dev.c 172.8 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

447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487

/**
 *	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.
 */
void __dev_remove_offload(struct packet_offload *po)
{
	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);
500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
}
EXPORT_SYMBOL(__dev_remove_offload);

/**
 *	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
523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
/******************************************************************************

		      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));
550
			strlcpy(s[i].name, name, IFNAMSIZ);
L
Linus Torvalds 已提交
551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574
			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] != ' ' &&
575
		    !strcmp(dev->name, s[i].name)) {
L
Linus Torvalds 已提交
576 577 578 579 580 581 582 583 584
			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
585
EXPORT_SYMBOL(netdev_boot_setup_check);
L
Linus Torvalds 已提交
586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609


/**
 *	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
	 */
610
	if (__dev_get_by_name(&init_net, name))
L
Linus Torvalds 已提交
611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655
		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
656
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
657 658 659 660 661 662 663 664 665
 *	@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.
 */

666
struct net_device *__dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
667
{
E
Eric Dumazet 已提交
668 669
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);
L
Linus Torvalds 已提交
670

671
	hlist_for_each_entry(dev, head, name_hlist)
L
Linus Torvalds 已提交
672 673
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;
E
Eric Dumazet 已提交
674

L
Linus Torvalds 已提交
675 676
	return NULL;
}
E
Eric Dumazet 已提交
677
EXPORT_SYMBOL(__dev_get_by_name);
L
Linus Torvalds 已提交
678

679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695
/**
 *	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);

696
	hlist_for_each_entry_rcu(dev, head, name_hlist)
697 698 699 700 701 702 703
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_name_rcu);

L
Linus Torvalds 已提交
704 705
/**
 *	dev_get_by_name		- find a device by its name
706
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
707 708 709 710 711 712 713 714 715
 *	@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.
 */

716
struct net_device *dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
717 718 719
{
	struct net_device *dev;

720 721
	rcu_read_lock();
	dev = dev_get_by_name_rcu(net, name);
L
Linus Torvalds 已提交
722 723
	if (dev)
		dev_hold(dev);
724
	rcu_read_unlock();
L
Linus Torvalds 已提交
725 726
	return dev;
}
E
Eric Dumazet 已提交
727
EXPORT_SYMBOL(dev_get_by_name);
L
Linus Torvalds 已提交
728 729 730

/**
 *	__dev_get_by_index - find a device by its ifindex
731
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
732 733 734 735 736 737 738 739 740
 *	@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.
 */

741
struct net_device *__dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
742
{
E
Eric Dumazet 已提交
743 744
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);
L
Linus Torvalds 已提交
745

746
	hlist_for_each_entry(dev, head, index_hlist)
L
Linus Torvalds 已提交
747 748
		if (dev->ifindex == ifindex)
			return dev;
E
Eric Dumazet 已提交
749

L
Linus Torvalds 已提交
750 751
	return NULL;
}
E
Eric Dumazet 已提交
752
EXPORT_SYMBOL(__dev_get_by_index);
L
Linus Torvalds 已提交
753

754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769
/**
 *	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);

770
	hlist_for_each_entry_rcu(dev, head, index_hlist)
771 772 773 774 775 776 777
		if (dev->ifindex == ifindex)
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_index_rcu);

L
Linus Torvalds 已提交
778 779 780

/**
 *	dev_get_by_index - find a device by its ifindex
781
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
782 783 784 785 786 787 788 789
 *	@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.
 */

790
struct net_device *dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
791 792 793
{
	struct net_device *dev;

794 795
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, ifindex);
L
Linus Torvalds 已提交
796 797
	if (dev)
		dev_hold(dev);
798
	rcu_read_unlock();
L
Linus Torvalds 已提交
799 800
	return dev;
}
E
Eric Dumazet 已提交
801
EXPORT_SYMBOL(dev_get_by_index);
L
Linus Torvalds 已提交
802

803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836
/**
 *	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
837
/**
838
 *	dev_getbyhwaddr_rcu - find a device by its hardware address
839
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
840 841 842 843
 *	@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
844 845
 *	is not found or a pointer to the device.
 *	The caller must hold RCU or RTNL.
846
 *	The returned device has not had its ref count increased
L
Linus Torvalds 已提交
847 848 849 850
 *	and the caller must therefore be careful about locking
 *
 */

851 852
struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
				       const char *ha)
L
Linus Torvalds 已提交
853 854 855
{
	struct net_device *dev;

856
	for_each_netdev_rcu(net, dev)
L
Linus Torvalds 已提交
857 858
		if (dev->type == type &&
		    !memcmp(dev->dev_addr, ha, dev->addr_len))
859 860 861
			return dev;

	return NULL;
L
Linus Torvalds 已提交
862
}
863
EXPORT_SYMBOL(dev_getbyhwaddr_rcu);
864

865
struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type)
L
Linus Torvalds 已提交
866 867 868
{
	struct net_device *dev;

869
	ASSERT_RTNL();
870
	for_each_netdev(net, dev)
871
		if (dev->type == type)
872 873 874
			return dev;

	return NULL;
875 876 877
}
EXPORT_SYMBOL(__dev_getfirstbyhwtype);

878
struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type)
879
{
880
	struct net_device *dev, *ret = NULL;
881

882 883 884 885 886 887 888 889 890
	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
891 892 893 894
}
EXPORT_SYMBOL(dev_getfirstbyhwtype);

/**
E
Eric Dumazet 已提交
895
 *	dev_get_by_flags_rcu - find any device with given flags
896
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
897 898 899 900
 *	@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
901 902
 *	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
903 904
 */

E
Eric Dumazet 已提交
905
struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short if_flags,
E
Eric Dumazet 已提交
906
				    unsigned short mask)
L
Linus Torvalds 已提交
907
{
908
	struct net_device *dev, *ret;
L
Linus Torvalds 已提交
909

910
	ret = NULL;
911
	for_each_netdev_rcu(net, dev) {
L
Linus Torvalds 已提交
912
		if (((dev->flags ^ if_flags) & mask) == 0) {
913
			ret = dev;
L
Linus Torvalds 已提交
914 915 916
			break;
		}
	}
917
	return ret;
L
Linus Torvalds 已提交
918
}
E
Eric Dumazet 已提交
919
EXPORT_SYMBOL(dev_get_by_flags_rcu);
L
Linus Torvalds 已提交
920 921 922 923 924 925

/**
 *	dev_valid_name - check if name is okay for network device
 *	@name: name string
 *
 *	Network device names need to be valid file names to
926 927
 *	to allow sysfs to work.  We also disallow any kind of
 *	whitespace.
L
Linus Torvalds 已提交
928
 */
929
bool dev_valid_name(const char *name)
L
Linus Torvalds 已提交
930
{
931
	if (*name == '\0')
932
		return false;
933
	if (strlen(name) >= IFNAMSIZ)
934
		return false;
935
	if (!strcmp(name, ".") || !strcmp(name, ".."))
936
		return false;
937 938 939

	while (*name) {
		if (*name == '/' || isspace(*name))
940
			return false;
941 942
		name++;
	}
943
	return true;
L
Linus Torvalds 已提交
944
}
E
Eric Dumazet 已提交
945
EXPORT_SYMBOL(dev_valid_name);
L
Linus Torvalds 已提交
946 947

/**
948 949
 *	__dev_alloc_name - allocate a name for a device
 *	@net: network namespace to allocate the device name in
L
Linus Torvalds 已提交
950
 *	@name: name format string
951
 *	@buf:  scratch buffer and result name string
L
Linus Torvalds 已提交
952 953
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
S
Stephen Hemminger 已提交
954 955 956 957 958 959
 *	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
960 961
 */

962
static int __dev_alloc_name(struct net *net, const char *name, char *buf)
L
Linus Torvalds 已提交
963 964 965 966
{
	int i = 0;
	const char *p;
	const int max_netdevices = 8*PAGE_SIZE;
S
Stephen Hemminger 已提交
967
	unsigned long *inuse;
L
Linus Torvalds 已提交
968 969 970 971 972 973 974 975 976 977 978 979 980
	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
981
		inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC);
L
Linus Torvalds 已提交
982 983 984
		if (!inuse)
			return -ENOMEM;

985
		for_each_netdev(net, d) {
L
Linus Torvalds 已提交
986 987 988 989 990 991
			if (!sscanf(d->name, name, &i))
				continue;
			if (i < 0 || i >= max_netdevices)
				continue;

			/*  avoid cases where sscanf is not exact inverse of printf */
992
			snprintf(buf, IFNAMSIZ, name, i);
L
Linus Torvalds 已提交
993 994 995 996 997 998 999 1000
			if (!strncmp(buf, d->name, IFNAMSIZ))
				set_bit(i, inuse);
		}

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

1001 1002
	if (buf != name)
		snprintf(buf, IFNAMSIZ, name, i);
1003
	if (!__dev_get_by_name(net, buf))
L
Linus Torvalds 已提交
1004 1005 1006 1007 1008 1009 1010 1011 1012
		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;
}

1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032
/**
 *	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;

1033 1034
	BUG_ON(!dev_net(dev));
	net = dev_net(dev);
1035 1036 1037 1038 1039
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
}
E
Eric Dumazet 已提交
1040
EXPORT_SYMBOL(dev_alloc_name);
1041

1042 1043 1044
static int dev_alloc_name_ns(struct net *net,
			     struct net_device *dev,
			     const char *name)
1045
{
1046 1047
	char buf[IFNAMSIZ];
	int ret;
1048

1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
	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);
1060

1061 1062 1063
	if (!dev_valid_name(name))
		return -EINVAL;

1064
	if (strchr(name, '%'))
1065
		return dev_alloc_name_ns(net, dev, name);
1066 1067
	else if (__dev_get_by_name(net, name))
		return -EEXIST;
1068 1069
	else if (dev->name != name)
		strlcpy(dev->name, name, IFNAMSIZ);
1070 1071 1072

	return 0;
}
L
Linus Torvalds 已提交
1073 1074 1075 1076 1077 1078 1079 1080 1081

/**
 *	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.
 */
1082
int dev_change_name(struct net_device *dev, const char *newname)
L
Linus Torvalds 已提交
1083
{
1084
	char oldname[IFNAMSIZ];
L
Linus Torvalds 已提交
1085
	int err = 0;
1086
	int ret;
1087
	struct net *net;
L
Linus Torvalds 已提交
1088 1089

	ASSERT_RTNL();
1090
	BUG_ON(!dev_net(dev));
L
Linus Torvalds 已提交
1091

1092
	net = dev_net(dev);
L
Linus Torvalds 已提交
1093 1094 1095
	if (dev->flags & IFF_UP)
		return -EBUSY;

1096
	write_seqcount_begin(&devnet_rename_seq);
1097 1098

	if (strncmp(newname, dev->name, IFNAMSIZ) == 0) {
1099
		write_seqcount_end(&devnet_rename_seq);
1100
		return 0;
1101
	}
1102

1103 1104
	memcpy(oldname, dev->name, IFNAMSIZ);

1105
	err = dev_get_valid_name(net, dev, newname);
1106
	if (err < 0) {
1107
		write_seqcount_end(&devnet_rename_seq);
1108
		return err;
1109
	}
L
Linus Torvalds 已提交
1110

1111
rollback:
1112 1113 1114
	ret = device_rename(&dev->dev, dev->name);
	if (ret) {
		memcpy(dev->name, oldname, IFNAMSIZ);
1115
		write_seqcount_end(&devnet_rename_seq);
1116
		return ret;
1117
	}
1118

1119
	write_seqcount_end(&devnet_rename_seq);
1120

1121
	write_lock_bh(&dev_base_lock);
1122
	hlist_del_rcu(&dev->name_hlist);
1123 1124 1125 1126 1127 1128
	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));
1129 1130
	write_unlock_bh(&dev_base_lock);

1131
	ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
1132 1133 1134
	ret = notifier_to_errno(ret);

	if (ret) {
1135 1136
		/* err >= 0 after dev_alloc_name() or stores the first errno */
		if (err >= 0) {
1137
			err = ret;
1138
			write_seqcount_begin(&devnet_rename_seq);
1139 1140
			memcpy(dev->name, oldname, IFNAMSIZ);
			goto rollback;
1141
		} else {
1142
			pr_err("%s: name change rollback failed: %d\n",
1143
			       dev->name, ret);
1144 1145
		}
	}
L
Linus Torvalds 已提交
1146 1147 1148 1149

	return err;
}

1150 1151 1152 1153
/**
 *	dev_set_alias - change ifalias of a device
 *	@dev: device
 *	@alias: name up to IFALIASZ
1154
 *	@len: limit of bytes to copy from info
1155 1156 1157 1158 1159
 *
 *	Set ifalias for a device,
 */
int dev_set_alias(struct net_device *dev, const char *alias, size_t len)
{
1160 1161
	char *new_ifalias;

1162 1163 1164 1165 1166
	ASSERT_RTNL();

	if (len >= IFALIASZ)
		return -EINVAL;

1167
	if (!len) {
1168 1169
		kfree(dev->ifalias);
		dev->ifalias = NULL;
1170 1171 1172
		return 0;
	}

1173 1174
	new_ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL);
	if (!new_ifalias)
1175
		return -ENOMEM;
1176
	dev->ifalias = new_ifalias;
1177 1178 1179 1180 1181 1182

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


1183
/**
S
Stephen Hemminger 已提交
1184
 *	netdev_features_change - device changes features
1185 1186 1187 1188 1189 1190
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed features.
 */
void netdev_features_change(struct net_device *dev)
{
1191
	call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev);
1192 1193 1194
}
EXPORT_SYMBOL(netdev_features_change);

L
Linus Torvalds 已提交
1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205
/**
 *	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) {
1206
		call_netdevice_notifiers(NETDEV_CHANGE, dev);
1207
		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
L
Linus Torvalds 已提交
1208 1209
	}
}
E
Eric Dumazet 已提交
1210
EXPORT_SYMBOL(netdev_state_change);
L
Linus Torvalds 已提交
1211

1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222
/**
 * 	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)
1223
{
1224 1225 1226
	rtnl_lock();
	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, dev);
	rtnl_unlock();
1227
}
1228
EXPORT_SYMBOL(netdev_notify_peers);
1229

1230
static int __dev_open(struct net_device *dev)
L
Linus Torvalds 已提交
1231
{
1232
	const struct net_device_ops *ops = dev->netdev_ops;
1233
	int ret;
L
Linus Torvalds 已提交
1234

1235 1236
	ASSERT_RTNL();

L
Linus Torvalds 已提交
1237 1238 1239
	if (!netif_device_present(dev))
		return -ENODEV;

1240 1241 1242 1243
	/* 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
	 */
1244
	netpoll_rx_disable(dev);
1245

1246 1247 1248 1249 1250
	ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

L
Linus Torvalds 已提交
1251
	set_bit(__LINK_STATE_START, &dev->state);
1252

1253 1254
	if (ops->ndo_validate_addr)
		ret = ops->ndo_validate_addr(dev);
1255

1256 1257
	if (!ret && ops->ndo_open)
		ret = ops->ndo_open(dev);
L
Linus Torvalds 已提交
1258

1259 1260
	netpoll_rx_enable(dev);

1261 1262 1263
	if (ret)
		clear_bit(__LINK_STATE_START, &dev->state);
	else {
L
Linus Torvalds 已提交
1264
		dev->flags |= IFF_UP;
1265
		net_dmaengine_get();
1266
		dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
1267
		dev_activate(dev);
1268
		add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
1269
	}
1270

L
Linus Torvalds 已提交
1271 1272 1273 1274
	return ret;
}

/**
1275 1276
 *	dev_open	- prepare an interface for use.
 *	@dev:	device to open
L
Linus Torvalds 已提交
1277
 *
1278 1279 1280 1281 1282 1283 1284
 *	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
1285
 */
1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296
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;

1297
	rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1298 1299 1300 1301 1302 1303
	call_netdevice_notifiers(NETDEV_UP, dev);

	return ret;
}
EXPORT_SYMBOL(dev_open);

1304
static int __dev_close_many(struct list_head *head)
L
Linus Torvalds 已提交
1305
{
1306
	struct net_device *dev;
1307

1308
	ASSERT_RTNL();
1309 1310
	might_sleep();

1311
	list_for_each_entry(dev, head, close_list) {
1312
		call_netdevice_notifiers(NETDEV_GOING_DOWN, dev);
L
Linus Torvalds 已提交
1313

1314
		clear_bit(__LINK_STATE_START, &dev->state);
L
Linus Torvalds 已提交
1315

1316 1317 1318 1319 1320 1321 1322 1323
		/* 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
1324

1325
	dev_deactivate_many(head);
1326

1327
	list_for_each_entry(dev, head, close_list) {
1328
		const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
1329

1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348
		/*
		 *	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)
{
1349
	int retval;
1350 1351
	LIST_HEAD(single);

1352
	/* Temporarily disable netpoll until the interface is down */
1353
	netpoll_rx_disable(dev);
1354

1355
	list_add(&dev->close_list, &single);
1356 1357
	retval = __dev_close_many(&single);
	list_del(&single);
1358 1359

	netpoll_rx_enable(dev);
1360
	return retval;
1361 1362
}

E
Eric Dumazet 已提交
1363
static int dev_close_many(struct list_head *head)
1364 1365
{
	struct net_device *dev, *tmp;
L
Linus Torvalds 已提交
1366

1367 1368
	/* Remove the devices that don't need to be closed */
	list_for_each_entry_safe(dev, tmp, head, close_list)
1369
		if (!(dev->flags & IFF_UP))
1370
			list_del_init(&dev->close_list);
1371 1372

	__dev_close_many(head);
L
Linus Torvalds 已提交
1373

1374
	list_for_each_entry_safe(dev, tmp, head, close_list) {
1375
		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1376
		call_netdevice_notifiers(NETDEV_DOWN, dev);
1377
		list_del_init(&dev->close_list);
1378
	}
1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393

	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)
{
1394 1395
	if (dev->flags & IFF_UP) {
		LIST_HEAD(single);
L
Linus Torvalds 已提交
1396

1397
		/* Block netpoll rx while the interface is going down */
1398
		netpoll_rx_disable(dev);
1399

1400
		list_add(&dev->close_list, &single);
1401 1402
		dev_close_many(&single);
		list_del(&single);
1403 1404

		netpoll_rx_enable(dev);
1405
	}
1406
	return 0;
L
Linus Torvalds 已提交
1407
}
E
Eric Dumazet 已提交
1408
EXPORT_SYMBOL(dev_close);
L
Linus Torvalds 已提交
1409 1410


1411 1412 1413 1414 1415 1416 1417 1418 1419 1420
/**
 *	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)
{
1421 1422 1423 1424 1425 1426 1427
	/*
	 * 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);

1428 1429 1430 1431
	/* the same for macvlan devices */
	if (netif_is_macvlan(dev))
		dev = macvlan_dev_real_dev(dev);

M
Michał Mirosław 已提交
1432 1433
	dev->wanted_features &= ~NETIF_F_LRO;
	netdev_update_features(dev);
1434

1435 1436
	if (unlikely(dev->features & NETIF_F_LRO))
		netdev_WARN(dev, "failed to disable LRO!\n");
1437 1438 1439
}
EXPORT_SYMBOL(dev_disable_lro);

1440 1441 1442 1443 1444 1445 1446 1447
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);
}
1448

1449 1450
static int dev_boot_phase = 1;

