dev.c 184.1 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>
L
Linus Torvalds 已提交
138

139 140
#include "net-sysfs.h"

141 142 143
/* Instead of increasing this, you should create a hash table. */
#define MAX_GRO_SKBS 8

H
Herbert Xu 已提交
144 145 146
/* 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
147
static DEFINE_SPINLOCK(ptype_lock);
148
static DEFINE_SPINLOCK(offload_lock);
149 150
struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
struct list_head ptype_all __read_mostly;	/* Taps */
151
static struct list_head offload_base __read_mostly;
L
Linus Torvalds 已提交
152

153
static int netif_rx_internal(struct sk_buff *skb);
154 155 156
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info);
157

L
Linus Torvalds 已提交
158
/*
159
 * The @dev_base_head list is protected by @dev_base_lock and the rtnl
L
Linus Torvalds 已提交
160 161
 * semaphore.
 *
162
 * Pure readers hold dev_base_lock for reading, or rcu_read_lock()
L
Linus Torvalds 已提交
163 164
 *
 * Writers must hold the rtnl semaphore while they loop through the
165
 * dev_base_head list, and hold dev_base_lock for writing when they do the
L
Linus Torvalds 已提交
166 167 168 169 170 171 172 173 174 175 176 177 178 179
 * 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
180 181 182 183 184 185
/* 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);

186
static seqcount_t devnet_rename_seq;
187

188 189 190 191 192
static inline void dev_base_seq_inc(struct net *net)
{
	while (++net->dev_base_seq == 0);
}

193
static inline struct hlist_head *dev_name_hash(struct net *net, const char *name)
L
Linus Torvalds 已提交
194
{
195 196
	unsigned int hash = full_name_hash(name, strnlen(name, IFNAMSIZ));

197
	return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)];
L
Linus Torvalds 已提交
198 199
}

200
static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex)
L
Linus Torvalds 已提交
201
{
202
	return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)];
L
Linus Torvalds 已提交
203 204
}

E
Eric Dumazet 已提交
205
static inline void rps_lock(struct softnet_data *sd)
206 207
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
208
	spin_lock(&sd->input_pkt_queue.lock);
209 210 211
#endif
}

E
Eric Dumazet 已提交
212
static inline void rps_unlock(struct softnet_data *sd)
213 214
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
215
	spin_unlock(&sd->input_pkt_queue.lock);
216 217 218
#endif
}

219
/* Device list insertion */
220
static void list_netdevice(struct net_device *dev)
221
{
222
	struct net *net = dev_net(dev);
223 224 225 226

	ASSERT_RTNL();

	write_lock_bh(&dev_base_lock);
227
	list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
228
	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
229 230
	hlist_add_head_rcu(&dev->index_hlist,
			   dev_index_hash(net, dev->ifindex));
231
	write_unlock_bh(&dev_base_lock);
232 233

	dev_base_seq_inc(net);
234 235
}

236 237 238
/* Device list removal
 * caller must respect a RCU grace period before freeing/reusing dev
 */
239 240 241 242 243 244
static void unlist_netdevice(struct net_device *dev)
{
	ASSERT_RTNL();

	/* Unlink dev from the device chain */
	write_lock_bh(&dev_base_lock);
245
	list_del_rcu(&dev->dev_list);
246
	hlist_del_rcu(&dev->name_hlist);
247
	hlist_del_rcu(&dev->index_hlist);
248
	write_unlock_bh(&dev_base_lock);
249 250

	dev_base_seq_inc(dev_net(dev));
251 252
}

L
Linus Torvalds 已提交
253 254 255 256
/*
 *	Our notifier list
 */

257
static RAW_NOTIFIER_HEAD(netdev_chain);
L
Linus Torvalds 已提交
258 259 260 261 262

/*
 *	Device drivers call our routines to queue packets here. We empty the
 *	queue in the local softnet handler.
 */
263

264
DEFINE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
E
Eric Dumazet 已提交
265
EXPORT_PER_CPU_SYMBOL(softnet_data);
L
Linus Torvalds 已提交
266

267
#ifdef CONFIG_LOCKDEP
268
/*
269
 * register_netdevice() inits txq->_xmit_lock and sets lockdep class
270 271 272 273 274 275 276 277 278 279 280 281 282 283 284
 * 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,
285 286 287
	 ARPHRD_FCFABRIC, ARPHRD_IEEE80211, ARPHRD_IEEE80211_PRISM,
	 ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, ARPHRD_PHONET_PIPE,
	 ARPHRD_IEEE802154, ARPHRD_VOID, ARPHRD_NONE};
288

289
static const char *const netdev_lock_name[] =
290 291 292 293 294 295 296 297 298 299 300 301
	{"_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",
302 303 304
	 "_xmit_FCFABRIC", "_xmit_IEEE80211", "_xmit_IEEE80211_PRISM",
	 "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", "_xmit_PHONET_PIPE",
	 "_xmit_IEEE802154", "_xmit_VOID", "_xmit_NONE"};
305 306

static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)];
307
static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)];
308 309 310 311 312 313 314 315 316 317 318 319

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;
}

320 321
static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
						 unsigned short dev_type)
322 323 324 325 326 327 328
{
	int i;

	i = netdev_lock_pos(dev_type);
	lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i],
				   netdev_lock_name[i]);
}
329 330 331 332 333 334 335 336 337 338

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]);
}
339
#else
340 341 342 343 344
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)
345 346 347
{
}
#endif
L
Linus Torvalds 已提交
348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370

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

		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)
 */

371 372 373 374 375 376 377 378
static inline struct list_head *ptype_head(const struct packet_type *pt)
{
	if (pt->type == htons(ETH_P_ALL))
		return &ptype_all;
	else
		return &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK];
}

L
Linus Torvalds 已提交
379 380 381 382 383 384 385 386
/**
 *	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.
 *
387
 *	This call does not sleep therefore it can not
L
Linus Torvalds 已提交
388 389 390 391 392 393
 *	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)
{
394
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
395

396 397 398
	spin_lock(&ptype_lock);
	list_add_rcu(&pt->list, head);
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
399
}
E
Eric Dumazet 已提交
400
EXPORT_SYMBOL(dev_add_pack);
L
Linus Torvalds 已提交
401 402 403 404 405 406 407 408

/**
 *	__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
409
 *	returns.
L
Linus Torvalds 已提交
410 411 412 413 414 415 416
 *
 *      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)
{
417
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
418 419
	struct packet_type *pt1;

420
	spin_lock(&ptype_lock);
L
Linus Torvalds 已提交
421 422 423 424 425 426 427 428

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

429
	pr_warn("dev_remove_pack: %p not found\n", pt);
L
Linus Torvalds 已提交
430
out:
431
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
432
}
E
Eric Dumazet 已提交
433 434
EXPORT_SYMBOL(__dev_remove_pack);

L
Linus Torvalds 已提交
435 436 437 438 439 440 441 442 443 444 445 446 447 448 449
/**
 *	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);
450

L
Linus Torvalds 已提交
451 452
	synchronize_net();
}
E
Eric Dumazet 已提交
453
EXPORT_SYMBOL(dev_remove_pack);
L
Linus Torvalds 已提交
454

455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490

/**
 *	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)
{
	struct list_head *head = &offload_base;

	spin_lock(&offload_lock);
	list_add_rcu(&po->list, head);
	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
491
static void __dev_remove_offload(struct packet_offload *po)
492 493 494 495
{
	struct list_head *head = &offload_base;
	struct packet_offload *po1;

496
	spin_lock(&offload_lock);
497 498 499 500 501 502 503 504 505 506

	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:
507
	spin_unlock(&offload_lock);
508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529
}

/**
 *	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
530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556
/******************************************************************************

		      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));
557
			strlcpy(s[i].name, name, IFNAMSIZ);
L
Linus Torvalds 已提交
558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581
			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] != ' ' &&
582
		    !strcmp(dev->name, s[i].name)) {
L
Linus Torvalds 已提交
583 584 585 586 587 588 589 590 591
			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
592
EXPORT_SYMBOL(netdev_boot_setup_check);
L
Linus Torvalds 已提交
593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616


/**
 *	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
	 */
617
	if (__dev_get_by_name(&init_net, name))
L
Linus Torvalds 已提交
618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662
		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

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

/**
 *	__dev_get_by_name	- find a device by its name
663
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
664 665 666 667 668 669 670 671 672
 *	@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.
 */

673
struct net_device *__dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
674
{
E
Eric Dumazet 已提交
675 676
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);
L
Linus Torvalds 已提交
677

678
	hlist_for_each_entry(dev, head, name_hlist)
L
Linus Torvalds 已提交
679 680
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;
E
Eric Dumazet 已提交
681

L
Linus Torvalds 已提交
682 683
	return NULL;
}
E
Eric Dumazet 已提交
684
EXPORT_SYMBOL(__dev_get_by_name);
L
Linus Torvalds 已提交
685

686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702
/**
 *	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);

703
	hlist_for_each_entry_rcu(dev, head, name_hlist)
704 705 706 707 708 709 710
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_name_rcu);

L
Linus Torvalds 已提交
711 712
/**
 *	dev_get_by_name		- find a device by its name
713
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
714 715 716 717 718 719 720 721 722
 *	@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.
 */

723
struct net_device *dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
724 725 726
{
	struct net_device *dev;

727 728
	rcu_read_lock();
	dev = dev_get_by_name_rcu(net, name);
L
Linus Torvalds 已提交
729 730
	if (dev)
		dev_hold(dev);
731
	rcu_read_unlock();
L
Linus Torvalds 已提交
732 733
	return dev;
}
E
Eric Dumazet 已提交
734
EXPORT_SYMBOL(dev_get_by_name);
L
Linus Torvalds 已提交
735 736 737

/**
 *	__dev_get_by_index - find a device by its ifindex
738
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
739 740 741 742 743 744 745 746 747
 *	@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.
 */

748
struct net_device *__dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
749
{
E
Eric Dumazet 已提交
750 751
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);
L
Linus Torvalds 已提交
752

753
	hlist_for_each_entry(dev, head, index_hlist)
L
Linus Torvalds 已提交
754 755
		if (dev->ifindex == ifindex)
			return dev;
E
Eric Dumazet 已提交
756

L
Linus Torvalds 已提交
757 758
	return NULL;
}
E
Eric Dumazet 已提交
759
EXPORT_SYMBOL(__dev_get_by_index);
L
Linus Torvalds 已提交
760

761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776
/**
 *	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);

777
	hlist_for_each_entry_rcu(dev, head, index_hlist)
778 779 780 781 782 783 784
		if (dev->ifindex == ifindex)
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_index_rcu);

L
Linus Torvalds 已提交
785 786 787

/**
 *	dev_get_by_index - find a device by its ifindex
788
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
789 790 791 792 793 794 795 796
 *	@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.
 */

797
struct net_device *dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
798 799 800
{
	struct net_device *dev;

801 802
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, ifindex);
L
Linus Torvalds 已提交
803 804
	if (dev)
		dev_hold(dev);
805
	rcu_read_unlock();
L
Linus Torvalds 已提交
806 807
	return dev;
}
E
Eric Dumazet 已提交
808
EXPORT_SYMBOL(dev_get_by_index);
L
Linus Torvalds 已提交
809

810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843
/**
 *	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
844
/**
845
 *	dev_getbyhwaddr_rcu - find a device by its hardware address
846
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
847 848 849 850
 *	@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
851 852
 *	is not found or a pointer to the device.
 *	The caller must hold RCU or RTNL.
853
 *	The returned device has not had its ref count increased
L
Linus Torvalds 已提交
854 855 856 857
 *	and the caller must therefore be careful about locking
 *
 */

858 859
struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
				       const char *ha)
L
Linus Torvalds 已提交
860 861 862
{
	struct net_device *dev;

863
	for_each_netdev_rcu(net, dev)
L
Linus Torvalds 已提交
864 865
		if (dev->type == type &&
		    !memcmp(dev->dev_addr, ha, dev->addr_len))
866 867 868
			return dev;

	return NULL;
L
Linus Torvalds 已提交
869
}
870
EXPORT_SYMBOL(dev_getbyhwaddr_rcu);
871

872
struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type)
L
Linus Torvalds 已提交
873 874 875
{
	struct net_device *dev;

876
	ASSERT_RTNL();
877
	for_each_netdev(net, dev)
878
		if (dev->type == type)
879 880 881
			return dev;

	return NULL;
882 883 884
}
EXPORT_SYMBOL(__dev_getfirstbyhwtype);

885
struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type)
886
{
887
	struct net_device *dev, *ret = NULL;
888

889 890 891 892 893 894 895 896 897
	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
898 899 900 901
}
EXPORT_SYMBOL(dev_getfirstbyhwtype);

/**
902
 *	__dev_get_by_flags - find any device with given flags
903
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
904 905 906 907
 *	@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
908
 *	is not found or a pointer to the device. Must be called inside
909
 *	rtnl_lock(), and result refcount is unchanged.
L
Linus Torvalds 已提交
910 911
 */

912 913
struct net_device *__dev_get_by_flags(struct net *net, unsigned short if_flags,
				      unsigned short mask)
L
Linus Torvalds 已提交
914
{
915
	struct net_device *dev, *ret;
L
Linus Torvalds 已提交
916

917 918
	ASSERT_RTNL();

919
	ret = NULL;
920
	for_each_netdev(net, dev) {
L
Linus Torvalds 已提交
921
		if (((dev->flags ^ if_flags) & mask) == 0) {
922
			ret = dev;
L
Linus Torvalds 已提交
923 924 925
			break;
		}
	}
926
	return ret;
L
Linus Torvalds 已提交
927
}
928
EXPORT_SYMBOL(__dev_get_by_flags);
L
Linus Torvalds 已提交
929 930 931 932 933 934

/**
 *	dev_valid_name - check if name is okay for network device
 *	@name: name string
 *
 *	Network device names need to be valid file names to
935 936
 *	to allow sysfs to work.  We also disallow any kind of
 *	whitespace.
L
Linus Torvalds 已提交
937
 */
938
bool dev_valid_name(const char *name)
L
Linus Torvalds 已提交
939
{
940
	if (*name == '\0')
941
		return false;
942
	if (strlen(name) >= IFNAMSIZ)
943
		return false;
944
	if (!strcmp(name, ".") || !strcmp(name, ".."))
945
		return false;
946 947 948

	while (*name) {
		if (*name == '/' || isspace(*name))
949
			return false;
950 951
		name++;
	}
952
	return true;
L
Linus Torvalds 已提交
953
}
E
Eric Dumazet 已提交
954
EXPORT_SYMBOL(dev_valid_name);
L
Linus Torvalds 已提交
955 956

/**
957 958
 *	__dev_alloc_name - allocate a name for a device
 *	@net: network namespace to allocate the device name in
L
Linus Torvalds 已提交
959
 *	@name: name format string
960
 *	@buf:  scratch buffer and result name string
L
Linus Torvalds 已提交
961 962
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
S
Stephen Hemminger 已提交
963 964 965 966 967 968
 *	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
969 970
 */

971
static int __dev_alloc_name(struct net *net, const char *name, char *buf)
L
Linus Torvalds 已提交
972 973 974 975
{
	int i = 0;
	const char *p;
	const int max_netdevices = 8*PAGE_SIZE;
S
Stephen Hemminger 已提交
976
	unsigned long *inuse;
L
Linus Torvalds 已提交
977 978 979 980 981 982 983 984 985 986 987 988 989
	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
990
		inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC);
L
Linus Torvalds 已提交
991 992 993
		if (!inuse)
			return -ENOMEM;

994
		for_each_netdev(net, d) {
L
Linus Torvalds 已提交
995 996 997 998 999 1000
			if (!sscanf(d->name, name, &i))
				continue;
			if (i < 0 || i >= max_netdevices)
				continue;

			/*  avoid cases where sscanf is not exact inverse of printf */
1001
			snprintf(buf, IFNAMSIZ, name, i);
L
Linus Torvalds 已提交
1002 1003 1004 1005 1006 1007 1008 1009
			if (!strncmp(buf, d->name, IFNAMSIZ))
				set_bit(i, inuse);
		}

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

1010 1011
	if (buf != name)
		snprintf(buf, IFNAMSIZ, name, i);
1012
	if (!__dev_get_by_name(net, buf))
L
Linus Torvalds 已提交
1013 1014 1015 1016 1017 1018 1019 1020 1021
		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;
}

1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
/**
 *	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;

1042 1043
	BUG_ON(!dev_net(dev));
	net = dev_net(dev);
1044 1045 1046 1047 1048
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
}
E
Eric Dumazet 已提交
1049
EXPORT_SYMBOL(dev_alloc_name);
1050

1051 1052 1053
static int dev_alloc_name_ns(struct net *net,
			     struct net_device *dev,
			     const char *name)
1054
{
1055 1056
	char buf[IFNAMSIZ];
	int ret;
1057

1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068
	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);
1069

1070 1071 1072
	if (!dev_valid_name(name))
		return -EINVAL;

1073
	if (strchr(name, '%'))
1074
		return dev_alloc_name_ns(net, dev, name);
1075 1076
	else if (__dev_get_by_name(net, name))
		return -EEXIST;
1077 1078
	else if (dev->name != name)
		strlcpy(dev->name, name, IFNAMSIZ);
1079 1080 1081

	return 0;
}
L
Linus Torvalds 已提交
1082 1083 1084 1085 1086 1087 1088 1089 1090

/**
 *	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.
 */
1091
int dev_change_name(struct net_device *dev, const char *newname)
L
Linus Torvalds 已提交
1092
{
1093
	unsigned char old_assign_type;
1094
	char oldname[IFNAMSIZ];
L
Linus Torvalds 已提交
1095
	int err = 0;
1096
	int ret;
1097
	struct net *net;
L
Linus Torvalds 已提交
1098 1099

	ASSERT_RTNL();
1100
	BUG_ON(!dev_net(dev));
L
Linus Torvalds 已提交
1101

1102
	net = dev_net(dev);
L
Linus Torvalds 已提交
1103 1104 1105
	if (dev->flags & IFF_UP)
		return -EBUSY;

1106
	write_seqcount_begin(&devnet_rename_seq);
1107 1108

	if (strncmp(newname, dev->name, IFNAMSIZ) == 0) {
1109
		write_seqcount_end(&devnet_rename_seq);
1110
		return 0;
1111
	}
1112

1113 1114
	memcpy(oldname, dev->name, IFNAMSIZ);

1115
	err = dev_get_valid_name(net, dev, newname);
1116
	if (err < 0) {
1117
		write_seqcount_end(&devnet_rename_seq);
1118
		return err;
1119
	}
L
Linus Torvalds 已提交
1120

1121 1122 1123
	if (oldname[0] && !strchr(oldname, '%'))
		netdev_info(dev, "renamed from %s\n", oldname);

1124 1125 1126
	old_assign_type = dev->name_assign_type;
	dev->name_assign_type = NET_NAME_RENAMED;

1127
rollback:
1128 1129 1130
	ret = device_rename(&dev->dev, dev->name);
	if (ret) {
		memcpy(dev->name, oldname, IFNAMSIZ);
1131
		dev->name_assign_type = old_assign_type;
1132
		write_seqcount_end(&devnet_rename_seq);
1133
		return ret;
1134
	}
1135

1136
	write_seqcount_end(&devnet_rename_seq);
1137

1138 1139
	netdev_adjacent_rename_links(dev, oldname);

1140
	write_lock_bh(&dev_base_lock);
1141
	hlist_del_rcu(&dev->name_hlist);
1142 1143 1144 1145 1146 1147
	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));
1148 1149
	write_unlock_bh(&dev_base_lock);

1150
	ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
1151 1152 1153
	ret = notifier_to_errno(ret);

	if (ret) {
1154 1155
		/* err >= 0 after dev_alloc_name() or stores the first errno */
		if (err >= 0) {
1156
			err = ret;
1157
			write_seqcount_begin(&devnet_rename_seq);
1158
			memcpy(dev->name, oldname, IFNAMSIZ);
1159
			memcpy(oldname, newname, IFNAMSIZ);
1160 1161
			dev->name_assign_type = old_assign_type;
			old_assign_type = NET_NAME_RENAMED;
1162
			goto rollback;
1163
		} else {
1164
			pr_err("%s: name change rollback failed: %d\n",
1165
			       dev->name, ret);
1166 1167
		}
	}
L
Linus Torvalds 已提交
1168 1169 1170 1171

	return err;
}

1172 1173 1174 1175
/**
 *	dev_set_alias - change ifalias of a device
 *	@dev: device
 *	@alias: name up to IFALIASZ
1176
 *	@len: limit of bytes to copy from info
1177 1178 1179 1180 1181
 *
 *	Set ifalias for a device,
 */
int dev_set_alias(struct net_device *dev, const char *alias, size_t len)
{
1182 1183
	char *new_ifalias;

1184 1185 1186 1187 1188
	ASSERT_RTNL();

	if (len >= IFALIASZ)
		return -EINVAL;

1189
	if (!len) {
1190 1191
		kfree(dev->ifalias);
		dev->ifalias = NULL;
1192 1193 1194
		return 0;
	}

1195 1196
	new_ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL);
	if (!new_ifalias)
1197
		return -ENOMEM;
1198
	dev->ifalias = new_ifalias;
1199 1200 1201 1202 1203 1204

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


1205
/**
S
Stephen Hemminger 已提交
1206
 *	netdev_features_change - device changes features
1207 1208 1209 1210 1211 1212
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed features.
 */
void netdev_features_change(struct net_device *dev)
{
1213
	call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev);
1214 1215 1216
}
EXPORT_SYMBOL(netdev_features_change);

L
Linus Torvalds 已提交
1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227
/**
 *	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) {
1228 1229 1230 1231 1232
		struct netdev_notifier_change_info change_info;

		change_info.flags_changed = 0;
		call_netdevice_notifiers_info(NETDEV_CHANGE, dev,
					      &change_info.info);
1233
		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
L
Linus Torvalds 已提交
1234 1235
	}
}
E
Eric Dumazet 已提交
1236
EXPORT_SYMBOL(netdev_state_change);
L
Linus Torvalds 已提交
1237

