dev.c 205.0 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
#include <net/sock.h>
E
Eric Dumazet 已提交
99
#include <net/busy_poll.h>
L
Linus Torvalds 已提交
100 101 102
#include <linux/rtnetlink.h>
#include <linux/stat.h>
#include <net/dst.h>
103
#include <net/dst_metadata.h>
L
Linus Torvalds 已提交
104 105
#include <net/pkt_sched.h>
#include <net/checksum.h>
106
#include <net/xfrm.h>
L
Linus Torvalds 已提交
107 108 109 110 111 112 113 114
#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
115
#include <linux/audit.h>
116
#include <linux/dmaengine.h>
117
#include <linux/err.h>
118
#include <linux/ctype.h>
119
#include <linux/if_arp.h>
120
#include <linux/if_vlan.h>
121
#include <linux/ip.h>
122
#include <net/ip.h>
123
#include <net/mpls.h>
124 125
#include <linux/ipv6.h>
#include <linux/in.h>
D
David S. Miller 已提交
126 127
#include <linux/jhash.h>
#include <linux/random.h>
128
#include <trace/events/napi.h>
129
#include <trace/events/net.h>
130
#include <trace/events/skb.h>
131
#include <linux/pci.h>
132
#include <linux/inetdevice.h>
133
#include <linux/cpu_rmap.h>
134
#include <linux/static_key.h>
E
Eliezer Tamir 已提交
135
#include <linux/hashtable.h>
136
#include <linux/vmalloc.h>
137
#include <linux/if_macvlan.h>
138
#include <linux/errqueue.h>
139
#include <linux/hrtimer.h>
140
#include <linux/netfilter_ingress.h>
141
#include <linux/sctp.h>
142
#include <linux/crash_dump.h>
L
Linus Torvalds 已提交
143

144 145
#include "net-sysfs.h"

146 147 148
/* Instead of increasing this, you should create a hash table. */
#define MAX_GRO_SKBS 8

H
Herbert Xu 已提交
149 150 151
/* 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
152
static DEFINE_SPINLOCK(ptype_lock);
153
static DEFINE_SPINLOCK(offload_lock);
154 155
struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
struct list_head ptype_all __read_mostly;	/* Taps */
156
static struct list_head offload_base __read_mostly;
L
Linus Torvalds 已提交
157

158
static int netif_rx_internal(struct sk_buff *skb);
159 160 161
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info);
162

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

188
static unsigned int napi_gen_id = NR_CPUS;
189
static DEFINE_READ_MOSTLY_HASHTABLE(napi_hash, 8);
E
Eliezer Tamir 已提交
190

191
static seqcount_t devnet_rename_seq;
192

193 194 195 196 197
static inline void dev_base_seq_inc(struct net *net)
{
	while (++net->dev_base_seq == 0);
}

198
static inline struct hlist_head *dev_name_hash(struct net *net, const char *name)
L
Linus Torvalds 已提交
199
{
200 201
	unsigned int hash = full_name_hash(name, strnlen(name, IFNAMSIZ));

202
	return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)];
L
Linus Torvalds 已提交
203 204
}

205
static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex)
L
Linus Torvalds 已提交
206
{
207
	return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)];
L
Linus Torvalds 已提交
208 209
}

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

E
Eric Dumazet 已提交
217
static inline void rps_unlock(struct softnet_data *sd)
218 219
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
220
	spin_unlock(&sd->input_pkt_queue.lock);
221 222 223
#endif
}

224
/* Device list insertion */
225
static void list_netdevice(struct net_device *dev)
226
{
227
	struct net *net = dev_net(dev);
228 229 230 231

	ASSERT_RTNL();

	write_lock_bh(&dev_base_lock);
232
	list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
233
	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
234 235
	hlist_add_head_rcu(&dev->index_hlist,
			   dev_index_hash(net, dev->ifindex));
236
	write_unlock_bh(&dev_base_lock);
237 238

	dev_base_seq_inc(net);
239 240
}

241 242 243
/* Device list removal
 * caller must respect a RCU grace period before freeing/reusing dev
 */
244 245 246 247 248 249
static void unlist_netdevice(struct net_device *dev)
{
	ASSERT_RTNL();

	/* Unlink dev from the device chain */
	write_lock_bh(&dev_base_lock);
250
	list_del_rcu(&dev->dev_list);
251
	hlist_del_rcu(&dev->name_hlist);
252
	hlist_del_rcu(&dev->index_hlist);
253
	write_unlock_bh(&dev_base_lock);
254 255

	dev_base_seq_inc(dev_net(dev));
256 257
}

L
Linus Torvalds 已提交
258 259 260 261
/*
 *	Our notifier list
 */

262
static RAW_NOTIFIER_HEAD(netdev_chain);
L
Linus Torvalds 已提交
263 264 265 266 267

/*
 *	Device drivers call our routines to queue packets here. We empty the
 *	queue in the local softnet handler.
 */
268

269
DEFINE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
E
Eric Dumazet 已提交
270
EXPORT_PER_CPU_SYMBOL(softnet_data);
L
Linus Torvalds 已提交
271

272
#ifdef CONFIG_LOCKDEP
273
/*
274
 * register_netdevice() inits txq->_xmit_lock and sets lockdep class
275 276 277 278 279 280 281 282 283 284 285 286 287 288 289
 * 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,
290 291 292
	 ARPHRD_FCFABRIC, ARPHRD_IEEE80211, ARPHRD_IEEE80211_PRISM,
	 ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, ARPHRD_PHONET_PIPE,
	 ARPHRD_IEEE802154, ARPHRD_VOID, ARPHRD_NONE};
293

294
static const char *const netdev_lock_name[] =
295 296 297 298 299 300 301 302 303 304 305 306
	{"_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",
307 308 309
	 "_xmit_FCFABRIC", "_xmit_IEEE80211", "_xmit_IEEE80211_PRISM",
	 "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", "_xmit_PHONET_PIPE",
	 "_xmit_IEEE802154", "_xmit_VOID", "_xmit_NONE"};
310 311

static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)];
312
static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)];
313 314 315 316 317 318 319 320 321 322 323 324

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;
}

325 326
static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
						 unsigned short dev_type)
327 328 329 330 331 332 333
{
	int i;

	i = netdev_lock_pos(dev_type);
	lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i],
				   netdev_lock_name[i]);
}
334 335 336 337 338 339 340 341 342 343

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]);
}
344
#else
345 346 347 348 349
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)
350 351 352
{
}
#endif
L
Linus Torvalds 已提交
353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375

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

		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)
 */

376 377 378
static inline struct list_head *ptype_head(const struct packet_type *pt)
{
	if (pt->type == htons(ETH_P_ALL))
379
		return pt->dev ? &pt->dev->ptype_all : &ptype_all;
380
	else
381 382
		return pt->dev ? &pt->dev->ptype_specific :
				 &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK];
383 384
}

L
Linus Torvalds 已提交
385 386 387 388 389 390 391 392
/**
 *	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.
 *
393
 *	This call does not sleep therefore it can not
L
Linus Torvalds 已提交
394 395 396 397 398 399
 *	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)
{
400
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
401

402 403 404
	spin_lock(&ptype_lock);
	list_add_rcu(&pt->list, head);
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
405
}
E
Eric Dumazet 已提交
406
EXPORT_SYMBOL(dev_add_pack);
L
Linus Torvalds 已提交
407 408 409 410 411 412 413 414

/**
 *	__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
415
 *	returns.
L
Linus Torvalds 已提交
416 417 418 419 420 421 422
 *
 *      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)
{
423
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
424 425
	struct packet_type *pt1;

426
	spin_lock(&ptype_lock);
L
Linus Torvalds 已提交
427 428 429 430 431 432 433 434

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

435
	pr_warn("dev_remove_pack: %p not found\n", pt);
L
Linus Torvalds 已提交
436
out:
437
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
438
}
E
Eric Dumazet 已提交
439 440
EXPORT_SYMBOL(__dev_remove_pack);

L
Linus Torvalds 已提交
441 442 443 444 445 446 447 448 449 450 451 452 453 454 455
/**
 *	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);
456

L
Linus Torvalds 已提交
457 458
	synchronize_net();
}
E
Eric Dumazet 已提交
459
EXPORT_SYMBOL(dev_remove_pack);
L
Linus Torvalds 已提交
460

461 462 463 464 465 466 467 468 469 470 471 472 473 474 475

/**
 *	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)
{
476
	struct packet_offload *elem;
477 478

	spin_lock(&offload_lock);
479 480 481 482 483
	list_for_each_entry(elem, &offload_base, list) {
		if (po->priority < elem->priority)
			break;
	}
	list_add_rcu(&po->list, elem->list.prev);
484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
	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
501
static void __dev_remove_offload(struct packet_offload *po)
502 503 504 505
{
	struct list_head *head = &offload_base;
	struct packet_offload *po1;

506
	spin_lock(&offload_lock);
507 508 509 510 511 512 513 514 515 516

	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:
517
	spin_unlock(&offload_lock);
518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539
}

/**
 *	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
540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566
/******************************************************************************

		      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));
567
			strlcpy(s[i].name, name, IFNAMSIZ);
L
Linus Torvalds 已提交
568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591
			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] != ' ' &&
592
		    !strcmp(dev->name, s[i].name)) {
L
Linus Torvalds 已提交
593 594 595 596 597 598 599 600 601
			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
602
EXPORT_SYMBOL(netdev_boot_setup_check);
L
Linus Torvalds 已提交
603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626


/**
 *	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
	 */
627
	if (__dev_get_by_name(&init_net, name))
L
Linus Torvalds 已提交
628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670
		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

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

671 672 673 674 675 676 677 678 679 680 681 682 683
/**
 *	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);

684
	return dev->ifindex;
685 686 687
}
EXPORT_SYMBOL(dev_get_iflink);

688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713
/**
 *	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
714 715
/**
 *	__dev_get_by_name	- find a device by its name
716
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
717 718 719 720 721 722 723 724 725
 *	@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.
 */

726
struct net_device *__dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
727
{
E
Eric Dumazet 已提交
728 729
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);
L
Linus Torvalds 已提交
730

731
	hlist_for_each_entry(dev, head, name_hlist)
L
Linus Torvalds 已提交
732 733
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;
E
Eric Dumazet 已提交
734

L
Linus Torvalds 已提交
735 736
	return NULL;
}
E
Eric Dumazet 已提交
737
EXPORT_SYMBOL(__dev_get_by_name);
L
Linus Torvalds 已提交
738

739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755
/**
 *	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);

756
	hlist_for_each_entry_rcu(dev, head, name_hlist)
757 758 759 760 761 762 763
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_name_rcu);

L
Linus Torvalds 已提交
764 765
/**
 *	dev_get_by_name		- find a device by its name
766
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
767 768 769 770 771 772 773 774 775
 *	@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.
 */

776
struct net_device *dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
777 778 779
{
	struct net_device *dev;

780 781
	rcu_read_lock();
	dev = dev_get_by_name_rcu(net, name);
L
Linus Torvalds 已提交
782 783
	if (dev)
		dev_hold(dev);
784
	rcu_read_unlock();
L
Linus Torvalds 已提交
785 786
	return dev;
}
E
Eric Dumazet 已提交
787
EXPORT_SYMBOL(dev_get_by_name);
L
Linus Torvalds 已提交
788 789 790

/**
 *	__dev_get_by_index - find a device by its ifindex
791
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
792 793 794 795 796 797 798 799 800
 *	@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.
 */

801
struct net_device *__dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
802
{
E
Eric Dumazet 已提交
803 804
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);
L
Linus Torvalds 已提交
805

806
	hlist_for_each_entry(dev, head, index_hlist)
L
Linus Torvalds 已提交
807 808
		if (dev->ifindex == ifindex)
			return dev;
E
Eric Dumazet 已提交
809

L
Linus Torvalds 已提交
810 811
	return NULL;
}
E
Eric Dumazet 已提交
812
EXPORT_SYMBOL(__dev_get_by_index);
L
Linus Torvalds 已提交
813

814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829
/**
 *	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);

830
	hlist_for_each_entry_rcu(dev, head, index_hlist)
831 832 833 834 835 836 837
		if (dev->ifindex == ifindex)
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_index_rcu);

L
Linus Torvalds 已提交
838 839 840

/**
 *	dev_get_by_index - find a device by its ifindex
841
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
842 843 844 845 846 847 848 849
 *	@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.
 */

850
struct net_device *dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
851 852 853
{
	struct net_device *dev;

854 855
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, ifindex);
L
Linus Torvalds 已提交
856 857
	if (dev)
		dev_hold(dev);
858
	rcu_read_unlock();
L
Linus Torvalds 已提交
859 860
	return dev;
}
E
Eric Dumazet 已提交
861
EXPORT_SYMBOL(dev_get_by_index);
L
Linus Torvalds 已提交
862

863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896
/**
 *	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
897
/**
898
 *	dev_getbyhwaddr_rcu - find a device by its hardware address
899
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
900 901 902 903
 *	@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
904 905
 *	is not found or a pointer to the device.
 *	The caller must hold RCU or RTNL.
906
 *	The returned device has not had its ref count increased
L
Linus Torvalds 已提交
907 908 909 910
 *	and the caller must therefore be careful about locking
 *
 */

911 912
struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
				       const char *ha)
L
Linus Torvalds 已提交
913 914 915
{
	struct net_device *dev;

916
	for_each_netdev_rcu(net, dev)
L
Linus Torvalds 已提交
917 918
		if (dev->type == type &&
		    !memcmp(dev->dev_addr, ha, dev->addr_len))
919 920 921
			return dev;

	return NULL;
L
Linus Torvalds 已提交
922
}
923
EXPORT_SYMBOL(dev_getbyhwaddr_rcu);
924

925
struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type)
L
Linus Torvalds 已提交
926 927 928
{
	struct net_device *dev;

929
	ASSERT_RTNL();
930
	for_each_netdev(net, dev)
931
		if (dev->type == type)
932 933 934
			return dev;

	return NULL;
935 936 937
}
EXPORT_SYMBOL(__dev_getfirstbyhwtype);

938
struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type)
939
{
940
	struct net_device *dev, *ret = NULL;
941

942 943 944 945 946 947 948 949 950
	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
951 952 953 954
}
EXPORT_SYMBOL(dev_getfirstbyhwtype);

/**
955
 *	__dev_get_by_flags - find any device with given flags
956
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
957 958 959 960
 *	@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
961
 *	is not found or a pointer to the device. Must be called inside
962
 *	rtnl_lock(), and result refcount is unchanged.
L
Linus Torvalds 已提交
963 964
 */

965 966
struct net_device *__dev_get_by_flags(struct net *net, unsigned short if_flags,
				      unsigned short mask)
L
Linus Torvalds 已提交
967
{
968
	struct net_device *dev, *ret;
L
Linus Torvalds 已提交
969

970 971
	ASSERT_RTNL();

972
	ret = NULL;
973
	for_each_netdev(net, dev) {
L
Linus Torvalds 已提交
974
		if (((dev->flags ^ if_flags) & mask) == 0) {
975
			ret = dev;
L
Linus Torvalds 已提交
976 977 978
			break;
		}
	}
979
	return ret;
L
Linus Torvalds 已提交
980
}
981
EXPORT_SYMBOL(__dev_get_by_flags);
L
Linus Torvalds 已提交
982 983 984 985 986 987

/**
 *	dev_valid_name - check if name is okay for network device
 *	@name: name string
 *
 *	Network device names need to be valid file names to
988 989
 *	to allow sysfs to work.  We also disallow any kind of
 *	whitespace.
L
Linus Torvalds 已提交
990
 */
991
bool dev_valid_name(const char *name)
L
Linus Torvalds 已提交
992
{
993
	if (*name == '\0')
994
		return false;
995
	if (strlen(name) >= IFNAMSIZ)
996
		return false;
997
	if (!strcmp(name, ".") || !strcmp(name, ".."))
998
		return false;
999 1000

	while (*name) {
1001
		if (*name == '/' || *name == ':' || isspace(*name))
1002
			return false;
1003 1004
		name++;
	}
1005
	return true;
L
Linus Torvalds 已提交
1006
}
E
Eric Dumazet 已提交
1007
EXPORT_SYMBOL(dev_valid_name);
L
Linus Torvalds 已提交
1008 1009

/**
1010 1011
 *	__dev_alloc_name - allocate a name for a device
 *	@net: network namespace to allocate the device name in
L
Linus Torvalds 已提交
1012
 *	@name: name format string
1013
 *	@buf:  scratch buffer and result name string
L
Linus Torvalds 已提交
1014 1015
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
S
Stephen Hemminger 已提交
1016 1017 1018 1019 1020 1021
 *	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
1022 1023
 */

1024
static int __dev_alloc_name(struct net *net, const char *name, char *buf)
L
Linus Torvalds 已提交
1025 1026 1027 1028
{
	int i = 0;
	const char *p;
	const int max_netdevices = 8*PAGE_SIZE;
S
Stephen Hemminger 已提交
1029
	unsigned long *inuse;
L
Linus Torvalds 已提交
1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042
	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
1043
		inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC);
L
Linus Torvalds 已提交
1044 1045 1046
		if (!inuse)
			return -ENOMEM;

1047
		for_each_netdev(net, d) {
L
Linus Torvalds 已提交
1048 1049 1050 1051 1052 1053
			if (!sscanf(d->name, name, &i))
				continue;
			if (i < 0 || i >= max_netdevices)
				continue;

			/*  avoid cases where sscanf is not exact inverse of printf */
1054
			snprintf(buf, IFNAMSIZ, name, i);
L
Linus Torvalds 已提交
1055 1056 1057 1058 1059 1060 1061 1062
			if (!strncmp(buf, d->name, IFNAMSIZ))
				set_bit(i, inuse);
		}

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

1063 1064
	if (buf != name)
		snprintf(buf, IFNAMSIZ, name, i);
1065
	if (!__dev_get_by_name(net, buf))
L
Linus Torvalds 已提交
1066 1067 1068 1069 1070 1071 1072 1073 1074
		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;
}

1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
/**
 *	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;

1095 1096
	BUG_ON(!dev_net(dev));
	net = dev_net(dev);
1097 1098 1099 1100 1101
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
}
E
Eric Dumazet 已提交
1102
EXPORT_SYMBOL(dev_alloc_name);
1103

1104 1105 1106
static int dev_alloc_name_ns(struct net *net,
			     struct net_device *dev,
			     const char *name)
1107
{
1108 1109
	char buf[IFNAMSIZ];
	int ret;
1110

1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121
	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);
1122

1123 1124 1125
	if (!dev_valid_name(name))
		return -EINVAL;

1126
	if (strchr(name, '%'))
1127
		return dev_alloc_name_ns(net, dev, name);
1128 1129
	else if (__dev_get_by_name(net, name))
		return -EEXIST;
1130 1131
	else if (dev->name != name)
		strlcpy(dev->name, name, IFNAMSIZ);
1132 1133 1134

	return 0;
}
L
Linus Torvalds 已提交
1135 1136 1137 1138 1139 1140 1141 1142 1143

/**
 *	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.
 */
1144
int dev_change_name(struct net_device *dev, const char *newname)
L
Linus Torvalds 已提交
1145
{
1146
	unsigned char old_assign_type;
1147
	char oldname[IFNAMSIZ];
L
Linus Torvalds 已提交
1148
	int err = 0;
1149
	int ret;
1150
	struct net *net;
L
Linus Torvalds 已提交
1151 1152

	ASSERT_RTNL();
1153
	BUG_ON(!dev_net(dev));
L
Linus Torvalds 已提交
1154

1155
	net = dev_net(dev);
L
Linus Torvalds 已提交
1156 1157 1158
	if (dev->flags & IFF_UP)
		return -EBUSY;

1159
	write_seqcount_begin(&devnet_rename_seq);
1160 1161

	if (strncmp(newname, dev->name, IFNAMSIZ) == 0) {
1162
		write_seqcount_end(&devnet_rename_seq);
1163
		return 0;
1164
	}
1165

1166 1167
	memcpy(oldname, dev->name, IFNAMSIZ);

1168
	err = dev_get_valid_name(net, dev, newname);
1169
	if (err < 0) {
1170
		write_seqcount_end(&devnet_rename_seq);
1171
		return err;
1172
	}
L
Linus Torvalds 已提交
1173

1174 1175 1176
	if (oldname[0] && !strchr(oldname, '%'))
		netdev_info(dev, "renamed from %s\n", oldname);

1177 1178 1179
	old_assign_type = dev->name_assign_type;
	dev->name_assign_type = NET_NAME_RENAMED;

1180
rollback:
1181 1182 1183
	ret = device_rename(&dev->dev, dev->name);
	if (ret) {
		memcpy(dev->name, oldname, IFNAMSIZ);
1184
		dev->name_assign_type = old_assign_type;
1185
		write_seqcount_end(&devnet_rename_seq);
1186
		return ret;
1187
	}
1188

1189
	write_seqcount_end(&devnet_rename_seq);
1190

1191 1192
	netdev_adjacent_rename_links(dev, oldname);

1193
	write_lock_bh(&dev_base_lock);
1194
	hlist_del_rcu(&dev->name_hlist);
1195 1196 1197 1198 1199 1200
	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));
1201 1202
	write_unlock_bh(&dev_base_lock);

1203
	ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
1204 1205 1206
	ret = notifier_to_errno(ret);

	if (ret) {
1207 1208
		/* err >= 0 after dev_alloc_name() or stores the first errno */
		if (err >= 0) {
1209
			err = ret;
1210
			write_seqcount_begin(&devnet_rename_seq);
1211
			memcpy(dev->name, oldname, IFNAMSIZ);
1212
			memcpy(oldname, newname, IFNAMSIZ);
1213 1214
			dev->name_assign_type = old_assign_type;
			old_assign_type = NET_NAME_RENAMED;
1215
			goto rollback;
1216
		} else {
1217
			pr_err("%s: name change rollback failed: %d\n",
1218
			       dev->name, ret);
1219 1220
		}
	}
L
Linus Torvalds 已提交
1221 1222 1223 1224

	return err;
}

1225 1226 1227 1228
/**
 *	dev_set_alias - change ifalias of a device
 *	@dev: device
 *	@alias: name up to IFALIASZ
1229
 *	@len: limit of bytes to copy from info
1230 1231 1232 1233 1234
 *
 *	Set ifalias for a device,
 */
int dev_set_alias(struct net_device *dev, const char *alias, size_t len)
{
1235 1236
	char *new_ifalias;

1237 1238 1239 1240 1241
	ASSERT_RTNL();

	if (len >= IFALIASZ)
		return -EINVAL;

1242
	if (!len) {
1243 1244
		kfree(dev->ifalias);
		dev->ifalias = NULL;
1245 1246 1247
		return 0;
	}

1248 1249
	new_ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL);
	if (!new_ifalias)
1250
		return -ENOMEM;
1251
	dev->ifalias = new_ifalias;
1252 1253 1254 1255 1256 1257

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


1258
/**
S
Stephen Hemminger 已提交
1259
 *	netdev_features_change - device changes features
1260 1261 1262 1263 1264 1265
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed features.
 */
void netdev_features_change(struct net_device *dev)
{
1266
	call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev);
1267 1268 1269
}
EXPORT_SYMBOL(netdev_features_change);

L
Linus Torvalds 已提交
1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
/**
 *	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) {
1281 1282 1283 1284 1285
		struct netdev_notifier_change_info change_info;

		change_info.flags_changed = 0;
		call_netdevice_notifiers_info(NETDEV_CHANGE, dev,
					      &change_info.info);
1286
		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
L
Linus Torvalds 已提交
1287 1288
	}
}
E
Eric Dumazet 已提交
1289
EXPORT_SYMBOL(netdev_state_change);
L
Linus Torvalds 已提交
1290

1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301
/**
 * 	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)
1302
{
1303 1304 1305
	rtnl_lock();
	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, dev);
	rtnl_unlock();
1306
}
1307
EXPORT_SYMBOL(netdev_notify_peers);
1308

1309
static int __dev_open(struct net_device *dev)
L
Linus Torvalds 已提交
1310
{
1311
	const struct net_device_ops *ops = dev->netdev_ops;
1312
	int ret;
L
Linus Torvalds 已提交
1313

1314 1315
	ASSERT_RTNL();

L
Linus Torvalds 已提交
1316 1317 1318
	if (!netif_device_present(dev))
		return -ENODEV;

1319 1320 1321 1322
	/* 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
	 */
1323
	netpoll_poll_disable(dev);
1324

1325 1326 1327 1328 1329
	ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

L
Linus Torvalds 已提交
1330
	set_bit(__LINK_STATE_START, &dev->state);
1331

1332 1333
	if (ops->ndo_validate_addr)
		ret = ops->ndo_validate_addr(dev);
1334

1335 1336
	if (!ret && ops->ndo_open)
		ret = ops->ndo_open(dev);
L
Linus Torvalds 已提交
1337

1338
	netpoll_poll_enable(dev);
1339

1340 1341 1342
	if (ret)
		clear_bit(__LINK_STATE_START, &dev->state);
	else {
L
Linus Torvalds 已提交
1343
		dev->flags |= IFF_UP;
1344
		dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
1345
		dev_activate(dev);
1346
		add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
1347
	}
1348

L
Linus Torvalds 已提交
1349 1350 1351 1352
	return ret;
}

/**
1353 1354
 *	dev_open	- prepare an interface for use.
 *	@dev:	device to open
L
Linus Torvalds 已提交
1355
 *
1356 1357 1358 1359 1360 1361 1362
 *	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
1363
 */
1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374
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;

