hv_netvsc: Copy packets sent by Hyper-V out of the receive buffer

Pointers to receive-buffer packets sent by Hyper-V are used within the
guest VM. Hyper-V can send packets with erroneous values or modify
packet fields after they are processed by the guest. To defend against
these scenarios, copy (sections of) the incoming packet after validating
their length and offset fields in netvsc_filter_receive(). In this way,
the packet can no longer be modified by the host.

Reported-by: Juan Vazquez <[e-mail]>
Signed-off-by: Andrea Parri (Microsoft) <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]
Signed-off-by: Jakub Kicinski <[e-mail]>

drivers/net/hyperv/hyperv_net.h

···
 	u32 processor_masks_entry_size;
 };
 
-/* Fwd declaration */
-struct ndis_tcp_ip_checksum_info;
-struct ndis_pkt_8021q_info;
+struct ndis_tcp_ip_checksum_info {
+	union {
+		struct {
+			u32 is_ipv4:1;
+			u32 is_ipv6:1;
+			u32 tcp_checksum:1;
+			u32 udp_checksum:1;
+			u32 ip_header_checksum:1;
+			u32 reserved:11;
+			u32 tcp_header_offset:10;
+		} transmit;
+		struct {
+			u32 tcp_checksum_failed:1;
+			u32 udp_checksum_failed:1;
+			u32 ip_checksum_failed:1;
+			u32 tcp_checksum_succeeded:1;
+			u32 udp_checksum_succeeded:1;
+			u32 ip_checksum_succeeded:1;
+			u32 loopback:1;
+			u32 tcp_checksum_value_invalid:1;
+			u32 ip_checksum_value_invalid:1;
+		} receive;
+		u32  value;
+	};
+};
+
+struct ndis_pkt_8021q_info {
+	union {
+		struct {
+			u32 pri:3; /* User Priority */
+			u32 cfi:1; /* Canonical Format ID */
+			u32 vlanid:12; /* VLAN ID */
+			u32 reserved:16;
+		};
+		u32 value;
+	};
+};
 
 /*
  * Represent netvsc packet which contains 1 RNDIS and 1 ethernet frame
···
 		struct sk_buff *skb,
 		bool xdp_tx);
 void netvsc_linkstatus_callback(struct net_device *net,
-				struct rndis_message *resp);
+				struct rndis_message *resp,
+				void *data);
 int netvsc_recv_callback(struct net_device *net,
 			 struct netvsc_device *nvdev,
 			 struct netvsc_channel *nvchan);
···
 #define NVSP_RSC_MAX 562 /* Max #RSC frags in a vmbus xfer page pkt */
 
 struct nvsc_rsc {
-	const struct ndis_pkt_8021q_info *vlan;
-	const struct ndis_tcp_ip_checksum_info *csum_info;
-	const u32 *hash_info;
+	struct ndis_pkt_8021q_info vlan;
+	struct ndis_tcp_ip_checksum_info csum_info;
+	u32 hash_info;
+	u8 ppi_flags; /* valid/present bits for the above PPIs */
 	u8 is_last; /* last RNDIS msg in a vmtransfer_page */
 	u32 cnt; /* #fragments in an RSC packet */
 	u32 pktlen; /* Full packet length */
 	void *data[NVSP_RSC_MAX];
 	u32 len[NVSP_RSC_MAX];
 };
+
+#define NVSC_RSC_VLAN		BIT(0)	/* valid/present bit for 'vlan' */
+#define NVSC_RSC_CSUM_INFO	BIT(1)	/* valid/present bit for 'csum_info' */
+#define NVSC_RSC_HASH_INFO	BIT(2)	/* valid/present bit for 'hash_info' */
 
 struct netvsc_stats {
 	u64 packets;
···
 struct netvsc_channel {
 	struct vmbus_channel *channel;
 	struct netvsc_device *net_device;
+	void *recv_buf; /* buffer to copy packets out from the receive buffer */
 	const struct vmpacket_descriptor *desc;
 	struct napi_struct napi;
 	struct multi_send_data msd;
···
 	u16 pkt_id;
 };
 
