1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Ethernet-type device handling.
7 *
8 * Version: @(#)eth.c 1.0.7 05/25/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Mark Evans, <evansmp@uhura.aston.ac.uk>
13 * Florian La Roche, <rzsfl@rz.uni-sb.de>
14 * Alan Cox, <gw4pts@gw4pts.ampr.org>
15 *
16 * Fixes:
17 * Mr Linux : Arp problems
18 * Alan Cox : Generic queue tidyup (very tiny here)
19 * Alan Cox : eth_header ntohs should be htons
20 * Alan Cox : eth_rebuild_header missing an htons and
21 * minor other things.
22 * Tegge : Arp bug fixes.
23 * Florian : Removed many unnecessary functions, code cleanup
24 * and changes for new arp and skbuff.
25 * Alan Cox : Redid header building to reflect new format.
26 * Alan Cox : ARP only when compiled with CONFIG_INET
27 * Greg Page : 802.2 and SNAP stuff.
28 * Alan Cox : MAC layer pointers/new format.
29 * Paul Gortmaker : eth_copy_and_sum shouldn't csum padding.
30 * Alan Cox : Protect against forwarding explosions with
31 * older network drivers and IFF_ALLMULTI.
32 * Christer Weinigel : Better rebuild header message.
33 * Andrew Morton : 26Feb01: kill ether_setup() - use netdev_boot_setup().
34 *
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
39 */
40 #include <linux/module.h>
41 #include <linux/types.h>
42 #include <linux/kernel.h>
43 #include <linux/string.h>
44 #include <linux/mm.h>
45 #include <linux/socket.h>
46 #include <linux/in.h>
47 #include <linux/inet.h>
48 #include <linux/ip.h>
49 #include <linux/netdevice.h>
50 #include <linux/etherdevice.h>
51 #include <linux/skbuff.h>
52 #include <linux/errno.h>
53 #include <linux/init.h>
54 #include <linux/if_ether.h>
55 #include <linux/of_net.h>
56 #include <linux/pci.h>
57 #include <net/dst.h>
58 #include <net/arp.h>
59 #include <net/sock.h>
60 #include <net/ipv6.h>
61 #include <net/ip.h>
62 #include <net/dsa.h>
63 #include <net/flow_dissector.h>
64 #include <linux/uaccess.h>
65 #include <net/pkt_sched.h>
66
67 __setup("ether=", netdev_boot_setup);
68
69 /**
70 * eth_header - create the Ethernet header
71 * @skb: buffer to alter
72 * @dev: source device
73 * @type: Ethernet type field
74 * @daddr: destination address (NULL leave destination address)
75 * @saddr: source address (NULL use device source address)
76 * @len: packet length (<= skb->len)
77 *
78 *
79 * Set the protocol type. For a packet of type ETH_P_802_3/2 we put the length
80 * in here instead.
81 */
eth_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)82 int eth_header(struct sk_buff *skb, struct net_device *dev,
83 unsigned short type,
84 const void *daddr, const void *saddr, unsigned int len)
85 {
86 struct ethhdr *eth = skb_push(skb, ETH_HLEN);
87
88 if (type != ETH_P_802_3 && type != ETH_P_802_2)
89 eth->h_proto = htons(type);
90 else
91 eth->h_proto = htons(len);
92
93 /*
94 * Set the source hardware address.
95 */
96
97 if (!saddr)
98 saddr = dev->dev_addr;
99 memcpy(eth->h_source, saddr, ETH_ALEN);
100
101 if (daddr) {
102 memcpy(eth->h_dest, daddr, ETH_ALEN);
103 return ETH_HLEN;
104 }
105
106 /*
107 * Anyway, the loopback-device should never use this function...
108 */
109
110 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP)) {
111 eth_zero_addr(eth->h_dest);
112 return ETH_HLEN;
113 }
114
115 return -ETH_HLEN;
116 }
117 EXPORT_SYMBOL(eth_header);
118
119 /**
120 * eth_get_headlen - determine the length of header for an ethernet frame
121 * @data: pointer to start of frame
122 * @len: total length of frame
123 *
124 * Make a best effort attempt to pull the length for all of the headers for
125 * a given frame in a linear buffer.