1375
	rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1376 1377 1378 1379 1380 1381
	call_netdevice_notifiers(NETDEV_UP, dev);

	return ret;
}
EXPORT_SYMBOL(dev_open);

1382
static int __dev_close_many(struct list_head *head)
L
Linus Torvalds 已提交
1383
{
1384
	struct net_device *dev;
1385

1386
	ASSERT_RTNL();
1387 1388
	might_sleep();

1389
	list_for_each_entry(dev, head, close_list) {
1390
		/* Temporarily disable netpoll until the interface is down */
1391
		netpoll_poll_disable(dev);
1392

1393
		call_netdevice_notifiers(NETDEV_GOING_DOWN, dev);
L
Linus Torvalds 已提交
1394

1395
		clear_bit(__LINK_STATE_START, &dev->state);
L
Linus Torvalds 已提交
1396

1397 1398 1399 1400 1401 1402
		/* 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.
		 */
1403
		smp_mb__after_atomic(); /* Commit netif_running(). */
1404
	}
L
Linus Torvalds 已提交
1405

1406
	dev_deactivate_many(head);
1407

1408
	list_for_each_entry(dev, head, close_list) {
1409
		const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
1410

1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421
		/*
		 *	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;
1422
		netpoll_poll_enable(dev);
1423 1424 1425 1426 1427 1428 1429
	}

	return 0;
}

static int __dev_close(struct net_device *dev)
{
1430
	int retval;
1431 1432
	LIST_HEAD(single);

1433
	list_add(&dev->close_list, &single);
1434 1435
	retval = __dev_close_many(&single);
	list_del(&single);
1436

1437
	return retval;
1438 1439
}

1440
int dev_close_many(struct list_head *head, bool unlink)
1441 1442
{
	struct net_device *dev, *tmp;
L
Linus Torvalds 已提交
1443

1444 1445
	/* Remove the devices that don't need to be closed */
	list_for_each_entry_safe(dev, tmp, head, close_list)
1446
		if (!(dev->flags & IFF_UP))
1447
			list_del_init(&dev->close_list);
1448 1449

	__dev_close_many(head);
L
Linus Torvalds 已提交
1450

1451
	list_for_each_entry_safe(dev, tmp, head, close_list) {
1452
		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1453
		call_netdevice_notifiers(NETDEV_DOWN, dev);
1454 1455
		if (unlink)
			list_del_init(&dev->close_list);
1456
	}
1457 1458 1459

	return 0;
}
1460
EXPORT_SYMBOL(dev_close_many);
1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472

/**
 *	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)
{
1473 1474
	if (dev->flags & IFF_UP) {
		LIST_HEAD(single);
L
Linus Torvalds 已提交
1475

1476
		list_add(&dev->close_list, &single);
1477
		dev_close_many(&single, true);
1478 1479
		list_del(&single);
	}
1480
	return 0;
L
Linus Torvalds 已提交
1481
}
E
Eric Dumazet 已提交
1482
EXPORT_SYMBOL(dev_close);
L
Linus Torvalds 已提交
1483 1484


1485 1486 1487 1488 1489 1490 1491 1492 1493 1494
/**
 *	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)
{
1495 1496
	struct net_device *lower_dev;
	struct list_head *iter;
1497

M
Michał Mirosław 已提交
1498 1499
	dev->wanted_features &= ~NETIF_F_LRO;
	netdev_update_features(dev);
1500

1501 1502
	if (unlikely(dev->features & NETIF_F_LRO))
		netdev_WARN(dev, "failed to disable LRO!\n");
1503 1504 1505

	netdev_for_each_lower_dev(dev, lower_dev, iter)
		dev_disable_lro(lower_dev);
1506 1507 1508
}
EXPORT_SYMBOL(dev_disable_lro);

1509 1510 1511 1512 1513 1514 1515 1516
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);
}
1517

1518 1519
static int dev_boot_phase = 1;

L
Linus Torvalds 已提交
1520 1521 1522 1523 1524 1525 1526 1527 1528 1529
/**
 *	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
1530
 *	to the new notifier to allow device to have a race free
L
Linus Torvalds 已提交
1531 1532 1533 1534 1535 1536
 *	view of the network device list.
 */

int register_netdevice_notifier(struct notifier_block *nb)
{
	struct net_device *dev;
1537
	struct net_device *last;
1538
	struct net *net;
L
Linus Torvalds 已提交
1539 1540 1541
	int err;

	rtnl_lock();
1542
	err = raw_notifier_chain_register(&netdev_chain, nb);
1543 1544
	if (err)
		goto unlock;
1545 1546 1547 1548
	if (dev_boot_phase)
		goto unlock;
	for_each_net(net) {
		for_each_netdev(net, dev) {
1549
			err = call_netdevice_notifier(nb, NETDEV_REGISTER, dev);
1550 1551 1552 1553 1554 1555
			err = notifier_to_errno(err);
			if (err)
				goto rollback;

			if (!(dev->flags & IFF_UP))
				continue;
L
Linus Torvalds 已提交
1556

1557
			call_netdevice_notifier(nb, NETDEV_UP, dev);
1558
		}
L
Linus Torvalds 已提交
1559
	}
1560 1561

unlock:
L
Linus Torvalds 已提交
1562 1563
	rtnl_unlock();
	return err;
1564 1565 1566

rollback:
	last = dev;
1567 1568 1569
	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev == last)
1570
				goto outroll;
1571

1572
			if (dev->flags & IFF_UP) {
1573 1574 1575
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1576
			}
1577
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1578 1579
		}
	}
1580

1581
outroll:
1582
	raw_notifier_chain_unregister(&netdev_chain, nb);
1583
	goto unlock;
L
Linus Torvalds 已提交
1584
}
E
Eric Dumazet 已提交
1585
EXPORT_SYMBOL(register_netdevice_notifier);
L
Linus Torvalds 已提交
1586 1587 1588 1589 1590 1591 1592 1593 1594

/**
 *	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.
1595 1596 1597 1598
 *
 * 	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
1599 1600 1601 1602
 */

int unregister_netdevice_notifier(struct notifier_block *nb)
{
1603 1604
	struct net_device *dev;
	struct net *net;
1605 1606 1607
	int err;

	rtnl_lock();
1608
	err = raw_notifier_chain_unregister(&netdev_chain, nb);
1609 1610 1611 1612 1613 1614
	if (err)
		goto unlock;

	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev->flags & IFF_UP) {
1615 1616 1617
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1618
			}
1619
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1620 1621 1622
		}
	}
unlock:
1623 1624
	rtnl_unlock();
	return err;
L
Linus Torvalds 已提交
1625
}
E
Eric Dumazet 已提交
1626
EXPORT_SYMBOL(unregister_netdevice_notifier);
L
Linus Torvalds 已提交
1627

1628 1629 1630 1631 1632 1633 1634 1635 1636 1637
/**
 *	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
1638 1639 1640
static int call_netdevice_notifiers_info(unsigned long val,
					 struct net_device *dev,
					 struct netdev_notifier_info *info)
1641 1642 1643 1644 1645 1646
{
	ASSERT_RTNL();
	netdev_notifier_info_init(info, dev);
	return raw_notifier_call_chain(&netdev_chain, val, info);
}

L
Linus Torvalds 已提交
1647 1648 1649
/**
 *	call_netdevice_notifiers - call all network notifier blocks
 *      @val: value passed unmodified to notifier function
1650
 *      @dev: net_device pointer passed unmodified to notifier function
L
Linus Torvalds 已提交
1651 1652
 *
 *	Call all network notifier blocks.  Parameters and return value
1653
 *	are as for raw_notifier_call_chain().
L
Linus Torvalds 已提交
1654 1655
 */

1656
int call_netdevice_notifiers(unsigned long val, struct net_device *dev)
L
Linus Torvalds 已提交
1657
{
1658 1659 1660
	struct netdev_notifier_info info;

	return call_netdevice_notifiers_info(val, dev, &info);
L
Linus Torvalds 已提交
1661
}
1662
EXPORT_SYMBOL(call_netdevice_notifiers);
L
Linus Torvalds 已提交
1663

1664
#ifdef CONFIG_NET_INGRESS
1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679
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

D
Daniel Borkmann 已提交
1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695
#ifdef CONFIG_NET_EGRESS
static struct static_key egress_needed __read_mostly;

void net_inc_egress_queue(void)
{
	static_key_slow_inc(&egress_needed);
}
EXPORT_SYMBOL_GPL(net_inc_egress_queue);

void net_dec_egress_queue(void)
{
	static_key_slow_dec(&egress_needed);
}
EXPORT_SYMBOL_GPL(net_dec_egress_queue);
#endif

1696
static struct static_key netstamp_needed __read_mostly;
1697
#ifdef HAVE_JUMP_LABEL
1698
/* We are not allowed to call static_key_slow_dec() from irq context
1699
 * If net_disable_timestamp() is called from irq context, defer the
1700
 * static_key_slow_dec() calls.
1701 1702 1703
 */
static atomic_t netstamp_needed_deferred;
#endif
L
Linus Torvalds 已提交
1704 1705 1706

void net_enable_timestamp(void)
{
1707 1708 1709 1710 1711
#ifdef HAVE_JUMP_LABEL
	int deferred = atomic_xchg(&netstamp_needed_deferred, 0);

	if (deferred) {
		while (--deferred)
1712
			static_key_slow_dec(&netstamp_needed);
1713 1714 1715
		return;
	}
#endif
1716
	static_key_slow_inc(&netstamp_needed);
L
Linus Torvalds 已提交
1717
}
E
Eric Dumazet 已提交
1718
EXPORT_SYMBOL(net_enable_timestamp);
L
Linus Torvalds 已提交
1719 1720 1721

void net_disable_timestamp(void)
{
1722 1723 1724 1725 1726 1727
#ifdef HAVE_JUMP_LABEL
	if (in_interrupt()) {
		atomic_inc(&netstamp_needed_deferred);
		return;
	}
#endif
1728
	static_key_slow_dec(&netstamp_needed);
L
Linus Torvalds 已提交
1729
}
E
Eric Dumazet 已提交
1730
EXPORT_SYMBOL(net_disable_timestamp);
L
Linus Torvalds 已提交
1731

E
Eric Dumazet 已提交
1732
static inline void net_timestamp_set(struct sk_buff *skb)
L
Linus Torvalds 已提交
1733
{
1734
	skb->tstamp.tv64 = 0;
1735
	if (static_key_false(&netstamp_needed))
1736
		__net_timestamp(skb);
L
Linus Torvalds 已提交
1737 1738
}

1739
#define net_timestamp_check(COND, SKB)			\
1740
	if (static_key_false(&netstamp_needed)) {		\
1741 1742 1743
		if ((COND) && !(SKB)->tstamp.tv64)	\
			__net_timestamp(SKB);		\
	}						\
E
Eric Dumazet 已提交
1744

1745
bool is_skb_forwardable(const struct net_device *dev, const struct sk_buff *skb)
1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763
{
	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;
}
1764
EXPORT_SYMBOL_GPL(is_skb_forwardable);
1765

H
Herbert Xu 已提交
1766 1767
int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb)
{
1768 1769
	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
	    unlikely(!is_skb_forwardable(dev, skb))) {
H
Herbert Xu 已提交
1770 1771 1772 1773 1774 1775
		atomic_long_inc(&dev->rx_dropped);
		kfree_skb(skb);
		return NET_RX_DROP;
	}

	skb_scrub_packet(skb, true);
1776
	skb->priority = 0;
H
Herbert Xu 已提交
1777
	skb->protocol = eth_type_trans(skb, dev);
1778
	skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
H
Herbert Xu 已提交
1779 1780 1781 1782 1783

	return 0;
}
EXPORT_SYMBOL_GPL(__dev_forward_skb);

1784 1785 1786 1787 1788 1789 1790 1791
/**
 * dev_forward_skb - loopback an skb to another netif
 *
 * @dev: destination network device
 * @skb: buffer to forward
 *
 * return values:
 *	NET_RX_SUCCESS	(no congestion)
1792
 *	NET_RX_DROP     (packet was dropped, but freed)
1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803
 *
 * 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
1804
	return __dev_forward_skb(dev, skb) ?: netif_rx_internal(skb);
1805 1806 1807
}
EXPORT_SYMBOL_GPL(dev_forward_skb);

1808 1809 1810 1811
static inline int deliver_skb(struct sk_buff *skb,
			      struct packet_type *pt_prev,
			      struct net_device *orig_dev)
{
1812 1813
	if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
		return -ENOMEM;
1814 1815 1816 1817
	atomic_inc(&skb->users);
	return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
}

1818 1819
static inline void deliver_ptype_list_skb(struct sk_buff *skb,
					  struct packet_type **pt,
1820 1821
					  struct net_device *orig_dev,
					  __be16 type,
1822 1823 1824 1825 1826 1827 1828 1829
					  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)
1830
			deliver_skb(skb, pt_prev, orig_dev);
1831 1832 1833 1834 1835
		pt_prev = ptype;
	}
	*pt = pt_prev;
}

1836 1837
static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
{
1838
	if (!ptype->af_packet_priv || !skb->sk)
1839 1840 1841 1842 1843 1844 1845 1846 1847 1848
		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
1849 1850 1851 1852 1853
/*
 *	Support routine. Sends outgoing frames to any network
 *	taps currently in use.
 */

1854
void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
L
Linus Torvalds 已提交
1855 1856
{
	struct packet_type *ptype;
1857 1858
	struct sk_buff *skb2 = NULL;
	struct packet_type *pt_prev = NULL;
1859
	struct list_head *ptype_list = &ptype_all;
1860

L
Linus Torvalds 已提交
1861
	rcu_read_lock();
1862 1863
again:
	list_for_each_entry_rcu(ptype, ptype_list, list) {
L
Linus Torvalds 已提交
1864 1865 1866
		/* Never send packets back to the socket
		 * they originated from - MvS (miquels@drinkel.ow.org)
		 */
1867 1868
		if (skb_loop_sk(ptype, skb))
			continue;
1869

1870 1871 1872 1873 1874
		if (pt_prev) {
			deliver_skb(skb2, pt_prev, skb->dev);
			pt_prev = ptype;
			continue;
		}
L
Linus Torvalds 已提交
1875

1876 1877 1878 1879
		/* need to clone skb, done only once */
		skb2 = skb_clone(skb, GFP_ATOMIC);
		if (!skb2)
			goto out_unlock;
1880

1881
		net_timestamp_set(skb2);
L
Linus Torvalds 已提交
1882

1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894
		/* 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
1895
		}
1896 1897 1898 1899 1900 1901 1902 1903 1904

		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
1905
	}
1906
out_unlock:
1907 1908
	if (pt_prev)
		pt_prev->func(skb2, skb->dev, pt_prev, skb->dev);
L
Linus Torvalds 已提交
1909 1910
	rcu_read_unlock();
}
1911
EXPORT_SYMBOL_GPL(dev_queue_xmit_nit);
L
Linus Torvalds 已提交
1912

1913 1914
/**
 * netif_setup_tc - Handle tc mappings on real_num_tx_queues change
1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925
 * @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
1926
static void netif_setup_tc(struct net_device *dev, unsigned int txq)
1927 1928 1929 1930 1931 1932
{
	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) {
1933
		pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n");
1934 1935 1936 1937 1938 1939 1940 1941 1942 1943
		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) {
1944 1945
			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);
1946 1947 1948 1949 1950
			netdev_set_prio_tc_map(dev, i, 0);
		}
	}
}

1951 1952 1953 1954 1955
#ifdef CONFIG_XPS
static DEFINE_MUTEX(xps_map_mutex);
#define xmap_dereference(P)		\
	rcu_dereference_protected((P), lockdep_is_held(&xps_map_mutex))

1956 1957
static struct xps_map *remove_xps_queue(struct xps_dev_maps *dev_maps,
					int cpu, u16 index)
1958
{
1959 1960
	struct xps_map *map = NULL;
	int pos;
1961

1962 1963
	if (dev_maps)
		map = xmap_dereference(dev_maps->cpu_map[cpu]);
1964

1965 1966
	for (pos = 0; map && pos < map->len; pos++) {
		if (map->queues[pos] == index) {
1967 1968 1969
			if (map->len > 1) {
				map->queues[pos] = map->queues[--map->len];
			} else {
1970
				RCU_INIT_POINTER(dev_maps->cpu_map[cpu], NULL);
1971 1972 1973
				kfree_rcu(map, rcu);
				map = NULL;
			}
1974
			break;
1975 1976 1977
		}
	}

1978 1979 1980
	return map;
}

1981
static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
1982 1983
{
	struct xps_dev_maps *dev_maps;
1984
	int cpu, i;
1985 1986 1987 1988 1989 1990 1991 1992 1993
	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) {
1994 1995 1996 1997 1998
		for (i = index; i < dev->num_tx_queues; i++) {
			if (!remove_xps_queue(dev_maps, cpu, i))
				break;
		}
		if (i == dev->num_tx_queues)
1999 2000 2001 2002
			active = true;
	}

	if (!active) {
2003 2004 2005 2006
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

2007 2008 2009 2010
	for (i = index; i < dev->num_tx_queues; i++)
		netdev_queue_numa_node_write(netdev_get_tx_queue(dev, i),
					     NUMA_NO_NODE);

2011 2012 2013 2014
out_no_maps:
	mutex_unlock(&xps_map_mutex);
}

2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049
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;
}

2050 2051
int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
			u16 index)
2052
{
2053
	struct xps_dev_maps *dev_maps, *new_dev_maps = NULL;
2054 2055
	struct xps_map *map, *new_map;
	int maps_sz = max_t(unsigned int, XPS_DEV_MAPS_SIZE, L1_CACHE_BYTES);
2056 2057
	int cpu, numa_node_id = -2;
	bool active = false;
2058 2059 2060 2061 2062

	mutex_lock(&xps_map_mutex);

	dev_maps = xmap_dereference(dev->xps_maps);

2063 2064 2065 2066 2067 2068 2069
	/* 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);
2070 2071
		if (!new_dev_maps) {
			mutex_unlock(&xps_map_mutex);
2072
			return -ENOMEM;
2073
		}
2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087

		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;

2088
	for_each_possible_cpu(cpu) {
2089 2090 2091 2092 2093 2094 2095 2096 2097 2098
		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;
2099 2100 2101 2102 2103 2104
#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
2105 2106 2107 2108
		} 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);
2109
		}
2110

2111 2112
	}

2113 2114
	rcu_assign_pointer(dev->xps_maps, new_dev_maps);

2115
	/* Cleanup old maps */
2116 2117 2118 2119 2120 2121 2122
	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);
		}
2123

2124
		kfree_rcu(dev_maps, rcu);
2125 2126
	}

2127 2128
	dev_maps = new_dev_maps;
	active = true;
2129

2130 2131
out_no_new_maps:
	/* update Tx queue numa node */
2132 2133 2134 2135
	netdev_queue_numa_node_write(netdev_get_tx_queue(dev, index),
				     (numa_node_id >= 0) ? numa_node_id :
				     NUMA_NO_NODE);

2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154
	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:
2155 2156 2157 2158
	mutex_unlock(&xps_map_mutex);

	return 0;
error:
2159 2160 2161 2162 2163 2164 2165 2166 2167
	/* 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);
	}

2168 2169 2170 2171 2172 2173 2174 2175
	mutex_unlock(&xps_map_mutex);

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

#endif
2176 2177 2178 2179
/*
 * 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.
 */
2180
int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
2181
{
T
Tom Herbert 已提交
2182 2183
	int rc;

2184 2185
	if (txq < 1 || txq > dev->num_tx_queues)
		return -EINVAL;
2186

2187 2188
	if (dev->reg_state == NETREG_REGISTERED ||
	    dev->reg_state == NETREG_UNREGISTERING) {
2189 2190
		ASSERT_RTNL();

T
Tom Herbert 已提交
2191 2192
		rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
						  txq);
T
Tom Herbert 已提交
2193 2194 2195
		if (rc)
			return rc;

2196 2197 2198
		if (dev->num_tc)
			netif_setup_tc(dev, txq);

2199
		if (txq < dev->real_num_tx_queues) {
2200
			qdisc_reset_all_tx_gt(dev, txq);
2201 2202 2203 2204
#ifdef CONFIG_XPS
			netif_reset_xps_queues_gt(dev, txq);
#endif
		}
2205
	}
2206 2207 2208

	dev->real_num_tx_queues = txq;
	return 0;
2209 2210
}
EXPORT_SYMBOL(netif_set_real_num_tx_queues);
2211

2212
#ifdef CONFIG_SYSFS
2213 2214 2215 2216 2217 2218 2219
/**
 *	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
2220 2221
 *	negative error code.  If called before registration, it always
 *	succeeds.
2222 2223 2224 2225 2226
 */
int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
{
	int rc;

T
Tom Herbert 已提交
2227 2228 2229
	if (rxq < 1 || rxq > dev->num_rx_queues)
		return -EINVAL;

2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244
	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

2245 2246
/**
 * netif_get_num_default_rss_queues - default number of RSS queues
2247 2248 2249 2250
 *
 * This routine should set an upper limit on the number of RSS queues
 * used by default by multiqueue devices.
 */
2251
int netif_get_num_default_rss_queues(void)
2252
{
2253 2254
	return is_kdump_kernel() ?
		1 : min_t(int, DEFAULT_MAX_NUM_RSS_QUEUES, num_online_cpus());
2255 2256 2257
}
EXPORT_SYMBOL(netif_get_num_default_rss_queues);

2258
static void __netif_reschedule(struct Qdisc *q)
2259
{
2260 2261
	struct softnet_data *sd;
	unsigned long flags;
2262

2263
	local_irq_save(flags);
2264
	sd = this_cpu_ptr(&softnet_data);
2265 2266 2267
	q->next_sched = NULL;
	*sd->output_queue_tailp = q;
	sd->output_queue_tailp = &q->next_sched;
2268 2269 2270 2271 2272 2273 2274 2275
	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);
2276 2277 2278
}
EXPORT_SYMBOL(__netif_schedule);

2279 2280 2281 2282 2283
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)
2284
{
2285 2286 2287
	return (struct dev_kfree_skb_cb *)skb->cb;
}

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 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334
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);

2335
void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason)
2336
{
2337
	unsigned long flags;
2338

2339 2340 2341 2342 2343
	if (likely(atomic_read(&skb->users) == 1)) {
		smp_rmb();
		atomic_set(&skb->users, 0);
	} else if (likely(!atomic_dec_and_test(&skb->users))) {
		return;
2344
	}
2345 2346 2347 2348 2349 2350
	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);
2351
}
2352
EXPORT_SYMBOL(__dev_kfree_skb_irq);
2353

2354
void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason)
2355 2356
{
	if (in_irq() || irqs_disabled())
2357
		__dev_kfree_skb_irq(skb, reason);
2358 2359 2360
	else
		dev_kfree_skb(skb);
}
2361
EXPORT_SYMBOL(__dev_kfree_skb_any);
2362 2363


2364 2365 2366 2367 2368 2369
/**
 * netif_device_detach - mark device as removed
 * @dev: network device
 *
 * Mark device as removed from system and therefore no longer available.
 */
2370 2371 2372 2373
void netif_device_detach(struct net_device *dev)
{
	if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2374
		netif_tx_stop_all_queues(dev);
2375 2376 2377 2378
	}
}
EXPORT_SYMBOL(netif_device_detach);

2379 2380 2381 2382 2383 2384
/**
 * netif_device_attach - mark device as attached
 * @dev: network device
 *
 * Mark device as attached from system and restart if needed.
 */
2385 2386 2387 2388
void netif_device_attach(struct net_device *dev)
{
	if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2389
		netif_tx_wake_all_queues(dev);
2390
		__netdev_watchdog_up(dev);
2391 2392 2393 2394
	}
}
EXPORT_SYMBOL(netif_device_attach);

J
Jiri Pirko 已提交
2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422
/*
 * 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);

2423 2424
static void skb_warn_bad_offload(const struct sk_buff *skb)
{
W
Wei Tang 已提交
2425
	static const netdev_features_t null_features;
2426
	struct net_device *dev = skb->dev;
2427
	const char *name = "";
2428

2429 2430 2431
	if (!net_ratelimit())
		return;

2432 2433 2434 2435 2436 2437
	if (dev) {
		if (dev->dev.parent)
			name = dev_driver_string(dev->dev.parent);
		else
			name = netdev_name(dev);
	}
2438 2439
	WARN(1, "%s: caps=(%pNF, %pNF) len=%d data_len=%d gso_size=%d "
	     "gso_type=%d ip_summed=%d\n",
2440
	     name, dev ? &dev->features : &null_features,
2441
	     skb->sk ? &skb->sk->sk_route_caps : &null_features,
2442 2443 2444 2445
	     skb->len, skb->data_len, skb_shinfo(skb)->gso_size,
	     skb_shinfo(skb)->gso_type, skb->ip_summed);
}

L
Linus Torvalds 已提交
2446 2447 2448 2449
/*
 * Invalidate hardware checksum when packet is to be mangled, and
 * complete checksum manually on outgoing path.
 */
2450
int skb_checksum_help(struct sk_buff *skb)
L
Linus Torvalds 已提交
2451
{
2452
	__wsum csum;
2453
	int ret = 0, offset;
L
Linus Torvalds 已提交
2454

2455
	if (skb->ip_summed == CHECKSUM_COMPLETE)
2456 2457 2458
		goto out_set_summed;

	if (unlikely(skb_shinfo(skb)->gso_size)) {
2459 2460
		skb_warn_bad_offload(skb);
		return -EINVAL;
L
Linus Torvalds 已提交
2461 2462
	}

