| Message ID | 20180212082635.GA25314@hirez.programming.kicks-ass.net (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, 12 Feb 2018, Peter Zijlstra wrote: > On Sun, Feb 11, 2018 at 10:13:44AM -0800, Linus Torvalds wrote: > > > That actually sounds like we could just > > > > (a) make gcc 4.5 be the minimum required version > > > > (b) actually error out if we find a bad compiler > > So the unofficial plan was to enforce asm-goto and -fentry support by > hard failure to build, which would get us at gcc-4.6 and then remove all Has gcc-4.6 a (planned) retpoline backport? IIRC the cutoff for that was gcc 4.9 Thanks, tglx -- To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, Feb 12, 2018 at 11:27:25AM +0100, Thomas Gleixner wrote: > On Mon, 12 Feb 2018, Peter Zijlstra wrote: > > > On Sun, Feb 11, 2018 at 10:13:44AM -0800, Linus Torvalds wrote: > > > > > That actually sounds like we could just > > > > > > (a) make gcc 4.5 be the minimum required version > > > > > > (b) actually error out if we find a bad compiler > > > > So the unofficial plan was to enforce asm-goto and -fentry support by > > hard failure to build, which would get us at gcc-4.6 and then remove all > > Has gcc-4.6 a (planned) retpoline backport? IIRC the cutoff for that was > gcc 4.9 Official GCC will not do retpoline before 4.9 AFAIK. But if someone were to want to build a RETPOLINE=n kernel I don't see why we should mandate retpoline. Also, distro's have backported retpoline much further back already. -- To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, 2018-02-12 at 09:26 +0100, Peter Zijlstra wrote: > On Sun, Feb 11, 2018 at 10:13:44AM -0800, Linus Torvalds wrote: > > > That actually sounds like we could just > > > > (a) make gcc 4.5 be the minimum required version > > > > (b) actually error out if we find a bad compiler > > So the unofficial plan was to enforce asm-goto and -fentry support by > hard failure to build, which would get us at gcc-4.6 and then remove all > the fallback cruft needed for those features -- for x86. If we want to > do this tree wide, that's obviously OK with me too ;-) This would also kill clang support, right?
On Mon, Feb 12, 2018 at 8:19 AM, David Woodhouse <dwmw2@infradead.org> wrote: > On Mon, 2018-02-12 at 09:26 +0100, Peter Zijlstra wrote: >> On Sun, Feb 11, 2018 at 10:13:44AM -0800, Linus Torvalds wrote: >> >> > That actually sounds like we could just >> > >> > (a) make gcc 4.5 be the minimum required version >> > >> > (b) actually error out if we find a bad compiler >> >> So the unofficial plan was to enforce asm-goto and -fentry support by >> hard failure to build, which would get us at gcc-4.6 and then remove all >> the fallback cruft needed for those features -- for x86. If we want to >> do this tree wide, that's obviously OK with me too ;-) > > This would also kill clang support, right? That would be bad: Android exclusively builds with clang. -Kees
On Mon, Feb 12, 2018 at 04:19:22PM +0000, David Woodhouse wrote: > On Mon, 2018-02-12 at 09:26 +0100, Peter Zijlstra wrote: > > On Sun, Feb 11, 2018 at 10:13:44AM -0800, Linus Torvalds wrote: > > > > > That actually sounds like we could just > > > > > > (a) make gcc 4.5 be the minimum required version > > > > > > (b) actually error out if we find a bad compiler > > > > So the unofficial plan was to enforce asm-goto and -fentry support by > > hard failure to build, which would get us at gcc-4.6 and then remove all > > the fallback cruft needed for those features -- for x86. If we want to > > do this tree wide, that's obviously OK with me too ;-) > > This would also kill clang support, right? Yes, not sure why I should care about that though. -- To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, Feb 12, 2018 at 08:56:31AM -0800, Kees Cook wrote:
> That would be bad: Android exclusively builds with clang.
So implement asm-goto already, and do asm-cc-output while you're at it.
The whole asm-goto/jump_label stuff really does make a measureable
difference in performance, and its bloody rediculous LLVM doesn't have
it.
And no, a different interface to do something similar is not OK. We're
not going to have a second implementation of jump-labels and everything
else that uses it.
