[Bug 246206] dev/sound: SNDCTL_SYSINFO does not report PCM_CAP_VIRTUAL
bugzilla-noreply at freebsd.org
bugzilla-noreply at freebsd.org
Tue May 5 01:42:49 UTC 2020
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=246206
Bug ID: 246206
Summary: dev/sound: SNDCTL_SYSINFO does not report
PCM_CAP_VIRTUAL
Product: Base System
Version: 12.1-RELEASE
Hardware: Any
OS: Any
Status: New
Severity: Affects Only Me
Priority: ---
Component: kern
Assignee: bugs at FreeBSD.org
Reporter: kevinz5000 at gmail.com
Attachment #214136 text/plain
mime type:
Created attachment 214136
--> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=214136&action=edit
Demonstration of issue
In sys/soundcard.h, PCM_CAP_VIRTUAL is documented as "Virtual device":
# define PCM_CAP_VIRTUAL 0x00040000 /* Virtual device */
When SNDCTL_AUDIOINFO is called on /dev/mixer, no virtual devices have have
PCM_CAP_VIRTUAL set. This means, apart from being smart, consumers of the OSS4
API can't tell virtual devices from real devices. This leads to a confusing
choice in, say, audio programs that allow selection of vchans.
A short example program that demonstrates the issue is attached. Example output
on my computer is:
$ clang vpctest.c -o vpctest
$ ./vpctest
dev /dev/dsp0.p0 virtual 0
dev /dev/dsp0.vp0 virtual 0
dev /dev/dsp0.vp1 virtual 0
dev /dev/dsp0.r0 virtual 0
dev /dev/dsp0.vr0 virtual 0
dev /dev/dsp0.vr1 virtual 0
The vp* and vr* devices are virtual, but PCM_CAP_VIRTUAL is not set.
The root cause of this issue appears to be in dev/sound/pcm/dsp.c:
ai->caps = PCM_CAP_REALTIME | PCM_CAP_MMAP | PCM_CAP_TRIGGER |
CHANNEL_GETCAPS(ch)
((ch->direction == PCMDIR_PLAY) ? PCM_CAP_OUTPUT :
PCM_CAP_INPUT);
Most notably, PCM_CAP_VIRTUAL is never set. Further, the channel capabilities
queried earlier:
caps = chn_getcaps(ch);
Do not appear to be used, either.
I suggest that virtual channels return PCM_CAP_VIRTUAL in vchan_getcaps()
(dev/sound/pcm/vchan.c), and then channel capabilities get OR'ed into ai->caps.
--
You are receiving this mail because:
You are the assignee for the bug.
More information about the freebsd-bugs
mailing list