dev.c 182.5 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>
135
#include <linux/errqueue.h>
L
Linus Torvalds 已提交
136

137 138
#include "net-sysfs.h"

139 140 141
/* Instead of increasing this, you should create a hash table. */
#define MAX_GRO_SKBS 8

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

151
static int netif_rx_internal(struct sk_buff *skb);
152 153 154
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info);
155

L
Linus Torvalds 已提交
156
/*
157
 * The @dev_base_head list is protected by @dev_base_lock and the rtnl
L
Linus Torvalds 已提交
158 159
 * semaphore.
 *
160
 * Pure readers hold dev_base_lock for reading, or rcu_read_lock()
L
Linus Torvalds 已提交
161 162
 *
 * Writers must hold the rtnl semaphore while they loop through the
163
 * dev_base_head list, and hold dev_base_lock for writing when they do the
L
Linus Torvalds 已提交
164 165 166 167 168 169 170 171 172 173 174 175 176 177
 * 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
178 179 180 181 182 183
/* 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);

184
static seqcount_t devnet_rename_seq;
185

186 187 188 189 190
static inline void dev_base_seq_inc(struct net *net)
{
	while (++net->dev_base_seq == 0);
}

191
static inline struct hlist_head *dev_name_hash(struct net *net, const char *name)
L
Linus Torvalds 已提交
192
{
193 194
	unsigned int hash = full_name_hash(name, strnlen(name, IFNAMSIZ));

195
	return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)];
L
Linus Torvalds 已提交
196 197
}

198
static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex)
L
Linus Torvalds 已提交
199
{
200
	return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)];
L
Linus Torvalds 已提交
201 202
}

E
Eric Dumazet 已提交
203
static inline void rps_lock(struct softnet_data *sd)
204 205
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
206
	spin_lock(&sd->input_pkt_queue.lock);
207 208 209
#endif
}

E
Eric Dumazet 已提交
210
static inline void rps_unlock(struct softnet_data *sd)
211 212
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
213
	spin_unlock(&sd->input_pkt_queue.lock);
214 215 216
#endif
}

217
/* Device list insertion */
218
static void list_netdevice(struct net_device *dev)
219
{
220
	struct net *net = dev_net(dev);
221 222 223 224

	ASSERT_RTNL();

	write_lock_bh(&dev_base_lock);
225
	list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
226
	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
227 228
	hlist_add_head_rcu(&dev->index_hlist,
			   dev_index_hash(net, dev->ifindex));
229
	write_unlock_bh(&dev_base_lock);
230 231

	dev_base_seq_inc(net);
232 233
}

234 235 236
/* Device list removal
 * caller must respect a RCU grace period before freeing/reusing dev
 */
237 238 239 240 241 242
static void unlist_netdevice(struct net_device *dev)
{
	ASSERT_RTNL();

	/* Unlink dev from the device chain */
	write_lock_bh(&dev_base_lock);
243
	list_del_rcu(&dev->dev_list);
244
	hlist_del_rcu(&dev->name_hlist);
245
	hlist_del_rcu(&dev->index_hlist);
246
	write_unlock_bh(&dev_base_lock);
247 248

	dev_base_seq_inc(dev_net(dev));
249 250
}

L
Linus Torvalds 已提交
251 252 253 254
/*
 *	Our notifier list
 */

255
static RAW_NOTIFIER_HEAD(netdev_chain);
L
Linus Torvalds 已提交
256 257 258 259 260

/*
 *	Device drivers call our routines to queue packets here. We empty the
 *	queue in the local softnet handler.
 */
261

262
DEFINE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
E
Eric Dumazet 已提交
263
EXPORT_PER_CPU_SYMBOL(softnet_data);
L
Linus Torvalds 已提交
264

265
#ifdef CONFIG_LOCKDEP
266
/*
267
 * register_netdevice() inits txq->_xmit_lock and sets lockdep class
268 269 270 271 272 273 274 275 276 277 278 279 280 281 282
 * 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,
283 284 285
	 ARPHRD_FCFABRIC, ARPHRD_IEEE80211, ARPHRD_IEEE80211_PRISM,
	 ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, ARPHRD_PHONET_PIPE,
	 ARPHRD_IEEE802154, ARPHRD_VOID, ARPHRD_NONE};
286

287
static const char *const netdev_lock_name[] =
288 289 290 291 292 293 294 295 296 297 298 299
	{"_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",
300 301 302
	 "_xmit_FCFABRIC", "_xmit_IEEE80211", "_xmit_IEEE80211_PRISM",
	 "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", "_xmit_PHONET_PIPE",
	 "_xmit_IEEE802154", "_xmit_VOID", "_xmit_NONE"};
303 304

static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)];
305
static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)];
306 307 308 309 310 311 312 313 314 315 316 317

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;
}

318 319
static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
						 unsigned short dev_type)
320 321 322 323 324 325 326
{
	int i;

	i = netdev_lock_pos(dev_type);
	lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i],
				   netdev_lock_name[i]);
}
327 328 329 330 331 332 333 334 335 336

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]);
}
337
#else
338 339 340 341 342
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)
343 344 345
{
}
#endif
L
Linus Torvalds 已提交
346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368

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

		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)
 */

369 370 371 372 373 374 375 376
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
377 378 379 380 381 382 383 384
/**
 *	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.
 *
385
 *	This call does not sleep therefore it can not
L
Linus Torvalds 已提交
386 387 388 389 390 391
 *	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)
{
392
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
393

394 395 396
	spin_lock(&ptype_lock);
	list_add_rcu(&pt->list, head);
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
397
}
E
Eric Dumazet 已提交
398
EXPORT_SYMBOL(dev_add_pack);
L
Linus Torvalds 已提交
399 400 401 402 403 404 405 406

/**
 *	__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
407
 *	returns.
L
Linus Torvalds 已提交
408 409 410 411 412 413 414
 *
 *      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)
{
415
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
416 417
	struct packet_type *pt1;

418
	spin_lock(&ptype_lock);
L
Linus Torvalds 已提交
419 420 421 422 423 424 425 426

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

427
	pr_warn("dev_remove_pack: %p not found\n", pt);
L
Linus Torvalds 已提交
428
out:
429
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
430
}
E
Eric Dumazet 已提交
431 432
EXPORT_SYMBOL(__dev_remove_pack);

L
Linus Torvalds 已提交
433 434 435 436 437 438 439 440 441 442 443 444 445 446 447
/**
 *	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);
448

L
Linus Torvalds 已提交
449 450
	synchronize_net();
}
E
Eric Dumazet 已提交
451
EXPORT_SYMBOL(dev_remove_pack);
L
Linus Torvalds 已提交
452

453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488

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

	spin_lock(&offload_lock);
	list_add_rcu(&po->list, head);
	spin_unlock(&offload_lock);
}
EXPORT_SYMBOL(dev_add_offload);

/**
 *	__dev_remove_offload	 - remove offload handler
 *	@po: packet offload declaration
 *
 *	Remove a protocol offload handler that was previously added to the
 *	kernel offload handlers by dev_add_offload(). The passed &offload_type
 *	is removed from the kernel lists and can be freed or reused once this
 *	function returns.
 *
 *      The packet type might still be in use by receivers
 *	and must not be freed until after all the CPU's have gone
 *	through a quiescent state.
 */
S
stephen hemminger 已提交
489
static void __dev_remove_offload(struct packet_offload *po)
490 491 492 493
{
	struct list_head *head = &offload_base;
	struct packet_offload *po1;

494
	spin_lock(&offload_lock);
495 496 497 498 499 500 501 502 503 504

	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:
505
	spin_unlock(&offload_lock);
506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527
}

/**
 *	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
528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554
/******************************************************************************

		      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));
555
			strlcpy(s[i].name, name, IFNAMSIZ);
L
Linus Torvalds 已提交
556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579
			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] != ' ' &&
580
		    !strcmp(dev->name, s[i].name)) {
L
Linus Torvalds 已提交
581 582 583 584 585 586 587 588 589
			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
590
EXPORT_SYMBOL(netdev_boot_setup_check);
L
Linus Torvalds 已提交
591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614


/**
 *	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
	 */
615
	if (__dev_get_by_name(&init_net, name))
L
Linus Torvalds 已提交
616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660
		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
661
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
662 663 664 665 666 667 668 669 670
 *	@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.
 */

671
struct net_device *__dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
672
{
E
Eric Dumazet 已提交
673 674
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);
L
Linus Torvalds 已提交
675

676
	hlist_for_each_entry(dev, head, name_hlist)
L
Linus Torvalds 已提交
677 678
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;
E
Eric Dumazet 已提交
679

L
Linus Torvalds 已提交
680 681
	return NULL;
}
E
Eric Dumazet 已提交
682
EXPORT_SYMBOL(__dev_get_by_name);
L
Linus Torvalds 已提交
683

684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700
/**
 *	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);

701
	hlist_for_each_entry_rcu(dev, head, name_hlist)
702 703 704 705 706 707 708
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_name_rcu);

L
Linus Torvalds 已提交
709 710
/**
 *	dev_get_by_name		- find a device by its name
711
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
712 713 714 715 716 717 718 719 720
 *	@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.
 */

721
struct net_device *dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
722 723 724
{
	struct net_device *dev;

725 726
	rcu_read_lock();
	dev = dev_get_by_name_rcu(net, name);
L
Linus Torvalds 已提交
727 728
	if (dev)
		dev_hold(dev);
729
	rcu_read_unlock();
L
Linus Torvalds 已提交
730 731
	return dev;
}
E
Eric Dumazet 已提交
732
EXPORT_SYMBOL(dev_get_by_name);
L
Linus Torvalds 已提交
733 734 735

/**
 *	__dev_get_by_index - find a device by its ifindex
736
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
737 738 739 740 741 742 743 744 745
 *	@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.
 */

746
struct net_device *__dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
747
{
E
Eric Dumazet 已提交
748 749
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);
L
Linus Torvalds 已提交
750

751
	hlist_for_each_entry(dev, head, index_hlist)
L
Linus Torvalds 已提交
752 753
		if (dev->ifindex == ifindex)
			return dev;
E
Eric Dumazet 已提交
754

L
Linus Torvalds 已提交
755 756
	return NULL;
}
E
Eric Dumazet 已提交
757
EXPORT_SYMBOL(__dev_get_by_index);
L
Linus Torvalds 已提交
758

759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774
/**
 *	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);

775
	hlist_for_each_entry_rcu(dev, head, index_hlist)
776 777 778 779 780 781 782
		if (dev->ifindex == ifindex)
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_index_rcu);

L
Linus Torvalds 已提交
783 784 785

/**
 *	dev_get_by_index - find a device by its ifindex
786
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
787 788 789 790 791 792 793 794
 *	@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.
 */

795
struct net_device *dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
796 797 798
{
	struct net_device *dev;

799 800
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, ifindex);
L
Linus Torvalds 已提交
801 802
	if (dev)
		dev_hold(dev);
803
	rcu_read_unlock();
L
Linus Torvalds 已提交
804 805
	return dev;
}
E
Eric Dumazet 已提交
806
EXPORT_SYMBOL(dev_get_by_index);
L
Linus Torvalds 已提交
807

808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841
/**
 *	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
842
/**
843
 *	dev_getbyhwaddr_rcu - find a device by its hardware address
844
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
845 846 847 848
 *	@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
849 850
 *	is not found or a pointer to the device.
 *	The caller must hold RCU or RTNL.
851
 *	The returned device has not had its ref count increased
L
Linus Torvalds 已提交
852 853 854 855
 *	and the caller must therefore be careful about locking
 *
 */

856 857
struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
				       const char *ha)
L
Linus Torvalds 已提交
858 859 860
{
	struct net_device *dev;

861
	for_each_netdev_rcu(net, dev)
L
Linus Torvalds 已提交
862 863
		if (dev->type == type &&
		    !memcmp(dev->dev_addr, ha, dev->addr_len))
864 865 866
			return dev;

	return NULL;
L
Linus Torvalds 已提交
867
}
868
EXPORT_SYMBOL(dev_getbyhwaddr_rcu);
869

870
struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type)
L
Linus Torvalds 已提交
871 872 873
{
	struct net_device *dev;

874
	ASSERT_RTNL();
875
	for_each_netdev(net, dev)
876
		if (dev->type == type)
877 878 879
			return dev;

	return NULL;
880 881 882
}
EXPORT_SYMBOL(__dev_getfirstbyhwtype);

883
struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type)
884
{
885
	struct net_device *dev, *ret = NULL;
886

887 888 889 890 891 892 893 894 895
	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
896 897 898 899
}
EXPORT_SYMBOL(dev_getfirstbyhwtype);

/**
900
 *	__dev_get_by_flags - find any device with given flags
901
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
902 903 904 905
 *	@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
906
 *	is not found or a pointer to the device. Must be called inside
907
 *	rtnl_lock(), and result refcount is unchanged.
L
Linus Torvalds 已提交
908 909
 */

910 911
struct net_device *__dev_get_by_flags(struct net *net, unsigned short if_flags,
				      unsigned short mask)
L
Linus Torvalds 已提交
912
{
913
	struct net_device *dev, *ret;
L
Linus Torvalds 已提交
914

915 916
	ASSERT_RTNL();

917
	ret = NULL;
918
	for_each_netdev(net, dev) {
L
Linus Torvalds 已提交
919
		if (((dev->flags ^ if_flags) & mask) == 0) {
920
			ret = dev;
L
Linus Torvalds 已提交
921 922 923
			break;
		}
	}
924
	return ret;
L
Linus Torvalds 已提交
925
}
926
EXPORT_SYMBOL(__dev_get_by_flags);
L
Linus Torvalds 已提交
927 928 929 930 931 932

/**
 *	dev_valid_name - check if name is okay for network device
 *	@name: name string
 *
 *	Network device names need to be valid file names to
933 934
 *	to allow sysfs to work.  We also disallow any kind of
 *	whitespace.
L
Linus Torvalds 已提交
935
 */
936
bool dev_valid_name(const char *name)
L
Linus Torvalds 已提交
937
{
938
	if (*name == '\0')
939
		return false;
940
	if (strlen(name) >= IFNAMSIZ)
941
		return false;
942
	if (!strcmp(name, ".") || !strcmp(name, ".."))
943
		return false;
944 945 946

	while (*name) {
		if (*name == '/' || isspace(*name))
947
			return false;
948 949
		name++;
	}
950
	return true;
L
Linus Torvalds 已提交
951
}
E
Eric Dumazet 已提交
952
EXPORT_SYMBOL(dev_valid_name);
L
Linus Torvalds 已提交
953 954

/**
955 956
 *	__dev_alloc_name - allocate a name for a device
 *	@net: network namespace to allocate the device name in
L
Linus Torvalds 已提交
957
 *	@name: name format string
958
 *	@buf:  scratch buffer and result name string
L
Linus Torvalds 已提交
959 960
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
S
Stephen Hemminger 已提交
961 962 963 964 965 966
 *	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
967 968
 */

969
static int __dev_alloc_name(struct net *net, const char *name, char *buf)
L
Linus Torvalds 已提交
970 971 972 973
{
	int i = 0;
	const char *p;
	const int max_netdevices = 8*PAGE_SIZE;
S
Stephen Hemminger 已提交
974
	unsigned long *inuse;
L
Linus Torvalds 已提交
975 976 977 978 979 980 981 982 983 984 985 986 987
	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
988
		inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC);
L
Linus Torvalds 已提交
989 990 991
		if (!inuse)
			return -ENOMEM;

992
		for_each_netdev(net, d) {
L
Linus Torvalds 已提交
993 994 995 996 997 998
			if (!sscanf(d->name, name, &i))
				continue;
			if (i < 0 || i >= max_netdevices)
				continue;

			/*  avoid cases where sscanf is not exact inverse of printf */
999
			snprintf(buf, IFNAMSIZ, name, i);
L
Linus Torvalds 已提交
1000 1001 1002 1003 1004 1005 1006 1007
			if (!strncmp(buf, d->name, IFNAMSIZ))
				set_bit(i, inuse);
		}

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

1008 1009
	if (buf != name)
		snprintf(buf, IFNAMSIZ, name, i);
1010
	if (!__dev_get_by_name(net, buf))
L
Linus Torvalds 已提交
1011 1012 1013 1014 1015 1016 1017 1018 1019
		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;
}

1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039
/**
 *	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;

1040 1041
	BUG_ON(!dev_net(dev));
	net = dev_net(dev);
1042 1043 1044 1045 1046
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
}
E
Eric Dumazet 已提交
1047
EXPORT_SYMBOL(dev_alloc_name);
1048

1049 1050 1051
static int dev_alloc_name_ns(struct net *net,
			     struct net_device *dev,
			     const char *name)
1052
{
1053 1054
	char buf[IFNAMSIZ];
	int ret;
1055

1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
	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);
1067

1068 1069 1070
	if (!dev_valid_name(name))
		return -EINVAL;

1071
	if (strchr(name, '%'))
1072
		return dev_alloc_name_ns(net, dev, name);
1073 1074
	else if (__dev_get_by_name(net, name))
		return -EEXIST;
1075 1076
	else if (dev->name != name)
		strlcpy(dev->name, name, IFNAMSIZ);
1077 1078 1079

	return 0;
}
L
Linus Torvalds 已提交
1080 1081 1082 1083 1084 1085 1086 1087 1088

/**
 *	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.
 */
1089
int dev_change_name(struct net_device *dev, const char *newname)
L
Linus Torvalds 已提交
1090
{
1091
	unsigned char old_assign_type;
1092
	char oldname[IFNAMSIZ];
L
Linus Torvalds 已提交
1093
	int err = 0;
1094
	int ret;
1095
	struct net *net;
L
Linus Torvalds 已提交
1096 1097

	ASSERT_RTNL();
1098
	BUG_ON(!dev_net(dev));
L
Linus Torvalds 已提交
1099

1100
	net = dev_net(dev);
L
Linus Torvalds 已提交
1101 1102 1103
	if (dev->flags & IFF_UP)
		return -EBUSY;

1104
	write_seqcount_begin(&devnet_rename_seq);
1105 1106

	if (strncmp(newname, dev->name, IFNAMSIZ) == 0) {
1107
		write_seqcount_end(&devnet_rename_seq);
1108
		return 0;
1109
	}
1110

1111 1112
	memcpy(oldname, dev->name, IFNAMSIZ);

1113
	err = dev_get_valid_name(net, dev, newname);
1114
	if (err < 0) {
1115
		write_seqcount_end(&devnet_rename_seq);
1116
		return err;
1117
	}
L
Linus Torvalds 已提交
1118

1119 1120 1121
	if (oldname[0] && !strchr(oldname, '%'))
		netdev_info(dev, "renamed from %s\n", oldname);

1122 1123 1124
	old_assign_type = dev->name_assign_type;
	dev->name_assign_type = NET_NAME_RENAMED;

1125
rollback:
1126 1127 1128
	ret = device_rename(&dev->dev, dev->name);
	if (ret) {
		memcpy(dev->name, oldname, IFNAMSIZ);
1129
		dev->name_assign_type = old_assign_type;
1130
		write_seqcount_end(&devnet_rename_seq);
1131
		return ret;
1132
	}
1133

1134
	write_seqcount_end(&devnet_rename_seq);
1135

1136 1137
	netdev_adjacent_rename_links(dev, oldname);

1138
	write_lock_bh(&dev_base_lock);
1139
	hlist_del_rcu(&dev->name_hlist);
1140 1141 1142 1143 1144 1145
	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));
1146 1147
	write_unlock_bh(&dev_base_lock);

1148
	ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
1149 1150 1151
	ret = notifier_to_errno(ret);

	if (ret) {
1152 1153
		/* err >= 0 after dev_alloc_name() or stores the first errno */
		if (err >= 0) {
1154
			err = ret;
1155
			write_seqcount_begin(&devnet_rename_seq);
1156
			memcpy(dev->name, oldname, IFNAMSIZ);
1157
			memcpy(oldname, newname, IFNAMSIZ);
1158 1159
			dev->name_assign_type = old_assign_type;
			old_assign_type = NET_NAME_RENAMED;
1160
			goto rollback;
1161
		} else {
1162
			pr_err("%s: name change rollback failed: %d\n",
1163
			       dev->name, ret);
1164 1165
		}
	}
L
Linus Torvalds 已提交
1166 1167 1168 1169

	return err;
}

1170 1171 1172 1173
/**
 *	dev_set_alias - change ifalias of a device
 *	@dev: device
 *	@alias: name up to IFALIASZ
1174
 *	@len: limit of bytes to copy from info
1175 1176 1177 1178 1179
 *
 *	Set ifalias for a device,
 */
int dev_set_alias(struct net_device *dev, const char *alias, size_t len)
{
1180 1181
	char *new_ifalias;

1182 1183 1184 1185 1186
	ASSERT_RTNL();

	if (len >= IFALIASZ)
		return -EINVAL;

1187
	if (!len) {
1188 1189
		kfree(dev->ifalias);
		dev->ifalias = NULL;
1190 1191 1192
		return 0;
	}

1193 1194
	new_ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL);
	if (!new_ifalias)
1195
		return -ENOMEM;
1196
	dev->ifalias = new_ifalias;
1197 1198 1199 1200 1201 1202

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


1203
/**
S
Stephen Hemminger 已提交
1204
 *	netdev_features_change - device changes features
1205 1206 1207 1208 1209 1210
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed features.
 */
void netdev_features_change(struct net_device *dev)
{
1211
	call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev);
1212 1213 1214
}
EXPORT_SYMBOL(netdev_features_change);

L
Linus Torvalds 已提交
1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225
/**
 *	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) {
1226 1227 1228 1229 1230
		struct netdev_notifier_change_info change_info;

		change_info.flags_changed = 0;
		call_netdevice_notifiers_info(NETDEV_CHANGE, dev,
					      &change_info.info);
1231
		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
L
Linus Torvalds 已提交
1232 1233
	}
}
E
Eric Dumazet 已提交
1234
EXPORT_SYMBOL(netdev_state_change);
L
Linus Torvalds 已提交
1235

1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246
/**
 * 	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)
1247
{
1248 1249 1250
	rtnl_lock();
	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, dev);
	rtnl_unlock();
1251
}
1252
EXPORT_SYMBOL(netdev_notify_peers);
1253

1254
static int __dev_open(struct net_device *dev)
L
Linus Torvalds 已提交
1255
{
1256
	const struct net_device_ops *ops = dev->netdev_ops;
1257
	int ret;
L
Linus Torvalds 已提交
1258

1259 1260
	ASSERT_RTNL();

