Hi all,
We are having a problem running tksurfer on several Centos 5 machines. We are getting the dreaded GLXBadLargeRequest and just the front sliver of the brain displayed. By installing the Nvidia driver manually, we made it work.
We have one Centos 5 machine where it is working (machine01). That one machine has an Nvidia card and the Nvidia driver installed.
We find that machine01 runs tksurfer correctly locally, and other machines can connect to machine01 with VNC and run tksurfer remotely on machine01. None of the other machines work correctly.
Following up on the Nvidia driver difference, I notice that the Nvidia driver package includes a number of GL libraries. On the other machines, standard Centos 5 installs, I see:
$ ls -l /usr/lib*/libGL* lrwxrwxrwx 1 root root 10 Jan 7 14:57 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 12 Jan 7 14:57 /usr/lib64/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 492720 Nov 10 11:23 /usr/lib64/libGL.so.1.2 lrwxrwxrwx 1 root root 11 Jan 7 14:57 /usr/lib64/libGLU.so -> libGLU.so.1 lrwxrwxrwx 1 root root 20 Jan 7 14:57 /usr/lib64/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 526048 Nov 10 11:23 /usr/lib64/libGLU.so.1.3.060501 lrwxrwxrwx 1 root root 12 Apr 23 14:49 /usr/lib/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 432016 Nov 10 11:24 /usr/lib/libGL.so.1.2 lrwxrwxrwx 1 root root 20 Feb 3 16:07 /usr/lib/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 524000 Nov 10 11:24 /usr/lib/libGLU.so.1.3.060501
I see (based on the modification time of the install) that there are several new libraries installed from Nvidia. That's a new libGL and 10mb libGLcore, and my standard libGL.so.1.2 has been zapped.
$ ls -l /usr/lib*/lib* | grep 09:21 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib64/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 9777704 Apr 16 09:21 /usr/lib64/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 661 Apr 16 09:21 /usr/lib64/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 822960 Apr 16 09:21 /usr/lib64/libGL.so.100.14.11 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so -> libnvidia-cfg.so.1 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.1 -> libnvidia-cfg.so.100.14.11 -rwxr-xr-x 1 root root 127096 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 3016 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.100.14.11 -r--r--r-- 1 root root 185148 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.a lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA_dynamic.so.1 -> libXvMCNVIDIA.so.100.14.11 -rwxr-xr-x 1 root root 137944 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.so.100.14.11 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 10080488 Apr 16 09:21 /usr/lib/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 659 Apr 16 09:21 /usr/lib/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 608400 Apr 16 09:21 /usr/lib/libGL.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 2352 Apr 16 09:21 /usr/lib/libnvidia-tls.so.100.14.11
For a test, I temporarily copied the libraries and make the appropriate links on another machine - basically installing the driver manually. I find that now the entire brain displays and I don't get the GLXBadLargeRequest.
I suppose I could have installed an Nvidia card and installed the driver to accomplish the same thing. You can use the -x option to the NVidia driver package to just extract it without running the installer. Be sure to check the LICENSE file, sections 2.1.2 and 2.1.3, before deciding how to proceed with this information.
So the problem lies in functionality available in the Nvidia libGL (etc) libraries that is not available in the standard Centos 5 libGL. What GL/GLX libraries do I install on a non-Nvidia machine in order to run tksurfer? Does tksurfer depend specifically on having Nvidia cards or other proprietary drivers installed? The wiki states:
A 2GHz or faster processor, at least 2GB of RAM, and a 3D graphics card (with a GPU and its own graphics memory) with accelerated OpenGL drivers, is recommended. Freesurfer is highly CPU and memory intensive (and moderately disk intensive), so concentrate on boosting those performance aspects (more memory is better). Most modern video graphics card will perform fine, but be aware that graphics cards that use CPU memory as video memory will have a noticeably slow redraw rate.
Thanks,
Brian Hanna CMRR University of Minnesota
The problem with the NVIDIA driver installation is that it overwrites the GL shared lib links and removes the default CentOS GL libs. If you then do an update via up2date, yum, or whatever you use to keep CentOS up to date and one of the packages updated is the GL one, then the NVIDIA ones get overwritten with the defaults ones again. The only fix is to reinstall the NVIDIA driver.
So you have to monitor your upgrades closely and anytime you see an 'xorg' package upgrage check the GL libs and reinstall NVIDIA if needed.
On Thu, 24 Apr 2008, Brian Hanna wrote:
Hi all,
We are having a problem running tksurfer on several Centos 5 machines. We are getting the dreaded GLXBadLargeRequest and just the front sliver of the brain displayed. By installing the Nvidia driver manually, we made it work.
We have one Centos 5 machine where it is working (machine01). That one machine has an Nvidia card and the Nvidia driver installed.
We find that machine01 runs tksurfer correctly locally, and other machines can connect to machine01 with VNC and run tksurfer remotely on machine01. None of the other machines work correctly.
