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 <linux/property.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 <net/flow_dissector.h>
 62#include <net/gro.h>
 63#include <linux/uaccess.h>
 64#include <net/pkt_sched.h>
 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, const void *data, u32 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	const unsigned short *sap;
158	const struct ethhdr *eth;
159	__be16 res;
160
161	skb->dev = dev;
162	skb_reset_mac_header(skb);
163
164	eth = eth_skb_pull_mac(skb);
165	eth_skb_pkt_type(skb, dev);
166
167	/*
168	 * Some variants of DSA tagging don't have an ethertype field
169	 * at all, so we check here whether one of those tagging
170	 * variants has been configured on the receiving interface,
171	 * and if so, set skb->protocol without looking at the packet.
172	 */
173	if (unlikely(netdev_uses_dsa(dev)))
174		return htons(ETH_P_XDSA);
175
176	if (likely(eth_proto_is_802_3(eth->h_proto)))
177		return eth->h_proto;
178
179	/*
180	 *      This is a magic hack to spot IPX packets. Older Novell breaks
181	 *      the protocol design and runs IPX over 802.3 without an 802.2 LLC
182	 *      layer. We look for FFFF which isn't a used 802.2 SSAP/DSAP. This
183	 *      won't work for fault tolerant netware but does for the rest.
184	 *	We use skb->dev as temporary storage to not hit
185	 *	CONFIG_STACKPROTECTOR_STRONG=y costs on some platforms.
186	 */
187	sap = skb_header_pointer(skb, 0, sizeof(*sap), &skb->dev);
188	res = (sap && *sap == 0xFFFF) ? htons(ETH_P_802_3) : htons(ETH_P_802_2);
189
190	/* restore skb->dev in case it was mangled by skb_header_pointer(). */
191	skb->dev = dev;
192	return res;
193}
194EXPORT_SYMBOL(eth_type_trans);
195
196int eth_header_parse(const struct sk_buff *skb, const struct net_device *dev,
197		     unsigned char *haddr)
198{
199	const struct ethhdr *eth = eth_hdr(skb);
200
201	memcpy(haddr, eth->h_source, ETH_ALEN);
202	return ETH_ALEN;
203}
204EXPORT_SYMBOL(eth_header_parse);
205
206/**
207 * eth_header_cache - fill cache entry from neighbour
208 * @neigh: source neighbour
209 * @hh: destination cache entry
210 * @type: Ethernet type field
211 *
212 * Create an Ethernet header template from the neighbour.
213 */
214int eth_header_cache(const struct neighbour *neigh, struct hh_cache *hh, __be16 type)
215{
216	struct ethhdr *eth;
217	const struct net_device *dev = neigh->dev;
218
219	eth = (struct ethhdr *)
220	    (((u8 *) hh->hh_data) + (HH_DATA_OFF(sizeof(*eth))));
221
222	if (type == htons(ETH_P_802_3))
223		return -1;
224
225	eth->h_proto = type;
226	memcpy(eth->h_source, dev->dev_addr, ETH_ALEN);
227	memcpy(eth->h_dest, neigh->ha, ETH_ALEN);
228
229	/* Pairs with READ_ONCE() in neigh_resolve_output(),
230	 * neigh_hh_output() and neigh_update_hhs().
231	 */
232	smp_store_release(&hh->hh_len, ETH_HLEN);
233
234	return 0;
235}
236EXPORT_SYMBOL(eth_header_cache);
237
238/**
239 * eth_header_cache_update - update cache entry
240 * @hh: destination cache entry
241 * @dev: network device
242 * @haddr: new hardware address
243 *
244 * Called by Address Resolution module to notify changes in address.