L
Linus Torvalds 已提交
1261 1262 1263
	if (!netif_device_present(dev))
		return -ENODEV;

1264 1265 1266 1267
	/* 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
	 */
1268
	netpoll_poll_disable(dev);
1269

1270 1271 1272 1273 1274
	ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

L
Linus Torvalds 已提交
1275
	set_bit(__LINK_STATE_START, &dev->state);
1276

1277 1278
	if (ops->ndo_validate_addr)
		ret = ops->ndo_validate_addr(dev);
1279

1280 1281
	if (!ret && ops->ndo_open)
		ret = ops->ndo_open(dev);
L
Linus Torvalds 已提交
1282

1283
	netpoll_poll_enable(dev);
1284

1285 1286 1287
	if (ret)
		clear_bit(__LINK_STATE_START, &dev->state);
	else {
L
Linus Torvalds 已提交
1288
		dev->flags |= IFF_UP;
1289
		net_dmaengine_get();
1290
		dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
1291
		dev_activate(dev);
1292
		add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
1293
	}
1294

L
Linus Torvalds 已提交
1295 1296 1297 1298
	return ret;
}

/**
1299 1300
 *	dev_open	- prepare an interface for use.
 *	@dev:	device to open
L
Linus Torvalds 已提交
1301
 *
1302 1303 1304 1305 1306 1307 1308
 *	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
1309
 */
1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320
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;

1321
	rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1322 1323 1324 1325 1326 1327
	call_netdevice_notifiers(NETDEV_UP, dev);

	return ret;
}
EXPORT_SYMBOL(dev_open);

1328
static int __dev_close_many(struct list_head *head)
L
Linus Torvalds 已提交
1329
{
1330
	struct net_device *dev;
1331

1332
	ASSERT_RTNL();
1333 1334
	might_sleep();

1335
	list_for_each_entry(dev, head, close_list) {
1336
		/* Temporarily disable netpoll until the interface is down */
1337
		netpoll_poll_disable(dev);
1338

1339
		call_netdevice_notifiers(NETDEV_GOING_DOWN, dev);
L
Linus Torvalds 已提交
1340

1341
		clear_bit(__LINK_STATE_START, &dev->state);
L
Linus Torvalds 已提交
1342

1343 1344 1345 1346 1347 1348
		/* 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.
		 */
1349
		smp_mb__after_atomic(); /* Commit netif_running(). */
1350
	}
L
Linus Torvalds 已提交
1351

1352
	dev_deactivate_many(head);
1353

1354
	list_for_each_entry(dev, head, close_list) {
1355
		const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
1356

1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368
		/*
		 *	Call the device specific close. This cannot fail.
		 *	Only if device is UP
		 *
		 *	We allow it to be called even after a DETACH hot-plug
		 *	event.
		 */
		if (ops->ndo_stop)
			ops->ndo_stop(dev);

		dev->flags &= ~IFF_UP;
		net_dmaengine_put();
1369
		netpoll_poll_enable(dev);
1370 1371 1372 1373 1374 1375 1376
	}

	return 0;
}

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

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

1384
	return retval;
1385 1386
}

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

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

	__dev_close_many(head);
L
Linus Torvalds 已提交
1397

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

	return 0;
}

/**
 *	dev_close - shutdown an interface.
 *	@dev: device to shutdown
 *
 *	This function moves an active device into down state. A
 *	%NETDEV_GOING_DOWN is sent to the netdev notifier chain. The device
 *	is then deactivated and finally a %NETDEV_DOWN is sent to the notifier
 *	chain.
 */
int dev_close(struct net_device *dev)
{
1418 1419
	if (dev->flags & IFF_UP) {
		LIST_HEAD(single);
L
Linus Torvalds 已提交
1420

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


1430 1431 1432 1433 1434 1435 1436 1437 1438 1439
/**
 *	dev_disable_lro - disable Large Receive Offload on a device
 *	@dev: device
 *
 *	Disable Large Receive Offload (LRO) on a net device.  Must be
 *	called under RTNL.  This is needed if received packets may be
 *	forwarded to another interface.
 */
void dev_disable_lro(struct net_device *dev)
{
1440 1441 1442 1443 1444 1445 1446
	/*
	 * 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);

1447 1448 1449 1450
	/* the same for macvlan devices */
	if (netif_is_macvlan(dev))
		dev = macvlan_dev_real_dev(dev);

M
Michał Mirosław 已提交
1451 1452
	dev->wanted_features &= ~NETIF_F_LRO;
	netdev_update_features(dev);
1453

1454 1455
	if (unlikely(dev->features & NETIF_F_LRO))
		netdev_WARN(dev, "failed to disable LRO!\n");
1456 1457 1458
}
EXPORT_SYMBOL(dev_disable_lro);

1459 1460 1461 1462 1463 1464 1465 1466
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);
}
1467

1468 1469
static int dev_boot_phase = 1;

L
Linus Torvalds 已提交
1470 1471 1472 1473 1474 1475 1476 1477 1478 1479
/**
 *	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
1480
 *	to the new notifier to allow device to have a race free
L
Linus Torvalds 已提交
1481 1482 1483 1484 1485 1486
 *	view of the network device list.
 */

int register_netdevice_notifier(struct notifier_block *nb)
{
	struct net_device *dev;
1487
	struct net_device *last;
1488
	struct net *net;
L
Linus Torvalds 已提交
1489 1490 1491
	int err;

	rtnl_lock();
1492
	err = raw_notifier_chain_register(&netdev_chain, nb);
1493 1494
	if (err)
		goto unlock;
1495 1496 1497 1498
	if (dev_boot_phase)
		goto unlock;
	for_each_net(net) {
		for_each_netdev(net, dev) {
1499
			err = call_netdevice_notifier(nb, NETDEV_REGISTER, dev);
1500 1501 1502 1503 1504 1505
			err = notifier_to_errno(err);
			if (err)
				goto rollback;

			if (!(dev->flags & IFF_UP))
				continue;
L
Linus Torvalds 已提交
1506

1507
			call_netdevice_notifier(nb, NETDEV_UP, dev);
1508
		}
L
Linus Torvalds 已提交
1509
	}
1510 1511

unlock:
L
Linus Torvalds 已提交
1512 1513
	rtnl_unlock();
	return err;
1514 1515 1516

rollback:
	last = dev;
1517 1518 1519
	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev == last)
1520
				goto outroll;
1521

1522
			if (dev->flags & IFF_UP) {
1523 1524 1525
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1526
			}
1527
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1528 1529
		}
	}
1530

1531
outroll:
1532
	raw_notifier_chain_unregister(&netdev_chain, nb);
1533
	goto unlock;
L
Linus Torvalds 已提交
1534
}
E
Eric Dumazet 已提交
1535
EXPORT_SYMBOL(register_netdevice_notifier);
L
Linus Torvalds 已提交
1536 1537 1538 1539 1540 1541 1542 1543 1544

/**
 *	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.
1545 1546 1547 1548
 *
 * 	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
1549 1550 1551 1552
 */

int unregister_netdevice_notifier(struct notifier_block *nb)
{
1553 1554
	struct net_device *dev;
	struct net *net;
1555 1556 1557
	int err;

	rtnl_lock();
1558
	err = raw_notifier_chain_unregister(&netdev_chain, nb);
1559 1560 1561 1562 1563 1564
	if (err)
		goto unlock;

	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev->flags & IFF_UP) {
1565 1566 1567
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1568
			}
1569
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1570 1571 1572
		}
	}
unlock:
1573 1574
	rtnl_unlock();
	return err;
L
Linus Torvalds 已提交
1575
}
E
Eric Dumazet 已提交
1576
EXPORT_SYMBOL(unregister_netdevice_notifier);
L
Linus Torvalds 已提交
1577

1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
/**
 *	call_netdevice_notifiers_info - call all network notifier blocks
 *	@val: value passed unmodified to notifier function
 *	@dev: net_device pointer passed unmodified to notifier function
 *	@info: notifier information data
 *
 *	Call all network notifier blocks.  Parameters and return value
 *	are as for raw_notifier_call_chain().
 */

S
stephen hemminger 已提交
1588 1589 1590
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info)
1591 1592 1593 1594 1595 1596
{
	ASSERT_RTNL();
	netdev_notifier_info_init(info, dev);
	return raw_notifier_call_chain(&netdev_chain, val, info);
}

L
Linus Torvalds 已提交
1597 1598 1599
/**
 *	call_netdevice_notifiers - call all network notifier blocks
 *      @val: value passed unmodified to notifier function
1600
 *      @dev: net_device pointer passed unmodified to notifier function
L
Linus Torvalds 已提交
1601 1602
 *
 *	Call all network notifier blocks.  Parameters and return value
1603
 *	are as for raw_notifier_call_chain().
L
Linus Torvalds 已提交
1604 1605
 */

1606
int call_netdevice_notifiers(unsigned long val, struct net_device *dev)
L
Linus Torvalds 已提交
1607
{
1608 1609 1610
	struct netdev_notifier_info info;

	return call_netdevice_notifiers_info(val, dev, &info);
L
Linus Torvalds 已提交
1611
}
1612
EXPORT_SYMBOL(call_netdevice_notifiers);
L
Linus Torvalds 已提交
1613

1614
static struct static_key netstamp_needed __read_mostly;
1615
#ifdef HAVE_JUMP_LABEL
1616
/* We are not allowed to call static_key_slow_dec() from irq context
1617
 * If net_disable_timestamp() is called from irq context, defer the
1618
 * static_key_slow_dec() calls.
1619 1620 1621
 */
static atomic_t netstamp_needed_deferred;
#endif
L
Linus Torvalds 已提交
1622 1623 1624

void net_enable_timestamp(void)
{
1625 1626 1627 1628 1629
#ifdef HAVE_JUMP_LABEL
	int deferred = atomic_xchg(&netstamp_needed_deferred, 0);

	if (deferred) {
		while (--deferred)
1630
			static_key_slow_dec(&netstamp_needed);
1631 1632 1633
		return;
	}
#endif
1634
	static_key_slow_inc(&netstamp_needed);
L
Linus Torvalds 已提交
1635
}
E
Eric Dumazet 已提交
1636
EXPORT_SYMBOL(net_enable_timestamp);
L
Linus Torvalds 已提交
1637 1638 1639

void net_disable_timestamp(void)
{
1640 1641 1642 1643 1644 1645
#ifdef HAVE_JUMP_LABEL
	if (in_interrupt()) {
		atomic_inc(&netstamp_needed_deferred);
		return;
	}
#endif
1646
	static_key_slow_dec(&netstamp_needed);
L
Linus Torvalds 已提交
1647
}
E
Eric Dumazet 已提交
1648
EXPORT_SYMBOL(net_disable_timestamp);
L
Linus Torvalds 已提交
1649

E
Eric Dumazet 已提交
1650
static inline void net_timestamp_set(struct sk_buff *skb)
L
Linus Torvalds 已提交
1651
{
1652
	skb->tstamp.tv64 = 0;
1653
	if (static_key_false(&netstamp_needed))
1654
		__net_timestamp(skb);
L
Linus Torvalds 已提交
1655 1656
}

1657
#define net_timestamp_check(COND, SKB)			\
1658
	if (static_key_false(&netstamp_needed)) {		\
1659 1660 1661
		if ((COND) && !(SKB)->tstamp.tv64)	\
			__net_timestamp(SKB);		\
	}						\
E
Eric Dumazet 已提交
1662

1663
bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb)
1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681
{
	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;
}
1682
EXPORT_SYMBOL_GPL(is_skb_forwardable);
1683

H
Herbert Xu 已提交
1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706
int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb)
{
	if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY) {
		if (skb_copy_ubufs(skb, GFP_ATOMIC)) {
			atomic_long_inc(&dev->rx_dropped);
			kfree_skb(skb);
			return NET_RX_DROP;
		}
	}

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

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

	return 0;
}
EXPORT_SYMBOL_GPL(__dev_forward_skb);

1707 1708 1709 1710 1711 1712 1713 1714
/**
 * dev_forward_skb - loopback an skb to another netif
 *
 * @dev: destination network device
 * @skb: buffer to forward
 *
 * return values:
 *	NET_RX_SUCCESS	(no congestion)
1715
 *	NET_RX_DROP     (packet was dropped, but freed)
1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726
 *
 * dev_forward_skb can be used for injecting an skb from the
 * start_xmit function of one device into the receive queue
 * of another device.
 *
 * The receiving device may be in another namespace, so
 * we have to clear all information in the skb that could
 * impact namespace isolation.
 */
int dev_forward_skb(struct net_device *dev, struct sk_buff *skb)
{
H
Herbert Xu 已提交
1727
	return __dev_forward_skb(dev, skb) ?: netif_rx_internal(skb);
1728 1729 1730
}
EXPORT_SYMBOL_GPL(dev_forward_skb);

1731 1732 1733 1734
static inline int deliver_skb(struct sk_buff *skb,
			      struct packet_type *pt_prev,
			      struct net_device *orig_dev)
{
1735 1736
	if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
		return -ENOMEM;
1737 1738 1739 1740
	atomic_inc(&skb->users);
	return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
}

1741 1742
static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
{
1743
	if (!ptype->af_packet_priv || !skb->sk)
1744 1745 1746 1747 1748 1749 1750 1751 1752 1753
		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
1754 1755 1756 1757 1758
/*
 *	Support routine. Sends outgoing frames to any network
 *	taps currently in use.
 */

1759
static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
L
Linus Torvalds 已提交
1760 1761
{
	struct packet_type *ptype;
1762 1763
	struct sk_buff *skb2 = NULL;
	struct packet_type *pt_prev = NULL;
1764

L
Linus Torvalds 已提交
1765 1766 1767 1768 1769 1770
	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) &&
1771
		    (!skb_loop_sk(ptype, skb))) {
1772 1773 1774 1775 1776 1777 1778
			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
1779 1780 1781
			if (!skb2)
				break;

1782 1783
			net_timestamp_set(skb2);

L
Linus Torvalds 已提交
1784 1785 1786 1787
			/* skb->nh should be correctly
			   set by sender, so that the second statement is
			   just protection against buggy protocols.
			 */
1788
			skb_reset_mac_header(skb2);
L
Linus Torvalds 已提交
1789

1790
			if (skb_network_header(skb2) < skb2->data ||
1791
			    skb_network_header(skb2) > skb_tail_pointer(skb2)) {
1792 1793 1794
				net_crit_ratelimited("protocol %04x is buggy, dev %s\n",
						     ntohs(skb2->protocol),
						     dev->name);
1795
				skb_reset_network_header(skb2);
L
Linus Torvalds 已提交
1796 1797
			}

1798
			skb2->transport_header = skb2->network_header;
L
Linus Torvalds 已提交
1799
			skb2->pkt_type = PACKET_OUTGOING;
1800
			pt_prev = ptype;
L
Linus Torvalds 已提交
1801 1802
		}
	}
1803 1804
	if (pt_prev)
		pt_prev->func(skb2, skb->dev, pt_prev, skb->dev);
L
Linus Torvalds 已提交
1805 1806 1807
	rcu_read_unlock();
}

1808 1809
/**
 * netif_setup_tc - Handle tc mappings on real_num_tx_queues change
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820
 * @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
1821
static void netif_setup_tc(struct net_device *dev, unsigned int txq)
1822 1823 1824 1825 1826 1827
{
	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) {
1828
		pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n");
1829 1830 1831 1832 1833 1834 1835 1836 1837 1838
		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) {
1839 1840
			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);
1841 1842 1843 1844 1845
			netdev_set_prio_tc_map(dev, i, 0);
		}
	}
}

1846 1847 1848 1849 1850
#ifdef CONFIG_XPS
static DEFINE_MUTEX(xps_map_mutex);
#define xmap_dereference(P)		\
	rcu_dereference_protected((P), lockdep_is_held(&xps_map_mutex))

1851 1852
static struct xps_map *remove_xps_queue(struct xps_dev_maps *dev_maps,
					int cpu, u16 index)
1853
{
1854 1855
	struct xps_map *map = NULL;
	int pos;
1856

1857 1858
	if (dev_maps)
		map = xmap_dereference(dev_maps->cpu_map[cpu]);
1859

1860 1861
	for (pos = 0; map && pos < map->len; pos++) {
		if (map->queues[pos] == index) {
1862 1863 1864
			if (map->len > 1) {
				map->queues[pos] = map->queues[--map->len];
			} else {
1865
				RCU_INIT_POINTER(dev_maps->cpu_map[cpu], NULL);
1866 1867 1868
				kfree_rcu(map, rcu);
				map = NULL;
			}
1869
			break;
1870 1871 1872
		}
	}

1873 1874 1875
	return map;
}

1876
static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
1877 1878
{
	struct xps_dev_maps *dev_maps;
1879
	int cpu, i;
1880 1881 1882 1883 1884 1885 1886 1887 1888
	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) {
1889 1890 1891 1892 1893
		for (i = index; i < dev->num_tx_queues; i++) {
			if (!remove_xps_queue(dev_maps, cpu, i))
				break;
		}
		if (i == dev->num_tx_queues)
1894 1895 1896 1897
			active = true;
	}

	if (!active) {
1898 1899 1900 1901
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

1902 1903 1904 1905
	for (i = index; i < dev->num_tx_queues; i++)
		netdev_queue_numa_node_write(netdev_get_tx_queue(dev, i),
					     NUMA_NO_NODE);

1906 1907 1908 1909
out_no_maps:
	mutex_unlock(&xps_map_mutex);
}

1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944
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;
}

1945 1946
int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
			u16 index)
1947
{
1948
	struct xps_dev_maps *dev_maps, *new_dev_maps = NULL;
1949 1950
	struct xps_map *map, *new_map;
	int maps_sz = max_t(unsigned int, XPS_DEV_MAPS_SIZE, L1_CACHE_BYTES);
1951 1952
	int cpu, numa_node_id = -2;
	bool active = false;
1953 1954 1955 1956 1957

	mutex_lock(&xps_map_mutex);

	dev_maps = xmap_dereference(dev->xps_maps);

1958 1959 1960 1961 1962 1963 1964
	/* 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);
1965 1966
		if (!new_dev_maps) {
			mutex_unlock(&xps_map_mutex);
1967
			return -ENOMEM;
1968
		}
1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982

		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;

1983
	for_each_possible_cpu(cpu) {
1984 1985 1986 1987 1988 1989 1990 1991 1992 1993
		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;
1994 1995 1996 1997 1998 1999
#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
2000 2001 2002 2003
		} 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);
2004
		}
2005

2006 2007
	}

2008 2009
	rcu_assign_pointer(dev->xps_maps, new_dev_maps);

2010
	/* Cleanup old maps */
2011 2012 2013 2014 2015 2016 2017
	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);
		}
2018

2019
		kfree_rcu(dev_maps, rcu);
2020 2021
	}

2022 2023
	dev_maps = new_dev_maps;
	active = true;
2024

2025 2026
out_no_new_maps:
	/* update Tx queue numa node */
2027 2028 2029 2030
	netdev_queue_numa_node_write(netdev_get_tx_queue(dev, index),
				     (numa_node_id >= 0) ? numa_node_id :
				     NUMA_NO_NODE);

2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049
	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:
2050 2051 2052 2053
	mutex_unlock(&xps_map_mutex);

	return 0;
error:
2054 2055 2056 2057 2058 2059 2060 2061 2062
	/* 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);
	}

2063 2064 2065 2066 2067 2068 2069 2070
	mutex_unlock(&xps_map_mutex);

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

#endif
2071 2072 2073 2074
/*
 * 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.
 */
2075
int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
2076
{
T
Tom Herbert 已提交
2077 2078
	int rc;

2079 2080
	if (txq < 1 || txq > dev->num_tx_queues)
		return -EINVAL;
2081

2082 2083
	if (dev->reg_state == NETREG_REGISTERED ||
	    dev->reg_state == NETREG_UNREGISTERING) {
2084 2085
		ASSERT_RTNL();

T
Tom Herbert 已提交
2086 2087
		rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
						  txq);
T
Tom Herbert 已提交
2088 2089 2090
		if (rc)
			return rc;

2091 2092 2093
		if (dev->num_tc)
			netif_setup_tc(dev, txq);

2094
		if (txq < dev->real_num_tx_queues) {
2095
			qdisc_reset_all_tx_gt(dev, txq);
2096 2097 2098 2099
#ifdef CONFIG_XPS
			netif_reset_xps_queues_gt(dev, txq);
#endif
		}
2100
	}
2101 2102 2103

	dev->real_num_tx_queues = txq;
	return 0;
2104 2105
}
EXPORT_SYMBOL(netif_set_real_num_tx_queues);
2106

2107
#ifdef CONFIG_SYSFS
2108 2109 2110 2111 2112 2113 2114
/**
 *	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
2115 2116
 *	negative error code.  If called before registration, it always
 *	succeeds.
2117 2118 2119 2120 2121
 */
int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
{
	int rc;

T
Tom Herbert 已提交
2122 2123 2124
	if (rxq < 1 || rxq > dev->num_rx_queues)
		return -EINVAL;

2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139
	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

2140 2141
/**
 * netif_get_num_default_rss_queues - default number of RSS queues
2142 2143 2144 2145
 *
 * This routine should set an upper limit on the number of RSS queues
 * used by default by multiqueue devices.
 */
2146
int netif_get_num_default_rss_queues(void)
2147 2148 2149 2150 2151
{
	return min_t(int, DEFAULT_MAX_NUM_RSS_QUEUES, num_online_cpus());
}
EXPORT_SYMBOL(netif_get_num_default_rss_queues);

2152
static inline void __netif_reschedule(struct Qdisc *q)
2153
{
2154 2155
	struct softnet_data *sd;
	unsigned long flags;
2156

2157 2158
	local_irq_save(flags);
	sd = &__get_cpu_var(softnet_data);
2159 2160 2161
	q->next_sched = NULL;
	*sd->output_queue_tailp = q;
	sd->output_queue_tailp = &q->next_sched;
2162 2163 2164 2165 2166 2167 2168 2169
	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);
2170 2171 2172
}
EXPORT_SYMBOL(__netif_schedule);

2173 2174 2175 2176 2177
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)
2178
{
2179 2180 2181
	return (struct dev_kfree_skb_cb *)skb->cb;
}

2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228
void netif_schedule_queue(struct netdev_queue *txq)
{
	rcu_read_lock();
	if (!(txq->state & QUEUE_STATE_ANY_XOFF)) {
		struct Qdisc *q = rcu_dereference(txq->qdisc);

