[MPlayer-dev-eng] [PATCH] mga_vid and linux kernel 2.6

[Jan Knutar]

On Saturday 26 June 2004 17:03, Alex Beregszaszi wrote:
> Hi,
> 
> > I have found a more recent version a the patch discussed in this
> > thread:
> > 
http://mplayerhq.hu/pipermail/mplayer-dev-eng/2003-October/021261.html
> > 
> > The more recent version is here:
> > http://www.linuxops.net/~pw/mga_vid/mga_vid-2.6.patch
> Committed this ^^
> 

This mga_vid doesn't really work for me, and atleast one other person 
that I talked to in #mplayer.

Upon inserting the module, I get a "Bad Address" from modprobe. I've 
tried to track down this a bit, and it just made no sense...

With some printk debugging, I've discovered that 
pci_find_device(PCI_VENDOR_ID_MATROX, PCI_ANY_ID, dev) on line 1372 
always returns 0. This despite that PCI_VENDOR_ID_MATROX /is/ listed 
in /proc/pci.

When no cards are found, mga_vid eventually exits with return(-EFAULT) 
at line 1799, hence modprobe giving me "Bad Address".

I'm puzzled about pci_find_device not finding the matrox card... 
However, I have a theory.

Looking at /proc/interrupts, I see
11:   41152778          XT-PIC  mga at PCI:1:0:0, Ensoniq AudioPCI
And when not playing audio on the shared interrupt, the interrupt 
advances by about 75 IRQ's per second, the refreshrate of my monitor...

I'm guessing Xorg is camping on the graphics board with interrupts 
enabled. How xvidix:mga_vid.so works when /IT/ thinks it disables 
interrupts, I have no idea, and that the whole thing doesn't just blow 
up in a spectacular way, is impressive, and frightening.