Re: [PATCH v3 14/21] vhost: dequeue for packed queues

[Maxime Coquelin <maxime.coquelin@redhat.com>]

On 04/05/2018 12:10 PM, Jens Freimann wrote:
> Implement code to dequeue and process descriptors from
> the vring if VIRTIO_F_RING_PACKED is enabled.
> 
> Check if descriptor was made available by driver by looking at
> VIRTIO_F_DESC_AVAIL flag in descriptor. If so dequeue and set
> the used flag VIRTIO_F_DESC_USED to the current value of the
> used wrap counter.
> 
> Used ring wrap counter needs to be toggled when last descriptor is
> written out. This allows the host/guest to detect new descriptors even
> after the ring has wrapped.
> 
> Signed-off-by: Jens Freimann <jfreimann@redhat.com>
> ---
>   lib/librte_vhost/vhost.c      |   1 +
>   lib/librte_vhost/vhost.h      |   1 +
>   lib/librte_vhost/virtio_net.c | 228 ++++++++++++++++++++++++++++++++++++++++++
>   3 files changed, 230 insertions(+)
> 
> diff --git a/lib/librte_vhost/vhost.c b/lib/librte_vhost/vhost.c
> index 1f17cdd75..eb5a98875 100644
> --- a/lib/librte_vhost/vhost.c
> +++ b/lib/librte_vhost/vhost.c
> @@ -185,6 +185,7 @@ init_vring_queue(struct virtio_net *dev, uint32_t vring_idx)
>   
>   	vq->kickfd = VIRTIO_UNINITIALIZED_EVENTFD;
>   	vq->callfd = VIRTIO_UNINITIALIZED_EVENTFD;
> +	vq->used_wrap_counter = 1;
>   
>   	vhost_user_iotlb_init(dev, vring_idx);
>   	/* Backends are set to -1 indicating an inactive device. */
> diff --git a/lib/librte_vhost/vhost.h b/lib/librte_vhost/vhost.h
> index 20d78f883..c8aa946fd 100644
> --- a/lib/librte_vhost/vhost.h
> +++ b/lib/librte_vhost/vhost.h
> @@ -112,6 +112,7 @@ struct vhost_virtqueue {
>   
>   	struct batch_copy_elem	*batch_copy_elems;
>   	uint16_t		batch_copy_nb_elems;
> +	uint16_t 		used_wrap_counter;
>   
>   	rte_rwlock_t	iotlb_lock;
>   	rte_rwlock_t	iotlb_pending_lock;
> diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c
> index ed7198dbb..7eea1da04 100644
> --- a/lib/librte_vhost/virtio_net.c
> +++ b/lib/librte_vhost/virtio_net.c
> @@ -19,6 +19,7 @@
>   
>   #include "iotlb.h"
>   #include "vhost.h"
> +#include "virtio-1.1.h"
>   
>   #define MAX_PKT_BURST 32
>   
> @@ -1118,6 +1119,233 @@ restore_mbuf(struct rte_mbuf *m)
>   	}
>   }
>   
> +static inline uint16_t
> +dequeue_desc_packed(struct virtio_net *dev, struct vhost_virtqueue *vq,
> +	     struct rte_mempool *mbuf_pool, struct rte_mbuf *m,
> +	     struct vring_desc_packed *descs)
> +{
> +	struct vring_desc_packed *desc;
> +	uint64_t desc_addr;
> +	uint32_t desc_avail, desc_offset;
> +	uint32_t mbuf_avail, mbuf_offset;
> +	uint32_t cpy_len;
> +	struct rte_mbuf *cur = m, *prev = m;
> +	struct virtio_net_hdr *hdr = NULL;
> +	uint16_t head_idx = vq->last_used_idx & (vq->size - 1);
> +	int wrap_counter = vq->used_wrap_counter;
> +	int rc = 0;
> +
> +	rte_spinlock_lock(&vq->access_lock);
> +
> +	if (unlikely(vq->enabled == 0))
> +		goto out;

It is unbalanced as it would unlock iotlb_rd_lock that isn't locked yet.