1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248
/**
 * 	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)
1249
{
1250 1251 1252
	rtnl_lock();
	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, dev);
	rtnl_unlock();
1253
}
1254
EXPORT_SYMBOL(netdev_notify_peers);
1255

1256
static int __dev_open(struct net_device *dev)
L
Linus Torvalds 已提交
1257
{
1258
	const struct net_device_ops *ops = dev->netdev_ops;
1259
	int ret;
L
Linus Torvalds 已提交
1260

1261 1262
	ASSERT_RTNL();

L
Linus Torvalds 已提交
1263 1264 1265
	if (!netif_device_present(dev))
		return -ENODEV;

1266 1267 1268 1269
	/* 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
	 */
1270
	netpoll_poll_disable(dev);
1271

1272 1273 1274 1275 1276
	ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

L
Linus Torvalds 已提交
1277
	set_bit(__LINK_STATE_START, &dev->state);
1278

1279 1280
	if (ops->ndo_validate_addr)
		ret = ops->ndo_validate_addr(dev);
1281

1282 1283
	if (!ret && ops->ndo_open)
		ret = ops->ndo_open(dev);
L
Linus Torvalds 已提交
1284

1285
	netpoll_poll_enable(dev);
1286

1287 1288 1289
	if (ret)
		clear_bit(__LINK_STATE_START, &dev->state);
	else {
L
Linus Torvalds 已提交
1290
		dev->flags |= IFF_UP;
1291
		dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
1292
		dev_activate(dev);
1293
		add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
1294
	}
1295

L
Linus Torvalds 已提交
1296 1297 1298 1299
	return ret;
}

/**
1300 1301
 *	dev_open	- prepare an interface for use.
 *	@dev:	device to open
L
Linus Torvalds 已提交
1302
 *
1303 1304 1305 1306 1307 1308 1309
 *	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
1310
 */
1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321
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;

1322
	rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1323 1324 1325 1326 1327 1328
	call_netdevice_notifiers(NETDEV_UP, dev);

	return ret;
}
EXPORT_SYMBOL(dev_open);

1329
static int __dev_close_many(struct list_head *head)
L
Linus Torvalds 已提交
1330
{
1331
	struct net_device *dev;
1332

1333
	ASSERT_RTNL();
1334 1335
	might_sleep();

1336
	list_for_each_entry(dev, head, close_list) {
1337
		/* Temporarily disable netpoll until the interface is down */
1338
		netpoll_poll_disable(dev);
1339

1340
		call_netdevice_notifiers(NETDEV_GOING_DOWN, dev);
L
Linus Torvalds 已提交
1341

1342
		clear_bit(__LINK_STATE_START, &dev->state);
L
Linus Torvalds 已提交
1343

1344 1345 1346 1347 1348 1349
		/* 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.
		 */
1350
		smp_mb__after_atomic(); /* Commit netif_running(). */
1351
	}
L
Linus Torvalds 已提交
1352

1353
	dev_deactivate_many(head);
1354

1355
	list_for_each_entry(dev, head, close_list) {
1356
		const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
1357

1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368
		/*
		 *	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;
1369
		netpoll_poll_enable(dev);
1370 1371 1372 1373 1374 1375 1376
	}

	return 0;
}

static int __dev_close(struct net_device *dev)
{
1377
	int retval;
1378 1379
	LIST_HEAD(single);

1380
	list_add(&dev->close_list, &single);
1381 1382
	retval = __dev_close_many(&single);
	list_del(&single);
1383

1384
	return retval;
1385 1386
}

E
Eric Dumazet 已提交
1387
static int dev_close_many(struct list_head *head)
1388 1389
{
	struct net_device *dev, *tmp;
L
Linus Torvalds 已提交
1390

1391 1392
	/* Remove the devices that don't need to be closed */
	list_for_each_entry_safe(dev, tmp, head, close_list)
1393
		if (!(dev->flags & IFF_UP))
1394
			list_del_init(&dev->close_list);
1395 1396

	__dev_close_many(head);
L
Linus Torvalds 已提交
1397

1398
	list_for_each_entry_safe(dev, tmp, head, close_list) {
1399
		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1400
		call_netdevice_notifiers(NETDEV_DOWN, dev);
1401
		list_del_init(&dev->close_list);
1402
	}
1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417

	return 0;
}

/**
 *	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)
{
1418 1419
	if (dev->flags & IFF_UP) {
		LIST_HEAD(single);
L
Linus Torvalds 已提交
1420

1421
		list_add(&dev->close_list, &single);
1422 1423 1424
		dev_close_many(&single);
		list_del(&single);
	}
1425
	return 0;
L
Linus Torvalds 已提交
1426
}
E
Eric Dumazet 已提交
1427
EXPORT_SYMBOL(dev_close);
L
Linus Torvalds 已提交
1428 1429


1430 1431 1432 1433 1434 1435 1436 1437 1438 1439
/**
 *	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)
{
1440 1441
	struct net_device *lower_dev;
	struct list_head *iter;
1442

M
Michał Mirosław 已提交
1443 1444
	dev->wanted_features &= ~NETIF_F_LRO;
	netdev_update_features(dev);
1445

1446 1447
	if (unlikely(dev->features & NETIF_F_LRO))
		netdev_WARN(dev, "failed to disable LRO!\n");
1448 1449 1450

	netdev_for_each_lower_dev(dev, lower_dev, iter)
		dev_disable_lro(lower_dev);
1451 1452 1453
}
EXPORT_SYMBOL(dev_disable_lro);

1454 1455 1456 1457 1458 1459 1460 1461
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);
}
1462

1463 1464
static int dev_boot_phase = 1;

L
Linus Torvalds 已提交
1465 1466 1467 1468 1469 1470 1471 1472 1473 1474
/**
 *	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
1475
 *	to the new notifier to allow device to have a race free
L
Linus Torvalds 已提交
1476 1477 1478 1479 1480 1481
 *	view of the network device list.
 */

int register_netdevice_notifier(struct notifier_block *nb)
{
	struct net_device *dev;
1482
	struct net_device *last;
1483
	struct net *net;
L
Linus Torvalds 已提交
1484 1485 1486
	int err;

	rtnl_lock();
1487
	err = raw_notifier_chain_register(&netdev_chain, nb);
1488 1489
	if (err)
		goto unlock;
1490 1491 1492 1493
	if (dev_boot_phase)
		goto unlock;
	for_each_net(net) {
		for_each_netdev(net, dev) {
1494
			err = call_netdevice_notifier(nb, NETDEV_REGISTER, dev);
1495 1496 1497 1498 1499 1500
			err = notifier_to_errno(err);
			if (err)
				goto rollback;

			if (!(dev->flags & IFF_UP))
				continue;
L
Linus Torvalds 已提交
1501

1502
			call_netdevice_notifier(nb, NETDEV_UP, dev);
1503
		}
L
Linus Torvalds 已提交
1504
	}
1505 1506

unlock:
L
Linus Torvalds 已提交
1507 1508
	rtnl_unlock();
	return err;
1509 1510 1511

rollback:
	last = dev;
1512 1513 1514
	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev == last)
1515
				goto outroll;
1516

1517
			if (dev->flags & IFF_UP) {
1518 1519 1520
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1521
			}
1522
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1523 1524
		}
	}
1525

1526
outroll:
1527
	raw_notifier_chain_unregister(&netdev_chain, nb);
1528
	goto unlock;
L
Linus Torvalds 已提交
1529
}
E
Eric Dumazet 已提交
1530
EXPORT_SYMBOL(register_netdevice_notifier);
L
Linus Torvalds 已提交
1531 1532 1533 1534 1535 1536 1537 1538 1539

/**
 *	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.
1540 1541 1542 1543
 *
 * 	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
1544 1545 1546 1547
 */

int unregister_netdevice_notifier(struct notifier_block *nb)
{
1548 1549
	struct net_device *dev;
	struct net *net;
1550 1551 1552
	int err;

	rtnl_lock();
1553
	err = raw_notifier_chain_unregister(&netdev_chain, nb);
1554 1555 1556 1557 1558 1559
	if (err)
		goto unlock;

	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev->flags & IFF_UP) {
1560 1561 1562
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1563
			}
1564
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1565 1566 1567
		}
	}
unlock:
1568 1569
	rtnl_unlock();
	return err;
L
Linus Torvalds 已提交
1570
}
E
Eric Dumazet 已提交
1571
EXPORT_SYMBOL(unregister_netdevice_notifier);
L
Linus Torvalds 已提交
1572

1573 1574 1575 1576 1577 1578 1579 1580 1581 1582
/**
 *	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
1583 1584 1585
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info)
1586 1587 1588 1589 1590 1591
{
	ASSERT_RTNL();
	netdev_notifier_info_init(info, dev);
	return raw_notifier_call_chain(&netdev_chain, val, info);
}

L
Linus Torvalds 已提交
1592 1593 1594
/**
 *	call_netdevice_notifiers - call all network notifier blocks
 *      @val: value passed unmodified to notifier function
1595
 *      @dev: net_device pointer passed unmodified to notifier function
L
Linus Torvalds 已提交
1596 1597
 *
 *	Call all network notifier blocks.  Parameters and return value
1598
 *	are as for raw_notifier_call_chain().
L
Linus Torvalds 已提交
1599 1600
 */

1601
int call_netdevice_notifiers(unsigned long val, struct net_device *dev)
L
Linus Torvalds 已提交
1602
{
1603 1604 1605
	struct netdev_notifier_info info;

	return call_netdevice_notifiers_info(val, dev, &info);
L
Linus Torvalds 已提交
1606
}
1607
EXPORT_SYMBOL(call_netdevice_notifiers);
L
Linus Torvalds 已提交
1608

1609
static struct static_key netstamp_needed __read_mostly;
1610
#ifdef HAVE_JUMP_LABEL
1611
/* We are not allowed to call static_key_slow_dec() from irq context
1612
 * If net_disable_timestamp() is called from irq context, defer the
1613
 * static_key_slow_dec() calls.
1614 1615 1616
 */
static atomic_t netstamp_needed_deferred;
#endif
L
Linus Torvalds 已提交
1617 1618 1619

void net_enable_timestamp(void)
{
1620 1621 1622 1623 1624
#ifdef HAVE_JUMP_LABEL
	int deferred = atomic_xchg(&netstamp_needed_deferred, 0);

	if (deferred) {
		while (--deferred)
1625
			static_key_slow_dec(&netstamp_needed);
1626 1627 1628
		return;
	}
#endif
1629
	static_key_slow_inc(&netstamp_needed);
L
Linus Torvalds 已提交
1630
}
E
Eric Dumazet 已提交
1631
EXPORT_SYMBOL(net_enable_timestamp);
L
Linus Torvalds 已提交
1632 1633 1634

void net_disable_timestamp(void)
{
1635 1636 1637 1638 1639 1640
#ifdef HAVE_JUMP_LABEL
	if (in_interrupt()) {
		atomic_inc(&netstamp_needed_deferred);
		return;
	}
#endif
1641
	static_key_slow_dec(&netstamp_needed);
L
Linus Torvalds 已提交
1642
}
E
Eric Dumazet 已提交
1643
EXPORT_SYMBOL(net_disable_timestamp);
L
Linus Torvalds 已提交
1644

E
Eric Dumazet 已提交
1645
static inline void net_timestamp_set(struct sk_buff *skb)
L
Linus Torvalds 已提交
1646
{
1647
	skb->tstamp.tv64 = 0;
1648
	if (static_key_false(&netstamp_needed))
1649
		__net_timestamp(skb);
L
Linus Torvalds 已提交
1650 1651
}

1652
#define net_timestamp_check(COND, SKB)			\
1653
	if (static_key_false(&netstamp_needed)) {		\
1654 1655 1656
		if ((COND) && !(SKB)->tstamp.tv64)	\
			__net_timestamp(SKB);		\
	}						\
E
Eric Dumazet 已提交
1657

1658
bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb)
1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676
{
	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;
}
1677
EXPORT_SYMBOL_GPL(is_skb_forwardable);
1678

H
Herbert Xu 已提交
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701
int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb)
{
	if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY) {
		if (skb_copy_ubufs(skb, GFP_ATOMIC)) {
			atomic_long_inc(&dev->rx_dropped);
			kfree_skb(skb);
			return NET_RX_DROP;
		}
	}

	if (unlikely(!is_skb_forwardable(dev, skb))) {
		atomic_long_inc(&dev->rx_dropped);
		kfree_skb(skb);
		return NET_RX_DROP;
	}

	skb_scrub_packet(skb, true);
	skb->protocol = eth_type_trans(skb, dev);

	return 0;
}
EXPORT_SYMBOL_GPL(__dev_forward_skb);

1702 1703 1704 1705 1706 1707 1708 1709
/**
 * dev_forward_skb - loopback an skb to another netif
 *
 * @dev: destination network device
 * @skb: buffer to forward
 *
 * return values:
 *	NET_RX_SUCCESS	(no congestion)
1710
 *	NET_RX_DROP     (packet was dropped, but freed)
1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721
 *
 * 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
1722
	return __dev_forward_skb(dev, skb) ?: netif_rx_internal(skb);
1723 1724 1725
}
EXPORT_SYMBOL_GPL(dev_forward_skb);

1726 1727 1728 1729
static inline int deliver_skb(struct sk_buff *skb,
			      struct packet_type *pt_prev,
			      struct net_device *orig_dev)
{
1730 1731
	if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
		return -ENOMEM;
1732 1733 1734 1735
	atomic_inc(&skb->users);
	return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
}

1736 1737
static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
{
1738
	if (!ptype->af_packet_priv || !skb->sk)
1739 1740 1741 1742 1743 1744 1745 1746 1747 1748
		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
1749 1750 1751 1752 1753
/*
 *	Support routine. Sends outgoing frames to any network
 *	taps currently in use.
 */

1754
static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
L
Linus Torvalds 已提交
1755 1756
{
	struct packet_type *ptype;
1757 1758
	struct sk_buff *skb2 = NULL;
	struct packet_type *pt_prev = NULL;
1759

L
Linus Torvalds 已提交
1760 1761 1762 1763 1764 1765
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
		/* Never send packets back to the socket
		 * they originated from - MvS (miquels@drinkel.ow.org)
		 */
		if ((ptype->dev == dev || !ptype->dev) &&
1766
		    (!skb_loop_sk(ptype, skb))) {
1767 1768 1769 1770 1771 1772 1773
			if (pt_prev) {
				deliver_skb(skb2, pt_prev, skb->dev);
				pt_prev = ptype;
				continue;
			}

			skb2 = skb_clone(skb, GFP_ATOMIC);
L
Linus Torvalds 已提交
1774 1775 1776
			if (!skb2)
				break;

1777 1778
			net_timestamp_set(skb2);

L
Linus Torvalds 已提交
1779 1780 1781 1782
			/* skb->nh should be correctly
			   set by sender, so that the second statement is
			   just protection against buggy protocols.
			 */
1783
			skb_reset_mac_header(skb2);
L
Linus Torvalds 已提交
1784

1785
			if (skb_network_header(skb2) < skb2->data ||
1786
			    skb_network_header(skb2) > skb_tail_pointer(skb2)) {
1787 1788 1789
				net_crit_ratelimited("protocol %04x is buggy, dev %s\n",
						     ntohs(skb2->protocol),
						     dev->name);
1790
				skb_reset_network_header(skb2);
L
Linus Torvalds 已提交
1791 1792
			}

1793
			skb2->transport_header = skb2->network_header;
L
Linus Torvalds 已提交
1794
			skb2->pkt_type = PACKET_OUTGOING;
1795
			pt_prev = ptype;
L
Linus Torvalds 已提交
1796 1797
		}
	}
1798 1799
	if (pt_prev)
		pt_prev->func(skb2, skb->dev, pt_prev, skb->dev);
L
Linus Torvalds 已提交
1800 1801 1802
	rcu_read_unlock();
}

1803 1804
/**
 * netif_setup_tc - Handle tc mappings on real_num_tx_queues change
1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815
 * @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
1816
static void netif_setup_tc(struct net_device *dev, unsigned int txq)
1817 1818 1819 1820 1821 1822
{
	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) {
1823
		pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n");
1824 1825 1826 1827 1828 1829 1830 1831 1832 1833
		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) {
1834 1835
			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);
1836 1837 1838 1839 1840
			netdev_set_prio_tc_map(dev, i, 0);
		}
	}
}

1841 1842 1843 1844 1845
#ifdef CONFIG_XPS
static DEFINE_MUTEX(xps_map_mutex);
#define xmap_dereference(P)		\
	rcu_dereference_protected((P), lockdep_is_held(&xps_map_mutex))

1846 1847
static struct xps_map *remove_xps_queue(struct xps_dev_maps *dev_maps,
					int cpu, u16 index)
1848
{
1849 1850
	struct xps_map *map = NULL;
	int pos;
1851

1852 1853
	if (dev_maps)
		map = xmap_dereference(dev_maps->cpu_map[cpu]);
1854

1855 1856
	for (pos = 0; map && pos < map->len; pos++) {
		if (map->queues[pos] == index) {
1857 1858 1859
			if (map->len > 1) {
				map->queues[pos] = map->queues[--map->len];
			} else {
1860
				RCU_INIT_POINTER(dev_maps->cpu_map[cpu], NULL);
1861 1862 1863
				kfree_rcu(map, rcu);
				map = NULL;
			}
1864
			break;
1865 1866 1867
		}
	}

1868 1869 1870
	return map;
}

1871
static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
1872 1873
{
	struct xps_dev_maps *dev_maps;
1874
	int cpu, i;
1875 1876 1877 1878 1879 1880 1881 1882 1883
	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) {
1884 1885 1886 1887 1888
		for (i = index; i < dev->num_tx_queues; i++) {
			if (!remove_xps_queue(dev_maps, cpu, i))
				break;
		}
		if (i == dev->num_tx_queues)
1889 1890 1891 1892
			active = true;
	}

	if (!active) {
1893 1894 1895 1896
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

1897 1898 1899 1900
	for (i = index; i < dev->num_tx_queues; i++)
		netdev_queue_numa_node_write(netdev_get_tx_queue(dev, i),
					     NUMA_NO_NODE);

1901 1902 1903 1904
out_no_maps:
	mutex_unlock(&xps_map_mutex);
}

1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939
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;
}

1940 1941
int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
			u16 index)
1942
{
1943
	struct xps_dev_maps *dev_maps, *new_dev_maps = NULL;
1944 1945
	struct xps_map *map, *new_map;
	int maps_sz = max_t(unsigned int, XPS_DEV_MAPS_SIZE, L1_CACHE_BYTES);
1946 1947
	int cpu, numa_node_id = -2;
	bool active = false;
1948 1949 1950 1951 1952

	mutex_lock(&xps_map_mutex);

	dev_maps = xmap_dereference(dev->xps_maps);

1953 1954 1955 1956 1957 1958 1959
	/* 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);
1960 1961
		if (!new_dev_maps) {
			mutex_unlock(&xps_map_mutex);
1962
			return -ENOMEM;
1963
		}
1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977

		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;

1978
	for_each_possible_cpu(cpu) {
1979 1980 1981 1982 1983 1984 1985 1986 1987 1988
		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;
1989 1990 1991 1992 1993 1994
#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
1995 1996 1997 1998
		} 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);
1999
		}
2000

2001 2002
	}

2003 2004
	rcu_assign_pointer(dev->xps_maps, new_dev_maps);

2005
	/* Cleanup old maps */
2006 2007 2008 2009 2010 2011 2012
	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);
		}
2013

2014
		kfree_rcu(dev_maps, rcu);
2015 2016
	}

2017 2018
	dev_maps = new_dev_maps;
	active = true;
2019

2020 2021
out_no_new_maps:
	/* update Tx queue numa node */
2022 2023 2024 2025
	netdev_queue_numa_node_write(netdev_get_tx_queue(dev, index),
				     (numa_node_id >= 0) ? numa_node_id :
				     NUMA_NO_NODE);

2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044
	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:
2045 2046 2047 2048
	mutex_unlock(&xps_map_mutex);

	return 0;
error:
2049 2050 2051 2052 2053 2054 2055 2056 2057
	/* 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);
	}

2058 2059 2060 2061 2062 2063 2064 2065
	mutex_unlock(&xps_map_mutex);

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

#endif
2066 2067 2068 2069
/*
 * 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.
 */
2070
int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
2071
{
T
Tom Herbert 已提交
2072 2073
	int rc;

2074 2075
	if (txq < 1 || txq > dev->num_tx_queues)
		return -EINVAL;
2076

2077 2078
	if (dev->reg_state == NETREG_REGISTERED ||
	    dev->reg_state == NETREG_UNREGISTERING) {
2079 2080
		ASSERT_RTNL();

T
Tom Herbert 已提交
2081 2082
		rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
						  txq);
T
Tom Herbert 已提交
2083 2084 2085
		if (rc)
			return rc;

2086 2087 2088
		if (dev->num_tc)
			netif_setup_tc(dev, txq);

2089
		if (txq < dev->real_num_tx_queues) {
2090
			qdisc_reset_all_tx_gt(dev, txq);
2091 2092 2093 2094
#ifdef CONFIG_XPS
			netif_reset_xps_queues_gt(dev, txq);
#endif
		}
2095
	}
2096 2097 2098

	dev->real_num_tx_queues = txq;
	return 0;
2099 2100
}
EXPORT_SYMBOL(netif_set_real_num_tx_queues);
2101

2102
#ifdef CONFIG_SYSFS
2103 2104 2105 2106 2107 2108 2109
/**
 *	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
2110 2111
 *	negative error code.  If called before registration, it always
 *	succeeds.
2112 2113 2114 2115 2116
 */
int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
{
	int rc;

T
Tom Herbert 已提交
2117 2118 2119
	if (rxq < 1 || rxq > dev->num_rx_queues)
		return -EINVAL;

2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134
	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

2135 2136
/**
 * netif_get_num_default_rss_queues - default number of RSS queues
2137 2138 2139 2140
 *
 * This routine should set an upper limit on the number of RSS queues
 * used by default by multiqueue devices.
 */
2141
int netif_get_num_default_rss_queues(void)
2142 2143 2144 2145 2146
{
	return min_t(int, DEFAULT_MAX_NUM_RSS_QUEUES, num_online_cpus());
}
EXPORT_SYMBOL(netif_get_num_default_rss_queues);

