The GCC '-maccumulate-outgoing-args' flag is enabled for most configs, mostly because of issues which are no longer relevant. For most configs, and with most recent versions of GCC, it's no longer needed. Clarify which cases need it, and only enable it for those cases. Also produce a compile-time error for the ftrace graph + mcount + '-Os' case, which will otherwise cause runtime failures. The main benefit of '-maccumulate-outgoing-args' is that it prevents an ugly prologue for functions which have aligned stacks. But removing the option also has some benefits: more readable argument saves, smaller text size, and (presumably) slightly improved performance. Here are the object size savings for 32-bit and 64-bit defconfig kernels: text data bss dec hex filename 10006710 3543328 1773568 15323606 e9d1d6 vmlinux.x86-32.before 9706358 3547424 1773568 15027350 e54c96 vmlinux.x86-32.after text data bss dec hex filename 10652105 4537576 843776 16033457 f4a6b1 vmlinux.x86-64.before 10639629 4537576 843776 16020981 f475f5 vmlinux.x86-64.after That comes out to a 3% text size improvement on x86-32 and a 0.1% text size improvement on x86-64.
Unified · Split
| @@ -120,10 +120,6 @@ else · 119 unmodified lines | |||
| 120 | # -funit-at-a-time shrinks the kernel .text considerably | 120 | # -funit-at-a-time shrinks the kernel .text considerably |
| 121 | # unfortunately it makes reading oopses harder. | 121 | # unfortunately it makes reading oopses harder. |
| 122 | KBUILD_CFLAGS += $(call cc-option,-funit-at-a-time) | 122 | KBUILD_CFLAGS += $(call cc-option,-funit-at-a-time) |
| 123 | |||
| 124 | # this works around some issues with generating unwind tables in older gccs | ||
| 125 | # newer gccs do it by default | ||
| 126 | KBUILD_CFLAGS += $(call cc-option,-maccumulate-outgoing-args) | ||
| 127 | endif | 123 | endif |
| 128 | 124 | ||
| 129 | ifdef CONFIG_X86_X32 | 125 | ifdef CONFIG_X86_X32 |
| @@ -147,6 +143,37 @@ ifeq ($(CONFIG_KMEMCHECK),y) · 17 unmodified lines | |||
| 147 | KBUILD_CFLAGS += $(call cc-option,-fno-builtin-memcpy) | 143 | KBUILD_CFLAGS += $(call cc-option,-fno-builtin-memcpy) |
| 148 | endif | 144 | endif |
| 149 | 145 | ||
| 146 | # | ||
| 147 | # If the function graph tracer is used with mcount instead of fentry, | ||
| 148 | # '-maccumulate-outgoing-args' is needed to prevent a GCC bug | ||
| 149 | # (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=42109) | ||
| 150 | # | ||
| 151 | ifdef CONFIG_FUNCTION_GRAPH_TRACER | ||
| 152 | ifndef CONFIG_HAVE_FENTRY | ||
| 153 | ACCUMULATE_OUTGOING_ARGS := 1 | ||
| 154 | else | ||
| 155 | ifeq ($(call cc-option-yn, -mfentry), n) | ||
| 156 | ACCUMULATE_OUTGOING_ARGS := 1 | ||
| 157 | endif | ||
| 158 | endif | ||
| 159 | endif | ||
| 160 | |||
| 161 | # | ||
| 162 | # Jump labels need '-maccumulate-outgoing-args' for gcc < 4.5.2 to prevent a | ||
| 163 | # GCC bug (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46226). There's no way | ||
| 164 | # to test for this bug at compile-time because the test case needs to execute, | ||
| 165 | # which is a no-go for cross compilers. So check the GCC version instead. | ||
| 166 | # | ||
| 167 | ifdef CONFIG_JUMP_LABEL | ||
| 168 | ifneq ($(ACCUMULATE_OUTGOING_ARGS), 1) | ||
| 169 | ACCUMULATE_OUTGOING_ARGS = $(call cc-if-fullversion, -lt, 040502, 1) | ||
| 170 | endif | ||
| 171 | endif | ||
| 172 | |||
| 173 | ifeq ($(ACCUMULATE_OUTGOING_ARGS), 1) | ||
| 174 | KBUILD_CFLAGS += -maccumulate-outgoing-args | ||
| 175 | endif | ||
| 176 | |||
