dev.c 172.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9
/*
 * 	NET3	Protocol independent device support routines.
 *
 *		This program is free software; you can redistribute it and/or
 *		modify it under the terms of the GNU General Public License
 *		as published by the Free Software Foundation; either version
 *		2 of the License, or (at your option) any later version.
 *
 *	Derived from the non IP parts of dev.c 1.0.19
10
 * 		Authors:	Ross Biro
L
Linus Torvalds 已提交
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76
 *				Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 *				Mark Evans, <evansmp@uhura.aston.ac.uk>
 *
 *	Additional Authors:
 *		Florian la Roche <rzsfl@rz.uni-sb.de>
 *		Alan Cox <gw4pts@gw4pts.ampr.org>
 *		David Hinds <dahinds@users.sourceforge.net>
 *		Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
 *		Adam Sulmicki <adam@cfar.umd.edu>
 *              Pekka Riikonen <priikone@poesidon.pspt.fi>
 *
 *	Changes:
 *              D.J. Barrow     :       Fixed bug where dev->refcnt gets set
 *              			to 2 if register_netdev gets called
 *              			before net_dev_init & also removed a
 *              			few lines of code in the process.
 *		Alan Cox	:	device private ioctl copies fields back.
 *		Alan Cox	:	Transmit queue code does relevant
 *					stunts to keep the queue safe.
 *		Alan Cox	:	Fixed double lock.
 *		Alan Cox	:	Fixed promisc NULL pointer trap
 *		????????	:	Support the full private ioctl range
 *		Alan Cox	:	Moved ioctl permission check into
 *					drivers
 *		Tim Kordas	:	SIOCADDMULTI/SIOCDELMULTI
 *		Alan Cox	:	100 backlog just doesn't cut it when
 *					you start doing multicast video 8)
 *		Alan Cox	:	Rewrote net_bh and list manager.
 *		Alan Cox	: 	Fix ETH_P_ALL echoback lengths.
 *		Alan Cox	:	Took out transmit every packet pass
 *					Saved a few bytes in the ioctl handler
 *		Alan Cox	:	Network driver sets packet type before
 *					calling netif_rx. Saves a function
 *					call a packet.
 *		Alan Cox	:	Hashed net_bh()
 *		Richard Kooijman:	Timestamp fixes.
 *		Alan Cox	:	Wrong field in SIOCGIFDSTADDR
 *		Alan Cox	:	Device lock protection.
 *		Alan Cox	: 	Fixed nasty side effect of device close
 *					changes.
 *		Rudi Cilibrasi	:	Pass the right thing to
 *					set_mac_address()
 *		Dave Miller	:	32bit quantity for the device lock to
 *					make it work out on a Sparc.
 *		Bjorn Ekwall	:	Added KERNELD hack.
 *		Alan Cox	:	Cleaned up the backlog initialise.
 *		Craig Metz	:	SIOCGIFCONF fix if space for under
 *					1 device.
 *	    Thomas Bogendoerfer :	Return ENODEV for dev_open, if there
 *					is no device open function.
 *		Andi Kleen	:	Fix error reporting for SIOCGIFCONF
 *	    Michael Chastain	:	Fix signed/unsigned for SIOCGIFCONF
 *		Cyrus Durgin	:	Cleaned for KMOD
 *		Adam Sulmicki   :	Bug Fix : Network Device Unload
 *					A network device unload needs to purge
 *					the backlog queue.
 *	Paul Rusty Russell	:	SIOCSIFNAME
 *              Pekka Riikonen  :	Netdev boot-time settings code
 *              Andrew Morton   :       Make unregister_netdevice wait
 *              			indefinitely on dev->refcnt
 * 		J Hadi Salim	:	- Backlog queue sampling
 *				        - netif_rx() feedback
 */

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

136 137
#include "net-sysfs.h"

138 139 140
/* Instead of increasing this, you should create a hash table. */
#define MAX_GRO_SKBS 8

H
Herbert Xu 已提交
141 142 143
/* 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
144
static DEFINE_SPINLOCK(ptype_lock);
145
static DEFINE_SPINLOCK(offload_lock);
146 147
struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
struct list_head ptype_all __read_mostly;	/* Taps */
148
static struct list_head offload_base __read_mostly;
L
Linus Torvalds 已提交
149 150

/*
151
 * The @dev_base_head list is protected by @dev_base_lock and the rtnl
L
Linus Torvalds 已提交
152 153
 * semaphore.
 *
154
 * Pure readers hold dev_base_lock for reading, or rcu_read_lock()
L
Linus Torvalds 已提交
155 156
 *
 * Writers must hold the rtnl semaphore while they loop through the
157
 * dev_base_head list, and hold dev_base_lock for writing when they do the
L
Linus Torvalds 已提交
158 159 160 161 162 163 164 165 166 167 168 169 170 171
 * 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
172 173 174 175 176 177
/* 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);

178
static seqcount_t devnet_rename_seq;
179

180 181 182 183 184
static inline void dev_base_seq_inc(struct net *net)
{
	while (++net->dev_base_seq == 0);
}

185
static inline struct hlist_head *dev_name_hash(struct net *net, const char *name)
L
Linus Torvalds 已提交
186
{
187 188
	unsigned int hash = full_name_hash(name, strnlen(name, IFNAMSIZ));

189
	return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)];
L
Linus Torvalds 已提交
190 191
}

192
static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex)
L
Linus Torvalds 已提交
193
{
194
	return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)];
L
Linus Torvalds 已提交
195 196
}

E
Eric Dumazet 已提交
197
static inline void rps_lock(struct softnet_data *sd)
198 199
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
200
	spin_lock(&sd->input_pkt_queue.lock);
201 202 203
#endif
}

E
Eric Dumazet 已提交
204
static inline void rps_unlock(struct softnet_data *sd)
205 206
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
207
	spin_unlock(&sd->input_pkt_queue.lock);
208 209 210
#endif
}

211
/* Device list insertion */
212
static void list_netdevice(struct net_device *dev)
213
{
214
	struct net *net = dev_net(dev);
215 216 217 218

	ASSERT_RTNL();

	write_lock_bh(&dev_base_lock);
219
	list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
220
	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
221 222
	hlist_add_head_rcu(&dev->index_hlist,
			   dev_index_hash(net, dev->ifindex));
223
	write_unlock_bh(&dev_base_lock);
224 225

	dev_base_seq_inc(net);
226 227
}

228 229 230
/* Device list removal
 * caller must respect a RCU grace period before freeing/reusing dev
 */
231 232 233 234 235 236
static void unlist_netdevice(struct net_device *dev)
{
	ASSERT_RTNL();

	/* Unlink dev from the device chain */
	write_lock_bh(&dev_base_lock);
237
	list_del_rcu(&dev->dev_list);
238
	hlist_del_rcu(&dev->name_hlist);
239
	hlist_del_rcu(&dev->index_hlist);
240
	write_unlock_bh(&dev_base_lock);
241 242

	dev_base_seq_inc(dev_net(dev));
243 244
}

L
Linus Torvalds 已提交
245 246 247 248
/*
 *	Our notifier list
 */

249
static RAW_NOTIFIER_HEAD(netdev_chain);
L
Linus Torvalds 已提交
250 251 252 253 254

/*
 *	Device drivers call our routines to queue packets here. We empty the
 *	queue in the local softnet handler.
 */
255

256
DEFINE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
E
Eric Dumazet 已提交
257
EXPORT_PER_CPU_SYMBOL(softnet_data);
L
Linus Torvalds 已提交
258

259
#ifdef CONFIG_LOCKDEP
260
/*
261
 * register_netdevice() inits txq->_xmit_lock and sets lockdep class
262 263 264 265 266 267 268 269 270 271 272 273 274 275 276
 * 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,
277 278 279
	 ARPHRD_FCFABRIC, ARPHRD_IEEE80211, ARPHRD_IEEE80211_PRISM,
	 ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, ARPHRD_PHONET_PIPE,
	 ARPHRD_IEEE802154, ARPHRD_VOID, ARPHRD_NONE};
280

281
static const char *const netdev_lock_name[] =
282 283 284 285 286 287 288 289 290 291 292 293
	{"_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",
294 295 296
	 "_xmit_FCFABRIC", "_xmit_IEEE80211", "_xmit_IEEE80211_PRISM",
	 "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", "_xmit_PHONET_PIPE",
	 "_xmit_IEEE802154", "_xmit_VOID", "_xmit_NONE"};
297 298

static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)];
299
static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)];
300 301 302 303 304 305 306 307 308 309 310 311

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;
}

312 313
static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
						 unsigned short dev_type)
314 315 316 317 318 319 320
{
	int i;

	i = netdev_lock_pos(dev_type);
	lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i],
				   netdev_lock_name[i]);
}
321 322 323 324 325 326 327 328 329 330

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]);
}
331
#else
332 333 334 335 336
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)
337 338 339
{
}
#endif
L
Linus Torvalds 已提交
340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362

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

		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)
 */

363 364 365 366 367 368 369 370
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
371 372 373 374 375 376 377 378
/**
 *	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.
 *
379
 *	This call does not sleep therefore it can not
L
Linus Torvalds 已提交
380 381 382 383 384 385
 *	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)
{
386
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
387

388 389 390
	spin_lock(&ptype_lock);
	list_add_rcu(&pt->list, head);
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
391
}
E
Eric Dumazet 已提交
392
EXPORT_SYMBOL(dev_add_pack);
L
Linus Torvalds 已提交
393 394 395 396 397 398 399 400

/**
 *	__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
401
 *	returns.
L
Linus Torvalds 已提交
402 403 404 405 406 407 408
 *
 *      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)
{
409
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
410 411
	struct packet_type *pt1;

412
	spin_lock(&ptype_lock);
L
Linus Torvalds 已提交
413 414 415 416 417 418 419 420

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

421
	pr_warn("dev_remove_pack: %p not found\n", pt);
L
Linus Torvalds 已提交
422
out:
423
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
424
}
E
Eric Dumazet 已提交
425 426
EXPORT_SYMBOL(__dev_remove_pack);

L
Linus Torvalds 已提交
427 428 429 430 431 432 433 434 435 436 437 438 439 440 441
/**
 *	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);
442

L
Linus Torvalds 已提交
443 444
	synchronize_net();
}
E
Eric Dumazet 已提交
445
EXPORT_SYMBOL(dev_remove_pack);
L
Linus Torvalds 已提交
446

447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482

/**
 *	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
483
static void __dev_remove_offload(struct packet_offload *po)
484 485 486 487
{
	struct list_head *head = &offload_base;
	struct packet_offload *po1;

488
	spin_lock(&offload_lock);
489 490 491 492 493 494 495 496 497 498

	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:
499
	spin_unlock(&offload_lock);
500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521
}

/**
 *	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
522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548
/******************************************************************************

		      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));
549
			strlcpy(s[i].name, name, IFNAMSIZ);
L
Linus Torvalds 已提交
550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573
			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] != ' ' &&
574
		    !strcmp(dev->name, s[i].name)) {
L
Linus Torvalds 已提交
575 576 577 578 579 580 581 582 583
			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
584
EXPORT_SYMBOL(netdev_boot_setup_check);
L
Linus Torvalds 已提交
585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608


/**
 *	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
	 */
609
	if (__dev_get_by_name(&init_net, name))
L
Linus Torvalds 已提交
610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654
		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
655
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
656 657 658 659 660 661 662 663 664
 *	@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.
 */

665
struct net_device *__dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
666
{
E
Eric Dumazet 已提交
667 668
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);
L
Linus Torvalds 已提交
669

670
	hlist_for_each_entry(dev, head, name_hlist)
L
Linus Torvalds 已提交
671 672
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;
E
Eric Dumazet 已提交
673

L
Linus Torvalds 已提交
674 675
	return NULL;
}
E
Eric Dumazet 已提交
676
EXPORT_SYMBOL(__dev_get_by_name);
L
Linus Torvalds 已提交
677

678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694
/**
 *	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);

695
	hlist_for_each_entry_rcu(dev, head, name_hlist)
696 697 698 699 700 701 702
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_name_rcu);

L
Linus Torvalds 已提交
703 704
/**
 *	dev_get_by_name		- find a device by its name
705
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
706 707 708 709 710 711 712 713 714
 *	@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.
 */

715
struct net_device *dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
716 717 718
{
	struct net_device *dev;

719 720
	rcu_read_lock();
	dev = dev_get_by_name_rcu(net, name);
L
Linus Torvalds 已提交
721 722
	if (dev)
		dev_hold(dev);
723
	rcu_read_unlock();
L
Linus Torvalds 已提交
724 725
	return dev;
}
E
Eric Dumazet 已提交
726
EXPORT_SYMBOL(dev_get_by_name);
L
Linus Torvalds 已提交
727 728 729

/**
 *	__dev_get_by_index - find a device by its ifindex
730
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
731 732 733 734 735 736 737 738 739
 *	@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.
 */

740
struct net_device *__dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
741
{
E
Eric Dumazet 已提交
742 743
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);
L
Linus Torvalds 已提交
744

745
	hlist_for_each_entry(dev, head, index_hlist)
L
Linus Torvalds 已提交
746 747
		if (dev->ifindex == ifindex)
			return dev;
E
Eric Dumazet 已提交
748

L
Linus Torvalds 已提交
749 750
	return NULL;
}
E
Eric Dumazet 已提交
751
EXPORT_SYMBOL(__dev_get_by_index);
L
Linus Torvalds 已提交
752

753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768
/**
 *	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);

769
	hlist_for_each_entry_rcu(dev, head, index_hlist)
770 771 772 773 774 775 776
		if (dev->ifindex == ifindex)
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_index_rcu);

L
Linus Torvalds 已提交
777 778 779

/**
 *	dev_get_by_index - find a device by its ifindex
780
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
781 782 783 784 785 786 787 788
 *	@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.
 */

789
struct net_device *dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
790 791 792
{
	struct net_device *dev;

793 794
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, ifindex);
L
Linus Torvalds 已提交
795 796
	if (dev)
		dev_hold(dev);
797
	rcu_read_unlock();
L
Linus Torvalds 已提交
798 799
	return dev;
}
E
Eric Dumazet 已提交
800
EXPORT_SYMBOL(dev_get_by_index);
L
Linus Torvalds 已提交
801

802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835
/**
 *	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
836
/**
837
 *	dev_getbyhwaddr_rcu - find a device by its hardware address
838
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
839 840 841 842
 *	@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
843 844
 *	is not found or a pointer to the device.
 *	The caller must hold RCU or RTNL.
845
 *	The returned device has not had its ref count increased
L
Linus Torvalds 已提交
846 847 848 849
 *	and the caller must therefore be careful about locking
 *
 */

850 851
struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
				       const char *ha)
L
Linus Torvalds 已提交
852 853 854
{
	struct net_device *dev;

855
	for_each_netdev_rcu(net, dev)
L
Linus Torvalds 已提交
856 857
		if (dev->type == type &&
		    !memcmp(dev->dev_addr, ha, dev->addr_len))
858 859 860
			return dev;

	return NULL;
L
Linus Torvalds 已提交
861
}
862
EXPORT_SYMBOL(dev_getbyhwaddr_rcu);
863

864
struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type)
L
Linus Torvalds 已提交
865 866 867
{
	struct net_device *dev;

868
	ASSERT_RTNL();
869
	for_each_netdev(net, dev)
870
		if (dev->type == type)
871 872 873
			return dev;

	return NULL;
874 875 876
}
EXPORT_SYMBOL(__dev_getfirstbyhwtype);

877
struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type)
878
{
879
	struct net_device *dev, *ret = NULL;
880

881 882 883 884 885 886 887 888 889
	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
890 891 892 893
}
EXPORT_SYMBOL(dev_getfirstbyhwtype);

/**
E
Eric Dumazet 已提交
894
 *	dev_get_by_flags_rcu - find any device with given flags
895
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
896 897 898 899
 *	@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
900 901
 *	is not found or a pointer to the device. Must be called inside
 *	rcu_read_lock(), and result refcount is unchanged.
L
Linus Torvalds 已提交
902 903
 */

E
Eric Dumazet 已提交
904
struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short if_flags,
E
Eric Dumazet 已提交
905
				    unsigned short mask)
L
Linus Torvalds 已提交
906
{
907
	struct net_device *dev, *ret;
L
Linus Torvalds 已提交
908

909
	ret = NULL;
910
	for_each_netdev_rcu(net, dev) {
L
Linus Torvalds 已提交
911
		if (((dev->flags ^ if_flags) & mask) == 0) {
912
			ret = dev;
L
Linus Torvalds 已提交
913 914 915
			break;
		}
	}
916
	return ret;
L
Linus Torvalds 已提交
917
}
E
Eric Dumazet 已提交
918
EXPORT_SYMBOL(dev_get_by_flags_rcu);
L
Linus Torvalds 已提交
919 920 921 922 923 924

/**
 *	dev_valid_name - check if name is okay for network device
 *	@name: name string
 *
 *	Network device names need to be valid file names to
925 926
 *	to allow sysfs to work.  We also disallow any kind of
 *	whitespace.
L
Linus Torvalds 已提交
927
 */
928
bool dev_valid_name(const char *name)
L
Linus Torvalds 已提交
929
{
930
	if (*name == '\0')
931
		return false;
932
	if (strlen(name) >= IFNAMSIZ)
933
		return false;
934
	if (!strcmp(name, ".") || !strcmp(name, ".."))
935
		return false;
936 937 938

	while (*name) {
		if (*name == '/' || isspace(*name))
939
			return false;
940 941
		name++;
	}
942
	return true;
L
Linus Torvalds 已提交
943
}
E
Eric Dumazet 已提交
944
EXPORT_SYMBOL(dev_valid_name);
L
Linus Torvalds 已提交
945 946

/**
947 948
 *	__dev_alloc_name - allocate a name for a device
 *	@net: network namespace to allocate the device name in
L
Linus Torvalds 已提交
949
 *	@name: name format string
950
 *	@buf:  scratch buffer and result name string
L
Linus Torvalds 已提交
951 952
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
S
Stephen Hemminger 已提交
953 954 955 956 957 958
 *	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
959 960
 */

961
static int __dev_alloc_name(struct net *net, const char *name, char *buf)
L
Linus Torvalds 已提交
962 963 964 965
{
	int i = 0;
	const char *p;
	const int max_netdevices = 8*PAGE_SIZE;
S
Stephen Hemminger 已提交
966
	unsigned long *inuse;
L
Linus Torvalds 已提交
967 968 969 970 971 972 973 974 975 976 977 978 979
	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
980
		inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC);
L
Linus Torvalds 已提交
981 982 983
		if (!inuse)
			return -ENOMEM;

984
		for_each_netdev(net, d) {
L
Linus Torvalds 已提交
985 986 987 988 989 990
			if (!sscanf(d->name, name, &i))
				continue;
			if (i < 0 || i >= max_netdevices)
				continue;

			/*  avoid cases where sscanf is not exact inverse of printf */
991
			snprintf(buf, IFNAMSIZ, name, i);
L
Linus Torvalds 已提交
992 993 994 995 996 997 998 999
			if (!strncmp(buf, d->name, IFNAMSIZ))
				set_bit(i, inuse);
		}

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

1000 1001
	if (buf != name)
		snprintf(buf, IFNAMSIZ, name, i);
1002
	if (!__dev_get_by_name(net, buf))
L
Linus Torvalds 已提交
1003 1004 1005 1006 1007 1008 1009 1010 1011
		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;
}

1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031
/**
 *	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;

1032 1033
	BUG_ON(!dev_net(dev));
	net = dev_net(dev);
1034 1035 1036 1037 1038
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
}
E
Eric Dumazet 已提交
1039
EXPORT_SYMBOL(dev_alloc_name);
1040

1041 1042 1043
static int dev_alloc_name_ns(struct net *net,
			     struct net_device *dev,
			     const char *name)
1044
{
1045 1046
	char buf[IFNAMSIZ];
	int ret;
1047

1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058
	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);
1059

1060 1061 1062
	if (!dev_valid_name(name))
		return -EINVAL;

1063
	if (strchr(name, '%'))
1064
		return dev_alloc_name_ns(net, dev, name);
1065 1066
	else if (__dev_get_by_name(net, name))
		return -EEXIST;
1067 1068
	else if (dev->name != name)
		strlcpy(dev->name, name, IFNAMSIZ);
1069 1070 1071

	return 0;
}
L
Linus Torvalds 已提交
1072 1073 1074 1075 1076 1077 1078 1079 1080

/**
 *	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.
 */
1081
int dev_change_name(struct net_device *dev, const char *newname)
L
Linus Torvalds 已提交
1082
{
1083
	char oldname[IFNAMSIZ];
L
Linus Torvalds 已提交
1084
	int err = 0;
1085
	int ret;
1086
	struct net *net;
L
Linus Torvalds 已提交
1087 1088

	ASSERT_RTNL();
1089
	BUG_ON(!dev_net(dev));
L
Linus Torvalds 已提交
1090

1091
	net = dev_net(dev);
L
Linus Torvalds 已提交
1092 1093 1094
	if (dev->flags & IFF_UP)
		return -EBUSY;

1095
	write_seqcount_begin(&devnet_rename_seq);
1096 1097

	if (strncmp(newname, dev->name, IFNAMSIZ) == 0) {
1098
		write_seqcount_end(&devnet_rename_seq);
1099
		return 0;
1100
	}
1101

1102 1103
	memcpy(oldname, dev->name, IFNAMSIZ);

1104
	err = dev_get_valid_name(net, dev, newname);
1105
	if (err < 0) {
1106
		write_seqcount_end(&devnet_rename_seq);
1107
		return err;
1108
	}
L
Linus Torvalds 已提交
1109

1110
rollback:
1111 1112 1113
	ret = device_rename(&dev->dev, dev->name);
	if (ret) {
		memcpy(dev->name, oldname, IFNAMSIZ);
1114
		write_seqcount_end(&devnet_rename_seq);
1115
		return ret;
1116
	}
1117

