D2517.id5333.diff
No OneTemporary
Actions

Size

59 KB

Referenced Files

None

Subscribers

None

D2517.id5333.diff
View Options

	Index: sys/dev/hyperv/include/hyperv.h
	===================================================================
	--- sys/dev/hyperv/include/hyperv.h
	+++ sys/dev/hyperv/include/hyperv.h
	@@ -107,7 +107,7 @@
	#define HV_MAX_PIPE_USER_DEFINED_BYTES 116


	-#define HV_MAX_PAGE_BUFFER_COUNT 16
	+#define HV_MAX_PAGE_BUFFER_COUNT 32
	#define HV_MAX_MULTIPAGE_BUFFER_COUNT 32

	#define HV_ALIGN_UP(value, align) \
	Index: sys/dev/hyperv/netvsc/hv_net_vsc.h
	===================================================================
	--- sys/dev/hyperv/netvsc/hv_net_vsc.h
	+++ sys/dev/hyperv/netvsc/hv_net_vsc.h
	@@ -50,11 +50,15 @@

	#define NVSP_PROTOCOL_VERSION_1 2
	#define NVSP_PROTOCOL_VERSION_2 0x30002
	+#define NVSP_PROTOCOL_VERSION_4 0x40000
	+#define NVSP_PROTOCOL_VERSION_5 0x50000
	#define NVSP_MIN_PROTOCOL_VERSION (NVSP_PROTOCOL_VERSION_1)
	#define NVSP_MAX_PROTOCOL_VERSION (NVSP_PROTOCOL_VERSION_2)

	#define NVSP_PROTOCOL_VERSION_CURRENT NVSP_PROTOCOL_VERSION_2

	+#define VERSION_4_OFFLOAD_SIZE 22
	+
	#define NVSP_OPERATIONAL_STATUS_OK (0x00000000)
	#define NVSP_OPERATIONAL_STATUS_DEGRADED (0x00000001)
	#define NVSP_OPERATIONAL_STATUS_NONRECOVERABLE (0x00000002)
	@@ -544,7 +548,7 @@


	#define NVSP_1_CHIMNEY_SEND_INVALID_OOB_INDEX 0xffffu
	-#define NVSP_1_CHIMNEY_SEND_INVALID_SECTION_INDEX 0xffffu
	+#define NVSP_1_CHIMNEY_SEND_INVALID_SECTION_INDEX 0xffffffff

	/*
	* NvspMessage2TypeSendChimneyPacket
	@@ -842,11 +846,11 @@
	* Defines
	*/

	-#define NETVSC_SEND_BUFFER_SIZE (641024) / 64K */
	+#define NETVSC_SEND_BUFFER_SIZE (1024102415) /* 15M */
	#define NETVSC_SEND_BUFFER_ID 0xface


	-#define NETVSC_RECEIVE_BUFFER_SIZE (10241024) / 1MB */
	+#define NETVSC_RECEIVE_BUFFER_SIZE (1024102416) /* 16MB */

	#define NETVSC_RECEIVE_BUFFER_ID 0xcafe

	@@ -862,6 +866,8 @@
	*/
	#define NETVSC_MAX_CONFIGURABLE_MTU (9 * 1024)

	+#define NETVSC_PACKET_SIZE 4096
	+
	/*
	* Data types
	*/
	@@ -873,15 +879,14 @@
	struct hv_device *dev;
	int num_outstanding_sends;

	- /* List of free preallocated NETVSC_PACKET to represent RX packet */
	- STAILQ_HEAD(PQ, netvsc_packet_) myrx_packet_list;
	- struct mtx rx_pkt_list_lock;
	-
	/* Send buffer allocated by us but manages by NetVSP */
	void *send_buf;
	uint32_t send_buf_size;
	uint32_t send_buf_gpadl_handle;
	uint32_t send_section_size;
	+ uint32_t send_section_count;
	+ unsigned long bitsmap_words;
	+ unsigned long *send_section_bitsmap;

	/* Receive buffer allocated by us but managed by NetVSP */
	void *rx_buf;
	@@ -903,35 +908,43 @@
	hv_bool_uint8_t destroy;
	/* Negotiated NVSP version */
	uint32_t nvsp_version;
	+
	+ unsigned char callback_buf[NETVSC_PACKET_SIZE];
	} netvsc_dev;


	typedef void (pfn_on_send_rx_completion)(void );

	-#define NETVSC_DEVICE_RING_BUFFER_SIZE (64 * PAGE_SIZE)
	-#define NETVSC_PACKET_MAXPAGE 16
	+#define NETVSC_DEVICE_RING_BUFFER_SIZE (128 * PAGE_SIZE)
	+#define NETVSC_PACKET_MAXPAGE 32


	-typedef struct xfer_page_packet_ {
	- /*
	- * This needs to be here because the network RX code casts
	- * an instantiation of this structure to a netvsc_packet.
	- */
	- STAILQ_ENTRY(netvsc_packet_) mylist_entry;
	+#define NETVSC_VLAN_PRIO_MASK 0xe000
	+#define NETVSC_VLAN_PRIO_SHIFT 13
	+#define NETVSC_VLAN_VID_MASK 0x0fff
	+
	+#define TYPE_IPV4 2
	+#define TYPE_IPV6 4
	+#define TYPE_TCP 2
	+#define TYPE_UDP 4

	- uint32_t count;
	-} xfer_page_packet;
	+#define TRANSPORT_TYPE_NOT_IP 0
	+#define TRANSPORT_TYPE_IPV4_TCP ((TYPE_IPV4 << 16) \| TYPE_TCP)
	+#define TRANSPORT_TYPE_IPV4_UDP ((TYPE_IPV4 << 16) \| TYPE_UDP)
	+#define TRANSPORT_TYPE_IPV6_TCP ((TYPE_IPV6 << 16) \| TYPE_TCP)
	+#define TRANSPORT_TYPE_IPV6_UDP ((TYPE_IPV6 << 16) \| TYPE_UDP)
	+
	+#ifdef __i386__
	+#define BITS_PER_LONG 32
	+#else
	+#define BITS_PER_LONG 64
	+#endif

	typedef struct netvsc_packet_ {
	- /*
	- * List used when enqueued on &net_dev->rx_packet_list,
	- * and when enqueued within the netvsc code
	- */
	- STAILQ_ENTRY(netvsc_packet_) mylist_entry;
	struct hv_device *device;
	hv_bool_uint8_t is_data_pkt; /* One byte */
	uint16_t vlan_tci;
	- xfer_page_packet *xfer_page_pkt;
	+ uint32_t status;

	/* Completion */
	union {
	@@ -948,9 +961,12 @@
	pfn_on_send_rx_completion on_send_completion;
	} send;
	} compl;
	+ uint32_t send_buf_section_idx;
	+ uint32_t send_buf_section_size;

	- void *extension;
	+ void *rndis_mesg;
	uint32_t tot_data_buf_len;
	+ void *data;
	uint32_t page_buf_count;
	hv_vmbus_page_buffer page_buffers[NETVSC_PACKET_MAXPAGE];
	} netvsc_packet;
	@@ -984,15 +1000,18 @@
	extern int hv_promisc_mode;

	extern void netvsc_linkstatus_callback(struct hv_device *device_obj,
	- uint32_t status);
	-extern int netvsc_recv(struct hv_device device_obj, netvsc_packet packet);
	+ uint32_t status);
	extern void netvsc_xmit_completion(void *context);
	-
	-extern void hv_nv_on_receive_completion(void *context);
	+extern void
	+hv_nv_on_receive_completion(struct hv_device *device,
	+ uint64_t tid,
	+ uint32_t status);
	extern netvsc_dev hv_nv_on_device_add(struct hv_device device, void *additional_info);
	extern int hv_nv_on_device_remove(struct hv_device *device,
	- boolean_t destroy_channel);
	+ boolean_t destroy_channel);
	extern int hv_nv_on_send(struct hv_device device, netvsc_packet pkt);

	+extern int hv_nv_get_next_send_section(netvsc_dev *net_dev);
	+
	#endif /* __HV_NET_VSC_H__ */

	Index: sys/dev/hyperv/netvsc/hv_net_vsc.c
	===================================================================
	--- sys/dev/hyperv/netvsc/hv_net_vsc.c
	+++ sys/dev/hyperv/netvsc/hv_net_vsc.c
	@@ -59,13 +59,12 @@
	static int hv_nv_destroy_send_buffer(netvsc_dev *net_dev);
	static int hv_nv_destroy_rx_buffer(netvsc_dev *net_dev);
	static int hv_nv_connect_to_vsp(struct hv_device *device);
	-static void hv_nv_on_send_completion(struct hv_device *device,
	- hv_vm_packet_descriptor *pkt);
	-static void hv_nv_on_receive(struct hv_device *device,
	- hv_vm_packet_descriptor *pkt);
	-static void hv_nv_send_receive_completion(struct hv_device *device,
	- uint64_t tid);
	-
	+static void hv_nv_on_send_completion(netvsc_dev *net_dev,
	+ struct hv_device *device,
	+ hv_vm_packet_descriptor *pkt);
	+static void hv_nv_on_receive(netvsc_dev *net_dev,
	+ struct hv_device *device,
	+ hv_vm_packet_descriptor *pkt);

	/*
	*
	@@ -128,6 +127,35 @@
	return (net_dev);
	}

	+extern int
	+hv_nv_get_next_send_section(netvsc_dev *net_dev)
	+{
	+ unsigned long bitsmap_words = net_dev->bitsmap_words;
	+ unsigned long *bitsmap = net_dev->send_section_bitsmap;
	+ unsigned long idx;
	+ int ret = NVSP_1_CHIMNEY_SEND_INVALID_SECTION_INDEX;
	+ int i;
	+
	+ for (i = 0; i < bitsmap_words; i++) {
	+ idx = ffs(~bitsmap[i]);
	+ if (0 == idx) {
	+ continue;
	+ }
	+ idx--;
	+ if (i * BITS_PER_LONG + idx >= net_dev->send_section_count) {
	+ return ret;
	+ }
	+
	+ if (synch_test_and_set_bit(idx, &bitsmap[i])) {
	+ continue;
	+ }
	+ ret = i * BITS_PER_LONG + idx;
	+ break;
	+ }
	+
	+ return ret;
	+}
	+
	/*
	* Net VSC initialize receive buffer with net VSP
	*
	@@ -259,7 +287,7 @@
	* channel to establish the gpadl handle.
	*/
	ret = hv_vmbus_channel_establish_gpadl(device->channel,
	- net_dev->send_buf, net_dev->send_buf_size,
	+ net_dev->send_buf, net_dev->send_buf_size,
	&net_dev->send_buf_gpadl_handle);
	if (ret != 0) {
	goto cleanup;
	@@ -280,7 +308,7 @@
	/* Send the gpadl notification request */

	ret = hv_vmbus_channel_send_packet(device->channel, init_pkt,
	- sizeof(nvsp_msg), (uint64_t)(uintptr_t)init_pkt,
	+ sizeof(nvsp_msg), (uint64_t)(uintptr_t)init_pkt,
	HV_VMBUS_PACKET_TYPE_DATA_IN_BAND,
	HV_VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
	if (ret != 0) {
	@@ -298,6 +326,18 @@

	net_dev->send_section_size =
	init_pkt->msgs.vers_1_msgs.send_send_buf_complete.section_size;
	+ net_dev->send_section_count =
	+ net_dev->send_buf_size / net_dev->send_section_size;
	+ net_dev->bitsmap_words =
	+ (net_dev->send_section_count + BITS_PER_LONG - 1) /
	+ BITS_PER_LONG;
	+ net_dev->send_section_bitsmap =
	+ malloc(net_dev->bitsmap_words * sizeof(long), M_DEVBUF,
	+ M_NOWAIT \| M_ZERO);
	+ if (NULL == net_dev->send_section_bitsmap) {
	+ ret = ENOMEM;
	+ goto cleanup;
	+ }

	goto exit;

