New upstream version 17.11.2upstream/17.11.2

Change-Id: Iccf80e68222e7fc45af045bfed265a962b5ccca6
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>

7 files changed, 473 insertions, 97 deletions

diff --git a/lib/librte_eal/common/include/rte_version.h b/lib/librte_eal/common/include/rte_version.h
index a0611f3d..4616a080 100644
--- a/lib/librte_eal/common/include/rte_version.h
+++ b/lib/librte_eal/common/include/rte_version.h
@@ -66,7 +66,7 @@ extern "C" {
 /**
  * Patch level number i.e. the z in yy.mm.z
  */
-#define RTE_VER_MINOR 1
+#define RTE_VER_MINOR 2
 
 /**
  * Extra string to be appended to version number
diff --git a/lib/librte_vhost/rte_vhost.h b/lib/librte_vhost/rte_vhost.h
index f6536449..3fc6034d 100644
--- a/lib/librte_vhost/rte_vhost.h
+++ b/lib/librte_vhost/rte_vhost.h
@@ -117,6 +117,11 @@ struct vhost_device_ops {
 /**
  * Convert guest physical address to host virtual address
  *
+ * This function is deprecated because unsafe.
+ * New rte_vhost_va_from_guest_pa() should be used instead to ensure
+ * guest physical ranges are fully and contiguously mapped into
+ * process virtual address space.
+ *
  * @param mem
  *  the guest memory regions
  * @param gpa
@@ -124,6 +129,7 @@ struct vhost_device_ops {
  * @return
  *  the host virtual address on success, 0 on failure
  */
+__rte_deprecated
 static __rte_always_inline uint64_t
 rte_vhost_gpa_to_vva(struct rte_vhost_memory *mem, uint64_t gpa)
 {
@@ -142,6 +148,46 @@ rte_vhost_gpa_to_vva(struct rte_vhost_memory *mem, uint64_t gpa)
 	return 0;
 }
 
+/**
+ * Convert guest physical address to host virtual address safely
+ *
+ * This variant of rte_vhost_gpa_to_vva() takes care all the
+ * requested length is mapped and contiguous in process address
+ * space.
+ *
+ * @param mem
+ *  the guest memory regions
+ * @param gpa
+ *  the guest physical address for querying
+ * @param len
+ *  the size of the requested area to map, updated with actual size mapped
+ * @return
+ *  the host virtual address on success, 0 on failure
+ */
+static __rte_always_inline uint64_t
+rte_vhost_va_from_guest_pa(struct rte_vhost_memory *mem,
+						   uint64_t gpa, uint64_t *len)
+{
+	struct rte_vhost_mem_region *r;
+	uint32_t i;
+
+	for (i = 0; i < mem->nregions; i++) {
+		r = &mem->regions[i];
+		if (gpa >= r->guest_phys_addr &&
+		    gpa <  r->guest_phys_addr + r->size) {
+
+			if (unlikely(*len > r->guest_phys_addr + r->size - gpa))
+				*len = r->guest_phys_addr + r->size - gpa;
+
+			return gpa - r->guest_phys_addr +
+			       r->host_user_addr;
+		}
+	}
+	*len = 0;
+
+	return 0;
+}
+
 #define RTE_VHOST_NEED_LOG(features)	((features) & (1ULL << VHOST_F_LOG_ALL))
 
 /**
diff --git a/lib/librte_vhost/rte_vhost_version.map b/lib/librte_vhost/rte_vhost_version.map
index 1e704953..9cb1d8ca 100644
--- a/lib/librte_vhost/rte_vhost_version.map
+++ b/lib/librte_vhost/rte_vhost_version.map
@@ -52,3 +52,9 @@ DPDK_17.08 {
 	rte_vhost_rx_queue_count;
 
 } DPDK_17.05;
+
+DPDK_17.11.2 {
+	global;
+
+	rte_vhost_va_from_guest_pa;
+} DPDK_17.08;
diff --git a/lib/librte_vhost/vhost.c b/lib/librte_vhost/vhost.c
index 51ea720a..a8ed40b1 100644
--- a/lib/librte_vhost/vhost.c
+++ b/lib/librte_vhost/vhost.c
@@ -58,17 +58,17 @@ struct virtio_net *vhost_devices[MAX_VHOST_DEVICE];
 /* Called with iotlb_lock read-locked */
 uint64_t
 __vhost_iova_to_vva(struct virtio_net *dev, struct vhost_virtqueue *vq,
-		    uint64_t iova, uint64_t size, uint8_t perm)
+		    uint64_t iova, uint64_t *size, uint8_t perm)
 {
 	uint64_t vva, tmp_size;
 
-	if (unlikely(!size))
+	if (unlikely(!*size))
 		return 0;
 
-	tmp_size = size;
+	tmp_size = *size;
 
 	vva = vhost_user_iotlb_cache_find(vq, iova, &tmp_size, perm);
-	if (tmp_size == size)
+	if (tmp_size == *size)
 		return vva;
 
 	iova += tmp_size;
@@ -158,32 +158,39 @@ free_device(struct virtio_net *dev)
 int
 vring_translate(struct virtio_net *dev, struct vhost_virtqueue *vq)
 {
-	uint64_t size;
+	uint64_t req_size, size;
 
 	if (!(dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)))
 		goto out;
 
-	size = sizeof(struct vring_desc) * vq->size;
+	req_size = sizeof(struct vring_desc) * vq->size;
+	size = req_size;
 	vq->desc = (struct vring_desc *)(uintptr_t)vhost_iova_to_vva(dev, vq,
 						vq->ring_addrs.desc_user_addr,
-						size, VHOST_ACCESS_RW);
-	if (!vq->desc)
+						&size, VHOST_ACCESS_RW);
+	if (!vq->desc || size != req_size)
 		return -1;
 
