Skip to content

Conversation

@sdimitro
Copy link

@sdimitro sdimitro commented Dec 1, 2025

Relevant Tickets:

  • NVIDIA ID: 5610224
  • CoreWeave ID: 789

We first hit this in the 570 releases and it seems like it hasn't been fixed in the 580 ones either.

Control flow that triggers the issue:

uvm_ioctl
  -> uvm_api_register_gpu
    -> uvm_va_space_register_gpu
      -> uvm_gpu_retain_by_uuid
        -> gpu_retain_by_uuid_locked
          -> add_gpu
            -> alloc_parent_gpu
              ** allocated parent_gpu but didn't initialize parent_gpu->pci_dev yet **
            <- alloc_parent_gpu
            ...
            -> init_parent_gpu
              ** returns error before initializing parent_gpu->pci_dev **
            <- ini_parent_gpu
            -> remove_gpu
              ...
              -> deinit_gpu
                ...
                -> uvm_pmm_gpu_device_p2p_deinit
                  -> uvm_device_p2p_static_bar
                    ** dereferences gpu->parent->pci_dev and crashes the system **

…rror path

NVIDIA ID: 5610224
CoreWeave ID: 789
@CLAassistant
Copy link

CLAassistant commented Dec 1, 2025

CLA assistant check
All committers have signed the CLA.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants