dev.c 192.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9
/*
 * 	NET3	Protocol independent device support routines.
 *
 *		This program is free software; you can redistribute it and/or
 *		modify it under the terms of the GNU General Public License
 *		as published by the Free Software Foundation; either version
 *		2 of the License, or (at your option) any later version.
 *
 *	Derived from the non IP parts of dev.c 1.0.19
10
 * 		Authors:	Ross Biro
L
Linus Torvalds 已提交
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76
 *				Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 *				Mark Evans, <evansmp@uhura.aston.ac.uk>
 *
 *	Additional Authors:
 *		Florian la Roche <rzsfl@rz.uni-sb.de>
 *		Alan Cox <gw4pts@gw4pts.ampr.org>
 *		David Hinds <dahinds@users.sourceforge.net>
 *		Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
 *		Adam Sulmicki <adam@cfar.umd.edu>
 *              Pekka Riikonen <priikone@poesidon.pspt.fi>
 *
 *	Changes:
 *              D.J. Barrow     :       Fixed bug where dev->refcnt gets set
 *              			to 2 if register_netdev gets called
 *              			before net_dev_init & also removed a
 *              			few lines of code in the process.
 *		Alan Cox	:	device private ioctl copies fields back.
 *		Alan Cox	:	Transmit queue code does relevant
 *					stunts to keep the queue safe.
 *		Alan Cox	:	Fixed double lock.
 *		Alan Cox	:	Fixed promisc NULL pointer trap
 *		????????	:	Support the full private ioctl range
 *		Alan Cox	:	Moved ioctl permission check into
 *					drivers
 *		Tim Kordas	:	SIOCADDMULTI/SIOCDELMULTI
 *		Alan Cox	:	100 backlog just doesn't cut it when
 *					you start doing multicast video 8)
 *		Alan Cox	:	Rewrote net_bh and list manager.
 *		Alan Cox	: 	Fix ETH_P_ALL echoback lengths.
 *		Alan Cox	:	Took out transmit every packet pass
 *					Saved a few bytes in the ioctl handler
 *		Alan Cox	:	Network driver sets packet type before
 *					calling netif_rx. Saves a function
 *					call a packet.
 *		Alan Cox	:	Hashed net_bh()
 *		Richard Kooijman:	Timestamp fixes.
 *		Alan Cox	:	Wrong field in SIOCGIFDSTADDR
 *		Alan Cox	:	Device lock protection.
 *		Alan Cox	: 	Fixed nasty side effect of device close
 *					changes.
 *		Rudi Cilibrasi	:	Pass the right thing to
 *					set_mac_address()
 *		Dave Miller	:	32bit quantity for the device lock to
 *					make it work out on a Sparc.
 *		Bjorn Ekwall	:	Added KERNELD hack.
 *		Alan Cox	:	Cleaned up the backlog initialise.
 *		Craig Metz	:	SIOCGIFCONF fix if space for under
 *					1 device.
 *	    Thomas Bogendoerfer :	Return ENODEV for dev_open, if there
 *					is no device open function.
 *		Andi Kleen	:	Fix error reporting for SIOCGIFCONF
 *	    Michael Chastain	:	Fix signed/unsigned for SIOCGIFCONF
 *		Cyrus Durgin	:	Cleaned for KMOD
 *		Adam Sulmicki   :	Bug Fix : Network Device Unload
 *					A network device unload needs to purge
 *					the backlog queue.
 *	Paul Rusty Russell	:	SIOCSIFNAME
 *              Pekka Riikonen  :	Netdev boot-time settings code
 *              Andrew Morton   :       Make unregister_netdevice wait
 *              			indefinitely on dev->refcnt
 * 		J Hadi Salim	:	- Backlog queue sampling
 *				        - netif_rx() feedback
 */

#include <asm/uaccess.h>
#include <linux/bitops.h>
77
#include <linux/capability.h>
L
Linus Torvalds 已提交
78 79 80
#include <linux/cpu.h>
#include <linux/types.h>
#include <linux/kernel.h>
81
#include <linux/hash.h>
82
#include <linux/slab.h>
L
Linus Torvalds 已提交
83
#include <linux/sched.h>
A
Arjan van de Ven 已提交
84
#include <linux/mutex.h>
L
Linus Torvalds 已提交
85 86 87 88 89 90 91 92 93
#include <linux/string.h>
#include <linux/mm.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/errno.h>
#include <linux/interrupt.h>
#include <linux/if_ether.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
94
#include <linux/ethtool.h>
L
Linus Torvalds 已提交
95 96
#include <linux/notifier.h>
#include <linux/skbuff.h>
97
#include <net/net_namespace.h>
L
Linus Torvalds 已提交
98 99 100 101
#include <net/sock.h>
#include <linux/rtnetlink.h>
#include <linux/stat.h>
#include <net/dst.h>
102
#include <net/dst_metadata.h>
L
Linus Torvalds 已提交
103 104
#include <net/pkt_sched.h>
#include <net/checksum.h>
105
#include <net/xfrm.h>
L
Linus Torvalds 已提交
106 107 108 109 110 111 112 113
#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
114
#include <linux/audit.h>
115
#include <linux/dmaengine.h>
116
#include <linux/err.h>
117
#include <linux/ctype.h>
118
#include <linux/if_arp.h>
119
#include <linux/if_vlan.h>
120
#include <linux/ip.h>
121
#include <net/ip.h>
122
#include <net/mpls.h>
123 124
#include <linux/ipv6.h>
#include <linux/in.h>
D
David S. Miller 已提交
125 126
#include <linux/jhash.h>
#include <linux/random.h>
127
#include <trace/events/napi.h>
128
#include <trace/events/net.h>
129
#include <trace/events/skb.h>
130
#include <linux/pci.h>
131
#include <linux/inetdevice.h>
132
#include <linux/cpu_rmap.h>
133
#include <linux/static_key.h>
E
Eliezer Tamir 已提交
134
#include <linux/hashtable.h>
135
#include <linux/vmalloc.h>
136
#include <linux/if_macvlan.h>
137
#include <linux/errqueue.h>
138
#include <linux/hrtimer.h>
139
#include <linux/netfilter_ingress.h>
L
Linus Torvalds 已提交
140

141 142
#include "net-sysfs.h"

143 144 145
/* Instead of increasing this, you should create a hash table. */
#define MAX_GRO_SKBS 8

H
Herbert Xu 已提交
146 147 148
/* 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
149
static DEFINE_SPINLOCK(ptype_lock);
150
static DEFINE_SPINLOCK(offload_lock);
151 152
struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
struct list_head ptype_all __read_mostly;	/* Taps */
153
static struct list_head offload_base __read_mostly;
L
Linus Torvalds 已提交
154

155
static int netif_rx_internal(struct sk_buff *skb);
156 157 158
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info);
159

L
Linus Torvalds 已提交
160
/*
161
 * The @dev_base_head list is protected by @dev_base_lock and the rtnl
L
Linus Torvalds 已提交
162 163
 * semaphore.
 *
164
 * Pure readers hold dev_base_lock for reading, or rcu_read_lock()
L
Linus Torvalds 已提交
165 166
 *
 * Writers must hold the rtnl semaphore while they loop through the
167
 * dev_base_head list, and hold dev_base_lock for writing when they do the
L
Linus Torvalds 已提交
168 169 170 171 172 173 174 175 176 177 178 179 180 181
 * 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
182 183 184 185 186 187
/* 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);

188
static seqcount_t devnet_rename_seq;
189

190 191 192 193 194
static inline void dev_base_seq_inc(struct net *net)
{
	while (++net->dev_base_seq == 0);
}

195
static inline struct hlist_head *dev_name_hash(struct net *net, const char *name)
L
Linus Torvalds 已提交
196
{
197 198
	unsigned int hash = full_name_hash(name, strnlen(name, IFNAMSIZ));

199
	return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)];
L
Linus Torvalds 已提交
200 201
}

202
static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex)
L
Linus Torvalds 已提交
203
{
204
	return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)];
L
Linus Torvalds 已提交
205 206
}

E
Eric Dumazet 已提交
207
static inline void rps_lock(struct softnet_data *sd)
208 209
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
210
	spin_lock(&sd->input_pkt_queue.lock);
211 212 213
#endif
}

E
Eric Dumazet 已提交
214
static inline void rps_unlock(struct softnet_data *sd)
215 216
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
217
	spin_unlock(&sd->input_pkt_queue.lock);
218 219 220
#endif
}

221
/* Device list insertion */
222
static void list_netdevice(struct net_device *dev)
223
{
224
	struct net *net = dev_net(dev);
225 226 227 228

	ASSERT_RTNL();

	write_lock_bh(&dev_base_lock);
229
	list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
230
	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
231 232
	hlist_add_head_rcu(&dev->index_hlist,
			   dev_index_hash(net, dev->ifindex));
233
	write_unlock_bh(&dev_base_lock);
234 235

	dev_base_seq_inc(net);
236 237
}

238 239 240
/* Device list removal
 * caller must respect a RCU grace period before freeing/reusing dev
 */
241 242 243 244 245 246
static void unlist_netdevice(struct net_device *dev)
{
	ASSERT_RTNL();

	/* Unlink dev from the device chain */
	write_lock_bh(&dev_base_lock);
247
	list_del_rcu(&dev->dev_list);
248
	hlist_del_rcu(&dev->name_hlist);
249
	hlist_del_rcu(&dev->index_hlist);
250
	write_unlock_bh(&dev_base_lock);
251 252

	dev_base_seq_inc(dev_net(dev));
253 254
}

L
Linus Torvalds 已提交
255 256 257 258
/*
 *	Our notifier list
 */

259
static RAW_NOTIFIER_HEAD(netdev_chain);
L
Linus Torvalds 已提交
260 261 262 263 264

/*
 *	Device drivers call our routines to queue packets here. We empty the
 *	queue in the local softnet handler.
 */
265

266
DEFINE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
E
Eric Dumazet 已提交
267
EXPORT_PER_CPU_SYMBOL(softnet_data);
L
Linus Torvalds 已提交
268

269
#ifdef CONFIG_LOCKDEP
270
/*
271
 * register_netdevice() inits txq->_xmit_lock and sets lockdep class
272 273 274 275 276 277 278 279 280 281 282 283 284 285 286
 * 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,
287 288 289
	 ARPHRD_FCFABRIC, ARPHRD_IEEE80211, ARPHRD_IEEE80211_PRISM,
	 ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, ARPHRD_PHONET_PIPE,
	 ARPHRD_IEEE802154, ARPHRD_VOID, ARPHRD_NONE};
290

291
static const char *const netdev_lock_name[] =
292 293 294 295 296 297 298 299 300 301 302 303
	{"_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",
304 305 306
	 "_xmit_FCFABRIC", "_xmit_IEEE80211", "_xmit_IEEE80211_PRISM",
	 "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", "_xmit_PHONET_PIPE",
	 "_xmit_IEEE802154", "_xmit_VOID", "_xmit_NONE"};
307 308

static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)];
309
static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)];
310 311 312 313 314 315 316 317 318 319 320 321

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;
}

322 323
static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
						 unsigned short dev_type)
324 325 326 327 328 329 330
{
	int i;

	i = netdev_lock_pos(dev_type);
	lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i],
				   netdev_lock_name[i]);
}
331 332 333 334 335 336 337 338 339 340

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]);
}
341
#else
342 343 344 345 346
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)
347 348 349
{
}
#endif
L
Linus Torvalds 已提交
350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372

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

		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)
 */

373 374 375
static inline struct list_head *ptype_head(const struct packet_type *pt)
{
	if (pt->type == htons(ETH_P_ALL))
376
		return pt->dev ? &pt->dev->ptype_all : &ptype_all;
377
	else
378 379
		return pt->dev ? &pt->dev->ptype_specific :
				 &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK];
380 381
}

L
Linus Torvalds 已提交
382 383 384 385 386 387 388 389
/**
 *	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.
 *
390
 *	This call does not sleep therefore it can not
L
Linus Torvalds 已提交
391 392 393 394 395 396
 *	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)
{
397
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
398

399 400 401
	spin_lock(&ptype_lock);
	list_add_rcu(&pt->list, head);
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
402
}
E
Eric Dumazet 已提交
403
EXPORT_SYMBOL(dev_add_pack);
L
Linus Torvalds 已提交
404 405 406 407 408 409 410 411

/**
 *	__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
412
 *	returns.
L
Linus Torvalds 已提交
413 414 415 416 417 418 419
 *
 *      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)
{
420
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
421 422
	struct packet_type *pt1;

423
	spin_lock(&ptype_lock);
L
Linus Torvalds 已提交
424 425 426 427 428 429 430 431

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

432
	pr_warn("dev_remove_pack: %p not found\n", pt);
L
Linus Torvalds 已提交
433
out:
434
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
435
}
E
Eric Dumazet 已提交
436 437
EXPORT_SYMBOL(__dev_remove_pack);

L
Linus Torvalds 已提交
438 439 440 441 442 443 444 445 446 447 448 449 450 451 452
/**
 *	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);
453

L
Linus Torvalds 已提交
454 455
	synchronize_net();
}
E
Eric Dumazet 已提交
456
EXPORT_SYMBOL(dev_remove_pack);
L
Linus Torvalds 已提交
457

458 459 460 461 462 463 464 465 466 467 468 469 470 471 472

/**
 *	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)
{
473
	struct packet_offload *elem;
474 475

	spin_lock(&offload_lock);
476 477 478 479 480
	list_for_each_entry(elem, &offload_base, list) {
		if (po->priority < elem->priority)
			break;
	}
	list_add_rcu(&po->list, elem->list.prev);
481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497
	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
498
static void __dev_remove_offload(struct packet_offload *po)
499 500 501 502
{
	struct list_head *head = &offload_base;
	struct packet_offload *po1;

503
	spin_lock(&offload_lock);
504 505 506 507 508 509 510 511 512 513

	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:
514
	spin_unlock(&offload_lock);
515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536
}

/**
 *	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
537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563
/******************************************************************************

		      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));
564
			strlcpy(s[i].name, name, IFNAMSIZ);
L
Linus Torvalds 已提交
565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588
			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] != ' ' &&
589
		    !strcmp(dev->name, s[i].name)) {
L
Linus Torvalds 已提交
590 591 592 593 594 595 596 597 598
			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
599
EXPORT_SYMBOL(netdev_boot_setup_check);
L
Linus Torvalds 已提交
600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623


/**
 *	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
	 */
624
	if (__dev_get_by_name(&init_net, name))
L
Linus Torvalds 已提交
625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667
		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

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

668 669 670 671 672 673 674 675 676 677 678 679 680
/**
 *	dev_get_iflink	- get 'iflink' value of a interface
 *	@dev: targeted interface
 *
 *	Indicates the ifindex the interface is linked to.
 *	Physical interfaces have the same 'ifindex' and 'iflink' values.
 */

int dev_get_iflink(const struct net_device *dev)
{
	if (dev->netdev_ops && dev->netdev_ops->ndo_get_iflink)
		return dev->netdev_ops->ndo_get_iflink(dev);

681
	return dev->ifindex;
682 683 684
}
EXPORT_SYMBOL(dev_get_iflink);

685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710
/**
 *	dev_fill_metadata_dst - Retrieve tunnel egress information.
 *	@dev: targeted interface
 *	@skb: The packet.
 *
 *	For better visibility of tunnel traffic OVS needs to retrieve
 *	egress tunnel information for a packet. Following API allows
 *	user to get this info.
 */
int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
{
	struct ip_tunnel_info *info;

	if (!dev->netdev_ops  || !dev->netdev_ops->ndo_fill_metadata_dst)
		return -EINVAL;

	info = skb_tunnel_info_unclone(skb);
	if (!info)
		return -ENOMEM;
	if (unlikely(!(info->mode & IP_TUNNEL_INFO_TX)))
		return -EINVAL;

	return dev->netdev_ops->ndo_fill_metadata_dst(dev, skb);
}
EXPORT_SYMBOL_GPL(dev_fill_metadata_dst);

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

723
struct net_device *__dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
724
{
E
Eric Dumazet 已提交
725 726
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);
L
Linus Torvalds 已提交
727

728
	hlist_for_each_entry(dev, head, name_hlist)
L
Linus Torvalds 已提交
729 730
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;
E
Eric Dumazet 已提交
731

L
Linus Torvalds 已提交
732 733
	return NULL;
}
E
Eric Dumazet 已提交
734
EXPORT_SYMBOL(__dev_get_by_name);
L
Linus Torvalds 已提交
735

736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752
/**
 *	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);

753
	hlist_for_each_entry_rcu(dev, head, name_hlist)
754 755 756 757 758 759 760
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_name_rcu);

L
Linus Torvalds 已提交
761 762
/**
 *	dev_get_by_name		- find a device by its name
763
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
764 765 766 767 768 769 770 771 772
 *	@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.
 */

773
struct net_device *dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
774 775 776
{
	struct net_device *dev;

777 778
	rcu_read_lock();
	dev = dev_get_by_name_rcu(net, name);
L
Linus Torvalds 已提交
779 780
	if (dev)
		dev_hold(dev);
781
	rcu_read_unlock();
L
Linus Torvalds 已提交
782 783
	return dev;
}
E
Eric Dumazet 已提交
784
EXPORT_SYMBOL(dev_get_by_name);
L
Linus Torvalds 已提交
785 786 787

/**
 *	__dev_get_by_index - find a device by its ifindex
788
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
789 790 791 792 793 794 795 796 797
 *	@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.
 */

798
struct net_device *__dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
799
{
E
Eric Dumazet 已提交
800 801
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);
L
Linus Torvalds 已提交
802

803
	hlist_for_each_entry(dev, head, index_hlist)
L
Linus Torvalds 已提交
804 805
		if (dev->ifindex == ifindex)
			return dev;
E
Eric Dumazet 已提交
806

L
Linus Torvalds 已提交
807 808
	return NULL;
}
E
Eric Dumazet 已提交
809
EXPORT_SYMBOL(__dev_get_by_index);
L
Linus Torvalds 已提交
810

811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826
/**
 *	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);

827
	hlist_for_each_entry_rcu(dev, head, index_hlist)
828 829 830 831 832 833 834
		if (dev->ifindex == ifindex)
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_index_rcu);

L
Linus Torvalds 已提交
835 836 837

/**
 *	dev_get_by_index - find a device by its ifindex
838
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
839 840 841 842 843 844 845 846
 *	@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.
 */

847
struct net_device *dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
848 849 850
{
	struct net_device *dev;

851 852
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, ifindex);
L
Linus Torvalds 已提交
853 854
	if (dev)
		dev_hold(dev);
855
	rcu_read_unlock();
L
Linus Torvalds 已提交
856 857
	return dev;
}
E
Eric Dumazet 已提交
858
EXPORT_SYMBOL(dev_get_by_index);
L
Linus Torvalds 已提交
859

860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893
/**
 *	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
894
/**
895
 *	dev_getbyhwaddr_rcu - find a device by its hardware address
896
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
897 898 899 900
 *	@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
901 902
 *	is not found or a pointer to the device.
 *	The caller must hold RCU or RTNL.
903
 *	The returned device has not had its ref count increased
L
Linus Torvalds 已提交
904 905 906 907
 *	and the caller must therefore be careful about locking
 *
 */

908 909
struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
				       const char *ha)
L
Linus Torvalds 已提交
910 911 912
{
	struct net_device *dev;

913
	for_each_netdev_rcu(net, dev)
L
Linus Torvalds 已提交
914 915
		if (dev->type == type &&
		    !memcmp(dev->dev_addr, ha, dev->addr_len))
916 917 918
			return dev;

	return NULL;
L
Linus Torvalds 已提交
919
}
920
EXPORT_SYMBOL(dev_getbyhwaddr_rcu);
921

922
struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type)
L
Linus Torvalds 已提交
923 924 925
{
	struct net_device *dev;

926
	ASSERT_RTNL();
927
	for_each_netdev(net, dev)
928
		if (dev->type == type)
929 930 931
			return dev;

	return NULL;
932 933 934
}
EXPORT_SYMBOL(__dev_getfirstbyhwtype);

935
struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type)
936
{
937
	struct net_device *dev, *ret = NULL;
938

939 940 941 942 943 944 945 946 947
	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
948 949 950 951
}
EXPORT_SYMBOL(dev_getfirstbyhwtype);

/**
952
 *	__dev_get_by_flags - find any device with given flags
953
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
954 955 956 957
 *	@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
958
 *	is not found or a pointer to the device. Must be called inside
959
 *	rtnl_lock(), and result refcount is unchanged.
L
Linus Torvalds 已提交
960 961
 */

962 963
struct net_device *__dev_get_by_flags(struct net *net, unsigned short if_flags,
				      unsigned short mask)
L
Linus Torvalds 已提交
964
{
965
	struct net_device *dev, *ret;
L
Linus Torvalds 已提交
966

967 968
	ASSERT_RTNL();

969
	ret = NULL;
970
	for_each_netdev(net, dev) {
L
Linus Torvalds 已提交
971
		if (((dev->flags ^ if_flags) & mask) == 0) {
972
			ret = dev;
L
Linus Torvalds 已提交
973 974 975
			break;
		}
	}
976
	return ret;
L
Linus Torvalds 已提交
977
}
978
EXPORT_SYMBOL(__dev_get_by_flags);
L
Linus Torvalds 已提交
979 980 981 982 983 984

/**
 *	dev_valid_name - check if name is okay for network device
 *	@name: name string
 *
 *	Network device names need to be valid file names to
985 986
 *	to allow sysfs to work.  We also disallow any kind of
 *	whitespace.
L
Linus Torvalds 已提交
987
 */
988
bool dev_valid_name(const char *name)
L
Linus Torvalds 已提交
989
{
990
	if (*name == '\0')
991
		return false;
992
	if (strlen(name) >= IFNAMSIZ)
993
		return false;
994
	if (!strcmp(name, ".") || !strcmp(name, ".."))
995
		return false;
996 997

	while (*name) {
998
		if (*name == '/' || *name == ':' || isspace(*name))
999
			return false;
1000 1001
		name++;
	}
1002
	return true;
L
Linus Torvalds 已提交
1003
}
E
Eric Dumazet 已提交
1004
EXPORT_SYMBOL(dev_valid_name);
L
Linus Torvalds 已提交
1005 1006

/**
1007 1008
 *	__dev_alloc_name - allocate a name for a device
 *	@net: network namespace to allocate the device name in
L
Linus Torvalds 已提交
1009
 *	@name: name format string
1010
 *	@buf:  scratch buffer and result name string
L
Linus Torvalds 已提交
1011 1012
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
S
Stephen Hemminger 已提交
1013 1014 1015 1016 1017 1018
 *	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
1019 1020
 */

1021
static int __dev_alloc_name(struct net *net, const char *name, char *buf)
L
Linus Torvalds 已提交
1022 1023 1024 1025
{
	int i = 0;
	const char *p;
	const int max_netdevices = 8*PAGE_SIZE;
S
Stephen Hemminger 已提交
1026
	unsigned long *inuse;
L
Linus Torvalds 已提交
1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039
	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
1040
		inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC);
L
Linus Torvalds 已提交
1041 1042 1043
		if (!inuse)
			return -ENOMEM;

1044
		for_each_netdev(net, d) {
L
Linus Torvalds 已提交
1045 1046 1047 1048 1049 1050
			if (!sscanf(d->name, name, &i))
				continue;
			if (i < 0 || i >= max_netdevices)
				continue;

			/*  avoid cases where sscanf is not exact inverse of printf */
1051
			snprintf(buf, IFNAMSIZ, name, i);
L
Linus Torvalds 已提交
1052 1053 1054 1055 1056 1057 1058 1059
			if (!strncmp(buf, d->name, IFNAMSIZ))
				set_bit(i, inuse);
		}

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

1060 1061
	if (buf != name)
		snprintf(buf, IFNAMSIZ, name, i);
1062
	if (!__dev_get_by_name(net, buf))
L
Linus Torvalds 已提交
1063 1064 1065 1066 1067 1068 1069 1070 1071
		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;
}

1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
/**
 *	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;

1092 1093
	BUG_ON(!dev_net(dev));
	net = dev_net(dev);
1094 1095 1096 1097 1098
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
}
E
Eric Dumazet 已提交
1099
EXPORT_SYMBOL(dev_alloc_name);
1100

1101 1102 1103
static int dev_alloc_name_ns(struct net *net,
			     struct net_device *dev,
			     const char *name)
1104
{
1105 1106
	char buf[IFNAMSIZ];
	int ret;
1107

1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
	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);
1119

1120 1121 1122
	if (!dev_valid_name(name))
		return -EINVAL;

1123
	if (strchr(name, '%'))
1124
		return dev_alloc_name_ns(net, dev, name);
1125 1126
	else if (__dev_get_by_name(net, name))
		return -EEXIST;
1127 1128
	else if (dev->name != name)
		strlcpy(dev->name, name, IFNAMSIZ);
1129 1130 1131

	return 0;
}
L
Linus Torvalds 已提交
1132 1133 1134 1135 1136 1137 1138 1139 1140

/**
 *	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.
 */
1141
int dev_change_name(struct net_device *dev, const char *newname)
L
Linus Torvalds 已提交
1142
{
1143
	unsigned char old_assign_type;
1144
	char oldname[IFNAMSIZ];
L
Linus Torvalds 已提交
1145
	int err = 0;
1146
	int ret;
1147
	struct net *net;
L
Linus Torvalds 已提交
1148 1149

	ASSERT_RTNL();
1150
	BUG_ON(!dev_net(dev));
L
Linus Torvalds 已提交
1151

1152
	net = dev_net(dev);
L
Linus Torvalds 已提交
1153 1154 1155
	if (dev->flags & IFF_UP)
		return -EBUSY;

1156
	write_seqcount_begin(&devnet_rename_seq);
1157 1158

	if (strncmp(newname, dev->name, IFNAMSIZ) == 0) {
1159
		write_seqcount_end(&devnet_rename_seq);
1160
		return 0;
1161
	}
1162

1163 1164
	memcpy(oldname, dev->name, IFNAMSIZ);

1165
	err = dev_get_valid_name(net, dev, newname);
1166
	if (err < 0) {
1167
		write_seqcount_end(&devnet_rename_seq);
1168
		return err;
1169
	}
L
Linus Torvalds 已提交
1170

1171 1172 1173
	if (oldname[0] && !strchr(oldname, '%'))
		netdev_info(dev, "renamed from %s\n", oldname);

1174 1175 1176
	old_assign_type = dev->name_assign_type;
	dev->name_assign_type = NET_NAME_RENAMED;

1177
rollback:
1178 1179 1180
	ret = device_rename(&dev->dev, dev->name);
	if (ret) {
		memcpy(dev->name, oldname, IFNAMSIZ);
1181
		dev->name_assign_type = old_assign_type;
1182
		write_seqcount_end(&devnet_rename_seq);
1183
		return ret;
1184
	}
1185

1186
	write_seqcount_end(&devnet_rename_seq);
1187

1188 1189
	netdev_adjacent_rename_links(dev, oldname);

1190
	write_lock_bh(&dev_base_lock);
1191
	hlist_del_rcu(&dev->name_hlist);
1192 1193 1194 1195 1196 1197
	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));
1198 1199
	write_unlock_bh(&dev_base_lock);

1200
	ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
1201 1202 1203
	ret = notifier_to_errno(ret);

	if (ret) {
1204 1205
		/* err >= 0 after dev_alloc_name() or stores the first errno */
		if (err >= 0) {
1206
			err = ret;
1207
			write_seqcount_begin(&devnet_rename_seq);
1208
			memcpy(dev->name, oldname, IFNAMSIZ);
1209
			memcpy(oldname, newname, IFNAMSIZ);
1210 1211
			dev->name_assign_type = old_assign_type;
			old_assign_type = NET_NAME_RENAMED;
1212
			goto rollback;
1213
		} else {
1214
			pr_err("%s: name change rollback failed: %d\n",
1215
			       dev->name, ret);
1216 1217
		}
	}
L
Linus Torvalds 已提交
1218 1219 1220 1221

	return err;
}

1222 1223 1224 1225
/**
 *	dev_set_alias - change ifalias of a device
 *	@dev: device
 *	@alias: name up to IFALIASZ
1226
 *	@len: limit of bytes to copy from info
1227 1228 1229 1230 1231
 *
 *	Set ifalias for a device,
 */
int dev_set_alias(struct net_device *dev, const char *alias, size_t len)
{
1232 1233
	char *new_ifalias;

1234 1235 1236 1237 1238
	ASSERT_RTNL();

	if (len >= IFALIASZ)
		return -EINVAL;

1239
	if (!len) {
1240 1241
		kfree(dev->ifalias);
		dev->ifalias = NULL;
1242 1243 1244
		return 0;
	}

1245 1246
	new_ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL);
	if (!new_ifalias)
1247
		return -ENOMEM;
1248
	dev->ifalias = new_ifalias;
1249 1250 1251 1252 1253 1254

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


1255
/**
S
Stephen Hemminger 已提交
1256
 *	netdev_features_change - device changes features
1257 1258 1259 1260 1261 1262
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed features.
 */
void netdev_features_change(struct net_device *dev)
{
1263
	call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev);
1264 1265 1266
}
EXPORT_SYMBOL(netdev_features_change);

L
Linus Torvalds 已提交
1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277
/**
 *	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) {
1278 1279 1280 1281 1282
		struct netdev_notifier_change_info change_info;

		change_info.flags_changed = 0;
		call_netdevice_notifiers_info(NETDEV_CHANGE, dev,
					      &change_info.info);
1283
		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
L
Linus Torvalds 已提交
1284 1285
	}
}
E
Eric Dumazet 已提交
1286
EXPORT_SYMBOL(netdev_state_change);
L
Linus Torvalds 已提交
1287

1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298
/**
 * 	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)
1299
{
1300 1301 1302
	rtnl_lock();
	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, dev);
	rtnl_unlock();
1303
}
1304
EXPORT_SYMBOL(netdev_notify_peers);
1305

1306
static int __dev_open(struct net_device *dev)
L
Linus Torvalds 已提交
1307
{
1308
	const struct net_device_ops *ops = dev->netdev_ops;
1309
	int ret;
L
Linus Torvalds 已提交
1310

1311 1312
	ASSERT_RTNL();

L
Linus Torvalds 已提交
1313 1314 1315
	if (!netif_device_present(dev))
		return -ENODEV;

1316 1317 1318 1319
	/* 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
	 */
1320
	netpoll_poll_disable(dev);
1321

1322 1323 1324 1325 1326
	ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

L
Linus Torvalds 已提交
1327
	set_bit(__LINK_STATE_START, &dev->state);
1328

1329 1330
	if (ops->ndo_validate_addr)
		ret = ops->ndo_validate_addr(dev);
1331

1332 1333
	if (!ret && ops->ndo_open)
		ret = ops->ndo_open(dev);
L
Linus Torvalds 已提交
1334

1335
	netpoll_poll_enable(dev);
1336

1337 1338 1339
	if (ret)
		clear_bit(__LINK_STATE_START, &dev->state);
	else {
L
Linus Torvalds 已提交
1340
		dev->flags |= IFF_UP;
1341
		dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
1342
		dev_activate(dev);
1343
		add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
1344
	}
1345

L
Linus Torvalds 已提交
1346 1347 1348 1349
	return ret;
}

/**
1350 1351
 *	dev_open	- prepare an interface for use.
 *	@dev:	device to open
L
Linus Torvalds 已提交
1352
 *
1353 1354 1355 1356 1357 1358 1359
 *	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
1360
 */
1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371
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;

1372
	rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1373 1374 1375 1376 1377 1378
	call_netdevice_notifiers(NETDEV_UP, dev);

	return ret;
}
EXPORT_SYMBOL(dev_open);

1379
static int __dev_close_many(struct list_head *head)
L
Linus Torvalds 已提交
1380
{
1381
	struct net_device *dev;
1382

1383
	ASSERT_RTNL();
1384 1385
	might_sleep();

1386
	list_for_each_entry(dev, head, close_list) {
1387
		/* Temporarily disable netpoll until the interface is down */
1388
		netpoll_poll_disable(dev);
1389

1390
		call_netdevice_notifiers(NETDEV_GOING_DOWN, dev);
L
Linus Torvalds 已提交
1391

1392
		clear_bit(__LINK_STATE_START, &dev->state);
L
Linus Torvalds 已提交
1393

1394 1395 1396 1397 1398 1399
		/* 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.
		 */
1400
		smp_mb__after_atomic(); /* Commit netif_running(). */
1401
	}
L
Linus Torvalds 已提交
1402

1403
	dev_deactivate_many(head);
1404

1405
	list_for_each_entry(dev, head, close_list) {
1406
		const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
1407

1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418
		/*
		 *	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;
1419
		netpoll_poll_enable(dev);
1420 1421 1422 1423 1424 1425 1426
	}

	return 0;
}

static int __dev_close(struct net_device *dev)
{
1427
	int retval;
1428 1429
	LIST_HEAD(single);

1430
	list_add(&dev->close_list, &single);
1431 1432
	retval = __dev_close_many(&single);
	list_del(&single);
1433

1434
	return retval;
1435 1436
}

1437
int dev_close_many(struct list_head *head, bool unlink)
1438 1439
{
	struct net_device *dev, *tmp;
L
Linus Torvalds 已提交
1440

1441 1442
	/* Remove the devices that don't need to be closed */
	list_for_each_entry_safe(dev, tmp, head, close_list)
1443
		if (!(dev->flags & IFF_UP))
1444
			list_del_init(&dev->close_list);
1445 1446

	__dev_close_many(head);
L
Linus Torvalds 已提交
1447

1448
	list_for_each_entry_safe(dev, tmp, head, close_list) {
1449
		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1450
		call_netdevice_notifiers(NETDEV_DOWN, dev);
1451 1452
		if (unlink)
			list_del_init(&dev->close_list);
1453
	}
1454 1455 1456

	return 0;
}
1457
EXPORT_SYMBOL(dev_close_many);
1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469

/**
 *	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)
{
1470 1471
	if (dev->flags & IFF_UP) {
		LIST_HEAD(single);
L
Linus Torvalds 已提交
1472

1473
		list_add(&dev->close_list, &single);
1474
		dev_close_many(&single, true);
1475 1476
		list_del(&single);
	}
1477
	return 0;
L
Linus Torvalds 已提交
1478
}
E
Eric Dumazet 已提交
1479
EXPORT_SYMBOL(dev_close);
L
Linus Torvalds 已提交
1480 1481


1482 1483 1484 1485 1486 1487 1488 1489 1490 1491
/**
 *	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)
{
1492 1493
	struct net_device *lower_dev;
	struct list_head *iter;
1494

M
Michał Mirosław 已提交
1495 1496
	dev->wanted_features &= ~NETIF_F_LRO;
	netdev_update_features(dev);
1497

1498 1499
	if (unlikely(dev->features & NETIF_F_LRO))
		netdev_WARN(dev, "failed to disable LRO!\n");
1500 1501 1502

	netdev_for_each_lower_dev(dev, lower_dev, iter)
		dev_disable_lro(lower_dev);
1503 1504 1505
}
EXPORT_SYMBOL(dev_disable_lro);

1506 1507 1508 1509 1510 1511 1512 1513
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);
}
1514

1515 1516
static int dev_boot_phase = 1;

L
Linus Torvalds 已提交
1517 1518 1519 1520 1521 1522 1523 1524 1525 1526
/**
 *	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
1527
 *	to the new notifier to allow device to have a race free
L
Linus Torvalds 已提交
1528 1529 1530 1531 1532 1533
 *	view of the network device list.
 */

int register_netdevice_notifier(struct notifier_block *nb)
{
	struct net_device *dev;
1534
	struct net_device *last;
1535
	struct net *net;
L
Linus Torvalds 已提交
1536 1537 1538
	int err;

	rtnl_lock();
1539
	err = raw_notifier_chain_register(&netdev_chain, nb);
1540 1541
	if (err)
		goto unlock;
1542 1543 1544 1545
	if (dev_boot_phase)
		goto unlock;
	for_each_net(net) {
		for_each_netdev(net, dev) {
1546
			err = call_netdevice_notifier(nb, NETDEV_REGISTER, dev);
1547 1548 1549 1550 1551 1552
			err = notifier_to_errno(err);
			if (err)
				goto rollback;

			if (!(dev->flags & IFF_UP))
				continue;
L
Linus Torvalds 已提交
1553

1554
			call_netdevice_notifier(nb, NETDEV_UP, dev);
1555
		}
L
Linus Torvalds 已提交
1556
	}
1557 1558

unlock:
L
Linus Torvalds 已提交
1559 1560
	rtnl_unlock();
	return err;
1561 1562 1563

rollback:
	last = dev;
1564 1565 1566
	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev == last)
1567
				goto outroll;
1568

1569
			if (dev->flags & IFF_UP) {
1570 1571 1572
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1573
			}
1574
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1575 1576
		}
	}
1577

1578
outroll:
1579
	raw_notifier_chain_unregister(&netdev_chain, nb);
1580
	goto unlock;
L
Linus Torvalds 已提交
1581
}
E
Eric Dumazet 已提交
1582
EXPORT_SYMBOL(register_netdevice_notifier);
L
Linus Torvalds 已提交
1583 1584 1585 1586 1587 1588 1589 1590 1591

/**
 *	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.
1592 1593 1594 1595
 *
 * 	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
1596 1597 1598 1599
 */

int unregister_netdevice_notifier(struct notifier_block *nb)
{
1600 1601
	struct net_device *dev;
	struct net *net;
1602 1603 1604
	int err;

	rtnl_lock();
1605
	err = raw_notifier_chain_unregister(&netdev_chain, nb);
1606 1607 1608 1609 1610 1611
	if (err)
		goto unlock;

	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev->flags & IFF_UP) {
1612 1613 1614
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1615
			}
1616
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1617 1618 1619
		}
	}
unlock:
1620 1621
	rtnl_unlock();
	return err;
L
Linus Torvalds 已提交
1622
}
E
Eric Dumazet 已提交
1623
EXPORT_SYMBOL(unregister_netdevice_notifier);
L
Linus Torvalds 已提交
1624

1625 1626 1627 1628 1629 1630 1631 1632 1633 1634
/**
 *	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
1635 1636 1637
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info)
1638 1639 1640 1641 1642 1643
{
	ASSERT_RTNL();
	netdev_notifier_info_init(info, dev);
	return raw_notifier_call_chain(&netdev_chain, val, info);
}

L
Linus Torvalds 已提交
1644 1645 1646
/**
 *	call_netdevice_notifiers - call all network notifier blocks
 *      @val: value passed unmodified to notifier function
1647
 *      @dev: net_device pointer passed unmodified to notifier function
L
Linus Torvalds 已提交
1648 1649
 *
 *	Call all network notifier blocks.  Parameters and return value
1650
 *	are as for raw_notifier_call_chain().
L
Linus Torvalds 已提交
1651 1652
 */

1653
int call_netdevice_notifiers(unsigned long val, struct net_device *dev)
L
Linus Torvalds 已提交
1654
{
1655 1656 1657
	struct netdev_notifier_info info;

	return call_netdevice_notifiers_info(val, dev, &info);
L
Linus Torvalds 已提交
1658
}
1659
EXPORT_SYMBOL(call_netdevice_notifiers);
L
Linus Torvalds 已提交
1660

1661
#ifdef CONFIG_NET_INGRESS
1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676
static struct static_key ingress_needed __read_mostly;

void net_inc_ingress_queue(void)
{
	static_key_slow_inc(&ingress_needed);
}
EXPORT_SYMBOL_GPL(net_inc_ingress_queue);

void net_dec_ingress_queue(void)
{
	static_key_slow_dec(&ingress_needed);
}
EXPORT_SYMBOL_GPL(net_dec_ingress_queue);
#endif

1677
static struct static_key netstamp_needed __read_mostly;
1678
#ifdef HAVE_JUMP_LABEL
1679
/* We are not allowed to call static_key_slow_dec() from irq context
1680
 * If net_disable_timestamp() is called from irq context, defer the
1681
 * static_key_slow_dec() calls.
1682 1683 1684
 */
static atomic_t netstamp_needed_deferred;
#endif
L
Linus Torvalds 已提交
1685 1686 1687

void net_enable_timestamp(void)
{
1688 1689 1690 1691 1692
#ifdef HAVE_JUMP_LABEL
	int deferred = atomic_xchg(&netstamp_needed_deferred, 0);

	if (deferred) {
		while (--deferred)
1693
			static_key_slow_dec(&netstamp_needed);
1694 1695 1696
		return;
	}
#endif
1697
	static_key_slow_inc(&netstamp_needed);
L
Linus Torvalds 已提交
1698
}
E
Eric Dumazet 已提交
1699
EXPORT_SYMBOL(net_enable_timestamp);
L
Linus Torvalds 已提交
1700 1701 1702

void net_disable_timestamp(void)
{
1703 1704 1705 1706 1707 1708
#ifdef HAVE_JUMP_LABEL
	if (in_interrupt()) {
		atomic_inc(&netstamp_needed_deferred);
		return;
	}
#endif
1709
	static_key_slow_dec(&netstamp_needed);
L
Linus Torvalds 已提交
1710
}
E
Eric Dumazet 已提交
1711
EXPORT_SYMBOL(net_disable_timestamp);
L
Linus Torvalds 已提交
1712

E
Eric Dumazet 已提交
1713
static inline void net_timestamp_set(struct sk_buff *skb)
L
Linus Torvalds 已提交
1714
{
1715
	skb->tstamp.tv64 = 0;
1716
	if (static_key_false(&netstamp_needed))
1717
		__net_timestamp(skb);
L
Linus Torvalds 已提交
1718 1719
}

1720
#define net_timestamp_check(COND, SKB)			\
1721
	if (static_key_false(&netstamp_needed)) {		\
1722 1723 1724
		if ((COND) && !(SKB)->tstamp.tv64)	\
			__net_timestamp(SKB);		\
	}						\
E
Eric Dumazet 已提交
1725

1726
bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb)
1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744
{
	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;
}
1745
EXPORT_SYMBOL_GPL(is_skb_forwardable);
1746

H
Herbert Xu 已提交
1747 1748
int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb)
{
1749 1750
	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
	    unlikely(!is_skb_forwardable(dev, skb))) {
H
Herbert Xu 已提交
1751 1752 1753 1754 1755 1756
		atomic_long_inc(&dev->rx_dropped);
		kfree_skb(skb);
		return NET_RX_DROP;
	}

	skb_scrub_packet(skb, true);
1757
	skb->priority = 0;
H
Herbert Xu 已提交
1758
	skb->protocol = eth_type_trans(skb, dev);
1759
	skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
H
Herbert Xu 已提交
1760 1761 1762 1763 1764

	return 0;
}
EXPORT_SYMBOL_GPL(__dev_forward_skb);

1765 1766 1767 1768 1769 1770 1771 1772
/**
 * dev_forward_skb - loopback an skb to another netif
 *
 * @dev: destination network device
 * @skb: buffer to forward
 *
 * return values:
 *	NET_RX_SUCCESS	(no congestion)
1773
 *	NET_RX_DROP     (packet was dropped, but freed)
1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784
 *
 * 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
1785
	return __dev_forward_skb(dev, skb) ?: netif_rx_internal(skb);
1786 1787 1788
}
EXPORT_SYMBOL_GPL(dev_forward_skb);

1789 1790 1791 1792
static inline int deliver_skb(struct sk_buff *skb,
			      struct packet_type *pt_prev,
			      struct net_device *orig_dev)
{
1793 1794
	if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
		return -ENOMEM;
1795 1796 1797 1798
	atomic_inc(&skb->users);
	return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
}

1799 1800
static inline void deliver_ptype_list_skb(struct sk_buff *skb,
					  struct packet_type **pt,
1801 1802
					  struct net_device *orig_dev,
					  __be16 type,
1803 1804 1805 1806 1807 1808 1809 1810
					  struct list_head *ptype_list)
{
	struct packet_type *ptype, *pt_prev = *pt;

	list_for_each_entry_rcu(ptype, ptype_list, list) {
		if (ptype->type != type)
			continue;
		if (pt_prev)
1811
			deliver_skb(skb, pt_prev, orig_dev);
1812 1813 1814 1815 1816
		pt_prev = ptype;
	}
	*pt = pt_prev;
}

1817 1818
static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
{
1819
	if (!ptype->af_packet_priv || !skb->sk)
1820 1821 1822 1823 1824 1825 1826 1827 1828 1829
		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
1830 1831 1832 1833 1834
/*
 *	Support routine. Sends outgoing frames to any network
 *	taps currently in use.
 */

1835
static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
L
Linus Torvalds 已提交
1836 1837
{
	struct packet_type *ptype;
1838 1839
	struct sk_buff *skb2 = NULL;
	struct packet_type *pt_prev = NULL;
1840
	struct list_head *ptype_list = &ptype_all;
1841

L
Linus Torvalds 已提交
1842
	rcu_read_lock();
1843 1844
again:
	list_for_each_entry_rcu(ptype, ptype_list, list) {
L
Linus Torvalds 已提交
1845 1846 1847
		/* Never send packets back to the socket
		 * they originated from - MvS (miquels@drinkel.ow.org)
		 */
1848 1849
		if (skb_loop_sk(ptype, skb))
			continue;
1850

1851 1852 1853 1854 1855
		if (pt_prev) {
			deliver_skb(skb2, pt_prev, skb->dev);
			pt_prev = ptype;
			continue;
		}
L
Linus Torvalds 已提交
1856

1857 1858 1859 1860
		/* need to clone skb, done only once */
		skb2 = skb_clone(skb, GFP_ATOMIC);
		if (!skb2)
			goto out_unlock;
1861

1862
		net_timestamp_set(skb2);
L
Linus Torvalds 已提交
1863

1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875
		/* skb->nh should be correctly
		 * set by sender, so that the second statement is
		 * just protection against buggy protocols.
		 */
		skb_reset_mac_header(skb2);

		if (skb_network_header(skb2) < skb2->data ||
		    skb_network_header(skb2) > skb_tail_pointer(skb2)) {
			net_crit_ratelimited("protocol %04x is buggy, dev %s\n",
					     ntohs(skb2->protocol),
					     dev->name);
			skb_reset_network_header(skb2);
L
Linus Torvalds 已提交
1876
		}
1877 1878 1879 1880 1881 1882 1883 1884 1885

		skb2->transport_header = skb2->network_header;
		skb2->pkt_type = PACKET_OUTGOING;
		pt_prev = ptype;
	}

	if (ptype_list == &ptype_all) {
		ptype_list = &dev->ptype_all;
		goto again;
L
Linus Torvalds 已提交
1886
	}
1887
out_unlock:
1888 1889
	if (pt_prev)
		pt_prev->func(skb2, skb->dev, pt_prev, skb->dev);
L
Linus Torvalds 已提交
1890 1891 1892
	rcu_read_unlock();
}

1893 1894
/**
 * netif_setup_tc - Handle tc mappings on real_num_tx_queues change
1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905
 * @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
1906
static void netif_setup_tc(struct net_device *dev, unsigned int txq)
1907 1908 1909 1910 1911 1912
{
	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) {
1913
		pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n");
1914 1915 1916 1917 1918 1919 1920 1921 1922 1923
		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) {
1924 1925
			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);
1926 1927 1928 1929 1930
			netdev_set_prio_tc_map(dev, i, 0);
		}
	}
}

1931 1932 1933 1934 1935
#ifdef CONFIG_XPS
static DEFINE_MUTEX(xps_map_mutex);
#define xmap_dereference(P)		\
	rcu_dereference_protected((P), lockdep_is_held(&xps_map_mutex))

1936 1937
static struct xps_map *remove_xps_queue(struct xps_dev_maps *dev_maps,
					int cpu, u16 index)
1938
{
1939 1940
	struct xps_map *map = NULL;
	int pos;
1941

1942 1943
	if (dev_maps)
		map = xmap_dereference(dev_maps->cpu_map[cpu]);
1944

1945 1946
	for (pos = 0; map && pos < map->len; pos++) {
		if (map->queues[pos] == index) {
1947 1948 1949
			if (map->len > 1) {
				map->queues[pos] = map->queues[--map->len];
			} else {
1950
				RCU_INIT_POINTER(dev_maps->cpu_map[cpu], NULL);
1951 1952 1953
				kfree_rcu(map, rcu);
				map = NULL;
			}
1954
			break;
1955 1956 1957
		}
	}

1958 1959 1960
	return map;
}

1961
static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
1962 1963
{
	struct xps_dev_maps *dev_maps;
1964
	int cpu, i;
1965 1966 1967 1968 1969 1970 1971 1972 1973
	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) {
1974 1975 1976 1977 1978
		for (i = index; i < dev->num_tx_queues; i++) {
			if (!remove_xps_queue(dev_maps, cpu, i))
				break;
		}
		if (i == dev->num_tx_queues)
1979 1980 1981 1982
			active = true;
	}

	if (!active) {
1983 1984 1985 1986
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

1987 1988 1989 1990
	for (i = index; i < dev->num_tx_queues; i++)
		netdev_queue_numa_node_write(netdev_get_tx_queue(dev, i),
					     NUMA_NO_NODE);

1991 1992 1993 1994
out_no_maps:
	mutex_unlock(&xps_map_mutex);
}

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029
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;
}

2030 2031
int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
			u16 index)
2032
{
2033
	struct xps_dev_maps *dev_maps, *new_dev_maps = NULL;
2034 2035
	struct xps_map *map, *new_map;
	int maps_sz = max_t(unsigned int, XPS_DEV_MAPS_SIZE, L1_CACHE_BYTES);
2036 2037
	int cpu, numa_node_id = -2;
	bool active = false;
2038 2039 2040 2041 2042

	mutex_lock(&xps_map_mutex);

	dev_maps = xmap_dereference(dev->xps_maps);

2043 2044 2045 2046 2047 2048 2049
	/* 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);
2050 2051
		if (!new_dev_maps) {
			mutex_unlock(&xps_map_mutex);
2052
			return -ENOMEM;
2053
		}
2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067

		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;

2068
	for_each_possible_cpu(cpu) {
2069 2070 2071 2072 2073 2074 2075 2076 2077 2078
		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;
2079 2080 2081 2082 2083 2084
#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
2085 2086 2087 2088
		} 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);
2089
		}
2090

2091 2092
	}

2093 2094
	rcu_assign_pointer(dev->xps_maps, new_dev_maps);

2095
	/* Cleanup old maps */
2096 2097 2098 2099 2100 2101 2102
	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);
		}
2103

2104
		kfree_rcu(dev_maps, rcu);
2105 2106
	}

2107 2108
	dev_maps = new_dev_maps;
	active = true;
2109

2110 2111
out_no_new_maps:
	/* update Tx queue numa node */
2112 2113 2114 2115
	netdev_queue_numa_node_write(netdev_get_tx_queue(dev, index),
				     (numa_node_id >= 0) ? numa_node_id :
				     NUMA_NO_NODE);

2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134
	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:
2135 2136 2137 2138
	mutex_unlock(&xps_map_mutex);

	return 0;
error:
2139 2140 2141 2142 2143 2144 2145 2146 2147
	/* 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);
	}

2148 2149 2150 2151 2152 2153 2154 2155
	mutex_unlock(&xps_map_mutex);

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

#endif
2156 2157 2158 2159
/*
 * 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.
 */
2160
int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
2161
{
T
Tom Herbert 已提交
2162 2163
	int rc;

2164 2165
	if (txq < 1 || txq > dev->num_tx_queues)
		return -EINVAL;
2166

2167 2168
	if (dev->reg_state == NETREG_REGISTERED ||
	    dev->reg_state == NETREG_UNREGISTERING) {
2169 2170
		ASSERT_RTNL();

T
Tom Herbert 已提交
2171 2172
		rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
						  txq);
T
Tom Herbert 已提交
2173 2174 2175
		if (rc)
			return rc;

2176 2177 2178
		if (dev->num_tc)
			netif_setup_tc(dev, txq);

2179
		if (txq < dev->real_num_tx_queues) {
2180
			qdisc_reset_all_tx_gt(dev, txq);
2181 2182 2183 2184
#ifdef CONFIG_XPS
			netif_reset_xps_queues_gt(dev, txq);
#endif
		}
2185
	}
2186 2187 2188

	dev->real_num_tx_queues = txq;
	return 0;
2189 2190
}
EXPORT_SYMBOL(netif_set_real_num_tx_queues);
2191

2192
#ifdef CONFIG_SYSFS
2193 2194 2195 2196 2197 2198 2199
/**
 *	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
2200 2201
 *	negative error code.  If called before registration, it always
 *	succeeds.
2202 2203 2204 2205 2206
 */
int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
{
	int rc;

T
Tom Herbert 已提交
2207 2208 2209
	if (rxq < 1 || rxq > dev->num_rx_queues)
		return -EINVAL;

2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224
	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

2225 2226
/**
 * netif_get_num_default_rss_queues - default number of RSS queues
2227 2228 2229 2230
 *
 * This routine should set an upper limit on the number of RSS queues
 * used by default by multiqueue devices.
 */
2231
int netif_get_num_default_rss_queues(void)
2232 2233 2234 2235 2236
{
	return min_t(int, DEFAULT_MAX_NUM_RSS_QUEUES, num_online_cpus());
}
EXPORT_SYMBOL(netif_get_num_default_rss_queues);

2237
static inline void __netif_reschedule(struct Qdisc *q)
2238
{
2239 2240
	struct softnet_data *sd;
	unsigned long flags;
2241

2242
	local_irq_save(flags);
2243
	sd = this_cpu_ptr(&softnet_data);
2244 2245 2246
	q->next_sched = NULL;
	*sd->output_queue_tailp = q;
	sd->output_queue_tailp = &q->next_sched;
2247 2248 2249 2250 2251 2252 2253 2254
	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);
2255 2256 2257
}
EXPORT_SYMBOL(__netif_schedule);

2258 2259 2260 2261 2262
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)
2263
{
2264 2265 2266
	return (struct dev_kfree_skb_cb *)skb->cb;
}

2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313
void netif_schedule_queue(struct netdev_queue *txq)
{
	rcu_read_lock();
	if (!(txq->state & QUEUE_STATE_ANY_XOFF)) {
		struct Qdisc *q = rcu_dereference(txq->qdisc);

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

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

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

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

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

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

2314
void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason)
2315
{
2316
	unsigned long flags;
2317

2318 2319 2320 2321 2322
	if (likely(atomic_read(&skb->users) == 1)) {
		smp_rmb();
		atomic_set(&skb->users, 0);
	} else if (likely(!atomic_dec_and_test(&skb->users))) {
		return;
2323
	}
2324 2325 2326 2327 2328 2329
	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);
2330
}
2331
EXPORT_SYMBOL(__dev_kfree_skb_irq);
2332

2333
void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason)
2334 2335
{
	if (in_irq() || irqs_disabled())
2336
		__dev_kfree_skb_irq(skb, reason);
2337 2338 2339
	else
		dev_kfree_skb(skb);
}
2340
EXPORT_SYMBOL(__dev_kfree_skb_any);
2341 2342


2343 2344 2345 2346 2347 2348
/**
 * netif_device_detach - mark device as removed
 * @dev: network device
 *
 * Mark device as removed from system and therefore no longer available.
 */
2349 2350 2351 2352
void netif_device_detach(struct net_device *dev)
{
	if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2353
		netif_tx_stop_all_queues(dev);
2354 2355 2356 2357
	}
}
EXPORT_SYMBOL(netif_device_detach);

2358 2359 2360 2361 2362 2363
/**
 * netif_device_attach - mark device as attached
 * @dev: network device
 *
 * Mark device as attached from system and restart if needed.
 */
2364 2365 2366 2367
void netif_device_attach(struct net_device *dev)
{
	if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2368
		netif_tx_wake_all_queues(dev);
2369
		__netdev_watchdog_up(dev);
2370 2371 2372 2373
	}
}
EXPORT_SYMBOL(netif_device_attach);

J
Jiri Pirko 已提交
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 2400 2401
/*
 * Returns a Tx hash based on the given packet descriptor a Tx queues' number
 * to be used as a distribution range.
 */
u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
		  unsigned int num_tx_queues)
{
	u32 hash;
	u16 qoffset = 0;
	u16 qcount = num_tx_queues;

	if (skb_rx_queue_recorded(skb)) {
		hash = skb_get_rx_queue(skb);
		while (unlikely(hash >= num_tx_queues))
			hash -= num_tx_queues;
		return hash;
	}

	if (dev->num_tc) {
		u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
		qoffset = dev->tc_to_txq[tc].offset;
		qcount = dev->tc_to_txq[tc].count;
	}

	return (u16) reciprocal_scale(skb_get_hash(skb), qcount) + qoffset;
}
EXPORT_SYMBOL(__skb_tx_hash);

2402 2403
static void skb_warn_bad_offload(const struct sk_buff *skb)
{
2404
	static const netdev_features_t null_features = 0;
2405 2406 2407
	struct net_device *dev = skb->dev;
	const char *driver = "";

2408 2409 2410
	if (!net_ratelimit())
		return;

2411 2412 2413 2414 2415
	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",
2416 2417
	     driver, dev ? &dev->features : &null_features,
	     skb->sk ? &skb->sk->sk_route_caps : &null_features,
2418 2419 2420 2421
	     skb->len, skb->data_len, skb_shinfo(skb)->gso_size,
	     skb_shinfo(skb)->gso_type, skb->ip_summed);
}

L
Linus Torvalds 已提交
2422 2423 2424 2425
/*
 * Invalidate hardware checksum when packet is to be mangled, and
 * complete checksum manually on outgoing path.
 */
2426
int skb_checksum_help(struct sk_buff *skb)
L
Linus Torvalds 已提交
2427
{
2428
	__wsum csum;
2429
	int ret = 0, offset;
L
Linus Torvalds 已提交
2430

2431
	if (skb->ip_summed == CHECKSUM_COMPLETE)
2432 2433 2434
		goto out_set_summed;

	if (unlikely(skb_shinfo(skb)->gso_size)) {
2435 2436
		skb_warn_bad_offload(skb);
		return -EINVAL;
L
Linus Torvalds 已提交
2437 2438
	}

2439 2440 2441 2442 2443 2444 2445 2446 2447
	/* 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;
	}

2448
	offset = skb_checksum_start_offset(skb);
2449 2450 2451 2452 2453 2454 2455 2456
	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
2457 2458 2459 2460 2461
		ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
		if (ret)
			goto out;
	}

2462
	*(__sum16 *)(skb->data + offset) = csum_fold(csum);
2463
out_set_summed:
L
Linus Torvalds 已提交
2464
	skb->ip_summed = CHECKSUM_NONE;
2465
out:
L
Linus Torvalds 已提交
2466 2467
	return ret;
}
E
Eric Dumazet 已提交
2468
EXPORT_SYMBOL(skb_checksum_help);
L
Linus Torvalds 已提交
2469

2470
__be16 skb_network_protocol(struct sk_buff *skb, int *depth)
2471
{
A
Al Viro 已提交
2472
	__be16 type = skb->protocol;
2473

2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484
	/* 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;
	}

2485
	return __vlan_get_protocol(skb, type, depth);
2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497
}

/**
 *	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;
2498 2499
	int vlan_depth = skb->mac_len;
	__be16 type = skb_network_protocol(skb, &vlan_depth);
2500 2501 2502 2503

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

2504
	__skb_pull(skb, vlan_depth);
2505 2506

	rcu_read_lock();
2507
	list_for_each_entry_rcu(ptype, &offload_base, list) {
2508 2509
		if (ptype->type == type && ptype->callbacks.gso_segment) {
			segs = ptype->callbacks.gso_segment(skb, features);
2510 2511 2512 2513 2514
			break;
		}
	}
	rcu_read_unlock();

2515
	__skb_push(skb, skb->data - skb_mac_header(skb));
2516

2517 2518
	return segs;
}
2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550
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);

2551 2552
		err = skb_cow_head(skb, 0);
		if (err < 0)
2553 2554 2555
			return ERR_PTR(err);
	}

2556
	SKB_GSO_CB(skb)->mac_offset = skb_headroom(skb);
2557 2558
	SKB_GSO_CB(skb)->encap_level = 0;

2559 2560 2561 2562 2563
	skb_reset_mac_header(skb);
	skb_reset_mac_len(skb);

	return skb_mac_gso_segment(skb, features);
}
2564
EXPORT_SYMBOL(__skb_gso_segment);
2565

2566 2567 2568 2569 2570
/* Take action when hardware reception checksum errors are detected. */
#ifdef CONFIG_BUG
void netdev_rx_csum_fault(struct net_device *dev)
{
	if (net_ratelimit()) {
2571
		pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>");
2572 2573 2574 2575 2576 2577
		dump_stack();
	}
}
EXPORT_SYMBOL(netdev_rx_csum_fault);
#endif

L
Linus Torvalds 已提交
2578 2579 2580 2581 2582
/* 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.
 */

2583
static int illegal_highdma(struct net_device *dev, struct sk_buff *skb)
L
Linus Torvalds 已提交
2584
{
2585
#ifdef CONFIG_HIGHMEM
L
Linus Torvalds 已提交
2586
	int i;
2587
	if (!(dev->features & NETIF_F_HIGHDMA)) {
2588 2589 2590
		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)))
2591
				return 1;
2592
		}
2593
	}
L
Linus Torvalds 已提交
2594

2595 2596
	if (PCI_DMA_BUS_IS_PHYS) {
		struct device *pdev = dev->dev.parent;
L
Linus Torvalds 已提交
2597

E
Eric Dumazet 已提交
2598 2599
		if (!pdev)
			return 0;
2600
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2601 2602
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			dma_addr_t addr = page_to_phys(skb_frag_page(frag));
2603 2604 2605 2606
			if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask)
				return 1;
		}
	}
2607
#endif
L
Linus Torvalds 已提交
2608 2609 2610
	return 0;
}

2611 2612 2613
/* If MPLS offload request, verify we are testing hardware MPLS features
 * instead of standard features for the netdev.
 */
P
Pravin B Shelar 已提交
2614
#if IS_ENABLED(CONFIG_NET_MPLS_GSO)
2615 2616 2617 2618
static netdev_features_t net_mpls_features(struct sk_buff *skb,
					   netdev_features_t features,
					   __be16 type)
{
2619
	if (eth_p_mpls(type))
2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632
		features &= skb->dev->mpls_features;

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

2633
static netdev_features_t harmonize_features(struct sk_buff *skb,
2634
	netdev_features_t features)
2635
{
2636
	int tmp;
2637 2638 2639 2640
	__be16 type;

	type = skb_network_protocol(skb, &tmp);
	features = net_mpls_features(skb, features, type);
2641

2642
	if (skb->ip_summed != CHECKSUM_NONE &&
2643
	    !can_checksum_protocol(features, type)) {
2644
		features &= ~NETIF_F_ALL_CSUM;
2645
	} else if (illegal_highdma(skb->dev, skb)) {
2646 2647 2648 2649 2650 2651
		features &= ~NETIF_F_SG;
	}

	return features;
}

2652 2653 2654 2655 2656 2657 2658 2659
netdev_features_t passthru_features_check(struct sk_buff *skb,
					  struct net_device *dev,
					  netdev_features_t features)
{
	return features;
}
EXPORT_SYMBOL(passthru_features_check);

2660 2661 2662 2663 2664 2665 2666
static netdev_features_t dflt_features_check(const struct sk_buff *skb,
					     struct net_device *dev,
					     netdev_features_t features)
{
	return vlan_features_check(skb, features);
}

2667
netdev_features_t netif_skb_features(struct sk_buff *skb)
2668
{
2669
	struct net_device *dev = skb->dev;
2670 2671
	netdev_features_t features = dev->features;
	u16 gso_segs = skb_shinfo(skb)->gso_segs;
2672

2673
	if (gso_segs > dev->gso_max_segs || gso_segs < dev->gso_min_segs)
2674 2675
		features &= ~NETIF_F_GSO_MASK;

2676 2677 2678 2679 2680 2681 2682
	/* 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;

2683 2684 2685 2686 2687
	if (skb_vlan_tagged(skb))
		features = netdev_intersect_features(features,
						     dev->vlan_features |
						     NETIF_F_HW_VLAN_CTAG_TX |
						     NETIF_F_HW_VLAN_STAG_TX);
2688

2689 2690 2691
	if (dev->netdev_ops->ndo_features_check)
		features &= dev->netdev_ops->ndo_features_check(skb, dev,
								features);
2692 2693
	else
		features &= dflt_features_check(skb, dev, features);
2694

2695
	return harmonize_features(skb, features);
2696
}
2697
EXPORT_SYMBOL(netif_skb_features);
2698

2699
static int xmit_one(struct sk_buff *skb, struct net_device *dev,
2700
		    struct netdev_queue *txq, bool more)
2701
{
2702 2703
	unsigned int len;
	int rc;
2704

2705
	if (!list_empty(&ptype_all) || !list_empty(&dev->ptype_all))
2706
		dev_queue_xmit_nit(skb, dev);
2707

2708 2709
	len = skb->len;
	trace_net_dev_start_xmit(skb, dev);
2710
	rc = netdev_start_xmit(skb, dev, txq, more);
2711
	trace_net_dev_xmit(skb, rc, dev, len);
E
Eric Dumazet 已提交
2712

2713 2714
	return rc;
}
2715

2716 2717
struct sk_buff *dev_hard_start_xmit(struct sk_buff *first, struct net_device *dev,
				    struct netdev_queue *txq, int *ret)
2718 2719 2720
{
	struct sk_buff *skb = first;
	int rc = NETDEV_TX_OK;
2721

2722 2723
	while (skb) {
		struct sk_buff *next = skb->next;
2724

2725
		skb->next = NULL;
2726
		rc = xmit_one(skb, dev, txq, next != NULL);
2727 2728 2729 2730
		if (unlikely(!dev_xmit_complete(rc))) {
			skb->next = next;
			goto out;
		}
2731

2732 2733 2734 2735
		skb = next;
		if (netif_xmit_stopped(txq) && skb) {
			rc = NETDEV_TX_BUSY;
			break;
2736
		}
2737
	}
2738

2739 2740 2741 2742
out:
	*ret = rc;
	return skb;
}
2743

E
Eric Dumazet 已提交
2744 2745
static struct sk_buff *validate_xmit_vlan(struct sk_buff *skb,
					  netdev_features_t features)
2746
{
2747
	if (skb_vlan_tag_present(skb) &&
2748 2749
	    !vlan_hw_offload_capable(features, skb->vlan_proto))
		skb = __vlan_hwaccel_push_inside(skb);
2750 2751
	return skb;
}
2752

2753
static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev)
2754 2755
{
	netdev_features_t features;
2756

2757 2758
	if (skb->next)
		return skb;
2759

2760 2761 2762 2763
	features = netif_skb_features(skb);
	skb = validate_xmit_vlan(skb, features);
	if (unlikely(!skb))
		goto out_null;
2764

2765
	if (netif_needs_gso(skb, features)) {
2766 2767 2768
		struct sk_buff *segs;

		segs = skb_gso_segment(skb, features);
2769
		if (IS_ERR(segs)) {
2770
			goto out_kfree_skb;
2771 2772 2773
		} else if (segs) {
			consume_skb(skb);
			skb = segs;
2774
		}
2775 2776 2777 2778
	} else {
		if (skb_needs_linearize(skb, features) &&
		    __skb_linearize(skb))
			goto out_kfree_skb;
2779

2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793
		/* If packet is not checksummed and device does not
		 * support checksumming for this protocol, complete
		 * checksumming here.
		 */
		if (skb->ip_summed == CHECKSUM_PARTIAL) {
			if (skb->encapsulation)
				skb_set_inner_transport_header(skb,
							       skb_checksum_start_offset(skb));
			else
				skb_set_transport_header(skb,
							 skb_checksum_start_offset(skb));
			if (!(features & NETIF_F_ALL_CSUM) &&
			    skb_checksum_help(skb))
				goto out_kfree_skb;
2794
		}
2795
	}
2796

2797
	return skb;
2798

2799 2800
out_kfree_skb:
	kfree_skb(skb);
2801 2802 2803
out_null:
	return NULL;
}
2804

2805 2806 2807 2808
struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev)
{
	struct sk_buff *next, *head = NULL, *tail;

2809
	for (; skb != NULL; skb = next) {
2810 2811
		next = skb->next;
		skb->next = NULL;
2812 2813 2814 2815

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

2816
		skb = validate_xmit_skb(skb, dev);
2817 2818
		if (!skb)
			continue;
2819

2820 2821 2822 2823 2824 2825 2826 2827
		if (!head)
			head = skb;
		else
			tail->next = skb;
		/* If skb was segmented, skb->prev points to
		 * the last segment. If not, it still contains skb.
		 */
		tail = skb->prev;
2828 2829
	}
	return head;
2830 2831
}

2832 2833 2834 2835 2836 2837 2838 2839 2840 2841
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)  {
2842
		unsigned int hdr_len;
2843
		u16 gso_segs = shinfo->gso_segs;
2844

2845 2846 2847 2848
		/* mac layer + network layer */
		hdr_len = skb_transport_header(skb) - skb_mac_header(skb);

		/* + transport layer */
2849 2850 2851 2852
		if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)))
			hdr_len += tcp_hdrlen(skb);
		else
			hdr_len += sizeof(struct udphdr);
2853 2854 2855 2856 2857 2858

		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;
2859 2860 2861
	}
}

2862 2863 2864 2865 2866
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
2867
	bool contended;
2868 2869
	int rc;

2870
	qdisc_pkt_len_init(skb);
E
Eric Dumazet 已提交
2871
	qdisc_calculate_pkt_len(skb, q);
2872 2873 2874
	/*
	 * Heuristic to force contended enqueues to serialize on a
	 * separate lock before trying to get qdisc main lock.
Y
Ying Xue 已提交
2875 2876
	 * This permits __QDISC___STATE_RUNNING owner to get the lock more
	 * often and dequeue packets faster.
2877
	 */
E
Eric Dumazet 已提交
2878
	contended = qdisc_is_running(q);
2879 2880 2881
	if (unlikely(contended))
		spin_lock(&q->busylock);

2882 2883 2884 2885 2886
	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) &&
2887
		   qdisc_run_begin(q)) {
2888 2889 2890 2891 2892
		/*
		 * 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.
		 */
2893 2894 2895

		qdisc_bstats_update(q, skb);

2896
		if (sch_direct_xmit(skb, q, dev, txq, root_lock, true)) {
2897 2898 2899 2900
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
2901
			__qdisc_run(q);
2902
		} else
2903
			qdisc_run_end(q);
2904 2905 2906

		rc = NET_XMIT_SUCCESS;
	} else {
E
Eric Dumazet 已提交
2907
		rc = q->enqueue(skb, q) & NET_XMIT_MASK;
2908 2909 2910 2911 2912 2913 2914
		if (qdisc_run_begin(q)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
			__qdisc_run(q);
		}
2915 2916
	}
	spin_unlock(root_lock);
2917 2918
	if (unlikely(contended))
		spin_unlock(&q->busylock);
2919 2920 2921
	return rc;
}

2922
#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2923 2924
static void skb_update_prio(struct sk_buff *skb)
{
I
Igor Maravic 已提交
2925
	struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap);
2926

2927 2928 2929 2930 2931 2932
	if (!skb->priority && skb->sk && map) {
		unsigned int prioidx = skb->sk->sk_cgrp_prioidx;

		if (prioidx < map->priomap_len)
			skb->priority = map->priomap[prioidx];
	}
2933 2934 2935 2936 2937
}
#else
#define skb_update_prio(skb)
#endif

2938 2939 2940
DEFINE_PER_CPU(int, xmit_recursion);
EXPORT_SYMBOL(xmit_recursion);

2941
#define RECURSION_LIMIT 10
2942

2943 2944
/**
 *	dev_loopback_xmit - loop back @skb
2945 2946
 *	@net: network namespace this loopback is happening in
 *	@sk:  sk needed to be a netfilter okfn
2947 2948
 *	@skb: buffer to transmit
 */
2949
int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961
{
	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);

2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003
static inline int get_xps_queue(struct net_device *dev, struct sk_buff *skb)
{
#ifdef CONFIG_XPS
	struct xps_dev_maps *dev_maps;
	struct xps_map *map;
	int queue_index = -1;

	rcu_read_lock();
	dev_maps = rcu_dereference(dev->xps_maps);
	if (dev_maps) {
		map = rcu_dereference(
		    dev_maps->cpu_map[skb->sender_cpu - 1]);
		if (map) {
			if (map->len == 1)
				queue_index = map->queues[0];
			else
				queue_index = map->queues[reciprocal_scale(skb_get_hash(skb),
									   map->len)];
			if (unlikely(queue_index >= dev->real_num_tx_queues))
				queue_index = -1;
		}
	}
	rcu_read_unlock();

	return queue_index;
#else
	return -1;
#endif
}

static u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb)
{
	struct sock *sk = skb->sk;
	int queue_index = sk_tx_queue_get(sk);

	if (queue_index < 0 || skb->ooo_okay ||
	    queue_index >= dev->real_num_tx_queues) {
		int new_index = get_xps_queue(dev, skb);
		if (new_index < 0)
			new_index = skb_tx_hash(dev, skb);

		if (queue_index != new_index && sk &&
3004
		    sk_fullsock(sk) &&
3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040
		    rcu_access_pointer(sk->sk_dst_cache))
			sk_tx_queue_set(sk, new_index);

		queue_index = new_index;
	}

	return queue_index;
}

struct netdev_queue *netdev_pick_tx(struct net_device *dev,
				    struct sk_buff *skb,
				    void *accel_priv)
{
	int queue_index = 0;

#ifdef CONFIG_XPS
	if (skb->sender_cpu == 0)
		skb->sender_cpu = raw_smp_processor_id() + 1;
#endif

	if (dev->real_num_tx_queues != 1) {
		const struct net_device_ops *ops = dev->netdev_ops;
		if (ops->ndo_select_queue)
			queue_index = ops->ndo_select_queue(dev, skb, accel_priv,
							    __netdev_pick_tx);
		else
			queue_index = __netdev_pick_tx(dev, skb);

		if (!accel_priv)
			queue_index = netdev_cap_txqueue(dev, queue_index);
	}

	skb_set_queue_mapping(skb, queue_index);
	return netdev_get_tx_queue(dev, queue_index);
}

3041
/**
3042
 *	__dev_queue_xmit - transmit a buffer
3043
 *	@skb: buffer to transmit
3044
 *	@accel_priv: private data used for L2 forwarding offload
3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066
 *
 *	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
 */
3067
static int __dev_queue_xmit(struct sk_buff *skb, void *accel_priv)
L
Linus Torvalds 已提交
3068 3069
{
	struct net_device *dev = skb->dev;
3070
	struct netdev_queue *txq;
L
Linus Torvalds 已提交
3071 3072 3073
	struct Qdisc *q;
	int rc = -ENOMEM;

3074 3075
	skb_reset_mac_header(skb);

3076 3077 3078
	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_SCHED_TSTAMP))
		__skb_tstamp_tx(skb, NULL, skb->sk, SCM_TSTAMP_SCHED);

3079 3080
	/* Disable soft irqs for various locks below. Also
	 * stops preemption for RCU.
L
Linus Torvalds 已提交
3081
	 */
3082
	rcu_read_lock_bh();
L
Linus Torvalds 已提交
3083

3084 3085
	skb_update_prio(skb);

3086 3087 3088 3089 3090 3091 3092 3093
	/* If device/qdisc don't need skb->dst, release it right now while
	 * its hot in this cpu cache.
	 */
	if (dev->priv_flags & IFF_XMIT_DST_RELEASE)
		skb_dst_drop(skb);
	else
		skb_dst_force(skb);

3094 3095 3096 3097 3098 3099 3100 3101 3102 3103
#ifdef CONFIG_NET_SWITCHDEV
	/* Don't forward if offload device already forwarded */
	if (skb->offload_fwd_mark &&
	    skb->offload_fwd_mark == dev->offload_fwd_mark) {
		consume_skb(skb);
		rc = NET_XMIT_SUCCESS;
		goto out;
	}
#endif

3104
	txq = netdev_pick_tx(dev, skb, accel_priv);
3105
	q = rcu_dereference_bh(txq->qdisc);
3106

L
Linus Torvalds 已提交
3107
#ifdef CONFIG_NET_CLS_ACT
E
Eric Dumazet 已提交
3108
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS);
L
Linus Torvalds 已提交
3109
#endif
3110
	trace_net_dev_queue(skb);
L
Linus Torvalds 已提交
3111
	if (q->enqueue) {
3112
		rc = __dev_xmit_skb(skb, q, dev, txq);
3113
		goto out;
L
Linus Torvalds 已提交
3114 3115 3116 3117 3118
	}

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

H
Herbert Xu 已提交
3119 3120
	   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
3121 3122 3123 3124 3125 3126 3127 3128 3129 3130
	   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 */

3131
		if (txq->xmit_lock_owner != cpu) {
L
Linus Torvalds 已提交
3132

3133 3134 3135
			if (__this_cpu_read(xmit_recursion) > RECURSION_LIMIT)
				goto recursion_alert;

3136 3137 3138 3139
			skb = validate_xmit_skb(skb, dev);
			if (!skb)
				goto drop;

3140
			HARD_TX_LOCK(dev, txq, cpu);
L
Linus Torvalds 已提交
3141

3142
			if (!netif_xmit_stopped(txq)) {
3143
				__this_cpu_inc(xmit_recursion);
3144
				skb = dev_hard_start_xmit(skb, dev, txq, &rc);
3145
				__this_cpu_dec(xmit_recursion);
3146
				if (dev_xmit_complete(rc)) {
3147
					HARD_TX_UNLOCK(dev, txq);
L
Linus Torvalds 已提交
3148 3149 3150
					goto out;
				}
			}
3151
			HARD_TX_UNLOCK(dev, txq);
3152 3153
			net_crit_ratelimited("Virtual device %s asks to queue packet!\n",
					     dev->name);
L
Linus Torvalds 已提交
3154 3155
		} else {
			/* Recursion is detected! It is possible,
3156 3157 3158
			 * unfortunately
			 */
recursion_alert:
3159 3160
			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
					     dev->name);
L
Linus Torvalds 已提交
3161 3162 3163 3164
		}
	}

	rc = -ENETDOWN;
3165
drop:
3166
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
3167

3168
	atomic_long_inc(&dev->tx_dropped);
3169
	kfree_skb_list(skb);
L
Linus Torvalds 已提交
3170 3171
	return rc;
out:
3172
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
3173 3174
	return rc;
}
3175

3176
int dev_queue_xmit(struct sk_buff *skb)
3177 3178 3179
{
	return __dev_queue_xmit(skb, NULL);
}
3180
EXPORT_SYMBOL(dev_queue_xmit);
L
Linus Torvalds 已提交
3181

3182 3183 3184 3185 3186 3187
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
3188 3189 3190 3191 3192

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

3193
int netdev_max_backlog __read_mostly = 1000;
E
Eric Dumazet 已提交
3194 3195
EXPORT_SYMBOL(netdev_max_backlog);

E
Eric Dumazet 已提交
3196
int netdev_tstamp_prequeue __read_mostly = 1;
3197 3198
int netdev_budget __read_mostly = 300;
int weight_p __read_mostly = 64;            /* old backlog weight */
L
Linus Torvalds 已提交
3199

E
Eric Dumazet 已提交
3200 3201 3202 3203 3204 3205 3206 3207
/* 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);
}

3208 3209 3210
#ifdef CONFIG_RPS

/* One global table that all flow-based protocols share. */
E
Eric Dumazet 已提交
3211
struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
3212
EXPORT_SYMBOL(rps_sock_flow_table);
3213 3214
u32 rps_cpu_mask __read_mostly;
EXPORT_SYMBOL(rps_cpu_mask);
3215

3216
struct static_key rps_needed __read_mostly;
3217

3218 3219 3220 3221
static struct rps_dev_flow *
set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
	    struct rps_dev_flow *rflow, u16 next_cpu)
{
3222
	if (next_cpu < nr_cpu_ids) {
3223 3224 3225 3226 3227 3228 3229 3230 3231
#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? */
3232 3233
		if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
		    !(dev->features & NETIF_F_NTUPLE))
3234 3235 3236 3237 3238 3239 3240 3241 3242
			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;
3243
		flow_id = skb_get_hash(skb) & flow_table->mask;
3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255
		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 =
3256
			per_cpu(softnet_data, next_cpu).input_queue_head;
3257 3258
	}

3259
	rflow->cpu = next_cpu;
3260 3261 3262
	return rflow;
}

3263 3264 3265 3266 3267 3268 3269 3270
/*
 * 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)
{
3271 3272
	const struct rps_sock_flow_table *sock_flow_table;
	struct netdev_rx_queue *rxqueue = dev->_rx;
3273
	struct rps_dev_flow_table *flow_table;
3274
	struct rps_map *map;
3275
	int cpu = -1;
3276
	u32 tcpu;
3277
	u32 hash;
3278 3279 3280

	if (skb_rx_queue_recorded(skb)) {
		u16 index = skb_get_rx_queue(skb);
3281

3282 3283 3284 3285 3286
		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);
3287 3288
			goto done;
		}
3289 3290
		rxqueue += index;
	}
3291

3292 3293 3294
	/* Avoid computing hash if RFS/RPS is not active for this rxqueue */

	flow_table = rcu_dereference(rxqueue->rps_flow_table);
E
Eric Dumazet 已提交
3295
	map = rcu_dereference(rxqueue->rps_map);
3296
	if (!flow_table && !map)
3297 3298
		goto done;

3299
	skb_reset_network_header(skb);
3300 3301
	hash = skb_get_hash(skb);
	if (!hash)
3302 3303
		goto done;

T
Tom Herbert 已提交
3304 3305 3306
	sock_flow_table = rcu_dereference(rps_sock_flow_table);
	if (flow_table && sock_flow_table) {
		struct rps_dev_flow *rflow;
3307 3308 3309 3310 3311 3312 3313
		u32 next_cpu;
		u32 ident;

		/* First check into global flow table if there is a match */
		ident = sock_flow_table->ents[hash & sock_flow_table->mask];
		if ((ident ^ hash) & ~rps_cpu_mask)
			goto try_rps;
T
Tom Herbert 已提交
3314

3315 3316 3317 3318 3319
		next_cpu = ident & rps_cpu_mask;

		/* OK, now we know there is a match,
		 * we can look at the local (per receive queue) flow table
		 */
3320
		rflow = &flow_table->flows[hash & flow_table->mask];
T
Tom Herbert 已提交
3321 3322 3323 3324 3325 3326
		tcpu = rflow->cpu;

		/*
		 * 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:
3327
		 *   - Current CPU is unset (>= nr_cpu_ids).
T
Tom Herbert 已提交
3328 3329 3330 3331 3332 3333 3334
		 *   - 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) &&
3335
		    (tcpu >= nr_cpu_ids || !cpu_online(tcpu) ||
T
Tom Herbert 已提交
3336
		     ((int)(per_cpu(softnet_data, tcpu).input_queue_head -
3337 3338
		      rflow->last_qtail)) >= 0)) {
			tcpu = next_cpu;
3339
			rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
3340
		}
3341

3342
		if (tcpu < nr_cpu_ids && cpu_online(tcpu)) {
T
Tom Herbert 已提交
3343 3344 3345 3346 3347 3348
			*rflowp = rflow;
			cpu = tcpu;
			goto done;
		}
	}

3349 3350
try_rps:

T
Tom Herbert 已提交
3351
	if (map) {
3352
		tcpu = map->cpus[reciprocal_scale(hash, map->len)];
T
Tom Herbert 已提交
3353 3354 3355 3356 3357 3358 3359 3360 3361 3362
		if (cpu_online(tcpu)) {
			cpu = tcpu;
			goto done;
		}
	}

done:
	return cpu;
}

3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382
#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;
3383
	unsigned int cpu;
3384 3385 3386 3387 3388 3389

	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);
3390
		if (rflow->filter == filter_id && cpu < nr_cpu_ids &&
3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402
		    ((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
3403
/* Called from hardirq (IPI) context */
E
Eric Dumazet 已提交
3404
static void rps_trigger_softirq(void *data)
T
Tom Herbert 已提交
3405
{
E
Eric Dumazet 已提交
3406 3407
	struct softnet_data *sd = data;

E
Eric Dumazet 已提交
3408
	____napi_schedule(sd, &sd->backlog);
C
Changli Gao 已提交
3409
	sd->received_rps++;
T
Tom Herbert 已提交
3410
}
E
Eric Dumazet 已提交
3411

T
Tom Herbert 已提交
3412
#endif /* CONFIG_RPS */
T
Tom Herbert 已提交
3413

E
Eric Dumazet 已提交
3414 3415 3416 3417 3418 3419 3420 3421
/*
 * 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
3422
	struct softnet_data *mysd = this_cpu_ptr(&softnet_data);
E
Eric Dumazet 已提交
3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434

	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;
}

3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448
#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;

3449
	sd = this_cpu_ptr(&softnet_data);
3450 3451 3452 3453

	rcu_read_lock();
	fl = rcu_dereference(sd->flow_limit);
	if (fl) {
3454
		new_flow = skb_get_hash(skb) & (fl->num_buckets - 1);
3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474
		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
3475 3476 3477 3478
/*
 * 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
3479 3480
static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
			      unsigned int *qtail)
T
Tom Herbert 已提交
3481
{
E
Eric Dumazet 已提交
3482
	struct softnet_data *sd;
T
Tom Herbert 已提交
3483
	unsigned long flags;
3484
	unsigned int qlen;
T
Tom Herbert 已提交
3485

E
Eric Dumazet 已提交
3486
	sd = &per_cpu(softnet_data, cpu);
T
Tom Herbert 已提交
3487 3488 3489

	local_irq_save(flags);

E
Eric Dumazet 已提交
3490
	rps_lock(sd);
3491 3492
	if (!netif_running(skb->dev))
		goto drop;
3493 3494
	qlen = skb_queue_len(&sd->input_pkt_queue);
	if (qlen <= netdev_max_backlog && !skb_flow_limit(skb, qlen)) {
3495
		if (qlen) {
T
Tom Herbert 已提交
3496
enqueue:
E
Eric Dumazet 已提交
3497
			__skb_queue_tail(&sd->input_pkt_queue, skb);
3498
			input_queue_tail_incr_save(sd, qtail);
E
Eric Dumazet 已提交
3499
			rps_unlock(sd);
3500
			local_irq_restore(flags);
T
Tom Herbert 已提交
3501 3502 3503
			return NET_RX_SUCCESS;
		}

3504 3505 3506 3507
		/* 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
3508
			if (!rps_ipi_queued(sd))
E
Eric Dumazet 已提交
3509
				____napi_schedule(sd, &sd->backlog);
T
Tom Herbert 已提交
3510 3511 3512 3513
		}
		goto enqueue;
	}

3514
drop:
C
Changli Gao 已提交
3515
	sd->dropped++;
E
Eric Dumazet 已提交
3516
	rps_unlock(sd);
T
Tom Herbert 已提交
3517 3518 3519

	local_irq_restore(flags);

3520
	atomic_long_inc(&skb->dev->rx_dropped);
T
Tom Herbert 已提交
3521 3522 3523
	kfree_skb(skb);
	return NET_RX_DROP;
}
L
Linus Torvalds 已提交
3524

3525
static int netif_rx_internal(struct sk_buff *skb)
L
Linus Torvalds 已提交
3526
{
3527
	int ret;
L
Linus Torvalds 已提交
3528

3529
	net_timestamp_check(netdev_tstamp_prequeue, skb);
L
Linus Torvalds 已提交
3530

3531
	trace_netif_rx(skb);
E
Eric Dumazet 已提交
3532
#ifdef CONFIG_RPS
3533
	if (static_key_false(&rps_needed)) {
T
Tom Herbert 已提交
3534
		struct rps_dev_flow voidflow, *rflow = &voidflow;
3535 3536
		int cpu;

3537
		preempt_disable();
3538
		rcu_read_lock();
T
Tom Herbert 已提交
3539 3540

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
3541 3542
		if (cpu < 0)
			cpu = smp_processor_id();
T
Tom Herbert 已提交
3543 3544 3545

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

3546
		rcu_read_unlock();
3547
		preempt_enable();
3548 3549
	} else
#endif
T
Tom Herbert 已提交
3550 3551 3552 3553 3554
	{
		unsigned int qtail;
		ret = enqueue_to_backlog(skb, get_cpu(), &qtail);
		put_cpu();
	}
3555
	return ret;
L
Linus Torvalds 已提交
3556
}
3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578

/**
 *	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
3579
EXPORT_SYMBOL(netif_rx);
L
Linus Torvalds 已提交
3580 3581 3582 3583 3584

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

3585 3586
	trace_netif_rx_ni_entry(skb);

L
Linus Torvalds 已提交
3587
	preempt_disable();
3588
	err = netif_rx_internal(skb);
L
Linus Torvalds 已提交
3589 3590 3591 3592 3593 3594 3595 3596 3597 3598
	if (local_softirq_pending())
		do_softirq();
	preempt_enable();

	return err;
}
EXPORT_SYMBOL(netif_rx_ni);

static void net_tx_action(struct softirq_action *h)
{
3599
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
L
Linus Torvalds 已提交
3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612

	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;

3613
			WARN_ON(atomic_read(&skb->users));
3614 3615 3616 3617
			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
3618 3619 3620 3621 3622
			__kfree_skb(skb);
		}
	}

	if (sd->output_queue) {
3623
		struct Qdisc *head;
L
Linus Torvalds 已提交
3624 3625 3626 3627

		local_irq_disable();
		head = sd->output_queue;
		sd->output_queue = NULL;
3628
		sd->output_queue_tailp = &sd->output_queue;
L
Linus Torvalds 已提交
3629 3630 3631
		local_irq_enable();

		while (head) {
3632 3633 3634
			struct Qdisc *q = head;
			spinlock_t *root_lock;

L
Linus Torvalds 已提交
3635 3636
			head = head->next_sched;

3637
			root_lock = qdisc_lock(q);
3638
			if (spin_trylock(root_lock)) {
3639
				smp_mb__before_atomic();
3640 3641
				clear_bit(__QDISC_STATE_SCHED,
					  &q->state);
3642 3643
				qdisc_run(q);
				spin_unlock(root_lock);
L
Linus Torvalds 已提交
3644
			} else {
3645
				if (!test_bit(__QDISC_STATE_DEACTIVATED,
3646
					      &q->state)) {
3647
					__netif_reschedule(q);
3648
				} else {
3649
					smp_mb__before_atomic();
3650 3651 3652
					clear_bit(__QDISC_STATE_SCHED,
						  &q->state);
				}
L
Linus Torvalds 已提交
3653 3654 3655 3656 3657
			}
		}
	}
}

3658 3659
#if (defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)) && \
    (defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE))
3660 3661 3662
/* This hook is defined here for ATM LANE */
int (*br_fdb_test_addr_hook)(struct net_device *dev,
			     unsigned char *addr) __read_mostly;
3663
EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
3664
#endif
L
Linus Torvalds 已提交
3665

3666 3667 3668 3669
static inline struct sk_buff *handle_ing(struct sk_buff *skb,
					 struct packet_type **pt_prev,
					 int *ret, struct net_device *orig_dev)
{
3670
#ifdef CONFIG_NET_CLS_ACT
3671 3672
	struct tcf_proto *cl = rcu_dereference_bh(skb->dev->ingress_cl_list);
	struct tcf_result cl_res;
3673

3674 3675 3676
	/* If there's at least one ingress present somewhere (so
	 * we get here via enabled static key), remaining devices
	 * that are not configured with an ingress qdisc will bail
3677
	 * out here.
3678
	 */
3679
	if (!cl)
3680
		return skb;
3681 3682 3683
	if (*pt_prev) {
		*ret = deliver_skb(skb, *pt_prev, orig_dev);
		*pt_prev = NULL;
L
Linus Torvalds 已提交
3684 3685
	}

3686
	qdisc_skb_cb(skb)->pkt_len = skb->len;
3687
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS);
3688
	qdisc_bstats_cpu_update(cl->q, skb);
3689

3690
	switch (tc_classify(skb, cl, &cl_res, false)) {
3691 3692 3693 3694 3695
	case TC_ACT_OK:
	case TC_ACT_RECLASSIFY:
		skb->tc_index = TC_H_MIN(cl_res.classid);
		break;
	case TC_ACT_SHOT:
3696
		qdisc_qstats_cpu_drop(cl->q);
3697 3698 3699 3700
	case TC_ACT_STOLEN:
	case TC_ACT_QUEUED:
		kfree_skb(skb);
		return NULL;
3701 3702 3703 3704 3705 3706 3707 3708
	case TC_ACT_REDIRECT:
		/* skb_mac_header check was done by cls/act_bpf, so
		 * we can safely push the L2 header back before
		 * redirecting to another netdev
		 */
		__skb_push(skb, skb->mac_len);
		skb_do_redirect(skb);
		return NULL;
3709 3710
	default:
		break;
3711
	}
3712
#endif /* CONFIG_NET_CLS_ACT */
3713 3714
	return skb;
}
L
Linus Torvalds 已提交
3715

3716 3717 3718 3719
/**
 *	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
3720
 *	@rx_handler_data: data pointer that is used by rx handler
3721
 *
3722
 *	Register a receive handler for a device. This handler will then be
3723 3724 3725 3726
 *	called from __netif_receive_skb. A negative errno code is returned
 *	on a failure.
 *
 *	The caller must hold the rtnl_mutex.
3727 3728
 *
 *	For a general description of rx_handler, see enum rx_handler_result.
3729 3730
 */
int netdev_rx_handler_register(struct net_device *dev,
J
Jiri Pirko 已提交
3731 3732
			       rx_handler_func_t *rx_handler,
			       void *rx_handler_data)
3733 3734 3735 3736 3737 3738
{
	ASSERT_RTNL();

	if (dev->rx_handler)
		return -EBUSY;

3739
	/* Note: rx_handler_data must be set before rx_handler */
J
Jiri Pirko 已提交
3740
	rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
3741 3742 3743 3744 3745 3746 3747 3748 3749 3750
	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
3751
 *	Unregister a receive handler from a device.
3752 3753 3754 3755 3756 3757 3758
 *
 *	The caller must hold the rtnl_mutex.
 */
void netdev_rx_handler_unregister(struct net_device *dev)
{

	ASSERT_RTNL();
3759
	RCU_INIT_POINTER(dev->rx_handler, NULL);
3760 3761 3762 3763 3764
	/* 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();
3765
	RCU_INIT_POINTER(dev->rx_handler_data, NULL);
3766 3767 3768
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);

3769 3770 3771 3772 3773 3774 3775
/*
 * 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) {
3776 3777 3778 3779 3780
	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):
3781 3782 3783 3784 3785 3786
		return true;
	default:
		return false;
	}
}

3787 3788 3789
static inline int nf_ingress(struct sk_buff *skb, struct packet_type **pt_prev,
			     int *ret, struct net_device *orig_dev)
{
3790
#ifdef CONFIG_NETFILTER_INGRESS
3791 3792 3793 3794 3795 3796 3797 3798
	if (nf_hook_ingress_active(skb)) {
		if (*pt_prev) {
			*ret = deliver_skb(skb, *pt_prev, orig_dev);
			*pt_prev = NULL;
		}

		return nf_hook_ingress(skb);
	}
3799
#endif /* CONFIG_NETFILTER_INGRESS */
3800 3801 3802
	return 0;
}

3803
static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc)
L
Linus Torvalds 已提交
3804 3805
{
	struct packet_type *ptype, *pt_prev;
3806
	rx_handler_func_t *rx_handler;
D
David S. Miller 已提交
3807
	struct net_device *orig_dev;
3808
	bool deliver_exact = false;
L
Linus Torvalds 已提交
3809
	int ret = NET_RX_DROP;
A
Al Viro 已提交
3810
	__be16 type;
L
Linus Torvalds 已提交
3811

3812
	net_timestamp_check(!netdev_tstamp_prequeue, skb);
3813

3814
	trace_netif_receive_skb(skb);
3815

J
Joe Eykholt 已提交
3816
	orig_dev = skb->dev;
3817

3818
	skb_reset_network_header(skb);
3819 3820
	if (!skb_transport_header_was_set(skb))
		skb_reset_transport_header(skb);
3821
	skb_reset_mac_len(skb);
L
Linus Torvalds 已提交
3822 3823 3824

	pt_prev = NULL;

3825
another_round:
3826
	skb->skb_iif = skb->dev->ifindex;
3827 3828 3829

	__this_cpu_inc(softnet_data.processed);

P
Patrick McHardy 已提交
3830 3831
	if (skb->protocol == cpu_to_be16(ETH_P_8021Q) ||
	    skb->protocol == cpu_to_be16(ETH_P_8021AD)) {
3832
		skb = skb_vlan_untag(skb);
3833
		if (unlikely(!skb))
3834
			goto out;
3835 3836
	}

L
Linus Torvalds 已提交
3837 3838 3839 3840 3841 3842 3843
#ifdef CONFIG_NET_CLS_ACT
	if (skb->tc_verd & TC_NCLS) {
		skb->tc_verd = CLR_TC_NCLS(skb->tc_verd);
		goto ncls;
	}
#endif

3844
	if (pfmemalloc)
3845 3846
		goto skip_taps;

L
Linus Torvalds 已提交
3847
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
3848 3849 3850 3851 3852 3853 3854 3855 3856
		if (pt_prev)
			ret = deliver_skb(skb, pt_prev, orig_dev);
		pt_prev = ptype;
	}

	list_for_each_entry_rcu(ptype, &skb->dev->ptype_all, list) {
		if (pt_prev)
			ret = deliver_skb(skb, pt_prev, orig_dev);
		pt_prev = ptype;
L
Linus Torvalds 已提交
3857 3858
	}

3859
skip_taps:
3860
#ifdef CONFIG_NET_INGRESS
3861 3862 3863
	if (static_key_false(&ingress_needed)) {
		skb = handle_ing(skb, &pt_prev, &ret, orig_dev);
		if (!skb)
3864
			goto out;
3865 3866

		if (nf_ingress(skb, &pt_prev, &ret, orig_dev) < 0)
3867
			goto out;
3868
	}
3869 3870
#endif
#ifdef CONFIG_NET_CLS_ACT
3871
	skb->tc_verd = 0;
L
Linus Torvalds 已提交
3872 3873
ncls:
#endif
3874
	if (pfmemalloc && !skb_pfmemalloc_protocol(skb))
3875 3876
		goto drop;

3877
	if (skb_vlan_tag_present(skb)) {
3878 3879 3880 3881
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3882
		if (vlan_do_receive(&skb))
3883 3884
			goto another_round;
		else if (unlikely(!skb))
3885
			goto out;
3886 3887
	}

3888
	rx_handler = rcu_dereference(skb->dev->rx_handler);
3889 3890 3891 3892 3893
	if (rx_handler) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3894 3895
		switch (rx_handler(&skb)) {
		case RX_HANDLER_CONSUMED:
3896
			ret = NET_RX_SUCCESS;
3897
			goto out;
3898
		case RX_HANDLER_ANOTHER:
3899
			goto another_round;
3900 3901 3902 3903 3904 3905 3906
		case RX_HANDLER_EXACT:
			deliver_exact = true;
		case RX_HANDLER_PASS:
			break;
		default:
			BUG();
		}
3907
	}
L
Linus Torvalds 已提交
3908

3909 3910
	if (unlikely(skb_vlan_tag_present(skb))) {
		if (skb_vlan_tag_get_id(skb))
E
Eric Dumazet 已提交
3911 3912 3913 3914 3915 3916 3917
			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;
	}
3918

3919 3920
	type = skb->protocol;

3921
	/* deliver only exact match when indicated */
3922 3923 3924 3925 3926
	if (likely(!deliver_exact)) {
		deliver_ptype_list_skb(skb, &pt_prev, orig_dev, type,
				       &ptype_base[ntohs(type) &
						   PTYPE_HASH_MASK]);
	}
3927

3928 3929 3930 3931 3932 3933
	deliver_ptype_list_skb(skb, &pt_prev, orig_dev, type,
			       &orig_dev->ptype_specific);

	if (unlikely(skb->dev != orig_dev)) {
		deliver_ptype_list_skb(skb, &pt_prev, orig_dev, type,
				       &skb->dev->ptype_specific);
L
Linus Torvalds 已提交
3934 3935 3936
	}

	if (pt_prev) {
3937
		if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
3938
			goto drop;
3939 3940
		else
			ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3941
	} else {
3942
drop:
3943
		atomic_long_inc(&skb->dev->rx_dropped);
L
Linus Torvalds 已提交
3944 3945 3946 3947 3948 3949 3950
		kfree_skb(skb);
		/* Jamal, now you will not able to escape explaining
		 * me how you were going to use this. :-)
		 */
		ret = NET_RX_DROP;
	}

3951
out:
3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976
	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
3977 3978
	return ret;
}
T
Tom Herbert 已提交
3979

3980
static int netif_receive_skb_internal(struct sk_buff *skb)
T
Tom Herbert 已提交
3981
{
3982 3983
	int ret;

3984
	net_timestamp_check(netdev_tstamp_prequeue, skb);
E
Eric Dumazet 已提交
3985

3986 3987 3988
	if (skb_defer_rx_timestamp(skb))
		return NET_RX_SUCCESS;

3989 3990
	rcu_read_lock();

E
Eric Dumazet 已提交
3991
#ifdef CONFIG_RPS
3992
	if (static_key_false(&rps_needed)) {
E
Eric Dumazet 已提交
3993
		struct rps_dev_flow voidflow, *rflow = &voidflow;
3994
		int cpu = get_rps_cpu(skb->dev, skb, &rflow);
T
Tom Herbert 已提交
3995

E
Eric Dumazet 已提交
3996 3997 3998
		if (cpu >= 0) {
			ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
			rcu_read_unlock();
3999
			return ret;
E
Eric Dumazet 已提交
4000
		}
T
Tom Herbert 已提交
4001
	}
4002
#endif
4003 4004 4005
	ret = __netif_receive_skb(skb);
	rcu_read_unlock();
	return ret;
T
Tom Herbert 已提交
4006
}
4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022

/**
 *	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
 */
4023
int netif_receive_skb(struct sk_buff *skb)
4024 4025 4026 4027 4028
{
	trace_netif_receive_skb_entry(skb);

	return netif_receive_skb_internal(skb);
}
4029
EXPORT_SYMBOL(netif_receive_skb);
L
Linus Torvalds 已提交
4030

E
Eric Dumazet 已提交
4031 4032 4033
/* Network device is going away, flush any packets still pending
 * Called with irqs disabled.
 */
4034
static void flush_backlog(void *arg)
4035
{
4036
	struct net_device *dev = arg;
4037
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
4038 4039
	struct sk_buff *skb, *tmp;

E
Eric Dumazet 已提交
4040
	rps_lock(sd);
4041
	skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
4042
		if (skb->dev == dev) {
E
Eric Dumazet 已提交
4043
			__skb_unlink(skb, &sd->input_pkt_queue);
4044
			kfree_skb(skb);
4045
			input_queue_head_incr(sd);
4046
		}
4047
	}
E
Eric Dumazet 已提交
4048
	rps_unlock(sd);
4049 4050 4051 4052 4053

	skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
		if (skb->dev == dev) {
			__skb_unlink(skb, &sd->process_queue);
			kfree_skb(skb);
4054
			input_queue_head_incr(sd);
4055 4056
		}
	}
4057 4058
}

4059 4060
static int napi_gro_complete(struct sk_buff *skb)
{
4061
	struct packet_offload *ptype;
4062
	__be16 type = skb->protocol;
4063
	struct list_head *head = &offload_base;
4064 4065
	int err = -ENOENT;

4066 4067
	BUILD_BUG_ON(sizeof(struct napi_gro_cb) > sizeof(skb->cb));

4068 4069
	if (NAPI_GRO_CB(skb)->count == 1) {
		skb_shinfo(skb)->gso_size = 0;
4070
		goto out;
4071
	}
4072 4073 4074

	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
4075
		if (ptype->type != type || !ptype->callbacks.gro_complete)
4076 4077
			continue;

4078
		err = ptype->callbacks.gro_complete(skb, 0);
4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089
		break;
	}
	rcu_read_unlock();

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

out:
4090
	return netif_receive_skb_internal(skb);
4091 4092
}

4093 4094 4095 4096 4097
/* 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)
4098
{
4099
	struct sk_buff *skb, *prev = NULL;
4100

4101 4102 4103 4104 4105 4106 4107
	/* 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) {
4108
		skb->next = NULL;
4109 4110 4111 4112 4113

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

		prev = skb->prev;
4114
		napi_gro_complete(skb);
4115
		napi->gro_count--;
4116 4117 4118 4119
	}

	napi->gro_list = NULL;
}
E
Eric Dumazet 已提交
4120
EXPORT_SYMBOL(napi_gro_flush);
4121

E
Eric Dumazet 已提交
4122 4123 4124 4125
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;
4126
	u32 hash = skb_get_hash_raw(skb);
E
Eric Dumazet 已提交
4127 4128 4129 4130

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

4131 4132 4133 4134 4135 4136 4137
		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
4138 4139 4140 4141
		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),
4142
						      skb_mac_header(skb));
E
Eric Dumazet 已提交
4143 4144
		else if (!diffs)
			diffs = memcmp(skb_mac_header(p),
4145
				       skb_mac_header(skb),
E
Eric Dumazet 已提交
4146 4147 4148 4149 4150
				       maclen);
		NAPI_GRO_CB(p)->same_flow = !diffs;
	}
}

4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164
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
4165 4166 4167
	}
}

4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188
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
4189
static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
4190 4191
{
	struct sk_buff **pp = NULL;
4192
	struct packet_offload *ptype;
4193
	__be16 type = skb->protocol;
4194
	struct list_head *head = &offload_base;
H
Herbert Xu 已提交
4195
	int same_flow;
4196
	enum gro_result ret;
4197
	int grow;
4198

4199
	if (!(skb->dev->features & NETIF_F_GRO))
4200 4201
		goto normal;

4202
	if (skb_is_gso(skb) || skb_has_frag_list(skb) || skb->csum_bad)
4203 4204
		goto normal;

E
Eric Dumazet 已提交
4205 4206
	gro_list_prepare(napi, skb);

4207 4208
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
4209
		if (ptype->type != type || !ptype->callbacks.gro_receive)
4210 4211
			continue;

4212
		skb_set_network_header(skb, skb_gro_offset(skb));
4213
		skb_reset_mac_len(skb);
4214 4215
		NAPI_GRO_CB(skb)->same_flow = 0;
		NAPI_GRO_CB(skb)->flush = 0;
H
Herbert Xu 已提交
4216
		NAPI_GRO_CB(skb)->free = 0;
4217
		NAPI_GRO_CB(skb)->udp_mark = 0;
4218
		NAPI_GRO_CB(skb)->gro_remcsum_start = 0;
4219

4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234
		/* Setup for GRO checksum validation */
		switch (skb->ip_summed) {
		case CHECKSUM_COMPLETE:
			NAPI_GRO_CB(skb)->csum = skb->csum;
			NAPI_GRO_CB(skb)->csum_valid = 1;
			NAPI_GRO_CB(skb)->csum_cnt = 0;
			break;
		case CHECKSUM_UNNECESSARY:
			NAPI_GRO_CB(skb)->csum_cnt = skb->csum_level + 1;
			NAPI_GRO_CB(skb)->csum_valid = 0;
			break;
		default:
			NAPI_GRO_CB(skb)->csum_cnt = 0;
			NAPI_GRO_CB(skb)->csum_valid = 0;
		}
4235

4236
		pp = ptype->callbacks.gro_receive(&napi->gro_list, skb);
4237 4238 4239 4240 4241 4242 4243
		break;
	}
	rcu_read_unlock();

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

H
Herbert Xu 已提交
4244
	same_flow = NAPI_GRO_CB(skb)->same_flow;
4245
	ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED;
H
Herbert Xu 已提交
4246

4247 4248 4249 4250 4251 4252
	if (pp) {
		struct sk_buff *nskb = *pp;

		*pp = nskb->next;
		nskb->next = NULL;
		napi_gro_complete(nskb);
4253
		napi->gro_count--;
4254 4255
	}

H
Herbert Xu 已提交
4256
	if (same_flow)
4257 4258
		goto ok;

4259
	if (NAPI_GRO_CB(skb)->flush)
4260 4261
		goto normal;

4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275
	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++;
	}
4276
	NAPI_GRO_CB(skb)->count = 1;
4277
	NAPI_GRO_CB(skb)->age = jiffies;
4278
	NAPI_GRO_CB(skb)->last = skb;
4279
	skb_shinfo(skb)->gso_size = skb_gro_len(skb);
4280 4281
	skb->next = napi->gro_list;
	napi->gro_list = skb;
4282
	ret = GRO_HELD;
4283

4284
pull:
4285 4286 4287
	grow = skb_gro_offset(skb) - skb_headlen(skb);
	if (grow > 0)
		gro_pull_from_frag0(skb, grow);
4288
ok:
4289
	return ret;
4290 4291

normal:
4292 4293
	ret = GRO_NORMAL;
	goto pull;
H
Herbert Xu 已提交
4294
}
4295

4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307
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;
}
4308
EXPORT_SYMBOL(gro_find_receive_by_type);
4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321

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;
}
4322
EXPORT_SYMBOL(gro_find_complete_by_type);
H
Herbert Xu 已提交
4323

R
Rami Rosen 已提交
4324
static gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
H
Herbert Xu 已提交
4325
{
4326 4327
	switch (ret) {
	case GRO_NORMAL:
4328
		if (netif_receive_skb_internal(skb))
4329 4330
			ret = GRO_DROP;
		break;
H
Herbert Xu 已提交
4331

4332
	case GRO_DROP:
H
Herbert Xu 已提交
4333 4334
		kfree_skb(skb);
		break;
4335

4336
	case GRO_MERGED_FREE:
4337 4338 4339 4340
		if (NAPI_GRO_CB(skb)->free == NAPI_GRO_FREE_STOLEN_HEAD)
			kmem_cache_free(skbuff_head_cache, skb);
		else
			__kfree_skb(skb);
4341 4342
		break;

4343 4344 4345
	case GRO_HELD:
	case GRO_MERGED:
		break;
H
Herbert Xu 已提交
4346 4347
	}

4348
	return ret;
4349 4350
}

4351
gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
4352
{
4353
	trace_napi_gro_receive_entry(skb);
4354

4355 4356
	skb_gro_reset_offset(skb);

E
Eric Dumazet 已提交
4357
	return napi_skb_finish(dev_gro_receive(napi, skb), skb);
4358 4359 4360
}
EXPORT_SYMBOL(napi_gro_receive);

S
stephen hemminger 已提交
4361
static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb)
4362
{
4363 4364 4365 4366
	if (unlikely(skb->pfmemalloc)) {
		consume_skb(skb);
		return;
	}
4367
	__skb_pull(skb, skb_headlen(skb));
4368 4369
	/* restore the reserve we had after netdev_alloc_skb_ip_align() */
	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN - skb_headroom(skb));
4370
	skb->vlan_tci = 0;
H
Herbert Xu 已提交
4371
	skb->dev = napi->dev;
A
Andy Gospodarek 已提交
4372
	skb->skb_iif = 0;
4373 4374
	skb->encapsulation = 0;
	skb_shinfo(skb)->gso_type = 0;
4375
	skb->truesize = SKB_TRUESIZE(skb_end_offset(skb));
4376 4377 4378 4379

	napi->skb = skb;
}

4380
struct sk_buff *napi_get_frags(struct napi_struct *napi)
H
Herbert Xu 已提交
4381 4382 4383 4384
{
	struct sk_buff *skb = napi->skb;

	if (!skb) {
4385
		skb = napi_alloc_skb(napi, GRO_MAX_HEAD);
E
Eric Dumazet 已提交
4386
		napi->skb = skb;
4387
	}
4388 4389
	return skb;
}
4390
EXPORT_SYMBOL(napi_get_frags);
4391

4392 4393 4394
static gro_result_t napi_frags_finish(struct napi_struct *napi,
				      struct sk_buff *skb,
				      gro_result_t ret)
4395
{
4396 4397
	switch (ret) {
	case GRO_NORMAL:
4398 4399 4400 4401
	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))
4402
			ret = GRO_DROP;
4403
		break;
H
Herbert Xu 已提交
4404

4405 4406 4407 4408
	case GRO_DROP:
	case GRO_MERGED_FREE:
		napi_reuse_skb(napi, skb);
		break;
4409 4410 4411

	case GRO_MERGED:
		break;
4412
	}
H
Herbert Xu 已提交
4413

4414
	return ret;
H
Herbert Xu 已提交
4415
}
4416

4417 4418 4419 4420
/* 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
4421
static struct sk_buff *napi_frags_skb(struct napi_struct *napi)
4422 4423
{
	struct sk_buff *skb = napi->skb;
4424 4425
	const struct ethhdr *eth;
	unsigned int hlen = sizeof(*eth);
4426 4427 4428

	napi->skb = NULL;

4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442
	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;
4443
	}
4444 4445 4446 4447 4448 4449 4450 4451
	__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;
4452 4453 4454 4455

	return skb;
}

4456
gro_result_t napi_gro_frags(struct napi_struct *napi)
4457
{
4458
	struct sk_buff *skb = napi_frags_skb(napi);
4459 4460

	if (!skb)
4461
		return GRO_DROP;
4462

4463 4464
	trace_napi_gro_frags_entry(skb);

E
Eric Dumazet 已提交
4465
	return napi_frags_finish(napi, skb, dev_gro_receive(napi, skb));
4466
}
H
Herbert Xu 已提交
4467 4468
EXPORT_SYMBOL(napi_gro_frags);

T
Tom Herbert 已提交
4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493
/* Compute the checksum from gro_offset and return the folded value
 * after adding in any pseudo checksum.
 */
__sum16 __skb_gro_checksum_complete(struct sk_buff *skb)
{
	__wsum wsum;
	__sum16 sum;

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

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

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

	return sum;
}
EXPORT_SYMBOL(__skb_gro_checksum_complete);

4494
/*
4495
 * net_rps_action_and_irq_enable sends any pending IPI's for rps.
4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512
 * 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))
4513
				smp_call_function_single_async(remsd->cpu,
4514
							   &remsd->csd);
4515 4516 4517 4518 4519 4520 4521
			remsd = next;
		}
	} else
#endif
		local_irq_enable();
}

E
Eric Dumazet 已提交
4522 4523 4524 4525 4526 4527 4528 4529 4530
static bool sd_has_rps_ipi_waiting(struct softnet_data *sd)
{
#ifdef CONFIG_RPS
	return sd->rps_ipi_list != NULL;
#else
	return false;
#endif
}

4531
static int process_backlog(struct napi_struct *napi, int quota)
L
Linus Torvalds 已提交
4532 4533
{
	int work = 0;
E
Eric Dumazet 已提交
4534
	struct softnet_data *sd = container_of(napi, struct softnet_data, backlog);
L
Linus Torvalds 已提交
4535

4536 4537 4538
	/* Check if we have pending ipi, its better to send them now,
	 * not waiting net_rx_action() end.
	 */
E
Eric Dumazet 已提交
4539
	if (sd_has_rps_ipi_waiting(sd)) {
4540 4541 4542
		local_irq_disable();
		net_rps_action_and_irq_enable(sd);
	}
E
Eric Dumazet 已提交
4543

4544
	napi->weight = weight_p;
4545
	local_irq_disable();
4546
	while (1) {
L
Linus Torvalds 已提交
4547
		struct sk_buff *skb;
4548 4549

		while ((skb = __skb_dequeue(&sd->process_queue))) {
4550
			rcu_read_lock();
4551 4552
			local_irq_enable();
			__netif_receive_skb(skb);
4553
			rcu_read_unlock();
4554
			local_irq_disable();
4555 4556 4557 4558 4559
			input_queue_head_incr(sd);
			if (++work >= quota) {
				local_irq_enable();
				return work;
			}
4560
		}
L
Linus Torvalds 已提交
4561

E
Eric Dumazet 已提交
4562
		rps_lock(sd);
4563
		if (skb_queue_empty(&sd->input_pkt_queue)) {
E
Eric Dumazet 已提交
4564 4565 4566
			/*
			 * Inline a custom version of __napi_complete().
			 * only current cpu owns and manipulates this napi,
4567 4568 4569
			 * and NAPI_STATE_SCHED is the only possible flag set
			 * on backlog.
			 * We can use a plain write instead of clear_bit(),
E
Eric Dumazet 已提交
4570 4571 4572
			 * and we dont need an smp_mb() memory barrier.
			 */
			napi->state = 0;
4573
			rps_unlock(sd);
E
Eric Dumazet 已提交
4574

4575
			break;
4576
		}
4577 4578 4579

		skb_queue_splice_tail_init(&sd->input_pkt_queue,
					   &sd->process_queue);
E
Eric Dumazet 已提交
4580
		rps_unlock(sd);
4581 4582
	}
	local_irq_enable();
L
Linus Torvalds 已提交
4583

4584 4585
	return work;
}
L
Linus Torvalds 已提交
4586

4587 4588
/**
 * __napi_schedule - schedule for receive
4589
 * @n: entry to schedule
4590
 *
4591 4592
 * The entry's receive function will be scheduled to run.
 * Consider using __napi_schedule_irqoff() if hard irqs are masked.
4593
 */
H
Harvey Harrison 已提交
4594
void __napi_schedule(struct napi_struct *n)
4595 4596
{
	unsigned long flags;
L
Linus Torvalds 已提交
4597

4598
	local_irq_save(flags);
4599
	____napi_schedule(this_cpu_ptr(&softnet_data), n);
4600
	local_irq_restore(flags);
L
Linus Torvalds 已提交
4601
}
4602 4603
EXPORT_SYMBOL(__napi_schedule);

4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615
/**
 * __napi_schedule_irqoff - schedule for receive
 * @n: entry to schedule
 *
 * Variant of __napi_schedule() assuming hard irqs are masked
 */
void __napi_schedule_irqoff(struct napi_struct *n)
{
	____napi_schedule(this_cpu_ptr(&softnet_data), n);
}
EXPORT_SYMBOL(__napi_schedule_irqoff);

4616 4617 4618 4619
void __napi_complete(struct napi_struct *n)
{
	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));

E
Eric Dumazet 已提交
4620
	list_del_init(&n->poll_list);
4621
	smp_mb__before_atomic();
4622 4623 4624 4625
	clear_bit(NAPI_STATE_SCHED, &n->state);
}
EXPORT_SYMBOL(__napi_complete);

4626
void napi_complete_done(struct napi_struct *n, int work_done)
4627 4628 4629 4630 4631 4632 4633 4634 4635 4636
{
	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;

4637 4638
	if (n->gro_list) {
		unsigned long timeout = 0;
E
Eric Dumazet 已提交
4639

4640 4641 4642 4643 4644 4645 4646 4647 4648
		if (work_done)
			timeout = n->dev->gro_flush_timeout;

		if (timeout)
			hrtimer_start(&n->timer, ns_to_ktime(timeout),
				      HRTIMER_MODE_REL_PINNED);
		else
			napi_gro_flush(n, false);
	}
E
Eric Dumazet 已提交
4649 4650 4651 4652 4653 4654 4655 4656
	if (likely(list_empty(&n->poll_list))) {
		WARN_ON_ONCE(!test_and_clear_bit(NAPI_STATE_SCHED, &n->state));
	} else {
		/* If n->poll_list is not empty, we need to mask irqs */
		local_irq_save(flags);
		__napi_complete(n);
		local_irq_restore(flags);
	}
4657
}
4658
EXPORT_SYMBOL(napi_complete_done);
4659

E
Eliezer Tamir 已提交
4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711
/* 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);

4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722
static enum hrtimer_restart napi_watchdog(struct hrtimer *timer)
{
	struct napi_struct *napi;

	napi = container_of(timer, struct napi_struct, timer);
	if (napi->gro_list)
		napi_schedule(napi);

	return HRTIMER_NORESTART;
}

4723 4724 4725 4726
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);
4727 4728
	hrtimer_init(&napi->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_PINNED);
	napi->timer.function = napi_watchdog;
4729
	napi->gro_count = 0;
4730
	napi->gro_list = NULL;
H
Herbert Xu 已提交
4731
	napi->skb = NULL;
4732
	napi->poll = poll;
E
Eric Dumazet 已提交
4733 4734 4735
	if (weight > NAPI_POLL_WEIGHT)
		pr_err_once("netif_napi_add() called with weight %d on device %s\n",
			    weight, dev->name);
4736 4737 4738
	napi->weight = weight;
	list_add(&napi->dev_list, &dev->napi_list);
	napi->dev = dev;
H
Herbert Xu 已提交
4739
#ifdef CONFIG_NETPOLL
4740 4741 4742 4743 4744 4745 4746
	spin_lock_init(&napi->poll_lock);
	napi->poll_owner = -1;
#endif
	set_bit(NAPI_STATE_SCHED, &napi->state);
}
EXPORT_SYMBOL(netif_napi_add);

4747 4748 4749 4750 4751 4752 4753
void napi_disable(struct napi_struct *n)
{
	might_sleep();
	set_bit(NAPI_STATE_DISABLE, &n->state);

	while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
		msleep(1);
4754 4755
	while (test_and_set_bit(NAPI_STATE_NPSVC, &n->state))
		msleep(1);
4756 4757 4758 4759 4760 4761 4762

	hrtimer_cancel(&n->timer);

	clear_bit(NAPI_STATE_DISABLE, &n->state);
}
EXPORT_SYMBOL(napi_disable);

4763 4764
void netif_napi_del(struct napi_struct *napi)
{
4765
	list_del_init(&napi->dev_list);
4766
	napi_free_frags(napi);
4767

E
Eric Dumazet 已提交
4768
	kfree_skb_list(napi->gro_list);
4769
	napi->gro_list = NULL;
4770
	napi->gro_count = 0;
4771 4772 4773
}
EXPORT_SYMBOL(netif_napi_del);

4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818
static int napi_poll(struct napi_struct *n, struct list_head *repoll)
{
	void *have;
	int work, weight;

	list_del_init(&n->poll_list);

	have = netpoll_poll_lock(n);

	weight = n->weight;

	/* This NAPI_STATE_SCHED test is for avoiding a race
	 * with netpoll's poll_napi().  Only the entity which
	 * obtains the lock and sees NAPI_STATE_SCHED set will
	 * actually make the ->poll() call.  Therefore we avoid
	 * accidentally calling ->poll() when NAPI is not scheduled.
	 */
	work = 0;
	if (test_bit(NAPI_STATE_SCHED, &n->state)) {
		work = n->poll(n, weight);
		trace_napi_poll(n);
	}

	WARN_ON_ONCE(work > weight);

	if (likely(work < weight))
		goto out_unlock;

	/* Drivers must not modify the NAPI state if they
	 * consume the entire weight.  In such cases this code
	 * still "owns" the NAPI instance and therefore can
	 * move the instance around on the list at-will.
	 */
	if (unlikely(napi_disable_pending(n))) {
		napi_complete(n);
		goto out_unlock;
	}

	if (n->gro_list) {
		/* flush too old packets
		 * If HZ < 1000, flush all packets.
		 */
		napi_gro_flush(n, HZ >= 1000);
	}

4819 4820 4821 4822 4823 4824 4825 4826 4827
	/* Some drivers may have called napi_schedule
	 * prior to exhausting their budget.
	 */
	if (unlikely(!list_empty(&n->poll_list))) {
		pr_warn_once("%s: Budget exhausted after napi rescheduled\n",
			     n->dev ? n->dev->name : "backlog");
		goto out_unlock;
	}

4828 4829 4830 4831 4832 4833 4834 4835
	list_add_tail(&n->poll_list, repoll);

out_unlock:
	netpoll_poll_unlock(have);

	return work;
}

L
Linus Torvalds 已提交
4836 4837
static void net_rx_action(struct softirq_action *h)
{
4838
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
4839
	unsigned long time_limit = jiffies + 2;
4840
	int budget = netdev_budget;
E
Eric Dumazet 已提交
4841 4842
	LIST_HEAD(list);
	LIST_HEAD(repoll);
4843

L
Linus Torvalds 已提交
4844
	local_irq_disable();
E
Eric Dumazet 已提交
4845 4846
	list_splice_init(&sd->poll_list, &list);
	local_irq_enable();
L
Linus Torvalds 已提交
4847

H
Herbert Xu 已提交
4848
	for (;;) {
4849
		struct napi_struct *n;
L
Linus Torvalds 已提交
4850

H
Herbert Xu 已提交
4851 4852 4853 4854 4855 4856
		if (list_empty(&list)) {
			if (!sd_has_rps_ipi_waiting(sd) && list_empty(&repoll))
				return;
			break;
		}

4857 4858 4859
		n = list_first_entry(&list, struct napi_struct, poll_list);
		budget -= napi_poll(n, &repoll);

E
Eric Dumazet 已提交
4860
		/* If softirq window is exhausted then punt.
4861 4862
		 * Allow this to run for 2 jiffies since which will allow
		 * an average latency of 1.5/HZ.
4863
		 */
H
Herbert Xu 已提交
4864 4865 4866 4867 4868
		if (unlikely(budget <= 0 ||
			     time_after_eq(jiffies, time_limit))) {
			sd->time_squeeze++;
			break;
		}
L
Linus Torvalds 已提交
4869
	}
E
Eric Dumazet 已提交
4870 4871 4872 4873 4874 4875 4876 4877 4878

	local_irq_disable();

	list_splice_tail_init(&sd->poll_list, &list);
	list_splice_tail(&repoll, &list);
	list_splice(&list, &sd->poll_list);
	if (!list_empty(&sd->poll_list))
		__raise_softirq_irqoff(NET_RX_SOFTIRQ);