2147
static inline void __netif_reschedule(struct Qdisc *q)
2148
{
2149 2150
	struct softnet_data *sd;
	unsigned long flags;
2151

2152
	local_irq_save(flags);
2153
	sd = this_cpu_ptr(&softnet_data);
2154 2155 2156
	q->next_sched = NULL;
	*sd->output_queue_tailp = q;
	sd->output_queue_tailp = &q->next_sched;
2157 2158 2159 2160 2161 2162 2163 2164
	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);
2165 2166 2167
}
EXPORT_SYMBOL(__netif_schedule);

2168 2169 2170 2171 2172
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)
2173
{
2174 2175 2176
	return (struct dev_kfree_skb_cb *)skb->cb;
}

2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223
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);

2224
void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason)
2225
{
2226
	unsigned long flags;
2227

2228 2229 2230 2231 2232
	if (likely(atomic_read(&skb->users) == 1)) {
		smp_rmb();
		atomic_set(&skb->users, 0);
	} else if (likely(!atomic_dec_and_test(&skb->users))) {
		return;
2233
	}
2234 2235 2236 2237 2238 2239
	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);
2240
}
2241
EXPORT_SYMBOL(__dev_kfree_skb_irq);
2242

2243
void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason)
2244 2245
{
	if (in_irq() || irqs_disabled())
2246
		__dev_kfree_skb_irq(skb, reason);
2247 2248 2249
	else
		dev_kfree_skb(skb);
}
2250
EXPORT_SYMBOL(__dev_kfree_skb_any);
2251 2252


2253 2254 2255 2256 2257 2258
/**
 * netif_device_detach - mark device as removed
 * @dev: network device
 *
 * Mark device as removed from system and therefore no longer available.
 */
2259 2260 2261 2262
void netif_device_detach(struct net_device *dev)
{
	if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2263
		netif_tx_stop_all_queues(dev);
2264 2265 2266 2267
	}
}
EXPORT_SYMBOL(netif_device_detach);

2268 2269 2270 2271 2272 2273
/**
 * netif_device_attach - mark device as attached
 * @dev: network device
 *
 * Mark device as attached from system and restart if needed.
 */
2274 2275 2276 2277
void netif_device_attach(struct net_device *dev)
{
	if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2278
		netif_tx_wake_all_queues(dev);
2279
		__netdev_watchdog_up(dev);
2280 2281 2282 2283
	}
}
EXPORT_SYMBOL(netif_device_attach);

2284 2285
static void skb_warn_bad_offload(const struct sk_buff *skb)
{
2286
	static const netdev_features_t null_features = 0;
2287 2288 2289
	struct net_device *dev = skb->dev;
	const char *driver = "";

2290 2291 2292
	if (!net_ratelimit())
		return;

2293 2294 2295 2296 2297
	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",
2298 2299
	     driver, dev ? &dev->features : &null_features,
	     skb->sk ? &skb->sk->sk_route_caps : &null_features,
2300 2301 2302 2303
	     skb->len, skb->data_len, skb_shinfo(skb)->gso_size,
	     skb_shinfo(skb)->gso_type, skb->ip_summed);
}

L
Linus Torvalds 已提交
2304 2305 2306 2307
/*
 * Invalidate hardware checksum when packet is to be mangled, and
 * complete checksum manually on outgoing path.
 */
2308
int skb_checksum_help(struct sk_buff *skb)
L
Linus Torvalds 已提交
2309
{
2310
	__wsum csum;
2311
	int ret = 0, offset;
L
Linus Torvalds 已提交
2312

2313
	if (skb->ip_summed == CHECKSUM_COMPLETE)
2314 2315 2316
		goto out_set_summed;

	if (unlikely(skb_shinfo(skb)->gso_size)) {
2317 2318
		skb_warn_bad_offload(skb);
		return -EINVAL;
L
Linus Torvalds 已提交
2319 2320
	}

2321 2322 2323 2324 2325 2326 2327 2328 2329
	/* 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;
	}

2330
	offset = skb_checksum_start_offset(skb);
2331 2332 2333 2334 2335 2336 2337 2338
	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
2339 2340 2341 2342 2343
		ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
		if (ret)
			goto out;
	}

2344
	*(__sum16 *)(skb->data + offset) = csum_fold(csum);
2345
out_set_summed:
L
Linus Torvalds 已提交
2346
	skb->ip_summed = CHECKSUM_NONE;
2347
out:
L
Linus Torvalds 已提交
2348 2349
	return ret;
}
E
Eric Dumazet 已提交
2350
EXPORT_SYMBOL(skb_checksum_help);
L
Linus Torvalds 已提交
2351

2352
__be16 skb_network_protocol(struct sk_buff *skb, int *depth)
2353
{
2354
	unsigned int vlan_depth = skb->mac_len;
A
Al Viro 已提交
2355
	__be16 type = skb->protocol;
2356

2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367
	/* 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;
	}

2368 2369 2370 2371 2372 2373
	/* if skb->protocol is 802.1Q/AD then the header should already be
	 * present at mac_len - VLAN_HLEN (if mac_len > 0), or at
	 * ETH_HLEN otherwise
	 */
	if (type == htons(ETH_P_8021Q) || type == htons(ETH_P_8021AD)) {
		if (vlan_depth) {
2374
			if (WARN_ON(vlan_depth < VLAN_HLEN))
2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391
				return 0;
			vlan_depth -= VLAN_HLEN;
		} else {
			vlan_depth = ETH_HLEN;
		}
		do {
			struct vlan_hdr *vh;

			if (unlikely(!pskb_may_pull(skb,
						    vlan_depth + VLAN_HLEN)))
				return 0;

			vh = (struct vlan_hdr *)(skb->data + vlan_depth);
			type = vh->h_vlan_encapsulated_proto;
			vlan_depth += VLAN_HLEN;
		} while (type == htons(ETH_P_8021Q) ||
			 type == htons(ETH_P_8021AD));
2392 2393
	}

2394 2395
	*depth = vlan_depth;

2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408
	return type;
}

/**
 *	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;
2409 2410
	int vlan_depth = skb->mac_len;
	__be16 type = skb_network_protocol(skb, &vlan_depth);
2411 2412 2413 2414

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

2415
	__skb_pull(skb, vlan_depth);
2416 2417

	rcu_read_lock();
2418
	list_for_each_entry_rcu(ptype, &offload_base, list) {
2419 2420
		if (ptype->type == type && ptype->callbacks.gso_segment) {
			segs = ptype->callbacks.gso_segment(skb, features);
2421 2422 2423 2424 2425
			break;
		}
	}
	rcu_read_unlock();

2426
	__skb_push(skb, skb->data - skb_mac_header(skb));
2427

2428 2429
	return segs;
}
2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461
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);

2462 2463
		err = skb_cow_head(skb, 0);
		if (err < 0)
2464 2465 2466
			return ERR_PTR(err);
	}

2467
	SKB_GSO_CB(skb)->mac_offset = skb_headroom(skb);
2468 2469
	SKB_GSO_CB(skb)->encap_level = 0;

2470 2471 2472 2473 2474
	skb_reset_mac_header(skb);
	skb_reset_mac_len(skb);

	return skb_mac_gso_segment(skb, features);
}
2475
EXPORT_SYMBOL(__skb_gso_segment);
2476

2477 2478 2479 2480 2481
/* Take action when hardware reception checksum errors are detected. */
#ifdef CONFIG_BUG
void netdev_rx_csum_fault(struct net_device *dev)
{
	if (net_ratelimit()) {
2482
		pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>");
2483 2484 2485 2486 2487 2488
		dump_stack();
	}
}
EXPORT_SYMBOL(netdev_rx_csum_fault);
#endif

L
Linus Torvalds 已提交
2489 2490 2491 2492 2493
/* 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.
 */

2494
static int illegal_highdma(struct net_device *dev, struct sk_buff *skb)
L
Linus Torvalds 已提交
2495
{
2496
#ifdef CONFIG_HIGHMEM
L
Linus Torvalds 已提交
2497
	int i;
2498
	if (!(dev->features & NETIF_F_HIGHDMA)) {
2499 2500 2501
		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)))
2502
				return 1;
2503
		}
2504
	}
L
Linus Torvalds 已提交
2505

2506 2507
	if (PCI_DMA_BUS_IS_PHYS) {
		struct device *pdev = dev->dev.parent;
L
Linus Torvalds 已提交
2508

E
Eric Dumazet 已提交
2509 2510
		if (!pdev)
			return 0;
2511
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2512 2513
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			dma_addr_t addr = page_to_phys(skb_frag_page(frag));
2514 2515 2516 2517
			if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask)
				return 1;
		}
	}
2518
#endif
L
Linus Torvalds 已提交
2519 2520 2521
	return 0;
}

2522 2523 2524 2525 2526 2527 2528 2529
/* If MPLS offload request, verify we are testing hardware MPLS features
 * instead of standard features for the netdev.
 */
#ifdef CONFIG_NET_MPLS_GSO
static netdev_features_t net_mpls_features(struct sk_buff *skb,
					   netdev_features_t features,
					   __be16 type)
{
2530
	if (eth_p_mpls(type))
2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543
		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

2544
static netdev_features_t harmonize_features(struct sk_buff *skb,
2545
	netdev_features_t features)
2546
{
2547
	int tmp;
2548 2549 2550 2551
	__be16 type;

	type = skb_network_protocol(skb, &tmp);
	features = net_mpls_features(skb, features, type);
2552

2553
	if (skb->ip_summed != CHECKSUM_NONE &&
2554
	    !can_checksum_protocol(features, type)) {
2555
		features &= ~NETIF_F_ALL_CSUM;
2556
	} else if (illegal_highdma(skb->dev, skb)) {
2557 2558 2559 2560 2561 2562
		features &= ~NETIF_F_SG;
	}

	return features;
}

2563
netdev_features_t netif_skb_features(struct sk_buff *skb)
2564
{
2565 2566 2567
	const struct net_device *dev = skb->dev;
	netdev_features_t features = dev->features;
	u16 gso_segs = skb_shinfo(skb)->gso_segs;
2568 2569
	__be16 protocol = skb->protocol;

2570
	if (gso_segs > dev->gso_max_segs || gso_segs < dev->gso_min_segs)
2571 2572
		features &= ~NETIF_F_GSO_MASK;

P
Patrick McHardy 已提交
2573
	if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD)) {
2574 2575
		struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
		protocol = veh->h_vlan_encapsulated_proto;
2576
	} else if (!vlan_tx_tag_present(skb)) {
2577
		return harmonize_features(skb, features);
2578
	}
2579

2580
	features = netdev_intersect_features(features,
2581
					     dev->vlan_features |
2582 2583
					     NETIF_F_HW_VLAN_CTAG_TX |
					     NETIF_F_HW_VLAN_STAG_TX);
2584

2585
	if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD))
2586 2587 2588 2589 2590 2591 2592
		features = netdev_intersect_features(features,
						     NETIF_F_SG |
						     NETIF_F_HIGHDMA |
						     NETIF_F_FRAGLIST |
						     NETIF_F_GEN_CSUM |
						     NETIF_F_HW_VLAN_CTAG_TX |
						     NETIF_F_HW_VLAN_STAG_TX);
2593

2594
	return harmonize_features(skb, features);
2595
}
2596
EXPORT_SYMBOL(netif_skb_features);
2597

2598
static int xmit_one(struct sk_buff *skb, struct net_device *dev,
2599
		    struct netdev_queue *txq, bool more)
2600
{
2601 2602
	unsigned int len;
	int rc;
2603

2604 2605
	if (!list_empty(&ptype_all))
		dev_queue_xmit_nit(skb, dev);
2606

2607 2608
	len = skb->len;
	trace_net_dev_start_xmit(skb, dev);
2609
	rc = netdev_start_xmit(skb, dev, txq, more);
2610
	trace_net_dev_xmit(skb, rc, dev, len);
E
Eric Dumazet 已提交
2611

2612 2613
	return rc;
}
2614

2615 2616
struct sk_buff *dev_hard_start_xmit(struct sk_buff *first, struct net_device *dev,
				    struct netdev_queue *txq, int *ret)
2617 2618 2619
{
	struct sk_buff *skb = first;
	int rc = NETDEV_TX_OK;
2620

2621 2622
	while (skb) {
		struct sk_buff *next = skb->next;
2623

2624
		skb->next = NULL;
2625
		rc = xmit_one(skb, dev, txq, next != NULL);
2626 2627 2628 2629
		if (unlikely(!dev_xmit_complete(rc))) {
			skb->next = next;
			goto out;
		}
2630

2631 2632 2633 2634
		skb = next;
		if (netif_xmit_stopped(txq) && skb) {
			rc = NETDEV_TX_BUSY;
			break;
2635
		}
2636
	}
2637

2638 2639 2640 2641
out:
	*ret = rc;
	return skb;
}
2642

E
Eric Dumazet 已提交
2643 2644
static struct sk_buff *validate_xmit_vlan(struct sk_buff *skb,
					  netdev_features_t features)
2645
{
2646
	if (vlan_tx_tag_present(skb) &&
2647 2648
	    !vlan_hw_offload_capable(features, skb->vlan_proto))
		skb = __vlan_hwaccel_push_inside(skb);
2649 2650
	return skb;
}
2651

2652
static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev)
2653 2654
{
	netdev_features_t features;
2655

2656 2657
	if (skb->next)
		return skb;
2658

2659 2660 2661 2662
	features = netif_skb_features(skb);
	skb = validate_xmit_vlan(skb, features);
	if (unlikely(!skb))
		goto out_null;
2663

2664 2665 2666 2667 2668 2669 2670
	/* 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;

T
Tom Herbert 已提交
2671
	if (netif_needs_gso(dev, skb, features)) {
2672 2673 2674
		struct sk_buff *segs;

		segs = skb_gso_segment(skb, features);
2675
		if (IS_ERR(segs)) {
2676
			goto out_kfree_skb;
2677 2678 2679
		} else if (segs) {
			consume_skb(skb);
			skb = segs;
2680
		}
2681 2682 2683 2684
	} else {
		if (skb_needs_linearize(skb, features) &&
		    __skb_linearize(skb))
			goto out_kfree_skb;
2685

2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699
		/* 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;
2700
		}
2701
	}
2702

2703
	return skb;
2704

2705 2706
out_kfree_skb:
	kfree_skb(skb);
2707 2708 2709
out_null:
	return NULL;
}
2710

2711 2712 2713 2714
struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev)
{
	struct sk_buff *next, *head = NULL, *tail;

2715
	for (; skb != NULL; skb = next) {
2716 2717
		next = skb->next;
		skb->next = NULL;
2718 2719 2720 2721

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

2722
		skb = validate_xmit_skb(skb, dev);
2723 2724
		if (!skb)
			continue;
2725

2726 2727 2728 2729 2730 2731 2732 2733
		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;
2734 2735
	}
	return head;
2736 2737
}

2738 2739 2740 2741 2742 2743 2744 2745 2746 2747
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)  {
2748
		unsigned int hdr_len;
2749
		u16 gso_segs = shinfo->gso_segs;
2750

2751 2752 2753 2754
		/* mac layer + network layer */
		hdr_len = skb_transport_header(skb) - skb_mac_header(skb);

		/* + transport layer */
2755 2756 2757 2758
		if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)))
			hdr_len += tcp_hdrlen(skb);
		else
			hdr_len += sizeof(struct udphdr);
2759 2760 2761 2762 2763 2764

		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;
2765 2766 2767
	}
}

2768 2769 2770 2771 2772
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
2773
	bool contended;
2774 2775
	int rc;

2776
	qdisc_pkt_len_init(skb);
E
Eric Dumazet 已提交
2777
	qdisc_calculate_pkt_len(skb, q);
2778 2779 2780
	/*
	 * Heuristic to force contended enqueues to serialize on a
	 * separate lock before trying to get qdisc main lock.
Y
Ying Xue 已提交
2781 2782
	 * This permits __QDISC___STATE_RUNNING owner to get the lock more
	 * often and dequeue packets faster.
2783
	 */
E
Eric Dumazet 已提交
2784
	contended = qdisc_is_running(q);
2785 2786 2787
	if (unlikely(contended))
		spin_lock(&q->busylock);

2788 2789 2790 2791 2792
	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) &&
2793
		   qdisc_run_begin(q)) {
2794 2795 2796 2797 2798
		/*
		 * 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.
		 */
2799 2800 2801

		qdisc_bstats_update(q, skb);

2802
		if (sch_direct_xmit(skb, q, dev, txq, root_lock, true)) {
2803 2804 2805 2806
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
2807
			__qdisc_run(q);
2808
		} else
2809
			qdisc_run_end(q);
2810 2811 2812

		rc = NET_XMIT_SUCCESS;
	} else {
E
Eric Dumazet 已提交
2813
		rc = q->enqueue(skb, q) & NET_XMIT_MASK;
2814 2815 2816 2817 2818 2819 2820
		if (qdisc_run_begin(q)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
			__qdisc_run(q);
		}
2821 2822
	}
	spin_unlock(root_lock);
2823 2824
	if (unlikely(contended))
		spin_unlock(&q->busylock);
2825 2826 2827
	return rc;
}

2828
#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2829 2830
static void skb_update_prio(struct sk_buff *skb)
{
I
Igor Maravic 已提交
2831
	struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap);
2832

2833 2834 2835 2836 2837 2838
	if (!skb->priority && skb->sk && map) {
		unsigned int prioidx = skb->sk->sk_cgrp_prioidx;

		if (prioidx < map->priomap_len)
			skb->priority = map->priomap[prioidx];
	}
2839 2840 2841 2842 2843
}
#else
#define skb_update_prio(skb)
#endif

2844
static DEFINE_PER_CPU(int, xmit_recursion);
2845
#define RECURSION_LIMIT 10
2846

2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863
/**
 *	dev_loopback_xmit - loop back @skb
 *	@skb: buffer to transmit
 */
int dev_loopback_xmit(struct sk_buff *skb)
{
	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);

2864
/**
2865
 *	__dev_queue_xmit - transmit a buffer
2866
 *	@skb: buffer to transmit
2867
 *	@accel_priv: private data used for L2 forwarding offload
2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889
 *
 *	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
 */
2890
static int __dev_queue_xmit(struct sk_buff *skb, void *accel_priv)
L
Linus Torvalds 已提交
2891 2892
{
	struct net_device *dev = skb->dev;
2893
	struct netdev_queue *txq;
L
Linus Torvalds 已提交
2894 2895 2896
	struct Qdisc *q;
	int rc = -ENOMEM;

2897 2898
	skb_reset_mac_header(skb);

2899 2900 2901
	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_SCHED_TSTAMP))
		__skb_tstamp_tx(skb, NULL, skb->sk, SCM_TSTAMP_SCHED);

2902 2903
	/* Disable soft irqs for various locks below. Also
	 * stops preemption for RCU.
L
Linus Torvalds 已提交
2904
	 */
2905
	rcu_read_lock_bh();
L
Linus Torvalds 已提交
2906

2907 2908
	skb_update_prio(skb);

2909 2910 2911 2912 2913 2914 2915 2916
	/* 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);

2917
	txq = netdev_pick_tx(dev, skb, accel_priv);
2918
	q = rcu_dereference_bh(txq->qdisc);
2919

L
Linus Torvalds 已提交
2920
#ifdef CONFIG_NET_CLS_ACT
E
Eric Dumazet 已提交
2921
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS);
L
Linus Torvalds 已提交
2922
#endif
2923
	trace_net_dev_queue(skb);
L
Linus Torvalds 已提交
2924
	if (q->enqueue) {
2925
		rc = __dev_xmit_skb(skb, q, dev, txq);
2926
		goto out;
L
Linus Torvalds 已提交
2927 2928 2929 2930 2931
	}

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

H
Herbert Xu 已提交
2932 2933
	   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
2934 2935 2936 2937 2938 2939 2940 2941 2942 2943
	   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 */

2944
		if (txq->xmit_lock_owner != cpu) {
L
Linus Torvalds 已提交
2945

2946 2947 2948
			if (__this_cpu_read(xmit_recursion) > RECURSION_LIMIT)
				goto recursion_alert;

2949 2950 2951 2952
			skb = validate_xmit_skb(skb, dev);
			if (!skb)
				goto drop;

2953
			HARD_TX_LOCK(dev, txq, cpu);
L
Linus Torvalds 已提交
2954

2955
			if (!netif_xmit_stopped(txq)) {
2956
				__this_cpu_inc(xmit_recursion);
2957
				skb = dev_hard_start_xmit(skb, dev, txq, &rc);
2958
				__this_cpu_dec(xmit_recursion);
2959
				if (dev_xmit_complete(rc)) {
2960
					HARD_TX_UNLOCK(dev, txq);
L
Linus Torvalds 已提交
2961 2962 2963
					goto out;
				}
			}
2964
			HARD_TX_UNLOCK(dev, txq);
2965 2966
			net_crit_ratelimited("Virtual device %s asks to queue packet!\n",
					     dev->name);
L
Linus Torvalds 已提交
2967 2968
		} else {
			/* Recursion is detected! It is possible,
2969 2970 2971
			 * unfortunately
			 */
recursion_alert:
2972 2973
			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
					     dev->name);
L
Linus Torvalds 已提交
2974 2975 2976 2977
		}
	}

	rc = -ENETDOWN;
2978
drop:
2979
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
2980

2981
	atomic_long_inc(&dev->tx_dropped);
2982
	kfree_skb_list(skb);
L
Linus Torvalds 已提交
2983 2984
	return rc;
out:
2985
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
2986 2987
	return rc;
}
2988 2989 2990 2991 2992

int dev_queue_xmit(struct sk_buff *skb)
{
	return __dev_queue_xmit(skb, NULL);
}
E
Eric Dumazet 已提交
2993
EXPORT_SYMBOL(dev_queue_xmit);
L
Linus Torvalds 已提交
2994