L
Linus Torvalds 已提交
1451 1452 1453 1454 1455 1456 1457 1458 1459 1460
/**
 *	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
1461
 *	to the new notifier to allow device to have a race free
L
Linus Torvalds 已提交
1462 1463 1464 1465 1466 1467
 *	view of the network device list.
 */

int register_netdevice_notifier(struct notifier_block *nb)
{
	struct net_device *dev;
1468
	struct net_device *last;
1469
	struct net *net;
L
Linus Torvalds 已提交
1470 1471 1472
	int err;

	rtnl_lock();
1473
	err = raw_notifier_chain_register(&netdev_chain, nb);
1474 1475
	if (err)
		goto unlock;
1476 1477 1478 1479
	if (dev_boot_phase)
		goto unlock;
	for_each_net(net) {
		for_each_netdev(net, dev) {
1480
			err = call_netdevice_notifier(nb, NETDEV_REGISTER, dev);
1481 1482 1483 1484 1485 1486
			err = notifier_to_errno(err);
			if (err)
				goto rollback;

			if (!(dev->flags & IFF_UP))
				continue;
L
Linus Torvalds 已提交
1487

1488
			call_netdevice_notifier(nb, NETDEV_UP, dev);
1489
		}
L
Linus Torvalds 已提交
1490
	}
1491 1492

unlock:
L
Linus Torvalds 已提交
1493 1494
	rtnl_unlock();
	return err;
1495 1496 1497

rollback:
	last = dev;
1498 1499 1500
	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev == last)
1501
				goto outroll;
1502

1503
			if (dev->flags & IFF_UP) {
1504 1505 1506
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1507
			}
1508
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1509 1510
		}
	}
1511

1512
outroll:
1513
	raw_notifier_chain_unregister(&netdev_chain, nb);
1514
	goto unlock;
L
Linus Torvalds 已提交
1515
}
E
Eric Dumazet 已提交
1516
EXPORT_SYMBOL(register_netdevice_notifier);
L
Linus Torvalds 已提交
1517 1518 1519 1520 1521 1522 1523 1524 1525

/**
 *	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.
1526 1527 1528 1529
 *
 * 	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
1530 1531 1532 1533
 */

int unregister_netdevice_notifier(struct notifier_block *nb)
{
1534 1535
	struct net_device *dev;
	struct net *net;
1536 1537 1538
	int err;

	rtnl_lock();
1539
	err = raw_notifier_chain_unregister(&netdev_chain, nb);
1540 1541 1542 1543 1544 1545
	if (err)
		goto unlock;

	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev->flags & IFF_UP) {
1546 1547 1548
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1549
			}
1550
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1551 1552 1553
		}
	}
unlock:
1554 1555
	rtnl_unlock();
	return err;
L
Linus Torvalds 已提交
1556
}
E
Eric Dumazet 已提交
1557
EXPORT_SYMBOL(unregister_netdevice_notifier);
L
Linus Torvalds 已提交
1558

1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577
/**
 *	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().
 */

int call_netdevice_notifiers_info(unsigned long val, struct net_device *dev,
				  struct netdev_notifier_info *info)
{
	ASSERT_RTNL();
	netdev_notifier_info_init(info, dev);
	return raw_notifier_call_chain(&netdev_chain, val, info);
}
EXPORT_SYMBOL(call_netdevice_notifiers_info);

L
Linus Torvalds 已提交
1578 1579 1580
/**
 *	call_netdevice_notifiers - call all network notifier blocks
 *      @val: value passed unmodified to notifier function
1581
 *      @dev: net_device pointer passed unmodified to notifier function
L
Linus Torvalds 已提交
1582 1583
 *
 *	Call all network notifier blocks.  Parameters and return value
1584
 *	are as for raw_notifier_call_chain().
L
Linus Torvalds 已提交
1585 1586
 */

1587
int call_netdevice_notifiers(unsigned long val, struct net_device *dev)
L
Linus Torvalds 已提交
1588
{
1589 1590 1591
	struct netdev_notifier_info info;

	return call_netdevice_notifiers_info(val, dev, &info);
L
Linus Torvalds 已提交
1592
}
1593
EXPORT_SYMBOL(call_netdevice_notifiers);
L
Linus Torvalds 已提交
1594

1595
static struct static_key netstamp_needed __read_mostly;
1596
#ifdef HAVE_JUMP_LABEL
1597
/* We are not allowed to call static_key_slow_dec() from irq context
1598
 * If net_disable_timestamp() is called from irq context, defer the
1599
 * static_key_slow_dec() calls.
1600 1601 1602
 */
static atomic_t netstamp_needed_deferred;
#endif
L
Linus Torvalds 已提交
1603 1604 1605

void net_enable_timestamp(void)
{
1606 1607 1608 1609 1610
#ifdef HAVE_JUMP_LABEL
	int deferred = atomic_xchg(&netstamp_needed_deferred, 0);

	if (deferred) {
		while (--deferred)
1611
			static_key_slow_dec(&netstamp_needed);
1612 1613 1614
		return;
	}
#endif
1615
	static_key_slow_inc(&netstamp_needed);
L
Linus Torvalds 已提交
1616
}
E
Eric Dumazet 已提交
1617
EXPORT_SYMBOL(net_enable_timestamp);
L
Linus Torvalds 已提交
1618 1619 1620

void net_disable_timestamp(void)
{
1621 1622 1623 1624 1625 1626
#ifdef HAVE_JUMP_LABEL
	if (in_interrupt()) {
		atomic_inc(&netstamp_needed_deferred);
		return;
	}
#endif
1627
	static_key_slow_dec(&netstamp_needed);
L
Linus Torvalds 已提交
1628
}
E
Eric Dumazet 已提交
1629
EXPORT_SYMBOL(net_disable_timestamp);
L
Linus Torvalds 已提交
1630

E
Eric Dumazet 已提交
1631
static inline void net_timestamp_set(struct sk_buff *skb)
L
Linus Torvalds 已提交
1632
{
1633
	skb->tstamp.tv64 = 0;
1634
	if (static_key_false(&netstamp_needed))
1635
		__net_timestamp(skb);
L
Linus Torvalds 已提交
1636 1637
}

1638
#define net_timestamp_check(COND, SKB)			\
1639
	if (static_key_false(&netstamp_needed)) {		\
1640 1641 1642
		if ((COND) && !(SKB)->tstamp.tv64)	\
			__net_timestamp(SKB);		\
	}						\
E
Eric Dumazet 已提交
1643

1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664
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;
}

1665 1666 1667 1668 1669 1670 1671 1672
/**
 * dev_forward_skb - loopback an skb to another netif
 *
 * @dev: destination network device
 * @skb: buffer to forward
 *
 * return values:
 *	NET_RX_SUCCESS	(no congestion)
1673
 *	NET_RX_DROP     (packet was dropped, but freed)
1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684
 *
 * 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)
{
1685 1686 1687 1688 1689 1690 1691 1692
	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;
		}
	}

1693
	if (unlikely(!is_skb_forwardable(dev, skb))) {
1694
		atomic_long_inc(&dev->rx_dropped);
1695
		kfree_skb(skb);
1696
		return NET_RX_DROP;
1697
	}
1698

1699
	skb_scrub_packet(skb, true);
1700
	skb->protocol = eth_type_trans(skb, dev);
1701

1702 1703 1704 1705
	return netif_rx(skb);
}
EXPORT_SYMBOL_GPL(dev_forward_skb);

1706 1707 1708 1709
static inline int deliver_skb(struct sk_buff *skb,
			      struct packet_type *pt_prev,
			      struct net_device *orig_dev)
{
1710 1711
	if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
		return -ENOMEM;
1712 1713 1714 1715
	atomic_inc(&skb->users);
	return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
}

1716 1717
static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
{
1718
	if (!ptype->af_packet_priv || !skb->sk)
1719 1720 1721 1722 1723 1724 1725 1726 1727 1728
		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
1729 1730 1731 1732 1733
/*
 *	Support routine. Sends outgoing frames to any network
 *	taps currently in use.
 */

1734
static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
L
Linus Torvalds 已提交
1735 1736
{
	struct packet_type *ptype;
1737 1738
	struct sk_buff *skb2 = NULL;
	struct packet_type *pt_prev = NULL;
1739

L
Linus Torvalds 已提交
1740 1741 1742 1743 1744 1745
	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) &&
1746
		    (!skb_loop_sk(ptype, skb))) {
1747 1748 1749 1750 1751 1752 1753
			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
1754 1755 1756
			if (!skb2)
				break;

1757 1758
			net_timestamp_set(skb2);

L
Linus Torvalds 已提交
1759 1760 1761 1762
			/* skb->nh should be correctly
			   set by sender, so that the second statement is
			   just protection against buggy protocols.
			 */
1763
			skb_reset_mac_header(skb2);
L
Linus Torvalds 已提交
1764

1765
			if (skb_network_header(skb2) < skb2->data ||
1766
			    skb_network_header(skb2) > skb_tail_pointer(skb2)) {
1767 1768 1769
				net_crit_ratelimited("protocol %04x is buggy, dev %s\n",
						     ntohs(skb2->protocol),
						     dev->name);
1770
				skb_reset_network_header(skb2);
L
Linus Torvalds 已提交
1771 1772
			}

1773
			skb2->transport_header = skb2->network_header;
L
Linus Torvalds 已提交
1774
			skb2->pkt_type = PACKET_OUTGOING;
1775
			pt_prev = ptype;
L
Linus Torvalds 已提交
1776 1777
		}
	}
1778 1779
	if (pt_prev)
		pt_prev->func(skb2, skb->dev, pt_prev, skb->dev);
L
Linus Torvalds 已提交
1780 1781 1782
	rcu_read_unlock();
}

1783 1784
/**
 * netif_setup_tc - Handle tc mappings on real_num_tx_queues change
1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795
 * @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
1796
static void netif_setup_tc(struct net_device *dev, unsigned int txq)
1797 1798 1799 1800 1801 1802
{
	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) {
1803
		pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n");
1804 1805 1806 1807 1808 1809 1810 1811 1812 1813
		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) {
1814 1815
			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);
1816 1817 1818 1819 1820
			netdev_set_prio_tc_map(dev, i, 0);
		}
	}
}

1821 1822 1823 1824 1825
#ifdef CONFIG_XPS
static DEFINE_MUTEX(xps_map_mutex);
#define xmap_dereference(P)		\
	rcu_dereference_protected((P), lockdep_is_held(&xps_map_mutex))

1826 1827
static struct xps_map *remove_xps_queue(struct xps_dev_maps *dev_maps,
					int cpu, u16 index)
1828
{
1829 1830
	struct xps_map *map = NULL;
	int pos;
1831

1832 1833
	if (dev_maps)
		map = xmap_dereference(dev_maps->cpu_map[cpu]);
1834

1835 1836
	for (pos = 0; map && pos < map->len; pos++) {
		if (map->queues[pos] == index) {
1837 1838 1839
			if (map->len > 1) {
				map->queues[pos] = map->queues[--map->len];
			} else {
1840
				RCU_INIT_POINTER(dev_maps->cpu_map[cpu], NULL);
1841 1842 1843
				kfree_rcu(map, rcu);
				map = NULL;
			}
1844
			break;
1845 1846 1847
		}
	}

1848 1849 1850
	return map;
}

1851
static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
1852 1853
{
	struct xps_dev_maps *dev_maps;
1854
	int cpu, i;
1855 1856 1857 1858 1859 1860 1861 1862 1863
	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) {
1864 1865 1866 1867 1868
		for (i = index; i < dev->num_tx_queues; i++) {
			if (!remove_xps_queue(dev_maps, cpu, i))
				break;
		}
		if (i == dev->num_tx_queues)
1869 1870 1871 1872
			active = true;
	}

	if (!active) {
1873 1874 1875 1876
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

1877 1878 1879 1880
	for (i = index; i < dev->num_tx_queues; i++)
		netdev_queue_numa_node_write(netdev_get_tx_queue(dev, i),
					     NUMA_NO_NODE);

1881 1882 1883 1884
out_no_maps:
	mutex_unlock(&xps_map_mutex);
}

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 1919
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;
}

1920 1921
int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
			u16 index)
1922
{
1923
	struct xps_dev_maps *dev_maps, *new_dev_maps = NULL;
1924 1925
	struct xps_map *map, *new_map;
	int maps_sz = max_t(unsigned int, XPS_DEV_MAPS_SIZE, L1_CACHE_BYTES);
1926 1927
	int cpu, numa_node_id = -2;
	bool active = false;
1928 1929 1930 1931 1932

	mutex_lock(&xps_map_mutex);

	dev_maps = xmap_dereference(dev->xps_maps);

1933 1934 1935 1936 1937 1938 1939
	/* 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);
1940 1941
		if (!new_dev_maps) {
			mutex_unlock(&xps_map_mutex);
1942
			return -ENOMEM;
1943
		}
1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957

		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;

1958
	for_each_possible_cpu(cpu) {
1959 1960 1961 1962 1963 1964 1965 1966 1967 1968
		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;
1969 1970 1971 1972 1973 1974
#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
1975 1976 1977 1978
		} 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);
1979
		}
1980

1981 1982
	}

1983 1984
	rcu_assign_pointer(dev->xps_maps, new_dev_maps);

1985
	/* Cleanup old maps */
1986 1987 1988 1989 1990 1991 1992
	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);
		}
1993

1994
		kfree_rcu(dev_maps, rcu);
1995 1996
	}

1997 1998
	dev_maps = new_dev_maps;
	active = true;
1999

2000 2001
out_no_new_maps:
	/* update Tx queue numa node */
2002 2003 2004 2005
	netdev_queue_numa_node_write(netdev_get_tx_queue(dev, index),
				     (numa_node_id >= 0) ? numa_node_id :
				     NUMA_NO_NODE);

2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024
	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:
2025 2026 2027 2028
	mutex_unlock(&xps_map_mutex);

	return 0;
error:
2029 2030 2031 2032 2033 2034 2035 2036 2037
	/* 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);
	}

2038 2039 2040 2041 2042 2043 2044 2045
	mutex_unlock(&xps_map_mutex);

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

#endif
2046 2047 2048 2049
/*
 * 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.
 */
2050
int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
2051
{
T
Tom Herbert 已提交
2052 2053
	int rc;

2054 2055
	if (txq < 1 || txq > dev->num_tx_queues)
		return -EINVAL;
2056

2057 2058
	if (dev->reg_state == NETREG_REGISTERED ||
	    dev->reg_state == NETREG_UNREGISTERING) {
2059 2060
		ASSERT_RTNL();

T
Tom Herbert 已提交
2061 2062
		rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
						  txq);
T
Tom Herbert 已提交
2063 2064 2065
		if (rc)
			return rc;

2066 2067 2068
		if (dev->num_tc)
			netif_setup_tc(dev, txq);

2069
		if (txq < dev->real_num_tx_queues) {
2070
			qdisc_reset_all_tx_gt(dev, txq);
2071 2072 2073 2074
#ifdef CONFIG_XPS
			netif_reset_xps_queues_gt(dev, txq);
#endif
		}
2075
	}
2076 2077 2078

	dev->real_num_tx_queues = txq;
	return 0;
2079 2080
}
EXPORT_SYMBOL(netif_set_real_num_tx_queues);
2081

2082 2083 2084 2085 2086 2087 2088 2089
#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
2090 2091
 *	negative error code.  If called before registration, it always
 *	succeeds.
2092 2093 2094 2095 2096
 */
int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
{
	int rc;

T
Tom Herbert 已提交
2097 2098 2099
	if (rxq < 1 || rxq > dev->num_rx_queues)
		return -EINVAL;

2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114
	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

2115 2116
/**
 * netif_get_num_default_rss_queues - default number of RSS queues
2117 2118 2119 2120
 *
 * This routine should set an upper limit on the number of RSS queues
 * used by default by multiqueue devices.
 */
2121
int netif_get_num_default_rss_queues(void)
2122 2123 2124 2125 2126
{
	return min_t(int, DEFAULT_MAX_NUM_RSS_QUEUES, num_online_cpus());
}
EXPORT_SYMBOL(netif_get_num_default_rss_queues);

2127
static inline void __netif_reschedule(struct Qdisc *q)
2128
{
2129 2130
	struct softnet_data *sd;
	unsigned long flags;
2131

2132 2133
	local_irq_save(flags);
	sd = &__get_cpu_var(softnet_data);
2134 2135 2136
	q->next_sched = NULL;
	*sd->output_queue_tailp = q;
	sd->output_queue_tailp = &q->next_sched;
2137 2138 2139 2140 2141 2142 2143 2144
	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);
2145 2146 2147
}
EXPORT_SYMBOL(__netif_schedule);

2148 2149 2150 2151 2152 2153 2154 2155 2156 2157
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)
2158
{
2159
	unsigned long flags;
2160

2161 2162 2163 2164 2165
	if (likely(atomic_read(&skb->users) == 1)) {
		smp_rmb();
		atomic_set(&skb->users, 0);
	} else if (likely(!atomic_dec_and_test(&skb->users))) {
		return;
2166
	}
2167 2168 2169 2170 2171 2172
	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);
2173
}
2174
EXPORT_SYMBOL(__dev_kfree_skb_irq);
2175

2176
void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason)
2177 2178
{
	if (in_irq() || irqs_disabled())
2179
		__dev_kfree_skb_irq(skb, reason);
2180 2181 2182
	else
		dev_kfree_skb(skb);
}
2183
EXPORT_SYMBOL(__dev_kfree_skb_any);
2184 2185


2186 2187 2188 2189 2190 2191
/**
 * netif_device_detach - mark device as removed
 * @dev: network device
 *
 * Mark device as removed from system and therefore no longer available.
 */
2192 2193 2194 2195
void netif_device_detach(struct net_device *dev)
{
	if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2196
		netif_tx_stop_all_queues(dev);
2197 2198 2199 2200
	}
}
EXPORT_SYMBOL(netif_device_detach);

2201 2202 2203 2204 2205 2206
/**
 * netif_device_attach - mark device as attached
 * @dev: network device
 *
 * Mark device as attached from system and restart if needed.
 */
2207 2208 2209 2210
void netif_device_attach(struct net_device *dev)
{
	if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2211
		netif_tx_wake_all_queues(dev);
2212
		__netdev_watchdog_up(dev);
2213 2214 2215 2216
	}
}
EXPORT_SYMBOL(netif_device_attach);

2217 2218
static void skb_warn_bad_offload(const struct sk_buff *skb)
{
2219
	static const netdev_features_t null_features = 0;
2220 2221 2222
	struct net_device *dev = skb->dev;
	const char *driver = "";

2223 2224 2225
	if (!net_ratelimit())
		return;

2226 2227 2228 2229 2230
	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",
2231 2232
	     driver, dev ? &dev->features : &null_features,
	     skb->sk ? &skb->sk->sk_route_caps : &null_features,
2233 2234 2235 2236
	     skb->len, skb->data_len, skb_shinfo(skb)->gso_size,
	     skb_shinfo(skb)->gso_type, skb->ip_summed);
}

L
Linus Torvalds 已提交
2237 2238 2239 2240
/*
 * Invalidate hardware checksum when packet is to be mangled, and
 * complete checksum manually on outgoing path.
 */
2241
int skb_checksum_help(struct sk_buff *skb)
L
Linus Torvalds 已提交
2242
{
2243
	__wsum csum;
2244
	int ret = 0, offset;
L
Linus Torvalds 已提交
2245

