[dpdk-stable] [PATCH v17.08 5/9] vhost: handle virtually non-contiguous buffers in Rx

[Maxime Coquelin]

This patch enables the handling of buffers non-contiguous in
process virtual address space in the enqueue path when mergeable
buffers aren't used.

When virtio-net header doesn't fit in a single chunck, it is
computed in a local variable and copied to the buffer chuncks
afterwards.

For packet content, the copy length is limited to the chunck
size, next chuncks VAs being fetched afterward.

This issue has been assigned CVE-2018-1059.

Signed-off-by: Maxime Coquelin <maxime.coquelin at redhat.com>
---
 lib/librte_vhost/virtio_net.c | 85 ++++++++++++++++++++++++++++++++++---------
 1 file changed, 68 insertions(+), 17 deletions(-)

diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c
index 22fcbb3fb..b3070c9df 100644
--- a/lib/librte_vhost/virtio_net.c
+++ b/lib/librte_vhost/virtio_net.c
@@ -213,33 +213,72 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vring_desc *descs,
 	uint32_t desc_avail, desc_offset;
 	uint32_t mbuf_avail, mbuf_offset;
 	uint32_t cpy_len;
-	uint64_t dlen;
+	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;
 
 	desc = &descs[desc_idx];
-	dlen = desc->len;
-	desc_addr = rte_vhost_va_from_guest_pa(dev->mem, desc->addr,
-					&dlen);
+	desc_chunck_len = desc->len;
+	desc_gaddr = desc->addr;
+	desc_addr = rte_vhost_va_from_guest_pa(dev->mem, desc_gaddr,
+					&desc_chunck_len);
 	/*
 	 * 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
 	 * otherwise stores offset on the stack instead of in a register.
 	 */
-	if (unlikely(dlen != desc->len || desc->len < dev->vhost_hlen) ||
-			!desc_addr)
+	if (unlikely(desc->len < dev->vhost_hlen) || !desc_addr)
 		return -1;
 
 	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 = rte_vhost_va_from_guest_pa(dev->mem, guest_addr, &len);
+			if (unlikely(!dst || !len))
+				return -1;
+
+			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 = rte_vhost_va_from_guest_pa(dev->mem, desc_gaddr,
+				&desc_chunck_len);
+		if (unlikely(!desc_addr))
+			return -1;
+
+		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;
@@ -262,21 +301,32 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vring_desc *descs,
 				return -1;
 
 			desc = &descs[desc->next];
-			dlen = desc->len;
-			desc_addr = rte_vhost_va_from_guest_pa(dev->mem, desc->addr,
-					&dlen);
-			if (unlikely(!desc_addr || dlen != desc->len))
+			desc_chunck_len = desc->len;
+			desc_gaddr = desc->addr;
+			desc_addr = rte_vhost_va_from_guest_pa(dev->mem,
+							desc_gaddr,
+							&desc_chunck_len);
+			if (unlikely(!desc_addr))
 				return -1;
 
 			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 = rte_vhost_va_from_guest_pa(dev->mem,
+					desc_gaddr,
+					&desc_chunck_len);
+			if (unlikely(!desc_addr))
+				return -1;
+			desc_offset = 0;
 		}
 
-		cpy_len = RTE_MIN(desc_avail, mbuf_avail);
+		cpy_len = RTE_MIN(desc_chunck_len, mbuf_avail);
 		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);
 
@@ -284,6 +334,7 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vring_desc *descs,
 		mbuf_offset += cpy_len;
 		desc_avail  -= cpy_len;
 		desc_offset += cpy_len;
+		desc_chunck_len -= cpy_len;
 	}
 
 	return 0;
-- 
2.14.3