dev.c 175.8 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9
/*
 * 	NET3	Protocol independent device support routines.
 *
 *		This program is free software; you can redistribute it and/or
 *		modify it under the terms of the GNU General Public License
 *		as published by the Free Software Foundation; either version
 *		2 of the License, or (at your option) any later version.
 *
 *	Derived from the non IP parts of dev.c 1.0.19
10
 * 		Authors:	Ross Biro
L
Linus Torvalds 已提交
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76
 *				Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 *				Mark Evans, <evansmp@uhura.aston.ac.uk>
 *
 *	Additional Authors:
 *		Florian la Roche <rzsfl@rz.uni-sb.de>
 *		Alan Cox <gw4pts@gw4pts.ampr.org>
 *		David Hinds <dahinds@users.sourceforge.net>
 *		Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
 *		Adam Sulmicki <adam@cfar.umd.edu>
 *              Pekka Riikonen <priikone@poesidon.pspt.fi>
 *
 *	Changes:
 *              D.J. Barrow     :       Fixed bug where dev->refcnt gets set
 *              			to 2 if register_netdev gets called
 *              			before net_dev_init & also removed a
 *              			few lines of code in the process.
 *		Alan Cox	:	device private ioctl copies fields back.
 *		Alan Cox	:	Transmit queue code does relevant
 *					stunts to keep the queue safe.
 *		Alan Cox	:	Fixed double lock.
 *		Alan Cox	:	Fixed promisc NULL pointer trap
 *		????????	:	Support the full private ioctl range
 *		Alan Cox	:	Moved ioctl permission check into
 *					drivers
 *		Tim Kordas	:	SIOCADDMULTI/SIOCDELMULTI
 *		Alan Cox	:	100 backlog just doesn't cut it when
 *					you start doing multicast video 8)
 *		Alan Cox	:	Rewrote net_bh and list manager.
 *		Alan Cox	: 	Fix ETH_P_ALL echoback lengths.
 *		Alan Cox	:	Took out transmit every packet pass
 *					Saved a few bytes in the ioctl handler
 *		Alan Cox	:	Network driver sets packet type before
 *					calling netif_rx. Saves a function
 *					call a packet.
 *		Alan Cox	:	Hashed net_bh()
 *		Richard Kooijman:	Timestamp fixes.
 *		Alan Cox	:	Wrong field in SIOCGIFDSTADDR
 *		Alan Cox	:	Device lock protection.
 *		Alan Cox	: 	Fixed nasty side effect of device close
 *					changes.
 *		Rudi Cilibrasi	:	Pass the right thing to
 *					set_mac_address()
 *		Dave Miller	:	32bit quantity for the device lock to
 *					make it work out on a Sparc.
 *		Bjorn Ekwall	:	Added KERNELD hack.
 *		Alan Cox	:	Cleaned up the backlog initialise.
 *		Craig Metz	:	SIOCGIFCONF fix if space for under
 *					1 device.
 *	    Thomas Bogendoerfer :	Return ENODEV for dev_open, if there
 *					is no device open function.
 *		Andi Kleen	:	Fix error reporting for SIOCGIFCONF
 *	    Michael Chastain	:	Fix signed/unsigned for SIOCGIFCONF
 *		Cyrus Durgin	:	Cleaned for KMOD
 *		Adam Sulmicki   :	Bug Fix : Network Device Unload
 *					A network device unload needs to purge
 *					the backlog queue.
 *	Paul Rusty Russell	:	SIOCSIFNAME
 *              Pekka Riikonen  :	Netdev boot-time settings code
 *              Andrew Morton   :       Make unregister_netdevice wait
 *              			indefinitely on dev->refcnt
 * 		J Hadi Salim	:	- Backlog queue sampling
 *				        - netif_rx() feedback
 */

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

136 137
#include "net-sysfs.h"

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

H
Herbert Xu 已提交
141 142 143
/* This should be increased if a protocol with a bigger head is added. */
#define GRO_MAX_HEAD (MAX_HEADER + 128)

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

150 151
static int netif_rx_internal(struct sk_buff *skb);

L
Linus Torvalds 已提交
152
/*
153
 * The @dev_base_head list is protected by @dev_base_lock and the rtnl
L
Linus Torvalds 已提交
154 155
 * semaphore.
 *
156
 * Pure readers hold dev_base_lock for reading, or rcu_read_lock()
L
Linus Torvalds 已提交
157 158
 *
 * Writers must hold the rtnl semaphore while they loop through the
159
 * dev_base_head list, and hold dev_base_lock for writing when they do the
L
Linus Torvalds 已提交
160 161 162 163 164 165 166 167 168 169 170 171 172 173
 * 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
174 175 176 177 178 179
/* 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);

180
static seqcount_t devnet_rename_seq;
181

182 183 184 185 186
static inline void dev_base_seq_inc(struct net *net)
{
	while (++net->dev_base_seq == 0);
}

187
static inline struct hlist_head *dev_name_hash(struct net *net, const char *name)
L
Linus Torvalds 已提交
188
{
189 190
	unsigned int hash = full_name_hash(name, strnlen(name, IFNAMSIZ));

191
	return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)];
L
Linus Torvalds 已提交
192 193
}

194
static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex)
L
Linus Torvalds 已提交
195
{
196
	return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)];
L
Linus Torvalds 已提交
197 198
}

E
Eric Dumazet 已提交
199
static inline void rps_lock(struct softnet_data *sd)
200 201
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
202
	spin_lock(&sd->input_pkt_queue.lock);
203 204 205
#endif
}

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

213
/* Device list insertion */
214
static void list_netdevice(struct net_device *dev)
215
{
216
	struct net *net = dev_net(dev);
217 218 219 220

	ASSERT_RTNL();

	write_lock_bh(&dev_base_lock);
221
	list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
222
	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
223 224
	hlist_add_head_rcu(&dev->index_hlist,
			   dev_index_hash(net, dev->ifindex));
225
	write_unlock_bh(&dev_base_lock);
226 227

	dev_base_seq_inc(net);
228 229
}

230 231 232
/* Device list removal
 * caller must respect a RCU grace period before freeing/reusing dev
 */
233 234 235 236 237 238
static void unlist_netdevice(struct net_device *dev)
{
	ASSERT_RTNL();

	/* Unlink dev from the device chain */
	write_lock_bh(&dev_base_lock);
239
	list_del_rcu(&dev->dev_list);
240
	hlist_del_rcu(&dev->name_hlist);
241
	hlist_del_rcu(&dev->index_hlist);
242
	write_unlock_bh(&dev_base_lock);
243 244

	dev_base_seq_inc(dev_net(dev));
245 246
}

L
Linus Torvalds 已提交
247 248 249 250
/*
 *	Our notifier list
 */

251
static RAW_NOTIFIER_HEAD(netdev_chain);
L
Linus Torvalds 已提交
252 253 254 255 256

/*
 *	Device drivers call our routines to queue packets here. We empty the
 *	queue in the local softnet handler.
 */
257

258
DEFINE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
E
Eric Dumazet 已提交
259
EXPORT_PER_CPU_SYMBOL(softnet_data);
L
Linus Torvalds 已提交
260

261
#ifdef CONFIG_LOCKDEP
262
/*
263
 * register_netdevice() inits txq->_xmit_lock and sets lockdep class
264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
 * 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,
279 280 281
	 ARPHRD_FCFABRIC, ARPHRD_IEEE80211, ARPHRD_IEEE80211_PRISM,
	 ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, ARPHRD_PHONET_PIPE,
	 ARPHRD_IEEE802154, ARPHRD_VOID, ARPHRD_NONE};
282

283
static const char *const netdev_lock_name[] =
284 285 286 287 288 289 290 291 292 293 294 295
	{"_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",
296 297 298
	 "_xmit_FCFABRIC", "_xmit_IEEE80211", "_xmit_IEEE80211_PRISM",
	 "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", "_xmit_PHONET_PIPE",
	 "_xmit_IEEE802154", "_xmit_VOID", "_xmit_NONE"};
299 300

static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)];
301
static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)];
302 303 304 305 306 307 308 309 310 311 312 313

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;
}

314 315
static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
						 unsigned short dev_type)
316 317 318 319 320 321 322
{
	int i;

	i = netdev_lock_pos(dev_type);
	lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i],
				   netdev_lock_name[i]);
}
323 324 325 326 327 328 329 330 331 332

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]);
}
333
#else
334 335 336 337 338
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)
339 340 341
{
}
#endif
L
Linus Torvalds 已提交
342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364

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

		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)
 */

365 366 367 368 369 370 371 372
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
373 374 375 376 377 378 379 380
/**
 *	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.
 *
381
 *	This call does not sleep therefore it can not
L
Linus Torvalds 已提交
382 383 384 385 386 387
 *	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)
{
388
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
389

390 391 392
	spin_lock(&ptype_lock);
	list_add_rcu(&pt->list, head);
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
393
}
E
Eric Dumazet 已提交
394
EXPORT_SYMBOL(dev_add_pack);
L
Linus Torvalds 已提交
395 396 397 398 399 400 401 402

/**
 *	__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
403
 *	returns.
L
Linus Torvalds 已提交
404 405 406 407 408 409 410
 *
 *      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)
{
411
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
412 413
	struct packet_type *pt1;

414
	spin_lock(&ptype_lock);
L
Linus Torvalds 已提交
415 416 417 418 419 420 421 422

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

423
	pr_warn("dev_remove_pack: %p not found\n", pt);
L
Linus Torvalds 已提交
424
out:
425
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
426
}
E
Eric Dumazet 已提交
427 428
EXPORT_SYMBOL(__dev_remove_pack);

L
Linus Torvalds 已提交
429 430 431 432 433 434 435 436 437 438 439 440 441 442 443
/**
 *	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);
444

L
Linus Torvalds 已提交
445 446
	synchronize_net();
}
E
Eric Dumazet 已提交
447
EXPORT_SYMBOL(dev_remove_pack);
L
Linus Torvalds 已提交
448

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

/**
 *	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
485
static void __dev_remove_offload(struct packet_offload *po)
486 487 488 489
{
	struct list_head *head = &offload_base;
	struct packet_offload *po1;

490
	spin_lock(&offload_lock);
491 492 493 494 495 496 497 498 499 500

	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:
501
	spin_unlock(&offload_lock);
502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523
}

/**
 *	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
524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550
/******************************************************************************

		      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));
551
			strlcpy(s[i].name, name, IFNAMSIZ);
L
Linus Torvalds 已提交
552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
			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] != ' ' &&
576
		    !strcmp(dev->name, s[i].name)) {
L
Linus Torvalds 已提交
577 578 579 580 581 582 583 584 585
			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
586
EXPORT_SYMBOL(netdev_boot_setup_check);
L
Linus Torvalds 已提交
587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610


/**
 *	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
	 */
611
	if (__dev_get_by_name(&init_net, name))
L
Linus Torvalds 已提交
612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656
		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
657
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
658 659 660 661 662 663 664 665 666
 *	@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.
 */

667
struct net_device *__dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
668
{
E
Eric Dumazet 已提交
669 670
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);
L
Linus Torvalds 已提交
671

672
	hlist_for_each_entry(dev, head, name_hlist)
L
Linus Torvalds 已提交
673 674
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;
E
Eric Dumazet 已提交
675

L
Linus Torvalds 已提交
676 677
	return NULL;
}
E
Eric Dumazet 已提交
678
EXPORT_SYMBOL(__dev_get_by_name);
L
Linus Torvalds 已提交
679

680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696
/**
 *	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);

697
	hlist_for_each_entry_rcu(dev, head, name_hlist)
698 699 700 701 702 703 704
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_name_rcu);

L
Linus Torvalds 已提交
705 706
/**
 *	dev_get_by_name		- find a device by its name
707
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
708 709 710 711 712 713 714 715 716
 *	@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.
 */

717
struct net_device *dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
718 719 720
{
	struct net_device *dev;

721 722
	rcu_read_lock();
	dev = dev_get_by_name_rcu(net, name);
L
Linus Torvalds 已提交
723 724
	if (dev)
		dev_hold(dev);
725
	rcu_read_unlock();
L
Linus Torvalds 已提交
726 727
	return dev;
}
E
Eric Dumazet 已提交
728
EXPORT_SYMBOL(dev_get_by_name);
L
Linus Torvalds 已提交
729 730 731

/**
 *	__dev_get_by_index - find a device by its ifindex
732
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
733 734 735 736 737 738 739 740 741
 *	@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.
 */

742
struct net_device *__dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
743
{
E
Eric Dumazet 已提交
744 745
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);
L
Linus Torvalds 已提交
746

747
	hlist_for_each_entry(dev, head, index_hlist)
L
Linus Torvalds 已提交
748 749
		if (dev->ifindex == ifindex)
			return dev;
E
Eric Dumazet 已提交
750

L
Linus Torvalds 已提交
751 752
	return NULL;
}
E
Eric Dumazet 已提交
753
EXPORT_SYMBOL(__dev_get_by_index);
L
Linus Torvalds 已提交
754

755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770
/**
 *	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);

771
	hlist_for_each_entry_rcu(dev, head, index_hlist)
772 773 774 775 776 777 778
		if (dev->ifindex == ifindex)
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_index_rcu);

L
Linus Torvalds 已提交
779 780 781

/**
 *	dev_get_by_index - find a device by its ifindex
782
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
783 784 785 786 787 788 789 790
 *	@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.
 */

791
struct net_device *dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
792 793 794
{
	struct net_device *dev;

795 796
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, ifindex);
L
Linus Torvalds 已提交
797 798
	if (dev)
		dev_hold(dev);
799
	rcu_read_unlock();
L
Linus Torvalds 已提交
800 801
	return dev;
}
E
Eric Dumazet 已提交
802
EXPORT_SYMBOL(dev_get_by_index);
L
Linus Torvalds 已提交
803

804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837
/**
 *	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
838
/**
839
 *	dev_getbyhwaddr_rcu - find a device by its hardware address
840
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
841 842 843 844
 *	@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
845 846
 *	is not found or a pointer to the device.
 *	The caller must hold RCU or RTNL.
847
 *	The returned device has not had its ref count increased
L
Linus Torvalds 已提交
848 849 850 851
 *	and the caller must therefore be careful about locking
 *
 */

852 853
struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
				       const char *ha)
L
Linus Torvalds 已提交
854 855 856
{
	struct net_device *dev;

857
	for_each_netdev_rcu(net, dev)
L
Linus Torvalds 已提交
858 859
		if (dev->type == type &&
		    !memcmp(dev->dev_addr, ha, dev->addr_len))
860 861 862
			return dev;

	return NULL;
L
Linus Torvalds 已提交
863
}
864
EXPORT_SYMBOL(dev_getbyhwaddr_rcu);
865

866
struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type)
L
Linus Torvalds 已提交
867 868 869
{
	struct net_device *dev;

870
	ASSERT_RTNL();
871
	for_each_netdev(net, dev)
872
		if (dev->type == type)
873 874 875
			return dev;

	return NULL;
876 877 878
}
EXPORT_SYMBOL(__dev_getfirstbyhwtype);

879
struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type)
880
{
881
	struct net_device *dev, *ret = NULL;
882

883 884 885 886 887 888 889 890 891
	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
892 893 894 895
}
EXPORT_SYMBOL(dev_getfirstbyhwtype);

/**
E
Eric Dumazet 已提交
896
 *	dev_get_by_flags_rcu - find any device with given flags
897
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
898 899 900 901
 *	@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
902 903
 *	is not found or a pointer to the device. Must be called inside
 *	rcu_read_lock(), and result refcount is unchanged.
L
Linus Torvalds 已提交
904 905
 */

E
Eric Dumazet 已提交
906
struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short if_flags,
E
Eric Dumazet 已提交
907
				    unsigned short mask)
L
Linus Torvalds 已提交
908
{
909
	struct net_device *dev, *ret;
L
Linus Torvalds 已提交
910

911
	ret = NULL;
912
	for_each_netdev_rcu(net, dev) {
L
Linus Torvalds 已提交
913
		if (((dev->flags ^ if_flags) & mask) == 0) {
914
			ret = dev;
L
Linus Torvalds 已提交
915 916 917
			break;
		}
	}
918
	return ret;
L
Linus Torvalds 已提交
919
}
E
Eric Dumazet 已提交
920
EXPORT_SYMBOL(dev_get_by_flags_rcu);
L
Linus Torvalds 已提交
921 922 923 924 925 926

/**
 *	dev_valid_name - check if name is okay for network device
 *	@name: name string
 *
 *	Network device names need to be valid file names to
927 928
 *	to allow sysfs to work.  We also disallow any kind of
 *	whitespace.
L
Linus Torvalds 已提交
929
 */
930
bool dev_valid_name(const char *name)
L
Linus Torvalds 已提交
931
{
932
	if (*name == '\0')
933
		return false;
934
	if (strlen(name) >= IFNAMSIZ)
935
		return false;
936
	if (!strcmp(name, ".") || !strcmp(name, ".."))
937
		return false;
938 939 940

	while (*name) {
		if (*name == '/' || isspace(*name))
941
			return false;
942 943
		name++;
	}
944
	return true;
L
Linus Torvalds 已提交
945
}
E
Eric Dumazet 已提交
946
EXPORT_SYMBOL(dev_valid_name);
L
Linus Torvalds 已提交
947 948

/**
949 950
 *	__dev_alloc_name - allocate a name for a device
 *	@net: network namespace to allocate the device name in
L
Linus Torvalds 已提交
951
 *	@name: name format string
952
 *	@buf:  scratch buffer and result name string
L
Linus Torvalds 已提交
953 954
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
S
Stephen Hemminger 已提交
955 956 957 958 959 960
 *	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
961 962
 */

963
static int __dev_alloc_name(struct net *net, const char *name, char *buf)
L
Linus Torvalds 已提交
964 965 966 967
{
	int i = 0;
	const char *p;
	const int max_netdevices = 8*PAGE_SIZE;
S
Stephen Hemminger 已提交
968
	unsigned long *inuse;
L
Linus Torvalds 已提交
969 970 971 972 973 974 975 976 977 978 979 980 981
	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
982
		inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC);
L
Linus Torvalds 已提交
983 984 985
		if (!inuse)
			return -ENOMEM;

986
		for_each_netdev(net, d) {
L
Linus Torvalds 已提交
987 988 989 990 991 992
			if (!sscanf(d->name, name, &i))
				continue;
			if (i < 0 || i >= max_netdevices)
				continue;

			/*  avoid cases where sscanf is not exact inverse of printf */
993
			snprintf(buf, IFNAMSIZ, name, i);
L
Linus Torvalds 已提交
994 995 996 997 998 999 1000 1001
			if (!strncmp(buf, d->name, IFNAMSIZ))
				set_bit(i, inuse);
		}

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

1002 1003
	if (buf != name)
		snprintf(buf, IFNAMSIZ, name, i);
1004
	if (!__dev_get_by_name(net, buf))
L
Linus Torvalds 已提交
1005 1006 1007 1008 1009 1010 1011 1012 1013
		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;
}

1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
/**
 *	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;

1034 1035
	BUG_ON(!dev_net(dev));
	net = dev_net(dev);
1036 1037 1038 1039 1040
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
}
E
Eric Dumazet 已提交
1041
EXPORT_SYMBOL(dev_alloc_name);
1042

1043 1044 1045
static int dev_alloc_name_ns(struct net *net,
			     struct net_device *dev,
			     const char *name)
1046
{
1047 1048
	char buf[IFNAMSIZ];
	int ret;
1049

1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
	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);
1061

1062 1063 1064
	if (!dev_valid_name(name))
		return -EINVAL;

1065
	if (strchr(name, '%'))
1066
		return dev_alloc_name_ns(net, dev, name);
1067 1068
	else if (__dev_get_by_name(net, name))
		return -EEXIST;
1069 1070
	else if (dev->name != name)
		strlcpy(dev->name, name, IFNAMSIZ);
1071 1072 1073

	return 0;
}
L
Linus Torvalds 已提交
1074 1075 1076 1077 1078 1079 1080 1081 1082

/**
 *	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.
 */
1083
int dev_change_name(struct net_device *dev, const char *newname)
L
Linus Torvalds 已提交
1084
{
1085
	char oldname[IFNAMSIZ];
L
Linus Torvalds 已提交
1086
	int err = 0;
1087
	int ret;
1088
	struct net *net;
L
Linus Torvalds 已提交
1089 1090

	ASSERT_RTNL();
1091
	BUG_ON(!dev_net(dev));
L
Linus Torvalds 已提交
1092

1093
	net = dev_net(dev);
L
Linus Torvalds 已提交
1094 1095 1096
	if (dev->flags & IFF_UP)
		return -EBUSY;

1097
	write_seqcount_begin(&devnet_rename_seq);
1098 1099

	if (strncmp(newname, dev->name, IFNAMSIZ) == 0) {
1100
		write_seqcount_end(&devnet_rename_seq);
1101
		return 0;
1102
	}
1103

1104 1105
	memcpy(oldname, dev->name, IFNAMSIZ);

