On Mon, Jan 16, 2012 at 01:45:24PM -0500, Andy Walls wrote: > On Mon, 2012-01-16 at 11:55 -0600, David Engel wrote: > > Hi, > > > > I have a MythTV backend with a Hauppauge HVR-2250 (dual tuner, > > ATSC?QAM, PCIe) and a Ceton InfiniTV4 (quad tuner, QAM/cable card, > > PCIe). Over the weekend, I started intermittently seeing corrupt > > recordings that were painful to watch. > > > > I eventually narrowed the problem down to when both tuners on the 2250 > > are active at the same time. In this case, both recordings have > > corruption (CRC errors, etc.). The InfiniTV4 does not appear to be > > affected by anything going on on the 2250. Likewise, the 2250 does > > not appear to be affected by anything going on on the InfiniTV4. > > > > I noticed something strange while diagnosinig the problem. When the > > 2250 is busy recording, top reports the CPU as being in wait for an > > abnormally high amount of time (~30% for one tuner busy and ~50% for > > both tuners busy). I don't recall seeing that before. I quickly > > tried a KWorld ATSC 110 on a different system and it showed no, or > > negligible wait time. > > > > Thinking that the 2250 was going bad, I replaced it with two KWorld > > ATSC 110s (single tuner, ATSC/QAM, PCI). The two 110s had the same > > problem as the 2250 -- corruption when both tuners are busy and > > unusually high wait time when either is busy. > > > > At this point, I'm suspecting a motherboard, memory or grounding > > issue, but would like some feedback in case there's anything I'm > > missing. The high wait time seems extremely odd to me. Perhaps it > > means something to those of you who are much more familiar with the > > cards and drivers. > > A long time in IO Wait and (presumably) dropped video data buffers for a > TV card driver usually means: > > a. The devices are not producing interrupts > b. Your system is dropping interrupts from the devices > c. Something in your system keeps interrupt processing disabled for a > long time. I tried the crude test of monitoring /proc/interrupts while recording. Using a third KWorld ATSC 110 on my development system, which isn't showing any problems, the driver serviced ~1200 interrupts per minute. I saw the same amount per card using two other 110s on my MythTV system, which is exhibiting the problem. > Of the above c. is unlikely. > > The root causes of a. and b. are many and varied. However most of those > root causes are under control of kernel software: core IRQ handling, > core PCI/e bus setup, or tv card driver modules and their supporting > modules. > > Hardware related root causes would be buggy PCI chipsets, poor RF > signal, a motherboard problem, or a power problem. > > I tend to blame the software before the hardware. > > > Sometimes, dust in the shared, conventional PCI bus can cause problems. > If you have conventional PCI slots in the system - whether you use them > or not - unplug the conventional PCI cards, blow the dust out of all > the slots, and reseat the cards. You should not need to do this for any > PCIe card slots, but it doesn't hurt. I also tried the HVR-2250 in my development system this evening. It worked fine. It's definitely something specific to my MythTV system. I can't do any further testing until Wednesday. Thoroughly blowing out the case might have to wait until the weekend. Thanks for you help. David > Regards, > Andy > > > Oh, the problem appear shortly after switching to the 3.1.9 kernel. I > > also tried the 3.1.8 and 3.0.14 kernels to rule out software and there > > was no effect on the problem. > > > David > > > -- > To unsubscribe from this list: send the line "unsubscribe linux-media" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html -- David Engel david@xxxxxxxxxx -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
