[v4,2/2] virtio: one way barrier for split vring avail idx
Checks
Commit Message
In case VIRTIO_F_ORDER_PLATFORM(36) is not negotiated, then the frontend
and backend are assumed to be implemented in software, that is they can
run on identical CPUs in an SMP configuration.
Thus a weak form of memory barriers like rte_smp_r/wmb, other than
rte_cio_r/wmb, is sufficient for this case(vq->hw->weak_barriers == 1)
and yields better performance.
For the above case, this patch helps yielding even better performance
by replacing the two-way barriers with C11 one-way barriers for avail
index in split ring.
Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Gavin Hu <gavin.hu@arm.com>
---
drivers/net/virtio/virtqueue.h | 20 ++++++++++++++++++--
lib/librte_vhost/virtio_net.c | 14 +++++---------
2 files changed, 23 insertions(+), 11 deletions(-)
Comments
On 4/24/20 5:39 AM, Joyce Kong wrote:
> In case VIRTIO_F_ORDER_PLATFORM(36) is not negotiated, then the frontend
> and backend are assumed to be implemented in software, that is they can
> run on identical CPUs in an SMP configuration.
> Thus a weak form of memory barriers like rte_smp_r/wmb, other than
> rte_cio_r/wmb, is sufficient for this case(vq->hw->weak_barriers == 1)
> and yields better performance.
> For the above case, this patch helps yielding even better performance
> by replacing the two-way barriers with C11 one-way barriers for avail
> index in split ring.
>
> Signed-off-by: Joyce Kong <joyce.kong@arm.com>
> Reviewed-by: Gavin Hu <gavin.hu@arm.com>
> ---
> drivers/net/virtio/virtqueue.h | 20 ++++++++++++++++++--
> lib/librte_vhost/virtio_net.c | 14 +++++---------
> 2 files changed, 23 insertions(+), 11 deletions(-)
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Thanks,
Maxime
@@ -500,8 +500,24 @@ void vq_ring_free_inorder(struct virtqueue *vq, uint16_t desc_idx,
static inline void
vq_update_avail_idx(struct virtqueue *vq)
{
- virtio_wmb(vq->hw->weak_barriers);
- vq->vq_split.ring.avail->idx = vq->vq_avail_idx;
+ if (vq->hw->weak_barriers) {
+ /* x86 prefers to using rte_smp_wmb over __atomic_store_n as
+ * it reports a slightly better perf, which comes from the
+ * saved branch by the compiler.
+ * The if and else branches are identical with the smp and
+ * cio barriers both defined as compiler barriers on x86.
+ */
+#ifdef RTE_ARCH_X86_64
+ rte_smp_wmb();
+ vq->vq_split.ring.avail->idx = vq->vq_avail_idx;
+#else
+ __atomic_store_n(&vq->vq_split.ring.avail->idx,
+ vq->vq_avail_idx, __ATOMIC_RELEASE);
+#endif
+ } else {
+ rte_cio_wmb();
+ vq->vq_split.ring.avail->idx = vq->vq_avail_idx;
+ }
}
static inline void
@@ -991,13 +991,11 @@ virtio_dev_rx_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
struct buf_vector buf_vec[BUF_VECTOR_MAX];
uint16_t avail_head;
- avail_head = *((volatile uint16_t *)&vq->avail->idx);
-
/*
* The ordering between avail index and
* desc reads needs to be enforced.
*/
- rte_smp_rmb();
+ avail_head = __atomic_load_n(&vq->avail->idx, __ATOMIC_ACQUIRE);
rte_prefetch0(&vq->avail->ring[vq->last_avail_idx & (vq->size - 1)]);
@@ -1712,16 +1710,14 @@ virtio_dev_tx_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
}
}
- free_entries = *((volatile uint16_t *)&vq->avail->idx) -
- vq->last_avail_idx;
- if (free_entries == 0)
- return 0;
-
/*
* The ordering between avail index and
* desc reads needs to be enforced.
*/
- rte_smp_rmb();
+ free_entries = __atomic_load_n(&vq->avail->idx, __ATOMIC_ACQUIRE) -
+ vq->last_avail_idx;
+ if (free_entries == 0)
+ return 0;
rte_prefetch0(&vq->avail->ring[vq->last_avail_idx & (vq->size - 1)]);