1/*
  2 * Copyright (c) 2009, Microsoft Corporation.
  3 *
  4 * This program is free software; you can redistribute it and/or modify it
  5 * under the terms and conditions of the GNU General Public License,
  6 * version 2, as published by the Free Software Foundation.
  7 *
  8 * This program is distributed in the hope it will be useful, but WITHOUT
  9 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 10 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 11 * more details.
 12 *
 13 * You should have received a copy of the GNU General Public License along with
 14 * this program; if not, see <http://www.gnu.org/licenses/>.
 15 *
 16 * Authors:
 17 *   Haiyang Zhang <haiyangz@microsoft.com>
 18 *   Hank Janssen  <hjanssen@microsoft.com>
 19 */
 20#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 21
 22#include <linux/init.h>
 23#include <linux/atomic.h>
 
 24#include <linux/module.h>
 25#include <linux/highmem.h>
 26#include <linux/device.h>
 27#include <linux/io.h>
 28#include <linux/delay.h>
 29#include <linux/netdevice.h>
 30#include <linux/inetdevice.h>
 31#include <linux/etherdevice.h>
 
 32#include <linux/skbuff.h>
 33#include <linux/if_vlan.h>
 34#include <linux/in.h>
 35#include <linux/slab.h>
 
 
 
 
 36#include <net/arp.h>
 37#include <net/route.h>
 38#include <net/sock.h>
 39#include <net/pkt_sched.h>
 
 
 40
 41#include "hyperv_net.h"
 42
 43struct net_device_context {
 44	/* point back to our device context */
 45	struct hv_device *device_ctx;
 46	struct delayed_work dwork;
 47	struct work_struct work;
 48};
 
 
 49
 50#define RING_SIZE_MIN 64
 51static int ring_size = 128;
 52module_param(ring_size, int, S_IRUGO);
 53MODULE_PARM_DESC(ring_size, "Ring buffer size (# of pages)");
 54
 55static void do_set_multicast(struct work_struct *w)
 
 
 
 
 
 
 
 
 
 
 
 56{
 57	struct net_device_context *ndevctx =
 58		container_of(w, struct net_device_context, work);
 59	struct netvsc_device *nvdev;
 60	struct rndis_device *rdev;
 61
 62	nvdev = hv_get_drvdata(ndevctx->device_ctx);
 63	if (nvdev == NULL || nvdev->ndev == NULL)
 64		return;
 65
 66	rdev = nvdev->extension;
 67	if (rdev == NULL)
 68		return;
 
 69
 70	if (nvdev->ndev->flags & IFF_PROMISC)
 71		rndis_filter_set_packet_filter(rdev,
 72			NDIS_PACKET_TYPE_PROMISCUOUS);
 73	else
 74		rndis_filter_set_packet_filter(rdev,
 75			NDIS_PACKET_TYPE_BROADCAST |
 76			NDIS_PACKET_TYPE_ALL_MULTICAST |
 77			NDIS_PACKET_TYPE_DIRECTED);
 78}
 79
 80static void netvsc_set_multicast_list(struct net_device *net)
 81{
 82	struct net_device_context *net_device_ctx = netdev_priv(net);
 
 
 83
 84	schedule_work(&net_device_ctx->work);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 85}
 86
 87static int netvsc_open(struct net_device *net)
 88{
 89	struct net_device_context *net_device_ctx = netdev_priv(net);
 90	struct hv_device *device_obj = net_device_ctx->device_ctx;
 91	struct netvsc_device *nvdev;
 92	struct rndis_device *rdev;
 93	int ret = 0;
 94
 95	netif_carrier_off(net);
 96
 97	/* Open up the device */
 98	ret = rndis_filter_open(device_obj);
 99	if (ret != 0) {
100		netdev_err(net, "unable to open device (ret %d).\n", ret);
101		return ret;
102	}
103
104	netif_start_queue(net);
105
106	nvdev = hv_get_drvdata(device_obj);
107	rdev = nvdev->extension;
108	if (!rdev->link_state)
109		netif_carrier_on(net);
 
 
110
111	return ret;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
112}
113
114static int netvsc_close(struct net_device *net)
115{
116	struct net_device_context *net_device_ctx = netdev_priv(net);
117	struct hv_device *device_obj = net_device_ctx->device_ctx;
 
 
118	int ret;
119
120	netif_tx_disable(net);
121
122	/* Make sure netvsc_set_multicast_list doesn't re-enable filter! */
123	cancel_work_sync(&net_device_ctx->work);
124	ret = rndis_filter_close(device_obj);
125	if (ret != 0)
 
 
126		netdev_err(net, "unable to close device (ret %d).\n", ret);
 
 
 
 
 
 
 
 
 
127
128	return ret;
129}
130
131static void *init_ppi_data(struct rndis_message *msg, u32 ppi_size,
132				int pkt_type)
133{
134	struct rndis_packet *rndis_pkt;
135	struct rndis_per_packet_info *ppi;
136
137	rndis_pkt = &msg->msg.pkt;
138	rndis_pkt->data_offset += ppi_size;
139
140	ppi = (struct rndis_per_packet_info *)((void *)rndis_pkt +
141		rndis_pkt->per_pkt_info_offset + rndis_pkt->per_pkt_info_len);
142
143	ppi->size = ppi_size;
144	ppi->type = pkt_type;
 
145	ppi->ppi_offset = sizeof(struct rndis_per_packet_info);
146
147	rndis_pkt->per_pkt_info_len += ppi_size;
148
149	return ppi;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
150}
151
152static void netvsc_xmit_completion(void *context)
 
153{
154	struct hv_netvsc_packet *packet = (struct hv_netvsc_packet *)context;
155	struct sk_buff *skb = (struct sk_buff *)
156		(unsigned long)packet->completion.send.send_completion_tid;
157
158	kfree(packet);
 
 
 
 
 
 
 
 
159
160	if (skb)
161		dev_kfree_skb_any(skb);
 
 
 
 
 
 
 
 
 
 
 
 
162}
163
164static u32 fill_pg_buf(struct page *page, u32 offset, u32 len,
165			struct hv_page_buffer *pb)
166{
167	int j = 0;
168
169	/* Deal with compund pages by ignoring unused part
170	 * of the page.
171	 */
172	page += (offset >> PAGE_SHIFT);
173	offset &= ~PAGE_MASK;
174
175	while (len > 0) {
176		unsigned long bytes;
177
178		bytes = PAGE_SIZE - offset;
179		if (bytes > len)
180			bytes = len;
181		pb[j].pfn = page_to_pfn(page);
182		pb[j].offset = offset;
183		pb[j].len = bytes;
184
185		offset += bytes;
186		len -= bytes;
187
188		if (offset == PAGE_SIZE && len) {
189			page++;
190			offset = 0;
191			j++;
192		}
193	}
194
195	return j + 1;
196}
197
198static u32 init_page_array(void *hdr, u32 len, struct sk_buff *skb,
 
199			   struct hv_page_buffer *pb)
200{
201	u32 slots_used = 0;
202	char *data = skb->data;
203	int frags = skb_shinfo(skb)->nr_frags;
204	int i;
205
206	/* The packet is laid out thus:
207	 * 1. hdr
208	 * 2. skb linear data
209	 * 3. skb fragment data
210	 */
211	if (hdr != NULL)
212		slots_used += fill_pg_buf(virt_to_page(hdr),
213					offset_in_page(hdr),
214					len, &pb[slots_used]);
215
216	slots_used += fill_pg_buf(virt_to_page(data),
217				offset_in_page(data),
218				skb_headlen(skb), &pb[slots_used]);
 
 
 
 
219
220	for (i = 0; i < frags; i++) {
221		skb_frag_t *frag = skb_shinfo(skb)->frags + i;
222
223		slots_used += fill_pg_buf(skb_frag_page(frag),
224					frag->page_offset,
225					skb_frag_size(frag), &pb[slots_used]);
 
