memory leak - 4191

[Hairball]

I'm developing an opengl application in RH8.0 with nvidia's linux drivers (v4191). Actually, the application uses VESS (http://vess.ist.ucf.edu) which is built on top of Open Scene Graph (OSG) or SGI's performer which use OpenGL.

Anyways, under performer and OSG, after running the application for a while, there's a huge memory leak. I developed my own memory debugging code (which grabs the malloc hooks inside of glibc) and my output tells me that the biggest leak is in nvidia's /usr/lib/libGL.so.1 at the symbol __nvsym17216. Unfortunately, I can't get a stack backtrace to find out what in my application is calling that. Any idea how I can debug this further? Is this an issue with nvidia's drivers?

I tried running my application in an older version of nvidia's drivers (2xxx something, can't tell right now because that computer is being used for something else at the moment) and the problem was not nearly as bad.

Any suggestions would be very much appreciated!

[Hairball]

One more thing, when I set a breakpoint (in gdb) on an offending symbol in /usr/lib/libGL.so.1 (like __nv17216 or __nv17217), I can't get a backtrace in gdb. info stack only goes back like one or two levels (other __nv.* symbols) and then gives some error message like "Cannot access memory at 0x2". Normally, I would suspect stack corruption, but everything works fine except for this memory leak.

[kappax]

well, i am not to sure if it is in the nvidia drivers, I was told that the older kernels have some mem leak in agpgart or someing (some where, because i was getting nasty fps after a while and what not ) to fix it i upgraded to a 2.4.20 kernel, if you have a stock kernel you should have a 2.4.18 kernel or so so i would try that.

[Noth]

I've had my box up with the 4191s for >45 days now running a few OGL apps, pretty much almost daily, and I havn't seen any memory leak unless it's really small and hasn't hurt me yet, I have 1.2G memory so it might take a while for me to notice =)

[spreelanka]

hairball, it's not just you. While i haven't followed it as much as you have, I'm pretty sure that there is a leak somewhere in the nvidia drivers. I loose ~250 of my 512mb of ram. While i didn't, a friend of mine tried using old drivers as well, with similar results. Can't help you much right now, just letting you know that you're right.
i'll post here if i find anything out.

[jblouin]

Hello, I sometimes get my opengl application to hang on some _nvsym*** and also I can't get a stack back to my application code which make debugging very difficult.

I wonder did you ever find a solution to your problem?

Jean