[pxc] A couple issues.

[Alessandro Rubini]

Hello

> 	dma.c:348: too few arguments to function `remap_page_range'

This is a redhat-specific issue. While looking in the 2.4 patches (and
not finding the extra argument anywhere) I found sis_main.c that
explicitly mentions it.

But this is an issue with 2.6 too, so I'll need to add a hack there.
Meanwhile, I'd better not support the redhat variant. If you want
to suggest a not for the manual, I'll gladly apply your patch.

> After configuring my kernel's memory usage to open 5MB of memory and
> loading the module, I tried to run pxc_xgrab.  The produced video
> image is only horizontal lines of gibberish.

According to the pattern you get, you can see whether DMA takes place
or not. If you try "cat /dev/pxc0pgm | od -t x1" you can see that. If
no DMA happens, you'll see a repeating test string (as defined
in allocator.c).

> 1.) Verified IRQ correlation between /proc/pci and /proc/interrupts.
> Initially, the PXC200 was sharing an IRQ with the USB bus, and
> didn't show up in /proc/interrupts.

It should show up only then the device is in use (for example, try
sleep 10000 < /dev/pxc0). And sharing should not be a problem (i.e.,
I think it's supported fine).

> I tried moving it to a different PCI slot and it then showed up on
> its own interrupt in /proc/interrupts.  The PXC200 now shows up in
> /proc/interrupts, and has the same IRQ in each file.

Ok
 
> 2.) Tried a different PXC200.

They usually work. But good you tried it; last year I got a bugged
board and lost a few months trying to find the bug in my driver,
before I finally got my hands on another identical card and found the
bug was in the hardware (then I had to wait another few months before
the italian dealer got it replaced).

> 3.) Tried using pxc_control to change the video format, and the mux.

When the video format is wrong, you see the image moving on the
screen but line length is wrong. About the mux, you can use the "More"
button in pxc_live.

> 4.) Tried it with and without the camera connected.
> 
> 5.) Compiled, installed, and susequently uninstalled pacco.

Not an issue, really.

I'm interested in whether interrupts flow in or not. Is the counter 
in /proc/interrupts increasing? (yes, avoiding a shared IRQ line is good
to this aim) -- but I suppose it does, or the read() system call will
block.

Another idea is compiling with DEBUG=y in the environment (or "make"
command line), and seeing in the kernel messages if something interesting
shows up.

/alessandro