From: Xiubo Li <xiubli@redhat.com> There has one race case for ceph's rbd-nbd tool. When do mapping it may fail with EBUSY from ioctl(nbd, NBD_DO_IT), but actually the nbd device has already unmaped. It dues to if just after the wake_up(), the recv_work() is scheduled out and defers calling the nbd_config_put(), though the map process has exited the "nbd->recv_task" is not cleared. Signed-off-by: Xiubo Li <xiubli@redhat.com> --- drivers/block/nbd.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c index edf8b632e3d2..f46e26c9d9b3 100644 --- a/drivers/block/nbd.c +++ b/drivers/block/nbd.c @@ -801,9 +801,9 @@ static void recv_work(struct work_struct *work) if (likely(!blk_should_fake_timeout(rq->q))) blk_mq_complete_request(rq); } + nbd_config_put(nbd); atomic_dec(&config->recv_threads); wake_up(&config->recv_wq); - nbd_config_put(nbd); kfree(args); } -- 2.18.4
On 10/13/20 10:45 PM, xiubli@redhat.com wrote:
> From: Xiubo Li <xiubli@redhat.com>
>
> There has one race case for ceph's rbd-nbd tool. When do mapping
> it may fail with EBUSY from ioctl(nbd, NBD_DO_IT), but actually
> the nbd device has already unmaped.
>
> It dues to if just after the wake_up(), the recv_work() is scheduled
> out and defers calling the nbd_config_put(), though the map process
> has exited the "nbd->recv_task" is not cleared.
>
> Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Thanks,
Josef
On 10/13/20 8:45 PM, xiubli@redhat.com wrote:
> From: Xiubo Li <xiubli@redhat.com>
>
> There has one race case for ceph's rbd-nbd tool. When do mapping
> it may fail with EBUSY from ioctl(nbd, NBD_DO_IT), but actually
> the nbd device has already unmaped.
>
> It dues to if just after the wake_up(), the recv_work() is scheduled
> out and defers calling the nbd_config_put(), though the map process
> has exited the "nbd->recv_task" is not cleared.
Applied, thanks.
--
Jens Axboe