2995 2996 2997 2998 2999 3000
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
3001 3002 3003 3004 3005

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

3006
int netdev_max_backlog __read_mostly = 1000;
E
Eric Dumazet 已提交
3007 3008
EXPORT_SYMBOL(netdev_max_backlog);

E
Eric Dumazet 已提交
3009
int netdev_tstamp_prequeue __read_mostly = 1;
3010 3011
int netdev_budget __read_mostly = 300;
int weight_p __read_mostly = 64;            /* old backlog weight */
L
Linus Torvalds 已提交
3012

E
Eric Dumazet 已提交
3013 3014 3015 3016 3017 3018 3019 3020
/* 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);
}

3021 3022 3023
#ifdef CONFIG_RPS

/* One global table that all flow-based protocols share. */
E
Eric Dumazet 已提交
3024
struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
3025 3026
EXPORT_SYMBOL(rps_sock_flow_table);

3027
struct static_key rps_needed __read_mostly;
3028

3029 3030 3031 3032
static struct rps_dev_flow *
set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
	    struct rps_dev_flow *rflow, u16 next_cpu)
{
3033
	if (next_cpu != RPS_NO_CPU) {
3034 3035 3036 3037 3038 3039 3040 3041 3042
#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? */
3043 3044
		if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
		    !(dev->features & NETIF_F_NTUPLE))
3045 3046 3047 3048 3049 3050 3051 3052 3053
			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;
3054
		flow_id = skb_get_hash(skb) & flow_table->mask;
3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066
		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 =
3067
			per_cpu(softnet_data, next_cpu).input_queue_head;
3068 3069
	}

3070
	rflow->cpu = next_cpu;
3071 3072 3073
	return rflow;
}

3074 3075 3076 3077 3078 3079 3080 3081 3082
/*
 * 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)
{
	struct netdev_rx_queue *rxqueue;
E
Eric Dumazet 已提交
3083
	struct rps_map *map;
3084 3085 3086 3087
	struct rps_dev_flow_table *flow_table;
	struct rps_sock_flow_table *sock_flow_table;
	int cpu = -1;
	u16 tcpu;
3088
	u32 hash;
3089 3090 3091

	if (skb_rx_queue_recorded(skb)) {
		u16 index = skb_get_rx_queue(skb);
3092 3093 3094 3095 3096
		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);
3097 3098 3099 3100 3101 3102
			goto done;
		}
		rxqueue = dev->_rx + index;
	} else
		rxqueue = dev->_rx;

E
Eric Dumazet 已提交
3103 3104
	map = rcu_dereference(rxqueue->rps_map);
	if (map) {
3105
		if (map->len == 1 &&
3106
		    !rcu_access_pointer(rxqueue->rps_flow_table)) {
3107 3108 3109 3110 3111
			tcpu = map->cpus[0];
			if (cpu_online(tcpu))
				cpu = tcpu;
			goto done;
		}
3112
	} else if (!rcu_access_pointer(rxqueue->rps_flow_table)) {
3113
		goto done;
3114
	}
3115

3116
	skb_reset_network_header(skb);
3117 3118
	hash = skb_get_hash(skb);
	if (!hash)
3119 3120
		goto done;

T
Tom Herbert 已提交
3121 3122 3123 3124 3125 3126
	flow_table = rcu_dereference(rxqueue->rps_flow_table);
	sock_flow_table = rcu_dereference(rps_sock_flow_table);
	if (flow_table && sock_flow_table) {
		u16 next_cpu;
		struct rps_dev_flow *rflow;

3127
		rflow = &flow_table->flows[hash & flow_table->mask];
T
Tom Herbert 已提交
3128 3129
		tcpu = rflow->cpu;

3130
		next_cpu = sock_flow_table->ents[hash & sock_flow_table->mask];
T
Tom Herbert 已提交
3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145

		/*
		 * 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:
		 *   - Current CPU is unset (equal to RPS_NO_CPU).
		 *   - 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) &&
		    (tcpu == RPS_NO_CPU || !cpu_online(tcpu) ||
		     ((int)(per_cpu(softnet_data, tcpu).input_queue_head -
3146 3147
		      rflow->last_qtail)) >= 0)) {
			tcpu = next_cpu;
3148
			rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
3149
		}
3150

T
Tom Herbert 已提交
3151 3152 3153 3154 3155 3156 3157
		if (tcpu != RPS_NO_CPU && cpu_online(tcpu)) {
			*rflowp = rflow;
			cpu = tcpu;
			goto done;
		}
	}

T
Tom Herbert 已提交
3158
	if (map) {
3159
		tcpu = map->cpus[reciprocal_scale(hash, map->len)];
T
Tom Herbert 已提交
3160 3161 3162 3163 3164 3165 3166 3167 3168 3169
		if (cpu_online(tcpu)) {
			cpu = tcpu;
			goto done;
		}
	}

done:
	return cpu;
}

3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209
#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;
	int cpu;

	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);
		if (rflow->filter == filter_id && cpu != RPS_NO_CPU &&
		    ((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
3210
/* Called from hardirq (IPI) context */
E
Eric Dumazet 已提交
3211
static void rps_trigger_softirq(void *data)
T
Tom Herbert 已提交
3212
{
E
Eric Dumazet 已提交
3213 3214
	struct softnet_data *sd = data;

E
Eric Dumazet 已提交
3215
	____napi_schedule(sd, &sd->backlog);
C
Changli Gao 已提交
3216
	sd->received_rps++;
T
Tom Herbert 已提交
3217
}
E
Eric Dumazet 已提交
3218

T
Tom Herbert 已提交
3219
#endif /* CONFIG_RPS */
T
Tom Herbert 已提交
3220

E
Eric Dumazet 已提交
3221 3222 3223 3224 3225 3226 3227 3228
/*
 * 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
3229
	struct softnet_data *mysd = this_cpu_ptr(&softnet_data);
E
Eric Dumazet 已提交
3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241

	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;
}

3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255
#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;

3256
	sd = this_cpu_ptr(&softnet_data);
3257 3258 3259 3260

	rcu_read_lock();
	fl = rcu_dereference(sd->flow_limit);
	if (fl) {
3261
		new_flow = skb_get_hash(skb) & (fl->num_buckets - 1);
3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281
		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
3282 3283 3284 3285
/*
 * 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
3286 3287
static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
			      unsigned int *qtail)
T
Tom Herbert 已提交
3288
{
E
Eric Dumazet 已提交
3289
	struct softnet_data *sd;
T
Tom Herbert 已提交
3290
	unsigned long flags;
3291
	unsigned int qlen;
T
Tom Herbert 已提交
3292

E
Eric Dumazet 已提交
3293
	sd = &per_cpu(softnet_data, cpu);
T
Tom Herbert 已提交
3294 3295 3296

	local_irq_save(flags);

E
Eric Dumazet 已提交
3297
	rps_lock(sd);
3298 3299
	qlen = skb_queue_len(&sd->input_pkt_queue);
	if (qlen <= netdev_max_backlog && !skb_flow_limit(skb, qlen)) {
3300
		if (qlen) {
T
Tom Herbert 已提交
3301
enqueue:
E
Eric Dumazet 已提交
3302
			__skb_queue_tail(&sd->input_pkt_queue, skb);
3303
			input_queue_tail_incr_save(sd, qtail);
E
Eric Dumazet 已提交
3304
			rps_unlock(sd);
3305
			local_irq_restore(flags);
T
Tom Herbert 已提交
3306 3307 3308
			return NET_RX_SUCCESS;
		}

3309 3310 3311 3312
		/* 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
3313
			if (!rps_ipi_queued(sd))
E
Eric Dumazet 已提交
3314
				____napi_schedule(sd, &sd->backlog);
T
Tom Herbert 已提交
3315 3316 3317 3318
		}
		goto enqueue;
	}

C
Changli Gao 已提交
3319
	sd->dropped++;
E
Eric Dumazet 已提交
3320
	rps_unlock(sd);
T
Tom Herbert 已提交
3321 3322 3323

	local_irq_restore(flags);

3324
	atomic_long_inc(&skb->dev->rx_dropped);
T
Tom Herbert 已提交
3325 3326 3327
	kfree_skb(skb);
	return NET_RX_DROP;
}
L
Linus Torvalds 已提交
3328

3329
static int netif_rx_internal(struct sk_buff *skb)
L
Linus Torvalds 已提交
3330
{
3331
	int ret;
L
Linus Torvalds 已提交
3332

3333
	net_timestamp_check(netdev_tstamp_prequeue, skb);
L
Linus Torvalds 已提交
3334

3335
	trace_netif_rx(skb);
E
Eric Dumazet 已提交
3336
#ifdef CONFIG_RPS
3337
	if (static_key_false(&rps_needed)) {
T
Tom Herbert 已提交
3338
		struct rps_dev_flow voidflow, *rflow = &voidflow;
3339 3340
		int cpu;

3341
		preempt_disable();
3342
		rcu_read_lock();
T
Tom Herbert 已提交
3343 3344

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
3345 3346
		if (cpu < 0)
			cpu = smp_processor_id();
T
Tom Herbert 已提交
3347 3348 3349

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

3350
		rcu_read_unlock();
3351
		preempt_enable();
3352 3353
	} else
#endif
T
Tom Herbert 已提交
3354 3355 3356 3357 3358
	{
		unsigned int qtail;
		ret = enqueue_to_backlog(skb, get_cpu(), &qtail);
		put_cpu();
	}
3359
	return ret;
L
Linus Torvalds 已提交
3360
}
3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382

/**
 *	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
3383
EXPORT_SYMBOL(netif_rx);
L
Linus Torvalds 已提交
3384 3385 3386 3387 3388

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

3389 3390
	trace_netif_rx_ni_entry(skb);

L
Linus Torvalds 已提交
3391
	preempt_disable();
3392
	err = netif_rx_internal(skb);
L
Linus Torvalds 已提交
3393 3394 3395 3396 3397 3398 3399 3400 3401 3402
	if (local_softirq_pending())
		do_softirq();
	preempt_enable();

	return err;
}
EXPORT_SYMBOL(netif_rx_ni);

static void net_tx_action(struct softirq_action *h)
{
3403
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
L
Linus Torvalds 已提交
3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416

	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;

3417
			WARN_ON(atomic_read(&skb->users));
3418 3419 3420 3421
			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
3422 3423 3424 3425 3426
			__kfree_skb(skb);
		}
	}

	if (sd->output_queue) {
3427
		struct Qdisc *head;
L
Linus Torvalds 已提交
3428 3429 3430 3431

		local_irq_disable();
		head = sd->output_queue;
		sd->output_queue = NULL;
3432
		sd->output_queue_tailp = &sd->output_queue;
L
Linus Torvalds 已提交
3433 3434 3435
		local_irq_enable();

		while (head) {
3436 3437 3438
			struct Qdisc *q = head;
			spinlock_t *root_lock;

L
Linus Torvalds 已提交
3439 3440
			head = head->next_sched;

3441
			root_lock = qdisc_lock(q);
3442
			if (spin_trylock(root_lock)) {
3443
				smp_mb__before_atomic();
3444 3445
				clear_bit(__QDISC_STATE_SCHED,
					  &q->state);
3446 3447
				qdisc_run(q);
				spin_unlock(root_lock);
L
Linus Torvalds 已提交
3448
			} else {
3449
				if (!test_bit(__QDISC_STATE_DEACTIVATED,
3450
					      &q->state)) {
3451
					__netif_reschedule(q);
3452
				} else {
3453
					smp_mb__before_atomic();
3454 3455 3456
					clear_bit(__QDISC_STATE_SCHED,
						  &q->state);
				}
L
Linus Torvalds 已提交
3457 3458 3459 3460 3461
			}
		}
	}
}

3462 3463
#if (defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)) && \
    (defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE))
3464 3465 3466
/* This hook is defined here for ATM LANE */
int (*br_fdb_test_addr_hook)(struct net_device *dev,
			     unsigned char *addr) __read_mostly;
3467
EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
3468
#endif
L
Linus Torvalds 已提交
3469 3470 3471 3472 3473 3474

#ifdef CONFIG_NET_CLS_ACT
/* TODO: Maybe we should just force sch_ingress to be compiled in
 * when CONFIG_NET_CLS_ACT is? otherwise some useless instructions
 * a compare and 2 stores extra right now if we dont have it on
 * but have CONFIG_NET_CLS_ACT
L
Lucas De Marchi 已提交
3475 3476
 * NOTE: This doesn't stop any functionality; if you dont have
 * the ingress scheduler, you just can't add policies on ingress.
L
Linus Torvalds 已提交
3477 3478
 *
 */
3479
static int ing_filter(struct sk_buff *skb, struct netdev_queue *rxq)
L
Linus Torvalds 已提交
3480 3481
{
	struct net_device *dev = skb->dev;
3482
	u32 ttl = G_TC_RTTL(skb->tc_verd);
3483 3484
	int result = TC_ACT_OK;
	struct Qdisc *q;
3485

3486
	if (unlikely(MAX_RED_LOOP < ttl++)) {
3487 3488
		net_warn_ratelimited("Redir loop detected Dropping packet (%d->%d)\n",
				     skb->skb_iif, dev->ifindex);
3489 3490
		return TC_ACT_SHOT;
	}
L
Linus Torvalds 已提交
3491

3492 3493
	skb->tc_verd = SET_TC_RTTL(skb->tc_verd, ttl);
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS);
L
Linus Torvalds 已提交
3494

3495
	q = rcu_dereference(rxq->qdisc);
3496
	if (q != &noop_qdisc) {
3497
		spin_lock(qdisc_lock(q));
3498 3499
		if (likely(!test_bit(__QDISC_STATE_DEACTIVATED, &q->state)))
			result = qdisc_enqueue_root(skb, q);
3500 3501
		spin_unlock(qdisc_lock(q));
	}
3502 3503 3504

	return result;
}
3505

3506 3507 3508 3509
static inline struct sk_buff *handle_ing(struct sk_buff *skb,
					 struct packet_type **pt_prev,
					 int *ret, struct net_device *orig_dev)
{
3510 3511
	struct netdev_queue *rxq = rcu_dereference(skb->dev->ingress_queue);

3512
	if (!rxq || rcu_access_pointer(rxq->qdisc) == &noop_qdisc)
3513
		goto out;
L
Linus Torvalds 已提交
3514

3515 3516 3517
	if (*pt_prev) {
		*ret = deliver_skb(skb, *pt_prev, orig_dev);
		*pt_prev = NULL;
L
Linus Torvalds 已提交
3518 3519
	}

3520
	switch (ing_filter(skb, rxq)) {
3521 3522 3523 3524 3525 3526 3527 3528 3529
	case TC_ACT_SHOT:
	case TC_ACT_STOLEN:
		kfree_skb(skb);
		return NULL;
	}

out:
	skb->tc_verd = 0;
	return skb;
L
Linus Torvalds 已提交
3530 3531 3532
}
#endif

3533 3534 3535 3536
/**
 *	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
3537
 *	@rx_handler_data: data pointer that is used by rx handler
3538
 *
3539
 *	Register a receive handler for a device. This handler will then be
3540 3541 3542 3543
 *	called from __netif_receive_skb. A negative errno code is returned
 *	on a failure.
 *
 *	The caller must hold the rtnl_mutex.
3544 3545
 *
 *	For a general description of rx_handler, see enum rx_handler_result.
3546 3547
 */
int netdev_rx_handler_register(struct net_device *dev,
J
Jiri Pirko 已提交
3548 3549
			       rx_handler_func_t *rx_handler,
			       void *rx_handler_data)
3550 3551 3552 3553 3554 3555
{
	ASSERT_RTNL();

	if (dev->rx_handler)
		return -EBUSY;

3556
	/* Note: rx_handler_data must be set before rx_handler */
J
Jiri Pirko 已提交
3557
	rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
3558 3559 3560 3561 3562 3563 3564 3565 3566 3567
	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
3568
 *	Unregister a receive handler from a device.
3569 3570 3571 3572 3573 3574 3575
 *
 *	The caller must hold the rtnl_mutex.
 */
void netdev_rx_handler_unregister(struct net_device *dev)
{

	ASSERT_RTNL();
3576
	RCU_INIT_POINTER(dev->rx_handler, NULL);
3577 3578 3579 3580 3581
	/* 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();
3582
	RCU_INIT_POINTER(dev->rx_handler_data, NULL);
3583 3584 3585
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);

3586 3587 3588 3589 3590 3591 3592
/*
 * 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) {
3593 3594 3595 3596 3597
	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):
3598 3599 3600 3601 3602 3603
		return true;
	default:
		return false;
	}
}

3604
static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc)
L
Linus Torvalds 已提交
3605 3606
{
	struct packet_type *ptype, *pt_prev;
3607
	rx_handler_func_t *rx_handler;
D
David S. Miller 已提交
3608
	struct net_device *orig_dev;
3609
	struct net_device *null_or_dev;
3610
	bool deliver_exact = false;
L
Linus Torvalds 已提交
3611
	int ret = NET_RX_DROP;
A
Al Viro 已提交
3612
	__be16 type;
L
Linus Torvalds 已提交
3613

3614
	net_timestamp_check(!netdev_tstamp_prequeue, skb);
3615

3616
	trace_netif_receive_skb(skb);
3617

J
Joe Eykholt 已提交
3618
	orig_dev = skb->dev;
3619

3620
	skb_reset_network_header(skb);
3621 3622
	if (!skb_transport_header_was_set(skb))
		skb_reset_transport_header(skb);
3623
	skb_reset_mac_len(skb);
L
Linus Torvalds 已提交
3624 3625 3626 3627 3628

	pt_prev = NULL;

	rcu_read_lock();

3629
another_round:
3630
	skb->skb_iif = skb->dev->ifindex;
3631 3632 3633

	__this_cpu_inc(softnet_data.processed);

P
Patrick McHardy 已提交
3634 3635
	if (skb->protocol == cpu_to_be16(ETH_P_8021Q) ||
	    skb->protocol == cpu_to_be16(ETH_P_8021AD)) {
3636
		skb = skb_vlan_untag(skb);
3637
		if (unlikely(!skb))
3638
			goto unlock;
3639 3640
	}

L
Linus Torvalds 已提交
3641 3642 3643 3644 3645 3646 3647
#ifdef CONFIG_NET_CLS_ACT
	if (skb->tc_verd & TC_NCLS) {
		skb->tc_verd = CLR_TC_NCLS(skb->tc_verd);
		goto ncls;
	}
#endif

3648
	if (pfmemalloc)
3649 3650
		goto skip_taps;

L
Linus Torvalds 已提交
3651
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
3652
		if (!ptype->dev || ptype->dev == skb->dev) {
3653
			if (pt_prev)
D
David S. Miller 已提交
3654
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3655 3656 3657 3658
			pt_prev = ptype;
		}
	}

3659
skip_taps:
L
Linus Torvalds 已提交
3660
#ifdef CONFIG_NET_CLS_ACT
3661 3662
	skb = handle_ing(skb, &pt_prev, &ret, orig_dev);
	if (!skb)
3663
		goto unlock;
L
Linus Torvalds 已提交
3664 3665 3666
ncls:
#endif

3667
	if (pfmemalloc && !skb_pfmemalloc_protocol(skb))
3668 3669
		goto drop;

3670 3671 3672 3673 3674
	if (vlan_tx_tag_present(skb)) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3675
		if (vlan_do_receive(&skb))
3676 3677
			goto another_round;
		else if (unlikely(!skb))
3678
			goto unlock;
3679 3680
	}

3681
	rx_handler = rcu_dereference(skb->dev->rx_handler);
3682 3683 3684 3685 3686
	if (rx_handler) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3687 3688
		switch (rx_handler(&skb)) {
		case RX_HANDLER_CONSUMED:
3689
			ret = NET_RX_SUCCESS;
3690
			goto unlock;
3691
		case RX_HANDLER_ANOTHER:
3692
			goto another_round;
3693 3694 3695 3696 3697 3698 3699
		case RX_HANDLER_EXACT:
			deliver_exact = true;
		case RX_HANDLER_PASS:
			break;
		default:
			BUG();
		}
3700
	}
L
Linus Torvalds 已提交
3701

E
Eric Dumazet 已提交
3702 3703 3704 3705 3706 3707 3708 3709 3710
	if (unlikely(vlan_tx_tag_present(skb))) {
		if (vlan_tx_tag_get_id(skb))
			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;
	}
3711

3712
	/* deliver only exact match when indicated */
3713
	null_or_dev = deliver_exact ? skb->dev : NULL;
3714

L
Linus Torvalds 已提交
3715
	type = skb->protocol;
3716 3717
	list_for_each_entry_rcu(ptype,
			&ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) {
3718
		if (ptype->type == type &&
3719 3720
		    (ptype->dev == null_or_dev || ptype->dev == skb->dev ||
		     ptype->dev == orig_dev)) {
3721
			if (pt_prev)
D
David S. Miller 已提交
3722
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3723 3724 3725 3726 3727
			pt_prev = ptype;
		}
	}

	if (pt_prev) {
3728
		if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
3729
			goto drop;
3730 3731
		else
			ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3732
	} else {
3733
drop:
3734
		atomic_long_inc(&skb->dev->rx_dropped);
L
Linus Torvalds 已提交
3735 3736 3737 3738 3739 3740 3741
		kfree_skb(skb);
		/* Jamal, now you will not able to escape explaining
		 * me how you were going to use this. :-)
		 */
		ret = NET_RX_DROP;
	}

3742
unlock:
L
Linus Torvalds 已提交
3743
	rcu_read_unlock();
3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768
	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
3769 3770
	return ret;
}
T
Tom Herbert 已提交
3771

3772
static int netif_receive_skb_internal(struct sk_buff *skb)
T
Tom Herbert 已提交
3773
{
3774
	net_timestamp_check(netdev_tstamp_prequeue, skb);
E
Eric Dumazet 已提交
3775

3776 3777 3778
	if (skb_defer_rx_timestamp(skb))
		return NET_RX_SUCCESS;

E
Eric Dumazet 已提交
3779
#ifdef CONFIG_RPS
3780
	if (static_key_false(&rps_needed)) {
E
Eric Dumazet 已提交
3781 3782
		struct rps_dev_flow voidflow, *rflow = &voidflow;
		int cpu, ret;
T
Tom Herbert 已提交
3783

