Commit | Line | Data |
---|---|---|
fceaf24a | 1 | /* |
fceaf24a HJ |
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 | |
adf8d3ff | 14 | * this program; if not, see <http://www.gnu.org/licenses/>. |
fceaf24a HJ |
15 | * |
16 | * Authors: | |
d0e94d17 | 17 | * Haiyang Zhang <haiyangz@microsoft.com> |
fceaf24a | 18 | * Hank Janssen <hjanssen@microsoft.com> |
fceaf24a | 19 | */ |
eb335bc4 HJ |
20 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
21 | ||
fceaf24a | 22 | #include <linux/init.h> |
9079ce69 | 23 | #include <linux/atomic.h> |
fceaf24a HJ |
24 | #include <linux/module.h> |
25 | #include <linux/highmem.h> | |
26 | #include <linux/device.h> | |
fceaf24a | 27 | #include <linux/io.h> |
fceaf24a HJ |
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> | |
c802db11 | 33 | #include <linux/if_vlan.h> |
fceaf24a | 34 | #include <linux/in.h> |
5a0e3ad6 | 35 | #include <linux/slab.h> |
fceaf24a HJ |
36 | #include <net/arp.h> |
37 | #include <net/route.h> | |
38 | #include <net/sock.h> | |
39 | #include <net/pkt_sched.h> | |
3f335ea2 | 40 | |
5ca7252a | 41 | #include "hyperv_net.h" |
fceaf24a | 42 | |
fceaf24a | 43 | |
fa85a6c2 | 44 | #define RING_SIZE_MIN 64 |
99c8da0f | 45 | static int ring_size = 128; |
450d7a4b SH |
46 | module_param(ring_size, int, S_IRUGO); |
47 | MODULE_PARM_DESC(ring_size, "Ring buffer size (# of pages)"); | |
fceaf24a | 48 | |
3f300ff4 SX |
49 | static const u32 default_msg = NETIF_MSG_DRV | NETIF_MSG_PROBE | |
50 | NETIF_MSG_LINK | NETIF_MSG_IFUP | | |
51 | NETIF_MSG_IFDOWN | NETIF_MSG_RX_ERR | | |
52 | NETIF_MSG_TX_ERR; | |
53 | ||
54 | static int debug = -1; | |
55 | module_param(debug, int, S_IRUGO); | |
56 | MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)"); | |
57 | ||
d426b2e3 HZ |
58 | static void do_set_multicast(struct work_struct *w) |
59 | { | |
792df872 WM |
60 | struct net_device_context *ndevctx = |
61 | container_of(w, struct net_device_context, work); | |
d426b2e3 HZ |
62 | struct netvsc_device *nvdev; |
63 | struct rndis_device *rdev; | |
64 | ||
65 | nvdev = hv_get_drvdata(ndevctx->device_ctx); | |
792df872 WM |
66 | if (nvdev == NULL || nvdev->ndev == NULL) |
67 | return; | |
d426b2e3 HZ |
68 | |
69 | rdev = nvdev->extension; | |
70 | if (rdev == NULL) | |
792df872 | 71 | return; |
d426b2e3 | 72 | |
792df872 | 73 | if (nvdev->ndev->flags & IFF_PROMISC) |
d426b2e3 HZ |
74 | rndis_filter_set_packet_filter(rdev, |
75 | NDIS_PACKET_TYPE_PROMISCUOUS); | |
76 | else | |
77 | rndis_filter_set_packet_filter(rdev, | |
78 | NDIS_PACKET_TYPE_BROADCAST | | |
79 | NDIS_PACKET_TYPE_ALL_MULTICAST | | |
80 | NDIS_PACKET_TYPE_DIRECTED); | |
d426b2e3 HZ |
81 | } |
82 | ||
4e9bfefa | 83 | static void netvsc_set_multicast_list(struct net_device *net) |
fceaf24a | 84 | { |
792df872 | 85 | struct net_device_context *net_device_ctx = netdev_priv(net); |
d426b2e3 | 86 | |
792df872 | 87 | schedule_work(&net_device_ctx->work); |
fceaf24a HJ |
88 | } |
89 | ||
fceaf24a HJ |
90 | static int netvsc_open(struct net_device *net) |
91 | { | |
fceaf24a | 92 | struct net_device_context *net_device_ctx = netdev_priv(net); |
6bad88da | 93 | struct hv_device *device_obj = net_device_ctx->device_ctx; |
891de74d HZ |
94 | struct netvsc_device *nvdev; |
95 | struct rndis_device *rdev; | |
02fafbc6 | 96 | int ret = 0; |
fceaf24a | 97 | |
891de74d HZ |
98 | netif_carrier_off(net); |
99 | ||
d515d0ff HZ |
100 | /* Open up the device */ |
101 | ret = rndis_filter_open(device_obj); | |
102 | if (ret != 0) { | |
103 | netdev_err(net, "unable to open device (ret %d).\n", ret); | |
104 | return ret; | |
fceaf24a HJ |
105 | } |
106 | ||
5b54dac8 | 107 | netif_tx_start_all_queues(net); |
d515d0ff | 108 | |
891de74d HZ |
109 | nvdev = hv_get_drvdata(device_obj); |
110 | rdev = nvdev->extension; | |
111 | if (!rdev->link_state) | |
112 | netif_carrier_on(net); | |
113 | ||
fceaf24a HJ |
114 | return ret; |
115 | } | |
116 | ||
fceaf24a HJ |
117 | static int netvsc_close(struct net_device *net) |
118 | { | |
fceaf24a | 119 | struct net_device_context *net_device_ctx = netdev_priv(net); |
6bad88da | 120 | struct hv_device *device_obj = net_device_ctx->device_ctx; |
02fafbc6 | 121 | int ret; |
fceaf24a | 122 | |
0a282538 | 123 | netif_tx_disable(net); |
fceaf24a | 124 | |
792df872 WM |
125 | /* Make sure netvsc_set_multicast_list doesn't re-enable filter! */ |
126 | cancel_work_sync(&net_device_ctx->work); | |
9c26aa0d | 127 | ret = rndis_filter_close(device_obj); |
fceaf24a | 128 | if (ret != 0) |
eb335bc4 | 129 | netdev_err(net, "unable to close device (ret %d).\n", ret); |
fceaf24a | 130 | |
fceaf24a HJ |
131 | return ret; |
132 | } | |
133 | ||
8a00251a KS |
134 | static void *init_ppi_data(struct rndis_message *msg, u32 ppi_size, |
135 | int pkt_type) | |
136 | { | |
137 | struct rndis_packet *rndis_pkt; | |
138 | struct rndis_per_packet_info *ppi; | |
139 | ||
140 | rndis_pkt = &msg->msg.pkt; | |
141 | rndis_pkt->data_offset += ppi_size; | |
142 | ||
143 | ppi = (struct rndis_per_packet_info *)((void *)rndis_pkt + | |