Following up on the Nvidia driver difference, I notice that the Nvidia driver package includes a number of GL libraries. On the other machines, standard Centos 5 installs, I see:
$ ls -l /usr/lib*/libGL* lrwxrwxrwx 1 root root 10 Jan 7 14:57 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 12 Jan 7 14:57 /usr/lib64/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 492720 Nov 10 11:23 /usr/lib64/libGL.so.1.2 lrwxrwxrwx 1 root root 11 Jan 7 14:57 /usr/lib64/libGLU.so -> libGLU.so.1 lrwxrwxrwx 1 root root 20 Jan 7 14:57 /usr/lib64/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 526048 Nov 10 11:23 /usr/lib64/libGLU.so.1.3.060501 lrwxrwxrwx 1 root root 12 Apr 23 14:49 /usr/lib/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 432016 Nov 10 11:24 /usr/lib/libGL.so.1.2 lrwxrwxrwx 1 root root 20 Feb 3 16:07 /usr/lib/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 524000 Nov 10 11:24 /usr/lib/libGLU.so.1.3.060501
I see (based on the modification time of the install) that there are several new libraries installed from Nvidia. That's a new libGL and 10mb libGLcore, and my standard libGL.so.1.2 has been zapped.
$ ls -l /usr/lib*/lib* | grep 09:21 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib64/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 9777704 Apr 16 09:21 /usr/lib64/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 661 Apr 16 09:21 /usr/lib64/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 822960 Apr 16 09:21 /usr/lib64/libGL.so.100.14.11 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so -> libnvidia-cfg.so.1 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.1 -> libnvidia-cfg.so.100.14.11 -rwxr-xr-x 1 root root 127096 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 3016 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.100.14.11 -r--r--r-- 1 root root 185148 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.a lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA_dynamic.so.1 -> libXvMCNVIDIA.so.100.14.11 -rwxr-xr-x 1 root root 137944 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.so.100.14.11 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 10080488 Apr 16 09:21 /usr/lib/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 659 Apr 16 09:21 /usr/lib/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 608400 Apr 16 09:21 /usr/lib/libGL.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 2352 Apr 16 09:21 /usr/lib/libnvidia-tls.so.100.14.11
For a test, I temporarily copied the libraries and make the appropriate links on another machine - basically installing the driver manually. I find that now the entire brain displays and I don't get the GLXBadLargeRequest.
I suppose I could have installed an Nvidia card and installed the driver to accomplish the same thing. You can use the -x option to the NVidia driver package to just extract it without running the installer. Be sure to check the LICENSE file, sections 2.1.2 and 2.1.3, before deciding how to proceed with this information.
So the problem lies in functionality available in the Nvidia libGL (etc) libraries that is not available in the standard Centos 5 libGL. What GL/GLX libraries do I install on a non-Nvidia machine in order to run tksurfer? Does tksurfer depend specifically on having Nvidia cards or other proprietary drivers installed? The wiki states:
A 2GHz or faster processor, at least 2GB of RAM, and a 3D graphics card (with a GPU and its own graphics memory) with accelerated OpenGL drivers, is recommended. Freesurfer is highly CPU and memory intensive (and moderately disk intensive), so concentrate on boosting those performance aspects (more memory is better). Most modern video graphics card will perform fine, but be aware that graphics cards that use CPU memory as video memory will have a noticeably slow redraw rate.
Thanks,
Brian Hanna CMRR University of Minnesota _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Yes, that makes sense. But what libraries can I use to run tksurfer -without- installing the Nvidia libraries?
Paul Raines wrote:
The problem with the NVIDIA driver installation is that it overwrites the GL shared lib links and removes the default CentOS GL libs. If you then do an update via up2date, yum, or whatever you use to keep CentOS up to date and one of the packages updated is the GL one, then the NVIDIA ones get overwritten with the defaults ones again. The only fix is to reinstall the NVIDIA driver.
So you have to monitor your upgrades closely and anytime you see an 'xorg' package upgrage check the GL libs and reinstall NVIDIA if needed.
On Thu, 24 Apr 2008, Brian Hanna wrote:
Hi all,
We are having a problem running tksurfer on several Centos 5 machines. We are getting the dreaded GLXBadLargeRequest and just the front sliver of the brain displayed. By installing the Nvidia driver manually, we made it work.
We have one Centos 5 machine where it is working (machine01). That one machine has an Nvidia card and the Nvidia driver installed.
We find that machine01 runs tksurfer correctly locally, and other machines can connect to machine01 with VNC and run tksurfer remotely on machine01. None of the other machines work correctly.
Following up on the Nvidia driver difference, I notice that the Nvidia driver package includes a number of GL libraries. On the other machines, standard Centos 5 installs, I see:
$ ls -l /usr/lib*/libGL* lrwxrwxrwx 1 root root 10 Jan 7 14:57 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 12 Jan 7 14:57 /usr/lib64/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 492720 Nov 10 11:23 /usr/lib64/libGL.so.1.2 lrwxrwxrwx 1 root root 11 Jan 7 14:57 /usr/lib64/libGLU.so -> libGLU.so.1 lrwxrwxrwx 1 root root 20 Jan 7 14:57 /usr/lib64/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 526048 Nov 10 11:23 /usr/lib64/libGLU.so.1.3.060501 lrwxrwxrwx 1 root root 12 Apr 23 14:49 /usr/lib/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 432016 Nov 10 11:24 /usr/lib/libGL.so.1.2 lrwxrwxrwx 1 root root 20 Feb 3 16:07 /usr/lib/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 524000 Nov 10 11:24 /usr/lib/libGLU.so.1.3.060501
I see (based on the modification time of the install) that there are several new libraries installed from Nvidia. That's a new libGL and 10mb libGLcore, and my standard libGL.so.1.2 has been zapped.
