlist) : evsel__next(evsel));
1522		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1523		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
	1621	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1622	+ if (ret)
	1623	+ return ret;
	1624	+
1524	1625	TEST_ASSERT_VAL("wrong group name", !evsel->group_name);
1525	1626	TEST_ASSERT_VAL("wrong leader", evsel__has_leader(evsel, leader));
1526	1627	/* TODO: The group modifier is not copied to the split group leader. */
@@ -1529,13 +1630,18 @@ static int test__exclusive_group(struct evlist *evlist) · 2 unmodified lines
1529	1630
1530	1631	/* cache-misses - can not be pinned, but will go on with the leader */
1531	1632	evsel = evsel__next(evsel);
1532		− TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
1533		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CACHE_MISSES));
	1633	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_CACHE_MISSES, "cache-misses");
	1634	+ if (ret)
	1635	+ return ret;
	1636	+
1534	1637	TEST_ASSERT_VAL("wrong exclusive", !evsel->core.attr.exclusive);
1535	1638
1536	1639	/* branch-misses - ditto */
1537	1640	evsel = evsel__next(evsel);
1538		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_BRANCH_MISSES));
	1641	+ ret = assert_hw(&evsel->core, PERF_COUNT_HW_BRANCH_MISSES, "branch-misses");
	1642	+ if (ret)
	1643	+ return ret;
	1644	+
1539	1645	TEST_ASSERT_VAL("wrong exclusive", !evsel->core.attr.exclusive);
1540	1646	}
1541	1647	return TEST_OK;
@@ -1677,9 +1783,11 @@ static int test__checkevent_raw_pmu(struct evlist *evlist) · 135 unmodified lines
1677	1783	static int test__sym_event_slash(struct evlist *evlist)
1678	1784	{
1679	1785	struct evsel *evsel = evlist__first(evlist);
	1786	+ int ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1787	+
	1788	+ if (ret)
	1789	+ return ret;
1680	1790
1681		− TEST_ASSERT_VAL("wrong type", evsel->core.attr.type == PERF_TYPE_HARDWARE);
1682		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
1683	1791	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
1684	1792	return TEST_OK;
1685	1793	}
@@ -1687,9 +1795,11 @@ static int test__sym_event_slash(struct evlist *evlist) · 1 unmodified line
1687	1795	static int test__sym_event_dc(struct evlist *evlist)
1688	1796	{
1689	1797	struct evsel *evsel = evlist__first(evlist);
	1798	+ int ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1799	+
	1800	+ if (ret)
	1801	+ return ret;
1690	1802
1691		− TEST_ASSERT_VAL("wrong type", evsel->core.attr.type == PERF_TYPE_HARDWARE);
1692		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
1693	1803	TEST_ASSERT_VAL("wrong exclude_user", evsel->core.attr.exclude_user);
1694	1804	return TEST_OK;
1695	1805	}
@@ -1697,9 +1807,11 @@ static int test__sym_event_dc(struct evlist *evlist) · 1 unmodified line
1697	1807	static int test__term_equal_term(struct evlist *evlist)
1698	1808	{
1699	1809	struct evsel *evsel = evlist__first(evlist);
	1810	+ int ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1811	+
	1812	+ if (ret)
	1813	+ return ret;
1700	1814
1701		− TEST_ASSERT_VAL("wrong type", evsel->core.attr.type == PERF_TYPE_HARDWARE);
1702		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
1703	1815	TEST_ASSERT_VAL("wrong name setting", strcmp(evsel->name, "name") == 0);
1704	1816	return TEST_OK;
1705	1817	}
@@ -1707,9 +1819,11 @@ static int test__term_equal_term(struct evlist *evlist) · 1 unmodified line
1707	1819	static int test__term_equal_legacy(struct evlist *evlist)
1708	1820	{
1709	1821	struct evsel *evsel = evlist__first(evlist);
	1822	+ int ret = assert_hw(&evsel->core, PERF_COUNT_HW_CPU_CYCLES, "cycles");
	1823	+
	1824	+ if (ret)
	1825	+ return ret;
1710	1826
1711		− TEST_ASSERT_VAL("wrong type", evsel->core.attr.type == PERF_TYPE_HARDWARE);
1712		− TEST_ASSERT_VAL("wrong config", test_config(evsel, PERF_COUNT_HW_CPU_CYCLES));
1713	1827	TEST_ASSERT_VAL("wrong name setting", strcmp(evsel->name, "l1d") == 0);
1714	1828	return TEST_OK;
1715	1829	}
@@ -2549,7 +2663,7 @@ static int test__pmu_events(struct test_suite *test __maybe_unused, int subtest · 833 unmodified lines
2549	2663	if (strchr(ent->d_name, '.'))
2550	2664	continue;
2551	2665
2552		− /* exclude parametrized ones (name contains '?') */
	2666	+ /* exclude parameterized ones (name contains '?') */
2553	2667	n = snprintf(pmu_event, sizeof(pmu_event), "%s%s", path, ent->d_name);
2554	2668	if (n >= PATH_MAX) {
2555	2669	pr_err("pmu event name crossed PATH_MAX(%d) size\n", PATH_MAX);
@@ -2578,7 +2692,7 @@ static int test__pmu_events(struct test_suite *test __maybe_unused, int subtest · 22 unmodified lines
2578	2692	fclose(file);
2579	2693
2580	2694	if (is_event_parameterized == 1) {
2581		− pr_debug("skipping parametrized PMU event: %s which contains ?\n", pmu_event);
	2695	+ pr_debug("skipping parameterized PMU event: %s which contains ?\n", pmu_event);
2582	2696	continue;
2583	2697	}
2584	2698

tools/perf/tests/perf-time-to-tsc.c+1−1

@@ -93,7 +93,7 @@ static int test__perf_time_to_tsc(struct test_suite *test __maybe_unused, int su · 92 unmodified lines
93	93	threads = thread_map__new(-1, getpid(), UINT_MAX);
94	94	CHECK_NOT_NULL__(threads);
95	95
96		− cpus = perf_cpu_map__new(NULL);
	96	+ cpus = perf_cpu_map__new_online_cpus();
97	97	CHECK_NOT_NULL__(cpus);
98	98
99	99	evlist = evlist__new();

tools/perf/tests/shell/coresight/memcpy_thread/memcpy_thread.c+0−1

@@ -42,7 +42,6 @@ static pthread_t new_thr(void (fn) (void arg), void arg) · 41 unmodified lines
42	42	int main(int argc, char **argv)
43	43	{
44	44	unsigned long i, len, size, thr;
45		− pthread_t threads[256];
46	45	struct args args[256];
47	46	long long v;
48	47

tools/perf/tests/shell/coresight/thread_loop/thread_loop.c+0−1

@@ -57,7 +57,6 @@ static pthread_t new_thr(void (fn) (void arg), void arg) · 56 unmodified lines
57	57	int main(int argc, char **argv)
58	58	{
59	59	unsigned int i, len, thr;
60		− pthread_t threads[256];
61	60	struct args args[256];
62	61
63	62	if (argc < 3) {

tools/perf/tests/shell/coresight/unroll_loop_thread/unroll_loop_thread.c+0−1

@@ -51,7 +51,6 @@ static pthread_t new_thr(void (fn) (void arg), void arg) · 50 unmodified lines
51	51	int main(int argc, char **argv)
52	52	{
53	53	unsigned int i, thr;
54		− pthread_t threads[256];
55	54	struct args args[256];
56	55
57	56	if (argc < 2) {

tools/perf/tests/shell/diff.sh+108−0

@@ -0,0 +1,108 @@
	1	+#!/bin/sh
	2	+# perf diff tests
	3	+# SPDX-License-Identifier: GPL-2.0
	4	+
	5	+set -e
	6	+
	7	+err=0
	8	+perfdata1=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
	9	+perfdata2=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
	10	+perfdata3=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
	11	+testprog="perf test -w thloop"
	12	+
	13	+shelldir=$(dirname "$0")
	14	+# shellcheck source=lib/perf_has_symbol.sh
	15	+. "${shelldir}"/lib/perf_has_symbol.sh
	16	+
	17	+testsym="test_loop"
	18	+
	19	+skip_test_missing_symbol ${testsym}
	20	+
	21	+cleanup() {
	22	+ rm -rf "${perfdata1}"
	23	+ rm -rf "${perfdata1}".old
	24	+ rm -rf "${perfdata2}"
	25	+ rm -rf "${perfdata2}".old
	26	+ rm -rf "${perfdata3}"
	27	+ rm -rf "${perfdata3}".old
	28	+
	29	+ trap - EXIT TERM INT
	30	+}
	31	+
	32	+trap_cleanup() {
	33	+ cleanup
	34	+ exit 1
	35	+}
	36	+trap trap_cleanup EXIT TERM INT
	37	+
	38	+make_data() {
	39	+ file="$1"
	40	+ if ! perf record -o "${file}" ${testprog} 2> /dev/null
	41	+ then
	42	+ echo "Workload record [Failed record]"
	43	+ echo 1
	44	+ return
	45	+ fi
	46	+ if ! perf report -i "${file}" -q \| grep -q "${testsym}"
	47	+ then
	48	+ echo "Workload record [Failed missing output]"
	49	+ echo 1
	50	+ return
	51	+ fi
	52	+ echo 0
	53	+}
	54	+
	55	+test_two_files() {
	56	+ echo "Basic two file diff test"
	57	+ err=$(make_data "${perfdata1}")
	58	+ if [ $err != 0 ]
	59	+ then
	60	+ return
	61	+ fi
	62	+ err=$(make_data "${perfdata2}")
	63	+ if [ $err != 0 ]
	64	+ then
	65	+ return
	66	+ fi
	67	+
	68	+ if ! perf diff "${perfdata1}" "${perfdata2}" \| grep -q "${testsym}"
	69	+ then
	70	+ echo "Basic two file diff test [Failed diff]"
	71	+ err=1
	72	+ return
	73	+ fi
	74	+ echo "Basic two file diff test [Success]"
	75	+}
	76	+
	77	+test_three_files() {
	78	+ echo "Basic three file diff test"
	79	+ err=$(make_data "${perfdata1}")
	80	+ if [ $err != 0 ]
	81	+ then
	82	+ return
	83	+ fi
	84	+ err=$(make_data "${perfdata2}")
	85	+ if [ $err != 0 ]
	86	+ then
	87	+ return
	88	+ fi
	89	+ err=$(make_data "${perfdata3}")
	90	+ if [ $err != 0 ]
	91	+ then
	92	+ return
	93	+ fi
	94	+
	95	+ if ! perf diff "${perfdata1}" "${perfdata2}" "${perfdata3}" \| grep -q "${testsym}"
	96	+ then
	97	+ echo "Basic three file diff test [Failed diff]"
	98	+ err=1
	99	+ return
	100	+ fi
	101	+ echo "Basic three file diff test [Success]"
	102	+}
	103	+
	104	+test_two_files
	105	+test_three_files
	106	+
	107	+cleanup
	108	+exit $err

tools/perf/tests/shell/lib/perf_has_symbol.sh+21−0

@@ -0,0 +1,21 @@
	1	+#!/bin/sh
	2	+# SPDX-License-Identifier: GPL-2.0
	3	+
	4	+perf_has_symbol()
	5	+{
	6	+ if perf test -vv "Symbols" 2>&1 \| grep "[[:space:]]$1$"; then
	7	+ echo "perf does have symbol '$1'"
	8	+ return 0
	9	+ fi
	10	+ echo "perf does not have symbol '$1'"
	11	+ return 1
	12	+}
	13	+
	14	+skip_test_missing_symbol()
	15	+{
	16	+ if ! perf_has_symbol "$1" ; then
	17	+ echo "perf is missing symbols - skipping test"
	18	+ exit 2
	19	+ fi
	20	+ return 0
	21	+}

tools/perf/tests/shell/lib/setup_python.sh+16−0

@@ -0,0 +1,16 @@
	1	+#!/bin/sh
	2	+# SPDX-License-Identifier: GPL-2.0
	3	+
	4	+if [ "x$PYTHON" = "x" ]
	5	+then
	6	+ python3 --version >/dev/null 2>&1 && PYTHON=python3
	7	+fi
	8	+if [ "x$PYTHON" = "x" ]
	9	+then
	10	+ python --version >/dev/null 2>&1 && PYTHON=python
	11	+fi
	12	+if [ "x$PYTHON" = "x" ]
	13	+then
	14	+ echo Skipping test, python not detected please set environment variable PYTHON.
	15	+ exit 2
	16	+fi

tools/perf/tests/shell/list.sh+19−0

@@ -0,0 +1,19 @@
	1	+#!/bin/sh
	2	+# perf list tests
	3	+# SPDX-License-Identifier: GPL-2.0
	4	+
	5	+set -e
	6	+err=0
	7	+
	8	+shelldir=$(dirname "$0")
	9	+# shellcheck source=lib/setup_python.sh
	10	+. "${shelldir}"/lib/setup_python.sh
	11	+
	12	+test_list_json() {
	13	+ echo "Json output test"
	14	+ perf list -j \| $PYTHON -m json.tool
	15	+ echo "Json output test [Success]"
	16	+}
	17	+
	18	+test_list_json
	19	+exit $err

tools/perf/tests/shell/pipe_test.sh+8−1

@@ -2,10 +2,17 @@ · 1 unmodified line
2	2	# perf pipe recording and injection test
3	3	# SPDX-License-Identifier: GPL-2.0
4	4
	5	+shelldir=$(dirname "$0")
	6	+# shellcheck source=lib/perf_has_symbol.sh
	7	+. "${shelldir}"/lib/perf_has_symbol.sh
	8	+
	9	+sym="noploop"
	10	+
	11	+skip_test_missing_symbol ${sym}
	12	+
5	13	data=$(mktemp /tmp/perf.data.XXXXXX)
6	14	prog="perf test -w noploop"
7	15	task="perf"
8		−sym="noploop"
9	16
10	17	if ! perf record -e task-clock:u -o - ${prog} \| perf report -i - --task \| grep ${task}; then
11	18	echo "cannot find the test file in the perf report"

tools/perf/tests/shell/record+probe_libc_inet_pton.sh+4−1

@@ -45,7 +45,10 @@ trace_libc_inet_pton_backtrace() { · 44 unmodified lines
45	45	;;
46	46	ppc64\|ppc64le)
47	47	eventattr='max-stack=4'
48		− echo "gaih_inet.*\+0x[[:xdigit:]]+[[:space:]]$$libc$$" >> $expected
	48	+ # Add gaih_inet to expected backtrace only if it is part of libc.
	49	+ if nm $libc \| grep -F -q gaih_inet.; then
	50	+ echo "gaih_inet.*\+0x[[:xdigit:]]+[[:space:]]$$libc$$" >> $expected
	51	+ fi
49	52	echo "getaddrinfo\+0x[[:xdigit:]]+[[:space:]]$$libc$$" >> $expected
50	53	echo ".(\+0x[[:xdigit:]]+\|\[unknown\])[[:space:]]$./bin/ping.*$$" >> $expected
51	54	;;

tools/perf/tests/shell/record.sh+37−1

@@ -8,10 +8,19 @@ shelldir=$(dirname "$0") · 7 unmodified lines
8	8	# shellcheck source=lib/waiting.sh
9	9	. "${shelldir}"/lib/waiting.sh
10	10
	11	+# shellcheck source=lib/perf_has_symbol.sh
	12	+. "${shelldir}"/lib/perf_has_symbol.sh
	13	+
	14	+testsym="test_loop"
	15	+
	16	+skip_test_missing_symbol ${testsym}
	17	+
11	18	err=0
12	19	perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
13	20	testprog="perf test -w thloop"
14		−testsym="test_loop"
	21	+cpu_pmu_dir="/sys/bus/event_source/devices/cpu*"
	22	+br_cntr_file="/caps/branch_counter_nr"
	23	+br_cntr_output="branch stack counters"
15	24
16	25	cleanup() {
17	26	rm -rf "${perfdata}"
@@ -155,10 +164,37 @@ test_workload() { · 137 unmodified lines
155	164	echo "Basic target workload test [Success]"
156	165	}
157	166
	167	+test_branch_counter() {
	168	+ echo "Basic branch counter test"
	169	+ # Check if the branch counter feature is supported
	170	+ for dir in $cpu_pmu_dir
	171	+ do
	172	+ if [ ! -e "$dir$br_cntr_file" ]
	173	+ then
	174	+ echo "branch counter feature not supported on all core PMUs ($dir) [Skipped]"
	175	+ return
	176	+ fi
	177	+ done
	178	+ if ! perf record -o "${perfdata}" -j any,counter ${testprog} 2> /dev/null
	179	+ then
	180	+ echo "Basic branch counter test [Failed record]"
	181	+ err=1
	182	+ return
	183	+ fi
	184	+ if ! perf report -i "${perfdata}" -D -q \| grep -q "$br_cntr_output"
	185	+ then
	186	+ echo "Basic branch record test [Failed missing output]"
	187	+ err=1
	188	+ return
	189	+ fi
	190	+ echo "Basic branch counter test [Success]"
	191	+}
	192	+
158	193	test_per_thread
159	194	test_register_capture
160	195	test_system_wide
161	196	test_workload
	197	+test_branch_counter
162	198
163	199	cleanup
164	200	exit $err

tools/perf/tests/shell/record_offcpu.sh+2−2

@@ -77,9 +77,9 @@ test_offcpu_child() { · 76 unmodified lines
77	77	err=1
78	78	return
79	79	fi
80		− # each process waits for read and write, so it should be more than 800 events
	80	+ # each process waits at least for poll, so it should be more than 400 events
81	81	if ! perf report -i ${perfdata} -s comm -q -n -t ';' --percent-limit=90 \| \
82		− awk -F ";" '{ if (NF > 3 && int($3) < 800) exit 1; }'
	82	+ awk -F ";" '{ if (NF > 3 && int($3) < 400) exit 1; }'
83	83	then
84	84	echo "Child task off-cpu test [Failed invalid output]"
85	85	err=1

tools/perf/tests/shell/script.sh+66−0

@@ -0,0 +1,66 @@
	1	+#!/bin/sh
	2	+# perf script tests
	3	+# SPDX-License-Identifier: GPL-2.0
	4	+
	5	+set -e
	6	+
	7	+temp_dir=$(mktemp -d /tmp/perf-test-script.XXXXXXXXXX)
	8	+
	9	+perfdatafile="${temp_dir}/perf.data"
	10	+db_test="${temp_dir}/db_test.py"
	11	+
	12	+err=0
	13	+
	14	+cleanup()
	15	+{
	16	+ trap - EXIT TERM INT
	17	+ sane=$(echo "${temp_dir}" \| cut -b 1-21)
	18	+ if [ "${sane}" = "/tmp/perf-test-script" ] ; then
	19	+ echo "--- Cleaning up ---"
	20	+ rm -f "${temp_dir}/"*
	21	+ rmdir "${temp_dir}"
	22	+ fi
	23	+}
	24	+
	25	+trap_cleanup()
	26	+{
	27	+ cleanup
	28	+ exit 1
	29	+}
	30	+
	31	+trap trap_cleanup EXIT TERM INT
	32	+
	33	+
	34	+test_db()
	35	+{
	36	+ echo "DB test"
	37	+
	38	+ # Check if python script is supported
	39	+ libpython=$(perf version --build-options \| grep python \| grep -cv OFF)
	40	+ if [ "${libpython}" != "1" ] ; then
	41	+ echo "SKIP: python scripting is not supported"
	42	+ err=2
	43	+ return
	44	+ fi
	45	+
	46	+ cat << "_end_of_file_" > "${db_test}"
	47	+perf_db_export_mode = True
	48	+perf_db_export_calls = False
	49	+perf_db_export_callchains = True
	50	+
	51	+def sample_table(*args):
	52	+ print(f'sample_table({args})')
	53	+
	54	+def call_path_table(*args):
	55	+ print(f'call_path_table({args}')
	56	+_end_of_file_
	57	+ perf record -g -o "${perfdatafile}" true
	58	+ perf script -i "${perfdatafile}" -s "${db_test}"
	59	+ echo "DB test [Success]"
	60	+}
	61	+
	62	+test_db
	63	+
	64	+cleanup
	65	+
	66	+exit $err

tools/perf/tests/shell/stat+json_output.sh+3−13

@@ -8,20 +8,10 @@ set -e · 7 unmodified lines
8	8
9	9	skip_test=0
10	10
	11	+shelldir=$(dirname "$0")
	12	+# shellcheck source=lib/setup_python.sh
	13	+. "${shelldir}"/lib/setup_python.sh
11	14	pythonchecker=$(dirname $0)/lib/perf_json_output_lint.py
12		−if [ "x$PYTHON" == "x" ]
13		−then
14		− if which python3 > /dev/null
15		− then
16		− PYTHON=python3
17		− elif which python > /dev/null
18		− then
19		− PYTHON=python
20		− else
21		− echo Skipping test, python not detected please set environment variable PYTHON.
22		− exit 2
23		− fi
24		−fi
25	15
26	16	stat_output=$(mktemp /tmp/__perf_test.stat_output.json.XXXXX)
27	17

tools/perf/tests/shell/stat_all_pmu.sh+1−1

@@ -4,7 +4,7 @@ · 3 unmodified lines
4	4
5	5	set -e
6	6
7		−# Test all PMU events; however exclude parametrized ones (name contains '?')
	7	+# Test all PMU events; however exclude parameterized ones (name contains '?')
8	8	for p in $(perf list --raw-dump pmu \| sed 's/[[:graph:]]\+?[[:graph:]]\+[[:space:]]//g'); do
9	9	echo "Testing $p"
10	10	result=$(perf stat -e "$p" true 2>&1)

tools/perf/tests/shell/stat_metrics_values.sh+4−10

@@ -1,16 +1,10 @@
1	1	#!/bin/bash
2	2	# perf metrics value validation
3	3	# SPDX-License-Identifier: GPL-2.0
4		−if [ "x$PYTHON" == "x" ]
5		−then
6		− if which python3 > /dev/null
7		− then
8		− PYTHON=python3
9		− else
10		− echo Skipping test, python3 not detected please set environment variable PYTHON.
11		− exit 2
12		− fi
13		−fi
	4	+
	5	+shelldir=$(dirname "$0")
	6	+# shellcheck source=lib/setup_python.sh
	7	+. "${shelldir}"/lib/setup_python.sh
14	8
15	9	grep -q GenuineIntel /proc/cpuinfo \|\| { echo Skipping non-Intel; exit 2; }
16	10

tools/perf/tests/shell/test_arm_callgraph_fp.sh+6−0

@@ -2,8 +2,14 @@ · 1 unmodified line
2	2	# Check Arm64 callgraphs are complete in fp mode
3	3	# SPDX-License-Identifier: GPL-2.0
4	4
	5	+shelldir=$(dirname "$0")
	6	+# shellcheck source=lib/perf_has_symbol.sh
	7	+. "${shelldir}"/lib/perf_has_symbol.sh
	8	+
5	9	lscpu \| grep -q "aarch64" \|\| exit 2
6	10
	11	+skip_test_missing_symbol leafloop
	12	+
7	13	PERF_DATA=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
8	14	TEST_PROGRAM="perf test -w leafloop"
9	15

tools/perf/tests/shell/test_brstack.sh+6−0

@@ -4,6 +4,10 @@ · 3 unmodified lines
4	4	# SPDX-License-Identifier: GPL-2.0
5	5	# German Gomez <german.gomez@arm.com>, 2022
6	6
	7	+shelldir=$(dirname "$0")
	8	+# shellcheck source=lib/perf_has_symbol.sh
	9	+. "${shelldir}"/lib/perf_has_symbol.sh
	10	+
7	11	# skip the test if the hardware doesn't support branch stack sampling
8	12	# and if the architecture doesn't support filter types: any,save_type,u
9	13	if ! perf record -o- --no-buildid --branch-filter any,save_type,u -- true > /dev/null 2>&1 ; then
@@ -11,6 +15,8 @@ if ! perf record -o- --no-buildid --branch-filter any,save_type,u -- true > /dev · 1 unmodified line
11	15	exit 2
12	16	fi
13	17
	18	+skip_test_missing_symbol brstack_bench
	19	+
14	20	TMPDIR=$(mktemp -d /tmp/__perf_test.program.XXXXX)
15	21	TESTPROG="perf test -w brstack"
16	22

tools/perf/tests/shell/test_data_symbol.sh+14−2

@@ -4,6 +4,13 @@ · 3 unmodified lines
4	4	# SPDX-License-Identifier: GPL-2.0
5	5	# Leo Yan <leo.yan@linaro.org>, 2022
6	6
	7	+shelldir=$(dirname "$0")
	8	+# shellcheck source=lib/waiting.sh
	9	+. "${shelldir}"/lib/waiting.sh
	10	+
	11	+# shellcheck source=lib/perf_has_symbol.sh
	12	+. "${shelldir}"/lib/perf_has_symbol.sh
	13	+
7	14	skip_if_no_mem_event() {
8	15	perf mem record -e list 2>&1 \| grep -E -q 'available' && return 0
9	16	return 2
@@ -11,8 +18,11 @@ skip_if_no_mem_event() { · 1 unmodified line
11	18
12	19	skip_if_no_mem_event \|\| exit 2
13	20
	21	+skip_test_missing_symbol buf1
	22	+
14	23	TEST_PROGRAM="perf test -w datasym"
15	24	PERF_DATA=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
	25	+ERR_FILE=$(mktemp /tmp/__perf_test.stderr.XXXXX)
16	26
17	27	check_result() {
18	28	# The memory report format is as below:
@@ -50,13 +60,15 @@ echo "Recording workload..." · 31 unmodified lines
50	60	# specific CPU and test in per-CPU mode.
51	61	is_amd=$(grep -E -c 'vendor_id.*AuthenticAMD' /proc/cpuinfo)
52	62	if (($is_amd >= 1)); then
53		− perf mem record -o ${PERF_DATA} -C 0 -- taskset -c 0 $TEST_PROGRAM &
	63	+ perf mem record -vvv -o ${PERF_DATA} -C 0 -- taskset -c 0 $TEST_PROGRAM 2>"${ERR_FILE}" &
54	64	else
55		− perf mem record --all-user -o ${PERF_DATA} -- $TEST_PROGRAM &
	65	+ perf mem record -vvv --all-user -o ${PERF_DATA} -- $TEST_PROGRAM 2>"${ERR_FILE}" &
56	66	fi
57	67
58	68	PERFPID=$!
59	69
	70	+wait_for_perf_to_start ${PERFPID} "${ERR_FILE}"
	71	+
60	72	sleep 1
61	73
62	74	kill $PERFPID

tools/perf/tests/shell/test_perf_data_converter_json.sh+3−10

@@ -6,16 +6,9 @@ set -e · 5 unmodified lines
6	6
7	7	err=0
8	8
9		−if [ "$PYTHON" = "" ] ; then
10		− if which python3 > /dev/null ; then
11		− PYTHON=python3
12		− elif which python > /dev/null ; then
13		− PYTHON=python
14		− else
15		− echo Skipping test, python not detected please set environment variable PYTHON.
16		− exit 2
17		− fi
18		−fi
	9	+shelldir=$(dirname "$0")
	10	+# shellcheck source=lib/setup_python.sh
	11	+. "${shelldir}"/lib/setup_python.sh
19	12
20	13	perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
21	14	result=$(mktemp /tmp/__perf_test.output.json.XXXXX)

tools/perf/tests/sigtrap.c+84−22

@@ -57,36 +57,79 @@ static struct perf_event_attr make_event_attr(void) · 56 unmodified lines
57	57	#ifdef HAVE_BPF_SKEL
58	58	#include <bpf/btf.h>
59	59
60		−static bool attr_has_sigtrap(void)
	60	+static struct btf *btf;
	61	+
	62	+static bool btf__available(void)
61	63	{
62		− bool ret = false;
63		− struct btf *btf;
64		− const struct btf_type *t;
	64	+ if (btf == NULL)
	65	+ btf = btf__load_vmlinux_btf();
	66	+
	67	+ return btf != NULL;
	68	+}
	69	+
	70	+static void btf__exit(void)
	71	+{
	72	+ btf__free(btf);
	73	+ btf = NULL;
	74	+}
	75	+
	76	+static const struct btf_member __btf_type__find_member_by_name(int type_id, const char member_name)
	77	+{
	78	+ const struct btf_type *t = btf__type_by_id(btf, type_id);
65	79	const struct btf_member *m;
66		− const char *name;
67		− int i, id;
	80	+ int i;
	81	+
	82	+ for (i = 0, m = btf_members(t); i < btf_vlen(t); i++, m++) {
	83	+ const char *current_member_name = btf__name_by_offset(btf, m->name_off);
	84	+ if (!strcmp(current_member_name, member_name))
	85	+ return m;
	86	+ }
68	87
69		− btf = btf__load_vmlinux_btf();
70		− if (btf == NULL) {
	88	+ return NULL;
	89	+}
	90	+
	91	+static bool attr_has_sigtrap(void)
	92	+{
	93	+ int id;
	94	+
	95	+ if (!btf__available()) {
71	96	/* should be an old kernel */
72	97	return false;
73	98	}
74	99
75	100	id = btf__find_by_name_kind(btf, "perf_event_attr", BTF_KIND_STRUCT);
76	101	if (id < 0)
77		− goto out;
	102	+ return false;
78	103
79		− t = btf__type_by_id(btf, id);
80		− for (i = 0, m = btf_members(t); i < btf_vlen(t); i++, m++) {
81		− name = btf__name_by_offset(btf, m->name_off);
82		− if (!strcmp(name, "sigtrap")) {
83		− ret = true;
84		− break;
85		− }
86		− }
87		−out:
88		− btf__free(btf);
89		− return ret;
	104	+ return __btf_type__find_member_by_name(id, "sigtrap") != NULL;
	105	+}
	106	+
	107	+static bool kernel_with_sleepable_spinlocks(void)
	108	+{
	109	+ const struct btf_member *member;
	110	+ const struct btf_type *type;
	111	+ const char *type_name;
	112	+ int id;
	113	+
	114	+ if (!btf__available())
	115	+ return false;
	116	+
	117	+ id = btf__find_by_name_kind(btf, "spinlock", BTF_KIND_STRUCT);
	118	+ if (id < 0)
	119	+ return false;
	120	+
	121	+ // Only RT has a "lock" member for "struct spinlock"
	122	+ member = __btf_type__find_member_by_name(id, "lock");
	123	+ if (member == NULL)
	124	+ return false;
	125	+
	126	+ // But check its type as well
	127	+ type = btf__type_by_id(btf, member->type);
	128	+ if (!type \|\| !btf_is_struct(type))
	129	+ return false;
	130	+
	131	+ type_name = btf__name_by_offset(btf, type->name_off);
	132	+ return type_name && !strcmp(type_name, "rt_mutex_base");
90	133	}
91	134	#else /* !HAVE_BPF_SKEL */
92	135	static bool attr_has_sigtrap(void)
@@ -109,6 +152,15 @@ static bool attr_has_sigtrap(void) · 16 unmodified lines
109	152
110	153	return ret;
111	154	}
	155	+
	156	+static bool kernel_with_sleepable_spinlocks(void)
	157	+{
	158	+ return false;
	159	+}
	160	+
	161	+static void btf__exit(void)
	162	+{
	163	+}
112	164	#endif /* HAVE_BPF_SKEL */
113	165
114	166	static void
@@ -147,7 +199,7 @@ static int run_test_threads(pthread_t threads, pthread_barrier_t barrier) · 32 unmodified lines
147	199
148	200	static int run_stress_test(int fd, pthread_t threads, pthread_barrier_t barrier)
149	201	{
150		− int ret;
	202	+ int ret, expected_sigtraps;
151	203
152	204	ctx.iterate_on = 3000;
153	205
@@ -156,7 +208,16 @@ static int run_stress_test(int fd, pthread_t threads, pthread_barrier_t barrie · 2 unmodified lines
156	208	ret = run_test_threads(threads, barrier);
157	209	TEST_ASSERT_EQUAL("disable failed", ioctl(fd, PERF_EVENT_IOC_DISABLE, 0), 0);
158	210
159		− TEST_ASSERT_EQUAL("unexpected sigtraps", ctx.signal_count, NUM_THREADS * ctx.iterate_on);
	211	+ expected_sigtraps = NUM_THREADS * ctx.iterate_on;
	212	+
	213	+ if (ctx.signal_count < expected_sigtraps && kernel_with_sleepable_spinlocks()) {
	214	+ pr_debug("Expected %d sigtraps, got %d, running on a kernel with sleepable spinlocks.\n",
	215	+ expected_sigtraps, ctx.signal_count);
	216	+ pr_debug("See https://lore.kernel.org/all/e368f2c848d77fbc8d259f44e2055fe469c219cf.camel@gmx.de/\n");
	217	+ return TEST_SKIP;
	218	+ } else
	219	+ TEST_ASSERT_EQUAL("unexpected sigtraps", ctx.signal_count, expected_sigtraps);
	220	+
160	221	TEST_ASSERT_EQUAL("missing signals or incorrectly delivered", ctx.tids_want_signal, 0);
161	222	TEST_ASSERT_VAL("unexpected si_addr", ctx.first_siginfo.si_addr == &ctx.iterate_on);
162	223	#if 0 /* FIXME: enable when libc's signal.h has si_perf_{type,data} */
@@ -221,6 +282,7 @@ out_restore_sigaction: · 58 unmodified lines
221	282	sigaction(SIGTRAP, &oldact, NULL);
222	283	out:
223	284	pthread_barrier_destroy(&barrier);
	285	+ btf__exit();
224	286	return ret;
225	287	}
226	288

tools/perf/tests/sw-clock.c+1−1

@@ -62,7 +62,7 @@ static int __test__sw_clock_freq(enum perf_sw_ids clock_id) · 61 unmodified lines
62	62	}
63	63	evlist__add(evlist, evsel);
64	64
65		− cpus = perf_cpu_map__dummy_new();
	65	+ cpus = perf_cpu_map__new_any_cpu();
66	66	threads = thread_map__new_by_tid(getpid());
67	67	if (!cpus \|\| !threads) {
68	68	err = -ENOMEM;

tools/perf/tests/switch-tracking.c+1−1

@@ -351,7 +351,7 @@ static int test__switch_tracking(struct test_suite *test __maybe_unused, int sub · 350 unmodified lines
351	351	goto out_err;
352	352	}
353	353
354		− cpus = perf_cpu_map__new(NULL);
	354	+ cpus = perf_cpu_map__new_online_cpus();
355	355	if (!cpus) {
356	356	pr_debug("perf_cpu_map__new failed!\n");
357	357	goto out_err;

tools/perf/tests/task-exit.c+1−1

@@ -70,7 +70,7 @@ static int test__task_exit(struct test_suite *test __maybe_unused, int subtest _ · 69 unmodified lines
70	70	* evlist__prepare_workload we'll fill in the only thread
71	71	* we're monitoring, the one forked there.
72	72	*/
73		− cpus = perf_cpu_map__dummy_new();
	73	+ cpus = perf_cpu_map__new_any_cpu();
74	74	threads = thread_map__new_by_tid(-1);
75	75	if (!cpus \|\| !threads) {
76	76	err = -ENOMEM;

tools/perf/tests/tests.h+1−0

@@ -207,5 +207,6 @@ DECLARE_WORKLOAD(brstack); · 206 unmodified lines
207	207	DECLARE_WORKLOAD(datasym);
208	208
209	209	extern const char *dso_to_test;
	210	+extern const char *test_objdump_path;
210	211
211	212	#endif /* TESTS_H */

tools/perf/tests/topology.c+1−1

@@ -215,7 +215,7 @@ static int test__session_topology(struct test_suite *test __maybe_unused, int su · 214 unmodified lines
215	215	if (session_write_header(path))
216	216	goto free_path;
217	217
218		− map = perf_cpu_map__new(NULL);
	218	+ map = perf_cpu_map__new_online_cpus();
219	219	if (map == NULL) {
220	220	pr_debug("failed to get system cpumap\n");
221	221	goto free_path;

tools/perf/tests/vmlinux-kallsyms.c+98−83

@@ -112,18 +112,92 @@ static bool is_ignored_symbol(const char *name, char type) · 111 unmodified lines
112	112	return false;
113	113	}
114	114
	115	+struct test__vmlinux_matches_kallsyms_cb_args {
	116	+ struct machine kallsyms;
	117	+ struct map *vmlinux_map;
	118	+ bool header_printed;
	119	+};
	120	+
	121	+static int test__vmlinux_matches_kallsyms_cb1(struct map map, void data)
	122	+{
	123	+ struct test__vmlinux_matches_kallsyms_cb_args *args = data;
	124	+ struct dso *dso = map__dso(map);
	125	+ /*
	126	+ * If it is the kernel, kallsyms is always "[kernel.kallsyms]", while
	127	+ * the kernel will have the path for the vmlinux file being used, so use
	128	+ * the short name, less descriptive but the same ("[kernel]" in both
	129	+ * cases.
	130	+ */
	131	+ struct map *pair = maps__find_by_name(args->kallsyms.kmaps,
	132	+ (dso->kernel ? dso->short_name : dso->name));
	133	+
	134	+ if (pair)
	135	+ map__set_priv(pair, 1);
	136	+ else {
	137	+ if (!args->header_printed) {
	138	+ pr_info("WARN: Maps only in vmlinux:\n");
	139	+ args->header_printed = true;
	140	+ }
	141	+ map__fprintf(map, stderr);
	142	+ }
	143	+ return 0;
	144	+}
	145	+
	146	+static int test__vmlinux_matches_kallsyms_cb2(struct map map, void data)
	147	+{
	148	+ struct test__vmlinux_matches_kallsyms_cb_args *args = data;
	149	+ struct map *pair;
	150	+ u64 mem_start = map__unmap_ip(args->vmlinux_map, map__start(map));
	151	+ u64 mem_end = map__unmap_ip(args->vmlinux_map, map__end(map));
	152	+
	153	+ pair = maps__find(args->kallsyms.kmaps, mem_start);
	154	+ if (pair == NULL \|\| map__priv(pair))
	155	+ return 0;
	156	+
	157	+ if (map__start(pair) == mem_start) {
	158	+ struct dso *dso = map__dso(map);
	159	+
	160	+ if (!args->header_printed) {
	161	+ pr_info("WARN: Maps in vmlinux with a different name in kallsyms:\n");
	162	+ args->header_printed = true;
	163	+ }
	164	+
	165	+ pr_info("WARN: %" PRIx64 "-%" PRIx64 " %" PRIx64 " %s in kallsyms as",
	166	+ map__start(map), map__end(map), map__pgoff(map), dso->name);
	167	+ if (mem_end != map__end(pair))
	168	+ pr_info(":\nWARN: *%" PRIx64 "-%" PRIx64 " %" PRIx64,
	169	+ map__start(pair), map__end(pair), map__pgoff(pair));
	170	+ pr_info(" %s\n", dso->name);
	171	+ map__set_priv(pair, 1);
	172	+ }
	173	+ return 0;
	174	+}
	175	+
	176	+static int test__vmlinux_matches_kallsyms_cb3(struct map map, void data)
	177	+{
	178	+ struct test__vmlinux_matches_kallsyms_cb_args *args = data;
	179	+
	180	+ if (!map__priv(map)) {
	181	+ if (!args->header_printed) {
	182	+ pr_info("WARN: Maps only in kallsyms:\n");
	183	+ args->header_printed = true;
	184	+ }
	185	+ map__fprintf(map, stderr);
	186	+ }
	187	+ return 0;
	188	+}
	189	+
115	190	static int test__vmlinux_matches_kallsyms(struct test_suite *test __maybe_unused,
116	191	int subtest __maybe_unused)
117	192	{
118	193	int err = TEST_FAIL;
119	194	struct rb_node *nd;
120	195	struct symbol *sym;
121		− struct map kallsyms_map, vmlinux_map;
122		− struct map_rb_node *rb_node;
123		− struct machine kallsyms, vmlinux;
	196	+ struct map *kallsyms_map;
	197	+ struct machine vmlinux;
124	198	struct maps *maps;
125	199	u64 mem_start, mem_end;
126		− bool header_printed;
	200	+ struct test__vmlinux_matches_kallsyms_cb_args args;
127	201
128	202	/*
129	203	* Step 1:
@@ -131,7 +205,7 @@ static int test__vmlinux_matches_kallsyms(struct test_suite *test __maybe_unused · 1 unmodified line
131	205	* Init the machines that will hold kernel, modules obtained from
132	206	* both vmlinux + .ko files and from /proc/kallsyms split by modules.
133	207	*/
134		− machine__init(&kallsyms, "", HOST_KERNEL_ID);
	208	+ machine__init(&args.kallsyms, "", HOST_KERNEL_ID);
135	209	machine__init(&vmlinux, "", HOST_KERNEL_ID);
136	210
137	211	maps = machine__kernel_maps(&vmlinux);
@@ -143,7 +217,7 @@ static int test__vmlinux_matches_kallsyms(struct test_suite *test __maybe_unused · 5 unmodified lines
143	217	* load /proc/kallsyms. Also create the modules maps from /proc/modules
144	218	* and find the .ko files that match them in /lib/modules/`uname -r`/.
145	219	*/
146		− if (machine__create_kernel_maps(&kallsyms) < 0) {
	220	+ if (machine__create_kernel_maps(&args.kallsyms) < 0) {
147	221	pr_debug("machine__create_kernel_maps failed");
148	222	err = TEST_SKIP;
149	223	goto out;
@@ -160,7 +234,7 @@ static int test__vmlinux_matches_kallsyms(struct test_suite *test __maybe_unused · 10 unmodified lines
160	234	* be compacted against the list of modules found in the "vmlinux"
161	235	* code and with the one got from /proc/modules from the "kallsyms" code.
162	236	*/
163		− if (machine__load_kallsyms(&kallsyms, "/proc/kallsyms") <= 0) {
	237	+ if (machine__load_kallsyms(&args.kallsyms, "/proc/kallsyms") <= 0) {
164	238	pr_debug("machine__load_kallsyms failed");
165	239	err = TEST_SKIP;
166	240	goto out;
@@ -174,7 +248,7 @@ static int test__vmlinux_matches_kallsyms(struct test_suite *test __maybe_unused · 7 unmodified lines
174	248	* to see if the running kernel was relocated by checking if it has the
175	249	* same value in the vmlinux file we load.
176	250	*/
177		− kallsyms_map = machine__kernel_map(&kallsyms);
	251	+ kallsyms_map = machine__kernel_map(&args.kallsyms);
178	252
179	253	/*
180	254	* Step 5:
@@ -186,7 +260,7 @@ static int test__vmlinux_matches_kallsyms(struct test_suite *test __maybe_unused · 5 unmodified lines
186	260	goto out;
187	261	}
188	262
189		− vmlinux_map = machine__kernel_map(&vmlinux);
	263	+ args.vmlinux_map = machine__kernel_map(&vmlinux);
190	264
191	265	/*
192	266	* Step 6:
@@ -213,7 +287,7 @@ static int test__vmlinux_matches_kallsyms(struct test_suite *test __maybe_unused · 20 unmodified lines
213	287	* in the kallsyms dso. For the ones that are in both, check its names and
214	288	* end addresses too.
215	289	*/
216		− map__for_each_symbol(vmlinux_map, sym, nd) {
	290	+ map__for_each_symbol(args.vmlinux_map, sym, nd) {
217	291	struct symbol pair, first_pair;
218	292
219	293	sym = rb_entry(nd, struct symbol, rb_node);
@@ -221,10 +295,10 @@ static int test__vmlinux_matches_kallsyms(struct test_suite *test __maybe_unused · 1 unmodified line
221	295	if (sym->start == sym->end)
222	296	continue;
223	297
224		− mem_start = map__unmap_ip(vmlinux_map, sym->start);
225		− mem_end = map__unmap_ip(vmlinux_map, sym->end);
	298	+ mem_start = map__unmap_ip(args.vmlinux_map, sym->start);
	299	+ mem_end = map__unmap_ip(args.vmlinux_map, sym->end);
226	300
227		− first_pair = machine__find_kernel_symbol(&kallsyms, mem_start, NULL);
	301	+ first_pair = machine__find_kernel_symbol(&args.kallsyms, mem_start, NULL);
228	302	pair = first_pair;
229	303
230	304	if (pair && UM(pair->start) == mem_start) {
@@ -253,7 +327,8 @@ next_pair: · 22 unmodified lines
253	327	*/
254	328	continue;
255	329	} else {
256		− pair = machine__find_kernel_symbol_by_name(&kallsyms, sym->name, NULL);
	330	+ pair = machine__find_kernel_symbol_by_name(&args.kallsyms,
	331	+ sym->name, NULL);
257	332	if (pair) {
258	333	if (UM(pair->start) == mem_start)
259	334	goto next_pair;
@@ -267,7 +342,7 @@ next_pair: · 7 unmodified lines
267	342
268	343	continue;
269	344	}
270		− } else if (mem_start == map__end(kallsyms.vmlinux_map)) {
	345	+ } else if (mem_start == map__end(args.kallsyms.vmlinux_map)) {
271	346	/*
272	347	* Ignore aliases to _etext, i.e. to the end of the kernel text area,
273	348	* such as __indirect_thunk_end.
@@ -289,78 +364,18 @@ next_pair: · 15 unmodified lines
289	364	if (verbose <= 0)
290	365	goto out;
291	366
292		− header_printed = false;
293		−
294		− maps__for_each_entry(maps, rb_node) {
295		− struct map *map = rb_node->map;
296		− struct dso *dso = map__dso(map);
297		− /*
298		− * If it is the kernel, kallsyms is always "[kernel.kallsyms]", while
299		− * the kernel will have the path for the vmlinux file being used,
300		− * so use the short name, less descriptive but the same ("[kernel]" in
301		− * both cases.
302		− */
303		− struct map *pair = maps__find_by_name(kallsyms.kmaps, (dso->kernel ?
304		− dso->short_name :
305		− dso->name));
306		− if (pair) {
307		− map__set_priv(pair, 1);
308		− } else {
309		− if (!header_printed) {
310		− pr_info("WARN: Maps only in vmlinux:\n");
311		− header_printed = true;
312		− }
313		− map__fprintf(map, stderr);
314		− }
315		− }
316		−
317		− header_printed = false;
318		−
319		− maps__for_each_entry(maps, rb_node) {
320		− struct map pair, map = rb_node->map;
321		−
322		− mem_start = map__unmap_ip(vmlinux_map, map__start(map));
323		− mem_end = map__unmap_ip(vmlinux_map, map__end(map));
	367	+ args.header_printed = false;
	368	+ maps__for_each_map(maps, test__vmlinux_matches_kallsyms_cb1, &args);
324	369
325		− pair = maps__find(kallsyms.kmaps, mem_start);
326		− if (pair == NULL \|\| map__priv(pair))
327		− continue;
328		−
329		− if (map__start(pair) == mem_start) {
330		− struct dso *dso = map__dso(map);
331		−
332		− if (!header_printed) {
333		− pr_info("WARN: Maps in vmlinux with a different name in kallsyms:\n");
334		− header_printed = true;
335		− }
336		−
337		− pr_info("WARN: %" PRIx64 "-%" PRIx64 " %" PRIx64 " %s in kallsyms as",
338		− map__start(map), map__end(map), map__pgoff(map), dso->name);
339		− if (mem_end != map__end(pair))
340		− pr_info(":\nWARN: *%" PRIx64 "-%" PRIx64 " %" PRIx64,
341		− map__start(pair), map__end(pair), map__pgoff(pair));
342		− pr_info(" %s\n", dso->name);
343		− map__set_priv(pair, 1);
344		− }
345		− }
346		−
347		− header_printed = false;
348		−
349		− maps = machine__kernel_maps(&kallsyms);
	370	+ args.header_printed = false;
	371	+ maps__for_each_map(maps, test__vmlinux_matches_kallsyms_cb2, &args);
350	372
351		− maps__for_each_entry(maps, rb_node) {
352		− struct map *map = rb_node->map;
	373	+ args.header_printed = false;
	374	+ maps = machine__kernel_maps(&args.kallsyms);
	375	+ maps__for_each_map(maps, test__vmlinux_matches_kallsyms_cb3, &args);
353	376
354		− if (!map__priv(map)) {
355		− if (!header_printed) {
356		− pr_info("WARN: Maps only in kallsyms:\n");
357		− header_printed = true;
358		− }
359		− map__fprintf(map, stderr);
360		− }
361		− }
362	377	out:
363		− machine__exit(&kallsyms);
	378	+ machine__exit(&args.kallsyms);
364	379	machine__exit(&vmlinux);
365	380	return err;
366	381	}

tools/perf/tests/workloads/thloop.c+1−3

@@ -7,7 +7,6 @@ · 6 unmodified lines
7	7	#include "../tests.h"
8	8
9	9	static volatile sig_atomic_t done;
10		−static volatile unsigned count;
11	10
12	11	/* We want to check this symbol in perf report */
13	12	noinline void test_loop(void);
@@ -19,8 +18,7 @@ static void sighandler(int sig __maybe_unused) · 5 unmodified lines
19	18
20	19	noinline void test_loop(void)
21	20	{
22		− while (!done)
23		− __atomic_fetch_add(&count, 1, __ATOMIC_RELAXED);
	21	+ while (!done);
24	22	}
25	23
26	24	static void thfunc(void arg)

tools/perf/trace/beauty/arch_errno_names.sh+6−4

@@ -57,13 +57,13 @@ create_arch_errno_table_func() · 56 unmodified lines
57	57	archlist="$1"
58	58	default="$2"
59	59
60		− printf 'const char arch_syscalls__strerrno(const char arch, int err)\n'
	60	+ printf 'arch_syscalls__strerrno_t arch_syscalls__strerrno_function(const char arch)\n'
61	61	printf '{\n'
62	62	for arch in $archlist; do
63	63	printf '\tif (!strcmp(arch, "%s"))\n' $(arch_string "$arch")
64		− printf '\t\treturn errno_to_name__%s(err);\n' $(arch_string "$arch")
	64	+ printf '\t\treturn errno_to_name__%s;\n' $(arch_string "$arch")
65	65	done
66		− printf '\treturn errno_to_name__%s(err);\n' $(arch_string "$default")
	66	+ printf '\treturn errno_to_name__%s;\n' $(arch_string "$default")
67	67	printf '}\n'
68	68	}
69	69
@@ -76,7 +76,9 @@ EoHEADER · 6 unmodified lines
76	76
77	77	# Create list of architectures that have a specific errno.h.
78	78	archlist=""
79		−for arch in $(find $toolsdir/arch -maxdepth 1 -mindepth 1 -type d -printf "%f\n" \| sort -r); do
	79	+for f in $toolsdir/arch/*/include/uapi/asm/errno.h; do
	80	+ d=${f%/include/uapi/asm/errno.h}
	81	+ arch="${d##*/}"
80	82	test -f $toolsdir/arch/$arch/include/uapi/asm/errno.h && archlist="$archlist $arch"
81	83	done
82	84

