2005-11-01 21:00:30

by Nick Warne

[permalink] [raw]
Subject: Trail/trace /dev/agpgart access

Hi all,

Sorry for the noise, but need experts :-)

This is most probably nothing to do with the kernel, but I am at a total lose
and have exhausted all avenues after 5 days of finally getting around to
looking into this to find out WHY I am the _only person in the world_ that
seems to get this issue - two accesses to /dev/agpgart on startx - or X raw
(or any other combination of X)

Please refer to this original mail:

http://lkml.org/lkml/2005/6/22/295

2.6.14 - agpgart and via-agp built in - use nVidia driver (I know, I know) on
Geforce4 MX440 SE.

I have traced startx->xinit->startkde to no avail.

I have been on to the xorg people in irc - no-one there has ever seen similar
(get on to nVidia)...

I coded in a pci AGP device counter in generic.c, and it does report that it
gets accessed twice (counter resets after leaving that routine):

agpgart: Found an AGP 2.0 compliant device at 0000:00:00.0.
agpgart: Putting AGP V2 device at 0000:00:00.0 into 4x mode - count 1
agpgart: Putting AGP V2 device at 0000:01:00.0 into 4x mode - count 2
agpgart: Found an AGP 2.0 compliant device at 0000:00:00.0.
agpgart: Putting AGP V2 device at 0000:00:00.0 into 4x mode - count 1
agpgart: Putting AGP V2 device at 0000:01:00.0 into 4x mode - count 2

xorg logs here -> http://www.nick.ukfsn.org/other/Xorg.0.log

So, is there anyway I can 'watch' /dev/agpgart to see what does this?

Many thanks,

Nick
--
http://sourceforge.net/projects/quake2plus

"Person who say it cannot be done should not interrupt person doing it."
-Chinese Proverb