Re: sb16 gets wedged when doing overlapped i/o (full duplex)

["John S. Denker" <jsd@monmouth.com>]

Uros Bizjak wrote:

>   From the error message you posted about this bug, it looks that driver
> is trying to call prepare() operator on a substream that is still
> in SNDRV_PCM_STATE_RUNNING state. Perhaps Jaroslav could tell us if this
> can happen if hardware pointers are not updated anymore (because sb16 DMA
> engine gets stucked somehow). Do you get an XRUN message in
> /var/log/messages?

Excellent question!

1) No xrun messages in /var/log/messages, but....

2) My program intentionally causes an overrun on input once
every 0.5 seconds.  It catches this condition, and recovers
by calling snd_pcm_prepare().  It does not print a message
about this (unless super-specially requested).

I realize that other programs may cause far fewer overruns,
which may (or may not :-) explain why others don't suffer
quite so badly from this bug.  Even so, it remains a bug.
An overrun shouldn't take down the whole system.

3) As previously indicated, if I load the system, I can cause
a few audible underruns on output.  But doing this has no
discernible effect on the incidence of the bug in question.



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf