dev.c 221.8 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 *      NET3    Protocol independent device support routines.
L
Linus Torvalds 已提交
3 4 5 6 7 8 9
 *
 *		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
 *				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
24 25 26
 *                                      to 2 if register_netdev gets called
 *                                      before net_dev_init & also removed a
 *                                      few lines of code in the process.
L
Linus Torvalds 已提交
27 28 29 30 31 32 33 34 35 36 37 38
 *		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.
39
 *              Alan Cox        :       Fix ETH_P_ALL echoback lengths.
L
Linus Torvalds 已提交
40 41 42 43 44 45 46 47 48
 *		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.
49
 *              Alan Cox        :       Fixed nasty side effect of device close
L
Linus Torvalds 已提交
50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69
 *					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
70 71
 *                                      indefinitely on dev->refcnt
 *              J Hadi Salim    :       - Backlog queue sampling
L
Linus Torvalds 已提交
72 73 74
 *				        - netif_rx() feedback
 */

75
#include <linux/uaccess.h>
L
Linus Torvalds 已提交
76
#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>
84
#include <linux/sched/mm.h>
A
Arjan van de Ven 已提交
85
#include <linux/mutex.h>
L
Linus Torvalds 已提交
86 87 88 89 90 91 92 93 94
#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>
95
#include <linux/ethtool.h>
L
Linus Torvalds 已提交
96 97
#include <linux/notifier.h>
#include <linux/skbuff.h>
98
#include <linux/bpf.h>
D
David S. Miller 已提交
99
#include <linux/bpf_trace.h>
100
#include <net/net_namespace.h>
L
Linus Torvalds 已提交
101
#include <net/sock.h>
E
Eric Dumazet 已提交
102
#include <net/busy_poll.h>
L
Linus Torvalds 已提交
103 104 105
#include <linux/rtnetlink.h>
#include <linux/stat.h>
#include <net/dst.h>
106
#include <net/dst_metadata.h>
L
Linus Torvalds 已提交
107
#include <net/pkt_sched.h>
108
#include <net/pkt_cls.h>
L
Linus Torvalds 已提交
109
#include <net/checksum.h>
110
#include <net/xfrm.h>
L
Linus Torvalds 已提交
111 112 113 114 115 116 117 118
#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
119
#include <linux/audit.h>
120
#include <linux/dmaengine.h>
121
#include <linux/err.h>
122
#include <linux/ctype.h>
123
#include <linux/if_arp.h>
124
#include <linux/if_vlan.h>
125
#include <linux/ip.h>
126
#include <net/ip.h>
127
#include <net/mpls.h>
128 129
#include <linux/ipv6.h>
#include <linux/in.h>
D
David S. Miller 已提交
130 131
#include <linux/jhash.h>
#include <linux/random.h>
132
#include <trace/events/napi.h>
133
#include <trace/events/net.h>
134
#include <trace/events/skb.h>
135
#include <linux/pci.h>
136
#include <linux/inetdevice.h>
137
#include <linux/cpu_rmap.h>
138
#include <linux/static_key.h>
E
Eliezer Tamir 已提交
139
#include <linux/hashtable.h>
140
#include <linux/vmalloc.h>
141
#include <linux/if_macvlan.h>
142
#include <linux/errqueue.h>
143
#include <linux/hrtimer.h>
144
#include <linux/netfilter_ingress.h>
145
#include <linux/crash_dump.h>
146
#include <linux/sctp.h>
147
#include <net/udp_tunnel.h>
148
#include <linux/net_namespace.h>
L
Linus Torvalds 已提交
149

150 151
#include "net-sysfs.h"

152 153 154
/* Instead of increasing this, you should create a hash table. */
#define MAX_GRO_SKBS 8

H
Herbert Xu 已提交
155 156 157
/* 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
158
static DEFINE_SPINLOCK(ptype_lock);
159
static DEFINE_SPINLOCK(offload_lock);
160 161
struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
struct list_head ptype_all __read_mostly;	/* Taps */
162
static struct list_head offload_base __read_mostly;
L
Linus Torvalds 已提交
163

164
static int netif_rx_internal(struct sk_buff *skb);
165 166
static int call_netdevice_notifiers_info(unsigned long val,
					 struct netdev_notifier_info *info);
167
static struct napi_struct *napi_by_id(unsigned int napi_id);
168

L
Linus Torvalds 已提交
169
/*
170
 * The @dev_base_head list is protected by @dev_base_lock and the rtnl
L
Linus Torvalds 已提交
171 172
 * semaphore.
 *
173
 * Pure readers hold dev_base_lock for reading, or rcu_read_lock()
L
Linus Torvalds 已提交
174 175
 *
 * Writers must hold the rtnl semaphore while they loop through the
176
 * dev_base_head list, and hold dev_base_lock for writing when they do the
L
Linus Torvalds 已提交
177 178 179 180 181 182 183 184 185 186 187 188 189 190
 * 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);

191 192
static DEFINE_MUTEX(ifalias_mutex);

E
Eliezer Tamir 已提交
193 194 195
/* protects napi_hash addition/deletion and napi_gen_id */
static DEFINE_SPINLOCK(napi_hash_lock);

196
static unsigned int napi_gen_id = NR_CPUS;
197
static DEFINE_READ_MOSTLY_HASHTABLE(napi_hash, 8);
E
Eliezer Tamir 已提交
198

199
static seqcount_t devnet_rename_seq;
200

201 202
static inline void dev_base_seq_inc(struct net *net)
{
203 204
	while (++net->dev_base_seq == 0)
		;
205 206
}

207
static inline struct hlist_head *dev_name_hash(struct net *net, const char *name)
L
Linus Torvalds 已提交
208
{
209
	unsigned int hash = full_name_hash(net, name, strnlen(name, IFNAMSIZ));
210

211
	return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)];
L
Linus Torvalds 已提交
212 213
}

214
static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex)
L
Linus Torvalds 已提交
215
{
216
	return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)];
L
Linus Torvalds 已提交
217 218
}

E
Eric Dumazet 已提交
219
static inline void rps_lock(struct softnet_data *sd)
220 221
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
222
	spin_lock(&sd->input_pkt_queue.lock);
223 224 225
#endif
}

E
Eric Dumazet 已提交
226
static inline void rps_unlock(struct softnet_data *sd)
227 228
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
229
	spin_unlock(&sd->input_pkt_queue.lock);
230 231 232
#endif
}

233
/* Device list insertion */
234
static void list_netdevice(struct net_device *dev)
235
{
236
	struct net *net = dev_net(dev);
237 238 239 240

	ASSERT_RTNL();

	write_lock_bh(&dev_base_lock);
241
	list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
242
	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
243 244
	hlist_add_head_rcu(&dev->index_hlist,
			   dev_index_hash(net, dev->ifindex));
245
	write_unlock_bh(&dev_base_lock);
246 247

	dev_base_seq_inc(net);
248 249
}

250 251 252
/* Device list removal
 * caller must respect a RCU grace period before freeing/reusing dev
 */
253 254 255 256 257 258
static void unlist_netdevice(struct net_device *dev)
{
	ASSERT_RTNL();

	/* Unlink dev from the device chain */
	write_lock_bh(&dev_base_lock);
259
	list_del_rcu(&dev->dev_list);
260
	hlist_del_rcu(&dev->name_hlist);
261
	hlist_del_rcu(&dev->index_hlist);
262
	write_unlock_bh(&dev_base_lock);
263 264

	dev_base_seq_inc(dev_net(dev));
265 266
}

L
Linus Torvalds 已提交
267 268 269 270
/*
 *	Our notifier list
 */

271
static RAW_NOTIFIER_HEAD(netdev_chain);
L
Linus Torvalds 已提交
272 273 274 275 276

/*
 *	Device drivers call our routines to queue packets here. We empty the
 *	queue in the local softnet handler.
 */
277

278
DEFINE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
E
Eric Dumazet 已提交
279
EXPORT_PER_CPU_SYMBOL(softnet_data);
L
Linus Torvalds 已提交
280

281
#ifdef CONFIG_LOCKDEP
282
/*
283
 * register_netdevice() inits txq->_xmit_lock and sets lockdep class
284 285
 * according to dev->type
 */
286 287
static const unsigned short netdev_lock_type[] = {
	 ARPHRD_NETROM, ARPHRD_ETHER, ARPHRD_EETHER, ARPHRD_AX25,
288 289 290 291 292 293 294 295 296 297 298
	 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,
299 300 301
	 ARPHRD_FCFABRIC, ARPHRD_IEEE80211, ARPHRD_IEEE80211_PRISM,
	 ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, ARPHRD_PHONET_PIPE,
	 ARPHRD_IEEE802154, ARPHRD_VOID, ARPHRD_NONE};
302

303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318
static const char *const netdev_lock_name[] = {
	"_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",
	"_xmit_FCFABRIC", "_xmit_IEEE80211", "_xmit_IEEE80211_PRISM",
	"_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", "_xmit_PHONET_PIPE",
	"_xmit_IEEE802154", "_xmit_VOID", "_xmit_NONE"};
319 320

static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)];
321
static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)];
322 323 324 325 326 327 328 329 330 331 332 333

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;
}

334 335
static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
						 unsigned short dev_type)
336 337 338 339 340 341 342
{
	int i;

	i = netdev_lock_pos(dev_type);
	lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i],
				   netdev_lock_name[i]);
}
343 344 345 346 347 348 349 350 351 352

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]);
}
353
#else
354 355 356 357 358
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)
359 360 361
{
}
#endif
L
Linus Torvalds 已提交
362 363

/*******************************************************************************
364 365 366 367
 *
 *		Protocol management and registration routines
 *
 *******************************************************************************/
L
Linus Torvalds 已提交
368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385


/*
 *	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)
 */

386 387 388
static inline struct list_head *ptype_head(const struct packet_type *pt)
{
	if (pt->type == htons(ETH_P_ALL))
389
		return pt->dev ? &pt->dev->ptype_all : &ptype_all;
390
	else
391 392
		return pt->dev ? &pt->dev->ptype_specific :
				 &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK];
393 394
}

L
Linus Torvalds 已提交
395 396 397 398 399 400 401 402
/**
 *	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.
 *
403
 *	This call does not sleep therefore it can not
L
Linus Torvalds 已提交
404 405 406 407 408 409
 *	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)
{
410
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
411

412 413 414
	spin_lock(&ptype_lock);
	list_add_rcu(&pt->list, head);
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
415
}
E
Eric Dumazet 已提交
416
EXPORT_SYMBOL(dev_add_pack);
L
Linus Torvalds 已提交
417 418 419 420 421 422 423 424

/**
 *	__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
425
 *	returns.
L
Linus Torvalds 已提交
426 427 428 429 430 431 432
 *
 *      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)
{
433
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
434 435
	struct packet_type *pt1;

436
	spin_lock(&ptype_lock);
L
Linus Torvalds 已提交
437 438 439 440 441 442 443 444

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

445
	pr_warn("dev_remove_pack: %p not found\n", pt);
L
Linus Torvalds 已提交
446
out:
447
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
448
}
E
Eric Dumazet 已提交
449 450
EXPORT_SYMBOL(__dev_remove_pack);

L
Linus Torvalds 已提交
451 452 453 454 455 456 457 458 459 460 461 462 463 464 465
/**
 *	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);
466

L
Linus Torvalds 已提交
467 468
	synchronize_net();
}
E
Eric Dumazet 已提交
469
EXPORT_SYMBOL(dev_remove_pack);
L
Linus Torvalds 已提交
470

471 472 473 474 475 476 477 478 479 480 481 482 483 484 485

/**
 *	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)
{
486
	struct packet_offload *elem;
487 488

	spin_lock(&offload_lock);
489 490 491 492 493
	list_for_each_entry(elem, &offload_base, list) {
		if (po->priority < elem->priority)
			break;
	}
	list_add_rcu(&po->list, elem->list.prev);
494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510
	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
511
static void __dev_remove_offload(struct packet_offload *po)
512 513 514 515
{
	struct list_head *head = &offload_base;
	struct packet_offload *po1;

516
	spin_lock(&offload_lock);
517 518 519 520 521 522 523 524 525 526

	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:
527
	spin_unlock(&offload_lock);
528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
}

/**
 *	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
550
/******************************************************************************
551 552 553 554
 *
 *		      Device Boot-time Settings Routines
 *
 ******************************************************************************/
L
Linus Torvalds 已提交
555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576

/* 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));
577
			strlcpy(s[i].name, name, IFNAMSIZ);
L
Linus Torvalds 已提交
578 579 580 581 582 583 584 585 586
			memcpy(&s[i].map, map, sizeof(s[i].map));
			break;
		}
	}

	return i >= NETDEV_BOOT_SETUP_MAX ? 0 : 1;
}

/**
587 588
 * netdev_boot_setup_check	- check boot time settings
 * @dev: the netdevice
L
Linus Torvalds 已提交
589
 *
590 591 592 593
 * 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.
L
Linus Torvalds 已提交
594 595 596 597 598 599 600 601
 */
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] != ' ' &&
602
		    !strcmp(dev->name, s[i].name)) {
603 604 605 606
			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;
L
Linus Torvalds 已提交
607 608 609 610 611
			return 1;
		}
	}
	return 0;
}
E
Eric Dumazet 已提交
612
EXPORT_SYMBOL(netdev_boot_setup_check);
L
Linus Torvalds 已提交
613 614 615


/**
616 617 618 619 620 621 622 623
 * netdev_boot_base	- get address from boot time settings
 * @prefix: prefix for network device
 * @unit: id for network device
 *
 * Check boot time settings for the base address of device.
 * The found settings are set for the device to be used
 * later in the device probing.
 * Returns 0 if no settings found.
L
Linus Torvalds 已提交
624 625 626 627 628 629 630 631 632 633 634 635 636
 */
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
	 */
637
	if (__dev_get_by_name(&init_net, name))
L
Linus Torvalds 已提交
638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675
		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);

/*******************************************************************************
676 677 678 679
 *
 *			    Device Interface Subroutines
 *
 *******************************************************************************/
L
Linus Torvalds 已提交
680

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

694
	return dev->ifindex;
695 696 697
}
EXPORT_SYMBOL(dev_get_iflink);

698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723
/**
 *	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
724 725
/**
 *	__dev_get_by_name	- find a device by its name
726
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
727 728 729 730 731 732 733 734 735
 *	@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.
 */

736
struct net_device *__dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
737
{
E
Eric Dumazet 已提交
738 739
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);
L
Linus Torvalds 已提交
740

741
	hlist_for_each_entry(dev, head, name_hlist)
L
Linus Torvalds 已提交
742 743
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;
E
Eric Dumazet 已提交
744

L
Linus Torvalds 已提交
745 746
	return NULL;
}
E
Eric Dumazet 已提交
747
EXPORT_SYMBOL(__dev_get_by_name);
L
Linus Torvalds 已提交
748

749
/**
750 751 752 753 754 755 756 757 758
 * 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.
759 760 761 762 763 764 765
 */

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);

766
	hlist_for_each_entry_rcu(dev, head, name_hlist)
767 768 769 770 771 772 773
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_name_rcu);

L
Linus Torvalds 已提交
774 775
/**
 *	dev_get_by_name		- find a device by its name
776
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
777 778 779 780 781 782 783 784 785
 *	@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.
 */

786
struct net_device *dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
787 788 789
{
	struct net_device *dev;

790 791
	rcu_read_lock();
	dev = dev_get_by_name_rcu(net, name);
L
Linus Torvalds 已提交
792 793
	if (dev)
		dev_hold(dev);
794
	rcu_read_unlock();
L
Linus Torvalds 已提交
795 796
	return dev;
}
E
Eric Dumazet 已提交
797
EXPORT_SYMBOL(dev_get_by_name);
L
Linus Torvalds 已提交
798 799 800

/**
 *	__dev_get_by_index - find a device by its ifindex
801
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
802 803 804 805 806 807 808 809 810
 *	@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.
 */

811
struct net_device *__dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
812
{
E
Eric Dumazet 已提交
813 814
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);
L
Linus Torvalds 已提交
815

816
	hlist_for_each_entry(dev, head, index_hlist)
L
Linus Torvalds 已提交
817 818
		if (dev->ifindex == ifindex)
			return dev;
E
Eric Dumazet 已提交
819

L
Linus Torvalds 已提交
820 821
	return NULL;
}
E
Eric Dumazet 已提交
822
EXPORT_SYMBOL(__dev_get_by_index);
L
Linus Torvalds 已提交
823

824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839
/**
 *	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);

840
	hlist_for_each_entry_rcu(dev, head, index_hlist)
841 842 843 844 845 846 847
		if (dev->ifindex == ifindex)
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_index_rcu);

L
Linus Torvalds 已提交
848 849 850

/**
 *	dev_get_by_index - find a device by its ifindex
851
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
852 853 854 855 856 857 858 859
 *	@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.
 */

860
struct net_device *dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
861 862 863
{
	struct net_device *dev;

864 865
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, ifindex);
L
Linus Torvalds 已提交
866 867
	if (dev)
		dev_hold(dev);
868
	rcu_read_unlock();
L
Linus Torvalds 已提交
869 870
	return dev;
}
E
Eric Dumazet 已提交
871
EXPORT_SYMBOL(dev_get_by_index);
L
Linus Torvalds 已提交
872

873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897
/**
 *	dev_get_by_napi_id - find a device by napi_id
 *	@napi_id: ID of the NAPI struct
 *
 *	Search for an interface by NAPI ID. 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_napi_id(unsigned int napi_id)
{
	struct napi_struct *napi;

	WARN_ON_ONCE(!rcu_read_lock_held());

	if (napi_id < MIN_NAPI_ID)
		return NULL;

	napi = napi_by_id(napi_id);

	return napi ? napi->dev : NULL;
}
EXPORT_SYMBOL(dev_get_by_napi_id);

898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931
/**
 *	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
932
/**
933
 *	dev_getbyhwaddr_rcu - find a device by its hardware address
934
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
935 936 937 938
 *	@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
939 940
 *	is not found or a pointer to the device.
 *	The caller must hold RCU or RTNL.
941
 *	The returned device has not had its ref count increased
L
Linus Torvalds 已提交
942 943 944 945
 *	and the caller must therefore be careful about locking
 *
 */

946 947
struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
				       const char *ha)
L
Linus Torvalds 已提交
948 949 950
{
	struct net_device *dev;

951
	for_each_netdev_rcu(net, dev)
L
Linus Torvalds 已提交
952 953
		if (dev->type == type &&
		    !memcmp(dev->dev_addr, ha, dev->addr_len))
954 955 956
			return dev;

	return NULL;
L
Linus Torvalds 已提交
957
}
958
EXPORT_SYMBOL(dev_getbyhwaddr_rcu);
959

960
struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type)
L
Linus Torvalds 已提交
961 962 963
{
	struct net_device *dev;

964
	ASSERT_RTNL();
965
	for_each_netdev(net, dev)
966
		if (dev->type == type)
967 968 969
			return dev;

	return NULL;
970 971 972
}
EXPORT_SYMBOL(__dev_getfirstbyhwtype);

973
struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type)
974
{
975
	struct net_device *dev, *ret = NULL;
976

977 978 979 980 981 982 983 984 985
	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
986 987 988 989
}
EXPORT_SYMBOL(dev_getfirstbyhwtype);

/**
990
 *	__dev_get_by_flags - find any device with given flags
991
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
992 993 994 995
 *	@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
996
 *	is not found or a pointer to the device. Must be called inside
997
 *	rtnl_lock(), and result refcount is unchanged.
L
Linus Torvalds 已提交
998 999
 */

1000 1001
struct net_device *__dev_get_by_flags(struct net *net, unsigned short if_flags,
				      unsigned short mask)
L
Linus Torvalds 已提交
1002
{
1003
	struct net_device *dev, *ret;
L
Linus Torvalds 已提交
1004

1005 1006
	ASSERT_RTNL();

1007
	ret = NULL;
1008
	for_each_netdev(net, dev) {
L
Linus Torvalds 已提交
1009
		if (((dev->flags ^ if_flags) & mask) == 0) {
1010
			ret = dev;
L
Linus Torvalds 已提交
1011 1012 1013
			break;
		}
	}
1014
	return ret;
L
Linus Torvalds 已提交
1015
}
1016
EXPORT_SYMBOL(__dev_get_by_flags);
L
Linus Torvalds 已提交
1017 1018 1019 1020 1021 1022

/**
 *	dev_valid_name - check if name is okay for network device
 *	@name: name string
 *
 *	Network device names need to be valid file names to
1023 1024
 *	to allow sysfs to work.  We also disallow any kind of
 *	whitespace.
L
Linus Torvalds 已提交
1025
 */
1026
bool dev_valid_name(const char *name)
L
Linus Torvalds 已提交
1027
{
1028
	if (*name == '\0')
1029
		return false;
1030
	if (strlen(name) >= IFNAMSIZ)
1031
		return false;
1032
	if (!strcmp(name, ".") || !strcmp(name, ".."))
1033
		return false;
1034 1035

	while (*name) {
1036
		if (*name == '/' || *name == ':' || isspace(*name))
1037
			return false;
1038 1039
		name++;
	}
1040
	return true;
L
Linus Torvalds 已提交
1041
}
E
Eric Dumazet 已提交
1042
EXPORT_SYMBOL(dev_valid_name);
L
Linus Torvalds 已提交
1043 1044

/**
1045 1046
 *	__dev_alloc_name - allocate a name for a device
 *	@net: network namespace to allocate the device name in
L
Linus Torvalds 已提交
1047
 *	@name: name format string
1048
 *	@buf:  scratch buffer and result name string
L
Linus Torvalds 已提交
1049 1050
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
S
Stephen Hemminger 已提交
1051 1052 1053 1054 1055 1056
 *	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
1057 1058
 */

1059
static int __dev_alloc_name(struct net *net, const char *name, char *buf)
L
Linus Torvalds 已提交
1060 1061 1062 1063
{
	int i = 0;
	const char *p;
	const int max_netdevices = 8*PAGE_SIZE;
S
Stephen Hemminger 已提交
1064
	unsigned long *inuse;
L
Linus Torvalds 已提交
1065 1066
	struct net_device *d;

1067 1068 1069
	if (!dev_valid_name(name))
		return -EINVAL;

1070
	p = strchr(name, '%');
L
Linus Torvalds 已提交
1071 1072 1073 1074 1075 1076 1077 1078 1079 1080
	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
1081
		inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC);
L
Linus Torvalds 已提交
1082 1083 1084
		if (!inuse)
			return -ENOMEM;

1085
		for_each_netdev(net, d) {
L
Linus Torvalds 已提交
1086 1087 1088 1089 1090 1091
			if (!sscanf(d->name, name, &i))
				continue;
			if (i < 0 || i >= max_netdevices)
				continue;

			/*  avoid cases where sscanf is not exact inverse of printf */
1092
			snprintf(buf, IFNAMSIZ, name, i);
L
Linus Torvalds 已提交
1093 1094 1095 1096 1097 1098 1099 1100
			if (!strncmp(buf, d->name, IFNAMSIZ))
				set_bit(i, inuse);
		}

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

1101
	snprintf(buf, IFNAMSIZ, name, i);
1102
	if (!__dev_get_by_name(net, buf))
L
Linus Torvalds 已提交
1103 1104 1105 1106 1107 1108
		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.
	 */
1109
	return -ENFILE;
L
Linus Torvalds 已提交
1110 1111
}

1112 1113 1114 1115 1116 1117 1118
static int dev_alloc_name_ns(struct net *net,
			     struct net_device *dev,
			     const char *name)
{
	char buf[IFNAMSIZ];
	int ret;

1119
	BUG_ON(!net);
1120 1121 1122 1123
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
L
Linus Torvalds 已提交
1124 1125
}

1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141
/**
 *	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)
{
1142
	return dev_alloc_name_ns(dev_net(dev), dev, name);
1143
}
E
Eric Dumazet 已提交
1144
EXPORT_SYMBOL(dev_alloc_name);
1145

1146 1147
int dev_get_valid_name(struct net *net, struct net_device *dev,
		       const char *name)
1148
{
1149
	return dev_alloc_name_ns(net, dev, name);
1150
}
1151
EXPORT_SYMBOL(dev_get_valid_name);
L
Linus Torvalds 已提交
1152 1153 1154 1155 1156 1157 1158 1159 1160

/**
 *	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.
 */
1161
int dev_change_name(struct net_device *dev, const char *newname)
L
Linus Torvalds 已提交
1162
{
1163
	unsigned char old_assign_type;
1164
	char oldname[IFNAMSIZ];
L
Linus Torvalds 已提交
1165
	int err = 0;
1166
	int ret;
1167
	struct net *net;
L
Linus Torvalds 已提交
1168 1169

	ASSERT_RTNL();
1170
	BUG_ON(!dev_net(dev));
L
Linus Torvalds 已提交
1171

1172
	net = dev_net(dev);
L
Linus Torvalds 已提交
1173 1174 1175
	if (dev->flags & IFF_UP)
		return -EBUSY;

1176
	write_seqcount_begin(&devnet_rename_seq);
1177 1178

	if (strncmp(newname, dev->name, IFNAMSIZ) == 0) {
1179
		write_seqcount_end(&devnet_rename_seq);
1180
		return 0;
1181
	}
1182

1183 1184
	memcpy(oldname, dev->name, IFNAMSIZ);

1185
	err = dev_get_valid_name(net, dev, newname);
1186
	if (err < 0) {
1187
		write_seqcount_end(&devnet_rename_seq);
1188
		return err;
1189
	}
L
Linus Torvalds 已提交
1190

1191 1192 1193
	if (oldname[0] && !strchr(oldname, '%'))
		netdev_info(dev, "renamed from %s\n", oldname);

1194 1195 1196
	old_assign_type = dev->name_assign_type;
	dev->name_assign_type = NET_NAME_RENAMED;

1197
rollback:
1198 1199 1200
	ret = device_rename(&dev->dev, dev->name);
	if (ret) {
		memcpy(dev->name, oldname, IFNAMSIZ);
1201
		dev->name_assign_type = old_assign_type;
1202
		write_seqcount_end(&devnet_rename_seq);
1203
		return ret;
1204
	}
1205

1206
	write_seqcount_end(&devnet_rename_seq);
1207

1208 1209
	netdev_adjacent_rename_links(dev, oldname);

1210
	write_lock_bh(&dev_base_lock);
1211
	hlist_del_rcu(&dev->name_hlist);
1212 1213 1214 1215 1216 1217
	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));
1218 1219
	write_unlock_bh(&dev_base_lock);

1220
	ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
1221 1222 1223
	ret = notifier_to_errno(ret);

	if (ret) {
1224 1225
		/* err >= 0 after dev_alloc_name() or stores the first errno */
		if (err >= 0) {
1226
			err = ret;
1227
			write_seqcount_begin(&devnet_rename_seq);
1228
			memcpy(dev->name, oldname, IFNAMSIZ);
1229
			memcpy(oldname, newname, IFNAMSIZ);
1230 1231
			dev->name_assign_type = old_assign_type;
			old_assign_type = NET_NAME_RENAMED;
1232
			goto rollback;
1233
		} else {
1234
			pr_err("%s: name change rollback failed: %d\n",
1235
			       dev->name, ret);
1236 1237
		}
	}
L
Linus Torvalds 已提交
1238 1239 1240 1241

	return err;
}

1242 1243 1244 1245
/**
 *	dev_set_alias - change ifalias of a device
 *	@dev: device
 *	@alias: name up to IFALIASZ
1246
 *	@len: limit of bytes to copy from info
1247 1248 1249 1250 1251
 *
 *	Set ifalias for a device,
 */
int dev_set_alias(struct net_device *dev, const char *alias, size_t len)
{
1252
	struct dev_ifalias *new_alias = NULL;
1253 1254 1255 1256

	if (len >= IFALIASZ)
		return -EINVAL;

1257 1258 1259 1260 1261 1262 1263
	if (len) {
		new_alias = kmalloc(sizeof(*new_alias) + len + 1, GFP_KERNEL);
		if (!new_alias)
			return -ENOMEM;

		memcpy(new_alias->ifalias, alias, len);
		new_alias->ifalias[len] = 0;
1264 1265
	}

1266 1267 1268 1269 1270 1271 1272
	mutex_lock(&ifalias_mutex);
	rcu_swap_protected(dev->ifalias, new_alias,
			   mutex_is_locked(&ifalias_mutex));
	mutex_unlock(&ifalias_mutex);

	if (new_alias)
		kfree_rcu(new_alias, rcuhead);
1273 1274 1275 1276

	return len;
}

1277 1278 1279
/**
 *	dev_get_alias - get ifalias of a device
 *	@dev: device
1280
 *	@name: buffer to store name of ifalias
1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298
 *	@len: size of buffer
 *
 *	get ifalias for a device.  Caller must make sure dev cannot go
 *	away,  e.g. rcu read lock or own a reference count to device.
 */
int dev_get_alias(const struct net_device *dev, char *name, size_t len)
{
	const struct dev_ifalias *alias;
	int ret = 0;

	rcu_read_lock();
	alias = rcu_dereference(dev->ifalias);
	if (alias)
		ret = snprintf(name, len, "%s", alias->ifalias);
	rcu_read_unlock();

	return ret;
}
1299

1300
/**
S
Stephen Hemminger 已提交
1301
 *	netdev_features_change - device changes features
1302 1303 1304 1305 1306 1307
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed features.
 */
void netdev_features_change(struct net_device *dev)
{
1308
	call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev);
1309 1310 1311
}
EXPORT_SYMBOL(netdev_features_change);

L
Linus Torvalds 已提交
1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322
/**
 *	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) {
1323 1324 1325
		struct netdev_notifier_change_info change_info = {
			.info.dev = dev,
		};
1326

1327
		call_netdevice_notifiers_info(NETDEV_CHANGE,
1328
					      &change_info.info);
1329
		rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
L
Linus Torvalds 已提交
1330 1331
	}
}
E
Eric Dumazet 已提交
1332
EXPORT_SYMBOL(netdev_state_change);
L
Linus Torvalds 已提交
1333

1334
/**
1335 1336
 * netdev_notify_peers - notify network peers about existence of @dev
 * @dev: network device
1337 1338 1339 1340 1341 1342 1343 1344
 *
 * 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)
1345
{
1346 1347
	rtnl_lock();
	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, dev);
1348
	call_netdevice_notifiers(NETDEV_RESEND_IGMP, dev);
1349
	rtnl_unlock();
1350
}
1351
EXPORT_SYMBOL(netdev_notify_peers);
1352

1353
static int __dev_open(struct net_device *dev)
L
Linus Torvalds 已提交
1354
{
1355
	const struct net_device_ops *ops = dev->netdev_ops;
1356
	int ret;
L
Linus Torvalds 已提交
1357