	@@ -434,6 +474,10 @@
	net_dev->send_buf = NULL;
	}

	+ if (net_dev->send_section_bitsmap) {
	+ free(net_dev->send_section_bitsmap, M_DEVBUF);
	+ }
	+
	return (ret);
	}

	@@ -447,7 +491,7 @@
	*/
	static int
	hv_nv_negotiate_nvsp_protocol(struct hv_device device, netvsc_dev net_dev,
	- uint32_t nvsp_ver)
	+ uint32_t nvsp_ver)
	{
	nvsp_msg *init_pkt;
	int ret;
	@@ -524,8 +568,13 @@
	{
	netvsc_dev *net_dev;
	nvsp_msg *init_pkt;
	- uint32_t nvsp_vers;
	uint32_t ndis_version;
	+ uint32_t protocol_list[] = { NVSP_PROTOCOL_VERSION_1,
	+ NVSP_PROTOCOL_VERSION_2,
	+ NVSP_PROTOCOL_VERSION_4,
	+ NVSP_PROTOCOL_VERSION_5 };
	+ int i;
	+ int protocol_number = 4;
	int ret = 0;
	device_t dev = device->device;
	hn_softc_t *sc = device_get_softc(dev);
	@@ -537,26 +586,30 @@
	}

	/*
	- * Negotiate the NVSP version. Try NVSP v2 first.
	+ * Negotiate the NVSP version. Try the latest NVSP first.
	*/
	- nvsp_vers = NVSP_PROTOCOL_VERSION_2;
	- ret = hv_nv_negotiate_nvsp_protocol(device, net_dev, nvsp_vers);
	- if (ret != 0) {
	- /* NVSP v2 failed, try NVSP v1 */
	- nvsp_vers = NVSP_PROTOCOL_VERSION_1;
	- ret = hv_nv_negotiate_nvsp_protocol(device, net_dev, nvsp_vers);
	- if (ret != 0) {
	- /* NVSP v1 failed, return bad status */
	- return (ret);
	+ for (i = protocol_number - 1; i >= 0; i--)
	+ if (hv_nv_negotiate_nvsp_protocol(device, net_dev,
	+ protocol_list[i]) == 0) {
	+ net_dev->nvsp_version = protocol_list[i];
	+ if (bootverbose)
	+ printf("Netvsc: got version 0x%x\n",
	+ net_dev->nvsp_version);
	+ break;
	}
	+
	+ if (i < 0) {
	+ if (bootverbose)
	+ printf("Netvsc: failed to negotiate a valid "
	+ "protocol.\n");
	+ return (EPROTO);
	}
	- net_dev->nvsp_version = nvsp_vers;

	/*
	* Set the MTU if supported by this NVSP protocol version
	* This needs to be right after the NVSP init message per Haiyang
	*/
	- if (nvsp_vers >= NVSP_PROTOCOL_VERSION_2)
	+ if (net_dev->nvsp_version >= NVSP_PROTOCOL_VERSION_2)
	ret = hv_nv_send_ndis_config(device, ifp->if_mtu);

	/*
	@@ -566,10 +619,11 @@

	memset(init_pkt, 0, sizeof(nvsp_msg));

	- /*
	- * Updated to version 5.1, minimum, for VLAN per Haiyang
	- */
	- ndis_version = NDIS_VERSION;
	+ if (net_dev->nvsp_version <= NVSP_PROTOCOL_VERSION_4) {
	+ ndis_version = NDIS_VERSION_6_1;
	+ } else {
	+ ndis_version = NDIS_VERSION_6_30;
	+ }

	init_pkt->hdr.msg_type = nvsp_msg_1_type_send_ndis_vers;
	init_pkt->msgs.vers_1_msgs.send_ndis_vers.ndis_major_vers =
	@@ -621,9 +675,7 @@
	hv_nv_on_device_add(struct hv_device device, void additional_info)
	{
	netvsc_dev *net_dev;
	- netvsc_packet *packet;
	- netvsc_packet *next_packet;
	- int i, ret = 0;
	+ int ret = 0;

	net_dev = hv_nv_alloc_net_device(device);
	if (!net_dev)
	@@ -631,29 +683,9 @@

	/* Initialize the NetVSC channel extension */
	net_dev->rx_buf_size = NETVSC_RECEIVE_BUFFER_SIZE;
	- mtx_init(&net_dev->rx_pkt_list_lock, "HV-RPL", NULL,
	- MTX_SPIN \| MTX_RECURSE);

	net_dev->send_buf_size = NETVSC_SEND_BUFFER_SIZE;

	- /* Same effect as STAILQ_HEAD_INITIALIZER() static initializer */
	- STAILQ_INIT(&net_dev->myrx_packet_list);
	-
	- /*
	- * malloc a sufficient number of netvsc_packet buffers to hold
	- * a packet list. Add them to the netvsc device packet queue.
	- */
	- for (i=0; i < NETVSC_RECEIVE_PACKETLIST_COUNT; i++) {
	- packet = malloc(sizeof(netvsc_packet) +
	- (NETVSC_RECEIVE_SG_COUNT * sizeof(hv_vmbus_page_buffer)),
	- M_DEVBUF, M_NOWAIT \| M_ZERO);
	- if (!packet) {
	- break;
	- }
	- STAILQ_INSERT_TAIL(&net_dev->myrx_packet_list, packet,
	- mylist_entry);
	- }
	-
	sema_init(&net_dev->channel_init_sema, 0, "netdev_sema");

	/*
	@@ -686,18 +718,6 @@
	*/
	if (net_dev) {
	sema_destroy(&net_dev->channel_init_sema);
	-
	- packet = STAILQ_FIRST(&net_dev->myrx_packet_list);
	- while (packet != NULL) {
	- next_packet = STAILQ_NEXT(packet, mylist_entry);
	- free(packet, M_DEVBUF);
	- packet = next_packet;
	- }
	- /* Reset the list to initial state */
	- STAILQ_INIT(&net_dev->myrx_packet_list);
	-
	- mtx_destroy(&net_dev->rx_pkt_list_lock);
	-
	free(net_dev, M_DEVBUF);
	}

	@@ -710,8 +730,6 @@
	int
	hv_nv_on_device_remove(struct hv_device *device, boolean_t destroy_channel)
	{
	- netvsc_packet *net_vsc_pkt;
	- netvsc_packet *next_net_vsc_pkt;
	hn_softc_t *sc = device_get_softc(device->device);
	netvsc_dev *net_dev = sc->net_dev;;

	@@ -738,18 +756,6 @@

	hv_vmbus_channel_close(device->channel);

	- /* Release all resources */
	- net_vsc_pkt = STAILQ_FIRST(&net_dev->myrx_packet_list);
	- while (net_vsc_pkt != NULL) {
	- next_net_vsc_pkt = STAILQ_NEXT(net_vsc_pkt, mylist_entry);
	- free(net_vsc_pkt, M_DEVBUF);
	- net_vsc_pkt = next_net_vsc_pkt;
	- }
	-
	- /* Reset the list to initial state */
	- STAILQ_INIT(&net_dev->myrx_packet_list);
	-
	- mtx_destroy(&net_dev->rx_pkt_list_lock);
	sema_destroy(&net_dev->channel_init_sema);
	free(net_dev, M_DEVBUF);

	@@ -759,18 +765,14 @@
	/*
	* Net VSC on send completion
	*/
	-static void
	-hv_nv_on_send_completion(struct hv_device device, hv_vm_packet_descriptor pkt)
	+static void
	+hv_nv_on_send_completion(netvsc_dev *net_dev,
	+ struct hv_device *device,
	+ hv_vm_packet_descriptor *pkt)
	{
	- netvsc_dev *net_dev;
	nvsp_msg *nvsp_msg_pkt;
	netvsc_packet *net_vsc_pkt;

	- net_dev = hv_nv_get_inbound_net_device(device);
	- if (!net_dev) {
	- return;
	- }
	-
	nvsp_msg_pkt =
	(nvsp_msg *)((unsigned long)pkt + (pkt->data_offset8 << 3));

	@@ -781,17 +783,25 @@
	== nvsp_msg_1_type_send_send_buf_complete) {
	/* Copy the response back */
	memcpy(&net_dev->channel_init_packet, nvsp_msg_pkt,
	- sizeof(nvsp_msg));
	+ sizeof(nvsp_msg));
	sema_post(&net_dev->channel_init_sema);
	} else if (nvsp_msg_pkt->hdr.msg_type ==
	- nvsp_msg_1_type_send_rndis_pkt_complete) {
	+ nvsp_msg_1_type_send_rndis_pkt_complete) {
	/* Get the send context */
	net_vsc_pkt =
	(netvsc_packet *)(unsigned long)pkt->transaction_id;
	+ if (NULL != net_vsc_pkt) {
	+ if (net_vsc_pkt->send_buf_section_idx !=
	+ NVSP_1_CHIMNEY_SEND_INVALID_SECTION_INDEX) {
	+ synch_change_bit(net_vsc_pkt->send_buf_section_idx,
	+ net_dev->send_section_bitsmap);
	+ }
	+
	+ /* Notify the layer above us */
	+ net_vsc_pkt->compl.send.on_send_completion(
	+ net_vsc_pkt->compl.send.send_completion_context);

	- /* Notify the layer above us */
	- net_vsc_pkt->compl.send.on_send_completion(
	- net_vsc_pkt->compl.send.send_completion_context);
	+ }

	atomic_subtract_int(&net_dev->num_outstanding_sends, 1);
	}
	@@ -822,10 +832,10 @@
	send_msg.msgs.vers_1_msgs.send_rndis_pkt.chan_type = 1;
	}