-	size = sizeof(struct vring_avail);
-	size += sizeof(uint16_t) * vq->size;
+	req_size = sizeof(struct vring_avail);
+	req_size += sizeof(uint16_t) * vq->size;
+	if (dev->features & (1ULL << VIRTIO_RING_F_EVENT_IDX))
+		req_size += sizeof(uint16_t);
+	size = req_size;
 	vq->avail = (struct vring_avail *)(uintptr_t)vhost_iova_to_vva(dev, vq,
 						vq->ring_addrs.avail_user_addr,
-						size, VHOST_ACCESS_RW);
-	if (!vq->avail)
+						&size, VHOST_ACCESS_RW);
+	if (!vq->avail || size != req_size)
 		return -1;
 
-	size = sizeof(struct vring_used);
-	size += sizeof(struct vring_used_elem) * vq->size;
+	req_size = sizeof(struct vring_used);
+	req_size += sizeof(struct vring_used_elem) * vq->size;
+	if (dev->features & (1ULL << VIRTIO_RING_F_EVENT_IDX))
+		req_size += sizeof(uint16_t);
+	size = req_size;
 	vq->used = (struct vring_used *)(uintptr_t)vhost_iova_to_vva(dev, vq,
 						vq->ring_addrs.used_user_addr,
-						size, VHOST_ACCESS_RW);
-	if (!vq->used)
+						&size, VHOST_ACCESS_RW);
+	if (!vq->used || size != req_size)
 		return -1;
 
 out:
diff --git a/lib/librte_vhost/vhost.h b/lib/librte_vhost/vhost.h
index c8f2a817..16d6b891 100644
--- a/lib/librte_vhost/vhost.h
+++ b/lib/librte_vhost/vhost.h
@@ -381,18 +381,18 @@ struct vhost_device_ops const *vhost_driver_callback_get(const char *path);
 void vhost_backend_cleanup(struct virtio_net *dev);
 
 uint64_t __vhost_iova_to_vva(struct virtio_net *dev, struct vhost_virtqueue *vq,
-			uint64_t iova, uint64_t size, uint8_t perm);
+			uint64_t iova, uint64_t *len, uint8_t perm);
 int vring_translate(struct virtio_net *dev, struct vhost_virtqueue *vq);
 void vring_invalidate(struct virtio_net *dev, struct vhost_virtqueue *vq);
 
 static __rte_always_inline uint64_t
 vhost_iova_to_vva(struct virtio_net *dev, struct vhost_virtqueue *vq,
-			uint64_t iova, uint64_t size, uint8_t perm)
+			uint64_t iova, uint64_t *len, uint8_t perm)
 {
 	if (!(dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)))
-		return rte_vhost_gpa_to_vva(dev->mem, iova);
+		return rte_vhost_va_from_guest_pa(dev->mem, iova, len);
 
-	return __vhost_iova_to_vva(dev, vq, iova, size, perm);
+	return __vhost_iova_to_vva(dev, vq, iova, len, perm);
 }
 
 #endif /* _VHOST_NET_CDEV_H_ */
diff --git a/lib/librte_vhost/vhost_user.c b/lib/librte_vhost/vhost_user.c
index 3acaacf5..50e654db 100644
--- a/lib/librte_vhost/vhost_user.c
+++ b/lib/librte_vhost/vhost_user.c
@@ -329,21 +329,26 @@ numa_realloc(struct virtio_net *dev, int index __rte_unused)
 
 /* Converts QEMU virtual address to Vhost virtual address. */
 static uint64_t
-qva_to_vva(struct virtio_net *dev, uint64_t qva)
+qva_to_vva(struct virtio_net *dev, uint64_t qva, uint64_t *len)
 {
-	struct rte_vhost_mem_region *reg;
+	struct rte_vhost_mem_region *r;
 	uint32_t i;
 
 	/* Find the region where the address lives. */
 	for (i = 0; i < dev->mem->nregions; i++) {
-		reg = &dev->mem->regions[i];
+		r = &dev->mem->regions[i];
+
+		if (qva >= r->guest_user_addr &&
+		    qva <  r->guest_user_addr + r->size) {
+
+			if (unlikely(*len > r->guest_user_addr + r->size - qva))
+				*len = r->guest_user_addr + r->size - qva;
 
-		if (qva >= reg->guest_user_addr &&
-		    qva <  reg->guest_user_addr + reg->size) {
-			return qva - reg->guest_user_addr +
-			       reg->host_user_addr;
+			return qva - r->guest_user_addr +
+			       r->host_user_addr;
 		}
 	}
+	*len = 0;
 
 	return 0;
 }
@@ -356,20 +361,20 @@ qva_to_vva(struct virtio_net *dev, uint64_t qva)
  */
 static uint64_t
 ring_addr_to_vva(struct virtio_net *dev, struct vhost_virtqueue *vq,
-		uint64_t ra, uint64_t size)
+		uint64_t ra, uint64_t *size)
 {
 	if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) {
 		uint64_t vva;
 
 		vva = vhost_user_iotlb_cache_find(vq, ra,
-					&size, VHOST_ACCESS_RW);
+					size, VHOST_ACCESS_RW);
 		if (!vva)
 			vhost_user_iotlb_miss(dev, ra, VHOST_ACCESS_RW);
 
 		return vva;
 	}
 
-	return qva_to_vva(dev, ra);
+	return qva_to_vva(dev, ra, size);
 }
 
 static struct virtio_net *
@@ -377,16 +382,18 @@ translate_ring_addresses(struct virtio_net *dev, int vq_index)
 {
 	struct vhost_virtqueue *vq = dev->virtqueue[vq_index];
 	struct vhost_vring_addr *addr = &vq->ring_addrs;
+	uint64_t len;
 
