eth.c 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560
  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 <linux/uaccess.h>
  62. #include <net/pkt_sched.h>
  63. __setup("ether=", netdev_boot_setup);
  64. /**
  65. * eth_header - create the Ethernet header
  66. * @skb: buffer to alter
  67. * @dev: source device
  68. * @type: Ethernet type field
  69. * @daddr: destination address (NULL leave destination address)
  70. * @saddr: source address (NULL use device source address)
  71. * @len: packet length (<= skb->len)
  72. *
  73. *
  74. * Set the protocol type. For a packet of type ETH_P_802_3/2 we put the length
  75. * in here instead.
  76. */
  77. int eth_header(struct sk_buff *skb, struct net_device *dev,
  78. unsigned short type,
  79. const void *daddr, const void *saddr, unsigned int len)
  80. {
  81. struct ethhdr *eth = skb_push(skb, ETH_HLEN);
  82. if (type != ETH_P_802_3 && type != ETH_P_802_2)
  83. eth->h_proto = htons(type);
  84. else
  85. eth->h_proto = htons(len);
  86. /*
  87. * Set the source hardware address.
  88. */
  89. if (!saddr)
  90. saddr = dev->dev_addr;
  91. memcpy(eth->h_source, saddr, ETH_ALEN);
  92. if (daddr) {
  93. memcpy(eth->h_dest, daddr, ETH_ALEN);
  94. return ETH_HLEN;
  95. }
  96. /*
  97. * Anyway, the loopback-device should never use this function...
  98. */
  99. if (dev->flags & (IFF_LOOPBACK | IFF_NOARP)) {
  100. eth_zero_addr(eth->h_dest);
  101. return ETH_HLEN;
  102. }
  103. return -ETH_HLEN;
  104. }
  105. EXPORT_SYMBOL(eth_header);
  106. /**
  107. * eth_get_headlen - determine the length of header for an ethernet frame
  108. * @dev: pointer to network device
  109. * @data: pointer to start of frame
  110. * @len: total length of frame
  111. *
  112. * Make a best effort attempt to pull the length for all of the headers for
  113. * a given frame in a linear buffer.
  114. */
  115. u32 eth_get_headlen(const struct net_device *dev, void *data, unsigned int len)
  116. {
  117. const unsigned int flags = FLOW_DISSECTOR_F_PARSE_1ST_FRAG;
  118. const struct ethhdr *eth = (const struct ethhdr *)data;
  119. struct flow_keys_basic keys;
  120. /* this should never happen, but better safe than sorry */
  121. if (unlikely(len < sizeof(*eth)))
  122. return len;
  123. /* parse any remaining L2/L3 headers, check for L4 */
  124. if (!skb_flow_dissect_flow_keys_basic(dev_net(dev), NULL, &keys, data,
  125. eth->h_proto, sizeof(*eth),
  126. len, flags))
  127. return max_t(u32, keys.control.thoff, sizeof(*eth));
  128. /* parse for any L4 headers */
  129. return min_t(u32, __skb_get_poff(NULL, data, &keys, len), len);
  130. }
  131. EXPORT_SYMBOL(eth_get_headlen);
  132. /**
  133. * eth_type_trans - determine the packet's protocol ID.
  134. * @skb: received socket data
  135. * @dev: receiving network device
  136. *
  137. * The rule here is that we
  138. * assume 802.3 if the type field is short enough to be a length.
  139. * This is normal practice and works for any 'now in use' protocol.
  140. */
  141. __be16 eth_type_trans(struct sk_buff *skb, struct net_device *dev)
  142. {
  143. unsigned short _service_access_point;
  144. const unsigned short *sap;
  145. const struct ethhdr *eth;
  146. skb->dev = dev;
  147. skb_reset_mac_header(skb);
  148. eth = (struct ethhdr *)skb->data;
  149. skb_pull_inline(skb, ETH_HLEN);
  150. if (unlikely(!ether_addr_equal_64bits(eth->h_dest,
  151. dev->dev_addr))) {
  152. if (unlikely(is_multicast_ether_addr_64bits(eth->h_dest))) {
  153. if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
  154. skb->pkt_type = PACKET_BROADCAST;
  155. else
  156. skb->pkt_type = PACKET_MULTICAST;
  157. } else {
  158. skb->pkt_type = PACKET_OTHERHOST;
  159. }
  160. }
  161. /*
  162. * Some variants of DSA tagging don't have an ethertype field
  163. * at all, so we check here whether one of those tagging
  164. * variants has been configured on the receiving interface,
  165. * and if so, set skb->protocol without looking at the packet.