$ ls -l /usr/lib*/lib* | grep 09:21 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib64/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 9777704 Apr 16 09:21 /usr/lib64/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 661 Apr 16 09:21 /usr/lib64/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 822960 Apr 16 09:21 /usr/lib64/libGL.so.100.14.11 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so -> libnvidia-cfg.so.1 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.1 -> libnvidia-cfg.so.100.14.11 -rwxr-xr-x 1 root root 127096 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 3016 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.100.14.11 -r--r--r-- 1 root root 185148 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.a lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA_dynamic.so.1 -> libXvMCNVIDIA.so.100.14.11 -rwxr-xr-x 1 root root 137944 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.so.100.14.11 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 10080488 Apr 16 09:21 /usr/lib/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 659 Apr 16 09:21 /usr/lib/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 608400 Apr 16 09:21 /usr/lib/libGL.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 2352 Apr 16 09:21 /usr/lib/libnvidia-tls.so.100.14.11
For a test, I temporarily copied the libraries and make the appropriate links on another machine - basically installing the driver manually. I find that now the entire brain displays and I don't get the GLXBadLargeRequest.
I suppose I could have installed an Nvidia card and installed the driver to accomplish the same thing. You can use the -x option to the NVidia driver package to just extract it without running the installer. Be sure to check the LICENSE file, sections 2.1.2 and 2.1.3, before deciding how to proceed with this information.
So the problem lies in functionality available in the Nvidia libGL (etc) libraries that is not available in the standard Centos 5 libGL. What GL/GLX libraries do I install on a non-Nvidia machine in order to run tksurfer? Does tksurfer depend specifically on having Nvidia cards or other proprietary drivers installed? The wiki states:
A 2GHz or faster processor, at least 2GB of RAM, and a 3D graphics card (with a GPU and its own graphics memory) with accelerated OpenGL drivers, is recommended. Freesurfer is highly CPU and memory intensive (and moderately disk intensive), so concentrate on boosting those performance aspects (more memory is better). Most modern video graphics card will perform fine, but be aware that graphics cards that use CPU memory as video memory will have a noticeably slow redraw rate.
Thanks,
Brian Hanna CMRR University of Minnesota _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Wait, you are copying the NVIDIA GL replacement libraries to machines that do not have NVIDIA cards and the proprietary nvidia driver installed? ANd it works? I would think no GL app would run in that case.
Maybe the NVIDIA GL libs are smart enough to handle the case they find no NVIDIA card/driver. Anyway, that points to some kind of bug in the default CentOS GL libs from Mesa.
One thing you can also try is getting the GL libs from some older version of the default Mesa ones, such as these from the original release of 5.0
mesa-libGL-6.5.1-7.2.el5.i386.rpm mesa-libGL-devel-6.5.1-7.2.el5.i386.rpm
Looks like the current ones are 7.5. SO if the 7.2 ones work, then something was introduced since then.
On Thu, 24 Apr 2008, Brian Hanna wrote:
Yes, that makes sense. But what libraries can I use to run tksurfer -without- installing the Nvidia libraries?
Paul Raines wrote:
The problem with the NVIDIA driver installation is that it overwrites the GL shared lib links and removes the default CentOS GL libs. If you then do an update via up2date, yum, or whatever you use to keep CentOS up to date and one of the packages updated is the GL one, then the NVIDIA ones get overwritten with the defaults ones again. The only fix is to reinstall the NVIDIA driver.
So you have to monitor your upgrades closely and anytime you see an 'xorg' package upgrage check the GL libs and reinstall NVIDIA if needed.
On Thu, 24 Apr 2008, Brian Hanna wrote:
Hi all,
We are having a problem running tksurfer on several Centos 5 machines. We are getting the dreaded GLXBadLargeRequest and just the front sliver of the brain displayed. By installing the Nvidia driver manually, we made it work.
We have one Centos 5 machine where it is working (machine01). That one machine has an Nvidia card and the Nvidia driver installed.
We find that machine01 runs tksurfer correctly locally, and other machines can connect to machine01 with VNC and run tksurfer remotely on machine01. None of the other machines work correctly.
Following up on the Nvidia driver difference, I notice that the Nvidia driver package includes a number of GL libraries. On the other machines, standard Centos 5 installs, I see:
$ ls -l /usr/lib*/libGL* lrwxrwxrwx 1 root root 10 Jan 7 14:57 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 12 Jan 7 14:57 /usr/lib64/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 492720 Nov 10 11:23 /usr/lib64/libGL.so.1.2 lrwxrwxrwx 1 root root 11 Jan 7 14:57 /usr/lib64/libGLU.so -> libGLU.so.1 lrwxrwxrwx 1 root root 20 Jan 7 14:57 /usr/lib64/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 526048 Nov 10 11:23 /usr/lib64/libGLU.so.1.3.060501 lrwxrwxrwx 1 root root 12 Apr 23 14:49 /usr/lib/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 432016 Nov 10 11:24 /usr/lib/libGL.so.1.2 lrwxrwxrwx 1 root root 20 Feb 3 16:07 /usr/lib/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 524000 Nov 10 11:24 /usr/lib/libGLU.so.1.3.060501
I see (based on the modification time of the install) that there are several new libraries installed from Nvidia. That's a new libGL and 10mb libGLcore, and my standard libGL.so.1.2 has been zapped.