 	/* The addresses are converted from QEMU virtual to Vhost virtual. */
 	if (vq->desc && vq->avail && vq->used)
 		return dev;
 
+	len = sizeof(struct vring_desc) * vq->size;
 	vq->desc = (struct vring_desc *)(uintptr_t)ring_addr_to_vva(dev,
-			vq, addr->desc_user_addr, sizeof(struct vring_desc));
-	if (vq->desc == 0) {
+			vq, addr->desc_user_addr, &len);
+	if (vq->desc == 0 || len != sizeof(struct vring_desc) * vq->size) {
 		RTE_LOG(DEBUG, VHOST_CONFIG,
-			"(%d) failed to find desc ring address.\n",
+			"(%d) failed to map desc ring.\n",
 			dev->vid);
 		return dev;
 	}
@@ -395,20 +402,26 @@ translate_ring_addresses(struct virtio_net *dev, int vq_index)
 	vq = dev->virtqueue[vq_index];
 	addr = &vq->ring_addrs;
 
+	len = sizeof(struct vring_avail) + sizeof(uint16_t) * vq->size;
 	vq->avail = (struct vring_avail *)(uintptr_t)ring_addr_to_vva(dev,
-			vq, addr->avail_user_addr, sizeof(struct vring_avail));
-	if (vq->avail == 0) {
+			vq, addr->avail_user_addr, &len);
+	if (vq->avail == 0 ||
+			len != sizeof(struct vring_avail) +
+			sizeof(uint16_t) * vq->size) {
 		RTE_LOG(DEBUG, VHOST_CONFIG,
-			"(%d) failed to find avail ring address.\n",
+			"(%d) failed to map avail ring.\n",
 			dev->vid);
 		return dev;
 	}
 
+	len = sizeof(struct vring_used) +
+		sizeof(struct vring_used_elem) * vq->size;
 	vq->used = (struct vring_used *)(uintptr_t)ring_addr_to_vva(dev,
-			vq, addr->used_user_addr, sizeof(struct vring_used));
-	if (vq->used == 0) {
+			vq, addr->used_user_addr, &len);
+	if (vq->used == 0 || len != sizeof(struct vring_used) +
+			sizeof(struct vring_used_elem) * vq->size) {
 		RTE_LOG(DEBUG, VHOST_CONFIG,
-			"(%d) failed to find used ring address.\n",
+			"(%d) failed to map used ring.\n",
 			dev->vid);
 		return dev;
 	}
@@ -1094,11 +1107,12 @@ vhost_user_iotlb_msg(struct virtio_net **pdev, struct VhostUserMsg *msg)
 	struct virtio_net *dev = *pdev;
 	struct vhost_iotlb_msg *imsg = &msg->payload.iotlb;
 	uint16_t i;
-	uint64_t vva;
+	uint64_t vva, len;
 
 	switch (imsg->type) {
 	case VHOST_IOTLB_UPDATE:
-		vva = qva_to_vva(dev, imsg->uaddr);
+		len = imsg->size;
+		vva = qva_to_vva(dev, imsg->uaddr, &len);
 		if (!vva)
 			return -1;
 
@@ -1106,7 +1120,7 @@ vhost_user_iotlb_msg(struct virtio_net **pdev, struct VhostUserMsg *msg)
 			struct vhost_virtqueue *vq = dev->virtqueue[i];
 
 			vhost_user_iotlb_cache_insert(vq, imsg->iova, vva,
-					imsg->size, imsg->perm);
+					len, imsg->perm);
 
 			if (is_vring_iotlb_update(vq, imsg))
 				*pdev = dev = translate_ring_addresses(dev, i);
diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c
index d3470307..a013c07b 100644
--- a/lib/librte_vhost/virtio_net.c
+++ b/lib/librte_vhost/virtio_net.c
@@ -45,6 +45,7 @@
 #include <rte_sctp.h>
 #include <rte_arp.h>
 #include <rte_spinlock.h>
+#include <rte_malloc.h>
 
 #include "iotlb.h"
 #include "vhost.h"
@@ -59,6 +60,46 @@ is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t nr_vring)
 	return (is_tx ^ (idx & 1)) == 0 && idx < nr_vring;
 }
 
+static __rte_always_inline struct vring_desc *
+alloc_copy_ind_table(struct virtio_net *dev, struct vhost_virtqueue *vq,
+					 struct vring_desc *desc)
+{
+	struct vring_desc *idesc;
+	uint64_t src, dst;
+	uint64_t len, remain = desc->len;
+	uint64_t desc_addr = desc->addr;
+
+	idesc = rte_malloc(__func__, desc->len, 0);
+	if (unlikely(!idesc))
+		return 0;
+
+	dst = (uint64_t)(uintptr_t)idesc;
+
+	while (remain) {
+		len = remain;
+		src = vhost_iova_to_vva(dev, vq, desc_addr, &len,
+				VHOST_ACCESS_RO);
+		if (unlikely(!src || !len)) {
+			rte_free(idesc);
+			return 0;
+		}
+
+		rte_memcpy((void *)(uintptr_t)dst, (void *)(uintptr_t)src, len);
+
+		remain -= len;
+		dst += len;
+		desc_addr += len;
+	}
+
+	return idesc;
+}
+
+static __rte_always_inline void
+free_ind_table(struct vring_desc *idesc)
+{
+	rte_free(idesc);
+}
+
 static __rte_always_inline void
 do_flush_shadow_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq,
 			  uint16_t to, uint16_t from, uint16_t size)
@@ -204,8 +245,9 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq,
 	uint32_t desc_avail, desc_offset;
 	uint32_t mbuf_avail, mbuf_offset;
 	uint32_t cpy_len;
+	uint64_t desc_chunck_len;
 	struct vring_desc *desc;
-	uint64_t desc_addr;
+	uint64_t desc_addr, desc_gaddr;
 	/* A counter to avoid desc dead loop chain */
 	uint16_t nr_desc = 1;
 	struct batch_copy_elem *batch_copy = vq->batch_copy_elems;
@@ -213,8 +255,10 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq,
 	int error = 0;
 
