Page MenuHomeFreeBSD

uma: Avoid excessive per-CPU draining
ClosedPublic

Authored by des on Mar 14 2025, 12:56 AM.
Tags
None
Referenced Files
Unknown Object (File)
Fri, Sep 18, 2:51 AM
Unknown Object (File)
Thu, Sep 17, 4:13 PM
Unknown Object (File)
Wed, Sep 16, 11:32 AM
Unknown Object (File)
Sun, Sep 13, 12:57 PM
Unknown Object (File)
Sat, Sep 12, 12:51 AM
Unknown Object (File)
Fri, Sep 11, 9:11 AM
Unknown Object (File)
Fri, Sep 11, 8:49 AM
Unknown Object (File)
Wed, Sep 9, 2:16 PM

Details

Summary

After commit 389a3fa693ef, uma_reclaim_domain(UMA_RECLAIM_DRAIN_CPU)
calls uma_zone_reclaim_domain(UMA_RECLAIM_DRAIN_CPU) twice on each zone
in addition to globally draining per-CPU caches. This was unintended
and is unnecessarily slow; in particular, draining per-CPU caches
requires binding to each CPU.

Stop draining per-CPU caches when visiting each zone, just do it once in
pcpu_cache_drain_safe() to minimize the amount of expensive sched_bind()
calls.

Fixes: 389a3fa693ef ("uma: Add UMA_ZONE_UNMANAGED")

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Not Applicable
Unit
Tests Not Applicable

Event Timeline

This revision is now accepted and ready to land.Mar 14 2025, 9:03 AM

FWIW, internal testing with an equivalent diff shows the intermittent (~8%) OOM panics seen with 389a3fa693ef are now resolved. Thanks!

des edited reviewers, added: markj; removed: des.

commandeering since @markj is afk

This revision was automatically updated to reflect the committed changes.