	- /* Not using send buffer section */
	send_msg.msgs.vers_1_msgs.send_rndis_pkt.send_buf_section_idx =
	- 0xFFFFFFFF;
	- send_msg.msgs.vers_1_msgs.send_rndis_pkt.send_buf_section_size = 0;
	+ pkt->send_buf_section_idx;
	+ send_msg.msgs.vers_1_msgs.send_rndis_pkt.send_buf_section_size =
	+ pkt->send_buf_section_size;

	if (pkt->page_buf_count) {
	ret = hv_vmbus_channel_send_packet_pagebuffer(device->channel,
	@@ -851,157 +861,84 @@
	* In the FreeBSD Hyper-V virtual world, this function deals exclusively
	* with virtual addresses.
	*/
	-static void
	-hv_nv_on_receive(struct hv_device device, hv_vm_packet_descriptor pkt)
	+static void
	+hv_nv_on_receive
	+(
	+ netvsc_dev *net_dev,
	+ struct hv_device *device,
	+ hv_vm_packet_descriptor *pkt
	+)
	{
	- netvsc_dev *net_dev;
	hv_vm_transfer_page_packet_header *vm_xfer_page_pkt;
	nvsp_msg *nvsp_msg_pkt;
	- netvsc_packet *net_vsc_pkt = NULL;
	- unsigned long start;
	- xfer_page_packet *xfer_page_pkt = NULL;
	- STAILQ_HEAD(PKT_LIST, netvsc_packet_) mylist_head =
	- STAILQ_HEAD_INITIALIZER(mylist_head);
	+ netvsc_packet vsc_pkt;
	+ netvsc_packet *net_vsc_pkt = &vsc_pkt;
	int count = 0;
	int i = 0;
	-
	- net_dev = hv_nv_get_inbound_net_device(device);
	- if (!net_dev)
	- return;
	+ int status = nvsp_status_success;

	/*
	* All inbound packets other than send completion should be
	* xfer page packet.
	*/
	- if (pkt->type != HV_VMBUS_PACKET_TYPE_DATA_USING_TRANSFER_PAGES)
	+ if (pkt->type != HV_VMBUS_PACKET_TYPE_DATA_USING_TRANSFER_PAGES) {
	+ printf("packet type %d is invalid!\n", pkt->type);
	return;
	+ }

	nvsp_msg_pkt = (nvsp_msg *)((unsigned long)pkt
	+ (pkt->data_offset8 << 3));

	/* Make sure this is a valid nvsp packet */
	- if (nvsp_msg_pkt->hdr.msg_type != nvsp_msg_1_type_send_rndis_pkt)
	+ if (nvsp_msg_pkt->hdr.msg_type != nvsp_msg_1_type_send_rndis_pkt) {
	+ printf("packet hdr type %d is invalid!\n", pkt->type);
	return;
	+ }

	vm_xfer_page_pkt = (hv_vm_transfer_page_packet_header *)pkt;

	- if (vm_xfer_page_pkt->transfer_page_set_id
	- != NETVSC_RECEIVE_BUFFER_ID) {
	- return;
	- }
	-
	- STAILQ_INIT(&mylist_head);
	-
	- /*
	- * Grab free packets (range count + 1) to represent this xfer page
	- * packet. +1 to represent the xfer page packet itself. We grab it
	- * here so that we know exactly how many we can fulfill.
	- */
	- mtx_lock_spin(&net_dev->rx_pkt_list_lock);
	- while (!STAILQ_EMPTY(&net_dev->myrx_packet_list)) {
	- net_vsc_pkt = STAILQ_FIRST(&net_dev->myrx_packet_list);
	- STAILQ_REMOVE_HEAD(&net_dev->myrx_packet_list, mylist_entry);
	-
	- STAILQ_INSERT_TAIL(&mylist_head, net_vsc_pkt, mylist_entry);
	-
	- if (++count == vm_xfer_page_pkt->range_count + 1)
	- break;
	- }
	-
	- mtx_unlock_spin(&net_dev->rx_pkt_list_lock);
	-
	- /*
	- * We need at least 2 netvsc pkts (1 to represent the xfer page
	- * and at least 1 for the range) i.e. we can handle some of the
	- * xfer page packet ranges...
	- */
	- if (count < 2) {
	- /* Return netvsc packet to the freelist */
	- mtx_lock_spin(&net_dev->rx_pkt_list_lock);
	- for (i=count; i != 0; i--) {
	- net_vsc_pkt = STAILQ_FIRST(&mylist_head);
	- STAILQ_REMOVE_HEAD(&mylist_head, mylist_entry);
	-
	- STAILQ_INSERT_TAIL(&net_dev->myrx_packet_list,
	- net_vsc_pkt, mylist_entry);
	- }
	- mtx_unlock_spin(&net_dev->rx_pkt_list_lock);
	-
	- hv_nv_send_receive_completion(device,
	- vm_xfer_page_pkt->d.transaction_id);
	-
	+ if (vm_xfer_page_pkt->transfer_page_set_id !=
	+ NETVSC_RECEIVE_BUFFER_ID) {
	+ printf("transfer_page_set_id %d is invalid!\n",
	+ vm_xfer_page_pkt->transfer_page_set_id);
	return;
	}

	- /* Take the first packet in the list */
	- xfer_page_pkt = (xfer_page_packet *)STAILQ_FIRST(&mylist_head);
	- STAILQ_REMOVE_HEAD(&mylist_head, mylist_entry);
	-
	- /* This is how many data packets we can supply */
	- xfer_page_pkt->count = count - 1;
	+ count = vm_xfer_page_pkt->range_count;
	+ net_vsc_pkt->device = device;

	/* Each range represents 1 RNDIS pkt that contains 1 Ethernet frame */
	- for (i=0; i < (count - 1); i++) {
	- net_vsc_pkt = STAILQ_FIRST(&mylist_head);
	- STAILQ_REMOVE_HEAD(&mylist_head, mylist_entry);
	-
	- /*
	- * Initialize the netvsc packet
	- */
	- net_vsc_pkt->xfer_page_pkt = xfer_page_pkt;
	- net_vsc_pkt->compl.rx.rx_completion_context = net_vsc_pkt;
	- net_vsc_pkt->device = device;
	- /* Save this so that we can send it back */
	- net_vsc_pkt->compl.rx.rx_completion_tid =
	- vm_xfer_page_pkt->d.transaction_id;
	-
	- net_vsc_pkt->tot_data_buf_len =
	- vm_xfer_page_pkt->ranges[i].byte_count;
	- net_vsc_pkt->page_buf_count = 1;
	-
	- net_vsc_pkt->page_buffers[0].length =
	- vm_xfer_page_pkt->ranges[i].byte_count;
	-
	- /* The virtual address of the packet in the receive buffer */
	- start = ((unsigned long)net_dev->rx_buf +
	+ for (i = 0; i < count; i++) {
	+ net_vsc_pkt->status = nvsp_status_success;
	+ net_vsc_pkt->data = (void *)((unsigned long)net_dev->rx_buf +
	vm_xfer_page_pkt->ranges[i].byte_offset);
	- start = ((unsigned long)start) & ~(PAGE_SIZE - 1);
	-
	- /* Page number of the virtual page containing packet start */
	- net_vsc_pkt->page_buffers[0].pfn = start >> PAGE_SHIFT;
	-
	- /* Calculate the page relative offset */
	- net_vsc_pkt->page_buffers[0].offset =
	- vm_xfer_page_pkt->ranges[i].byte_offset & (PAGE_SIZE - 1);
	-
	- /*
	- * In this implementation, we are dealing with virtual
	- * addresses exclusively. Since we aren't using physical
	- * addresses at all, we don't care if a packet crosses a
	- * page boundary. For this reason, the original code to
	- * check for and handle page crossings has been removed.
	- */
	-
	- /*
	- * Pass it to the upper layer. The receive completion call
	- * has been moved into this function.
	- */
	- hv_rf_on_receive(device, net_vsc_pkt);
	+ net_vsc_pkt->tot_data_buf_len =
	+ vm_xfer_page_pkt->ranges[i].byte_count;

	- /*
	- * Moved completion call back here so that all received
	- * messages (not just data messages) will trigger a response
	- * message back to the host.
	- */
	- hv_nv_on_receive_completion(net_vsc_pkt);
	+ hv_rf_on_receive(net_dev, device, net_vsc_pkt);
	+ if (net_vsc_pkt->status != nvsp_status_success) {
	+ status = nvsp_status_failure;
	+ }
	}
	+
	+ /*
	+ * Moved completion call back here so that all received
	+ * messages (not just data messages) will trigger a response
	+ * message back to the host.
	+ */
	+ hv_nv_on_receive_completion(device, vm_xfer_page_pkt->d.transaction_id,
	+ status);
	}

	/*
	- * Net VSC send receive completion
	+ * Net VSC on receive completion
	+ *
	+ * Send a receive completion packet to RNDIS device (ie NetVsp)
	*/
	-static void
	-hv_nv_send_receive_completion(struct hv_device *device, uint64_t tid)
	+void
	+hv_nv_on_receive_completion(struct hv_device *device,
	+ uint64_t tid,
	+ uint32_t status)
	{
	nvsp_msg rx_comp_msg;
	int retries = 0;
	@@ -1011,7 +948,7 @@

	/* Pass in the status */
	rx_comp_msg.msgs.vers_1_msgs.send_rndis_pkt_complete.status =
	- nvsp_status_success;
	+ status;

	retry_send_cmplt:
	/* Send the completion */
	@@ -1032,81 +969,25 @@
	}

	/*
	- * Net VSC on receive completion
	- *
	- * Send a receive completion packet to RNDIS device (ie NetVsp)
	- */
	-void
	-hv_nv_on_receive_completion(void *context)
	-{
	- netvsc_packet packet = (netvsc_packet )context;
	- struct hv_device device = (struct hv_device )packet->device;
	- netvsc_dev *net_dev;
	- uint64_t tid = 0;
	- boolean_t send_rx_completion = FALSE;
	-
	- /*
	- * Even though it seems logical to do a hv_nv_get_outbound_net_device()
	- * here to send out receive completion, we are using
	- * hv_nv_get_inbound_net_device() since we may have disabled
	- * outbound traffic already.
	- */
	- net_dev = hv_nv_get_inbound_net_device(device);
	- if (net_dev == NULL)
	- return;
	-
	- /* Overloading use of the lock. */
	- mtx_lock_spin(&net_dev->rx_pkt_list_lock);
	-
	- packet->xfer_page_pkt->count--;
	-
	- /*
	- * Last one in the line that represent 1 xfer page packet.
	- * Return the xfer page packet itself to the free list.
	- */
	- if (packet->xfer_page_pkt->count == 0) {
	- send_rx_completion = TRUE;
	- tid = packet->compl.rx.rx_completion_tid;
	- STAILQ_INSERT_TAIL(&net_dev->myrx_packet_list,
	- (netvsc_packet *)(packet->xfer_page_pkt), mylist_entry);
	- }
	-
	- /* Put the packet back on the free list */
	- STAILQ_INSERT_TAIL(&net_dev->myrx_packet_list, packet, mylist_entry);
	- mtx_unlock_spin(&net_dev->rx_pkt_list_lock);
	-
	- /* Send a receive completion for the xfer page packet */
	- if (send_rx_completion)
	- hv_nv_send_receive_completion(device, tid);
	-}
	-
	-/*
	* Net VSC on channel callback
	*/
	static void
	hv_nv_on_channel_callback(void *context)
	{
	- /* Fixme: Magic number */
	- const int net_pkt_size = 2048;
	struct hv_device device = (struct hv_device )context;
	netvsc_dev *net_dev;
	uint32_t bytes_rxed;
	uint64_t request_id;
	- uint8_t *packet;
	- hv_vm_packet_descriptor *desc;
	+ hv_vm_packet_descriptor *desc;
	uint8_t *buffer;
	- int bufferlen = net_pkt_size;
	- int ret = 0;
	-
	- packet = malloc(net_pkt_size * sizeof(uint8_t), M_DEVBUF, M_NOWAIT);
	- if (!packet)
	- return;
	-
	- buffer = packet;
	+ int bufferlen = NETVSC_PACKET_SIZE;
	+ int ret = 0;

