net: deinline netif_tx_stop_all_queues(), remove WARN_ON in netif_tx_stop_queue()
[deliverable/linux.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
e8db0be1 28#include <linux/pm_qos.h>
d7fe0f24 29#include <linux/timer.h>
187f1882 30#include <linux/bug.h>
bea3348e 31#include <linux/delay.h>
60063497 32#include <linux/atomic.h>
53511453 33#include <linux/prefetch.h>
1da177e4
LT
34#include <asm/cache.h>
35#include <asm/byteorder.h>
36
1da177e4 37#include <linux/percpu.h>
4d5b78c0 38#include <linux/rculist.h>
db217334 39#include <linux/dmaengine.h>
bea3348e 40#include <linux/workqueue.h>
114cf580 41#include <linux/dynamic_queue_limits.h>
1da177e4 42
b1b67dd4 43#include <linux/ethtool.h>
a050c33f 44#include <net/net_namespace.h>
cf85d08f 45#include <net/dsa.h>
7a6b6f51 46#ifdef CONFIG_DCB
2f90b865
AD
47#include <net/dcbnl.h>
48#endif
5bc1421e 49#include <net/netprio_cgroup.h>
a050c33f 50
a59e2ecb 51#include <linux/netdev_features.h>
77162022 52#include <linux/neighbour.h>
607ca46e 53#include <uapi/linux/netdevice.h>
61bd3857 54#include <uapi/linux/if_bonding.h>
a59e2ecb 55
115c1d6e 56struct netpoll_info;
313162d0 57struct device;
c1f19b51 58struct phy_device;
704232c2
JB
59/* 802.11 specific */
60struct wireless_dev;
98a18b6f
AA
61/* 802.15.4 specific */
62struct wpan_dev;
03c57747 63struct mpls_dev;
1da177e4 64
f629d208
JP
65void netdev_set_default_ethtool_ops(struct net_device *dev,
66 const struct ethtool_ops *ops);
d07d7507 67
9a1654ba
JP
68/* Backlog congestion levels */
69#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
70#define NET_RX_DROP 1 /* packet dropped */
71
572a9d7b
PM
72/*
73 * Transmit return codes: transmit return codes originate from three different
74 * namespaces:
75 *
76 * - qdisc return codes
77 * - driver transmit return codes
78 * - errno values
79 *
80 * Drivers are allowed to return any one of those in their hard_start_xmit()
81 * function. Real network devices commonly used with qdiscs should only return
82 * the driver transmit return codes though - when qdiscs are used, the actual
83 * transmission happens asynchronously, so the value is not propagated to
84 * higher layers. Virtual network devices transmit synchronously, in this case
85 * the driver transmit return codes are consumed by dev_queue_xmit(), all
86 * others are propagated to higher layers.
87 */
88
89/* qdisc ->enqueue() return codes. */
90#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
91#define NET_XMIT_DROP 0x01 /* skb dropped */
92#define NET_XMIT_CN 0x02 /* congestion notification */
93#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
94#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 95
b9df3cb8
GR
96/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
97 * indicates that the device will soon be dropping packets, or already drops
98 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 99#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
100#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101
dc1f8bf6 102/* Driver transmit return codes */
9a1654ba 103#define NETDEV_TX_MASK 0xf0
572a9d7b 104
dc1f8bf6 105enum netdev_tx {
572a9d7b 106 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
107 NETDEV_TX_OK = 0x00, /* driver took care of packet */
108 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
109 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
dc1f8bf6
SH
110};
111typedef enum netdev_tx netdev_tx_t;
112
9a1654ba
JP
113/*
114 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
115 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 */
117static inline bool dev_xmit_complete(int rc)
118{
119 /*
120 * Positive cases with an skb consumed by a driver:
121 * - successful transmission (rc == NETDEV_TX_OK)
122 * - error while transmitting (rc < 0)
123 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 */
125 if (likely(rc < NET_XMIT_MASK))
126 return true;
127
128 return false;
129}
130
1da177e4
LT
131/*
132 * Compute the worst case header length according to the protocols
133 * used.
134 */
fe2918b0 135
d11ead75 136#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
137# if defined(CONFIG_MAC80211_MESH)
138# define LL_MAX_HEADER 128
139# else
140# define LL_MAX_HEADER 96
141# endif
1da177e4 142#else
8388e3da 143# define LL_MAX_HEADER 32
1da177e4
LT
144#endif
145
d11ead75
BH
146#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
147 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
148#define MAX_HEADER LL_MAX_HEADER
149#else
150#define MAX_HEADER (LL_MAX_HEADER + 48)
151#endif
152
153/*
be1f3c2c
BH
154 * Old network device statistics. Fields are native words
155 * (unsigned long) so they can be read and written atomically.
1da177e4 156 */
fe2918b0 157
d94d9fee 158struct net_device_stats {
3cfde79c
BH
159 unsigned long rx_packets;
160 unsigned long tx_packets;
161 unsigned long rx_bytes;
162 unsigned long tx_bytes;
163 unsigned long rx_errors;
164 unsigned long tx_errors;
165 unsigned long rx_dropped;
166 unsigned long tx_dropped;
167 unsigned long multicast;
1da177e4 168 unsigned long collisions;
1da177e4 169 unsigned long rx_length_errors;
3cfde79c
BH
170 unsigned long rx_over_errors;
171 unsigned long rx_crc_errors;
172 unsigned long rx_frame_errors;
173 unsigned long rx_fifo_errors;
174 unsigned long rx_missed_errors;
1da177e4
LT
175 unsigned long tx_aborted_errors;
176 unsigned long tx_carrier_errors;
177 unsigned long tx_fifo_errors;
178 unsigned long tx_heartbeat_errors;
179 unsigned long tx_window_errors;
1da177e4
LT
180 unsigned long rx_compressed;
181 unsigned long tx_compressed;
182};
183
1da177e4
LT
184
185#include <linux/cache.h>
186#include <linux/skbuff.h>
187
adc9300e 188#ifdef CONFIG_RPS
c5905afb
IM
189#include <linux/static_key.h>
190extern struct static_key rps_needed;
adc9300e
ED
191#endif
192
1da177e4
LT
193struct neighbour;
194struct neigh_parms;
195struct sk_buff;
196
f001fde5
JP
197struct netdev_hw_addr {
198 struct list_head list;
199 unsigned char addr[MAX_ADDR_LEN];
200 unsigned char type;
ccffad25
JP
201#define NETDEV_HW_ADDR_T_LAN 1
202#define NETDEV_HW_ADDR_T_SAN 2
203#define NETDEV_HW_ADDR_T_SLAVE 3
204#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 205#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 206 bool global_use;
4cd729b0 207 int sync_cnt;
8f8f103d 208 int refcount;
4543fbef 209 int synced;
f001fde5
JP
210 struct rcu_head rcu_head;
211};
212
31278e71
JP
213struct netdev_hw_addr_list {
214 struct list_head list;
215 int count;
216};
217
22bedad3
JP
218#define netdev_hw_addr_list_count(l) ((l)->count)
219#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
220#define netdev_hw_addr_list_for_each(ha, l) \
221 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 222
22bedad3
JP
223#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
224#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
225#define netdev_for_each_uc_addr(ha, dev) \
226 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 227
22bedad3
JP
228#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
229#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 230#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 231 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 232
d94d9fee 233struct hh_cache {
f6b72b62 234 u16 hh_len;
5c25f686 235 u16 __pad;
3644f0ce 236 seqlock_t hh_lock;
1da177e4
LT
237
238 /* cached hardware header; allow for machine alignment needs. */
239#define HH_DATA_MOD 16
240#define HH_DATA_OFF(__len) \
5ba0eac6 241 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
242#define HH_DATA_ALIGN(__len) \
243 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
244 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
245};
246
247/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
248 * Alternative is:
249 * dev->hard_header_len ? (dev->hard_header_len +
250 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
251 *
252 * We could use other alignment values, but we must maintain the
253 * relationship HH alignment <= LL alignment.
254 */
255#define LL_RESERVED_SPACE(dev) \
f5184d26 256 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 257#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 258 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 259
3b04ddde
SH
260struct header_ops {
261 int (*create) (struct sk_buff *skb, struct net_device *dev,
262 unsigned short type, const void *daddr,
95c96174 263 const void *saddr, unsigned int len);
3b04ddde 264 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 265 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
266 void (*cache_update)(struct hh_cache *hh,
267 const struct net_device *dev,
268 const unsigned char *haddr);
269};
270
1da177e4
LT
271/* These flag bits are private to the generic network queueing
272 * layer, they may not be explicitly referenced by any other
273 * code.
274 */
275
d94d9fee 276enum netdev_state_t {
1da177e4
LT
277 __LINK_STATE_START,
278 __LINK_STATE_PRESENT,
1da177e4 279 __LINK_STATE_NOCARRIER,
b00055aa
SR
280 __LINK_STATE_LINKWATCH_PENDING,
281 __LINK_STATE_DORMANT,
1da177e4
LT
282};
283
284
285/*
286 * This structure holds at boot time configured netdevice settings. They
fe2918b0 287 * are then used in the device probing.
