vhost: implemented indirect descriptors

Every descriptor may now contain
a separate descriptor table.

This increases the maximum number
of concurrent IO in a virtqueue
from about (vq->size / 3) to
(vq->size * SPDK_VHOST_IOVS_MAX).

Knowing that SPDK_VHOST_IOVS_MAX
is 128, and assuming that vq->size
is 128 (current hardcoded limit
for QEMU), this gives us over 16k
concurrent possible iovectors
on the fly

Change-Id: I0853d80f6f90d53f8774231972b430a5bf05460e
Signed-off-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-on: https://review.gerrithub.io/373703
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>

lib/vhost/vhost.c

@@ -114,15 +114,30 @@ spdk_vhost_vq_avail_ring_get(struct rte_vhost_vring *vq, uint16_t *reqs, uint16_
 	return count;
 }
 
+static bool
+spdk_vhost_vring_desc_is_indirect(struct vring_desc *cur_desc)
+{
+	return !!(cur_desc->flags & VRING_DESC_F_INDIRECT);
+}
+
 int
-spdk_vhost_vq_get_desc(struct rte_vhost_vring *vq, uint16_t req_idx, struct vring_desc **desc,
-		       struct vring_desc **desc_table, uint32_t *desc_table_size)
+spdk_vhost_vq_get_desc(struct spdk_vhost_dev *vdev, struct rte_vhost_vring *vq, uint16_t req_idx,
+		       struct vring_desc **desc, struct vring_desc **desc_table, uint32_t *desc_table_size)
 {
 	if (spdk_unlikely(req_idx >= vq->size)) {
 		return -1;
 	}
 
 	*desc = &vq->desc[req_idx];
+
+	if (spdk_vhost_vring_desc_is_indirect(*desc)) {
+		assert(spdk_vhost_dev_has_feature(vdev, VIRTIO_RING_F_INDIRECT_DESC));
+		*desc_table = spdk_vhost_gpa_to_vva(vdev, (*desc)->addr);
+		*desc_table_size = (*desc)->len / sizeof(**desc);
+		*desc = *desc_table;
+		return 0;
+	}
+
 	*desc_table = vq->desc;
 	*desc_table_size = vq->size;
 

lib/vhost/vhost_blk.c

@@ -130,7 +130,7 @@ blk_iovs_setup(struct spdk_vhost_dev *vdev, struct rte_vhost_vring *vq, uint16_t
 	uint32_t desc_table_size, len = 0;
 	int rc;
 
-	rc = spdk_vhost_vq_get_desc(vq, req_idx, &desc, &desc_table, &desc_table_size);
+	rc = spdk_vhost_vq_get_desc(vdev, vq, req_idx, &desc, &desc_table, &desc_table_size);
 	if (rc != 0) {
 		SPDK_ERRLOG("%s: Invalid descriptor at index %"PRIu16".\n", vdev->name, req_idx);
 		return -1;

lib/vhost/vhost_internal.h

@@ -74,8 +74,7 @@
 	(1ULL << VIRTIO_RING_F_INDIRECT_DESC))
 
 #define SPDK_VHOST_DISABLED_FEATURES ((1ULL << VHOST_F_LOG_ALL) | \
-	(1ULL << VIRTIO_RING_F_EVENT_IDX) | \
-	(1ULL << VIRTIO_RING_F_INDIRECT_DESC))
+	(1ULL << VIRTIO_RING_F_EVENT_IDX))
 
 enum spdk_vhost_dev_type {
 	SPDK_VHOST_DEV_T_SCSI,
@@ -133,6 +132,7 @@ bool spdk_vhost_vq_should_notify(struct spdk_vhost_dev *vdev, struct rte_vhost_v
  * The descriptor will provide access to the entire descriptor
  * chain. The subsequent descriptors are accesible via
  * \c spdk_vhost_vring_desc_get_next.
+ * \param vdev vhost device
  * \param vq virtqueue
  * \param req_idx descriptor index
  * \param desc pointer to be set to the descriptor
@@ -144,8 +144,9 @@ bool spdk_vhost_vq_should_notify(struct spdk_vhost_dev *vdev, struct rte_vhost_v
  * \return 0 on success, -1 if given index is invalid.
  * If -1 is returned, the params won't be changed.
  */
-int spdk_vhost_vq_get_desc(struct rte_vhost_vring *vq, uint16_t req_idx, struct vring_desc **desc,
-			   struct vring_desc **desc_table, uint32_t *desc_table_size);
+int spdk_vhost_vq_get_desc(struct spdk_vhost_dev *vdev, struct rte_vhost_vring *vq,
+			   uint16_t req_idx,
+			   struct vring_desc **desc, struct vring_desc **desc_table, uint32_t *desc_table_size);
 void spdk_vhost_vq_used_ring_enqueue(struct spdk_vhost_dev *vdev, struct rte_vhost_vring *vq,
 				     uint16_t id, uint32_t len);
 