-struct ndis_pkt_8021q_info {
-	union {
-		struct {
-			u32 pri:3; /* User Priority */
-			u32 cfi:1; /* Canonical Format ID */
-			u32 vlanid:12; /* VLAN ID */
-			u32 reserved:16;
-		};
-		u32 value;
-	};
-};
-
 struct ndis_object_header {
 	u8 type;
 	u8 revision;
···
 	struct {
 		u8 encapsulated_packet_task_offload;
 		u8 encapsulation_types;
-	};
-};
-
-struct ndis_tcp_ip_checksum_info {
-	union {
-		struct {
-			u32 is_ipv4:1;
-			u32 is_ipv6:1;
-			u32 tcp_checksum:1;
-			u32 udp_checksum:1;
-			u32 ip_header_checksum:1;
-			u32 reserved:11;
-			u32 tcp_header_offset:10;
-		} transmit;
-		struct {
-			u32 tcp_checksum_failed:1;
-			u32 udp_checksum_failed:1;
-			u32 ip_checksum_failed:1;
-			u32 tcp_checksum_succeeded:1;
-			u32 udp_checksum_succeeded:1;
-			u32 ip_checksum_succeeded:1;
-			u32 loopback:1;
-			u32 tcp_checksum_value_invalid:1;
-			u32 ip_checksum_value_invalid:1;
-		} receive;
-		u32  value;
 	};
 };
 

drivers/net/hyperv/netvsc.c

···
 
 	for (i = 0; i < VRSS_CHANNEL_MAX; i++) {
 		xdp_rxq_info_unreg(&nvdev->chan_table[i].xdp_rxq);
+		kfree(nvdev->chan_table[i].recv_buf);
 		vfree(nvdev->chan_table[i].mrc.slots);
 	}
 
···
 			continue;
 		}
 
+		/* We're going to copy (sections of) the packet into nvchan->recv_buf;
+		 * make sure that nvchan->recv_buf is large enough to hold the packet.
+		 */
+		if (unlikely(buflen > net_device->recv_section_size)) {
+			nvchan->rsc.cnt = 0;
+			status = NVSP_STAT_FAIL;
+			netif_err(net_device_ctx, rx_err, ndev,
+				  "Packet too big: buflen=%u recv_section_size=%u\n",
+				  buflen, net_device->recv_section_size);
+
+			continue;
+		}
+
 		data = recv_buf + offset;
 
 		nvchan->rsc.is_last = (i == count - 1);
···
 
 	for (i = 0; i < VRSS_CHANNEL_MAX; i++) {
 		struct netvsc_channel *nvchan = &net_device->chan_table[i];
+
+		nvchan->recv_buf = kzalloc(device_info->recv_section_size, GFP_KERNEL);
+		if (nvchan->recv_buf == NULL) {
+			ret = -ENOMEM;
+			goto cleanup2;
+		}
 
 		nvchan->channel = device->channel;
 		nvchan->net_device = net_device;

drivers/net/hyperv/netvsc_drv.c

···
  * netvsc_linkstatus_callback - Link up/down notification
  */
 void netvsc_linkstatus_callback(struct net_device *net,
-				struct rndis_message *resp)
+				struct rndis_message *resp,
+				void *data)
 {
 	struct rndis_indicate_status *indicate = &resp->msg.indicate_status;
 	struct net_device_context *ndev_ctx = netdev_priv(net);
···
 			   resp->msg_len);
 		return;
 	}
+
+	/* Copy the RNDIS indicate status into nvchan->recv_buf */
+	memcpy(indicate, data + RNDIS_HEADER_SIZE, sizeof(*indicate));
 
 	/* Update the physical link speed when changing to another vSwitch */
 	if (indicate->status == RNDIS_STATUS_LINK_SPEED_CHANGE) {
···
 			return;
 		}
 
-		speed = *(u32 *)((void *)indicate
-				 + indicate->status_buf_offset) / 10000;
+		speed = *(u32 *)(data + RNDIS_HEADER_SIZE + indicate->status_buf_offset) / 10000;
 		ndev_ctx->speed = speed;
 		return;
 	}
···
 					     struct xdp_buff *xdp)
 {
 	struct napi_struct *napi = &nvchan->napi;
-	const struct ndis_pkt_8021q_info *vlan = nvchan->rsc.vlan;
+	const struct ndis_pkt_8021q_info *vlan = &nvchan->rsc.vlan;
 	const struct ndis_tcp_ip_checksum_info *csum_info =
-						nvchan->rsc.csum_info;
-	const u32 *hash_info = nvchan->rsc.hash_info;
+						&nvchan->rsc.csum_info;
+	const u32 *hash_info = &nvchan->rsc.hash_info;
+	u8 ppi_flags = nvchan->rsc.ppi_flags;
 	struct sk_buff *skb;
 	void *xbuf = xdp->data_hard_start;
 	int i;
···
 	 * We compute it here if the flags are set, because on Linux, the IP
 	 * checksum is always checked.
 	 */
-	if (csum_info && csum_info->receive.ip_checksum_value_invalid &&
+	if ((ppi_flags & NVSC_RSC_CSUM_INFO) && csum_info->receive.ip_checksum_value_invalid &&
 	    csum_info->receive.ip_checksum_succeeded &&
 	    skb->protocol == htons(ETH_P_IP)) {
 		/* Check that there is enough space to hold the IP header. */
···
 	}
 