4879
	net_rps_action_and_irq_enable(sd);
L
Linus Torvalds 已提交
4880 4881
}

4882
struct netdev_adjacent {
J
Jiri Pirko 已提交
4883
	struct net_device *dev;
4884 4885

	/* upper master flag, there can only be one master device per list */
J
Jiri Pirko 已提交
4886
	bool master;
4887 4888 4889 4890

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

4891 4892 4893
	/* private field for the users */
	void *private;

J
Jiri Pirko 已提交
4894 4895 4896 4897
	struct list_head list;
	struct rcu_head rcu;
};

4898
static struct netdev_adjacent *__netdev_find_adj(struct net_device *adj_dev,
4899
						 struct list_head *adj_list)
J
Jiri Pirko 已提交
4900
{
4901 4902
	struct netdev_adjacent *adj;

4903
	list_for_each_entry(adj, adj_list, list) {
4904 4905
		if (adj->dev == adj_dev)
			return adj;
J
Jiri Pirko 已提交
4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923
	}
	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();

4924
	return __netdev_find_adj(upper_dev, &dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4925 4926 4927 4928 4929 4930 4931 4932 4933 4934
}
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
4935
static bool netdev_has_any_upper_dev(struct net_device *dev)
J
Jiri Pirko 已提交
4936 4937 4938
{
	ASSERT_RTNL();

4939
	return !list_empty(&dev->all_adj_list.upper);
J
Jiri Pirko 已提交
4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950
}

/**
 * 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)
{
4951
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
4952 4953 4954

	ASSERT_RTNL();

4955
	if (list_empty(&dev->adj_list.upper))
J
Jiri Pirko 已提交
4956 4957
		return NULL;

4958
	upper = list_first_entry(&dev->adj_list.upper,
4959
				 struct netdev_adjacent, list);
J
Jiri Pirko 已提交
4960 4961 4962 4963 4964 4965
	if (likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get);

4966 4967 4968 4969 4970 4971 4972 4973 4974 4975
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);

4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001
/**
 * netdev_upper_get_next_dev_rcu - Get the next dev from upper list
 * @dev: device
 * @iter: list_head ** of the current position
 *
 * Gets the next device from the dev's upper list, starting from iter
 * position. The caller must hold RCU read lock.
 */
struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
						 struct list_head **iter)
{
	struct netdev_adjacent *upper;

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

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

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

	*iter = &upper->list;

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

5002 5003
/**
 * netdev_all_upper_get_next_dev_rcu - Get the next dev from upper list
5004 5005 5006 5007 5008 5009
 * @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.
 */
5010 5011
struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
						     struct list_head **iter)
5012 5013 5014
{
	struct netdev_adjacent *upper;

5015
	WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_rtnl_is_held());
5016 5017 5018

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

5019
	if (&upper->list == &dev->all_adj_list.upper)
5020 5021 5022 5023 5024 5025
		return NULL;

	*iter = &upper->list;

	return upper->dev;
}
5026
EXPORT_SYMBOL(netdev_all_upper_get_next_dev_rcu);
5027

5028 5029 5030 5031 5032 5033 5034 5035 5036
/**
 * 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
5037
 * list will remain unchanged.
5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048
 */
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;

5049
	*iter = lower->list.next;
5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076

	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;

5077
	*iter = &lower->list;
5078 5079 5080 5081 5082

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

5083 5084 5085 5086 5087 5088 5089 5090 5091
/**
 * netdev_lower_get_next - Get the next device from the lower neighbour
 *                         list
 * @dev: device
 * @iter: list_head ** of the current position
 *
 * Gets the next netdev_adjacent from the dev's lower neighbour
 * list, starting from iter position. The caller must hold RTNL lock or
 * its own locking that guarantees that the neighbour lower
5092
 * list will remain unchanged.
5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108
 */
void *netdev_lower_get_next(struct net_device *dev, struct list_head **iter)
{
	struct netdev_adjacent *lower;

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

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

	*iter = &lower->list;

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

5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129
/**
 * 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
5130 5131 5132 5133 5134 5135 5136 5137 5138
/**
 * 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)
{
5139
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
5140

5141
	upper = list_first_or_null_rcu(&dev->adj_list.upper,
5142
				       struct netdev_adjacent, list);
J
Jiri Pirko 已提交
5143 5144 5145 5146 5147 5148
	if (upper && likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get_rcu);

5149
static int netdev_adjacent_sysfs_add(struct net_device *dev,
5150 5151 5152 5153 5154 5155 5156 5157 5158
			      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);
}
5159
static void netdev_adjacent_sysfs_del(struct net_device *dev,
5160 5161 5162 5163 5164 5165 5166 5167 5168
			       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);
}

5169 5170 5171 5172 5173 5174 5175 5176
static inline bool netdev_adjacent_is_neigh_list(struct net_device *dev,
						 struct net_device *adj_dev,
						 struct list_head *dev_list)
{
	return (dev_list == &dev->adj_list.upper ||
		dev_list == &dev->adj_list.lower) &&
		net_eq(dev_net(dev), dev_net(adj_dev));
}
5177

5178 5179
static int __netdev_adjacent_dev_insert(struct net_device *dev,
					struct net_device *adj_dev,
5180
					struct list_head *dev_list,
5181
					void *private, bool master)
5182 5183
{
	struct netdev_adjacent *adj;
5184
	int ret;
5185

5186
	adj = __netdev_find_adj(adj_dev, dev_list);
5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199

	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;
5200
	adj->private = private;
5201
	dev_hold(adj_dev);
5202 5203 5204

	pr_debug("dev_hold for %s, because of link added from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
5205

5206
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list)) {
5207
		ret = netdev_adjacent_sysfs_add(dev, adj_dev, dev_list);
5208 5209 5210 5211
		if (ret)
			goto free_adj;
	}

5212
	/* Ensure that master link is always the first item in list. */
5213 5214 5215 5216
	if (master) {
		ret = sysfs_create_link(&(dev->dev.kobj),
					&(adj_dev->dev.kobj), "master");
		if (ret)
5217
			goto remove_symlinks;
5218

5219
		list_add_rcu(&adj->list, dev_list);
5220
	} else {
5221
		list_add_tail_rcu(&adj->list, dev_list);
5222
	}
5223 5224

	return 0;
5225

5226
remove_symlinks:
5227
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
5228
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
5229 5230
free_adj:
	kfree(adj);
5231
	dev_put(adj_dev);
5232 5233

	return ret;
5234 5235
}

S
stephen hemminger 已提交
5236 5237 5238
static void __netdev_adjacent_dev_remove(struct net_device *dev,
					 struct net_device *adj_dev,
					 struct list_head *dev_list)
5239 5240 5241
{
	struct netdev_adjacent *adj;

5242
	adj = __netdev_find_adj(adj_dev, dev_list);
5243

5244 5245 5246
	if (!adj) {
		pr_err("tried to remove device %s from %s\n",
		       dev->name, adj_dev->name);
5247
		BUG();
5248
	}
5249 5250

	if (adj->ref_nr > 1) {
5251 5252
		pr_debug("%s to %s ref_nr-- = %d\n", dev->name, adj_dev->name,
			 adj->ref_nr-1);
5253 5254 5255 5256
		adj->ref_nr--;
		return;
	}

5257 5258 5259
	if (adj->master)
		sysfs_remove_link(&(dev->dev.kobj), "master");

5260
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
5261
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
5262

5263
	list_del_rcu(&adj->list);
5264 5265
	pr_debug("dev_put for %s, because link removed from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
5266 5267 5268 5269
	dev_put(adj_dev);
	kfree_rcu(adj, rcu);
}

S
stephen hemminger 已提交
5270 5271 5272 5273 5274
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)
5275 5276 5277
{
	int ret;

5278 5279
	ret = __netdev_adjacent_dev_insert(dev, upper_dev, up_list, private,
					   master);
5280 5281 5282
	if (ret)
		return ret;

5283 5284
	ret = __netdev_adjacent_dev_insert(upper_dev, dev, down_list, private,
					   false);
5285
	if (ret) {
5286
		__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
5287 5288 5289 5290 5291 5292
		return ret;
	}

	return 0;
}

S
stephen hemminger 已提交
5293 5294
static int __netdev_adjacent_dev_link(struct net_device *dev,
				      struct net_device *upper_dev)
5295
{
5296 5297 5298
	return __netdev_adjacent_dev_link_lists(dev, upper_dev,
						&dev->all_adj_list.upper,
						&upper_dev->all_adj_list.lower,
5299
						NULL, false);
5300 5301
}

S
stephen hemminger 已提交
5302 5303 5304 5305
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)
5306
{
5307 5308
	__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
	__netdev_adjacent_dev_remove(upper_dev, dev, down_list);
5309 5310
}

S
stephen hemminger 已提交
5311 5312
static void __netdev_adjacent_dev_unlink(struct net_device *dev,
					 struct net_device *upper_dev)
5313
{
5314 5315 5316 5317 5318
	__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
5319 5320 5321
static int __netdev_adjacent_dev_link_neighbour(struct net_device *dev,
						struct net_device *upper_dev,
						void *private, bool master)
5322 5323 5324 5325 5326 5327 5328 5329 5330
{
	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,
5331
					       private, master);
5332 5333 5334 5335 5336 5337
	if (ret) {
		__netdev_adjacent_dev_unlink(dev, upper_dev);
		return ret;
	}

	return 0;
5338 5339
}

S
stephen hemminger 已提交
5340 5341
static void __netdev_adjacent_dev_unlink_neighbour(struct net_device *dev,
						   struct net_device *upper_dev)
5342 5343 5344 5345 5346 5347
{
	__netdev_adjacent_dev_unlink(dev, upper_dev);
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev,
					   &dev->adj_list.upper,
					   &upper_dev->adj_list.lower);
}
5348

J
Jiri Pirko 已提交
5349
static int __netdev_upper_dev_link(struct net_device *dev,
5350 5351
				   struct net_device *upper_dev, bool master,
				   void *private)
J
Jiri Pirko 已提交
5352
{
5353
	struct netdev_notifier_changeupper_info changeupper_info;
5354 5355
	struct netdev_adjacent *i, *j, *to_i, *to_j;
	int ret = 0;
J
Jiri Pirko 已提交
5356 5357 5358 5359 5360 5361 5362

	ASSERT_RTNL();

	if (dev == upper_dev)
		return -EBUSY;

	/* To prevent loops, check if dev is not upper device to upper_dev. */
5363
	if (__netdev_find_adj(dev, &upper_dev->all_adj_list.upper))
J
Jiri Pirko 已提交
5364 5365
		return -EBUSY;

5366
	if (__netdev_find_adj(upper_dev, &dev->adj_list.upper))
J
Jiri Pirko 已提交
5367 5368 5369 5370 5371
		return -EEXIST;

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

5372 5373 5374 5375
	changeupper_info.upper_dev = upper_dev;
	changeupper_info.master = master;
	changeupper_info.linking = true;

5376 5377 5378 5379 5380 5381
	ret = call_netdevice_notifiers_info(NETDEV_PRECHANGEUPPER, dev,
					    &changeupper_info.info);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

5382 5383
	ret = __netdev_adjacent_dev_link_neighbour(dev, upper_dev, private,
						   master);
5384 5385
	if (ret)
		return ret;
J
Jiri Pirko 已提交
5386

5387
	/* Now that we linked these devs, make all the upper_dev's
5388
	 * all_adj_list.upper visible to every dev's all_adj_list.lower an
5389 5390 5391
	 * versa, and don't forget the devices itself. All of these
	 * links are non-neighbours.
	 */
5392 5393 5394 5395
	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);
5396 5397 5398 5399 5400 5401 5402
			ret = __netdev_adjacent_dev_link(i->dev, j->dev);
			if (ret)
				goto rollback_mesh;
		}
	}

	/* add dev to every upper_dev's upper device */
5403 5404 5405
	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);
5406 5407 5408 5409 5410 5411
		ret = __netdev_adjacent_dev_link(dev, i->dev);
		if (ret)
			goto rollback_upper_mesh;
	}

	/* add upper_dev to every dev's lower device */
5412 5413 5414
	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);
5415 5416 5417 5418
		ret = __netdev_adjacent_dev_link(i->dev, upper_dev);
		if (ret)
			goto rollback_lower_mesh;
	}
J
Jiri Pirko 已提交
5419

5420 5421
	call_netdevice_notifiers_info(NETDEV_CHANGEUPPER, dev,
				      &changeupper_info.info);
J
Jiri Pirko 已提交
5422
	return 0;
5423 5424 5425

rollback_lower_mesh:
	to_i = i;
5426
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
5427 5428 5429 5430 5431 5432 5433 5434 5435
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);
	}

	i = NULL;

rollback_upper_mesh:
	to_i = i;
5436
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list) {
5437 5438 5439 5440 5441 5442 5443 5444 5445 5446
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(dev, i->dev);
	}

	i = j = NULL;

rollback_mesh:
	to_i = i;
	to_j = j;
5447 5448
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list) {
5449 5450 5451 5452 5453 5454 5455 5456
			if (i == to_i && j == to_j)
				break;
			__netdev_adjacent_dev_unlink(i->dev, j->dev);
		}
		if (i == to_i)
			break;
	}

5457
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
5458 5459

	return ret;
J
Jiri Pirko 已提交
5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474
}

/**
 * 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)
{
5475
	return __netdev_upper_dev_link(dev, upper_dev, false, NULL);
J
Jiri Pirko 已提交
5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492
}
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)
{
5493
	return __netdev_upper_dev_link(dev, upper_dev, true, NULL);
J
Jiri Pirko 已提交
5494 5495 5496
}
EXPORT_SYMBOL(netdev_master_upper_dev_link);

5497 5498 5499 5500 5501 5502 5503 5504
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
5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515
/**
 * 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)
{
5516
	struct netdev_notifier_changeupper_info changeupper_info;
5517
	struct netdev_adjacent *i, *j;
J
Jiri Pirko 已提交
5518 5519
	ASSERT_RTNL();

5520 5521 5522 5523
	changeupper_info.upper_dev = upper_dev;
	changeupper_info.master = netdev_master_upper_dev_get(dev) == upper_dev;
	changeupper_info.linking = false;

5524 5525 5526
	call_netdevice_notifiers_info(NETDEV_PRECHANGEUPPER, dev,
				      &changeupper_info.info);

5527
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
5528 5529 5530 5531 5532

	/* 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.
	 */
5533 5534
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list)
5535 5536 5537 5538 5539
			__netdev_adjacent_dev_unlink(i->dev, j->dev);

	/* remove also the devices itself from lower/upper device
	 * list
	 */
5540
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
5541 5542
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);

5543
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list)
5544 5545
		__netdev_adjacent_dev_unlink(dev, i->dev);

5546 5547
	call_netdevice_notifiers_info(NETDEV_CHANGEUPPER, dev,
				      &changeupper_info.info);
J
Jiri Pirko 已提交
5548 5549 5550
}
EXPORT_SYMBOL(netdev_upper_dev_unlink);

5551 5552 5553
/**
 * netdev_bonding_info_change - Dispatch event about slave change
 * @dev: device
5554
 * @bonding_info: info to dispatch
5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570
 *
 * Send NETDEV_BONDING_INFO to netdev notifiers with info.
 * The caller must hold the RTNL lock.
 */
void netdev_bonding_info_change(struct net_device *dev,
				struct netdev_bonding_info *bonding_info)
{
	struct netdev_notifier_bonding_info	info;

	memcpy(&info.bonding_info, bonding_info,
	       sizeof(struct netdev_bonding_info));
	call_netdevice_notifiers_info(NETDEV_BONDING_INFO, dev,
				      &info.info);
}
EXPORT_SYMBOL(netdev_bonding_info_change);

E
Eric Dumazet 已提交
5571
static void netdev_adjacent_add_links(struct net_device *dev)
5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595
{
	struct netdev_adjacent *iter;

	struct net *net = dev_net(dev);

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

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

E
Eric Dumazet 已提交
5596
static void netdev_adjacent_del_links(struct net_device *dev)
5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620
{
	struct netdev_adjacent *iter;

	struct net *net = dev_net(dev);

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

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

5621
void netdev_adjacent_rename_links(struct net_device *dev, char *oldname)
5622
{
5623
	struct netdev_adjacent *iter;
5624

5625 5626
	struct net *net = dev_net(dev);

5627
	list_for_each_entry(iter, &dev->adj_list.upper, list) {
5628 5629
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
5630 5631 5632 5633 5634
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.lower);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.lower);
	}
5635

5636
	list_for_each_entry(iter, &dev->adj_list.lower, list) {
5637 5638
		if (!net_eq(net,dev_net(iter->dev)))
			continue;
5639 5640 5641 5642 5643
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.upper);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.upper);
	}
5644 5645 5646 5647 5648 5649 5650 5651 5652
}

void *netdev_lower_dev_get_private(struct net_device *dev,
				   struct net_device *lower_dev)
{
	struct netdev_adjacent *lower;

	if (!lower_dev)
		return NULL;
5653
	lower = __netdev_find_adj(lower_dev, &dev->adj_list.lower);
5654 5655 5656 5657 5658 5659 5660
	if (!lower)
		return NULL;

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

5661 5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684

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

	ASSERT_RTNL();

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

	if (type_check(dev))
		max_nest++;

	return max_nest;
}
EXPORT_SYMBOL(dev_get_nest_level);

5685 5686
static void dev_change_rx_flags(struct net_device *dev, int flags)
{
5687 5688
	const struct net_device_ops *ops = dev->netdev_ops;

5689
	if (ops->ndo_change_rx_flags)
5690
		ops->ndo_change_rx_flags(dev, flags);
5691 5692
}

5693
static int __dev_set_promiscuity(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5694
{
5695
	unsigned int old_flags = dev->flags;
5696 5697
	kuid_t uid;
	kgid_t gid;
L
Linus Torvalds 已提交
5698

5699 5700
	ASSERT_RTNL();

5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711
	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;
5712 5713
			pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n",
				dev->name);
5714 5715 5716
			return -EOVERFLOW;
		}
	}
5717
	if (dev->flags != old_flags) {
5718 5719 5720
		pr_info("device %s %s promiscuous mode\n",
			dev->name,
			dev->flags & IFF_PROMISC ? "entered" : "left");
5721 5722
		if (audit_enabled) {
			current_uid_gid(&uid, &gid);
5723 5724 5725 5726 5727
			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),
5728
				from_kuid(&init_user_ns, audit_get_loginuid(current)),
5729 5730
				from_kuid(&init_user_ns, uid),
				from_kgid(&init_user_ns, gid),
5731
				audit_get_sessionid(current));
5732
		}
5733

5734
		dev_change_rx_flags(dev, IFF_PROMISC);
L
Linus Torvalds 已提交
5735
	}
5736 5737
	if (notify)
		__dev_notify_flags(dev, old_flags, IFF_PROMISC);
5738
	return 0;
L
Linus Torvalds 已提交
5739 5740
}

5741 5742 5743 5744 5745 5746 5747 5748 5749
/**
 *	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.
5750
 *	Return 0 if successful or a negative errno code on error.
5751
 */
5752
int dev_set_promiscuity(struct net_device *dev, int inc)
5753
{
5754
	unsigned int old_flags = dev->flags;
5755
	int err;
5756

5757
	err = __dev_set_promiscuity(dev, inc, true);
5758
	if (err < 0)
5759
		return err;
5760 5761
	if (dev->flags != old_flags)
		dev_set_rx_mode(dev);
5762
	return err;
5763
}
E
Eric Dumazet 已提交
5764
EXPORT_SYMBOL(dev_set_promiscuity);
5765

5766
static int __dev_set_allmulti(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
5767
{
5768
	unsigned int old_flags = dev->flags, old_gflags = dev->gflags;
L
Linus Torvalds 已提交
5769

5770 5771
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5772
	dev->flags |= IFF_ALLMULTI;
5773 5774 5775 5776 5777 5778 5779 5780 5781 5782
	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;
5783 5784
			pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n",
				dev->name);
5785 5786 5787
			return -EOVERFLOW;
		}
	}
5788
	if (dev->flags ^ old_flags) {
5789
		dev_change_rx_flags(dev, IFF_ALLMULTI);
5790
		dev_set_rx_mode(dev);
5791 5792 5793
		if (notify)
			__dev_notify_flags(dev, old_flags,
					   dev->gflags ^ old_gflags);
5794
	}
5795
	return 0;
5796
}
5797 5798 5799 5800 5801 5802 5803 5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814

/**
 *	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
5815
EXPORT_SYMBOL(dev_set_allmulti);
5816 5817 5818 5819

/*
 *	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
5820
 *	filtering it is put in promiscuous mode while unicast addresses
5821 5822 5823 5824
 *	are present.
 */
void __dev_set_rx_mode(struct net_device *dev)
{
5825 5826
	const struct net_device_ops *ops = dev->netdev_ops;

5827 5828 5829 5830 5831
	/* dev_open will call this function so the list will stay sane. */
	if (!(dev->flags&IFF_UP))
		return;

	if (!netif_device_present(dev))
5832
		return;
5833

5834
	if (!(dev->priv_flags & IFF_UNICAST_FLT)) {
5835 5836 5837
		/* Unicast addresses changes may only happen under the rtnl,
		 * therefore calling __dev_set_promiscuity here is safe.
		 */
5838
		if (!netdev_uc_empty(dev) && !dev->uc_promisc) {
5839
			__dev_set_promiscuity(dev, 1, false);
5840
			dev->uc_promisc = true;
5841
		} else if (netdev_uc_empty(dev) && dev->uc_promisc) {
5842
			__dev_set_promiscuity(dev, -1, false);
5843
			dev->uc_promisc = false;
5844 5845
		}
	}
5846 5847 5848

	if (ops->ndo_set_rx_mode)
		ops->ndo_set_rx_mode(dev);
5849 5850 5851 5852
}

void dev_set_rx_mode(struct net_device *dev)
{
5853
	netif_addr_lock_bh(dev);
5854
	__dev_set_rx_mode(dev);
5855
	netif_addr_unlock_bh(dev);
L
Linus Torvalds 已提交
5856 5857
}

5858 5859 5860 5861 5862 5863
/**
 *	dev_get_flags - get flags reported to userspace
 *	@dev: device
 *
 *	Get the combination of flag bits exported through APIs to userspace.
 */
5864
unsigned int dev_get_flags(const struct net_device *dev)
L
Linus Torvalds 已提交
5865
{
5866
	unsigned int flags;
L
Linus Torvalds 已提交
5867 5868 5869

	flags = (dev->flags & ~(IFF_PROMISC |
				IFF_ALLMULTI |
S
Stefan Rompf 已提交
5870 5871 5872
				IFF_RUNNING |
				IFF_LOWER_UP |
				IFF_DORMANT)) |
L
Linus Torvalds 已提交
5873 5874 5875
		(dev->gflags & (IFF_PROMISC |
				IFF_ALLMULTI));

S
Stefan Rompf 已提交
5876 5877 5878 5879 5880 5881 5882 5883
	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
5884 5885 5886

	return flags;
}
E
Eric Dumazet 已提交
5887
EXPORT_SYMBOL(dev_get_flags);
L
Linus Torvalds 已提交
5888

5889
int __dev_change_flags(struct net_device *dev, unsigned int flags)
L
Linus Torvalds 已提交
5890
{
5891
	unsigned int old_flags = dev->flags;
5892
	int ret;
L
Linus Torvalds 已提交
5893

5894 5895
	ASSERT_RTNL();

L
Linus Torvalds 已提交
5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909
	/*
	 *	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.
	 */

5910 5911
	if ((old_flags ^ flags) & IFF_MULTICAST)
		dev_change_rx_flags(dev, IFF_MULTICAST);
5912

5913
	dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5914 5915 5916 5917 5918 5919 5920 5921

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

	ret = 0;
5922
	if ((old_flags ^ flags) & IFF_UP)
5923
		ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev);
L
Linus Torvalds 已提交
5924 5925

	if ((flags ^ dev->gflags) & IFF_PROMISC) {
E
Eric Dumazet 已提交
5926
		int inc = (flags & IFF_PROMISC) ? 1 : -1;
5927
		unsigned int old_flags = dev->flags;
E
Eric Dumazet 已提交
5928

L
Linus Torvalds 已提交
5929
		dev->gflags ^= IFF_PROMISC;
5930 5931 5932 5933

		if (__dev_set_promiscuity(dev, inc, false) >= 0)
			if (dev->flags != old_flags)
				dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
5934 5935 5936 5937 5938 5939 5940
	}

	/* 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
5941 5942
		int inc = (flags & IFF_ALLMULTI) ? 1 : -1;

L
Linus Torvalds 已提交
5943
		dev->gflags ^= IFF_ALLMULTI;
5944
		__dev_set_allmulti(dev, inc, false);
L
Linus Torvalds 已提交
5945 5946
	}

5947 5948 5949
	return ret;
}

5950 5951
void __dev_notify_flags(struct net_device *dev, unsigned int old_flags,
			unsigned int gchanges)
5952 5953 5954
{
	unsigned int changes = dev->flags ^ old_flags;

