ktls: Improve handling of the bind_threads tunable a bit
ClosedPublic
Actions

Authored by markj on Jan 14 2021, 4:36 PM.

Details

Reviewers

jhb
gallatin

Commits

rG4dc1b17dbb9f: ktls: Improve handling of the bind_threads tunable a bit

Summary

Only check for empty domains if we actually tried to configure domain affinity in the first place. Otherwise setting bind_threads=1 will always cause the sysctl value to be reported as zero. This is harmless since the threads end up being bound, but it's confusing.
Try to improve the sysctl description a bit.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

markj created this revision.Jan 14 2021, 4:36 PM

Herald added a subscriber: imp. · View Herald TranscriptJan 14 2021, 4:36 PM

markj requested review of this revision.Jan 14 2021, 4:36 PM

Harbormaster completed remote builds in B36204: Diff 82272.Jan 14 2021, 4:36 PM

markj added a parent revision: D28160: netipsec: Fix handling of unmapped mbufs.Jan 14 2021, 4:36 PM

markj added a child revision: D28162: arm64: Stop setting VM_BCACHE_SIZE_MAX.

markj added reviewers: jhb, gallatin.Jan 14 2021, 4:37 PM

markj removed a child revision: D28162: arm64: Stop setting VM_BCACHE_SIZE_MAX.Jan 14 2021, 4:45 PM

gallatin accepted this revision.Jan 14 2021, 6:14 PM

This revision is now accepted and ready to land.Jan 14 2021, 6:14 PM

I saw this the other day and almost thought that we should have separate knobs for "bind threads to cores" and "bind sessions to NUMA domains" rather than have ktls_bind_threads serve as a kind of tri-state. However, I think this approach is fine.

Closed by commit rG4dc1b17dbb9f: ktls: Improve handling of the bind_threads tunable a bit (authored by markj). · Explain WhyJan 20 2021, 2:33 AM

This revision was automatically updated to reflect the committed changes.

markj added a commit: rG4dc1b17dbb9f: ktls: Improve handling of the bind_threads tunable a bit.