> +	if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
> +		vhost_user_iotlb_rd_lock(vq);
> +
> +	desc = &descs[vq->last_used_idx & (vq->size - 1)];
> +	if (unlikely((desc->len < dev->vhost_hlen)) ||
> +			(desc->flags & VRING_DESC_F_INDIRECT)) {
> +			RTE_LOG(ERR, VHOST_DATA,
> +				"INDIRECT not supported yet\n");
> +			rc = -1;
> +			goto out;
> +	}
> +
> +	desc_addr = vhost_iova_to_vva(dev, vq, desc->addr,
> +				      sizeof(*desc), VHOST_ACCESS_RO);
> +
> +	if (unlikely(!desc_addr)) {
> +		rc = -1;
> +		goto out;
> +	}
> +
> +	if (virtio_net_with_host_offload(dev)) {
> +		hdr = (struct virtio_net_hdr *)((uintptr_t)desc_addr);
> +		rte_prefetch0(hdr);
> +	}
> +
> +	/*
> +	 * A virtio driver normally uses at least 2 desc buffers
> +	 * for Tx: the first for storing the header, and others
> +	 * for storing the data.
> +	 */
> +	if (likely((desc->len == dev->vhost_hlen) &&
> +		   (desc->flags & VRING_DESC_F_NEXT) != 0)) {
> +		if ((++vq->last_used_idx & (vq->size - 1)) == 0)
> +			toggle_wrap_counter(vq);
> +
> +		desc = &descs[vq->last_used_idx & (vq->size - 1)];
> +
> +		if (unlikely(desc->flags & VRING_DESC_F_INDIRECT)) {
> +			RTE_LOG(ERR, VHOST_DATA,
> +				"INDIRECT not supported yet\n");
> +			rc = -1;
> +			goto out;
> +		}
> +
> +		desc_addr = vhost_iova_to_vva(dev, vq, desc->addr,
> +					      sizeof(*desc), VHOST_ACCESS_RO);
> +		if (unlikely(!desc_addr)) {
> +			rc = -1;
> +			goto out;
> +		}
> +
> +		desc_offset = 0;
> +		desc_avail  = desc->len;
> +	} else {
> +		desc_avail  = desc->len - dev->vhost_hlen;
> +		desc_offset = dev->vhost_hlen;
> +	}
> +
> +	rte_prefetch0((void *)(uintptr_t)(desc_addr + desc_offset));
> +
> +	PRINT_PACKET(dev, (uintptr_t)(desc_addr + desc_offset), desc_avail, 0);
> +
> +	mbuf_offset = 0;
> +	mbuf_avail  = m->buf_len - RTE_PKTMBUF_HEADROOM;
> +	while (1) {
> +		uint64_t hpa;
> +
> +		cpy_len = RTE_MIN(desc_avail, mbuf_avail);
> +
> +		/*
> +		 * A desc buf might across two host physical pages that are
> +		 * not continuous. In such case (gpa_to_hpa returns 0), data
> +		 * will be copied even though zero copy is enabled.
> +		 */
> +		if (unlikely(dev->dequeue_zero_copy && (hpa = gpa_to_hpa(dev,
> +					desc->addr + desc_offset, cpy_len)))) {
> +			cur->data_len = cpy_len;
> +			cur->data_off = 0;
> +			cur->buf_addr = (void *)(uintptr_t)desc_addr;
> +			cur->buf_physaddr = hpa;
> +
> +			/*
> +			 * In zero copy mode, one mbuf can only reference data
> +			 * for one or partial of one desc buff.
> +			 */
> +			mbuf_avail = cpy_len;
> +		} else {
> +			rte_memcpy(rte_pktmbuf_mtod_offset(cur, void *,
> +							   mbuf_offset),
> +				(void *)((uintptr_t)(desc_addr + desc_offset)),
> +				cpy_len);
> +		}
> +
> +		mbuf_avail  -= cpy_len;
> +		mbuf_offset += cpy_len;
> +		desc_avail  -= cpy_len;
> +		desc_offset += cpy_len;
> +
> +		/* This desc reaches to its end, get the next one */
> +		if (desc_avail == 0) {
> +			if ((desc->flags & VRING_DESC_F_NEXT) == 0)