1358 1359
	ASSERT_RTNL();

L
Linus Torvalds 已提交
1360 1361 1362
	if (!netif_device_present(dev))
		return -ENODEV;

1363 1364 1365 1366
	/* 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
	 */
1367
	netpoll_poll_disable(dev);
1368

1369 1370 1371 1372 1373
	ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

L
Linus Torvalds 已提交
1374
	set_bit(__LINK_STATE_START, &dev->state);
1375

1376 1377
	if (ops->ndo_validate_addr)
		ret = ops->ndo_validate_addr(dev);
1378

1379 1380
	if (!ret && ops->ndo_open)
		ret = ops->ndo_open(dev);
L
Linus Torvalds 已提交
1381

1382
	netpoll_poll_enable(dev);
1383

1384 1385 1386
	if (ret)
		clear_bit(__LINK_STATE_START, &dev->state);
	else {
L
Linus Torvalds 已提交
1387
		dev->flags |= IFF_UP;
1388
		dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
1389
		dev_activate(dev);
1390
		add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
1391
	}
1392

L
Linus Torvalds 已提交
1393 1394 1395 1396
	return ret;
}

/**
1397 1398
 *	dev_open	- prepare an interface for use.
 *	@dev:	device to open
L
Linus Torvalds 已提交
1399
 *
1400 1401 1402 1403 1404 1405 1406
 *	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
1407
 */
1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418
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;

1419
	rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1420 1421 1422 1423 1424 1425
	call_netdevice_notifiers(NETDEV_UP, dev);

	return ret;
}
EXPORT_SYMBOL(dev_open);

1426
static void __dev_close_many(struct list_head *head)
L
Linus Torvalds 已提交
1427
{
1428
	struct net_device *dev;
1429

1430
	ASSERT_RTNL();
1431 1432
	might_sleep();

1433
	list_for_each_entry(dev, head, close_list) {
1434
		/* Temporarily disable netpoll until the interface is down */
1435
		netpoll_poll_disable(dev);
1436

1437
		call_netdevice_notifiers(NETDEV_GOING_DOWN, dev);
L
Linus Torvalds 已提交
1438

1439
		clear_bit(__LINK_STATE_START, &dev->state);
L
Linus Torvalds 已提交
1440

1441 1442 1443 1444 1445 1446
		/* 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.
		 */
1447
		smp_mb__after_atomic(); /* Commit netif_running(). */
1448
	}
L
Linus Torvalds 已提交
1449

1450
	dev_deactivate_many(head);
1451

1452
	list_for_each_entry(dev, head, close_list) {
1453
		const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
1454

1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465
		/*
		 *	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;
1466
		netpoll_poll_enable(dev);
1467 1468 1469
	}
}

1470
static void __dev_close(struct net_device *dev)
1471 1472 1473
{
	LIST_HEAD(single);

1474
	list_add(&dev->close_list, &single);
1475
	__dev_close_many(&single);
1476
	list_del(&single);
1477 1478
}

1479
void dev_close_many(struct list_head *head, bool unlink)
1480 1481
{
	struct net_device *dev, *tmp;
L
Linus Torvalds 已提交
1482

1483 1484
	/* Remove the devices that don't need to be closed */
	list_for_each_entry_safe(dev, tmp, head, close_list)
1485
		if (!(dev->flags & IFF_UP))
1486
			list_del_init(&dev->close_list);
1487 1488

	__dev_close_many(head);
L
Linus Torvalds 已提交
1489

1490
	list_for_each_entry_safe(dev, tmp, head, close_list) {
1491
		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING, GFP_KERNEL);
1492
		call_netdevice_notifiers(NETDEV_DOWN, dev);
1493 1494
		if (unlink)
			list_del_init(&dev->close_list);
1495
	}
1496
}
1497
EXPORT_SYMBOL(dev_close_many);
1498 1499 1500 1501 1502 1503 1504 1505 1506 1507

/**
 *	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.
 */
1508
void dev_close(struct net_device *dev)
1509
{
1510 1511
	if (dev->flags & IFF_UP) {
		LIST_HEAD(single);
L
Linus Torvalds 已提交
1512

1513
		list_add(&dev->close_list, &single);
1514
		dev_close_many(&single, true);
1515 1516
		list_del(&single);
	}
L
Linus Torvalds 已提交
1517
}
E
Eric Dumazet 已提交
1518
EXPORT_SYMBOL(dev_close);
L
Linus Torvalds 已提交
1519 1520


1521 1522 1523 1524 1525 1526 1527 1528 1529 1530
/**
 *	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)
{
1531 1532
	struct net_device *lower_dev;
	struct list_head *iter;
1533

M
Michał Mirosław 已提交
1534 1535
	dev->wanted_features &= ~NETIF_F_LRO;
	netdev_update_features(dev);
1536

1537 1538
	if (unlikely(dev->features & NETIF_F_LRO))
		netdev_WARN(dev, "failed to disable LRO!\n");
1539 1540 1541

	netdev_for_each_lower_dev(dev, lower_dev, iter)
		dev_disable_lro(lower_dev);
1542 1543 1544
}
EXPORT_SYMBOL(dev_disable_lro);

1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
/**
 *	dev_disable_gro_hw - disable HW Generic Receive Offload on a device
 *	@dev: device
 *
 *	Disable HW Generic Receive Offload (GRO_HW) on a net device.  Must be
 *	called under RTNL.  This is needed if Generic XDP is installed on
 *	the device.
 */
static void dev_disable_gro_hw(struct net_device *dev)
{
	dev->wanted_features &= ~NETIF_F_GRO_HW;
	netdev_update_features(dev);

	if (unlikely(dev->features & NETIF_F_GRO_HW))
		netdev_WARN(dev, "failed to disable GRO_HW!\n");
}

1562 1563 1564
static int call_netdevice_notifier(struct notifier_block *nb, unsigned long val,
				   struct net_device *dev)
{
1565 1566 1567
	struct netdev_notifier_info info = {
		.dev = dev,
	};
1568 1569 1570

	return nb->notifier_call(nb, val, &info);
}
1571

1572 1573
static int dev_boot_phase = 1;

L
Linus Torvalds 已提交
1574
/**
1575 1576
 * register_netdevice_notifier - register a network notifier block
 * @nb: notifier
L
Linus Torvalds 已提交
1577
 *
1578 1579 1580 1581
 * 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.
L
Linus Torvalds 已提交
1582
 *
1583 1584 1585
 * When registered all registration and up events are replayed
 * to the new notifier to allow device to have a race free
 * view of the network device list.
L
Linus Torvalds 已提交
1586 1587 1588 1589 1590
 */

int register_netdevice_notifier(struct notifier_block *nb)
{
	struct net_device *dev;
1591
	struct net_device *last;
1592
	struct net *net;
L
Linus Torvalds 已提交
1593 1594 1595
	int err;

	rtnl_lock();
1596
	err = raw_notifier_chain_register(&netdev_chain, nb);
1597 1598
	if (err)
		goto unlock;
1599 1600 1601 1602
	if (dev_boot_phase)
		goto unlock;
	for_each_net(net) {
		for_each_netdev(net, dev) {
1603
			err = call_netdevice_notifier(nb, NETDEV_REGISTER, dev);
1604 1605 1606 1607 1608 1609
			err = notifier_to_errno(err);
			if (err)
				goto rollback;

			if (!(dev->flags & IFF_UP))
				continue;
L
Linus Torvalds 已提交
1610

1611
			call_netdevice_notifier(nb, NETDEV_UP, dev);
1612
		}
L
Linus Torvalds 已提交
1613
	}
1614 1615

unlock:
L
Linus Torvalds 已提交
1616 1617
	rtnl_unlock();
	return err;
1618 1619 1620

rollback:
	last = dev;
1621 1622 1623
	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev == last)
1624
				goto outroll;
1625

1626
			if (dev->flags & IFF_UP) {
1627 1628 1629
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1630
			}
1631
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1632 1633
		}
	}
1634

1635
outroll:
1636
	raw_notifier_chain_unregister(&netdev_chain, nb);
1637
	goto unlock;
L
Linus Torvalds 已提交
1638
}
E
Eric Dumazet 已提交
1639
EXPORT_SYMBOL(register_netdevice_notifier);
L
Linus Torvalds 已提交
1640 1641

/**
1642 1643
 * unregister_netdevice_notifier - unregister a network notifier block
 * @nb: notifier
L
Linus Torvalds 已提交
1644
 *
1645 1646 1647 1648
 * 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.
1649
 *
1650 1651 1652
 * 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
1653 1654 1655 1656
 */

int unregister_netdevice_notifier(struct notifier_block *nb)
{
1657 1658
	struct net_device *dev;
	struct net *net;
1659 1660 1661
	int err;

	rtnl_lock();
1662
	err = raw_notifier_chain_unregister(&netdev_chain, nb);
1663 1664 1665 1666 1667 1668
	if (err)
		goto unlock;

	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev->flags & IFF_UP) {
1669 1670 1671
				call_netdevice_notifier(nb, NETDEV_GOING_DOWN,
							dev);
				call_netdevice_notifier(nb, NETDEV_DOWN, dev);
1672
			}
1673
			call_netdevice_notifier(nb, NETDEV_UNREGISTER, dev);
1674 1675 1676
		}
	}
unlock:
1677 1678
	rtnl_unlock();
	return err;
L
Linus Torvalds 已提交
1679
}
E
Eric Dumazet 已提交
1680
EXPORT_SYMBOL(unregister_netdevice_notifier);
L
Linus Torvalds 已提交
1681

1682 1683 1684 1685 1686 1687 1688 1689 1690 1691
/**
 *	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
1692 1693
static int call_netdevice_notifiers_info(unsigned long val,
					 struct netdev_notifier_info *info)
1694 1695 1696 1697 1698
{
	ASSERT_RTNL();
	return raw_notifier_call_chain(&netdev_chain, val, info);
}

L
Linus Torvalds 已提交
1699 1700 1701
/**
 *	call_netdevice_notifiers - call all network notifier blocks
 *      @val: value passed unmodified to notifier function
1702
 *      @dev: net_device pointer passed unmodified to notifier function
L
Linus Torvalds 已提交
1703 1704
 *
 *	Call all network notifier blocks.  Parameters and return value
1705
 *	are as for raw_notifier_call_chain().
L
Linus Torvalds 已提交
1706 1707
 */

1708
int call_netdevice_notifiers(unsigned long val, struct net_device *dev)
L
Linus Torvalds 已提交
1709
{
1710 1711 1712
	struct netdev_notifier_info info = {
		.dev = dev,
	};
1713

1714
	return call_netdevice_notifiers_info(val, &info);
L
Linus Torvalds 已提交
1715
}
1716
EXPORT_SYMBOL(call_netdevice_notifiers);
L
Linus Torvalds 已提交
1717

1718
#ifdef CONFIG_NET_INGRESS
1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733
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
1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749
#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

1750
static struct static_key netstamp_needed __read_mostly;
1751 1752
#ifdef HAVE_JUMP_LABEL
static atomic_t netstamp_needed_deferred;
1753
static atomic_t netstamp_wanted;
1754
static void netstamp_clear(struct work_struct *work)
L
Linus Torvalds 已提交
1755
{
1756
	int deferred = atomic_xchg(&netstamp_needed_deferred, 0);
1757
	int wanted;
1758

1759 1760 1761 1762 1763
	wanted = atomic_add_return(deferred, &netstamp_wanted);
	if (wanted > 0)
		static_key_enable(&netstamp_needed);
	else
		static_key_disable(&netstamp_needed);
1764 1765
}
static DECLARE_WORK(netstamp_work, netstamp_clear);
1766
#endif
1767 1768 1769

void net_enable_timestamp(void)
{
1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782
#ifdef HAVE_JUMP_LABEL
	int wanted;

	while (1) {
		wanted = atomic_read(&netstamp_wanted);
		if (wanted <= 0)
			break;
		if (atomic_cmpxchg(&netstamp_wanted, wanted, wanted + 1) == wanted)
			return;
	}
	atomic_inc(&netstamp_needed_deferred);
	schedule_work(&netstamp_work);
#else
1783
	static_key_slow_inc(&netstamp_needed);
1784
#endif
L
Linus Torvalds 已提交
1785
}
E
Eric Dumazet 已提交
1786
EXPORT_SYMBOL(net_enable_timestamp);
L
Linus Torvalds 已提交
1787 1788 1789

void net_disable_timestamp(void)
{
1790
#ifdef HAVE_JUMP_LABEL
1791 1792 1793 1794 1795 1796 1797 1798 1799 1800
	int wanted;

	while (1) {
		wanted = atomic_read(&netstamp_wanted);
		if (wanted <= 1)
			break;
		if (atomic_cmpxchg(&netstamp_wanted, wanted, wanted - 1) == wanted)
			return;
	}
	atomic_dec(&netstamp_needed_deferred);
1801 1802
	schedule_work(&netstamp_work);
#else
1803
	static_key_slow_dec(&netstamp_needed);
1804
#endif
L
Linus Torvalds 已提交
1805
}
E
Eric Dumazet 已提交
1806
EXPORT_SYMBOL(net_disable_timestamp);
L
Linus Torvalds 已提交
1807

E
Eric Dumazet 已提交
1808
static inline void net_timestamp_set(struct sk_buff *skb)
L
Linus Torvalds 已提交
1809
{
T
Thomas Gleixner 已提交
1810
	skb->tstamp = 0;
1811
	if (static_key_false(&netstamp_needed))
1812
		__net_timestamp(skb);
L
Linus Torvalds 已提交
1813 1814
}

1815
#define net_timestamp_check(COND, SKB)			\
1816
	if (static_key_false(&netstamp_needed)) {		\
T
Thomas Gleixner 已提交
1817
		if ((COND) && !(SKB)->tstamp)	\
1818 1819
			__net_timestamp(SKB);		\
	}						\
E
Eric Dumazet 已提交
1820

1821
bool is_skb_forwardable(const struct net_device *dev, const struct sk_buff *skb)
1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839
{
	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;
}
1840
EXPORT_SYMBOL_GPL(is_skb_forwardable);
1841

H
Herbert Xu 已提交
1842 1843
int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb)
{
1844
	int ret = ____dev_forward_skb(dev, skb);
H
Herbert Xu 已提交
1845

1846 1847 1848 1849
	if (likely(!ret)) {
		skb->protocol = eth_type_trans(skb, dev);
		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
	}
H
Herbert Xu 已提交
1850

1851
	return ret;
H
Herbert Xu 已提交
1852 1853 1854
}
EXPORT_SYMBOL_GPL(__dev_forward_skb);

1855 1856 1857 1858 1859 1860 1861 1862
/**
 * dev_forward_skb - loopback an skb to another netif
 *
 * @dev: destination network device
 * @skb: buffer to forward
 *
 * return values:
 *	NET_RX_SUCCESS	(no congestion)
1863
 *	NET_RX_DROP     (packet was dropped, but freed)
1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874
 *
 * 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
1875
	return __dev_forward_skb(dev, skb) ?: netif_rx_internal(skb);
1876 1877 1878
}
EXPORT_SYMBOL_GPL(dev_forward_skb);

1879 1880 1881 1882
static inline int deliver_skb(struct sk_buff *skb,
			      struct packet_type *pt_prev,
			      struct net_device *orig_dev)
{
W
Willem de Bruijn 已提交
1883
	if (unlikely(skb_orphan_frags_rx(skb, GFP_ATOMIC)))
1884
		return -ENOMEM;
1885
	refcount_inc(&skb->users);
1886 1887 1888
	return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
}

1889 1890
static inline void deliver_ptype_list_skb(struct sk_buff *skb,
					  struct packet_type **pt,
1891 1892
					  struct net_device *orig_dev,
					  __be16 type,
1893 1894 1895 1896 1897 1898 1899 1900
					  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)
1901
			deliver_skb(skb, pt_prev, orig_dev);
1902 1903 1904 1905 1906
		pt_prev = ptype;
	}
	*pt = pt_prev;
}

1907 1908
static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
{
1909
	if (!ptype->af_packet_priv || !skb->sk)
1910 1911 1912 1913 1914 1915 1916 1917 1918 1919
		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
1920 1921 1922 1923 1924
/*
 *	Support routine. Sends outgoing frames to any network
 *	taps currently in use.
 */

1925
void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
L
Linus Torvalds 已提交
1926 1927
{
	struct packet_type *ptype;
1928 1929
	struct sk_buff *skb2 = NULL;
	struct packet_type *pt_prev = NULL;
1930
	struct list_head *ptype_list = &ptype_all;
1931

L
Linus Torvalds 已提交
1932
	rcu_read_lock();
1933 1934
again:
	list_for_each_entry_rcu(ptype, ptype_list, list) {
L
Linus Torvalds 已提交
1935 1936 1937
		/* Never send packets back to the socket
		 * they originated from - MvS (miquels@drinkel.ow.org)
		 */
1938 1939
		if (skb_loop_sk(ptype, skb))
			continue;
1940

1941 1942 1943 1944 1945
		if (pt_prev) {
			deliver_skb(skb2, pt_prev, skb->dev);
			pt_prev = ptype;
			continue;
		}
L
Linus Torvalds 已提交
1946

1947 1948 1949 1950
		/* need to clone skb, done only once */
		skb2 = skb_clone(skb, GFP_ATOMIC);
		if (!skb2)
			goto out_unlock;
1951

1952
		net_timestamp_set(skb2);
L
Linus Torvalds 已提交
1953

1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965
		/* 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
1966
		}
1967 1968 1969 1970 1971 1972 1973 1974 1975

		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
1976
	}
1977
out_unlock:
1978 1979 1980 1981 1982 1983
	if (pt_prev) {
		if (!skb_orphan_frags_rx(skb2, GFP_ATOMIC))
			pt_prev->func(skb2, skb->dev, pt_prev, skb->dev);
		else
			kfree_skb(skb2);
	}
L
Linus Torvalds 已提交
1984 1985
	rcu_read_unlock();
}
1986
EXPORT_SYMBOL_GPL(dev_queue_xmit_nit);
L
Linus Torvalds 已提交
1987

1988 1989
/**
 * netif_setup_tc - Handle tc mappings on real_num_tx_queues change
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
 * @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
2001
static void netif_setup_tc(struct net_device *dev, unsigned int txq)
2002 2003 2004 2005 2006 2007
{
	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) {
2008
		pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n");
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
		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) {
2019 2020
			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);
2021 2022 2023 2024 2025
			netdev_set_prio_tc_map(dev, i, 0);
		}
	}
}

2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041
int netdev_txq_to_tc(struct net_device *dev, unsigned int txq)
{
	if (dev->num_tc) {
		struct netdev_tc_txq *tc = &dev->tc_to_txq[0];
		int i;

		for (i = 0; i < TC_MAX_QUEUE; i++, tc++) {
			if ((txq - tc->offset) < tc->count)
				return i;
		}

		return -1;
	}

	return 0;
}
2042
EXPORT_SYMBOL(netdev_txq_to_tc);
2043

2044 2045 2046 2047 2048
#ifdef CONFIG_XPS
static DEFINE_MUTEX(xps_map_mutex);
#define xmap_dereference(P)		\
	rcu_dereference_protected((P), lockdep_is_held(&xps_map_mutex))

2049 2050
static bool remove_xps_queue(struct xps_dev_maps *dev_maps,
			     int tci, u16 index)
2051
{
2052 2053
	struct xps_map *map = NULL;
	int pos;
2054

2055
	if (dev_maps)
2056 2057 2058
		map = xmap_dereference(dev_maps->cpu_map[tci]);
	if (!map)
		return false;
2059

2060 2061 2062 2063 2064 2065
	for (pos = map->len; pos--;) {
		if (map->queues[pos] != index)
			continue;

		if (map->len > 1) {
			map->queues[pos] = map->queues[--map->len];
2066
			break;
2067
		}
2068 2069 2070 2071

		RCU_INIT_POINTER(dev_maps->cpu_map[tci], NULL);
		kfree_rcu(map, rcu);
		return false;
2072 2073
	}

2074
	return true;
2075 2076
}

2077 2078 2079 2080
static bool remove_xps_queue_cpu(struct net_device *dev,
				 struct xps_dev_maps *dev_maps,
				 int cpu, u16 offset, u16 count)
{
2081 2082 2083
	int num_tc = dev->num_tc ? : 1;
	bool active = false;
	int tci;
2084

2085 2086 2087 2088 2089 2090 2091 2092 2093
	for (tci = cpu * num_tc; num_tc--; tci++) {
		int i, j;

		for (i = count, j = offset; i--; j++) {
			if (!remove_xps_queue(dev_maps, cpu, j))
				break;
		}

		active |= i < 0;
2094 2095
	}

2096
	return active;
2097 2098 2099 2100
}

static void netif_reset_xps_queues(struct net_device *dev, u16 offset,
				   u16 count)
2101 2102
{
	struct xps_dev_maps *dev_maps;
2103
	int cpu, i;
2104 2105 2106 2107 2108 2109 2110 2111
	bool active = false;

	mutex_lock(&xps_map_mutex);
	dev_maps = xmap_dereference(dev->xps_maps);

	if (!dev_maps)
		goto out_no_maps;

2112 2113 2114
	for_each_possible_cpu(cpu)
		active |= remove_xps_queue_cpu(dev, dev_maps, cpu,
					       offset, count);
2115 2116

	if (!active) {
2117 2118 2119 2120
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

2121
	for (i = offset + (count - 1); count--; i--)
2122 2123 2124
		netdev_queue_numa_node_write(netdev_get_tx_queue(dev, i),
					     NUMA_NO_NODE);

2125 2126 2127 2128
out_no_maps:
	mutex_unlock(&xps_map_mutex);
}

2129 2130 2131 2132 2133
static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
{
	netif_reset_xps_queues(dev, index, dev->num_tx_queues - index);
}

2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168
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;
}

2169 2170
int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
			u16 index)
2171
{
2172
	struct xps_dev_maps *dev_maps, *new_dev_maps = NULL;
2173 2174
	int i, cpu, tci, numa_node_id = -2;
	int maps_sz, num_tc = 1, tc = 0;
2175
	struct xps_map *map, *new_map;
2176
	bool active = false;
2177

2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188
	if (dev->num_tc) {
		num_tc = dev->num_tc;
		tc = netdev_txq_to_tc(dev, index);
		if (tc < 0)
			return -EINVAL;
	}

	maps_sz = XPS_DEV_MAPS_SIZE(num_tc);
	if (maps_sz < L1_CACHE_BYTES)
		maps_sz = L1_CACHE_BYTES;

2189 2190 2191 2192
	mutex_lock(&xps_map_mutex);

	dev_maps = xmap_dereference(dev->xps_maps);

2193
	/* allocate memory for queue storage */
2194
	for_each_cpu_and(cpu, cpu_online_mask, mask) {
2195 2196
		if (!new_dev_maps)
			new_dev_maps = kzalloc(maps_sz, GFP_KERNEL);
2197 2198
		if (!new_dev_maps) {
			mutex_unlock(&xps_map_mutex);
2199
			return -ENOMEM;
2200
		}
2201

2202 2203
		tci = cpu * num_tc + tc;
		map = dev_maps ? xmap_dereference(dev_maps->cpu_map[tci]) :
2204 2205 2206 2207 2208 2209
				 NULL;

		map = expand_xps_map(map, cpu, index);
		if (!map)
			goto error;

2210
		RCU_INIT_POINTER(new_dev_maps->cpu_map[tci], map);
2211 2212 2213 2214 2215
	}

	if (!new_dev_maps)
		goto out_no_new_maps;

2216
	for_each_possible_cpu(cpu) {
2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228
		/* copy maps belonging to foreign traffic classes */
		for (i = tc, tci = cpu * num_tc; dev_maps && i--; tci++) {
			/* fill in the new device map from the old device map */
			map = xmap_dereference(dev_maps->cpu_map[tci]);
			RCU_INIT_POINTER(new_dev_maps->cpu_map[tci], map);
		}

		/* We need to explicitly update tci as prevous loop
		 * could break out early if dev_maps is NULL.
		 */
		tci = cpu * num_tc + tc;

2229 2230 2231 2232
		if (cpumask_test_cpu(cpu, mask) && cpu_online(cpu)) {
			/* add queue to CPU maps */
			int pos = 0;

2233
			map = xmap_dereference(new_dev_maps->cpu_map[tci]);
2234 2235 2236 2237 2238
			while ((pos < map->len) && (map->queues[pos] != index))
				pos++;

			if (pos == map->len)
				map->queues[map->len++] = index;
2239 2240 2241 2242 2243 2244
#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
2245 2246
		} else if (dev_maps) {
			/* fill in the new device map from the old device map */
2247 2248
			map = xmap_dereference(dev_maps->cpu_map[tci]);
			RCU_INIT_POINTER(new_dev_maps->cpu_map[tci], map);
2249
		}
2250

2251 2252 2253 2254 2255 2256
		/* copy maps belonging to foreign traffic classes */
		for (i = num_tc - tc, tci++; dev_maps && --i; tci++) {
			/* fill in the new device map from the old device map */
			map = xmap_dereference(dev_maps->cpu_map[tci]);
			RCU_INIT_POINTER(new_dev_maps->cpu_map[tci], map);
		}
2257 2258
	}

2259 2260
	rcu_assign_pointer(dev->xps_maps, new_dev_maps);

2261
	/* Cleanup old maps */
2262 2263 2264 2265 2266 2267 2268
	if (!dev_maps)
		goto out_no_old_maps;

	for_each_possible_cpu(cpu) {
		for (i = num_tc, tci = cpu * num_tc; i--; tci++) {
			new_map = xmap_dereference(new_dev_maps->cpu_map[tci]);
			map = xmap_dereference(dev_maps->cpu_map[tci]);
2269 2270 2271
			if (map && map != new_map)
				kfree_rcu(map, rcu);
		}
2272 2273
	}

2274 2275 2276
	kfree_rcu(dev_maps, rcu);

out_no_old_maps:
2277 2278
	dev_maps = new_dev_maps;
	active = true;
2279

2280 2281
out_no_new_maps:
	/* update Tx queue numa node */
2282 2283 2284 2285
	netdev_queue_numa_node_write(netdev_get_tx_queue(dev, index),
				     (numa_node_id >= 0) ? numa_node_id :
				     NUMA_NO_NODE);

2286 2287 2288 2289 2290
	if (!dev_maps)
		goto out_no_maps;

	/* removes queue from unused CPUs */
	for_each_possible_cpu(cpu) {
2291 2292 2293 2294 2295 2296
		for (i = tc, tci = cpu * num_tc; i--; tci++)
			active |= remove_xps_queue(dev_maps, tci, index);
		if (!cpumask_test_cpu(cpu, mask) || !cpu_online(cpu))
			active |= remove_xps_queue(dev_maps, tci, index);
		for (i = num_tc - tc, tci++; --i; tci++)
			active |= remove_xps_queue(dev_maps, tci, index);
2297 2298 2299 2300 2301 2302 2303 2304 2305
	}

	/* free map if not active */
	if (!active) {
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

out_no_maps:
2306 2307 2308 2309
	mutex_unlock(&xps_map_mutex);

	return 0;
error:
2310 2311
	/* remove any maps that we added */
	for_each_possible_cpu(cpu) {
2312 2313 2314 2315 2316 2317 2318 2319
		for (i = num_tc, tci = cpu * num_tc; i--; tci++) {
			new_map = xmap_dereference(new_dev_maps->cpu_map[tci]);
			map = dev_maps ?
			      xmap_dereference(dev_maps->cpu_map[tci]) :
			      NULL;
			if (new_map && new_map != map)
				kfree(new_map);
		}
2320 2321
	}

2322 2323 2324 2325 2326 2327 2328 2329
	mutex_unlock(&xps_map_mutex);

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

#endif
2330 2331
void netdev_reset_tc(struct net_device *dev)
{
2332 2333 2334
#ifdef CONFIG_XPS
	netif_reset_xps_queues_gt(dev, 0);
#endif
2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345
	dev->num_tc = 0;
	memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
	memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
}
EXPORT_SYMBOL(netdev_reset_tc);

int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
{
	if (tc >= dev->num_tc)
		return -EINVAL;

2346 2347 2348
#ifdef CONFIG_XPS
	netif_reset_xps_queues(dev, offset, count);
#endif
2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359
	dev->tc_to_txq[tc].count = count;
	dev->tc_to_txq[tc].offset = offset;
	return 0;
}
EXPORT_SYMBOL(netdev_set_tc_queue);

int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
{
	if (num_tc > TC_MAX_QUEUE)
		return -EINVAL;

2360 2361 2362
#ifdef CONFIG_XPS
	netif_reset_xps_queues_gt(dev, 0);
#endif
2363 2364 2365 2366 2367
	dev->num_tc = num_tc;
	return 0;
}
EXPORT_SYMBOL(netdev_set_num_tc);

2368 2369 2370 2371
/*
 * 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.
 */
2372
int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
2373
{
T
Tom Herbert 已提交
2374 2375
	int rc;

2376 2377
	if (txq < 1 || txq > dev->num_tx_queues)
		return -EINVAL;
2378

2379 2380
	if (dev->reg_state == NETREG_REGISTERED ||
	    dev->reg_state == NETREG_UNREGISTERING) {
2381 2382
		ASSERT_RTNL();

T
Tom Herbert 已提交
2383 2384
		rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
						  txq);
T
Tom Herbert 已提交
2385 2386 2387
		if (rc)
			return rc;

2388 2389 2390
		if (dev->num_tc)
			netif_setup_tc(dev, txq);

2391
		if (txq < dev->real_num_tx_queues) {
2392
			qdisc_reset_all_tx_gt(dev, txq);
2393 2394 2395 2396
#ifdef CONFIG_XPS
			netif_reset_xps_queues_gt(dev, txq);
#endif
		}
2397
	}
2398 2399 2400

	dev->real_num_tx_queues = txq;
	return 0;
2401 2402
}
EXPORT_SYMBOL(netif_set_real_num_tx_queues);
2403

2404
#ifdef CONFIG_SYSFS
2405 2406 2407 2408 2409 2410 2411
/**
 *	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
2412 2413
 *	negative error code.  If called before registration, it always
 *	succeeds.
2414 2415 2416 2417 2418
 */
int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
{
	int rc;

T
Tom Herbert 已提交
2419 2420 2421
	if (rxq < 1 || rxq > dev->num_rx_queues)
		return -EINVAL;

2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436
	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

2437 2438
/**
 * netif_get_num_default_rss_queues - default number of RSS queues
2439 2440 2441 2442
 *
 * This routine should set an upper limit on the number of RSS queues
 * used by default by multiqueue devices.
 */
2443
int netif_get_num_default_rss_queues(void)
2444
{
2445 2446
	return is_kdump_kernel() ?
		1 : min_t(int, DEFAULT_MAX_NUM_RSS_QUEUES, num_online_cpus());
2447 2448 2449
}
EXPORT_SYMBOL(netif_get_num_default_rss_queues);