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

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

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

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

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

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

2229
void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason)
2230
{
2231
	unsigned long flags;
2232

2233 2234 2235 2236 2237
	if (likely(atomic_read(&skb->users) == 1)) {
		smp_rmb();
		atomic_set(&skb->users, 0);
	} else if (likely(!atomic_dec_and_test(&skb->users))) {
		return;
2238
	}
2239 2240 2241 2242 2243 2244
	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);
2245
}
2246
EXPORT_SYMBOL(__dev_kfree_skb_irq);
2247

2248
void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason)
2249 2250
{
	if (in_irq() || irqs_disabled())
2251
		__dev_kfree_skb_irq(skb, reason);
2252 2253 2254
	else
		dev_kfree_skb(skb);
}
2255
EXPORT_SYMBOL(__dev_kfree_skb_any);
2256 2257


2258 2259 2260 2261 2262 2263
/**
 * netif_device_detach - mark device as removed
 * @dev: network device
 *
 * Mark device as removed from system and therefore no longer available.
 */
2264 2265 2266 2267
void netif_device_detach(struct net_device *dev)
{
	if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2268
		netif_tx_stop_all_queues(dev);
2269 2270 2271 2272
	}
}
EXPORT_SYMBOL(netif_device_detach);

2273 2274 2275 2276 2277 2278
/**
 * netif_device_attach - mark device as attached
 * @dev: network device
 *
 * Mark device as attached from system and restart if needed.
 */
2279 2280 2281 2282
void netif_device_attach(struct net_device *dev)
{
	if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2283
		netif_tx_wake_all_queues(dev);
2284
		__netdev_watchdog_up(dev);
2285 2286 2287 2288
	}
}
EXPORT_SYMBOL(netif_device_attach);

2289 2290
static void skb_warn_bad_offload(const struct sk_buff *skb)
{
2291
	static const netdev_features_t null_features = 0;
2292 2293 2294
	struct net_device *dev = skb->dev;
	const char *driver = "";

2295 2296 2297
	if (!net_ratelimit())
		return;

2298 2299 2300 2301 2302
	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",
2303 2304
	     driver, dev ? &dev->features : &null_features,
	     skb->sk ? &skb->sk->sk_route_caps : &null_features,
2305 2306 2307 2308
	     skb->len, skb->data_len, skb_shinfo(skb)->gso_size,
	     skb_shinfo(skb)->gso_type, skb->ip_summed);
}

L
Linus Torvalds 已提交
2309 2310 2311 2312
/*
 * Invalidate hardware checksum when packet is to be mangled, and
 * complete checksum manually on outgoing path.
 */
2313
int skb_checksum_help(struct sk_buff *skb)
L
Linus Torvalds 已提交
2314
{
2315
	__wsum csum;
2316
	int ret = 0, offset;
L
Linus Torvalds 已提交
2317

2318
	if (skb->ip_summed == CHECKSUM_COMPLETE)
2319 2320 2321
		goto out_set_summed;

	if (unlikely(skb_shinfo(skb)->gso_size)) {
2322 2323
		skb_warn_bad_offload(skb);
		return -EINVAL;
L
Linus Torvalds 已提交
2324 2325
	}

2326 2327 2328 2329 2330 2331 2332 2333 2334
	/* 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;
	}

2335
	offset = skb_checksum_start_offset(skb);
2336 2337 2338 2339 2340 2341 2342 2343
	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
2344 2345 2346 2347 2348
		ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
		if (ret)
			goto out;
	}

2349
	*(__sum16 *)(skb->data + offset) = csum_fold(csum);
2350
out_set_summed:
L
Linus Torvalds 已提交
2351
	skb->ip_summed = CHECKSUM_NONE;
2352
out:
L
Linus Torvalds 已提交
2353 2354
	return ret;
}
E
Eric Dumazet 已提交
2355
EXPORT_SYMBOL(skb_checksum_help);
L
Linus Torvalds 已提交
2356

2357
__be16 skb_network_protocol(struct sk_buff *skb, int *depth)
2358
{
2359
	unsigned int vlan_depth = skb->mac_len;
A
Al Viro 已提交
2360
	__be16 type = skb->protocol;
2361

2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372
	/* 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;
	}

2373 2374 2375 2376 2377 2378
	/* if skb->protocol is 802.1Q/AD then the header should already be
	 * present at mac_len - VLAN_HLEN (if mac_len > 0), or at
	 * ETH_HLEN otherwise
	 */
	if (type == htons(ETH_P_8021Q) || type == htons(ETH_P_8021AD)) {
		if (vlan_depth) {
2379
			if (WARN_ON(vlan_depth < VLAN_HLEN))
2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396
				return 0;
			vlan_depth -= VLAN_HLEN;
		} else {
			vlan_depth = ETH_HLEN;
		}
		do {
			struct vlan_hdr *vh;

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

			vh = (struct vlan_hdr *)(skb->data + vlan_depth);
			type = vh->h_vlan_encapsulated_proto;
			vlan_depth += VLAN_HLEN;
		} while (type == htons(ETH_P_8021Q) ||
			 type == htons(ETH_P_8021AD));
2397 2398
	}

2399 2400
	*depth = vlan_depth;

2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413
	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;
2414 2415
	int vlan_depth = skb->mac_len;
	__be16 type = skb_network_protocol(skb, &vlan_depth);
2416 2417 2418 2419

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

2420
	__skb_pull(skb, vlan_depth);
2421 2422

	rcu_read_lock();
2423
	list_for_each_entry_rcu(ptype, &offload_base, list) {
2424 2425
		if (ptype->type == type && ptype->callbacks.gso_segment) {
			segs = ptype->callbacks.gso_segment(skb, features);
2426 2427 2428 2429 2430
			break;
		}
	}
	rcu_read_unlock();

2431
	__skb_push(skb, skb->data - skb_mac_header(skb));
2432

2433 2434
	return segs;
}
2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466
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);

2467 2468
		err = skb_cow_head(skb, 0);
		if (err < 0)
2469 2470 2471
			return ERR_PTR(err);
	}

2472
	SKB_GSO_CB(skb)->mac_offset = skb_headroom(skb);
2473 2474
	SKB_GSO_CB(skb)->encap_level = 0;

2475 2476 2477 2478 2479
	skb_reset_mac_header(skb);
	skb_reset_mac_len(skb);

	return skb_mac_gso_segment(skb, features);
}
2480
EXPORT_SYMBOL(__skb_gso_segment);
2481

2482 2483 2484 2485 2486
/* Take action when hardware reception checksum errors are detected. */
#ifdef CONFIG_BUG
void netdev_rx_csum_fault(struct net_device *dev)
{
	if (net_ratelimit()) {
2487
		pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>");
2488 2489 2490 2491 2492 2493
		dump_stack();
	}
}
EXPORT_SYMBOL(netdev_rx_csum_fault);
#endif

L
Linus Torvalds 已提交
2494 2495 2496 2497 2498
/* 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.
 */

2499
static int illegal_highdma(struct net_device *dev, struct sk_buff *skb)
L
Linus Torvalds 已提交
2500
{
2501
#ifdef CONFIG_HIGHMEM
L
Linus Torvalds 已提交
2502
	int i;
2503
	if (!(dev->features & NETIF_F_HIGHDMA)) {
2504 2505 2506
		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)))
2507
				return 1;
2508
		}
2509
	}
L
Linus Torvalds 已提交
2510

2511 2512
	if (PCI_DMA_BUS_IS_PHYS) {
		struct device *pdev = dev->dev.parent;
L
Linus Torvalds 已提交
2513

E
Eric Dumazet 已提交
2514 2515
		if (!pdev)
			return 0;
2516
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2517 2518
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			dma_addr_t addr = page_to_phys(skb_frag_page(frag));
2519 2520 2521 2522
			if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask)
				return 1;
		}
	}
2523
#endif
L
Linus Torvalds 已提交
2524 2525 2526
	return 0;
}

2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548
/* If MPLS offload request, verify we are testing hardware MPLS features
 * instead of standard features for the netdev.
 */
#ifdef CONFIG_NET_MPLS_GSO
static netdev_features_t net_mpls_features(struct sk_buff *skb,
					   netdev_features_t features,
					   __be16 type)
{
	if (type == htons(ETH_P_MPLS_UC) || type == htons(ETH_P_MPLS_MC))
		features &= skb->dev->mpls_features;

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

2549
static netdev_features_t harmonize_features(struct sk_buff *skb,
2550
	netdev_features_t features)
2551
{
2552
	int tmp;
2553 2554 2555 2556
	__be16 type;

	type = skb_network_protocol(skb, &tmp);
	features = net_mpls_features(skb, features, type);
2557

2558
	if (skb->ip_summed != CHECKSUM_NONE &&
2559
	    !can_checksum_protocol(features, type)) {
2560
		features &= ~NETIF_F_ALL_CSUM;
2561
	} else if (illegal_highdma(skb->dev, skb)) {
2562 2563 2564 2565 2566 2567
		features &= ~NETIF_F_SG;
	}

	return features;
}

2568
netdev_features_t netif_skb_features(struct sk_buff *skb)
2569 2570
{
	__be16 protocol = skb->protocol;
2571
	netdev_features_t features = skb->dev->features;
2572

2573
	if (skb_shinfo(skb)->gso_segs > skb->dev->gso_max_segs)
2574 2575
		features &= ~NETIF_F_GSO_MASK;

P
Patrick McHardy 已提交
2576
	if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD)) {
2577 2578
		struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
		protocol = veh->h_vlan_encapsulated_proto;
2579
	} else if (!vlan_tx_tag_present(skb)) {
2580
		return harmonize_features(skb, features);
2581
	}
2582

2583 2584 2585 2586
	features = netdev_intersect_features(features,
					     skb->dev->vlan_features |
					     NETIF_F_HW_VLAN_CTAG_TX |
					     NETIF_F_HW_VLAN_STAG_TX);
2587

2588
	if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD))
2589 2590 2591 2592 2593 2594 2595
		features = netdev_intersect_features(features,
						     NETIF_F_SG |
						     NETIF_F_HIGHDMA |
						     NETIF_F_FRAGLIST |
						     NETIF_F_GEN_CSUM |
						     NETIF_F_HW_VLAN_CTAG_TX |
						     NETIF_F_HW_VLAN_STAG_TX);
2596

2597
	return harmonize_features(skb, features);
2598
}
2599
EXPORT_SYMBOL(netif_skb_features);
2600

2601
static int xmit_one(struct sk_buff *skb, struct net_device *dev,
2602
		    struct netdev_queue *txq, bool more)
2603 2604 2605 2606 2607 2608 2609 2610 2611
{
	unsigned int len;
	int rc;

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

	len = skb->len;
	trace_net_dev_start_xmit(skb, dev);
2612
	rc = netdev_start_xmit(skb, dev, txq, more);
2613 2614 2615 2616 2617
	trace_net_dev_xmit(skb, rc, dev, len);

	return rc;
}

2618 2619
struct sk_buff *dev_hard_start_xmit(struct sk_buff *first, struct net_device *dev,
				    struct netdev_queue *txq, int *ret)
2620 2621 2622 2623 2624 2625 2626 2627
{
	struct sk_buff *skb = first;
	int rc = NETDEV_TX_OK;

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

		skb->next = NULL;
2628
		rc = xmit_one(skb, dev, txq, next != NULL);
2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645
		if (unlikely(!dev_xmit_complete(rc))) {
			skb->next = next;
			goto out;
		}

		skb = next;
		if (netif_xmit_stopped(txq) && skb) {
			rc = NETDEV_TX_BUSY;
			break;
		}
	}

out:
	*ret = rc;
	return skb;
}

2646
struct sk_buff *validate_xmit_vlan(struct sk_buff *skb, netdev_features_t features)
2647
{
2648 2649 2650 2651 2652 2653 2654 2655 2656
	if (vlan_tx_tag_present(skb) &&
	    !vlan_hw_offload_capable(features, skb->vlan_proto)) {
		skb = __vlan_put_tag(skb, skb->vlan_proto,
				     vlan_tx_tag_get(skb));
		if (skb)
			skb->vlan_tci = 0;
	}
	return skb;
}
2657

2658
static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev)
2659 2660
{
	netdev_features_t features;
2661

2662 2663
	if (skb->next)
		return skb;
E
Eric Dumazet 已提交
2664

2665 2666 2667 2668 2669
	/* If device doesn't need skb->dst, release it right now while
	 * its hot in this cpu cache
	 */
	if (dev->priv_flags & IFF_XMIT_DST_RELEASE)
		skb_dst_drop(skb);
2670

2671 2672 2673 2674
	features = netif_skb_features(skb);
	skb = validate_xmit_vlan(skb, features);
	if (unlikely(!skb))
		goto out_null;
2675

2676 2677 2678 2679 2680 2681 2682 2683
	/* 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;

	if (netif_needs_gso(skb, features)) {
2684 2685 2686
		struct sk_buff *segs;

		segs = skb_gso_segment(skb, features);
2687
		if (IS_ERR(segs)) {
2688
			segs = NULL;
2689 2690 2691 2692
		} else if (segs) {
			consume_skb(skb);
			skb = segs;
		}
2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711
	} else {
		if (skb_needs_linearize(skb, features) &&
		    __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) {
			if (skb->encapsulation)
				skb_set_inner_transport_header(skb,
							       skb_checksum_start_offset(skb));
			else
				skb_set_transport_header(skb,
							 skb_checksum_start_offset(skb));
			if (!(features & NETIF_F_ALL_CSUM) &&
			    skb_checksum_help(skb))
				goto out_kfree_skb;
2712
		}
2713
	}
2714

2715
	return skb;
2716

2717 2718 2719 2720 2721
out_kfree_skb:
	kfree_skb(skb);
out_null:
	return NULL;
}
2722

2723 2724 2725 2726
struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev)
{
	struct sk_buff *next, *head = NULL, *tail;

2727
	for (; skb != NULL; skb = next) {
2728 2729
		next = skb->next;
		skb->next = NULL;
2730 2731 2732 2733

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

2734
		skb = validate_xmit_skb(skb, dev);
2735 2736
		if (!skb)
			continue;
2737

2738 2739 2740 2741 2742 2743 2744 2745
		if (!head)
			head = skb;
		else
			tail->next = skb;
		/* If skb was segmented, skb->prev points to
		 * the last segment. If not, it still contains skb.
		 */
		tail = skb->prev;
2746 2747 2748 2749
	}
	return head;
}

2750 2751 2752 2753 2754 2755 2756 2757 2758 2759
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)  {
2760
		unsigned int hdr_len;
2761
		u16 gso_segs = shinfo->gso_segs;
2762

2763 2764 2765 2766
		/* mac layer + network layer */
		hdr_len = skb_transport_header(skb) - skb_mac_header(skb);

		/* + transport layer */
2767 2768 2769 2770
		if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)))
			hdr_len += tcp_hdrlen(skb);
		else
			hdr_len += sizeof(struct udphdr);
2771 2772 2773 2774 2775 2776

		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;
2777 2778 2779
	}
}

2780 2781 2782 2783 2784
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
2785
	bool contended;
2786 2787
	int rc;

2788
	qdisc_pkt_len_init(skb);
E
Eric Dumazet 已提交
2789
	qdisc_calculate_pkt_len(skb, q);
2790 2791 2792
	/*
	 * Heuristic to force contended enqueues to serialize on a
	 * separate lock before trying to get qdisc main lock.
Y
Ying Xue 已提交
2793 2794
	 * This permits __QDISC___STATE_RUNNING owner to get the lock more
	 * often and dequeue packets faster.
2795
	 */
E
Eric Dumazet 已提交
2796
	contended = qdisc_is_running(q);
2797 2798 2799
	if (unlikely(contended))
		spin_lock(&q->busylock);

2800 2801 2802 2803 2804
	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) &&
2805
		   qdisc_run_begin(q)) {
2806 2807 2808 2809 2810
		/*
		 * 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
2811 2812
		if (!(dev->priv_flags & IFF_XMIT_DST_RELEASE))
			skb_dst_force(skb);
2813 2814 2815

		qdisc_bstats_update(q, skb);

2816
		if (sch_direct_xmit(skb, q, dev, txq, root_lock, true)) {
2817 2818 2819 2820
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
2821
			__qdisc_run(q);
2822
		} else
2823
			qdisc_run_end(q);
2824 2825 2826

		rc = NET_XMIT_SUCCESS;
	} else {
E
Eric Dumazet 已提交
2827
		skb_dst_force(skb);
E
Eric Dumazet 已提交
2828
		rc = q->enqueue(skb, q) & NET_XMIT_MASK;
2829 2830 2831 2832 2833 2834 2835
		if (qdisc_run_begin(q)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
			__qdisc_run(q);
		}
2836 2837
	}
	spin_unlock(root_lock);
2838 2839
	if (unlikely(contended))
		spin_unlock(&q->busylock);
2840 2841 2842
	return rc;
}

2843
#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2844 2845
static void skb_update_prio(struct sk_buff *skb)
{
I
Igor Maravic 已提交
2846
	struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap);
2847

2848 2849 2850 2851 2852 2853
	if (!skb->priority && skb->sk && map) {
		unsigned int prioidx = skb->sk->sk_cgrp_prioidx;

		if (prioidx < map->priomap_len)
			skb->priority = map->priomap[prioidx];
	}
2854 2855 2856 2857 2858
}
#else
#define skb_update_prio(skb)
#endif

2859
static DEFINE_PER_CPU(int, xmit_recursion);
2860
#define RECURSION_LIMIT 10
2861

2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878
/**
 *	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);

2879
/**
2880
 *	__dev_queue_xmit - transmit a buffer
2881
 *	@skb: buffer to transmit
2882
 *	@accel_priv: private data used for L2 forwarding offload
2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904
 *
 *	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
 */
2905
static int __dev_queue_xmit(struct sk_buff *skb, void *accel_priv)
L
Linus Torvalds 已提交
2906 2907
{
	struct net_device *dev = skb->dev;
2908
	struct netdev_queue *txq;
L
Linus Torvalds 已提交
2909 2910 2911
	struct Qdisc *q;
	int rc = -ENOMEM;

2912 2913
	skb_reset_mac_header(skb);

2914 2915 2916
	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_SCHED_TSTAMP))
		__skb_tstamp_tx(skb, NULL, skb->sk, SCM_TSTAMP_SCHED);

2917 2918
	/* Disable soft irqs for various locks below. Also
	 * stops preemption for RCU.
L
Linus Torvalds 已提交
2919
	 */
2920
	rcu_read_lock_bh();
L
Linus Torvalds 已提交
2921

2922 2923
	skb_update_prio(skb);

2924
	txq = netdev_pick_tx(dev, skb, accel_priv);
2925
	q = rcu_dereference_bh(txq->qdisc);
2926

L
Linus Torvalds 已提交
2927
#ifdef CONFIG_NET_CLS_ACT
E
Eric Dumazet 已提交
2928
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS);
L
Linus Torvalds 已提交
2929
#endif
2930
	trace_net_dev_queue(skb);
L
Linus Torvalds 已提交
2931
	if (q->enqueue) {
2932
		rc = __dev_xmit_skb(skb, q, dev, txq);
2933
		goto out;
L
Linus Torvalds 已提交
2934 2935 2936 2937 2938
	}

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

H
Herbert Xu 已提交
2939 2940
	   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
2941 2942 2943 2944 2945 2946 2947 2948 2949 2950
	   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 */

2951
		if (txq->xmit_lock_owner != cpu) {
L
Linus Torvalds 已提交
2952

2953 2954 2955
			if (__this_cpu_read(xmit_recursion) > RECURSION_LIMIT)
				goto recursion_alert;

2956 2957 2958 2959
			skb = validate_xmit_skb(skb, dev);
			if (!skb)
				goto drop;

2960
			HARD_TX_LOCK(dev, txq, cpu);
L
Linus Torvalds 已提交
2961

2962
			if (!netif_xmit_stopped(txq)) {
2963
				__this_cpu_inc(xmit_recursion);
2964
				skb = dev_hard_start_xmit(skb, dev, txq, &rc);
2965
				__this_cpu_dec(xmit_recursion);
2966
				if (dev_xmit_complete(rc)) {
2967
					HARD_TX_UNLOCK(dev, txq);
L
Linus Torvalds 已提交
2968 2969 2970
					goto out;
				}
			}
2971
			HARD_TX_UNLOCK(dev, txq);
2972 2973
			net_crit_ratelimited("Virtual device %s asks to queue packet!\n",
					     dev->name);
L
Linus Torvalds 已提交
2974 2975
		} else {
			/* Recursion is detected! It is possible,
2976 2977 2978
			 * unfortunately
			 */
recursion_alert:
2979 2980
			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
					     dev->name);
L
Linus Torvalds 已提交
2981 2982 2983 2984
		}
	}

	rc = -ENETDOWN;
2985
drop:
2986
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
2987

2988
	atomic_long_inc(&dev->tx_dropped);
2989
	kfree_skb_list(skb);
L
Linus Torvalds 已提交
2990 2991
	return rc;
out:
2992
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
2993 2994
	return rc;
}
2995 2996 2997 2998 2999

int dev_queue_xmit(struct sk_buff *skb)
{
	return __dev_queue_xmit(skb, NULL);
}
E
Eric Dumazet 已提交
3000
EXPORT_SYMBOL(dev_queue_xmit);
L
Linus Torvalds 已提交
3001

3002 3003 3004 3005 3006 3007
int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv)
{
	return __dev_queue_xmit(skb, accel_priv);
}
EXPORT_SYMBOL(dev_queue_xmit_accel);

L
Linus Torvalds 已提交
3008 3009 3010 3011 3012

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

3013
int netdev_max_backlog __read_mostly = 1000;
E
Eric Dumazet 已提交
3014 3015
EXPORT_SYMBOL(netdev_max_backlog);

E
Eric Dumazet 已提交
3016
int netdev_tstamp_prequeue __read_mostly = 1;
3017 3018
int netdev_budget __read_mostly = 300;
int weight_p __read_mostly = 64;            /* old backlog weight */
L
Linus Torvalds 已提交
3019

E
Eric Dumazet 已提交
3020 3021 3022 3023 3024 3025 3026 3027
/* 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);
}

3028 3029 3030
#ifdef CONFIG_RPS

/* One global table that all flow-based protocols share. */
E
Eric Dumazet 已提交
3031
struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
3032 3033
EXPORT_SYMBOL(rps_sock_flow_table);

