The vhost-scsi driver currently does not handle any control queue
operations. In particular, vhost_scsi_ctl_handle_kick, merely prints out
a debug message but does nothing else. This can cause guest VMs to hang.
As part of SCSI recovery from an error, e.g., an I/O timeout, the SCSI
midlayer attempts to abort the failed operation. The SCSI virtio driver
translates the abort to a SCSI TMF request that gets put on the control
queue (virtscsi_abort -> virtscsi_tmf). The SCSI virtio driver then
waits indefinitely for this request to be completed, but it never will
because vhost-scsi never responds to that request.
To avoid a hang, always respond to control queue operations; explicitly
reject TMF requests, and return a no-op response to event requests.
Signed-off-by: Bijan Mottahedeh <bijan.mottahedeh@xxxxxxxxxx>
---
drivers/vhost/scsi.c | 190 +++++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 190 insertions(+)
diff --git a/drivers/vhost/scsi.c b/drivers/vhost/scsi.c
index c24bb69..faf0dcf 100644
--- a/drivers/vhost/scsi.c
+++ b/drivers/vhost/scsi.c
@@ -1048,9 +1048,199 @@ static void vhost_scsi_submission_work(struct work_struct *work)
mutex_unlock(&vq->mutex);
}
+static void
+vhost_scsi_send_tmf_resp(struct vhost_scsi *vs,
+ struct vhost_virtqueue *vq,
+ int head, unsigned int out)
+{
+ struct virtio_scsi_ctrl_tmf_resp __user *resp;
+ struct virtio_scsi_ctrl_tmf_resp rsp;
+ int ret;
+
+ pr_debug("%s\n", __func__);
+ memset(&rsp, 0, sizeof(rsp));
+ rsp.response = VIRTIO_SCSI_S_FUNCTION_REJECTED;
+ resp = vq->iov[out].iov_base;
+ ret = __copy_to_user(resp, &rsp, sizeof(rsp));
+ if (!ret)
+ vhost_add_used_and_signal(&vs->dev, vq, head, 0);
+ else
+ pr_err("Faulted on virtio_scsi_ctrl_tmf_resp\n");
+}
+
+static void
+vhost_scsi_send_an_resp(struct vhost_scsi *vs,
+ struct vhost_virtqueue *vq,
+ int head, unsigned int out)
+{
+ struct virtio_scsi_ctrl_an_resp __user *resp;
+ struct virtio_scsi_ctrl_an_resp rsp;
+ int ret;
+
+ pr_debug("%s\n", __func__);
+ memset(&rsp, 0, sizeof(rsp)); /* event_actual = 0 */
+ rsp.response = VIRTIO_SCSI_S_OK;
+ resp = vq->iov[out].iov_base;
+ ret = __copy_to_user(resp, &rsp, sizeof(rsp));
+ if (!ret)
+ vhost_add_used_and_signal(&vs->dev, vq, head, 0);
+ else
+ pr_err("Faulted on virtio_scsi_ctrl_an_resp\n");
+}
+
+static void
+vhost_scsi_ctl_handle_vq(struct vhost_scsi *vs, struct vhost_virtqueue *vq)
+{
+ union {
+ __virtio32 type;
+ struct virtio_scsi_ctrl_an_req an;
+ struct virtio_scsi_ctrl_tmf_req tmf;
+ } v_req;
+ struct iov_iter out_iter;
+ unsigned int out = 0, in = 0;
+ int head;
+ size_t req_size, rsp_size, typ_size;
+ size_t out_size, in_size;
+ u8 *lunp;
+ void *req;
+
+ mutex_lock(&vq->mutex);
+ /*
+ * We can handle the vq only after the endpoint is setup by calling the
+ * VHOST_SCSI_SET_ENDPOINT ioctl.
+ */
+ if (!vq->private_data)
+ goto out;
+
+ vhost_disable_notify(&vs->dev, vq);
+
+ for (;;) {
+ head = vhost_get_vq_desc(vq, vq->iov,
+ ARRAY_SIZE(vq->iov), &out, &in,
+ NULL, NULL);
+ pr_debug("vhost_get_vq_desc: head: %d, out: %u in: %u\n",
+ head, out, in);
+ /* On error, stop handling until the next kick. */
+ if (unlikely(head < 0))
+ break;
+ /* Nothing new? Wait for eventfd to tell us they refilled. */
+ if (head == vq->num) {
+ if (unlikely(vhost_enable_notify(&vs->dev, vq))) {
+ vhost_disable_notify(&vs->dev, vq);
+ continue;
+ }
+ break;
+ }
+
+ /*
+ * Get the size of request and response buffers.