1118
	write_seqcount_end(&devnet_rename_seq);
1119

1120
	write_lock_bh(&dev_base_lock);
1121
	hlist_del_rcu(&dev->name_hlist);
1122 1123 1124 1125 1126 1127
	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));
1128 1129
	write_unlock_bh(&dev_base_lock);

1130
	ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
1131 1132 1133
	ret = notifier_to_errno(ret);

	if (ret) {
1134 1135
		/* err >= 0 after dev_alloc_name() or stores the first errno */
		if (err >= 0) {
1136
			err = ret;
1137
			write_seqcount_begin(&devnet_rename_seq);
1138 1139
			memcpy(dev->name, oldname, IFNAMSIZ);
			goto rollback;
1140
		} else {
1141
			pr_err("%s: name change rollback failed: %d\n",
1142
			       dev->name, ret);
1143 1144
		}
	}
L
Linus Torvalds 已提交
1145 1146 1147 1148

	return err;
}

1149 1150 1151 1152
/**
 *	dev_set_alias - change ifalias of a device
 *	@dev: device
 *	@alias: name up to IFALIASZ
1153
 *	@len: limit of bytes to copy from info
1154 1155 1156 1157 1158
 *
 *	Set ifalias for a device,
 */
int dev_set_alias(struct net_device *dev, const char *alias, size_t len)
{
1159 1160
	char *new_ifalias;

1161 1162 1163 1164 1165
	ASSERT_RTNL();

	if (len >= IFALIASZ)
		return -EINVAL;

1166
	if (!len) {
1167 1168
		kfree(dev->ifalias);
		dev->ifalias = NULL;
1169 1170 1171
		return 0;
	}

1172 1173
	new_ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL);
	if (!new_ifalias)
1174
		return -ENOMEM;
1175
	dev->ifalias = new_ifalias;
1176 1177 1178 1179 1180 1181

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


1182
/**
S
Stephen Hemminger 已提交
1183
 *	netdev_features_change - device changes features
1184 1185 1186 1187 1188 1189
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed features.
 */
void netdev_features_change(struct net_device *dev)
{
1190
	call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev);
1191 1192 1193
}
EXPORT_SYMBOL(netdev_features_change);

L
Linus Torvalds 已提交
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
/**
 *	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) {
1205
		call_netdevice_notifiers(NETDEV_CHANGE, dev);
1206
		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
L
Linus Torvalds 已提交
1207 1208
	}
}
E
Eric Dumazet 已提交
1209
EXPORT_SYMBOL(netdev_state_change);
L
Linus Torvalds 已提交
1210

1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221
/**
 * 	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)
1222
{
1223 1224 1225
	rtnl_lock();
	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, dev);
	rtnl_unlock();
1226
}
1227
EXPORT_SYMBOL(netdev_notify_peers);
1228

1229
static int __dev_open(struct net_device *dev)
L
Linus Torvalds 已提交
1230
{
1231
	const struct net_device_ops *ops = dev->netdev_ops;
1232
	int ret;
L
Linus Torvalds 已提交
1233

1234 1235
	ASSERT_RTNL();

L
Linus Torvalds 已提交
1236 1237 1238
	if (!netif_device_present(dev))
		return -ENODEV;

1239 1240 1241 1242
	/* 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
	 */
1243
	netpoll_rx_disable(dev);
1244

1245 1246 1247 1248 1249
	ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

L
Linus Torvalds 已提交
1250
	set_bit(__LINK_STATE_START, &dev->state);
1251

1252 1253
	if (ops->ndo_validate_addr)
		ret = ops->ndo_validate_addr(dev);
1254

1255 1256
	if (!ret && ops->ndo_open)
		ret = ops->ndo_open(dev);
L
Linus Torvalds 已提交
1257

1258 1259
	netpoll_rx_enable(dev);

1260 1261 1262
	if (ret)
		clear_bit(__LINK_STATE_START, &dev->state);
	else {
L
Linus Torvalds 已提交
1263
		dev->flags |= IFF_UP;
1264
		net_dmaengine_get();
1265
		dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
1266
		dev_activate(dev);
1267
		add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
1268
	}
1269

L
Linus Torvalds 已提交
1270 1271 1272 1273
	return ret;
}

/**
1274 1275
 *	dev_open	- prepare an interface for use.
 *	@dev:	device to open
L
Linus Torvalds 已提交
1276
 *
1277 1278 1279 1280 1281 1282 1283
 *	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
1284
 */
1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
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;

1296
	rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1297 1298 1299 1300 1301 1302
	call_netdevice_notifiers(NETDEV_UP, dev);

	return ret;
}
EXPORT_SYMBOL(dev_open);

1303
static int __dev_close_many(struct list_head *head)
L
Linus Torvalds 已提交
1304
{
1305
	struct net_device *dev;
1306

1307
	ASSERT_RTNL();
1308 1309
	might_sleep();

1310
	list_for_each_entry(dev, head, close_list) {
1311
		call_netdevice_notifiers(NETDEV_GOING_DOWN, dev);
L
Linus Torvalds 已提交
1312

1313
		clear_bit(__LINK_STATE_START, &dev->state);
L
Linus Torvalds 已提交
1314

1315 1316 1317 1318 1319 1320 1321 1322
		/* 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.
		 */
		smp_mb__after_clear_bit(); /* Commit netif_running(). */
	}
L
Linus Torvalds 已提交
1323

1324
	dev_deactivate_many(head);
1325

1326
	list_for_each_entry(dev, head, close_list) {
1327
		const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
1328

1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347
		/*
		 *	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;
		net_dmaengine_put();
	}

	return 0;
}

static int __dev_close(struct net_device *dev)
{
1348
	int retval;
1349 1350
	LIST_HEAD(single);

1351
	/* Temporarily disable netpoll until the interface is down */
1352
	netpoll_rx_disable(dev);
1353

1354
	list_add(&dev->close_list, &single);
1355 1356
	retval = __dev_close_many(&single);
	list_del(&single);
1357 1358

	netpoll_rx_enable(dev);
1359
	return retval;
1360 1361
}

E
Eric Dumazet 已提交
1362
static int dev_close_many(struct list_head *head)
1363 1364
{
	struct net_device *dev, *tmp;
L
Linus Torvalds 已提交
1365

1366 1367
	/* Remove the devices that don't need to be closed */
	list_for_each_entry_safe(dev, tmp, head, close_list)
1368
		if (!(dev->flags & IFF_UP))
1369
			list_del_init(&dev->close_list);
1370 1371

	__dev_close_many(head);
L
Linus Torvalds 已提交
1372

1373
	list_for_each_entry_safe(dev, tmp, head, close_list) {
1374
		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1375
		call_netdevice_notifiers(NETDEV_DOWN, dev);
1376
		list_del_init(&dev->close_list);
1377
	}
1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392

	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)
{
1393 1394
	if (dev->flags & IFF_UP) {
		LIST_HEAD(single);
L
Linus Torvalds 已提交
1395

1396
		/* Block netpoll rx while the interface is going down */
1397
		netpoll_rx_disable(dev);
1398

1399
		list_add(&dev->close_list, &single);
1400 1401
		dev_close_many(&single);
		list_del(&single);
1402 1403

		netpoll_rx_enable(dev);
1404
	}
1405
	return 0;
L
Linus Torvalds 已提交
1406
}
E
Eric Dumazet 已提交
1407
EXPORT_SYMBOL(dev_close);
L
Linus Torvalds 已提交
1408 1409


1410 1411 1412 1413 1414 1415 1416 1417 1418 1419
/**
 *	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)
{
1420 1421 1422 1423 1424 1425 1426
	/*
	 * If we're trying to disable lro on a vlan device
	 * use the underlying physical device instead
	 */
	if (is_vlan_dev(dev))
		dev = vlan_dev_real_dev(dev);

1427 1428 1429 1430
	/* the same for macvlan devices */
	if (netif_is_macvlan(dev))
		dev = macvlan_dev_real_dev(dev);

M
Michał Mirosław 已提交
1431 1432
	dev->wanted_features &= ~NETIF_F_LRO;
	netdev_update_features(dev);
1433

1434 1435
	if (unlikely(dev->features & NETIF_F_LRO))
		netdev_WARN(dev, "failed to disable LRO!\n");
1436 1437 1438
}
EXPORT_SYMBOL(dev_disable_lro);

1439 1440 1441 1442 1443 1444 1445 1446
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);
}
1447

1448 1449
static int dev_boot_phase = 1;

L
Linus Torvalds 已提交
1450 1451 1452 1453 1454 1455 1456 1457 1458 1459
/**
 *	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
1460
 *	to the new notifier to allow device to have a race free
L
Linus Torvalds 已提交
1461 1462 1463 1464 1465 1466
 *	view of the network device list.
 */

int register_netdevice_notifier(struct notifier_block *nb)
{
	struct net_device *dev;
1467
	struct net_device *last;
1468
	struct net *net;
L
Linus Torvalds 已提交
1469 1470 1471
	int err;

	rtnl_lock();
1472
	err = raw_notifier_chain_register(&netdev_chain, nb);
1473 1474
	if (err)
		goto unlock;
1475 1476 1477 1478
	if (dev_boot_phase)
		goto unlock;
	for_each_net(net) {
		for_each_netdev(net, dev) {
1479
			err = call_netdevice_notifier(nb, NETDEV_REGISTER, dev);
1480 1481 1482 1483 1484 1485
			err = notifier_to_errno(err);
			if (err)
				goto rollback;

			if (!(dev->flags & IFF_UP))
				continue;
L
Linus Torvalds 已提交
1486

1487
			call_netdevice_notifier(nb, NETDEV_UP, dev);
1488
		}
L
Linus Torvalds 已提交
1489
	}
1490 1491

unlock:
L
Linus Torvalds 已提交
1492 1493
	rtnl_unlock();
	return err;
1494 1495 1496

rollback:
	last = dev;
1497 1498 1499
	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev == last)
1500
				goto outroll;
1501

1502
			if (dev->flags & IFF_UP) {
1503 1504 1505
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1506
			}
1507
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1508 1509
		}
	}
1510

1511
outroll:
1512
	raw_notifier_chain_unregister(&netdev_chain, nb);
1513
	goto unlock;
L
Linus Torvalds 已提交
1514
}
E
Eric Dumazet 已提交
1515
EXPORT_SYMBOL(register_netdevice_notifier);
L
Linus Torvalds 已提交
1516 1517 1518 1519 1520 1521 1522 1523 1524

/**
 *	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.
1525 1526 1527 1528
 *
 * 	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
1529 1530 1531 1532
 */

int unregister_netdevice_notifier(struct notifier_block *nb)
{
1533 1534
	struct net_device *dev;
	struct net *net;
1535 1536 1537
	int err;

	rtnl_lock();
1538
	err = raw_notifier_chain_unregister(&netdev_chain, nb);
1539 1540 1541 1542 1543 1544
	if (err)
		goto unlock;

	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev->flags & IFF_UP) {
1545 1546 1547
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1548
			}
1549
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1550 1551 1552
		}
	}
unlock:
1553 1554
	rtnl_unlock();
	return err;
L
Linus Torvalds 已提交
1555
}
E
Eric Dumazet 已提交
1556
EXPORT_SYMBOL(unregister_netdevice_notifier);
L
Linus Torvalds 已提交
1557

1558 1559 1560 1561 1562 1563 1564 1565 1566 1567
/**
 *	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
1568 1569 1570
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info)
1571 1572 1573 1574 1575 1576
{
	ASSERT_RTNL();
	netdev_notifier_info_init(info, dev);
	return raw_notifier_call_chain(&netdev_chain, val, info);
}

L
Linus Torvalds 已提交
1577 1578 1579
/**
 *	call_netdevice_notifiers - call all network notifier blocks
 *      @val: value passed unmodified to notifier function
1580
 *      @dev: net_device pointer passed unmodified to notifier function
L
Linus Torvalds 已提交
1581 1582
 *
 *	Call all network notifier blocks.  Parameters and return value
1583
 *	are as for raw_notifier_call_chain().
L
Linus Torvalds 已提交
1584 1585
 */

1586
int call_netdevice_notifiers(unsigned long val, struct net_device *dev)
L
Linus Torvalds 已提交
1587
{
1588 1589 1590
	struct netdev_notifier_info info;

	return call_netdevice_notifiers_info(val, dev, &info);
L
Linus Torvalds 已提交
1591
}
1592
EXPORT_SYMBOL(call_netdevice_notifiers);
L
Linus Torvalds 已提交
1593

1594
static struct static_key netstamp_needed __read_mostly;
1595
#ifdef HAVE_JUMP_LABEL
1596
/* We are not allowed to call static_key_slow_dec() from irq context
1597
 * If net_disable_timestamp() is called from irq context, defer the
1598
 * static_key_slow_dec() calls.
1599 1600 1601
 */
static atomic_t netstamp_needed_deferred;
#endif
L
Linus Torvalds 已提交
1602 1603 1604

void net_enable_timestamp(void)
{
1605 1606 1607 1608 1609
#ifdef HAVE_JUMP_LABEL
	int deferred = atomic_xchg(&netstamp_needed_deferred, 0);

	if (deferred) {
		while (--deferred)
1610
			static_key_slow_dec(&netstamp_needed);
1611 1612 1613
		return;
	}
#endif
1614
	static_key_slow_inc(&netstamp_needed);
L
Linus Torvalds 已提交
1615
}
E
Eric Dumazet 已提交
1616
EXPORT_SYMBOL(net_enable_timestamp);
L
Linus Torvalds 已提交
1617 1618 1619

void net_disable_timestamp(void)
{
1620 1621 1622 1623 1624 1625
#ifdef HAVE_JUMP_LABEL
	if (in_interrupt()) {
		atomic_inc(&netstamp_needed_deferred);
		return;
	}
#endif
1626
	static_key_slow_dec(&netstamp_needed);
L
Linus Torvalds 已提交
1627
}
E
Eric Dumazet 已提交
1628
EXPORT_SYMBOL(net_disable_timestamp);
L
Linus Torvalds 已提交
1629

E
Eric Dumazet 已提交
1630
static inline void net_timestamp_set(struct sk_buff *skb)
L
Linus Torvalds 已提交
1631
{
1632
	skb->tstamp.tv64 = 0;
1633
	if (static_key_false(&netstamp_needed))
1634
		__net_timestamp(skb);
L
Linus Torvalds 已提交
1635 1636
}

1637
#define net_timestamp_check(COND, SKB)			\
1638
	if (static_key_false(&netstamp_needed)) {		\
1639 1640 1641
		if ((COND) && !(SKB)->tstamp.tv64)	\
			__net_timestamp(SKB);		\
	}						\
E
Eric Dumazet 已提交
1642

1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663
static inline bool is_skb_forwardable(struct net_device *dev,
				      struct sk_buff *skb)
{
	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;
}

1664 1665 1666 1667 1668 1669 1670 1671
/**
 * dev_forward_skb - loopback an skb to another netif
 *
 * @dev: destination network device
 * @skb: buffer to forward
 *
 * return values:
 *	NET_RX_SUCCESS	(no congestion)
1672
 *	NET_RX_DROP     (packet was dropped, but freed)
1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683
 *
 * 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)
{
1684 1685 1686 1687 1688 1689 1690 1691
	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;
		}
	}

1692
	if (unlikely(!is_skb_forwardable(dev, skb))) {
1693
		atomic_long_inc(&dev->rx_dropped);
1694
		kfree_skb(skb);
1695
		return NET_RX_DROP;
1696
	}
1697

1698
	skb_scrub_packet(skb, true);
1699
	skb->protocol = eth_type_trans(skb, dev);
1700

1701 1702 1703 1704
	return netif_rx(skb);
}
EXPORT_SYMBOL_GPL(dev_forward_skb);

1705 1706 1707 1708
static inline int deliver_skb(struct sk_buff *skb,
			      struct packet_type *pt_prev,
			      struct net_device *orig_dev)
{
1709 1710
	if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
		return -ENOMEM;
1711 1712 1713 1714
	atomic_inc(&skb->users);
	return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
}

1715 1716
static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
{
1717
	if (!ptype->af_packet_priv || !skb->sk)
1718 1719 1720 1721 1722 1723 1724 1725 1726 1727
		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
1728 1729 1730 1731 1732
/*
 *	Support routine. Sends outgoing frames to any network
 *	taps currently in use.
 */

1733
static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
L
Linus Torvalds 已提交
1734 1735
{
	struct packet_type *ptype;
1736 1737
	struct sk_buff *skb2 = NULL;
	struct packet_type *pt_prev = NULL;
1738

L
Linus Torvalds 已提交
1739 1740 1741 1742 1743 1744
	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) &&
1745
		    (!skb_loop_sk(ptype, skb))) {
1746 1747 1748 1749 1750 1751 1752
			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
1753 1754 1755
			if (!skb2)
				break;

1756 1757
			net_timestamp_set(skb2);

L
Linus Torvalds 已提交
1758 1759 1760 1761
			/* skb->nh should be correctly
			   set by sender, so that the second statement is
			   just protection against buggy protocols.
			 */
1762
			skb_reset_mac_header(skb2);
L
Linus Torvalds 已提交
1763

1764
			if (skb_network_header(skb2) < skb2->data ||
1765
			    skb_network_header(skb2) > skb_tail_pointer(skb2)) {
1766 1767 1768
				net_crit_ratelimited("protocol %04x is buggy, dev %s\n",
						     ntohs(skb2->protocol),
						     dev->name);
1769
				skb_reset_network_header(skb2);
L
Linus Torvalds 已提交
1770 1771
			}

1772
			skb2->transport_header = skb2->network_header;
L
Linus Torvalds 已提交
1773
			skb2->pkt_type = PACKET_OUTGOING;
1774
			pt_prev = ptype;
L
Linus Torvalds 已提交
1775 1776
		}
	}
1777 1778
	if (pt_prev)
		pt_prev->func(skb2, skb->dev, pt_prev, skb->dev);
L
Linus Torvalds 已提交
1779 1780 1781
	rcu_read_unlock();
}

1782 1783
/**
 * netif_setup_tc - Handle tc mappings on real_num_tx_queues change
1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794
 * @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
1795
static void netif_setup_tc(struct net_device *dev, unsigned int txq)
1796 1797 1798 1799 1800 1801
{
	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) {
1802
		pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n");
1803 1804 1805 1806 1807 1808 1809 1810 1811 1812
		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) {
1813 1814
			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);
1815 1816 1817 1818 1819
			netdev_set_prio_tc_map(dev, i, 0);
		}
	}
}

1820 1821 1822 1823 1824
#ifdef CONFIG_XPS
static DEFINE_MUTEX(xps_map_mutex);
#define xmap_dereference(P)		\
	rcu_dereference_protected((P), lockdep_is_held(&xps_map_mutex))

1825 1826
static struct xps_map *remove_xps_queue(struct xps_dev_maps *dev_maps,
					int cpu, u16 index)
1827
{
1828 1829
	struct xps_map *map = NULL;
	int pos;
1830

1831 1832
	if (dev_maps)
		map = xmap_dereference(dev_maps->cpu_map[cpu]);
1833

1834 1835
	for (pos = 0; map && pos < map->len; pos++) {
		if (map->queues[pos] == index) {
1836 1837 1838
			if (map->len > 1) {
				map->queues[pos] = map->queues[--map->len];
			} else {
1839
				RCU_INIT_POINTER(dev_maps->cpu_map[cpu], NULL);
1840 1841 1842
				kfree_rcu(map, rcu);
				map = NULL;
			}
1843
			break;
1844 1845 1846
		}
	}

1847 1848 1849
	return map;
}

1850
static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
1851 1852
{
	struct xps_dev_maps *dev_maps;
1853
	int cpu, i;
1854 1855 1856 1857 1858 1859 1860 1861 1862
	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) {
1863 1864 1865 1866 1867
		for (i = index; i < dev->num_tx_queues; i++) {
			if (!remove_xps_queue(dev_maps, cpu, i))
				break;
		}
		if (i == dev->num_tx_queues)
1868 1869 1870 1871
			active = true;
	}

	if (!active) {
1872 1873 1874 1875
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

1876 1877 1878 1879
	for (i = index; i < dev->num_tx_queues; i++)
		netdev_queue_numa_node_write(netdev_get_tx_queue(dev, i),
					     NUMA_NO_NODE);

1880 1881 1882 1883
out_no_maps:
	mutex_unlock(&xps_map_mutex);
}

1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918
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;
}

1919 1920
int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
			u16 index)
1921
{
1922
	struct xps_dev_maps *dev_maps, *new_dev_maps = NULL;
1923 1924
	struct xps_map *map, *new_map;
	int maps_sz = max_t(unsigned int, XPS_DEV_MAPS_SIZE, L1_CACHE_BYTES);
1925 1926
	int cpu, numa_node_id = -2;
	bool active = false;
1927 1928 1929 1930 1931

	mutex_lock(&xps_map_mutex);

	dev_maps = xmap_dereference(dev->xps_maps);

1932 1933 1934 1935 1936 1937 1938
	/* 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);
1939 1940
		if (!new_dev_maps) {
			mutex_unlock(&xps_map_mutex);
1941
			return -ENOMEM;
1942
		}
1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956

		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;

1957
	for_each_possible_cpu(cpu) {
1958 1959 1960 1961 1962 1963 1964 1965 1966 1967
		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;
1968 1969 1970 1971 1972 1973
#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
1974 1975 1976 1977
		} 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);
1978
		}
1979

1980 1981
	}

1982 1983
	rcu_assign_pointer(dev->xps_maps, new_dev_maps);

1984
	/* Cleanup old maps */
1985 1986 1987 1988 1989 1990 1991
	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);
		}
1992

1993
		kfree_rcu(dev_maps, rcu);
1994 1995
	}

1996 1997
	dev_maps = new_dev_maps;
	active = true;
1998

1999 2000
out_no_new_maps:
	/* update Tx queue numa node */
2001 2002 2003 2004
	netdev_queue_numa_node_write(netdev_get_tx_queue(dev, index),
				     (numa_node_id >= 0) ? numa_node_id :
				     NUMA_NO_NODE);

