linux/net/ethernet/eth.c
<<
>>
Prefs
   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
  64__setup("ether=", netdev_boot_setup);
  65
  66/**
  67 * eth_header - create the Ethernet header
  68 * @skb:        buffer to alter
  69 * @dev:        source device
  70 * @type:       Ethernet type field
  71 * @daddr: destination address (NULL leave destination address)
  72 * @saddr: source address (NULL use device source address)
  73 * @len:   packet length (<= skb->len)
  74 *
  75 *
  76 * Set the protocol type. For a packet of type ETH_P_802_3/2 we put the length
  77 * in here instead.
  78 */
  79int eth_header(struct sk_buff *skb, struct net_device *dev,
  80               unsigned short type,
  81               const void *daddr, const void *saddr, unsigned int len)
  82{
  83        struct ethhdr *eth = skb_push(skb, ETH_HLEN);
  84
  85        if (type != ETH_P_802_3 && type != ETH_P_802_2)
  86                eth->h_proto = htons(type);
  87        else
  88                eth->h_proto = htons(len);
  89
  90        /*
  91         *      Set the source hardware address.
  92         */
  93
  94        if (!saddr)
  95                saddr = dev->dev_addr;
  96        memcpy(eth->h_source, saddr, ETH_ALEN);
  97
  98        if (daddr) {
  99                memcpy(eth->h_dest, daddr, ETH_ALEN);
 100                return ETH_HLEN;
 101        }
 102
 103        /*
 104         *      Anyway, the loopback-device should never use this function...
 105         */
 106
 107        if (dev->flags & (IFF_LOOPBACK | IFF_NOARP)) {
 108                eth_zero_addr(eth->h_dest);
 109                return ETH_HLEN;
 110        }
 111
 112        return -ETH_HLEN;
 113}
 114EXPORT_SYMBOL(eth_header);
 115
 116/**
 117 * eth_get_headlen - determine the length of header for an ethernet frame
 118 * @dev: pointer to network device
 119 * @data: pointer to start of frame
 120 * @len: total length of frame
 121 *
 122 * Make a best effort attempt to pull the length for all of the headers for
 123 * a given frame in a linear buffer.
 124 */
 125u32 eth_get_headlen(const struct net_device *dev, void *data, unsigned int len)
 126{
 127        const unsigned int flags = FLOW_DISSECTOR_F_PARSE_1ST_FRAG;
 128        const struct ethhdr *eth = (const struct ethhdr *)data;
 129        struct flow_keys_basic keys;
 130
 131        /* this should never happen, but better safe than sorry */
 132        if (unlikely(len < sizeof(*eth)))
 133                return len;
 134
 135        /* parse any remaining L2/L3 headers, check for L4 */
 136        if (!skb_flow_dissect_flow_keys_basic(dev_net(dev), NULL, &keys, data,
 137                                              eth->h_proto, sizeof(*eth),
 138                                              len, flags))
 139                return max_t(u32, keys.control.thoff, sizeof(*eth));
 140
 141        /* parse for any L4 headers */
 142        return min_t(u32, __skb_get_poff(NULL, data, &keys, len), len);
 143}
 144EXPORT_SYMBOL(eth_get_headlen);
 145
 146/**
 147 * eth_type_trans - determine the packet's protocol ID.
 148 * @skb: received socket data
 149 * @dev: receiving network device
 150 *
 151 * The rule here is that we
 152 * assume 802.3 if the type field is short enough to be a length.
 153 * This is normal practice and works for any 'now in use' protocol.
 154 */
 155__be16 eth_type_trans(struct sk_buff *skb, struct net_device *dev)
 156{
 157        unsigned short _service_access_point;
 158        const unsigned short *sap;
 159        const struct ethhdr *eth;
 160
 161        skb->dev = dev;
 162        skb_reset_mac_header(skb);
 163
 164        eth = (struct ethhdr *)skb->data;
 165        skb_pull_inline(skb, ETH_HLEN);
 166
 167        if (unlikely(!ether_addr_equal_64bits(eth->h_dest,
 168                                              dev->dev_addr))) {
 169                if (unlikely(is_multicast_ether_addr_64bits(eth->h_dest))) {
 170                        if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
 171                                skb->pkt_type = PACKET_BROADCAST;
 172                        else
 173                                skb->pkt_type = PACKET_MULTICAST;
 174                } else {
 175                        skb->pkt_type = PACKET_OTHERHOST;
 176                }
 177        }
 178
 179        /*
 180         * Some variants of DSA tagging don't have an ethertype field
 181         * at all, so we check here whether one of those tagging
 182         * variants has been configured on the receiving interface,
 183         * and if so, set skb->protocol without looking at the packet.