2450
static void __netif_reschedule(struct Qdisc *q)
2451
{
2452 2453
	struct softnet_data *sd;
	unsigned long flags;
2454

2455
	local_irq_save(flags);
2456
	sd = this_cpu_ptr(&softnet_data);
2457 2458 2459
	q->next_sched = NULL;
	*sd->output_queue_tailp = q;
	sd->output_queue_tailp = &q->next_sched;
2460 2461 2462 2463 2464 2465 2466 2467
	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);
2468 2469 2470
}
EXPORT_SYMBOL(__netif_schedule);

2471 2472 2473 2474 2475
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)
2476
{
2477 2478 2479
	return (struct dev_kfree_skb_cb *)skb->cb;
}

2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504
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);

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);

2505
void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason)
2506
{
2507
	unsigned long flags;
2508

2509 2510 2511
	if (unlikely(!skb))
		return;

2512
	if (likely(refcount_read(&skb->users) == 1)) {
2513
		smp_rmb();
2514 2515
		refcount_set(&skb->users, 0);
	} else if (likely(!refcount_dec_and_test(&skb->users))) {
2516
		return;
2517
	}
2518 2519 2520 2521 2522 2523
	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);
2524
}
2525
EXPORT_SYMBOL(__dev_kfree_skb_irq);
2526

2527
void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason)
2528 2529
{
	if (in_irq() || irqs_disabled())
2530
		__dev_kfree_skb_irq(skb, reason);
2531 2532 2533
	else
		dev_kfree_skb(skb);
}
2534
EXPORT_SYMBOL(__dev_kfree_skb_any);
2535 2536


2537 2538 2539 2540 2541 2542
/**
 * netif_device_detach - mark device as removed
 * @dev: network device
 *
 * Mark device as removed from system and therefore no longer available.
 */
2543 2544 2545 2546
void netif_device_detach(struct net_device *dev)
{
	if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2547
		netif_tx_stop_all_queues(dev);
2548 2549 2550 2551
	}
}
EXPORT_SYMBOL(netif_device_detach);

2552 2553 2554 2555 2556 2557
/**
 * netif_device_attach - mark device as attached
 * @dev: network device
 *
 * Mark device as attached from system and restart if needed.
 */
2558 2559 2560 2561
void netif_device_attach(struct net_device *dev)
{
	if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2562
		netif_tx_wake_all_queues(dev);
2563
		__netdev_watchdog_up(dev);
2564 2565 2566 2567
	}
}
EXPORT_SYMBOL(netif_device_attach);

J
Jiri Pirko 已提交
2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587
/*
 * 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);
2588

J
Jiri Pirko 已提交
2589 2590 2591 2592 2593 2594 2595 2596
		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);

2597 2598
static void skb_warn_bad_offload(const struct sk_buff *skb)
{
W
Wei Tang 已提交
2599
	static const netdev_features_t null_features;
2600
	struct net_device *dev = skb->dev;
2601
	const char *name = "";
2602

2603 2604 2605
	if (!net_ratelimit())
		return;

2606 2607 2608 2609 2610 2611
	if (dev) {
		if (dev->dev.parent)
			name = dev_driver_string(dev->dev.parent);
		else
			name = netdev_name(dev);
	}
2612 2613
	WARN(1, "%s: caps=(%pNF, %pNF) len=%d data_len=%d gso_size=%d "
	     "gso_type=%d ip_summed=%d\n",
2614
	     name, dev ? &dev->features : &null_features,
2615
	     skb->sk ? &skb->sk->sk_route_caps : &null_features,
2616 2617 2618 2619
	     skb->len, skb->data_len, skb_shinfo(skb)->gso_size,
	     skb_shinfo(skb)->gso_type, skb->ip_summed);
}

L
Linus Torvalds 已提交
2620 2621 2622 2623
/*
 * Invalidate hardware checksum when packet is to be mangled, and
 * complete checksum manually on outgoing path.
 */
2624
int skb_checksum_help(struct sk_buff *skb)
L
Linus Torvalds 已提交
2625
{
2626
	__wsum csum;
2627
	int ret = 0, offset;
L
Linus Torvalds 已提交
2628

2629
	if (skb->ip_summed == CHECKSUM_COMPLETE)
2630 2631 2632
		goto out_set_summed;

	if (unlikely(skb_shinfo(skb)->gso_size)) {
2633 2634
		skb_warn_bad_offload(skb);
		return -EINVAL;
L
Linus Torvalds 已提交
2635 2636
	}

2637 2638 2639 2640 2641 2642 2643 2644 2645
	/* 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;
	}

2646
	offset = skb_checksum_start_offset(skb);
2647 2648 2649 2650 2651 2652 2653 2654
	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
2655 2656 2657 2658 2659
		ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
		if (ret)
			goto out;
	}

2660
	*(__sum16 *)(skb->data + offset) = csum_fold(csum) ?: CSUM_MANGLED_0;
2661
out_set_summed:
L
Linus Torvalds 已提交
2662
	skb->ip_summed = CHECKSUM_NONE;
2663
out:
L
Linus Torvalds 已提交
2664 2665
	return ret;
}
E
Eric Dumazet 已提交
2666
EXPORT_SYMBOL(skb_checksum_help);
L
Linus Torvalds 已提交
2667

2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703
int skb_crc32c_csum_help(struct sk_buff *skb)
{
	__le32 crc32c_csum;
	int ret = 0, offset, start;

	if (skb->ip_summed != CHECKSUM_PARTIAL)
		goto out;

	if (unlikely(skb_is_gso(skb)))
		goto out;

	/* Before computing a checksum, we should make sure no frag could
	 * be modified by an external entity : checksum could be wrong.
	 */
	if (unlikely(skb_has_shared_frag(skb))) {
		ret = __skb_linearize(skb);
		if (ret)
			goto out;
	}
	start = skb_checksum_start_offset(skb);
	offset = start + offsetof(struct sctphdr, checksum);
	if (WARN_ON_ONCE(offset >= skb_headlen(skb))) {
		ret = -EINVAL;
		goto out;
	}
	if (skb_cloned(skb) &&
	    !skb_clone_writable(skb, offset + sizeof(__le32))) {
		ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
		if (ret)
			goto out;
	}
	crc32c_csum = cpu_to_le32(~__skb_checksum(skb, start,
						  skb->len - start, ~(__u32)0,
						  crc32c_csum_stub));
	*(__le32 *)(skb->data + offset) = crc32c_csum;
	skb->ip_summed = CHECKSUM_NONE;
2704
	skb->csum_not_inet = 0;
2705 2706 2707 2708
out:
	return ret;
}

2709
__be16 skb_network_protocol(struct sk_buff *skb, int *depth)
2710
{
A
Al Viro 已提交
2711
	__be16 type = skb->protocol;
2712

2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723
	/* 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;
	}

2724
	return __vlan_get_protocol(skb, type, depth);
2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736
}

/**
 *	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;
2737 2738
	int vlan_depth = skb->mac_len;
	__be16 type = skb_network_protocol(skb, &vlan_depth);
2739 2740 2741 2742

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

2743
	__skb_pull(skb, vlan_depth);
2744 2745

	rcu_read_lock();
2746
	list_for_each_entry_rcu(ptype, &offload_base, list) {
2747 2748
		if (ptype->type == type && ptype->callbacks.gso_segment) {
			segs = ptype->callbacks.gso_segment(skb, features);
2749 2750 2751 2752 2753
			break;
		}
	}
	rcu_read_unlock();

2754
	__skb_push(skb, skb->data - skb_mac_header(skb));
2755

2756 2757
	return segs;
}
2758 2759 2760 2761 2762 2763 2764 2765
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)
2766 2767
		return skb->ip_summed != CHECKSUM_PARTIAL &&
		       skb->ip_summed != CHECKSUM_UNNECESSARY;
2768 2769

	return skb->ip_summed == CHECKSUM_NONE;
2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781
}

/**
 *	__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.
2782 2783
 *
 *	Segmentation preserves SKB_SGO_CB_OFFSET bytes of previous skb cb.
2784 2785 2786 2787
 */
struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
				  netdev_features_t features, bool tx_path)
{
2788 2789
	struct sk_buff *segs;

2790 2791 2792
	if (unlikely(skb_needs_check(skb, tx_path))) {
		int err;

2793
		/* We're going to init ->check field in TCP or UDP header */
2794 2795
		err = skb_cow_head(skb, 0);
		if (err < 0)
2796 2797 2798
			return ERR_PTR(err);
	}

2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811
	/* 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;
	}

2812 2813 2814
	BUILD_BUG_ON(SKB_SGO_CB_OFFSET +
		     sizeof(*SKB_GSO_CB(skb)) > sizeof(skb->cb));

2815
	SKB_GSO_CB(skb)->mac_offset = skb_headroom(skb);
2816 2817
	SKB_GSO_CB(skb)->encap_level = 0;

2818 2819 2820
	skb_reset_mac_header(skb);
	skb_reset_mac_len(skb);

2821 2822
	segs = skb_mac_gso_segment(skb, features);

2823
	if (unlikely(skb_needs_check(skb, tx_path) && !IS_ERR(segs)))
2824 2825 2826
		skb_warn_bad_offload(skb);

	return segs;
2827
}
2828
EXPORT_SYMBOL(__skb_gso_segment);
2829

2830 2831 2832 2833 2834
/* Take action when hardware reception checksum errors are detected. */
#ifdef CONFIG_BUG
void netdev_rx_csum_fault(struct net_device *dev)
{
	if (net_ratelimit()) {
2835
		pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>");
2836 2837 2838 2839 2840 2841
		dump_stack();
	}
}
EXPORT_SYMBOL(netdev_rx_csum_fault);
#endif

L
Linus Torvalds 已提交
2842 2843 2844 2845 2846
/* 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.
 */

2847
static int illegal_highdma(struct net_device *dev, struct sk_buff *skb)
L
Linus Torvalds 已提交
2848
{
2849
#ifdef CONFIG_HIGHMEM
L
Linus Torvalds 已提交
2850
	int i;
2851

2852
	if (!(dev->features & NETIF_F_HIGHDMA)) {
2853 2854
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2855

2856
			if (PageHighMem(skb_frag_page(frag)))
2857
				return 1;
2858
		}
2859
	}
L
Linus Torvalds 已提交
2860

2861 2862
	if (PCI_DMA_BUS_IS_PHYS) {
		struct device *pdev = dev->dev.parent;
L
Linus Torvalds 已提交
2863

E
Eric Dumazet 已提交
2864 2865
		if (!pdev)
			return 0;
2866
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2867 2868
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			dma_addr_t addr = page_to_phys(skb_frag_page(frag));
2869

2870 2871 2872 2873
			if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask)
				return 1;
		}
	}
2874
#endif
L
Linus Torvalds 已提交
2875 2876 2877
	return 0;
}

2878 2879 2880
/* 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
2881
#if IS_ENABLED(CONFIG_NET_MPLS_GSO)
2882 2883 2884 2885
static netdev_features_t net_mpls_features(struct sk_buff *skb,
					   netdev_features_t features,
					   __be16 type)
{
2886
	if (eth_p_mpls(type))
2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899
		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

2900
static netdev_features_t harmonize_features(struct sk_buff *skb,
2901
	netdev_features_t features)
2902
{
2903
	int tmp;
2904 2905 2906 2907
	__be16 type;

	type = skb_network_protocol(skb, &tmp);
	features = net_mpls_features(skb, features, type);
2908

2909
	if (skb->ip_summed != CHECKSUM_NONE &&
2910
	    !can_checksum_protocol(features, type)) {
2911
		features &= ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
2912
	}
2913 2914
	if (illegal_highdma(skb->dev, skb))
		features &= ~NETIF_F_SG;
2915 2916 2917 2918

	return features;
}

2919 2920 2921 2922 2923 2924 2925 2926
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);

2927 2928 2929 2930 2931 2932 2933
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);
}

2934 2935 2936 2937 2938 2939 2940 2941 2942
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;

2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953
	/* 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.
2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965
	 */
	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;
}

2966
netdev_features_t netif_skb_features(struct sk_buff *skb)
2967
{
2968
	struct net_device *dev = skb->dev;
2969
	netdev_features_t features = dev->features;
2970

2971 2972
	if (skb_is_gso(skb))
		features = gso_features_check(skb, dev, features);
2973

2974 2975 2976 2977 2978 2979 2980
	/* 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;

2981 2982 2983 2984 2985
	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);
2986

2987 2988 2989
	if (dev->netdev_ops->ndo_features_check)
		features &= dev->netdev_ops->ndo_features_check(skb, dev,
								features);
2990 2991
	else
		features &= dflt_features_check(skb, dev, features);
2992

2993
	return harmonize_features(skb, features);
2994
}
2995
EXPORT_SYMBOL(netif_skb_features);
2996

2997
static int xmit_one(struct sk_buff *skb, struct net_device *dev,
2998
		    struct netdev_queue *txq, bool more)
2999
{
3000 3001
	unsigned int len;
	int rc;
3002

3003
	if (!list_empty(&ptype_all) || !list_empty(&dev->ptype_all))
3004
		dev_queue_xmit_nit(skb, dev);
3005

3006 3007
	len = skb->len;
	trace_net_dev_start_xmit(skb, dev);
3008
	rc = netdev_start_xmit(skb, dev, txq, more);
3009
	trace_net_dev_xmit(skb, rc, dev, len);
E
Eric Dumazet 已提交
3010

3011 3012
	return rc;
}
3013

3014 3015
struct sk_buff *dev_hard_start_xmit(struct sk_buff *first, struct net_device *dev,
				    struct netdev_queue *txq, int *ret)
3016 3017 3018
{
	struct sk_buff *skb = first;
	int rc = NETDEV_TX_OK;
3019

3020 3021
	while (skb) {
		struct sk_buff *next = skb->next;
3022

3023
		skb->next = NULL;
3024
		rc = xmit_one(skb, dev, txq, next != NULL);
3025 3026 3027 3028
		if (unlikely(!dev_xmit_complete(rc))) {
			skb->next = next;
			goto out;
		}
3029

3030 3031 3032 3033
		skb = next;
		if (netif_xmit_stopped(txq) && skb) {
			rc = NETDEV_TX_BUSY;
			break;
3034
		}
3035
	}
3036

3037 3038 3039 3040
out:
	*ret = rc;
	return skb;
}
3041

E
Eric Dumazet 已提交
3042 3043
static struct sk_buff *validate_xmit_vlan(struct sk_buff *skb,
					  netdev_features_t features)
3044
{
3045
	if (skb_vlan_tag_present(skb) &&
3046 3047
	    !vlan_hw_offload_capable(features, skb->vlan_proto))
		skb = __vlan_hwaccel_push_inside(skb);
3048 3049
	return skb;
}
3050

3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061
int skb_csum_hwoffload_help(struct sk_buff *skb,
			    const netdev_features_t features)
{
	if (unlikely(skb->csum_not_inet))
		return !!(features & NETIF_F_SCTP_CRC) ? 0 :
			skb_crc32c_csum_help(skb);

	return !!(features & NETIF_F_CSUM_MASK) ? 0 : skb_checksum_help(skb);
}
EXPORT_SYMBOL(skb_csum_hwoffload_help);

3062
static struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev)
3063 3064
{
	netdev_features_t features;
3065

3066 3067 3068 3069
	features = netif_skb_features(skb);
	skb = validate_xmit_vlan(skb, features);
	if (unlikely(!skb))
		goto out_null;
3070

3071
	if (netif_needs_gso(skb, features)) {
3072 3073 3074
		struct sk_buff *segs;

		segs = skb_gso_segment(skb, features);
3075
		if (IS_ERR(segs)) {
3076
			goto out_kfree_skb;
3077 3078 3079
		} else if (segs) {
			consume_skb(skb);
			skb = segs;
3080
		}
3081 3082 3083 3084
	} else {
		if (skb_needs_linearize(skb, features) &&
		    __skb_linearize(skb))
			goto out_kfree_skb;
3085

3086 3087 3088
		if (validate_xmit_xfrm(skb, features))
			goto out_kfree_skb;

3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099
		/* 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));
3100
			if (skb_csum_hwoffload_help(skb, features))
3101
				goto out_kfree_skb;
3102
		}
3103
	}
3104

3105
	return skb;
3106

3107 3108
out_kfree_skb:
	kfree_skb(skb);
3109
out_null:
E
Eric Dumazet 已提交
3110
	atomic_long_inc(&dev->tx_dropped);
3111 3112
	return NULL;
}
3113

3114 3115 3116 3117
struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev)
{
	struct sk_buff *next, *head = NULL, *tail;

3118
	for (; skb != NULL; skb = next) {
3119 3120
		next = skb->next;
		skb->next = NULL;
3121 3122 3123 3124

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

3125
		skb = validate_xmit_skb(skb, dev);
3126 3127
		if (!skb)
			continue;
3128

3129 3130 3131 3132 3133 3134 3135 3136
		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;
3137 3138
	}
	return head;
3139
}
3140
EXPORT_SYMBOL_GPL(validate_xmit_skb_list);
3141

3142 3143 3144 3145 3146 3147 3148 3149 3150 3151
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)  {
3152
		unsigned int hdr_len;
3153
		u16 gso_segs = shinfo->gso_segs;
3154

3155 3156 3157 3158
		/* mac layer + network layer */
		hdr_len = skb_transport_header(skb) - skb_mac_header(skb);

		/* + transport layer */
3159 3160 3161 3162
		if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)))
			hdr_len += tcp_hdrlen(skb);
		else
			hdr_len += sizeof(struct udphdr);
3163 3164 3165 3166 3167 3168

		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;
3169 3170 3171
	}
}

3172 3173 3174 3175 3176
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);
3177
	struct sk_buff *to_free = NULL;
E
Eric Dumazet 已提交
3178
	bool contended;
3179 3180
	int rc;

E
Eric Dumazet 已提交
3181
	qdisc_calculate_pkt_len(skb, q);
3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196

	if (q->flags & TCQ_F_NOLOCK) {
		if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) {
			__qdisc_drop(skb, &to_free);
			rc = NET_XMIT_DROP;
		} else {
			rc = q->enqueue(skb, q, &to_free) & NET_XMIT_MASK;
			__qdisc_run(q);
		}

		if (unlikely(to_free))
			kfree_skb_list(to_free);
		return rc;
	}

3197 3198 3199
	/*
	 * Heuristic to force contended enqueues to serialize on a
	 * separate lock before trying to get qdisc main lock.
3200
	 * This permits qdisc->running owner to get the lock more
Y
Ying Xue 已提交
3201
	 * often and dequeue packets faster.
3202
	 */
E
Eric Dumazet 已提交
3203
	contended = qdisc_is_running(q);
3204 3205 3206
	if (unlikely(contended))
		spin_lock(&q->busylock);

3207 3208
	spin_lock(root_lock);
	if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) {
3209
		__qdisc_drop(skb, &to_free);
3210 3211
		rc = NET_XMIT_DROP;
	} else if ((q->flags & TCQ_F_CAN_BYPASS) && !qdisc_qlen(q) &&
3212
		   qdisc_run_begin(q)) {
3213 3214 3215 3216 3217
		/*
		 * 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.
		 */
3218 3219 3220

		qdisc_bstats_update(q, skb);

3221
		if (sch_direct_xmit(skb, q, dev, txq, root_lock, true)) {
3222 3223 3224 3225
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
3226
			__qdisc_run(q);
3227
		}
3228

3229
		qdisc_run_end(q);
3230 3231
		rc = NET_XMIT_SUCCESS;
	} else {
3232
		rc = q->enqueue(skb, q, &to_free) & NET_XMIT_MASK;
3233 3234 3235 3236 3237 3238
		if (qdisc_run_begin(q)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
			__qdisc_run(q);
3239
			qdisc_run_end(q);
3240
		}
3241 3242
	}
	spin_unlock(root_lock);
3243 3244
	if (unlikely(to_free))
		kfree_skb_list(to_free);
3245 3246
	if (unlikely(contended))
		spin_unlock(&q->busylock);
3247 3248 3249
	return rc;
}

3250
#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
3251 3252
static void skb_update_prio(struct sk_buff *skb)
{
I
Igor Maravic 已提交
3253
	struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap);
3254

3255
	if (!skb->priority && skb->sk && map) {
3256 3257
		unsigned int prioidx =
			sock_cgroup_prioidx(&skb->sk->sk_cgrp_data);
3258 3259 3260 3261

		if (prioidx < map->priomap_len)
			skb->priority = map->priomap[prioidx];
	}
3262 3263 3264 3265 3266
}
#else
#define skb_update_prio(skb)
#endif

3267 3268 3269
DEFINE_PER_CPU(int, xmit_recursion);
EXPORT_SYMBOL(xmit_recursion);

3270 3271
/**
 *	dev_loopback_xmit - loop back @skb
3272 3273
 *	@net: network namespace this loopback is happening in
 *	@sk:  sk needed to be a netfilter okfn
3274 3275
 *	@skb: buffer to transmit
 */
3276
int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288
{
	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
3289 3290 3291 3292
#ifdef CONFIG_NET_EGRESS
static struct sk_buff *
sch_handle_egress(struct sk_buff *skb, int *ret, struct net_device *dev)
{
3293
	struct mini_Qdisc *miniq = rcu_dereference_bh(dev->miniq_egress);
D
Daniel Borkmann 已提交
3294 3295
	struct tcf_result cl_res;

3296
	if (!miniq)
D
Daniel Borkmann 已提交
3297 3298
		return skb;

3299
	/* qdisc_skb_cb(skb)->pkt_len was already set by the caller. */
3300
	mini_qdisc_bstats_cpu_update(miniq, skb);
D
Daniel Borkmann 已提交
3301

3302
	switch (tcf_classify(skb, miniq->filter_list, &cl_res, false)) {
D
Daniel Borkmann 已提交
3303 3304 3305 3306 3307
	case TC_ACT_OK:
	case TC_ACT_RECLASSIFY:
		skb->tc_index = TC_H_MIN(cl_res.classid);
		break;
	case TC_ACT_SHOT:
3308
		mini_qdisc_qstats_cpu_drop(miniq);
D
Daniel Borkmann 已提交
3309
		*ret = NET_XMIT_DROP;
3310 3311
		kfree_skb(skb);
		return NULL;
D
Daniel Borkmann 已提交
3312 3313
	case TC_ACT_STOLEN:
	case TC_ACT_QUEUED:
3314
	case TC_ACT_TRAP:
D
Daniel Borkmann 已提交
3315
		*ret = NET_XMIT_SUCCESS;
3316
		consume_skb(skb);
D
Daniel Borkmann 已提交
3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330
		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 */

3331 3332 3333 3334 3335 3336 3337 3338 3339 3340
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) {
3341 3342 3343 3344 3345 3346 3347 3348
		unsigned int tci = skb->sender_cpu - 1;

		if (dev->num_tc) {
			tci *= dev->num_tc;
			tci += netdev_get_prio_tc_map(dev, skb->priority);
		}

		map = rcu_dereference(dev_maps->cpu_map[tci]);
3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374
		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);
3375

3376 3377 3378 3379
		if (new_index < 0)
			new_index = skb_tx_hash(dev, skb);

		if (queue_index != new_index && sk &&
3380
		    sk_fullsock(sk) &&
3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396
		    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
3397 3398 3399
	u32 sender_cpu = skb->sender_cpu - 1;

	if (sender_cpu >= (u32)NR_CPUS)
3400 3401 3402 3403 3404
		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;
3405

3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419
		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);
}

3420
/**
3421
 *	__dev_queue_xmit - transmit a buffer
3422
 *	@skb: buffer to transmit
3423
 *	@accel_priv: private data used for L2 forwarding offload
3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445
 *
 *	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
 */
3446
static int __dev_queue_xmit(struct sk_buff *skb, void *accel_priv)
L
Linus Torvalds 已提交
3447 3448
{
	struct net_device *dev = skb->dev;
3449
	struct netdev_queue *txq;
L
Linus Torvalds 已提交
3450 3451 3452
	struct Qdisc *q;
	int rc = -ENOMEM;

3453 3454
	skb_reset_mac_header(skb);

3455 3456 3457
	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_SCHED_TSTAMP))
		__skb_tstamp_tx(skb, NULL, skb->sk, SCM_TSTAMP_SCHED);

3458 3459
	/* Disable soft irqs for various locks below. Also
	 * stops preemption for RCU.
L
Linus Torvalds 已提交
3460
	 */
3461
	rcu_read_lock_bh();
L
Linus Torvalds 已提交
3462

3463 3464
	skb_update_prio(skb);

D
Daniel Borkmann 已提交
3465 3466
	qdisc_pkt_len_init(skb);
#ifdef CONFIG_NET_CLS_ACT
3467
	skb->tc_at_ingress = 0;
D
Daniel Borkmann 已提交
3468 3469 3470 3471 3472 3473 3474 3475
# ifdef CONFIG_NET_EGRESS
	if (static_key_false(&egress_needed)) {
		skb = sch_handle_egress(skb, &rc, dev);
		if (!skb)
			goto out;
	}
# endif
#endif
3476 3477 3478 3479 3480 3481 3482 3483
	/* 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);

3484
	txq = netdev_pick_tx(dev, skb, accel_priv);
3485
	q = rcu_dereference_bh(txq->qdisc);
3486

3487
	trace_net_dev_queue(skb);
L
Linus Torvalds 已提交
3488
	if (q->enqueue) {
3489
		rc = __dev_xmit_skb(skb, q, dev, txq);
3490
		goto out;
L
Linus Torvalds 已提交
3491 3492 3493
	}

	/* The device has no queue. Common case for software devices:
3494
	 * loopback, all the sorts of tunnels...
L
Linus Torvalds 已提交
3495

3496 3497 3498 3499 3500
	 * Really, it is unlikely that netif_tx_lock protection is necessary
	 * here.  (f.e. loopback and IP tunnels are clean ignoring statistics
	 * counters.)
	 * However, it is possible, that they rely on protection
	 * made by us here.
L
Linus Torvalds 已提交
3501

3502 3503
	 * Check this and shot the lock. It is not prone from deadlocks.
	 *Either shot noqueue qdisc, it is even simpler 8)
L
Linus Torvalds 已提交
3504 3505 3506 3507
	 */
	if (dev->flags & IFF_UP) {
		int cpu = smp_processor_id(); /* ok because BHs are off */

3508
		if (txq->xmit_lock_owner != cpu) {
3509 3510
			if (unlikely(__this_cpu_read(xmit_recursion) >
				     XMIT_RECURSION_LIMIT))
3511 3512
				goto recursion_alert;

3513 3514
			skb = validate_xmit_skb(skb, dev);
			if (!skb)
E
Eric Dumazet 已提交
3515
				goto out;
3516

3517
			HARD_TX_LOCK(dev, txq, cpu);
L
Linus Torvalds 已提交
3518

3519
			if (!netif_xmit_stopped(txq)) {
3520
				__this_cpu_inc(xmit_recursion);
3521
				skb = dev_hard_start_xmit(skb, dev, txq, &rc);
3522
				__this_cpu_dec(xmit_recursion);
3523
				if (dev_xmit_complete(rc)) {
3524
					HARD_TX_UNLOCK(dev, txq);
L
Linus Torvalds 已提交
3525 3526 3527
					goto out;
				}
			}
3528
			HARD_TX_UNLOCK(dev, txq);
3529 3530
			net_crit_ratelimited("Virtual device %s asks to queue packet!\n",
					     dev->name);
L
Linus Torvalds 已提交
3531 3532
		} else {
			/* Recursion is detected! It is possible,
3533 3534 3535
			 * unfortunately
			 */
recursion_alert:
3536 3537
			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
					     dev->name);
L
Linus Torvalds 已提交
3538 3539 3540 3541
		}
	}

	rc = -ENETDOWN;
3542
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
3543

3544
	atomic_long_inc(&dev->tx_dropped);
3545
	kfree_skb_list(skb);
L
Linus Torvalds 已提交
3546 3547
	return rc;
out:
3548
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
3549 3550
	return rc;
}
3551

3552
int dev_queue_xmit(struct sk_buff *skb)
3553 3554 3555
{
	return __dev_queue_xmit(skb, NULL);
}
3556
EXPORT_SYMBOL(dev_queue_xmit);
L
Linus Torvalds 已提交
3557

3558 3559 3560 3561 3562 3563
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
3564

3565 3566 3567
/*************************************************************************
 *			Receiver routines
 *************************************************************************/
L
Linus Torvalds 已提交
3568

3569
int netdev_max_backlog __read_mostly = 1000;
E
Eric Dumazet 已提交
3570 3571
EXPORT_SYMBOL(netdev_max_backlog);

E
Eric Dumazet 已提交
3572
int netdev_tstamp_prequeue __read_mostly = 1;
3573
int netdev_budget __read_mostly = 300;
3574
unsigned int __read_mostly netdev_budget_usecs = 2000;
3575 3576 3577 3578 3579
int weight_p __read_mostly = 64;           /* old backlog weight */
int dev_weight_rx_bias __read_mostly = 1;  /* bias for backlog weight */
int dev_weight_tx_bias __read_mostly = 1;  /* bias for output_queue quota */
int dev_rx_weight __read_mostly = 64;
int dev_tx_weight __read_mostly = 64;
L
Linus Torvalds 已提交
3580

E
Eric Dumazet 已提交
3581 3582 3583 3584 3585 3586 3587 3588
/* 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);
}

3589 3590 3591
#ifdef CONFIG_RPS

/* One global table that all flow-based protocols share. */
E
Eric Dumazet 已提交
3592
struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
3593
EXPORT_SYMBOL(rps_sock_flow_table);
3594 3595
u32 rps_cpu_mask __read_mostly;
EXPORT_SYMBOL(rps_cpu_mask);
3596

3597
struct static_key rps_needed __read_mostly;
3598
EXPORT_SYMBOL(rps_needed);
E
Eric Dumazet 已提交
3599 3600
struct static_key rfs_needed __read_mostly;
EXPORT_SYMBOL(rfs_needed);
3601

3602 3603 3604 3605
static struct rps_dev_flow *
set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
	    struct rps_dev_flow *rflow, u16 next_cpu)
{
3606
	if (next_cpu < nr_cpu_ids) {
3607 3608 3609 3610 3611 3612 3613 3614 3615
#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? */
3616 3617
		if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
		    !(dev->features & NETIF_F_NTUPLE))
3618 3619 3620 3621 3622 3623 3624 3625 3626
			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;
3627
		flow_id = skb_get_hash(skb) & flow_table->mask;
3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639
		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 =
3640
			per_cpu(softnet_data, next_cpu).input_queue_head;
3641 3642
	}