E
Eric Dumazet 已提交
3784 3785 3786
		rcu_read_lock();

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
T
Tom Herbert 已提交
3787

E
Eric Dumazet 已提交
3788 3789 3790
		if (cpu >= 0) {
			ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
			rcu_read_unlock();
3791
			return ret;
E
Eric Dumazet 已提交
3792
		}
3793
		rcu_read_unlock();
T
Tom Herbert 已提交
3794
	}
3795
#endif
3796
	return __netif_receive_skb(skb);
T
Tom Herbert 已提交
3797
}
3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819

/**
 *	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
 */
int netif_receive_skb(struct sk_buff *skb)
{
	trace_netif_receive_skb_entry(skb);

	return netif_receive_skb_internal(skb);
}
E
Eric Dumazet 已提交
3820
EXPORT_SYMBOL(netif_receive_skb);
L
Linus Torvalds 已提交
3821

E
Eric Dumazet 已提交
3822 3823 3824
/* Network device is going away, flush any packets still pending
 * Called with irqs disabled.
 */
3825
static void flush_backlog(void *arg)
3826
{
3827
	struct net_device *dev = arg;
3828
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
3829 3830
	struct sk_buff *skb, *tmp;

E
Eric Dumazet 已提交
3831
	rps_lock(sd);
3832
	skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
3833
		if (skb->dev == dev) {
E
Eric Dumazet 已提交
3834
			__skb_unlink(skb, &sd->input_pkt_queue);
3835
			kfree_skb(skb);
3836
			input_queue_head_incr(sd);
3837
		}
3838
	}
E
Eric Dumazet 已提交
3839
	rps_unlock(sd);
3840 3841 3842 3843 3844

	skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
		if (skb->dev == dev) {
			__skb_unlink(skb, &sd->process_queue);
			kfree_skb(skb);
3845
			input_queue_head_incr(sd);
3846 3847
		}
	}
3848 3849
}

3850 3851
static int napi_gro_complete(struct sk_buff *skb)
{
3852
	struct packet_offload *ptype;
3853
	__be16 type = skb->protocol;
3854
	struct list_head *head = &offload_base;
3855 3856
	int err = -ENOENT;

3857 3858
	BUILD_BUG_ON(sizeof(struct napi_gro_cb) > sizeof(skb->cb));

3859 3860
	if (NAPI_GRO_CB(skb)->count == 1) {
		skb_shinfo(skb)->gso_size = 0;
3861
		goto out;
3862
	}
3863 3864 3865

	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
3866
		if (ptype->type != type || !ptype->callbacks.gro_complete)
3867 3868
			continue;

3869
		err = ptype->callbacks.gro_complete(skb, 0);
3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880
		break;
	}
	rcu_read_unlock();

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

out:
3881
	return netif_receive_skb_internal(skb);
3882 3883
}

3884 3885 3886 3887 3888
/* 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)
3889
{
3890
	struct sk_buff *skb, *prev = NULL;
3891

3892 3893 3894 3895 3896 3897 3898
	/* 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) {
3899
		skb->next = NULL;
3900 3901 3902 3903 3904

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

		prev = skb->prev;
3905
		napi_gro_complete(skb);
3906
		napi->gro_count--;
3907 3908 3909 3910
	}

	napi->gro_list = NULL;
}
E
Eric Dumazet 已提交
3911
EXPORT_SYMBOL(napi_gro_flush);
3912

E
Eric Dumazet 已提交
3913 3914 3915 3916
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;
3917
	u32 hash = skb_get_hash_raw(skb);
E
Eric Dumazet 已提交
3918 3919 3920 3921

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

3922 3923 3924 3925 3926 3927 3928
		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
3929 3930 3931 3932
		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),
3933
						      skb_mac_header(skb));
E
Eric Dumazet 已提交
3934 3935
		else if (!diffs)
			diffs = memcmp(skb_mac_header(p),
3936
				       skb_mac_header(skb),
E
Eric Dumazet 已提交
3937 3938 3939 3940 3941
				       maclen);
		NAPI_GRO_CB(p)->same_flow = !diffs;
	}
}

3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955
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
3956 3957 3958
	}
}

3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979
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
3980
static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3981 3982
{
	struct sk_buff **pp = NULL;
3983
	struct packet_offload *ptype;
3984
	__be16 type = skb->protocol;
3985
	struct list_head *head = &offload_base;
H
Herbert Xu 已提交
3986
	int same_flow;
3987
	enum gro_result ret;
3988
	int grow;
3989

3990
	if (!(skb->dev->features & NETIF_F_GRO))
3991 3992
		goto normal;

3993
	if (skb_is_gso(skb) || skb_has_frag_list(skb) || skb->csum_bad)
3994 3995
		goto normal;

E
Eric Dumazet 已提交
3996 3997
	gro_list_prepare(napi, skb);

3998 3999
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
4000
		if (ptype->type != type || !ptype->callbacks.gro_receive)
4001 4002
			continue;

4003
		skb_set_network_header(skb, skb_gro_offset(skb));
4004
		skb_reset_mac_len(skb);
4005 4006
		NAPI_GRO_CB(skb)->same_flow = 0;
		NAPI_GRO_CB(skb)->flush = 0;
H
Herbert Xu 已提交
4007
		NAPI_GRO_CB(skb)->free = 0;
4008
		NAPI_GRO_CB(skb)->udp_mark = 0;
4009

4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024
		/* 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;
		}
4025

4026
		pp = ptype->callbacks.gro_receive(&napi->gro_list, skb);
4027 4028 4029 4030 4031 4032 4033
		break;
	}
	rcu_read_unlock();

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

H
Herbert Xu 已提交
4034
	same_flow = NAPI_GRO_CB(skb)->same_flow;
4035
	ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED;
H
Herbert Xu 已提交
4036

4037 4038 4039 4040 4041 4042
	if (pp) {
		struct sk_buff *nskb = *pp;

		*pp = nskb->next;
		nskb->next = NULL;
		napi_gro_complete(nskb);
4043
		napi->gro_count--;
4044 4045
	}

H
Herbert Xu 已提交
4046
	if (same_flow)
4047 4048
		goto ok;

4049
	if (NAPI_GRO_CB(skb)->flush)
4050 4051
		goto normal;

4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065
	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++;
	}
4066
	NAPI_GRO_CB(skb)->count = 1;
4067
	NAPI_GRO_CB(skb)->age = jiffies;
4068
	NAPI_GRO_CB(skb)->last = skb;
4069
	skb_shinfo(skb)->gso_size = skb_gro_len(skb);
4070 4071
	skb->next = napi->gro_list;
	napi->gro_list = skb;
4072
	ret = GRO_HELD;
4073

4074
pull:
4075 4076 4077
	grow = skb_gro_offset(skb) - skb_headlen(skb);
	if (grow > 0)
		gro_pull_from_frag0(skb, grow);
4078
ok:
4079
	return ret;
4080 4081

normal:
4082 4083
	ret = GRO_NORMAL;
	goto pull;
H
Herbert Xu 已提交
4084
}
4085

4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097
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;
}
4098
EXPORT_SYMBOL(gro_find_receive_by_type);
4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111

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;
}
4112
EXPORT_SYMBOL(gro_find_complete_by_type);
H
Herbert Xu 已提交
4113

R
Rami Rosen 已提交
4114
static gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
H
Herbert Xu 已提交
4115
{
4116 4117
	switch (ret) {
	case GRO_NORMAL:
4118
		if (netif_receive_skb_internal(skb))
4119 4120
			ret = GRO_DROP;
		break;
H
Herbert Xu 已提交
4121

4122
	case GRO_DROP:
H
Herbert Xu 已提交
4123 4124
		kfree_skb(skb);
		break;
4125

4126
	case GRO_MERGED_FREE:
4127 4128 4129 4130
		if (NAPI_GRO_CB(skb)->free == NAPI_GRO_FREE_STOLEN_HEAD)
			kmem_cache_free(skbuff_head_cache, skb);
		else
			__kfree_skb(skb);
4131 4132
		break;

4133 4134 4135
	case GRO_HELD:
	case GRO_MERGED:
		break;
H
Herbert Xu 已提交
4136 4137
	}

4138
	return ret;
4139 4140
}

4141
gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
4142
{
4143
	trace_napi_gro_receive_entry(skb);
4144

4145 4146
	skb_gro_reset_offset(skb);

E
Eric Dumazet 已提交
4147
	return napi_skb_finish(dev_gro_receive(napi, skb), skb);
4148 4149 4150
}
EXPORT_SYMBOL(napi_gro_receive);

S
stephen hemminger 已提交
4151
static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb)
4152
{
4153 4154 4155 4156
	if (unlikely(skb->pfmemalloc)) {
		consume_skb(skb);
		return;
	}
4157
	__skb_pull(skb, skb_headlen(skb));
4158 4159
	/* restore the reserve we had after netdev_alloc_skb_ip_align() */
	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN - skb_headroom(skb));
4160
	skb->vlan_tci = 0;
H
Herbert Xu 已提交
4161
	skb->dev = napi->dev;
A
Andy Gospodarek 已提交
4162
	skb->skb_iif = 0;
4163 4164
	skb->encapsulation = 0;
	skb_shinfo(skb)->gso_type = 0;
4165
	skb->truesize = SKB_TRUESIZE(skb_end_offset(skb));
4166 4167 4168 4169

	napi->skb = skb;
}

4170
struct sk_buff *napi_get_frags(struct napi_struct *napi)
H
Herbert Xu 已提交
4171 4172 4173 4174
{
	struct sk_buff *skb = napi->skb;

	if (!skb) {
4175
		skb = napi_alloc_skb(napi, GRO_MAX_HEAD);
E
Eric Dumazet 已提交
4176
		napi->skb = skb;
4177
	}
4178 4179
	return skb;
}
4180
EXPORT_SYMBOL(napi_get_frags);
4181

4182 4183 4184
static gro_result_t napi_frags_finish(struct napi_struct *napi,
				      struct sk_buff *skb,
				      gro_result_t ret)
4185
{
4186 4187
	switch (ret) {
	case GRO_NORMAL:
4188 4189 4190 4191
	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))
4192
			ret = GRO_DROP;
4193
		break;
H
Herbert Xu 已提交
4194

4195 4196 4197 4198
	case GRO_DROP:
	case GRO_MERGED_FREE:
		napi_reuse_skb(napi, skb);
		break;
4199 4200 4201

	case GRO_MERGED:
		break;
4202
	}
H
Herbert Xu 已提交
4203

4204
	return ret;
H
Herbert Xu 已提交
4205
}
4206

4207 4208 4209 4210
/* 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
4211
static struct sk_buff *napi_frags_skb(struct napi_struct *napi)
4212 4213
{
	struct sk_buff *skb = napi->skb;
4214 4215
	const struct ethhdr *eth;
	unsigned int hlen = sizeof(*eth);
4216 4217 4218

	napi->skb = NULL;

4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232
	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;
4233
	}
4234 4235 4236 4237 4238 4239 4240 4241
	__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;
4242 4243 4244 4245

	return skb;
}

4246
gro_result_t napi_gro_frags(struct napi_struct *napi)
4247
{
4248
	struct sk_buff *skb = napi_frags_skb(napi);
4249 4250

	if (!skb)
4251
		return GRO_DROP;
4252

4253 4254
	trace_napi_gro_frags_entry(skb);

E
Eric Dumazet 已提交
4255
	return napi_frags_finish(napi, skb, dev_gro_receive(napi, skb));
4256
}
H
Herbert Xu 已提交
4257 4258
EXPORT_SYMBOL(napi_gro_frags);

T
Tom Herbert 已提交
4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283
/* 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);

4284
/*
4285
 * net_rps_action_and_irq_enable sends any pending IPI's for rps.
4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302
 * 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))
4303
				smp_call_function_single_async(remsd->cpu,
4304
							   &remsd->csd);
4305 4306 4307 4308 4309 4310 4311
			remsd = next;
		}
	} else
#endif
		local_irq_enable();
}

E
Eric Dumazet 已提交
4312 4313 4314 4315 4316 4317 4318 4319 4320
static bool sd_has_rps_ipi_waiting(struct softnet_data *sd)
{
#ifdef CONFIG_RPS
	return sd->rps_ipi_list != NULL;
#else
	return false;
#endif
}

4321
static int process_backlog(struct napi_struct *napi, int quota)
L
Linus Torvalds 已提交
4322 4323
{
	int work = 0;
E
Eric Dumazet 已提交
4324
	struct softnet_data *sd = container_of(napi, struct softnet_data, backlog);
L
Linus Torvalds 已提交
4325

4326 4327 4328
	/* Check if we have pending ipi, its better to send them now,
	 * not waiting net_rx_action() end.
	 */
E
Eric Dumazet 已提交
4329
	if (sd_has_rps_ipi_waiting(sd)) {
4330 4331 4332
		local_irq_disable();
		net_rps_action_and_irq_enable(sd);
	}
E
Eric Dumazet 已提交
4333

4334
	napi->weight = weight_p;
4335
	local_irq_disable();
4336
	while (1) {
L
Linus Torvalds 已提交
4337
		struct sk_buff *skb;
4338 4339 4340 4341 4342

		while ((skb = __skb_dequeue(&sd->process_queue))) {
			local_irq_enable();
			__netif_receive_skb(skb);
			local_irq_disable();
4343 4344 4345 4346 4347
			input_queue_head_incr(sd);
			if (++work >= quota) {
				local_irq_enable();
				return work;
			}
4348
		}
L
Linus Torvalds 已提交
4349

E
Eric Dumazet 已提交
4350
		rps_lock(sd);
4351
		if (skb_queue_empty(&sd->input_pkt_queue)) {
E
Eric Dumazet 已提交
4352 4353 4354
			/*
			 * Inline a custom version of __napi_complete().
			 * only current cpu owns and manipulates this napi,
4355 4356 4357
			 * 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
4358 4359 4360
			 * and we dont need an smp_mb() memory barrier.
			 */
			napi->state = 0;
4361
			rps_unlock(sd);
E
Eric Dumazet 已提交
4362

4363
			break;
4364
		}
4365 4366 4367

		skb_queue_splice_tail_init(&sd->input_pkt_queue,
					   &sd->process_queue);
E
Eric Dumazet 已提交
4368
		rps_unlock(sd);
4369 4370
	}
	local_irq_enable();
L
Linus Torvalds 已提交
4371

4372 4373
	return work;
}
L
Linus Torvalds 已提交
4374

4375 4376
/**
 * __napi_schedule - schedule for receive
4377
 * @n: entry to schedule
4378
 *
4379 4380
 * The entry's receive function will be scheduled to run.
 * Consider using __napi_schedule_irqoff() if hard irqs are masked.
4381
 */
H
Harvey Harrison 已提交
4382
void __napi_schedule(struct napi_struct *n)
4383 4384
{
	unsigned long flags;
L
Linus Torvalds 已提交
4385

4386
	local_irq_save(flags);
4387
	____napi_schedule(this_cpu_ptr(&softnet_data), n);
4388
	local_irq_restore(flags);
L
Linus Torvalds 已提交
4389
}
4390 4391
EXPORT_SYMBOL(__napi_schedule);

4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403
/**
 * __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);

4404 4405 4406 4407
void __napi_complete(struct napi_struct *n)
{
	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));

E
Eric Dumazet 已提交
4408
	list_del_init(&n->poll_list);
4409
	smp_mb__before_atomic();
4410 4411 4412 4413
	clear_bit(NAPI_STATE_SCHED, &n->state);
}
EXPORT_SYMBOL(__napi_complete);

4414
void napi_complete_done(struct napi_struct *n, int work_done)
4415 4416 4417 4418 4419 4420 4421 4422 4423 4424
{
	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;

4425 4426
	if (n->gro_list) {
		unsigned long timeout = 0;
E
Eric Dumazet 已提交
4427

4428 4429 4430 4431 4432 4433 4434 4435 4436
		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
4437 4438 4439 4440 4441 4442 4443 4444
	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);
	}
4445
}
4446
EXPORT_SYMBOL(napi_complete_done);
4447

E
Eliezer Tamir 已提交
4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499
/* 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);

4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510
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;
}

4511 4512 4513 4514
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);
4515 4516
	hrtimer_init(&napi->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_PINNED);
	napi->timer.function = napi_watchdog;
4517
	napi->gro_count = 0;
4518
	napi->gro_list = NULL;
H
Herbert Xu 已提交
4519
	napi->skb = NULL;
4520
	napi->poll = poll;
E
Eric Dumazet 已提交
4521 4522 4523
	if (weight > NAPI_POLL_WEIGHT)
		pr_err_once("netif_napi_add() called with weight %d on device %s\n",
			    weight, dev->name);
4524 4525 4526
	napi->weight = weight;
	list_add(&napi->dev_list, &dev->napi_list);
	napi->dev = dev;
H
Herbert Xu 已提交
4527
#ifdef CONFIG_NETPOLL
4528 4529 4530 4531 4532 4533 4534
	spin_lock_init(&napi->poll_lock);
	napi->poll_owner = -1;
#endif
	set_bit(NAPI_STATE_SCHED, &napi->state);
}
EXPORT_SYMBOL(netif_napi_add);

4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548
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);

4549 4550
void netif_napi_del(struct napi_struct *napi)
{
4551
	list_del_init(&napi->dev_list);
4552
	napi_free_frags(napi);
4553

E
Eric Dumazet 已提交
4554
	kfree_skb_list(napi->gro_list);
4555
	napi->gro_list = NULL;
4556
	napi->gro_count = 0;
4557 4558 4559
}
EXPORT_SYMBOL(netif_napi_del);

4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604
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);
	}

4605 4606 4607 4608 4609 4610 4611 4612 4613
	/* 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;
	}

4614 4615 4616 4617 4618 4619 4620 4621
	list_add_tail(&n->poll_list, repoll);

out_unlock:
	netpoll_poll_unlock(have);

	return work;
}

L
Linus Torvalds 已提交
4622 4623
static void net_rx_action(struct softirq_action *h)
{
4624
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
4625
	unsigned long time_limit = jiffies + 2;
4626
	int budget = netdev_budget;
E
Eric Dumazet 已提交
4627 4628
	LIST_HEAD(list);
	LIST_HEAD(repoll);
4629

L
Linus Torvalds 已提交
4630
	local_irq_disable();
E
Eric Dumazet 已提交
4631 4632
	list_splice_init(&sd->poll_list, &list);
	local_irq_enable();
L
Linus Torvalds 已提交
4633

E
Eric Dumazet 已提交
4634
	while (!list_empty(&list)) {
4635
		struct napi_struct *n;
L
Linus Torvalds 已提交
4636

E
Eric Dumazet 已提交
4637
		/* If softirq window is exhausted then punt.
4638 4639
		 * Allow this to run for 2 jiffies since which will allow
		 * an average latency of 1.5/HZ.
4640
		 */
4641
		if (unlikely(budget <= 0 || time_after_eq(jiffies, time_limit)))
L
Linus Torvalds 已提交
4642 4643 4644
			goto softnet_break;


E
Eric Dumazet 已提交
4645
		n = list_first_entry(&list, struct napi_struct, poll_list);
4646
		budget -= napi_poll(n, &repoll);
L
Linus Torvalds 已提交
4647
	}
E
Eric Dumazet 已提交
4648 4649 4650 4651 4652

	if (!sd_has_rps_ipi_waiting(sd) &&
	    list_empty(&list) &&
	    list_empty(&repoll))
		return;
L
Linus Torvalds 已提交
4653
out:
E
Eric Dumazet 已提交
4654 4655 4656 4657 4658 4659 4660 4661
	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);

4662
	net_rps_action_and_irq_enable(sd);
T
Tom Herbert 已提交
4663

L
Linus Torvalds 已提交
4664 4665 4666
	return;

softnet_break:
C
Changli Gao 已提交
4667
	sd->time_squeeze++;
L
Linus Torvalds 已提交
4668 4669 4670
	goto out;
}

4671
struct netdev_adjacent {
J
Jiri Pirko 已提交
4672
	struct net_device *dev;
4673 4674

	/* upper master flag, there can only be one master device per list */
J
Jiri Pirko 已提交
4675
	bool master;
4676 4677 4678 4679

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

4680 4681 4682
	/* private field for the users */
	void *private;

J
Jiri Pirko 已提交
4683 4684 4685 4686
	struct list_head list;
	struct rcu_head rcu;
};

4687 4688
static struct netdev_adjacent *__netdev_find_adj(struct net_device *dev,
						 struct net_device *adj_dev,
4689
						 struct list_head *adj_list)
J
Jiri Pirko 已提交
4690
{
4691 4692
	struct netdev_adjacent *adj;

4693
	list_for_each_entry(adj, adj_list, list) {
4694 4695
		if (adj->dev == adj_dev)
			return adj;
J
Jiri Pirko 已提交
4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713
	}
	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();

4714
	return __netdev_find_adj(dev, upper_dev, &dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4715 4716 4717 4718 4719 4720 4721 4722 4723 4724
}
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
4725
static bool netdev_has_any_upper_dev(struct net_device *dev)
J
Jiri Pirko 已提交
4726 4727 4728
{
	ASSERT_RTNL();

4729
	return !list_empty(&dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740
}

/**
 * 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)
{
4741
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4742 4743 4744

	ASSERT_RTNL();

4745
	if (list_empty(&dev->adj_list.upper))
J
Jiri Pirko 已提交
4746 4747
		return NULL;

4748
	upper = list_first_entry(&dev->adj_list.upper,
4749
				 struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4750 4751 4752 4753 4754 4755
	if (likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get);

4756 4757 4758 4759 4760 4761 4762 4763 4764 4765
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);

4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791
/**
 * 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);

4792 4793
/**
 * netdev_all_upper_get_next_dev_rcu - Get the next dev from upper list
4794 4795 4796 4797 4798 4799
 * @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.
 */
4800 4801
struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
						     struct list_head **iter)
4802 4803 4804
{
	struct netdev_adjacent *upper;