1106
	err = dev_get_valid_name(net, dev, newname);
1107
	if (err < 0) {
1108
		write_seqcount_end(&devnet_rename_seq);
1109
		return err;
1110
	}
L
Linus Torvalds 已提交
1111

1112
rollback:
1113 1114 1115
	ret = device_rename(&dev->dev, dev->name);
	if (ret) {
		memcpy(dev->name, oldname, IFNAMSIZ);
1116
		write_seqcount_end(&devnet_rename_seq);
1117
		return ret;
1118
	}
1119

1120
	write_seqcount_end(&devnet_rename_seq);
1121

1122 1123
	netdev_adjacent_rename_links(dev, oldname);

1124
	write_lock_bh(&dev_base_lock);
1125
	hlist_del_rcu(&dev->name_hlist);
1126 1127 1128 1129 1130 1131
	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));
1132 1133
	write_unlock_bh(&dev_base_lock);

1134
	ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
1135 1136 1137
	ret = notifier_to_errno(ret);

	if (ret) {
1138 1139
		/* err >= 0 after dev_alloc_name() or stores the first errno */
		if (err >= 0) {
1140
			err = ret;
1141
			write_seqcount_begin(&devnet_rename_seq);
1142
			memcpy(dev->name, oldname, IFNAMSIZ);
1143
			memcpy(oldname, newname, IFNAMSIZ);
1144
			goto rollback;
1145
		} else {
1146
			pr_err("%s: name change rollback failed: %d\n",
1147
			       dev->name, ret);
1148 1149
		}
	}
L
Linus Torvalds 已提交
1150 1151 1152 1153

	return err;
}

1154 1155 1156 1157
/**
 *	dev_set_alias - change ifalias of a device
 *	@dev: device
 *	@alias: name up to IFALIASZ
1158
 *	@len: limit of bytes to copy from info
1159 1160 1161 1162 1163
 *
 *	Set ifalias for a device,
 */
int dev_set_alias(struct net_device *dev, const char *alias, size_t len)
{
1164 1165
	char *new_ifalias;

1166 1167 1168 1169 1170
	ASSERT_RTNL();

	if (len >= IFALIASZ)
		return -EINVAL;

1171
	if (!len) {
1172 1173
		kfree(dev->ifalias);
		dev->ifalias = NULL;
1174 1175 1176
		return 0;
	}

1177 1178
	new_ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL);
	if (!new_ifalias)
1179
		return -ENOMEM;
1180
	dev->ifalias = new_ifalias;
1181 1182 1183 1184 1185 1186

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


1187
/**
S
Stephen Hemminger 已提交
1188
 *	netdev_features_change - device changes features
1189 1190 1191 1192 1193 1194
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed features.
 */
void netdev_features_change(struct net_device *dev)
{
1195
	call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev);
1196 1197 1198
}
EXPORT_SYMBOL(netdev_features_change);

L
Linus Torvalds 已提交
1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209
/**
 *	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) {
1210
		call_netdevice_notifiers(NETDEV_CHANGE, dev);
1211
		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
L
Linus Torvalds 已提交
1212 1213
	}
}
E
Eric Dumazet 已提交
1214
EXPORT_SYMBOL(netdev_state_change);
L
Linus Torvalds 已提交
1215

1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226
/**
 * 	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)
1227
{
1228 1229 1230
	rtnl_lock();
	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, dev);
	rtnl_unlock();
1231
}
1232
EXPORT_SYMBOL(netdev_notify_peers);
1233

1234
static int __dev_open(struct net_device *dev)
L
Linus Torvalds 已提交
1235
{
1236
	const struct net_device_ops *ops = dev->netdev_ops;
1237
	int ret;
L
Linus Torvalds 已提交
1238

1239 1240
	ASSERT_RTNL();

L
Linus Torvalds 已提交
1241 1242 1243
	if (!netif_device_present(dev))
		return -ENODEV;

1244 1245 1246 1247
	/* 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
	 */
1248
	netpoll_poll_disable(dev);
1249

1250 1251 1252 1253 1254
	ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

L
Linus Torvalds 已提交
1255
	set_bit(__LINK_STATE_START, &dev->state);
1256

1257 1258
	if (ops->ndo_validate_addr)
		ret = ops->ndo_validate_addr(dev);
1259

1260 1261
	if (!ret && ops->ndo_open)
		ret = ops->ndo_open(dev);
L
Linus Torvalds 已提交
1262

1263
	netpoll_poll_enable(dev);
1264

1265 1266 1267
	if (ret)
		clear_bit(__LINK_STATE_START, &dev->state);
	else {
L
Linus Torvalds 已提交
1268
		dev->flags |= IFF_UP;
1269
		net_dmaengine_get();
1270
		dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
1271
		dev_activate(dev);
1272
		add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
1273
	}
1274

L
Linus Torvalds 已提交
1275 1276 1277 1278
	return ret;
}

/**
1279 1280
 *	dev_open	- prepare an interface for use.
 *	@dev:	device to open
L
Linus Torvalds 已提交
1281
 *
1282 1283 1284 1285 1286 1287 1288
 *	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
1289
 */
1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300
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;

1301
	rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1302 1303 1304 1305 1306 1307
	call_netdevice_notifiers(NETDEV_UP, dev);

	return ret;
}
EXPORT_SYMBOL(dev_open);

1308
static int __dev_close_many(struct list_head *head)
L
Linus Torvalds 已提交
1309
{
1310
	struct net_device *dev;
1311

1312
	ASSERT_RTNL();
1313 1314
	might_sleep();

1315
	list_for_each_entry(dev, head, close_list) {
1316
		/* Temporarily disable netpoll until the interface is down */
1317
		netpoll_poll_disable(dev);
1318

1319
		call_netdevice_notifiers(NETDEV_GOING_DOWN, dev);
L
Linus Torvalds 已提交
1320

1321
		clear_bit(__LINK_STATE_START, &dev->state);
L
Linus Torvalds 已提交
1322

1323 1324 1325 1326 1327 1328 1329 1330
		/* Synchronize to scheduled poll. We cannot touch poll list, it
		 * can be even on different cpu. So just clear netif_running().
		 *
		 * dev->stop() will invoke napi_disable() on all of it's
		 * napi_struct instances on this device.
		 */
		smp_mb__after_clear_bit(); /* Commit netif_running(). */
	}
L
Linus Torvalds 已提交
1331

1332
	dev_deactivate_many(head);
1333

1334
	list_for_each_entry(dev, head, close_list) {
1335
		const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
1336

1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348
		/*
		 *	Call the device specific close. This cannot fail.
		 *	Only if device is UP
		 *
		 *	We allow it to be called even after a DETACH hot-plug
		 *	event.
		 */
		if (ops->ndo_stop)
			ops->ndo_stop(dev);

		dev->flags &= ~IFF_UP;
		net_dmaengine_put();
1349
		netpoll_poll_enable(dev);
1350 1351 1352 1353 1354 1355 1356
	}

	return 0;
}

static int __dev_close(struct net_device *dev)
{
1357
	int retval;
1358 1359
	LIST_HEAD(single);

1360
	list_add(&dev->close_list, &single);
1361 1362
	retval = __dev_close_many(&single);
	list_del(&single);
1363

1364
	return retval;
1365 1366
}

E
Eric Dumazet 已提交
1367
static int dev_close_many(struct list_head *head)
1368 1369
{
	struct net_device *dev, *tmp;
L
Linus Torvalds 已提交
1370

1371 1372
	/* Remove the devices that don't need to be closed */
	list_for_each_entry_safe(dev, tmp, head, close_list)
1373
		if (!(dev->flags & IFF_UP))
1374
			list_del_init(&dev->close_list);
1375 1376

	__dev_close_many(head);
L
Linus Torvalds 已提交
1377

1378
	list_for_each_entry_safe(dev, tmp, head, close_list) {
1379
		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1380
		call_netdevice_notifiers(NETDEV_DOWN, dev);
1381
		list_del_init(&dev->close_list);
1382
	}
1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397

	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)
{
1398 1399
	if (dev->flags & IFF_UP) {
		LIST_HEAD(single);
L
Linus Torvalds 已提交
1400

1401
		list_add(&dev->close_list, &single);
1402 1403 1404
		dev_close_many(&single);
		list_del(&single);
	}
1405
	return 0;
L
Linus Torvalds 已提交
1406
}
E
Eric Dumazet 已提交
1407
EXPORT_SYMBOL(dev_close);
L
Linus Torvalds 已提交
1408 1409


1410 1411 1412 1413 1414 1415 1416 1417 1418 1419
/**
 *	dev_disable_lro - disable Large Receive Offload on a device
 *	@dev: device
 *
 *	Disable Large Receive Offload (LRO) on a net device.  Must be
 *	called under RTNL.  This is needed if received packets may be
 *	forwarded to another interface.
 */
void dev_disable_lro(struct net_device *dev)
{
1420 1421 1422 1423 1424 1425 1426
	/*
	 * If we're trying to disable lro on a vlan device
	 * use the underlying physical device instead
	 */
	if (is_vlan_dev(dev))
		dev = vlan_dev_real_dev(dev);

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

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

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

1439 1440 1441 1442 1443 1444 1445 1446
static int call_netdevice_notifier(struct notifier_block *nb, unsigned long val,
				   struct net_device *dev)
{
	struct netdev_notifier_info info;

	netdev_notifier_info_init(&info, dev);
	return nb->notifier_call(nb, val, &info);
}
1447

1448 1449
static int dev_boot_phase = 1;

L
Linus Torvalds 已提交
1450 1451 1452 1453 1454 1455 1456 1457 1458 1459
/**
 *	register_netdevice_notifier - register a network notifier block
 *	@nb: notifier
 *
 *	Register a notifier to be called when network device events occur.
 *	The notifier passed is linked into the kernel structures and must
 *	not be reused until it has been unregistered. A negative errno code
 *	is returned on a failure.
 *
 * 	When registered all registration and up events are replayed
1460
 *	to the new notifier to allow device to have a race free
L
Linus Torvalds 已提交
1461 1462 1463 1464 1465 1466
 *	view of the network device list.
 */

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

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

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

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

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

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

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

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

/**
 *	unregister_netdevice_notifier - unregister a network notifier block
 *	@nb: notifier
 *
 *	Unregister a notifier previously registered by
 *	register_netdevice_notifier(). The notifier is unlinked into the
 *	kernel structures and may then be reused. A negative errno code
 *	is returned on a failure.
1525 1526 1527 1528
 *
 * 	After unregistering unregister and down device events are synthesized
 *	for all devices on the device list to the removed notifier to remove
 *	the need for special case cleanup code.
L
Linus Torvalds 已提交
1529 1530 1531 1532
 */

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

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

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

1558 1559 1560 1561 1562 1563 1564 1565 1566 1567
/**
 *	call_netdevice_notifiers_info - call all network notifier blocks
 *	@val: value passed unmodified to notifier function
 *	@dev: net_device pointer passed unmodified to notifier function
 *	@info: notifier information data
 *
 *	Call all network notifier blocks.  Parameters and return value
 *	are as for raw_notifier_call_chain().
 */

S
stephen hemminger 已提交
1568 1569 1570
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info)
1571 1572 1573 1574 1575 1576
{
	ASSERT_RTNL();
	netdev_notifier_info_init(info, dev);
	return raw_notifier_call_chain(&netdev_chain, val, info);
}

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

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

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

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

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

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

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

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

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

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

H
Herbert Xu 已提交
1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686
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);

1687 1688 1689 1690 1691 1692 1693 1694
/**
 * dev_forward_skb - loopback an skb to another netif
 *
 * @dev: destination network device
 * @skb: buffer to forward
 *
 * return values:
 *	NET_RX_SUCCESS	(no congestion)
1695
 *	NET_RX_DROP     (packet was dropped, but freed)
1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706
 *
 * 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
1707
	return __dev_forward_skb(dev, skb) ?: netif_rx_internal(skb);
1708 1709 1710
}
EXPORT_SYMBOL_GPL(dev_forward_skb);

1711 1712 1713 1714
static inline int deliver_skb(struct sk_buff *skb,
			      struct packet_type *pt_prev,
			      struct net_device *orig_dev)
{
1715 1716
	if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
		return -ENOMEM;
1717 1718 1719 1720
	atomic_inc(&skb->users);
	return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
}

1721 1722
static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
{
1723
	if (!ptype->af_packet_priv || !skb->sk)
1724 1725 1726 1727 1728 1729 1730 1731 1732 1733
		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
1734 1735 1736 1737 1738
/*
 *	Support routine. Sends outgoing frames to any network
 *	taps currently in use.
 */

1739
static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
L
Linus Torvalds 已提交
1740 1741
{
	struct packet_type *ptype;
1742 1743
	struct sk_buff *skb2 = NULL;
	struct packet_type *pt_prev = NULL;
1744

L
Linus Torvalds 已提交
1745 1746 1747 1748 1749 1750
	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) &&
1751
		    (!skb_loop_sk(ptype, skb))) {
1752 1753 1754 1755 1756 1757 1758
			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
1759 1760 1761
			if (!skb2)
				break;

1762 1763
			net_timestamp_set(skb2);

L
Linus Torvalds 已提交
1764 1765 1766 1767
			/* skb->nh should be correctly
			   set by sender, so that the second statement is
			   just protection against buggy protocols.
			 */
1768
			skb_reset_mac_header(skb2);
L
Linus Torvalds 已提交
1769

1770
			if (skb_network_header(skb2) < skb2->data ||
1771
			    skb_network_header(skb2) > skb_tail_pointer(skb2)) {
1772 1773 1774
				net_crit_ratelimited("protocol %04x is buggy, dev %s\n",
						     ntohs(skb2->protocol),
						     dev->name);
1775
				skb_reset_network_header(skb2);
L
Linus Torvalds 已提交
1776 1777
			}

1778
			skb2->transport_header = skb2->network_header;
L
Linus Torvalds 已提交
1779
			skb2->pkt_type = PACKET_OUTGOING;
1780
			pt_prev = ptype;
L
Linus Torvalds 已提交
1781 1782
		}
	}
1783 1784
	if (pt_prev)
		pt_prev->func(skb2, skb->dev, pt_prev, skb->dev);
L
Linus Torvalds 已提交
1785 1786 1787
	rcu_read_unlock();
}

1788 1789
/**
 * netif_setup_tc - Handle tc mappings on real_num_tx_queues change
1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800
 * @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
1801
static void netif_setup_tc(struct net_device *dev, unsigned int txq)
1802 1803 1804 1805 1806 1807
{
	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) {
1808
		pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n");
1809 1810 1811 1812 1813 1814 1815 1816 1817 1818
		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) {
1819 1820
			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);
1821 1822 1823 1824 1825
			netdev_set_prio_tc_map(dev, i, 0);
		}
	}
}

1826 1827 1828 1829 1830
#ifdef CONFIG_XPS
static DEFINE_MUTEX(xps_map_mutex);
#define xmap_dereference(P)		\
	rcu_dereference_protected((P), lockdep_is_held(&xps_map_mutex))

1831 1832
static struct xps_map *remove_xps_queue(struct xps_dev_maps *dev_maps,
					int cpu, u16 index)
1833
{
1834 1835
	struct xps_map *map = NULL;
	int pos;
1836

1837 1838
	if (dev_maps)
		map = xmap_dereference(dev_maps->cpu_map[cpu]);
1839

1840 1841
	for (pos = 0; map && pos < map->len; pos++) {
		if (map->queues[pos] == index) {
1842 1843 1844
			if (map->len > 1) {
				map->queues[pos] = map->queues[--map->len];
			} else {
1845
				RCU_INIT_POINTER(dev_maps->cpu_map[cpu], NULL);
1846 1847 1848
				kfree_rcu(map, rcu);
				map = NULL;
			}
1849
			break;
1850 1851 1852
		}
	}

1853 1854 1855
	return map;
}

1856
static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
1857 1858
{
	struct xps_dev_maps *dev_maps;
1859
	int cpu, i;
1860 1861 1862 1863 1864 1865 1866 1867 1868
	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) {
1869 1870 1871 1872 1873
		for (i = index; i < dev->num_tx_queues; i++) {
			if (!remove_xps_queue(dev_maps, cpu, i))
				break;
		}
		if (i == dev->num_tx_queues)
1874 1875 1876 1877
			active = true;
	}

	if (!active) {
1878 1879 1880 1881
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

1882 1883 1884 1885
	for (i = index; i < dev->num_tx_queues; i++)
		netdev_queue_numa_node_write(netdev_get_tx_queue(dev, i),
					     NUMA_NO_NODE);

1886 1887 1888 1889
out_no_maps:
	mutex_unlock(&xps_map_mutex);
}

1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924
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;
}

1925 1926
int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
			u16 index)
1927
{
1928
	struct xps_dev_maps *dev_maps, *new_dev_maps = NULL;
1929 1930
	struct xps_map *map, *new_map;
	int maps_sz = max_t(unsigned int, XPS_DEV_MAPS_SIZE, L1_CACHE_BYTES);
1931 1932
	int cpu, numa_node_id = -2;
	bool active = false;
1933 1934 1935 1936 1937

	mutex_lock(&xps_map_mutex);

	dev_maps = xmap_dereference(dev->xps_maps);

1938 1939 1940 1941 1942 1943 1944
	/* 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);
1945 1946
		if (!new_dev_maps) {
			mutex_unlock(&xps_map_mutex);
1947
			return -ENOMEM;
1948
		}
1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962

		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;

1963
	for_each_possible_cpu(cpu) {
1964 1965 1966 1967 1968 1969 1970 1971 1972 1973
		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;
1974 1975 1976 1977 1978 1979
#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
1980 1981 1982 1983
		} 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);
1984
		}
1985

1986 1987
	}

1988 1989
	rcu_assign_pointer(dev->xps_maps, new_dev_maps);

1990
	/* Cleanup old maps */
1991 1992 1993 1994 1995 1996 1997
	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);
		}
1998

1999
		kfree_rcu(dev_maps, rcu);
2000 2001
	}

2002 2003
	dev_maps = new_dev_maps;
	active = true;
2004

2005 2006
out_no_new_maps:
	/* update Tx queue numa node */
2007 2008 2009 2010
	netdev_queue_numa_node_write(netdev_get_tx_queue(dev, index),
				     (numa_node_id >= 0) ? numa_node_id :
				     NUMA_NO_NODE);

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029
	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:
2030 2031 2032 2033
	mutex_unlock(&xps_map_mutex);

	return 0;
error:
2034 2035 2036 2037 2038 2039 2040 2041 2042
	/* 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);
	}

2043 2044 2045 2046 2047 2048 2049 2050
	mutex_unlock(&xps_map_mutex);

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

#endif
2051 2052 2053 2054
/*
 * 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.
 */
2055
int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
2056
{
T
Tom Herbert 已提交
2057 2058
	int rc;

2059 2060
	if (txq < 1 || txq > dev->num_tx_queues)
		return -EINVAL;
2061

2062 2063
	if (dev->reg_state == NETREG_REGISTERED ||
	    dev->reg_state == NETREG_UNREGISTERING) {
2064 2065
		ASSERT_RTNL();

T
Tom Herbert 已提交
2066 2067
		rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
						  txq);
T
Tom Herbert 已提交
2068 2069 2070
		if (rc)
			return rc;

2071 2072 2073
		if (dev->num_tc)
			netif_setup_tc(dev, txq);

2074
		if (txq < dev->real_num_tx_queues) {
2075
			qdisc_reset_all_tx_gt(dev, txq);
2076 2077 2078 2079
#ifdef CONFIG_XPS
			netif_reset_xps_queues_gt(dev, txq);
#endif
		}
2080
	}
2081 2082 2083

	dev->real_num_tx_queues = txq;
	return 0;
2084 2085
}
EXPORT_SYMBOL(netif_set_real_num_tx_queues);
2086

2087
#ifdef CONFIG_SYSFS
2088 2089 2090 2091 2092 2093 2094
/**
 *	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
2095 2096
 *	negative error code.  If called before registration, it always
 *	succeeds.
2097 2098 2099 2100 2101
 */
int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
{
	int rc;

T
Tom Herbert 已提交
2102 2103 2104
	if (rxq < 1 || rxq > dev->num_rx_queues)
		return -EINVAL;

2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119
	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

2120 2121
/**
 * netif_get_num_default_rss_queues - default number of RSS queues
2122 2123 2124 2125
 *
 * This routine should set an upper limit on the number of RSS queues
 * used by default by multiqueue devices.
 */
2126
int netif_get_num_default_rss_queues(void)
2127 2128 2129 2130 2131
{
	return min_t(int, DEFAULT_MAX_NUM_RSS_QUEUES, num_online_cpus());
}
EXPORT_SYMBOL(netif_get_num_default_rss_queues);

2132
static inline void __netif_reschedule(struct Qdisc *q)
2133
{
2134 2135
	struct softnet_data *sd;
	unsigned long flags;
2136