2463 2464 2465 2466 2467 2468 2469 2470 2471
	/* 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;
	}

2472
	offset = skb_checksum_start_offset(skb);
2473 2474 2475 2476 2477 2478 2479 2480
	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
2481 2482 2483 2484 2485
		ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
		if (ret)
			goto out;
	}

2486
	*(__sum16 *)(skb->data + offset) = csum_fold(csum);
2487
out_set_summed:
L
Linus Torvalds 已提交
2488
	skb->ip_summed = CHECKSUM_NONE;
2489
out:
L
Linus Torvalds 已提交
2490 2491
	return ret;
}
E
Eric Dumazet 已提交
2492
EXPORT_SYMBOL(skb_checksum_help);
L
Linus Torvalds 已提交
2493

2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 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 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628
/* skb_csum_offload_check - Driver helper function to determine if a device
 * with limited checksum offload capabilities is able to offload the checksum
 * for a given packet.
 *
 * Arguments:
 *   skb - sk_buff for the packet in question
 *   spec - contains the description of what device can offload
 *   csum_encapped - returns true if the checksum being offloaded is
 *	      encpasulated. That is it is checksum for the transport header
 *	      in the inner headers.
 *   checksum_help - when set indicates that helper function should
 *	      call skb_checksum_help if offload checks fail
 *
 * Returns:
 *   true: Packet has passed the checksum checks and should be offloadable to
 *	   the device (a driver may still need to check for additional
 *	   restrictions of its device)
 *   false: Checksum is not offloadable. If checksum_help was set then
 *	   skb_checksum_help was called to resolve checksum for non-GSO
 *	   packets and when IP protocol is not SCTP
 */
bool __skb_csum_offload_chk(struct sk_buff *skb,
			    const struct skb_csum_offl_spec *spec,
			    bool *csum_encapped,
			    bool csum_help)
{
	struct iphdr *iph;
	struct ipv6hdr *ipv6;
	void *nhdr;
	int protocol;
	u8 ip_proto;

	if (skb->protocol == htons(ETH_P_8021Q) ||
	    skb->protocol == htons(ETH_P_8021AD)) {
		if (!spec->vlan_okay)
			goto need_help;
	}

	/* We check whether the checksum refers to a transport layer checksum in
	 * the outermost header or an encapsulated transport layer checksum that
	 * corresponds to the inner headers of the skb. If the checksum is for
	 * something else in the packet we need help.
	 */
	if (skb_checksum_start_offset(skb) == skb_transport_offset(skb)) {
		/* Non-encapsulated checksum */
		protocol = eproto_to_ipproto(vlan_get_protocol(skb));
		nhdr = skb_network_header(skb);
		*csum_encapped = false;
		if (spec->no_not_encapped)
			goto need_help;
	} else if (skb->encapsulation && spec->encap_okay &&
		   skb_checksum_start_offset(skb) ==
		   skb_inner_transport_offset(skb)) {
		/* Encapsulated checksum */
		*csum_encapped = true;
		switch (skb->inner_protocol_type) {
		case ENCAP_TYPE_ETHER:
			protocol = eproto_to_ipproto(skb->inner_protocol);
			break;
		case ENCAP_TYPE_IPPROTO:
			protocol = skb->inner_protocol;
			break;
		}
		nhdr = skb_inner_network_header(skb);
	} else {
		goto need_help;
	}

	switch (protocol) {
	case IPPROTO_IP:
		if (!spec->ipv4_okay)
			goto need_help;
		iph = nhdr;
		ip_proto = iph->protocol;
		if (iph->ihl != 5 && !spec->ip_options_okay)
			goto need_help;
		break;
	case IPPROTO_IPV6:
		if (!spec->ipv6_okay)
			goto need_help;
		if (spec->no_encapped_ipv6 && *csum_encapped)
			goto need_help;
		ipv6 = nhdr;
		nhdr += sizeof(*ipv6);
		ip_proto = ipv6->nexthdr;
		break;
	default:
		goto need_help;
	}

ip_proto_again:
	switch (ip_proto) {
	case IPPROTO_TCP:
		if (!spec->tcp_okay ||
		    skb->csum_offset != offsetof(struct tcphdr, check))
			goto need_help;
		break;
	case IPPROTO_UDP:
		if (!spec->udp_okay ||
		    skb->csum_offset != offsetof(struct udphdr, check))
			goto need_help;
		break;
	case IPPROTO_SCTP:
		if (!spec->sctp_okay ||
		    skb->csum_offset != offsetof(struct sctphdr, checksum))
			goto cant_help;
		break;
	case NEXTHDR_HOP:
	case NEXTHDR_ROUTING:
	case NEXTHDR_DEST: {
		u8 *opthdr = nhdr;

		if (protocol != IPPROTO_IPV6 || !spec->ext_hdrs_okay)
			goto need_help;

		ip_proto = opthdr[0];
		nhdr += (opthdr[1] + 1) << 3;

		goto ip_proto_again;
	}
	default:
		goto need_help;
	}

	/* Passed the tests for offloading checksum */
	return true;

need_help:
	if (csum_help && !skb_shinfo(skb)->gso_size)
		skb_checksum_help(skb);
cant_help:
	return false;
}
EXPORT_SYMBOL(__skb_csum_offload_chk);

2629
__be16 skb_network_protocol(struct sk_buff *skb, int *depth)
2630
{
A
Al Viro 已提交
2631
	__be16 type = skb->protocol;
2632

2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643
	/* 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;
	}

2644
	return __vlan_get_protocol(skb, type, depth);
2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656
}

/**
 *	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;
2657 2658
	int vlan_depth = skb->mac_len;
	__be16 type = skb_network_protocol(skb, &vlan_depth);
2659 2660 2661 2662

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

2663
	__skb_pull(skb, vlan_depth);
2664 2665

	rcu_read_lock();
2666
	list_for_each_entry_rcu(ptype, &offload_base, list) {
2667 2668
		if (ptype->type == type && ptype->callbacks.gso_segment) {
			segs = ptype->callbacks.gso_segment(skb, features);
2669 2670 2671 2672 2673
			break;
		}
	}
	rcu_read_unlock();

2674
	__skb_push(skb, skb->data - skb_mac_header(skb));
2675

2676 2677
	return segs;
}
2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700
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.
2701 2702
 *
 *	Segmentation preserves SKB_SGO_CB_OFFSET bytes of previous skb cb.
2703 2704 2705 2706 2707 2708 2709 2710 2711
 */
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);

2712 2713
		err = skb_cow_head(skb, 0);
		if (err < 0)
2714 2715 2716
			return ERR_PTR(err);
	}

2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729
	/* Only report GSO partial support if it will enable us to
	 * support segmentation on this frame without needing additional
	 * work.
	 */
	if (features & NETIF_F_GSO_PARTIAL) {
		netdev_features_t partial_features = NETIF_F_GSO_ROBUST;
		struct net_device *dev = skb->dev;

		partial_features |= dev->features & dev->gso_partial_features;
		if (!skb_gso_ok(skb, features | partial_features))
			features &= ~NETIF_F_GSO_PARTIAL;
	}

2730 2731 2732
	BUILD_BUG_ON(SKB_SGO_CB_OFFSET +
		     sizeof(*SKB_GSO_CB(skb)) > sizeof(skb->cb));

2733
	SKB_GSO_CB(skb)->mac_offset = skb_headroom(skb);
2734 2735
	SKB_GSO_CB(skb)->encap_level = 0;

2736 2737 2738 2739 2740
	skb_reset_mac_header(skb);
	skb_reset_mac_len(skb);

	return skb_mac_gso_segment(skb, features);
}
2741
EXPORT_SYMBOL(__skb_gso_segment);
2742

2743 2744 2745 2746 2747
/* Take action when hardware reception checksum errors are detected. */
#ifdef CONFIG_BUG
void netdev_rx_csum_fault(struct net_device *dev)
{
	if (net_ratelimit()) {
2748
		pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>");
2749 2750 2751 2752 2753 2754
		dump_stack();
	}
}
EXPORT_SYMBOL(netdev_rx_csum_fault);
#endif

L
Linus Torvalds 已提交
2755 2756 2757 2758 2759
/* 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.
 */

2760
static int illegal_highdma(struct net_device *dev, struct sk_buff *skb)
L
Linus Torvalds 已提交
2761
{
2762
#ifdef CONFIG_HIGHMEM
L
Linus Torvalds 已提交
2763
	int i;
2764
	if (!(dev->features & NETIF_F_HIGHDMA)) {
2765 2766 2767
		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)))
2768
				return 1;
2769
		}
2770
	}
L
Linus Torvalds 已提交
2771

2772 2773
	if (PCI_DMA_BUS_IS_PHYS) {
		struct device *pdev = dev->dev.parent;
L
Linus Torvalds 已提交
2774

E
Eric Dumazet 已提交
2775 2776
		if (!pdev)
			return 0;
2777
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2778 2779
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			dma_addr_t addr = page_to_phys(skb_frag_page(frag));
2780 2781 2782 2783
			if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask)
				return 1;
		}
	}
2784
#endif
L
Linus Torvalds 已提交
2785 2786 2787
	return 0;
}

2788 2789 2790
/* 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
2791
#if IS_ENABLED(CONFIG_NET_MPLS_GSO)
2792 2793 2794 2795
static netdev_features_t net_mpls_features(struct sk_buff *skb,
					   netdev_features_t features,
					   __be16 type)
{
2796
	if (eth_p_mpls(type))
2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809
		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

2810
static netdev_features_t harmonize_features(struct sk_buff *skb,
2811
	netdev_features_t features)
2812
{
2813
	int tmp;
2814 2815 2816 2817
	__be16 type;

	type = skb_network_protocol(skb, &tmp);
	features = net_mpls_features(skb, features, type);
2818

2819
	if (skb->ip_summed != CHECKSUM_NONE &&
2820
	    !can_checksum_protocol(features, type)) {
2821
		features &= ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
2822
	} else if (illegal_highdma(skb->dev, skb)) {
2823 2824 2825 2826 2827 2828
		features &= ~NETIF_F_SG;
	}

	return features;
}

2829 2830 2831 2832 2833 2834 2835 2836
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);

2837 2838 2839 2840 2841 2842 2843
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);
}

2844 2845 2846 2847 2848 2849 2850 2851 2852
static netdev_features_t gso_features_check(const struct sk_buff *skb,
					    struct net_device *dev,
					    netdev_features_t features)
{
	u16 gso_segs = skb_shinfo(skb)->gso_segs;

	if (gso_segs > dev->gso_max_segs)
		return features & ~NETIF_F_GSO_MASK;

2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863
	/* Support for GSO partial features requires software
	 * intervention before we can actually process the packets
	 * so we need to strip support for any partial features now
	 * and we can pull them back in after we have partially
	 * segmented the frame.
	 */
	if (!(skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL))
		features &= ~dev->gso_partial_features;

	/* Make sure to clear the IPv4 ID mangling feature if the
	 * IPv4 header has the potential to be fragmented.
2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875
	 */
	if (skb_shinfo(skb)->gso_type & SKB_GSO_TCPV4) {
		struct iphdr *iph = skb->encapsulation ?
				    inner_ip_hdr(skb) : ip_hdr(skb);

		if (!(iph->frag_off & htons(IP_DF)))
			features &= ~NETIF_F_TSO_MANGLEID;
	}

	return features;
}

2876
netdev_features_t netif_skb_features(struct sk_buff *skb)
2877
{
2878
	struct net_device *dev = skb->dev;
2879
	netdev_features_t features = dev->features;
2880

2881 2882
	if (skb_is_gso(skb))
		features = gso_features_check(skb, dev, features);
2883

2884 2885 2886 2887 2888 2889 2890
	/* 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;

2891 2892 2893 2894 2895
	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);
2896

2897 2898 2899
	if (dev->netdev_ops->ndo_features_check)
		features &= dev->netdev_ops->ndo_features_check(skb, dev,
								features);
2900 2901
	else
		features &= dflt_features_check(skb, dev, features);
2902

2903
	return harmonize_features(skb, features);
2904
}
2905
EXPORT_SYMBOL(netif_skb_features);
2906

2907
static int xmit_one(struct sk_buff *skb, struct net_device *dev,
2908
		    struct netdev_queue *txq, bool more)
2909
{
2910 2911
	unsigned int len;
	int rc;
2912

2913
	if (!list_empty(&ptype_all) || !list_empty(&dev->ptype_all))
2914
		dev_queue_xmit_nit(skb, dev);
2915

2916 2917
	len = skb->len;
	trace_net_dev_start_xmit(skb, dev);
2918
	rc = netdev_start_xmit(skb, dev, txq, more);
2919
	trace_net_dev_xmit(skb, rc, dev, len);
E
Eric Dumazet 已提交
2920

2921 2922
	return rc;
}
2923

2924 2925
struct sk_buff *dev_hard_start_xmit(struct sk_buff *first, struct net_device *dev,
				    struct netdev_queue *txq, int *ret)
2926 2927 2928
{
	struct sk_buff *skb = first;
	int rc = NETDEV_TX_OK;
2929

2930 2931
	while (skb) {
		struct sk_buff *next = skb->next;
2932

2933
		skb->next = NULL;
2934
		rc = xmit_one(skb, dev, txq, next != NULL);
2935 2936 2937 2938
		if (unlikely(!dev_xmit_complete(rc))) {
			skb->next = next;
			goto out;
		}
2939

2940 2941 2942 2943
		skb = next;
		if (netif_xmit_stopped(txq) && skb) {
			rc = NETDEV_TX_BUSY;
			break;
2944
		}
2945
	}
2946

2947 2948 2949 2950
out:
	*ret = rc;
	return skb;
}
2951

E
Eric Dumazet 已提交
2952 2953
static struct sk_buff *validate_xmit_vlan(struct sk_buff *skb,
					  netdev_features_t features)
2954
{
2955
	if (skb_vlan_tag_present(skb) &&
2956 2957
	    !vlan_hw_offload_capable(features, skb->vlan_proto))
		skb = __vlan_hwaccel_push_inside(skb);
2958 2959
	return skb;
}
2960

2961
static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev)
2962 2963
{
	netdev_features_t features;
2964

2965 2966 2967 2968
	features = netif_skb_features(skb);
	skb = validate_xmit_vlan(skb, features);
	if (unlikely(!skb))
		goto out_null;
2969

2970
	if (netif_needs_gso(skb, features)) {
2971 2972 2973
		struct sk_buff *segs;

		segs = skb_gso_segment(skb, features);
2974
		if (IS_ERR(segs)) {
2975
			goto out_kfree_skb;
2976 2977 2978
		} else if (segs) {
			consume_skb(skb);
			skb = segs;
2979
		}
2980 2981 2982 2983
	} else {
		if (skb_needs_linearize(skb, features) &&
		    __skb_linearize(skb))
			goto out_kfree_skb;
2984

2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995
		/* 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));
2996
			if (!(features & NETIF_F_CSUM_MASK) &&
2997 2998
			    skb_checksum_help(skb))
				goto out_kfree_skb;
2999
		}
3000
	}
3001

3002
	return skb;
3003

3004 3005
out_kfree_skb:
	kfree_skb(skb);
3006
out_null:
E
Eric Dumazet 已提交
3007
	atomic_long_inc(&dev->tx_dropped);
3008 3009
	return NULL;
}
3010

3011 3012 3013 3014
struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev)
{
	struct sk_buff *next, *head = NULL, *tail;

3015
	for (; skb != NULL; skb = next) {
3016 3017
		next = skb->next;
		skb->next = NULL;
3018 3019 3020 3021

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

3022
		skb = validate_xmit_skb(skb, dev);
3023 3024
		if (!skb)
			continue;
3025

3026 3027 3028 3029 3030 3031 3032 3033
		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;
3034 3035
	}
	return head;
3036 3037
}

3038 3039 3040 3041 3042 3043 3044 3045 3046 3047
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)  {
3048
		unsigned int hdr_len;
3049
		u16 gso_segs = shinfo->gso_segs;
3050

3051 3052 3053 3054
		/* mac layer + network layer */
		hdr_len = skb_transport_header(skb) - skb_mac_header(skb);

		/* + transport layer */
3055 3056 3057 3058
		if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)))
			hdr_len += tcp_hdrlen(skb);
		else
			hdr_len += sizeof(struct udphdr);
3059 3060 3061 3062 3063 3064

		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;
3065 3066 3067
	}
}

3068 3069 3070 3071 3072
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);
3073
	struct sk_buff *to_free = NULL;
E
Eric Dumazet 已提交
3074
	bool contended;
3075 3076
	int rc;

E
Eric Dumazet 已提交
3077
	qdisc_calculate_pkt_len(skb, q);
3078 3079 3080
	/*
	 * Heuristic to force contended enqueues to serialize on a
	 * separate lock before trying to get qdisc main lock.
3081
	 * This permits qdisc->running owner to get the lock more
Y
Ying Xue 已提交
3082
	 * often and dequeue packets faster.
3083
	 */
E
Eric Dumazet 已提交
3084
	contended = qdisc_is_running(q);
3085 3086 3087
	if (unlikely(contended))
		spin_lock(&q->busylock);

3088 3089
	spin_lock(root_lock);
	if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) {
3090
		__qdisc_drop(skb, &to_free);
3091 3092
		rc = NET_XMIT_DROP;
	} else if ((q->flags & TCQ_F_CAN_BYPASS) && !qdisc_qlen(q) &&
3093
		   qdisc_run_begin(q)) {
3094 3095 3096 3097 3098
		/*
		 * 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.
		 */
3099 3100 3101

		qdisc_bstats_update(q, skb);

3102
		if (sch_direct_xmit(skb, q, dev, txq, root_lock, true)) {
3103 3104 3105 3106
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
3107
			__qdisc_run(q);
3108
		} else
3109
			qdisc_run_end(q);
3110 3111 3112

		rc = NET_XMIT_SUCCESS;
	} else {
3113
		rc = q->enqueue(skb, q, &to_free) & NET_XMIT_MASK;
3114 3115 3116 3117 3118 3119 3120
		if (qdisc_run_begin(q)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
			__qdisc_run(q);
		}
3121 3122
	}
	spin_unlock(root_lock);
3123 3124
	if (unlikely(to_free))
		kfree_skb_list(to_free);
3125 3126
	if (unlikely(contended))
		spin_unlock(&q->busylock);
3127 3128 3129
	return rc;
}

3130
#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
3131 3132
static void skb_update_prio(struct sk_buff *skb)
{
I
Igor Maravic 已提交
3133
	struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap);
3134

3135
	if (!skb->priority && skb->sk && map) {
3136 3137
		unsigned int prioidx =
			sock_cgroup_prioidx(&skb->sk->sk_cgrp_data);
3138 3139 3140 3141

		if (prioidx < map->priomap_len)
			skb->priority = map->priomap[prioidx];
	}
3142 3143 3144 3145 3146
}
#else
#define skb_update_prio(skb)
#endif

3147 3148 3149
DEFINE_PER_CPU(int, xmit_recursion);
EXPORT_SYMBOL(xmit_recursion);

3150 3151
/**
 *	dev_loopback_xmit - loop back @skb
3152 3153
 *	@net: network namespace this loopback is happening in
 *	@sk:  sk needed to be a netfilter okfn
3154 3155
 *	@skb: buffer to transmit
 */
3156
int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168
{
	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);

D
Daniel Borkmann 已提交
3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191
#ifdef CONFIG_NET_EGRESS
static struct sk_buff *
sch_handle_egress(struct sk_buff *skb, int *ret, struct net_device *dev)
{
	struct tcf_proto *cl = rcu_dereference_bh(dev->egress_cl_list);
	struct tcf_result cl_res;

	if (!cl)
		return skb;

	/* skb->tc_verd and qdisc_skb_cb(skb)->pkt_len were already set
	 * earlier by the caller.
	 */
	qdisc_bstats_cpu_update(cl->q, skb);

	switch (tc_classify(skb, cl, &cl_res, false)) {
	case TC_ACT_OK:
	case TC_ACT_RECLASSIFY:
		skb->tc_index = TC_H_MIN(cl_res.classid);
		break;
	case TC_ACT_SHOT:
		qdisc_qstats_cpu_drop(cl->q);
		*ret = NET_XMIT_DROP;
3192 3193
		kfree_skb(skb);
		return NULL;
D
Daniel Borkmann 已提交
3194 3195 3196
	case TC_ACT_STOLEN:
	case TC_ACT_QUEUED:
		*ret = NET_XMIT_SUCCESS;
3197
		consume_skb(skb);
D
Daniel Borkmann 已提交
3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211
		return NULL;
	case TC_ACT_REDIRECT:
		/* No need to push/pop skb's mac_header here on egress! */
		skb_do_redirect(skb);
		*ret = NET_XMIT_SUCCESS;
		return NULL;
	default:
		break;
	}

	return skb;
}
#endif /* CONFIG_NET_EGRESS */

3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253
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 &&
3254
		    sk_fullsock(sk) &&
3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270
		    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
3271 3272 3273
	u32 sender_cpu = skb->sender_cpu - 1;

	if (sender_cpu >= (u32)NR_CPUS)
3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292
		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);
}

3293
/**
3294
 *	__dev_queue_xmit - transmit a buffer
3295
 *	@skb: buffer to transmit
3296
 *	@accel_priv: private data used for L2 forwarding offload
3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318
 *
 *	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
 */
3319
static int __dev_queue_xmit(struct sk_buff *skb, void *accel_priv)
L
Linus Torvalds 已提交
3320 3321
{
	struct net_device *dev = skb->dev;
3322
	struct netdev_queue *txq;
L
Linus Torvalds 已提交
3323 3324 3325
	struct Qdisc *q;
	int rc = -ENOMEM;

3326 3327
	skb_reset_mac_header(skb);

3328 3329 3330
	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_SCHED_TSTAMP))
		__skb_tstamp_tx(skb, NULL, skb->sk, SCM_TSTAMP_SCHED);

3331 3332
	/* Disable soft irqs for various locks below. Also
	 * stops preemption for RCU.
L
Linus Torvalds 已提交
3333
	 */
3334
	rcu_read_lock_bh();
L
Linus Torvalds 已提交
3335

3336 3337
	skb_update_prio(skb);

D
Daniel Borkmann 已提交
3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348
	qdisc_pkt_len_init(skb);
#ifdef CONFIG_NET_CLS_ACT
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS);
# ifdef CONFIG_NET_EGRESS
	if (static_key_false(&egress_needed)) {
		skb = sch_handle_egress(skb, &rc, dev);
		if (!skb)
			goto out;
	}
# endif
#endif
3349 3350 3351 3352 3353 3354 3355 3356
	/* 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);

3357 3358 3359 3360 3361 3362 3363 3364 3365 3366
#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

3367
	txq = netdev_pick_tx(dev, skb, accel_priv);
3368
	q = rcu_dereference_bh(txq->qdisc);
3369

3370
	trace_net_dev_queue(skb);
L
Linus Torvalds 已提交
3371
	if (q->enqueue) {
3372
		rc = __dev_xmit_skb(skb, q, dev, txq);
3373
		goto out;
L
Linus Torvalds 已提交
3374 3375 3376 3377 3378
	}

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

H
Herbert Xu 已提交
3379 3380
	   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
3381 3382 3383 3384 3385 3386 3387 3388 3389 3390
	   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 */

3391
		if (txq->xmit_lock_owner != cpu) {
3392 3393
			if (unlikely(__this_cpu_read(xmit_recursion) >
				     XMIT_RECURSION_LIMIT))
3394 3395
				goto recursion_alert;

3396 3397
			skb = validate_xmit_skb(skb, dev);
			if (!skb)
E
Eric Dumazet 已提交
3398
				goto out;
3399

3400
			HARD_TX_LOCK(dev, txq, cpu);
L
Linus Torvalds 已提交
3401

3402
			if (!netif_xmit_stopped(txq)) {
3403
				__this_cpu_inc(xmit_recursion);
3404
				skb = dev_hard_start_xmit(skb, dev, txq, &rc);
3405
				__this_cpu_dec(xmit_recursion);
3406
				if (dev_xmit_complete(rc)) {
3407
					HARD_TX_UNLOCK(dev, txq);
L
Linus Torvalds 已提交
3408 3409 3410
					goto out;
				}
			}
3411
			HARD_TX_UNLOCK(dev, txq);
3412 3413
			net_crit_ratelimited("Virtual device %s asks to queue packet!\n",
					     dev->name);
L
Linus Torvalds 已提交
3414 3415
		} else {
			/* Recursion is detected! It is possible,
3416 3417 3418
			 * unfortunately
			 */
recursion_alert:
3419 3420
			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
					     dev->name);
L
Linus Torvalds 已提交
3421 3422 3423 3424
		}
	}

	rc = -ENETDOWN;
3425
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
3426

3427
	atomic_long_inc(&dev->tx_dropped);
3428
	kfree_skb_list(skb);
L
Linus Torvalds 已提交
3429 3430
	return rc;
out:
3431
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
3432 3433
	return rc;
}
3434

3435
int dev_queue_xmit(struct sk_buff *skb)
3436 3437 3438
{
	return __dev_queue_xmit(skb, NULL);
}
3439
EXPORT_SYMBOL(dev_queue_xmit);
L
Linus Torvalds 已提交
3440

3441 3442 3443 3444 3445 3446
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
3447 3448 3449 3450 3451