 	desc = &descs[desc_idx];
-	desc_addr = vhost_iova_to_vva(dev, vq, desc->addr,
-					desc->len, VHOST_ACCESS_RW);
+	desc_chunck_len = desc->len;
+	desc_gaddr = desc->addr;
+	desc_addr = vhost_iova_to_vva(dev, vq, desc_gaddr,
+					&desc_chunck_len, VHOST_ACCESS_RW);
 	/*
 	 * Checking of 'desc_addr' placed outside of 'unlikely' macro to avoid
 	 * performance issue with some versions of gcc (4.8.4 and 5.3.0) which
@@ -227,12 +271,58 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq,
 
 	rte_prefetch0((void *)(uintptr_t)desc_addr);
 
-	virtio_enqueue_offload(m, (struct virtio_net_hdr *)(uintptr_t)desc_addr);
-	vhost_log_write(dev, desc->addr, dev->vhost_hlen);
-	PRINT_PACKET(dev, (uintptr_t)desc_addr, dev->vhost_hlen, 0);
+	if (likely(desc_chunck_len >= dev->vhost_hlen)) {
+		virtio_enqueue_offload(m,
+				(struct virtio_net_hdr *)(uintptr_t)desc_addr);
+		PRINT_PACKET(dev, (uintptr_t)desc_addr, dev->vhost_hlen, 0);
+		vhost_log_write(dev, desc_gaddr, dev->vhost_hlen);
+	} else {
+		struct virtio_net_hdr vnet_hdr;
+		uint64_t remain = dev->vhost_hlen;
+		uint64_t len;
+		uint64_t src = (uint64_t)(uintptr_t)&vnet_hdr, dst;
+		uint64_t guest_addr = desc_gaddr;
+
+		virtio_enqueue_offload(m, &vnet_hdr);
+
+		while (remain) {
+			len = remain;
+			dst = vhost_iova_to_vva(dev, vq, guest_addr,
+					&len, VHOST_ACCESS_RW);
+			if (unlikely(!dst || !len)) {
+				error = -1;
+				goto out;
+			}
+
+			rte_memcpy((void *)(uintptr_t)dst,
+					(void *)(uintptr_t)src, len);
+
+			PRINT_PACKET(dev, (uintptr_t)dst, len, 0);
+			vhost_log_write(dev, guest_addr, len);
+			remain -= len;
+			guest_addr += len;
+			dst += len;
+		}
+	}
 
-	desc_offset = dev->vhost_hlen;
 	desc_avail  = desc->len - dev->vhost_hlen;
+	if (unlikely(desc_chunck_len < dev->vhost_hlen)) {
+		desc_chunck_len = desc_avail;
+		desc_gaddr = desc->addr + dev->vhost_hlen;
+		desc_addr = vhost_iova_to_vva(dev,
+				vq, desc_gaddr,
+				&desc_chunck_len,
+				VHOST_ACCESS_RW);
+		if (unlikely(!desc_addr)) {
+			error = -1;
+			goto out;
+		}
+
+		desc_offset = 0;
+	} else {
+		desc_offset = dev->vhost_hlen;
+		desc_chunck_len -= dev->vhost_hlen;
+	}
 
 	mbuf_avail  = rte_pktmbuf_data_len(m);
 	mbuf_offset = 0;
@@ -258,8 +348,10 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq,
 			}
 
 			desc = &descs[desc->next];
-			desc_addr = vhost_iova_to_vva(dev, vq, desc->addr,
-							desc->len,
+			desc_chunck_len = desc->len;
+			desc_gaddr = desc->addr;
+			desc_addr = vhost_iova_to_vva(dev, vq, desc_gaddr,
+							&desc_chunck_len,
 							VHOST_ACCESS_RW);
 			if (unlikely(!desc_addr)) {
 				error = -1;
@@ -268,15 +360,26 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq,
 
 			desc_offset = 0;
 			desc_avail  = desc->len;
+		} else if (unlikely(desc_chunck_len == 0)) {
+			desc_chunck_len = desc_avail;
+			desc_gaddr += desc_offset;
+			desc_addr = vhost_iova_to_vva(dev,
+					vq, desc_gaddr,
+					&desc_chunck_len, VHOST_ACCESS_RW);
+			if (unlikely(!desc_addr)) {
+				error = -1;
+				goto out;
+			}
+			desc_offset = 0;
 		}
 
