]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - net/ethernet/eth.c
Merge branch 'next' into for-linus
[mirror_ubuntu-jammy-kernel.git] / net / ethernet / eth.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Ethernet-type device handling.
8 *
9 * Version: @(#)eth.c 1.0.7 05/25/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Mark Evans, <evansmp@uhura.aston.ac.uk>
14 * Florian La Roche, <rzsfl@rz.uni-sb.de>
15 * Alan Cox, <gw4pts@gw4pts.ampr.org>
16 *
17 * Fixes:
18 * Mr Linux : Arp problems
19 * Alan Cox : Generic queue tidyup (very tiny here)
20 * Alan Cox : eth_header ntohs should be htons
21 * Alan Cox : eth_rebuild_header missing an htons and
22 * minor other things.
23 * Tegge : Arp bug fixes.
24 * Florian : Removed many unnecessary functions, code cleanup
25 * and changes for new arp and skbuff.
26 * Alan Cox : Redid header building to reflect new format.
27 * Alan Cox : ARP only when compiled with CONFIG_INET
28 * Greg Page : 802.2 and SNAP stuff.
29 * Alan Cox : MAC layer pointers/new format.
30 * Paul Gortmaker : eth_copy_and_sum shouldn't csum padding.
31 * Alan Cox : Protect against forwarding explosions with
32 * older network drivers and IFF_ALLMULTI.
33 * Christer Weinigel : Better rebuild header message.
34 * Andrew Morton : 26Feb01: kill ether_setup() - use netdev_boot_setup().
35 */
36 #include <linux/module.h>
37 #include <linux/types.h>
38 #include <linux/kernel.h>
39 #include <linux/string.h>
40 #include <linux/mm.h>
41 #include <linux/socket.h>
42 #include <linux/in.h>
43 #include <linux/inet.h>
44 #include <linux/ip.h>
45 #include <linux/netdevice.h>
46 #include <linux/nvmem-consumer.h>
47 #include <linux/etherdevice.h>
48 #include <linux/skbuff.h>
49 #include <linux/errno.h>
50 #include <linux/init.h>
51 #include <linux/if_ether.h>
52 #include <linux/of_net.h>
53 #include <linux/pci.h>
54 #include <net/dst.h>
55 #include <net/arp.h>
56 #include <net/sock.h>
57 #include <net/ipv6.h>
58 #include <net/ip.h>
59 #include <net/dsa.h>
60 #include <net/flow_dissector.h>
61 #include <net/gro.h>
62 #include <linux/uaccess.h>
63 #include <net/pkt_sched.h>
64
65 __setup("ether=", netdev_boot_setup);
66
67 /**
68 * eth_header - create the Ethernet header
69 * @skb: buffer to alter
70 * @dev: source device
71 * @type: Ethernet type field
72 * @daddr: destination address (NULL leave destination address)
73 * @saddr: source address (NULL use device source address)
74 * @len: packet length (<= skb->len)
75 *
76 *
77 * Set the protocol type. For a packet of type ETH_P_802_3/2 we put the length
78 * in here instead.
79 */
80 int eth_header(struct sk_buff *skb, struct net_device *dev,
81 unsigned short type,
82 const void *daddr, const void *saddr, unsigned int len)
83 {
84 struct ethhdr *eth = skb_push(skb, ETH_HLEN);
85
86 if (type != ETH_P_802_3 && type != ETH_P_802_2)
87 eth->h_proto = htons(type);
88 else
89 eth->h_proto = htons(len);
90
91 /*
92 * Set the source hardware address.
93 */
94
95 if (!saddr)
96 saddr = dev->dev_addr;
97 memcpy(eth->h_source, saddr, ETH_ALEN);
98
99 if (daddr) {
100 memcpy(eth->h_dest, daddr, ETH_ALEN);
101 return ETH_HLEN;
102 }
103
104 /*
105 * Anyway, the loopback-device should never use this function...
106 */
107
108 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP)) {
109 eth_zero_addr(eth->h_dest);
110 return ETH_HLEN;
111 }
112
113 return -ETH_HLEN;
114 }
115 EXPORT_SYMBOL(eth_header);
116
117 /**
118 * eth_get_headlen - determine the length of header for an ethernet frame
119 * @dev: pointer to network device
120 * @data: pointer to start of frame
121 * @len: total length of frame
122 *
123 * Make a best effort attempt to pull the length for all of the headers for
124 * a given frame in a linear buffer.
