On Thu, 2008-11-27 at 17:46 +0000, Adam Sampson wrote:
Fernando Lopez-Lezcano
<nando(a)ccrma.Stanford.EDU> writes:
A caveat is that the realtime preemption patches
for 2.6.26.x don't
seem to work too well (midi problems when used with external
controllers).
Could you be more specific? It sounds like the problem I'm currently
seeing with 2.6.26.6-rt11.
I had problems with earlier RT kernels such as 2.6.25.8-rt7 occasionally
locking up on my machine (in a way that das_watchdog wouldn't recover
from but the kernel software watchdog would detect), but 2.6.26.6-rt11
has been running happily for three weeks now.
I have seen it lock up hard. Not frequently but it happens.
The downside is that the MIDI port on my Audiophile
2496 doesn't work:
trying to open /dev/snd/midiC0D0 always returns EBUSY, even when nothing
else has it open, and events from it don't seem to make it into the
sequencer. In addition, the snd_ice1712 module seems to have a reference
count of 2 even when nothing has any of the sound devices open...
What I'm seeing is that the alsa sequencer does not work when there are
rawmidi ports involved. That is, it works as long as the sequencer ports
are software, not hardware.
And rawmidi works fine by itself. But as soon as you try to use an
external device through the alsa sequencer the client hangs. And it
hangs hard, you need a reboot to get rid of it (the rest of the machine
keeps running just fine - this appears to be a deadlock inside the
sequencer code).
I posted a long thread on lkml, with printk information and so on, and
there were some suggestions (one made midi input work) but no kernel
guru stepped up to fix the problem or request more information. I latter
discovered that the problem was not solved, the internal timer of the
alsa sequencer seem to be broken. I have not had time to repost a
message to lkml and try to debug this.
-- Fernando