2137 2138
	local_irq_save(flags);
	sd = &__get_cpu_var(softnet_data);
2139 2140 2141
	q->next_sched = NULL;
	*sd->output_queue_tailp = q;
	sd->output_queue_tailp = &q->next_sched;
2142 2143 2144 2145 2146 2147 2148 2149
	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);
2150 2151 2152
}
EXPORT_SYMBOL(__netif_schedule);

2153 2154 2155 2156 2157
struct dev_kfree_skb_cb {
	enum skb_free_reason reason;
};

static struct dev_kfree_skb_cb *get_kfree_skb_cb(const struct sk_buff *skb)
2158
{
2159 2160 2161 2162
	return (struct dev_kfree_skb_cb *)skb->cb;
}

void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason)
2163
{
2164
	unsigned long flags;
2165

2166 2167 2168 2169 2170
	if (likely(atomic_read(&skb->users) == 1)) {
		smp_rmb();
		atomic_set(&skb->users, 0);
	} else if (likely(!atomic_dec_and_test(&skb->users))) {
		return;
2171
	}
2172 2173 2174 2175 2176 2177
	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);
2178
}
2179
EXPORT_SYMBOL(__dev_kfree_skb_irq);
2180

2181
void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason)
2182 2183
{
	if (in_irq() || irqs_disabled())
2184
		__dev_kfree_skb_irq(skb, reason);
2185 2186 2187
	else
		dev_kfree_skb(skb);
}
2188
EXPORT_SYMBOL(__dev_kfree_skb_any);
2189 2190


2191 2192 2193 2194 2195 2196
/**
 * netif_device_detach - mark device as removed
 * @dev: network device
 *
 * Mark device as removed from system and therefore no longer available.
 */
2197 2198 2199 2200
void netif_device_detach(struct net_device *dev)
{
	if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2201
		netif_tx_stop_all_queues(dev);
2202 2203 2204 2205
	}
}
EXPORT_SYMBOL(netif_device_detach);

2206 2207 2208 2209 2210 2211
/**
 * netif_device_attach - mark device as attached
 * @dev: network device
 *
 * Mark device as attached from system and restart if needed.
 */
2212 2213 2214 2215
void netif_device_attach(struct net_device *dev)
{
	if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2216
		netif_tx_wake_all_queues(dev);
2217
		__netdev_watchdog_up(dev);
2218 2219 2220 2221
	}
}
EXPORT_SYMBOL(netif_device_attach);

2222 2223
static void skb_warn_bad_offload(const struct sk_buff *skb)
{
2224
	static const netdev_features_t null_features = 0;
2225 2226 2227
	struct net_device *dev = skb->dev;
	const char *driver = "";

2228 2229 2230
	if (!net_ratelimit())
		return;

2231 2232 2233 2234 2235
	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",
2236 2237
	     driver, dev ? &dev->features : &null_features,
	     skb->sk ? &skb->sk->sk_route_caps : &null_features,
2238 2239 2240 2241
	     skb->len, skb->data_len, skb_shinfo(skb)->gso_size,
	     skb_shinfo(skb)->gso_type, skb->ip_summed);
}

L
Linus Torvalds 已提交
2242 2243 2244 2245
/*
 * Invalidate hardware checksum when packet is to be mangled, and
 * complete checksum manually on outgoing path.
 */
2246
int skb_checksum_help(struct sk_buff *skb)
L
Linus Torvalds 已提交
2247
{
2248
	__wsum csum;
2249
	int ret = 0, offset;
L
Linus Torvalds 已提交
2250

2251
	if (skb->ip_summed == CHECKSUM_COMPLETE)
2252 2253 2254
		goto out_set_summed;

	if (unlikely(skb_shinfo(skb)->gso_size)) {
2255 2256
		skb_warn_bad_offload(skb);
		return -EINVAL;
L
Linus Torvalds 已提交
2257 2258
	}

2259 2260 2261 2262 2263 2264 2265 2266 2267
	/* 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;
	}

2268
	offset = skb_checksum_start_offset(skb);
2269 2270 2271 2272 2273 2274 2275 2276
	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
2277 2278 2279 2280 2281
		ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
		if (ret)
			goto out;
	}

2282
	*(__sum16 *)(skb->data + offset) = csum_fold(csum);
2283
out_set_summed:
L
Linus Torvalds 已提交
2284
	skb->ip_summed = CHECKSUM_NONE;
2285
out:
L
Linus Torvalds 已提交
2286 2287
	return ret;
}
E
Eric Dumazet 已提交
2288
EXPORT_SYMBOL(skb_checksum_help);
L
Linus Torvalds 已提交
2289

2290
__be16 skb_network_protocol(struct sk_buff *skb, int *depth)
2291
{
A
Al Viro 已提交
2292
	__be16 type = skb->protocol;
2293
	int vlan_depth = skb->mac_len;
2294

2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305
	/* Tunnel gso handlers can set protocol to ethernet. */
	if (type == htons(ETH_P_TEB)) {
		struct ethhdr *eth;

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

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

P
Patrick McHardy 已提交
2306
	while (type == htons(ETH_P_8021Q) || type == htons(ETH_P_8021AD)) {
2307
		struct vlan_hdr *vh;
2308

2309
		if (unlikely(!pskb_may_pull(skb, vlan_depth + VLAN_HLEN)))
2310
			return 0;
2311

2312 2313 2314
		vh = (struct vlan_hdr *)(skb->data + vlan_depth);
		type = vh->h_vlan_encapsulated_proto;
		vlan_depth += VLAN_HLEN;
2315 2316
	}

2317 2318
	*depth = vlan_depth;

2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331
	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;
2332 2333
	int vlan_depth = skb->mac_len;
	__be16 type = skb_network_protocol(skb, &vlan_depth);
2334 2335 2336 2337

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

2338
	__skb_pull(skb, vlan_depth);
2339 2340

	rcu_read_lock();
2341
	list_for_each_entry_rcu(ptype, &offload_base, list) {
2342
		if (ptype->type == type && ptype->callbacks.gso_segment) {
2343
			if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) {
2344 2345
				int err;

2346
				err = ptype->callbacks.gso_send_check(skb);
2347 2348 2349
				segs = ERR_PTR(err);
				if (err || skb_gso_ok(skb, features))
					break;
2350 2351
				__skb_push(skb, (skb->data -
						 skb_network_header(skb)));
2352
			}
2353
			segs = ptype->callbacks.gso_segment(skb, features);
2354 2355 2356 2357 2358
			break;
		}
	}
	rcu_read_unlock();

2359
	__skb_push(skb, skb->data - skb_mac_header(skb));
2360

2361 2362
	return segs;
}
2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399
EXPORT_SYMBOL(skb_mac_gso_segment);


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

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

		skb_warn_bad_offload(skb);

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

2400
	SKB_GSO_CB(skb)->mac_offset = skb_headroom(skb);
2401 2402
	SKB_GSO_CB(skb)->encap_level = 0;

2403 2404 2405 2406 2407
	skb_reset_mac_header(skb);
	skb_reset_mac_len(skb);

	return skb_mac_gso_segment(skb, features);
}
2408
EXPORT_SYMBOL(__skb_gso_segment);
2409

2410 2411 2412 2413 2414
/* Take action when hardware reception checksum errors are detected. */
#ifdef CONFIG_BUG
void netdev_rx_csum_fault(struct net_device *dev)
{
	if (net_ratelimit()) {
2415
		pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>");
2416 2417 2418 2419 2420 2421
		dump_stack();
	}
}
EXPORT_SYMBOL(netdev_rx_csum_fault);
#endif

L
Linus Torvalds 已提交
2422 2423 2424 2425 2426
/* 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.
 */

2427
static int illegal_highdma(const struct net_device *dev, struct sk_buff *skb)
L
Linus Torvalds 已提交
2428
{
2429
#ifdef CONFIG_HIGHMEM
L
Linus Torvalds 已提交
2430
	int i;
2431
	if (!(dev->features & NETIF_F_HIGHDMA)) {
2432 2433 2434
		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)))
2435
				return 1;
2436
		}
2437
	}
L
Linus Torvalds 已提交
2438

2439 2440
	if (PCI_DMA_BUS_IS_PHYS) {
		struct device *pdev = dev->dev.parent;
L
Linus Torvalds 已提交
2441

E
Eric Dumazet 已提交
2442 2443
		if (!pdev)
			return 0;
2444
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2445 2446
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			dma_addr_t addr = page_to_phys(skb_frag_page(frag));
2447 2448 2449 2450
			if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask)
				return 1;
		}
	}
2451
#endif
L
Linus Torvalds 已提交
2452 2453 2454
	return 0;
}

2455 2456 2457 2458 2459 2460 2461 2462 2463 2464
struct dev_gso_cb {
	void (*destructor)(struct sk_buff *skb);
};

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

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

E
Eric Dumazet 已提交
2465 2466
	kfree_skb_list(skb->next);
	skb->next = NULL;
2467 2468 2469 2470 2471 2472 2473 2474 2475

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

/**
 *	dev_gso_segment - Perform emulated hardware segmentation on skb.
 *	@skb: buffer to segment
2476
 *	@features: device features as applicable to this skb
2477 2478 2479 2480
 *
 *	This function segments the given skb and stores the list of segments
 *	in skb->next.
 */
2481
static int dev_gso_segment(struct sk_buff *skb, netdev_features_t features)
2482 2483
{
	struct sk_buff *segs;
2484 2485 2486 2487 2488 2489

	segs = skb_gso_segment(skb, features);

	/* Verifying header integrity only. */
	if (!segs)
		return 0;
2490

2491
	if (IS_ERR(segs))
2492 2493 2494 2495 2496 2497 2498 2499 2500
		return PTR_ERR(segs);

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

	return 0;
}

2501
static netdev_features_t harmonize_features(struct sk_buff *skb,
2502 2503
					    const struct net_device *dev,
					    netdev_features_t features)
2504
{
2505 2506
	int tmp;

2507
	if (skb->ip_summed != CHECKSUM_NONE &&
2508
	    !can_checksum_protocol(features, skb_network_protocol(skb, &tmp))) {
2509
		features &= ~NETIF_F_ALL_CSUM;
2510
	} else if (illegal_highdma(dev, skb)) {
2511 2512 2513 2514 2515 2516
		features &= ~NETIF_F_SG;
	}

	return features;
}

2517 2518
netdev_features_t netif_skb_dev_features(struct sk_buff *skb,
					 const struct net_device *dev)
2519 2520
{
	__be16 protocol = skb->protocol;
2521
	netdev_features_t features = dev->features;
2522

2523
	if (skb_shinfo(skb)->gso_segs > dev->gso_max_segs)
2524 2525
		features &= ~NETIF_F_GSO_MASK;

P
Patrick McHardy 已提交
2526
	if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD)) {
2527 2528
		struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
		protocol = veh->h_vlan_encapsulated_proto;
2529
	} else if (!vlan_tx_tag_present(skb)) {
2530
		return harmonize_features(skb, dev, features);
2531
	}
2532

2533
	features &= (dev->vlan_features | NETIF_F_HW_VLAN_CTAG_TX |
P
Patrick McHardy 已提交
2534
					       NETIF_F_HW_VLAN_STAG_TX);
2535

2536
	if (protocol == htons(ETH_P_8021Q) || protocol == htons(ETH_P_8021AD))
2537
		features &= NETIF_F_SG | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST |
P
Patrick McHardy 已提交
2538 2539
				NETIF_F_GEN_CSUM | NETIF_F_HW_VLAN_CTAG_TX |
				NETIF_F_HW_VLAN_STAG_TX;
2540

2541
	return harmonize_features(skb, dev, features);
2542
}
2543
EXPORT_SYMBOL(netif_skb_dev_features);
2544

2545
int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2546
			struct netdev_queue *txq)
2547
{
2548
	const struct net_device_ops *ops = dev->netdev_ops;
2549
	int rc = NETDEV_TX_OK;
2550
	unsigned int skb_len;
2551

2552
	if (likely(!skb->next)) {
2553
		netdev_features_t features;
2554

2555
		/*
L
Lucas De Marchi 已提交
2556
		 * If device doesn't need skb->dst, release it right now while
2557 2558
		 * its hot in this cpu cache
		 */
E
Eric Dumazet 已提交
2559 2560 2561
		if (dev->priv_flags & IFF_XMIT_DST_RELEASE)
			skb_dst_drop(skb);

2562 2563
		features = netif_skb_features(skb);

2564
		if (vlan_tx_tag_present(skb) &&
2565 2566 2567
		    !vlan_hw_offload_capable(features, skb->vlan_proto)) {
			skb = __vlan_put_tag(skb, skb->vlan_proto,
					     vlan_tx_tag_get(skb));
2568 2569 2570 2571 2572 2573
			if (unlikely(!skb))
				goto out;

			skb->vlan_tci = 0;
		}

2574 2575 2576 2577 2578 2579 2580
		/* 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;

2581
		if (netif_needs_gso(skb, features)) {
2582
			if (unlikely(dev_gso_segment(skb, features)))
2583 2584 2585
				goto out_kfree_skb;
			if (skb->next)
				goto gso;
2586
		} else {
2587
			if (skb_needs_linearize(skb, features) &&
2588 2589 2590 2591 2592 2593 2594 2595
			    __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) {
2596 2597 2598 2599 2600 2601
				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));
2602
				if (!(features & NETIF_F_ALL_CSUM) &&
2603 2604 2605
				     skb_checksum_help(skb))
					goto out_kfree_skb;
			}
2606 2607
		}

2608 2609 2610
		if (!list_empty(&ptype_all))
			dev_queue_xmit_nit(skb, dev);

2611
		skb_len = skb->len;
2612
		trace_net_dev_start_xmit(skb, dev);
2613
		rc = ops->ndo_start_xmit(skb, dev);
2614
		trace_net_dev_xmit(skb, rc, dev, skb_len);
2615
		if (rc == NETDEV_TX_OK)
E
Eric Dumazet 已提交
2616
			txq_trans_update(txq);
2617
		return rc;
2618 2619
	}

2620
gso:
2621 2622 2623 2624 2625
	do {
		struct sk_buff *nskb = skb->next;

		skb->next = nskb->next;
		nskb->next = NULL;
2626

2627 2628 2629
		if (!list_empty(&ptype_all))
			dev_queue_xmit_nit(nskb, dev);

2630
		skb_len = nskb->len;
2631
		trace_net_dev_start_xmit(nskb, dev);
2632
		rc = ops->ndo_start_xmit(nskb, dev);
2633
		trace_net_dev_xmit(nskb, rc, dev, skb_len);
2634
		if (unlikely(rc != NETDEV_TX_OK)) {
2635 2636
			if (rc & ~NETDEV_TX_MASK)
				goto out_kfree_gso_skb;
2637
			nskb->next = skb->next;
2638 2639 2640
			skb->next = nskb;
			return rc;
		}
E
Eric Dumazet 已提交
2641
		txq_trans_update(txq);
2642
		if (unlikely(netif_xmit_stopped(txq) && skb->next))
2643
			return NETDEV_TX_BUSY;
2644
	} while (skb->next);
2645

2646
out_kfree_gso_skb:
2647
	if (likely(skb->next == NULL)) {
2648
		skb->destructor = DEV_GSO_CB(skb)->destructor;
2649 2650 2651
		consume_skb(skb);
		return rc;
	}
2652 2653
out_kfree_skb:
	kfree_skb(skb);
2654
out:
2655
	return rc;
2656
}
2657
EXPORT_SYMBOL_GPL(dev_hard_start_xmit);
2658

2659 2660 2661 2662 2663 2664 2665 2666 2667 2668
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)  {
2669
		unsigned int hdr_len;
2670
		u16 gso_segs = shinfo->gso_segs;
2671

2672 2673 2674 2675
		/* mac layer + network layer */
		hdr_len = skb_transport_header(skb) - skb_mac_header(skb);

		/* + transport layer */
2676 2677 2678 2679
		if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)))
			hdr_len += tcp_hdrlen(skb);
		else
			hdr_len += sizeof(struct udphdr);
2680 2681 2682 2683 2684 2685

		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;
2686 2687 2688
	}
}

2689 2690 2691 2692 2693
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
2694
	bool contended;
2695 2696
	int rc;

2697
	qdisc_pkt_len_init(skb);
E
Eric Dumazet 已提交
2698
	qdisc_calculate_pkt_len(skb, q);
2699 2700 2701 2702 2703 2704
	/*
	 * Heuristic to force contended enqueues to serialize on a
	 * separate lock before trying to get qdisc main lock.
	 * This permits __QDISC_STATE_RUNNING owner to get the lock more often
	 * and dequeue packets faster.
	 */
E
Eric Dumazet 已提交
2705
	contended = qdisc_is_running(q);
2706 2707 2708
	if (unlikely(contended))
		spin_lock(&q->busylock);

2709 2710 2711 2712 2713
	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) &&
2714
		   qdisc_run_begin(q)) {
2715 2716 2717 2718 2719
		/*
		 * 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
2720 2721
		if (!(dev->priv_flags & IFF_XMIT_DST_RELEASE))
			skb_dst_force(skb);
2722 2723 2724

		qdisc_bstats_update(q, skb);

2725 2726 2727 2728 2729
		if (sch_direct_xmit(skb, q, dev, txq, root_lock)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
2730
			__qdisc_run(q);
2731
		} else
2732
			qdisc_run_end(q);
2733 2734 2735

		rc = NET_XMIT_SUCCESS;
	} else {
E
Eric Dumazet 已提交
2736
		skb_dst_force(skb);
E
Eric Dumazet 已提交
2737
		rc = q->enqueue(skb, q) & NET_XMIT_MASK;
2738 2739 2740 2741 2742 2743 2744
		if (qdisc_run_begin(q)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
			__qdisc_run(q);
		}
2745 2746
	}
	spin_unlock(root_lock);
2747 2748
	if (unlikely(contended))
		spin_unlock(&q->busylock);
2749 2750 2751
	return rc;
}

2752
#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2753 2754
static void skb_update_prio(struct sk_buff *skb)
{
I
Igor Maravic 已提交
2755
	struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap);
2756

2757 2758 2759 2760 2761 2762
	if (!skb->priority && skb->sk && map) {
		unsigned int prioidx = skb->sk->sk_cgrp_prioidx;

		if (prioidx < map->priomap_len)
			skb->priority = map->priomap[prioidx];
	}
2763 2764 2765 2766 2767
}
#else
#define skb_update_prio(skb)
#endif

2768
static DEFINE_PER_CPU(int, xmit_recursion);
2769
#define RECURSION_LIMIT 10
2770

2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787
/**
 *	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);

2788
/**
2789
 *	__dev_queue_xmit - transmit a buffer
2790
 *	@skb: buffer to transmit
2791
 *	@accel_priv: private data used for L2 forwarding offload
2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813
 *
 *	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
 */
2814
static int __dev_queue_xmit(struct sk_buff *skb, void *accel_priv)
L
Linus Torvalds 已提交
2815 2816
{
	struct net_device *dev = skb->dev;
2817
	struct netdev_queue *txq;
L
Linus Torvalds 已提交
2818 2819 2820
	struct Qdisc *q;
	int rc = -ENOMEM;

2821 2822
	skb_reset_mac_header(skb);

2823 2824
	/* Disable soft irqs for various locks below. Also
	 * stops preemption for RCU.
L
Linus Torvalds 已提交
2825
	 */
2826
	rcu_read_lock_bh();
L
Linus Torvalds 已提交
2827

2828 2829
	skb_update_prio(skb);

2830
	txq = netdev_pick_tx(dev, skb, accel_priv);
2831
	q = rcu_dereference_bh(txq->qdisc);
2832

L
Linus Torvalds 已提交
2833
#ifdef CONFIG_NET_CLS_ACT
E
Eric Dumazet 已提交
2834
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS);
L
Linus Torvalds 已提交
2835
#endif
2836
	trace_net_dev_queue(skb);
L
Linus Torvalds 已提交
2837
	if (q->enqueue) {
2838
		rc = __dev_xmit_skb(skb, q, dev, txq);
2839
		goto out;
L
Linus Torvalds 已提交
2840 2841 2842 2843 2844
	}

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

H
Herbert Xu 已提交
2845 2846
	   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
2847 2848 2849 2850 2851 2852 2853 2854 2855 2856
	   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 */

2857
		if (txq->xmit_lock_owner != cpu) {
L
Linus Torvalds 已提交
2858

2859 2860 2861
			if (__this_cpu_read(xmit_recursion) > RECURSION_LIMIT)
				goto recursion_alert;

2862
			HARD_TX_LOCK(dev, txq, cpu);
L
Linus Torvalds 已提交
2863

2864
			if (!netif_xmit_stopped(txq)) {
2865
				__this_cpu_inc(xmit_recursion);
2866
				rc = dev_hard_start_xmit(skb, dev, txq);
2867
				__this_cpu_dec(xmit_recursion);
2868
				if (dev_xmit_complete(rc)) {
2869
					HARD_TX_UNLOCK(dev, txq);
L
Linus Torvalds 已提交
2870 2871 2872
					goto out;
				}
			}
2873
			HARD_TX_UNLOCK(dev, txq);
2874 2875
			net_crit_ratelimited("Virtual device %s asks to queue packet!\n",
					     dev->name);
L
Linus Torvalds 已提交
2876 2877
		} else {
			/* Recursion is detected! It is possible,
2878 2879 2880
			 * unfortunately
			 */
recursion_alert:
2881 2882
			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
					     dev->name);
L
Linus Torvalds 已提交
2883 2884 2885 2886
		}
	}