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

3452
int netdev_max_backlog __read_mostly = 1000;
E
Eric Dumazet 已提交
3453 3454
EXPORT_SYMBOL(netdev_max_backlog);

E
Eric Dumazet 已提交
3455
int netdev_tstamp_prequeue __read_mostly = 1;
3456 3457
int netdev_budget __read_mostly = 300;
int weight_p __read_mostly = 64;            /* old backlog weight */
L
Linus Torvalds 已提交
3458

E
Eric Dumazet 已提交
3459 3460 3461 3462 3463 3464 3465 3466
/* 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);
}

3467 3468 3469
#ifdef CONFIG_RPS

/* One global table that all flow-based protocols share. */
E
Eric Dumazet 已提交
3470
struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
3471
EXPORT_SYMBOL(rps_sock_flow_table);
3472 3473
u32 rps_cpu_mask __read_mostly;
EXPORT_SYMBOL(rps_cpu_mask);
3474

3475
struct static_key rps_needed __read_mostly;
3476
EXPORT_SYMBOL(rps_needed);
3477

3478 3479 3480 3481
static struct rps_dev_flow *
set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
	    struct rps_dev_flow *rflow, u16 next_cpu)
{
3482
	if (next_cpu < nr_cpu_ids) {
3483 3484 3485 3486 3487 3488 3489 3490 3491
#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? */
3492 3493
		if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
		    !(dev->features & NETIF_F_NTUPLE))
3494 3495 3496 3497 3498 3499 3500 3501 3502
			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;
3503
		flow_id = skb_get_hash(skb) & flow_table->mask;
3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515
		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 =
3516
			per_cpu(softnet_data, next_cpu).input_queue_head;
3517 3518
	}

3519
	rflow->cpu = next_cpu;
3520 3521 3522
	return rflow;
}

3523 3524 3525 3526 3527 3528 3529 3530
/*
 * 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)
{
3531 3532
	const struct rps_sock_flow_table *sock_flow_table;
	struct netdev_rx_queue *rxqueue = dev->_rx;
3533
	struct rps_dev_flow_table *flow_table;
3534
	struct rps_map *map;
3535
	int cpu = -1;
3536
	u32 tcpu;
3537
	u32 hash;
3538 3539 3540

	if (skb_rx_queue_recorded(skb)) {
		u16 index = skb_get_rx_queue(skb);
3541

3542 3543 3544 3545 3546
		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);
3547 3548
			goto done;
		}
3549 3550
		rxqueue += index;
	}
3551

3552 3553 3554
	/* 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
3555
	map = rcu_dereference(rxqueue->rps_map);
3556
	if (!flow_table && !map)
3557 3558
		goto done;

3559
	skb_reset_network_header(skb);
3560 3561
	hash = skb_get_hash(skb);
	if (!hash)
3562 3563
		goto done;

T
Tom Herbert 已提交
3564 3565 3566
	sock_flow_table = rcu_dereference(rps_sock_flow_table);
	if (flow_table && sock_flow_table) {
		struct rps_dev_flow *rflow;
3567 3568 3569 3570 3571 3572 3573
		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
3574

3575 3576 3577 3578 3579
		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
		 */
3580
		rflow = &flow_table->flows[hash & flow_table->mask];
T
Tom Herbert 已提交
3581 3582 3583 3584 3585 3586
		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:
3587
		 *   - Current CPU is unset (>= nr_cpu_ids).
T
Tom Herbert 已提交
3588 3589 3590 3591 3592 3593 3594
		 *   - 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) &&
3595
		    (tcpu >= nr_cpu_ids || !cpu_online(tcpu) ||
T
Tom Herbert 已提交
3596
		     ((int)(per_cpu(softnet_data, tcpu).input_queue_head -
3597 3598
		      rflow->last_qtail)) >= 0)) {
			tcpu = next_cpu;
3599
			rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
3600
		}
3601

3602
		if (tcpu < nr_cpu_ids && cpu_online(tcpu)) {
T
Tom Herbert 已提交
3603 3604 3605 3606 3607 3608
			*rflowp = rflow;
			cpu = tcpu;
			goto done;
		}
	}

3609 3610
try_rps:

T
Tom Herbert 已提交
3611
	if (map) {
3612
		tcpu = map->cpus[reciprocal_scale(hash, map->len)];
T
Tom Herbert 已提交
3613 3614 3615 3616 3617 3618 3619 3620 3621 3622
		if (cpu_online(tcpu)) {
			cpu = tcpu;
			goto done;
		}
	}

done:
	return cpu;
}

3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642
#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;
3643
	unsigned int cpu;
3644 3645 3646 3647 3648 3649

	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);
3650
		if (rflow->filter == filter_id && cpu < nr_cpu_ids &&
3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662
		    ((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
3663
/* Called from hardirq (IPI) context */
E
Eric Dumazet 已提交
3664
static void rps_trigger_softirq(void *data)
T
Tom Herbert 已提交
3665
{
E
Eric Dumazet 已提交
3666 3667
	struct softnet_data *sd = data;

E
Eric Dumazet 已提交
3668
	____napi_schedule(sd, &sd->backlog);
C
Changli Gao 已提交
3669
	sd->received_rps++;
T
Tom Herbert 已提交
3670
}
E
Eric Dumazet 已提交
3671

T
Tom Herbert 已提交
3672
#endif /* CONFIG_RPS */
T
Tom Herbert 已提交
3673

E
Eric Dumazet 已提交
3674 3675 3676 3677 3678 3679 3680 3681
/*
 * 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
3682
	struct softnet_data *mysd = this_cpu_ptr(&softnet_data);
E
Eric Dumazet 已提交
3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694

	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;
}

3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708
#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;

3709
	sd = this_cpu_ptr(&softnet_data);
3710 3711 3712 3713

	rcu_read_lock();
	fl = rcu_dereference(sd->flow_limit);
	if (fl) {
3714
		new_flow = skb_get_hash(skb) & (fl->num_buckets - 1);
3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734
		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
3735 3736 3737 3738
/*
 * 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
3739 3740
static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
			      unsigned int *qtail)
T
Tom Herbert 已提交
3741
{
E
Eric Dumazet 已提交
3742
	struct softnet_data *sd;
T
Tom Herbert 已提交
3743
	unsigned long flags;
3744
	unsigned int qlen;
T
Tom Herbert 已提交
3745

E
Eric Dumazet 已提交
3746
	sd = &per_cpu(softnet_data, cpu);
T
Tom Herbert 已提交
3747 3748 3749

	local_irq_save(flags);

E
Eric Dumazet 已提交
3750
	rps_lock(sd);
3751 3752
	if (!netif_running(skb->dev))
		goto drop;
3753 3754
	qlen = skb_queue_len(&sd->input_pkt_queue);
	if (qlen <= netdev_max_backlog && !skb_flow_limit(skb, qlen)) {
3755
		if (qlen) {
T
Tom Herbert 已提交
3756
enqueue:
E
Eric Dumazet 已提交
3757
			__skb_queue_tail(&sd->input_pkt_queue, skb);
3758
			input_queue_tail_incr_save(sd, qtail);
E
Eric Dumazet 已提交
3759
			rps_unlock(sd);
3760
			local_irq_restore(flags);
T
Tom Herbert 已提交
3761 3762 3763
			return NET_RX_SUCCESS;
		}

3764 3765 3766 3767
		/* 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
3768
			if (!rps_ipi_queued(sd))
E
Eric Dumazet 已提交
3769
				____napi_schedule(sd, &sd->backlog);
T
Tom Herbert 已提交
3770 3771 3772 3773
		}
		goto enqueue;
	}

3774
drop:
C
Changli Gao 已提交
3775
	sd->dropped++;
E
Eric Dumazet 已提交
3776
	rps_unlock(sd);
T
Tom Herbert 已提交
3777 3778 3779

	local_irq_restore(flags);

3780
	atomic_long_inc(&skb->dev->rx_dropped);
T
Tom Herbert 已提交
3781 3782 3783
	kfree_skb(skb);
	return NET_RX_DROP;
}
L
Linus Torvalds 已提交
3784

3785
static int netif_rx_internal(struct sk_buff *skb)
L
Linus Torvalds 已提交
3786
{
3787
	int ret;
L
Linus Torvalds 已提交
3788

3789
	net_timestamp_check(netdev_tstamp_prequeue, skb);
L
Linus Torvalds 已提交
3790

3791
	trace_netif_rx(skb);
E
Eric Dumazet 已提交
3792
#ifdef CONFIG_RPS
3793
	if (static_key_false(&rps_needed)) {
T
Tom Herbert 已提交
3794
		struct rps_dev_flow voidflow, *rflow = &voidflow;
3795 3796
		int cpu;

3797
		preempt_disable();
3798
		rcu_read_lock();
T
Tom Herbert 已提交
3799 3800

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
3801 3802
		if (cpu < 0)
			cpu = smp_processor_id();
T
Tom Herbert 已提交
3803 3804 3805

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

3806
		rcu_read_unlock();
3807
		preempt_enable();
3808 3809
	} else
#endif
T
Tom Herbert 已提交
3810 3811 3812 3813 3814
	{
		unsigned int qtail;
		ret = enqueue_to_backlog(skb, get_cpu(), &qtail);
		put_cpu();
	}
3815
	return ret;
L
Linus Torvalds 已提交
3816
}
3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838

/**
 *	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
3839
EXPORT_SYMBOL(netif_rx);
L
Linus Torvalds 已提交
3840 3841 3842 3843 3844

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

3845 3846
	trace_netif_rx_ni_entry(skb);

L
Linus Torvalds 已提交
3847
	preempt_disable();
3848
	err = netif_rx_internal(skb);
L
Linus Torvalds 已提交
3849 3850 3851 3852 3853 3854 3855 3856 3857 3858
	if (local_softirq_pending())
		do_softirq();
	preempt_enable();

	return err;
}
EXPORT_SYMBOL(netif_rx_ni);

static void net_tx_action(struct softirq_action *h)
{
3859
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
L
Linus Torvalds 已提交
3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872

	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;

3873
			WARN_ON(atomic_read(&skb->users));
3874 3875 3876 3877
			if (likely(get_kfree_skb_cb(skb)->reason == SKB_REASON_CONSUMED))
				trace_consume_skb(skb);
			else
				trace_kfree_skb(skb, net_tx_action);
3878 3879 3880 3881 3882

			if (skb->fclone != SKB_FCLONE_UNAVAILABLE)
				__kfree_skb(skb);
			else
				__kfree_skb_defer(skb);
L
Linus Torvalds 已提交
3883
		}
3884 3885

		__kfree_skb_flush();
L
Linus Torvalds 已提交
3886 3887 3888
	}

	if (sd->output_queue) {
3889
		struct Qdisc *head;
L
Linus Torvalds 已提交
3890 3891 3892 3893

		local_irq_disable();
		head = sd->output_queue;
		sd->output_queue = NULL;
3894
		sd->output_queue_tailp = &sd->output_queue;
L
Linus Torvalds 已提交
3895 3896 3897
		local_irq_enable();

		while (head) {
3898 3899 3900
			struct Qdisc *q = head;
			spinlock_t *root_lock;

L
Linus Torvalds 已提交
3901 3902
			head = head->next_sched;

3903
			root_lock = qdisc_lock(q);
3904 3905 3906 3907 3908 3909 3910 3911
			spin_lock(root_lock);
			/* We need to make sure head->next_sched is read
			 * before clearing __QDISC_STATE_SCHED
			 */
			smp_mb__before_atomic();
			clear_bit(__QDISC_STATE_SCHED, &q->state);
			qdisc_run(q);
			spin_unlock(root_lock);
L
Linus Torvalds 已提交
3912 3913 3914 3915
		}
	}
}

3916 3917
#if (defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)) && \
    (defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE))
3918 3919 3920
/* This hook is defined here for ATM LANE */
int (*br_fdb_test_addr_hook)(struct net_device *dev,
			     unsigned char *addr) __read_mostly;
3921
EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
3922
#endif
L
Linus Torvalds 已提交
3923

D
Daniel Borkmann 已提交
3924 3925 3926
static inline struct sk_buff *
sch_handle_ingress(struct sk_buff *skb, struct packet_type **pt_prev, int *ret,
		   struct net_device *orig_dev)
3927
{
3928
#ifdef CONFIG_NET_CLS_ACT
3929 3930
	struct tcf_proto *cl = rcu_dereference_bh(skb->dev->ingress_cl_list);
	struct tcf_result cl_res;
3931

3932 3933 3934
	/* 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
3935
	 * out here.
3936
	 */
3937
	if (!cl)
3938
		return skb;
3939 3940 3941
	if (*pt_prev) {
		*ret = deliver_skb(skb, *pt_prev, orig_dev);
		*pt_prev = NULL;
L
Linus Torvalds 已提交
3942 3943
	}

3944
	qdisc_skb_cb(skb)->pkt_len = skb->len;
3945
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS);
3946
	qdisc_bstats_cpu_update(cl->q, skb);
3947

3948
	switch (tc_classify(skb, cl, &cl_res, false)) {
3949 3950 3951 3952 3953
	case TC_ACT_OK:
	case TC_ACT_RECLASSIFY:
		skb->tc_index = TC_H_MIN(cl_res.classid);
		break;
	case TC_ACT_SHOT:
3954
		qdisc_qstats_cpu_drop(cl->q);
3955 3956
		kfree_skb(skb);
		return NULL;
3957 3958
	case TC_ACT_STOLEN:
	case TC_ACT_QUEUED:
3959
		consume_skb(skb);
3960
		return NULL;
3961 3962 3963 3964 3965 3966 3967 3968
	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;
3969 3970
	default:
		break;
3971
	}
3972
#endif /* CONFIG_NET_CLS_ACT */
3973 3974
	return skb;
}
L
Linus Torvalds 已提交
3975

3976 3977 3978 3979
/**
 *	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
3980
 *	@rx_handler_data: data pointer that is used by rx handler
3981
 *
3982
 *	Register a receive handler for a device. This handler will then be
3983 3984 3985 3986
 *	called from __netif_receive_skb. A negative errno code is returned
 *	on a failure.
 *
 *	The caller must hold the rtnl_mutex.
3987 3988
 *
 *	For a general description of rx_handler, see enum rx_handler_result.
3989 3990
 */
int netdev_rx_handler_register(struct net_device *dev,
J
Jiri Pirko 已提交
3991 3992
			       rx_handler_func_t *rx_handler,
			       void *rx_handler_data)
3993 3994 3995 3996 3997 3998
{
	ASSERT_RTNL();

	if (dev->rx_handler)
		return -EBUSY;

3999
	/* Note: rx_handler_data must be set before rx_handler */
J
Jiri Pirko 已提交
4000
	rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
4001 4002 4003 4004 4005 4006 4007 4008 4009 4010
	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
4011
 *	Unregister a receive handler from a device.
4012 4013 4014 4015 4016 4017 4018
 *
 *	The caller must hold the rtnl_mutex.
 */
void netdev_rx_handler_unregister(struct net_device *dev)
{

	ASSERT_RTNL();
4019
	RCU_INIT_POINTER(dev->rx_handler, NULL);
4020 4021 4022 4023 4024
	/* 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();
4025
	RCU_INIT_POINTER(dev->rx_handler_data, NULL);
4026 4027 4028
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);

4029 4030 4031 4032 4033 4034 4035
/*
 * 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) {
4036 4037 4038 4039 4040
	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):
4041 4042 4043 4044 4045 4046
		return true;
	default:
		return false;
	}
}

4047 4048 4049
static inline int nf_ingress(struct sk_buff *skb, struct packet_type **pt_prev,
			     int *ret, struct net_device *orig_dev)
{
4050
#ifdef CONFIG_NETFILTER_INGRESS
4051 4052 4053 4054 4055 4056 4057 4058
	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);
	}
4059
#endif /* CONFIG_NETFILTER_INGRESS */
4060 4061 4062
	return 0;
}

4063
static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc)
L
Linus Torvalds 已提交
4064 4065
{
	struct packet_type *ptype, *pt_prev;
4066
	rx_handler_func_t *rx_handler;
D
David S. Miller 已提交
4067
	struct net_device *orig_dev;
4068
	bool deliver_exact = false;
L
Linus Torvalds 已提交
4069
	int ret = NET_RX_DROP;
A
Al Viro 已提交
4070
	__be16 type;
L
Linus Torvalds 已提交
4071

4072
	net_timestamp_check(!netdev_tstamp_prequeue, skb);
4073

4074
	trace_netif_receive_skb(skb);
4075

J
Joe Eykholt 已提交
4076
	orig_dev = skb->dev;
4077

4078
	skb_reset_network_header(skb);
4079 4080
	if (!skb_transport_header_was_set(skb))
		skb_reset_transport_header(skb);
4081
	skb_reset_mac_len(skb);
L
Linus Torvalds 已提交
4082 4083 4084

	pt_prev = NULL;

4085
another_round:
4086
	skb->skb_iif = skb->dev->ifindex;
4087 4088 4089

	__this_cpu_inc(softnet_data.processed);

P
Patrick McHardy 已提交
4090 4091
	if (skb->protocol == cpu_to_be16(ETH_P_8021Q) ||
	    skb->protocol == cpu_to_be16(ETH_P_8021AD)) {
4092
		skb = skb_vlan_untag(skb);
4093
		if (unlikely(!skb))
4094
			goto out;
4095 4096
	}

L
Linus Torvalds 已提交
4097 4098 4099 4100 4101 4102 4103
#ifdef CONFIG_NET_CLS_ACT
	if (skb->tc_verd & TC_NCLS) {
		skb->tc_verd = CLR_TC_NCLS(skb->tc_verd);
		goto ncls;
	}
#endif

4104
	if (pfmemalloc)
4105 4106
		goto skip_taps;

L
Linus Torvalds 已提交
4107
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
4108 4109 4110 4111 4112 4113 4114 4115 4116
		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
4117 4118
	}

4119
skip_taps:
4120
#ifdef CONFIG_NET_INGRESS
4121
	if (static_key_false(&ingress_needed)) {
D
Daniel Borkmann 已提交
4122
		skb = sch_handle_ingress(skb, &pt_prev, &ret, orig_dev);
4123
		if (!skb)
4124
			goto out;
4125 4126

		if (nf_ingress(skb, &pt_prev, &ret, orig_dev) < 0)
4127
			goto out;
4128
	}
4129 4130
#endif
#ifdef CONFIG_NET_CLS_ACT
4131
	skb->tc_verd = 0;
L
Linus Torvalds 已提交
4132 4133
ncls:
#endif
4134
	if (pfmemalloc && !skb_pfmemalloc_protocol(skb))
4135 4136
		goto drop;

4137
	if (skb_vlan_tag_present(skb)) {
4138 4139 4140 4141
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
4142
		if (vlan_do_receive(&skb))
4143 4144
			goto another_round;
		else if (unlikely(!skb))
4145
			goto out;
4146 4147
	}

4148
	rx_handler = rcu_dereference(skb->dev->rx_handler);
4149 4150 4151 4152 4153
	if (rx_handler) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
4154 4155
		switch (rx_handler(&skb)) {
		case RX_HANDLER_CONSUMED:
4156
			ret = NET_RX_SUCCESS;
4157
			goto out;
4158
		case RX_HANDLER_ANOTHER:
4159
			goto another_round;
4160 4161 4162 4163 4164 4165 4166
		case RX_HANDLER_EXACT:
			deliver_exact = true;
		case RX_HANDLER_PASS:
			break;
		default:
			BUG();
		}
4167
	}
L
Linus Torvalds 已提交
4168

4169 4170
	if (unlikely(skb_vlan_tag_present(skb))) {
		if (skb_vlan_tag_get_id(skb))
E
Eric Dumazet 已提交
4171 4172 4173 4174 4175 4176 4177
			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;
	}
4178

4179 4180
	type = skb->protocol;

4181
	/* deliver only exact match when indicated */
4182 4183 4184 4185 4186
	if (likely(!deliver_exact)) {
		deliver_ptype_list_skb(skb, &pt_prev, orig_dev, type,
				       &ptype_base[ntohs(type) &
						   PTYPE_HASH_MASK]);
	}
4187

4188 4189 4190 4191 4192 4193
	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
4194 4195 4196
	}

	if (pt_prev) {
4197
		if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
4198
			goto drop;
4199 4200
		else
			ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
L
Linus Torvalds 已提交
4201
	} else {
4202
drop:
4203 4204 4205 4206
		if (!deliver_exact)
			atomic_long_inc(&skb->dev->rx_dropped);
		else
			atomic_long_inc(&skb->dev->rx_nohandler);
L
Linus Torvalds 已提交
4207 4208 4209 4210 4211 4212 4213
		kfree_skb(skb);
		/* Jamal, now you will not able to escape explaining
		 * me how you were going to use this. :-)
		 */
		ret = NET_RX_DROP;
	}

4214
out:
4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239
	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
4240 4241
	return ret;
}
T
Tom Herbert 已提交
4242

4243
static int netif_receive_skb_internal(struct sk_buff *skb)
T
Tom Herbert 已提交
4244
{
4245 4246
	int ret;

4247
	net_timestamp_check(netdev_tstamp_prequeue, skb);
E
Eric Dumazet 已提交
4248

4249 4250 4251
	if (skb_defer_rx_timestamp(skb))
		return NET_RX_SUCCESS;

4252 4253
	rcu_read_lock();

E
Eric Dumazet 已提交
4254
#ifdef CONFIG_RPS
4255
	if (static_key_false(&rps_needed)) {
E
Eric Dumazet 已提交
4256
		struct rps_dev_flow voidflow, *rflow = &voidflow;
4257
		int cpu = get_rps_cpu(skb->dev, skb, &rflow);
T
Tom Herbert 已提交
4258

E
Eric Dumazet 已提交
4259 4260 4261
		if (cpu >= 0) {
			ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
			rcu_read_unlock();
4262
			return ret;
E
Eric Dumazet 已提交
4263
		}
T
Tom Herbert 已提交
4264
	}
4265
#endif
4266 4267 4268
	ret = __netif_receive_skb(skb);
	rcu_read_unlock();
	return ret;
T
Tom Herbert 已提交
4269
}
4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285

/**
 *	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
 */
4286
int netif_receive_skb(struct sk_buff *skb)
4287 4288 4289 4290 4291
{
	trace_netif_receive_skb_entry(skb);

	return netif_receive_skb_internal(skb);
}
4292
EXPORT_SYMBOL(netif_receive_skb);
L
Linus Torvalds 已提交
4293

E
Eric Dumazet 已提交
4294 4295 4296
/* Network device is going away, flush any packets still pending
 * Called with irqs disabled.
 */
4297
static void flush_backlog(void *arg)
4298
{
4299
	struct net_device *dev = arg;
4300
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
4301 4302
	struct sk_buff *skb, *tmp;

E
Eric Dumazet 已提交
4303
	rps_lock(sd);
4304
	skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
4305
		if (skb->dev == dev) {
E
Eric Dumazet 已提交
4306
			__skb_unlink(skb, &sd->input_pkt_queue);
4307
			kfree_skb(skb);
4308
			input_queue_head_incr(sd);
4309
		}
4310
	}
E
Eric Dumazet 已提交
4311
	rps_unlock(sd);
4312 4313 4314 4315 4316

	skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
		if (skb->dev == dev) {
			__skb_unlink(skb, &sd->process_queue);
			kfree_skb(skb);
4317
			input_queue_head_incr(sd);
4318 4319
		}
	}
4320 4321
}

4322 4323
static int napi_gro_complete(struct sk_buff *skb)
{
4324
	struct packet_offload *ptype;
4325
	__be16 type = skb->protocol;
4326
	struct list_head *head = &offload_base;
4327 4328
	int err = -ENOENT;

4329 4330
	BUILD_BUG_ON(sizeof(struct napi_gro_cb) > sizeof(skb->cb));

4331 4332
	if (NAPI_GRO_CB(skb)->count == 1) {
		skb_shinfo(skb)->gso_size = 0;
4333
		goto out;
4334
	}
4335 4336 4337

	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
4338
		if (ptype->type != type || !ptype->callbacks.gro_complete)
4339 4340
			continue;

4341
		err = ptype->callbacks.gro_complete(skb, 0);
4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352
		break;
	}
	rcu_read_unlock();

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

out:
4353
	return netif_receive_skb_internal(skb);
4354 4355
}

4356 4357 4358 4359 4360
/* 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)
4361
{
4362
	struct sk_buff *skb, *prev = NULL;
4363

4364 4365 4366 4367 4368 4369 4370
	/* 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) {
4371
		skb->next = NULL;
4372 4373 4374 4375 4376

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

		prev = skb->prev;
4377
		napi_gro_complete(skb);
4378
		napi->gro_count--;
4379 4380 4381 4382
	}

	napi->gro_list = NULL;
}
E
Eric Dumazet 已提交
4383
EXPORT_SYMBOL(napi_gro_flush);
4384

E
Eric Dumazet 已提交
4385 4386 4387 4388
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;
4389
	u32 hash = skb_get_hash_raw(skb);
E
Eric Dumazet 已提交
4390 4391 4392 4393

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

4394 4395 4396 4397 4398 4399 4400
		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
4401 4402
		diffs = (unsigned long)p->dev ^ (unsigned long)skb->dev;
		diffs |= p->vlan_tci ^ skb->vlan_tci;
4403
		diffs |= skb_metadata_dst_cmp(p, skb);
E
Eric Dumazet 已提交
4404 4405
		if (maclen == ETH_HLEN)
			diffs |= compare_ether_header(skb_mac_header(p),
4406
						      skb_mac_header(skb));
E
Eric Dumazet 已提交
4407 4408
		else if (!diffs)
			diffs = memcmp(skb_mac_header(p),
4409
				       skb_mac_header(skb),
E
Eric Dumazet 已提交
4410 4411 4412 4413 4414
				       maclen);
		NAPI_GRO_CB(p)->same_flow = !diffs;
	}
}

