[mythtv-users] freeze while seeking

[R. G. Newbury]

Pavol Droba wrote:
> On Thu, Apr 13, 2006 at 02:57:08PM -0400, R. G. Newbury wrote:
> 
>>Pavol Droba wrote:
>>
>>>Hi,
>>>
>>>I have a problem with mythtv and I wonder if somebody have experienced something similar.
>>>
>>>I have VIA EPIA SP13000 system runnig epios (gentoo based distro for via boards)  
>>>with latest updates. This means mythtv 0.19 and openchrome driver (snapshot from 2006-03-06).
>>>
>>>Mythtv is configured to use via xvmc. Everything works smoothly, except of one problem.
>>>Occasionally when seeking the recorded shows with FF or Rewind buttons, graphical system freezes.
>>>
>>>Somethimes I'm able to switch to a text console, but it is not always the case. If I remotely 
>>>kill whole X system, it restars ok, but any attempt to play a video using xvmc, gives instantly 
>>>another freeze. This time it just shows a green screen and doesn't allow to switch to text console. 
>>>So I have to reboot the mashine remotely.
>>>
>>>I'm not sure if the problem resides in mythtv or in openchrome or somewhere else. So far I have
>>>encountered it only when using mythtv. I haven't seen anything unusual in logs.
>>
>>I have the same hardware now running Fedora Core 5 with the openchrome 
>>driver (washington.kelkoo.net/epia/FC5/rpms).
>>
>>I have experienced this problem as well, but only when rapidly moving 
>>backwards and forwards in a recording. I have not tested it but it seems 
>>that ff and rewind are much more fragile in recordings which have NOT 
>>been commflagged and sometimes they do not work on such files.
>>
>>Generally ff/rwd even up to 20x works fine, but multiple back and forth 
>>attempts will lock up the frontend. I have heard the backend continuing 
>>to record the scheduled program on a couple of ocassions, and I have 
>>just done nothing until that program ended and the recording was fine., 
>>but it took a 'big-red-switch' recovery.
>>
>>I suspect that this is actually an X system error, probably somewhere in 
>>the openchrome driver. Since it only rarely allows any recovery I have 
>>not seen error messages very often but I have seen some 'texture' or 
>>'surface' references in error output.
>>
>>I will have to set up the laptop to ssh in and see if I can induce some 
>>crashes...
>>
> 
> 
> As I have mentioned, only graphical system goes down on the crash. I was always able to 
> log in remotely. Backend is working fine. However, scanning through logs I have not
> find anything that mentioned a problem.
> 
> The only thing that might be slightly related are following lines from the Xorg.0.log
> 
> Fulfilled via DRI at 27316064
> Freed 0 (pool 0)
> Fulfilled via DRI at 27740032
> viaWaitVideoCommandFire: Timeout. 
> 
> These are the last lines in the log after the freeze.

Yes, the backend continues to operate. This is clearly a display level 
freeze and almost surely X related only.
I have not looked at the xorg log output, but I will.

You might try running mythcommflag --rebuild on a file, then try using 
ff/rwd to see if it makes a difference.

Geoff