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

2889
	atomic_long_inc(&dev->tx_dropped);
L
Linus Torvalds 已提交
2890 2891 2892
	kfree_skb(skb);
	return rc;
out:
2893
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
2894 2895
	return rc;
}
2896 2897 2898 2899 2900

int dev_queue_xmit(struct sk_buff *skb)
{
	return __dev_queue_xmit(skb, NULL);
}
E
Eric Dumazet 已提交
2901
EXPORT_SYMBOL(dev_queue_xmit);
L
Linus Torvalds 已提交
2902

2903 2904 2905 2906 2907 2908
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
2909 2910 2911 2912 2913

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

2914
int netdev_max_backlog __read_mostly = 1000;
E
Eric Dumazet 已提交
2915 2916
EXPORT_SYMBOL(netdev_max_backlog);

E
Eric Dumazet 已提交
2917
int netdev_tstamp_prequeue __read_mostly = 1;
2918 2919
int netdev_budget __read_mostly = 300;
int weight_p __read_mostly = 64;            /* old backlog weight */
L
Linus Torvalds 已提交
2920

E
Eric Dumazet 已提交
2921 2922 2923 2924 2925 2926 2927 2928
/* 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);
}

2929 2930 2931
#ifdef CONFIG_RPS

/* One global table that all flow-based protocols share. */
E
Eric Dumazet 已提交
2932
struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
2933 2934
EXPORT_SYMBOL(rps_sock_flow_table);

2935
struct static_key rps_needed __read_mostly;
2936

2937 2938 2939 2940
static struct rps_dev_flow *
set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
	    struct rps_dev_flow *rflow, u16 next_cpu)
{
2941
	if (next_cpu != RPS_NO_CPU) {
2942 2943 2944 2945 2946 2947 2948 2949 2950
#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? */
2951 2952
		if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
		    !(dev->features & NETIF_F_NTUPLE))
2953 2954 2955 2956 2957 2958 2959 2960 2961
			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;
2962
		flow_id = skb_get_hash(skb) & flow_table->mask;
2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974
		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 =
2975
			per_cpu(softnet_data, next_cpu).input_queue_head;
2976 2977
	}

2978
	rflow->cpu = next_cpu;
2979 2980 2981
	return rflow;
}

2982 2983 2984 2985 2986 2987 2988 2989 2990
/*
 * 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
2991
	struct rps_map *map;
2992 2993 2994 2995
	struct rps_dev_flow_table *flow_table;
	struct rps_sock_flow_table *sock_flow_table;
	int cpu = -1;
	u16 tcpu;
2996
	u32 hash;
2997 2998 2999

	if (skb_rx_queue_recorded(skb)) {
		u16 index = skb_get_rx_queue(skb);
3000 3001 3002 3003 3004
		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);
3005 3006 3007 3008 3009 3010
			goto done;
		}
		rxqueue = dev->_rx + index;
	} else
		rxqueue = dev->_rx;

E
Eric Dumazet 已提交
3011 3012
	map = rcu_dereference(rxqueue->rps_map);
	if (map) {
3013
		if (map->len == 1 &&
3014
		    !rcu_access_pointer(rxqueue->rps_flow_table)) {
3015 3016 3017 3018 3019
			tcpu = map->cpus[0];
			if (cpu_online(tcpu))
				cpu = tcpu;
			goto done;
		}
3020
	} else if (!rcu_access_pointer(rxqueue->rps_flow_table)) {
3021
		goto done;
3022
	}
3023

3024
	skb_reset_network_header(skb);
3025 3026
	hash = skb_get_hash(skb);
	if (!hash)
3027 3028
		goto done;

T
Tom Herbert 已提交
3029 3030 3031 3032 3033 3034
	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;

3035
		rflow = &flow_table->flows[hash & flow_table->mask];
T
Tom Herbert 已提交
3036 3037
		tcpu = rflow->cpu;

3038
		next_cpu = sock_flow_table->ents[hash & sock_flow_table->mask];
T
Tom Herbert 已提交
3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053

		/*
		 * 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 -
3054 3055
		      rflow->last_qtail)) >= 0)) {
			tcpu = next_cpu;
3056
			rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
3057
		}
3058

T
Tom Herbert 已提交
3059 3060 3061 3062 3063 3064 3065
		if (tcpu != RPS_NO_CPU && cpu_online(tcpu)) {
			*rflowp = rflow;
			cpu = tcpu;
			goto done;
		}
	}

T
Tom Herbert 已提交
3066
	if (map) {
3067
		tcpu = map->cpus[((u64) hash * map->len) >> 32];
T
Tom Herbert 已提交
3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078

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

done:
	return cpu;
}

3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118
#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
3119
/* Called from hardirq (IPI) context */
E
Eric Dumazet 已提交
3120
static void rps_trigger_softirq(void *data)
T
Tom Herbert 已提交
3121
{
E
Eric Dumazet 已提交
3122 3123
	struct softnet_data *sd = data;

E
Eric Dumazet 已提交
3124
	____napi_schedule(sd, &sd->backlog);
C
Changli Gao 已提交
3125
	sd->received_rps++;
T
Tom Herbert 已提交
3126
}
E
Eric Dumazet 已提交
3127

T
Tom Herbert 已提交
3128
#endif /* CONFIG_RPS */
T
Tom Herbert 已提交
3129

E
Eric Dumazet 已提交
3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150
/*
 * 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;
}

3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169
#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) {
3170
		new_flow = skb_get_hash(skb) & (fl->num_buckets - 1);
3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190
		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
3191 3192 3193 3194
/*
 * 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
3195 3196
static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
			      unsigned int *qtail)
T
Tom Herbert 已提交
3197
{
E
Eric Dumazet 已提交
3198
	struct softnet_data *sd;
T
Tom Herbert 已提交
3199
	unsigned long flags;
3200
	unsigned int qlen;
T
Tom Herbert 已提交
3201

E
Eric Dumazet 已提交
3202
	sd = &per_cpu(softnet_data, cpu);
T
Tom Herbert 已提交
3203 3204 3205

	local_irq_save(flags);

E
Eric Dumazet 已提交
3206
	rps_lock(sd);
3207 3208
	qlen = skb_queue_len(&sd->input_pkt_queue);
	if (qlen <= netdev_max_backlog && !skb_flow_limit(skb, qlen)) {
3209
		if (skb_queue_len(&sd->input_pkt_queue)) {
T
Tom Herbert 已提交
3210
enqueue:
E
Eric Dumazet 已提交
3211
			__skb_queue_tail(&sd->input_pkt_queue, skb);
3212
			input_queue_tail_incr_save(sd, qtail);
E
Eric Dumazet 已提交
3213
			rps_unlock(sd);
3214
			local_irq_restore(flags);
T
Tom Herbert 已提交
3215 3216 3217
			return NET_RX_SUCCESS;
		}

3218 3219 3220 3221
		/* 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
3222
			if (!rps_ipi_queued(sd))
E
Eric Dumazet 已提交
3223
				____napi_schedule(sd, &sd->backlog);
T
Tom Herbert 已提交
3224 3225 3226 3227
		}
		goto enqueue;
	}

C
Changli Gao 已提交
3228
	sd->dropped++;
E
Eric Dumazet 已提交
3229
	rps_unlock(sd);
T
Tom Herbert 已提交
3230 3231 3232

	local_irq_restore(flags);

3233
	atomic_long_inc(&skb->dev->rx_dropped);
T
Tom Herbert 已提交
3234 3235 3236
	kfree_skb(skb);
	return NET_RX_DROP;
}
L
Linus Torvalds 已提交
3237

3238
static int netif_rx_internal(struct sk_buff *skb)
L
Linus Torvalds 已提交
3239
{
3240
	int ret;
L
Linus Torvalds 已提交
3241

3242
	net_timestamp_check(netdev_tstamp_prequeue, skb);
L
Linus Torvalds 已提交
3243

3244
	trace_netif_rx(skb);
E
Eric Dumazet 已提交
3245
#ifdef CONFIG_RPS
3246
	if (static_key_false(&rps_needed)) {
T
Tom Herbert 已提交
3247
		struct rps_dev_flow voidflow, *rflow = &voidflow;
3248 3249
		int cpu;

3250
		preempt_disable();
3251
		rcu_read_lock();
T
Tom Herbert 已提交
3252 3253

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
3254 3255
		if (cpu < 0)
			cpu = smp_processor_id();
T
Tom Herbert 已提交
3256 3257 3258

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

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

/**
 *	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
3292
EXPORT_SYMBOL(netif_rx);
L
Linus Torvalds 已提交
3293 3294 3295 3296 3297

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

3298 3299
	trace_netif_rx_ni_entry(skb);

L
Linus Torvalds 已提交
3300
	preempt_disable();
3301
	err = netif_rx_internal(skb);
L
Linus Torvalds 已提交
3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325
	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;

3326
			WARN_ON(atomic_read(&skb->users));
3327 3328 3329 3330
			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
3331 3332 3333 3334 3335
			__kfree_skb(skb);
		}
	}

	if (sd->output_queue) {
3336
		struct Qdisc *head;
L
Linus Torvalds 已提交
3337 3338 3339 3340

		local_irq_disable();
		head = sd->output_queue;
		sd->output_queue = NULL;
3341
		sd->output_queue_tailp = &sd->output_queue;
L
Linus Torvalds 已提交
3342 3343 3344
		local_irq_enable();

		while (head) {
3345 3346 3347
			struct Qdisc *q = head;
			spinlock_t *root_lock;

L
Linus Torvalds 已提交
3348 3349
			head = head->next_sched;

3350
			root_lock = qdisc_lock(q);
3351
			if (spin_trylock(root_lock)) {
3352 3353 3354
				smp_mb__before_clear_bit();
				clear_bit(__QDISC_STATE_SCHED,
					  &q->state);
3355 3356
				qdisc_run(q);
				spin_unlock(root_lock);
L
Linus Torvalds 已提交
3357
			} else {
3358
				if (!test_bit(__QDISC_STATE_DEACTIVATED,
3359
					      &q->state)) {
3360
					__netif_reschedule(q);
3361 3362 3363 3364 3365
				} else {
					smp_mb__before_clear_bit();
					clear_bit(__QDISC_STATE_SCHED,
						  &q->state);
				}
L
Linus Torvalds 已提交
3366 3367 3368 3369 3370
			}
		}
	}
}

3371 3372
#if (defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)) && \
    (defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE))
3373 3374 3375
/* This hook is defined here for ATM LANE */
int (*br_fdb_test_addr_hook)(struct net_device *dev,
			     unsigned char *addr) __read_mostly;
3376
EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
3377
#endif
L
Linus Torvalds 已提交
3378 3379 3380 3381 3382 3383

#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
3384 3385
 * 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
3386 3387
 *
 */
3388
static int ing_filter(struct sk_buff *skb, struct netdev_queue *rxq)
L
Linus Torvalds 已提交
3389 3390
{
	struct net_device *dev = skb->dev;
3391
	u32 ttl = G_TC_RTTL(skb->tc_verd);
3392 3393
	int result = TC_ACT_OK;
	struct Qdisc *q;
3394

3395
	if (unlikely(MAX_RED_LOOP < ttl++)) {
3396 3397
		net_warn_ratelimited("Redir loop detected Dropping packet (%d->%d)\n",
				     skb->skb_iif, dev->ifindex);
3398 3399
		return TC_ACT_SHOT;
	}
L
Linus Torvalds 已提交
3400

3401 3402
	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
3403

3404
	q = rxq->qdisc;
3405
	if (q != &noop_qdisc) {
3406
		spin_lock(qdisc_lock(q));
3407 3408
		if (likely(!test_bit(__QDISC_STATE_DEACTIVATED, &q->state)))
			result = qdisc_enqueue_root(skb, q);
3409 3410
		spin_unlock(qdisc_lock(q));
	}
3411 3412 3413

	return result;
}
3414

3415 3416 3417 3418
static inline struct sk_buff *handle_ing(struct sk_buff *skb,
					 struct packet_type **pt_prev,
					 int *ret, struct net_device *orig_dev)
{
3419 3420 3421
	struct netdev_queue *rxq = rcu_dereference(skb->dev->ingress_queue);

	if (!rxq || rxq->qdisc == &noop_qdisc)
3422
		goto out;
L
Linus Torvalds 已提交
3423

3424 3425 3426
	if (*pt_prev) {
		*ret = deliver_skb(skb, *pt_prev, orig_dev);
		*pt_prev = NULL;
L
Linus Torvalds 已提交
3427 3428
	}

3429
	switch (ing_filter(skb, rxq)) {
3430 3431 3432 3433 3434 3435 3436 3437 3438
	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
3439 3440 3441
}
#endif

3442 3443 3444 3445
/**
 *	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
3446
 *	@rx_handler_data: data pointer that is used by rx handler
3447
 *
3448
 *	Register a receive handler for a device. This handler will then be
3449 3450 3451 3452
 *	called from __netif_receive_skb. A negative errno code is returned
 *	on a failure.
 *
 *	The caller must hold the rtnl_mutex.
3453 3454
 *
 *	For a general description of rx_handler, see enum rx_handler_result.
3455 3456
 */
int netdev_rx_handler_register(struct net_device *dev,
J
Jiri Pirko 已提交
3457 3458
			       rx_handler_func_t *rx_handler,
			       void *rx_handler_data)
3459 3460 3461 3462 3463 3464
{
	ASSERT_RTNL();

	if (dev->rx_handler)
		return -EBUSY;

3465
	/* Note: rx_handler_data must be set before rx_handler */
J
Jiri Pirko 已提交
3466
	rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
3467 3468 3469 3470 3471 3472 3473 3474 3475 3476
	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
3477
 *	Unregister a receive handler from a device.
3478 3479 3480 3481 3482 3483 3484
 *
 *	The caller must hold the rtnl_mutex.
 */
void netdev_rx_handler_unregister(struct net_device *dev)
{

	ASSERT_RTNL();
3485
	RCU_INIT_POINTER(dev->rx_handler, NULL);
3486 3487 3488 3489 3490
	/* 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();
3491
	RCU_INIT_POINTER(dev->rx_handler_data, NULL);
3492 3493 3494
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);

3495 3496 3497 3498 3499 3500 3501
/*
 * 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) {
3502 3503 3504 3505 3506
	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):
3507 3508 3509 3510 3511 3512
		return true;
	default:
		return false;
	}
}

3513
static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc)
L
Linus Torvalds 已提交
3514 3515
{
	struct packet_type *ptype, *pt_prev;
3516
	rx_handler_func_t *rx_handler;
D
David S. Miller 已提交
3517
	struct net_device *orig_dev;
3518
	struct net_device *null_or_dev;
3519
	bool deliver_exact = false;
L
Linus Torvalds 已提交
3520
	int ret = NET_RX_DROP;
A
Al Viro 已提交
3521
	__be16 type;
L
Linus Torvalds 已提交
3522

3523
	net_timestamp_check(!netdev_tstamp_prequeue, skb);
3524

3525
	trace_netif_receive_skb(skb);
3526

J
Joe Eykholt 已提交
3527
	orig_dev = skb->dev;
3528

3529
	skb_reset_network_header(skb);
3530 3531
	if (!skb_transport_header_was_set(skb))
		skb_reset_transport_header(skb);
3532
	skb_reset_mac_len(skb);
L
Linus Torvalds 已提交
3533 3534 3535 3536 3537

	pt_prev = NULL;

	rcu_read_lock();

3538
another_round:
3539
	skb->skb_iif = skb->dev->ifindex;
3540 3541 3542

	__this_cpu_inc(softnet_data.processed);

P
Patrick McHardy 已提交
3543 3544
	if (skb->protocol == cpu_to_be16(ETH_P_8021Q) ||
	    skb->protocol == cpu_to_be16(ETH_P_8021AD)) {
3545 3546
		skb = vlan_untag(skb);
		if (unlikely(!skb))
3547
			goto unlock;
3548 3549
	}

L
Linus Torvalds 已提交
3550 3551 3552 3553 3554 3555 3556
#ifdef CONFIG_NET_CLS_ACT
	if (skb->tc_verd & TC_NCLS) {
		skb->tc_verd = CLR_TC_NCLS(skb->tc_verd);
		goto ncls;
	}
#endif

3557
	if (pfmemalloc)
3558 3559
		goto skip_taps;

L
Linus Torvalds 已提交
3560
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
3561
		if (!ptype->dev || ptype->dev == skb->dev) {
3562
			if (pt_prev)
D
David S. Miller 已提交
3563
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3564 3565 3566 3567
			pt_prev = ptype;
		}
	}

3568
skip_taps:
L
Linus Torvalds 已提交
3569
#ifdef CONFIG_NET_CLS_ACT
3570 3571
	skb = handle_ing(skb, &pt_prev, &ret, orig_dev);
	if (!skb)
3572
		goto unlock;
L
Linus Torvalds 已提交
3573 3574 3575
ncls:
#endif

3576
	if (pfmemalloc && !skb_pfmemalloc_protocol(skb))
3577 3578
		goto drop;

3579 3580 3581 3582 3583
	if (vlan_tx_tag_present(skb)) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3584
		if (vlan_do_receive(&skb))
3585 3586
			goto another_round;
		else if (unlikely(!skb))
3587
			goto unlock;
3588 3589
	}

3590
	rx_handler = rcu_dereference(skb->dev->rx_handler);
3591 3592 3593 3594 3595
	if (rx_handler) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3596 3597
		switch (rx_handler(&skb)) {
		case RX_HANDLER_CONSUMED:
3598
			ret = NET_RX_SUCCESS;
3599
			goto unlock;
3600
		case RX_HANDLER_ANOTHER:
3601
			goto another_round;
3602 3603 3604 3605 3606 3607 3608
		case RX_HANDLER_EXACT:
			deliver_exact = true;
		case RX_HANDLER_PASS:
			break;
		default:
			BUG();
		}
3609
	}
L
Linus Torvalds 已提交
3610

E
Eric Dumazet 已提交
3611 3612 3613 3614 3615 3616 3617 3618 3619
	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;
	}
3620

3621
	/* deliver only exact match when indicated */
3622
	null_or_dev = deliver_exact ? skb->dev : NULL;
3623

L
Linus Torvalds 已提交
3624
	type = skb->protocol;
3625 3626
	list_for_each_entry_rcu(ptype,
			&ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) {
3627
		if (ptype->type == type &&
3628 3629
		    (ptype->dev == null_or_dev || ptype->dev == skb->dev ||
		     ptype->dev == orig_dev)) {
3630
			if (pt_prev)
D
David S. Miller 已提交
3631
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3632 3633 3634 3635 3636
			pt_prev = ptype;
		}
	}

	if (pt_prev) {
3637
		if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
3638
			goto drop;
3639 3640
		else
			ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3641
	} else {
3642
drop:
3643
		atomic_long_inc(&skb->dev->rx_dropped);
L
Linus Torvalds 已提交
3644 3645 3646 3647 3648 3649 3650
		kfree_skb(skb);
		/* Jamal, now you will not able to escape explaining
		 * me how you were going to use this. :-)
		 */
		ret = NET_RX_DROP;
	}

3651
unlock:
L
Linus Torvalds 已提交
3652
	rcu_read_unlock();
3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677
	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
3678 3679
	return ret;
}
T
Tom Herbert 已提交
3680

3681
static int netif_receive_skb_internal(struct sk_buff *skb)
T
Tom Herbert 已提交
3682
{
3683
	net_timestamp_check(netdev_tstamp_prequeue, skb);
E
Eric Dumazet 已提交
3684

3685 3686 3687
	if (skb_defer_rx_timestamp(skb))
		return NET_RX_SUCCESS;

E
Eric Dumazet 已提交
3688
#ifdef CONFIG_RPS
3689
	if (static_key_false(&rps_needed)) {
E
Eric Dumazet 已提交
3690 3691
		struct rps_dev_flow voidflow, *rflow = &voidflow;
		int cpu, ret;
T
Tom Herbert 已提交
3692

E
Eric Dumazet 已提交
3693 3694 3695
		rcu_read_lock();

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
T
Tom Herbert 已提交
3696

E
Eric Dumazet 已提交
3697 3698 3699
		if (cpu >= 0) {
			ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
			rcu_read_unlock();
3700
			return ret;
E
Eric Dumazet 已提交
3701
		}
3702
		rcu_read_unlock();
T
Tom Herbert 已提交
3703
	}
3704
#endif
3705
	return __netif_receive_skb(skb);
T
Tom Herbert 已提交
3706
}
3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728

