Re: [PATCH] Fix race in ring_buffer_consume(): Replace ring_buffer_consume and ring_buffer_peek with ring_buffer_get_event

[Ingo Molnar <mingo@elte.hu>]

* Indan Zupancic <indan@nul.nu> wrote:

> On Mon, December 29, 2008 13:24, Ingo Molnar wrote:
> >
> > * Indan Zupancic <indan@nul.nu> wrote:
> >
> >> Original mail was mangled, patch resent via git.
> >>
> >> Signed-off-by: Indan Zupancic <indan@nul.nu>
> >> ---
> >>  include/linux/ring_buffer.h   |    4 +---
> >>  kernel/trace/ring_buffer.c    |   39
> >> ++++++++-------------------------------
> >>  kernel/trace/trace.c          |   15 ++++++++-------
> >>  kernel/trace/trace_selftest.c |    2 +-
> >>  4 files changed, 18 insertions(+), 42 deletions(-)
> >
> > there's been a number of updates here - could you please do a patch
> > against tip/master:
> >
> >   http://people.redhat.com/mingo/tip.git/README
> 
> Thanks for that readme, now I discovered git remote, which is not
> mentioned often enough for some reason.
> 
> A lot changed indeed, most importantly the race that I hit is
> fixed in tip by improved locking. Replace ring_buffer_consume()
> and ring_buffer_peek() with ring_buffer_get_event() or not is
> just a matter of taste now. Are you still interested in such a
> patch?

the code gets simpler and more readable, so why not? Steve, any 
objections?

> (34 EXPORT_SYMBOL_GPLs in ring_buffer.c seems a bit excessive though.)

agreed.

	Ingo