 184         * The DSA tagging protocol may be able to decode some but not all
 185         * traffic (for example only for management). In that case give it the
 186         * option to filter the packets from which it can decode source port
 187         * information.
 188         */
 189        if (unlikely(netdev_uses_dsa(dev)) && dsa_can_decode(skb, dev))
 190                return htons(ETH_P_XDSA);
 191
 192        if (likely(eth_proto_is_802_3(eth->h_proto)))
 193                return eth->h_proto;
 194
 195        /*
 196         *      This is a magic hack to spot IPX packets. Older Novell breaks
 197         *      the protocol design and runs IPX over 802.3 without an 802.2 LLC
 198         *      layer. We look for FFFF which isn't a used 802.2 SSAP/DSAP. This
 199         *      won't work for fault tolerant netware but does for the rest.
 200         */
 201        sap = skb_header_pointer(skb, 0, sizeof(*sap), &_service_access_point);
 202        if (sap && *sap == 0xFFFF)
 203                return htons(ETH_P_802_3);
 204
 205        /*
 206         *      Real 802.2 LLC
 207         */
 208        return htons(ETH_P_802_2);
 209}
 210EXPORT_SYMBOL(eth_type_trans);
 211
 212/**
 213 * eth_header_parse - extract hardware address from packet
 214 * @skb: packet to extract header from
 215 * @haddr: destination buffer
 216 */
 217int eth_header_parse(const struct sk_buff *skb, unsigned char *haddr)
 218{
 219        const struct ethhdr *eth = eth_hdr(skb);
 220        memcpy(haddr, eth->h_source, ETH_ALEN);
 221        return ETH_ALEN;
 222}
 223EXPORT_SYMBOL(eth_header_parse);
 224
 225/**
 226 * eth_header_cache - fill cache entry from neighbour
 227 * @neigh: source neighbour
 228 * @hh: destination cache entry
 229 * @type: Ethernet type field
 230 *
 231 * Create an Ethernet header template from the neighbour.
 232 */
 233int eth_header_cache(const struct neighbour *neigh, struct hh_cache *hh, __be16 type)
 234{
 235        struct ethhdr *eth;
 236        const struct net_device *dev = neigh->dev;
 237
 238        eth = (struct ethhdr *)
 239            (((u8 *) hh->hh_data) + (HH_DATA_OFF(sizeof(*eth))));
 240
 241        if (type == htons(ETH_P_802_3))
 242                return -1;
 243
 244        eth->h_proto = type;
 245        memcpy(eth->h_source, dev->dev_addr, ETH_ALEN);
 246        memcpy(eth->h_dest, neigh->ha, ETH_ALEN);
 247
 248        /* Pairs with READ_ONCE() in neigh_resolve_output(),
 249         * neigh_hh_output() and neigh_update_hhs().
 250         */
 251        smp_store_release(&hh->hh_len, ETH_HLEN);
 252
 253        return 0;
 254}
 255EXPORT_SYMBOL(eth_header_cache);
 256
 257/**
 258 * eth_header_cache_update - update cache entry
 259 * @hh: destination cache entry
 260 * @dev: network device
 261 * @haddr: new hardware address
 262 *
 263 * Called by Address Resolution module to notify changes in address.
 264 */
 265void eth_header_cache_update(struct hh_cache *hh,
 266                             const struct net_device *dev,
 267                             const unsigned char *haddr)
 268{
 269        memcpy(((u8 *) hh->hh_data) + HH_DATA_OFF(sizeof(struct ethhdr)),
 270               haddr, ETH_ALEN);
 271}
 272EXPORT_SYMBOL(eth_header_cache_update);
 273
 274/**
 275 * eth_header_parser_protocol - extract protocol from L2 header
 276 * @skb: packet to extract protocol from
 277 */
 278__be16 eth_header_parse_protocol(const struct sk_buff *skb)
 279{
 280        const struct ethhdr *eth = eth_hdr(skb);
 281
 282        return eth->h_proto;
 283}
 284EXPORT_SYMBOL(eth_header_parse_protocol);
 285
 286/**
 287 * eth_prepare_mac_addr_change - prepare for mac change
 288 * @dev: network device
 289 * @p: socket address
 290 */
 291int eth_prepare_mac_addr_change(struct net_device *dev, void *p)
 292{
 293        struct sockaddr *addr = p;
 294
 295        if (!(dev->priv_flags & IFF_LIVE_ADDR_CHANGE) && netif_running(dev))
 296                return -EBUSY;
 297        if (!is_valid_ether_addr(addr->sa_data))
 298                return -EADDRNOTAVAIL;
 299        return 0;
 300}
 301EXPORT_SYMBOL(eth_prepare_mac_addr_change);
 302
 303/**
 304 * eth_commit_mac_addr_change - commit mac change
 305 * @dev: network device
 306 * @p: socket address
 307 */
 308void eth_commit_mac_addr_change(struct net_device *dev, void *p)
 309{
 310        struct sockaddr *addr = p;
 311
 312        memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
 313}
 314EXPORT_SYMBOL(eth_commit_mac_addr_change);
 315
 316/**
 317 * eth_mac_addr - set new Ethernet hardware address
 318 * @dev: network device
 319 * @p: socket address
 320 *
 321 * Change hardware address of device.