144 | rndis_pkt->per_pkt_info_offset + rndis_pkt->per_pkt_info_len); | |
145 | ||
146 | ppi->size = ppi_size; | |
147 | ppi->type = pkt_type; | |
148 | ppi->ppi_offset = sizeof(struct rndis_per_packet_info); | |
149 | ||
150 | rndis_pkt->per_pkt_info_len += ppi_size; | |
151 | ||
152 | return ppi; | |
153 | } | |
154 | ||
5b54dac8 HZ |
155 | union sub_key { |
156 | u64 k; | |
157 | struct { | |
158 | u8 pad[3]; | |
159 | u8 kb; | |
160 | u32 ka; | |
161 | }; | |
162 | }; | |
163 | ||
164 | /* Toeplitz hash function | |
165 | * data: network byte order | |
166 | * return: host byte order | |
167 | */ | |
f88e6714 | 168 | static u32 comp_hash(u8 *key, int klen, void *data, int dlen) |
5b54dac8 HZ |
169 | { |
170 | union sub_key subk; | |
171 | int k_next = 4; | |
172 | u8 dt; | |
173 | int i, j; | |
174 | u32 ret = 0; | |
175 | ||
176 | subk.k = 0; | |
177 | subk.ka = ntohl(*(u32 *)key); | |
178 | ||
179 | for (i = 0; i < dlen; i++) { | |
180 | subk.kb = key[k_next]; | |
181 | k_next = (k_next + 1) % klen; | |
f88e6714 | 182 | dt = ((u8 *)data)[i]; |
5b54dac8 HZ |
183 | for (j = 0; j < 8; j++) { |
184 | if (dt & 0x80) | |
185 | ret ^= subk.ka; | |
186 | dt <<= 1; | |
187 | subk.k <<= 1; | |
188 | } | |
189 | } | |
190 | ||
191 | return ret; | |
192 | } | |
193 | ||
194 | static bool netvsc_set_hash(u32 *hash, struct sk_buff *skb) | |
195 | { | |
f88e6714 | 196 | struct flow_keys flow; |
5b54dac8 | 197 | int data_len; |
5b54dac8 | 198 | |
06635a35 JP |
199 | if (!skb_flow_dissect_flow_keys(skb, &flow) || |
200 | !(flow.basic.n_proto == htons(ETH_P_IP) || | |
201 | flow.basic.n_proto == htons(ETH_P_IPV6))) | |
5b54dac8 HZ |
202 | return false; |
203 | ||
06635a35 | 204 | if (flow.basic.ip_proto == IPPROTO_TCP) |
f88e6714 HZ |
205 | data_len = 12; |
206 | else | |
207 | data_len = 8; | |
5b54dac8 | 208 | |
f88e6714 | 209 | *hash = comp_hash(netvsc_hash_key, HASH_KEYLEN, &flow, data_len); |
5b54dac8 | 210 | |
f88e6714 | 211 | return true; |
5b54dac8 HZ |
212 | } |
213 | ||
214 | static u16 netvsc_select_queue(struct net_device *ndev, struct sk_buff *skb, | |
215 | void *accel_priv, select_queue_fallback_t fallback) | |
216 | { | |
217 | struct net_device_context *net_device_ctx = netdev_priv(ndev); | |
218 | struct hv_device *hdev = net_device_ctx->device_ctx; | |
219 | struct netvsc_device *nvsc_dev = hv_get_drvdata(hdev); | |
220 | u32 hash; | |
221 | u16 q_idx = 0; | |
222 | ||
223 | if (nvsc_dev == NULL || ndev->real_num_tx_queues <= 1) | |
224 | return 0; | |
225 | ||
307f0995 | 226 | if (netvsc_set_hash(&hash, skb)) { |
5b54dac8 HZ |
227 | q_idx = nvsc_dev->send_table[hash % VRSS_SEND_TAB_SIZE] % |
228 | ndev->real_num_tx_queues; | |
307f0995 HZ |
229 | skb_set_hash(skb, hash, PKT_HASH_TYPE_L3); |
230 | } | |
5b54dac8 HZ |
231 | |
232 | return q_idx; | |
233 | } | |
234 | ||
ee90b812 | 235 | void netvsc_xmit_completion(void *context) |
fceaf24a | 236 | { |
4193d4f4 | 237 | struct hv_netvsc_packet *packet = (struct hv_netvsc_packet *)context; |
02fafbc6 | 238 | struct sk_buff *skb = (struct sk_buff *) |
893f6627 | 239 | (unsigned long)packet->send_completion_tid; |
fceaf24a | 240 | |
cbacec76 | 241 | if (skb) |
b220f5f9 | 242 | dev_kfree_skb_any(skb); |
fceaf24a HJ |
243 | } |
244 | ||
54a7357f KS |
245 | static u32 fill_pg_buf(struct page *page, u32 offset, u32 len, |
246 | struct hv_page_buffer *pb) | |
247 | { | |
248 | int j = 0; | |
249 | ||
250 | /* Deal with compund pages by ignoring unused part | |
251 | * of the page. | |
252 | */ | |
253 | page += (offset >> PAGE_SHIFT); | |
254 | offset &= ~PAGE_MASK; | |
255 | ||
256 | while (len > 0) { | |
257 | unsigned long bytes; | |
258 | ||
259 | bytes = PAGE_SIZE - offset; | |
260 | if (bytes > len) | |
261 | bytes = len; | |
262 | pb[j].pfn = page_to_pfn(page); | |
263 | pb[j].offset = offset; | |
264 | pb[j].len = bytes; | |
265 | ||
266 | offset += bytes; | |
267 | len -= bytes; | |
268 | ||
269 | if (offset == PAGE_SIZE && len) { | |
270 | page++; | |
271 | offset = 0; | |
272 | j++; | |
273 | } | |
274 | } | |
275 | ||
276 | return j + 1; | |
277 | } | |
278 | ||
8a00251a | 279 | static u32 init_page_array(void *hdr, u32 len, struct sk_buff *skb, |
aa0a34be | 280 | struct hv_netvsc_packet *packet) |
54a7357f | 281 | { |
aa0a34be | 282 | struct hv_page_buffer *pb = packet->page_buf; |
54a7357f KS |
283 | u32 slots_used = 0; |
284 | char *data = skb->data; | |
285 | int frags = skb_shinfo(skb)->nr_frags; | |
286 | int i; | |
287 | ||
288 | /* The packet is laid out thus: | |
aa0a34be | 289 | * 1. hdr: RNDIS header and PPI |
54a7357f KS |
290 | * 2. skb linear data |
291 | * 3. skb fragment data | |
292 | */ | |
293 | if (hdr != NULL) | |
294 | slots_used += fill_pg_buf(virt_to_page(hdr), | |
295 | offset_in_page(hdr), | |
296 | len, &pb[slots_used]); | |
297 | ||
aa0a34be HZ |
298 | packet->rmsg_size = len; |
299 | packet->rmsg_pgcnt = slots_used; | |
300 | ||
54a7357f KS |
301 | slots_used += fill_pg_buf(virt_to_page(data), |
302 | offset_in_page(data), | |
