Linux won't boot after MIG enabled

Graphics / Linux Linux
1

Hi,

I have RHEL9.5 running a Dell R7525 server that has PCIe A100 GPU.
After enabling mig on the gpu the system won’t boot anymore

  • Nvidia Driver Version: 570.86.15 CUDA Version: 12.8
  • BIOS Firmware (2.18.1) latest

rpviewer (4)
rpviewer (4)1024×768 8.02 KB

How can I resolve this?

Updates:

  • cant get in rescue mode
  • as was suggested in a diff thread, tried booting with both kernel params below and still oops in [nvidia_modeset]
    • pci=realloc=off realloc=off
    • pci=realloc=on realloc=on
  • managed to turn mig off and at least can boot into linux
    (added devie to pci-stub, logged into linux, unbined device, modprobe nvidia, which didn’t cause any error(!?), disabled mig, reboot and all back working)

still would like to resolve the issue and enable mig, any help would be appreciated.

3

Same here. The problem is also with the new nvidia-driver 570.124.06.

Linux xyz 6.1.131-1.el9.elrepo.x86_64 #1 SMP PREEMPT_DYNAMIC Thu Mar 13 13:25:58 EDT 2025 x86_64 x86_64 x86_64 GNU/Linux

We are using AlmaLinux 9.5 and it’s a HPE ProLiant DL380 Gen10.

Thanks for the “pci-stub”-hint, I was able to revive the machine again.

As a workaround I modified to disable the nvidia-modeset-crashing-the-system-“feature”
/etc/modprobe.d/nvidia-modeset.conf

#options nvidia-drm modeset=0
install nvidia-modeset /bin/false

then

depmod -a
dracut -f
reboot

We are still doing it manually with pci-stub
and then loading the nvidia-driver by hand now,
and also the mig-mode seems to work.
I haven’t done another reboot yet.