1da177e4
LT
288 */
289struct netdev_boot_setup {
290 char name[IFNAMSIZ];
291 struct ifmap map;
292};
293#define NETDEV_BOOT_SETUP_MAX 8
294
f629d208 295int __init netdev_boot_setup(char *str);
1da177e4 296
bea3348e
SH
297/*
298 * Structure for NAPI scheduling similar to tasklet but with weighting
299 */
300struct napi_struct {
301 /* The poll_list must only be managed by the entity which
302 * changes the state of the NAPI_STATE_SCHED bit. This means
303 * whoever atomically sets that bit can add this napi_struct
304 * to the per-cpu poll_list, and whoever clears that bit
305 * can remove from the list right before clearing the bit.
306 */
307 struct list_head poll_list;
308
309 unsigned long state;
310 int weight;
404f7c9e 311 unsigned int gro_count;
bea3348e
SH
312 int (*poll)(struct napi_struct *, int);
313#ifdef CONFIG_NETPOLL
314 spinlock_t poll_lock;
315 int poll_owner;
bea3348e 316#endif
5d38a079 317 struct net_device *dev;
d565b0a1 318 struct sk_buff *gro_list;
5d38a079 319 struct sk_buff *skb;
3b47d303 320 struct hrtimer timer;
404f7c9e 321 struct list_head dev_list;
af12fa6e
ET
322 struct hlist_node napi_hash_node;
323 unsigned int napi_id;
bea3348e
SH
324};
325
d94d9fee 326enum {
bea3348e 327 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 328 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 329 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
af12fa6e 330 NAPI_STATE_HASHED, /* In NAPI hash */
bea3348e
SH
331};
332
5b252f0c 333enum gro_result {
d1c76af9
HX
334 GRO_MERGED,
335 GRO_MERGED_FREE,
336 GRO_HELD,
337 GRO_NORMAL,
338 GRO_DROP,
339};
5b252f0c 340typedef enum gro_result gro_result_t;
d1c76af9 341
8a4eb573
JP
342/*
343 * enum rx_handler_result - Possible return values for rx_handlers.
344 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
345 * further.
346 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
347 * case skb->dev was changed by rx_handler.
348 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
349 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
350 *
351 * rx_handlers are functions called from inside __netif_receive_skb(), to do
352 * special processing of the skb, prior to delivery to protocol handlers.
353 *
354 * Currently, a net_device can only have a single rx_handler registered. Trying
355 * to register a second rx_handler will return -EBUSY.
356 *
357 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
358 * To unregister a rx_handler on a net_device, use
359 * netdev_rx_handler_unregister().
360 *
361 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
362 * do with the skb.
363 *
364 * If the rx_handler consumed to skb in some way, it should return
365 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
366 * the skb to be delivered in some other ways.
367 *
368 * If the rx_handler changed skb->dev, to divert the skb to another
369 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
370 * new device will be called if it exists.
371 *
372 * If the rx_handler consider the skb should be ignored, it should return
373 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 374 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
375 *
376 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
377 * delivered, it should return RX_HANDLER_PASS.
378 *
379 * A device without a registered rx_handler will behave as if rx_handler
380 * returned RX_HANDLER_PASS.
381 */
382
383enum rx_handler_result {
384 RX_HANDLER_CONSUMED,
385 RX_HANDLER_ANOTHER,
386 RX_HANDLER_EXACT,
387 RX_HANDLER_PASS,
388};
389typedef enum rx_handler_result rx_handler_result_t;
390typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 391
f629d208 392void __napi_schedule(struct napi_struct *n);
bc9ad166 393void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 394
4d29515f 395static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
396{
397 return test_bit(NAPI_STATE_DISABLE, &n->state);
398}
399
bea3348e
SH
400/**
401 * napi_schedule_prep - check if napi can be scheduled
402 * @n: napi context
403 *
404 * Test if NAPI routine is already running, and if not mark
405 * it as running. This is used as a condition variable
a0a46196
DM
406 * insure only one NAPI poll instance runs. We also make
407 * sure there is no pending NAPI disable.
bea3348e 408 */
4d29515f 409static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 410{
a0a46196
DM
411 return !napi_disable_pending(n) &&
412 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
413}
414
415/**
416 * napi_schedule - schedule NAPI poll
417 * @n: napi context
418 *
419 * Schedule NAPI poll routine to be called if it is not already
420 * running.
421 */
422static inline void napi_schedule(struct napi_struct *n)
423{
424 if (napi_schedule_prep(n))
425 __napi_schedule(n);
426}
427
bc9ad166
ED
428/**
429 * napi_schedule_irqoff - schedule NAPI poll
430 * @n: napi context
431 *
432 * Variant of napi_schedule(), assuming hard irqs are masked.
433 */
434static inline void napi_schedule_irqoff(struct napi_struct *n)
435{
436 if (napi_schedule_prep(n))
437 __napi_schedule_irqoff(n);
438}
439
bfe13f54 440/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 441static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
442{
443 if (napi_schedule_prep(napi)) {
444 __napi_schedule(napi);
4d29515f 445 return true;
bfe13f54 446 }
4d29515f 447 return false;
bfe13f54
RD
448}
449
3b47d303
ED
450void __napi_complete(struct napi_struct *n);
451void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
452/**
453 * napi_complete - NAPI processing complete
454 * @n: napi context
455 *
456 * Mark NAPI processing as complete.
3b47d303 457 * Consider using napi_complete_done() instead.
bea3348e 458 */
3b47d303
ED
459static inline void napi_complete(struct napi_struct *n)
460{
461 return napi_complete_done(n, 0);
462}
bea3348e 463
af12fa6e
ET
464/**
465 * napi_by_id - lookup a NAPI by napi_id
466 * @napi_id: hashed napi_id
467 *
468 * lookup @napi_id in napi_hash table
469 * must be called under rcu_read_lock()
470 */
f629d208 471struct napi_struct *napi_by_id(unsigned int napi_id);
af12fa6e
ET
472
473/**
474 * napi_hash_add - add a NAPI to global hashtable
475 * @napi: napi context
476 *
477 * generate a new napi_id and store a @napi under it in napi_hash
478 */
f629d208 479void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
480
481/**
482 * napi_hash_del - remove a NAPI from global table
483 * @napi: napi context
484 *
485 * Warning: caller must observe rcu grace period
486 * before freeing memory containing @napi
487 */
f629d208 488void napi_hash_del(struct napi_struct *napi);
af12fa6e 489
bea3348e
SH
490/**
491 * napi_disable - prevent NAPI from scheduling
492 * @n: napi context
493 *
494 * Stop NAPI from being scheduled on this context.
495 * Waits till any outstanding processing completes.
496 */
3b47d303 497void napi_disable(struct napi_struct *n);
bea3348e
SH
498
499/**
500 * napi_enable - enable NAPI scheduling
501 * @n: napi context
502 *
503 * Resume NAPI from being scheduled on this context.
504 * Must be paired with napi_disable.
505 */
506static inline void napi_enable(struct napi_struct *n)
507{
508 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 509 smp_mb__before_atomic();
bea3348e
SH
510 clear_bit(NAPI_STATE_SCHED, &n->state);
511}
512
c264c3de
SH
513#ifdef CONFIG_SMP
514/**
515 * napi_synchronize - wait until NAPI is not running
516 * @n: napi context
517 *
518 * Wait until NAPI is done being scheduled on this context.
519 * Waits till any outstanding processing completes but
520 * does not disable future activations.
521 */
522static inline void napi_synchronize(const struct napi_struct *n)
523{
524 while (test_bit(NAPI_STATE_SCHED, &n->state))
525 msleep(1);
526}
527#else
528# define napi_synchronize(n) barrier()
529#endif
530
d94d9fee 531enum netdev_queue_state_t {
73466498
TH
532 __QUEUE_STATE_DRV_XOFF,
533 __QUEUE_STATE_STACK_XOFF,
c3f26a26 534 __QUEUE_STATE_FROZEN,
79d16385 535};
8e2f1a63
DB
536
537#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
538#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
539#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
540
541#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
542#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
543 QUEUE_STATE_FROZEN)
544#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
545 QUEUE_STATE_FROZEN)
546
73466498
TH
547/*
548 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
549 * netif_tx_* functions below are used to manipulate this flag. The
550 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
551 * queue independently. The netif_xmit_*stopped functions below are called
552 * to check if the queue has been stopped by the driver or stack (either
553 * of the XOFF bits are set in the state). Drivers should not need to call
554 * netif_xmit*stopped functions, they should only be using netif_tx_*.
