Re: [PATCH 03/23] perf: Add ability to attach user level registersdump to sample

[Frederic Weisbecker]

On Fri, Jun 29, 2012 at 01:30:12PM +0200, Jiri Olsa wrote:
> On Fri, Jun 29, 2012 at 01:23:51PM +0200, Frederic Weisbecker wrote:
> 
> SNIP
> 
> > > 
> > > well, thats basically what we have now..
> > > 
> > > when the kernel is compiled for 32 bits, the bitmask allows to store:
> > > 
> > > ax, bx, cx, dx, si, di, bp, sp, ip, flags, cs, ds, es, fs, gs, ss
> > > 
> > > and when  kernel is compiled for 64 bits, the bitmask adds 64bit stuff:
> > > 
> > > ax, bx, cx, dx, si, di, bp, sp, ip, flags, cs, ds, es, fs, gs, ss
> > > r8, r9, r10, r11, r12, r13, r14, r15
> > > 
> > > 
> > > - 32 bits kernel is straightforward
> > > 
> > > - for 64 bits kernel we store whatever bitmask instructs to,
> > >   regardless if we are in compat task or native 64,
> > >   user space will deal with that in post processing
> > 
> > I can think of the crazy scenario where perf itself is compat and there
> > are 64 bits apps that will be profiled by perf. It means perf must check
> > if the kernel is 32 or 64 and depending on this, request r8-r15 or not.
> > 
> > May be that's just too unlikely to matter.
> 
> hopefully ;) anyway the kernel interface stays the same in this case as well
> 
> we can add this check later when/if we deal with compat tasks

Right!
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/