Catalyst/fglrx startx freeze - blank/black screen

Question

Sometimes, usually the second or third time I use startx after a reboot, video is not displaying.  I get a black screen (blank) with no cursor and no windows of any sort.  Ctrl+Alt+F[1-12] does not work at all.  The system is unresponsive.  I'm not sure whether the system is locking up or if video is just not displaying.
My graphics card is a "Mobility Radeon 9600 M10"
The Xorg doesn't appear to have any errors either.  Here's the log from when Xorg failed:
http://www.lastholdout.com/foo/xorg/Xorg.fail
Also, here's my xorg.conf file.  I generated it using 'sudo aticonfig --initial':
http://www.lastholdout.com/foo/xorg/xorg.conf
Notice that I added:
   Option      "UseInternalAGPGART"         "no"
   Option      "KernelModuleParm"           "agplock=0"
around line 135 for good measure since Arch's Wiki says that those options may help prevent lockups.
I've had this problem at least since 2.6.22.  Back in December, I switched back to the open source ATI driver.  As of late, system performance has been a real drag, and other than this strange behavior, catalyst seems to do a better job.  Is anyone else having this problem?
Last edited by Infracephas (2008-02-17 19:52:53)

Answer 1

Sorry,
I'd pretty much given up on this - I know it sounds strange, but this only affects me at home, when I turn on the laptop at work, it works perfectly every time.  If I try using the work network profile at home it sometimes randomly works.
I did a bit more digging into this, and running a strace produced lots of ioctl messages:
15692 open("/dev/dri/card0", O_RDWR)    = 10
15692 write(0, "drmOpenDevice: open result is 10,"..., 39) = 39
15692 write(0, "drmOpenByBusid: drmOpenMinor retu"..., 40) = 40
15692 ioctl(10, 0xc03064a6, 0x7fffc8705010) = 0
15692 ioctl(10, 0xc03064a6, 0x7fffc8705010) = 0
15692 ioctl(10, 0xc03064a6, 0x7fffc8705010) = 0
15692 ioctl(10, 0xc03064a6, 0x7fffc8705010) = 0
15692 ioctl(10, 0xc03064a6, 0x7fffc8705010) = 0
(about 2,000,000 or so before I managed to kill X)
i don't think this is a wireless issue any more, it is more than likely a fglrx problem, but I doubt we will get any luck trying to get support out of ATI, in the meantime I will just end up constantly rebooting my laptop until it decides to work.
BTW I have also seen this happen under Mandriva when network profiles were used, but never got a chance to dig into it very much.
As I type this I am looking at it a bit more closely and I can see that in the same strace I also get:
15692 open("/dev/dri/card0", O_RDWR)    = 11
15692 write(0, "drmOpenDevice: open result is 11,"..., 39) = 39
15692 write(0, "drmOpenByBusid: drmOpenMinor retu"..., 40) = 40
which makes me think that X has opened the DRI device twice and there may be some kind of race condition going on here, at which point I am in way too deep and if there is anyone here with clue, please help.
Cheers
Jammib