-		cpy_len = RTE_MIN(desc_avail, mbuf_avail);
+		cpy_len = RTE_MIN(desc_chunck_len, mbuf_avail);
 		if (likely(cpy_len > MAX_BATCH_LEN || copy_nb >= vq->size)) {
 			rte_memcpy((void *)((uintptr_t)(desc_addr +
 							desc_offset)),
 				rte_pktmbuf_mtod_offset(m, void *, mbuf_offset),
 				cpy_len);
-			vhost_log_write(dev, desc->addr + desc_offset, cpy_len);
+			vhost_log_write(dev, desc_gaddr + desc_offset, cpy_len);
 			PRINT_PACKET(dev, (uintptr_t)(desc_addr + desc_offset),
 				     cpy_len, 0);
 		} else {
@@ -284,7 +387,7 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq,
 				(void *)((uintptr_t)(desc_addr + desc_offset));
 			batch_copy[copy_nb].src =
 				rte_pktmbuf_mtod_offset(m, void *, mbuf_offset);
-			batch_copy[copy_nb].log_addr = desc->addr + desc_offset;
+			batch_copy[copy_nb].log_addr = desc_gaddr + desc_offset;
 			batch_copy[copy_nb].len = cpy_len;
 			copy_nb++;
 		}
@@ -293,6 +396,7 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		mbuf_offset += cpy_len;
 		desc_avail  -= cpy_len;
 		desc_offset += cpy_len;
+		desc_chunck_len -= cpy_len;
 	}
 
 out:
@@ -371,20 +475,34 @@ virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id,
 
 	rte_prefetch0(&vq->desc[desc_indexes[0]]);
 	for (i = 0; i < count; i++) {
+		struct vring_desc *idesc = NULL;
 		uint16_t desc_idx = desc_indexes[i];
 		int err;
 
 		if (vq->desc[desc_idx].flags & VRING_DESC_F_INDIRECT) {
+			uint64_t dlen = vq->desc[desc_idx].len;
 			descs = (struct vring_desc *)(uintptr_t)
 				vhost_iova_to_vva(dev,
 						vq, vq->desc[desc_idx].addr,
-						vq->desc[desc_idx].len,
-						VHOST_ACCESS_RO);
+						&dlen, VHOST_ACCESS_RO);
 			if (unlikely(!descs)) {
 				count = i;
 				break;
 			}
 
+			if (unlikely(dlen < vq->desc[desc_idx].len)) {
+				/*
+				 * The indirect desc table is not contiguous
+				 * in process VA space, we have to copy it.
+				 */
+				idesc = alloc_copy_ind_table(dev, vq,
+							&vq->desc[desc_idx]);
+				if (unlikely(!idesc))
+					break;
+
+				descs = idesc;
+			}
+
 			desc_idx = 0;
 			sz = vq->desc[desc_idx].len / sizeof(*descs);
 		} else {
@@ -395,11 +513,15 @@ virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id,
 		err = copy_mbuf_to_desc(dev, vq, descs, pkts[i], desc_idx, sz);
 		if (unlikely(err)) {
 			count = i;
+			free_ind_table(idesc);
 			break;
 		}
 
 		if (i + 1 < count)
 			rte_prefetch0(&vq->desc[desc_indexes[i+1]]);
+
+		if (unlikely(!!idesc))
+			free_ind_table(idesc);
 	}
 
 	do_data_copy_enqueue(dev, vq);
@@ -438,24 +560,41 @@ fill_vec_buf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 	uint16_t idx = vq->avail->ring[avail_idx & (vq->size - 1)];
 	uint32_t vec_id = *vec_idx;
 	uint32_t len    = 0;
+	uint64_t dlen;
 	struct vring_desc *descs = vq->desc;
+	struct vring_desc *idesc = NULL;
 
 	*desc_chain_head = idx;
 
 	if (vq->desc[idx].flags & VRING_DESC_F_INDIRECT) {
+		dlen = vq->desc[idx].len;
 		descs = (struct vring_desc *)(uintptr_t)
 			vhost_iova_to_vva(dev, vq, vq->desc[idx].addr,
-						vq->desc[idx].len,
+						&dlen,
 						VHOST_ACCESS_RO);
 		if (unlikely(!descs))
 			return -1;
 
+		if (unlikely(dlen < vq->desc[idx].len)) {
+			/*
+			 * The indirect desc table is not contiguous
+			 * in process VA space, we have to copy it.
+			 */
+			idesc = alloc_copy_ind_table(dev, vq, &vq->desc[idx]);
+			if (unlikely(!idesc))
+				return -1;
+
+			descs = idesc;
+		}
+
 		idx = 0;
 	}
 
 	while (1) {
-		if (unlikely(vec_id >= BUF_VECTOR_MAX || idx >= vq->size))
+		if (unlikely(vec_id >= BUF_VECTOR_MAX || idx >= vq->size)) {
+			free_ind_table(idesc);
 			return -1;
+		}
 
 		len += descs[idx].len;
 		buf_vec[vec_id].buf_addr = descs[idx].addr;
@@ -472,6 +611,9 @@ fill_vec_buf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 	*desc_chain_len = len;
 	*vec_idx = vec_id;
 
+	if (unlikely(!!idesc))
+		free_ind_table(idesc);
+
 	return 0;
 }
 
@@ -526,13 +668,15 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
 			    uint16_t num_buffers)
 {
 	uint32_t vec_idx = 0;
-	uint64_t desc_addr;
+	uint64_t desc_addr, desc_gaddr;
 	uint32_t mbuf_offset, mbuf_avail;
 	uint32_t desc_offset, desc_avail;
 	uint32_t cpy_len;
+	uint64_t desc_chunck_len;
 	uint64_t hdr_addr, hdr_phys_addr;
 	struct rte_mbuf *hdr_mbuf;
 	struct batch_copy_elem *batch_copy = vq->batch_copy_elems;
+	struct virtio_net_hdr_mrg_rxbuf tmp_hdr, *hdr = NULL;
 	uint16_t copy_nb = vq->batch_copy_nb_elems;
 	int error = 0;
 