 	/* Do L4 checksum offload if enabled and present. */
-	if (csum_info && (net->features & NETIF_F_RXCSUM)) {
+	if ((ppi_flags & NVSC_RSC_CSUM_INFO) && (net->features & NETIF_F_RXCSUM)) {
 		if (csum_info->receive.tcp_checksum_succeeded ||
 		    csum_info->receive.udp_checksum_succeeded)
 			skb->ip_summed = CHECKSUM_UNNECESSARY;
 	}
 
-	if (hash_info && (net->features & NETIF_F_RXHASH))
+	if ((ppi_flags & NVSC_RSC_HASH_INFO) && (net->features & NETIF_F_RXHASH))
 		skb_set_hash(skb, *hash_info, PKT_HASH_TYPE_L4);
 
-	if (vlan) {
+	if (ppi_flags & NVSC_RSC_VLAN) {
 		u16 vlan_tci = vlan->vlanid | (vlan->pri << VLAN_PRIO_SHIFT) |
 			(vlan->cfi ? VLAN_CFI_MASK : 0);
 

drivers/net/hyperv/rndis_filter.c

···
 }
 
 static void dump_rndis_message(struct net_device *netdev,
-			       const struct rndis_message *rndis_msg)
+			       const struct rndis_message *rndis_msg,
+			       const void *data)
 {
 	switch (rndis_msg->ndis_msg_type) {
 	case RNDIS_MSG_PACKET:
 		if (rndis_msg->msg_len - RNDIS_HEADER_SIZE >= sizeof(struct rndis_packet)) {
-			const struct rndis_packet *pkt = &rndis_msg->msg.pkt;
+			const struct rndis_packet *pkt = data + RNDIS_HEADER_SIZE;
 			netdev_dbg(netdev, "RNDIS_MSG_PACKET (len %u, "
 				   "data offset %u data len %u, # oob %u, "
 				   "oob offset %u, oob len %u, pkt offset %u, "
···
 		if (rndis_msg->msg_len - RNDIS_HEADER_SIZE >=
 				sizeof(struct rndis_initialize_complete)) {
 			const struct rndis_initialize_complete *init_complete =
-				&rndis_msg->msg.init_complete;
+				data + RNDIS_HEADER_SIZE;
 			netdev_dbg(netdev, "RNDIS_MSG_INIT_C "
 				"(len %u, id 0x%x, status 0x%x, major %d, minor %d, "
 				"device flags %d, max xfer size 0x%x, max pkts %u, "
···
 		if (rndis_msg->msg_len - RNDIS_HEADER_SIZE >=
 				sizeof(struct rndis_query_complete)) {
 			const struct rndis_query_complete *query_complete =
-				&rndis_msg->msg.query_complete;
+				data + RNDIS_HEADER_SIZE;
 			netdev_dbg(netdev, "RNDIS_MSG_QUERY_C "
 				"(len %u, id 0x%x, status 0x%x, buf len %u, "
 				"buf offset %u)\n",
···
 	case RNDIS_MSG_SET_C:
 		if (rndis_msg->msg_len - RNDIS_HEADER_SIZE + sizeof(struct rndis_set_complete)) {
 			const struct rndis_set_complete *set_complete =
-				&rndis_msg->msg.set_complete;
+				data + RNDIS_HEADER_SIZE;
 			netdev_dbg(netdev,
 				"RNDIS_MSG_SET_C (len %u, id 0x%x, status 0x%x)\n",
 				rndis_msg->msg_len,
···
 		if (rndis_msg->msg_len - RNDIS_HEADER_SIZE >=
 				sizeof(struct rndis_indicate_status)) {
 			const struct rndis_indicate_status *indicate_status =
-				&rndis_msg->msg.indicate_status;
+				data + RNDIS_HEADER_SIZE;
 			netdev_dbg(netdev, "RNDIS_MSG_INDICATE "
 				"(len %u, status 0x%x, buf len %u, buf offset %u)\n",
 				rndis_msg->msg_len,
···
 
 static void rndis_filter_receive_response(struct net_device *ndev,
 					  struct netvsc_device *nvdev,
-					  const struct rndis_message *resp)
+					  struct rndis_message *resp,
+					  void *data)
 {
+	u32 *req_id = &resp->msg.init_complete.req_id;
 	struct rndis_device *dev = nvdev->extension;
 	struct rndis_request *request = NULL;
 	bool found = false;
···
 		return;
 	}
 
