History

kbuild/Makefile: Prepare for using macros in inline assembly code to work around asm() related GCC inlining bugs

Using macros in inline assembly allows us to work around bugs in GCC's inlining decisions. Compile macros.S and use it to assemble all C files. Currently only x86 will use it. Background: The inlining pass of GCC doesn't include an assembler, so it's not aware of basic properties of the generated code, such as its size in bytes, or that there are such things as discontiuous blocks of code and data due to the newfangled linker feature called 'sections' ... Instead GCC uses a lazy and fragile heuristic: it does a linear count of certain syntactic and whitespace elements in inlined assembly block source code, such as a count of new-lines and semicolons (!), as a poor substitute for "code size and complexity". Unsurprisingly this heuristic falls over and breaks its neck whith certain common types of kernel code that use inline assembly, such as the frequent practice of putting useful information into alternative sections. As a result of this fresh, 20+ years old GCC bug, GCC's inlining decisions are effectively disabled for inlined functions that make use of such asm() blocks, because GCC thinks those sections of code are "large" - when in reality they are often result in just a very low number of machine instructions. This absolute lack of inlining provess when GCC comes across such asm() blocks both increases generated kernel code size and causes performance overhead, which is particularly noticeable on paravirt kernels, which make frequent use of these inlining facilities in attempt to stay out of the way when running on baremetal hardware. Instead of fixing the compiler we use a workaround: we set an assembly macro and call it from the inlined assembly block. As a result GCC considers the inline assembly block as a single instruction. (Which it often isn't but I digress.) This uglifies and bloats the source code - for example just the refcount related changes have this impact: Makefile | 9 +++++++-- arch/x86/Makefile | 7 +++++++ arch/x86/kernel/macros.S | 7 +++++++ scripts/Kbuild.include | 4 +++- scripts/mod/Makefile | 2 ++ 5 files changed, 26 insertions(+), 3 deletions(-) Yay readability and maintainability, it's not like assembly code is hard to read and maintain ... We also hope that GCC will eventually get fixed, but we are not holding our breath for that. Yet we are optimistic, it might still happen, any decade now. [ mingo: Wrote new changelog describing the background. ]

Nadav Amit committed 8 years ago

Commit 77b0bf51 parent 35e76b95 files · +26 −3Browse files

Unified · Split

5 files
  1. Makefile +7 −2
  2. arch/x86/Makefile +7 −0
  3. arch/x86/kernel/macros.S +7 −0
  4. scripts/Kbuild.include +3 −1
  5. scripts/mod/Makefile +2 −0
Makefile+7−2
@@ -1071,7 +1071,7 @@ scripts: scripts_basic asm-generic gcc-plugins $(autoksyms_h) · 1070 unmodified lines
10711071 # version.h and scripts_basic is processed / created.
10721072
10731073 # Listed in dependency order
1074−PHONY += prepare archprepare prepare0 prepare1 prepare2 prepare3
1074+PHONY += prepare archprepare macroprepare prepare0 prepare1 prepare2 prepare3
10751075
10761076 # prepare3 is used to check if we are building in a separate output directory,
10771077 # and if so do:
@@ -1094,7 +1094,9 @@ prepare2: prepare3 outputmakefile asm-generic · 16 unmodified lines
10941094 prepare1: prepare2 $(version_h) $(autoksyms_h) include/generated/utsrelease.h
10951095 $(cmd_crmodverdir)
10961096
1097−archprepare: archheaders archscripts prepare1 scripts_basic
1097+macroprepare: prepare1 archmacros
1098+
1099+archprepare: archheaders archscripts macroprepare scripts_basic
10981100
10991101 prepare0: archprepare gcc-plugins
11001102 $(Q)$(MAKE) $(build)=.
@@ -1162,6 +1164,9 @@ archheaders: · 61 unmodified lines
11621164 PHONY += archscripts
11631165 archscripts:
11641166
1167+PHONY += archmacros
1168+archmacros:
1169+
11651170 PHONY += __headers
11661171 __headers: $(version_h) scripts_basic uapi-asm-generic archheaders archscripts
11671172 $(Q)$(MAKE) $(build)=scripts build_unifdef
arch/x86/Makefile+7−0
@@ -236,6 +236,13 @@ archscripts: scripts_basic · 235 unmodified lines
236236 archheaders:
237237 $(Q)$(MAKE) $(build)=arch/x86/entry/syscalls all
238238
239+archmacros:
240+ $(Q)$(MAKE) $(build)=arch/x86/kernel arch/x86/kernel/macros.s
241+
242+ASM_MACRO_FLAGS = -Wa,arch/x86/kernel/macros.s -Wa,-
243+export ASM_MACRO_FLAGS
244+KBUILD_CFLAGS += $(ASM_MACRO_FLAGS)
245+
239246 ###
240247 # Kernel objects
241248
arch/x86/kernel/macros.S+7−0
@@ -0,0 +1,7 @@
1+/* SPDX-License-Identifier: GPL-2.0 */
2+
3+/*
4+ * This file includes headers whose assembly part includes macros which are
5+ * commonly used. The macros are precompiled into assmebly file which is later
6+ * assembled together with each compiled file.
7+ */
scripts/Kbuild.include+3−1
@@ -115,7 +115,9 @@ __cc-option = $(call try-run,\ · 114 unmodified lines
115115
116116 # Do not attempt to build with gcc plugins during cc-option tests.
117117 # (And this uses delayed resolution so the flags will be up to date.)
118−CC_OPTION_CFLAGS = $(filter-out $(GCC_PLUGINS_CFLAGS),$(KBUILD_CFLAGS))
118+# In addition, do not include the asm macros which are built later.
119+CC_OPTION_FILTERED = $(GCC_PLUGINS_CFLAGS) $(ASM_MACRO_FLAGS)
120+CC_OPTION_CFLAGS = $(filter-out $(CC_OPTION_FILTERED),$(KBUILD_CFLAGS))
119121
120122 # cc-option
121123 # Usage: cflags-y += $(call cc-option,-march=winchip-c6,-march=i586)
scripts/mod/Makefile+2−0
@@ -4,6 +4,8 @@ OBJECT_FILES_NON_STANDARD := y · 3 unmodified lines
44 hostprogs-y := modpost mk_elfconfig
55 always := $(hostprogs-y) empty.o
66
7+CFLAGS_REMOVE_empty.o := $(ASM_MACRO_FLAGS)
8+
79 modpost-objs := modpost.o file2alias.o sumversion.o
810
911 devicetable-offsets-file := devicetable-offsets.h