4805
	WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_rtnl_is_held());
4806 4807 4808

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

4809
	if (&upper->list == &dev->all_adj_list.upper)
4810 4811 4812 4813 4814 4815
		return NULL;

	*iter = &upper->list;

	return upper->dev;
}
4816
EXPORT_SYMBOL(netdev_all_upper_get_next_dev_rcu);
4817

4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838
/**
 * 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
 * list will remain unchainged.
 */
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;

4839
	*iter = lower->list.next;
4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866

	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;

4867
	*iter = &lower->list;
4868 4869 4870 4871 4872

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

4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898
/**
 * 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
 * list will remain unchainged.
 */
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);

4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919
/**
 * 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
4920 4921 4922 4923 4924 4925 4926 4927 4928
/**
 * 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)
{
4929
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4930

4931
	upper = list_first_or_null_rcu(&dev->adj_list.upper,
4932
				       struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4933 4934 4935 4936 4937 4938
	if (upper && likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get_rcu);

4939
static int netdev_adjacent_sysfs_add(struct net_device *dev,
4940 4941 4942 4943 4944 4945 4946 4947 4948
			      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);
}
4949
static void netdev_adjacent_sysfs_del(struct net_device *dev,
4950 4951 4952 4953 4954 4955 4956 4957 4958
			       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);
}

4959 4960 4961 4962 4963 4964 4965 4966
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));
}
4967

4968 4969
static int __netdev_adjacent_dev_insert(struct net_device *dev,
					struct net_device *adj_dev,
4970
					struct list_head *dev_list,
4971
					void *private, bool master)
4972 4973
{
	struct netdev_adjacent *adj;
4974
	int ret;
4975

4976
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989

	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;
4990
	adj->private = private;
4991
	dev_hold(adj_dev);
4992 4993 4994

	pr_debug("dev_hold for %s, because of link added from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
4995

4996
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list)) {
4997
		ret = netdev_adjacent_sysfs_add(dev, adj_dev, dev_list);
4998 4999 5000 5001
		if (ret)
			goto free_adj;
	}

5002
	/* Ensure that master link is always the first item in list. */
5003 5004 5005 5006
	if (master) {
		ret = sysfs_create_link(&(dev->dev.kobj),
					&(adj_dev->dev.kobj), "master");
		if (ret)
5007
			goto remove_symlinks;
5008

5009
		list_add_rcu(&adj->list, dev_list);
5010
	} else {
5011
		list_add_tail_rcu(&adj->list, dev_list);
5012
	}
5013 5014

	return 0;
5015

5016
remove_symlinks:
5017
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
5018
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
5019 5020
free_adj:
	kfree(adj);
5021
	dev_put(adj_dev);
5022 5023

	return ret;
5024 5025
}

S
stephen hemminger 已提交
5026 5027 5028
static void __netdev_adjacent_dev_remove(struct net_device *dev,
					 struct net_device *adj_dev,
					 struct list_head *dev_list)
5029 5030 5031
{
	struct netdev_adjacent *adj;

5032
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
5033

5034 5035 5036
	if (!adj) {
		pr_err("tried to remove device %s from %s\n",
		       dev->name, adj_dev->name);
5037
		BUG();
5038
	}
5039 5040

	if (adj->ref_nr > 1) {
5041 5042
		pr_debug("%s to %s ref_nr-- = %d\n", dev->name, adj_dev->name,
			 adj->ref_nr-1);
5043 5044 5045 5046
		adj->ref_nr--;
		return;
	}

5047 5048 5049
	if (adj->master)
		sysfs_remove_link(&(dev->dev.kobj), "master");

5050
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
5051
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
5052

5053
	list_del_rcu(&adj->list);
5054 5055
	pr_debug("dev_put for %s, because link removed from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
5056 5057 5058 5059
	dev_put(adj_dev);
	kfree_rcu(adj, rcu);
}

S
stephen hemminger 已提交
5060 5061 5062 5063 5064
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)
5065 5066 5067
{
	int ret;

5068 5069
	ret = __netdev_adjacent_dev_insert(dev, upper_dev, up_list, private,
					   master);
5070 5071 5072
	if (ret)
		return ret;

5073 5074
	ret = __netdev_adjacent_dev_insert(upper_dev, dev, down_list, private,
					   false);
5075
	if (ret) {
5076
		__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
5077 5078 5079 5080 5081 5082
		return ret;
	}

	return 0;
}

S
stephen hemminger 已提交
5083 5084
static int __netdev_adjacent_dev_link(struct net_device *dev,
				      struct net_device *upper_dev)
5085
{
5086 5087 5088
	return __netdev_adjacent_dev_link_lists(dev, upper_dev,
						&dev->all_adj_list.upper,
						&upper_dev->all_adj_list.lower,
5089
						NULL, false);
5090 5091
}

S
stephen hemminger 已提交
5092 5093 5094 5095
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)
5096
{
5097 5098
	__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
	__netdev_adjacent_dev_remove(upper_dev, dev, down_list);
5099 5100
}

S
stephen hemminger 已提交
5101 5102
static void __netdev_adjacent_dev_unlink(struct net_device *dev,
					 struct net_device *upper_dev)
5103
{
5104 5105 5106 5107 5108
	__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
5109 5110 5111
static int __netdev_adjacent_dev_link_neighbour(struct net_device *dev,
						struct net_device *upper_dev,
						void *private, bool master)
5112 5113 5114 5115 5116 5117 5118 5119 5120
{
	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,
5121
					       private, master);
5122 5123 5124 5125 5126 5127
	if (ret) {
		__netdev_adjacent_dev_unlink(dev, upper_dev);
		return ret;
	}

	return 0;
5128 5129
}

S
stephen hemminger 已提交
5130 5131
static void __netdev_adjacent_dev_unlink_neighbour(struct net_device *dev,
						   struct net_device *upper_dev)
5132 5133 5134 5135 5136 5137
{
	__netdev_adjacent_dev_unlink(dev, upper_dev);
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev,
					   &dev->adj_list.upper,
					   &upper_dev->adj_list.lower);
}
5138

J
Jiri Pirko 已提交
5139
static int __netdev_upper_dev_link(struct net_device *dev,
5140 5141
				   struct net_device *upper_dev, bool master,
				   void *private)
J
Jiri Pirko 已提交
5142
{
5143 5144
	struct netdev_adjacent *i, *j, *to_i, *to_j;
	int ret = 0;
J
Jiri Pirko 已提交
5145 5146 5147 5148 5149 5150 5151

	ASSERT_RTNL();

	if (dev == upper_dev)
		return -EBUSY;

	/* To prevent loops, check if dev is not upper device to upper_dev. */
5152
	if (__netdev_find_adj(upper_dev, dev, &upper_dev->all_adj_list.upper))
J
Jiri Pirko 已提交
5153 5154
		return -EBUSY;

5155
	if (__netdev_find_adj(dev, upper_dev, &dev->all_adj_list.upper))
J
Jiri Pirko 已提交
5156 5157 5158 5159 5160
		return -EEXIST;

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

5161 5162
	ret = __netdev_adjacent_dev_link_neighbour(dev, upper_dev, private,
						   master);
5163 5164
	if (ret)
		return ret;
J
Jiri Pirko 已提交
5165

5166
	/* Now that we linked these devs, make all the upper_dev's
5167
	 * all_adj_list.upper visible to every dev's all_adj_list.lower an
5168 5169 5170
	 * versa, and don't forget the devices itself. All of these
	 * links are non-neighbours.
	 */
5171 5172 5173 5174
	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);
5175 5176 5177 5178 5179 5180 5181
			ret = __netdev_adjacent_dev_link(i->dev, j->dev);
			if (ret)
				goto rollback_mesh;
		}
	}

	/* add dev to every upper_dev's upper device */
5182 5183 5184
	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);
5185 5186 5187 5188 5189 5190
		ret = __netdev_adjacent_dev_link(dev, i->dev);
		if (ret)
			goto rollback_upper_mesh;
	}

	/* add upper_dev to every dev's lower device */
5191 5192 5193
	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);
5194 5195 5196 5197
		ret = __netdev_adjacent_dev_link(i->dev, upper_dev);
		if (ret)
			goto rollback_lower_mesh;
	}
J
Jiri Pirko 已提交
5198

5199
	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
J
Jiri Pirko 已提交
5200
	return 0;
5201 5202 5203

rollback_lower_mesh:
	to_i = i;
5204
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
5205 5206 5207 5208 5209 5210 5211 5212 5213
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);
	}

	i = NULL;

rollback_upper_mesh:
	to_i = i;
5214
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list) {
5215 5216 5217 5218 5219 5220 5221 5222 5223 5224
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(dev, i->dev);
	}

	i = j = NULL;

rollback_mesh:
	to_i = i;
	to_j = j;
5225 5226
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list) {
5227 5228 5229 5230 5231 5232 5233 5234
			if (i == to_i && j == to_j)
				break;
			__netdev_adjacent_dev_unlink(i->dev, j->dev);
		}
		if (i == to_i)
			break;
	}

5235
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
5236 5237

	return ret;
J
Jiri Pirko 已提交
5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252
}

/**
 * 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)
{
5253
	return __netdev_upper_dev_link(dev, upper_dev, false, NULL);
J
Jiri Pirko 已提交
5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270
}
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)
{
5271
	return __netdev_upper_dev_link(dev, upper_dev, true, NULL);
J
Jiri Pirko 已提交
5272 5273 5274
}
EXPORT_SYMBOL(netdev_master_upper_dev_link);

5275 5276 5277 5278 5279 5280 5281 5282
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
5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293
/**
 * 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)
{
5294
	struct netdev_adjacent *i, *j;
J
Jiri Pirko 已提交
5295 5296
	ASSERT_RTNL();

5297
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
5298 5299 5300 5301 5302

	/* 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.
	 */
5303 5304
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list)
5305 5306 5307 5308 5309
			__netdev_adjacent_dev_unlink(i->dev, j->dev);

	/* remove also the devices itself from lower/upper device
	 * list
	 */
5310
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
5311 5312
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);

5313
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list)
5314 5315
		__netdev_adjacent_dev_unlink(dev, i->dev);

5316
	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
J
Jiri Pirko 已提交
5317 5318 5319
}
EXPORT_SYMBOL(netdev_upper_dev_unlink);

5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369
void netdev_adjacent_add_links(struct net_device *dev)
{
	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);
	}
}

void netdev_adjacent_del_links(struct net_device *dev)
{
	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);
	}
}

5370
void netdev_adjacent_rename_links(struct net_device *dev, char *oldname)
5371
{
5372
	struct netdev_adjacent *iter;
5373

5374 5375
	struct net *net = dev_net(dev);

5376
	list_for_each_entry(iter, &dev->adj_list.upper, list) {
5377 5378
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
5379 5380 5381 5382 5383
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.lower);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.lower);
	}
5384

5385
	list_for_each_entry(iter, &dev->adj_list.lower, list) {
5386 5387
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
5388 5389 5390 5391 5392
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.upper);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.upper);
	}
5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409
}

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);

5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433

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);

5434 5435
static void dev_change_rx_flags(struct net_device *dev, int flags)
{
5436 5437
	const struct net_device_ops *ops = dev->netdev_ops;

5438
	if (ops->ndo_change_rx_flags)
5439
		ops->ndo_change_rx_flags(dev, flags);
5440 5441
}

5442
static int __dev_set_promiscuity(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5443
{
5444
	unsigned int old_flags = dev->flags;
5445 5446
	kuid_t uid;
	kgid_t gid;
L
Linus Torvalds 已提交
5447

5448 5449
	ASSERT_RTNL();

5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460
	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;
5461 5462
			pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n",
				dev->name);
5463 5464 5465
			return -EOVERFLOW;
		}
	}
5466
	if (dev->flags != old_flags) {
5467 5468 5469
		pr_info("device %s %s promiscuous mode\n",
			dev->name,
			dev->flags & IFF_PROMISC ? "entered" : "left");
5470 5471
		if (audit_enabled) {
			current_uid_gid(&uid, &gid);
5472 5473 5474 5475 5476
			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),
5477
				from_kuid(&init_user_ns, audit_get_loginuid(current)),
5478 5479
				from_kuid(&init_user_ns, uid),
				from_kgid(&init_user_ns, gid),
5480
				audit_get_sessionid(current));
5481
		}
5482

5483
		dev_change_rx_flags(dev, IFF_PROMISC);
L
Linus Torvalds 已提交
5484
	}
5485 5486
	if (notify)
		__dev_notify_flags(dev, old_flags, IFF_PROMISC);
5487
	return 0;
L
Linus Torvalds 已提交
5488 5489
}

5490 5491 5492 5493 5494 5495 5496 5497 5498
/**
 *	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.
5499
 *	Return 0 if successful or a negative errno code on error.
5500
 */
5501
int dev_set_promiscuity(struct net_device *dev, int inc)
5502
{
5503
	unsigned int old_flags = dev->flags;
5504
	int err;
5505

5506
	err = __dev_set_promiscuity(dev, inc, true);
5507
	if (err < 0)
5508
		return err;
5509 5510
	if (dev->flags != old_flags)
		dev_set_rx_mode(dev);
5511
	return err;
5512
}
E
Eric Dumazet 已提交
5513
EXPORT_SYMBOL(dev_set_promiscuity);
5514

5515
static int __dev_set_allmulti(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5516
{
5517
	unsigned int old_flags = dev->flags, old_gflags = dev->gflags;
L
Linus Torvalds 已提交
5518

5519 5520
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5521
	dev->flags |= IFF_ALLMULTI;
5522 5523 5524 5525 5526 5527 5528 5529 5530 5531
	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;
5532 5533
			pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n",
				dev->name);
5534 5535 5536
			return -EOVERFLOW;
		}
	}
5537
	if (dev->flags ^ old_flags) {
5538
		dev_change_rx_flags(dev, IFF_ALLMULTI);
5539
		dev_set_rx_mode(dev);
5540 5541 5542
		if (notify)
			__dev_notify_flags(dev, old_flags,
					   dev->gflags ^ old_gflags);
5543
	}
5544
	return 0;
5545
}
5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563

/**
 *	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
5564
EXPORT_SYMBOL(dev_set_allmulti);
5565 5566 5567 5568

/*
 *	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
5569
 *	filtering it is put in promiscuous mode while unicast addresses
5570 5571 5572 5573
 *	are present.
 */
void __dev_set_rx_mode(struct net_device *dev)
{
5574 5575
	const struct net_device_ops *ops = dev->netdev_ops;

5576 5577 5578 5579 5580
	/* dev_open will call this function so the list will stay sane. */
	if (!(dev->flags&IFF_UP))
		return;

	if (!netif_device_present(dev))
5581
		return;
5582

5583
	if (!(dev->priv_flags & IFF_UNICAST_FLT)) {
5584 5585 5586
		/* Unicast addresses changes may only happen under the rtnl,
		 * therefore calling __dev_set_promiscuity here is safe.
		 */
5587
		if (!netdev_uc_empty(dev) && !dev->uc_promisc) {
5588
			__dev_set_promiscuity(dev, 1, false);
5589
			dev->uc_promisc = true;
5590
		} else if (netdev_uc_empty(dev) && dev->uc_promisc) {
5591
			__dev_set_promiscuity(dev, -1, false);
5592
			dev->uc_promisc = false;
5593 5594
		}
	}
5595 5596 5597

	if (ops->ndo_set_rx_mode)
		ops->ndo_set_rx_mode(dev);
5598 5599 5600 5601
}

void dev_set_rx_mode(struct net_device *dev)
{
5602
	netif_addr_lock_bh(dev);
5603
	__dev_set_rx_mode(dev);
5604
	netif_addr_unlock_bh(dev);
L
Linus Torvalds 已提交
5605 5606
}

5607 5608 5609 5610 5611 5612
/**
 *	dev_get_flags - get flags reported to userspace
 *	@dev: device
 *
 *	Get the combination of flag bits exported through APIs to userspace.
 */
5613
unsigned int dev_get_flags(const struct net_device *dev)
L
Linus Torvalds 已提交
5614
{
5615
	unsigned int flags;
L
Linus Torvalds 已提交
5616 5617 5618

	flags = (dev->flags & ~(IFF_PROMISC |
				IFF_ALLMULTI |
S
Stefan Rompf 已提交
5619 5620 5621
				IFF_RUNNING |
				IFF_LOWER_UP |
				IFF_DORMANT)) |
L
Linus Torvalds 已提交
5622 5623 5624
		(dev->gflags & (IFF_PROMISC |
				IFF_ALLMULTI));

S
Stefan Rompf 已提交
5625 5626 5627 5628 5629 5630 5631 5632
	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
5633 5634 5635

	return flags;
}
E
Eric Dumazet 已提交
5636
EXPORT_SYMBOL(dev_get_flags);
L
Linus Torvalds 已提交
5637

5638
int __dev_change_flags(struct net_device *dev, unsigned int flags)
L
Linus Torvalds 已提交
5639
{
5640
	unsigned int old_flags = dev->flags;
5641
	int ret;
L
Linus Torvalds 已提交
5642

5643 5644
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658
	/*
	 *	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.
	 */

5659 5660
	if ((old_flags ^ flags) & IFF_MULTICAST)
		dev_change_rx_flags(dev, IFF_MULTICAST);
5661

5662
	dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5663 5664 5665 5666 5667 5668 5669 5670

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

	ret = 0;
5671
	if ((old_flags ^ flags) & IFF_UP)
5672
		ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev);
L
Linus Torvalds 已提交
5673 5674

	if ((flags ^ dev->gflags) & IFF_PROMISC) {
E
Eric Dumazet 已提交
5675
		int inc = (flags & IFF_PROMISC) ? 1 : -1;
5676
		unsigned int old_flags = dev->flags;
E
Eric Dumazet 已提交
5677

L
Linus Torvalds 已提交
5678
		dev->gflags ^= IFF_PROMISC;
5679 5680 5681 5682

		if (__dev_set_promiscuity(dev, inc, false) >= 0)
			if (dev->flags != old_flags)
				dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5683 5684 5685 5686 5687 5688 5689
	}

	/* 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
5690 5691
		int inc = (flags & IFF_ALLMULTI) ? 1 : -1;

L
Linus Torvalds 已提交
5692
		dev->gflags ^= IFF_ALLMULTI;
5693
		__dev_set_allmulti(dev, inc, false);
L
Linus Torvalds 已提交
5694 5695
	}

5696 5697 5698
	return ret;
}

5699 5700
void __dev_notify_flags(struct net_device *dev, unsigned int old_flags,
			unsigned int gchanges)
5701 5702 5703
{
	unsigned int changes = dev->flags ^ old_flags;

5704
	if (gchanges)
5705
		rtmsg_ifinfo(RTM_NEWLINK, dev, gchanges, GFP_ATOMIC);
5706

5707 5708 5709 5710 5711 5712 5713 5714
	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 &&
5715 5716 5717 5718 5719 5720 5721
	    (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);
	}
5722 5723 5724 5725 5726 5727 5728 5729 5730 5731
}

/**
 *	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.
 */
5732
int dev_change_flags(struct net_device *dev, unsigned int flags)
5733
{
5734
	int ret;
5735
	unsigned int changes, old_flags = dev->flags, old_gflags = dev->gflags;
5736 5737 5738 5739 5740

	ret = __dev_change_flags(dev, flags);
	if (ret < 0)
		return ret;

5741
	changes = (old_flags ^ dev->flags) | (old_gflags ^ dev->gflags);
5742
	__dev_notify_flags(dev, old_flags, changes);
L
Linus Torvalds 已提交
5743 5744
	return ret;
}
E
Eric Dumazet 已提交
5745
EXPORT_SYMBOL(dev_change_flags);
L
Linus Torvalds 已提交
5746

5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757
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;
}

5758 5759 5760 5761 5762 5763 5764
/**
 *	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
5765 5766
int dev_set_mtu(struct net_device *dev, int new_mtu)
{
5767
	int err, orig_mtu;
L
Linus Torvalds 已提交
5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778

	if (new_mtu == dev->mtu)
		return 0;

	/*	MTU must be positive.	 */
	if (new_mtu < 0)
		return -EINVAL;

	if (!netif_device_present(dev))
		return -ENODEV;

5779 5780 5781 5782
	err = call_netdevice_notifiers(NETDEV_PRECHANGEMTU, dev);
	err = notifier_to_errno(err);
	if (err)
		return err;
5783

5784 5785
	orig_mtu = dev->mtu;
	err = __dev_set_mtu(dev, new_mtu);
5786

5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797
	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
5798 5799
	return err;
}
E
Eric Dumazet 已提交
5800
EXPORT_SYMBOL(dev_set_mtu);
L
Linus Torvalds 已提交
5801

5802 5803 5804 5805 5806 5807 5808 5809 5810 5811 5812
/**
 *	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);

5813 5814 5815 5816 5817 5818 5819
/**
 *	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
5820 5821
int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa)
{
5822
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
5823 5824
	int err;

5825
	if (!ops->ndo_set_mac_address)
L
Linus Torvalds 已提交
5826 5827 5828 5829 5830
		return -EOPNOTSUPP;
	if (sa->sa_family != dev->type)
		return -EINVAL;
	if (!netif_device_present(dev))
		return -ENODEV;
5831
	err = ops->ndo_set_mac_address(dev, sa);
5832 5833
	if (err)
		return err;
J
Jiri Pirko 已提交
5834
	dev->addr_assign_type = NET_ADDR_SET;
5835
	call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
5836
	add_device_randomness(dev->dev_addr, dev->addr_len);
5837
	return 0;
L
Linus Torvalds 已提交
5838
}
E
Eric Dumazet 已提交
5839
EXPORT_SYMBOL(dev_set_mac_address);
L
Linus Torvalds 已提交
5840

J
Jiri Pirko 已提交
5841 5842 5843
/**
 *	dev_change_carrier - Change device carrier
 *	@dev: device
5844
 *	@new_carrier: new value
J
Jiri Pirko 已提交
5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859
 *
 *	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);

5860 5861 5862 5863 5864 5865 5866 5867
/**
 *	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,
5868
			 struct netdev_phys_item_id *ppid)
5869 5870 5871 5872 5873 5874 5875 5876 5877
{
	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);

