Differential D60509

sound: Busy the device in sysctl_dev_pcm_mode()
Needs ReviewPublic
Actions

Authored by christos on Fri, Oct 9, 9:12 AM.

Project Tags

None

Referenced Files

None

Subscribers

imp

Details

Reviewers

markj
emaste
kib

Summary

PCM_LOCK does not protect anything here, since the counts are plain
ints, and the mixer pointer is set once in mixer_init() and never
cleared. It also does not protect against the device being freed.

Busy the device instead, as the other handlers do.

MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Skipped

Unit

Tests Skipped

Build Status

Buildable 77896
Build 74779: arc lint + arc unit

Event Timeline

christos created this revision.Fri, Oct 9, 9:12 AM

Herald added a subscriber: imp. · View Herald TranscriptFri, Oct 9, 9:12 AM

christos requested review of this revision.Fri, Oct 9, 9:12 AM

Harbormaster completed remote builds in B77896: Diff 189176.Fri, Oct 9, 9:12 AM

christos added a parent revision: D60507: sound: Busy devices before giving up the topology lock.Fri, Oct 9, 9:30 AM

christos added a child revision: D60512: sound: Stop taking PCM_LOCK to create and destroy channels.Fri, Oct 9, 9:33 AM

PCM_LOCK does not protect anything here, since the counts are plain ints,

I don't understand this reasoning. What does being a plain int have to do with anything?

and the mixer pointer is set once in mixer_init() and never cleared.

It is never cleared, but it may be freed before the sysctl is destroyed.

sys/dev/sound/pcm/sound.c
268	This check is racy, no? What happens if we're racing with pcm_unregister()?

Revision Contents
Changeset List

Path

Size

sys/

dev/

sound/

pcm/

sound.c

4 lines

Diff 189176

View Options

sound: Busy the device in sysctl_dev_pcm_mode()Needs ReviewPublicActions