3643
	rflow->cpu = next_cpu;
3644 3645 3646
	return rflow;
}

3647 3648 3649 3650 3651 3652 3653 3654
/*
 * 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)
{
3655 3656
	const struct rps_sock_flow_table *sock_flow_table;
	struct netdev_rx_queue *rxqueue = dev->_rx;
3657
	struct rps_dev_flow_table *flow_table;
3658
	struct rps_map *map;
3659
	int cpu = -1;
3660
	u32 tcpu;
3661
	u32 hash;
3662 3663 3664

	if (skb_rx_queue_recorded(skb)) {
		u16 index = skb_get_rx_queue(skb);
3665

3666 3667 3668 3669 3670
		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);
3671 3672
			goto done;
		}
3673 3674
		rxqueue += index;
	}
3675

3676 3677 3678
	/* 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
3679
	map = rcu_dereference(rxqueue->rps_map);
3680
	if (!flow_table && !map)
3681 3682
		goto done;

3683
	skb_reset_network_header(skb);
3684 3685
	hash = skb_get_hash(skb);
	if (!hash)
3686 3687
		goto done;

T
Tom Herbert 已提交
3688 3689 3690
	sock_flow_table = rcu_dereference(rps_sock_flow_table);
	if (flow_table && sock_flow_table) {
		struct rps_dev_flow *rflow;
3691 3692 3693 3694 3695 3696 3697
		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
3698

3699 3700 3701 3702 3703
		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
		 */
3704
		rflow = &flow_table->flows[hash & flow_table->mask];
T
Tom Herbert 已提交
3705 3706 3707 3708 3709 3710
		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:
3711
		 *   - Current CPU is unset (>= nr_cpu_ids).
T
Tom Herbert 已提交
3712 3713 3714 3715 3716 3717 3718
		 *   - 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) &&
3719
		    (tcpu >= nr_cpu_ids || !cpu_online(tcpu) ||
T
Tom Herbert 已提交
3720
		     ((int)(per_cpu(softnet_data, tcpu).input_queue_head -
3721 3722
		      rflow->last_qtail)) >= 0)) {
			tcpu = next_cpu;
3723
			rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
3724
		}
3725

3726
		if (tcpu < nr_cpu_ids && cpu_online(tcpu)) {
T
Tom Herbert 已提交
3727 3728 3729 3730 3731 3732
			*rflowp = rflow;
			cpu = tcpu;
			goto done;
		}
	}

3733 3734
try_rps:

T
Tom Herbert 已提交
3735
	if (map) {
3736
		tcpu = map->cpus[reciprocal_scale(hash, map->len)];
T
Tom Herbert 已提交
3737 3738 3739 3740 3741 3742 3743 3744 3745 3746
		if (cpu_online(tcpu)) {
			cpu = tcpu;
			goto done;
		}
	}

done:
	return cpu;
}

3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766
#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;
3767
	unsigned int cpu;
3768 3769 3770 3771 3772

	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];
3773
		cpu = READ_ONCE(rflow->cpu);
3774
		if (rflow->filter == filter_id && cpu < nr_cpu_ids &&
3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786
		    ((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
3787
/* Called from hardirq (IPI) context */
E
Eric Dumazet 已提交
3788
static void rps_trigger_softirq(void *data)
T
Tom Herbert 已提交
3789
{
E
Eric Dumazet 已提交
3790 3791
	struct softnet_data *sd = data;

E
Eric Dumazet 已提交
3792
	____napi_schedule(sd, &sd->backlog);
C
Changli Gao 已提交
3793
	sd->received_rps++;
T
Tom Herbert 已提交
3794
}
E
Eric Dumazet 已提交
3795

T
Tom Herbert 已提交
3796
#endif /* CONFIG_RPS */
T
Tom Herbert 已提交
3797

E
Eric Dumazet 已提交
3798 3799 3800 3801 3802 3803 3804 3805
/*
 * 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
3806
	struct softnet_data *mysd = this_cpu_ptr(&softnet_data);
E
Eric Dumazet 已提交
3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818

	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;
}

3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832
#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;

3833
	sd = this_cpu_ptr(&softnet_data);
3834 3835 3836 3837

	rcu_read_lock();
	fl = rcu_dereference(sd->flow_limit);
	if (fl) {
3838
		new_flow = skb_get_hash(skb) & (fl->num_buckets - 1);
3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858
		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
3859 3860 3861 3862
/*
 * 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
3863 3864
static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
			      unsigned int *qtail)
T
Tom Herbert 已提交
3865
{
E
Eric Dumazet 已提交
3866
	struct softnet_data *sd;
T
Tom Herbert 已提交
3867
	unsigned long flags;
3868
	unsigned int qlen;
T
Tom Herbert 已提交
3869

E
Eric Dumazet 已提交
3870
	sd = &per_cpu(softnet_data, cpu);
T
Tom Herbert 已提交
3871 3872 3873

	local_irq_save(flags);

E
Eric Dumazet 已提交
3874
	rps_lock(sd);
3875 3876
	if (!netif_running(skb->dev))
		goto drop;
3877 3878
	qlen = skb_queue_len(&sd->input_pkt_queue);
	if (qlen <= netdev_max_backlog && !skb_flow_limit(skb, qlen)) {
3879
		if (qlen) {
T
Tom Herbert 已提交
3880
enqueue:
E
Eric Dumazet 已提交
3881
			__skb_queue_tail(&sd->input_pkt_queue, skb);
3882
			input_queue_tail_incr_save(sd, qtail);
E
Eric Dumazet 已提交
3883
			rps_unlock(sd);
3884
			local_irq_restore(flags);
T
Tom Herbert 已提交
3885 3886 3887
			return NET_RX_SUCCESS;
		}

3888 3889 3890 3891
		/* 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
3892
			if (!rps_ipi_queued(sd))
E
Eric Dumazet 已提交
3893
				____napi_schedule(sd, &sd->backlog);
T
Tom Herbert 已提交
3894 3895 3896 3897
		}
		goto enqueue;
	}

3898
drop:
C
Changli Gao 已提交
3899
	sd->dropped++;
E
Eric Dumazet 已提交
3900
	rps_unlock(sd);
T
Tom Herbert 已提交
3901 3902 3903

	local_irq_restore(flags);

3904
	atomic_long_inc(&skb->dev->rx_dropped);
T
Tom Herbert 已提交
3905 3906 3907
	kfree_skb(skb);
	return NET_RX_DROP;
}
L
Linus Torvalds 已提交
3908

3909 3910 3911
static u32 netif_receive_generic_xdp(struct sk_buff *skb,
				     struct bpf_prog *xdp_prog)
{
3912
	u32 metalen, act = XDP_DROP;
3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923
	struct xdp_buff xdp;
	void *orig_data;
	int hlen, off;
	u32 mac_len;

	/* Reinjected packets coming from act_mirred or similar should
	 * not get XDP generic processing.
	 */
	if (skb_cloned(skb))
		return XDP_PASS;

3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942
	/* XDP packets must be linear and must have sufficient headroom
	 * of XDP_PACKET_HEADROOM bytes. This is the guarantee that also
	 * native XDP provides, thus we need to do it here as well.
	 */
	if (skb_is_nonlinear(skb) ||
	    skb_headroom(skb) < XDP_PACKET_HEADROOM) {
		int hroom = XDP_PACKET_HEADROOM - skb_headroom(skb);
		int troom = skb->tail + skb->data_len - skb->end;

		/* In case we have to go down the path and also linearize,
		 * then lets do the pskb_expand_head() work just once here.
		 */
		if (pskb_expand_head(skb,
				     hroom > 0 ? ALIGN(hroom, NET_SKB_PAD) : 0,
				     troom > 0 ? troom + 128 : 0, GFP_ATOMIC))
			goto do_drop;
		if (troom > 0 && __skb_linearize(skb))
			goto do_drop;
	}
3943 3944 3945 3946 3947 3948 3949

	/* The XDP program wants to see the packet starting at the MAC
	 * header.
	 */
	mac_len = skb->data - skb_mac_header(skb);
	hlen = skb_headlen(skb) + mac_len;
	xdp.data = skb->data - mac_len;
3950
	xdp.data_meta = xdp.data;
3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961
	xdp.data_end = xdp.data + hlen;
	xdp.data_hard_start = skb->data - skb_headroom(skb);
	orig_data = xdp.data;

	act = bpf_prog_run_xdp(xdp_prog, &xdp);

	off = xdp.data - orig_data;
	if (off > 0)
		__skb_pull(skb, off);
	else if (off < 0)
		__skb_push(skb, -off);
3962
	skb->mac_header += off;
3963 3964

	switch (act) {
3965
	case XDP_REDIRECT:
3966 3967
	case XDP_TX:
		__skb_push(skb, mac_len);
3968
		break;
3969
	case XDP_PASS:
3970 3971 3972
		metalen = xdp.data - xdp.data_meta;
		if (metalen)
			skb_metadata_set(skb, metalen);
3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991
		break;
	default:
		bpf_warn_invalid_xdp_action(act);
		/* fall through */
	case XDP_ABORTED:
		trace_xdp_exception(skb->dev, xdp_prog, act);
		/* fall through */
	case XDP_DROP:
	do_drop:
		kfree_skb(skb);
		break;
	}

	return act;
}

/* When doing generic XDP we have to bypass the qdisc layer and the
 * network taps in order to match in-driver-XDP behavior.
 */
3992
void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog)
3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012
{
	struct net_device *dev = skb->dev;
	struct netdev_queue *txq;
	bool free_skb = true;
	int cpu, rc;

	txq = netdev_pick_tx(dev, skb, NULL);
	cpu = smp_processor_id();
	HARD_TX_LOCK(dev, txq, cpu);
	if (!netif_xmit_stopped(txq)) {
		rc = netdev_start_xmit(skb, dev, txq, 0);
		if (dev_xmit_complete(rc))
			free_skb = false;
	}
	HARD_TX_UNLOCK(dev, txq);
	if (free_skb) {
		trace_xdp_exception(dev, xdp_prog, XDP_TX);
		kfree_skb(skb);
	}
}
4013
EXPORT_SYMBOL_GPL(generic_xdp_tx);
4014 4015 4016

static struct static_key generic_xdp_needed __read_mostly;

4017
int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb)
4018 4019 4020
{
	if (xdp_prog) {
		u32 act = netif_receive_generic_xdp(skb, xdp_prog);
4021
		int err;
4022 4023

		if (act != XDP_PASS) {
4024 4025
			switch (act) {
			case XDP_REDIRECT:
4026 4027
				err = xdp_do_generic_redirect(skb->dev, skb,
							      xdp_prog);
4028 4029 4030 4031
				if (err)
					goto out_redir;
			/* fallthru to submit skb */
			case XDP_TX:
4032
				generic_xdp_tx(skb, xdp_prog);
4033 4034
				break;
			}
4035 4036 4037 4038
			return XDP_DROP;
		}
	}
	return XDP_PASS;
4039 4040 4041
out_redir:
	kfree_skb(skb);
	return XDP_DROP;
4042
}
4043
EXPORT_SYMBOL_GPL(do_xdp_generic);
4044

4045
static int netif_rx_internal(struct sk_buff *skb)
L
Linus Torvalds 已提交
4046
{
4047
	int ret;
L
Linus Torvalds 已提交
4048

4049
	net_timestamp_check(netdev_tstamp_prequeue, skb);
L
Linus Torvalds 已提交
4050

4051
	trace_netif_rx(skb);
4052 4053

	if (static_key_false(&generic_xdp_needed)) {
4054 4055 4056 4057 4058 4059 4060
		int ret;

		preempt_disable();
		rcu_read_lock();
		ret = do_xdp_generic(rcu_dereference(skb->dev->xdp_prog), skb);
		rcu_read_unlock();
		preempt_enable();
4061

4062 4063 4064 4065
		/* Consider XDP consuming the packet a success from
		 * the netdev point of view we do not want to count
		 * this as an error.
		 */
4066
		if (ret != XDP_PASS)
4067
			return NET_RX_SUCCESS;
4068 4069
	}

E
Eric Dumazet 已提交
4070
#ifdef CONFIG_RPS
4071
	if (static_key_false(&rps_needed)) {
T
Tom Herbert 已提交
4072
		struct rps_dev_flow voidflow, *rflow = &voidflow;
4073 4074
		int cpu;

4075
		preempt_disable();
4076
		rcu_read_lock();
T
Tom Herbert 已提交
4077 4078

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
4079 4080
		if (cpu < 0)
			cpu = smp_processor_id();
T
Tom Herbert 已提交
4081 4082 4083

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

4084
		rcu_read_unlock();
4085
		preempt_enable();
4086 4087
	} else
#endif
T
Tom Herbert 已提交
4088 4089
	{
		unsigned int qtail;
4090

T
Tom Herbert 已提交
4091 4092 4093
		ret = enqueue_to_backlog(skb, get_cpu(), &qtail);
		put_cpu();
	}
4094
	return ret;
L
Linus Torvalds 已提交
4095
}
4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117

/**
 *	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
4118
EXPORT_SYMBOL(netif_rx);
L
Linus Torvalds 已提交
4119 4120 4121 4122 4123

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

4124 4125
	trace_netif_rx_ni_entry(skb);

L
Linus Torvalds 已提交
4126
	preempt_disable();
4127
	err = netif_rx_internal(skb);
L
Linus Torvalds 已提交
4128 4129 4130 4131 4132 4133 4134 4135
	if (local_softirq_pending())
		do_softirq();
	preempt_enable();

	return err;
}
EXPORT_SYMBOL(netif_rx_ni);

4136
static __latent_entropy void net_tx_action(struct softirq_action *h)
L
Linus Torvalds 已提交
4137
{
4138
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
L
Linus Torvalds 已提交
4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149

	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;
4150

L
Linus Torvalds 已提交
4151 4152
			clist = clist->next;

4153
			WARN_ON(refcount_read(&skb->users));
4154 4155 4156 4157
			if (likely(get_kfree_skb_cb(skb)->reason == SKB_REASON_CONSUMED))
				trace_consume_skb(skb);
			else
				trace_kfree_skb(skb, net_tx_action);
4158 4159 4160 4161 4162

			if (skb->fclone != SKB_FCLONE_UNAVAILABLE)
				__kfree_skb(skb);
			else
				__kfree_skb_defer(skb);
L
Linus Torvalds 已提交
4163
		}
4164 4165

		__kfree_skb_flush();
L
Linus Torvalds 已提交
4166 4167 4168
	}

	if (sd->output_queue) {
4169
		struct Qdisc *head;
L
Linus Torvalds 已提交
4170 4171 4172 4173

		local_irq_disable();
		head = sd->output_queue;
		sd->output_queue = NULL;
4174
		sd->output_queue_tailp = &sd->output_queue;
L
Linus Torvalds 已提交
4175 4176 4177
		local_irq_enable();

		while (head) {
4178
			struct Qdisc *q = head;
4179
			spinlock_t *root_lock = NULL;
4180

L
Linus Torvalds 已提交
4181 4182
			head = head->next_sched;

4183 4184 4185 4186
			if (!(q->flags & TCQ_F_NOLOCK)) {
				root_lock = qdisc_lock(q);
				spin_lock(root_lock);
			}
4187 4188 4189 4190 4191 4192
			/* 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);
4193 4194
			if (root_lock)
				spin_unlock(root_lock);
L
Linus Torvalds 已提交
4195 4196 4197 4198
		}
	}
}

4199
#if IS_ENABLED(CONFIG_BRIDGE) && IS_ENABLED(CONFIG_ATM_LANE)
4200 4201 4202
/* This hook is defined here for ATM LANE */
int (*br_fdb_test_addr_hook)(struct net_device *dev,
			     unsigned char *addr) __read_mostly;
4203
EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
4204
#endif
L
Linus Torvalds 已提交
4205

D
Daniel Borkmann 已提交
4206 4207 4208
static inline struct sk_buff *
sch_handle_ingress(struct sk_buff *skb, struct packet_type **pt_prev, int *ret,
		   struct net_device *orig_dev)
4209
{
4210
#ifdef CONFIG_NET_CLS_ACT
4211
	struct mini_Qdisc *miniq = rcu_dereference_bh(skb->dev->miniq_ingress);
4212
	struct tcf_result cl_res;
4213

4214 4215 4216
	/* 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
4217
	 * out here.
4218
	 */
4219
	if (!miniq)
4220
		return skb;
4221

4222 4223 4224
	if (*pt_prev) {
		*ret = deliver_skb(skb, *pt_prev, orig_dev);
		*pt_prev = NULL;
L
Linus Torvalds 已提交
4225 4226
	}

4227
	qdisc_skb_cb(skb)->pkt_len = skb->len;
4228
	skb->tc_at_ingress = 1;
4229
	mini_qdisc_bstats_cpu_update(miniq, skb);
4230

4231
	switch (tcf_classify(skb, miniq->filter_list, &cl_res, false)) {
4232 4233 4234 4235 4236
	case TC_ACT_OK:
	case TC_ACT_RECLASSIFY:
		skb->tc_index = TC_H_MIN(cl_res.classid);
		break;
	case TC_ACT_SHOT:
4237
		mini_qdisc_qstats_cpu_drop(miniq);
4238 4239
		kfree_skb(skb);
		return NULL;
4240 4241
	case TC_ACT_STOLEN:
	case TC_ACT_QUEUED:
4242
	case TC_ACT_TRAP:
4243
		consume_skb(skb);
4244
		return NULL;
4245 4246 4247 4248 4249 4250 4251 4252
	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;
4253 4254
	default:
		break;
4255
	}
4256
#endif /* CONFIG_NET_CLS_ACT */
4257 4258
	return skb;
}
L
Linus Torvalds 已提交
4259

M
Mahesh Bandewar 已提交
4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275
/**
 *	netdev_is_rx_handler_busy - check if receive handler is registered
 *	@dev: device to check
 *
 *	Check if a receive handler is already registered for a given device.
 *	Return true if there one.
 *
 *	The caller must hold the rtnl_mutex.
 */
bool netdev_is_rx_handler_busy(struct net_device *dev)
{
	ASSERT_RTNL();
	return dev && rtnl_dereference(dev->rx_handler);
}
EXPORT_SYMBOL_GPL(netdev_is_rx_handler_busy);

4276 4277 4278 4279
/**
 *	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
4280
 *	@rx_handler_data: data pointer that is used by rx handler
4281
 *
4282
 *	Register a receive handler for a device. This handler will then be
4283 4284 4285 4286
 *	called from __netif_receive_skb. A negative errno code is returned
 *	on a failure.
 *
 *	The caller must hold the rtnl_mutex.
4287 4288
 *
 *	For a general description of rx_handler, see enum rx_handler_result.
4289 4290
 */
int netdev_rx_handler_register(struct net_device *dev,
J
Jiri Pirko 已提交
4291 4292
			       rx_handler_func_t *rx_handler,
			       void *rx_handler_data)
4293
{
M
Mahesh Bandewar 已提交
4294
	if (netdev_is_rx_handler_busy(dev))
4295 4296
		return -EBUSY;

4297
	/* Note: rx_handler_data must be set before rx_handler */
J
Jiri Pirko 已提交
4298
	rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
4299 4300 4301 4302 4303 4304 4305 4306 4307 4308
	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
4309
 *	Unregister a receive handler from a device.
4310 4311 4312 4313 4314 4315 4316
 *
 *	The caller must hold the rtnl_mutex.
 */
void netdev_rx_handler_unregister(struct net_device *dev)
{

	ASSERT_RTNL();
4317
	RCU_INIT_POINTER(dev->rx_handler, NULL);
4318 4319 4320 4321 4322
	/* 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();
4323
	RCU_INIT_POINTER(dev->rx_handler_data, NULL);
4324 4325 4326
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);

4327 4328 4329 4330 4331 4332 4333
/*
 * 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) {
4334 4335 4336 4337 4338
	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):
4339 4340 4341 4342 4343 4344
		return true;
	default:
		return false;
	}
}

4345 4346 4347
static inline int nf_ingress(struct sk_buff *skb, struct packet_type **pt_prev,
			     int *ret, struct net_device *orig_dev)
{
4348
#ifdef CONFIG_NETFILTER_INGRESS
4349
	if (nf_hook_ingress_active(skb)) {
4350 4351
		int ingress_retval;

4352 4353 4354 4355 4356
		if (*pt_prev) {
			*ret = deliver_skb(skb, *pt_prev, orig_dev);
			*pt_prev = NULL;
		}

4357 4358 4359 4360
		rcu_read_lock();
		ingress_retval = nf_hook_ingress(skb);
		rcu_read_unlock();
		return ingress_retval;
4361
	}
4362
#endif /* CONFIG_NETFILTER_INGRESS */
4363 4364 4365
	return 0;
}

4366
static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc)
L
Linus Torvalds 已提交
4367 4368
{
	struct packet_type *ptype, *pt_prev;
4369
	rx_handler_func_t *rx_handler;
D
David S. Miller 已提交
4370
	struct net_device *orig_dev;
4371
	bool deliver_exact = false;
L
Linus Torvalds 已提交
4372
	int ret = NET_RX_DROP;
A
Al Viro 已提交
4373
	__be16 type;
L
Linus Torvalds 已提交
4374

4375
	net_timestamp_check(!netdev_tstamp_prequeue, skb);
4376

4377
	trace_netif_receive_skb(skb);
4378

J
Joe Eykholt 已提交
4379
	orig_dev = skb->dev;
4380

4381
	skb_reset_network_header(skb);
4382 4383
	if (!skb_transport_header_was_set(skb))
		skb_reset_transport_header(skb);
4384
	skb_reset_mac_len(skb);
L
Linus Torvalds 已提交
4385 4386 4387

	pt_prev = NULL;

4388
another_round:
4389
	skb->skb_iif = skb->dev->ifindex;
4390 4391 4392

	__this_cpu_inc(softnet_data.processed);

P
Patrick McHardy 已提交
4393 4394
	if (skb->protocol == cpu_to_be16(ETH_P_8021Q) ||
	    skb->protocol == cpu_to_be16(ETH_P_8021AD)) {
4395
		skb = skb_vlan_untag(skb);
4396
		if (unlikely(!skb))
4397
			goto out;
4398 4399
	}

4400 4401
	if (skb_skip_tc_classify(skb))
		goto skip_classify;
L
Linus Torvalds 已提交
4402

4403
	if (pfmemalloc)
4404 4405
		goto skip_taps;

L
Linus Torvalds 已提交
4406
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
4407 4408 4409 4410 4411 4412 4413 4414 4415
		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
4416 4417
	}

4418
skip_taps:
4419
#ifdef CONFIG_NET_INGRESS
4420
	if (static_key_false(&ingress_needed)) {
D
Daniel Borkmann 已提交
4421
		skb = sch_handle_ingress(skb, &pt_prev, &ret, orig_dev);
4422
		if (!skb)
4423
			goto out;
4424 4425

		if (nf_ingress(skb, &pt_prev, &ret, orig_dev) < 0)
4426
			goto out;
4427
	}
4428
#endif
4429
	skb_reset_tc(skb);
4430
skip_classify:
4431
	if (pfmemalloc && !skb_pfmemalloc_protocol(skb))
4432 4433
		goto drop;

4434
	if (skb_vlan_tag_present(skb)) {
4435 4436 4437 4438
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
4439
		if (vlan_do_receive(&skb))
4440 4441
			goto another_round;
		else if (unlikely(!skb))
4442
			goto out;
4443 4444
	}

4445
	rx_handler = rcu_dereference(skb->dev->rx_handler);
4446 4447 4448 4449 4450
	if (rx_handler) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
4451 4452
		switch (rx_handler(&skb)) {
		case RX_HANDLER_CONSUMED:
4453
			ret = NET_RX_SUCCESS;
4454
			goto out;
4455
		case RX_HANDLER_ANOTHER:
4456
			goto another_round;
4457 4458 4459 4460 4461 4462 4463
		case RX_HANDLER_EXACT:
			deliver_exact = true;
		case RX_HANDLER_PASS:
			break;
		default:
			BUG();
		}
4464
	}
L
Linus Torvalds 已提交
4465

4466 4467
	if (unlikely(skb_vlan_tag_present(skb))) {
		if (skb_vlan_tag_get_id(skb))
E
Eric Dumazet 已提交
4468 4469 4470 4471 4472 4473 4474
			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;
	}
4475

4476 4477
	type = skb->protocol;

4478
	/* deliver only exact match when indicated */
4479 4480 4481 4482 4483
	if (likely(!deliver_exact)) {
		deliver_ptype_list_skb(skb, &pt_prev, orig_dev, type,
				       &ptype_base[ntohs(type) &
						   PTYPE_HASH_MASK]);
	}
4484

4485 4486 4487 4488 4489 4490
	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
4491 4492 4493
	}

	if (pt_prev) {
W
Willem de Bruijn 已提交
4494
		if (unlikely(skb_orphan_frags_rx(skb, GFP_ATOMIC)))
4495
			goto drop;
4496 4497
		else
			ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
L
Linus Torvalds 已提交
4498
	} else {
4499
drop:
4500 4501 4502 4503
		if (!deliver_exact)
			atomic_long_inc(&skb->dev->rx_dropped);
		else
			atomic_long_inc(&skb->dev->rx_nohandler);
L
Linus Torvalds 已提交
4504 4505 4506 4507 4508 4509 4510
		kfree_skb(skb);
		/* Jamal, now you will not able to escape explaining
		 * me how you were going to use this. :-)
		 */
		ret = NET_RX_DROP;
	}

4511
out:
4512 4513 4514
	return ret;
}

4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541
/**
 *	netif_receive_skb_core - special purpose version of netif_receive_skb
 *	@skb: buffer to process
 *
 *	More direct receive version of netif_receive_skb().  It should
 *	only be used by callers that have a need to skip RPS and Generic XDP.
 *	Caller must also take care of handling if (page_is_)pfmemalloc.
 *
 *	This function may only be called from softirq context and interrupts
 *	should be enabled.
 *
 *	Return values (usually ignored):
 *	NET_RX_SUCCESS: no congestion
 *	NET_RX_DROP: packet was dropped
 */
int netif_receive_skb_core(struct sk_buff *skb)
{
	int ret;

	rcu_read_lock();
	ret = __netif_receive_skb_core(skb, false);
	rcu_read_unlock();

	return ret;
}
EXPORT_SYMBOL(netif_receive_skb_core);

4542 4543 4544 4545 4546
static int __netif_receive_skb(struct sk_buff *skb)
{
	int ret;

	if (sk_memalloc_socks() && skb_pfmemalloc(skb)) {
4547
		unsigned int noreclaim_flag;
4548 4549 4550 4551 4552 4553 4554 4555 4556 4557

		/*
		 * 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.
		 */
4558
		noreclaim_flag = memalloc_noreclaim_save();
4559
		ret = __netif_receive_skb_core(skb, true);
4560
		memalloc_noreclaim_restore(noreclaim_flag);
4561 4562 4563
	} else
		ret = __netif_receive_skb_core(skb, false);

L
Linus Torvalds 已提交
4564 4565
	return ret;
}
T
Tom Herbert 已提交
4566

4567
static int generic_xdp_install(struct net_device *dev, struct netdev_bpf *xdp)
D
David S. Miller 已提交
4568
{
M
Martin KaFai Lau 已提交
4569
	struct bpf_prog *old = rtnl_dereference(dev->xdp_prog);
D
David S. Miller 已提交
4570 4571 4572 4573
	struct bpf_prog *new = xdp->prog;
	int ret = 0;

	switch (xdp->command) {
M
Martin KaFai Lau 已提交
4574
	case XDP_SETUP_PROG:
D
David S. Miller 已提交
4575 4576 4577 4578 4579 4580 4581 4582 4583
		rcu_assign_pointer(dev->xdp_prog, new);
		if (old)
			bpf_prog_put(old);

		if (old && !new) {
			static_key_slow_dec(&generic_xdp_needed);
		} else if (new && !old) {
			static_key_slow_inc(&generic_xdp_needed);
			dev_disable_lro(dev);
4584
			dev_disable_gro_hw(dev);
D
David S. Miller 已提交
4585 4586 4587 4588
		}
		break;

	case XDP_QUERY_PROG:
M
Martin KaFai Lau 已提交
4589 4590
		xdp->prog_attached = !!old;
		xdp->prog_id = old ? old->aux->id : 0;
D
David S. Miller 已提交
4591 4592 4593 4594 4595 4596 4597 4598 4599 4600
		break;

	default:
		ret = -EINVAL;
		break;
	}

	return ret;
}

4601
static int netif_receive_skb_internal(struct sk_buff *skb)
T
Tom Herbert 已提交
4602
{
4603 4604
	int ret;

4605
	net_timestamp_check(netdev_tstamp_prequeue, skb);
E
Eric Dumazet 已提交
4606

4607 4608 4609
	if (skb_defer_rx_timestamp(skb))
		return NET_RX_SUCCESS;

D
David S. Miller 已提交
4610
	if (static_key_false(&generic_xdp_needed)) {
4611
		int ret;
D
David S. Miller 已提交
4612

4613 4614 4615 4616 4617 4618 4619
		preempt_disable();
		rcu_read_lock();
		ret = do_xdp_generic(rcu_dereference(skb->dev->xdp_prog), skb);
		rcu_read_unlock();
		preempt_enable();

		if (ret != XDP_PASS)
4620
			return NET_RX_DROP;
D
David S. Miller 已提交
4621 4622
	}

4623
	rcu_read_lock();
E
Eric Dumazet 已提交
4624
#ifdef CONFIG_RPS
4625
	if (static_key_false(&rps_needed)) {
E
Eric Dumazet 已提交
4626
		struct rps_dev_flow voidflow, *rflow = &voidflow;
4627
		int cpu = get_rps_cpu(skb->dev, skb, &rflow);
T
Tom Herbert 已提交
4628

E
Eric Dumazet 已提交
4629 4630 4631
		if (cpu >= 0) {
			ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
			rcu_read_unlock();
4632
			return ret;
E
Eric Dumazet 已提交
4633
		}
T
Tom Herbert 已提交
4634
	}
4635
#endif
4636 4637 4638
	ret = __netif_receive_skb(skb);
	rcu_read_unlock();
	return ret;
T
Tom Herbert 已提交
4639
}
4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655

/**
 *	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
 */
4656
int netif_receive_skb(struct sk_buff *skb)
4657 4658 4659 4660 4661
{
	trace_netif_receive_skb_entry(skb);

	return netif_receive_skb_internal(skb);
}
4662
EXPORT_SYMBOL(netif_receive_skb);
L
Linus Torvalds 已提交
4663