3034
struct static_key rps_needed __read_mostly;
3035

3036 3037 3038 3039
static struct rps_dev_flow *
set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
	    struct rps_dev_flow *rflow, u16 next_cpu)
{
3040
	if (next_cpu != RPS_NO_CPU) {
3041 3042 3043 3044 3045 3046 3047 3048 3049
#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? */
3050 3051
		if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
		    !(dev->features & NETIF_F_NTUPLE))
3052 3053 3054 3055 3056 3057 3058 3059 3060
			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;
3061
		flow_id = skb_get_hash(skb) & flow_table->mask;
3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073
		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 =
3074
			per_cpu(softnet_data, next_cpu).input_queue_head;
3075 3076
	}

3077
	rflow->cpu = next_cpu;
3078 3079 3080
	return rflow;
}

3081 3082 3083 3084 3085 3086 3087 3088 3089
/*
 * 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
3090
	struct rps_map *map;
3091 3092 3093 3094
	struct rps_dev_flow_table *flow_table;
	struct rps_sock_flow_table *sock_flow_table;
	int cpu = -1;
	u16 tcpu;
3095
	u32 hash;
3096 3097 3098

	if (skb_rx_queue_recorded(skb)) {
		u16 index = skb_get_rx_queue(skb);
3099 3100 3101 3102 3103
		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);
3104 3105 3106 3107 3108 3109
			goto done;
		}
		rxqueue = dev->_rx + index;
	} else
		rxqueue = dev->_rx;

E
Eric Dumazet 已提交
3110 3111
	map = rcu_dereference(rxqueue->rps_map);
	if (map) {
3112
		if (map->len == 1 &&
3113
		    !rcu_access_pointer(rxqueue->rps_flow_table)) {
3114 3115 3116 3117 3118
			tcpu = map->cpus[0];
			if (cpu_online(tcpu))
				cpu = tcpu;
			goto done;
		}
3119
	} else if (!rcu_access_pointer(rxqueue->rps_flow_table)) {
3120
		goto done;
3121
	}
3122

3123
	skb_reset_network_header(skb);
3124 3125
	hash = skb_get_hash(skb);
	if (!hash)
3126 3127
		goto done;

T
Tom Herbert 已提交
3128 3129 3130 3131 3132 3133
	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;

3134
		rflow = &flow_table->flows[hash & flow_table->mask];
T
Tom Herbert 已提交
3135 3136
		tcpu = rflow->cpu;

3137
		next_cpu = sock_flow_table->ents[hash & sock_flow_table->mask];
T
Tom Herbert 已提交
3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152

		/*
		 * 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 -
3153 3154
		      rflow->last_qtail)) >= 0)) {
			tcpu = next_cpu;
3155
			rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
3156
		}
3157

T
Tom Herbert 已提交
3158 3159 3160 3161 3162 3163 3164
		if (tcpu != RPS_NO_CPU && cpu_online(tcpu)) {
			*rflowp = rflow;
			cpu = tcpu;
			goto done;
		}
	}

T
Tom Herbert 已提交
3165
	if (map) {
3166
		tcpu = map->cpus[reciprocal_scale(hash, map->len)];
T
Tom Herbert 已提交
3167 3168 3169 3170 3171 3172 3173 3174 3175 3176
		if (cpu_online(tcpu)) {
			cpu = tcpu;
			goto done;
		}
	}

done:
	return cpu;
}

3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216
#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
3217
/* Called from hardirq (IPI) context */
E
Eric Dumazet 已提交
3218
static void rps_trigger_softirq(void *data)
T
Tom Herbert 已提交
3219
{
E
Eric Dumazet 已提交
3220 3221
	struct softnet_data *sd = data;

E
Eric Dumazet 已提交
3222
	____napi_schedule(sd, &sd->backlog);
C
Changli Gao 已提交
3223
	sd->received_rps++;
T
Tom Herbert 已提交
3224
}
E
Eric Dumazet 已提交
3225

T
Tom Herbert 已提交
3226
#endif /* CONFIG_RPS */
T
Tom Herbert 已提交
3227

E
Eric Dumazet 已提交
3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248
/*
 * 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;
}

3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267
#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) {
3268
		new_flow = skb_get_hash(skb) & (fl->num_buckets - 1);
3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288
		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
3289 3290 3291 3292
/*
 * 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
3293 3294
static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
			      unsigned int *qtail)
T
Tom Herbert 已提交
3295
{
E
Eric Dumazet 已提交
3296
	struct softnet_data *sd;
T
Tom Herbert 已提交
3297
	unsigned long flags;
3298
	unsigned int qlen;
T
Tom Herbert 已提交
3299

E
Eric Dumazet 已提交
3300
	sd = &per_cpu(softnet_data, cpu);
T
Tom Herbert 已提交
3301 3302 3303

	local_irq_save(flags);

E
Eric Dumazet 已提交
3304
	rps_lock(sd);
3305 3306
	qlen = skb_queue_len(&sd->input_pkt_queue);
	if (qlen <= netdev_max_backlog && !skb_flow_limit(skb, qlen)) {
3307
		if (skb_queue_len(&sd->input_pkt_queue)) {
T
Tom Herbert 已提交
3308
enqueue:
E
Eric Dumazet 已提交
3309
			__skb_queue_tail(&sd->input_pkt_queue, skb);
3310
			input_queue_tail_incr_save(sd, qtail);
E
Eric Dumazet 已提交
3311
			rps_unlock(sd);
3312
			local_irq_restore(flags);
T
Tom Herbert 已提交
3313 3314 3315
			return NET_RX_SUCCESS;
		}

3316 3317 3318 3319
		/* 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
3320
			if (!rps_ipi_queued(sd))
E
Eric Dumazet 已提交
3321
				____napi_schedule(sd, &sd->backlog);
T
Tom Herbert 已提交
3322 3323 3324 3325
		}
		goto enqueue;
	}

C
Changli Gao 已提交
3326
	sd->dropped++;
E
Eric Dumazet 已提交
3327
	rps_unlock(sd);
T
Tom Herbert 已提交
3328 3329 3330

	local_irq_restore(flags);

3331
	atomic_long_inc(&skb->dev->rx_dropped);
T
Tom Herbert 已提交
3332 3333 3334
	kfree_skb(skb);
	return NET_RX_DROP;
}
L
Linus Torvalds 已提交
3335

3336
static int netif_rx_internal(struct sk_buff *skb)
L
Linus Torvalds 已提交
3337
{
3338
	int ret;
L
Linus Torvalds 已提交
3339

3340
	net_timestamp_check(netdev_tstamp_prequeue, skb);
L
Linus Torvalds 已提交
3341

3342
	trace_netif_rx(skb);
E
Eric Dumazet 已提交
3343
#ifdef CONFIG_RPS
3344
	if (static_key_false(&rps_needed)) {
T
Tom Herbert 已提交
3345
		struct rps_dev_flow voidflow, *rflow = &voidflow;
3346 3347
		int cpu;

3348
		preempt_disable();
3349
		rcu_read_lock();
T
Tom Herbert 已提交
3350 3351

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
3352 3353
		if (cpu < 0)
			cpu = smp_processor_id();
T
Tom Herbert 已提交
3354 3355 3356

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

3357
		rcu_read_unlock();
3358
		preempt_enable();
3359 3360
	} else
#endif
T
Tom Herbert 已提交
3361 3362 3363 3364 3365
	{
		unsigned int qtail;
		ret = enqueue_to_backlog(skb, get_cpu(), &qtail);
		put_cpu();
	}
3366
	return ret;
L
Linus Torvalds 已提交
3367
}
3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389

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

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

	return netif_rx_internal(skb);
}
E
Eric Dumazet 已提交
3390
EXPORT_SYMBOL(netif_rx);
L
Linus Torvalds 已提交
3391 3392 3393 3394 3395

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

3396 3397
	trace_netif_rx_ni_entry(skb);

L
Linus Torvalds 已提交
3398
	preempt_disable();
3399
	err = netif_rx_internal(skb);
L
Linus Torvalds 已提交
3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423
	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;

3424
			WARN_ON(atomic_read(&skb->users));
3425 3426 3427 3428
			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
3429 3430 3431 3432 3433
			__kfree_skb(skb);
		}
	}

	if (sd->output_queue) {
3434
		struct Qdisc *head;
L
Linus Torvalds 已提交
3435 3436 3437 3438

		local_irq_disable();
		head = sd->output_queue;
		sd->output_queue = NULL;
3439
		sd->output_queue_tailp = &sd->output_queue;
L
Linus Torvalds 已提交
3440 3441 3442
		local_irq_enable();

		while (head) {
3443 3444 3445
			struct Qdisc *q = head;
			spinlock_t *root_lock;

L
Linus Torvalds 已提交
3446 3447
			head = head->next_sched;

3448
			root_lock = qdisc_lock(q);
3449
			if (spin_trylock(root_lock)) {
3450
				smp_mb__before_atomic();
3451 3452
				clear_bit(__QDISC_STATE_SCHED,
					  &q->state);
3453 3454
				qdisc_run(q);
				spin_unlock(root_lock);
L
Linus Torvalds 已提交
3455
			} else {
3456
				if (!test_bit(__QDISC_STATE_DEACTIVATED,
3457
					      &q->state)) {
3458
					__netif_reschedule(q);
3459
				} else {
3460
					smp_mb__before_atomic();
3461 3462 3463
					clear_bit(__QDISC_STATE_SCHED,
						  &q->state);
				}
L
Linus Torvalds 已提交
3464 3465 3466 3467 3468
			}
		}
	}
}

3469 3470
#if (defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)) && \
    (defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE))
3471 3472 3473
/* This hook is defined here for ATM LANE */
int (*br_fdb_test_addr_hook)(struct net_device *dev,
			     unsigned char *addr) __read_mostly;
3474
EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
3475
#endif
L
Linus Torvalds 已提交
3476 3477 3478 3479 3480 3481

#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
3482 3483
 * 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
3484 3485
 *
 */
3486
static int ing_filter(struct sk_buff *skb, struct netdev_queue *rxq)
L
Linus Torvalds 已提交
3487 3488
{
	struct net_device *dev = skb->dev;
3489
	u32 ttl = G_TC_RTTL(skb->tc_verd);
3490 3491
	int result = TC_ACT_OK;
	struct Qdisc *q;
3492

3493
	if (unlikely(MAX_RED_LOOP < ttl++)) {
3494 3495
		net_warn_ratelimited("Redir loop detected Dropping packet (%d->%d)\n",
				     skb->skb_iif, dev->ifindex);
3496 3497
		return TC_ACT_SHOT;
	}
L
Linus Torvalds 已提交
3498

3499 3500
	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
3501

3502
	q = rcu_dereference(rxq->qdisc);
3503
	if (q != &noop_qdisc) {
3504
		spin_lock(qdisc_lock(q));
3505 3506
		if (likely(!test_bit(__QDISC_STATE_DEACTIVATED, &q->state)))
			result = qdisc_enqueue_root(skb, q);
3507 3508
		spin_unlock(qdisc_lock(q));
	}
3509 3510 3511

	return result;
}
3512

3513 3514 3515 3516
static inline struct sk_buff *handle_ing(struct sk_buff *skb,
					 struct packet_type **pt_prev,
					 int *ret, struct net_device *orig_dev)
{
3517 3518
	struct netdev_queue *rxq = rcu_dereference(skb->dev->ingress_queue);

3519
	if (!rxq || rcu_access_pointer(rxq->qdisc) == &noop_qdisc)
3520
		goto out;
L
Linus Torvalds 已提交
3521

3522 3523 3524
	if (*pt_prev) {
		*ret = deliver_skb(skb, *pt_prev, orig_dev);
		*pt_prev = NULL;
L
Linus Torvalds 已提交
3525 3526
	}

3527
	switch (ing_filter(skb, rxq)) {
3528 3529 3530 3531 3532 3533 3534 3535 3536
	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
3537 3538 3539
}
#endif

3540 3541 3542 3543
/**
 *	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
3544
 *	@rx_handler_data: data pointer that is used by rx handler
3545
 *
3546
 *	Register a receive handler for a device. This handler will then be
3547 3548 3549 3550
 *	called from __netif_receive_skb. A negative errno code is returned
 *	on a failure.
 *
 *	The caller must hold the rtnl_mutex.
3551 3552
 *
 *	For a general description of rx_handler, see enum rx_handler_result.
3553 3554
 */
int netdev_rx_handler_register(struct net_device *dev,
J
Jiri Pirko 已提交
3555 3556
			       rx_handler_func_t *rx_handler,
			       void *rx_handler_data)
3557 3558 3559 3560 3561 3562
{
	ASSERT_RTNL();

	if (dev->rx_handler)
		return -EBUSY;

3563
	/* Note: rx_handler_data must be set before rx_handler */
J
Jiri Pirko 已提交
3564
	rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
3565 3566 3567 3568 3569 3570 3571 3572 3573 3574
	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
3575
 *	Unregister a receive handler from a device.
3576 3577 3578 3579 3580 3581 3582
 *
 *	The caller must hold the rtnl_mutex.
 */
void netdev_rx_handler_unregister(struct net_device *dev)
{

	ASSERT_RTNL();
3583
	RCU_INIT_POINTER(dev->rx_handler, NULL);
3584 3585 3586 3587 3588
	/* 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();
3589
	RCU_INIT_POINTER(dev->rx_handler_data, NULL);
3590 3591 3592
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);

3593 3594 3595 3596 3597 3598 3599
/*
 * 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) {
3600 3601 3602 3603 3604
	case htons(ETH_P_ARP):
	case htons(ETH_P_IP):
	case htons(ETH_P_IPV6):
	case htons(ETH_P_8021Q):
	case htons(ETH_P_8021AD):
3605 3606 3607 3608 3609 3610
		return true;
	default:
		return false;
	}
}

3611
static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc)
L
Linus Torvalds 已提交
3612 3613
{
	struct packet_type *ptype, *pt_prev;
3614
	rx_handler_func_t *rx_handler;
D
David S. Miller 已提交
3615
	struct net_device *orig_dev;
3616
	struct net_device *null_or_dev;
3617
	bool deliver_exact = false;
L
Linus Torvalds 已提交
3618
	int ret = NET_RX_DROP;
A
Al Viro 已提交
3619
	__be16 type;
L
Linus Torvalds 已提交
3620

3621
	net_timestamp_check(!netdev_tstamp_prequeue, skb);
3622

3623
	trace_netif_receive_skb(skb);
3624

J
Joe Eykholt 已提交
3625
	orig_dev = skb->dev;
3626

3627
	skb_reset_network_header(skb);
3628 3629
	if (!skb_transport_header_was_set(skb))
		skb_reset_transport_header(skb);
3630
	skb_reset_mac_len(skb);
L
Linus Torvalds 已提交
3631 3632 3633 3634 3635

	pt_prev = NULL;

	rcu_read_lock();

3636
another_round:
3637
	skb->skb_iif = skb->dev->ifindex;
3638 3639 3640

	__this_cpu_inc(softnet_data.processed);

P
Patrick McHardy 已提交
3641 3642
	if (skb->protocol == cpu_to_be16(ETH_P_8021Q) ||
	    skb->protocol == cpu_to_be16(ETH_P_8021AD)) {
3643
		skb = skb_vlan_untag(skb);
3644
		if (unlikely(!skb))
3645
			goto unlock;
3646 3647
	}

L
Linus Torvalds 已提交
3648 3649 3650 3651 3652 3653 3654
#ifdef CONFIG_NET_CLS_ACT
	if (skb->tc_verd & TC_NCLS) {
		skb->tc_verd = CLR_TC_NCLS(skb->tc_verd);
		goto ncls;
	}
#endif

3655
	if (pfmemalloc)
3656 3657
		goto skip_taps;

L
Linus Torvalds 已提交
3658
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
3659
		if (!ptype->dev || ptype->dev == skb->dev) {
3660
			if (pt_prev)
D
David S. Miller 已提交
3661
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3662 3663 3664 3665
			pt_prev = ptype;
		}
	}

3666
skip_taps:
L
Linus Torvalds 已提交
3667
#ifdef CONFIG_NET_CLS_ACT
3668 3669
	skb = handle_ing(skb, &pt_prev, &ret, orig_dev);
	if (!skb)
3670
		goto unlock;
L
Linus Torvalds 已提交
3671 3672 3673
ncls:
#endif

3674
	if (pfmemalloc && !skb_pfmemalloc_protocol(skb))
3675 3676
		goto drop;

3677 3678 3679 3680 3681
	if (vlan_tx_tag_present(skb)) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3682
		if (vlan_do_receive(&skb))
3683 3684
			goto another_round;
		else if (unlikely(!skb))
3685
			goto unlock;
3686 3687
	}

3688
	rx_handler = rcu_dereference(skb->dev->rx_handler);
3689 3690 3691 3692 3693
	if (rx_handler) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3694 3695
		switch (rx_handler(&skb)) {
		case RX_HANDLER_CONSUMED:
3696
			ret = NET_RX_SUCCESS;
3697
			goto unlock;
3698
		case RX_HANDLER_ANOTHER:
3699
			goto another_round;
3700 3701 3702 3703 3704 3705 3706
		case RX_HANDLER_EXACT:
			deliver_exact = true;
		case RX_HANDLER_PASS:
			break;
		default:
			BUG();
		}
3707
	}
L
Linus Torvalds 已提交
3708

E
Eric Dumazet 已提交
3709 3710 3711 3712 3713 3714 3715 3716 3717
	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;
	}
3718

3719
	/* deliver only exact match when indicated */
3720
	null_or_dev = deliver_exact ? skb->dev : NULL;
3721

L
Linus Torvalds 已提交
3722
	type = skb->protocol;
3723 3724
	list_for_each_entry_rcu(ptype,
			&ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) {
3725
		if (ptype->type == type &&
3726 3727
		    (ptype->dev == null_or_dev || ptype->dev == skb->dev ||
		     ptype->dev == orig_dev)) {
3728
			if (pt_prev)
D
David S. Miller 已提交
3729
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3730 3731 3732 3733 3734
			pt_prev = ptype;
		}
	}

	if (pt_prev) {
3735
		if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
3736
			goto drop;
3737 3738
		else
			ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3739
	} else {
3740
drop:
3741
		atomic_long_inc(&skb->dev->rx_dropped);
L
Linus Torvalds 已提交
3742 3743 3744 3745 3746 3747 3748
		kfree_skb(skb);
		/* Jamal, now you will not able to escape explaining
		 * me how you were going to use this. :-)
		 */
		ret = NET_RX_DROP;
	}

3749
unlock:
L
Linus Torvalds 已提交
3750
	rcu_read_unlock();
3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775
	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
3776 3777
	return ret;
}
T
Tom Herbert 已提交
3778

3779
static int netif_receive_skb_internal(struct sk_buff *skb)
T
Tom Herbert 已提交
3780
{
3781
	net_timestamp_check(netdev_tstamp_prequeue, skb);
E
Eric Dumazet 已提交
3782

3783 3784 3785
	if (skb_defer_rx_timestamp(skb))
		return NET_RX_SUCCESS;

E
Eric Dumazet 已提交
3786
#ifdef CONFIG_RPS
3787
	if (static_key_false(&rps_needed)) {
E
Eric Dumazet 已提交
3788 3789
		struct rps_dev_flow voidflow, *rflow = &voidflow;
		int cpu, ret;
T
Tom Herbert 已提交
3790

E
Eric Dumazet 已提交
3791 3792 3793
		rcu_read_lock();

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
T
Tom Herbert 已提交
3794

E
Eric Dumazet 已提交
3795 3796 3797
		if (cpu >= 0) {
			ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
			rcu_read_unlock();
3798
			return ret;
E
Eric Dumazet 已提交
3799
		}
3800
		rcu_read_unlock();
T
Tom Herbert 已提交
3801
	}
3802
#endif
3803
	return __netif_receive_skb(skb);
T
Tom Herbert 已提交
3804
}
3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826

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

	return netif_receive_skb_internal(skb);
}
E
Eric Dumazet 已提交
3827
EXPORT_SYMBOL(netif_receive_skb);
L
Linus Torvalds 已提交
3828

E
Eric Dumazet 已提交
3829 3830 3831
/* Network device is going away, flush any packets still pending
 * Called with irqs disabled.
 */
3832
static void flush_backlog(void *arg)
3833
{
3834
	struct net_device *dev = arg;
E
Eric Dumazet 已提交
3835
	struct softnet_data *sd = &__get_cpu_var(softnet_data);
3836 3837
	struct sk_buff *skb, *tmp;

E
Eric Dumazet 已提交
3838
	rps_lock(sd);
3839
	skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
3840
		if (skb->dev == dev) {
E
Eric Dumazet 已提交
3841
			__skb_unlink(skb, &sd->input_pkt_queue);
3842
			kfree_skb(skb);
3843
			input_queue_head_incr(sd);
3844
		}
3845
	}
E
Eric Dumazet 已提交
3846
	rps_unlock(sd);
3847 3848 3849 3850 3851

	skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
		if (skb->dev == dev) {
			__skb_unlink(skb, &sd->process_queue);
			kfree_skb(skb);
3852
			input_queue_head_incr(sd);
3853 3854
		}
	}
3855 3856
}

3857 3858
static int napi_gro_complete(struct sk_buff *skb)
{
3859
	struct packet_offload *ptype;
3860
	__be16 type = skb->protocol;
3861
	struct list_head *head = &offload_base;
3862 3863
	int err = -ENOENT;

3864 3865
	BUILD_BUG_ON(sizeof(struct napi_gro_cb) > sizeof(skb->cb));

3866 3867
	if (NAPI_GRO_CB(skb)->count == 1) {
		skb_shinfo(skb)->gso_size = 0;
3868
		goto out;
3869
	}
3870 3871 3872

	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
3873
		if (ptype->type != type || !ptype->callbacks.gro_complete)
3874 3875
			continue;

3876
		err = ptype->callbacks.gro_complete(skb, 0);
3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887
		break;
	}
	rcu_read_unlock();

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

out:
3888
	return netif_receive_skb_internal(skb);
3889 3890
}

3891 3892 3893 3894 3895
/* 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)
3896
{
3897
	struct sk_buff *skb, *prev = NULL;
3898

3899 3900 3901 3902 3903 3904 3905
	/* 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) {
3906
		skb->next = NULL;
3907 3908 3909 3910 3911

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

		prev = skb->prev;
3912
		napi_gro_complete(skb);
3913
		napi->gro_count--;
3914 3915 3916 3917
	}

	napi->gro_list = NULL;
}
E
Eric Dumazet 已提交
3918
EXPORT_SYMBOL(napi_gro_flush);
3919

