Move the fault busy sleep into its own function.
ClosedPublic
Actions

Authored by jeff on Dec 15 2019, 2:49 AM.

Details

Reviewers

alc
kib
markj
dougm

Commits

rS355997: Move vm_fault busy logic into its own function for clarity and re-use by

Summary

The motivation here is two fold; Eventually I will want to sleep from more than a single place in vm_fault(). Secondly, fault has become a 900 line function that was only comprehensible to me after weeks. I have a series of patches that split the stages off into other functions simply to improve readability and control flow.

Diff Detail

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 28281
Build 26391: arc lint + arc unit

Event Timeline

jeff created this revision.Dec 15 2019, 2:49 AM

Harbormaster completed remote builds in B28158: Diff 65674.Dec 15 2019, 2:50 AM

jeff edited the summary of this revision. (Show Details)Dec 15 2019, 2:55 AM

jeff added reviewers: alc, kib, markj, dougm.

jeff set the repository for this revision to rS FreeBSD src repository - subversion.

Herald added a subscriber: imp. · View Herald TranscriptDec 15 2019, 2:55 AM

dougm added inline comments.Dec 15 2019, 3:06 AM

sys/vm/vm_fault.c
686	You might as well reformat the comment blocks so that they fill 79 columns.

Ignoring the KASSERT added, it looks fine.

sys/vm/vm_fault.c
847	Is this supposed to be committed ? And why ?

This revision is now accepted and ready to land.Dec 15 2019, 8:22 PM

jeff added inline comments.Dec 15 2019, 9:04 PM

sys/vm/vm_fault.c
847	I didn't realize that snuck into this patch. I believe it is impossible. OBJ_DEAD is possible but not OBJT_DEAD. We pip_wait before OBJT_DEAD so it should not be visible. This branch in aggregate has seen probably a week of stress2 without hitting this assert but I will remove it for a later commit.