555 */
79d16385 556
bb949fbd 557struct netdev_queue {
6a321cb3
ED
558/*
559 * read mostly part
560 */
bb949fbd 561 struct net_device *dev;
46e5da40 562 struct Qdisc __rcu *qdisc;
b0e1e646 563 struct Qdisc *qdisc_sleeping;
ccf5ff69 564#ifdef CONFIG_SYSFS
1d24eb48
TH
565 struct kobject kobj;
566#endif
f2cd2d3e
ED
567#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
568 int numa_node;
569#endif
6a321cb3
ED
570/*
571 * write mostly part
572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
9d21493b
ED
575 /*
576 * please use this field instead of dev->trans_start
577 */
578 unsigned long trans_start;
ccf5ff69 579
580 /*
581 * Number of TX timeouts for this queue
582 * (/sys/class/net/DEV/Q/trans_timeout)
583 */
584 unsigned long trans_timeout;
114cf580
TH
585
586 unsigned long state;
587
588#ifdef CONFIG_BQL
589 struct dql dql;
590#endif
822b3b2e 591 unsigned long tx_maxrate;
e8a0464c 592} ____cacheline_aligned_in_smp;
bb949fbd 593
f2cd2d3e
ED
594static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
595{
596#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
597 return q->numa_node;
598#else
b236da69 599 return NUMA_NO_NODE;
f2cd2d3e
ED
600#endif
601}
602
603static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
604{
605#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 q->numa_node = node;
607#endif
608}
609
df334545 610#ifdef CONFIG_RPS
0a9627f2
TH
611/*
612 * This structure holds an RPS map which can be of variable length. The
613 * map is an array of CPUs.
614 */
615struct rps_map {
616 unsigned int len;
617 struct rcu_head rcu;
618 u16 cpus[0];
619};
60b778ce 620#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 621
fec5e652 622/*
c445477d
BH
623 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
624 * tail pointer for that CPU's input queue at the time of last enqueue, and
625 * a hardware filter index.
fec5e652
TH
626 */
627struct rps_dev_flow {
628 u16 cpu;
c445477d 629 u16 filter;
fec5e652
TH
630 unsigned int last_qtail;
631};
c445477d 632#define RPS_NO_FILTER 0xffff
fec5e652
TH
633
634/*
635 * The rps_dev_flow_table structure contains a table of flow mappings.
636 */
637struct rps_dev_flow_table {
638 unsigned int mask;
639 struct rcu_head rcu;
fec5e652
TH
640 struct rps_dev_flow flows[0];
641};
642#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 643 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
644
645/*
646 * The rps_sock_flow_table contains mappings of flows to the last CPU
647 * on which they were processed by the application (set in recvmsg).
567e4b79
ED
648 * Each entry is a 32bit value. Upper part is the high order bits
649 * of flow hash, lower part is cpu number.
650 * rps_cpu_mask is used to partition the space, depending on number of
651 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
652 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
653 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
654 */
655struct rps_sock_flow_table {
567e4b79 656 u32 mask;
93c1af6c
ED
657
658 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 659};
567e4b79 660#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
661
662#define RPS_NO_CPU 0xffff
663
567e4b79
ED
664extern u32 rps_cpu_mask;
665extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
666
fec5e652
TH
667static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
668 u32 hash)
669{
670 if (table && hash) {
567e4b79
ED
671 unsigned int index = hash & table->mask;
672 u32 val = hash & ~rps_cpu_mask;
fec5e652
TH
673
674 /* We only give a hint, preemption can change cpu under us */
567e4b79 675 val |= raw_smp_processor_id();
fec5e652 676
567e4b79
ED
677 if (table->ents[index] != val)
678 table->ents[index] = val;
fec5e652
TH
679 }
680}
681
c445477d 682#ifdef CONFIG_RFS_ACCEL
f629d208
JP
683bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
684 u16 filter_id);
c445477d 685#endif
a953be53 686#endif /* CONFIG_RPS */
c445477d 687
0a9627f2
TH
688/* This structure contains an instance of an RX queue. */
689struct netdev_rx_queue {
a953be53 690#ifdef CONFIG_RPS
6e3f7faf
ED
691 struct rps_map __rcu *rps_map;
692 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 693#endif
6e3f7faf 694 struct kobject kobj;
fe822240 695 struct net_device *dev;
0a9627f2 696} ____cacheline_aligned_in_smp;
a953be53
MD
697
698/*
699 * RX queue sysfs structures and functions.
700 */
701struct rx_queue_attribute {
702 struct attribute attr;
703 ssize_t (*show)(struct netdev_rx_queue *queue,
704 struct rx_queue_attribute *attr, char *buf);
705 ssize_t (*store)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, const char *buf, size_t len);
707};
d314774c 708
bf264145
TH
709#ifdef CONFIG_XPS
710/*
711 * This structure holds an XPS map which can be of variable length. The
712 * map is an array of queues.
713 */
714struct xps_map {
715 unsigned int len;
716 unsigned int alloc_len;
717 struct rcu_head rcu;
718 u16 queues[0];
719};
60b778ce 720#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
bf264145
TH
721#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
722 / sizeof(u16))
723
724/*
725 * This structure holds all XPS maps for device. Maps are indexed by CPU.
726 */
727struct xps_dev_maps {
728 struct rcu_head rcu;
a4177869 729 struct xps_map __rcu *cpu_map[0];
bf264145
TH
730};
731#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
732 (nr_cpu_ids * sizeof(struct xps_map *)))
733#endif /* CONFIG_XPS */
734
4f57c087
JF
735#define TC_MAX_QUEUE 16
736#define TC_BITMASK 15
737/* HW offloaded queuing disciplines txq count and offset maps */
738struct netdev_tc_txq {
739 u16 count;
740 u16 offset;
741};
742
68bad94e
NP
743#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
744/*
745 * This structure is to hold information about the device
746 * configured to run FCoE protocol stack.
747 */
748struct netdev_fcoe_hbainfo {
749 char manufacturer[64];
750 char serial_number[64];
751 char hardware_version[64];
752 char driver_version[64];
753 char optionrom_version[64];
754 char firmware_version[64];
755 char model[256];
756 char model_description[256];
757};
758#endif
759
02637fce 760#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 761
02637fce
JP
762/* This structure holds a unique identifier to identify some
763 * physical item (port for example) used by a netdevice.
66b52b0d 764 */
02637fce
JP
765struct netdev_phys_item_id {
766 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
767 unsigned char id_len;
768};
769
99932d4f
DB
770typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
771 struct sk_buff *skb);
772
d314774c
SH
773/*
774 * This structure defines the management hooks for network devices.
00829823
SH
775 * The following hooks can be defined; unless noted otherwise, they are
776 * optional and can be filled with a null pointer.
d314774c
SH
777 *
778 * int (*ndo_init)(struct net_device *dev);
779 * This function is called once when network device is registered.
780 * The network device can use this to any late stage initializaton
781 * or semantic validattion. It can fail with an error code which will
782 * be propogated back to register_netdev
783 *
784 * void (*ndo_uninit)(struct net_device *dev);
785 * This function is called when device is unregistered or when registration
786 * fails. It is not called if init fails.
787 *
788 * int (*ndo_open)(struct net_device *dev);
789 * This function is called when network device transistions to the up
790 * state.
791 *
792 * int (*ndo_stop)(struct net_device *dev);
793 * This function is called when network device transistions to the down
794 * state.
795 *
dc1f8bf6
SH
796 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
797 * struct net_device *dev);
00829823 798 * Called when a packet needs to be transmitted.
e79d8429
RR
799 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
800 * the queue before that can happen; it's for obsolete devices and weird
801 * corner cases, but the stack really does a non-trivial amount
802 * of useless work if you return NETDEV_TX_BUSY.
dc1f8bf6 803 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
804 * Required can not be NULL.
805 *
f663dd9a 806 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 807 * void *accel_priv, select_queue_fallback_t fallback);
00829823
SH
808 * Called to decide which queue to when device supports multiple
809 * transmit queues.
810 *
d314774c
SH
811 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
812 * This function is called to allow device receiver to make
813 * changes to configuration when multicast or promiscious is enabled.
814 *
815 * void (*ndo_set_rx_mode)(struct net_device *dev);
816 * This function is called device changes address list filtering.
01789349
JP
817 * If driver handles unicast address filtering, it should set
818 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
819 *
820 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
821 * This function is called when the Media Access Control address
37b607c5 822 * needs to be changed. If this interface is not defined, the
d314774c
SH
823 * mac address can not be changed.
824 *
825 * int (*ndo_validate_addr)(struct net_device *dev);
826 * Test if Media Access Control address is valid for the device.
827 *
828 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
829 * Called when a user request an ioctl which can't be handled by
830 * the generic interface code. If not defined ioctl's return
831 * not supported error code.
832 *
833 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
834 * Used to set network devices bus interface parameters. This interface
835 * is retained for legacy reason, new devices should use the bus
836 * interface (PCI) for low level management.
837 *
838 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
839 * Called when a user wants to change the Maximum Transfer Unit
840 * of a device. If not defined, any request to change MTU will
841 * will return an error.
842 *
00829823 843 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
844 * Callback uses when the transmitter has not made any progress
845 * for dev->watchdog ticks.
