On Tue, Jun 24, 2025 at 02:56:22PM -0400, Stefan Hajnoczi wrote:
> On Mon, Jun 02, 2025 at 02:44:33AM +0000, Parav Pandit wrote:
> > When the PCI device is surprise removed, requests may not complete
> > the device as the VQ is marked as broken. Due to this, the disk
> > deletion hangs.
>
> There are loops in the core virtio driver code that expect device
> register reads to eventually return 0:
> drivers/virtio/virtio_pci_modern.c:vp_reset()
> drivers/virtio/virtio_pci_modern_dev.c:vp_modern_set_queue_reset()
>
> Is there a hang if these loops are hit when a device has been surprise
> removed?
Probably, yes. We can't really check broken status there though
I think - reset is called with a broken device normally.
> I'm trying to understand whether surprise removal is fully
> supported or whether this patch is one step in that direction.
I think it's a step.
> Apart from that, I'm happy with the virtio_blk.c aspects of the patch:
> Reviewed-by: Stefan Hajnoczi <stefa...@redhat.com>
>
> >
> > Fix it by aborting the requests when the VQ is broken.
> >
> > With this fix now fio completes swiftly.
> > An alternative of IO timeout has been considered, however
> > when the driver knows about unresponsive block device, swiftly clearing
> > them enables users and upper layers to react quickly.
> >
> > Verified with multiple device unplug iterations with pending requests in
> > virtio used ring and some pending with the device.
> >
> > Fixes: 43bb40c5b926 ("virtio_pci: Support surprise removal of virtio pci
> > device")
> > Cc: sta...@vger.kernel.org
> > Reported-by: Li RongQing <lirongq...@baidu.com>
> > Closes:
> > https://lore.kernel.org/virtualization/c45dd68698cd47238c55fb73ca9b4...@baidu.com/
> > Reviewed-by: Max Gurtovoy <mgurto...@nvidia.com>
> > Reviewed-by: Israel Rukshin <isra...@nvidia.com>
> > Signed-off-by: Parav Pandit <pa...@nvidia.com>
> >
> > ---
> > v4->v5:
> > - fixed comment style where comment to start with one empty line at start
> > - Addressed comments from Alok
> > - fixed typo in broken vq check
> > v3->v4:
> > - Addressed comments from Michael
> > - renamed virtblk_request_cancel() to
> > virtblk_complete_request_with_ioerr()
> > - Added comments for virtblk_complete_request_with_ioerr()
> > - Renamed virtblk_broken_device_cleanup() to
> > virtblk_cleanup_broken_device()
> > - Added comments for virtblk_cleanup_broken_device()
> > - Moved the broken vq check in virtblk_remove()
> > - Fixed comment style to have first empty line
> > - replaced freezed to frozen
> > - Fixed comments rephrased
> >
> > v2->v3:
> > - Addressed comments from Michael
> > - updated comment for synchronizing with callbacks
> >
> > v1->v2:
> > - Addressed comments from Stephan
> > - fixed spelling to 'waiting'
> > - Addressed comments from Michael
> > - Dropped checking broken vq from queue_rq() and queue_rqs()
> > because it is checked in lower layer routines in virtio core
> >
> > v0->v1:
> > - Fixed comments from Stefan to rename a cleanup function
> > - Improved logic for handling any outstanding requests
> > in bio layer
> > - improved cancel callback to sync with ongoing done()
> > ---
> > drivers/block/virtio_blk.c | 95 ++++++++++++++++++++++++++++++++++++++
> > 1 file changed, 95 insertions(+)
> >
> > diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
> > index 7cffea01d868..c5e383c0ac48 100644
> > --- a/drivers/block/virtio_blk.c
> > +++ b/drivers/block/virtio_blk.c
> > @@ -1554,6 +1554,98 @@ static int virtblk_probe(struct virtio_device *vdev)
> > return err;
> > }
> >
> > +/*
> > + * If the vq is broken, device will not complete requests.
> > + * So we do it for the device.
