eth.c 13 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518
  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 <net/dst.h>
  56. #include <net/arp.h>
  57. #include <net/sock.h>
  58. #include <net/ipv6.h>
  59. #include <net/ip.h>
  60. #include <net/dsa.h>
  61. #include <linux/uaccess.h>
  62. __setup("ether=", netdev_boot_setup);
  63. /**
  64. * eth_header - create the Ethernet header
  65. * @skb: buffer to alter
  66. * @dev: source device
  67. * @type: Ethernet type field
  68. * @daddr: destination address (NULL leave destination address)
  69. * @saddr: source address (NULL use device source address)
  70. * @len: packet length (<= skb->len)
  71. *
  72. *
  73. * Set the protocol type. For a packet of type ETH_P_802_3/2 we put the length
  74. * in here instead.
  75. */
  76. int eth_header(struct sk_buff *skb, struct net_device *dev,
  77. unsigned short type,
  78. const void *daddr, const void *saddr, unsigned int len)
  79. {
  80. struct ethhdr *eth = (struct ethhdr *)skb_push(skb, ETH_HLEN);
  81. if (type != ETH_P_802_3 && type != ETH_P_802_2)
  82. eth->h_proto = htons(type);
  83. else
  84. eth->h_proto = htons(len);
  85. /*
  86. * Set the source hardware address.
  87. */
  88. if (!saddr)
  89. saddr = dev->dev_addr;
  90. memcpy(eth->h_source, saddr, ETH_ALEN);
  91. if (daddr) {
  92. memcpy(eth->h_dest, daddr, ETH_ALEN);
  93. return ETH_HLEN;
  94. }
  95. /*
  96. * Anyway, the loopback-device should never use this function...
  97. */
  98. if (dev->flags & (IFF_LOOPBACK | IFF_NOARP)) {
  99. memset(eth->h_dest, 0, ETH_ALEN);
  100. return ETH_HLEN;
  101. }
  102. return -ETH_HLEN;
  103. }
  104. EXPORT_SYMBOL(eth_header);
  105. /**
  106. * eth_rebuild_header- rebuild the Ethernet MAC header.
  107. * @skb: socket buffer to update
  108. *
  109. * This is called after an ARP or IPV6 ndisc it's resolution on this
  110. * sk_buff. We now let protocol (ARP) fill in the other fields.
  111. *
  112. * This routine CANNOT use cached dst->neigh!
  113. * Really, it is used only when dst->neigh is wrong.
  114. */
  115. int eth_rebuild_header(struct sk_buff *skb)
  116. {
  117. struct ethhdr *eth = (struct ethhdr *)skb->data;
  118. struct net_device *dev = skb->dev;
  119. switch (eth->h_proto) {
  120. #ifdef CONFIG_INET
  121. case htons(ETH_P_IP):
  122. return arp_find(eth->h_dest, skb);
  123. #endif
  124. default:
  125. netdev_dbg(dev,
  126. "%s: unable to resolve type %X addresses.\n",
  127. dev->name, ntohs(eth->h_proto));
  128. memcpy(eth->h_source, dev->dev_addr, ETH_ALEN);
  129. break;
  130. }
  131. return 0;
  132. }
  133. EXPORT_SYMBOL(eth_rebuild_header);
  134. /**
  135. * eth_get_headlen - determine the the length of header for an ethernet frame
  136. * @data: pointer to start of frame
  137. * @len: total length of frame
  138. *
  139. * Make a best effort attempt to pull the length for all of the headers for
  140. * a given frame in a linear buffer.
  141. */
  142. u32 eth_get_headlen(void *data, unsigned int len)
  143. {
  144. const struct ethhdr *eth = (const struct ethhdr *)data;
  145. struct flow_keys keys;
  146. /* this should never happen, but better safe than sorry */
  147. if (len < sizeof(*eth))
  148. return len;
  149. /* parse any remaining L2/L3 headers, check for L4 */
  150. if (!__skb_flow_dissect(NULL, &keys, data,
  151. eth->h_proto, sizeof(*eth), len))
  152. return max_t(u32, keys.thoff, sizeof(*eth));
  153. /* parse for any L4 headers */
  154. return min_t(u32, __skb_get_poff(NULL, data, &keys, len), len);
  155. }
  156. EXPORT_SYMBOL(eth_get_headlen);
  157. /**
  158. * eth_type_trans - determine the packet's protocol ID.