L
Linus Torvalds 已提交
5878 5879
/**
 *	dev_new_index	-	allocate an ifindex
5880
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
5881 5882 5883 5884 5885
 *
 *	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.
 */
5886
static int dev_new_index(struct net *net)
L
Linus Torvalds 已提交
5887
{
5888
	int ifindex = net->ifindex;
L
Linus Torvalds 已提交
5889 5890 5891
	for (;;) {
		if (++ifindex <= 0)
			ifindex = 1;
5892
		if (!__dev_get_by_index(net, ifindex))
5893
			return net->ifindex = ifindex;
L
Linus Torvalds 已提交
5894 5895 5896 5897
	}
}

/* Delayed registration/unregisteration */
5898
static LIST_HEAD(net_todo_list);
5899
DECLARE_WAIT_QUEUE_HEAD(netdev_unregistering_wq);
L
Linus Torvalds 已提交
5900

5901
static void net_set_todo(struct net_device *dev)
L
Linus Torvalds 已提交
5902 5903
{
	list_add_tail(&dev->todo_list, &net_todo_list);
5904
	dev_net(dev)->dev_unreg_count++;
L
Linus Torvalds 已提交
5905 5906
}

5907
static void rollback_registered_many(struct list_head *head)
5908
{
5909
	struct net_device *dev, *tmp;
5910
	LIST_HEAD(close_head);
5911

5912 5913 5914
	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

5915
	list_for_each_entry_safe(dev, tmp, head, unreg_list) {
5916
		/* Some devices call without registering
5917 5918
		 * for initialization unwind. Remove those
		 * devices and proceed with the remaining.
5919 5920
		 */
		if (dev->reg_state == NETREG_UNINITIALIZED) {
5921 5922
			pr_debug("unregister_netdevice: device %s/%p never was registered\n",
				 dev->name, dev);
5923

5924
			WARN_ON(1);
5925 5926
			list_del(&dev->unreg_list);
			continue;
5927
		}
5928
		dev->dismantle = true;
5929
		BUG_ON(dev->reg_state != NETREG_REGISTERED);
5930
	}
5931

5932
	/* If device is running, close it first. */
5933 5934 5935
	list_for_each_entry(dev, head, unreg_list)
		list_add_tail(&dev->close_list, &close_head);
	dev_close_many(&close_head);
5936

5937
	list_for_each_entry(dev, head, unreg_list) {
5938 5939
		/* And unlink it from device chain. */
		unlist_netdevice(dev);
5940

5941 5942
		dev->reg_state = NETREG_UNREGISTERING;
	}
5943 5944 5945

	synchronize_net();

5946
	list_for_each_entry(dev, head, unreg_list) {
5947 5948
		struct sk_buff *skb = NULL;

5949 5950
		/* Shutdown queueing discipline. */
		dev_shutdown(dev);
5951 5952


5953 5954 5955 5956
		/* Notify protocols, that we are about to destroy
		   this device. They should clean all the things.
		*/
		call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
5957

5958 5959 5960 5961 5962
		if (!dev->rtnl_link_ops ||
		    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
			skb = rtmsg_ifinfo_build_skb(RTM_DELLINK, dev, ~0U,
						     GFP_KERNEL);

5963 5964 5965
		/*
		 *	Flush the unicast and multicast chains
		 */
5966
		dev_uc_flush(dev);
5967
		dev_mc_flush(dev);
5968

5969 5970
		if (dev->netdev_ops->ndo_uninit)
			dev->netdev_ops->ndo_uninit(dev);
5971

5972 5973
		if (skb)
			rtmsg_ifinfo_send(skb, dev, GFP_KERNEL);
5974

J
Jiri Pirko 已提交
5975 5976
		/* Notifier chain MUST detach us all upper devices. */
		WARN_ON(netdev_has_any_upper_dev(dev));
5977

5978 5979
		/* Remove entries from kobject tree */
		netdev_unregister_kobject(dev);
5980 5981 5982 5983
#ifdef CONFIG_XPS
		/* Remove XPS queueing entries */
		netif_reset_xps_queues_gt(dev, 0);
#endif
5984
	}
5985

5986
	synchronize_net();
5987

5988
	list_for_each_entry(dev, head, unreg_list)
5989 5990 5991 5992 5993 5994 5995 5996 5997
		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
5998
	list_del(&single);
5999 6000
}

6001 6002
static netdev_features_t netdev_fix_features(struct net_device *dev,
	netdev_features_t features)
6003
{
6004 6005 6006
	/* Fix illegal checksum combinations */
	if ((features & NETIF_F_HW_CSUM) &&
	    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
6007
		netdev_warn(dev, "mixed HW and IP checksum settings.\n");
6008 6009 6010
		features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
	}

6011
	/* TSO requires that SG is present as well. */
6012
	if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
6013
		netdev_dbg(dev, "Dropping TSO features since no SG feature.\n");
6014
		features &= ~NETIF_F_ALL_TSO;
6015 6016
	}

6017 6018 6019 6020 6021 6022 6023 6024 6025 6026 6027 6028 6029
	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;
	}

6030 6031 6032 6033
	/* TSO ECN requires that TSO is present as well. */
	if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN)
		features &= ~NETIF_F_TSO_ECN;

6034 6035
	/* Software GSO depends on SG. */
	if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) {
6036
		netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n");
6037 6038 6039
		features &= ~NETIF_F_GSO;
	}

6040
	/* UFO needs SG and checksumming */
6041
	if (features & NETIF_F_UFO) {
6042 6043 6044 6045
		/* 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))) {
6046
			netdev_dbg(dev,
6047
				"Dropping NETIF_F_UFO since no checksum offload features.\n");
6048 6049 6050 6051
			features &= ~NETIF_F_UFO;
		}

		if (!(features & NETIF_F_SG)) {
6052
			netdev_dbg(dev,
6053
				"Dropping NETIF_F_UFO since no NETIF_F_SG feature.\n");
6054 6055 6056 6057
			features &= ~NETIF_F_UFO;
		}
	}

J
Jiri Pirko 已提交
6058 6059 6060 6061 6062 6063 6064
#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;

6065 6066 6067
	return features;
}

6068
int __netdev_update_features(struct net_device *dev)
6069
{
6070
	netdev_features_t features;
6071 6072
	int err = 0;

6073 6074
	ASSERT_RTNL();

6075 6076 6077 6078 6079 6080 6081 6082 6083
	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)
6084
		return 0;
6085

6086 6087
	netdev_dbg(dev, "Features changed: %pNF -> %pNF\n",
		&dev->features, &features);
6088 6089 6090 6091

	if (dev->netdev_ops->ndo_set_features)
		err = dev->netdev_ops->ndo_set_features(dev, features);

6092
	if (unlikely(err < 0)) {
6093
		netdev_err(dev,
6094 6095
			"set_features() failed (%d); wanted %pNF, left %pNF\n",
			err, &features, &dev->features);
6096 6097 6098 6099 6100 6101 6102 6103 6104
		return -1;
	}

	if (!err)
		dev->features = features;

	return 1;
}

6105 6106 6107 6108 6109 6110 6111 6112
/**
 *	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.
 */
6113 6114 6115 6116
void netdev_update_features(struct net_device *dev)
{
	if (__netdev_update_features(dev))
		netdev_features_change(dev);
6117 6118 6119
}
EXPORT_SYMBOL(netdev_update_features);

6120 6121 6122 6123 6124 6125 6126 6127 6128 6129 6130 6131 6132 6133 6134 6135 6136
/**
 *	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);

6137 6138 6139 6140 6141 6142 6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163
/**
 *	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);

6164
#ifdef CONFIG_SYSFS
6165 6166 6167
static int netif_alloc_rx_queues(struct net_device *dev)
{
	unsigned int i, count = dev->num_rx_queues;
T
Tom Herbert 已提交
6168
	struct netdev_rx_queue *rx;
6169

T
Tom Herbert 已提交
6170
	BUG_ON(count < 1);
6171

T
Tom Herbert 已提交
6172
	rx = kcalloc(count, sizeof(struct netdev_rx_queue), GFP_KERNEL);
6173
	if (!rx)
T
Tom Herbert 已提交
6174
		return -ENOMEM;
6175

T
Tom Herbert 已提交
6176 6177 6178
	dev->_rx = rx;

	for (i = 0; i < count; i++)
T
Tom Herbert 已提交
6179
		rx[i].dev = dev;
6180 6181
	return 0;
}
T
Tom Herbert 已提交
6182
#endif
6183

C
Changli Gao 已提交
6184 6185 6186 6187 6188 6189 6190
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;
6191
	netdev_queue_numa_node_write(queue, NUMA_NO_NODE);
C
Changli Gao 已提交
6192
	queue->dev = dev;
T
Tom Herbert 已提交
6193 6194 6195
#ifdef CONFIG_BQL
	dql_init(&queue->dql, HZ);
#endif
C
Changli Gao 已提交
6196 6197
}

6198 6199
static void netif_free_tx_queues(struct net_device *dev)
{
W
WANG Cong 已提交
6200
	kvfree(dev->_tx);
6201 6202
}

6203 6204 6205 6206
static int netif_alloc_netdev_queues(struct net_device *dev)
{
	unsigned int count = dev->num_tx_queues;
	struct netdev_queue *tx;
6207
	size_t sz = count * sizeof(*tx);
6208

6209
	BUG_ON(count < 1 || count > 0xffff);
6210

6211 6212 6213 6214 6215 6216
	tx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!tx) {
		tx = vzalloc(sz);
		if (!tx)
			return -ENOMEM;
	}
6217
	dev->_tx = tx;
T
Tom Herbert 已提交
6218

6219 6220
	netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
	spin_lock_init(&dev->tx_global_lock);
C
Changli Gao 已提交
6221 6222

	return 0;
6223 6224
}

L
Linus Torvalds 已提交
6225 6226 6227 6228 6229 6230 6231 6232 6233 6234 6235 6236 6237 6238 6239 6240 6241 6242 6243 6244
/**
 *	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;
6245
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
6246 6247 6248 6249

	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

6250 6251
	might_sleep();

L
Linus Torvalds 已提交
6252 6253
	/* When net_device's are persistent, this will be fatal. */
	BUG_ON(dev->reg_state != NETREG_UNINITIALIZED);
6254
	BUG_ON(!net);
L
Linus Torvalds 已提交
6255

6256
	spin_lock_init(&dev->addr_list_lock);
6257
	netdev_set_addr_lockdep_class(dev);
L
Linus Torvalds 已提交
6258 6259 6260

	dev->iflink = -1;

6261
	ret = dev_get_valid_name(net, dev, dev->name);
6262 6263 6264
	if (ret < 0)
		goto out;

L
Linus Torvalds 已提交
6265
	/* Init, if this function is available */
6266 6267
	if (dev->netdev_ops->ndo_init) {
		ret = dev->netdev_ops->ndo_init(dev);
L
Linus Torvalds 已提交
6268 6269 6270
		if (ret) {
			if (ret > 0)
				ret = -EIO;
6271
			goto out;
L
Linus Torvalds 已提交
6272 6273
		}
	}
6274

6275 6276
	if (((dev->hw_features | dev->features) &
	     NETIF_F_HW_VLAN_CTAG_FILTER) &&
6277 6278 6279 6280 6281 6282 6283
	    (!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;
	}

6284 6285 6286 6287 6288 6289
	ret = -EBUSY;
	if (!dev->ifindex)
		dev->ifindex = dev_new_index(net);
	else if (__dev_get_by_index(net, dev->ifindex))
		goto err_uninit;

L
Linus Torvalds 已提交
6290 6291 6292
	if (dev->iflink == -1)
		dev->iflink = dev->ifindex;

6293 6294 6295 6296
	/* Transfer changeable features to wanted_features and enable
	 * software offloads (GSO and GRO).
	 */
	dev->hw_features |= NETIF_F_SOFT_FEATURES;
6297 6298
	dev->features |= NETIF_F_SOFT_FEATURES;
	dev->wanted_features = dev->features & dev->hw_features;
L
Linus Torvalds 已提交
6299

6300 6301
	if (!(dev->flags & IFF_LOOPBACK)) {
		dev->hw_features |= NETIF_F_NOCACHE_COPY;
6302 6303
	}

6304
	/* Make NETIF_F_HIGHDMA inheritable to VLAN devices.
6305
	 */
6306
	dev->vlan_features |= NETIF_F_HIGHDMA;
6307

6308 6309 6310 6311
	/* Make NETIF_F_SG inheritable to tunnel devices.
	 */
	dev->hw_enc_features |= NETIF_F_SG;

S
Simon Horman 已提交
6312 6313 6314 6315
	/* Make NETIF_F_SG inheritable to MPLS.
	 */
	dev->mpls_features |= NETIF_F_SG;

6316 6317 6318 6319 6320
	ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		goto err_uninit;

6321
	ret = netdev_register_kobject(dev);
6322
	if (ret)
6323
		goto err_uninit;
6324 6325
	dev->reg_state = NETREG_REGISTERED;

6326
	__netdev_update_features(dev);
6327

L
Linus Torvalds 已提交
6328 6329 6330 6331 6332 6333 6334
	/*
	 *	Default initial state at registry is that the
	 *	device is present.
	 */

	set_bit(__LINK_STATE_PRESENT, &dev->state);

6335 6336
	linkwatch_init_dev(dev);

L
Linus Torvalds 已提交
6337 6338
	dev_init_scheduler(dev);
	dev_hold(dev);
6339
	list_netdevice(dev);
6340
	add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
6341

6342 6343 6344 6345 6346 6347 6348
	/* 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
6349
	/* Notify protocols, that a new device appeared. */
6350
	ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
6351
	ret = notifier_to_errno(ret);
6352 6353 6354 6355
	if (ret) {
		rollback_registered(dev);
		dev->reg_state = NETREG_UNREGISTERED;
	}
6356 6357 6358 6359
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
6360 6361
	if (!dev->rtnl_link_ops ||
	    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
6362
		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
L
Linus Torvalds 已提交
6363 6364 6365

out:
	return ret;
6366 6367

err_uninit:
6368 6369
	if (dev->netdev_ops->ndo_uninit)
		dev->netdev_ops->ndo_uninit(dev);
6370
	goto out;
L
Linus Torvalds 已提交
6371
}
E
Eric Dumazet 已提交
6372
EXPORT_SYMBOL(register_netdevice);
L
Linus Torvalds 已提交
6373

6374 6375 6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395 6396 6397 6398 6399 6400 6401 6402 6403 6404
/**
 *	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
6405 6406 6407 6408 6409
	/* Note : We dont allocate pcpu_refcnt for dummy devices,
	 * because users of this 'device' dont need to change
	 * its refcount.
	 */

6410 6411 6412 6413 6414
	return 0;
}
EXPORT_SYMBOL_GPL(init_dummy_netdev);


L
Linus Torvalds 已提交
6415 6416 6417 6418 6419 6420 6421 6422 6423
/**
 *	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.
 *
6424
 *	This is a wrapper around register_netdevice that takes the rtnl semaphore
L
Linus Torvalds 已提交
6425 6426 6427 6428 6429 6430 6431 6432 6433 6434 6435 6436 6437 6438
 *	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
6439 6440 6441 6442 6443 6444 6445 6446 6447 6448
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);

6449
/**
L
Linus Torvalds 已提交
6450
 * netdev_wait_allrefs - wait until all references are gone.
6451
 * @dev: target net_device
L
Linus Torvalds 已提交
6452 6453 6454 6455 6456 6457 6458
 *
 * 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
6459
 * call dev_put.
L
Linus Torvalds 已提交
6460 6461 6462 6463
 */
static void netdev_wait_allrefs(struct net_device *dev)
{
	unsigned long rebroadcast_time, warning_time;
E
Eric Dumazet 已提交
6464
	int refcnt;
L
Linus Torvalds 已提交
6465

6466 6467
	linkwatch_forget_dev(dev);

L
Linus Torvalds 已提交
6468
	rebroadcast_time = warning_time = jiffies;
E
Eric Dumazet 已提交
6469 6470 6471
	refcnt = netdev_refcnt_read(dev);

	while (refcnt != 0) {
L
Linus Torvalds 已提交
6472
		if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
6473
			rtnl_lock();
L
Linus Torvalds 已提交
6474 6475

			/* Rebroadcast unregister notification */
6476
			call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
L
Linus Torvalds 已提交
6477

6478
			__rtnl_unlock();
6479
			rcu_barrier();
6480 6481
			rtnl_lock();

6482
			call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
L
Linus Torvalds 已提交
6483 6484 6485 6486 6487 6488 6489 6490 6491 6492 6493
			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();
			}

6494
			__rtnl_unlock();
L
Linus Torvalds 已提交
6495 6496 6497 6498 6499 6500

			rebroadcast_time = jiffies;
		}

		msleep(250);

E
Eric Dumazet 已提交
6501 6502
		refcnt = netdev_refcnt_read(dev);

L
Linus Torvalds 已提交
6503
		if (time_after(jiffies, warning_time + 10 * HZ)) {
6504 6505
			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
				 dev->name, refcnt);
L
Linus Torvalds 已提交
6506 6507 6508 6509 6510 6511 6512 6513 6514 6515 6516 6517 6518 6519 6520 6521 6522 6523 6524
			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
6525
 * We are invoked by rtnl_unlock().
L
Linus Torvalds 已提交
6526
 * This allows us to deal with problems:
6527
 * 1) We can delete sysfs objects which invoke hotplug
L
Linus Torvalds 已提交
6528 6529 6530
 *    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
6531 6532 6533
 *
 * 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
6534 6535 6536
 */
void netdev_run_todo(void)
{
6537
	struct list_head list;
L
Linus Torvalds 已提交
6538 6539

	/* Snapshot list, allow later requests */
6540
	list_replace_init(&net_todo_list, &list);
H
Herbert Xu 已提交
6541 6542

	__rtnl_unlock();
6543

6544 6545

	/* Wait for rcu callbacks to finish before next phase */
6546 6547 6548
	if (!list_empty(&list))
		rcu_barrier();

L
Linus Torvalds 已提交
6549 6550
	while (!list_empty(&list)) {
		struct net_device *dev
6551
			= list_first_entry(&list, struct net_device, todo_list);
L
Linus Torvalds 已提交
6552 6553
		list_del(&dev->todo_list);

6554
		rtnl_lock();
6555
		call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6556
		__rtnl_unlock();
6557

6558
		if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
6559
			pr_err("network todo '%s' but state %d\n",
6560 6561 6562 6563
			       dev->name, dev->reg_state);
			dump_stack();
			continue;
		}
L
Linus Torvalds 已提交
6564

6565
		dev->reg_state = NETREG_UNREGISTERED;
L
Linus Torvalds 已提交
6566

6567
		on_each_cpu(flush_backlog, dev, 1);
6568

6569
		netdev_wait_allrefs(dev);
L
Linus Torvalds 已提交
6570

6571
		/* paranoia */
E
Eric Dumazet 已提交
6572
		BUG_ON(netdev_refcnt_read(dev));
6573 6574
		WARN_ON(rcu_access_pointer(dev->ip_ptr));
		WARN_ON(rcu_access_pointer(dev->ip6_ptr));
6575
		WARN_ON(dev->dn_ptr);
L
Linus Torvalds 已提交
6576

6577 6578
		if (dev->destructor)
			dev->destructor(dev);
6579

6580 6581 6582 6583 6584 6585
		/* Report a network device has been unregistered */
		rtnl_lock();
		dev_net(dev)->dev_unreg_count--;
		__rtnl_unlock();
		wake_up(&netdev_unregistering_wq);

6586 6587
		/* Free network device */
		kobject_put(&dev->dev.kobj);
L
Linus Torvalds 已提交
6588 6589 6590
	}
}

6591 6592 6593
/* Convert net_device_stats to rtnl_link_stats64.  They have the same
 * fields in the same order, with only the type differing.
 */
6594 6595
void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
			     const struct net_device_stats *netdev_stats)
6596 6597
{
#if BITS_PER_LONG == 64
6598 6599
	BUILD_BUG_ON(sizeof(*stats64) != sizeof(*netdev_stats));
	memcpy(stats64, netdev_stats, sizeof(*stats64));
6600 6601 6602 6603 6604 6605 6606 6607 6608 6609 6610
#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
}
6611
EXPORT_SYMBOL(netdev_stats_to_stats64);
6612

6613 6614 6615
/**
 *	dev_get_stats	- get network device statistics
 *	@dev: device to get statistics from
6616
 *	@storage: place to store stats
6617
 *
6618 6619 6620 6621
 *	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.
6622
 */
6623 6624
struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
					struct rtnl_link_stats64 *storage)
6625
{
6626 6627
	const struct net_device_ops *ops = dev->netdev_ops;

6628 6629
	if (ops->ndo_get_stats64) {
		memset(storage, 0, sizeof(*storage));
6630 6631
		ops->ndo_get_stats64(dev, storage);
	} else if (ops->ndo_get_stats) {
6632
		netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev));
6633 6634
	} else {
		netdev_stats_to_stats64(storage, &dev->stats);
6635
	}
6636
	storage->rx_dropped += atomic_long_read(&dev->rx_dropped);
6637
	storage->tx_dropped += atomic_long_read(&dev->tx_dropped);
6638
	return storage;
R
Rusty Russell 已提交
6639
}
6640
EXPORT_SYMBOL(dev_get_stats);
R
Rusty Russell 已提交
6641

6642
struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
6643
{
6644
	struct netdev_queue *queue = dev_ingress_queue(dev);
6645

6646 6647 6648 6649 6650 6651 6652 6653 6654 6655 6656 6657
#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);
	queue->qdisc = &noop_qdisc;
	queue->qdisc_sleeping = &noop_qdisc;
	rcu_assign_pointer(dev->ingress_queue, queue);
#endif
	return queue;
6658 6659
}

6660 6661
static const struct ethtool_ops default_ethtool_ops;

6662 6663 6664 6665 6666 6667 6668 6669
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);

6670 6671 6672 6673
void netdev_freemem(struct net_device *dev)
{
	char *addr = (char *)dev - dev->padded;

