Environment

• RAID Model: SR1000 / SR1001 / SR1010
• Host Hardware: Multi-GPU system (e.g., NVIDIA A-series, with or without PCIe switch such as PLX/PEX)
• Operating System: RHEL / Rocky / AlmaLinux, Ubuntu / Debian (UEFI boot)

• Software: SupremeRAID Pre-installer (v1.7.x+)
  
  

• Issue

  After installing the SupremeRAID pre-installer on a multi-GPU system, the RAID card is not detected by nvidia-smi.
  When checking kernel logs (dmesg), the following error messages appear:

  NVRM: This PCI I/O region assigned to your NVIDIA device is invalid:
  NVRM: BAR0 is 0M @ 0x0 (PCI:0001:05:00.0)
  nvidia: probe of 0001:05:00.0 failed with error -1
  
  

  This issue typically occurs during the boot sequence, before the NVIDIA or SupremeRAID drivers are fully initialized.

  ---

  Root Cause

  On systems with multiple GPUs or complex PCIe topologies, the Linux kernel may reallocate PCI resources during boot.
  When PCI reallocation is enabled, certain devices may receive incorrect or zero-sized BAR assignments (e.g., BAR0 = 0M), causing the NVIDIA probe routine to fail.

  Because the SupremeRAID controller relies on proper PCIe enumeration, this failure can also prevent the RAID card from being detected.

  ---

  Resolution

  Disable PCI resource reallocation during boot by adding the following parameter to the GRUB configuration:

  pci=realloc=off
  

  This preserves the original BAR assignments made by the system firmware (BIOS/UEFI) and prevents the kernel from overwriting them.

  ---

  Steps for Ubuntu / Debian

  1. Edit the GRUB configuration:

     sudo nano /etc/default/grub
     

  2. Add pci=realloc=off to the GRUB_CMDLINE_LINUX line (append to existing parameters):

     GRUB_CMDLINE_LINUX="quiet splash pci=realloc=off"
     

  3. Update GRUB:

     sudo update-grub
     

  4. Reboot the system:

     sudo reboot
     

  ---

  Steps for RHEL / Rocky / AlmaLinux

  1. Edit the GRUB configuration:

     sudo nano /etc/default/grub
     

  2. Add pci=realloc=off to the kernel command line:

     GRUB_CMDLINE_LINUX="... pci=realloc=off"
     

  3. Rebuild the GRUB configuration (UEFI examples):

     sudo grub2-mkconfig -o /boot/efi/EFI/redhat/grub.cfg
     # or for Rocky / Alma:
     sudo grub2-mkconfig -o /boot/efi/EFI/rocky/grub.cfg
     sudo grub2-mkconfig -o /boot/efi/EFI/almalinux/grub.cfg
     

     (For BIOS mode, use /boot/grub2/grub.cfg.)

  4. Reboot:

     sudo reboot
     

  ---

  Verification

  After rebooting:

  1. Check dmesg to confirm the BAR0 error no longer appears.

  2. Run nvidia-smi to ensure all GPUs are detected without probe failures.

  3. Verify SupremeRAID controller detection:

     nvidia-smi
     

  4. If the issue persists, also verify:

     • Above 4G Decoding and Resizable BAR are enabled in BIOS.

     • PCIe slots and switch topology are correctly configured.

     • IOMMU settings are not causing conflicts.

  ---

  Notes

  • The pci=realloc=off parameter prevents the kernel from reallocating PCI resources. This is generally safe for systems where firmware already assigns resources correctly (typical in GPU + RAID configurations).

  • On systems with many hot-plug devices or dynamic PCI topologies, this setting may limit some flexibility.

  • If you add or move devices later, re-evaluate BIOS and kernel parameters accordingly.