126 */
eth_get_headlen(void * data,unsigned int len)127 u32 eth_get_headlen(void *data, unsigned int len)
128 {
129 const unsigned int flags = FLOW_DISSECTOR_F_PARSE_1ST_FRAG;
130 const struct ethhdr *eth = (const struct ethhdr *)data;
131 struct flow_keys_basic keys;
132
133 /* this should never happen, but better safe than sorry */
134 if (unlikely(len < sizeof(*eth)))
135 return len;
136
137 /* parse any remaining L2/L3 headers, check for L4 */
138 if (!skb_flow_dissect_flow_keys_basic(NULL, &keys, data, eth->h_proto,
139 sizeof(*eth), len, flags))
140 return max_t(u32, keys.control.thoff, sizeof(*eth));
141
142 /* parse for any L4 headers */
143 return min_t(u32, __skb_get_poff(NULL, data, &keys, len), len);
144 }
145 EXPORT_SYMBOL(eth_get_headlen);
146
147 /**
148 * eth_type_trans - determine the packet's protocol ID.
149 * @skb: received socket data
150 * @dev: receiving network device
151 *
152 * The rule here is that we
153 * assume 802.3 if the type field is short enough to be a length.
154 * This is normal practice and works for any 'now in use' protocol.
155 */
eth_type_trans(struct sk_buff * skb,struct net_device * dev)156 __be16 eth_type_trans(struct sk_buff *skb, struct net_device *dev)
157 {
158 unsigned short _service_access_point;
159 const unsigned short *sap;
160 const struct ethhdr *eth;
161
162 skb->dev = dev;
163 skb_reset_mac_header(skb);
164
165 eth = (struct ethhdr *)skb->data;
166 skb_pull_inline(skb, ETH_HLEN);
167
168 if (unlikely(is_multicast_ether_addr_64bits(eth->h_dest))) {
169 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
170 skb->pkt_type = PACKET_BROADCAST;
171 else
172 skb->pkt_type = PACKET_MULTICAST;
173 }
174 else if (unlikely(!ether_addr_equal_64bits(eth->h_dest,
175 dev->dev_addr)))
176 skb->pkt_type = PACKET_OTHERHOST;
177
178 /*
179 * Some variants of DSA tagging don't have an ethertype field
180 * at all, so we check here whether one of those tagging
181 * variants has been configured on the receiving interface,
182 * and if so, set skb->protocol without looking at the packet.
183 */
184 if (unlikely(netdev_uses_dsa(dev)))
185 return htons(ETH_P_XDSA);
186
187 if (likely(eth_proto_is_802_3(eth->h_proto)))
188 return eth->h_proto;
189
190 /*
191 * This is a magic hack to spot IPX packets. Older Novell breaks
192 * the protocol design and runs IPX over 802.3 without an 802.2 LLC
193 * layer. We look for FFFF which isn't a used 802.2 SSAP/DSAP. This
194 * won't work for fault tolerant netware but does for the rest.
195 */
196 sap = skb_header_pointer(skb, 0, sizeof(*sap), &_service_access_point);
197 if (sap && *sap == 0xFFFF)
198 return htons(ETH_P_802_3);
199
200 /*
201 * Real 802.2 LLC
202 */
203 return htons(ETH_P_802_2);
204 }
205 EXPORT_SYMBOL(eth_type_trans);
206
207 /**
208 * eth_header_parse - extract hardware address from packet
209 * @skb: packet to extract header from
210 * @haddr: destination buffer
211 */
eth_header_parse(const struct sk_buff * skb,unsigned char * haddr)212 int eth_header_parse(const struct sk_buff *skb, unsigned char *haddr)
213 {
214 const struct ethhdr *eth = eth_hdr(skb);
215 memcpy(haddr, eth->h_source, ETH_ALEN);
216 return ETH_ALEN;
217 }
218 EXPORT_SYMBOL(eth_header_parse);
219
220 /**
221 * eth_header_cache - fill cache entry from neighbour
222 * @neigh: source neighbour
223 * @hh: destination cache entry
224 * @type: Ethernet type field
225 *
226 * Create an Ethernet header template from the neighbour.
227 */
eth_header_cache(const struct neighbour * neigh,struct hh_cache * hh,__be16 type)228 int eth_header_cache(const struct neighbour *neigh, struct hh_cache *hh, __be16 type)
229 {
230 struct ethhdr *eth;
231 const struct net_device *dev = neigh->dev;
232
233 eth = (struct ethhdr *)
234 (((u8 *) hh->hh_data) + (HH_DATA_OFF(sizeof(*eth))));
235
236 if (type == htons(ETH_P_802_3))
237 return -1;
238
239 eth->h_proto = type;
240 memcpy(eth->h_source, dev->dev_addr, ETH_ALEN);
241 memcpy(eth->h_dest, neigh->ha, ETH_ALEN);
242 hh->hh_len = ETH_HLEN;
243 return 0;
244 }
245 EXPORT_SYMBOL(eth_header_cache);
246
247 /**
248 * eth_header_cache_update - update cache entry
249 * @hh: destination cache entry
250 * @dev: network device
251 * @haddr: new hardware address
252 *
253 * Called by Address Resolution module to notify changes in address.
254 */
eth_header_cache_update(struct hh_cache * hh,const struct net_device * dev,const unsigned char * haddr)255 void eth_header_cache_update(struct hh_cache *hh,
256 const struct net_device *dev,
257 const unsigned char *haddr)
258 {
259 memcpy(((u8 *) hh->hh_data) + HH_DATA_OFF(sizeof(struct ethhdr)),
260 haddr, ETH_ALEN);
261 }
262 EXPORT_SYMBOL(eth_header_cache_update);
263
264 /**
265 * eth_prepare_mac_addr_change - prepare for mac change
266 * @dev: network device
267 * @p: socket address
268 */
eth_prepare_mac_addr_change(struct net_device * dev,void * p)269 int eth_prepare_mac_addr_change(struct net_device *dev, void *p)
270 {
271 struct sockaddr *addr = p;
272
273 if (!(dev->priv_flags & IFF_LIVE_ADDR_CHANGE) && netif_running(dev))
274 return -EBUSY;
275 if (!is_valid_ether_addr(addr->sa_data))
276 return -EADDRNOTAVAIL;
277 return 0;
278 }
279 EXPORT_SYMBOL(eth_prepare_mac_addr_change);
280
281 /**
282 * eth_commit_mac_addr_change - commit mac change
283 * @dev: network device
284 * @p: socket address
285 */
eth_commit_mac_addr_change(struct net_device * dev,void * p)286 void eth_commit_mac_addr_change(struct net_device *dev, void *p)
287 {
288 struct sockaddr *addr = p;
289
290 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
291 }
292 EXPORT_SYMBOL(eth_commit_mac_addr_change);
293
294 /**
295 * eth_mac_addr - set new Ethernet hardware address
296 * @dev: network device
297 * @p: socket address
298 *
299 * Change hardware address of device.
300 *
301 * This doesn't change hardware matching, so needs to be overridden
302 * for most real devices.
303 */
eth_mac_addr(struct net_device * dev,void * p)304 int eth_mac_addr(struct net_device *dev, void *p)
305 {
306 int ret;
307
308 ret = eth_prepare_mac_addr_change(dev, p);
309 if (ret < 0)
310 return ret;
311 eth_commit_mac_addr_change(dev, p);
312 return 0;
313 }
314 EXPORT_SYMBOL(eth_mac_addr);
315
316 /**
317 * eth_change_mtu - set new MTU size
318 * @dev: network device
319 * @new_mtu: new Maximum Transfer Unit
320 *
321 * Allow changing MTU size. Needs to be overridden for devices
322 * supporting jumbo frames.