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023
	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:
2024 2025 2026 2027
	mutex_unlock(&xps_map_mutex);

	return 0;
error:
2028 2029 2030 2031 2032 2033 2034 2035 2036
	/* 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);
	}

2037 2038 2039 2040 2041 2042 2043 2044
	mutex_unlock(&xps_map_mutex);

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

#endif
2045 2046 2047 2048
/*
 * 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.
 */
2049
int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
2050
{
T
Tom Herbert 已提交
2051 2052
	int rc;

2053 2054
	if (txq < 1 || txq > dev->num_tx_queues)
		return -EINVAL;
2055

2056 2057
	if (dev->reg_state == NETREG_REGISTERED ||
	    dev->reg_state == NETREG_UNREGISTERING) {
2058 2059
		ASSERT_RTNL();

T
Tom Herbert 已提交
2060 2061
		rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
						  txq);
T
Tom Herbert 已提交
2062 2063 2064
		if (rc)
			return rc;

2065 2066 2067
		if (dev->num_tc)
			netif_setup_tc(dev, txq);

2068
		if (txq < dev->real_num_tx_queues) {
2069
			qdisc_reset_all_tx_gt(dev, txq);
2070 2071 2072 2073
#ifdef CONFIG_XPS
			netif_reset_xps_queues_gt(dev, txq);
#endif
		}
2074
	}
2075 2076 2077

	dev->real_num_tx_queues = txq;
	return 0;
2078 2079
}
EXPORT_SYMBOL(netif_set_real_num_tx_queues);
2080

2081 2082 2083 2084 2085 2086 2087 2088
#ifdef CONFIG_RPS
/**
 *	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
2089 2090
 *	negative error code.  If called before registration, it always
 *	succeeds.
2091 2092 2093 2094 2095
 */
int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
{
	int rc;

T
Tom Herbert 已提交
2096 2097 2098
	if (rxq < 1 || rxq > dev->num_rx_queues)
		return -EINVAL;

2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113
	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

2114 2115
/**
 * netif_get_num_default_rss_queues - default number of RSS queues
2116 2117 2118 2119
 *
 * This routine should set an upper limit on the number of RSS queues
 * used by default by multiqueue devices.
 */
2120
int netif_get_num_default_rss_queues(void)
2121 2122 2123 2124 2125
{
	return min_t(int, DEFAULT_MAX_NUM_RSS_QUEUES, num_online_cpus());
}
EXPORT_SYMBOL(netif_get_num_default_rss_queues);

2126
static inline void __netif_reschedule(struct Qdisc *q)
2127
{
2128 2129
	struct softnet_data *sd;
	unsigned long flags;
2130

2131 2132
	local_irq_save(flags);
	sd = &__get_cpu_var(softnet_data);
2133 2134 2135
	q->next_sched = NULL;
	*sd->output_queue_tailp = q;
	sd->output_queue_tailp = &q->next_sched;
2136 2137 2138 2139 2140 2141 2142 2143
	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);
2144 2145 2146
}
EXPORT_SYMBOL(__netif_schedule);

2147 2148 2149 2150 2151 2152 2153 2154 2155 2156
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)
{
	return (struct dev_kfree_skb_cb *)skb->cb;
}

void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason)
2157
{
2158
	unsigned long flags;
2159

2160 2161 2162 2163 2164
	if (likely(atomic_read(&skb->users) == 1)) {
		smp_rmb();
		atomic_set(&skb->users, 0);
	} else if (likely(!atomic_dec_and_test(&skb->users))) {
		return;
2165
	}
2166 2167 2168 2169 2170 2171
	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);
2172
}
2173
EXPORT_SYMBOL(__dev_kfree_skb_irq);
2174

2175
void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason)
2176 2177
{
	if (in_irq() || irqs_disabled())
2178
		__dev_kfree_skb_irq(skb, reason);
2179 2180 2181
	else
		dev_kfree_skb(skb);
}
2182
EXPORT_SYMBOL(__dev_kfree_skb_any);
2183 2184


2185 2186 2187 2188 2189 2190
/**
 * netif_device_detach - mark device as removed
 * @dev: network device
 *
 * Mark device as removed from system and therefore no longer available.
 */
2191 2192 2193 2194
void netif_device_detach(struct net_device *dev)
{
	if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2195
		netif_tx_stop_all_queues(dev);
2196 2197 2198 2199
	}
}
EXPORT_SYMBOL(netif_device_detach);

2200 2201 2202 2203 2204 2205
/**
 * netif_device_attach - mark device as attached
 * @dev: network device
 *
 * Mark device as attached from system and restart if needed.
 */
2206 2207 2208 2209
void netif_device_attach(struct net_device *dev)
{
	if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2210
		netif_tx_wake_all_queues(dev);
2211
		__netdev_watchdog_up(dev);
2212 2213 2214 2215
	}
}
EXPORT_SYMBOL(netif_device_attach);

2216 2217
static void skb_warn_bad_offload(const struct sk_buff *skb)
{
2218
	static const netdev_features_t null_features = 0;
2219 2220 2221
	struct net_device *dev = skb->dev;
	const char *driver = "";

2222 2223 2224
	if (!net_ratelimit())
		return;

2225 2226 2227 2228 2229
	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",
2230 2231
	     driver, dev ? &dev->features : &null_features,
	     skb->sk ? &skb->sk->sk_route_caps : &null_features,
2232 2233 2234 2235
	     skb->len, skb->data_len, skb_shinfo(skb)->gso_size,
	     skb_shinfo(skb)->gso_type, skb->ip_summed);
}

L
Linus Torvalds 已提交
2236 2237 2238 2239
/*
 * Invalidate hardware checksum when packet is to be mangled, and
 * complete checksum manually on outgoing path.
 */
2240
int skb_checksum_help(struct sk_buff *skb)
L
Linus Torvalds 已提交
2241
{
2242
	__wsum csum;
2243
	int ret = 0, offset;
L
Linus Torvalds 已提交
2244

2245
	if (skb->ip_summed == CHECKSUM_COMPLETE)
2246 2247 2248
		goto out_set_summed;

	if (unlikely(skb_shinfo(skb)->gso_size)) {
2249 2250
		skb_warn_bad_offload(skb);
		return -EINVAL;
L
Linus Torvalds 已提交
2251 2252
	}

2253 2254 2255 2256 2257 2258 2259 2260 2261
	/* 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;
	}

2262
	offset = skb_checksum_start_offset(skb);
2263 2264 2265 2266 2267 2268 2269 2270
	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
2271 2272 2273 2274 2275
		ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
		if (ret)
			goto out;
	}

2276
	*(__sum16 *)(skb->data + offset) = csum_fold(csum);
2277
out_set_summed:
L
Linus Torvalds 已提交
2278
	skb->ip_summed = CHECKSUM_NONE;
2279
out:
L
Linus Torvalds 已提交
2280 2281
	return ret;
}
E
Eric Dumazet 已提交
2282
EXPORT_SYMBOL(skb_checksum_help);
L
Linus Torvalds 已提交
2283

2284
__be16 skb_network_protocol(struct sk_buff *skb)
2285
{
A
Al Viro 已提交
2286
	__be16 type = skb->protocol;
2287
	int vlan_depth = ETH_HLEN;
2288

2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299
	/* 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;
	}

P
Patrick McHardy 已提交
2300
	while (type == htons(ETH_P_8021Q) || type == htons(ETH_P_8021AD)) {
2301
		struct vlan_hdr *vh;
2302

2303
		if (unlikely(!pskb_may_pull(skb, vlan_depth + VLAN_HLEN)))
2304
			return 0;
2305

2306 2307 2308
		vh = (struct vlan_hdr *)(skb->data + vlan_depth);
		type = vh->h_vlan_encapsulated_proto;
		vlan_depth += VLAN_HLEN;
2309 2310
	}

2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328
	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;
	__be16 type = skb_network_protocol(skb);

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

2329 2330 2331
	__skb_pull(skb, skb->mac_len);

	rcu_read_lock();
2332
	list_for_each_entry_rcu(ptype, &offload_base, list) {
2333
		if (ptype->type == type && ptype->callbacks.gso_segment) {
2334
			if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) {
2335 2336
				int err;

2337
				err = ptype->callbacks.gso_send_check(skb);
2338 2339 2340
				segs = ERR_PTR(err);
				if (err || skb_gso_ok(skb, features))
					break;
2341 2342
				__skb_push(skb, (skb->data -
						 skb_network_header(skb)));
2343
			}
2344
			segs = ptype->callbacks.gso_segment(skb, features);
2345 2346 2347 2348 2349
			break;
		}
	}
	rcu_read_unlock();

2350
	__skb_push(skb, skb->data - skb_mac_header(skb));
2351

2352 2353
	return segs;
}
2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390
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);

		if (skb_header_cloned(skb) &&
		    (err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
			return ERR_PTR(err);
	}

2391
	SKB_GSO_CB(skb)->mac_offset = skb_headroom(skb);
2392 2393
	SKB_GSO_CB(skb)->encap_level = 0;

2394 2395 2396 2397 2398
	skb_reset_mac_header(skb);
	skb_reset_mac_len(skb);

	return skb_mac_gso_segment(skb, features);
}
2399
EXPORT_SYMBOL(__skb_gso_segment);
2400

2401 2402 2403 2404 2405
/* Take action when hardware reception checksum errors are detected. */
#ifdef CONFIG_BUG
void netdev_rx_csum_fault(struct net_device *dev)
{
	if (net_ratelimit()) {
2406
		pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>");
2407 2408 2409 2410 2411 2412
		dump_stack();
	}
}
EXPORT_SYMBOL(netdev_rx_csum_fault);
#endif

L
Linus Torvalds 已提交
2413 2414 2415 2416 2417
/* 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.
 */

E
Eric Dumazet 已提交
2418
static int illegal_highdma(struct net_device *dev, struct sk_buff *skb)
L
Linus Torvalds 已提交
2419
{
2420
#ifdef CONFIG_HIGHMEM
L
Linus Torvalds 已提交
2421
	int i;
2422
	if (!(dev->features & NETIF_F_HIGHDMA)) {
2423 2424 2425
		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)))
2426
				return 1;
2427
		}
2428
	}
L
Linus Torvalds 已提交
2429

2430 2431
	if (PCI_DMA_BUS_IS_PHYS) {
		struct device *pdev = dev->dev.parent;
L
Linus Torvalds 已提交
2432

E
Eric Dumazet 已提交
2433 2434
		if (!pdev)
			return 0;
2435
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2436 2437
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			dma_addr_t addr = page_to_phys(skb_frag_page(frag));
2438 2439 2440 2441
			if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask)
				return 1;
		}
	}
2442
#endif
L
Linus Torvalds 已提交
2443 2444 2445
	return 0;
}

2446 2447 2448 2449 2450 2451 2452 2453 2454 2455
struct dev_gso_cb {
	void (*destructor)(struct sk_buff *skb);
};

#define DEV_GSO_CB(skb) ((struct dev_gso_cb *)(skb)->cb)

static void dev_gso_skb_destructor(struct sk_buff *skb)
{
	struct dev_gso_cb *cb;

E
Eric Dumazet 已提交
2456 2457
	kfree_skb_list(skb->next);
	skb->next = NULL;
2458 2459 2460 2461 2462 2463 2464 2465 2466

	cb = DEV_GSO_CB(skb);
	if (cb->destructor)
		cb->destructor(skb);
}

/**
 *	dev_gso_segment - Perform emulated hardware segmentation on skb.
 *	@skb: buffer to segment
2467
 *	@features: device features as applicable to this skb
2468 2469 2470 2471
 *
 *	This function segments the given skb and stores the list of segments
 *	in skb->next.
 */
2472
static int dev_gso_segment(struct sk_buff *skb, netdev_features_t features)
2473 2474
{
	struct sk_buff *segs;
2475 2476 2477 2478 2479 2480

	segs = skb_gso_segment(skb, features);

	/* Verifying header integrity only. */
	if (!segs)
		return 0;
2481

2482
	if (IS_ERR(segs))
2483 2484 2485 2486 2487 2488 2489 2490 2491
		return PTR_ERR(segs);

	skb->next = segs;
	DEV_GSO_CB(skb)->destructor = skb->destructor;
	skb->destructor = dev_gso_skb_destructor;

	return 0;
}

2492
static netdev_features_t harmonize_features(struct sk_buff *skb,
2493
	netdev_features_t features)
2494
{
2495
	if (skb->ip_summed != CHECKSUM_NONE &&
2496
	    !can_checksum_protocol(features, skb_network_protocol(skb))) {
2497 2498 2499 2500 2501 2502 2503 2504
		features &= ~NETIF_F_ALL_CSUM;
	} else if (illegal_highdma(skb->dev, skb)) {
		features &= ~NETIF_F_SG;
	}

	return features;
}

2505
netdev_features_t netif_skb_features(struct sk_buff *skb)
2506 2507
{
	__be16 protocol = skb->protocol;
2508
	netdev_features_t features = skb->dev->features;
2509

2510 2511 2512
	if (skb_shinfo(skb)->gso_segs > skb->dev->gso_max_segs)
		features &= ~NETIF_F_GSO_MASK;

P
Patrick McHardy 已提交
2513
	if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD)) {
2514 2515
		struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
		protocol = veh->h_vlan_encapsulated_proto;
2516
	} else if (!vlan_tx_tag_present(skb)) {
2517
		return harmonize_features(skb, features);
2518
	}
2519

P
Patrick McHardy 已提交
2520 2521
	features &= (skb->dev->vlan_features | NETIF_F_HW_VLAN_CTAG_TX |
					       NETIF_F_HW_VLAN_STAG_TX);
2522

2523
	if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD))
2524
		features &= NETIF_F_SG | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST |
P
Patrick McHardy 已提交
2525 2526
				NETIF_F_GEN_CSUM | NETIF_F_HW_VLAN_CTAG_TX |
				NETIF_F_HW_VLAN_STAG_TX;
2527 2528

	return harmonize_features(skb, features);
2529
}
2530
EXPORT_SYMBOL(netif_skb_features);
2531

2532
int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2533
			struct netdev_queue *txq, void *accel_priv)
2534
{
2535
	const struct net_device_ops *ops = dev->netdev_ops;
2536
	int rc = NETDEV_TX_OK;
2537
	unsigned int skb_len;
2538

2539
	if (likely(!skb->next)) {
2540
		netdev_features_t features;
2541

2542
		/*
L
Lucas De Marchi 已提交
2543
		 * If device doesn't need skb->dst, release it right now while
2544 2545
		 * its hot in this cpu cache
		 */
E
Eric Dumazet 已提交
2546 2547 2548
		if (dev->priv_flags & IFF_XMIT_DST_RELEASE)
			skb_dst_drop(skb);

2549 2550
		features = netif_skb_features(skb);

2551
		if (vlan_tx_tag_present(skb) &&
2552 2553 2554
		    !vlan_hw_offload_capable(features, skb->vlan_proto)) {
			skb = __vlan_put_tag(skb, skb->vlan_proto,
					     vlan_tx_tag_get(skb));
2555 2556 2557 2558 2559 2560
			if (unlikely(!skb))
				goto out;

			skb->vlan_tci = 0;
		}

2561 2562 2563 2564 2565 2566 2567
		/* 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;

2568
		if (netif_needs_gso(skb, features)) {
2569
			if (unlikely(dev_gso_segment(skb, features)))
2570 2571 2572
				goto out_kfree_skb;
			if (skb->next)
				goto gso;
2573
		} else {
2574
			if (skb_needs_linearize(skb, features) &&
2575 2576 2577 2578 2579 2580 2581 2582
			    __skb_linearize(skb))
				goto out_kfree_skb;

			/* If packet is not checksummed and device does not
			 * support checksumming for this protocol, complete
			 * checksumming here.
			 */
			if (skb->ip_summed == CHECKSUM_PARTIAL) {
2583 2584 2585 2586 2587 2588
				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));
2589
				if (!(features & NETIF_F_ALL_CSUM) &&
2590 2591 2592
				     skb_checksum_help(skb))
					goto out_kfree_skb;
			}
2593 2594
		}

2595 2596 2597
		if (!list_empty(&ptype_all))
			dev_queue_xmit_nit(skb, dev);

2598
		skb_len = skb->len;
2599 2600 2601 2602 2603
		if (accel_priv)
			rc = ops->ndo_dfwd_start_xmit(skb, dev, accel_priv);
		else
			rc = ops->ndo_start_xmit(skb, dev);

2604
		trace_net_dev_xmit(skb, rc, dev, skb_len);
2605
		if (rc == NETDEV_TX_OK && txq)
E
Eric Dumazet 已提交
2606
			txq_trans_update(txq);
2607
		return rc;
2608 2609
	}

2610
gso:
2611 2612 2613 2614 2615
	do {
		struct sk_buff *nskb = skb->next;

		skb->next = nskb->next;
		nskb->next = NULL;
2616

2617 2618 2619
		if (!list_empty(&ptype_all))
			dev_queue_xmit_nit(nskb, dev);

2620
		skb_len = nskb->len;
2621 2622 2623 2624
		if (accel_priv)
			rc = ops->ndo_dfwd_start_xmit(nskb, dev, accel_priv);
		else
			rc = ops->ndo_start_xmit(nskb, dev);
2625
		trace_net_dev_xmit(nskb, rc, dev, skb_len);
2626
		if (unlikely(rc != NETDEV_TX_OK)) {
2627 2628
			if (rc & ~NETDEV_TX_MASK)
				goto out_kfree_gso_skb;
2629
			nskb->next = skb->next;
2630 2631 2632
			skb->next = nskb;
			return rc;
		}
E
Eric Dumazet 已提交
2633
		txq_trans_update(txq);
2634
		if (unlikely(netif_xmit_stopped(txq) && skb->next))
2635
			return NETDEV_TX_BUSY;
2636
	} while (skb->next);
2637

2638
out_kfree_gso_skb:
2639
	if (likely(skb->next == NULL)) {
2640
		skb->destructor = DEV_GSO_CB(skb)->destructor;
2641 2642 2643
		consume_skb(skb);
		return rc;
	}
2644 2645
out_kfree_skb:
	kfree_skb(skb);
2646
out:
2647
	return rc;
2648
}
2649
EXPORT_SYMBOL_GPL(dev_hard_start_xmit);
2650

2651 2652 2653 2654 2655 2656 2657 2658 2659 2660
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)  {
2661
		unsigned int hdr_len;
2662
		u16 gso_segs = shinfo->gso_segs;
2663

2664 2665 2666 2667
		/* mac layer + network layer */
		hdr_len = skb_transport_header(skb) - skb_mac_header(skb);

		/* + transport layer */
2668 2669 2670 2671
		if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)))
			hdr_len += tcp_hdrlen(skb);
		else
			hdr_len += sizeof(struct udphdr);
2672 2673 2674 2675 2676 2677

		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;
2678 2679 2680
	}
}

2681 2682 2683 2684 2685
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
2686
	bool contended;
2687 2688
	int rc;

2689
	qdisc_pkt_len_init(skb);
E
Eric Dumazet 已提交
2690
	qdisc_calculate_pkt_len(skb, q);
2691 2692 2693 2694 2695 2696
	/*
	 * Heuristic to force contended enqueues to serialize on a
	 * separate lock before trying to get qdisc main lock.
	 * This permits __QDISC_STATE_RUNNING owner to get the lock more often
	 * and dequeue packets faster.
	 */
E
Eric Dumazet 已提交
2697
	contended = qdisc_is_running(q);
2698 2699 2700
	if (unlikely(contended))
		spin_lock(&q->busylock);

2701 2702 2703 2704 2705
	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) &&
2706
		   qdisc_run_begin(q)) {
2707 2708 2709 2710 2711
		/*
		 * 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.
		 */
E
Eric Dumazet 已提交
2712 2713
		if (!(dev->priv_flags & IFF_XMIT_DST_RELEASE))
			skb_dst_force(skb);
2714 2715 2716

		qdisc_bstats_update(q, skb);

2717 2718 2719 2720 2721
		if (sch_direct_xmit(skb, q, dev, txq, root_lock)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
2722
			__qdisc_run(q);
2723
		} else
2724
			qdisc_run_end(q);
2725 2726 2727

		rc = NET_XMIT_SUCCESS;
	} else {
E
Eric Dumazet 已提交
2728
		skb_dst_force(skb);
E
Eric Dumazet 已提交
2729
		rc = q->enqueue(skb, q) & NET_XMIT_MASK;
2730 2731 2732 2733 2734 2735 2736
		if (qdisc_run_begin(q)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
			__qdisc_run(q);
		}
2737 2738
	}
	spin_unlock(root_lock);
2739 2740
	if (unlikely(contended))
		spin_unlock(&q->busylock);
2741 2742 2743
	return rc;
}

2744 2745 2746
#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
static void skb_update_prio(struct sk_buff *skb)
{
I
Igor Maravic 已提交
2747
	struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap);
2748

2749 2750 2751 2752 2753 2754
	if (!skb->priority && skb->sk && map) {
		unsigned int prioidx = skb->sk->sk_cgrp_prioidx;

		if (prioidx < map->priomap_len)
			skb->priority = map->priomap[prioidx];
	}
2755 2756 2757 2758 2759
}
#else
#define skb_update_prio(skb)
#endif

2760
static DEFINE_PER_CPU(int, xmit_recursion);
2761
#define RECURSION_LIMIT 10
2762

2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779
/**
 *	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);

2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804
/**
 *	dev_queue_xmit - transmit a buffer
 *	@skb: buffer to transmit
 *
 *	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
 */
L
Linus Torvalds 已提交
2805 2806 2807
int dev_queue_xmit(struct sk_buff *skb)
{
	struct net_device *dev = skb->dev;
2808
	struct netdev_queue *txq;
L
Linus Torvalds 已提交
2809 2810 2811
	struct Qdisc *q;
	int rc = -ENOMEM;