tools/perf/trace/beauty/beauty.h+0−2

@@ -251,6 +251,4 @@ size_t open__scnprintf_flags(unsigned long flags, char *bf, size_t size, bool sh · 250 unmodified lines
251	251	void syscall_arg__set_ret_scnprintf(struct syscall_arg *arg,
252	252	size_t (ret_scnprintf)(char bf, size_t size, struct syscall_arg *arg));
253	253
254		−const char arch_syscalls__strerrno(const char arch, int err);
255		−
256	254	#endif /* _PERF_TRACE_BEAUTY_H */

tools/perf/trace/beauty/prctl_option.sh+2−2

@@ -4,9 +4,9 @@ · 3 unmodified lines
4	4	[ $# -eq 1 ] && header_dir=$1 \|\| header_dir=tools/include/uapi/linux/
5	5
6	6	printf "static const char *prctl_options[] = {\n"
7		−regex='^#define[[:space:]]{1}PR_(\w+)[[:space:]]([[:xdigit:]]+)([[:space:]]\/.*)?$'
	7	+regex='^#define[[:space:]]{1}PR_(\w+)[[:space:]]([[:xdigit:]]+)([[:space:]]/.*)?$'
8	8	grep -E $regex ${header_dir}/prctl.h \| grep -v PR_SET_PTRACER \| \
9		− sed -r "s/$regex/\2 \1/g" \| \
	9	+ sed -E "s%$regex%\2 \1%g" \| \
10	10	sort -n \| xargs printf "\t[%s] = \"%s\",\n"
11	11	printf "};\n"
12	12

tools/perf/trace/beauty/socket.sh+2−2

@@ -18,10 +18,10 @@ grep -E $ipproto_regex ${uapi_header_dir}/in.h \| \ · 17 unmodified lines
18	18	printf "};\n\n"
19	19
20	20	printf "static const char *socket_level[] = {\n"
21		−socket_level_regex='^#define[[:space:]]+SOL_(\w+)[[:space:]]+([[:digit:]]+)([[:space:]]+\/.*)?'
	21	+socket_level_regex='^#define[[:space:]]+SOL_(\w+)[[:space:]]+([[:digit:]]+)([[:space:]]+/.*)?'
22	22
23	23	grep -E $socket_level_regex ${beauty_header_dir}/socket.h \| \
24		− sed -r "s/$socket_level_regex/\2 \1/g" \| \
	24	+ sed -E "s%$socket_level_regex%\2 \1%g" \| \
25	25	sort -n \| xargs printf "\t[%s] = \"%s\",\n"
26	26	printf "};\n\n"
27	27

tools/perf/ui/browsers/annotate.c+48−56

@@ -27,7 +27,6 @@ struct annotate_browser { · 26 unmodified lines
27	27	struct rb_node *curr_hot;
28	28	struct annotation_line *selection;
29	29	struct arch *arch;
30		− struct annotation_options *opts;
31	30	bool searching_backwards;
32	31	char search_bf[128];
33	32	};
@@ -38,11 +37,10 @@ static inline struct annotation browser__annotation(struct ui_browser browser) · 4 unmodified lines
38	37	return symbol__annotation(ms->sym);
39	38	}
40	39
41		−static bool disasm_line__filter(struct ui_browser browser, void entry)
	40	+static bool disasm_line__filter(struct ui_browser browser __maybe_unused, void entry)
42	41	{
43		− struct annotation *notes = browser__annotation(browser);
44	42	struct annotation_line *al = list_entry(entry, struct annotation_line, node);
45		− return annotation_line__filter(al, notes);
	43	+ return annotation_line__filter(al);
46	44	}
47	45
48	46	static int ui_browser__jumps_percent_color(struct ui_browser *browser, int nr, bool current)
@@ -97,7 +95,7 @@ static void annotate_browser__write(struct ui_browser browser, void entry, int · 48 unmodified lines
97	95	struct annotation_write_ops ops = {
98	96	.first_line = row == 0,
99	97	.current_entry = is_current_entry,
100		− .change_color = (!notes->options->hide_src_code &&
	98	+ .change_color = (!annotate_opts.hide_src_code &&
101	99	(!is_current_entry \|\|
102	100	(browser->use_navkeypressed &&
103	101	!browser->navkeypressed))),
@@ -114,7 +112,7 @@ static void annotate_browser__write(struct ui_browser browser, void entry, int · 10 unmodified lines
114	112	if (!browser->navkeypressed)
115	113	ops.width += 1;
116	114
117		− annotation_line__write(al, notes, &ops, ab->opts);
	115	+ annotation_line__write(al, notes, &ops);
118	116
119	117	if (ops.current_entry)
120	118	ab->selection = al;
@@ -128,7 +126,7 @@ static int is_fused(struct annotate_browser ab, struct disasm_line cursor) · 7 unmodified lines
128	126
129	127	while (pos && pos->al.offset == -1) {
130	128	pos = list_prev_entry(pos, al.node);
131		− if (!ab->opts->hide_src_code)
	129	+ if (!annotate_opts.hide_src_code)
132	130	diff++;
133	131	}
134	132
@@ -188,14 +186,14 @@ static void annotate_browser__draw_current_jump(struct ui_browser *browser) · 53 unmodified lines
188	186	* name right after the '<' token and probably treating this like a
189	187	* 'call' instruction.
190	188	*/
191		− target = notes->offsets[cursor->ops.target.offset];
	189	+ target = notes->src->offsets[cursor->ops.target.offset];
192	190	if (target == NULL) {
193	191	ui_helpline__printf("WARN: jump target inconsistency, press 'o', notes->offsets[%#x] = NULL\n",
194	192	cursor->ops.target.offset);
195	193	return;
196	194	}
197	195
198		− if (notes->options->hide_src_code) {
	196	+ if (annotate_opts.hide_src_code) {
199	197	from = cursor->al.idx_asm;
200	198	to = target->idx_asm;
201	199	} else {
@@ -224,7 +222,7 @@ static unsigned int annotate_browser__refresh(struct ui_browser *browser) · 22 unmodified lines
224	222	int ret = ui_browser__list_head_refresh(browser);
225	223	int pcnt_width = annotation__pcnt_width(notes);
226	224
227		− if (notes->options->jump_arrows)
	225	+ if (annotate_opts.jump_arrows)
228	226	annotate_browser__draw_current_jump(browser);
229	227
230	228	ui_browser__set_color(browser, HE_COLORSET_NORMAL);
@@ -258,7 +256,7 @@ static void disasm_rb_tree__insert(struct annotate_browser *browser, · 27 unmodified lines
258	256	parent = *p;
259	257	l = rb_entry(parent, struct annotation_line, rb_node);
260	258
261		− if (disasm__cmp(al, l, browser->opts->percent_type) < 0)
	259	+ if (disasm__cmp(al, l, annotate_opts.percent_type) < 0)
262	260	p = &(*p)->rb_left;
263	261	else
264	262	p = &(*p)->rb_right;
@@ -270,7 +268,6 @@ static void disasm_rb_tree__insert(struct annotate_browser *browser, · 5 unmodified lines
270	268	static void annotate_browser__set_top(struct annotate_browser *browser,
271	269	struct annotation_line *pos, u32 idx)
272	270	{
273		− struct annotation *notes = browser__annotation(&browser->b);
274	271	unsigned back;
275	272
276	273	ui_browser__refresh_dimensions(&browser->b);
@@ -280,7 +277,7 @@ static void annotate_browser__set_top(struct annotate_browser *browser, · 3 unmodified lines
280	277	while (browser->b.top_idx != 0 && back != 0) {
281	278	pos = list_entry(pos->node.prev, struct annotation_line, node);
282	279
283		− if (annotation_line__filter(pos, notes))
	280	+ if (annotation_line__filter(pos))
284	281	continue;
285	282
286	283	--browser->b.top_idx;
@@ -294,11 +291,10 @@ static void annotate_browser__set_top(struct annotate_browser *browser, · 7 unmodified lines
294	291	static void annotate_browser__set_rb_top(struct annotate_browser *browser,
295	292	struct rb_node *nd)
296	293	{
297		− struct annotation *notes = browser__annotation(&browser->b);
298	294	struct annotation_line * pos = rb_entry(nd, struct annotation_line, rb_node);
299	295	u32 idx = pos->idx;
300	296
301		− if (notes->options->hide_src_code)
	297	+ if (annotate_opts.hide_src_code)
302	298	idx = pos->idx_asm;
303	299	annotate_browser__set_top(browser, pos, idx);
304	300	browser->curr_hot = nd;
@@ -331,13 +327,13 @@ static void annotate_browser__calc_percent(struct annotate_browser *browser, · 26 unmodified lines
331	327	double percent;
332	328
333	329	percent = annotation_data__percent(&pos->al.data[i],
334		− browser->opts->percent_type);
	330	+ annotate_opts.percent_type);
335	331
336	332	if (max_percent < percent)
337	333	max_percent = percent;
338	334	}
339	335
340		− if (max_percent < 0.01 && pos->al.ipc == 0) {
	336	+ if (max_percent < 0.01 && (!pos->al.cycles \|\| pos->al.cycles->ipc == 0)) {
341	337	RB_CLEAR_NODE(&pos->al.rb_node);
342	338	continue;
343	339	}
@@ -380,12 +376,12 @@ static bool annotate_browser__toggle_source(struct annotate_browser *browser) · 36 unmodified lines
380	376	browser->b.seek(&browser->b, offset, SEEK_CUR);
381	377	al = list_entry(browser->b.top, struct annotation_line, node);
382	378
383		− if (notes->options->hide_src_code) {
	379	+ if (annotate_opts.hide_src_code) {
384	380	if (al->idx_asm < offset)
385	381	offset = al->idx;
386	382
387		− browser->b.nr_entries = notes->nr_entries;
388		− notes->options->hide_src_code = false;
	383	+ browser->b.nr_entries = notes->src->nr_entries;
	384	+ annotate_opts.hide_src_code = false;
389	385	browser->b.seek(&browser->b, -offset, SEEK_CUR);
390	386	browser->b.top_idx = al->idx - offset;
391	387	browser->b.index = al->idx;
@@ -402,8 +398,8 @@ static bool annotate_browser__toggle_source(struct annotate_browser *browser) · 10 unmodified lines
402	398	if (al->idx_asm < offset)
403	399	offset = al->idx_asm;
404	400
405		− browser->b.nr_entries = notes->nr_asm_entries;
406		− notes->options->hide_src_code = true;
	401	+ browser->b.nr_entries = notes->src->nr_asm_entries;
	402	+ annotate_opts.hide_src_code = true;
407	403	browser->b.seek(&browser->b, -offset, SEEK_CUR);
408	404	browser->b.top_idx = al->idx_asm - offset;
409	405	browser->b.index = al->idx_asm;
@@ -435,7 +431,7 @@ static void ui_browser__init_asm_mode(struct ui_browser *browser) · 25 unmodified lines
435	431	{
436	432	struct annotation *notes = browser__annotation(browser);
437	433	ui_browser__reset_index(browser);
438		− browser->nr_entries = notes->nr_asm_entries;
	434	+ browser->nr_entries = notes->src->nr_asm_entries;
439	435	}
440	436
441	437	static int sym_title(struct symbol sym, struct map map, char *title,
@@ -483,8 +479,8 @@ static bool annotate_browser__callq(struct annotate_browser *browser, · 41 unmodified lines
483	479	target_ms.map = ms->map;
484	480	target_ms.sym = dl->ops.target.sym;
485	481	annotation__unlock(notes);
486		− symbol__tui_annotate(&target_ms, evsel, hbt, browser->opts);
487		− sym_title(ms->sym, ms->map, title, sizeof(title), browser->opts->percent_type);
	482	+ symbol__tui_annotate(&target_ms, evsel, hbt);
	483	+ sym_title(ms->sym, ms->map, title, sizeof(title), annotate_opts.percent_type);
488	484	ui_browser__show_title(&browser->b, title);
489	485	return true;
490	486	}
@@ -500,7 +496,7 @@ struct disasm_line annotate_browser__find_offset(struct annotate_browser brows · 9 unmodified lines
500	496	list_for_each_entry(pos, &notes->src->source, al.node) {
501	497	if (pos->al.offset == offset)
502	498	return pos;
503		− if (!annotation_line__filter(&pos->al, notes))
	499	+ if (!annotation_line__filter(&pos->al))
504	500	++*idx;
505	501	}
506	502
@@ -544,7 +540,7 @@ struct annotation_line annotate_browser__find_string(struct annotate_browser b · 37 unmodified lines
544	540
545	541	*idx = browser->b.index;
546	542	list_for_each_entry_continue(al, &notes->src->source, node) {
547		− if (annotation_line__filter(al, notes))
	543	+ if (annotation_line__filter(al))
548	544	continue;
549	545
550	546	++*idx;
@@ -581,7 +577,7 @@ struct annotation_line *annotate_browser__find_string_reverse(struct annotate_br · 30 unmodified lines
581	577
582	578	*idx = browser->b.index;
583	579	list_for_each_entry_continue_reverse(al, &notes->src->source, node) {
584		− if (annotation_line__filter(al, notes))
	580	+ if (annotation_line__filter(al))
585	581	continue;
586	582
587	583	--*idx;
@@ -659,7 +655,6 @@ bool annotate_browser__continue_search_reverse(struct annotate_browser *browser, · 71 unmodified lines
659	655
660	656	static int annotate_browser__show(struct ui_browser browser, char title, const char *help)
661	657	{
662		− struct annotate_browser *ab = container_of(browser, struct annotate_browser, b);
663	658	struct map_symbol *ms = browser->priv;
664	659	struct symbol *sym = ms->sym;
665	660	char symbol_dso[SYM_TITLE_MAX_SIZE];
@@ -667,7 +662,7 @@ static int annotate_browser__show(struct ui_browser browser, char title, const · 1 unmodified line
667	662	if (ui_browser__show(browser, title, help) < 0)
668	663	return -1;
669	664
670		− sym_title(sym, ms->map, symbol_dso, sizeof(symbol_dso), ab->opts->percent_type);
	665	+ sym_title(sym, ms->map, symbol_dso, sizeof(symbol_dso), annotate_opts.percent_type);
671	666
672	667	ui_browser__gotorc_title(browser, 0, 0);
673	668	ui_browser__set_color(browser, HE_COLORSET_ROOT);
@@ -809,7 +804,7 @@ static int annotate_browser__run(struct annotate_browser *browser, · 135 unmodified lines
809	804	annotate_browser__show(&browser->b, title, help);
810	805	continue;
811	806	case 'k':
812		− notes->options->show_linenr = !notes->options->show_linenr;
	807	+ annotate_opts.show_linenr = !annotate_opts.show_linenr;
813	808	continue;
814	809	case 'l':
815	810	annotate_browser__show_full_location (&browser->b);
@@ -822,18 +817,18 @@ static int annotate_browser__run(struct annotate_browser *browser, · 6 unmodified lines
822	817	ui_helpline__puts(help);
823	818	continue;
824	819	case 'o':
825		− notes->options->use_offset = !notes->options->use_offset;
	820	+ annotate_opts.use_offset = !annotate_opts.use_offset;
826	821	annotation__update_column_widths(notes);
827	822	continue;
828	823	case 'O':
829		− if (++notes->options->offset_level > ANNOTATION__MAX_OFFSET_LEVEL)
830		− notes->options->offset_level = ANNOTATION__MIN_OFFSET_LEVEL;
	824	+ if (++annotate_opts.offset_level > ANNOTATION__MAX_OFFSET_LEVEL)
	825	+ annotate_opts.offset_level = ANNOTATION__MIN_OFFSET_LEVEL;
831	826	continue;
832	827	case 'j':
833		− notes->options->jump_arrows = !notes->options->jump_arrows;
	828	+ annotate_opts.jump_arrows = !annotate_opts.jump_arrows;
834	829	continue;
835	830	case 'J':
836		− notes->options->show_nr_jumps = !notes->options->show_nr_jumps;
	831	+ annotate_opts.show_nr_jumps = !annotate_opts.show_nr_jumps;
837	832	annotation__update_column_widths(notes);
838	833	continue;
839	834	case '/':
@@ -860,7 +855,7 @@ show_help: · 20 unmodified lines
860	855	browser->b.height,
861	856	browser->b.index,
862	857	browser->b.top_idx,
863		− notes->nr_asm_entries);
	858	+ notes->src->nr_asm_entries);
864	859	}
865	860	continue;
866	861	case K_ENTER:
@@ -884,7 +879,7 @@ show_sup_ins: · 17 unmodified lines
884	879	continue;
885	880	}
886	881	case 'P':
887		− map_symbol__annotation_dump(ms, evsel, browser->opts);
	882	+ map_symbol__annotation_dump(ms, evsel);
888	883	continue;
889	884	case 't':
890	885	if (symbol_conf.show_total_period) {
@@ -897,15 +892,15 @@ show_sup_ins: · 6 unmodified lines
897	892	annotation__update_column_widths(notes);
898	893	continue;
899	894	case 'c':
900		− if (notes->options->show_minmax_cycle)
901		− notes->options->show_minmax_cycle = false;
	895	+ if (annotate_opts.show_minmax_cycle)
	896	+ annotate_opts.show_minmax_cycle = false;
902	897	else
903		− notes->options->show_minmax_cycle = true;
	898	+ annotate_opts.show_minmax_cycle = true;
904	899	annotation__update_column_widths(notes);
905	900	continue;
906	901	case 'p':
907	902	case 'b':
908		− switch_percent_type(browser->opts, key == 'b');
	903	+ switch_percent_type(&annotate_opts, key == 'b');
909	904	hists__scnprintf_title(hists, title, sizeof(title));
910	905	annotate_browser__show(&browser->b, title, help);
911	906	continue;
@@ -932,26 +927,24 @@ out: · 20 unmodified lines
932	927	}
933	928
934	929	int map_symbol__tui_annotate(struct map_symbol ms, struct evsel evsel,
935		− struct hist_browser_timer *hbt,
936		− struct annotation_options *opts)
	930	+ struct hist_browser_timer *hbt)
937	931	{
938		− return symbol__tui_annotate(ms, evsel, hbt, opts);
	932	+ return symbol__tui_annotate(ms, evsel, hbt);
939	933	}
940	934
941	935	int hist_entry__tui_annotate(struct hist_entry he, struct evsel evsel,
942		− struct hist_browser_timer *hbt,
943		− struct annotation_options *opts)
	936	+ struct hist_browser_timer *hbt)
944	937	{
945	938	/* reset abort key so that it can get Ctrl-C as a key */
946	939	SLang_reset_tty();
947	940	SLang_init_tty(0, 0, 0);
	941	+ SLtty_set_suspend_state(true);
948	942
949		− return map_symbol__tui_annotate(&he->ms, evsel, hbt, opts);
	943	+ return map_symbol__tui_annotate(&he->ms, evsel, hbt);
950	944	}
951	945
952	946	int symbol__tui_annotate(struct map_symbol ms, struct evsel evsel,
953		− struct hist_browser_timer *hbt,
954		− struct annotation_options *opts)
	947	+ struct hist_browser_timer *hbt)
955	948	{
956	949	struct symbol *sym = ms->sym;
957	950	struct annotation *notes = symbol__annotation(sym);
@@ -965,7 +958,6 @@ int symbol__tui_annotate(struct map_symbol ms, struct evsel evsel, · 7 unmodified lines
965	958	.priv = ms,
966	959	.use_navkeypressed = true,
967	960	},
968		− .opts = opts,
969	961	};
970	962	struct dso *dso;
971	963	int ret = -1, err;
@@ -979,7 +971,7 @@ int symbol__tui_annotate(struct map_symbol ms, struct evsel evsel, · 7 unmodified lines
979	971	return -1;
980	972
981	973	if (not_annotated) {
982		− err = symbol__annotate2(ms, evsel, opts, &browser.arch);
	974	+ err = symbol__annotate2(ms, evsel, &browser.arch);
983	975	if (err) {
984	976	char msg[BUFSIZ];
985	977	dso->annotate_warned = true;
@@ -991,12 +983,12 @@ int symbol__tui_annotate(struct map_symbol ms, struct evsel evsel, · 5 unmodified lines
991	983
992	984	ui_helpline__push("Press ESC to exit");
993	985
994		− browser.b.width = notes->max_line_len;
995		− browser.b.nr_entries = notes->nr_entries;
	986	+ browser.b.width = notes->src->max_line_len;
	987	+ browser.b.nr_entries = notes->src->nr_entries;
996	988	browser.b.entries = &notes->src->source,
997	989	browser.b.width += 18; /* Percentage */
998	990
999		− if (notes->options->hide_src_code)
	991	+ if (annotate_opts.hide_src_code)
1000	992	ui_browser__init_asm_mode(&browser.b);
1001	993
1002	994	ret = annotate_browser__run(&browser, evsel, hbt);
@@ -1006,6 +998,6 @@ int symbol__tui_annotate(struct map_symbol ms, struct evsel evsel, · 3 unmodified lines
1006	998
1007	999	out_free_offsets:
1008	1000	if(not_annotated)
1009		− zfree(&notes->offsets);
	1001	+ zfree(&notes->src->offsets);
1010	1002	return ret;
1011	1003	}

tools/perf/ui/browsers/hists.c+14−22

@@ -2250,8 +2250,7 @@ struct hist_browser hist_browser__new(struct hists hists) · 2249 unmodified lines
2250	2250	static struct hist_browser *
2251	2251	perf_evsel_browser__new(struct evsel *evsel,
2252	2252	struct hist_browser_timer *hbt,
2253		− struct perf_env *env,
2254		− struct annotation_options *annotation_opts)
	2253	+ struct perf_env *env)
2255	2254	{
2256	2255	struct hist_browser *browser = hist_browser__new(evsel__hists(evsel));
2257	2256
@@ -2259,7 +2258,6 @@ perf_evsel_browser__new(struct evsel *evsel, · 1 unmodified line
2259	2258	browser->hbt = hbt;
2260	2259	browser->env = env;
2261	2260	browser->title = hists_browser__scnprintf_title;
2262		− browser->annotation_opts = annotation_opts;
2263	2261	}
2264	2262	return browser;
2265	2263	}
@@ -2432,8 +2430,8 @@ do_annotate(struct hist_browser browser, struct popup_action act) · 166 unmodified lines
2432	2430	struct hist_entry *he;
2433	2431	int err;
2434	2432
2435		− if (!browser->annotation_opts->objdump_path &&
2436		− perf_env__lookup_objdump(browser->env, &browser->annotation_opts->objdump_path))
	2433	+ if (!annotate_opts.objdump_path &&
	2434	+ perf_env__lookup_objdump(browser->env, &annotate_opts.objdump_path))
2437	2435	return 0;
2438	2436
2439	2437	notes = symbol__annotation(act->ms.sym);
@@ -2445,8 +2443,7 @@ do_annotate(struct hist_browser browser, struct popup_action act) · 5 unmodified lines
2445	2443	else
2446	2444	evsel = hists_to_evsel(browser->hists);
2447	2445
2448		− err = map_symbol__tui_annotate(&act->ms, evsel, browser->hbt,
2449		− browser->annotation_opts);
	2446	+ err = map_symbol__tui_annotate(&act->ms, evsel, browser->hbt);
2450	2447	he = hist_browser__selected_entry(browser);
2451	2448	/*
2452	2449	* offer option to annotate the other branch source or target
@@ -2943,11 +2940,10 @@ next: · 490 unmodified lines
2943	2940
2944	2941	static int evsel__hists_browse(struct evsel evsel, int nr_events, const char helpline,
2945	2942	bool left_exits, struct hist_browser_timer *hbt, float min_pcnt,
2946		− struct perf_env *env, bool warn_lost_event,
2947		− struct annotation_options *annotation_opts)
	2943	+ struct perf_env *env, bool warn_lost_event)
2948	2944	{
2949	2945	struct hists *hists = evsel__hists(evsel);
2950		− struct hist_browser *browser = perf_evsel_browser__new(evsel, hbt, env, annotation_opts);
	2946	+ struct hist_browser *browser = perf_evsel_browser__new(evsel, hbt, env);
2951	2947	struct branch_info *bi = NULL;
2952	2948	#define MAX_OPTIONS 16
2953	2949	char *options[MAX_OPTIONS];
@@ -3004,6 +3000,7 @@ static int evsel__hists_browse(struct evsel evsel, int nr_events, const char h · 50 unmodified lines
3004	3000	/* reset abort key so that it can get Ctrl-C as a key */
3005	3001	SLang_reset_tty();
3006	3002	SLang_init_tty(0, 0, 0);
	3003	+ SLtty_set_suspend_state(true);
3007	3004
3008	3005	if (min_pcnt)
3009	3006	browser->min_pcnt = min_pcnt;
@@ -3398,7 +3395,6 @@ out: · 388 unmodified lines
3398	3395	struct evsel_menu {
3399	3396	struct ui_browser b;
3400	3397	struct evsel *selection;
3401		− struct annotation_options *annotation_opts;
3402	3398	bool lost_events, lost_events_warned;
3403	3399	float min_pcnt;
3404	3400	struct perf_env *env;
@@ -3499,8 +3495,7 @@ browse_hists: · 94 unmodified lines
3499	3495	hbt->timer(hbt->arg);
3500	3496	key = evsel__hists_browse(pos, nr_events, help, true, hbt,
3501	3497	menu->min_pcnt, menu->env,
3502		− warn_lost_event,
3503		− menu->annotation_opts);
	3498	+ warn_lost_event);
3504	3499	ui_browser__show_title(&menu->b, title);
3505	3500	switch (key) {
3506	3501	case K_TAB:
@@ -3557,7 +3552,7 @@ static bool filter_group_entries(struct ui_browser *browser __maybe_unused, · 50 unmodified lines
3557	3552
3558	3553	static int __evlist__tui_browse_hists(struct evlist evlist, int nr_entries, const char help,
3559	3554	struct hist_browser_timer hbt, float min_pcnt, struct perf_env env,
3560		− bool warn_lost_event, struct annotation_options *annotation_opts)
	3555	+ bool warn_lost_event)
3561	3556	{
3562	3557	struct evsel *pos;
3563	3558	struct evsel_menu menu = {
@@ -3572,7 +3567,6 @@ static int __evlist__tui_browse_hists(struct evlist *evlist, int nr_entries, con · 8 unmodified lines
3572	3567	},
3573	3568	.min_pcnt = min_pcnt,
3574	3569	.env = env,
3575		− .annotation_opts = annotation_opts,
3576	3570	};
3577	3571
3578	3572	ui_helpline__push("Press ESC to exit");
@@ -3607,8 +3601,7 @@ static bool evlist__single_entry(struct evlist *evlist) · 28 unmodified lines
3607	3601	}
3608	3602
3609	3603	int evlist__tui_browse_hists(struct evlist evlist, const char help, struct hist_browser_timer *hbt,
3610		− float min_pcnt, struct perf_env *env, bool warn_lost_event,
3611		− struct annotation_options *annotation_opts)
	3604	+ float min_pcnt, struct perf_env *env, bool warn_lost_event)
3612	3605	{
3613	3606	int nr_entries = evlist->core.nr_entries;
3614	3607
@@ -3617,7 +3610,7 @@ single_entry: { · 2 unmodified lines
3617	3610	struct evsel *first = evlist__first(evlist);
3618	3611
3619	3612	return evsel__hists_browse(first, nr_entries, help, false, hbt, min_pcnt,
3620		− env, warn_lost_event, annotation_opts);
	3613	+ env, warn_lost_event);
3621	3614	}
3622	3615	}
3623	3616
@@ -3635,7 +3628,7 @@ single_entry: { · 11 unmodified lines
3635	3628	}
3636	3629
3637	3630	return __evlist__tui_browse_hists(evlist, nr_entries, help, hbt, min_pcnt, env,
3638		− warn_lost_event, annotation_opts);
	3631	+ warn_lost_event);
3639	3632	}
3640	3633
3641	3634	static int block_hists_browser__title(struct hist_browser browser, char bf,
@@ -3654,8 +3647,7 @@ static int block_hists_browser__title(struct hist_browser browser, char bf, · 12 unmodified lines
3654	3647	}
3655	3648
3656	3649	int block_hists_tui_browse(struct block_hist bh, struct evsel evsel,
3657		− float min_percent, struct perf_env *env,
3658		− struct annotation_options *annotation_opts)
	3650	+ float min_percent, struct perf_env *env)
3659	3651	{
3660	3652	struct hists *hists = &bh->block_hists;
3661	3653	struct hist_browser *browser;
@@ -3672,11 +3664,11 @@ int block_hists_tui_browse(struct block_hist bh, struct evsel evsel, · 10 unmodified lines
3672	3664	browser->title = block_hists_browser__title;
3673	3665	browser->min_pcnt = min_percent;
3674	3666	browser->env = env;
3675		− browser->annotation_opts = annotation_opts;
3676	3667
3677	3668	/* reset abort key so that it can get Ctrl-C as a key */
3678	3669	SLang_reset_tty();
3679	3670	SLang_init_tty(0, 0, 0);
	3671	+ SLtty_set_suspend_state(true);
3680	3672
3681	3673	memset(&action, 0, sizeof(action));
3682	3674

tools/perf/ui/browsers/hists.h+0−2

@@ -4,7 +4,6 @@ · 3 unmodified lines
4	4
5	5	#include "ui/browser.h"
6	6
7		−struct annotation_options;
8	7	struct evsel;
9	8
10	9	struct hist_browser {
@@ -15,7 +14,6 @@ struct hist_browser { · 4 unmodified lines
15	14	struct hist_browser_timer *hbt;
16	15	struct pstack *pstack;
17	16	struct perf_env *env;
18		− struct annotation_options *annotation_opts;
19	17	struct evsel *block_evsel;
20	18	int print_seq;
21	19	bool show_dso;

tools/perf/ui/browsers/scripts.c+1−0

@@ -166,6 +166,7 @@ void run_script(char *cmd) · 165 unmodified lines
166	166	printf("\033[c\033[H\033[J");
167	167	fflush(stdout);
168	168	SLang_init_tty(0, 0, 0);
	169	+ SLtty_set_suspend_state(true);
169	170	SLsmg_refresh();
170	171	}
171	172

tools/perf/ui/gtk/annotate.c+2−4

@@ -162,7 +162,6 @@ static int perf_gtk__annotate_symbol(GtkWidget window, struct map_symbol ms, · 161 unmodified lines
162	162	}
163	163
164	164	static int symbol__gtk_annotate(struct map_symbol ms, struct evsel evsel,
165		− struct annotation_options *options,
166	165	struct hist_browser_timer *hbt)
167	166	{
168	167	struct dso *dso = map__dso(ms->map);
@@ -176,7 +175,7 @@ static int symbol__gtk_annotate(struct map_symbol ms, struct evsel evsel, · 7 unmodified lines
176	175	if (dso->annotate_warned)
177	176	return -1;
178	177
179		− err = symbol__annotate(ms, evsel, options, NULL);
	178	+ err = symbol__annotate(ms, evsel, NULL);
180	179	if (err) {
181	180	char msg[BUFSIZ];
182	181	dso->annotate_warned = true;
@@ -244,10 +243,9 @@ static int symbol__gtk_annotate(struct map_symbol ms, struct evsel evsel, · 61 unmodified lines
244	243
245	244	int hist_entry__gtk_annotate(struct hist_entry *he,
246	245	struct evsel *evsel,
247		− struct annotation_options *options,
248	246	struct hist_browser_timer *hbt)
249	247	{
250		− return symbol__gtk_annotate(&he->ms, evsel, options, hbt);
	248	+ return symbol__gtk_annotate(&he->ms, evsel, hbt);
251	249	}
252	250
253	251	void perf_gtk__show_annotations(void)

tools/perf/ui/gtk/gtk.h+0−2

@@ -56,13 +56,11 @@ struct evsel; · 55 unmodified lines
56	56	struct evlist;
57	57	struct hist_entry;
58	58	struct hist_browser_timer;
59		−struct annotation_options;
60	59
61	60	int evlist__gtk_browse_hists(struct evlist evlist, const char help,
62	61	struct hist_browser_timer *hbt, float min_pcnt);
63	62	int hist_entry__gtk_annotate(struct hist_entry *he,
64	63	struct evsel *evsel,
65		− struct annotation_options *options,
66	64	struct hist_browser_timer *hbt);
67	65	void perf_gtk__show_annotations(void);
68	66

tools/perf/ui/tui/setup.c+22−0

@@ -2,12 +2,14 @@ · 1 unmodified line
2	2	#include <signal.h>
3	3	#include <stdbool.h>
4	4	#include <stdlib.h>
	5	+#include <termios.h>
5	6	#include <unistd.h>
6	7	#include <linux/kernel.h>
7	8	#ifdef HAVE_BACKTRACE_SUPPORT
8	9	#include <execinfo.h>
9	10	#endif
10	11
	12	+#include "../../util/color.h"
11	13	#include "../../util/debug.h"
12	14	#include "../browser.h"
13	15	#include "../helpline.h"
@@ -121,6 +123,23 @@ static void ui__signal(int sig) · 107 unmodified lines
121	123	exit(0);
122	124	}
123	125
	126	+static void ui__sigcont(int sig)
	127	+{
	128	+ static struct termios tty;
	129	+
	130	+ if (sig == SIGTSTP) {
	131	+ while (tcgetattr(SLang_TT_Read_FD, &tty) == -1 && errno == EINTR)
	132	+ ;
	133	+ while (write(SLang_TT_Read_FD, PERF_COLOR_RESET, sizeof(PERF_COLOR_RESET) - 1) == -1 && errno == EINTR)
	134	+ ;
	135	+ raise(SIGSTOP);
	136	+ } else {
	137	+ while (tcsetattr(SLang_TT_Read_FD, TCSADRAIN, &tty) == -1 && errno == EINTR)
	138	+ ;
	139	+ raise(SIGWINCH);
	140	+ }
	141	+}
	142	+
124	143	int ui__init(void)
125	144	{
126	145	int err;
@@ -135,6 +154,7 @@ int ui__init(void) · 8 unmodified lines
135	154	err = SLang_init_tty(-1, 0, 0);
136	155	if (err < 0)
137	156	goto out;
	157	+ SLtty_set_suspend_state(true);
138	158
139	159	err = SLkp_init();
140	160	if (err < 0) {
@@ -149,6 +169,8 @@ int ui__init(void) · 8 unmodified lines
149	169	signal(SIGINT, ui__signal);
150	170	signal(SIGQUIT, ui__signal);
151	171	signal(SIGTERM, ui__signal);
	172	+ signal(SIGTSTP, ui__sigcont);
	173	+ signal(SIGCONT, ui__sigcont);
152	174
153	175	perf_error__register(&perf_tui_eops);
154	176

tools/perf/util/Build+2−0

@@ -195,6 +195,8 @@ endif · 194 unmodified lines
195	195	perf-$(CONFIG_DWARF) += probe-finder.o
196	196	perf-$(CONFIG_DWARF) += dwarf-aux.o
197	197	perf-$(CONFIG_DWARF) += dwarf-regs.o
	198	+perf-$(CONFIG_DWARF) += debuginfo.o
	199	+perf-$(CONFIG_DWARF) += annotate-data.o
198	200
199	201	perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
200	202	perf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind-local.o

tools/perf/util/annotate-data.c+405−0

@@ -0,0 +1,405 @@
	1	+/* SPDX-License-Identifier: GPL-2.0 */
	2	+/*
	3	+ * Convert sample address to data type using DWARF debug info.
	4	+ *
	5	+ * Written by Namhyung Kim <namhyung@kernel.org>
	6	+ */
	7	+
	8	+#include <stdio.h>
	9	+#include <stdlib.h>
	10	+#include <inttypes.h>
	11	+
	12	+#include "annotate-data.h"
	13	+#include "debuginfo.h"
	14	+#include "debug.h"
	15	+#include "dso.h"
	16	+#include "evsel.h"
	17	+#include "evlist.h"
	18	+#include "map.h"
	19	+#include "map_symbol.h"
	20	+#include "strbuf.h"
	21	+#include "symbol.h"
	22	+#include "symbol_conf.h"
	23	+
	24	+/*
	25	+ * Compare type name and size to maintain them in a tree.
	26	+ * I'm not sure if DWARF would have information of a single type in many
	27	+ * different places (compilation units). If not, it could compare the
	28	+ * offset of the type entry in the .debug_info section.
	29	+ */
	30	+static int data_type_cmp(const void _key, const struct rb_node node)
	31	+{
	32	+ const struct annotated_data_type *key = _key;
	33	+ struct annotated_data_type *type;
	34	+
	35	+ type = rb_entry(node, struct annotated_data_type, node);
	36	+
	37	+ if (key->self.size != type->self.size)
	38	+ return key->self.size - type->self.size;
	39	+ return strcmp(key->self.type_name, type->self.type_name);
	40	+}
	41	+
	42	+static bool data_type_less(struct rb_node node_a, const struct rb_node node_b)
	43	+{
	44	+ struct annotated_data_type a, b;
	45	+
	46	+ a = rb_entry(node_a, struct annotated_data_type, node);
	47	+ b = rb_entry(node_b, struct annotated_data_type, node);
	48	+
	49	+ if (a->self.size != b->self.size)
	50	+ return a->self.size < b->self.size;
	51	+ return strcmp(a->self.type_name, b->self.type_name) < 0;
	52	+}
	53	+
	54	+/* Recursively add new members for struct/union */
	55	+static int __add_member_cb(Dwarf_Die die, void arg)
	56	+{
	57	+ struct annotated_member *parent = arg;
	58	+ struct annotated_member *member;
	59	+ Dwarf_Die member_type, die_mem;
	60	+ Dwarf_Word size, loc;
	61	+ Dwarf_Attribute attr;
	62	+ struct strbuf sb;
	63	+ int tag;
	64	+
	65	+ if (dwarf_tag(die) != DW_TAG_member)
	66	+ return DIE_FIND_CB_SIBLING;
	67	+
	68	+ member = zalloc(sizeof(*member));
	69	+ if (member == NULL)
	70	+ return DIE_FIND_CB_END;
	71	+
	72	+ strbuf_init(&sb, 32);
	73	+ die_get_typename(die, &sb);
	74	+
	75	+ die_get_real_type(die, &member_type);
	76	+ if (dwarf_aggregate_size(&member_type, &size) < 0)
	77	+ size = 0;
	78	+
	79	+ if (!dwarf_attr_integrate(die, DW_AT_data_member_location, &attr))
	80	+ loc = 0;
	81	+ else
	82	+ dwarf_formudata(&attr, &loc);
	83	+
	84	+ member->type_name = strbuf_detach(&sb, NULL);
	85	+ /* member->var_name can be NULL */
	86	+ if (dwarf_diename(die))
	87	+ member->var_name = strdup(dwarf_diename(die));
	88	+ member->size = size;
	89	+ member->offset = loc + parent->offset;
	90	+ INIT_LIST_HEAD(&member->children);
	91	+ list_add_tail(&member->node, &parent->children);
	92	+
	93	+ tag = dwarf_tag(&member_type);
	94	+ switch (tag) {
	95	+ case DW_TAG_structure_type:
	96	+ case DW_TAG_union_type:
	97	+ die_find_child(&member_type, __add_member_cb, member, &die_mem);
	98	+ break;
	99	+ default:
	100	+ break;
	101	+ }
	102	+ return DIE_FIND_CB_SIBLING;
	103	+}
	104	+
	105	+static void add_member_types(struct annotated_data_type parent, Dwarf_Die type)
	106	+{
	107	+ Dwarf_Die die_mem;
	108	+
	109	+ die_find_child(type, __add_member_cb, &parent->self, &die_mem);
	110	+}
	111	+
	112	+static void delete_members(struct annotated_member *member)
	113	+{
	114	+ struct annotated_member child, tmp;
	115	+
	116	+ list_for_each_entry_safe(child, tmp, &member->children, node) {
	117	+ list_del(&child->node);
	118	+ delete_members(child);
	119	+ free(child->type_name);
	120	+ free(child->var_name);
	121	+ free(child);
	122	+ }
	123	+}
	124	+
	125	+static struct annotated_data_type dso__findnew_data_type(struct dso dso,
	126	+ Dwarf_Die *type_die)
	127	+{
	128	+ struct annotated_data_type *result = NULL;
	129	+ struct annotated_data_type key;
	130	+ struct rb_node *node;
	131	+ struct strbuf sb;
	132	+ char *type_name;
	133	+ Dwarf_Word size;
	134	+
	135	+ strbuf_init(&sb, 32);
	136	+ if (die_get_typename_from_type(type_die, &sb) < 0)
	137	+ strbuf_add(&sb, "(unknown type)", 14);
	138	+ type_name = strbuf_detach(&sb, NULL);
	139	+ dwarf_aggregate_size(type_die, &size);
	140	+
	141	+ /* Check existing nodes in dso->data_types tree */
	142	+ key.self.type_name = type_name;
	143	+ key.self.size = size;
	144	+ node = rb_find(&key, &dso->data_types, data_type_cmp);
	145	+ if (node) {
	146	+ result = rb_entry(node, struct annotated_data_type, node);
	147	+ free(type_name);
	148	+ return result;
	149	+ }
	150	+
	151	+ /* If not, add a new one */
	152	+ result = zalloc(sizeof(*result));
	153	+ if (result == NULL) {
	154	+ free(type_name);
	155	+ return NULL;
	156	+ }
	157	+
	158	+ result->self.type_name = type_name;
	159	+ result->self.size = size;
	160	+ INIT_LIST_HEAD(&result->self.children);
	161	+
	162	+ if (symbol_conf.annotate_data_member)
	163	+ add_member_types(result, type_die);
	164	+
	165	+ rb_add(&result->node, &dso->data_types, data_type_less);
	166	+ return result;
	167	+}
	168	+
	169	+static bool find_cu_die(struct debuginfo di, u64 pc, Dwarf_Die cu_die)
	170	+{
	171	+ Dwarf_Off off, next_off;
	172	+ size_t header_size;
	173	+
	174	+ if (dwarf_addrdie(di->dbg, pc, cu_die) != NULL)
	175	+ return cu_die;
	176	+
	177	+ /*
	178	+ * There are some kernels don't have full aranges and contain only a few
	179	+ * aranges entries. Fallback to iterate all CU entries in .debug_info
	180	+ * in case it's missing.
	181	+ */
	182	+ off = 0;
	183	+ while (dwarf_nextcu(di->dbg, off, &next_off, &header_size,
	184	+ NULL, NULL, NULL) == 0) {
	185	+ if (dwarf_offdie(di->dbg, off + header_size, cu_die) &&
	186	+ dwarf_haspc(cu_die, pc))
	187	+ return true;
	188	+
	189	+ off = next_off;
	190	+ }
	191	+ return false;
	192	+}
	193	+
	194	+/* The type info will be saved in @type_die */
	195	+static int check_variable(Dwarf_Die var_die, Dwarf_Die type_die, int offset)
	196	+{
	197	+ Dwarf_Word size;
	198	+
	199	+ /* Get the type of the variable */
	200	+ if (die_get_real_type(var_die, type_die) == NULL) {
	201	+ pr_debug("variable has no type\n");
	202	+ ann_data_stat.no_typeinfo++;
	203	+ return -1;
	204	+ }
	205	+
	206	+ /*
	207	+ * It expects a pointer type for a memory access.
	208	+ * Convert to a real type it points to.
	209	+ */
	210	+ if (dwarf_tag(type_die) != DW_TAG_pointer_type \|\|
	211	+ die_get_real_type(type_die, type_die) == NULL) {
	212	+ pr_debug("no pointer or no type\n");
	213	+ ann_data_stat.no_typeinfo++;
	214	+ return -1;
	215	+ }
	216	+
	217	+ /* Get the size of the actual type */
	218	+ if (dwarf_aggregate_size(type_die, &size) < 0) {
	219	+ pr_debug("type size is unknown\n");
	220	+ ann_data_stat.invalid_size++;
	221	+ return -1;
	222	+ }
	223	+
	224	+ /* Minimal sanity check */
	225	+ if ((unsigned)offset >= size) {
	226	+ pr_debug("offset: %d is bigger than size: %" PRIu64 "\n", offset, size);
	227	+ ann_data_stat.bad_offset++;
	228	+ return -1;
	229	+ }
	230	+
	231	+ return 0;
	232	+}
	233	+
	234	+/* The result will be saved in @type_die */
	235	+static int find_data_type_die(struct debuginfo *di, u64 pc,
	236	+ int reg, int offset, Dwarf_Die *type_die)
	237	+{
	238	+ Dwarf_Die cu_die, var_die;
	239	+ Dwarf_Die *scopes = NULL;
	240	+ int ret = -1;
	241	+ int i, nr_scopes;
	242	+
	243	+ /* Get a compile_unit for this address */
	244	+ if (!find_cu_die(di, pc, &cu_die)) {
	245	+ pr_debug("cannot find CU for address %" PRIx64 "\n", pc);
	246	+ ann_data_stat.no_cuinfo++;
	247	+ return -1;
	248	+ }
	249	+
	250	+ /* Get a list of nested scopes - i.e. (inlined) functions and blocks. */
	251	+ nr_scopes = die_get_scopes(&cu_die, pc, &scopes);
	252	+
	253	+ /* Search from the inner-most scope to the outer */
	254	+ for (i = nr_scopes - 1; i >= 0; i--) {
	255	+ /* Look up variables/parameters in this scope */
	256	+ if (!die_find_variable_by_reg(&scopes[i], pc, reg, &var_die))
	257	+ continue;
	258	+
	259	+ /* Found a variable, see if it's correct */
	260	+ ret = check_variable(&var_die, type_die, offset);
	261	+ goto out;
	262	+ }
	263	+ if (ret < 0)
	264	+ ann_data_stat.no_var++;
	265	+
	266	+out:
	267	+ free(scopes);
	268	+ return ret;
	269	+}
	270	+
	271	+/**
	272	+ * find_data_type - Return a data type at the location
	273	+ * @ms: map and symbol at the location
	274	+ * @ip: instruction address of the memory access
	275	+ * @reg: register that holds the base address
	276	+ * @offset: offset from the base address
	277	+ *
	278	+ * This functions searches the debug information of the binary to get the data
	279	+ * type it accesses. The exact location is expressed by (ip, reg, offset).
	280	+ * It return %NULL if not found.
	281	+ */
	282	+struct annotated_data_type find_data_type(struct map_symbol ms, u64 ip,
	283	+ int reg, int offset)
	284	+{
	285	+ struct annotated_data_type *result = NULL;
	286	+ struct dso *dso = map__dso(ms->map);
	287	+ struct debuginfo *di;
	288	+ Dwarf_Die type_die;
	289	+ u64 pc;
	290	+
	291	+ di = debuginfo__new(dso->long_name);
	292	+ if (di == NULL) {
	293	+ pr_debug("cannot get the debug info\n");
	294	+ return NULL;
	295	+ }
	296	+
	297	+ /*
	298	+ * IP is a relative instruction address from the start of the map, as
	299	+ * it can be randomized/relocated, it needs to translate to PC which is
	300	+ * a file address for DWARF processing.
	301	+ */
	302	+ pc = map__rip_2objdump(ms->map, ip);
	303	+ if (find_data_type_die(di, pc, reg, offset, &type_die) < 0)
	304	+ goto out;
	305	+
	306	+ result = dso__findnew_data_type(dso, &type_die);
	307	+
	308	+out:
	309	+ debuginfo__delete(di);
	310	+ return result;
	311	+}
	312	+
	313	+static int alloc_data_type_histograms(struct annotated_data_type *adt, int nr_entries)
	314	+{
	315	+ int i;
	316	+ size_t sz = sizeof(struct type_hist);
	317	+
	318	+ sz += sizeof(struct type_hist_entry) * adt->self.size;
	319	+
	320	+ /* Allocate a table of pointers for each event */
	321	+ adt->nr_histograms = nr_entries;
	322	+ adt->histograms = calloc(nr_entries, sizeof(*adt->histograms));
	323	+ if (adt->histograms == NULL)
	324	+ return -ENOMEM;
	325	+
	326	+ /*
	327	+ * Each histogram is allocated for the whole size of the type.
	328	+ * TODO: Probably we can move the histogram to members.
	329	+ */
	330	+ for (i = 0; i < nr_entries; i++) {
	331	+ adt->histograms[i] = zalloc(sz);
	332	+ if (adt->histograms[i] == NULL)
	333	+ goto err;
	334	+ }
	335	+ return 0;
	336	+
	337	+err:
	338	+ while (--i >= 0)
	339	+ free(adt->histograms[i]);
	340	+ free(adt->histograms);
	341	+ return -ENOMEM;
	342	+}
	343	+
	344	+static void delete_data_type_histograms(struct annotated_data_type *adt)
	345	+{
	346	+ for (int i = 0; i < adt->nr_histograms; i++)
	347	+ free(adt->histograms[i]);
	348	+ free(adt->histograms);
	349	+}
	350	+
	351	+void annotated_data_type__tree_delete(struct rb_root *root)
	352	+{
	353	+ struct annotated_data_type *pos;
	354	+
	355	+ while (!RB_EMPTY_ROOT(root)) {
	356	+ struct rb_node *node = rb_first(root);
	357	+
	358	+ rb_erase(node, root);
	359	+ pos = rb_entry(node, struct annotated_data_type, node);
	360	+ delete_members(&pos->self);
	361	+ delete_data_type_histograms(pos);
	362	+ free(pos->self.type_name);
	363	+ free(pos);
	364	+ }
	365	+}
	366	+
	367	+/**
	368	+ * annotated_data_type__update_samples - Update histogram
	369	+ * @adt: Data type to update
	370	+ * @evsel: Event to update
	371	+ * @offset: Offset in the type
	372	+ * @nr_samples: Number of samples at this offset
	373	+ * @period: Event count at this offset
	374	+ *
	375	+ * This function updates type histogram at @ofs for @evsel. Samples are
	376	+ * aggregated before calling this function so it can be called with more
	377	+ * than one samples at a certain offset.
	378	+ */
	379	+int annotated_data_type__update_samples(struct annotated_data_type *adt,
	380	+ struct evsel *evsel, int offset,
	381	+ int nr_samples, u64 period)
	382	+{
	383	+ struct type_hist *h;
	384	+
	385	+ if (adt == NULL)
	386	+ return 0;
	387	+
	388	+ if (adt->histograms == NULL) {
	389	+ int nr = evsel->evlist->core.nr_entries;
	390	+
	391	+ if (alloc_data_type_histograms(adt, nr) < 0)
	392	+ return -1;
	393	+ }
	394	+
	395	+ if (offset < 0 \|\| offset >= adt->self.size)
	396	+ return -1;
	397	+
	398	+ h = adt->histograms[evsel->core.idx];
	399	+
	400	+ h->nr_samples += nr_samples;
	401	+ h->addr[offset].nr_samples += nr_samples;
	402	+ h->period += period;
	403	+ h->addr[offset].period += period;
	404	+ return 0;
	405	+}