$ ls -l /usr/lib*/lib* | grep 09:21 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib64/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 9777704 Apr 16 09:21 /usr/lib64/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 661 Apr 16 09:21 /usr/lib64/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 822960 Apr 16 09:21 /usr/lib64/libGL.so.100.14.11 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so -> libnvidia-cfg.so.1 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.1 -> libnvidia-cfg.so.100.14.11 -rwxr-xr-x 1 root root 127096 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 3016 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.100.14.11 -r--r--r-- 1 root root 185148 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.a lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA_dynamic.so.1 -> libXvMCNVIDIA.so.100.14.11 -rwxr-xr-x 1 root root 137944 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.so.100.14.11 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 10080488 Apr 16 09:21 /usr/lib/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 659 Apr 16 09:21 /usr/lib/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 608400 Apr 16 09:21 /usr/lib/libGL.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 2352 Apr 16 09:21 /usr/lib/libnvidia-tls.so.100.14.11
For a test, I temporarily copied the libraries and make the appropriate links on another machine - basically installing the driver manually. I find that now the entire brain displays and I don't get the GLXBadLargeRequest.
I suppose I could have installed an Nvidia card and installed the driver to accomplish the same thing. You can use the -x option to the NVidia driver package to just extract it without running the installer. Be sure to check the LICENSE file, sections 2.1.2 and 2.1.3, before deciding how to proceed with this information.
So the problem lies in functionality available in the Nvidia libGL (etc) libraries that is not available in the standard Centos 5 libGL. What GL/GLX libraries do I install on a non-Nvidia machine in order to run tksurfer? Does tksurfer depend specifically on having Nvidia cards or other proprietary drivers installed? The wiki states:
A 2GHz or faster processor, at least 2GB of RAM, and a 3D graphics card (with a GPU and its own graphics memory) with accelerated OpenGL drivers, is recommended. Freesurfer is highly CPU and memory intensive (and moderately disk intensive), so concentrate on boosting those performance aspects (more memory is better). Most modern video graphics card will perform fine, but be aware that graphics cards that use CPU memory as video memory will have a noticeably slow redraw rate.
Thanks,
Brian Hanna CMRR University of Minnesota _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Well, yes, as a test I copied in the Nvidia GL libraries. For remote VNC, it works. Locally, though, glxgears and glxinfo now give segfaults.
On the wiki, I see a downloadable version of vncserver.glx.
http://surfer.nmr.mgh.harvard.edu/fswiki/RemoteAccess
This version sets up which refers to some special GLoverride libraries at mgh.
if ($ENV{LD_LIBRARY_PATH}) { $ENV{LD_LIBRARY_PATH} = '/usr/lib64/GLoverride:/usr/lib/GLoverride:' . $ENV{LD_LIBRARY_PATH}; } else { $ENV{LD_LIBRARY_PATH} = '/usr/lib64/GLoverride:/usr/lib/GLoverride'; }
What is in those libraries - where did they come from? Perhaps this is the approach I need to use - leave the standard distro GL libraries in place, and put the Nvidia libraries into another location for use only by vnc.
Thanks,
Brian
Paul Raines wrote:
Wait, you are copying the NVIDIA GL replacement libraries to machines that do not have NVIDIA cards and the proprietary nvidia driver installed? ANd it works? I would think no GL app would run in that case.
Maybe the NVIDIA GL libs are smart enough to handle the case they find no NVIDIA card/driver. Anyway, that points to some kind of bug in the default CentOS GL libs from Mesa.
One thing you can also try is getting the GL libs from some older version of the default Mesa ones, such as these from the original release of 5.0
mesa-libGL-6.5.1-7.2.el5.i386.rpm mesa-libGL-devel-6.5.1-7.2.el5.i386.rpm
Looks like the current ones are 7.5. SO if the 7.2 ones work, then something was introduced since then.
On Thu, 24 Apr 2008, Brian Hanna wrote:
Yes, that makes sense. But what libraries can I use to run tksurfer -without- installing the Nvidia libraries?
Paul Raines wrote:
The problem with the NVIDIA driver installation is that it overwrites the GL shared lib links and removes the default CentOS GL libs. If you then do an update via up2date, yum, or whatever you use to keep CentOS up to date and one of the packages updated is the GL one, then the NVIDIA ones get overwritten with the defaults ones again. The only fix is to reinstall the NVIDIA driver.
So you have to monitor your upgrades closely and anytime you see an 'xorg' package upgrage check the GL libs and reinstall NVIDIA if needed.
On Thu, 24 Apr 2008, Brian Hanna wrote:
Hi all,
We are having a problem running tksurfer on several Centos 5 machines. We are getting the dreaded GLXBadLargeRequest and just the front sliver of the brain displayed. By installing the Nvidia driver manually, we made it work.
We have one Centos 5 machine where it is working (machine01). That one machine has an Nvidia card and the Nvidia driver installed.
We find that machine01 runs tksurfer correctly locally, and other machines can connect to machine01 with VNC and run tksurfer remotely on machine01. None of the other machines work correctly.
