Re: [PATCH V7 3/6] perf, x86: handle multiple records in PEBS buffer

[Peter Zijlstra]

On Wed, May 06, 2015 at 03:01:16PM +0200, Andi Kleen wrote:
> > > > +		if (p->status != (1 << bit)) {
> > > > +			u64 pebs_status;
> > > > +
> > > > +			/* slow path */
> > > > +			pebs_status = p->status & cpuc->pebs_enabled;
> > > > +			pebs_status &= (1ULL << MAX_PEBS_EVENTS) - 1;
> > > > +			if (pebs_status != (1 << bit)) {
> > > > +				perf_log_lost(event);
> > > 
> > > Does it make sense to keep an error[bit] count and only log once with the
> > > actual number in? -- when !0 obviously.
> > 
> > Yes, will do it.
> 
> If you use anything but u8 for the array member it would be too large
> for the NMI stack, and u8 is lilkely overflow prone. Would not do it.

If we're so close that 4*8=32 bytes would overflow the stack we're in
trouble already.

But we could just switch counts[] over to short and have another short
array for errors[], that way no extra bytes of stack used.
--
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/