lkml.org 
[lkml]   [2004]   [Nov]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: Real-Time Preemption, -RT-2.6.10-rc2-mm3-V0.7.31-7

* Ingo Molnar <mingo@elte.hu> wrote:

> --- ./drivers/alsa/alsa_driver.c.orig 2004-11-26 14:11:26.000000000 +0100
> +++ ./drivers/alsa/alsa_driver.c 2004-11-26 14:24:41.000000000 +0100

i think the one below is a better approach - it will only trace events
that the ALSA driver has reported to be an xrun. I.e. the full latency
path from the point where poll() is called, up to the point where jackd
[after the latency has occured] considers it an xrun worth
counting/reporting. The tracing restarts at every poll(), so only the
latency is captured. If you up your tracebuffer to 40K+ entries then i'd
suggest to use the following trace settings:

echo 0 > /proc/asound/card0/pcm0p/xrun_debug
echo 1 > /proc/sys/kernel/trace_user_triggered
echo 0 > /proc/sys/kernel/trace_freerunning
echo 0 > /proc/sys/kernel/preempt_max_latency
echo 0 > /proc/sys/kernel/preempt_thresh
echo 0 > /proc/sys/kernel/preempt_wakeup_timing
i.e. dont use trace_freerunning - this will give much easier to parse
traces.

a suggestion wrt. the format of the .trc files: it would be nice if you
could dump the PIDs of all relevant tasks into it too, to make it easier
to identify who causes what latency. Ideally it would be useful to
have a more symbolic trace - i.e. instead of:

3570 00000000 254981.991ms (+0.000ms): up (ext3_orphan_del)
it would be:

jackd-3570 00000000 254981.991ms (+0.000ms): up (ext3_orphan_del)
but this is easier done in the kernel - some of the tasks involved in a
latency might be long gone by the time you detect the xrun.

Ingo

--- ./drivers/alsa/alsa_driver.c.orig 2004-11-26 14:11:26.000000000 +0100
+++ ./drivers/alsa/alsa_driver.c 2004-11-26 15:31:37.000000000 +0100
@@ -1077,13 +1077,16 @@ alsa_driver_xrun_recovery (alsa_driver_t
&& driver->process_count > XRUN_REPORT_DELAY) {
struct timeval now, diff, tstamp;
driver->xrun_count++;
+ gettimeofday(0,0);
gettimeofday(&now, 0);
snd_pcm_status_get_trigger_tstamp(status, &tstamp);
timersub(&now, &tstamp, &diff);
*delayed_usecs = diff.tv_sec * 1000000.0 + diff.tv_usec;
+#if 0
fprintf(stderr, "\n\n**** alsa_pcm: xrun of at least %.3f "
"msecs\n\n",
*delayed_usecs / 1000.0);
+#endif
}

if (alsa_driver_stop (driver) ||
@@ -1185,6 +1188,7 @@ alsa_driver_wait (alsa_driver_t *driver,
nfds++;
}

+ gettimeofday(0,1);
poll_enter = jack_get_microseconds ();

if (poll (driver->pfd, nfds, driver->poll_timeout) < 0) {
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
\
 
 \ /
  Last update: 2005-03-22 14:08    [W:0.233 / U:0.880 seconds]
©2003-2014 Jasper Spaans. Advertise on this site