2812 2813
	skb_reset_mac_header(skb);

2814 2815
	/* Disable soft irqs for various locks below. Also
	 * stops preemption for RCU.
L
Linus Torvalds 已提交
2816
	 */
2817
	rcu_read_lock_bh();
L
Linus Torvalds 已提交
2818

2819 2820
	skb_update_prio(skb);

2821
	txq = netdev_pick_tx(dev, skb);
2822
	q = rcu_dereference_bh(txq->qdisc);
2823

L
Linus Torvalds 已提交
2824
#ifdef CONFIG_NET_CLS_ACT
E
Eric Dumazet 已提交
2825
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS);
L
Linus Torvalds 已提交
2826
#endif
2827
	trace_net_dev_queue(skb);
L
Linus Torvalds 已提交
2828
	if (q->enqueue) {
2829
		rc = __dev_xmit_skb(skb, q, dev, txq);
2830
		goto out;
L
Linus Torvalds 已提交
2831 2832 2833 2834 2835
	}

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

H
Herbert Xu 已提交
2836 2837
	   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
2838 2839 2840 2841 2842 2843 2844 2845 2846 2847
	   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 */

2848
		if (txq->xmit_lock_owner != cpu) {
L
Linus Torvalds 已提交
2849

2850 2851 2852
			if (__this_cpu_read(xmit_recursion) > RECURSION_LIMIT)
				goto recursion_alert;

2853
			HARD_TX_LOCK(dev, txq, cpu);
L
Linus Torvalds 已提交
2854

2855
			if (!netif_xmit_stopped(txq)) {
2856
				__this_cpu_inc(xmit_recursion);
2857
				rc = dev_hard_start_xmit(skb, dev, txq, NULL);
2858
				__this_cpu_dec(xmit_recursion);
2859
				if (dev_xmit_complete(rc)) {
2860
					HARD_TX_UNLOCK(dev, txq);
L
Linus Torvalds 已提交
2861 2862 2863
					goto out;
				}
			}
2864
			HARD_TX_UNLOCK(dev, txq);
2865 2866
			net_crit_ratelimited("Virtual device %s asks to queue packet!\n",
					     dev->name);
L
Linus Torvalds 已提交
2867 2868
		} else {
			/* Recursion is detected! It is possible,
2869 2870 2871
			 * unfortunately
			 */
recursion_alert:
2872 2873
			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
					     dev->name);
L
Linus Torvalds 已提交
2874 2875 2876 2877
		}
	}

	rc = -ENETDOWN;
2878
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
2879 2880 2881 2882

	kfree_skb(skb);
	return rc;
out:
2883
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
2884 2885
	return rc;
}
E
Eric Dumazet 已提交
2886
EXPORT_SYMBOL(dev_queue_xmit);
L
Linus Torvalds 已提交
2887 2888 2889 2890 2891 2892


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

2893
int netdev_max_backlog __read_mostly = 1000;
E
Eric Dumazet 已提交
2894 2895
EXPORT_SYMBOL(netdev_max_backlog);

E
Eric Dumazet 已提交
2896
int netdev_tstamp_prequeue __read_mostly = 1;
2897 2898
int netdev_budget __read_mostly = 300;
int weight_p __read_mostly = 64;            /* old backlog weight */
L
Linus Torvalds 已提交
2899

E
Eric Dumazet 已提交
2900 2901 2902 2903 2904 2905 2906 2907
/* 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);
}

2908 2909 2910
#ifdef CONFIG_RPS

/* One global table that all flow-based protocols share. */
E
Eric Dumazet 已提交
2911
struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
2912 2913
EXPORT_SYMBOL(rps_sock_flow_table);

2914
struct static_key rps_needed __read_mostly;
2915

2916 2917 2918 2919
static struct rps_dev_flow *
set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
	    struct rps_dev_flow *rflow, u16 next_cpu)
{
2920
	if (next_cpu != RPS_NO_CPU) {
2921 2922 2923 2924 2925 2926 2927 2928 2929
#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? */
2930 2931
		if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
		    !(dev->features & NETIF_F_NTUPLE))
2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953
			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;
		flow_id = skb->rxhash & flow_table->mask;
		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 =
2954
			per_cpu(softnet_data, next_cpu).input_queue_head;
2955 2956
	}

2957
	rflow->cpu = next_cpu;
2958 2959 2960
	return rflow;
}

2961 2962 2963 2964 2965 2966 2967 2968 2969
/*
 * 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
2970
	struct rps_map *map;
2971 2972 2973 2974 2975 2976 2977
	struct rps_dev_flow_table *flow_table;
	struct rps_sock_flow_table *sock_flow_table;
	int cpu = -1;
	u16 tcpu;

	if (skb_rx_queue_recorded(skb)) {
		u16 index = skb_get_rx_queue(skb);
2978 2979 2980 2981 2982
		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);
2983 2984 2985 2986 2987 2988
			goto done;
		}
		rxqueue = dev->_rx + index;
	} else
		rxqueue = dev->_rx;

E
Eric Dumazet 已提交
2989 2990
	map = rcu_dereference(rxqueue->rps_map);
	if (map) {
2991
		if (map->len == 1 &&
2992
		    !rcu_access_pointer(rxqueue->rps_flow_table)) {
2993 2994 2995 2996 2997
			tcpu = map->cpus[0];
			if (cpu_online(tcpu))
				cpu = tcpu;
			goto done;
		}
2998
	} else if (!rcu_access_pointer(rxqueue->rps_flow_table)) {
2999
		goto done;
3000
	}
3001

3002
	skb_reset_network_header(skb);
3003
	if (!skb_get_hash(skb))
3004 3005
		goto done;

T
Tom Herbert 已提交
3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031
	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;

		rflow = &flow_table->flows[skb->rxhash & flow_table->mask];
		tcpu = rflow->cpu;

		next_cpu = sock_flow_table->ents[skb->rxhash &
		    sock_flow_table->mask];

		/*
		 * 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 -
3032 3033
		      rflow->last_qtail)) >= 0)) {
			tcpu = next_cpu;
3034
			rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
3035
		}
3036

T
Tom Herbert 已提交
3037 3038 3039 3040 3041 3042 3043
		if (tcpu != RPS_NO_CPU && cpu_online(tcpu)) {
			*rflowp = rflow;
			cpu = tcpu;
			goto done;
		}
	}

T
Tom Herbert 已提交
3044
	if (map) {
T
Tom Herbert 已提交
3045
		tcpu = map->cpus[((u64) skb->rxhash * map->len) >> 32];
T
Tom Herbert 已提交
3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056

		if (cpu_online(tcpu)) {
			cpu = tcpu;
			goto done;
		}
	}

done:
	return cpu;
}

3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096
#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
3097
/* Called from hardirq (IPI) context */
E
Eric Dumazet 已提交
3098
static void rps_trigger_softirq(void *data)
T
Tom Herbert 已提交
3099
{
E
Eric Dumazet 已提交
3100 3101
	struct softnet_data *sd = data;

E
Eric Dumazet 已提交
3102
	____napi_schedule(sd, &sd->backlog);
C
Changli Gao 已提交
3103
	sd->received_rps++;
T
Tom Herbert 已提交
3104
}
E
Eric Dumazet 已提交
3105

T
Tom Herbert 已提交
3106
#endif /* CONFIG_RPS */
T
Tom Herbert 已提交
3107

E
Eric Dumazet 已提交
3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128
/*
 * 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
	struct softnet_data *mysd = &__get_cpu_var(softnet_data);

	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;
}

3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147
#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;

	sd = &__get_cpu_var(softnet_data);

	rcu_read_lock();
	fl = rcu_dereference(sd->flow_limit);
	if (fl) {
3148
		new_flow = skb_get_hash(skb) & (fl->num_buckets - 1);
3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168
		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
3169 3170 3171 3172
/*
 * 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
3173 3174
static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
			      unsigned int *qtail)
T
Tom Herbert 已提交
3175
{
E
Eric Dumazet 已提交
3176
	struct softnet_data *sd;
T
Tom Herbert 已提交
3177
	unsigned long flags;
3178
	unsigned int qlen;
T
Tom Herbert 已提交
3179

E
Eric Dumazet 已提交
3180
	sd = &per_cpu(softnet_data, cpu);
T
Tom Herbert 已提交
3181 3182 3183

	local_irq_save(flags);

E
Eric Dumazet 已提交
3184
	rps_lock(sd);
3185 3186
	qlen = skb_queue_len(&sd->input_pkt_queue);
	if (qlen <= netdev_max_backlog && !skb_flow_limit(skb, qlen)) {
3187
		if (skb_queue_len(&sd->input_pkt_queue)) {
T
Tom Herbert 已提交
3188
enqueue:
E
Eric Dumazet 已提交
3189
			__skb_queue_tail(&sd->input_pkt_queue, skb);
3190
			input_queue_tail_incr_save(sd, qtail);
E
Eric Dumazet 已提交
3191
			rps_unlock(sd);
3192
			local_irq_restore(flags);
T
Tom Herbert 已提交
3193 3194 3195
			return NET_RX_SUCCESS;
		}

3196 3197 3198 3199
		/* 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
3200
			if (!rps_ipi_queued(sd))
E
Eric Dumazet 已提交
3201
				____napi_schedule(sd, &sd->backlog);
T
Tom Herbert 已提交
3202 3203 3204 3205
		}
		goto enqueue;
	}

C
Changli Gao 已提交
3206
	sd->dropped++;
E
Eric Dumazet 已提交
3207
	rps_unlock(sd);
T
Tom Herbert 已提交
3208 3209 3210

	local_irq_restore(flags);

3211
	atomic_long_inc(&skb->dev->rx_dropped);
T
Tom Herbert 已提交
3212 3213 3214
	kfree_skb(skb);
	return NET_RX_DROP;
}
L
Linus Torvalds 已提交
3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232

/**
 *	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)
{
3233
	int ret;
L
Linus Torvalds 已提交
3234 3235 3236 3237 3238

	/* if netpoll wants it, pretend we never saw it */
	if (netpoll_rx(skb))
		return NET_RX_DROP;

3239
	net_timestamp_check(netdev_tstamp_prequeue, skb);
L
Linus Torvalds 已提交
3240

3241
	trace_netif_rx(skb);
E
Eric Dumazet 已提交
3242
#ifdef CONFIG_RPS
3243
	if (static_key_false(&rps_needed)) {
T
Tom Herbert 已提交
3244
		struct rps_dev_flow voidflow, *rflow = &voidflow;
3245 3246
		int cpu;

3247
		preempt_disable();
3248
		rcu_read_lock();
T
Tom Herbert 已提交
3249 3250

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
3251 3252
		if (cpu < 0)
			cpu = smp_processor_id();
T
Tom Herbert 已提交
3253 3254 3255

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

3256
		rcu_read_unlock();
3257
		preempt_enable();
3258 3259
	} else
#endif
T
Tom Herbert 已提交
3260 3261 3262 3263 3264
	{
		unsigned int qtail;
		ret = enqueue_to_backlog(skb, get_cpu(), &qtail);
		put_cpu();
	}
3265
	return ret;
L
Linus Torvalds 已提交
3266
}
E
Eric Dumazet 已提交
3267
EXPORT_SYMBOL(netif_rx);
L
Linus Torvalds 已提交
3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298

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

	preempt_disable();
	err = netif_rx(skb);
	if (local_softirq_pending())
		do_softirq();
	preempt_enable();

	return err;
}
EXPORT_SYMBOL(netif_rx_ni);

static void net_tx_action(struct softirq_action *h)
{
	struct softnet_data *sd = &__get_cpu_var(softnet_data);

	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;

3299
			WARN_ON(atomic_read(&skb->users));
3300 3301 3302 3303
			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
3304 3305 3306 3307 3308
			__kfree_skb(skb);
		}
	}

	if (sd->output_queue) {
3309
		struct Qdisc *head;
L
Linus Torvalds 已提交
3310 3311 3312 3313

		local_irq_disable();
		head = sd->output_queue;
		sd->output_queue = NULL;
3314
		sd->output_queue_tailp = &sd->output_queue;
L
Linus Torvalds 已提交
3315 3316 3317
		local_irq_enable();

		while (head) {
3318 3319 3320
			struct Qdisc *q = head;
			spinlock_t *root_lock;

L
Linus Torvalds 已提交
3321 3322
			head = head->next_sched;

3323
			root_lock = qdisc_lock(q);
3324
			if (spin_trylock(root_lock)) {
3325 3326 3327
				smp_mb__before_clear_bit();
				clear_bit(__QDISC_STATE_SCHED,
					  &q->state);
3328 3329
				qdisc_run(q);
				spin_unlock(root_lock);
L
Linus Torvalds 已提交
3330
			} else {
3331
				if (!test_bit(__QDISC_STATE_DEACTIVATED,
3332
					      &q->state)) {
3333
					__netif_reschedule(q);
3334 3335 3336 3337 3338
				} else {
					smp_mb__before_clear_bit();
					clear_bit(__QDISC_STATE_SCHED,
						  &q->state);
				}
L
Linus Torvalds 已提交
3339 3340 3341 3342 3343
			}
		}
	}
}

3344 3345
#if (defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)) && \
    (defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE))
3346 3347 3348
/* This hook is defined here for ATM LANE */
int (*br_fdb_test_addr_hook)(struct net_device *dev,
			     unsigned char *addr) __read_mostly;
3349
EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
3350
#endif
L
Linus Torvalds 已提交
3351 3352 3353 3354 3355 3356

#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
3357 3358
 * 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
3359 3360
 *
 */
3361
static int ing_filter(struct sk_buff *skb, struct netdev_queue *rxq)
L
Linus Torvalds 已提交
3362 3363
{
	struct net_device *dev = skb->dev;
3364
	u32 ttl = G_TC_RTTL(skb->tc_verd);
3365 3366
	int result = TC_ACT_OK;
	struct Qdisc *q;
3367

3368
	if (unlikely(MAX_RED_LOOP < ttl++)) {
3369 3370
		net_warn_ratelimited("Redir loop detected Dropping packet (%d->%d)\n",
				     skb->skb_iif, dev->ifindex);
3371 3372
		return TC_ACT_SHOT;
	}
L
Linus Torvalds 已提交
3373

3374 3375
	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
3376

3377
	q = rxq->qdisc;
3378
	if (q != &noop_qdisc) {
3379
		spin_lock(qdisc_lock(q));
3380 3381
		if (likely(!test_bit(__QDISC_STATE_DEACTIVATED, &q->state)))
			result = qdisc_enqueue_root(skb, q);
3382 3383
		spin_unlock(qdisc_lock(q));
	}
3384 3385 3386

	return result;
}
3387

3388 3389 3390 3391
static inline struct sk_buff *handle_ing(struct sk_buff *skb,
					 struct packet_type **pt_prev,
					 int *ret, struct net_device *orig_dev)
{
3392 3393 3394
	struct netdev_queue *rxq = rcu_dereference(skb->dev->ingress_queue);

	if (!rxq || rxq->qdisc == &noop_qdisc)
3395
		goto out;
L
Linus Torvalds 已提交
3396

3397 3398 3399
	if (*pt_prev) {
		*ret = deliver_skb(skb, *pt_prev, orig_dev);
		*pt_prev = NULL;
L
Linus Torvalds 已提交
3400 3401
	}

3402
	switch (ing_filter(skb, rxq)) {
3403 3404 3405 3406 3407 3408 3409 3410 3411
	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
3412 3413 3414
}
#endif

3415 3416 3417 3418
/**
 *	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
3419
 *	@rx_handler_data: data pointer that is used by rx handler
3420 3421 3422 3423 3424 3425
 *
 *	Register a receive hander for a device. This handler will then be
 *	called from __netif_receive_skb. A negative errno code is returned
 *	on a failure.
 *
 *	The caller must hold the rtnl_mutex.
3426 3427
 *
 *	For a general description of rx_handler, see enum rx_handler_result.
3428 3429
 */
int netdev_rx_handler_register(struct net_device *dev,
J
Jiri Pirko 已提交
3430 3431
			       rx_handler_func_t *rx_handler,
			       void *rx_handler_data)
3432 3433 3434 3435 3436 3437
{
	ASSERT_RTNL();

	if (dev->rx_handler)
		return -EBUSY;

3438
	/* Note: rx_handler_data must be set before rx_handler */
J
Jiri Pirko 已提交
3439
	rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
3440 3441 3442 3443 3444 3445 3446 3447 3448 3449
	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
3450
 *	Unregister a receive handler from a device.
3451 3452 3453 3454 3455 3456 3457
 *
 *	The caller must hold the rtnl_mutex.
 */
void netdev_rx_handler_unregister(struct net_device *dev)
{

	ASSERT_RTNL();
3458
	RCU_INIT_POINTER(dev->rx_handler, NULL);
3459 3460 3461 3462 3463
	/* 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();
3464
	RCU_INIT_POINTER(dev->rx_handler_data, NULL);
3465 3466 3467
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);

3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478
/*
 * 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) {
	case __constant_htons(ETH_P_ARP):
	case __constant_htons(ETH_P_IP):
	case __constant_htons(ETH_P_IPV6):
	case __constant_htons(ETH_P_8021Q):
P
Patrick McHardy 已提交
3479
	case __constant_htons(ETH_P_8021AD):
3480 3481 3482 3483 3484 3485
		return true;
	default:
		return false;
	}
}

3486
static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc)
L
Linus Torvalds 已提交
3487 3488
{
	struct packet_type *ptype, *pt_prev;
3489
	rx_handler_func_t *rx_handler;
D
David S. Miller 已提交
3490
	struct net_device *orig_dev;
3491
	struct net_device *null_or_dev;
3492
	bool deliver_exact = false;
L
Linus Torvalds 已提交
3493
	int ret = NET_RX_DROP;
A
Al Viro 已提交
3494
	__be16 type;
L
Linus Torvalds 已提交
3495

3496
	net_timestamp_check(!netdev_tstamp_prequeue, skb);
3497

3498
	trace_netif_receive_skb(skb);
3499

L
Linus Torvalds 已提交
3500
	/* if we've gotten here through NAPI, check netpoll */
3501
	if (netpoll_receive_skb(skb))
3502
		goto out;
L
Linus Torvalds 已提交
3503

J
Joe Eykholt 已提交
3504
	orig_dev = skb->dev;
3505

3506
	skb_reset_network_header(skb);
3507 3508
	if (!skb_transport_header_was_set(skb))
		skb_reset_transport_header(skb);
3509
	skb_reset_mac_len(skb);
L
Linus Torvalds 已提交
3510 3511 3512 3513 3514

	pt_prev = NULL;

	rcu_read_lock();

3515
another_round:
3516
	skb->skb_iif = skb->dev->ifindex;
3517 3518 3519

	__this_cpu_inc(softnet_data.processed);

P
Patrick McHardy 已提交
3520 3521
	if (skb->protocol == cpu_to_be16(ETH_P_8021Q) ||
	    skb->protocol == cpu_to_be16(ETH_P_8021AD)) {
3522 3523
		skb = vlan_untag(skb);
		if (unlikely(!skb))
3524
			goto unlock;
3525 3526
	}

L
Linus Torvalds 已提交
3527 3528 3529 3530 3531 3532 3533
#ifdef CONFIG_NET_CLS_ACT
	if (skb->tc_verd & TC_NCLS) {
		skb->tc_verd = CLR_TC_NCLS(skb->tc_verd);
		goto ncls;
	}
#endif

3534
	if (pfmemalloc)
3535 3536
		goto skip_taps;

L
Linus Torvalds 已提交
3537
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
3538
		if (!ptype->dev || ptype->dev == skb->dev) {
3539
			if (pt_prev)
D
David S. Miller 已提交
3540
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3541 3542 3543 3544
			pt_prev = ptype;
		}
	}

3545
skip_taps:
L
Linus Torvalds 已提交
3546
#ifdef CONFIG_NET_CLS_ACT
3547 3548
	skb = handle_ing(skb, &pt_prev, &ret, orig_dev);
	if (!skb)
3549
		goto unlock;
L
Linus Torvalds 已提交
3550 3551 3552
ncls:
#endif

3553
	if (pfmemalloc && !skb_pfmemalloc_protocol(skb))
3554 3555
		goto drop;

3556 3557 3558 3559 3560
	if (vlan_tx_tag_present(skb)) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3561
		if (vlan_do_receive(&skb))
3562 3563
			goto another_round;
		else if (unlikely(!skb))
3564
			goto unlock;
3565 3566
	}

3567
	rx_handler = rcu_dereference(skb->dev->rx_handler);
3568 3569 3570 3571 3572
	if (rx_handler) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3573 3574
		switch (rx_handler(&skb)) {
		case RX_HANDLER_CONSUMED:
3575
			ret = NET_RX_SUCCESS;
3576
			goto unlock;
3577
		case RX_HANDLER_ANOTHER:
3578
			goto another_round;
3579 3580 3581 3582 3583 3584 3585
		case RX_HANDLER_EXACT:
			deliver_exact = true;
		case RX_HANDLER_PASS:
			break;
		default:
			BUG();
		}
3586
	}
L
Linus Torvalds 已提交
3587

E
Eric Dumazet 已提交
3588 3589 3590 3591 3592 3593 3594 3595 3596
	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;
	}
3597

3598
	/* deliver only exact match when indicated */
3599
	null_or_dev = deliver_exact ? skb->dev : NULL;
3600