846 *
3cfde79c 847 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 848 * struct rtnl_link_stats64 *storage);
d308e38f 849 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 850 * Called when a user wants to get the network device usage
be1f3c2c 851 * statistics. Drivers must do one of the following:
3cfde79c
BH
852 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
853 * rtnl_link_stats64 structure passed by the caller.
82695d9b 854 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
855 * (which should normally be dev->stats) and return a pointer to
856 * it. The structure may be changed asynchronously only if each
857 * field is written atomically.
858 * 3. Update dev->stats asynchronously and atomically, and define
859 * neither operation.
d314774c 860 *
5d632cb7 861 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
862 * If device support VLAN filtering this function is called when a
863 * VLAN id is registered.
d314774c 864 *
5d632cb7 865 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
866 * If device support VLAN filtering this function is called when a
867 * VLAN id is unregistered.
d314774c
SH
868 *
869 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
870 *
871 * SR-IOV management functions.
872 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
873 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
874 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
875 * int max_tx_rate);
5f8444a3 876 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
877 * int (*ndo_get_vf_config)(struct net_device *dev,
878 * int vf, struct ifla_vf_info *ivf);
1d8faf48 879 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
880 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
881 * struct nlattr *port[]);
01a3d796
VZ
882 *
883 * Enable or disable the VF ability to query its RSS Redirection Table and
884 * Hash Key. This is needed since on some devices VF share this information
885 * with PF and querying it may adduce a theoretical security risk.
886 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 887 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
888 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
889 * Called to setup 'tc' number of traffic classes in the net device. This
890 * is always called from the stack with the rtnl lock held and netif tx
891 * queues stopped. This allows the netdevice to perform queue management
892 * safely.
c445477d 893 *
e9bce845
YZ
894 * Fiber Channel over Ethernet (FCoE) offload functions.
895 * int (*ndo_fcoe_enable)(struct net_device *dev);
896 * Called when the FCoE protocol stack wants to start using LLD for FCoE
897 * so the underlying device can perform whatever needed configuration or
898 * initialization to support acceleration of FCoE traffic.
899 *
900 * int (*ndo_fcoe_disable)(struct net_device *dev);
901 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
902 * so the underlying device can perform whatever needed clean-ups to
903 * stop supporting acceleration of FCoE traffic.
904 *
905 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
906 * struct scatterlist *sgl, unsigned int sgc);
907 * Called when the FCoE Initiator wants to initialize an I/O that
908 * is a possible candidate for Direct Data Placement (DDP). The LLD can
909 * perform necessary setup and returns 1 to indicate the device is set up
910 * successfully to perform DDP on this I/O, otherwise this returns 0.
911 *
912 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
913 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
914 * indicated by the FC exchange id 'xid', so the underlying device can
915 * clean up and reuse resources for later DDP requests.
916 *
917 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
918 * struct scatterlist *sgl, unsigned int sgc);
919 * Called when the FCoE Target wants to initialize an I/O that
920 * is a possible candidate for Direct Data Placement (DDP). The LLD can
921 * perform necessary setup and returns 1 to indicate the device is set up
922 * successfully to perform DDP on this I/O, otherwise this returns 0.
923 *
68bad94e
NP
924 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
925 * struct netdev_fcoe_hbainfo *hbainfo);
926 * Called when the FCoE Protocol stack wants information on the underlying
927 * device. This information is utilized by the FCoE protocol stack to
928 * register attributes with Fiber Channel management service as per the
929 * FC-GS Fabric Device Management Information(FDMI) specification.
930 *
e9bce845
YZ
931 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
932 * Called when the underlying device wants to override default World Wide
933 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
934 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
935 * protocol stack to use.
936 *
c445477d
BH
937 * RFS acceleration.
938 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
939 * u16 rxq_index, u32 flow_id);
940 * Set hardware filter for RFS. rxq_index is the target queue index;
941 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
942 * Return the filter ID on success, or a negative error code.
fbaec0ea 943 *
8b98a70c 944 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
945 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
946 * Called to make another netdev an underling.
947 *
948 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
949 * Called to release previously enslaved netdev.
5455c699
MM
950 *
951 * Feature/offload setting functions.
c8f44aff
MM
952 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
953 * netdev_features_t features);
5455c699
MM
954 * Adjusts the requested feature flags according to device-specific
955 * constraints, and returns the resulting flags. Must not modify
956 * the device state.
957 *
c8f44aff 958 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
959 * Called to update device configuration to new features. Passed
960 * feature set might be less than what was returned by ndo_fix_features()).
961 * Must return >0 or -errno if it changed dev->features itself.
962 *
edc7d573 963 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
964 * struct net_device *dev,
f6f6424b 965 * const unsigned char *addr, u16 vid, u16 flags)
77162022 966 * Adds an FDB entry to dev for addr.
1690be63
VY
967 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
968 * struct net_device *dev,
f6f6424b 969 * const unsigned char *addr, u16 vid)
77162022
JF
970 * Deletes the FDB entry from dev coresponding to addr.
971 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
972 * struct net_device *dev, struct net_device *filter_dev,
973 * int idx)
77162022
JF
974 * Used to add FDB entries to dump requests. Implementers should add
975 * entries to skb and update idx with the number of entries.
e5a55a89 976 *
ad41faa8
ND
977 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
978 * u16 flags)
e5a55a89 979 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
980 * struct net_device *dev, u32 filter_mask,
981 * int nlflags)
ad41faa8
ND
982 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
983 * u16 flags);
4bf84c35
JP
984 *
985 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
986 * Called to change device carrier. Soft-devices (like dummy, team, etc)
987 * which do not represent real hardware may define this to allow their
988 * userspace components to manage their virtual carrier state. Devices
989 * that determine carrier state from physical hardware properties (eg
990 * network cables) or protocol-dependent mechanisms (eg
991 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
992 *
993 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 994 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
995 * Called to get ID of physical port of this device. If driver does
996 * not implement this, it is assumed that the hw is not able to have
997 * multiple net devices on single physical port.
53cf5275
JG
998 *
999 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 1000 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1001 * Called by vxlan to notiy a driver about the UDP port and socket
1002 * address family that vxlan is listnening to. It is called only when
1003 * a new port starts listening. The operation is protected by the
1004 * vxlan_net->sock_lock.
1005 *
1006 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 1007 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1008 * Called by vxlan to notify the driver about a UDP port and socket
1009 * address family that vxlan is not listening to anymore. The operation
1010 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
1011 *
1012 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1013 * struct net_device *dev)
1014 * Called by upper layer devices to accelerate switching or other
1015 * station functionality into hardware. 'pdev is the lowerdev
1016 * to use for the offload and 'dev' is the net device that will
1017 * back the offload. Returns a pointer to the private structure
1018 * the upper layer will maintain.
1019 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1020 * Called by upper layer device to delete the station created
1021 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1022 * the station and priv is the structure returned by the add
1023 * operation.
1024 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1025 * struct net_device *dev,
1026 * void *priv);
1027 * Callback to use for xmit over the accelerated station. This
1028 * is used in place of ndo_start_xmit on accelerated net
1029 * devices.
5f35227e
JG
1030 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1031 * struct net_device *dev
1032 * netdev_features_t features);
04ffcb25 1033 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1034 * performing offload operations on a given packet. This is to give
1035 * the device an opportunity to implement any restrictions that cannot
1036 * be otherwise expressed by feature flags. The check is called with
1037 * the set of features that the stack has calculated and it returns
1038 * those the driver believes to be appropriate.
822b3b2e
JF
1039 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1040 * int queue_index, u32 maxrate);
1041 * Called when a user wants to set a max-rate limitation of specific
1042 * TX queue.
a54acb3a
ND
1043 * int (*ndo_get_iflink)(const struct net_device *dev);
1044 * Called to get the iflink value of this device.