@@ -541,9 +685,12 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		goto out;
 	}
 
-	desc_addr = vhost_iova_to_vva(dev, vq, buf_vec[vec_idx].buf_addr,
-						buf_vec[vec_idx].buf_len,
-						VHOST_ACCESS_RW);
+	desc_chunck_len = buf_vec[vec_idx].buf_len;
+	desc_gaddr = buf_vec[vec_idx].buf_addr;
+	desc_addr = vhost_iova_to_vva(dev, vq,
+					desc_gaddr,
+					&desc_chunck_len,
+					VHOST_ACCESS_RW);
 	if (buf_vec[vec_idx].buf_len < dev->vhost_hlen || !desc_addr) {
 		error = -1;
 		goto out;
@@ -551,14 +698,35 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
 
 	hdr_mbuf = m;
 	hdr_addr = desc_addr;
-	hdr_phys_addr = buf_vec[vec_idx].buf_addr;
+	if (unlikely(desc_chunck_len < dev->vhost_hlen))
+		hdr = &tmp_hdr;
+	else
+		hdr = (struct virtio_net_hdr_mrg_rxbuf *)(uintptr_t)hdr_addr;
+	hdr_phys_addr = desc_gaddr;
 	rte_prefetch0((void *)(uintptr_t)hdr_addr);
 
 	LOG_DEBUG(VHOST_DATA, "(%d) RX: num merge buffers %d\n",
 		dev->vid, num_buffers);
 
 	desc_avail  = buf_vec[vec_idx].buf_len - dev->vhost_hlen;
-	desc_offset = dev->vhost_hlen;
+	if (unlikely(desc_chunck_len < dev->vhost_hlen)) {
+		desc_chunck_len = desc_avail;
+		desc_gaddr += dev->vhost_hlen;
+		desc_addr = vhost_iova_to_vva(dev, vq,
+				desc_gaddr,
+				&desc_chunck_len,
+				VHOST_ACCESS_RW);
+		if (unlikely(!desc_addr)) {
+			error = -1;
+			goto out;
+		}
+
+		desc_offset = 0;
+	} else {
+		desc_offset = dev->vhost_hlen;
+		desc_chunck_len -= dev->vhost_hlen;
+	}
+
 
 	mbuf_avail  = rte_pktmbuf_data_len(m);
 	mbuf_offset = 0;
@@ -566,10 +734,12 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		/* done with current desc buf, get the next one */
 		if (desc_avail == 0) {
 			vec_idx++;
+			desc_chunck_len = buf_vec[vec_idx].buf_len;
+			desc_gaddr = buf_vec[vec_idx].buf_addr;
 			desc_addr =
 				vhost_iova_to_vva(dev, vq,
-					buf_vec[vec_idx].buf_addr,
-					buf_vec[vec_idx].buf_len,
+					desc_gaddr,
+					&desc_chunck_len,
 					VHOST_ACCESS_RW);
 			if (unlikely(!desc_addr)) {
 				error = -1;
@@ -580,6 +750,17 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
 			rte_prefetch0((void *)(uintptr_t)desc_addr);
 			desc_offset = 0;
 			desc_avail  = buf_vec[vec_idx].buf_len;
+		} else if (unlikely(desc_chunck_len == 0)) {
+			desc_chunck_len = desc_avail;
+			desc_gaddr += desc_offset;
+			desc_addr = vhost_iova_to_vva(dev, vq,
+					desc_gaddr,
+					&desc_chunck_len, VHOST_ACCESS_RW);
+			if (unlikely(!desc_addr)) {
+				error = -1;
+				goto out;
+			}
+			desc_offset = 0;
 		}
 
 		/* done with current mbuf, get the next one */
@@ -591,30 +772,55 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		}
 
 		if (hdr_addr) {
-			struct virtio_net_hdr_mrg_rxbuf *hdr;
-
-			hdr = (struct virtio_net_hdr_mrg_rxbuf *)(uintptr_t)
-				hdr_addr;
 			virtio_enqueue_offload(hdr_mbuf, &hdr->hdr);
 			ASSIGN_UNLESS_EQUAL(hdr->num_buffers, num_buffers);
 
-			vhost_log_write(dev, hdr_phys_addr, dev->vhost_hlen);
-			PRINT_PACKET(dev, (uintptr_t)hdr_addr,
-				     dev->vhost_hlen, 0);
+			if (unlikely(hdr == &tmp_hdr)) {
+				uint64_t len;
+				uint64_t remain = dev->vhost_hlen;
+				uint64_t src = (uint64_t)(uintptr_t)hdr, dst;
+				uint64_t guest_addr = hdr_phys_addr;
+
+				while (remain) {
+					len = remain;
+					dst = vhost_iova_to_vva(dev, vq,
+							guest_addr, &len,
+							VHOST_ACCESS_RW);
+					if (unlikely(!dst || !len)) {
+						error = -1;
+						goto out;
+					}
+
+					rte_memcpy((void *)(uintptr_t)dst,
+							(void *)(uintptr_t)src,
+							len);
+
+					PRINT_PACKET(dev, (uintptr_t)dst,
+							len, 0);
+					vhost_log_write(dev, guest_addr, len);
+
+					remain -= len;
+					guest_addr += len;
+					dst += len;
+				}
+			} else {
+				PRINT_PACKET(dev, (uintptr_t)hdr_addr,
+						dev->vhost_hlen, 0);
+				vhost_log_write(dev, hdr_phys_addr,
+						dev->vhost_hlen);
+			}
 