4664
DEFINE_PER_CPU(struct work_struct, flush_works);
4665 4666 4667

/* Network device is going away, flush any packets still pending */
static void flush_backlog(struct work_struct *work)
4668 4669
{
	struct sk_buff *skb, *tmp;
4670 4671 4672 4673
	struct softnet_data *sd;

	local_bh_disable();
	sd = this_cpu_ptr(&softnet_data);
4674

4675
	local_irq_disable();
E
Eric Dumazet 已提交
4676
	rps_lock(sd);
4677
	skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
4678
		if (skb->dev->reg_state == NETREG_UNREGISTERING) {
E
Eric Dumazet 已提交
4679
			__skb_unlink(skb, &sd->input_pkt_queue);
4680
			kfree_skb(skb);
4681
			input_queue_head_incr(sd);
4682
		}
4683
	}
E
Eric Dumazet 已提交
4684
	rps_unlock(sd);
4685
	local_irq_enable();
4686 4687

	skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
4688
		if (skb->dev->reg_state == NETREG_UNREGISTERING) {
4689 4690
			__skb_unlink(skb, &sd->process_queue);
			kfree_skb(skb);
4691
			input_queue_head_incr(sd);
4692 4693
		}
	}
4694 4695 4696
	local_bh_enable();
}

4697
static void flush_all_backlogs(void)
4698 4699 4700 4701 4702
{
	unsigned int cpu;

	get_online_cpus();

4703 4704 4705
	for_each_online_cpu(cpu)
		queue_work_on(cpu, system_highpri_wq,
			      per_cpu_ptr(&flush_works, cpu));
4706 4707

	for_each_online_cpu(cpu)
4708
		flush_work(per_cpu_ptr(&flush_works, cpu));
4709 4710

	put_online_cpus();
4711 4712
}

4713 4714
static int napi_gro_complete(struct sk_buff *skb)
{
4715
	struct packet_offload *ptype;
4716
	__be16 type = skb->protocol;
4717
	struct list_head *head = &offload_base;
4718 4719
	int err = -ENOENT;

4720 4721
	BUILD_BUG_ON(sizeof(struct napi_gro_cb) > sizeof(skb->cb));

4722 4723
	if (NAPI_GRO_CB(skb)->count == 1) {
		skb_shinfo(skb)->gso_size = 0;
4724
		goto out;
4725
	}
4726 4727 4728

	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
4729
		if (ptype->type != type || !ptype->callbacks.gro_complete)
4730 4731
			continue;

4732
		err = ptype->callbacks.gro_complete(skb, 0);
4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743
		break;
	}
	rcu_read_unlock();

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

out:
4744
	return netif_receive_skb_internal(skb);
4745 4746
}

4747 4748 4749 4750 4751
/* 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)
4752
{
4753
	struct sk_buff *skb, *prev = NULL;
4754

4755 4756 4757 4758 4759 4760 4761
	/* 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) {
4762
		skb->next = NULL;
4763 4764 4765 4766 4767

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

		prev = skb->prev;
4768
		napi_gro_complete(skb);
4769
		napi->gro_count--;
4770 4771 4772 4773
	}

	napi->gro_list = NULL;
}
E
Eric Dumazet 已提交
4774
EXPORT_SYMBOL(napi_gro_flush);
4775

E
Eric Dumazet 已提交
4776 4777 4778 4779
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;
4780
	u32 hash = skb_get_hash_raw(skb);
E
Eric Dumazet 已提交
4781 4782 4783 4784

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

4785 4786 4787 4788 4789 4790 4791
		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
4792 4793
		diffs = (unsigned long)p->dev ^ (unsigned long)skb->dev;
		diffs |= p->vlan_tci ^ skb->vlan_tci;
4794
		diffs |= skb_metadata_dst_cmp(p, skb);
4795
		diffs |= skb_metadata_differs(p, skb);
E
Eric Dumazet 已提交
4796 4797
		if (maclen == ETH_HLEN)
			diffs |= compare_ether_header(skb_mac_header(p),
4798
						      skb_mac_header(skb));
E
Eric Dumazet 已提交
4799 4800
		else if (!diffs)
			diffs = memcmp(skb_mac_header(p),
4801
				       skb_mac_header(skb),
E
Eric Dumazet 已提交
4802 4803 4804 4805 4806
				       maclen);
		NAPI_GRO_CB(p)->same_flow = !diffs;
	}
}

4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819
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);
4820 4821 4822
		NAPI_GRO_CB(skb)->frag0_len = min_t(unsigned int,
						    skb_frag_size(frag0),
						    skb->end - skb->tail);
E
Eric Dumazet 已提交
4823 4824 4825
	}
}

4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846
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
4847
static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
4848 4849
{
	struct sk_buff **pp = NULL;
4850
	struct packet_offload *ptype;
4851
	__be16 type = skb->protocol;
4852
	struct list_head *head = &offload_base;
H
Herbert Xu 已提交
4853
	int same_flow;
4854
	enum gro_result ret;
4855
	int grow;
4856

D
David S. Miller 已提交
4857
	if (netif_elide_gro(skb->dev))
4858 4859
		goto normal;

E
Eric Dumazet 已提交
4860 4861
	gro_list_prepare(napi, skb);

4862 4863
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
4864
		if (ptype->type != type || !ptype->callbacks.gro_receive)
4865 4866
			continue;

4867
		skb_set_network_header(skb, skb_gro_offset(skb));
4868
		skb_reset_mac_len(skb);
4869
		NAPI_GRO_CB(skb)->same_flow = 0;
E
Eric Dumazet 已提交
4870
		NAPI_GRO_CB(skb)->flush = skb_is_gso(skb) || skb_has_frag_list(skb);
H
Herbert Xu 已提交
4871
		NAPI_GRO_CB(skb)->free = 0;
4872
		NAPI_GRO_CB(skb)->encap_mark = 0;
S
Sabrina Dubroca 已提交
4873
		NAPI_GRO_CB(skb)->recursion_counter = 0;
4874
		NAPI_GRO_CB(skb)->is_fou = 0;
4875
		NAPI_GRO_CB(skb)->is_atomic = 1;
4876
		NAPI_GRO_CB(skb)->gro_remcsum_start = 0;
4877

4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892
		/* 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;
		}
4893

4894
		pp = ptype->callbacks.gro_receive(&napi->gro_list, skb);
4895 4896 4897 4898 4899 4900 4901
		break;
	}
	rcu_read_unlock();

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

4902 4903 4904 4905 4906
	if (IS_ERR(pp) && PTR_ERR(pp) == -EINPROGRESS) {
		ret = GRO_CONSUMED;
		goto ok;
	}

H
Herbert Xu 已提交
4907
	same_flow = NAPI_GRO_CB(skb)->same_flow;
4908
	ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED;
H
Herbert Xu 已提交
4909

4910 4911 4912 4913 4914 4915
	if (pp) {
		struct sk_buff *nskb = *pp;

		*pp = nskb->next;
		nskb->next = NULL;
		napi_gro_complete(nskb);
4916
		napi->gro_count--;
4917 4918
	}

H
Herbert Xu 已提交
4919
	if (same_flow)
4920 4921
		goto ok;

4922
	if (NAPI_GRO_CB(skb)->flush)
4923 4924
		goto normal;

4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938
	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++;
	}
4939
	NAPI_GRO_CB(skb)->count = 1;
4940
	NAPI_GRO_CB(skb)->age = jiffies;
4941
	NAPI_GRO_CB(skb)->last = skb;
4942
	skb_shinfo(skb)->gso_size = skb_gro_len(skb);
4943 4944
	skb->next = napi->gro_list;
	napi->gro_list = skb;
4945
	ret = GRO_HELD;
4946

4947
pull:
4948 4949 4950
	grow = skb_gro_offset(skb) - skb_headlen(skb);
	if (grow > 0)
		gro_pull_from_frag0(skb, grow);
4951
ok:
4952
	return ret;
4953 4954

normal:
4955 4956
	ret = GRO_NORMAL;
	goto pull;
H
Herbert Xu 已提交
4957
}
4958

4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970
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;
}
4971
EXPORT_SYMBOL(gro_find_receive_by_type);
4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984

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;
}
4985
EXPORT_SYMBOL(gro_find_complete_by_type);
H
Herbert Xu 已提交
4986

4987 4988 4989 4990 4991 4992 4993
static void napi_skb_free_stolen_head(struct sk_buff *skb)
{
	skb_dst_drop(skb);
	secpath_reset(skb);
	kmem_cache_free(skbuff_head_cache, skb);
}

R
Rami Rosen 已提交
4994
static gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
H
Herbert Xu 已提交
4995
{
4996 4997
	switch (ret) {
	case GRO_NORMAL:
4998
		if (netif_receive_skb_internal(skb))
4999 5000
			ret = GRO_DROP;
		break;
H
Herbert Xu 已提交
5001

5002
	case GRO_DROP:
H
Herbert Xu 已提交
5003 5004
		kfree_skb(skb);
		break;
5005

5006
	case GRO_MERGED_FREE:
5007 5008 5009
		if (NAPI_GRO_CB(skb)->free == NAPI_GRO_FREE_STOLEN_HEAD)
			napi_skb_free_stolen_head(skb);
		else
5010
			__kfree_skb(skb);
5011 5012
		break;

5013 5014
	case GRO_HELD:
	case GRO_MERGED:
5015
	case GRO_CONSUMED:
5016
		break;
H
Herbert Xu 已提交
5017 5018
	}

5019
	return ret;
5020 5021
}

5022
gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
5023
{
5024
	skb_mark_napi_id(skb, napi);
5025
	trace_napi_gro_receive_entry(skb);
5026

5027 5028
	skb_gro_reset_offset(skb);

E
Eric Dumazet 已提交
5029
	return napi_skb_finish(dev_gro_receive(napi, skb), skb);
5030 5031 5032
}
EXPORT_SYMBOL(napi_gro_receive);

S
stephen hemminger 已提交
5033
static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb)
5034
{
5035 5036 5037 5038
	if (unlikely(skb->pfmemalloc)) {
		consume_skb(skb);
		return;
	}
5039
	__skb_pull(skb, skb_headlen(skb));
5040 5041
	/* restore the reserve we had after netdev_alloc_skb_ip_align() */
	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN - skb_headroom(skb));
5042
	skb->vlan_tci = 0;
H
Herbert Xu 已提交
5043
	skb->dev = napi->dev;
A
Andy Gospodarek 已提交
5044
	skb->skb_iif = 0;
5045 5046
	skb->encapsulation = 0;
	skb_shinfo(skb)->gso_type = 0;
5047
	skb->truesize = SKB_TRUESIZE(skb_end_offset(skb));
5048
	secpath_reset(skb);
5049 5050 5051 5052

	napi->skb = skb;
}

5053
struct sk_buff *napi_get_frags(struct napi_struct *napi)
H
Herbert Xu 已提交
5054 5055 5056 5057
{
	struct sk_buff *skb = napi->skb;

	if (!skb) {
5058
		skb = napi_alloc_skb(napi, GRO_MAX_HEAD);
5059 5060 5061 5062
		if (skb) {
			napi->skb = skb;
			skb_mark_napi_id(skb, napi);
		}
5063
	}
5064 5065
	return skb;
}
5066
EXPORT_SYMBOL(napi_get_frags);
5067

5068 5069 5070
static gro_result_t napi_frags_finish(struct napi_struct *napi,
				      struct sk_buff *skb,
				      gro_result_t ret)
5071
{
5072 5073
	switch (ret) {
	case GRO_NORMAL:
5074 5075 5076 5077
	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))
5078
			ret = GRO_DROP;
5079
		break;
H
Herbert Xu 已提交
5080

5081 5082 5083
	case GRO_DROP:
		napi_reuse_skb(napi, skb);
		break;
5084

5085 5086 5087 5088 5089 5090 5091
	case GRO_MERGED_FREE:
		if (NAPI_GRO_CB(skb)->free == NAPI_GRO_FREE_STOLEN_HEAD)
			napi_skb_free_stolen_head(skb);
		else
			napi_reuse_skb(napi, skb);
		break;

5092
	case GRO_MERGED:
5093
	case GRO_CONSUMED:
5094
		break;
5095
	}
H
Herbert Xu 已提交
5096

5097
	return ret;
H
Herbert Xu 已提交
5098
}
5099

5100 5101 5102 5103
/* 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
5104
static struct sk_buff *napi_frags_skb(struct napi_struct *napi)
5105 5106
{
	struct sk_buff *skb = napi->skb;
5107 5108
	const struct ethhdr *eth;
	unsigned int hlen = sizeof(*eth);
5109 5110 5111

	napi->skb = NULL;

5112 5113 5114 5115 5116 5117 5118
	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)) {
5119 5120
			net_warn_ratelimited("%s: dropping impossible skb from %s\n",
					     __func__, napi->dev->name);
5121 5122 5123 5124 5125 5126 5127
			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;
5128
	}
5129 5130 5131 5132 5133 5134 5135 5136
	__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;
5137 5138 5139 5140

	return skb;
}

5141
gro_result_t napi_gro_frags(struct napi_struct *napi)
5142
{
5143
	struct sk_buff *skb = napi_frags_skb(napi);
5144 5145

	if (!skb)
5146
		return GRO_DROP;
5147

5148 5149
	trace_napi_gro_frags_entry(skb);

E
Eric Dumazet 已提交
5150
	return napi_frags_finish(napi, skb, dev_gro_receive(napi, skb));
5151
}
H
Herbert Xu 已提交
5152 5153
EXPORT_SYMBOL(napi_gro_frags);

T
Tom Herbert 已提交
5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178
/* 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);

5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191
static void net_rps_send_ipi(struct softnet_data *remsd)
{
#ifdef CONFIG_RPS
	while (remsd) {
		struct softnet_data *next = remsd->rps_ipi_next;

		if (cpu_online(remsd->cpu))
			smp_call_function_single_async(remsd->cpu, &remsd->csd);
		remsd = next;
	}
#endif
}

5192
/*
5193
 * net_rps_action_and_irq_enable sends any pending IPI's for rps.
5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206
 * 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. */
5207
		net_rps_send_ipi(remsd);
5208 5209 5210 5211 5212
	} else
#endif
		local_irq_enable();
}

E
Eric Dumazet 已提交
5213 5214 5215 5216 5217 5218 5219 5220 5221
static bool sd_has_rps_ipi_waiting(struct softnet_data *sd)
{
#ifdef CONFIG_RPS
	return sd->rps_ipi_list != NULL;
#else
	return false;
#endif
}

5222
static int process_backlog(struct napi_struct *napi, int quota)
L
Linus Torvalds 已提交
5223
{
E
Eric Dumazet 已提交
5224
	struct softnet_data *sd = container_of(napi, struct softnet_data, backlog);
5225 5226
	bool again = true;
	int work = 0;
L
Linus Torvalds 已提交
5227

5228 5229 5230
	/* Check if we have pending ipi, its better to send them now,
	 * not waiting net_rx_action() end.
	 */
E
Eric Dumazet 已提交
5231
	if (sd_has_rps_ipi_waiting(sd)) {
5232 5233 5234
		local_irq_disable();
		net_rps_action_and_irq_enable(sd);
	}
E
Eric Dumazet 已提交
5235

5236
	napi->weight = dev_rx_weight;
5237
	while (again) {
L
Linus Torvalds 已提交
5238
		struct sk_buff *skb;
5239 5240

		while ((skb = __skb_dequeue(&sd->process_queue))) {
5241
			rcu_read_lock();
5242
			__netif_receive_skb(skb);
5243
			rcu_read_unlock();
5244
			input_queue_head_incr(sd);
5245
			if (++work >= quota)
5246
				return work;
5247

5248
		}
L
Linus Torvalds 已提交
5249

5250
		local_irq_disable();
E
Eric Dumazet 已提交
5251
		rps_lock(sd);
5252
		if (skb_queue_empty(&sd->input_pkt_queue)) {
E
Eric Dumazet 已提交
5253 5254 5255
			/*
			 * Inline a custom version of __napi_complete().
			 * only current cpu owns and manipulates this napi,
5256 5257 5258
			 * 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
5259 5260 5261
			 * and we dont need an smp_mb() memory barrier.
			 */
			napi->state = 0;
5262 5263 5264 5265
			again = false;
		} else {
			skb_queue_splice_tail_init(&sd->input_pkt_queue,
						   &sd->process_queue);
5266
		}
E
Eric Dumazet 已提交
5267
		rps_unlock(sd);
5268
		local_irq_enable();
5269
	}
L
Linus Torvalds 已提交
5270

5271 5272
	return work;
}
L
Linus Torvalds 已提交
5273

5274 5275
/**
 * __napi_schedule - schedule for receive
5276
 * @n: entry to schedule
5277
 *
5278 5279
 * The entry's receive function will be scheduled to run.
 * Consider using __napi_schedule_irqoff() if hard irqs are masked.
5280
 */
H
Harvey Harrison 已提交
5281
void __napi_schedule(struct napi_struct *n)
5282 5283
{
	unsigned long flags;
L
Linus Torvalds 已提交
5284

5285
	local_irq_save(flags);
5286
	____napi_schedule(this_cpu_ptr(&softnet_data), n);
5287
	local_irq_restore(flags);
L
Linus Torvalds 已提交
5288
}
5289 5290
EXPORT_SYMBOL(__napi_schedule);

E
Eric Dumazet 已提交
5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318 5319 5320 5321 5322 5323
/**
 *	napi_schedule_prep - check if napi can be scheduled
 *	@n: napi context
 *
 * Test if NAPI routine is already running, and if not mark
 * it as running.  This is used as a condition variable
 * insure only one NAPI poll instance runs.  We also make
 * sure there is no pending NAPI disable.
 */
bool napi_schedule_prep(struct napi_struct *n)
{
	unsigned long val, new;

	do {
		val = READ_ONCE(n->state);
		if (unlikely(val & NAPIF_STATE_DISABLE))
			return false;
		new = val | NAPIF_STATE_SCHED;

		/* Sets STATE_MISSED bit if STATE_SCHED was already set
		 * This was suggested by Alexander Duyck, as compiler
		 * emits better code than :
		 * if (val & NAPIF_STATE_SCHED)
		 *     new |= NAPIF_STATE_MISSED;
		 */
		new |= (val & NAPIF_STATE_SCHED) / NAPIF_STATE_SCHED *
						   NAPIF_STATE_MISSED;
	} while (cmpxchg(&n->state, val, new) != val);

	return !(val & NAPIF_STATE_SCHED);
}
EXPORT_SYMBOL(napi_schedule_prep);

5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335
/**
 * __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);

5336
bool napi_complete_done(struct napi_struct *n, int work_done)
5337
{
E
Eric Dumazet 已提交
5338
	unsigned long flags, val, new;
5339 5340

	/*
5341 5342 5343 5344
	 * 1) Don't let napi dequeue from the cpu poll list
	 *    just in case its running on a different cpu.
	 * 2) If we are busy polling, do nothing here, we have
	 *    the guarantee we will be called later.
5345
	 */
5346 5347
	if (unlikely(n->state & (NAPIF_STATE_NPSVC |
				 NAPIF_STATE_IN_BUSY_POLL)))
5348
		return false;
5349

5350 5351
	if (n->gro_list) {
		unsigned long timeout = 0;
E
Eric Dumazet 已提交
5352

5353 5354 5355 5356 5357 5358 5359 5360 5361
		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
5362
	if (unlikely(!list_empty(&n->poll_list))) {
E
Eric Dumazet 已提交
5363 5364
		/* If n->poll_list is not empty, we need to mask irqs */
		local_irq_save(flags);
E
Eric Dumazet 已提交
5365
		list_del_init(&n->poll_list);
E
Eric Dumazet 已提交
5366 5367
		local_irq_restore(flags);
	}
E
Eric Dumazet 已提交
5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388

	do {
		val = READ_ONCE(n->state);

		WARN_ON_ONCE(!(val & NAPIF_STATE_SCHED));

		new = val & ~(NAPIF_STATE_MISSED | NAPIF_STATE_SCHED);

		/* If STATE_MISSED was set, leave STATE_SCHED set,
		 * because we will call napi->poll() one more time.
		 * This C code was suggested by Alexander Duyck to help gcc.
		 */
		new |= (val & NAPIF_STATE_MISSED) / NAPIF_STATE_MISSED *
						    NAPIF_STATE_SCHED;
	} while (cmpxchg(&n->state, val, new) != val);

	if (unlikely(val & NAPIF_STATE_MISSED)) {
		__napi_schedule(n);
		return false;
	}

5389
	return true;
5390
}
5391
EXPORT_SYMBOL(napi_complete_done);
5392

E
Eliezer Tamir 已提交
5393
/* must be called under rcu_read_lock(), as we dont take a reference */
E
Eric Dumazet 已提交
5394
static struct napi_struct *napi_by_id(unsigned int napi_id)
E
Eliezer Tamir 已提交
5395 5396 5397 5398 5399 5400 5401 5402 5403 5404
{
	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
5405 5406

#if defined(CONFIG_NET_RX_BUSY_POLL)
5407

5408
#define BUSY_POLL_BUDGET 8
5409 5410 5411 5412 5413

static void busy_poll_stop(struct napi_struct *napi, void *have_poll_lock)
{
	int rc;

E
Eric Dumazet 已提交
5414 5415 5416 5417 5418 5419 5420 5421 5422 5423
	/* Busy polling means there is a high chance device driver hard irq
	 * could not grab NAPI_STATE_SCHED, and that NAPI_STATE_MISSED was
	 * set in napi_schedule_prep().
	 * Since we are about to call napi->poll() once more, we can safely
	 * clear NAPI_STATE_MISSED.
	 *
	 * Note: x86 could use a single "lock and ..." instruction
	 * to perform these two clear_bit()
	 */
	clear_bit(NAPI_STATE_MISSED, &napi->state);
5424 5425 5426 5427 5428 5429 5430 5431
	clear_bit(NAPI_STATE_IN_BUSY_POLL, &napi->state);

	local_bh_disable();

	/* All we really want here is to re-enable device interrupts.
	 * Ideally, a new ndo_busy_poll_stop() could avoid another round.
	 */
	rc = napi->poll(napi, BUSY_POLL_BUDGET);
5432
	trace_napi_poll(napi, rc, BUSY_POLL_BUDGET);
5433 5434 5435 5436 5437 5438
	netpoll_poll_unlock(have_poll_lock);
	if (rc == BUSY_POLL_BUDGET)
		__napi_schedule(napi);
	local_bh_enable();
}

5439 5440 5441
void napi_busy_loop(unsigned int napi_id,
		    bool (*loop_end)(void *, unsigned long),
		    void *loop_end_arg)
E
Eric Dumazet 已提交
5442
{
5443
	unsigned long start_time = loop_end ? busy_loop_current_time() : 0;
5444 5445
	int (*napi_poll)(struct napi_struct *napi, int budget);
	void *have_poll_lock = NULL;
E
Eric Dumazet 已提交
5446
	struct napi_struct *napi;
5447 5448 5449

restart:
	napi_poll = NULL;
E
Eric Dumazet 已提交
5450

5451
	rcu_read_lock();
E
Eric Dumazet 已提交
5452

5453
	napi = napi_by_id(napi_id);
E
Eric Dumazet 已提交
5454 5455 5456
	if (!napi)
		goto out;

5457 5458
	preempt_disable();
	for (;;) {
5459 5460
		int work = 0;

5461
		local_bh_disable();
5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477
		if (!napi_poll) {
			unsigned long val = READ_ONCE(napi->state);

			/* If multiple threads are competing for this napi,
			 * we avoid dirtying napi->state as much as we can.
			 */
			if (val & (NAPIF_STATE_DISABLE | NAPIF_STATE_SCHED |
				   NAPIF_STATE_IN_BUSY_POLL))
				goto count;
			if (cmpxchg(&napi->state, val,
				    val | NAPIF_STATE_IN_BUSY_POLL |
					  NAPIF_STATE_SCHED) != val)
				goto count;
			have_poll_lock = netpoll_poll_lock(napi);
			napi_poll = napi->poll;
		}
5478 5479
		work = napi_poll(napi, BUSY_POLL_BUDGET);
		trace_napi_poll(napi, work, BUSY_POLL_BUDGET);
5480
count:
5481
		if (work > 0)
5482
			__NET_ADD_STATS(dev_net(napi->dev),
5483
					LINUX_MIB_BUSYPOLLRXPACKETS, work);
5484
		local_bh_enable();
E
Eric Dumazet 已提交
5485

5486
		if (!loop_end || loop_end(loop_end_arg, start_time))
5487
			break;
E
Eric Dumazet 已提交
5488

5489 5490 5491 5492 5493 5494
		if (unlikely(need_resched())) {
			if (napi_poll)
				busy_poll_stop(napi, have_poll_lock);
			preempt_enable();
			rcu_read_unlock();
			cond_resched();
5495
			if (loop_end(loop_end_arg, start_time))
5496
				return;
5497 5498
			goto restart;
		}
5499
		cpu_relax();
5500 5501 5502 5503
	}
	if (napi_poll)
		busy_poll_stop(napi, have_poll_lock);
	preempt_enable();
E
Eric Dumazet 已提交
5504
out:
5505
	rcu_read_unlock();
E
Eric Dumazet 已提交
5506
}
5507
EXPORT_SYMBOL(napi_busy_loop);
E
Eric Dumazet 已提交
5508 5509

#endif /* CONFIG_NET_RX_BUSY_POLL */
E
Eliezer Tamir 已提交
5510

5511
static void napi_hash_add(struct napi_struct *napi)
E
Eliezer Tamir 已提交
5512
{
E
Eric Dumazet 已提交
5513 5514
	if (test_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state) ||
	    test_and_set_bit(NAPI_STATE_HASHED, &napi->state))
5515
		return;
E
Eliezer Tamir 已提交
5516

5517
	spin_lock(&napi_hash_lock);
E
Eliezer Tamir 已提交
5518

5519
	/* 0..NR_CPUS range is reserved for sender_cpu use */
5520
	do {
5521 5522
		if (unlikely(++napi_gen_id < MIN_NAPI_ID))
			napi_gen_id = MIN_NAPI_ID;
5523 5524
	} while (napi_by_id(napi_gen_id));
	napi->napi_id = napi_gen_id;
E
Eliezer Tamir 已提交
5525

5526 5527
	hlist_add_head_rcu(&napi->napi_hash_node,
			   &napi_hash[napi->napi_id % HASH_SIZE(napi_hash)]);
E
Eliezer Tamir 已提交
5528

5529
	spin_unlock(&napi_hash_lock);
E
Eliezer Tamir 已提交
5530 5531 5532 5533 5534
}

/* Warning : caller is responsible to make sure rcu grace period
 * is respected before freeing memory containing @napi
 */
5535
bool napi_hash_del(struct napi_struct *napi)
E
Eliezer Tamir 已提交
5536
{
5537 5538
	bool rcu_sync_needed = false;

E
Eliezer Tamir 已提交
5539 5540
	spin_lock(&napi_hash_lock);

5541 5542
	if (test_and_clear_bit(NAPI_STATE_HASHED, &napi->state)) {
		rcu_sync_needed = true;
E
Eliezer Tamir 已提交
5543
		hlist_del_rcu(&napi->napi_hash_node);
5544
	}
E
Eliezer Tamir 已提交
5545
	spin_unlock(&napi_hash_lock);
5546
	return rcu_sync_needed;
E
Eliezer Tamir 已提交
5547 5548 5549
}
EXPORT_SYMBOL_GPL(napi_hash_del);

5550 5551 5552 5553 5554
static enum hrtimer_restart napi_watchdog(struct hrtimer *timer)
{
	struct napi_struct *napi;

	napi = container_of(timer, struct napi_struct, timer);
E
Eric Dumazet 已提交
5555 5556 5557 5558 5559 5560 5561

	/* Note : we use a relaxed variant of napi_schedule_prep() not setting
	 * NAPI_STATE_MISSED, since we do not react to a device IRQ.
	 */
	if (napi->gro_list && !napi_disable_pending(napi) &&
	    !test_and_set_bit(NAPI_STATE_SCHED, &napi->state))
		__napi_schedule_irqoff(napi);
5562 5563 5564 5565

	return HRTIMER_NORESTART;
}

5566 5567 5568 5569
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);
5570 5571
	hrtimer_init(&napi->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_PINNED);
	napi->timer.function = napi_watchdog;
5572
	napi->gro_count = 0;
5573
	napi->gro_list = NULL;
H
Herbert Xu 已提交
5574
	napi->skb = NULL;
5575
	napi->poll = poll;
E
Eric Dumazet 已提交
5576 5577 5578
	if (weight > NAPI_POLL_WEIGHT)
		pr_err_once("netif_napi_add() called with weight %d on device %s\n",
			    weight, dev->name);
5579 5580 5581
	napi->weight = weight;
	list_add(&napi->dev_list, &dev->napi_list);
	napi->dev = dev;
H
Herbert Xu 已提交
5582
#ifdef CONFIG_NETPOLL
5583 5584 5585
	napi->poll_owner = -1;
#endif
	set_bit(NAPI_STATE_SCHED, &napi->state);
5586
	napi_hash_add(napi);
5587 5588 5589
}
EXPORT_SYMBOL(netif_napi_add);

5590 5591 5592 5593 5594 5595 5596
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);
5597 5598
	while (test_and_set_bit(NAPI_STATE_NPSVC, &n->state))
		msleep(1);
5599 5600 5601 5602 5603 5604 5605

	hrtimer_cancel(&n->timer);

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

5606
/* Must be called in process context */
5607 5608
void netif_napi_del(struct napi_struct *napi)
{
5609 5610 5611
	might_sleep();
	if (napi_hash_del(napi))
		synchronize_net();
5612
	list_del_init(&napi->dev_list);
5613
	napi_free_frags(napi);
5614

E
Eric Dumazet 已提交
5615
	kfree_skb_list(napi->gro_list);
5616
	napi->gro_list = NULL;
5617
	napi->gro_count = 0;
5618 5619 5620
}
EXPORT_SYMBOL(netif_napi_del);

5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640
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);
5641
		trace_napi_poll(n, work, weight);
5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665
	}

	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);
	}

5666 5667 5668 5669 5670 5671 5672 5673 5674
	/* 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;
	}

5675 5676 5677 5678 5679 5680 5681 5682
	list_add_tail(&n->poll_list, repoll);

out_unlock:
	netpoll_poll_unlock(have);

	return work;
}

5683
static __latent_entropy void net_rx_action(struct softirq_action *h)
L
Linus Torvalds 已提交
5684
{
5685
	struct softnet_data *sd = this_cpu_ptr(&softnet_data);
5686 5687
	unsigned long time_limit = jiffies +
		usecs_to_jiffies(netdev_budget_usecs);
5688
	int budget = netdev_budget;
E
Eric Dumazet 已提交
5689 5690
	LIST_HEAD(list);
	LIST_HEAD(repoll);
5691