245 */
246void eth_header_cache_update(struct hh_cache *hh,
247			     const struct net_device *dev,
248			     const unsigned char *haddr)
249{
250	memcpy(((u8 *) hh->hh_data) + HH_DATA_OFF(sizeof(struct ethhdr)),
251	       haddr, ETH_ALEN);
252}
253EXPORT_SYMBOL(eth_header_cache_update);
254
255/**
256 * eth_header_parse_protocol - extract protocol from L2 header
257 * @skb: packet to extract protocol from
258 */
259__be16 eth_header_parse_protocol(const struct sk_buff *skb)
260{
261	const struct ethhdr *eth = eth_hdr(skb);
262
263	return eth->h_proto;
264}
265EXPORT_SYMBOL(eth_header_parse_protocol);
266
267/**
268 * eth_prepare_mac_addr_change - prepare for mac change
269 * @dev: network device
270 * @p: socket address
271 */
272int eth_prepare_mac_addr_change(struct net_device *dev, void *p)
273{
274	struct sockaddr *addr = p;
275
276	if (!(dev->priv_flags & IFF_LIVE_ADDR_CHANGE) && netif_running(dev))
277		return -EBUSY;
278	if (!is_valid_ether_addr(addr->sa_data))
279		return -EADDRNOTAVAIL;
280	return 0;
281}
282EXPORT_SYMBOL(eth_prepare_mac_addr_change);
283
284/**
285 * eth_commit_mac_addr_change - commit mac change
286 * @dev: network device
287 * @p: socket address
288 */
289void eth_commit_mac_addr_change(struct net_device *dev, void *p)
290{
291	struct sockaddr *addr = p;
292
293	eth_hw_addr_set(dev, addr->sa_data);
294}
295EXPORT_SYMBOL(eth_commit_mac_addr_change);
296
297/**
298 * eth_mac_addr - set new Ethernet hardware address
299 * @dev: network device
300 * @p: socket address
301 *
302 * Change hardware address of device.
303 *
304 * This doesn't change hardware matching, so needs to be overridden
305 * for most real devices.
306 */
307int eth_mac_addr(struct net_device *dev, void *p)
308{
309	int ret;
310
311	ret = eth_prepare_mac_addr_change(dev, p);
312	if (ret < 0)
313		return ret;
314	eth_commit_mac_addr_change(dev, p);
315	return 0;
316}
317EXPORT_SYMBOL(eth_mac_addr);
318
319int eth_validate_addr(struct net_device *dev)
320{
321	if (!is_valid_ether_addr(dev->dev_addr))
322		return -EADDRNOTAVAIL;
323
324	return 0;
325}
326EXPORT_SYMBOL(eth_validate_addr);
327
328const struct header_ops eth_header_ops ____cacheline_aligned = {
329	.create		= eth_header,
330	.parse		= eth_header_parse,
331	.cache		= eth_header_cache,
332	.cache_update	= eth_header_cache_update,
333	.parse_protocol	= eth_header_parse_protocol,
334};
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 */
342void 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->min_header_len	= ETH_HLEN;
348	dev->mtu		= ETH_DATA_LEN;
349	dev->min_mtu		= ETH_MIN_MTU;
350	dev->max_mtu		= ETH_DATA_LEN;
351	dev->addr_len		= ETH_ALEN;
352	dev->tx_queue_len	= DEFAULT_TX_QUEUE_LEN;
353	dev->flags		= IFF_BROADCAST|IFF_MULTICAST;
354	dev->priv_flags		|= IFF_TX_SKB_SHARING;
355
356	eth_broadcast_addr(dev->broadcast);
357
358}
359EXPORT_SYMBOL(ether_setup);
360
361/**
362 * alloc_etherdev_mqs - Allocates and sets up an Ethernet device
363 * @sizeof_priv: Size of additional driver-private structure to be allocated
364 *	for this Ethernet device
365 * @txqs: The number of TX queues this device has.
366 * @rxqs: The number of RX queues this device has.
367 *
368 * Fill in the fields of the device structure with Ethernet-generic
369 * values. Basically does everything except registering the device.
370 *
371 * Constructs a new net device, complete with a private data area of
372 * size (sizeof_priv).  A 32-byte (not bit) alignment is enforced for
373 * this private data area.