tools/perf/util/annotate-data.h+143−0

@@ -0,0 +1,143 @@
	1	+/* SPDX-License-Identifier: GPL-2.0 */
	2	+#ifndef _PERF_ANNOTATE_DATA_H
	3	+#define _PERF_ANNOTATE_DATA_H
	4	+
	5	+#include <errno.h>
	6	+#include <linux/compiler.h>
	7	+#include <linux/rbtree.h>
	8	+#include <linux/types.h>
	9	+
	10	+struct evsel;
	11	+struct map_symbol;
	12	+
	13	+/**
	14	+ * struct annotated_member - Type of member field
	15	+ * @node: List entry in the parent list
	16	+ * @children: List head for child nodes
	17	+ * @type_name: Name of the member type
	18	+ * @var_name: Name of the member variable
	19	+ * @offset: Offset from the outer data type
	20	+ * @size: Size of the member field
	21	+ *
	22	+ * This represents a member type in a data type.
	23	+ */
	24	+struct annotated_member {
	25	+ struct list_head node;
	26	+ struct list_head children;
	27	+ char *type_name;
	28	+ char *var_name;
	29	+ int offset;
	30	+ int size;
	31	+};
	32	+
	33	+/**
	34	+ * struct type_hist_entry - Histogram entry per offset
	35	+ * @nr_samples: Number of samples
	36	+ * @period: Count of event
	37	+ */
	38	+struct type_hist_entry {
	39	+ int nr_samples;
	40	+ u64 period;
	41	+};
	42	+
	43	+/**
	44	+ * struct type_hist - Type histogram for each event
	45	+ * @nr_samples: Total number of samples in this data type
	46	+ * @period: Total count of the event in this data type
	47	+ * @offset: Array of histogram entry
	48	+ */
	49	+struct type_hist {
	50	+ u64 nr_samples;
	51	+ u64 period;
	52	+ struct type_hist_entry addr[];
	53	+};
	54	+
	55	+/**
	56	+ * struct annotated_data_type - Data type to profile
	57	+ * @node: RB-tree node for dso->type_tree
	58	+ * @self: Actual type information
	59	+ * @nr_histogram: Number of histogram entries
	60	+ * @histograms: An array of pointers to histograms
	61	+ *
	62	+ * This represents a data type accessed by samples in the profile data.
	63	+ */
	64	+struct annotated_data_type {
	65	+ struct rb_node node;
	66	+ struct annotated_member self;
	67	+ int nr_histograms;
	68	+ struct type_hist **histograms;
	69	+};
	70	+
	71	+extern struct annotated_data_type unknown_type;
	72	+
	73	+/**
	74	+ * struct annotated_data_stat - Debug statistics
	75	+ * @total: Total number of entry
	76	+ * @no_sym: No symbol or map found
	77	+ * @no_insn: Failed to get disasm line
	78	+ * @no_insn_ops: The instruction has no operands
	79	+ * @no_mem_ops: The instruction has no memory operands
	80	+ * @no_reg: Failed to extract a register from the operand
	81	+ * @no_dbginfo: The binary has no debug information
	82	+ * @no_cuinfo: Failed to find a compile_unit
	83	+ * @no_var: Failed to find a matching variable
	84	+ * @no_typeinfo: Failed to get a type info for the variable
	85	+ * @invalid_size: Failed to get a size info of the type
	86	+ * @bad_offset: The access offset is out of the type
	87	+ */
	88	+struct annotated_data_stat {
	89	+ int total;
	90	+ int no_sym;
	91	+ int no_insn;
	92	+ int no_insn_ops;
	93	+ int no_mem_ops;
	94	+ int no_reg;
	95	+ int no_dbginfo;
	96	+ int no_cuinfo;
	97	+ int no_var;
	98	+ int no_typeinfo;
	99	+ int invalid_size;
	100	+ int bad_offset;
	101	+};
	102	+extern struct annotated_data_stat ann_data_stat;
	103	+
	104	+#ifdef HAVE_DWARF_SUPPORT
	105	+
	106	+/* Returns data type at the location (ip, reg, offset) */
	107	+struct annotated_data_type find_data_type(struct map_symbol ms, u64 ip,
	108	+ int reg, int offset);
	109	+
	110	+/* Update type access histogram at the given offset */
	111	+int annotated_data_type__update_samples(struct annotated_data_type *adt,
	112	+ struct evsel *evsel, int offset,
	113	+ int nr_samples, u64 period);
	114	+
	115	+/* Release all data type information in the tree */
	116	+void annotated_data_type__tree_delete(struct rb_root *root);
	117	+
	118	+#else /* HAVE_DWARF_SUPPORT */
	119	+
	120	+static inline struct annotated_data_type *
	121	+find_data_type(struct map_symbol *ms __maybe_unused, u64 ip __maybe_unused,
	122	+ int reg __maybe_unused, int offset __maybe_unused)
	123	+{
	124	+ return NULL;
	125	+}
	126	+
	127	+static inline int
	128	+annotated_data_type__update_samples(struct annotated_data_type *adt __maybe_unused,
	129	+ struct evsel *evsel __maybe_unused,
	130	+ int offset __maybe_unused,
	131	+ int nr_samples __maybe_unused,
	132	+ u64 period __maybe_unused)
	133	+{
	134	+ return -1;
	135	+}
	136	+
	137	+static inline void annotated_data_type__tree_delete(struct rb_root *root __maybe_unused)
	138	+{
	139	+}
	140	+
	141	+#endif /* HAVE_DWARF_SUPPORT */
	142	+
	143	+#endif /* _PERF_ANNOTATE_DATA_H */

tools/perf/util/annotate.c+522−165

@@ -25,12 +25,14 @@ · 24 unmodified lines
25	25	#include "units.h"
26	26	#include "debug.h"
27	27	#include "annotate.h"
	28	+#include "annotate-data.h"
28	29	#include "evsel.h"
29	30	#include "evlist.h"
30	31	#include "bpf-event.h"
31	32	#include "bpf-utils.h"
32	33	#include "block-range.h"
33	34	#include "string2.h"
	35	+#include "dwarf-regs.h"
34	36	#include "util/event.h"
35	37	#include "util/sharded_mutex.h"
36	38	#include "arch/common.h"
@@ -57,6 +59,9 @@ · 20 unmodified lines
57	59
58	60	#include <linux/ctype.h>
59	61
	62	+/* global annotation options */
	63	+struct annotation_options annotate_opts;
	64	+
60	65	static regex_t file_lineno;
61	66
62	67	static struct ins_ops ins__find(struct arch arch, const char *name);
@@ -85,6 +90,8 @@ struct arch { · 22 unmodified lines
85	90	struct {
86	91	char comment_char;
87	92	char skip_functions_char;
	93	+ char register_char;
	94	+ char memory_ref_char;
88	95	} objdump;
89	96	};
90	97
@@ -96,6 +103,10 @@ static struct ins_ops nop_ops; · 5 unmodified lines
96	103	static struct ins_ops lock_ops;
97	104	static struct ins_ops ret_ops;
98	105
	106	+/* Data type collection debug statistics */
	107	+struct annotated_data_stat ann_data_stat;
	108	+LIST_HEAD(ann_insn_stat);
	109	+
99	110	static int arch__grow_instructions(struct arch *arch)
100	111	{
101	112	struct ins *new_instructions;
@@ -188,6 +199,8 @@ static struct arch architectures[] = { · 86 unmodified lines
188	199	.insn_suffix = "bwlq",
189	200	.objdump = {
190	201	.comment_char = '#',
	202	+ .register_char = '%',
	203	+ .memory_ref_char = '(',
191	204	},
192	205	},
193	206	{
@@ -340,10 +353,10 @@ bool ins__is_call(const struct ins *ins) · 146 unmodified lines
340	353	*/
341	354	static inline const char validate_comma(const char c, struct ins_operands *ops)
342	355	{
343		− if (ops->raw_comment && c > ops->raw_comment)
	356	+ if (ops->jump.raw_comment && c > ops->jump.raw_comment)
344	357	return NULL;
345	358
346		− if (ops->raw_func_start && c > ops->raw_func_start)
	359	+ if (ops->jump.raw_func_start && c > ops->jump.raw_func_start)
347	360	return NULL;
348	361
349	362	return c;
@@ -359,8 +372,8 @@ static int jump__parse(struct arch arch, struct ins_operands ops, struct map_s · 9 unmodified lines
359	372	const char *c = strchr(ops->raw, ',');
360	373	u64 start, end;
361	374
362		− ops->raw_comment = strchr(ops->raw, arch->objdump.comment_char);
363		− ops->raw_func_start = strchr(ops->raw, '<');
	375	+ ops->jump.raw_comment = strchr(ops->raw, arch->objdump.comment_char);
	376	+ ops->jump.raw_func_start = strchr(ops->raw, '<');
364	377
365	378	c = validate_comma(c, ops);
366	379
@@ -462,7 +475,16 @@ static int jump__scnprintf(struct ins ins, char bf, size_t size, · 95 unmodified lines
462	475	ops->target.offset);
463	476	}
464	477
	478	+static void jump__delete(struct ins_operands *ops __maybe_unused)
	479	+{
	480	+ /*
	481	+ * The ops->jump.raw_comment and ops->jump.raw_func_start belong to the
	482	+ * raw string, don't free them.
	483	+ */
	484	+}
	485	+
465	486	static struct ins_ops jump_ops = {
	487	+ .free = jump__delete,
466	488	.parse = jump__parse,
467	489	.scnprintf = jump__scnprintf,
468	490	};
@@ -557,6 +579,34 @@ static struct ins_ops lock_ops = { · 88 unmodified lines
557	579	.scnprintf = lock__scnprintf,
558	580	};
559	581
	582	+/*
	583	+ * Check if the operand has more than one registers like x86 SIB addressing:
	584	+ * 0x1234(%rax, %rbx, 8)
	585	+ *
	586	+ * But it doesn't care segment selectors like %gs:0x5678(%rcx), so just check
	587	+ * the input string after 'memory_ref_char' if exists.
	588	+ */
	589	+static bool check_multi_regs(struct arch arch, const char op)
	590	+{
	591	+ int count = 0;
	592	+
	593	+ if (arch->objdump.register_char == 0)
	594	+ return false;
	595	+
	596	+ if (arch->objdump.memory_ref_char) {
	597	+ op = strchr(op, arch->objdump.memory_ref_char);
	598	+ if (op == NULL)
	599	+ return false;
	600	+ }
	601	+
	602	+ while ((op = strchr(op, arch->objdump.register_char)) != NULL) {
	603	+ count++;
	604	+ op++;
	605	+ }
	606	+
	607	+ return count > 1;
	608	+}
	609	+
560	610	static int mov__parse(struct arch arch, struct ins_operands ops, struct map_symbol *ms __maybe_unused)
561	611	{
562	612	char s = strchr(ops->raw, ','), target, *comment, prev;
@@ -584,6 +634,8 @@ static int mov__parse(struct arch arch, struct ins_operands ops, struct map_sy · 21 unmodified lines
584	634	if (ops->source.raw == NULL)
585	635	return -1;
586	636
	637	+ ops->source.multi_regs = check_multi_regs(arch, ops->source.raw);
	638	+
587	639	target = skip_spaces(++s);
588	640	comment = strchr(s, arch->objdump.comment_char);
589	641
@@ -604,6 +656,8 @@ static int mov__parse(struct arch arch, struct ins_operands ops, struct map_sy · 14 unmodified lines
604	656	if (ops->target.raw == NULL)
605	657	goto out_free_source;
606	658
	659	+ ops->target.multi_regs = check_multi_regs(arch, ops->target.raw);
	660	+
607	661	if (comment == NULL)
608	662	return 0;
609	663
@@ -795,6 +849,11 @@ static struct arch arch__find(const char name) · 185 unmodified lines
795	849	return bsearch(name, architectures, nmemb, sizeof(struct arch), arch__key_cmp);
796	850	}
797	851
	852	+bool arch__is(struct arch arch, const char name)
	853	+{
	854	+ return !strcmp(arch->name, name);
	855	+}
	856	+
798	857	static struct annotated_source *annotated_source__new(void)
799	858	{
800	859	struct annotated_source src = zalloc(sizeof(src));
@@ -810,7 +869,6 @@ static __maybe_unused void annotated_source__delete(struct annotated_source *src · 9 unmodified lines
810	869	if (src == NULL)
811	870	return;
812	871	zfree(&src->histograms);
813		− zfree(&src->cycles_hist);
814	872	free(src);
815	873	}
816	874
@@ -845,18 +903,6 @@ static int annotated_source__alloc_histograms(struct annotated_source *src, · 28 unmodified lines
845	903	return src->histograms ? 0 : -1;
846	904	}
847	905
848		−/* The cycles histogram is lazily allocated. */
849		−static int symbol__alloc_hist_cycles(struct symbol *sym)
850		−{
851		− struct annotation *notes = symbol__annotation(sym);
852		− const size_t size = symbol__size(sym);
853		−
854		− notes->src->cycles_hist = calloc(size, sizeof(struct cyc_hist));
855		− if (notes->src->cycles_hist == NULL)
856		− return -1;
857		− return 0;
858		−}
859		−
860	906	void symbol__annotate_zero_histograms(struct symbol *sym)
861	907	{
862	908	struct annotation *notes = symbol__annotation(sym);
@@ -865,9 +911,10 @@ void symbol__annotate_zero_histograms(struct symbol *sym) · 2 unmodified lines
865	911	if (notes->src != NULL) {
866	912	memset(notes->src->histograms, 0,
867	913	notes->src->nr_histograms * notes->src->sizeof_sym_hist);
868		− if (notes->src->cycles_hist)
869		− memset(notes->src->cycles_hist, 0,
870		− symbol__size(sym) * sizeof(struct cyc_hist));
	914	+ }
	915	+ if (notes->branch && notes->branch->cycles_hist) {
	916	+ memset(notes->branch->cycles_hist, 0,
	917	+ symbol__size(sym) * sizeof(struct cyc_hist));
871	918	}
872	919	annotation__unlock(notes);
873	920	}
@@ -958,23 +1005,33 @@ static int __symbol__inc_addr_samples(struct map_symbol *ms, · 84 unmodified lines
958	1005	return 0;
959	1006	}
960	1007
	1008	+struct annotated_branch annotation__get_branch(struct annotation notes)
	1009	+{
	1010	+ if (notes == NULL)
	1011	+ return NULL;
	1012	+
	1013	+ if (notes->branch == NULL)
	1014	+ notes->branch = zalloc(sizeof(*notes->branch));
	1015	+
	1016	+ return notes->branch;
	1017	+}
	1018	+
961	1019	static struct cyc_hist symbol__cycles_hist(struct symbol sym)
962	1020	{
963	1021	struct annotation *notes = symbol__annotation(sym);
	1022	+ struct annotated_branch *branch;
964	1023
965		− if (notes->src == NULL) {
966		− notes->src = annotated_source__new();
967		− if (notes->src == NULL)
968		− return NULL;
969		− goto alloc_cycles_hist;
970		− }
	1024	+ branch = annotation__get_branch(notes);
	1025	+ if (branch == NULL)
	1026	+ return NULL;
	1027	+
	1028	+ if (branch->cycles_hist == NULL) {
	1029	+ const size_t size = symbol__size(sym);
971	1030
972		− if (!notes->src->cycles_hist) {
973		−alloc_cycles_hist:
974		− symbol__alloc_hist_cycles(sym);
	1031	+ branch->cycles_hist = calloc(size, sizeof(struct cyc_hist));
975	1032	}
976	1033
977		− return notes->src->cycles_hist;
	1034	+ return branch->cycles_hist;
978	1035	}
979	1036
980	1037	struct annotated_source symbol__hists(struct symbol sym, int nr_hists)
@@ -1077,12 +1134,20 @@ static unsigned annotation__count_insn(struct annotation *notes, u64 start, u64 · 96 unmodified lines
1077	1134	u64 offset;
1078	1135
1079	1136	for (offset = start; offset <= end; offset++) {
1080		− if (notes->offsets[offset])
	1137	+ if (notes->src->offsets[offset])
1081	1138	n_insn++;
1082	1139	}
1083	1140	return n_insn;
1084	1141	}
1085	1142
	1143	+static void annotated_branch__delete(struct annotated_branch *branch)
	1144	+{
	1145	+ if (branch) {
	1146	+ zfree(&branch->cycles_hist);
	1147	+ free(branch);
	1148	+ }
	1149	+}
	1150	+
1086	1151	static void annotation__count_and_fill(struct annotation notes, u64 start, u64 end, struct cyc_hist ch)
1087	1152	{
1088	1153	unsigned n_insn;
@@ -1091,6 +1156,7 @@ static void annotation__count_and_fill(struct annotation *notes, u64 start, u64 · 2 unmodified lines
1091	1156
1092	1157	n_insn = annotation__count_insn(notes, start, end);
1093	1158	if (n_insn && ch->num && ch->cycles) {
	1159	+ struct annotated_branch *branch;
1094	1160	float ipc = n_insn / ((double)ch->cycles / (double)ch->num);
1095	1161
1096	1162	/* Hide data when there are too many overlaps. */
@@ -1098,54 +1164,76 @@ static void annotation__count_and_fill(struct annotation *notes, u64 start, u64 · 1 unmodified line
1098	1164	return;
1099	1165
1100	1166	for (offset = start; offset <= end; offset++) {
1101		− struct annotation_line *al = notes->offsets[offset];
	1167	+ struct annotation_line *al = notes->src->offsets[offset];
1102	1168
1103		− if (al && al->ipc == 0.0) {
1104		− al->ipc = ipc;
	1169	+ if (al && al->cycles && al->cycles->ipc == 0.0) {
	1170	+ al->cycles->ipc = ipc;
1105	1171	cover_insn++;
1106	1172	}
1107	1173	}
1108	1174
1109		− if (cover_insn) {
1110		− notes->hit_cycles += ch->cycles;
1111		− notes->hit_insn += n_insn * ch->num;
1112		− notes->cover_insn += cover_insn;
	1175	+ branch = annotation__get_branch(notes);
	1176	+ if (cover_insn && branch) {
	1177	+ branch->hit_cycles += ch->cycles;
	1178	+ branch->hit_insn += n_insn * ch->num;
	1179	+ branch->cover_insn += cover_insn;
1113	1180	}
1114	1181	}
1115	1182	}
1116	1183
1117		−void annotation__compute_ipc(struct annotation *notes, size_t size)
	1184	+static int annotation__compute_ipc(struct annotation *notes, size_t size)
1118	1185	{
	1186	+ int err = 0;
1119	1187	s64 offset;
1120	1188
1121		− if (!notes->src \|\| !notes->src->cycles_hist)
1122		− return;
	1189	+ if (!notes->branch \|\| !notes->branch->cycles_hist)
	1190	+ return 0;
1123	1191
1124		− notes->total_insn = annotation__count_insn(notes, 0, size - 1);
1125		− notes->hit_cycles = 0;
1126		− notes->hit_insn = 0;
1127		− notes->cover_insn = 0;
	1192	+ notes->branch->total_insn = annotation__count_insn(notes, 0, size - 1);
	1193	+ notes->branch->hit_cycles = 0;
	1194	+ notes->branch->hit_insn = 0;
	1195	+ notes->branch->cover_insn = 0;
1128	1196
1129	1197	annotation__lock(notes);
1130	1198	for (offset = size - 1; offset >= 0; --offset) {
1131	1199	struct cyc_hist *ch;
1132	1200
1133		− ch = &notes->src->cycles_hist[offset];
	1201	+ ch = &notes->branch->cycles_hist[offset];
1134	1202	if (ch && ch->cycles) {
1135	1203	struct annotation_line *al;
1136	1204
	1205	+ al = notes->src->offsets[offset];
	1206	+ if (al && al->cycles == NULL) {
	1207	+ al->cycles = zalloc(sizeof(*al->cycles));
	1208	+ if (al->cycles == NULL) {
	1209	+ err = ENOMEM;
	1210	+ break;
	1211	+ }
	1212	+ }
1137	1213	if (ch->have_start)
1138	1214	annotation__count_and_fill(notes, ch->start, offset, ch);
1139		− al = notes->offsets[offset];
1140	1215	if (al && ch->num_aggr) {
1141		− al->cycles = ch->cycles_aggr / ch->num_aggr;
1142		− al->cycles_max = ch->cycles_max;
1143		− al->cycles_min = ch->cycles_min;
	1216	+ al->cycles->avg = ch->cycles_aggr / ch->num_aggr;
	1217	+ al->cycles->max = ch->cycles_max;
	1218	+ al->cycles->min = ch->cycles_min;
	1219	+ }
	1220	+ }
	1221	+ }
	1222	+
	1223	+ if (err) {
	1224	+ while (++offset < (s64)size) {
	1225	+ struct cyc_hist *ch = &notes->branch->cycles_hist[offset];
	1226	+
	1227	+ if (ch && ch->cycles) {
	1228	+ struct annotation_line *al = notes->src->offsets[offset];
	1229	+ if (al)
	1230	+ zfree(&al->cycles);
1144	1231	}
1145		− notes->have_cycles = true;
1146	1232	}
1147	1233	}
	1234	+
1148	1235	annotation__unlock(notes);
	1236	+ return 0;
1149	1237	}
1150	1238
1151	1239	int addr_map_symbol__inc_samples(struct addr_map_symbol ams, struct perf_sample sample,
@@ -1225,6 +1313,7 @@ static void annotation_line__exit(struct annotation_line *al) · 73 unmodified lines
1225	1313	{
1226	1314	zfree_srcline(&al->path);
1227	1315	zfree(&al->line);
	1316	+ zfree(&al->cycles);
1228	1317	}
1229	1318
1230	1319	static size_t disasm_line_size(int nr)
@@ -1299,6 +1388,7 @@ int disasm_line__scnprintf(struct disasm_line dl, char bf, size_t size, bool r · 68 unmodified lines
1299	1388	void annotation__exit(struct annotation *notes)
1300	1389	{
1301	1390	annotated_source__delete(notes->src);
	1391	+ annotated_branch__delete(notes->branch);
1302	1392	}
1303	1393
1304	1394	static struct sharded_mutex *sharded_mutex;
@@ -1817,7 +1907,6 @@ static int symbol__disassemble_bpf(struct symbol *sym, · 512 unmodified lines
1817	1907	struct annotate_args *args)
1818	1908	{
1819	1909	struct annotation *notes = symbol__annotation(sym);
1820		− struct annotation_options *opts = args->options;
1821	1910	struct bpf_prog_linfo *prog_linfo = NULL;
1822	1911	struct bpf_prog_info_node *info_node;
1823	1912	int len = sym->end - sym->start;
@@ -1927,7 +2016,7 @@ static int symbol__disassemble_bpf(struct symbol *sym, · 103 unmodified lines
1927	2016	prev_buf_size = buf_size;
1928	2017	fflush(s);
1929	2018
1930		− if (!opts->hide_src_code && srcline) {
	2019	+ if (!annotate_opts.hide_src_code && srcline) {
1931	2020	args->offset = -1;
1932	2021	args->line = strdup(srcline);
1933	2022	args->line_nr = 0;
@@ -2050,7 +2139,7 @@ static char expand_tabs(char line, char *storage, size_t storage_len) · 116 unmodified lines
2050	2139
2051	2140	static int symbol__disassemble(struct symbol sym, struct annotate_args args)
2052	2141	{
2053		− struct annotation_options *opts = args->options;
	2142	+ struct annotation_options *opts = &annotate_opts;
2054	2143	struct map *map = args->ms.map;
2055	2144	struct dso *dso = map__dso(map);
2056	2145	char *command;
@@ -2113,12 +2202,13 @@ static int symbol__disassemble(struct symbol sym, struct annotate_args args) · 56 unmodified lines
2113	2202	err = asprintf(&command,
2114	2203	"%s %s%s --start-address=0x%016" PRIx64
2115	2204	" --stop-address=0x%016" PRIx64
2116		− " -l -d %s %s %s %c%s%c %s%s -C \"$1\"",
	2205	+ " %s -d %s %s %s %c%s%c %s%s -C \"$1\"",
2117	2206	opts->objdump_path ?: "objdump",
2118	2207	opts->disassembler_style ? "-M " : "",
2119	2208	opts->disassembler_style ?: "",
2120	2209	map__rip_2objdump(map, sym->start),
2121	2210	map__rip_2objdump(map, sym->end),
	2211	+ opts->show_linenr ? "-l" : "",
2122	2212	opts->show_asm_raw ? "" : "--no-show-raw-insn",
2123	2213	opts->annotate_src ? "-S" : "",
2124	2214	opts->prefix ? "--prefix " : "",
@@ -2299,15 +2389,8 @@ void symbol__calc_percent(struct symbol sym, struct evsel evsel) · 174 unmodified lines
2299	2389	annotation__calc_percent(notes, evsel, symbol__size(sym));
2300	2390	}
2301	2391
2302		−int symbol__annotate(struct map_symbol ms, struct evsel evsel,
2303		− struct annotation_options options, struct arch *parch)
	2392	+static int evsel__get_arch(struct evsel evsel, struct arch *parch)
2304	2393	{
2305		− struct symbol *sym = ms->sym;
2306		− struct annotation *notes = symbol__annotation(sym);
2307		− struct annotate_args args = {
2308		− .evsel = evsel,
2309		− .options = options,
2310		− };
2311	2394	struct perf_env *env = evsel__env(evsel);
2312	2395	const char *arch_name = perf_env__arch(env);
2313	2396	struct arch *arch;
@@ -2316,25 +2399,45 @@ int symbol__annotate(struct map_symbol ms, struct evsel evsel, · 2 unmodified lines
2316	2399	if (!arch_name)
2317	2400	return errno;
2318	2401
2319		− args.arch = arch = arch__find(arch_name);
	2402	+ *parch = arch = arch__find(arch_name);
2320	2403	if (arch == NULL) {
2321	2404	pr_err("%s: unsupported arch %s\n", __func__, arch_name);
2322	2405	return ENOTSUP;
2323	2406	}
2324	2407
2325		− if (parch)
2326		− *parch = arch;
2327		−
2328	2408	if (arch->init) {
2329	2409	err = arch->init(arch, env ? env->cpuid : NULL);
2330	2410	if (err) {
2331		− pr_err("%s: failed to initialize %s arch priv area\n", __func__, arch->name);
	2411	+ pr_err("%s: failed to initialize %s arch priv area\n",
	2412	+ __func__, arch->name);
2332	2413	return err;
2333	2414	}
2334	2415	}
	2416	+ return 0;
	2417	+}
2335	2418
	2419	+int symbol__annotate(struct map_symbol ms, struct evsel evsel,
	2420	+ struct arch **parch)
	2421	+{
	2422	+ struct symbol *sym = ms->sym;
	2423	+ struct annotation *notes = symbol__annotation(sym);
	2424	+ struct annotate_args args = {
	2425	+ .evsel = evsel,
	2426	+ .options = &annotate_opts,
	2427	+ };
	2428	+ struct arch *arch = NULL;
	2429	+ int err;
	2430	+
	2431	+ err = evsel__get_arch(evsel, &arch);
	2432	+ if (err < 0)
	2433	+ return err;
	2434	+
	2435	+ if (parch)
	2436	+ *parch = arch;
	2437	+
	2438	+ args.arch = arch;
2336	2439	args.ms = *ms;
2337		− if (notes->options && notes->options->full_addr)
	2440	+ if (annotate_opts.full_addr)
2338	2441	notes->start = map__objdump_2mem(ms->map, ms->sym->start);
2339	2442	else
2340	2443	notes->start = map__rip_2objdump(ms->map, ms->sym->start);
@@ -2342,12 +2445,12 @@ int symbol__annotate(struct map_symbol ms, struct evsel evsel, · 1 unmodified line
2342	2445	return symbol__disassemble(sym, &args);
2343	2446	}
2344	2447
2345		−static void insert_source_line(struct rb_root root, struct annotation_line al,
2346		− struct annotation_options *opts)
	2448	+static void insert_source_line(struct rb_root root, struct annotation_line al)
2347	2449	{
2348	2450	struct annotation_line *iter;
2349	2451	struct rb_node **p = &root->rb_node;
2350	2452	struct rb_node *parent = NULL;
	2453	+ unsigned int percent_type = annotate_opts.percent_type;
2351	2454	int i, ret;
2352	2455
2353	2456	while (*p != NULL) {
@@ -2358,7 +2461,7 @@ static void insert_source_line(struct rb_root root, struct annotation_line al, · 4 unmodified lines
2358	2461	if (ret == 0) {
2359	2462	for (i = 0; i < al->data_nr; i++) {
2360	2463	iter->data[i].percent_sum += annotation_data__percent(&al->data[i],
2361		− opts->percent_type);
	2464	+ percent_type);
2362	2465	}
2363	2466	return;
2364	2467	}
@@ -2371,7 +2474,7 @@ static void insert_source_line(struct rb_root root, struct annotation_line al, · 6 unmodified lines
2371	2474
2372	2475	for (i = 0; i < al->data_nr; i++) {
2373	2476	al->data[i].percent_sum = annotation_data__percent(&al->data[i],
2374		− opts->percent_type);
	2477	+ percent_type);
2375	2478	}
2376	2479
2377	2480	rb_link_node(&al->rb_node, parent, p);
@@ -2493,8 +2596,7 @@ static int annotated_source__addr_fmt_width(struct list_head *lines, u64 start) · 115 unmodified lines
2493	2596	return 0;
2494	2597	}
2495	2598
2496		−int symbol__annotate_printf(struct map_symbol ms, struct evsel evsel,
2497		− struct annotation_options *opts)
	2599	+int symbol__annotate_printf(struct map_symbol ms, struct evsel evsel)
2498	2600	{
2499	2601	struct map *map = ms->map;
2500	2602	struct symbol *sym = ms->sym;
@@ -2505,6 +2607,7 @@ int symbol__annotate_printf(struct map_symbol ms, struct evsel evsel, · 4 unmodified lines
2505	2607	struct annotation *notes = symbol__annotation(sym);
2506	2608	struct sym_hist *h = annotation__histogram(notes, evsel->core.idx);
2507	2609	struct annotation_line pos, queue = NULL;
	2610	+ struct annotation_options *opts = &annotate_opts;
2508	2611	u64 start = map__rip_2objdump(map, sym->start);
2509	2612	int printed = 2, queue_len = 0, addr_fmt_width;
2510	2613	int more = 0;
@@ -2633,8 +2736,7 @@ static void FILE__write_graph(void *fp, int graph) · 122 unmodified lines
2633	2736	fputs(s, fp);
2634	2737	}
2635	2738
2636		−static int symbol__annotate_fprintf2(struct symbol sym, FILE fp,
2637		− struct annotation_options *opts)
	2739	+static int symbol__annotate_fprintf2(struct symbol sym, FILE fp)
2638	2740	{
2639	2741	struct annotation *notes = symbol__annotation(sym);
2640	2742	struct annotation_write_ops wops = {
@@ -2649,9 +2751,9 @@ static int symbol__annotate_fprintf2(struct symbol sym, FILE fp, · 8 unmodified lines
2649	2751	struct annotation_line *al;
2650	2752
2651	2753	list_for_each_entry(al, &notes->src->source, node) {
2652		− if (annotation_line__filter(al, notes))
	2754	+ if (annotation_line__filter(al))
2653	2755	continue;
2654		− annotation_line__write(al, notes, &wops, opts);
	2756	+ annotation_line__write(al, notes, &wops);
2655	2757	fputc('\n', fp);
2656	2758	wops.first_line = false;
2657	2759	}
@@ -2659,8 +2761,7 @@ static int symbol__annotate_fprintf2(struct symbol sym, FILE fp, · 1 unmodified line
2659	2761	return 0;
2660	2762	}
2661	2763
2662		−int map_symbol__annotation_dump(struct map_symbol ms, struct evsel evsel,
2663		− struct annotation_options *opts)
	2764	+int map_symbol__annotation_dump(struct map_symbol ms, struct evsel evsel)
2664	2765	{
2665	2766	const char *ev_name = evsel__name(evsel);
2666	2767	char buf[1024];
@@ -2682,7 +2783,7 @@ int map_symbol__annotation_dump(struct map_symbol ms, struct evsel evsel, · 15 unmodified lines
2682	2783
2683	2784	fprintf(fp, "%s() %s\nEvent: %s\n\n",
2684	2785	ms->sym->name, map__dso(ms->map)->long_name, ev_name);
2685		− symbol__annotate_fprintf2(ms->sym, fp, opts);
	2786	+ symbol__annotate_fprintf2(ms->sym, fp);
2686	2787
2687	2788	fclose(fp);
2688	2789	err = 0;
@@ -2769,7 +2870,7 @@ void annotation__mark_jump_targets(struct annotation notes, struct symbol sym) · 80 unmodified lines
2769	2870	return;
2770	2871
2771	2872	for (offset = 0; offset < size; ++offset) {
2772		− struct annotation_line *al = notes->offsets[offset];
	2873	+ struct annotation_line *al = notes->src->offsets[offset];
2773	2874	struct disasm_line *dl;
2774	2875
2775	2876	dl = disasm_line(al);
@@ -2777,7 +2878,7 @@ void annotation__mark_jump_targets(struct annotation notes, struct symbol sym) · 1 unmodified line
2777	2878	if (!disasm_line__is_valid_local_jump(dl, sym))
2778	2879	continue;
2779	2880
2780		− al = notes->offsets[dl->ops.target.offset];
	2881	+ al = notes->src->offsets[dl->ops.target.offset];
2781	2882
2782	2883	/*
2783	2884	* FIXME: Oops, no jump target? Buggy disassembler? Or do we
@@ -2794,19 +2895,20 @@ void annotation__mark_jump_targets(struct annotation notes, struct symbol sym) · 10 unmodified lines
2794	2895	void annotation__set_offsets(struct annotation *notes, s64 size)
2795	2896	{
2796	2897	struct annotation_line *al;
	2898	+ struct annotated_source *src = notes->src;
2797	2899
2798		− notes->max_line_len = 0;
2799		− notes->nr_entries = 0;
2800		− notes->nr_asm_entries = 0;
	2900	+ src->max_line_len = 0;
	2901	+ src->nr_entries = 0;
	2902	+ src->nr_asm_entries = 0;
2801	2903
2802		− list_for_each_entry(al, &notes->src->source, node) {
	2904	+ list_for_each_entry(al, &src->source, node) {
2803	2905	size_t line_len = strlen(al->line);
2804	2906
2805		− if (notes->max_line_len < line_len)
2806		− notes->max_line_len = line_len;
2807		− al->idx = notes->nr_entries++;
	2907	+ if (src->max_line_len < line_len)
	2908	+ src->max_line_len = line_len;
	2909	+ al->idx = src->nr_entries++;
2808	2910	if (al->offset != -1) {
2809		− al->idx_asm = notes->nr_asm_entries++;
	2911	+ al->idx_asm = src->nr_asm_entries++;
2810	2912	/*
2811	2913	* FIXME: short term bandaid to cope with assembly
2812	2914	* routines that comes with labels in the same column
@@ -2815,7 +2917,7 @@ void annotation__set_offsets(struct annotation *notes, s64 size) · 2 unmodified lines
2815	2917	* E.g. copy_user_generic_unrolled
2816	2918	*/
2817	2919	if (al->offset < size)
2818		− notes->offsets[al->offset] = al;
	2920	+ notes->src->offsets[al->offset] = al;
2819	2921	} else
2820	2922	al->idx_asm = -1;
2821	2923	}
@@ -2858,24 +2960,24 @@ void annotation__init_column_widths(struct annotation notes, struct symbol sym · 36 unmodified lines
2858	2960
2859	2961	void annotation__update_column_widths(struct annotation *notes)
2860	2962	{
2861		− if (notes->options->use_offset)
	2963	+ if (annotate_opts.use_offset)
2862	2964	notes->widths.target = notes->widths.min_addr;
2863		− else if (notes->options->full_addr)
	2965	+ else if (annotate_opts.full_addr)
2864	2966	notes->widths.target = BITS_PER_LONG / 4;
2865	2967	else
2866	2968	notes->widths.target = notes->widths.max_addr;
2867	2969
2868	2970	notes->widths.addr = notes->widths.target;
2869	2971
2870		− if (notes->options->show_nr_jumps)
	2972	+ if (annotate_opts.show_nr_jumps)
2871	2973	notes->widths.addr += notes->widths.jumps + 1;
2872	2974	}
2873	2975
2874	2976	void annotation__toggle_full_addr(struct annotation notes, struct map_symbol ms)
2875	2977	{
2876		− notes->options->full_addr = !notes->options->full_addr;
	2978	+ annotate_opts.full_addr = !annotate_opts.full_addr;
2877	2979
2878		− if (notes->options->full_addr)
	2980	+ if (annotate_opts.full_addr)
2879	2981	notes->start = map__objdump_2mem(ms->map, ms->sym->start);
2880	2982	else
2881	2983	notes->start = map__rip_2objdump(ms->map, ms->sym->start);
@@ -2884,8 +2986,7 @@ void annotation__toggle_full_addr(struct annotation notes, struct map_symbol m · 2 unmodified lines
2884	2986	}
2885	2987
2886	2988	static void annotation__calc_lines(struct annotation notes, struct map map,
2887		− struct rb_root *root,
2888		− struct annotation_options *opts)
	2989	+ struct rb_root *root)
2889	2990	{
2890	2991	struct annotation_line *al;
2891	2992	struct rb_root tmp_root = RB_ROOT;
@@ -2898,7 +2999,7 @@ static void annotation__calc_lines(struct annotation notes, struct map map, · 6 unmodified lines
2898	2999	double percent;
2899	3000
2900	3001	percent = annotation_data__percent(&al->data[i],
2901		− opts->percent_type);
	3002	+ annotate_opts.percent_type);
2902	3003
2903	3004	if (percent > percent_max)
2904	3005	percent_max = percent;
@@ -2909,22 +3010,20 @@ static void annotation__calc_lines(struct annotation notes, struct map map, · 4 unmodified lines
2909	3010
2910	3011	al->path = get_srcline(map__dso(map), notes->start + al->offset, NULL,
2911	3012	false, true, notes->start + al->offset);
2912		− insert_source_line(&tmp_root, al, opts);
	3013	+ insert_source_line(&tmp_root, al);
2913	3014	}
2914	3015
2915	3016	resort_source_line(root, &tmp_root);
2916	3017	}
2917	3018
2918		−static void symbol__calc_lines(struct map_symbol ms, struct rb_root root,
2919		− struct annotation_options *opts)
	3019	+static void symbol__calc_lines(struct map_symbol ms, struct rb_root root)
2920	3020	{
2921	3021	struct annotation *notes = symbol__annotation(ms->sym);
2922	3022
2923		− annotation__calc_lines(notes, ms->map, root, opts);
	3023	+ annotation__calc_lines(notes, ms->map, root);
2924	3024	}
2925	3025
2926		−int symbol__tty_annotate2(struct map_symbol ms, struct evsel evsel,
2927		− struct annotation_options *opts)
	3026	+int symbol__tty_annotate2(struct map_symbol ms, struct evsel evsel)
2928	3027	{
2929	3028	struct dso *dso = map__dso(ms->map);
2930	3029	struct symbol *sym = ms->sym;
@@ -2933,7 +3032,7 @@ int symbol__tty_annotate2(struct map_symbol ms, struct evsel evsel, · 2 unmodified lines
2933	3032	char buf[1024];
2934	3033	int err;
2935	3034
2936		− err = symbol__annotate2(ms, evsel, opts, NULL);
	3035	+ err = symbol__annotate2(ms, evsel, NULL);
2937	3036	if (err) {
2938	3037	char msg[BUFSIZ];
2939	3038
@@ -2943,31 +3042,31 @@ int symbol__tty_annotate2(struct map_symbol ms, struct evsel evsel, · 3 unmodified lines
2943	3042	return -1;
2944	3043	}
2945	3044
2946		− if (opts->print_lines) {
2947		− srcline_full_filename = opts->full_path;
2948		− symbol__calc_lines(ms, &source_line, opts);
	3045	+ if (annotate_opts.print_lines) {
	3046	+ srcline_full_filename = annotate_opts.full_path;
	3047	+ symbol__calc_lines(ms, &source_line);
2949	3048	print_summary(&source_line, dso->long_name);
2950	3049	}
2951	3050
2952	3051	hists__scnprintf_title(hists, buf, sizeof(buf));
2953	3052	fprintf(stdout, "%s, [percent: %s]\n%s() %s\n",
2954		− buf, percent_type_str(opts->percent_type), sym->name, dso->long_name);
2955		− symbol__annotate_fprintf2(sym, stdout, opts);
	3053	+ buf, percent_type_str(annotate_opts.percent_type), sym->name,
	3054	+ dso->long_name);
	3055	+ symbol__annotate_fprintf2(sym, stdout);
2956	3056
2957	3057	annotated_source__purge(symbol__annotation(sym)->src);
2958	3058
2959	3059	return 0;
2960	3060	}
2961	3061
2962		−int symbol__tty_annotate(struct map_symbol ms, struct evsel evsel,
2963		− struct annotation_options *opts)
	3062	+int symbol__tty_annotate(struct map_symbol ms, struct evsel evsel)
2964	3063	{
2965	3064	struct dso *dso = map__dso(ms->map);
2966	3065	struct symbol *sym = ms->sym;
2967	3066	struct rb_root source_line = RB_ROOT;
2968	3067	int err;
2969	3068
2970		− err = symbol__annotate(ms, evsel, opts, NULL);
	3069	+ err = symbol__annotate(ms, evsel, NULL);
2971	3070	if (err) {
2972	3071	char msg[BUFSIZ];
2973	3072
@@ -2979,13 +3078,13 @@ int symbol__tty_annotate(struct map_symbol ms, struct evsel evsel, · 5 unmodified lines
2979	3078
2980	3079	symbol__calc_percent(sym, evsel);
2981	3080
2982		− if (opts->print_lines) {
2983		− srcline_full_filename = opts->full_path;
2984		− symbol__calc_lines(ms, &source_line, opts);
	3081	+ if (annotate_opts.print_lines) {
	3082	+ srcline_full_filename = annotate_opts.full_path;
	3083	+ symbol__calc_lines(ms, &source_line);
2985	3084	print_summary(&source_line, dso->long_name);
2986	3085	}
2987	3086
2988		− symbol__annotate_printf(ms, evsel, opts);
	3087	+ symbol__annotate_printf(ms, evsel);
2989	3088
2990	3089	annotated_source__purge(symbol__annotation(sym)->src);
2991	3090
@@ -3046,19 +3145,20 @@ call_like: · 54 unmodified lines
3046	3145	obj__printf(obj, " ");
3047	3146	}
3048	3147
3049		− disasm_line__scnprintf(dl, bf, size, !notes->options->use_offset, notes->widths.max_ins_name);
	3148	+ disasm_line__scnprintf(dl, bf, size, !annotate_opts.use_offset, notes->widths.max_ins_name);
3050	3149	}
3051	3150
3052	3151	static void ipc_coverage_string(char bf, int size, struct annotation notes)
3053	3152	{
3054	3153	double ipc = 0.0, coverage = 0.0;
	3154	+ struct annotated_branch *branch = annotation__get_branch(notes);
3055	3155
3056		− if (notes->hit_cycles)
3057		− ipc = notes->hit_insn / ((double)notes->hit_cycles);
	3156	+ if (branch && branch->hit_cycles)
	3157	+ ipc = branch->hit_insn / ((double)branch->hit_cycles);
3058	3158
3059		− if (notes->total_insn) {
3060		− coverage = notes->cover_insn * 100.0 /
3061		− ((double)notes->total_insn);
	3159	+ if (branch && branch->total_insn) {
	3160	+ coverage = branch->cover_insn * 100.0 /
	3161	+ ((double)branch->total_insn);
3062	3162	}
3063	3163
3064	3164	scnprintf(bf, size, "(Average IPC: %.2f, IPC Coverage: %.1f%%)",
@@ -3083,8 +3183,8 @@ static void __annotation_line__write(struct annotation_line *al, struct annotati · 18 unmodified lines
3083	3183	int printed;
3084	3184
3085	3185	if (first_line && (al->offset == -1 \|\| percent_max == 0.0)) {
3086		− if (notes->have_cycles) {
3087		− if (al->ipc == 0.0 && al->cycles == 0)
	3186	+ if (notes->branch && al->cycles) {
	3187	+ if (al->cycles->ipc == 0.0 && al->cycles->avg == 0)
3088	3188	show_title = true;
3089	3189	} else
3090	3190	show_title = true;
@@ -3120,18 +3220,18 @@ static void __annotation_line__write(struct annotation_line *al, struct annotati · 29 unmodified lines
3120	3220	}
3121	3221	}
3122	3222
3123		− if (notes->have_cycles) {
3124		− if (al->ipc)
3125		− obj__printf(obj, "%*.2f ", ANNOTATION__IPC_WIDTH - 1, al->ipc);
	3223	+ if (notes->branch) {
	3224	+ if (al->cycles && al->cycles->ipc)
	3225	+ obj__printf(obj, "%*.2f ", ANNOTATION__IPC_WIDTH - 1, al->cycles->ipc);
3126	3226	else if (!show_title)
3127	3227	obj__printf(obj, "%*s", ANNOTATION__IPC_WIDTH, " ");
3128	3228	else
3129	3229	obj__printf(obj, "%*s ", ANNOTATION__IPC_WIDTH - 1, "IPC");
3130	3230
3131		− if (!notes->options->show_minmax_cycle) {
3132		− if (al->cycles)
	3231	+ if (!annotate_opts.show_minmax_cycle) {
	3232	+ if (al->cycles && al->cycles->avg)
3133	3233	obj__printf(obj, "%*" PRIu64 " ",
3134		− ANNOTATION__CYCLES_WIDTH - 1, al->cycles);
	3234	+ ANNOTATION__CYCLES_WIDTH - 1, al->cycles->avg);
3135	3235	else if (!show_title)
3136	3236	obj__printf(obj, "%*s",
3137	3237	ANNOTATION__CYCLES_WIDTH, " ");
@@ -3145,8 +3245,8 @@ static void __annotation_line__write(struct annotation_line *al, struct annotati · 7 unmodified lines
3145	3245
3146	3246	scnprintf(str, sizeof(str),
3147	3247	"%" PRIu64 "(%" PRIu64 "/%" PRIu64 ")",
3148		− al->cycles, al->cycles_min,
3149		− al->cycles_max);
	3248	+ al->cycles->avg, al->cycles->min,
	3249	+ al->cycles->max);
3150	3250
3151	3251	obj__printf(obj, "%*s ",
3152	3252	ANNOTATION__MINMAX_CYCLES_WIDTH - 1,
@@ -3172,7 +3272,7 @@ static void __annotation_line__write(struct annotation_line *al, struct annotati · 19 unmodified lines
3172	3272	if (!*al->line)
3173	3273	obj__printf(obj, "%-*s", width - pcnt_width - cycles_width, " ");
3174	3274	else if (al->offset == -1) {
3175		− if (al->line_nr && notes->options->show_linenr)
	3275	+ if (al->line_nr && annotate_opts.show_linenr)
3176	3276	printed = scnprintf(bf, sizeof(bf), "%-*d ", notes->widths.addr + 1, al->line_nr);
3177	3277	else
3178	3278	printed = scnprintf(bf, sizeof(bf), "%-*s ", notes->widths.addr, " ");
@@ -3182,15 +3282,15 @@ static void __annotation_line__write(struct annotation_line *al, struct annotati · 3 unmodified lines
3182	3282	u64 addr = al->offset;
3183	3283	int color = -1;
3184	3284
3185		− if (!notes->options->use_offset)
	3285	+ if (!annotate_opts.use_offset)
3186	3286	addr += notes->start;
3187	3287
3188		− if (!notes->options->use_offset) {
	3288	+ if (!annotate_opts.use_offset) {
3189	3289	printed = scnprintf(bf, sizeof(bf), "%" PRIx64 ": ", addr);
3190	3290	} else {
3191	3291	if (al->jump_sources &&
3192		− notes->options->offset_level >= ANNOTATION__OFFSET_JUMP_TARGETS) {
3193		− if (notes->options->show_nr_jumps) {
	3292	+ annotate_opts.offset_level >= ANNOTATION__OFFSET_JUMP_TARGETS) {
	3293	+ if (annotate_opts.show_nr_jumps) {
3194	3294	int prev;
3195	3295	printed = scnprintf(bf, sizeof(bf), "%*d ",
3196	3296	notes->widths.jumps,
@@ -3204,9 +3304,9 @@ print_addr: · 7 unmodified lines
3204	3304	printed = scnprintf(bf, sizeof(bf), "%*" PRIx64 ": ",
3205	3305	notes->widths.target, addr);
3206	3306	} else if (ins__is_call(&disasm_line(al)->ins) &&
3207		− notes->options->offset_level >= ANNOTATION__OFFSET_CALL) {
	3307	+ annotate_opts.offset_level >= ANNOTATION__OFFSET_CALL) {
3208	3308	goto print_addr;
3209		− } else if (notes->options->offset_level == ANNOTATION__MAX_OFFSET_LEVEL) {
	3309	+ } else if (annotate_opts.offset_level == ANNOTATION__MAX_OFFSET_LEVEL) {
3210	3310	goto print_addr;
3211	3311	} else {
3212	3312	printed = scnprintf(bf, sizeof(bf), "%-*s ",
@@ -3228,43 +3328,44 @@ print_addr: · 15 unmodified lines
3228	3328	}
3229	3329
3230	3330	void annotation_line__write(struct annotation_line al, struct annotation notes,
3231		− struct annotation_write_ops *wops,
3232		− struct annotation_options *opts)
	3331	+ struct annotation_write_ops *wops)
3233	3332	{
3234	3333	__annotation_line__write(al, notes, wops->first_line, wops->current_entry,
3235	3334	wops->change_color, wops->width, wops->obj,
3236		− opts->percent_type,
	3335	+ annotate_opts.percent_type,
3237	3336	wops->set_color, wops->set_percent_color,
3238	3337	wops->set_jumps_percent_color, wops->printf,
3239	3338	wops->write_graph);
3240	3339	}
3241	3340
3242	3341	int symbol__annotate2(struct map_symbol ms, struct evsel evsel,
3243		− struct annotation_options options, struct arch *parch)
	3342	+ struct arch **parch)
3244	3343	{
3245	3344	struct symbol *sym = ms->sym;
3246	3345	struct annotation *notes = symbol__annotation(sym);
3247	3346	size_t size = symbol__size(sym);
3248	3347	int nr_pcnt = 1, err;
3249	3348
3250		− notes->offsets = zalloc(size * sizeof(struct annotation_line *));
3251		− if (notes->offsets == NULL)
	3349	+ notes->src->offsets = zalloc(size * sizeof(struct annotation_line *));
	3350	+ if (notes->src->offsets == NULL)
3252	3351	return ENOMEM;
3253	3352
3254	3353	if (evsel__is_group_event(evsel))
3255	3354	nr_pcnt = evsel->core.nr_members;
3256	3355
3257		− err = symbol__annotate(ms, evsel, options, parch);
	3356	+ err = symbol__annotate(ms, evsel, parch);
3258	3357	if (err)
3259	3358	goto out_free_offsets;
3260	3359
3261		− notes->options = options;
3262		−
3263	3360	symbol__calc_percent(sym, evsel);
3264	3361
3265	3362	annotation__set_offsets(notes, size);
3266	3363	annotation__mark_jump_targets(notes, sym);
3267		− annotation__compute_ipc(notes, size);
	3364	+
	3365	+ err = annotation__compute_ipc(notes, size);
	3366	+ if (err)
	3367	+ goto out_free_offsets;
	3368	+
3268	3369	annotation__init_column_widths(notes, sym);
3269	3370	notes->nr_events = nr_pcnt;
3270	3371
@@ -3274,7 +3375,7 @@ int symbol__annotate2(struct map_symbol ms, struct evsel evsel, · 3 unmodified lines
3274	3375	return 0;
3275	3376
3276	3377	out_free_offsets:
3277		− zfree(&notes->offsets);
	3378	+ zfree(&notes->src->offsets);
3278	3379	return err;
3279	3380	}
3280	3381
@@ -3337,8 +3438,10 @@ static int annotation__config(const char var, const char value, void *data) · 56 unmodified lines
3337	3438	return 0;
3338	3439	}
3339	3440
3340		−void annotation_options__init(struct annotation_options *opt)
	3441	+void annotation_options__init(void)
3341	3442	{
	3443	+ struct annotation_options *opt = &annotate_opts;
	3444	+
3342	3445	memset(opt, 0, sizeof(*opt));
3343	3446
3344	3447	/* Default values. */
@@ -3349,16 +3452,15 @@ void annotation_options__init(struct annotation_options *opt) · 4 unmodified lines
3349	3452	opt->percent_type = PERCENT_PERIOD_LOCAL;
3350	3453	}
3351	3454
3352		−
3353		−void annotation_options__exit(struct annotation_options *opt)
	3455	+void annotation_options__exit(void)
3354	3456	{
3355		− zfree(&opt->disassembler_style);
3356		− zfree(&opt->objdump_path);
	3457	+ zfree(&annotate_opts.disassembler_style);
	3458	+ zfree(&annotate_opts.objdump_path);
3357	3459	}
3358	3460
3359		−void annotation_config__init(struct annotation_options *opt)
	3461	+void annotation_config__init(void)
3360	3462	{
3361		− perf_config(annotation__config, opt);
	3463	+ perf_config(annotation__config, &annotate_opts);
3362	3464	}
3363	3465
3364	3466	static unsigned int parse_percent_type(char str1, char str2)
@@ -3382,10 +3484,9 @@ static unsigned int parse_percent_type(char str1, char str2) · 17 unmodified lines
3382	3484	return type;
3383	3485	}
3384	3486
3385		−int annotate_parse_percent_type(const struct option opt, const char _str,
	3487	+int annotate_parse_percent_type(const struct option opt __maybe_unused, const char _str,
3386	3488	int unset __maybe_unused)
3387	3489	{
3388		− struct annotation_options *opts = opt->value;
3389	3490	unsigned int type;
3390	3491	char str1, str2;
3391	3492	int err = -1;
@@ -3404,7 +3505,7 @@ int annotate_parse_percent_type(const struct option opt, const char _str, · 12 unmodified lines
3404	3505	if (type == (unsigned int) -1)
3405	3506	type = parse_percent_type(str2, str1);
3406	3507	if (type != (unsigned int) -1) {
3407		− opts->percent_type = type;
	3508	+ annotate_opts.percent_type = type;
3408	3509	err = 0;
3409	3510	}
3410	3511
@@ -3413,11 +3514,267 @@ out: · 2 unmodified lines
3413	3514	return err;
3414	3515	}
3415	3516
3416		−int annotate_check_args(struct annotation_options *args)
	3517	+int annotate_check_args(void)
3417	3518	{
	3519	+ struct annotation_options *args = &annotate_opts;
	3520	+
3418	3521	if (args->prefix_strip && !args->prefix) {
3419	3522	pr_err("--prefix-strip requires --prefix\n");
3420	3523	return -1;
3421	3524	}
3422	3525	return 0;
3423	3526	}
	3527	+
	3528	+/*
	3529	+ * Get register number and access offset from the given instruction.
	3530	+ * It assumes AT&T x86 asm format like OFFSET(REG). Maybe it needs
	3531	+ * to revisit the format when it handles different architecture.
	3532	+ * Fills @reg and @offset when return 0.
	3533	+ */
	3534	+static int extract_reg_offset(struct arch arch, const char str,
	3535	+ struct annotated_op_loc *op_loc)
	3536	+{
	3537	+ char *p;
	3538	+ char *regname;
	3539	+
	3540	+ if (arch->objdump.register_char == 0)
	3541	+ return -1;
	3542	+
	3543	+ /*
	3544	+ * It should start from offset, but it's possible to skip 0
	3545	+ * in the asm. So 0(%rax) should be same as (%rax).
	3546	+ *
	3547	+ * However, it also start with a segment select register like
	3548	+ * %gs:0x18(%rbx). In that case it should skip the part.
	3549	+ */
	3550	+ if (*str == arch->objdump.register_char) {
	3551	+ while (str && !isdigit(str) &&
	3552	+ *str != arch->objdump.memory_ref_char)
	3553	+ str++;
	3554	+ }
	3555	+
	3556	+ op_loc->offset = strtol(str, &p, 0);
	3557	+
	3558	+ p = strchr(p, arch->objdump.register_char);
	3559	+ if (p == NULL)
	3560	+ return -1;
	3561	+
	3562	+ regname = strdup(p);
	3563	+ if (regname == NULL)
	3564	+ return -1;
	3565	+
	3566	+ op_loc->reg = get_dwarf_regnum(regname, 0);
	3567	+ free(regname);
	3568	+ return 0;
	3569	+}
	3570	+
	3571	+/**
	3572	+ * annotate_get_insn_location - Get location of instruction
	3573	+ * @arch: the architecture info
	3574	+ * @dl: the target instruction
	3575	+ * @loc: a buffer to save the data
	3576	+ *
	3577	+ * Get detailed location info (register and offset) in the instruction.
	3578	+ * It needs both source and target operand and whether it accesses a
	3579	+ * memory location. The offset field is meaningful only when the
	3580	+ * corresponding mem flag is set.
	3581	+ *
	3582	+ * Some examples on x86:
	3583	+ *
	3584	+ * mov (%rax), %rcx # src_reg = rax, src_mem = 1, src_offset = 0
	3585	+ * # dst_reg = rcx, dst_mem = 0
	3586	+ *
	3587	+ * mov 0x18, %r8 # src_reg = -1, dst_reg = r8
	3588	+ */
	3589	+int annotate_get_insn_location(struct arch arch, struct disasm_line dl,
	3590	+ struct annotated_insn_loc *loc)
	3591	+{
	3592	+ struct ins_operands *ops;
	3593	+ struct annotated_op_loc *op_loc;
	3594	+ int i;
	3595	+
	3596	+ if (!strcmp(dl->ins.name, "lock"))
	3597	+ ops = dl->ops.locked.ops;
	3598	+ else
	3599	+ ops = &dl->ops;
	3600	+
	3601	+ if (ops == NULL)
	3602	+ return -1;
	3603	+
	3604	+ memset(loc, 0, sizeof(*loc));
	3605	+
	3606	+ for_each_insn_op_loc(loc, i, op_loc) {
	3607	+ const char *insn_str = ops->source.raw;
	3608	+
	3609	+ if (i == INSN_OP_TARGET)
	3610	+ insn_str = ops->target.raw;
	3611	+
	3612	+ /* Invalidate the register by default */
	3613	+ op_loc->reg = -1;
	3614	+
	3615	+ if (insn_str == NULL)
	3616	+ continue;
	3617	+
	3618	+ if (strchr(insn_str, arch->objdump.memory_ref_char)) {
	3619	+ op_loc->mem_ref = true;
	3620	+ extract_reg_offset(arch, insn_str, op_loc);
	3621	+ } else {
	3622	+ char *s = strdup(insn_str);
	3623	+
	3624	+ if (s) {
	3625	+ op_loc->reg = get_dwarf_regnum(s, 0);
	3626	+ free(s);
	3627	+ }
	3628	+ }
	3629	+ }
	3630	+
	3631	+ return 0;
	3632	+}
	3633	+
	3634	+static void symbol__ensure_annotate(struct map_symbol ms, struct evsel evsel)
	3635	+{
	3636	+ struct disasm_line dl, tmp_dl;
	3637	+ struct annotation *notes;
	3638	+
	3639	+ notes = symbol__annotation(ms->sym);
	3640	+ if (!list_empty(&notes->src->source))
	3641	+ return;
	3642	+
	3643	+ if (symbol__annotate(ms, evsel, NULL) < 0)
	3644	+ return;
	3645	+
	3646	+ /* remove non-insn disasm lines for simplicity */
	3647	+ list_for_each_entry_safe(dl, tmp_dl, &notes->src->source, al.node) {
	3648	+ if (dl->al.offset == -1) {
	3649	+ list_del(&dl->al.node);
	3650	+ free(dl);
	3651	+ }
	3652	+ }
	3653	+}
	3654	+
	3655	+static struct disasm_line find_disasm_line(struct symbol sym, u64 ip)
	3656	+{
	3657	+ struct disasm_line *dl;
	3658	+ struct annotation *notes;
	3659	+
	3660	+ notes = symbol__annotation(sym);
	3661	+
	3662	+ list_for_each_entry(dl, &notes->src->source, al.node) {
	3663	+ if (sym->start + dl->al.offset == ip)
	3664	+ return dl;
	3665	+ }
	3666	+ return NULL;
	3667	+}
	3668	+
	3669	+static struct annotated_item_stat annotate_data_stat(struct list_head head,
	3670	+ const char *name)
	3671	+{
	3672	+ struct annotated_item_stat *istat;
	3673	+
	3674	+ list_for_each_entry(istat, head, list) {
	3675	+ if (!strcmp(istat->name, name))
	3676	+ return istat;
	3677	+ }
	3678	+
	3679	+ istat = zalloc(sizeof(*istat));
	3680	+ if (istat == NULL)
	3681	+ return NULL;
	3682	+
	3683	+ istat->name = strdup(name);
	3684	+ if (istat->name == NULL) {
	3685	+ free(istat);
	3686	+ return NULL;
	3687	+ }
	3688	+
	3689	+ list_add_tail(&istat->list, head);
	3690	+ return istat;
	3691	+}
	3692	+
	3693	+/**
	3694	+ * hist_entry__get_data_type - find data type for given hist entry
	3695	+ * @he: hist entry
	3696	+ *
	3697	+ * This function first annotates the instruction at @he->ip and extracts
	3698	+ * register and offset info from it. Then it searches the DWARF debug
	3699	+ * info to get a variable and type information using the address, register,
	3700	+ * and offset.
	3701	+ */
	3702	+struct annotated_data_type hist_entry__get_data_type(struct hist_entry he)
	3703	+{
	3704	+ struct map_symbol *ms = &he->ms;
	3705	+ struct evsel *evsel = hists_to_evsel(he->hists);
	3706	+ struct arch *arch;
	3707	+ struct disasm_line *dl;
	3708	+ struct annotated_insn_loc loc;
	3709	+ struct annotated_op_loc *op_loc;
	3710	+ struct annotated_data_type *mem_type;
	3711	+ struct annotated_item_stat *istat;
	3712	+ u64 ip = he->ip;
	3713	+ int i;
	3714	+
	3715	+ ann_data_stat.total++;
	3716	+
	3717	+ if (ms->map == NULL \|\| ms->sym == NULL) {
	3718	+ ann_data_stat.no_sym++;
	3719	+ return NULL;
	3720	+ }
	3721	+
	3722	+ if (!symbol_conf.init_annotation) {
	3723	+ ann_data_stat.no_sym++;
	3724	+ return NULL;
	3725	+ }
	3726	+
	3727	+ if (evsel__get_arch(evsel, &arch) < 0) {
	3728	+ ann_data_stat.no_insn++;
	3729	+ return NULL;
	3730	+ }
	3731	+
	3732	+ /* Make sure it runs objdump to get disasm of the function */
	3733	+ symbol__ensure_annotate(ms, evsel);
	3734	+
	3735	+ /*
	3736	+ * Get a disasm to extract the location from the insn.
	3737	+ * This is too slow...
	3738	+ */
	3739	+ dl = find_disasm_line(ms->sym, ip);
	3740	+ if (dl == NULL) {
	3741	+ ann_data_stat.no_insn++;
	3742	+ return NULL;
	3743	+ }
	3744	+
	3745	+ istat = annotate_data_stat(&ann_insn_stat, dl->ins.name);
	3746	+ if (istat == NULL) {
	3747	+ ann_data_stat.no_insn++;
	3748	+ return NULL;
	3749	+ }
	3750	+
	3751	+ if (annotate_get_insn_location(arch, dl, &loc) < 0) {
	3752	+ ann_data_stat.no_insn_ops++;
	3753	+ istat->bad++;
	3754	+ return NULL;
	3755	+ }
	3756	+
	3757	+ for_each_insn_op_loc(&loc, i, op_loc) {
	3758	+ if (!op_loc->mem_ref)
	3759	+ continue;
	3760	+
	3761	+ mem_type = find_data_type(ms, ip, op_loc->reg, op_loc->offset);
	3762	+ if (mem_type)
	3763	+ istat->good++;
	3764	+ else
	3765	+ istat->bad++;
	3766	+
	3767	+ if (symbol_conf.annotate_data_sample) {
	3768	+ annotated_data_type__update_samples(mem_type, evsel,
	3769	+ op_loc->offset,
	3770	+ he->stat.nr_events,
	3771	+ he->stat.period);
	3772	+ }
	3773	+ he->mem_type_off = op_loc->offset;
	3774	+ return mem_type;
	3775	+ }
	3776	+
	3777	+ ann_data_stat.no_mem_ops++;
	3778	+ istat->bad++;
	3779	+ return NULL;
	3780	+}