5955
	if (gchanges)
5956
		rtmsg_ifinfo(RTM_NEWLINK, dev, gchanges, GFP_ATOMIC);
5957

5958 5959 5960 5961 5962 5963 5964 5965
	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 &&
5966 5967 5968 5969 5970 5971 5972
	    (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);
	}
5973 5974 5975 5976 5977 5978 5979 5980 5981 5982
}

/**
 *	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.
 */
5983
int dev_change_flags(struct net_device *dev, unsigned int flags)
5984
{
5985
	int ret;
5986
	unsigned int changes, old_flags = dev->flags, old_gflags = dev->gflags;
5987 5988 5989 5990 5991

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

5992
	changes = (old_flags ^ dev->flags) | (old_gflags ^ dev->gflags);
5993
	__dev_notify_flags(dev, old_flags, changes);
L
Linus Torvalds 已提交
5994 5995
	return ret;
}
E
Eric Dumazet 已提交
5996
EXPORT_SYMBOL(dev_change_flags);
L
Linus Torvalds 已提交
5997

5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008
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;
}

6009 6010 6011 6012 6013 6014 6015
/**
 *	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
6016 6017
int dev_set_mtu(struct net_device *dev, int new_mtu)
{
6018
	int err, orig_mtu;
L
Linus Torvalds 已提交
6019 6020 6021 6022 6023 6024 6025 6026 6027 6028 6029

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

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

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

6030 6031 6032 6033
	err = call_netdevice_notifiers(NETDEV_PRECHANGEMTU, dev);
	err = notifier_to_errno(err);
	if (err)
		return err;
6034

6035 6036
	orig_mtu = dev->mtu;
	err = __dev_set_mtu(dev, new_mtu);
6037

6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048
	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
6049 6050
	return err;
}
E
Eric Dumazet 已提交
6051
EXPORT_SYMBOL(dev_set_mtu);
L
Linus Torvalds 已提交
6052

6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063
/**
 *	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);

6064 6065 6066 6067 6068 6069 6070
/**
 *	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
6071 6072
int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa)
{
6073
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
6074 6075
	int err;

6076
	if (!ops->ndo_set_mac_address)
L
Linus Torvalds 已提交
6077 6078 6079 6080 6081
		return -EOPNOTSUPP;
	if (sa->sa_family != dev->type)
		return -EINVAL;
	if (!netif_device_present(dev))
		return -ENODEV;
6082
	err = ops->ndo_set_mac_address(dev, sa);
6083 6084
	if (err)
		return err;
J
Jiri Pirko 已提交
6085
	dev->addr_assign_type = NET_ADDR_SET;
6086
	call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
6087
	add_device_randomness(dev->dev_addr, dev->addr_len);
6088
	return 0;
L
Linus Torvalds 已提交
6089
}
E
Eric Dumazet 已提交
6090
EXPORT_SYMBOL(dev_set_mac_address);
L
Linus Torvalds 已提交
6091

J
Jiri Pirko 已提交
6092 6093 6094
/**
 *	dev_change_carrier - Change device carrier
 *	@dev: device
6095
 *	@new_carrier: new value
J
Jiri Pirko 已提交
6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110
 *
 *	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);

6111 6112 6113 6114 6115 6116 6117 6118
/**
 *	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,
6119
			 struct netdev_phys_item_id *ppid)
6120 6121 6122 6123 6124 6125 6126 6127 6128
{
	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);

6129 6130 6131 6132 6133 6134 6135 6136 6137 6138 6139 6140 6141 6142 6143 6144 6145 6146
/**
 *	dev_get_phys_port_name - Get device physical port name
 *	@dev: device
 *	@name: port name
 *
 *	Get device physical port name
 */
int dev_get_phys_port_name(struct net_device *dev,
			   char *name, size_t len)
{
	const struct net_device_ops *ops = dev->netdev_ops;

	if (!ops->ndo_get_phys_port_name)
		return -EOPNOTSUPP;
	return ops->ndo_get_phys_port_name(dev, name, len);
}
EXPORT_SYMBOL(dev_get_phys_port_name);

6147 6148 6149 6150 6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163 6164 6165 6166
/**
 *	dev_change_proto_down - update protocol port state information
 *	@dev: device
 *	@proto_down: new value
 *
 *	This info can be used by switch drivers to set the phys state of the
 *	port.
 */
int dev_change_proto_down(struct net_device *dev, bool proto_down)
{
	const struct net_device_ops *ops = dev->netdev_ops;

	if (!ops->ndo_change_proto_down)
		return -EOPNOTSUPP;
	if (!netif_device_present(dev))
		return -ENODEV;
	return ops->ndo_change_proto_down(dev, proto_down);
}
EXPORT_SYMBOL(dev_change_proto_down);

L
Linus Torvalds 已提交
6167 6168
/**
 *	dev_new_index	-	allocate an ifindex
6169
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
6170 6171 6172 6173 6174
 *
 *	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.
 */
6175
static int dev_new_index(struct net *net)
L
Linus Torvalds 已提交
6176
{
6177
	int ifindex = net->ifindex;
L
Linus Torvalds 已提交
6178 6179 6180
	for (;;) {
		if (++ifindex <= 0)
			ifindex = 1;
6181
		if (!__dev_get_by_index(net, ifindex))
6182
			return net->ifindex = ifindex;
L
Linus Torvalds 已提交
6183 6184 6185 6186
	}
}

/* Delayed registration/unregisteration */
6187
static LIST_HEAD(net_todo_list);
6188
DECLARE_WAIT_QUEUE_HEAD(netdev_unregistering_wq);
L
Linus Torvalds 已提交
6189

6190
static void net_set_todo(struct net_device *dev)
L
Linus Torvalds 已提交
6191 6192
{
	list_add_tail(&dev->todo_list, &net_todo_list);
6193
	dev_net(dev)->dev_unreg_count++;
L
Linus Torvalds 已提交
6194 6195
}

6196
static void rollback_registered_many(struct list_head *head)
6197
{
6198
	struct net_device *dev, *tmp;
6199
	LIST_HEAD(close_head);
6200

6201 6202 6203
	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

6204
	list_for_each_entry_safe(dev, tmp, head, unreg_list) {
6205
		/* Some devices call without registering
6206 6207
		 * for initialization unwind. Remove those
		 * devices and proceed with the remaining.
6208 6209
		 */
		if (dev->reg_state == NETREG_UNINITIALIZED) {
6210 6211
			pr_debug("unregister_netdevice: device %s/%p never was registered\n",
				 dev->name, dev);
6212

6213
			WARN_ON(1);
6214 6215
			list_del(&dev->unreg_list);
			continue;
6216
		}
6217
		dev->dismantle = true;
6218
		BUG_ON(dev->reg_state != NETREG_REGISTERED);
6219
	}
6220

6221
	/* If device is running, close it first. */
6222 6223
	list_for_each_entry(dev, head, unreg_list)
		list_add_tail(&dev->close_list, &close_head);
6224
	dev_close_many(&close_head, true);
6225

6226
	list_for_each_entry(dev, head, unreg_list) {
6227 6228
		/* And unlink it from device chain. */
		unlist_netdevice(dev);
6229

6230
		dev->reg_state = NETREG_UNREGISTERING;
6231
		on_each_cpu(flush_backlog, dev, 1);
6232
	}
6233 6234 6235

	synchronize_net();

6236
	list_for_each_entry(dev, head, unreg_list) {
6237 6238
		struct sk_buff *skb = NULL;

6239 6240
		/* Shutdown queueing discipline. */
		dev_shutdown(dev);
6241 6242


6243 6244 6245 6246
		/* Notify protocols, that we are about to destroy
		   this device. They should clean all the things.
		*/
		call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
6247

6248 6249 6250 6251 6252
		if (!dev->rtnl_link_ops ||
		    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
			skb = rtmsg_ifinfo_build_skb(RTM_DELLINK, dev, ~0U,
						     GFP_KERNEL);

6253 6254 6255
		/*
		 *	Flush the unicast and multicast chains
		 */
6256
		dev_uc_flush(dev);
6257
		dev_mc_flush(dev);
6258

6259 6260
		if (dev->netdev_ops->ndo_uninit)
			dev->netdev_ops->ndo_uninit(dev);
6261

6262 6263
		if (skb)
			rtmsg_ifinfo_send(skb, dev, GFP_KERNEL);
6264

J
Jiri Pirko 已提交
6265 6266
		/* Notifier chain MUST detach us all upper devices. */
		WARN_ON(netdev_has_any_upper_dev(dev));
6267

6268 6269
		/* Remove entries from kobject tree */
		netdev_unregister_kobject(dev);
6270 6271 6272 6273
#ifdef CONFIG_XPS
		/* Remove XPS queueing entries */
		netif_reset_xps_queues_gt(dev, 0);
#endif
6274
	}
6275

6276
	synchronize_net();
6277

6278
	list_for_each_entry(dev, head, unreg_list)
6279 6280 6281 6282 6283 6284 6285 6286 6287
		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
6288
	list_del(&single);
6289 6290
}

6291 6292
static netdev_features_t netdev_fix_features(struct net_device *dev,
	netdev_features_t features)
6293
{
6294 6295 6296
	/* Fix illegal checksum combinations */
	if ((features & NETIF_F_HW_CSUM) &&
	    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
6297
		netdev_warn(dev, "mixed HW and IP checksum settings.\n");
6298 6299 6300
		features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
	}

6301
	/* TSO requires that SG is present as well. */
6302
	if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
6303
		netdev_dbg(dev, "Dropping TSO features since no SG feature.\n");
6304
		features &= ~NETIF_F_ALL_TSO;
6305 6306
	}

6307 6308 6309 6310 6311 6312 6313 6314 6315 6316 6317 6318 6319
	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;
	}

6320 6321 6322 6323
	/* TSO ECN requires that TSO is present as well. */
	if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN)
		features &= ~NETIF_F_TSO_ECN;

6324 6325
	/* Software GSO depends on SG. */
	if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) {
6326
		netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n");
6327 6328 6329
		features &= ~NETIF_F_GSO;
	}

6330
	/* UFO needs SG and checksumming */
6331
	if (features & NETIF_F_UFO) {
6332 6333 6334 6335
		/* 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))) {
6336
			netdev_dbg(dev,
6337
				"Dropping NETIF_F_UFO since no checksum offload features.\n");
6338 6339 6340 6341
			features &= ~NETIF_F_UFO;
		}

		if (!(features & NETIF_F_SG)) {
6342
			netdev_dbg(dev,
6343
				"Dropping NETIF_F_UFO since no NETIF_F_SG feature.\n");
6344 6345 6346 6347
			features &= ~NETIF_F_UFO;
		}
	}

J
Jiri Pirko 已提交
6348 6349 6350 6351 6352 6353 6354
#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;

6355 6356 6357
	return features;
}

6358
int __netdev_update_features(struct net_device *dev)
6359
{
6360
	netdev_features_t features;
6361 6362
	int err = 0;

6363 6364
	ASSERT_RTNL();

6365 6366 6367 6368 6369 6370 6371 6372 6373
	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)
6374
		return 0;
6375

6376 6377
	netdev_dbg(dev, "Features changed: %pNF -> %pNF\n",
		&dev->features, &features);
6378 6379 6380 6381

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

6382
	if (unlikely(err < 0)) {
6383
		netdev_err(dev,
6384 6385
			"set_features() failed (%d); wanted %pNF, left %pNF\n",
			err, &features, &dev->features);
6386 6387 6388 6389 6390 6391 6392 6393 6394
		return -1;
	}

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

	return 1;
}

6395 6396 6397 6398 6399 6400 6401 6402
/**
 *	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.
 */
6403 6404 6405 6406
void netdev_update_features(struct net_device *dev)
{
	if (__netdev_update_features(dev))
		netdev_features_change(dev);
6407 6408 6409
}
EXPORT_SYMBOL(netdev_update_features);

6410 6411 6412 6413 6414 6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426
/**
 *	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);

6427 6428 6429 6430 6431 6432 6433 6434 6435 6436 6437 6438 6439 6440 6441 6442 6443 6444 6445 6446 6447 6448 6449 6450 6451 6452 6453
/**
 *	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);

6454
#ifdef CONFIG_SYSFS
6455 6456 6457
static int netif_alloc_rx_queues(struct net_device *dev)
{
	unsigned int i, count = dev->num_rx_queues;
T
Tom Herbert 已提交
6458
	struct netdev_rx_queue *rx;
6459
	size_t sz = count * sizeof(*rx);
6460

T
Tom Herbert 已提交
6461
	BUG_ON(count < 1);
6462

6463 6464 6465 6466 6467 6468
	rx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!rx) {
		rx = vzalloc(sz);
		if (!rx)
			return -ENOMEM;
	}
T
Tom Herbert 已提交
6469 6470 6471
	dev->_rx = rx;

	for (i = 0; i < count; i++)
T
Tom Herbert 已提交
6472
		rx[i].dev = dev;
6473 6474
	return 0;
}
T
Tom Herbert 已提交
6475
#endif
6476

C
Changli Gao 已提交
6477 6478 6479 6480 6481 6482 6483
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;
6484
	netdev_queue_numa_node_write(queue, NUMA_NO_NODE);
C
Changli Gao 已提交
6485
	queue->dev = dev;
T
Tom Herbert 已提交
6486 6487 6488
#ifdef CONFIG_BQL
	dql_init(&queue->dql, HZ);
#endif
C
Changli Gao 已提交
6489 6490
}

6491 6492
static void netif_free_tx_queues(struct net_device *dev)
{
W
WANG Cong 已提交
6493
	kvfree(dev->_tx);
6494 6495
}

6496 6497 6498 6499
static int netif_alloc_netdev_queues(struct net_device *dev)
{
	unsigned int count = dev->num_tx_queues;
	struct netdev_queue *tx;
6500
	size_t sz = count * sizeof(*tx);
6501

6502 6503
	if (count < 1 || count > 0xffff)
		return -EINVAL;
6504

6505 6506 6507 6508 6509 6510
	tx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!tx) {
		tx = vzalloc(sz);
		if (!tx)
			return -ENOMEM;
	}
6511
	dev->_tx = tx;
T
Tom Herbert 已提交
6512

6513 6514
	netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
	spin_lock_init(&dev->tx_global_lock);
C
Changli Gao 已提交
6515 6516

	return 0;
6517 6518
}

6519 6520 6521 6522 6523 6524 6525 6526 6527 6528 6529
void netif_tx_stop_all_queues(struct net_device *dev)
{
	unsigned int i;

	for (i = 0; i < dev->num_tx_queues; i++) {
		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
		netif_tx_stop_queue(txq);
	}
}
EXPORT_SYMBOL(netif_tx_stop_all_queues);

L
Linus Torvalds 已提交
6530 6531 6532 6533 6534 6535 6536 6537 6538 6539 6540 6541 6542 6543 6544 6545 6546 6547 6548 6549
/**
 *	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;
6550
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
6551 6552 6553 6554

	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

6555 6556
	might_sleep();

L
Linus Torvalds 已提交
6557 6558
	/* When net_device's are persistent, this will be fatal. */
	BUG_ON(dev->reg_state != NETREG_UNINITIALIZED);
6559
	BUG_ON(!net);
L
Linus Torvalds 已提交
6560

6561
	spin_lock_init(&dev->addr_list_lock);
6562
	netdev_set_addr_lockdep_class(dev);
L
Linus Torvalds 已提交
6563

6564
	ret = dev_get_valid_name(net, dev, dev->name);
6565 6566 6567
	if (ret < 0)
		goto out;

L
Linus Torvalds 已提交
6568
	/* Init, if this function is available */
6569 6570
	if (dev->netdev_ops->ndo_init) {
		ret = dev->netdev_ops->ndo_init(dev);
L
Linus Torvalds 已提交
6571 6572 6573
		if (ret) {
			if (ret > 0)
				ret = -EIO;
6574
			goto out;
L
Linus Torvalds 已提交
6575 6576
		}
	}
6577

6578 6579
	if (((dev->hw_features | dev->features) &
	     NETIF_F_HW_VLAN_CTAG_FILTER) &&
6580 6581 6582 6583 6584 6585 6586
	    (!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;
	}

6587 6588 6589 6590 6591 6592
	ret = -EBUSY;
	if (!dev->ifindex)
		dev->ifindex = dev_new_index(net);
	else if (__dev_get_by_index(net, dev->ifindex))
		goto err_uninit;

6593 6594 6595 6596
	/* Transfer changeable features to wanted_features and enable
	 * software offloads (GSO and GRO).
	 */
	dev->hw_features |= NETIF_F_SOFT_FEATURES;
6597 6598
	dev->features |= NETIF_F_SOFT_FEATURES;
	dev->wanted_features = dev->features & dev->hw_features;
L
Linus Torvalds 已提交
6599

6600 6601
	if (!(dev->flags & IFF_LOOPBACK)) {
		dev->hw_features |= NETIF_F_NOCACHE_COPY;
6602 6603
	}

6604
	/* Make NETIF_F_HIGHDMA inheritable to VLAN devices.
6605
	 */
6606
	dev->vlan_features |= NETIF_F_HIGHDMA;
6607

6608 6609 6610 6611
	/* Make NETIF_F_SG inheritable to tunnel devices.
	 */
	dev->hw_enc_features |= NETIF_F_SG;

S
Simon Horman 已提交
6612 6613 6614 6615
	/* Make NETIF_F_SG inheritable to MPLS.
	 */
	dev->mpls_features |= NETIF_F_SG;

6616 6617 6618 6619 6620
	ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		goto err_uninit;

6621
	ret = netdev_register_kobject(dev);
6622
	if (ret)
6623
		goto err_uninit;
6624 6625
	dev->reg_state = NETREG_REGISTERED;

6626
	__netdev_update_features(dev);
6627

L
Linus Torvalds 已提交
6628 6629 6630 6631 6632 6633 6634
	/*
	 *	Default initial state at registry is that the
	 *	device is present.
	 */

	set_bit(__LINK_STATE_PRESENT, &dev->state);

6635 6636
	linkwatch_init_dev(dev);

L
Linus Torvalds 已提交
6637 6638
	dev_init_scheduler(dev);
	dev_hold(dev);
6639
	list_netdevice(dev);
6640
	add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
6641

6642 6643 6644 6645 6646 6647 6648
	/* 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
6649
	/* Notify protocols, that a new device appeared. */
6650
	ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
6651
	ret = notifier_to_errno(ret);
6652 6653 6654 6655
	if (ret) {
		rollback_registered(dev);
		dev->reg_state = NETREG_UNREGISTERED;
	}
6656 6657 6658 6659
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
6660 6661
	if (!dev->rtnl_link_ops ||
	    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
6662
		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
L
Linus Torvalds 已提交
6663 6664 6665

out:
	return ret;
6666 6667

err_uninit:
6668 6669
	if (dev->netdev_ops->ndo_uninit)
		dev->netdev_ops->ndo_uninit(dev);
6670
	goto out;
L
Linus Torvalds 已提交
6671
}
E
Eric Dumazet 已提交
6672
EXPORT_SYMBOL(register_netdevice);
L
Linus Torvalds 已提交
6673

6674 6675 6676 6677 6678 6679 6680 6681 6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693 6694 6695 6696 6697 6698 6699 6700 6701 6702 6703 6704
/**
 *	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
6705 6706 6707 6708 6709
	/* Note : We dont allocate pcpu_refcnt for dummy devices,
	 * because users of this 'device' dont need to change
	 * its refcount.
	 */

6710 6711 6712 6713 6714
	return 0;
}
EXPORT_SYMBOL_GPL(init_dummy_netdev);


L
Linus Torvalds 已提交
6715 6716 6717 6718 6719 6720 6721 6722 6723
/**
 *	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.
 *
6724
 *	This is a wrapper around register_netdevice that takes the rtnl semaphore
L
Linus Torvalds 已提交
6725 6726 6727 6728 6729 6730 6731 6732 6733 6734 6735 6736 6737 6738
 *	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
6739 6740 6741 6742 6743 6744 6745 6746 6747 6748
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);

6749
/**
L
Linus Torvalds 已提交
6750
 * netdev_wait_allrefs - wait until all references are gone.
6751
 * @dev: target net_device
L
Linus Torvalds 已提交
6752 6753 6754 6755 6756 6757 6758
 *
 * 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
6759
 * call dev_put.
L
Linus Torvalds 已提交
6760 6761 6762 6763
 */
static void netdev_wait_allrefs(struct net_device *dev)
{
	unsigned long rebroadcast_time, warning_time;
E
Eric Dumazet 已提交
6764
	int refcnt;
L
Linus Torvalds 已提交
6765

6766 6767
	linkwatch_forget_dev(dev);

L
Linus Torvalds 已提交
6768
	rebroadcast_time = warning_time = jiffies;
E
Eric Dumazet 已提交
6769 6770 6771
	refcnt = netdev_refcnt_read(dev);

	while (refcnt != 0) {
L
Linus Torvalds 已提交
6772
		if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
6773
			rtnl_lock();
L
Linus Torvalds 已提交
6774 6775

			/* Rebroadcast unregister notification */
6776
			call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
L
Linus Torvalds 已提交
6777

6778
			__rtnl_unlock();
6779
			rcu_barrier();
6780 6781
			rtnl_lock();

6782
			call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
L
Linus Torvalds 已提交
6783 6784 6785 6786 6787 6788 6789 6790 6791 6792 6793
			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();
			}

6794
			__rtnl_unlock();
L
Linus Torvalds 已提交
6795 6796 6797 6798 6799 6800

			rebroadcast_time = jiffies;
		}

		msleep(250);

E
Eric Dumazet 已提交
6801 6802
		refcnt = netdev_refcnt_read(dev);

L
Linus Torvalds 已提交
6803
		if (time_after(jiffies, warning_time + 10 * HZ)) {
6804 6805
			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
				 dev->name, refcnt);
L
Linus Torvalds 已提交
6806 6807 6808 6809 6810 6811 6812 6813 6814 6815 6816 6817 6818 6819 6820 6821 6822 6823 6824
			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
6825
 * We are invoked by rtnl_unlock().
L
Linus Torvalds 已提交
6826
 * This allows us to deal with problems:
6827
 * 1) We can delete sysfs objects which invoke hotplug
L
Linus Torvalds 已提交
6828 6829 6830
 *    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
6831 6832 6833
 *
 * 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
6834 6835 6836
 */
void netdev_run_todo(void)
{
6837
	struct list_head list;
L
Linus Torvalds 已提交
6838 6839

	/* Snapshot list, allow later requests */
6840
	list_replace_init(&net_todo_list, &list);
H
Herbert Xu 已提交
6841 6842

	__rtnl_unlock();
6843

6844 6845

	/* Wait for rcu callbacks to finish before next phase */
6846 6847 6848
	if (!list_empty(&list))
		rcu_barrier();

L
Linus Torvalds 已提交
6849 6850
	while (!list_empty(&list)) {
		struct net_device *dev
6851
			= list_first_entry(&list, struct net_device, todo_list);
L
Linus Torvalds 已提交
6852 6853
		list_del(&dev->todo_list);

6854
		rtnl_lock();
6855
		call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
6856
		__rtnl_unlock();
6857

6858
		if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
6859
			pr_err("network todo '%s' but state %d\n",
6860 6861 6862 6863
			       dev->name, dev->reg_state);
			dump_stack();
			continue;
		}
L
Linus Torvalds 已提交
6864

6865
		dev->reg_state = NETREG_UNREGISTERED;
L
Linus Torvalds 已提交
6866

6867
		netdev_wait_allrefs(dev);
L
Linus Torvalds 已提交
6868

6869
		/* paranoia */
E
Eric Dumazet 已提交
6870
		BUG_ON(netdev_refcnt_read(dev));
6871 6872
		BUG_ON(!list_empty(&dev->ptype_all));
		BUG_ON(!list_empty(&dev->ptype_specific));
6873 6874
		WARN_ON(rcu_access_pointer(dev->ip_ptr));
		WARN_ON(rcu_access_pointer(dev->ip6_ptr));
6875
		WARN_ON(dev->dn_ptr);
L
Linus Torvalds 已提交
6876

6877 6878
		if (dev->destructor)
			dev->destructor(dev);
6879

6880 6881 6882 6883 6884 6885
		/* Report a network device has been unregistered */
		rtnl_lock();
		dev_net(dev)->dev_unreg_count--;
		__rtnl_unlock();
		wake_up(&netdev_unregistering_wq);

6886 6887
		/* Free network device */
		kobject_put(&dev->dev.kobj);
L
Linus Torvalds 已提交
6888 6889 6890
	}
}

6891 6892 6893
/* Convert net_device_stats to rtnl_link_stats64.  They have the same
 * fields in the same order, with only the type differing.
 */
6894 6895
void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
			     const struct net_device_stats *netdev_stats)
6896 6897
{
#if BITS_PER_LONG == 64
6898 6899
	BUILD_BUG_ON(sizeof(*stats64) != sizeof(*netdev_stats));
	memcpy(stats64, netdev_stats, sizeof(*stats64));
6900 6901 6902 6903 6904 6905 6906 6907 6908 6909 6910
#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
}
6911
EXPORT_SYMBOL(netdev_stats_to_stats64);
6912