lib/vhost/vhost_scsi.c

@@ -190,7 +190,7 @@ eventq_enqueue(struct spdk_vhost_scsi_dev *svdev, unsigned scsi_dev_num, uint32_
 		return;
 	}
 
-	rc = spdk_vhost_vq_get_desc(vq, req, &desc, &desc_table, &desc_table_size);
+	rc = spdk_vhost_vq_get_desc(&svdev->vdev, vq, req, &desc, &desc_table, &desc_table_size);
 	if (rc != 0 || desc->len < sizeof(*desc_ev)) {
 		SPDK_ERRLOG("Controller %s: Invalid eventq descriptor at index %"PRIu16".\n",
 			    svdev->vdev.name, req);
@@ -314,6 +314,7 @@ spdk_vhost_scsi_task_init_target(struct spdk_vhost_scsi_task *task, const __u8 *
 static void
 process_ctrl_request(struct spdk_vhost_scsi_task *task)
 {
+	struct spdk_vhost_dev *vdev = &task->svdev->vdev;
 	struct vring_desc *desc, *desc_table;
 	struct virtio_scsi_ctrl_tmf_req *ctrl_req;
 	struct virtio_scsi_ctrl_an_resp *an_resp;
@@ -322,14 +323,14 @@ process_ctrl_request(struct spdk_vhost_scsi_task *task)
 
 	spdk_scsi_task_construct(&task->scsi, spdk_vhost_scsi_task_mgmt_cpl, spdk_vhost_scsi_task_free_cb,
 				 NULL);
-	rc = spdk_vhost_vq_get_desc(task->vq, task->req_idx, &desc, &desc_table, &desc_table_size);
+	rc = spdk_vhost_vq_get_desc(vdev, task->vq, task->req_idx, &desc, &desc_table, &desc_table_size);
 	if (spdk_unlikely(rc != 0)) {
 		SPDK_ERRLOG("%s: Invalid controlq descriptor at index %d.\n",
-			    task->svdev->vdev.name, task->req_idx);
+			    vdev->name, task->req_idx);
 		goto out;
 	}
 
-	ctrl_req = spdk_vhost_gpa_to_vva(&task->svdev->vdev, desc->addr);
+	ctrl_req = spdk_vhost_gpa_to_vva(vdev, desc->addr);
 
 	SPDK_DEBUGLOG(SPDK_TRACE_VHOST_SCSI_QUEUE,
 		      "Processing controlq descriptor: desc %d/%p, desc_addr %p, len %d, flags %d, last_used_idx %d; kickfd %d; size %d\n",
@@ -343,14 +344,14 @@ process_ctrl_request(struct spdk_vhost_scsi_task *task)
 	spdk_vhost_vring_desc_get_next(&desc, desc_table, desc_table_size);
 	if (spdk_unlikely(desc == NULL)) {
 		SPDK_ERRLOG("%s: No response descriptor for controlq request %d.\n",
-			    task->svdev->vdev.name, task->req_idx);
+			    vdev->name, task->req_idx);
 		goto out;
 	}
 
 	/* Process the TMF request */
 	switch (ctrl_req->type) {
 	case VIRTIO_SCSI_T_TMF:
-		task->tmf_resp = spdk_vhost_gpa_to_vva(&task->svdev->vdev, desc->addr);
+		task->tmf_resp = spdk_vhost_gpa_to_vva(vdev, desc->addr);
 