226	}
227	return slots_used;
228}
229
230static int count_skb_frag_slots(struct sk_buff *skb)
231{
232	int i, frags = skb_shinfo(skb)->nr_frags;
233	int pages = 0;
234
235	for (i = 0; i < frags; i++) {
236		skb_frag_t *frag = skb_shinfo(skb)->frags + i;
237		unsigned long size = skb_frag_size(frag);
238		unsigned long offset = frag->page_offset;
239
240		/* Skip unused frames from start of page */
241		offset &= ~PAGE_MASK;
242		pages += PFN_UP(offset + size);
243	}
244	return pages;
245}
246
247static int netvsc_get_slots(struct sk_buff *skb)
248{
249	char *data = skb->data;
250	unsigned int offset = offset_in_page(data);
251	unsigned int len = skb_headlen(skb);
252	int slots;
253	int frag_slots;
254
255	slots = DIV_ROUND_UP(offset + len, PAGE_SIZE);
256	frag_slots = count_skb_frag_slots(skb);
257	return slots + frag_slots;
258}
259
260static u32 get_net_transport_info(struct sk_buff *skb, u32 *trans_off)
261{
262	u32 ret_val = TRANSPORT_INFO_NOT_IP;
 
263
264	if ((eth_hdr(skb)->h_proto != htons(ETH_P_IP)) &&
265		(eth_hdr(skb)->h_proto != htons(ETH_P_IPV6))) {
266		goto not_ip;
 
 
 
 
 
 
 
 
267	}
268
269	*trans_off = skb_transport_offset(skb);
 
270
271	if ((eth_hdr(skb)->h_proto == htons(ETH_P_IP))) {
272		struct iphdr *iphdr = ip_hdr(skb);
 
 
 
 
 
 
 
 
 
 
 
 
 
273
274		if (iphdr->protocol == IPPROTO_TCP)
275			ret_val = TRANSPORT_INFO_IPV4_TCP;
276		else if (iphdr->protocol == IPPROTO_UDP)
277			ret_val = TRANSPORT_INFO_IPV4_UDP;
278	} else {
279		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
280			ret_val = TRANSPORT_INFO_IPV6_TCP;
281		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
282			ret_val = TRANSPORT_INFO_IPV6_UDP;
283	}
284
285not_ip:
286	return ret_val;
287}
288
289static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
290{
291	struct net_device_context *net_device_ctx = netdev_priv(net);
292	struct hv_netvsc_packet *packet;
293	int ret;
294	unsigned int num_data_pgs;
295	struct rndis_message *rndis_msg;
296	struct rndis_packet *rndis_pkt;
297	u32 rndis_msg_size;
298	bool isvlan;
299	struct rndis_per_packet_info *ppi;
300	struct ndis_tcp_ip_checksum_info *csum_info;
301	struct ndis_tcp_lso_info *lso_info;
302	int  hdr_offset;
303	u32 net_trans_info;
304
 
 
 
 
 
 
 
 
 
305
306	/* We will atmost need two pages to describe the rndis
307	 * header. We can only transmit MAX_PAGE_BUFFER_COUNT number
308	 * of pages in a single packet.
 
309	 */
 
310	num_data_pgs = netvsc_get_slots(skb) + 2;
311	if (num_data_pgs > MAX_PAGE_BUFFER_COUNT) {
312		netdev_err(net, "Packet too big: %u\n", skb->len);
313		dev_kfree_skb(skb);
314		net->stats.tx_dropped++;
315		return NETDEV_TX_OK;
316	}
317
318	/* Allocate a netvsc packet based on # of frags. */
319	packet = kzalloc(sizeof(struct hv_netvsc_packet) +
320			 (num_data_pgs * sizeof(struct hv_page_buffer)) +
321			 sizeof(struct rndis_message) +
322			 NDIS_VLAN_PPI_SIZE +
323			 NDIS_CSUM_PPI_SIZE +
324			 NDIS_LSO_PPI_SIZE, GFP_ATOMIC);
325	if (!packet) {
326		/* out of memory, drop packet */
327		netdev_err(net, "unable to allocate hv_netvsc_packet\n");
328
329		dev_kfree_skb(skb);
330		net->stats.tx_dropped++;
331		return NETDEV_TX_OK;
 