  159. * @skb: received socket data
  160. * @dev: receiving network device
  161. *
  162. * The rule here is that we
  163. * assume 802.3 if the type field is short enough to be a length.
  164. * This is normal practice and works for any 'now in use' protocol.
  165. */
  166. __be16 eth_type_trans(struct sk_buff *skb, struct net_device *dev)
  167. {
  168. unsigned short _service_access_point;
  169. const unsigned short *sap;
  170. const struct ethhdr *eth;
  171. skb->dev = dev;
  172. skb_reset_mac_header(skb);
  173. skb_pull_inline(skb, ETH_HLEN);
  174. eth = eth_hdr(skb);
  175. if (unlikely(is_multicast_ether_addr(eth->h_dest))) {
  176. if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
  177. skb->pkt_type = PACKET_BROADCAST;
  178. else
  179. skb->pkt_type = PACKET_MULTICAST;
  180. }
  181. else if (unlikely(!ether_addr_equal_64bits(eth->h_dest,
  182. dev->dev_addr)))
  183. skb->pkt_type = PACKET_OTHERHOST;
  184. /*
  185. * Some variants of DSA tagging don't have an ethertype field
  186. * at all, so we check here whether one of those tagging
  187. * variants has been configured on the receiving interface,
  188. * and if so, set skb->protocol without looking at the packet.
  189. */
  190. if (unlikely(netdev_uses_dsa(dev)))
  191. return htons(ETH_P_XDSA);
  192. if (likely(ntohs(eth->h_proto) >= ETH_P_802_3_MIN))
  193. return eth->h_proto;
  194. /*
  195. * This is a magic hack to spot IPX packets. Older Novell breaks
  196. * the protocol design and runs IPX over 802.3 without an 802.2 LLC
  197. * layer. We look for FFFF which isn't a used 802.2 SSAP/DSAP. This
  198. * won't work for fault tolerant netware but does for the rest.
  199. */
  200. sap = skb_header_pointer(skb, 0, sizeof(*sap), &_service_access_point);
  201. if (sap && *sap == 0xFFFF)
  202. return htons(ETH_P_802_3);
  203. /*
  204. * Real 802.2 LLC
  205. */
  206. return htons(ETH_P_802_2);
  207. }
  208. EXPORT_SYMBOL(eth_type_trans);
  209. /**
  210. * eth_header_parse - extract hardware address from packet
  211. * @skb: packet to extract header from
  212. * @haddr: destination buffer
  213. */
  214. int eth_header_parse(const struct sk_buff *skb, unsigned char *haddr)
  215. {
  216. const struct ethhdr *eth = eth_hdr(skb);
  217. memcpy(haddr, eth->h_source, ETH_ALEN);
  218. return ETH_ALEN;
  219. }
  220. EXPORT_SYMBOL(eth_header_parse);
  221. /**
  222. * eth_header_cache - fill cache entry from neighbour
  223. * @neigh: source neighbour
  224. * @hh: destination cache entry
  225. * @type: Ethernet type field
  226. *
  227. * Create an Ethernet header template from the neighbour.
  228. */
  229. int eth_header_cache(const struct neighbour *neigh, struct hh_cache *hh, __be16 type)
  230. {
  231. struct ethhdr *eth;
  232. const struct net_device *dev = neigh->dev;
  233. eth = (struct ethhdr *)
  234. (((u8 *) hh->hh_data) + (HH_DATA_OFF(sizeof(*eth))));
  235. if (type == htons(ETH_P_802_3))
  236. return -1;
  237. eth->h_proto = type;
  238. memcpy(eth->h_source, dev->dev_addr, ETH_ALEN);
  239. memcpy(eth->h_dest, neigh->ha, ETH_ALEN);
  240. hh->hh_len = ETH_HLEN;
  241. return 0;
  242. }
  243. EXPORT_SYMBOL(eth_header_cache);
  244. /**
  245. * eth_header_cache_update - update cache entry
  246. * @hh: destination cache entry
  247. * @dev: network device
  248. * @haddr: new hardware address
  249. *
  250. * Called by Address Resolution module to notify changes in address.