4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428
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
4429 4430 4431
	}
}

4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452
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
4453
static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
4454 4455
{
	struct sk_buff **pp = NULL;
4456
	struct packet_offload *ptype;
4457
	__be16 type = skb->protocol;
4458
	struct list_head *head = &offload_base;
H
Herbert Xu 已提交
4459
	int same_flow;
4460
	enum gro_result ret;
4461
	int grow;
4462

4463
	if (!(skb->dev->features & NETIF_F_GRO))
4464 4465
		goto normal;

4466
	if (skb_is_gso(skb) || skb_has_frag_list(skb) || skb->csum_bad)
4467 4468
		goto normal;

E
Eric Dumazet 已提交
4469 4470
	gro_list_prepare(napi, skb);

4471 4472
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
4473
		if (ptype->type != type || !ptype->callbacks.gro_receive)
4474 4475
			continue;

4476
		skb_set_network_header(skb, skb_gro_offset(skb));
4477
		skb_reset_mac_len(skb);
4478 4479
		NAPI_GRO_CB(skb)->same_flow = 0;
		NAPI_GRO_CB(skb)->flush = 0;
H
Herbert Xu 已提交
4480
		NAPI_GRO_CB(skb)->free = 0;
4481
		NAPI_GRO_CB(skb)->encap_mark = 0;
4482
		NAPI_GRO_CB(skb)->is_fou = 0;
4483
		NAPI_GRO_CB(skb)->is_atomic = 1;
4484
		NAPI_GRO_CB(skb)->gro_remcsum_start = 0;
4485

4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500
		/* 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;
		}
4501

4502
		pp = ptype->callbacks.gro_receive(&napi->gro_list, skb);
4503 4504 4505 4506 4507 4508 4509
		break;
	}
	rcu_read_unlock();

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

H
Herbert Xu 已提交
4510
	same_flow = NAPI_GRO_CB(skb)->same_flow;
4511
	ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED;
H
Herbert Xu 已提交
4512

4513 4514 4515 4516 4517 4518
	if (pp) {
		struct sk_buff *nskb = *pp;

		*pp = nskb->next;
		nskb->next = NULL;
		napi_gro_complete(nskb);
4519
		napi->gro_count--;
4520 4521
	}

H
Herbert Xu 已提交
4522
	if (same_flow)
4523 4524
		goto ok;

4525
	if (NAPI_GRO_CB(skb)->flush)
4526 4527
		goto normal;

4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541
	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++;
	}
4542
	NAPI_GRO_CB(skb)->count = 1;
4543
	NAPI_GRO_CB(skb)->age = jiffies;
4544
	NAPI_GRO_CB(skb)->last = skb;
4545
	skb_shinfo(skb)->gso_size = skb_gro_len(skb);
4546 4547
	skb->next = napi->gro_list;
	napi->gro_list = skb;
4548
	ret = GRO_HELD;
4549

4550
pull:
4551 4552 4553
	grow = skb_gro_offset(skb) - skb_headlen(skb);
	if (grow > 0)
		gro_pull_from_frag0(skb, grow);
4554
ok:
4555
	return ret;
4556 4557

normal:
4558 4559
	ret = GRO_NORMAL;
	goto pull;
H
Herbert Xu 已提交
4560
}
4561

4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573
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;
}
4574
EXPORT_SYMBOL(gro_find_receive_by_type);
4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587

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;
}
4588
EXPORT_SYMBOL(gro_find_complete_by_type);
H
Herbert Xu 已提交
4589

R
Rami Rosen 已提交
4590
static gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
H
Herbert Xu 已提交
4591
{
4592 4593
	switch (ret) {
	case GRO_NORMAL:
4594
		if (netif_receive_skb_internal(skb))
4595 4596
			ret = GRO_DROP;
		break;
H
Herbert Xu 已提交
4597

4598
	case GRO_DROP:
H
Herbert Xu 已提交
4599 4600
		kfree_skb(skb);
		break;
4601

4602
	case GRO_MERGED_FREE:
4603 4604
		if (NAPI_GRO_CB(skb)->free == NAPI_GRO_FREE_STOLEN_HEAD) {
			skb_dst_drop(skb);
4605
			kmem_cache_free(skbuff_head_cache, skb);
4606
		} else {
4607
			__kfree_skb(skb);
4608
		}
4609 4610
		break;

4611 4612 4613
	case GRO_HELD:
	case GRO_MERGED:
		break;
H
Herbert Xu 已提交
4614 4615
	}

4616
	return ret;
4617 4618
}

4619
gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
4620
{
4621
	skb_mark_napi_id(skb, napi);
4622
	trace_napi_gro_receive_entry(skb);
4623

4624 4625
	skb_gro_reset_offset(skb);

E
Eric Dumazet 已提交
4626
	return napi_skb_finish(dev_gro_receive(napi, skb), skb);
4627 4628 4629
}
EXPORT_SYMBOL(napi_gro_receive);

S
stephen hemminger 已提交
4630
static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb)
4631
{
4632 4633 4634 4635
	if (unlikely(skb->pfmemalloc)) {
		consume_skb(skb);
		return;
	}
4636
	__skb_pull(skb, skb_headlen(skb));
4637 4638
	/* restore the reserve we had after netdev_alloc_skb_ip_align() */
	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN - skb_headroom(skb));
4639
	skb->vlan_tci = 0;
H
Herbert Xu 已提交
4640
	skb->dev = napi->dev;
A
Andy Gospodarek 已提交
4641
	skb->skb_iif = 0;
4642 4643
	skb->encapsulation = 0;
	skb_shinfo(skb)->gso_type = 0;
4644
	skb->truesize = SKB_TRUESIZE(skb_end_offset(skb));
4645 4646 4647 4648

	napi->skb = skb;
}

4649
struct sk_buff *napi_get_frags(struct napi_struct *napi)
H
Herbert Xu 已提交
4650 4651 4652 4653
{
	struct sk_buff *skb = napi->skb;

	if (!skb) {
4654
		skb = napi_alloc_skb(napi, GRO_MAX_HEAD);
4655 4656 4657 4658
		if (skb) {
			napi->skb = skb;
			skb_mark_napi_id(skb, napi);
		}
4659
	}
4660 4661
	return skb;
}
4662
EXPORT_SYMBOL(napi_get_frags);
4663

4664 4665 4666
static gro_result_t napi_frags_finish(struct napi_struct *napi,
				      struct sk_buff *skb,
				      gro_result_t ret)
4667
{
4668 4669
	switch (ret) {
	case GRO_NORMAL:
4670 4671 4672 4673
	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))
4674
			ret = GRO_DROP;
4675
		break;
H
Herbert Xu 已提交
4676

4677 4678 4679 4680
	case GRO_DROP:
	case GRO_MERGED_FREE:
		napi_reuse_skb(napi, skb);
		break;
4681 4682 4683

	case GRO_MERGED:
		break;
4684
	}
H
Herbert Xu 已提交
4685

4686
	return ret;
H
Herbert Xu 已提交
4687
}
4688

4689 4690 4691 4692
/* 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
4693
static struct sk_buff *napi_frags_skb(struct napi_struct *napi)
4694 4695
{
	struct sk_buff *skb = napi->skb;
4696 4697
	const struct ethhdr *eth;
	unsigned int hlen = sizeof(*eth);
4698 4699 4700

	napi->skb = NULL;

4701 4702 4703 4704 4705 4706 4707
	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)) {
4708 4709
			net_warn_ratelimited("%s: dropping impossible skb from %s\n",
					     __func__, napi->dev->name);
4710 4711 4712 4713 4714 4715 4716
			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;
4717
	}
4718 4719 4720 4721 4722 4723 4724 4725
	__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;
4726 4727 4728 4729

	return skb;
}

4730
gro_result_t napi_gro_frags(struct napi_struct *napi)
4731
{
4732
	struct sk_buff *skb = napi_frags_skb(napi);
4733 4734

	if (!skb)
4735
		return GRO_DROP;
4736

4737 4738
	trace_napi_gro_frags_entry(skb);

E
Eric Dumazet 已提交
4739
	return napi_frags_finish(napi, skb, dev_gro_receive(napi, skb));
4740
}
H
Herbert Xu 已提交
4741 4742
EXPORT_SYMBOL(napi_gro_frags);

T
Tom Herbert 已提交
4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767
/* 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);

4768
/*
4769
 * net_rps_action_and_irq_enable sends any pending IPI's for rps.
4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786
 * 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))
4787
				smp_call_function_single_async(remsd->cpu,
4788
							   &remsd->csd);
4789 4790 4791 4792 4793 4794 4795
			remsd = next;
		}
	} else
#endif
		local_irq_enable();
}

E
Eric Dumazet 已提交
4796 4797 4798 4799 4800 4801 4802 4803 4804
static bool sd_has_rps_ipi_waiting(struct softnet_data *sd)
{
#ifdef CONFIG_RPS
	return sd->rps_ipi_list != NULL;
#else
	return false;
#endif
}

4805
static int process_backlog(struct napi_struct *napi, int quota)
L
Linus Torvalds 已提交
4806 4807
{
	int work = 0;
E
Eric Dumazet 已提交
4808
	struct softnet_data *sd = container_of(napi, struct softnet_data, backlog);
L
Linus Torvalds 已提交
4809

4810 4811 4812
	/* Check if we have pending ipi, its better to send them now,
	 * not waiting net_rx_action() end.
	 */
E
Eric Dumazet 已提交
4813
	if (sd_has_rps_ipi_waiting(sd)) {
4814 4815 4816
		local_irq_disable();
		net_rps_action_and_irq_enable(sd);
	}
E
Eric Dumazet 已提交
4817

4818
	napi->weight = weight_p;
4819
	local_irq_disable();
4820
	while (1) {
L
Linus Torvalds 已提交
4821
		struct sk_buff *skb;
4822 4823

		while ((skb = __skb_dequeue(&sd->process_queue))) {
4824
			rcu_read_lock();
4825 4826
			local_irq_enable();
			__netif_receive_skb(skb);
4827
			rcu_read_unlock();
4828
			local_irq_disable();
4829 4830 4831 4832 4833
			input_queue_head_incr(sd);
			if (++work >= quota) {
				local_irq_enable();
				return work;
			}
4834
		}
L
Linus Torvalds 已提交
4835

E
Eric Dumazet 已提交
4836
		rps_lock(sd);
4837
		if (skb_queue_empty(&sd->input_pkt_queue)) {
E
Eric Dumazet 已提交
4838 4839 4840
			/*
			 * Inline a custom version of __napi_complete().
			 * only current cpu owns and manipulates this napi,
4841 4842 4843
			 * 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
4844 4845 4846
			 * and we dont need an smp_mb() memory barrier.
			 */
			napi->state = 0;
4847
			rps_unlock(sd);
E
Eric Dumazet 已提交
4848

4849
			break;
4850
		}
4851 4852 4853

		skb_queue_splice_tail_init(&sd->input_pkt_queue,
					   &sd->process_queue);
E
Eric Dumazet 已提交
4854
		rps_unlock(sd);
4855 4856
	}
	local_irq_enable();
L
Linus Torvalds 已提交
4857

4858 4859
	return work;
}
L
Linus Torvalds 已提交
4860

4861 4862
/**
 * __napi_schedule - schedule for receive
4863
 * @n: entry to schedule
4864
 *
4865 4866
 * The entry's receive function will be scheduled to run.
 * Consider using __napi_schedule_irqoff() if hard irqs are masked.
4867
 */
H
Harvey Harrison 已提交
4868
void __napi_schedule(struct napi_struct *n)
4869 4870
{
	unsigned long flags;
L
Linus Torvalds 已提交
4871

4872
	local_irq_save(flags);
4873
	____napi_schedule(this_cpu_ptr(&softnet_data), n);
4874
	local_irq_restore(flags);
L
Linus Torvalds 已提交
4875
}
4876 4877
EXPORT_SYMBOL(__napi_schedule);

4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889
/**
 * __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);

4890 4891 4892 4893
void __napi_complete(struct napi_struct *n)
{
	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));

E
Eric Dumazet 已提交
4894
	list_del_init(&n->poll_list);
4895
	smp_mb__before_atomic();
4896 4897 4898 4899
	clear_bit(NAPI_STATE_SCHED, &n->state);
}
EXPORT_SYMBOL(__napi_complete);

4900
void napi_complete_done(struct napi_struct *n, int work_done)
4901 4902 4903 4904 4905 4906 4907 4908 4909 4910
{
	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;

4911 4912
	if (n->gro_list) {
		unsigned long timeout = 0;
E
Eric Dumazet 已提交
4913

4914 4915 4916 4917 4918 4919 4920 4921 4922
		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
4923 4924 4925 4926 4927 4928 4929 4930
	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);
	}
4931
}
4932
EXPORT_SYMBOL(napi_complete_done);
4933

E
Eliezer Tamir 已提交
4934
/* must be called under rcu_read_lock(), as we dont take a reference */
E
Eric Dumazet 已提交
4935
static struct napi_struct *napi_by_id(unsigned int napi_id)
E
Eliezer Tamir 已提交
4936 4937 4938 4939 4940 4941 4942 4943 4944 4945
{
	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;
}
E
Eric Dumazet 已提交
4946 4947

#if defined(CONFIG_NET_RX_BUSY_POLL)
4948
#define BUSY_POLL_BUDGET 8
E
Eric Dumazet 已提交
4949 4950 4951
bool sk_busy_loop(struct sock *sk, int nonblock)
{
	unsigned long end_time = !nonblock ? sk_busy_loop_end_time(sk) : 0;
4952
	int (*busy_poll)(struct napi_struct *dev);
E
Eric Dumazet 已提交
4953 4954 4955
	struct napi_struct *napi;
	int rc = false;

4956
	rcu_read_lock();
E
Eric Dumazet 已提交
4957 4958 4959 4960 4961

	napi = napi_by_id(sk->sk_napi_id);
	if (!napi)
		goto out;

4962 4963
	/* Note: ndo_busy_poll method is optional in linux-4.5 */
	busy_poll = napi->dev->netdev_ops->ndo_busy_poll;
E
Eric Dumazet 已提交
4964 4965

	do {
4966
		rc = 0;
4967
		local_bh_disable();
4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982
		if (busy_poll) {
			rc = busy_poll(napi);
		} else if (napi_schedule_prep(napi)) {
			void *have = netpoll_poll_lock(napi);

			if (test_bit(NAPI_STATE_SCHED, &napi->state)) {
				rc = napi->poll(napi, BUSY_POLL_BUDGET);
				trace_napi_poll(napi);
				if (rc == BUSY_POLL_BUDGET) {
					napi_complete_done(napi, rc);
					napi_schedule(napi);
				}
			}
			netpoll_poll_unlock(have);
		}
4983
		if (rc > 0)
4984 4985
			__NET_ADD_STATS(sock_net(sk),
					LINUX_MIB_BUSYPOLLRXPACKETS, rc);
4986
		local_bh_enable();
E
Eric Dumazet 已提交
4987 4988 4989 4990 4991 4992 4993 4994 4995 4996

		if (rc == LL_FLUSH_FAILED)
			break; /* permanent failure */

		cpu_relax();
	} while (!nonblock && skb_queue_empty(&sk->sk_receive_queue) &&
		 !need_resched() && !busy_loop_timeout(end_time));

	rc = !skb_queue_empty(&sk->sk_receive_queue);
out:
4997
	rcu_read_unlock();
E
Eric Dumazet 已提交
4998 4999 5000 5001 5002
	return rc;
}
EXPORT_SYMBOL(sk_busy_loop);

#endif /* CONFIG_NET_RX_BUSY_POLL */
E
Eliezer Tamir 已提交
5003 5004 5005

void napi_hash_add(struct napi_struct *napi)
{
E
Eric Dumazet 已提交
5006 5007
	if (test_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state) ||
	    test_and_set_bit(NAPI_STATE_HASHED, &napi->state))
5008
		return;
E
Eliezer Tamir 已提交
5009

5010
	spin_lock(&napi_hash_lock);
E
Eliezer Tamir 已提交
5011

5012 5013 5014 5015 5016 5017
	/* 0..NR_CPUS+1 range is reserved for sender_cpu use */
	do {
		if (unlikely(++napi_gen_id < NR_CPUS + 1))
			napi_gen_id = NR_CPUS + 1;
	} while (napi_by_id(napi_gen_id));
	napi->napi_id = napi_gen_id;
E
Eliezer Tamir 已提交
5018

5019 5020
	hlist_add_head_rcu(&napi->napi_hash_node,
			   &napi_hash[napi->napi_id % HASH_SIZE(napi_hash)]);
E
Eliezer Tamir 已提交
5021

5022
	spin_unlock(&napi_hash_lock);
E
Eliezer Tamir 已提交
5023 5024 5025 5026 5027 5028
}
EXPORT_SYMBOL_GPL(napi_hash_add);

/* Warning : caller is responsible to make sure rcu grace period
 * is respected before freeing memory containing @napi
 */
5029
bool napi_hash_del(struct napi_struct *napi)
E
Eliezer Tamir 已提交
5030
{
5031 5032
	bool rcu_sync_needed = false;

E
Eliezer Tamir 已提交
5033 5034
	spin_lock(&napi_hash_lock);

5035 5036
	if (test_and_clear_bit(NAPI_STATE_HASHED, &napi->state)) {
		rcu_sync_needed = true;
E
Eliezer Tamir 已提交
5037
		hlist_del_rcu(&napi->napi_hash_node);
5038
	}
E
Eliezer Tamir 已提交
5039
	spin_unlock(&napi_hash_lock);
5040
	return rcu_sync_needed;
E
Eliezer Tamir 已提交
5041 5042 5043
}
EXPORT_SYMBOL_GPL(napi_hash_del);

5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054
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;
}

5055 5056 5057 5058
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);
5059 5060
	hrtimer_init(&napi->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_PINNED);
	napi->timer.function = napi_watchdog;
5061
	napi->gro_count = 0;
5062
	napi->gro_list = NULL;
H
Herbert Xu 已提交
5063
	napi->skb = NULL;
5064
	napi->poll = poll;
E
Eric Dumazet 已提交
5065 5066 5067
	if (weight > NAPI_POLL_WEIGHT)
		pr_err_once("netif_napi_add() called with weight %d on device %s\n",
			    weight, dev->name);
5068 5069 5070
	napi->weight = weight;
	list_add(&napi->dev_list, &dev->napi_list);
	napi->dev = dev;
H
Herbert Xu 已提交
5071
#ifdef CONFIG_NETPOLL
5072 5073 5074 5075
	spin_lock_init(&napi->poll_lock);
	napi->poll_owner = -1;
#endif
	set_bit(NAPI_STATE_SCHED, &napi->state);
5076
	napi_hash_add(napi);
5077 5078 5079
}
EXPORT_SYMBOL(netif_napi_add);

5080 5081 5082 5083 5084 5085 5086
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);
5087 5088
	while (test_and_set_bit(NAPI_STATE_NPSVC, &n->state))
		msleep(1);
5089 5090 5091 5092 5093 5094 5095

	hrtimer_cancel(&n->timer);

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

5096
/* Must be called in process context */
5097 5098
void netif_napi_del(struct napi_struct *napi)
{
5099 5100 5101
	might_sleep();
	if (napi_hash_del(napi))
		synchronize_net();
5102
	list_del_init(&napi->dev_list);
5103
	napi_free_frags(napi);
5104

E
Eric Dumazet 已提交
5105
	kfree_skb_list(napi->gro_list);
5106
	napi->gro_list = NULL;
5107
	napi->gro_count = 0;
5108 5109 5110
}
EXPORT_SYMBOL(netif_napi_del);

5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155
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);
	}

5156 5157 5158 5159 5160 5161 5162 5163 5164
	/* 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;
	}

5165 5166 5167 5168 5169 5170 5171 5172
	list_add_tail(&n->poll_list, repoll);

out_unlock:
	netpoll_poll_unlock(have);

	return work;
}

L
Linus Torvalds 已提交
5173 5174
static void net_rx_action(struct softirq_action *h)
{
5175
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
5176
	unsigned long time_limit = jiffies + 2;
5177
	int budget = netdev_budget;
E
Eric Dumazet 已提交
5178 5179
	LIST_HEAD(list);
	LIST_HEAD(repoll);
5180

L
Linus Torvalds 已提交
5181
	local_irq_disable();
E
Eric Dumazet 已提交
5182 5183
	list_splice_init(&sd->poll_list, &list);
	local_irq_enable();
L
Linus Torvalds 已提交
5184

H
Herbert Xu 已提交
5185
	for (;;) {
5186
		struct napi_struct *n;
L
Linus Torvalds 已提交
5187

H
Herbert Xu 已提交
5188 5189 5190 5191 5192 5193
		if (list_empty(&list)) {
			if (!sd_has_rps_ipi_waiting(sd) && list_empty(&repoll))
				return;
			break;
		}

5194 5195 5196
		n = list_first_entry(&list, struct napi_struct, poll_list);
		budget -= napi_poll(n, &repoll);

E
Eric Dumazet 已提交
5197
		/* If softirq window is exhausted then punt.
5198 5199
		 * Allow this to run for 2 jiffies since which will allow
		 * an average latency of 1.5/HZ.
5200
		 */
H
Herbert Xu 已提交
5201 5202 5203 5204 5205
		if (unlikely(budget <= 0 ||
			     time_after_eq(jiffies, time_limit))) {
			sd->time_squeeze++;
			break;
		}
L
Linus Torvalds 已提交
5206
	}
E
Eric Dumazet 已提交
5207

5208
	__kfree_skb_flush();
E
Eric Dumazet 已提交
5209 5210 5211 5212 5213 5214 5215 5216
	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);

5217
	net_rps_action_and_irq_enable(sd);
L
Linus Torvalds 已提交
5218 5219
}

5220
struct netdev_adjacent {
J
Jiri Pirko 已提交
5221
	struct net_device *dev;
5222 5223

	/* upper master flag, there can only be one master device per list */
J
Jiri Pirko 已提交
5224
	bool master;
5225 5226 5227 5228

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

5229 5230 5231
	/* private field for the users */
	void *private;

J
Jiri Pirko 已提交
5232 5233 5234 5235
	struct list_head list;
	struct rcu_head rcu;
};

5236
static struct netdev_adjacent *__netdev_find_adj(struct net_device *adj_dev,
5237
						 struct list_head *adj_list)
J
Jiri Pirko 已提交
5238
{
5239 5240
	struct netdev_adjacent *adj;

5241
	list_for_each_entry(adj, adj_list, list) {
5242 5243
		if (adj->dev == adj_dev)
			return adj;
J
Jiri Pirko 已提交
5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261
	}
	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();

5262
	return __netdev_find_adj(upper_dev, &dev->all_adj_list.upper);
J
Jiri Pirko 已提交
5263 5264 5265 5266 5267 5268 5269 5270 5271 5272
}
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
5273
static bool netdev_has_any_upper_dev(struct net_device *dev)
J
Jiri Pirko 已提交
5274 5275 5276
{
	ASSERT_RTNL();

5277
	return !list_empty(&dev->all_adj_list.upper);
J
Jiri Pirko 已提交
5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288
}

/**
 * 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)
{
5289
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
5290 5291 5292

	ASSERT_RTNL();

5293
	if (list_empty(&dev->adj_list.upper))
J
Jiri Pirko 已提交
5294 5295
		return NULL;

5296
	upper = list_first_entry(&dev->adj_list.upper,
5297
				 struct netdev_adjacent, list);
J
Jiri Pirko 已提交
5298 5299 5300 5301 5302 5303
	if (likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get);

5304 5305 5306 5307 5308 5309 5310 5311 5312 5313
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);

5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339
/**
 * 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);

5340 5341
/**
 * netdev_all_upper_get_next_dev_rcu - Get the next dev from upper list
5342 5343 5344 5345 5346 5347
 * @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.
 */
5348 5349
struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
						     struct list_head **iter)
5350 5351 5352
{
	struct netdev_adjacent *upper;

5353
	WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_rtnl_is_held());
5354 5355 5356

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

5357
	if (&upper->list == &dev->all_adj_list.upper)
5358 5359 5360 5361 5362 5363
		return NULL;

	*iter = &upper->list;

	return upper->dev;
}
5364
EXPORT_SYMBOL(netdev_all_upper_get_next_dev_rcu);
5365

5366 5367 5368 5369 5370 5371 5372 5373 5374
/**
 * 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
5375
 * list will remain unchanged.
5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386
 */
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;

5387
	*iter = lower->list.next;
5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414

	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;

5415
	*iter = &lower->list;
5416 5417 5418 5419 5420

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

5421 5422 5423 5424 5425 5426 5427 5428 5429
/**
 * 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
5430
 * list will remain unchanged.
5431 5432 5433 5434 5435
 */
void *netdev_lower_get_next(struct net_device *dev, struct list_head **iter)
{
	struct netdev_adjacent *lower;

5436
	lower = list_entry(*iter, struct netdev_adjacent, list);
5437 5438 5439 5440

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

5441
	*iter = lower->list.next;
5442 5443 5444 5445 5446

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

5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467
/**
 * 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
5468 5469 5470 5471 5472 5473 5474 5475 5476
/**
 * 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)
{
5477
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
5478

5479
	upper = list_first_or_null_rcu(&dev->adj_list.upper,
5480
				       struct netdev_adjacent, list);
J
Jiri Pirko 已提交
5481 5482 5483 5484 5485 5486
	if (upper && likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get_rcu);

5487
static int netdev_adjacent_sysfs_add(struct net_device *dev,
5488 5489 5490 5491 5492 5493 5494 5495 5496
			      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);
}
5497
static void netdev_adjacent_sysfs_del(struct net_device *dev,
5498 5499 5500 5501 5502 5503 5504 5505 5506
			       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);
}

5507 5508 5509 5510 5511 5512 5513 5514
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));
}
5515

5516 5517
static int __netdev_adjacent_dev_insert(struct net_device *dev,
					struct net_device *adj_dev,
5518
					struct list_head *dev_list,
5519
					void *private, bool master)
5520 5521
{
	struct netdev_adjacent *adj;
5522
	int ret;
5523

5524
	adj = __netdev_find_adj(adj_dev, dev_list);
5525 5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537

	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;
5538
	adj->private = private;
5539
	dev_hold(adj_dev);
5540 5541 5542

	pr_debug("dev_hold for %s, because of link added from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
5543

5544
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list)) {
5545
		ret = netdev_adjacent_sysfs_add(dev, adj_dev, dev_list);
5546 5547 5548 5549
		if (ret)
			goto free_adj;
	}

5550
	/* Ensure that master link is always the first item in list. */
5551 5552 5553 5554
	if (master) {
		ret = sysfs_create_link(&(dev->dev.kobj),
					&(adj_dev->dev.kobj), "master");
		if (ret)
5555
			goto remove_symlinks;
5556

5557
		list_add_rcu(&adj->list, dev_list);
5558
	} else {
5559
		list_add_tail_rcu(&adj->list, dev_list);
5560
	}
5561 5562

	return 0;
5563

5564
remove_symlinks:
5565
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
5566
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
5567 5568
free_adj:
	kfree(adj);
5569
	dev_put(adj_dev);
5570 5571

	return ret;
5572 5573
}

S
stephen hemminger 已提交
5574 5575 5576
static void __netdev_adjacent_dev_remove(struct net_device *dev,
					 struct net_device *adj_dev,
					 struct list_head *dev_list)
5577 5578 5579
{
	struct netdev_adjacent *adj;

5580
	adj = __netdev_find_adj(adj_dev, dev_list);
5581

5582 5583 5584
	if (!adj) {
		pr_err("tried to remove device %s from %s\n",
		       dev->name, adj_dev->name);
5585
		BUG();
5586
	}
5587 5588

	if (adj->ref_nr > 1) {
5589 5590
		pr_debug("%s to %s ref_nr-- = %d\n", dev->name, adj_dev->name,
			 adj->ref_nr-1);
5591 5592 5593 5594
		adj->ref_nr--;
		return;
	}

5595 5596 5597
	if (adj->master)
		sysfs_remove_link(&(dev->dev.kobj), "master");

5598
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
5599
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
5600

5601
	list_del_rcu(&adj->list);
5602 5603
	pr_debug("dev_put for %s, because link removed from %s to %s\n",
		 adj_dev->name, dev->name, adj_dev->name);
5604 5605 5606 5607
	dev_put(adj_dev);
	kfree_rcu(adj, rcu);
}

S
stephen hemminger 已提交
5608 5609 5610 5611 5612
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)
5613 5614 5615
{
	int ret;

5616 5617
	ret = __netdev_adjacent_dev_insert(dev, upper_dev, up_list, private,
					   master);
5618 5619 5620
	if (ret)
		return ret;

5621 5622
	ret = __netdev_adjacent_dev_insert(upper_dev, dev, down_list, private,
					   false);
5623
	if (ret) {
5624
		__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
5625 5626 5627 5628 5629 5630
		return ret;
	}

	return 0;
}

S
stephen hemminger 已提交
5631 5632
static int __netdev_adjacent_dev_link(struct net_device *dev,
				      struct net_device *upper_dev)
5633
{
5634 5635 5636
	return __netdev_adjacent_dev_link_lists(dev, upper_dev,
						&dev->all_adj_list.upper,
						&upper_dev->all_adj_list.lower,
5637
						NULL, false);
5638 5639
}

S
stephen hemminger 已提交
5640 5641 5642 5643
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)
5644
{
5645 5646
	__netdev_adjacent_dev_remove(dev, upper_dev, up_list);
	__netdev_adjacent_dev_remove(upper_dev, dev, down_list);
5647 5648
}

S
stephen hemminger 已提交
5649 5650
static void __netdev_adjacent_dev_unlink(struct net_device *dev,
					 struct net_device *upper_dev)
5651
{
5652 5653 5654 5655 5656
	__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
5657 5658 5659
static int __netdev_adjacent_dev_link_neighbour(struct net_device *dev,
						struct net_device *upper_dev,
						void *private, bool master)
5660 5661 5662 5663 5664 5665 5666 5667 5668
{
	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,
5669
					       private, master);
5670 5671 5672 5673 5674 5675
	if (ret) {
		__netdev_adjacent_dev_unlink(dev, upper_dev);
		return ret;
	}

	return 0;
5676 5677
}

S
stephen hemminger 已提交
5678 5679
static void __netdev_adjacent_dev_unlink_neighbour(struct net_device *dev,
						   struct net_device *upper_dev)
5680 5681 5682 5683 5684 5685
{
	__netdev_adjacent_dev_unlink(dev, upper_dev);
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev,
					   &dev->adj_list.upper,
					   &upper_dev->adj_list.lower);
}
5686

J
Jiri Pirko 已提交
5687
static int __netdev_upper_dev_link(struct net_device *dev,
5688
				   struct net_device *upper_dev, bool master,
5689
				   void *upper_priv, void *upper_info)
J
Jiri Pirko 已提交
5690
{
5691
	struct netdev_notifier_changeupper_info changeupper_info;
5692 5693
	struct netdev_adjacent *i, *j, *to_i, *to_j;
	int ret = 0;
J
Jiri Pirko 已提交
5694 5695 5696 5697 5698 5699 5700

	ASSERT_RTNL();

	if (dev == upper_dev)
		return -EBUSY;

	/* To prevent loops, check if dev is not upper device to upper_dev. */
5701
	if (__netdev_find_adj(dev, &upper_dev->all_adj_list.upper))
J
Jiri Pirko 已提交
5702 5703
		return -EBUSY;

5704
	if (__netdev_find_adj(upper_dev, &dev->adj_list.upper))
J
Jiri Pirko 已提交
5705 5706 5707 5708 5709
		return -EEXIST;

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

5710 5711 5712
	changeupper_info.upper_dev = upper_dev;
	changeupper_info.master = master;
	changeupper_info.linking = true;
5713
	changeupper_info.upper_info = upper_info;
5714

5715 5716 5717 5718 5719 5720
	ret = call_netdevice_notifiers_info(NETDEV_PRECHANGEUPPER, dev,
					    &changeupper_info.info);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

5721
	ret = __netdev_adjacent_dev_link_neighbour(dev, upper_dev, upper_priv,
5722
						   master);
5723 5724
	if (ret)
		return ret;
J
Jiri Pirko 已提交
5725

5726
	/* Now that we linked these devs, make all the upper_dev's
5727
	 * all_adj_list.upper visible to every dev's all_adj_list.lower an
5728 5729 5730
	 * versa, and don't forget the devices itself. All of these
	 * links are non-neighbours.
	 */
5731 5732 5733 5734
	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);
5735 5736 5737 5738 5739 5740 5741
			ret = __netdev_adjacent_dev_link(i->dev, j->dev);
			if (ret)
				goto rollback_mesh;
		}
	}

	/* add dev to every upper_dev's upper device */
5742 5743 5744
	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);
5745 5746 5747 5748 5749 5750
		ret = __netdev_adjacent_dev_link(dev, i->dev);
		if (ret)
			goto rollback_upper_mesh;
	}

	/* add upper_dev to every dev's lower device */
5751 5752 5753
	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);
5754 5755 5756 5757
		ret = __netdev_adjacent_dev_link(i->dev, upper_dev);
		if (ret)
			goto rollback_lower_mesh;
	}
J
Jiri Pirko 已提交
5758

5759 5760 5761 5762 5763 5764
	ret = call_netdevice_notifiers_info(NETDEV_CHANGEUPPER, dev,
					    &changeupper_info.info);
	ret = notifier_to_errno(ret);
	if (ret)
		goto rollback_lower_mesh;

J
Jiri Pirko 已提交
5765
	return 0;
5766 5767 5768

rollback_lower_mesh:
	to_i = i;
5769
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
5770 5771 5772 5773 5774 5775 5776 5777 5778
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);
	}

	i = NULL;

rollback_upper_mesh:
	to_i = i;
5779
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list) {
5780 5781 5782 5783 5784 5785 5786 5787 5788 5789
		if (i == to_i)
			break;
		__netdev_adjacent_dev_unlink(dev, i->dev);
	}

	i = j = NULL;

rollback_mesh:
	to_i = i;
	to_j = j;
5790 5791
	list_for_each_entry(i, &dev->all_adj_list.lower, list) {
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list) {
5792 5793 5794 5795 5796 5797 5798 5799
			if (i == to_i && j == to_j)
				break;
			__netdev_adjacent_dev_unlink(i->dev, j->dev);
		}
		if (i == to_i)
			break;
	}

5800
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
5801 5802

	return ret;
J
Jiri Pirko 已提交
5803 5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817
}

/**
 * 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)
{
5818
	return __netdev_upper_dev_link(dev, upper_dev, false, NULL, NULL);
J
Jiri Pirko 已提交
5819 5820 5821 5822 5823 5824 5825
}
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
5826
 * @upper_priv: upper device private
5827
 * @upper_info: upper info to be passed down via notifier
J
Jiri Pirko 已提交
5828 5829 5830 5831 5832 5833 5834 5835
 *
 * 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,
5836
				 struct net_device *upper_dev,
5837
				 void *upper_priv, void *upper_info)
J
Jiri Pirko 已提交
5838
{
5839 5840
	return __netdev_upper_dev_link(dev, upper_dev, true,
				       upper_priv, upper_info);
J
Jiri Pirko 已提交
5841 5842 5843 5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854
}
EXPORT_SYMBOL(netdev_master_upper_dev_link);

/**
 * 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)
{
5855
	struct netdev_notifier_changeupper_info changeupper_info;
5856
	struct netdev_adjacent *i, *j;
J
Jiri Pirko 已提交
5857 5858
	ASSERT_RTNL();

5859 5860 5861 5862
	changeupper_info.upper_dev = upper_dev;
	changeupper_info.master = netdev_master_upper_dev_get(dev) == upper_dev;
	changeupper_info.linking = false;

5863 5864 5865
	call_netdevice_notifiers_info(NETDEV_PRECHANGEUPPER, dev,
				      &changeupper_info.info);

5866
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
5867 5868 5869 5870 5871

	/* 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.
	 */
5872 5873
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
		list_for_each_entry(j, &upper_dev->all_adj_list.upper, list)
5874 5875 5876 5877 5878
			__netdev_adjacent_dev_unlink(i->dev, j->dev);

	/* remove also the devices itself from lower/upper device
	 * list
	 */
5879
	list_for_each_entry(i, &dev->all_adj_list.lower, list)
5880 5881
		__netdev_adjacent_dev_unlink(i->dev, upper_dev);

5882
	list_for_each_entry(i, &upper_dev->all_adj_list.upper, list)
5883 5884
		__netdev_adjacent_dev_unlink(dev, i->dev);

5885 5886
	call_netdevice_notifiers_info(NETDEV_CHANGEUPPER, dev,
				      &changeupper_info.info);
J
Jiri Pirko 已提交
5887 5888 5889
}
EXPORT_SYMBOL(netdev_upper_dev_unlink);

5890 5891 5892
/**
 * netdev_bonding_info_change - Dispatch event about slave change
 * @dev: device
5893
 * @bonding_info: info to dispatch
5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909
 *
 * 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
5910
static void netdev_adjacent_add_links(struct net_device *dev)
5911 5912 5913 5914 5915 5916
{
	struct netdev_adjacent *iter;

	struct net *net = dev_net(dev);

	list_for_each_entry(iter, &dev->adj_list.upper, list) {
W
Wei Tang 已提交
5917
		if (!net_eq(net, dev_net(iter->dev)))
5918 5919 5920 5921 5922 5923 5924 5925
			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) {
W
Wei Tang 已提交
5926
		if (!net_eq(net, dev_net(iter->dev)))
5927 5928 5929 5930 5931 5932 5933 5934
			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
5935
static void netdev_adjacent_del_links(struct net_device *dev)
5936 5937 5938 5939 5940 5941
{
	struct netdev_adjacent *iter;

	struct net *net = dev_net(dev);

	list_for_each_entry(iter, &dev->adj_list.upper, list) {
W
Wei Tang 已提交
5942
		if (!net_eq(net, dev_net(iter->dev)))
5943 5944 5945 5946 5947 5948 5949 5950
			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) {
W
Wei Tang 已提交
5951
		if (!net_eq(net, dev_net(iter->dev)))
5952 5953 5954 5955 5956 5957 5958 5959
			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);
	}
}

5960
void netdev_adjacent_rename_links(struct net_device *dev, char *oldname)
5961
{
5962
	struct netdev_adjacent *iter;
5963

5964 5965
	struct net *net = dev_net(dev);

5966
	list_for_each_entry(iter, &dev->adj_list.upper, list) {
W
Wei Tang 已提交
5967
		if (!net_eq(net, dev_net(iter->dev)))
5968
			continue;
5969 5970 5971 5972 5973
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.lower);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.lower);
	}
5974

5975
	list_for_each_entry(iter, &dev->adj_list.lower, list) {
W
Wei Tang 已提交
5976
		if (!net_eq(net, dev_net(iter->dev)))
5977
			continue;
5978 5979 5980 5981 5982
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.upper);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.upper);
	}
5983 5984 5985 5986 5987 5988 5989 5990 5991
}

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

	if (!lower_dev)
		return NULL;
5992
	lower = __netdev_find_adj(lower_dev, &dev->adj_list.lower);
5993 5994 5995 5996 5997 5998 5999
	if (!lower)
		return NULL;

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

6000 6001

int dev_get_nest_level(struct net_device *dev,
6002
		       bool (*type_check)(const struct net_device *dev))
6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018 6019 6020 6021 6022 6023
{
	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);

6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043
/**
 * netdev_lower_change - Dispatch event about lower device state change
 * @lower_dev: device
 * @lower_state_info: state to dispatch
 *
 * Send NETDEV_CHANGELOWERSTATE to netdev notifiers with info.
 * The caller must hold the RTNL lock.
 */
void netdev_lower_state_changed(struct net_device *lower_dev,
				void *lower_state_info)
{
	struct netdev_notifier_changelowerstate_info changelowerstate_info;

	ASSERT_RTNL();
	changelowerstate_info.lower_state_info = lower_state_info;
	call_netdevice_notifiers_info(NETDEV_CHANGELOWERSTATE, lower_dev,
				      &changelowerstate_info.info);
}
EXPORT_SYMBOL(netdev_lower_state_changed);

6044 6045
static void dev_change_rx_flags(struct net_device *dev, int flags)
{
6046 6047
	const struct net_device_ops *ops = dev->netdev_ops;

6048
	if (ops->ndo_change_rx_flags)
6049
		ops->ndo_change_rx_flags(dev, flags);
6050 6051
}

6052
static int __dev_set_promiscuity(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
6053
{
6054
	unsigned int old_flags = dev->flags;
6055 6056
	kuid_t uid;
	kgid_t gid;
L
Linus Torvalds 已提交
6057

6058 6059
	ASSERT_RTNL();

6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070
	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;
6071 6072
			pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n",
				dev->name);
6073 6074 6075
			return -EOVERFLOW;
		}
	}
6076
	if (dev->flags != old_flags) {
6077 6078 6079
		pr_info("device %s %s promiscuous mode\n",
			dev->name,
			dev->flags & IFF_PROMISC ? "entered" : "left");
6080 6081
		if (audit_enabled) {
			current_uid_gid(&uid, &gid);
6082 6083 6084 6085 6086
			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),
6087
				from_kuid(&init_user_ns, audit_get_loginuid(current)),
6088 6089
				from_kuid(&init_user_ns, uid),
				from_kgid(&init_user_ns, gid),
6090
				audit_get_sessionid(current));
6091
		}
6092

6093
		dev_change_rx_flags(dev, IFF_PROMISC);
L
Linus Torvalds 已提交
6094
	}
6095 6096
	if (notify)
		__dev_notify_flags(dev, old_flags, IFF_PROMISC);
6097
	return 0;
L
Linus Torvalds 已提交
6098 6099
}

6100 6101 6102 6103 6104 6105 6106 6107 6108
/**
 *	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.
6109
 *	Return 0 if successful or a negative errno code on error.
6110
 */
6111
int dev_set_promiscuity(struct net_device *dev, int inc)
6112
{
6113
	unsigned int old_flags = dev->flags;
6114
	int err;
6115

6116
	err = __dev_set_promiscuity(dev, inc, true);
6117
	if (err < 0)
6118
		return err;
6119 6120
	if (dev->flags != old_flags)
		dev_set_rx_mode(dev);
6121
	return err;
6122
}
E
Eric Dumazet 已提交
6123
EXPORT_SYMBOL(dev_set_promiscuity);
6124

6125
static int __dev_set_allmulti(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
6126
{
6127
	unsigned int old_flags = dev->flags, old_gflags = dev->gflags;
L
Linus Torvalds 已提交
6128

6129 6130
	ASSERT_RTNL();

L
Linus Torvalds 已提交
6131
	dev->flags |= IFF_ALLMULTI;
6132 6133 6134 6135 6136 6137 6138 6139 6140 6141
	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;
6142 6143
			pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n",
				dev->name);
6144 6145 6146
			return -EOVERFLOW;
		}
	}
6147
	if (dev->flags ^ old_flags) {
6148
		dev_change_rx_flags(dev, IFF_ALLMULTI);
6149
		dev_set_rx_mode(dev);
6150 6151 6152
		if (notify)
			__dev_notify_flags(dev, old_flags,
					   dev->gflags ^ old_gflags);
6153
	}
6154
	return 0;
6155
}
6156 6157 6158 6159 6160 6161 6162 6163 6164 6165 6166 6167 6168 6169 6170 6171 6172 6173

/**
 *	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
6174
EXPORT_SYMBOL(dev_set_allmulti);
6175 6176 6177 6178

/*
 *	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
6179
 *	filtering it is put in promiscuous mode while unicast addresses
6180 6181 6182 6183
 *	are present.
 */
void __dev_set_rx_mode(struct net_device *dev)
{
6184 6185
	const struct net_device_ops *ops = dev->netdev_ops;

6186 6187 6188 6189 6190
	/* dev_open will call this function so the list will stay sane. */
	if (!(dev->flags&IFF_UP))
		return;

	if (!netif_device_present(dev))
6191
		return;
6192

6193
	if (!(dev->priv_flags & IFF_UNICAST_FLT)) {
6194 6195 6196
		/* Unicast addresses changes may only happen under the rtnl,
		 * therefore calling __dev_set_promiscuity here is safe.
		 */
6197
		if (!netdev_uc_empty(dev) && !dev->uc_promisc) {
6198
			__dev_set_promiscuity(dev, 1, false);
6199
			dev->uc_promisc = true;
6200
		} else if (netdev_uc_empty(dev) && dev->uc_promisc) {
6201
			__dev_set_promiscuity(dev, -1, false);
6202
			dev->uc_promisc = false;
6203 6204
		}
	}
6205 6206 6207

	if (ops->ndo_set_rx_mode)
		ops->ndo_set_rx_mode(dev);
6208 6209 6210 6211
}

void dev_set_rx_mode(struct net_device *dev)
{
6212
	netif_addr_lock_bh(dev);
6213
	__dev_set_rx_mode(dev);
6214
	netif_addr_unlock_bh(dev);
L
Linus Torvalds 已提交
6215 6216
}

6217 6218 6219 6220 6221 6222
/**
 *	dev_get_flags - get flags reported to userspace
 *	@dev: device
 *
 *	Get the combination of flag bits exported through APIs to userspace.
 */
6223
unsigned int dev_get_flags(const struct net_device *dev)
L
Linus Torvalds 已提交
6224
{
6225
	unsigned int flags;
L
Linus Torvalds 已提交
6226 6227 6228

	flags = (dev->flags & ~(IFF_PROMISC |
				IFF_ALLMULTI |
S
Stefan Rompf 已提交
6229 6230 6231
				IFF_RUNNING |
				IFF_LOWER_UP |
				IFF_DORMANT)) |
L
Linus Torvalds 已提交
6232 6233 6234
		(dev->gflags & (IFF_PROMISC |
				IFF_ALLMULTI));

S
Stefan Rompf 已提交
6235 6236 6237 6238 6239 6240 6241 6242
	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
6243 6244 6245

	return flags;
}
E
Eric Dumazet 已提交
6246
EXPORT_SYMBOL(dev_get_flags);
L
Linus Torvalds 已提交
6247

6248
int __dev_change_flags(struct net_device *dev, unsigned int flags)
L
Linus Torvalds 已提交
6249
{
6250
	unsigned int old_flags = dev->flags;
6251
	int ret;
L
Linus Torvalds 已提交
6252

6253 6254
	ASSERT_RTNL();

L
Linus Torvalds 已提交
6255 6256 6257 6258 6259 6260 6261 6262 6263 6264 6265 6266 6267 6268
	/*
	 *	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.
	 */

6269 6270
	if ((old_flags ^ flags) & IFF_MULTICAST)
		dev_change_rx_flags(dev, IFF_MULTICAST);
6271

6272
	dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
6273 6274 6275 6276 6277 6278 6279 6280

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

	ret = 0;
6281
	if ((old_flags ^ flags) & IFF_UP)
6282
		ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev);
L
Linus Torvalds 已提交
6283 6284

	if ((flags ^ dev->gflags) & IFF_PROMISC) {
E
Eric Dumazet 已提交
6285
		int inc = (flags & IFF_PROMISC) ? 1 : -1;
6286
		unsigned int old_flags = dev->flags;
E
Eric Dumazet 已提交
6287

L
Linus Torvalds 已提交
6288
		dev->gflags ^= IFF_PROMISC;
6289 6290 6291 6292

		if (__dev_set_promiscuity(dev, inc, false) >= 0)
			if (dev->flags != old_flags)
				dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
6293 6294 6295 6296 6297 6298 6299
	}

	/* 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
6300 6301
		int inc = (flags & IFF_ALLMULTI) ? 1 : -1;

L
Linus Torvalds 已提交
6302
		dev->gflags ^= IFF_ALLMULTI;
6303
		__dev_set_allmulti(dev, inc, false);
L
Linus Torvalds 已提交
6304 6305
	}

6306 6307 6308
	return ret;
}

6309 6310
void __dev_notify_flags(struct net_device *dev, unsigned int old_flags,
			unsigned int gchanges)
6311 6312 6313
{
	unsigned int changes = dev->flags ^ old_flags;

6314
	if (gchanges)
6315
		rtmsg_ifinfo(RTM_NEWLINK, dev, gchanges, GFP_ATOMIC);
6316

6317 6318 6319 6320 6321 6322 6323 6324
	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 &&
6325 6326 6327 6328 6329 6330 6331
	    (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);
	}
6332 6333 6334 6335 6336 6337 6338 6339 6340 6341
}

/**
 *	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.
 */
6342
int dev_change_flags(struct net_device *dev, unsigned int flags)
6343
{
6344
	int ret;
6345
	unsigned int changes, old_flags = dev->flags, old_gflags = dev->gflags;
6346 6347 6348 6349 6350

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

6351
	changes = (old_flags ^ dev->flags) | (old_gflags ^ dev->gflags);
6352
	__dev_notify_flags(dev, old_flags, changes);
L
Linus Torvalds 已提交
6353 6354
	return ret;
}
E
Eric Dumazet 已提交
6355
EXPORT_SYMBOL(dev_change_flags);
L
Linus Torvalds 已提交
6356

6357 6358 6359 6360 6361 6362 6363 6364 6365 6366 6367
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;
}

6368 6369 6370 6371 6372 6373 6374
/**
 *	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
6375 6376
int dev_set_mtu(struct net_device *dev, int new_mtu)
{
6377
	int err, orig_mtu;
L
Linus Torvalds 已提交
6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388

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

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

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

6389 6390 6391 6392
	err = call_netdevice_notifiers(NETDEV_PRECHANGEMTU, dev);
	err = notifier_to_errno(err);
	if (err)
		return err;
6393

6394 6395
	orig_mtu = dev->mtu;
	err = __dev_set_mtu(dev, new_mtu);
6396

6397 6398 6399 6400 6401 6402 6403 6404 6405 6406 6407
	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
6408 6409
	return err;
}
E
Eric Dumazet 已提交
6410
EXPORT_SYMBOL(dev_set_mtu);
L
Linus Torvalds 已提交
6411

6412 6413 6414 6415 6416 6417 6418 6419 6420 6421 6422
/**
 *	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);

6423 6424 6425 6426 6427 6428 6429
/**
 *	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
6430 6431
int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa)
{
6432
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
6433 6434
	int err;

