dev.c 191.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9
/*
 * 	NET3	Protocol independent device support routines.
 *
 *		This program is free software; you can redistribute it and/or
 *		modify it under the terms of the GNU General Public License
 *		as published by the Free Software Foundation; either version
 *		2 of the License, or (at your option) any later version.
 *
 *	Derived from the non IP parts of dev.c 1.0.19
10
 * 		Authors:	Ross Biro
L
Linus Torvalds 已提交
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76
 *				Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 *				Mark Evans, <evansmp@uhura.aston.ac.uk>
 *
 *	Additional Authors:
 *		Florian la Roche <rzsfl@rz.uni-sb.de>
 *		Alan Cox <gw4pts@gw4pts.ampr.org>
 *		David Hinds <dahinds@users.sourceforge.net>
 *		Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
 *		Adam Sulmicki <adam@cfar.umd.edu>
 *              Pekka Riikonen <priikone@poesidon.pspt.fi>
 *
 *	Changes:
 *              D.J. Barrow     :       Fixed bug where dev->refcnt gets set
 *              			to 2 if register_netdev gets called
 *              			before net_dev_init & also removed a
 *              			few lines of code in the process.
 *		Alan Cox	:	device private ioctl copies fields back.
 *		Alan Cox	:	Transmit queue code does relevant
 *					stunts to keep the queue safe.
 *		Alan Cox	:	Fixed double lock.
 *		Alan Cox	:	Fixed promisc NULL pointer trap
 *		????????	:	Support the full private ioctl range
 *		Alan Cox	:	Moved ioctl permission check into
 *					drivers
 *		Tim Kordas	:	SIOCADDMULTI/SIOCDELMULTI
 *		Alan Cox	:	100 backlog just doesn't cut it when
 *					you start doing multicast video 8)
 *		Alan Cox	:	Rewrote net_bh and list manager.
 *		Alan Cox	: 	Fix ETH_P_ALL echoback lengths.
 *		Alan Cox	:	Took out transmit every packet pass
 *					Saved a few bytes in the ioctl handler
 *		Alan Cox	:	Network driver sets packet type before
 *					calling netif_rx. Saves a function
 *					call a packet.
 *		Alan Cox	:	Hashed net_bh()
 *		Richard Kooijman:	Timestamp fixes.
 *		Alan Cox	:	Wrong field in SIOCGIFDSTADDR
 *		Alan Cox	:	Device lock protection.
 *		Alan Cox	: 	Fixed nasty side effect of device close
 *					changes.
 *		Rudi Cilibrasi	:	Pass the right thing to
 *					set_mac_address()
 *		Dave Miller	:	32bit quantity for the device lock to
 *					make it work out on a Sparc.
 *		Bjorn Ekwall	:	Added KERNELD hack.
 *		Alan Cox	:	Cleaned up the backlog initialise.
 *		Craig Metz	:	SIOCGIFCONF fix if space for under
 *					1 device.
 *	    Thomas Bogendoerfer :	Return ENODEV for dev_open, if there
 *					is no device open function.
 *		Andi Kleen	:	Fix error reporting for SIOCGIFCONF
 *	    Michael Chastain	:	Fix signed/unsigned for SIOCGIFCONF
 *		Cyrus Durgin	:	Cleaned for KMOD
 *		Adam Sulmicki   :	Bug Fix : Network Device Unload
 *					A network device unload needs to purge
 *					the backlog queue.
 *	Paul Rusty Russell	:	SIOCSIFNAME
 *              Pekka Riikonen  :	Netdev boot-time settings code
 *              Andrew Morton   :       Make unregister_netdevice wait
 *              			indefinitely on dev->refcnt
 * 		J Hadi Salim	:	- Backlog queue sampling
 *				        - netif_rx() feedback
 */

#include <asm/uaccess.h>
#include <linux/bitops.h>
77
#include <linux/capability.h>
L
Linus Torvalds 已提交
78 79 80
#include <linux/cpu.h>
#include <linux/types.h>
#include <linux/kernel.h>
81
#include <linux/hash.h>
82
#include <linux/slab.h>
L
Linus Torvalds 已提交
83
#include <linux/sched.h>
A
Arjan van de Ven 已提交
84
#include <linux/mutex.h>
L
Linus Torvalds 已提交
85 86 87 88 89 90 91 92 93
#include <linux/string.h>
#include <linux/mm.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/errno.h>
#include <linux/interrupt.h>
#include <linux/if_ether.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
94
#include <linux/ethtool.h>
L
Linus Torvalds 已提交
95 96
#include <linux/notifier.h>
#include <linux/skbuff.h>
97
#include <net/net_namespace.h>
L
Linus Torvalds 已提交
98 99 100 101 102 103
#include <net/sock.h>
#include <linux/rtnetlink.h>
#include <linux/stat.h>
#include <net/dst.h>
#include <net/pkt_sched.h>
#include <net/checksum.h>
104
#include <net/xfrm.h>
L
Linus Torvalds 已提交
105 106 107 108 109 110 111 112
#include <linux/highmem.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/netpoll.h>
#include <linux/rcupdate.h>
#include <linux/delay.h>
#include <net/iw_handler.h>
#include <asm/current.h>
S
Steve Grubb 已提交
113
#include <linux/audit.h>
114
#include <linux/dmaengine.h>
115
#include <linux/err.h>
116
#include <linux/ctype.h>
117
#include <linux/if_arp.h>
118
#include <linux/if_vlan.h>
119
#include <linux/ip.h>
120
#include <net/ip.h>
121
#include <net/mpls.h>
122 123
#include <linux/ipv6.h>
#include <linux/in.h>
D
David S. Miller 已提交
124 125
#include <linux/jhash.h>
#include <linux/random.h>
126
#include <trace/events/napi.h>
127
#include <trace/events/net.h>
128
#include <trace/events/skb.h>
129
#include <linux/pci.h>
130
#include <linux/inetdevice.h>
131
#include <linux/cpu_rmap.h>
132
#include <linux/static_key.h>
E
Eliezer Tamir 已提交
133
#include <linux/hashtable.h>
134
#include <linux/vmalloc.h>
135
#include <linux/if_macvlan.h>
136
#include <linux/errqueue.h>
137
#include <linux/hrtimer.h>
138
#include <linux/netfilter_ingress.h>
L
Linus Torvalds 已提交
139

140 141
#include "net-sysfs.h"

142 143 144
/* Instead of increasing this, you should create a hash table. */
#define MAX_GRO_SKBS 8

H
Herbert Xu 已提交
145 146 147
/* This should be increased if a protocol with a bigger head is added. */
#define GRO_MAX_HEAD (MAX_HEADER + 128)

L
Linus Torvalds 已提交
148
static DEFINE_SPINLOCK(ptype_lock);
149
static DEFINE_SPINLOCK(offload_lock);
150 151
struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
struct list_head ptype_all __read_mostly;	/* Taps */
152
static struct list_head offload_base __read_mostly;
L
Linus Torvalds 已提交
153

154
static int netif_rx_internal(struct sk_buff *skb);
155 156 157
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info);
158

L
Linus Torvalds 已提交
159
/*
160
 * The @dev_base_head list is protected by @dev_base_lock and the rtnl
L
Linus Torvalds 已提交
161 162
 * semaphore.
 *
163
 * Pure readers hold dev_base_lock for reading, or rcu_read_lock()
L
Linus Torvalds 已提交
164 165
 *
 * Writers must hold the rtnl semaphore while they loop through the
166
 * dev_base_head list, and hold dev_base_lock for writing when they do the
L
Linus Torvalds 已提交
167 168 169 170 171 172 173 174 175 176 177 178 179 180
 * actual updates.  This allows pure readers to access the list even
 * while a writer is preparing to update it.
 *
 * To put it another way, dev_base_lock is held for writing only to
 * protect against pure readers; the rtnl semaphore provides the
 * protection against other writers.
 *
 * See, for example usages, register_netdevice() and
 * unregister_netdevice(), which must be called with the rtnl
 * semaphore held.
 */
DEFINE_RWLOCK(dev_base_lock);
EXPORT_SYMBOL(dev_base_lock);

E
Eliezer Tamir 已提交
181 182 183 184 185 186
/* protects napi_hash addition/deletion and napi_gen_id */
static DEFINE_SPINLOCK(napi_hash_lock);

static unsigned int napi_gen_id;
static DEFINE_HASHTABLE(napi_hash, 8);

187
static seqcount_t devnet_rename_seq;
188

189 190 191 192 193
static inline void dev_base_seq_inc(struct net *net)
{
	while (++net->dev_base_seq == 0);
}

194
static inline struct hlist_head *dev_name_hash(struct net *net, const char *name)
L
Linus Torvalds 已提交
195
{
196 197
	unsigned int hash = full_name_hash(name, strnlen(name, IFNAMSIZ));

198
	return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)];
L
Linus Torvalds 已提交
199 200
}

201
static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex)
L
Linus Torvalds 已提交
202
{
203
	return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)];
L
Linus Torvalds 已提交
204 205
}

E
Eric Dumazet 已提交
206
static inline void rps_lock(struct softnet_data *sd)
207 208
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
209
	spin_lock(&sd->input_pkt_queue.lock);
210 211 212
#endif
}

E
Eric Dumazet 已提交
213
static inline void rps_unlock(struct softnet_data *sd)
214 215
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
216
	spin_unlock(&sd->input_pkt_queue.lock);
217 218 219
#endif
}

220
/* Device list insertion */
221
static void list_netdevice(struct net_device *dev)
222
{
223
	struct net *net = dev_net(dev);
224 225 226 227

	ASSERT_RTNL();

	write_lock_bh(&dev_base_lock);
228
	list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
229
	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
230 231
	hlist_add_head_rcu(&dev->index_hlist,
			   dev_index_hash(net, dev->ifindex));
232
	write_unlock_bh(&dev_base_lock);
233 234

	dev_base_seq_inc(net);
235 236
}

237 238 239
/* Device list removal
 * caller must respect a RCU grace period before freeing/reusing dev
 */
240 241 242 243 244 245
static void unlist_netdevice(struct net_device *dev)
{
	ASSERT_RTNL();

	/* Unlink dev from the device chain */
	write_lock_bh(&dev_base_lock);
246
	list_del_rcu(&dev->dev_list);
247
	hlist_del_rcu(&dev->name_hlist);
248
	hlist_del_rcu(&dev->index_hlist);
249
	write_unlock_bh(&dev_base_lock);
250 251

	dev_base_seq_inc(dev_net(dev));
252 253
}

L
Linus Torvalds 已提交
254 255 256 257
/*
 *	Our notifier list
 */

258
static RAW_NOTIFIER_HEAD(netdev_chain);
L
Linus Torvalds 已提交
259 260 261 262 263

/*
 *	Device drivers call our routines to queue packets here. We empty the
 *	queue in the local softnet handler.
 */
264

265
DEFINE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
E
Eric Dumazet 已提交
266
EXPORT_PER_CPU_SYMBOL(softnet_data);
L
Linus Torvalds 已提交
267

268
#ifdef CONFIG_LOCKDEP
269
/*
270
 * register_netdevice() inits txq->_xmit_lock and sets lockdep class
271 272 273 274 275 276 277 278 279 280 281 282 283 284 285
 * according to dev->type
 */
static const unsigned short netdev_lock_type[] =
	{ARPHRD_NETROM, ARPHRD_ETHER, ARPHRD_EETHER, ARPHRD_AX25,
	 ARPHRD_PRONET, ARPHRD_CHAOS, ARPHRD_IEEE802, ARPHRD_ARCNET,
	 ARPHRD_APPLETLK, ARPHRD_DLCI, ARPHRD_ATM, ARPHRD_METRICOM,
	 ARPHRD_IEEE1394, ARPHRD_EUI64, ARPHRD_INFINIBAND, ARPHRD_SLIP,
	 ARPHRD_CSLIP, ARPHRD_SLIP6, ARPHRD_CSLIP6, ARPHRD_RSRVD,
	 ARPHRD_ADAPT, ARPHRD_ROSE, ARPHRD_X25, ARPHRD_HWX25,
	 ARPHRD_PPP, ARPHRD_CISCO, ARPHRD_LAPB, ARPHRD_DDCMP,
	 ARPHRD_RAWHDLC, ARPHRD_TUNNEL, ARPHRD_TUNNEL6, ARPHRD_FRAD,
	 ARPHRD_SKIP, ARPHRD_LOOPBACK, ARPHRD_LOCALTLK, ARPHRD_FDDI,
	 ARPHRD_BIF, ARPHRD_SIT, ARPHRD_IPDDP, ARPHRD_IPGRE,
	 ARPHRD_PIMREG, ARPHRD_HIPPI, ARPHRD_ASH, ARPHRD_ECONET,
	 ARPHRD_IRDA, ARPHRD_FCPP, ARPHRD_FCAL, ARPHRD_FCPL,
286 287 288
	 ARPHRD_FCFABRIC, ARPHRD_IEEE80211, ARPHRD_IEEE80211_PRISM,
	 ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, ARPHRD_PHONET_PIPE,
	 ARPHRD_IEEE802154, ARPHRD_VOID, ARPHRD_NONE};
289

290
static const char *const netdev_lock_name[] =
291 292 293 294 295 296 297 298 299 300 301 302
	{"_xmit_NETROM", "_xmit_ETHER", "_xmit_EETHER", "_xmit_AX25",
	 "_xmit_PRONET", "_xmit_CHAOS", "_xmit_IEEE802", "_xmit_ARCNET",
	 "_xmit_APPLETLK", "_xmit_DLCI", "_xmit_ATM", "_xmit_METRICOM",
	 "_xmit_IEEE1394", "_xmit_EUI64", "_xmit_INFINIBAND", "_xmit_SLIP",
	 "_xmit_CSLIP", "_xmit_SLIP6", "_xmit_CSLIP6", "_xmit_RSRVD",
	 "_xmit_ADAPT", "_xmit_ROSE", "_xmit_X25", "_xmit_HWX25",
	 "_xmit_PPP", "_xmit_CISCO", "_xmit_LAPB", "_xmit_DDCMP",
	 "_xmit_RAWHDLC", "_xmit_TUNNEL", "_xmit_TUNNEL6", "_xmit_FRAD",
	 "_xmit_SKIP", "_xmit_LOOPBACK", "_xmit_LOCALTLK", "_xmit_FDDI",
	 "_xmit_BIF", "_xmit_SIT", "_xmit_IPDDP", "_xmit_IPGRE",
	 "_xmit_PIMREG", "_xmit_HIPPI", "_xmit_ASH", "_xmit_ECONET",
	 "_xmit_IRDA", "_xmit_FCPP", "_xmit_FCAL", "_xmit_FCPL",
303 304 305
	 "_xmit_FCFABRIC", "_xmit_IEEE80211", "_xmit_IEEE80211_PRISM",
	 "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", "_xmit_PHONET_PIPE",
	 "_xmit_IEEE802154", "_xmit_VOID", "_xmit_NONE"};
306 307

static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)];
308
static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)];
309 310 311 312 313 314 315 316 317 318 319 320

static inline unsigned short netdev_lock_pos(unsigned short dev_type)
{
	int i;

	for (i = 0; i < ARRAY_SIZE(netdev_lock_type); i++)
		if (netdev_lock_type[i] == dev_type)
			return i;
	/* the last key is used by default */
	return ARRAY_SIZE(netdev_lock_type) - 1;
}

321 322
static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
						 unsigned short dev_type)
323 324 325 326 327 328 329
{
	int i;

	i = netdev_lock_pos(dev_type);
	lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i],
				   netdev_lock_name[i]);
}
330 331 332 333 334 335 336 337 338 339

static inline void netdev_set_addr_lockdep_class(struct net_device *dev)
{
	int i;

	i = netdev_lock_pos(dev->type);
	lockdep_set_class_and_name(&dev->addr_list_lock,
				   &netdev_addr_lock_key[i],
				   netdev_lock_name[i]);
}
340
#else
341 342 343 344 345
static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
						 unsigned short dev_type)
{
}
static inline void netdev_set_addr_lockdep_class(struct net_device *dev)
346 347 348
{
}
#endif
L
Linus Torvalds 已提交
349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371

/*******************************************************************************

		Protocol management and registration routines

*******************************************************************************/

/*
 *	Add a protocol ID to the list. Now that the input handler is
 *	smarter we can dispense with all the messy stuff that used to be
 *	here.
 *
 *	BEWARE!!! Protocol handlers, mangling input packets,
 *	MUST BE last in hash buckets and checking protocol handlers
 *	MUST start from promiscuous ptype_all chain in net_bh.
 *	It is true now, do not change it.
 *	Explanation follows: if protocol handler, mangling packet, will
 *	be the first on list, it is not able to sense, that packet
 *	is cloned and should be copied-on-write, so that it will
 *	change it and subsequent readers will get broken packet.
 *							--ANK (980803)
 */

372 373 374
static inline struct list_head *ptype_head(const struct packet_type *pt)
{
	if (pt->type == htons(ETH_P_ALL))
375
		return pt->dev ? &pt->dev->ptype_all : &ptype_all;
376
	else
377 378
		return pt->dev ? &pt->dev->ptype_specific :
				 &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK];
379 380
}

L
Linus Torvalds 已提交
381 382 383 384 385 386 387 388
/**
 *	dev_add_pack - add packet handler
 *	@pt: packet type declaration
 *
 *	Add a protocol handler to the networking stack. The passed &packet_type
 *	is linked into kernel lists and may not be freed until it has been
 *	removed from the kernel lists.
 *
389
 *	This call does not sleep therefore it can not
L
Linus Torvalds 已提交
390 391 392 393 394 395
 *	guarantee all CPU's that are in middle of receiving packets
 *	will see the new packet type (until the next received packet).
 */

void dev_add_pack(struct packet_type *pt)
{
396
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
397

398 399 400
	spin_lock(&ptype_lock);
	list_add_rcu(&pt->list, head);
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
401
}
E
Eric Dumazet 已提交
402
EXPORT_SYMBOL(dev_add_pack);
L
Linus Torvalds 已提交
403 404 405 406 407 408 409 410

/**
 *	__dev_remove_pack	 - remove packet handler
 *	@pt: packet type declaration
 *
 *	Remove a protocol handler that was previously added to the kernel
 *	protocol handlers by dev_add_pack(). The passed &packet_type is removed
 *	from the kernel lists and can be freed or reused once this function
411
 *	returns.
L
Linus Torvalds 已提交
412 413 414 415 416 417 418
 *
 *      The packet type might still be in use by receivers
 *	and must not be freed until after all the CPU's have gone
 *	through a quiescent state.
 */
void __dev_remove_pack(struct packet_type *pt)
{
419
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
420 421
	struct packet_type *pt1;

422
	spin_lock(&ptype_lock);
L
Linus Torvalds 已提交
423 424 425 426 427 428 429 430

	list_for_each_entry(pt1, head, list) {
		if (pt == pt1) {
			list_del_rcu(&pt->list);
			goto out;
		}
	}

431
	pr_warn("dev_remove_pack: %p not found\n", pt);
L
Linus Torvalds 已提交
432
out:
433
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
434
}
E
Eric Dumazet 已提交
435 436
EXPORT_SYMBOL(__dev_remove_pack);

L
Linus Torvalds 已提交
437 438 439 440 441 442 443 444 445 446 447 448 449 450 451
/**
 *	dev_remove_pack	 - remove packet handler
 *	@pt: packet type declaration
 *
 *	Remove a protocol handler that was previously added to the kernel
 *	protocol handlers by dev_add_pack(). The passed &packet_type is removed
 *	from the kernel lists and can be freed or reused once this function
 *	returns.
 *
 *	This call sleeps to guarantee that no CPU is looking at the packet
 *	type after return.
 */
void dev_remove_pack(struct packet_type *pt)
{
	__dev_remove_pack(pt);
452

L
Linus Torvalds 已提交
453 454
	synchronize_net();
}
E
Eric Dumazet 已提交
455
EXPORT_SYMBOL(dev_remove_pack);
L
Linus Torvalds 已提交
456

457 458 459 460 461 462 463 464 465 466 467 468 469 470 471

/**
 *	dev_add_offload - register offload handlers
 *	@po: protocol offload declaration
 *
 *	Add protocol offload handlers to the networking stack. The passed
 *	&proto_offload is linked into kernel lists and may not be freed until
 *	it has been removed from the kernel lists.
 *
 *	This call does not sleep therefore it can not
 *	guarantee all CPU's that are in middle of receiving packets
 *	will see the new offload handlers (until the next received packet).
 */
void dev_add_offload(struct packet_offload *po)
{
472
	struct packet_offload *elem;
473 474

	spin_lock(&offload_lock);
475 476 477 478 479
	list_for_each_entry(elem, &offload_base, list) {
		if (po->priority < elem->priority)
			break;
	}
	list_add_rcu(&po->list, elem->list.prev);
480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496
	spin_unlock(&offload_lock);
}
EXPORT_SYMBOL(dev_add_offload);

/**
 *	__dev_remove_offload	 - remove offload handler
 *	@po: packet offload declaration
 *
 *	Remove a protocol offload handler that was previously added to the
 *	kernel offload handlers by dev_add_offload(). The passed &offload_type
 *	is removed from the kernel lists and can be freed or reused once this
 *	function returns.
 *
 *      The packet type might still be in use by receivers
 *	and must not be freed until after all the CPU's have gone
 *	through a quiescent state.
 */
S
stephen hemminger 已提交
497
static void __dev_remove_offload(struct packet_offload *po)
498 499 500 501
{
	struct list_head *head = &offload_base;
	struct packet_offload *po1;

502
	spin_lock(&offload_lock);
503 504 505 506 507 508 509 510 511 512

	list_for_each_entry(po1, head, list) {
		if (po == po1) {
			list_del_rcu(&po->list);
			goto out;
		}
	}

	pr_warn("dev_remove_offload: %p not found\n", po);
out:
513
	spin_unlock(&offload_lock);
514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535
}

/**
 *	dev_remove_offload	 - remove packet offload handler
 *	@po: packet offload declaration
 *
 *	Remove a packet offload handler that was previously added to the kernel
 *	offload handlers by dev_add_offload(). The passed &offload_type is
 *	removed from the kernel lists and can be freed or reused once this
 *	function returns.
 *
 *	This call sleeps to guarantee that no CPU is looking at the packet
 *	type after return.
 */
void dev_remove_offload(struct packet_offload *po)
{
	__dev_remove_offload(po);

	synchronize_net();
}
EXPORT_SYMBOL(dev_remove_offload);

L
Linus Torvalds 已提交
536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562
/******************************************************************************

		      Device Boot-time Settings Routines

*******************************************************************************/

/* Boot time configuration table */
static struct netdev_boot_setup dev_boot_setup[NETDEV_BOOT_SETUP_MAX];

/**
 *	netdev_boot_setup_add	- add new setup entry
 *	@name: name of the device
 *	@map: configured settings for the device
 *
 *	Adds new setup entry to the dev_boot_setup list.  The function
 *	returns 0 on error and 1 on success.  This is a generic routine to
 *	all netdevices.
 */
static int netdev_boot_setup_add(char *name, struct ifmap *map)
{
	struct netdev_boot_setup *s;
	int i;

	s = dev_boot_setup;
	for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) {
		if (s[i].name[0] == '\0' || s[i].name[0] == ' ') {
			memset(s[i].name, 0, sizeof(s[i].name));
563
			strlcpy(s[i].name, name, IFNAMSIZ);
L
Linus Torvalds 已提交
564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587
			memcpy(&s[i].map, map, sizeof(s[i].map));
			break;
		}
	}

	return i >= NETDEV_BOOT_SETUP_MAX ? 0 : 1;
}

/**
 *	netdev_boot_setup_check	- check boot time settings
 *	@dev: the netdevice
 *
 * 	Check boot time settings for the device.
 *	The found settings are set for the device to be used
 *	later in the device probing.
 *	Returns 0 if no settings found, 1 if they are.
 */
int netdev_boot_setup_check(struct net_device *dev)
{
	struct netdev_boot_setup *s = dev_boot_setup;
	int i;

	for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) {
		if (s[i].name[0] != '\0' && s[i].name[0] != ' ' &&
588
		    !strcmp(dev->name, s[i].name)) {
L
Linus Torvalds 已提交
589 590 591 592 593 594 595 596 597
			dev->irq 	= s[i].map.irq;
			dev->base_addr 	= s[i].map.base_addr;
			dev->mem_start 	= s[i].map.mem_start;
			dev->mem_end 	= s[i].map.mem_end;
			return 1;
		}
	}
	return 0;
}
E
Eric Dumazet 已提交
598
EXPORT_SYMBOL(netdev_boot_setup_check);
L
Linus Torvalds 已提交
599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622


/**
 *	netdev_boot_base	- get address from boot time settings
 *	@prefix: prefix for network device
 *	@unit: id for network device
 *
 * 	Check boot time settings for the base address of device.
 *	The found settings are set for the device to be used
 *	later in the device probing.
 *	Returns 0 if no settings found.
 */
unsigned long netdev_boot_base(const char *prefix, int unit)
{
	const struct netdev_boot_setup *s = dev_boot_setup;
	char name[IFNAMSIZ];
	int i;

	sprintf(name, "%s%d", prefix, unit);

	/*
	 * If device already registered then return base of 1
	 * to indicate not to probe for this interface
	 */
623
	if (__dev_get_by_name(&init_net, name))
L
Linus Torvalds 已提交
624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666
		return 1;

	for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++)
		if (!strcmp(name, s[i].name))
			return s[i].map.base_addr;
	return 0;
}

/*
 * Saves at boot time configured settings for any netdevice.
 */
int __init netdev_boot_setup(char *str)
{
	int ints[5];
	struct ifmap map;

	str = get_options(str, ARRAY_SIZE(ints), ints);
	if (!str || !*str)
		return 0;

	/* Save settings */
	memset(&map, 0, sizeof(map));
	if (ints[0] > 0)
		map.irq = ints[1];
	if (ints[0] > 1)
		map.base_addr = ints[2];
	if (ints[0] > 2)
		map.mem_start = ints[3];
	if (ints[0] > 3)
		map.mem_end = ints[4];

	/* Add new entry to the list */
	return netdev_boot_setup_add(str, &map);
}

__setup("netdev=", netdev_boot_setup);

/*******************************************************************************

			    Device Interface Subroutines

*******************************************************************************/

667 668 669 670 671 672 673 674 675 676 677 678 679
/**
 *	dev_get_iflink	- get 'iflink' value of a interface
 *	@dev: targeted interface
 *
 *	Indicates the ifindex the interface is linked to.
 *	Physical interfaces have the same 'ifindex' and 'iflink' values.
 */

int dev_get_iflink(const struct net_device *dev)
{
	if (dev->netdev_ops && dev->netdev_ops->ndo_get_iflink)
		return dev->netdev_ops->ndo_get_iflink(dev);

680
	return dev->ifindex;
681 682 683
}
EXPORT_SYMBOL(dev_get_iflink);

L
Linus Torvalds 已提交
684 685
/**
 *	__dev_get_by_name	- find a device by its name
686
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
687 688 689 690 691 692 693 694 695
 *	@name: name to find
 *
 *	Find an interface by name. Must be called under RTNL semaphore
 *	or @dev_base_lock. If the name is found a pointer to the device
 *	is returned. If the name is not found then %NULL is returned. The
 *	reference counters are not incremented so the caller must be
 *	careful with locks.
 */

696
struct net_device *__dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
697
{
E
Eric Dumazet 已提交
698 699
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);
L
Linus Torvalds 已提交
700

701
	hlist_for_each_entry(dev, head, name_hlist)
L
Linus Torvalds 已提交
702 703
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;
E
Eric Dumazet 已提交
704

L
Linus Torvalds 已提交
705 706
	return NULL;
}
E
Eric Dumazet 已提交
707
EXPORT_SYMBOL(__dev_get_by_name);
L
Linus Torvalds 已提交
708

709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725
/**
 *	dev_get_by_name_rcu	- find a device by its name
 *	@net: the applicable net namespace
 *	@name: name to find
 *
 *	Find an interface by name.
 *	If the name is found a pointer to the device is returned.
 * 	If the name is not found then %NULL is returned.
 *	The reference counters are not incremented so the caller must be
 *	careful with locks. The caller must hold RCU lock.
 */

struct net_device *dev_get_by_name_rcu(struct net *net, const char *name)
{
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);

726
	hlist_for_each_entry_rcu(dev, head, name_hlist)
727 728 729 730 731 732 733
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_name_rcu);

L
Linus Torvalds 已提交
734 735
/**
 *	dev_get_by_name		- find a device by its name
736
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
737 738 739 740 741 742 743 744 745
 *	@name: name to find
 *
 *	Find an interface by name. This can be called from any
 *	context and does its own locking. The returned handle has
 *	the usage count incremented and the caller must use dev_put() to
 *	release it when it is no longer needed. %NULL is returned if no
 *	matching device is found.
 */

746
struct net_device *dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
747 748 749
{
	struct net_device *dev;

750 751
	rcu_read_lock();
	dev = dev_get_by_name_rcu(net, name);
L
Linus Torvalds 已提交
752 753
	if (dev)
		dev_hold(dev);
754
	rcu_read_unlock();
L
Linus Torvalds 已提交
755 756
	return dev;
}
E
Eric Dumazet 已提交
757
EXPORT_SYMBOL(dev_get_by_name);
L
Linus Torvalds 已提交
758 759 760

/**
 *	__dev_get_by_index - find a device by its ifindex
761
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
762 763 764 765 766 767 768 769 770
 *	@ifindex: index of device
 *
 *	Search for an interface by index. Returns %NULL if the device
 *	is not found or a pointer to the device. The device has not
 *	had its reference counter increased so the caller must be careful
 *	about locking. The caller must hold either the RTNL semaphore
 *	or @dev_base_lock.
 */

771
struct net_device *__dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
772
{
E
Eric Dumazet 已提交
773 774
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);
L
Linus Torvalds 已提交
775

776
	hlist_for_each_entry(dev, head, index_hlist)
L
Linus Torvalds 已提交
777 778
		if (dev->ifindex == ifindex)
			return dev;
E
Eric Dumazet 已提交
779

L
Linus Torvalds 已提交
780 781
	return NULL;
}
E
Eric Dumazet 已提交
782
EXPORT_SYMBOL(__dev_get_by_index);
L
Linus Torvalds 已提交
783

784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799
/**
 *	dev_get_by_index_rcu - find a device by its ifindex
 *	@net: the applicable net namespace
 *	@ifindex: index of device
 *
 *	Search for an interface by index. Returns %NULL if the device
 *	is not found or a pointer to the device. The device has not
 *	had its reference counter increased so the caller must be careful
 *	about locking. The caller must hold RCU lock.
 */

struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex)
{
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);

800
	hlist_for_each_entry_rcu(dev, head, index_hlist)
801 802 803 804 805 806 807
		if (dev->ifindex == ifindex)
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_index_rcu);

L
Linus Torvalds 已提交
808 809 810

/**
 *	dev_get_by_index - find a device by its ifindex
811
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
812 813 814 815 816 817 818 819
 *	@ifindex: index of device
 *
 *	Search for an interface by index. Returns NULL if the device
 *	is not found or a pointer to the device. The device returned has
 *	had a reference added and the pointer is safe until the user calls
 *	dev_put to indicate they have finished with it.
 */

820
struct net_device *dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
821 822 823
{
	struct net_device *dev;

824 825
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, ifindex);
L
Linus Torvalds 已提交
826 827
	if (dev)
		dev_hold(dev);
828
	rcu_read_unlock();
L
Linus Torvalds 已提交
829 830
	return dev;
}
E
Eric Dumazet 已提交
831
EXPORT_SYMBOL(dev_get_by_index);
L
Linus Torvalds 已提交
832

833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866
/**
 *	netdev_get_name - get a netdevice name, knowing its ifindex.
 *	@net: network namespace
 *	@name: a pointer to the buffer where the name will be stored.
 *	@ifindex: the ifindex of the interface to get the name from.
 *
 *	The use of raw_seqcount_begin() and cond_resched() before
 *	retrying is required as we want to give the writers a chance
 *	to complete when CONFIG_PREEMPT is not set.
 */
int netdev_get_name(struct net *net, char *name, int ifindex)
{
	struct net_device *dev;
	unsigned int seq;

retry:
	seq = raw_seqcount_begin(&devnet_rename_seq);
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, ifindex);
	if (!dev) {
		rcu_read_unlock();
		return -ENODEV;
	}

	strcpy(name, dev->name);
	rcu_read_unlock();
	if (read_seqcount_retry(&devnet_rename_seq, seq)) {
		cond_resched();
		goto retry;
	}

	return 0;
}

L
Linus Torvalds 已提交
867
/**
868
 *	dev_getbyhwaddr_rcu - find a device by its hardware address
869
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
870 871 872 873
 *	@type: media type of device
 *	@ha: hardware address
 *
 *	Search for an interface by MAC address. Returns NULL if the device
E
Eric Dumazet 已提交
874 875
 *	is not found or a pointer to the device.
 *	The caller must hold RCU or RTNL.
876
 *	The returned device has not had its ref count increased
L
Linus Torvalds 已提交
877 878 879 880
 *	and the caller must therefore be careful about locking
 *
 */

881 882
struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
				       const char *ha)
L
Linus Torvalds 已提交
883 884 885
{
	struct net_device *dev;

886
	for_each_netdev_rcu(net, dev)
L
Linus Torvalds 已提交
887 888
		if (dev->type == type &&
		    !memcmp(dev->dev_addr, ha, dev->addr_len))
889 890 891
			return dev;

	return NULL;
L
Linus Torvalds 已提交
892
}
893
EXPORT_SYMBOL(dev_getbyhwaddr_rcu);
894

895
struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type)
L
Linus Torvalds 已提交
896 897 898
{
	struct net_device *dev;

899
	ASSERT_RTNL();
900
	for_each_netdev(net, dev)
901
		if (dev->type == type)
902 903 904
			return dev;

	return NULL;
905 906 907
}
EXPORT_SYMBOL(__dev_getfirstbyhwtype);

908
struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type)
909
{
910
	struct net_device *dev, *ret = NULL;
911

912 913 914 915 916 917 918 919 920
	rcu_read_lock();
	for_each_netdev_rcu(net, dev)
		if (dev->type == type) {
			dev_hold(dev);
			ret = dev;
			break;
		}
	rcu_read_unlock();
	return ret;
L
Linus Torvalds 已提交
921 922 923 924
}
EXPORT_SYMBOL(dev_getfirstbyhwtype);

/**
925
 *	__dev_get_by_flags - find any device with given flags
926
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
927 928 929 930
 *	@if_flags: IFF_* values
 *	@mask: bitmask of bits in if_flags to check
 *
 *	Search for any interface with the given flags. Returns NULL if a device
E
Eric Dumazet 已提交
931
 *	is not found or a pointer to the device. Must be called inside
932
 *	rtnl_lock(), and result refcount is unchanged.
L
Linus Torvalds 已提交
933 934
 */

935 936
struct net_device *__dev_get_by_flags(struct net *net, unsigned short if_flags,
				      unsigned short mask)
L
Linus Torvalds 已提交
937
{
938
	struct net_device *dev, *ret;
L
Linus Torvalds 已提交
939

940 941
	ASSERT_RTNL();