Following up on the Nvidia driver difference, I notice that the Nvidia driver package includes a number of GL libraries. On the other machines, standard Centos 5 installs, I see:
$ ls -l /usr/lib*/libGL* lrwxrwxrwx 1 root root 10 Jan 7 14:57 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 12 Jan 7 14:57 /usr/lib64/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 492720 Nov 10 11:23 /usr/lib64/libGL.so.1.2 lrwxrwxrwx 1 root root 11 Jan 7 14:57 /usr/lib64/libGLU.so -> libGLU.so.1 lrwxrwxrwx 1 root root 20 Jan 7 14:57 /usr/lib64/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 526048 Nov 10 11:23 /usr/lib64/libGLU.so.1.3.060501 lrwxrwxrwx 1 root root 12 Apr 23 14:49 /usr/lib/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 432016 Nov 10 11:24 /usr/lib/libGL.so.1.2 lrwxrwxrwx 1 root root 20 Feb 3 16:07 /usr/lib/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 524000 Nov 10 11:24 /usr/lib/libGLU.so.1.3.060501
I see (based on the modification time of the install) that there are several new libraries installed from Nvidia. That's a new libGL and 10mb libGLcore, and my standard libGL.so.1.2 has been zapped.
$ ls -l /usr/lib*/lib* | grep 09:21 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib64/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 9777704 Apr 16 09:21 /usr/lib64/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 661 Apr 16 09:21 /usr/lib64/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 822960 Apr 16 09:21 /usr/lib64/libGL.so.100.14.11 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so -> libnvidia-cfg.so.1 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.1 -> libnvidia-cfg.so.100.14.11 -rwxr-xr-x 1 root root 127096 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 3016 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.100.14.11 -r--r--r-- 1 root root 185148 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.a lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA_dynamic.so.1 -> libXvMCNVIDIA.so.100.14.11 -rwxr-xr-x 1 root root 137944 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.so.100.14.11 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 10080488 Apr 16 09:21 /usr/lib/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 659 Apr 16 09:21 /usr/lib/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 608400 Apr 16 09:21 /usr/lib/libGL.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 2352 Apr 16 09:21 /usr/lib/libnvidia-tls.so.100.14.11
For a test, I temporarily copied the libraries and make the appropriate links on another machine - basically installing the driver manually. I find that now the entire brain displays and I don't get the GLXBadLargeRequest.
I suppose I could have installed an Nvidia card and installed the driver to accomplish the same thing. You can use the -x option to the NVidia driver package to just extract it without running the installer. Be sure to check the LICENSE file, sections 2.1.2 and 2.1.3, before deciding how to proceed with this information.
So the problem lies in functionality available in the Nvidia libGL (etc) libraries that is not available in the standard Centos 5 libGL. What GL/GLX libraries do I install on a non-Nvidia machine in order to run tksurfer? Does tksurfer depend specifically on having Nvidia cards or other proprietary drivers installed? The wiki states:
A 2GHz or faster processor, at least 2GB of RAM, and a 3D graphics card (with a GPU and its own graphics memory) with accelerated OpenGL drivers, is recommended. Freesurfer is highly CPU and memory intensive (and moderately disk intensive), so concentrate on boosting those performance aspects (more memory is better). Most modern video graphics card will perform fine, but be aware that graphics cards that use CPU memory as video memory will have a noticeably slow redraw rate.
Thanks,
Brian Hanna CMRR University of Minnesota _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
At the Martinos Center, I have extracted the GL libraries from the Mesa rpms and put them in those directories. When some one is using VNC (or even doing remote X), it is better to use those libraries than the ones that come with NVIDIA. Otherwise they do get errors. Which is pretty much the opposite of what you are describing.
The libs I have in /usr/lib64/GLoverride:/usr/lib/GLoverride were extraced long ago and I have not kept them up to date with new Mesa updates that come with CentOS unless a problem has arisen
So you putting the NVIDIA libraries there would be the complete opposite of what we are doing.
On Thu, 24 Apr 2008, Brian Hanna wrote:
Well, yes, as a test I copied in the Nvidia GL libraries. For remote VNC, it works. Locally, though, glxgears and glxinfo now give segfaults.
On the wiki, I see a downloadable version of vncserver.glx.
http://surfer.nmr.mgh.harvard.edu/fswiki/RemoteAccess
This version sets up which refers to some special GLoverride libraries at mgh.
if ($ENV{LD_LIBRARY_PATH}) { $ENV{LD_LIBRARY_PATH} = '/usr/lib64/GLoverride:/usr/lib/GLoverride:' . $ENV{LD_LIBRARY_PATH}; } else { $ENV{LD_LIBRARY_PATH} = '/usr/lib64/GLoverride:/usr/lib/GLoverride'; }
What is in those libraries - where did they come from? Perhaps this is the approach I need to use - leave the standard distro GL libraries in place, and put the Nvidia libraries into another location for use only by vnc.
Thanks,
Brian
Paul Raines wrote:
Wait, you are copying the NVIDIA GL replacement libraries to machines that do not have NVIDIA cards and the proprietary nvidia driver installed? ANd it works? I would think no GL app would run in that case.