 			hdr_addr = 0;
 		}
 
-		cpy_len = RTE_MIN(desc_avail, mbuf_avail);
+		cpy_len = RTE_MIN(desc_chunck_len, mbuf_avail);
 
 		if (likely(cpy_len > MAX_BATCH_LEN || copy_nb >= vq->size)) {
 			rte_memcpy((void *)((uintptr_t)(desc_addr +
 							desc_offset)),
 				rte_pktmbuf_mtod_offset(m, void *, mbuf_offset),
 				cpy_len);
-			vhost_log_write(dev,
-				buf_vec[vec_idx].buf_addr + desc_offset,
-				cpy_len);
+			vhost_log_write(dev, desc_gaddr + desc_offset, cpy_len);
 			PRINT_PACKET(dev, (uintptr_t)(desc_addr + desc_offset),
 				cpy_len, 0);
 		} else {
@@ -622,8 +828,7 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
 				(void *)((uintptr_t)(desc_addr + desc_offset));
 			batch_copy[copy_nb].src =
 				rte_pktmbuf_mtod_offset(m, void *, mbuf_offset);
-			batch_copy[copy_nb].log_addr =
-				buf_vec[vec_idx].buf_addr + desc_offset;
+			batch_copy[copy_nb].log_addr = desc_gaddr + desc_offset;
 			batch_copy[copy_nb].len = cpy_len;
 			copy_nb++;
 		}
@@ -632,6 +837,7 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		mbuf_offset += cpy_len;
 		desc_avail  -= cpy_len;
 		desc_offset += cpy_len;
+		desc_chunck_len -= cpy_len;
 	}
 
 out:
@@ -907,11 +1113,13 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		  struct rte_mempool *mbuf_pool)
 {
 	struct vring_desc *desc;
-	uint64_t desc_addr;
+	uint64_t desc_addr, desc_gaddr;
 	uint32_t desc_avail, desc_offset;
 	uint32_t mbuf_avail, mbuf_offset;
 	uint32_t cpy_len;
+	uint64_t desc_chunck_len;
 	struct rte_mbuf *cur = m, *prev = m;
+	struct virtio_net_hdr tmp_hdr;
 	struct virtio_net_hdr *hdr = NULL;
 	/* A counter to avoid desc dead loop chain */
 	uint32_t nr_desc = 1;
@@ -926,9 +1134,11 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		goto out;
 	}
 
+	desc_chunck_len = desc->len;
+	desc_gaddr = desc->addr;
 	desc_addr = vhost_iova_to_vva(dev,
-					vq, desc->addr,
-					desc->len,
+					vq, desc_gaddr,
+					&desc_chunck_len,
 					VHOST_ACCESS_RO);
 	if (unlikely(!desc_addr)) {
 		error = -1;
@@ -936,8 +1146,40 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 	}
 
 	if (virtio_net_with_host_offload(dev)) {
-		hdr = (struct virtio_net_hdr *)((uintptr_t)desc_addr);
-		rte_prefetch0(hdr);
+		if (unlikely(desc_chunck_len < sizeof(struct virtio_net_hdr))) {
+			uint64_t len = desc_chunck_len;
+			uint64_t remain = sizeof(struct virtio_net_hdr);
+			uint64_t src = desc_addr;
+			uint64_t dst = (uint64_t)(uintptr_t)&tmp_hdr;
+			uint64_t guest_addr = desc_gaddr;
+
+			/*
+			 * No luck, the virtio-net header doesn't fit
+			 * in a contiguous virtual area.
+			 */
+			while (remain) {
+				len = remain;
+				src = vhost_iova_to_vva(dev, vq,
+						guest_addr, &len,
+						VHOST_ACCESS_RO);
+				if (unlikely(!src || !len)) {
+					error = -1;
+					goto out;
+				}
+
+				rte_memcpy((void *)(uintptr_t)dst,
+						   (void *)(uintptr_t)src, len);
+
+				guest_addr += len;
+				remain -= len;
+				dst += len;
+			}
+
+			hdr = &tmp_hdr;
+		} else {
+			hdr = (struct virtio_net_hdr *)((uintptr_t)desc_addr);
+			rte_prefetch0(hdr);
+		}
 	}
 
 	/*
@@ -953,9 +1195,11 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 			goto out;
 		}
 
+		desc_chunck_len = desc->len;
+		desc_gaddr = desc->addr;
 		desc_addr = vhost_iova_to_vva(dev,
-							vq, desc->addr,
-							desc->len,
+							vq, desc_gaddr,
+							&desc_chunck_len,
 							VHOST_ACCESS_RO);
 		if (unlikely(!desc_addr)) {
 			error = -1;
@@ -967,19 +1211,37 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		nr_desc    += 1;
 	} else {
 		desc_avail  = desc->len - dev->vhost_hlen;
-		desc_offset = dev->vhost_hlen;
+
+		if (unlikely(desc_chunck_len < dev->vhost_hlen)) {
+			desc_chunck_len = desc_avail;
+			desc_gaddr += dev->vhost_hlen;
+			desc_addr = vhost_iova_to_vva(dev,
+					vq, desc_gaddr,
+					&desc_chunck_len,
+					VHOST_ACCESS_RO);
+			if (unlikely(!desc_addr)) {
+				error = -1;
+				goto out;
+			}
+
+			desc_offset = 0;
+		} else {
+			desc_offset = dev->vhost_hlen;
+			desc_chunck_len -= dev->vhost_hlen;
+		}
 	}
 
 	rte_prefetch0((void *)(uintptr_t)(desc_addr + desc_offset));
 