942
	ret = NULL;
943
	for_each_netdev(net, dev) {
L
Linus Torvalds 已提交
944
		if (((dev->flags ^ if_flags) & mask) == 0) {
945
			ret = dev;
L
Linus Torvalds 已提交
946 947 948
			break;
		}
	}
949
	return ret;
L
Linus Torvalds 已提交
950
}
951
EXPORT_SYMBOL(__dev_get_by_flags);
L
Linus Torvalds 已提交
952 953 954 955 956 957

/**
 *	dev_valid_name - check if name is okay for network device
 *	@name: name string
 *
 *	Network device names need to be valid file names to
958 959
 *	to allow sysfs to work.  We also disallow any kind of
 *	whitespace.
L
Linus Torvalds 已提交
960
 */
961
bool dev_valid_name(const char *name)
L
Linus Torvalds 已提交
962
{
963
	if (*name == '\0')
964
		return false;
965
	if (strlen(name) >= IFNAMSIZ)
966
		return false;
967
	if (!strcmp(name, ".") || !strcmp(name, ".."))
968
		return false;
969 970

	while (*name) {
971
		if (*name == '/' || *name == ':' || isspace(*name))
972
			return false;
973 974
		name++;
	}
975
	return true;
L
Linus Torvalds 已提交
976
}
E
Eric Dumazet 已提交
977
EXPORT_SYMBOL(dev_valid_name);
L
Linus Torvalds 已提交
978 979

/**
980 981
 *	__dev_alloc_name - allocate a name for a device
 *	@net: network namespace to allocate the device name in
L
Linus Torvalds 已提交
982
 *	@name: name format string
983
 *	@buf:  scratch buffer and result name string
L
Linus Torvalds 已提交
984 985
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
S
Stephen Hemminger 已提交
986 987 988 989 990 991
 *	id. It scans list of devices to build up a free map, then chooses
 *	the first empty slot. The caller must hold the dev_base or rtnl lock
 *	while allocating the name and adding the device in order to avoid
 *	duplicates.
 *	Limited to bits_per_byte * page size devices (ie 32K on most platforms).
 *	Returns the number of the unit assigned or a negative errno code.
L
Linus Torvalds 已提交
992 993
 */

994
static int __dev_alloc_name(struct net *net, const char *name, char *buf)
L
Linus Torvalds 已提交
995 996 997 998
{
	int i = 0;
	const char *p;
	const int max_netdevices = 8*PAGE_SIZE;
S
Stephen Hemminger 已提交
999
	unsigned long *inuse;
L
Linus Torvalds 已提交
1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012
	struct net_device *d;

	p = strnchr(name, IFNAMSIZ-1, '%');
	if (p) {
		/*
		 * Verify the string as this thing may have come from
		 * the user.  There must be either one "%d" and no other "%"
		 * characters.
		 */
		if (p[1] != 'd' || strchr(p + 2, '%'))
			return -EINVAL;

		/* Use one page as a bit array of possible slots */
S
Stephen Hemminger 已提交
1013
		inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC);
L
Linus Torvalds 已提交
1014 1015 1016
		if (!inuse)
			return -ENOMEM;

1017
		for_each_netdev(net, d) {
L
Linus Torvalds 已提交
1018 1019 1020 1021 1022 1023
			if (!sscanf(d->name, name, &i))
				continue;
			if (i < 0 || i >= max_netdevices)
				continue;

			/*  avoid cases where sscanf is not exact inverse of printf */
1024
			snprintf(buf, IFNAMSIZ, name, i);
L
Linus Torvalds 已提交
1025 1026 1027 1028 1029 1030 1031 1032
			if (!strncmp(buf, d->name, IFNAMSIZ))
				set_bit(i, inuse);
		}

		i = find_first_zero_bit(inuse, max_netdevices);
		free_page((unsigned long) inuse);
	}

1033 1034
	if (buf != name)
		snprintf(buf, IFNAMSIZ, name, i);
1035
	if (!__dev_get_by_name(net, buf))
L
Linus Torvalds 已提交
1036 1037 1038 1039 1040 1041 1042 1043 1044
		return i;

	/* It is possible to run out of possible slots
	 * when the name is long and there isn't enough space left
	 * for the digits, or if all bits are used.
	 */
	return -ENFILE;
}

1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064
/**
 *	dev_alloc_name - allocate a name for a device
 *	@dev: device
 *	@name: name format string
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
 *	id. It scans list of devices to build up a free map, then chooses
 *	the first empty slot. The caller must hold the dev_base or rtnl lock
 *	while allocating the name and adding the device in order to avoid
 *	duplicates.
 *	Limited to bits_per_byte * page size devices (ie 32K on most platforms).
 *	Returns the number of the unit assigned or a negative errno code.
 */

int dev_alloc_name(struct net_device *dev, const char *name)
{
	char buf[IFNAMSIZ];
	struct net *net;
	int ret;

1065 1066
	BUG_ON(!dev_net(dev));
	net = dev_net(dev);
1067 1068 1069 1070 1071
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
}
E
Eric Dumazet 已提交
1072
EXPORT_SYMBOL(dev_alloc_name);
1073

1074 1075 1076
static int dev_alloc_name_ns(struct net *net,
			     struct net_device *dev,
			     const char *name)
1077
{
1078 1079
	char buf[IFNAMSIZ];
	int ret;
1080

1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
}

static int dev_get_valid_name(struct net *net,
			      struct net_device *dev,
			      const char *name)
{
	BUG_ON(!net);
1092

1093 1094 1095
	if (!dev_valid_name(name))
		return -EINVAL;

1096
	if (strchr(name, '%'))
1097
		return dev_alloc_name_ns(net, dev, name);
1098 1099
	else if (__dev_get_by_name(net, name))
		return -EEXIST;
1100 1101
	else if (dev->name != name)
		strlcpy(dev->name, name, IFNAMSIZ);
1102 1103 1104

	return 0;
}
L
Linus Torvalds 已提交
1105 1106 1107 1108 1109 1110 1111 1112 1113

/**
 *	dev_change_name - change name of a device
 *	@dev: device
 *	@newname: name (or format string) must be at least IFNAMSIZ
 *
 *	Change name of a device, can pass format strings "eth%d".
 *	for wildcarding.
 */
1114
int dev_change_name(struct net_device *dev, const char *newname)
L
Linus Torvalds 已提交
1115
{
1116
	unsigned char old_assign_type;
1117
	char oldname[IFNAMSIZ];
L
Linus Torvalds 已提交
1118
	int err = 0;
1119
	int ret;
1120
	struct net *net;
L
Linus Torvalds 已提交
1121 1122

	ASSERT_RTNL();
1123
	BUG_ON(!dev_net(dev));
L
Linus Torvalds 已提交
1124

1125
	net = dev_net(dev);
L
Linus Torvalds 已提交
1126 1127 1128
	if (dev->flags & IFF_UP)
		return -EBUSY;

1129
	write_seqcount_begin(&devnet_rename_seq);
1130 1131

	if (strncmp(newname, dev->name, IFNAMSIZ) == 0) {
1132
		write_seqcount_end(&devnet_rename_seq);
1133
		return 0;
1134
	}
1135

1136 1137
	memcpy(oldname, dev->name, IFNAMSIZ);

1138
	err = dev_get_valid_name(net, dev, newname);
1139
	if (err < 0) {
1140
		write_seqcount_end(&devnet_rename_seq);
1141
		return err;
1142
	}
L
Linus Torvalds 已提交
1143

1144 1145 1146
	if (oldname[0] && !strchr(oldname, '%'))
		netdev_info(dev, "renamed from %s\n", oldname);

1147 1148 1149
	old_assign_type = dev->name_assign_type;
	dev->name_assign_type = NET_NAME_RENAMED;

1150
rollback:
1151 1152 1153
	ret = device_rename(&dev->dev, dev->name);
	if (ret) {
		memcpy(dev->name, oldname, IFNAMSIZ);
1154
		dev->name_assign_type = old_assign_type;
1155
		write_seqcount_end(&devnet_rename_seq);
1156
		return ret;
1157
	}
1158

1159
	write_seqcount_end(&devnet_rename_seq);
1160

1161 1162
	netdev_adjacent_rename_links(dev, oldname);

1163
	write_lock_bh(&dev_base_lock);
1164
	hlist_del_rcu(&dev->name_hlist);
1165 1166 1167 1168 1169 1170
	write_unlock_bh(&dev_base_lock);

	synchronize_rcu();

	write_lock_bh(&dev_base_lock);
	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
1171 1172
	write_unlock_bh(&dev_base_lock);

1173
	ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
1174 1175 1176
	ret = notifier_to_errno(ret);

	if (ret) {
1177 1178
		/* err >= 0 after dev_alloc_name() or stores the first errno */
		if (err >= 0) {
1179
			err = ret;
1180
			write_seqcount_begin(&devnet_rename_seq);
1181
			memcpy(dev->name, oldname, IFNAMSIZ);
1182
			memcpy(oldname, newname, IFNAMSIZ);
1183 1184
			dev->name_assign_type = old_assign_type;
			old_assign_type = NET_NAME_RENAMED;
1185
			goto rollback;
1186
		} else {
1187
			pr_err("%s: name change rollback failed: %d\n",
1188
			       dev->name, ret);
1189 1190
		}
	}
L
Linus Torvalds 已提交
1191 1192 1193 1194

	return err;
}

1195 1196 1197 1198
/**
 *	dev_set_alias - change ifalias of a device
 *	@dev: device
 *	@alias: name up to IFALIASZ
1199
 *	@len: limit of bytes to copy from info
1200 1201 1202 1203 1204
 *
 *	Set ifalias for a device,
 */
int dev_set_alias(struct net_device *dev, const char *alias, size_t len)
{
1205 1206
	char *new_ifalias;

1207 1208 1209 1210 1211
	ASSERT_RTNL();

	if (len >= IFALIASZ)
		return -EINVAL;

1212
	if (!len) {
1213 1214
		kfree(dev->ifalias);
		dev->ifalias = NULL;
1215 1216 1217
		return 0;
	}

1218 1219
	new_ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL);
	if (!new_ifalias)
1220
		return -ENOMEM;
1221
	dev->ifalias = new_ifalias;
1222 1223 1224 1225 1226 1227

	strlcpy(dev->ifalias, alias, len+1);
	return len;
}


1228
/**
S
Stephen Hemminger 已提交
1229
 *	netdev_features_change - device changes features
1230 1231 1232 1233 1234 1235
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed features.
 */
void netdev_features_change(struct net_device *dev)
{
1236
	call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev);
1237 1238 1239
}
EXPORT_SYMBOL(netdev_features_change);

L
Linus Torvalds 已提交
1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250
/**
 *	netdev_state_change - device changes state
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed state. This function calls
 *	the notifier chains for netdev_chain and sends a NEWLINK message
 *	to the routing socket.
 */
void netdev_state_change(struct net_device *dev)
{
	if (dev->flags & IFF_UP) {
1251 1252 1253 1254 1255
		struct netdev_notifier_change_info change_info;

		change_info.flags_changed = 0;
		call_netdevice_notifiers_info(NETDEV_CHANGE, dev,
					      &change_info.info);
1256
		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
L
Linus Torvalds 已提交
1257 1258
	}
}
E
Eric Dumazet 已提交
1259
EXPORT_SYMBOL(netdev_state_change);
L
Linus Torvalds 已提交
1260

1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271
/**
 * 	netdev_notify_peers - notify network peers about existence of @dev
 * 	@dev: network device
 *
 * Generate traffic such that interested network peers are aware of
 * @dev, such as by generating a gratuitous ARP. This may be used when
 * a device wants to inform the rest of the network about some sort of
 * reconfiguration such as a failover event or virtual machine
 * migration.
 */
void netdev_notify_peers(struct net_device *dev)
1272
{
1273 1274 1275
	rtnl_lock();
	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, dev);
	rtnl_unlock();
1276
}
1277
EXPORT_SYMBOL(netdev_notify_peers);
1278

1279
static int __dev_open(struct net_device *dev)
L
Linus Torvalds 已提交
1280
{
1281
	const struct net_device_ops *ops = dev->netdev_ops;
1282
	int ret;
L
Linus Torvalds 已提交
1283

1284 1285
	ASSERT_RTNL();

L
Linus Torvalds 已提交
1286 1287 1288
	if (!netif_device_present(dev))
		return -ENODEV;

1289 1290 1291 1292
	/* Block netpoll from trying to do any rx path servicing.
	 * If we don't do this there is a chance ndo_poll_controller
	 * or ndo_poll may be running while we open the device
	 */
1293
	netpoll_poll_disable(dev);
1294

1295 1296 1297 1298 1299
	ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

L
Linus Torvalds 已提交
1300
	set_bit(__LINK_STATE_START, &dev->state);
1301

1302 1303
	if (ops->ndo_validate_addr)
		ret = ops->ndo_validate_addr(dev);
1304

1305 1306
	if (!ret && ops->ndo_open)
		ret = ops->ndo_open(dev);
L
Linus Torvalds 已提交
1307

1308
	netpoll_poll_enable(dev);
1309

1310 1311 1312
	if (ret)
		clear_bit(__LINK_STATE_START, &dev->state);
	else {
L
Linus Torvalds 已提交
1313
		dev->flags |= IFF_UP;
1314
		dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
1315
		dev_activate(dev);
1316
		add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
1317
	}
1318

L
Linus Torvalds 已提交
1319 1320 1321 1322
	return ret;
}

/**
1323 1324
 *	dev_open	- prepare an interface for use.
 *	@dev:	device to open
L
Linus Torvalds 已提交
1325
 *
1326 1327 1328 1329 1330 1331 1332
 *	Takes a device from down to up state. The device's private open
 *	function is invoked and then the multicast lists are loaded. Finally
 *	the device is moved into the up state and a %NETDEV_UP message is
 *	sent to the netdev notifier chain.
 *
 *	Calling this function on an active interface is a nop. On a failure
 *	a negative errno code is returned.
L
Linus Torvalds 已提交
1333
 */
1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344
int dev_open(struct net_device *dev)
{
	int ret;

	if (dev->flags & IFF_UP)
		return 0;

	ret = __dev_open(dev);
	if (ret < 0)
		return ret;

1345
	rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1346 1347 1348 1349 1350 1351
	call_netdevice_notifiers(NETDEV_UP, dev);

	return ret;
}
EXPORT_SYMBOL(dev_open);

1352
static int __dev_close_many(struct list_head *head)
L
Linus Torvalds 已提交
1353
{
1354
	struct net_device *dev;
1355

1356
	ASSERT_RTNL();
1357 1358
	might_sleep();

1359
	list_for_each_entry(dev, head, close_list) {
1360
		/* Temporarily disable netpoll until the interface is down */
1361
		netpoll_poll_disable(dev);
1362

1363
		call_netdevice_notifiers(NETDEV_GOING_DOWN, dev);
L
Linus Torvalds 已提交
1364

1365
		clear_bit(__LINK_STATE_START, &dev->state);
L
Linus Torvalds 已提交
1366

1367 1368 1369 1370 1371 1372
		/* Synchronize to scheduled poll. We cannot touch poll list, it
		 * can be even on different cpu. So just clear netif_running().
		 *
		 * dev->stop() will invoke napi_disable() on all of it's
		 * napi_struct instances on this device.
		 */
1373
		smp_mb__after_atomic(); /* Commit netif_running(). */
1374
	}
L
Linus Torvalds 已提交
1375

1376
	dev_deactivate_many(head);
1377

1378
	list_for_each_entry(dev, head, close_list) {
1379
		const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
1380

1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391
		/*
		 *	Call the device specific close. This cannot fail.
		 *	Only if device is UP
		 *
		 *	We allow it to be called even after a DETACH hot-plug
		 *	event.
		 */
		if (ops->ndo_stop)
			ops->ndo_stop(dev);

		dev->flags &= ~IFF_UP;
1392
		netpoll_poll_enable(dev);
1393 1394 1395 1396 1397 1398 1399
	}

	return 0;
}

static int __dev_close(struct net_device *dev)
{
1400
	int retval;
1401 1402
	LIST_HEAD(single);

1403
	list_add(&dev->close_list, &single);
1404 1405
	retval = __dev_close_many(&single);
	list_del(&single);
1406

1407
	return retval;
1408 1409
}

1410
int dev_close_many(struct list_head *head, bool unlink)
1411 1412
{
	struct net_device *dev, *tmp;
L
Linus Torvalds 已提交
1413

1414 1415
	/* Remove the devices that don't need to be closed */
	list_for_each_entry_safe(dev, tmp, head, close_list)
1416
		if (!(dev->flags & IFF_UP))
1417
			list_del_init(&dev->close_list);
1418 1419

	__dev_close_many(head);
L
Linus Torvalds 已提交
1420

1421
	list_for_each_entry_safe(dev, tmp, head, close_list) {
1422
		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1423
		call_netdevice_notifiers(NETDEV_DOWN, dev);
1424 1425
		if (unlink)
			list_del_init(&dev->close_list);
1426
	}
1427 1428 1429

	return 0;
}
1430
EXPORT_SYMBOL(dev_close_many);
1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442

/**
 *	dev_close - shutdown an interface.
 *	@dev: device to shutdown
 *
 *	This function moves an active device into down state. A
 *	%NETDEV_GOING_DOWN is sent to the netdev notifier chain. The device
 *	is then deactivated and finally a %NETDEV_DOWN is sent to the notifier
 *	chain.
 */
int dev_close(struct net_device *dev)
{
1443 1444
	if (dev->flags & IFF_UP) {
		LIST_HEAD(single);
L
Linus Torvalds 已提交
1445

1446
		list_add(&dev->close_list, &single);
1447
		dev_close_many(&single, true);
1448 1449
		list_del(&single);
	}
1450
	return 0;
L
Linus Torvalds 已提交
1451
}
E
Eric Dumazet 已提交
1452
EXPORT_SYMBOL(dev_close);
L
Linus Torvalds 已提交
1453 1454


1455 1456 1457 1458 1459 1460 1461 1462 1463 1464
/**
 *	dev_disable_lro - disable Large Receive Offload on a device
 *	@dev: device
 *
 *	Disable Large Receive Offload (LRO) on a net device.  Must be
 *	called under RTNL.  This is needed if received packets may be
 *	forwarded to another interface.
 */
void dev_disable_lro(struct net_device *dev)
{
1465 1466
	struct net_device *lower_dev;
	struct list_head *iter;
1467

M
Michał Mirosław 已提交
1468 1469
	dev->wanted_features &= ~NETIF_F_LRO;
	netdev_update_features(dev);
1470

1471 1472
	if (unlikely(dev->features & NETIF_F_LRO))
		netdev_WARN(dev, "failed to disable LRO!\n");
1473 1474 1475

	netdev_for_each_lower_dev(dev, lower_dev, iter)
		dev_disable_lro(lower_dev);
1476 1477 1478
}
EXPORT_SYMBOL(dev_disable_lro);

1479 1480 1481 1482 1483 1484 1485 1486
static int call_netdevice_notifier(struct notifier_block *nb, unsigned long val,
				   struct net_device *dev)
{
	struct netdev_notifier_info info;

	netdev_notifier_info_init(&info, dev);
	return nb->notifier_call(nb, val, &info);
}
1487

1488 1489
static int dev_boot_phase = 1;

L
Linus Torvalds 已提交
1490 1491 1492 1493 1494 1495 1496 1497 1498 1499
/**
 *	register_netdevice_notifier - register a network notifier block
 *	@nb: notifier
 *
 *	Register a notifier to be called when network device events occur.
 *	The notifier passed is linked into the kernel structures and must
 *	not be reused until it has been unregistered. A negative errno code
 *	is returned on a failure.
 *
 * 	When registered all registration and up events are replayed
1500
 *	to the new notifier to allow device to have a race free
L
Linus Torvalds 已提交
1501 1502 1503 1504 1505 1506
 *	view of the network device list.
 */

int register_netdevice_notifier(struct notifier_block *nb)
{
	struct net_device *dev;
1507
	struct net_device *last;
1508
	struct net *net;
L
Linus Torvalds 已提交
1509 1510 1511
	int err;

	rtnl_lock();
1512
	err = raw_notifier_chain_register(&netdev_chain, nb);
1513 1514
	if (err)
		goto unlock;
1515 1516 1517 1518
	if (dev_boot_phase)
		goto unlock;
	for_each_net(net) {
		for_each_netdev(net, dev) {
1519
			err = call_netdevice_notifier(nb, NETDEV_REGISTER, dev);
1520 1521 1522 1523 1524 1525
			err = notifier_to_errno(err);
			if (err)
				goto rollback;

			if (!(dev->flags & IFF_UP))
				continue;
L
Linus Torvalds 已提交
1526

1527
			call_netdevice_notifier(nb, NETDEV_UP, dev);
1528
		}
L
Linus Torvalds 已提交
1529
	}
1530 1531

unlock:
L
Linus Torvalds 已提交
1532 1533
	rtnl_unlock();
	return err;
1534 1535 1536

rollback:
	last = dev;
1537 1538 1539
	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev == last)
1540
				goto outroll;
1541

1542
			if (dev->flags & IFF_UP) {
1543 1544 1545
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1546
			}
1547
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1548 1549
		}
	}
1550

1551
outroll:
1552
	raw_notifier_chain_unregister(&netdev_chain, nb);
1553
	goto unlock;
L
Linus Torvalds 已提交
1554
}
E
Eric Dumazet 已提交
1555
EXPORT_SYMBOL(register_netdevice_notifier);
L
Linus Torvalds 已提交
1556 1557 1558 1559 1560 1561 1562 1563 1564

/**
 *	unregister_netdevice_notifier - unregister a network notifier block
 *	@nb: notifier
 *
 *	Unregister a notifier previously registered by
 *	register_netdevice_notifier(). The notifier is unlinked into the
 *	kernel structures and may then be reused. A negative errno code
 *	is returned on a failure.
1565 1566 1567 1568
 *
 * 	After unregistering unregister and down device events are synthesized
 *	for all devices on the device list to the removed notifier to remove
 *	the need for special case cleanup code.
L
Linus Torvalds 已提交
1569 1570 1571 1572
 */

int unregister_netdevice_notifier(struct notifier_block *nb)
{
1573 1574
	struct net_device *dev;
	struct net *net;
1575 1576 1577
	int err;

	rtnl_lock();
1578
	err = raw_notifier_chain_unregister(&netdev_chain, nb);
1579 1580 1581 1582 1583 1584
	if (err)
		goto unlock;

	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev->flags & IFF_UP) {
1585 1586 1587
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1588
			}
1589
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1590 1591 1592
		}
	}
unlock:
1593 1594
	rtnl_unlock();
	return err;
L
Linus Torvalds 已提交
1595
}
E
Eric Dumazet 已提交
1596
EXPORT_SYMBOL(unregister_netdevice_notifier);
L
Linus Torvalds 已提交
1597

1598 1599 1600 1601 1602 1603 1604 1605 1606 1607
/**
 *	call_netdevice_notifiers_info - call all network notifier blocks
 *	@val: value passed unmodified to notifier function
 *	@dev: net_device pointer passed unmodified to notifier function
 *	@info: notifier information data
 *
 *	Call all network notifier blocks.  Parameters and return value
 *	are as for raw_notifier_call_chain().
 */

S
stephen hemminger 已提交
1608 1609 1610
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info)
1611 1612 1613 1614 1615 1616
{
	ASSERT_RTNL();
	netdev_notifier_info_init(info, dev);
	return raw_notifier_call_chain(&netdev_chain, val, info);
}

L
Linus Torvalds 已提交
1617 1618 1619
/**
 *	call_netdevice_notifiers - call all network notifier blocks
 *      @val: value passed unmodified to notifier function
1620
 *      @dev: net_device pointer passed unmodified to notifier function
L
Linus Torvalds 已提交
1621 1622
 *
 *	Call all network notifier blocks.  Parameters and return value
1623
 *	are as for raw_notifier_call_chain().
L
Linus Torvalds 已提交
1624 1625
 */

1626
int call_netdevice_notifiers(unsigned long val, struct net_device *dev)
L
Linus Torvalds 已提交
1627
{
1628 1629 1630
	struct netdev_notifier_info info;

	return call_netdevice_notifiers_info(val, dev, &info);
L
Linus Torvalds 已提交
1631
}
1632
EXPORT_SYMBOL(call_netdevice_notifiers);
L
Linus Torvalds 已提交
1633

1634
#ifdef CONFIG_NET_INGRESS
1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649
static struct static_key ingress_needed __read_mostly;

void net_inc_ingress_queue(void)
{
	static_key_slow_inc(&ingress_needed);
}
EXPORT_SYMBOL_GPL(net_inc_ingress_queue);

void net_dec_ingress_queue(void)
{
	static_key_slow_dec(&ingress_needed);
}
EXPORT_SYMBOL_GPL(net_dec_ingress_queue);
#endif

1650
static struct static_key netstamp_needed __read_mostly;
1651
#ifdef HAVE_JUMP_LABEL
1652
/* We are not allowed to call static_key_slow_dec() from irq context
1653
 * If net_disable_timestamp() is called from irq context, defer the
1654
 * static_key_slow_dec() calls.
1655 1656 1657
 */
static atomic_t netstamp_needed_deferred;
#endif
L
Linus Torvalds 已提交
1658 1659 1660

void net_enable_timestamp(void)
{
1661 1662 1663 1664 1665
#ifdef HAVE_JUMP_LABEL
	int deferred = atomic_xchg(&netstamp_needed_deferred, 0);

	if (deferred) {
		while (--deferred)
1666
			static_key_slow_dec(&netstamp_needed);
1667 1668 1669
		return;
	}
#endif
1670
	static_key_slow_inc(&netstamp_needed);
L
Linus Torvalds 已提交
1671
}
E
Eric Dumazet 已提交
1672
EXPORT_SYMBOL(net_enable_timestamp);
L
Linus Torvalds 已提交
1673 1674 1675

void net_disable_timestamp(void)
{
1676 1677 1678 1679 1680 1681
#ifdef HAVE_JUMP_LABEL
	if (in_interrupt()) {
		atomic_inc(&netstamp_needed_deferred);
		return;
	}
#endif
1682
	static_key_slow_dec(&netstamp_needed);
L
Linus Torvalds 已提交
1683
}
E
Eric Dumazet 已提交
1684
EXPORT_SYMBOL(net_disable_timestamp);
L
Linus Torvalds 已提交
1685

E
Eric Dumazet 已提交
1686
static inline void net_timestamp_set(struct sk_buff *skb)
L
Linus Torvalds 已提交
1687
{
1688
	skb->tstamp.tv64 = 0;
1689
	if (static_key_false(&netstamp_needed))
1690
		__net_timestamp(skb);
L
Linus Torvalds 已提交
1691 1692
}

1693
#define net_timestamp_check(COND, SKB)			\
1694
	if (static_key_false(&netstamp_needed)) {		\
1695 1696 1697
		if ((COND) && !(SKB)->tstamp.tv64)	\
			__net_timestamp(SKB);		\
	}						\
E
Eric Dumazet 已提交
1698

1699
bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb)
1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717
{
	unsigned int len;

	if (!(dev->flags & IFF_UP))
		return false;

	len = dev->mtu + dev->hard_header_len + VLAN_HLEN;
	if (skb->len <= len)
		return true;

	/* if TSO is enabled, we don't care about the length as the packet
	 * could be forwarded without being segmented before
	 */
	if (skb_is_gso(skb))
		return true;

	return false;
}
1718
EXPORT_SYMBOL_GPL(is_skb_forwardable);
1719

H
Herbert Xu 已提交
1720 1721
int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb)
{
1722 1723
	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
	    unlikely(!is_skb_forwardable(dev, skb))) {
H
Herbert Xu 已提交
1724 1725 1726 1727 1728 1729
		atomic_long_inc(&dev->rx_dropped);
		kfree_skb(skb);
		return NET_RX_DROP;
	}

	skb_scrub_packet(skb, true);
1730
	skb->priority = 0;
H
Herbert Xu 已提交
1731
	skb->protocol = eth_type_trans(skb, dev);
1732
	skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
H
Herbert Xu 已提交
1733 1734 1735 1736 1737

	return 0;
}
EXPORT_SYMBOL_GPL(__dev_forward_skb);

1738 1739 1740 1741 1742 1743 1744 1745
/**
 * dev_forward_skb - loopback an skb to another netif
 *
 * @dev: destination network device
 * @skb: buffer to forward
 *
 * return values:
 *	NET_RX_SUCCESS	(no congestion)
1746
 *	NET_RX_DROP     (packet was dropped, but freed)
1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757
 *
 * dev_forward_skb can be used for injecting an skb from the
 * start_xmit function of one device into the receive queue
 * of another device.
 *
 * The receiving device may be in another namespace, so
 * we have to clear all information in the skb that could
 * impact namespace isolation.
 */
int dev_forward_skb(struct net_device *dev, struct sk_buff *skb)
{
H
Herbert Xu 已提交
1758
	return __dev_forward_skb(dev, skb) ?: netif_rx_internal(skb);
1759 1760 1761
}
EXPORT_SYMBOL_GPL(dev_forward_skb);

1762 1763 1764 1765
static inline int deliver_skb(struct sk_buff *skb,
			      struct packet_type *pt_prev,
			      struct net_device *orig_dev)
{
1766 1767
	if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
		return -ENOMEM;
1768 1769 1770 1771
	atomic_inc(&skb->users);
	return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
}

1772 1773
static inline void deliver_ptype_list_skb(struct sk_buff *skb,
					  struct packet_type **pt,
1774 1775
					  struct net_device *orig_dev,
					  __be16 type,
1776 1777 1778 1779 1780 1781 1782 1783
					  struct list_head *ptype_list)
{
	struct packet_type *ptype, *pt_prev = *pt;

	list_for_each_entry_rcu(ptype, ptype_list, list) {
		if (ptype->type != type)
			continue;
		if (pt_prev)
1784
			deliver_skb(skb, pt_prev, orig_dev);
1785 1786 1787 1788 1789
		pt_prev = ptype;
	}
	*pt = pt_prev;
}

1790 1791
static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
{
1792
	if (!ptype->af_packet_priv || !skb->sk)
1793 1794 1795 1796 1797 1798 1799 1800 1801 1802
		return false;

	if (ptype->id_match)
		return ptype->id_match(ptype, skb->sk);
	else if ((struct sock *)ptype->af_packet_priv == skb->sk)
		return true;

	return false;
}

L
Linus Torvalds 已提交
1803 1804 1805 1806 1807
/*
 *	Support routine. Sends outgoing frames to any network
 *	taps currently in use.
 */

1808
static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
L
Linus Torvalds 已提交
1809 1810
{
	struct packet_type *ptype;
1811 1812
	struct sk_buff *skb2 = NULL;
	struct packet_type *pt_prev = NULL;
1813
	struct list_head *ptype_list = &ptype_all;
1814

L
Linus Torvalds 已提交
1815
	rcu_read_lock();
1816 1817
again:
	list_for_each_entry_rcu(ptype, ptype_list, list) {
L
Linus Torvalds 已提交
1818 1819 1820
		/* Never send packets back to the socket
		 * they originated from - MvS (miquels@drinkel.ow.org)
		 */
1821 1822
		if (skb_loop_sk(ptype, skb))
			continue;
1823

1824 1825 1826 1827 1828
		if (pt_prev) {
			deliver_skb(skb2, pt_prev, skb->dev);
			pt_prev = ptype;
			continue;
		}
L
Linus Torvalds 已提交
1829

1830 1831 1832 1833
		/* need to clone skb, done only once */
		skb2 = skb_clone(skb, GFP_ATOMIC);
		if (!skb2)
			goto out_unlock;
1834

1835
		net_timestamp_set(skb2);
L
Linus Torvalds 已提交
1836

1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848
		/* skb->nh should be correctly
		 * set by sender, so that the second statement is
		 * just protection against buggy protocols.
		 */
		skb_reset_mac_header(skb2);

		if (skb_network_header(skb2) < skb2->data ||
		    skb_network_header(skb2) > skb_tail_pointer(skb2)) {
			net_crit_ratelimited("protocol %04x is buggy, dev %s\n",
					     ntohs(skb2->protocol),
					     dev->name);
			skb_reset_network_header(skb2);
L
Linus Torvalds 已提交
1849
		}
1850 1851 1852 1853 1854 1855 1856 1857 1858

		skb2->transport_header = skb2->network_header;
		skb2->pkt_type = PACKET_OUTGOING;
		pt_prev = ptype;
	}

	if (ptype_list == &ptype_all) {
		ptype_list = &dev->ptype_all;
		goto again;
L
Linus Torvalds 已提交
1859
	}
1860
out_unlock:
1861 1862
	if (pt_prev)
		pt_prev->func(skb2, skb->dev, pt_prev, skb->dev);
L
Linus Torvalds 已提交
1863 1864 1865
	rcu_read_unlock();
}

1866 1867
/**
 * netif_setup_tc - Handle tc mappings on real_num_tx_queues change
1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878
 * @dev: Network device
 * @txq: number of queues available
 *
 * If real_num_tx_queues is changed the tc mappings may no longer be
 * valid. To resolve this verify the tc mapping remains valid and if
 * not NULL the mapping. With no priorities mapping to this
 * offset/count pair it will no longer be used. In the worst case TC0
 * is invalid nothing can be done so disable priority mappings. If is
 * expected that drivers will fix this mapping if they can before
 * calling netif_set_real_num_tx_queues.
 */
E
Eric Dumazet 已提交
1879
static void netif_setup_tc(struct net_device *dev, unsigned int txq)
1880 1881 1882 1883 1884 1885
{
	int i;
	struct netdev_tc_txq *tc = &dev->tc_to_txq[0];

	/* If TC0 is invalidated disable TC mapping */
	if (tc->offset + tc->count > txq) {
1886
		pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n");
1887 1888 1889 1890 1891 1892 1893 1894 1895 1896
		dev->num_tc = 0;
		return;
	}

	/* Invalidated prio to tc mappings set to TC0 */
	for (i = 1; i < TC_BITMASK + 1; i++) {
		int q = netdev_get_prio_tc_map(dev, i);

		tc = &dev->tc_to_txq[q];
		if (tc->offset + tc->count > txq) {
1897 1898
			pr_warn("Number of in use tx queues changed. Priority %i to tc mapping %i is no longer valid. Setting map to 0\n",
				i, q);
1899 1900 1901 1902 1903
			netdev_set_prio_tc_map(dev, i, 0);
		}
	}
}

1904 1905 1906 1907 1908
#ifdef CONFIG_XPS
static DEFINE_MUTEX(xps_map_mutex);
#define xmap_dereference(P)		\
	rcu_dereference_protected((P), lockdep_is_held(&xps_map_mutex))

1909 1910
static struct xps_map *remove_xps_queue(struct xps_dev_maps *dev_maps,
					int cpu, u16 index)
1911
{
1912 1913
	struct xps_map *map = NULL;
	int pos;
1914

1915 1916
	if (dev_maps)
		map = xmap_dereference(dev_maps->cpu_map[cpu]);
1917

1918 1919
	for (pos = 0; map && pos < map->len; pos++) {
		if (map->queues[pos] == index) {
1920 1921 1922
			if (map->len > 1) {
				map->queues[pos] = map->queues[--map->len];
			} else {
1923
				RCU_INIT_POINTER(dev_maps->cpu_map[cpu], NULL);
1924 1925 1926
				kfree_rcu(map, rcu);
				map = NULL;
			}
1927
			break;
1928 1929 1930
		}
	}