2246
	if (skb->ip_summed == CHECKSUM_COMPLETE)
2247 2248 2249
		goto out_set_summed;

	if (unlikely(skb_shinfo(skb)->gso_size)) {
2250 2251
		skb_warn_bad_offload(skb);
		return -EINVAL;
L
Linus Torvalds 已提交
2252 2253
	}

2254 2255 2256 2257 2258 2259 2260 2261 2262
	/* 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;
	}

2263
	offset = skb_checksum_start_offset(skb);
2264 2265 2266 2267 2268 2269 2270 2271
	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
2272 2273 2274 2275 2276
		ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
		if (ret)
			goto out;
	}

2277
	*(__sum16 *)(skb->data + offset) = csum_fold(csum);
2278
out_set_summed:
L
Linus Torvalds 已提交
2279
	skb->ip_summed = CHECKSUM_NONE;
2280
out:
L
Linus Torvalds 已提交
2281 2282
	return ret;
}
E
Eric Dumazet 已提交
2283
EXPORT_SYMBOL(skb_checksum_help);
L
Linus Torvalds 已提交
2284

2285
__be16 skb_network_protocol(struct sk_buff *skb)
2286
{
A
Al Viro 已提交
2287
	__be16 type = skb->protocol;
2288
	int vlan_depth = ETH_HLEN;
2289

2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300
	/* 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
2301
	while (type == htons(ETH_P_8021Q) || type == htons(ETH_P_8021AD)) {
2302
		struct vlan_hdr *vh;
2303

2304
		if (unlikely(!pskb_may_pull(skb, vlan_depth + VLAN_HLEN)))
2305
			return 0;
2306

2307 2308 2309
		vh = (struct vlan_hdr *)(skb->data + vlan_depth);
		type = vh->h_vlan_encapsulated_proto;
		vlan_depth += VLAN_HLEN;
2310 2311
	}

2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329
	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);

2330 2331 2332
	__skb_pull(skb, skb->mac_len);

	rcu_read_lock();
2333
	list_for_each_entry_rcu(ptype, &offload_base, list) {
2334
		if (ptype->type == type && ptype->callbacks.gso_segment) {
2335
			if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) {
2336 2337
				int err;

2338
				err = ptype->callbacks.gso_send_check(skb);
2339 2340 2341
				segs = ERR_PTR(err);
				if (err || skb_gso_ok(skb, features))
					break;
2342 2343
				__skb_push(skb, (skb->data -
						 skb_network_header(skb)));
2344
			}
2345
			segs = ptype->callbacks.gso_segment(skb, features);
2346 2347 2348 2349 2350
			break;
		}
	}
	rcu_read_unlock();

2351
	__skb_push(skb, skb->data - skb_mac_header(skb));
2352

2353 2354
	return segs;
}
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 2391
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);
	}

2392
	SKB_GSO_CB(skb)->mac_offset = skb_headroom(skb);
2393 2394
	SKB_GSO_CB(skb)->encap_level = 0;

2395 2396 2397 2398 2399
	skb_reset_mac_header(skb);
	skb_reset_mac_len(skb);

	return skb_mac_gso_segment(skb, features);
}
2400
EXPORT_SYMBOL(__skb_gso_segment);
2401

2402 2403 2404 2405 2406
/* Take action when hardware reception checksum errors are detected. */
#ifdef CONFIG_BUG
void netdev_rx_csum_fault(struct net_device *dev)
{
	if (net_ratelimit()) {
2407
		pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>");
2408 2409 2410 2411 2412 2413
		dump_stack();
	}
}
EXPORT_SYMBOL(netdev_rx_csum_fault);
#endif

L
Linus Torvalds 已提交
2414 2415 2416 2417 2418
/* 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
2419
static int illegal_highdma(struct net_device *dev, struct sk_buff *skb)
L
Linus Torvalds 已提交
2420
{
2421
#ifdef CONFIG_HIGHMEM
L
Linus Torvalds 已提交
2422
	int i;
2423
	if (!(dev->features & NETIF_F_HIGHDMA)) {
2424 2425 2426
		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)))
2427
				return 1;
2428
		}
2429
	}
L
Linus Torvalds 已提交
2430

2431 2432
	if (PCI_DMA_BUS_IS_PHYS) {
		struct device *pdev = dev->dev.parent;
L
Linus Torvalds 已提交
2433

E
Eric Dumazet 已提交
2434 2435
		if (!pdev)
			return 0;
2436
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2437 2438
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			dma_addr_t addr = page_to_phys(skb_frag_page(frag));
2439 2440 2441 2442
			if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask)
				return 1;
		}
	}
2443
#endif
L
Linus Torvalds 已提交
2444 2445 2446
	return 0;
}

2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472
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;

	do {
		struct sk_buff *nskb = skb->next;

		skb->next = nskb->next;
		nskb->next = NULL;
		kfree_skb(nskb);
	} while (skb->next);

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

/**
 *	dev_gso_segment - Perform emulated hardware segmentation on skb.
 *	@skb: buffer to segment
2473
 *	@features: device features as applicable to this skb
2474 2475 2476 2477
 *
 *	This function segments the given skb and stores the list of segments
 *	in skb->next.
 */
2478
static int dev_gso_segment(struct sk_buff *skb, netdev_features_t features)
2479 2480
{
	struct sk_buff *segs;
2481 2482 2483 2484 2485 2486

	segs = skb_gso_segment(skb, features);

	/* Verifying header integrity only. */
	if (!segs)
		return 0;
2487

2488
	if (IS_ERR(segs))
2489 2490 2491 2492 2493 2494 2495 2496 2497
		return PTR_ERR(segs);

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

	return 0;
}

2498
static netdev_features_t harmonize_features(struct sk_buff *skb,
2499
	netdev_features_t features)
2500
{
2501
	if (skb->ip_summed != CHECKSUM_NONE &&
2502
	    !can_checksum_protocol(features, skb_network_protocol(skb))) {
2503 2504 2505 2506 2507 2508 2509 2510
		features &= ~NETIF_F_ALL_CSUM;
	} else if (illegal_highdma(skb->dev, skb)) {
		features &= ~NETIF_F_SG;
	}

	return features;
}

2511
netdev_features_t netif_skb_features(struct sk_buff *skb)
2512 2513
{
	__be16 protocol = skb->protocol;
2514
	netdev_features_t features = skb->dev->features;
2515

2516 2517 2518
	if (skb_shinfo(skb)->gso_segs > skb->dev->gso_max_segs)
		features &= ~NETIF_F_GSO_MASK;

P
Patrick McHardy 已提交
2519
	if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD)) {
2520 2521
		struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
		protocol = veh->h_vlan_encapsulated_proto;
2522
	} else if (!vlan_tx_tag_present(skb)) {
2523
		return harmonize_features(skb, features);
2524
	}
2525

P
Patrick McHardy 已提交
2526 2527
	features &= (skb->dev->vlan_features | NETIF_F_HW_VLAN_CTAG_TX |
					       NETIF_F_HW_VLAN_STAG_TX);
2528

2529
	if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD))
2530
		features &= NETIF_F_SG | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST |
P
Patrick McHardy 已提交
2531 2532
				NETIF_F_GEN_CSUM | NETIF_F_HW_VLAN_CTAG_TX |
				NETIF_F_HW_VLAN_STAG_TX;
2533 2534

	return harmonize_features(skb, features);
2535
}
2536
EXPORT_SYMBOL(netif_skb_features);
2537

2538
int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2539
			struct netdev_queue *txq, void *accel_priv)
2540
{
2541
	const struct net_device_ops *ops = dev->netdev_ops;
2542
	int rc = NETDEV_TX_OK;
2543
	unsigned int skb_len;
2544

2545
	if (likely(!skb->next)) {
2546
		netdev_features_t features;
2547

2548
		/*
L
Lucas De Marchi 已提交
2549
		 * If device doesn't need skb->dst, release it right now while
2550 2551
		 * its hot in this cpu cache
		 */
E
Eric Dumazet 已提交
2552 2553 2554
		if (dev->priv_flags & IFF_XMIT_DST_RELEASE)
			skb_dst_drop(skb);

2555 2556
		features = netif_skb_features(skb);

2557
		if (vlan_tx_tag_present(skb) &&
2558 2559 2560
		    !vlan_hw_offload_capable(features, skb->vlan_proto)) {
			skb = __vlan_put_tag(skb, skb->vlan_proto,
					     vlan_tx_tag_get(skb));
2561 2562 2563 2564 2565 2566
			if (unlikely(!skb))
				goto out;

			skb->vlan_tci = 0;
		}

2567 2568 2569 2570 2571 2572 2573
		/* 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;

2574
		if (netif_needs_gso(skb, features)) {
2575
			if (unlikely(dev_gso_segment(skb, features)))
2576 2577 2578
				goto out_kfree_skb;
			if (skb->next)
				goto gso;
2579
		} else {
2580
			if (skb_needs_linearize(skb, features) &&
2581 2582 2583 2584 2585 2586 2587 2588
			    __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) {
2589 2590 2591 2592 2593 2594
				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));
2595
				if (!(features & NETIF_F_ALL_CSUM) &&
2596 2597 2598
				     skb_checksum_help(skb))
					goto out_kfree_skb;
			}
2599 2600
		}

2601 2602 2603
		if (!list_empty(&ptype_all))
			dev_queue_xmit_nit(skb, dev);

2604
		skb_len = skb->len;
2605 2606 2607 2608 2609
		if (accel_priv)
			rc = ops->ndo_dfwd_start_xmit(skb, dev, accel_priv);
		else
			rc = ops->ndo_start_xmit(skb, dev);

2610
		trace_net_dev_xmit(skb, rc, dev, skb_len);
2611
		if (rc == NETDEV_TX_OK && txq)
E
Eric Dumazet 已提交
2612
			txq_trans_update(txq);
2613
		return rc;
2614 2615
	}

2616
gso:
2617 2618 2619 2620 2621
	do {
		struct sk_buff *nskb = skb->next;

		skb->next = nskb->next;
		nskb->next = NULL;
2622

2623 2624 2625
		if (!list_empty(&ptype_all))
			dev_queue_xmit_nit(nskb, dev);

2626
		skb_len = nskb->len;
2627 2628 2629 2630
		if (accel_priv)
			rc = ops->ndo_dfwd_start_xmit(nskb, dev, accel_priv);
		else
			rc = ops->ndo_start_xmit(nskb, dev);
2631
		trace_net_dev_xmit(nskb, rc, dev, skb_len);
2632
		if (unlikely(rc != NETDEV_TX_OK)) {
2633 2634
			if (rc & ~NETDEV_TX_MASK)
				goto out_kfree_gso_skb;
2635
			nskb->next = skb->next;
2636 2637 2638
			skb->next = nskb;
			return rc;
		}
E
Eric Dumazet 已提交
2639
		txq_trans_update(txq);
2640
		if (unlikely(netif_xmit_stopped(txq) && skb->next))
2641
			return NETDEV_TX_BUSY;
2642
	} while (skb->next);
2643

2644
out_kfree_gso_skb:
2645
	if (likely(skb->next == NULL)) {
2646
		skb->destructor = DEV_GSO_CB(skb)->destructor;
2647 2648 2649
		consume_skb(skb);
		return rc;
	}
2650 2651
out_kfree_skb:
	kfree_skb(skb);
2652
out:
2653
	return rc;
2654
}
2655
EXPORT_SYMBOL_GPL(dev_hard_start_xmit);
2656

2657 2658 2659 2660 2661 2662 2663 2664 2665 2666
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)  {
2667
		unsigned int hdr_len;
2668
		u16 gso_segs = shinfo->gso_segs;
2669

2670 2671 2672 2673
		/* mac layer + network layer */
		hdr_len = skb_transport_header(skb) - skb_mac_header(skb);

		/* + transport layer */
2674 2675 2676 2677
		if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)))
			hdr_len += tcp_hdrlen(skb);
		else
			hdr_len += sizeof(struct udphdr);
2678 2679 2680 2681 2682 2683

		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;
2684 2685 2686
	}
}

2687 2688 2689 2690 2691
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
2692
	bool contended;
2693 2694
	int rc;

2695
	qdisc_pkt_len_init(skb);
E
Eric Dumazet 已提交
2696
	qdisc_calculate_pkt_len(skb, q);
2697 2698 2699 2700 2701 2702
	/*
	 * 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
2703
	contended = qdisc_is_running(q);
2704 2705 2706
	if (unlikely(contended))
		spin_lock(&q->busylock);

2707 2708 2709 2710 2711
	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) &&
2712
		   qdisc_run_begin(q)) {
2713 2714 2715 2716 2717
		/*
		 * 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
2718 2719
		if (!(dev->priv_flags & IFF_XMIT_DST_RELEASE))
			skb_dst_force(skb);
2720 2721 2722

		qdisc_bstats_update(q, skb);

2723 2724 2725 2726 2727
		if (sch_direct_xmit(skb, q, dev, txq, root_lock)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
2728
			__qdisc_run(q);
2729
		} else
2730
			qdisc_run_end(q);
2731 2732 2733

		rc = NET_XMIT_SUCCESS;
	} else {
E
Eric Dumazet 已提交
2734
		skb_dst_force(skb);
E
Eric Dumazet 已提交
2735
		rc = q->enqueue(skb, q) & NET_XMIT_MASK;
2736 2737 2738 2739 2740 2741 2742
		if (qdisc_run_begin(q)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
			__qdisc_run(q);
		}
2743 2744
	}
	spin_unlock(root_lock);
2745 2746
	if (unlikely(contended))
		spin_unlock(&q->busylock);
2747 2748 2749
	return rc;
}

2750 2751 2752
#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
static void skb_update_prio(struct sk_buff *skb)
{
I
Igor Maravic 已提交
2753
	struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap);
2754

2755 2756 2757 2758 2759 2760
	if (!skb->priority && skb->sk && map) {
		unsigned int prioidx = skb->sk->sk_cgrp_prioidx;

		if (prioidx < map->priomap_len)
			skb->priority = map->priomap[prioidx];
	}
2761 2762 2763 2764 2765
}
#else
#define skb_update_prio(skb)
#endif

2766
static DEFINE_PER_CPU(int, xmit_recursion);
2767
#define RECURSION_LIMIT 10
2768

2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785
/**
 *	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);

2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810
/**
 *	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
2811 2812 2813
int dev_queue_xmit(struct sk_buff *skb)
{
	struct net_device *dev = skb->dev;
2814
	struct netdev_queue *txq;
L
Linus Torvalds 已提交
2815 2816 2817
	struct Qdisc *q;
	int rc = -ENOMEM;

2818 2819
	skb_reset_mac_header(skb);

2820 2821
	/* Disable soft irqs for various locks below. Also
	 * stops preemption for RCU.
L
Linus Torvalds 已提交
2822
	 */
2823
	rcu_read_lock_bh();
L
Linus Torvalds 已提交
2824

2825 2826
	skb_update_prio(skb);

2827
	txq = netdev_pick_tx(dev, skb);
2828
	q = rcu_dereference_bh(txq->qdisc);
2829

L
Linus Torvalds 已提交
2830
#ifdef CONFIG_NET_CLS_ACT
E
Eric Dumazet 已提交
2831
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS);
L
Linus Torvalds 已提交
2832
#endif
2833
	trace_net_dev_queue(skb);
L
Linus Torvalds 已提交
2834
	if (q->enqueue) {
2835
		rc = __dev_xmit_skb(skb, q, dev, txq);
2836
		goto out;
L
Linus Torvalds 已提交
2837 2838 2839 2840 2841
	}

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

H
Herbert Xu 已提交
2842 2843
	   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
2844 2845 2846 2847 2848 2849 2850 2851 2852 2853
	   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 */

2854
		if (txq->xmit_lock_owner != cpu) {
L
Linus Torvalds 已提交
2855

2856 2857 2858
			if (__this_cpu_read(xmit_recursion) > RECURSION_LIMIT)
				goto recursion_alert;

2859
			HARD_TX_LOCK(dev, txq, cpu);
L
Linus Torvalds 已提交
2860

2861
			if (!netif_xmit_stopped(txq)) {
2862
				__this_cpu_inc(xmit_recursion);
2863
				rc = dev_hard_start_xmit(skb, dev, txq, NULL);
2864
				__this_cpu_dec(xmit_recursion);
2865
				if (dev_xmit_complete(rc)) {
2866
					HARD_TX_UNLOCK(dev, txq);
L
Linus Torvalds 已提交
2867 2868 2869
					goto out;
				}
			}
2870
			HARD_TX_UNLOCK(dev, txq);
2871 2872
			net_crit_ratelimited("Virtual device %s asks to queue packet!\n",
					     dev->name);
L
Linus Torvalds 已提交
2873 2874
		} else {
			/* Recursion is detected! It is possible,
2875 2876 2877
			 * unfortunately
			 */
recursion_alert:
2878 2879
			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
					     dev->name);
L
Linus Torvalds 已提交
2880 2881 2882 2883
		}
	}

	rc = -ENETDOWN;
2884
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
2885 2886 2887 2888

	kfree_skb(skb);
	return rc;
out:
2889
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
2890 2891
	return rc;
}
E
Eric Dumazet 已提交
2892
EXPORT_SYMBOL(dev_queue_xmit);
L
Linus Torvalds 已提交
2893 2894 2895 2896 2897 2898


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

2899
int netdev_max_backlog __read_mostly = 1000;
E
Eric Dumazet 已提交
2900 2901
EXPORT_SYMBOL(netdev_max_backlog);

E
Eric Dumazet 已提交
2902
int netdev_tstamp_prequeue __read_mostly = 1;
2903 2904
int netdev_budget __read_mostly = 300;
int weight_p __read_mostly = 64;            /* old backlog weight */
L
Linus Torvalds 已提交
2905

E
Eric Dumazet 已提交
2906 2907 2908 2909 2910 2911 2912 2913
/* 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);
}

2914 2915 2916
#ifdef CONFIG_RPS

/* One global table that all flow-based protocols share. */
E
Eric Dumazet 已提交
2917
struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
2918 2919
EXPORT_SYMBOL(rps_sock_flow_table);

2920
struct static_key rps_needed __read_mostly;
2921

2922 2923 2924 2925
static struct rps_dev_flow *
set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
	    struct rps_dev_flow *rflow, u16 next_cpu)
{
2926
	if (next_cpu != RPS_NO_CPU) {
2927 2928 2929 2930 2931 2932 2933 2934 2935
#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? */
2936 2937
		if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
		    !(dev->features & NETIF_F_NTUPLE))
2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959
			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 =
2960
			per_cpu(softnet_data, next_cpu).input_queue_head;
2961 2962
	}

2963
	rflow->cpu = next_cpu;
2964 2965 2966
	return rflow;
}

2967 2968 2969 2970 2971 2972 2973 2974 2975
/*
 * 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
2976
	struct rps_map *map;
2977 2978 2979 2980 2981 2982 2983
	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);
2984 2985 2986 2987 2988
		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);
2989 2990 2991 2992 2993 2994
			goto done;
		}
		rxqueue = dev->_rx + index;
	} else
		rxqueue = dev->_rx;

E
Eric Dumazet 已提交
2995 2996
	map = rcu_dereference(rxqueue->rps_map);
	if (map) {
2997
		if (map->len == 1 &&
2998
		    !rcu_access_pointer(rxqueue->rps_flow_table)) {
2999 3000 3001 3002 3003
			tcpu = map->cpus[0];
			if (cpu_online(tcpu))
				cpu = tcpu;
			goto done;
		}
3004
	} else if (!rcu_access_pointer(rxqueue->rps_flow_table)) {
3005
		goto done;
3006
	}
3007