374 */
375
376struct net_device *alloc_etherdev_mqs(int sizeof_priv, unsigned int txqs,
377				      unsigned int rxqs)
378{
379	return alloc_netdev_mqs(sizeof_priv, "eth%d", NET_NAME_ENUM,
380				ether_setup, txqs, rxqs);
381}
382EXPORT_SYMBOL(alloc_etherdev_mqs);
383
384ssize_t sysfs_format_mac(char *buf, const unsigned char *addr, int len)
385{
386	return sysfs_emit(buf, "%*phC\n", len, addr);
387}
388EXPORT_SYMBOL(sysfs_format_mac);
389
390struct sk_buff *eth_gro_receive(struct list_head *head, struct sk_buff *skb)
391{
392	const struct packet_offload *ptype;
393	unsigned int hlen, off_eth;
394	struct sk_buff *pp = NULL;
395	struct ethhdr *eh, *eh2;
396	struct sk_buff *p;
397	__be16 type;
398	int flush = 1;
399
400	off_eth = skb_gro_offset(skb);
401	hlen = off_eth + sizeof(*eh);
402	eh = skb_gro_header(skb, hlen, off_eth);
403	if (unlikely(!eh))
404		goto out;
405
406	flush = 0;
407
408	list_for_each_entry(p, head, list) {
409		if (!NAPI_GRO_CB(p)->same_flow)
410			continue;
411
412		eh2 = (struct ethhdr *)(p->data + off_eth);
413		if (compare_ether_header(eh, eh2)) {
414			NAPI_GRO_CB(p)->same_flow = 0;
415			continue;
416		}
417	}
418
419	type = eh->h_proto;
420
421	ptype = gro_find_receive_by_type(type);
422	if (ptype == NULL) {
423		flush = 1;
424		goto out;
425	}
426
427	skb_gro_pull(skb, sizeof(*eh));
428	skb_gro_postpull_rcsum(skb, eh, sizeof(*eh));
429
430	pp = indirect_call_gro_receive_inet(ptype->callbacks.gro_receive,
431					    ipv6_gro_receive, inet_gro_receive,
432					    head, skb);
433
434out:
435	skb_gro_flush_final(skb, pp, flush);
436
437	return pp;
438}
439EXPORT_SYMBOL(eth_gro_receive);
440
441int eth_gro_complete(struct sk_buff *skb, int nhoff)
442{
443	struct ethhdr *eh = (struct ethhdr *)(skb->data + nhoff);
444	__be16 type = eh->h_proto;
445	struct packet_offload *ptype;
446	int err = -ENOSYS;
447
448	if (skb->encapsulation)
449		skb_set_inner_mac_header(skb, nhoff);
450
451	ptype = gro_find_complete_by_type(type);
452	if (ptype != NULL)
453		err = INDIRECT_CALL_INET(ptype->callbacks.gro_complete,
454					 ipv6_gro_complete, inet_gro_complete,
455					 skb, nhoff + sizeof(*eh));
456
457	return err;
458}
459EXPORT_SYMBOL(eth_gro_complete);
460
461static struct packet_offload eth_packet_offload __read_mostly = {
462	.type = cpu_to_be16(ETH_P_TEB),
463	.priority = 10,
464	.callbacks = {
465		.gro_receive = eth_gro_receive,
466		.gro_complete = eth_gro_complete,
467	},
468};
469
470static int __init eth_offload_init(void)
471{
472	dev_add_offload(&eth_packet_offload);
473
474	return 0;
475}
476
477fs_initcall(eth_offload_init);
478
479unsigned char * __weak arch_get_platform_mac_address(void)
480{
481	return NULL;
482}
483
484int eth_platform_get_mac_address(struct device *dev, u8 *mac_addr)
485{
486	unsigned char *addr;
487	int ret;
488
489	ret = of_get_mac_address(dev->of_node, mac_addr);
490	if (!ret)
491		return 0;
492
493	addr = arch_get_platform_mac_address();
494	if (!addr)
495		return -ENODEV;
496
497	ether_addr_copy(mac_addr, addr);
498
499	return 0;
500}
501EXPORT_SYMBOL(eth_platform_get_mac_address);
502
503/**
504 * platform_get_ethdev_address - Set netdev's MAC address from a given device
505 * @dev:	Pointer to the device
506 * @netdev:	Pointer to netdev to write the address to
507 *
508 * Wrapper around eth_platform_get_mac_address() which writes the address
509 * directly to netdev->dev_addr.
510 */
511int platform_get_ethdev_address(struct device *dev, struct net_device *netdev)
512{
513	u8 addr[ETH_ALEN] __aligned(2);
514	int ret;
515
516	ret = eth_platform_get_mac_address(dev, addr);
517	if (!ret)
518		eth_hw_addr_set(netdev, addr);
519	return ret;
520}
521EXPORT_SYMBOL(platform_get_ethdev_address);
522
523/**
524 * nvmem_get_mac_address - Obtain the MAC address from an nvmem cell named
525 * 'mac-address' associated with given device.