	net_dev = hv_nv_get_inbound_net_device(device);
	if (net_dev == NULL)
	- goto out;
	+ return;
	+
	+ buffer = net_dev->callback_buf;

	do {
	ret = hv_vmbus_channel_recv_packet_raw(device->channel,
	@@ -1116,12 +997,13 @@
	desc = (hv_vm_packet_descriptor *)buffer;
	switch (desc->type) {
	case HV_VMBUS_PACKET_TYPE_COMPLETION:
	- hv_nv_on_send_completion(device, desc);
	+ hv_nv_on_send_completion(net_dev, device, desc);
	break;
	case HV_VMBUS_PACKET_TYPE_DATA_USING_TRANSFER_PAGES:
	- hv_nv_on_receive(device, desc);
	+ hv_nv_on_receive(net_dev, device, desc);
	break;
	default:
	+ printf("hv_cb recv unknow packet\n");
	break;
	}
	} else {
	@@ -1129,16 +1011,22 @@
	}
	} else if (ret == ENOBUFS) {
	/* Handle large packet */
	- free(buffer, M_DEVBUF);
	+ if (bufferlen > NETVSC_PACKET_SIZE){
	+ free(buffer, M_DEVBUF);
	+ buffer = NULL;
	+ }
	+
	+ /* alloc new buffer */
	buffer = malloc(bytes_rxed, M_DEVBUF, M_NOWAIT);
	if (buffer == NULL) {
	+ printf("hv_cb malloc buffer failed, len=%u\n", bytes_rxed);
	+ bufferlen = 0;
	break;
	}
	bufferlen = bytes_rxed;
	}
	} while (1);

	-out:
	- free(buffer, M_DEVBUF);
	+ if (bufferlen > NETVSC_PACKET_SIZE)
	+ free(buffer, M_DEVBUF);
	}
	-
	Index: sys/dev/hyperv/netvsc/hv_netvsc_drv_freebsd.c
	===================================================================
	--- sys/dev/hyperv/netvsc/hv_netvsc_drv_freebsd.c
	+++ sys/dev/hyperv/netvsc/hv_netvsc_drv_freebsd.c
	@@ -83,6 +83,9 @@
	#include <netinet/in.h>
	#include <netinet/ip.h>
	#include <netinet/if_ether.h>
	+#include <netinet/tcp.h>
	+#include <netinet/udp.h>
	+#include <netinet/ip6.h>

	#include <vm/vm.h>
	#include <vm/vm_param.h>
	@@ -103,6 +106,8 @@

	#include <machine/intr_machdep.h>

	+#include <machine/in_cksum.h>
	+
	#include <dev/hyperv/include/hyperv.h>
	#include "hv_net_vsc.h"
	#include "hv_rndis.h"
	@@ -165,6 +170,52 @@
	static int hn_start_locked(struct ifnet *ifp);
	static void hn_start(struct ifnet *ifp);

	+/*
	+ * NetVsc get message transport protocol type
	+ */
	+static uint32_t get_transport_proto_type(struct mbuf *m_head)
	+{
	+ uint32_t ret_val = TRANSPORT_TYPE_NOT_IP;
	+ uint16_t ether_type = 0;
	+ int ether_len = 0;
	+ struct ether_vlan_header *eh;
	+ struct ip *iph;
	+ struct ip6_hdr *ip6;
	+
	+ eh = mtod(m_head, struct ether_vlan_header*);
	+ if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
	+ ether_len = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
	+ ether_type = eh->evl_proto;
	+ } else {
	+ ether_len = ETHER_HDR_LEN;
	+ ether_type = eh->evl_encap_proto;
	+ }
	+
	+ switch (ntohs(ether_type)) {
	+ case ETHERTYPE_IPV6:
	+ ip6 = (struct ip6_hdr *)(m_head->m_data + ether_len);
	+
	+ if (IPPROTO_TCP == ip6->ip6_nxt) {
	+ ret_val = TRANSPORT_TYPE_IPV6_TCP;
	+ } else if (IPPROTO_UDP == ip6->ip6_nxt) {
	+ ret_val = TRANSPORT_TYPE_IPV6_UDP;
	+ }
	+ break;
	+ case ETHERTYPE_IP:
	+ iph = (struct ip *)(m_head->m_data + ether_len);
	+
	+ if (IPPROTO_TCP == iph->ip_p) {
	+ ret_val = TRANSPORT_TYPE_IPV4_TCP;
	+ } else if (IPPROTO_UDP == iph->ip_p) {
	+ ret_val = TRANSPORT_TYPE_IPV4_UDP;
	+ }
	+ break;
	+ default:
	+ break;
	+ }
	+
	+ return ret_val;
	+}

	/*
	* NetVsc driver initialization
	@@ -276,8 +327,11 @@
	* Tell upper layers that we support full VLAN capability.
	*/
	ifp->if_hdrlen = sizeof(struct ether_vlan_header);
	- ifp->if_capabilities \|= IFCAP_VLAN_HWTAGGING \| IFCAP_VLAN_MTU;
	- ifp->if_capenable \|= IFCAP_VLAN_HWTAGGING \| IFCAP_VLAN_MTU;
	+ ifp->if_capabilities \|=
	+ IFCAP_VLAN_HWTAGGING \| IFCAP_VLAN_MTU \| IFCAP_HWCSUM \| IFCAP_TSO;
	+ ifp->if_capenable \|=
	+ IFCAP_VLAN_HWTAGGING \| IFCAP_VLAN_MTU \| IFCAP_HWCSUM \| IFCAP_TSO;
	+ ifp->if_hwassist = CSUM_TCP \| CSUM_UDP \| CSUM_TSO;

	ret = hv_rf_on_device_add(device_ctx, &device_info);
	if (ret != 0) {
	@@ -362,17 +416,28 @@
	{
	hn_softc_t *sc = ifp->if_softc;
	struct hv_device *device_ctx = vmbus_get_devctx(sc->hn_dev);
	+ netvsc_dev *net_dev = sc->net_dev;
	uint8_t *buf;
	netvsc_packet *packet;
	struct mbuf m_head, m;
	struct mbuf *mc_head = NULL;
	+ struct ether_vlan_header *eh;
	+ rndis_msg *rndis_mesg;
	+ rndis_packet *rndis_pkt;
	+ rndis_per_packet_info *rppi;
	+ ndis_8021q_info *rppi_vlan_info;
	+ rndis_tcp_ip_csum_info *csum_info;
	+ rndis_tcp_tso_info *tso_info;
	+ int ether_len;
	int i;
	int num_frags;
	int len;
	- int xlen;
	- int rppi_size;
	int retries = 0;
	- int ret = 0;
	+ int ret = 0;
	+ uint32_t rndis_msg_size = 0;
	+ uint32_t trans_proto_type;
	+ uint32_t send_buf_section_idx =
	+ NVSP_1_CHIMNEY_SEND_INVALID_SECTION_INDEX;

	while (!IFQ_DRV_IS_EMPTY(&sc->hn_ifp->if_snd)) {
	IFQ_DRV_DEQUEUE(&sc->hn_ifp->if_snd, m_head);
	@@ -382,7 +447,6 @@

	len = 0;
	num_frags = 0;
	- xlen = 0;

	/* Walk the mbuf list computing total length and num frags */
	for (m = m_head; m != NULL; m = m->m_next) {
	@@ -401,66 +465,216 @@

	/* If exceeds # page_buffers in netvsc_packet */
	if (num_frags > NETVSC_PACKET_MAXPAGE) {
	- m_freem(m);
	-
	+ printf("exceed max page buffers,%d,%d\n",
	+ num_frags, NETVSC_PACKET_MAXPAGE);
	+ m_freem(m_head);
	+ if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
	return (EINVAL);
	}

	- rppi_size = 0;
	- if (m_head->m_flags & M_VLANTAG) {
	- rppi_size = sizeof(rndis_per_packet_info) +
	- sizeof(ndis_8021q_info);
	- }
	-
	/*
	* Allocate a buffer with space for a netvsc packet plus a
	* number of reserved areas. First comes a (currently 16
	* bytes, currently unused) reserved data area. Second is
	- * the netvsc_packet, which includes (currently 4) page
	- * buffers. Third (optional) is a rndis_per_packet_info
	- * struct, but only if a VLAN tag should be inserted into the
	- * Ethernet frame by the Hyper-V infrastructure. Fourth is
	- * an area reserved for an rndis_filter_packet struct.
	+ * the netvsc_packet. Third is an area reserved for an
	+ * rndis_filter_packet struct. Fourth (optional) is a
	+ * rndis_per_packet_info struct.
	* Changed malloc to M_NOWAIT to avoid sleep under spin lock.
	* No longer reserving extra space for page buffers, as they
	* are already part of the netvsc_packet.
	*/
	buf = malloc(HV_NV_PACKET_OFFSET_IN_BUF +
	- sizeof(netvsc_packet) + rppi_size +
	- sizeof(rndis_filter_packet),
	- M_DEVBUF, M_ZERO \| M_NOWAIT);
	+ sizeof(netvsc_packet) +
	+ sizeof(rndis_msg) +
	+ RNDIS_VLAN_PPI_SIZE +
	+ RNDIS_TSO_PPI_SIZE +
	+ RNDIS_CSUM_PPI_SIZE,
	+ M_DEVBUF, M_ZERO \| M_NOWAIT);
	if (buf == NULL) {
	- m_freem(m);
	-
	+ printf("hn:malloc packet failed\n");
	+ m_freem(m_head);
	+ if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
	return (ENOMEM);
	}

	packet = (netvsc_packet *)(buf + HV_NV_PACKET_OFFSET_IN_BUF);
	(vm_offset_t )buf = HV_NV_SC_PTR_OFFSET_IN_BUF;