d314774c
SH
1045 */
1046struct net_device_ops {
1047 int (*ndo_init)(struct net_device *dev);
1048 void (*ndo_uninit)(struct net_device *dev);
1049 int (*ndo_open)(struct net_device *dev);
1050 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 1051 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
1052 struct net_device *dev);
1053 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1054 struct sk_buff *skb,
99932d4f
DB
1055 void *accel_priv,
1056 select_queue_fallback_t fallback);
d314774c
SH
1057 void (*ndo_change_rx_flags)(struct net_device *dev,
1058 int flags);
d314774c 1059 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1060 int (*ndo_set_mac_address)(struct net_device *dev,
1061 void *addr);
d314774c 1062 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1063 int (*ndo_do_ioctl)(struct net_device *dev,
1064 struct ifreq *ifr, int cmd);
d314774c
SH
1065 int (*ndo_set_config)(struct net_device *dev,
1066 struct ifmap *map);
00829823
SH
1067 int (*ndo_change_mtu)(struct net_device *dev,
1068 int new_mtu);
1069 int (*ndo_neigh_setup)(struct net_device *dev,
1070 struct neigh_parms *);
d314774c
SH
1071 void (*ndo_tx_timeout) (struct net_device *dev);
1072
28172739
ED
1073 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1074 struct rtnl_link_stats64 *storage);
d314774c
SH
1075 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1076
8e586137 1077 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1078 __be16 proto, u16 vid);
8e586137 1079 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1080 __be16 proto, u16 vid);
d314774c 1081#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1082 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1083 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1084 struct netpoll_info *info);
0e34e931 1085 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1086#endif
e0d1095a 1087#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1088 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1089#endif
95c26df8
WM
1090 int (*ndo_set_vf_mac)(struct net_device *dev,
1091 int queue, u8 *mac);
1092 int (*ndo_set_vf_vlan)(struct net_device *dev,
1093 int queue, u16 vlan, u8 qos);
ed616689
SC
1094 int (*ndo_set_vf_rate)(struct net_device *dev,
1095 int vf, int min_tx_rate,
1096 int max_tx_rate);
5f8444a3
GR
1097 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1098 int vf, bool setting);
95c26df8
WM
1099 int (*ndo_get_vf_config)(struct net_device *dev,
1100 int vf,
1101 struct ifla_vf_info *ivf);
1d8faf48
RE
1102 int (*ndo_set_vf_link_state)(struct net_device *dev,
1103 int vf, int link_state);
57b61080
SF
1104 int (*ndo_set_vf_port)(struct net_device *dev,
1105 int vf,
1106 struct nlattr *port[]);
1107 int (*ndo_get_vf_port)(struct net_device *dev,
1108 int vf, struct sk_buff *skb);
01a3d796
VZ
1109 int (*ndo_set_vf_rss_query_en)(
1110 struct net_device *dev,
1111 int vf, bool setting);
4f57c087 1112 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
d11ead75 1113#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1114 int (*ndo_fcoe_enable)(struct net_device *dev);
1115 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1116 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1117 u16 xid,
1118 struct scatterlist *sgl,
1119 unsigned int sgc);
1120 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1121 u16 xid);
6247e086
YZ
1122 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1123 u16 xid,
1124 struct scatterlist *sgl,
1125 unsigned int sgc);
68bad94e
NP
1126 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1127 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1128#endif
1129
d11ead75 1130#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1131#define NETDEV_FCOE_WWNN 0
1132#define NETDEV_FCOE_WWPN 1
1133 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1134 u64 *wwn, int type);
4d288d57 1135#endif
3c9c36bc 1136
c445477d
BH
1137#ifdef CONFIG_RFS_ACCEL
1138 int (*ndo_rx_flow_steer)(struct net_device *dev,
1139 const struct sk_buff *skb,
1140 u16 rxq_index,
1141 u32 flow_id);
1142#endif
fbaec0ea
JP
1143 int (*ndo_add_slave)(struct net_device *dev,
1144 struct net_device *slave_dev);
1145 int (*ndo_del_slave)(struct net_device *dev,
1146 struct net_device *slave_dev);
c8f44aff
MM
1147 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1148 netdev_features_t features);
5455c699 1149 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1150 netdev_features_t features);
da6a8fa0 1151 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1152 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1153
1154 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1155 struct nlattr *tb[],
77162022 1156 struct net_device *dev,
6b6e2725 1157 const unsigned char *addr,
f6f6424b 1158 u16 vid,
77162022
JF
1159 u16 flags);
1160 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1161 struct nlattr *tb[],
77162022 1162 struct net_device *dev,
f6f6424b
JP
1163 const unsigned char *addr,
1164 u16 vid);
77162022
JF
1165 int (*ndo_fdb_dump)(struct sk_buff *skb,
1166 struct netlink_callback *cb,
1167 struct net_device *dev,
5d5eacb3 1168 struct net_device *filter_dev,
77162022 1169 int idx);
e5a55a89
JF
1170
1171 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1172 struct nlmsghdr *nlh,
1173 u16 flags);
e5a55a89
JF
1174 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1175 u32 pid, u32 seq,
6cbdceeb 1176 struct net_device *dev,
46c264da
ND
1177 u32 filter_mask,
1178 int nlflags);
407af329 1179 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1180 struct nlmsghdr *nlh,
1181 u16 flags);
4bf84c35
JP
1182 int (*ndo_change_carrier)(struct net_device *dev,
1183 bool new_carrier);
66b52b0d 1184 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1185 struct netdev_phys_item_id *ppid);
db24a904
DA
1186 int (*ndo_get_phys_port_name)(struct net_device *dev,
1187 char *name, size_t len);
53cf5275
JG
1188 void (*ndo_add_vxlan_port)(struct net_device *dev,
1189 sa_family_t sa_family,
35e42379 1190 __be16 port);
53cf5275
JG
1191 void (*ndo_del_vxlan_port)(struct net_device *dev,
1192 sa_family_t sa_family,
35e42379 1193 __be16 port);
a6cc0cfa
JF
1194
1195 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1196 struct net_device *dev);
1197 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1198 void *priv);
1199
1200 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1201 struct net_device *dev,
1202 void *priv);
25175ba5 1203 int (*ndo_get_lock_subclass)(struct net_device *dev);
5f35227e
JG
1204 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1205 struct net_device *dev,
1206 netdev_features_t features);
822b3b2e
JF
1207 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1208 int queue_index,
1209 u32 maxrate);
a54acb3a 1210 int (*ndo_get_iflink)(const struct net_device *dev);
d314774c
SH
1211};
1212
7aa98047
LR
1213/**
1214 * enum net_device_priv_flags - &struct net_device priv_flags
1215 *
1216 * These are the &struct net_device, they are only set internally
1217 * by drivers and used in the kernel. These flags are invisible to
1218 * userspace, this means that the order of these flags can change
1219 * during any kernel release.
1220 *
1221 * You should have a pretty good reason to be extending these flags.
1222 *
1223 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1224 * @IFF_EBRIDGE: Ethernet bridging device
1225 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1226 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1227 * @IFF_MASTER_ALB: bonding master, balance-alb
1228 * @IFF_BONDING: bonding master or slave
1229 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1230 * @IFF_ISATAP: ISATAP interface (RFC4214)
1231 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1232 * @IFF_WAN_HDLC: WAN HDLC device
1233 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1234 * release skb->dst
1235 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1236 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1237 * @IFF_MACVLAN_PORT: device used as macvlan port
1238 * @IFF_BRIDGE_PORT: device used as bridge port
1239 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1240 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1241 * @IFF_UNICAST_FLT: Supports unicast filtering
1242 * @IFF_TEAM_PORT: device used as team port
1243 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1244 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1245 * change when it's running
1246 * @IFF_MACVLAN: Macvlan device
1247 */
1248enum netdev_priv_flags {
1249 IFF_802_1Q_VLAN = 1<<0,
1250 IFF_EBRIDGE = 1<<1,
1251 IFF_SLAVE_INACTIVE = 1<<2,
1252 IFF_MASTER_8023AD = 1<<3,
1253 IFF_MASTER_ALB = 1<<4,
1254 IFF_BONDING = 1<<5,
1255 IFF_SLAVE_NEEDARP = 1<<6,
1256 IFF_ISATAP = 1<<7,
1257 IFF_MASTER_ARPMON = 1<<8,
1258 IFF_WAN_HDLC = 1<<9,
1259 IFF_XMIT_DST_RELEASE = 1<<10,
1260 IFF_DONT_BRIDGE = 1<<11,
1261 IFF_DISABLE_NETPOLL = 1<<12,
1262 IFF_MACVLAN_PORT = 1<<13,
1263 IFF_BRIDGE_PORT = 1<<14,
1264 IFF_OVS_DATAPATH = 1<<15,
1265 IFF_TX_SKB_SHARING = 1<<16,
1266 IFF_UNICAST_FLT = 1<<17,
1267 IFF_TEAM_PORT = 1<<18,
1268 IFF_SUPP_NOFCS = 1<<19,
1269 IFF_LIVE_ADDR_CHANGE = 1<<20,
1270 IFF_MACVLAN = 1<<21,
02875878 1271 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
2ad7bf36
MB
1272 IFF_IPVLAN_MASTER = 1<<23,
1273 IFF_IPVLAN_SLAVE = 1<<24,
7aa98047
LR
1274};
1275
1276#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1277#define IFF_EBRIDGE IFF_EBRIDGE
1278#define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1279#define IFF_MASTER_8023AD IFF_MASTER_8023AD
1280#define IFF_MASTER_ALB IFF_MASTER_ALB
1281#define IFF_BONDING IFF_BONDING
1282#define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1283#define IFF_ISATAP IFF_ISATAP
1284#define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1285#define IFF_WAN_HDLC IFF_WAN_HDLC
1286#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1287#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1288#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1289#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1290#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1291#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1292#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1293#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1294#define IFF_TEAM_PORT IFF_TEAM_PORT
1295#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1296#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1297#define IFF_MACVLAN IFF_MACVLAN
02875878 1298#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1299#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1300#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
7aa98047 1301
536721b1
KK
1302/**
1303 * struct net_device - The DEVICE structure.
1304 * Actually, this whole structure is a big mistake. It mixes I/O
1305 * data with strictly "high-level" data, and it has to know about
1306 * almost every data structure used in the INET module.
