Re: [PATCH] x86: mostly disable '-maccumulate-outgoing-args'

[Josh Poimboeuf]

On Thu, Mar 16, 2017 at 01:32:01PM -0400, Steven Rostedt wrote:
> On Thu, 16 Mar 2017 10:42:08 -0500
> Josh Poimboeuf <jpoimboe@xxxxxxxxxx> wrote:
> 
> > Signed-off-by: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
> > ---
> >  arch/x86/Makefile        | 29 +++++++++++++++++++++++++----
> >  arch/x86/Makefile_32.cpu | 18 ------------------
> >  arch/x86/kernel/ftrace.c |  6 ++++++
> >  scripts/Kbuild.include   |  4 ++++
> >  4 files changed, 35 insertions(+), 22 deletions(-)
> > 
> > diff --git a/arch/x86/Makefile b/arch/x86/Makefile
> > index 2d44933..fa45989b 100644
> > --- a/arch/x86/Makefile
> > +++ b/arch/x86/Makefile
> > @@ -120,10 +120,6 @@ else
> >          # -funit-at-a-time shrinks the kernel .text considerably
> >          # unfortunately it makes reading oopses harder.
> >          KBUILD_CFLAGS += $(call cc-option,-funit-at-a-time)
> > -
> > -        # this works around some issues with generating unwind tables in older gccs
> > -        # newer gccs do it by default
> > -        KBUILD_CFLAGS += $(call cc-option,-maccumulate-outgoing-args)
> >  endif
> >  
> >  ifdef CONFIG_X86_X32
> > @@ -147,6 +143,31 @@ ifeq ($(CONFIG_KMEMCHECK),y)
> >  	KBUILD_CFLAGS += $(call cc-option,-fno-builtin-memcpy)
> >  endif
> >  
> > +# If the function graph tracer is used with mcount instead of fentry,
> > +# '-maccumulate-outgoing-args' is needed to prevent gcc bug
> 
> 			"to prevent a gcc bug"

It was

  "to prevent gcc bug https://gcc.gnu.org/bugzilla/show_bug.cgi?id=42109";

where "gcc bug" was an adjective and the URL was a noun.  But yeah,
that's kind of confusing, and the line wrap made it more so.  Maybe I'll
change it to

  "to prevent a gcc bug (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=42109)"

and a similar change for the jump label bug comment.

> > +# 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 ($(call cc-option, -mfentry),)
> 
> Hmm, the blank entry makes me nervous. I wonder if it would be better
> if we had ifneq ($(call cc-option-yn, -mfentry),y)
> 
> Unfortunately, there's one of each in the existing kernel, so there is
> really no precedence.

Either way seems fine.  I'll go with your suggested change.

> > +	ACCUMULATE_OUTGOING_ARGS := 1
> > +    endif
> > +  endif
> > +endif
> > +
> > +# Jump labels need '-maccumulate-outgoing-args' for gcc < 4.5.2 to prevent
> 
> Can we make a test instead? I hate testing versions, and things get
> backported all the time. We usually like to have a test case instead of
> relying on versions. Not to mention, a newer gcc may one day break.

Tests are generally better, but I'm not sure how to test for this
cleanly.  The test is rather big for embedding in a makefile:

  https://gcc.gnu.org/bugzilla/attachment.cgi?id=22199

Any ideas?

-- 
Josh