Created attachment 25407 [details] Xorg.conf For Nouvea EXA Hello, I would love to be able to use the nouveau driver on my hp tx2000z laptop. I can do so today, but must explicitly configure the driver to use shadowfb. If I allow the driver to attempt to run using EXA one of two things happen: 1) I see a horribly corrupted X display 2) X fails to start, giving me the message "Fatal server error: Detected GPU lockup" I am currently running Fedora 10, with version 1.5.3 of the X-server and fairly recent GIT builds of the nouveau driver and drm. (I tried these thinking that perhaps the standard F10 packages were not new enough). Currently running 2.6.28.7 SMP 64-bit kernel, but I've tried several other variants also. The machine is an HP Pavillion tx2000z laptop with an NV4E chipset (Geforce 6150 Go). 4GB of RAM, and an AMD Turion processor. I am running the x86_64 version of Fedora 10. As a longshot, I also downloaded a Fedora 11 Preview live CD, after hearing of their successful nouveau test day. I hoped the newer revision of the X server included there would help. Unfortunately, F11 also gives me the GPU lockup message. I am attaching X.org.log files of two startups of X (one with corruption, and one with the gpu lockup message) and my xorg.conf file. If I can provide you with any other useful information (or if I've somehow missed a very obvious configuration option), please let me know! Thanks, Jeremy
Created attachment 25408 [details] Xorg log with corrupt display
Created attachment 25409 [details] Xorg log with GPU lockup
Reading around, something stuck out about having more 2GB of memory in a system being problematic. So, I had a few minutes last night to power down my laptop, remove one 2GB DIMM, and start back up. EXA works in this configuration. Turning on desktop effects with Xrender under KDE is PAINFULLY slow--slower than running the nouveau driver with shadowfb--but I no longer get the error messages or screen corruption I previously reported. Is this behavior likely to change? I reinstalled the second DIMM and let the memtest application run all night. The memory checks out just fine. Obviously, the built-in video adapter does not have dedicated VRAM, but why would having 4GB of system memory cause the accelerated driver to fail?
Created attachment 32337 [details] Xorg.log from F12 still shows lockup problem Hello All, I decided it was time to upgrade to Fedora 12, and again would really like to use nouveau so that I don't have to do the nvidia compile (and maybe download) dance whenever I update my OS kernel... and, because I hear such good things about nouveau. The nvidia driver worked for me under F10, and if I must, I'll go back to it, but that just seems so WRONG! Anyway, 7 months since my last attempt and I'm still getting the GPU lockup error, both from X and from the KMS driver. Right now, I've disabled kernel modesetting, and I'm using nouveau with the shadowfb. Please help! I'll gather whatever info you might need to help diagnose/fix the bug, but here's where I'm at now: Still an HP tx2000z laptop with an NV4E... here's what lspci -vnn says about it: 00:05.0 VGA compatible controller [0300]: nVidia Corporation C51 [Geforce 6150 Go] [10de:0244] (rev a2) (prog-if 00 [VGA controller]) Subsystem: Hewlett-Packard Company Device [103c:30e5] Flags: bus master, 66MHz, fast devsel, latency 0, IRQ 18 Memory at c2000000 (32-bit, non-prefetchable) [size=16M] Memory at d0000000 (64-bit, prefetchable) [size=256M] Memory at c1000000 (64-bit, non-prefetchable) [size=16M] [virtual] Expansion ROM at c0020000 [disabled] [size=128K] Capabilities: [48] Power Management version 2 Capabilities: [50] MSI: Enable- Count=1/1 Maskable- 64bit+ Kernel modules: nouveau, nvidiafb I'm running F12 x86_64 with kernel kernel-2.6.31.9-174.fc12 and nouveau rpm: xorg-x11-drv-nouveau-0.0.15-18.20091105gite1c2efd.fc12. I've double-checked, and though it's listed as available, I'm not loading the nvidiafb module. Please let me know what information will be helpful to you in diagnosing the issue, and I will try to provide it ASAP. Thank you! Jeremy
I believe it would be beneficial to 1. install the latest git version of DRM modules, libdrm and DDX (the X driver) 2. use the minimalistic xorg.conf from http://nouveau.freedesktop.org/wiki/InstallNouveau 3. Use KMS (enabled by default nowadays) 4. provide a full kernel log from boot up to the GPU lockup, and X log Step 1 probably requires you to upgrade your kernel to at least 2.6.32 for the DRM modules to build. It seems you have two nvidia devices, according to your X log. The kernel log should give more information if that is a problem. Related to this, IIRC the latest DRM code contains changes in multi-device support, although it is not complete. If this is too much, you should at least provide the kernel log.
Created attachment 32347 [details] Xorg.0.log - EQ overflowing message
Created attachment 32348 [details] dmesg from boot of 2.6.32.2
Created attachment 32349 [details] /var/log/messages with lots of kernel complaints
Created attachment 32359 [details] New Xorg log
Created attachment 32360 [details] dmesg from boot of 2.6.32.2
Created attachment 32361 [details] /var/log/messages
Hi Pekka, Thanks so much for responding. I've done as you asked: 1) Booted with kernel 2.6.32.2 2) replaced xorg.conf with the minimal version you referenced 3) checked out and compiled libdrm from git 4) downloaded nightly nouveau drm mods and compiled 5) checked out and compiled the nouveau ddx from git At first, I got some nasty results (the first three new attachments) but that's because I forgot to actually -install- the kernel nouveau drm mods... After fixing that, I produced the latest three logfiles. With KMS enabled, everything works perfectly--for one X login. With KDM running, logging out gets you a blank screen and there doesn't seem to be a way to restore the local display. Restarting from ssh, and trying again gets the same result, a single good login and then a blank screen. If I disable KMS, however with nomodeset, (and for now, just trying things from init 3), I can start and stop X as many times as I like. I do notice that on line 636 of the dmesg from a KMS-enabled boot, we have this message: [drm] nouveau 0000:00:05.0: GPU lockup - switching to software fbcon Suggestions on where to go from here would be welcome! Thanks again, Jeremy
Ok, I've had a couple of days to play with things... definitely not there yet. I still have weirdness when I end an X-session and try to start a new one. Sometimes it works, and sometimes it doesn't. I'll try to produce some logs with useful info... sometimes I see a GPU lockup message, sometimes not. Also xv is not working properly. For example, xine with xshm looks fine, but play the same video with xv and there's lots of flickering and funky coloring. Hopefully useful logs will follow. If anyone can tell me what will produce the most useful info for diagnosis, I will work to supply it. Jeremy
An update for anyone interested... probably my last on this ticket, but I seem to have things mostly working. The snapshots I loaded on from 2009-12-29 may have been ok, but I don't think I had the drm modules loading properly from the init ram disk. So, I tried again with the drm mod snapshot from 15 January and also compiled the mesa libdrm and xorg nouveau drivers from about that time. I've been running with them since then and things seem generally stable on my laptop. The only trouble I have seen so far is when I try to turn on Xrender in KDE, but I'll save that for debugging later. As of now, however, I can report that kms, exa, and xvideo seem to be working generally ok on my hp tx2000. So, should anyone be tracking things on this almost year-old bug still marked as "NEW," please close it out if you wish. If I have new bugs to report, I'll open a new one.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.