3008
	skb_reset_network_header(skb);
3009 3010 3011
	if (!skb_get_rxhash(skb))
		goto done;

T
Tom Herbert 已提交
3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037
	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 -
3038 3039
		      rflow->last_qtail)) >= 0)) {
			tcpu = next_cpu;
3040
			rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
3041
		}
3042

T
Tom Herbert 已提交
3043 3044 3045 3046 3047 3048 3049
		if (tcpu != RPS_NO_CPU && cpu_online(tcpu)) {
			*rflowp = rflow;
			cpu = tcpu;
			goto done;
		}
	}

T
Tom Herbert 已提交
3050
	if (map) {
T
Tom Herbert 已提交
3051
		tcpu = map->cpus[((u64) skb->rxhash * map->len) >> 32];
T
Tom Herbert 已提交
3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062

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

done:
	return cpu;
}

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 3097 3098 3099 3100 3101 3102
#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
3103
/* Called from hardirq (IPI) context */
E
Eric Dumazet 已提交
3104
static void rps_trigger_softirq(void *data)
T
Tom Herbert 已提交
3105
{
E
Eric Dumazet 已提交
3106 3107
	struct softnet_data *sd = data;

E
Eric Dumazet 已提交
3108
	____napi_schedule(sd, &sd->backlog);
C
Changli Gao 已提交
3109
	sd->received_rps++;
T
Tom Herbert 已提交
3110
}
E
Eric Dumazet 已提交
3111

T
Tom Herbert 已提交
3112
#endif /* CONFIG_RPS */
T
Tom Herbert 已提交
3113

E
Eric Dumazet 已提交
3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134
/*
 * 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;
}

3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174
#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) {
		new_flow = skb_get_rxhash(skb) & (fl->num_buckets - 1);
		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
3175 3176 3177 3178
/*
 * 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
3179 3180
static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
			      unsigned int *qtail)
T
Tom Herbert 已提交
3181
{
E
Eric Dumazet 已提交
3182
	struct softnet_data *sd;
T
Tom Herbert 已提交
3183
	unsigned long flags;
3184
	unsigned int qlen;
T
Tom Herbert 已提交
3185

E
Eric Dumazet 已提交
3186
	sd = &per_cpu(softnet_data, cpu);
T
Tom Herbert 已提交
3187 3188 3189

	local_irq_save(flags);

E
Eric Dumazet 已提交
3190
	rps_lock(sd);
3191 3192
	qlen = skb_queue_len(&sd->input_pkt_queue);
	if (qlen <= netdev_max_backlog && !skb_flow_limit(skb, qlen)) {
3193
		if (skb_queue_len(&sd->input_pkt_queue)) {
T
Tom Herbert 已提交
3194
enqueue:
E
Eric Dumazet 已提交
3195
			__skb_queue_tail(&sd->input_pkt_queue, skb);
3196
			input_queue_tail_incr_save(sd, qtail);
E
Eric Dumazet 已提交
3197
			rps_unlock(sd);
3198
			local_irq_restore(flags);
T
Tom Herbert 已提交
3199 3200 3201
			return NET_RX_SUCCESS;
		}

3202 3203 3204 3205
		/* 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
3206
			if (!rps_ipi_queued(sd))
E
Eric Dumazet 已提交
3207
				____napi_schedule(sd, &sd->backlog);
T
Tom Herbert 已提交
3208 3209 3210 3211
		}
		goto enqueue;
	}

C
Changli Gao 已提交
3212
	sd->dropped++;
E
Eric Dumazet 已提交
3213
	rps_unlock(sd);
T
Tom Herbert 已提交
3214 3215 3216

	local_irq_restore(flags);

3217
	atomic_long_inc(&skb->dev->rx_dropped);
T
Tom Herbert 已提交
3218 3219 3220
	kfree_skb(skb);
	return NET_RX_DROP;
}
L
Linus Torvalds 已提交
3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238

/**
 *	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)
{
3239
	int ret;
L
Linus Torvalds 已提交
3240 3241 3242 3243 3244

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

3245
	net_timestamp_check(netdev_tstamp_prequeue, skb);
L
Linus Torvalds 已提交
3246

3247
	trace_netif_rx(skb);
E
Eric Dumazet 已提交
3248
#ifdef CONFIG_RPS
3249
	if (static_key_false(&rps_needed)) {
T
Tom Herbert 已提交
3250
		struct rps_dev_flow voidflow, *rflow = &voidflow;
3251 3252
		int cpu;

3253
		preempt_disable();
3254
		rcu_read_lock();
T
Tom Herbert 已提交
3255 3256

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
3257 3258
		if (cpu < 0)
			cpu = smp_processor_id();
T
Tom Herbert 已提交
3259 3260 3261

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

3262
		rcu_read_unlock();
3263
		preempt_enable();
3264 3265
	} else
#endif
T
Tom Herbert 已提交
3266 3267 3268 3269 3270
	{
		unsigned int qtail;
		ret = enqueue_to_backlog(skb, get_cpu(), &qtail);
		put_cpu();
	}
3271
	return ret;
L
Linus Torvalds 已提交
3272
}
E
Eric Dumazet 已提交
3273
EXPORT_SYMBOL(netif_rx);
L
Linus Torvalds 已提交
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 3299 3300 3301 3302 3303 3304

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;

3305
			WARN_ON(atomic_read(&skb->users));
3306 3307 3308 3309
			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
3310 3311 3312 3313 3314
			__kfree_skb(skb);
		}
	}

	if (sd->output_queue) {
3315
		struct Qdisc *head;
L
Linus Torvalds 已提交
3316 3317 3318 3319

		local_irq_disable();
		head = sd->output_queue;
		sd->output_queue = NULL;
3320
		sd->output_queue_tailp = &sd->output_queue;
L
Linus Torvalds 已提交
3321 3322 3323
		local_irq_enable();

		while (head) {
3324 3325 3326
			struct Qdisc *q = head;
			spinlock_t *root_lock;

L
Linus Torvalds 已提交
3327 3328
			head = head->next_sched;

3329
			root_lock = qdisc_lock(q);
3330
			if (spin_trylock(root_lock)) {
3331 3332 3333
				smp_mb__before_clear_bit();
				clear_bit(__QDISC_STATE_SCHED,
					  &q->state);
3334 3335
				qdisc_run(q);
				spin_unlock(root_lock);
L
Linus Torvalds 已提交
3336
			} else {
3337
				if (!test_bit(__QDISC_STATE_DEACTIVATED,
3338
					      &q->state)) {
3339
					__netif_reschedule(q);
3340 3341 3342 3343 3344
				} else {
					smp_mb__before_clear_bit();
					clear_bit(__QDISC_STATE_SCHED,
						  &q->state);
				}
L
Linus Torvalds 已提交
3345 3346 3347 3348 3349
			}
		}
	}
}

3350 3351
#if (defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)) && \
    (defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE))
3352 3353 3354
/* This hook is defined here for ATM LANE */
int (*br_fdb_test_addr_hook)(struct net_device *dev,
			     unsigned char *addr) __read_mostly;
3355
EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
3356
#endif
L
Linus Torvalds 已提交
3357 3358 3359 3360 3361 3362

#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
3363 3364
 * 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
3365 3366
 *
 */
3367
static int ing_filter(struct sk_buff *skb, struct netdev_queue *rxq)
L
Linus Torvalds 已提交
3368 3369
{
	struct net_device *dev = skb->dev;
3370
	u32 ttl = G_TC_RTTL(skb->tc_verd);
3371 3372
	int result = TC_ACT_OK;
	struct Qdisc *q;
3373

3374
	if (unlikely(MAX_RED_LOOP < ttl++)) {
3375 3376
		net_warn_ratelimited("Redir loop detected Dropping packet (%d->%d)\n",
				     skb->skb_iif, dev->ifindex);
3377 3378
		return TC_ACT_SHOT;
	}
L
Linus Torvalds 已提交
3379

3380 3381
	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
3382

3383
	q = rxq->qdisc;
3384
	if (q != &noop_qdisc) {
3385
		spin_lock(qdisc_lock(q));
3386 3387
		if (likely(!test_bit(__QDISC_STATE_DEACTIVATED, &q->state)))
			result = qdisc_enqueue_root(skb, q);
3388 3389
		spin_unlock(qdisc_lock(q));
	}
3390 3391 3392

	return result;
}
3393

3394 3395 3396 3397
static inline struct sk_buff *handle_ing(struct sk_buff *skb,
					 struct packet_type **pt_prev,
					 int *ret, struct net_device *orig_dev)
{
3398 3399 3400
	struct netdev_queue *rxq = rcu_dereference(skb->dev->ingress_queue);

	if (!rxq || rxq->qdisc == &noop_qdisc)
3401
		goto out;
L
Linus Torvalds 已提交
3402

3403 3404 3405
	if (*pt_prev) {
		*ret = deliver_skb(skb, *pt_prev, orig_dev);
		*pt_prev = NULL;
L
Linus Torvalds 已提交
3406 3407
	}

3408
	switch (ing_filter(skb, rxq)) {
3409 3410 3411 3412 3413 3414 3415 3416 3417
	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
3418 3419 3420
}
#endif

3421 3422 3423 3424
/**
 *	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
3425
 *	@rx_handler_data: data pointer that is used by rx handler
3426 3427 3428 3429 3430 3431
 *
 *	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.
3432 3433
 *
 *	For a general description of rx_handler, see enum rx_handler_result.
3434 3435
 */
int netdev_rx_handler_register(struct net_device *dev,
J
Jiri Pirko 已提交
3436 3437
			       rx_handler_func_t *rx_handler,
			       void *rx_handler_data)
3438 3439 3440 3441 3442 3443
{
	ASSERT_RTNL();

	if (dev->rx_handler)
		return -EBUSY;

3444
	/* Note: rx_handler_data must be set before rx_handler */
J
Jiri Pirko 已提交
3445
	rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
3446 3447 3448 3449 3450 3451 3452 3453 3454 3455
	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
3456
 *	Unregister a receive handler from a device.
3457 3458 3459 3460 3461 3462 3463
 *
 *	The caller must hold the rtnl_mutex.
 */
void netdev_rx_handler_unregister(struct net_device *dev)
{

	ASSERT_RTNL();
3464
	RCU_INIT_POINTER(dev->rx_handler, NULL);
3465 3466 3467 3468 3469
	/* 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();
3470
	RCU_INIT_POINTER(dev->rx_handler_data, NULL);
3471 3472 3473
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);

3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484
/*
 * 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
3485
	case __constant_htons(ETH_P_8021AD):
3486 3487 3488 3489 3490 3491
		return true;
	default:
		return false;
	}
}

3492
static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc)
L
Linus Torvalds 已提交
3493 3494
{
	struct packet_type *ptype, *pt_prev;
3495
	rx_handler_func_t *rx_handler;
D
David S. Miller 已提交
3496
	struct net_device *orig_dev;
3497
	struct net_device *null_or_dev;
3498
	bool deliver_exact = false;
L
Linus Torvalds 已提交
3499
	int ret = NET_RX_DROP;
A
Al Viro 已提交
3500
	__be16 type;
L
Linus Torvalds 已提交
3501

3502
	net_timestamp_check(!netdev_tstamp_prequeue, skb);
3503

3504
	trace_netif_receive_skb(skb);
3505

L
Linus Torvalds 已提交
3506
	/* if we've gotten here through NAPI, check netpoll */
3507
	if (netpoll_receive_skb(skb))
3508
		goto out;
L
Linus Torvalds 已提交
3509

J
Joe Eykholt 已提交
3510
	orig_dev = skb->dev;
3511

3512
	skb_reset_network_header(skb);
3513 3514
	if (!skb_transport_header_was_set(skb))
		skb_reset_transport_header(skb);
3515
	skb_reset_mac_len(skb);
L
Linus Torvalds 已提交
3516 3517 3518 3519 3520

	pt_prev = NULL;

	rcu_read_lock();

3521
another_round:
3522
	skb->skb_iif = skb->dev->ifindex;
3523 3524 3525

	__this_cpu_inc(softnet_data.processed);

P
Patrick McHardy 已提交
3526 3527
	if (skb->protocol == cpu_to_be16(ETH_P_8021Q) ||
	    skb->protocol == cpu_to_be16(ETH_P_8021AD)) {
3528 3529
		skb = vlan_untag(skb);
		if (unlikely(!skb))
3530
			goto unlock;
3531 3532
	}

L
Linus Torvalds 已提交
3533 3534 3535 3536 3537 3538 3539
#ifdef CONFIG_NET_CLS_ACT
	if (skb->tc_verd & TC_NCLS) {
		skb->tc_verd = CLR_TC_NCLS(skb->tc_verd);
		goto ncls;
	}
#endif

3540
	if (pfmemalloc)
3541 3542
		goto skip_taps;

L
Linus Torvalds 已提交
3543
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
3544
		if (!ptype->dev || ptype->dev == skb->dev) {
3545
			if (pt_prev)
D
David S. Miller 已提交
3546
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3547 3548 3549 3550
			pt_prev = ptype;
		}
	}

3551
skip_taps:
L
Linus Torvalds 已提交
3552
#ifdef CONFIG_NET_CLS_ACT
3553 3554
	skb = handle_ing(skb, &pt_prev, &ret, orig_dev);
	if (!skb)
3555
		goto unlock;
L
Linus Torvalds 已提交
3556 3557 3558
ncls:
#endif

3559
	if (pfmemalloc && !skb_pfmemalloc_protocol(skb))
3560 3561
		goto drop;

3562 3563 3564 3565 3566
	if (vlan_tx_tag_present(skb)) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3567
		if (vlan_do_receive(&skb))
3568 3569
			goto another_round;
		else if (unlikely(!skb))
3570
			goto unlock;
3571 3572
	}

3573
	rx_handler = rcu_dereference(skb->dev->rx_handler);
3574 3575 3576 3577 3578
	if (rx_handler) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3579 3580
		switch (rx_handler(&skb)) {
		case RX_HANDLER_CONSUMED:
3581
			ret = NET_RX_SUCCESS;
3582
			goto unlock;
3583
		case RX_HANDLER_ANOTHER:
3584
			goto another_round;
3585 3586 3587 3588 3589 3590 3591
		case RX_HANDLER_EXACT:
			deliver_exact = true;
		case RX_HANDLER_PASS:
			break;
		default:
			BUG();
		}
3592
	}
L
Linus Torvalds 已提交
3593

E
Eric Dumazet 已提交
3594 3595 3596 3597 3598 3599 3600 3601 3602
	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;
	}
3603

3604
	/* deliver only exact match when indicated */
3605
	null_or_dev = deliver_exact ? skb->dev : NULL;
3606

L
Linus Torvalds 已提交
3607
	type = skb->protocol;
3608 3609
	list_for_each_entry_rcu(ptype,
			&ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) {
3610
		if (ptype->type == type &&
3611 3612
		    (ptype->dev == null_or_dev || ptype->dev == skb->dev ||
		     ptype->dev == orig_dev)) {
3613
			if (pt_prev)
D
David S. Miller 已提交
3614
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3615 3616 3617 3618 3619
			pt_prev = ptype;
		}
	}

	if (pt_prev) {
3620
		if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
3621
			goto drop;
3622 3623
		else
			ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3624
	} else {
3625
drop:
3626
		atomic_long_inc(&skb->dev->rx_dropped);
L
Linus Torvalds 已提交
3627 3628 3629 3630 3631 3632 3633
		kfree_skb(skb);
		/* Jamal, now you will not able to escape explaining
		 * me how you were going to use this. :-)
		 */
		ret = NET_RX_DROP;
	}

3634
unlock:
L
Linus Torvalds 已提交
3635
	rcu_read_unlock();
3636
out:
3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661
	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
3662 3663
	return ret;
}
T
Tom Herbert 已提交
3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681

/**
 *	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)
{
3682
	net_timestamp_check(netdev_tstamp_prequeue, skb);
E
Eric Dumazet 已提交
3683

3684 3685 3686
	if (skb_defer_rx_timestamp(skb))
		return NET_RX_SUCCESS;

E
Eric Dumazet 已提交
3687
#ifdef CONFIG_RPS
3688
	if (static_key_false(&rps_needed)) {
E
Eric Dumazet 已提交
3689 3690
		struct rps_dev_flow voidflow, *rflow = &voidflow;
		int cpu, ret;
T
Tom Herbert 已提交
3691

E
Eric Dumazet 已提交
3692 3693 3694
		rcu_read_lock();

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
T
Tom Herbert 已提交
3695

E
Eric Dumazet 已提交
3696 3697 3698
		if (cpu >= 0) {
			ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
			rcu_read_unlock();
3699
			return ret;
E
Eric Dumazet 已提交
3700
		}
3701
		rcu_read_unlock();
T
Tom Herbert 已提交
3702
	}
3703
#endif
3704
	return __netif_receive_skb(skb);
T
Tom Herbert 已提交
3705
}
E
Eric Dumazet 已提交
3706
EXPORT_SYMBOL(netif_receive_skb);
L
Linus Torvalds 已提交
3707

E
Eric Dumazet 已提交
3708 3709 3710
/* Network device is going away, flush any packets still pending
 * Called with irqs disabled.
 */
3711
static void flush_backlog(void *arg)
3712
{
3713
	struct net_device *dev = arg;
E
Eric Dumazet 已提交
3714
	struct softnet_data *sd = &__get_cpu_var(softnet_data);
3715 3716
	struct sk_buff *skb, *tmp;

E
Eric Dumazet 已提交
3717
	rps_lock(sd);
3718
	skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
3719
		if (skb->dev == dev) {
E
Eric Dumazet 已提交
3720
			__skb_unlink(skb, &sd->input_pkt_queue);
3721
			kfree_skb(skb);
3722
			input_queue_head_incr(sd);
3723
		}
3724
	}
E
Eric Dumazet 已提交
3725
	rps_unlock(sd);
3726 3727 3728 3729 3730

	skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
		if (skb->dev == dev) {
			__skb_unlink(skb, &sd->process_queue);
			kfree_skb(skb);
3731
			input_queue_head_incr(sd);
3732 3733
		}
	}
3734 3735
}

3736 3737
static int napi_gro_complete(struct sk_buff *skb)
{
3738
	struct packet_offload *ptype;
3739
	__be16 type = skb->protocol;
3740
	struct list_head *head = &offload_base;
3741 3742
	int err = -ENOENT;

3743 3744
	BUILD_BUG_ON(sizeof(struct napi_gro_cb) > sizeof(skb->cb));

3745 3746
	if (NAPI_GRO_CB(skb)->count == 1) {
		skb_shinfo(skb)->gso_size = 0;
3747
		goto out;
3748
	}
3749 3750 3751

	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
3752
		if (ptype->type != type || !ptype->callbacks.gro_complete)
3753 3754
			continue;

3755
		err = ptype->callbacks.gro_complete(skb, 0);
3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769
		break;
	}
	rcu_read_unlock();

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

out:
	return netif_receive_skb(skb);
}

3770 3771 3772 3773 3774
/* 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)
3775
{
3776
	struct sk_buff *skb, *prev = NULL;
3777

3778 3779 3780 3781 3782 3783 3784
	/* 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) {
3785
		skb->next = NULL;
3786 3787 3788 3789 3790