  251. */
  252. void eth_header_cache_update(struct hh_cache *hh,
  253. const struct net_device *dev,
  254. const unsigned char *haddr)
  255. {
  256. memcpy(((u8 *) hh->hh_data) + HH_DATA_OFF(sizeof(struct ethhdr)),
  257. haddr, ETH_ALEN);
  258. }
  259. EXPORT_SYMBOL(eth_header_cache_update);
  260. /**
  261. * eth_prepare_mac_addr_change - prepare for mac change
  262. * @dev: network device
  263. * @p: socket address
  264. */
  265. int eth_prepare_mac_addr_change(struct net_device *dev, void *p)
  266. {
  267. struct sockaddr *addr = p;
  268. if (!(dev->priv_flags & IFF_LIVE_ADDR_CHANGE) && netif_running(dev))
  269. return -EBUSY;
  270. if (!is_valid_ether_addr(addr->sa_data))
  271. return -EADDRNOTAVAIL;
  272. return 0;
  273. }
  274. EXPORT_SYMBOL(eth_prepare_mac_addr_change);
  275. /**
  276. * eth_commit_mac_addr_change - commit mac change
  277. * @dev: network device
  278. * @p: socket address
  279. */
  280. void eth_commit_mac_addr_change(struct net_device *dev, void *p)
  281. {
  282. struct sockaddr *addr = p;
  283. memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
  284. }
  285. EXPORT_SYMBOL(eth_commit_mac_addr_change);
  286. /**
  287. * eth_mac_addr - set new Ethernet hardware address
  288. * @dev: network device
  289. * @p: socket address
  290. *
  291. * Change hardware address of device.
  292. *
  293. * This doesn't change hardware matching, so needs to be overridden
  294. * for most real devices.
  295. */
  296. int eth_mac_addr(struct net_device *dev, void *p)
  297. {
  298. int ret;
  299. ret = eth_prepare_mac_addr_change(dev, p);
  300. if (ret < 0)
  301. return ret;
  302. eth_commit_mac_addr_change(dev, p);
  303. return 0;
  304. }
  305. EXPORT_SYMBOL(eth_mac_addr);
  306. /**
  307. * eth_change_mtu - set new MTU size
  308. * @dev: network device
  309. * @new_mtu: new Maximum Transfer Unit
  310. *
  311. * Allow changing MTU size. Needs to be overridden for devices
  312. * supporting jumbo frames.
  313. */
  314. int eth_change_mtu(struct net_device *dev, int new_mtu)
  315. {
  316. if (new_mtu < 68 || new_mtu > ETH_DATA_LEN)
  317. return -EINVAL;
  318. dev->mtu = new_mtu;
  319. return 0;
  320. }
  321. EXPORT_SYMBOL(eth_change_mtu);
  322. int eth_validate_addr(struct net_device *dev)
  323. {
  324. if (!is_valid_ether_addr(dev->dev_addr))
  325. return -EADDRNOTAVAIL;
  326. return 0;
  327. }
  328. EXPORT_SYMBOL(eth_validate_addr);
  329. const struct header_ops eth_header_ops ____cacheline_aligned = {
  330. .create = eth_header,
  331. .parse = eth_header_parse,
  332. .rebuild = eth_rebuild_header,
  333. .cache = eth_header_cache,
  334. .cache_update = eth_header_cache_update,
  335. };
  336. /**
  337. * ether_setup - setup Ethernet network device
  338. * @dev: network device
  339. *
  340. * Fill in the fields of the device structure with Ethernet-generic values.
  341. */
  342. void ether_setup(struct net_device *dev)
  343. {
  344. dev->header_ops = &eth_header_ops;
  345. dev->type = ARPHRD_ETHER;
  346. dev->hard_header_len = ETH_HLEN;
  347. dev->mtu = ETH_DATA_LEN;
  348. dev->addr_len = ETH_ALEN;
  349. dev->tx_queue_len = 1000; /* Ethernet wants good queues */
  350. dev->flags = IFF_BROADCAST|IFF_MULTICAST;
  351. dev->priv_flags |= IFF_TX_SKB_SHARING;
  352. memset(dev->broadcast, 0xFF, ETH_ALEN);
  353. }
  354. EXPORT_SYMBOL(ether_setup);
  355. /**
  356. * alloc_etherdev_mqs - Allocates and sets up an Ethernet device
  357. * @sizeof_priv: Size of additional driver-private structure to be allocated
  358. * for this Ethernet device
  359. * @txqs: The number of TX queues this device has.