/**
 *	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
3729
EXPORT_SYMBOL(netif_receive_skb);
L
Linus Torvalds 已提交
3730

E
Eric Dumazet 已提交
3731 3732 3733
/* Network device is going away, flush any packets still pending
 * Called with irqs disabled.
 */
3734
static void flush_backlog(void *arg)
3735
{
3736
	struct net_device *dev = arg;
E
Eric Dumazet 已提交
3737
	struct softnet_data *sd = &__get_cpu_var(softnet_data);
3738 3739
	struct sk_buff *skb, *tmp;

E
Eric Dumazet 已提交
3740
	rps_lock(sd);
3741
	skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
3742
		if (skb->dev == dev) {
E
Eric Dumazet 已提交
3743
			__skb_unlink(skb, &sd->input_pkt_queue);
3744
			kfree_skb(skb);
3745
			input_queue_head_incr(sd);
3746
		}
3747
	}
E
Eric Dumazet 已提交
3748
	rps_unlock(sd);
3749 3750 3751 3752 3753

	skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
		if (skb->dev == dev) {
			__skb_unlink(skb, &sd->process_queue);
			kfree_skb(skb);
3754
			input_queue_head_incr(sd);
3755 3756
		}
	}
3757 3758
}

3759 3760
static int napi_gro_complete(struct sk_buff *skb)
{
3761
	struct packet_offload *ptype;
3762
	__be16 type = skb->protocol;
3763
	struct list_head *head = &offload_base;
3764 3765
	int err = -ENOENT;

3766 3767
	BUILD_BUG_ON(sizeof(struct napi_gro_cb) > sizeof(skb->cb));

3768 3769
	if (NAPI_GRO_CB(skb)->count == 1) {
		skb_shinfo(skb)->gso_size = 0;
3770
		goto out;
3771
	}
3772 3773 3774

	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
3775
		if (ptype->type != type || !ptype->callbacks.gro_complete)
3776 3777
			continue;

3778
		err = ptype->callbacks.gro_complete(skb, 0);
3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789
		break;
	}
	rcu_read_unlock();

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

out:
3790
	return netif_receive_skb_internal(skb);
3791 3792
}

3793 3794 3795 3796 3797
/* 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)
3798
{
3799
	struct sk_buff *skb, *prev = NULL;
3800

3801 3802 3803 3804 3805 3806 3807
	/* 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) {
3808
		skb->next = NULL;
3809 3810 3811 3812 3813

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

		prev = skb->prev;
3814
		napi_gro_complete(skb);
3815
		napi->gro_count--;
3816 3817 3818 3819
	}

	napi->gro_list = NULL;
}
E
Eric Dumazet 已提交
3820
EXPORT_SYMBOL(napi_gro_flush);
3821

E
Eric Dumazet 已提交
3822 3823 3824 3825
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;
3826
	u32 hash = skb_get_hash_raw(skb);
E
Eric Dumazet 已提交
3827 3828 3829 3830

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

3831 3832 3833 3834 3835 3836 3837
		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
3838 3839 3840 3841
		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),
3842
						      skb_mac_header(skb));
E
Eric Dumazet 已提交
3843 3844
		else if (!diffs)
			diffs = memcmp(skb_mac_header(p),
3845
				       skb_mac_header(skb),
E
Eric Dumazet 已提交
3846 3847 3848 3849 3850
				       maclen);
		NAPI_GRO_CB(p)->same_flow = !diffs;
	}
}

3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864
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
3865 3866 3867
	}
}

3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888
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
3889
static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3890 3891
{
	struct sk_buff **pp = NULL;
3892
	struct packet_offload *ptype;
3893
	__be16 type = skb->protocol;
3894
	struct list_head *head = &offload_base;
H
Herbert Xu 已提交
3895
	int same_flow;
3896
	enum gro_result ret;
3897
	int grow;
3898

3899
	if (!(skb->dev->features & NETIF_F_GRO))
3900 3901
		goto normal;

3902
	if (skb_is_gso(skb) || skb_has_frag_list(skb))
3903 3904
		goto normal;

E
Eric Dumazet 已提交
3905
	gro_list_prepare(napi, skb);
3906
	NAPI_GRO_CB(skb)->csum = skb->csum; /* Needed for CHECKSUM_COMPLETE */
E
Eric Dumazet 已提交
3907

3908 3909
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
3910
		if (ptype->type != type || !ptype->callbacks.gro_receive)
3911 3912
			continue;

3913
		skb_set_network_header(skb, skb_gro_offset(skb));
3914
		skb_reset_mac_len(skb);
3915 3916
		NAPI_GRO_CB(skb)->same_flow = 0;
		NAPI_GRO_CB(skb)->flush = 0;
H
Herbert Xu 已提交
3917
		NAPI_GRO_CB(skb)->free = 0;
3918
		NAPI_GRO_CB(skb)->udp_mark = 0;
3919

3920
		pp = ptype->callbacks.gro_receive(&napi->gro_list, skb);
3921 3922 3923 3924 3925 3926 3927
		break;
	}
	rcu_read_unlock();

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

H
Herbert Xu 已提交
3928
	same_flow = NAPI_GRO_CB(skb)->same_flow;
3929
	ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED;
H
Herbert Xu 已提交
3930

3931 3932 3933 3934 3935 3936
	if (pp) {
		struct sk_buff *nskb = *pp;

		*pp = nskb->next;
		nskb->next = NULL;
		napi_gro_complete(nskb);
3937
		napi->gro_count--;
3938 3939
	}

H
Herbert Xu 已提交
3940
	if (same_flow)
3941 3942
		goto ok;

3943
	if (NAPI_GRO_CB(skb)->flush)
3944 3945
		goto normal;

3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959
	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++;
	}
3960
	NAPI_GRO_CB(skb)->count = 1;
3961
	NAPI_GRO_CB(skb)->age = jiffies;
3962
	skb_shinfo(skb)->gso_size = skb_gro_len(skb);
3963 3964
	skb->next = napi->gro_list;
	napi->gro_list = skb;
3965
	ret = GRO_HELD;
3966

3967
pull:
3968 3969 3970
	grow = skb_gro_offset(skb) - skb_headlen(skb);
	if (grow > 0)
		gro_pull_from_frag0(skb, grow);
3971
ok:
3972
	return ret;
3973 3974

normal:
3975 3976
	ret = GRO_NORMAL;
	goto pull;
H
Herbert Xu 已提交
3977
}
3978

3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990
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;
}
3991
EXPORT_SYMBOL(gro_find_receive_by_type);
3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004

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;
}
4005
EXPORT_SYMBOL(gro_find_complete_by_type);
H
Herbert Xu 已提交
4006

R
Rami Rosen 已提交
4007
static gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
H
Herbert Xu 已提交
4008
{
4009 4010
	switch (ret) {
	case GRO_NORMAL:
4011
		if (netif_receive_skb_internal(skb))
4012 4013
			ret = GRO_DROP;
		break;
H
Herbert Xu 已提交
4014

4015
	case GRO_DROP:
H
Herbert Xu 已提交
4016 4017
		kfree_skb(skb);
		break;
4018

4019
	case GRO_MERGED_FREE:
4020 4021 4022 4023
		if (NAPI_GRO_CB(skb)->free == NAPI_GRO_FREE_STOLEN_HEAD)
			kmem_cache_free(skbuff_head_cache, skb);
		else
			__kfree_skb(skb);
4024 4025
		break;

4026 4027 4028
	case GRO_HELD:
	case GRO_MERGED:
		break;
H
Herbert Xu 已提交
4029 4030
	}

4031
	return ret;
4032 4033
}

4034
gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
4035
{
4036
	trace_napi_gro_receive_entry(skb);
4037

4038 4039
	skb_gro_reset_offset(skb);

E
Eric Dumazet 已提交
4040
	return napi_skb_finish(dev_gro_receive(napi, skb), skb);
4041 4042 4043
}
EXPORT_SYMBOL(napi_gro_receive);

S
stephen hemminger 已提交
4044
static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb)
4045 4046
{
	__skb_pull(skb, skb_headlen(skb));
4047 4048
	/* restore the reserve we had after netdev_alloc_skb_ip_align() */
	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN - skb_headroom(skb));
4049
	skb->vlan_tci = 0;
H
Herbert Xu 已提交
4050
	skb->dev = napi->dev;
A
Andy Gospodarek 已提交
4051
	skb->skb_iif = 0;
4052
	skb->truesize = SKB_TRUESIZE(skb_end_offset(skb));
4053 4054 4055 4056

	napi->skb = skb;
}

4057
struct sk_buff *napi_get_frags(struct napi_struct *napi)
H
Herbert Xu 已提交
4058 4059 4060 4061
{
	struct sk_buff *skb = napi->skb;

	if (!skb) {
4062
		skb = netdev_alloc_skb_ip_align(napi->dev, GRO_MAX_HEAD);
E
Eric Dumazet 已提交
4063
		napi->skb = skb;
4064
	}
4065 4066
	return skb;
}
4067
EXPORT_SYMBOL(napi_get_frags);
4068

4069 4070 4071
static gro_result_t napi_frags_finish(struct napi_struct *napi,
				      struct sk_buff *skb,
				      gro_result_t ret)
4072
{
4073 4074
	switch (ret) {
	case GRO_NORMAL:
4075 4076 4077 4078
	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))
4079
			ret = GRO_DROP;
4080
		break;
H
Herbert Xu 已提交
4081

4082 4083 4084 4085
	case GRO_DROP:
	case GRO_MERGED_FREE:
		napi_reuse_skb(napi, skb);
		break;
4086 4087 4088

	case GRO_MERGED:
		break;
4089
	}
H
Herbert Xu 已提交
4090

4091
	return ret;
H
Herbert Xu 已提交
4092
}
4093

4094 4095 4096 4097
/* 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
4098
static struct sk_buff *napi_frags_skb(struct napi_struct *napi)
4099 4100
{
	struct sk_buff *skb = napi->skb;
4101 4102
	const struct ethhdr *eth;
	unsigned int hlen = sizeof(*eth);
4103 4104 4105

	napi->skb = NULL;

4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119
	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;
4120
	}
4121 4122 4123 4124 4125 4126 4127 4128
	__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;
4129 4130 4131 4132

	return skb;
}

4133
gro_result_t napi_gro_frags(struct napi_struct *napi)
4134
{
4135
	struct sk_buff *skb = napi_frags_skb(napi);
4136 4137

	if (!skb)
4138
		return GRO_DROP;
4139

4140 4141
	trace_napi_gro_frags_entry(skb);

E
Eric Dumazet 已提交
4142
	return napi_frags_finish(napi, skb, dev_gro_receive(napi, skb));
4143
}
H
Herbert Xu 已提交
4144 4145
EXPORT_SYMBOL(napi_gro_frags);

4146
/*
4147
 * net_rps_action_and_irq_enable sends any pending IPI's for rps.
4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164
 * 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))
4165
				smp_call_function_single_async(remsd->cpu,
4166
							   &remsd->csd);
4167 4168 4169 4170 4171 4172 4173
			remsd = next;
		}
	} else
#endif
		local_irq_enable();
}

4174
static int process_backlog(struct napi_struct *napi, int quota)
L
Linus Torvalds 已提交
4175 4176
{
	int work = 0;
E
Eric Dumazet 已提交
4177
	struct softnet_data *sd = container_of(napi, struct softnet_data, backlog);
L
Linus Torvalds 已提交
4178

4179 4180 4181 4182 4183 4184 4185 4186 4187
#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
4188
	napi->weight = weight_p;
4189 4190
	local_irq_disable();
	while (work < quota) {
L
Linus Torvalds 已提交
4191
		struct sk_buff *skb;
4192 4193 4194 4195 4196 4197
		unsigned int qlen;

		while ((skb = __skb_dequeue(&sd->process_queue))) {
			local_irq_enable();
			__netif_receive_skb(skb);
			local_irq_disable();
4198 4199 4200 4201 4202
			input_queue_head_incr(sd);
			if (++work >= quota) {
				local_irq_enable();
				return work;
			}
4203
		}
L
Linus Torvalds 已提交
4204

E
Eric Dumazet 已提交
4205
		rps_lock(sd);
4206
		qlen = skb_queue_len(&sd->input_pkt_queue);
4207
		if (qlen)
4208 4209
			skb_queue_splice_tail_init(&sd->input_pkt_queue,
						   &sd->process_queue);
4210

4211
		if (qlen < quota - work) {
E
Eric Dumazet 已提交
4212 4213 4214 4215 4216 4217 4218 4219 4220 4221
			/*
			 * Inline a custom version of __napi_complete().
			 * only current cpu owns and manipulates this napi,
			 * and NAPI_STATE_SCHED is the only possible flag set on backlog.
			 * we can use a plain write instead of clear_bit(),
			 * and we dont need an smp_mb() memory barrier.
			 */
			list_del(&napi->poll_list);
			napi->state = 0;

4222
			quota = work + qlen;
4223
		}
E
Eric Dumazet 已提交
4224
		rps_unlock(sd);
4225 4226
	}
	local_irq_enable();
L
Linus Torvalds 已提交
4227

4228 4229
	return work;
}
L
Linus Torvalds 已提交
4230

4231 4232
/**
 * __napi_schedule - schedule for receive
4233
 * @n: entry to schedule
4234 4235 4236
 *
 * The entry's receive function will be scheduled to run
 */
H
Harvey Harrison 已提交
4237
void __napi_schedule(struct napi_struct *n)
4238 4239
{
	unsigned long flags;
L
Linus Torvalds 已提交
4240

4241
	local_irq_save(flags);
E
Eric Dumazet 已提交
4242
	____napi_schedule(&__get_cpu_var(softnet_data), n);
4243
	local_irq_restore(flags);
L
Linus Torvalds 已提交
4244
}
4245 4246
EXPORT_SYMBOL(__napi_schedule);

4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268
void __napi_complete(struct napi_struct *n)
{
	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
	BUG_ON(n->gro_list);

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

void napi_complete(struct napi_struct *n)
{
	unsigned long flags;

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

4269
	napi_gro_flush(n, false);
4270 4271 4272 4273 4274 4275
	local_irq_save(flags);
	__napi_complete(n);
	local_irq_restore(flags);
}
EXPORT_SYMBOL(napi_complete);

E
Eliezer Tamir 已提交
4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327
/* 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);

4328 4329 4330 4331
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);
4332
	napi->gro_count = 0;
4333
	napi->gro_list = NULL;
H
Herbert Xu 已提交
4334
	napi->skb = NULL;
4335
	napi->poll = poll;
E
Eric Dumazet 已提交
4336 4337 4338
	if (weight > NAPI_POLL_WEIGHT)
		pr_err_once("netif_napi_add() called with weight %d on device %s\n",
			    weight, dev->name);
4339 4340 4341
	napi->weight = weight;
	list_add(&napi->dev_list, &dev->napi_list);
	napi->dev = dev;
H
Herbert Xu 已提交
4342
#ifdef CONFIG_NETPOLL
4343 4344 4345 4346 4347 4348 4349 4350 4351
	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)
{
4352
	list_del_init(&napi->dev_list);
4353
	napi_free_frags(napi);
4354

E
Eric Dumazet 已提交
4355
	kfree_skb_list(napi->gro_list);
4356
	napi->gro_list = NULL;
4357
	napi->gro_count = 0;
4358 4359 4360
}
EXPORT_SYMBOL(netif_napi_del);

L
Linus Torvalds 已提交
4361 4362
static void net_rx_action(struct softirq_action *h)
{
4363
	struct softnet_data *sd = &__get_cpu_var(softnet_data);
4364
	unsigned long time_limit = jiffies + 2;
4365
	int budget = netdev_budget;
4366 4367
	void *have;

L
Linus Torvalds 已提交
4368 4369
	local_irq_disable();

4370
	while (!list_empty(&sd->poll_list)) {
4371 4372
		struct napi_struct *n;
		int work, weight;
L
Linus Torvalds 已提交
4373

4374
		/* If softirq window is exhuasted then punt.
4375 4376
		 * Allow this to run for 2 jiffies since which will allow
		 * an average latency of 1.5/HZ.
4377
		 */
4378
		if (unlikely(budget <= 0 || time_after_eq(jiffies, time_limit)))
L
Linus Torvalds 已提交
4379 4380 4381 4382
			goto softnet_break;

		local_irq_enable();

4383 4384 4385 4386 4387
		/* 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.
		 */
4388
		n = list_first_entry(&sd->poll_list, struct napi_struct, poll_list);
L
Linus Torvalds 已提交
4389

4390 4391 4392 4393
		have = netpoll_poll_lock(n);

		weight = n->weight;

4394 4395 4396 4397
		/* 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
4398
		 * accidentally calling ->poll() when NAPI is not scheduled.
4399 4400
		 */
		work = 0;
4401
		if (test_bit(NAPI_STATE_SCHED, &n->state)) {
4402
			work = n->poll(n, weight);
4403 4404
			trace_napi_poll(n);
		}
4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416

		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.
		 */
4417
		if (unlikely(work == weight)) {
4418 4419 4420 4421
			if (unlikely(napi_disable_pending(n))) {
				local_irq_enable();
				napi_complete(n);
				local_irq_disable();
4422 4423 4424 4425 4426 4427 4428 4429 4430
			} 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();
				}
4431
				list_move_tail(&n->poll_list, &sd->poll_list);
4432
			}
4433
		}
4434 4435

		netpoll_poll_unlock(have);
L
Linus Torvalds 已提交
4436 4437
	}
out:
4438
	net_rps_action_and_irq_enable(sd);
T
Tom Herbert 已提交
4439

4440 4441 4442 4443 4444
#ifdef CONFIG_NET_DMA
	/*
	 * There may not be any more sk_buffs coming right now, so push
	 * any pending DMA copies to hardware
	 */
4445
	dma_issue_pending_all();
4446
#endif
4447

L
Linus Torvalds 已提交
4448 4449 4450
	return;

softnet_break:
C
Changli Gao 已提交
4451
	sd->time_squeeze++;
L
Linus Torvalds 已提交
4452 4453 4454 4455
	__raise_softirq_irqoff(NET_RX_SOFTIRQ);
	goto out;
}

4456
struct netdev_adjacent {
J
Jiri Pirko 已提交
4457
	struct net_device *dev;
4458 4459

	/* upper master flag, there can only be one master device per list */
J
Jiri Pirko 已提交
4460
	bool master;
4461 4462 4463 4464

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

4465 4466 4467
	/* private field for the users */
	void *private;

J
Jiri Pirko 已提交
4468 4469 4470 4471
	struct list_head list;
	struct rcu_head rcu;
};

4472 4473
static struct netdev_adjacent *__netdev_find_adj(struct net_device *dev,
						 struct net_device *adj_dev,
4474
						 struct list_head *adj_list)
J
Jiri Pirko 已提交
4475
{
4476 4477
	struct netdev_adjacent *adj;

4478
	list_for_each_entry(adj, adj_list, list) {
4479 4480
		if (adj->dev == adj_dev)
			return adj;
J
Jiri Pirko 已提交
4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498
	}
	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();

4499
	return __netdev_find_adj(dev, upper_dev, &dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4500 4501 4502 4503 4504 4505 4506 4507 4508 4509
}
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
4510
static bool netdev_has_any_upper_dev(struct net_device *dev)
J
Jiri Pirko 已提交
4511 4512 4513
{
	ASSERT_RTNL();

4514
	return !list_empty(&dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525
}

/**
 * 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)
{
4526
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4527 4528 4529

	ASSERT_RTNL();

4530
	if (list_empty(&dev->adj_list.upper))
J
Jiri Pirko 已提交
4531 4532
		return NULL;

4533
	upper = list_first_entry(&dev->adj_list.upper,
4534
				 struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4535 4536 4537 4538 4539 4540
	if (likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get);

4541 4542 4543 4544 4545 4546 4547 4548 4549 4550
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);

4551 4552
/**
 * netdev_all_upper_get_next_dev_rcu - Get the next dev from upper list
4553 4554 4555 4556 4557 4558
 * @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.
 */
4559 4560
struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
						     struct list_head **iter)
4561 4562 4563
{
	struct netdev_adjacent *upper;

4564
	WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_rtnl_is_held());
4565 4566 4567

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

4568
	if (&upper->list == &dev->all_adj_list.upper)
4569 4570 4571 4572 4573 4574
		return NULL;

	*iter = &upper->list;

	return upper->dev;
}
4575
EXPORT_SYMBOL(netdev_all_upper_get_next_dev_rcu);
4576

4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597
/**
 * 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;

4598
	*iter = lower->list.next;
4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625

	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;

4626
	*iter = &lower->list;
4627 4628 4629 4630 4631

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

4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652
/**
 * 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
4653 4654 4655 4656 4657 4658 4659 4660 4661
/**
 * 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)
{
4662
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4663

4664
	upper = list_first_or_null_rcu(&dev->adj_list.upper,
4665
				       struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4666 4667 4668 4669 4670 4671
	if (upper && likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get_rcu);