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

		prev = skb->prev;
3791
		napi_gro_complete(skb);
3792
		napi->gro_count--;
3793 3794 3795 3796
	}

	napi->gro_list = NULL;
}
E
Eric Dumazet 已提交
3797
EXPORT_SYMBOL(napi_gro_flush);
3798

E
Eric Dumazet 已提交
3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820
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;
	}
}

3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837
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
3838
static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3839 3840
{
	struct sk_buff **pp = NULL;
3841
	struct packet_offload *ptype;
3842
	__be16 type = skb->protocol;
3843
	struct list_head *head = &offload_base;
H
Herbert Xu 已提交
3844
	int same_flow;
3845
	enum gro_result ret;
3846

3847
	if (!(skb->dev->features & NETIF_F_GRO) || netpoll_rx_on(skb))
3848 3849
		goto normal;

3850
	if (skb_is_gso(skb) || skb_has_frag_list(skb))
3851 3852
		goto normal;

3853
	skb_gro_reset_offset(skb);
E
Eric Dumazet 已提交
3854 3855
	gro_list_prepare(napi, skb);

3856 3857
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
3858
		if (ptype->type != type || !ptype->callbacks.gro_receive)
3859 3860
			continue;

3861
		skb_set_network_header(skb, skb_gro_offset(skb));
3862
		skb_reset_mac_len(skb);
3863 3864
		NAPI_GRO_CB(skb)->same_flow = 0;
		NAPI_GRO_CB(skb)->flush = 0;
H
Herbert Xu 已提交
3865
		NAPI_GRO_CB(skb)->free = 0;
3866

3867
		pp = ptype->callbacks.gro_receive(&napi->gro_list, skb);
3868 3869 3870 3871 3872 3873 3874
		break;
	}
	rcu_read_unlock();

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

H
Herbert Xu 已提交
3875
	same_flow = NAPI_GRO_CB(skb)->same_flow;
3876
	ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED;
H
Herbert Xu 已提交
3877

3878 3879 3880 3881 3882 3883
	if (pp) {
		struct sk_buff *nskb = *pp;

		*pp = nskb->next;
		nskb->next = NULL;
		napi_gro_complete(nskb);
3884
		napi->gro_count--;
3885 3886
	}

H
Herbert Xu 已提交
3887
	if (same_flow)
3888 3889
		goto ok;

3890
	if (NAPI_GRO_CB(skb)->flush || napi->gro_count >= MAX_GRO_SKBS)
3891 3892
		goto normal;

3893
	napi->gro_count++;
3894
	NAPI_GRO_CB(skb)->count = 1;
3895
	NAPI_GRO_CB(skb)->age = jiffies;
3896
	skb_shinfo(skb)->gso_size = skb_gro_len(skb);
3897 3898
	skb->next = napi->gro_list;
	napi->gro_list = skb;
3899
	ret = GRO_HELD;
3900

3901
pull:
H
Herbert Xu 已提交
3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912
	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
3913
		skb_frag_size_sub(&skb_shinfo(skb)->frags[0], grow);
H
Herbert Xu 已提交
3914

E
Eric Dumazet 已提交
3915
		if (unlikely(!skb_frag_size(&skb_shinfo(skb)->frags[0]))) {
3916
			skb_frag_unref(skb, 0);
H
Herbert Xu 已提交
3917 3918
			memmove(skb_shinfo(skb)->frags,
				skb_shinfo(skb)->frags + 1,
3919
				--skb_shinfo(skb)->nr_frags * sizeof(skb_frag_t));
H
Herbert Xu 已提交
3920
		}
3921 3922
	}

3923
ok:
3924
	return ret;
3925 3926

normal:
3927 3928
	ret = GRO_NORMAL;
	goto pull;
H
Herbert Xu 已提交
3929
}
3930

H
Herbert Xu 已提交
3931

R
Rami Rosen 已提交
3932
static gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
H
Herbert Xu 已提交
3933
{
3934 3935
	switch (ret) {
	case GRO_NORMAL:
3936 3937 3938
		if (netif_receive_skb(skb))
			ret = GRO_DROP;
		break;
H
Herbert Xu 已提交
3939

3940
	case GRO_DROP:
H
Herbert Xu 已提交
3941 3942
		kfree_skb(skb);
		break;
3943

3944
	case GRO_MERGED_FREE:
3945 3946 3947 3948
		if (NAPI_GRO_CB(skb)->free == NAPI_GRO_FREE_STOLEN_HEAD)
			kmem_cache_free(skbuff_head_cache, skb);
		else
			__kfree_skb(skb);
3949 3950
		break;

3951 3952 3953
	case GRO_HELD:
	case GRO_MERGED:
		break;
H
Herbert Xu 已提交
3954 3955
	}

3956
	return ret;
3957 3958
}

3959
gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3960
{
E
Eric Dumazet 已提交
3961
	return napi_skb_finish(dev_gro_receive(napi, skb), skb);
3962 3963 3964
}
EXPORT_SYMBOL(napi_gro_receive);

S
stephen hemminger 已提交
3965
static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb)
3966 3967
{
	__skb_pull(skb, skb_headlen(skb));
3968 3969
	/* restore the reserve we had after netdev_alloc_skb_ip_align() */
	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN - skb_headroom(skb));
3970
	skb->vlan_tci = 0;
H
Herbert Xu 已提交
3971
	skb->dev = napi->dev;
A
Andy Gospodarek 已提交
3972
	skb->skb_iif = 0;
3973 3974 3975 3976

	napi->skb = skb;
}

3977
struct sk_buff *napi_get_frags(struct napi_struct *napi)
H
Herbert Xu 已提交
3978 3979 3980 3981
{
	struct sk_buff *skb = napi->skb;

	if (!skb) {
3982
		skb = netdev_alloc_skb_ip_align(napi->dev, GRO_MAX_HEAD);
E
Eric Dumazet 已提交
3983
		napi->skb = skb;
3984
	}
3985 3986
	return skb;
}
3987
EXPORT_SYMBOL(napi_get_frags);
3988

R
Rami Rosen 已提交
3989
static gro_result_t napi_frags_finish(struct napi_struct *napi, struct sk_buff *skb,
3990
			       gro_result_t ret)
3991
{
3992 3993
	switch (ret) {
	case GRO_NORMAL:
3994
		if (netif_receive_skb(skb))
3995
			ret = GRO_DROP;
3996
		break;
H
Herbert Xu 已提交
3997

3998 3999 4000 4001
	case GRO_DROP:
	case GRO_MERGED_FREE:
		napi_reuse_skb(napi, skb);
		break;
4002

4003
	case GRO_HELD:
4004 4005
	case GRO_MERGED:
		break;
4006
	}
H
Herbert Xu 已提交
4007

4008
	return ret;
H
Herbert Xu 已提交
4009
}
4010

E
Eric Dumazet 已提交
4011
static struct sk_buff *napi_frags_skb(struct napi_struct *napi)
4012 4013 4014 4015 4016
{
	struct sk_buff *skb = napi->skb;

	napi->skb = NULL;

4017 4018 4019
	if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr)))) {
		napi_reuse_skb(napi, skb);
		return NULL;
4020
	}
4021
	skb->protocol = eth_type_trans(skb, skb->dev);
4022 4023 4024 4025

	return skb;
}

4026
gro_result_t napi_gro_frags(struct napi_struct *napi)
4027
{
4028
	struct sk_buff *skb = napi_frags_skb(napi);
4029 4030

	if (!skb)
4031
		return GRO_DROP;
4032

E
Eric Dumazet 已提交
4033
	return napi_frags_finish(napi, skb, dev_gro_receive(napi, skb));
4034
}
H
Herbert Xu 已提交
4035 4036
EXPORT_SYMBOL(napi_gro_frags);

4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064
/*
 * net_rps_action sends any pending IPI's for rps.
 * 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();
}

4065
static int process_backlog(struct napi_struct *napi, int quota)
L
Linus Torvalds 已提交
4066 4067
{
	int work = 0;
E
Eric Dumazet 已提交
4068
	struct softnet_data *sd = container_of(napi, struct softnet_data, backlog);
L
Linus Torvalds 已提交
4069

4070 4071 4072 4073 4074 4075 4076 4077 4078
#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
4079
	napi->weight = weight_p;
4080 4081
	local_irq_disable();
	while (work < quota) {
L
Linus Torvalds 已提交
4082
		struct sk_buff *skb;
4083 4084 4085 4086 4087 4088
		unsigned int qlen;

		while ((skb = __skb_dequeue(&sd->process_queue))) {
			local_irq_enable();
			__netif_receive_skb(skb);
			local_irq_disable();
4089 4090 4091 4092 4093
			input_queue_head_incr(sd);
			if (++work >= quota) {
				local_irq_enable();
				return work;
			}
4094
		}
L
Linus Torvalds 已提交
4095

E
Eric Dumazet 已提交
4096
		rps_lock(sd);
4097
		qlen = skb_queue_len(&sd->input_pkt_queue);
4098
		if (qlen)
4099 4100
			skb_queue_splice_tail_init(&sd->input_pkt_queue,
						   &sd->process_queue);
4101

4102
		if (qlen < quota - work) {
E
Eric Dumazet 已提交
4103 4104 4105 4106 4107 4108 4109 4110 4111 4112
			/*
			 * 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;

4113
			quota = work + qlen;
4114
		}
E
Eric Dumazet 已提交
4115
		rps_unlock(sd);
4116 4117
	}
	local_irq_enable();
L
Linus Torvalds 已提交
4118

4119 4120
	return work;
}
L
Linus Torvalds 已提交
4121

4122 4123
/**
 * __napi_schedule - schedule for receive
4124
 * @n: entry to schedule
4125 4126 4127
 *
 * The entry's receive function will be scheduled to run
 */
H
Harvey Harrison 已提交
4128
void __napi_schedule(struct napi_struct *n)
4129 4130
{
	unsigned long flags;
L
Linus Torvalds 已提交
4131

4132
	local_irq_save(flags);
E
Eric Dumazet 已提交
4133
	____napi_schedule(&__get_cpu_var(softnet_data), n);
4134
	local_irq_restore(flags);
L
Linus Torvalds 已提交
4135
}
4136 4137
EXPORT_SYMBOL(__napi_schedule);

4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159
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;

4160
	napi_gro_flush(n, false);
4161 4162 4163 4164 4165 4166
	local_irq_save(flags);
	__napi_complete(n);
	local_irq_restore(flags);
}
EXPORT_SYMBOL(napi_complete);

E
Eliezer Tamir 已提交
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 4213 4214 4215 4216 4217 4218
/* 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);

4219 4220 4221 4222
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);
4223
	napi->gro_count = 0;
4224
	napi->gro_list = NULL;
H
Herbert Xu 已提交
4225
	napi->skb = NULL;
4226
	napi->poll = poll;
E
Eric Dumazet 已提交
4227 4228 4229
	if (weight > NAPI_POLL_WEIGHT)
		pr_err_once("netif_napi_add() called with weight %d on device %s\n",
			    weight, dev->name);
4230 4231 4232
	napi->weight = weight;
	list_add(&napi->dev_list, &dev->napi_list);
	napi->dev = dev;
H
Herbert Xu 已提交
4233
#ifdef CONFIG_NETPOLL
4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244
	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)
{
	struct sk_buff *skb, *next;

4245
	list_del_init(&napi->dev_list);
4246
	napi_free_frags(napi);
4247 4248 4249 4250 4251 4252 4253 4254

	for (skb = napi->gro_list; skb; skb = next) {
		next = skb->next;
		skb->next = NULL;
		kfree_skb(skb);
	}

	napi->gro_list = NULL;
4255
	napi->gro_count = 0;
4256 4257 4258
}
EXPORT_SYMBOL(netif_napi_del);

L
Linus Torvalds 已提交
4259 4260
static void net_rx_action(struct softirq_action *h)
{
4261
	struct softnet_data *sd = &__get_cpu_var(softnet_data);
4262
	unsigned long time_limit = jiffies + 2;
4263
	int budget = netdev_budget;
4264 4265
	void *have;

L
Linus Torvalds 已提交
4266 4267
	local_irq_disable();

4268
	while (!list_empty(&sd->poll_list)) {
4269 4270
		struct napi_struct *n;
		int work, weight;
L
Linus Torvalds 已提交
4271

4272
		/* If softirq window is exhuasted then punt.
4273 4274
		 * Allow this to run for 2 jiffies since which will allow
		 * an average latency of 1.5/HZ.
4275
		 */
4276
		if (unlikely(budget <= 0 || time_after_eq(jiffies, time_limit)))
L
Linus Torvalds 已提交
4277 4278 4279 4280
			goto softnet_break;

		local_irq_enable();

4281 4282 4283 4284 4285
		/* 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.
		 */
4286
		n = list_first_entry(&sd->poll_list, struct napi_struct, poll_list);
L
Linus Torvalds 已提交
4287

4288 4289 4290 4291
		have = netpoll_poll_lock(n);

		weight = n->weight;

4292 4293 4294 4295
		/* 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
4296
		 * accidentally calling ->poll() when NAPI is not scheduled.
4297 4298
		 */
		work = 0;
4299
		if (test_bit(NAPI_STATE_SCHED, &n->state)) {
4300
			work = n->poll(n, weight);
4301 4302
			trace_napi_poll(n);
		}
4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314

		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.
		 */
4315
		if (unlikely(work == weight)) {
4316 4317 4318 4319
			if (unlikely(napi_disable_pending(n))) {
				local_irq_enable();
				napi_complete(n);
				local_irq_disable();
4320 4321 4322 4323 4324 4325 4326 4327 4328
			} 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();
				}
4329
				list_move_tail(&n->poll_list, &sd->poll_list);
4330
			}
4331
		}
4332 4333

		netpoll_poll_unlock(have);
L
Linus Torvalds 已提交
4334 4335
	}
out:
4336
	net_rps_action_and_irq_enable(sd);
T
Tom Herbert 已提交
4337

4338 4339 4340 4341 4342
#ifdef CONFIG_NET_DMA
	/*
	 * There may not be any more sk_buffs coming right now, so push
	 * any pending DMA copies to hardware
	 */
4343
	dma_issue_pending_all();
4344
#endif
4345

L
Linus Torvalds 已提交
4346 4347 4348
	return;

softnet_break:
C
Changli Gao 已提交
4349
	sd->time_squeeze++;
L
Linus Torvalds 已提交
4350 4351 4352 4353
	__raise_softirq_irqoff(NET_RX_SOFTIRQ);
	goto out;
}

4354
struct netdev_adjacent {
J
Jiri Pirko 已提交
4355
	struct net_device *dev;
4356 4357

	/* upper master flag, there can only be one master device per list */
J
Jiri Pirko 已提交
4358
	bool master;
4359 4360 4361 4362

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

4363 4364 4365
	/* private field for the users */
	void *private;

J
Jiri Pirko 已提交
4366 4367 4368 4369
	struct list_head list;
	struct rcu_head rcu;
};

4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382
static struct netdev_adjacent *__netdev_find_adj_rcu(struct net_device *dev,
						     struct net_device *adj_dev,
						     struct list_head *adj_list)
{
	struct netdev_adjacent *adj;

	list_for_each_entry_rcu(adj, adj_list, list) {
		if (adj->dev == adj_dev)
			return adj;
	}
	return NULL;
}

4383 4384
static struct netdev_adjacent *__netdev_find_adj(struct net_device *dev,
						 struct net_device *adj_dev,
4385
						 struct list_head *adj_list)
J
Jiri Pirko 已提交
4386
{
4387 4388
	struct netdev_adjacent *adj;

4389
	list_for_each_entry(adj, adj_list, list) {
4390 4391
		if (adj->dev == adj_dev)
			return adj;
J
Jiri Pirko 已提交
4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409
	}
	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();

4410
	return __netdev_find_adj(dev, upper_dev, &dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424
}
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.
 */
bool netdev_has_any_upper_dev(struct net_device *dev)
{
	ASSERT_RTNL();

4425
	return !list_empty(&dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437
}
EXPORT_SYMBOL(netdev_has_any_upper_dev);

/**
 * 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)
{
4438
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4439 4440 4441

	ASSERT_RTNL();

4442
	if (list_empty(&dev->adj_list.upper))
J
Jiri Pirko 已提交
4443 4444
		return NULL;

4445
	upper = list_first_entry(&dev->adj_list.upper,
4446
				 struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4447 4448 4449 4450 4451 4452
	if (likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get);

4453 4454 4455 4456 4457 4458 4459 4460 4461 4462
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);

4463 4464
/**
 * netdev_all_upper_get_next_dev_rcu - Get the next dev from upper list
4465 4466 4467 4468 4469 4470
 * @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.
 */
4471 4472
struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
						     struct list_head **iter)
4473 4474 4475 4476 4477 4478 4479
{
	struct netdev_adjacent *upper;

	WARN_ON_ONCE(!rcu_read_lock_held());

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

4480
	if (&upper->list == &dev->all_adj_list.upper)
4481 4482 4483 4484 4485 4486
		return NULL;

	*iter = &upper->list;

	return upper->dev;
}
4487
EXPORT_SYMBOL(netdev_all_upper_get_next_dev_rcu);
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 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545
/**
 * 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);

4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566
/**
 * 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
4567 4568 4569 4570 4571 4572 4573 4574 4575
/**
 * 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)
{
4576
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4577

4578
	upper = list_first_or_null_rcu(&dev->adj_list.upper,
4579
				       struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4580 4581 4582 4583 4584 4585
	if (upper && likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get_rcu);

4586 4587
static int __netdev_adjacent_dev_insert(struct net_device *dev,
					struct net_device *adj_dev,
4588
					struct list_head *dev_list,
4589
					void *private, bool master)
4590 4591
{
	struct netdev_adjacent *adj;
4592
	char linkname[IFNAMSIZ+7];
4593
	int ret;
4594

4595
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608

	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;
4609
	adj->private = private;
4610
	dev_hold(adj_dev);
4611 4612 4613

	pr_debug("dev_hold for %s, because of link added from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
4614

4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628
	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;
	}

4629
	/* Ensure that master link is always the first item in list. */
4630 4631 4632 4633
	if (master) {
		ret = sysfs_create_link(&(dev->dev.kobj),
					&(adj_dev->dev.kobj), "master");
		if (ret)
4634
			goto remove_symlinks;
4635

4636
		list_add_rcu(&adj->list, dev_list);
4637
	} else {
4638
		list_add_tail_rcu(&adj->list, dev_list);
4639
	}
4640 4641

	return 0;
4642

4643 4644 4645 4646 4647 4648 4649 4650 4651
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);
	}

4652 4653
free_adj:
	kfree(adj);
4654
	dev_put(adj_dev);
4655 4656

	return ret;
4657 4658 4659
}

void __netdev_adjacent_dev_remove(struct net_device *dev,
4660 4661
				  struct net_device *adj_dev,
				  struct list_head *dev_list)
4662 4663
{
	struct netdev_adjacent *adj;
4664
	char linkname[IFNAMSIZ+7];
4665

4666
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
4667

4668 4669 4670
	if (!adj) {
		pr_err("tried to remove device %s from %s\n",
		       dev->name, adj_dev->name);
4671
		BUG();
4672
	}
4673 4674

	if (adj->ref_nr > 1) {
4675 4676
		pr_debug("%s to %s ref_nr-- = %d\n", dev->name, adj_dev->name,
			 adj->ref_nr-1);
4677 4678 4679 4680
		adj->ref_nr--;
		return;
	}

