Re: [ecasound] double buffer and kernel realtime

From: Kai Vehmanen <kvehmanen@email-addr-hidden>
Date: Sat Aug 20 2005 - 14:57:18 EEST

On Wed, 17 Aug 2005, Pedro Antonio Fructuoso Merino wrote:

>> user is root, when i use the double buffer the computer is crash. I had tryed
> This sounds like a bug, and one related to SCHED_FIFO (real-time) scheduling mode. But the
> Can`t you reproduce the problem? I use -z:db,661500 , try it. With -o rtnull the problem persists. I
> could see that computer block only happen when send commad "stop" (i didn`t test with cs-disconnect
> because i don`t want restart de computer... again).

Ok, now I could reproduce the problem. Seems like it is related to really
large db buffersize, and stopping the disk i/o threads (that's the hang
happens when the processing is stopped and with explicit "stop" and
"cs-disconnect" commands).

While waiting for the fix, add -r:-1 to the command-line to avoid the

> I could see, when i didn`t use a big buffer, that when send stop the
> computer blocks for 2-5 seconds. So is possible that it (ecasound, alsa,
> kernel, mpg321...) go into an infinite bucle.

Yes, and when run with SCHED_FIFO, hanging in system calls
(communication with mpg321 and sound device drivers) or inter-thread
synchronization (ecasound internal bugs) will result in system free.
Add -r:-1 to avoid SCHED_FIFO in all cases (of course, this increases
the risk of buffer underruns/overruns, but still better than a frozen

  links, my public keys, etc at
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement *
Ecasound-list mailing list
Received on Sat Aug 20 16:15:08 2005

This archive was generated by hypermail 2.1.8 : Sat Aug 20 2005 - 16:15:08 EEST