303 | skb_headlen(skb), &pb[slots_used]); | |
304 | ||
305 | for (i = 0; i < frags; i++) { | |
306 | skb_frag_t *frag = skb_shinfo(skb)->frags + i; | |
307 | ||
308 | slots_used += fill_pg_buf(skb_frag_page(frag), | |
309 | frag->page_offset, | |
310 | skb_frag_size(frag), &pb[slots_used]); | |
311 | } | |
8a00251a | 312 | return slots_used; |
54a7357f KS |
313 | } |
314 | ||
315 | static int count_skb_frag_slots(struct sk_buff *skb) | |
316 | { | |
317 | int i, frags = skb_shinfo(skb)->nr_frags; | |
318 | int pages = 0; | |
319 | ||
320 | for (i = 0; i < frags; i++) { | |
321 | skb_frag_t *frag = skb_shinfo(skb)->frags + i; | |
322 | unsigned long size = skb_frag_size(frag); | |
323 | unsigned long offset = frag->page_offset; | |
324 | ||
325 | /* Skip unused frames from start of page */ | |
326 | offset &= ~PAGE_MASK; | |
327 | pages += PFN_UP(offset + size); | |
328 | } | |
329 | return pages; | |
330 | } | |
331 | ||
332 | static int netvsc_get_slots(struct sk_buff *skb) | |
333 | { | |
334 | char *data = skb->data; | |
335 | unsigned int offset = offset_in_page(data); | |
336 | unsigned int len = skb_headlen(skb); | |
337 | int slots; | |
338 | int frag_slots; | |
339 | ||
340 | slots = DIV_ROUND_UP(offset + len, PAGE_SIZE); | |
341 | frag_slots = count_skb_frag_slots(skb); | |
342 | return slots + frag_slots; | |
343 | } | |
344 | ||
08cd04bf KS |
345 | static u32 get_net_transport_info(struct sk_buff *skb, u32 *trans_off) |
346 | { | |
347 | u32 ret_val = TRANSPORT_INFO_NOT_IP; | |
348 | ||
349 | if ((eth_hdr(skb)->h_proto != htons(ETH_P_IP)) && | |
350 | (eth_hdr(skb)->h_proto != htons(ETH_P_IPV6))) { | |
351 | goto not_ip; | |
352 | } | |
353 | ||
354 | *trans_off = skb_transport_offset(skb); | |
355 | ||
356 | if ((eth_hdr(skb)->h_proto == htons(ETH_P_IP))) { | |
357 | struct iphdr *iphdr = ip_hdr(skb); | |
358 | ||
359 | if (iphdr->protocol == IPPROTO_TCP) | |
360 | ret_val = TRANSPORT_INFO_IPV4_TCP; | |
361 | else if (iphdr->protocol == IPPROTO_UDP) | |
362 | ret_val = TRANSPORT_INFO_IPV4_UDP; | |
363 | } else { | |
364 | if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP) | |
365 | ret_val = TRANSPORT_INFO_IPV6_TCP; | |
366 | else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP) | |
367 | ret_val = TRANSPORT_INFO_IPV6_UDP; | |
368 | } | |
369 | ||
370 | not_ip: | |
371 | return ret_val; | |
372 | } | |
373 | ||
02fafbc6 | 374 | static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net) |
fceaf24a | 375 | { |
fceaf24a | 376 | struct net_device_context *net_device_ctx = netdev_priv(net); |
981a1bd8 | 377 | struct hv_netvsc_packet *packet = NULL; |
02fafbc6 | 378 | int ret; |
8a00251a KS |
379 | unsigned int num_data_pgs; |
380 | struct rndis_message *rndis_msg; | |
381 | struct rndis_packet *rndis_pkt; | |
382 | u32 rndis_msg_size; | |
383 | bool isvlan; | |
e88f7e07 | 384 | bool linear = false; |
8a00251a | 385 | struct rndis_per_packet_info *ppi; |
08cd04bf | 386 | struct ndis_tcp_ip_checksum_info *csum_info; |
77bf5487 | 387 | struct ndis_tcp_lso_info *lso_info; |
08cd04bf KS |
388 | int hdr_offset; |
389 | u32 net_trans_info; | |
307f0995 | 390 | u32 hash; |
e88f7e07 | 391 | u32 skb_length; |
b08cc791 KS |
392 | u32 pkt_sz; |
393 | struct hv_page_buffer page_buf[MAX_PAGE_BUFFER_COUNT]; | |
7eafd9b4 | 394 | struct netvsc_stats *tx_stats = this_cpu_ptr(net_device_ctx->tx_stats); |
fceaf24a | 395 | |
54a7357f KS |
396 | /* We will atmost need two pages to describe the rndis |
397 | * header. We can only transmit MAX_PAGE_BUFFER_COUNT number | |
e88f7e07 VK |
398 | * of pages in a single packet. If skb is scattered around |
399 | * more pages we try linearizing it. | |
54a7357f | 400 | */ |
e88f7e07 VK |
401 | |
402 | check_size: | |
403 | skb_length = skb->len; | |
8a00251a | 404 | num_data_pgs = netvsc_get_slots(skb) + 2; |
e88f7e07 VK |
405 | if (num_data_pgs > MAX_PAGE_BUFFER_COUNT && linear) { |
406 | net_alert_ratelimited("packet too big: %u pages (%u bytes)\n", | |
407 | num_data_pgs, skb->len); | |
981a1bd8 VK |
408 | ret = -EFAULT; |
409 | goto drop; | |
e88f7e07 VK |
410 | } else if (num_data_pgs > MAX_PAGE_BUFFER_COUNT) { |
411 | if (skb_linearize(skb)) { | |
412 | net_alert_ratelimited("failed to linearize skb\n"); | |
413 | ret = -ENOMEM; | |
414 | goto drop; | |
415 | } | |
416 | linear = true; | |
417 | goto check_size; | |
54a7357f | 418 | } |
fceaf24a | 419 | |
72151422 | 420 | pkt_sz = sizeof(struct hv_netvsc_packet) + RNDIS_AND_PPI_SIZE; |
b08cc791 | 421 | |
b56fc3c5 KS |
422 | ret = skb_cow_head(skb, pkt_sz); |
423 | if (ret) { | |
424 | netdev_err(net, "unable to alloc hv_netvsc_packet\n"); | |
425 | ret = -ENOMEM; | |
426 | goto drop; | |
fceaf24a | 427 | } |
b56fc3c5 KS |
428 | /* Use the headroom for building up the packet */ |
429 | packet = (struct hv_netvsc_packet *)skb->head; | |
fceaf24a | 430 | |
b08cc791 | 431 | packet->status = 0; |
7c3877f2 HZ |
432 | packet->xmit_more = skb->xmit_more; |
433 | ||
1f5f3a75 | 434 | packet->vlan_tci = skb->vlan_tci; |
b08cc791 | 435 | packet->page_buf = page_buf; |