L
Linus Torvalds 已提交
3601
	type = skb->protocol;
3602 3603
	list_for_each_entry_rcu(ptype,
			&ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) {
3604
		if (ptype->type == type &&
3605 3606
		    (ptype->dev == null_or_dev || ptype->dev == skb->dev ||
		     ptype->dev == orig_dev)) {
3607
			if (pt_prev)
D
David S. Miller 已提交
3608
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3609 3610 3611 3612 3613
			pt_prev = ptype;
		}
	}

	if (pt_prev) {
3614
		if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
3615
			goto drop;
3616 3617
		else
			ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3618
	} else {
3619
drop:
3620
		atomic_long_inc(&skb->dev->rx_dropped);
L
Linus Torvalds 已提交
3621 3622 3623 3624 3625 3626 3627
		kfree_skb(skb);
		/* Jamal, now you will not able to escape explaining
		 * me how you were going to use this. :-)
		 */
		ret = NET_RX_DROP;
	}

3628
unlock:
L
Linus Torvalds 已提交
3629
	rcu_read_unlock();
3630
out:
3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655
	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
3656 3657
	return ret;
}
T
Tom Herbert 已提交
3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675

/**
 *	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)
{
3676
	net_timestamp_check(netdev_tstamp_prequeue, skb);
E
Eric Dumazet 已提交
3677

3678 3679 3680
	if (skb_defer_rx_timestamp(skb))
		return NET_RX_SUCCESS;

E
Eric Dumazet 已提交
3681
#ifdef CONFIG_RPS
3682
	if (static_key_false(&rps_needed)) {
E
Eric Dumazet 已提交
3683 3684
		struct rps_dev_flow voidflow, *rflow = &voidflow;
		int cpu, ret;
T
Tom Herbert 已提交
3685

E
Eric Dumazet 已提交
3686 3687 3688
		rcu_read_lock();

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
T
Tom Herbert 已提交
3689

E
Eric Dumazet 已提交
3690 3691 3692
		if (cpu >= 0) {
			ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
			rcu_read_unlock();
3693
			return ret;
E
Eric Dumazet 已提交
3694
		}
3695
		rcu_read_unlock();
T
Tom Herbert 已提交
3696
	}
3697
#endif
3698
	return __netif_receive_skb(skb);
T
Tom Herbert 已提交
3699
}
E
Eric Dumazet 已提交
3700
EXPORT_SYMBOL(netif_receive_skb);
L
Linus Torvalds 已提交
3701

E
Eric Dumazet 已提交
3702 3703 3704
/* Network device is going away, flush any packets still pending
 * Called with irqs disabled.
 */
3705
static void flush_backlog(void *arg)
3706
{
3707
	struct net_device *dev = arg;
E
Eric Dumazet 已提交
3708
	struct softnet_data *sd = &__get_cpu_var(softnet_data);
3709 3710
	struct sk_buff *skb, *tmp;

E
Eric Dumazet 已提交
3711
	rps_lock(sd);
3712
	skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
3713
		if (skb->dev == dev) {
E
Eric Dumazet 已提交
3714
			__skb_unlink(skb, &sd->input_pkt_queue);
3715
			kfree_skb(skb);
3716
			input_queue_head_incr(sd);
3717
		}
3718
	}
E
Eric Dumazet 已提交
3719
	rps_unlock(sd);
3720 3721 3722 3723 3724

	skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
		if (skb->dev == dev) {
			__skb_unlink(skb, &sd->process_queue);
			kfree_skb(skb);
3725
			input_queue_head_incr(sd);
3726 3727
		}
	}
3728 3729
}

3730 3731
static int napi_gro_complete(struct sk_buff *skb)
{
3732
	struct packet_offload *ptype;
3733
	__be16 type = skb->protocol;
3734
	struct list_head *head = &offload_base;
3735 3736
	int err = -ENOENT;

3737 3738
	BUILD_BUG_ON(sizeof(struct napi_gro_cb) > sizeof(skb->cb));

3739 3740
	if (NAPI_GRO_CB(skb)->count == 1) {
		skb_shinfo(skb)->gso_size = 0;
3741
		goto out;
3742
	}
3743 3744 3745

	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
3746
		if (ptype->type != type || !ptype->callbacks.gro_complete)
3747 3748
			continue;

3749
		err = ptype->callbacks.gro_complete(skb, 0);
3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763
		break;
	}
	rcu_read_unlock();

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

out:
	return netif_receive_skb(skb);
}

3764 3765 3766 3767 3768
/* 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)
3769
{
3770
	struct sk_buff *skb, *prev = NULL;
3771

3772 3773 3774 3775 3776 3777 3778
	/* 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) {
3779
		skb->next = NULL;
3780 3781 3782 3783 3784

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

		prev = skb->prev;
3785
		napi_gro_complete(skb);
3786
		napi->gro_count--;
3787 3788 3789 3790
	}

	napi->gro_list = NULL;
}
E
Eric Dumazet 已提交
3791
EXPORT_SYMBOL(napi_gro_flush);
3792

E
Eric Dumazet 已提交
3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814
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;

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

		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),
						      skb_gro_mac_header(skb));
		else if (!diffs)
			diffs = memcmp(skb_mac_header(p),
				       skb_gro_mac_header(skb),
				       maclen);
		NAPI_GRO_CB(p)->same_flow = !diffs;
		NAPI_GRO_CB(p)->flush = 0;
	}
}

3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831
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);
	}
}

R
Rami Rosen 已提交
3832
static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3833 3834
{
	struct sk_buff **pp = NULL;
3835
	struct packet_offload *ptype;
3836
	__be16 type = skb->protocol;
3837
	struct list_head *head = &offload_base;
H
Herbert Xu 已提交
3838
	int same_flow;
3839
	enum gro_result ret;
3840

3841
	if (!(skb->dev->features & NETIF_F_GRO) || netpoll_rx_on(skb))
3842 3843
		goto normal;

3844
	if (skb_is_gso(skb) || skb_has_frag_list(skb))
3845 3846
		goto normal;

3847
	skb_gro_reset_offset(skb);
E
Eric Dumazet 已提交
3848 3849
	gro_list_prepare(napi, skb);

3850 3851
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
3852
		if (ptype->type != type || !ptype->callbacks.gro_receive)
3853 3854
			continue;

3855
		skb_set_network_header(skb, skb_gro_offset(skb));
3856
		skb_reset_mac_len(skb);
3857 3858
		NAPI_GRO_CB(skb)->same_flow = 0;
		NAPI_GRO_CB(skb)->flush = 0;
H
Herbert Xu 已提交
3859
		NAPI_GRO_CB(skb)->free = 0;
3860

3861
		pp = ptype->callbacks.gro_receive(&napi->gro_list, skb);
3862 3863 3864 3865 3866 3867 3868
		break;
	}
	rcu_read_unlock();

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

H
Herbert Xu 已提交
3869
	same_flow = NAPI_GRO_CB(skb)->same_flow;
3870
	ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED;
H
Herbert Xu 已提交
3871

3872 3873 3874 3875 3876 3877
	if (pp) {
		struct sk_buff *nskb = *pp;

		*pp = nskb->next;
		nskb->next = NULL;
		napi_gro_complete(nskb);
3878
		napi->gro_count--;
3879 3880
	}

H
Herbert Xu 已提交
3881
	if (same_flow)
3882 3883
		goto ok;

3884
	if (NAPI_GRO_CB(skb)->flush || napi->gro_count >= MAX_GRO_SKBS)
3885 3886
		goto normal;

3887
	napi->gro_count++;
3888
	NAPI_GRO_CB(skb)->count = 1;
3889
	NAPI_GRO_CB(skb)->age = jiffies;
3890
	skb_shinfo(skb)->gso_size = skb_gro_len(skb);
3891 3892
	skb->next = napi->gro_list;
	napi->gro_list = skb;
3893
	ret = GRO_HELD;
3894

3895
pull:
H
Herbert Xu 已提交
3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906
	if (skb_headlen(skb) < skb_gro_offset(skb)) {
		int grow = skb_gro_offset(skb) - skb_headlen(skb);

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

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

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

		skb_shinfo(skb)->frags[0].page_offset += grow;
E
Eric Dumazet 已提交
3907
		skb_frag_size_sub(&skb_shinfo(skb)->frags[0], grow);
H
Herbert Xu 已提交
3908

E
Eric Dumazet 已提交
3909
		if (unlikely(!skb_frag_size(&skb_shinfo(skb)->frags[0]))) {
3910
			skb_frag_unref(skb, 0);
H
Herbert Xu 已提交
3911 3912
			memmove(skb_shinfo(skb)->frags,
				skb_shinfo(skb)->frags + 1,
3913
				--skb_shinfo(skb)->nr_frags * sizeof(skb_frag_t));
H
Herbert Xu 已提交
3914
		}
3915 3916
	}

3917
ok:
3918
	return ret;
3919 3920

normal:
3921 3922
	ret = GRO_NORMAL;
	goto pull;
H
Herbert Xu 已提交
3923
}
3924

H
Herbert Xu 已提交
3925

R
Rami Rosen 已提交
3926
static gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
H
Herbert Xu 已提交
3927
{
3928 3929
	switch (ret) {
	case GRO_NORMAL:
3930 3931 3932
		if (netif_receive_skb(skb))
			ret = GRO_DROP;
		break;
H
Herbert Xu 已提交
3933

3934
	case GRO_DROP:
H
Herbert Xu 已提交
3935 3936
		kfree_skb(skb);
		break;
3937

3938
	case GRO_MERGED_FREE:
3939 3940 3941 3942
		if (NAPI_GRO_CB(skb)->free == NAPI_GRO_FREE_STOLEN_HEAD)
			kmem_cache_free(skbuff_head_cache, skb);
		else
			__kfree_skb(skb);
3943 3944
		break;

3945 3946 3947
	case GRO_HELD:
	case GRO_MERGED:
		break;
H
Herbert Xu 已提交
3948 3949
	}

3950
	return ret;
3951 3952
}

3953
gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3954
{
E
Eric Dumazet 已提交
3955
	return napi_skb_finish(dev_gro_receive(napi, skb), skb);
3956 3957 3958
}
EXPORT_SYMBOL(napi_gro_receive);

S
stephen hemminger 已提交
3959
static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb)
3960 3961
{
	__skb_pull(skb, skb_headlen(skb));
3962 3963
	/* restore the reserve we had after netdev_alloc_skb_ip_align() */
	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN - skb_headroom(skb));
3964
	skb->vlan_tci = 0;
H
Herbert Xu 已提交
3965
	skb->dev = napi->dev;
A
Andy Gospodarek 已提交
3966
	skb->skb_iif = 0;
3967 3968 3969 3970

	napi->skb = skb;
}

3971
struct sk_buff *napi_get_frags(struct napi_struct *napi)
H
Herbert Xu 已提交
3972 3973 3974 3975
{
	struct sk_buff *skb = napi->skb;

	if (!skb) {
3976
		skb = netdev_alloc_skb_ip_align(napi->dev, GRO_MAX_HEAD);
E
Eric Dumazet 已提交
3977
		napi->skb = skb;
3978
	}
3979 3980
	return skb;
}
3981
EXPORT_SYMBOL(napi_get_frags);
3982

R
Rami Rosen 已提交
3983
static gro_result_t napi_frags_finish(struct napi_struct *napi, struct sk_buff *skb,
3984
			       gro_result_t ret)
3985
{
3986 3987
	switch (ret) {
	case GRO_NORMAL:
3988
		if (netif_receive_skb(skb))
3989
			ret = GRO_DROP;
3990
		break;
H
Herbert Xu 已提交
3991

3992 3993 3994 3995
	case GRO_DROP:
	case GRO_MERGED_FREE:
		napi_reuse_skb(napi, skb);
		break;
3996

3997
	case GRO_HELD:
3998 3999
	case GRO_MERGED:
		break;
4000
	}
H
Herbert Xu 已提交
4001

4002
	return ret;
H
Herbert Xu 已提交
4003
}
4004

E
Eric Dumazet 已提交
4005
static struct sk_buff *napi_frags_skb(struct napi_struct *napi)
4006 4007 4008 4009 4010
{
	struct sk_buff *skb = napi->skb;

	napi->skb = NULL;

4011 4012 4013
	if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr)))) {
		napi_reuse_skb(napi, skb);
		return NULL;
4014
	}
4015
	skb->protocol = eth_type_trans(skb, skb->dev);
4016 4017 4018 4019

	return skb;
}

4020
gro_result_t napi_gro_frags(struct napi_struct *napi)
4021
{
4022
	struct sk_buff *skb = napi_frags_skb(napi);
4023 4024

	if (!skb)
4025
		return GRO_DROP;
4026

E
Eric Dumazet 已提交
4027
	return napi_frags_finish(napi, skb, dev_gro_receive(napi, skb));
4028
}
H
Herbert Xu 已提交
4029 4030
EXPORT_SYMBOL(napi_gro_frags);

4031
/*
4032
 * net_rps_action_and_irq_enable sends any pending IPI's for rps.
4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058
 * 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))
				__smp_call_function_single(remsd->cpu,
							   &remsd->csd, 0);
			remsd = next;
		}
	} else
#endif
		local_irq_enable();
}

4059
static int process_backlog(struct napi_struct *napi, int quota)
L
Linus Torvalds 已提交
4060 4061
{
	int work = 0;
E
Eric Dumazet 已提交
4062
	struct softnet_data *sd = container_of(napi, struct softnet_data, backlog);
L
Linus Torvalds 已提交
4063

4064 4065 4066 4067 4068 4069 4070 4071 4072
#ifdef CONFIG_RPS
	/* Check if we have pending ipi, its better to send them now,
	 * not waiting net_rx_action() end.
	 */
	if (sd->rps_ipi_list) {
		local_irq_disable();
		net_rps_action_and_irq_enable(sd);
	}
#endif
4073
	napi->weight = weight_p;
4074 4075
	local_irq_disable();
	while (work < quota) {
L
Linus Torvalds 已提交
4076
		struct sk_buff *skb;
4077 4078 4079 4080 4081 4082
		unsigned int qlen;

		while ((skb = __skb_dequeue(&sd->process_queue))) {
			local_irq_enable();
			__netif_receive_skb(skb);
			local_irq_disable();
4083 4084 4085 4086 4087
			input_queue_head_incr(sd);
			if (++work >= quota) {
				local_irq_enable();
				return work;
			}
4088
		}
L
Linus Torvalds 已提交
4089

E
Eric Dumazet 已提交
4090
		rps_lock(sd);
4091
		qlen = skb_queue_len(&sd->input_pkt_queue);
4092
		if (qlen)
4093 4094
			skb_queue_splice_tail_init(&sd->input_pkt_queue,
						   &sd->process_queue);
4095

4096
		if (qlen < quota - work) {
E
Eric Dumazet 已提交
4097 4098 4099 4100 4101 4102 4103 4104 4105 4106
			/*
			 * Inline a custom version of __napi_complete().
			 * only current cpu owns and manipulates this napi,
			 * and NAPI_STATE_SCHED is the only possible flag set on backlog.
			 * we can use a plain write instead of clear_bit(),
			 * and we dont need an smp_mb() memory barrier.
			 */
			list_del(&napi->poll_list);
			napi->state = 0;

4107
			quota = work + qlen;
4108
		}
E
Eric Dumazet 已提交
4109
		rps_unlock(sd);
4110 4111
	}
	local_irq_enable();
L
Linus Torvalds 已提交
4112

4113 4114
	return work;
}
L
Linus Torvalds 已提交
4115

4116 4117
/**
 * __napi_schedule - schedule for receive
4118
 * @n: entry to schedule
4119 4120 4121
 *
 * The entry's receive function will be scheduled to run
 */
H
Harvey Harrison 已提交
4122
void __napi_schedule(struct napi_struct *n)
4123 4124
{
	unsigned long flags;
L
Linus Torvalds 已提交
4125

4126
	local_irq_save(flags);
E
Eric Dumazet 已提交
4127
	____napi_schedule(&__get_cpu_var(softnet_data), n);
4128
	local_irq_restore(flags);
L
Linus Torvalds 已提交
4129
}
4130 4131
EXPORT_SYMBOL(__napi_schedule);

4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153
void __napi_complete(struct napi_struct *n)
{
	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
	BUG_ON(n->gro_list);

	list_del(&n->poll_list);
	smp_mb__before_clear_bit();
	clear_bit(NAPI_STATE_SCHED, &n->state);
}
EXPORT_SYMBOL(__napi_complete);

void napi_complete(struct napi_struct *n)
{
	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;

4154
	napi_gro_flush(n, false);
4155 4156 4157 4158 4159 4160
	local_irq_save(flags);
	__napi_complete(n);
	local_irq_restore(flags);
}
EXPORT_SYMBOL(napi_complete);

E
Eliezer Tamir 已提交
4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212
/* 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);

4213 4214 4215 4216
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);
4217
	napi->gro_count = 0;
4218
	napi->gro_list = NULL;
H
Herbert Xu 已提交
4219
	napi->skb = NULL;
4220
	napi->poll = poll;
E
Eric Dumazet 已提交
4221 4222 4223
	if (weight > NAPI_POLL_WEIGHT)
		pr_err_once("netif_napi_add() called with weight %d on device %s\n",
			    weight, dev->name);
4224 4225 4226
	napi->weight = weight;
	list_add(&napi->dev_list, &dev->napi_list);
	napi->dev = dev;
H
Herbert Xu 已提交
4227
#ifdef CONFIG_NETPOLL
4228 4229 4230 4231 4232 4233 4234 4235 4236
	spin_lock_init(&napi->poll_lock);
	napi->poll_owner = -1;
#endif
	set_bit(NAPI_STATE_SCHED, &napi->state);
}
EXPORT_SYMBOL(netif_napi_add);

void netif_napi_del(struct napi_struct *napi)
{
4237
	list_del_init(&napi->dev_list);
4238
	napi_free_frags(napi);
4239

E
Eric Dumazet 已提交
4240
	kfree_skb_list(napi->gro_list);
4241
	napi->gro_list = NULL;
4242
	napi->gro_count = 0;
4243 4244 4245
}
EXPORT_SYMBOL(netif_napi_del);

L
Linus Torvalds 已提交
4246 4247
static void net_rx_action(struct softirq_action *h)
{
4248
	struct softnet_data *sd = &__get_cpu_var(softnet_data);
4249
	unsigned long time_limit = jiffies + 2;
4250
	int budget = netdev_budget;
4251 4252
	void *have;

L
Linus Torvalds 已提交
4253 4254
	local_irq_disable();

4255
	while (!list_empty(&sd->poll_list)) {
4256 4257
		struct napi_struct *n;
		int work, weight;
L
Linus Torvalds 已提交
4258

4259
		/* If softirq window is exhuasted then punt.
4260 4261
		 * Allow this to run for 2 jiffies since which will allow
		 * an average latency of 1.5/HZ.
4262
		 */
4263
		if (unlikely(budget <= 0 || time_after_eq(jiffies, time_limit)))
L
Linus Torvalds 已提交
4264 4265 4266 4267
			goto softnet_break;

		local_irq_enable();

4268 4269 4270 4271 4272
		/* Even though interrupts have been re-enabled, this
		 * access is safe because interrupts can only add new
		 * entries to the tail of this list, and only ->poll()
		 * calls can remove this head entry from the list.
		 */
4273
		n = list_first_entry(&sd->poll_list, struct napi_struct, poll_list);
L
Linus Torvalds 已提交
4274

4275 4276 4277 4278
		have = netpoll_poll_lock(n);

		weight = n->weight;

4279 4280 4281 4282
		/* 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
L
Lucas De Marchi 已提交
4283
		 * accidentally calling ->poll() when NAPI is not scheduled.
4284 4285
		 */
		work = 0;
4286
		if (test_bit(NAPI_STATE_SCHED, &n->state)) {
4287
			work = n->poll(n, weight);
4288 4289
			trace_napi_poll(n);
		}
4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301

		WARN_ON_ONCE(work > weight);

		budget -= work;

		local_irq_disable();

		/* 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.
		 */
4302
		if (unlikely(work == weight)) {
4303 4304 4305 4306
			if (unlikely(napi_disable_pending(n))) {
				local_irq_enable();
				napi_complete(n);
				local_irq_disable();
4307 4308 4309 4310 4311 4312 4313 4314 4315
			} else {
				if (n->gro_list) {
					/* flush too old packets
					 * If HZ < 1000, flush all packets.
					 */
					local_irq_enable();
					napi_gro_flush(n, HZ >= 1000);
					local_irq_disable();
				}
4316
				list_move_tail(&n->poll_list, &sd->poll_list);
4317
			}
4318
		}
4319 4320

		netpoll_poll_unlock(have);
L
Linus Torvalds 已提交
4321 4322
	}
out:
4323
	net_rps_action_and_irq_enable(sd);
T
Tom Herbert 已提交
4324

4325 4326 4327 4328 4329
#ifdef CONFIG_NET_DMA
	/*
	 * There may not be any more sk_buffs coming right now, so push
	 * any pending DMA copies to hardware
	 */
4330
	dma_issue_pending_all();
4331
#endif
4332

L
Linus Torvalds 已提交
4333 4334 4335
	return;

softnet_break:
C
Changli Gao 已提交
4336
	sd->time_squeeze++;
L
Linus Torvalds 已提交
4337 4338 4339 4340
	__raise_softirq_irqoff(NET_RX_SOFTIRQ);
	goto out;
}

4341
struct netdev_adjacent {
J
Jiri Pirko 已提交
4342
	struct net_device *dev;
4343 4344

	/* upper master flag, there can only be one master device per list */
J
Jiri Pirko 已提交
4345
	bool master;
4346 4347 4348 4349

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

4350 4351 4352
	/* private field for the users */
	void *private;

J
Jiri Pirko 已提交
4353 4354 4355 4356
	struct list_head list;
	struct rcu_head rcu;
};

4357 4358
static struct netdev_adjacent *__netdev_find_adj(struct net_device *dev,
						 struct net_device *adj_dev,
4359
						 struct list_head *adj_list)
J
Jiri Pirko 已提交
4360
{
4361 4362
	struct netdev_adjacent *adj;

4363
	list_for_each_entry(adj, adj_list, list) {
4364 4365
		if (adj->dev == adj_dev)
			return adj;
J
Jiri Pirko 已提交
4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383
	}
	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();