323 */
eth_change_mtu(struct net_device * dev,int new_mtu)324 int eth_change_mtu(struct net_device *dev, int new_mtu)
325 {
326 netdev_warn(dev, "%s is deprecated\n", __func__);
327 dev->mtu = new_mtu;
328 return 0;
329 }
330 EXPORT_SYMBOL(eth_change_mtu);
331
eth_validate_addr(struct net_device * dev)332 int eth_validate_addr(struct net_device *dev)
333 {
334 if (!is_valid_ether_addr(dev->dev_addr))
335 return -EADDRNOTAVAIL;
336
337 return 0;
338 }
339 EXPORT_SYMBOL(eth_validate_addr);
340
341 const struct header_ops eth_header_ops ____cacheline_aligned = {
342 .create = eth_header,
343 .parse = eth_header_parse,
344 .cache = eth_header_cache,
345 .cache_update = eth_header_cache_update,
346 };
347
348 /**
349 * ether_setup - setup Ethernet network device
350 * @dev: network device
351 *
352 * Fill in the fields of the device structure with Ethernet-generic values.
353 */
ether_setup(struct net_device * dev)354 void ether_setup(struct net_device *dev)
355 {
356 dev->header_ops = ð_header_ops;
357 dev->type = ARPHRD_ETHER;
358 dev->hard_header_len = ETH_HLEN;
359 dev->min_header_len = ETH_HLEN;
360 dev->mtu = ETH_DATA_LEN;
361 dev->min_mtu = ETH_MIN_MTU;
362 dev->max_mtu = ETH_DATA_LEN;
363 dev->addr_len = ETH_ALEN;
364 dev->tx_queue_len = DEFAULT_TX_QUEUE_LEN;
365 dev->flags = IFF_BROADCAST|IFF_MULTICAST;
366 dev->priv_flags |= IFF_TX_SKB_SHARING;
367
368 eth_broadcast_addr(dev->broadcast);
369
370 }
371 EXPORT_SYMBOL(ether_setup);
372
373 /**
374 * alloc_etherdev_mqs - Allocates and sets up an Ethernet device
375 * @sizeof_priv: Size of additional driver-private structure to be allocated
376 * for this Ethernet device
377 * @txqs: The number of TX queues this device has.
378 * @rxqs: The number of RX queues this device has.
379 *
380 * Fill in the fields of the device structure with Ethernet-generic
381 * values. Basically does everything except registering the device.
382 *
383 * Constructs a new net device, complete with a private data area of
384 * size (sizeof_priv). A 32-byte (not bit) alignment is enforced for
385 * this private data area.
386 */
387
alloc_etherdev_mqs(int sizeof_priv,unsigned int txqs,unsigned int rxqs)388 struct net_device *alloc_etherdev_mqs(int sizeof_priv, unsigned int txqs,
389 unsigned int rxqs)
390 {
391 return alloc_netdev_mqs(sizeof_priv, "eth%d", NET_NAME_UNKNOWN,
392 ether_setup, txqs, rxqs);
393 }
394 EXPORT_SYMBOL(alloc_etherdev_mqs);
395
devm_free_netdev(struct device * dev,void * res)396 static void devm_free_netdev(struct device *dev, void *res)
397 {
398 free_netdev(*(struct net_device **)res);
399 }
400
devm_alloc_etherdev_mqs(struct device * dev,int sizeof_priv,unsigned int txqs,unsigned int rxqs)401 struct net_device *devm_alloc_etherdev_mqs(struct device *dev, int sizeof_priv,
402 unsigned int txqs, unsigned int rxqs)
403 {
404 struct net_device **dr;
405 struct net_device *netdev;
406
407 dr = devres_alloc(devm_free_netdev, sizeof(*dr), GFP_KERNEL);
408 if (!dr)
409 return NULL;
410
411 netdev = alloc_etherdev_mqs(sizeof_priv, txqs, rxqs);
412 if (!netdev) {
413 devres_free(dr);
414 return NULL;
415 }
416
417 *dr = netdev;
418 devres_add(dev, dr);
419
420 return netdev;
421 }
422 EXPORT_SYMBOL(devm_alloc_etherdev_mqs);
423