1f5f3a75 | 436 | |
5b54dac8 HZ |
437 | packet->q_idx = skb_get_queue_mapping(skb); |
438 | ||
8a00251a | 439 | packet->is_data_pkt = true; |
4d447c9a | 440 | packet->total_data_buflen = skb->len; |
fceaf24a | 441 | |
8a00251a | 442 | packet->rndis_msg = (struct rndis_message *)((unsigned long)packet + |
b08cc791 KS |
443 | sizeof(struct hv_netvsc_packet)); |
444 | ||
72151422 | 445 | memset(packet->rndis_msg, 0, RNDIS_AND_PPI_SIZE); |
fceaf24a | 446 | |
454f18a9 | 447 | /* Set the completion routine */ |
893f6627 HZ |
448 | packet->send_completion = netvsc_xmit_completion; |
449 | packet->send_completion_ctx = packet; | |
450 | packet->send_completion_tid = (unsigned long)skb; | |
fceaf24a | 451 | |
8a00251a KS |
452 | isvlan = packet->vlan_tci & VLAN_TAG_PRESENT; |
453 | ||
454 | /* Add the rndis header */ | |
455 | rndis_msg = packet->rndis_msg; | |
456 | rndis_msg->ndis_msg_type = RNDIS_MSG_PACKET; | |
457 | rndis_msg->msg_len = packet->total_data_buflen; | |
458 | rndis_pkt = &rndis_msg->msg.pkt; | |
459 | rndis_pkt->data_offset = sizeof(struct rndis_packet); | |
460 | rndis_pkt->data_len = packet->total_data_buflen; | |
461 | rndis_pkt->per_pkt_info_offset = sizeof(struct rndis_packet); | |
462 | ||
463 | rndis_msg_size = RNDIS_MESSAGE_SIZE(struct rndis_packet); | |
464 | ||
307f0995 HZ |
465 | hash = skb_get_hash_raw(skb); |
466 | if (hash != 0 && net->real_num_tx_queues > 1) { | |
467 | rndis_msg_size += NDIS_HASH_PPI_SIZE; | |
468 | ppi = init_ppi_data(rndis_msg, NDIS_HASH_PPI_SIZE, | |
469 | NBL_HASH_VALUE); | |
470 | *(u32 *)((void *)ppi + ppi->ppi_offset) = hash; | |
471 | } | |
472 | ||
8a00251a KS |
473 | if (isvlan) { |
474 | struct ndis_pkt_8021q_info *vlan; | |
475 | ||
476 | rndis_msg_size += NDIS_VLAN_PPI_SIZE; | |
477 | ppi = init_ppi_data(rndis_msg, NDIS_VLAN_PPI_SIZE, | |
478 | IEEE_8021Q_INFO); | |
479 | vlan = (struct ndis_pkt_8021q_info *)((void *)ppi + | |
480 | ppi->ppi_offset); | |
481 | vlan->vlanid = packet->vlan_tci & VLAN_VID_MASK; | |
482 | vlan->pri = (packet->vlan_tci & VLAN_PRIO_MASK) >> | |
483 | VLAN_PRIO_SHIFT; | |
484 | } | |
485 | ||
08cd04bf KS |
486 | net_trans_info = get_net_transport_info(skb, &hdr_offset); |
487 | if (net_trans_info == TRANSPORT_INFO_NOT_IP) | |
488 | goto do_send; | |
489 | ||
490 | /* | |
491 | * Setup the sendside checksum offload only if this is not a | |
492 | * GSO packet. | |
493 | */ | |
494 | if (skb_is_gso(skb)) | |
77bf5487 | 495 | goto do_lso; |
08cd04bf | 496 | |
22041fb0 KS |
497 | if ((skb->ip_summed == CHECKSUM_NONE) || |
498 | (skb->ip_summed == CHECKSUM_UNNECESSARY)) | |
499 | goto do_send; | |
500 | ||
08cd04bf KS |
501 | rndis_msg_size += NDIS_CSUM_PPI_SIZE; |
502 | ppi = init_ppi_data(rndis_msg, NDIS_CSUM_PPI_SIZE, | |
503 | TCPIP_CHKSUM_PKTINFO); | |
504 | ||
505 | csum_info = (struct ndis_tcp_ip_checksum_info *)((void *)ppi + | |
506 | ppi->ppi_offset); | |
507 | ||
508 | if (net_trans_info & (INFO_IPV4 << 16)) | |
509 | csum_info->transmit.is_ipv4 = 1; | |
510 | else | |
511 | csum_info->transmit.is_ipv6 = 1; | |
512 | ||
513 | if (net_trans_info & INFO_TCP) { | |
514 | csum_info->transmit.tcp_checksum = 1; | |
515 | csum_info->transmit.tcp_header_offset = hdr_offset; | |
516 | } else if (net_trans_info & INFO_UDP) { | |
af9893a3 KS |
517 | /* UDP checksum offload is not supported on ws2008r2. |
518 | * Furthermore, on ws2012 and ws2012r2, there are some | |
519 | * issues with udp checksum offload from Linux guests. | |
520 | * (these are host issues). | |
521 | * For now compute the checksum here. | |
522 | */ | |
523 | struct udphdr *uh; | |
524 | u16 udp_len; | |
525 | ||
526 | ret = skb_cow_head(skb, 0); | |
527 | if (ret) | |
528 | goto drop; | |
529 | ||
530 | uh = udp_hdr(skb); | |
531 | udp_len = ntohs(uh->len); | |
532 | uh->check = 0; | |
533 | uh->check = csum_tcpudp_magic(ip_hdr(skb)->saddr, | |
534 | ip_hdr(skb)->daddr, | |
535 | udp_len, IPPROTO_UDP, | |
536 | csum_partial(uh, udp_len, 0)); | |
537 | if (uh->check == 0) | |
538 | uh->check = CSUM_MANGLED_0; | |
539 | ||
540 | csum_info->transmit.udp_checksum = 0; | |
08cd04bf | 541 | } |
77bf5487 KS |
542 | goto do_send; |
543 | ||
544 | do_lso: | |
545 | rndis_msg_size += NDIS_LSO_PPI_SIZE; | |
546 | ppi = init_ppi_data(rndis_msg, NDIS_LSO_PPI_SIZE, | |
547 | TCP_LARGESEND_PKTINFO); | |
548 | ||
549 | lso_info = (struct ndis_tcp_lso_info *)((void *)ppi + | |
550 | ppi->ppi_offset); | |
551 | ||
552 | lso_info->lso_v2_transmit.type = NDIS_TCP_LARGE_SEND_OFFLOAD_V2_TYPE; | |
553 | if (net_trans_info & (INFO_IPV4 << 16)) { | |
554 | lso_info->lso_v2_transmit.ip_version = | |
555 | NDIS_TCP_LARGE_SEND_OFFLOAD_IPV4; | |
556 | ip_hdr(skb)->tot_len = 0; | |
557 | ip_hdr(skb)->check = 0; | |
558 | tcp_hdr(skb)->check = | |
559 | ~csum_tcpudp_magic(ip_hdr(skb)->saddr, | |
560 | ip_hdr(skb)->daddr, 0, IPPROTO_TCP, 0); | |
561 | } else { | |
562 | lso_info->lso_v2_transmit.ip_version = | |
563 | NDIS_TCP_LARGE_SEND_OFFLOAD_IPV6; | |
564 | ipv6_hdr(skb)->payload_len = 0; | |
565 | tcp_hdr(skb)->check = | |