1307 *
1308 * @name: This is the first field of the "visible" part of this structure
1309 * (i.e. as seen by users in the "Space.c" file). It is the name
1310 * of the interface.
1311 *
1312 * @name_hlist: Device name hash chain, please keep it close to name[]
1313 * @ifalias: SNMP alias
1314 * @mem_end: Shared memory end
1315 * @mem_start: Shared memory start
1316 * @base_addr: Device I/O address
1317 * @irq: Device IRQ number
1318 *
14ffbbb8
TG
1319 * @carrier_changes: Stats to monitor carrier on<->off transitions
1320 *
536721b1
KK
1321 * @state: Generic network queuing layer state, see netdev_state_t
1322 * @dev_list: The global list of network devices
1323 * @napi_list: List entry, that is used for polling napi devices
1324 * @unreg_list: List entry, that is used, when we are unregistering the
1325 * device, see the function unregister_netdev
1326 * @close_list: List entry, that is used, when we are closing the device
1327 *
1328 * @adj_list: Directly linked devices, like slaves for bonding
1329 * @all_adj_list: All linked devices, *including* neighbours
1330 * @features: Currently active device features
1331 * @hw_features: User-changeable features
1332 *
1333 * @wanted_features: User-requested features
1334 * @vlan_features: Mask of features inheritable by VLAN devices
1335 *
1336 * @hw_enc_features: Mask of features inherited by encapsulating devices
1337 * This field indicates what encapsulation
1338 * offloads the hardware is capable of doing,
1339 * and drivers will need to set them appropriately.
1340 *
1341 * @mpls_features: Mask of features inheritable by MPLS
1342 *
1343 * @ifindex: interface index
388069d3 1344 * @group: The group, that the device belongs to
536721b1
KK
1345 *
1346 * @stats: Statistics struct, which was left as a legacy, use
1347 * rtnl_link_stats64 instead
1348 *
1349 * @rx_dropped: Dropped packets by core network,
1350 * do not use this in drivers
1351 * @tx_dropped: Dropped packets by core network,
1352 * do not use this in drivers
1353 *
536721b1
KK
1354 * @wireless_handlers: List of functions to handle Wireless Extensions,
1355 * instead of ioctl,
1356 * see <net/iw_handler.h> for details.
1357 * @wireless_data: Instance data managed by the core of wireless extensions
1358 *
1359 * @netdev_ops: Includes several pointers to callbacks,
1360 * if one wants to override the ndo_*() functions
1361 * @ethtool_ops: Management operations
d476059e 1362 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1363 * of Layer 2 headers.
1364 *
1365 * @flags: Interface flags (a la BSD)
1366 * @priv_flags: Like 'flags' but invisible to userspace,
1367 * see if.h for the definitions
1368 * @gflags: Global flags ( kept as legacy )
1369 * @padded: How much padding added by alloc_netdev()
1370 * @operstate: RFC2863 operstate
1371 * @link_mode: Mapping policy to operstate
1372 * @if_port: Selectable AUI, TP, ...
1373 * @dma: DMA channel
1374 * @mtu: Interface MTU value
1375 * @type: Interface hardware type
1376 * @hard_header_len: Hardware header length
1377 *
1378 * @needed_headroom: Extra headroom the hardware may need, but not in all
1379 * cases can this be guaranteed
1380 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1381 * cases can this be guaranteed. Some cases also use
1382 * LL_MAX_HEADER instead to allocate the skb
1383 *
1384 * interface address info:
1385 *
1386 * @perm_addr: Permanent hw address
1387 * @addr_assign_type: Hw address assignment type
1388 * @addr_len: Hardware address length
1389 * @neigh_priv_len; Used in neigh_alloc(),
1390 * initialized only in atm/clip.c
1391 * @dev_id: Used to differentiate devices that share
1392 * the same link layer address
1393 * @dev_port: Used to differentiate devices that share
1394 * the same function
1395 * @addr_list_lock: XXX: need comments on this one
536721b1
KK
1396 * @uc_promisc: Counter, that indicates, that promiscuous mode
1397 * has been enabled due to the need to listen to
1398 * additional unicast addresses in a device that
1399 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1400 * @uc: unicast mac addresses
1401 * @mc: multicast mac addresses
1402 * @dev_addrs: list of device hw addresses
1403 * @queues_kset: Group of all Kobjects in the Tx and RX queues
536721b1
KK
1404 * @promiscuity: Number of times, the NIC is told to work in
1405 * Promiscuous mode, if it becomes 0 the NIC will
1406 * exit from working in Promiscuous mode
1407 * @allmulti: Counter, enables or disables allmulticast mode
1408 *
1409 * @vlan_info: VLAN info
1410 * @dsa_ptr: dsa specific data
1411 * @tipc_ptr: TIPC specific data
1412 * @atalk_ptr: AppleTalk link
1413 * @ip_ptr: IPv4 specific data
1414 * @dn_ptr: DECnet specific data
1415 * @ip6_ptr: IPv6 specific data
1416 * @ax25_ptr: AX.25 specific data
1417 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1418 *
1419 * @last_rx: Time of last Rx
1420 * @dev_addr: Hw address (before bcast,
1421 * because most packets are unicast)
1422 *
1423 * @_rx: Array of RX queues
1424 * @num_rx_queues: Number of RX queues
1425 * allocated at register_netdev() time
1426 * @real_num_rx_queues: Number of RX queues currently active in device
1427 *
1428 * @rx_handler: handler for received packets
1429 * @rx_handler_data: XXX: need comments on this one
1430 * @ingress_queue: XXX: need comments on this one
1431 * @broadcast: hw bcast address
1432 *
14ffbbb8
TG
1433 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1434 * indexed by RX queue number. Assigned by driver.
1435 * This must only be set if the ndo_rx_flow_steer
1436 * operation is defined
1437 * @index_hlist: Device index hash chain
1438 *
536721b1
KK
1439 * @_tx: Array of TX queues
1440 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1441 * @real_num_tx_queues: Number of TX queues currently active in device
1442 * @qdisc: Root qdisc from userspace point of view
1443 * @tx_queue_len: Max frames per queue allowed
1444 * @tx_global_lock: XXX: need comments on this one
1445 *
1446 * @xps_maps: XXX: need comments on this one
1447 *
536721b1
KK
1448 * @trans_start: Time (in jiffies) of last Tx
1449 * @watchdog_timeo: Represents the timeout that is used by
1450 * the watchdog ( see dev_watchdog() )
1451 * @watchdog_timer: List of timers
1452 *
1453 * @pcpu_refcnt: Number of references to this device
1454 * @todo_list: Delayed register/unregister
536721b1
KK
1455 * @link_watch_list: XXX: need comments on this one
1456 *
1457 * @reg_state: Register/unregister state machine
1458 * @dismantle: Device is going to be freed
1459 * @rtnl_link_state: This enum represents the phases of creating
1460 * a new link
1461 *
1462 * @destructor: Called from unregister,
1463 * can be used to call free_netdev
1464 * @npinfo: XXX: need comments on this one
1465 * @nd_net: Network namespace this network device is inside
1466 *
1467 * @ml_priv: Mid-layer private
1468 * @lstats: Loopback statistics
1469 * @tstats: Tunnel statistics
1470 * @dstats: Dummy statistics
1471 * @vstats: Virtual ethernet statistics
1472 *
1473 * @garp_port: GARP
1474 * @mrp_port: MRP
1475 *
1476 * @dev: Class/net/name entry
1477 * @sysfs_groups: Space for optional device, statistics and wireless
1478 * sysfs groups
1479 *
1480 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1481 * @rtnl_link_ops: Rtnl_link_ops
1482 *
1483 * @gso_max_size: Maximum size of generic segmentation offload
1484 * @gso_max_segs: Maximum number of segments that can be passed to the
1485 * NIC for GSO
fcbeb976
ED
1486 * @gso_min_segs: Minimum number of segments that can be passed to the
1487 * NIC for GSO
536721b1
KK
1488 *
1489 * @dcbnl_ops: Data Center Bridging netlink ops
1490 * @num_tc: Number of traffic classes in the net device
1491 * @tc_to_txq: XXX: need comments on this one
1492 * @prio_tc_map XXX: need comments on this one
1493 *
1494 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1495 *
1496 * @priomap: XXX: need comments on this one
1497 * @phydev: Physical device may attach itself
1498 * for hardware timestamping
1499 *
1500 * @qdisc_tx_busylock: XXX: need comments on this one
1501 *
536721b1 1502 * @pm_qos_req: Power Management QoS object
1da177e4
LT
1503 *
1504 * FIXME: cleanup struct net_device such that network protocol info
1505 * moves out.
