Messages in this thread | | | Date | Tue, 26 May 2015 06:44:31 -0700 | Subject | Re: [PATCH v2 01/11] perf,x86: Fix event/group validation | From | Stephane Eranian <> |
| |
On Tue, May 26, 2015 at 6:22 AM, Peter Zijlstra <peterz@infradead.org> wrote: > On Tue, May 26, 2015 at 05:25:59AM -0700, Stephane Eranian wrote: >> > IIRC the problem was that the copy from c2 into c1: >> > >> > if (c1 && (c1->flags & PERF_X86_EVENT_DYNAMIC)) { >> > bitmap_copy(c1->idxmsk, c2->idxmsk, X86_PMC_IDX_MAX); >> > c1->weight = c2->weight; >> > c2 = c1; >> > } >> > >> > is incomplete. For instance, flags is not copied, and some code down the >> > line might check that and get wrong flags. >> > >> Ok, now I remember this code. It has to do with incremental scheduling. >> Suppose E1, E2, E3 events where E1 is exclusive. The first call is >> for scheduling E1. It gets to get_event_constraint() which "allocates" a >> dynamic constraint. The second call tries to schedule E1, E2. But the >> second time for E1, you already have the dynamic constraint allocated, so >> this code is reusing the constraint storage and just updates the bitmask >> and weight. >> >> Now, that the storage is not actually dynamic (kmalloc'd), but taken from a >> fixed size array in cpuc, I believe we can simplify this and "re-allocate" >> the constraint for each incremental call to intel_get_event_constraints(). >> Do you agree? > > That would probably work, the whole incremental thing seems superfluous > to me.
Well, you could pass the whole lists (4) of events in one call and let the arch level sort it out. It could do one pass and stop at first error to get a fixed bound like today. But it would need to deal with groups as well...so not so easy and why repeat the group processing for each arch?
| |