tools/perf/util/annotate.h+100−44

@@ -23,6 +23,7 @@ struct option; · 22 unmodified lines
23	23	struct perf_sample;
24	24	struct evsel;
25	25	struct symbol;
	26	+struct annotated_data_type;
26	27
27	28	struct ins {
28	29	const char *name;
@@ -31,8 +32,6 @@ struct ins { · 2 unmodified lines
31	32
32	33	struct ins_operands {
33	34	char *raw;
34		− char *raw_comment;
35		− char *raw_func_start;
36	35	struct {
37	36	char *raw;
38	37	char *name;
@@ -41,22 +40,30 @@ struct ins_operands { · 2 unmodified lines
41	40	s64 offset;
42	41	bool offset_avail;
43	42	bool outside;
	43	+ bool multi_regs;
44	44	} target;
45	45	union {
46	46	struct {
47	47	char *raw;
48	48	char *name;
49	49	u64 addr;
	50	+ bool multi_regs;
50	51	} source;
51	52	struct {
52	53	struct ins ins;
53	54	struct ins_operands *ops;
54	55	} locked;
	56	+ struct {
	57	+ char *raw_comment;
	58	+ char *raw_func_start;
	59	+ } jump;
55	60	};
56	61	};
57	62
58	63	struct arch;
59	64
	65	+bool arch__is(struct arch arch, const char name);
	66	+
60	67	struct ins_ops {
61	68	void (free)(struct ins_operands ops);
62	69	int (parse)(struct arch arch, struct ins_operands ops, struct map_symbol ms);
@@ -101,6 +108,8 @@ struct annotation_options { · 38 unmodified lines
101	108	unsigned int percent_type;
102	109	};
103	110
	111	+extern struct annotation_options annotate_opts;
	112	+
104	113	enum {
105	114	ANNOTATION__OFFSET_JUMP_TARGETS = 1,
106	115	ANNOTATION__OFFSET_CALL,
@@ -130,6 +139,13 @@ struct annotation_data { · 23 unmodified lines
130	139	struct sym_hist_entry he;
131	140	};
132	141
	142	+struct cycles_info {
	143	+ float ipc;
	144	+ u64 avg;
	145	+ u64 max;
	146	+ u64 min;
	147	+};
	148	+
133	149	struct annotation_line {
134	150	struct list_head node;
135	151	struct rb_node rb_node;
@@ -137,12 +153,9 @@ struct annotation_line { · 1 unmodified line
137	153	char *line;
138	154	int line_nr;
139	155	char *fileloc;
140		− int jump_sources;
141		− float ipc;
142		− u64 cycles;
143		− u64 cycles_max;
144		− u64 cycles_min;
145	156	char *path;
	157	+ struct cycles_info *cycles;
	158	+ int jump_sources;
146	159	u32 idx;
147	160	int idx_asm;
148	161	int data_nr;
@@ -214,8 +227,7 @@ struct annotation_write_ops { · 65 unmodified lines
214	227	};
215	228
216	229	void annotation_line__write(struct annotation_line al, struct annotation notes,
217		− struct annotation_write_ops *ops,
218		− struct annotation_options *opts);
	230	+ struct annotation_write_ops *ops);
219	231
220	232	int __annotation__scnprintf_samples_period(struct annotation *notes,
221	233	char *bf, size_t size,
@@ -264,27 +276,29 @@ struct cyc_hist { · 42 unmodified lines
264	276	* returns.
265	277	*/
266	278	struct annotated_source {
267		− struct list_head source;
268		− int nr_histograms;
269		− size_t sizeof_sym_hist;
270		− struct cyc_hist *cycles_hist;
271		− struct sym_hist *histograms;
	279	+ struct list_head source;
	280	+ size_t sizeof_sym_hist;
	281	+ struct sym_hist *histograms;
	282	+ struct annotation_line **offsets;
	283	+ int nr_histograms;
	284	+ int nr_entries;
	285	+ int nr_asm_entries;
	286	+ u16 max_line_len;
272	287	};
273	288
274		−struct LOCKABLE annotation {
275		− u64 max_coverage;
276		− u64 start;
	289	+struct annotated_branch {
277	290	u64 hit_cycles;
278	291	u64 hit_insn;
279	292	unsigned int total_insn;
280	293	unsigned int cover_insn;
281		− struct annotation_options *options;
282		− struct annotation_line **offsets;
	294	+ struct cyc_hist *cycles_hist;
	295	+ u64 max_coverage;
	296	+};
	297	+
	298	+struct LOCKABLE annotation {
	299	+ u64 start;
283	300	int nr_events;
284	301	int max_jump_sources;
285		− int nr_entries;
286		− int nr_asm_entries;
287		− u16 max_line_len;
288	302	struct {
289	303	u8 addr;
290	304	u8 jumps;
@@ -293,8 +307,8 @@ struct LOCKABLE annotation { · 2 unmodified lines
293	307	u8 max_addr;
294	308	u8 max_ins_name;
295	309	} widths;
296		− bool have_cycles;
297	310	struct annotated_source *src;
	311	+ struct annotated_branch *branch;
298	312	};
299	313
300	314	static inline void annotation__init(struct annotation *notes __maybe_unused)
@@ -308,10 +322,10 @@ bool annotation__trylock(struct annotation *notes) EXCLUSIVE_TRYLOCK_FUNCTION(tr · 7 unmodified lines
308	322
309	323	static inline int annotation__cycles_width(struct annotation *notes)
310	324	{
311		− if (notes->have_cycles && notes->options->show_minmax_cycle)
	325	+ if (notes->branch && annotate_opts.show_minmax_cycle)
312	326	return ANNOTATION__IPC_WIDTH + ANNOTATION__MINMAX_CYCLES_WIDTH;
313	327
314		− return notes->have_cycles ? ANNOTATION__IPC_WIDTH + ANNOTATION__CYCLES_WIDTH : 0;
	328	+ return notes->branch ? ANNOTATION__IPC_WIDTH + ANNOTATION__CYCLES_WIDTH : 0;
315	329	}
316	330
317	331	static inline int annotation__pcnt_width(struct annotation *notes)
@@ -319,13 +333,12 @@ static inline int annotation__pcnt_width(struct annotation *notes) · 1 unmodified line
319	333	return (symbol_conf.show_total_period ? 12 : 7) * notes->nr_events;
320	334	}
321	335
322		−static inline bool annotation_line__filter(struct annotation_line al, struct annotation notes)
	336	+static inline bool annotation_line__filter(struct annotation_line *al)
323	337	{
324		− return notes->options->hide_src_code && al->offset == -1;
	338	+ return annotate_opts.hide_src_code && al->offset == -1;
325	339	}
326	340
327	341	void annotation__set_offsets(struct annotation *notes, s64 size);
328		−void annotation__compute_ipc(struct annotation *notes, size_t size);
329	342	void annotation__mark_jump_targets(struct annotation notes, struct symbol sym);
330	343	void annotation__update_column_widths(struct annotation *notes);
331	344	void annotation__init_column_widths(struct annotation notes, struct symbol sym);
@@ -349,6 +362,8 @@ static inline struct annotation symbol__annotation(struct symbol sym) · 17 unmodified lines
349	362	int addr_map_symbol__inc_samples(struct addr_map_symbol ams, struct perf_sample sample,
350	363	struct evsel *evsel);
351	364
	365	+struct annotated_branch annotation__get_branch(struct annotation notes);
	366	+
352	367	int addr_map_symbol__account_cycles(struct addr_map_symbol *ams,
353	368	struct addr_map_symbol *start,
354	369	unsigned cycles);
@@ -361,11 +376,9 @@ void symbol__annotate_zero_histograms(struct symbol *sym); · 6 unmodified lines
361	376
362	377	int symbol__annotate(struct map_symbol *ms,
363	378	struct evsel *evsel,
364		− struct annotation_options *options,
365	379	struct arch **parch);
366	380	int symbol__annotate2(struct map_symbol *ms,
367	381	struct evsel *evsel,
368		− struct annotation_options *options,
369	382	struct arch **parch);
370	383
371	384	enum symbol_disassemble_errno {
@@ -392,43 +405,86 @@ enum symbol_disassemble_errno { · 20 unmodified lines
392	405
393	406	int symbol__strerror_disassemble(struct map_symbol ms, int errnum, char buf, size_t buflen);
394	407
395		−int symbol__annotate_printf(struct map_symbol ms, struct evsel evsel,
396		− struct annotation_options *options);
	408	+int symbol__annotate_printf(struct map_symbol ms, struct evsel evsel);
397	409	void symbol__annotate_zero_histogram(struct symbol *sym, int evidx);
398	410	void symbol__annotate_decay_histogram(struct symbol *sym, int evidx);
399	411	void annotated_source__purge(struct annotated_source *as);
400	412
401		−int map_symbol__annotation_dump(struct map_symbol ms, struct evsel evsel,
402		− struct annotation_options *opts);
	413	+int map_symbol__annotation_dump(struct map_symbol ms, struct evsel evsel);
403	414
404	415	bool ui__has_annotation(void);
405	416
406		−int symbol__tty_annotate(struct map_symbol ms, struct evsel evsel, struct annotation_options *opts);
	417	+int symbol__tty_annotate(struct map_symbol ms, struct evsel evsel);
407	418
408		−int symbol__tty_annotate2(struct map_symbol ms, struct evsel evsel, struct annotation_options *opts);
	419	+int symbol__tty_annotate2(struct map_symbol ms, struct evsel evsel);
409	420
410	421	#ifdef HAVE_SLANG_SUPPORT
411	422	int symbol__tui_annotate(struct map_symbol ms, struct evsel evsel,
412		− struct hist_browser_timer *hbt,
413		− struct annotation_options *opts);
	423	+ struct hist_browser_timer *hbt);
414	424	#else
415	425	static inline int symbol__tui_annotate(struct map_symbol *ms __maybe_unused,
416	426	struct evsel *evsel __maybe_unused,
417		− struct hist_browser_timer *hbt __maybe_unused,
418		− struct annotation_options *opts __maybe_unused)
	427	+ struct hist_browser_timer *hbt __maybe_unused)
419	428	{
420	429	return 0;
421	430	}
422	431	#endif
423	432
424		−void annotation_options__init(struct annotation_options *opt);
425		−void annotation_options__exit(struct annotation_options *opt);
	433	+void annotation_options__init(void);
	434	+void annotation_options__exit(void);
426	435
427		−void annotation_config__init(struct annotation_options *opt);
	436	+void annotation_config__init(void);
428	437
429	438	int annotate_parse_percent_type(const struct option opt, const char _str,
430	439	int unset);
431	440
432		−int annotate_check_args(struct annotation_options *args);
	441	+int annotate_check_args(void);
	442	+
	443	+/**
	444	+ * struct annotated_op_loc - Location info of instruction operand
	445	+ * @reg: Register in the operand
	446	+ * @offset: Memory access offset in the operand
	447	+ * @mem_ref: Whether the operand accesses memory
	448	+ */
	449	+struct annotated_op_loc {
	450	+ int reg;
	451	+ int offset;
	452	+ bool mem_ref;
	453	+};
	454	+
	455	+enum annotated_insn_ops {
	456	+ INSN_OP_SOURCE = 0,
	457	+ INSN_OP_TARGET = 1,
	458	+
	459	+ INSN_OP_MAX,
	460	+};
	461	+
	462	+/**
	463	+ * struct annotated_insn_loc - Location info of instruction
	464	+ * @ops: Array of location info for source and target operands
	465	+ */
	466	+struct annotated_insn_loc {
	467	+ struct annotated_op_loc ops[INSN_OP_MAX];
	468	+};
	469	+
	470	+#define for_each_insn_op_loc(insn_loc, i, op_loc) \
	471	+ for (i = INSN_OP_SOURCE, op_loc = &(insn_loc)->ops[i]; \
	472	+ i < INSN_OP_MAX; \
	473	+ i++, op_loc++)
	474	+
	475	+/* Get detailed location info in the instruction */
	476	+int annotate_get_insn_location(struct arch arch, struct disasm_line dl,
	477	+ struct annotated_insn_loc *loc);
	478	+
	479	+/* Returns a data type from the sample instruction (if any) */
	480	+struct annotated_data_type hist_entry__get_data_type(struct hist_entry he);
	481	+
	482	+struct annotated_item_stat {
	483	+ struct list_head list;
	484	+ char *name;
	485	+ int good;
	486	+ int bad;
	487	+};
	488	+extern struct list_head ann_insn_stat;
433	489
434	490	#endif /* __PERF_ANNOTATE_H */

tools/perf/util/auxtrace.c+5−2

@@ -174,7 +174,7 @@ void auxtrace_mmap_params__set_idx(struct auxtrace_mmap_params *mp, · 173 unmodified lines
174	174	struct evlist *evlist,
175	175	struct evsel *evsel, int idx)
176	176	{
177		− bool per_cpu = !perf_cpu_map__empty(evlist->core.user_requested_cpus);
	177	+ bool per_cpu = !perf_cpu_map__has_any_cpu_or_is_empty(evlist->core.user_requested_cpus);
178	178
179	179	mp->mmap_needed = evsel->needs_auxtrace_mmap;
180	180
@@ -648,7 +648,7 @@ int auxtrace_parse_snapshot_options(struct auxtrace_record *itr, · 467 unmodified lines
648	648
649	649	static int evlist__enable_event_idx(struct evlist evlist, struct evsel evsel, int idx)
650	650	{
651		− bool per_cpu_mmaps = !perf_cpu_map__empty(evlist->core.user_requested_cpus);
	651	+ bool per_cpu_mmaps = !perf_cpu_map__has_any_cpu_or_is_empty(evlist->core.user_requested_cpus);
652	652
653	653	if (per_cpu_mmaps) {
654	654	struct perf_cpu evlist_cpu = perf_cpu_map__cpu(evlist->core.all_cpus, idx);
@@ -1638,6 +1638,9 @@ int itrace_do_parse_synth_opts(struct itrace_synth_opts *synth_opts, · 983 unmodified lines
1638	1638	case 'Z':
1639	1639	synth_opts->timeless_decoding = true;
1640	1640	break;
	1641	+ case 'T':
	1642	+ synth_opts->use_timestamp = true;
	1643	+ break;
1641	1644	case ' ':
1642	1645	case ',':
1643	1646	break;

tools/perf/util/auxtrace.h+3−0

@@ -99,6 +99,7 @@ enum itrace_period_type { · 98 unmodified lines
99	99	* @remote_access: whether to synthesize remote access events
100	100	* @mem: whether to synthesize memory events
101	101	* @timeless_decoding: prefer "timeless" decoding i.e. ignore timestamps
	102	+ * @use_timestamp: use the timestamp trace as kernel time
102	103	* @vm_time_correlation: perform VM Time Correlation
103	104	* @vm_tm_corr_dry_run: VM Time Correlation dry-run
104	105	* @vm_tm_corr_args: VM Time Correlation implementation-specific arguments
@@ -146,6 +147,7 @@ struct itrace_synth_opts { · 41 unmodified lines
146	147	bool remote_access;
147	148	bool mem;
148	149	bool timeless_decoding;
	150	+ bool use_timestamp;
149	151	bool vm_time_correlation;
150	152	bool vm_tm_corr_dry_run;
151	153	char *vm_tm_corr_args;
@@ -678,6 +680,7 @@ bool auxtrace__evsel_is_auxtrace(struct perf_session *session, · 526 unmodified lines
678	680	" q: quicker (less detailed) decoding\n" \
679	681	" A: approximate IPC\n" \
680	682	" Z: prefer to ignore timestamps (so-called \"timeless\" decoding)\n" \
	683	+" T: use the timestamp trace as kernel time\n" \
681	684	" PERIOD[ns\|us\|ms\|i\|t]: specify period to sample stream\n" \
682	685	" concatenate multiple options. Default is iybxwpe or cewp\n"
683	686

tools/perf/util/block-info.c+4−6

@@ -129,9 +129,9 @@ int block_info__process_sym(struct hist_entry he, struct block_hist bh, · 128 unmodified lines
129	129	al.sym = he->ms.sym;
130	130
131	131	notes = symbol__annotation(he->ms.sym);
132		− if (!notes \|\| !notes->src \|\| !notes->src->cycles_hist)
	132	+ if (!notes \|\| !notes->branch \|\| !notes->branch->cycles_hist)
133	133	return 0;
134		− ch = notes->src->cycles_hist;
	134	+ ch = notes->branch->cycles_hist;
135	135	for (unsigned int i = 0; i < symbol__size(he->ms.sym); i++) {
136	136	if (ch[i].num_aggr) {
137	137	struct block_info *bi;
@@ -464,8 +464,7 @@ void block_info__free_report(struct block_report *reps, int nr_reps) · 326 unmodified lines
464	464	}
465	465
466	466	int report__browse_block_hists(struct block_hist *bh, float min_percent,
467		− struct evsel evsel, struct perf_env env,
468		− struct annotation_options *annotation_opts)
	467	+ struct evsel evsel, struct perf_env env)
469	468	{
470	469	int ret;
471	470
@@ -477,8 +476,7 @@ int report__browse_block_hists(struct block_hist *bh, float min_percent, · 5 unmodified lines
477	476	return 0;
478	477	case 1:
479	478	symbol_conf.report_individual_block = true;
480		− ret = block_hists_tui_browse(bh, evsel, min_percent,
481		− env, annotation_opts);
	479	+ ret = block_hists_tui_browse(bh, evsel, min_percent, env);
482	480	return ret;
483	481	default:
484	482	return -1;

tools/perf/util/block-info.h+1−2

@@ -78,8 +78,7 @@ struct block_report block_info__create_report(struct evlist evlist, · 77 unmodified lines
78	78	void block_info__free_report(struct block_report *reps, int nr_reps);
79	79
80	80	int report__browse_block_hists(struct block_hist *bh, float min_percent,
81		− struct evsel evsel, struct perf_env env,
82		− struct annotation_options *annotation_opts);
	81	+ struct evsel evsel, struct perf_env env);
83	82
84	83	float block_info__total_cycles_percent(struct hist_entry *he);
85	84

tools/perf/util/block-range.c+6−1

@@ -311,6 +311,7 @@ done: · 310 unmodified lines
311	311	double block_range__coverage(struct block_range *br)
312	312	{
313	313	struct symbol *sym;
	314	+ struct annotated_branch *branch;
314	315
315	316	if (!br) {
316	317	if (block_ranges.blocks)
@@ -323,5 +324,9 @@ double block_range__coverage(struct block_range *br) · 6 unmodified lines
323	324	if (!sym)
324	325	return -1;
325	326
326		− return (double)br->coverage / symbol__annotation(sym)->max_coverage;
	327	+ branch = symbol__annotation(sym)->branch;
	328	+ if (!branch)
	329	+ return -1;
	330	+
	331	+ return (double)br->coverage / branch->max_coverage;
327	332	}

tools/perf/util/bpf-event.c+7−4

@@ -386,6 +386,9 @@ int perf_event__synthesize_bpf_events(struct perf_session *session, · 385 unmodified lines
386	386	int err;
387	387	int fd;
388	388
	389	+ if (opts->no_bpf_event)
	390	+ return 0;
	391	+
389	392	event = malloc(sizeof(event->bpf) + KSYM_NAME_LEN + machine->id_hdr_size);
390	393	if (!event)
391	394	return -1;
@@ -542,9 +545,9 @@ int evlist__add_bpf_sb_event(struct evlist evlist, struct perf_env env) · 150 unmodified lines
542	545	return evlist__add_sb_event(evlist, &attr, bpf_event__sb_cb, env);
543	546	}
544	547
545		−void bpf_event__print_bpf_prog_info(struct bpf_prog_info *info,
546		− struct perf_env *env,
547		− FILE *fp)
	548	+void __bpf_event__print_bpf_prog_info(struct bpf_prog_info *info,
	549	+ struct perf_env *env,
	550	+ FILE *fp)
548	551	{
549	552	__u32 prog_lens = (__u32 )(uintptr_t)(info->jited_func_lens);
550	553	__u64 prog_addrs = (__u64 )(uintptr_t)(info->jited_ksyms);
@@ -560,7 +563,7 @@ void bpf_event__print_bpf_prog_info(struct bpf_prog_info *info, · 9 unmodified lines
560	563	if (info->btf_id) {
561	564	struct btf_node *node;
562	565
563		− node = perf_env__find_btf(env, info->btf_id);
	566	+ node = __perf_env__find_btf(env, info->btf_id);
564	567	if (node)
565	568	btf = btf__new((__u8 *)(node->data),
566	569	node->data_size);

tools/perf/util/bpf-event.h+6−6

@@ -33,9 +33,9 @@ struct btf_node { · 32 unmodified lines
33	33	int machine__process_bpf(struct machine machine, union perf_event event,
34	34	struct perf_sample *sample);
35	35	int evlist__add_bpf_sb_event(struct evlist evlist, struct perf_env env);
36		−void bpf_event__print_bpf_prog_info(struct bpf_prog_info *info,
37		− struct perf_env *env,
38		− FILE *fp);
	36	+void __bpf_event__print_bpf_prog_info(struct bpf_prog_info *info,
	37	+ struct perf_env *env,
	38	+ FILE *fp);
39	39	#else
40	40	static inline int machine__process_bpf(struct machine *machine __maybe_unused,
41	41	union perf_event *event __maybe_unused,
@@ -50,9 +50,9 @@ static inline int evlist__add_bpf_sb_event(struct evlist *evlist __maybe_unused, · 8 unmodified lines
50	50	return 0;
51	51	}
52	52
53		−static inline void bpf_event__print_bpf_prog_info(struct bpf_prog_info *info __maybe_unused,
54		− struct perf_env *env __maybe_unused,
55		− FILE *fp __maybe_unused)
	53	+static inline void __bpf_event__print_bpf_prog_info(struct bpf_prog_info *info __maybe_unused,
	54	+ struct perf_env *env __maybe_unused,
	55	+ FILE *fp __maybe_unused)
56	56	{
57	57
58	58	}

tools/perf/util/bpf_counter.c+1−1

@@ -455,7 +455,7 @@ static int bperf__load(struct evsel evsel, struct target target) · 454 unmodified lines
455	455	return -1;
456	456
457	457	if (!all_cpu_map) {
458		− all_cpu_map = perf_cpu_map__new(NULL);
	458	+ all_cpu_map = perf_cpu_map__new_online_cpus();
459	459	if (!all_cpu_map)
460	460	return -1;
461	461	}

tools/perf/util/bpf_lock_contention.c+1−1

@@ -318,7 +318,7 @@ int lock_contention_read(struct lock_contention *con) · 317 unmodified lines
318	318	}
319	319
320	320	/* make sure it loads the kernel map */
321		− map__load(maps__first(machine->kmaps)->map);
	321	+ maps__load_first(machine->kmaps);
322	322
323	323	prev_key = NULL;
324	324	while (!bpf_map_get_next_key(fd, prev_key, &key)) {

tools/perf/util/compress.h+5−2

@@ -3,6 +3,8 @@ · 2 unmodified lines
3	3	#define PERF_COMPRESS_H
4	4
5	5	#include <stdbool.h>
	6	+#include <stddef.h>
	7	+#include <sys/types.h>
6	8	#ifdef HAVE_ZSTD_SUPPORT
7	9	#include <zstd.h>
8	10	#endif
@@ -21,6 +23,7 @@ struct zstd_data { · 12 unmodified lines
21	23	#ifdef HAVE_ZSTD_SUPPORT
22	24	ZSTD_CStream *cstream;
23	25	ZSTD_DStream *dstream;
	26	+ int comp_level;
24	27	#endif
25	28	};
26	29
@@ -29,7 +32,7 @@ struct zstd_data { · 2 unmodified lines
29	32	int zstd_init(struct zstd_data *data, int level);
30	33	int zstd_fini(struct zstd_data *data);
31	34
32		−size_t zstd_compress_stream_to_records(struct zstd_data data, void dst, size_t dst_size,
	35	+ssize_t zstd_compress_stream_to_records(struct zstd_data data, void dst, size_t dst_size,
33	36	void *src, size_t src_size, size_t max_record_size,
34	37	size_t process_header(void *record, size_t increment));
35	38
@@ -48,7 +51,7 @@ static inline int zstd_fini(struct zstd_data *data __maybe_unused) · 12 unmodified lines
48	51	}
49	52
50	53	static inline
51		−size_t zstd_compress_stream_to_records(struct zstd_data *data __maybe_unused,
	54	+ssize_t zstd_compress_stream_to_records(struct zstd_data *data __maybe_unused,
52	55	void *dst __maybe_unused, size_t dst_size __maybe_unused,
53	56	void *src __maybe_unused, size_t src_size __maybe_unused,
54	57	size_t max_record_size __maybe_unused,

tools/perf/util/cpumap.c+1−1

@@ -672,7 +672,7 @@ struct perf_cpu_map cpu_map__online(void) / thread unsafe */ · 671 unmodified lines
672	672	static struct perf_cpu_map *online;
673	673
674	674	if (!online)
675		− online = perf_cpu_map__new(NULL); /* from /sys/devices/system/cpu/online */
	675	+ online = perf_cpu_map__new_online_cpus(); /* from /sys/devices/system/cpu/online */
676	676
677	677	return online;
678	678	}

tools/perf/util/cputopo.c+1−1

@@ -267,7 +267,7 @@ struct cpu_topology *cpu_topology__new(void) · 266 unmodified lines
267	267	ncpus = cpu__max_present_cpu().cpu;
268	268
269	269	/* build online CPU map */
270		− map = perf_cpu_map__new(NULL);
	270	+ map = perf_cpu_map__new_online_cpus();
271	271	if (map == NULL) {
272	272	pr_debug("failed to get system cpumap\n");
273	273	return NULL;

tools/perf/util/cs-etm.c+18−3

@@ -3346,12 +3346,27 @@ int cs_etm__process_auxtrace_info_full(union perf_event *event, · 3345 unmodified lines
3346	3346	etm->metadata = metadata;
3347	3347	etm->auxtrace_type = auxtrace_info->type;
3348	3348
3349		− /* Use virtual timestamps if all ETMs report ts_source = 1 */
3350		− etm->has_virtual_ts = cs_etm__has_virtual_ts(metadata, num_cpu);
	3349	+ if (etm->synth_opts.use_timestamp)
	3350	+ /*
	3351	+ * Prior to Armv8.4, Arm CPUs don't support FEAT_TRF feature,
	3352	+ * therefore the decoder cannot know if the timestamp trace is
	3353	+ * same with the kernel time.
	3354	+ *
	3355	+ * If a user has knowledge for the working platform and can
	3356	+ * specify itrace option 'T' to tell decoder to forcely use the
	3357	+ * traced timestamp as the kernel time.
	3358	+ */
	3359	+ etm->has_virtual_ts = true;
	3360	+ else
	3361	+ /* Use virtual timestamps if all ETMs report ts_source = 1 */
	3362	+ etm->has_virtual_ts = cs_etm__has_virtual_ts(metadata, num_cpu);
3351	3363
3352	3364	if (!etm->has_virtual_ts)
3353	3365	ui__warning("Virtual timestamps are not enabled, or not supported by the traced system.\n"
3354		− "The time field of the samples will not be set accurately.\n\n");
	3366	+ "The time field of the samples will not be set accurately.\n"
	3367	+ "For Arm CPUs prior to Armv8.4 or without support FEAT_TRF,\n"
	3368	+ "you can specify the itrace option 'T' for timestamp decoding\n"
	3369	+ "if the Coresight timestamp on the platform is same with the kernel time.\n\n");
3355	3370
3356	3371	etm->auxtrace.process_event = cs_etm__process_event;
3357	3372	etm->auxtrace.process_auxtrace_event = cs_etm__process_auxtrace_event;

tools/perf/util/db-export.c+2−2

@@ -253,8 +253,8 @@ static struct call_path call_path_from_sample(struct db_export dbe, · 252 unmodified lines
253	253	*/
254	254	addr_location__init(&al);
255	255	al.sym = node->ms.sym;
256		− al.map = node->ms.map;
257		− al.maps = thread__maps(thread);
	256	+ al.map = map__get(node->ms.map);
	257	+ al.maps = maps__get(thread__maps(thread));
258	258	al.addr = node->ip;
259	259
260	260	if (al.map && !al.sym)

tools/perf/util/debug.c+15−7

@@ -38,12 +38,21 @@ bool dump_trace = false, quiet = false; · 37 unmodified lines
38	38	int debug_ordered_events;
39	39	static int redirect_to_stderr;
40	40	int debug_data_convert;
41		−static FILE *debug_file;
	41	+static FILE *_debug_file;
42	42	bool debug_display_time;
43	43
	44	+FILE *debug_file(void)
	45	+{
	46	+ if (!_debug_file) {
	47	+ pr_warning_once("debug_file not set");
	48	+ debug_set_file(stderr);
	49	+ }
	50	+ return _debug_file;
	51	+}
	52	+
44	53	void debug_set_file(FILE *file)
45	54	{
46		− debug_file = file;
	55	+ _debug_file = file;
47	56	}
48	57
49	58	void debug_set_display_time(bool set)
@@ -78,8 +87,8 @@ int veprintf(int level, int var, const char *fmt, va_list args) · 28 unmodified lines
78	87	if (use_browser >= 1 && !redirect_to_stderr) {
79	88	ui_helpline__vshow(fmt, args);
80	89	} else {
81		− ret = fprintf_time(debug_file);
82		− ret += vfprintf(debug_file, fmt, args);
	90	+ ret = fprintf_time(debug_file());
	91	+ ret += vfprintf(debug_file(), fmt, args);
83	92	}
84	93	}
85	94
@@ -107,9 +116,8 @@ static int veprintf_time(u64 t, const char *fmt, va_list args) · 21 unmodified lines
107	116	nsecs -= secs * NSEC_PER_SEC;
108	117	usecs = nsecs / NSEC_PER_USEC;
109	118
110		− ret = fprintf(stderr, "[%13" PRIu64 ".%06" PRIu64 "] ",
111		− secs, usecs);
112		− ret += vfprintf(stderr, fmt, args);
	119	+ ret = fprintf(debug_file(), "[%13" PRIu64 ".%06" PRIu64 "] ", secs, usecs);
	120	+ ret += vfprintf(debug_file(), fmt, args);
113	121	return ret;
114	122	}
115	123

tools/perf/util/debug.h+1−0

@@ -77,6 +77,7 @@ int eprintf_time(int level, int var, u64 t, const char *fmt, ...) __printf(4, 5) · 76 unmodified lines
77	77	int veprintf(int level, int var, const char *fmt, va_list args);
78	78
79	79	int perf_debug_option(const char *str);
	80	+FILE *debug_file(void);
80	81	void debug_set_file(FILE *file);
81	82	void debug_set_display_time(bool set);
82	83	void perf_debug_setup(void);

tools/perf/util/debuginfo.c+205−0

@@ -0,0 +1,205 @@
	1	+// SPDX-License-Identifier: GPL-2.0-or-later
	2	+/*
	3	+ * DWARF debug information handling code. Copied from probe-finder.c.
	4	+ *
	5	+ * Written by Masami Hiramatsu <mhiramat@redhat.com>
	6	+ */
	7	+
	8	+#include <errno.h>
	9	+#include <fcntl.h>
	10	+#include <stdio.h>
	11	+#include <stdlib.h>
	12	+#include <string.h>
	13	+#include <unistd.h>
	14	+#include <linux/zalloc.h>
	15	+
	16	+#include "build-id.h"
	17	+#include "dso.h"
	18	+#include "debug.h"
	19	+#include "debuginfo.h"
	20	+#include "symbol.h"
	21	+
	22	+#ifdef HAVE_DEBUGINFOD_SUPPORT
	23	+#include <elfutils/debuginfod.h>
	24	+#endif
	25	+
	26	+/* Dwarf FL wrappers */
	27	+static char debuginfo_path; / Currently dummy */
	28	+
	29	+static const Dwfl_Callbacks offline_callbacks = {
	30	+ .find_debuginfo = dwfl_standard_find_debuginfo,
	31	+ .debuginfo_path = &debuginfo_path,
	32	+
	33	+ .section_address = dwfl_offline_section_address,
	34	+
	35	+ /* We use this table for core files too. */
	36	+ .find_elf = dwfl_build_id_find_elf,
	37	+};
	38	+
	39	+/* Get a Dwarf from offline image */
	40	+static int debuginfo__init_offline_dwarf(struct debuginfo *dbg,
	41	+ const char *path)
	42	+{
	43	+ GElf_Addr dummy;
	44	+ int fd;
	45	+
	46	+ fd = open(path, O_RDONLY);
	47	+ if (fd < 0)
	48	+ return fd;
	49	+
	50	+ dbg->dwfl = dwfl_begin(&offline_callbacks);
	51	+ if (!dbg->dwfl)
	52	+ goto error;
	53	+
	54	+ dwfl_report_begin(dbg->dwfl);
	55	+ dbg->mod = dwfl_report_offline(dbg->dwfl, "", "", fd);
	56	+ if (!dbg->mod)
	57	+ goto error;
	58	+
	59	+ dbg->dbg = dwfl_module_getdwarf(dbg->mod, &dbg->bias);
	60	+ if (!dbg->dbg)
	61	+ goto error;
	62	+
	63	+ dwfl_module_build_id(dbg->mod, &dbg->build_id, &dummy);
	64	+
	65	+ dwfl_report_end(dbg->dwfl, NULL, NULL);
	66	+
	67	+ return 0;
	68	+error:
	69	+ if (dbg->dwfl)
	70	+ dwfl_end(dbg->dwfl);
	71	+ else
	72	+ close(fd);
	73	+ memset(dbg, 0, sizeof(*dbg));
	74	+
	75	+ return -ENOENT;
	76	+}
	77	+
	78	+static struct debuginfo __debuginfo__new(const char path)
	79	+{
	80	+ struct debuginfo dbg = zalloc(sizeof(dbg));
	81	+ if (!dbg)
	82	+ return NULL;
	83	+
	84	+ if (debuginfo__init_offline_dwarf(dbg, path) < 0)
	85	+ zfree(&dbg);
	86	+ if (dbg)
	87	+ pr_debug("Open Debuginfo file: %s\n", path);
	88	+ return dbg;
	89	+}
	90	+
	91	+enum dso_binary_type distro_dwarf_types[] = {
	92	+ DSO_BINARY_TYPE__FEDORA_DEBUGINFO,
	93	+ DSO_BINARY_TYPE__UBUNTU_DEBUGINFO,
	94	+ DSO_BINARY_TYPE__OPENEMBEDDED_DEBUGINFO,
	95	+ DSO_BINARY_TYPE__BUILDID_DEBUGINFO,
	96	+ DSO_BINARY_TYPE__MIXEDUP_UBUNTU_DEBUGINFO,
	97	+ DSO_BINARY_TYPE__NOT_FOUND,
	98	+};
	99	+
	100	+struct debuginfo debuginfo__new(const char path)
	101	+{
	102	+ enum dso_binary_type *type;
	103	+ char buf[PATH_MAX], nil = '\0';
	104	+ struct dso *dso;
	105	+ struct debuginfo *dinfo = NULL;
	106	+ struct build_id bid;
	107	+
	108	+ /* Try to open distro debuginfo files */
	109	+ dso = dso__new(path);
	110	+ if (!dso)
	111	+ goto out;
	112	+
	113	+ /* Set the build id for DSO_BINARY_TYPE__BUILDID_DEBUGINFO */
	114	+ if (is_regular_file(path) && filename__read_build_id(path, &bid) > 0)
	115	+ dso__set_build_id(dso, &bid);
	116	+
	117	+ for (type = distro_dwarf_types;
	118	+ !dinfo && *type != DSO_BINARY_TYPE__NOT_FOUND;
	119	+ type++) {
	120	+ if (dso__read_binary_type_filename(dso, *type, &nil,
	121	+ buf, PATH_MAX) < 0)
	122	+ continue;
	123	+ dinfo = __debuginfo__new(buf);
	124	+ }
	125	+ dso__put(dso);
	126	+
	127	+out:
	128	+ /* if failed to open all distro debuginfo, open given binary */
	129	+ return dinfo ? : __debuginfo__new(path);
	130	+}
	131	+
	132	+void debuginfo__delete(struct debuginfo *dbg)
	133	+{
	134	+ if (dbg) {
	135	+ if (dbg->dwfl)
	136	+ dwfl_end(dbg->dwfl);
	137	+ free(dbg);
	138	+ }
	139	+}
	140	+
	141	+/* For the kernel module, we need a special code to get a DIE */
	142	+int debuginfo__get_text_offset(struct debuginfo dbg, Dwarf_Addr offs,
	143	+ bool adjust_offset)
	144	+{
	145	+ int n, i;
	146	+ Elf32_Word shndx;
	147	+ Elf_Scn *scn;
	148	+ Elf *elf;
	149	+ GElf_Shdr mem, *shdr;
	150	+ const char *p;
	151	+
	152	+ elf = dwfl_module_getelf(dbg->mod, &dbg->bias);
	153	+ if (!elf)
	154	+ return -EINVAL;
	155	+
	156	+ /* Get the number of relocations */
	157	+ n = dwfl_module_relocations(dbg->mod);
	158	+ if (n < 0)
	159	+ return -ENOENT;
	160	+ /* Search the relocation related .text section */
	161	+ for (i = 0; i < n; i++) {
	162	+ p = dwfl_module_relocation_info(dbg->mod, i, &shndx);
	163	+ if (strcmp(p, ".text") == 0) {
	164	+ /* OK, get the section header */
	165	+ scn = elf_getscn(elf, shndx);
	166	+ if (!scn)
	167	+ return -ENOENT;
	168	+ shdr = gelf_getshdr(scn, &mem);
	169	+ if (!shdr)
	170	+ return -ENOENT;
	171	+ *offs = shdr->sh_addr;
	172	+ if (adjust_offset)
	173	+ *offs -= shdr->sh_offset;
	174	+ }
	175	+ }
	176	+ return 0;
	177	+}
	178	+
	179	+#ifdef HAVE_DEBUGINFOD_SUPPORT
	180	+int get_source_from_debuginfod(const char *raw_path,
	181	+ const char sbuild_id, char *new_path)
	182	+{
	183	+ debuginfod_client *c = debuginfod_begin();
	184	+ const char *p = raw_path;
	185	+ int fd;
	186	+
	187	+ if (!c)
	188	+ return -ENOMEM;
	189	+
	190	+ fd = debuginfod_find_source(c, (const unsigned char *)sbuild_id,
	191	+ 0, p, new_path);
	192	+ pr_debug("Search %s from debuginfod -> %d\n", p, fd);
	193	+ if (fd >= 0)
	194	+ close(fd);
	195	+ debuginfod_end(c);
	196	+ if (fd < 0) {
	197	+ pr_debug("Failed to find %s in debuginfod (%s)\n",
	198	+ raw_path, sbuild_id);
	199	+ return -ENOENT;
	200	+ }
	201	+ pr_debug("Got a source %s\n", *new_path);
	202	+
	203	+ return 0;
	204	+}
	205	+#endif /* HAVE_DEBUGINFOD_SUPPORT */