1931 1932 1933
	return map;
}

1934
static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
1935 1936
{
	struct xps_dev_maps *dev_maps;
1937
	int cpu, i;
1938 1939 1940 1941 1942 1943 1944 1945 1946
	bool active = false;

	mutex_lock(&xps_map_mutex);
	dev_maps = xmap_dereference(dev->xps_maps);

	if (!dev_maps)
		goto out_no_maps;

	for_each_possible_cpu(cpu) {
1947 1948 1949 1950 1951
		for (i = index; i < dev->num_tx_queues; i++) {
			if (!remove_xps_queue(dev_maps, cpu, i))
				break;
		}
		if (i == dev->num_tx_queues)
1952 1953 1954 1955
			active = true;
	}

	if (!active) {
1956 1957 1958 1959
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

1960 1961 1962 1963
	for (i = index; i < dev->num_tx_queues; i++)
		netdev_queue_numa_node_write(netdev_get_tx_queue(dev, i),
					     NUMA_NO_NODE);

1964 1965 1966 1967
out_no_maps:
	mutex_unlock(&xps_map_mutex);
}

1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
static struct xps_map *expand_xps_map(struct xps_map *map,
				      int cpu, u16 index)
{
	struct xps_map *new_map;
	int alloc_len = XPS_MIN_MAP_ALLOC;
	int i, pos;

	for (pos = 0; map && pos < map->len; pos++) {
		if (map->queues[pos] != index)
			continue;
		return map;
	}

	/* Need to add queue to this CPU's existing map */
	if (map) {
		if (pos < map->alloc_len)
			return map;

		alloc_len = map->alloc_len * 2;
	}

	/* Need to allocate new map to store queue on this CPU's map */
	new_map = kzalloc_node(XPS_MAP_SIZE(alloc_len), GFP_KERNEL,
			       cpu_to_node(cpu));
	if (!new_map)
		return NULL;

	for (i = 0; i < pos; i++)
		new_map->queues[i] = map->queues[i];
	new_map->alloc_len = alloc_len;
	new_map->len = pos;

	return new_map;
}

2003 2004
int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
			u16 index)
2005
{
2006
	struct xps_dev_maps *dev_maps, *new_dev_maps = NULL;
2007 2008
	struct xps_map *map, *new_map;
	int maps_sz = max_t(unsigned int, XPS_DEV_MAPS_SIZE, L1_CACHE_BYTES);
2009 2010
	int cpu, numa_node_id = -2;
	bool active = false;
2011 2012 2013 2014 2015

	mutex_lock(&xps_map_mutex);

	dev_maps = xmap_dereference(dev->xps_maps);

2016 2017 2018 2019 2020 2021 2022
	/* allocate memory for queue storage */
	for_each_online_cpu(cpu) {
		if (!cpumask_test_cpu(cpu, mask))
			continue;

		if (!new_dev_maps)
			new_dev_maps = kzalloc(maps_sz, GFP_KERNEL);
2023 2024
		if (!new_dev_maps) {
			mutex_unlock(&xps_map_mutex);
2025
			return -ENOMEM;
2026
		}
2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040

		map = dev_maps ? xmap_dereference(dev_maps->cpu_map[cpu]) :
				 NULL;

		map = expand_xps_map(map, cpu, index);
		if (!map)
			goto error;

		RCU_INIT_POINTER(new_dev_maps->cpu_map[cpu], map);
	}

	if (!new_dev_maps)
		goto out_no_new_maps;

2041
	for_each_possible_cpu(cpu) {
2042 2043 2044 2045 2046 2047 2048 2049 2050 2051
		if (cpumask_test_cpu(cpu, mask) && cpu_online(cpu)) {
			/* add queue to CPU maps */
			int pos = 0;

			map = xmap_dereference(new_dev_maps->cpu_map[cpu]);
			while ((pos < map->len) && (map->queues[pos] != index))
				pos++;

			if (pos == map->len)
				map->queues[map->len++] = index;
2052 2053 2054 2055 2056 2057
#ifdef CONFIG_NUMA
			if (numa_node_id == -2)
				numa_node_id = cpu_to_node(cpu);
			else if (numa_node_id != cpu_to_node(cpu))
				numa_node_id = -1;
#endif
2058 2059 2060 2061
		} else if (dev_maps) {
			/* fill in the new device map from the old device map */
			map = xmap_dereference(dev_maps->cpu_map[cpu]);
			RCU_INIT_POINTER(new_dev_maps->cpu_map[cpu], map);
2062
		}
2063

2064 2065
	}

2066 2067
	rcu_assign_pointer(dev->xps_maps, new_dev_maps);

2068
	/* Cleanup old maps */
2069 2070 2071 2072 2073 2074 2075
	if (dev_maps) {
		for_each_possible_cpu(cpu) {
			new_map = xmap_dereference(new_dev_maps->cpu_map[cpu]);
			map = xmap_dereference(dev_maps->cpu_map[cpu]);
			if (map && map != new_map)
				kfree_rcu(map, rcu);
		}
2076

2077
		kfree_rcu(dev_maps, rcu);
2078 2079
	}

2080 2081
	dev_maps = new_dev_maps;
	active = true;
2082

2083 2084
out_no_new_maps:
	/* update Tx queue numa node */
2085 2086 2087 2088
	netdev_queue_numa_node_write(netdev_get_tx_queue(dev, index),
				     (numa_node_id >= 0) ? numa_node_id :
				     NUMA_NO_NODE);

2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107
	if (!dev_maps)
		goto out_no_maps;

	/* removes queue from unused CPUs */
	for_each_possible_cpu(cpu) {
		if (cpumask_test_cpu(cpu, mask) && cpu_online(cpu))
			continue;

		if (remove_xps_queue(dev_maps, cpu, index))
			active = true;
	}

	/* free map if not active */
	if (!active) {
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

out_no_maps:
2108 2109 2110 2111
	mutex_unlock(&xps_map_mutex);

	return 0;
error:
2112 2113 2114 2115 2116 2117 2118 2119 2120
	/* remove any maps that we added */
	for_each_possible_cpu(cpu) {
		new_map = xmap_dereference(new_dev_maps->cpu_map[cpu]);
		map = dev_maps ? xmap_dereference(dev_maps->cpu_map[cpu]) :
				 NULL;
		if (new_map && new_map != map)
			kfree(new_map);
	}

2121 2122 2123 2124 2125 2126 2127 2128
	mutex_unlock(&xps_map_mutex);

	kfree(new_dev_maps);
	return -ENOMEM;
}
EXPORT_SYMBOL(netif_set_xps_queue);

#endif
2129 2130 2131 2132
/*
 * Routine to help set real_num_tx_queues. To avoid skbs mapped to queues
 * greater then real_num_tx_queues stale skbs on the qdisc must be flushed.
 */
2133
int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
2134
{
T
Tom Herbert 已提交
2135 2136
	int rc;

2137 2138
	if (txq < 1 || txq > dev->num_tx_queues)
		return -EINVAL;
2139

2140 2141
	if (dev->reg_state == NETREG_REGISTERED ||
	    dev->reg_state == NETREG_UNREGISTERING) {
2142 2143
		ASSERT_RTNL();

T
Tom Herbert 已提交
2144 2145
		rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
						  txq);
T
Tom Herbert 已提交
2146 2147 2148
		if (rc)
			return rc;

2149 2150 2151
		if (dev->num_tc)
			netif_setup_tc(dev, txq);

2152
		if (txq < dev->real_num_tx_queues) {
2153
			qdisc_reset_all_tx_gt(dev, txq);
2154 2155 2156 2157
#ifdef CONFIG_XPS
			netif_reset_xps_queues_gt(dev, txq);
#endif
		}
2158
	}
2159 2160 2161

	dev->real_num_tx_queues = txq;
	return 0;
2162 2163
}
EXPORT_SYMBOL(netif_set_real_num_tx_queues);
2164

2165
#ifdef CONFIG_SYSFS
2166 2167 2168 2169 2170 2171 2172
/**
 *	netif_set_real_num_rx_queues - set actual number of RX queues used
 *	@dev: Network device
 *	@rxq: Actual number of RX queues
 *
 *	This must be called either with the rtnl_lock held or before
 *	registration of the net device.  Returns 0 on success, or a
2173 2174
 *	negative error code.  If called before registration, it always
 *	succeeds.
2175 2176 2177 2178 2179
 */
int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
{
	int rc;

T
Tom Herbert 已提交
2180 2181 2182
	if (rxq < 1 || rxq > dev->num_rx_queues)
		return -EINVAL;

2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197
	if (dev->reg_state == NETREG_REGISTERED) {
		ASSERT_RTNL();

		rc = net_rx_queue_update_kobjects(dev, dev->real_num_rx_queues,
						  rxq);
		if (rc)
			return rc;
	}

	dev->real_num_rx_queues = rxq;
	return 0;
}
EXPORT_SYMBOL(netif_set_real_num_rx_queues);
#endif

2198 2199
/**
 * netif_get_num_default_rss_queues - default number of RSS queues
2200 2201 2202 2203
 *
 * This routine should set an upper limit on the number of RSS queues
 * used by default by multiqueue devices.
 */
2204
int netif_get_num_default_rss_queues(void)
2205 2206 2207 2208 2209
{
	return min_t(int, DEFAULT_MAX_NUM_RSS_QUEUES, num_online_cpus());
}
EXPORT_SYMBOL(netif_get_num_default_rss_queues);

2210
static inline void __netif_reschedule(struct Qdisc *q)
2211
{
2212 2213
	struct softnet_data *sd;
	unsigned long flags;
2214

2215
	local_irq_save(flags);
2216
	sd = this_cpu_ptr(&softnet_data);
2217 2218 2219
	q->next_sched = NULL;
	*sd->output_queue_tailp = q;
	sd->output_queue_tailp = &q->next_sched;
2220 2221 2222 2223 2224 2225 2226 2227
	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_restore(flags);
}

void __netif_schedule(struct Qdisc *q)
{
	if (!test_and_set_bit(__QDISC_STATE_SCHED, &q->state))
		__netif_reschedule(q);
2228 2229 2230
}
EXPORT_SYMBOL(__netif_schedule);

2231 2232 2233 2234 2235
struct dev_kfree_skb_cb {
	enum skb_free_reason reason;
};

static struct dev_kfree_skb_cb *get_kfree_skb_cb(const struct sk_buff *skb)
2236
{
2237 2238 2239
	return (struct dev_kfree_skb_cb *)skb->cb;
}

2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
void netif_schedule_queue(struct netdev_queue *txq)
{
	rcu_read_lock();
	if (!(txq->state & QUEUE_STATE_ANY_XOFF)) {
		struct Qdisc *q = rcu_dereference(txq->qdisc);

		__netif_schedule(q);
	}
	rcu_read_unlock();
}
EXPORT_SYMBOL(netif_schedule_queue);

/**
 *	netif_wake_subqueue - allow sending packets on subqueue
 *	@dev: network device
 *	@queue_index: sub queue index
 *
 * Resume individual transmit queue of a device with multiple transmit queues.
 */
void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
{
	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);

	if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state)) {
		struct Qdisc *q;

		rcu_read_lock();
		q = rcu_dereference(txq->qdisc);
		__netif_schedule(q);
		rcu_read_unlock();
	}
}
EXPORT_SYMBOL(netif_wake_subqueue);

void netif_tx_wake_queue(struct netdev_queue *dev_queue)
{
	if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state)) {
		struct Qdisc *q;

		rcu_read_lock();
		q = rcu_dereference(dev_queue->qdisc);
		__netif_schedule(q);
		rcu_read_unlock();
	}
}
EXPORT_SYMBOL(netif_tx_wake_queue);

2287
void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason)
2288
{
2289
	unsigned long flags;
2290

2291 2292 2293 2294 2295
	if (likely(atomic_read(&skb->users) == 1)) {
		smp_rmb();
		atomic_set(&skb->users, 0);
	} else if (likely(!atomic_dec_and_test(&skb->users))) {
		return;
2296
	}
2297 2298 2299 2300 2301 2302
	get_kfree_skb_cb(skb)->reason = reason;
	local_irq_save(flags);
	skb->next = __this_cpu_read(softnet_data.completion_queue);
	__this_cpu_write(softnet_data.completion_queue, skb);
	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_restore(flags);
2303
}
2304
EXPORT_SYMBOL(__dev_kfree_skb_irq);
2305

2306
void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason)
2307 2308
{
	if (in_irq() || irqs_disabled())
2309
		__dev_kfree_skb_irq(skb, reason);
2310 2311 2312
	else
		dev_kfree_skb(skb);
}
2313
EXPORT_SYMBOL(__dev_kfree_skb_any);
2314 2315


2316 2317 2318 2319 2320 2321
/**
 * netif_device_detach - mark device as removed
 * @dev: network device
 *
 * Mark device as removed from system and therefore no longer available.
 */
2322 2323 2324 2325
void netif_device_detach(struct net_device *dev)
{
	if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2326
		netif_tx_stop_all_queues(dev);
2327 2328 2329 2330
	}
}
EXPORT_SYMBOL(netif_device_detach);

2331 2332 2333 2334 2335 2336
/**
 * netif_device_attach - mark device as attached
 * @dev: network device
 *
 * Mark device as attached from system and restart if needed.
 */
2337 2338 2339 2340
void netif_device_attach(struct net_device *dev)
{
	if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2341
		netif_tx_wake_all_queues(dev);
2342
		__netdev_watchdog_up(dev);
2343 2344 2345 2346
	}
}
EXPORT_SYMBOL(netif_device_attach);

J
Jiri Pirko 已提交
2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374
/*
 * Returns a Tx hash based on the given packet descriptor a Tx queues' number
 * to be used as a distribution range.
 */
u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
		  unsigned int num_tx_queues)
{
	u32 hash;
	u16 qoffset = 0;
	u16 qcount = num_tx_queues;

	if (skb_rx_queue_recorded(skb)) {
		hash = skb_get_rx_queue(skb);
		while (unlikely(hash >= num_tx_queues))
			hash -= num_tx_queues;
		return hash;
	}

	if (dev->num_tc) {
		u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
		qoffset = dev->tc_to_txq[tc].offset;
		qcount = dev->tc_to_txq[tc].count;
	}

	return (u16) reciprocal_scale(skb_get_hash(skb), qcount) + qoffset;
}
EXPORT_SYMBOL(__skb_tx_hash);

2375 2376
static void skb_warn_bad_offload(const struct sk_buff *skb)
{
2377
	static const netdev_features_t null_features = 0;
2378 2379 2380
	struct net_device *dev = skb->dev;
	const char *driver = "";

2381 2382 2383
	if (!net_ratelimit())
		return;

2384 2385 2386 2387 2388
	if (dev && dev->dev.parent)
		driver = dev_driver_string(dev->dev.parent);

	WARN(1, "%s: caps=(%pNF, %pNF) len=%d data_len=%d gso_size=%d "
	     "gso_type=%d ip_summed=%d\n",
2389 2390
	     driver, dev ? &dev->features : &null_features,
	     skb->sk ? &skb->sk->sk_route_caps : &null_features,
2391 2392 2393 2394
	     skb->len, skb->data_len, skb_shinfo(skb)->gso_size,
	     skb_shinfo(skb)->gso_type, skb->ip_summed);
}

L
Linus Torvalds 已提交
2395 2396 2397 2398
/*
 * Invalidate hardware checksum when packet is to be mangled, and
 * complete checksum manually on outgoing path.
 */
2399
int skb_checksum_help(struct sk_buff *skb)
L
Linus Torvalds 已提交
2400
{
2401
	__wsum csum;
2402
	int ret = 0, offset;
L
Linus Torvalds 已提交
2403

2404
	if (skb->ip_summed == CHECKSUM_COMPLETE)
2405 2406 2407
		goto out_set_summed;

	if (unlikely(skb_shinfo(skb)->gso_size)) {
2408 2409
		skb_warn_bad_offload(skb);
		return -EINVAL;
L
Linus Torvalds 已提交
2410 2411
	}

2412 2413 2414 2415 2416 2417 2418 2419 2420
	/* Before computing a checksum, we should make sure no frag could
	 * be modified by an external entity : checksum could be wrong.
	 */
	if (skb_has_shared_frag(skb)) {
		ret = __skb_linearize(skb);
		if (ret)
			goto out;
	}

2421
	offset = skb_checksum_start_offset(skb);
2422 2423 2424 2425 2426 2427 2428 2429
	BUG_ON(offset >= skb_headlen(skb));
	csum = skb_checksum(skb, offset, skb->len - offset, 0);

	offset += skb->csum_offset;
	BUG_ON(offset + sizeof(__sum16) > skb_headlen(skb));

	if (skb_cloned(skb) &&
	    !skb_clone_writable(skb, offset + sizeof(__sum16))) {
L
Linus Torvalds 已提交
2430 2431 2432 2433 2434
		ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
		if (ret)
			goto out;
	}

2435
	*(__sum16 *)(skb->data + offset) = csum_fold(csum);
2436
out_set_summed:
L
Linus Torvalds 已提交
2437
	skb->ip_summed = CHECKSUM_NONE;
2438
out:
L
Linus Torvalds 已提交
2439 2440
	return ret;
}
E
Eric Dumazet 已提交
2441
EXPORT_SYMBOL(skb_checksum_help);
L
Linus Torvalds 已提交
2442

2443
__be16 skb_network_protocol(struct sk_buff *skb, int *depth)
2444
{
A
Al Viro 已提交
2445
	__be16 type = skb->protocol;
2446

2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457
	/* Tunnel gso handlers can set protocol to ethernet. */
	if (type == htons(ETH_P_TEB)) {
		struct ethhdr *eth;

		if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
			return 0;

		eth = (struct ethhdr *)skb_mac_header(skb);
		type = eth->h_proto;
	}

2458
	return __vlan_get_protocol(skb, type, depth);
2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470
}

/**
 *	skb_mac_gso_segment - mac layer segmentation handler.
 *	@skb: buffer to segment
 *	@features: features for the output path (see dev->features)
 */
struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
				    netdev_features_t features)
{
	struct sk_buff *segs = ERR_PTR(-EPROTONOSUPPORT);
	struct packet_offload *ptype;
2471 2472
	int vlan_depth = skb->mac_len;
	__be16 type = skb_network_protocol(skb, &vlan_depth);
2473 2474 2475 2476

	if (unlikely(!type))
		return ERR_PTR(-EINVAL);

2477
	__skb_pull(skb, vlan_depth);
2478 2479

	rcu_read_lock();
2480
	list_for_each_entry_rcu(ptype, &offload_base, list) {
2481 2482
		if (ptype->type == type && ptype->callbacks.gso_segment) {
			segs = ptype->callbacks.gso_segment(skb, features);
2483 2484 2485 2486 2487
			break;
		}
	}
	rcu_read_unlock();

2488
	__skb_push(skb, skb->data - skb_mac_header(skb));
2489

2490 2491
	return segs;
}
2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523
EXPORT_SYMBOL(skb_mac_gso_segment);


/* openvswitch calls this on rx path, so we need a different check.
 */
static inline bool skb_needs_check(struct sk_buff *skb, bool tx_path)
{
	if (tx_path)
		return skb->ip_summed != CHECKSUM_PARTIAL;
	else
		return skb->ip_summed == CHECKSUM_NONE;
}

/**
 *	__skb_gso_segment - Perform segmentation on skb.
 *	@skb: buffer to segment
 *	@features: features for the output path (see dev->features)
 *	@tx_path: whether it is called in TX path
 *
 *	This function segments the given skb and returns a list of segments.
 *
 *	It may return NULL if the skb requires no segmentation.  This is
 *	only possible when GSO is used for verifying header integrity.
 */
struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
				  netdev_features_t features, bool tx_path)
{
	if (unlikely(skb_needs_check(skb, tx_path))) {
		int err;

		skb_warn_bad_offload(skb);

2524 2525
		err = skb_cow_head(skb, 0);
		if (err < 0)
2526 2527 2528
			return ERR_PTR(err);
	}

2529
	SKB_GSO_CB(skb)->mac_offset = skb_headroom(skb);
2530 2531
	SKB_GSO_CB(skb)->encap_level = 0;

2532 2533 2534 2535 2536
	skb_reset_mac_header(skb);
	skb_reset_mac_len(skb);

	return skb_mac_gso_segment(skb, features);
}
2537
EXPORT_SYMBOL(__skb_gso_segment);
2538

2539 2540 2541 2542 2543
/* Take action when hardware reception checksum errors are detected. */
#ifdef CONFIG_BUG
void netdev_rx_csum_fault(struct net_device *dev)
{
	if (net_ratelimit()) {
2544
		pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>");
2545 2546 2547 2548 2549 2550
		dump_stack();
	}
}
EXPORT_SYMBOL(netdev_rx_csum_fault);
#endif

L
Linus Torvalds 已提交
2551 2552 2553 2554 2555
/* Actually, we should eliminate this check as soon as we know, that:
 * 1. IOMMU is present and allows to map all the memory.
 * 2. No high memory really exists on this machine.
 */

2556
static int illegal_highdma(struct net_device *dev, struct sk_buff *skb)
L
Linus Torvalds 已提交
2557
{
2558
#ifdef CONFIG_HIGHMEM
L
Linus Torvalds 已提交
2559
	int i;
2560
	if (!(dev->features & NETIF_F_HIGHDMA)) {
2561 2562 2563
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			if (PageHighMem(skb_frag_page(frag)))
2564
				return 1;
2565
		}
2566
	}
L
Linus Torvalds 已提交
2567

2568 2569
	if (PCI_DMA_BUS_IS_PHYS) {
		struct device *pdev = dev->dev.parent;
L
Linus Torvalds 已提交
2570

E
Eric Dumazet 已提交
2571 2572
		if (!pdev)
			return 0;
2573
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2574 2575
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			dma_addr_t addr = page_to_phys(skb_frag_page(frag));
2576 2577 2578 2579
			if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask)
				return 1;
		}
	}
2580
#endif
L
Linus Torvalds 已提交
2581 2582 2583
	return 0;
}

2584 2585 2586
/* If MPLS offload request, verify we are testing hardware MPLS features
 * instead of standard features for the netdev.
 */
P
Pravin B Shelar 已提交
2587
#if IS_ENABLED(CONFIG_NET_MPLS_GSO)
2588 2589 2590 2591
static netdev_features_t net_mpls_features(struct sk_buff *skb,
					   netdev_features_t features,
					   __be16 type)
{
2592
	if (eth_p_mpls(type))
2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605
		features &= skb->dev->mpls_features;

	return features;
}
#else
static netdev_features_t net_mpls_features(struct sk_buff *skb,
					   netdev_features_t features,
					   __be16 type)
{
	return features;
}
#endif

2606
static netdev_features_t harmonize_features(struct sk_buff *skb,
2607
	netdev_features_t features)
2608
{
2609
	int tmp;
2610 2611 2612 2613
	__be16 type;

	type = skb_network_protocol(skb, &tmp);
	features = net_mpls_features(skb, features, type);
2614

2615
	if (skb->ip_summed != CHECKSUM_NONE &&
2616
	    !can_checksum_protocol(features, type)) {
2617
		features &= ~NETIF_F_ALL_CSUM;
2618
	} else if (illegal_highdma(skb->dev, skb)) {
2619 2620 2621 2622 2623 2624
		features &= ~NETIF_F_SG;
	}

	return features;
}

2625 2626 2627 2628 2629 2630 2631 2632
netdev_features_t passthru_features_check(struct sk_buff *skb,
					  struct net_device *dev,
					  netdev_features_t features)
{
	return features;
}
EXPORT_SYMBOL(passthru_features_check);

2633 2634 2635 2636 2637 2638 2639
static netdev_features_t dflt_features_check(const struct sk_buff *skb,
					     struct net_device *dev,
					     netdev_features_t features)
{
	return vlan_features_check(skb, features);
}

2640
netdev_features_t netif_skb_features(struct sk_buff *skb)
2641
{
2642
	struct net_device *dev = skb->dev;
2643 2644
	netdev_features_t features = dev->features;
	u16 gso_segs = skb_shinfo(skb)->gso_segs;
2645

2646
	if (gso_segs > dev->gso_max_segs || gso_segs < dev->gso_min_segs)
2647 2648
		features &= ~NETIF_F_GSO_MASK;

2649 2650 2651 2652 2653 2654 2655
	/* If encapsulation offload request, verify we are testing
	 * hardware encapsulation features instead of standard
	 * features for the netdev
	 */
	if (skb->encapsulation)
		features &= dev->hw_enc_features;

2656 2657 2658 2659 2660
	if (skb_vlan_tagged(skb))
		features = netdev_intersect_features(features,
						     dev->vlan_features |
						     NETIF_F_HW_VLAN_CTAG_TX |
						     NETIF_F_HW_VLAN_STAG_TX);
2661

2662 2663 2664
	if (dev->netdev_ops->ndo_features_check)
		features &= dev->netdev_ops->ndo_features_check(skb, dev,
								features);
2665 2666
	else
		features &= dflt_features_check(skb, dev, features);
2667

2668
	return harmonize_features(skb, features);
2669
}
2670
EXPORT_SYMBOL(netif_skb_features);
2671

2672
static int xmit_one(struct sk_buff *skb, struct net_device *dev,
2673
		    struct netdev_queue *txq, bool more)
2674
{
2675 2676
	unsigned int len;
	int rc;
2677

2678
	if (!list_empty(&ptype_all) || !list_empty(&dev->ptype_all))
2679
		dev_queue_xmit_nit(skb, dev);
2680

2681 2682
	len = skb->len;
	trace_net_dev_start_xmit(skb, dev);
2683
	rc = netdev_start_xmit(skb, dev, txq, more);
2684
	trace_net_dev_xmit(skb, rc, dev, len);
E
Eric Dumazet 已提交
2685

2686 2687
	return rc;
}
2688

2689 2690
struct sk_buff *dev_hard_start_xmit(struct sk_buff *first, struct net_device *dev,
				    struct netdev_queue *txq, int *ret)
2691 2692 2693
{
	struct sk_buff *skb = first;
	int rc = NETDEV_TX_OK;
2694

2695 2696
	while (skb) {
		struct sk_buff *next = skb->next;
2697

2698
		skb->next = NULL;
2699
		rc = xmit_one(skb, dev, txq, next != NULL);
2700 2701 2702 2703
		if (unlikely(!dev_xmit_complete(rc))) {
			skb->next = next;
			goto out;
		}
2704

2705 2706 2707 2708
		skb = next;
		if (netif_xmit_stopped(txq) && skb) {
			rc = NETDEV_TX_BUSY;
			break;
2709
		}
2710
	}
2711

2712 2713 2714 2715
out:
	*ret = rc;
	return skb;
}
2716

E
Eric Dumazet 已提交
2717 2718
static struct sk_buff *validate_xmit_vlan(struct sk_buff *skb,
					  netdev_features_t features)
2719
{
2720
	if (skb_vlan_tag_present(skb) &&
2721 2722
	    !vlan_hw_offload_capable(features, skb->vlan_proto))
		skb = __vlan_hwaccel_push_inside(skb);
2723 2724
	return skb;
}
2725

2726
static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev)
2727 2728
{
	netdev_features_t features;
2729

2730 2731
	if (skb->next)
		return skb;
2732

2733 2734 2735 2736
	features = netif_skb_features(skb);
	skb = validate_xmit_vlan(skb, features);
	if (unlikely(!skb))
		goto out_null;
2737

2738
	if (netif_needs_gso(skb, features)) {
2739 2740 2741
		struct sk_buff *segs;

		segs = skb_gso_segment(skb, features);
2742
		if (IS_ERR(segs)) {
2743
			goto out_kfree_skb;
2744 2745 2746
		} else if (segs) {
			consume_skb(skb);
			skb = segs;
2747
		}
2748 2749 2750 2751
	} else {
		if (skb_needs_linearize(skb, features) &&
		    __skb_linearize(skb))
			goto out_kfree_skb;
2752

2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766
		/* If packet is not checksummed and device does not
		 * support checksumming for this protocol, complete
		 * checksumming here.
		 */
		if (skb->ip_summed == CHECKSUM_PARTIAL) {
			if (skb->encapsulation)
				skb_set_inner_transport_header(skb,
							       skb_checksum_start_offset(skb));
			else
				skb_set_transport_header(skb,
							 skb_checksum_start_offset(skb));
			if (!(features & NETIF_F_ALL_CSUM) &&
			    skb_checksum_help(skb))
				goto out_kfree_skb;
2767
		}
2768
	}
2769

2770
	return skb;
2771

2772 2773
out_kfree_skb:
	kfree_skb(skb);
2774 2775 2776
out_null:
	return NULL;
}
2777

2778 2779 2780 2781
struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev)
{
	struct sk_buff *next, *head = NULL, *tail;

2782
	for (; skb != NULL; skb = next) {
2783 2784
		next = skb->next;
		skb->next = NULL;
2785 2786 2787 2788

		/* in case skb wont be segmented, point to itself */
		skb->prev = skb;

2789
		skb = validate_xmit_skb(skb, dev);
2790 2791
		if (!skb)
			continue;
2792

2793 2794 2795 2796 2797 2798 2799 2800
		if (!head)
			head = skb;
		else
			tail->next = skb;
		/* If skb was segmented, skb->prev points to
		 * the last segment. If not, it still contains skb.
		 */
		tail = skb->prev;
2801 2802
	}
	return head;
2803 2804
}

2805 2806 2807 2808 2809 2810 2811 2812 2813 2814
static void qdisc_pkt_len_init(struct sk_buff *skb)
{
	const struct skb_shared_info *shinfo = skb_shinfo(skb);

	qdisc_skb_cb(skb)->pkt_len = skb->len;

	/* To get more precise estimation of bytes sent on wire,
	 * we add to pkt_len the headers size of all segments
	 */
	if (shinfo->gso_size)  {
2815
		unsigned int hdr_len;
2816
		u16 gso_segs = shinfo->gso_segs;
2817

2818 2819 2820 2821
		/* mac layer + network layer */
		hdr_len = skb_transport_header(skb) - skb_mac_header(skb);

		/* + transport layer */
2822 2823 2824 2825
		if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)))
			hdr_len += tcp_hdrlen(skb);
		else
			hdr_len += sizeof(struct udphdr);
2826 2827 2828 2829 2830 2831

		if (shinfo->gso_type & SKB_GSO_DODGY)
			gso_segs = DIV_ROUND_UP(skb->len - hdr_len,
						shinfo->gso_size);

		qdisc_skb_cb(skb)->pkt_len += (gso_segs - 1) * hdr_len;
2832 2833 2834
	}
}

2835 2836 2837 2838 2839
static inline int __dev_xmit_skb(struct sk_buff *skb, struct Qdisc *q,
				 struct net_device *dev,
				 struct netdev_queue *txq)
{
	spinlock_t *root_lock = qdisc_lock(q);
E
Eric Dumazet 已提交
2840
	bool contended;
2841 2842
	int rc;

2843
	qdisc_pkt_len_init(skb);
E
Eric Dumazet 已提交
2844
	qdisc_calculate_pkt_len(skb, q);
2845 2846 2847
	/*
	 * Heuristic to force contended enqueues to serialize on a
	 * separate lock before trying to get qdisc main lock.
Y
Ying Xue 已提交
2848 2849
	 * This permits __QDISC___STATE_RUNNING owner to get the lock more
	 * often and dequeue packets faster.
2850
	 */
E
Eric Dumazet 已提交
2851
	contended = qdisc_is_running(q);
2852 2853 2854
	if (unlikely(contended))
		spin_lock(&q->busylock);

2855 2856 2857 2858 2859
	spin_lock(root_lock);
	if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) {
		kfree_skb(skb);
		rc = NET_XMIT_DROP;
	} else if ((q->flags & TCQ_F_CAN_BYPASS) && !qdisc_qlen(q) &&
2860
		   qdisc_run_begin(q)) {
2861 2862 2863 2864 2865
		/*
		 * This is a work-conserving queue; there are no old skbs
		 * waiting to be sent out; and the qdisc is not running -
		 * xmit the skb directly.
		 */
2866 2867 2868

		qdisc_bstats_update(q, skb);

2869
		if (sch_direct_xmit(skb, q, dev, txq, root_lock, true)) {
2870 2871 2872 2873
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
2874
			__qdisc_run(q);
2875
		} else
2876
			qdisc_run_end(q);
2877 2878 2879

		rc = NET_XMIT_SUCCESS;
	} else {
E
Eric Dumazet 已提交
2880
		rc = q->enqueue(skb, q) & NET_XMIT_MASK;
2881 2882 2883 2884 2885 2886 2887
		if (qdisc_run_begin(q)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
			__qdisc_run(q);
		}
2888 2889
	}
	spin_unlock(root_lock);
2890 2891
	if (unlikely(contended))
		spin_unlock(&q->busylock);
2892 2893 2894
	return rc;
}

2895
#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2896 2897
static void skb_update_prio(struct sk_buff *skb)
{
I
Igor Maravic 已提交
2898
	struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap);
2899

2900 2901 2902 2903 2904 2905
	if (!skb->priority && skb->sk && map) {
		unsigned int prioidx = skb->sk->sk_cgrp_prioidx;

		if (prioidx < map->priomap_len)
			skb->priority = map->priomap[prioidx];
	}
2906 2907 2908 2909 2910
}
#else
#define skb_update_prio(skb)
#endif

2911 2912 2913
DEFINE_PER_CPU(int, xmit_recursion);
EXPORT_SYMBOL(xmit_recursion);

2914
#define RECURSION_LIMIT 10
2915

2916 2917
/**
 *	dev_loopback_xmit - loop back @skb
2918 2919
 *	@net: network namespace this loopback is happening in
 *	@sk:  sk needed to be a netfilter okfn
2920 2921
 *	@skb: buffer to transmit
 */
2922
int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934
{
	skb_reset_mac_header(skb);
	__skb_pull(skb, skb_network_offset(skb));
	skb->pkt_type = PACKET_LOOPBACK;
	skb->ip_summed = CHECKSUM_UNNECESSARY;
	WARN_ON(!skb_dst(skb));
	skb_dst_force(skb);
	netif_rx_ni(skb);
	return 0;
}
EXPORT_SYMBOL(dev_loopback_xmit);

2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012
static inline int get_xps_queue(struct net_device *dev, struct sk_buff *skb)
{
#ifdef CONFIG_XPS
	struct xps_dev_maps *dev_maps;
	struct xps_map *map;
	int queue_index = -1;

	rcu_read_lock();
	dev_maps = rcu_dereference(dev->xps_maps);
	if (dev_maps) {
		map = rcu_dereference(
		    dev_maps->cpu_map[skb->sender_cpu - 1]);
		if (map) {
			if (map->len == 1)
				queue_index = map->queues[0];
			else
				queue_index = map->queues[reciprocal_scale(skb_get_hash(skb),
									   map->len)];
			if (unlikely(queue_index >= dev->real_num_tx_queues))
				queue_index = -1;
		}
	}
	rcu_read_unlock();

	return queue_index;
#else
	return -1;
#endif
}

static u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb)
{
	struct sock *sk = skb->sk;
	int queue_index = sk_tx_queue_get(sk);

	if (queue_index < 0 || skb->ooo_okay ||
	    queue_index >= dev->real_num_tx_queues) {
		int new_index = get_xps_queue(dev, skb);
		if (new_index < 0)
			new_index = skb_tx_hash(dev, skb);

		if (queue_index != new_index && sk &&
		    rcu_access_pointer(sk->sk_dst_cache))
			sk_tx_queue_set(sk, new_index);

		queue_index = new_index;
	}

	return queue_index;
}

struct netdev_queue *netdev_pick_tx(struct net_device *dev,
				    struct sk_buff *skb,
				    void *accel_priv)
{
	int queue_index = 0;

#ifdef CONFIG_XPS
	if (skb->sender_cpu == 0)
		skb->sender_cpu = raw_smp_processor_id() + 1;
#endif

	if (dev->real_num_tx_queues != 1) {
		const struct net_device_ops *ops = dev->netdev_ops;
		if (ops->ndo_select_queue)
			queue_index = ops->ndo_select_queue(dev, skb, accel_priv,
							    __netdev_pick_tx);
		else
			queue_index = __netdev_pick_tx(dev, skb);

		if (!accel_priv)
			queue_index = netdev_cap_txqueue(dev, queue_index);
	}

	skb_set_queue_mapping(skb, queue_index);
	return netdev_get_tx_queue(dev, queue_index);
}

3013
/**
3014
 *	__dev_queue_xmit - transmit a buffer
3015
 *	@skb: buffer to transmit
3016
 *	@accel_priv: private data used for L2 forwarding offload
3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038
 *
 *	Queue a buffer for transmission to a network device. The caller must
 *	have set the device and priority and built the buffer before calling
 *	this function. The function can be called from an interrupt.
 *
 *	A negative errno code is returned on a failure. A success does not
 *	guarantee the frame will be transmitted as it may be dropped due
 *	to congestion or traffic shaping.
 *
 * -----------------------------------------------------------------------------------
 *      I notice this method can also return errors from the queue disciplines,
 *      including NET_XMIT_DROP, which is a positive value.  So, errors can also
 *      be positive.
 *
 *      Regardless of the return value, the skb is consumed, so it is currently
 *      difficult to retry a send to this method.  (You can bump the ref count
 *      before sending to hold a reference for retry if you are careful.)
 *
 *      When calling this method, interrupts MUST be enabled.  This is because
 *      the BH enable code must have IRQs enabled so that it will not deadlock.
 *          --BLG
 */
3039
static int __dev_queue_xmit(struct sk_buff *skb, void *accel_priv)
L
Linus Torvalds 已提交
3040 3041
{
	struct net_device *dev = skb->dev;
3042
	struct netdev_queue *txq;
L
Linus Torvalds 已提交
3043 3044 3045
	struct Qdisc *q;
	int rc = -ENOMEM;

3046 3047
	skb_reset_mac_header(skb);

3048 3049 3050
	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_SCHED_TSTAMP))
		__skb_tstamp_tx(skb, NULL, skb->sk, SCM_TSTAMP_SCHED);

3051 3052
	/* Disable soft irqs for various locks below. Also
	 * stops preemption for RCU.
L
Linus Torvalds 已提交
3053
	 */
3054
	rcu_read_lock_bh();
L
Linus Torvalds 已提交
3055

3056 3057
	skb_update_prio(skb);

3058 3059 3060 3061 3062 3063 3064 3065
	/* If device/qdisc don't need skb->dst, release it right now while
	 * its hot in this cpu cache.
	 */
	if (dev->priv_flags & IFF_XMIT_DST_RELEASE)
		skb_dst_drop(skb);
	else
		skb_dst_force(skb);

3066 3067 3068 3069 3070 3071 3072 3073 3074 3075
#ifdef CONFIG_NET_SWITCHDEV
	/* Don't forward if offload device already forwarded */
	if (skb->offload_fwd_mark &&
	    skb->offload_fwd_mark == dev->offload_fwd_mark) {
		consume_skb(skb);
		rc = NET_XMIT_SUCCESS;
		goto out;
	}
#endif

3076
	txq = netdev_pick_tx(dev, skb, accel_priv);
3077
	q = rcu_dereference_bh(txq->qdisc);
3078

L
Linus Torvalds 已提交
3079
#ifdef CONFIG_NET_CLS_ACT
E
Eric Dumazet 已提交
3080
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS);
L
Linus Torvalds 已提交
3081
#endif
3082
	trace_net_dev_queue(skb);
L
Linus Torvalds 已提交
3083
	if (q->enqueue) {
3084
		rc = __dev_xmit_skb(skb, q, dev, txq);
3085
		goto out;
L
Linus Torvalds 已提交
3086 3087 3088 3089 3090
	}

	/* The device has no queue. Common case for software devices:
	   loopback, all the sorts of tunnels...

