RDMA/mlx5: Fix crash when unbind multiport slave

Fix the below crash when deleting a slave from the unaffiliated list
twice. First time when the slave is bound to the master and the second
when the slave is unloaded.

Fix it by checking if slave is unaffiliated (doesn't have ib device)
before removing from the list.

  RIP: 0010:mlx5r_mp_remove+0x4e/0xa0 [mlx5_ib]
  Call Trace:
   auxiliary_bus_remove+0x18/0x30
   __device_release_driver+0x177/x220
   device_release_driver+0x24/0x30
   bus_remove_device+0xd8/0x140
   device_del+0x18a/0x3e0
   mlx5_rescan_drivers_locked+0xa9/0x210 [mlx5_core]
   mlx5_unregister_device+0x34/0x60 [mlx5_core]
   mlx5_uninit_one+0x32/0x100 [mlx5_core]
   remove_one+0x6e/0xe0 [mlx5_core]
   pci_device_remove+0x36/0xa0
   __device_release_driver+0x177/0x220
   device_driver_detach+0x3c/0xa0
   unbind_store+0x113/0x130
   kernfs_fop_write_iter+0x110/0x1a0
   new_sync_write+0x116/0x1a0
   vfs_write+0x1ba/0x260
   ksys_write+0x5f/0xe0
   do_syscall_64+0x3d/0x90
   entry_SYSCALL_64_after_hwframe+0x44/0xae

Fixes: 93f8244431 ("RDMA/mlx5: Convert mlx5_ib to use auxiliary bus")
Link: https://lore.kernel.org/r/17ec98989b0ba88f7adfbad68eb20bce8d567b44.1628587493.git.leonro@nvidia.com
Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
This commit is contained in:
Maor Gottlieb 2021-08-10 12:25:11 +03:00 committed by Jason Gunthorpe
parent 7c60610d47
commit da78fe5fb3

View File

@ -4454,7 +4454,8 @@ static void mlx5r_mp_remove(struct auxiliary_device *adev)
mutex_lock(&mlx5_ib_multiport_mutex);
if (mpi->ibdev)
mlx5_ib_unbind_slave_port(mpi->ibdev, mpi);
list_del(&mpi->list);
else
list_del(&mpi->list);
mutex_unlock(&mlx5_ib_multiport_mutex);
kfree(mpi);
}