On Wed, May 27, 2020 at 8:57 AM Dongli Zhang <dongli.zh...@oracle.com> wrote:
>
> There may be a race between nvme_reap_pending_cqes() and nvme_poll(), e.g.,
> when doing live reset while polling the nvme device.
>
> CPU X CPU Y
> nvme_poll()
> nvme_dev_disable()
> -> nvme_stop_queues()
> -> nvme_suspend_io_queues()
> -> nvme_suspend_queue()
> -> spin_lock(&nvmeq->cq_poll_lock);
> -> nvme_reap_pending_cqes()
> -> nvme_process_cq() -> nvme_process_cq()
>
> In the above scenario, the nvme_process_cq() for the same queue may be
> running on both CPU X and CPU Y concurrently.
>
> It is much more easier to reproduce the issue when CONFIG_PREEMPT is
> enabled in kernel. When CONFIG_PREEMPT is disabled, it would take longer
> time for nvme_stop_queues()-->blk_mq_quiesce_queue() to wait for grace
> period.
>
> This patch protects nvme_process_cq() with nvmeq->cq_poll_lock in
> nvme_reap_pending_cqes().
>
> Signed-off-by: Dongli Zhang <dongli.zh...@oracle.com>
> ---
> drivers/nvme/host/pci.c | 11 +++++++----
> 1 file changed, 7 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
> index 3726dc780d15..cc46e250fcac 100644
> --- a/drivers/nvme/host/pci.c
> +++ b/drivers/nvme/host/pci.c
> @@ -1382,16 +1382,19 @@ static void nvme_disable_admin_queue(struct nvme_dev
> *dev, bool shutdown)
>
> /*
> * Called only on a device that has been disabled and after all other threads
> - * that can check this device's completion queues have synced. This is the
> - * last chance for the driver to see a natural completion before
> - * nvme_cancel_request() terminates all incomplete requests.
> + * that can check this device's completion queues have synced, except
> + * nvme_poll(). This is the last chance for the driver to see a natural
> + * completion before nvme_cancel_request() terminates all incomplete
> requests.
> */
> static void nvme_reap_pending_cqes(struct nvme_dev *dev)
> {
> int i;
>
> - for (i = dev->ctrl.queue_count - 1; i > 0; i--)
> + for (i = dev->ctrl.queue_count - 1; i > 0; i--) {
> + spin_lock(&dev->queues[i].cq_poll_lock);
> nvme_process_cq(&dev->queues[i]);
> + spin_unlock(&dev->queues[i].cq_poll_lock);
> + }
> }
Looks a real race, and the fix is fine:
Reviewed-by: Ming Lei <ming....@redhat.com>
thanks,
Ming Lei
