kern/52442: x68k's vs(4) is broken after in-kernel mixing

[isaki%NetBSD.org@localhost]

>Number:         52442
>Category:       kern
>Synopsis:       x68k's vs(4) is broken after in-kernel mixing
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sun Jul 30 12:55:00 +0000 2017
>Originator:     Tetsuya Isaki
>Release:        NetBSD-current
>Organization:
NetBSD
>Environment:
NetBSD 8.99.1 x68k
>Description:
After merging in-kernel mixing, x68k's vs(4) has several
different problems concurrently.
This is -current matter but is the same with netbsd-8 branch.

#1 device autoconfig failed. [RESOLVED in -current]

  The combination of encoding parameters inquired by the
  MI audio layer has changed.

#2 filter parameter (or something) mismatch.

  MSM6258 can only play its own ADPCM and current MI audio layer
  only supprots SLINEAR devices.
  Therefore some conversion between MI and MD needed.
  But that parameter or something seems to be incorrect now.

#3 ring buffer (or something) is strange.

  Even if problem #2 is solved,
  when playback starts, white noise is output first.
  After a while, the audio is played.  Probably, playback is
  started even though there is no output data yet.
  Then after a while, the same place is played many times because
  it runs the ring buffer even though there is no new data.

#4 frequency mismatch.

  [Though this is not mandatory]
  Now MD layer reports 16kHz as our supported frequency for MI
  layer but set hardware 15.6kHz.
  MSM6258 can only play 15.6kHz, 10.4kHz, 7.8kHz, 5.2kHz, and
  3.9kHz.  and MI audio layer does not support any of them.
  The audio layer has aurateconv, so I'm happy if I can use it.

>How-To-Repeat:
See dmesg on x68k.
And run audioplay(1) on x68k.
>Fix:
N/A