4672
static int netdev_adjacent_sysfs_add(struct net_device *dev,
4673 4674 4675 4676 4677 4678 4679 4680 4681
			      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);
}
4682
static void netdev_adjacent_sysfs_del(struct net_device *dev,
4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695
			       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);
}

#define netdev_adjacent_is_neigh_list(dev, dev_list) \
		(dev_list == &dev->adj_list.upper || \
		 dev_list == &dev->adj_list.lower)

4696 4697
static int __netdev_adjacent_dev_insert(struct net_device *dev,
					struct net_device *adj_dev,
4698
					struct list_head *dev_list,
4699
					void *private, bool master)
4700 4701
{
	struct netdev_adjacent *adj;
4702
	int ret;
4703

4704
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717

	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;
4718
	adj->private = private;
4719
	dev_hold(adj_dev);
4720 4721 4722

	pr_debug("dev_hold for %s, because of link added from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
4723

4724 4725
	if (netdev_adjacent_is_neigh_list(dev, dev_list)) {
		ret = netdev_adjacent_sysfs_add(dev, adj_dev, dev_list);
4726 4727 4728 4729
		if (ret)
			goto free_adj;
	}

4730
	/* Ensure that master link is always the first item in list. */
4731 4732 4733 4734
	if (master) {
		ret = sysfs_create_link(&(dev->dev.kobj),
					&(adj_dev->dev.kobj), "master");
		if (ret)
4735
			goto remove_symlinks;
4736

4737
		list_add_rcu(&adj->list, dev_list);
4738
	} else {
4739
		list_add_tail_rcu(&adj->list, dev_list);
4740
	}
4741 4742

	return 0;
4743

4744
remove_symlinks:
4745 4746
	if (netdev_adjacent_is_neigh_list(dev, dev_list))
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
4747 4748
free_adj:
	kfree(adj);
4749
	dev_put(adj_dev);
4750 4751

	return ret;
4752 4753
}

S
stephen hemminger 已提交
4754 4755 4756
static void __netdev_adjacent_dev_remove(struct net_device *dev,
					 struct net_device *adj_dev,
					 struct list_head *dev_list)
4757 4758 4759
{
	struct netdev_adjacent *adj;

4760
	adj = __netdev_find_adj(dev, adj_dev, dev_list);
4761

4762 4763 4764
	if (!adj) {
		pr_err("tried to remove device %s from %s\n",
		       dev->name, adj_dev->name);
4765
		BUG();
4766
	}
4767 4768

	if (adj->ref_nr > 1) {
4769 4770
		pr_debug("%s to %s ref_nr-- = %d\n", dev->name, adj_dev->name,
			 adj->ref_nr-1);
4771 4772 4773 4774
		adj->ref_nr--;
		return;
	}

4775 4776 4777
	if (adj->master)
		sysfs_remove_link(&(dev->dev.kobj), "master");

4778 4779
	if (netdev_adjacent_is_neigh_list(dev, dev_list))
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
4780

4781
	list_del_rcu(&adj->list);
4782 4783
	pr_debug("dev_put for %s, because link removed from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
4784 4785 4786 4787
	dev_put(adj_dev);
	kfree_rcu(adj, rcu);
}

S
stephen hemminger 已提交
4788 4789 4790 4791 4792
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)
4793 4794 4795
{
	int ret;

4796 4797
	ret = __netdev_adjacent_dev_insert(dev, upper_dev, up_list, private,
					   master);
4798 4799 4800
	if (ret)
		return ret;

4801 4802
	ret = __netdev_adjacent_dev_insert(upper_dev, dev, down_list, private,
					   false);
4803
	if (ret) {
4804
		__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
4805 4806 4807 4808 4809 4810
		return ret;
	}

	return 0;
}

S
stephen hemminger 已提交
4811 4812
static int __netdev_adjacent_dev_link(struct net_device *dev,
				      struct net_device *upper_dev)
4813
{
4814 4815 4816
	return __netdev_adjacent_dev_link_lists(dev, upper_dev,
						&dev->all_adj_list.upper,
						&upper_dev->all_adj_list.lower,
4817
						NULL, false);
4818 4819
}

S
stephen hemminger 已提交
4820 4821 4822 4823
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)
4824
{
4825 4826
	__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
	__netdev_adjacent_dev_remove(upper_dev, dev, down_list);
4827 4828
}

S
stephen hemminger 已提交
4829 4830
static void __netdev_adjacent_dev_unlink(struct net_device *dev,
					 struct net_device *upper_dev)
4831
{
4832 4833 4834 4835 4836
	__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
4837 4838 4839
static int __netdev_adjacent_dev_link_neighbour(struct net_device *dev,
						struct net_device *upper_dev,
						void *private, bool master)
4840 4841 4842 4843 4844 4845 4846 4847 4848
{
	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,
4849
					       private, master);
4850 4851 4852 4853 4854 4855
	if (ret) {
		__netdev_adjacent_dev_unlink(dev, upper_dev);
		return ret;
	}

	return 0;
4856 4857
}

S
stephen hemminger 已提交
4858 4859
static void __netdev_adjacent_dev_unlink_neighbour(struct net_device *dev,
						   struct net_device *upper_dev)
4860 4861 4862 4863 4864 4865
{
	__netdev_adjacent_dev_unlink(dev, upper_dev);
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev,
					   &dev->adj_list.upper,
					   &upper_dev->adj_list.lower);
}
4866

J
Jiri Pirko 已提交
4867
static int __netdev_upper_dev_link(struct net_device *dev,
4868 4869
				   struct net_device *upper_dev, bool master,
				   void *private)
J
Jiri Pirko 已提交
4870
{
4871 4872
	struct netdev_adjacent *i, *j, *to_i, *to_j;
	int ret = 0;
J
Jiri Pirko 已提交
4873 4874 4875 4876 4877 4878 4879

	ASSERT_RTNL();

	if (dev == upper_dev)
		return -EBUSY;

	/* To prevent loops, check if dev is not upper device to upper_dev. */
4880
	if (__netdev_find_adj(upper_dev, dev, &upper_dev->all_adj_list.upper))
J
Jiri Pirko 已提交
4881 4882
		return -EBUSY;

4883
	if (__netdev_find_adj(dev, upper_dev, &dev->all_adj_list.upper))
J
Jiri Pirko 已提交
4884 4885 4886 4887 4888
		return -EEXIST;

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

4889 4890
	ret = __netdev_adjacent_dev_link_neighbour(dev, upper_dev, private,
						   master);
4891 4892
	if (ret)
		return ret;
J
Jiri Pirko 已提交
4893

4894
	/* Now that we linked these devs, make all the upper_dev's
4895
	 * all_adj_list.upper visible to every dev's all_adj_list.lower an
4896 4897 4898
	 * versa, and don't forget the devices itself. All of these
	 * links are non-neighbours.
	 */
4899 4900 4901 4902
	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);
4903 4904 4905 4906 4907 4908 4909
			ret = __netdev_adjacent_dev_link(i->dev, j->dev);
			if (ret)
				goto rollback_mesh;
		}
	}

	/* add dev to every upper_dev's upper device */
4910 4911 4912
	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);
4913 4914 4915 4916 4917 4918
		ret = __netdev_adjacent_dev_link(dev, i->dev);
		if (ret)
			goto rollback_upper_mesh;
	}

	/* add upper_dev to every dev's lower device */
4919 4920 4921
	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);
4922 4923 4924 4925
		ret = __netdev_adjacent_dev_link(i->dev, upper_dev);
		if (ret)
			goto rollback_lower_mesh;
	}
J
Jiri Pirko 已提交
4926

4927
	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
J
Jiri Pirko 已提交
4928
	return 0;
4929 4930 4931

rollback_lower_mesh:
	to_i = i;
4932
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
4933 4934 4935 4936 4937 4938 4939 4940 4941
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);
	}

	i = NULL;

rollback_upper_mesh:
	to_i = i;
4942
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list) {
4943 4944 4945 4946 4947 4948 4949 4950 4951 4952
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(dev, i->dev);
	}

	i = j = NULL;

rollback_mesh:
	to_i = i;
	to_j = j;
4953 4954
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list) {
4955 4956 4957 4958 4959 4960 4961 4962
			if (i == to_i && j == to_j)
				break;
			__netdev_adjacent_dev_unlink(i->dev, j->dev);
		}
		if (i == to_i)
			break;
	}

4963
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
4964 4965

	return ret;
J
Jiri Pirko 已提交
4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980
}

/**
 * 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)
{
4981
	return __netdev_upper_dev_link(dev, upper_dev, false, NULL);
J
Jiri Pirko 已提交
4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998
}
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)
{
4999
	return __netdev_upper_dev_link(dev, upper_dev, true, NULL);
J
Jiri Pirko 已提交
5000 5001 5002
}
EXPORT_SYMBOL(netdev_master_upper_dev_link);

5003 5004 5005 5006 5007 5008 5009 5010
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
5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021
/**
 * 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)
{
5022
	struct netdev_adjacent *i, *j;
J
Jiri Pirko 已提交
5023 5024
	ASSERT_RTNL();

5025
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
5026 5027 5028 5029 5030

	/* 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.
	 */
5031 5032
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list)
5033 5034 5035 5036 5037
			__netdev_adjacent_dev_unlink(i->dev, j->dev);

	/* remove also the devices itself from lower/upper device
	 * list
	 */
5038
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
5039 5040
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);

5041
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list)
5042 5043
		__netdev_adjacent_dev_unlink(dev, i->dev);

5044
	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
J
Jiri Pirko 已提交
5045 5046 5047
}
EXPORT_SYMBOL(netdev_upper_dev_unlink);

5048
void netdev_adjacent_rename_links(struct net_device *dev, char *oldname)
5049
{
5050
	struct netdev_adjacent *iter;
5051

5052 5053 5054 5055 5056 5057
	list_for_each_entry(iter, &dev->adj_list.upper, list) {
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.lower);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.lower);
	}
5058

5059 5060 5061 5062 5063 5064
	list_for_each_entry(iter, &dev->adj_list.lower, list) {
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.upper);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.upper);
	}
5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081
}

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);

5082 5083
static void dev_change_rx_flags(struct net_device *dev, int flags)
{
5084 5085
	const struct net_device_ops *ops = dev->netdev_ops;

5086
	if (ops->ndo_change_rx_flags)
5087
		ops->ndo_change_rx_flags(dev, flags);
5088 5089
}

5090
static int __dev_set_promiscuity(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5091
{
5092
	unsigned int old_flags = dev->flags;
5093 5094
	kuid_t uid;
	kgid_t gid;
L
Linus Torvalds 已提交
5095

5096 5097
	ASSERT_RTNL();

5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108
	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;
5109 5110
			pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n",
				dev->name);
5111 5112 5113
			return -EOVERFLOW;
		}
	}
5114
	if (dev->flags != old_flags) {
5115 5116 5117
		pr_info("device %s %s promiscuous mode\n",
			dev->name,
			dev->flags & IFF_PROMISC ? "entered" : "left");
5118 5119
		if (audit_enabled) {
			current_uid_gid(&uid, &gid);
5120 5121 5122 5123 5124
			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),
5125
				from_kuid(&init_user_ns, audit_get_loginuid(current)),
5126 5127
				from_kuid(&init_user_ns, uid),
				from_kgid(&init_user_ns, gid),
5128
				audit_get_sessionid(current));
5129
		}
5130

5131
		dev_change_rx_flags(dev, IFF_PROMISC);
L
Linus Torvalds 已提交
5132
	}
5133 5134
	if (notify)
		__dev_notify_flags(dev, old_flags, IFF_PROMISC);
5135
	return 0;
L
Linus Torvalds 已提交
5136 5137
}

5138 5139 5140 5141 5142 5143 5144 5145 5146
/**
 *	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.
5147
 *	Return 0 if successful or a negative errno code on error.
5148
 */
5149
int dev_set_promiscuity(struct net_device *dev, int inc)
5150
{
5151
	unsigned int old_flags = dev->flags;
5152
	int err;
5153

5154
	err = __dev_set_promiscuity(dev, inc, true);
5155
	if (err < 0)
5156
		return err;
5157 5158
	if (dev->flags != old_flags)
		dev_set_rx_mode(dev);
5159
	return err;
5160
}
E
Eric Dumazet 已提交
5161
EXPORT_SYMBOL(dev_set_promiscuity);
5162

5163
static int __dev_set_allmulti(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5164
{
5165
	unsigned int old_flags = dev->flags, old_gflags = dev->gflags;
L
Linus Torvalds 已提交
5166

5167 5168
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5169
	dev->flags |= IFF_ALLMULTI;
5170 5171 5172 5173 5174 5175 5176 5177 5178 5179
	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;
5180 5181
			pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n",
				dev->name);
5182 5183 5184
			return -EOVERFLOW;
		}
	}
5185
	if (dev->flags ^ old_flags) {
5186
		dev_change_rx_flags(dev, IFF_ALLMULTI);
5187
		dev_set_rx_mode(dev);
5188 5189 5190
		if (notify)
			__dev_notify_flags(dev, old_flags,
					   dev->gflags ^ old_gflags);
5191
	}
5192
	return 0;
5193
}
5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211

/**
 *	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
5212
EXPORT_SYMBOL(dev_set_allmulti);
5213 5214 5215 5216

/*
 *	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
5217
 *	filtering it is put in promiscuous mode while unicast addresses
5218 5219 5220 5221
 *	are present.
 */
void __dev_set_rx_mode(struct net_device *dev)
{
5222 5223
	const struct net_device_ops *ops = dev->netdev_ops;

5224 5225 5226 5227 5228
	/* dev_open will call this function so the list will stay sane. */
	if (!(dev->flags&IFF_UP))
		return;

	if (!netif_device_present(dev))
5229
		return;
5230

5231
	if (!(dev->priv_flags & IFF_UNICAST_FLT)) {
5232 5233 5234
		/* Unicast addresses changes may only happen under the rtnl,
		 * therefore calling __dev_set_promiscuity here is safe.
		 */
5235
		if (!netdev_uc_empty(dev) && !dev->uc_promisc) {
5236
			__dev_set_promiscuity(dev, 1, false);
5237
			dev->uc_promisc = true;
5238
		} else if (netdev_uc_empty(dev) && dev->uc_promisc) {
5239
			__dev_set_promiscuity(dev, -1, false);
5240
			dev->uc_promisc = false;
5241 5242
		}
	}
5243 5244 5245

	if (ops->ndo_set_rx_mode)
		ops->ndo_set_rx_mode(dev);
5246
}
5247
EXPORT_SYMBOL(__dev_set_rx_mode);
5248 5249 5250

void dev_set_rx_mode(struct net_device *dev)
{
5251
	netif_addr_lock_bh(dev);
5252
	__dev_set_rx_mode(dev);
5253
	netif_addr_unlock_bh(dev);
L
Linus Torvalds 已提交
5254 5255
}

5256 5257 5258 5259 5260 5261
/**
 *	dev_get_flags - get flags reported to userspace
 *	@dev: device
 *
 *	Get the combination of flag bits exported through APIs to userspace.
 */
5262
unsigned int dev_get_flags(const struct net_device *dev)
L
Linus Torvalds 已提交
5263
{
5264
	unsigned int flags;
L
Linus Torvalds 已提交
5265 5266 5267

	flags = (dev->flags & ~(IFF_PROMISC |
				IFF_ALLMULTI |
S
Stefan Rompf 已提交
5268 5269 5270
				IFF_RUNNING |
				IFF_LOWER_UP |
				IFF_DORMANT)) |
L
Linus Torvalds 已提交
5271 5272 5273
		(dev->gflags & (IFF_PROMISC |
				IFF_ALLMULTI));

S
Stefan Rompf 已提交
5274 5275 5276 5277 5278 5279 5280 5281
	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
5282 5283 5284

	return flags;
}
E
Eric Dumazet 已提交
5285
EXPORT_SYMBOL(dev_get_flags);
L
Linus Torvalds 已提交
5286

5287
int __dev_change_flags(struct net_device *dev, unsigned int flags)
L
Linus Torvalds 已提交
5288
{
5289
	unsigned int old_flags = dev->flags;
5290
	int ret;
L
Linus Torvalds 已提交
5291

5292 5293
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307
	/*
	 *	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.
	 */

5308 5309
	if ((old_flags ^ flags) & IFF_MULTICAST)
		dev_change_rx_flags(dev, IFF_MULTICAST);
5310

5311
	dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5312 5313 5314 5315 5316 5317 5318 5319 5320

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

	ret = 0;
	if ((old_flags ^ flags) & IFF_UP) {	/* Bit is different  ? */
5321
		ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev);
L
Linus Torvalds 已提交
5322 5323

		if (!ret)
5324
			dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5325 5326 5327
	}

	if ((flags ^ dev->gflags) & IFF_PROMISC) {
E
Eric Dumazet 已提交
5328
		int inc = (flags & IFF_PROMISC) ? 1 : -1;
5329
		unsigned int old_flags = dev->flags;
E
Eric Dumazet 已提交
5330

L
Linus Torvalds 已提交
5331
		dev->gflags ^= IFF_PROMISC;
5332 5333 5334 5335

		if (__dev_set_promiscuity(dev, inc, false) >= 0)
			if (dev->flags != old_flags)
				dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5336 5337 5338 5339 5340 5341 5342
	}

	/* 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
5343 5344
		int inc = (flags & IFF_ALLMULTI) ? 1 : -1;

L
Linus Torvalds 已提交
5345
		dev->gflags ^= IFF_ALLMULTI;
5346
		__dev_set_allmulti(dev, inc, false);
L
Linus Torvalds 已提交
5347 5348
	}

5349 5350 5351
	return ret;
}

5352 5353
void __dev_notify_flags(struct net_device *dev, unsigned int old_flags,
			unsigned int gchanges)
5354 5355 5356
{
	unsigned int changes = dev->flags ^ old_flags;

5357
	if (gchanges)
5358
		rtmsg_ifinfo(RTM_NEWLINK, dev, gchanges, GFP_ATOMIC);
5359

5360 5361 5362 5363 5364 5365 5366 5367
	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 &&
5368 5369 5370 5371 5372 5373 5374
	    (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);
	}
5375 5376 5377 5378 5379 5380 5381 5382 5383 5384
}

/**
 *	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.
 */
5385
int dev_change_flags(struct net_device *dev, unsigned int flags)
5386
{
5387
	int ret;
5388
	unsigned int changes, old_flags = dev->flags, old_gflags = dev->gflags;
5389 5390 5391 5392 5393

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

5394
	changes = (old_flags ^ dev->flags) | (old_gflags ^ dev->gflags);
5395
	__dev_notify_flags(dev, old_flags, changes);
L
Linus Torvalds 已提交
5396 5397
	return ret;
}
E
Eric Dumazet 已提交
5398
EXPORT_SYMBOL(dev_change_flags);
L
Linus Torvalds 已提交
5399

5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410
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;
}

5411 5412 5413 5414 5415 5416 5417
/**
 *	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
5418 5419
int dev_set_mtu(struct net_device *dev, int new_mtu)
{
5420
	int err, orig_mtu;
L
Linus Torvalds 已提交
5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431

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

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

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

5432 5433 5434 5435
	err = call_netdevice_notifiers(NETDEV_PRECHANGEMTU, dev);
	err = notifier_to_errno(err);
	if (err)
		return err;
5436

5437 5438
	orig_mtu = dev->mtu;
	err = __dev_set_mtu(dev, new_mtu);
5439

5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450
	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
5451 5452
	return err;
}
E
Eric Dumazet 已提交
5453
EXPORT_SYMBOL(dev_set_mtu);
L
Linus Torvalds 已提交
5454

5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465
/**
 *	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);

5466 5467 5468 5469 5470 5471 5472
/**
 *	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
5473 5474
int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa)
{
5475
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
5476 5477
	int err;

5478
	if (!ops->ndo_set_mac_address)
L
Linus Torvalds 已提交
5479 5480 5481 5482 5483
		return -EOPNOTSUPP;
	if (sa->sa_family != dev->type)
		return -EINVAL;
	if (!netif_device_present(dev))
		return -ENODEV;
5484
	err = ops->ndo_set_mac_address(dev, sa);
5485 5486
	if (err)
		return err;
J
Jiri Pirko 已提交
5487
	dev->addr_assign_type = NET_ADDR_SET;
5488
	call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
5489
	add_device_randomness(dev->dev_addr, dev->addr_len);
5490
	return 0;
L
Linus Torvalds 已提交
5491
}
E
Eric Dumazet 已提交
5492
EXPORT_SYMBOL(dev_set_mac_address);
L
Linus Torvalds 已提交
5493

J
Jiri Pirko 已提交
5494 5495 5496
/**
 *	dev_change_carrier - Change device carrier
 *	@dev: device
5497
 *	@new_carrier: new value
J
Jiri Pirko 已提交
5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512
 *
 *	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);