 322 *
 323 * This doesn't change hardware matching, so needs to be overridden
 324 * for most real devices.
 325 */
 326int eth_mac_addr(struct net_device *dev, void *p)
 327{
 328        int ret;
 329
 330        ret = eth_prepare_mac_addr_change(dev, p);
 331        if (ret < 0)
 332                return ret;
 333        eth_commit_mac_addr_change(dev, p);
 334        return 0;
 335}
 336EXPORT_SYMBOL(eth_mac_addr);
 337
 338int eth_validate_addr(struct net_device *dev)
 339{
 340        if (!is_valid_ether_addr(dev->dev_addr))
 341                return -EADDRNOTAVAIL;
 342
 343        return 0;
 344}
 345EXPORT_SYMBOL(eth_validate_addr);
 346
 347const struct header_ops eth_header_ops ____cacheline_aligned = {
 348        .create         = eth_header,
 349        .parse          = eth_header_parse,
 350        .cache          = eth_header_cache,
 351        .cache_update   = eth_header_cache_update,
 352        .parse_protocol = eth_header_parse_protocol,
 353};
 354
 355/**
 356 * ether_setup - setup Ethernet network device
 357 * @dev: network device
 358 *
 359 * Fill in the fields of the device structure with Ethernet-generic values.
 360 */
 361void ether_setup(struct net_device *dev)
 362{
 363        dev->header_ops         = &eth_header_ops;
 364        dev->type               = ARPHRD_ETHER;
 365        dev->hard_header_len    = ETH_HLEN;
 366        dev->min_header_len     = ETH_HLEN;
 367        dev->mtu                = ETH_DATA_LEN;
 368        dev->min_mtu            = ETH_MIN_MTU;
 369        dev->max_mtu            = ETH_DATA_LEN;
 370        dev->addr_len           = ETH_ALEN;
 371        dev->tx_queue_len       = DEFAULT_TX_QUEUE_LEN;
 372        dev->flags              = IFF_BROADCAST|IFF_MULTICAST;
 373        dev->priv_flags         |= IFF_TX_SKB_SHARING;
 374
 375        eth_broadcast_addr(dev->broadcast);
 376
 377}
 378EXPORT_SYMBOL(ether_setup);
 379
 380/**
 381 * alloc_etherdev_mqs - Allocates and sets up an Ethernet device
 382 * @sizeof_priv: Size of additional driver-private structure to be allocated
 383 *      for this Ethernet device
 384 * @txqs: The number of TX queues this device has.
 385 * @rxqs: The number of RX queues this device has.
 386 *
 387 * Fill in the fields of the device structure with Ethernet-generic
 388 * values. Basically does everything except registering the device.
 389 *
 390 * Constructs a new net device, complete with a private data area of
 391 * size (sizeof_priv).  A 32-byte (not bit) alignment is enforced for
 392 * this private data area.