 
 
 
 
332	}
333
334	packet->vlan_tci = skb->vlan_tci;
 
 
 
 
 
 
 
 
 
 
 
 
335
336	packet->is_data_pkt = true;
337	packet->total_data_buflen = skb->len;
338
339	packet->rndis_msg = (struct rndis_message *)((unsigned long)packet +
340				sizeof(struct hv_netvsc_packet) +
341				(num_data_pgs * sizeof(struct hv_page_buffer)));
342
343	/* Set the completion routine */
344	packet->completion.send.send_completion = netvsc_xmit_completion;
345	packet->completion.send.send_completion_ctx = packet;
346	packet->completion.send.send_completion_tid = (unsigned long)skb;
347
348	isvlan = packet->vlan_tci & VLAN_TAG_PRESENT;
349
350	/* Add the rndis header */
351	rndis_msg = packet->rndis_msg;
352	rndis_msg->ndis_msg_type = RNDIS_MSG_PACKET;
353	rndis_msg->msg_len = packet->total_data_buflen;
354	rndis_pkt = &rndis_msg->msg.pkt;
355	rndis_pkt->data_offset = sizeof(struct rndis_packet);
356	rndis_pkt->data_len = packet->total_data_buflen;
357	rndis_pkt->per_pkt_info_offset = sizeof(struct rndis_packet);
358
359	rndis_msg_size = RNDIS_MESSAGE_SIZE(struct rndis_packet);
 
 
 
 
360
361	if (isvlan) {
362		struct ndis_pkt_8021q_info *vlan;
363
364		rndis_msg_size += NDIS_VLAN_PPI_SIZE;
365		ppi = init_ppi_data(rndis_msg, NDIS_VLAN_PPI_SIZE,
366					IEEE_8021Q_INFO);
367		vlan = (struct ndis_pkt_8021q_info *)((void *)ppi +
368						ppi->ppi_offset);
369		vlan->vlanid = packet->vlan_tci & VLAN_VID_MASK;
370		vlan->pri = (packet->vlan_tci & VLAN_PRIO_MASK) >>
371				VLAN_PRIO_SHIFT;
372	}
373
374	net_trans_info = get_net_transport_info(skb, &hdr_offset);
375	if (net_trans_info == TRANSPORT_INFO_NOT_IP)
376		goto do_send;
377
378	/*
379	 * Setup the sendside checksum offload only if this is not a
380	 * GSO packet.
381	 */
382	if (skb_is_gso(skb))
383		goto do_lso;
384
385	if ((skb->ip_summed == CHECKSUM_NONE) ||
386	    (skb->ip_summed == CHECKSUM_UNNECESSARY))
387		goto do_send;
388
389	rndis_msg_size += NDIS_CSUM_PPI_SIZE;
390	ppi = init_ppi_data(rndis_msg, NDIS_CSUM_PPI_SIZE,
391			    TCPIP_CHKSUM_PKTINFO);
392
393	csum_info = (struct ndis_tcp_ip_checksum_info *)((void *)ppi +
394			ppi->ppi_offset);
395
396	if (net_trans_info & (INFO_IPV4 << 16))
397		csum_info->transmit.is_ipv4 = 1;
398	else
399		csum_info->transmit.is_ipv6 = 1;
 
 
 
 
 
 
 
 
 
 
 