E
Eric Dumazet 已提交
3920 3921 3922 3923
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;
3924
	u32 hash = skb_get_hash_raw(skb);
E
Eric Dumazet 已提交
3925 3926 3927 3928

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

3929 3930 3931 3932 3933 3934 3935
		NAPI_GRO_CB(p)->flush = 0;

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

E
Eric Dumazet 已提交
3936 3937 3938 3939
		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),
3940
						      skb_mac_header(skb));
E
Eric Dumazet 已提交
3941 3942
		else if (!diffs)
			diffs = memcmp(skb_mac_header(p),
3943
				       skb_mac_header(skb),
E
Eric Dumazet 已提交
3944 3945 3946 3947 3948
				       maclen);
		NAPI_GRO_CB(p)->same_flow = !diffs;
	}
}

3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962
static void skb_gro_reset_offset(struct sk_buff *skb)
{
	const struct skb_shared_info *pinfo = skb_shinfo(skb);
	const skb_frag_t *frag0 = &pinfo->frags[0];

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

	if (skb_mac_header(skb) == skb_tail_pointer(skb) &&
	    pinfo->nr_frags &&
	    !PageHighMem(skb_frag_page(frag0))) {
		NAPI_GRO_CB(skb)->frag0 = skb_frag_address(frag0);
		NAPI_GRO_CB(skb)->frag0_len = skb_frag_size(frag0);
E
Eric Dumazet 已提交
3963 3964 3965
	}
}

3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986
static void gro_pull_from_frag0(struct sk_buff *skb, int grow)
{
	struct skb_shared_info *pinfo = skb_shinfo(skb);

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

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

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

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

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

R
Rami Rosen 已提交
3987
static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3988 3989
{
	struct sk_buff **pp = NULL;
3990
	struct packet_offload *ptype;
3991
	__be16 type = skb->protocol;
3992
	struct list_head *head = &offload_base;
H
Herbert Xu 已提交
3993
	int same_flow;
3994
	enum gro_result ret;
3995
	int grow;
3996

3997
	if (!(skb->dev->features & NETIF_F_GRO))
3998 3999
		goto normal;

4000
	if (skb_is_gso(skb) || skb_has_frag_list(skb) || skb->csum_bad)
4001 4002
		goto normal;

E
Eric Dumazet 已提交
4003
	gro_list_prepare(napi, skb);
T
Tom Herbert 已提交
4004

4005 4006
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
4007
		if (ptype->type != type || !ptype->callbacks.gro_receive)
4008 4009
			continue;

4010
		skb_set_network_header(skb, skb_gro_offset(skb));
4011
		skb_reset_mac_len(skb);
4012 4013
		NAPI_GRO_CB(skb)->same_flow = 0;
		NAPI_GRO_CB(skb)->flush = 0;
H
Herbert Xu 已提交
4014
		NAPI_GRO_CB(skb)->free = 0;
4015
		NAPI_GRO_CB(skb)->udp_mark = 0;
4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031

		/* Setup for GRO checksum validation */
		switch (skb->ip_summed) {
		case CHECKSUM_COMPLETE:
			NAPI_GRO_CB(skb)->csum = skb->csum;
			NAPI_GRO_CB(skb)->csum_valid = 1;
			NAPI_GRO_CB(skb)->csum_cnt = 0;
			break;
		case CHECKSUM_UNNECESSARY:
			NAPI_GRO_CB(skb)->csum_cnt = skb->csum_level + 1;
			NAPI_GRO_CB(skb)->csum_valid = 0;
			break;
		default:
			NAPI_GRO_CB(skb)->csum_cnt = 0;
			NAPI_GRO_CB(skb)->csum_valid = 0;
		}
4032

4033
		pp = ptype->callbacks.gro_receive(&napi->gro_list, skb);
4034 4035 4036 4037 4038 4039 4040
		break;
	}
	rcu_read_unlock();

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

H
Herbert Xu 已提交
4041
	same_flow = NAPI_GRO_CB(skb)->same_flow;
4042
	ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED;
H
Herbert Xu 已提交
4043

4044 4045 4046 4047 4048 4049
	if (pp) {
		struct sk_buff *nskb = *pp;

		*pp = nskb->next;
		nskb->next = NULL;
		napi_gro_complete(nskb);
4050
		napi->gro_count--;
4051 4052
	}

H
Herbert Xu 已提交
4053
	if (same_flow)
4054 4055
		goto ok;

4056
	if (NAPI_GRO_CB(skb)->flush)
4057 4058
		goto normal;

4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072
	if (unlikely(napi->gro_count >= MAX_GRO_SKBS)) {
		struct sk_buff *nskb = napi->gro_list;

		/* locate the end of the list to select the 'oldest' flow */
		while (nskb->next) {
			pp = &nskb->next;
			nskb = *pp;
		}
		*pp = NULL;
		nskb->next = NULL;
		napi_gro_complete(nskb);
	} else {
		napi->gro_count++;
	}
4073
	NAPI_GRO_CB(skb)->count = 1;
4074
	NAPI_GRO_CB(skb)->age = jiffies;
4075
	NAPI_GRO_CB(skb)->last = skb;
4076
	skb_shinfo(skb)->gso_size = skb_gro_len(skb);
4077 4078
	skb->next = napi->gro_list;
	napi->gro_list = skb;
4079
	ret = GRO_HELD;
4080

4081
pull:
4082 4083 4084
	grow = skb_gro_offset(skb) - skb_headlen(skb);
	if (grow > 0)
		gro_pull_from_frag0(skb, grow);
4085
ok:
4086
	return ret;
4087 4088

normal:
4089 4090
	ret = GRO_NORMAL;
	goto pull;
H
Herbert Xu 已提交
4091
}
4092

4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104
struct packet_offload *gro_find_receive_by_type(__be16 type)
{
	struct list_head *offload_head = &offload_base;
	struct packet_offload *ptype;

	list_for_each_entry_rcu(ptype, offload_head, list) {
		if (ptype->type != type || !ptype->callbacks.gro_receive)
			continue;
		return ptype;
	}
	return NULL;
}
4105
EXPORT_SYMBOL(gro_find_receive_by_type);
4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118

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

	list_for_each_entry_rcu(ptype, offload_head, list) {
		if (ptype->type != type || !ptype->callbacks.gro_complete)
			continue;
		return ptype;
	}
	return NULL;
}
4119
EXPORT_SYMBOL(gro_find_complete_by_type);
H
Herbert Xu 已提交
4120

R
Rami Rosen 已提交
4121
static gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
H
Herbert Xu 已提交
4122
{
4123 4124
	switch (ret) {
	case GRO_NORMAL:
4125
		if (netif_receive_skb_internal(skb))
4126 4127
			ret = GRO_DROP;
		break;
H
Herbert Xu 已提交
4128

4129
	case GRO_DROP:
H
Herbert Xu 已提交
4130 4131
		kfree_skb(skb);
		break;
4132

4133
	case GRO_MERGED_FREE:
4134 4135 4136 4137
		if (NAPI_GRO_CB(skb)->free == NAPI_GRO_FREE_STOLEN_HEAD)
			kmem_cache_free(skbuff_head_cache, skb);
		else
			__kfree_skb(skb);
4138 4139
		break;

4140 4141 4142
	case GRO_HELD:
	case GRO_MERGED:
		break;
H
Herbert Xu 已提交
4143 4144
	}

4145
	return ret;
4146 4147
}

4148
gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
4149
{
4150
	trace_napi_gro_receive_entry(skb);
4151

4152 4153
	skb_gro_reset_offset(skb);

E
Eric Dumazet 已提交
4154
	return napi_skb_finish(dev_gro_receive(napi, skb), skb);
4155 4156 4157
}
EXPORT_SYMBOL(napi_gro_receive);

S
stephen hemminger 已提交
4158
static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb)
4159 4160
{
	__skb_pull(skb, skb_headlen(skb));
4161 4162
	/* restore the reserve we had after netdev_alloc_skb_ip_align() */
	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN - skb_headroom(skb));
4163
	skb->vlan_tci = 0;
H
Herbert Xu 已提交
4164
	skb->dev = napi->dev;
A
Andy Gospodarek 已提交
4165
	skb->skb_iif = 0;
4166 4167
	skb->encapsulation = 0;
	skb_shinfo(skb)->gso_type = 0;
4168
	skb->truesize = SKB_TRUESIZE(skb_end_offset(skb));
4169 4170 4171 4172

	napi->skb = skb;
}

4173
struct sk_buff *napi_get_frags(struct napi_struct *napi)
H
Herbert Xu 已提交
4174 4175 4176 4177
{
	struct sk_buff *skb = napi->skb;

	if (!skb) {
4178
		skb = netdev_alloc_skb_ip_align(napi->dev, GRO_MAX_HEAD);
E
Eric Dumazet 已提交
4179
		napi->skb = skb;
4180
	}
4181 4182
	return skb;
}
4183
EXPORT_SYMBOL(napi_get_frags);
4184

4185 4186 4187
static gro_result_t napi_frags_finish(struct napi_struct *napi,
				      struct sk_buff *skb,
				      gro_result_t ret)
4188
{
4189 4190
	switch (ret) {
	case GRO_NORMAL:
4191 4192 4193 4194
	case GRO_HELD:
		__skb_push(skb, ETH_HLEN);
		skb->protocol = eth_type_trans(skb, skb->dev);
		if (ret == GRO_NORMAL && netif_receive_skb_internal(skb))
4195
			ret = GRO_DROP;
4196
		break;
H
Herbert Xu 已提交
4197

4198 4199 4200 4201
	case GRO_DROP:
	case GRO_MERGED_FREE:
		napi_reuse_skb(napi, skb);
		break;
4202 4203 4204

	case GRO_MERGED:
		break;
4205
	}
H
Herbert Xu 已提交
4206

4207
	return ret;
H
Herbert Xu 已提交
4208
}
4209

4210 4211 4212 4213
/* Upper GRO stack assumes network header starts at gro_offset=0
 * Drivers could call both napi_gro_frags() and napi_gro_receive()
 * We copy ethernet header into skb->data to have a common layout.
 */
E
Eric Dumazet 已提交
4214
static struct sk_buff *napi_frags_skb(struct napi_struct *napi)
4215 4216
{
	struct sk_buff *skb = napi->skb;
4217 4218
	const struct ethhdr *eth;
	unsigned int hlen = sizeof(*eth);
4219 4220 4221

	napi->skb = NULL;

4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235
	skb_reset_mac_header(skb);
	skb_gro_reset_offset(skb);

	eth = skb_gro_header_fast(skb, 0);
	if (unlikely(skb_gro_header_hard(skb, hlen))) {
		eth = skb_gro_header_slow(skb, hlen, 0);
		if (unlikely(!eth)) {
			napi_reuse_skb(napi, skb);
			return NULL;
		}
	} else {
		gro_pull_from_frag0(skb, hlen);
		NAPI_GRO_CB(skb)->frag0 += hlen;
		NAPI_GRO_CB(skb)->frag0_len -= hlen;
4236
	}
4237 4238 4239 4240 4241 4242 4243 4244
	__skb_pull(skb, hlen);

	/*
	 * This works because the only protocols we care about don't require
	 * special handling.
	 * We'll fix it up properly in napi_frags_finish()
	 */
	skb->protocol = eth->h_proto;
4245 4246 4247 4248

	return skb;
}

4249
gro_result_t napi_gro_frags(struct napi_struct *napi)
4250
{
4251
	struct sk_buff *skb = napi_frags_skb(napi);
4252 4253

	if (!skb)
4254
		return GRO_DROP;
4255

4256 4257
	trace_napi_gro_frags_entry(skb);

E
Eric Dumazet 已提交
4258
	return napi_frags_finish(napi, skb, dev_gro_receive(napi, skb));
4259
}
H
Herbert Xu 已提交
4260 4261
EXPORT_SYMBOL(napi_gro_frags);

T
Tom Herbert 已提交
4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286
/* Compute the checksum from gro_offset and return the folded value
 * after adding in any pseudo checksum.
 */
__sum16 __skb_gro_checksum_complete(struct sk_buff *skb)
{
	__wsum wsum;
	__sum16 sum;

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

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

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

	return sum;
}
EXPORT_SYMBOL(__skb_gro_checksum_complete);

4287
/*
4288
 * net_rps_action_and_irq_enable sends any pending IPI's for rps.
4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305
 * 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))
4306
				smp_call_function_single_async(remsd->cpu,
4307
							   &remsd->csd);
4308 4309 4310 4311 4312 4313 4314
			remsd = next;
		}
	} else
#endif
		local_irq_enable();
}

4315
static int process_backlog(struct napi_struct *napi, int quota)
L
Linus Torvalds 已提交
4316 4317
{
	int work = 0;
E
Eric Dumazet 已提交
4318
	struct softnet_data *sd = container_of(napi, struct softnet_data, backlog);
L
Linus Torvalds 已提交
4319

4320 4321 4322 4323 4324 4325 4326 4327 4328
#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
4329
	napi->weight = weight_p;
4330
	local_irq_disable();
4331
	while (1) {
L
Linus Torvalds 已提交
4332
		struct sk_buff *skb;
4333 4334 4335 4336 4337

		while ((skb = __skb_dequeue(&sd->process_queue))) {
			local_irq_enable();
			__netif_receive_skb(skb);
			local_irq_disable();
4338 4339 4340 4341 4342
			input_queue_head_incr(sd);
			if (++work >= quota) {
				local_irq_enable();
				return work;
			}
4343
		}
L
Linus Torvalds 已提交
4344

E
Eric Dumazet 已提交
4345
		rps_lock(sd);
4346
		if (skb_queue_empty(&sd->input_pkt_queue)) {
E
Eric Dumazet 已提交
4347 4348 4349
			/*
			 * Inline a custom version of __napi_complete().
			 * only current cpu owns and manipulates this napi,
4350 4351 4352
			 * and NAPI_STATE_SCHED is the only possible flag set
			 * on backlog.
			 * We can use a plain write instead of clear_bit(),
E
Eric Dumazet 已提交
4353 4354 4355 4356
			 * and we dont need an smp_mb() memory barrier.
			 */
			list_del(&napi->poll_list);
			napi->state = 0;
4357
			rps_unlock(sd);
E
Eric Dumazet 已提交
4358

4359
			break;
4360
		}
4361 4362 4363

		skb_queue_splice_tail_init(&sd->input_pkt_queue,
					   &sd->process_queue);
E
Eric Dumazet 已提交
4364
		rps_unlock(sd);
4365 4366
	}
	local_irq_enable();
L
Linus Torvalds 已提交
4367

4368 4369
	return work;
}
L
Linus Torvalds 已提交
4370

4371 4372
/**
 * __napi_schedule - schedule for receive
4373
 * @n: entry to schedule
4374 4375 4376
 *
 * The entry's receive function will be scheduled to run
 */
H
Harvey Harrison 已提交
4377
void __napi_schedule(struct napi_struct *n)
4378 4379
{
	unsigned long flags;
L
Linus Torvalds 已提交
4380

4381
	local_irq_save(flags);
E
Eric Dumazet 已提交
4382
	____napi_schedule(&__get_cpu_var(softnet_data), n);
4383
	local_irq_restore(flags);
L
Linus Torvalds 已提交
4384
}
4385 4386
EXPORT_SYMBOL(__napi_schedule);

4387 4388 4389 4390 4391 4392
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);
4393
	smp_mb__before_atomic();
4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408
	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;

4409
	napi_gro_flush(n, false);
4410 4411 4412 4413 4414 4415
	local_irq_save(flags);
	__napi_complete(n);
	local_irq_restore(flags);
}
EXPORT_SYMBOL(napi_complete);

E
Eliezer Tamir 已提交
4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467
/* 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);

4468 4469 4470 4471
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);
4472
	napi->gro_count = 0;
4473
	napi->gro_list = NULL;
H
Herbert Xu 已提交
4474
	napi->skb = NULL;
4475
	napi->poll = poll;
E
Eric Dumazet 已提交
4476 4477 4478
	if (weight > NAPI_POLL_WEIGHT)
		pr_err_once("netif_napi_add() called with weight %d on device %s\n",
			    weight, dev->name);
4479 4480 4481
	napi->weight = weight;
	list_add(&napi->dev_list, &dev->napi_list);
	napi->dev = dev;
H
Herbert Xu 已提交
4482
#ifdef CONFIG_NETPOLL
4483 4484 4485 4486 4487 4488 4489 4490 4491
	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)
{
4492
	list_del_init(&napi->dev_list);
4493
	napi_free_frags(napi);
4494

E
Eric Dumazet 已提交
4495
	kfree_skb_list(napi->gro_list);
4496
	napi->gro_list = NULL;
4497
	napi->gro_count = 0;
4498 4499 4500
}
EXPORT_SYMBOL(netif_napi_del);

L
Linus Torvalds 已提交
4501 4502
static void net_rx_action(struct softirq_action *h)
{
4503
	struct softnet_data *sd = &__get_cpu_var(softnet_data);
4504
	unsigned long time_limit = jiffies + 2;
4505
	int budget = netdev_budget;
4506 4507
	void *have;

L
Linus Torvalds 已提交
4508 4509
	local_irq_disable();

4510
	while (!list_empty(&sd->poll_list)) {
4511 4512
		struct napi_struct *n;
		int work, weight;
L
Linus Torvalds 已提交
4513

4514
		/* If softirq window is exhuasted then punt.
4515 4516
		 * Allow this to run for 2 jiffies since which will allow
		 * an average latency of 1.5/HZ.
4517
		 */
4518
		if (unlikely(budget <= 0 || time_after_eq(jiffies, time_limit)))
L
Linus Torvalds 已提交
4519 4520 4521 4522
			goto softnet_break;

		local_irq_enable();

4523 4524 4525 4526 4527
		/* 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.
		 */
4528
		n = list_first_entry(&sd->poll_list, struct napi_struct, poll_list);
L
Linus Torvalds 已提交
4529

4530 4531 4532 4533
		have = netpoll_poll_lock(n);

		weight = n->weight;

4534 4535 4536 4537
		/* 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
4538
		 * accidentally calling ->poll() when NAPI is not scheduled.
4539 4540
		 */
		work = 0;
4541
		if (test_bit(NAPI_STATE_SCHED, &n->state)) {
4542
			work = n->poll(n, weight);
4543 4544
			trace_napi_poll(n);
		}
4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556

		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.
		 */
4557
		if (unlikely(work == weight)) {
4558 4559 4560 4561
			if (unlikely(napi_disable_pending(n))) {
				local_irq_enable();
				napi_complete(n);
				local_irq_disable();
4562 4563 4564 4565 4566 4567 4568 4569 4570
			} 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();
				}
4571
				list_move_tail(&n->poll_list, &sd->poll_list);
4572
			}
4573
		}
4574 4575

		netpoll_poll_unlock(have);
L
Linus Torvalds 已提交
4576 4577
	}
out:
4578
	net_rps_action_and_irq_enable(sd);
T
Tom Herbert 已提交
4579

4580 4581 4582 4583 4584
#ifdef CONFIG_NET_DMA
	/*
	 * There may not be any more sk_buffs coming right now, so push
	 * any pending DMA copies to hardware
	 */
4585
	dma_issue_pending_all();
4586
#endif
4587

L
Linus Torvalds 已提交
4588 4589 4590
	return;

softnet_break:
C
Changli Gao 已提交
4591
	sd->time_squeeze++;
L
Linus Torvalds 已提交
4592 4593 4594 4595
	__raise_softirq_irqoff(NET_RX_SOFTIRQ);
	goto out;
}

4596
struct netdev_adjacent {
J
Jiri Pirko 已提交
4597
	struct net_device *dev;
4598 4599

	/* upper master flag, there can only be one master device per list */
J
Jiri Pirko 已提交
4600
	bool master;
4601 4602 4603 4604

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

4605 4606 4607
	/* private field for the users */
	void *private;

J
Jiri Pirko 已提交
4608 4609 4610 4611
	struct list_head list;
	struct rcu_head rcu;
};

4612 4613
static struct netdev_adjacent *__netdev_find_adj(struct net_device *dev,
						 struct net_device *adj_dev,
4614
						 struct list_head *adj_list)
J
Jiri Pirko 已提交
4615
{
4616 4617
	struct netdev_adjacent *adj;

4618
	list_for_each_entry(adj, adj_list, list) {
4619 4620
		if (adj->dev == adj_dev)
			return adj;
J
Jiri Pirko 已提交
4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638
	}
	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();

4639
	return __netdev_find_adj(dev, upper_dev, &dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4640 4641 4642 4643 4644 4645 4646 4647 4648 4649
}
EXPORT_SYMBOL(netdev_has_upper_dev);

/**
 * netdev_has_any_upper_dev - Check if device is linked to some device
 * @dev: device
 *
 * Find out if a device is linked to an upper device and return true in case
 * it is. The caller must hold the RTNL lock.
 */
S
stephen hemminger 已提交
4650
static bool netdev_has_any_upper_dev(struct net_device *dev)
J
Jiri Pirko 已提交
4651 4652 4653
{
	ASSERT_RTNL();

4654
	return !list_empty(&dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665
}

/**
 * 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)
{
4666
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4667 4668 4669

	ASSERT_RTNL();

4670
	if (list_empty(&dev->adj_list.upper))
J
Jiri Pirko 已提交
4671 4672
		return NULL;

4673
	upper = list_first_entry(&dev->adj_list.upper,
4674
				 struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4675 4676 4677 4678 4679 4680
	if (likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get);

4681 4682 4683 4684 4685 4686 4687 4688 4689 4690
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);

4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716
/**
 * netdev_upper_get_next_dev_rcu - Get the next dev from upper list
 * @dev: device
 * @iter: list_head ** of the current position
 *
 * Gets the next device from the dev's upper list, starting from iter
 * position. The caller must hold RCU read lock.
 */
struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
						 struct list_head **iter)
{
	struct netdev_adjacent *upper;

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

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

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

	*iter = &upper->list;

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

4717 4718
/**
 * netdev_all_upper_get_next_dev_rcu - Get the next dev from upper list
4719 4720 4721 4722 4723 4724
 * @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.
 */
4725 4726
struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
						     struct list_head **iter)
4727 4728 4729
{
	struct netdev_adjacent *upper;

4730
	WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_rtnl_is_held());
4731 4732 4733

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

4734
	if (&upper->list == &dev->all_adj_list.upper)
4735 4736 4737 4738 4739 4740
		return NULL;

	*iter = &upper->list;

	return upper->dev;
}
4741
EXPORT_SYMBOL(netdev_all_upper_get_next_dev_rcu);
4742

4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763
/**
 * 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;