566 | ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr, | |
567 | &ipv6_hdr(skb)->daddr, 0, IPPROTO_TCP, 0); | |
568 | } | |
569 | lso_info->lso_v2_transmit.tcp_header_offset = hdr_offset; | |
570 | lso_info->lso_v2_transmit.mss = skb_shinfo(skb)->gso_size; | |
08cd04bf KS |
571 | |
572 | do_send: | |
8a00251a KS |
573 | /* Start filling in the page buffers with the rndis hdr */ |
574 | rndis_msg->msg_len += rndis_msg_size; | |
942396b0 | 575 | packet->total_data_buflen = rndis_msg->msg_len; |
8a00251a | 576 | packet->page_buf_cnt = init_page_array(rndis_msg, rndis_msg_size, |
aa0a34be | 577 | skb, packet); |
8a00251a KS |
578 | |
579 | ret = netvsc_send(net_device_ctx->device_ctx, packet); | |
580 | ||
af9893a3 | 581 | drop: |
02fafbc6 | 582 | if (ret == 0) { |
7eafd9b4 | 583 | u64_stats_update_begin(&tx_stats->s_sync); |
584 | tx_stats->packets++; | |
585 | tx_stats->bytes += skb_length; | |
586 | u64_stats_update_end(&tx_stats->s_sync); | |
b220f5f9 | 587 | } else { |
33be96e4 HZ |
588 | if (ret != -EAGAIN) { |
589 | dev_kfree_skb_any(skb); | |
590 | net->stats.tx_dropped++; | |
591 | } | |
fceaf24a HJ |
592 | } |
593 | ||
33be96e4 | 594 | return (ret == -EAGAIN) ? NETDEV_TX_BUSY : NETDEV_TX_OK; |
fceaf24a HJ |
595 | } |
596 | ||
3e189519 | 597 | /* |
02fafbc6 GKH |
598 | * netvsc_linkstatus_callback - Link up/down notification |
599 | */ | |
90ef117a | 600 | void netvsc_linkstatus_callback(struct hv_device *device_obj, |
3a494e71 | 601 | struct rndis_message *resp) |
fceaf24a | 602 | { |
3a494e71 | 603 | struct rndis_indicate_status *indicate = &resp->msg.indicate_status; |
2ddd5e5f | 604 | struct net_device *net; |
c996edcf | 605 | struct net_device_context *ndev_ctx; |
2ddd5e5f | 606 | struct netvsc_device *net_device; |
891de74d | 607 | struct rndis_device *rdev; |
2ddd5e5f S |
608 | |
609 | net_device = hv_get_drvdata(device_obj); | |
891de74d HZ |
610 | rdev = net_device->extension; |
611 | ||
3a494e71 HZ |
612 | switch (indicate->status) { |
613 | case RNDIS_STATUS_MEDIA_CONNECT: | |
614 | rdev->link_state = false; | |
615 | break; | |
616 | case RNDIS_STATUS_MEDIA_DISCONNECT: | |
617 | rdev->link_state = true; | |
618 | break; | |
619 | case RNDIS_STATUS_NETWORK_CHANGE: | |
620 | rdev->link_change = true; | |
621 | break; | |
622 | default: | |
623 | return; | |
624 | } | |
891de74d | 625 | |
2ddd5e5f | 626 | net = net_device->ndev; |
fceaf24a | 627 | |
891de74d | 628 | if (!net || net->reg_state != NETREG_REGISTERED) |
fceaf24a | 629 | return; |
fceaf24a | 630 | |
891de74d | 631 | ndev_ctx = netdev_priv(net); |
3a494e71 | 632 | if (!rdev->link_state) { |
c4b6a2ea | 633 | schedule_delayed_work(&ndev_ctx->dwork, 0); |
122a5f64 | 634 | schedule_delayed_work(&ndev_ctx->dwork, msecs_to_jiffies(20)); |
02fafbc6 | 635 | } else { |
891de74d | 636 | schedule_delayed_work(&ndev_ctx->dwork, 0); |
fceaf24a | 637 | } |
fceaf24a HJ |
638 | } |
639 | ||
3e189519 HJ |
640 | /* |
641 | * netvsc_recv_callback - Callback when we receive a packet from the | |
642 | * "wire" on the specified device. | |
02fafbc6 | 643 | */ |
f79adf8f | 644 | int netvsc_recv_callback(struct hv_device *device_obj, |
e3d605ed KS |
645 | struct hv_netvsc_packet *packet, |
646 | struct ndis_tcp_ip_checksum_info *csum_info) | |
fceaf24a | 647 | { |
6f4c4446 | 648 | struct net_device *net; |
7eafd9b4 | 649 | struct net_device_context *net_device_ctx; |
fceaf24a | 650 | struct sk_buff *skb; |
7eafd9b4 | 651 | struct netvsc_stats *rx_stats; |
fceaf24a | 652 | |
6f4c4446 | 653 | net = ((struct netvsc_device *)hv_get_drvdata(device_obj))->ndev; |
a68f9614 | 654 | if (!net || net->reg_state != NETREG_REGISTERED) { |
63f6921d | 655 | packet->status = NVSP_STAT_FAIL; |
fceaf24a HJ |
656 | return 0; |
657 | } | |
7eafd9b4 | 658 | net_device_ctx = netdev_priv(net); |
659 | rx_stats = this_cpu_ptr(net_device_ctx->rx_stats); | |
fceaf24a | 660 | |
9495c282 | 661 | /* Allocate a skb - TODO direct I/O to pages? */ |
72a2f5bd | 662 | skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen); |
9495c282 SH |
663 | if (unlikely(!skb)) { |
664 | ++net->stats.rx_dropped; | |
63f6921d | 665 | packet->status = NVSP_STAT_FAIL; |
9495c282 SH |
666 | return 0; |
667 | } | |
fceaf24a | 668 | |
02fafbc6 GKH |
669 | /* |
670 | * Copy to skb. This copy is needed here since the memory pointed by | |
671 | * hv_netvsc_packet cannot be deallocated | |
672 | */ | |
45326342 HZ |
673 | memcpy(skb_put(skb, packet->total_data_buflen), packet->data, |
674 | packet->total_data_buflen); | |
fceaf24a HJ |
675 | |
676 | skb->protocol = eth_type_trans(skb, net); | |
e3d605ed KS |
677 | if (csum_info) { |
678 | /* We only look at the IP checksum here. | |
679 | * Should we be dropping the packet if checksum | |
680 | * failed? How do we deal with other checksums - TCP/UDP? | |
681 | */ | |
682 | if (csum_info->receive.ip_checksum_succeeded) | |
683 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
684 | else | |
685 | skb->ip_summed = CHECKSUM_NONE; | |
686 | } | |
687 | ||
93725cbd HZ |
