vmm: Dynamically allocate a couple of per-CPU state save areas
ClosedPublic
Actions

Authored by markj on Apr 25 2023, 5:22 PM.

Details

Reviewers

rew
corvink

Group Reviewers

bhyve

Commits

rG0a300a95c10f: vmm: Dynamically allocate a couple of per-CPU state save areas
rG74ac712f72cf: vmm: Dynamically allocate a couple of per-CPU state save areas

Summary

This avoids bloating the BSS when MAXCPU is large.

No functional change intended.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Skipped

Unit

Tests Skipped

Build Status

Buildable 51179
Build 48070: arc lint + arc unit

Event Timeline

markj created this revision.Apr 25 2023, 5:22 PM

Herald added subscribers: bcran, rgrimes, imp. · View Herald TranscriptApr 25 2023, 5:22 PM

markj requested review of this revision.Apr 25 2023, 5:22 PM

Harbormaster completed remote builds in B51179: Diff 121012.Apr 25 2023, 5:22 PM

Would anyone be willing to test this on an AMD system?

In D39805#906206, @markj wrote:

Would anyone be willing to test this on an AMD system?

What kind of the tests? Just run VM? I can test it on Ryzen 3700x.

In D39805#906206, @markj wrote:

Would anyone be willing to test this on an AMD system?

works for me

sys/amd64/vmm/amd/svm.c
257–266	should MAXCPU be pruned at some point?

This revision is now accepted and ready to land.Apr 25 2023, 7:24 PM

In D39805#906277, @rew wrote:

In D39805#906206, @markj wrote:

Would anyone be willing to test this on an AMD system?

works for me

Thanks.

In D39805#906265, @afedorov wrote:

In D39805#906206, @markj wrote:

Would anyone be willing to test this on an AMD system?

What kind of the tests? Just run VM? I can test it on Ryzen 3700x.

Yes, booting a guest ought to be enough. Not really necessary now that Rob tried it.

sys/amd64/vmm/amd/svm.c
257–266	Yeah, that's a follow-up step that I just didn't take since `asid[]` is much smaller.

corvink accepted this revision.Apr 26 2023, 6:08 AM

Closed by commit rG74ac712f72cf: vmm: Dynamically allocate a couple of per-CPU state save areas (authored by markj). · Explain WhyApr 26 2023, 2:14 PM

This revision was automatically updated to reflect the committed changes.

markj added a commit: rG74ac712f72cf: vmm: Dynamically allocate a couple of per-CPU state save areas.

somebody attempted a different kind of test: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=271251 — Kernel panics when loading vmm module with VT-x disabled

markj added a commit: rG0a300a95c10f: vmm: Dynamically allocate a couple of per-CPU state save areas.May 11 2023, 1:33 PM