| 150 | # Stackpointer is addressed different for 32 bit and 64 bit x86 | 177 | # Stackpointer is addressed different for 32 bit and 64 bit x86 |
| 151 | sp-$(CONFIG_X86_32) := esp | 178 | sp-$(CONFIG_X86_32) := esp |
| 152 | sp-$(CONFIG_X86_64) := rsp | 179 | sp-$(CONFIG_X86_64) := rsp |
| @@ -45,24 +45,6 @@ cflags-$(CONFIG_MGEODE_LX) += $(call cc-option,-march=geode,-march=pentium-mmx) · 44 unmodified lines | |||
| 45 | # cpu entries | 45 | # cpu entries |
| 46 | cflags-$(CONFIG_X86_GENERIC) += $(call tune,generic,$(call tune,i686)) | 46 | cflags-$(CONFIG_X86_GENERIC) += $(call tune,generic,$(call tune,i686)) |
| 47 | 47 | ||
| 48 | # Work around the pentium-mmx code generator madness of gcc4.4.x which | ||
| 49 | # does stack alignment by generating horrible code _before_ the mcount | ||
| 50 | # prologue (push %ebp, mov %esp, %ebp) which breaks the function graph | ||
| 51 | # tracer assumptions. For i686, generic, core2 this is set by the | ||
| 52 | # compiler anyway | ||
| 53 | ifeq ($(CONFIG_FUNCTION_GRAPH_TRACER), y) | ||
| 54 | ADD_ACCUMULATE_OUTGOING_ARGS := y | ||
| 55 | endif | ||
| 56 | |||
| 57 | # Work around to a bug with asm goto with first implementations of it | ||
| 58 | # in gcc causing gcc to mess up the push and pop of the stack in some | ||
| 59 | # uses of asm goto. | ||
| 60 | ifeq ($(CONFIG_JUMP_LABEL), y) | ||
| 61 | ADD_ACCUMULATE_OUTGOING_ARGS := y | ||
| 62 | endif | ||
| 63 | |||
| 64 | cflags-$(ADD_ACCUMULATE_OUTGOING_ARGS) += $(call cc-option,-maccumulate-outgoing-args) | ||
| 65 | |||
| 66 | # Bug fix for binutils: this option is required in order to keep | 48 | # Bug fix for binutils: this option is required in order to keep |
| 67 | # binutils from generating NOPL instructions against our will. | 49 | # binutils from generating NOPL instructions against our will. |
| 68 | ifneq ($(CONFIG_X86_P6_NOP),y) | 50 | ifneq ($(CONFIG_X86_P6_NOP),y) |
| @@ -29,6 +29,12 @@ · 28 unmodified lines | |||
| 29 | #include <asm/ftrace.h> | 29 | #include <asm/ftrace.h> |
| 30 | #include <asm/nops.h> | 30 | #include <asm/nops.h> |
| 31 | 31 | ||
| 32 | #if defined(CONFIG_FUNCTION_GRAPH_TRACER) && \ | ||
| 33 | !defined(CC_USING_FENTRY) && \ | ||
| 34 | !defined(CONFIG_CC_OPTIMIZE_FOR_PERFORMANCE) | ||
| 35 | # error The following combination is not supported: ((compiler missing -mfentry) || (CONFIG_X86_32 and !CONFIG_DYNAMIC_FTRACE)) && CONFIG_FUNCTION_GRAPH_TRACER && CONFIG_CC_OPTIMIZE_FOR_SIZE | ||
| 36 | #endif | ||
| 37 | |||
| 32 | #ifdef CONFIG_DYNAMIC_FTRACE | 38 | #ifdef CONFIG_DYNAMIC_FTRACE |
| 33 | 39 | ||
| 34 | int ftrace_arch_code_modify_prepare(void) | 40 | int ftrace_arch_code_modify_prepare(void) |
| @@ -148,6 +148,10 @@ cc-fullversion = $(shell $(CONFIG_SHELL) \ · 147 unmodified lines | |||
| 148 | # Usage: EXTRA_CFLAGS += $(call cc-ifversion, -lt, 0402, -O1) | 148 | # Usage: EXTRA_CFLAGS += $(call cc-ifversion, -lt, 0402, -O1) |
| 149 | cc-ifversion = $(shell [ $(cc-version) $(1) $(2) ] && echo $(3) || echo $(4)) | 149 | cc-ifversion = $(shell [ $(cc-version) $(1) $(2) ] && echo $(3) || echo $(4)) |
| 150 | 150 | ||
| 151 | # cc-if-fullversion | ||
| 152 | # Usage: EXTRA_CFLAGS += $(call cc-if-fullversion, -lt, 040502, -O1) | ||
| 153 | cc-if-fullversion = $(shell [ $(cc-fullversion) $(1) $(2) ] && echo $(3) || echo $(4)) | ||
| 154 | |||
| 151 | # cc-ldoption | 155 | # cc-ldoption |
| 152 | # Usage: ldflags += $(call cc-ldoption, -Wl$(comma)--hash-style=both) | 156 | # Usage: ldflags += $(call cc-ldoption, -Wl$(comma)--hash-style=both) |
| 153 | cc-ldoption = $(call try-run,\ | 157 | cc-ldoption = $(call try-run,\ |