526 *
527 * @dev:	Device with which the mac-address cell is associated.
528 * @addrbuf:	Buffer to which the MAC address will be copied on success.
529 *
530 * Returns 0 on success or a negative error number on failure.
531 */
532int nvmem_get_mac_address(struct device *dev, void *addrbuf)
533{
534	struct nvmem_cell *cell;
535	const void *mac;
536	size_t len;
537
538	cell = nvmem_cell_get(dev, "mac-address");
539	if (IS_ERR(cell))
540		return PTR_ERR(cell);
541
542	mac = nvmem_cell_read(cell, &len);
543	nvmem_cell_put(cell);
544
545	if (IS_ERR(mac))
546		return PTR_ERR(mac);
547
548	if (len != ETH_ALEN || !is_valid_ether_addr(mac)) {
549		kfree(mac);
550		return -EINVAL;
551	}
552
553	ether_addr_copy(addrbuf, mac);
554	kfree(mac);
555
556	return 0;
557}
558
559static int fwnode_get_mac_addr(struct fwnode_handle *fwnode,
560			       const char *name, char *addr)
561{
562	int ret;
563
564	ret = fwnode_property_read_u8_array(fwnode, name, addr, ETH_ALEN);
565	if (ret)
566		return ret;
567
568	if (!is_valid_ether_addr(addr))
569		return -EINVAL;
570	return 0;
571}
572
573/**
574 * fwnode_get_mac_address - Get the MAC from the firmware node
575 * @fwnode:	Pointer to the firmware node
576 * @addr:	Address of buffer to store the MAC in
577 *
578 * Search the firmware node for the best MAC address to use.  'mac-address' is
579 * checked first, because that is supposed to contain to "most recent" MAC
580 * address. If that isn't set, then 'local-mac-address' is checked next,
581 * because that is the default address.  If that isn't set, then the obsolete
582 * 'address' is checked, just in case we're using an old device tree.
583 *
584 * Note that the 'address' property is supposed to contain a virtual address of
585 * the register set, but some DTS files have redefined that property to be the
586 * MAC address.
587 *
588 * All-zero MAC addresses are rejected, because those could be properties that
589 * exist in the firmware tables, but were not updated by the firmware.  For
590 * example, the DTS could define 'mac-address' and 'local-mac-address', with
591 * zero MAC addresses.  Some older U-Boots only initialized 'local-mac-address'.
592 * In this case, the real MAC is in 'local-mac-address', and 'mac-address'
593 * exists but is all zeros.
594 */
595int fwnode_get_mac_address(struct fwnode_handle *fwnode, char *addr)
596{
597	if (!fwnode_get_mac_addr(fwnode, "mac-address", addr) ||
598	    !fwnode_get_mac_addr(fwnode, "local-mac-address", addr) ||
599	    !fwnode_get_mac_addr(fwnode, "address", addr))
600		return 0;
601
602	return -ENOENT;
603}
604EXPORT_SYMBOL(fwnode_get_mac_address);
605
606/**
607 * device_get_mac_address - Get the MAC for a given device
608 * @dev:	Pointer to the device
609 * @addr:	Address of buffer to store the MAC in
610 */
611int device_get_mac_address(struct device *dev, char *addr)
612{
613	if (!fwnode_get_mac_address(dev_fwnode(dev), addr))
614		return 0;
615
616	return nvmem_get_mac_address(dev, addr);
617}
618EXPORT_SYMBOL(device_get_mac_address);
619
620/**
621 * device_get_ethdev_address - Set netdev's MAC address from a given device
622 * @dev:	Pointer to the device
623 * @netdev:	Pointer to netdev to write the address to
624 *
625 * Wrapper around device_get_mac_address() which writes the address
626 * directly to netdev->dev_addr.
627 */
628int device_get_ethdev_address(struct device *dev, struct net_device *netdev)
629{
630	u8 addr[ETH_ALEN];
631	int ret;
632
633	ret = device_get_mac_address(dev, addr);
634	if (!ret)
635		eth_hw_addr_set(netdev, addr);
636	return ret;
637}
638EXPORT_SYMBOL(device_get_ethdev_address);