Maybe the NVIDIA GL libs are smart enough to handle the case they find no NVIDIA card/driver. Anyway, that points to some kind of bug in the default CentOS GL libs from Mesa.
One thing you can also try is getting the GL libs from some older version of the default Mesa ones, such as these from the original release of 5.0
mesa-libGL-6.5.1-7.2.el5.i386.rpm mesa-libGL-devel-6.5.1-7.2.el5.i386.rpm
Looks like the current ones are 7.5. SO if the 7.2 ones work, then something was introduced since then.
On Thu, 24 Apr 2008, Brian Hanna wrote:
Yes, that makes sense. But what libraries can I use to run tksurfer -without- installing the Nvidia libraries?
Paul Raines wrote:
The problem with the NVIDIA driver installation is that it overwrites the GL shared lib links and removes the default CentOS GL libs. If you then do an update via up2date, yum, or whatever you use to keep CentOS up to date and one of the packages updated is the GL one, then the NVIDIA ones get overwritten with the defaults ones again. The only fix is to reinstall the NVIDIA driver.
So you have to monitor your upgrades closely and anytime you see an 'xorg' package upgrage check the GL libs and reinstall NVIDIA if needed.
On Thu, 24 Apr 2008, Brian Hanna wrote:
Hi all,
We are having a problem running tksurfer on several Centos 5 machines. We are getting the dreaded GLXBadLargeRequest and just the front sliver of the brain displayed. By installing the Nvidia driver manually, we made it work.
We have one Centos 5 machine where it is working (machine01). That one machine has an Nvidia card and the Nvidia driver installed.
We find that machine01 runs tksurfer correctly locally, and other machines can connect to machine01 with VNC and run tksurfer remotely on machine01. None of the other machines work correctly.
Following up on the Nvidia driver difference, I notice that the Nvidia driver package includes a number of GL libraries. On the other machines, standard Centos 5 installs, I see:
$ ls -l /usr/lib*/libGL* lrwxrwxrwx 1 root root 10 Jan 7 14:57 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 12 Jan 7 14:57 /usr/lib64/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 492720 Nov 10 11:23 /usr/lib64/libGL.so.1.2 lrwxrwxrwx 1 root root 11 Jan 7 14:57 /usr/lib64/libGLU.so -> libGLU.so.1 lrwxrwxrwx 1 root root 20 Jan 7 14:57 /usr/lib64/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 526048 Nov 10 11:23 /usr/lib64/libGLU.so.1.3.060501 lrwxrwxrwx 1 root root 12 Apr 23 14:49 /usr/lib/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 432016 Nov 10 11:24 /usr/lib/libGL.so.1.2 lrwxrwxrwx 1 root root 20 Feb 3 16:07 /usr/lib/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 524000 Nov 10 11:24 /usr/lib/libGLU.so.1.3.060501
I see (based on the modification time of the install) that there are several new libraries installed from Nvidia. That's a new libGL and 10mb libGLcore, and my standard libGL.so.1.2 has been zapped.
$ ls -l /usr/lib*/lib* | grep 09:21 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib64/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 9777704 Apr 16 09:21 /usr/lib64/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 661 Apr 16 09:21 /usr/lib64/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 822960 Apr 16 09:21 /usr/lib64/libGL.so.100.14.11 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so -> libnvidia-cfg.so.1 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.1 -> libnvidia-cfg.so.100.14.11 -rwxr-xr-x 1 root root 127096 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 3016 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.100.14.11 -r--r--r-- 1 root root 185148 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.a lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA_dynamic.so.1 -> libXvMCNVIDIA.so.100.14.11 -rwxr-xr-x 1 root root 137944 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.so.100.14.11 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 10080488 Apr 16 09:21 /usr/lib/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 659 Apr 16 09:21 /usr/lib/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 608400 Apr 16 09:21 /usr/lib/libGL.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 2352 Apr 16 09:21 /usr/lib/libnvidia-tls.so.100.14.11
For a test, I temporarily copied the libraries and make the appropriate links on another machine - basically installing the driver manually. I find that now the entire brain displays and I don't get the GLXBadLargeRequest.
I suppose I could have installed an Nvidia card and installed the driver to accomplish the same thing. You can use the -x option to the NVidia driver package to just extract it without running the installer. Be sure to check the LICENSE file, sections 2.1.2 and 2.1.3, before deciding how to proceed with this information.
So the problem lies in functionality available in the Nvidia libGL (etc) libraries that is not available in the standard Centos 5 libGL. What GL/GLX libraries do I install on a non-Nvidia machine in order to run tksurfer? Does tksurfer depend specifically on having Nvidia cards or other proprietary drivers installed? The wiki states:
A 2GHz or faster processor, at least 2GB of RAM, and a 3D graphics card (with a GPU and its own graphics memory) with accelerated OpenGL drivers, is recommended. Freesurfer is highly CPU and memory intensive (and moderately disk intensive), so concentrate on boosting those performance aspects (more memory is better). Most modern video graphics card will perform fine, but be aware that graphics cards that use CPU memory as video memory will have a noticeably slow redraw rate.
Thanks,
Brian Hanna CMRR University of Minnesota _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Brian,
Does the 'qdec' application run for you on Centos5?