H
Herbert Xu 已提交
3091 3092
	   Really, it is unlikely that netif_tx_lock protection is necessary
	   here.  (f.e. loopback and IP tunnels are clean ignoring statistics
L
Linus Torvalds 已提交
3093 3094 3095 3096 3097 3098 3099 3100 3101 3102
	   counters.)
	   However, it is possible, that they rely on protection
	   made by us here.

	   Check this and shot the lock. It is not prone from deadlocks.
	   Either shot noqueue qdisc, it is even simpler 8)
	 */
	if (dev->flags & IFF_UP) {
		int cpu = smp_processor_id(); /* ok because BHs are off */

3103
		if (txq->xmit_lock_owner != cpu) {
L
Linus Torvalds 已提交
3104

3105 3106 3107
			if (__this_cpu_read(xmit_recursion) > RECURSION_LIMIT)
				goto recursion_alert;

3108 3109 3110 3111
			skb = validate_xmit_skb(skb, dev);
			if (!skb)
				goto drop;

3112
			HARD_TX_LOCK(dev, txq, cpu);
L
Linus Torvalds 已提交
3113

3114
			if (!netif_xmit_stopped(txq)) {
3115
				__this_cpu_inc(xmit_recursion);
3116
				skb = dev_hard_start_xmit(skb, dev, txq, &rc);
3117
				__this_cpu_dec(xmit_recursion);
3118
				if (dev_xmit_complete(rc)) {
3119
					HARD_TX_UNLOCK(dev, txq);
L
Linus Torvalds 已提交
3120 3121 3122
					goto out;
				}
			}
3123
			HARD_TX_UNLOCK(dev, txq);
3124 3125
			net_crit_ratelimited("Virtual device %s asks to queue packet!\n",
					     dev->name);
L
Linus Torvalds 已提交
3126 3127
		} else {
			/* Recursion is detected! It is possible,
3128 3129 3130
			 * unfortunately
			 */
recursion_alert:
3131 3132
			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
					     dev->name);
L
Linus Torvalds 已提交
3133 3134 3135 3136
		}
	}

	rc = -ENETDOWN;
3137
drop:
3138
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
3139

3140
	atomic_long_inc(&dev->tx_dropped);
3141
	kfree_skb_list(skb);
L
Linus Torvalds 已提交
3142 3143
	return rc;
out:
3144
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
3145 3146
	return rc;
}
3147

3148
int dev_queue_xmit(struct sk_buff *skb)
3149 3150 3151
{
	return __dev_queue_xmit(skb, NULL);
}
3152
EXPORT_SYMBOL(dev_queue_xmit);
L
Linus Torvalds 已提交
3153

3154 3155 3156 3157 3158 3159
int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv)
{
	return __dev_queue_xmit(skb, accel_priv);
}
EXPORT_SYMBOL(dev_queue_xmit_accel);

L
Linus Torvalds 已提交
3160 3161 3162 3163 3164

/*=======================================================================
			Receiver routines
  =======================================================================*/

3165
int netdev_max_backlog __read_mostly = 1000;
E
Eric Dumazet 已提交
3166 3167
EXPORT_SYMBOL(netdev_max_backlog);

E
Eric Dumazet 已提交
3168
int netdev_tstamp_prequeue __read_mostly = 1;
3169 3170
int netdev_budget __read_mostly = 300;
int weight_p __read_mostly = 64;            /* old backlog weight */
L
Linus Torvalds 已提交
3171

E
Eric Dumazet 已提交
3172 3173 3174 3175 3176 3177 3178 3179
/* Called with irq disabled */
static inline void ____napi_schedule(struct softnet_data *sd,
				     struct napi_struct *napi)
{
	list_add_tail(&napi->poll_list, &sd->poll_list);
	__raise_softirq_irqoff(NET_RX_SOFTIRQ);
}

3180 3181 3182
#ifdef CONFIG_RPS

/* One global table that all flow-based protocols share. */
E
Eric Dumazet 已提交
3183
struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
3184
EXPORT_SYMBOL(rps_sock_flow_table);
3185 3186
u32 rps_cpu_mask __read_mostly;
EXPORT_SYMBOL(rps_cpu_mask);
3187

3188
struct static_key rps_needed __read_mostly;
3189

3190 3191 3192 3193
static struct rps_dev_flow *
set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
	    struct rps_dev_flow *rflow, u16 next_cpu)
{
3194
	if (next_cpu < nr_cpu_ids) {
3195 3196 3197 3198 3199 3200 3201 3202 3203
#ifdef CONFIG_RFS_ACCEL
		struct netdev_rx_queue *rxqueue;
		struct rps_dev_flow_table *flow_table;
		struct rps_dev_flow *old_rflow;
		u32 flow_id;
		u16 rxq_index;
		int rc;

		/* Should we steer this flow to a different hardware queue? */
3204 3205
		if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
		    !(dev->features & NETIF_F_NTUPLE))
3206 3207 3208 3209 3210 3211 3212 3213 3214
			goto out;
		rxq_index = cpu_rmap_lookup_index(dev->rx_cpu_rmap, next_cpu);
		if (rxq_index == skb_get_rx_queue(skb))
			goto out;

		rxqueue = dev->_rx + rxq_index;
		flow_table = rcu_dereference(rxqueue->rps_flow_table);
		if (!flow_table)
			goto out;
3215
		flow_id = skb_get_hash(skb) & flow_table->mask;
3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227
		rc = dev->netdev_ops->ndo_rx_flow_steer(dev, skb,
							rxq_index, flow_id);
		if (rc < 0)
			goto out;
		old_rflow = rflow;
		rflow = &flow_table->flows[flow_id];
		rflow->filter = rc;
		if (old_rflow->filter == rflow->filter)
			old_rflow->filter = RPS_NO_FILTER;
	out:
#endif
		rflow->last_qtail =
3228
			per_cpu(softnet_data, next_cpu).input_queue_head;
3229 3230
	}

3231
	rflow->cpu = next_cpu;
3232 3233 3234
	return rflow;
}

3235 3236 3237 3238 3239 3240 3241 3242
/*
 * get_rps_cpu is called from netif_receive_skb and returns the target
 * CPU from the RPS map of the receiving queue for a given skb.
 * rcu_read_lock must be held on entry.
 */
static int get_rps_cpu(struct net_device *dev, struct sk_buff *skb,
		       struct rps_dev_flow **rflowp)
{
3243 3244
	const struct rps_sock_flow_table *sock_flow_table;
	struct netdev_rx_queue *rxqueue = dev->_rx;
3245
	struct rps_dev_flow_table *flow_table;
3246
	struct rps_map *map;
3247
	int cpu = -1;
3248
	u32 tcpu;
3249
	u32 hash;
3250 3251 3252

	if (skb_rx_queue_recorded(skb)) {
		u16 index = skb_get_rx_queue(skb);
3253

3254 3255 3256 3257 3258
		if (unlikely(index >= dev->real_num_rx_queues)) {
			WARN_ONCE(dev->real_num_rx_queues > 1,
				  "%s received packet on queue %u, but number "
				  "of RX queues is %u\n",
				  dev->name, index, dev->real_num_rx_queues);
3259 3260
			goto done;
		}
3261 3262
		rxqueue += index;
	}
3263

3264 3265 3266
	/* Avoid computing hash if RFS/RPS is not active for this rxqueue */

	flow_table = rcu_dereference(rxqueue->rps_flow_table);
E
Eric Dumazet 已提交
3267
	map = rcu_dereference(rxqueue->rps_map);
3268
	if (!flow_table && !map)
3269 3270
		goto done;

3271
	skb_reset_network_header(skb);
3272 3273
	hash = skb_get_hash(skb);
	if (!hash)
3274 3275
		goto done;

T
Tom Herbert 已提交
3276 3277 3278
	sock_flow_table = rcu_dereference(rps_sock_flow_table);
	if (flow_table && sock_flow_table) {
		struct rps_dev_flow *rflow;
3279 3280 3281 3282 3283 3284 3285
		u32 next_cpu;
		u32 ident;

		/* First check into global flow table if there is a match */
		ident = sock_flow_table->ents[hash & sock_flow_table->mask];
		if ((ident ^ hash) & ~rps_cpu_mask)
			goto try_rps;
T
Tom Herbert 已提交
3286

3287 3288 3289 3290 3291
		next_cpu = ident & rps_cpu_mask;

		/* OK, now we know there is a match,
		 * we can look at the local (per receive queue) flow table
		 */
3292
		rflow = &flow_table->flows[hash & flow_table->mask];
T
Tom Herbert 已提交
3293 3294 3295 3296 3297 3298
		tcpu = rflow->cpu;

		/*
		 * If the desired CPU (where last recvmsg was done) is
		 * different from current CPU (one in the rx-queue flow
		 * table entry), switch if one of the following holds:
3299
		 *   - Current CPU is unset (>= nr_cpu_ids).
T
Tom Herbert 已提交
3300 3301 3302 3303 3304 3305 3306
		 *   - Current CPU is offline.
		 *   - The current CPU's queue tail has advanced beyond the
		 *     last packet that was enqueued using this table entry.
		 *     This guarantees that all previous packets for the flow
		 *     have been dequeued, thus preserving in order delivery.
		 */
		if (unlikely(tcpu != next_cpu) &&
3307
		    (tcpu >= nr_cpu_ids || !cpu_online(tcpu) ||
T
Tom Herbert 已提交
3308
		     ((int)(per_cpu(softnet_data, tcpu).input_queue_head -
3309 3310
		      rflow->last_qtail)) >= 0)) {
			tcpu = next_cpu;
3311
			rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
3312
		}
3313

3314
		if (tcpu < nr_cpu_ids && cpu_online(tcpu)) {
T
Tom Herbert 已提交
3315 3316 3317 3318 3319 3320
			*rflowp = rflow;
			cpu = tcpu;
			goto done;
		}
	}

3321 3322
try_rps:

T
Tom Herbert 已提交
3323
	if (map) {
3324
		tcpu = map->cpus[reciprocal_scale(hash, map->len)];
T
Tom Herbert 已提交
3325 3326 3327 3328 3329 3330 3331 3332 3333 3334
		if (cpu_online(tcpu)) {
			cpu = tcpu;
			goto done;
		}
	}

done:
	return cpu;
}

3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354
#ifdef CONFIG_RFS_ACCEL

/**
 * rps_may_expire_flow - check whether an RFS hardware filter may be removed
 * @dev: Device on which the filter was set
 * @rxq_index: RX queue index
 * @flow_id: Flow ID passed to ndo_rx_flow_steer()
 * @filter_id: Filter ID returned by ndo_rx_flow_steer()
 *
 * Drivers that implement ndo_rx_flow_steer() should periodically call
 * this function for each installed filter and remove the filters for
 * which it returns %true.
 */
bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
			 u32 flow_id, u16 filter_id)
{
	struct netdev_rx_queue *rxqueue = dev->_rx + rxq_index;
	struct rps_dev_flow_table *flow_table;
	struct rps_dev_flow *rflow;
	bool expire = true;
3355
	unsigned int cpu;
3356 3357 3358 3359 3360 3361

	rcu_read_lock();
	flow_table = rcu_dereference(rxqueue->rps_flow_table);
	if (flow_table && flow_id <= flow_table->mask) {
		rflow = &flow_table->flows[flow_id];
		cpu = ACCESS_ONCE(rflow->cpu);
3362
		if (rflow->filter == filter_id && cpu < nr_cpu_ids &&
3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374
		    ((int)(per_cpu(softnet_data, cpu).input_queue_head -
			   rflow->last_qtail) <
		     (int)(10 * flow_table->mask)))
			expire = false;
	}
	rcu_read_unlock();
	return expire;
}
EXPORT_SYMBOL(rps_may_expire_flow);

#endif /* CONFIG_RFS_ACCEL */

T
Tom Herbert 已提交
3375
/* Called from hardirq (IPI) context */
E
Eric Dumazet 已提交
3376
static void rps_trigger_softirq(void *data)
T
Tom Herbert 已提交
3377
{
E
Eric Dumazet 已提交
3378 3379
	struct softnet_data *sd = data;

E
Eric Dumazet 已提交
3380
	____napi_schedule(sd, &sd->backlog);
C
Changli Gao 已提交
3381
	sd->received_rps++;
T
Tom Herbert 已提交
3382
}
E
Eric Dumazet 已提交
3383

T
Tom Herbert 已提交
3384
#endif /* CONFIG_RPS */
T
Tom Herbert 已提交
3385

E
Eric Dumazet 已提交
3386 3387 3388 3389 3390 3391 3392 3393
/*
 * Check if this softnet_data structure is another cpu one
 * If yes, queue it to our IPI list and return 1
 * If no, return 0
 */
static int rps_ipi_queued(struct softnet_data *sd)
{
#ifdef CONFIG_RPS
3394
	struct softnet_data *mysd = this_cpu_ptr(&softnet_data);
E
Eric Dumazet 已提交
3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406

	if (sd != mysd) {
		sd->rps_ipi_next = mysd->rps_ipi_list;
		mysd->rps_ipi_list = sd;

		__raise_softirq_irqoff(NET_RX_SOFTIRQ);
		return 1;
	}
#endif /* CONFIG_RPS */
	return 0;
}

3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420
#ifdef CONFIG_NET_FLOW_LIMIT
int netdev_flow_limit_table_len __read_mostly = (1 << 12);
#endif

static bool skb_flow_limit(struct sk_buff *skb, unsigned int qlen)
{
#ifdef CONFIG_NET_FLOW_LIMIT
	struct sd_flow_limit *fl;
	struct softnet_data *sd;
	unsigned int old_flow, new_flow;

	if (qlen < (netdev_max_backlog >> 1))
		return false;

3421
	sd = this_cpu_ptr(&softnet_data);
3422 3423 3424 3425

	rcu_read_lock();
	fl = rcu_dereference(sd->flow_limit);
	if (fl) {
3426
		new_flow = skb_get_hash(skb) & (fl->num_buckets - 1);
3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446
		old_flow = fl->history[fl->history_head];
		fl->history[fl->history_head] = new_flow;

		fl->history_head++;
		fl->history_head &= FLOW_LIMIT_HISTORY - 1;

		if (likely(fl->buckets[old_flow]))
			fl->buckets[old_flow]--;

		if (++fl->buckets[new_flow] > (FLOW_LIMIT_HISTORY >> 1)) {
			fl->count++;
			rcu_read_unlock();
			return true;
		}
	}
	rcu_read_unlock();
#endif
	return false;
}

T
Tom Herbert 已提交
3447 3448 3449 3450
/*
 * enqueue_to_backlog is called to queue an skb to a per CPU backlog
 * queue (may be a remote CPU queue).
 */
T
Tom Herbert 已提交
3451 3452
static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
			      unsigned int *qtail)
T
Tom Herbert 已提交
3453
{
E
Eric Dumazet 已提交
3454
	struct softnet_data *sd;
T
Tom Herbert 已提交
3455
	unsigned long flags;
3456
	unsigned int qlen;
T
Tom Herbert 已提交
3457

E
Eric Dumazet 已提交
3458
	sd = &per_cpu(softnet_data, cpu);
T
Tom Herbert 已提交
3459 3460 3461

	local_irq_save(flags);

E
Eric Dumazet 已提交
3462
	rps_lock(sd);
3463 3464
	if (!netif_running(skb->dev))
		goto drop;
3465 3466
	qlen = skb_queue_len(&sd->input_pkt_queue);
	if (qlen <= netdev_max_backlog && !skb_flow_limit(skb, qlen)) {
3467
		if (qlen) {
T
Tom Herbert 已提交
3468
enqueue:
E
Eric Dumazet 已提交
3469
			__skb_queue_tail(&sd->input_pkt_queue, skb);
3470
			input_queue_tail_incr_save(sd, qtail);
E
Eric Dumazet 已提交
3471
			rps_unlock(sd);
3472
			local_irq_restore(flags);
T
Tom Herbert 已提交
3473 3474 3475
			return NET_RX_SUCCESS;
		}

3476 3477 3478 3479
		/* Schedule NAPI for backlog device
		 * We can use non atomic operation since we own the queue lock
		 */
		if (!__test_and_set_bit(NAPI_STATE_SCHED, &sd->backlog.state)) {
E
Eric Dumazet 已提交
3480
			if (!rps_ipi_queued(sd))
E
Eric Dumazet 已提交
3481
				____napi_schedule(sd, &sd->backlog);
T
Tom Herbert 已提交
3482 3483 3484 3485
		}
		goto enqueue;
	}

3486
drop:
C
Changli Gao 已提交
3487
	sd->dropped++;
E
Eric Dumazet 已提交
3488
	rps_unlock(sd);
T
Tom Herbert 已提交
3489 3490 3491

	local_irq_restore(flags);

3492
	atomic_long_inc(&skb->dev->rx_dropped);
T
Tom Herbert 已提交
3493 3494 3495
	kfree_skb(skb);
	return NET_RX_DROP;
}
L
Linus Torvalds 已提交
3496

3497
static int netif_rx_internal(struct sk_buff *skb)
L
Linus Torvalds 已提交
3498
{
3499
	int ret;
L
Linus Torvalds 已提交
3500

3501
	net_timestamp_check(netdev_tstamp_prequeue, skb);
L
Linus Torvalds 已提交
3502

3503
	trace_netif_rx(skb);
E
Eric Dumazet 已提交
3504
#ifdef CONFIG_RPS
3505
	if (static_key_false(&rps_needed)) {
T
Tom Herbert 已提交
3506
		struct rps_dev_flow voidflow, *rflow = &voidflow;
3507 3508
		int cpu;

3509
		preempt_disable();
3510
		rcu_read_lock();
T
Tom Herbert 已提交
3511 3512

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
3513 3514
		if (cpu < 0)
			cpu = smp_processor_id();
T
Tom Herbert 已提交
3515 3516 3517

		ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);

3518
		rcu_read_unlock();
3519
		preempt_enable();
3520 3521
	} else
#endif
T
Tom Herbert 已提交
3522 3523 3524 3525 3526
	{
		unsigned int qtail;
		ret = enqueue_to_backlog(skb, get_cpu(), &qtail);
		put_cpu();
	}
3527
	return ret;
L
Linus Torvalds 已提交
3528
}
3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550

/**
 *	netif_rx	-	post buffer to the network code
 *	@skb: buffer to post
 *
 *	This function receives a packet from a device driver and queues it for
 *	the upper (protocol) levels to process.  It always succeeds. The buffer
 *	may be dropped during processing for congestion control or by the
 *	protocol layers.
 *
 *	return values:
 *	NET_RX_SUCCESS	(no congestion)
 *	NET_RX_DROP     (packet was dropped)
 *
 */

int netif_rx(struct sk_buff *skb)
{
	trace_netif_rx_entry(skb);

	return netif_rx_internal(skb);
}
E
Eric Dumazet 已提交
3551
EXPORT_SYMBOL(netif_rx);
L
Linus Torvalds 已提交
3552 3553 3554 3555 3556

int netif_rx_ni(struct sk_buff *skb)
{
	int err;

3557 3558
	trace_netif_rx_ni_entry(skb);

L
Linus Torvalds 已提交
3559
	preempt_disable();
3560
	err = netif_rx_internal(skb);
L
Linus Torvalds 已提交
3561 3562 3563 3564 3565 3566 3567 3568 3569 3570
	if (local_softirq_pending())
		do_softirq();
	preempt_enable();

	return err;
}
EXPORT_SYMBOL(netif_rx_ni);

static void net_tx_action(struct softirq_action *h)
{
3571
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
L
Linus Torvalds 已提交
3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584

	if (sd->completion_queue) {
		struct sk_buff *clist;

		local_irq_disable();
		clist = sd->completion_queue;
		sd->completion_queue = NULL;
		local_irq_enable();

		while (clist) {
			struct sk_buff *skb = clist;
			clist = clist->next;

3585
			WARN_ON(atomic_read(&skb->users));
3586 3587 3588 3589
			if (likely(get_kfree_skb_cb(skb)->reason == SKB_REASON_CONSUMED))
				trace_consume_skb(skb);
			else
				trace_kfree_skb(skb, net_tx_action);
L
Linus Torvalds 已提交
3590 3591 3592 3593 3594
			__kfree_skb(skb);
		}
	}

	if (sd->output_queue) {
3595
		struct Qdisc *head;
L
Linus Torvalds 已提交
3596 3597 3598 3599

		local_irq_disable();
		head = sd->output_queue;
		sd->output_queue = NULL;
3600
		sd->output_queue_tailp = &sd->output_queue;
L
Linus Torvalds 已提交
3601 3602 3603
		local_irq_enable();

		while (head) {
3604 3605 3606
			struct Qdisc *q = head;
			spinlock_t *root_lock;

L
Linus Torvalds 已提交
3607 3608
			head = head->next_sched;

3609
			root_lock = qdisc_lock(q);
3610
			if (spin_trylock(root_lock)) {
3611
				smp_mb__before_atomic();
3612 3613
				clear_bit(__QDISC_STATE_SCHED,
					  &q->state);
3614 3615
				qdisc_run(q);
				spin_unlock(root_lock);
L
Linus Torvalds 已提交
3616
			} else {
3617
				if (!test_bit(__QDISC_STATE_DEACTIVATED,
3618
					      &q->state)) {
3619
					__netif_reschedule(q);
3620
				} else {
3621
					smp_mb__before_atomic();
3622 3623 3624
					clear_bit(__QDISC_STATE_SCHED,
						  &q->state);
				}
L
Linus Torvalds 已提交
3625 3626 3627 3628 3629
			}
		}
	}
}

3630 3631
#if (defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)) && \
    (defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE))
3632 3633 3634
/* This hook is defined here for ATM LANE */
int (*br_fdb_test_addr_hook)(struct net_device *dev,
			     unsigned char *addr) __read_mostly;
3635
EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
3636
#endif
L
Linus Torvalds 已提交
3637

3638 3639 3640 3641
static inline struct sk_buff *handle_ing(struct sk_buff *skb,
					 struct packet_type **pt_prev,
					 int *ret, struct net_device *orig_dev)
{
3642
#ifdef CONFIG_NET_CLS_ACT
3643 3644
	struct tcf_proto *cl = rcu_dereference_bh(skb->dev->ingress_cl_list);
	struct tcf_result cl_res;
3645

3646 3647 3648
	/* If there's at least one ingress present somewhere (so
	 * we get here via enabled static key), remaining devices
	 * that are not configured with an ingress qdisc will bail
3649
	 * out here.
3650
	 */
3651
	if (!cl)
3652
		return skb;
3653 3654 3655
	if (*pt_prev) {
		*ret = deliver_skb(skb, *pt_prev, orig_dev);
		*pt_prev = NULL;
L
Linus Torvalds 已提交
3656 3657
	}

3658
	qdisc_skb_cb(skb)->pkt_len = skb->len;
3659
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS);
3660
	qdisc_bstats_cpu_update(cl->q, skb);
3661

3662
	switch (tc_classify(skb, cl, &cl_res, false)) {
3663 3664 3665 3666 3667
	case TC_ACT_OK:
	case TC_ACT_RECLASSIFY:
		skb->tc_index = TC_H_MIN(cl_res.classid);
		break;
	case TC_ACT_SHOT:
3668
		qdisc_qstats_cpu_drop(cl->q);
3669 3670 3671 3672
	case TC_ACT_STOLEN:
	case TC_ACT_QUEUED:
		kfree_skb(skb);
		return NULL;
3673 3674 3675 3676 3677 3678 3679 3680
	case TC_ACT_REDIRECT:
		/* skb_mac_header check was done by cls/act_bpf, so
		 * we can safely push the L2 header back before
		 * redirecting to another netdev
		 */
		__skb_push(skb, skb->mac_len);
		skb_do_redirect(skb);
		return NULL;
3681 3682
	default:
		break;
3683
	}
3684
#endif /* CONFIG_NET_CLS_ACT */
3685 3686
	return skb;
}
L
Linus Torvalds 已提交
3687

3688 3689 3690 3691
/**
 *	netdev_rx_handler_register - register receive handler
 *	@dev: device to register a handler for
 *	@rx_handler: receive handler to register
J
Jiri Pirko 已提交
3692
 *	@rx_handler_data: data pointer that is used by rx handler
3693
 *
3694
 *	Register a receive handler for a device. This handler will then be
3695 3696 3697 3698
 *	called from __netif_receive_skb. A negative errno code is returned
 *	on a failure.
 *
 *	The caller must hold the rtnl_mutex.
3699 3700
 *
 *	For a general description of rx_handler, see enum rx_handler_result.
3701 3702
 */
int netdev_rx_handler_register(struct net_device *dev,
J
Jiri Pirko 已提交
3703 3704
			       rx_handler_func_t *rx_handler,
			       void *rx_handler_data)
3705 3706 3707 3708 3709 3710
{
	ASSERT_RTNL();

	if (dev->rx_handler)
		return -EBUSY;

3711
	/* Note: rx_handler_data must be set before rx_handler */
J
Jiri Pirko 已提交
3712
	rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
3713 3714 3715 3716 3717 3718 3719 3720 3721 3722
	rcu_assign_pointer(dev->rx_handler, rx_handler);

	return 0;
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_register);

/**
 *	netdev_rx_handler_unregister - unregister receive handler
 *	@dev: device to unregister a handler from
 *
K
Kusanagi Kouichi 已提交
3723
 *	Unregister a receive handler from a device.
3724 3725 3726 3727 3728 3729 3730
 *
 *	The caller must hold the rtnl_mutex.
 */
void netdev_rx_handler_unregister(struct net_device *dev)
{

	ASSERT_RTNL();
3731
	RCU_INIT_POINTER(dev->rx_handler, NULL);
3732 3733 3734 3735 3736
	/* a reader seeing a non NULL rx_handler in a rcu_read_lock()
	 * section has a guarantee to see a non NULL rx_handler_data
	 * as well.
	 */
	synchronize_net();
3737
	RCU_INIT_POINTER(dev->rx_handler_data, NULL);
3738 3739 3740
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);

3741 3742 3743 3744 3745 3746 3747
/*
 * Limit the use of PFMEMALLOC reserves to those protocols that implement
 * the special handling of PFMEMALLOC skbs.
 */
static bool skb_pfmemalloc_protocol(struct sk_buff *skb)
{
	switch (skb->protocol) {
3748 3749 3750 3751 3752
	case htons(ETH_P_ARP):
	case htons(ETH_P_IP):
	case htons(ETH_P_IPV6):
	case htons(ETH_P_8021Q):
	case htons(ETH_P_8021AD):
3753 3754 3755 3756 3757 3758
		return true;
	default:
		return false;
	}
}

3759 3760 3761
static inline int nf_ingress(struct sk_buff *skb, struct packet_type **pt_prev,
			     int *ret, struct net_device *orig_dev)
{
3762
#ifdef CONFIG_NETFILTER_INGRESS
3763 3764 3765 3766 3767 3768 3769 3770
	if (nf_hook_ingress_active(skb)) {
		if (*pt_prev) {
			*ret = deliver_skb(skb, *pt_prev, orig_dev);
			*pt_prev = NULL;
		}

		return nf_hook_ingress(skb);
	}
3771
#endif /* CONFIG_NETFILTER_INGRESS */
3772 3773 3774
	return 0;
}

3775
static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc)
L
Linus Torvalds 已提交
3776 3777
{
	struct packet_type *ptype, *pt_prev;
3778
	rx_handler_func_t *rx_handler;
D
David S. Miller 已提交
3779
	struct net_device *orig_dev;
3780
	bool deliver_exact = false;
L
Linus Torvalds 已提交
3781
	int ret = NET_RX_DROP;
A
Al Viro 已提交
3782
	__be16 type;
L
Linus Torvalds 已提交
3783

3784
	net_timestamp_check(!netdev_tstamp_prequeue, skb);
3785

3786
	trace_netif_receive_skb(skb);
3787

J
Joe Eykholt 已提交
3788
	orig_dev = skb->dev;
3789

3790
	skb_reset_network_header(skb);
3791 3792
	if (!skb_transport_header_was_set(skb))
		skb_reset_transport_header(skb);
3793
	skb_reset_mac_len(skb);
L
Linus Torvalds 已提交
3794 3795 3796

	pt_prev = NULL;

3797
another_round:
3798
	skb->skb_iif = skb->dev->ifindex;
3799 3800 3801

	__this_cpu_inc(softnet_data.processed);

P
Patrick McHardy 已提交
3802 3803
	if (skb->protocol == cpu_to_be16(ETH_P_8021Q) ||
	    skb->protocol == cpu_to_be16(ETH_P_8021AD)) {
3804
		skb = skb_vlan_untag(skb);
3805
		if (unlikely(!skb))
3806
			goto out;
3807 3808
	}

L
Linus Torvalds 已提交
3809 3810 3811 3812 3813 3814 3815
#ifdef CONFIG_NET_CLS_ACT
	if (skb->tc_verd & TC_NCLS) {
		skb->tc_verd = CLR_TC_NCLS(skb->tc_verd);
		goto ncls;
	}
#endif

3816
	if (pfmemalloc)
3817 3818
		goto skip_taps;

L
Linus Torvalds 已提交
3819
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
3820 3821 3822 3823 3824 3825 3826 3827 3828
		if (pt_prev)
			ret = deliver_skb(skb, pt_prev, orig_dev);
		pt_prev = ptype;
	}

	list_for_each_entry_rcu(ptype, &skb->dev->ptype_all, list) {
		if (pt_prev)
			ret = deliver_skb(skb, pt_prev, orig_dev);
		pt_prev = ptype;
L
Linus Torvalds 已提交
3829 3830
	}

3831
skip_taps:
3832
#ifdef CONFIG_NET_INGRESS
3833 3834 3835
	if (static_key_false(&ingress_needed)) {
		skb = handle_ing(skb, &pt_prev, &ret, orig_dev);
		if (!skb)
3836
			goto out;
3837 3838

		if (nf_ingress(skb, &pt_prev, &ret, orig_dev) < 0)
3839
			goto out;
3840
	}
3841 3842
#endif
#ifdef CONFIG_NET_CLS_ACT
3843
	skb->tc_verd = 0;
L
Linus Torvalds 已提交
3844 3845
ncls:
#endif
3846
	if (pfmemalloc && !skb_pfmemalloc_protocol(skb))
3847 3848
		goto drop;

3849
	if (skb_vlan_tag_present(skb)) {
3850 3851 3852 3853
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3854
		if (vlan_do_receive(&skb))
3855 3856
			goto another_round;
		else if (unlikely(!skb))
3857
			goto out;
3858 3859
	}

3860
	rx_handler = rcu_dereference(skb->dev->rx_handler);
3861 3862 3863 3864 3865
	if (rx_handler) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3866 3867
		switch (rx_handler(&skb)) {
		case RX_HANDLER_CONSUMED:
3868
			ret = NET_RX_SUCCESS;
3869
			goto out;
3870
		case RX_HANDLER_ANOTHER:
3871
			goto another_round;
3872 3873 3874 3875 3876 3877 3878
		case RX_HANDLER_EXACT:
			deliver_exact = true;
		case RX_HANDLER_PASS:
			break;
		default:
			BUG();
		}
3879
	}
L
Linus Torvalds 已提交
3880

3881 3882
	if (unlikely(skb_vlan_tag_present(skb))) {
		if (skb_vlan_tag_get_id(skb))
E
Eric Dumazet 已提交
3883 3884 3885 3886 3887 3888 3889
			skb->pkt_type = PACKET_OTHERHOST;
		/* Note: we might in the future use prio bits
		 * and set skb->priority like in vlan_do_receive()
		 * For the time being, just ignore Priority Code Point
		 */
		skb->vlan_tci = 0;
	}
3890