688 | if (packet->vlan_tci & VLAN_TAG_PRESENT) |
689 | __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), | |
690 | packet->vlan_tci); | |
fceaf24a | 691 | |
4baab261 | 692 | skb_record_rx_queue(skb, packet->channel-> |
e565e803 | 693 | offermsg.offer.sub_channel_index); |
5b54dac8 | 694 | |
7eafd9b4 | 695 | u64_stats_update_begin(&rx_stats->s_sync); |
696 | rx_stats->packets++; | |
697 | rx_stats->bytes += packet->total_data_buflen; | |
698 | u64_stats_update_end(&rx_stats->s_sync); | |
9495c282 | 699 | |
02fafbc6 GKH |
700 | /* |
701 | * Pass the skb back up. Network stack will deallocate the skb when it | |
9495c282 SH |
702 | * is done. |
703 | * TODO - use NAPI? | |
02fafbc6 | 704 | */ |
9495c282 | 705 | netif_rx(skb); |
fceaf24a | 706 | |
fceaf24a HJ |
707 | return 0; |
708 | } | |
709 | ||
f82f4ad7 SH |
710 | static void netvsc_get_drvinfo(struct net_device *net, |
711 | struct ethtool_drvinfo *info) | |
712 | { | |
7826d43f | 713 | strlcpy(info->driver, KBUILD_MODNAME, sizeof(info->driver)); |
7826d43f | 714 | strlcpy(info->fw_version, "N/A", sizeof(info->fw_version)); |
f82f4ad7 SH |
715 | } |
716 | ||
59995370 AS |
717 | static void netvsc_get_channels(struct net_device *net, |
718 | struct ethtool_channels *channel) | |
719 | { | |
720 | struct net_device_context *net_device_ctx = netdev_priv(net); | |
721 | struct hv_device *dev = net_device_ctx->device_ctx; | |
722 | struct netvsc_device *nvdev = hv_get_drvdata(dev); | |
723 | ||
724 | if (nvdev) { | |
725 | channel->max_combined = nvdev->max_chn; | |
726 | channel->combined_count = nvdev->num_chn; | |
727 | } | |
728 | } | |
729 | ||
4d447c9a HZ |
730 | static int netvsc_change_mtu(struct net_device *ndev, int mtu) |
731 | { | |
732 | struct net_device_context *ndevctx = netdev_priv(ndev); | |
733 | struct hv_device *hdev = ndevctx->device_ctx; | |
734 | struct netvsc_device *nvdev = hv_get_drvdata(hdev); | |
735 | struct netvsc_device_info device_info; | |
736 | int limit = ETH_DATA_LEN; | |
737 | ||
738 | if (nvdev == NULL || nvdev->destroy) | |
739 | return -ENODEV; | |
740 | ||
a1eabb01 | 741 | if (nvdev->nvsp_version >= NVSP_PROTOCOL_VERSION_2) |
4d3c9d37 | 742 | limit = NETVSC_MTU - ETH_HLEN; |
4d447c9a | 743 | |
4d3c9d37 HZ |
744 | /* Hyper-V hosts don't support MTU < ETH_DATA_LEN (1500) */ |
745 | if (mtu < ETH_DATA_LEN || mtu > limit) | |
4d447c9a HZ |
746 | return -EINVAL; |
747 | ||
748 | nvdev->start_remove = true; | |
792df872 | 749 | cancel_work_sync(&ndevctx->work); |
0a282538 | 750 | netif_tx_disable(ndev); |
4d447c9a HZ |
751 | rndis_filter_device_remove(hdev); |
752 | ||
753 | ndev->mtu = mtu; | |
754 | ||
755 | ndevctx->device_ctx = hdev; | |
756 | hv_set_drvdata(hdev, ndev); | |
757 | device_info.ring_size = ring_size; | |
758 | rndis_filter_device_add(hdev, &device_info); | |
5b54dac8 | 759 | netif_tx_wake_all_queues(ndev); |
4d447c9a HZ |
760 | |
761 | return 0; | |
762 | } | |
763 | ||
7eafd9b4 | 764 | static struct rtnl_link_stats64 *netvsc_get_stats64(struct net_device *net, |
765 | struct rtnl_link_stats64 *t) | |
766 | { | |
767 | struct net_device_context *ndev_ctx = netdev_priv(net); | |
768 | int cpu; | |
769 | ||
770 | for_each_possible_cpu(cpu) { | |
771 | struct netvsc_stats *tx_stats = per_cpu_ptr(ndev_ctx->tx_stats, | |
772 | cpu); | |
773 | struct netvsc_stats *rx_stats = per_cpu_ptr(ndev_ctx->rx_stats, | |
774 | cpu); | |
775 | u64 tx_packets, tx_bytes, rx_packets, rx_bytes; | |
776 | unsigned int start; | |
777 | ||
778 | do { | |
779 | start = u64_stats_fetch_begin_irq(&tx_stats->s_sync); | |
780 | tx_packets = tx_stats->packets; | |
781 | tx_bytes = tx_stats->bytes; | |
782 | } while (u64_stats_fetch_retry_irq(&tx_stats->s_sync, start)); | |
783 | ||
784 | do { | |
785 | start = u64_stats_fetch_begin_irq(&rx_stats->s_sync); | |
786 | rx_packets = rx_stats->packets; | |
787 | rx_bytes = rx_stats->bytes; | |
788 | } while (u64_stats_fetch_retry_irq(&rx_stats->s_sync, start)); | |
789 | ||
790 | t->tx_bytes += tx_bytes; | |
791 | t->tx_packets += tx_packets; | |
792 | t->rx_bytes += rx_bytes; | |
793 | t->rx_packets += rx_packets; | |
794 | } | |
795 | ||
796 | t->tx_dropped = net->stats.tx_dropped; | |
797 | t->tx_errors = net->stats.tx_dropped; | |
798 | ||
799 | t->rx_dropped = net->stats.rx_dropped; | |
800 | t->rx_errors = net->stats.rx_errors; | |
801 | ||
802 | return t; | |
803 | } | |
1ce09e89 HZ |
804 | |
805 | static int netvsc_set_mac_addr(struct net_device *ndev, void *p) | |
806 | { | |
807 | struct net_device_context *ndevctx = netdev_priv(ndev); | |
808 | struct hv_device *hdev = ndevctx->device_ctx; | |
809 | struct sockaddr *addr = p; | |
9a4c831e | 810 | char save_adr[ETH_ALEN]; |
1ce09e89 HZ |
811 | unsigned char save_aatype; |
812 | int err; | |
813 | ||
814 | memcpy(save_adr, ndev->dev_addr, ETH_ALEN); | |
815 | save_aatype = ndev->addr_assign_type; | |
816 | ||
817 | err = eth_mac_addr(ndev, p); | |
818 | if (err != 0) | |
819 | return err; | |
820 | ||
821 | err = rndis_filter_set_device_mac(hdev, addr->sa_data); | |
