ATI 9.2 Linux Driver REBOOTS system at xdm start

[drankinatty]

Listmates:

I really have to ask WTF is up with ATI and its linux driver? I have installed every driver since October 2007, liven through the horrible days until a reasonably stable 8-4 driver was released, enjoyed all the drivers through 8-9, an then I have been absolutely flabbergasted and disappointed with every release since 8-9. On both my Toshiba laptops (old P35D, and a newer P205D). Every driver since the 8-9 driver has shown a 40-50% reduction in glxgears frame rates and has made my boxes notably slower.

Both are running openSuSE 11.0 with KDE 3.5.

Reverting to the 8-9 driver restores performance and all is well. Today I got around to installing the 9.2 driver (why the hell ATI used a '.' as a separator instead of a '-' this time -- who knows). I installed the driver the same as I always have. Build the package from the installer with --buildpkg, drop to runlevel 3, remove the old driver, prep the kernel with:

make mrproper
make cloneconfig > /dev/null 2>&1
make modules_prepare
make clean

Install the new fglrx rpm, reboot and done. (yes I save my xorg.conf). To my utter dismay, when I boot the system and it gets to the point where X is started, my box REBOOTS -- WTF? I wiped the system clean and tried again -- same results. I wiped the system again and reinstalled the 8-9 driver, and everthing is fine.

I have filed no less than 5 LinuxDFeedback tickets detailing the problems to ATI along with glxgears frame rates, amdpcsdb, dmesg, xorg.conf and Xorg.0.log and have got nothing but a slap in the face by ATI/AMD with one ATI tech idiot closing my feedback ticket saying that there was no support for Linux and that I should have submitted my report through the LinuxDFeedback mechanism ("Hello dipshit", that's what I did in the first place....)

So having been forced to revert to the 8-9 driver once again (which isn't an option for anyone running openSuSE 11.1) I have to ask, "What is going on with ATI?" What gives with the 4 releases of drivers with declining performance? Did the code for the 2XXX, 3XXX and 4XXX series of cards have cause unintended negative consequences for the older Radeon cards?

Also, what is the best way to get ATI to respond and fix the problems that have been introduced in the Linux driver over the past 7 months? ATI has proven time and again that the Linux Driver Feedback is nothing but a joke or something immediately redirected to > /dev/numbnuts/

Anybody else got any inroads with ATI? In Nov. Dec. 2007 we had good response from them fixing a SONAME problem in the driver, but ever since then -- nothing.

Anybody else having a different experience?

[Wyatt]

I believe I've encountered this as well. Does your dmesg push to your syslog? Look in there for something like so:

BUG: using smp_processor_id() in preemptible [00000000] code: X/26646
caller is KCL_SPINLOCK_STATIC_Grab+0x35/0x60 [fglrx]

I've yet to find a solution that doesn't involve switching to a different driver or rolling back to an older kernel, but I believe there's a bug up on the tracker about this, at least. I didn't try, but building a non-preemptible kernel might work, too. That's sort of not an option on my end in this case, so I took the easy route.

Note that if you switch to a VT before your login manager loads fully (at least in my case), you can log in and kill X. At least from there, you can operate via SSH or screen.

[drankinatty]

I don't know if my logs push it, but what I did get in syslog was:

Mar 4 16:41:20 alchemy kernel: fglrx: Unknown symbol flush_tlb_page

This is really frustrating. Thank God I can still use the 8-9 driver or I would be looking for an Nvidia based laptop right about now..