I have been trying to add a third 1920x1200 monitor to my workstation using an additional 9500gt graphics card in the extra graphics slot on my dp45sg motherboard. I have tried many different configurations using twinview, xinerama and just multiple xservers. I all cases, whenever the 2nd gpu is active in the xorg.conf, after gdm comes up and I sign in, gnome launches but Xorg crashes after the mouse is moved to any boarder or if the mouse is clicked. Two monitors on the same gpu have no issue but 1 monitor on each gpu causes a crash. I am running xorg-server-1.7.4 in a completely up to date ~amd64 system. This bug, while similar to some of the xinerama issues that have been reported with xorg-server-1.7.4 is different because it happens with or without xinerama active. Reproducible: Always
Please attach the output of nvidia-bug-report.sh (or whatever the script is call, I can never remember its name). Thanks
As I explained in the bug report X locks up requiring a reboot as soon as the mouse is moved. This makes it impossible to run anything under the configuration in question. I could run the report before X starts or under a configuration that doesn't use the display attached to the GPU but I don't believe that would tell you anything. I have watched Xorg.0.log while X was starting from another computer remotely logged in, but it didn't show anything consistent. Most times it showed no error at all. Other times it showed a few lines of alternating NVIDIA(2) Wait, 1 ... NVIDIA(2) Wait, 2 after signing on but before bringing up the gnome desktop. A few times Xorg seqfaulted with a backtrace. The variation of logs leads me to believe that the screen mapping is broken and memory is being corrupted. I have worked around the problem by buying a Matrox TripleHead2Go and running all 3 displays off 1 gpu. The other gpu is still installed so I can test during off hours if you need any other info.
nvidia-bug-report.sh does not require the X server to be running. Its explicitly meant for debug purposes so you can and should still run it and attach that info here and forward it on to linux-bugs@nvidia.com If you can not get it to work I would send an e-mail to linux-bugs@nvidia.com and include as many bits of information as possible. A nicer way than attaching a dozen files to the bug report would be to attach them here and include a reference to this bug. Items that jump out at me that are missing from this would be: 1) first GPU 2) exact make/model of motherboard 3) any and all overclock settings 4) details whether different versions of xorg-server work 5) exact version of nvidia-drivers package 6) your xorg.conf 7) your Xorg.0.log 8) copies of the backtraces you have received 9) exactly what kernel you used 10) your .config for that kernel 11) did any other kernels work? 12) dmesg when loading the nvidia.ko Basically, need tons of info. If there's any information you've gathered but I haven't listed it, attach it. Even if you think its useless.
This should have been marked as UPSTREAM.