> +				break;
> +
> +			if ((++vq->last_used_idx & (vq->size - 1)) == 0)
> +				toggle_wrap_counter(vq);
> +
> +			desc = &descs[vq->last_used_idx & (vq->size - 1)];
> +			if (unlikely(desc->flags & VRING_DESC_F_INDIRECT)) {
> +				RTE_LOG(ERR, VHOST_DATA, "INDIRECT not supported yet");
> +			}
> +
> +			desc_addr = vhost_iova_to_vva(dev, vq, desc->addr,
> +						      sizeof(*desc), VHOST_ACCESS_RO);
> +			if (unlikely(!desc_addr)) {
> +				rc = -1;
> +				goto out;
Not sure, but, shouldn't you break here instead so rings gets updated if
some of the descs have been processed?
> +			}
> +
> +			rte_prefetch0((void *)(uintptr_t)desc_addr);
> +
> +			desc_offset = 0;
> +			desc_avail  = desc->len;
> +
> +			PRINT_PACKET(dev, (uintptr_t)desc_addr, desc->len, 0);
> +		}
> +
> +		/*
> +		 * This mbuf reaches to its end, get a new one
> +		 * to hold more data.
> +		 */
> +		if (mbuf_avail == 0) {
> +			cur = rte_pktmbuf_alloc(mbuf_pool);
> +			if (unlikely(cur == NULL)) {
> +				RTE_LOG(ERR, VHOST_DATA, "Failed to "
> +					"allocate memory for mbuf.\n");
> +				rc = -1;
> +				goto out;
> +			}
> +
> +			prev->next = cur;
> +			prev->data_len = mbuf_offset;
> +			m->nb_segs += 1;
> +			m->pkt_len += mbuf_offset;
> +			prev = cur;
> +
> +			mbuf_offset = 0;
> +			mbuf_avail  = cur->buf_len - RTE_PKTMBUF_HEADROOM;
> +		}
> +	}
> +
> +	if (hdr)
> +		vhost_dequeue_offload(hdr, m);
> +
> +	if ((++vq->last_used_idx & (vq->size - 1)) == 0)
> +		toggle_wrap_counter(vq);
> +
> +	rte_smp_wmb();
> +	_set_desc_used(&descs[head_idx], wrap_counter);
> +
> +	prev->data_len = mbuf_offset;
> +	m->pkt_len    += mbuf_offset;
> +
> +out:
> +	if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM))
> +		vhost_user_iotlb_rd_unlock(vq);
> +	rte_spinlock_unlock(&vq->access_lock);
> +
> +	return rc;
> +}
> +
> +static inline uint16_t
> +vhost_dequeue_burst_packed(struct virtio_net *dev, struct vhost_virtqueue *vq,
> +			struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts,
> +			uint16_t count)
> +{
> +	uint16_t i;
> +	uint16_t idx;
> +	struct vring_desc_packed *desc = vq->desc_packed;
> +	int err;
> +
> +	count = RTE_MIN(MAX_PKT_BURST, count);
> +	for (i = 0; i < count; i++) {
> +		idx = vq->last_used_idx & (vq->size - 1);
> +		if (!desc_is_avail(vq, &desc[idx]))
> +			break;
> +		rte_smp_rmb();
> +
> +		pkts[i] = rte_pktmbuf_alloc(mbuf_pool);
> +		if (unlikely(pkts[i] == NULL)) {
> +			RTE_LOG(ERR, VHOST_DATA,
> +				"Failed to allocate memory for mbuf.\n");
> +			break;
> +		}
> +
> +		err = dequeue_desc_packed(dev, vq, mbuf_pool, pkts[i], desc);
> +		if (unlikely(err)) {
> +			rte_pktmbuf_free(pkts[i]);
> +			break;
> +		}
> +	}
> +
> +	rte_spinlock_unlock(&vq->access_lock);
I think you forgot to remove this line.

> +
> +	return i;
> +}
> +
>   uint16_t
>   rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
>   	struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count)
>