4681 4682 4683
	if (adj->master)
		sysfs_remove_link(&(dev->dev.kobj), "master");

4684 4685 4686 4687 4688 4689 4690 4691
	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);
	}

4692
	list_del_rcu(&adj->list);
4693 4694
	pr_debug("dev_put for %s, because link removed from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
4695 4696 4697 4698
	dev_put(adj_dev);
	kfree_rcu(adj, rcu);
}

4699 4700 4701 4702
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,
4703
				     void *private, bool master)
4704 4705 4706
{
	int ret;

4707 4708
	ret = __netdev_adjacent_dev_insert(dev, upper_dev, up_list, private,
					   master);
4709 4710 4711
	if (ret)
		return ret;

4712 4713
	ret = __netdev_adjacent_dev_insert(upper_dev, dev, down_list, private,
					   false);
4714
	if (ret) {
4715
		__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
4716 4717 4718 4719 4720 4721
		return ret;
	}

	return 0;
}

4722 4723
int __netdev_adjacent_dev_link(struct net_device *dev,
			       struct net_device *upper_dev)
4724
{
4725 4726 4727
	return __netdev_adjacent_dev_link_lists(dev, upper_dev,
						&dev->all_adj_list.upper,
						&upper_dev->all_adj_list.lower,
4728
						NULL, false);
4729 4730
}

4731 4732 4733 4734
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)
4735
{
4736 4737
	__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
	__netdev_adjacent_dev_remove(upper_dev, dev, down_list);
4738 4739 4740 4741 4742
}

void __netdev_adjacent_dev_unlink(struct net_device *dev,
				  struct net_device *upper_dev)
{
4743 4744 4745 4746 4747 4748 4749
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev,
					   &dev->all_adj_list.upper,
					   &upper_dev->all_adj_list.lower);
}

int __netdev_adjacent_dev_link_neighbour(struct net_device *dev,
					 struct net_device *upper_dev,
4750
					 void *private, bool master)
4751 4752 4753 4754 4755 4756 4757 4758 4759
{
	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,
4760
					       private, master);
4761 4762 4763 4764 4765 4766
	if (ret) {
		__netdev_adjacent_dev_unlink(dev, upper_dev);
		return ret;
	}

	return 0;
4767 4768
}

4769 4770 4771 4772 4773 4774 4775 4776
void __netdev_adjacent_dev_unlink_neighbour(struct net_device *dev,
					    struct net_device *upper_dev)
{
	__netdev_adjacent_dev_unlink(dev, upper_dev);
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev,
					   &dev->adj_list.upper,
					   &upper_dev->adj_list.lower);
}
4777

J
Jiri Pirko 已提交
4778
static int __netdev_upper_dev_link(struct net_device *dev,
4779 4780
				   struct net_device *upper_dev, bool master,
				   void *private)
J
Jiri Pirko 已提交
4781
{
4782 4783
	struct netdev_adjacent *i, *j, *to_i, *to_j;
	int ret = 0;
J
Jiri Pirko 已提交
4784 4785 4786 4787 4788 4789 4790

	ASSERT_RTNL();

	if (dev == upper_dev)
		return -EBUSY;

	/* To prevent loops, check if dev is not upper device to upper_dev. */
4791
	if (__netdev_find_adj(upper_dev, dev, &upper_dev->all_adj_list.upper))
J
Jiri Pirko 已提交
4792 4793
		return -EBUSY;

4794
	if (__netdev_find_adj(dev, upper_dev, &dev->all_adj_list.upper))
J
Jiri Pirko 已提交
4795 4796 4797 4798 4799
		return -EEXIST;

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

4800 4801
	ret = __netdev_adjacent_dev_link_neighbour(dev, upper_dev, private,
						   master);
4802 4803
	if (ret)
		return ret;
J
Jiri Pirko 已提交
4804

4805
	/* Now that we linked these devs, make all the upper_dev's
4806
	 * all_adj_list.upper visible to every dev's all_adj_list.lower an
4807 4808 4809
	 * versa, and don't forget the devices itself. All of these
	 * links are non-neighbours.
	 */
4810 4811 4812 4813
	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);
4814 4815 4816 4817 4818 4819 4820
			ret = __netdev_adjacent_dev_link(i->dev, j->dev);
			if (ret)
				goto rollback_mesh;
		}
	}

	/* add dev to every upper_dev's upper device */
4821 4822 4823
	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);
4824 4825 4826 4827 4828 4829
		ret = __netdev_adjacent_dev_link(dev, i->dev);
		if (ret)
			goto rollback_upper_mesh;
	}

	/* add upper_dev to every dev's lower device */
4830 4831 4832
	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);
4833 4834 4835 4836
		ret = __netdev_adjacent_dev_link(i->dev, upper_dev);
		if (ret)
			goto rollback_lower_mesh;
	}
J
Jiri Pirko 已提交
4837

4838
	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
J
Jiri Pirko 已提交
4839
	return 0;
4840 4841 4842

rollback_lower_mesh:
	to_i = i;
4843
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
4844 4845 4846 4847 4848 4849 4850 4851 4852
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);
	}

	i = NULL;

rollback_upper_mesh:
	to_i = i;
4853
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list) {
4854 4855 4856 4857 4858 4859 4860 4861 4862 4863
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(dev, i->dev);
	}

	i = j = NULL;

rollback_mesh:
	to_i = i;
	to_j = j;
4864 4865
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list) {
4866 4867 4868 4869 4870 4871 4872 4873
			if (i == to_i && j == to_j)
				break;
			__netdev_adjacent_dev_unlink(i->dev, j->dev);
		}
		if (i == to_i)
			break;
	}

4874
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
4875 4876

	return ret;
J
Jiri Pirko 已提交
4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891
}

/**
 * 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)
{
4892
	return __netdev_upper_dev_link(dev, upper_dev, false, NULL);
J
Jiri Pirko 已提交
4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909
}
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)
{
4910
	return __netdev_upper_dev_link(dev, upper_dev, true, NULL);
J
Jiri Pirko 已提交
4911 4912 4913
}
EXPORT_SYMBOL(netdev_master_upper_dev_link);

4914 4915 4916 4917 4918 4919 4920 4921
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
4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932
/**
 * 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)
{
4933
	struct netdev_adjacent *i, *j;
J
Jiri Pirko 已提交
4934 4935
	ASSERT_RTNL();

4936
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
4937 4938 4939 4940 4941

	/* 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.
	 */
4942 4943
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list)
4944 4945 4946 4947 4948
			__netdev_adjacent_dev_unlink(i->dev, j->dev);

	/* remove also the devices itself from lower/upper device
	 * list
	 */
4949
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
4950 4951
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);

4952
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list)
4953 4954
		__netdev_adjacent_dev_unlink(dev, i->dev);

4955
	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
J
Jiri Pirko 已提交
4956 4957 4958
}
EXPORT_SYMBOL(netdev_upper_dev_unlink);

4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988
void *netdev_lower_dev_get_private_rcu(struct net_device *dev,
				       struct net_device *lower_dev)
{
	struct netdev_adjacent *lower;

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

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

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);

4989 4990
static void dev_change_rx_flags(struct net_device *dev, int flags)
{
4991 4992
	const struct net_device_ops *ops = dev->netdev_ops;

4993
	if (ops->ndo_change_rx_flags)
4994
		ops->ndo_change_rx_flags(dev, flags);
4995 4996
}

4997
static int __dev_set_promiscuity(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
4998
{
4999
	unsigned int old_flags = dev->flags;
5000 5001
	kuid_t uid;
	kgid_t gid;
L
Linus Torvalds 已提交
5002

5003 5004
	ASSERT_RTNL();

5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015
	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;
5016 5017
			pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n",
				dev->name);
5018 5019 5020
			return -EOVERFLOW;
		}
	}
5021
	if (dev->flags != old_flags) {
5022 5023 5024
		pr_info("device %s %s promiscuous mode\n",
			dev->name,
			dev->flags & IFF_PROMISC ? "entered" : "left");
5025 5026
		if (audit_enabled) {
			current_uid_gid(&uid, &gid);
5027 5028 5029 5030 5031
			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),
5032
				from_kuid(&init_user_ns, audit_get_loginuid(current)),
5033 5034
				from_kuid(&init_user_ns, uid),
				from_kgid(&init_user_ns, gid),
5035
				audit_get_sessionid(current));
5036
		}
5037

5038
		dev_change_rx_flags(dev, IFF_PROMISC);
L
Linus Torvalds 已提交
5039
	}
5040 5041
	if (notify)
		__dev_notify_flags(dev, old_flags, IFF_PROMISC);
5042
	return 0;
L
Linus Torvalds 已提交
5043 5044
}

5045 5046 5047 5048 5049 5050 5051 5052 5053
/**
 *	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.
5054
 *	Return 0 if successful or a negative errno code on error.
5055
 */
5056
int dev_set_promiscuity(struct net_device *dev, int inc)
5057
{
5058
	unsigned int old_flags = dev->flags;
5059
	int err;
5060

5061
	err = __dev_set_promiscuity(dev, inc, true);
5062
	if (err < 0)
5063
		return err;
5064 5065
	if (dev->flags != old_flags)
		dev_set_rx_mode(dev);
5066
	return err;
5067
}
E
Eric Dumazet 已提交
5068
EXPORT_SYMBOL(dev_set_promiscuity);
5069

5070
static int __dev_set_allmulti(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5071
{
5072
	unsigned int old_flags = dev->flags, old_gflags = dev->gflags;
L
Linus Torvalds 已提交
5073

5074 5075
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5076
	dev->flags |= IFF_ALLMULTI;
5077 5078 5079 5080 5081 5082 5083 5084 5085 5086
	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;
5087 5088
			pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n",
				dev->name);
5089 5090 5091
			return -EOVERFLOW;
		}
	}
5092
	if (dev->flags ^ old_flags) {
5093
		dev_change_rx_flags(dev, IFF_ALLMULTI);
5094
		dev_set_rx_mode(dev);
5095 5096 5097
		if (notify)
			__dev_notify_flags(dev, old_flags,
					   dev->gflags ^ old_gflags);
5098
	}
5099
	return 0;
5100
}
5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118

/**
 *	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
5119
EXPORT_SYMBOL(dev_set_allmulti);
5120 5121 5122 5123

/*
 *	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
5124
 *	filtering it is put in promiscuous mode while unicast addresses
5125 5126 5127 5128
 *	are present.
 */
void __dev_set_rx_mode(struct net_device *dev)
{
5129 5130
	const struct net_device_ops *ops = dev->netdev_ops;

5131 5132 5133 5134 5135
	/* dev_open will call this function so the list will stay sane. */
	if (!(dev->flags&IFF_UP))
		return;

	if (!netif_device_present(dev))
5136
		return;
5137

5138
	if (!(dev->priv_flags & IFF_UNICAST_FLT)) {
5139 5140 5141
		/* Unicast addresses changes may only happen under the rtnl,
		 * therefore calling __dev_set_promiscuity here is safe.
		 */
5142
		if (!netdev_uc_empty(dev) && !dev->uc_promisc) {
5143
			__dev_set_promiscuity(dev, 1, false);
5144
			dev->uc_promisc = true;
5145
		} else if (netdev_uc_empty(dev) && dev->uc_promisc) {
5146
			__dev_set_promiscuity(dev, -1, false);
5147
			dev->uc_promisc = false;
5148 5149
		}
	}
5150 5151 5152

	if (ops->ndo_set_rx_mode)
		ops->ndo_set_rx_mode(dev);
5153 5154 5155 5156
}

void dev_set_rx_mode(struct net_device *dev)
{
5157
	netif_addr_lock_bh(dev);
5158
	__dev_set_rx_mode(dev);
5159
	netif_addr_unlock_bh(dev);
L
Linus Torvalds 已提交
5160 5161
}

5162 5163 5164 5165 5166 5167
/**
 *	dev_get_flags - get flags reported to userspace
 *	@dev: device
 *
 *	Get the combination of flag bits exported through APIs to userspace.
 */
5168
unsigned int dev_get_flags(const struct net_device *dev)
L
Linus Torvalds 已提交
5169
{
5170
	unsigned int flags;
L
Linus Torvalds 已提交
5171 5172 5173

	flags = (dev->flags & ~(IFF_PROMISC |
				IFF_ALLMULTI |
S
Stefan Rompf 已提交
5174 5175 5176
				IFF_RUNNING |
				IFF_LOWER_UP |
				IFF_DORMANT)) |
L
Linus Torvalds 已提交
5177 5178 5179
		(dev->gflags & (IFF_PROMISC |
				IFF_ALLMULTI));

S
Stefan Rompf 已提交
5180 5181 5182 5183 5184 5185 5186 5187
	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
5188 5189 5190

	return flags;
}
E
Eric Dumazet 已提交
5191
EXPORT_SYMBOL(dev_get_flags);
L
Linus Torvalds 已提交
5192

5193
int __dev_change_flags(struct net_device *dev, unsigned int flags)
L
Linus Torvalds 已提交
5194
{
5195
	unsigned int old_flags = dev->flags;
5196
	int ret;
L
Linus Torvalds 已提交
5197

5198 5199
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213
	/*
	 *	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.
	 */

5214 5215
	if ((old_flags ^ flags) & IFF_MULTICAST)
		dev_change_rx_flags(dev, IFF_MULTICAST);
5216

5217
	dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5218 5219 5220 5221 5222 5223 5224 5225 5226

	/*
	 *	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  ? */
5227
		ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev);
L
Linus Torvalds 已提交
5228 5229

		if (!ret)
5230
			dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5231 5232 5233
	}

	if ((flags ^ dev->gflags) & IFF_PROMISC) {
E
Eric Dumazet 已提交
5234
		int inc = (flags & IFF_PROMISC) ? 1 : -1;
5235
		unsigned int old_flags = dev->flags;
E
Eric Dumazet 已提交
5236

L
Linus Torvalds 已提交
5237
		dev->gflags ^= IFF_PROMISC;
5238 5239 5240 5241

		if (__dev_set_promiscuity(dev, inc, false) >= 0)
			if (dev->flags != old_flags)
				dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5242 5243 5244 5245 5246 5247 5248
	}

	/* 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
5249 5250
		int inc = (flags & IFF_ALLMULTI) ? 1 : -1;

L
Linus Torvalds 已提交
5251
		dev->gflags ^= IFF_ALLMULTI;
5252
		__dev_set_allmulti(dev, inc, false);
L
Linus Torvalds 已提交
5253 5254
	}

5255 5256 5257
	return ret;
}

5258 5259
void __dev_notify_flags(struct net_device *dev, unsigned int old_flags,
			unsigned int gchanges)
5260 5261 5262
{
	unsigned int changes = dev->flags ^ old_flags;

5263
	if (gchanges)
5264
		rtmsg_ifinfo(RTM_NEWLINK, dev, gchanges, GFP_ATOMIC);
5265

5266 5267 5268 5269 5270 5271 5272 5273
	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 &&
5274 5275 5276 5277 5278 5279 5280
	    (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);
	}
5281 5282 5283 5284 5285 5286 5287 5288 5289 5290
}

/**
 *	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.
 */
5291
int dev_change_flags(struct net_device *dev, unsigned int flags)
5292
{
5293
	int ret;
5294
	unsigned int changes, old_flags = dev->flags, old_gflags = dev->gflags;
5295 5296 5297 5298 5299

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

5300
	changes = (old_flags ^ dev->flags) | (old_gflags ^ dev->gflags);
5301
	__dev_notify_flags(dev, old_flags, changes);
L
Linus Torvalds 已提交
5302 5303
	return ret;
}
E
Eric Dumazet 已提交
5304
EXPORT_SYMBOL(dev_change_flags);
L
Linus Torvalds 已提交
5305

5306 5307 5308 5309 5310 5311 5312
/**
 *	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
5313 5314
int dev_set_mtu(struct net_device *dev, int new_mtu)
{
5315
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328
	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;
5329 5330
	if (ops->ndo_change_mtu)
		err = ops->ndo_change_mtu(dev, new_mtu);
L
Linus Torvalds 已提交
5331 5332
	else
		dev->mtu = new_mtu;
5333

5334
	if (!err)
5335
		call_netdevice_notifiers(NETDEV_CHANGEMTU, dev);
L
Linus Torvalds 已提交
5336 5337
	return err;
}
E
Eric Dumazet 已提交
5338
EXPORT_SYMBOL(dev_set_mtu);
L
Linus Torvalds 已提交
5339

5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350
/**
 *	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);

5351 5352 5353 5354 5355 5356 5357
/**
 *	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
5358 5359
int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa)
{
5360
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
5361 5362
	int err;

5363
	if (!ops->ndo_set_mac_address)
L
Linus Torvalds 已提交
5364 5365 5366 5367 5368
		return -EOPNOTSUPP;
	if (sa->sa_family != dev->type)
		return -EINVAL;
	if (!netif_device_present(dev))
		return -ENODEV;
5369
	err = ops->ndo_set_mac_address(dev, sa);
5370 5371
	if (err)
		return err;
J
Jiri Pirko 已提交
5372
	dev->addr_assign_type = NET_ADDR_SET;
5373
	call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
5374
	add_device_randomness(dev->dev_addr, dev->addr_len);
5375
	return 0;
L
Linus Torvalds 已提交
5376
}
E
Eric Dumazet 已提交
5377
EXPORT_SYMBOL(dev_set_mac_address);
L
Linus Torvalds 已提交
5378

J
Jiri Pirko 已提交
5379 5380 5381
/**
 *	dev_change_carrier - Change device carrier
 *	@dev: device
5382
 *	@new_carrier: new value
J
Jiri Pirko 已提交
5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397
 *
 *	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);

5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415
/**
 *	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
5416 5417
/**
 *	dev_new_index	-	allocate an ifindex
5418
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
5419 5420 5421 5422 5423
 *
 *	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.
 */
5424
static int dev_new_index(struct net *net)
L
Linus Torvalds 已提交
5425
{
5426
	int ifindex = net->ifindex;
L
Linus Torvalds 已提交
5427 5428 5429
	for (;;) {
		if (++ifindex <= 0)
			ifindex = 1;
5430
		if (!__dev_get_by_index(net, ifindex))
5431
			return net->ifindex = ifindex;
L
Linus Torvalds 已提交
5432 5433 5434 5435
	}
}

/* Delayed registration/unregisteration */
5436
static LIST_HEAD(net_todo_list);
5437
static DECLARE_WAIT_QUEUE_HEAD(netdev_unregistering_wq);
L
Linus Torvalds 已提交
5438

5439
static void net_set_todo(struct net_device *dev)
L
Linus Torvalds 已提交
5440 5441
{
	list_add_tail(&dev->todo_list, &net_todo_list);
5442
	dev_net(dev)->dev_unreg_count++;
L
Linus Torvalds 已提交
5443 5444
}

5445
static void rollback_registered_many(struct list_head *head)
5446
{
5447
	struct net_device *dev, *tmp;
5448
	LIST_HEAD(close_head);
5449

5450 5451 5452
	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

5453
	list_for_each_entry_safe(dev, tmp, head, unreg_list) {
5454
		/* Some devices call without registering
5455 5456
		 * for initialization unwind. Remove those
		 * devices and proceed with the remaining.
5457 5458
		 */
		if (dev->reg_state == NETREG_UNINITIALIZED) {
5459 5460
			pr_debug("unregister_netdevice: device %s/%p never was registered\n",
				 dev->name, dev);
5461