W
WANG Cong 已提交
6674
	kvfree(addr);
6675 6676
}

L
Linus Torvalds 已提交
6677
/**
T
Tom Herbert 已提交
6678
 *	alloc_netdev_mqs - allocate network device
6679 6680 6681 6682 6683 6684
 *	@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
6685 6686
 *
 *	Allocates a struct net_device with private data area for driver use
L
Li Zhong 已提交
6687
 *	and performs basic initialization.  Also allocates subqueue structs
T
Tom Herbert 已提交
6688
 *	for each queue on the device.
L
Linus Torvalds 已提交
6689
 */
T
Tom Herbert 已提交
6690
struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
6691
		unsigned char name_assign_type,
T
Tom Herbert 已提交
6692 6693
		void (*setup)(struct net_device *),
		unsigned int txqs, unsigned int rxqs)
L
Linus Torvalds 已提交
6694 6695
{
	struct net_device *dev;
6696
	size_t alloc_size;
6697
	struct net_device *p;
L
Linus Torvalds 已提交
6698

6699 6700
	BUG_ON(strlen(name) >= sizeof(dev->name));

T
Tom Herbert 已提交
6701
	if (txqs < 1) {
6702
		pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
6703 6704 6705
		return NULL;
	}

6706
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
6707
	if (rxqs < 1) {
6708
		pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
T
Tom Herbert 已提交
6709 6710 6711 6712
		return NULL;
	}
#endif

6713
	alloc_size = sizeof(struct net_device);
6714 6715
	if (sizeof_priv) {
		/* ensure 32-byte alignment of private area */
6716
		alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
6717 6718 6719
		alloc_size += sizeof_priv;
	}
	/* ensure 32-byte alignment of whole construct */
6720
	alloc_size += NETDEV_ALIGN - 1;
L
Linus Torvalds 已提交
6721

6722 6723 6724
	p = kzalloc(alloc_size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!p)
		p = vzalloc(alloc_size);
6725
	if (!p)
L
Linus Torvalds 已提交
6726 6727
		return NULL;

6728
	dev = PTR_ALIGN(p, NETDEV_ALIGN);
L
Linus Torvalds 已提交
6729
	dev->padded = (char *)dev - (char *)p;
6730

E
Eric Dumazet 已提交
6731 6732
	dev->pcpu_refcnt = alloc_percpu(int);
	if (!dev->pcpu_refcnt)
6733
		goto free_dev;
6734 6735

	if (dev_addr_init(dev))
E
Eric Dumazet 已提交
6736
		goto free_pcpu;
6737

6738
	dev_mc_init(dev);
6739
	dev_uc_init(dev);
J
Jiri Pirko 已提交
6740

6741
	dev_net_set(dev, &init_net);
L
Linus Torvalds 已提交
6742

6743
	dev->gso_max_size = GSO_MAX_SIZE;
6744
	dev->gso_max_segs = GSO_MAX_SEGS;
6745
	dev->gso_min_segs = 0;
6746 6747 6748

	INIT_LIST_HEAD(&dev->napi_list);
	INIT_LIST_HEAD(&dev->unreg_list);
6749
	INIT_LIST_HEAD(&dev->close_list);
6750
	INIT_LIST_HEAD(&dev->link_watch_list);
6751 6752 6753 6754
	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);
6755
	dev->priv_flags = IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM;
6756 6757
	setup(dev);

T
Tom Herbert 已提交
6758 6759
	dev->num_tx_queues = txqs;
	dev->real_num_tx_queues = txqs;
6760
	if (netif_alloc_netdev_queues(dev))
6761
		goto free_all;
6762

6763
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
6764 6765
	dev->num_rx_queues = rxqs;
	dev->real_num_rx_queues = rxqs;
T
Tom Herbert 已提交
6766
	if (netif_alloc_rx_queues(dev))
6767
		goto free_all;
E
Eric Dumazet 已提交
6768
#endif
T
Tom Herbert 已提交
6769

L
Linus Torvalds 已提交
6770
	strcpy(dev->name, name);
6771
	dev->name_assign_type = name_assign_type;
6772
	dev->group = INIT_NETDEV_GROUP;
6773 6774
	if (!dev->ethtool_ops)
		dev->ethtool_ops = &default_ethtool_ops;
L
Linus Torvalds 已提交
6775
	return dev;
6776

6777 6778 6779 6780
free_all:
	free_netdev(dev);
	return NULL;

E
Eric Dumazet 已提交
6781 6782
free_pcpu:
	free_percpu(dev->pcpu_refcnt);
6783 6784
free_dev:
	netdev_freemem(dev);
6785
	return NULL;
L
Linus Torvalds 已提交
6786
}
T
Tom Herbert 已提交
6787
EXPORT_SYMBOL(alloc_netdev_mqs);
L
Linus Torvalds 已提交
6788 6789 6790 6791 6792

/**
 *	free_netdev - free network device
 *	@dev: device
 *
6793 6794
 *	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
6795 6796 6797 6798
 *	If this is the last reference then it will be freed.
 */
void free_netdev(struct net_device *dev)
{
6799 6800
	struct napi_struct *p, *n;

6801 6802
	release_net(dev_net(dev));

6803
	netif_free_tx_queues(dev);
6804
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
6805 6806
	kfree(dev->_rx);
#endif
6807

6808
	kfree(rcu_dereference_protected(dev->ingress_queue, 1));
6809

6810 6811 6812
	/* Flush device addresses */
	dev_addr_flush(dev);

6813 6814 6815
	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
		netif_napi_del(p);

E
Eric Dumazet 已提交
6816 6817 6818
	free_percpu(dev->pcpu_refcnt);
	dev->pcpu_refcnt = NULL;

S
Stephen Hemminger 已提交
6819
	/*  Compatibility with error handling in drivers */
L
Linus Torvalds 已提交
6820
	if (dev->reg_state == NETREG_UNINITIALIZED) {
6821
		netdev_freemem(dev);
L
Linus Torvalds 已提交
6822 6823 6824 6825 6826 6827
		return;
	}

	BUG_ON(dev->reg_state != NETREG_UNREGISTERED);
	dev->reg_state = NETREG_RELEASED;

6828 6829
	/* will free via device release */
	put_device(&dev->dev);
L
Linus Torvalds 已提交
6830
}
E
Eric Dumazet 已提交
6831
EXPORT_SYMBOL(free_netdev);
6832

6833 6834 6835 6836 6837 6838
/**
 *	synchronize_net -  Synchronize with packet receive processing
 *
 *	Wait for packets currently being received to be done.
 *	Does not block later packets from starting.
 */
6839
void synchronize_net(void)
L
Linus Torvalds 已提交
6840 6841
{
	might_sleep();
6842 6843 6844 6845
	if (rtnl_is_locked())
		synchronize_rcu_expedited();
	else
		synchronize_rcu();
L
Linus Torvalds 已提交
6846
}
E
Eric Dumazet 已提交
6847
EXPORT_SYMBOL(synchronize_net);
L
Linus Torvalds 已提交
6848 6849

/**
6850
 *	unregister_netdevice_queue - remove device from the kernel
L
Linus Torvalds 已提交
6851
 *	@dev: device
6852
 *	@head: list
6853
 *
L
Linus Torvalds 已提交
6854
 *	This function shuts down a device interface and removes it
6855
 *	from the kernel tables.
6856
 *	If head not NULL, device is queued to be unregistered later.
L
Linus Torvalds 已提交
6857 6858 6859 6860 6861
 *
 *	Callers must hold the rtnl semaphore.  You may want
 *	unregister_netdev() instead of this.
 */

6862
void unregister_netdevice_queue(struct net_device *dev, struct list_head *head)
L
Linus Torvalds 已提交
6863
{
6864 6865
	ASSERT_RTNL();

6866
	if (head) {
6867
		list_move_tail(&dev->unreg_list, head);
6868 6869 6870 6871 6872
	} else {
		rollback_registered(dev);
		/* Finish processing unregister after unlock */
		net_set_todo(dev);
	}
L
Linus Torvalds 已提交
6873
}
6874
EXPORT_SYMBOL(unregister_netdevice_queue);
L
Linus Torvalds 已提交
6875

6876 6877 6878
/**
 *	unregister_netdevice_many - unregister many devices
 *	@head: list of devices
6879 6880 6881
 *
 *  Note: As most callers use a stack allocated list_head,
 *  we force a list_del() to make sure stack wont be corrupted later.
6882 6883 6884 6885 6886 6887 6888 6889 6890
 */
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);
6891
		list_del(head);
6892 6893
	}
}
6894
EXPORT_SYMBOL(unregister_netdevice_many);
6895

L
Linus Torvalds 已提交
6896 6897 6898 6899 6900
/**
 *	unregister_netdev - remove device from the kernel
 *	@dev: device
 *
 *	This function shuts down a device interface and removes it
6901
 *	from the kernel tables.
L
Linus Torvalds 已提交
6902 6903 6904 6905 6906 6907 6908 6909 6910 6911 6912 6913 6914
 *
 *	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);

6915 6916 6917 6918 6919 6920 6921 6922 6923 6924 6925 6926 6927 6928 6929 6930 6931 6932 6933 6934 6935 6936 6937 6938 6939 6940 6941 6942 6943 6944 6945
/**
 *	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;
6946
	if (net_eq(dev_net(dev), net))
6947 6948 6949 6950 6951 6952
		goto out;

	/* Pick the destination device name, and ensure
	 * we can use it in the destination network namespace.
	 */
	err = -EEXIST;
6953
	if (__dev_get_by_name(net, dev->name)) {
6954 6955 6956
		/* We get here if we can't use the current device name */
		if (!pat)
			goto out;
6957
		if (dev_get_valid_name(net, dev, pat) < 0)
6958 6959 6960 6961 6962 6963 6964 6965
			goto out;
	}

	/*
	 * And now a mini version of register_netdevice unregister_netdevice.
	 */

	/* If device is running close it first. */
6966
	dev_close(dev);
6967 6968 6969 6970 6971 6972 6973 6974 6975 6976 6977 6978

	/* 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.
6979 6980 6981 6982

	   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.
6983 6984
	*/
	call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
6985 6986
	rcu_barrier();
	call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6987
	rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);
6988 6989 6990 6991

	/*
	 *	Flush the unicast and multicast chains
	 */
6992
	dev_uc_flush(dev);
6993
	dev_mc_flush(dev);
6994

6995 6996
	/* Send a netdev-removed uevent to the old namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_REMOVE);
6997
	netdev_adjacent_del_links(dev);
6998

6999
	/* Actually switch the network namespace */
7000
	dev_net_set(dev, net);
7001 7002 7003 7004 7005 7006 7007 7008 7009

	/* If there is an ifindex conflict assign a new one */
	if (__dev_get_by_index(net, dev->ifindex)) {
		int iflink = (dev->iflink == dev->ifindex);
		dev->ifindex = dev_new_index(net);
		if (iflink)
			dev->iflink = dev->ifindex;
	}

7010 7011
	/* Send a netdev-add uevent to the new namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_ADD);
7012
	netdev_adjacent_add_links(dev);
7013

7014
	/* Fixup kobjects */
7015
	err = device_rename(&dev->dev, dev->name);
7016
	WARN_ON(err);
7017 7018 7019 7020 7021 7022 7023

	/* Add the device back in the hashes */
	list_netdevice(dev);

	/* Notify protocols, that a new device appeared. */
	call_netdevice_notifiers(NETDEV_REGISTER, dev);

7024 7025 7026 7027
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
7028
	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
7029

7030 7031 7032 7033 7034
	synchronize_net();
	err = 0;
out:
	return err;
}
7035
EXPORT_SYMBOL_GPL(dev_change_net_namespace);
7036

L
Linus Torvalds 已提交
7037 7038 7039 7040 7041 7042 7043 7044 7045
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;

7046
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
L
Linus Torvalds 已提交
7047 7048 7049 7050 7051 7052 7053 7054 7055 7056 7057 7058 7059 7060 7061 7062
		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. */
7063 7064 7065 7066 7067 7068
	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;
	}
7069 7070 7071 7072 7073
	/* Append NAPI poll list from offline CPU. */
	if (!list_empty(&oldsd->poll_list)) {
		list_splice_init(&oldsd->poll_list, &sd->poll_list);
		raise_softirq_irqoff(NET_RX_SOFTIRQ);
	}
L
Linus Torvalds 已提交
7074 7075 7076 7077 7078

	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_enable();

	/* Process offline CPU's input_pkt_queue */
7079
	while ((skb = __skb_dequeue(&oldsd->process_queue))) {
7080
		netif_rx_internal(skb);
7081
		input_queue_head_incr(oldsd);
T
Tom Herbert 已提交
7082
	}
7083
	while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) {
7084
		netif_rx_internal(skb);
7085 7086
		input_queue_head_incr(oldsd);
	}
L
Linus Torvalds 已提交
7087 7088 7089 7090 7091

	return NOTIFY_OK;
}


7092
/**
7093 7094 7095 7096
 *	netdev_increment_features - increment feature set by one
 *	@all: current feature set
 *	@one: new feature set
 *	@mask: mask feature set
7097 7098
 *
 *	Computes a new feature set after adding a device with feature set
7099 7100
 *	@one to the master device with current feature set @all.  Will not
 *	enable anything that is off in @mask. Returns the new feature set.
7101
 */
7102 7103
netdev_features_t netdev_increment_features(netdev_features_t all,
	netdev_features_t one, netdev_features_t mask)
7104
{
7105 7106 7107
	if (mask & NETIF_F_GEN_CSUM)
		mask |= NETIF_F_ALL_CSUM;
	mask |= NETIF_F_VLAN_CHALLENGED;
7108

7109 7110
	all |= one & (NETIF_F_ONE_FOR_ALL|NETIF_F_ALL_CSUM) & mask;
	all &= one | ~NETIF_F_ALL_FOR_ALL;
7111

7112 7113 7114
	/* If one device supports hw checksumming, set for all. */
	if (all & NETIF_F_GEN_CSUM)
		all &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
7115 7116 7117

	return all;
}
7118
EXPORT_SYMBOL(netdev_increment_features);
7119

7120
static struct hlist_head * __net_init netdev_create_hash(void)
7121 7122 7123 7124 7125 7126 7127 7128 7129 7130 7131 7132
{
	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;
}

7133
/* Initialize per network namespace state */
7134
static int __net_init netdev_init(struct net *net)
7135
{
7136 7137
	if (net != &init_net)
		INIT_LIST_HEAD(&net->dev_base_head);
7138

7139 7140 7141
	net->dev_name_head = netdev_create_hash();
	if (net->dev_name_head == NULL)
		goto err_name;
7142

7143 7144 7145
	net->dev_index_head = netdev_create_hash();
	if (net->dev_index_head == NULL)
		goto err_idx;
7146 7147

	return 0;
7148 7149 7150 7151 7152

err_idx:
	kfree(net->dev_name_head);
err_name:
	return -ENOMEM;
7153 7154
}

7155 7156 7157 7158 7159 7160
/**
 *	netdev_drivername - network driver for the device
 *	@dev: network device
 *
 *	Determine network driver for device.
 */
7161
const char *netdev_drivername(const struct net_device *dev)
7162
{
7163 7164
	const struct device_driver *driver;
	const struct device *parent;
7165
	const char *empty = "";
7166 7167 7168

	parent = dev->dev.parent;
	if (!parent)
7169
		return empty;
7170 7171 7172

	driver = parent->driver;
	if (driver && driver->name)
7173 7174
		return driver->name;
	return empty;
7175 7176
}

7177 7178
static void __netdev_printk(const char *level, const struct net_device *dev,
			    struct va_format *vaf)
7179
{
7180
	if (dev && dev->dev.parent) {
7181 7182 7183 7184 7185 7186 7187
		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);
7188
	} else if (dev) {
7189 7190
		printk("%s%s%s: %pV",
		       level, netdev_name(dev), netdev_reg_state(dev), vaf);
7191
	} else {
7192
		printk("%s(NULL net_device): %pV", level, vaf);
7193
	}
7194 7195
}

7196 7197
void netdev_printk(const char *level, const struct net_device *dev,
		   const char *format, ...)
7198 7199 7200 7201 7202 7203 7204 7205 7206
{
	struct va_format vaf;
	va_list args;

	va_start(args, format);

	vaf.fmt = format;
	vaf.va = &args;

7207
	__netdev_printk(level, dev, &vaf);
7208

7209 7210 7211 7212 7213
	va_end(args);
}
EXPORT_SYMBOL(netdev_printk);

#define define_netdev_printk_level(func, level)			\
7214
void func(const struct net_device *dev, const char *fmt, ...)	\
7215 7216 7217 7218 7219 7220 7221 7222 7223
{								\
	struct va_format vaf;					\
	va_list args;						\
								\
	va_start(args, fmt);					\
								\
	vaf.fmt = fmt;						\
	vaf.va = &args;						\
								\
7224
	__netdev_printk(level, dev, &vaf);			\
7225
								\
7226 7227 7228 7229 7230 7231 7232 7233 7234 7235 7236 7237
	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);

7238
static void __net_exit netdev_exit(struct net *net)
7239 7240 7241 7242 7243
{
	kfree(net->dev_name_head);
	kfree(net->dev_index_head);
}

7244
static struct pernet_operations __net_initdata netdev_net_ops = {
7245 7246 7247 7248
	.init = netdev_init,
	.exit = netdev_exit,
};

7249
static void __net_exit default_device_exit(struct net *net)
7250
{
7251
	struct net_device *dev, *aux;
7252
	/*
7253
	 * Push all migratable network devices back to the
7254 7255 7256
	 * initial network namespace
	 */
	rtnl_lock();
7257
	for_each_netdev_safe(net, dev, aux) {
7258
		int err;
7259
		char fb_name[IFNAMSIZ];
7260 7261 7262 7263 7264

		/* Ignore unmoveable devices (i.e. loopback) */
		if (dev->features & NETIF_F_NETNS_LOCAL)
			continue;

7265 7266 7267
		/* Leave virtual devices for the generic cleanup */
		if (dev->rtnl_link_ops)
			continue;
7268

L
Lucas De Marchi 已提交
7269
		/* Push remaining network devices to init_net */
7270 7271
		snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex);
		err = dev_change_net_namespace(dev, &init_net, fb_name);
7272
		if (err) {
7273 7274
			pr_emerg("%s: failed to move %s to init_net: %d\n",
				 __func__, dev->name, err);
7275
			BUG();
7276 7277 7278 7279 7280
		}
	}
	rtnl_unlock();
}

7281 7282 7283 7284 7285 7286 7287
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;
7288
	DEFINE_WAIT_FUNC(wait, woken_wake_function);
7289

7290
	add_wait_queue(&netdev_unregistering_wq, &wait);
7291 7292 7293 7294 7295 7296 7297 7298 7299 7300 7301 7302
	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();
7303 7304

		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
7305
	}
7306
	remove_wait_queue(&netdev_unregistering_wq, &wait);
7307 7308
}

7309 7310 7311
static void __net_exit default_device_exit_batch(struct list_head *net_list)
{
	/* At exit all network devices most be removed from a network
7312
	 * namespace.  Do this in the reverse order of registration.
7313 7314 7315 7316 7317 7318 7319
	 * 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);

7320 7321 7322 7323 7324 7325 7326 7327 7328 7329 7330 7331
	/* 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);
7332 7333
	list_for_each_entry(net, net_list, exit_list) {
		for_each_netdev_reverse(net, dev) {
7334
			if (dev->rtnl_link_ops && dev->rtnl_link_ops->dellink)
7335 7336 7337 7338 7339 7340 7341 7342 7343
				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();
}

7344
static struct pernet_operations __net_initdata default_device_ops = {
7345
	.exit = default_device_exit,
7346
	.exit_batch = default_device_exit_batch,
7347 7348
};

L
Linus Torvalds 已提交
7349 7350 7351 7352 7353 7354 7355 7356 7357 7358 7359 7360 7361 7362 7363 7364 7365 7366 7367 7368
/*
 *	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;

7369
	if (netdev_kobject_init())
L
Linus Torvalds 已提交
7370 7371 7372
		goto out;

	INIT_LIST_HEAD(&ptype_all);
7373
	for (i = 0; i < PTYPE_HASH_SIZE; i++)
L
Linus Torvalds 已提交
7374 7375
		INIT_LIST_HEAD(&ptype_base[i]);

7376 7377
	INIT_LIST_HEAD(&offload_base);

7378 7379
	if (register_pernet_subsys(&netdev_net_ops))
		goto out;
L
Linus Torvalds 已提交
7380 7381 7382 7383 7384

	/*
	 *	Initialise the packet receive queues.
	 */

7385
	for_each_possible_cpu(i) {
E
Eric Dumazet 已提交
7386
		struct softnet_data *sd = &per_cpu(softnet_data, i);
L
Linus Torvalds 已提交
7387

E
Eric Dumazet 已提交
7388
		skb_queue_head_init(&sd->input_pkt_queue);
7389
		skb_queue_head_init(&sd->process_queue);
E
Eric Dumazet 已提交
7390
		INIT_LIST_HEAD(&sd->poll_list);
7391
		sd->output_queue_tailp = &sd->output_queue;
E
Eric Dumazet 已提交
7392
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
7393 7394 7395
		sd->csd.func = rps_trigger_softirq;
		sd->csd.info = sd;
		sd->cpu = i;
7396
#endif
T
Tom Herbert 已提交
7397

E
Eric Dumazet 已提交
7398 7399
		sd->backlog.poll = process_backlog;
		sd->backlog.weight = weight_p;
L
Linus Torvalds 已提交
7400 7401 7402 7403
	}

	dev_boot_phase = 0;

7404 7405 7406 7407 7408 7409 7410 7411 7412 7413 7414 7415 7416 7417 7418
	/* 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;

7419 7420
	open_softirq(NET_TX_SOFTIRQ, net_tx_action);
	open_softirq(NET_RX_SOFTIRQ, net_rx_action);
L
Linus Torvalds 已提交
7421 7422 7423 7424 7425 7426 7427 7428 7429

	hotcpu_notifier(dev_cpu_callback, 0);
	dst_init();
	rc = 0;
out:
	return rc;
}

subsys_initcall(net_dev_init);