hwpmc_amd: improve assertions in amd_pcpu_fini()
AbandonedPublic
Actions

Authored by mhorne on Aug 1 2023, 4:24 PM.

Details

Reviewers

jkoshy
gnn
ali_mashtizadeh.com

Group Reviewers

pmc

Summary

We don't need to stop the counters here, as the comment already implies.
We can rely on the stop method being called for every active PMC before
finalization and assert this.

Test Plan

Unload hwpmc module with a mixed set of system and process-mode counters running.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 52936
Build 49827: arc lint + arc unit

Event Timeline

mhorne created this revision.Aug 1 2023, 4:24 PM

Herald added a reviewer: gnn. · View Herald TranscriptAug 1 2023, 4:24 PM

Herald added a subscriber: imp. · View Herald Transcript

mhorne requested review of this revision.Aug 1 2023, 4:24 PM

Harbormaster completed remote builds in B52936: Diff 125432.Aug 1 2023, 4:24 PM

mhorne added a parent revision: D41277: hwpmc_amd: check extended CPUID bits.Aug 1 2023, 4:24 PM

mhorne added a child revision: D41279: pmc.amd(3): rename from pmc.k8(3).

mhorne edited the test plan for this revision. (Show Details)Aug 1 2023, 4:29 PM

jkoshy added inline comments.Aug 2 2023, 7:36 AM

sys/dev/hwpmc/hwpmc_amd.c
774	We might want to check the actual number of PMCs on the CPU too? E.g. this loop assumes that AMD CPUs always have PMCs implemented at indices 0..5, but older CPUs only have 4 (per memory).
778	Add symbolic constants here for the '6', '12', and '16'?