3891 3892
	type = skb->protocol;

3893
	/* deliver only exact match when indicated */
3894 3895 3896 3897 3898
	if (likely(!deliver_exact)) {
		deliver_ptype_list_skb(skb, &pt_prev, orig_dev, type,
				       &ptype_base[ntohs(type) &
						   PTYPE_HASH_MASK]);
	}
3899

3900 3901 3902 3903 3904 3905
	deliver_ptype_list_skb(skb, &pt_prev, orig_dev, type,
			       &orig_dev->ptype_specific);

	if (unlikely(skb->dev != orig_dev)) {
		deliver_ptype_list_skb(skb, &pt_prev, orig_dev, type,
				       &skb->dev->ptype_specific);
L
Linus Torvalds 已提交
3906 3907 3908
	}

	if (pt_prev) {
3909
		if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
3910
			goto drop;
3911 3912
		else
			ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3913
	} else {
3914
drop:
3915
		atomic_long_inc(&skb->dev->rx_dropped);
L
Linus Torvalds 已提交
3916 3917 3918 3919 3920 3921 3922
		kfree_skb(skb);
		/* Jamal, now you will not able to escape explaining
		 * me how you were going to use this. :-)
		 */
		ret = NET_RX_DROP;
	}

3923
out:
3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948
	return ret;
}

static int __netif_receive_skb(struct sk_buff *skb)
{
	int ret;

	if (sk_memalloc_socks() && skb_pfmemalloc(skb)) {
		unsigned long pflags = current->flags;

		/*
		 * PFMEMALLOC skbs are special, they should
		 * - be delivered to SOCK_MEMALLOC sockets only
		 * - stay away from userspace
		 * - have bounded memory usage
		 *
		 * Use PF_MEMALLOC as this saves us from propagating the allocation
		 * context down to all allocation sites.
		 */
		current->flags |= PF_MEMALLOC;
		ret = __netif_receive_skb_core(skb, true);
		tsk_restore_flags(current, pflags, PF_MEMALLOC);
	} else
		ret = __netif_receive_skb_core(skb, false);

L
Linus Torvalds 已提交
3949 3950
	return ret;
}
T
Tom Herbert 已提交
3951

3952
static int netif_receive_skb_internal(struct sk_buff *skb)
T
Tom Herbert 已提交
3953
{
3954 3955
	int ret;

3956
	net_timestamp_check(netdev_tstamp_prequeue, skb);
E
Eric Dumazet 已提交
3957

3958 3959 3960
	if (skb_defer_rx_timestamp(skb))
		return NET_RX_SUCCESS;

3961 3962
	rcu_read_lock();

E
Eric Dumazet 已提交
3963
#ifdef CONFIG_RPS
3964
	if (static_key_false(&rps_needed)) {
E
Eric Dumazet 已提交
3965
		struct rps_dev_flow voidflow, *rflow = &voidflow;
3966
		int cpu = get_rps_cpu(skb->dev, skb, &rflow);
T
Tom Herbert 已提交
3967

E
Eric Dumazet 已提交
3968 3969 3970
		if (cpu >= 0) {
			ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
			rcu_read_unlock();
3971
			return ret;
E
Eric Dumazet 已提交
3972
		}
T
Tom Herbert 已提交
3973
	}
3974
#endif
3975 3976 3977
	ret = __netif_receive_skb(skb);
	rcu_read_unlock();
	return ret;
T
Tom Herbert 已提交
3978
}
3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994

/**
 *	netif_receive_skb - process receive buffer from network
 *	@skb: buffer to process
 *
 *	netif_receive_skb() is the main receive data processing function.
 *	It always succeeds. The buffer may be dropped during processing
 *	for congestion control or by the protocol layers.
 *
 *	This function may only be called from softirq context and interrupts
 *	should be enabled.
 *
 *	Return values (usually ignored):
 *	NET_RX_SUCCESS: no congestion
 *	NET_RX_DROP: packet was dropped
 */
3995
int netif_receive_skb(struct sk_buff *skb)
3996 3997 3998 3999 4000
{
	trace_netif_receive_skb_entry(skb);

	return netif_receive_skb_internal(skb);
}
4001
EXPORT_SYMBOL(netif_receive_skb);
L
Linus Torvalds 已提交
4002

E
Eric Dumazet 已提交
4003 4004 4005
/* Network device is going away, flush any packets still pending
 * Called with irqs disabled.
 */
4006
static void flush_backlog(void *arg)
4007
{
4008
	struct net_device *dev = arg;
4009
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
4010 4011
	struct sk_buff *skb, *tmp;

E
Eric Dumazet 已提交
4012
	rps_lock(sd);
4013
	skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
4014
		if (skb->dev == dev) {
E
Eric Dumazet 已提交
4015
			__skb_unlink(skb, &sd->input_pkt_queue);
4016
			kfree_skb(skb);
4017
			input_queue_head_incr(sd);
4018
		}
4019
	}
E
Eric Dumazet 已提交
4020
	rps_unlock(sd);
4021 4022 4023 4024 4025

	skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
		if (skb->dev == dev) {
			__skb_unlink(skb, &sd->process_queue);
			kfree_skb(skb);
4026
			input_queue_head_incr(sd);
4027 4028
		}
	}
4029 4030
}

4031 4032
static int napi_gro_complete(struct sk_buff *skb)
{
4033
	struct packet_offload *ptype;
4034
	__be16 type = skb->protocol;
4035
	struct list_head *head = &offload_base;
4036 4037
	int err = -ENOENT;

4038 4039
	BUILD_BUG_ON(sizeof(struct napi_gro_cb) > sizeof(skb->cb));

4040 4041
	if (NAPI_GRO_CB(skb)->count == 1) {
		skb_shinfo(skb)->gso_size = 0;
4042
		goto out;
4043
	}
4044 4045 4046

	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
4047
		if (ptype->type != type || !ptype->callbacks.gro_complete)
4048 4049
			continue;

4050
		err = ptype->callbacks.gro_complete(skb, 0);
4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061
		break;
	}
	rcu_read_unlock();

	if (err) {
		WARN_ON(&ptype->list == head);
		kfree_skb(skb);
		return NET_RX_SUCCESS;
	}

out:
4062
	return netif_receive_skb_internal(skb);
4063 4064
}

4065 4066 4067 4068 4069
/* napi->gro_list contains packets ordered by age.
 * youngest packets at the head of it.
 * Complete skbs in reverse order to reduce latencies.
 */
void napi_gro_flush(struct napi_struct *napi, bool flush_old)
4070
{
4071
	struct sk_buff *skb, *prev = NULL;
4072

4073 4074 4075 4076 4077 4078 4079
	/* scan list and build reverse chain */
	for (skb = napi->gro_list; skb != NULL; skb = skb->next) {
		skb->prev = prev;
		prev = skb;
	}

	for (skb = prev; skb; skb = prev) {
4080
		skb->next = NULL;
4081 4082 4083 4084 4085

		if (flush_old && NAPI_GRO_CB(skb)->age == jiffies)
			return;

		prev = skb->prev;
4086
		napi_gro_complete(skb);
4087
		napi->gro_count--;
4088 4089 4090 4091
	}

	napi->gro_list = NULL;
}
E
Eric Dumazet 已提交
4092
EXPORT_SYMBOL(napi_gro_flush);
4093

E
Eric Dumazet 已提交
4094 4095 4096 4097
static void gro_list_prepare(struct napi_struct *napi, struct sk_buff *skb)
{
	struct sk_buff *p;
	unsigned int maclen = skb->dev->hard_header_len;
4098
	u32 hash = skb_get_hash_raw(skb);
E
Eric Dumazet 已提交
4099 4100 4101 4102

	for (p = napi->gro_list; p; p = p->next) {
		unsigned long diffs;

4103 4104 4105 4106 4107 4108 4109
		NAPI_GRO_CB(p)->flush = 0;

		if (hash != skb_get_hash_raw(p)) {
			NAPI_GRO_CB(p)->same_flow = 0;
			continue;
		}

E
Eric Dumazet 已提交
4110 4111 4112 4113
		diffs = (unsigned long)p->dev ^ (unsigned long)skb->dev;
		diffs |= p->vlan_tci ^ skb->vlan_tci;
		if (maclen == ETH_HLEN)
			diffs |= compare_ether_header(skb_mac_header(p),
4114
						      skb_mac_header(skb));
E
Eric Dumazet 已提交
4115 4116
		else if (!diffs)
			diffs = memcmp(skb_mac_header(p),
4117
				       skb_mac_header(skb),
E
Eric Dumazet 已提交
4118 4119 4120 4121 4122
				       maclen);
		NAPI_GRO_CB(p)->same_flow = !diffs;
	}
}

4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136
static void skb_gro_reset_offset(struct sk_buff *skb)
{
	const struct skb_shared_info *pinfo = skb_shinfo(skb);
	const skb_frag_t *frag0 = &pinfo->frags[0];

	NAPI_GRO_CB(skb)->data_offset = 0;
	NAPI_GRO_CB(skb)->frag0 = NULL;
	NAPI_GRO_CB(skb)->frag0_len = 0;

	if (skb_mac_header(skb) == skb_tail_pointer(skb) &&
	    pinfo->nr_frags &&
	    !PageHighMem(skb_frag_page(frag0))) {
		NAPI_GRO_CB(skb)->frag0 = skb_frag_address(frag0);
		NAPI_GRO_CB(skb)->frag0_len = skb_frag_size(frag0);
E
Eric Dumazet 已提交
4137 4138 4139
	}
}

4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160
static void gro_pull_from_frag0(struct sk_buff *skb, int grow)
{
	struct skb_shared_info *pinfo = skb_shinfo(skb);

	BUG_ON(skb->end - skb->tail < grow);

	memcpy(skb_tail_pointer(skb), NAPI_GRO_CB(skb)->frag0, grow);

	skb->data_len -= grow;
	skb->tail += grow;

	pinfo->frags[0].page_offset += grow;
	skb_frag_size_sub(&pinfo->frags[0], grow);

	if (unlikely(!skb_frag_size(&pinfo->frags[0]))) {
		skb_frag_unref(skb, 0);
		memmove(pinfo->frags, pinfo->frags + 1,
			--pinfo->nr_frags * sizeof(pinfo->frags[0]));
	}
}

R
Rami Rosen 已提交
4161
static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
4162 4163
{
	struct sk_buff **pp = NULL;
4164
	struct packet_offload *ptype;
4165
	__be16 type = skb->protocol;
4166
	struct list_head *head = &offload_base;
H
Herbert Xu 已提交
4167
	int same_flow;
4168
	enum gro_result ret;
4169
	int grow;
4170

4171
	if (!(skb->dev->features & NETIF_F_GRO))
4172 4173
		goto normal;

4174
	if (skb_is_gso(skb) || skb_has_frag_list(skb) || skb->csum_bad)
4175 4176
		goto normal;

E
Eric Dumazet 已提交
4177 4178
	gro_list_prepare(napi, skb);

4179 4180
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
4181
		if (ptype->type != type || !ptype->callbacks.gro_receive)
4182 4183
			continue;

4184
		skb_set_network_header(skb, skb_gro_offset(skb));
4185
		skb_reset_mac_len(skb);
4186 4187
		NAPI_GRO_CB(skb)->same_flow = 0;
		NAPI_GRO_CB(skb)->flush = 0;
H
Herbert Xu 已提交
4188
		NAPI_GRO_CB(skb)->free = 0;
4189
		NAPI_GRO_CB(skb)->udp_mark = 0;
4190
		NAPI_GRO_CB(skb)->gro_remcsum_start = 0;
4191

4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206
		/* Setup for GRO checksum validation */
		switch (skb->ip_summed) {
		case CHECKSUM_COMPLETE:
			NAPI_GRO_CB(skb)->csum = skb->csum;
			NAPI_GRO_CB(skb)->csum_valid = 1;
			NAPI_GRO_CB(skb)->csum_cnt = 0;
			break;
		case CHECKSUM_UNNECESSARY:
			NAPI_GRO_CB(skb)->csum_cnt = skb->csum_level + 1;
			NAPI_GRO_CB(skb)->csum_valid = 0;
			break;
		default:
			NAPI_GRO_CB(skb)->csum_cnt = 0;
			NAPI_GRO_CB(skb)->csum_valid = 0;
		}
4207

4208
		pp = ptype->callbacks.gro_receive(&napi->gro_list, skb);
4209 4210 4211 4212 4213 4214 4215
		break;
	}
	rcu_read_unlock();

	if (&ptype->list == head)
		goto normal;

H
Herbert Xu 已提交
4216
	same_flow = NAPI_GRO_CB(skb)->same_flow;
4217
	ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED;
H
Herbert Xu 已提交
4218

4219 4220 4221 4222 4223 4224
	if (pp) {
		struct sk_buff *nskb = *pp;

		*pp = nskb->next;
		nskb->next = NULL;
		napi_gro_complete(nskb);
4225
		napi->gro_count--;
4226 4227
	}

H
Herbert Xu 已提交
4228
	if (same_flow)
4229 4230
		goto ok;

4231
	if (NAPI_GRO_CB(skb)->flush)
4232 4233
		goto normal;

4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247
	if (unlikely(napi->gro_count >= MAX_GRO_SKBS)) {
		struct sk_buff *nskb = napi->gro_list;

		/* locate the end of the list to select the 'oldest' flow */
		while (nskb->next) {
			pp = &nskb->next;
			nskb = *pp;
		}
		*pp = NULL;
		nskb->next = NULL;
		napi_gro_complete(nskb);
	} else {
		napi->gro_count++;
	}
4248
	NAPI_GRO_CB(skb)->count = 1;
4249
	NAPI_GRO_CB(skb)->age = jiffies;
4250
	NAPI_GRO_CB(skb)->last = skb;
4251
	skb_shinfo(skb)->gso_size = skb_gro_len(skb);
4252 4253
	skb->next = napi->gro_list;
	napi->gro_list = skb;
4254
	ret = GRO_HELD;
4255

4256
pull:
4257 4258 4259
	grow = skb_gro_offset(skb) - skb_headlen(skb);
	if (grow > 0)
		gro_pull_from_frag0(skb, grow);
4260
ok:
4261
	return ret;
4262 4263

normal:
4264 4265
	ret = GRO_NORMAL;
	goto pull;
H
Herbert Xu 已提交
4266
}
4267

4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279
struct packet_offload *gro_find_receive_by_type(__be16 type)
{
	struct list_head *offload_head = &offload_base;
	struct packet_offload *ptype;

	list_for_each_entry_rcu(ptype, offload_head, list) {
		if (ptype->type != type || !ptype->callbacks.gro_receive)
			continue;
		return ptype;
	}
	return NULL;
}
4280
EXPORT_SYMBOL(gro_find_receive_by_type);
4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293

struct packet_offload *gro_find_complete_by_type(__be16 type)
{
	struct list_head *offload_head = &offload_base;
	struct packet_offload *ptype;

	list_for_each_entry_rcu(ptype, offload_head, list) {
		if (ptype->type != type || !ptype->callbacks.gro_complete)
			continue;
		return ptype;
	}
	return NULL;
}
4294
EXPORT_SYMBOL(gro_find_complete_by_type);
H
Herbert Xu 已提交
4295

R
Rami Rosen 已提交
4296
static gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
H
Herbert Xu 已提交
4297
{
4298 4299
	switch (ret) {
	case GRO_NORMAL:
4300
		if (netif_receive_skb_internal(skb))
4301 4302
			ret = GRO_DROP;
		break;
H
Herbert Xu 已提交
4303

4304
	case GRO_DROP:
H
Herbert Xu 已提交
4305 4306
		kfree_skb(skb);
		break;
4307

4308
	case GRO_MERGED_FREE:
4309 4310 4311 4312
		if (NAPI_GRO_CB(skb)->free == NAPI_GRO_FREE_STOLEN_HEAD)
			kmem_cache_free(skbuff_head_cache, skb);
		else
			__kfree_skb(skb);
4313 4314
		break;

4315 4316 4317
	case GRO_HELD:
	case GRO_MERGED:
		break;
H
Herbert Xu 已提交
4318 4319
	}

4320
	return ret;
4321 4322
}

4323
gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
4324
{
4325
	trace_napi_gro_receive_entry(skb);
4326

4327 4328
	skb_gro_reset_offset(skb);

E
Eric Dumazet 已提交
4329
	return napi_skb_finish(dev_gro_receive(napi, skb), skb);
4330 4331 4332
}
EXPORT_SYMBOL(napi_gro_receive);

S
stephen hemminger 已提交
4333
static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb)
4334
{
4335 4336 4337 4338
	if (unlikely(skb->pfmemalloc)) {
		consume_skb(skb);
		return;
	}
4339
	__skb_pull(skb, skb_headlen(skb));
4340 4341
	/* restore the reserve we had after netdev_alloc_skb_ip_align() */
	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN - skb_headroom(skb));
4342
	skb->vlan_tci = 0;
H
Herbert Xu 已提交
4343
	skb->dev = napi->dev;
A
Andy Gospodarek 已提交
4344
	skb->skb_iif = 0;
4345 4346
	skb->encapsulation = 0;
	skb_shinfo(skb)->gso_type = 0;
4347
	skb->truesize = SKB_TRUESIZE(skb_end_offset(skb));
4348 4349 4350 4351

	napi->skb = skb;
}

4352
struct sk_buff *napi_get_frags(struct napi_struct *napi)
H
Herbert Xu 已提交
4353 4354 4355 4356
{
	struct sk_buff *skb = napi->skb;

	if (!skb) {
4357
		skb = napi_alloc_skb(napi, GRO_MAX_HEAD);
E
Eric Dumazet 已提交
4358
		napi->skb = skb;
4359
	}
4360 4361
	return skb;
}
4362
EXPORT_SYMBOL(napi_get_frags);
4363

4364 4365 4366
static gro_result_t napi_frags_finish(struct napi_struct *napi,
				      struct sk_buff *skb,
				      gro_result_t ret)
4367
{
4368 4369
	switch (ret) {
	case GRO_NORMAL:
4370 4371 4372 4373
	case GRO_HELD:
		__skb_push(skb, ETH_HLEN);
		skb->protocol = eth_type_trans(skb, skb->dev);
		if (ret == GRO_NORMAL && netif_receive_skb_internal(skb))
4374
			ret = GRO_DROP;
4375
		break;
H
Herbert Xu 已提交
4376

4377 4378 4379 4380
	case GRO_DROP:
	case GRO_MERGED_FREE:
		napi_reuse_skb(napi, skb);
		break;
4381 4382 4383

	case GRO_MERGED:
		break;
4384
	}
H
Herbert Xu 已提交
4385

4386
	return ret;
H
Herbert Xu 已提交
4387
}
4388

4389 4390 4391 4392
/* Upper GRO stack assumes network header starts at gro_offset=0
 * Drivers could call both napi_gro_frags() and napi_gro_receive()
 * We copy ethernet header into skb->data to have a common layout.
 */
E
Eric Dumazet 已提交
4393
static struct sk_buff *napi_frags_skb(struct napi_struct *napi)
4394 4395
{
	struct sk_buff *skb = napi->skb;
4396 4397
	const struct ethhdr *eth;
	unsigned int hlen = sizeof(*eth);
4398 4399 4400

	napi->skb = NULL;

4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414
	skb_reset_mac_header(skb);
	skb_gro_reset_offset(skb);

	eth = skb_gro_header_fast(skb, 0);
	if (unlikely(skb_gro_header_hard(skb, hlen))) {
		eth = skb_gro_header_slow(skb, hlen, 0);
		if (unlikely(!eth)) {
			napi_reuse_skb(napi, skb);
			return NULL;
		}
	} else {
		gro_pull_from_frag0(skb, hlen);
		NAPI_GRO_CB(skb)->frag0 += hlen;
		NAPI_GRO_CB(skb)->frag0_len -= hlen;
4415
	}
4416 4417 4418 4419 4420 4421 4422 4423
	__skb_pull(skb, hlen);

	/*
	 * This works because the only protocols we care about don't require
	 * special handling.
	 * We'll fix it up properly in napi_frags_finish()
	 */
	skb->protocol = eth->h_proto;
4424 4425 4426 4427

	return skb;
}

4428
gro_result_t napi_gro_frags(struct napi_struct *napi)
4429
{
4430
	struct sk_buff *skb = napi_frags_skb(napi);
4431 4432

	if (!skb)
4433
		return GRO_DROP;
4434

4435 4436
	trace_napi_gro_frags_entry(skb);

E
Eric Dumazet 已提交
4437
	return napi_frags_finish(napi, skb, dev_gro_receive(napi, skb));
4438
}
H
Herbert Xu 已提交
4439 4440
EXPORT_SYMBOL(napi_gro_frags);

T
Tom Herbert 已提交
4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465
/* Compute the checksum from gro_offset and return the folded value
 * after adding in any pseudo checksum.
 */
__sum16 __skb_gro_checksum_complete(struct sk_buff *skb)
{
	__wsum wsum;
	__sum16 sum;

	wsum = skb_checksum(skb, skb_gro_offset(skb), skb_gro_len(skb), 0);

	/* NAPI_GRO_CB(skb)->csum holds pseudo checksum */
	sum = csum_fold(csum_add(NAPI_GRO_CB(skb)->csum, wsum));
	if (likely(!sum)) {
		if (unlikely(skb->ip_summed == CHECKSUM_COMPLETE) &&
		    !skb->csum_complete_sw)
			netdev_rx_csum_fault(skb->dev);
	}

	NAPI_GRO_CB(skb)->csum = wsum;
	NAPI_GRO_CB(skb)->csum_valid = 1;

	return sum;
}
EXPORT_SYMBOL(__skb_gro_checksum_complete);

4466
/*
4467
 * net_rps_action_and_irq_enable sends any pending IPI's for rps.
4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484
 * Note: called with local irq disabled, but exits with local irq enabled.
 */
static void net_rps_action_and_irq_enable(struct softnet_data *sd)
{
#ifdef CONFIG_RPS
	struct softnet_data *remsd = sd->rps_ipi_list;

	if (remsd) {
		sd->rps_ipi_list = NULL;

		local_irq_enable();

		/* Send pending IPI's to kick RPS processing on remote cpus. */
		while (remsd) {
			struct softnet_data *next = remsd->rps_ipi_next;

			if (cpu_online(remsd->cpu))
4485
				smp_call_function_single_async(remsd->cpu,
4486
							   &remsd->csd);
4487 4488 4489 4490 4491 4492 4493
			remsd = next;
		}
	} else
#endif
		local_irq_enable();
}

E
Eric Dumazet 已提交
4494 4495 4496 4497 4498 4499 4500 4501 4502
static bool sd_has_rps_ipi_waiting(struct softnet_data *sd)
{
#ifdef CONFIG_RPS
	return sd->rps_ipi_list != NULL;
#else
	return false;
#endif
}

4503
static int process_backlog(struct napi_struct *napi, int quota)
L
Linus Torvalds 已提交
4504 4505
{
	int work = 0;
E
Eric Dumazet 已提交
4506
	struct softnet_data *sd = container_of(napi, struct softnet_data, backlog);
L
Linus Torvalds 已提交
4507

4508 4509 4510
	/* Check if we have pending ipi, its better to send them now,
	 * not waiting net_rx_action() end.
	 */
E
Eric Dumazet 已提交
4511
	if (sd_has_rps_ipi_waiting(sd)) {
4512 4513 4514
		local_irq_disable();
		net_rps_action_and_irq_enable(sd);
	}
E
Eric Dumazet 已提交
4515

4516
	napi->weight = weight_p;
4517
	local_irq_disable();
4518
	while (1) {
L
Linus Torvalds 已提交
4519
		struct sk_buff *skb;
4520 4521

		while ((skb = __skb_dequeue(&sd->process_queue))) {
4522
			rcu_read_lock();
4523 4524
			local_irq_enable();
			__netif_receive_skb(skb);
4525
			rcu_read_unlock();
4526
			local_irq_disable();
4527 4528 4529 4530 4531
			input_queue_head_incr(sd);
			if (++work >= quota) {
				local_irq_enable();
				return work;
			}
4532
		}
L
Linus Torvalds 已提交
4533

E
Eric Dumazet 已提交
4534
		rps_lock(sd);
4535
		if (skb_queue_empty(&sd->input_pkt_queue)) {
E
Eric Dumazet 已提交
4536 4537 4538
			/*
			 * Inline a custom version of __napi_complete().
			 * only current cpu owns and manipulates this napi,
4539 4540 4541
			 * and NAPI_STATE_SCHED is the only possible flag set
			 * on backlog.
			 * We can use a plain write instead of clear_bit(),
E
Eric Dumazet 已提交
4542 4543 4544
			 * and we dont need an smp_mb() memory barrier.
			 */
			napi->state = 0;
4545
			rps_unlock(sd);
E
Eric Dumazet 已提交
4546

4547
			break;
4548
		}
4549 4550 4551

		skb_queue_splice_tail_init(&sd->input_pkt_queue,
					   &sd->process_queue);
E
Eric Dumazet 已提交
4552
		rps_unlock(sd);
4553 4554
	}
	local_irq_enable();
L
Linus Torvalds 已提交
4555

4556 4557
	return work;
}
L
Linus Torvalds 已提交
4558

4559 4560
/**
 * __napi_schedule - schedule for receive
4561
 * @n: entry to schedule
4562
 *
4563 4564
 * The entry's receive function will be scheduled to run.
 * Consider using __napi_schedule_irqoff() if hard irqs are masked.
4565
 */
H
Harvey Harrison 已提交
4566
void __napi_schedule(struct napi_struct *n)
4567 4568
{
	unsigned long flags;
L
Linus Torvalds 已提交
4569

4570
	local_irq_save(flags);
4571
	____napi_schedule(this_cpu_ptr(&softnet_data), n);
4572
	local_irq_restore(flags);
L
Linus Torvalds 已提交
4573
}
4574 4575
EXPORT_SYMBOL(__napi_schedule);

4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587
/**
 * __napi_schedule_irqoff - schedule for receive
 * @n: entry to schedule
 *
 * Variant of __napi_schedule() assuming hard irqs are masked
 */
void __napi_schedule_irqoff(struct napi_struct *n)
{
	____napi_schedule(this_cpu_ptr(&softnet_data), n);
}
EXPORT_SYMBOL(__napi_schedule_irqoff);

4588 4589 4590 4591
void __napi_complete(struct napi_struct *n)
{
	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));

E
Eric Dumazet 已提交
4592
	list_del_init(&n->poll_list);
4593
	smp_mb__before_atomic();
4594 4595 4596 4597
	clear_bit(NAPI_STATE_SCHED, &n->state);
}
EXPORT_SYMBOL(__napi_complete);

4598
void napi_complete_done(struct napi_struct *n, int work_done)
4599 4600 4601 4602 4603 4604 4605 4606 4607 4608
{
	unsigned long flags;

	/*
	 * don't let napi dequeue from the cpu poll list
	 * just in case its running on a different cpu
	 */
	if (unlikely(test_bit(NAPI_STATE_NPSVC, &n->state)))
		return;

4609 4610
	if (n->gro_list) {
		unsigned long timeout = 0;
E
Eric Dumazet 已提交
4611

4612 4613 4614 4615 4616 4617 4618 4619 4620
		if (work_done)
			timeout = n->dev->gro_flush_timeout;

		if (timeout)
			hrtimer_start(&n->timer, ns_to_ktime(timeout),
				      HRTIMER_MODE_REL_PINNED);
		else
			napi_gro_flush(n, false);
	}
E
Eric Dumazet 已提交
4621 4622 4623 4624 4625 4626 4627 4628
	if (likely(list_empty(&n->poll_list))) {
		WARN_ON_ONCE(!test_and_clear_bit(NAPI_STATE_SCHED, &n->state));
	} else {
		/* If n->poll_list is not empty, we need to mask irqs */
		local_irq_save(flags);
		__napi_complete(n);
		local_irq_restore(flags);
	}
4629
}
4630
EXPORT_SYMBOL(napi_complete_done);
4631

E
Eliezer Tamir 已提交
4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683
/* must be called under rcu_read_lock(), as we dont take a reference */
struct napi_struct *napi_by_id(unsigned int napi_id)
{
	unsigned int hash = napi_id % HASH_SIZE(napi_hash);
	struct napi_struct *napi;

	hlist_for_each_entry_rcu(napi, &napi_hash[hash], napi_hash_node)
		if (napi->napi_id == napi_id)
			return napi;

	return NULL;
}
EXPORT_SYMBOL_GPL(napi_by_id);

void napi_hash_add(struct napi_struct *napi)
{
	if (!test_and_set_bit(NAPI_STATE_HASHED, &napi->state)) {

		spin_lock(&napi_hash_lock);

		/* 0 is not a valid id, we also skip an id that is taken
		 * we expect both events to be extremely rare
		 */
		napi->napi_id = 0;
		while (!napi->napi_id) {
			napi->napi_id = ++napi_gen_id;
			if (napi_by_id(napi->napi_id))
				napi->napi_id = 0;
		}

		hlist_add_head_rcu(&napi->napi_hash_node,
			&napi_hash[napi->napi_id % HASH_SIZE(napi_hash)]);

		spin_unlock(&napi_hash_lock);
	}
}
EXPORT_SYMBOL_GPL(napi_hash_add);

/* Warning : caller is responsible to make sure rcu grace period
 * is respected before freeing memory containing @napi
 */
void napi_hash_del(struct napi_struct *napi)
{
	spin_lock(&napi_hash_lock);

	if (test_and_clear_bit(NAPI_STATE_HASHED, &napi->state))
		hlist_del_rcu(&napi->napi_hash_node);

	spin_unlock(&napi_hash_lock);
}
EXPORT_SYMBOL_GPL(napi_hash_del);

4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694
static enum hrtimer_restart napi_watchdog(struct hrtimer *timer)
{
	struct napi_struct *napi;

	napi = container_of(timer, struct napi_struct, timer);
	if (napi->gro_list)
		napi_schedule(napi);

	return HRTIMER_NORESTART;
}

4695 4696 4697 4698
void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
		    int (*poll)(struct napi_struct *, int), int weight)
{
	INIT_LIST_HEAD(&napi->poll_list);
4699 4700
	hrtimer_init(&napi->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_PINNED);
	napi->timer.function = napi_watchdog;
4701
	napi->gro_count = 0;
4702
	napi->gro_list = NULL;
H
Herbert Xu 已提交
4703
	napi->skb = NULL;
4704
	napi->poll = poll;
E
Eric Dumazet 已提交
4705 4706 4707
	if (weight > NAPI_POLL_WEIGHT)
		pr_err_once("netif_napi_add() called with weight %d on device %s\n",
			    weight, dev->name);
4708 4709 4710
	napi->weight = weight;
	list_add(&napi->dev_list, &dev->napi_list);
	napi->dev = dev;
H
Herbert Xu 已提交
4711
#ifdef CONFIG_NETPOLL
4712 4713 4714 4715 4716 4717 4718
	spin_lock_init(&napi->poll_lock);
	napi->poll_owner = -1;
#endif
	set_bit(NAPI_STATE_SCHED, &napi->state);
}
EXPORT_SYMBOL(netif_napi_add);

4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732
void napi_disable(struct napi_struct *n)
{
	might_sleep();
	set_bit(NAPI_STATE_DISABLE, &n->state);

	while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
		msleep(1);

	hrtimer_cancel(&n->timer);

	clear_bit(NAPI_STATE_DISABLE, &n->state);
}
EXPORT_SYMBOL(napi_disable);

4733 4734
void netif_napi_del(struct napi_struct *napi)
{
4735
	list_del_init(&napi->dev_list);
4736
	napi_free_frags(napi);
4737

E
Eric Dumazet 已提交
4738
	kfree_skb_list(napi->gro_list);
4739
	napi->gro_list = NULL;
4740
	napi->gro_count = 0;
4741 4742 4743
}
EXPORT_SYMBOL(netif_napi_del);

4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787 4788
static int napi_poll(struct napi_struct *n, struct list_head *repoll)
{
	void *have;
	int work, weight;

	list_del_init(&n->poll_list);

	have = netpoll_poll_lock(n);

	weight = n->weight;

	/* This NAPI_STATE_SCHED test is for avoiding a race
	 * with netpoll's poll_napi().  Only the entity which
	 * obtains the lock and sees NAPI_STATE_SCHED set will
	 * actually make the ->poll() call.  Therefore we avoid
	 * accidentally calling ->poll() when NAPI is not scheduled.
	 */
	work = 0;
	if (test_bit(NAPI_STATE_SCHED, &n->state)) {
		work = n->poll(n, weight);
		trace_napi_poll(n);
	}

	WARN_ON_ONCE(work > weight);

	if (likely(work < weight))
		goto out_unlock;

	/* Drivers must not modify the NAPI state if they
	 * consume the entire weight.  In such cases this code
	 * still "owns" the NAPI instance and therefore can
	 * move the instance around on the list at-will.
	 */
	if (unlikely(napi_disable_pending(n))) {
		napi_complete(n);
		goto out_unlock;
	}

	if (n->gro_list) {
		/* flush too old packets
		 * If HZ < 1000, flush all packets.
		 */
		napi_gro_flush(n, HZ >= 1000);
	}

4789 4790 4791 4792 4793 4794 4795 4796 4797
	/* Some drivers may have called napi_schedule
	 * prior to exhausting their budget.
	 */
	if (unlikely(!list_empty(&n->poll_list))) {
		pr_warn_once("%s: Budget exhausted after napi rescheduled\n",
			     n->dev ? n->dev->name : "backlog");
		goto out_unlock;
	}

4798 4799 4800 4801 4802 4803 4804 4805
	list_add_tail(&n->poll_list, repoll);

out_unlock:
	netpoll_poll_unlock(have);

	return work;
}

L
Linus Torvalds 已提交
4806 4807
static void net_rx_action(struct softirq_action *h)
{
4808
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
4809
	unsigned long time_limit = jiffies + 2;
4810
	int budget = netdev_budget;
E
Eric Dumazet 已提交
4811 4812
	LIST_HEAD(list);
	LIST_HEAD(repoll);
4813

L
Linus Torvalds 已提交
4814
	local_irq_disable();
E
Eric Dumazet 已提交
4815 4816
	list_splice_init(&sd->poll_list, &list);
	local_irq_enable();
L
Linus Torvalds 已提交
4817

H
Herbert Xu 已提交
4818
	for (;;) {
4819
		struct napi_struct *n;
L
Linus Torvalds 已提交
4820

H
Herbert Xu 已提交
4821 4822 4823 4824 4825 4826
		if (list_empty(&list)) {
			if (!sd_has_rps_ipi_waiting(sd) && list_empty(&repoll))
				return;
			break;
		}

4827 4828 4829
		n = list_first_entry(&list, struct napi_struct, poll_list);
		budget -= napi_poll(n, &repoll);

E
Eric Dumazet 已提交
4830
		/* If softirq window is exhausted then punt.
4831 4832
		 * Allow this to run for 2 jiffies since which will allow
		 * an average latency of 1.5/HZ.
4833
		 */
H
Herbert Xu 已提交
4834 4835 4836 4837 4838
		if (unlikely(budget <= 0 ||
			     time_after_eq(jiffies, time_limit))) {
			sd->time_squeeze++;
			break;
		}
L
Linus Torvalds 已提交
4839
	}
E
Eric Dumazet 已提交
4840 4841 4842 4843 4844 4845 4846 4847 4848

	local_irq_disable();

	list_splice_tail_init(&sd->poll_list, &list);
	list_splice_tail(&repoll, &list);
	list_splice(&list, &sd->poll_list);
	if (!list_empty(&sd->poll_list))
		__raise_softirq_irqoff(NET_RX_SOFTIRQ);