6913 6914 6915
/**
 *	dev_get_stats	- get network device statistics
 *	@dev: device to get statistics from
6916
 *	@storage: place to store stats
6917
 *
6918 6919 6920 6921
 *	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.
6922
 */
6923 6924
struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
					struct rtnl_link_stats64 *storage)
6925
{
6926 6927
	const struct net_device_ops *ops = dev->netdev_ops;

6928 6929
	if (ops->ndo_get_stats64) {
		memset(storage, 0, sizeof(*storage));
6930 6931
		ops->ndo_get_stats64(dev, storage);
	} else if (ops->ndo_get_stats) {
6932
		netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev));
6933 6934
	} else {
		netdev_stats_to_stats64(storage, &dev->stats);
6935
	}
6936
	storage->rx_dropped += atomic_long_read(&dev->rx_dropped);
6937
	storage->tx_dropped += atomic_long_read(&dev->tx_dropped);
6938
	return storage;
R
Rusty Russell 已提交
6939
}
6940
EXPORT_SYMBOL(dev_get_stats);
R
Rusty Russell 已提交
6941

6942
struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
6943
{
6944
	struct netdev_queue *queue = dev_ingress_queue(dev);
6945

6946 6947 6948 6949 6950 6951 6952
#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);
E
Eric Dumazet 已提交
6953
	RCU_INIT_POINTER(queue->qdisc, &noop_qdisc);
6954 6955 6956 6957
	queue->qdisc_sleeping = &noop_qdisc;
	rcu_assign_pointer(dev->ingress_queue, queue);
#endif
	return queue;
6958 6959
}

6960 6961
static const struct ethtool_ops default_ethtool_ops;

6962 6963 6964 6965 6966 6967 6968 6969
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);

6970 6971 6972 6973
void netdev_freemem(struct net_device *dev)
{
	char *addr = (char *)dev - dev->padded;

W
WANG Cong 已提交
6974
	kvfree(addr);
6975 6976
}

L
Linus Torvalds 已提交
6977
/**
T
Tom Herbert 已提交
6978
 *	alloc_netdev_mqs - allocate network device
6979 6980 6981 6982 6983 6984
 *	@sizeof_priv:		size of private data to allocate space for
 *	@name:			device name format string
 *	@name_assign_type: 	origin of device name
 *	@setup:			callback to initialize device
 *	@txqs:			the number of TX subqueues to allocate
 *	@rxqs:			the number of RX subqueues to allocate
L
Linus Torvalds 已提交
6985 6986
 *
 *	Allocates a struct net_device with private data area for driver use
L
Li Zhong 已提交
6987
 *	and performs basic initialization.  Also allocates subqueue structs
T
Tom Herbert 已提交
6988
 *	for each queue on the device.
L
Linus Torvalds 已提交
6989
 */
T
Tom Herbert 已提交
6990
struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
6991
		unsigned char name_assign_type,
T
Tom Herbert 已提交
6992 6993
		void (*setup)(struct net_device *),
		unsigned int txqs, unsigned int rxqs)
L
Linus Torvalds 已提交
6994 6995
{
	struct net_device *dev;
6996
	size_t alloc_size;
6997
	struct net_device *p;
L
Linus Torvalds 已提交
6998

6999 7000
	BUG_ON(strlen(name) >= sizeof(dev->name));

T
Tom Herbert 已提交
7001
	if (txqs < 1) {
7002
		pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
7003 7004 7005
		return NULL;
	}

7006
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
7007
	if (rxqs < 1) {
7008
		pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
T
Tom Herbert 已提交
7009 7010 7011 7012
		return NULL;
	}
#endif

7013
	alloc_size = sizeof(struct net_device);
7014 7015
	if (sizeof_priv) {
		/* ensure 32-byte alignment of private area */
7016
		alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
7017 7018 7019
		alloc_size += sizeof_priv;
	}
	/* ensure 32-byte alignment of whole construct */
7020
	alloc_size += NETDEV_ALIGN - 1;
L
Linus Torvalds 已提交
7021

7022 7023 7024
	p = kzalloc(alloc_size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!p)
		p = vzalloc(alloc_size);
7025
	if (!p)
L
Linus Torvalds 已提交
7026 7027
		return NULL;

7028
	dev = PTR_ALIGN(p, NETDEV_ALIGN);
L
Linus Torvalds 已提交
7029
	dev->padded = (char *)dev - (char *)p;
7030

E
Eric Dumazet 已提交
7031 7032
	dev->pcpu_refcnt = alloc_percpu(int);
	if (!dev->pcpu_refcnt)
7033
		goto free_dev;
7034 7035

	if (dev_addr_init(dev))
E
Eric Dumazet 已提交
7036
		goto free_pcpu;
7037

7038
	dev_mc_init(dev);
7039
	dev_uc_init(dev);
J
Jiri Pirko 已提交
7040

7041
	dev_net_set(dev, &init_net);
L
Linus Torvalds 已提交
7042

7043
	dev->gso_max_size = GSO_MAX_SIZE;
7044
	dev->gso_max_segs = GSO_MAX_SEGS;
7045
	dev->gso_min_segs = 0;
7046 7047 7048

	INIT_LIST_HEAD(&dev->napi_list);
	INIT_LIST_HEAD(&dev->unreg_list);
7049
	INIT_LIST_HEAD(&dev->close_list);
7050
	INIT_LIST_HEAD(&dev->link_watch_list);
7051 7052 7053 7054
	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);
7055 7056
	INIT_LIST_HEAD(&dev->ptype_all);
	INIT_LIST_HEAD(&dev->ptype_specific);
7057
	dev->priv_flags = IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM;
7058 7059
	setup(dev);

7060
	if (!dev->tx_queue_len)
7061
		dev->priv_flags |= IFF_NO_QUEUE;
7062

T
Tom Herbert 已提交
7063 7064
	dev->num_tx_queues = txqs;
	dev->real_num_tx_queues = txqs;
7065
	if (netif_alloc_netdev_queues(dev))
7066
		goto free_all;
7067

7068
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
7069 7070
	dev->num_rx_queues = rxqs;
	dev->real_num_rx_queues = rxqs;
T
Tom Herbert 已提交
7071
	if (netif_alloc_rx_queues(dev))
7072
		goto free_all;
E
Eric Dumazet 已提交
7073
#endif
T
Tom Herbert 已提交
7074

L
Linus Torvalds 已提交
7075
	strcpy(dev->name, name);
7076
	dev->name_assign_type = name_assign_type;
7077
	dev->group = INIT_NETDEV_GROUP;
7078 7079
	if (!dev->ethtool_ops)
		dev->ethtool_ops = &default_ethtool_ops;
7080 7081 7082

	nf_hook_ingress_init(dev);

L
Linus Torvalds 已提交
7083
	return dev;
7084

7085 7086 7087 7088
free_all:
	free_netdev(dev);
	return NULL;

E
Eric Dumazet 已提交
7089 7090
free_pcpu:
	free_percpu(dev->pcpu_refcnt);
7091 7092
free_dev:
	netdev_freemem(dev);
7093
	return NULL;
L
Linus Torvalds 已提交
7094
}
T
Tom Herbert 已提交
7095
EXPORT_SYMBOL(alloc_netdev_mqs);
L
Linus Torvalds 已提交
7096 7097 7098 7099 7100

/**
 *	free_netdev - free network device
 *	@dev: device
 *
7101 7102
 *	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
7103 7104 7105 7106
 *	If this is the last reference then it will be freed.
 */
void free_netdev(struct net_device *dev)
{
7107 7108
	struct napi_struct *p, *n;

7109
	netif_free_tx_queues(dev);
7110
#ifdef CONFIG_SYSFS
7111
	kvfree(dev->_rx);
T
Tom Herbert 已提交
7112
#endif
7113

7114
	kfree(rcu_dereference_protected(dev->ingress_queue, 1));
7115

7116 7117 7118
	/* Flush device addresses */
	dev_addr_flush(dev);

7119 7120 7121
	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
		netif_napi_del(p);

E
Eric Dumazet 已提交
7122 7123 7124
	free_percpu(dev->pcpu_refcnt);
	dev->pcpu_refcnt = NULL;

S
Stephen Hemminger 已提交
7125
	/*  Compatibility with error handling in drivers */
L
Linus Torvalds 已提交
7126
	if (dev->reg_state == NETREG_UNINITIALIZED) {
7127
		netdev_freemem(dev);
L
Linus Torvalds 已提交
7128 7129 7130 7131 7132 7133
		return;
	}

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

7134 7135
	/* will free via device release */
	put_device(&dev->dev);
L
Linus Torvalds 已提交
7136
}
E
Eric Dumazet 已提交
7137
EXPORT_SYMBOL(free_netdev);
7138

7139 7140 7141 7142 7143 7144
/**
 *	synchronize_net -  Synchronize with packet receive processing
 *
 *	Wait for packets currently being received to be done.
 *	Does not block later packets from starting.
 */
7145
void synchronize_net(void)
L
Linus Torvalds 已提交
7146 7147
{
	might_sleep();
7148 7149 7150 7151
	if (rtnl_is_locked())
		synchronize_rcu_expedited();
	else
		synchronize_rcu();
L
Linus Torvalds 已提交
7152
}
E
Eric Dumazet 已提交
7153
EXPORT_SYMBOL(synchronize_net);
L
Linus Torvalds 已提交
7154 7155

/**
7156
 *	unregister_netdevice_queue - remove device from the kernel
L
Linus Torvalds 已提交
7157
 *	@dev: device
7158
 *	@head: list
7159
 *
L
Linus Torvalds 已提交
7160
 *	This function shuts down a device interface and removes it
7161
 *	from the kernel tables.
7162
 *	If head not NULL, device is queued to be unregistered later.
L
Linus Torvalds 已提交
7163 7164 7165 7166 7167
 *
 *	Callers must hold the rtnl semaphore.  You may want
 *	unregister_netdev() instead of this.
 */

7168
void unregister_netdevice_queue(struct net_device *dev, struct list_head *head)
L
Linus Torvalds 已提交
7169
{
7170 7171
	ASSERT_RTNL();

7172
	if (head) {
7173
		list_move_tail(&dev->unreg_list, head);
7174 7175 7176 7177 7178
	} else {
		rollback_registered(dev);
		/* Finish processing unregister after unlock */
		net_set_todo(dev);
	}
L
Linus Torvalds 已提交
7179
}
7180
EXPORT_SYMBOL(unregister_netdevice_queue);
L
Linus Torvalds 已提交
7181

7182 7183 7184
/**
 *	unregister_netdevice_many - unregister many devices
 *	@head: list of devices
7185 7186 7187
 *
 *  Note: As most callers use a stack allocated list_head,
 *  we force a list_del() to make sure stack wont be corrupted later.
7188 7189 7190 7191 7192 7193 7194 7195 7196
 */
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);
7197
		list_del(head);
7198 7199
	}
}
7200
EXPORT_SYMBOL(unregister_netdevice_many);
7201

L
Linus Torvalds 已提交
7202 7203 7204 7205 7206
/**
 *	unregister_netdev - remove device from the kernel
 *	@dev: device
 *
 *	This function shuts down a device interface and removes it
7207
 *	from the kernel tables.
L
Linus Torvalds 已提交
7208 7209 7210 7211 7212 7213 7214 7215 7216 7217 7218 7219 7220
 *
 *	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);

7221 7222 7223 7224 7225 7226 7227 7228 7229 7230 7231 7232 7233 7234 7235 7236 7237 7238 7239 7240 7241 7242 7243 7244 7245 7246 7247 7248 7249 7250 7251
/**
 *	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;
7252
	if (net_eq(dev_net(dev), net))
7253 7254 7255 7256 7257 7258
		goto out;

	/* Pick the destination device name, and ensure
	 * we can use it in the destination network namespace.
	 */
	err = -EEXIST;
7259
	if (__dev_get_by_name(net, dev->name)) {
7260 7261 7262
		/* We get here if we can't use the current device name */
		if (!pat)
			goto out;
7263
		if (dev_get_valid_name(net, dev, pat) < 0)
7264 7265 7266 7267 7268 7269 7270 7271
			goto out;
	}

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

	/* If device is running close it first. */
7272
	dev_close(dev);
7273 7274 7275 7276 7277 7278 7279 7280 7281 7282 7283 7284

	/* 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.
7285 7286 7287 7288

	   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.
7289 7290
	*/
	call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
7291 7292
	rcu_barrier();
	call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
7293
	rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);
7294 7295 7296 7297

	/*
	 *	Flush the unicast and multicast chains
	 */
7298
	dev_uc_flush(dev);
7299
	dev_mc_flush(dev);
7300

7301 7302
	/* Send a netdev-removed uevent to the old namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_REMOVE);
7303
	netdev_adjacent_del_links(dev);
7304

7305
	/* Actually switch the network namespace */
7306
	dev_net_set(dev, net);
7307 7308

	/* If there is an ifindex conflict assign a new one */
7309
	if (__dev_get_by_index(net, dev->ifindex))
7310 7311
		dev->ifindex = dev_new_index(net);

7312 7313
	/* Send a netdev-add uevent to the new namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_ADD);
7314
	netdev_adjacent_add_links(dev);
7315

7316
	/* Fixup kobjects */
7317
	err = device_rename(&dev->dev, dev->name);
7318
	WARN_ON(err);
7319 7320 7321 7322 7323 7324 7325

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

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

7326 7327 7328 7329
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
7330
	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
7331

7332 7333 7334 7335 7336
	synchronize_net();
	err = 0;
out:
	return err;
}
7337
EXPORT_SYMBOL_GPL(dev_change_net_namespace);
7338

L
Linus Torvalds 已提交
7339 7340 7341 7342 7343 7344 7345 7346 7347
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;

7348
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
L
Linus Torvalds 已提交
7349 7350 7351 7352 7353 7354 7355 7356 7357 7358 7359 7360 7361 7362 7363 7364
		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. */
7365 7366 7367 7368 7369 7370
	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;
	}
E
Eric Dumazet 已提交
7371 7372 7373 7374 7375 7376 7377 7378 7379 7380 7381 7382 7383 7384
	/* Append NAPI poll list from offline CPU, with one exception :
	 * process_backlog() must be called by cpu owning percpu backlog.
	 * We properly handle process_queue & input_pkt_queue later.
	 */
	while (!list_empty(&oldsd->poll_list)) {
		struct napi_struct *napi = list_first_entry(&oldsd->poll_list,
							    struct napi_struct,
							    poll_list);

		list_del_init(&napi->poll_list);
		if (napi->poll == process_backlog)
			napi->state = 0;
		else
			____napi_schedule(sd, napi);
7385
	}
L
Linus Torvalds 已提交
7386 7387 7388 7389 7390

	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_enable();

	/* Process offline CPU's input_pkt_queue */
7391
	while ((skb = __skb_dequeue(&oldsd->process_queue))) {
7392
		netif_rx_ni(skb);
7393
		input_queue_head_incr(oldsd);
T
Tom Herbert 已提交
7394
	}
E
Eric Dumazet 已提交
7395
	while ((skb = skb_dequeue(&oldsd->input_pkt_queue))) {
7396
		netif_rx_ni(skb);
7397 7398
		input_queue_head_incr(oldsd);
	}
L
Linus Torvalds 已提交
7399 7400 7401 7402 7403

	return NOTIFY_OK;
}


7404
/**
7405 7406 7407 7408
 *	netdev_increment_features - increment feature set by one
 *	@all: current feature set
 *	@one: new feature set
 *	@mask: mask feature set
7409 7410
 *
 *	Computes a new feature set after adding a device with feature set
7411 7412
 *	@one to the master device with current feature set @all.  Will not
 *	enable anything that is off in @mask. Returns the new feature set.
7413
 */
7414 7415
netdev_features_t netdev_increment_features(netdev_features_t all,
	netdev_features_t one, netdev_features_t mask)
7416
{
7417 7418 7419
	if (mask & NETIF_F_GEN_CSUM)
		mask |= NETIF_F_ALL_CSUM;
	mask |= NETIF_F_VLAN_CHALLENGED;
7420

7421 7422
	all |= one & (NETIF_F_ONE_FOR_ALL|NETIF_F_ALL_CSUM) & mask;
	all &= one | ~NETIF_F_ALL_FOR_ALL;
7423

7424 7425 7426
	/* If one device supports hw checksumming, set for all. */
	if (all & NETIF_F_GEN_CSUM)
		all &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
7427 7428 7429

	return all;
}
7430
EXPORT_SYMBOL(netdev_increment_features);
7431

7432
static struct hlist_head * __net_init netdev_create_hash(void)
7433 7434 7435 7436 7437 7438 7439 7440 7441 7442 7443 7444
{
	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;
}

7445
/* Initialize per network namespace state */
7446
static int __net_init netdev_init(struct net *net)
7447
{
7448 7449
	if (net != &init_net)
		INIT_LIST_HEAD(&net->dev_base_head);
7450

7451 7452 7453
	net->dev_name_head = netdev_create_hash();
	if (net->dev_name_head == NULL)
		goto err_name;
7454

7455 7456 7457
	net->dev_index_head = netdev_create_hash();
	if (net->dev_index_head == NULL)
		goto err_idx;
7458 7459

	return 0;
7460 7461 7462 7463 7464

err_idx:
	kfree(net->dev_name_head);
err_name:
	return -ENOMEM;
7465 7466
}

7467 7468 7469 7470 7471 7472
/**
 *	netdev_drivername - network driver for the device
 *	@dev: network device
 *
 *	Determine network driver for device.
 */
7473
const char *netdev_drivername(const struct net_device *dev)
7474
{
7475 7476
	const struct device_driver *driver;
	const struct device *parent;
7477
	const char *empty = "";
7478 7479 7480

	parent = dev->dev.parent;
	if (!parent)
7481
		return empty;
7482 7483 7484

	driver = parent->driver;
	if (driver && driver->name)
7485 7486
		return driver->name;
	return empty;
7487 7488
}

7489 7490
static void __netdev_printk(const char *level, const struct net_device *dev,
			    struct va_format *vaf)
7491
{
7492
	if (dev && dev->dev.parent) {
7493 7494 7495 7496 7497 7498 7499
		dev_printk_emit(level[1] - '0',
				dev->dev.parent,
				"%s %s %s%s: %pV",
				dev_driver_string(dev->dev.parent),
				dev_name(dev->dev.parent),
				netdev_name(dev), netdev_reg_state(dev),
				vaf);
7500
	} else if (dev) {
7501 7502
		printk("%s%s%s: %pV",
		       level, netdev_name(dev), netdev_reg_state(dev), vaf);
7503
	} else {
7504
		printk("%s(NULL net_device): %pV", level, vaf);
7505
	}
7506 7507
}

7508 7509
void netdev_printk(const char *level, const struct net_device *dev,
		   const char *format, ...)
7510 7511 7512 7513 7514 7515 7516 7517 7518
{
	struct va_format vaf;
	va_list args;

	va_start(args, format);

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

7519
	__netdev_printk(level, dev, &vaf);
7520

7521 7522 7523 7524 7525
	va_end(args);
}
EXPORT_SYMBOL(netdev_printk);

#define define_netdev_printk_level(func, level)			\
7526
void func(const struct net_device *dev, const char *fmt, ...)	\
7527 7528 7529 7530 7531 7532 7533 7534 7535
{								\
	struct va_format vaf;					\
	va_list args;						\
								\
	va_start(args, fmt);					\
								\
	vaf.fmt = fmt;						\
	vaf.va = &args;						\
								\
7536
	__netdev_printk(level, dev, &vaf);			\
7537
								\
7538 7539 7540 7541 7542 7543 7544 7545 7546 7547 7548 7549
	va_end(args);						\
}								\
EXPORT_SYMBOL(func);

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

7550
static void __net_exit netdev_exit(struct net *net)
7551 7552 7553 7554 7555
{
	kfree(net->dev_name_head);
	kfree(net->dev_index_head);
}

7556
static struct pernet_operations __net_initdata netdev_net_ops = {
7557 7558 7559 7560
	.init = netdev_init,
	.exit = netdev_exit,
};

7561
static void __net_exit default_device_exit(struct net *net)
7562
{
7563
	struct net_device *dev, *aux;
7564
	/*
7565
	 * Push all migratable network devices back to the
7566 7567 7568
	 * initial network namespace
	 */
	rtnl_lock();
7569
	for_each_netdev_safe(net, dev, aux) {
7570
		int err;
7571
		char fb_name[IFNAMSIZ];
7572 7573 7574 7575 7576

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

7577 7578 7579
		/* Leave virtual devices for the generic cleanup */
		if (dev->rtnl_link_ops)
			continue;
7580

L
Lucas De Marchi 已提交
7581
		/* Push remaining network devices to init_net */
7582 7583
		snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex);
		err = dev_change_net_namespace(dev, &init_net, fb_name);
7584
		if (err) {
7585 7586
			pr_emerg("%s: failed to move %s to init_net: %d\n",
				 __func__, dev->name, err);
7587
			BUG();
7588 7589 7590 7591 7592
		}
	}
	rtnl_unlock();
}

7593 7594 7595 7596 7597 7598 7599
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;
7600
	DEFINE_WAIT_FUNC(wait, woken_wake_function);
7601

7602
	add_wait_queue(&netdev_unregistering_wq, &wait);
7603 7604 7605 7606 7607 7608 7609 7610 7611 7612 7613 7614
	for (;;) {
		unregistering = false;
		rtnl_lock();
		list_for_each_entry(net, net_list, exit_list) {
			if (net->dev_unreg_count > 0) {
				unregistering = true;
				break;
			}
		}
		if (!unregistering)
			break;
		__rtnl_unlock();
7615 7616

		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
7617
	}
7618
	remove_wait_queue(&netdev_unregistering_wq, &wait);
7619 7620
}

7621 7622 7623
static void __net_exit default_device_exit_batch(struct list_head *net_list)
{
	/* At exit all network devices most be removed from a network
7624
	 * namespace.  Do this in the reverse order of registration.
7625 7626 7627 7628 7629 7630 7631
	 * 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);

7632 7633 7634 7635 7636 7637 7638 7639 7640 7641 7642 7643
	/* 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);
7644 7645
	list_for_each_entry(net, net_list, exit_list) {
		for_each_netdev_reverse(net, dev) {
7646
			if (dev->rtnl_link_ops && dev->rtnl_link_ops->dellink)
7647 7648 7649 7650 7651 7652 7653 7654 7655
				dev->rtnl_link_ops->dellink(dev, &dev_kill_list);
			else
				unregister_netdevice_queue(dev, &dev_kill_list);
		}
	}
	unregister_netdevice_many(&dev_kill_list);
	rtnl_unlock();
}

7656
static struct pernet_operations __net_initdata default_device_ops = {
7657
	.exit = default_device_exit,
7658
	.exit_batch = default_device_exit_batch,
7659 7660
};

L
Linus Torvalds 已提交
7661 7662 7663 7664 7665 7666 7667 7668 7669 7670 7671 7672 7673 7674 7675 7676 7677 7678 7679 7680
/*
 *	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;

7681
	if (netdev_kobject_init())
L
Linus Torvalds 已提交
7682 7683 7684
		goto out;

	INIT_LIST_HEAD(&ptype_all);
7685
	for (i = 0; i < PTYPE_HASH_SIZE; i++)
L
Linus Torvalds 已提交
7686 7687
		INIT_LIST_HEAD(&ptype_base[i]);

7688 7689
	INIT_LIST_HEAD(&offload_base);

7690 7691
	if (register_pernet_subsys(&netdev_net_ops))
		goto out;
L
Linus Torvalds 已提交
7692 7693 7694 7695 7696

	/*
	 *	Initialise the packet receive queues.
	 */

7697
	for_each_possible_cpu(i) {
E
Eric Dumazet 已提交
7698
		struct softnet_data *sd = &per_cpu(softnet_data, i);
L
Linus Torvalds 已提交
7699

E
Eric Dumazet 已提交
7700
		skb_queue_head_init(&sd->input_pkt_queue);
7701
		skb_queue_head_init(&sd->process_queue);
E
Eric Dumazet 已提交
7702
		INIT_LIST_HEAD(&sd->poll_list);
7703
		sd->output_queue_tailp = &sd->output_queue;
E
Eric Dumazet 已提交
7704
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
7705 7706 7707
		sd->csd.func = rps_trigger_softirq;
		sd->csd.info = sd;
		sd->cpu = i;
7708
#endif
T
Tom Herbert 已提交
7709

E
Eric Dumazet 已提交
7710 7711
		sd->backlog.poll = process_backlog;
		sd->backlog.weight = weight_p;
L
Linus Torvalds 已提交
7712 7713 7714 7715
	}

	dev_boot_phase = 0;

7716 7717 7718 7719 7720 7721 7722 7723 7724 7725 7726 7727 7728 7729 7730
	/* 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;

7731 7732
	open_softirq(NET_TX_SOFTIRQ, net_tx_action);
	open_softirq(NET_RX_SOFTIRQ, net_rx_action);
L
Linus Torvalds 已提交
7733 7734

	hotcpu_notifier(dev_cpu_callback, 0);
T
Thomas Graf 已提交
7735
	dst_subsys_init();
L
Linus Torvalds 已提交
7736 7737 7738 7739 7740 7741
	rc = 0;
out:
	return rc;
}

subsys_initcall(net_dev_init);