L
Linus Torvalds 已提交
5692
	local_irq_disable();
E
Eric Dumazet 已提交
5693 5694
	list_splice_init(&sd->poll_list, &list);
	local_irq_enable();
L
Linus Torvalds 已提交
5695

H
Herbert Xu 已提交
5696
	for (;;) {
5697
		struct napi_struct *n;
L
Linus Torvalds 已提交
5698

H
Herbert Xu 已提交
5699 5700
		if (list_empty(&list)) {
			if (!sd_has_rps_ipi_waiting(sd) && list_empty(&repoll))
5701
				goto out;
H
Herbert Xu 已提交
5702 5703 5704
			break;
		}

5705 5706 5707
		n = list_first_entry(&list, struct napi_struct, poll_list);
		budget -= napi_poll(n, &repoll);

E
Eric Dumazet 已提交
5708
		/* If softirq window is exhausted then punt.
5709 5710
		 * Allow this to run for 2 jiffies since which will allow
		 * an average latency of 1.5/HZ.
5711
		 */
H
Herbert Xu 已提交
5712 5713 5714 5715 5716
		if (unlikely(budget <= 0 ||
			     time_after_eq(jiffies, time_limit))) {
			sd->time_squeeze++;
			break;
		}
L
Linus Torvalds 已提交
5717
	}
E
Eric Dumazet 已提交
5718 5719 5720 5721 5722 5723 5724 5725 5726

	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);

5727
	net_rps_action_and_irq_enable(sd);
5728 5729
out:
	__kfree_skb_flush();
L
Linus Torvalds 已提交
5730 5731
}

5732
struct netdev_adjacent {
J
Jiri Pirko 已提交
5733
	struct net_device *dev;
5734 5735

	/* upper master flag, there can only be one master device per list */
J
Jiri Pirko 已提交
5736
	bool master;
5737 5738 5739 5740

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

5741 5742 5743
	/* private field for the users */
	void *private;

J
Jiri Pirko 已提交
5744 5745 5746 5747
	struct list_head list;
	struct rcu_head rcu;
};

5748
static struct netdev_adjacent *__netdev_find_adj(struct net_device *adj_dev,
5749
						 struct list_head *adj_list)
J
Jiri Pirko 已提交
5750
{
5751 5752
	struct netdev_adjacent *adj;

5753
	list_for_each_entry(adj, adj_list, list) {
5754 5755
		if (adj->dev == adj_dev)
			return adj;
J
Jiri Pirko 已提交
5756 5757 5758 5759
	}
	return NULL;
}

5760 5761 5762 5763 5764 5765 5766
static int __netdev_has_upper_dev(struct net_device *upper_dev, void *data)
{
	struct net_device *dev = data;

	return upper_dev == dev;
}

J
Jiri Pirko 已提交
5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780
/**
 * 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();

5781 5782
	return netdev_walk_all_upper_dev_rcu(dev, __netdev_has_upper_dev,
					     upper_dev);
J
Jiri Pirko 已提交
5783 5784 5785
}
EXPORT_SYMBOL(netdev_has_upper_dev);

5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803
/**
 * netdev_has_upper_dev_all - 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 the entire upper device chain.
 * The caller must hold rcu lock.
 */

bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
				  struct net_device *upper_dev)
{
	return !!netdev_walk_all_upper_dev_rcu(dev, __netdev_has_upper_dev,
					       upper_dev);
}
EXPORT_SYMBOL(netdev_has_upper_dev_all_rcu);

J
Jiri Pirko 已提交
5804 5805 5806 5807 5808 5809 5810
/**
 * 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.
 */
5811
bool netdev_has_any_upper_dev(struct net_device *dev)
J
Jiri Pirko 已提交
5812 5813 5814
{
	ASSERT_RTNL();

5815
	return !list_empty(&dev->adj_list.upper);
J
Jiri Pirko 已提交
5816
}
5817
EXPORT_SYMBOL(netdev_has_any_upper_dev);
J
Jiri Pirko 已提交
5818 5819 5820 5821 5822 5823 5824 5825 5826 5827

/**
 * 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)
{
5828
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
5829 5830 5831

	ASSERT_RTNL();

5832
	if (list_empty(&dev->adj_list.upper))
J
Jiri Pirko 已提交
5833 5834
		return NULL;

5835
	upper = list_first_entry(&dev->adj_list.upper,
5836
				 struct netdev_adjacent, list);
J
Jiri Pirko 已提交
5837 5838 5839 5840 5841 5842
	if (likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get);

5843 5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856
/**
 * netdev_has_any_lower_dev - Check if device is linked to some device
 * @dev: device
 *
 * Find out if a device is linked to a lower device and return true in case
 * it is. The caller must hold the RTNL lock.
 */
static bool netdev_has_any_lower_dev(struct net_device *dev)
{
	ASSERT_RTNL();

	return !list_empty(&dev->adj_list.lower);
}

5857 5858 5859 5860 5861 5862 5863 5864 5865 5866
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);

5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892
/**
 * 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);

5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937
static struct net_device *netdev_next_upper_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;
}

int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
				  int (*fn)(struct net_device *dev,
					    void *data),
				  void *data)
{
	struct net_device *udev;
	struct list_head *iter;
	int ret;

	for (iter = &dev->adj_list.upper,
	     udev = netdev_next_upper_dev_rcu(dev, &iter);
	     udev;
	     udev = netdev_next_upper_dev_rcu(dev, &iter)) {
		/* first is the upper device itself */
		ret = fn(udev, data);
		if (ret)
			return ret;

		/* then look at all of its upper devices */
		ret = netdev_walk_all_upper_dev_rcu(udev, fn, data);
		if (ret)
			return ret;
	}

	return 0;
}
EXPORT_SYMBOL_GPL(netdev_walk_all_upper_dev_rcu);

5938 5939 5940 5941 5942 5943 5944 5945 5946
/**
 * 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
5947
 * list will remain unchanged.
5948 5949 5950 5951 5952 5953 5954 5955 5956 5957 5958
 */
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;

5959
	*iter = lower->list.next;
5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985 5986

	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;

5987
	*iter = &lower->list;
5988 5989 5990 5991 5992

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

5993 5994 5995 5996 5997 5998 5999 6000 6001
/**
 * 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
6002
 * list will remain unchanged.
6003 6004 6005 6006 6007
 */
void *netdev_lower_get_next(struct net_device *dev, struct list_head **iter)
{
	struct netdev_adjacent *lower;

6008
	lower = list_entry(*iter, struct netdev_adjacent, list);
6009 6010 6011 6012

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

6013
	*iter = lower->list.next;
6014 6015 6016 6017 6018

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

6019 6020 6021 6022 6023
static struct net_device *netdev_next_lower_dev(struct net_device *dev,
						struct list_head **iter)
{
	struct netdev_adjacent *lower;

6024
	lower = list_entry((*iter)->next, struct netdev_adjacent, list);
6025 6026 6027 6028

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

6029
	*iter = &lower->list;
6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087 6088 6089 6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103

	return lower->dev;
}

int netdev_walk_all_lower_dev(struct net_device *dev,
			      int (*fn)(struct net_device *dev,
					void *data),
			      void *data)
{
	struct net_device *ldev;
	struct list_head *iter;
	int ret;

	for (iter = &dev->adj_list.lower,
	     ldev = netdev_next_lower_dev(dev, &iter);
	     ldev;
	     ldev = netdev_next_lower_dev(dev, &iter)) {
		/* first is the lower device itself */
		ret = fn(ldev, data);
		if (ret)
			return ret;

		/* then look at all of its lower devices */
		ret = netdev_walk_all_lower_dev(ldev, fn, data);
		if (ret)
			return ret;
	}

	return 0;
}
EXPORT_SYMBOL_GPL(netdev_walk_all_lower_dev);

static struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
						    struct list_head **iter)
{
	struct netdev_adjacent *lower;

	lower = list_entry_rcu((*iter)->next, struct netdev_adjacent, list);
	if (&lower->list == &dev->adj_list.lower)
		return NULL;

	*iter = &lower->list;

	return lower->dev;
}

int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
				  int (*fn)(struct net_device *dev,
					    void *data),
				  void *data)
{
	struct net_device *ldev;
	struct list_head *iter;
	int ret;

	for (iter = &dev->adj_list.lower,
	     ldev = netdev_next_lower_dev_rcu(dev, &iter);
	     ldev;
	     ldev = netdev_next_lower_dev_rcu(dev, &iter)) {
		/* first is the lower device itself */
		ret = fn(ldev, data);
		if (ret)
			return ret;

		/* then look at all of its lower devices */
		ret = netdev_walk_all_lower_dev_rcu(ldev, fn, data);
		if (ret)
			return ret;
	}

	return 0;
}
EXPORT_SYMBOL_GPL(netdev_walk_all_lower_dev_rcu);

6104 6105 6106 6107 6108 6109 6110 6111 6112 6113 6114 6115 6116 6117 6118 6119 6120 6121 6122 6123 6124
/**
 * 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
6125 6126 6127 6128 6129 6130 6131 6132 6133
/**
 * 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)
{
6134
	struct netdev_adjacent *upper;
J
Jiri Pirko 已提交
6135

6136
	upper = list_first_or_null_rcu(&dev->adj_list.upper,
6137
				       struct netdev_adjacent, list);
J
Jiri Pirko 已提交
6138 6139 6140 6141 6142 6143
	if (upper && likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get_rcu);

6144
static int netdev_adjacent_sysfs_add(struct net_device *dev,
6145 6146 6147 6148
			      struct net_device *adj_dev,
			      struct list_head *dev_list)
{
	char linkname[IFNAMSIZ+7];
6149

6150 6151 6152 6153 6154
	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);
}
6155
static void netdev_adjacent_sysfs_del(struct net_device *dev,
6156 6157 6158 6159
			       char *name,
			       struct list_head *dev_list)
{
	char linkname[IFNAMSIZ+7];
6160

6161 6162 6163 6164 6165
	sprintf(linkname, dev_list == &dev->adj_list.upper ?
		"upper_%s" : "lower_%s", name);
	sysfs_remove_link(&(dev->dev.kobj), linkname);
}

6166 6167 6168 6169 6170 6171 6172 6173
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));
}
6174

6175 6176
static int __netdev_adjacent_dev_insert(struct net_device *dev,
					struct net_device *adj_dev,
6177
					struct list_head *dev_list,
6178
					void *private, bool master)
6179 6180
{
	struct netdev_adjacent *adj;
6181
	int ret;
6182

6183
	adj = __netdev_find_adj(adj_dev, dev_list);
6184 6185

	if (adj) {
6186
		adj->ref_nr += 1;
6187 6188 6189
		pr_debug("Insert adjacency: dev %s adj_dev %s adj->ref_nr %d\n",
			 dev->name, adj_dev->name, adj->ref_nr);

6190 6191 6192 6193 6194 6195 6196 6197 6198
		return 0;
	}

	adj = kmalloc(sizeof(*adj), GFP_KERNEL);
	if (!adj)
		return -ENOMEM;

	adj->dev = adj_dev;
	adj->master = master;
6199
	adj->ref_nr = 1;
6200
	adj->private = private;
6201
	dev_hold(adj_dev);
6202

6203 6204
	pr_debug("Insert adjacency: dev %s adj_dev %s adj->ref_nr %d; dev_hold on %s\n",
		 dev->name, adj_dev->name, adj->ref_nr, adj_dev->name);
6205

6206
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list)) {
6207
		ret = netdev_adjacent_sysfs_add(dev, adj_dev, dev_list);
6208 6209 6210 6211
		if (ret)
			goto free_adj;
	}

6212
	/* Ensure that master link is always the first item in list. */
6213 6214 6215 6216
	if (master) {
		ret = sysfs_create_link(&(dev->dev.kobj),
					&(adj_dev->dev.kobj), "master");
		if (ret)
6217
			goto remove_symlinks;
6218

6219
		list_add_rcu(&adj->list, dev_list);
6220
	} else {
6221
		list_add_tail_rcu(&adj->list, dev_list);
6222
	}
6223 6224

	return 0;
6225

6226
remove_symlinks:
6227
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
6228
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
6229 6230
free_adj:
	kfree(adj);
6231
	dev_put(adj_dev);
6232 6233

	return ret;
6234 6235
}

S
stephen hemminger 已提交
6236 6237
static void __netdev_adjacent_dev_remove(struct net_device *dev,
					 struct net_device *adj_dev,
6238
					 u16 ref_nr,
S
stephen hemminger 已提交
6239
					 struct list_head *dev_list)
6240 6241 6242
{
	struct netdev_adjacent *adj;

6243 6244 6245
	pr_debug("Remove adjacency: dev %s adj_dev %s ref_nr %d\n",
		 dev->name, adj_dev->name, ref_nr);

6246
	adj = __netdev_find_adj(adj_dev, dev_list);
6247

6248
	if (!adj) {
6249
		pr_err("Adjacency does not exist for device %s from %s\n",
6250
		       dev->name, adj_dev->name);
6251 6252
		WARN_ON(1);
		return;
6253
	}
6254

6255
	if (adj->ref_nr > ref_nr) {
6256 6257 6258
		pr_debug("adjacency: %s to %s ref_nr - %d = %d\n",
			 dev->name, adj_dev->name, ref_nr,
			 adj->ref_nr - ref_nr);
6259
		adj->ref_nr -= ref_nr;
6260 6261 6262
		return;
	}

6263 6264 6265
	if (adj->master)
		sysfs_remove_link(&(dev->dev.kobj), "master");

6266
	if (netdev_adjacent_is_neigh_list(dev, adj_dev, dev_list))
6267
		netdev_adjacent_sysfs_del(dev, adj_dev->name, dev_list);
6268

6269
	list_del_rcu(&adj->list);
6270
	pr_debug("adjacency: dev_put for %s, because link removed from %s to %s\n",
6271
		 adj_dev->name, dev->name, adj_dev->name);
6272 6273 6274 6275
	dev_put(adj_dev);
	kfree_rcu(adj, rcu);
}

S
stephen hemminger 已提交
6276 6277 6278 6279 6280
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)
6281 6282 6283
{
	int ret;

6284
	ret = __netdev_adjacent_dev_insert(dev, upper_dev, up_list,
6285
					   private, master);
6286 6287 6288
	if (ret)
		return ret;

6289
	ret = __netdev_adjacent_dev_insert(upper_dev, dev, down_list,
6290
					   private, false);
6291
	if (ret) {
6292
		__netdev_adjacent_dev_remove(dev, upper_dev, 1, up_list);
6293 6294 6295 6296 6297 6298
		return ret;
	}

	return 0;
}

S
stephen hemminger 已提交
6299 6300
static void __netdev_adjacent_dev_unlink_lists(struct net_device *dev,
					       struct net_device *upper_dev,
6301
					       u16 ref_nr,
S
stephen hemminger 已提交
6302 6303
					       struct list_head *up_list,
					       struct list_head *down_list)
6304
{
6305 6306
	__netdev_adjacent_dev_remove(dev, upper_dev, ref_nr, up_list);
	__netdev_adjacent_dev_remove(upper_dev, dev, ref_nr, down_list);
6307 6308
}

S
stephen hemminger 已提交
6309 6310 6311
static int __netdev_adjacent_dev_link_neighbour(struct net_device *dev,
						struct net_device *upper_dev,
						void *private, bool master)
6312
{
6313 6314 6315 6316
	return __netdev_adjacent_dev_link_lists(dev, upper_dev,
						&dev->adj_list.upper,
						&upper_dev->adj_list.lower,
						private, master);
6317 6318
}

S
stephen hemminger 已提交
6319 6320
static void __netdev_adjacent_dev_unlink_neighbour(struct net_device *dev,
						   struct net_device *upper_dev)
6321
{
6322
	__netdev_adjacent_dev_unlink_lists(dev, upper_dev, 1,
6323 6324 6325
					   &dev->adj_list.upper,
					   &upper_dev->adj_list.lower);
}
6326

J
Jiri Pirko 已提交
6327
static int __netdev_upper_dev_link(struct net_device *dev,
6328
				   struct net_device *upper_dev, bool master,
6329 6330
				   void *upper_priv, void *upper_info,
				   struct netlink_ext_ack *extack)
J
Jiri Pirko 已提交
6331
{
6332 6333 6334
	struct netdev_notifier_changeupper_info changeupper_info = {
		.info = {
			.dev = dev,
6335
			.extack = extack,
6336 6337 6338 6339 6340 6341
		},
		.upper_dev = upper_dev,
		.master = master,
		.linking = true,
		.upper_info = upper_info,
	};
6342
	int ret = 0;
J
Jiri Pirko 已提交
6343 6344 6345 6346 6347 6348 6349

	ASSERT_RTNL();

	if (dev == upper_dev)
		return -EBUSY;

	/* To prevent loops, check if dev is not upper device to upper_dev. */
6350
	if (netdev_has_upper_dev(upper_dev, dev))
J
Jiri Pirko 已提交
6351 6352
		return -EBUSY;

6353
	if (netdev_has_upper_dev(dev, upper_dev))
J
Jiri Pirko 已提交
6354 6355 6356 6357 6358
		return -EEXIST;

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

6359
	ret = call_netdevice_notifiers_info(NETDEV_PRECHANGEUPPER,
6360 6361 6362 6363 6364
					    &changeupper_info.info);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

6365
	ret = __netdev_adjacent_dev_link_neighbour(dev, upper_dev, upper_priv,
6366
						   master);
6367 6368
	if (ret)
		return ret;
J
Jiri Pirko 已提交
6369

6370
	ret = call_netdevice_notifiers_info(NETDEV_CHANGEUPPER,
6371 6372 6373
					    &changeupper_info.info);
	ret = notifier_to_errno(ret);
	if (ret)
6374
		goto rollback;
6375

J
Jiri Pirko 已提交
6376
	return 0;
6377

6378
rollback:
6379
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
6380 6381

	return ret;
J
Jiri Pirko 已提交
6382 6383 6384 6385 6386 6387 6388 6389 6390 6391 6392 6393 6394
}

/**
 * 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,
6395 6396
			  struct net_device *upper_dev,
			  struct netlink_ext_ack *extack)
J
Jiri Pirko 已提交
6397
{
6398 6399
	return __netdev_upper_dev_link(dev, upper_dev, false,
				       NULL, NULL, extack);
J
Jiri Pirko 已提交
6400 6401 6402 6403 6404 6405 6406
}
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
6407
 * @upper_priv: upper device private
6408
 * @upper_info: upper info to be passed down via notifier
J
Jiri Pirko 已提交
6409 6410 6411 6412 6413 6414 6415 6416
 *
 * 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,
6417
				 struct net_device *upper_dev,
6418 6419
				 void *upper_priv, void *upper_info,
				 struct netlink_ext_ack *extack)
J
Jiri Pirko 已提交
6420
{
6421
	return __netdev_upper_dev_link(dev, upper_dev, true,
6422
				       upper_priv, upper_info, extack);
J
Jiri Pirko 已提交
6423 6424 6425 6426 6427 6428 6429 6430 6431 6432 6433 6434 6435 6436
}
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)
{
6437 6438 6439 6440 6441 6442 6443
	struct netdev_notifier_changeupper_info changeupper_info = {
		.info = {
			.dev = dev,
		},
		.upper_dev = upper_dev,
		.linking = false,
	};
6444

J
Jiri Pirko 已提交
6445 6446
	ASSERT_RTNL();

6447 6448
	changeupper_info.master = netdev_master_upper_dev_get(dev) == upper_dev;

6449
	call_netdevice_notifiers_info(NETDEV_PRECHANGEUPPER,
6450 6451
				      &changeupper_info.info);

6452
	__netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
6453

6454
	call_netdevice_notifiers_info(NETDEV_CHANGEUPPER,
6455
				      &changeupper_info.info);
J
Jiri Pirko 已提交
6456 6457 6458
}
EXPORT_SYMBOL(netdev_upper_dev_unlink);

6459 6460 6461
/**
 * netdev_bonding_info_change - Dispatch event about slave change
 * @dev: device
6462
 * @bonding_info: info to dispatch
6463 6464 6465 6466 6467 6468 6469
 *
 * 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)
{
6470 6471 6472
	struct netdev_notifier_bonding_info info = {
		.info.dev = dev,
	};
6473 6474 6475

	memcpy(&info.bonding_info, bonding_info,
	       sizeof(struct netdev_bonding_info));
6476
	call_netdevice_notifiers_info(NETDEV_BONDING_INFO,
6477 6478 6479 6480
				      &info.info);
}
EXPORT_SYMBOL(netdev_bonding_info_change);

E
Eric Dumazet 已提交
6481
static void netdev_adjacent_add_links(struct net_device *dev)
6482 6483 6484 6485 6486 6487
{
	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
6488
		if (!net_eq(net, dev_net(iter->dev)))
6489 6490 6491 6492 6493 6494 6495 6496
			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
6497
		if (!net_eq(net, dev_net(iter->dev)))
6498 6499 6500 6501 6502 6503 6504 6505
			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
6506
static void netdev_adjacent_del_links(struct net_device *dev)
6507 6508 6509 6510 6511 6512
{
	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
6513
		if (!net_eq(net, dev_net(iter->dev)))
6514 6515 6516 6517 6518 6519 6520 6521
			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
6522
		if (!net_eq(net, dev_net(iter->dev)))
6523 6524 6525 6526 6527 6528 6529 6530
			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);
	}
}

6531
void netdev_adjacent_rename_links(struct net_device *dev, char *oldname)
6532
{
6533
	struct netdev_adjacent *iter;
6534

6535 6536
	struct net *net = dev_net(dev);

6537
	list_for_each_entry(iter, &dev->adj_list.upper, list) {
W
Wei Tang 已提交
6538
		if (!net_eq(net, dev_net(iter->dev)))
6539
			continue;
6540 6541 6542 6543 6544
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.lower);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.lower);
	}
6545

6546
	list_for_each_entry(iter, &dev->adj_list.lower, list) {
W
Wei Tang 已提交
6547
		if (!net_eq(net, dev_net(iter->dev)))
6548
			continue;
6549 6550 6551 6552 6553
		netdev_adjacent_sysfs_del(iter->dev, oldname,
					  &iter->dev->adj_list.upper);
		netdev_adjacent_sysfs_add(iter->dev, dev,
					  &iter->dev->adj_list.upper);
	}
6554 6555 6556 6557 6558 6559 6560 6561 6562
}

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

	if (!lower_dev)
		return NULL;
6563
	lower = __netdev_find_adj(lower_dev, &dev->adj_list.lower);
6564 6565 6566 6567 6568 6569 6570
	if (!lower)
		return NULL;

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

6571

6572
int dev_get_nest_level(struct net_device *dev)
6573 6574 6575 6576 6577 6578 6579 6580 6581
{
	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) {
6582
		nest = dev_get_nest_level(lower);
6583 6584 6585 6586
		if (max_nest < nest)
			max_nest = nest;
	}

6587
	return max_nest + 1;
6588 6589 6590
}
EXPORT_SYMBOL(dev_get_nest_level);

6591 6592 6593 6594 6595 6596 6597 6598 6599 6600 6601
/**
 * 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)
{
6602 6603 6604
	struct netdev_notifier_changelowerstate_info changelowerstate_info = {
		.info.dev = lower_dev,
	};
6605 6606 6607

	ASSERT_RTNL();
	changelowerstate_info.lower_state_info = lower_state_info;
6608
	call_netdevice_notifiers_info(NETDEV_CHANGELOWERSTATE,
6609 6610 6611 6612
				      &changelowerstate_info.info);
}
EXPORT_SYMBOL(netdev_lower_state_changed);

6613 6614
static void dev_change_rx_flags(struct net_device *dev, int flags)
{
6615 6616
	const struct net_device_ops *ops = dev->netdev_ops;

6617
	if (ops->ndo_change_rx_flags)
6618
		ops->ndo_change_rx_flags(dev, flags);
6619 6620
}

6621
static int __dev_set_promiscuity(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
6622
{
6623
	unsigned int old_flags = dev->flags;
6624 6625
	kuid_t uid;
	kgid_t gid;
L
Linus Torvalds 已提交
6626

6627 6628
	ASSERT_RTNL();

6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639
	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;
6640 6641
			pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n",
				dev->name);
6642 6643 6644
			return -EOVERFLOW;
		}
	}
6645
	if (dev->flags != old_flags) {
6646 6647 6648
		pr_info("device %s %s promiscuous mode\n",
			dev->name,
			dev->flags & IFF_PROMISC ? "entered" : "left");
6649 6650
		if (audit_enabled) {
			current_uid_gid(&uid, &gid);
6651 6652 6653 6654 6655
			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),
6656
				from_kuid(&init_user_ns, audit_get_loginuid(current)),
6657 6658
				from_kuid(&init_user_ns, uid),
				from_kgid(&init_user_ns, gid),
6659
				audit_get_sessionid(current));
6660
		}
6661

6662
		dev_change_rx_flags(dev, IFF_PROMISC);
L
Linus Torvalds 已提交
6663
	}
6664 6665
	if (notify)
		__dev_notify_flags(dev, old_flags, IFF_PROMISC);
6666
	return 0;
L
Linus Torvalds 已提交
6667 6668
}

6669 6670 6671 6672 6673 6674 6675 6676 6677
/**
 *	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.
6678
 *	Return 0 if successful or a negative errno code on error.
6679
 */
6680
int dev_set_promiscuity(struct net_device *dev, int inc)
6681
{
6682
	unsigned int old_flags = dev->flags;
6683
	int err;
6684

6685
	err = __dev_set_promiscuity(dev, inc, true);
6686
	if (err < 0)
6687
		return err;
6688 6689
	if (dev->flags != old_flags)
		dev_set_rx_mode(dev);
6690
	return err;
6691
}
E
Eric Dumazet 已提交
6692
EXPORT_SYMBOL(dev_set_promiscuity);
6693

6694
static int __dev_set_allmulti(struct net_device *dev, int inc, bool notify)
L
Linus Torvalds 已提交
6695
{
6696
	unsigned int old_flags = dev->flags, old_gflags = dev->gflags;
L
Linus Torvalds 已提交
6697

6698 6699
	ASSERT_RTNL();

L
Linus Torvalds 已提交
6700
	dev->flags |= IFF_ALLMULTI;
6701 6702 6703 6704 6705 6706 6707 6708 6709 6710
	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;
6711 6712
			pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n",
				dev->name);
6713 6714 6715
			return -EOVERFLOW;
		}
	}
6716
	if (dev->flags ^ old_flags) {
6717
		dev_change_rx_flags(dev, IFF_ALLMULTI);
6718
		dev_set_rx_mode(dev);
6719 6720 6721
		if (notify)
			__dev_notify_flags(dev, old_flags,
					   dev->gflags ^ old_gflags);
6722
	}
6723
	return 0;
6724
}
6725 6726 6727 6728 6729 6730 6731 6732 6733 6734 6735 6736 6737 6738 6739 6740 6741 6742

/**
 *	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
6743
EXPORT_SYMBOL(dev_set_allmulti);
6744 6745 6746 6747

/*
 *	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
6748
 *	filtering it is put in promiscuous mode while unicast addresses
6749 6750 6751 6752
 *	are present.
 */
void __dev_set_rx_mode(struct net_device *dev)
{
6753 6754
	const struct net_device_ops *ops = dev->netdev_ops;

6755 6756 6757 6758 6759
	/* dev_open will call this function so the list will stay sane. */
	if (!(dev->flags&IFF_UP))
		return;

	if (!netif_device_present(dev))
6760
		return;
6761

6762
	if (!(dev->priv_flags & IFF_UNICAST_FLT)) {
6763 6764 6765
		/* Unicast addresses changes may only happen under the rtnl,
		 * therefore calling __dev_set_promiscuity here is safe.
		 */
6766
		if (!netdev_uc_empty(dev) && !dev->uc_promisc) {
6767
			__dev_set_promiscuity(dev, 1, false);
6768
			dev->uc_promisc = true;
6769
		} else if (netdev_uc_empty(dev) && dev->uc_promisc) {
6770
			__dev_set_promiscuity(dev, -1, false);
6771
			dev->uc_promisc = false;
6772 6773
		}
	}
6774 6775 6776

	if (ops->ndo_set_rx_mode)
		ops->ndo_set_rx_mode(dev);
6777 6778 6779 6780
}

void dev_set_rx_mode(struct net_device *dev)
{
6781
	netif_addr_lock_bh(dev);
6782
	__dev_set_rx_mode(dev);
6783
	netif_addr_unlock_bh(dev);
L
Linus Torvalds 已提交
6784 6785
}

6786 6787 6788 6789 6790 6791
/**
 *	dev_get_flags - get flags reported to userspace
 *	@dev: device
 *
 *	Get the combination of flag bits exported through APIs to userspace.
 */
6792
unsigned int dev_get_flags(const struct net_device *dev)
L
Linus Torvalds 已提交
6793
{
6794
	unsigned int flags;
L
Linus Torvalds 已提交
6795 6796 6797

	flags = (dev->flags & ~(IFF_PROMISC |
				IFF_ALLMULTI |
S
Stefan Rompf 已提交
6798 6799 6800
				IFF_RUNNING |
				IFF_LOWER_UP |
				IFF_DORMANT)) |
L
Linus Torvalds 已提交
6801 6802 6803
		(dev->gflags & (IFF_PROMISC |
				IFF_ALLMULTI));

S
Stefan Rompf 已提交
6804 6805 6806 6807 6808 6809 6810 6811
	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
6812 6813 6814

	return flags;
}
E
Eric Dumazet 已提交
6815
EXPORT_SYMBOL(dev_get_flags);
L
Linus Torvalds 已提交
6816

6817
int __dev_change_flags(struct net_device *dev, unsigned int flags)
L
Linus Torvalds 已提交
6818
{
6819
	unsigned int old_flags = dev->flags;
6820
	int ret;
L
Linus Torvalds 已提交
6821

6822 6823
	ASSERT_RTNL();

L
Linus Torvalds 已提交
6824 6825 6826 6827 6828 6829 6830 6831 6832 6833 6834 6835 6836 6837
	/*
	 *	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.
	 */

6838 6839
	if ((old_flags ^ flags) & IFF_MULTICAST)
		dev_change_rx_flags(dev, IFF_MULTICAST);
6840

6841
	dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
6842 6843 6844 6845 6846 6847 6848 6849