822 | if (err != 0) { | |
823 | /* roll back to saved MAC */ | |
824 | memcpy(ndev->dev_addr, save_adr, ETH_ALEN); | |
825 | ndev->addr_assign_type = save_aatype; | |
826 | } | |
827 | ||
828 | return err; | |
829 | } | |
830 | ||
316158fe RW |
831 | #ifdef CONFIG_NET_POLL_CONTROLLER |
832 | static void netvsc_poll_controller(struct net_device *net) | |
833 | { | |
834 | /* As netvsc_start_xmit() works synchronous we don't have to | |
835 | * trigger anything here. | |
836 | */ | |
837 | } | |
838 | #endif | |
1ce09e89 | 839 | |
f82f4ad7 SH |
840 | static const struct ethtool_ops ethtool_ops = { |
841 | .get_drvinfo = netvsc_get_drvinfo, | |
f82f4ad7 | 842 | .get_link = ethtool_op_get_link, |
59995370 | 843 | .get_channels = netvsc_get_channels, |
f82f4ad7 SH |
844 | }; |
845 | ||
df2fff28 GKH |
846 | static const struct net_device_ops device_ops = { |
847 | .ndo_open = netvsc_open, | |
848 | .ndo_stop = netvsc_close, | |
849 | .ndo_start_xmit = netvsc_start_xmit, | |
afc4b13d | 850 | .ndo_set_rx_mode = netvsc_set_multicast_list, |
4d447c9a | 851 | .ndo_change_mtu = netvsc_change_mtu, |
b681b588 | 852 | .ndo_validate_addr = eth_validate_addr, |
1ce09e89 | 853 | .ndo_set_mac_address = netvsc_set_mac_addr, |
5b54dac8 | 854 | .ndo_select_queue = netvsc_select_queue, |
7eafd9b4 | 855 | .ndo_get_stats64 = netvsc_get_stats64, |
316158fe RW |
856 | #ifdef CONFIG_NET_POLL_CONTROLLER |
857 | .ndo_poll_controller = netvsc_poll_controller, | |
858 | #endif | |
df2fff28 GKH |
859 | }; |
860 | ||
c996edcf HZ |
861 | /* |
862 | * Send GARP packet to network peers after migrations. | |
863 | * After Quick Migration, the network is not immediately operational in the | |
864 | * current context when receiving RNDIS_STATUS_MEDIA_CONNECT event. So, add | |
122a5f64 | 865 | * another netif_notify_peers() into a delayed work, otherwise GARP packet |
c996edcf | 866 | * will not be sent after quick migration, and cause network disconnection. |
891de74d | 867 | * Also, we update the carrier status here. |
c996edcf | 868 | */ |
891de74d | 869 | static void netvsc_link_change(struct work_struct *w) |
c996edcf HZ |
870 | { |
871 | struct net_device_context *ndev_ctx; | |
872 | struct net_device *net; | |
2ddd5e5f | 873 | struct netvsc_device *net_device; |
891de74d | 874 | struct rndis_device *rdev; |
3a494e71 HZ |
875 | bool notify, refresh = false; |
876 | char *argv[] = { "/etc/init.d/network", "restart", NULL }; | |
877 | char *envp[] = { "HOME=/", "PATH=/sbin:/usr/sbin:/bin:/usr/bin", NULL }; | |
891de74d HZ |
878 | |
879 | rtnl_lock(); | |
c996edcf | 880 | |
122a5f64 | 881 | ndev_ctx = container_of(w, struct net_device_context, dwork.work); |
2ddd5e5f | 882 | net_device = hv_get_drvdata(ndev_ctx->device_ctx); |
891de74d | 883 | rdev = net_device->extension; |
2ddd5e5f | 884 | net = net_device->ndev; |
891de74d HZ |
885 | |
886 | if (rdev->link_state) { | |
887 | netif_carrier_off(net); | |
888 | notify = false; | |
889 | } else { | |
890 | netif_carrier_on(net); | |
891 | notify = true; | |
3a494e71 HZ |
892 | if (rdev->link_change) { |
893 | rdev->link_change = false; | |
894 | refresh = true; | |
895 | } | |
891de74d HZ |
896 | } |
897 | ||
898 | rtnl_unlock(); | |
899 | ||
3a494e71 HZ |
900 | if (refresh) |
901 | call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC); | |
902 | ||
891de74d HZ |
903 | if (notify) |
904 | netdev_notify_peers(net); | |
c996edcf HZ |
905 | } |
906 | ||
7eafd9b4 | 907 | static void netvsc_free_netdev(struct net_device *netdev) |
908 | { | |
909 | struct net_device_context *net_device_ctx = netdev_priv(netdev); | |
910 | ||
911 | free_percpu(net_device_ctx->tx_stats); | |
912 | free_percpu(net_device_ctx->rx_stats); | |
913 | free_netdev(netdev); | |
914 | } | |
c996edcf | 915 | |
84946899 S |
916 | static int netvsc_probe(struct hv_device *dev, |
917 | const struct hv_vmbus_device_id *dev_id) | |
df2fff28 | 918 | { |
df2fff28 GKH |
919 | struct net_device *net = NULL; |
920 | struct net_device_context *net_device_ctx; | |
921 | struct netvsc_device_info device_info; | |
5b54dac8 | 922 | struct netvsc_device *nvdev; |
df2fff28 | 923 | int ret; |
b08cc791 | 924 | u32 max_needed_headroom; |
df2fff28 | 925 | |
5b54dac8 HZ |
926 | net = alloc_etherdev_mq(sizeof(struct net_device_context), |
927 | num_online_cpus()); | |
df2fff28 | 928 | if (!net) |
51a805d0 | 929 | return -ENOMEM; |
df2fff28 | 930 | |
b08cc791 | 931 | max_needed_headroom = sizeof(struct hv_netvsc_packet) + |
72151422 | 932 | RNDIS_AND_PPI_SIZE; |
b08cc791 | 933 | |
1b07da51 HZ |
934 | netif_carrier_off(net); |
935 | ||
df2fff28 | 936 | net_device_ctx = netdev_priv(net); |
9efd21e1 | 937 | net_device_ctx->device_ctx = dev; |
3f300ff4 SX |
938 | net_device_ctx->msg_enable = netif_msg_init(debug, default_msg); |
939 | if (netif_msg_probe(net_device_ctx)) | |
940 | netdev_dbg(net, "netvsc msg_enable: %d\n", | |
941 | net_device_ctx->msg_enable); | |
942 | ||
7eafd9b4 | 943 | net_device_ctx->tx_stats = netdev_alloc_pcpu_stats(struct netvsc_stats); |