125 */
126 u32 eth_get_headlen(const struct net_device *dev, const void *data, u32 len)
127 {
128 const unsigned int flags = FLOW_DISSECTOR_F_PARSE_1ST_FRAG;
129 const struct ethhdr *eth = (const struct ethhdr *)data;
130 struct flow_keys_basic keys;
131
132 /* this should never happen, but better safe than sorry */
133 if (unlikely(len < sizeof(*eth)))
134 return len;
135
136 /* parse any remaining L2/L3 headers, check for L4 */
137 if (!skb_flow_dissect_flow_keys_basic(dev_net(dev), NULL, &keys, data,
138 eth->h_proto, sizeof(*eth),
139 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 */
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(!ether_addr_equal_64bits(eth->h_dest,
169 dev->dev_addr))) {
170 if (unlikely(is_multicast_ether_addr_64bits(eth->h_dest))) {
171 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
172 skb->pkt_type = PACKET_BROADCAST;
173 else
174 skb->pkt_type = PACKET_MULTICAST;
175 } else {
176 skb->pkt_type = PACKET_OTHERHOST;
177 }
178 }
179
180 /*
181 * Some variants of DSA tagging don't have an ethertype field
182 * at all, so we check here whether one of those tagging
183 * variants has been configured on the receiving interface,
184 * and if so, set skb->protocol without looking at the packet.
185 * The DSA tagging protocol may be able to decode some but not all
186 * traffic (for example only for management). In that case give it the
187 * option to filter the packets from which it can decode source port
188 * information.
189 */
190 if (unlikely(netdev_uses_dsa(dev)) && dsa_can_decode(skb, dev))
191 return htons(ETH_P_XDSA);
192
193 if (likely(eth_proto_is_802_3(eth->h_proto)))
194 return eth->h_proto;
195
196 /*
197 * This is a magic hack to spot IPX packets. Older Novell breaks
198 * the protocol design and runs IPX over 802.3 without an 802.2 LLC
199 * layer. We look for FFFF which isn't a used 802.2 SSAP/DSAP. This
200 * won't work for fault tolerant netware but does for the rest.
201 */
202 sap = skb_header_pointer(skb, 0, sizeof(*sap), &_service_access_point);
203 if (sap && *sap == 0xFFFF)
204 return htons(ETH_P_802_3);
205
206 /*
207 * Real 802.2 LLC
208 */
209 return htons(ETH_P_802_2);
210 }
211 EXPORT_SYMBOL(eth_type_trans);
212
213 /**
214 * eth_header_parse - extract hardware address from packet
215 * @skb: packet to extract header from
216 * @haddr: destination buffer
217 */
218 int eth_header_parse(const struct sk_buff *skb, unsigned char *haddr)
219 {
220 const struct ethhdr *eth = eth_hdr(skb);
221 memcpy(haddr, eth->h_source, ETH_ALEN);
222 return ETH_ALEN;
223 }
224 EXPORT_SYMBOL(eth_header_parse);
225
226 /**
227 * eth_header_cache - fill cache entry from neighbour
228 * @neigh: source neighbour
229 * @hh: destination cache entry
230 * @type: Ethernet type field
231 *
232 * Create an Ethernet header template from the neighbour.
233 */
234 int eth_header_cache(const struct neighbour *neigh, struct hh_cache *hh, __be16 type)
235 {
236 struct ethhdr *eth;
237 const struct net_device *dev = neigh->dev;
238
239 eth = (struct ethhdr *)
240 (((u8 *) hh->hh_data) + (HH_DATA_OFF(sizeof(*eth))));
241
242 if (type == htons(ETH_P_802_3))
243 return -1;
244
245 eth->h_proto = type;
246 memcpy(eth->h_source, dev->dev_addr, ETH_ALEN);
247 memcpy(eth->h_dest, neigh->ha, ETH_ALEN);
248
249 /* Pairs with READ_ONCE() in neigh_resolve_output(),
250 * neigh_hh_output() and neigh_update_hhs().
251 */
252 smp_store_release(&hh->hh_len, ETH_HLEN);
253
254 return 0;
255 }
256 EXPORT_SYMBOL(eth_header_cache);
257
258 /**
259 * eth_header_cache_update - update cache entry
260 * @hh: destination cache entry
261 * @dev: network device
262 * @haddr: new hardware address
263 *
264 * Called by Address Resolution module to notify changes in address.
