Messages in this thread | | | Subject | Re: Funny perf | From | Peter Zijlstra <> | Date | Wed, 23 Feb 2011 11:21:48 +0100 |
| |
On Thu, 2011-02-17 at 19:26 +0100, Borislav Petkov wrote: > > so this looks like some of the samples do not belong in the perf.data > output with that garbled cpu field. It lets me also assume that the > sample->time field is also crap so the timestamp check is falsely > tripped.
Most interesting, I haven't tried to reproduce, but it does sound smelly.
I did have a quick peek at some code I suspected but that all seems in order. I guess someone will have to sit down and figure this out proper, if nobody steps up I guess it'll have to be me, but that also means it might take a while since I've got some other pending stuff to stare at before this :/
| |