5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530
/**
 *	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
5531 5532
/**
 *	dev_new_index	-	allocate an ifindex
5533
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
5534 5535 5536 5537 5538
 *
 *	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.
 */
5539
static int dev_new_index(struct net *net)
L
Linus Torvalds 已提交
5540
{
5541
	int ifindex = net->ifindex;
L
Linus Torvalds 已提交
5542 5543 5544
	for (;;) {
		if (++ifindex <= 0)
			ifindex = 1;
5545
		if (!__dev_get_by_index(net, ifindex))
5546
			return net->ifindex = ifindex;
L
Linus Torvalds 已提交
5547 5548 5549 5550
	}
}

/* Delayed registration/unregisteration */
5551
static LIST_HEAD(net_todo_list);
5552
static DECLARE_WAIT_QUEUE_HEAD(netdev_unregistering_wq);
L
Linus Torvalds 已提交
5553

5554
static void net_set_todo(struct net_device *dev)
L
Linus Torvalds 已提交
5555 5556
{
	list_add_tail(&dev->todo_list, &net_todo_list);
5557
	dev_net(dev)->dev_unreg_count++;
L
Linus Torvalds 已提交
5558 5559
}

5560
static void rollback_registered_many(struct list_head *head)
5561
{
5562
	struct net_device *dev, *tmp;
5563
	LIST_HEAD(close_head);
5564

5565 5566 5567
	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

5568
	list_for_each_entry_safe(dev, tmp, head, unreg_list) {
5569
		/* Some devices call without registering
5570 5571
		 * for initialization unwind. Remove those
		 * devices and proceed with the remaining.
5572 5573
		 */
		if (dev->reg_state == NETREG_UNINITIALIZED) {
5574 5575
			pr_debug("unregister_netdevice: device %s/%p never was registered\n",
				 dev->name, dev);
5576

5577
			WARN_ON(1);
5578 5579
			list_del(&dev->unreg_list);
			continue;
5580
		}
5581
		dev->dismantle = true;
5582
		BUG_ON(dev->reg_state != NETREG_REGISTERED);
5583
	}
5584

5585
	/* If device is running, close it first. */
5586 5587 5588
	list_for_each_entry(dev, head, unreg_list)
		list_add_tail(&dev->close_list, &close_head);
	dev_close_many(&close_head);
5589

5590
	list_for_each_entry(dev, head, unreg_list) {
5591 5592
		/* And unlink it from device chain. */
		unlist_netdevice(dev);
5593

5594 5595
		dev->reg_state = NETREG_UNREGISTERING;
	}
5596 5597 5598

	synchronize_net();

5599 5600 5601
	list_for_each_entry(dev, head, unreg_list) {
		/* Shutdown queueing discipline. */
		dev_shutdown(dev);
5602 5603


5604 5605 5606 5607
		/* Notify protocols, that we are about to destroy
		   this device. They should clean all the things.
		*/
		call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
5608

5609 5610
		if (!dev->rtnl_link_ops ||
		    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
5611
			rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);
5612

5613 5614 5615
		/*
		 *	Flush the unicast and multicast chains
		 */
5616
		dev_uc_flush(dev);
5617
		dev_mc_flush(dev);
5618

5619 5620
		if (dev->netdev_ops->ndo_uninit)
			dev->netdev_ops->ndo_uninit(dev);
5621

J
Jiri Pirko 已提交
5622 5623
		/* Notifier chain MUST detach us all upper devices. */
		WARN_ON(netdev_has_any_upper_dev(dev));
5624

5625 5626
		/* Remove entries from kobject tree */
		netdev_unregister_kobject(dev);
5627 5628 5629 5630
#ifdef CONFIG_XPS
		/* Remove XPS queueing entries */
		netif_reset_xps_queues_gt(dev, 0);
#endif
5631
	}
5632

5633
	synchronize_net();
5634

5635
	list_for_each_entry(dev, head, unreg_list)
5636 5637 5638 5639 5640 5641 5642 5643 5644
		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
5645
	list_del(&single);
5646 5647
}

5648 5649
static netdev_features_t netdev_fix_features(struct net_device *dev,
	netdev_features_t features)
5650
{
5651 5652 5653
	/* Fix illegal checksum combinations */
	if ((features & NETIF_F_HW_CSUM) &&
	    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
5654
		netdev_warn(dev, "mixed HW and IP checksum settings.\n");
5655 5656 5657
		features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
	}

5658
	/* TSO requires that SG is present as well. */
5659
	if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
5660
		netdev_dbg(dev, "Dropping TSO features since no SG feature.\n");
5661
		features &= ~NETIF_F_ALL_TSO;
5662 5663
	}

5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676
	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;
	}

5677 5678 5679 5680
	/* TSO ECN requires that TSO is present as well. */
	if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN)
		features &= ~NETIF_F_TSO_ECN;

5681 5682
	/* Software GSO depends on SG. */
	if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) {
5683
		netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n");
5684 5685 5686
		features &= ~NETIF_F_GSO;
	}

5687
	/* UFO needs SG and checksumming */
5688
	if (features & NETIF_F_UFO) {
5689 5690 5691 5692
		/* 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))) {
5693
			netdev_dbg(dev,
5694
				"Dropping NETIF_F_UFO since no checksum offload features.\n");
5695 5696 5697 5698
			features &= ~NETIF_F_UFO;
		}

		if (!(features & NETIF_F_SG)) {
5699
			netdev_dbg(dev,
5700
				"Dropping NETIF_F_UFO since no NETIF_F_SG feature.\n");
5701 5702 5703 5704
			features &= ~NETIF_F_UFO;
		}
	}

J
Jiri Pirko 已提交
5705 5706 5707 5708 5709 5710 5711
#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;

5712 5713 5714
	return features;
}

5715
int __netdev_update_features(struct net_device *dev)
5716
{
5717
	netdev_features_t features;
5718 5719
	int err = 0;

5720 5721
	ASSERT_RTNL();

5722 5723 5724 5725 5726 5727 5728 5729 5730
	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)
5731
		return 0;
5732

5733 5734
	netdev_dbg(dev, "Features changed: %pNF -> %pNF\n",
		&dev->features, &features);
5735 5736 5737 5738

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

5739
	if (unlikely(err < 0)) {
5740
		netdev_err(dev,
5741 5742
			"set_features() failed (%d); wanted %pNF, left %pNF\n",
			err, &features, &dev->features);
5743 5744 5745 5746 5747 5748 5749 5750 5751
		return -1;
	}

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

	return 1;
}

5752 5753 5754 5755 5756 5757 5758 5759
/**
 *	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.
 */
5760 5761 5762 5763
void netdev_update_features(struct net_device *dev)
{
	if (__netdev_update_features(dev))
		netdev_features_change(dev);
5764 5765 5766
}
EXPORT_SYMBOL(netdev_update_features);

5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780 5781 5782 5783
/**
 *	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);

5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803 5804 5805 5806 5807 5808 5809 5810
/**
 *	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);

5811
#ifdef CONFIG_SYSFS
5812 5813 5814
static int netif_alloc_rx_queues(struct net_device *dev)
{
	unsigned int i, count = dev->num_rx_queues;
T
Tom Herbert 已提交
5815
	struct netdev_rx_queue *rx;
5816

T
Tom Herbert 已提交
5817
	BUG_ON(count < 1);
5818

T
Tom Herbert 已提交
5819
	rx = kcalloc(count, sizeof(struct netdev_rx_queue), GFP_KERNEL);
5820
	if (!rx)
T
Tom Herbert 已提交
5821
		return -ENOMEM;
5822

T
Tom Herbert 已提交
5823 5824 5825
	dev->_rx = rx;

	for (i = 0; i < count; i++)
T
Tom Herbert 已提交
5826
		rx[i].dev = dev;
5827 5828
	return 0;
}
T
Tom Herbert 已提交
5829
#endif
5830

C
Changli Gao 已提交
5831 5832 5833 5834 5835 5836 5837
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;
5838
	netdev_queue_numa_node_write(queue, NUMA_NO_NODE);
C
Changli Gao 已提交
5839
	queue->dev = dev;
T
Tom Herbert 已提交
5840 5841 5842
#ifdef CONFIG_BQL
	dql_init(&queue->dql, HZ);
#endif
C
Changli Gao 已提交
5843 5844
}

5845 5846 5847 5848 5849 5850 5851 5852
static void netif_free_tx_queues(struct net_device *dev)
{
	if (is_vmalloc_addr(dev->_tx))
		vfree(dev->_tx);
	else
		kfree(dev->_tx);
}

5853 5854 5855 5856
static int netif_alloc_netdev_queues(struct net_device *dev)
{
	unsigned int count = dev->num_tx_queues;
	struct netdev_queue *tx;
5857
	size_t sz = count * sizeof(*tx);
5858

5859
	BUG_ON(count < 1 || count > 0xffff);
5860

5861 5862 5863 5864 5865 5866
	tx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!tx) {
		tx = vzalloc(sz);
		if (!tx)
			return -ENOMEM;
	}
5867
	dev->_tx = tx;
T
Tom Herbert 已提交
5868

5869 5870
	netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
	spin_lock_init(&dev->tx_global_lock);
C
Changli Gao 已提交
5871 5872

	return 0;
5873 5874
}

L
Linus Torvalds 已提交
5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894
/**
 *	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;
5895
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
5896 5897 5898 5899

	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

5900 5901
	might_sleep();

L
Linus Torvalds 已提交
5902 5903
	/* When net_device's are persistent, this will be fatal. */
	BUG_ON(dev->reg_state != NETREG_UNINITIALIZED);
5904
	BUG_ON(!net);
L
Linus Torvalds 已提交
5905

5906
	spin_lock_init(&dev->addr_list_lock);
5907
	netdev_set_addr_lockdep_class(dev);
L
Linus Torvalds 已提交
5908 5909 5910

	dev->iflink = -1;

5911
	ret = dev_get_valid_name(net, dev, dev->name);
5912 5913 5914
	if (ret < 0)
		goto out;

L
Linus Torvalds 已提交
5915
	/* Init, if this function is available */
5916 5917
	if (dev->netdev_ops->ndo_init) {
		ret = dev->netdev_ops->ndo_init(dev);
L
Linus Torvalds 已提交
5918 5919 5920
		if (ret) {
			if (ret > 0)
				ret = -EIO;
5921
			goto out;
L
Linus Torvalds 已提交
5922 5923
		}
	}
5924

5925 5926
	if (((dev->hw_features | dev->features) &
	     NETIF_F_HW_VLAN_CTAG_FILTER) &&
5927 5928 5929 5930 5931 5932 5933
	    (!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;
	}

5934 5935 5936 5937 5938 5939
	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
5940 5941 5942
	if (dev->iflink == -1)
		dev->iflink = dev->ifindex;

5943 5944 5945 5946
	/* Transfer changeable features to wanted_features and enable
	 * software offloads (GSO and GRO).
	 */
	dev->hw_features |= NETIF_F_SOFT_FEATURES;
5947 5948
	dev->features |= NETIF_F_SOFT_FEATURES;
	dev->wanted_features = dev->features & dev->hw_features;
L
Linus Torvalds 已提交
5949

5950 5951
	if (!(dev->flags & IFF_LOOPBACK)) {
		dev->hw_features |= NETIF_F_NOCACHE_COPY;
5952 5953
	}

5954
	/* Make NETIF_F_HIGHDMA inheritable to VLAN devices.
5955
	 */
5956
	dev->vlan_features |= NETIF_F_HIGHDMA;
5957

5958 5959 5960 5961
	/* Make NETIF_F_SG inheritable to tunnel devices.
	 */
	dev->hw_enc_features |= NETIF_F_SG;

S
Simon Horman 已提交
5962 5963 5964 5965
	/* Make NETIF_F_SG inheritable to MPLS.
	 */
	dev->mpls_features |= NETIF_F_SG;

5966 5967 5968 5969 5970
	ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		goto err_uninit;

5971
	ret = netdev_register_kobject(dev);
5972
	if (ret)
5973
		goto err_uninit;
5974 5975
	dev->reg_state = NETREG_REGISTERED;

5976
	__netdev_update_features(dev);
5977

L
Linus Torvalds 已提交
5978 5979 5980 5981 5982 5983 5984
	/*
	 *	Default initial state at registry is that the
	 *	device is present.
	 */

	set_bit(__LINK_STATE_PRESENT, &dev->state);

5985 5986
	linkwatch_init_dev(dev);

L
Linus Torvalds 已提交
5987 5988
	dev_init_scheduler(dev);
	dev_hold(dev);
5989
	list_netdevice(dev);
5990
	add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
5991

5992 5993 5994 5995 5996 5997 5998
	/* 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
5999
	/* Notify protocols, that a new device appeared. */
6000
	ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
6001
	ret = notifier_to_errno(ret);
6002 6003 6004 6005
	if (ret) {
		rollback_registered(dev);
		dev->reg_state = NETREG_UNREGISTERED;
	}
6006 6007 6008 6009
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
6010 6011
	if (!dev->rtnl_link_ops ||
	    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
6012
		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
L
Linus Torvalds 已提交
6013 6014 6015

out:
	return ret;
6016 6017

err_uninit:
6018 6019
	if (dev->netdev_ops->ndo_uninit)
		dev->netdev_ops->ndo_uninit(dev);
6020
	goto out;
L
Linus Torvalds 已提交
6021
}
E
Eric Dumazet 已提交
6022
EXPORT_SYMBOL(register_netdevice);
L
Linus Torvalds 已提交
6023

6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054
/**
 *	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
6055 6056 6057 6058 6059
	/* Note : We dont allocate pcpu_refcnt for dummy devices,
	 * because users of this 'device' dont need to change
	 * its refcount.
	 */

6060 6061 6062 6063 6064
	return 0;
}
EXPORT_SYMBOL_GPL(init_dummy_netdev);


L
Linus Torvalds 已提交
6065 6066 6067 6068 6069 6070 6071 6072 6073
/**
 *	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.
 *
6074
 *	This is a wrapper around register_netdevice that takes the rtnl semaphore
L
Linus Torvalds 已提交
6075 6076 6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087 6088
 *	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
6089 6090 6091 6092 6093 6094 6095 6096 6097 6098
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);

6099
/**
L
Linus Torvalds 已提交
6100
 * netdev_wait_allrefs - wait until all references are gone.
6101
 * @dev: target net_device
L
Linus Torvalds 已提交
6102 6103 6104 6105 6106 6107 6108
 *
 * 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
6109
 * call dev_put.
L
Linus Torvalds 已提交
6110 6111 6112 6113
 */
static void netdev_wait_allrefs(struct net_device *dev)
{
	unsigned long rebroadcast_time, warning_time;
E
Eric Dumazet 已提交
6114
	int refcnt;
L
Linus Torvalds 已提交
6115

6116 6117
	linkwatch_forget_dev(dev);

L
Linus Torvalds 已提交
6118
	rebroadcast_time = warning_time = jiffies;
E
Eric Dumazet 已提交
6119 6120 6121
	refcnt = netdev_refcnt_read(dev);

	while (refcnt != 0) {
L
Linus Torvalds 已提交
6122
		if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
6123
			rtnl_lock();
L
Linus Torvalds 已提交
6124 6125

			/* Rebroadcast unregister notification */
6126
			call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
L
Linus Torvalds 已提交
6127

6128
			__rtnl_unlock();
6129
			rcu_barrier();
6130 6131
			rtnl_lock();

6132
			call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
L
Linus Torvalds 已提交
6133 6134 6135 6136 6137 6138 6139 6140 6141 6142 6143
			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();
			}

6144
			__rtnl_unlock();
L
Linus Torvalds 已提交
6145 6146 6147 6148 6149 6150

			rebroadcast_time = jiffies;
		}

		msleep(250);

E
Eric Dumazet 已提交
6151 6152
		refcnt = netdev_refcnt_read(dev);

L
Linus Torvalds 已提交
6153
		if (time_after(jiffies, warning_time + 10 * HZ)) {
6154 6155
			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
				 dev->name, refcnt);
L
Linus Torvalds 已提交
6156 6157 6158 6159 6160 6161 6162 6163 6164 6165 6166 6167 6168 6169 6170 6171 6172 6173 6174
			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
6175
 * We are invoked by rtnl_unlock().
L
Linus Torvalds 已提交
6176
 * This allows us to deal with problems:
6177
 * 1) We can delete sysfs objects which invoke hotplug
L
Linus Torvalds 已提交
6178 6179 6180
 *    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
6181 6182 6183
 *
 * 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
6184 6185 6186
 */
void netdev_run_todo(void)
{
6187
	struct list_head list;
L
Linus Torvalds 已提交
6188 6189

	/* Snapshot list, allow later requests */
6190
	list_replace_init(&net_todo_list, &list);
H
Herbert Xu 已提交
6191 6192

	__rtnl_unlock();
6193

6194 6195

	/* Wait for rcu callbacks to finish before next phase */
6196 6197 6198
	if (!list_empty(&list))
		rcu_barrier();

L
Linus Torvalds 已提交
6199 6200
	while (!list_empty(&list)) {
		struct net_device *dev
6201
			= list_first_entry(&list, struct net_device, todo_list);
L
Linus Torvalds 已提交
6202 6203
		list_del(&dev->todo_list);

6204
		rtnl_lock();
6205
		call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6206
		__rtnl_unlock();
6207

6208
		if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
6209
			pr_err("network todo '%s' but state %d\n",
6210 6211 6212 6213
			       dev->name, dev->reg_state);
			dump_stack();
			continue;
		}
L
Linus Torvalds 已提交
6214

6215
		dev->reg_state = NETREG_UNREGISTERED;
L
Linus Torvalds 已提交
6216

6217
		on_each_cpu(flush_backlog, dev, 1);
6218

6219
		netdev_wait_allrefs(dev);
L
Linus Torvalds 已提交
6220

6221
		/* paranoia */
E
Eric Dumazet 已提交
6222
		BUG_ON(netdev_refcnt_read(dev));
6223 6224
		WARN_ON(rcu_access_pointer(dev->ip_ptr));
		WARN_ON(rcu_access_pointer(dev->ip6_ptr));
6225
		WARN_ON(dev->dn_ptr);
L
Linus Torvalds 已提交
6226

6227 6228
		if (dev->destructor)
			dev->destructor(dev);
6229

6230 6231 6232 6233 6234 6235
		/* Report a network device has been unregistered */
		rtnl_lock();
		dev_net(dev)->dev_unreg_count--;
		__rtnl_unlock();
		wake_up(&netdev_unregistering_wq);

6236 6237
		/* Free network device */
		kobject_put(&dev->dev.kobj);
L
Linus Torvalds 已提交
6238 6239 6240
	}
}

6241 6242 6243
/* Convert net_device_stats to rtnl_link_stats64.  They have the same
 * fields in the same order, with only the type differing.
 */
6244 6245
void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
			     const struct net_device_stats *netdev_stats)
6246 6247
{
#if BITS_PER_LONG == 64
6248 6249
	BUILD_BUG_ON(sizeof(*stats64) != sizeof(*netdev_stats));
	memcpy(stats64, netdev_stats, sizeof(*stats64));
6250 6251 6252 6253 6254 6255 6256 6257 6258 6259 6260
#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
}
6261
EXPORT_SYMBOL(netdev_stats_to_stats64);
6262

6263 6264 6265
/**
 *	dev_get_stats	- get network device statistics
 *	@dev: device to get statistics from
6266
 *	@storage: place to store stats
6267
 *
6268 6269 6270 6271
 *	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.
6272
 */
6273 6274
struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
					struct rtnl_link_stats64 *storage)
6275
{
6276 6277
	const struct net_device_ops *ops = dev->netdev_ops;

6278 6279
	if (ops->ndo_get_stats64) {
		memset(storage, 0, sizeof(*storage));
6280 6281
		ops->ndo_get_stats64(dev, storage);
	} else if (ops->ndo_get_stats) {
6282
		netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev));
6283 6284
	} else {
		netdev_stats_to_stats64(storage, &dev->stats);
6285
	}
6286
	storage->rx_dropped += atomic_long_read(&dev->rx_dropped);
6287
	storage->tx_dropped += atomic_long_read(&dev->tx_dropped);
6288
	return storage;
R
Rusty Russell 已提交
6289
}
6290
EXPORT_SYMBOL(dev_get_stats);
R
Rusty Russell 已提交
6291

6292
struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
6293
{
6294
	struct netdev_queue *queue = dev_ingress_queue(dev);
6295

6296 6297 6298 6299 6300 6301 6302 6303 6304 6305 6306 6307
#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;
6308 6309
}

6310 6311
static const struct ethtool_ops default_ethtool_ops;

6312 6313 6314 6315 6316 6317 6318 6319
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);

6320 6321 6322 6323 6324 6325 6326 6327 6328 6329
void netdev_freemem(struct net_device *dev)
{
	char *addr = (char *)dev - dev->padded;

