[vdr] Re: ARM crashed while cutting, VDR watchdog timer failed.

[Klaus Schmidinger <Klaus.Schmidinger@cadsoft.de>]

Carsten Koch wrote:
> 
> I did a simple vdr cutting: setting two marks and starting
> the cutting process. Then I wanted to see how far the new
> recording was, so I brought up the recordings menu, selected
> the %name entry and pressed OK twice.
> The picture on my screen disappeared, but vdr's watchdog
> timer did not catch the error.

VDR's watchdog timer only catches cases where the main program loop
isn't processed for an extended period of time. If the main loop
is processed regularly, VDR has no way of telling if something is
wrong at the moment (at least none that I know of).

> Here is what the logs say:
> 
> Dec 30 02:22:02 vdr vdr[3053]: creating directory /video/%Humoristen_des_Nordens
> Dec 30 02:22:02 vdr vdr[3053]: creating directory /video/%Humoristen_des_Nordens/2001-12-26.18:28.50.98.rec
> Dec 30 02:22:02 vdr vdr[3053]: loading /video/Humoristen_des_Nordens/2001-12-26.18:28.50.98.rec//marks.vdr
> Dec 30 02:22:02 vdr vdr[3053]: playing '/video/Humoristen_des_Nordens/2001-12-26.18:28.50.98.rec/001.vdr'
> Dec 30 02:22:02 vdr vdr[3053]: recording to '/video/%Humoristen_des_Nordens/2001-12-26.18:28.50.98.rec/001.vdr'
> Dec 30 02:22:03 vdr vdr[3169]: video cutting thread started (pid=3169)
> Dec 30 02:22:03 vdr vdr[3053]: info: Editing process started
> Dec 30 02:22:05 vdr vdr[3053]: max. latency time 3 seconds
> Dec 30 02:22:10 vdr vdr[3105]: output thread ended (pid=3105)
> Dec 30 02:22:10 vdr vdr[3106]: input thread ended (pid=3106)
> Dec 30 02:22:17 vdr vdr[3053]: max. latency time 6 seconds
> Dec 30 02:23:36 vdr vdr[3053]: loading /video/%Humoristen_des_Nordens/2001-12-26.18:28.50.98.rec//marks.vdr
> Dec 30 02:23:36 vdr vdr[3053]: replay /video/%Humoristen_des_Nordens/2001-12-26.18:28.50.98.rec
> Dec 30 02:23:36 vdr vdr[3053]: playing '/video/%Humoristen_des_Nordens/2001-12-26.18:28.50.98.rec/001.vdr'
> Dec 30 02:23:39 vdr vdr[3178]: output thread started (pid=3178)
> Dec 30 02:23:39 vdr vdr[3179]: input thread started (pid=3179)
> Dec 30 02:23:53 vdr kernel: dvb0: ARM crashed!
> Dec 30 02:26:31 vdr vdr[3169]: end video cutting thread
> 
> after this, I pressed Ctrl-C in the vdr shell window
> and VDR still did not go away.
> So I entered "killproc vdr" in another shell, which finally
> did give the following log entry (over 6 minutes late, my
> watchdog timer is set to 15 seconds):
> 
> Dec 30 02:30:36 vdr vdr[3053]: PANIC: watchdog timer expired - exiting!
> 
> I have observed this exact same behaviour before, so I guess
> it should be easy to reproduce.

I haven't had any ARM crashes in months - I wouldn't know how to
easily reproduce this...

> Assuming that the firmware bug that causes the ARM to crash
> hard to fix, is there a workaround in vdr that could be
> implemented? Maybe an extra usleep somewhere?
> 
> And why does the watchdog timer not work?

See above.

> I am using kernel 2.4.16-xfs.

I'm running kernel 2.4.10 without any problems.
The only thing that happens sometimes (rarely) is that a recording tries to
start, but then the driver doesn't deliver any data. After a while VDR
recognizes this and performs a restart (with driver reload). After that, the
recording works just fine.

Klaus
-- 
_______________________________________________________________

Klaus Schmidinger                       Phone: +49-8635-6989-10
CadSoft Computer GmbH                   Fax:   +49-8635-6989-40
Hofmark 2                               Email:   kls@cadsoft.de
D-84568 Pleiskirchen, Germany           URL:     www.cadsoft.de
_______________________________________________________________