4849
	net_rps_action_and_irq_enable(sd);
L
Linus Torvalds 已提交
4850 4851
}

4852
struct netdev_adjacent {
J
Jiri Pirko 已提交
4853
	struct net_device *dev;
4854 4855

	/* upper master flag, there can only be one master device per list */
J
Jiri Pirko 已提交
4856
	bool master;
4857 4858 4859 4860

	/* counter for the number of times this device was added to us */
	u16 ref_nr;

4861 4862 4863
	/* private field for the users */
	void *private;

J
Jiri Pirko 已提交
4864 4865 4866 4867
	struct list_head list;
	struct rcu_head rcu;
};

4868 4869
static struct netdev_adjacent *__netdev_find_adj(struct net_device *dev,
						 struct net_device *adj_dev,
4870
						 struct list_head *adj_list)
J
Jiri Pirko 已提交
4871
{
4872 4873
	struct netdev_adjacent *adj;

4874
	list_for_each_entry(adj, adj_list, list) {
4875 4876
		if (adj->dev == adj_dev)
			return adj;
J
Jiri Pirko 已提交
4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894
	}
	return NULL;
}

/**
 * netdev_has_upper_dev - Check if device is linked to an upper device
 * @dev: device
 * @upper_dev: upper device to check
 *
 * Find out if a device is linked to specified upper device and return true
 * in case it is. Note that this checks only immediate upper device,
 * not through a complete stack of devices. The caller must hold the RTNL lock.
 */
bool netdev_has_upper_dev(struct net_device *dev,
			  struct net_device *upper_dev)
{
	ASSERT_RTNL();

4895
	return __netdev_find_adj(dev, upper_dev, &dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4896 4897 4898 4899 4900 4901 4902 4903 4904 4905
}
EXPORT_SYMBOL(netdev_has_upper_dev);

/**
 * netdev_has_any_upper_dev - Check if device is linked to some device
 * @dev: device
 *
 * Find out if a device is linked to an upper device and return true in case
 * it is. The caller must hold the RTNL lock.
 */
S
stephen hemminger 已提交
4906
static bool netdev_has_any_upper_dev(struct net_device *dev)
J
Jiri Pirko 已提交
4907 4908 4909
{
	ASSERT_RTNL();

4910
	return !list_empty(&dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921
}

/**
 * netdev_master_upper_dev_get - Get master upper device
 * @dev: device
 *
 * Find a master upper device and return pointer to it or NULL in case
 * it's not there. The caller must hold the RTNL lock.
 */
struct net_device *netdev_master_upper_dev_get(struct net_device *dev)
{
4922
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4923 4924 4925

	ASSERT_RTNL();

4926
	if (list_empty(&dev->adj_list.upper))
J
Jiri Pirko 已提交
4927 4928
		return NULL;

4929
	upper = list_first_entry(&dev->adj_list.upper,
4930
				 struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4931 4932 4933 4934 4935 4936
	if (likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get);

4937 4938 4939 4940 4941 4942 4943 4944 4945 4946
void *netdev_adjacent_get_private(struct list_head *adj_list)
{
	struct netdev_adjacent *adj;

	adj = list_entry(adj_list, struct netdev_adjacent, list);

	return adj->private;
}
EXPORT_SYMBOL(netdev_adjacent_get_private);

4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972
/**
 * netdev_upper_get_next_dev_rcu - Get the next dev from upper list
 * @dev: device
 * @iter: list_head ** of the current position
 *
 * Gets the next device from the dev's upper list, starting from iter
 * position. The caller must hold RCU read lock.
 */
struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
						 struct list_head **iter)
{
	struct netdev_adjacent *upper;

	WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_rtnl_is_held());

	upper = list_entry_rcu((*iter)->next, struct netdev_adjacent, list);

	if (&upper->list == &dev->adj_list.upper)
		return NULL;

	*iter = &upper->list;

	return upper->dev;
}
EXPORT_SYMBOL(netdev_upper_get_next_dev_rcu);

4973 4974
/**
 * netdev_all_upper_get_next_dev_rcu - Get the next dev from upper list
4975 4976 4977 4978 4979 4980
 * @dev: device
 * @iter: list_head ** of the current position
 *
 * Gets the next device from the dev's upper list, starting from iter
 * position. The caller must hold RCU read lock.
 */
4981 4982
struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
						     struct list_head **iter)
4983 4984 4985
{
	struct netdev_adjacent *upper;

4986
	WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_rtnl_is_held());
4987 4988 4989

	upper = list_entry_rcu((*iter)->next, struct netdev_adjacent, list);

4990
	if (&upper->list == &dev->all_adj_list.upper)
4991 4992 4993 4994 4995 4996
		return NULL;

	*iter = &upper->list;

	return upper->dev;
}
4997
EXPORT_SYMBOL(netdev_all_upper_get_next_dev_rcu);
4998

4999 5000 5001 5002 5003 5004 5005 5006 5007
/**
 * netdev_lower_get_next_private - Get the next ->private from the
 *				   lower neighbour list
 * @dev: device
 * @iter: list_head ** of the current position
 *
 * Gets the next netdev_adjacent->private from the dev's lower neighbour
 * list, starting from iter position. The caller must hold either hold the
 * RTNL lock or its own locking that guarantees that the neighbour lower
5008
 * list will remain unchanged.
5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019
 */
void *netdev_lower_get_next_private(struct net_device *dev,
				    struct list_head **iter)
{
	struct netdev_adjacent *lower;

	lower = list_entry(*iter, struct netdev_adjacent, list);

	if (&lower->list == &dev->adj_list.lower)
		return NULL;

5020
	*iter = lower->list.next;
5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047

	return lower->private;
}
EXPORT_SYMBOL(netdev_lower_get_next_private);

/**
 * netdev_lower_get_next_private_rcu - Get the next ->private from the
 *				       lower neighbour list, RCU
 *				       variant
 * @dev: device
 * @iter: list_head ** of the current position
 *
 * Gets the next netdev_adjacent->private from the dev's lower neighbour
 * list, starting from iter position. The caller must hold RCU read lock.
 */
void *netdev_lower_get_next_private_rcu(struct net_device *dev,
					struct list_head **iter)
{
	struct netdev_adjacent *lower;

	WARN_ON_ONCE(!rcu_read_lock_held());

	lower = list_entry_rcu((*iter)->next, struct netdev_adjacent, list);

	if (&lower->list == &dev->adj_list.lower)
		return NULL;

5048
	*iter = &lower->list;
5049 5050 5051 5052 5053

	return lower->private;
}
EXPORT_SYMBOL(netdev_lower_get_next_private_rcu);

5054 5055 5056 5057 5058 5059 5060 5061 5062
/**
 * netdev_lower_get_next - Get the next device from the lower neighbour
 *                         list
 * @dev: device
 * @iter: list_head ** of the current position
 *
 * Gets the next netdev_adjacent from the dev's lower neighbour
 * list, starting from iter position. The caller must hold RTNL lock or
 * its own locking that guarantees that the neighbour lower
5063
 * list will remain unchanged.
5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079
 */
void *netdev_lower_get_next(struct net_device *dev, struct list_head **iter)
{
	struct netdev_adjacent *lower;

	lower = list_entry((*iter)->next, struct netdev_adjacent, list);

	if (&lower->list == &dev->adj_list.lower)
		return NULL;

	*iter = &lower->list;

	return lower->dev;
}
EXPORT_SYMBOL(netdev_lower_get_next);

5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100
/**
 * netdev_lower_get_first_private_rcu - Get the first ->private from the
 *				       lower neighbour list, RCU
 *				       variant
 * @dev: device
 *
 * Gets the first netdev_adjacent->private from the dev's lower neighbour
 * list. The caller must hold RCU read lock.
 */
void *netdev_lower_get_first_private_rcu(struct net_device *dev)
{
	struct netdev_adjacent *lower;

	lower = list_first_or_null_rcu(&dev->adj_list.lower,
			struct netdev_adjacent, list);
	if (lower)
		return lower->private;
	return NULL;
}
EXPORT_SYMBOL(netdev_lower_get_first_private_rcu);

J
Jiri Pirko 已提交
5101 5102 5103 5104 5105 5106 5107 5108 5109
/**
 * netdev_master_upper_dev_get_rcu - Get master upper device
 * @dev: device
 *
 * Find a master upper device and return pointer to it or NULL in case
 * it's not there. The caller must hold the RCU read lock.
 */
struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev)
{
5110
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
5111

5112
	upper = list_first_or_null_rcu(&dev->adj_list.upper,
5113
				       struct netdev_adjacent, list);
J
Jiri Pirko 已提交
5114 5115 5116 5117 5118 5119
	if (upper && likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get_rcu);

5120
static int netdev_adjacent_sysfs_add(struct net_device *dev,
5121 5122 5123 5124 5125 5126 5127 5128 5129
			      struct net_device *adj_dev,
			      struct list_head *dev_list)
{
	char linkname[IFNAMSIZ+7];
	sprintf(linkname, dev_list == &dev->adj_list.upper ?
		"upper_%s" : "lower_%s", adj_dev->name);
	return sysfs_create_link(&(dev->dev.kobj), &(adj_dev->dev.kobj),
				 linkname);
}
5130
static void netdev_adjacent_sysfs_del(struct net_device *dev,
5131 5132 5133 5134 5135 5136 5137 5138 5139
			       char *name,
			       struct list_head *dev_list)
{
	char linkname[IFNAMSIZ+7];
	sprintf(linkname, dev_list == &dev->adj_list.upper ?
		"upper_%s" : "lower_%s", name);
	sysfs_remove_link(&(dev->dev.kobj), linkname);
}

5140 5141 5142 5143 5144 5145 5146 5147
static inline bool netdev_adjacent_is_neigh_list(struct net_device *dev,
						 struct net_device *adj_dev,
						 struct list_head *dev_list)
{
	return (dev_list == &dev->adj_list.upper ||
		dev_list == &dev->adj_list.lower) &&
		net_eq(dev_net(dev), dev_net(adj_dev));
}
5148

5149 5150
static int __netdev_adjacent_dev_insert(struct net_device *dev,
					struct net_device *adj_dev,
5151
					struct list_head *dev_list,
5152
					void *private, bool master)
5153 5154
{
	struct netdev_adjacent *adj;
5155
	int ret;
5156

5157
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170

	if (adj) {
		adj->ref_nr++;
		return 0;
	}

	adj = kmalloc(sizeof(*adj), GFP_KERNEL);
	if (!adj)
		return -ENOMEM;

	adj->dev = adj_dev;
	adj->master = master;
	adj->ref_nr = 1;
5171
	adj->private = private;
5172
	dev_hold(adj_dev);
5173 5174 5175

	pr_debug("dev_hold for %s, because of link added from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
5176

5177
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list)) {
5178
		ret = netdev_adjacent_sysfs_add(dev, adj_dev, dev_list);
5179 5180 5181 5182
		if (ret)
			goto free_adj;
	}

5183
	/* Ensure that master link is always the first item in list. */
5184 5185 5186 5187
	if (master) {
		ret = sysfs_create_link(&(dev->dev.kobj),
					&(adj_dev->dev.kobj), "master");
		if (ret)
5188
			goto remove_symlinks;
5189

5190
		list_add_rcu(&adj->list, dev_list);
5191
	} else {
5192
		list_add_tail_rcu(&adj->list, dev_list);
5193
	}
5194 5195

	return 0;
5196

5197
remove_symlinks:
5198
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
5199
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
5200 5201
free_adj:
	kfree(adj);
5202
	dev_put(adj_dev);
5203 5204

	return ret;
5205 5206
}

S
stephen hemminger 已提交
5207 5208 5209
static void __netdev_adjacent_dev_remove(struct net_device *dev,
					 struct net_device *adj_dev,
					 struct list_head *dev_list)
5210 5211 5212
{
	struct netdev_adjacent *adj;

5213
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
5214

5215 5216 5217
	if (!adj) {
		pr_err("tried to remove device %s from %s\n",
		       dev->name, adj_dev->name);
5218
		BUG();
5219
	}
5220 5221

	if (adj->ref_nr > 1) {
5222 5223
		pr_debug("%s to %s ref_nr-- = %d\n", dev->name, adj_dev->name,
			 adj->ref_nr-1);
5224 5225 5226 5227
		adj->ref_nr--;
		return;
	}

5228 5229 5230
	if (adj->master)
		sysfs_remove_link(&(dev->dev.kobj), "master");

5231
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
5232
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
5233

5234
	list_del_rcu(&adj->list);
5235 5236
	pr_debug("dev_put for %s, because link removed from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
5237 5238 5239 5240
	dev_put(adj_dev);
	kfree_rcu(adj, rcu);
}

S
stephen hemminger 已提交
5241 5242 5243 5244 5245
static int __netdev_adjacent_dev_link_lists(struct net_device *dev,
					    struct net_device *upper_dev,
					    struct list_head *up_list,
					    struct list_head *down_list,
					    void *private, bool master)
5246 5247 5248
{
	int ret;

5249 5250
	ret = __netdev_adjacent_dev_insert(dev, upper_dev, up_list, private,
					   master);
5251 5252 5253
	if (ret)
		return ret;

5254 5255
	ret = __netdev_adjacent_dev_insert(upper_dev, dev, down_list, private,
					   false);
5256
	if (ret) {
5257
		__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
5258 5259 5260 5261 5262 5263
		return ret;
	}

	return 0;
}

S
stephen hemminger 已提交
5264 5265
static int __netdev_adjacent_dev_link(struct net_device *dev,
				      struct net_device *upper_dev)
5266
{
5267 5268 5269
	return __netdev_adjacent_dev_link_lists(dev, upper_dev,
						&dev->all_adj_list.upper,
						&upper_dev->all_adj_list.lower,
5270
						NULL, false);
5271 5272
}

S
stephen hemminger 已提交
5273 5274 5275 5276
static void __netdev_adjacent_dev_unlink_lists(struct net_device *dev,
					       struct net_device *upper_dev,
					       struct list_head *up_list,
					       struct list_head *down_list)
5277
{
5278 5279
	__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
	__netdev_adjacent_dev_remove(upper_dev, dev, down_list);
5280 5281
}

S
stephen hemminger 已提交
5282 5283
static void __netdev_adjacent_dev_unlink(struct net_device *dev,
					 struct net_device *upper_dev)
5284
{
5285 5286 5287 5288 5289
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev,
					   &dev->all_adj_list.upper,
					   &upper_dev->all_adj_list.lower);
}

S
stephen hemminger 已提交
5290 5291 5292
static int __netdev_adjacent_dev_link_neighbour(struct net_device *dev,
						struct net_device *upper_dev,
						void *private, bool master)
5293 5294 5295 5296 5297 5298 5299 5300 5301
{
	int ret = __netdev_adjacent_dev_link(dev, upper_dev);

	if (ret)
		return ret;

	ret = __netdev_adjacent_dev_link_lists(dev, upper_dev,
					       &dev->adj_list.upper,
					       &upper_dev->adj_list.lower,
5302
					       private, master);
5303 5304 5305 5306 5307 5308
	if (ret) {
		__netdev_adjacent_dev_unlink(dev, upper_dev);
		return ret;
	}

	return 0;
5309 5310
}

S
stephen hemminger 已提交
5311 5312
static void __netdev_adjacent_dev_unlink_neighbour(struct net_device *dev,
						   struct net_device *upper_dev)
5313 5314 5315 5316 5317 5318
{
	__netdev_adjacent_dev_unlink(dev, upper_dev);
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev,
					   &dev->adj_list.upper,
					   &upper_dev->adj_list.lower);
}
5319

J
Jiri Pirko 已提交
5320
static int __netdev_upper_dev_link(struct net_device *dev,
5321 5322
				   struct net_device *upper_dev, bool master,
				   void *private)
J
Jiri Pirko 已提交
5323
{
5324
	struct netdev_notifier_changeupper_info changeupper_info;
5325 5326
	struct netdev_adjacent *i, *j, *to_i, *to_j;
	int ret = 0;
J
Jiri Pirko 已提交
5327 5328 5329 5330 5331 5332 5333

	ASSERT_RTNL();

	if (dev == upper_dev)
		return -EBUSY;

	/* To prevent loops, check if dev is not upper device to upper_dev. */
5334
	if (__netdev_find_adj(upper_dev, dev, &upper_dev->all_adj_list.upper))
J
Jiri Pirko 已提交
5335 5336
		return -EBUSY;

5337
	if (__netdev_find_adj(dev, upper_dev, &dev->adj_list.upper))
J
Jiri Pirko 已提交
5338 5339 5340 5341 5342
		return -EEXIST;

	if (master && netdev_master_upper_dev_get(dev))
		return -EBUSY;

5343 5344 5345 5346
	changeupper_info.upper_dev = upper_dev;
	changeupper_info.master = master;
	changeupper_info.linking = true;

5347 5348
	ret = __netdev_adjacent_dev_link_neighbour(dev, upper_dev, private,
						   master);
5349 5350
	if (ret)
		return ret;
J
Jiri Pirko 已提交
5351

5352
	/* Now that we linked these devs, make all the upper_dev's
5353
	 * all_adj_list.upper visible to every dev's all_adj_list.lower an
5354 5355 5356
	 * versa, and don't forget the devices itself. All of these
	 * links are non-neighbours.
	 */
5357 5358 5359 5360
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list) {
			pr_debug("Interlinking %s with %s, non-neighbour\n",
				 i->dev->name, j->dev->name);
5361 5362 5363 5364 5365 5366 5367
			ret = __netdev_adjacent_dev_link(i->dev, j->dev);
			if (ret)
				goto rollback_mesh;
		}
	}

	/* add dev to every upper_dev's upper device */
5368 5369 5370
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list) {
		pr_debug("linking %s's upper device %s with %s\n",
			 upper_dev->name, i->dev->name, dev->name);
5371 5372 5373 5374 5375 5376
		ret = __netdev_adjacent_dev_link(dev, i->dev);
		if (ret)
			goto rollback_upper_mesh;
	}

	/* add upper_dev to every dev's lower device */
5377 5378 5379
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
		pr_debug("linking %s's lower device %s with %s\n", dev->name,
			 i->dev->name, upper_dev->name);
5380 5381 5382 5383
		ret = __netdev_adjacent_dev_link(i->dev, upper_dev);
		if (ret)
			goto rollback_lower_mesh;
	}
J
Jiri Pirko 已提交
5384

5385 5386
	call_netdevice_notifiers_info(NETDEV_CHANGEUPPER, dev,
				      &changeupper_info.info);
J
Jiri Pirko 已提交
5387
	return 0;
5388 5389 5390

rollback_lower_mesh:
	to_i = i;
5391
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
5392 5393 5394 5395 5396 5397 5398 5399 5400
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);
	}

	i = NULL;

rollback_upper_mesh:
	to_i = i;
5401
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list) {
5402 5403 5404 5405 5406 5407 5408 5409 5410 5411
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(dev, i->dev);
	}

	i = j = NULL;

rollback_mesh:
	to_i = i;
	to_j = j;
5412 5413
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list) {
5414 5415 5416 5417 5418 5419 5420 5421
			if (i == to_i && j == to_j)
				break;
			__netdev_adjacent_dev_unlink(i->dev, j->dev);
		}
		if (i == to_i)
			break;
	}

5422
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
5423 5424

	return ret;
J
Jiri Pirko 已提交
5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439
}

/**
 * netdev_upper_dev_link - Add a link to the upper device
 * @dev: device
 * @upper_dev: new upper device
 *
 * Adds a link to device which is upper to this one. The caller must hold
 * the RTNL lock. On a failure a negative errno code is returned.
 * On success the reference counts are adjusted and the function
 * returns zero.
 */
int netdev_upper_dev_link(struct net_device *dev,
			  struct net_device *upper_dev)
{
5440
	return __netdev_upper_dev_link(dev, upper_dev, false, NULL);
J
Jiri Pirko 已提交
5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457
}
EXPORT_SYMBOL(netdev_upper_dev_link);

/**
 * netdev_master_upper_dev_link - Add a master link to the upper device
 * @dev: device
 * @upper_dev: new upper device
 *
 * Adds a link to device which is upper to this one. In this case, only
 * one master upper device can be linked, although other non-master devices
 * might be linked as well. The caller must hold the RTNL lock.
 * On a failure a negative errno code is returned. On success the reference
 * counts are adjusted and the function returns zero.
 */
int netdev_master_upper_dev_link(struct net_device *dev,
				 struct net_device *upper_dev)
{
5458
	return __netdev_upper_dev_link(dev, upper_dev, true, NULL);
J
Jiri Pirko 已提交
5459 5460 5461
}
EXPORT_SYMBOL(netdev_master_upper_dev_link);

5462 5463 5464 5465 5466 5467 5468 5469
int netdev_master_upper_dev_link_private(struct net_device *dev,
					 struct net_device *upper_dev,
					 void *private)
{
	return __netdev_upper_dev_link(dev, upper_dev, true, private);
}
EXPORT_SYMBOL(netdev_master_upper_dev_link_private);

J
Jiri Pirko 已提交
5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480
/**
 * netdev_upper_dev_unlink - Removes a link to upper device
 * @dev: device
 * @upper_dev: new upper device
 *
 * Removes a link to device which is upper to this one. The caller must hold
 * the RTNL lock.
 */
void netdev_upper_dev_unlink(struct net_device *dev,
			     struct net_device *upper_dev)
{
5481
	struct netdev_notifier_changeupper_info changeupper_info;
5482
	struct netdev_adjacent *i, *j;
J
Jiri Pirko 已提交
5483 5484
	ASSERT_RTNL();

5485 5486 5487 5488
	changeupper_info.upper_dev = upper_dev;
	changeupper_info.master = netdev_master_upper_dev_get(dev) == upper_dev;
	changeupper_info.linking = false;

5489
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
5490 5491 5492 5493 5494

	/* Here is the tricky part. We must remove all dev's lower
	 * devices from all upper_dev's upper devices and vice
	 * versa, to maintain the graph relationship.
	 */
5495 5496
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list)
5497 5498 5499 5500 5501
			__netdev_adjacent_dev_unlink(i->dev, j->dev);

	/* remove also the devices itself from lower/upper device
	 * list
	 */
5502
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
5503 5504
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);

5505
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list)
5506 5507
		__netdev_adjacent_dev_unlink(dev, i->dev);

5508 5509
	call_netdevice_notifiers_info(NETDEV_CHANGEUPPER, dev,
				      &changeupper_info.info);
J
Jiri Pirko 已提交
5510 5511 5512
}
EXPORT_SYMBOL(netdev_upper_dev_unlink);

5513 5514 5515
/**
 * netdev_bonding_info_change - Dispatch event about slave change
 * @dev: device
5516
 * @bonding_info: info to dispatch
5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532
 *
 * Send NETDEV_BONDING_INFO to netdev notifiers with info.
 * The caller must hold the RTNL lock.
 */
void netdev_bonding_info_change(struct net_device *dev,
				struct netdev_bonding_info *bonding_info)
{
	struct netdev_notifier_bonding_info	info;

	memcpy(&info.bonding_info, bonding_info,
	       sizeof(struct netdev_bonding_info));
	call_netdevice_notifiers_info(NETDEV_BONDING_INFO, dev,
				      &info.info);
}
EXPORT_SYMBOL(netdev_bonding_info_change);

E
Eric Dumazet 已提交
5533
static void netdev_adjacent_add_links(struct net_device *dev)
5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557
{
	struct netdev_adjacent *iter;

	struct net *net = dev_net(dev);

	list_for_each_entry(iter, &dev->adj_list.upper, list) {
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.lower);
		netdev_adjacent_sysfs_add(dev, iter->dev,
					  &dev->adj_list.upper);
	}

	list_for_each_entry(iter, &dev->adj_list.lower, list) {
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.upper);
		netdev_adjacent_sysfs_add(dev, iter->dev,
					  &dev->adj_list.lower);
	}
}

E
Eric Dumazet 已提交
5558
static void netdev_adjacent_del_links(struct net_device *dev)
5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582
{
	struct netdev_adjacent *iter;

	struct net *net = dev_net(dev);

	list_for_each_entry(iter, &dev->adj_list.upper, list) {
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
		netdev_adjacent_sysfs_del(iter->dev, dev->name,
					  &iter->dev->adj_list.lower);
		netdev_adjacent_sysfs_del(dev, iter->dev->name,
					  &dev->adj_list.upper);
	}

	list_for_each_entry(iter, &dev->adj_list.lower, list) {
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
		netdev_adjacent_sysfs_del(iter->dev, dev->name,
					  &iter->dev->adj_list.upper);
		netdev_adjacent_sysfs_del(dev, iter->dev->name,
					  &dev->adj_list.lower);
	}
}

5583
void netdev_adjacent_rename_links(struct net_device *dev, char *oldname)
5584
{
5585
	struct netdev_adjacent *iter;
5586

5587 5588
	struct net *net = dev_net(dev);

5589
	list_for_each_entry(iter, &dev->adj_list.upper, list) {
5590 5591
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
5592 5593 5594 5595 5596
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.lower);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.lower);
	}
5597

5598
	list_for_each_entry(iter, &dev->adj_list.lower, list) {
5599 5600
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
5601 5602 5603 5604 5605
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.upper);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.upper);
	}
5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622
}

void *netdev_lower_dev_get_private(struct net_device *dev,
				   struct net_device *lower_dev)
{
	struct netdev_adjacent *lower;

	if (!lower_dev)
		return NULL;
	lower = __netdev_find_adj(dev, lower_dev, &dev->adj_list.lower);
	if (!lower)
		return NULL;

	return lower->private;
}
EXPORT_SYMBOL(netdev_lower_dev_get_private);

5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646

int dev_get_nest_level(struct net_device *dev,
		       bool (*type_check)(struct net_device *dev))
{
	struct net_device *lower = NULL;
	struct list_head *iter;
	int max_nest = -1;
	int nest;

	ASSERT_RTNL();

	netdev_for_each_lower_dev(dev, lower, iter) {
		nest = dev_get_nest_level(lower, type_check);
		if (max_nest < nest)
			max_nest = nest;
	}

	if (type_check(dev))
		max_nest++;

	return max_nest;
}
EXPORT_SYMBOL(dev_get_nest_level);

5647 5648
static void dev_change_rx_flags(struct net_device *dev, int flags)
{
5649 5650
	const struct net_device_ops *ops = dev->netdev_ops;

5651
	if (ops->ndo_change_rx_flags)
5652
		ops->ndo_change_rx_flags(dev, flags);
5653 5654
}

5655
static int __dev_set_promiscuity(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5656
{
5657
	unsigned int old_flags = dev->flags;
5658 5659
	kuid_t uid;
	kgid_t gid;
L
Linus Torvalds 已提交
5660

5661 5662
	ASSERT_RTNL();

5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673
	dev->flags |= IFF_PROMISC;
	dev->promiscuity += inc;
	if (dev->promiscuity == 0) {
		/*
		 * Avoid overflow.
		 * If inc causes overflow, untouch promisc and return error.
		 */
		if (inc < 0)
			dev->flags &= ~IFF_PROMISC;
		else {
			dev->promiscuity -= inc;
5674 5675
			pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n",
				dev->name);
5676 5677 5678
			return -EOVERFLOW;
		}
	}
5679
	if (dev->flags != old_flags) {
5680 5681 5682
		pr_info("device %s %s promiscuous mode\n",
			dev->name,
			dev->flags & IFF_PROMISC ? "entered" : "left");
5683 5684
		if (audit_enabled) {
			current_uid_gid(&uid, &gid);
5685 5686 5687 5688 5689
			audit_log(current->audit_context, GFP_ATOMIC,
				AUDIT_ANOM_PROMISCUOUS,
				"dev=%s prom=%d old_prom=%d auid=%u uid=%u gid=%u ses=%u",
				dev->name, (dev->flags & IFF_PROMISC),
				(old_flags & IFF_PROMISC),
5690
				from_kuid(&init_user_ns, audit_get_loginuid(current)),
5691 5692
				from_kuid(&init_user_ns, uid),
				from_kgid(&init_user_ns, gid),
5693
				audit_get_sessionid(current));
5694
		}
5695

5696
		dev_change_rx_flags(dev, IFF_PROMISC);
L
Linus Torvalds 已提交
5697
	}
5698 5699
	if (notify)
		__dev_notify_flags(dev, old_flags, IFF_PROMISC);
5700
	return 0;
L
Linus Torvalds 已提交
5701 5702
}

5703 5704 5705 5706 5707 5708 5709 5710 5711
/**
 *	dev_set_promiscuity	- update promiscuity count on a device
 *	@dev: device
 *	@inc: modifier
 *
 *	Add or remove promiscuity from a device. While the count in the device
 *	remains above zero the interface remains promiscuous. Once it hits zero
 *	the device reverts back to normal filtering operation. A negative inc
 *	value is used to drop promiscuity on the device.
5712
 *	Return 0 if successful or a negative errno code on error.
5713
 */
5714
int dev_set_promiscuity(struct net_device *dev, int inc)
5715
{
5716
	unsigned int old_flags = dev->flags;
5717
	int err;
5718

5719
	err = __dev_set_promiscuity(dev, inc, true);
5720
	if (err < 0)
5721
		return err;
5722 5723
	if (dev->flags != old_flags)
		dev_set_rx_mode(dev);
5724
	return err;
5725
}
E
Eric Dumazet 已提交
5726
EXPORT_SYMBOL(dev_set_promiscuity);
5727

5728
static int __dev_set_allmulti(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5729
{
5730
	unsigned int old_flags = dev->flags, old_gflags = dev->gflags;
L
Linus Torvalds 已提交
5731

5732 5733
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5734
	dev->flags |= IFF_ALLMULTI;
5735 5736 5737 5738 5739 5740 5741 5742 5743 5744
	dev->allmulti += inc;
	if (dev->allmulti == 0) {
		/*
		 * Avoid overflow.
		 * If inc causes overflow, untouch allmulti and return error.
		 */
		if (inc < 0)
			dev->flags &= ~IFF_ALLMULTI;
		else {
			dev->allmulti -= inc;
5745 5746
			pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n",
				dev->name);
5747 5748 5749
			return -EOVERFLOW;
		}
	}
5750
	if (dev->flags ^ old_flags) {
5751
		dev_change_rx_flags(dev, IFF_ALLMULTI);
5752
		dev_set_rx_mode(dev);
5753 5754 5755
		if (notify)
			__dev_notify_flags(dev, old_flags,
					   dev->gflags ^ old_gflags);
5756
	}
5757
	return 0;
5758
}
5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774 5775 5776

/**
 *	dev_set_allmulti	- update allmulti count on a device
 *	@dev: device
 *	@inc: modifier
 *
 *	Add or remove reception of all multicast frames to a device. While the
 *	count in the device remains above zero the interface remains listening
 *	to all interfaces. Once it hits zero the device reverts back to normal
 *	filtering operation. A negative @inc value is used to drop the counter
 *	when releasing a resource needing all multicasts.
 *	Return 0 if successful or a negative errno code on error.
 */

int dev_set_allmulti(struct net_device *dev, int inc)
{
	return __dev_set_allmulti(dev, inc, true);
}
E
Eric Dumazet 已提交
5777
EXPORT_SYMBOL(dev_set_allmulti);
5778 5779 5780 5781

/*
 *	Upload unicast and multicast address lists to device and
 *	configure RX filtering. When the device doesn't support unicast
J
Joe Perches 已提交
5782
 *	filtering it is put in promiscuous mode while unicast addresses
5783 5784 5785 5786
 *	are present.
 */
void __dev_set_rx_mode(struct net_device *dev)
{
5787 5788
	const struct net_device_ops *ops = dev->netdev_ops;

5789 5790 5791 5792 5793
	/* dev_open will call this function so the list will stay sane. */
	if (!(dev->flags&IFF_UP))
		return;

	if (!netif_device_present(dev))
5794
		return;
5795

5796
	if (!(dev->priv_flags & IFF_UNICAST_FLT)) {
5797 5798 5799
		/* Unicast addresses changes may only happen under the rtnl,
		 * therefore calling __dev_set_promiscuity here is safe.
		 */
5800
		if (!netdev_uc_empty(dev) && !dev->uc_promisc) {
5801
			__dev_set_promiscuity(dev, 1, false);
5802
			dev->uc_promisc = true;
5803
		} else if (netdev_uc_empty(dev) && dev->uc_promisc) {
5804
			__dev_set_promiscuity(dev, -1, false);
5805
			dev->uc_promisc = false;
5806 5807
		}
	}
5808 5809 5810

	if (ops->ndo_set_rx_mode)
		ops->ndo_set_rx_mode(dev);
5811 5812 5813 5814
}

void dev_set_rx_mode(struct net_device *dev)
{
5815
	netif_addr_lock_bh(dev);
5816
	__dev_set_rx_mode(dev);
5817
	netif_addr_unlock_bh(dev);
L
Linus Torvalds 已提交
5818 5819
}

5820 5821 5822 5823 5824 5825
/**
 *	dev_get_flags - get flags reported to userspace
 *	@dev: device
 *
 *	Get the combination of flag bits exported through APIs to userspace.
 */
5826
unsigned int dev_get_flags(const struct net_device *dev)
L
Linus Torvalds 已提交
5827
{
5828
	unsigned int flags;
L
Linus Torvalds 已提交
5829 5830 5831

	flags = (dev->flags & ~(IFF_PROMISC |
				IFF_ALLMULTI |
S
Stefan Rompf 已提交
5832 5833 5834
				IFF_RUNNING |
				IFF_LOWER_UP |
				IFF_DORMANT)) |
L
Linus Torvalds 已提交
5835 5836 5837
		(dev->gflags & (IFF_PROMISC |
				IFF_ALLMULTI));

S
Stefan Rompf 已提交
5838 5839 5840 5841 5842 5843 5844 5845
	if (netif_running(dev)) {
		if (netif_oper_up(dev))
			flags |= IFF_RUNNING;
		if (netif_carrier_ok(dev))
			flags |= IFF_LOWER_UP;
		if (netif_dormant(dev))
			flags |= IFF_DORMANT;
	}
L
Linus Torvalds 已提交
5846 5847 5848

	return flags;
}
E
Eric Dumazet 已提交
5849
EXPORT_SYMBOL(dev_get_flags);
L
Linus Torvalds 已提交
5850

5851
int __dev_change_flags(struct net_device *dev, unsigned int flags)
L
Linus Torvalds 已提交
5852
{
5853
	unsigned int old_flags = dev->flags;
5854
	int ret;
L
Linus Torvalds 已提交
5855

5856 5857
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871
	/*
	 *	Set the flags on our device.
	 */

	dev->flags = (flags & (IFF_DEBUG | IFF_NOTRAILERS | IFF_NOARP |
			       IFF_DYNAMIC | IFF_MULTICAST | IFF_PORTSEL |
			       IFF_AUTOMEDIA)) |
		     (dev->flags & (IFF_UP | IFF_VOLATILE | IFF_PROMISC |
				    IFF_ALLMULTI));

	/*
	 *	Load in the correct multicast list now the flags have changed.
	 */

5872 5873
	if ((old_flags ^ flags) & IFF_MULTICAST)
		dev_change_rx_flags(dev, IFF_MULTICAST);
5874