	+ packet->is_data_pkt = TRUE;
	+
	+ /* Set up the rndis header */
	+ packet->page_buf_count = num_frags;
	+
	+ /* Initialize it from the mbuf */
	+ packet->tot_data_buf_len = len;
	+
	/*
	* extension points to the area reserved for the
	* rndis_filter_packet, which is placed just after
	* the netvsc_packet (and rppi struct, if present;
	* length is updated later).
	*/
	- packet->extension = packet + 1;
	+ packet->rndis_mesg = packet + 1;
	+ rndis_mesg = (rndis_msg *)packet->rndis_mesg;
	+ rndis_mesg->ndis_msg_type = REMOTE_NDIS_PACKET_MSG;

	- /* Set up the rndis header */
	- packet->page_buf_count = num_frags;
	+ rndis_pkt = &rndis_mesg->msg.packet;
	+ rndis_pkt->data_offset = sizeof(rndis_packet);
	+ rndis_pkt->data_length = packet->tot_data_buf_len;
	+ rndis_pkt->per_pkt_info_offset = sizeof(rndis_packet);

	- /* Initialize it from the mbuf */
	- packet->tot_data_buf_len = len;
	+ rndis_msg_size = RNDIS_MESSAGE_SIZE(rndis_packet);

	/*
	* If the Hyper-V infrastructure needs to embed a VLAN tag,
	* initialize netvsc_packet and rppi struct values as needed.
	*/
	- if (rppi_size) {
	- /* Lower layers need the VLAN TCI */
	+ if (m_head->m_flags & M_VLANTAG) {
	+ /*
	+ * set up some additional fields so the Hyper-V infrastructure will stuff the VLAN tag
	+ * into the frame.
	+ */
	packet->vlan_tci = m_head->m_pkthdr.ether_vtag;
	+
	+ rndis_msg_size += RNDIS_VLAN_PPI_SIZE;
	+
	+ rppi = hv_set_rppi_data(rndis_mesg, RNDIS_VLAN_PPI_SIZE,
	+ ieee_8021q_info);
	+
	+ /* VLAN info immediately follows rppi struct */
	+ rppi_vlan_info = (ndis_8021q_info )((char)rppi +
	+ rppi->per_packet_info_offset);
	+ /* FreeBSD does not support CFI or priority */
	+ rppi_vlan_info->u1.s1.vlan_id =
	+ packet->vlan_tci & 0xfff;
	+ }
	+
	+ if (0 == m_head->m_pkthdr.csum_flags) {
	+ goto pre_send;
	+ }
	+
	+ eh = mtod(m_head, struct ether_vlan_header*);
	+ if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
	+ ether_len = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
	+ } else {
	+ ether_len = ETHER_HDR_LEN;
	+ }
	+
	+ trans_proto_type = get_transport_proto_type(m_head);
	+ if (TRANSPORT_TYPE_NOT_IP == trans_proto_type) {
	+ goto pre_send;
	+ }
	+
	+ /*
	+ * TSO packet needless to setup the send side checksum
	+ * offload.
	+ */
	+ if (m_head->m_pkthdr.csum_flags & CSUM_TSO) {
	+ goto do_tso;
	+ }
	+
	+ /* setup checksum offload */
	+ rndis_msg_size += RNDIS_CSUM_PPI_SIZE;
	+ rppi = hv_set_rppi_data(rndis_mesg, RNDIS_CSUM_PPI_SIZE,
	+ tcpip_chksum_info);
	+ csum_info = (rndis_tcp_ip_csum_info )((char)rppi +
	+ rppi->per_packet_info_offset);
	+
	+ if (trans_proto_type & (TYPE_IPV4 << 16)) {
	+ csum_info->xmit.is_ipv4 = 1;
	+ } else {
	+ csum_info->xmit.is_ipv6 = 1;
	+ }
	+
	+ if (trans_proto_type & TYPE_TCP) {
	+ csum_info->xmit.tcp_csum = 1;
	+ csum_info->xmit.tcp_header_offset = 0;
	+ } else if (trans_proto_type & TYPE_UDP) {
	+ csum_info->xmit.udp_csum = 1;
	+ }
	+
	+ goto pre_send;
	+
	+do_tso:
	+ /* setup TCP segmentation offload */
	+ rndis_msg_size += RNDIS_TSO_PPI_SIZE;
	+ rppi = hv_set_rppi_data(rndis_mesg, RNDIS_TSO_PPI_SIZE,
	+ tcp_large_send_info);
	+
	+ tso_info = (rndis_tcp_tso_info )((char )rppi +
	+ rppi->per_packet_info_offset);
	+ tso_info->lso_v2_xmit.type =
	+ RNDIS_TCP_LARGE_SEND_OFFLOAD_V2_TYPE;
	+
	+ if (trans_proto_type & (TYPE_IPV4 << 16)) {
	+ struct ip *ip =
	+ (struct ip *)(m_head->m_data + ether_len);
	+ unsigned long iph_len = ip->ip_hl << 2;
	+ struct tcphdr *th =
	+ (struct tcphdr *)((caddr_t)ip + iph_len);
	+
	+ tso_info->lso_v2_xmit.ip_version =
	+ RNDIS_TCP_LARGE_SEND_OFFLOAD_IPV4;
	+ ip->ip_len = 0;
	+ ip->ip_sum = 0;
	+
	+ th->th_sum = in_pseudo(ip->ip_src.s_addr,
	+ ip->ip_dst.s_addr,
	+ htons(IPPROTO_TCP));
	+ } else {
	+ struct ip6_hdr *ip6 =
	+ (struct ip6_hdr *)(m_head->m_data + ether_len);
	+ struct tcphdr th = (struct tcphdr )(ip6 + 1);
	+
	+ tso_info->lso_v2_xmit.ip_version =
	+ RNDIS_TCP_LARGE_SEND_OFFLOAD_IPV6;
	+ ip6->ip6_plen = 0;
	+ th->th_sum = in6_cksum_pseudo(ip6, 0, IPPROTO_TCP, 0);
	+ }
	+ tso_info->lso_v2_xmit.tcp_header_offset = 0;
	+ tso_info->lso_v2_xmit.mss = m_head->m_pkthdr.tso_segsz;
	+
	+pre_send:
	+ rndis_mesg->msg_len = packet->tot_data_buf_len + rndis_msg_size;
	+ packet->tot_data_buf_len = rndis_mesg->msg_len;
	+
	+ /* send packet with send buffer */
	+ if (packet->tot_data_buf_len < net_dev->send_section_size) {
	+ send_buf_section_idx =
	+ hv_nv_get_next_send_section(net_dev);
	+ if (send_buf_section_idx !=
	+ NVSP_1_CHIMNEY_SEND_INVALID_SECTION_INDEX) {
	+ char dest = ((char )net_dev->send_buf +
	+ send_buf_section_idx *
	+ net_dev->send_section_size);
	+
	+ memcpy(dest, rndis_mesg, rndis_msg_size);
	+ dest += rndis_msg_size;
	+ for (m = m_head; m != NULL; m = m->m_next) {
	+ if (m->m_len) {
	+ memcpy(dest,
	+ (void *)mtod(m, vm_offset_t),
	+ m->m_len);
	+ dest += m->m_len;
	+ }
	+ }
	+
	+ packet->send_buf_section_idx =
	+ send_buf_section_idx;
	+ packet->send_buf_section_size =
	+ packet->tot_data_buf_len;
	+ packet->page_buf_count = 0;
	+ goto do_send;
	+ }
	}

	+ /* send packet with page buffer */
	+ packet->page_buffers[0].pfn =
	+ hv_get_phys_addr(rndis_mesg) >> PAGE_SHIFT;
	+ packet->page_buffers[0].offset =
	+ (unsigned long)rndis_mesg & (PAGE_SIZE - 1);
	+ packet->page_buffers[0].length = rndis_msg_size;
	+
	/*
	* Fill the page buffers with mbuf info starting at index
	* HV_RF_NUM_TX_RESERVED_PAGE_BUFS.
	@@ -479,6 +693,12 @@
	}
	}

	+ packet->send_buf_section_idx =
	+ NVSP_1_CHIMNEY_SEND_INVALID_SECTION_INDEX;
	+ packet->send_buf_section_size = 0;
	+
	+do_send:
	+
	/*
	* If bpf, copy the mbuf chain. This is less expensive than
	* it appears; the mbuf clusters are not copied, only their
	@@ -497,8 +717,7 @@
	packet->compl.send.send_completion_tid = (uint64_t)(uintptr_t)m_head;

	/* Removed critical_enter(), does not appear necessary */
	- ret = hv_rf_on_send(device_ctx, packet);
	-
	+ ret = hv_nv_on_send(device_ctx, packet);
	if (ret == 0) {
	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
	/* if bpf && mc_head, call bpf_mtap code */
	@@ -526,6 +745,7 @@
	* send completion
	*/
	netvsc_xmit_completion(packet);
	+ if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
	}

	/* if bpf && mc_head, free the mbuf chain copy */
	@@ -621,13 +841,17 @@
	* Note: This is no longer used as a callback
	*/
	int
	-netvsc_recv(struct hv_device device_ctx, netvsc_packet packet)
	+netvsc_recv
	+(
	+ struct hv_device *device_ctx,
	+ netvsc_packet *packet,
	+ rndis_tcp_ip_csum_info *csum_info
	+)
	{
	hn_softc_t sc = (hn_softc_t )device_get_softc(device_ctx->device);
	struct mbuf *m_new;
	struct ifnet *ifp;
	int size;
	- int i;

	if (sc == NULL) {
	return (0); /* TODO: KYS how can this be! */
	@@ -661,36 +885,35 @@

	m_new = m_getjcl(M_NOWAIT, MT_DATA, M_PKTHDR, size);

	- if (m_new == NULL)
	+ if (m_new == NULL) {
	+ printf("alloc mbuf failed!!!\n");
	return (0);
	+ }

	- /*
	- * Remove trailing junk from RX data buffer.
	- * Fixme: This will not work for multiple Hyper-V RX buffers.
	- * Fortunately, the channel gathers all RX data into one buffer.
	- *
	- * L2 frame length, with L2 header, not including CRC
	- */
	- packet->page_buffers[0].length = packet->tot_data_buf_len;
	+ hv_m_append(m_new, packet->tot_data_buf_len,
	+ packet->data);

	- /*
	- * Copy the received packet to one or more mbufs.
	- * The copy is required since the memory pointed to by netvsc_packet
	- * cannot be deallocated
	- */
	- for (i=0; i < packet->page_buf_count; i++) {
	- /* Shift virtual page number to form virtual page address */
	- uint8_t vaddr = (uint8_t )(uintptr_t)
	- (packet->page_buffers[i].pfn << PAGE_SHIFT);
	+ m_new->m_pkthdr.rcvif = ifp;

