Strange X artifacts and possible bug using NVS-400

[chrispayne]

I'm seeing some strange response from an NVidia Quadro NVS-400 (PNY) running on Intel based Motherboards. We run Scientific Linux (https://www.scientificlinux.org/) on our machines, a free version based on RedHat Enterprise Linux. (It's going to be obvious but...) I'm not sure where the problem is, driver, kernel, bios..... but here is what I have.

I have seen the symptoms on several Intel motherboards, 865, 915, 945 chipset, with several different video cards, so I don't think it's bad hardware. I have updated the motherboard BIOS, as well as adjusted many settings therein as part of troubleshooting this problem. The CPUs are Pentium 4 with Hyper-threading, which is why I even fell into using the SMP kernel (if you are not aware, Hyper-threading makes the CPU look like an SMP system to Linux).

Also, I am using the latest (8178) version of the driver.

For SL3 (== RHEL3) using a 2.4.21-37.EL kernel :

If I use a NON-SMP kernel, the second GPU on the NVS-400 is not initialized, it does not get an interrupt assigned, and it is unusable (no matter what I do in the bios, kernel boot params, etc).

If I use the SMP kernel, (and recompile the NVidia driver of course) both GPUs become usable. I would stop here, but I have reports from one user (although no actual data or logs) of the X display becoming corrupted, "little pieces of windows left on other windows" kind of things. Shortly after these effects appear, the X server will crash with this error in the XFree86.0.log:

Fatal server error:
Freeing resource id=40E706C2 which isn't there

I suspect this may be due to using an SMP kernel, but that is more of a gut feeling due to previous SMP driver problems (raid).

For SL4.2 (== RHEL4) using a 2.6.9-22.0.2.EL kernel.

Both the NON-SMP kernel and the SMP kernel work fine for both GPUs.

I guess if someone can reassure me that the strange X response seen in the SL3 systems noted above will not reoccur under SL42, I'd be a happier camper as I upgrade the half dozen machines with these cards. Other changes to try to make the SL3 systems happy are also welcome. I have been holding off upgrading the OS for other reasons, but will upgrade if required to resolve the issue.

Attached are the 4 nvidia-bug-report.log files from the 4 scenarios above.

PS. I'm also trying to get multiple NVS-400s working in one computer, but that is probably another issue and probably another thread in a bit.

Cheers
Chris

[netllama]

Was the 305_smp bug report generated while the corruption was present? I'm asking because I don't see anything unusual.

Is there anything specific that triggers the corruption?

Thanks,
Lonni

[chrispayne]

I have not been able to get at the system when the strange behavior is happening, they are supposed to let me know the next time, during the wierdness.

From their description, it seems to start occuring after some amount of time, and moderately heavy usage (lots of X windows open, some open for days etc).

I will try to get a bug report when it is in a bad state, for now, the above is all I've got. These crashes are extremely inconvneint for these guys, which is why I'm fishing for information without the bad state bug report file.

Cheers
Chris