6435
	if (!ops->ndo_set_mac_address)
L
Linus Torvalds 已提交
6436 6437 6438 6439 6440
		return -EOPNOTSUPP;
	if (sa->sa_family != dev->type)
		return -EINVAL;
	if (!netif_device_present(dev))
		return -ENODEV;
6441
	err = ops->ndo_set_mac_address(dev, sa);
6442 6443
	if (err)
		return err;
J
Jiri Pirko 已提交
6444
	dev->addr_assign_type = NET_ADDR_SET;
6445
	call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
6446
	add_device_randomness(dev->dev_addr, dev->addr_len);
6447
	return 0;
L
Linus Torvalds 已提交
6448
}
E
Eric Dumazet 已提交
6449
EXPORT_SYMBOL(dev_set_mac_address);
L
Linus Torvalds 已提交
6450

J
Jiri Pirko 已提交
6451 6452 6453
/**
 *	dev_change_carrier - Change device carrier
 *	@dev: device
6454
 *	@new_carrier: new value
J
Jiri Pirko 已提交
6455 6456 6457 6458 6459 6460 6461 6462 6463 6464 6465 6466 6467 6468 6469
 *
 *	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);

6470 6471 6472 6473 6474 6475 6476 6477
/**
 *	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,
6478
			 struct netdev_phys_item_id *ppid)
6479 6480 6481 6482 6483 6484 6485 6486 6487
{
	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);

6488 6489 6490 6491
/**
 *	dev_get_phys_port_name - Get device physical port name
 *	@dev: device
 *	@name: port name
6492
 *	@len: limit of bytes to copy to name
6493 6494 6495 6496 6497 6498 6499 6500 6501 6502 6503 6504 6505 6506
 *
 *	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);

6507 6508 6509 6510 6511 6512 6513 6514 6515 6516 6517 6518 6519 6520 6521 6522 6523 6524 6525 6526
/**
 *	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
6527 6528
/**
 *	dev_new_index	-	allocate an ifindex
6529
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
6530 6531 6532 6533 6534
 *
 *	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.
 */
6535
static int dev_new_index(struct net *net)
L
Linus Torvalds 已提交
6536
{
6537
	int ifindex = net->ifindex;
L
Linus Torvalds 已提交
6538 6539 6540
	for (;;) {
		if (++ifindex <= 0)
			ifindex = 1;
6541
		if (!__dev_get_by_index(net, ifindex))
6542
			return net->ifindex = ifindex;
L
Linus Torvalds 已提交
6543 6544 6545 6546
	}
}

/* Delayed registration/unregisteration */
6547
static LIST_HEAD(net_todo_list);
6548
DECLARE_WAIT_QUEUE_HEAD(netdev_unregistering_wq);
L
Linus Torvalds 已提交
6549

6550
static void net_set_todo(struct net_device *dev)
L
Linus Torvalds 已提交
6551 6552
{
	list_add_tail(&dev->todo_list, &net_todo_list);
6553
	dev_net(dev)->dev_unreg_count++;
L
Linus Torvalds 已提交
6554 6555
}

6556
static void rollback_registered_many(struct list_head *head)
6557
{
6558
	struct net_device *dev, *tmp;
6559
	LIST_HEAD(close_head);
6560

6561 6562 6563
	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

6564
	list_for_each_entry_safe(dev, tmp, head, unreg_list) {
6565
		/* Some devices call without registering
6566 6567
		 * for initialization unwind. Remove those
		 * devices and proceed with the remaining.
6568 6569
		 */
		if (dev->reg_state == NETREG_UNINITIALIZED) {
6570 6571
			pr_debug("unregister_netdevice: device %s/%p never was registered\n",
				 dev->name, dev);
6572

6573
			WARN_ON(1);
6574 6575
			list_del(&dev->unreg_list);
			continue;
6576
		}
6577
		dev->dismantle = true;
6578
		BUG_ON(dev->reg_state != NETREG_REGISTERED);
6579
	}
6580

6581
	/* If device is running, close it first. */
6582 6583
	list_for_each_entry(dev, head, unreg_list)
		list_add_tail(&dev->close_list, &close_head);
6584
	dev_close_many(&close_head, true);
6585

6586
	list_for_each_entry(dev, head, unreg_list) {
6587 6588
		/* And unlink it from device chain. */
		unlist_netdevice(dev);
6589

6590
		dev->reg_state = NETREG_UNREGISTERING;
6591
		on_each_cpu(flush_backlog, dev, 1);
6592
	}
6593 6594 6595

	synchronize_net();

6596
	list_for_each_entry(dev, head, unreg_list) {
6597 6598
		struct sk_buff *skb = NULL;

6599 6600
		/* Shutdown queueing discipline. */
		dev_shutdown(dev);
6601 6602


6603 6604 6605 6606
		/* Notify protocols, that we are about to destroy
		   this device. They should clean all the things.
		*/
		call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
6607

6608 6609 6610 6611 6612
		if (!dev->rtnl_link_ops ||
		    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
			skb = rtmsg_ifinfo_build_skb(RTM_DELLINK, dev, ~0U,
						     GFP_KERNEL);

6613 6614 6615
		/*
		 *	Flush the unicast and multicast chains
		 */
6616
		dev_uc_flush(dev);
6617
		dev_mc_flush(dev);
6618

6619 6620
		if (dev->netdev_ops->ndo_uninit)
			dev->netdev_ops->ndo_uninit(dev);
6621

6622 6623
		if (skb)
			rtmsg_ifinfo_send(skb, dev, GFP_KERNEL);
6624

J
Jiri Pirko 已提交
6625 6626
		/* Notifier chain MUST detach us all upper devices. */
		WARN_ON(netdev_has_any_upper_dev(dev));
6627

6628 6629
		/* Remove entries from kobject tree */
		netdev_unregister_kobject(dev);
6630 6631 6632 6633
#ifdef CONFIG_XPS
		/* Remove XPS queueing entries */
		netif_reset_xps_queues_gt(dev, 0);
#endif
6634
	}
6635

6636
	synchronize_net();
6637

6638
	list_for_each_entry(dev, head, unreg_list)
6639 6640 6641 6642 6643 6644 6645 6646 6647
		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
6648
	list_del(&single);
6649 6650
}

6651 6652 6653 6654 6655
static netdev_features_t netdev_sync_upper_features(struct net_device *lower,
	struct net_device *upper, netdev_features_t features)
{
	netdev_features_t upper_disables = NETIF_F_UPPER_DISABLES;
	netdev_features_t feature;
6656
	int feature_bit;
6657

6658 6659
	for_each_netdev_feature(&upper_disables, feature_bit) {
		feature = __NETIF_F_BIT(feature_bit);
6660 6661 6662 6663 6664 6665 6666 6667 6668 6669 6670 6671 6672 6673 6674 6675
		if (!(upper->wanted_features & feature)
		    && (features & feature)) {
			netdev_dbg(lower, "Dropping feature %pNF, upper dev %s has it off.\n",
				   &feature, upper->name);
			features &= ~feature;
		}
	}

	return features;
}

static void netdev_sync_lower_features(struct net_device *upper,
	struct net_device *lower, netdev_features_t features)
{
	netdev_features_t upper_disables = NETIF_F_UPPER_DISABLES;
	netdev_features_t feature;
6676
	int feature_bit;
6677

6678 6679
	for_each_netdev_feature(&upper_disables, feature_bit) {
		feature = __NETIF_F_BIT(feature_bit);
6680 6681 6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692
		if (!(features & feature) && (lower->features & feature)) {
			netdev_dbg(upper, "Disabling feature %pNF on lower dev %s.\n",
				   &feature, lower->name);
			lower->wanted_features &= ~feature;
			netdev_update_features(lower);

			if (unlikely(lower->features & feature))
				netdev_WARN(upper, "failed to disable %pNF on %s!\n",
					    &feature, lower->name);
		}
	}
}

6693 6694
static netdev_features_t netdev_fix_features(struct net_device *dev,
	netdev_features_t features)
6695
{
6696 6697 6698
	/* Fix illegal checksum combinations */
	if ((features & NETIF_F_HW_CSUM) &&
	    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
6699
		netdev_warn(dev, "mixed HW and IP checksum settings.\n");
6700 6701 6702
		features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
	}

6703
	/* TSO requires that SG is present as well. */
6704
	if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
6705
		netdev_dbg(dev, "Dropping TSO features since no SG feature.\n");
6706
		features &= ~NETIF_F_ALL_TSO;
6707 6708
	}

6709 6710 6711 6712 6713 6714 6715 6716 6717 6718 6719 6720 6721
	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;
	}

6722 6723 6724 6725
	/* TSO with IPv4 ID mangling requires IPv4 TSO be enabled */
	if ((features & NETIF_F_TSO_MANGLEID) && !(features & NETIF_F_TSO))
		features &= ~NETIF_F_TSO_MANGLEID;

6726 6727 6728 6729
	/* TSO ECN requires that TSO is present as well. */
	if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN)
		features &= ~NETIF_F_TSO_ECN;

6730 6731
	/* Software GSO depends on SG. */
	if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) {
6732
		netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n");
6733 6734 6735
		features &= ~NETIF_F_GSO;
	}

6736
	/* UFO needs SG and checksumming */
6737
	if (features & NETIF_F_UFO) {
6738
		/* maybe split UFO into V4 and V6? */
6739 6740 6741
		if (!(features & NETIF_F_HW_CSUM) &&
		    ((features & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)) !=
		     (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM))) {
6742
			netdev_dbg(dev,
6743
				"Dropping NETIF_F_UFO since no checksum offload features.\n");
6744 6745 6746 6747
			features &= ~NETIF_F_UFO;
		}

		if (!(features & NETIF_F_SG)) {
6748
			netdev_dbg(dev,
6749
				"Dropping NETIF_F_UFO since no NETIF_F_SG feature.\n");
6750 6751 6752 6753
			features &= ~NETIF_F_UFO;
		}
	}

6754 6755 6756 6757 6758 6759 6760 6761
	/* GSO partial features require GSO partial be set */
	if ((features & dev->gso_partial_features) &&
	    !(features & NETIF_F_GSO_PARTIAL)) {
		netdev_dbg(dev,
			   "Dropping partially supported GSO features since no GSO partial.\n");
		features &= ~dev->gso_partial_features;
	}

J
Jiri Pirko 已提交
6762 6763 6764 6765 6766 6767 6768
#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;

6769 6770 6771
	return features;
}

6772
int __netdev_update_features(struct net_device *dev)
6773
{
6774
	struct net_device *upper, *lower;
6775
	netdev_features_t features;
6776
	struct list_head *iter;
6777
	int err = -1;
6778

6779 6780
	ASSERT_RTNL();

6781 6782 6783 6784 6785 6786 6787 6788
	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);

6789 6790 6791 6792
	/* some features can't be enabled if they're off an an upper device */
	netdev_for_each_upper_dev_rcu(dev, upper, iter)
		features = netdev_sync_upper_features(dev, upper, features);

6793
	if (dev->features == features)
6794
		goto sync_lower;
6795

6796 6797
	netdev_dbg(dev, "Features changed: %pNF -> %pNF\n",
		&dev->features, &features);
6798 6799 6800

	if (dev->netdev_ops->ndo_set_features)
		err = dev->netdev_ops->ndo_set_features(dev, features);
6801 6802
	else
		err = 0;
6803

6804
	if (unlikely(err < 0)) {
6805
		netdev_err(dev,
6806 6807
			"set_features() failed (%d); wanted %pNF, left %pNF\n",
			err, &features, &dev->features);
6808 6809 6810 6811
		/* return non-0 since some features might have changed and
		 * it's better to fire a spurious notification than miss it
		 */
		return -1;
6812 6813
	}

6814
sync_lower:
6815 6816 6817 6818 6819 6820
	/* some features must be disabled on lower devices when disabled
	 * on an upper device (think: bonding master or bridge)
	 */
	netdev_for_each_lower_dev(dev, lower, iter)
		netdev_sync_lower_features(dev, lower, features);

6821 6822 6823
	if (!err)
		dev->features = features;

6824
	return err < 0 ? 0 : 1;
6825 6826
}

6827 6828 6829 6830 6831 6832 6833 6834
/**
 *	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.
 */
6835 6836 6837 6838
void netdev_update_features(struct net_device *dev)
{
	if (__netdev_update_features(dev))
		netdev_features_change(dev);
6839 6840 6841
}
EXPORT_SYMBOL(netdev_update_features);

6842 6843 6844 6845 6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857 6858
/**
 *	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);

6859 6860 6861 6862 6863 6864 6865 6866 6867 6868 6869 6870 6871 6872 6873 6874 6875 6876 6877 6878 6879 6880 6881 6882 6883 6884 6885
/**
 *	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);

6886
#ifdef CONFIG_SYSFS
6887 6888 6889
static int netif_alloc_rx_queues(struct net_device *dev)
{
	unsigned int i, count = dev->num_rx_queues;
T
Tom Herbert 已提交
6890
	struct netdev_rx_queue *rx;
6891
	size_t sz = count * sizeof(*rx);
6892

T
Tom Herbert 已提交
6893
	BUG_ON(count < 1);
6894

6895 6896 6897 6898 6899 6900
	rx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!rx) {
		rx = vzalloc(sz);
		if (!rx)
			return -ENOMEM;
	}
T
Tom Herbert 已提交
6901 6902 6903
	dev->_rx = rx;

	for (i = 0; i < count; i++)
T
Tom Herbert 已提交
6904
		rx[i].dev = dev;
6905 6906
	return 0;
}
T
Tom Herbert 已提交
6907
#endif
6908

C
Changli Gao 已提交
6909 6910 6911 6912 6913 6914 6915
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;
6916
	netdev_queue_numa_node_write(queue, NUMA_NO_NODE);
C
Changli Gao 已提交
6917
	queue->dev = dev;
T
Tom Herbert 已提交
6918 6919 6920
#ifdef CONFIG_BQL
	dql_init(&queue->dql, HZ);
#endif
C
Changli Gao 已提交
6921 6922
}

6923 6924
static void netif_free_tx_queues(struct net_device *dev)
{
W
WANG Cong 已提交
6925
	kvfree(dev->_tx);
6926 6927
}

6928 6929 6930 6931
static int netif_alloc_netdev_queues(struct net_device *dev)
{
	unsigned int count = dev->num_tx_queues;
	struct netdev_queue *tx;
6932
	size_t sz = count * sizeof(*tx);
6933

6934 6935
	if (count < 1 || count > 0xffff)
		return -EINVAL;
6936

6937 6938 6939 6940 6941 6942
	tx = kzalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!tx) {
		tx = vzalloc(sz);
		if (!tx)
			return -ENOMEM;
	}
6943
	dev->_tx = tx;
T
Tom Herbert 已提交
6944

6945 6946
	netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
	spin_lock_init(&dev->tx_global_lock);
C
Changli Gao 已提交
6947 6948

	return 0;
6949 6950
}

6951 6952 6953 6954 6955 6956 6957 6958 6959 6960 6961
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
6962 6963 6964 6965 6966 6967 6968 6969 6970 6971 6972 6973 6974 6975 6976 6977 6978 6979 6980 6981
/**
 *	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;
6982
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
6983 6984 6985 6986

	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

6987 6988
	might_sleep();

L
Linus Torvalds 已提交
6989 6990
	/* When net_device's are persistent, this will be fatal. */
	BUG_ON(dev->reg_state != NETREG_UNINITIALIZED);
6991
	BUG_ON(!net);
L
Linus Torvalds 已提交
6992

6993
	spin_lock_init(&dev->addr_list_lock);
6994
	netdev_set_addr_lockdep_class(dev);
L
Linus Torvalds 已提交
6995

6996
	ret = dev_get_valid_name(net, dev, dev->name);
6997 6998 6999
	if (ret < 0)
		goto out;

L
Linus Torvalds 已提交
7000
	/* Init, if this function is available */
7001 7002
	if (dev->netdev_ops->ndo_init) {
		ret = dev->netdev_ops->ndo_init(dev);
L
Linus Torvalds 已提交
7003 7004 7005
		if (ret) {
			if (ret > 0)
				ret = -EIO;
7006
			goto out;
L
Linus Torvalds 已提交
7007 7008
		}
	}
7009

7010 7011
	if (((dev->hw_features | dev->features) &
	     NETIF_F_HW_VLAN_CTAG_FILTER) &&
7012 7013 7014 7015 7016 7017 7018
	    (!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;
	}

7019 7020 7021 7022 7023 7024
	ret = -EBUSY;
	if (!dev->ifindex)
		dev->ifindex = dev_new_index(net);
	else if (__dev_get_by_index(net, dev->ifindex))
		goto err_uninit;

7025 7026 7027 7028
	/* Transfer changeable features to wanted_features and enable
	 * software offloads (GSO and GRO).
	 */
	dev->hw_features |= NETIF_F_SOFT_FEATURES;
7029 7030
	dev->features |= NETIF_F_SOFT_FEATURES;
	dev->wanted_features = dev->features & dev->hw_features;
L
Linus Torvalds 已提交
7031

7032
	if (!(dev->flags & IFF_LOOPBACK))
7033
		dev->hw_features |= NETIF_F_NOCACHE_COPY;
7034

7035 7036 7037 7038 7039
	/* If IPv4 TCP segmentation offload is supported we should also
	 * allow the device to enable segmenting the frame with the option
	 * of ignoring a static IP ID value.  This doesn't enable the
	 * feature itself but allows the user to enable it later.
	 */
7040 7041
	if (dev->hw_features & NETIF_F_TSO)
		dev->hw_features |= NETIF_F_TSO_MANGLEID;
7042 7043 7044 7045 7046 7047
	if (dev->vlan_features & NETIF_F_TSO)
		dev->vlan_features |= NETIF_F_TSO_MANGLEID;
	if (dev->mpls_features & NETIF_F_TSO)
		dev->mpls_features |= NETIF_F_TSO_MANGLEID;
	if (dev->hw_enc_features & NETIF_F_TSO)
		dev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
7048

7049
	/* Make NETIF_F_HIGHDMA inheritable to VLAN devices.
7050
	 */
7051
	dev->vlan_features |= NETIF_F_HIGHDMA;
7052

7053 7054
	/* Make NETIF_F_SG inheritable to tunnel devices.
	 */
7055
	dev->hw_enc_features |= NETIF_F_SG | NETIF_F_GSO_PARTIAL;
7056

S
Simon Horman 已提交
7057 7058 7059 7060
	/* Make NETIF_F_SG inheritable to MPLS.
	 */
	dev->mpls_features |= NETIF_F_SG;

7061 7062 7063 7064 7065
	ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		goto err_uninit;

7066
	ret = netdev_register_kobject(dev);
7067
	if (ret)
7068
		goto err_uninit;
7069 7070
	dev->reg_state = NETREG_REGISTERED;

7071
	__netdev_update_features(dev);
7072

L
Linus Torvalds 已提交
7073 7074 7075 7076 7077 7078 7079
	/*
	 *	Default initial state at registry is that the
	 *	device is present.
	 */

	set_bit(__LINK_STATE_PRESENT, &dev->state);

7080 7081
	linkwatch_init_dev(dev);

L
Linus Torvalds 已提交
7082 7083
	dev_init_scheduler(dev);
	dev_hold(dev);
7084
	list_netdevice(dev);
7085
	add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
7086

7087 7088 7089 7090 7091 7092 7093
	/* 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
7094
	/* Notify protocols, that a new device appeared. */
7095
	ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
7096
	ret = notifier_to_errno(ret);
7097 7098 7099 7100
	if (ret) {
		rollback_registered(dev);
		dev->reg_state = NETREG_UNREGISTERED;
	}
7101 7102 7103 7104
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
7105 7106
	if (!dev->rtnl_link_ops ||
	    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
7107
		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
L
Linus Torvalds 已提交
7108 7109 7110

out:
	return ret;
7111 7112

err_uninit:
7113 7114
	if (dev->netdev_ops->ndo_uninit)
		dev->netdev_ops->ndo_uninit(dev);
7115
	goto out;
L
Linus Torvalds 已提交
7116
}
E
Eric Dumazet 已提交
7117
EXPORT_SYMBOL(register_netdevice);
L
Linus Torvalds 已提交
7118

7119 7120 7121 7122 7123 7124 7125 7126 7127 7128 7129 7130 7131 7132 7133 7134 7135 7136 7137 7138 7139 7140 7141 7142 7143 7144 7145 7146 7147 7148 7149
/**
 *	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
7150 7151 7152 7153 7154
	/* Note : We dont allocate pcpu_refcnt for dummy devices,
	 * because users of this 'device' dont need to change
	 * its refcount.
	 */

7155 7156 7157 7158 7159
	return 0;
}
EXPORT_SYMBOL_GPL(init_dummy_netdev);


L
Linus Torvalds 已提交
7160 7161 7162 7163 7164 7165 7166 7167 7168
/**
 *	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.
 *
7169
 *	This is a wrapper around register_netdevice that takes the rtnl semaphore
L
Linus Torvalds 已提交
7170 7171 7172 7173 7174 7175 7176 7177 7178 7179 7180 7181 7182 7183
 *	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
7184 7185 7186 7187 7188 7189 7190 7191 7192 7193
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);

7194
/**
L
Linus Torvalds 已提交
7195
 * netdev_wait_allrefs - wait until all references are gone.
7196
 * @dev: target net_device
L
Linus Torvalds 已提交
7197 7198 7199 7200 7201 7202 7203
 *
 * 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
7204
 * call dev_put.
L
Linus Torvalds 已提交
7205 7206 7207 7208
 */
static void netdev_wait_allrefs(struct net_device *dev)
{
	unsigned long rebroadcast_time, warning_time;
E
Eric Dumazet 已提交
7209
	int refcnt;
L
Linus Torvalds 已提交
7210

7211 7212
	linkwatch_forget_dev(dev);

L
Linus Torvalds 已提交
7213
	rebroadcast_time = warning_time = jiffies;
E
Eric Dumazet 已提交
7214 7215 7216
	refcnt = netdev_refcnt_read(dev);

	while (refcnt != 0) {
L
Linus Torvalds 已提交
7217
		if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
7218
			rtnl_lock();
L
Linus Torvalds 已提交
7219 7220

			/* Rebroadcast unregister notification */
7221
			call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
L
Linus Torvalds 已提交
7222

7223
			__rtnl_unlock();
7224
			rcu_barrier();
7225 7226
			rtnl_lock();

7227
			call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
L
Linus Torvalds 已提交
7228 7229 7230 7231 7232 7233 7234 7235 7236 7237 7238
			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();
			}

7239
			__rtnl_unlock();
L
Linus Torvalds 已提交
7240 7241 7242 7243 7244 7245

			rebroadcast_time = jiffies;
		}

		msleep(250);

E
Eric Dumazet 已提交
7246 7247
		refcnt = netdev_refcnt_read(dev);

L
Linus Torvalds 已提交
7248
		if (time_after(jiffies, warning_time + 10 * HZ)) {
7249 7250
			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
				 dev->name, refcnt);
L
Linus Torvalds 已提交
7251 7252 7253 7254 7255 7256 7257 7258 7259 7260 7261 7262 7263 7264 7265 7266 7267 7268 7269
			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
7270
 * We are invoked by rtnl_unlock().
L
Linus Torvalds 已提交
7271
 * This allows us to deal with problems:
7272
 * 1) We can delete sysfs objects which invoke hotplug
L
Linus Torvalds 已提交
7273 7274 7275
 *    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
7276 7277 7278
 *
 * 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
7279 7280 7281
 */
void netdev_run_todo(void)
{
7282
	struct list_head list;
L
Linus Torvalds 已提交
7283 7284

	/* Snapshot list, allow later requests */
7285
	list_replace_init(&net_todo_list, &list);
H
Herbert Xu 已提交
7286 7287

	__rtnl_unlock();
7288

7289 7290

	/* Wait for rcu callbacks to finish before next phase */
7291 7292 7293
	if (!list_empty(&list))
		rcu_barrier();

L
Linus Torvalds 已提交
7294 7295
	while (!list_empty(&list)) {
		struct net_device *dev
7296
			= list_first_entry(&list, struct net_device, todo_list);
L
Linus Torvalds 已提交
7297 7298
		list_del(&dev->todo_list);

7299
		rtnl_lock();
7300
		call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
7301
		__rtnl_unlock();
7302

7303
		if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
7304
			pr_err("network todo '%s' but state %d\n",
7305 7306 7307 7308
			       dev->name, dev->reg_state);
			dump_stack();
			continue;
		}
L
Linus Torvalds 已提交
7309

7310
		dev->reg_state = NETREG_UNREGISTERED;
L
Linus Torvalds 已提交
7311

7312
		netdev_wait_allrefs(dev);
L
Linus Torvalds 已提交
7313

7314
		/* paranoia */
E
Eric Dumazet 已提交
7315
		BUG_ON(netdev_refcnt_read(dev));
7316 7317
		BUG_ON(!list_empty(&dev->ptype_all));
		BUG_ON(!list_empty(&dev->ptype_specific));
7318 7319
		WARN_ON(rcu_access_pointer(dev->ip_ptr));
		WARN_ON(rcu_access_pointer(dev->ip6_ptr));
7320
		WARN_ON(dev->dn_ptr);
L
Linus Torvalds 已提交
7321

7322 7323
		if (dev->destructor)
			dev->destructor(dev);
7324

7325 7326 7327 7328 7329 7330
		/* Report a network device has been unregistered */
		rtnl_lock();
		dev_net(dev)->dev_unreg_count--;
		__rtnl_unlock();
		wake_up(&netdev_unregistering_wq);

7331 7332
		/* Free network device */
		kobject_put(&dev->dev.kobj);
L
Linus Torvalds 已提交
7333 7334 7335
	}
}

