View Issue Details

IDProjectCategoryView StatusLast Update
0000724channel: elrepo/el7kmod-nvidiapublic2017-05-10 13:25
Reporterwiad Assigned Topperry  
PrioritynormalSeveritymajorReproducibilityalways
Status resolvedResolutionfixed 
Summary0000724: kmod-nvidia-375.39-2 not working with NVIDIA GRID K1
DescriptionOn RHEL7, after updating kmod-nvidia to latest version (375.39, required for the 3.10.0-514.10.2 kernel), the nvidia kernel modules fails to load.

dmesg output:

NVRM: The NVIDIA GRID K1 GPU installed in this system is
NVRM: supported through the NVIDIA 367.xx Legacy drivers. Please
NVRM: visit http://www.nvidia.com/object/unix.html for more
NVRM: information. The 375.39 NVIDIA driver will ignore
NVRM: this GPU. Continuing probe...


Additional InformationNot tried with NVIDIA GRID K2, but supposedly the same issue there?

nvidia packages:
nvidia-x11-drv-375.39-1.el7.elrepo.x86_64
kmod-nvidia-375.39-2.el7.elrepo.x86_64

kernel:
3.10.0-514.10.2.el7.x86_64


TagsNo tags attached.
Reported upstream

Activities

wiad

2017-03-29 04:55

reporter   ~0005109

Ok, so nvidia has stopped supporting GRID K1/K2 cards with their new driver branch. Question is, will there be a legacy 367 driver available from elrepo that we can use with our K1/K2 cards?

pperry

2017-03-29 10:55

administrator   ~0005110

Yes, looks like you are correct. The GRID K1/K2/K340/K520 chipsets will need the 367.xx driver. Here's what the nvidia driver page says:

These GPUs will continue to be maintained through the special legacy NVIDIA GPU driver releases. The 367.xx driver supports the following set of GPUs:


So you will need to revert to the last 367.xx driver, but that won't work with the very latest RHEL7 kernel due to a change in the ABI, so I'll need to rebuild kmod-nvidia for you.

We can then fork it to a kmod-nvidia-367xx legacy package as we have done for other legacy branches as it looks like nvidia intend to maintain this branch separately for your chipset.

I'll try to get the initial kmod-nvidia update rebuilt for you asap so we can at least get you functional again. I'll post back here later this evening once done.

Thanks for the report.

wiad

2017-03-29 11:12

reporter   ~0005111

Hey, thats great! I was worried we would have to lock kernel version on these systems. Right now I've downgraded the kernel and installed the 367 kmod-nvidia version so we are in a functional state - so there's no panic on our behalf.

pperry

2017-03-29 12:52

administrator   ~0005112

Last edited: 2017-03-29 12:53

I have rebuilt the following package against the latest RHEL7 kernel (kernel-3.10.0-514.10.2.el7.x86_64) and released them to the main elrepo repository. They are syncing to the mirrors and should be available shortly.

kmod-nvidia-367.57-3.el7.elrepo.x86_64.rpm
nvidia-kmod-367.57-3.el7.elrepo.nosrc.rpm

So for now, you can install:

kmod-nvidia-367.57-3.el7.elrepo.x86_64.rpm <-- Note it's the -3.el7 release
nvidia-x11-drv-367.57-1.el7.elrepo.x86_64.rpm

and that combination should support the GRID K1/K2/K340/K520 devices on the latest RHEL7 kernel.

Please note, for kmod-nvidia-367.57, the -1.el7 release supports 7.2, -2.el7 supports early 7.3 kernels and the -3.el7 release is required for the current rhel7 kernel and above. The corresponding nvidia-x11-drv-367.57-1.el7.elrepo.x86_64.rpm package may be used with any of the above kmod-nvidia packages.

When you are ready to update, I would uninstall your current nvidia packages (yum erase kmod-nvidia), then install:

yum install kmod-nvidia-367.57-3.el7.elrepo.x86_64.rpm (making sure it's the correct release)

and reboot to the latest RHEL7 kernel (3.10.0-514.10.2.el7.x86_64). Yum will pull in nvidia-x11-drv for you as a dependency.

Then don't update your nvidia packages for now (you can add an exclude line to the repo file).

I will fork a legacy branch for you when nvidia next updates the 367.xx drivers.

I will also blacklist the GRID K1/K2/K340/K520 chipsets in the next 375.xx release and remove the old packages from the repo, so after that point yum-plugin-nvidia will prevent you from installing the unsupported 375.xx or above drivers.

Let me know if you have any issues and I'll update this bug as necessary

wiad

2017-03-31 00:09

reporter   ~0005113

This is great, thanks!

pperry

2017-05-10 12:42

administrator   ~0005177

Just to note, I've just released 375.66 drivers which now blacklist these unsupported devices so yum should no longer offer 375.xx updates (once packages have sync'd to all the mirror sites).

I'm still waiting for nvidia to make a 367.xx legacy release.

pperry

2017-05-10 13:25

administrator   ~0005178

I've just noticed in the changelog for 375.66:

Restored support for the following GPU: GRID K520

and sure enough, it's listed here as now supported:

http://us.download.nvidia.com/XFree86/Linux-x86_64/375.66/README/supportedchips.html

So we will need to un-blacklist GRID K520 [10DE:118A] on the next release.

Issue History

Date Modified Username Field Change
2017-03-29 01:55 wiad New Issue
2017-03-29 01:55 wiad Status new => assigned
2017-03-29 01:55 wiad Assigned To => pperry
2017-03-29 04:55 wiad Note Added: 0005109
2017-03-29 10:55 pperry Note Added: 0005110
2017-03-29 11:12 wiad Note Added: 0005111
2017-03-29 12:52 pperry Note Added: 0005112
2017-03-29 12:53 pperry Note Edited: 0005112
2017-03-31 00:09 wiad Note Added: 0005113
2017-04-01 07:50 pperry Status assigned => resolved
2017-04-01 07:50 pperry Resolution open => fixed
2017-05-10 12:42 pperry Note Added: 0005177
2017-05-10 13:25 pperry Note Added: 0005178