tools/perf/util/debuginfo.h+64−0

@@ -0,0 +1,64 @@
	1	+/* SPDX-License-Identifier: GPL-2.0-or-later */
	2	+#ifndef _PERF_DEBUGINFO_H
	3	+#define _PERF_DEBUGINFO_H
	4	+
	5	+#include <errno.h>
	6	+#include <linux/compiler.h>
	7	+
	8	+#ifdef HAVE_DWARF_SUPPORT
	9	+
	10	+#include "dwarf-aux.h"
	11	+
	12	+/* debug information structure */
	13	+struct debuginfo {
	14	+ Dwarf *dbg;
	15	+ Dwfl_Module *mod;
	16	+ Dwfl *dwfl;
	17	+ Dwarf_Addr bias;
	18	+ const unsigned char *build_id;
	19	+};
	20	+
	21	+/* This also tries to open distro debuginfo */
	22	+struct debuginfo debuginfo__new(const char path);
	23	+void debuginfo__delete(struct debuginfo *dbg);
	24	+
	25	+int debuginfo__get_text_offset(struct debuginfo dbg, Dwarf_Addr offs,
	26	+ bool adjust_offset);
	27	+
	28	+#else /* HAVE_DWARF_SUPPORT */
	29	+
	30	+/* dummy debug information structure */
	31	+struct debuginfo {
	32	+};
	33	+
	34	+static inline struct debuginfo debuginfo__new(const char path __maybe_unused)
	35	+{
	36	+ return NULL;
	37	+}
	38	+
	39	+static inline void debuginfo__delete(struct debuginfo *dbg __maybe_unused)
	40	+{
	41	+}
	42	+
	43	+static inline int debuginfo__get_text_offset(struct debuginfo *dbg __maybe_unused,
	44	+ Dwarf_Addr *offs __maybe_unused,
	45	+ bool adjust_offset __maybe_unused)
	46	+{
	47	+ return -EINVAL;
	48	+}
	49	+
	50	+#endif /* HAVE_DWARF_SUPPORT */
	51	+
	52	+#ifdef HAVE_DEBUGINFOD_SUPPORT
	53	+int get_source_from_debuginfod(const char raw_path, const char sbuild_id,
	54	+ char **new_path);
	55	+#else /* HAVE_DEBUGINFOD_SUPPORT */
	56	+static inline int get_source_from_debuginfod(const char *raw_path __maybe_unused,
	57	+ const char *sbuild_id __maybe_unused,
	58	+ char **new_path __maybe_unused)
	59	+{
	60	+ return -ENOTSUP;
	61	+}
	62	+#endif /* HAVE_DEBUGINFOD_SUPPORT */
	63	+
	64	+#endif /* _PERF_DEBUGINFO_H */

tools/perf/util/dso.c+4−0

@@ -31,6 +31,7 @@ · 30 unmodified lines
31	31	#include "debug.h"
32	32	#include "string2.h"
33	33	#include "vdso.h"
	34	+#include "annotate-data.h"
34	35
35	36	static const char * const debuglink_paths[] = {
36	37	"%.0s%s",
@@ -1327,6 +1328,7 @@ struct dso dso__new_id(const char name, struct dso_id *id) · 1290 unmodified lines
1327	1328	dso->data.cache = RB_ROOT;
1328	1329	dso->inlined_nodes = RB_ROOT_CACHED;
1329	1330	dso->srclines = RB_ROOT_CACHED;
	1331	+ dso->data_types = RB_ROOT;
1330	1332	dso->data.fd = -1;
1331	1333	dso->data.status = DSO_DATA_STATUS_UNKNOWN;
1332	1334	dso->symtab_type = DSO_BINARY_TYPE__NOT_FOUND;
@@ -1370,6 +1372,8 @@ void dso__delete(struct dso *dso) · 37 unmodified lines
1370	1372	symbols__delete(&dso->symbols);
1371	1373	dso->symbol_names_len = 0;
1372	1374	zfree(&dso->symbol_names);
	1375	+ annotated_data_type__tree_delete(&dso->data_types);
	1376	+
1373	1377	if (dso->short_name_allocated) {
1374	1378	zfree((char **)&dso->short_name);
1375	1379	dso->short_name_allocated = false;

tools/perf/util/dso.h+2−0

@@ -154,6 +154,8 @@ struct dso { · 153 unmodified lines
154	154	size_t symbol_names_len;
155	155	struct rb_root_cached inlined_nodes;
156	156	struct rb_root_cached srclines;
	157	+ struct rb_root data_types;
	158	+
157	159	struct {
158	160	u64 addr;
159	161	struct symbol *symbol;

tools/perf/util/dwarf-aux.c+232−17

@@ -1051,32 +1051,28 @@ Dwarf_Die die_find_member(Dwarf_Die st_die, const char *name, · 1050 unmodified lines
1051	1051	}
1052	1052
1053	1053	/**
1054		− * die_get_typename - Get the name of given variable DIE
1055		− * @vr_die: a variable DIE
	1054	+ * die_get_typename_from_type - Get the name of given type DIE
	1055	+ * @type_die: a type DIE
1056	1056	* @buf: a strbuf for result type name
1057	1057	*
1058		− * Get the name of @vr_die and stores it to @buf. Return 0 if succeeded.
	1058	+ * Get the name of @type_die and stores it to @buf. Return 0 if succeeded.
1059	1059	* and Return -ENOENT if failed to find type name.
1060	1060	* Note that the result will stores typedef name if possible, and stores
1061	1061	* "*(function_type)" if the type is a function pointer.
1062	1062	*/
1063		−int die_get_typename(Dwarf_Die vr_die, struct strbuf buf)
	1063	+int die_get_typename_from_type(Dwarf_Die type_die, struct strbuf buf)
1064	1064	{
1065		− Dwarf_Die type;
1066	1065	int tag, ret;
1067	1066	const char *tmp = "";
1068	1067
1069		− if (__die_get_real_type(vr_die, &type) == NULL)
1070		− return -ENOENT;
1071		−
1072		− tag = dwarf_tag(&type);
	1068	+ tag = dwarf_tag(type_die);
1073	1069	if (tag == DW_TAG_array_type \|\| tag == DW_TAG_pointer_type)
1074	1070	tmp = "*";
1075	1071	else if (tag == DW_TAG_subroutine_type) {
1076	1072	/* Function pointer */
1077	1073	return strbuf_add(buf, "(function_type)", 15);
1078	1074	} else {
1079		− const char *name = dwarf_diename(&type);
	1075	+ const char *name = dwarf_diename(type_die);
1080	1076
1081	1077	if (tag == DW_TAG_union_type)
1082	1078	tmp = "union ";
@@ -1089,8 +1085,35 @@ int die_get_typename(Dwarf_Die vr_die, struct strbuf buf) · 6 unmodified lines
1089	1085	/* Write a base name */
1090	1086	return strbuf_addf(buf, "%s%s", tmp, name ?: "");
1091	1087	}
1092		− ret = die_get_typename(&type, buf);
1093		− return ret ? ret : strbuf_addstr(buf, tmp);
	1088	+ ret = die_get_typename(type_die, buf);
	1089	+ if (ret < 0) {
	1090	+ /* void pointer has no type attribute */
	1091	+ if (tag == DW_TAG_pointer_type && ret == -ENOENT)
	1092	+ return strbuf_addf(buf, "void*");
	1093	+
	1094	+ return ret;
	1095	+ }
	1096	+ return strbuf_addstr(buf, tmp);
	1097	+}
	1098	+
	1099	+/**
	1100	+ * die_get_typename - Get the name of given variable DIE
	1101	+ * @vr_die: a variable DIE
	1102	+ * @buf: a strbuf for result type name
	1103	+ *
	1104	+ * Get the name of @vr_die and stores it to @buf. Return 0 if succeeded.
	1105	+ * and Return -ENOENT if failed to find type name.
	1106	+ * Note that the result will stores typedef name if possible, and stores
	1107	+ * "*(function_type)" if the type is a function pointer.
	1108	+ */
	1109	+int die_get_typename(Dwarf_Die vr_die, struct strbuf buf)
	1110	+{
	1111	+ Dwarf_Die type;
	1112	+
	1113	+ if (__die_get_real_type(vr_die, &type) == NULL)
	1114	+ return -ENOENT;
	1115	+
	1116	+ return die_get_typename_from_type(&type, buf);
1094	1117	}
1095	1118
1096	1119	/**
@@ -1238,12 +1261,151 @@ int die_get_var_range(Dwarf_Die sp_die, Dwarf_Die vr_die, struct strbuf *buf) · 141 unmodified lines
1238	1261	out:
1239	1262	return ret;
1240	1263	}
1241		−#else
1242		−int die_get_var_range(Dwarf_Die *sp_die __maybe_unused,
1243		− Dwarf_Die *vr_die __maybe_unused,
1244		− struct strbuf *buf __maybe_unused)
	1264	+
	1265	+/* Interval parameters for __die_find_var_reg_cb() */
	1266	+struct find_var_data {
	1267	+ /* Target instruction address */
	1268	+ Dwarf_Addr pc;
	1269	+ /* Target memory address (for global data) */
	1270	+ Dwarf_Addr addr;
	1271	+ /* Target register */
	1272	+ unsigned reg;
	1273	+ /* Access offset, set for global data */
	1274	+ int offset;
	1275	+};
	1276	+
	1277	+/* Max number of registers DW_OP_regN supports */
	1278	+#define DWARF_OP_DIRECT_REGS 32
	1279	+
	1280	+/* Only checks direct child DIEs in the given scope. */
	1281	+static int __die_find_var_reg_cb(Dwarf_Die die_mem, void arg)
	1282	+{
	1283	+ struct find_var_data *data = arg;
	1284	+ int tag = dwarf_tag(die_mem);
	1285	+ ptrdiff_t off = 0;
	1286	+ Dwarf_Attribute attr;
	1287	+ Dwarf_Addr base, start, end;
	1288	+ Dwarf_Op *ops;
	1289	+ size_t nops;
	1290	+
	1291	+ if (tag != DW_TAG_variable && tag != DW_TAG_formal_parameter)
	1292	+ return DIE_FIND_CB_SIBLING;
	1293	+
	1294	+ if (dwarf_attr(die_mem, DW_AT_location, &attr) == NULL)
	1295	+ return DIE_FIND_CB_SIBLING;
	1296	+
	1297	+ while ((off = dwarf_getlocations(&attr, off, &base, &start, &end, &ops, &nops)) > 0) {
	1298	+ /* Assuming the location list is sorted by address */
	1299	+ if (end < data->pc)
	1300	+ continue;
	1301	+ if (start > data->pc)
	1302	+ break;
	1303	+
	1304	+ /* Only match with a simple case */
	1305	+ if (data->reg < DWARF_OP_DIRECT_REGS) {
	1306	+ if (ops->atom == (DW_OP_reg0 + data->reg) && nops == 1)
	1307	+ return DIE_FIND_CB_END;
	1308	+ } else {
	1309	+ if (ops->atom == DW_OP_regx && ops->number == data->reg &&
	1310	+ nops == 1)
	1311	+ return DIE_FIND_CB_END;
	1312	+ }
	1313	+ }
	1314	+ return DIE_FIND_CB_SIBLING;
	1315	+}
	1316	+
	1317	+/**
	1318	+ * die_find_variable_by_reg - Find a variable saved in a register
	1319	+ * @sc_die: a scope DIE
	1320	+ * @pc: the program address to find
	1321	+ * @reg: the register number to find
	1322	+ * @die_mem: a buffer to save the resulting DIE
	1323	+ *
	1324	+ * Find the variable DIE accessed by the given register.
	1325	+ */
	1326	+Dwarf_Die die_find_variable_by_reg(Dwarf_Die sc_die, Dwarf_Addr pc, int reg,
	1327	+ Dwarf_Die *die_mem)
	1328	+{
	1329	+ struct find_var_data data = {
	1330	+ .pc = pc,
	1331	+ .reg = reg,
	1332	+ };
	1333	+ return die_find_child(sc_die, __die_find_var_reg_cb, &data, die_mem);
	1334	+}
	1335	+
	1336	+/* Only checks direct child DIEs in the given scope */
	1337	+static int __die_find_var_addr_cb(Dwarf_Die die_mem, void arg)
	1338	+{
	1339	+ struct find_var_data *data = arg;
	1340	+ int tag = dwarf_tag(die_mem);
	1341	+ ptrdiff_t off = 0;
	1342	+ Dwarf_Attribute attr;
	1343	+ Dwarf_Addr base, start, end;
	1344	+ Dwarf_Word size;
	1345	+ Dwarf_Die type_die;
	1346	+ Dwarf_Op *ops;
	1347	+ size_t nops;
	1348	+
	1349	+ if (tag != DW_TAG_variable)
	1350	+ return DIE_FIND_CB_SIBLING;
	1351	+
	1352	+ if (dwarf_attr(die_mem, DW_AT_location, &attr) == NULL)
	1353	+ return DIE_FIND_CB_SIBLING;
	1354	+
	1355	+ while ((off = dwarf_getlocations(&attr, off, &base, &start, &end, &ops, &nops)) > 0) {
	1356	+ if (ops->atom != DW_OP_addr)
	1357	+ continue;
	1358	+
	1359	+ if (data->addr < ops->number)
	1360	+ continue;
	1361	+
	1362	+ if (data->addr == ops->number) {
	1363	+ /* Update offset relative to the start of the variable */
	1364	+ data->offset = 0;
	1365	+ return DIE_FIND_CB_END;
	1366	+ }
	1367	+
	1368	+ if (die_get_real_type(die_mem, &type_die) == NULL)
	1369	+ continue;
	1370	+
	1371	+ if (dwarf_aggregate_size(&type_die, &size) < 0)
	1372	+ continue;
	1373	+
	1374	+ if (data->addr >= ops->number + size)
	1375	+ continue;
	1376	+
	1377	+ /* Update offset relative to the start of the variable */
	1378	+ data->offset = data->addr - ops->number;
	1379	+ return DIE_FIND_CB_END;
	1380	+ }
	1381	+ return DIE_FIND_CB_SIBLING;
	1382	+}
	1383	+
	1384	+/**
	1385	+ * die_find_variable_by_addr - Find variable located at given address
	1386	+ * @sc_die: a scope DIE
	1387	+ * @pc: the program address to find
	1388	+ * @addr: the data address to find
	1389	+ * @die_mem: a buffer to save the resulting DIE
	1390	+ * @offset: the offset in the resulting type
	1391	+ *
	1392	+ * Find the variable DIE located at the given address (in PC-relative mode).
	1393	+ * This is usually for global variables.
	1394	+ */
	1395	+Dwarf_Die die_find_variable_by_addr(Dwarf_Die sc_die, Dwarf_Addr pc,
	1396	+ Dwarf_Addr addr, Dwarf_Die *die_mem,
	1397	+ int *offset)
1245	1398	{
1246		− return -ENOTSUP;
	1399	+ struct find_var_data data = {
	1400	+ .pc = pc,
	1401	+ .addr = addr,
	1402	+ };
	1403	+ Dwarf_Die *result;
	1404	+
	1405	+ result = die_find_child(sc_die, __die_find_var_addr_cb, &data, die_mem);
	1406	+ if (result)
	1407	+ *offset = data.offset;
	1408	+ return result;
1247	1409	}
1248	1410	#endif
1249	1411
@@ -1425,3 +1587,56 @@ void die_skip_prologue(Dwarf_Die sp_die, Dwarf_Die cu_die, · 175 unmodified lines
1425	1587
1426	1588	*entrypc = postprologue_addr;
1427	1589	}
	1590	+
	1591	+/* Internal parameters for __die_find_scope_cb() */
	1592	+struct find_scope_data {
	1593	+ /* Target instruction address */
	1594	+ Dwarf_Addr pc;
	1595	+ /* Number of scopes found [output] */
	1596	+ int nr;
	1597	+ /* Array of scopes found, 0 for the outermost one. [output] */
	1598	+ Dwarf_Die *scopes;
	1599	+};
	1600	+
	1601	+static int __die_find_scope_cb(Dwarf_Die die_mem, void arg)
	1602	+{
	1603	+ struct find_scope_data *data = arg;
	1604	+
	1605	+ if (dwarf_haspc(die_mem, data->pc)) {
	1606	+ Dwarf_Die *tmp;
	1607	+
	1608	+ tmp = realloc(data->scopes, (data->nr + 1) * sizeof(*tmp));
	1609	+ if (tmp == NULL)
	1610	+ return DIE_FIND_CB_END;
	1611	+
	1612	+ memcpy(tmp + data->nr, die_mem, sizeof(*die_mem));
	1613	+ data->scopes = tmp;
	1614	+ data->nr++;
	1615	+ return DIE_FIND_CB_CHILD;
	1616	+ }
	1617	+ return DIE_FIND_CB_SIBLING;
	1618	+}
	1619	+
	1620	+/**
	1621	+ * die_get_scopes - Return a list of scopes including the address
	1622	+ * @cu_die: a compile unit DIE
	1623	+ * @pc: the address to find
	1624	+ * @scopes: the array of DIEs for scopes (result)
	1625	+ *
	1626	+ * This function does the same as the dwarf_getscopes() but doesn't follow
	1627	+ * the origins of inlined functions. It returns the number of scopes saved
	1628	+ * in the @scopes argument. The outer scope will be saved first (index 0) and
	1629	+ * the last one is the innermost scope at the @pc.
	1630	+ */
	1631	+int die_get_scopes(Dwarf_Die cu_die, Dwarf_Addr pc, Dwarf_Die *scopes)
	1632	+{
	1633	+ struct find_scope_data data = {
	1634	+ .pc = pc,
	1635	+ };
	1636	+ Dwarf_Die die_mem;
	1637	+
	1638	+ die_find_child(cu_die, __die_find_scope_cb, &data, &die_mem);
	1639	+
	1640	+ *scopes = data.scopes;
	1641	+ return data.nr;
	1642	+}

tools/perf/util/dwarf-aux.h+49−2

@@ -116,12 +116,14 @@ Dwarf_Die die_find_variable_at(Dwarf_Die sp_die, const char *name, · 115 unmodified lines
116	116	Dwarf_Die die_find_member(Dwarf_Die st_die, const char *name,
117	117	Dwarf_Die *die_mem);
118	118
	119	+/* Get the name of given type DIE */
	120	+int die_get_typename_from_type(Dwarf_Die type_die, struct strbuf buf);
	121	+
119	122	/* Get the name of given variable DIE */
120	123	int die_get_typename(Dwarf_Die vr_die, struct strbuf buf);
121	124
122	125	/* Get the name and type of given variable DIE, stored as "type\tname" */
123	126	int die_get_varname(Dwarf_Die vr_die, struct strbuf buf);
124		−int die_get_var_range(Dwarf_Die sp_die, Dwarf_Die vr_die, struct strbuf *buf);
125	127
126	128	/* Check if target program is compiled with optimization */
127	129	bool die_is_optimized_target(Dwarf_Die *cu_die);
@@ -130,4 +132,49 @@ bool die_is_optimized_target(Dwarf_Die *cu_die); · 2 unmodified lines
130	132	void die_skip_prologue(Dwarf_Die sp_die, Dwarf_Die cu_die,
131	133	Dwarf_Addr *entrypc);
132	134
133		−#endif
	135	+/* Get the list of including scopes */
	136	+int die_get_scopes(Dwarf_Die cu_die, Dwarf_Addr pc, Dwarf_Die *scopes);
	137	+
	138	+#ifdef HAVE_DWARF_GETLOCATIONS_SUPPORT
	139	+
	140	+/* Get byte offset range of given variable DIE */
	141	+int die_get_var_range(Dwarf_Die sp_die, Dwarf_Die vr_die, struct strbuf *buf);
	142	+
	143	+/* Find a variable saved in the 'reg' at given address */
	144	+Dwarf_Die die_find_variable_by_reg(Dwarf_Die sc_die, Dwarf_Addr pc, int reg,
	145	+ Dwarf_Die *die_mem);
	146	+
	147	+/* Find a (global) variable located in the 'addr' */
	148	+Dwarf_Die die_find_variable_by_addr(Dwarf_Die sc_die, Dwarf_Addr pc,
	149	+ Dwarf_Addr addr, Dwarf_Die *die_mem,
	150	+ int *offset);
	151	+
	152	+#else /* HAVE_DWARF_GETLOCATIONS_SUPPORT */
	153	+
	154	+static inline int die_get_var_range(Dwarf_Die *sp_die __maybe_unused,
	155	+ Dwarf_Die *vr_die __maybe_unused,
	156	+ struct strbuf *buf __maybe_unused)
	157	+{
	158	+ return -ENOTSUP;
	159	+}
	160	+
	161	+static inline Dwarf_Die die_find_variable_by_reg(Dwarf_Die sc_die __maybe_unused,
	162	+ Dwarf_Addr pc __maybe_unused,
	163	+ int reg __maybe_unused,
	164	+ Dwarf_Die *die_mem __maybe_unused)
	165	+{
	166	+ return NULL;
	167	+}
	168	+
	169	+static inline Dwarf_Die die_find_variable_by_addr(Dwarf_Die sc_die __maybe_unused,
	170	+ Dwarf_Addr pc __maybe_unused,
	171	+ Dwarf_Addr addr __maybe_unused,
	172	+ Dwarf_Die *die_mem __maybe_unused,
	173	+ int *offset __maybe_unused)
	174	+{
	175	+ return NULL;
	176	+}
	177	+
	178	+#endif /* HAVE_DWARF_GETLOCATIONS_SUPPORT */
	179	+
	180	+#endif /* _DWARF_AUX_H */

tools/perf/util/dwarf-regs.c+34−0

@@ -5,9 +5,12 @@ · 4 unmodified lines
5	5	* Written by: Masami Hiramatsu <mhiramat@kernel.org>
6	6	*/
7	7
	8	+#include <stdlib.h>
	9	+#include <string.h>
8	10	#include <debug.h>
9	11	#include <dwarf-regs.h>
10	12	#include <elf.h>
	13	+#include <errno.h>
11	14	#include <linux/kernel.h>
12	15
13	16	#ifndef EM_AARCH64
@@ -68,3 +71,34 @@ const char *get_dwarf_regstr(unsigned int n, unsigned int machine) · 54 unmodified lines
68	71	}
69	72	return NULL;
70	73	}
	74	+
	75	+__weak int get_arch_regnum(const char *name __maybe_unused)
	76	+{
	77	+ return -ENOTSUP;
	78	+}
	79	+
	80	+/* Return DWARF register number from architecture register name */
	81	+int get_dwarf_regnum(const char *name, unsigned int machine)
	82	+{
	83	+ char *regname = strdup(name);
	84	+ int reg = -1;
	85	+ char *p;
	86	+
	87	+ if (regname == NULL)
	88	+ return -EINVAL;
	89	+
	90	+ /* For convenience, remove trailing characters */
	91	+ p = strpbrk(regname, " ,)");
	92	+ if (p)
	93	+ *p = '\0';
	94	+
	95	+ switch (machine) {
	96	+ case EM_NONE: /* Generic arch - use host arch */
	97	+ reg = get_arch_regnum(regname);
	98	+ break;
	99	+ default:
	100	+ pr_err("ELF MACHINE %x is not supported.\n", machine);
	101	+ }
	102	+ free(regname);
	103	+ return reg;
	104	+}

tools/perf/util/env.c+64−18

@@ -3,6 +3,7 @@ · 2 unmodified lines
3	3	#include "debug.h"
4	4	#include "env.h"
5	5	#include "util/header.h"
	6	+#include "linux/compiler.h"
6	7	#include <linux/ctype.h>
7	8	#include <linux/zalloc.h>
8	9	#include "cgroup.h"
@@ -12,6 +13,7 @@ · 3 unmodified lines
12	13	#include <string.h>
13	14	#include "pmus.h"
14	15	#include "strbuf.h"
	16	+#include "trace/beauty/beauty.h"
15	17
16	18	struct perf_env perf_env;
17	19
@@ -22,13 +24,19 @@ struct perf_env perf_env; · 4 unmodified lines
22	24
23	25	void perf_env__insert_bpf_prog_info(struct perf_env *env,
24	26	struct bpf_prog_info_node *info_node)
	27	+{
	28	+ down_write(&env->bpf_progs.lock);
	29	+ __perf_env__insert_bpf_prog_info(env, info_node);
	30	+ up_write(&env->bpf_progs.lock);
	31	+}
	32	+
	33	+void __perf_env__insert_bpf_prog_info(struct perf_env env, struct bpf_prog_info_node info_node)
25	34	{
26	35	__u32 prog_id = info_node->info_linear->info.id;
27	36	struct bpf_prog_info_node *node;
28	37	struct rb_node *parent = NULL;
29	38	struct rb_node **p;
30	39
31		− down_write(&env->bpf_progs.lock);
32	40	p = &env->bpf_progs.infos.rb_node;
33	41
34	42	while (*p != NULL) {
@@ -40,15 +48,13 @@ void perf_env__insert_bpf_prog_info(struct perf_env *env, · 5 unmodified lines
40	48	p = &(*p)->rb_right;
41	49	} else {
42	50	pr_debug("duplicated bpf prog info %u\n", prog_id);
43		− goto out;
	51	+ return;
44	52	}
45	53	}
46	54
47	55	rb_link_node(&info_node->rb_node, parent, p);
48	56	rb_insert_color(&info_node->rb_node, &env->bpf_progs.infos);
49	57	env->bpf_progs.infos_cnt++;
50		−out:
51		− up_write(&env->bpf_progs.lock);
52	58	}
53	59
54	60	struct bpf_prog_info_node perf_env__find_bpf_prog_info(struct perf_env env,
@@ -77,14 +83,22 @@ out: · 22 unmodified lines
77	83	}
78	84
79	85	bool perf_env__insert_btf(struct perf_env env, struct btf_node btf_node)
	86	+{
	87	+ bool ret;
	88	+
	89	+ down_write(&env->bpf_progs.lock);
	90	+ ret = __perf_env__insert_btf(env, btf_node);
	91	+ up_write(&env->bpf_progs.lock);
	92	+ return ret;
	93	+}
	94	+
	95	+bool __perf_env__insert_btf(struct perf_env env, struct btf_node btf_node)
80	96	{
81	97	struct rb_node *parent = NULL;
82	98	__u32 btf_id = btf_node->id;
83	99	struct btf_node *node;
84	100	struct rb_node **p;
85		− bool ret = true;
86	101
87		− down_write(&env->bpf_progs.lock);
88	102	p = &env->bpf_progs.btfs.rb_node;
89	103
90	104	while (*p != NULL) {
@@ -96,25 +110,31 @@ bool perf_env__insert_btf(struct perf_env env, struct btf_node btf_node) · 5 unmodified lines
96	110	p = &(*p)->rb_right;
97	111	} else {
98	112	pr_debug("duplicated btf %u\n", btf_id);
99		− ret = false;
100		− goto out;
	113	+ return false;
101	114	}
102	115	}
103	116
104	117	rb_link_node(&btf_node->rb_node, parent, p);
105	118	rb_insert_color(&btf_node->rb_node, &env->bpf_progs.btfs);
106	119	env->bpf_progs.btfs_cnt++;
107		−out:
108		− up_write(&env->bpf_progs.lock);
109		− return ret;
	120	+ return true;
110	121	}
111	122
112	123	struct btf_node perf_env__find_btf(struct perf_env env, __u32 btf_id)
	124	+{
	125	+ struct btf_node *res;
	126	+
	127	+ down_read(&env->bpf_progs.lock);
	128	+ res = __perf_env__find_btf(env, btf_id);
	129	+ up_read(&env->bpf_progs.lock);
	130	+ return res;
	131	+}
	132	+
	133	+struct btf_node __perf_env__find_btf(struct perf_env env, __u32 btf_id)
113	134	{
114	135	struct btf_node *node = NULL;
115	136	struct rb_node *n;
116	137
117		− down_read(&env->bpf_progs.lock);
118	138	n = env->bpf_progs.btfs.rb_node;
119	139
120	140	while (n) {
@@ -124,13 +144,9 @@ struct btf_node perf_env__find_btf(struct perf_env env, __u32 btf_id) · 3 unmodified lines
124	144	else if (btf_id > node->id)
125	145	n = n->rb_right;
126	146	else
127		− goto out;
	147	+ return node;
128	148	}
129		− node = NULL;
130		−
131		−out:
132		− up_read(&env->bpf_progs.lock);
133		− return node;
	149	+ return NULL;
134	150	}
135	151
136	152	/* purge data in bpf_progs.infos tree */
@@ -453,6 +469,18 @@ const char perf_env__arch(struct perf_env env) · 316 unmodified lines
453	469	return normalize_arch(arch_name);
454	470	}
455	471
	472	+const char perf_env__arch_strerrno(struct perf_env env __maybe_unused, int err __maybe_unused)
	473	+{
	474	+#if defined(HAVE_SYSCALL_TABLE_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
	475	+ if (env->arch_strerrno == NULL)
	476	+ env->arch_strerrno = arch_syscalls__strerrno_function(perf_env__arch(env));
	477	+
	478	+ return env->arch_strerrno ? env->arch_strerrno(err) : "no arch specific strerrno function";
	479	+#else
	480	+ return "!(HAVE_SYSCALL_TABLE_SUPPORT && HAVE_LIBTRACEEVENT)";
	481	+#endif
	482	+}
	483	+
456	484	const char perf_env__cpuid(struct perf_env env)
457	485	{
458	486	int status;
@@ -531,6 +559,24 @@ int perf_env__numa_node(struct perf_env *env, struct perf_cpu cpu) · 72 unmodified lines
531	559	return cpu.cpu >= 0 && cpu.cpu < env->nr_numa_map ? env->numa_map[cpu.cpu] : -1;
532	560	}
533	561
	562	+bool perf_env__has_pmu_mapping(struct perf_env env, const char pmu_name)
	563	+{
	564	+ char pmu_mapping = env->pmu_mappings, colon;
	565	+
	566	+ for (int i = 0; i < env->nr_pmu_mappings; ++i) {
	567	+ if (strtoul(pmu_mapping, &colon, 0) == ULONG_MAX \|\| *colon != ':')
	568	+ goto out_error;
	569	+
	570	+ pmu_mapping = colon + 1;
	571	+ if (strcmp(pmu_mapping, pmu_name) == 0)
	572	+ return true;
	573	+
	574	+ pmu_mapping += strlen(pmu_mapping) + 1;
	575	+ }
	576	+out_error:
	577	+ return false;
	578	+}
	579	+
534	580	char perf_env__find_pmu_cap(struct perf_env env, const char *pmu_name,
535	581	const char *cap)
536	582	{

tools/perf/util/env.h+17−0

@@ -46,10 +46,17 @@ struct hybrid_node { · 45 unmodified lines
46	46	struct pmu_caps {
47	47	int nr_caps;
48	48	unsigned int max_branches;
	49	+ unsigned int br_cntr_nr;
	50	+ unsigned int br_cntr_width;
	51	+
49	52	char **caps;
50	53	char *pmu_name;
51	54	};
52	55
	56	+typedef const char *(arch_syscalls__strerrno_t)(int err);
	57	+
	58	+arch_syscalls__strerrno_t arch_syscalls__strerrno_function(const char arch);
	59	+
53	60	struct perf_env {
54	61	char *hostname;
55	62	char *os_release;
@@ -62,6 +69,8 @@ struct perf_env { · 6 unmodified lines
62	69	unsigned long long total_mem;
63	70	unsigned int msr_pmu_type;
64	71	unsigned int max_branches;
	72	+ unsigned int br_cntr_nr;
	73	+ unsigned int br_cntr_width;
65	74	int kernel_is_64_bit;
66	75
67	76	int nr_cmdline;
@@ -130,6 +139,7 @@ struct perf_env { · 62 unmodified lines
130	139	*/
131	140	bool enabled;
132	141	} clock;
	142	+ arch_syscalls__strerrno_t *arch_strerrno;
133	143	};
134	144
135	145	enum perf_compress_type {
@@ -159,19 +169,26 @@ int perf_env__read_cpu_topology_map(struct perf_env *env); · 23 unmodified lines
159	169	void cpu_cache_level__free(struct cpu_cache_level *cache);
160	170
161	171	const char perf_env__arch(struct perf_env env);
	172	+const char perf_env__arch_strerrno(struct perf_env env, int err);
162	173	const char perf_env__cpuid(struct perf_env env);
163	174	const char perf_env__raw_arch(struct perf_env env);
164	175	int perf_env__nr_cpus_avail(struct perf_env *env);
165	176
166	177	void perf_env__init(struct perf_env *env);
	178	+void __perf_env__insert_bpf_prog_info(struct perf_env *env,
	179	+ struct bpf_prog_info_node *info_node);
167	180	void perf_env__insert_bpf_prog_info(struct perf_env *env,
168	181	struct bpf_prog_info_node *info_node);
169	182	struct bpf_prog_info_node perf_env__find_bpf_prog_info(struct perf_env env,
170	183	__u32 prog_id);
171	184	bool perf_env__insert_btf(struct perf_env env, struct btf_node btf_node);
	185	+bool __perf_env__insert_btf(struct perf_env env, struct btf_node btf_node);
172	186	struct btf_node perf_env__find_btf(struct perf_env env, __u32 btf_id);
	187	+struct btf_node __perf_env__find_btf(struct perf_env env, __u32 btf_id);
173	188
174	189	int perf_env__numa_node(struct perf_env *env, struct perf_cpu cpu);
175	190	char perf_env__find_pmu_cap(struct perf_env env, const char *pmu_name,
176	191	const char *cap);
	192	+
	193	+bool perf_env__has_pmu_mapping(struct perf_env env, const char pmu_name);
177	194	#endif /* __PERF_ENV_H */

tools/perf/util/event.c+2−2

@@ -617,13 +617,13 @@ struct map thread__find_map(struct thread thread, u8 cpumode, u64 addr, · 616 unmodified lines
617	617	if (cpumode == PERF_RECORD_MISC_KERNEL && perf_host) {
618	618	al->level = 'k';
619	619	maps = machine__kernel_maps(machine);
620		− load_map = true;
	620	+ load_map = !symbol_conf.lazy_load_kernel_maps;
621	621	} else if (cpumode == PERF_RECORD_MISC_USER && perf_host) {
622	622	al->level = '.';
623	623	} else if (cpumode == PERF_RECORD_MISC_GUEST_KERNEL && perf_guest) {
624	624	al->level = 'g';
625	625	maps = machine__kernel_maps(machine);
626		− load_map = true;
	626	+ load_map = !symbol_conf.lazy_load_kernel_maps;
627	627	} else if (cpumode == PERF_RECORD_MISC_GUEST_USER && perf_guest) {
628	628	al->level = 'u';
629	629	} else {

tools/perf/util/evlist.c+32−2

@@ -1056,7 +1056,7 @@ int evlist__create_maps(struct evlist evlist, struct target target) · 1055 unmodified lines
1056	1056	return -1;
1057	1057
1058	1058	if (target__uses_dummy_map(target))
1059		− cpus = perf_cpu_map__dummy_new();
	1059	+ cpus = perf_cpu_map__new_any_cpu();
1060	1060	else
1061	1061	cpus = perf_cpu_map__new(target->cpu_list);
1062	1062
@@ -1352,7 +1352,7 @@ static int evlist__create_syswide_maps(struct evlist *evlist) · 289 unmodified lines
1352	1352	* error, and we may not want to do that fallback to a
1353	1353	* default cpu identity map :-\
1354	1354	*/
1355		− cpus = perf_cpu_map__new(NULL);
	1355	+ cpus = perf_cpu_map__new_online_cpus();
1356	1356	if (!cpus)
1357	1357	goto out;
1358	1358
@@ -2518,3 +2518,33 @@ void evlist__warn_user_requested_cpus(struct evlist evlist, const char cpu_lis · 1159 unmodified lines
2518	2518	}
2519	2519	perf_cpu_map__put(user_requested_cpus);
2520	2520	}
	2521	+
	2522	+void evlist__uniquify_name(struct evlist *evlist)
	2523	+{
	2524	+ char new_name, empty_attributes[2] = ":", attributes;
	2525	+ struct evsel *pos;
	2526	+
	2527	+ if (perf_pmus__num_core_pmus() == 1)
	2528	+ return;
	2529	+
	2530	+ evlist__for_each_entry(evlist, pos) {
	2531	+ if (!evsel__is_hybrid(pos))
	2532	+ continue;
	2533	+
	2534	+ if (strchr(pos->name, '/'))
	2535	+ continue;
	2536	+
	2537	+ attributes = strchr(pos->name, ':');
	2538	+ if (attributes)
	2539	+ *attributes = '\0';
	2540	+ else
	2541	+ attributes = empty_attributes;
	2542	+
	2543	+ if (asprintf(&new_name, "%s/%s/%s", pos->pmu_name, pos->name, attributes + 1)) {
	2544	+ free(pos->name);
	2545	+ pos->name = new_name;
	2546	+ } else {
	2547	+ *attributes = ':';
	2548	+ }
	2549	+ }
	2550	+}

tools/perf/util/evlist.h+1−0

@@ -442,5 +442,6 @@ struct evsel evlist__find_evsel(struct evlist evlist, int idx); · 441 unmodified lines
442	442	int evlist__scnprintf_evsels(struct evlist evlist, size_t size, char bf);
443	443	void evlist__check_mem_load_aux(struct evlist *evlist);
444	444	void evlist__warn_user_requested_cpus(struct evlist evlist, const char cpu_list);
	445	+void evlist__uniquify_name(struct evlist *evlist);
445	446
446	447	#endif /* __PERF_EVLIST_H */

tools/perf/util/evsel.c+45−10

@@ -1801,7 +1801,7 @@ static int __evsel__prepare_open(struct evsel evsel, struct perf_cpu_map cpus, · 1800 unmodified lines
1801	1801
1802	1802	if (cpus == NULL) {
1803	1803	if (empty_cpu_map == NULL) {
1804		− empty_cpu_map = perf_cpu_map__dummy_new();
	1804	+ empty_cpu_map = perf_cpu_map__new_any_cpu();
1805	1805	if (empty_cpu_map == NULL)
1806	1806	return -ENOMEM;
1807	1807	}
@@ -1832,6 +1832,8 @@ static int __evsel__prepare_open(struct evsel evsel, struct perf_cpu_map cpus, · 24 unmodified lines
1832	1832
1833	1833	static void evsel__disable_missing_features(struct evsel *evsel)
1834	1834	{
	1835	+ if (perf_missing_features.branch_counters)
	1836	+ evsel->core.attr.branch_sample_type &= ~PERF_SAMPLE_BRANCH_COUNTERS;
1835	1837	if (perf_missing_features.read_lost)
1836	1838	evsel->core.attr.read_format &= ~PERF_FORMAT_LOST;
1837	1839	if (perf_missing_features.weight_struct) {
@@ -1885,7 +1887,12 @@ bool evsel__detect_missing_features(struct evsel *evsel) · 47 unmodified lines
1885	1887	* Must probe features in the order they were added to the
1886	1888	* perf_event_attr interface.
1887	1889	*/
1888		− if (!perf_missing_features.read_lost &&
	1890	+ if (!perf_missing_features.branch_counters &&
	1891	+ (evsel->core.attr.branch_sample_type & PERF_SAMPLE_BRANCH_COUNTERS)) {
	1892	+ perf_missing_features.branch_counters = true;
	1893	+ pr_debug2("switching off branch counters support\n");
	1894	+ return true;
	1895	+ } else if (!perf_missing_features.read_lost &&
1889	1896	(evsel->core.attr.read_format & PERF_FORMAT_LOST)) {
1890	1897	perf_missing_features.read_lost = true;
1891	1898	pr_debug2("switching off PERF_FORMAT_LOST support\n");
@@ -2318,6 +2325,22 @@ u64 evsel__bitfield_swap_branch_flags(u64 value) · 426 unmodified lines
2318	2325	return new_val;
2319	2326	}
2320	2327
	2328	+static inline bool evsel__has_branch_counters(const struct evsel *evsel)
	2329	+{
	2330	+ struct evsel cur, leader = evsel__leader(evsel);
	2331	+
	2332	+ /* The branch counters feature only supports group */
	2333	+ if (!leader \|\| !evsel->evlist)
	2334	+ return false;
	2335	+
	2336	+ evlist__for_each_entry(evsel->evlist, cur) {
	2337	+ if ((leader == evsel__leader(cur)) &&
	2338	+ (cur->core.attr.branch_sample_type & PERF_SAMPLE_BRANCH_COUNTERS))
	2339	+ return true;
	2340	+ }
	2341	+ return false;
	2342	+}
	2343	+
2321	2344	int evsel__parse_sample(struct evsel evsel, union perf_event event,
2322	2345	struct perf_sample *data)
2323	2346	{
@@ -2551,6 +2574,16 @@ int evsel__parse_sample(struct evsel evsel, union perf_event event, · 227 unmodified lines
2551	2574
2552	2575	OVERFLOW_CHECK(array, sz, max_size);
2553	2576	array = (void *)array + sz;
	2577	+
	2578	+ if (evsel__has_branch_counters(evsel)) {
	2579	+ OVERFLOW_CHECK_u64(array);
	2580	+
	2581	+ data->branch_stack_cntr = (u64 *)array;
	2582	+ sz = data->branch_stack->nr * sizeof(u64);
	2583	+
	2584	+ OVERFLOW_CHECK(array, sz, max_size);
	2585	+ array = (void *)array + sz;
	2586	+ }
2554	2587	}
2555	2588
2556	2589	if (type & PERF_SAMPLE_REGS_USER) {
@@ -2820,7 +2853,8 @@ u64 evsel__intval_common(struct evsel evsel, struct perf_sample sample, const · 263 unmodified lines
2820	2853
2821	2854	#endif
2822	2855
2823		−bool evsel__fallback(struct evsel evsel, int err, char msg, size_t msgsize)
	2856	+bool evsel__fallback(struct evsel evsel, struct target target, int err,
	2857	+ char *msg, size_t msgsize)
2824	2858	{
2825	2859	int paranoid;
2826	2860
@@ -2828,18 +2862,19 @@ bool evsel__fallback(struct evsel evsel, int err, char msg, size_t msgsize) · 1 unmodified line
2828	2862	evsel->core.attr.type == PERF_TYPE_HARDWARE &&
2829	2863	evsel->core.attr.config == PERF_COUNT_HW_CPU_CYCLES) {
2830	2864	/*
2831		− * If it's cycles then fall back to hrtimer based
2832		− * cpu-clock-tick sw counter, which is always available even if
2833		− * no PMU support.
	2865	+ * If it's cycles then fall back to hrtimer based cpu-clock sw
	2866	+ * counter, which is always available even if no PMU support.
2834	2867	*
2835	2868	* PPC returns ENXIO until 2.6.37 (behavior changed with commit
2836	2869	* b0a873e).
2837	2870	*/
2838		− scnprintf(msg, msgsize, "%s",
2839		−"The cycles event is not supported, trying to fall back to cpu-clock-ticks");
2840		−
2841	2871	evsel->core.attr.type = PERF_TYPE_SOFTWARE;
2842		− evsel->core.attr.config = PERF_COUNT_SW_CPU_CLOCK;
	2872	+ evsel->core.attr.config = target__has_cpu(target)
	2873	+ ? PERF_COUNT_SW_CPU_CLOCK
	2874	+ : PERF_COUNT_SW_TASK_CLOCK;
	2875	+ scnprintf(msg, msgsize,
	2876	+ "The cycles event is not supported, trying to fall back to %s",
	2877	+ target__has_cpu(target) ? "cpu-clock" : "task-clock");
2843	2878
2844	2879	zfree(&evsel->name);
2845	2880	return true;

tools/perf/util/evsel.h+3−1

@@ -191,6 +191,7 @@ struct perf_missing_features { · 190 unmodified lines
191	191	bool code_page_size;
192	192	bool weight_struct;
193	193	bool read_lost;
	194	+ bool branch_counters;
194	195	};
195	196
196	197	extern struct perf_missing_features perf_missing_features;
@@ -459,7 +460,8 @@ static inline bool evsel__is_clock(const struct evsel *evsel) · 262 unmodified lines
459	460	evsel__match(evsel, SOFTWARE, SW_TASK_CLOCK);
460	461	}
461	462
462		−bool evsel__fallback(struct evsel evsel, int err, char msg, size_t msgsize);
	463	+bool evsel__fallback(struct evsel evsel, struct target target, int err,
	464	+ char *msg, size_t msgsize);
463	465	int evsel__open_strerror(struct evsel evsel, struct target target,
464	466	int err, char *msg, size_t size);
465	467

tools/perf/util/genelf.c+3−3

@@ -293,9 +293,9 @@ jit_write_elf(int fd, uint64_t load_addr, const char *sym, · 292 unmodified lines
293	293	*/
294	294	phdr = elf_newphdr(e, 1);
295	295	phdr[0].p_type = PT_LOAD;
296		− phdr[0].p_offset = 0;
297		− phdr[0].p_vaddr = 0;
298		− phdr[0].p_paddr = 0;
	296	+ phdr[0].p_offset = GEN_ELF_TEXT_OFFSET;
	297	+ phdr[0].p_vaddr = GEN_ELF_TEXT_OFFSET;
	298	+ phdr[0].p_paddr = GEN_ELF_TEXT_OFFSET;
299	299	phdr[0].p_filesz = csize;
300	300	phdr[0].p_memsz = csize;
301	301	phdr[0].p_flags = PF_X \| PF_R;

tools/perf/util/header.c+33−10

@@ -1444,7 +1444,9 @@ static int build_mem_topology(struct memory_node *nodesp, u64 cntp) · 1443 unmodified lines
1444	1444	nodes = new_nodes;
1445	1445	size += 4;
1446	1446	}
1447		− ret = memory_node__read(&nodes[cnt++], idx);
	1447	+ ret = memory_node__read(&nodes[cnt], idx);
	1448	+ if (!ret)
	1449	+ cnt += 1;
1448	1450	}
1449	1451	out:
1450	1452	closedir(dir);
@@ -1847,8 +1849,8 @@ static void print_bpf_prog_info(struct feat_fd ff, FILE fp) · 396 unmodified lines
1847	1849	node = rb_entry(next, struct bpf_prog_info_node, rb_node);
1848	1850	next = rb_next(&node->rb_node);
1849	1851
1850		− bpf_event__print_bpf_prog_info(&node->info_linear->info,
1851		− env, fp);
	1852	+ __bpf_event__print_bpf_prog_info(&node->info_linear->info,
	1853	+ env, fp);
1852	1854	}
1853	1855
1854	1856	up_read(&env->bpf_progs.lock);
@@ -2145,6 +2147,14 @@ static void print_pmu_caps(struct feat_fd ff, FILE fp) · 290 unmodified lines
2145	2147	__print_pmu_caps(fp, pmu_caps->nr_caps, pmu_caps->caps,
2146	2148	pmu_caps->pmu_name);
2147	2149	}
	2150	+
	2151	+ if (strcmp(perf_env__arch(&ff->ph->env), "x86") == 0 &&
	2152	+ perf_env__has_pmu_mapping(&ff->ph->env, "ibs_op")) {
	2153	+ char *max_precise = perf_env__find_pmu_cap(&ff->ph->env, "cpu", "max_precise");
	2154	+
	2155	+ if (max_precise != NULL && atoi(max_precise) == 0)
	2156	+ fprintf(fp, "# AMD systems uses ibs_op// PMU for some precise events, e.g.: cycles:p, see the 'perf list' man page for further details.\n");
	2157	+ }
2148	2158	}
2149	2159
2150	2160	static void print_pmu_mappings(struct feat_fd ff, FILE fp)
@@ -3178,7 +3188,7 @@ static int process_bpf_prog_info(struct feat_fd ff, void data __maybe_unused) · 1027 unmodified lines
3178	3188	/* after reading from file, translate offset to address */
3179	3189	bpil_offs_to_addr(info_linear);
3180	3190	info_node->info_linear = info_linear;
3181		− perf_env__insert_bpf_prog_info(env, info_node);
	3191	+ __perf_env__insert_bpf_prog_info(env, info_node);
3182	3192	}
3183	3193
3184	3194	up_write(&env->bpf_progs.lock);
@@ -3225,7 +3235,7 @@ static int process_bpf_btf(struct feat_fd ff, void data __maybe_unused) · 40 unmodified lines
3225	3235	if (__do_read(ff, node->data, data_size))
3226	3236	goto out;
3227	3237
3228		− perf_env__insert_btf(env, node);
	3238	+ __perf_env__insert_btf(env, node);
3229	3239	node = NULL;
3230	3240	}
3231	3241
@@ -3259,7 +3269,9 @@ static int process_compressed(struct feat_fd *ff, · 27 unmodified lines
3259	3269	}
3260	3270
3261	3271	static int __process_pmu_caps(struct feat_fd ff, int nr_caps,
3262		− char **caps, unsigned int max_branches)
	3272	+ char **caps, unsigned int max_branches,
	3273	+ unsigned int *br_cntr_nr,
	3274	+ unsigned int *br_cntr_width)
3263	3275	{
3264	3276	char name, value, *ptr;
3265	3277	u32 nr_pmu_caps, i;
@@ -3294,6 +3306,12 @@ static int __process_pmu_caps(struct feat_fd ff, int nr_caps, · 28 unmodified lines
3294	3306	if (!strcmp(name, "branches"))
3295	3307	*max_branches = atoi(value);
3296	3308
	3309	+ if (!strcmp(name, "branch_counter_nr"))
	3310	+ *br_cntr_nr = atoi(value);
	3311	+
	3312	+ if (!strcmp(name, "branch_counter_width"))
	3313	+ *br_cntr_width = atoi(value);
	3314	+
3297	3315	free(value);
3298	3316	free(name);
3299	3317	}
@@ -3318,7 +3336,9 @@ static int process_cpu_pmu_caps(struct feat_fd *ff, · 18 unmodified lines
3318	3336	{
3319	3337	int ret = __process_pmu_caps(ff, &ff->ph->env.nr_cpu_pmu_caps,
3320	3338	&ff->ph->env.cpu_pmu_caps,
3321		− &ff->ph->env.max_branches);
	3339	+ &ff->ph->env.max_branches,
	3340	+ &ff->ph->env.br_cntr_nr,
	3341	+ &ff->ph->env.br_cntr_width);
3322	3342
3323	3343	if (!ret && !ff->ph->env.cpu_pmu_caps)
3324	3344	pr_debug("cpu pmu capabilities not available\n");
@@ -3347,7 +3367,9 @@ static int process_pmu_caps(struct feat_fd ff, void data __maybe_unused) · 22 unmodified lines
3347	3367	for (i = 0; i < nr_pmu; i++) {
3348	3368	ret = __process_pmu_caps(ff, &pmu_caps[i].nr_caps,
3349	3369	&pmu_caps[i].caps,
3350		− &pmu_caps[i].max_branches);
	3370	+ &pmu_caps[i].max_branches,
	3371	+ &pmu_caps[i].br_cntr_nr,
	3372	+ &pmu_caps[i].br_cntr_width);
3351	3373	if (ret)
3352	3374	goto err;
3353	3375
@@ -4369,9 +4391,10 @@ size_t perf_event__fprintf_event_update(union perf_event event, FILE fp) · 1015 unmodified lines
4369	4391	ret += fprintf(fp, "... ");
4370	4392
4371	4393	map = cpu_map__new_data(&ev->cpus.cpus);
4372		− if (map)
	4394	+ if (map) {
4373	4395	ret += cpu_map__fprintf(map, fp);
4374		− else
	4396	+ perf_cpu_map__put(map);
	4397	+ } else
4375	4398	ret += fprintf(fp, "failed to get cpus\n");
4376	4399	break;
4377	4400	default:

tools/perf/util/hisi-ptt.c+1−0

@@ -123,6 +123,7 @@ static int hisi_ptt_process_auxtrace_event(struct perf_session *session, · 122 unmodified lines
123	123	if (dump_trace)
124	124	hisi_ptt_dump_event(ptt, data, size);
125	125
	126	+ free(data);
126	127	return 0;
127	128	}
128	129

tools/perf/util/hist.h+11−17

@@ -82,6 +82,9 @@ enum hist_column { · 81 unmodified lines
82	82	HISTC_ADDR_TO,
83	83	HISTC_ADDR,
84	84	HISTC_SIMD,
	85	+ HISTC_TYPE,
	86	+ HISTC_TYPE_OFFSET,
	87	+ HISTC_SYMBOL_OFFSET,
85	88	HISTC_NR_COLS, /* Last entry */
86	89	};
87	90
@@ -457,7 +460,6 @@ struct hist_browser_timer { · 369 unmodified lines
457	460	int refresh;
458	461	};
459	462
460		−struct annotation_options;
461	463	struct res_sample;
462	464
463	465	enum rstype {
@@ -473,16 +475,13 @@ struct block_hist; · 9 unmodified lines
473	475	void attr_to_script(char buf, struct perf_event_attr attr);
474	476
475	477	int map_symbol__tui_annotate(struct map_symbol ms, struct evsel evsel,
476		− struct hist_browser_timer *hbt,
477		− struct annotation_options *annotation_opts);
	478	+ struct hist_browser_timer *hbt);
478	479
479	480	int hist_entry__tui_annotate(struct hist_entry he, struct evsel evsel,
480		− struct hist_browser_timer *hbt,
481		− struct annotation_options *annotation_opts);
	481	+ struct hist_browser_timer *hbt);
482	482
483	483	int evlist__tui_browse_hists(struct evlist evlist, const char help, struct hist_browser_timer *hbt,
484		− float min_pcnt, struct perf_env *env, bool warn_lost_event,
485		− struct annotation_options *annotation_options);
	484	+ float min_pcnt, struct perf_env *env, bool warn_lost_event);
486	485
487	486	int script_browse(const char script_opt, struct evsel evsel);
488	487
@@ -492,8 +491,7 @@ int res_sample_browse(struct res_sample *res_samples, int num_res, · 3 unmodified lines
492	491	void res_sample_init(void);
493	492
494	493	int block_hists_tui_browse(struct block_hist bh, struct evsel evsel,
495		− float min_percent, struct perf_env *env,
496		− struct annotation_options *annotation_opts);
	494	+ float min_percent, struct perf_env *env);
497	495	#else
498	496	static inline
499	497	int evlist__tui_browse_hists(struct evlist *evlist __maybe_unused,
@@ -501,23 +499,20 @@ int evlist__tui_browse_hists(struct evlist *evlist __maybe_unused, · 1 unmodified line
501	499	struct hist_browser_timer *hbt __maybe_unused,
502	500	float min_pcnt __maybe_unused,
503	501	struct perf_env *env __maybe_unused,
504		− bool warn_lost_event __maybe_unused,
505		− struct annotation_options *annotation_options __maybe_unused)
	502	+ bool warn_lost_event __maybe_unused)
506	503	{
507	504	return 0;
508	505	}
509	506	static inline int map_symbol__tui_annotate(struct map_symbol *ms __maybe_unused,
510	507	struct evsel *evsel __maybe_unused,
511		− struct hist_browser_timer *hbt __maybe_unused,
512		− struct annotation_options *annotation_options __maybe_unused)
	508	+ struct hist_browser_timer *hbt __maybe_unused)
513	509	{
514	510	return 0;
515	511	}
516	512
517	513	static inline int hist_entry__tui_annotate(struct hist_entry *he __maybe_unused,
518	514	struct evsel *evsel __maybe_unused,
519		− struct hist_browser_timer *hbt __maybe_unused,
520		− struct annotation_options *annotation_opts __maybe_unused)
	515	+ struct hist_browser_timer *hbt __maybe_unused)
521	516	{
522	517	return 0;
523	518	}
@@ -541,8 +536,7 @@ static inline void res_sample_init(void) {} · 17 unmodified lines
541	536	static inline int block_hists_tui_browse(struct block_hist *bh __maybe_unused,
542	537	struct evsel *evsel __maybe_unused,
543	538	float min_percent __maybe_unused,
544		− struct perf_env *env __maybe_unused,
545		− struct annotation_options *annotation_opts __maybe_unused)
	539	+ struct perf_env *env __maybe_unused)
546	540	{
547	541	return 0;
548	542	}

tools/perf/util/include/dwarf-regs.h+19−0

@@ -2,6 +2,9 @@ · 1 unmodified line
2	2	#ifndef _PERF_DWARF_REGS_H_
3	3	#define _PERF_DWARF_REGS_H_
4	4
	5	+#define DWARF_REG_PC 0xd3af9c /* random number */
	6	+#define DWARF_REG_FB 0xd3affb /* random number */
	7	+
5	8	#ifdef HAVE_DWARF_SUPPORT
6	9	const char *get_arch_regstr(unsigned int n);
7	10	/*
@@ -10,6 +13,22 @@ const char *get_arch_regstr(unsigned int n); · 2 unmodified lines
10	13	* machine: ELF machine signature (EM_*)
11	14	*/
12	15	const char *get_dwarf_regstr(unsigned int n, unsigned int machine);
	16	+
	17	+int get_arch_regnum(const char *name);
	18	+/*
	19	+ * get_dwarf_regnum - Returns DWARF regnum from register name
	20	+ * name: architecture register name
	21	+ * machine: ELF machine signature (EM_*)
	22	+ */
	23	+int get_dwarf_regnum(const char *name, unsigned int machine);
	24	+
	25	+#else /* HAVE_DWARF_SUPPORT */
	26	+
	27	+static inline int get_dwarf_regnum(const char *name __maybe_unused,
	28	+ unsigned int machine __maybe_unused)
	29	+{
	30	+ return -1;
	31	+}
13	32	#endif
14	33
15	34	#ifdef HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET

tools/perf/util/machine.c+41−34

@@ -453,7 +453,7 @@ static struct thread findnew_guest_code(struct machine machine, · 452 unmodified lines
453	453	* Guest code can be found in hypervisor process at the same address
454	454	* so copy host maps.
455	455	*/
456		− err = maps__clone(thread, thread__maps(host_thread));
	456	+ err = maps__copy_from(thread__maps(thread), thread__maps(host_thread));
457	457	thread__put(host_thread);
458	458	if (err)
459	459	goto out_err;
@@ -1285,33 +1285,46 @@ static u64 find_entry_trampoline(struct dso *dso) · 825 unmodified lines
1285	1285	#define X86_64_CPU_ENTRY_AREA_SIZE 0x2c000
1286	1286	#define X86_64_ENTRY_TRAMPOLINE 0x6000
1287	1287
	1288	+struct machine__map_x86_64_entry_trampolines_args {
	1289	+ struct maps *kmaps;
	1290	+ bool found;
	1291	+};
	1292	+
	1293	+static int machine__map_x86_64_entry_trampolines_cb(struct map map, void data)
	1294	+{
	1295	+ struct machine__map_x86_64_entry_trampolines_args *args = data;
	1296	+ struct map *dest_map;
	1297	+ struct kmap *kmap = __map__kmap(map);
	1298	+
	1299	+ if (!kmap \|\| !is_entry_trampoline(kmap->name))
	1300	+ return 0;
	1301	+
	1302	+ dest_map = maps__find(args->kmaps, map__pgoff(map));
	1303	+ if (dest_map != map)
	1304	+ map__set_pgoff(map, map__map_ip(dest_map, map__pgoff(map)));
	1305	+
	1306	+ args->found = true;
	1307	+ return 0;
	1308	+}
	1309	+
1288	1310	/* Map x86_64 PTI entry trampolines */
1289	1311	int machine__map_x86_64_entry_trampolines(struct machine *machine,
1290	1312	struct dso *kernel)
1291	1313	{
1292		− struct maps *kmaps = machine__kernel_maps(machine);
	1314	+ struct machine__map_x86_64_entry_trampolines_args args = {
	1315	+ .kmaps = machine__kernel_maps(machine),
	1316	+ .found = false,
	1317	+ };
1293	1318	int nr_cpus_avail, cpu;
1294		− bool found = false;
1295		− struct map_rb_node *rb_node;
1296	1319	u64 pgoff;
1297	1320
1298	1321	/*
1299	1322	* In the vmlinux case, pgoff is a virtual address which must now be
1300	1323	* mapped to a vmlinux offset.
1301	1324	*/
1302		− maps__for_each_entry(kmaps, rb_node) {
1303		− struct map dest_map, map = rb_node->map;
1304		− struct kmap *kmap = __map__kmap(map);
1305		−
1306		− if (!kmap \|\| !is_entry_trampoline(kmap->name))
1307		− continue;
	1325	+ maps__for_each_map(args.kmaps, machine__map_x86_64_entry_trampolines_cb, &args);
1308	1326
1309		− dest_map = maps__find(kmaps, map__pgoff(map));
1310		− if (dest_map != map)
1311		− map__set_pgoff(map, map__map_ip(dest_map, map__pgoff(map)));
1312		− found = true;
1313		− }
1314		− if (found \|\| machine->trampolines_mapped)
	1327	+ if (args.found \|\| machine->trampolines_mapped)
1315	1328	return 0;
1316	1329
1317	1330	pgoff = find_entry_trampoline(kernel);
@@ -1359,8 +1372,7 @@ __machine__create_kernel_maps(struct machine machine, struct dso kernel) · 41 unmodified lines
1359	1372	if (machine->vmlinux_map == NULL)
1360	1373	return -ENOMEM;
1361	1374
1362		− map__set_map_ip(machine->vmlinux_map, identity__map_ip);
1363		− map__set_unmap_ip(machine->vmlinux_map, identity__map_ip);
	1375	+ map__set_mapping_type(machine->vmlinux_map, MAPPING_TYPE__IDENTITY);
1364	1376	return maps__insert(machine__kernel_maps(machine), machine->vmlinux_map);
1365	1377	}
1366	1378
@@ -1750,12 +1762,11 @@ int machine__create_kernel_maps(struct machine *machine) · 383 unmodified lines
1750	1762
1751	1763	if (end == ~0ULL) {
1752	1764	/* update end address of the kernel map using adjacent module address */
1753		− struct map_rb_node *rb_node = maps__find_node(machine__kernel_maps(machine),
1754		− machine__kernel_map(machine));
1755		− struct map_rb_node *next = map_rb_node__next(rb_node);
	1765	+ struct map *next = maps__find_next_entry(machine__kernel_maps(machine),
	1766	+ machine__kernel_map(machine));
1756	1767
1757	1768	if (next)
1758		− machine__set_kernel_mmap(machine, start, map__start(next->map));
	1769	+ machine__set_kernel_mmap(machine, start, map__start(next));
1759	1770	}
1760	1771
1761	1772	out_put:
@@ -2157,9 +2168,13 @@ int machine__process_exit_event(struct machine machine, union perf_event event · 395 unmodified lines
2157	2168	if (dump_trace)
2158	2169	perf_event__fprintf_task(event, stdout);
2159	2170
2160		− if (thread != NULL)
2161		− thread__put(thread);
2162		−
	2171	+ if (thread != NULL) {
	2172	+ if (symbol_conf.keep_exited_threads)
	2173	+ thread__set_exited(thread, /exited=/true);
	2174	+ else
	2175	+ machine__remove_thread(machine, thread);
	2176	+ }
	2177	+ thread__put(thread);
2163	2178	return 0;
2164	2179	}
2165	2180
@@ -3395,16 +3410,8 @@ int machine__for_each_dso(struct machine machine, machine__dso_t fn, void priv · 1229 unmodified lines
3395	3410	int machine__for_each_kernel_map(struct machine machine, machine__map_t fn, void priv)
3396	3411	{
3397	3412	struct maps *maps = machine__kernel_maps(machine);
3398		− struct map_rb_node *pos;
3399		− int err = 0;
3400	3413
3401		− maps__for_each_entry(maps, pos) {
3402		− err = fn(pos->map, priv);
3403		− if (err != 0) {
3404		− break;
3405		− }
3406		− }
3407		− return err;
	3414	+ return maps__for_each_map(maps, fn, priv);
3408	3415	}
3409	3416
3410	3417	bool machine__is_lock_function(struct machine *machine, u64 addr)

tools/perf/util/map.c+2−18

@@ -109,8 +109,7 @@ void map__init(struct map map, u64 start, u64 end, u64 pgoff, struct dso dso) · 108 unmodified lines
109	109	map__set_pgoff(map, pgoff);
110	110	map__set_reloc(map, 0);
111	111	map__set_dso(map, dso__get(dso));
112		− map__set_map_ip(map, map__dso_map_ip);
113		− map__set_unmap_ip(map, map__dso_unmap_ip);
	112	+ map__set_mapping_type(map, MAPPING_TYPE__DSO);
114	113	map__set_erange_warned(map, false);
115	114	refcount_set(map__refcnt(map), 1);
116	115	}
@@ -172,7 +171,7 @@ struct map map__new(struct machine machine, u64 start, u64 len, · 55 unmodified lines
172	171	map__init(result, start, start + len, pgoff, dso);
173	172
174	173	if (anon \|\| no_dso) {
175		− map->map_ip = map->unmap_ip = identity__map_ip;
	174	+ map->mapping_type = MAPPING_TYPE__IDENTITY;
176	175
177	176	/*
178	177	* Set memory without DSO as loaded. All map__find_*
@@ -630,18 +629,3 @@ struct maps map__kmaps(struct map map) · 451 unmodified lines
630	629	}
631	630	return kmap->kmaps;
632	631	}
633		−
634		−u64 map__dso_map_ip(const struct map *map, u64 ip)
635		−{
636		− return ip - map__start(map) + map__pgoff(map);
637		−}
638		−
639		−u64 map__dso_unmap_ip(const struct map *map, u64 ip)
640		−{
641		− return ip + map__start(map) - map__pgoff(map);
642		−}
643		−
644		−u64 identity__map_ip(const struct map *map __maybe_unused, u64 ip)
645		−{
646		− return ip;
647		−}

tools/perf/util/map.h+43−40

@@ -16,23 +16,25 @@ struct dso; · 15 unmodified lines
16	16	struct maps;
17	17	struct machine;
18	18
	19	+enum mapping_type {
	20	+ /* map__map_ip/map__unmap_ip are given as offsets in the DSO. */
	21	+ MAPPING_TYPE__DSO,
	22	+ /* map__map_ip/map__unmap_ip are just the given ip value. */
	23	+ MAPPING_TYPE__IDENTITY,
	24	+};
	25	+
19	26	DECLARE_RC_STRUCT(map) {
20	27	u64 start;
21	28	u64 end;
22		− bool erange_warned:1;
23		− bool priv:1;
24		− u32 prot;
25	29	u64 pgoff;
26	30	u64 reloc;
27		−
28		− /* ip -> dso rip */
29		− u64 (map_ip)(const struct map , u64);
30		− /* dso rip -> ip */
31		− u64 (unmap_ip)(const struct map , u64);
32		−
33	31	struct dso *dso;
34	32	refcount_t refcnt;
	33	+ u32 prot;
35	34	u32 flags;
	35	+ enum mapping_type mapping_type:8;
	36	+ bool erange_warned;
	37	+ bool priv;
36	38	};
37	39
38	40	struct kmap;
@@ -41,38 +43,11 @@ struct kmap __map__kmap(struct map map); · 2 unmodified lines
41	43	struct kmap map__kmap(struct map map);
42	44	struct maps map__kmaps(struct map map);
43	45
44		−/* ip -> dso rip */
45		−u64 map__dso_map_ip(const struct map *map, u64 ip);
46		−/* dso rip -> ip */
47		−u64 map__dso_unmap_ip(const struct map *map, u64 ip);
48		−/* Returns ip */
49		−u64 identity__map_ip(const struct map *map __maybe_unused, u64 ip);
50		−
51	46	static inline struct dso map__dso(const struct map map)
52	47	{
53	48	return RC_CHK_ACCESS(map)->dso;
54	49	}
55	50
56		−static inline u64 map__map_ip(const struct map *map, u64 ip)
57		−{
58		− return RC_CHK_ACCESS(map)->map_ip(map, ip);
59		−}
60		−
61		−static inline u64 map__unmap_ip(const struct map *map, u64 ip)
62		−{
63		− return RC_CHK_ACCESS(map)->unmap_ip(map, ip);
64		−}
65		−
66		−static inline void map__map_ip_ptr(struct map map)
67		−{
68		− return RC_CHK_ACCESS(map)->map_ip;
69		−}
70		−
71		−static inline void* map__unmap_ip_ptr(struct map *map)
72		−{
73		− return RC_CHK_ACCESS(map)->unmap_ip;
74		−}
75		−
76	51	static inline u64 map__start(const struct map *map)
77	52	{
78	53	return RC_CHK_ACCESS(map)->start;
@@ -123,6 +98,34 @@ static inline size_t map__size(const struct map *map) · 44 unmodified lines
123	98	return map__end(map) - map__start(map);
124	99	}
125	100
	101	+/* ip -> dso rip */
	102	+static inline u64 map__dso_map_ip(const struct map *map, u64 ip)
	103	+{
	104	+ return ip - map__start(map) + map__pgoff(map);
	105	+}
	106	+
	107	+/* dso rip -> ip */
	108	+static inline u64 map__dso_unmap_ip(const struct map *map, u64 rip)
	109	+{
	110	+ return rip + map__start(map) - map__pgoff(map);
	111	+}
	112	+
	113	+static inline u64 map__map_ip(const struct map *map, u64 ip_or_rip)
	114	+{
	115	+ if ((RC_CHK_ACCESS(map)->mapping_type) == MAPPING_TYPE__DSO)
	116	+ return map__dso_map_ip(map, ip_or_rip);
	117	+ else
	118	+ return ip_or_rip;
	119	+}
	120	+
	121	+static inline u64 map__unmap_ip(const struct map *map, u64 ip_or_rip)
	122	+{
	123	+ if ((RC_CHK_ACCESS(map)->mapping_type) == MAPPING_TYPE__DSO)
	124	+ return map__dso_unmap_ip(map, ip_or_rip);
	125	+ else
	126	+ return ip_or_rip;
	127	+}
	128	+
126	129	/* rip/ip <-> addr suitable for passing to `objdump --start-address=` */
127	130	u64 map__rip_2objdump(struct map *map, u64 rip);
128	131
@@ -294,13 +297,13 @@ static inline void map__set_dso(struct map map, struct dso dso) · 165 unmodified lines
294	297	RC_CHK_ACCESS(map)->dso = dso;
295	298	}
296	299
297		−static inline void map__set_map_ip(struct map map, u64 (map_ip)(const struct map *map, u64 ip))
	300	+static inline void map__set_mapping_type(struct map *map, enum mapping_type type)
298	301	{
299		− RC_CHK_ACCESS(map)->map_ip = map_ip;
	302	+ RC_CHK_ACCESS(map)->mapping_type = type;
300	303	}
301	304
302		−static inline void map__set_unmap_ip(struct map map, u64 (unmap_ip)(const struct map *map, u64 rip))
	305	+static inline enum mapping_type map__mapping_type(struct map *map)
303	306	{
304		− RC_CHK_ACCESS(map)->unmap_ip = unmap_ip;
	307	+ return RC_CHK_ACCESS(map)->mapping_type;
305	308	}
306	309	#endif /* __PERF_MAP_H */

tools/perf/util/maps.c+445−83

@@ -10,6 +10,68 @@ · 9 unmodified lines
10	10	#include "ui/ui.h"
11	11	#include "unwind.h"
12	12
	13	+struct map_rb_node {
	14	+ struct rb_node rb_node;
	15	+ struct map *map;
	16	+};
	17	+
	18	+#define maps__for_each_entry(maps, map) \
	19	+ for (map = maps__first(maps); map; map = map_rb_node__next(map))
	20	+
	21	+#define maps__for_each_entry_safe(maps, map, next) \
	22	+ for (map = maps__first(maps), next = map_rb_node__next(map); map; \
	23	+ map = next, next = map_rb_node__next(map))
	24	+
	25	+static struct rb_root maps__entries(struct maps maps)
	26	+{
	27	+ return &RC_CHK_ACCESS(maps)->entries;
	28	+}
	29	+
	30	+static struct rw_semaphore maps__lock(struct maps maps)
	31	+{
	32	+ return &RC_CHK_ACCESS(maps)->lock;
	33	+}
	34	+
	35	+static struct map *maps__maps_by_name(struct maps maps)
	36	+{
	37	+ return RC_CHK_ACCESS(maps)->maps_by_name;
	38	+}
	39	+
	40	+static struct map_rb_node maps__first(struct maps maps)
	41	+{
	42	+ struct rb_node *first = rb_first(maps__entries(maps));
	43	+
	44	+ if (first)
	45	+ return rb_entry(first, struct map_rb_node, rb_node);
	46	+ return NULL;
	47	+}
	48	+
	49	+static struct map_rb_node map_rb_node__next(struct map_rb_node node)
	50	+{
	51	+ struct rb_node *next;
	52	+
	53	+ if (!node)
	54	+ return NULL;
	55	+
	56	+ next = rb_next(&node->rb_node);
	57	+
	58	+ if (!next)
	59	+ return NULL;
	60	+
	61	+ return rb_entry(next, struct map_rb_node, rb_node);
	62	+}
	63	+
	64	+static struct map_rb_node maps__find_node(struct maps maps, struct map *map)
	65	+{
	66	+ struct map_rb_node *rb_node;
	67	+
	68	+ maps__for_each_entry(maps, rb_node) {
	69	+ if (rb_node->RC_CHK_ACCESS(map) == RC_CHK_ACCESS(map))
	70	+ return rb_node;
	71	+ }
	72	+ return NULL;
	73	+}
	74	+
13	75	static void maps__init(struct maps maps, struct machine machine)
14	76	{
15	77	refcount_set(maps__refcnt(maps), 1);
@@ -196,6 +258,41 @@ void maps__put(struct maps *maps) · 180 unmodified lines
196	258	RC_CHK_PUT(maps);
197	259	}
198	260
	261	+int maps__for_each_map(struct maps maps, int (cb)(struct map map, void data), void *data)
	262	+{
	263	+ struct map_rb_node *pos;
	264	+ int ret = 0;
	265	+
	266	+ down_read(maps__lock(maps));
	267	+ maps__for_each_entry(maps, pos) {
	268	+ ret = cb(pos->map, data);
	269	+ if (ret)
	270	+ break;
	271	+ }
	272	+ up_read(maps__lock(maps));
	273	+ return ret;
	274	+}
	275	+
	276	+void maps__remove_maps(struct maps maps, bool (cb)(struct map map, void data), void *data)
	277	+{
	278	+ struct map_rb_node pos, next;
	279	+ unsigned int start_nr_maps;
	280	+
	281	+ down_write(maps__lock(maps));
	282	+
	283	+ start_nr_maps = maps__nr_maps(maps);
	284	+ maps__for_each_entry_safe(maps, pos, next) {
	285	+ if (cb(pos->map, data)) {
	286	+ __maps__remove(maps, pos);
	287	+ --RC_CHK_ACCESS(maps)->nr_maps;
	288	+ }
	289	+ }
	290	+ if (maps__maps_by_name(maps) && start_nr_maps != maps__nr_maps(maps))
	291	+ __maps__free_maps_by_name(maps);
	292	+
	293	+ up_write(maps__lock(maps));
	294	+}
	295	+
199	296	struct symbol maps__find_symbol(struct maps maps, u64 addr, struct map **mapp)
200	297	{
201	298	struct map *map = maps__find(maps, addr);
@@ -210,31 +307,40 @@ struct symbol maps__find_symbol(struct maps maps, u64 addr, struct map **mapp) · 8 unmodified lines
210	307	return NULL;
211	308	}
212	309
213		−struct symbol maps__find_symbol_by_name(struct maps maps, const char name, struct map *mapp)
214		−{
	310	+struct maps__find_symbol_by_name_args {
	311	+ struct map **mapp;
	312	+ const char *name;
215	313	struct symbol *sym;
216		− struct map_rb_node *pos;
	314	+};
217	315
218		− down_read(maps__lock(maps));
	316	+static int maps__find_symbol_by_name_cb(struct map map, void data)
	317	+{
	318	+ struct maps__find_symbol_by_name_args *args = data;
219	319
220		− maps__for_each_entry(maps, pos) {
221		− sym = map__find_symbol_by_name(pos->map, name);
	320	+ args->sym = map__find_symbol_by_name(map, args->name);
	321	+ if (!args->sym)
	322	+ return 0;
222	323
223		− if (sym == NULL)
224		− continue;
225		− if (!map__contains_symbol(pos->map, sym)) {
226		− sym = NULL;
227		− continue;
228		− }
229		− if (mapp != NULL)
230		− *mapp = pos->map;
231		− goto out;
	324	+ if (!map__contains_symbol(map, args->sym)) {
	325	+ args->sym = NULL;
	326	+ return 0;
232	327	}
233	328
234		− sym = NULL;
235		−out:
236		− up_read(maps__lock(maps));
237		− return sym;
	329	+ if (args->mapp != NULL)
	330	+ *args->mapp = map__get(map);
	331	+ return 1;
	332	+}
	333	+
	334	+struct symbol maps__find_symbol_by_name(struct maps maps, const char name, struct map *mapp)
	335	+{
	336	+ struct maps__find_symbol_by_name_args args = {
	337	+ .mapp = mapp,
	338	+ .name = name,
	339	+ .sym = NULL,
	340	+ };
	341	+
	342	+ maps__for_each_map(maps, maps__find_symbol_by_name_cb, &args);
	343	+ return args.sym;
238	344	}
239	345
240	346	int maps__find_ams(struct maps maps, struct addr_map_symbol ams)
@@ -253,41 +359,46 @@ int maps__find_ams(struct maps maps, struct addr_map_symbol ams) · 12 unmodified lines
253	359	return ams->ms.sym ? 0 : -1;
254	360	}
255	361
256		−size_t maps__fprintf(struct maps maps, FILE fp)
257		−{
258		− size_t printed = 0;
259		− struct map_rb_node *pos;
	362	+struct maps__fprintf_args {
	363	+ FILE *fp;
	364	+ size_t printed;
	365	+};
260	366
261		− down_read(maps__lock(maps));
	367	+static int maps__fprintf_cb(struct map map, void data)
	368	+{
	369	+ struct maps__fprintf_args *args = data;
262	370
263		− maps__for_each_entry(maps, pos) {
264		− printed += fprintf(fp, "Map:");
265		− printed += map__fprintf(pos->map, fp);
266		− if (verbose > 2) {
267		− printed += dso__fprintf(map__dso(pos->map), fp);
268		− printed += fprintf(fp, "--\n");
269		− }
	371	+ args->printed += fprintf(args->fp, "Map:");
	372	+ args->printed += map__fprintf(map, args->fp);
	373	+ if (verbose > 2) {
	374	+ args->printed += dso__fprintf(map__dso(map), args->fp);
	375	+ args->printed += fprintf(args->fp, "--\n");
270	376	}
	377	+ return 0;
	378	+}
271	379
272		− up_read(maps__lock(maps));
	380	+size_t maps__fprintf(struct maps maps, FILE fp)
	381	+{
	382	+ struct maps__fprintf_args args = {
	383	+ .fp = fp,
	384	+ .printed = 0,
	385	+ };
	386	+
	387	+ maps__for_each_map(maps, maps__fprintf_cb, &args);
273	388
274		− return printed;
	389	+ return args.printed;
275	390	}
276	391
277		−int maps__fixup_overlappings(struct maps maps, struct map map, FILE *fp)
	392	+/*
	393	+ * Find first map where end > map->start.
	394	+ * Same as find_vma() in kernel.
	395	+ */
	396	+static struct rb_node first_ending_after(struct maps maps, const struct map *map)
278	397	{
279	398	struct rb_root *root;
280	399	struct rb_node next, first;
281		− int err = 0;
282		−
283		− down_write(maps__lock(maps));
284	400
285	401	root = maps__entries(maps);
286		−
287		− /*
288		− * Find first map where end > map->start.
289		− * Same as find_vma() in kernel.
290		− */
291	402	next = root->rb_node;
292	403	first = NULL;
293	404	while (next) {
@@ -301,8 +412,23 @@ int maps__fixup_overlappings(struct maps maps, struct map map, FILE *fp) · 7 unmodified lines
301	412	} else
302	413	next = next->rb_right;
303	414	}
	415	+ return first;
	416	+}
304	417
305		− next = first;
	418	+/*
	419	+ * Adds new to maps, if new overlaps existing entries then the existing maps are
	420	+ * adjusted or removed so that new fits without overlapping any entries.
	421	+ */
	422	+int maps__fixup_overlap_and_insert(struct maps maps, struct map new)
	423	+{
	424	+
	425	+ struct rb_node *next;
	426	+ int err = 0;
	427	+ FILE *fp = debug_file();
	428	+
	429	+ down_write(maps__lock(maps));
	430	+
	431	+ next = first_ending_after(maps, new);
306	432	while (next && !err) {
307	433	struct map_rb_node *pos = rb_entry(next, struct map_rb_node, rb_node);
308	434	next = rb_next(&pos->rb_node);
@@ -311,27 +437,27 @@ int maps__fixup_overlappings(struct maps maps, struct map map, FILE *fp) · 2 unmodified lines
311	437	* Stop if current map starts after map->end.
312	438	* Maps are ordered by start: next will not overlap for sure.
313	439	*/
314		− if (map__start(pos->map) >= map__end(map))
	440	+ if (map__start(pos->map) >= map__end(new))
315	441	break;
316	442
317	443	if (verbose >= 2) {
318	444
319	445	if (use_browser) {
320	446	pr_debug("overlapping maps in %s (disable tui for more info)\n",
321		− map__dso(map)->name);
	447	+ map__dso(new)->name);
322	448	} else {
323		− fputs("overlapping maps:\n", fp);
324		− map__fprintf(map, fp);
	449	+ pr_debug("overlapping maps:\n");
	450	+ map__fprintf(new, fp);
325	451	map__fprintf(pos->map, fp);
326	452	}
327	453	}
328	454
329		− rb_erase_init(&pos->rb_node, root);
	455	+ rb_erase_init(&pos->rb_node, maps__entries(maps));
330	456	/*
331	457	* Now check if we need to create new maps for areas not
332	458	* overlapped by the new map:
333	459	*/
334		− if (map__start(map) > map__start(pos->map)) {
	460	+ if (map__start(new) > map__start(pos->map)) {
335	461	struct map *before = map__clone(pos->map);
336	462
337	463	if (before == NULL) {
@@ -339,7 +465,7 @@ int maps__fixup_overlappings(struct maps maps, struct map map, FILE *fp) · 1 unmodified line
339	465	goto put_map;
340	466	}
341	467
342		− map__set_end(before, map__start(map));
	468	+ map__set_end(before, map__start(new));
343	469	err = __maps__insert(maps, before);
344	470	if (err) {
345	471	map__put(before);
@@ -351,7 +477,7 @@ int maps__fixup_overlappings(struct maps maps, struct map map, FILE *fp) · 5 unmodified lines
351	477	map__put(before);
352	478	}
353	479
354		− if (map__end(map) < map__end(pos->map)) {
	480	+ if (map__end(new) < map__end(pos->map)) {
355	481	struct map *after = map__clone(pos->map);
356	482
357	483	if (after == NULL) {
@@ -359,10 +485,10 @@ int maps__fixup_overlappings(struct maps maps, struct map map, FILE *fp) · 1 unmodified line
359	485	goto put_map;
360	486	}
361	487
362		− map__set_start(after, map__end(map));
363		− map__add_pgoff(after, map__end(map) - map__start(pos->map));
364		− assert(map__map_ip(pos->map, map__end(map)) ==
365		− map__map_ip(after, map__end(map)));
	488	+ map__set_start(after, map__end(new));
	489	+ map__add_pgoff(after, map__end(new) - map__start(pos->map));
	490	+ assert(map__map_ip(pos->map, map__end(new)) ==
	491	+ map__map_ip(after, map__end(new)));
366	492	err = __maps__insert(maps, after);
367	493	if (err) {
368	494	map__put(after);
@@ -376,16 +502,14 @@ put_map: · 7 unmodified lines
376	502	map__put(pos->map);
377	503	free(pos);
378	504	}
	505	+ /* Add the map. */
	506	+ err = __maps__insert(maps, new);
379	507	up_write(maps__lock(maps));
380	508	return err;
381	509	}
382	510
383		−/*
384		− * XXX This should not really _copy_ te maps, but refcount them.
385		− */
386		−int maps__clone(struct thread thread, struct maps parent)
	511	+int maps__copy_from(struct maps maps, struct maps parent)
387	512	{
388		− struct maps *maps = thread__maps(thread);
389	513	int err;
390	514	struct map_rb_node *rb_node;
391	515
@@ -416,17 +540,6 @@ out_unlock: · 24 unmodified lines
416	540	return err;
417	541	}
418	542
419		−struct map_rb_node maps__find_node(struct maps maps, struct map *map)
420		−{
421		− struct map_rb_node *rb_node;
422		−
423		− maps__for_each_entry(maps, rb_node) {
424		− if (rb_node->RC_CHK_ACCESS(map) == RC_CHK_ACCESS(map))
425		− return rb_node;
426		− }
427		− return NULL;
428		−}
429		−
430	543	struct map maps__find(struct maps maps, u64 ip)
431	544	{
432	545	struct rb_node *p;
@@ -452,26 +565,275 @@ out: · 19 unmodified lines
452	565	return m ? m->map : NULL;
453	566	}
454	567
455		−struct map_rb_node maps__first(struct maps maps)
	568	+static int map__strcmp(const void a, const void b)
456	569	{
457		− struct rb_node *first = rb_first(maps__entries(maps));
	570	+ const struct map map_a = (const struct map **)a;
	571	+ const struct map map_b = (const struct map **)b;
	572	+ const struct dso *dso_a = map__dso(map_a);
	573	+ const struct dso *dso_b = map__dso(map_b);
	574	+ int ret = strcmp(dso_a->short_name, dso_b->short_name);
458	575
459		− if (first)
460		− return rb_entry(first, struct map_rb_node, rb_node);
461		− return NULL;
	576	+ if (ret == 0 && map_a != map_b) {
	577	+ /*
	578	+ * Ensure distinct but name equal maps have an order in part to
	579	+ * aid reference counting.
	580	+ */
	581	+ ret = (int)map__start(map_a) - (int)map__start(map_b);
	582	+ if (ret == 0)
	583	+ ret = (int)((intptr_t)map_a - (intptr_t)map_b);
	584	+ }
	585	+
	586	+ return ret;
462	587	}
463	588
464		−struct map_rb_node map_rb_node__next(struct map_rb_node node)
	589	+static int map__strcmp_name(const void name, const void b)
465	590	{
466		− struct rb_node *next;
	591	+ const struct dso dso = map__dso((const struct map **)b);
467	592
468		− if (!node)
469		− return NULL;
	593	+ return strcmp(name, dso->short_name);
	594	+}
470	595
471		− next = rb_next(&node->rb_node);
	596	+void __maps__sort_by_name(struct maps *maps)
	597	+{
	598	+ qsort(maps__maps_by_name(maps), maps__nr_maps(maps), sizeof(struct map *), map__strcmp);
	599	+}
472	600
473		− if (!next)
	601	+static int map__groups__sort_by_name_from_rbtree(struct maps *maps)
	602	+{
	603	+ struct map_rb_node *rb_node;
	604	+ struct map **maps_by_name = realloc(maps__maps_by_name(maps),
	605	+ maps__nr_maps(maps) * sizeof(struct map *));
	606	+ int i = 0;
	607	+
	608	+ if (maps_by_name == NULL)
	609	+ return -1;
	610	+
	611	+ up_read(maps__lock(maps));
	612	+ down_write(maps__lock(maps));
	613	+
	614	+ RC_CHK_ACCESS(maps)->maps_by_name = maps_by_name;
	615	+ RC_CHK_ACCESS(maps)->nr_maps_allocated = maps__nr_maps(maps);
	616	+
	617	+ maps__for_each_entry(maps, rb_node)
	618	+ maps_by_name[i++] = map__get(rb_node->map);
	619	+
	620	+ __maps__sort_by_name(maps);
	621	+
	622	+ up_write(maps__lock(maps));
	623	+ down_read(maps__lock(maps));
	624	+
	625	+ return 0;
	626	+}
	627	+
	628	+static struct map __maps__find_by_name(struct maps maps, const char *name)
	629	+{
	630	+ struct map **mapp;
	631	+
	632	+ if (maps__maps_by_name(maps) == NULL &&
	633	+ map__groups__sort_by_name_from_rbtree(maps))
474	634	return NULL;
475	635
476		− return rb_entry(next, struct map_rb_node, rb_node);
	636	+ mapp = bsearch(name, maps__maps_by_name(maps), maps__nr_maps(maps),
	637	+ sizeof(*mapp), map__strcmp_name);
	638	+ if (mapp)
	639	+ return *mapp;
	640	+ return NULL;
	641	+}
	642	+
	643	+struct map maps__find_by_name(struct maps maps, const char *name)
	644	+{
	645	+ struct map_rb_node *rb_node;
	646	+ struct map *map;
	647	+
	648	+ down_read(maps__lock(maps));
	649	+
	650	+
	651	+ if (RC_CHK_ACCESS(maps)->last_search_by_name) {
	652	+ const struct dso *dso = map__dso(RC_CHK_ACCESS(maps)->last_search_by_name);
	653	+
	654	+ if (strcmp(dso->short_name, name) == 0) {
	655	+ map = RC_CHK_ACCESS(maps)->last_search_by_name;
	656	+ goto out_unlock;
	657	+ }
	658	+ }
	659	+ /*
	660	+ * If we have maps->maps_by_name, then the name isn't in the rbtree,
	661	+ * as maps->maps_by_name mirrors the rbtree when lookups by name are
	662	+ * made.
	663	+ */
	664	+ map = __maps__find_by_name(maps, name);
	665	+ if (map \|\| maps__maps_by_name(maps) != NULL)
	666	+ goto out_unlock;
	667	+
	668	+ /* Fallback to traversing the rbtree... */
	669	+ maps__for_each_entry(maps, rb_node) {
	670	+ struct dso *dso;
	671	+
	672	+ map = rb_node->map;
	673	+ dso = map__dso(map);
	674	+ if (strcmp(dso->short_name, name) == 0) {
	675	+ RC_CHK_ACCESS(maps)->last_search_by_name = map;
	676	+ goto out_unlock;
	677	+ }
	678	+ }
	679	+ map = NULL;
	680	+
	681	+out_unlock:
	682	+ up_read(maps__lock(maps));
	683	+ return map;
	684	+}
	685	+
	686	+struct map maps__find_next_entry(struct maps maps, struct map *map)
	687	+{
	688	+ struct map_rb_node *rb_node = maps__find_node(maps, map);
	689	+ struct map_rb_node *next = map_rb_node__next(rb_node);
	690	+
	691	+ if (next)
	692	+ return next->map;
	693	+
	694	+ return NULL;
	695	+}
	696	+
	697	+void maps__fixup_end(struct maps *maps)
	698	+{
	699	+ struct map_rb_node prev = NULL, curr;
	700	+
	701	+ down_write(maps__lock(maps));
	702	+
	703	+ maps__for_each_entry(maps, curr) {
	704	+ if (prev && (!map__end(prev->map) \|\| map__end(prev->map) > map__start(curr->map)))
	705	+ map__set_end(prev->map, map__start(curr->map));
	706	+
	707	+ prev = curr;
	708	+ }
	709	+
	710	+ /*
	711	+ * We still haven't the actual symbols, so guess the
	712	+ * last map final address.
	713	+ */
	714	+ if (curr && !map__end(curr->map))
	715	+ map__set_end(curr->map, ~0ULL);
	716	+
	717	+ up_write(maps__lock(maps));
	718	+}
	719	+
	720	+/*
	721	+ * Merges map into maps by splitting the new map within the existing map
	722	+ * regions.
	723	+ */
	724	+int maps__merge_in(struct maps kmaps, struct map new_map)
	725	+{
	726	+ struct map_rb_node *rb_node;
	727	+ struct rb_node *first;
	728	+ bool overlaps;
	729	+ LIST_HEAD(merged);
	730	+ int err = 0;
	731	+
	732	+ down_read(maps__lock(kmaps));
	733	+ first = first_ending_after(kmaps, new_map);
	734	+ rb_node = first ? rb_entry(first, struct map_rb_node, rb_node) : NULL;
	735	+ overlaps = rb_node && map__start(rb_node->map) < map__end(new_map);
	736	+ up_read(maps__lock(kmaps));
	737	+
	738	+ if (!overlaps)
	739	+ return maps__insert(kmaps, new_map);
	740	+
	741	+ maps__for_each_entry(kmaps, rb_node) {
	742	+ struct map *old_map = rb_node->map;
	743	+
	744	+ /* no overload with this one */
	745	+ if (map__end(new_map) < map__start(old_map) \|\|
	746	+ map__start(new_map) >= map__end(old_map))
	747	+ continue;
	748	+
	749	+ if (map__start(new_map) < map__start(old_map)) {
	750	+ /*
	751	+ * \|new......
	752	+ * \|old....
	753	+ */
	754	+ if (map__end(new_map) < map__end(old_map)) {
	755	+ /*
	756	+ * \|new......\| -> \|new..\|
	757	+ * \|old....\| -> \|old....\|
	758	+ */
	759	+ map__set_end(new_map, map__start(old_map));
	760	+ } else {
	761	+ /*
	762	+ * \|new.............\| -> \|new..\| \|new..\|
	763	+ * \|old....\| -> \|old....\|
	764	+ */
	765	+ struct map_list_node *m = map_list_node__new();
	766	+
	767	+ if (!m) {
	768	+ err = -ENOMEM;
	769	+ goto out;
	770	+ }
	771	+
	772	+ m->map = map__clone(new_map);
	773	+ if (!m->map) {
	774	+ free(m);
	775	+ err = -ENOMEM;
	776	+ goto out;
	777	+ }
	778	+
	779	+ map__set_end(m->map, map__start(old_map));
	780	+ list_add_tail(&m->node, &merged);
	781	+ map__add_pgoff(new_map, map__end(old_map) - map__start(new_map));
	782	+ map__set_start(new_map, map__end(old_map));
	783	+ }
	784	+ } else {
	785	+ /*
	786	+ * \|new......
	787	+ * \|old....
	788	+ */
	789	+ if (map__end(new_map) < map__end(old_map)) {
	790	+ /*
	791	+ * \|new..\| -> x
	792	+ * \|old.........\| -> \|old.........\|
	793	+ */
	794	+ map__put(new_map);
	795	+ new_map = NULL;
	796	+ break;
	797	+ } else {
	798	+ /*
	799	+ * \|new......\| -> \|new...\|
	800	+ * \|old....\| -> \|old....\|
	801	+ */
	802	+ map__add_pgoff(new_map, map__end(old_map) - map__start(new_map));
	803	+ map__set_start(new_map, map__end(old_map));
	804	+ }
	805	+ }
	806	+ }
	807	+
	808	+out:
	809	+ while (!list_empty(&merged)) {
	810	+ struct map_list_node *old_node;
	811	+
	812	+ old_node = list_entry(merged.next, struct map_list_node, node);
	813	+ list_del_init(&old_node->node);
	814	+ if (!err)
	815	+ err = maps__insert(kmaps, old_node->map);
	816	+ map__put(old_node->map);
	817	+ free(old_node);
	818	+ }
	819	+
	820	+ if (new_map) {
	821	+ if (!err)
	822	+ err = maps__insert(kmaps, new_map);
	823	+ map__put(new_map);
	824	+ }
	825	+ return err;
	826	+}
	827	+
	828	+void maps__load_first(struct maps *maps)
	829	+{
	830	+ struct map_rb_node *first;
	831	+
	832	+ down_read(maps__lock(maps));
	833	+
	834	+ first = maps__first(maps);
	835	+ if (first)
	836	+ map__load(first->map);
	837	+
	838	+ up_read(maps__lock(maps));
477	839	}

tools/perf/util/maps.h+19−29

@@ -14,24 +14,18 @@ struct ref_reloc_sym; · 13 unmodified lines
14	14	struct machine;
15	15	struct map;
16	16	struct maps;
17		−struct thread;
18	17
19		−struct map_rb_node {
20		− struct rb_node rb_node;
	18	+struct map_list_node {
	19	+ struct list_head node;
21	20	struct map *map;
22	21	};
23	22
24		−struct map_rb_node maps__first(struct maps maps);
25		−struct map_rb_node map_rb_node__next(struct map_rb_node node);
26		−struct map_rb_node maps__find_node(struct maps maps, struct map *map);
27		−struct map maps__find(struct maps maps, u64 addr);
28		−
29		−#define maps__for_each_entry(maps, map) \
30		− for (map = maps__first(maps); map; map = map_rb_node__next(map))
	23	+static inline struct map_list_node *map_list_node__new(void)
	24	+{
	25	+ return malloc(sizeof(struct map_list_node));
	26	+}
31	27
32		−#define maps__for_each_entry_safe(maps, map, next) \
33		− for (map = maps__first(maps), next = map_rb_node__next(map); map; \
34		− map = next, next = map_rb_node__next(map))
	28	+struct map maps__find(struct maps maps, u64 addr);
35	29
36	30	DECLARE_RC_STRUCT(maps) {
37	31	struct rb_root entries;
@@ -58,7 +52,7 @@ struct kmap { · 20 unmodified lines
58	52
59	53	struct maps maps__new(struct machine machine);
60	54	bool maps__empty(struct maps *maps);
61		−int maps__clone(struct thread thread, struct maps parent);
	55	+int maps__copy_from(struct maps maps, struct maps parent);
62	56
63	57	struct maps maps__get(struct maps maps);
64	58	void maps__put(struct maps *maps);
@@ -71,26 +65,16 @@ static inline void __maps__zput(struct maps **map) · 6 unmodified lines
71	65
72	66	#define maps__zput(map) __maps__zput(&map)
73	67
74		−static inline struct rb_root maps__entries(struct maps maps)
75		−{
76		− return &RC_CHK_ACCESS(maps)->entries;
77		−}
	68	+/* Iterate over map calling cb for each entry. */
	69	+int maps__for_each_map(struct maps maps, int (cb)(struct map map, void data), void *data);
	70	+/* Iterate over map removing an entry if cb returns true. */
	71	+void maps__remove_maps(struct maps maps, bool (cb)(struct map map, void data), void *data);
78	72
79	73	static inline struct machine maps__machine(struct maps maps)
80	74	{
81	75	return RC_CHK_ACCESS(maps)->machine;
82	76	}
83	77
84		−static inline struct rw_semaphore maps__lock(struct maps maps)
85		−{
86		− return &RC_CHK_ACCESS(maps)->lock;
87		−}
88		−
89		−static inline struct map *maps__maps_by_name(struct maps maps)
90		−{
91		− return RC_CHK_ACCESS(maps)->maps_by_name;
92		−}
93		−
94	78	static inline unsigned int maps__nr_maps(const struct maps *maps)
95	79	{
96	80	return RC_CHK_ACCESS(maps)->nr_maps;
@@ -125,12 +109,18 @@ struct addr_map_symbol; · 28 unmodified lines
125	109
126	110	int maps__find_ams(struct maps maps, struct addr_map_symbol ams);
127	111
128		−int maps__fixup_overlappings(struct maps maps, struct map map, FILE *fp);
	112	+int maps__fixup_overlap_and_insert(struct maps maps, struct map new);
129	113
130	114	struct map maps__find_by_name(struct maps maps, const char *name);
131	115
	116	+struct map maps__find_next_entry(struct maps maps, struct map *map);
	117	+
132	118	int maps__merge_in(struct maps kmaps, struct map new_map);
133	119
134	120	void __maps__sort_by_name(struct maps *maps);
135	121
	122	+void maps__fixup_end(struct maps *maps);
	123	+
	124	+void maps__load_first(struct maps *maps);
	125	+
136	126	#endif // __PERF_MAPS_H