265 */
266 void eth_header_cache_update(struct hh_cache *hh,
267 const struct net_device *dev,
268 const unsigned char *haddr)
269 {
270 memcpy(((u8 *) hh->hh_data) + HH_DATA_OFF(sizeof(struct ethhdr)),
271 haddr, ETH_ALEN);
272 }
273 EXPORT_SYMBOL(eth_header_cache_update);
274
275 /**
276 * eth_header_parse_protocol - extract protocol from L2 header
277 * @skb: packet to extract protocol from
278 */
279 __be16 eth_header_parse_protocol(const struct sk_buff *skb)
280 {
281 const struct ethhdr *eth = eth_hdr(skb);
282
283 return eth->h_proto;
284 }
285 EXPORT_SYMBOL(eth_header_parse_protocol);
286
287 /**
288 * eth_prepare_mac_addr_change - prepare for mac change
289 * @dev: network device
290 * @p: socket address
291 */
292 int eth_prepare_mac_addr_change(struct net_device *dev, void *p)
293 {
294 struct sockaddr *addr = p;
295
296 if (!(dev->priv_flags & IFF_LIVE_ADDR_CHANGE) && netif_running(dev))
297 return -EBUSY;
298 if (!is_valid_ether_addr(addr->sa_data))
299 return -EADDRNOTAVAIL;
300 return 0;
301 }
302 EXPORT_SYMBOL(eth_prepare_mac_addr_change);
303
304 /**
305 * eth_commit_mac_addr_change - commit mac change
306 * @dev: network device
307 * @p: socket address
308 */
309 void eth_commit_mac_addr_change(struct net_device *dev, void *p)
310 {
311 struct sockaddr *addr = p;
312
313 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
314 }
315 EXPORT_SYMBOL(eth_commit_mac_addr_change);
316
317 /**
318 * eth_mac_addr - set new Ethernet hardware address
319 * @dev: network device
320 * @p: socket address
321 *
322 * Change hardware address of device.
323 *
324 * This doesn't change hardware matching, so needs to be overridden
325 * for most real devices.
326 */
327 int eth_mac_addr(struct net_device *dev, void *p)
328 {
329 int ret;
330
331 ret = eth_prepare_mac_addr_change(dev, p);
332 if (ret < 0)
333 return ret;
334 eth_commit_mac_addr_change(dev, p);
335 return 0;
336 }
337 EXPORT_SYMBOL(eth_mac_addr);
338
339 int eth_validate_addr(struct net_device *dev)
340 {
341 if (!is_valid_ether_addr(dev->dev_addr))
342 return -EADDRNOTAVAIL;
343
344 return 0;
345 }
346 EXPORT_SYMBOL(eth_validate_addr);
347
348 const struct header_ops eth_header_ops ____cacheline_aligned = {
349 .create = eth_header,
350 .parse = eth_header_parse,
351 .cache = eth_header_cache,
352 .cache_update = eth_header_cache_update,
353 .parse_protocol = eth_header_parse_protocol,
354 };
355
356 /**
357 * ether_setup - setup Ethernet network device
358 * @dev: network device
359 *
360 * Fill in the fields of the device structure with Ethernet-generic values.
361 */
362 void ether_setup(struct net_device *dev)
363 {
364 dev->header_ops = &eth_header_ops;
365 dev->type = ARPHRD_ETHER;
366 dev->hard_header_len = ETH_HLEN;
367 dev->min_header_len = ETH_HLEN;
368 dev->mtu = ETH_DATA_LEN;
369 dev->min_mtu = ETH_MIN_MTU;
370 dev->max_mtu = ETH_DATA_LEN;
371 dev->addr_len = ETH_ALEN;
372 dev->tx_queue_len = DEFAULT_TX_QUEUE_LEN;
373 dev->flags = IFF_BROADCAST|IFF_MULTICAST;
374 dev->priv_flags |= IFF_TX_SKB_SHARING;
375
376 eth_broadcast_addr(dev->broadcast);
377
378 }
379 EXPORT_SYMBOL(ether_setup);
380
381 /**
382 * alloc_etherdev_mqs - Allocates and sets up an Ethernet device
383 * @sizeof_priv: Size of additional driver-private structure to be allocated
384 * for this Ethernet device
385 * @txqs: The number of TX queues this device has.
386 * @rxqs: The number of RX queues this device has.
387 *
388 * Fill in the fields of the device structure with Ethernet-generic
389 * values. Basically does everything except registering the device.
390 *
391 * Constructs a new net device, complete with a private data area of
392 * size (sizeof_priv). A 32-byte (not bit) alignment is enforced for
393 * this private data area.