400
401	if (net_trans_info & INFO_TCP) {
402		csum_info->transmit.tcp_checksum = 1;
403		csum_info->transmit.tcp_header_offset = hdr_offset;
404	} else if (net_trans_info & INFO_UDP) {
405		/* UDP checksum offload is not supported on ws2008r2.
406		 * Furthermore, on ws2012 and ws2012r2, there are some
407		 * issues with udp checksum offload from Linux guests.
408		 * (these are host issues).
409		 * For now compute the checksum here.
410		 */
411		struct udphdr *uh;
412		u16 udp_len;
413
414		ret = skb_cow_head(skb, 0);
415		if (ret)
416			goto drop;
417
418		uh = udp_hdr(skb);
419		udp_len = ntohs(uh->len);
420		uh->check = 0;
421		uh->check = csum_tcpudp_magic(ip_hdr(skb)->saddr,
422					      ip_hdr(skb)->daddr,
423					      udp_len, IPPROTO_UDP,
424					      csum_partial(uh, udp_len, 0));
425		if (uh->check == 0)
426			uh->check = CSUM_MANGLED_0;
427
428		csum_info->transmit.udp_checksum = 0;
429	}
430	goto do_send;
431
432do_lso:
433	rndis_msg_size += NDIS_LSO_PPI_SIZE;
434	ppi = init_ppi_data(rndis_msg, NDIS_LSO_PPI_SIZE,
435			    TCP_LARGESEND_PKTINFO);
436
437	lso_info = (struct ndis_tcp_lso_info *)((void *)ppi +
438			ppi->ppi_offset);
439
440	lso_info->lso_v2_transmit.type = NDIS_TCP_LARGE_SEND_OFFLOAD_V2_TYPE;
441	if (net_trans_info & (INFO_IPV4 << 16)) {
442		lso_info->lso_v2_transmit.ip_version =
443			NDIS_TCP_LARGE_SEND_OFFLOAD_IPV4;
444		ip_hdr(skb)->tot_len = 0;
445		ip_hdr(skb)->check = 0;
446		tcp_hdr(skb)->check =
447		~csum_tcpudp_magic(ip_hdr(skb)->saddr,
448				   ip_hdr(skb)->daddr, 0, IPPROTO_TCP, 0);
449	} else {
450		lso_info->lso_v2_transmit.ip_version =
451			NDIS_TCP_LARGE_SEND_OFFLOAD_IPV6;
452		ipv6_hdr(skb)->payload_len = 0;
453		tcp_hdr(skb)->check =
454		~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
455				&ipv6_hdr(skb)->daddr, 0, IPPROTO_TCP, 0);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
456	}
457	lso_info->lso_v2_transmit.tcp_header_offset = hdr_offset;
458	lso_info->lso_v2_transmit.mss = skb_shinfo(skb)->gso_size;
459
460do_send:
461	/* Start filling in the page buffers with the rndis hdr */
462	rndis_msg->msg_len += rndis_msg_size;
 
463	packet->page_buf_cnt = init_page_array(rndis_msg, rndis_msg_size,
464					skb, &packet->page_buf[0]);
465
466	ret = netvsc_send(net_device_ctx->device_ctx, packet);
 
467
468drop:
469	if (ret == 0) {
470		net->stats.tx_bytes += skb->len;
471		net->stats.tx_packets++;
472	} else {
473		kfree(packet);
474		if (ret != -EAGAIN) {
475			dev_kfree_skb_any(skb);
476			net->stats.tx_dropped++;
477		}
478	}
479
480	return (ret == -EAGAIN) ? NETDEV_TX_BUSY : NETDEV_TX_OK;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
481}
482
483/*
484 * netvsc_linkstatus_callback - Link up/down notification
485 */
486void netvsc_linkstatus_callback(struct hv_device *device_obj,
487				       unsigned int status)
488{
489	struct net_device *net;
490	struct net_device_context *ndev_ctx;
491	struct netvsc_device *net_device;
492	struct rndis_device *rdev;
 
 
 
 
 
 
 
 
493
494	net_device = hv_get_drvdata(device_obj);
495	rdev = net_device->extension;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
496
497	rdev->link_state = status != 1;
 