	- hv_m_append(m_new, packet->page_buffers[i].length,
	- vaddr + packet->page_buffers[i].offset);
	- }
	+ /* receive side checksum offload */
	+ m_new->m_pkthdr.csum_flags = 0;
	+ if (NULL != csum_info) {
	+ /* IP csum offload */
	+ if (csum_info->receive.ip_csum_succeeded) {
	+ m_new->m_pkthdr.csum_flags \|=
	+ (CSUM_IP_CHECKED \| CSUM_IP_VALID);
	+ }

	- m_new->m_pkthdr.rcvif = ifp;
	+ /* TCP csum offload */
	+ if (csum_info->receive.tcp_csum_succeeded) {
	+ m_new->m_pkthdr.csum_flags \|=
	+ (CSUM_DATA_VALID \| CSUM_PSEUDO_HDR);
	+ m_new->m_pkthdr.csum_data = 0xffff;
	+ }
	+ }

	if ((packet->vlan_tci != 0) &&
	- (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0) {
	+ (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0) {
	m_new->m_pkthdr.ether_vtag = packet->vlan_tci;
	m_new->m_flags \|= M_VLANTAG;
	}
	@@ -736,15 +959,26 @@
	switch(cmd) {

	case SIOCSIFADDR:
	+ case SIOCGIFADDR:
	#ifdef INET
	+ NV_LOCK(sc);
	if (ifa->ifa_addr->sa_family == AF_INET) {
	ifp->if_flags \|= IFF_UP;
	- if (!(ifp->if_drv_flags & IFF_DRV_RUNNING))
	- hn_ifinit(sc);
	- arp_ifinit(ifp, ifa);
	- } else
	-#endif
	+ if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
	+ ifp->if_drv_flags &= ~(IFF_DRV_RUNNING \| IFF_DRV_OACTIVE);
	+ if_link_state_change(ifp, LINK_STATE_DOWN);
	+ ifp->if_drv_flags \|= IFF_DRV_RUNNING;
	+ ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
	+ if_link_state_change(ifp, LINK_STATE_UP);
	+ }
	+ NV_UNLOCK(sc);
	+ } else {
	+ NV_UNLOCK(sc);
	+#endif
	error = ether_ioctl(ifp, cmd, data);
	+#ifdef INET
	+ }
	+#endif
	break;
	case SIOCSIFMTU:
	hn_dev = vmbus_get_devctx(sc->hn_dev);
	@@ -858,13 +1092,34 @@
	break;
	case SIOCSIFCAP:
	mask = ifr->ifr_reqcap ^ ifp->if_capenable;
	- if (mask & IFCAP_HWCSUM) {
	- if (IFCAP_HWCSUM & ifp->if_capenable) {
	- ifp->if_capenable &= ~IFCAP_HWCSUM;
	+ if (mask & IFCAP_TXCSUM) {
	+ if (IFCAP_TXCSUM & ifp->if_capenable) {
	+ ifp->if_capenable &= ~IFCAP_TXCSUM;
	+ ifp->if_hwassist &= ~(CSUM_TCP \| CSUM_UDP);
	+ } else {
	+ ifp->if_capenable \|= IFCAP_TXCSUM;
	+ ifp->if_hwassist \|= (CSUM_TCP \| CSUM_UDP);
	+ }
	+ }
	+
	+ if (mask & IFCAP_RXCSUM) {
	+ if (IFCAP_RXCSUM & ifp->if_capenable) {
	+ ifp->if_capenable &= ~IFCAP_RXCSUM;
	} else {
	- ifp->if_capenable \|= IFCAP_HWCSUM;
	+ ifp->if_capenable \|= IFCAP_RXCSUM;
	}
	}
	+
	+ if (mask & IFCAP_TSO4) {
	+ ifp->if_capenable ^= IFCAP_TSO4;
	+ ifp->if_hwassist ^= CSUM_IP_TSO;
	+ }
	+
	+ if (mask & IFCAP_TSO6) {
	+ ifp->if_capenable ^= IFCAP_TSO6;
	+ ifp->if_hwassist ^= CSUM_IP6_TSO;
	+ }
	+
	error = 0;
	break;
	case SIOCADDMULTI:
	Index: sys/dev/hyperv/netvsc/hv_rndis.h
	===================================================================
	--- sys/dev/hyperv/netvsc/hv_rndis.h
	+++ sys/dev/hyperv/netvsc/hv_rndis.h
	@@ -36,6 +36,9 @@
	#define NDIS_VERSION_5_0 0x00050000
	#define NDIS_VERSION_5_1 0x00050001
	#define NDIS_VERSION_6_0 0x00060000
	+#define NDIS_VERSION_6_1 0x00060001
	+#define NDIS_VERSION_6_30 0x0006001e
	+
	#define NDIS_VERSION (NDIS_VERSION_5_1)

	/*
	@@ -347,6 +350,36 @@
	#define RNDIS_MAJOR_VERSION 0x00000001
	#define RNDIS_MINOR_VERSION 0x00000000

	+
	+/*
	+ * Remote NDIS offload parameters
	+ */
	+#define RNDIS_OBJECT_TYPE_DEFAULT 0x80
	+
	+#define RNDIS_OFFLOAD_PARAMETERS_REVISION_3 3
	+#define RNDIS_OFFLOAD_PARAMETERS_NO_CHANGE 0
	+#define RNDIS_OFFLOAD_PARAMETERS_LSOV2_DISABLED 1
	+#define RNDIS_OFFLOAD_PARAMETERS_LSOV2_ENABLED 2
	+#define RNDIS_OFFLOAD_PARAMETERS_LSOV1_ENABLED 2
	+#define RNDIS_OFFLOAD_PARAMETERS_RSC_DISABLED 1
	+#define RNDIS_OFFLOAD_PARAMETERS_RSC_ENABLED 2
	+#define RNDIS_OFFLOAD_PARAMETERS_TX_RX_DISABLED 1
	+#define RNDIS_OFFLOAD_PARAMETERS_TX_ENABLED_RX_DISABLED 2
	+#define RNDIS_OFFLOAD_PARAMETERS_RX_ENABLED_TX_DISABLED 3
	+#define RNDIS_OFFLOAD_PARAMETERS_TX_RX_ENABLED 4
	+
	+#define RNDIS_TCP_LARGE_SEND_OFFLOAD_V2_TYPE 1
	+#define RNDIS_TCP_LARGE_SEND_OFFLOAD_IPV4 0
	+#define RNDIS_TCP_LARGE_SEND_OFFLOAD_IPV6 1
	+
	+
	+#define RNDIS_OID_TCP_OFFLOAD_CURRENT_CONFIG 0xFC01020B /* query only */
	+#define RNDIS_OID_TCP_OFFLOAD_PARAMETERS 0xFC01020C /* set only */
	+#define RNDIS_OID_TCP_OFFLOAD_HARDWARE_CAPABILITIES 0xFC01020D/* query only */
	+#define RNDIS_OID_TCP_CONNECTION_OFFLOAD_CURRENT_CONFIG 0xFC01020E /* query only */
	+#define RNDIS_OID_TCP_CONNECTION_OFFLOAD_HARDWARE_CAPABILITIES 0xFC01020F /* query */
	+#define RNDIS_OID_OFFLOAD_ENCAPSULATION 0x0101010A /* set/query */
	+
	/*
	* NdisInitialize message
	*/
	@@ -585,6 +618,108 @@
	} u1;
	} ndis_8021q_info;

	+struct rndis_object_header {
	+ uint8_t type;
	+ uint8_t revision;
	+ uint16_t size;
	+};
	+
	+typedef struct rndis_offload_params_ {
	+ struct rndis_object_header header;
	+ uint8_t ipv4_csum;
	+ uint8_t tcp_ipv4_csum;
	+ uint8_t udp_ipv4_csum;
	+ uint8_t tcp_ipv6_csum;
	+ uint8_t udp_ipv6_csum;
	+ uint8_t lso_v1;
	+ uint8_t ip_sec_v1;
	+ uint8_t lso_v2_ipv4;
	+ uint8_t lso_v2_ipv6;
	+ uint8_t tcp_connection_ipv4;
	+ uint8_t tcp_connection_ipv6;
	+ uint32_t flags;
	+ uint8_t ip_sec_v2;
	+ uint8_t ip_sec_v2_ipv4;
	+ struct {
	+ uint8_t rsc_ipv4;
	+ uint8_t rsc_ipv6;
	+ };
	+ struct {
	+ uint8_t encapsulated_packet_task_offload;
	+ uint8_t encapsulation_types;
	+ };
	+
	+}rndis_offload_params;
	+
	+
	+typedef struct rndis_tcp_ip_csum_info_ {
	+ union {
	+ struct {
	+ uint32_t is_ipv4:1;
	+ uint32_t is_ipv6:1;
	+ uint32_t tcp_csum:1;
	+ uint32_t udp_csum:1;
	+ uint32_t ip_header_csum:1;
	+ uint32_t reserved:11;
	+ uint32_t tcp_header_offset:10;
	+ } xmit;
	+ struct {
	+ uint32_t tcp_csum_failed:1;
	+ uint32_t udp_csum_failed:1;
	+ uint32_t ip_csum_failed:1;
	+ uint32_t tcp_csum_succeeded:1;
	+ uint32_t udp_csum_succeeded:1;
	+ uint32_t ip_csum_succeeded:1;
	+ uint32_t loopback:1;
	+ uint32_t tcp_csum_value_invalid:1;
	+ uint32_t ip_csum_value_invalid:1;
	+ } receive;
	+ uint32_t value;
	+ };
	+} rndis_tcp_ip_csum_info;
	+
	+typedef struct rndis_tcp_tso_info_ {
	+ union {
	+ struct {
	+ uint32_t unused:30;
	+ uint32_t type:1;
	+ uint32_t reserved2:1;
	+ } xmit;
	+ struct {
	+ uint32_t mss:20;
	+ uint32_t tcp_header_offset:10;
	+ uint32_t type:1;
	+ uint32_t reserved2:1;
	+ } lso_v1_xmit;
	+ struct {
	+ uint32_t tcp_payload:30;
	+ uint32_t type:1;
	+ uint32_t reserved2:1;
	+ } lso_v1_xmit_complete;
	+ struct {
	+ uint32_t mss:20;
	+ uint32_t tcp_header_offset:10;
	+ uint32_t type:1;
	+ uint32_t ip_version:1;
	+ } lso_v2_xmit;
	+ struct {
	+ uint32_t reserved:30;
	+ uint32_t type:1;
	+ uint32_t reserved2:1;
	+ } lso_v2_xmit_complete;
	+ uint32_t value;
	+ };
	+} rndis_tcp_tso_info;
	+
	+#define RNDIS_VLAN_PPI_SIZE (sizeof(rndis_per_packet_info) + \
	+ sizeof(ndis_8021q_info))
	+
	+#define RNDIS_CSUM_PPI_SIZE (sizeof(rndis_per_packet_info) + \
	+ sizeof(rndis_tcp_ip_csum_info))
	+
	+#define RNDIS_TSO_PPI_SIZE (sizeof(rndis_per_packet_info) + \
	+ sizeof(rndis_tcp_tso_info))
	+
	/*
	* Format of Information buffer passed in a SetRequest for the OID
	* OID_GEN_RNDIS_CONFIG_PARAMETER.
	@@ -906,6 +1041,22 @@
	#define NDIS_PACKET_TYPE_FUNCTIONAL 0x00000400
	#define NDIS_PACKET_TYPE_MAC_FRAME 0x00000800