5462
			WARN_ON(1);
5463 5464
			list_del(&dev->unreg_list);
			continue;
5465
		}
5466
		dev->dismantle = true;
5467
		BUG_ON(dev->reg_state != NETREG_REGISTERED);
5468
	}
5469

5470
	/* If device is running, close it first. */
5471 5472 5473
	list_for_each_entry(dev, head, unreg_list)
		list_add_tail(&dev->close_list, &close_head);
	dev_close_many(&close_head);
5474

5475
	list_for_each_entry(dev, head, unreg_list) {
5476 5477
		/* And unlink it from device chain. */
		unlist_netdevice(dev);
5478

5479 5480
		dev->reg_state = NETREG_UNREGISTERING;
	}
5481 5482 5483

	synchronize_net();

5484 5485 5486
	list_for_each_entry(dev, head, unreg_list) {
		/* Shutdown queueing discipline. */
		dev_shutdown(dev);
5487 5488


5489 5490 5491 5492
		/* Notify protocols, that we are about to destroy
		   this device. They should clean all the things.
		*/
		call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
5493

5494 5495
		if (!dev->rtnl_link_ops ||
		    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
5496
			rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);
5497

5498 5499 5500
		/*
		 *	Flush the unicast and multicast chains
		 */
5501
		dev_uc_flush(dev);
5502
		dev_mc_flush(dev);
5503

5504 5505
		if (dev->netdev_ops->ndo_uninit)
			dev->netdev_ops->ndo_uninit(dev);
5506

J
Jiri Pirko 已提交
5507 5508
		/* Notifier chain MUST detach us all upper devices. */
		WARN_ON(netdev_has_any_upper_dev(dev));
5509

5510 5511
		/* Remove entries from kobject tree */
		netdev_unregister_kobject(dev);
5512 5513 5514 5515
#ifdef CONFIG_XPS
		/* Remove XPS queueing entries */
		netif_reset_xps_queues_gt(dev, 0);
#endif
5516
	}
5517

5518
	synchronize_net();
5519

5520
	list_for_each_entry(dev, head, unreg_list)
5521 5522 5523 5524 5525 5526 5527 5528 5529
		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
5530
	list_del(&single);
5531 5532
}

5533 5534
static netdev_features_t netdev_fix_features(struct net_device *dev,
	netdev_features_t features)
5535
{
5536 5537 5538
	/* Fix illegal checksum combinations */
	if ((features & NETIF_F_HW_CSUM) &&
	    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
5539
		netdev_warn(dev, "mixed HW and IP checksum settings.\n");
5540 5541 5542
		features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
	}

5543
	/* TSO requires that SG is present as well. */
5544
	if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
5545
		netdev_dbg(dev, "Dropping TSO features since no SG feature.\n");
5546
		features &= ~NETIF_F_ALL_TSO;
5547 5548
	}

5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561
	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;
	}

5562 5563 5564 5565
	/* TSO ECN requires that TSO is present as well. */
	if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN)
		features &= ~NETIF_F_TSO_ECN;

5566 5567
	/* Software GSO depends on SG. */
	if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) {
5568
		netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n");
5569 5570 5571
		features &= ~NETIF_F_GSO;
	}

5572
	/* UFO needs SG and checksumming */
5573
	if (features & NETIF_F_UFO) {
5574 5575 5576 5577
		/* 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))) {
5578
			netdev_dbg(dev,
5579
				"Dropping NETIF_F_UFO since no checksum offload features.\n");
5580 5581 5582 5583
			features &= ~NETIF_F_UFO;
		}

		if (!(features & NETIF_F_SG)) {
5584
			netdev_dbg(dev,
5585
				"Dropping NETIF_F_UFO since no NETIF_F_SG feature.\n");
5586 5587 5588 5589 5590 5591 5592
			features &= ~NETIF_F_UFO;
		}
	}

	return features;
}

5593
int __netdev_update_features(struct net_device *dev)
5594
{
5595
	netdev_features_t features;
5596 5597
	int err = 0;

5598 5599
	ASSERT_RTNL();

5600 5601 5602 5603 5604 5605 5606 5607 5608
	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)
5609
		return 0;
5610

5611 5612
	netdev_dbg(dev, "Features changed: %pNF -> %pNF\n",
		&dev->features, &features);
5613 5614 5615 5616

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

5617
	if (unlikely(err < 0)) {
5618
		netdev_err(dev,
5619 5620
			"set_features() failed (%d); wanted %pNF, left %pNF\n",
			err, &features, &dev->features);
5621 5622 5623 5624 5625 5626 5627 5628 5629
		return -1;
	}

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

	return 1;
}

5630 5631 5632 5633 5634 5635 5636 5637
/**
 *	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.
 */
5638 5639 5640 5641
void netdev_update_features(struct net_device *dev)
{
	if (__netdev_update_features(dev))
		netdev_features_change(dev);
5642 5643 5644
}
EXPORT_SYMBOL(netdev_update_features);

5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661
/**
 *	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);

5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688
/**
 *	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
5689
#ifdef CONFIG_RPS
5690 5691 5692
static int netif_alloc_rx_queues(struct net_device *dev)
{
	unsigned int i, count = dev->num_rx_queues;
T
Tom Herbert 已提交
5693
	struct netdev_rx_queue *rx;
5694

T
Tom Herbert 已提交
5695
	BUG_ON(count < 1);
5696

T
Tom Herbert 已提交
5697
	rx = kcalloc(count, sizeof(struct netdev_rx_queue), GFP_KERNEL);
5698
	if (!rx)
T
Tom Herbert 已提交
5699
		return -ENOMEM;
5700

T
Tom Herbert 已提交
5701 5702 5703
	dev->_rx = rx;

	for (i = 0; i < count; i++)
T
Tom Herbert 已提交
5704
		rx[i].dev = dev;
5705 5706
	return 0;
}
T
Tom Herbert 已提交
5707
#endif
5708

C
Changli Gao 已提交
5709 5710 5711 5712 5713 5714 5715
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;
5716
	netdev_queue_numa_node_write(queue, NUMA_NO_NODE);
C
Changli Gao 已提交
5717
	queue->dev = dev;
T
Tom Herbert 已提交
5718 5719 5720
#ifdef CONFIG_BQL
	dql_init(&queue->dql, HZ);
#endif
C
Changli Gao 已提交
5721 5722
}

5723 5724 5725 5726 5727 5728 5729 5730
static void netif_free_tx_queues(struct net_device *dev)
{
	if (is_vmalloc_addr(dev->_tx))
		vfree(dev->_tx);
	else
		kfree(dev->_tx);
}

5731 5732 5733 5734
static int netif_alloc_netdev_queues(struct net_device *dev)
{
	unsigned int count = dev->num_tx_queues;
	struct netdev_queue *tx;
5735
	size_t sz = count * sizeof(*tx);
5736

5737
	BUG_ON(count < 1 || count > 0xffff);
5738

5739 5740 5741 5742 5743 5744
	tx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!tx) {
		tx = vzalloc(sz);
		if (!tx)
			return -ENOMEM;
	}
5745
	dev->_tx = tx;
T
Tom Herbert 已提交
5746

5747 5748
	netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
	spin_lock_init(&dev->tx_global_lock);
C
Changli Gao 已提交
5749 5750

	return 0;
5751 5752
}

L
Linus Torvalds 已提交
5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772
/**
 *	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;
5773
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
5774 5775 5776 5777

	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

5778 5779
	might_sleep();

L
Linus Torvalds 已提交
5780 5781
	/* When net_device's are persistent, this will be fatal. */
	BUG_ON(dev->reg_state != NETREG_UNINITIALIZED);
5782
	BUG_ON(!net);
L
Linus Torvalds 已提交
5783

5784
	spin_lock_init(&dev->addr_list_lock);
5785
	netdev_set_addr_lockdep_class(dev);
L
Linus Torvalds 已提交
5786 5787 5788

	dev->iflink = -1;

5789
	ret = dev_get_valid_name(net, dev, dev->name);
5790 5791 5792
	if (ret < 0)
		goto out;

L
Linus Torvalds 已提交
5793
	/* Init, if this function is available */
5794 5795
	if (dev->netdev_ops->ndo_init) {
		ret = dev->netdev_ops->ndo_init(dev);
L
Linus Torvalds 已提交
5796 5797 5798
		if (ret) {
			if (ret > 0)
				ret = -EIO;
5799
			goto out;
L
Linus Torvalds 已提交
5800 5801
		}
	}
5802

5803 5804
	if (((dev->hw_features | dev->features) &
	     NETIF_F_HW_VLAN_CTAG_FILTER) &&
5805 5806 5807 5808 5809 5810 5811
	    (!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;
	}

5812 5813 5814 5815 5816 5817
	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
5818 5819 5820
	if (dev->iflink == -1)
		dev->iflink = dev->ifindex;

5821 5822 5823 5824
	/* Transfer changeable features to wanted_features and enable
	 * software offloads (GSO and GRO).
	 */
	dev->hw_features |= NETIF_F_SOFT_FEATURES;
5825 5826
	dev->features |= NETIF_F_SOFT_FEATURES;
	dev->wanted_features = dev->features & dev->hw_features;
L
Linus Torvalds 已提交
5827

5828
	/* Turn on no cache copy if HW is doing checksum */
5829 5830 5831 5832 5833 5834
	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;
		}
5835 5836
	}

5837
	/* Make NETIF_F_HIGHDMA inheritable to VLAN devices.
5838
	 */
5839
	dev->vlan_features |= NETIF_F_HIGHDMA;
5840

5841 5842 5843 5844
	/* Make NETIF_F_SG inheritable to tunnel devices.
	 */
	dev->hw_enc_features |= NETIF_F_SG;

S
Simon Horman 已提交
5845 5846 5847 5848
	/* Make NETIF_F_SG inheritable to MPLS.
	 */
	dev->mpls_features |= NETIF_F_SG;

5849 5850 5851 5852 5853
	ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		goto err_uninit;

5854
	ret = netdev_register_kobject(dev);
5855
	if (ret)
5856
		goto err_uninit;
5857 5858
	dev->reg_state = NETREG_REGISTERED;

5859
	__netdev_update_features(dev);
5860

L
Linus Torvalds 已提交
5861 5862 5863 5864 5865 5866 5867
	/*
	 *	Default initial state at registry is that the
	 *	device is present.
	 */

	set_bit(__LINK_STATE_PRESENT, &dev->state);

5868 5869
	linkwatch_init_dev(dev);

L
Linus Torvalds 已提交
5870 5871
	dev_init_scheduler(dev);
	dev_hold(dev);
5872
	list_netdevice(dev);
5873
	add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
5874

5875 5876 5877 5878 5879 5880 5881
	/* 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
5882
	/* Notify protocols, that a new device appeared. */
5883
	ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
5884
	ret = notifier_to_errno(ret);
5885 5886 5887 5888
	if (ret) {
		rollback_registered(dev);
		dev->reg_state = NETREG_UNREGISTERED;
	}
5889 5890 5891 5892
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
5893 5894
	if (!dev->rtnl_link_ops ||
	    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
5895
		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
L
Linus Torvalds 已提交
5896 5897 5898

out:
	return ret;
5899 5900

err_uninit:
5901 5902
	if (dev->netdev_ops->ndo_uninit)
		dev->netdev_ops->ndo_uninit(dev);
5903
	goto out;
L
Linus Torvalds 已提交
5904
}
E
Eric Dumazet 已提交
5905
EXPORT_SYMBOL(register_netdevice);
L
Linus Torvalds 已提交
5906

5907 5908 5909 5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937
/**
 *	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
5938 5939 5940 5941 5942
	/* Note : We dont allocate pcpu_refcnt for dummy devices,
	 * because users of this 'device' dont need to change
	 * its refcount.
	 */

5943 5944 5945 5946 5947
	return 0;
}
EXPORT_SYMBOL_GPL(init_dummy_netdev);


L
Linus Torvalds 已提交
5948 5949 5950 5951 5952 5953 5954 5955 5956
/**
 *	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.
 *
5957
 *	This is a wrapper around register_netdevice that takes the rtnl semaphore
L
Linus Torvalds 已提交
5958 5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971
 *	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
5972 5973 5974 5975 5976 5977 5978 5979 5980 5981
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);

5982
/**
L
Linus Torvalds 已提交
5983
 * netdev_wait_allrefs - wait until all references are gone.
5984
 * @dev: target net_device
L
Linus Torvalds 已提交
5985 5986 5987 5988 5989 5990 5991
 *
 * 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
5992
 * call dev_put.
L
Linus Torvalds 已提交
5993 5994 5995 5996
 */
static void netdev_wait_allrefs(struct net_device *dev)
{
	unsigned long rebroadcast_time, warning_time;
E
Eric Dumazet 已提交
5997
	int refcnt;
L
Linus Torvalds 已提交
5998

5999 6000
	linkwatch_forget_dev(dev);

L
Linus Torvalds 已提交
6001
	rebroadcast_time = warning_time = jiffies;
E
Eric Dumazet 已提交
6002 6003 6004
	refcnt = netdev_refcnt_read(dev);

	while (refcnt != 0) {
L
Linus Torvalds 已提交
6005
		if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
6006
			rtnl_lock();
L
Linus Torvalds 已提交
6007 6008

			/* Rebroadcast unregister notification */
6009
			call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
L
Linus Torvalds 已提交
6010

6011
			__rtnl_unlock();
6012
			rcu_barrier();
6013 6014
			rtnl_lock();

6015
			call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
L
Linus Torvalds 已提交
6016 6017 6018 6019 6020 6021 6022 6023 6024 6025 6026
			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();
			}

6027
			__rtnl_unlock();
L
Linus Torvalds 已提交
6028 6029 6030 6031 6032 6033

			rebroadcast_time = jiffies;
		}

		msleep(250);

E
Eric Dumazet 已提交
6034 6035
		refcnt = netdev_refcnt_read(dev);

L
Linus Torvalds 已提交
6036
		if (time_after(jiffies, warning_time + 10 * HZ)) {
6037 6038
			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
				 dev->name, refcnt);
L
Linus Torvalds 已提交
6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057
			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
6058
 * We are invoked by rtnl_unlock().
L
Linus Torvalds 已提交
6059
 * This allows us to deal with problems:
6060
 * 1) We can delete sysfs objects which invoke hotplug
L
Linus Torvalds 已提交
6061 6062 6063
 *    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
6064 6065 6066
 *
 * 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
6067 6068 6069
 */
void netdev_run_todo(void)
{
6070
	struct list_head list;
L
Linus Torvalds 已提交
6071 6072

	/* Snapshot list, allow later requests */
6073
	list_replace_init(&net_todo_list, &list);
H
Herbert Xu 已提交
6074 6075

	__rtnl_unlock();
6076

6077 6078

	/* Wait for rcu callbacks to finish before next phase */
6079 6080 6081
	if (!list_empty(&list))
		rcu_barrier();

L
Linus Torvalds 已提交
6082 6083
	while (!list_empty(&list)) {
		struct net_device *dev
6084
			= list_first_entry(&list, struct net_device, todo_list);
L
Linus Torvalds 已提交
6085 6086
		list_del(&dev->todo_list);

6087
		rtnl_lock();
6088
		call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6089
		__rtnl_unlock();
6090

6091
		if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
6092
			pr_err("network todo '%s' but state %d\n",
6093 6094 6095 6096
			       dev->name, dev->reg_state);
			dump_stack();
			continue;
		}
L
Linus Torvalds 已提交
6097

6098
		dev->reg_state = NETREG_UNREGISTERED;
L
Linus Torvalds 已提交
6099

6100
		on_each_cpu(flush_backlog, dev, 1);
6101

6102
		netdev_wait_allrefs(dev);
L
Linus Torvalds 已提交
6103

6104
		/* paranoia */
E
Eric Dumazet 已提交
6105
		BUG_ON(netdev_refcnt_read(dev));
6106 6107
		WARN_ON(rcu_access_pointer(dev->ip_ptr));
		WARN_ON(rcu_access_pointer(dev->ip6_ptr));
6108
		WARN_ON(dev->dn_ptr);
L
Linus Torvalds 已提交
6109

6110 6111
		if (dev->destructor)
			dev->destructor(dev);
6112

6113 6114 6115 6116 6117 6118
		/* Report a network device has been unregistered */
		rtnl_lock();
		dev_net(dev)->dev_unreg_count--;
		__rtnl_unlock();
		wake_up(&netdev_unregistering_wq);

6119 6120
		/* Free network device */
		kobject_put(&dev->dev.kobj);
L
Linus Torvalds 已提交
6121 6122 6123
	}
}

6124 6125 6126
/* Convert net_device_stats to rtnl_link_stats64.  They have the same
 * fields in the same order, with only the type differing.
 */
6127 6128
void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
			     const struct net_device_stats *netdev_stats)
6129 6130
{
#if BITS_PER_LONG == 64
6131 6132
	BUILD_BUG_ON(sizeof(*stats64) != sizeof(*netdev_stats));
	memcpy(stats64, netdev_stats, sizeof(*stats64));
6133 6134 6135 6136 6137 6138 6139 6140 6141 6142 6143
#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
}
6144
EXPORT_SYMBOL(netdev_stats_to_stats64);
6145

6146 6147 6148
/**
 *	dev_get_stats	- get network device statistics
 *	@dev: device to get statistics from
6149
 *	@storage: place to store stats
6150
 *
6151 6152 6153 6154
 *	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.
6155
 */
6156 6157
struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
					struct rtnl_link_stats64 *storage)
6158
{
6159 6160
	const struct net_device_ops *ops = dev->netdev_ops;

6161 6162
	if (ops->ndo_get_stats64) {
		memset(storage, 0, sizeof(*storage));
6163 6164
		ops->ndo_get_stats64(dev, storage);
	} else if (ops->ndo_get_stats) {
6165
		netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev));
6166 6167
	} else {
		netdev_stats_to_stats64(storage, &dev->stats);
6168
	}
6169
	storage->rx_dropped += atomic_long_read(&dev->rx_dropped);
6170
	return storage;
R
Rusty Russell 已提交
6171
}
6172
EXPORT_SYMBOL(dev_get_stats);
R
Rusty Russell 已提交
6173

6174
struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
6175
{
6176
	struct netdev_queue *queue = dev_ingress_queue(dev);
6177

6178 6179 6180 6181 6182 6183 6184 6185 6186 6187 6188 6189
#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;
6190 6191
}

6192 6193
static const struct ethtool_ops default_ethtool_ops;

6194 6195 6196 6197 6198 6199 6200 6201
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);

6202 6203 6204 6205 6206 6207 6208 6209 6210 6211
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
6212
/**
T
Tom Herbert 已提交
6213
 *	alloc_netdev_mqs - allocate network device
L
Linus Torvalds 已提交
6214 6215 6216
 *	@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
6217 6218
 *	@txqs:		the number of TX subqueues to allocate
 *	@rxqs:		the number of RX subqueues to allocate
L
Linus Torvalds 已提交
6219 6220
 *
 *	Allocates a struct net_device with private data area for driver use
6221
 *	and performs basic initialization.  Also allocates subquue structs
T
Tom Herbert 已提交
6222
 *	for each queue on the device.
L
Linus Torvalds 已提交
6223
 */
T
Tom Herbert 已提交
6224 6225 6226
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
6227 6228
{
	struct net_device *dev;
6229
	size_t alloc_size;
6230
	struct net_device *p;
L
Linus Torvalds 已提交
6231