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

	ret = 0;
6850 6851 6852 6853 6854 6855
	if ((old_flags ^ flags) & IFF_UP) {
		if (old_flags & IFF_UP)
			__dev_close(dev);
		else
			ret = __dev_open(dev);
	}
L
Linus Torvalds 已提交
6856 6857

	if ((flags ^ dev->gflags) & IFF_PROMISC) {
E
Eric Dumazet 已提交
6858
		int inc = (flags & IFF_PROMISC) ? 1 : -1;
6859
		unsigned int old_flags = dev->flags;
E
Eric Dumazet 已提交
6860

L
Linus Torvalds 已提交
6861
		dev->gflags ^= IFF_PROMISC;
6862 6863 6864 6865

		if (__dev_set_promiscuity(dev, inc, false) >= 0)
			if (dev->flags != old_flags)
				dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
6866 6867 6868
	}

	/* NOTE: order of synchronization of IFF_PROMISC and IFF_ALLMULTI
6869 6870
	 * is important. Some (broken) drivers set IFF_PROMISC, when
	 * IFF_ALLMULTI is requested not asking us and not reporting.
L
Linus Torvalds 已提交
6871 6872
	 */
	if ((flags ^ dev->gflags) & IFF_ALLMULTI) {
E
Eric Dumazet 已提交
6873 6874
		int inc = (flags & IFF_ALLMULTI) ? 1 : -1;

L
Linus Torvalds 已提交
6875
		dev->gflags ^= IFF_ALLMULTI;
6876
		__dev_set_allmulti(dev, inc, false);
L
Linus Torvalds 已提交
6877 6878
	}

6879 6880 6881
	return ret;
}

6882 6883
void __dev_notify_flags(struct net_device *dev, unsigned int old_flags,
			unsigned int gchanges)
6884 6885 6886
{
	unsigned int changes = dev->flags ^ old_flags;

6887
	if (gchanges)
6888
		rtmsg_ifinfo(RTM_NEWLINK, dev, gchanges, GFP_ATOMIC);
6889

6890 6891 6892 6893 6894 6895 6896 6897
	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 &&
6898
	    (changes & ~(IFF_UP | IFF_PROMISC | IFF_ALLMULTI | IFF_VOLATILE))) {
6899 6900 6901 6902 6903 6904
		struct netdev_notifier_change_info change_info = {
			.info = {
				.dev = dev,
			},
			.flags_changed = changes,
		};
6905

6906
		call_netdevice_notifiers_info(NETDEV_CHANGE, &change_info.info);
6907
	}
6908 6909 6910 6911 6912 6913 6914 6915 6916 6917
}

/**
 *	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.
 */
6918
int dev_change_flags(struct net_device *dev, unsigned int flags)
6919
{
6920
	int ret;
6921
	unsigned int changes, old_flags = dev->flags, old_gflags = dev->gflags;
6922 6923 6924 6925 6926

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

6927
	changes = (old_flags ^ dev->flags) | (old_gflags ^ dev->gflags);
6928
	__dev_notify_flags(dev, old_flags, changes);
L
Linus Torvalds 已提交
6929 6930
	return ret;
}
E
Eric Dumazet 已提交
6931
EXPORT_SYMBOL(dev_change_flags);
L
Linus Torvalds 已提交
6932

6933
int __dev_set_mtu(struct net_device *dev, int new_mtu)
6934 6935 6936 6937 6938 6939 6940 6941 6942
{
	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;
}
6943
EXPORT_SYMBOL(__dev_set_mtu);
6944

6945 6946 6947 6948 6949 6950 6951
/**
 *	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
6952 6953
int dev_set_mtu(struct net_device *dev, int new_mtu)
{
6954
	int err, orig_mtu;
L
Linus Torvalds 已提交
6955 6956 6957 6958

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

6959 6960 6961 6962
	/* MTU must be positive, and in range */
	if (new_mtu < 0 || new_mtu < dev->min_mtu) {
		net_err_ratelimited("%s: Invalid MTU %d requested, hw min %d\n",
				    dev->name, new_mtu, dev->min_mtu);
L
Linus Torvalds 已提交
6963
		return -EINVAL;
6964 6965 6966 6967
	}

	if (dev->max_mtu > 0 && new_mtu > dev->max_mtu) {
		net_err_ratelimited("%s: Invalid MTU %d requested, hw max %d\n",
6968
				    dev->name, new_mtu, dev->max_mtu);
6969 6970
		return -EINVAL;
	}
L
Linus Torvalds 已提交
6971 6972 6973 6974

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

6975 6976 6977 6978
	err = call_netdevice_notifiers(NETDEV_PRECHANGEMTU, dev);
	err = notifier_to_errno(err);
	if (err)
		return err;
6979

6980 6981
	orig_mtu = dev->mtu;
	err = __dev_set_mtu(dev, new_mtu);
6982

6983 6984 6985 6986 6987 6988 6989 6990 6991 6992 6993
	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
6994 6995
	return err;
}
E
Eric Dumazet 已提交
6996
EXPORT_SYMBOL(dev_set_mtu);
L
Linus Torvalds 已提交
6997

6998 6999 7000 7001 7002 7003 7004 7005 7006 7007 7008
/**
 *	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);

7009 7010 7011 7012 7013 7014 7015
/**
 *	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
7016 7017
int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa)
{
7018
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
7019 7020
	int err;

7021
	if (!ops->ndo_set_mac_address)
L
Linus Torvalds 已提交
7022 7023 7024 7025 7026
		return -EOPNOTSUPP;
	if (sa->sa_family != dev->type)
		return -EINVAL;
	if (!netif_device_present(dev))
		return -ENODEV;
7027
	err = ops->ndo_set_mac_address(dev, sa);
7028 7029
	if (err)
		return err;
J
Jiri Pirko 已提交
7030
	dev->addr_assign_type = NET_ADDR_SET;
7031
	call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
7032
	add_device_randomness(dev->dev_addr, dev->addr_len);
7033
	return 0;
L
Linus Torvalds 已提交
7034
}
E
Eric Dumazet 已提交
7035
EXPORT_SYMBOL(dev_set_mac_address);
L
Linus Torvalds 已提交
7036

J
Jiri Pirko 已提交
7037 7038 7039
/**
 *	dev_change_carrier - Change device carrier
 *	@dev: device
7040
 *	@new_carrier: new value
J
Jiri Pirko 已提交
7041 7042 7043 7044 7045 7046 7047 7048 7049 7050 7051 7052 7053 7054 7055
 *
 *	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);

7056 7057 7058 7059 7060 7061 7062 7063
/**
 *	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,
7064
			 struct netdev_phys_item_id *ppid)
7065 7066 7067 7068 7069 7070 7071 7072 7073
{
	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);

7074 7075 7076 7077
/**
 *	dev_get_phys_port_name - Get device physical port name
 *	@dev: device
 *	@name: port name
7078
 *	@len: limit of bytes to copy to name
7079 7080 7081 7082 7083 7084 7085 7086 7087 7088 7089 7090 7091 7092
 *
 *	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);

7093 7094 7095 7096 7097 7098 7099 7100 7101 7102 7103 7104 7105 7106 7107 7108 7109 7110 7111 7112
/**
 *	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);

7113 7114
void __dev_xdp_query(struct net_device *dev, bpf_op_t bpf_op,
		     struct netdev_bpf *xdp)
7115
{
7116 7117
	memset(xdp, 0, sizeof(*xdp));
	xdp->command = XDP_QUERY_PROG;
7118 7119

	/* Query must always succeed. */
7120 7121 7122 7123 7124 7125 7126 7127
	WARN_ON(bpf_op(dev, xdp) < 0);
}

static u8 __dev_xdp_attached(struct net_device *dev, bpf_op_t bpf_op)
{
	struct netdev_bpf xdp;

	__dev_xdp_query(dev, bpf_op, &xdp);
M
Martin KaFai Lau 已提交
7128

7129 7130 7131
	return xdp.prog_attached;
}

7132
static int dev_xdp_install(struct net_device *dev, bpf_op_t bpf_op,
7133
			   struct netlink_ext_ack *extack, u32 flags,
7134 7135
			   struct bpf_prog *prog)
{
7136
	struct netdev_bpf xdp;
7137 7138

	memset(&xdp, 0, sizeof(xdp));
7139 7140 7141 7142
	if (flags & XDP_FLAGS_HW_MODE)
		xdp.command = XDP_SETUP_PROG_HW;
	else
		xdp.command = XDP_SETUP_PROG;
7143
	xdp.extack = extack;
7144
	xdp.flags = flags;
7145 7146
	xdp.prog = prog;

7147
	return bpf_op(dev, &xdp);
7148 7149
}

7150 7151 7152 7153 7154 7155 7156 7157 7158 7159 7160 7161 7162 7163 7164 7165 7166 7167 7168 7169 7170
static void dev_xdp_uninstall(struct net_device *dev)
{
	struct netdev_bpf xdp;
	bpf_op_t ndo_bpf;

	/* Remove generic XDP */
	WARN_ON(dev_xdp_install(dev, generic_xdp_install, NULL, 0, NULL));

	/* Remove from the driver */
	ndo_bpf = dev->netdev_ops->ndo_bpf;
	if (!ndo_bpf)
		return;

	__dev_xdp_query(dev, ndo_bpf, &xdp);
	if (xdp.prog_attached == XDP_ATTACHED_NONE)
		return;

	/* Program removal should always succeed */
	WARN_ON(dev_xdp_install(dev, ndo_bpf, NULL, xdp.prog_flags, NULL));
}

7171 7172 7173
/**
 *	dev_change_xdp_fd - set or clear a bpf program for a device rx path
 *	@dev: device
7174
 *	@extack: netlink extended ack
7175
 *	@fd: new program fd or negative value to clear
7176
 *	@flags: xdp-related flags
7177 7178 7179
 *
 *	Set or clear a bpf program for a device
 */
7180 7181
int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
		      int fd, u32 flags)
7182 7183 7184
{
	const struct net_device_ops *ops = dev->netdev_ops;
	struct bpf_prog *prog = NULL;
7185
	bpf_op_t bpf_op, bpf_chk;
7186 7187
	int err;

7188 7189
	ASSERT_RTNL();

7190 7191
	bpf_op = bpf_chk = ops->ndo_bpf;
	if (!bpf_op && (flags & (XDP_FLAGS_DRV_MODE | XDP_FLAGS_HW_MODE)))
7192
		return -EOPNOTSUPP;
7193 7194 7195 7196
	if (!bpf_op || (flags & XDP_FLAGS_SKB_MODE))
		bpf_op = generic_xdp_install;
	if (bpf_op == bpf_chk)
		bpf_chk = generic_xdp_install;
D
David S. Miller 已提交
7197

7198
	if (fd >= 0) {
7199
		if (bpf_chk && __dev_xdp_attached(dev, bpf_chk))
7200 7201
			return -EEXIST;
		if ((flags & XDP_FLAGS_UPDATE_IF_NOEXIST) &&
7202
		    __dev_xdp_attached(dev, bpf_op))
7203
			return -EBUSY;
7204

7205 7206
		prog = bpf_prog_get_type_dev(fd, BPF_PROG_TYPE_XDP,
					     bpf_op == ops->ndo_bpf);
7207 7208
		if (IS_ERR(prog))
			return PTR_ERR(prog);
7209 7210 7211 7212 7213 7214 7215

		if (!(flags & XDP_FLAGS_HW_MODE) &&
		    bpf_prog_is_dev_bound(prog->aux)) {
			NL_SET_ERR_MSG(extack, "using device-bound program without HW_MODE flag is not supported");
			bpf_prog_put(prog);
			return -EINVAL;
		}
7216 7217
	}

7218
	err = dev_xdp_install(dev, bpf_op, extack, flags, prog);
7219 7220 7221 7222 7223 7224
	if (err < 0 && prog)
		bpf_prog_put(prog);

	return err;
}

L
Linus Torvalds 已提交
7225 7226
/**
 *	dev_new_index	-	allocate an ifindex
7227
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
7228 7229 7230 7231 7232
 *
 *	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.
 */
7233
static int dev_new_index(struct net *net)
L
Linus Torvalds 已提交
7234
{
7235
	int ifindex = net->ifindex;
7236

L
Linus Torvalds 已提交
7237 7238 7239
	for (;;) {
		if (++ifindex <= 0)
			ifindex = 1;
7240
		if (!__dev_get_by_index(net, ifindex))
7241
			return net->ifindex = ifindex;
L
Linus Torvalds 已提交
7242 7243 7244 7245
	}
}

/* Delayed registration/unregisteration */
7246
static LIST_HEAD(net_todo_list);
7247
DECLARE_WAIT_QUEUE_HEAD(netdev_unregistering_wq);
L
Linus Torvalds 已提交
7248

7249
static void net_set_todo(struct net_device *dev)
L
Linus Torvalds 已提交
7250 7251
{
	list_add_tail(&dev->todo_list, &net_todo_list);
7252
	dev_net(dev)->dev_unreg_count++;
L
Linus Torvalds 已提交
7253 7254
}

7255
static void rollback_registered_many(struct list_head *head)
7256
{
7257
	struct net_device *dev, *tmp;
7258
	LIST_HEAD(close_head);
7259

7260 7261 7262
	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

7263
	list_for_each_entry_safe(dev, tmp, head, unreg_list) {
7264
		/* Some devices call without registering
7265 7266
		 * for initialization unwind. Remove those
		 * devices and proceed with the remaining.
7267 7268
		 */
		if (dev->reg_state == NETREG_UNINITIALIZED) {
7269 7270
			pr_debug("unregister_netdevice: device %s/%p never was registered\n",
				 dev->name, dev);
7271

7272
			WARN_ON(1);
7273 7274
			list_del(&dev->unreg_list);
			continue;
7275
		}
7276
		dev->dismantle = true;
7277
		BUG_ON(dev->reg_state != NETREG_REGISTERED);
7278
	}
7279

7280
	/* If device is running, close it first. */
7281 7282
	list_for_each_entry(dev, head, unreg_list)
		list_add_tail(&dev->close_list, &close_head);
7283
	dev_close_many(&close_head, true);
7284

7285
	list_for_each_entry(dev, head, unreg_list) {
7286 7287
		/* And unlink it from device chain. */
		unlist_netdevice(dev);
7288

7289 7290
		dev->reg_state = NETREG_UNREGISTERING;
	}
7291
	flush_all_backlogs();
7292 7293 7294

	synchronize_net();

7295
	list_for_each_entry(dev, head, unreg_list) {
7296 7297
		struct sk_buff *skb = NULL;

7298 7299
		/* Shutdown queueing discipline. */
		dev_shutdown(dev);
7300

7301
		dev_xdp_uninstall(dev);
7302

7303
		/* Notify protocols, that we are about to destroy
7304 7305
		 * this device. They should clean all the things.
		 */
7306
		call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
7307

7308 7309
		if (!dev->rtnl_link_ops ||
		    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
7310
			skb = rtmsg_ifinfo_build_skb(RTM_DELLINK, dev, ~0U, 0,
7311
						     GFP_KERNEL, NULL);
7312

7313 7314 7315
		/*
		 *	Flush the unicast and multicast chains
		 */
7316
		dev_uc_flush(dev);
7317
		dev_mc_flush(dev);
7318

7319 7320
		if (dev->netdev_ops->ndo_uninit)
			dev->netdev_ops->ndo_uninit(dev);
7321

7322 7323
		if (skb)
			rtmsg_ifinfo_send(skb, dev, GFP_KERNEL);
7324

J
Jiri Pirko 已提交
7325 7326
		/* Notifier chain MUST detach us all upper devices. */
		WARN_ON(netdev_has_any_upper_dev(dev));
7327
		WARN_ON(netdev_has_any_lower_dev(dev));
7328

7329 7330
		/* Remove entries from kobject tree */
		netdev_unregister_kobject(dev);
7331 7332 7333 7334
#ifdef CONFIG_XPS
		/* Remove XPS queueing entries */
		netif_reset_xps_queues_gt(dev, 0);
#endif
7335
	}
7336

7337
	synchronize_net();
7338

7339
	list_for_each_entry(dev, head, unreg_list)
7340 7341 7342 7343 7344 7345 7346 7347 7348
		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
7349
	list_del(&single);
7350 7351
}

7352 7353 7354 7355 7356
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;
7357
	int feature_bit;
7358

7359 7360
	for_each_netdev_feature(&upper_disables, feature_bit) {
		feature = __NETIF_F_BIT(feature_bit);
7361 7362 7363 7364 7365 7366 7367 7368 7369 7370 7371 7372 7373 7374 7375 7376
		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;
7377
	int feature_bit;
7378

7379 7380
	for_each_netdev_feature(&upper_disables, feature_bit) {
		feature = __NETIF_F_BIT(feature_bit);
7381 7382 7383 7384 7385 7386 7387 7388 7389 7390 7391 7392 7393
		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);
		}
	}
}

7394 7395
static netdev_features_t netdev_fix_features(struct net_device *dev,
	netdev_features_t features)
7396
{
7397 7398 7399
	/* Fix illegal checksum combinations */
	if ((features & NETIF_F_HW_CSUM) &&
	    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
7400
		netdev_warn(dev, "mixed HW and IP checksum settings.\n");
7401 7402 7403
		features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
	}

7404
	/* TSO requires that SG is present as well. */
7405
	if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
7406
		netdev_dbg(dev, "Dropping TSO features since no SG feature.\n");
7407
		features &= ~NETIF_F_ALL_TSO;
7408 7409
	}

7410 7411 7412 7413 7414 7415 7416 7417 7418 7419 7420 7421 7422
	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;
	}

7423 7424 7425 7426
	/* 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;

7427 7428 7429 7430
	/* TSO ECN requires that TSO is present as well. */
	if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN)
		features &= ~NETIF_F_TSO_ECN;

7431 7432
	/* Software GSO depends on SG. */
	if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) {
7433
		netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n");
7434 7435 7436
		features &= ~NETIF_F_GSO;
	}

7437 7438 7439 7440 7441 7442 7443 7444
	/* 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;
	}

M
Michael Chan 已提交
7445 7446 7447 7448 7449 7450 7451 7452 7453 7454 7455 7456
	if (!(features & NETIF_F_RXCSUM)) {
		/* NETIF_F_GRO_HW implies doing RXCSUM since every packet
		 * successfully merged by hardware must also have the
		 * checksum verified by hardware.  If the user does not
		 * want to enable RXCSUM, logically, we should disable GRO_HW.
		 */
		if (features & NETIF_F_GRO_HW) {
			netdev_dbg(dev, "Dropping NETIF_F_GRO_HW since no RXCSUM feature.\n");
			features &= ~NETIF_F_GRO_HW;
		}
	}

7457 7458 7459
	return features;
}

7460
int __netdev_update_features(struct net_device *dev)
7461
{
7462
	struct net_device *upper, *lower;
7463
	netdev_features_t features;
7464
	struct list_head *iter;
7465
	int err = -1;
7466

7467 7468
	ASSERT_RTNL();

7469 7470 7471 7472 7473 7474 7475 7476
	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);

7477 7478 7479 7480
	/* 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);

7481
	if (dev->features == features)
7482
		goto sync_lower;
7483

7484 7485
	netdev_dbg(dev, "Features changed: %pNF -> %pNF\n",
		&dev->features, &features);
7486 7487 7488

	if (dev->netdev_ops->ndo_set_features)
		err = dev->netdev_ops->ndo_set_features(dev, features);
7489 7490
	else
		err = 0;
7491

7492
	if (unlikely(err < 0)) {
7493
		netdev_err(dev,
7494 7495
			"set_features() failed (%d); wanted %pNF, left %pNF\n",
			err, &features, &dev->features);
7496 7497 7498 7499
		/* return non-0 since some features might have changed and
		 * it's better to fire a spurious notification than miss it
		 */
		return -1;
7500 7501
	}

7502
sync_lower:
7503 7504 7505 7506 7507 7508
	/* 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);

7509 7510 7511 7512 7513 7514 7515 7516 7517 7518 7519 7520 7521 7522 7523 7524 7525 7526 7527
	if (!err) {
		netdev_features_t diff = features ^ dev->features;

		if (diff & NETIF_F_RX_UDP_TUNNEL_PORT) {
			/* udp_tunnel_{get,drop}_rx_info both need
			 * NETIF_F_RX_UDP_TUNNEL_PORT enabled on the
			 * device, or they won't do anything.
			 * Thus we need to update dev->features
			 * *before* calling udp_tunnel_get_rx_info,
			 * but *after* calling udp_tunnel_drop_rx_info.
			 */
			if (features & NETIF_F_RX_UDP_TUNNEL_PORT) {
				dev->features = features;
				udp_tunnel_get_rx_info(dev);
			} else {
				udp_tunnel_drop_rx_info(dev);
			}
		}

7528
		dev->features = features;
7529
	}
7530

7531
	return err < 0 ? 0 : 1;
7532 7533
}

7534 7535 7536 7537 7538 7539 7540 7541
/**
 *	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.
 */
7542 7543 7544 7545
void netdev_update_features(struct net_device *dev)
{
	if (__netdev_update_features(dev))
		netdev_features_change(dev);
7546 7547 7548
}
EXPORT_SYMBOL(netdev_update_features);

7549 7550 7551 7552 7553 7554 7555 7556 7557 7558 7559 7560 7561 7562 7563 7564 7565
/**
 *	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);

7566 7567 7568 7569 7570 7571 7572 7573 7574 7575 7576 7577 7578 7579 7580 7581 7582
/**
 *	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);

7583 7584 7585 7586
	if (netif_carrier_ok(rootdev))
		netif_carrier_on(dev);
	else
		netif_carrier_off(dev);
7587 7588 7589
}
EXPORT_SYMBOL(netif_stacked_transfer_operstate);

7590
#ifdef CONFIG_SYSFS
7591 7592 7593
static int netif_alloc_rx_queues(struct net_device *dev)
{
	unsigned int i, count = dev->num_rx_queues;
T
Tom Herbert 已提交
7594
	struct netdev_rx_queue *rx;
7595
	size_t sz = count * sizeof(*rx);
7596

T
Tom Herbert 已提交
7597
	BUG_ON(count < 1);
7598

7599
	rx = kvzalloc(sz, GFP_KERNEL | __GFP_RETRY_MAYFAIL);
7600 7601 7602
	if (!rx)
		return -ENOMEM;

T
Tom Herbert 已提交
7603 7604 7605
	dev->_rx = rx;

	for (i = 0; i < count; i++)
T
Tom Herbert 已提交
7606
		rx[i].dev = dev;
7607 7608
	return 0;
}
T
Tom Herbert 已提交
7609
#endif
7610

C
Changli Gao 已提交
7611 7612 7613 7614 7615 7616 7617
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;
7618
	netdev_queue_numa_node_write(queue, NUMA_NO_NODE);
C
Changli Gao 已提交
7619
	queue->dev = dev;
T
Tom Herbert 已提交
7620 7621 7622
#ifdef CONFIG_BQL
	dql_init(&queue->dql, HZ);
#endif
C
Changli Gao 已提交
7623 7624
}

7625 7626
static void netif_free_tx_queues(struct net_device *dev)
{
W
WANG Cong 已提交
7627
	kvfree(dev->_tx);
7628 7629
}

7630 7631 7632 7633
static int netif_alloc_netdev_queues(struct net_device *dev)
{
	unsigned int count = dev->num_tx_queues;
	struct netdev_queue *tx;
7634
	size_t sz = count * sizeof(*tx);
7635

7636 7637
	if (count < 1 || count > 0xffff)
		return -EINVAL;
7638

7639
	tx = kvzalloc(sz, GFP_KERNEL | __GFP_RETRY_MAYFAIL);
7640 7641 7642
	if (!tx)
		return -ENOMEM;

7643
	dev->_tx = tx;
T
Tom Herbert 已提交
7644

7645 7646
	netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
	spin_lock_init(&dev->tx_global_lock);
C
Changli Gao 已提交
7647 7648

	return 0;
7649 7650
}

7651 7652 7653 7654 7655 7656
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);
7657

7658 7659 7660 7661 7662
		netif_tx_stop_queue(txq);
	}
}
EXPORT_SYMBOL(netif_tx_stop_all_queues);

L
Linus Torvalds 已提交
7663 7664 7665 7666 7667 7668 7669 7670 7671 7672 7673 7674 7675 7676 7677 7678 7679 7680 7681 7682
/**
 *	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;
7683
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
7684 7685 7686 7687

	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

7688 7689
	might_sleep();

L
Linus Torvalds 已提交
7690 7691
	/* When net_device's are persistent, this will be fatal. */
	BUG_ON(dev->reg_state != NETREG_UNINITIALIZED);
7692
	BUG_ON(!net);
L
Linus Torvalds 已提交
7693

7694
	spin_lock_init(&dev->addr_list_lock);
7695
	netdev_set_addr_lockdep_class(dev);
L
Linus Torvalds 已提交
7696

7697
	ret = dev_get_valid_name(net, dev, dev->name);
7698 7699 7700
	if (ret < 0)
		goto out;

L
Linus Torvalds 已提交
7701
	/* Init, if this function is available */
7702 7703
	if (dev->netdev_ops->ndo_init) {
		ret = dev->netdev_ops->ndo_init(dev);
L
Linus Torvalds 已提交
7704 7705 7706
		if (ret) {
			if (ret > 0)
				ret = -EIO;
7707
			goto out;
L
Linus Torvalds 已提交
7708 7709
		}
	}
7710

7711 7712
	if (((dev->hw_features | dev->features) &
	     NETIF_F_HW_VLAN_CTAG_FILTER) &&
7713 7714 7715 7716 7717 7718 7719
	    (!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;
	}

7720 7721 7722 7723 7724 7725
	ret = -EBUSY;
	if (!dev->ifindex)
		dev->ifindex = dev_new_index(net);
	else if (__dev_get_by_index(net, dev->ifindex))
		goto err_uninit;

7726 7727 7728 7729
	/* Transfer changeable features to wanted_features and enable
	 * software offloads (GSO and GRO).
	 */
	dev->hw_features |= NETIF_F_SOFT_FEATURES;
7730
	dev->features |= NETIF_F_SOFT_FEATURES;
7731 7732 7733 7734 7735 7736

	if (dev->netdev_ops->ndo_udp_tunnel_add) {
		dev->features |= NETIF_F_RX_UDP_TUNNEL_PORT;
		dev->hw_features |= NETIF_F_RX_UDP_TUNNEL_PORT;
	}

7737
	dev->wanted_features = dev->features & dev->hw_features;
L
Linus Torvalds 已提交
7738

7739
	if (!(dev->flags & IFF_LOOPBACK))
7740
		dev->hw_features |= NETIF_F_NOCACHE_COPY;
7741

7742 7743 7744 7745 7746
	/* 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.
	 */
7747 7748
	if (dev->hw_features & NETIF_F_TSO)
		dev->hw_features |= NETIF_F_TSO_MANGLEID;
7749 7750 7751 7752 7753 7754
	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;
7755

7756
	/* Make NETIF_F_HIGHDMA inheritable to VLAN devices.
7757
	 */
7758
	dev->vlan_features |= NETIF_F_HIGHDMA;
7759

7760 7761
	/* Make NETIF_F_SG inheritable to tunnel devices.
	 */
7762
	dev->hw_enc_features |= NETIF_F_SG | NETIF_F_GSO_PARTIAL;
7763

S
Simon Horman 已提交
7764 7765 7766 7767
	/* Make NETIF_F_SG inheritable to MPLS.
	 */
	dev->mpls_features |= NETIF_F_SG;

7768 7769 7770 7771 7772
	ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		goto err_uninit;

7773
	ret = netdev_register_kobject(dev);
7774
	if (ret)
7775
		goto err_uninit;
7776 7777
	dev->reg_state = NETREG_REGISTERED;

7778
	__netdev_update_features(dev);
7779

L
Linus Torvalds 已提交
7780 7781 7782 7783 7784 7785 7786
	/*
	 *	Default initial state at registry is that the
	 *	device is present.
	 */

	set_bit(__LINK_STATE_PRESENT, &dev->state);

7787 7788
	linkwatch_init_dev(dev);

L
Linus Torvalds 已提交
7789 7790
	dev_init_scheduler(dev);
	dev_hold(dev);
7791
	list_netdevice(dev);
7792
	add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
7793

7794 7795 7796 7797 7798 7799 7800
	/* 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
7801
	/* Notify protocols, that a new device appeared. */
7802
	ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
7803
	ret = notifier_to_errno(ret);
7804 7805 7806 7807
	if (ret) {
		rollback_registered(dev);
		dev->reg_state = NETREG_UNREGISTERED;
	}
7808 7809 7810 7811
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
7812 7813
	if (!dev->rtnl_link_ops ||
	    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
7814
		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
L
Linus Torvalds 已提交
7815 7816 7817

out:
	return ret;
7818 7819

err_uninit:
7820 7821
	if (dev->netdev_ops->ndo_uninit)
		dev->netdev_ops->ndo_uninit(dev);
7822 7823
	if (dev->priv_destructor)
		dev->priv_destructor(dev);
7824
	goto out;
L
Linus Torvalds 已提交
7825
}
E
Eric Dumazet 已提交
7826
EXPORT_SYMBOL(register_netdevice);
L
Linus Torvalds 已提交
7827

7828 7829 7830 7831 7832 7833 7834 7835 7836 7837 7838 7839 7840 7841 7842 7843 7844 7845 7846 7847 7848 7849 7850 7851 7852 7853 7854 7855 7856 7857 7858
/**
 *	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
7859 7860 7861 7862 7863
	/* Note : We dont allocate pcpu_refcnt for dummy devices,
	 * because users of this 'device' dont need to change
	 * its refcount.
	 */

7864 7865 7866 7867 7868
	return 0;
}
EXPORT_SYMBOL_GPL(init_dummy_netdev);


L
Linus Torvalds 已提交
7869 7870 7871 7872 7873 7874 7875 7876 7877
/**
 *	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.
 *
7878
 *	This is a wrapper around register_netdevice that takes the rtnl semaphore
L
Linus Torvalds 已提交
7879 7880 7881 7882 7883 7884 7885 7886 7887 7888 7889 7890 7891 7892
 *	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
7893 7894 7895 7896 7897 7898 7899 7900 7901 7902
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);

7903
/**
L
Linus Torvalds 已提交
7904
 * netdev_wait_allrefs - wait until all references are gone.
7905
 * @dev: target net_device
L
Linus Torvalds 已提交
7906 7907 7908 7909 7910 7911 7912
 *
 * 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
7913
 * call dev_put.
L
Linus Torvalds 已提交
7914 7915 7916 7917
 */
static void netdev_wait_allrefs(struct net_device *dev)
{
	unsigned long rebroadcast_time, warning_time;
E
Eric Dumazet 已提交
7918
	int refcnt;
L
Linus Torvalds 已提交
7919

7920 7921
	linkwatch_forget_dev(dev);