 
 
498
499	net = net_device->ndev;
 
 
 
 
500
501	if (!net || net->reg_state != NETREG_REGISTERED)
502		return;
503
504	ndev_ctx = netdev_priv(net);
505	if (status == 1) {
506		schedule_delayed_work(&ndev_ctx->dwork, 0);
507		schedule_delayed_work(&ndev_ctx->dwork, msecs_to_jiffies(20));
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
508	} else {
509		schedule_delayed_work(&ndev_ctx->dwork, 0);
 
 
 
 
 
 
 
 
 
 
510	}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
511}
512
513/*
514 * netvsc_recv_callback -  Callback when we receive a packet from the
515 * "wire" on the specified device.
516 */
517int netvsc_recv_callback(struct hv_device *device_obj,
518				struct hv_netvsc_packet *packet,
519				struct ndis_tcp_ip_checksum_info *csum_info)
520{
521	struct net_device *net;
 
 
522	struct sk_buff *skb;
 
 
 
523
524	net = ((struct netvsc_device *)hv_get_drvdata(device_obj))->ndev;
525	if (!net || net->reg_state != NETREG_REGISTERED) {
526		packet->status = NVSP_STAT_FAIL;
527		return 0;
 
 
 
 
 
 
 
 
 
 
528	}
529
530	/* Allocate a skb - TODO direct I/O to pages? */
531	skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen);
 
532	if (unlikely(!skb)) {
533		++net->stats.rx_dropped;
534		packet->status = NVSP_STAT_FAIL;
535		return 0;
536	}
537
 
 
538	/*
539	 * Copy to skb. This copy is needed here since the memory pointed by
540	 * hv_netvsc_packet cannot be deallocated
 
541	 */
542	memcpy(skb_put(skb, packet->total_data_buflen), packet->data,
543		packet->total_data_buflen);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
544
545	skb->protocol = eth_type_trans(skb, net);
546	if (csum_info) {
547		/* We only look at the IP checksum here.
548		 * Should we be dropping the packet if checksum
549		 * failed? How do we deal with other checksums - TCP/UDP?
550		 */
551		if (csum_info->receive.ip_checksum_succeeded)
552			skb->ip_summed = CHECKSUM_UNNECESSARY;
553		else
554			skb->ip_summed = CHECKSUM_NONE;
 
 
 
 
 
 
 
 
 
 
555	}
 
556
557	if (packet->vlan_tci & VLAN_TAG_PRESENT)
558		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
559				       packet->vlan_tci);
 
 
 
 
 
560
561	net->stats.rx_packets++;
562	net->stats.rx_bytes += packet->total_data_buflen;
563
564	/*
565	 * Pass the skb back up. Network stack will deallocate the skb when it
566	 * is done.
567	 * TODO - use NAPI?
568	 */
569	netif_rx(skb);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
570
571	return 0;
572}
573
574static void netvsc_get_drvinfo(struct net_device *net,
575			       struct ethtool_drvinfo *info)
576{
577	strlcpy(info->driver, KBUILD_MODNAME, sizeof(info->driver));
578	strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
579}
580
581static int netvsc_change_mtu(struct net_device *ndev, int mtu)
582{
583	struct net_device_context *ndevctx = netdev_priv(ndev);
584	struct hv_device *hdev =  ndevctx->device_ctx;
585	struct netvsc_device *nvdev = hv_get_drvdata(hdev);
586	struct netvsc_device_info device_info;
587	int limit = ETH_DATA_LEN;
 
588
589	if (nvdev == NULL || nvdev->destroy)
590		return -ENODEV;
591
592	if (nvdev->nvsp_version >= NVSP_PROTOCOL_VERSION_2)
593		limit = NETVSC_MTU;
594
595	if (mtu < 68 || mtu > limit)
596		return -EINVAL;
597
598	nvdev->start_remove = true;
599	cancel_work_sync(&ndevctx->work);
600	netif_tx_disable(ndev);
601	rndis_filter_device_remove(hdev);
 
