[dpdk-dev] [PATCH v2 1/6] vhost: rewrite enqueue
On Thu, Aug 18, 2016 at 02:33:06AM -0400, Zhihong Wang wrote:
> This patch implements the vhost logic from scratch into a single function
> designed for high performance and better maintainability.
>
> Signed-off-by: Zhihong Wang
> ---
> lib/librte_vhost/vhost_rxtx.c | 212
> --
> 1 file changed, 205 insertions(+), 7 deletions(-)
>
> diff --git a/lib/librte_vhost/vhost_rxtx.c b/lib/librte_vhost/vhost_rxtx.c
> index 08a73fd..8e6d782 100644
> --- a/lib/librte_vhost/vhost_rxtx.c
> +++ b/lib/librte_vhost/vhost_rxtx.c
> @@ -91,7 +91,7 @@ is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t
> qp_nb)
> return (is_tx ^ (idx & 1)) == 0 && idx < qp_nb * VIRTIO_QNUM;
> }
>
> -static void
> +static inline void __attribute__((always_inline))
> virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr
> *net_hdr)
> {
> if (m_buf->ol_flags & PKT_TX_L4_MASK) {
> @@ -533,19 +533,217 @@ virtio_dev_merge_rx(struct virtio_net *dev, uint16_t
> queue_id,
> return pkt_idx;
> }
>
> +static inline uint32_t __attribute__((always_inline))
> +loop_check(struct vhost_virtqueue *vq, uint16_t avail_idx, uint32_t pkt_left)
> +{
> + if (pkt_left == 0 || avail_idx == vq->last_used_idx)
> + return 1;
> +
> + return 0;
> +}
Hmmm, I don't see any benifit from making such simple check into a
function.
> +static inline uint32_t __attribute__((always_inline))
> +enqueue_packet(struct virtio_net *dev, struct vhost_virtqueue *vq,
> + uint16_t avail_idx, struct rte_mbuf *mbuf,
> + uint32_t is_mrg_rxbuf)
> +{
> + struct virtio_net_hdr_mrg_rxbuf *virtio_hdr;
> + struct vring_desc *desc;
> + uint64_t desc_host_write_addr = 0;
> + uint32_t desc_chain_head = 0;
> + uint32_t desc_chain_len = 0;
> + uint32_t desc_current = 0;
> + uint32_t desc_write_offset = 0;
> + uint32_t mbuf_len = 0;
> + uint32_t mbuf_len_left = 0;
> + uint32_t copy_len = 0;
The dequeue function uses var like desc_addr, desc_avail, desc_offset,
mbuf_avail, ..., I see no reason to use something different here. This
breaks the code consistency. Besides that, var name like desc_host_write_addr
looks redundant; desc_addr is much cleaner.
--yliu
[dpdk-dev] [PATCH v2 1/6] vhost: rewrite enqueue
> -Original Message-
> From: Yuanhan Liu [mailto:yuanhan.liu at linux.intel.com]
> Sent: Friday, August 19, 2016 10:39 AM
> To: Wang, Zhihong
> Cc: dev at dpdk.org; maxime.coquelin at redhat.com
> Subject: Re: [PATCH v2 1/6] vhost: rewrite enqueue
>
> On Thu, Aug 18, 2016 at 02:33:06AM -0400, Zhihong Wang wrote:
> > This patch implements the vhost logic from scratch into a single function
> > designed for high performance and better maintainability.
> >
> > Signed-off-by: Zhihong Wang
> > ---
> > lib/librte_vhost/vhost_rxtx.c | 212
> --
> > 1 file changed, 205 insertions(+), 7 deletions(-)
> >
> > diff --git a/lib/librte_vhost/vhost_rxtx.c b/lib/librte_vhost/vhost_rxtx.c
> > index 08a73fd..8e6d782 100644
> > --- a/lib/librte_vhost/vhost_rxtx.c
> > +++ b/lib/librte_vhost/vhost_rxtx.c
> > @@ -91,7 +91,7 @@ is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t
> qp_nb)
> > return (is_tx ^ (idx & 1)) == 0 && idx < qp_nb * VIRTIO_QNUM;
> > }
> >
> > -static void
> > +static inline void __attribute__((always_inline))
> > virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr
> *net_hdr)
> > {
> > if (m_buf->ol_flags & PKT_TX_L4_MASK) {
> > @@ -533,19 +533,217 @@ virtio_dev_merge_rx(struct virtio_net *dev,
> uint16_t queue_id,
> > return pkt_idx;
> > }
> >
> > +static inline uint32_t __attribute__((always_inline))
> > +loop_check(struct vhost_virtqueue *vq, uint16_t avail_idx, uint32_t
> > pkt_left)
> > +{
> > + if (pkt_left == 0 || avail_idx == vq->last_used_idx)
> > + return 1;
> > +
> > + return 0;
> > +}
>
> Hmmm, I don't see any benifit from making such simple check into a
> function.
It's for prefetch code later to be merged into this function.
>
> > +static inline uint32_t __attribute__((always_inline))
> > +enqueue_packet(struct virtio_net *dev, struct vhost_virtqueue *vq,
> > + uint16_t avail_idx, struct rte_mbuf *mbuf,
> > + uint32_t is_mrg_rxbuf)
> > +{
> > + struct virtio_net_hdr_mrg_rxbuf *virtio_hdr;
> > + struct vring_desc *desc;
> > + uint64_t desc_host_write_addr = 0;
> > + uint32_t desc_chain_head = 0;
> > + uint32_t desc_chain_len = 0;
> > + uint32_t desc_current = 0;
> > + uint32_t desc_write_offset = 0;
> > + uint32_t mbuf_len = 0;
> > + uint32_t mbuf_len_left = 0;
> > + uint32_t copy_len = 0;
>
> The dequeue function uses var like desc_addr, desc_avail, desc_offset,
> mbuf_avail, ..., I see no reason to use something different here. This
> breaks the code consistency. Besides that, var name like desc_host_write_addr
> looks redundant; desc_addr is much cleaner.
