From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.5) Gecko/20011013 Description of problem: X-Windows locks up under concurrent OpenGL (Mesa) and I/O on Athlon based motherboards with VIA and AMD chipsets and AGP enabled. This problem does NOT occur on Pentium III and Pentium IV based systems with Intel chipsets. Nor, if AGP is disabled. Using either OpenGL sample applications (morph3d, teapot), a solar system simulation app called ssystem, or our own application IMOD along with concurrent I/O causes all of our Athlon based systems to hang within a couple of hours. We can usually ssh into the hung machine from another machine and kill the application to bring back X. Top shows that the app is either using all of CPU cycles or splitting it evenly with X. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: Reproduction Procedure: Run an OpenGL app such ssystem http://linux.tucows.com/home/preview/9981.html or our IMOD ftp://bio3d.colorado.edu/pub/graphicstest/obj7.mod.gz . To run ssystem: tar -xzvf solartest.tar.gz cd solartest/ssystem ./ssystem resize thewindow greater than 800x800 To run IMOD: gunzip obj7.mod.gz imodv obj7.mod You should get a 512x512 window with a model displayed in it. Type 5 then 6 on the numeric keypad to start the model spinning. A simple way to do persistent concurrent I/O: Create a file called bigfile larger than the RAM size of the computer under tcsh while (1) \cp bigfile bigfile1 end Actual Results: X will appear to hang within 3 hours on all of our machines. Sometimes the display will hang until the next mouse or keyboard event. This is a bad sign and is usually a precursor to a full hang-up. Expected Results: Machine should have continued running with X responding normally Additional info: We have tried many hardware combinations that include both VIA and AMD chipsets, NVIdia and Matrox video cards, below is a list of motherboards, video cards and RedHat version for which we have observed and can reproduce this behavior: CPUs: Athlon 1.0 -1.4 GHz Athlon MP 1800+ Motherboards: IWILL KK-266 ABIT KT-7/KT-7a/KT-7A-Raid ABIT KG-7-RAID Tyan Tiger MP (S2460) Video cards: NVidia Geforce2 MX, Ultra NVidia Geforce3 and Geforce3 Titanium 500 Nvidia drivers: 1.0-1512 and 1.0-1541 Matrox G400 with 1.3 drivers and XFree86 4.0.3 drivers OS Versions: Red Hat 7.1 Red Hat 7.1 with kernel upgrades from redhat.com Red Hat Roswell Red Hat Roswell with ALL upgrades from Red Hat Network Red Hat 7.2 SUSE 7.1 Xfree86 versions 4.0.3 and 4.1.0 Typical system config: 40 GB IDE (ATA-100) hard disk Generic CD-ROM Generic network card (100BT) 1-1.5 GB RAM 400 watt or greater power supply
Red Hat Linux does not support 3D acceleration on any Nvidia hardware. Red Hat does not support Nvidia's binary only drivers, you'll have to contact Nvidia for any bugs encountered while using their drivers unless the problem can be reproduced without loading any non Red Hat supplied kernel modules with the stock supplied XFree86. The G400 should work with the supplied drivers. Usage of Matrox's website drivers and/or binary only hallib is not supported. If you are having a problem with the G400 using the Red Hat supplied XFree86 drivers, please attach your XFree86 config file, and your XFree86 server log using the link below, and I will try to reproduce. Also, please include your /var/log/messages.
It is possible the problem you are experiencing is the recently discovered Athlon PSE/AGP bug, which is discussed on the linux-kernel mailing list. I suggest you try booting your machine with the "nopentium" option which disables the feature in question affected by this hardware design flaw. Also, if you are using the Athlon kernel, please try to use the i386 kernel as well, as there are some other problems with Athlon systems and VIA motherboards. If you still encounter these problems, be sure you are using our most recent kernel update for the distro. Again, Nvidia binary drivers are not supported, so please test the problem on a supported system. Please reopen the bug report if the problem persists. Thanks.