1506 */
1507
d94d9fee 1508struct net_device {
1da177e4 1509 char name[IFNAMSIZ];
9356b8fc 1510 struct hlist_node name_hlist;
0b815a1a 1511 char *ifalias;
1da177e4
LT
1512 /*
1513 * I/O specific fields
1514 * FIXME: Merge these and struct ifmap into one
1515 */
536721b1
KK
1516 unsigned long mem_end;
1517 unsigned long mem_start;
1518 unsigned long base_addr;
1519 int irq;
1da177e4 1520
14ffbbb8
TG
1521 atomic_t carrier_changes;
1522
1da177e4 1523 /*
536721b1
KK
1524 * Some hardware also needs these fields (state,dev_list,
1525 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1526 * part of the usual set specified in Space.c.
1527 */
1528
1da177e4
LT
1529 unsigned long state;
1530
7562f876 1531 struct list_head dev_list;
bea3348e 1532 struct list_head napi_list;
44a0873d 1533 struct list_head unreg_list;
5cde2829 1534 struct list_head close_list;
7866a621
SN
1535 struct list_head ptype_all;
1536 struct list_head ptype_specific;
2f268f12 1537
2f268f12
VF
1538 struct {
1539 struct list_head upper;
1540 struct list_head lower;
1541 } adj_list;
1542
2f268f12
VF
1543 struct {
1544 struct list_head upper;
1545 struct list_head lower;
1546 } all_adj_list;
4c3d5e7b 1547
c8f44aff 1548 netdev_features_t features;
c8f44aff 1549 netdev_features_t hw_features;
c8f44aff 1550 netdev_features_t wanted_features;
c8f44aff 1551 netdev_features_t vlan_features;
6a674e9c 1552 netdev_features_t hw_enc_features;
0d89d203 1553 netdev_features_t mpls_features;
04ed3e74 1554
1da177e4 1555 int ifindex;
7a66bbc9 1556 int group;
1da177e4 1557
c45d286e 1558 struct net_device_stats stats;
015f0688 1559
015f0688
ED
1560 atomic_long_t rx_dropped;
1561 atomic_long_t tx_dropped;
1da177e4 1562
b86e0280 1563#ifdef CONFIG_WIRELESS_EXT
1da177e4 1564 const struct iw_handler_def * wireless_handlers;
1da177e4 1565 struct iw_public_data * wireless_data;
b86e0280 1566#endif
d314774c 1567 const struct net_device_ops *netdev_ops;
76fd8593 1568 const struct ethtool_ops *ethtool_ops;
4170604f 1569#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1570 const struct switchdev_ops *switchdev_ops;
4170604f 1571#endif
1da177e4 1572
3b04ddde
SH
1573 const struct header_ops *header_ops;
1574
536721b1
KK
1575 unsigned int flags;
1576 unsigned int priv_flags;
1577
1da177e4 1578 unsigned short gflags;
536721b1 1579 unsigned short padded;
1da177e4 1580
536721b1
KK
1581 unsigned char operstate;
1582 unsigned char link_mode;
b00055aa 1583
536721b1
KK
1584 unsigned char if_port;
1585 unsigned char dma;
bdc220da 1586
536721b1
KK
1587 unsigned int mtu;
1588 unsigned short type;
1589 unsigned short hard_header_len;
1da177e4 1590
f5184d26
JB
1591 unsigned short needed_headroom;
1592 unsigned short needed_tailroom;
1593
1da177e4 1594 /* Interface address info. */
536721b1
KK
1595 unsigned char perm_addr[MAX_ADDR_LEN];
1596 unsigned char addr_assign_type;
1597 unsigned char addr_len;
a0a9663d 1598 unsigned short neigh_priv_len;
536721b1
KK
1599 unsigned short dev_id;
1600 unsigned short dev_port;
ccffad25 1601 spinlock_t addr_list_lock;
14ffbbb8
TG
1602 unsigned char name_assign_type;
1603 bool uc_promisc;
536721b1
KK
1604 struct netdev_hw_addr_list uc;
1605 struct netdev_hw_addr_list mc;
1606 struct netdev_hw_addr_list dev_addrs;
1607
4c3d5e7b
ED
1608#ifdef CONFIG_SYSFS
1609 struct kset *queues_kset;
1610#endif
9d45abe1
WC
1611 unsigned int promiscuity;
1612 unsigned int allmulti;
1da177e4 1613
1da177e4
LT
1614
1615 /* Protocol specific pointers */
65ac6a5f 1616
d11ead75 1617#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1618 struct vlan_info __rcu *vlan_info;
65ac6a5f 1619#endif
34a430d7 1620#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1621 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1622#endif
1623#if IS_ENABLED(CONFIG_TIPC)
536721b1 1624 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1625#endif
536721b1
KK
1626 void *atalk_ptr;
1627 struct in_device __rcu *ip_ptr;
1628 struct dn_dev __rcu *dn_ptr;
1629 struct inet6_dev __rcu *ip6_ptr;
1630 void *ax25_ptr;
1631 struct wireless_dev *ieee80211_ptr;
98a18b6f 1632 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1633#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1634 struct mpls_dev __rcu *mpls_ptr;
1635#endif
1da177e4 1636
9356b8fc 1637/*
cd13539b 1638 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1639 */
536721b1 1640 unsigned long last_rx;
4dc89133 1641
9356b8fc 1642 /* Interface address info used in eth_type_trans() */
536721b1 1643 unsigned char *dev_addr;
f001fde5 1644
0a9627f2 1645
a953be53 1646#ifdef CONFIG_SYSFS
0a9627f2
TH
1647 struct netdev_rx_queue *_rx;
1648
0a9627f2 1649 unsigned int num_rx_queues;
62fe0b40 1650 unsigned int real_num_rx_queues;
c445477d 1651
df334545 1652#endif
0a9627f2 1653
3b47d303 1654 unsigned long gro_flush_timeout;
61391cde 1655 rx_handler_func_t __rcu *rx_handler;
1656 void __rcu *rx_handler_data;
e8a0464c 1657
4cda01e8 1658#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1659 struct tcf_proto __rcu *ingress_cl_list;
1660#endif
24824a09 1661 struct netdev_queue __rcu *ingress_queue;
d2788d34 1662
536721b1 1663 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1664#ifdef CONFIG_RFS_ACCEL
1665 struct cpu_rmap *rx_cpu_rmap;
1666#endif
1667 struct hlist_node index_hlist;
cd13539b
ED
1668
1669/*
1670 * Cache lines mostly used on transmit path
1671 */
e8a0464c
DM
1672 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1673 unsigned int num_tx_queues;
fd2ea0a7 1674 unsigned int real_num_tx_queues;
af356afa 1675 struct Qdisc *qdisc;
536721b1 1676 unsigned long tx_queue_len;
c3f26a26 1677 spinlock_t tx_global_lock;
14ffbbb8 1678 int watchdog_timeo;
cd13539b 1679
bf264145 1680#ifdef CONFIG_XPS
a4177869 1681 struct xps_dev_maps __rcu *xps_maps;
bf264145 1682#endif
1d24eb48 1683
9356b8fc 1684 /* These may be needed for future network-power-down code. */
9d21493b
ED
1685
1686 /*
1687 * trans_start here is expensive for high speed devices on SMP,
1688 * please use netdev_queue->trans_start instead.