+	/* Copy the request ID into nvchan->recv_buf */
+	*req_id = *(u32 *)(data + RNDIS_HEADER_SIZE);
+
 	spin_lock_irqsave(&dev->request_lock, flags);
 	list_for_each_entry(request, &dev->req_list, list_ent) {
 		/*
 		 * All request/response message contains RequestId as the 1st
 		 * field
 		 */
-		if (request->request_msg.msg.init_req.req_id
-		    == resp->msg.init_complete.req_id) {
+		if (request->request_msg.msg.init_req.req_id == *req_id) {
 			found = true;
 			break;
 		}
···
 	if (found) {
 		if (resp->msg_len <=
 		    sizeof(struct rndis_message) + RNDIS_EXT_LEN) {
-			memcpy(&request->response_msg, resp,
-			       resp->msg_len);
+			memcpy(&request->response_msg, resp, RNDIS_HEADER_SIZE + sizeof(*req_id));
+			memcpy((void *)&request->response_msg + RNDIS_HEADER_SIZE + sizeof(*req_id),
+			       data + RNDIS_HEADER_SIZE + sizeof(*req_id),
+			       resp->msg_len - RNDIS_HEADER_SIZE - sizeof(*req_id));
 			if (request->request_msg.ndis_msg_type ==
 			    RNDIS_MSG_QUERY && request->request_msg.msg.
 			    query_req.oid == RNDIS_OID_GEN_MEDIA_CONNECT_STATUS)
···
 		netdev_err(ndev,
 			"no rndis request found for this response "
 			"(id 0x%x res type 0x%x)\n",
-			resp->msg.init_complete.req_id,
+			*req_id,
 			resp->ndis_msg_type);
 	}
 }
···
 static inline void *rndis_get_ppi(struct net_device *ndev,
 				  struct rndis_packet *rpkt,
 				  u32 rpkt_len, u32 type, u8 internal,
-				  u32 ppi_size)
+				  u32 ppi_size, void *data)
 {
 	struct rndis_per_packet_info *ppi;
 	int len;
···
 
 	ppi = (struct rndis_per_packet_info *)((ulong)rpkt +
 		rpkt->per_pkt_info_offset);
+	/* Copy the PPIs into nvchan->recv_buf */
+	memcpy(ppi, data + RNDIS_HEADER_SIZE + rpkt->per_pkt_info_offset, rpkt->per_pkt_info_len);
 	len = rpkt->per_pkt_info_len;
 
 	while (len > 0) {
···
 	if (cnt) {
 		nvchan->rsc.pktlen += len;
 	} else {
-		nvchan->rsc.vlan = vlan;
-		nvchan->rsc.csum_info = csum_info;
+		/* The data/values pointed by vlan, csum_info and hash_info are shared
+		 * across the different 'fragments' of the RSC packet; store them into
+		 * the packet itself.
+		 */
+		if (vlan != NULL) {
+			memcpy(&nvchan->rsc.vlan, vlan, sizeof(*vlan));
+			nvchan->rsc.ppi_flags |= NVSC_RSC_VLAN;
+		} else {
+			nvchan->rsc.ppi_flags &= ~NVSC_RSC_VLAN;
+		}
+		if (csum_info != NULL) {
+			memcpy(&nvchan->rsc.csum_info, csum_info, sizeof(*csum_info));
+			nvchan->rsc.ppi_flags |= NVSC_RSC_CSUM_INFO;
+		} else {
+			nvchan->rsc.ppi_flags &= ~NVSC_RSC_CSUM_INFO;
+		}
 		nvchan->rsc.pktlen = len;
-		nvchan->rsc.hash_info = hash_info;
+		if (hash_info != NULL) {
+			nvchan->rsc.csum_info = *csum_info;
+			nvchan->rsc.ppi_flags |= NVSC_RSC_HASH_INFO;
+		} else {
+			nvchan->rsc.ppi_flags &= ~NVSC_RSC_HASH_INFO;
+		}
 	}
 
 	nvchan->rsc.data[cnt] = data;
···
 				     struct netvsc_device *nvdev,
 				     struct netvsc_channel *nvchan,
 				     struct rndis_message *msg,
-				     u32 data_buflen)
+				     void *data, u32 data_buflen)
 {
 	struct rndis_packet *rndis_pkt = &msg->msg.pkt;
 	const struct ndis_tcp_ip_checksum_info *csum_info;
···
 	const struct rndis_pktinfo_id *pktinfo_id;
 	const u32 *hash_info;
 	u32 data_offset, rpkt_len;
-	void *data;
 	bool rsc_more = false;
 	int ret;
 