sysfs_format_mac(char * buf,const unsigned char * addr,int len)424 ssize_t sysfs_format_mac(char *buf, const unsigned char *addr, int len)
425 {
426 return scnprintf(buf, PAGE_SIZE, "%*phC\n", len, addr);
427 }
428 EXPORT_SYMBOL(sysfs_format_mac);
429
eth_gro_receive(struct list_head * head,struct sk_buff * skb)430 struct sk_buff *eth_gro_receive(struct list_head *head, struct sk_buff *skb)
431 {
432 const struct packet_offload *ptype;
433 unsigned int hlen, off_eth;
434 struct sk_buff *pp = NULL;
435 struct ethhdr *eh, *eh2;
436 struct sk_buff *p;
437 __be16 type;
438 int flush = 1;
439
440 off_eth = skb_gro_offset(skb);
441 hlen = off_eth + sizeof(*eh);
442 eh = skb_gro_header_fast(skb, off_eth);
443 if (skb_gro_header_hard(skb, hlen)) {
444 eh = skb_gro_header_slow(skb, hlen, off_eth);
445 if (unlikely(!eh))
446 goto out;
447 }
448
449 flush = 0;
450
451 list_for_each_entry(p, head, list) {
452 if (!NAPI_GRO_CB(p)->same_flow)
453 continue;
454
455 eh2 = (struct ethhdr *)(p->data + off_eth);
456 if (compare_ether_header(eh, eh2)) {
457 NAPI_GRO_CB(p)->same_flow = 0;
458 continue;
459 }
460 }
461
462 type = eh->h_proto;
463
464 rcu_read_lock();
465 ptype = gro_find_receive_by_type(type);
466 if (ptype == NULL) {
467 flush = 1;
468 goto out_unlock;
469 }
470
471 skb_gro_pull(skb, sizeof(*eh));
472 skb_gro_postpull_rcsum(skb, eh, sizeof(*eh));
473 pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb);
474
475 out_unlock:
476 rcu_read_unlock();
477 out:
478 skb_gro_flush_final(skb, pp, flush);
479
480 return pp;
481 }
482 EXPORT_SYMBOL(eth_gro_receive);
483
eth_gro_complete(struct sk_buff * skb,int nhoff)484 int eth_gro_complete(struct sk_buff *skb, int nhoff)
485 {
486 struct ethhdr *eh = (struct ethhdr *)(skb->data + nhoff);
487 __be16 type = eh->h_proto;
488 struct packet_offload *ptype;
489 int err = -ENOSYS;
490
491 if (skb->encapsulation)
492 skb_set_inner_mac_header(skb, nhoff);
493
494 rcu_read_lock();
495 ptype = gro_find_complete_by_type(type);
496 if (ptype != NULL)
497 err = ptype->callbacks.gro_complete(skb, nhoff +
498 sizeof(struct ethhdr));
499
500 rcu_read_unlock();
501 return err;
502 }
503 EXPORT_SYMBOL(eth_gro_complete);
504
505 static struct packet_offload eth_packet_offload __read_mostly = {
506 .type = cpu_to_be16(ETH_P_TEB),
507 .priority = 10,
508 .callbacks = {
509 .gro_receive = eth_gro_receive,
510 .gro_complete = eth_gro_complete,
511 },
512 };
513
eth_offload_init(void)514 static int __init eth_offload_init(void)
515 {
516 dev_add_offload(ð_packet_offload);
517
518 return 0;
519 }
520
521 fs_initcall(eth_offload_init);
522
arch_get_platform_mac_address(void)523 unsigned char * __weak arch_get_platform_mac_address(void)
524 {
525 return NULL;
526 }
527
eth_platform_get_mac_address(struct device * dev,u8 * mac_addr)528 int eth_platform_get_mac_address(struct device *dev, u8 *mac_addr)
529 {
530 const unsigned char *addr;
531 struct device_node *dp;
532
533 if (dev_is_pci(dev))
534 dp = pci_device_to_OF_node(to_pci_dev(dev));
535 else
536 dp = dev->of_node;
537
538 addr = NULL;
539 if (dp)
540 addr = of_get_mac_address(dp);
541 if (!addr)
542 addr = arch_get_platform_mac_address();
543
544 if (!addr)
545 return -ENODEV;
546
547 ether_addr_copy(mac_addr, addr);
548 return 0;
549 }
550 EXPORT_SYMBOL(eth_platform_get_mac_address);
551