tools/perf/util/mem-events.c+14−11

@@ -100,11 +100,14 @@ int perf_mem_events__parse(const char *str) · 99 unmodified lines
100	100	return -1;
101	101	}
102	102
103		−static bool perf_mem_event__supported(const char mnt, char sysfs_name)
	103	+static bool perf_mem_event__supported(const char mnt, struct perf_pmu pmu,
	104	+ struct perf_mem_event *e)
104	105	{
	106	+ char sysfs_name[100];
105	107	char path[PATH_MAX];
106	108	struct stat st;
107	109
	110	+ scnprintf(sysfs_name, sizeof(sysfs_name), e->sysfs_name, pmu->name);
108	111	scnprintf(path, PATH_MAX, "%s/devices/%s", mnt, sysfs_name);
109	112	return !stat(path, &st);
110	113	}
@@ -120,7 +123,6 @@ int perf_mem_events__init(void) · 9 unmodified lines
120	123
121	124	for (j = 0; j < PERF_MEM_EVENTS__MAX; j++) {
122	125	struct perf_mem_event *e = perf_mem_events__ptr(j);
123		− char sysfs_name[100];
124	126	struct perf_pmu *pmu = NULL;
125	127
126	128	/*
@@ -136,12 +138,12 @@ int perf_mem_events__init(void) · 9 unmodified lines
136	138	* of core PMU.
137	139	*/
138	140	while ((pmu = perf_pmus__scan(pmu)) != NULL) {
139		− scnprintf(sysfs_name, sizeof(sysfs_name), e->sysfs_name, pmu->name);
140		− e->supported \|= perf_mem_event__supported(mnt, sysfs_name);
	141	+ e->supported \|= perf_mem_event__supported(mnt, pmu, e);
	142	+ if (e->supported) {
	143	+ found = true;
	144	+ break;
	145	+ }
141	146	}
142		−
143		− if (e->supported)
144		− found = true;
145	147	}
146	148
147	149	return found ? 0 : -ENOENT;
@@ -167,13 +169,10 @@ static void perf_mem_events__print_unsupport_hybrid(struct perf_mem_event *e, · 19 unmodified lines
167	169	int idx)
168	170	{
169	171	const char *mnt = sysfs__mount();
170		− char sysfs_name[100];
171	172	struct perf_pmu *pmu = NULL;
172	173
173	174	while ((pmu = perf_pmus__scan(pmu)) != NULL) {
174		− scnprintf(sysfs_name, sizeof(sysfs_name), e->sysfs_name,
175		− pmu->name);
176		− if (!perf_mem_event__supported(mnt, sysfs_name)) {
	175	+ if (!perf_mem_event__supported(mnt, pmu, e)) {
177	176	pr_err("failed: event '%s' not supported\n",
178	177	perf_mem_events__name(idx, pmu->name));
179	178	}
@@ -183,6 +182,7 @@ static void perf_mem_events__print_unsupport_hybrid(struct perf_mem_event *e, · 3 unmodified lines
183	182	int perf_mem_events__record_args(const char *rec_argv, int argv_nr,
184	183	char *rec_tmp, int tmp_nr)
185	184	{
	185	+ const char *mnt = sysfs__mount();
186	186	int i = *argv_nr, k = 0;
187	187	struct perf_mem_event *e;
188	188
@@ -211,6 +211,9 @@ int perf_mem_events__record_args(const char *rec_argv, int argv_nr, · 22 unmodified lines
211	211	while ((pmu = perf_pmus__scan(pmu)) != NULL) {
212	212	const char *s = perf_mem_events__name(j, pmu->name);
213	213
	214	+ if (!perf_mem_event__supported(mnt, pmu, e))
	215	+ continue;
	216	+
214	217	rec_argv[i++] = "-e";
215	218	if (s) {
216	219	char *copy = strdup(s);

tools/perf/util/mmap.c+2−3

@@ -295,15 +295,14 @@ int mmap__mmap(struct mmap map, struct mmap_params mp, int fd, struct perf_cpu · 294 unmodified lines
295	295
296	296	map->core.flush = mp->flush;
297	297
298		− map->comp_level = mp->comp_level;
299	298	#ifndef PYTHON_PERF
300		− if (zstd_init(&map->zstd_data, map->comp_level)) {
	299	+ if (zstd_init(&map->zstd_data, mp->comp_level)) {
301	300	pr_debug2("failed to init mmap compressor, error %d\n", errno);
302	301	return -1;
303	302	}
304	303	#endif
305	304
306		− if (map->comp_level && !perf_mmap__aio_enabled(map)) {
	305	+ if (mp->comp_level && !perf_mmap__aio_enabled(map)) {
307	306	map->data = mmap(NULL, mmap__mmap_len(map), PROT_READ\|PROT_WRITE,
308	307	MAP_PRIVATE\|MAP_ANONYMOUS, 0, 0);
309	308	if (map->data == MAP_FAILED) {

tools/perf/util/mmap.h+0−1

@@ -39,7 +39,6 @@ struct mmap { · 38 unmodified lines
39	39	#endif
40	40	struct mmap_cpu_mask affinity_mask;
41	41	void *data;
42		− int comp_level;
43	42	struct perf_data_file *file;
44	43	struct zstd_data zstd_data;
45	44	};

tools/perf/util/parse-branch-options.c+1−0

@@ -36,6 +36,7 @@ static const struct branch_mode branch_modes[] = { · 35 unmodified lines
36	36	BRANCH_OPT("stack", PERF_SAMPLE_BRANCH_CALL_STACK),
37	37	BRANCH_OPT("hw_index", PERF_SAMPLE_BRANCH_HW_INDEX),
38	38	BRANCH_OPT("priv", PERF_SAMPLE_BRANCH_PRIV_SAVE),
	39	+ BRANCH_OPT("counter", PERF_SAMPLE_BRANCH_COUNTERS),
39	40	BRANCH_END
40	41	};
41	42

tools/perf/util/parse-events.c+38−14

@@ -976,7 +976,7 @@ static int config_term_pmu(struct perf_event_attr *attr, · 975 unmodified lines
976	976	struct parse_events_error *err)
977	977	{
978	978	if (term->type_term == PARSE_EVENTS__TERM_TYPE_LEGACY_CACHE) {
979		− const struct perf_pmu *pmu = perf_pmus__find_by_type(attr->type);
	979	+ struct perf_pmu *pmu = perf_pmus__find_by_type(attr->type);
980	980
981	981	if (!pmu) {
982	982	char *err_str;
@@ -986,15 +986,23 @@ static int config_term_pmu(struct perf_event_attr *attr, · 3 unmodified lines
986	986	err_str, /help=/NULL);
987	987	return -EINVAL;
988	988	}
989		− if (perf_pmu__supports_legacy_cache(pmu)) {
	989	+ /*
	990	+ * Rewrite the PMU event to a legacy cache one unless the PMU
	991	+ * doesn't support legacy cache events or the event is present
	992	+ * within the PMU.
	993	+ */
	994	+ if (perf_pmu__supports_legacy_cache(pmu) &&
	995	+ !perf_pmu__have_event(pmu, term->config)) {
990	996	attr->type = PERF_TYPE_HW_CACHE;
991	997	return parse_events__decode_legacy_cache(term->config, pmu->type,
992	998	&attr->config);
993		− } else
	999	+ } else {
994	1000	term->type_term = PARSE_EVENTS__TERM_TYPE_USER;
	1001	+ term->no_value = true;
	1002	+ }
995	1003	}
996	1004	if (term->type_term == PARSE_EVENTS__TERM_TYPE_HARDWARE) {
997		− const struct perf_pmu *pmu = perf_pmus__find_by_type(attr->type);
	1005	+ struct perf_pmu *pmu = perf_pmus__find_by_type(attr->type);
998	1006
999	1007	if (!pmu) {
1000	1008	char *err_str;
@@ -1004,10 +1012,19 @@ static int config_term_pmu(struct perf_event_attr *attr, · 3 unmodified lines
1004	1012	err_str, /help=/NULL);
1005	1013	return -EINVAL;
1006	1014	}
1007		− attr->type = PERF_TYPE_HARDWARE;
1008		− attr->config = term->val.num;
1009		− if (perf_pmus__supports_extended_type())
1010		− attr->config \|= (__u64)pmu->type << PERF_PMU_TYPE_SHIFT;
	1015	+ /*
	1016	+ * If the PMU has a sysfs or json event prefer it over
	1017	+ * legacy. ARM requires this.
	1018	+ */
	1019	+ if (perf_pmu__have_event(pmu, term->config)) {
	1020	+ term->type_term = PARSE_EVENTS__TERM_TYPE_USER;
	1021	+ term->no_value = true;
	1022	+ } else {
	1023	+ attr->type = PERF_TYPE_HARDWARE;
	1024	+ attr->config = term->val.num;
	1025	+ if (perf_pmus__supports_extended_type())
	1026	+ attr->config \|= (__u64)pmu->type << PERF_PMU_TYPE_SHIFT;
	1027	+ }
1011	1028	return 0;
1012	1029	}
1013	1030	if (term->type_term == PARSE_EVENTS__TERM_TYPE_USER \|\|
@@ -1381,6 +1398,7 @@ int parse_events_add_pmu(struct parse_events_state *parse_state, · 367 unmodified lines
1381	1398	YYLTYPE *loc = loc_;
1382	1399	LIST_HEAD(config_terms);
1383	1400	struct parse_events_terms parsed_terms;
	1401	+ bool alias_rewrote_terms = false;
1384	1402
1385	1403	pmu = parse_state->fake_pmu ?: perf_pmus__find(name);
1386	1404
@@ -1433,7 +1451,15 @@ int parse_events_add_pmu(struct parse_events_state *parse_state, · 46 unmodified lines
1433	1451	return evsel ? 0 : -ENOMEM;
1434	1452	}
1435	1453
1436		− if (!parse_state->fake_pmu && perf_pmu__check_alias(pmu, &parsed_terms, &info, err)) {
	1454	+ /* Configure attr/terms with a known PMU, this will set hardcoded terms. */
	1455	+ if (config_attr(&attr, &parsed_terms, parse_state->error, config_term_pmu)) {
	1456	+ parse_events_terms__exit(&parsed_terms);
	1457	+ return -EINVAL;
	1458	+ }
	1459	+
	1460	+ /* Look for event names in the terms and rewrite into format based terms. */
	1461	+ if (!parse_state->fake_pmu && perf_pmu__check_alias(pmu, &parsed_terms,
	1462	+ &info, &alias_rewrote_terms, err)) {
1437	1463	parse_events_terms__exit(&parsed_terms);
1438	1464	return -EINVAL;
1439	1465	}
@@ -1447,11 +1473,9 @@ int parse_events_add_pmu(struct parse_events_state *parse_state, · 7 unmodified lines
1447	1473	strbuf_release(&sb);
1448	1474	}
1449	1475
1450		− /*
1451		− * Configure hardcoded terms first, no need to check
1452		− * return value when called with fail == 0 ;)
1453		− */
1454		− if (config_attr(&attr, &parsed_terms, parse_state->error, config_term_pmu)) {
	1476	+ /* Configure attr/terms again if an alias was expanded. */
	1477	+ if (alias_rewrote_terms &&
	1478	+ config_attr(&attr, &parsed_terms, parse_state->error, config_term_pmu)) {
1455	1479	parse_events_terms__exit(&parsed_terms);
1456	1480	return -EINVAL;
1457	1481	}

tools/perf/util/perf_api_probe.c+2−2

@@ -64,7 +64,7 @@ static bool perf_probe_api(setup_probe_fn_t fn) · 63 unmodified lines
64	64	struct perf_cpu cpu;
65	65	int ret, i = 0;
66	66
67		− cpus = perf_cpu_map__new(NULL);
	67	+ cpus = perf_cpu_map__new_online_cpus();
68	68	if (!cpus)
69	69	return false;
70	70	cpu = perf_cpu_map__cpu(cpus, 0);
@@ -140,7 +140,7 @@ bool perf_can_record_cpu_wide(void) · 69 unmodified lines
140	140	struct perf_cpu cpu;
141	141	int fd;
142	142
143		− cpus = perf_cpu_map__new(NULL);
	143	+ cpus = perf_cpu_map__new_online_cpus();
144	144	if (!cpus)
145	145	return false;
146	146

tools/perf/util/perf_event_attr_fprintf.c+1−0

@@ -55,6 +55,7 @@ static void __p_branch_sample_type(char *buf, size_t size, u64 value) · 54 unmodified lines
55	55	bit_name(COND), bit_name(CALL_STACK), bit_name(IND_JUMP),
56	56	bit_name(CALL), bit_name(NO_FLAGS), bit_name(NO_CYCLES),
57	57	bit_name(TYPE_SAVE), bit_name(HW_INDEX), bit_name(PRIV_SAVE),
	58	+ bit_name(COUNTERS),
58	59	{ .name = NULL, }
59	60	};
60	61	#undef bit_name

tools/perf/util/pmu.c+6−2

@@ -1494,12 +1494,14 @@ static int check_info_data(struct perf_pmu *pmu, · 1493 unmodified lines
1494	1494	* defined for the alias
1495	1495	*/
1496	1496	int perf_pmu__check_alias(struct perf_pmu pmu, struct parse_events_terms head_terms,
1497		− struct perf_pmu_info info, struct parse_events_error err)
	1497	+ struct perf_pmu_info info, bool rewrote_terms,
	1498	+ struct parse_events_error *err)
1498	1499	{
1499	1500	struct parse_events_term term, h;
1500	1501	struct perf_pmu_alias *alias;
1501	1502	int ret;
1502	1503
	1504	+ *rewrote_terms = false;
1503	1505	info->per_pkg = false;
1504	1506
1505	1507	/*
@@ -1521,7 +1523,7 @@ int perf_pmu__check_alias(struct perf_pmu pmu, struct parse_events_terms head_ · 15 unmodified lines
1521	1523	NULL);
1522	1524	return ret;
1523	1525	}
1524		−
	1526	+ *rewrote_terms = true;
1525	1527	ret = check_info_data(pmu, alias, info, err, term->err_term);
1526	1528	if (ret)
1527	1529	return ret;
@@ -1615,6 +1617,8 @@ bool perf_pmu__auto_merge_stats(const struct perf_pmu *pmu) · 87 unmodified lines
1615	1617
1616	1618	bool perf_pmu__have_event(struct perf_pmu pmu, const char name)
1617	1619	{
	1620	+ if (!name)
	1621	+ return false;
1618	1622	if (perf_pmu__find_alias(pmu, name, /load=/ true) != NULL)
1619	1623	return true;
1620	1624	if (pmu->cpu_aliases_added \|\| !pmu->events_table)

tools/perf/util/pmu.h+2−1

@@ -201,7 +201,8 @@ int perf_pmu__config_terms(const struct perf_pmu *pmu, · 200 unmodified lines
201	201	__u64 perf_pmu__format_bits(struct perf_pmu pmu, const char name);
202	202	int perf_pmu__format_type(struct perf_pmu pmu, const char name);
203	203	int perf_pmu__check_alias(struct perf_pmu pmu, struct parse_events_terms head_terms,
204		− struct perf_pmu_info info, struct parse_events_error err);
	204	+ struct perf_pmu_info info, bool rewrote_terms,
	205	+ struct parse_events_error *err);
205	206	int perf_pmu__find_event(struct perf_pmu pmu, const char event, void *state, pmu_event_callback cb);
206	207
207	208	int perf_pmu__format_parse(struct perf_pmu *pmu, int dirfd, bool eager_load);

tools/perf/util/probe-event.c+26−14

@@ -149,10 +149,32 @@ static int kernel_get_symbol_address_by_name(const char name, u64 addr, · 148 unmodified lines
149	149	return 0;
150	150	}
151	151
	152	+struct kernel_get_module_map_cb_args {
	153	+ const char *module;
	154	+ struct map *result;
	155	+};
	156	+
	157	+static int kernel_get_module_map_cb(struct map map, void data)
	158	+{
	159	+ struct kernel_get_module_map_cb_args *args = data;
	160	+ struct dso *dso = map__dso(map);
	161	+ const char short_name = dso->short_name; / short_name is "[module]" */
	162	+ u16 short_name_len = dso->short_name_len;
	163	+
	164	+ if (strncmp(short_name + 1, args->module, short_name_len - 2) == 0 &&
	165	+ args->module[short_name_len - 2] == '\0') {
	166	+ args->result = map__get(map);
	167	+ return 1;
	168	+ }
	169	+ return 0;
	170	+}
	171	+
152	172	static struct map kernel_get_module_map(const char module)
153	173	{
154		− struct maps *maps = machine__kernel_maps(host_machine);
155		− struct map_rb_node *pos;
	174	+ struct kernel_get_module_map_cb_args args = {
	175	+ .module = module,
	176	+ .result = NULL,
	177	+ };
156	178
157	179	/* A file path -- this is an offline module */
158	180	if (module && strchr(module, '/'))
@@ -164,19 +186,9 @@ static struct map kernel_get_module_map(const char module) · 5 unmodified lines
164	186	return map__get(map);
165	187	}
166	188
167		− maps__for_each_entry(maps, pos) {
168		− /* short_name is "[module]" */
169		− struct dso *dso = map__dso(pos->map);
170		− const char *short_name = dso->short_name;
171		− u16 short_name_len = dso->short_name_len;
	189	+ maps__for_each_map(machine__kernel_maps(host_machine), kernel_get_module_map_cb, &args);
172	190
173		− if (strncmp(short_name + 1, module,
174		− short_name_len - 2) == 0 &&
175		− module[short_name_len - 2] == '\0') {
176		− return map__get(pos->map);
177		− }
178		− }
179		− return NULL;
	191	+ return args.result;
180	192	}
181	193
182	194	struct map get_target_map(const char target, struct nsinfo *nsi, bool user)

tools/perf/util/probe-finder.c+5−196

@@ -23,6 +23,7 @@ · 22 unmodified lines
23	23	#include "event.h"
24	24	#include "dso.h"
25	25	#include "debug.h"
	26	+#include "debuginfo.h"
26	27	#include "intlist.h"
27	28	#include "strbuf.h"
28	29	#include "strlist.h"
@@ -31,128 +32,9 @@ · 2 unmodified lines
31	32	#include "probe-file.h"
32	33	#include "string2.h"
33	34
34		−#ifdef HAVE_DEBUGINFOD_SUPPORT
35		−#include <elfutils/debuginfod.h>
36		−#endif
37		−
38	35	/* Kprobe tracer basic type is up to u64 */
39	36	#define MAX_BASIC_TYPE_BITS 64
40	37
41		−/* Dwarf FL wrappers */
42		−static char debuginfo_path; / Currently dummy */
43		−
44		−static const Dwfl_Callbacks offline_callbacks = {
45		− .find_debuginfo = dwfl_standard_find_debuginfo,
46		− .debuginfo_path = &debuginfo_path,
47		−
48		− .section_address = dwfl_offline_section_address,
49		−
50		− /* We use this table for core files too. */
51		− .find_elf = dwfl_build_id_find_elf,
52		−};
53		−
54		−/* Get a Dwarf from offline image */
55		−static int debuginfo__init_offline_dwarf(struct debuginfo *dbg,
56		− const char *path)
57		−{
58		− GElf_Addr dummy;
59		− int fd;
60		−
61		− fd = open(path, O_RDONLY);
62		− if (fd < 0)
63		− return fd;
64		−
65		− dbg->dwfl = dwfl_begin(&offline_callbacks);
66		− if (!dbg->dwfl)
67		− goto error;
68		−
69		− dwfl_report_begin(dbg->dwfl);
70		− dbg->mod = dwfl_report_offline(dbg->dwfl, "", "", fd);
71		− if (!dbg->mod)
72		− goto error;
73		−
74		− dbg->dbg = dwfl_module_getdwarf(dbg->mod, &dbg->bias);
75		− if (!dbg->dbg)
76		− goto error;
77		−
78		− dwfl_module_build_id(dbg->mod, &dbg->build_id, &dummy);
79		−
80		− dwfl_report_end(dbg->dwfl, NULL, NULL);
81		−
82		− return 0;
83		−error:
84		− if (dbg->dwfl)
85		− dwfl_end(dbg->dwfl);
86		− else
87		− close(fd);
88		− memset(dbg, 0, sizeof(*dbg));
89		−
90		− return -ENOENT;
91		−}
92		−
93		−static struct debuginfo __debuginfo__new(const char path)
94		−{
95		− struct debuginfo dbg = zalloc(sizeof(dbg));
96		− if (!dbg)
97		− return NULL;
98		−
99		− if (debuginfo__init_offline_dwarf(dbg, path) < 0)
100		− zfree(&dbg);
101		− if (dbg)
102		− pr_debug("Open Debuginfo file: %s\n", path);
103		− return dbg;
104		−}
105		−
106		−enum dso_binary_type distro_dwarf_types[] = {
107		− DSO_BINARY_TYPE__FEDORA_DEBUGINFO,
108		− DSO_BINARY_TYPE__UBUNTU_DEBUGINFO,
109		− DSO_BINARY_TYPE__OPENEMBEDDED_DEBUGINFO,
110		− DSO_BINARY_TYPE__BUILDID_DEBUGINFO,
111		− DSO_BINARY_TYPE__MIXEDUP_UBUNTU_DEBUGINFO,
112		− DSO_BINARY_TYPE__NOT_FOUND,
113		−};
114		−
115		−struct debuginfo debuginfo__new(const char path)
116		−{
117		− enum dso_binary_type *type;
118		− char buf[PATH_MAX], nil = '\0';
119		− struct dso *dso;
120		− struct debuginfo *dinfo = NULL;
121		− struct build_id bid;
122		−
123		− /* Try to open distro debuginfo files */
124		− dso = dso__new(path);
125		− if (!dso)
126		− goto out;
127		−
128		− /* Set the build id for DSO_BINARY_TYPE__BUILDID_DEBUGINFO */
129		− if (is_regular_file(path) && filename__read_build_id(path, &bid) > 0)
130		− dso__set_build_id(dso, &bid);
131		−
132		− for (type = distro_dwarf_types;
133		− !dinfo && *type != DSO_BINARY_TYPE__NOT_FOUND;
134		− type++) {
135		− if (dso__read_binary_type_filename(dso, *type, &nil,
136		− buf, PATH_MAX) < 0)
137		− continue;
138		− dinfo = __debuginfo__new(buf);
139		− }
140		− dso__put(dso);
141		−
142		−out:
143		− /* if failed to open all distro debuginfo, open given binary */
144		− return dinfo ? : __debuginfo__new(path);
145		−}
146		−
147		−void debuginfo__delete(struct debuginfo *dbg)
148		−{
149		− if (dbg) {
150		− if (dbg->dwfl)
151		− dwfl_end(dbg->dwfl);
152		− free(dbg);
153		− }
154		−}
155		−
156	38	/*
157	39	* Probe finder related functions
158	40	*/
@@ -722,7 +604,7 @@ static int call_probe_finder(Dwarf_Die sc_die, struct probe_finder pf) · 563 unmodified lines
722	604	ret = dwarf_getlocation_addr(&fb_attr, pf->addr, &pf->fb_ops, &nops, 1);
723	605	if (ret <= 0 \|\| nops == 0) {
724	606	pf->fb_ops = NULL;
725		−#if _ELFUTILS_PREREQ(0, 142)
	607	+#ifdef HAVE_DWARF_CFI_SUPPORT
726	608	} else if (nops == 1 && pf->fb_ops[0].atom == DW_OP_call_frame_cfa &&
727	609	(pf->cfi_eh != NULL \|\| pf->cfi_dbg != NULL)) {
728	610	if ((dwarf_cfi_addrframe(pf->cfi_eh, pf->addr, &frame) != 0 &&
@@ -733,7 +615,7 @@ static int call_probe_finder(Dwarf_Die sc_die, struct probe_finder pf) · 4 unmodified lines
733	615	free(frame);
734	616	return -ENOENT;
735	617	}
736		−#endif
	618	+#endif /* HAVE_DWARF_CFI_SUPPORT */
737	619	}
738	620
739	621	/* Call finder's callback handler */
@@ -1258,7 +1140,7 @@ static int debuginfo__find_probes(struct debuginfo *dbg, · 518 unmodified lines
1258	1140
1259	1141	pf->machine = ehdr.e_machine;
1260	1142
1261		−#if _ELFUTILS_PREREQ(0, 142)
	1143	+#ifdef HAVE_DWARF_CFI_SUPPORT
1262	1144	do {
1263	1145	GElf_Shdr shdr;
1264	1146
@@ -1268,7 +1150,7 @@ static int debuginfo__find_probes(struct debuginfo *dbg, · 3 unmodified lines
1268	1150
1269	1151	pf->cfi_dbg = dwarf_getcfi(dbg->dbg);
1270	1152	} while (0);
1271		−#endif
	1153	+#endif /* HAVE_DWARF_CFI_SUPPORT */
1272	1154
1273	1155	ret = debuginfo__find_probe_location(dbg, pf);
1274	1156	return ret;
@@ -1677,44 +1559,6 @@ int debuginfo__find_available_vars_at(struct debuginfo *dbg, · 402 unmodified lines
1677	1559	return (ret < 0) ? ret : af.nvls;
1678	1560	}
1679	1561
1680		−/* For the kernel module, we need a special code to get a DIE */
1681		−int debuginfo__get_text_offset(struct debuginfo dbg, Dwarf_Addr offs,
1682		− bool adjust_offset)
1683		−{
1684		− int n, i;
1685		− Elf32_Word shndx;
1686		− Elf_Scn *scn;
1687		− Elf *elf;
1688		− GElf_Shdr mem, *shdr;
1689		− const char *p;
1690		−
1691		− elf = dwfl_module_getelf(dbg->mod, &dbg->bias);
1692		− if (!elf)
1693		− return -EINVAL;
1694		−
1695		− /* Get the number of relocations */
1696		− n = dwfl_module_relocations(dbg->mod);
1697		− if (n < 0)
1698		− return -ENOENT;
1699		− /* Search the relocation related .text section */
1700		− for (i = 0; i < n; i++) {
1701		− p = dwfl_module_relocation_info(dbg->mod, i, &shndx);
1702		− if (strcmp(p, ".text") == 0) {
1703		− /* OK, get the section header */
1704		− scn = elf_getscn(elf, shndx);
1705		− if (!scn)
1706		− return -ENOENT;
1707		− shdr = gelf_getshdr(scn, &mem);
1708		− if (!shdr)
1709		− return -ENOENT;
1710		− *offs = shdr->sh_addr;
1711		− if (adjust_offset)
1712		− *offs -= shdr->sh_offset;
1713		− }
1714		− }
1715		− return 0;
1716		−}
1717		−
1718	1562	/* Reverse search */
1719	1563	int debuginfo__find_probe_point(struct debuginfo *dbg, u64 addr,
1720	1564	struct perf_probe_point *ppt)
@@ -2009,41 +1853,6 @@ found: · 288 unmodified lines
2009	1853	return (ret < 0) ? ret : lf.found;
2010	1854	}
2011	1855
2012		−#ifdef HAVE_DEBUGINFOD_SUPPORT
2013		−/* debuginfod doesn't require the comp_dir but buildid is required */
2014		−static int get_source_from_debuginfod(const char *raw_path,
2015		− const char sbuild_id, char *new_path)
2016		−{
2017		− debuginfod_client *c = debuginfod_begin();
2018		− const char *p = raw_path;
2019		− int fd;
2020		−
2021		− if (!c)
2022		− return -ENOMEM;
2023		−
2024		− fd = debuginfod_find_source(c, (const unsigned char *)sbuild_id,
2025		− 0, p, new_path);
2026		− pr_debug("Search %s from debuginfod -> %d\n", p, fd);
2027		− if (fd >= 0)
2028		− close(fd);
2029		− debuginfod_end(c);
2030		− if (fd < 0) {
2031		− pr_debug("Failed to find %s in debuginfod (%s)\n",
2032		− raw_path, sbuild_id);
2033		− return -ENOENT;
2034		− }
2035		− pr_debug("Got a source %s\n", *new_path);
2036		−
2037		− return 0;
2038		−}
2039		−#else
2040		−static inline int get_source_from_debuginfod(const char *raw_path __maybe_unused,
2041		− const char *sbuild_id __maybe_unused,
2042		− char **new_path __maybe_unused)
2043		−{
2044		− return -ENOTSUP;
2045		−}
2046		−#endif
2047	1856	/*
2048	1857	* Find a src file from a DWARF tag path. Prepend optional source path prefix
2049	1858	* and chop off leading directories that do not exist. Result is passed back as

tools/perf/util/probe-finder.h+1−18

@@ -24,21 +24,7 @@ static inline int is_c_varname(const char *name) · 23 unmodified lines
24	24	#ifdef HAVE_DWARF_SUPPORT
25	25
26	26	#include "dwarf-aux.h"
27		−
28		−/* TODO: export debuginfo data structure even if no dwarf support */
29		−
30		−/* debug information structure */
31		−struct debuginfo {
32		− Dwarf *dbg;
33		− Dwfl_Module *mod;
34		− Dwfl *dwfl;
35		− Dwarf_Addr bias;
36		− const unsigned char *build_id;
37		−};
38		−
39		−/* This also tries to open distro debuginfo */
40		−struct debuginfo debuginfo__new(const char path);
41		−void debuginfo__delete(struct debuginfo *dbg);
	27	+#include "debuginfo.h"
42	28
43	29	/* Find probe_trace_events specified by perf_probe_event from debuginfo */
44	30	int debuginfo__find_trace_events(struct debuginfo *dbg,
@@ -49,9 +35,6 @@ int debuginfo__find_trace_events(struct debuginfo *dbg, · 4 unmodified lines
49	35	int debuginfo__find_probe_point(struct debuginfo *dbg, u64 addr,
50	36	struct perf_probe_point *ppt);
51	37
52		−int debuginfo__get_text_offset(struct debuginfo dbg, Dwarf_Addr offs,
53		− bool adjust_offset);
54		−
55	38	/* Find a line range */
56	39	int debuginfo__find_line_range(struct debuginfo dbg, struct line_range lr);
57	40

tools/perf/util/record.c+2−2

@@ -237,8 +237,8 @@ bool evlist__can_select_event(struct evlist evlist, const char str) · 236 unmodified lines
237	237
238	238	evsel = evlist__last(temp_evlist);
239	239
240		− if (!evlist \|\| perf_cpu_map__empty(evlist->core.user_requested_cpus)) {
241		− struct perf_cpu_map *cpus = perf_cpu_map__new(NULL);
	240	+ if (!evlist \|\| perf_cpu_map__has_any_cpu_or_is_empty(evlist->core.user_requested_cpus)) {
	241	+ struct perf_cpu_map *cpus = perf_cpu_map__new_online_cpus();
242	242
243	243	if (cpus)
244	244	cpu = perf_cpu_map__cpu(cpus, 0);

tools/perf/util/s390-cpumcf-kernel.h+2−0

@@ -12,6 +12,8 @@ · 11 unmodified lines
12	12	#define S390_CPUMCF_DIAG_DEF 0xfeef /* Counter diagnostic entry ID */
13	13	#define PERF_EVENT_CPUM_CF_DIAG 0xBC000 /* Event: Counter sets */
14	14	#define PERF_EVENT_CPUM_SF_DIAG 0xBD000 /* Event: Combined-sampling */
	15	+#define PERF_EVENT_PAI_CRYPTO_ALL 0x1000 /* Event: CRYPTO_ALL */
	16	+#define PERF_EVENT_PAI_NNPA_ALL 0x1800 /* Event: NNPA_ALL */
15	17
16	18	struct cf_ctrset_entry { /* CPU-M CF counter set entry (8 byte) */
17	19	unsigned int def:16; /* 0-15 Data Entry Format */

tools/perf/util/s390-sample-raw.c+104−10

@@ -51,8 +51,6 @@ static bool s390_cpumcfdg_testctr(struct perf_sample *sample) · 50 unmodified lines
51	51	struct cf_trailer_entry *te;
52	52	struct cf_ctrset_entry *cep, ce;
53	53
54		− if (!len)
55		− return false;
56	54	while (offset < len) {
57	55	cep = (struct cf_ctrset_entry *)(buf + offset);
58	56	ce.def = be16_to_cpu(cep->def);
@@ -125,6 +123,9 @@ static int get_counterset_start(int setnr) · 66 unmodified lines
125	123	return 128;
126	124	case CPUMF_CTR_SET_MT_DIAG: /* Diagnostic counter set */
127	125	return 448;
	126	+ case PERF_EVENT_PAI_NNPA_ALL: /* PAI NNPA counter set */
	127	+ case PERF_EVENT_PAI_CRYPTO_ALL: /* PAI CRYPTO counter set */
	128	+ return setnr;
128	129	default:
129	130	return -1;
130	131	}
@@ -212,27 +213,120 @@ static void s390_cpumcfdg_dump(struct perf_pmu pmu, struct perf_sample sample) · 81 unmodified lines
212	213	}
213	214	}
214	215
	216	+#pragma GCC diagnostic push
	217	+#pragma GCC diagnostic ignored "-Wpacked"
	218	+#pragma GCC diagnostic ignored "-Wattributes"
	219	+/*
	220	+ * Check for consistency of PAI_CRYPTO/PAI_NNPA raw data.
	221	+ */
	222	+struct pai_data { /* Event number and value */
	223	+ u16 event_nr;
	224	+ u64 event_val;
	225	+} __packed;
	226	+
	227	+#pragma GCC diagnostic pop
	228	+
	229	+/*
	230	+ * Test for valid raw data. At least one PAI event should be in the raw
	231	+ * data section.
	232	+ */
	233	+static bool s390_pai_all_test(struct perf_sample *sample)
	234	+{
	235	+ size_t len = sample->raw_size;
	236	+
	237	+ if (len < 0xa)
	238	+ return false;
	239	+ return true;
	240	+}
	241	+
	242	+static void s390_pai_all_dump(struct evsel evsel, struct perf_sample sample)
	243	+{
	244	+ size_t len = sample->raw_size, offset = 0;
	245	+ unsigned char *p = sample->raw_data;
	246	+ const char *color = PERF_COLOR_BLUE;
	247	+ struct pai_data pai_data;
	248	+ char *ev_name;
	249	+
	250	+ while (offset < len) {
	251	+ memcpy(&pai_data.event_nr, p, sizeof(pai_data.event_nr));
	252	+ pai_data.event_nr = be16_to_cpu(pai_data.event_nr);
	253	+ p += sizeof(pai_data.event_nr);
	254	+ offset += sizeof(pai_data.event_nr);
	255	+
	256	+ memcpy(&pai_data.event_val, p, sizeof(pai_data.event_val));
	257	+ pai_data.event_val = be64_to_cpu(pai_data.event_val);
	258	+ p += sizeof(pai_data.event_val);
	259	+ offset += sizeof(pai_data.event_val);
	260	+
	261	+ ev_name = get_counter_name(evsel->core.attr.config,
	262	+ pai_data.event_nr, evsel->pmu);
	263	+ color_fprintf(stdout, color, "\tCounter:%03d %s Value:%#018lx\n",
	264	+ pai_data.event_nr, ev_name ?: "<unknown>",
	265	+ pai_data.event_val);
	266	+ free(ev_name);
	267	+
	268	+ if (offset + 0xa > len)
	269	+ break;
	270	+ }
	271	+ color_fprintf(stdout, color, "\n");
	272	+}
	273	+
215	274	/* S390 specific trace event function. Check for PERF_RECORD_SAMPLE events
216		− * and if the event was triggered by a counter set diagnostic event display
217		− * its raw data.
	275	+ * and if the event was triggered by a
	276	+ * - counter set diagnostic event
	277	+ * - processor activity assist (PAI) crypto counter event
	278	+ * - processor activity assist (PAI) neural network processor assist (NNPA)
	279	+ * counter event
	280	+ * display its raw data.
218	281	* The function is only invoked when the dump flag -D is set.
	282	+ *
	283	+ * Function evlist__s390_sample_raw() is defined as call back after it has
	284	+ * been verified that the perf.data file was created on s390 platform.
219	285	*/
220		−void evlist__s390_sample_raw(struct evlist evlist, union perf_event event, struct perf_sample *sample)
	286	+void evlist__s390_sample_raw(struct evlist evlist, union perf_event event,
	287	+ struct perf_sample *sample)
221	288	{
	289	+ const char *pai_name;
222	290	struct evsel *evsel;
223	291
224	292	if (event->header.type != PERF_RECORD_SAMPLE)
225	293	return;
226	294
227	295	evsel = evlist__event2evsel(evlist, event);
228		− if (evsel == NULL \|\|
229		− evsel->core.attr.config != PERF_EVENT_CPUM_CF_DIAG)
	296	+ if (!evsel)
	297	+ return;
	298	+
	299	+ /* Check for raw data in sample */
	300	+ if (!sample->raw_size \|\| !sample->raw_data)
230	301	return;
231	302
232	303	/* Display raw data on screen */
233		− if (!s390_cpumcfdg_testctr(sample)) {
234		− pr_err("Invalid counter set data encountered\n");
	304	+ if (evsel->core.attr.config == PERF_EVENT_CPUM_CF_DIAG) {
	305	+ if (!evsel->pmu)
	306	+ evsel->pmu = perf_pmus__find("cpum_cf");
	307	+ if (!s390_cpumcfdg_testctr(sample))
	308	+ pr_err("Invalid counter set data encountered\n");
	309	+ else
	310	+ s390_cpumcfdg_dump(evsel->pmu, sample);
	311	+ return;
	312	+ }
	313	+
	314	+ switch (evsel->core.attr.config) {
	315	+ case PERF_EVENT_PAI_NNPA_ALL:
	316	+ pai_name = "NNPA_ALL";
	317	+ break;
	318	+ case PERF_EVENT_PAI_CRYPTO_ALL:
	319	+ pai_name = "CRYPTO_ALL";
	320	+ break;
	321	+ default:
235	322	return;
236	323	}
237		− s390_cpumcfdg_dump(evsel->pmu, sample);
	324	+
	325	+ if (!s390_pai_all_test(sample)) {
	326	+ pr_err("Invalid %s raw data encountered\n", pai_name);
	327	+ } else {
	328	+ if (!evsel->pmu)
	329	+ evsel->pmu = perf_pmus__find_by_type(evsel->core.attr.type);
	330	+ s390_pai_all_dump(evsel, sample);
	331	+ }
238	332	}

tools/perf/util/sample.h+1−0

@@ -113,6 +113,7 @@ struct perf_sample { · 112 unmodified lines
113	113	void *raw_data;
114	114	struct ip_callchain *callchain;
115	115	struct branch_stack *branch_stack;
	116	+ u64 *branch_stack_cntr;
116	117	struct regs_dump user_regs;
117	118	struct regs_dump intr_regs;
118	119	struct stack_dump user_stack;

tools/perf/util/scripting-engines/trace-event-perl.c+3−0

@@ -490,6 +490,9 @@ static int perl_start_script(const char script, int argc, const char *argv, · 489 unmodified lines
490	490	scripting_context->session = session;
491	491
492	492	command_line = malloc((argc + 2) * sizeof(const char *));
	493	+ if (!command_line)
	494	+ return -ENOMEM;
	495	+
493	496	command_line[0] = "";
494	497	command_line[1] = script;
495	498	for (i = 2; i < argc + 2; i++)

tools/perf/util/scripting-engines/trace-event-python.c+15−2

@@ -353,6 +353,8 @@ static PyObject get_field_numeric_entry(struct tep_event event, · 352 unmodified lines
353	353
354	354	if (is_array) {
355	355	list = PyList_New(field->arraylen);
	356	+ if (!list)
	357	+ Py_FatalError("couldn't create Python list");
356	358	item_size = field->size / field->arraylen;
357	359	n_items = field->arraylen;
358	360	} else {
@@ -754,7 +756,7 @@ static void regs_map(struct regs_dump regs, uint64_t mask, const char arch, ch · 395 unmodified lines
754	756	}
755	757	}
756	758
757		−static void set_regs_in_dict(PyObject *dict,
	759	+static int set_regs_in_dict(PyObject *dict,
758	760	struct perf_sample *sample,
759	761	struct evsel *evsel)
760	762	{
@@ -770,6 +772,8 @@ static void set_regs_in_dict(PyObject *dict, · 9 unmodified lines
770	772	*/
771	773	int size = __sw_hweight64(attr->sample_regs_intr) * 28;
772	774	char *bf = malloc(size);
	775	+ if (!bf)
	776	+ return -1;
773	777
774	778	regs_map(&sample->intr_regs, attr->sample_regs_intr, arch, bf, size);
775	779
@@ -781,6 +785,8 @@ static void set_regs_in_dict(PyObject *dict, · 5 unmodified lines
781	785	pydict_set_item_string_decref(dict, "uregs",
782	786	_PyUnicode_FromString(bf));
783	787	free(bf);
	788	+
	789	+ return 0;
784	790	}
785	791
786	792	static void set_sym_in_dict(PyObject dict, struct addr_location al,
@@ -920,7 +926,8 @@ static PyObject get_perf_sample_dict(struct perf_sample sample, · 133 unmodified lines
920	926	PyLong_FromUnsignedLongLong(sample->cyc_cnt));
921	927	}
922	928
923		− set_regs_in_dict(dict, sample, evsel);
	929	+ if (set_regs_in_dict(dict, sample, evsel))
	930	+ Py_FatalError("Failed to setting regs in dict");
924	931
925	932	return dict;
926	933	}
@@ -1918,12 +1925,18 @@ static int python_start_script(const char script, int argc, const char *argv, · 991 unmodified lines
1918	1925	scripting_context->session = session;
1919	1926	#if PY_MAJOR_VERSION < 3
1920	1927	command_line = malloc((argc + 1) * sizeof(const char *));
	1928	+ if (!command_line)
	1929	+ return -1;
	1930	+
1921	1931	command_line[0] = script;
1922	1932	for (i = 1; i < argc + 1; i++)
1923	1933	command_line[i] = argv[i - 1];
1924	1934	PyImport_AppendInittab(name, initperf_trace_context);
1925	1935	#else
1926	1936	command_line = malloc((argc + 1) * sizeof(wchar_t *));
	1937	+ if (!command_line)
	1938	+ return -1;
	1939	+
1927	1940	command_line[0] = Py_DecodeLocale(script, NULL);
1928	1941	for (i = 1; i < argc + 1; i++)
1929	1942	command_line[i] = Py_DecodeLocale(argv[i - 1], NULL);

tools/perf/util/session.c+18−2

@@ -115,6 +115,11 @@ static int perf_session__open(struct perf_session *session, int repipe_fd) · 114 unmodified lines
115	115	return -1;
116	116	}
117	117
	118	+ if (perf_header__has_feat(&session->header, HEADER_AUXTRACE)) {
	119	+ /* Auxiliary events may reference exited threads, hold onto dead ones. */
	120	+ symbol_conf.keep_exited_threads = true;
	121	+ }
	122	+
118	123	if (perf_data__is_pipe(data))
119	124	return 0;
120	125
@@ -1150,9 +1155,13 @@ static void callchain__printf(struct evsel *evsel, · 1029 unmodified lines
1150	1155	i, callchain->ips[i]);
1151	1156	}
1152	1157
1153		−static void branch_stack__printf(struct perf_sample *sample, bool callstack)
	1158	+static void branch_stack__printf(struct perf_sample *sample,
	1159	+ struct evsel *evsel)
1154	1160	{
1155	1161	struct branch_entry *entries = perf_sample__branch_entries(sample);
	1162	+ bool callstack = evsel__has_branch_callstack(evsel);
	1163	+ u64 *branch_stack_cntr = sample->branch_stack_cntr;
	1164	+ struct perf_env *env = evsel__env(evsel);
1156	1165	uint64_t i;
1157	1166
1158	1167	if (!callstack) {
@@ -1194,6 +1203,13 @@ static void branch_stack__printf(struct perf_sample *sample, bool callstack) · 35 unmodified lines
1194	1203	}
1195	1204	}
1196	1205	}
	1206	+
	1207	+ if (branch_stack_cntr) {
	1208	+ printf("... branch stack counters: nr:%" PRIu64 " (counter width: %u max counter nr:%u)\n",
	1209	+ sample->branch_stack->nr, env->br_cntr_width, env->br_cntr_nr);
	1210	+ for (i = 0; i < sample->branch_stack->nr; i++)
	1211	+ printf("..... %2"PRIu64": %016" PRIx64 "\n", i, branch_stack_cntr[i]);
	1212	+ }
1197	1213	}
1198	1214
1199	1215	static void regs_dump__printf(u64 mask, u64 regs, const char arch)
@@ -1355,7 +1371,7 @@ static void dump_sample(struct evsel evsel, union perf_event event, · 155 unmodified lines
1355	1371	callchain__printf(evsel, sample);
1356	1372
1357	1373	if (evsel__has_br_stack(evsel))
1358		− branch_stack__printf(sample, evsel__has_branch_callstack(evsel));
	1374	+ branch_stack__printf(sample, evsel);
1359	1375
1360	1376	if (sample_type & PERF_SAMPLE_REGS_USER)
1361	1377	regs_user__printf(sample, arch);

tools/perf/util/sort.c+207−9

@@ -24,6 +24,7 @@ · 23 unmodified lines
24	24	#include "strbuf.h"
25	25	#include "mem-events.h"
26	26	#include "annotate.h"
	27	+#include "annotate-data.h"
27	28	#include "event.h"
28	29	#include "time-utils.h"
29	30	#include "cgroup.h"
@@ -418,6 +419,52 @@ struct sort_entry sort_sym = { · 388 unmodified lines
418	419	.se_width_idx = HISTC_SYMBOL,
419	420	};
420	421
	422	+/* --sort symoff */
	423	+
	424	+static int64_t
	425	+sort__symoff_cmp(struct hist_entry left, struct hist_entry right)
	426	+{
	427	+ int64_t ret;
	428	+
	429	+ ret = sort__sym_cmp(left, right);
	430	+ if (ret)
	431	+ return ret;
	432	+
	433	+ return left->ip - right->ip;
	434	+}
	435	+
	436	+static int64_t
	437	+sort__symoff_sort(struct hist_entry left, struct hist_entry right)
	438	+{
	439	+ int64_t ret;
	440	+
	441	+ ret = sort__sym_sort(left, right);
	442	+ if (ret)
	443	+ return ret;
	444	+
	445	+ return left->ip - right->ip;
	446	+}
	447	+
	448	+static int
	449	+hist_entry__symoff_snprintf(struct hist_entry he, char bf, size_t size, unsigned int width)
	450	+{
	451	+ struct symbol *sym = he->ms.sym;
	452	+
	453	+ if (sym == NULL)
	454	+ return repsep_snprintf(bf, size, "[%c] %-#.*llx", he->level, width - 4, he->ip);
	455	+
	456	+ return repsep_snprintf(bf, size, "[%c] %s+0x%llx", he->level, sym->name, he->ip - sym->start);
	457	+}
	458	+
	459	+struct sort_entry sort_sym_offset = {
	460	+ .se_header = "Symbol Offset",
	461	+ .se_cmp = sort__symoff_cmp,
	462	+ .se_sort = sort__symoff_sort,
	463	+ .se_snprintf = hist_entry__symoff_snprintf,
	464	+ .se_filter = hist_entry__sym_filter,
	465	+ .se_width_idx = HISTC_SYMBOL_OFFSET,
	466	+};
	467	+
421	468	/* --sort srcline */
422	469
423	470	char hist_entry__srcline(struct hist_entry he)
@@ -583,21 +630,21 @@ static int hist_entry__sym_ipc_snprintf(struct hist_entry he, char bf, · 159 unmodified lines
583	630	{
584	631
585	632	struct symbol *sym = he->ms.sym;
586		− struct annotation *notes;
	633	+ struct annotated_branch *branch;
587	634	double ipc = 0.0, coverage = 0.0;
588	635	char tmp[64];
589	636
590	637	if (!sym)
591	638	return repsep_snprintf(bf, size, "%-*s", width, "-");
592	639
593		− notes = symbol__annotation(sym);
	640	+ branch = symbol__annotation(sym)->branch;
594	641
595		− if (notes->hit_cycles)
596		− ipc = notes->hit_insn / ((double)notes->hit_cycles);
	642	+ if (branch && branch->hit_cycles)
	643	+ ipc = branch->hit_insn / ((double)branch->hit_cycles);
597	644
598		− if (notes->total_insn) {
599		− coverage = notes->cover_insn * 100.0 /
600		− ((double)notes->total_insn);
	645	+ if (branch && branch->total_insn) {
	646	+ coverage = branch->cover_insn * 100.0 /
	647	+ ((double)branch->total_insn);
601	648	}
602	649
603	650	snprintf(tmp, sizeof(tmp), "%-5.2f [%5.1f%%]", ipc, coverage);
@@ -2094,7 +2141,7 @@ struct sort_entry sort_dso_size = { · 1490 unmodified lines
2094	2141	.se_width_idx = HISTC_DSO_SIZE,
2095	2142	};
2096	2143
2097		−/* --sort dso_size */
	2144	+/* --sort addr */
2098	2145
2099	2146	static int64_t
2100	2147	sort__addr_cmp(struct hist_entry left, struct hist_entry right)
@@ -2131,6 +2178,152 @@ struct sort_entry sort_addr = { · 30 unmodified lines
2131	2178	.se_width_idx = HISTC_ADDR,
2132	2179	};
2133	2180
	2181	+/* --sort type */
	2182	+
	2183	+struct annotated_data_type unknown_type = {
	2184	+ .self = {
	2185	+ .type_name = (char *)"(unknown)",
	2186	+ .children = LIST_HEAD_INIT(unknown_type.self.children),
	2187	+ },
	2188	+};
	2189	+
	2190	+static int64_t
	2191	+sort__type_cmp(struct hist_entry left, struct hist_entry right)
	2192	+{
	2193	+ return sort__addr_cmp(left, right);
	2194	+}
	2195	+
	2196	+static void sort__type_init(struct hist_entry *he)
	2197	+{
	2198	+ if (he->mem_type)
	2199	+ return;
	2200	+
	2201	+ he->mem_type = hist_entry__get_data_type(he);
	2202	+ if (he->mem_type == NULL) {
	2203	+ he->mem_type = &unknown_type;
	2204	+ he->mem_type_off = 0;
	2205	+ }
	2206	+}
	2207	+
	2208	+static int64_t
	2209	+sort__type_collapse(struct hist_entry left, struct hist_entry right)
	2210	+{
	2211	+ struct annotated_data_type *left_type = left->mem_type;
	2212	+ struct annotated_data_type *right_type = right->mem_type;
	2213	+
	2214	+ if (!left_type) {
	2215	+ sort__type_init(left);
	2216	+ left_type = left->mem_type;
	2217	+ }
	2218	+
	2219	+ if (!right_type) {
	2220	+ sort__type_init(right);
	2221	+ right_type = right->mem_type;
	2222	+ }
	2223	+
	2224	+ return strcmp(left_type->self.type_name, right_type->self.type_name);
	2225	+}
	2226	+
	2227	+static int64_t
	2228	+sort__type_sort(struct hist_entry left, struct hist_entry right)
	2229	+{
	2230	+ return sort__type_collapse(left, right);
	2231	+}
	2232	+
	2233	+static int hist_entry__type_snprintf(struct hist_entry he, char bf,
	2234	+ size_t size, unsigned int width)
	2235	+{
	2236	+ return repsep_snprintf(bf, size, "%-*s", width, he->mem_type->self.type_name);
	2237	+}
	2238	+
	2239	+struct sort_entry sort_type = {
	2240	+ .se_header = "Data Type",
	2241	+ .se_cmp = sort__type_cmp,
	2242	+ .se_collapse = sort__type_collapse,
	2243	+ .se_sort = sort__type_sort,
	2244	+ .se_init = sort__type_init,
	2245	+ .se_snprintf = hist_entry__type_snprintf,
	2246	+ .se_width_idx = HISTC_TYPE,
	2247	+};
	2248	+
	2249	+/* --sort typeoff */
	2250	+
	2251	+static int64_t
	2252	+sort__typeoff_sort(struct hist_entry left, struct hist_entry right)
	2253	+{
	2254	+ struct annotated_data_type *left_type = left->mem_type;
	2255	+ struct annotated_data_type *right_type = right->mem_type;
	2256	+ int64_t ret;
	2257	+
	2258	+ if (!left_type) {
	2259	+ sort__type_init(left);
	2260	+ left_type = left->mem_type;
	2261	+ }
	2262	+
	2263	+ if (!right_type) {
	2264	+ sort__type_init(right);
	2265	+ right_type = right->mem_type;
	2266	+ }
	2267	+
	2268	+ ret = strcmp(left_type->self.type_name, right_type->self.type_name);
	2269	+ if (ret)
	2270	+ return ret;
	2271	+ return left->mem_type_off - right->mem_type_off;
	2272	+}
	2273	+
	2274	+static void fill_member_name(char buf, size_t sz, struct annotated_member m,
	2275	+ int offset, bool first)
	2276	+{
	2277	+ struct annotated_member *child;
	2278	+
	2279	+ if (list_empty(&m->children))
	2280	+ return;
	2281	+
	2282	+ list_for_each_entry(child, &m->children, node) {
	2283	+ if (child->offset <= offset && offset < child->offset + child->size) {
	2284	+ int len = 0;
	2285	+
	2286	+ /* It can have anonymous struct/union members */
	2287	+ if (child->var_name) {
	2288	+ len = scnprintf(buf, sz, "%s%s",
	2289	+ first ? "" : ".", child->var_name);
	2290	+ first = false;
	2291	+ }
	2292	+
	2293	+ fill_member_name(buf + len, sz - len, child, offset, first);
	2294	+ return;
	2295	+ }
	2296	+ }
	2297	+}
	2298	+
	2299	+static int hist_entry__typeoff_snprintf(struct hist_entry he, char bf,
	2300	+ size_t size, unsigned int width __maybe_unused)
	2301	+{
	2302	+ struct annotated_data_type *he_type = he->mem_type;
	2303	+ char buf[4096];
	2304	+
	2305	+ buf[0] = '\0';
	2306	+ if (list_empty(&he_type->self.children))
	2307	+ snprintf(buf, sizeof(buf), "no field");
	2308	+ else
	2309	+ fill_member_name(buf, sizeof(buf), &he_type->self,
	2310	+ he->mem_type_off, true);
	2311	+ buf[4095] = '\0';
	2312	+
	2313	+ return repsep_snprintf(bf, size, "%s %+d (%s)", he_type->self.type_name,
	2314	+ he->mem_type_off, buf);
	2315	+}
	2316	+
	2317	+struct sort_entry sort_type_offset = {
	2318	+ .se_header = "Data Type Offset",
	2319	+ .se_cmp = sort__type_cmp,
	2320	+ .se_collapse = sort__typeoff_sort,
	2321	+ .se_sort = sort__typeoff_sort,
	2322	+ .se_init = sort__type_init,
	2323	+ .se_snprintf = hist_entry__typeoff_snprintf,
	2324	+ .se_width_idx = HISTC_TYPE_OFFSET,
	2325	+};
	2326	+
2134	2327
2135	2328	struct sort_dimension {
2136	2329	const char *name;
@@ -2185,7 +2378,10 @@ static struct sort_dimension common_sort_dimensions[] = { · 48 unmodified lines
2185	2378	DIM(SORT_ADDR, "addr", sort_addr),
2186	2379	DIM(SORT_LOCAL_RETIRE_LAT, "local_retire_lat", sort_local_p_stage_cyc),
2187	2380	DIM(SORT_GLOBAL_RETIRE_LAT, "retire_lat", sort_global_p_stage_cyc),
2188		− DIM(SORT_SIMD, "simd", sort_simd)
	2381	+ DIM(SORT_SIMD, "simd", sort_simd),
	2382	+ DIM(SORT_ANNOTATE_DATA_TYPE, "type", sort_type),
	2383	+ DIM(SORT_ANNOTATE_DATA_TYPE_OFFSET, "typeoff", sort_type_offset),
	2384	+ DIM(SORT_SYM_OFFSET, "symoff", sort_sym_offset),
2189	2385	};
2190	2386
2191	2387	#undef DIM
@@ -3205,6 +3401,8 @@ int sort_dimension__add(struct perf_hpp_list list, const char tok, · 1013 unmodified lines
3205	3401	list->thread = 1;
3206	3402	} else if (sd->entry == &sort_comm) {
3207	3403	list->comm = 1;
	3404	+ } else if (sd->entry == &sort_type_offset) {
	3405	+ symbol_conf.annotate_data_member = true;
3208	3406	}
3209	3407
3210	3408	return __sort_dimension__add(sd, list, level);

