I'm not expecting a solution to this problem as I've found a workaround and it does seem to apply to a very specific situation. I'm just raising the point here as a reference.
Just a small bit of history first. I've been trying to record the last prerecorded VHS tapes that I have with my current system (at the time running 5D1), so I could put them on DVD. I found on certain tapes the PVR350 would lock up, stopping the tape would sometimes restart the capture. I upgraded ivtv to the next version for that kernel, it still continued to do this. So putting those tapes to the side, I captured the rest of them and put them of the DVDs. Having a look at some of them, I noticed strange skips, where it looked like if someone had pressed "Pause" for second or so, looking at the recordings themselves they showed the same feature.
So I reached a conclusion that capturing from pre-recorded tapes using a PVR350 without some kind of video cleanup circuit is going to be unreliable.
So after a while I decide to set up one of my old boxes to capture using a WinTV Theater, so after some issues, it seemed to work OK, but there were capture quality issues. I tied these down to the system not being fast enough. So I transferred the WinTV card and PATA HDD over to my main workstation ending up with the following:
Athlon64 3700+ Asus A8N-SLI
1Gb PC 3200, 80Gb PATA HDD, 200GB SATA+250GB SATA (Not used),
DVD-RW GSA--4163B, Sony DVD-ROM, Gigabyte GV-NX66128DP
Hauppauage WinTV Theater, . KM R5E50
I needed to reinstall KM, without any problems, well, apart from me forgetting to plug in the link into the soundcard.
Test captures went OK, trying to get a good compromise for quality and space.
So on an actual capture, everything seemed to go OK until the end where it locked up solidly. Rebooting I found the recording was OK (I had the problem with the backend not restarting), so I carried out another capture and the same thing happened, except there was no sound on the recording.
So after further investigation I noticed in the logfiles that /dev/dsp1 had disappeared, so rebooting and going into live TV confirmed it was OK now.
I noticed that the system locked up when the tape reaches the unrecorded bit at the end, this seems to happen
every time this point is reached on whatever tape. After reboot, dev/dsp1 would disappear as well.
The work around is to keep an eye on the elapsed time and stop recording before this part of the tape is reached.
Bruce S.