Re: [PATCH] perf, tools: Always use non inlined file name for srcfile
Em Tue, Sep 01, 2015 at 11:48:39AM -0700, Andi Kleen escreveu: > On Tue, Sep 01, 2015 at 03:36:57PM -0300, Arnaldo Carvalho de Melo wrote: > > Em Tue, Sep 01, 2015 at 11:11:42AM -0700, Andi Kleen escreveu: > > > From: Andi Kleen > > > > > > When profiling the kernel with srcfile it's common to "get > > > stuck" in include. For example a lot of code uses current > > > or other inlines, so they get accounted to some random > > > include file. This is not very useful as a high level > > > categorization. > > > > Cool idea :-) > > Yes. > > It would be also nice to use this information for unwinding > (so to show the inline stack as part of the call graph) Yes, agreed. > > Why not the so much simpler: > > > > while (bfd_find_inliner_info(...)); > > > > But other than that, wouldn't be better to put an upper limit on this? > > > > Say, 1024 levels of unwinding to avoid tripping in some bfd lib bug that > > could make this function always return true and make addr2line get stuck > > in an infinite loop? > > Done. I sent a v2. Thanks, I applied v2. - Arnaldo -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH] perf, tools: Always use non inlined file name for srcfile
On Tue, Sep 01, 2015 at 03:36:57PM -0300, Arnaldo Carvalho de Melo wrote: > Em Tue, Sep 01, 2015 at 11:11:42AM -0700, Andi Kleen escreveu: > > From: Andi Kleen > > > > When profiling the kernel with srcfile it's common to "get > > stuck" in include. For example a lot of code uses current > > or other inlines, so they get accounted to some random > > include file. This is not very useful as a high level > > categorization. > > Cool idea :-) Yes. It would be also nice to use this information for unwinding (so to show the inline stack as part of the call graph) > Why not the so much simpler: > > while (bfd_find_inliner_info(...)); > > But other than that, wouldn't be better to put an upper limit on this? > > Say, 1024 levels of unwinding to avoid tripping in some bfd lib bug that > could make this function always return true and make addr2line get stuck > in an infinite loop? Done. I sent a v2. -Andi -- a...@linux.intel.com -- Speaking for myself only -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH] perf, tools: Always use non inlined file name for srcfile
Em Tue, Sep 01, 2015 at 11:11:42AM -0700, Andi Kleen escreveu:
> From: Andi Kleen
>
> When profiling the kernel with srcfile it's common to "get
> stuck" in include. For example a lot of code uses current
> or other inlines, so they get accounted to some random
> include file. This is not very useful as a high level
> categorization.
Cool idea :-)
> For example just profiling the idle loop usually shows
> mostly inlines, so you never see the actual cpuidle file.
> @@ -170,6 +171,15 @@ static int addr2line(const char *dso_name, u64 addr,
>
> bfd_map_over_sections(a2l->abfd, find_address_in_section, a2l);
>
> + if (a2l->found && unwind_inlines) {
> + bool found;
> +
> + do {
> + found = bfd_find_inliner_info(a2l->abfd, &a2l->filename,
> + &a2l->funcname,
> &a2l->line);
> + } while (found);
> + }
Why not the so much simpler:
while (bfd_find_inliner_info(...));
But other than that, wouldn't be better to put an upper limit on this?
Say, 1024 levels of unwinding to avoid tripping in some bfd lib bug that
could make this function always return true and make addr2line get stuck
in an infinite loop?
- Arnaldo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