4384
	return __netdev_find_adj(dev, upper_dev, &dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4385 4386 4387 4388 4389 4390 4391 4392 4393 4394
}
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
4395
static bool netdev_has_any_upper_dev(struct net_device *dev)
J
Jiri Pirko 已提交
4396 4397 4398
{
	ASSERT_RTNL();

4399
	return !list_empty(&dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410
}

/**
 * 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)
{
4411
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4412 4413 4414

	ASSERT_RTNL();

4415
	if (list_empty(&dev->adj_list.upper))
J
Jiri Pirko 已提交
4416 4417
		return NULL;

4418
	upper = list_first_entry(&dev->adj_list.upper,
4419
				 struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4420 4421 4422 4423 4424 4425
	if (likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get);

4426 4427 4428 4429 4430 4431 4432 4433 4434 4435
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);

4436 4437
/**
 * netdev_all_upper_get_next_dev_rcu - Get the next dev from upper list
4438 4439 4440 4441 4442 4443
 * @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.
 */
4444 4445
struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
						     struct list_head **iter)
4446 4447 4448 4449 4450 4451 4452
{
	struct netdev_adjacent *upper;

	WARN_ON_ONCE(!rcu_read_lock_held());

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

4453
	if (&upper->list == &dev->all_adj_list.upper)
4454 4455 4456 4457 4458 4459
		return NULL;

	*iter = &upper->list;

	return upper->dev;
}
4460
EXPORT_SYMBOL(netdev_all_upper_get_next_dev_rcu);
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 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518
/**
 * 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;

	if (iter)
		*iter = lower->list.next;

	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;

	if (iter)
		*iter = &lower->list;

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

4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539
/**
 * 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
4540 4541 4542 4543 4544 4545 4546 4547 4548
/**
 * 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)
{
4549
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4550

4551
	upper = list_first_or_null_rcu(&dev->adj_list.upper,
4552
				       struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4553 4554 4555 4556 4557 4558
	if (upper && likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get_rcu);

4559 4560
static int __netdev_adjacent_dev_insert(struct net_device *dev,
					struct net_device *adj_dev,
4561
					struct list_head *dev_list,
4562
					void *private, bool master)
4563 4564
{
	struct netdev_adjacent *adj;
4565
	char linkname[IFNAMSIZ+7];
4566
	int ret;
4567

4568
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581

	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;
4582
	adj->private = private;
4583
	dev_hold(adj_dev);
4584 4585 4586

	pr_debug("dev_hold for %s, because of link added from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
4587

4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601
	if (dev_list == &dev->adj_list.lower) {
		sprintf(linkname, "lower_%s", adj_dev->name);
		ret = sysfs_create_link(&(dev->dev.kobj),
					&(adj_dev->dev.kobj), linkname);
		if (ret)
			goto free_adj;
	} else if (dev_list == &dev->adj_list.upper) {
		sprintf(linkname, "upper_%s", adj_dev->name);
		ret = sysfs_create_link(&(dev->dev.kobj),
					&(adj_dev->dev.kobj), linkname);
		if (ret)
			goto free_adj;
	}

4602
	/* Ensure that master link is always the first item in list. */
4603 4604 4605 4606
	if (master) {
		ret = sysfs_create_link(&(dev->dev.kobj),
					&(adj_dev->dev.kobj), "master");
		if (ret)
4607
			goto remove_symlinks;
4608

4609
		list_add_rcu(&adj->list, dev_list);
4610
	} else {
4611
		list_add_tail_rcu(&adj->list, dev_list);
4612
	}
4613 4614

	return 0;
4615

4616 4617 4618 4619 4620 4621 4622 4623 4624
remove_symlinks:
	if (dev_list == &dev->adj_list.lower) {
		sprintf(linkname, "lower_%s", adj_dev->name);
		sysfs_remove_link(&(dev->dev.kobj), linkname);
	} else if (dev_list == &dev->adj_list.upper) {
		sprintf(linkname, "upper_%s", adj_dev->name);
		sysfs_remove_link(&(dev->dev.kobj), linkname);
	}

4625 4626
free_adj:
	kfree(adj);
4627
	dev_put(adj_dev);
4628 4629

	return ret;
4630 4631
}

S
stephen hemminger 已提交
4632 4633 4634
static void __netdev_adjacent_dev_remove(struct net_device *dev,
					 struct net_device *adj_dev,
					 struct list_head *dev_list)
4635 4636
{
	struct netdev_adjacent *adj;
4637
	char linkname[IFNAMSIZ+7];
4638

4639
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
4640

4641 4642 4643
	if (!adj) {
		pr_err("tried to remove device %s from %s\n",
		       dev->name, adj_dev->name);
4644
		BUG();
4645
	}
4646 4647

	if (adj->ref_nr > 1) {
4648 4649
		pr_debug("%s to %s ref_nr-- = %d\n", dev->name, adj_dev->name,
			 adj->ref_nr-1);
4650 4651 4652 4653
		adj->ref_nr--;
		return;
	}

4654 4655 4656
	if (adj->master)
		sysfs_remove_link(&(dev->dev.kobj), "master");

4657 4658 4659 4660 4661 4662 4663 4664
	if (dev_list == &dev->adj_list.lower) {
		sprintf(linkname, "lower_%s", adj_dev->name);
		sysfs_remove_link(&(dev->dev.kobj), linkname);
	} else if (dev_list == &dev->adj_list.upper) {
		sprintf(linkname, "upper_%s", adj_dev->name);
		sysfs_remove_link(&(dev->dev.kobj), linkname);
	}

4665
	list_del_rcu(&adj->list);
4666 4667
	pr_debug("dev_put for %s, because link removed from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
4668 4669 4670 4671
	dev_put(adj_dev);
	kfree_rcu(adj, rcu);
}

S
stephen hemminger 已提交
4672 4673 4674 4675 4676
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)
4677 4678 4679
{
	int ret;

4680 4681
	ret = __netdev_adjacent_dev_insert(dev, upper_dev, up_list, private,
					   master);
4682 4683 4684
	if (ret)
		return ret;

4685 4686
	ret = __netdev_adjacent_dev_insert(upper_dev, dev, down_list, private,
					   false);
4687
	if (ret) {
4688
		__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
4689 4690 4691 4692 4693 4694
		return ret;
	}

	return 0;
}

S
stephen hemminger 已提交
4695 4696
static int __netdev_adjacent_dev_link(struct net_device *dev,
				      struct net_device *upper_dev)
4697
{
4698 4699 4700
	return __netdev_adjacent_dev_link_lists(dev, upper_dev,
						&dev->all_adj_list.upper,
						&upper_dev->all_adj_list.lower,
4701
						NULL, false);
4702 4703
}

S
stephen hemminger 已提交
4704 4705 4706 4707
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)
4708
{
4709 4710
	__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
	__netdev_adjacent_dev_remove(upper_dev, dev, down_list);
4711 4712
}

S
stephen hemminger 已提交
4713 4714
static void __netdev_adjacent_dev_unlink(struct net_device *dev,
					 struct net_device *upper_dev)
4715
{
4716 4717 4718 4719 4720
	__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
4721 4722 4723
static int __netdev_adjacent_dev_link_neighbour(struct net_device *dev,
						struct net_device *upper_dev,
						void *private, bool master)
4724 4725 4726 4727 4728 4729 4730 4731 4732
{
	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,
4733
					       private, master);
4734 4735 4736 4737 4738 4739
	if (ret) {
		__netdev_adjacent_dev_unlink(dev, upper_dev);
		return ret;
	}

	return 0;
4740 4741
}

S
stephen hemminger 已提交
4742 4743
static void __netdev_adjacent_dev_unlink_neighbour(struct net_device *dev,
						   struct net_device *upper_dev)
4744 4745 4746 4747 4748 4749
{
	__netdev_adjacent_dev_unlink(dev, upper_dev);
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev,
					   &dev->adj_list.upper,
					   &upper_dev->adj_list.lower);
}
4750

J
Jiri Pirko 已提交
4751
static int __netdev_upper_dev_link(struct net_device *dev,
4752 4753
				   struct net_device *upper_dev, bool master,
				   void *private)
J
Jiri Pirko 已提交
4754
{
4755 4756
	struct netdev_adjacent *i, *j, *to_i, *to_j;
	int ret = 0;
J
Jiri Pirko 已提交
4757 4758 4759 4760 4761 4762 4763

	ASSERT_RTNL();

	if (dev == upper_dev)
		return -EBUSY;

	/* To prevent loops, check if dev is not upper device to upper_dev. */
4764
	if (__netdev_find_adj(upper_dev, dev, &upper_dev->all_adj_list.upper))
J
Jiri Pirko 已提交
4765 4766
		return -EBUSY;

4767
	if (__netdev_find_adj(dev, upper_dev, &dev->all_adj_list.upper))
J
Jiri Pirko 已提交
4768 4769 4770 4771 4772
		return -EEXIST;

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

4773 4774
	ret = __netdev_adjacent_dev_link_neighbour(dev, upper_dev, private,
						   master);
4775 4776
	if (ret)
		return ret;
J
Jiri Pirko 已提交
4777

4778
	/* Now that we linked these devs, make all the upper_dev's
4779
	 * all_adj_list.upper visible to every dev's all_adj_list.lower an
4780 4781 4782
	 * versa, and don't forget the devices itself. All of these
	 * links are non-neighbours.
	 */
4783 4784 4785 4786
	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);
4787 4788 4789 4790 4791 4792 4793
			ret = __netdev_adjacent_dev_link(i->dev, j->dev);
			if (ret)
				goto rollback_mesh;
		}
	}

	/* add dev to every upper_dev's upper device */
4794 4795 4796
	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);
4797 4798 4799 4800 4801 4802
		ret = __netdev_adjacent_dev_link(dev, i->dev);
		if (ret)
			goto rollback_upper_mesh;
	}

	/* add upper_dev to every dev's lower device */
4803 4804 4805
	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);
4806 4807 4808 4809
		ret = __netdev_adjacent_dev_link(i->dev, upper_dev);
		if (ret)
			goto rollback_lower_mesh;
	}
J
Jiri Pirko 已提交
4810

4811
	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
J
Jiri Pirko 已提交
4812
	return 0;
4813 4814 4815

rollback_lower_mesh:
	to_i = i;
4816
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
4817 4818 4819 4820 4821 4822 4823 4824 4825
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);
	}

	i = NULL;

rollback_upper_mesh:
	to_i = i;
4826
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list) {
4827 4828 4829 4830 4831 4832 4833 4834 4835 4836
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(dev, i->dev);
	}

	i = j = NULL;

rollback_mesh:
	to_i = i;
	to_j = j;
4837 4838
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list) {
4839 4840 4841 4842 4843 4844 4845 4846
			if (i == to_i && j == to_j)
				break;
			__netdev_adjacent_dev_unlink(i->dev, j->dev);
		}
		if (i == to_i)
			break;
	}

4847
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
4848 4849

	return ret;
J
Jiri Pirko 已提交
4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864
}

/**
 * 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)
{
4865
	return __netdev_upper_dev_link(dev, upper_dev, false, NULL);
J
Jiri Pirko 已提交
4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882
}
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)
{
4883
	return __netdev_upper_dev_link(dev, upper_dev, true, NULL);
J
Jiri Pirko 已提交
4884 4885 4886
}
EXPORT_SYMBOL(netdev_master_upper_dev_link);

4887 4888 4889 4890 4891 4892 4893 4894
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
4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905
/**
 * 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)
{
4906
	struct netdev_adjacent *i, *j;
J
Jiri Pirko 已提交
4907 4908
	ASSERT_RTNL();

4909
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
4910 4911 4912 4913 4914

	/* 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.
	 */
4915 4916
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list)
4917 4918 4919 4920 4921
			__netdev_adjacent_dev_unlink(i->dev, j->dev);

	/* remove also the devices itself from lower/upper device
	 * list
	 */
4922
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
4923 4924
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);

4925
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list)
4926 4927
		__netdev_adjacent_dev_unlink(dev, i->dev);

4928
	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
J
Jiri Pirko 已提交
4929 4930 4931
}
EXPORT_SYMBOL(netdev_upper_dev_unlink);

4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946
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);

4947 4948
static void dev_change_rx_flags(struct net_device *dev, int flags)
{
4949 4950
	const struct net_device_ops *ops = dev->netdev_ops;

4951
	if (ops->ndo_change_rx_flags)
4952
		ops->ndo_change_rx_flags(dev, flags);
4953 4954
}

4955
static int __dev_set_promiscuity(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
4956
{
4957
	unsigned int old_flags = dev->flags;
4958 4959
	kuid_t uid;
	kgid_t gid;
L
Linus Torvalds 已提交
4960

4961 4962
	ASSERT_RTNL();

4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973
	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;
4974 4975
			pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n",
				dev->name);
4976 4977 4978
			return -EOVERFLOW;
		}
	}
4979
	if (dev->flags != old_flags) {
4980 4981 4982
		pr_info("device %s %s promiscuous mode\n",
			dev->name,
			dev->flags & IFF_PROMISC ? "entered" : "left");
4983 4984
		if (audit_enabled) {
			current_uid_gid(&uid, &gid);
4985 4986 4987 4988 4989
			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),
4990
				from_kuid(&init_user_ns, audit_get_loginuid(current)),
4991 4992
				from_kuid(&init_user_ns, uid),
				from_kgid(&init_user_ns, gid),
4993
				audit_get_sessionid(current));
4994
		}
4995

4996
		dev_change_rx_flags(dev, IFF_PROMISC);
L
Linus Torvalds 已提交
4997
	}
4998 4999
	if (notify)
		__dev_notify_flags(dev, old_flags, IFF_PROMISC);
5000
	return 0;
L
Linus Torvalds 已提交
5001 5002
}

5003 5004 5005 5006 5007 5008 5009 5010 5011
/**
 *	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.
5012
 *	Return 0 if successful or a negative errno code on error.
5013
 */
5014
int dev_set_promiscuity(struct net_device *dev, int inc)
5015
{
5016
	unsigned int old_flags = dev->flags;
5017
	int err;
5018

5019
	err = __dev_set_promiscuity(dev, inc, true);
5020
	if (err < 0)
5021
		return err;
5022 5023
	if (dev->flags != old_flags)
		dev_set_rx_mode(dev);
5024
	return err;
5025
}
E
Eric Dumazet 已提交
5026
EXPORT_SYMBOL(dev_set_promiscuity);
5027

5028
static int __dev_set_allmulti(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5029
{
5030
	unsigned int old_flags = dev->flags, old_gflags = dev->gflags;
L
Linus Torvalds 已提交
5031

5032 5033
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5034
	dev->flags |= IFF_ALLMULTI;
5035 5036 5037 5038 5039 5040 5041 5042 5043 5044
	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;
5045 5046
			pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n",
				dev->name);
5047 5048 5049
			return -EOVERFLOW;
		}
	}
5050
	if (dev->flags ^ old_flags) {
5051
		dev_change_rx_flags(dev, IFF_ALLMULTI);
5052
		dev_set_rx_mode(dev);
5053 5054 5055
		if (notify)
			__dev_notify_flags(dev, old_flags,
					   dev->gflags ^ old_gflags);
5056
	}
5057
	return 0;
5058
}
5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076

/**
 *	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
5077
EXPORT_SYMBOL(dev_set_allmulti);
5078 5079 5080 5081

/*
 *	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
5082
 *	filtering it is put in promiscuous mode while unicast addresses
5083 5084 5085 5086
 *	are present.
 */
void __dev_set_rx_mode(struct net_device *dev)
{
5087 5088
	const struct net_device_ops *ops = dev->netdev_ops;

5089 5090 5091 5092 5093
	/* dev_open will call this function so the list will stay sane. */
	if (!(dev->flags&IFF_UP))
		return;

	if (!netif_device_present(dev))
5094
		return;
5095

5096
	if (!(dev->priv_flags & IFF_UNICAST_FLT)) {
5097 5098 5099
		/* Unicast addresses changes may only happen under the rtnl,
		 * therefore calling __dev_set_promiscuity here is safe.
		 */
5100
		if (!netdev_uc_empty(dev) && !dev->uc_promisc) {
5101
			__dev_set_promiscuity(dev, 1, false);
5102
			dev->uc_promisc = true;
5103
		} else if (netdev_uc_empty(dev) && dev->uc_promisc) {
5104
			__dev_set_promiscuity(dev, -1, false);
5105
			dev->uc_promisc = false;
5106 5107
		}
	}
5108 5109 5110

	if (ops->ndo_set_rx_mode)
		ops->ndo_set_rx_mode(dev);
5111 5112 5113 5114
}

void dev_set_rx_mode(struct net_device *dev)
{
5115
	netif_addr_lock_bh(dev);
5116
	__dev_set_rx_mode(dev);
5117
	netif_addr_unlock_bh(dev);
L
Linus Torvalds 已提交
5118 5119
}

5120 5121 5122 5123 5124 5125
/**
 *	dev_get_flags - get flags reported to userspace
 *	@dev: device
 *
 *	Get the combination of flag bits exported through APIs to userspace.
 */
5126
unsigned int dev_get_flags(const struct net_device *dev)
L
Linus Torvalds 已提交
5127
{
5128
	unsigned int flags;
L
Linus Torvalds 已提交
5129 5130 5131

	flags = (dev->flags & ~(IFF_PROMISC |
				IFF_ALLMULTI |
S
Stefan Rompf 已提交
5132 5133 5134
				IFF_RUNNING |
				IFF_LOWER_UP |
				IFF_DORMANT)) |
L
Linus Torvalds 已提交
5135 5136 5137
		(dev->gflags & (IFF_PROMISC |
				IFF_ALLMULTI));

S
Stefan Rompf 已提交
5138 5139 5140 5141 5142 5143 5144 5145
	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
5146 5147 5148

	return flags;
}
E
Eric Dumazet 已提交
5149
EXPORT_SYMBOL(dev_get_flags);
L
Linus Torvalds 已提交
5150

5151
int __dev_change_flags(struct net_device *dev, unsigned int flags)
L
Linus Torvalds 已提交
5152
{
5153
	unsigned int old_flags = dev->flags;
5154
	int ret;
L
Linus Torvalds 已提交
5155

5156 5157
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171
	/*
	 *	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.
	 */

5172 5173
	if ((old_flags ^ flags) & IFF_MULTICAST)
		dev_change_rx_flags(dev, IFF_MULTICAST);
5174

5175
	dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5176 5177 5178 5179 5180 5181 5182 5183 5184

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

	ret = 0;
	if ((old_flags ^ flags) & IFF_UP) {	/* Bit is different  ? */
5185
		ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev);
L
Linus Torvalds 已提交
5186 5187

		if (!ret)
5188
			dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5189 5190 5191
	}

	if ((flags ^ dev->gflags) & IFF_PROMISC) {
E
Eric Dumazet 已提交
5192
		int inc = (flags & IFF_PROMISC) ? 1 : -1;
5193
		unsigned int old_flags = dev->flags;
E
Eric Dumazet 已提交
5194

L
Linus Torvalds 已提交
5195
		dev->gflags ^= IFF_PROMISC;
5196 5197 5198 5199

		if (__dev_set_promiscuity(dev, inc, false) >= 0)
			if (dev->flags != old_flags)
				dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5200 5201 5202 5203 5204 5205 5206
	}

	/* 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
5207 5208
		int inc = (flags & IFF_ALLMULTI) ? 1 : -1;

L
Linus Torvalds 已提交
5209
		dev->gflags ^= IFF_ALLMULTI;
5210
		__dev_set_allmulti(dev, inc, false);
L
Linus Torvalds 已提交
5211 5212
	}

5213 5214 5215
	return ret;
}

5216 5217
void __dev_notify_flags(struct net_device *dev, unsigned int old_flags,
			unsigned int gchanges)
5218 5219 5220
{
	unsigned int changes = dev->flags ^ old_flags;

5221
	if (gchanges)
5222
		rtmsg_ifinfo(RTM_NEWLINK, dev, gchanges, GFP_ATOMIC);
5223

5224 5225 5226 5227 5228 5229 5230 5231
	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 &&
5232 5233 5234 5235 5236 5237 5238
	    (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);
	}
5239 5240 5241 5242 5243 5244 5245 5246 5247 5248
}

/**
 *	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.
 */
5249
int dev_change_flags(struct net_device *dev, unsigned int flags)
5250
{
5251
	int ret;
5252
	unsigned int changes, old_flags = dev->flags, old_gflags = dev->gflags;
5253 5254 5255 5256 5257

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

5258
	changes = (old_flags ^ dev->flags) | (old_gflags ^ dev->gflags);
5259
	__dev_notify_flags(dev, old_flags, changes);
L
Linus Torvalds 已提交
5260 5261
	return ret;
}
E
Eric Dumazet 已提交
5262
EXPORT_SYMBOL(dev_change_flags);
L
Linus Torvalds 已提交
5263

5264 5265 5266 5267 5268 5269 5270
/**
 *	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
5271 5272
int dev_set_mtu(struct net_device *dev, int new_mtu)
{
5273
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286
	int err;

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

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

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

	err = 0;
5287 5288
	if (ops->ndo_change_mtu)
		err = ops->ndo_change_mtu(dev, new_mtu);
L
Linus Torvalds 已提交
5289 5290
	else
		dev->mtu = new_mtu;
5291

5292
	if (!err)
5293
		call_netdevice_notifiers(NETDEV_CHANGEMTU, dev);
L
Linus Torvalds 已提交
5294 5295
	return err;
}
E
Eric Dumazet 已提交
5296
EXPORT_SYMBOL(dev_set_mtu);
L
Linus Torvalds 已提交
5297

5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308
/**
 *	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);

5309 5310 5311 5312 5313 5314 5315
/**
 *	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
5316 5317
int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa)
{
5318
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
5319 5320
	int err;