  166. * The DSA tagging protocol may be able to decode some but not all
  167. * traffic (for example only for management). In that case give it the
  168. * option to filter the packets from which it can decode source port
  169. * information.
  170. */
  171. if (unlikely(netdev_uses_dsa(dev)) && dsa_can_decode(skb, dev))
  172. return htons(ETH_P_XDSA);
  173. if (likely(eth_proto_is_802_3(eth->h_proto)))
  174. return eth->h_proto;
  175. /*
  176. * This is a magic hack to spot IPX packets. Older Novell breaks
  177. * the protocol design and runs IPX over 802.3 without an 802.2 LLC
  178. * layer. We look for FFFF which isn't a used 802.2 SSAP/DSAP. This
  179. * won't work for fault tolerant netware but does for the rest.
  180. */
  181. sap = skb_header_pointer(skb, 0, sizeof(*sap), &_service_access_point);
  182. if (sap && *sap == 0xFFFF)
  183. return htons(ETH_P_802_3);
  184. /*
  185. * Real 802.2 LLC
  186. */
  187. return htons(ETH_P_802_2);
  188. }
  189. EXPORT_SYMBOL(eth_type_trans);
  190. /**
  191. * eth_header_parse - extract hardware address from packet
  192. * @skb: packet to extract header from
  193. * @haddr: destination buffer
  194. */
  195. int eth_header_parse(const struct sk_buff *skb, unsigned char *haddr)
  196. {
  197. const struct ethhdr *eth = eth_hdr(skb);
  198. memcpy(haddr, eth->h_source, ETH_ALEN);
  199. return ETH_ALEN;
  200. }
  201. EXPORT_SYMBOL(eth_header_parse);
  202. /**
  203. * eth_header_cache - fill cache entry from neighbour
  204. * @neigh: source neighbour
  205. * @hh: destination cache entry
  206. * @type: Ethernet type field
  207. *
  208. * Create an Ethernet header template from the neighbour.
  209. */
  210. int eth_header_cache(const struct neighbour *neigh, struct hh_cache *hh, __be16 type)
  211. {
  212. struct ethhdr *eth;
  213. const struct net_device *dev = neigh->dev;
  214. eth = (struct ethhdr *)
  215. (((u8 *) hh->hh_data) + (HH_DATA_OFF(sizeof(*eth))));
  216. if (type == htons(ETH_P_802_3))
  217. return -1;
  218. eth->h_proto = type;
  219. memcpy(eth->h_source, dev->dev_addr, ETH_ALEN);
  220. memcpy(eth->h_dest, neigh->ha, ETH_ALEN);
  221. /* Pairs with READ_ONCE() in neigh_resolve_output(),
  222. * neigh_hh_output() and neigh_update_hhs().
  223. */
  224. smp_store_release(&hh->hh_len, ETH_HLEN);
  225. return 0;
  226. }
  227. EXPORT_SYMBOL(eth_header_cache);
  228. /**
  229. * eth_header_cache_update - update cache entry
  230. * @hh: destination cache entry
  231. * @dev: network device
  232. * @haddr: new hardware address
  233. *
  234. * Called by Address Resolution module to notify changes in address.
  235. */
  236. void eth_header_cache_update(struct hh_cache *hh,
  237. const struct net_device *dev,
  238. const unsigned char *haddr)
  239. {
  240. memcpy(((u8 *) hh->hh_data) + HH_DATA_OFF(sizeof(struct ethhdr)),
  241. haddr, ETH_ALEN);
  242. }
  243. EXPORT_SYMBOL(eth_header_cache_update);
  244. /**
  245. * eth_header_parser_protocol - extract protocol from L2 header
  246. * @skb: packet to extract protocol from
  247. */
  248. __be16 eth_header_parse_protocol(const struct sk_buff *skb)
  249. {
  250. const struct ethhdr *eth = eth_hdr(skb);
  251. return eth->h_proto;
  252. }
  253. EXPORT_SYMBOL(eth_header_parse_protocol);
  254. /**
  255. * eth_prepare_mac_addr_change - prepare for mac change
  256. * @dev: network device
  257. * @p: socket address
  258. */
  259. int eth_prepare_mac_addr_change(struct net_device *dev, void *p)
  260. {
  261. struct sockaddr *addr = p;
  262. if (!(dev->priv_flags & IFF_LIVE_ADDR_CHANGE) && netif_running(dev))
  263. return -EBUSY;
  264. if (!is_valid_ether_addr(addr->sa_data))
  265. return -EADDRNOTAVAIL;
  266. return 0;
  267. }
  268. EXPORT_SYMBOL(eth_prepare_mac_addr_change);
  269. /**
  270. * eth_commit_mac_addr_change - commit mac change
  271. * @dev: network device
  272. * @p: socket address
  273. */
  274. void eth_commit_mac_addr_change(struct net_device *dev, void *p)
  275. {
  276. struct sockaddr *addr = p;
  277. memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
  278. }
  279. EXPORT_SYMBOL(eth_commit_mac_addr_change);
  280. /**
  281. * eth_mac_addr - set new Ethernet hardware address
  282. * @dev: network device
  283. * @p: socket address
  284. *
  285. * Change hardware address of device.
