On Mon, 13 Apr 2026 15:36:03 +0800
"Jinhui Guo" <[email protected]> wrote:
> virtiovf_read_device_context_chunk() takes migf->list_lock with
> spin_lock() but releases it with spin_unlock_irq(). This mismatch
> can incorrectly enable interrupts if they were already disabled
> when the lock was acquired, leading to unbalanced IRQ state.
>
> Fix by using spin_lock_irq() to match spin_unlock_irq().
>
> Fixes: 0bbc82e4ec79 ("vfio/virtio: Add support for the basic live migration
> functionality")
> Cc: [email protected]
> Signed-off-by: Jinhui Guo <[email protected]>
> ---
> drivers/vfio/pci/virtio/migrate.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/vfio/pci/virtio/migrate.c
> b/drivers/vfio/pci/virtio/migrate.c
> index 35fa2d6ed611..9fc24788fc04 100644
> --- a/drivers/vfio/pci/virtio/migrate.c
> +++ b/drivers/vfio/pci/virtio/migrate.c
> @@ -621,7 +621,7 @@ virtiovf_read_device_context_chunk(struct
> virtiovf_migration_file *migf,
>
> buf->start_pos = buf->migf->max_pos;
> migf->max_pos += buf->length;
> - spin_lock(&migf->list_lock);
> + spin_lock_irq(&migf->list_lock);
> list_add_tail(&buf->buf_elm, &migf->buf_list);
> spin_unlock_irq(&migf->list_lock);
> return 0;
Yes, that fixes the bug, but why are we using a spinlock-irq here in
the first place? I think this just copied the mlx5 vfio-pci variant
driver, which does make use of their list_lock under hardirq context,
but no such use case exists in this virtio driver.
A more complete fix would be to to convert list_lock to a mutex.
Thanks,
Alex