6232 6233
	BUG_ON(strlen(name) >= sizeof(dev->name));

T
Tom Herbert 已提交
6234
	if (txqs < 1) {
6235
		pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
6236 6237 6238
		return NULL;
	}

T
Tom Herbert 已提交
6239 6240
#ifdef CONFIG_RPS
	if (rxqs < 1) {
6241
		pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
T
Tom Herbert 已提交
6242 6243 6244 6245
		return NULL;
	}
#endif

6246
	alloc_size = sizeof(struct net_device);
6247 6248
	if (sizeof_priv) {
		/* ensure 32-byte alignment of private area */
6249
		alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
6250 6251 6252
		alloc_size += sizeof_priv;
	}
	/* ensure 32-byte alignment of whole construct */
6253
	alloc_size += NETDEV_ALIGN - 1;
L
Linus Torvalds 已提交
6254

6255 6256 6257
	p = kzalloc(alloc_size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!p)
		p = vzalloc(alloc_size);
6258
	if (!p)
L
Linus Torvalds 已提交
6259 6260
		return NULL;

6261
	dev = PTR_ALIGN(p, NETDEV_ALIGN);
L
Linus Torvalds 已提交
6262
	dev->padded = (char *)dev - (char *)p;
6263

E
Eric Dumazet 已提交
6264 6265
	dev->pcpu_refcnt = alloc_percpu(int);
	if (!dev->pcpu_refcnt)
6266
		goto free_dev;
6267 6268

	if (dev_addr_init(dev))
E
Eric Dumazet 已提交
6269
		goto free_pcpu;
6270

6271
	dev_mc_init(dev);
6272
	dev_uc_init(dev);
J
Jiri Pirko 已提交
6273

6274
	dev_net_set(dev, &init_net);
L
Linus Torvalds 已提交
6275

6276
	dev->gso_max_size = GSO_MAX_SIZE;
6277
	dev->gso_max_segs = GSO_MAX_SEGS;
6278 6279 6280

	INIT_LIST_HEAD(&dev->napi_list);
	INIT_LIST_HEAD(&dev->unreg_list);
6281
	INIT_LIST_HEAD(&dev->close_list);
6282
	INIT_LIST_HEAD(&dev->link_watch_list);
6283 6284 6285 6286
	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);
6287 6288 6289
	dev->priv_flags = IFF_XMIT_DST_RELEASE;
	setup(dev);

T
Tom Herbert 已提交
6290 6291
	dev->num_tx_queues = txqs;
	dev->real_num_tx_queues = txqs;
6292
	if (netif_alloc_netdev_queues(dev))
6293
		goto free_all;
6294

E
Eric Dumazet 已提交
6295
#ifdef CONFIG_RPS
T
Tom Herbert 已提交
6296 6297
	dev->num_rx_queues = rxqs;
	dev->real_num_rx_queues = rxqs;
T
Tom Herbert 已提交
6298
	if (netif_alloc_rx_queues(dev))
6299
		goto free_all;
E
Eric Dumazet 已提交
6300
#endif
T
Tom Herbert 已提交
6301

L
Linus Torvalds 已提交
6302
	strcpy(dev->name, name);
6303
	dev->group = INIT_NETDEV_GROUP;
6304 6305
	if (!dev->ethtool_ops)
		dev->ethtool_ops = &default_ethtool_ops;
L
Linus Torvalds 已提交
6306
	return dev;
6307

6308 6309 6310 6311
free_all:
	free_netdev(dev);
	return NULL;

E
Eric Dumazet 已提交
6312 6313
free_pcpu:
	free_percpu(dev->pcpu_refcnt);
6314
	netif_free_tx_queues(dev);
T
Tom Herbert 已提交
6315 6316 6317 6318
#ifdef CONFIG_RPS
	kfree(dev->_rx);
#endif

6319 6320
free_dev:
	netdev_freemem(dev);
6321
	return NULL;
L
Linus Torvalds 已提交
6322
}
T
Tom Herbert 已提交
6323
EXPORT_SYMBOL(alloc_netdev_mqs);
L
Linus Torvalds 已提交
6324 6325 6326 6327 6328

/**
 *	free_netdev - free network device
 *	@dev: device
 *
6329 6330
 *	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
6331 6332 6333 6334
 *	If this is the last reference then it will be freed.
 */
void free_netdev(struct net_device *dev)
{
6335 6336
	struct napi_struct *p, *n;

6337 6338
	release_net(dev_net(dev));

6339
	netif_free_tx_queues(dev);
T
Tom Herbert 已提交
6340 6341 6342
#ifdef CONFIG_RPS
	kfree(dev->_rx);
#endif
6343

6344
	kfree(rcu_dereference_protected(dev->ingress_queue, 1));
6345

6346 6347 6348
	/* Flush device addresses */
	dev_addr_flush(dev);

6349 6350 6351
	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
		netif_napi_del(p);

E
Eric Dumazet 已提交
6352 6353 6354
	free_percpu(dev->pcpu_refcnt);
	dev->pcpu_refcnt = NULL;

S
Stephen Hemminger 已提交
6355
	/*  Compatibility with error handling in drivers */
L
Linus Torvalds 已提交
6356
	if (dev->reg_state == NETREG_UNINITIALIZED) {
6357
		netdev_freemem(dev);
L
Linus Torvalds 已提交
6358 6359 6360 6361 6362 6363
		return;
	}

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

6364 6365
	/* will free via device release */
	put_device(&dev->dev);
L
Linus Torvalds 已提交
6366
}
E
Eric Dumazet 已提交
6367
EXPORT_SYMBOL(free_netdev);
6368

6369 6370 6371 6372 6373 6374
/**
 *	synchronize_net -  Synchronize with packet receive processing
 *
 *	Wait for packets currently being received to be done.
 *	Does not block later packets from starting.
 */
6375
void synchronize_net(void)
L
Linus Torvalds 已提交
6376 6377
{
	might_sleep();
6378 6379 6380 6381
	if (rtnl_is_locked())
		synchronize_rcu_expedited();
	else
		synchronize_rcu();
L
Linus Torvalds 已提交
6382
}
E
Eric Dumazet 已提交
6383
EXPORT_SYMBOL(synchronize_net);
L
Linus Torvalds 已提交
6384 6385

/**
6386
 *	unregister_netdevice_queue - remove device from the kernel
L
Linus Torvalds 已提交
6387
 *	@dev: device
6388
 *	@head: list
6389
 *
L
Linus Torvalds 已提交
6390
 *	This function shuts down a device interface and removes it
6391
 *	from the kernel tables.
6392
 *	If head not NULL, device is queued to be unregistered later.
L
Linus Torvalds 已提交
6393 6394 6395 6396 6397
 *
 *	Callers must hold the rtnl semaphore.  You may want
 *	unregister_netdev() instead of this.
 */

6398
void unregister_netdevice_queue(struct net_device *dev, struct list_head *head)
L
Linus Torvalds 已提交
6399
{
6400 6401
	ASSERT_RTNL();

6402
	if (head) {
6403
		list_move_tail(&dev->unreg_list, head);
6404 6405 6406 6407 6408
	} else {
		rollback_registered(dev);
		/* Finish processing unregister after unlock */
		net_set_todo(dev);
	}
L
Linus Torvalds 已提交
6409
}
6410
EXPORT_SYMBOL(unregister_netdevice_queue);
L
Linus Torvalds 已提交
6411

6412 6413 6414 6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425
/**
 *	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);
	}
}
6426
EXPORT_SYMBOL(unregister_netdevice_many);
6427

L
Linus Torvalds 已提交
6428 6429 6430 6431 6432
/**
 *	unregister_netdev - remove device from the kernel
 *	@dev: device
 *
 *	This function shuts down a device interface and removes it
6433
 *	from the kernel tables.
L
Linus Torvalds 已提交
6434 6435 6436 6437 6438 6439 6440 6441 6442 6443 6444 6445 6446
 *
 *	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);

6447 6448 6449 6450 6451 6452 6453 6454 6455 6456 6457 6458 6459 6460 6461 6462 6463 6464 6465 6466 6467 6468 6469 6470 6471 6472 6473 6474 6475 6476 6477
/**
 *	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;
6478
	if (net_eq(dev_net(dev), net))
6479 6480 6481 6482 6483 6484
		goto out;

	/* Pick the destination device name, and ensure
	 * we can use it in the destination network namespace.
	 */
	err = -EEXIST;
6485
	if (__dev_get_by_name(net, dev->name)) {
6486 6487 6488
		/* We get here if we can't use the current device name */
		if (!pat)
			goto out;
6489
		if (dev_get_valid_name(net, dev, pat) < 0)
6490 6491 6492 6493 6494 6495 6496 6497
			goto out;
	}

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

	/* If device is running close it first. */
6498
	dev_close(dev);
6499 6500 6501 6502 6503 6504 6505 6506 6507 6508 6509 6510

	/* 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.
6511 6512 6513 6514

	   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.
6515 6516
	*/
	call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
6517 6518
	rcu_barrier();
	call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6519
	rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);
6520 6521 6522 6523

	/*
	 *	Flush the unicast and multicast chains
	 */
6524
	dev_uc_flush(dev);
6525
	dev_mc_flush(dev);
6526

6527 6528 6529
	/* Send a netdev-removed uevent to the old namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_REMOVE);

6530
	/* Actually switch the network namespace */
6531
	dev_net_set(dev, net);
6532 6533 6534 6535 6536 6537 6538 6539 6540

	/* 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;
	}

6541 6542 6543
	/* Send a netdev-add uevent to the new namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_ADD);

6544
	/* Fixup kobjects */
6545
	err = device_rename(&dev->dev, dev->name);
6546
	WARN_ON(err);
6547 6548 6549 6550 6551 6552 6553

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

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

6554 6555 6556 6557
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
6558
	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
6559

6560 6561 6562 6563 6564
	synchronize_net();
	err = 0;
out:
	return err;
}
6565
EXPORT_SYMBOL_GPL(dev_change_net_namespace);
6566

L
Linus Torvalds 已提交
6567 6568 6569 6570 6571 6572 6573 6574 6575
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;

6576
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
L
Linus Torvalds 已提交
6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590 6591 6592
		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. */
6593 6594 6595 6596 6597 6598
	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;
	}
6599 6600 6601 6602 6603
	/* 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
6604 6605 6606 6607 6608

	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_enable();

	/* Process offline CPU's input_pkt_queue */
6609
	while ((skb = __skb_dequeue(&oldsd->process_queue))) {
L
Linus Torvalds 已提交
6610
		netif_rx(skb);
6611
		input_queue_head_incr(oldsd);
T
Tom Herbert 已提交
6612
	}
6613
	while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) {
6614
		netif_rx(skb);
6615 6616
		input_queue_head_incr(oldsd);
	}
L
Linus Torvalds 已提交
6617 6618 6619 6620 6621

	return NOTIFY_OK;
}


6622
/**
6623 6624 6625 6626
 *	netdev_increment_features - increment feature set by one
 *	@all: current feature set
 *	@one: new feature set
 *	@mask: mask feature set
6627 6628
 *
 *	Computes a new feature set after adding a device with feature set
6629 6630
 *	@one to the master device with current feature set @all.  Will not
 *	enable anything that is off in @mask. Returns the new feature set.
6631
 */
6632 6633
netdev_features_t netdev_increment_features(netdev_features_t all,
	netdev_features_t one, netdev_features_t mask)
6634
{
6635 6636 6637
	if (mask & NETIF_F_GEN_CSUM)
		mask |= NETIF_F_ALL_CSUM;
	mask |= NETIF_F_VLAN_CHALLENGED;
6638

6639 6640
	all |= one & (NETIF_F_ONE_FOR_ALL|NETIF_F_ALL_CSUM) & mask;
	all &= one | ~NETIF_F_ALL_FOR_ALL;
6641

6642 6643 6644
	/* If one device supports hw checksumming, set for all. */
	if (all & NETIF_F_GEN_CSUM)
		all &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
6645 6646 6647

	return all;
}
6648
EXPORT_SYMBOL(netdev_increment_features);
6649

6650
static struct hlist_head * __net_init netdev_create_hash(void)
6651 6652 6653 6654 6655 6656 6657 6658 6659 6660 6661 6662
{
	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;
}

6663
/* Initialize per network namespace state */
6664
static int __net_init netdev_init(struct net *net)
6665
{
6666 6667
	if (net != &init_net)
		INIT_LIST_HEAD(&net->dev_base_head);
6668

6669 6670 6671
	net->dev_name_head = netdev_create_hash();
	if (net->dev_name_head == NULL)
		goto err_name;
6672

6673 6674 6675
	net->dev_index_head = netdev_create_hash();
	if (net->dev_index_head == NULL)
		goto err_idx;
6676 6677

	return 0;
6678 6679 6680 6681 6682

err_idx:
	kfree(net->dev_name_head);
err_name:
	return -ENOMEM;
6683 6684
}

6685 6686 6687 6688 6689 6690
/**
 *	netdev_drivername - network driver for the device
 *	@dev: network device
 *
 *	Determine network driver for device.
 */
6691
const char *netdev_drivername(const struct net_device *dev)
6692
{
6693 6694
	const struct device_driver *driver;
	const struct device *parent;
6695
	const char *empty = "";
6696 6697 6698

	parent = dev->dev.parent;
	if (!parent)
6699
		return empty;
6700 6701 6702

	driver = parent->driver;
	if (driver && driver->name)
6703 6704
		return driver->name;
	return empty;
6705 6706
}

6707
static int __netdev_printk(const char *level, const struct net_device *dev,
6708 6709 6710 6711
			   struct va_format *vaf)
{
	int r;

6712
	if (dev && dev->dev.parent) {
6713 6714 6715 6716 6717 6718
		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);
6719
	} else if (dev) {
6720
		r = printk("%s%s: %pV", level, netdev_name(dev), vaf);
6721
	} else {
6722
		r = printk("%s(NULL net_device): %pV", level, vaf);
6723
	}
6724 6725 6726 6727 6728 6729 6730 6731 6732 6733 6734 6735 6736 6737 6738 6739 6740

	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);
6741

6742 6743 6744 6745 6746 6747 6748 6749 6750 6751 6752 6753 6754 6755 6756 6757 6758 6759 6760
	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);			\
6761
								\
6762 6763 6764 6765 6766 6767 6768 6769 6770 6771 6772 6773 6774 6775
	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);

6776
static void __net_exit netdev_exit(struct net *net)
6777 6778 6779 6780 6781
{
	kfree(net->dev_name_head);
	kfree(net->dev_index_head);
}

6782
static struct pernet_operations __net_initdata netdev_net_ops = {
6783 6784 6785 6786
	.init = netdev_init,
	.exit = netdev_exit,
};

6787
static void __net_exit default_device_exit(struct net *net)
6788
{
6789
	struct net_device *dev, *aux;
6790
	/*
6791
	 * Push all migratable network devices back to the
6792 6793 6794
	 * initial network namespace
	 */
	rtnl_lock();
6795
	for_each_netdev_safe(net, dev, aux) {
6796
		int err;
6797
		char fb_name[IFNAMSIZ];
6798 6799 6800 6801 6802

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

6803 6804 6805
		/* Leave virtual devices for the generic cleanup */
		if (dev->rtnl_link_ops)
			continue;
6806

L
Lucas De Marchi 已提交
6807
		/* Push remaining network devices to init_net */
6808 6809
		snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex);
		err = dev_change_net_namespace(dev, &init_net, fb_name);
6810
		if (err) {
6811 6812
			pr_emerg("%s: failed to move %s to init_net: %d\n",
				 __func__, dev->name, err);
6813
			BUG();
6814 6815 6816 6817 6818
		}
	}
	rtnl_unlock();
}

6819 6820 6821 6822 6823 6824 6825 6826 6827 6828 6829 6830 6831 6832 6833 6834 6835 6836 6837 6838 6839 6840 6841 6842 6843 6844 6845 6846
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);
}

6847 6848 6849
static void __net_exit default_device_exit_batch(struct list_head *net_list)
{
	/* At exit all network devices most be removed from a network
6850
	 * namespace.  Do this in the reverse order of registration.
6851 6852 6853 6854 6855 6856 6857
	 * 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);

6858 6859 6860 6861 6862 6863 6864 6865 6866 6867 6868 6869
	/* 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);
6870 6871 6872 6873 6874 6875 6876 6877 6878
	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
6879
	list_del(&dev_kill_list);
6880 6881 6882
	rtnl_unlock();
}

6883
static struct pernet_operations __net_initdata default_device_ops = {
6884
	.exit = default_device_exit,
6885
	.exit_batch = default_device_exit_batch,
6886 6887
};

L
Linus Torvalds 已提交
6888 6889 6890 6891 6892 6893 6894 6895 6896 6897 6898 6899 6900 6901 6902 6903 6904 6905 6906 6907
/*
 *	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;

6908
	if (netdev_kobject_init())
L
Linus Torvalds 已提交
6909 6910 6911
		goto out;

	INIT_LIST_HEAD(&ptype_all);
6912
	for (i = 0; i < PTYPE_HASH_SIZE; i++)
L
Linus Torvalds 已提交
6913 6914
		INIT_LIST_HEAD(&ptype_base[i]);

6915 6916
	INIT_LIST_HEAD(&offload_base);

6917 6918
	if (register_pernet_subsys(&netdev_net_ops))
		goto out;
L
Linus Torvalds 已提交
6919 6920 6921 6922 6923

	/*
	 *	Initialise the packet receive queues.
	 */

6924
	for_each_possible_cpu(i) {
E
Eric Dumazet 已提交
6925
		struct softnet_data *sd = &per_cpu(softnet_data, i);
L
Linus Torvalds 已提交
6926

C
Changli Gao 已提交
6927
		memset(sd, 0, sizeof(*sd));
E
Eric Dumazet 已提交
6928
		skb_queue_head_init(&sd->input_pkt_queue);
6929
		skb_queue_head_init(&sd->process_queue);
E
Eric Dumazet 已提交
6930 6931
		sd->completion_queue = NULL;
		INIT_LIST_HEAD(&sd->poll_list);
6932 6933
		sd->output_queue = NULL;
		sd->output_queue_tailp = &sd->output_queue;
E
Eric Dumazet 已提交
6934
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
6935 6936 6937 6938
		sd->csd.func = rps_trigger_softirq;
		sd->csd.info = sd;
		sd->csd.flags = 0;
		sd->cpu = i;
6939
#endif
T
Tom Herbert 已提交
6940

E
Eric Dumazet 已提交
6941 6942 6943 6944
		sd->backlog.poll = process_backlog;
		sd->backlog.weight = weight_p;
		sd->backlog.gro_list = NULL;
		sd->backlog.gro_count = 0;
6945 6946 6947 6948

#ifdef CONFIG_NET_FLOW_LIMIT
		sd->flow_limit = NULL;
#endif
L
Linus Torvalds 已提交
6949 6950 6951 6952
	}

	dev_boot_phase = 0;

6953 6954 6955 6956 6957 6958 6959 6960 6961 6962 6963 6964 6965 6966 6967
	/* 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;

6968 6969
	open_softirq(NET_TX_SOFTIRQ, net_tx_action);
	open_softirq(NET_RX_SOFTIRQ, net_rx_action);
L
Linus Torvalds 已提交
6970 6971 6972 6973 6974 6975 6976 6977 6978

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

subsys_initcall(net_dev_init);