  360. * @rxqs: The number of RX queues this device has.
  361. *
  362. * Fill in the fields of the device structure with Ethernet-generic
  363. * values. Basically does everything except registering the device.
  364. *
  365. * Constructs a new net device, complete with a private data area of
  366. * size (sizeof_priv). A 32-byte (not bit) alignment is enforced for
  367. * this private data area.
  368. */
  369. struct net_device *alloc_etherdev_mqs(int sizeof_priv, unsigned int txqs,
  370. unsigned int rxqs)
  371. {
  372. return alloc_netdev_mqs(sizeof_priv, "eth%d", NET_NAME_UNKNOWN,
  373. ether_setup, txqs, rxqs);
  374. }
  375. EXPORT_SYMBOL(alloc_etherdev_mqs);
  376. ssize_t sysfs_format_mac(char *buf, const unsigned char *addr, int len)
  377. {
  378. return scnprintf(buf, PAGE_SIZE, "%*phC\n", len, addr);
  379. }
  380. EXPORT_SYMBOL(sysfs_format_mac);
  381. struct sk_buff **eth_gro_receive(struct sk_buff **head,
  382. struct sk_buff *skb)
  383. {
  384. struct sk_buff *p, **pp = NULL;
  385. struct ethhdr *eh, *eh2;
  386. unsigned int hlen, off_eth;
  387. const struct packet_offload *ptype;
  388. __be16 type;
  389. int flush = 1;
  390. off_eth = skb_gro_offset(skb);
  391. hlen = off_eth + sizeof(*eh);
  392. eh = skb_gro_header_fast(skb, off_eth);
  393. if (skb_gro_header_hard(skb, hlen)) {
  394. eh = skb_gro_header_slow(skb, hlen, off_eth);
  395. if (unlikely(!eh))
  396. goto out;
  397. }
  398. flush = 0;
  399. for (p = *head; p; p = p->next) {
  400. if (!NAPI_GRO_CB(p)->same_flow)
  401. continue;
  402. eh2 = (struct ethhdr *)(p->data + off_eth);
  403. if (compare_ether_header(eh, eh2)) {
  404. NAPI_GRO_CB(p)->same_flow = 0;
  405. continue;
  406. }
  407. }
  408. type = eh->h_proto;
  409. rcu_read_lock();
  410. ptype = gro_find_receive_by_type(type);
  411. if (ptype == NULL) {
  412. flush = 1;
  413. goto out_unlock;
  414. }
  415. skb_gro_pull(skb, sizeof(*eh));
  416. skb_gro_postpull_rcsum(skb, eh, sizeof(*eh));
  417. pp = ptype->callbacks.gro_receive(head, skb);
  418. out_unlock:
  419. rcu_read_unlock();
  420. out:
  421. NAPI_GRO_CB(skb)->flush |= flush;
  422. return pp;
  423. }
  424. EXPORT_SYMBOL(eth_gro_receive);
  425. int eth_gro_complete(struct sk_buff *skb, int nhoff)
  426. {
  427. struct ethhdr *eh = (struct ethhdr *)(skb->data + nhoff);
  428. __be16 type = eh->h_proto;
  429. struct packet_offload *ptype;
  430. int err = -ENOSYS;
  431. if (skb->encapsulation)
  432. skb_set_inner_mac_header(skb, nhoff);
  433. rcu_read_lock();
  434. ptype = gro_find_complete_by_type(type);
  435. if (ptype != NULL)
  436. err = ptype->callbacks.gro_complete(skb, nhoff +
  437. sizeof(struct ethhdr));
  438. rcu_read_unlock();
  439. return err;
  440. }
  441. EXPORT_SYMBOL(eth_gro_complete);
  442. static struct packet_offload eth_packet_offload __read_mostly = {
  443. .type = cpu_to_be16(ETH_P_TEB),
  444. .callbacks = {
  445. .gro_receive = eth_gro_receive,
  446. .gro_complete = eth_gro_complete,
  447. },
  448. };
  449. static int __init eth_offload_init(void)
  450. {
  451. dev_add_offload(&eth_packet_offload);
  452. return 0;
  453. }
  454. fs_initcall(eth_offload_init);