5321
	if (!ops->ndo_set_mac_address)
L
Linus Torvalds 已提交
5322 5323 5324 5325 5326
		return -EOPNOTSUPP;
	if (sa->sa_family != dev->type)
		return -EINVAL;
	if (!netif_device_present(dev))
		return -ENODEV;
5327
	err = ops->ndo_set_mac_address(dev, sa);
5328 5329
	if (err)
		return err;
J
Jiri Pirko 已提交
5330
	dev->addr_assign_type = NET_ADDR_SET;
5331
	call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
5332
	add_device_randomness(dev->dev_addr, dev->addr_len);
5333
	return 0;
L
Linus Torvalds 已提交
5334
}
E
Eric Dumazet 已提交
5335
EXPORT_SYMBOL(dev_set_mac_address);
L
Linus Torvalds 已提交
5336

J
Jiri Pirko 已提交
5337 5338 5339
/**
 *	dev_change_carrier - Change device carrier
 *	@dev: device
5340
 *	@new_carrier: new value
J
Jiri Pirko 已提交
5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355
 *
 *	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);

5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373
/**
 *	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,
			 struct netdev_phys_port_id *ppid)
{
	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
5374 5375
/**
 *	dev_new_index	-	allocate an ifindex
5376
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
5377 5378 5379 5380 5381
 *
 *	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.
 */
5382
static int dev_new_index(struct net *net)
L
Linus Torvalds 已提交
5383
{
5384
	int ifindex = net->ifindex;
L
Linus Torvalds 已提交
5385 5386 5387
	for (;;) {
		if (++ifindex <= 0)
			ifindex = 1;
5388
		if (!__dev_get_by_index(net, ifindex))
5389
			return net->ifindex = ifindex;
L
Linus Torvalds 已提交
5390 5391 5392 5393
	}
}

/* Delayed registration/unregisteration */
5394
static LIST_HEAD(net_todo_list);
5395
static DECLARE_WAIT_QUEUE_HEAD(netdev_unregistering_wq);
L
Linus Torvalds 已提交
5396

5397
static void net_set_todo(struct net_device *dev)
L
Linus Torvalds 已提交
5398 5399
{
	list_add_tail(&dev->todo_list, &net_todo_list);
5400
	dev_net(dev)->dev_unreg_count++;
L
Linus Torvalds 已提交
5401 5402
}

5403
static void rollback_registered_many(struct list_head *head)
5404
{
5405
	struct net_device *dev, *tmp;
5406
	LIST_HEAD(close_head);
5407

5408 5409 5410
	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

5411
	list_for_each_entry_safe(dev, tmp, head, unreg_list) {
5412
		/* Some devices call without registering
5413 5414
		 * for initialization unwind. Remove those
		 * devices and proceed with the remaining.
5415 5416
		 */
		if (dev->reg_state == NETREG_UNINITIALIZED) {
5417 5418
			pr_debug("unregister_netdevice: device %s/%p never was registered\n",
				 dev->name, dev);
5419

5420
			WARN_ON(1);
5421 5422
			list_del(&dev->unreg_list);
			continue;
5423
		}
5424
		dev->dismantle = true;
5425
		BUG_ON(dev->reg_state != NETREG_REGISTERED);
5426
	}
5427

5428
	/* If device is running, close it first. */
5429 5430 5431
	list_for_each_entry(dev, head, unreg_list)
		list_add_tail(&dev->close_list, &close_head);
	dev_close_many(&close_head);
5432

5433
	list_for_each_entry(dev, head, unreg_list) {
5434 5435
		/* And unlink it from device chain. */
		unlist_netdevice(dev);
5436

5437 5438
		dev->reg_state = NETREG_UNREGISTERING;
	}
5439 5440 5441

	synchronize_net();

5442 5443 5444
	list_for_each_entry(dev, head, unreg_list) {
		/* Shutdown queueing discipline. */
		dev_shutdown(dev);
5445 5446


5447 5448 5449 5450
		/* Notify protocols, that we are about to destroy
		   this device. They should clean all the things.
		*/
		call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
5451

5452 5453
		if (!dev->rtnl_link_ops ||
		    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
5454
			rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);
5455

5456 5457 5458
		/*
		 *	Flush the unicast and multicast chains
		 */
5459
		dev_uc_flush(dev);
5460
		dev_mc_flush(dev);
5461

5462 5463
		if (dev->netdev_ops->ndo_uninit)
			dev->netdev_ops->ndo_uninit(dev);
5464

J
Jiri Pirko 已提交
5465 5466
		/* Notifier chain MUST detach us all upper devices. */
		WARN_ON(netdev_has_any_upper_dev(dev));
5467

5468 5469
		/* Remove entries from kobject tree */
		netdev_unregister_kobject(dev);
5470 5471 5472 5473
#ifdef CONFIG_XPS
		/* Remove XPS queueing entries */
		netif_reset_xps_queues_gt(dev, 0);
#endif
5474
	}
5475

5476
	synchronize_net();
5477

5478
	list_for_each_entry(dev, head, unreg_list)
5479 5480 5481 5482 5483 5484 5485 5486 5487
		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
5488
	list_del(&single);
5489 5490
}

5491 5492
static netdev_features_t netdev_fix_features(struct net_device *dev,
	netdev_features_t features)
5493
{
5494 5495 5496
	/* Fix illegal checksum combinations */
	if ((features & NETIF_F_HW_CSUM) &&
	    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
5497
		netdev_warn(dev, "mixed HW and IP checksum settings.\n");
5498 5499 5500
		features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
	}

5501
	/* TSO requires that SG is present as well. */
5502
	if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
5503
		netdev_dbg(dev, "Dropping TSO features since no SG feature.\n");
5504
		features &= ~NETIF_F_ALL_TSO;
5505 5506
	}

5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519
	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;
	}

5520 5521 5522 5523
	/* TSO ECN requires that TSO is present as well. */
	if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN)
		features &= ~NETIF_F_TSO_ECN;

5524 5525
	/* Software GSO depends on SG. */
	if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) {
5526
		netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n");
5527 5528 5529
		features &= ~NETIF_F_GSO;
	}

5530
	/* UFO needs SG and checksumming */
5531
	if (features & NETIF_F_UFO) {
5532 5533 5534 5535
		/* 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))) {
5536
			netdev_dbg(dev,
5537
				"Dropping NETIF_F_UFO since no checksum offload features.\n");
5538 5539 5540 5541
			features &= ~NETIF_F_UFO;
		}

		if (!(features & NETIF_F_SG)) {
5542
			netdev_dbg(dev,
5543
				"Dropping NETIF_F_UFO since no NETIF_F_SG feature.\n");
5544 5545 5546 5547 5548 5549 5550
			features &= ~NETIF_F_UFO;
		}
	}

	return features;
}

5551
int __netdev_update_features(struct net_device *dev)
5552
{
5553
	netdev_features_t features;
5554 5555
	int err = 0;

5556 5557
	ASSERT_RTNL();

5558 5559 5560 5561 5562 5563 5564 5565 5566
	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)
5567
		return 0;
5568

5569 5570
	netdev_dbg(dev, "Features changed: %pNF -> %pNF\n",
		&dev->features, &features);
5571 5572 5573 5574

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

5575
	if (unlikely(err < 0)) {
5576
		netdev_err(dev,
5577 5578
			"set_features() failed (%d); wanted %pNF, left %pNF\n",
			err, &features, &dev->features);
5579 5580 5581 5582 5583 5584 5585 5586 5587
		return -1;
	}

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

	return 1;
}

5588 5589 5590 5591 5592 5593 5594 5595
/**
 *	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.
 */
5596 5597 5598 5599
void netdev_update_features(struct net_device *dev)
{
	if (__netdev_update_features(dev))
		netdev_features_change(dev);
5600 5601 5602
}
EXPORT_SYMBOL(netdev_update_features);

5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619
/**
 *	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);

5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646
/**
 *	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);

T
Tom Herbert 已提交
5647
#ifdef CONFIG_RPS
5648 5649 5650
static int netif_alloc_rx_queues(struct net_device *dev)
{
	unsigned int i, count = dev->num_rx_queues;
T
Tom Herbert 已提交
5651
	struct netdev_rx_queue *rx;
5652

T
Tom Herbert 已提交
5653
	BUG_ON(count < 1);
5654

T
Tom Herbert 已提交
5655
	rx = kcalloc(count, sizeof(struct netdev_rx_queue), GFP_KERNEL);
5656
	if (!rx)
T
Tom Herbert 已提交
5657
		return -ENOMEM;
5658

T
Tom Herbert 已提交
5659 5660 5661
	dev->_rx = rx;

	for (i = 0; i < count; i++)
T
Tom Herbert 已提交
5662
		rx[i].dev = dev;
5663 5664
	return 0;
}
T
Tom Herbert 已提交
5665
#endif
5666

C
Changli Gao 已提交
5667 5668 5669 5670 5671 5672 5673
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;
5674
	netdev_queue_numa_node_write(queue, NUMA_NO_NODE);
C
Changli Gao 已提交
5675
	queue->dev = dev;
T
Tom Herbert 已提交
5676 5677 5678
#ifdef CONFIG_BQL
	dql_init(&queue->dql, HZ);
#endif
C
Changli Gao 已提交
5679 5680
}

5681 5682 5683 5684 5685 5686 5687 5688
static void netif_free_tx_queues(struct net_device *dev)
{
	if (is_vmalloc_addr(dev->_tx))
		vfree(dev->_tx);
	else
		kfree(dev->_tx);
}

5689 5690 5691 5692
static int netif_alloc_netdev_queues(struct net_device *dev)
{
	unsigned int count = dev->num_tx_queues;
	struct netdev_queue *tx;
5693
	size_t sz = count * sizeof(*tx);
5694

5695
	BUG_ON(count < 1 || count > 0xffff);
5696

5697 5698 5699 5700 5701 5702
	tx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!tx) {
		tx = vzalloc(sz);
		if (!tx)
			return -ENOMEM;
	}
5703
	dev->_tx = tx;
T
Tom Herbert 已提交
5704

5705 5706
	netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
	spin_lock_init(&dev->tx_global_lock);
C
Changli Gao 已提交
5707 5708

	return 0;
5709 5710
}

L
Linus Torvalds 已提交
5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730
/**
 *	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;
5731
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
5732 5733 5734 5735

	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

5736 5737
	might_sleep();

L
Linus Torvalds 已提交
5738 5739
	/* When net_device's are persistent, this will be fatal. */
	BUG_ON(dev->reg_state != NETREG_UNINITIALIZED);
5740
	BUG_ON(!net);
L
Linus Torvalds 已提交
5741

5742
	spin_lock_init(&dev->addr_list_lock);
5743
	netdev_set_addr_lockdep_class(dev);
L
Linus Torvalds 已提交
5744 5745 5746

	dev->iflink = -1;

5747
	ret = dev_get_valid_name(net, dev, dev->name);
5748 5749 5750
	if (ret < 0)
		goto out;

L
Linus Torvalds 已提交
5751
	/* Init, if this function is available */
5752 5753
	if (dev->netdev_ops->ndo_init) {
		ret = dev->netdev_ops->ndo_init(dev);
L
Linus Torvalds 已提交
5754 5755 5756
		if (ret) {
			if (ret > 0)
				ret = -EIO;
5757
			goto out;
L
Linus Torvalds 已提交
5758 5759
		}
	}
5760

5761 5762
	if (((dev->hw_features | dev->features) &
	     NETIF_F_HW_VLAN_CTAG_FILTER) &&
5763 5764 5765 5766 5767 5768 5769
	    (!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;
	}

5770 5771 5772 5773 5774 5775
	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
5776 5777 5778
	if (dev->iflink == -1)
		dev->iflink = dev->ifindex;

5779 5780 5781 5782
	/* Transfer changeable features to wanted_features and enable
	 * software offloads (GSO and GRO).
	 */
	dev->hw_features |= NETIF_F_SOFT_FEATURES;
5783 5784
	dev->features |= NETIF_F_SOFT_FEATURES;
	dev->wanted_features = dev->features & dev->hw_features;
L
Linus Torvalds 已提交
5785

5786
	/* Turn on no cache copy if HW is doing checksum */
5787 5788 5789 5790 5791 5792
	if (!(dev->flags & IFF_LOOPBACK)) {
		dev->hw_features |= NETIF_F_NOCACHE_COPY;
		if (dev->features & NETIF_F_ALL_CSUM) {
			dev->wanted_features |= NETIF_F_NOCACHE_COPY;
			dev->features |= NETIF_F_NOCACHE_COPY;
		}
5793 5794
	}

5795
	/* Make NETIF_F_HIGHDMA inheritable to VLAN devices.
5796
	 */
5797
	dev->vlan_features |= NETIF_F_HIGHDMA;
5798

5799 5800 5801 5802
	/* Make NETIF_F_SG inheritable to tunnel devices.
	 */
	dev->hw_enc_features |= NETIF_F_SG;

S
Simon Horman 已提交
5803 5804 5805 5806
	/* Make NETIF_F_SG inheritable to MPLS.
	 */
	dev->mpls_features |= NETIF_F_SG;

5807 5808 5809 5810 5811
	ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		goto err_uninit;

5812
	ret = netdev_register_kobject(dev);
5813
	if (ret)
5814
		goto err_uninit;
5815 5816
	dev->reg_state = NETREG_REGISTERED;

5817
	__netdev_update_features(dev);
5818

L
Linus Torvalds 已提交
5819 5820 5821 5822 5823 5824 5825
	/*
	 *	Default initial state at registry is that the
	 *	device is present.
	 */

	set_bit(__LINK_STATE_PRESENT, &dev->state);

5826 5827
	linkwatch_init_dev(dev);

L
Linus Torvalds 已提交
5828 5829
	dev_init_scheduler(dev);
	dev_hold(dev);
5830
	list_netdevice(dev);
5831
	add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
5832

5833 5834 5835 5836 5837 5838 5839
	/* 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
5840
	/* Notify protocols, that a new device appeared. */
5841
	ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
5842
	ret = notifier_to_errno(ret);
5843 5844 5845 5846
	if (ret) {
		rollback_registered(dev);
		dev->reg_state = NETREG_UNREGISTERED;
	}
5847 5848 5849 5850
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
5851 5852
	if (!dev->rtnl_link_ops ||
	    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
5853
		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
L
Linus Torvalds 已提交
5854 5855 5856

out:
	return ret;
5857 5858

err_uninit:
5859 5860
	if (dev->netdev_ops->ndo_uninit)
		dev->netdev_ops->ndo_uninit(dev);
5861
	goto out;
L
Linus Torvalds 已提交
5862
}
E
Eric Dumazet 已提交
5863
EXPORT_SYMBOL(register_netdevice);
L
Linus Torvalds 已提交
5864

5865 5866 5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895
/**
 *	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
5896 5897 5898 5899 5900
	/* Note : We dont allocate pcpu_refcnt for dummy devices,
	 * because users of this 'device' dont need to change
	 * its refcount.
	 */

5901 5902 5903 5904 5905
	return 0;
}
EXPORT_SYMBOL_GPL(init_dummy_netdev);


L
Linus Torvalds 已提交
5906 5907 5908 5909 5910 5911 5912 5913 5914
/**
 *	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.
 *
5915
 *	This is a wrapper around register_netdevice that takes the rtnl semaphore
L
Linus Torvalds 已提交
5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929
 *	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
5930 5931 5932 5933 5934 5935 5936 5937 5938 5939
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);

5940
/**
L
Linus Torvalds 已提交
5941
 * netdev_wait_allrefs - wait until all references are gone.
5942
 * @dev: target net_device
L
Linus Torvalds 已提交
5943 5944 5945 5946 5947 5948 5949
 *
 * 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
5950
 * call dev_put.
L
Linus Torvalds 已提交
5951 5952 5953 5954
 */
static void netdev_wait_allrefs(struct net_device *dev)
{
	unsigned long rebroadcast_time, warning_time;
E
Eric Dumazet 已提交
5955
	int refcnt;
L
Linus Torvalds 已提交
5956

5957 5958
	linkwatch_forget_dev(dev);

L
Linus Torvalds 已提交
5959
	rebroadcast_time = warning_time = jiffies;
E
Eric Dumazet 已提交
5960 5961 5962
	refcnt = netdev_refcnt_read(dev);

	while (refcnt != 0) {
L
Linus Torvalds 已提交
5963
		if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
5964
			rtnl_lock();
L
Linus Torvalds 已提交
5965 5966

			/* Rebroadcast unregister notification */
5967
			call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
L
Linus Torvalds 已提交
5968

5969
			__rtnl_unlock();
5970
			rcu_barrier();
5971 5972
			rtnl_lock();

5973
			call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
L
Linus Torvalds 已提交
5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984
			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();
			}

5985
			__rtnl_unlock();
L
Linus Torvalds 已提交
5986 5987 5988 5989 5990 5991

			rebroadcast_time = jiffies;
		}

		msleep(250);

E
Eric Dumazet 已提交
5992 5993
		refcnt = netdev_refcnt_read(dev);

L
Linus Torvalds 已提交
5994
		if (time_after(jiffies, warning_time + 10 * HZ)) {
5995 5996
			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
				 dev->name, refcnt);
L
Linus Torvalds 已提交
5997 5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015
			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
6016
 * We are invoked by rtnl_unlock().
L
Linus Torvalds 已提交
6017
 * This allows us to deal with problems:
6018
 * 1) We can delete sysfs objects which invoke hotplug
L
Linus Torvalds 已提交
6019 6020 6021
 *    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
6022 6023 6024
 *
 * 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
6025 6026 6027
 */
void netdev_run_todo(void)
{
6028
	struct list_head list;
L
Linus Torvalds 已提交
6029 6030

	/* Snapshot list, allow later requests */
6031
	list_replace_init(&net_todo_list, &list);
H
Herbert Xu 已提交
6032 6033

	__rtnl_unlock();
6034

6035 6036

	/* Wait for rcu callbacks to finish before next phase */
6037 6038 6039
	if (!list_empty(&list))
		rcu_barrier();

L
Linus Torvalds 已提交
6040 6041
	while (!list_empty(&list)) {
		struct net_device *dev
6042
			= list_first_entry(&list, struct net_device, todo_list);
L
Linus Torvalds 已提交
6043 6044
		list_del(&dev->todo_list);

6045
		rtnl_lock();
6046
		call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6047
		__rtnl_unlock();
6048

6049
		if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
6050
			pr_err("network todo '%s' but state %d\n",
6051 6052 6053 6054
			       dev->name, dev->reg_state);
			dump_stack();
			continue;
		}
L
Linus Torvalds 已提交
6055

6056
		dev->reg_state = NETREG_UNREGISTERED;
L
Linus Torvalds 已提交
6057

6058
		on_each_cpu(flush_backlog, dev, 1);
6059

6060
		netdev_wait_allrefs(dev);
L
Linus Torvalds 已提交
6061

6062
		/* paranoia */
E
Eric Dumazet 已提交
6063
		BUG_ON(netdev_refcnt_read(dev));
6064 6065
		WARN_ON(rcu_access_pointer(dev->ip_ptr));
		WARN_ON(rcu_access_pointer(dev->ip6_ptr));
6066
		WARN_ON(dev->dn_ptr);
L
Linus Torvalds 已提交
6067

6068 6069
		if (dev->destructor)
			dev->destructor(dev);
6070

6071 6072 6073 6074 6075 6076
		/* Report a network device has been unregistered */
		rtnl_lock();
		dev_net(dev)->dev_unreg_count--;
		__rtnl_unlock();
		wake_up(&netdev_unregistering_wq);

6077 6078
		/* Free network device */
		kobject_put(&dev->dev.kobj);
L
Linus Torvalds 已提交
6079 6080 6081
	}
}

6082 6083 6084
/* Convert net_device_stats to rtnl_link_stats64.  They have the same
 * fields in the same order, with only the type differing.
 */
6085 6086
void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
			     const struct net_device_stats *netdev_stats)
6087 6088
{
#if BITS_PER_LONG == 64
6089 6090
	BUILD_BUG_ON(sizeof(*stats64) != sizeof(*netdev_stats));
	memcpy(stats64, netdev_stats, sizeof(*stats64));
6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101
#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
}
6102
EXPORT_SYMBOL(netdev_stats_to_stats64);
6103

6104 6105 6106
/**
 *	dev_get_stats	- get network device statistics
 *	@dev: device to get statistics from
6107
 *	@storage: place to store stats
6108
 *
6109 6110 6111 6112
 *	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.
6113
 */
6114 6115
struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
					struct rtnl_link_stats64 *storage)
6116
{
6117 6118
	const struct net_device_ops *ops = dev->netdev_ops;

6119 6120
	if (ops->ndo_get_stats64) {
		memset(storage, 0, sizeof(*storage));
6121 6122
		ops->ndo_get_stats64(dev, storage);
	} else if (ops->ndo_get_stats) {
6123
		netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev));
6124 6125
	} else {
		netdev_stats_to_stats64(storage, &dev->stats);
6126
	}
6127
	storage->rx_dropped += atomic_long_read(&dev->rx_dropped);
6128
	return storage;
R
Rusty Russell 已提交
6129
}
6130
EXPORT_SYMBOL(dev_get_stats);
R
Rusty Russell 已提交
6131

6132
struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
6133
{
6134
	struct netdev_queue *queue = dev_ingress_queue(dev);
6135

6136 6137 6138 6139 6140 6141 6142 6143 6144 6145 6146 6147
#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;
6148 6149
}

6150 6151
static const struct ethtool_ops default_ethtool_ops;

6152 6153 6154 6155 6156 6157 6158 6159
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);