944 | if (!net_device_ctx->tx_stats) { | |
945 | free_netdev(net); | |
946 | return -ENOMEM; | |
947 | } | |
948 | net_device_ctx->rx_stats = netdev_alloc_pcpu_stats(struct netvsc_stats); | |
949 | if (!net_device_ctx->rx_stats) { | |
950 | free_percpu(net_device_ctx->tx_stats); | |
951 | free_netdev(net); | |
952 | return -ENOMEM; | |
953 | } | |
954 | ||
2ddd5e5f | 955 | hv_set_drvdata(dev, net); |
891de74d | 956 | INIT_DELAYED_WORK(&net_device_ctx->dwork, netvsc_link_change); |
792df872 | 957 | INIT_WORK(&net_device_ctx->work, do_set_multicast); |
df2fff28 | 958 | |
df2fff28 GKH |
959 | net->netdev_ops = &device_ops; |
960 | ||
77bf5487 KS |
961 | net->hw_features = NETIF_F_RXCSUM | NETIF_F_SG | NETIF_F_IP_CSUM | |
962 | NETIF_F_TSO; | |
08cd04bf | 963 | net->features = NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_SG | NETIF_F_RXCSUM | |
77bf5487 | 964 | NETIF_F_IP_CSUM | NETIF_F_TSO; |
6048718d | 965 | |
7ad24ea4 | 966 | net->ethtool_ops = ðtool_ops; |
9efd21e1 | 967 | SET_NETDEV_DEV(net, &dev->device); |
df2fff28 | 968 | |
b08cc791 KS |
969 | /* |
970 | * Request additional head room in the skb. | |
971 | * We will use this space to build the rndis | |
972 | * heaser and other state we need to maintain. | |
973 | */ | |
974 | net->needed_headroom = max_needed_headroom; | |
975 | ||
692e084e HZ |
976 | /* Notify the netvsc driver of the new device */ |
977 | device_info.ring_size = ring_size; | |
978 | ret = rndis_filter_device_add(dev, &device_info); | |
979 | if (ret != 0) { | |
980 | netdev_err(net, "unable to add netvsc device (ret %d)\n", ret); | |
7eafd9b4 | 981 | netvsc_free_netdev(net); |
2ddd5e5f | 982 | hv_set_drvdata(dev, NULL); |
692e084e | 983 | return ret; |
df2fff28 | 984 | } |
692e084e HZ |
985 | memcpy(net->dev_addr, device_info.mac_adr, ETH_ALEN); |
986 | ||
5b54dac8 HZ |
987 | nvdev = hv_get_drvdata(dev); |
988 | netif_set_real_num_tx_queues(net, nvdev->num_chn); | |
989 | netif_set_real_num_rx_queues(net, nvdev->num_chn); | |
5b54dac8 | 990 | |
a68f9614 HZ |
991 | ret = register_netdev(net); |
992 | if (ret != 0) { | |
993 | pr_err("Unable to register netdev.\n"); | |
994 | rndis_filter_device_remove(dev); | |
7eafd9b4 | 995 | netvsc_free_netdev(net); |
1b07da51 HZ |
996 | } else { |
997 | schedule_delayed_work(&net_device_ctx->dwork, 0); | |
a68f9614 HZ |
998 | } |
999 | ||
df2fff28 GKH |
1000 | return ret; |
1001 | } | |
1002 | ||
415b023a | 1003 | static int netvsc_remove(struct hv_device *dev) |
df2fff28 | 1004 | { |
2ddd5e5f | 1005 | struct net_device *net; |
122a5f64 | 1006 | struct net_device_context *ndev_ctx; |
2ddd5e5f S |
1007 | struct netvsc_device *net_device; |
1008 | ||
1009 | net_device = hv_get_drvdata(dev); | |
1010 | net = net_device->ndev; | |
df2fff28 | 1011 | |
df2fff28 | 1012 | if (net == NULL) { |
415b023a | 1013 | dev_err(&dev->device, "No net device to remove\n"); |
df2fff28 GKH |
1014 | return 0; |
1015 | } | |
1016 | ||
4d447c9a HZ |
1017 | net_device->start_remove = true; |
1018 | ||
122a5f64 HZ |
1019 | ndev_ctx = netdev_priv(net); |
1020 | cancel_delayed_work_sync(&ndev_ctx->dwork); | |
792df872 | 1021 | cancel_work_sync(&ndev_ctx->work); |
122a5f64 | 1022 | |
df2fff28 | 1023 | /* Stop outbound asap */ |
0a282538 | 1024 | netif_tx_disable(net); |
df2fff28 GKH |
1025 | |
1026 | unregister_netdev(net); | |
1027 | ||
1028 | /* | |
1029 | * Call to the vsc driver to let it know that the device is being | |
1030 | * removed | |
1031 | */ | |
df06bcff | 1032 | rndis_filter_device_remove(dev); |
df2fff28 | 1033 | |
7eafd9b4 | 1034 | netvsc_free_netdev(net); |
df06bcff | 1035 | return 0; |
df2fff28 GKH |
1036 | } |
1037 | ||
345c4cc3 | 1038 | static const struct hv_vmbus_device_id id_table[] = { |
c45cf2d4 | 1039 | /* Network guid */ |
8f505944 | 1040 | { HV_NIC_GUID, }, |
c45cf2d4 | 1041 | { }, |
345c4cc3 S |
1042 | }; |
1043 | ||
1044 | MODULE_DEVICE_TABLE(vmbus, id_table); | |
1045 | ||
f1542a66 | 1046 | /* The one and only one */ |
fde0ef9b | 1047 | static struct hv_driver netvsc_drv = { |
d31b20fc | 1048 | .name = KBUILD_MODNAME, |
345c4cc3 | 1049 | .id_table = id_table, |
fde0ef9b S |
1050 | .probe = netvsc_probe, |
1051 | .remove = netvsc_remove, | |
d4890970 | 1052 | }; |
f1542a66 | 1053 | |
a9869c94 | 1054 | static void __exit netvsc_drv_exit(void) |
fceaf24a | 1055 | { |
768fa219 | 1056 | vmbus_driver_unregister(&netvsc_drv); |
fceaf24a HJ |
1057 | } |
1058 | ||
1fde28cf | 1059 | static int __init netvsc_drv_init(void) |
df2fff28 | 1060 | { |
fa85a6c2 HZ |
1061 | if (ring_size < RING_SIZE_MIN) { |
1062 | ring_size = RING_SIZE_MIN; | |
1063 | pr_info("Increased ring_size to %d (min allowed)\n", | |
1064 | ring_size); | |
1065 | } | |
768fa219 | 1066 | return vmbus_driver_register(&netvsc_drv); |
df2fff28 GKH |
1067 | } |
1068 | ||
26c14cc1 | 1069 | MODULE_LICENSE("GPL"); |
7880fc54 | 1070 | MODULE_DESCRIPTION("Microsoft Hyper-V network driver"); |
fceaf24a | 1071 | |
1fde28cf | 1072 | module_init(netvsc_drv_init); |
a9869c94 | 1073 | module_exit(netvsc_drv_exit); |