  286. *
  287. * This doesn't change hardware matching, so needs to be overridden
  288. * for most real devices.
  289. */
  290. int eth_mac_addr(struct net_device *dev, void *p)
  291. {
  292. int ret;
  293. ret = eth_prepare_mac_addr_change(dev, p);
  294. if (ret < 0)
  295. return ret;
  296. eth_commit_mac_addr_change(dev, p);
  297. return 0;
  298. }
  299. EXPORT_SYMBOL(eth_mac_addr);
  300. int eth_validate_addr(struct net_device *dev)
  301. {
  302. if (!is_valid_ether_addr(dev->dev_addr))
  303. return -EADDRNOTAVAIL;
  304. return 0;
  305. }
  306. EXPORT_SYMBOL(eth_validate_addr);
  307. const struct header_ops eth_header_ops ____cacheline_aligned = {
  308. .create = eth_header,
  309. .parse = eth_header_parse,
  310. .cache = eth_header_cache,
  311. .cache_update = eth_header_cache_update,
  312. .parse_protocol = eth_header_parse_protocol,
  313. };
  314. /**
  315. * ether_setup - setup Ethernet network device
  316. * @dev: network device
  317. *
  318. * Fill in the fields of the device structure with Ethernet-generic values.
  319. */
  320. void ether_setup(struct net_device *dev)
  321. {
  322. dev->header_ops = &eth_header_ops;
  323. dev->type = ARPHRD_ETHER;
  324. dev->hard_header_len = ETH_HLEN;
  325. dev->min_header_len = ETH_HLEN;
  326. dev->mtu = ETH_DATA_LEN;
  327. dev->min_mtu = ETH_MIN_MTU;
  328. dev->max_mtu = ETH_DATA_LEN;
  329. dev->addr_len = ETH_ALEN;
  330. dev->tx_queue_len = DEFAULT_TX_QUEUE_LEN;
  331. dev->flags = IFF_BROADCAST|IFF_MULTICAST;
  332. dev->priv_flags |= IFF_TX_SKB_SHARING;
  333. eth_broadcast_addr(dev->broadcast);
  334. }
  335. EXPORT_SYMBOL(ether_setup);
  336. /**
  337. * alloc_etherdev_mqs - Allocates and sets up an Ethernet device
  338. * @sizeof_priv: Size of additional driver-private structure to be allocated
  339. * for this Ethernet device
  340. * @txqs: The number of TX queues this device has.
  341. * @rxqs: The number of RX queues this device has.
  342. *
  343. * Fill in the fields of the device structure with Ethernet-generic
  344. * values. Basically does everything except registering the device.
  345. *
  346. * Constructs a new net device, complete with a private data area of
  347. * size (sizeof_priv). A 32-byte (not bit) alignment is enforced for
  348. * this private data area.