394 */
395
396 struct net_device *alloc_etherdev_mqs(int sizeof_priv, unsigned int txqs,
397 unsigned int rxqs)
398 {
399 return alloc_netdev_mqs(sizeof_priv, "eth%d", NET_NAME_UNKNOWN,
400 ether_setup, txqs, rxqs);
401 }
402 EXPORT_SYMBOL(alloc_etherdev_mqs);
403
404 ssize_t sysfs_format_mac(char *buf, const unsigned char *addr, int len)
405 {
406 return scnprintf(buf, PAGE_SIZE, "%*phC\n", len, addr);
407 }
408 EXPORT_SYMBOL(sysfs_format_mac);
409
410 struct sk_buff *eth_gro_receive(struct list_head *head, struct sk_buff *skb)
411 {
412 const struct packet_offload *ptype;
413 unsigned int hlen, off_eth;
414 struct sk_buff *pp = NULL;
415 struct ethhdr *eh, *eh2;
416 struct sk_buff *p;
417 __be16 type;
418 int flush = 1;
419
420 off_eth = skb_gro_offset(skb);
421 hlen = off_eth + sizeof(*eh);
422 eh = skb_gro_header_fast(skb, off_eth);
423 if (skb_gro_header_hard(skb, hlen)) {
424 eh = skb_gro_header_slow(skb, hlen, off_eth);
425 if (unlikely(!eh))
426 goto out;
427 }
428
429 flush = 0;
430
431 list_for_each_entry(p, head, list) {
432 if (!NAPI_GRO_CB(p)->same_flow)
433 continue;
434
435 eh2 = (struct ethhdr *)(p->data + off_eth);
436 if (compare_ether_header(eh, eh2)) {
437 NAPI_GRO_CB(p)->same_flow = 0;
438 continue;
439 }
440 }
441
442 type = eh->h_proto;
443
444 rcu_read_lock();
445 ptype = gro_find_receive_by_type(type);
446 if (ptype == NULL) {
447 flush = 1;
448 goto out_unlock;
449 }
450
451 skb_gro_pull(skb, sizeof(*eh));
452 skb_gro_postpull_rcsum(skb, eh, sizeof(*eh));
453
454 pp = indirect_call_gro_receive_inet(ptype->callbacks.gro_receive,
455 ipv6_gro_receive, inet_gro_receive,
456 head, skb);
457
458 out_unlock:
459 rcu_read_unlock();
460 out:
461 skb_gro_flush_final(skb, pp, flush);
462
463 return pp;
464 }
465 EXPORT_SYMBOL(eth_gro_receive);
466
467 int eth_gro_complete(struct sk_buff *skb, int nhoff)
468 {
469 struct ethhdr *eh = (struct ethhdr *)(skb->data + nhoff);
470 __be16 type = eh->h_proto;
471 struct packet_offload *ptype;
472 int err = -ENOSYS;
473
474 if (skb->encapsulation)
475 skb_set_inner_mac_header(skb, nhoff);
476
477 rcu_read_lock();
478 ptype = gro_find_complete_by_type(type);
479 if (ptype != NULL)
480 err = INDIRECT_CALL_INET(ptype->callbacks.gro_complete,
481 ipv6_gro_complete, inet_gro_complete,
482 skb, nhoff + sizeof(*eh));
483
484 rcu_read_unlock();
485 return err;
486 }
487 EXPORT_SYMBOL(eth_gro_complete);
488
489 static struct packet_offload eth_packet_offload __read_mostly = {
490 .type = cpu_to_be16(ETH_P_TEB),
491 .priority = 10,
492 .callbacks = {
493 .gro_receive = eth_gro_receive,
494 .gro_complete = eth_gro_complete,
495 },
496 };
497
498 static int __init eth_offload_init(void)
499 {
500 dev_add_offload(&eth_packet_offload);
501
502 return 0;
503 }
504
505 fs_initcall(eth_offload_init);
506
507 unsigned char * __weak arch_get_platform_mac_address(void)
508 {
509 return NULL;
510 }
511
512 int eth_platform_get_mac_address(struct device *dev, u8 *mac_addr)
513 {
514 unsigned char *addr;
515 int ret;
516
517 ret = of_get_mac_address(dev->of_node, mac_addr);
518 if (!ret)
519 return 0;
520
521 addr = arch_get_platform_mac_address();
522 if (!addr)
523 return -ENODEV;
524
525 ether_addr_copy(mac_addr, addr);
526
527 return 0;
528 }
529 EXPORT_SYMBOL(eth_platform_get_mac_address);
530
531 /**
532 * nvmem_get_mac_address - Obtain the MAC address from an nvmem cell named
533 * 'mac-address' associated with given device.
534 *
535 * @dev: Device with which the mac-address cell is associated.
536 * @addrbuf: Buffer to which the MAC address will be copied on success.
537 *
538 * Returns 0 on success or a negative error number on failure.
539 */
540 int nvmem_get_mac_address(struct device *dev, void *addrbuf)
541 {
542 struct nvmem_cell *cell;
543 const void *mac;
544 size_t len;
545
546 cell = nvmem_cell_get(dev, "mac-address");
547 if (IS_ERR(cell))
548 return PTR_ERR(cell);
549
550 mac = nvmem_cell_read(cell, &len);
551 nvmem_cell_put(cell);
552
553 if (IS_ERR(mac))
554 return PTR_ERR(mac);
555
556 if (len != ETH_ALEN || !is_valid_ether_addr(mac)) {
557 kfree(mac);
558 return -EINVAL;
559 }
560
561 ether_addr_copy(addrbuf, mac);
562 kfree(mac);
563
564 return 0;
565 }
566 EXPORT_SYMBOL(nvmem_get_mac_address);