L
Linus Torvalds 已提交
7922
	rebroadcast_time = warning_time = jiffies;
E
Eric Dumazet 已提交
7923 7924 7925
	refcnt = netdev_refcnt_read(dev);

	while (refcnt != 0) {
L
Linus Torvalds 已提交
7926
		if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
7927
			rtnl_lock();
L
Linus Torvalds 已提交
7928 7929

			/* Rebroadcast unregister notification */
7930
			call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
L
Linus Torvalds 已提交
7931

7932
			__rtnl_unlock();
7933
			rcu_barrier();
7934 7935
			rtnl_lock();

7936
			call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
L
Linus Torvalds 已提交
7937 7938 7939 7940 7941 7942 7943 7944 7945 7946 7947
			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();
			}

7948
			__rtnl_unlock();
L
Linus Torvalds 已提交
7949 7950 7951 7952 7953 7954

			rebroadcast_time = jiffies;
		}

		msleep(250);

E
Eric Dumazet 已提交
7955 7956
		refcnt = netdev_refcnt_read(dev);

L
Linus Torvalds 已提交
7957
		if (time_after(jiffies, warning_time + 10 * HZ)) {
7958 7959
			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
				 dev->name, refcnt);
L
Linus Torvalds 已提交
7960 7961 7962 7963 7964 7965 7966 7967 7968 7969 7970 7971 7972 7973 7974 7975 7976 7977 7978
			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
7979
 * We are invoked by rtnl_unlock().
L
Linus Torvalds 已提交
7980
 * This allows us to deal with problems:
7981
 * 1) We can delete sysfs objects which invoke hotplug
L
Linus Torvalds 已提交
7982 7983 7984
 *    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
7985 7986 7987
 *
 * 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
7988 7989 7990
 */
void netdev_run_todo(void)
{
7991
	struct list_head list;
L
Linus Torvalds 已提交
7992 7993

	/* Snapshot list, allow later requests */
7994
	list_replace_init(&net_todo_list, &list);
H
Herbert Xu 已提交
7995 7996

	__rtnl_unlock();
7997

7998 7999

	/* Wait for rcu callbacks to finish before next phase */
8000 8001 8002
	if (!list_empty(&list))
		rcu_barrier();

L
Linus Torvalds 已提交
8003 8004
	while (!list_empty(&list)) {
		struct net_device *dev
8005
			= list_first_entry(&list, struct net_device, todo_list);
L
Linus Torvalds 已提交
8006 8007
		list_del(&dev->todo_list);

8008
		rtnl_lock();
8009
		call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
8010
		__rtnl_unlock();
8011

8012
		if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
8013
			pr_err("network todo '%s' but state %d\n",
8014 8015 8016 8017
			       dev->name, dev->reg_state);
			dump_stack();
			continue;
		}
L
Linus Torvalds 已提交
8018

8019
		dev->reg_state = NETREG_UNREGISTERED;
L
Linus Torvalds 已提交
8020

8021
		netdev_wait_allrefs(dev);
L
Linus Torvalds 已提交
8022

8023
		/* paranoia */
E
Eric Dumazet 已提交
8024
		BUG_ON(netdev_refcnt_read(dev));
8025 8026
		BUG_ON(!list_empty(&dev->ptype_all));
		BUG_ON(!list_empty(&dev->ptype_specific));
8027 8028
		WARN_ON(rcu_access_pointer(dev->ip_ptr));
		WARN_ON(rcu_access_pointer(dev->ip6_ptr));
8029
		WARN_ON(dev->dn_ptr);
L
Linus Torvalds 已提交
8030

8031 8032 8033 8034
		if (dev->priv_destructor)
			dev->priv_destructor(dev);
		if (dev->needs_free_netdev)
			free_netdev(dev);
8035

8036 8037 8038 8039 8040 8041
		/* Report a network device has been unregistered */
		rtnl_lock();
		dev_net(dev)->dev_unreg_count--;
		__rtnl_unlock();
		wake_up(&netdev_unregistering_wq);

8042 8043
		/* Free network device */
		kobject_put(&dev->dev.kobj);
L
Linus Torvalds 已提交
8044 8045 8046
	}
}

8047 8048 8049 8050
/* 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.
8051
 */
8052 8053
void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
			     const struct net_device_stats *netdev_stats)
8054 8055
{
#if BITS_PER_LONG == 64
8056
	BUILD_BUG_ON(sizeof(*stats64) < sizeof(*netdev_stats));
8057
	memcpy(stats64, netdev_stats, sizeof(*netdev_stats));
8058 8059 8060
	/* zero out counters that only exist in rtnl_link_stats64 */
	memset((char *)stats64 + sizeof(*netdev_stats), 0,
	       sizeof(*stats64) - sizeof(*netdev_stats));
8061
#else
8062
	size_t i, n = sizeof(*netdev_stats) / sizeof(unsigned long);
8063 8064 8065
	const unsigned long *src = (const unsigned long *)netdev_stats;
	u64 *dst = (u64 *)stats64;

8066
	BUILD_BUG_ON(n > sizeof(*stats64) / sizeof(u64));
8067 8068
	for (i = 0; i < n; i++)
		dst[i] = src[i];
8069 8070 8071
	/* zero out counters that only exist in rtnl_link_stats64 */
	memset((char *)stats64 + n * sizeof(u64), 0,
	       sizeof(*stats64) - n * sizeof(u64));
8072 8073
#endif
}
8074
EXPORT_SYMBOL(netdev_stats_to_stats64);
8075

8076 8077 8078
/**
 *	dev_get_stats	- get network device statistics
 *	@dev: device to get statistics from
8079
 *	@storage: place to store stats
8080
 *
8081 8082 8083 8084
 *	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.
8085
 */
8086 8087
struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
					struct rtnl_link_stats64 *storage)
8088
{
8089 8090
	const struct net_device_ops *ops = dev->netdev_ops;

8091 8092
	if (ops->ndo_get_stats64) {
		memset(storage, 0, sizeof(*storage));
8093 8094
		ops->ndo_get_stats64(dev, storage);
	} else if (ops->ndo_get_stats) {
8095
		netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev));
8096 8097
	} else {
		netdev_stats_to_stats64(storage, &dev->stats);
8098
	}
8099 8100 8101
	storage->rx_dropped += (unsigned long)atomic_long_read(&dev->rx_dropped);
	storage->tx_dropped += (unsigned long)atomic_long_read(&dev->tx_dropped);
	storage->rx_nohandler += (unsigned long)atomic_long_read(&dev->rx_nohandler);
8102
	return storage;
R
Rusty Russell 已提交
8103
}
8104
EXPORT_SYMBOL(dev_get_stats);
R
Rusty Russell 已提交
8105

8106
struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
8107
{
8108
	struct netdev_queue *queue = dev_ingress_queue(dev);
8109

8110 8111 8112 8113 8114 8115 8116
#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
8117
	RCU_INIT_POINTER(queue->qdisc, &noop_qdisc);
8118 8119 8120 8121
	queue->qdisc_sleeping = &noop_qdisc;
	rcu_assign_pointer(dev->ingress_queue, queue);
#endif
	return queue;
8122 8123
}

8124 8125
static const struct ethtool_ops default_ethtool_ops;

8126 8127 8128 8129 8130 8131 8132 8133
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);

8134 8135 8136 8137
void netdev_freemem(struct net_device *dev)
{
	char *addr = (char *)dev - dev->padded;

W
WANG Cong 已提交
8138
	kvfree(addr);
8139 8140
}

L
Linus Torvalds 已提交
8141
/**
8142 8143 8144 8145 8146 8147 8148 8149 8150 8151 8152
 * alloc_netdev_mqs - allocate network device
 * @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
 *
 * Allocates a struct net_device with private data area for driver use
 * and performs basic initialization.  Also allocates subqueue structs
 * for each queue on the device.
L
Linus Torvalds 已提交
8153
 */
T
Tom Herbert 已提交
8154
struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
8155
		unsigned char name_assign_type,
T
Tom Herbert 已提交
8156 8157
		void (*setup)(struct net_device *),
		unsigned int txqs, unsigned int rxqs)
L
Linus Torvalds 已提交
8158 8159
{
	struct net_device *dev;
8160
	unsigned int alloc_size;
8161
	struct net_device *p;
L
Linus Torvalds 已提交
8162

8163 8164
	BUG_ON(strlen(name) >= sizeof(dev->name));

T
Tom Herbert 已提交
8165
	if (txqs < 1) {
8166
		pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
8167 8168 8169
		return NULL;
	}

8170
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
8171
	if (rxqs < 1) {
8172
		pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
T
Tom Herbert 已提交
8173 8174 8175 8176
		return NULL;
	}
#endif

8177
	alloc_size = sizeof(struct net_device);
8178 8179
	if (sizeof_priv) {
		/* ensure 32-byte alignment of private area */
8180
		alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
8181 8182 8183
		alloc_size += sizeof_priv;
	}
	/* ensure 32-byte alignment of whole construct */
8184
	alloc_size += NETDEV_ALIGN - 1;
L
Linus Torvalds 已提交
8185

8186
	p = kvzalloc(alloc_size, GFP_KERNEL | __GFP_RETRY_MAYFAIL);
8187
	if (!p)
L
Linus Torvalds 已提交
8188 8189
		return NULL;

8190
	dev = PTR_ALIGN(p, NETDEV_ALIGN);
L
Linus Torvalds 已提交
8191
	dev->padded = (char *)dev - (char *)p;
8192

E
Eric Dumazet 已提交
8193 8194
	dev->pcpu_refcnt = alloc_percpu(int);
	if (!dev->pcpu_refcnt)
8195
		goto free_dev;
8196 8197

	if (dev_addr_init(dev))
E
Eric Dumazet 已提交
8198
		goto free_pcpu;
8199

8200
	dev_mc_init(dev);
8201
	dev_uc_init(dev);
J
Jiri Pirko 已提交
8202

8203
	dev_net_set(dev, &init_net);
L
Linus Torvalds 已提交
8204

8205
	dev->gso_max_size = GSO_MAX_SIZE;
8206
	dev->gso_max_segs = GSO_MAX_SEGS;
8207 8208 8209

	INIT_LIST_HEAD(&dev->napi_list);
	INIT_LIST_HEAD(&dev->unreg_list);
8210
	INIT_LIST_HEAD(&dev->close_list);
8211
	INIT_LIST_HEAD(&dev->link_watch_list);
8212 8213
	INIT_LIST_HEAD(&dev->adj_list.upper);
	INIT_LIST_HEAD(&dev->adj_list.lower);
8214 8215
	INIT_LIST_HEAD(&dev->ptype_all);
	INIT_LIST_HEAD(&dev->ptype_specific);
8216 8217 8218
#ifdef CONFIG_NET_SCHED
	hash_init(dev->qdisc_hash);
#endif
8219
	dev->priv_flags = IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM;
8220 8221
	setup(dev);

8222
	if (!dev->tx_queue_len) {
8223
		dev->priv_flags |= IFF_NO_QUEUE;
8224
		dev->tx_queue_len = DEFAULT_TX_QUEUE_LEN;
8225
	}
8226

T
Tom Herbert 已提交
8227 8228
	dev->num_tx_queues = txqs;
	dev->real_num_tx_queues = txqs;
8229
	if (netif_alloc_netdev_queues(dev))
8230
		goto free_all;
8231

8232
#ifdef CONFIG_SYSFS
T
Tom Herbert 已提交
8233 8234
	dev->num_rx_queues = rxqs;
	dev->real_num_rx_queues = rxqs;
T
Tom Herbert 已提交
8235
	if (netif_alloc_rx_queues(dev))
8236
		goto free_all;
E
Eric Dumazet 已提交
8237
#endif
T
Tom Herbert 已提交
8238

L
Linus Torvalds 已提交
8239
	strcpy(dev->name, name);
8240
	dev->name_assign_type = name_assign_type;
8241
	dev->group = INIT_NETDEV_GROUP;
8242 8243
	if (!dev->ethtool_ops)
		dev->ethtool_ops = &default_ethtool_ops;
8244 8245 8246

	nf_hook_ingress_init(dev);

L
Linus Torvalds 已提交
8247
	return dev;
8248

8249 8250 8251 8252
free_all:
	free_netdev(dev);
	return NULL;

E
Eric Dumazet 已提交
8253 8254
free_pcpu:
	free_percpu(dev->pcpu_refcnt);
8255 8256
free_dev:
	netdev_freemem(dev);
8257
	return NULL;
L
Linus Torvalds 已提交
8258
}
T
Tom Herbert 已提交
8259
EXPORT_SYMBOL(alloc_netdev_mqs);
L
Linus Torvalds 已提交
8260 8261

/**
8262 8263
 * free_netdev - free network device
 * @dev: device
L
Linus Torvalds 已提交
8264
 *
8265 8266 8267 8268
 * This function does the last stage of destroying an allocated device
 * interface. The reference to the device object is released. If this
 * is the last reference then it will be freed.Must be called in process
 * context.
L
Linus Torvalds 已提交
8269 8270 8271
 */
void free_netdev(struct net_device *dev)
{
8272 8273
	struct napi_struct *p, *n;

8274
	might_sleep();
8275
	netif_free_tx_queues(dev);
8276
#ifdef CONFIG_SYSFS
8277
	kvfree(dev->_rx);
T
Tom Herbert 已提交
8278
#endif
8279

8280
	kfree(rcu_dereference_protected(dev->ingress_queue, 1));
8281

8282 8283 8284
	/* Flush device addresses */
	dev_addr_flush(dev);

8285 8286 8287
	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
		netif_napi_del(p);

E
Eric Dumazet 已提交
8288 8289 8290
	free_percpu(dev->pcpu_refcnt);
	dev->pcpu_refcnt = NULL;

S
Stephen Hemminger 已提交
8291
	/*  Compatibility with error handling in drivers */
L
Linus Torvalds 已提交
8292
	if (dev->reg_state == NETREG_UNINITIALIZED) {
8293
		netdev_freemem(dev);
L
Linus Torvalds 已提交
8294 8295 8296 8297 8298 8299
		return;
	}

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

8300 8301
	/* will free via device release */
	put_device(&dev->dev);
L
Linus Torvalds 已提交
8302
}
E
Eric Dumazet 已提交
8303
EXPORT_SYMBOL(free_netdev);
8304

8305 8306 8307 8308 8309 8310
/**
 *	synchronize_net -  Synchronize with packet receive processing
 *
 *	Wait for packets currently being received to be done.
 *	Does not block later packets from starting.
 */
8311
void synchronize_net(void)
L
Linus Torvalds 已提交
8312 8313
{
	might_sleep();
8314 8315 8316 8317
	if (rtnl_is_locked())
		synchronize_rcu_expedited();
	else
		synchronize_rcu();
L
Linus Torvalds 已提交
8318
}
E
Eric Dumazet 已提交
8319
EXPORT_SYMBOL(synchronize_net);
L
Linus Torvalds 已提交
8320 8321

/**
8322
 *	unregister_netdevice_queue - remove device from the kernel
L
Linus Torvalds 已提交
8323
 *	@dev: device
8324
 *	@head: list
8325
 *
L
Linus Torvalds 已提交
8326
 *	This function shuts down a device interface and removes it
8327
 *	from the kernel tables.
8328
 *	If head not NULL, device is queued to be unregistered later.
L
Linus Torvalds 已提交
8329 8330 8331 8332 8333
 *
 *	Callers must hold the rtnl semaphore.  You may want
 *	unregister_netdev() instead of this.
 */

8334
void unregister_netdevice_queue(struct net_device *dev, struct list_head *head)
L
Linus Torvalds 已提交
8335
{
8336 8337
	ASSERT_RTNL();

8338
	if (head) {
8339
		list_move_tail(&dev->unreg_list, head);
8340 8341 8342 8343 8344
	} else {
		rollback_registered(dev);
		/* Finish processing unregister after unlock */
		net_set_todo(dev);
	}
L
Linus Torvalds 已提交
8345
}
8346
EXPORT_SYMBOL(unregister_netdevice_queue);
L
Linus Torvalds 已提交
8347

8348 8349 8350
/**
 *	unregister_netdevice_many - unregister many devices
 *	@head: list of devices
8351 8352 8353
 *
 *  Note: As most callers use a stack allocated list_head,
 *  we force a list_del() to make sure stack wont be corrupted later.
8354 8355 8356 8357 8358 8359 8360 8361 8362
 */
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);
8363
		list_del(head);
8364 8365
	}
}
8366
EXPORT_SYMBOL(unregister_netdevice_many);
8367

L
Linus Torvalds 已提交
8368 8369 8370 8371 8372
/**
 *	unregister_netdev - remove device from the kernel
 *	@dev: device
 *
 *	This function shuts down a device interface and removes it
8373
 *	from the kernel tables.
L
Linus Torvalds 已提交
8374 8375 8376 8377 8378 8379 8380 8381 8382 8383 8384 8385 8386
 *
 *	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);

8387 8388 8389 8390 8391 8392 8393 8394 8395 8396 8397 8398 8399 8400 8401 8402
/**
 *	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)
{
8403
	int err, new_nsid;
8404 8405 8406 8407 8408 8409 8410 8411 8412 8413 8414 8415 8416 8417

	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;
8418
	if (net_eq(dev_net(dev), net))
8419 8420 8421 8422 8423 8424
		goto out;

	/* Pick the destination device name, and ensure
	 * we can use it in the destination network namespace.
	 */
	err = -EEXIST;
8425
	if (__dev_get_by_name(net, dev->name)) {
8426 8427 8428
		/* We get here if we can't use the current device name */
		if (!pat)
			goto out;
8429
		if (dev_get_valid_name(net, dev, pat) < 0)
8430 8431 8432 8433 8434 8435 8436 8437
			goto out;
	}

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

	/* If device is running close it first. */
8438
	dev_close(dev);
8439 8440 8441 8442 8443 8444 8445 8446 8447 8448 8449

	/* 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
8450 8451 8452 8453 8454 8455
	 * this device. They should clean all the things.
	 *
	 * 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.
	 */
8456
	call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
8457 8458
	rcu_barrier();
	call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
8459 8460 8461 8462 8463
	if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net)
		new_nsid = peernet2id_alloc(dev_net(dev), net);
	else
		new_nsid = peernet2id(dev_net(dev), net);
	rtmsg_ifinfo_newnet(RTM_DELLINK, dev, ~0U, GFP_KERNEL, &new_nsid);
8464 8465 8466 8467

	/*
	 *	Flush the unicast and multicast chains
	 */
8468
	dev_uc_flush(dev);
8469
	dev_mc_flush(dev);
8470

8471 8472
	/* Send a netdev-removed uevent to the old namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_REMOVE);
8473
	netdev_adjacent_del_links(dev);
8474

8475
	/* Actually switch the network namespace */
8476
	dev_net_set(dev, net);
8477 8478

	/* If there is an ifindex conflict assign a new one */
8479
	if (__dev_get_by_index(net, dev->ifindex))
8480 8481
		dev->ifindex = dev_new_index(net);

8482 8483
	/* Send a netdev-add uevent to the new namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_ADD);
8484
	netdev_adjacent_add_links(dev);
8485

8486
	/* Fixup kobjects */
8487
	err = device_rename(&dev->dev, dev->name);
8488
	WARN_ON(err);
8489 8490 8491 8492 8493 8494 8495

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

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

8496 8497 8498 8499
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
8500
	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U, GFP_KERNEL);
8501

8502 8503 8504 8505 8506
	synchronize_net();
	err = 0;
out:
	return err;
}
8507
EXPORT_SYMBOL_GPL(dev_change_net_namespace);
8508

8509
static int dev_cpu_dead(unsigned int oldcpu)
L
Linus Torvalds 已提交
8510 8511 8512
{
	struct sk_buff **list_skb;
	struct sk_buff *skb;
8513
	unsigned int cpu;
8514
	struct softnet_data *sd, *oldsd, *remsd = NULL;
L
Linus Torvalds 已提交
8515 8516 8517 8518 8519 8520 8521 8522 8523 8524 8525 8526 8527 8528 8529

	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. */
8530 8531 8532 8533 8534 8535
	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
8536 8537 8538 8539 8540 8541 8542 8543 8544 8545 8546 8547 8548 8549
	/* 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);
8550
	}
L
Linus Torvalds 已提交
8551 8552 8553 8554

	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_enable();

8555 8556 8557 8558 8559 8560 8561
#ifdef CONFIG_RPS
	remsd = oldsd->rps_ipi_list;
	oldsd->rps_ipi_list = NULL;
#endif
	/* send out pending IPI's on offline CPU */
	net_rps_send_ipi(remsd);

L
Linus Torvalds 已提交
8562
	/* Process offline CPU's input_pkt_queue */
8563
	while ((skb = __skb_dequeue(&oldsd->process_queue))) {
8564
		netif_rx_ni(skb);
8565
		input_queue_head_incr(oldsd);
T
Tom Herbert 已提交
8566
	}
E
Eric Dumazet 已提交
8567
	while ((skb = skb_dequeue(&oldsd->input_pkt_queue))) {
8568
		netif_rx_ni(skb);
8569 8570
		input_queue_head_incr(oldsd);
	}
L
Linus Torvalds 已提交
8571

8572
	return 0;
L
Linus Torvalds 已提交
8573 8574
}

8575
/**
8576 8577 8578 8579
 *	netdev_increment_features - increment feature set by one
 *	@all: current feature set
 *	@one: new feature set
 *	@mask: mask feature set
8580 8581
 *
 *	Computes a new feature set after adding a device with feature set
8582 8583
 *	@one to the master device with current feature set @all.  Will not
 *	enable anything that is off in @mask. Returns the new feature set.
8584
 */
8585 8586
netdev_features_t netdev_increment_features(netdev_features_t all,
	netdev_features_t one, netdev_features_t mask)
8587
{
8588
	if (mask & NETIF_F_HW_CSUM)
8589
		mask |= NETIF_F_CSUM_MASK;
8590
	mask |= NETIF_F_VLAN_CHALLENGED;
8591

8592
	all |= one & (NETIF_F_ONE_FOR_ALL | NETIF_F_CSUM_MASK) & mask;
8593
	all &= one | ~NETIF_F_ALL_FOR_ALL;
8594

8595
	/* If one device supports hw checksumming, set for all. */
8596 8597
	if (all & NETIF_F_HW_CSUM)
		all &= ~(NETIF_F_CSUM_MASK & ~NETIF_F_HW_CSUM);
8598 8599 8600

	return all;
}
8601
EXPORT_SYMBOL(netdev_increment_features);
8602

8603
static struct hlist_head * __net_init netdev_create_hash(void)
8604 8605 8606 8607 8608 8609 8610 8611 8612 8613 8614 8615
{
	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;
}

8616
/* Initialize per network namespace state */
8617
static int __net_init netdev_init(struct net *net)
8618
{
8619 8620
	if (net != &init_net)
		INIT_LIST_HEAD(&net->dev_base_head);
8621

8622 8623 8624
	net->dev_name_head = netdev_create_hash();
	if (net->dev_name_head == NULL)
		goto err_name;
8625

8626 8627 8628
	net->dev_index_head = netdev_create_hash();
	if (net->dev_index_head == NULL)
		goto err_idx;
8629 8630

	return 0;
8631 8632 8633 8634 8635

err_idx:
	kfree(net->dev_name_head);
err_name:
	return -ENOMEM;
8636 8637
}

8638 8639 8640 8641 8642 8643
/**
 *	netdev_drivername - network driver for the device
 *	@dev: network device
 *
 *	Determine network driver for device.
 */
8644
const char *netdev_drivername(const struct net_device *dev)
8645
{
8646 8647
	const struct device_driver *driver;
	const struct device *parent;
8648
	const char *empty = "";
8649 8650 8651

	parent = dev->dev.parent;
	if (!parent)
8652
		return empty;
8653 8654 8655

	driver = parent->driver;
	if (driver && driver->name)
8656 8657
		return driver->name;
	return empty;
8658 8659
}

8660 8661
static void __netdev_printk(const char *level, const struct net_device *dev,
			    struct va_format *vaf)
8662
{
8663
	if (dev && dev->dev.parent) {
8664 8665 8666 8667 8668 8669 8670
		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);
8671
	} else if (dev) {
8672 8673
		printk("%s%s%s: %pV",
		       level, netdev_name(dev), netdev_reg_state(dev), vaf);
8674
	} else {
8675
		printk("%s(NULL net_device): %pV", level, vaf);
8676
	}
8677 8678
}

8679 8680
void netdev_printk(const char *level, const struct net_device *dev,
		   const char *format, ...)
8681 8682 8683 8684 8685 8686 8687 8688 8689
{
	struct va_format vaf;
	va_list args;

	va_start(args, format);

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

8690
	__netdev_printk(level, dev, &vaf);
8691

8692 8693 8694 8695 8696
	va_end(args);
}
EXPORT_SYMBOL(netdev_printk);

#define define_netdev_printk_level(func, level)			\
8697
void func(const struct net_device *dev, const char *fmt, ...)	\
8698 8699 8700 8701 8702 8703 8704 8705 8706
{								\
	struct va_format vaf;					\
	va_list args;						\
								\
	va_start(args, fmt);					\
								\
	vaf.fmt = fmt;						\
	vaf.va = &args;						\
								\
8707
	__netdev_printk(level, dev, &vaf);			\
8708
								\
8709 8710 8711 8712 8713 8714 8715 8716 8717 8718 8719 8720
	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);

8721
static void __net_exit netdev_exit(struct net *net)
8722 8723 8724
{
	kfree(net->dev_name_head);
	kfree(net->dev_index_head);
8725 8726
	if (net != &init_net)
		WARN_ON_ONCE(!list_empty(&net->dev_base_head));
8727 8728
}

8729
static struct pernet_operations __net_initdata netdev_net_ops = {
8730 8731 8732 8733
	.init = netdev_init,
	.exit = netdev_exit,
};

8734
static void __net_exit default_device_exit(struct net *net)
8735
{
8736
	struct net_device *dev, *aux;
8737
	/*
8738
	 * Push all migratable network devices back to the
8739 8740 8741
	 * initial network namespace
	 */
	rtnl_lock();
8742
	for_each_netdev_safe(net, dev, aux) {
8743
		int err;
8744
		char fb_name[IFNAMSIZ];
8745 8746 8747 8748 8749

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

8750 8751 8752
		/* Leave virtual devices for the generic cleanup */
		if (dev->rtnl_link_ops)
			continue;
8753

L
Lucas De Marchi 已提交
8754
		/* Push remaining network devices to init_net */
8755 8756
		snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex);
		err = dev_change_net_namespace(dev, &init_net, fb_name);
8757
		if (err) {
8758 8759
			pr_emerg("%s: failed to move %s to init_net: %d\n",
				 __func__, dev->name, err);
8760
			BUG();
8761 8762 8763 8764 8765
		}
	}
	rtnl_unlock();
}

8766 8767 8768 8769 8770 8771 8772
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;
8773
	DEFINE_WAIT_FUNC(wait, woken_wake_function);
8774

8775
	add_wait_queue(&netdev_unregistering_wq, &wait);
8776 8777 8778 8779 8780 8781 8782 8783 8784 8785 8786 8787
	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();
8788 8789

		wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
8790
	}
8791
	remove_wait_queue(&netdev_unregistering_wq, &wait);
8792 8793
}

8794 8795 8796
static void __net_exit default_device_exit_batch(struct list_head *net_list)
{
	/* At exit all network devices most be removed from a network
8797
	 * namespace.  Do this in the reverse order of registration.
8798 8799 8800 8801 8802 8803 8804
	 * 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);

8805 8806 8807 8808 8809 8810 8811 8812 8813 8814 8815 8816
	/* 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);
8817 8818
	list_for_each_entry(net, net_list, exit_list) {
		for_each_netdev_reverse(net, dev) {
8819
			if (dev->rtnl_link_ops && dev->rtnl_link_ops->dellink)
8820 8821 8822 8823 8824 8825 8826 8827 8828
				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();
}

8829
static struct pernet_operations __net_initdata default_device_ops = {
8830
	.exit = default_device_exit,
8831
	.exit_batch = default_device_exit_batch,
8832 8833
};

L
Linus Torvalds 已提交
8834 8835 8836 8837 8838 8839 8840 8841 8842 8843 8844 8845 8846 8847 8848 8849 8850 8851 8852 8853
/*
 *	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;

8854
	if (netdev_kobject_init())
L
Linus Torvalds 已提交
8855 8856 8857
		goto out;

	INIT_LIST_HEAD(&ptype_all);
8858
	for (i = 0; i < PTYPE_HASH_SIZE; i++)
L
Linus Torvalds 已提交
8859 8860
		INIT_LIST_HEAD(&ptype_base[i]);

8861 8862
	INIT_LIST_HEAD(&offload_base);

8863 8864
	if (register_pernet_subsys(&netdev_net_ops))
		goto out;
L
Linus Torvalds 已提交
8865 8866 8867 8868 8869

	/*
	 *	Initialise the packet receive queues.
	 */

8870
	for_each_possible_cpu(i) {
8871
		struct work_struct *flush = per_cpu_ptr(&flush_works, i);
E
Eric Dumazet 已提交
8872
		struct softnet_data *sd = &per_cpu(softnet_data, i);
L
Linus Torvalds 已提交
8873

8874 8875
		INIT_WORK(flush, flush_backlog);

E
Eric Dumazet 已提交
8876
		skb_queue_head_init(&sd->input_pkt_queue);
8877
		skb_queue_head_init(&sd->process_queue);
E
Eric Dumazet 已提交
8878
		INIT_LIST_HEAD(&sd->poll_list);
8879
		sd->output_queue_tailp = &sd->output_queue;
E
Eric Dumazet 已提交
8880
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
8881 8882 8883
		sd->csd.func = rps_trigger_softirq;
		sd->csd.info = sd;
		sd->cpu = i;
8884
#endif
T
Tom Herbert 已提交
8885

E
Eric Dumazet 已提交
8886 8887
		sd->backlog.poll = process_backlog;
		sd->backlog.weight = weight_p;
L
Linus Torvalds 已提交
8888 8889 8890 8891
	}

	dev_boot_phase = 0;

8892 8893 8894 8895 8896 8897 8898 8899 8900 8901 8902 8903 8904 8905 8906
	/* 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;

8907 8908
	open_softirq(NET_TX_SOFTIRQ, net_tx_action);
	open_softirq(NET_RX_SOFTIRQ, net_rx_action);
L
Linus Torvalds 已提交
8909

8910 8911 8912
	rc = cpuhp_setup_state_nocalls(CPUHP_NET_DEV_DEAD, "net/dev:dead",
				       NULL, dev_cpu_dead);
	WARN_ON(rc < 0);
L
Linus Torvalds 已提交
8913 8914 8915 8916 8917 8918
	rc = 0;
out:
	return rc;
}

subsys_initcall(net_dev_init);