5875
	dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5876 5877 5878 5879 5880 5881 5882 5883

	/*
	 *	Have we downed the interface. We handle IFF_UP ourselves
	 *	according to user attempts to set it, rather than blindly
	 *	setting it.
	 */

	ret = 0;
5884
	if ((old_flags ^ flags) & IFF_UP)
5885
		ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev);
L
Linus Torvalds 已提交
5886 5887

	if ((flags ^ dev->gflags) & IFF_PROMISC) {
E
Eric Dumazet 已提交
5888
		int inc = (flags & IFF_PROMISC) ? 1 : -1;
5889
		unsigned int old_flags = dev->flags;
E
Eric Dumazet 已提交
5890

L
Linus Torvalds 已提交
5891
		dev->gflags ^= IFF_PROMISC;
5892 5893 5894 5895

		if (__dev_set_promiscuity(dev, inc, false) >= 0)
			if (dev->flags != old_flags)
				dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5896 5897 5898 5899 5900 5901 5902
	}

	/* NOTE: order of synchronization of IFF_PROMISC and IFF_ALLMULTI
	   is important. Some (broken) drivers set IFF_PROMISC, when
	   IFF_ALLMULTI is requested not asking us and not reporting.
	 */
	if ((flags ^ dev->gflags) & IFF_ALLMULTI) {
E
Eric Dumazet 已提交
5903 5904
		int inc = (flags & IFF_ALLMULTI) ? 1 : -1;

L
Linus Torvalds 已提交
5905
		dev->gflags ^= IFF_ALLMULTI;
5906
		__dev_set_allmulti(dev, inc, false);
L
Linus Torvalds 已提交
5907 5908
	}

5909 5910 5911
	return ret;
}

5912 5913
void __dev_notify_flags(struct net_device *dev, unsigned int old_flags,
			unsigned int gchanges)
5914 5915 5916
{
	unsigned int changes = dev->flags ^ old_flags;

5917
	if (gchanges)
5918
		rtmsg_ifinfo(RTM_NEWLINK, dev, gchanges, GFP_ATOMIC);
5919

5920 5921 5922 5923 5924 5925 5926 5927
	if (changes & IFF_UP) {
		if (dev->flags & IFF_UP)
			call_netdevice_notifiers(NETDEV_UP, dev);
		else
			call_netdevice_notifiers(NETDEV_DOWN, dev);
	}

	if (dev->flags & IFF_UP &&
5928 5929 5930 5931 5932 5933 5934
	    (changes & ~(IFF_UP | IFF_PROMISC | IFF_ALLMULTI | IFF_VOLATILE))) {
		struct netdev_notifier_change_info change_info;

		change_info.flags_changed = changes;
		call_netdevice_notifiers_info(NETDEV_CHANGE, dev,
					      &change_info.info);
	}
5935 5936 5937 5938 5939 5940 5941 5942 5943 5944
}

/**
 *	dev_change_flags - change device settings
 *	@dev: device
 *	@flags: device state flags
 *
 *	Change settings on device based state flags. The flags are
 *	in the userspace exported format.
 */
5945
int dev_change_flags(struct net_device *dev, unsigned int flags)
5946
{
5947
	int ret;
5948
	unsigned int changes, old_flags = dev->flags, old_gflags = dev->gflags;
5949 5950 5951 5952 5953

	ret = __dev_change_flags(dev, flags);
	if (ret < 0)
		return ret;

5954
	changes = (old_flags ^ dev->flags) | (old_gflags ^ dev->gflags);
5955
	__dev_notify_flags(dev, old_flags, changes);
L
Linus Torvalds 已提交
5956 5957
	return ret;
}
E
Eric Dumazet 已提交
5958
EXPORT_SYMBOL(dev_change_flags);
L
Linus Torvalds 已提交
5959

5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970
static int __dev_set_mtu(struct net_device *dev, int new_mtu)
{
	const struct net_device_ops *ops = dev->netdev_ops;

	if (ops->ndo_change_mtu)
		return ops->ndo_change_mtu(dev, new_mtu);

	dev->mtu = new_mtu;
	return 0;
}

5971 5972 5973 5974 5975 5976 5977
/**
 *	dev_set_mtu - Change maximum transfer unit
 *	@dev: device
 *	@new_mtu: new transfer unit
 *
 *	Change the maximum transfer size of the network device.
 */
L
Linus Torvalds 已提交
5978 5979
int dev_set_mtu(struct net_device *dev, int new_mtu)
{
5980
	int err, orig_mtu;
L
Linus Torvalds 已提交
5981 5982 5983 5984 5985 5986 5987 5988 5989 5990 5991

	if (new_mtu == dev->mtu)
		return 0;

	/*	MTU must be positive.	 */
	if (new_mtu < 0)
		return -EINVAL;

	if (!netif_device_present(dev))
		return -ENODEV;

5992 5993 5994 5995
	err = call_netdevice_notifiers(NETDEV_PRECHANGEMTU, dev);
	err = notifier_to_errno(err);
	if (err)
		return err;
5996

5997 5998
	orig_mtu = dev->mtu;
	err = __dev_set_mtu(dev, new_mtu);
5999

6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010
	if (!err) {
		err = call_netdevice_notifiers(NETDEV_CHANGEMTU, dev);
		err = notifier_to_errno(err);
		if (err) {
			/* setting mtu back and notifying everyone again,
			 * so that they have a chance to revert changes.
			 */
			__dev_set_mtu(dev, orig_mtu);
			call_netdevice_notifiers(NETDEV_CHANGEMTU, dev);
		}
	}
L
Linus Torvalds 已提交
6011 6012
	return err;
}
E
Eric Dumazet 已提交
6013
EXPORT_SYMBOL(dev_set_mtu);
L
Linus Torvalds 已提交
6014

6015 6016 6017 6018 6019 6020 6021 6022 6023 6024 6025
/**
 *	dev_set_group - Change group this device belongs to
 *	@dev: device
 *	@new_group: group this device should belong to
 */
void dev_set_group(struct net_device *dev, int new_group)
{
	dev->group = new_group;
}
EXPORT_SYMBOL(dev_set_group);

6026 6027 6028 6029 6030 6031 6032
/**
 *	dev_set_mac_address - Change Media Access Control Address
 *	@dev: device
 *	@sa: new address
 *
 *	Change the hardware (MAC) address of the device
 */
L
Linus Torvalds 已提交
6033 6034
int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa)
{
6035
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
6036 6037
	int err;

6038
	if (!ops->ndo_set_mac_address)
L
Linus Torvalds 已提交
6039 6040 6041 6042 6043
		return -EOPNOTSUPP;
	if (sa->sa_family != dev->type)
		return -EINVAL;
	if (!netif_device_present(dev))
		return -ENODEV;
6044
	err = ops->ndo_set_mac_address(dev, sa);
6045 6046
	if (err)
		return err;
J
Jiri Pirko 已提交
6047
	dev->addr_assign_type = NET_ADDR_SET;
6048
	call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
6049
	add_device_randomness(dev->dev_addr, dev->addr_len);
6050
	return 0;
L
Linus Torvalds 已提交
6051
}
E
Eric Dumazet 已提交
6052
EXPORT_SYMBOL(dev_set_mac_address);
L
Linus Torvalds 已提交
6053

J
Jiri Pirko 已提交
6054 6055 6056
/**
 *	dev_change_carrier - Change device carrier
 *	@dev: device
6057
 *	@new_carrier: new value
J
Jiri Pirko 已提交
6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072
 *
 *	Change device carrier
 */
int dev_change_carrier(struct net_device *dev, bool new_carrier)
{
	const struct net_device_ops *ops = dev->netdev_ops;

	if (!ops->ndo_change_carrier)
		return -EOPNOTSUPP;
	if (!netif_device_present(dev))
		return -ENODEV;
	return ops->ndo_change_carrier(dev, new_carrier);
}
EXPORT_SYMBOL(dev_change_carrier);

6073 6074 6075 6076 6077 6078 6079 6080
/**
 *	dev_get_phys_port_id - Get device physical port ID
 *	@dev: device
 *	@ppid: port ID
 *
 *	Get device physical port ID
 */
int dev_get_phys_port_id(struct net_device *dev,
6081
			 struct netdev_phys_item_id *ppid)
6082 6083 6084 6085 6086 6087 6088 6089 6090
{
	const struct net_device_ops *ops = dev->netdev_ops;

	if (!ops->ndo_get_phys_port_id)
		return -EOPNOTSUPP;
	return ops->ndo_get_phys_port_id(dev, ppid);
}
EXPORT_SYMBOL(dev_get_phys_port_id);

6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108
/**
 *	dev_get_phys_port_name - Get device physical port name
 *	@dev: device
 *	@name: port name
 *
 *	Get device physical port name
 */
int dev_get_phys_port_name(struct net_device *dev,
			   char *name, size_t len)
{
	const struct net_device_ops *ops = dev->netdev_ops;

	if (!ops->ndo_get_phys_port_name)
		return -EOPNOTSUPP;
	return ops->ndo_get_phys_port_name(dev, name, len);
}
EXPORT_SYMBOL(dev_get_phys_port_name);

6109 6110 6111 6112 6113 6114 6115 6116 6117 6118 6119 6120 6121 6122 6123 6124 6125 6126 6127 6128
/**
 *	dev_change_proto_down - update protocol port state information
 *	@dev: device
 *	@proto_down: new value
 *
 *	This info can be used by switch drivers to set the phys state of the
 *	port.
 */
int dev_change_proto_down(struct net_device *dev, bool proto_down)
{
	const struct net_device_ops *ops = dev->netdev_ops;

	if (!ops->ndo_change_proto_down)
		return -EOPNOTSUPP;
	if (!netif_device_present(dev))
		return -ENODEV;
	return ops->ndo_change_proto_down(dev, proto_down);
}
EXPORT_SYMBOL(dev_change_proto_down);

L
Linus Torvalds 已提交
6129 6130
/**
 *	dev_new_index	-	allocate an ifindex
6131
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
6132 6133 6134 6135 6136
 *
 *	Returns a suitable unique value for a new device interface
 *	number.  The caller must hold the rtnl semaphore or the
 *	dev_base_lock to be sure it remains unique.
 */
6137
static int dev_new_index(struct net *net)
L
Linus Torvalds 已提交
6138
{
6139
	int ifindex = net->ifindex;
L
Linus Torvalds 已提交
6140 6141 6142
	for (;;) {
		if (++ifindex <= 0)
			ifindex = 1;
6143
		if (!__dev_get_by_index(net, ifindex))
6144
			return net->ifindex = ifindex;
L
Linus Torvalds 已提交
6145 6146 6147 6148
	}
}

/* Delayed registration/unregisteration */
6149
static LIST_HEAD(net_todo_list);
6150
DECLARE_WAIT_QUEUE_HEAD(netdev_unregistering_wq);
L
Linus Torvalds 已提交
6151

6152
static void net_set_todo(struct net_device *dev)
L
Linus Torvalds 已提交
6153 6154
{
	list_add_tail(&dev->todo_list, &net_todo_list);
6155
	dev_net(dev)->dev_unreg_count++;
L
Linus Torvalds 已提交
6156 6157
}

6158
static void rollback_registered_many(struct list_head *head)
6159
{
6160
	struct net_device *dev, *tmp;
6161
	LIST_HEAD(close_head);
6162

6163 6164 6165
	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

6166
	list_for_each_entry_safe(dev, tmp, head, unreg_list) {
6167
		/* Some devices call without registering
6168 6169
		 * for initialization unwind. Remove those
		 * devices and proceed with the remaining.
6170 6171
		 */
		if (dev->reg_state == NETREG_UNINITIALIZED) {
6172 6173
			pr_debug("unregister_netdevice: device %s/%p never was registered\n",
				 dev->name, dev);
6174

6175
			WARN_ON(1);
6176 6177
			list_del(&dev->unreg_list);
			continue;
6178
		}
6179
		dev->dismantle = true;
6180
		BUG_ON(dev->reg_state != NETREG_REGISTERED);
6181
	}
6182

6183
	/* If device is running, close it first. */
6184 6185
	list_for_each_entry(dev, head, unreg_list)
		list_add_tail(&dev->close_list, &close_head);
6186
	dev_close_many(&close_head, true);
6187

6188
	list_for_each_entry(dev, head, unreg_list) {
6189 6190
		/* And unlink it from device chain. */
		unlist_netdevice(dev);
6191

6192
		dev->reg_state = NETREG_UNREGISTERING;
6193
		on_each_cpu(flush_backlog, dev, 1);
6194
	}
6195 6196 6197

	synchronize_net();

6198
	list_for_each_entry(dev, head, unreg_list) {
6199 6200
		struct sk_buff *skb = NULL;

6201 6202
		/* Shutdown queueing discipline. */
		dev_shutdown(dev);
6203 6204


6205 6206 6207 6208
		/* Notify protocols, that we are about to destroy
		   this device. They should clean all the things.
		*/
		call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
6209

6210 6211 6212 6213 6214
		if (!dev->rtnl_link_ops ||
		    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
			skb = rtmsg_ifinfo_build_skb(RTM_DELLINK, dev, ~0U,
						     GFP_KERNEL);

6215 6216 6217
		/*
		 *	Flush the unicast and multicast chains
		 */
6218
		dev_uc_flush(dev);
6219
		dev_mc_flush(dev);
6220

6221 6222
		if (dev->netdev_ops->ndo_uninit)
			dev->netdev_ops->ndo_uninit(dev);
6223

6224 6225
		if (skb)
			rtmsg_ifinfo_send(skb, dev, GFP_KERNEL);
6226

J
Jiri Pirko 已提交
6227 6228
		/* Notifier chain MUST detach us all upper devices. */
		WARN_ON(netdev_has_any_upper_dev(dev));
6229

6230 6231
		/* Remove entries from kobject tree */
		netdev_unregister_kobject(dev);
6232 6233 6234 6235
#ifdef CONFIG_XPS
		/* Remove XPS queueing entries */
		netif_reset_xps_queues_gt(dev, 0);
#endif
6236
	}
6237

6238
	synchronize_net();
6239

6240
	list_for_each_entry(dev, head, unreg_list)
6241 6242 6243 6244 6245 6246 6247 6248 6249
		dev_put(dev);
}

static void rollback_registered(struct net_device *dev)
{
	LIST_HEAD(single);

	list_add(&dev->unreg_list, &single);
	rollback_registered_many(&single);
E
Eric Dumazet 已提交
6250
	list_del(&single);
6251 6252
}

6253 6254
static netdev_features_t netdev_fix_features(struct net_device *dev,
	netdev_features_t features)
6255
{
6256 6257 6258
	/* Fix illegal checksum combinations */
	if ((features & NETIF_F_HW_CSUM) &&
	    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
6259
		netdev_warn(dev, "mixed HW and IP checksum settings.\n");
6260 6261 6262
		features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
	}

6263
	/* TSO requires that SG is present as well. */
6264
	if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
6265
		netdev_dbg(dev, "Dropping TSO features since no SG feature.\n");
6266
		features &= ~NETIF_F_ALL_TSO;
6267 6268
	}

6269 6270 6271 6272 6273 6274 6275 6276 6277 6278 6279 6280 6281
	if ((features & NETIF_F_TSO) && !(features & NETIF_F_HW_CSUM) &&
					!(features & NETIF_F_IP_CSUM)) {
		netdev_dbg(dev, "Dropping TSO features since no CSUM feature.\n");
		features &= ~NETIF_F_TSO;
		features &= ~NETIF_F_TSO_ECN;
	}

	if ((features & NETIF_F_TSO6) && !(features & NETIF_F_HW_CSUM) &&
					 !(features & NETIF_F_IPV6_CSUM)) {
		netdev_dbg(dev, "Dropping TSO6 features since no CSUM feature.\n");
		features &= ~NETIF_F_TSO6;
	}

6282 6283 6284 6285
	/* TSO ECN requires that TSO is present as well. */
	if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN)
		features &= ~NETIF_F_TSO_ECN;

6286 6287
	/* Software GSO depends on SG. */
	if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) {
6288
		netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n");
6289 6290 6291
		features &= ~NETIF_F_GSO;
	}

6292
	/* UFO needs SG and checksumming */
6293
	if (features & NETIF_F_UFO) {
6294 6295 6296 6297
		/* maybe split UFO into V4 and V6? */
		if (!((features & NETIF_F_GEN_CSUM) ||
		    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))
			    == (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
6298
			netdev_dbg(dev,
6299
				"Dropping NETIF_F_UFO since no checksum offload features.\n");
6300 6301 6302 6303
			features &= ~NETIF_F_UFO;
		}

		if (!(features & NETIF_F_SG)) {
6304
			netdev_dbg(dev,
6305
				"Dropping NETIF_F_UFO since no NETIF_F_SG feature.\n");
6306 6307 6308 6309
			features &= ~NETIF_F_UFO;
		}
	}

J
Jiri Pirko 已提交
6310 6311 6312 6313 6314 6315 6316
#ifdef CONFIG_NET_RX_BUSY_POLL
	if (dev->netdev_ops->ndo_busy_poll)
		features |= NETIF_F_BUSY_POLL;
	else
#endif
		features &= ~NETIF_F_BUSY_POLL;

6317 6318 6319
	return features;
}

6320
int __netdev_update_features(struct net_device *dev)
6321
{
6322
	netdev_features_t features;
6323 6324
	int err = 0;

6325 6326
	ASSERT_RTNL();

6327 6328 6329 6330 6331 6332 6333 6334 6335
	features = netdev_get_wanted_features(dev);

	if (dev->netdev_ops->ndo_fix_features)
		features = dev->netdev_ops->ndo_fix_features(dev, features);

	/* driver might be less strict about feature dependencies */
	features = netdev_fix_features(dev, features);

	if (dev->features == features)
6336
		return 0;
6337

6338 6339
	netdev_dbg(dev, "Features changed: %pNF -> %pNF\n",
		&dev->features, &features);
6340 6341 6342 6343

	if (dev->netdev_ops->ndo_set_features)
		err = dev->netdev_ops->ndo_set_features(dev, features);

6344
	if (unlikely(err < 0)) {
6345
		netdev_err(dev,
6346 6347
			"set_features() failed (%d); wanted %pNF, left %pNF\n",
			err, &features, &dev->features);
6348 6349 6350 6351 6352 6353 6354 6355 6356
		return -1;
	}

	if (!err)
		dev->features = features;

	return 1;
}

6357 6358 6359 6360 6361 6362 6363 6364
/**
 *	netdev_update_features - recalculate device features
 *	@dev: the device to check
 *
 *	Recalculate dev->features set and send notifications if it
 *	has changed. Should be called after driver or hardware dependent
 *	conditions might have changed that influence the features.
 */
6365 6366 6367 6368
void netdev_update_features(struct net_device *dev)
{
	if (__netdev_update_features(dev))
		netdev_features_change(dev);
6369 6370 6371
}
EXPORT_SYMBOL(netdev_update_features);

6372 6373 6374 6375 6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388
/**
 *	netdev_change_features - recalculate device features
 *	@dev: the device to check
 *
 *	Recalculate dev->features set and send notifications even
 *	if they have not changed. Should be called instead of
 *	netdev_update_features() if also dev->vlan_features might
 *	have changed to allow the changes to be propagated to stacked
 *	VLAN devices.
 */
void netdev_change_features(struct net_device *dev)
{
	__netdev_update_features(dev);
	netdev_features_change(dev);
}
EXPORT_SYMBOL(netdev_change_features);

6389 6390 6391 6392 6393 6394 6395 6396 6397 6398 6399 6400 6401 6402 6403 6404 6405 6406 6407 6408 6409 6410 6411 6412 6413 6414 6415
/**
 *	netif_stacked_transfer_operstate -	transfer operstate
 *	@rootdev: the root or lower level device to transfer state from
 *	@dev: the device to transfer operstate to
 *
 *	Transfer operational state from root to device. This is normally
 *	called when a stacking relationship exists between the root
 *	device and the device(a leaf device).
 */
void netif_stacked_transfer_operstate(const struct net_device *rootdev,
					struct net_device *dev)
{
	if (rootdev->operstate == IF_OPER_DORMANT)
		netif_dormant_on(dev);
	else
		netif_dormant_off(dev);

	if (netif_carrier_ok(rootdev)) {
		if (!netif_carrier_ok(dev))
			netif_carrier_on(dev);
	} else {
		if (netif_carrier_ok(dev))
			netif_carrier_off(dev);
	}
}
EXPORT_SYMBOL(netif_stacked_transfer_operstate);

6416
#ifdef CONFIG_SYSFS
6417 6418 6419
static int netif_alloc_rx_queues(struct net_device *dev)
{
	unsigned int i, count = dev->num_rx_queues;
T
Tom Herbert 已提交
6420
	struct netdev_rx_queue *rx;
6421
	size_t sz = count * sizeof(*rx);
6422

T
Tom Herbert 已提交
6423
	BUG_ON(count < 1);
6424

6425 6426 6427 6428 6429 6430
	rx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!rx) {
		rx = vzalloc(sz);
		if (!rx)
			return -ENOMEM;
	}
T
Tom Herbert 已提交
6431 6432 6433
	dev->_rx = rx;

	for (i = 0; i < count; i++)
T
Tom Herbert 已提交
6434
		rx[i].dev = dev;
6435 6436
	return 0;
}
T
Tom Herbert 已提交
6437
#endif
6438

C
Changli Gao 已提交
6439 6440 6441 6442 6443 6444 6445
static void netdev_init_one_queue(struct net_device *dev,
				  struct netdev_queue *queue, void *_unused)
{
	/* Initialize queue lock */
	spin_lock_init(&queue->_xmit_lock);
	netdev_set_xmit_lockdep_class(&queue->_xmit_lock, dev->type);
	queue->xmit_lock_owner = -1;
6446
	netdev_queue_numa_node_write(queue, NUMA_NO_NODE);
C
Changli Gao 已提交
6447
	queue->dev = dev;
T
Tom Herbert 已提交
6448 6449 6450
#ifdef CONFIG_BQL
	dql_init(&queue->dql, HZ);
#endif
C
Changli Gao 已提交
6451 6452
}

6453 6454
static void netif_free_tx_queues(struct net_device *dev)
{
W
WANG Cong 已提交
6455
	kvfree(dev->_tx);
6456 6457
}

6458 6459 6460 6461
static int netif_alloc_netdev_queues(struct net_device *dev)
{
	unsigned int count = dev->num_tx_queues;
	struct netdev_queue *tx;
6462
	size_t sz = count * sizeof(*tx);
6463

6464 6465
	if (count < 1 || count > 0xffff)
		return -EINVAL;
6466

6467 6468 6469 6470 6471 6472
	tx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!tx) {
		tx = vzalloc(sz);
		if (!tx)
			return -ENOMEM;
	}
6473
	dev->_tx = tx;
T
Tom Herbert 已提交
6474

6475 6476
	netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
	spin_lock_init(&dev->tx_global_lock);
C
Changli Gao 已提交
6477 6478

	return 0;
6479 6480
}

6481 6482 6483 6484 6485 6486 6487 6488 6489 6490 6491
void netif_tx_stop_all_queues(struct net_device *dev)
{
	unsigned int i;

	for (i = 0; i < dev->num_tx_queues; i++) {
		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
		netif_tx_stop_queue(txq);
	}
}
EXPORT_SYMBOL(netif_tx_stop_all_queues);

L
Linus Torvalds 已提交
6492 6493 6494 6495 6496 6497 6498 6499 6500 6501 6502 6503 6504 6505 6506 6507 6508 6509 6510 6511
/**
 *	register_netdevice	- register a network device
 *	@dev: device to register
 *
 *	Take a completed network device structure and add it to the kernel
 *	interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier
 *	chain. 0 is returned on success. A negative errno code is returned
 *	on a failure to set up the device, or if the name is a duplicate.
 *
 *	Callers must hold the rtnl semaphore. You may want
 *	register_netdev() instead of this.
 *
 *	BUGS:
 *	The locking appears insufficient to guarantee two parallel registers
 *	will not get the same name.
 */

int register_netdevice(struct net_device *dev)
{
	int ret;
6512
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
6513 6514 6515 6516

	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

6517 6518
	might_sleep();

L
Linus Torvalds 已提交
6519 6520
	/* When net_device's are persistent, this will be fatal. */
	BUG_ON(dev->reg_state != NETREG_UNINITIALIZED);
6521
	BUG_ON(!net);
L
Linus Torvalds 已提交
6522

6523
	spin_lock_init(&dev->addr_list_lock);
6524
	netdev_set_addr_lockdep_class(dev);
L
Linus Torvalds 已提交
6525

6526
	ret = dev_get_valid_name(net, dev, dev->name);
6527 6528 6529
	if (ret < 0)
		goto out;

L
Linus Torvalds 已提交
6530
	/* Init, if this function is available */
6531 6532
	if (dev->netdev_ops->ndo_init) {
		ret = dev->netdev_ops->ndo_init(dev);
L
Linus Torvalds 已提交
6533 6534 6535
		if (ret) {
			if (ret > 0)
				ret = -EIO;
6536
			goto out;
L
Linus Torvalds 已提交
6537 6538
		}
	}
6539

6540 6541
	if (((dev->hw_features | dev->features) &
	     NETIF_F_HW_VLAN_CTAG_FILTER) &&
6542 6543 6544 6545 6546 6547 6548
	    (!dev->netdev_ops->ndo_vlan_rx_add_vid ||
	     !dev->netdev_ops->ndo_vlan_rx_kill_vid)) {
		netdev_WARN(dev, "Buggy VLAN acceleration in driver!\n");
		ret = -EINVAL;
		goto err_uninit;
	}

6549 6550 6551 6552 6553 6554
	ret = -EBUSY;
	if (!dev->ifindex)
		dev->ifindex = dev_new_index(net);
	else if (__dev_get_by_index(net, dev->ifindex))
		goto err_uninit;

6555 6556 6557 6558
	/* Transfer changeable features to wanted_features and enable
	 * software offloads (GSO and GRO).
	 */
	dev->hw_features |= NETIF_F_SOFT_FEATURES;
6559 6560
	dev->features |= NETIF_F_SOFT_FEATURES;
	dev->wanted_features = dev->features & dev->hw_features;
L
Linus Torvalds 已提交
6561

6562 6563
	if (!(dev->flags & IFF_LOOPBACK)) {
		dev->hw_features |= NETIF_F_NOCACHE_COPY;
6564 6565
	}

6566
	/* Make NETIF_F_HIGHDMA inheritable to VLAN devices.
6567
	 */
6568
	dev->vlan_features |= NETIF_F_HIGHDMA;
6569

6570 6571 6572 6573
	/* Make NETIF_F_SG inheritable to tunnel devices.
	 */
	dev->hw_enc_features |= NETIF_F_SG;

S
Simon Horman 已提交
6574 6575 6576 6577
	/* Make NETIF_F_SG inheritable to MPLS.
	 */
	dev->mpls_features |= NETIF_F_SG;

6578 6579 6580 6581 6582
	ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		goto err_uninit;

6583
	ret = netdev_register_kobject(dev);
6584
	if (ret)
6585
		goto err_uninit;
6586 6587
	dev->reg_state = NETREG_REGISTERED;

6588
	__netdev_update_features(dev);
6589

L
Linus Torvalds 已提交
6590 6591 6592 6593 6594 6595 6596
	/*
	 *	Default initial state at registry is that the
	 *	device is present.
	 */

	set_bit(__LINK_STATE_PRESENT, &dev->state);

6597 6598
	linkwatch_init_dev(dev);

L
Linus Torvalds 已提交
6599 6600
	dev_init_scheduler(dev);
	dev_hold(dev);
6601
	list_netdevice(dev);
6602
	add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
6603

6604 6605 6606 6607 6608 6609 6610
	/* If the device has permanent device address, driver should
	 * set dev_addr and also addr_assign_type should be set to
	 * NET_ADDR_PERM (default value).
	 */
	if (dev->addr_assign_type == NET_ADDR_PERM)
		memcpy(dev->perm_addr, dev->dev_addr, dev->addr_len);

L
Linus Torvalds 已提交
6611
	/* Notify protocols, that a new device appeared. */
6612
	ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
6613
	ret = notifier_to_errno(ret);
6614 6615 6616 6617
	if (ret) {
		rollback_registered(dev);
		dev->reg_state = NETREG_UNREGISTERED;
	}
6618 6619 6620 6621
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
6622 6623
	if (!dev->rtnl_link_ops ||
	    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
6624
		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
L
Linus Torvalds 已提交
6625 6626 6627

out:
	return ret;
6628 6629

err_uninit:
6630 6631
	if (dev->netdev_ops->ndo_uninit)
		dev->netdev_ops->ndo_uninit(dev);
6632
	goto out;
L
Linus Torvalds 已提交
6633
}
E
Eric Dumazet 已提交
6634
EXPORT_SYMBOL(register_netdevice);
L
Linus Torvalds 已提交
6635

6636 6637 6638 6639 6640 6641 6642 6643 6644 6645 6646 6647 6648 6649 6650 6651 6652 6653 6654 6655 6656 6657 6658 6659 6660 6661 6662 6663 6664 6665 6666
/**
 *	init_dummy_netdev	- init a dummy network device for NAPI
 *	@dev: device to init
 *
 *	This takes a network device structure and initialize the minimum
 *	amount of fields so it can be used to schedule NAPI polls without
 *	registering a full blown interface. This is to be used by drivers
 *	that need to tie several hardware interfaces to a single NAPI
 *	poll scheduler due to HW limitations.
 */
int init_dummy_netdev(struct net_device *dev)
{
	/* Clear everything. Note we don't initialize spinlocks
	 * are they aren't supposed to be taken by any of the
	 * NAPI code and this dummy netdev is supposed to be
	 * only ever used for NAPI polls
	 */
	memset(dev, 0, sizeof(struct net_device));

	/* make sure we BUG if trying to hit standard
	 * register/unregister code path
	 */
	dev->reg_state = NETREG_DUMMY;

	/* NAPI wants this */
	INIT_LIST_HEAD(&dev->napi_list);

	/* a dummy interface is started by default */
	set_bit(__LINK_STATE_PRESENT, &dev->state);
	set_bit(__LINK_STATE_START, &dev->state);

E
Eric Dumazet 已提交
6667 6668 6669 6670 6671
	/* Note : We dont allocate pcpu_refcnt for dummy devices,
	 * because users of this 'device' dont need to change
	 * its refcount.
	 */

6672 6673 6674 6675 6676
	return 0;
}
EXPORT_SYMBOL_GPL(init_dummy_netdev);


L
Linus Torvalds 已提交
6677 6678 6679 6680 6681 6682 6683 6684 6685
/**
 *	register_netdev	- register a network device
 *	@dev: device to register
 *
 *	Take a completed network device structure and add it to the kernel
 *	interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier
 *	chain. 0 is returned on success. A negative errno code is returned
 *	on a failure to set up the device, or if the name is a duplicate.
 *
6686
 *	This is a wrapper around register_netdevice that takes the rtnl semaphore
L
Linus Torvalds 已提交
6687 6688 6689 6690 6691 6692 6693 6694 6695 6696 6697 6698 6699 6700
 *	and expands the device name if you passed a format string to
 *	alloc_netdev.
 */
int register_netdev(struct net_device *dev)
{
	int err;

	rtnl_lock();
	err = register_netdevice(dev);
	rtnl_unlock();
	return err;
}
EXPORT_SYMBOL(register_netdev);

E
Eric Dumazet 已提交
6701 6702 6703 6704 6705 6706 6707 6708 6709 6710
int netdev_refcnt_read(const struct net_device *dev)
{
	int i, refcnt = 0;

	for_each_possible_cpu(i)
		refcnt += *per_cpu_ptr(dev->pcpu_refcnt, i);
	return refcnt;
}
EXPORT_SYMBOL(netdev_refcnt_read);

6711
/**
L
Linus Torvalds 已提交
6712
 * netdev_wait_allrefs - wait until all references are gone.
6713
 * @dev: target net_device
L
Linus Torvalds 已提交
6714 6715 6716 6717 6718 6719 6720
 *
 * This is called when unregistering network devices.
 *
 * Any protocol or device that holds a reference should register
 * for netdevice notification, and cleanup and put back the
 * reference if they receive an UNREGISTER event.
 * We can get stuck here if buggy protocols don't correctly
6721
 * call dev_put.
L
Linus Torvalds 已提交
6722 6723 6724 6725
 */
static void netdev_wait_allrefs(struct net_device *dev)
{
	unsigned long rebroadcast_time, warning_time;
E
Eric Dumazet 已提交
6726
	int refcnt;
L
Linus Torvalds 已提交
6727

6728 6729
	linkwatch_forget_dev(dev);

L
Linus Torvalds 已提交
6730
	rebroadcast_time = warning_time = jiffies;
E
Eric Dumazet 已提交
6731 6732 6733
	refcnt = netdev_refcnt_read(dev);

	while (refcnt != 0) {
L
Linus Torvalds 已提交
6734
		if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
6735
			rtnl_lock();
L
Linus Torvalds 已提交
6736 6737

			/* Rebroadcast unregister notification */
6738
			call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
L
Linus Torvalds 已提交
6739

6740
			__rtnl_unlock();
6741
			rcu_barrier();
6742 6743
			rtnl_lock();

6744
			call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
L
Linus Torvalds 已提交
6745 6746 6747 6748 6749 6750 6751 6752 6753 6754 6755
			if (test_bit(__LINK_STATE_LINKWATCH_PENDING,
				     &dev->state)) {
				/* We must not have linkwatch events
				 * pending on unregister. If this
				 * happens, we simply run the queue
				 * unscheduled, resulting in a noop
				 * for this device.
				 */
				linkwatch_run_queue();
			}

6756
			__rtnl_unlock();
L
Linus Torvalds 已提交
6757 6758 6759 6760 6761 6762

			rebroadcast_time = jiffies;
		}

		msleep(250);

E
Eric Dumazet 已提交
6763 6764
		refcnt = netdev_refcnt_read(dev);

L
Linus Torvalds 已提交
6765
		if (time_after(jiffies, warning_time + 10 * HZ)) {
6766 6767
			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
				 dev->name, refcnt);
L
Linus Torvalds 已提交
6768 6769 6770 6771 6772 6773 6774 6775 6776 6777 6778 6779 6780 6781 6782 6783 6784 6785 6786
			warning_time = jiffies;
		}
	}
}

/* The sequence is:
 *
 *	rtnl_lock();
 *	...
 *	register_netdevice(x1);
 *	register_netdevice(x2);
 *	...
 *	unregister_netdevice(y1);
 *	unregister_netdevice(y2);
 *      ...
 *	rtnl_unlock();
 *	free_netdev(y1);
 *	free_netdev(y2);
 *
H
Herbert Xu 已提交
6787
 * We are invoked by rtnl_unlock().
L
Linus Torvalds 已提交
6788
 * This allows us to deal with problems:
6789
 * 1) We can delete sysfs objects which invoke hotplug
L
Linus Torvalds 已提交
6790 6791 6792
 *    without deadlocking with linkwatch via keventd.
 * 2) Since we run with the RTNL semaphore not held, we can sleep
 *    safely in order to wait for the netdev refcnt to drop to zero.
H
Herbert Xu 已提交
6793 6794 6795
 *
 * We must not return until all unregister events added during
 * the interval the lock was held have been completed.
L
Linus Torvalds 已提交
6796 6797 6798
 */
void netdev_run_todo(void)
{
6799
	struct list_head list;
L
Linus Torvalds 已提交
6800 6801