  349. */
  350. struct net_device *alloc_etherdev_mqs(int sizeof_priv, unsigned int txqs,
  351. unsigned int rxqs)
  352. {
  353. return alloc_netdev_mqs(sizeof_priv, "eth%d", NET_NAME_UNKNOWN,
  354. ether_setup, txqs, rxqs);
  355. }
  356. EXPORT_SYMBOL(alloc_etherdev_mqs);
  357. ssize_t sysfs_format_mac(char *buf, const unsigned char *addr, int len)
  358. {
  359. return scnprintf(buf, PAGE_SIZE, "%*phC\n", len, addr);
  360. }
  361. EXPORT_SYMBOL(sysfs_format_mac);
  362. struct sk_buff *eth_gro_receive(struct list_head *head, struct sk_buff *skb)
  363. {
  364. const struct packet_offload *ptype;
  365. unsigned int hlen, off_eth;
  366. struct sk_buff *pp = NULL;
  367. struct ethhdr *eh, *eh2;
  368. struct sk_buff *p;
  369. __be16 type;
  370. int flush = 1;
  371. off_eth = skb_gro_offset(skb);
  372. hlen = off_eth + sizeof(*eh);
  373. eh = skb_gro_header_fast(skb, off_eth);
  374. if (skb_gro_header_hard(skb, hlen)) {
  375. eh = skb_gro_header_slow(skb, hlen, off_eth);
  376. if (unlikely(!eh))
  377. goto out;
  378. }
  379. flush = 0;
  380. list_for_each_entry(p, head, list) {
  381. if (!NAPI_GRO_CB(p)->same_flow)
  382. continue;
  383. eh2 = (struct ethhdr *)(p->data + off_eth);
  384. if (compare_ether_header(eh, eh2)) {
  385. NAPI_GRO_CB(p)->same_flow = 0;
  386. continue;
  387. }
  388. }
  389. type = eh->h_proto;
  390. rcu_read_lock();
  391. ptype = gro_find_receive_by_type(type);
  392. if (ptype == NULL) {
  393. flush = 1;
  394. goto out_unlock;
  395. }
  396. skb_gro_pull(skb, sizeof(*eh));
  397. skb_gro_postpull_rcsum(skb, eh, sizeof(*eh));
  398. pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb);
  399. out_unlock:
  400. rcu_read_unlock();
  401. out:
  402. skb_gro_flush_final(skb, pp, flush);
  403. return pp;
  404. }
  405. EXPORT_SYMBOL(eth_gro_receive);
  406. int eth_gro_complete(struct sk_buff *skb, int nhoff)
  407. {
  408. struct ethhdr *eh = (struct ethhdr *)(skb->data + nhoff);
  409. __be16 type = eh->h_proto;
  410. struct packet_offload *ptype;
  411. int err = -ENOSYS;
  412. if (skb->encapsulation)
  413. skb_set_inner_mac_header(skb, nhoff);
  414. rcu_read_lock();
  415. ptype = gro_find_complete_by_type(type);
  416. if (ptype != NULL)
  417. err = ptype->callbacks.gro_complete(skb, nhoff +
  418. sizeof(struct ethhdr));
  419. rcu_read_unlock();
  420. return err;
  421. }
  422. EXPORT_SYMBOL(eth_gro_complete);
  423. static struct packet_offload eth_packet_offload __read_mostly = {
  424. .type = cpu_to_be16(ETH_P_TEB),
  425. .priority = 10,
  426. .callbacks = {
  427. .gro_receive = eth_gro_receive,
  428. .gro_complete = eth_gro_complete,
  429. },
  430. };
  431. static int __init eth_offload_init(void)
  432. {
  433. dev_add_offload(&eth_packet_offload);
  434. return 0;
  435. }
  436. fs_initcall(eth_offload_init);
  437. unsigned char * __weak arch_get_platform_mac_address(void)
  438. {
  439. return NULL;
  440. }
  441. int eth_platform_get_mac_address(struct device *dev, u8 *mac_addr)
  442. {
  443. const unsigned char *addr = NULL;
  444. if (dev->of_node)
  445. addr = of_get_mac_address(dev->of_node);
  446. if (IS_ERR_OR_NULL(addr))
  447. addr = arch_get_platform_mac_address();
  448. if (!addr)
  449. return -ENODEV;
  450. ether_addr_copy(mac_addr, addr);
  451. return 0;
  452. }
  453. EXPORT_SYMBOL(eth_platform_get_mac_address);
  454. /**
  455. * Obtain the MAC address from an nvmem cell named 'mac-address' associated
  456. * with given device.
  457. *
  458. * @dev: Device with which the mac-address cell is associated.
  459. * @addrbuf: Buffer to which the MAC address will be copied on success.
  460. *
  461. * Returns 0 on success or a negative error number on failure.
  462. */
  463. int nvmem_get_mac_address(struct device *dev, void *addrbuf)
  464. {
  465. struct nvmem_cell *cell;
  466. const void *mac;
  467. size_t len;
  468. cell = nvmem_cell_get(dev, "mac-address");
  469. if (IS_ERR(cell))
  470. return PTR_ERR(cell);
  471. mac = nvmem_cell_read(cell, &len);
  472. nvmem_cell_put(cell);
  473. if (IS_ERR(mac))
  474. return PTR_ERR(mac);
  475. if (len != ETH_ALEN || !is_valid_ether_addr(mac)) {
  476. kfree(mac);
  477. return -EINVAL;
  478. }
  479. ether_addr_copy(addrbuf, mac);
  480. kfree(mac);
  481. return 0;
  482. }
  483. EXPORT_SYMBOL(nvmem_get_mac_address);