Okay.
>
> --yliu
[dpdk-dev] [PATCH v2 1/6] vhost: rewrite enqueue
This patch implements the vhost logic from scratch into a single function
designed for high performance and better maintainability.
Signed-off-by: Zhihong Wang
---
lib/librte_vhost/vhost_rxtx.c | 212 --
1 file changed, 205 insertions(+), 7 deletions(-)
diff --git a/lib/librte_vhost/vhost_rxtx.c b/lib/librte_vhost/vhost_rxtx.c
index 08a73fd..8e6d782 100644
--- a/lib/librte_vhost/vhost_rxtx.c
+++ b/lib/librte_vhost/vhost_rxtx.c
@@ -91,7 +91,7 @@ is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t
qp_nb)
return (is_tx ^ (idx & 1)) == 0 && idx < qp_nb * VIRTIO_QNUM;
}
-static void
+static inline void __attribute__((always_inline))
virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr *net_hdr)
{
if (m_buf->ol_flags & PKT_TX_L4_MASK) {
@@ -533,19 +533,217 @@ virtio_dev_merge_rx(struct virtio_net *dev, uint16_t
queue_id,
return pkt_idx;
}
+static inline uint32_t __attribute__((always_inline))
+loop_check(struct vhost_virtqueue *vq, uint16_t avail_idx, uint32_t pkt_left)
+{
+ if (pkt_left == 0 || avail_idx == vq->last_used_idx)
+ return 1;
+
+ return 0;
+}
+
+static inline uint32_t __attribute__((always_inline))
+enqueue_packet(struct virtio_net *dev, struct vhost_virtqueue *vq,
+ uint16_t avail_idx, struct rte_mbuf *mbuf,
+ uint32_t is_mrg_rxbuf)
+{
+ struct virtio_net_hdr_mrg_rxbuf *virtio_hdr;
+ struct vring_desc *desc;
+ uint64_t desc_host_write_addr = 0;
+ uint32_t desc_chain_head = 0;
+ uint32_t desc_chain_len = 0;
+ uint32_t desc_current = 0;
+ uint32_t desc_write_offset = 0;
+ uint32_t mbuf_len = 0;
+ uint32_t mbuf_len_left = 0;
+ uint32_t copy_len = 0;
+ uint32_t extra_buffers = 0;
+ uint32_t used_idx_round = 0;
+
+ /* start with the first mbuf of the packet */
+ mbuf_len = rte_pktmbuf_data_len(mbuf);
+ mbuf_len_left = mbuf_len;
+
+ /* get the current desc */
+ desc_current = vq->avail->ring[(vq->last_used_idx) & (vq->size - 1)];
+ desc_chain_head = desc_current;
+ desc = >desc[desc_current];
+ desc_host_write_addr = gpa_to_vva(dev, desc->addr);
+ if (unlikely(!desc_host_write_addr))
+ goto error;
+
+ /* handle virtio header */
+ virtio_hdr = (struct virtio_net_hdr_mrg_rxbuf *)
+ (uintptr_t)desc_host_write_addr;
+ memset((void *)(uintptr_t)&(virtio_hdr->hdr),
+ 0, dev->vhost_hlen);
+ virtio_enqueue_offload(mbuf, &(virtio_hdr->hdr));
+ vhost_log_write(dev, desc->addr, dev->vhost_hlen);
+ desc_write_offset = dev->vhost_hlen;
+ desc_chain_len = desc_write_offset;
+ desc_host_write_addr += desc_write_offset;
+ if (is_mrg_rxbuf)
+ virtio_hdr->num_buffers = 1;
+
+ /* start copy from mbuf to desc */
+ while (1) {
+ /* get the next mbuf if the current done */
+ if (!mbuf_len_left) {
+ if (mbuf->next) {
+ mbuf = mbuf->next;
+ mbuf_len = rte_pktmbuf_data_len(mbuf);
+ mbuf_len_left = mbuf_len;
+ } else
+ break;
+ }
+
+ /* get the next desc if the current done */
+ if (desc->len <= desc_write_offset) {
+ if (desc->flags & VRING_DESC_F_NEXT) {
+ /* go on with the current desc chain */
+ desc_write_offset = 0;
+ desc_current = desc->next;
+ desc = >desc[desc_current];
+ desc_host_write_addr =
+ gpa_to_vva(dev, desc->addr);
+ if (unlikely(!desc_host_write_addr))
+ goto rollback;
+ } else if (is_mrg_rxbuf) {
+ /* start with the next desc chain */
+ used_idx_round = vq->last_used_idx
+ & (vq->size - 1);
+ vq->used->ring[used_idx_round].id =
+ desc_chain_head;
+ vq->used->ring[used_idx_round].len =
+ desc_chain_len;
+ vhost_log_used_vring(dev, vq,
+ offsetof(struct vring_used,
+ ring[used_idx_round]),
+ sizeof(vq->used->ring[
+ used_idx_round]));
+ vq->last_used_idx++;
+ extra_buffers++;
+
