On Sun, Aug 31, 2003 at 06:40:04PM +0100, Jon Burgess wrote: > Emard wrote: > > >This DMA corruption has been there long time around. > >We tried to fix it with many tweaks, that's the reason of > >vpeirq introduction but from your experience it seems some > >data loss hole still remained open. > > What was done before vpeirq? How far back in the code do I need to look > to find this?
Before March or so, the code is dirty and pure junk which easily can miss DMA at PCI latency. It can work well on some machines but PCI latency should be well tuned. The source or the problem was blamed on separate vsync detection chip found mostly on budget cards. This chip and 7146 can have slightly different timing of vsync and therefrom came the DMA blocks missing in the previous code. It could be that metzlerbro's are still using this type of code in their DVB drivers vpeirq() doesn't need to trigger on the vsync chip signal and is much better solution invented by a czech developer who had 3 budget cards downloading simultaneousy and had missed DMA interrupts which vpeirq() completely solved. I was using it on 2 budget cards simultaneous cards and worked fine for me too. Emard -- Info: To unsubscribe send a mail to [EMAIL PROTECTED] with "unsubscribe linux-dvb" as subject.
