vhost: fix packed ring descriptor update in async enqueue
Checks
Commit Message
For the packet using multiple descriptors, the flag of first
descriptor cannot be updated until finishing updating the flag
of remaining descriptors. However, if the packet's descriptor
index is wrapped, the first descriptor's flag is updated earlier
than the others.
Fixes: 873e8dad6f49 ("vhost: support packed ring in async datapath")
Cc: stable@dpdk.org
Signed-off-by: Jiayu Hu <jiayu.hu@intel.com>
---
lib/vhost/virtio_net.c | 122 ++++++++++++++++++-----------------------
1 file changed, 54 insertions(+), 68 deletions(-)
Comments
On 11/4/21 11:19, Jiayu Hu wrote:
> For the packet using multiple descriptors, the flag of first
> descriptor cannot be updated until finishing updating the flag
> of remaining descriptors. However, if the packet's descriptor
> index is wrapped, the first descriptor's flag is updated earlier
> than the others.
The commit message is ambiguous IMHO, it is not clear whether it
describes the bug or the fix. I understand it describes the problem,
as the first descriptor flags always have to be updated last, but it
please reword it to make it clear what iu the issue, and how this patch
fixes it.
> Fixes: 873e8dad6f49 ("vhost: support packed ring in async datapath")
> Cc: stable@dpdk.org
If you think it is worth backporting it to stable, please take care of
preparing the backport as I suspect it will not apply cleanly.
> Signed-off-by: Jiayu Hu <jiayu.hu@intel.com>
> ---
> lib/vhost/virtio_net.c | 122 ++++++++++++++++++-----------------------
> 1 file changed, 54 insertions(+), 68 deletions(-)
>
@@ -1549,60 +1549,6 @@ virtio_dev_rx_async_submit_split(struct virtio_net *dev,
return pkt_idx;
}
-static __rte_always_inline void
-vhost_update_used_packed(struct vhost_virtqueue *vq,
- struct vring_used_elem_packed *shadow_ring,
- uint16_t count)
-{
- int i;
- uint16_t used_idx = vq->last_used_idx;
- uint16_t head_idx = vq->last_used_idx;
- uint16_t head_flags = 0;
-
- if (count == 0)
- return;
-
- /* Split loop in two to save memory barriers */
- for (i = 0; i < count; i++) {
- vq->desc_packed[used_idx].id = shadow_ring[i].id;
- vq->desc_packed[used_idx].len = shadow_ring[i].len;
-
- used_idx += shadow_ring[i].count;
- if (used_idx >= vq->size)
- used_idx -= vq->size;
- }
-
- /* The ordering for storing desc flags needs to be enforced. */
- rte_atomic_thread_fence(__ATOMIC_RELEASE);
-
- for (i = 0; i < count; i++) {
- uint16_t flags;
-
- if (vq->shadow_used_packed[i].len)
- flags = VRING_DESC_F_WRITE;
- else
- flags = 0;
-
- if (vq->used_wrap_counter) {
- flags |= VRING_DESC_F_USED;
- flags |= VRING_DESC_F_AVAIL;
- } else {
- flags &= ~VRING_DESC_F_USED;
- flags &= ~VRING_DESC_F_AVAIL;
- }
-
- if (i > 0) {
- vq->desc_packed[vq->last_used_idx].flags = flags;
- } else {
- head_idx = vq->last_used_idx;
- head_flags = flags;
- }
-
- vq_inc_last_used_packed(vq, shadow_ring[i].count);
- }
-
- vq->desc_packed[head_idx].flags = head_flags;
-}
static __rte_always_inline int
vhost_enqueue_async_packed(struct virtio_net *dev,
@@ -1819,23 +1765,63 @@ write_back_completed_descs_packed(struct vhost_virtqueue *vq,
uint16_t n_buffers)
{
struct vhost_async *async = vq->async;
- uint16_t nr_left = n_buffers;
- uint16_t from, to;
+ uint16_t from = async->last_buffer_idx_packed;
+ uint16_t used_idx = vq->last_used_idx;
+ uint16_t head_idx = vq->last_used_idx;
+ uint16_t head_flags = 0;
+ uint16_t i;
- do {
- from = async->last_buffer_idx_packed;
- to = (from + nr_left) % vq->size;
- if (to > from) {
- vhost_update_used_packed(vq, async->buffers_packed + from, to - from);
- async->last_buffer_idx_packed += nr_left;
- nr_left = 0;
+ /* Split loop in two to save memory barriers */
+ for (i = 0; i < n_buffers; i++) {
+ vq->desc_packed[used_idx].id = async->buffers_packed[from].id;
+ vq->desc_packed[used_idx].len = async->buffers_packed[from].len;
+
+ used_idx += async->buffers_packed[from].count;
+ if (used_idx >= vq->size)
+ used_idx -= vq->size;
+
+ from++;
+ if (from >= vq->size)
+ from = 0;
+ }
+
+ /* The ordering for storing desc flags needs to be enforced. */
+ rte_atomic_thread_fence(__ATOMIC_RELEASE);
+
+ from = async->last_buffer_idx_packed;
+
+ for (i = 0; i < n_buffers; i++) {
+ uint16_t flags;
+
+ if (async->buffers_packed[from].len)
+ flags = VRING_DESC_F_WRITE;
+ else
+ flags = 0;
+
+ if (vq->used_wrap_counter) {
+ flags |= VRING_DESC_F_USED;
+ flags |= VRING_DESC_F_AVAIL;
} else {
- vhost_update_used_packed(vq, async->buffers_packed + from,
- vq->size - from);
- async->last_buffer_idx_packed = 0;
- nr_left -= vq->size - from;
+ flags &= ~VRING_DESC_F_USED;
+ flags &= ~VRING_DESC_F_AVAIL;
}
- } while (nr_left > 0);
+
+ if (i > 0) {
+ vq->desc_packed[vq->last_used_idx].flags = flags;
+ } else {
+ head_idx = vq->last_used_idx;
+ head_flags = flags;
+ }
+
+ vq_inc_last_used_packed(vq, async->buffers_packed[from].count);
+
+ from++;
+ if (from == vq->size)
+ from = 0;
+ }
+
+ vq->desc_packed[head_idx].flags = head_flags;
+ async->last_buffer_idx_packed = from;
}
static __rte_always_inline uint16_t