tools/perf/util/sort.h+7−0

@@ -15,6 +15,7 @@ · 14 unmodified lines
15	15
16	16	struct option;
17	17	struct thread;
	18	+struct annotated_data_type;
18	19
19	20	extern regex_t parent_regex;
20	21	extern const char *sort_order;
@@ -34,6 +35,7 @@ extern struct sort_entry sort_dso_to; · 13 unmodified lines
34	35	extern struct sort_entry sort_sym_from;
35	36	extern struct sort_entry sort_sym_to;
36	37	extern struct sort_entry sort_srcline;
	38	+extern struct sort_entry sort_type;
37	39	extern const char default_mem_sort_order[];
38	40	extern bool chk_double_cl;
39	41
@@ -111,6 +113,7 @@ struct hist_entry { · 71 unmodified lines
111	113	u64 p_stage_cyc;
112	114	u8 cpumode;
113	115	u8 depth;
	116	+ int mem_type_off;
114	117	struct simd_flags simd_flags;
115	118
116	119	/* We are added by hists__add_dummy_entry. */
@@ -154,6 +157,7 @@ struct hist_entry { · 37 unmodified lines
154	157	struct perf_hpp_list *hpp_list;
155	158	struct hist_entry *parent_he;
156	159	struct hist_entry_ops *ops;
	160	+ struct annotated_data_type *mem_type;
157	161	union {
158	162	/* this is for hierarchical entry structure */
159	163	struct {
@@ -243,6 +247,9 @@ enum sort_type { · 83 unmodified lines
243	247	SORT_LOCAL_RETIRE_LAT,
244	248	SORT_GLOBAL_RETIRE_LAT,
245	249	SORT_SIMD,
	250	+ SORT_ANNOTATE_DATA_TYPE,
	251	+ SORT_ANNOTATE_DATA_TYPE_OFFSET,
	252	+ SORT_SYM_OFFSET,
246	253
247	254	/* branch stack specific sort keys */
248	255	__SORT_BRANCH_STACK,

tools/perf/util/stat-display.c+1−1

@@ -898,7 +898,7 @@ static bool hybrid_uniquify(struct evsel evsel, struct perf_stat_config config · 897 unmodified lines
898	898
899	899	static void uniquify_counter(struct perf_stat_config config, struct evsel counter)
900	900	{
901		− if (config->no_merge \|\| hybrid_uniquify(counter, config))
	901	+ if (config->aggr_mode == AGGR_NONE \|\| hybrid_uniquify(counter, config))
902	902	uniquify_event_name(counter);
903	903	}
904	904

tools/perf/util/stat-shadow.c+1−1

@@ -264,7 +264,7 @@ static void print_ll_miss(struct perf_stat_config *config, · 263 unmodified lines
264	264	static const double color_ratios[3] = {20.0, 10.0, 5.0};
265	265
266	266	print_ratio(config, evsel, aggr_idx, misses, out, STAT_LL_CACHE, color_ratios,
267		− "of all L1-icache accesses");
	267	+ "of all LL-cache accesses");
268	268	}
269	269
270	270	static void print_dtlb_miss(struct perf_stat_config *config,

tools/perf/util/stat.c+2−2

@@ -315,7 +315,7 @@ static int check_per_pkg(struct evsel counter, struct perf_counts_values vals, · 314 unmodified lines
315	315	if (!counter->per_pkg)
316	316	return 0;
317	317
318		− if (perf_cpu_map__empty(cpus))
	318	+ if (perf_cpu_map__has_any_cpu_or_is_empty(cpus))
319	319	return 0;
320	320
321	321	if (!mask) {
@@ -592,7 +592,7 @@ void perf_stat_merge_counters(struct perf_stat_config config, struct evlist ev · 270 unmodified lines
592	592	{
593	593	struct evsel *evsel;
594	594
595		− if (config->no_merge)
	595	+ if (config->aggr_mode == AGGR_NONE)
596	596	return;
597	597
598	598	evlist__for_each_entry(evlist, evsel)

tools/perf/util/stat.h+0−1

@@ -76,7 +76,6 @@ struct perf_stat_config { · 75 unmodified lines
76	76	bool null_run;
77	77	bool ru_display;
78	78	bool big_num;
79		− bool no_merge;
80	79	bool hybrid_merge;
81	80	bool walltime_run_table;
82	81	bool all_kernel;

tools/perf/util/symbol-elf.c+2−4

@@ -1392,8 +1392,7 @@ static int dso__process_kernel_symbol(struct dso dso, struct map map, · 1391 unmodified lines
1392	1392	map__set_start(map, shdr->sh_addr + ref_reloc(kmap));
1393	1393	map__set_end(map, map__start(map) + shdr->sh_size);
1394	1394	map__set_pgoff(map, shdr->sh_offset);
1395		− map__set_map_ip(map, map__dso_map_ip);
1396		− map__set_unmap_ip(map, map__dso_unmap_ip);
	1395	+ map__set_mapping_type(map, MAPPING_TYPE__DSO);
1397	1396	/* Ensure maps are correctly ordered */
1398	1397	if (kmaps) {
1399	1398	int err;
@@ -1455,8 +1454,7 @@ static int dso__process_kernel_symbol(struct dso dso, struct map map, · 55 unmodified lines
1455	1454	map__set_end(curr_map, map__start(curr_map) + shdr->sh_size);
1456	1455	map__set_pgoff(curr_map, shdr->sh_offset);
1457	1456	} else {
1458		− map__set_map_ip(curr_map, identity__map_ip);
1459		− map__set_unmap_ip(curr_map, identity__map_ip);
	1457	+ map__set_mapping_type(curr_map, MAPPING_TYPE__IDENTITY);
1460	1458	}
1461	1459	curr_dso->symtab_type = dso->symtab_type;
1462	1460	if (maps__insert(kmaps, curr_map))

tools/perf/util/symbol-minimal.c+6−4

@@ -159,9 +159,10 @@ int filename__read_build_id(const char filename, struct build_id bid) · 158 unmodified lines
159	159	goto out_free;
160	160
161	161	ret = read_build_id(buf, buf_size, bid, need_swap);
162		− if (ret == 0)
	162	+ if (ret == 0) {
163	163	ret = bid->size;
164		− break;
	164	+ break;
	165	+ }
165	166	}
166	167	} else {
167	168	Elf64_Ehdr ehdr;
@@ -210,9 +211,10 @@ int filename__read_build_id(const char filename, struct build_id bid) · 42 unmodified lines
210	211	goto out_free;
211	212
212	213	ret = read_build_id(buf, buf_size, bid, need_swap);
213		− if (ret == 0)
	214	+ if (ret == 0) {
214	215	ret = bid->size;
215		− break;
	216	+ break;
	217	+ }
216	218	}
217	219	}
218	220	out_free:

tools/perf/util/symbol.c+29−277

@@ -48,11 +48,6 @@ static bool symbol__is_idle(const char *name); · 47 unmodified lines
48	48	int vmlinux_path__nr_entries;
49	49	char **vmlinux_path;
50	50
51		−struct map_list_node {
52		− struct list_head node;
53		− struct map *map;
54		−};
55		−
56	51	struct symbol_conf symbol_conf = {
57	52	.nanosecs = false,
58	53	.use_modules = true,
@@ -90,11 +85,6 @@ static enum dso_binary_type binary_type_symtab[] = { · 31 unmodified lines
90	85
91	86	#define DSO_BINARY_TYPE__SYMTAB_CNT ARRAY_SIZE(binary_type_symtab)
92	87
93		−static struct map_list_node *map_list_node__new(void)
94		−{
95		− return malloc(sizeof(struct map_list_node));
96		−}
97		−
98	88	static bool symbol_type__filter(char symbol_type)
99	89	{
100	90	symbol_type = toupper(symbol_type);
@@ -270,29 +260,6 @@ void symbols__fixup_end(struct rb_root_cached *symbols, bool is_kallsyms) · 169 unmodified lines
270	260	curr->end = roundup(curr->start, 4096) + 4096;
271	261	}
272	262
273		−void maps__fixup_end(struct maps *maps)
274		−{
275		− struct map_rb_node prev = NULL, curr;
276		−
277		− down_write(maps__lock(maps));
278		−
279		− maps__for_each_entry(maps, curr) {
280		− if (prev != NULL && !map__end(prev->map))
281		− map__set_end(prev->map, map__start(curr->map));
282		−
283		− prev = curr;
284		− }
285		−
286		− /*
287		− * We still haven't the actual symbols, so guess the
288		− * last map final address.
289		− */
290		− if (curr && !map__end(curr->map))
291		− map__set_end(curr->map, ~0ULL);
292		−
293		− up_write(maps__lock(maps));
294		−}
295		−
296	263	struct symbol symbol__new(u64 start, u64 len, u8 binding, u8 type, const char name)
297	264	{
298	265	size_t namelen = strlen(name) + 1;
@@ -956,8 +923,7 @@ static int maps__split_kallsyms(struct maps kmaps, struct dso dso, u64 delta, · 657 unmodified lines
956	923	return -1;
957	924	}
958	925
959		− map__set_map_ip(curr_map, identity__map_ip);
960		− map__set_unmap_ip(curr_map, identity__map_ip);
	926	+ map__set_mapping_type(curr_map, MAPPING_TYPE__IDENTITY);
961	927	if (maps__insert(kmaps, curr_map)) {
962	928	dso__put(ndso);
963	929	return -1;
@@ -1148,33 +1114,35 @@ out_delete_from: · 184 unmodified lines
1148	1114	return ret;
1149	1115	}
1150	1116
	1117	+static int do_validate_kcore_modules_cb(struct map old_map, void data)
	1118	+{
	1119	+ struct rb_root *modules = data;
	1120	+ struct module_info *mi;
	1121	+ struct dso *dso;
	1122	+
	1123	+ if (!__map__is_kmodule(old_map))
	1124	+ return 0;
	1125	+
	1126	+ dso = map__dso(old_map);
	1127	+ /* Module must be in memory at the same address */
	1128	+ mi = find_module(dso->short_name, modules);
	1129	+ if (!mi \|\| mi->start != map__start(old_map))
	1130	+ return -EINVAL;
	1131	+
	1132	+ return 0;
	1133	+}
	1134	+
1151	1135	static int do_validate_kcore_modules(const char filename, struct maps kmaps)
1152	1136	{
1153	1137	struct rb_root modules = RB_ROOT;
1154		− struct map_rb_node *old_node;
1155	1138	int err;
1156	1139
1157	1140	err = read_proc_modules(filename, &modules);
1158	1141	if (err)
1159	1142	return err;
1160	1143
1161		− maps__for_each_entry(kmaps, old_node) {
1162		− struct map *old_map = old_node->map;
1163		− struct module_info *mi;
1164		− struct dso *dso;
	1144	+ err = maps__for_each_map(kmaps, do_validate_kcore_modules_cb, &modules);
1165	1145
1166		− if (!__map__is_kmodule(old_map)) {
1167		− continue;
1168		− }
1169		− dso = map__dso(old_map);
1170		− /* Module must be in memory at the same address */
1171		− mi = find_module(dso->short_name, &modules);
1172		− if (!mi \|\| mi->start != map__start(old_map)) {
1173		− err = -EINVAL;
1174		− goto out;
1175		− }
1176		− }
1177		−out:
1178	1146	delete_modules(&modules);
1179	1147	return err;
1180	1148	}
@@ -1271,101 +1239,15 @@ static int kcore_mapfn(u64 start, u64 len, u64 pgoff, void *data) · 90 unmodified lines
1271	1239	return 0;
1272	1240	}
1273	1241
1274		−/*
1275		− * Merges map into maps by splitting the new map within the existing map
1276		− * regions.
1277		− */
1278		−int maps__merge_in(struct maps kmaps, struct map new_map)
	1242	+static bool remove_old_maps(struct map map, void data)
1279	1243	{
1280		− struct map_rb_node *rb_node;
1281		− LIST_HEAD(merged);
1282		− int err = 0;
1283		−
1284		− maps__for_each_entry(kmaps, rb_node) {
1285		− struct map *old_map = rb_node->map;
1286		−
1287		− /* no overload with this one */
1288		− if (map__end(new_map) < map__start(old_map) \|\|
1289		− map__start(new_map) >= map__end(old_map))
1290		− continue;
1291		−
1292		− if (map__start(new_map) < map__start(old_map)) {
1293		− /*
1294		− * \|new......
1295		− * \|old....
1296		− */
1297		− if (map__end(new_map) < map__end(old_map)) {
1298		− /*
1299		− * \|new......\| -> \|new..\|
1300		− * \|old....\| -> \|old....\|
1301		− */
1302		− map__set_end(new_map, map__start(old_map));
1303		− } else {
1304		− /*
1305		− * \|new.............\| -> \|new..\| \|new..\|
1306		− * \|old....\| -> \|old....\|
1307		− */
1308		− struct map_list_node *m = map_list_node__new();
1309		−
1310		− if (!m) {
1311		− err = -ENOMEM;
1312		− goto out;
1313		− }
1314		−
1315		− m->map = map__clone(new_map);
1316		− if (!m->map) {
1317		− free(m);
1318		− err = -ENOMEM;
1319		− goto out;
1320		− }
1321		−
1322		− map__set_end(m->map, map__start(old_map));
1323		− list_add_tail(&m->node, &merged);
1324		− map__add_pgoff(new_map, map__end(old_map) - map__start(new_map));
1325		− map__set_start(new_map, map__end(old_map));
1326		− }
1327		− } else {
1328		− /*
1329		− * \|new......
1330		− * \|old....
1331		− */
1332		− if (map__end(new_map) < map__end(old_map)) {
1333		− /*
1334		− * \|new..\| -> x
1335		− * \|old.........\| -> \|old.........\|
1336		− */
1337		− map__put(new_map);
1338		− new_map = NULL;
1339		− break;
1340		− } else {
1341		− /*
1342		− * \|new......\| -> \|new...\|
1343		− * \|old....\| -> \|old....\|
1344		− */
1345		− map__add_pgoff(new_map, map__end(old_map) - map__start(new_map));
1346		− map__set_start(new_map, map__end(old_map));
1347		− }
1348		− }
1349		− }
1350		−
1351		−out:
1352		− while (!list_empty(&merged)) {
1353		− struct map_list_node *old_node;
1354		−
1355		− old_node = list_entry(merged.next, struct map_list_node, node);
1356		− list_del_init(&old_node->node);
1357		− if (!err)
1358		− err = maps__insert(kmaps, old_node->map);
1359		− map__put(old_node->map);
1360		− free(old_node);
1361		− }
	1244	+ const struct map *map_to_save = data;
1362	1245
1363		− if (new_map) {
1364		− if (!err)
1365		− err = maps__insert(kmaps, new_map);
1366		− map__put(new_map);
1367		− }
1368		− return err;
	1246	+ /*
	1247	+ * We need to preserve eBPF maps even if they are covered by kcore,
	1248	+ * because we need to access eBPF dso for source data.
	1249	+ */
	1250	+ return !RC_CHK_EQUAL(map, map_to_save) && !__map__is_bpf_prog(map);
1369	1251	}
1370	1252
1371	1253	static int dso__load_kcore(struct dso dso, struct map map,
@@ -1374,7 +1256,6 @@ static int dso__load_kcore(struct dso dso, struct map map, · 2 unmodified lines
1374	1256	struct maps *kmaps = map__kmaps(map);
1375	1257	struct kcore_mapfn_data md;
1376	1258	struct map *replacement_map = NULL;
1377		− struct map_rb_node old_node, next;
1378	1259	struct machine *machine;
1379	1260	bool is_64_bit;
1380	1261	int err, fd;
@@ -1421,17 +1302,7 @@ static int dso__load_kcore(struct dso dso, struct map map, · 40 unmodified lines
1421	1302	}
1422	1303
1423	1304	/* Remove old maps */
1424		− maps__for_each_entry_safe(kmaps, old_node, next) {
1425		− struct map *old_map = old_node->map;
1426		−
1427		− /*
1428		− * We need to preserve eBPF maps even if they are
1429		− * covered by kcore, because we need to access
1430		− * eBPF dso for source data.
1431		− */
1432		− if (old_map != map && !__map__is_bpf_prog(old_map))
1433		− maps__remove(kmaps, old_map);
1434		− }
	1305	+ maps__remove_maps(kmaps, remove_old_maps, map);
1435	1306	machine->trampolines_mapped = false;
1436	1307
1437	1308	/* Find the kernel map using the '_stext' symbol */
@@ -1475,8 +1346,7 @@ static int dso__load_kcore(struct dso dso, struct map map, · 37 unmodified lines
1475	1346	map__set_start(map, map__start(new_map));
1476	1347	map__set_end(map, map__end(new_map));
1477	1348	map__set_pgoff(map, map__pgoff(new_map));
1478		− map__set_map_ip(map, map__map_ip_ptr(new_map));
1479		− map__set_unmap_ip(map, map__unmap_ip_ptr(new_map));
	1349	+ map__set_mapping_type(map, map__mapping_type(new_map));
1480	1350	/* Ensure maps are correctly ordered */
1481	1351	map_ref = map__get(map);
1482	1352	maps__remove(kmaps, map_ref);
@@ -2067,124 +1937,6 @@ out: · 584 unmodified lines
2067	1937	return ret;
2068	1938	}
2069	1939
2070		−static int map__strcmp(const void a, const void b)
2071		−{
2072		− const struct map map_a = (const struct map **)a;
2073		− const struct map map_b = (const struct map **)b;
2074		− const struct dso *dso_a = map__dso(map_a);
2075		− const struct dso *dso_b = map__dso(map_b);
2076		− int ret = strcmp(dso_a->short_name, dso_b->short_name);
2077		−
2078		− if (ret == 0 && map_a != map_b) {
2079		− /*
2080		− * Ensure distinct but name equal maps have an order in part to
2081		− * aid reference counting.
2082		− */
2083		− ret = (int)map__start(map_a) - (int)map__start(map_b);
2084		− if (ret == 0)
2085		− ret = (int)((intptr_t)map_a - (intptr_t)map_b);
2086		− }
2087		−
2088		− return ret;
2089		−}
2090		−
2091		−static int map__strcmp_name(const void name, const void b)
2092		−{
2093		− const struct dso dso = map__dso((const struct map **)b);
2094		−
2095		− return strcmp(name, dso->short_name);
2096		−}
2097		−
2098		−void __maps__sort_by_name(struct maps *maps)
2099		−{
2100		− qsort(maps__maps_by_name(maps), maps__nr_maps(maps), sizeof(struct map *), map__strcmp);
2101		−}
2102		−
2103		−static int map__groups__sort_by_name_from_rbtree(struct maps *maps)
2104		−{
2105		− struct map_rb_node *rb_node;
2106		− struct map **maps_by_name = realloc(maps__maps_by_name(maps),
2107		− maps__nr_maps(maps) * sizeof(struct map *));
2108		− int i = 0;
2109		−
2110		− if (maps_by_name == NULL)
2111		− return -1;
2112		−
2113		− up_read(maps__lock(maps));
2114		− down_write(maps__lock(maps));
2115		−
2116		− RC_CHK_ACCESS(maps)->maps_by_name = maps_by_name;
2117		− RC_CHK_ACCESS(maps)->nr_maps_allocated = maps__nr_maps(maps);
2118		−
2119		− maps__for_each_entry(maps, rb_node)
2120		− maps_by_name[i++] = map__get(rb_node->map);
2121		−
2122		− __maps__sort_by_name(maps);
2123		−
2124		− up_write(maps__lock(maps));
2125		− down_read(maps__lock(maps));
2126		−
2127		− return 0;
2128		−}
2129		−
2130		−static struct map __maps__find_by_name(struct maps maps, const char *name)
2131		−{
2132		− struct map **mapp;
2133		−
2134		− if (maps__maps_by_name(maps) == NULL &&
2135		− map__groups__sort_by_name_from_rbtree(maps))
2136		− return NULL;
2137		−
2138		− mapp = bsearch(name, maps__maps_by_name(maps), maps__nr_maps(maps),
2139		− sizeof(*mapp), map__strcmp_name);
2140		− if (mapp)
2141		− return *mapp;
2142		− return NULL;
2143		−}
2144		−
2145		−struct map maps__find_by_name(struct maps maps, const char *name)
2146		−{
2147		− struct map_rb_node *rb_node;
2148		− struct map *map;
2149		−
2150		− down_read(maps__lock(maps));
2151		−
2152		−
2153		− if (RC_CHK_ACCESS(maps)->last_search_by_name) {
2154		− const struct dso *dso = map__dso(RC_CHK_ACCESS(maps)->last_search_by_name);
2155		−
2156		− if (strcmp(dso->short_name, name) == 0) {
2157		− map = RC_CHK_ACCESS(maps)->last_search_by_name;
2158		− goto out_unlock;
2159		− }
2160		− }
2161		− /*
2162		− * If we have maps->maps_by_name, then the name isn't in the rbtree,
2163		− * as maps->maps_by_name mirrors the rbtree when lookups by name are
2164		− * made.
2165		− */
2166		− map = __maps__find_by_name(maps, name);
2167		− if (map \|\| maps__maps_by_name(maps) != NULL)
2168		− goto out_unlock;
2169		−
2170		− /* Fallback to traversing the rbtree... */
2171		− maps__for_each_entry(maps, rb_node) {
2172		− struct dso *dso;
2173		−
2174		− map = rb_node->map;
2175		− dso = map__dso(map);
2176		− if (strcmp(dso->short_name, name) == 0) {
2177		− RC_CHK_ACCESS(maps)->last_search_by_name = map;
2178		− goto out_unlock;
2179		− }
2180		− }
2181		− map = NULL;
2182		−
2183		−out_unlock:
2184		− up_read(maps__lock(maps));
2185		− return map;
2186		−}
2187		−
2188	1940	int dso__load_vmlinux(struct dso dso, struct map map,
2189	1941	const char *vmlinux, bool vmlinux_allocated)
2190	1942	{

tools/perf/util/symbol.h+0−1

@@ -189,7 +189,6 @@ void __symbols__insert(struct rb_root_cached symbols, struct symbol sym, · 188 unmodified lines
189	189	void symbols__insert(struct rb_root_cached symbols, struct symbol sym);
190	190	void symbols__fixup_duplicate(struct rb_root_cached *symbols);
191	191	void symbols__fixup_end(struct rb_root_cached *symbols, bool is_kallsyms);
192		−void maps__fixup_end(struct maps *maps);
193	192
194	193	typedef int (mapfn_t)(u64 start, u64 len, u64 pgoff, void data);
195	194	int file__read_maps(int fd, bool exe, mapfn_t mapfn, void *data,

tools/perf/util/symbol_conf.h+5−1

@@ -42,7 +42,11 @@ struct symbol_conf { · 41 unmodified lines
42	42	inline_name,
43	43	disable_add2line_warn,
44	44	buildid_mmap2,
45		− guest_code;
	45	+ guest_code,
	46	+ lazy_load_kernel_maps,
	47	+ keep_exited_threads,
	48	+ annotate_data_member,
	49	+ annotate_data_sample;
46	50	const char *vmlinux_name,
47	51	*kallsyms_name,
48	52	*source_prefix,

tools/perf/util/synthetic-events.c+67−51

@@ -665,18 +665,74 @@ int perf_event__synthesize_cgroups(struct perf_tool *tool __maybe_unused, · 664 unmodified lines
665	665	}
666	666	#endif
667	667
	668	+struct perf_event__synthesize_modules_maps_cb_args {
	669	+ struct perf_tool *tool;
	670	+ perf_event__handler_t process;
	671	+ struct machine *machine;
	672	+ union perf_event *event;
	673	+};
	674	+
	675	+static int perf_event__synthesize_modules_maps_cb(struct map map, void data)
	676	+{
	677	+ struct perf_event__synthesize_modules_maps_cb_args *args = data;
	678	+ union perf_event *event = args->event;
	679	+ struct dso *dso;
	680	+ size_t size;
	681	+
	682	+ if (!__map__is_kmodule(map))
	683	+ return 0;
	684	+
	685	+ dso = map__dso(map);
	686	+ if (symbol_conf.buildid_mmap2) {
	687	+ size = PERF_ALIGN(dso->long_name_len + 1, sizeof(u64));
	688	+ event->mmap2.header.type = PERF_RECORD_MMAP2;
	689	+ event->mmap2.header.size = (sizeof(event->mmap2) -
	690	+ (sizeof(event->mmap2.filename) - size));
	691	+ memset(event->mmap2.filename + size, 0, args->machine->id_hdr_size);
	692	+ event->mmap2.header.size += args->machine->id_hdr_size;
	693	+ event->mmap2.start = map__start(map);
	694	+ event->mmap2.len = map__size(map);
	695	+ event->mmap2.pid = args->machine->pid;
	696	+
	697	+ memcpy(event->mmap2.filename, dso->long_name, dso->long_name_len + 1);
	698	+
	699	+ perf_record_mmap2__read_build_id(&event->mmap2, args->machine, false);
	700	+ } else {
	701	+ size = PERF_ALIGN(dso->long_name_len + 1, sizeof(u64));
	702	+ event->mmap.header.type = PERF_RECORD_MMAP;
	703	+ event->mmap.header.size = (sizeof(event->mmap) -
	704	+ (sizeof(event->mmap.filename) - size));
	705	+ memset(event->mmap.filename + size, 0, args->machine->id_hdr_size);
	706	+ event->mmap.header.size += args->machine->id_hdr_size;
	707	+ event->mmap.start = map__start(map);
	708	+ event->mmap.len = map__size(map);
	709	+ event->mmap.pid = args->machine->pid;
	710	+
	711	+ memcpy(event->mmap.filename, dso->long_name, dso->long_name_len + 1);
	712	+ }
	713	+
	714	+ if (perf_tool__process_synth_event(args->tool, event, args->machine, args->process) != 0)
	715	+ return -1;
	716	+
	717	+ return 0;
	718	+}
	719	+
668	720	int perf_event__synthesize_modules(struct perf_tool *tool, perf_event__handler_t process,
669	721	struct machine *machine)
670	722	{
671		− int rc = 0;
672		− struct map_rb_node *pos;
	723	+ int rc;
673	724	struct maps *maps = machine__kernel_maps(machine);
674		− union perf_event *event;
675		− size_t size = symbol_conf.buildid_mmap2 ?
676		− sizeof(event->mmap2) : sizeof(event->mmap);
	725	+ struct perf_event__synthesize_modules_maps_cb_args args = {
	726	+ .tool = tool,
	727	+ .process = process,
	728	+ .machine = machine,
	729	+ };
	730	+ size_t size = symbol_conf.buildid_mmap2
	731	+ ? sizeof(args.event->mmap2)
	732	+ : sizeof(args.event->mmap);
677	733
678		− event = zalloc(size + machine->id_hdr_size);
679		− if (event == NULL) {
	734	+ args.event = zalloc(size + machine->id_hdr_size);
	735	+ if (args.event == NULL) {
680	736	pr_debug("Not enough memory synthesizing mmap event "
681	737	"for kernel modules\n");
682	738	return -1;
@@ -687,53 +743,13 @@ int perf_event__synthesize_modules(struct perf_tool *tool, perf_event__handler_t · 4 unmodified lines
687	743	* __perf_event_mmap
688	744	*/
689	745	if (machine__is_host(machine))
690		− event->header.misc = PERF_RECORD_MISC_KERNEL;
	746	+ args.event->header.misc = PERF_RECORD_MISC_KERNEL;
691	747	else
692		− event->header.misc = PERF_RECORD_MISC_GUEST_KERNEL;
693		−
694		− maps__for_each_entry(maps, pos) {
695		− struct map *map = pos->map;
696		− struct dso *dso;
	748	+ args.event->header.misc = PERF_RECORD_MISC_GUEST_KERNEL;
697	749
698		− if (!__map__is_kmodule(map))
699		− continue;
	750	+ rc = maps__for_each_map(maps, perf_event__synthesize_modules_maps_cb, &args);
700	751
701		− dso = map__dso(map);
702		− if (symbol_conf.buildid_mmap2) {
703		− size = PERF_ALIGN(dso->long_name_len + 1, sizeof(u64));
704		− event->mmap2.header.type = PERF_RECORD_MMAP2;
705		− event->mmap2.header.size = (sizeof(event->mmap2) -
706		− (sizeof(event->mmap2.filename) - size));
707		− memset(event->mmap2.filename + size, 0, machine->id_hdr_size);
708		− event->mmap2.header.size += machine->id_hdr_size;
709		− event->mmap2.start = map__start(map);
710		− event->mmap2.len = map__size(map);
711		− event->mmap2.pid = machine->pid;
712		−
713		− memcpy(event->mmap2.filename, dso->long_name, dso->long_name_len + 1);
714		−
715		− perf_record_mmap2__read_build_id(&event->mmap2, machine, false);
716		− } else {
717		− size = PERF_ALIGN(dso->long_name_len + 1, sizeof(u64));
718		− event->mmap.header.type = PERF_RECORD_MMAP;
719		− event->mmap.header.size = (sizeof(event->mmap) -
720		− (sizeof(event->mmap.filename) - size));
721		− memset(event->mmap.filename + size, 0, machine->id_hdr_size);
722		− event->mmap.header.size += machine->id_hdr_size;
723		− event->mmap.start = map__start(map);
724		− event->mmap.len = map__size(map);
725		− event->mmap.pid = machine->pid;
726		−
727		− memcpy(event->mmap.filename, dso->long_name, dso->long_name_len + 1);
728		− }
729		−
730		− if (perf_tool__process_synth_event(tool, event, machine, process) != 0) {
731		− rc = -1;
732		− break;
733		− }
734		− }
735		−
736		− free(event);
	752	+ free(args.event);
737	753	return rc;
738	754	}
739	755

tools/perf/util/thread.c+20−22

@@ -345,38 +345,36 @@ int thread__insert_map(struct thread thread, struct map map) · 344 unmodified lines
345	345	if (ret)
346	346	return ret;
347	347
348		− maps__fixup_overlappings(thread__maps(thread), map, stderr);
349		− return maps__insert(thread__maps(thread), map);
	348	+ return maps__fixup_overlap_and_insert(thread__maps(thread), map);
350	349	}
351	350
352		−static int __thread__prepare_access(struct thread *thread)
	351	+struct thread__prepare_access_maps_cb_args {
	352	+ int err;
	353	+ struct maps *maps;
	354	+};
	355	+
	356	+static int thread__prepare_access_maps_cb(struct map map, void data)
353	357	{
354	358	bool initialized = false;
355		− int err = 0;
356		− struct maps *maps = thread__maps(thread);
357		− struct map_rb_node *rb_node;
358		−
359		− down_read(maps__lock(maps));
360		−
361		− maps__for_each_entry(maps, rb_node) {
362		− err = unwind__prepare_access(thread__maps(thread), rb_node->map, &initialized);
363		− if (err \|\| initialized)
364		− break;
365		− }
	359	+ struct thread__prepare_access_maps_cb_args *args = data;
366	360
367		− up_read(maps__lock(maps));
	361	+ args->err = unwind__prepare_access(args->maps, map, &initialized);
368	362
369		− return err;
	363	+ return (args->err \|\| initialized) ? 1 : 0;
370	364	}
371	365
372	366	static int thread__prepare_access(struct thread *thread)
373	367	{
374		− int err = 0;
	368	+ struct thread__prepare_access_maps_cb_args args = {
	369	+ .err = 0,
	370	+ };
375	371
376		− if (dwarf_callchain_users)
377		− err = __thread__prepare_access(thread);
	372	+ if (dwarf_callchain_users) {
	373	+ args.maps = thread__maps(thread);
	374	+ maps__for_each_map(thread__maps(thread), thread__prepare_access_maps_cb, &args);
	375	+ }
378	376
379		− return err;
	377	+ return args.err;
380	378	}
381	379
382	380	static int thread__clone_maps(struct thread thread, struct thread parent, bool do_maps_clone)
@@ -385,14 +383,14 @@ static int thread__clone_maps(struct thread thread, struct thread parent, bool · 2 unmodified lines
385	383	if (thread__pid(thread) == thread__pid(parent))
386	384	return thread__prepare_access(thread);
387	385
388		− if (thread__maps(thread) == thread__maps(parent)) {
	386	+ if (RC_CHK_EQUAL(thread__maps(thread), thread__maps(parent))) {
389	387	pr_debug("broken map groups on thread %d/%d parent %d/%d\n",
390	388	thread__pid(thread), thread__tid(thread),
391	389	thread__pid(parent), thread__tid(parent));
392	390	return 0;
393	391	}
394	392	/* But this one is new process, copy maps. */
395		− return do_maps_clone ? maps__clone(thread, thread__maps(parent)) : 0;
	393	+ return do_maps_clone ? maps__copy_from(thread__maps(thread), thread__maps(parent)) : 0;
396	394	}
397	395
398	396	int thread__fork(struct thread thread, struct thread parent, u64 timestamp, bool do_maps_clone)

tools/perf/util/thread.h+14−0

@@ -36,13 +36,22 @@ struct thread_rb_node { · 35 unmodified lines
36	36	};
37	37
38	38	DECLARE_RC_STRUCT(thread) {
	39	+ /** @maps: mmaps associated with this thread. */
39	40	struct maps *maps;
40	41	pid_t pid_; /* Not all tools update this */
	42	+ /** @tid: thread ID number unique to a machine. */
41	43	pid_t tid;
	44	+ /** @ppid: parent process of the process this thread belongs to. */
42	45	pid_t ppid;
43	46	int cpu;
44	47	int guest_cpu; /* For QEMU thread */
45	48	refcount_t refcnt;
	49	+ /**
	50	+ * @exited: Has the thread had an exit event. Such threads are usually
	51	+ * removed from the machine's threads but some events/tools require
	52	+ * access to dead threads.
	53	+ */
	54	+ bool exited;
46	55	bool comm_set;
47	56	int comm_len;
48	57	struct list_head namespaces_list;
@@ -189,6 +198,11 @@ static inline refcount_t thread__refcnt(struct thread thread) · 140 unmodified lines
189	198	return &RC_CHK_ACCESS(thread)->refcnt;
190	199	}
191	200
	201	+static inline void thread__set_exited(struct thread *thread, bool exited)
	202	+{
	203	+ RC_CHK_ACCESS(thread)->exited = exited;
	204	+}
	205	+
192	206	static inline bool thread__comm_set(const struct thread *thread)
193	207	{
194	208	return RC_CHK_ACCESS(thread)->comm_set;

tools/perf/util/top.c+4−5

@@ -28,6 +28,7 @@ size_t perf_top__header_snprintf(struct perf_top top, char bf, size_t size) · 27 unmodified lines
28	28	struct record_opts *opts = &top->record_opts;
29	29	struct target *target = &opts->target;
30	30	size_t ret = 0;
	31	+ int nr_cpus;
31	32
32	33	if (top->samples) {
33	34	samples_per_sec = top->samples / top->delay_secs;
@@ -93,19 +94,17 @@ size_t perf_top__header_snprintf(struct perf_top top, char bf, size_t size) · 59 unmodified lines
93	94	else
94	95	ret += SNPRINTF(bf + ret, size - ret, " (all");
95	96
	97	+ nr_cpus = perf_cpu_map__nr(top->evlist->core.user_requested_cpus);
96	98	if (target->cpu_list)
97	99	ret += SNPRINTF(bf + ret, size - ret, ", CPU%s: %s)",
98		− perf_cpu_map__nr(top->evlist->core.user_requested_cpus) > 1
99		− ? "s" : "",
	100	+ nr_cpus > 1 ? "s" : "",
100	101	target->cpu_list);
101	102	else {
102	103	if (target->tid)
103	104	ret += SNPRINTF(bf + ret, size - ret, ")");
104	105	else
105	106	ret += SNPRINTF(bf + ret, size - ret, ", %d CPU%s)",
106		− perf_cpu_map__nr(top->evlist->core.user_requested_cpus),
107		− perf_cpu_map__nr(top->evlist->core.user_requested_cpus) > 1
108		− ? "s" : "");
	107	+ nr_cpus, nr_cpus > 1 ? "s" : "");
109	108	}
110	109
111	110	perf_top__reset_sample_counters(top);

tools/perf/util/top.h+0−1

@@ -21,7 +21,6 @@ struct perf_top { · 20 unmodified lines
21	21	struct perf_tool tool;
22	22	struct evlist evlist, sb_evlist;
23	23	struct record_opts record_opts;
24		− struct annotation_options annotation_opts;
25	24	struct evswitch evswitch;
26	25	/*
27	26	* Symbols will be added here in perf_event__process_sample and will

tools/perf/util/unwind-libdw.c+17−4

@@ -46,6 +46,7 @@ static int __report_module(struct addr_location *al, u64 ip, · 45 unmodified lines
46	46	{
47	47	Dwfl_Module *mod;
48	48	struct dso *dso = NULL;
	49	+ Dwarf_Addr base;
49	50	/*
50	51	* Some callers will use al->sym, so we can't just use the
51	52	* cheaper thread__find_map() here.
@@ -58,13 +59,25 @@ static int __report_module(struct addr_location *al, u64 ip, · 6 unmodified lines
58	59	if (!dso)
59	60	return 0;
60	61
	62	+ /*
	63	+ * The generated JIT DSO files only map the code segment without
	64	+ * ELF headers. Since JIT codes used to be packed in a memory
	65	+ * segment, calculating the base address using pgoff falls into
	66	+ * a different code in another DSO. So just use the map->start
	67	+ * directly to pick the correct one.
	68	+ */
	69	+ if (!strncmp(dso->long_name, "/tmp/jitted-", 12))
	70	+ base = map__start(al->map);
	71	+ else
	72	+ base = map__start(al->map) - map__pgoff(al->map);
	73	+
61	74	mod = dwfl_addrmodule(ui->dwfl, ip);
62	75	if (mod) {
63	76	Dwarf_Addr s;
64	77
65	78	dwfl_module_info(mod, NULL, &s, NULL, NULL, NULL, NULL, NULL);
66		− if (s != map__start(al->map) - map__pgoff(al->map))
67		− mod = 0;
	79	+ if (s != base)
	80	+ mod = NULL;
68	81	}
69	82
70	83	if (!mod) {
@@ -72,14 +85,14 @@ static int __report_module(struct addr_location *al, u64 ip, · 1 unmodified line
72	85
73	86	__symbol__join_symfs(filename, sizeof(filename), dso->long_name);
74	87	mod = dwfl_report_elf(ui->dwfl, dso->short_name, filename, -1,
75		− map__start(al->map) - map__pgoff(al->map), false);
	88	+ base, false);
76	89	}
77	90	if (!mod) {
78	91	char filename[PATH_MAX];
79	92
80	93	if (dso__build_id_filename(dso, filename, sizeof(filename), false))
81	94	mod = dwfl_report_elf(ui->dwfl, dso->short_name, filename, -1,
82		− map__start(al->map) - map__pgoff(al->map), false);
	95	+ base, false);
83	96	}
84	97
85	98	if (mod) {

tools/perf/util/unwind-libunwind-local.c+24−10

@@ -302,12 +302,31 @@ static int unwind_spec_ehframe(struct dso dso, struct machine machine, · 301 unmodified lines
302	302	return 0;
303	303	}
304	304
	305	+struct read_unwind_spec_eh_frame_maps_cb_args {
	306	+ struct dso *dso;
	307	+ u64 base_addr;
	308	+};
	309	+
	310	+static int read_unwind_spec_eh_frame_maps_cb(struct map map, void data)
	311	+{
	312	+
	313	+ struct read_unwind_spec_eh_frame_maps_cb_args *args = data;
	314	+
	315	+ if (map__dso(map) == args->dso && map__start(map) - map__pgoff(map) < args->base_addr)
	316	+ args->base_addr = map__start(map) - map__pgoff(map);
	317	+
	318	+ return 0;
	319	+}
	320	+
	321	+
305	322	static int read_unwind_spec_eh_frame(struct dso dso, struct unwind_info ui,
306	323	u64 table_data, u64 segbase,
307	324	u64 *fde_count)
308	325	{
309		− struct map_rb_node *map_node;
310		− u64 base_addr = UINT64_MAX;
	326	+ struct read_unwind_spec_eh_frame_maps_cb_args args = {
	327	+ .dso = dso,
	328	+ .base_addr = UINT64_MAX,
	329	+ };
311	330	int ret, fd;
312	331
313	332	if (dso->data.eh_frame_hdr_offset == 0) {
@@ -325,16 +344,11 @@ static int read_unwind_spec_eh_frame(struct dso dso, struct unwind_info ui, · 11 unmodified lines
325	344	return -EINVAL;
326	345	}
327	346
328		− maps__for_each_entry(thread__maps(ui->thread), map_node) {
329		− struct map *map = map_node->map;
330		− u64 start = map__start(map);
	347	+ maps__for_each_map(thread__maps(ui->thread), read_unwind_spec_eh_frame_maps_cb, &args);
331	348
332		− if (map__dso(map) == dso && start < base_addr)
333		− base_addr = start;
334		− }
335		− base_addr -= dso->data.elf_base_addr;
	349	+ args.base_addr -= dso->data.elf_base_addr;
336	350	/* Address of .eh_frame_hdr */
337		− *segbase = base_addr + dso->data.eh_frame_hdr_addr;
	351	+ *segbase = args.base_addr + dso->data.eh_frame_hdr_addr;
338	352	ret = unwind_spec_ehframe(dso, ui->machine, dso->data.eh_frame_hdr_offset,
339	353	table_data, fde_count);
340	354	if (ret)

tools/perf/util/vdso.c+23−12

@@ -140,23 +140,34 @@ static struct dso __machine__addnew_vdso(struct machine machine, const char *s · 139 unmodified lines
140	140	return dso;
141	141	}
142	142
	143	+struct machine__thread_dso_type_maps_cb_args {
	144	+ struct machine *machine;
	145	+ enum dso_type dso_type;
	146	+};
	147	+
	148	+static int machine__thread_dso_type_maps_cb(struct map map, void data)
	149	+{
	150	+ struct machine__thread_dso_type_maps_cb_args *args = data;
	151	+ struct dso *dso = map__dso(map);
	152	+
	153	+ if (!dso \|\| dso->long_name[0] != '/')
	154	+ return 0;
	155	+
	156	+ args->dso_type = dso__type(dso, args->machine);
	157	+ return (args->dso_type != DSO__TYPE_UNKNOWN) ? 1 : 0;
	158	+}
	159	+
143	160	static enum dso_type machine__thread_dso_type(struct machine *machine,
144	161	struct thread *thread)
145	162	{
146		− enum dso_type dso_type = DSO__TYPE_UNKNOWN;
147		− struct map_rb_node *rb_node;
148		−
149		− maps__for_each_entry(thread__maps(thread), rb_node) {
150		− struct dso *dso = map__dso(rb_node->map);
	163	+ struct machine__thread_dso_type_maps_cb_args args = {
	164	+ .machine = machine,
	165	+ .dso_type = DSO__TYPE_UNKNOWN,
	166	+ };
151	167
152		− if (!dso \|\| dso->long_name[0] != '/')
153		− continue;
154		− dso_type = dso__type(dso, machine);
155		− if (dso_type != DSO__TYPE_UNKNOWN)
156		− break;
157		− }
	168	+ maps__for_each_map(thread__maps(thread), machine__thread_dso_type_maps_cb, &args);
158	169
159		− return dso_type;
	170	+ return args.dso_type;
160	171	}
161	172
162	173	#if BITS_PER_LONG == 64

tools/perf/util/zstd.c+33−30

@@ -7,35 +7,9 @@ · 6 unmodified lines
7	7
8	8	int zstd_init(struct zstd_data *data, int level)
9	9	{
10		− size_t ret;
11		−
12		− data->dstream = ZSTD_createDStream();
13		− if (data->dstream == NULL) {
14		− pr_err("Couldn't create decompression stream.\n");
15		− return -1;
16		− }
17		−
18		− ret = ZSTD_initDStream(data->dstream);
19		− if (ZSTD_isError(ret)) {
20		− pr_err("Failed to initialize decompression stream: %s\n", ZSTD_getErrorName(ret));
21		− return -1;
22		− }
23		−
24		− if (!level)
25		− return 0;
26		−
27		− data->cstream = ZSTD_createCStream();
28		− if (data->cstream == NULL) {
29		− pr_err("Couldn't create compression stream.\n");
30		− return -1;
31		− }
32		−
33		− ret = ZSTD_initCStream(data->cstream, level);
34		− if (ZSTD_isError(ret)) {
35		− pr_err("Failed to initialize compression stream: %s\n", ZSTD_getErrorName(ret));
36		− return -1;
37		− }
38		−
	10	+ data->comp_level = level;
	11	+ data->dstream = NULL;
	12	+ data->cstream = NULL;
39	13	return 0;
40	14	}
41	15
@@ -54,7 +28,7 @@ int zstd_fini(struct zstd_data *data) · 12 unmodified lines
54	28	return 0;
55	29	}
56	30
57		−size_t zstd_compress_stream_to_records(struct zstd_data data, void dst, size_t dst_size,
	31	+ssize_t zstd_compress_stream_to_records(struct zstd_data data, void dst, size_t dst_size,
58	32	void *src, size_t src_size, size_t max_record_size,
59	33	size_t process_header(void *record, size_t increment))
60	34	{
@@ -63,6 +37,21 @@ size_t zstd_compress_stream_to_records(struct zstd_data data, void dst, size_t · 2 unmodified lines
63	37	ZSTD_outBuffer output;
64	38	void *record;
65	39
	40	+ if (!data->cstream) {
	41	+ data->cstream = ZSTD_createCStream();
	42	+ if (data->cstream == NULL) {
	43	+ pr_err("Couldn't create compression stream.\n");
	44	+ return -1;
	45	+ }
	46	+
	47	+ ret = ZSTD_initCStream(data->cstream, data->comp_level);
	48	+ if (ZSTD_isError(ret)) {
	49	+ pr_err("Failed to initialize compression stream: %s\n",
	50	+ ZSTD_getErrorName(ret));
	51	+ return -1;
	52	+ }
	53	+ }
	54	+
66	55	while (input.pos < input.size) {
67	56	record = dst;
68	57	size = process_header(record, 0);
@@ -96,6 +85,20 @@ size_t zstd_decompress_stream(struct zstd_data data, void src, size_t src_size · 27 unmodified lines
96	85	ZSTD_inBuffer input = { src, src_size, 0 };
97	86	ZSTD_outBuffer output = { dst, dst_size, 0 };
98	87
	88	+ if (!data->dstream) {
	89	+ data->dstream = ZSTD_createDStream();
	90	+ if (data->dstream == NULL) {
	91	+ pr_err("Couldn't create decompression stream.\n");
	92	+ return 0;
	93	+ }
	94	+
	95	+ ret = ZSTD_initDStream(data->dstream);
	96	+ if (ZSTD_isError(ret)) {
	97	+ pr_err("Failed to initialize decompression stream: %s\n",
	98	+ ZSTD_getErrorName(ret));
	99	+ return 0;
	100	+ }
	101	+ }
99	102	while (input.pos < input.size) {
100	103	ret = ZSTD_decompressStream(data->dstream, &output, &input);
101	104	if (ZSTD_isError(ret)) {