Our apps do not require any particular graphics card. Through experience, it seems the Nvidia cards give the least trouble (compared to ATI), but we have never been able to track down exactly why (something in the bowels of each manufactures OpenGL/GLX implementation). I think, but am not positive, that GLX is graphics- card specific (well I would think OpenGL is too), and tksurfer uses GLX (although I think there is a way to force software rendering), so I cannot say how the default Centos5 GL/GLX drivers will behave.
So for the systems that are not working, what graphics card is installed?
Nick
On Thu, 2008-04-24 at 11:54 -0500, Brian Hanna wrote:
Hi all,
We are having a problem running tksurfer on several Centos 5 machines. We are getting the dreaded GLXBadLargeRequest and just the front sliver of the brain displayed. By installing the Nvidia driver manually, we made it work.
We have one Centos 5 machine where it is working (machine01). That one machine has an Nvidia card and the Nvidia driver installed.
We find that machine01 runs tksurfer correctly locally, and other machines can connect to machine01 with VNC and run tksurfer remotely on machine01. None of the other machines work correctly.
Following up on the Nvidia driver difference, I notice that the Nvidia driver package includes a number of GL libraries. On the other machines, standard Centos 5 installs, I see:
$ ls -l /usr/lib*/libGL* lrwxrwxrwx 1 root root 10 Jan 7 14:57 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 12 Jan 7 14:57 /usr/lib64/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 492720 Nov 10 11:23 /usr/lib64/libGL.so.1.2 lrwxrwxrwx 1 root root 11 Jan 7 14:57 /usr/lib64/libGLU.so -> libGLU.so.1 lrwxrwxrwx 1 root root 20 Jan 7 14:57 /usr/lib64/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 526048 Nov 10 11:23 /usr/lib64/libGLU.so.1.3.060501 lrwxrwxrwx 1 root root 12 Apr 23 14:49 /usr/lib/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 432016 Nov 10 11:24 /usr/lib/libGL.so.1.2 lrwxrwxrwx 1 root root 20 Feb 3 16:07 /usr/lib/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 524000 Nov 10 11:24 /usr/lib/libGLU.so.1.3.060501
I see (based on the modification time of the install) that there are several new libraries installed from Nvidia. That's a new libGL and 10mb libGLcore, and my standard libGL.so.1.2 has been zapped.
$ ls -l /usr/lib*/lib* | grep 09:21 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib64/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 9777704 Apr 16 09:21 /usr/lib64/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 661 Apr 16 09:21 /usr/lib64/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 822960 Apr 16 09:21 /usr/lib64/libGL.so.100.14.11 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so -> libnvidia-cfg.so.1 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.1 -> libnvidia-cfg.so.100.14.11 -rwxr-xr-x 1 root root 127096 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 3016 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.100.14.11 -r--r--r-- 1 root root 185148 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.a lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA_dynamic.so.1 -> libXvMCNVIDIA.so.100.14.11 -rwxr-xr-x 1 root root 137944 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.so.100.14.11 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 10080488 Apr 16 09:21 /usr/lib/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 659 Apr 16 09:21 /usr/lib/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 608400 Apr 16 09:21 /usr/lib/libGL.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 2352 Apr 16 09:21 /usr/lib/libnvidia-tls.so.100.14.11
For a test, I temporarily copied the libraries and make the appropriate links on another machine - basically installing the driver manually. I find that now the entire brain displays and I don't get the GLXBadLargeRequest.
I suppose I could have installed an Nvidia card and installed the driver to accomplish the same thing. You can use the -x option to the NVidia driver package to just extract it without running the installer. Be sure to check the LICENSE file, sections 2.1.2 and 2.1.3, before deciding how to proceed with this information.
So the problem lies in functionality available in the Nvidia libGL (etc) libraries that is not available in the standard Centos 5 libGL. What GL/GLX libraries do I install on a non-Nvidia machine in order to run tksurfer? Does tksurfer depend specifically on having Nvidia cards or other proprietary drivers installed? The wiki states:
A 2GHz or faster processor, at least 2GB of RAM, and a 3D graphics card (with a GPU and its own graphics memory) with accelerated OpenGL drivers, is recommended. Freesurfer is highly CPU and memory intensive (and moderately disk intensive), so concentrate on boosting those performance aspects (more memory is better). Most modern video graphics card will perform fine, but be aware that graphics cards that use CPU memory as video memory will have a noticeably slow redraw rate.
Thanks,
Brian Hanna CMRR University of Minnesota _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Yes, qdec apparently works ok without the Nvidia driver libraries.
We want to run tksurfer remotely, so the video card won't really be used. But apparently we need special functionality that does not exist in the standard GL libraries.
Purchasing a video card just to get access to the driver libraries - is this what we need to do?
Or is there a standard, open-source GL library that supports everything that tksurfer requires?
Thanks,
Brian
Nick Schmansky wrote:
Brian,
Does the 'qdec' application run for you on Centos5?
Our apps do not require any particular graphics card. Through experience, it seems the Nvidia cards give the least trouble (compared to ATI), but we have never been able to track down exactly why (something in the bowels of each manufactures OpenGL/GLX implementation). I think, but am not positive, that GLX is graphics- card specific (well I would think OpenGL is too), and tksurfer uses GLX (although I think there is a way to force software rendering), so I cannot say how the default Centos5 GL/GLX drivers will behave.
So for the systems that are not working, what graphics card is installed?