	/* Snapshot list, allow later requests */
6802
	list_replace_init(&net_todo_list, &list);
H
Herbert Xu 已提交
6803 6804

	__rtnl_unlock();
6805

6806 6807

	/* Wait for rcu callbacks to finish before next phase */
6808 6809 6810
	if (!list_empty(&list))
		rcu_barrier();

L
Linus Torvalds 已提交
6811 6812
	while (!list_empty(&list)) {
		struct net_device *dev
6813
			= list_first_entry(&list, struct net_device, todo_list);
L
Linus Torvalds 已提交
6814 6815
		list_del(&dev->todo_list);

6816
		rtnl_lock();
6817
		call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6818
		__rtnl_unlock();
6819

6820
		if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
6821
			pr_err("network todo '%s' but state %d\n",
6822 6823 6824 6825
			       dev->name, dev->reg_state);
			dump_stack();
			continue;
		}
L
Linus Torvalds 已提交
6826

6827
		dev->reg_state = NETREG_UNREGISTERED;
L
Linus Torvalds 已提交
6828

6829
		netdev_wait_allrefs(dev);
L
Linus Torvalds 已提交
6830

6831
		/* paranoia */
E
Eric Dumazet 已提交
6832
		BUG_ON(netdev_refcnt_read(dev));
6833 6834
		BUG_ON(!list_empty(&dev->ptype_all));
		BUG_ON(!list_empty(&dev->ptype_specific));
6835 6836
		WARN_ON(rcu_access_pointer(dev->ip_ptr));
		WARN_ON(rcu_access_pointer(dev->ip6_ptr));
6837
		WARN_ON(dev->dn_ptr);
L
Linus Torvalds 已提交
6838

6839 6840
		if (dev->destructor)
			dev->destructor(dev);
6841

6842 6843 6844 6845 6846 6847
		/* Report a network device has been unregistered */
		rtnl_lock();
		dev_net(dev)->dev_unreg_count--;
		__rtnl_unlock();
		wake_up(&netdev_unregistering_wq);

6848 6849
		/* Free network device */
		kobject_put(&dev->dev.kobj);
L
Linus Torvalds 已提交
6850 6851 6852
	}
}

6853 6854 6855
/* Convert net_device_stats to rtnl_link_stats64.  They have the same
 * fields in the same order, with only the type differing.
 */
6856 6857
void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
			     const struct net_device_stats *netdev_stats)
6858 6859
{
#if BITS_PER_LONG == 64
6860 6861
	BUILD_BUG_ON(sizeof(*stats64) != sizeof(*netdev_stats));
	memcpy(stats64, netdev_stats, sizeof(*stats64));
6862 6863 6864 6865 6866 6867 6868 6869 6870 6871 6872
#else
	size_t i, n = sizeof(*stats64) / sizeof(u64);
	const unsigned long *src = (const unsigned long *)netdev_stats;
	u64 *dst = (u64 *)stats64;

	BUILD_BUG_ON(sizeof(*netdev_stats) / sizeof(unsigned long) !=
		     sizeof(*stats64) / sizeof(u64));
	for (i = 0; i < n; i++)
		dst[i] = src[i];
#endif
}
6873
EXPORT_SYMBOL(netdev_stats_to_stats64);
6874

6875 6876 6877
/**
 *	dev_get_stats	- get network device statistics
 *	@dev: device to get statistics from
6878
 *	@storage: place to store stats
6879
 *
6880 6881 6882 6883
 *	Get network statistics from device. Return @storage.
 *	The device driver may provide its own method by setting
 *	dev->netdev_ops->get_stats64 or dev->netdev_ops->get_stats;
 *	otherwise the internal statistics structure is used.
6884
 */
6885 6886
struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
					struct rtnl_link_stats64 *storage)
6887
{
6888 6889
	const struct net_device_ops *ops = dev->netdev_ops;

6890 6891
	if (ops->ndo_get_stats64) {
		memset(storage, 0, sizeof(*storage));
6892 6893
		ops->ndo_get_stats64(dev, storage);
	} else if (ops->ndo_get_stats) {
6894
		netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev));
6895 6896
	} else {
		netdev_stats_to_stats64(storage, &dev->stats);
6897
	}
6898
	storage->rx_dropped += atomic_long_read(&dev->rx_dropped);
6899
	storage->tx_dropped += atomic_long_read(&dev->tx_dropped);
6900
	return storage;
R
Rusty Russell 已提交
6901
}
6902
EXPORT_SYMBOL(dev_get_stats);
R
Rusty Russell 已提交
6903

6904
struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
6905
{
6906
	struct netdev_queue *queue = dev_ingress_queue(dev);
6907

6908 6909 6910 6911 6912 6913 6914
#ifdef CONFIG_NET_CLS_ACT
	if (queue)
		return queue;
	queue = kzalloc(sizeof(*queue), GFP_KERNEL);
	if (!queue)
		return NULL;
	netdev_init_one_queue(dev, queue, NULL);
E
Eric Dumazet 已提交
6915
	RCU_INIT_POINTER(queue->qdisc, &noop_qdisc);
6916 6917 6918 6919
	queue->qdisc_sleeping = &noop_qdisc;
	rcu_assign_pointer(dev->ingress_queue, queue);
#endif
	return queue;
6920 6921
}

6922 6923
static const struct ethtool_ops default_ethtool_ops;

6924 6925 6926 6927 6928 6929 6930 6931
void netdev_set_default_ethtool_ops(struct net_device *dev,
				    const struct ethtool_ops *ops)
{
	if (dev->ethtool_ops == &default_ethtool_ops)
		dev->ethtool_ops = ops;
}
EXPORT_SYMBOL_GPL(netdev_set_default_ethtool_ops);

6932 6933 6934 6935
void netdev_freemem(struct net_device *dev)
{
	char *addr = (char *)dev - dev->padded;

W
WANG Cong 已提交
6936
	kvfree(addr);
6937 6938
}

L
Linus Torvalds 已提交
6939
/**
T
Tom Herbert 已提交
6940
 *	alloc_netdev_mqs - allocate network device
6941 6942 6943 6944 6945 6946
 *	@sizeof_priv:		size of private data to allocate space for
 *	@name:			device name format string
 *	@name_assign_type: 	origin of device name
 *	@setup:			callback to initialize device
 *	@txqs:			the number of TX subqueues to allocate
 *	@rxqs:			the number of RX subqueues to allocate
L
Linus Torvalds 已提交
6947 6948
 *
 *	Allocates a struct net_device with private data area for driver use
L
Li Zhong 已提交
6949
 *	and performs basic initialization.  Also allocates subqueue structs
T
Tom Herbert 已提交
6950
 *	for each queue on the device.
L
Linus Torvalds 已提交
6951
 */
T
Tom Herbert 已提交
6952
struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
6953
		unsigned char name_assign_type,
T
Tom Herbert 已提交
6954 6955
		void (*setup)(struct net_device *),
		unsigned int txqs, unsigned int rxqs)
L
Linus Torvalds 已提交
6956 6957
{
	struct net_device *dev;
6958
	size_t alloc_size;
6959
	struct net_device *p;
L
Linus Torvalds 已提交
6960

6961 6962
	BUG_ON(strlen(name) >= sizeof(dev->name));

T
Tom Herbert 已提交
6963
	if (txqs < 1) {
6964
		pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
6965 6966 6967
		return NULL;
	}

6968
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
6969
	if (rxqs < 1) {
6970
		pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
T
Tom Herbert 已提交
6971 6972 6973 6974
		return NULL;
	}
#endif

6975
	alloc_size = sizeof(struct net_device);
6976 6977
	if (sizeof_priv) {
		/* ensure 32-byte alignment of private area */
6978
		alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
6979 6980 6981
		alloc_size += sizeof_priv;
	}
	/* ensure 32-byte alignment of whole construct */
6982
	alloc_size += NETDEV_ALIGN - 1;
L
Linus Torvalds 已提交
6983

6984 6985 6986
	p = kzalloc(alloc_size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!p)
		p = vzalloc(alloc_size);
6987
	if (!p)
L
Linus Torvalds 已提交
6988 6989
		return NULL;

6990
	dev = PTR_ALIGN(p, NETDEV_ALIGN);
L
Linus Torvalds 已提交
6991
	dev->padded = (char *)dev - (char *)p;
6992

E
Eric Dumazet 已提交
6993 6994
	dev->pcpu_refcnt = alloc_percpu(int);
	if (!dev->pcpu_refcnt)
6995
		goto free_dev;
6996 6997

	if (dev_addr_init(dev))
E
Eric Dumazet 已提交
6998
		goto free_pcpu;
6999

7000
	dev_mc_init(dev);
7001
	dev_uc_init(dev);
J
Jiri Pirko 已提交
7002

7003
	dev_net_set(dev, &init_net);
L
Linus Torvalds 已提交
7004

7005
	dev->gso_max_size = GSO_MAX_SIZE;
7006
	dev->gso_max_segs = GSO_MAX_SEGS;
7007
	dev->gso_min_segs = 0;
7008 7009 7010

	INIT_LIST_HEAD(&dev->napi_list);
	INIT_LIST_HEAD(&dev->unreg_list);
7011
	INIT_LIST_HEAD(&dev->close_list);
7012
	INIT_LIST_HEAD(&dev->link_watch_list);
7013 7014 7015 7016
	INIT_LIST_HEAD(&dev->adj_list.upper);
	INIT_LIST_HEAD(&dev->adj_list.lower);
	INIT_LIST_HEAD(&dev->all_adj_list.upper);
	INIT_LIST_HEAD(&dev->all_adj_list.lower);
7017 7018
	INIT_LIST_HEAD(&dev->ptype_all);
	INIT_LIST_HEAD(&dev->ptype_specific);
7019
	dev->priv_flags = IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM;
7020 7021
	setup(dev);

7022
	if (!dev->tx_queue_len)
7023
		dev->priv_flags |= IFF_NO_QUEUE;
7024

T
Tom Herbert 已提交
7025 7026
	dev->num_tx_queues = txqs;
	dev->real_num_tx_queues = txqs;
7027
	if (netif_alloc_netdev_queues(dev))
7028
		goto free_all;
7029

7030
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
7031 7032
	dev->num_rx_queues = rxqs;
	dev->real_num_rx_queues = rxqs;
T
Tom Herbert 已提交
7033
	if (netif_alloc_rx_queues(dev))
7034
		goto free_all;
E
Eric Dumazet 已提交
7035
#endif
T
Tom Herbert 已提交
7036

L
Linus Torvalds 已提交
7037
	strcpy(dev->name, name);
7038
	dev->name_assign_type = name_assign_type;
7039
	dev->group = INIT_NETDEV_GROUP;
7040 7041
	if (!dev->ethtool_ops)
		dev->ethtool_ops = &default_ethtool_ops;
7042 7043 7044

	nf_hook_ingress_init(dev);

L
Linus Torvalds 已提交
7045
	return dev;
7046

7047 7048 7049 7050
free_all:
	free_netdev(dev);
	return NULL;

E
Eric Dumazet 已提交
7051 7052
free_pcpu:
	free_percpu(dev->pcpu_refcnt);
7053 7054
free_dev:
	netdev_freemem(dev);
7055
	return NULL;
L
Linus Torvalds 已提交
7056
}
T
Tom Herbert 已提交
7057
EXPORT_SYMBOL(alloc_netdev_mqs);
L
Linus Torvalds 已提交
7058 7059 7060 7061 7062

/**
 *	free_netdev - free network device
 *	@dev: device
 *
7063 7064
 *	This function does the last stage of destroying an allocated device
 * 	interface. The reference to the device object is released.
L
Linus Torvalds 已提交
7065 7066 7067 7068
 *	If this is the last reference then it will be freed.
 */
void free_netdev(struct net_device *dev)
{
7069 7070
	struct napi_struct *p, *n;

7071
	netif_free_tx_queues(dev);
7072
#ifdef CONFIG_SYSFS
7073
	kvfree(dev->_rx);
T
Tom Herbert 已提交
7074
#endif
7075

7076
	kfree(rcu_dereference_protected(dev->ingress_queue, 1));
7077

7078 7079 7080
	/* Flush device addresses */
	dev_addr_flush(dev);

7081 7082 7083
	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
		netif_napi_del(p);

E
Eric Dumazet 已提交
7084 7085 7086
	free_percpu(dev->pcpu_refcnt);
	dev->pcpu_refcnt = NULL;

S
Stephen Hemminger 已提交
7087
	/*  Compatibility with error handling in drivers */
L
Linus Torvalds 已提交
7088
	if (dev->reg_state == NETREG_UNINITIALIZED) {
7089
		netdev_freemem(dev);
L
Linus Torvalds 已提交
7090 7091 7092 7093 7094 7095
		return;
	}

	BUG_ON(dev->reg_state != NETREG_UNREGISTERED);
	dev->reg_state = NETREG_RELEASED;

7096 7097
	/* will free via device release */
	put_device(&dev->dev);
L
Linus Torvalds 已提交
7098
}
E
Eric Dumazet 已提交
7099
EXPORT_SYMBOL(free_netdev);
7100

7101 7102 7103 7104 7105 7106
/**
 *	synchronize_net -  Synchronize with packet receive processing
 *
 *	Wait for packets currently being received to be done.
 *	Does not block later packets from starting.
 */
7107
void synchronize_net(void)
L
Linus Torvalds 已提交
7108 7109
{
	might_sleep();
7110 7111 7112 7113
	if (rtnl_is_locked())
		synchronize_rcu_expedited();
	else
		synchronize_rcu();
L
Linus Torvalds 已提交
7114
}
E
Eric Dumazet 已提交
7115
EXPORT_SYMBOL(synchronize_net);
L
Linus Torvalds 已提交
7116 7117

/**
7118
 *	unregister_netdevice_queue - remove device from the kernel
L
Linus Torvalds 已提交
7119
 *	@dev: device
7120
 *	@head: list
7121
 *
L
Linus Torvalds 已提交
7122
 *	This function shuts down a device interface and removes it
7123
 *	from the kernel tables.
7124
 *	If head not NULL, device is queued to be unregistered later.
L
Linus Torvalds 已提交
7125 7126 7127 7128 7129
 *
 *	Callers must hold the rtnl semaphore.  You may want
 *	unregister_netdev() instead of this.
 */

7130
void unregister_netdevice_queue(struct net_device *dev, struct list_head *head)
L
Linus Torvalds 已提交
7131
{
7132 7133
	ASSERT_RTNL();

7134
	if (head) {
7135
		list_move_tail(&dev->unreg_list, head);
7136 7137 7138 7139 7140
	} else {
		rollback_registered(dev);
		/* Finish processing unregister after unlock */
		net_set_todo(dev);
	}
L
Linus Torvalds 已提交
7141
}
7142
EXPORT_SYMBOL(unregister_netdevice_queue);
L
Linus Torvalds 已提交
7143

7144 7145 7146
/**
 *	unregister_netdevice_many - unregister many devices
 *	@head: list of devices
7147 7148 7149
 *
 *  Note: As most callers use a stack allocated list_head,
 *  we force a list_del() to make sure stack wont be corrupted later.
7150 7151 7152 7153 7154 7155 7156 7157 7158
 */
void unregister_netdevice_many(struct list_head *head)
{
	struct net_device *dev;

	if (!list_empty(head)) {
		rollback_registered_many(head);
		list_for_each_entry(dev, head, unreg_list)
			net_set_todo(dev);
7159
		list_del(head);
7160 7161
	}
}
7162
EXPORT_SYMBOL(unregister_netdevice_many);
7163

L
Linus Torvalds 已提交
7164 7165 7166 7167 7168
/**
 *	unregister_netdev - remove device from the kernel
 *	@dev: device
 *
 *	This function shuts down a device interface and removes it
7169
 *	from the kernel tables.
L
Linus Torvalds 已提交
7170 7171 7172 7173 7174 7175 7176 7177 7178 7179 7180 7181 7182
 *
 *	This is just a wrapper for unregister_netdevice that takes
 *	the rtnl semaphore.  In general you want to use this and not
 *	unregister_netdevice.
 */
void unregister_netdev(struct net_device *dev)
{
	rtnl_lock();
	unregister_netdevice(dev);
	rtnl_unlock();
}
EXPORT_SYMBOL(unregister_netdev);

7183 7184 7185 7186 7187 7188 7189 7190 7191 7192 7193 7194 7195 7196 7197 7198 7199 7200 7201 7202 7203 7204 7205 7206 7207 7208 7209 7210 7211 7212 7213
/**
 *	dev_change_net_namespace - move device to different nethost namespace
 *	@dev: device
 *	@net: network namespace
 *	@pat: If not NULL name pattern to try if the current device name
 *	      is already taken in the destination network namespace.
 *
 *	This function shuts down a device interface and moves it
 *	to a new network namespace. On success 0 is returned, on
 *	a failure a netagive errno code is returned.
 *
 *	Callers must hold the rtnl semaphore.
 */

int dev_change_net_namespace(struct net_device *dev, struct net *net, const char *pat)
{
	int err;

	ASSERT_RTNL();

	/* Don't allow namespace local devices to be moved. */
	err = -EINVAL;
	if (dev->features & NETIF_F_NETNS_LOCAL)
		goto out;

	/* Ensure the device has been registrered */
	if (dev->reg_state != NETREG_REGISTERED)
		goto out;

	/* Get out if there is nothing todo */
	err = 0;
7214
	if (net_eq(dev_net(dev), net))
7215 7216 7217 7218 7219 7220
		goto out;

	/* Pick the destination device name, and ensure
	 * we can use it in the destination network namespace.
	 */
	err = -EEXIST;
7221
	if (__dev_get_by_name(net, dev->name)) {
7222 7223 7224
		/* We get here if we can't use the current device name */
		if (!pat)
			goto out;
7225
		if (dev_get_valid_name(net, dev, pat) < 0)
7226 7227 7228 7229 7230 7231 7232 7233
			goto out;
	}

	/*
	 * And now a mini version of register_netdevice unregister_netdevice.
	 */

	/* If device is running close it first. */
7234
	dev_close(dev);
7235 7236 7237 7238 7239 7240 7241 7242 7243 7244 7245 7246

	/* And unlink it from device chain */
	err = -ENODEV;
	unlist_netdevice(dev);

	synchronize_net();

	/* Shutdown queueing discipline. */
	dev_shutdown(dev);

	/* Notify protocols, that we are about to destroy
	   this device. They should clean all the things.
7247 7248 7249 7250

	   Note that dev->reg_state stays at NETREG_REGISTERED.
	   This is wanted because this way 8021q and macvlan know
	   the device is just moving and can keep their slaves up.
7251 7252
	*/
	call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
7253 7254
	rcu_barrier();
	call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
7255
	rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);
7256 7257 7258 7259

	/*
	 *	Flush the unicast and multicast chains
	 */
7260
	dev_uc_flush(dev);
7261
	dev_mc_flush(dev);
7262

7263 7264
	/* Send a netdev-removed uevent to the old namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_REMOVE);
7265
	netdev_adjacent_del_links(dev);
7266

7267
	/* Actually switch the network namespace */
7268
	dev_net_set(dev, net);
7269 7270

	/* If there is an ifindex conflict assign a new one */
7271
	if (__dev_get_by_index(net, dev->ifindex))
7272 7273
		dev->ifindex = dev_new_index(net);

7274 7275
	/* Send a netdev-add uevent to the new namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_ADD);
7276
	netdev_adjacent_add_links(dev);
7277

7278
	/* Fixup kobjects */
7279
	err = device_rename(&dev->dev, dev->name);
7280
	WARN_ON(err);
7281 7282 7283 7284 7285 7286 7287

	/* Add the device back in the hashes */
	list_netdevice(dev);

	/* Notify protocols, that a new device appeared. */
	call_netdevice_notifiers(NETDEV_REGISTER, dev);

7288 7289 7290 7291
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
7292
	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
7293

7294 7295 7296 7297 7298
	synchronize_net();
	err = 0;
out:
	return err;
}
7299
EXPORT_SYMBOL_GPL(dev_change_net_namespace);
7300

L
Linus Torvalds 已提交
7301 7302 7303 7304 7305 7306 7307 7308 7309
static int dev_cpu_callback(struct notifier_block *nfb,
			    unsigned long action,
			    void *ocpu)
{
	struct sk_buff **list_skb;
	struct sk_buff *skb;
	unsigned int cpu, oldcpu = (unsigned long)ocpu;
	struct softnet_data *sd, *oldsd;

7310
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
L
Linus Torvalds 已提交
7311 7312 7313 7314 7315 7316 7317 7318 7319 7320 7321 7322 7323 7324 7325 7326
		return NOTIFY_OK;

	local_irq_disable();
	cpu = smp_processor_id();
	sd = &per_cpu(softnet_data, cpu);
	oldsd = &per_cpu(softnet_data, oldcpu);

	/* Find end of our completion_queue. */
	list_skb = &sd->completion_queue;
	while (*list_skb)
		list_skb = &(*list_skb)->next;
	/* Append completion queue from offline CPU. */
	*list_skb = oldsd->completion_queue;
	oldsd->completion_queue = NULL;

	/* Append output queue from offline CPU. */
7327 7328 7329 7330 7331 7332
	if (oldsd->output_queue) {
		*sd->output_queue_tailp = oldsd->output_queue;
		sd->output_queue_tailp = oldsd->output_queue_tailp;
		oldsd->output_queue = NULL;
		oldsd->output_queue_tailp = &oldsd->output_queue;
	}
E
Eric Dumazet 已提交
7333 7334 7335 7336 7337 7338 7339 7340 7341 7342 7343 7344 7345 7346
	/* Append NAPI poll list from offline CPU, with one exception :
	 * process_backlog() must be called by cpu owning percpu backlog.
	 * We properly handle process_queue & input_pkt_queue later.
	 */
	while (!list_empty(&oldsd->poll_list)) {
		struct napi_struct *napi = list_first_entry(&oldsd->poll_list,
							    struct napi_struct,
							    poll_list);

		list_del_init(&napi->poll_list);
		if (napi->poll == process_backlog)
			napi->state = 0;
		else
			____napi_schedule(sd, napi);
7347
	}
L
Linus Torvalds 已提交
7348 7349 7350 7351 7352

	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_enable();

	/* Process offline CPU's input_pkt_queue */
7353
	while ((skb = __skb_dequeue(&oldsd->process_queue))) {
7354
		netif_rx_ni(skb);
7355
		input_queue_head_incr(oldsd);
T
Tom Herbert 已提交
7356
	}
E
Eric Dumazet 已提交
7357
	while ((skb = skb_dequeue(&oldsd->input_pkt_queue))) {
7358
		netif_rx_ni(skb);
7359 7360
		input_queue_head_incr(oldsd);
	}
L
Linus Torvalds 已提交
7361 7362 7363 7364 7365

	return NOTIFY_OK;
}


7366
/**
7367 7368 7369 7370
 *	netdev_increment_features - increment feature set by one
 *	@all: current feature set
 *	@one: new feature set
 *	@mask: mask feature set
7371 7372
 *
 *	Computes a new feature set after adding a device with feature set
7373 7374
 *	@one to the master device with current feature set @all.  Will not
 *	enable anything that is off in @mask. Returns the new feature set.
7375
 */
7376 7377
netdev_features_t netdev_increment_features(netdev_features_t all,
	netdev_features_t one, netdev_features_t mask)
7378
{
7379 7380 7381
	if (mask & NETIF_F_GEN_CSUM)
		mask |= NETIF_F_ALL_CSUM;
	mask |= NETIF_F_VLAN_CHALLENGED;
7382

7383 7384
	all |= one & (NETIF_F_ONE_FOR_ALL|NETIF_F_ALL_CSUM) & mask;
	all &= one | ~NETIF_F_ALL_FOR_ALL;
7385

7386 7387 7388
	/* If one device supports hw checksumming, set for all. */
	if (all & NETIF_F_GEN_CSUM)
		all &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
7389 7390 7391

	return all;
}
7392
EXPORT_SYMBOL(netdev_increment_features);
7393

7394
static struct hlist_head * __net_init netdev_create_hash(void)
7395 7396 7397 7398 7399 7400 7401 7402 7403 7404 7405 7406
{
	int i;
	struct hlist_head *hash;

	hash = kmalloc(sizeof(*hash) * NETDEV_HASHENTRIES, GFP_KERNEL);
	if (hash != NULL)
		for (i = 0; i < NETDEV_HASHENTRIES; i++)
			INIT_HLIST_HEAD(&hash[i]);

	return hash;
}

7407
/* Initialize per network namespace state */
7408
static int __net_init netdev_init(struct net *net)
7409
{
7410 7411
	if (net != &init_net)
		INIT_LIST_HEAD(&net->dev_base_head);
7412

7413 7414 7415
	net->dev_name_head = netdev_create_hash();
	if (net->dev_name_head == NULL)
		goto err_name;
7416

7417 7418 7419
	net->dev_index_head = netdev_create_hash();
	if (net->dev_index_head == NULL)
		goto err_idx;
7420 7421

	return 0;
7422 7423 7424 7425 7426

err_idx:
	kfree(net->dev_name_head);
err_name:
	return -ENOMEM;
7427 7428
}

7429 7430 7431 7432 7433 7434
/**
 *	netdev_drivername - network driver for the device
 *	@dev: network device
 *
 *	Determine network driver for device.
 */
7435
const char *netdev_drivername(const struct net_device *dev)
7436
{
7437 7438
	const struct device_driver *driver;
	const struct device *parent;
7439
	const char *empty = "";
7440 7441 7442

	parent = dev->dev.parent;
	if (!parent)
7443
		return empty;
7444 7445 7446

	driver = parent->driver;
	if (driver && driver->name)
7447 7448
		return driver->name;
	return empty;
7449 7450
}

7451 7452
static void __netdev_printk(const char *level, const struct net_device *dev,
			    struct va_format *vaf)
7453
{
7454
	if (dev && dev->dev.parent) {
7455 7456 7457 7458 7459 7460 7461
		dev_printk_emit(level[1] - '0',
				dev->dev.parent,
				"%s %s %s%s: %pV",
				dev_driver_string(dev->dev.parent),
				dev_name(dev->dev.parent),
				netdev_name(dev), netdev_reg_state(dev),
				vaf);
7462
	} else if (dev) {
7463 7464
		printk("%s%s%s: %pV",
		       level, netdev_name(dev), netdev_reg_state(dev), vaf);
7465
	} else {
7466
		printk("%s(NULL net_device): %pV", level, vaf);
7467
	}
7468 7469
}

7470 7471
void netdev_printk(const char *level, const struct net_device *dev,
		   const char *format, ...)
7472 7473 7474 7475 7476 7477 7478 7479 7480
{
	struct va_format vaf;
	va_list args;

	va_start(args, format);

	vaf.fmt = format;
	vaf.va = &args;

7481
	__netdev_printk(level, dev, &vaf);
7482

7483 7484 7485 7486 7487
	va_end(args);
}
EXPORT_SYMBOL(netdev_printk);

#define define_netdev_printk_level(func, level)			\
7488
void func(const struct net_device *dev, const char *fmt, ...)	\
7489 7490 7491 7492 7493 7494 7495 7496 7497
{								\
	struct va_format vaf;					\
	va_list args;						\
								\
	va_start(args, fmt);					\
								\
	vaf.fmt = fmt;						\
	vaf.va = &args;						\
								\
7498
	__netdev_printk(level, dev, &vaf);			\
7499
								\
7500 7501 7502 7503 7504 7505 7506 7507 7508 7509 7510 7511
	va_end(args);						\
}								\
EXPORT_SYMBOL(func);

define_netdev_printk_level(netdev_emerg, KERN_EMERG);
define_netdev_printk_level(netdev_alert, KERN_ALERT);
define_netdev_printk_level(netdev_crit, KERN_CRIT);
define_netdev_printk_level(netdev_err, KERN_ERR);
define_netdev_printk_level(netdev_warn, KERN_WARNING);
define_netdev_printk_level(netdev_notice, KERN_NOTICE);
define_netdev_printk_level(netdev_info, KERN_INFO);

7512
static void __net_exit netdev_exit(struct net *net)
7513 7514 7515 7516 7517
{
	kfree(net->dev_name_head);
	kfree(net->dev_index_head);
}

7518
static struct pernet_operations __net_initdata netdev_net_ops = {
7519 7520 7521 7522
	.init = netdev_init,
	.exit = netdev_exit,
};

7523
static void __net_exit default_device_exit(struct net *net)
7524
{
7525
	struct net_device *dev, *aux;
7526
	/*
7527
	 * Push all migratable network devices back to the
7528 7529 7530
	 * initial network namespace
	 */
	rtnl_lock();
7531
	for_each_netdev_safe(net, dev, aux) {
7532
		int err;
7533
		char fb_name[IFNAMSIZ];
7534 7535 7536 7537 7538

		/* Ignore unmoveable devices (i.e. loopback) */
		if (dev->features & NETIF_F_NETNS_LOCAL)
			continue;

7539 7540 7541
		/* Leave virtual devices for the generic cleanup */
		if (dev->rtnl_link_ops)
			continue;
7542

L
Lucas De Marchi 已提交
7543
		/* Push remaining network devices to init_net */
7544 7545
		snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex);
		err = dev_change_net_namespace(dev, &init_net, fb_name);
7546
		if (err) {
7547 7548
			pr_emerg("%s: failed to move %s to init_net: %d\n",
				 __func__, dev->name, err);
7549
			BUG();
7550 7551 7552 7553 7554
		}
	}
	rtnl_unlock();
}

7555 7556 7557 7558 7559 7560 7561
static void __net_exit rtnl_lock_unregistering(struct list_head *net_list)
{
	/* Return with the rtnl_lock held when there are no network
	 * devices unregistering in any network namespace in net_list.
	 */
	struct net *net;
	bool unregistering;
7562
	DEFINE_WAIT_FUNC(wait, woken_wake_function);
7563

7564
	add_wait_queue(&netdev_unregistering_wq, &wait);
7565 7566 7567 7568 7569 7570 7571 7572 7573 7574 7575 7576
	for (;;) {
		unregistering = false;
		rtnl_lock();
		list_for_each_entry(net, net_list, exit_list) {
			if (net->dev_unreg_count > 0) {
				unregistering = true;
				break;
			}
		}
		if (!unregistering)
			break;
		__rtnl_unlock();
7577 7578

		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
7579
	}
7580
	remove_wait_queue(&netdev_unregistering_wq, &wait);
7581 7582
}

7583 7584 7585
static void __net_exit default_device_exit_batch(struct list_head *net_list)
{
	/* At exit all network devices most be removed from a network
7586
	 * namespace.  Do this in the reverse order of registration.
7587 7588 7589 7590 7591 7592 7593
	 * Do this across as many network namespaces as possible to
	 * improve batching efficiency.
	 */
	struct net_device *dev;
	struct net *net;
	LIST_HEAD(dev_kill_list);

7594 7595 7596 7597 7598 7599 7600 7601 7602 7603 7604 7605
	/* To prevent network device cleanup code from dereferencing
	 * loopback devices or network devices that have been freed
	 * wait here for all pending unregistrations to complete,
	 * before unregistring the loopback device and allowing the
	 * network namespace be freed.
	 *
	 * The netdev todo list containing all network devices
	 * unregistrations that happen in default_device_exit_batch
	 * will run in the rtnl_unlock() at the end of
	 * default_device_exit_batch.
	 */
	rtnl_lock_unregistering(net_list);
7606 7607
	list_for_each_entry(net, net_list, exit_list) {
		for_each_netdev_reverse(net, dev) {
7608
			if (dev->rtnl_link_ops && dev->rtnl_link_ops->dellink)
7609 7610 7611 7612 7613 7614 7615 7616 7617
				dev->rtnl_link_ops->dellink(dev, &dev_kill_list);
			else
				unregister_netdevice_queue(dev, &dev_kill_list);
		}
	}
	unregister_netdevice_many(&dev_kill_list);
	rtnl_unlock();
}

7618
static struct pernet_operations __net_initdata default_device_ops = {
7619
	.exit = default_device_exit,
7620
	.exit_batch = default_device_exit_batch,
7621 7622
};

L
Linus Torvalds 已提交
7623 7624 7625 7626 7627 7628 7629 7630 7631 7632 7633 7634 7635 7636 7637 7638 7639 7640 7641 7642
/*
 *	Initialize the DEV module. At boot time this walks the device list and
 *	unhooks any devices that fail to initialise (normally hardware not
 *	present) and leaves us with a valid list of present and active devices.
 *
 */

/*
 *       This is called single threaded during boot, so no need
 *       to take the rtnl semaphore.
 */
static int __init net_dev_init(void)
{
	int i, rc = -ENOMEM;

	BUG_ON(!dev_boot_phase);

	if (dev_proc_init())
		goto out;

7643
	if (netdev_kobject_init())
L
Linus Torvalds 已提交
7644 7645 7646
		goto out;

	INIT_LIST_HEAD(&ptype_all);
7647
	for (i = 0; i < PTYPE_HASH_SIZE; i++)
L
Linus Torvalds 已提交
7648 7649
		INIT_LIST_HEAD(&ptype_base[i]);

7650 7651
	INIT_LIST_HEAD(&offload_base);

7652 7653
	if (register_pernet_subsys(&netdev_net_ops))
		goto out;
L
Linus Torvalds 已提交
7654 7655 7656 7657 7658

	/*
	 *	Initialise the packet receive queues.
	 */

7659
	for_each_possible_cpu(i) {
E
Eric Dumazet 已提交
7660
		struct softnet_data *sd = &per_cpu(softnet_data, i);
L
Linus Torvalds 已提交
7661

E
Eric Dumazet 已提交
7662
		skb_queue_head_init(&sd->input_pkt_queue);
7663
		skb_queue_head_init(&sd->process_queue);
E
Eric Dumazet 已提交
7664
		INIT_LIST_HEAD(&sd->poll_list);
7665
		sd->output_queue_tailp = &sd->output_queue;
E
Eric Dumazet 已提交
7666
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
7667 7668 7669
		sd->csd.func = rps_trigger_softirq;
		sd->csd.info = sd;
		sd->cpu = i;
7670
#endif
T
Tom Herbert 已提交
7671

E
Eric Dumazet 已提交
7672 7673
		sd->backlog.poll = process_backlog;
		sd->backlog.weight = weight_p;
L
Linus Torvalds 已提交
7674 7675 7676 7677
	}

	dev_boot_phase = 0;

7678 7679 7680 7681 7682 7683 7684 7685 7686 7687 7688 7689 7690 7691 7692
	/* The loopback device is special if any other network devices
	 * is present in a network namespace the loopback device must
	 * be present. Since we now dynamically allocate and free the
	 * loopback device ensure this invariant is maintained by
	 * keeping the loopback device as the first device on the
	 * list of network devices.  Ensuring the loopback devices
	 * is the first device that appears and the last network device
	 * that disappears.
	 */
	if (register_pernet_device(&loopback_net_ops))
		goto out;

	if (register_pernet_device(&default_device_ops))
		goto out;

7693 7694
	open_softirq(NET_TX_SOFTIRQ, net_tx_action);
	open_softirq(NET_RX_SOFTIRQ, net_rx_action);
L
Linus Torvalds 已提交
7695 7696

	hotcpu_notifier(dev_cpu_callback, 0);
T
Thomas Graf 已提交
7697
	dst_subsys_init();
L
Linus Torvalds 已提交
7698 7699 7700 7701 7702 7703
	rc = 0;
out:
	return rc;
}

subsys_initcall(net_dev_init);