 393 */
 394
 395struct net_device *alloc_etherdev_mqs(int sizeof_priv, unsigned int txqs,
 396                                      unsigned int rxqs)
 397{
 398        return alloc_netdev_mqs(sizeof_priv, "eth%d", NET_NAME_UNKNOWN,
 399                                ether_setup, txqs, rxqs);
 400}
 401EXPORT_SYMBOL(alloc_etherdev_mqs);
 402
 403ssize_t sysfs_format_mac(char *buf, const unsigned char *addr, int len)
 404{
 405        return scnprintf(buf, PAGE_SIZE, "%*phC\n", len, addr);
 406}
 407EXPORT_SYMBOL(sysfs_format_mac);
 408
 409struct sk_buff *eth_gro_receive(struct list_head *head, struct sk_buff *skb)
 410{
 411        const struct packet_offload *ptype;
 412        unsigned int hlen, off_eth;
 413        struct sk_buff *pp = NULL;
 414        struct ethhdr *eh, *eh2;
 415        struct sk_buff *p;
 416        __be16 type;
 417        int flush = 1;
 418
 419        off_eth = skb_gro_offset(skb);
 420        hlen = off_eth + sizeof(*eh);
 421        eh = skb_gro_header_fast(skb, off_eth);
 422        if (skb_gro_header_hard(skb, hlen)) {
 423                eh = skb_gro_header_slow(skb, hlen, off_eth);
 424                if (unlikely(!eh))
 425                        goto out;
 426        }
 427
 428        flush = 0;
 429
 430        list_for_each_entry(p, head, list) {
 431                if (!NAPI_GRO_CB(p)->same_flow)
 432                        continue;
 433
 434                eh2 = (struct ethhdr *)(p->data + off_eth);
 435                if (compare_ether_header(eh, eh2)) {
 436                        NAPI_GRO_CB(p)->same_flow = 0;
 437                        continue;
 438                }
 439        }
 440
 441        type = eh->h_proto;
 442
 443        rcu_read_lock();
 444        ptype = gro_find_receive_by_type(type);
 445        if (ptype == NULL) {
 446                flush = 1;
 447                goto out_unlock;
 448        }
 449
 450        skb_gro_pull(skb, sizeof(*eh));
 451        skb_gro_postpull_rcsum(skb, eh, sizeof(*eh));
 452        pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb);
 453
 454out_unlock:
 455        rcu_read_unlock();
 456out:
 457        skb_gro_flush_final(skb, pp, flush);
 458
 459        return pp;
 460}
 461EXPORT_SYMBOL(eth_gro_receive);
 462
 463int eth_gro_complete(struct sk_buff *skb, int nhoff)
 464{
 465        struct ethhdr *eh = (struct ethhdr *)(skb->data + nhoff);
 466        __be16 type = eh->h_proto;
 467        struct packet_offload *ptype;
 468        int err = -ENOSYS;
 469
 470        if (skb->encapsulation)
 471                skb_set_inner_mac_header(skb, nhoff);
 472
 473        rcu_read_lock();
 474        ptype = gro_find_complete_by_type(type);
 475        if (ptype != NULL)
 476                err = ptype->callbacks.gro_complete(skb, nhoff +
 477                                                    sizeof(struct ethhdr));
 478
 479        rcu_read_unlock();
 480        return err;
 481}
 482EXPORT_SYMBOL(eth_gro_complete);
 483
 484static struct packet_offload eth_packet_offload __read_mostly = {
 485        .type = cpu_to_be16(ETH_P_TEB),
 486        .priority = 10,
 487        .callbacks = {
 488                .gro_receive = eth_gro_receive,
 489                .gro_complete = eth_gro_complete,
 490        },
 491};
 492
 493static int __init eth_offload_init(void)
 494{
 495        dev_add_offload(&eth_packet_offload);
 496
 497        return 0;
 498}
 499
 500fs_initcall(eth_offload_init);
 501
 502unsigned char * __weak arch_get_platform_mac_address(void)
 503{
 504        return NULL;
 505}
 506
 507int eth_platform_get_mac_address(struct device *dev, u8 *mac_addr)
 508{
 509        const unsigned char *addr = NULL;
 510
 511        if (dev->of_node)
 512                addr = of_get_mac_address(dev->of_node);
 513        if (IS_ERR_OR_NULL(addr))
 514                addr = arch_get_platform_mac_address();
 515
 516        if (!addr)
 517                return -ENODEV;
 518
 519        ether_addr_copy(mac_addr, addr);
 520
 521        return 0;
 522}
 523EXPORT_SYMBOL(eth_platform_get_mac_address);
 524
 525/**
 526 * Obtain the MAC address from an nvmem cell named 'mac-address' associated
 527 * with given device.
 528 *
 529 * @dev:        Device with which the mac-address cell is associated.
 530 * @addrbuf:    Buffer to which the MAC address will be copied on success.
 531 *
 532 * Returns 0 on success or a negative error number on failure.
 533 */
 534int nvmem_get_mac_address(struct device *dev, void *addrbuf)
 535{
 536        struct nvmem_cell *cell;
 537        const void *mac;
 538        size_t len;
 539
 540        cell = nvmem_cell_get(dev, "mac-address");
 541        if (IS_ERR(cell))
 542                return PTR_ERR(cell);
 543
 544        mac = nvmem_cell_read(cell, &len);
 545        nvmem_cell_put(cell);
 546
 547        if (IS_ERR(mac))
 548                return PTR_ERR(mac);
 549
 550        if (len != ETH_ALEN || !is_valid_ether_addr(mac)) {
 551                kfree(mac);
 552                return -EINVAL;
 553        }
 554
 555        ether_addr_copy(addrbuf, mac);
 556        kfree(mac);
 557
 558        return 0;
 559}
 560EXPORT_SYMBOL(nvmem_get_mac_address);
 561