	+/*
	+ * Externs
	+ */
	+extern int netvsc_recv
	+(
	+ struct hv_device *device_ctx,
	+ netvsc_packet *packet,
	+ rndis_tcp_ip_csum_info *csum_info
	+);
	+
	+extern void* hv_set_rppi_data(rndis_msg *rndis_mesg,
	+ uint32_t rppi_size,
	+ int pkt_type
	+);
	+
	+extern void* hv_get_ppi_data(rndis_packet *rpkt, uint32_t type);

	#endif /* __HV_RNDIS_H__ */

	Index: sys/dev/hyperv/netvsc/hv_rndis_filter.h
	===================================================================
	--- sys/dev/hyperv/netvsc/hv_rndis_filter.h
	+++ sys/dev/hyperv/netvsc/hv_rndis_filter.h
	@@ -90,27 +90,16 @@
	uint8_t hw_mac_addr[HW_MACADDR_LEN];
	} rndis_device;

	-typedef struct rndis_filter_packet_ {
	- void *completion_context;
	- /* No longer used */
	- pfn_on_send_rx_completion on_completion;
	-
	- rndis_msg message;
	-} rndis_filter_packet;
	-
	-
	/*
	* Externs
	*/

	-extern int hv_rf_on_receive(struct hv_device device, netvsc_packet pkt);
	+extern int hv_rf_on_receive(netvsc_dev net_dev, struct hv_device device, netvsc_packet *pkt);
	extern int hv_rf_on_device_add(struct hv_device device, void additl_info);
	extern int hv_rf_on_device_remove(struct hv_device *device,
	boolean_t destroy_channel);
	extern int hv_rf_on_open(struct hv_device *device);
	extern int hv_rf_on_close(struct hv_device *device);
	-extern int hv_rf_on_send(struct hv_device device, netvsc_packet pkt);
	-

	#endif /* __HV_RNDIS_FILTER_H__ */

	Index: sys/dev/hyperv/netvsc/hv_rndis_filter.c
	===================================================================
	--- sys/dev/hyperv/netvsc/hv_rndis_filter.c
	+++ sys/dev/hyperv/netvsc/hv_rndis_filter.c
	@@ -67,9 +67,68 @@
	static int hv_rf_init_device(rndis_device *device);
	static int hv_rf_open_device(rndis_device *device);
	static int hv_rf_close_device(rndis_device *device);
	-static void hv_rf_on_send_completion(void *context);
	static void hv_rf_on_send_request_completion(void *context);
	static void hv_rf_on_send_request_halt_completion(void *context);
	+int
	+hv_rf_send_offload_request(struct hv_device *device,
	+ rndis_offload_params *offloads);
	+/*
	+ * Set the Per-Packet-Info with the specified type
	+ */
	+extern void*
	+hv_set_rppi_data
	+(
	+ rndis_msg *rndis_mesg,
	+ uint32_t rppi_size,
	+ int pkt_type
	+)
	+{
	+ rndis_packet *rndis_pkt;
	+ rndis_per_packet_info *rppi;
	+
	+ rndis_pkt = &rndis_mesg->msg.packet;
	+ rndis_pkt->data_offset += rppi_size;
	+
	+ rppi = (rndis_per_packet_info )((char )rndis_pkt +
	+ rndis_pkt->per_pkt_info_offset + rndis_pkt->per_pkt_info_length);
	+
	+ rppi->size = rppi_size;
	+ rppi->type = pkt_type;
	+ rppi->per_packet_info_offset = sizeof(rndis_per_packet_info);
	+
	+ rndis_pkt->per_pkt_info_length += rppi_size;
	+
	+ return rppi;
	+}
	+
	+/*
	+ * Get the Per-Packet-Info with the specified type
	+ * return NULL if not found.
	+ */
	+extern void*
	+hv_get_ppi_data(rndis_packet *rpkt, uint32_t type)
	+{
	+ rndis_per_packet_info *ppi;
	+ int len;
	+
	+ if (rpkt->per_pkt_info_offset == 0)
	+ return NULL;
	+
	+ ppi = (rndis_per_packet_info *)((unsigned long)rpkt +
	+ rpkt->per_pkt_info_offset);
	+ len = rpkt->per_pkt_info_length;
	+
	+ while (len > 0) {
	+ if (ppi->type == type)
	+ return (void *)((unsigned long)ppi +
	+ ppi->per_packet_info_offset);
	+
	+ len -= ppi->size;
	+ ppi = (rndis_per_packet_info *)((unsigned long)ppi + ppi->size);
	+ }
	+
	+ return NULL;
	+}


	/*
	@@ -169,7 +228,7 @@
	*/
	static int
	hv_rf_send_request(rndis_device device, rndis_request request,
	- uint32_t message_type)
	+ uint32_t message_type)
	{
	int ret;
	netvsc_packet *packet;
	@@ -196,6 +255,9 @@
	hv_rf_on_send_request_halt_completion;
	}
	packet->compl.send.send_completion_tid = (unsigned long)device;
	+ packet->send_buf_section_idx =
	+ NVSP_1_CHIMNEY_SEND_INVALID_SECTION_INDEX;
	+ packet->send_buf_section_size = 0;

	ret = hv_nv_on_send(device->net_dev->dev, packet);

	@@ -248,6 +310,82 @@
	}
	}

	+int
	+hv_rf_send_offload_request(struct hv_device *device,
	+ rndis_offload_params *offloads)
	+{
	+ rndis_request *request;
	+ rndis_set_request *set;
	+ rndis_offload_params *offload_req;
	+ rndis_set_complete *set_complete;
	+ rndis_device *rndis_dev;
	+ hn_softc_t *sc = device_get_softc(device->device);
	+ netvsc_dev *net_dev = sc->net_dev;
	+ uint32_t vsp_version = net_dev->nvsp_version;
	+ uint32_t extlen = sizeof(rndis_offload_params);
	+ int ret;
	+
	+ if (vsp_version <= NVSP_PROTOCOL_VERSION_4) {
	+ extlen = VERSION_4_OFFLOAD_SIZE;
	+ /* On NVSP_PROTOCOL_VERSION_4 and below, we do not support
	+ * UDP checksum offload.
	+ */
	+ offloads->udp_ipv4_csum = 0;
	+ offloads->udp_ipv6_csum = 0;
	+ }
	+
	+ rndis_dev = net_dev->extension;
	+
	+ request = hv_rndis_request(rndis_dev, REMOTE_NDIS_SET_MSG,
	+ RNDIS_MESSAGE_SIZE(rndis_set_request) + extlen);
	+ if (!request)
	+ return -ENOMEM;
	+
	+ set = &request->request_msg.msg.set_request;
	+ set->oid = RNDIS_OID_TCP_OFFLOAD_PARAMETERS;
	+ set->info_buffer_length = extlen;
	+ set->info_buffer_offset = sizeof(rndis_set_request);
	+ set->device_vc_handle = 0;
	+
	+ offload_req = (rndis_offload_params *)((unsigned long)set +
	+ set->info_buffer_offset);
	+ offload_req = offloads;
	+ offload_req->header.type = RNDIS_OBJECT_TYPE_DEFAULT;
	+ offload_req->header.revision = RNDIS_OFFLOAD_PARAMETERS_REVISION_3;
	+ offload_req->header.size = extlen;
	+
	+ ret = hv_rf_send_request(rndis_dev, request, REMOTE_NDIS_SET_MSG);
	+ if (ret != 0) {
	+ printf("hv send offload request failed, ret=%d!\n", ret);
	+ goto cleanup;
	+ }
	+
	+ ret = sema_timedwait(&request->wait_sema, 500);
	+ if (ret != 0){
	+ printf("hv send offload request timeout\n");
	+ goto cleanup;
	+ }
	+
	+ set_complete = &request->response_msg.msg.set_complete;
	+ if (set_complete->status == RNDIS_STATUS_SUCCESS) {
	+ printf("hv send offload request success!!!\n");
	+ ret = 0;
	+ } else {
	+ if (set_complete->status == STATUS_NOT_SUPPORTED){
	+ printf("HV Not support offload\n");
	+ ret = 0;
	+ } else {
	+ ret = -1;
	+ }
	+ }
	+
	+cleanup:
	+ if (request)
	+ hv_put_rndis_request(rndis_dev, request);
	+
	+ return (ret);
	+}
	+
	/*
	* RNDIS filter receive indicate status
	*/
	@@ -272,9 +410,9 @@
	hv_rf_receive_data(rndis_device device, rndis_msg message, netvsc_packet *pkt)
	{
	rndis_packet *rndis_pkt;
	- rndis_per_packet_info *rppi;
	- ndis_8021q_info *rppi_vlan_info;
	+ ndis_8021q_info *rppi_vlan_info;
	uint32_t data_offset;
	+ rndis_tcp_ip_csum_info *csum_info = NULL;

	rndis_pkt = &message->msg.packet;

	@@ -286,88 +424,56 @@
	/* Remove rndis header, then pass data packet up the stack */
	data_offset = RNDIS_HEADER_SIZE + rndis_pkt->data_offset;

	- /* L2 frame length, with L2 header, not including CRC */
	- pkt->tot_data_buf_len = rndis_pkt->data_length;
	- pkt->page_buffers[0].offset += data_offset;
	- /* Buffer length now L2 frame length plus trailing junk */
	- pkt->page_buffers[0].length -= data_offset;
	-
	- pkt->is_data_pkt = TRUE;
	+ pkt->tot_data_buf_len -= data_offset;
	+ if (pkt->tot_data_buf_len < rndis_pkt->data_length) {
	+ pkt->status = nvsp_status_failure;
	+ printf("tot len %u is less than data len %u\n",
	+ pkt->tot_data_buf_len, rndis_pkt->data_length);
	+ return;
	+ }

	- pkt->vlan_tci = 0;
	+ pkt->tot_data_buf_len = rndis_pkt->data_length;
	+ pkt->data = (void *)((unsigned long)pkt->data + data_offset);