 		/* Check if we are processing a valid request */
 		if (task->scsi_dev == NULL) {
@@ -374,7 +375,7 @@ process_ctrl_request(struct spdk_vhost_scsi_task *task)
 		break;
 	case VIRTIO_SCSI_T_AN_QUERY:
 	case VIRTIO_SCSI_T_AN_SUBSCRIBE: {
-		an_resp = spdk_vhost_gpa_to_vva(&task->svdev->vdev, desc->addr);
+		an_resp = spdk_vhost_gpa_to_vva(vdev, desc->addr);
 		an_resp->response = VIRTIO_SCSI_S_ABORTED;
 		break;
 	}
@@ -384,7 +385,7 @@ process_ctrl_request(struct spdk_vhost_scsi_task *task)
 	}
 
 out:
-	spdk_vhost_vq_used_ring_enqueue(&task->svdev->vdev, task->vq, task->req_idx, 0);
+	spdk_vhost_vq_used_ring_enqueue(vdev, task->vq, task->req_idx, 0);
 	spdk_vhost_scsi_task_put(task);
 }
 
@@ -405,7 +406,7 @@ task_data_setup(struct spdk_vhost_scsi_task *task,
 	uint32_t desc_table_len, len = 0;
 	int rc;
 
-	rc = spdk_vhost_vq_get_desc(task->vq, task->req_idx, &desc, &desc_table, &desc_table_len);
+	rc = spdk_vhost_vq_get_desc(vdev, task->vq, task->req_idx, &desc, &desc_table, &desc_table_len);
 	/* First descriptor must be readable */
 	if (rc != 0 || spdk_unlikely(spdk_vhost_vring_desc_is_wr(desc))) {
 		SPDK_WARNLOG("Invalid first (request) descriptor.\n");

test/unit/lib/vhost/vhost_blk.c/vhost_blk_ut.c

@@ -63,8 +63,9 @@ DEFINE_STUB(spdk_ring_enqueue, size_t, (struct spdk_ring *ring, void **objs, siz
 DEFINE_STUB(spdk_ring_dequeue, size_t, (struct spdk_ring *ring, void **objs, size_t count), 0);
 DEFINE_STUB_V(spdk_vhost_vq_used_ring_enqueue, (struct spdk_vhost_dev *vdev,
 		struct rte_vhost_vring *vq, uint16_t id, uint32_t len));
-DEFINE_STUB(spdk_vhost_vq_get_desc, int, (struct rte_vhost_vring *vq, uint16_t req_idx,
-		struct vring_desc **desc, struct vring_desc **desc_table, uint32_t *desc_table_size), 0);
+DEFINE_STUB(spdk_vhost_vq_get_desc, int, (struct spdk_vhost_dev *vdev, struct rte_vhost_vring *vq,
+		uint16_t req_idx, struct vring_desc **desc, struct vring_desc **desc_table,
+		uint32_t *desc_table_size), 0);
 DEFINE_STUB(spdk_vhost_vring_desc_is_wr, bool, (struct vring_desc *cur_desc), false);
 DEFINE_STUB(spdk_vhost_vring_desc_to_iov, int, (struct spdk_vhost_dev *vdev, struct iovec *iov,
 		uint16_t *iov_index, const struct vring_desc *desc), 0);

test/unit/lib/vhost/vhost_scsi.c/vhost_scsi_ut.c

@@ -59,8 +59,9 @@ DEFINE_STUB_P(spdk_scsi_lun_get_name, const char, (const struct spdk_scsi_lun *l
 DEFINE_STUB(spdk_scsi_lun_get_id, int, (const struct spdk_scsi_lun *lun), 0);
 DEFINE_STUB(spdk_vhost_vq_avail_ring_get, uint16_t, (struct rte_vhost_vring *vq, uint16_t *reqs,
 		uint16_t reqs_len), 0);
-DEFINE_STUB(spdk_vhost_vq_get_desc, int, (struct rte_vhost_vring *vq, uint16_t req_idx,
-		struct vring_desc **desc, struct vring_desc **desc_table, uint32_t *desc_table_size), 0);
+DEFINE_STUB(spdk_vhost_vq_get_desc, int, (struct spdk_vhost_dev *vdev, struct rte_vhost_vring *vq,
+		uint16_t req_idx, struct vring_desc **desc, struct vring_desc **desc_table,
+		uint32_t *desc_table_size), 0);
 DEFINE_STUB_VP(spdk_vhost_gpa_to_vva, (struct spdk_vhost_dev *vdev, uint64_t addr), {0});
 DEFINE_STUB_V(spdk_vhost_vq_used_ring_enqueue, (struct spdk_vhost_dev *vdev,
 		struct rte_vhost_vring *vq, uint16_t id, uint32_t len));