6160 6161 6162 6163 6164 6165 6166 6167 6168 6169
void netdev_freemem(struct net_device *dev)
{
	char *addr = (char *)dev - dev->padded;

	if (is_vmalloc_addr(addr))
		vfree(addr);
	else
		kfree(addr);
}

L
Linus Torvalds 已提交
6170
/**
T
Tom Herbert 已提交
6171
 *	alloc_netdev_mqs - allocate network device
L
Linus Torvalds 已提交
6172 6173 6174
 *	@sizeof_priv:	size of private data to allocate space for
 *	@name:		device name format string
 *	@setup:		callback to initialize device
T
Tom Herbert 已提交
6175 6176
 *	@txqs:		the number of TX subqueues to allocate
 *	@rxqs:		the number of RX subqueues to allocate
L
Linus Torvalds 已提交
6177 6178
 *
 *	Allocates a struct net_device with private data area for driver use
6179
 *	and performs basic initialization.  Also allocates subquue structs
T
Tom Herbert 已提交
6180
 *	for each queue on the device.
L
Linus Torvalds 已提交
6181
 */
T
Tom Herbert 已提交
6182 6183 6184
struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
		void (*setup)(struct net_device *),
		unsigned int txqs, unsigned int rxqs)
L
Linus Torvalds 已提交
6185 6186
{
	struct net_device *dev;
6187
	size_t alloc_size;
6188
	struct net_device *p;
L
Linus Torvalds 已提交
6189

6190 6191
	BUG_ON(strlen(name) >= sizeof(dev->name));

T
Tom Herbert 已提交
6192
	if (txqs < 1) {
6193
		pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
6194 6195 6196
		return NULL;
	}

T
Tom Herbert 已提交
6197 6198
#ifdef CONFIG_RPS
	if (rxqs < 1) {
6199
		pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
T
Tom Herbert 已提交
6200 6201 6202 6203
		return NULL;
	}
#endif

6204
	alloc_size = sizeof(struct net_device);
6205 6206
	if (sizeof_priv) {
		/* ensure 32-byte alignment of private area */
6207
		alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
6208 6209 6210
		alloc_size += sizeof_priv;
	}
	/* ensure 32-byte alignment of whole construct */
6211
	alloc_size += NETDEV_ALIGN - 1;
L
Linus Torvalds 已提交
6212

6213 6214 6215
	p = kzalloc(alloc_size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!p)
		p = vzalloc(alloc_size);
6216
	if (!p)
L
Linus Torvalds 已提交
6217 6218
		return NULL;

6219
	dev = PTR_ALIGN(p, NETDEV_ALIGN);
L
Linus Torvalds 已提交
6220
	dev->padded = (char *)dev - (char *)p;
6221

E
Eric Dumazet 已提交
6222 6223
	dev->pcpu_refcnt = alloc_percpu(int);
	if (!dev->pcpu_refcnt)
6224
		goto free_dev;
6225 6226

	if (dev_addr_init(dev))
E
Eric Dumazet 已提交
6227
		goto free_pcpu;
6228

6229
	dev_mc_init(dev);
6230
	dev_uc_init(dev);
J
Jiri Pirko 已提交
6231

6232
	dev_net_set(dev, &init_net);
L
Linus Torvalds 已提交
6233

6234
	dev->gso_max_size = GSO_MAX_SIZE;
6235
	dev->gso_max_segs = GSO_MAX_SEGS;
6236 6237 6238

	INIT_LIST_HEAD(&dev->napi_list);
	INIT_LIST_HEAD(&dev->unreg_list);
6239
	INIT_LIST_HEAD(&dev->close_list);
6240
	INIT_LIST_HEAD(&dev->link_watch_list);
6241 6242 6243 6244
	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);
6245 6246 6247
	dev->priv_flags = IFF_XMIT_DST_RELEASE;
	setup(dev);

T
Tom Herbert 已提交
6248 6249
	dev->num_tx_queues = txqs;
	dev->real_num_tx_queues = txqs;
6250
	if (netif_alloc_netdev_queues(dev))
6251
		goto free_all;
6252

E
Eric Dumazet 已提交
6253
#ifdef CONFIG_RPS
T
Tom Herbert 已提交
6254 6255
	dev->num_rx_queues = rxqs;
	dev->real_num_rx_queues = rxqs;
T
Tom Herbert 已提交
6256
	if (netif_alloc_rx_queues(dev))
6257
		goto free_all;
E
Eric Dumazet 已提交
6258
#endif
T
Tom Herbert 已提交
6259

L
Linus Torvalds 已提交
6260
	strcpy(dev->name, name);
6261
	dev->group = INIT_NETDEV_GROUP;
6262 6263
	if (!dev->ethtool_ops)
		dev->ethtool_ops = &default_ethtool_ops;
L
Linus Torvalds 已提交
6264
	return dev;
6265

6266 6267 6268 6269
free_all:
	free_netdev(dev);
	return NULL;

E
Eric Dumazet 已提交
6270 6271
free_pcpu:
	free_percpu(dev->pcpu_refcnt);
6272
	netif_free_tx_queues(dev);
T
Tom Herbert 已提交
6273 6274 6275 6276
#ifdef CONFIG_RPS
	kfree(dev->_rx);
#endif

6277 6278
free_dev:
	netdev_freemem(dev);
6279
	return NULL;
L
Linus Torvalds 已提交
6280
}
T
Tom Herbert 已提交
6281
EXPORT_SYMBOL(alloc_netdev_mqs);
L
Linus Torvalds 已提交
6282 6283 6284 6285 6286

/**
 *	free_netdev - free network device
 *	@dev: device
 *
6287 6288
 *	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
6289 6290 6291 6292
 *	If this is the last reference then it will be freed.
 */
void free_netdev(struct net_device *dev)
{
6293 6294
	struct napi_struct *p, *n;

6295 6296
	release_net(dev_net(dev));

6297
	netif_free_tx_queues(dev);
T
Tom Herbert 已提交
6298 6299 6300
#ifdef CONFIG_RPS
	kfree(dev->_rx);
#endif
6301

6302
	kfree(rcu_dereference_protected(dev->ingress_queue, 1));
6303

6304 6305 6306
	/* Flush device addresses */
	dev_addr_flush(dev);

6307 6308 6309
	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
		netif_napi_del(p);

E
Eric Dumazet 已提交
6310 6311 6312
	free_percpu(dev->pcpu_refcnt);
	dev->pcpu_refcnt = NULL;

S
Stephen Hemminger 已提交
6313
	/*  Compatibility with error handling in drivers */
L
Linus Torvalds 已提交
6314
	if (dev->reg_state == NETREG_UNINITIALIZED) {
6315
		netdev_freemem(dev);
L
Linus Torvalds 已提交
6316 6317 6318 6319 6320 6321
		return;
	}

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

6322 6323
	/* will free via device release */
	put_device(&dev->dev);
L
Linus Torvalds 已提交
6324
}
E
Eric Dumazet 已提交
6325
EXPORT_SYMBOL(free_netdev);
6326

6327 6328 6329 6330 6331 6332
/**
 *	synchronize_net -  Synchronize with packet receive processing
 *
 *	Wait for packets currently being received to be done.
 *	Does not block later packets from starting.
 */
6333
void synchronize_net(void)
L
Linus Torvalds 已提交
6334 6335
{
	might_sleep();
6336 6337 6338 6339
	if (rtnl_is_locked())
		synchronize_rcu_expedited();
	else
		synchronize_rcu();
L
Linus Torvalds 已提交
6340
}
E
Eric Dumazet 已提交
6341
EXPORT_SYMBOL(synchronize_net);
L
Linus Torvalds 已提交
6342 6343

/**
6344
 *	unregister_netdevice_queue - remove device from the kernel
L
Linus Torvalds 已提交
6345
 *	@dev: device
6346
 *	@head: list
6347
 *
L
Linus Torvalds 已提交
6348
 *	This function shuts down a device interface and removes it
6349
 *	from the kernel tables.
6350
 *	If head not NULL, device is queued to be unregistered later.
L
Linus Torvalds 已提交
6351 6352 6353 6354 6355
 *
 *	Callers must hold the rtnl semaphore.  You may want
 *	unregister_netdev() instead of this.
 */

6356
void unregister_netdevice_queue(struct net_device *dev, struct list_head *head)
L
Linus Torvalds 已提交
6357
{
6358 6359
	ASSERT_RTNL();

6360
	if (head) {
6361
		list_move_tail(&dev->unreg_list, head);
6362 6363 6364 6365 6366
	} else {
		rollback_registered(dev);
		/* Finish processing unregister after unlock */
		net_set_todo(dev);
	}
L
Linus Torvalds 已提交
6367
}
6368
EXPORT_SYMBOL(unregister_netdevice_queue);
L
Linus Torvalds 已提交
6369

6370 6371 6372 6373 6374 6375 6376 6377 6378 6379 6380 6381 6382 6383
/**
 *	unregister_netdevice_many - unregister many devices
 *	@head: list of devices
 */
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);
	}
}
6384
EXPORT_SYMBOL(unregister_netdevice_many);
6385

L
Linus Torvalds 已提交
6386 6387 6388 6389 6390
/**
 *	unregister_netdev - remove device from the kernel
 *	@dev: device
 *
 *	This function shuts down a device interface and removes it
6391
 *	from the kernel tables.
L
Linus Torvalds 已提交
6392 6393 6394 6395 6396 6397 6398 6399 6400 6401 6402 6403 6404
 *
 *	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);

6405 6406 6407 6408 6409 6410 6411 6412 6413 6414 6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426 6427 6428 6429 6430 6431 6432 6433 6434 6435
/**
 *	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;
6436
	if (net_eq(dev_net(dev), net))
6437 6438 6439 6440 6441 6442
		goto out;

	/* Pick the destination device name, and ensure
	 * we can use it in the destination network namespace.
	 */
	err = -EEXIST;
6443
	if (__dev_get_by_name(net, dev->name)) {
6444 6445 6446
		/* We get here if we can't use the current device name */
		if (!pat)
			goto out;
6447
		if (dev_get_valid_name(net, dev, pat) < 0)
6448 6449 6450 6451 6452 6453 6454 6455
			goto out;
	}

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

	/* If device is running close it first. */
6456
	dev_close(dev);
6457 6458 6459 6460 6461 6462 6463 6464 6465 6466 6467 6468

	/* 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.
6469 6470 6471 6472

	   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.
6473 6474
	*/
	call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
6475 6476
	rcu_barrier();
	call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6477
	rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);
6478 6479 6480 6481

	/*
	 *	Flush the unicast and multicast chains
	 */
6482
	dev_uc_flush(dev);
6483
	dev_mc_flush(dev);
6484

6485 6486 6487
	/* Send a netdev-removed uevent to the old namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_REMOVE);

6488
	/* Actually switch the network namespace */
6489
	dev_net_set(dev, net);
6490 6491 6492 6493 6494 6495 6496 6497 6498

	/* 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;
	}

6499 6500 6501
	/* Send a netdev-add uevent to the new namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_ADD);

6502
	/* Fixup kobjects */
6503
	err = device_rename(&dev->dev, dev->name);
6504
	WARN_ON(err);
6505 6506 6507 6508 6509 6510 6511

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

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

6512 6513 6514 6515
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
6516
	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
6517

6518 6519 6520 6521 6522
	synchronize_net();
	err = 0;
out:
	return err;
}
6523
EXPORT_SYMBOL_GPL(dev_change_net_namespace);
6524

L
Linus Torvalds 已提交
6525 6526 6527 6528 6529 6530 6531 6532 6533
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;

6534
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
L
Linus Torvalds 已提交
6535 6536 6537 6538 6539 6540 6541 6542 6543 6544 6545 6546 6547 6548 6549 6550
		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. */
6551 6552 6553 6554 6555 6556
	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;
	}
6557 6558 6559 6560 6561
	/* 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
6562 6563 6564 6565 6566

	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_enable();

	/* Process offline CPU's input_pkt_queue */
6567
	while ((skb = __skb_dequeue(&oldsd->process_queue))) {
L
Linus Torvalds 已提交
6568
		netif_rx(skb);
6569
		input_queue_head_incr(oldsd);
T
Tom Herbert 已提交
6570
	}
6571
	while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) {
6572
		netif_rx(skb);
6573 6574
		input_queue_head_incr(oldsd);
	}
L
Linus Torvalds 已提交
6575 6576 6577 6578 6579

	return NOTIFY_OK;
}


6580
/**
6581 6582 6583 6584
 *	netdev_increment_features - increment feature set by one
 *	@all: current feature set
 *	@one: new feature set
 *	@mask: mask feature set
6585 6586
 *
 *	Computes a new feature set after adding a device with feature set
6587 6588
 *	@one to the master device with current feature set @all.  Will not
 *	enable anything that is off in @mask. Returns the new feature set.
6589
 */
6590 6591
netdev_features_t netdev_increment_features(netdev_features_t all,
	netdev_features_t one, netdev_features_t mask)
6592
{
6593 6594 6595
	if (mask & NETIF_F_GEN_CSUM)
		mask |= NETIF_F_ALL_CSUM;
	mask |= NETIF_F_VLAN_CHALLENGED;
6596

6597 6598
	all |= one & (NETIF_F_ONE_FOR_ALL|NETIF_F_ALL_CSUM) & mask;
	all &= one | ~NETIF_F_ALL_FOR_ALL;
6599

6600 6601 6602
	/* If one device supports hw checksumming, set for all. */
	if (all & NETIF_F_GEN_CSUM)
		all &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
6603 6604 6605

	return all;
}
6606
EXPORT_SYMBOL(netdev_increment_features);
6607

6608
static struct hlist_head * __net_init netdev_create_hash(void)
6609 6610 6611 6612 6613 6614 6615 6616 6617 6618 6619 6620
{
	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;
}

6621
/* Initialize per network namespace state */
6622
static int __net_init netdev_init(struct net *net)
6623
{
6624 6625
	if (net != &init_net)
		INIT_LIST_HEAD(&net->dev_base_head);
6626

6627 6628 6629
	net->dev_name_head = netdev_create_hash();
	if (net->dev_name_head == NULL)
		goto err_name;
6630

6631 6632 6633
	net->dev_index_head = netdev_create_hash();
	if (net->dev_index_head == NULL)
		goto err_idx;
6634 6635

	return 0;
6636 6637 6638 6639 6640

err_idx:
	kfree(net->dev_name_head);
err_name:
	return -ENOMEM;
6641 6642
}

6643 6644 6645 6646 6647 6648
/**
 *	netdev_drivername - network driver for the device
 *	@dev: network device
 *
 *	Determine network driver for device.
 */
6649
const char *netdev_drivername(const struct net_device *dev)
6650
{
6651 6652
	const struct device_driver *driver;
	const struct device *parent;
6653
	const char *empty = "";
6654 6655 6656

	parent = dev->dev.parent;
	if (!parent)
6657
		return empty;
6658 6659 6660

	driver = parent->driver;
	if (driver && driver->name)
6661 6662
		return driver->name;
	return empty;
6663 6664
}

6665
static int __netdev_printk(const char *level, const struct net_device *dev,
6666 6667 6668 6669
			   struct va_format *vaf)
{
	int r;

6670
	if (dev && dev->dev.parent) {
6671 6672 6673 6674 6675 6676
		r = dev_printk_emit(level[1] - '0',
				    dev->dev.parent,
				    "%s %s %s: %pV",
				    dev_driver_string(dev->dev.parent),
				    dev_name(dev->dev.parent),
				    netdev_name(dev), vaf);
6677
	} else if (dev) {
6678
		r = printk("%s%s: %pV", level, netdev_name(dev), vaf);
6679
	} else {
6680
		r = printk("%s(NULL net_device): %pV", level, vaf);
6681
	}
6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693 6694 6695 6696 6697 6698

	return r;
}

int netdev_printk(const char *level, const struct net_device *dev,
		  const char *format, ...)
{
	struct va_format vaf;
	va_list args;
	int r;

	va_start(args, format);

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

	r = __netdev_printk(level, dev, &vaf);
6699

6700 6701 6702 6703 6704 6705 6706 6707 6708 6709 6710 6711 6712 6713 6714 6715 6716 6717 6718
	va_end(args);

	return r;
}
EXPORT_SYMBOL(netdev_printk);

#define define_netdev_printk_level(func, level)			\
int func(const struct net_device *dev, const char *fmt, ...)	\
{								\
	int r;							\
	struct va_format vaf;					\
	va_list args;						\
								\
	va_start(args, fmt);					\
								\
	vaf.fmt = fmt;						\
	vaf.va = &args;						\
								\
	r = __netdev_printk(level, dev, &vaf);			\
6719
								\
6720 6721 6722 6723 6724 6725 6726 6727 6728 6729 6730 6731 6732 6733
	va_end(args);						\
								\
	return r;						\
}								\
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);

6734
static void __net_exit netdev_exit(struct net *net)
6735 6736 6737 6738 6739
{
	kfree(net->dev_name_head);
	kfree(net->dev_index_head);
}

6740
static struct pernet_operations __net_initdata netdev_net_ops = {
6741 6742 6743 6744
	.init = netdev_init,
	.exit = netdev_exit,
};

6745
static void __net_exit default_device_exit(struct net *net)
6746
{
6747
	struct net_device *dev, *aux;
6748
	/*
6749
	 * Push all migratable network devices back to the
6750 6751 6752
	 * initial network namespace
	 */
	rtnl_lock();
6753
	for_each_netdev_safe(net, dev, aux) {
6754
		int err;
6755
		char fb_name[IFNAMSIZ];
6756 6757 6758 6759 6760

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

6761 6762 6763
		/* Leave virtual devices for the generic cleanup */
		if (dev->rtnl_link_ops)
			continue;
6764

L
Lucas De Marchi 已提交
6765
		/* Push remaining network devices to init_net */
6766 6767
		snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex);
		err = dev_change_net_namespace(dev, &init_net, fb_name);
6768
		if (err) {
6769 6770
			pr_emerg("%s: failed to move %s to init_net: %d\n",
				 __func__, dev->name, err);
6771
			BUG();
6772 6773 6774 6775 6776
		}
	}
	rtnl_unlock();
}

6777 6778 6779 6780 6781 6782 6783 6784 6785 6786 6787 6788 6789 6790 6791 6792 6793 6794 6795 6796 6797 6798 6799 6800 6801 6802 6803 6804
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;
	DEFINE_WAIT(wait);

	for (;;) {
		prepare_to_wait(&netdev_unregistering_wq, &wait,
				TASK_UNINTERRUPTIBLE);
		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();
		schedule();
	}
	finish_wait(&netdev_unregistering_wq, &wait);
}

6805 6806 6807
static void __net_exit default_device_exit_batch(struct list_head *net_list)
{
	/* At exit all network devices most be removed from a network
6808
	 * namespace.  Do this in the reverse order of registration.
6809 6810 6811 6812 6813 6814 6815
	 * 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);

6816 6817 6818 6819 6820 6821 6822 6823 6824 6825 6826 6827
	/* 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);
6828 6829 6830 6831 6832 6833 6834 6835 6836
	list_for_each_entry(net, net_list, exit_list) {
		for_each_netdev_reverse(net, dev) {
			if (dev->rtnl_link_ops)
				dev->rtnl_link_ops->dellink(dev, &dev_kill_list);
			else
				unregister_netdevice_queue(dev, &dev_kill_list);
		}
	}
	unregister_netdevice_many(&dev_kill_list);
E
Eric Dumazet 已提交
6837
	list_del(&dev_kill_list);
6838 6839 6840
	rtnl_unlock();
}

6841
static struct pernet_operations __net_initdata default_device_ops = {
6842
	.exit = default_device_exit,
6843
	.exit_batch = default_device_exit_batch,
6844 6845
};

L
Linus Torvalds 已提交
6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857 6858 6859 6860 6861 6862 6863 6864 6865
/*
 *	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;

6866
	if (netdev_kobject_init())
L
Linus Torvalds 已提交
6867 6868 6869
		goto out;

	INIT_LIST_HEAD(&ptype_all);
6870
	for (i = 0; i < PTYPE_HASH_SIZE; i++)
L
Linus Torvalds 已提交
6871 6872
		INIT_LIST_HEAD(&ptype_base[i]);

6873 6874
	INIT_LIST_HEAD(&offload_base);

6875 6876
	if (register_pernet_subsys(&netdev_net_ops))
		goto out;
L
Linus Torvalds 已提交
6877 6878 6879 6880 6881

	/*
	 *	Initialise the packet receive queues.
	 */

6882
	for_each_possible_cpu(i) {
E
Eric Dumazet 已提交
6883
		struct softnet_data *sd = &per_cpu(softnet_data, i);
L
Linus Torvalds 已提交
6884

C
Changli Gao 已提交
6885
		memset(sd, 0, sizeof(*sd));
E
Eric Dumazet 已提交
6886
		skb_queue_head_init(&sd->input_pkt_queue);
6887
		skb_queue_head_init(&sd->process_queue);
E
Eric Dumazet 已提交
6888 6889
		sd->completion_queue = NULL;
		INIT_LIST_HEAD(&sd->poll_list);
6890 6891
		sd->output_queue = NULL;
		sd->output_queue_tailp = &sd->output_queue;
E
Eric Dumazet 已提交
6892
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
6893 6894 6895 6896
		sd->csd.func = rps_trigger_softirq;
		sd->csd.info = sd;
		sd->csd.flags = 0;
		sd->cpu = i;
6897
#endif
T
Tom Herbert 已提交
6898

E
Eric Dumazet 已提交
6899 6900 6901 6902
		sd->backlog.poll = process_backlog;
		sd->backlog.weight = weight_p;
		sd->backlog.gro_list = NULL;
		sd->backlog.gro_count = 0;
6903 6904 6905 6906

#ifdef CONFIG_NET_FLOW_LIMIT
		sd->flow_limit = NULL;
#endif
L
Linus Torvalds 已提交
6907 6908 6909 6910
	}

	dev_boot_phase = 0;

6911 6912 6913 6914 6915 6916 6917 6918 6919 6920 6921 6922 6923 6924 6925
	/* 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;

6926 6927
	open_softirq(NET_TX_SOFTIRQ, net_tx_action);
	open_softirq(NET_RX_SOFTIRQ, net_rx_action);
L
Linus Torvalds 已提交
6928 6929 6930 6931 6932 6933 6934 6935 6936

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

subsys_initcall(net_dev_init);