4764
	*iter = lower->list.next;
4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791

	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;

4792
	*iter = &lower->list;
4793 4794 4795 4796 4797

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

4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823
/**
 * netdev_lower_get_next - Get the next device from the lower neighbour
 *                         list
 * @dev: device
 * @iter: list_head ** of the current position
 *
 * Gets the next netdev_adjacent from the dev's lower neighbour
 * list, starting from iter position. The caller must hold RTNL lock or
 * its own locking that guarantees that the neighbour lower
 * list will remain unchainged.
 */
void *netdev_lower_get_next(struct net_device *dev, struct list_head **iter)
{
	struct netdev_adjacent *lower;

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

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

	*iter = &lower->list;

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

4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844
/**
 * 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
4845 4846 4847 4848 4849 4850 4851 4852 4853
/**
 * 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)
{
4854
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4855

4856
	upper = list_first_or_null_rcu(&dev->adj_list.upper,
4857
				       struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4858 4859 4860 4861 4862 4863
	if (upper && likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get_rcu);

4864
static int netdev_adjacent_sysfs_add(struct net_device *dev,
4865 4866 4867 4868 4869 4870 4871 4872 4873
			      struct net_device *adj_dev,
			      struct list_head *dev_list)
{
	char linkname[IFNAMSIZ+7];
	sprintf(linkname, dev_list == &dev->adj_list.upper ?
		"upper_%s" : "lower_%s", adj_dev->name);
	return sysfs_create_link(&(dev->dev.kobj), &(adj_dev->dev.kobj),
				 linkname);
}
4874
static void netdev_adjacent_sysfs_del(struct net_device *dev,
4875 4876 4877 4878 4879 4880 4881 4882 4883
			       char *name,
			       struct list_head *dev_list)
{
	char linkname[IFNAMSIZ+7];
	sprintf(linkname, dev_list == &dev->adj_list.upper ?
		"upper_%s" : "lower_%s", name);
	sysfs_remove_link(&(dev->dev.kobj), linkname);
}

4884 4885 4886 4887 4888 4889 4890 4891
static inline bool netdev_adjacent_is_neigh_list(struct net_device *dev,
						 struct net_device *adj_dev,
						 struct list_head *dev_list)
{
	return (dev_list == &dev->adj_list.upper ||
		dev_list == &dev->adj_list.lower) &&
		net_eq(dev_net(dev), dev_net(adj_dev));
}
4892

4893 4894
static int __netdev_adjacent_dev_insert(struct net_device *dev,
					struct net_device *adj_dev,
4895
					struct list_head *dev_list,
4896
					void *private, bool master)
4897 4898
{
	struct netdev_adjacent *adj;
4899
	int ret;
4900

4901
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914

	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;
4915
	adj->private = private;
4916
	dev_hold(adj_dev);
4917 4918 4919

	pr_debug("dev_hold for %s, because of link added from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
4920

4921
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list)) {
4922
		ret = netdev_adjacent_sysfs_add(dev, adj_dev, dev_list);
4923 4924 4925 4926
		if (ret)
			goto free_adj;
	}

4927
	/* Ensure that master link is always the first item in list. */
4928 4929 4930 4931
	if (master) {
		ret = sysfs_create_link(&(dev->dev.kobj),
					&(adj_dev->dev.kobj), "master");
		if (ret)
4932
			goto remove_symlinks;
4933

4934
		list_add_rcu(&adj->list, dev_list);
4935
	} else {
4936
		list_add_tail_rcu(&adj->list, dev_list);
4937
	}
4938 4939

	return 0;
4940

4941
remove_symlinks:
4942
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
4943
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
4944 4945
free_adj:
	kfree(adj);
4946
	dev_put(adj_dev);
4947 4948

	return ret;
4949 4950
}

S
stephen hemminger 已提交
4951 4952 4953
static void __netdev_adjacent_dev_remove(struct net_device *dev,
					 struct net_device *adj_dev,
					 struct list_head *dev_list)
4954 4955 4956
{
	struct netdev_adjacent *adj;

4957
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
4958

4959 4960 4961
	if (!adj) {
		pr_err("tried to remove device %s from %s\n",
		       dev->name, adj_dev->name);
4962
		BUG();
4963
	}
4964 4965

	if (adj->ref_nr > 1) {
4966 4967
		pr_debug("%s to %s ref_nr-- = %d\n", dev->name, adj_dev->name,
			 adj->ref_nr-1);
4968 4969 4970 4971
		adj->ref_nr--;
		return;
	}

4972 4973 4974
	if (adj->master)
		sysfs_remove_link(&(dev->dev.kobj), "master");

4975
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
4976
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
4977

4978
	list_del_rcu(&adj->list);
4979 4980
	pr_debug("dev_put for %s, because link removed from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
4981 4982 4983 4984
	dev_put(adj_dev);
	kfree_rcu(adj, rcu);
}

S
stephen hemminger 已提交
4985 4986 4987 4988 4989
static int __netdev_adjacent_dev_link_lists(struct net_device *dev,
					    struct net_device *upper_dev,
					    struct list_head *up_list,
					    struct list_head *down_list,
					    void *private, bool master)
4990 4991 4992
{
	int ret;

4993 4994
	ret = __netdev_adjacent_dev_insert(dev, upper_dev, up_list, private,
					   master);
4995 4996 4997
	if (ret)
		return ret;

4998 4999
	ret = __netdev_adjacent_dev_insert(upper_dev, dev, down_list, private,
					   false);
5000
	if (ret) {
5001
		__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
5002 5003 5004 5005 5006 5007
		return ret;
	}

	return 0;
}

S
stephen hemminger 已提交
5008 5009
static int __netdev_adjacent_dev_link(struct net_device *dev,
				      struct net_device *upper_dev)
5010
{
5011 5012 5013
	return __netdev_adjacent_dev_link_lists(dev, upper_dev,
						&dev->all_adj_list.upper,
						&upper_dev->all_adj_list.lower,
5014
						NULL, false);
5015 5016
}

S
stephen hemminger 已提交
5017 5018 5019 5020
static void __netdev_adjacent_dev_unlink_lists(struct net_device *dev,
					       struct net_device *upper_dev,
					       struct list_head *up_list,
					       struct list_head *down_list)
5021
{
5022 5023
	__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
	__netdev_adjacent_dev_remove(upper_dev, dev, down_list);
5024 5025
}

S
stephen hemminger 已提交
5026 5027
static void __netdev_adjacent_dev_unlink(struct net_device *dev,
					 struct net_device *upper_dev)
5028
{
5029 5030 5031 5032 5033
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev,
					   &dev->all_adj_list.upper,
					   &upper_dev->all_adj_list.lower);
}

S
stephen hemminger 已提交
5034 5035 5036
static int __netdev_adjacent_dev_link_neighbour(struct net_device *dev,
						struct net_device *upper_dev,
						void *private, bool master)
5037 5038 5039 5040 5041 5042 5043 5044 5045
{
	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,
5046
					       private, master);
5047 5048 5049 5050 5051 5052
	if (ret) {
		__netdev_adjacent_dev_unlink(dev, upper_dev);
		return ret;
	}

	return 0;
5053 5054
}

S
stephen hemminger 已提交
5055 5056
static void __netdev_adjacent_dev_unlink_neighbour(struct net_device *dev,
						   struct net_device *upper_dev)
5057 5058 5059 5060 5061 5062
{
	__netdev_adjacent_dev_unlink(dev, upper_dev);
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev,
					   &dev->adj_list.upper,
					   &upper_dev->adj_list.lower);
}
5063

J
Jiri Pirko 已提交
5064
static int __netdev_upper_dev_link(struct net_device *dev,
5065 5066
				   struct net_device *upper_dev, bool master,
				   void *private)
J
Jiri Pirko 已提交
5067
{
5068 5069
	struct netdev_adjacent *i, *j, *to_i, *to_j;
	int ret = 0;
J
Jiri Pirko 已提交
5070 5071 5072 5073 5074 5075 5076

	ASSERT_RTNL();

	if (dev == upper_dev)
		return -EBUSY;

	/* To prevent loops, check if dev is not upper device to upper_dev. */
5077
	if (__netdev_find_adj(upper_dev, dev, &upper_dev->all_adj_list.upper))
J
Jiri Pirko 已提交
5078 5079
		return -EBUSY;

5080
	if (__netdev_find_adj(dev, upper_dev, &dev->all_adj_list.upper))
J
Jiri Pirko 已提交
5081 5082 5083 5084 5085
		return -EEXIST;

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

5086 5087
	ret = __netdev_adjacent_dev_link_neighbour(dev, upper_dev, private,
						   master);
5088 5089
	if (ret)
		return ret;
J
Jiri Pirko 已提交
5090

5091
	/* Now that we linked these devs, make all the upper_dev's
5092
	 * all_adj_list.upper visible to every dev's all_adj_list.lower an
5093 5094 5095
	 * versa, and don't forget the devices itself. All of these
	 * links are non-neighbours.
	 */
5096 5097 5098 5099
	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);
5100 5101 5102 5103 5104 5105 5106
			ret = __netdev_adjacent_dev_link(i->dev, j->dev);
			if (ret)
				goto rollback_mesh;
		}
	}

	/* add dev to every upper_dev's upper device */
5107 5108 5109
	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);
5110 5111 5112 5113 5114 5115
		ret = __netdev_adjacent_dev_link(dev, i->dev);
		if (ret)
			goto rollback_upper_mesh;
	}

	/* add upper_dev to every dev's lower device */
5116 5117 5118
	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);
5119 5120 5121 5122
		ret = __netdev_adjacent_dev_link(i->dev, upper_dev);
		if (ret)
			goto rollback_lower_mesh;
	}
J
Jiri Pirko 已提交
5123

5124
	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
J
Jiri Pirko 已提交
5125
	return 0;
5126 5127 5128

rollback_lower_mesh:
	to_i = i;
5129
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
5130 5131 5132 5133 5134 5135 5136 5137 5138
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);
	}

	i = NULL;

rollback_upper_mesh:
	to_i = i;
5139
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list) {
5140 5141 5142 5143 5144 5145 5146 5147 5148 5149
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(dev, i->dev);
	}

	i = j = NULL;

rollback_mesh:
	to_i = i;
	to_j = j;
5150 5151
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list) {
5152 5153 5154 5155 5156 5157 5158 5159
			if (i == to_i && j == to_j)
				break;
			__netdev_adjacent_dev_unlink(i->dev, j->dev);
		}
		if (i == to_i)
			break;
	}

5160
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
5161 5162

	return ret;
J
Jiri Pirko 已提交
5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177
}

/**
 * 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)
{
5178
	return __netdev_upper_dev_link(dev, upper_dev, false, NULL);
J
Jiri Pirko 已提交
5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195
}
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)
{
5196
	return __netdev_upper_dev_link(dev, upper_dev, true, NULL);
J
Jiri Pirko 已提交
5197 5198 5199
}
EXPORT_SYMBOL(netdev_master_upper_dev_link);

5200 5201 5202 5203 5204 5205 5206 5207
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
5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218
/**
 * 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)
{
5219
	struct netdev_adjacent *i, *j;
J
Jiri Pirko 已提交
5220 5221
	ASSERT_RTNL();

5222
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
5223 5224 5225 5226 5227

	/* 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.
	 */
5228 5229
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list)
5230 5231 5232 5233 5234
			__netdev_adjacent_dev_unlink(i->dev, j->dev);

	/* remove also the devices itself from lower/upper device
	 * list
	 */
5235
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
5236 5237
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);

5238
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list)
5239 5240
		__netdev_adjacent_dev_unlink(dev, i->dev);

5241
	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
J
Jiri Pirko 已提交
5242 5243 5244
}
EXPORT_SYMBOL(netdev_upper_dev_unlink);

5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294
void netdev_adjacent_add_links(struct net_device *dev)
{
	struct netdev_adjacent *iter;

	struct net *net = dev_net(dev);

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

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

void netdev_adjacent_del_links(struct net_device *dev)
{
	struct netdev_adjacent *iter;

	struct net *net = dev_net(dev);

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

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

5295
void netdev_adjacent_rename_links(struct net_device *dev, char *oldname)
5296
{
5297
	struct netdev_adjacent *iter;
5298

5299 5300
	struct net *net = dev_net(dev);

5301
	list_for_each_entry(iter, &dev->adj_list.upper, list) {
5302 5303
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
5304 5305 5306 5307 5308
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.lower);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.lower);
	}
5309

5310
	list_for_each_entry(iter, &dev->adj_list.lower, list) {
5311 5312
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
5313 5314 5315 5316 5317
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.upper);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.upper);
	}
5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334
}

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);

5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358

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

	ASSERT_RTNL();

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

	if (type_check(dev))
		max_nest++;

	return max_nest;
}
EXPORT_SYMBOL(dev_get_nest_level);

5359 5360
static void dev_change_rx_flags(struct net_device *dev, int flags)
{
5361 5362
	const struct net_device_ops *ops = dev->netdev_ops;

5363
	if (ops->ndo_change_rx_flags)
5364
		ops->ndo_change_rx_flags(dev, flags);
5365 5366
}

5367
static int __dev_set_promiscuity(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5368
{
5369
	unsigned int old_flags = dev->flags;
5370 5371
	kuid_t uid;
	kgid_t gid;
L
Linus Torvalds 已提交
5372

5373 5374
	ASSERT_RTNL();

5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385
	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;
5386 5387
			pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n",
				dev->name);
5388 5389 5390
			return -EOVERFLOW;
		}
	}
5391
	if (dev->flags != old_flags) {
5392 5393 5394
		pr_info("device %s %s promiscuous mode\n",
			dev->name,
			dev->flags & IFF_PROMISC ? "entered" : "left");
5395 5396
		if (audit_enabled) {
			current_uid_gid(&uid, &gid);
5397 5398 5399 5400 5401
			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),
5402
				from_kuid(&init_user_ns, audit_get_loginuid(current)),
5403 5404
				from_kuid(&init_user_ns, uid),
				from_kgid(&init_user_ns, gid),
5405
				audit_get_sessionid(current));
5406
		}
5407

5408
		dev_change_rx_flags(dev, IFF_PROMISC);
L
Linus Torvalds 已提交
5409
	}
5410 5411
	if (notify)
		__dev_notify_flags(dev, old_flags, IFF_PROMISC);
5412
	return 0;
L
Linus Torvalds 已提交
5413 5414
}

5415 5416 5417 5418 5419 5420 5421 5422 5423
/**
 *	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.
5424
 *	Return 0 if successful or a negative errno code on error.
5425
 */
5426
int dev_set_promiscuity(struct net_device *dev, int inc)
5427
{
5428
	unsigned int old_flags = dev->flags;
5429
	int err;
5430

5431
	err = __dev_set_promiscuity(dev, inc, true);
5432
	if (err < 0)
5433
		return err;
5434 5435
	if (dev->flags != old_flags)
		dev_set_rx_mode(dev);
5436
	return err;
5437
}
E
Eric Dumazet 已提交
5438
EXPORT_SYMBOL(dev_set_promiscuity);
5439

5440
static int __dev_set_allmulti(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5441
{
5442
	unsigned int old_flags = dev->flags, old_gflags = dev->gflags;
L
Linus Torvalds 已提交
5443

5444 5445
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5446
	dev->flags |= IFF_ALLMULTI;
5447 5448 5449 5450 5451 5452 5453 5454 5455 5456
	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;
5457 5458
			pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n",
				dev->name);
5459 5460 5461
			return -EOVERFLOW;
		}
	}
5462
	if (dev->flags ^ old_flags) {
5463
		dev_change_rx_flags(dev, IFF_ALLMULTI);
5464
		dev_set_rx_mode(dev);
5465 5466 5467
		if (notify)
			__dev_notify_flags(dev, old_flags,
					   dev->gflags ^ old_gflags);
5468
	}
5469
	return 0;
5470
}
5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488

/**
 *	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
5489
EXPORT_SYMBOL(dev_set_allmulti);
5490 5491 5492 5493

/*
 *	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
5494
 *	filtering it is put in promiscuous mode while unicast addresses
5495 5496 5497 5498
 *	are present.
 */
void __dev_set_rx_mode(struct net_device *dev)
{
5499 5500
	const struct net_device_ops *ops = dev->netdev_ops;

5501 5502 5503 5504 5505
	/* dev_open will call this function so the list will stay sane. */
	if (!(dev->flags&IFF_UP))
		return;

	if (!netif_device_present(dev))
5506
		return;
5507

5508
	if (!(dev->priv_flags & IFF_UNICAST_FLT)) {
5509 5510 5511
		/* Unicast addresses changes may only happen under the rtnl,
		 * therefore calling __dev_set_promiscuity here is safe.
		 */
5512
		if (!netdev_uc_empty(dev) && !dev->uc_promisc) {
5513
			__dev_set_promiscuity(dev, 1, false);
5514
			dev->uc_promisc = true;
5515
		} else if (netdev_uc_empty(dev) && dev->uc_promisc) {
5516
			__dev_set_promiscuity(dev, -1, false);
5517
			dev->uc_promisc = false;
5518 5519
		}
	}
5520 5521 5522

	if (ops->ndo_set_rx_mode)
		ops->ndo_set_rx_mode(dev);
5523 5524 5525 5526
}

void dev_set_rx_mode(struct net_device *dev)
{
5527
	netif_addr_lock_bh(dev);
5528
	__dev_set_rx_mode(dev);
5529
	netif_addr_unlock_bh(dev);
L
Linus Torvalds 已提交
5530 5531
}

5532 5533 5534 5535 5536 5537
/**
 *	dev_get_flags - get flags reported to userspace
 *	@dev: device
 *
 *	Get the combination of flag bits exported through APIs to userspace.
 */
5538
unsigned int dev_get_flags(const struct net_device *dev)
L
Linus Torvalds 已提交
5539
{
5540
	unsigned int flags;
L
Linus Torvalds 已提交
5541 5542 5543

	flags = (dev->flags & ~(IFF_PROMISC |
				IFF_ALLMULTI |
S
Stefan Rompf 已提交
5544 5545 5546
				IFF_RUNNING |
				IFF_LOWER_UP |
				IFF_DORMANT)) |
L
Linus Torvalds 已提交
5547 5548 5549
		(dev->gflags & (IFF_PROMISC |
				IFF_ALLMULTI));

S
Stefan Rompf 已提交
5550 5551 5552 5553 5554 5555 5556 5557
	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
5558 5559 5560

	return flags;
}
E
Eric Dumazet 已提交
5561
EXPORT_SYMBOL(dev_get_flags);
L
Linus Torvalds 已提交
5562

5563
int __dev_change_flags(struct net_device *dev, unsigned int flags)
L
Linus Torvalds 已提交
5564
{
5565
	unsigned int old_flags = dev->flags;
5566
	int ret;
L
Linus Torvalds 已提交
5567

5568 5569
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583
	/*
	 *	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.
	 */

5584 5585
	if ((old_flags ^ flags) & IFF_MULTICAST)
		dev_change_rx_flags(dev, IFF_MULTICAST);
5586

5587
	dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5588 5589 5590 5591 5592 5593 5594 5595

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

	ret = 0;
5596
	if ((old_flags ^ flags) & IFF_UP)
5597
		ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev);
L
Linus Torvalds 已提交
5598 5599

	if ((flags ^ dev->gflags) & IFF_PROMISC) {
E
Eric Dumazet 已提交
5600
		int inc = (flags & IFF_PROMISC) ? 1 : -1;
5601
		unsigned int old_flags = dev->flags;
E
Eric Dumazet 已提交
5602

L
Linus Torvalds 已提交
5603
		dev->gflags ^= IFF_PROMISC;
5604 5605 5606 5607

		if (__dev_set_promiscuity(dev, inc, false) >= 0)
			if (dev->flags != old_flags)
				dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5608 5609 5610 5611 5612 5613 5614
	}

	/* 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
5615 5616
		int inc = (flags & IFF_ALLMULTI) ? 1 : -1;

L
Linus Torvalds 已提交
5617
		dev->gflags ^= IFF_ALLMULTI;
5618
		__dev_set_allmulti(dev, inc, false);
L
Linus Torvalds 已提交
5619 5620
	}

5621 5622 5623
	return ret;
}

5624 5625
void __dev_notify_flags(struct net_device *dev, unsigned int old_flags,
			unsigned int gchanges)
5626 5627 5628
{
	unsigned int changes = dev->flags ^ old_flags;

5629
	if (gchanges)
5630
		rtmsg_ifinfo(RTM_NEWLINK, dev, gchanges, GFP_ATOMIC);
5631

5632 5633 5634 5635 5636 5637 5638 5639
	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 &&
5640 5641 5642 5643 5644 5645 5646
	    (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);
	}
5647 5648 5649 5650 5651 5652 5653 5654 5655 5656
}

/**
 *	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.
 */
5657
int dev_change_flags(struct net_device *dev, unsigned int flags)
5658
{
5659
	int ret;
5660
	unsigned int changes, old_flags = dev->flags, old_gflags = dev->gflags;
5661 5662 5663 5664 5665

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

5666
	changes = (old_flags ^ dev->flags) | (old_gflags ^ dev->gflags);
5667
	__dev_notify_flags(dev, old_flags, changes);
L
Linus Torvalds 已提交
5668 5669
	return ret;
}
E
Eric Dumazet 已提交
5670
EXPORT_SYMBOL(dev_change_flags);
L
Linus Torvalds 已提交
5671

5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682
static int __dev_set_mtu(struct net_device *dev, int new_mtu)
{
	const struct net_device_ops *ops = dev->netdev_ops;

	if (ops->ndo_change_mtu)
		return ops->ndo_change_mtu(dev, new_mtu);

	dev->mtu = new_mtu;
	return 0;
}

5683 5684 5685 5686 5687 5688 5689
/**
 *	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
5690 5691
int dev_set_mtu(struct net_device *dev, int new_mtu)
{
5692
	int err, orig_mtu;
L
Linus Torvalds 已提交
5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703

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

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

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

5704 5705 5706 5707
	err = call_netdevice_notifiers(NETDEV_PRECHANGEMTU, dev);
	err = notifier_to_errno(err);
	if (err)
		return err;
5708

