Re: [RFC GIT PULL] x86 entry / perf stacktrace changes

[Ingo Molnar <mingo@elte.hu>]

* Frederic Weisbecker <fweisbec@gmail.com> wrote:

> Ingo,
> 
> Please pull the perf/stacktrace branch that can be found at:
> 
> git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing.git
> 	perf/stacktrace
> 
> Or may be not yet. It's still in RFC because I would like
> to ensure I did not break cfi annotations. I hope Jan beulich
> or others can have a look.
> 
> Sorry to mess up irq entry and perf stacktrace changes in the same
> topic. But the changes happen to be very interconnected.
> 
> Several cleanups, fixes and optimizations for the x86-64
> stacktraces and irq entry.
> 
> Last patch seems to add a bit of overhead in the irq entry
> (one more "push") but given the changes made in previous
> patches, the end result is a more optimized and more clear irq
> entry.
> 
> Thanks,
> 	Frederic
> ---
> 
> Frederic Weisbecker (6):
>       x86: Save stack pointer in perf live regs savings
>       x86: Fetch stack from regs when possible in dump_trace()
>       x86,64: Simplify save_regs()
>       x86,64: Separate arg1 from rbp handling in SAVE_REGS_IRQ
>       x86: Remove useless unwinder backlink from irq regs saving
>       x86: Don't use frame pointer to save old stack on irq entry
> 
> 
>  arch/x86/include/asm/perf_event.h |    5 +++
>  arch/x86/kernel/dumpstack_64.c    |   37 +++-----------------
>  arch/x86/kernel/entry_64.S        |   69 ++++++++++++++++--------------------
>  3 files changed, 41 insertions(+), 70 deletions(-)

Pulled, thanks a lot Frederic! The irq backtrace fixes look pretty 
important - if there's any CFI annotation detail to be fixed we can 
do it in a delta.

Thanks,

	Ingo