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

L
Linus Torvalds 已提交
6330
/**
T
Tom Herbert 已提交
6331
 *	alloc_netdev_mqs - allocate network device
L
Linus Torvalds 已提交
6332 6333 6334
 *	@sizeof_priv:	size of private data to allocate space for
 *	@name:		device name format string
 *	@setup:		callback to initialize device
T
Tom Herbert 已提交
6335 6336
 *	@txqs:		the number of TX subqueues to allocate
 *	@rxqs:		the number of RX subqueues to allocate
L
Linus Torvalds 已提交
6337 6338
 *
 *	Allocates a struct net_device with private data area for driver use
L
Li Zhong 已提交
6339
 *	and performs basic initialization.  Also allocates subqueue structs
T
Tom Herbert 已提交
6340
 *	for each queue on the device.
L
Linus Torvalds 已提交
6341
 */
T
Tom Herbert 已提交
6342 6343 6344
struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
		void (*setup)(struct net_device *),
		unsigned int txqs, unsigned int rxqs)
L
Linus Torvalds 已提交
6345 6346
{
	struct net_device *dev;
6347
	size_t alloc_size;
6348
	struct net_device *p;
L
Linus Torvalds 已提交
6349

6350 6351
	BUG_ON(strlen(name) >= sizeof(dev->name));

T
Tom Herbert 已提交
6352
	if (txqs < 1) {
6353
		pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
6354 6355 6356
		return NULL;
	}

6357
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
6358
	if (rxqs < 1) {
6359
		pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
T
Tom Herbert 已提交
6360 6361 6362 6363
		return NULL;
	}
#endif

6364
	alloc_size = sizeof(struct net_device);
6365 6366
	if (sizeof_priv) {
		/* ensure 32-byte alignment of private area */
6367
		alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
6368 6369 6370
		alloc_size += sizeof_priv;
	}
	/* ensure 32-byte alignment of whole construct */
6371
	alloc_size += NETDEV_ALIGN - 1;
L
Linus Torvalds 已提交
6372

6373 6374 6375
	p = kzalloc(alloc_size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!p)
		p = vzalloc(alloc_size);
6376
	if (!p)
L
Linus Torvalds 已提交
6377 6378
		return NULL;

6379
	dev = PTR_ALIGN(p, NETDEV_ALIGN);
L
Linus Torvalds 已提交
6380
	dev->padded = (char *)dev - (char *)p;
6381

E
Eric Dumazet 已提交
6382 6383
	dev->pcpu_refcnt = alloc_percpu(int);
	if (!dev->pcpu_refcnt)
6384
		goto free_dev;
6385 6386

	if (dev_addr_init(dev))
E
Eric Dumazet 已提交
6387
		goto free_pcpu;
6388

6389
	dev_mc_init(dev);
6390
	dev_uc_init(dev);
J
Jiri Pirko 已提交
6391

6392
	dev_net_set(dev, &init_net);
L
Linus Torvalds 已提交
6393

6394
	dev->gso_max_size = GSO_MAX_SIZE;
6395
	dev->gso_max_segs = GSO_MAX_SEGS;
6396 6397 6398

	INIT_LIST_HEAD(&dev->napi_list);
	INIT_LIST_HEAD(&dev->unreg_list);
6399
	INIT_LIST_HEAD(&dev->close_list);
6400
	INIT_LIST_HEAD(&dev->link_watch_list);
6401 6402 6403 6404
	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);
6405 6406 6407
	dev->priv_flags = IFF_XMIT_DST_RELEASE;
	setup(dev);

T
Tom Herbert 已提交
6408 6409
	dev->num_tx_queues = txqs;
	dev->real_num_tx_queues = txqs;
6410
	if (netif_alloc_netdev_queues(dev))
6411
		goto free_all;
6412

6413
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
6414 6415
	dev->num_rx_queues = rxqs;
	dev->real_num_rx_queues = rxqs;
T
Tom Herbert 已提交
6416
	if (netif_alloc_rx_queues(dev))
6417
		goto free_all;
E
Eric Dumazet 已提交
6418
#endif
T
Tom Herbert 已提交
6419

L
Linus Torvalds 已提交
6420
	strcpy(dev->name, name);
6421
	dev->group = INIT_NETDEV_GROUP;
6422 6423
	if (!dev->ethtool_ops)
		dev->ethtool_ops = &default_ethtool_ops;
L
Linus Torvalds 已提交
6424
	return dev;
6425

6426 6427 6428 6429
free_all:
	free_netdev(dev);
	return NULL;

E
Eric Dumazet 已提交
6430 6431
free_pcpu:
	free_percpu(dev->pcpu_refcnt);
6432
	netif_free_tx_queues(dev);
6433
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
6434 6435 6436
	kfree(dev->_rx);
#endif

6437 6438
free_dev:
	netdev_freemem(dev);
6439
	return NULL;
L
Linus Torvalds 已提交
6440
}
T
Tom Herbert 已提交
6441
EXPORT_SYMBOL(alloc_netdev_mqs);
L
Linus Torvalds 已提交
6442 6443 6444 6445 6446

/**
 *	free_netdev - free network device
 *	@dev: device
 *
6447 6448
 *	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
6449 6450 6451 6452
 *	If this is the last reference then it will be freed.
 */
void free_netdev(struct net_device *dev)
{
6453 6454
	struct napi_struct *p, *n;

6455 6456
	release_net(dev_net(dev));

6457
	netif_free_tx_queues(dev);
6458
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
6459 6460
	kfree(dev->_rx);
#endif
6461

6462
	kfree(rcu_dereference_protected(dev->ingress_queue, 1));
6463

6464 6465 6466
	/* Flush device addresses */
	dev_addr_flush(dev);

6467 6468 6469
	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
		netif_napi_del(p);

E
Eric Dumazet 已提交
6470 6471 6472
	free_percpu(dev->pcpu_refcnt);
	dev->pcpu_refcnt = NULL;

S
Stephen Hemminger 已提交
6473
	/*  Compatibility with error handling in drivers */
L
Linus Torvalds 已提交
6474
	if (dev->reg_state == NETREG_UNINITIALIZED) {
6475
		netdev_freemem(dev);
L
Linus Torvalds 已提交
6476 6477 6478 6479 6480 6481
		return;
	}

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

6482 6483
	/* will free via device release */
	put_device(&dev->dev);
L
Linus Torvalds 已提交
6484
}
E
Eric Dumazet 已提交
6485
EXPORT_SYMBOL(free_netdev);
6486

6487 6488 6489 6490 6491 6492
/**
 *	synchronize_net -  Synchronize with packet receive processing
 *
 *	Wait for packets currently being received to be done.
 *	Does not block later packets from starting.
 */
6493
void synchronize_net(void)
L
Linus Torvalds 已提交
6494 6495
{
	might_sleep();
6496 6497 6498 6499
	if (rtnl_is_locked())
		synchronize_rcu_expedited();
	else
		synchronize_rcu();
L
Linus Torvalds 已提交
6500
}
E
Eric Dumazet 已提交
6501
EXPORT_SYMBOL(synchronize_net);
L
Linus Torvalds 已提交
6502 6503

/**
6504
 *	unregister_netdevice_queue - remove device from the kernel
L
Linus Torvalds 已提交
6505
 *	@dev: device
6506
 *	@head: list
6507
 *
L
Linus Torvalds 已提交
6508
 *	This function shuts down a device interface and removes it
6509
 *	from the kernel tables.
6510
 *	If head not NULL, device is queued to be unregistered later.
L
Linus Torvalds 已提交
6511 6512 6513 6514 6515
 *
 *	Callers must hold the rtnl semaphore.  You may want
 *	unregister_netdev() instead of this.
 */

6516
void unregister_netdevice_queue(struct net_device *dev, struct list_head *head)
L
Linus Torvalds 已提交
6517
{
6518 6519
	ASSERT_RTNL();

6520
	if (head) {
6521
		list_move_tail(&dev->unreg_list, head);
6522 6523 6524 6525 6526
	} else {
		rollback_registered(dev);
		/* Finish processing unregister after unlock */
		net_set_todo(dev);
	}
L
Linus Torvalds 已提交
6527
}
6528
EXPORT_SYMBOL(unregister_netdevice_queue);
L
Linus Torvalds 已提交
6529

6530 6531 6532 6533 6534 6535 6536 6537 6538 6539 6540 6541 6542 6543
/**
 *	unregister_netdevice_many - unregister many devices
 *	@head: list of devices
 */
void unregister_netdevice_many(struct list_head *head)
{
	struct net_device *dev;

	if (!list_empty(head)) {
		rollback_registered_many(head);
		list_for_each_entry(dev, head, unreg_list)
			net_set_todo(dev);
	}
}
6544
EXPORT_SYMBOL(unregister_netdevice_many);
6545

L
Linus Torvalds 已提交
6546 6547 6548 6549 6550
/**
 *	unregister_netdev - remove device from the kernel
 *	@dev: device
 *
 *	This function shuts down a device interface and removes it
6551
 *	from the kernel tables.
L
Linus Torvalds 已提交
6552 6553 6554 6555 6556 6557 6558 6559 6560 6561 6562 6563 6564
 *
 *	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);

6565 6566 6567 6568 6569 6570 6571 6572 6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590 6591 6592 6593 6594 6595
/**
 *	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;
6596
	if (net_eq(dev_net(dev), net))
6597 6598 6599 6600 6601 6602
		goto out;

	/* Pick the destination device name, and ensure
	 * we can use it in the destination network namespace.
	 */
	err = -EEXIST;
6603
	if (__dev_get_by_name(net, dev->name)) {
6604 6605 6606
		/* We get here if we can't use the current device name */
		if (!pat)
			goto out;
6607
		if (dev_get_valid_name(net, dev, pat) < 0)
6608 6609 6610 6611 6612 6613 6614 6615
			goto out;
	}

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

	/* If device is running close it first. */
6616
	dev_close(dev);
6617 6618 6619 6620 6621 6622 6623 6624 6625 6626 6627 6628

	/* 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.
6629 6630 6631 6632

	   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.
6633 6634
	*/
	call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
6635 6636
	rcu_barrier();
	call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6637
	rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);
6638 6639 6640 6641

	/*
	 *	Flush the unicast and multicast chains
	 */
6642
	dev_uc_flush(dev);
6643
	dev_mc_flush(dev);
6644

6645 6646 6647
	/* Send a netdev-removed uevent to the old namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_REMOVE);

6648
	/* Actually switch the network namespace */
6649
	dev_net_set(dev, net);
6650 6651 6652 6653 6654 6655 6656 6657 6658

	/* 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;
	}

6659 6660 6661
	/* Send a netdev-add uevent to the new namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_ADD);

6662
	/* Fixup kobjects */
6663
	err = device_rename(&dev->dev, dev->name);
6664
	WARN_ON(err);
6665 6666 6667 6668 6669 6670 6671

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

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

6672 6673 6674 6675
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
6676
	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
6677

6678 6679 6680 6681 6682
	synchronize_net();
	err = 0;
out:
	return err;
}
6683
EXPORT_SYMBOL_GPL(dev_change_net_namespace);
6684

L
Linus Torvalds 已提交
6685 6686 6687 6688 6689 6690 6691 6692 6693
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;

6694
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
L
Linus Torvalds 已提交
6695 6696 6697 6698 6699 6700 6701 6702 6703 6704 6705 6706 6707 6708 6709 6710
		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. */
6711 6712 6713 6714 6715 6716
	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;
	}
6717 6718 6719 6720 6721
	/* 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
6722 6723 6724 6725 6726

	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_enable();

	/* Process offline CPU's input_pkt_queue */
6727
	while ((skb = __skb_dequeue(&oldsd->process_queue))) {
6728
		netif_rx_internal(skb);
6729
		input_queue_head_incr(oldsd);
T
Tom Herbert 已提交
6730
	}
6731
	while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) {
6732
		netif_rx_internal(skb);
6733 6734
		input_queue_head_incr(oldsd);
	}
L
Linus Torvalds 已提交
6735 6736 6737 6738 6739

	return NOTIFY_OK;
}


6740
/**
6741 6742 6743 6744
 *	netdev_increment_features - increment feature set by one
 *	@all: current feature set
 *	@one: new feature set
 *	@mask: mask feature set
6745 6746
 *
 *	Computes a new feature set after adding a device with feature set
6747 6748
 *	@one to the master device with current feature set @all.  Will not
 *	enable anything that is off in @mask. Returns the new feature set.
6749
 */
6750 6751
netdev_features_t netdev_increment_features(netdev_features_t all,
	netdev_features_t one, netdev_features_t mask)
6752
{
6753 6754 6755
	if (mask & NETIF_F_GEN_CSUM)
		mask |= NETIF_F_ALL_CSUM;
	mask |= NETIF_F_VLAN_CHALLENGED;
6756

6757 6758
	all |= one & (NETIF_F_ONE_FOR_ALL|NETIF_F_ALL_CSUM) & mask;
	all &= one | ~NETIF_F_ALL_FOR_ALL;
6759

6760 6761 6762
	/* If one device supports hw checksumming, set for all. */
	if (all & NETIF_F_GEN_CSUM)
		all &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
6763 6764 6765

	return all;
}
6766
EXPORT_SYMBOL(netdev_increment_features);
6767

6768
static struct hlist_head * __net_init netdev_create_hash(void)
6769 6770 6771 6772 6773 6774 6775 6776 6777 6778 6779 6780
{
	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;
}

6781
/* Initialize per network namespace state */
6782
static int __net_init netdev_init(struct net *net)
6783
{
6784 6785
	if (net != &init_net)
		INIT_LIST_HEAD(&net->dev_base_head);
6786

6787 6788 6789
	net->dev_name_head = netdev_create_hash();
	if (net->dev_name_head == NULL)
		goto err_name;
6790

6791 6792 6793
	net->dev_index_head = netdev_create_hash();
	if (net->dev_index_head == NULL)
		goto err_idx;
6794 6795

	return 0;
6796 6797 6798 6799 6800

err_idx:
	kfree(net->dev_name_head);
err_name:
	return -ENOMEM;
6801 6802
}

6803 6804 6805 6806 6807 6808
/**
 *	netdev_drivername - network driver for the device
 *	@dev: network device
 *
 *	Determine network driver for device.
 */
6809
const char *netdev_drivername(const struct net_device *dev)
6810
{
6811 6812
	const struct device_driver *driver;
	const struct device *parent;
6813
	const char *empty = "";
6814 6815 6816

	parent = dev->dev.parent;
	if (!parent)
6817
		return empty;
6818 6819 6820

	driver = parent->driver;
	if (driver && driver->name)
6821 6822
		return driver->name;
	return empty;
6823 6824
}

6825
static int __netdev_printk(const char *level, const struct net_device *dev,
6826 6827 6828 6829
			   struct va_format *vaf)
{
	int r;

6830
	if (dev && dev->dev.parent) {
6831 6832 6833 6834 6835 6836
		r = dev_printk_emit(level[1] - '0',
				    dev->dev.parent,
				    "%s %s %s: %pV",
				    dev_driver_string(dev->dev.parent),
				    dev_name(dev->dev.parent),
				    netdev_name(dev), vaf);
6837
	} else if (dev) {
6838
		r = printk("%s%s: %pV", level, netdev_name(dev), vaf);
6839
	} else {
6840
		r = printk("%s(NULL net_device): %pV", level, vaf);
6841
	}
6842 6843 6844 6845 6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857 6858

	return r;
}

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

	va_start(args, format);

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

	r = __netdev_printk(level, dev, &vaf);
6859

6860 6861 6862 6863 6864 6865 6866 6867 6868 6869 6870 6871 6872 6873 6874 6875 6876 6877 6878
	va_end(args);

	return r;
}
EXPORT_SYMBOL(netdev_printk);

#define define_netdev_printk_level(func, level)			\
int func(const struct net_device *dev, const char *fmt, ...)	\
{								\
	int r;							\
	struct va_format vaf;					\
	va_list args;						\
								\
	va_start(args, fmt);					\
								\
	vaf.fmt = fmt;						\
	vaf.va = &args;						\
								\
	r = __netdev_printk(level, dev, &vaf);			\
6879
								\
6880 6881 6882 6883 6884 6885 6886 6887 6888 6889 6890 6891 6892 6893
	va_end(args);						\
								\
	return r;						\
}								\
EXPORT_SYMBOL(func);

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

6894
static void __net_exit netdev_exit(struct net *net)
6895 6896 6897 6898 6899
{
	kfree(net->dev_name_head);
	kfree(net->dev_index_head);
}

6900
static struct pernet_operations __net_initdata netdev_net_ops = {
6901 6902 6903 6904
	.init = netdev_init,
	.exit = netdev_exit,
};

6905
static void __net_exit default_device_exit(struct net *net)
6906
{
6907
	struct net_device *dev, *aux;
6908
	/*
6909
	 * Push all migratable network devices back to the
6910 6911 6912
	 * initial network namespace
	 */
	rtnl_lock();
6913
	for_each_netdev_safe(net, dev, aux) {
6914
		int err;
6915
		char fb_name[IFNAMSIZ];
6916 6917 6918 6919 6920

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

6921 6922 6923
		/* Leave virtual devices for the generic cleanup */
		if (dev->rtnl_link_ops)
			continue;
6924

L
Lucas De Marchi 已提交
6925
		/* Push remaining network devices to init_net */
6926 6927
		snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex);
		err = dev_change_net_namespace(dev, &init_net, fb_name);
6928
		if (err) {
6929 6930
			pr_emerg("%s: failed to move %s to init_net: %d\n",
				 __func__, dev->name, err);
6931
			BUG();
6932 6933 6934 6935 6936
		}
	}
	rtnl_unlock();
}

6937 6938 6939 6940 6941 6942 6943 6944 6945 6946 6947 6948 6949 6950 6951 6952 6953 6954 6955 6956 6957 6958 6959 6960 6961 6962 6963 6964
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);
}

6965 6966 6967
static void __net_exit default_device_exit_batch(struct list_head *net_list)
{
	/* At exit all network devices most be removed from a network
6968
	 * namespace.  Do this in the reverse order of registration.
6969 6970 6971 6972 6973 6974 6975
	 * 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);

6976 6977 6978 6979 6980 6981 6982 6983 6984 6985 6986 6987
	/* 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);
6988 6989 6990 6991 6992 6993 6994 6995 6996
	list_for_each_entry(net, net_list, exit_list) {
		for_each_netdev_reverse(net, dev) {
			if (dev->rtnl_link_ops)
				dev->rtnl_link_ops->dellink(dev, &dev_kill_list);
			else
				unregister_netdevice_queue(dev, &dev_kill_list);
		}
	}
	unregister_netdevice_many(&dev_kill_list);
E
Eric Dumazet 已提交
6997
	list_del(&dev_kill_list);
6998 6999 7000
	rtnl_unlock();
}

7001
static struct pernet_operations __net_initdata default_device_ops = {
7002
	.exit = default_device_exit,
7003
	.exit_batch = default_device_exit_batch,
7004 7005
};

L
Linus Torvalds 已提交
7006 7007 7008 7009 7010 7011 7012 7013 7014 7015 7016 7017 7018 7019 7020 7021 7022 7023 7024 7025
/*
 *	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;

7026
	if (netdev_kobject_init())
L
Linus Torvalds 已提交
7027 7028 7029
		goto out;

	INIT_LIST_HEAD(&ptype_all);
7030
	for (i = 0; i < PTYPE_HASH_SIZE; i++)
L
Linus Torvalds 已提交
7031 7032
		INIT_LIST_HEAD(&ptype_base[i]);

7033 7034
	INIT_LIST_HEAD(&offload_base);

7035 7036
	if (register_pernet_subsys(&netdev_net_ops))
		goto out;
L
Linus Torvalds 已提交
7037 7038 7039 7040 7041

	/*
	 *	Initialise the packet receive queues.
	 */

7042
	for_each_possible_cpu(i) {
E
Eric Dumazet 已提交
7043
		struct softnet_data *sd = &per_cpu(softnet_data, i);
L
Linus Torvalds 已提交
7044

E
Eric Dumazet 已提交
7045
		skb_queue_head_init(&sd->input_pkt_queue);
7046
		skb_queue_head_init(&sd->process_queue);
E
Eric Dumazet 已提交
7047
		INIT_LIST_HEAD(&sd->poll_list);
7048
		sd->output_queue_tailp = &sd->output_queue;
E
Eric Dumazet 已提交
7049
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
7050 7051 7052
		sd->csd.func = rps_trigger_softirq;
		sd->csd.info = sd;
		sd->cpu = i;
7053
#endif
T
Tom Herbert 已提交
7054

E
Eric Dumazet 已提交
7055 7056
		sd->backlog.poll = process_backlog;
		sd->backlog.weight = weight_p;
L
Linus Torvalds 已提交
7057 7058 7059 7060
	}

	dev_boot_phase = 0;

7061 7062 7063 7064 7065 7066 7067 7068 7069 7070 7071 7072 7073 7074 7075
	/* 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;

7076 7077
	open_softirq(NET_TX_SOFTIRQ, net_tx_action);
	open_softirq(NET_RX_SOFTIRQ, net_rx_action);
L
Linus Torvalds 已提交
7078 7079 7080 7081 7082 7083 7084 7085 7086

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

subsys_initcall(net_dev_init);