5709 5710
	orig_mtu = dev->mtu;
	err = __dev_set_mtu(dev, new_mtu);
5711

5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722
	if (!err) {
		err = call_netdevice_notifiers(NETDEV_CHANGEMTU, dev);
		err = notifier_to_errno(err);
		if (err) {
			/* setting mtu back and notifying everyone again,
			 * so that they have a chance to revert changes.
			 */
			__dev_set_mtu(dev, orig_mtu);
			call_netdevice_notifiers(NETDEV_CHANGEMTU, dev);
		}
	}
L
Linus Torvalds 已提交
5723 5724
	return err;
}
E
Eric Dumazet 已提交
5725
EXPORT_SYMBOL(dev_set_mtu);
L
Linus Torvalds 已提交
5726

5727 5728 5729 5730 5731 5732 5733 5734 5735 5736 5737
/**
 *	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);

5738 5739 5740 5741 5742 5743 5744
/**
 *	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
5745 5746
int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa)
{
5747
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
5748 5749
	int err;

5750
	if (!ops->ndo_set_mac_address)
L
Linus Torvalds 已提交
5751 5752 5753 5754 5755
		return -EOPNOTSUPP;
	if (sa->sa_family != dev->type)
		return -EINVAL;
	if (!netif_device_present(dev))
		return -ENODEV;
5756
	err = ops->ndo_set_mac_address(dev, sa);
5757 5758
	if (err)
		return err;
J
Jiri Pirko 已提交
5759
	dev->addr_assign_type = NET_ADDR_SET;
5760
	call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
5761
	add_device_randomness(dev->dev_addr, dev->addr_len);
5762
	return 0;
L
Linus Torvalds 已提交
5763
}
E
Eric Dumazet 已提交
5764
EXPORT_SYMBOL(dev_set_mac_address);
L
Linus Torvalds 已提交
5765

J
Jiri Pirko 已提交
5766 5767 5768
/**
 *	dev_change_carrier - Change device carrier
 *	@dev: device
5769
 *	@new_carrier: new value
J
Jiri Pirko 已提交
5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784
 *
 *	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);

5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802
/**
 *	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
5803 5804
/**
 *	dev_new_index	-	allocate an ifindex
5805
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
5806 5807 5808 5809 5810
 *
 *	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.
 */
5811
static int dev_new_index(struct net *net)
L
Linus Torvalds 已提交
5812
{
5813
	int ifindex = net->ifindex;
L
Linus Torvalds 已提交
5814 5815 5816
	for (;;) {
		if (++ifindex <= 0)
			ifindex = 1;
5817
		if (!__dev_get_by_index(net, ifindex))
5818
			return net->ifindex = ifindex;
L
Linus Torvalds 已提交
5819 5820 5821 5822
	}
}

/* Delayed registration/unregisteration */
5823
static LIST_HEAD(net_todo_list);
5824
DECLARE_WAIT_QUEUE_HEAD(netdev_unregistering_wq);
L
Linus Torvalds 已提交
5825

5826
static void net_set_todo(struct net_device *dev)
L
Linus Torvalds 已提交
5827 5828
{
	list_add_tail(&dev->todo_list, &net_todo_list);
5829
	dev_net(dev)->dev_unreg_count++;
L
Linus Torvalds 已提交
5830 5831
}

5832
static void rollback_registered_many(struct list_head *head)
5833
{
5834
	struct net_device *dev, *tmp;
5835
	LIST_HEAD(close_head);
5836

5837 5838 5839
	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

5840
	list_for_each_entry_safe(dev, tmp, head, unreg_list) {
5841
		/* Some devices call without registering
5842 5843
		 * for initialization unwind. Remove those
		 * devices and proceed with the remaining.
5844 5845
		 */
		if (dev->reg_state == NETREG_UNINITIALIZED) {
5846 5847
			pr_debug("unregister_netdevice: device %s/%p never was registered\n",
				 dev->name, dev);
5848

5849
			WARN_ON(1);
5850 5851
			list_del(&dev->unreg_list);
			continue;
5852
		}
5853
		dev->dismantle = true;
5854
		BUG_ON(dev->reg_state != NETREG_REGISTERED);
5855
	}
5856

5857
	/* If device is running, close it first. */
5858 5859 5860
	list_for_each_entry(dev, head, unreg_list)
		list_add_tail(&dev->close_list, &close_head);
	dev_close_many(&close_head);
5861

5862
	list_for_each_entry(dev, head, unreg_list) {
5863 5864
		/* And unlink it from device chain. */
		unlist_netdevice(dev);
5865

5866 5867
		dev->reg_state = NETREG_UNREGISTERING;
	}
5868 5869 5870

	synchronize_net();

5871 5872 5873
	list_for_each_entry(dev, head, unreg_list) {
		/* Shutdown queueing discipline. */
		dev_shutdown(dev);
5874 5875


5876 5877 5878 5879
		/* Notify protocols, that we are about to destroy
		   this device. They should clean all the things.
		*/
		call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
5880

5881 5882 5883
		/*
		 *	Flush the unicast and multicast chains
		 */
5884
		dev_uc_flush(dev);
5885
		dev_mc_flush(dev);
5886

5887 5888
		if (dev->netdev_ops->ndo_uninit)
			dev->netdev_ops->ndo_uninit(dev);
5889

5890 5891 5892 5893
		if (!dev->rtnl_link_ops ||
		    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
			rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);

J
Jiri Pirko 已提交
5894 5895
		/* Notifier chain MUST detach us all upper devices. */
		WARN_ON(netdev_has_any_upper_dev(dev));
5896

5897 5898
		/* Remove entries from kobject tree */
		netdev_unregister_kobject(dev);
5899 5900 5901 5902
#ifdef CONFIG_XPS
		/* Remove XPS queueing entries */
		netif_reset_xps_queues_gt(dev, 0);
#endif
5903
	}
5904

5905
	synchronize_net();
5906

5907
	list_for_each_entry(dev, head, unreg_list)
5908 5909 5910 5911 5912 5913 5914 5915 5916
		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
5917
	list_del(&single);
5918 5919
}

5920 5921
static netdev_features_t netdev_fix_features(struct net_device *dev,
	netdev_features_t features)
5922
{
5923 5924 5925
	/* Fix illegal checksum combinations */
	if ((features & NETIF_F_HW_CSUM) &&
	    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
5926
		netdev_warn(dev, "mixed HW and IP checksum settings.\n");
5927 5928 5929
		features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
	}

5930
	/* TSO requires that SG is present as well. */
5931
	if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
5932
		netdev_dbg(dev, "Dropping TSO features since no SG feature.\n");
5933
		features &= ~NETIF_F_ALL_TSO;
5934 5935
	}

5936 5937 5938 5939 5940 5941 5942 5943 5944 5945 5946 5947 5948
	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;
	}

5949 5950 5951 5952
	/* TSO ECN requires that TSO is present as well. */
	if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN)
		features &= ~NETIF_F_TSO_ECN;

5953 5954
	/* Software GSO depends on SG. */
	if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) {
5955
		netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n");
5956 5957 5958
		features &= ~NETIF_F_GSO;
	}

5959
	/* UFO needs SG and checksumming */
5960
	if (features & NETIF_F_UFO) {
5961 5962 5963 5964
		/* 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))) {
5965
			netdev_dbg(dev,
5966
				"Dropping NETIF_F_UFO since no checksum offload features.\n");
5967 5968 5969 5970
			features &= ~NETIF_F_UFO;
		}

		if (!(features & NETIF_F_SG)) {
5971
			netdev_dbg(dev,
5972
				"Dropping NETIF_F_UFO since no NETIF_F_SG feature.\n");
5973 5974 5975 5976
			features &= ~NETIF_F_UFO;
		}
	}

J
Jiri Pirko 已提交
5977 5978 5979 5980 5981 5982 5983
#ifdef CONFIG_NET_RX_BUSY_POLL
	if (dev->netdev_ops->ndo_busy_poll)
		features |= NETIF_F_BUSY_POLL;
	else
#endif
		features &= ~NETIF_F_BUSY_POLL;

5984 5985 5986
	return features;
}

5987
int __netdev_update_features(struct net_device *dev)
5988
{
5989
	netdev_features_t features;
5990 5991
	int err = 0;

5992 5993
	ASSERT_RTNL();

5994 5995 5996 5997 5998 5999 6000 6001 6002
	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)
6003
		return 0;
6004

6005 6006
	netdev_dbg(dev, "Features changed: %pNF -> %pNF\n",
		&dev->features, &features);
6007 6008 6009 6010

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

6011
	if (unlikely(err < 0)) {
6012
		netdev_err(dev,
6013 6014
			"set_features() failed (%d); wanted %pNF, left %pNF\n",
			err, &features, &dev->features);
6015 6016 6017 6018 6019 6020 6021 6022 6023
		return -1;
	}

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

	return 1;
}

6024 6025 6026 6027 6028 6029 6030 6031
/**
 *	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.
 */
6032 6033 6034 6035
void netdev_update_features(struct net_device *dev)
{
	if (__netdev_update_features(dev))
		netdev_features_change(dev);
6036 6037 6038
}
EXPORT_SYMBOL(netdev_update_features);

6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055
/**
 *	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);

6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082
/**
 *	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);

6083
#ifdef CONFIG_SYSFS
6084 6085 6086
static int netif_alloc_rx_queues(struct net_device *dev)
{
	unsigned int i, count = dev->num_rx_queues;
T
Tom Herbert 已提交
6087
	struct netdev_rx_queue *rx;
6088

T
Tom Herbert 已提交
6089
	BUG_ON(count < 1);
6090

T
Tom Herbert 已提交
6091
	rx = kcalloc(count, sizeof(struct netdev_rx_queue), GFP_KERNEL);
6092
	if (!rx)
T
Tom Herbert 已提交
6093
		return -ENOMEM;
6094

T
Tom Herbert 已提交
6095 6096 6097
	dev->_rx = rx;

	for (i = 0; i < count; i++)
T
Tom Herbert 已提交
6098
		rx[i].dev = dev;
6099 6100
	return 0;
}
T
Tom Herbert 已提交
6101
#endif
6102

C
Changli Gao 已提交
6103 6104 6105 6106 6107 6108 6109
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;
6110
	netdev_queue_numa_node_write(queue, NUMA_NO_NODE);
C
Changli Gao 已提交
6111
	queue->dev = dev;
T
Tom Herbert 已提交
6112 6113 6114
#ifdef CONFIG_BQL
	dql_init(&queue->dql, HZ);
#endif
C
Changli Gao 已提交
6115 6116
}

6117 6118
static void netif_free_tx_queues(struct net_device *dev)
{
W
WANG Cong 已提交
6119
	kvfree(dev->_tx);
6120 6121
}

6122 6123 6124 6125
static int netif_alloc_netdev_queues(struct net_device *dev)
{
	unsigned int count = dev->num_tx_queues;
	struct netdev_queue *tx;
6126
	size_t sz = count * sizeof(*tx);
6127

6128
	BUG_ON(count < 1 || count > 0xffff);
6129

6130 6131 6132 6133 6134 6135
	tx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!tx) {
		tx = vzalloc(sz);
		if (!tx)
			return -ENOMEM;
	}
6136
	dev->_tx = tx;
T
Tom Herbert 已提交
6137

6138 6139
	netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
	spin_lock_init(&dev->tx_global_lock);
C
Changli Gao 已提交
6140 6141

	return 0;
6142 6143
}

L
Linus Torvalds 已提交
6144 6145 6146 6147 6148 6149 6150 6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163
/**
 *	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;
6164
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
6165 6166 6167 6168

	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

6169 6170
	might_sleep();

L
Linus Torvalds 已提交
6171 6172
	/* When net_device's are persistent, this will be fatal. */
	BUG_ON(dev->reg_state != NETREG_UNINITIALIZED);
6173
	BUG_ON(!net);
L
Linus Torvalds 已提交
6174

6175
	spin_lock_init(&dev->addr_list_lock);
6176
	netdev_set_addr_lockdep_class(dev);
L
Linus Torvalds 已提交
6177 6178 6179

	dev->iflink = -1;

6180
	ret = dev_get_valid_name(net, dev, dev->name);
6181 6182 6183
	if (ret < 0)
		goto out;

L
Linus Torvalds 已提交
6184
	/* Init, if this function is available */
6185 6186
	if (dev->netdev_ops->ndo_init) {
		ret = dev->netdev_ops->ndo_init(dev);
L
Linus Torvalds 已提交
6187 6188 6189
		if (ret) {
			if (ret > 0)
				ret = -EIO;
6190
			goto out;
L
Linus Torvalds 已提交
6191 6192
		}
	}
6193

6194 6195
	if (((dev->hw_features | dev->features) &
	     NETIF_F_HW_VLAN_CTAG_FILTER) &&
6196 6197 6198 6199 6200 6201 6202
	    (!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;
	}

6203 6204 6205 6206 6207 6208
	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
6209 6210 6211
	if (dev->iflink == -1)
		dev->iflink = dev->ifindex;

6212 6213 6214 6215
	/* Transfer changeable features to wanted_features and enable
	 * software offloads (GSO and GRO).
	 */
	dev->hw_features |= NETIF_F_SOFT_FEATURES;
6216 6217
	dev->features |= NETIF_F_SOFT_FEATURES;
	dev->wanted_features = dev->features & dev->hw_features;
L
Linus Torvalds 已提交
6218

6219 6220
	if (!(dev->flags & IFF_LOOPBACK)) {
		dev->hw_features |= NETIF_F_NOCACHE_COPY;
6221 6222
	}

6223
	/* Make NETIF_F_HIGHDMA inheritable to VLAN devices.
6224
	 */
6225
	dev->vlan_features |= NETIF_F_HIGHDMA;
6226

6227 6228 6229 6230
	/* Make NETIF_F_SG inheritable to tunnel devices.
	 */
	dev->hw_enc_features |= NETIF_F_SG;

S
Simon Horman 已提交
6231 6232 6233 6234
	/* Make NETIF_F_SG inheritable to MPLS.
	 */
	dev->mpls_features |= NETIF_F_SG;

6235 6236 6237 6238 6239
	ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		goto err_uninit;

6240
	ret = netdev_register_kobject(dev);
6241
	if (ret)
6242
		goto err_uninit;
6243 6244
	dev->reg_state = NETREG_REGISTERED;

6245
	__netdev_update_features(dev);
6246

L
Linus Torvalds 已提交
6247 6248 6249 6250 6251 6252 6253
	/*
	 *	Default initial state at registry is that the
	 *	device is present.
	 */

	set_bit(__LINK_STATE_PRESENT, &dev->state);

6254 6255
	linkwatch_init_dev(dev);

L
Linus Torvalds 已提交
6256 6257
	dev_init_scheduler(dev);
	dev_hold(dev);
6258
	list_netdevice(dev);
6259
	add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
6260

6261 6262 6263 6264 6265 6266 6267
	/* 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
6268
	/* Notify protocols, that a new device appeared. */
6269
	ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
6270
	ret = notifier_to_errno(ret);
6271 6272 6273 6274
	if (ret) {
		rollback_registered(dev);
		dev->reg_state = NETREG_UNREGISTERED;
	}
6275 6276 6277 6278
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
6279 6280
	if (!dev->rtnl_link_ops ||
	    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
6281
		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
L
Linus Torvalds 已提交
6282 6283 6284

out:
	return ret;
6285 6286

err_uninit:
6287 6288
	if (dev->netdev_ops->ndo_uninit)
		dev->netdev_ops->ndo_uninit(dev);
6289
	goto out;
L
Linus Torvalds 已提交
6290
}
E
Eric Dumazet 已提交
6291
EXPORT_SYMBOL(register_netdevice);
L
Linus Torvalds 已提交
6292

6293 6294 6295 6296 6297 6298 6299 6300 6301 6302 6303 6304 6305 6306 6307 6308 6309 6310 6311 6312 6313 6314 6315 6316 6317 6318 6319 6320 6321 6322 6323
/**
 *	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
6324 6325 6326 6327 6328
	/* Note : We dont allocate pcpu_refcnt for dummy devices,
	 * because users of this 'device' dont need to change
	 * its refcount.
	 */

6329 6330 6331 6332 6333
	return 0;
}
EXPORT_SYMBOL_GPL(init_dummy_netdev);


L
Linus Torvalds 已提交
6334 6335 6336 6337 6338 6339 6340 6341 6342
/**
 *	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.
 *
6343
 *	This is a wrapper around register_netdevice that takes the rtnl semaphore
L
Linus Torvalds 已提交
6344 6345 6346 6347 6348 6349 6350 6351 6352 6353 6354 6355 6356 6357
 *	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
6358 6359 6360 6361 6362 6363 6364 6365 6366 6367
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);

6368
/**
L
Linus Torvalds 已提交
6369
 * netdev_wait_allrefs - wait until all references are gone.
6370
 * @dev: target net_device
L
Linus Torvalds 已提交
6371 6372 6373 6374 6375 6376 6377
 *
 * 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
6378
 * call dev_put.
L
Linus Torvalds 已提交
6379 6380 6381 6382
 */
static void netdev_wait_allrefs(struct net_device *dev)
{
	unsigned long rebroadcast_time, warning_time;
E
Eric Dumazet 已提交
6383
	int refcnt;
L
Linus Torvalds 已提交
6384

6385 6386
	linkwatch_forget_dev(dev);

L
Linus Torvalds 已提交
6387
	rebroadcast_time = warning_time = jiffies;
E
Eric Dumazet 已提交
6388 6389 6390
	refcnt = netdev_refcnt_read(dev);

	while (refcnt != 0) {
L
Linus Torvalds 已提交
6391
		if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
6392
			rtnl_lock();
L
Linus Torvalds 已提交
6393 6394

			/* Rebroadcast unregister notification */
6395
			call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
L
Linus Torvalds 已提交
6396

6397
			__rtnl_unlock();
6398
			rcu_barrier();
6399 6400
			rtnl_lock();

6401
			call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
L
Linus Torvalds 已提交
6402 6403 6404 6405 6406 6407 6408 6409 6410 6411 6412
			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();
			}

6413
			__rtnl_unlock();
L
Linus Torvalds 已提交
6414 6415 6416 6417 6418 6419

			rebroadcast_time = jiffies;
		}

		msleep(250);

E
Eric Dumazet 已提交
6420 6421
		refcnt = netdev_refcnt_read(dev);

L
Linus Torvalds 已提交
6422
		if (time_after(jiffies, warning_time + 10 * HZ)) {
6423 6424
			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
				 dev->name, refcnt);
L
Linus Torvalds 已提交
6425 6426 6427 6428 6429 6430 6431 6432 6433 6434 6435 6436 6437 6438 6439 6440 6441 6442 6443
			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
6444
 * We are invoked by rtnl_unlock().
L
Linus Torvalds 已提交
6445
 * This allows us to deal with problems:
6446
 * 1) We can delete sysfs objects which invoke hotplug
L
Linus Torvalds 已提交
6447 6448 6449
 *    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
6450 6451 6452
 *
 * 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
6453 6454 6455
 */
void netdev_run_todo(void)
{
6456
	struct list_head list;
L
Linus Torvalds 已提交
6457 6458

	/* Snapshot list, allow later requests */
6459
	list_replace_init(&net_todo_list, &list);
H
Herbert Xu 已提交
6460 6461

	__rtnl_unlock();
6462

6463 6464

	/* Wait for rcu callbacks to finish before next phase */
6465 6466 6467
	if (!list_empty(&list))
		rcu_barrier();

L
Linus Torvalds 已提交
6468 6469
	while (!list_empty(&list)) {
		struct net_device *dev
6470
			= list_first_entry(&list, struct net_device, todo_list);
L
Linus Torvalds 已提交
6471 6472
		list_del(&dev->todo_list);

6473
		rtnl_lock();
6474
		call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6475
		__rtnl_unlock();
6476

6477
		if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
6478
			pr_err("network todo '%s' but state %d\n",
6479 6480 6481 6482
			       dev->name, dev->reg_state);
			dump_stack();
			continue;
		}
L
Linus Torvalds 已提交
6483

6484
		dev->reg_state = NETREG_UNREGISTERED;
L
Linus Torvalds 已提交
6485

6486
		on_each_cpu(flush_backlog, dev, 1);
6487

6488
		netdev_wait_allrefs(dev);
L
Linus Torvalds 已提交
6489

6490
		/* paranoia */
E
Eric Dumazet 已提交
6491
		BUG_ON(netdev_refcnt_read(dev));
6492 6493
		WARN_ON(rcu_access_pointer(dev->ip_ptr));
		WARN_ON(rcu_access_pointer(dev->ip6_ptr));
6494
		WARN_ON(dev->dn_ptr);
L
Linus Torvalds 已提交
6495

6496 6497
		if (dev->destructor)
			dev->destructor(dev);
6498

6499 6500 6501 6502 6503 6504
		/* Report a network device has been unregistered */
		rtnl_lock();
		dev_net(dev)->dev_unreg_count--;
		__rtnl_unlock();
		wake_up(&netdev_unregistering_wq);

6505 6506
		/* Free network device */
		kobject_put(&dev->dev.kobj);
L
Linus Torvalds 已提交
6507 6508 6509
	}
}

6510 6511 6512
/* Convert net_device_stats to rtnl_link_stats64.  They have the same
 * fields in the same order, with only the type differing.
 */
6513 6514
void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
			     const struct net_device_stats *netdev_stats)
6515 6516
{
#if BITS_PER_LONG == 64
6517 6518
	BUILD_BUG_ON(sizeof(*stats64) != sizeof(*netdev_stats));
	memcpy(stats64, netdev_stats, sizeof(*stats64));
6519 6520 6521 6522 6523 6524 6525 6526 6527 6528 6529
#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
}
6530
EXPORT_SYMBOL(netdev_stats_to_stats64);
6531

6532 6533 6534
/**
 *	dev_get_stats	- get network device statistics
 *	@dev: device to get statistics from
6535
 *	@storage: place to store stats
6536
 *
6537 6538 6539 6540
 *	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.
6541
 */
6542 6543
struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
					struct rtnl_link_stats64 *storage)
6544
{
6545 6546
	const struct net_device_ops *ops = dev->netdev_ops;

6547 6548
	if (ops->ndo_get_stats64) {
		memset(storage, 0, sizeof(*storage));
6549 6550
		ops->ndo_get_stats64(dev, storage);
	} else if (ops->ndo_get_stats) {
6551
		netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev));
6552 6553
	} else {
		netdev_stats_to_stats64(storage, &dev->stats);
6554
	}
6555
	storage->rx_dropped += atomic_long_read(&dev->rx_dropped);
6556
	storage->tx_dropped += atomic_long_read(&dev->tx_dropped);
6557
	return storage;
R
Rusty Russell 已提交
6558
}
6559
EXPORT_SYMBOL(dev_get_stats);
R
Rusty Russell 已提交
6560