+ */
+ out_size = iov_length(vq->iov, out);
+ in_size = iov_length(&vq->iov[out], in);
+
+ /*
+ * Copy over the virtio-scsi request header, which for a
+ * ANY_LAYOUT enabled guest may span multiple iovecs, or a
+ * single iovec may contain both the header + outgoing
+ * WRITE payloads.
+ *
+ * copy_from_iter() will advance out_iter, so that it will
+ * point at the start of the outgoing WRITE payload, if
+ * DMA_TO_DEVICE is set.
+ */
+ iov_iter_init(&out_iter, WRITE, vq->iov, out, out_size);
+
+ req = &v_req.type;
+ typ_size = sizeof(v_req.type);
+
+ if (unlikely(!copy_from_iter_full(req, typ_size, &out_iter))) {
+ vq_err(vq, "Faulted on copy_from_iter tmf type\n");
+ /*
+ * The size of the response buffer varies based on
+ * the request type and must be validated against it.
+ * Since the request type is not known, don't send
+ * a response.
+ */
+ continue;
+ }
+
+ switch (v_req.type) {
+ case VIRTIO_SCSI_T_TMF:
+ req = &v_req.tmf;
+ lunp = &v_req.tmf.lun[0];
+ req_size = sizeof(struct virtio_scsi_ctrl_tmf_req);
+ rsp_size = sizeof(struct virtio_scsi_ctrl_tmf_resp);
+ break;
+ case VIRTIO_SCSI_T_AN_QUERY:
+ case VIRTIO_SCSI_T_AN_SUBSCRIBE:
+ req = &v_req.an;
+ lunp = &v_req.an.lun[0];
+ req_size = sizeof(struct virtio_scsi_ctrl_an_req);
+ rsp_size = sizeof(struct virtio_scsi_ctrl_an_resp);
+ break;
+ default:
+ vq_err(vq, "Unknown control request %d", v_req.type);
+ continue;
+ }
+
+ /*
+ * Check for a sane response buffer so we can report early
+ * errors back to the guest.
+ */
+ if (unlikely(in_size < rsp_size)) {
+ vq_err(vq,
+ "Resp buf too small, need min %zu bytes got %zu",
+ rsp_size, in_size);
+ /*
+ * Notifications are disabled at this point;
+ * continue so they can be eventually enabled
+ * when processing terminates.
+ */
+ continue;
+ }
+
+ if (unlikely(out_size < req_size)) {
+ vq_err(vq,
+ "Req buf too small, need min %zu bytes got %zu",
+ req_size, out_size);
+ vhost_scsi_send_bad_target(vs, vq, head, out);
+ continue;
+ }
+
+ req += typ_size;
+ req_size -= typ_size;
+
+ if (unlikely(!copy_from_iter_full(req, req_size, &out_iter))) {
+ vq_err(vq, "Faulted on copy_from_iter\n");
+ vhost_scsi_send_bad_target(vs, vq, head, out);
+ continue;
+ }
+
+ /* virtio-scsi spec requires byte 0 of the lun to be 1 */
+ if (unlikely(*lunp != 1)) {
+ vq_err(vq, "Illegal virtio-scsi lun: %u\n", *lunp);
+ vhost_scsi_send_bad_target(vs, vq, head, out);
+ continue;
+ }
+
+ if (v_req.type == VIRTIO_SCSI_T_TMF) {
+ pr_debug("%s tmf %d\n", __func__, v_req.tmf.subtype);
+ vhost_scsi_send_tmf_resp(vs, vq, head, out);
+ } else
+ vhost_scsi_send_an_resp(vs, vq, head, out);
+ }
+out:
+ mutex_unlock(&vq->mutex);
+}
+
static void vhost_scsi_ctl_handle_kick(struct vhost_work *work)
{
+ struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
+ poll.work);
+ struct vhost_scsi *vs = container_of(vq->dev, struct vhost_scsi, dev);
+
pr_debug("%s: The handling func for control queue.\n", __func__);
+ vhost_scsi_ctl_handle_vq(vs, vq);
}
static void
--
1.8.3.1