7336 7337 7338 7339
/* Convert net_device_stats to rtnl_link_stats64. rtnl_link_stats64 has
 * all the same fields in the same order as net_device_stats, with only
 * the type differing, but rtnl_link_stats64 may have additional fields
 * at the end for newer counters.
7340
 */
7341 7342
void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
			     const struct net_device_stats *netdev_stats)
7343 7344
{
#if BITS_PER_LONG == 64
7345
	BUILD_BUG_ON(sizeof(*stats64) < sizeof(*netdev_stats));
7346
	memcpy(stats64, netdev_stats, sizeof(*stats64));
7347 7348 7349
	/* zero out counters that only exist in rtnl_link_stats64 */
	memset((char *)stats64 + sizeof(*netdev_stats), 0,
	       sizeof(*stats64) - sizeof(*netdev_stats));
7350
#else
7351
	size_t i, n = sizeof(*netdev_stats) / sizeof(unsigned long);
7352 7353 7354
	const unsigned long *src = (const unsigned long *)netdev_stats;
	u64 *dst = (u64 *)stats64;

7355
	BUILD_BUG_ON(n > sizeof(*stats64) / sizeof(u64));
7356 7357
	for (i = 0; i < n; i++)
		dst[i] = src[i];
7358 7359 7360
	/* zero out counters that only exist in rtnl_link_stats64 */
	memset((char *)stats64 + n * sizeof(u64), 0,
	       sizeof(*stats64) - n * sizeof(u64));
7361 7362
#endif
}
7363
EXPORT_SYMBOL(netdev_stats_to_stats64);
7364

7365 7366 7367
/**
 *	dev_get_stats	- get network device statistics
 *	@dev: device to get statistics from
7368
 *	@storage: place to store stats
7369
 *
7370 7371 7372 7373
 *	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.
7374
 */
7375 7376
struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
					struct rtnl_link_stats64 *storage)
7377
{
7378 7379
	const struct net_device_ops *ops = dev->netdev_ops;

7380 7381
	if (ops->ndo_get_stats64) {
		memset(storage, 0, sizeof(*storage));
7382 7383
		ops->ndo_get_stats64(dev, storage);
	} else if (ops->ndo_get_stats) {
7384
		netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev));
7385 7386
	} else {
		netdev_stats_to_stats64(storage, &dev->stats);
7387
	}
7388
	storage->rx_dropped += atomic_long_read(&dev->rx_dropped);
7389
	storage->tx_dropped += atomic_long_read(&dev->tx_dropped);
7390
	storage->rx_nohandler += atomic_long_read(&dev->rx_nohandler);
7391
	return storage;
R
Rusty Russell 已提交
7392
}
7393
EXPORT_SYMBOL(dev_get_stats);
R
Rusty Russell 已提交
7394

7395
struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
7396
{
7397
	struct netdev_queue *queue = dev_ingress_queue(dev);
7398

7399 7400 7401 7402 7403 7404 7405
#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
7406
	RCU_INIT_POINTER(queue->qdisc, &noop_qdisc);
7407 7408 7409 7410
	queue->qdisc_sleeping = &noop_qdisc;
	rcu_assign_pointer(dev->ingress_queue, queue);
#endif
	return queue;
7411 7412
}

7413 7414
static const struct ethtool_ops default_ethtool_ops;

7415 7416 7417 7418 7419 7420 7421 7422
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);

7423 7424 7425 7426
void netdev_freemem(struct net_device *dev)
{
	char *addr = (char *)dev - dev->padded;

W
WANG Cong 已提交
7427
	kvfree(addr);
7428 7429
}

L
Linus Torvalds 已提交
7430
/**
T
Tom Herbert 已提交
7431
 *	alloc_netdev_mqs - allocate network device
7432 7433 7434 7435 7436 7437
 *	@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
7438 7439
 *
 *	Allocates a struct net_device with private data area for driver use
L
Li Zhong 已提交
7440
 *	and performs basic initialization.  Also allocates subqueue structs
T
Tom Herbert 已提交
7441
 *	for each queue on the device.
L
Linus Torvalds 已提交
7442
 */
T
Tom Herbert 已提交
7443
struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
7444
		unsigned char name_assign_type,
T
Tom Herbert 已提交
7445 7446
		void (*setup)(struct net_device *),
		unsigned int txqs, unsigned int rxqs)
L
Linus Torvalds 已提交
7447 7448
{
	struct net_device *dev;
7449
	size_t alloc_size;
7450
	struct net_device *p;
L
Linus Torvalds 已提交
7451

7452 7453
	BUG_ON(strlen(name) >= sizeof(dev->name));

T
Tom Herbert 已提交
7454
	if (txqs < 1) {
7455
		pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
7456 7457 7458
		return NULL;
	}

7459
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
7460
	if (rxqs < 1) {
7461
		pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
T
Tom Herbert 已提交
7462 7463 7464 7465
		return NULL;
	}
#endif

7466
	alloc_size = sizeof(struct net_device);
7467 7468
	if (sizeof_priv) {
		/* ensure 32-byte alignment of private area */
7469
		alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
7470 7471 7472
		alloc_size += sizeof_priv;
	}
	/* ensure 32-byte alignment of whole construct */
7473
	alloc_size += NETDEV_ALIGN - 1;
L
Linus Torvalds 已提交
7474

7475 7476 7477
	p = kzalloc(alloc_size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
	if (!p)
		p = vzalloc(alloc_size);
7478
	if (!p)
L
Linus Torvalds 已提交
7479 7480
		return NULL;

7481
	dev = PTR_ALIGN(p, NETDEV_ALIGN);
L
Linus Torvalds 已提交
7482
	dev->padded = (char *)dev - (char *)p;
7483

E
Eric Dumazet 已提交
7484 7485
	dev->pcpu_refcnt = alloc_percpu(int);
	if (!dev->pcpu_refcnt)
7486
		goto free_dev;
7487 7488

	if (dev_addr_init(dev))
E
Eric Dumazet 已提交
7489
		goto free_pcpu;
7490

7491
	dev_mc_init(dev);
7492
	dev_uc_init(dev);
J
Jiri Pirko 已提交
7493

7494
	dev_net_set(dev, &init_net);
L
Linus Torvalds 已提交
7495

7496
	dev->gso_max_size = GSO_MAX_SIZE;
7497
	dev->gso_max_segs = GSO_MAX_SEGS;
7498 7499 7500

	INIT_LIST_HEAD(&dev->napi_list);
	INIT_LIST_HEAD(&dev->unreg_list);
7501
	INIT_LIST_HEAD(&dev->close_list);
7502
	INIT_LIST_HEAD(&dev->link_watch_list);
7503 7504 7505 7506
	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);
7507 7508
	INIT_LIST_HEAD(&dev->ptype_all);
	INIT_LIST_HEAD(&dev->ptype_specific);
7509
	dev->priv_flags = IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM;
7510 7511
	setup(dev);

7512
	if (!dev->tx_queue_len) {
7513
		dev->priv_flags |= IFF_NO_QUEUE;
7514 7515
		dev->tx_queue_len = 1;
	}
7516

T
Tom Herbert 已提交
7517 7518
	dev->num_tx_queues = txqs;
	dev->real_num_tx_queues = txqs;
7519
	if (netif_alloc_netdev_queues(dev))
7520
		goto free_all;
7521

7522
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
7523 7524
	dev->num_rx_queues = rxqs;
	dev->real_num_rx_queues = rxqs;
T
Tom Herbert 已提交
7525
	if (netif_alloc_rx_queues(dev))
7526
		goto free_all;
E
Eric Dumazet 已提交
7527
#endif
T
Tom Herbert 已提交
7528

L
Linus Torvalds 已提交
7529
	strcpy(dev->name, name);
7530
	dev->name_assign_type = name_assign_type;
7531
	dev->group = INIT_NETDEV_GROUP;
7532 7533
	if (!dev->ethtool_ops)
		dev->ethtool_ops = &default_ethtool_ops;
7534 7535 7536

	nf_hook_ingress_init(dev);

L
Linus Torvalds 已提交
7537
	return dev;
7538

7539 7540 7541 7542
free_all:
	free_netdev(dev);
	return NULL;

E
Eric Dumazet 已提交
7543 7544
free_pcpu:
	free_percpu(dev->pcpu_refcnt);
7545 7546
free_dev:
	netdev_freemem(dev);
7547
	return NULL;
L
Linus Torvalds 已提交
7548
}
T
Tom Herbert 已提交
7549
EXPORT_SYMBOL(alloc_netdev_mqs);
L
Linus Torvalds 已提交
7550 7551 7552 7553 7554

/**
 *	free_netdev - free network device
 *	@dev: device
 *
7555 7556
 *	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
7557
 *	If this is the last reference then it will be freed.
7558
 *	Must be called in process context.
L
Linus Torvalds 已提交
7559 7560 7561
 */
void free_netdev(struct net_device *dev)
{
7562 7563
	struct napi_struct *p, *n;

7564
	might_sleep();
7565
	netif_free_tx_queues(dev);
7566
#ifdef CONFIG_SYSFS
7567
	kvfree(dev->_rx);
T
Tom Herbert 已提交
7568
#endif
7569

7570
	kfree(rcu_dereference_protected(dev->ingress_queue, 1));
7571

7572 7573 7574
	/* Flush device addresses */
	dev_addr_flush(dev);

7575 7576 7577
	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
		netif_napi_del(p);

E
Eric Dumazet 已提交
7578 7579 7580
	free_percpu(dev->pcpu_refcnt);
	dev->pcpu_refcnt = NULL;

S
Stephen Hemminger 已提交
7581
	/*  Compatibility with error handling in drivers */
L
Linus Torvalds 已提交
7582
	if (dev->reg_state == NETREG_UNINITIALIZED) {
7583
		netdev_freemem(dev);
L
Linus Torvalds 已提交
7584 7585 7586 7587 7588 7589
		return;
	}

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

7590 7591
	/* will free via device release */
	put_device(&dev->dev);
L
Linus Torvalds 已提交
7592
}
E
Eric Dumazet 已提交
7593
EXPORT_SYMBOL(free_netdev);
7594

7595 7596 7597 7598 7599 7600
/**
 *	synchronize_net -  Synchronize with packet receive processing
 *
 *	Wait for packets currently being received to be done.
 *	Does not block later packets from starting.
 */
7601
void synchronize_net(void)
L
Linus Torvalds 已提交
7602 7603
{
	might_sleep();
7604 7605 7606 7607
	if (rtnl_is_locked())
		synchronize_rcu_expedited();
	else
		synchronize_rcu();
L
Linus Torvalds 已提交
7608
}
E
Eric Dumazet 已提交
7609
EXPORT_SYMBOL(synchronize_net);
L
Linus Torvalds 已提交
7610 7611

/**
7612
 *	unregister_netdevice_queue - remove device from the kernel
L
Linus Torvalds 已提交
7613
 *	@dev: device
7614
 *	@head: list
7615
 *
L
Linus Torvalds 已提交
7616
 *	This function shuts down a device interface and removes it
7617
 *	from the kernel tables.
7618
 *	If head not NULL, device is queued to be unregistered later.
L
Linus Torvalds 已提交
7619 7620 7621 7622 7623
 *
 *	Callers must hold the rtnl semaphore.  You may want
 *	unregister_netdev() instead of this.
 */

7624
void unregister_netdevice_queue(struct net_device *dev, struct list_head *head)
L
Linus Torvalds 已提交
7625
{
7626 7627
	ASSERT_RTNL();

7628
	if (head) {
7629
		list_move_tail(&dev->unreg_list, head);
7630 7631 7632 7633 7634
	} else {
		rollback_registered(dev);
		/* Finish processing unregister after unlock */
		net_set_todo(dev);
	}
L
Linus Torvalds 已提交
7635
}
7636
EXPORT_SYMBOL(unregister_netdevice_queue);
L
Linus Torvalds 已提交
7637

7638 7639 7640
/**
 *	unregister_netdevice_many - unregister many devices
 *	@head: list of devices
7641 7642 7643
 *
 *  Note: As most callers use a stack allocated list_head,
 *  we force a list_del() to make sure stack wont be corrupted later.
7644 7645 7646 7647 7648 7649 7650 7651 7652
 */
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);
7653
		list_del(head);
7654 7655
	}
}
7656
EXPORT_SYMBOL(unregister_netdevice_many);
7657

L
Linus Torvalds 已提交
7658 7659 7660 7661 7662
/**
 *	unregister_netdev - remove device from the kernel
 *	@dev: device
 *
 *	This function shuts down a device interface and removes it
7663
 *	from the kernel tables.
L
Linus Torvalds 已提交
7664 7665 7666 7667 7668 7669 7670 7671 7672 7673 7674 7675 7676
 *
 *	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);

7677 7678 7679 7680 7681 7682 7683 7684 7685 7686 7687 7688 7689 7690 7691 7692 7693 7694 7695 7696 7697 7698 7699 7700 7701 7702 7703 7704 7705 7706 7707
/**
 *	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;
7708
	if (net_eq(dev_net(dev), net))
7709 7710 7711 7712 7713 7714
		goto out;

	/* Pick the destination device name, and ensure
	 * we can use it in the destination network namespace.
	 */
	err = -EEXIST;
7715
	if (__dev_get_by_name(net, dev->name)) {
7716 7717 7718
		/* We get here if we can't use the current device name */
		if (!pat)
			goto out;
7719
		if (dev_get_valid_name(net, dev, pat) < 0)
7720 7721 7722 7723 7724 7725 7726 7727
			goto out;
	}

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

	/* If device is running close it first. */
7728
	dev_close(dev);
7729 7730 7731 7732 7733 7734 7735 7736 7737 7738 7739 7740

	/* 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.
7741 7742 7743 7744

	   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.
7745 7746
	*/
	call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
7747 7748
	rcu_barrier();
	call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
7749
	rtmsg_ifinfo(RTM_DELLINK, dev, ~0U, GFP_KERNEL);
7750 7751 7752 7753

	/*
	 *	Flush the unicast and multicast chains
	 */
7754
	dev_uc_flush(dev);
7755
	dev_mc_flush(dev);
7756

7757 7758
	/* Send a netdev-removed uevent to the old namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_REMOVE);
7759
	netdev_adjacent_del_links(dev);
7760

7761
	/* Actually switch the network namespace */
7762
	dev_net_set(dev, net);
7763 7764

	/* If there is an ifindex conflict assign a new one */
7765
	if (__dev_get_by_index(net, dev->ifindex))
7766 7767
		dev->ifindex = dev_new_index(net);

7768 7769
	/* Send a netdev-add uevent to the new namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_ADD);
7770
	netdev_adjacent_add_links(dev);
7771

7772
	/* Fixup kobjects */
7773
	err = device_rename(&dev->dev, dev->name);
7774
	WARN_ON(err);
7775 7776 7777 7778 7779 7780 7781

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

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

7782 7783 7784 7785
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
7786
	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
7787

7788 7789 7790 7791 7792
	synchronize_net();
	err = 0;
out:
	return err;
}
7793
EXPORT_SYMBOL_GPL(dev_change_net_namespace);
7794

L
Linus Torvalds 已提交
7795 7796 7797 7798 7799 7800 7801 7802 7803
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;

7804
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
L
Linus Torvalds 已提交
7805 7806 7807 7808 7809 7810 7811 7812 7813 7814 7815 7816 7817 7818 7819 7820
		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. */
7821 7822 7823 7824 7825 7826
	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
7827 7828 7829 7830 7831 7832 7833 7834 7835 7836 7837 7838 7839 7840
	/* 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);
7841
	}
L
Linus Torvalds 已提交
7842 7843 7844 7845 7846

	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_enable();

	/* Process offline CPU's input_pkt_queue */
7847
	while ((skb = __skb_dequeue(&oldsd->process_queue))) {
7848
		netif_rx_ni(skb);
7849
		input_queue_head_incr(oldsd);
T
Tom Herbert 已提交
7850
	}
E
Eric Dumazet 已提交
7851
	while ((skb = skb_dequeue(&oldsd->input_pkt_queue))) {
7852
		netif_rx_ni(skb);
7853 7854
		input_queue_head_incr(oldsd);
	}
L
Linus Torvalds 已提交
7855 7856 7857 7858 7859

	return NOTIFY_OK;
}


7860
/**
7861 7862 7863 7864
 *	netdev_increment_features - increment feature set by one
 *	@all: current feature set
 *	@one: new feature set
 *	@mask: mask feature set
7865 7866
 *
 *	Computes a new feature set after adding a device with feature set
7867 7868
 *	@one to the master device with current feature set @all.  Will not
 *	enable anything that is off in @mask. Returns the new feature set.
7869
 */
7870 7871
netdev_features_t netdev_increment_features(netdev_features_t all,
	netdev_features_t one, netdev_features_t mask)
7872
{
7873
	if (mask & NETIF_F_HW_CSUM)
7874
		mask |= NETIF_F_CSUM_MASK;
7875
	mask |= NETIF_F_VLAN_CHALLENGED;
7876

7877
	all |= one & (NETIF_F_ONE_FOR_ALL | NETIF_F_CSUM_MASK) & mask;
7878
	all &= one | ~NETIF_F_ALL_FOR_ALL;
7879

7880
	/* If one device supports hw checksumming, set for all. */
7881 7882
	if (all & NETIF_F_HW_CSUM)
		all &= ~(NETIF_F_CSUM_MASK & ~NETIF_F_HW_CSUM);
7883 7884 7885

	return all;
}
7886
EXPORT_SYMBOL(netdev_increment_features);
7887

7888
static struct hlist_head * __net_init netdev_create_hash(void)
7889 7890 7891 7892 7893 7894 7895 7896 7897 7898 7899 7900
{
	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;
}

7901
/* Initialize per network namespace state */
7902
static int __net_init netdev_init(struct net *net)
7903
{
7904 7905
	if (net != &init_net)
		INIT_LIST_HEAD(&net->dev_base_head);
7906

7907 7908 7909
	net->dev_name_head = netdev_create_hash();
	if (net->dev_name_head == NULL)
		goto err_name;
7910

7911 7912 7913
	net->dev_index_head = netdev_create_hash();
	if (net->dev_index_head == NULL)
		goto err_idx;
7914 7915

	return 0;
7916 7917 7918 7919 7920

err_idx:
	kfree(net->dev_name_head);
err_name:
	return -ENOMEM;
7921 7922
}

7923 7924 7925 7926 7927 7928
/**
 *	netdev_drivername - network driver for the device
 *	@dev: network device
 *
 *	Determine network driver for device.
 */
7929
const char *netdev_drivername(const struct net_device *dev)
7930
{
7931 7932
	const struct device_driver *driver;
	const struct device *parent;
7933
	const char *empty = "";
7934 7935 7936

	parent = dev->dev.parent;
	if (!parent)
7937
		return empty;
7938 7939 7940

	driver = parent->driver;
	if (driver && driver->name)
7941 7942
		return driver->name;
	return empty;
7943 7944
}

7945 7946
static void __netdev_printk(const char *level, const struct net_device *dev,
			    struct va_format *vaf)
7947
{
7948
	if (dev && dev->dev.parent) {
7949 7950 7951 7952 7953 7954 7955
		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);
7956
	} else if (dev) {
7957 7958
		printk("%s%s%s: %pV",
		       level, netdev_name(dev), netdev_reg_state(dev), vaf);
7959
	} else {
7960
		printk("%s(NULL net_device): %pV", level, vaf);
7961
	}
7962 7963
}

7964 7965
void netdev_printk(const char *level, const struct net_device *dev,
		   const char *format, ...)
7966 7967 7968 7969 7970 7971 7972 7973 7974
{
	struct va_format vaf;
	va_list args;

	va_start(args, format);

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

7975
	__netdev_printk(level, dev, &vaf);
7976

7977 7978 7979 7980 7981
	va_end(args);
}
EXPORT_SYMBOL(netdev_printk);

#define define_netdev_printk_level(func, level)			\
7982
void func(const struct net_device *dev, const char *fmt, ...)	\
7983 7984 7985 7986 7987 7988 7989 7990 7991
{								\
	struct va_format vaf;					\
	va_list args;						\
								\
	va_start(args, fmt);					\
								\
	vaf.fmt = fmt;						\
	vaf.va = &args;						\
								\
7992
	__netdev_printk(level, dev, &vaf);			\
7993
								\
7994 7995 7996 7997 7998 7999 8000 8001 8002 8003 8004 8005
	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);

8006
static void __net_exit netdev_exit(struct net *net)
8007 8008 8009 8010 8011
{
	kfree(net->dev_name_head);
	kfree(net->dev_index_head);
}

8012
static struct pernet_operations __net_initdata netdev_net_ops = {
8013 8014 8015 8016
	.init = netdev_init,
	.exit = netdev_exit,
};

8017
static void __net_exit default_device_exit(struct net *net)
8018
{
8019
	struct net_device *dev, *aux;
8020
	/*
8021
	 * Push all migratable network devices back to the
8022 8023 8024
	 * initial network namespace
	 */
	rtnl_lock();
8025
	for_each_netdev_safe(net, dev, aux) {
8026
		int err;
8027
		char fb_name[IFNAMSIZ];
8028 8029 8030 8031 8032

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

8033 8034 8035
		/* Leave virtual devices for the generic cleanup */
		if (dev->rtnl_link_ops)
			continue;
8036

L
Lucas De Marchi 已提交
8037
		/* Push remaining network devices to init_net */
8038 8039
		snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex);
		err = dev_change_net_namespace(dev, &init_net, fb_name);
8040
		if (err) {
8041 8042
			pr_emerg("%s: failed to move %s to init_net: %d\n",
				 __func__, dev->name, err);
8043
			BUG();
8044 8045 8046 8047 8048
		}
	}
	rtnl_unlock();
}

8049 8050 8051 8052 8053 8054 8055
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;
8056
	DEFINE_WAIT_FUNC(wait, woken_wake_function);
8057

8058
	add_wait_queue(&netdev_unregistering_wq, &wait);
8059 8060 8061 8062 8063 8064 8065 8066 8067 8068 8069 8070
	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();
8071 8072

		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
8073
	}
8074
	remove_wait_queue(&netdev_unregistering_wq, &wait);
8075 8076
}

8077 8078 8079
static void __net_exit default_device_exit_batch(struct list_head *net_list)
{
	/* At exit all network devices most be removed from a network
8080
	 * namespace.  Do this in the reverse order of registration.
8081 8082 8083 8084 8085 8086 8087
	 * 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);

8088 8089 8090 8091 8092 8093 8094 8095 8096 8097 8098 8099
	/* 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);
8100 8101
	list_for_each_entry(net, net_list, exit_list) {
		for_each_netdev_reverse(net, dev) {
8102
			if (dev->rtnl_link_ops && dev->rtnl_link_ops->dellink)
8103 8104 8105 8106 8107 8108 8109 8110 8111
				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();
}

8112
static struct pernet_operations __net_initdata default_device_ops = {
8113
	.exit = default_device_exit,
8114
	.exit_batch = default_device_exit_batch,
8115 8116
};

L
Linus Torvalds 已提交
8117 8118 8119 8120 8121 8122 8123 8124 8125 8126 8127 8128 8129 8130 8131 8132 8133 8134 8135 8136
/*
 *	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;

8137
	if (netdev_kobject_init())
L
Linus Torvalds 已提交
8138 8139 8140
		goto out;

	INIT_LIST_HEAD(&ptype_all);
8141
	for (i = 0; i < PTYPE_HASH_SIZE; i++)
L
Linus Torvalds 已提交
8142 8143
		INIT_LIST_HEAD(&ptype_base[i]);

8144 8145
	INIT_LIST_HEAD(&offload_base);

8146 8147
	if (register_pernet_subsys(&netdev_net_ops))
		goto out;
L
Linus Torvalds 已提交
8148 8149 8150 8151 8152

	/*
	 *	Initialise the packet receive queues.
	 */

8153
	for_each_possible_cpu(i) {
E
Eric Dumazet 已提交
8154
		struct softnet_data *sd = &per_cpu(softnet_data, i);
L
Linus Torvalds 已提交
8155

E
Eric Dumazet 已提交
8156
		skb_queue_head_init(&sd->input_pkt_queue);
8157
		skb_queue_head_init(&sd->process_queue);
E
Eric Dumazet 已提交
8158
		INIT_LIST_HEAD(&sd->poll_list);
8159
		sd->output_queue_tailp = &sd->output_queue;
E
Eric Dumazet 已提交
8160
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
8161 8162 8163
		sd->csd.func = rps_trigger_softirq;
		sd->csd.info = sd;
		sd->cpu = i;
8164
#endif
T
Tom Herbert 已提交
8165

E
Eric Dumazet 已提交
8166 8167
		sd->backlog.poll = process_backlog;
		sd->backlog.weight = weight_p;
L
Linus Torvalds 已提交
8168 8169 8170 8171
	}

	dev_boot_phase = 0;

8172 8173 8174 8175 8176 8177 8178 8179 8180 8181 8182 8183 8184 8185 8186
	/* 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;

8187 8188
	open_softirq(NET_TX_SOFTIRQ, net_tx_action);
	open_softirq(NET_RX_SOFTIRQ, net_rx_action);
L
Linus Torvalds 已提交
8189 8190

	hotcpu_notifier(dev_cpu_callback, 0);
T
Thomas Graf 已提交
8191
	dst_subsys_init();
L
Linus Torvalds 已提交
8192 8193 8194 8195 8196 8197
	rc = 0;
out:
	return rc;
}

subsys_initcall(net_dev_init);