--
To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, Feb 12, 2018 at 9:05 AM, Peter Zijlstra <peterz@infradead.org> wrote: > On Mon, Feb 12, 2018 at 08:56:31AM -0800, Kees Cook wrote: >> That would be bad: Android exclusively builds with clang. > > So implement asm-goto already, and do asm-cc-output while you're at it. Yup, I've already been asking for it. I'm hoping there will be more time/attention now that retpoline is "done". -Kees
On Mon, 2018-02-12 at 09:33 -0800, Kees Cook wrote: > On Mon, Feb 12, 2018 at 9:05 AM, Peter Zijlstra <peterz@infradead.org> wrote: > > On Mon, Feb 12, 2018 at 08:56:31AM -0800, Kees Cook wrote: > >> That would be bad: Android exclusively builds with clang. > > > > So implement asm-goto already, and do asm-cc-output while you're at it. > > Yup, I've already been asking for it. I'm hoping there will be more > time/attention now that retpoline is "done". Retpoline isn't done for 32-bit x86. https://bugs.llvm.org/show_bug.cgi?id=36329
On Mon, Feb 12, 2018 at 9:36 AM, David Woodhouse <dwmw2@infradead.org> wrote: > On Mon, 2018-02-12 at 09:33 -0800, Kees Cook wrote: >> On Mon, Feb 12, 2018 at 9:05 AM, Peter Zijlstra <peterz@infradead.org> wrote: >> > On Mon, Feb 12, 2018 at 08:56:31AM -0800, Kees Cook wrote: >> >> That would be bad: Android exclusively builds with clang. >> > >> > So implement asm-goto already, and do asm-cc-output while you're at it. >> >> Yup, I've already been asking for it. I'm hoping there will be more >> time/attention now that retpoline is "done". > > Retpoline isn't done for 32-bit x86. > > https://bugs.llvm.org/show_bug.cgi?id=36329 Understood. I should have said: s/now that/when/ -Kees
diff --git a/Makefile b/Makefile index d192dd826cce..1a46f23d0974 100644 --- a/Makefile +++ b/Makefile @@ -489,6 +489,17 @@ KBUILD_CFLAGS += $(CLANG_TARGET) $(CLANG_GCC_TC) KBUILD_AFLAGS += $(CLANG_TARGET) $(CLANG_GCC_TC) endif +# check for 'asm goto' +ifeq ($(shell $(srctree)/scripts/gcc-goto.sh $(CC) $(KBUILD_CFLAGS)), y) + CC_HAVE_ASM_GOTO := 1 + KBUILD_CFLAGS += -DCC_HAVE_ASM_GOTO + KBUILD_AFLAGS += -DCC_HAVE_ASM_GOTO +endif + +ifeq ($(call cc-option-yn, -mfentry), y) + CC_HAVE_FENTRY := 1 +endif + ifeq ($(config-targets),1) # =========================================================================== # *config targets only - make sure prerequisites are updated, and descend @@ -654,12 +665,6 @@ KBUILD_CFLAGS += $(call cc-ifversion, -lt, 0409, \ # Tell gcc to never replace conditional load with a non-conditional one KBUILD_CFLAGS += $(call cc-option,--param=allow-store-data-races=0) -# check for 'asm goto' -ifeq ($(call shell-cached,$(CONFIG_SHELL) $(srctree)/scripts/gcc-goto.sh $(CC) $(KBUILD_CFLAGS)), y) - KBUILD_CFLAGS += -DCC_HAVE_ASM_GOTO - KBUILD_AFLAGS += -DCC_HAVE_ASM_GOTO -endif - include scripts/Makefile.kcov include scripts/Makefile.gcc-plugins diff --git a/arch/x86/Makefile b/arch/x86/Makefile index fad55160dcb9..35cea458a7be 100644 --- a/arch/x86/Makefile +++ b/arch/x86/Makefile @@ -158,27 +158,12 @@ ifdef CONFIG_X86_X32 endif export CONFIG_X86_X32_ABI -# -# If the function graph tracer is used with mcount instead of fentry, -# '-maccumulate-outgoing-args' is needed to prevent a GCC bug -# (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=42109) -# -ifdef CONFIG_FUNCTION_GRAPH_TRACER - ifndef CONFIG_HAVE_FENTRY - ACCUMULATE_OUTGOING_ARGS := 1 - else - ifeq ($(call cc-option-yn, -mfentry), n) - ACCUMULATE_OUTGOING_ARGS := 1 - - # GCC ignores '-maccumulate-outgoing-args' when used with '-Os'. - # If '-Os' is enabled, disable it and print a warning. - ifdef CONFIG_CC_OPTIMIZE_FOR_SIZE - undefine CONFIG_CC_OPTIMIZE_FOR_SIZE - $(warning Disabling CONFIG_CC_OPTIMIZE_FOR_SIZE. Your compiler does not have -mfentry so you cannot optimize for size with CONFIG_FUNCTION_GRAPH_TRACER.) - endif +ifndef CC_HAVE_ASM_GOTO + $(error Compiler lacks asm-goto support.) +endif - endif - endif +ifndef CC_HAVE_FENTRY + $(error Compiler lacks -mfentry support.) endif #