#472 Radeon 7500 QW hangs on logout of KDE/GNOME
When the Xserver is reset from the logout of KDE or
GNOME, the system blackscreens with no keyboard a
cess at all requiring a hard reset of the system.
I'm using XFree86 CVS 4.3.0-20030218 built by
Mandrake for the 9.1 Cooker distribution.
As near as I can guess (since no info is found in either
my XFree86.0.log nor dmesg) I think that when the
logout occures it shuts X11 down, and then restarts it to
start KDM. However, when it shuts the card down, the
system locks due to the drm kernel module getting in a
race condition with the PCI bus. However I could be
barking up the wrong tree here...
My systems specs are:
AMD Athlon XP1700+
256 MB RAM
ASUS A7V266-E Motherboard
ATI All In Wonder Radeon 7500
Kernel: 2.4.21
glibc: 3.2.1
GCC: 3.2.2
XFree86: XFree86-4.3.0-20030218-1mdk.i586.rpm
Distribution: Mandrake Linux Cooker 9.1
Logged In: YES
user_id=762247
No, I am getting the same thing as well. I happen to be
running the ATI Radeon 7500 (Not the All in Wonder version)
PCI Bus model, and get the same thing when I try to
shutdown...Then again, Linux regonizes all QW boards as the
AGP model...maybe it is an error caused when you force it to
run in PCI mode...
Logged In: YES
user_id=212039
I'm getting the same thing on a Radeon 700 QY AGP, and I'm not
forcing it to run in PCI mode. This happens using 2.5.x kernels or the
latest radeon's drm module. Using 2.4.x kernels dri/drm doesn't work
at all, libgl gives: "initDriver failed" or something similar.
Logged In: YES
user_id=800592
Ok, I seem to have the same problem on my IBM ThinkPad R40
with ATI Mobility Radeon M7 LW (which is a Radeon 7500):
The first time I start X (and kdm) after booting, everyting
is fine. I can log into KDE and graphics run smoothly - I
can even play Quake3 without any problems.
But when I log out of KDE and X gets restarted I get a
"blank" screen that, beginning in the corners, slowly turns
a whitisch yello exhibiting a strange pattern. So clearly X
fails to restart correctly.
The same happens if I kill X by pressing CRTL-ALT-BACKSPACE
or by killing the X process. It does not matter if I logged
into KDE after booting - it's plainly the restarting of X
that fails.
Apart from that the machine keeps running nicely, all the
text consoles can be accessed. I just can't get X do display
anything useful without first rebooting the machine.
No error messages can be found in either syslog or XFree.0.log.
But when I compared the XFree.0.log from the first
(successful) start after boot with the second run (restart
of X), I noticed this:
-----snip from diff-----
491a492
> (WW) RADEON(0): Bad V_BIOS checksum
499,500c500,501
< (II) RADEON(0): Panel ID string: 1024x768
< (II) RADEON(0): Panel Size from BIOS: 1024x768
---
> (II) RADEON(0): Panel ID string:
> (II) RADEON(0): Panel Size from BIOS: 65535x65535
-----end of snip-----
Apparently the VIDEO-BIOS got messed up somehow during the
restart of X. Not good. The Panel ID and size information
are messed up as well and XFree86 can't find AGP now:
-----snip-----
604d599
< (II) RADEON(0): [drm] loaded kernel module for "radeon" driver
610,626c605,607
< (II) RADEON(0): [agp] Mode 0x1f000201 [AGP 0x8086/0x3340;
Card 0x1002/0x4c57]
[...]
---
> (WW) RADEON(0): [agp] AGP not available
> (II) RADEON(0): [drm] removed 1 reserved context for kernel
> (II) RADEON(0): [drm] unmapping 8192 bytes of SAREA
0x629d1000 at 0x15573000
-----snip-----
I can "dumb-fix" the entire problem by not loading the "DRI"
module, which lead me to the conclusion that this very
module should at least be somewhat involved in causing the
failure. But without DRI there's no 3D support... :-(
Can any of the developers speculate on the origin of this
problem?
I'll be glad to provide more detailed information if
requested. :-)
Cheers,
Enno (knutbert)
Hardware: IBM ThinkPad R40 (2722-3GG), Pentium-M (Centrino),
Intel 82801DB / 82855PM Chipset, ATI Radeon Mobility M7 LW
(7500 LW)
Software: SuSE Linux 8.2, kernel 2.4.20-SuSE, GCC 3.3,
XFree86 4.3.0, DRI radeon-20030613-linux
Logged In: YES
user_id=800592
Good news! DRI is not to blame for my problem!
After poking around for a while I found out that indeed
AGPGART was the culprit:
When I started X with no "Extensions" at all loaded and then
executed the "testgart" program, I could provoke that
strange phenomenon I described in my last post. So there was
the problem without DRI being involved.
I really should have gotten suspicios when AGPGART said
"agpgart: Trying generic Intel routines for device id: 3340"
Well, just for the record, here's the solution:
Just patch the kernel, so agpgart recognizes and correctly
initializes the Intel i855PM "Centrino" (82855PM) AGP
Controller.
Bill Nottingham provided a patch against 2.4.20 on the LKML:
http://lkml.org/archive/2003/4/23/193/index.html
Happy patching!
And sorry about the inconvenience...
Cheers,
Enno (knutbert)
Logged In: YES
user_id=423974
Are any of the people who have seen this bug been able to
reproduce it with a recent DRI build?