> > + */
> > +static bool virtblk_complete_request_with_ioerr(struct request *rq, void
> > *data)
> > +{
> > + struct virtblk_req *vbr = blk_mq_rq_to_pdu(rq);
> > + struct virtio_blk *vblk = data;
> > + struct virtio_blk_vq *vq;
> > + unsigned long flags;
> > +
> > + vq = &vblk->vqs[rq->mq_hctx->queue_num];
> > +
> > + spin_lock_irqsave(&vq->lock, flags);
> > +
> > + vbr->in_hdr.status = VIRTIO_BLK_S_IOERR;
> > + if (blk_mq_request_started(rq) && !blk_mq_request_completed(rq))
> > + blk_mq_complete_request(rq);
> > +
> > + spin_unlock_irqrestore(&vq->lock, flags);
> > + return true;
> > +}
> > +
> > +/*
> > + * If the device is broken, it will not use any buffers and waiting
> > + * for that to happen is pointless. We'll do the cleanup in the driver,
> > + * completing all requests for the device.
> > + */
> > +static void virtblk_cleanup_broken_device(struct virtio_blk *vblk)
> > +{
> > + struct request_queue *q = vblk->disk->queue;
> > +
> > + /*
> > + * Start freezing the queue, so that new requests keeps waiting at the
> > + * door of bio_queue_enter(). We cannot fully freeze the queue because
> > + * frozen queue is an empty queue and there are pending requests, so
> > + * only start freezing it.
> > + */
> > + blk_freeze_queue_start(q);
> > +
> > + /*
> > + * When quiescing completes, all ongoing dispatches have completed
> > + * and no new dispatch will happen towards the driver.
> > + */
> > + blk_mq_quiesce_queue(q);
> > +
> > + /*
> > + * Synchronize with any ongoing VQ callbacks that may have started
> > + * before the VQs were marked as broken. Any outstanding requests
> > + * will be completed by virtblk_complete_request_with_ioerr().
> > + */
> > + virtio_synchronize_cbs(vblk->vdev);
> > +
> > + /*
> > + * At this point, no new requests can enter the queue_rq() and
> > + * completion routine will not complete any new requests either for the
> > + * broken vq. Hence, it is safe to cancel all requests which are
> > + * started.
> > + */
> > + blk_mq_tagset_busy_iter(&vblk->tag_set,
> > + virtblk_complete_request_with_ioerr, vblk);
> > + blk_mq_tagset_wait_completed_request(&vblk->tag_set);
> > +
> > + /*
> > + * All pending requests are cleaned up. Time to resume so that disk
> > + * deletion can be smooth. Start the HW queues so that when queue is
> > + * unquiesced requests can again enter the driver.
> > + */
> > + blk_mq_start_stopped_hw_queues(q, true);
> > +
> > + /*
> > + * Unquiescing will trigger dispatching any pending requests to the
> > + * driver which has crossed bio_queue_enter() to the driver.
> > + */
> > + blk_mq_unquiesce_queue(q);
> > +
> > + /*
> > + * Wait for all pending dispatches to terminate which may have been
> > + * initiated after unquiescing.
> > + */
> > + blk_mq_freeze_queue_wait(q);
> > +
> > + /*
> > + * Mark the disk dead so that once we unfreeze the queue, requests
> > + * waiting at the door of bio_queue_enter() can be aborted right away.
> > + */
> > + blk_mark_disk_dead(vblk->disk);
> > +
> > + /* Unfreeze the queue so that any waiting requests will be aborted. */
> > + blk_mq_unfreeze_queue_nomemrestore(q);
> > +}
> > +
> > static void virtblk_remove(struct virtio_device *vdev)
> > {
> > struct virtio_blk *vblk = vdev->priv;
> > @@ -1561,6 +1653,9 @@ static void virtblk_remove(struct virtio_device *vdev)
> > /* Make sure no work handler is accessing the device. */
> > flush_work(&vblk->config_work);
> >
> > + if (virtqueue_is_broken(vblk->vqs[0].vq))
> > + virtblk_cleanup_broken_device(vblk);
> > +
> > del_gendisk(vblk->disk);
> > blk_mq_free_tag_set(&vblk->tag_set);
> >
> > --
> > 2.34.1
> >