Nick
On Thu, 2008-04-24 at 11:54 -0500, Brian Hanna wrote:
Hi all,
We are having a problem running tksurfer on several Centos 5 machines. We are getting the dreaded GLXBadLargeRequest and just the front sliver of the brain displayed. By installing the Nvidia driver manually, we made it work.
We have one Centos 5 machine where it is working (machine01). That one machine has an Nvidia card and the Nvidia driver installed.
We find that machine01 runs tksurfer correctly locally, and other machines can connect to machine01 with VNC and run tksurfer remotely on machine01. None of the other machines work correctly.
Following up on the Nvidia driver difference, I notice that the Nvidia driver package includes a number of GL libraries. On the other machines, standard Centos 5 installs, I see:
$ ls -l /usr/lib*/libGL* lrwxrwxrwx 1 root root 10 Jan 7 14:57 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 12 Jan 7 14:57 /usr/lib64/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 492720 Nov 10 11:23 /usr/lib64/libGL.so.1.2 lrwxrwxrwx 1 root root 11 Jan 7 14:57 /usr/lib64/libGLU.so -> libGLU.so.1 lrwxrwxrwx 1 root root 20 Jan 7 14:57 /usr/lib64/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 526048 Nov 10 11:23 /usr/lib64/libGLU.so.1.3.060501 lrwxrwxrwx 1 root root 12 Apr 23 14:49 /usr/lib/libGL.so.1 -> libGL.so.1.2 -rwxr-xr-x 1 root root 432016 Nov 10 11:24 /usr/lib/libGL.so.1.2 lrwxrwxrwx 1 root root 20 Feb 3 16:07 /usr/lib/libGLU.so.1 -> libGLU.so.1.3.060501 -rwxr-xr-x 1 root root 524000 Nov 10 11:24 /usr/lib/libGLU.so.1.3.060501
I see (based on the modification time of the install) that there are several new libraries installed from Nvidia. That's a new libGL and 10mb libGLcore, and my standard libGL.so.1.2 has been zapped.
$ ls -l /usr/lib*/lib* | grep 09:21 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib64/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 9777704 Apr 16 09:21 /usr/lib64/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 661 Apr 16 09:21 /usr/lib64/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib64/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 822960 Apr 16 09:21 /usr/lib64/libGL.so.100.14.11 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so -> libnvidia-cfg.so.1 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.1 -> libnvidia-cfg.so.100.14.11 -rwxr-xr-x 1 root root 127096 Apr 16 09:21 /usr/lib64/libnvidia-cfg.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 3016 Apr 16 09:21 /usr/lib64/libnvidia-tls.so.100.14.11 -r--r--r-- 1 root root 185148 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.a lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA_dynamic.so.1 -> libXvMCNVIDIA.so.100.14.11 -rwxr-xr-x 1 root root 137944 Apr 16 09:21 /usr/lib64/libXvMCNVIDIA.so.100.14.11 lrwxrwxrwx 1 root root 22 Apr 16 09:21 /usr/lib/libGLcore.so.1 -> libGLcore.so.100.14.11 -rwxr-xr-x 1 root root 10080488 Apr 16 09:21 /usr/lib/libGLcore.so.100.14.11 -rw-r--r-- 1 root root 659 Apr 16 09:21 /usr/lib/libGL.la lrwxrwxrwx 1 root root 10 Apr 16 09:21 /usr/lib/libGL.so -> libGL.so.1 lrwxrwxrwx 1 root root 18 Apr 16 09:21 /usr/lib/libGL.so.1 -> libGL.so.100.14.11 -rwxr-xr-x 1 root root 608400 Apr 16 09:21 /usr/lib/libGL.so.100.14.11 lrwxrwxrwx 1 root root 26 Apr 16 09:21 /usr/lib/libnvidia-tls.so.1 -> libnvidia-tls.so.100.14.11 -rwxr-xr-x 1 root root 2352 Apr 16 09:21 /usr/lib/libnvidia-tls.so.100.14.11
For a test, I temporarily copied the libraries and make the appropriate links on another machine - basically installing the driver manually. I find that now the entire brain displays and I don't get the GLXBadLargeRequest.
I suppose I could have installed an Nvidia card and installed the driver to accomplish the same thing. You can use the -x option to the NVidia driver package to just extract it without running the installer. Be sure to check the LICENSE file, sections 2.1.2 and 2.1.3, before deciding how to proceed with this information.
So the problem lies in functionality available in the Nvidia libGL (etc) libraries that is not available in the standard Centos 5 libGL. What GL/GLX libraries do I install on a non-Nvidia machine in order to run tksurfer? Does tksurfer depend specifically on having Nvidia cards or other proprietary drivers installed? The wiki states:
A 2GHz or faster processor, at least 2GB of RAM, and a 3D graphics card (with a GPU and its own graphics memory) with accelerated OpenGL drivers, is recommended. Freesurfer is highly CPU and memory intensive (and moderately disk intensive), so concentrate on boosting those performance aspects (more memory is better). Most modern video graphics card will perform fine, but be aware that graphics cards that use CPU memory as video memory will have a noticeably slow redraw rate.
Thanks,
Brian Hanna CMRR University of Minnesota _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Brian Hanna -
Nick Schmansky -
Paul Raines