[pvrusb2] My tuners occasionally hang

[Jan Ceuleers]

Hi there.

A few weeks ago I reported that occasionally my four HVR-1900 tuners
hang. The consensus of opinion at the time was that this was due to a
generic USB problem in the kernel. When I upgraded to a newer kernel
system stability seemed to be much better.

That is certainly the case, but the problem hasn't gone away completely.
I woke up this morning to an email alerting me to another zero-length
recording. When I checked the logs I saw a few pvrusb2-related oopses so
I thought I'd report those here.

The log is at the link below. It's just over 1MB.

http://filebin.ca/GJn8SLB6PB/pvrusb2.oops.log.2

You will see that the problem begins at 00:15:34. Six seconds later
mythtv calls the script that corresponds with the end of a recording,
which notices that the recorded file is empty. This script schedules a
tuner power cycle.

Tuner power cycling is done from a cronjob run by root (because the
mythtv user does not have privilege to do so itself). The first attempt
begins at 00:20:01. Power to the tuners is actually killed at 00:20:09
and restored at 00:20:20.

Four seconds later the kernel says that a pvrusb2 thread has been
blocked for more than 120 seconds. Perhaps my power cycling script
should also unload/reload a bunch of modules.

Since this power cycling attempt has not been successful, the cronjob
fires again at 00:30:01. Etc.

janc at zotac:~$ uname -a
Linux zotac 3.3.0+ #4 SMP Mon Apr 23 08:43:03 CEST 2012 x86_64 x86_64
x86_64 GNU/Linux

This is a self-compiled kernel with one RAID-related patch applied.
Otherwise it's vanilla 3.3.0. When I reboot after sending this email I
will boot a self-compiled 3.4 kernel because I now know that the kernel
I'm currently running contains a low-probability but high-impact
RAID-related bug which (if it occurs) manifests itself on shutdown.
Fingers crossed.

Thanks, Jan