 
 
 
 
 
602
603	ndev->mtu = mtu;
604
605	ndevctx->device_ctx = hdev;
606	hv_set_drvdata(hdev, ndev);
607	device_info.ring_size = ring_size;
608	rndis_filter_device_add(hdev, &device_info);
609	netif_wake_queue(ndev);
 
 
 
 
 
 
 
610
611	return 0;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
612}
613
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
614
615static int netvsc_set_mac_addr(struct net_device *ndev, void *p)
616{
617	struct net_device_context *ndevctx = netdev_priv(ndev);
618	struct hv_device *hdev =  ndevctx->device_ctx;
 
619	struct sockaddr *addr = p;
620	char save_adr[ETH_ALEN];
621	unsigned char save_aatype;
622	int err;
623
624	memcpy(save_adr, ndev->dev_addr, ETH_ALEN);
625	save_aatype = ndev->addr_assign_type;
626
627	err = eth_mac_addr(ndev, p);
628	if (err != 0)
629		return err;
630
631	err = rndis_filter_set_device_mac(hdev, addr->sa_data);
632	if (err != 0) {
633		/* roll back to saved MAC */
634		memcpy(ndev->dev_addr, save_adr, ETH_ALEN);
635		ndev->addr_assign_type = save_aatype;
 
 
 
 
 
 
 
 
 
 
 
636	}
637
638	return err;
639}
640
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
641
642static const struct ethtool_ops ethtool_ops = {
643	.get_drvinfo	= netvsc_get_drvinfo,
 
 
 
 
644	.get_link	= ethtool_op_get_link,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
645};
646
647static const struct net_device_ops device_ops = {
648	.ndo_open =			netvsc_open,
649	.ndo_stop =			netvsc_close,
650	.ndo_start_xmit =		netvsc_start_xmit,
651	.ndo_set_rx_mode =		netvsc_set_multicast_list,
 
 
 
652	.ndo_change_mtu =		netvsc_change_mtu,
653	.ndo_validate_addr =		eth_validate_addr,
654	.ndo_set_mac_address =		netvsc_set_mac_addr,
 
 
 
 
655};
656
657/*
658 * Send GARP packet to network peers after migrations.
659 * After Quick Migration, the network is not immediately operational in the
660 * current context when receiving RNDIS_STATUS_MEDIA_CONNECT event. So, add
661 * another netif_notify_peers() into a delayed work, otherwise GARP packet
662 * will not be sent after quick migration, and cause network disconnection.
663 * Also, we update the carrier status here.
664 */
665static void netvsc_link_change(struct work_struct *w)
666{
667	struct net_device_context *ndev_ctx;
668	struct net_device *net;
 
 
 
 
669	struct netvsc_device *net_device;
670	struct rndis_device *rdev;
671	bool notify;
672
673	rtnl_lock();
 
 
 
 
 
 
 
 
674
675	ndev_ctx = container_of(w, struct net_device_context, dwork.work);
676	net_device = hv_get_drvdata(ndev_ctx->device_ctx);
677	rdev = net_device->extension;
678	net = net_device->ndev;
679
680	if (rdev->link_state) {
681		netif_carrier_off(net);
682		notify = false;
683	} else {
684		netif_carrier_on(net);
685		notify = true;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
686	}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
687
688	rtnl_unlock();
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
689
690	if (notify)
691		netdev_notify_peers(net);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
692}
693
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
694
695static int netvsc_probe(struct hv_device *dev,
696			const struct hv_vmbus_device_id *dev_id)
697{
698	struct net_device *net = NULL;
699	struct net_device_context *net_device_ctx;
700	struct netvsc_device_info device_info;
701	int ret;
 
702
703	net = alloc_etherdev(sizeof(struct net_device_context));
 
704	if (!net)
705		return -ENOMEM;
706
707	netif_carrier_off(net);
708
 
 
709	net_device_ctx = netdev_priv(net);
710	net_device_ctx->device_ctx = dev;
 