···
 			   data_buflen);
 		return NVSP_STAT_FAIL;
 	}
+
+	/* Copy the RNDIS packet into nvchan->recv_buf */
+	memcpy(rndis_pkt, data + RNDIS_HEADER_SIZE, sizeof(*rndis_pkt));
 
 	/* Validate rndis_pkt offset */
 	if (rndis_pkt->data_offset >= data_buflen - RNDIS_HEADER_SIZE) {
···
 		return NVSP_STAT_FAIL;
 	}
 
-	vlan = rndis_get_ppi(ndev, rndis_pkt, rpkt_len, IEEE_8021Q_INFO, 0, sizeof(*vlan));
+	vlan = rndis_get_ppi(ndev, rndis_pkt, rpkt_len, IEEE_8021Q_INFO, 0, sizeof(*vlan),
+			     data);
 
 	csum_info = rndis_get_ppi(ndev, rndis_pkt, rpkt_len, TCPIP_CHKSUM_PKTINFO, 0,
-				  sizeof(*csum_info));
+				  sizeof(*csum_info), data);
 
 	hash_info = rndis_get_ppi(ndev, rndis_pkt, rpkt_len, NBL_HASH_VALUE, 0,
-				  sizeof(*hash_info));
+				  sizeof(*hash_info), data);
 
 	pktinfo_id = rndis_get_ppi(ndev, rndis_pkt, rpkt_len, RNDIS_PKTINFO_ID, 1,
-				   sizeof(*pktinfo_id));
-
-	data = (void *)msg + data_offset;
+				   sizeof(*pktinfo_id), data);
 
 	/* Identify RSC frags, drop erroneous packets */
 	if (pktinfo_id && (pktinfo_id->flag & RNDIS_PKTINFO_SUBALLOC)) {
···
 	 * the data packet to the stack, without the rndis trailer padding
 	 */
 	rsc_add_data(nvchan, vlan, csum_info, hash_info,
-		     data, rndis_pkt->data_len);
+		     data + data_offset, rndis_pkt->data_len);
 
 	if (rsc_more)
 		return NVSP_STAT_SUCCESS;
···
 			 void *data, u32 buflen)
 {
 	struct net_device_context *net_device_ctx = netdev_priv(ndev);
-	struct rndis_message *rndis_msg = data;
+	struct rndis_message *rndis_msg = nvchan->recv_buf;
+
+	if (buflen < RNDIS_HEADER_SIZE) {
+		netdev_err(ndev, "Invalid rndis_msg (buflen: %u)\n", buflen);
+		return NVSP_STAT_FAIL;
+	}
+
+	/* Copy the RNDIS msg header into nvchan->recv_buf */
+	memcpy(rndis_msg, data, RNDIS_HEADER_SIZE);
 
 	/* Validate incoming rndis_message packet */
-	if (buflen < RNDIS_HEADER_SIZE || rndis_msg->msg_len < RNDIS_HEADER_SIZE ||
+	if (rndis_msg->msg_len < RNDIS_HEADER_SIZE ||
 	    buflen < rndis_msg->msg_len) {
 		netdev_err(ndev, "Invalid rndis_msg (buflen: %u, msg_len: %u)\n",
 			   buflen, rndis_msg->msg_len);
···
 	}
 
 	if (netif_msg_rx_status(net_device_ctx))
-		dump_rndis_message(ndev, rndis_msg);
+		dump_rndis_message(ndev, rndis_msg, data);
 
 	switch (rndis_msg->ndis_msg_type) {
 	case RNDIS_MSG_PACKET:
 		return rndis_filter_receive_data(ndev, net_dev, nvchan,
-						 rndis_msg, buflen);
+						 rndis_msg, data, buflen);
 	case RNDIS_MSG_INIT_C:
 	case RNDIS_MSG_QUERY_C:
 	case RNDIS_MSG_SET_C:
 		/* completion msgs */
-		rndis_filter_receive_response(ndev, net_dev, rndis_msg);
+		rndis_filter_receive_response(ndev, net_dev, rndis_msg, data);
 		break;
 
 	case RNDIS_MSG_INDICATE:
 		/* notification msgs */
-		netvsc_linkstatus_callback(ndev, rndis_msg);
+		netvsc_linkstatus_callback(ndev, rndis_msg, data);
 		break;
 	default:
 		netdev_err(ndev,