fuse
ClosedPublic
Actions

Authored by asomers on Jun 16 2021, 12:59 AM.

Details

Reviewers

emaste
pfg

Commits

rG6b2266b050a6: fusefs: support EVFILT_WRITE on /dev/fuse
rG7b096a997fad: fusefs: support EVFILT_WRITE on /dev/fuse
rG7b8622fa220b: fusefs: support EVFILT_WRITE on /dev/fuse

Summary

fusefs: support EVFILT_WRITE on /dev/fuse

/dev/fuse is always ready for writing, so it's kind of dumb to poll it.
But some applications do it anyway. Better to return ready than EINVAL.

MFC after: 2 weeks

Test Plan

Existing kqueue test modified to use kqueue for writing as well as for reading. Tested with a real file system from https://github.com/Sherlock-Holo/fuse3

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

asomers created this revision.Jun 16 2021, 12:59 AM

Herald added a subscriber: imp. · View Herald TranscriptJun 16 2021, 12:59 AM

asomers requested review of this revision.Jun 16 2021, 12:59 AM

Harbormaster completed remote builds in B39927: Diff 90941.Jun 16 2021, 1:00 AM

emaste added inline comments.Jun 16 2021, 4:42 PM

tests/sys/fs/fusefs/mockfs.cc
870–872	Curious about this change

asomers added inline comments.Jun 16 2021, 4:58 PM

tests/sys/fs/fusefs/mockfs.cc
870–872	It's necessary to make the tests pass, though I could've done it a different way. The problem is that line 943 adds the EVFILT_WRITE filter to the same fd. When control comes back around to here, kevent() immediately returns because EVFILT_WRITE is still set, and and file descriptor is always writable. So adding EV_ONESHOT at 943 prevents kevent here from returning too soon. And I added EV_ONESHOT here for symmetry's sake, and because line 943 doesn't care about readability. Alternatively, I could've added EV_ONESHOT at line 943 only; that would've worked. Or, I could've set EV_DISABLE on EVFILT_READ at line 946 or here.