	- /*
	- * Read the VLAN ID if supplied by the Hyper-V infrastructure.
	- * Let higher-level driver code decide if it wants to use it.
	- * Ignore CFI, priority for now as FreeBSD does not support these.
	- */
	- if (rndis_pkt->per_pkt_info_offset != 0) {
	- /* rppi struct exists; compute its address */
	- rppi = (rndis_per_packet_info )((uint8_t )rndis_pkt +
	- rndis_pkt->per_pkt_info_offset);
	- /* if VLAN ppi struct, get the VLAN ID */
	- if (rppi->type == ieee_8021q_info) {
	- rppi_vlan_info = (ndis_8021q_info )((uint8_t )rppi
	- + rppi->per_packet_info_offset);
	- pkt->vlan_tci = rppi_vlan_info->u1.s1.vlan_id;
	- }
	+ rppi_vlan_info = hv_get_ppi_data(rndis_pkt, ieee_8021q_info);
	+ if (rppi_vlan_info) {
	+ pkt->vlan_tci = rppi_vlan_info->u1.s1.vlan_id;
	+ } else {
	+ pkt->vlan_tci = 0;
	}

	- netvsc_recv(device->net_dev->dev, pkt);
	+ csum_info = hv_get_ppi_data(rndis_pkt, tcpip_chksum_info);
	+ netvsc_recv(device->net_dev->dev, pkt, csum_info);
	}

	/*
	* RNDIS filter on receive
	*/
	int
	-hv_rf_on_receive(struct hv_device device, netvsc_packet pkt)
	+hv_rf_on_receive(netvsc_dev net_dev, struct hv_device device, netvsc_packet *pkt)
	{
	- hn_softc_t *sc = device_get_softc(device->device);
	- netvsc_dev *net_dev = sc->net_dev;
	rndis_device *rndis_dev;
	- rndis_msg rndis_mesg;
	rndis_msg *rndis_hdr;

	/* Make sure the rndis device state is initialized */
	- if (net_dev->extension == NULL)
	+ if (net_dev->extension == NULL) {
	+ pkt->status = nvsp_status_failure;
	return (ENODEV);
	+ }

	rndis_dev = (rndis_device *)net_dev->extension;
	- if (rndis_dev->state == RNDIS_DEV_UNINITIALIZED)
	+ if (rndis_dev->state == RNDIS_DEV_UNINITIALIZED) {
	+ pkt->status = nvsp_status_failure;
	return (EINVAL);
	-
	- /* Shift virtual page number to form virtual page address */
	- rndis_hdr = (rndis_msg *)(uintptr_t)(pkt->page_buffers[0].pfn << PAGE_SHIFT);
	-
	- rndis_hdr = (void *)((unsigned long)rndis_hdr
	- + pkt->page_buffers[0].offset);
	-
	- /*
	- * Make sure we got a valid rndis message
	- * Fixme: There seems to be a bug in set completion msg where
	- * its msg_len is 16 bytes but the byte_count field in the
	- * xfer page range shows 52 bytes
	- */
	-#if 0
	- if (pkt->tot_data_buf_len != rndis_hdr->msg_len) {
	- DPRINT_ERR(NETVSC, "invalid rndis message? (expected %u "
	- "bytes got %u)... dropping this message!",
	- rndis_hdr->msg_len, pkt->tot_data_buf_len);
	- DPRINT_EXIT(NETVSC);
	-
	- return (-1);
	}
	-#endif

	- memcpy(&rndis_mesg, rndis_hdr,
	- (rndis_hdr->msg_len > sizeof(rndis_msg)) ?
	- sizeof(rndis_msg) : rndis_hdr->msg_len);
	+ rndis_hdr = pkt->data;

	- switch (rndis_mesg.ndis_msg_type) {
	+ switch (rndis_hdr->ndis_msg_type) {

	/* data message */
	case REMOTE_NDIS_PACKET_MSG:
	- hv_rf_receive_data(rndis_dev, &rndis_mesg, pkt);
	+ hv_rf_receive_data(rndis_dev, rndis_hdr, pkt);
	break;
	/* completion messages */
	case REMOTE_NDIS_INITIALIZE_CMPLT:
	@@ -375,15 +481,15 @@
	case REMOTE_NDIS_SET_CMPLT:
	case REMOTE_NDIS_RESET_CMPLT:
	case REMOTE_NDIS_KEEPALIVE_CMPLT:
	- hv_rf_receive_response(rndis_dev, &rndis_mesg);
	+ hv_rf_receive_response(rndis_dev, rndis_hdr);
	break;
	/* notification message */
	case REMOTE_NDIS_INDICATE_STATUS_MSG:
	- hv_rf_receive_indicate_status(rndis_dev, &rndis_mesg);
	+ hv_rf_receive_indicate_status(rndis_dev, rndis_hdr);
	break;
	default:
	printf("hv_rf_on_receive(): Unknown msg_type 0x%x\n",
	- rndis_mesg.ndis_msg_type);
	+ rndis_hdr->ndis_msg_type);
	break;
	}

	@@ -711,6 +817,7 @@
	int ret;
	netvsc_dev *net_dev;
	rndis_device *rndis_dev;
	+ rndis_offload_params offloads;
	netvsc_device_info dev_info = (netvsc_device_info )additl_info;

	rndis_dev = hv_get_rndis_device();
	@@ -752,6 +859,21 @@
	if (ret != 0) {
	/* TODO: shut down rndis device and the channel */
	}
	+
	+ /* config csum offload and send request to host */
	+ memset(&offloads, 0, sizeof(offloads));
	+ offloads.ipv4_csum = RNDIS_OFFLOAD_PARAMETERS_TX_RX_ENABLED;
	+ offloads.tcp_ipv4_csum = RNDIS_OFFLOAD_PARAMETERS_TX_RX_ENABLED;
	+ offloads.udp_ipv4_csum = RNDIS_OFFLOAD_PARAMETERS_TX_RX_ENABLED;
	+ offloads.tcp_ipv6_csum = RNDIS_OFFLOAD_PARAMETERS_TX_RX_ENABLED;
	+ offloads.udp_ipv6_csum = RNDIS_OFFLOAD_PARAMETERS_TX_RX_ENABLED;
	+ offloads.lso_v2_ipv4 = RNDIS_OFFLOAD_PARAMETERS_LSOV2_ENABLED;
	+
	+ ret = hv_rf_send_offload_request(device, &offloads);
	+ if (ret != 0){
	+ /* TODO: shut down rndis device and the channel */
	+ printf("hv_rf_send_offload_request failed, ret=%d\n", ret);
	+ }

	memcpy(dev_info->mac_addr, rndis_dev->hw_mac_addr, HW_MACADDR_LEN);

	@@ -810,103 +932,6 @@
	}

	/*
	- * RNDIS filter on send
	- */
	-int
	-hv_rf_on_send(struct hv_device device, netvsc_packet pkt)
	-{
	- rndis_filter_packet *filter_pkt;
	- rndis_msg *rndis_mesg;
	- rndis_packet *rndis_pkt;
	- rndis_per_packet_info *rppi;
	- ndis_8021q_info *rppi_vlan_info;
	- uint32_t rndis_msg_size;
	- int ret = 0;
	-
	- /* Add the rndis header */
	- filter_pkt = (rndis_filter_packet *)pkt->extension;
	-
	- memset(filter_pkt, 0, sizeof(rndis_filter_packet));
	-
	- rndis_mesg = &filter_pkt->message;
	- rndis_msg_size = RNDIS_MESSAGE_SIZE(rndis_packet);
	-
	- if (pkt->vlan_tci != 0) {
	- rndis_msg_size += sizeof(rndis_per_packet_info) +
	- sizeof(ndis_8021q_info);
	- }
	-
	- rndis_mesg->ndis_msg_type = REMOTE_NDIS_PACKET_MSG;
	- rndis_mesg->msg_len = pkt->tot_data_buf_len + rndis_msg_size;
	-
	- rndis_pkt = &rndis_mesg->msg.packet;
	- rndis_pkt->data_offset = sizeof(rndis_packet);
	- rndis_pkt->data_length = pkt->tot_data_buf_len;
	-
	- pkt->is_data_pkt = TRUE;
	- pkt->page_buffers[0].pfn = hv_get_phys_addr(rndis_mesg) >> PAGE_SHIFT;
	- pkt->page_buffers[0].offset =
	- (unsigned long)rndis_mesg & (PAGE_SIZE - 1);
	- pkt->page_buffers[0].length = rndis_msg_size;
	-
	- /* Save the packet context */
	- filter_pkt->completion_context =
	- pkt->compl.send.send_completion_context;
	-
	- /* Use ours */
	- pkt->compl.send.on_send_completion = hv_rf_on_send_completion;
	- pkt->compl.send.send_completion_context = filter_pkt;
	-
	- /*
	- * If there is a VLAN tag, we need to set up some additional
	- * fields so the Hyper-V infrastructure will stuff the VLAN tag
	- * into the frame.
	- */
	- if (pkt->vlan_tci != 0) {
	- /* Move data offset past end of rppi + VLAN structs */
	- rndis_pkt->data_offset += sizeof(rndis_per_packet_info) +
	- sizeof(ndis_8021q_info);
	-
	- /* must be set when we have rppi, VLAN info */
	- rndis_pkt->per_pkt_info_offset = sizeof(rndis_packet);
	- rndis_pkt->per_pkt_info_length = sizeof(rndis_per_packet_info) +
	- sizeof(ndis_8021q_info);
	-
	- /* rppi immediately follows rndis_pkt */
	- rppi = (rndis_per_packet_info *)(rndis_pkt + 1);
	- rppi->size = sizeof(rndis_per_packet_info) +
	- sizeof(ndis_8021q_info);
	- rppi->type = ieee_8021q_info;
	- rppi->per_packet_info_offset = sizeof(rndis_per_packet_info);
	-
	- /* VLAN info immediately follows rppi struct */
	- rppi_vlan_info = (ndis_8021q_info *)(rppi + 1);
	- /* FreeBSD does not support CFI or priority */
	- rppi_vlan_info->u1.s1.vlan_id = pkt->vlan_tci & 0xfff;
	- }
	-
	- /*
	- * Invoke netvsc send. If return status is bad, the caller now
	- * resets the context pointers before retrying.
	- */
	- ret = hv_nv_on_send(device, pkt);
	-
	- return (ret);
	-}
	-
	-/*
	- * RNDIS filter on send completion callback
	- */
	-static void
	-hv_rf_on_send_completion(void *context)
	-{
	- rndis_filter_packet filter_pkt = (rndis_filter_packet )context;
	-
	- /* Pass it back to the original handler */
	- netvsc_xmit_completion(filter_pkt->completion_context);
	-}
	-
	-/*
	* RNDIS filter on send request completion callback
	*/
	static void

File Metadata

Mime Type: text/plain
Expires: Tue, Dec 3, 5:58 AM (19 h, 56 m)
Storage Engine: blob
Storage Format: Raw Data
Storage Handle: 15001761
Default Alt Text: D2517.id5333.diff (59 KB)

D2517.id5333.diffNo OneTemporaryActions

D2517.id5333.diffView Options

File Metadata

Event Timeline

D2517.id5333.diff
No OneTemporary
Actions

D2517.id5333.diff
View Options