 
 
 
 
711	hv_set_drvdata(dev, net);
 
712	INIT_DELAYED_WORK(&net_device_ctx->dwork, netvsc_link_change);
713	INIT_WORK(&net_device_ctx->work, do_set_multicast);
 
 
 
 
 
 
 
 
 
714
715	net->netdev_ops = &device_ops;
 
 
 
716
717	net->hw_features = NETIF_F_RXCSUM | NETIF_F_SG | NETIF_F_IP_CSUM |
718				NETIF_F_TSO;
719	net->features = NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_SG | NETIF_F_RXCSUM |
720			NETIF_F_IP_CSUM | NETIF_F_TSO;
721
722	SET_ETHTOOL_OPS(net, &ethtool_ops);
723	SET_NETDEV_DEV(net, &dev->device);
 
 
 
724
725	/* Notify the netvsc driver of the new device */
726	device_info.ring_size = ring_size;
727	ret = rndis_filter_device_add(dev, &device_info);
728	if (ret != 0) {
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
729		netdev_err(net, "unable to add netvsc device (ret %d)\n", ret);
730		free_netdev(net);
731		hv_set_drvdata(dev, NULL);
732		return ret;
733	}
734	memcpy(net->dev_addr, device_info.mac_adr, ETH_ALEN);
735
736	ret = register_netdev(net);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
737	if (ret != 0) {
738		pr_err("Unable to register netdev.\n");
739		rndis_filter_device_remove(dev);
740		free_netdev(net);
741	} else {
742		schedule_delayed_work(&net_device_ctx->dwork, 0);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
743	}
 
 
 
 
744
 
 
 
 
 
 
 
 
 
 
 
745	return ret;
746}
747
748static int netvsc_remove(struct hv_device *dev)
749{
750	struct net_device *net;
751	struct net_device_context *ndev_ctx;
752	struct netvsc_device *net_device;
753
754	net_device = hv_get_drvdata(dev);
755	net = net_device->ndev;
756
 
757	if (net == NULL) {
758		dev_err(&dev->device, "No net device to remove\n");
759		return 0;
760	}
761
762	net_device->start_remove = true;
763
764	ndev_ctx = netdev_priv(net);
765	cancel_delayed_work_sync(&ndev_ctx->dwork);
766	cancel_work_sync(&ndev_ctx->work);
767
768	/* Stop outbound asap */
769	netif_tx_disable(net);
770
771	unregister_netdev(net);
 
 
 
 
 
772
773	/*
774	 * Call to the vsc driver to let it know that the device is being
775	 * removed
776	 */
777	rndis_filter_device_remove(dev);
 
 
778
 
 
 
 
 
 
 
 
 
 
 
779	free_netdev(net);
780	return 0;
781}
782
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
783static const struct hv_vmbus_device_id id_table[] = {
784	/* Network guid */
785	{ HV_NIC_GUID, },
786	{ },
787};
788
789MODULE_DEVICE_TABLE(vmbus, id_table);
790
791/* The one and only one */
792static struct  hv_driver netvsc_drv = {
793	.name = KBUILD_MODNAME,
794	.id_table = id_table,
795	.probe = netvsc_probe,
796	.remove = netvsc_remove,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
797};
798
799static void __exit netvsc_drv_exit(void)
800{
 
801	vmbus_driver_unregister(&netvsc_drv);
802}
803
804static int __init netvsc_drv_init(void)
805{
 
 
806	if (ring_size < RING_SIZE_MIN) {
807		ring_size = RING_SIZE_MIN;
808		pr_info("Increased ring_size to %d (min allowed)\n",
809			ring_size);
810	}
811	return vmbus_driver_register(&netvsc_drv);
 
 
 
 
 
 
 
 
 
 
 
 
812}
813
814MODULE_LICENSE("GPL");
815MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
816
817module_init(netvsc_drv_init);
818module_exit(netvsc_drv_exit);