-	PRINT_PACKET(dev, (uintptr_t)(desc_addr + desc_offset), desc_avail, 0);
+	PRINT_PACKET(dev, (uintptr_t)(desc_addr + desc_offset),
+			desc_chunck_len, 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);
+		cpy_len = RTE_MIN(desc_chunck_len, mbuf_avail);
 
 		/*
 		 * A desc buf might across two host physical pages that are
@@ -987,7 +1249,7 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		 * 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)))) {
+					desc_gaddr + desc_offset, cpy_len)))) {
 			cur->data_len = cpy_len;
 			cur->data_off = 0;
 			cur->buf_addr = (void *)(uintptr_t)(desc_addr
@@ -1002,7 +1264,8 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		} else {
 			if (likely(cpy_len > MAX_BATCH_LEN ||
 				   copy_nb >= vq->size ||
-				   (hdr && cur == m))) {
+				   (hdr && cur == m) ||
+				   desc->len != desc_chunck_len)) {
 				rte_memcpy(rte_pktmbuf_mtod_offset(cur, void *,
 								   mbuf_offset),
 					   (void *)((uintptr_t)(desc_addr +
@@ -1023,6 +1286,7 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		mbuf_avail  -= cpy_len;
 		mbuf_offset += cpy_len;
 		desc_avail  -= cpy_len;
+		desc_chunck_len -= cpy_len;
 		desc_offset += cpy_len;
 
 		/* This desc reaches to its end, get the next one */
@@ -1041,9 +1305,11 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 				goto out;
 			}
 
+			desc_chunck_len = desc->len;
+			desc_gaddr = desc->addr;
 			desc_addr = vhost_iova_to_vva(dev,
-							vq, desc->addr,
-							desc->len,
+							vq, desc_gaddr,
+							&desc_chunck_len,
 							VHOST_ACCESS_RO);
 			if (unlikely(!desc_addr)) {
 				error = -1;
@@ -1055,7 +1321,23 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 			desc_offset = 0;
 			desc_avail  = desc->len;
 
-			PRINT_PACKET(dev, (uintptr_t)desc_addr, desc->len, 0);
+			PRINT_PACKET(dev, (uintptr_t)desc_addr,
+					desc_chunck_len, 0);
+		} else if (unlikely(desc_chunck_len == 0)) {
+			desc_chunck_len = desc_avail;
+			desc_gaddr += desc_offset;
+			desc_addr = vhost_iova_to_vva(dev, vq,
+					desc_gaddr,
+					&desc_chunck_len,
+					VHOST_ACCESS_RO);
+			if (unlikely(!desc_addr)) {
+				error = -1;
+				goto out;
+			}
+			desc_offset = 0;
+
+			PRINT_PACKET(dev, (uintptr_t)desc_addr,
+					desc_chunck_len, 0);
 		}
 
 		/*
@@ -1317,22 +1599,37 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
 	/* Prefetch descriptor index. */
 	rte_prefetch0(&vq->desc[desc_indexes[0]]);
 	for (i = 0; i < count; i++) {
-		struct vring_desc *desc;
+		struct vring_desc *desc, *idesc = NULL;
 		uint16_t sz, idx;
+		uint64_t dlen;
 		int err;
 
 		if (likely(i + 1 < count))
 			rte_prefetch0(&vq->desc[desc_indexes[i + 1]]);
 
 		if (vq->desc[desc_indexes[i]].flags & VRING_DESC_F_INDIRECT) {
+			dlen = vq->desc[desc_indexes[i]].len;
 			desc = (struct vring_desc *)(uintptr_t)
 				vhost_iova_to_vva(dev, vq,
 						vq->desc[desc_indexes[i]].addr,
-						sizeof(*desc),
+						&dlen,
 						VHOST_ACCESS_RO);
 			if (unlikely(!desc))
 				break;
 
+			if (unlikely(dlen < vq->desc[desc_indexes[i]].len)) {
+				/*
+				 * The indirect desc table is not contiguous
+				 * in process VA space, we have to copy it.
+				 */
+				idesc = alloc_copy_ind_table(dev, vq,
+						&vq->desc[desc_indexes[i]]);
+				if (unlikely(!idesc))
+					break;
+
+				desc = idesc;
+			}
+
 			rte_prefetch0(desc);
 			sz = vq->desc[desc_indexes[i]].len / sizeof(*desc);
 			idx = 0;
@@ -1346,6 +1643,7 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
 		if (unlikely(pkts[i] == NULL)) {
 			RTE_LOG(ERR, VHOST_DATA,
 				"Failed to allocate memory for mbuf.\n");
+			free_ind_table(idesc);
 			break;
 		}
 
@@ -1353,6 +1651,7 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
 					mbuf_pool);
 		if (unlikely(err)) {
 			rte_pktmbuf_free(pkts[i]);
+			free_ind_table(idesc);
 			break;
 		}
 
@@ -1362,6 +1661,7 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
 			zmbuf = get_zmbuf(vq);
 			if (!zmbuf) {
 				rte_pktmbuf_free(pkts[i]);
+				free_ind_table(idesc);
 				break;
 			}
 			zmbuf->mbuf = pkts[i];
@@ -1378,6 +1678,9 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
 			vq->nr_zmbuf += 1;
 			TAILQ_INSERT_TAIL(&vq->zmbuf_list, zmbuf, next);
 		}
+
+		if (unlikely(!!idesc))
+			free_ind_table(idesc);
 	}
 	vq->last_avail_idx += i;