6561
struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
6562
{
6563
	struct netdev_queue *queue = dev_ingress_queue(dev);
6564

6565 6566 6567 6568 6569 6570 6571 6572 6573 6574 6575 6576
#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;
6577 6578
}

6579 6580
static const struct ethtool_ops default_ethtool_ops;

6581 6582 6583 6584 6585 6586 6587 6588
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);

6589 6590 6591 6592
void netdev_freemem(struct net_device *dev)
{
	char *addr = (char *)dev - dev->padded;

W
WANG Cong 已提交
6593
	kvfree(addr);
6594 6595
}

L
Linus Torvalds 已提交
6596
/**
T
Tom Herbert 已提交
6597
 *	alloc_netdev_mqs - allocate network device
6598 6599 6600 6601 6602 6603
 *	@sizeof_priv:		size of private data to allocate space for
 *	@name:			device name format string
 *	@name_assign_type: 	origin of device name
 *	@setup:			callback to initialize device
 *	@txqs:			the number of TX subqueues to allocate
 *	@rxqs:			the number of RX subqueues to allocate
L
Linus Torvalds 已提交
6604 6605
 *
 *	Allocates a struct net_device with private data area for driver use
L
Li Zhong 已提交
6606
 *	and performs basic initialization.  Also allocates subqueue structs
T
Tom Herbert 已提交
6607
 *	for each queue on the device.
L
Linus Torvalds 已提交
6608
 */
T
Tom Herbert 已提交
6609
struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
6610
		unsigned char name_assign_type,
T
Tom Herbert 已提交
6611 6612
		void (*setup)(struct net_device *),
		unsigned int txqs, unsigned int rxqs)
L
Linus Torvalds 已提交
6613 6614
{
	struct net_device *dev;
6615
	size_t alloc_size;
6616
	struct net_device *p;
L
Linus Torvalds 已提交
6617

6618 6619
	BUG_ON(strlen(name) >= sizeof(dev->name));

T
Tom Herbert 已提交
6620
	if (txqs < 1) {
6621
		pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
6622 6623 6624
		return NULL;
	}

6625
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
6626
	if (rxqs < 1) {
6627
		pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
T
Tom Herbert 已提交
6628 6629 6630 6631
		return NULL;
	}
#endif

6632
	alloc_size = sizeof(struct net_device);
6633 6634
	if (sizeof_priv) {
		/* ensure 32-byte alignment of private area */
6635
		alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
6636 6637 6638
		alloc_size += sizeof_priv;
	}
	/* ensure 32-byte alignment of whole construct */
6639
	alloc_size += NETDEV_ALIGN - 1;
L
Linus Torvalds 已提交
6640

6641 6642 6643
	p = kzalloc(alloc_size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!p)
		p = vzalloc(alloc_size);
6644
	if (!p)
L
Linus Torvalds 已提交
6645 6646
		return NULL;

6647
	dev = PTR_ALIGN(p, NETDEV_ALIGN);
L
Linus Torvalds 已提交
6648
	dev->padded = (char *)dev - (char *)p;
6649

E
Eric Dumazet 已提交
6650 6651
	dev->pcpu_refcnt = alloc_percpu(int);
	if (!dev->pcpu_refcnt)
6652
		goto free_dev;
6653 6654

	if (dev_addr_init(dev))
E
Eric Dumazet 已提交
6655
		goto free_pcpu;
6656

6657
	dev_mc_init(dev);
6658
	dev_uc_init(dev);
J
Jiri Pirko 已提交
6659

6660
	dev_net_set(dev, &init_net);
L
Linus Torvalds 已提交
6661

6662
	dev->gso_max_size = GSO_MAX_SIZE;
6663
	dev->gso_max_segs = GSO_MAX_SEGS;
6664 6665 6666

	INIT_LIST_HEAD(&dev->napi_list);
	INIT_LIST_HEAD(&dev->unreg_list);
6667
	INIT_LIST_HEAD(&dev->close_list);
6668
	INIT_LIST_HEAD(&dev->link_watch_list);
6669 6670 6671 6672
	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);
6673 6674 6675
	dev->priv_flags = IFF_XMIT_DST_RELEASE;
	setup(dev);

T
Tom Herbert 已提交
6676 6677
	dev->num_tx_queues = txqs;
	dev->real_num_tx_queues = txqs;
6678
	if (netif_alloc_netdev_queues(dev))
6679
		goto free_all;
6680

6681
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
6682 6683
	dev->num_rx_queues = rxqs;
	dev->real_num_rx_queues = rxqs;
T
Tom Herbert 已提交
6684
	if (netif_alloc_rx_queues(dev))
6685
		goto free_all;
E
Eric Dumazet 已提交
6686
#endif
T
Tom Herbert 已提交
6687

L
Linus Torvalds 已提交
6688
	strcpy(dev->name, name);
6689
	dev->name_assign_type = name_assign_type;
6690
	dev->group = INIT_NETDEV_GROUP;
6691 6692
	if (!dev->ethtool_ops)
		dev->ethtool_ops = &default_ethtool_ops;
L
Linus Torvalds 已提交
6693
	return dev;
6694

6695 6696 6697 6698
free_all:
	free_netdev(dev);
	return NULL;

E
Eric Dumazet 已提交
6699 6700
free_pcpu:
	free_percpu(dev->pcpu_refcnt);
6701 6702
free_dev:
	netdev_freemem(dev);
6703
	return NULL;
L
Linus Torvalds 已提交
6704
}
T
Tom Herbert 已提交
6705
EXPORT_SYMBOL(alloc_netdev_mqs);
L
Linus Torvalds 已提交
6706 6707 6708 6709 6710

/**
 *	free_netdev - free network device
 *	@dev: device
 *
6711 6712
 *	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
6713 6714 6715 6716
 *	If this is the last reference then it will be freed.
 */
void free_netdev(struct net_device *dev)
{
6717 6718
	struct napi_struct *p, *n;

6719 6720
	release_net(dev_net(dev));

6721
	netif_free_tx_queues(dev);
6722
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
6723 6724
	kfree(dev->_rx);
#endif
6725

6726
	kfree(rcu_dereference_protected(dev->ingress_queue, 1));
6727

6728 6729 6730
	/* Flush device addresses */
	dev_addr_flush(dev);

6731 6732 6733
	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
		netif_napi_del(p);

E
Eric Dumazet 已提交
6734 6735 6736
	free_percpu(dev->pcpu_refcnt);
	dev->pcpu_refcnt = NULL;

S
Stephen Hemminger 已提交
6737
	/*  Compatibility with error handling in drivers */
L
Linus Torvalds 已提交
6738
	if (dev->reg_state == NETREG_UNINITIALIZED) {
6739
		netdev_freemem(dev);
L
Linus Torvalds 已提交
6740 6741 6742 6743 6744 6745
		return;
	}

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

6746 6747
	/* will free via device release */
	put_device(&dev->dev);
L
Linus Torvalds 已提交
6748
}
E
Eric Dumazet 已提交
6749
EXPORT_SYMBOL(free_netdev);
6750

6751 6752 6753 6754 6755 6756
/**
 *	synchronize_net -  Synchronize with packet receive processing
 *
 *	Wait for packets currently being received to be done.
 *	Does not block later packets from starting.
 */
6757
void synchronize_net(void)
L
Linus Torvalds 已提交
6758 6759
{
	might_sleep();
6760 6761 6762 6763
	if (rtnl_is_locked())
		synchronize_rcu_expedited();
	else
		synchronize_rcu();
L
Linus Torvalds 已提交
6764
}
E
Eric Dumazet 已提交
6765
EXPORT_SYMBOL(synchronize_net);
L
Linus Torvalds 已提交
6766 6767

/**
6768
 *	unregister_netdevice_queue - remove device from the kernel
L
Linus Torvalds 已提交
6769
 *	@dev: device
6770
 *	@head: list
6771
 *
L
Linus Torvalds 已提交
6772
 *	This function shuts down a device interface and removes it
6773
 *	from the kernel tables.
6774
 *	If head not NULL, device is queued to be unregistered later.
L
Linus Torvalds 已提交
6775 6776 6777 6778 6779
 *
 *	Callers must hold the rtnl semaphore.  You may want
 *	unregister_netdev() instead of this.
 */

6780
void unregister_netdevice_queue(struct net_device *dev, struct list_head *head)
L
Linus Torvalds 已提交
6781
{
6782 6783
	ASSERT_RTNL();

6784
	if (head) {
6785
		list_move_tail(&dev->unreg_list, head);
6786 6787 6788 6789 6790
	} else {
		rollback_registered(dev);
		/* Finish processing unregister after unlock */
		net_set_todo(dev);
	}
L
Linus Torvalds 已提交
6791
}
6792
EXPORT_SYMBOL(unregister_netdevice_queue);
L
Linus Torvalds 已提交
6793

6794 6795 6796
/**
 *	unregister_netdevice_many - unregister many devices
 *	@head: list of devices
6797 6798 6799
 *
 *  Note: As most callers use a stack allocated list_head,
 *  we force a list_del() to make sure stack wont be corrupted later.
6800 6801 6802 6803 6804 6805 6806 6807 6808
 */
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);
6809
		list_del(head);
6810 6811
	}
}
6812
EXPORT_SYMBOL(unregister_netdevice_many);
6813

L
Linus Torvalds 已提交
6814 6815 6816 6817 6818
/**
 *	unregister_netdev - remove device from the kernel
 *	@dev: device
 *
 *	This function shuts down a device interface and removes it
6819
 *	from the kernel tables.
L
Linus Torvalds 已提交
6820 6821 6822 6823 6824 6825 6826 6827 6828 6829 6830 6831 6832
 *
 *	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);

6833 6834 6835 6836 6837 6838 6839 6840 6841 6842 6843 6844 6845 6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857 6858 6859 6860 6861 6862 6863
/**
 *	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;
6864
	if (net_eq(dev_net(dev), net))
6865 6866 6867 6868 6869 6870
		goto out;

	/* Pick the destination device name, and ensure
	 * we can use it in the destination network namespace.
	 */
	err = -EEXIST;
6871
	if (__dev_get_by_name(net, dev->name)) {
6872 6873 6874
		/* We get here if we can't use the current device name */
		if (!pat)
			goto out;
6875
		if (dev_get_valid_name(net, dev, pat) < 0)
6876 6877 6878 6879 6880 6881 6882 6883
			goto out;
	}

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

	/* If device is running close it first. */
6884
	dev_close(dev);
6885 6886 6887 6888 6889 6890 6891 6892 6893 6894 6895 6896

	/* 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.
6897 6898 6899 6900

	   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.
6901 6902
	*/
	call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
6903 6904
	rcu_barrier();
	call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6905
	rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);
6906 6907 6908 6909

	/*
	 *	Flush the unicast and multicast chains
	 */
6910
	dev_uc_flush(dev);
6911
	dev_mc_flush(dev);
6912

6913 6914
	/* Send a netdev-removed uevent to the old namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_REMOVE);
6915
	netdev_adjacent_del_links(dev);
6916

6917
	/* Actually switch the network namespace */
6918
	dev_net_set(dev, net);
6919 6920 6921 6922 6923 6924 6925 6926 6927

	/* 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;
	}

6928 6929
	/* Send a netdev-add uevent to the new namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_ADD);
6930
	netdev_adjacent_add_links(dev);
6931

6932
	/* Fixup kobjects */
6933
	err = device_rename(&dev->dev, dev->name);
6934
	WARN_ON(err);
6935 6936 6937 6938 6939 6940 6941

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

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

6942 6943 6944 6945
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
6946
	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
6947

6948 6949 6950 6951 6952
	synchronize_net();
	err = 0;
out:
	return err;
}
6953
EXPORT_SYMBOL_GPL(dev_change_net_namespace);
6954

L
Linus Torvalds 已提交
6955 6956 6957 6958 6959 6960 6961 6962 6963
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;

6964
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
L
Linus Torvalds 已提交
6965 6966 6967 6968 6969 6970 6971 6972 6973 6974 6975 6976 6977 6978 6979 6980
		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. */
6981 6982 6983 6984 6985 6986
	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;
	}
6987 6988 6989 6990 6991
	/* 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
6992 6993 6994 6995 6996

	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_enable();

	/* Process offline CPU's input_pkt_queue */
6997
	while ((skb = __skb_dequeue(&oldsd->process_queue))) {
6998
		netif_rx_internal(skb);
6999
		input_queue_head_incr(oldsd);
T
Tom Herbert 已提交
7000
	}
7001
	while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) {
7002
		netif_rx_internal(skb);
7003 7004
		input_queue_head_incr(oldsd);
	}
L
Linus Torvalds 已提交
7005 7006 7007 7008 7009

	return NOTIFY_OK;
}


7010
/**
7011 7012 7013 7014
 *	netdev_increment_features - increment feature set by one
 *	@all: current feature set
 *	@one: new feature set
 *	@mask: mask feature set
7015 7016
 *
 *	Computes a new feature set after adding a device with feature set
7017 7018
 *	@one to the master device with current feature set @all.  Will not
 *	enable anything that is off in @mask. Returns the new feature set.
7019
 */
7020 7021
netdev_features_t netdev_increment_features(netdev_features_t all,
	netdev_features_t one, netdev_features_t mask)
7022
{
7023 7024 7025
	if (mask & NETIF_F_GEN_CSUM)
		mask |= NETIF_F_ALL_CSUM;
	mask |= NETIF_F_VLAN_CHALLENGED;
7026

7027 7028
	all |= one & (NETIF_F_ONE_FOR_ALL|NETIF_F_ALL_CSUM) & mask;
	all &= one | ~NETIF_F_ALL_FOR_ALL;
7029

7030 7031 7032
	/* If one device supports hw checksumming, set for all. */
	if (all & NETIF_F_GEN_CSUM)
		all &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
7033 7034 7035

	return all;
}
7036
EXPORT_SYMBOL(netdev_increment_features);
7037

7038
static struct hlist_head * __net_init netdev_create_hash(void)
7039 7040 7041 7042 7043 7044 7045 7046 7047 7048 7049 7050
{
	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;
}

7051
/* Initialize per network namespace state */
7052
static int __net_init netdev_init(struct net *net)
7053
{
7054 7055
	if (net != &init_net)
		INIT_LIST_HEAD(&net->dev_base_head);
7056

7057 7058 7059
	net->dev_name_head = netdev_create_hash();
	if (net->dev_name_head == NULL)
		goto err_name;
7060

7061 7062 7063
	net->dev_index_head = netdev_create_hash();
	if (net->dev_index_head == NULL)
		goto err_idx;
7064 7065

	return 0;
7066 7067 7068 7069 7070

err_idx:
	kfree(net->dev_name_head);
err_name:
	return -ENOMEM;
7071 7072
}

7073 7074 7075 7076 7077 7078
/**
 *	netdev_drivername - network driver for the device
 *	@dev: network device
 *
 *	Determine network driver for device.
 */
7079
const char *netdev_drivername(const struct net_device *dev)
7080
{
7081 7082
	const struct device_driver *driver;
	const struct device *parent;
7083
	const char *empty = "";
7084 7085 7086

	parent = dev->dev.parent;
	if (!parent)
7087
		return empty;
7088 7089 7090

	driver = parent->driver;
	if (driver && driver->name)
7091 7092
		return driver->name;
	return empty;
7093 7094
}

7095 7096
static void __netdev_printk(const char *level, const struct net_device *dev,
			    struct va_format *vaf)
7097
{
7098
	if (dev && dev->dev.parent) {
7099 7100 7101 7102 7103 7104 7105
		dev_printk_emit(level[1] - '0',
				dev->dev.parent,
				"%s %s %s%s: %pV",
				dev_driver_string(dev->dev.parent),
				dev_name(dev->dev.parent),
				netdev_name(dev), netdev_reg_state(dev),
				vaf);
7106
	} else if (dev) {
7107 7108
		printk("%s%s%s: %pV",
		       level, netdev_name(dev), netdev_reg_state(dev), vaf);
7109
	} else {
7110
		printk("%s(NULL net_device): %pV", level, vaf);
7111
	}
7112 7113
}

7114 7115
void netdev_printk(const char *level, const struct net_device *dev,
		   const char *format, ...)
7116 7117 7118 7119 7120 7121 7122 7123 7124
{
	struct va_format vaf;
	va_list args;

	va_start(args, format);

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

7125
	__netdev_printk(level, dev, &vaf);
7126

7127 7128 7129 7130 7131
	va_end(args);
}
EXPORT_SYMBOL(netdev_printk);

#define define_netdev_printk_level(func, level)			\
7132
void func(const struct net_device *dev, const char *fmt, ...)	\
7133 7134 7135 7136 7137 7138 7139 7140 7141
{								\
	struct va_format vaf;					\
	va_list args;						\
								\
	va_start(args, fmt);					\
								\
	vaf.fmt = fmt;						\
	vaf.va = &args;						\
								\
7142
	__netdev_printk(level, dev, &vaf);			\
7143
								\
7144 7145 7146 7147 7148 7149 7150 7151 7152 7153 7154 7155
	va_end(args);						\
}								\
EXPORT_SYMBOL(func);

define_netdev_printk_level(netdev_emerg, KERN_EMERG);
define_netdev_printk_level(netdev_alert, KERN_ALERT);
define_netdev_printk_level(netdev_crit, KERN_CRIT);
define_netdev_printk_level(netdev_err, KERN_ERR);
define_netdev_printk_level(netdev_warn, KERN_WARNING);
define_netdev_printk_level(netdev_notice, KERN_NOTICE);
define_netdev_printk_level(netdev_info, KERN_INFO);

7156
static void __net_exit netdev_exit(struct net *net)
7157 7158 7159 7160 7161
{
	kfree(net->dev_name_head);
	kfree(net->dev_index_head);
}

7162
static struct pernet_operations __net_initdata netdev_net_ops = {
7163 7164 7165 7166
	.init = netdev_init,
	.exit = netdev_exit,
};

7167
static void __net_exit default_device_exit(struct net *net)
7168
{
7169
	struct net_device *dev, *aux;
7170
	/*
7171
	 * Push all migratable network devices back to the
7172 7173 7174
	 * initial network namespace
	 */
	rtnl_lock();
7175
	for_each_netdev_safe(net, dev, aux) {
7176
		int err;
7177
		char fb_name[IFNAMSIZ];
7178 7179 7180 7181 7182

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

7183 7184 7185
		/* Leave virtual devices for the generic cleanup */
		if (dev->rtnl_link_ops)
			continue;
7186

L
Lucas De Marchi 已提交
7187
		/* Push remaining network devices to init_net */
7188 7189
		snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex);
		err = dev_change_net_namespace(dev, &init_net, fb_name);
7190
		if (err) {
7191 7192
			pr_emerg("%s: failed to move %s to init_net: %d\n",
				 __func__, dev->name, err);
7193
			BUG();
7194 7195 7196 7197 7198
		}
	}
	rtnl_unlock();
}

7199 7200 7201 7202 7203 7204 7205 7206 7207 7208 7209 7210 7211 7212 7213 7214 7215 7216 7217 7218 7219 7220 7221 7222 7223 7224 7225 7226
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);
}

7227 7228 7229
static void __net_exit default_device_exit_batch(struct list_head *net_list)
{
	/* At exit all network devices most be removed from a network
7230
	 * namespace.  Do this in the reverse order of registration.
7231 7232 7233 7234 7235 7236 7237
	 * 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);

7238 7239 7240 7241 7242 7243 7244 7245 7246 7247 7248 7249
	/* 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);
7250 7251
	list_for_each_entry(net, net_list, exit_list) {
		for_each_netdev_reverse(net, dev) {
7252
			if (dev->rtnl_link_ops && dev->rtnl_link_ops->dellink)
7253 7254 7255 7256 7257 7258 7259 7260 7261
				dev->rtnl_link_ops->dellink(dev, &dev_kill_list);
			else
				unregister_netdevice_queue(dev, &dev_kill_list);
		}
	}
	unregister_netdevice_many(&dev_kill_list);
	rtnl_unlock();
}

7262
static struct pernet_operations __net_initdata default_device_ops = {
7263
	.exit = default_device_exit,
7264
	.exit_batch = default_device_exit_batch,
7265 7266
};

L
Linus Torvalds 已提交
7267 7268 7269 7270 7271 7272 7273 7274 7275 7276 7277 7278 7279 7280 7281 7282 7283 7284 7285 7286
/*
 *	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;

7287
	if (netdev_kobject_init())
L
Linus Torvalds 已提交
7288 7289 7290
		goto out;

	INIT_LIST_HEAD(&ptype_all);
7291
	for (i = 0; i < PTYPE_HASH_SIZE; i++)
L
Linus Torvalds 已提交
7292 7293
		INIT_LIST_HEAD(&ptype_base[i]);

7294 7295
	INIT_LIST_HEAD(&offload_base);

7296 7297
	if (register_pernet_subsys(&netdev_net_ops))
		goto out;
L
Linus Torvalds 已提交
7298 7299 7300 7301 7302

	/*
	 *	Initialise the packet receive queues.
	 */

7303
	for_each_possible_cpu(i) {
E
Eric Dumazet 已提交
7304
		struct softnet_data *sd = &per_cpu(softnet_data, i);
L
Linus Torvalds 已提交
7305

E
Eric Dumazet 已提交
7306
		skb_queue_head_init(&sd->input_pkt_queue);
7307
		skb_queue_head_init(&sd->process_queue);
E
Eric Dumazet 已提交
7308
		INIT_LIST_HEAD(&sd->poll_list);
7309
		sd->output_queue_tailp = &sd->output_queue;
E
Eric Dumazet 已提交
7310
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
7311 7312 7313
		sd->csd.func = rps_trigger_softirq;
		sd->csd.info = sd;
		sd->cpu = i;
7314
#endif
T
Tom Herbert 已提交
7315

E
Eric Dumazet 已提交
7316 7317
		sd->backlog.poll = process_backlog;
		sd->backlog.weight = weight_p;
L
Linus Torvalds 已提交
7318 7319 7320 7321
	}

	dev_boot_phase = 0;

7322 7323 7324 7325 7326 7327 7328 7329 7330 7331 7332 7333 7334 7335 7336
	/* 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;

7337 7338
	open_softirq(NET_TX_SOFTIRQ, net_tx_action);
	open_softirq(NET_RX_SOFTIRQ, net_rx_action);
L
Linus Torvalds 已提交
7339 7340 7341 7342 7343 7344 7345 7346 7347

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

subsys_initcall(net_dev_init);