1689 */
536721b1 1690 unsigned long trans_start;
9356b8fc 1691
9356b8fc
ED
1692 struct timer_list watchdog_timer;
1693
29b4433d 1694 int __percpu *pcpu_refcnt;
1da177e4 1695 struct list_head todo_list;
1da177e4 1696
e014debe 1697 struct list_head link_watch_list;
572a103d 1698
1da177e4 1699 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1700 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1701 NETREG_UNREGISTERING, /* called unregister_netdevice */
1702 NETREG_UNREGISTERED, /* completed unregister todo */
1703 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1704 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1705 } reg_state:8;
1706
536721b1 1707 bool dismantle;
a2835763
PM
1708
1709 enum {
1710 RTNL_LINK_INITIALIZED,
1711 RTNL_LINK_INITIALIZING,
1712 } rtnl_link_state:16;
1da177e4 1713
d314774c 1714 void (*destructor)(struct net_device *dev);
1da177e4 1715
1da177e4 1716#ifdef CONFIG_NETPOLL
5fbee843 1717 struct netpoll_info __rcu *npinfo;
1da177e4 1718#endif
eae792b7 1719
0c5c9fb5 1720 possible_net_t nd_net;
4a1c5371 1721
4951704b 1722 /* mid-layer private */
a7855c78 1723 union {
536721b1
KK
1724 void *ml_priv;
1725 struct pcpu_lstats __percpu *lstats;
8f84985f 1726 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1727 struct pcpu_dstats __percpu *dstats;
1728 struct pcpu_vstats __percpu *vstats;
a7855c78 1729 };
536721b1 1730
3cc77ec7 1731 struct garp_port __rcu *garp_port;
febf018d 1732 struct mrp_port __rcu *mrp_port;
1da177e4 1733
536721b1 1734 struct device dev;
0c509a6c 1735 const struct attribute_group *sysfs_groups[4];
a953be53 1736 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1737
38f7b870 1738 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1739
82cc1a7a
PWJ
1740 /* for setting kernel sock attribute on TCP connection setup */
1741#define GSO_MAX_SIZE 65536
1742 unsigned int gso_max_size;
30b678d8
BH
1743#define GSO_MAX_SEGS 65535
1744 u16 gso_max_segs;
fcbeb976 1745 u16 gso_min_segs;
7a6b6f51 1746#ifdef CONFIG_DCB
32953543 1747 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1748#endif
4f57c087
JF
1749 u8 num_tc;
1750 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1751 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1752
d11ead75 1753#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1754 unsigned int fcoe_ddp_xid;
5bc1421e 1755#endif
86f8515f 1756#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1757 struct netprio_map __rcu *priomap;
4d288d57 1758#endif
c1f19b51 1759 struct phy_device *phydev;
23d3b8bf 1760 struct lock_class_key *qdisc_tx_busylock;
1da177e4 1761};
43cb76d9 1762#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1763
1764#define NETDEV_ALIGN 32
1da177e4 1765
4f57c087
JF
1766static inline
1767int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1768{
1769 return dev->prio_tc_map[prio & TC_BITMASK];
1770}
1771
1772static inline
1773int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1774{
1775 if (tc >= dev->num_tc)
1776 return -EINVAL;
1777
1778 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1779 return 0;
1780}
1781
1782static inline
1783void netdev_reset_tc(struct net_device *dev)
1784{
1785 dev->num_tc = 0;
1786 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1787 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1788}
1789
1790static inline
1791int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1792{
1793 if (tc >= dev->num_tc)
1794 return -EINVAL;
1795
1796 dev->tc_to_txq[tc].count = count;
1797 dev->tc_to_txq[tc].offset = offset;
1798 return 0;
1799}
1800
1801static inline
1802int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1803{
1804 if (num_tc > TC_MAX_QUEUE)
1805 return -EINVAL;
1806
1807 dev->num_tc = num_tc;
1808 return 0;
1809}
1810
1811static inline
1812int netdev_get_num_tc(struct net_device *dev)
1813{
1814 return dev->num_tc;
1815}
1816
e8a0464c
DM
1817static inline
1818struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1819 unsigned int index)
1820{
1821 return &dev->_tx[index];
1822}
1823
10c51b56
DB
1824static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1825 const struct sk_buff *skb)
1826{
1827 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1828}
1829
e8a0464c
DM
1830static inline void netdev_for_each_tx_queue(struct net_device *dev,
1831 void (*f)(struct net_device *,
1832 struct netdev_queue *,
1833 void *),
1834 void *arg)
1835{
1836 unsigned int i;
1837
1838 for (i = 0; i < dev->num_tx_queues; i++)
1839 f(dev, &dev->_tx[i], arg);
1840}
1841
f629d208 1842struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1843 struct sk_buff *skb,
1844 void *accel_priv);
8c4c49df 1845
c346dca1
YH
1846/*
1847 * Net namespace inlines
1848 */
1849static inline
1850struct net *dev_net(const struct net_device *dev)
1851{
c2d9ba9b 1852 return read_pnet(&dev->nd_net);
c346dca1
YH
1853}
1854
1855static inline
f5aa23fd 1856void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 1857{
0c5c9fb5 1858 write_pnet(&dev->nd_net, net);
c346dca1
YH
1859}
1860
3e8a72d1 1861static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 1862{
3fc88677 1863#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
1864 if (dev->dsa_ptr != NULL)
1865 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 1866#endif
5aed85ce 1867 return false;
396138f0
LB
1868}
1869
bea3348e
SH
1870/**
1871 * netdev_priv - access network device private data
1872 * @dev: network device
1873 *
1874 * Get network device private data
1875 */
6472ce60 1876static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1877{
1ce8e7b5 1878 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1879}
1880
1da177e4
LT
1881/* Set the sysfs physical device reference for the network logical device
1882 * if set prior to registration will cause a symlink during initialization.
1883 */
43cb76d9 1884#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1885
384912ed 1886/* Set the sysfs device type for the network logical device to allow
3f79410c 1887 * fine-grained identification of different network device types. For
384912ed
MH
1888 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1889 */
1890#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1891
82dc3c63
ED
1892/* Default NAPI poll() weight
1893 * Device drivers are strongly advised to not use bigger value
1894 */
1895#define NAPI_POLL_WEIGHT 64
1896
3b582cc1
SH
1897/**
1898 * netif_napi_add - initialize a napi context
1899 * @dev: network device
1900 * @napi: napi context
1901 * @poll: polling function
1902 * @weight: default weight
1903 *
1904 * netif_napi_add() must be used to initialize a napi context prior to calling
1905 * *any* of the other napi related functions.
1906 */
d565b0a1
HX
1907void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1908 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1909
d8156534
AD
1910/**
1911 * netif_napi_del - remove a napi context
1912 * @napi: napi context
1913 *
1914 * netif_napi_del() removes a napi context from the network device napi list
1915 */
d565b0a1
HX
1916void netif_napi_del(struct napi_struct *napi);
1917
1918struct napi_gro_cb {
78a478d0
HX
1919 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1920 void *frag0;
1921
7489594c
HX
1922 /* Length of frag0. */
1923 unsigned int frag0_len;
1924
86911732
HX
1925 /* This indicates where we are processing relative to skb->data. */
1926 int data_offset;
1927
d565b0a1 1928 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
1929 u16 flush;
1930
1931 /* Save the IP ID here and check when we get to the transport layer */
1932 u16 flush_id;
d565b0a1
HX
1933
1934 /* Number of segments aggregated. */
2e71a6f8
ED
1935 u16 count;
1936
15e2396d
TH
1937 /* Start offset for remote checksum offload */
1938 u16 gro_remcsum_start;
1939
2e71a6f8
ED
1940 /* jiffies when first packet was created/queued */
1941 unsigned long age;
86347245 1942
afe93325 1943 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
1944 u16 proto;
1945
baa32ff4
TH
1946 /* This is non-zero if the packet may be of the same flow. */
1947 u8 same_flow:1;
1948
b582ef09 1949 /* Used in udp_gro_receive */
573e8fca
TH
1950 u8 udp_mark:1;
1951
1952 /* GRO checksum is valid */
1953 u8 csum_valid:1;
1954
662880f4
TH
1955 /* Number of checksums via CHECKSUM_UNNECESSARY */
1956 u8 csum_cnt:3;
c3c7c254 1957
baa32ff4
TH
1958 /* Free the skb? */
1959 u8 free:2;
1960#define NAPI_GRO_FREE 1
1961#define NAPI_GRO_FREE_STOLEN_HEAD 2
1962
efc98d08
TH
1963 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1964 u8 is_ipv6:1;
1965
baa32ff4
TH
1966 /* 7 bit hole */
1967
bf5a755f
JC
1968 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1969 __wsum csum;
1970
c3c7c254
ED
1971 /* used in skb_gro_receive() slow path */
1972 struct sk_buff *last;
d565b0a1
HX
1973};
1974
1975#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 1976
1da177e4 1977struct packet_type {
f2ccd8fa
DM
1978 __be16 type; /* This is really htons(ether_type). */
1979 struct net_device *dev; /* NULL is wildcarded here */
1980 int (*func) (struct sk_buff *,
1981 struct net_device *,
1982 struct packet_type *,
1983 struct net_device *);
c0de08d0
EL
1984 bool (*id_match)(struct packet_type *ptype,
1985 struct sock *sk);
1da177e4
LT
1986 void *af_packet_priv;
1987 struct list_head list;
1988};
1989
f191a1d1 1990struct offload_callbacks {
576a30eb 1991 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 1992 netdev_features_t features);
d565b0a1 1993 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 1994 struct sk_buff *skb);
299603e8 1995 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
1996};
1997
1998struct packet_offload {
1999 __be16 type; /* This is really htons(ether_type). */
2000 struct offload_callbacks callbacks;
2001 struct list_head list;
1da177e4
LT
2002};
2003
a2b12f3c
TH
2004struct udp_offload;
2005
2006struct udp_offload_callbacks {
2007 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2008 struct sk_buff *skb,
2009 struct udp_offload *uoff);
2010 int (*gro_complete)(struct sk_buff *skb,
2011 int nhoff,
2012 struct udp_offload *uoff);
2013};
2014
b582ef09
OG
2015struct udp_offload {
2016 __be16 port;
afe93325 2017 u8 ipproto;
a2b12f3c 2018 struct udp_offload_callbacks callbacks;
b582ef09
OG
2019};
2020
8f84985f
LR
2021/* often modified stats are per cpu, other are shared (netdev->stats) */
2022struct pcpu_sw_netstats {
2023 u64 rx_packets;
2024 u64 rx_bytes;
2025 u64 tx_packets;
2026 u64 tx_bytes;
2027 struct u64_stats_sync syncp;
2028};
2029
1c213bd2
WC
2030#define netdev_alloc_pcpu_stats(type) \
2031({ \
693350c2 2032 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1c213bd2 2033 if (pcpu_stats) { \
ec65aafb
JB
2034 int __cpu; \
2035 for_each_possible_cpu(__cpu) { \
1c213bd2 2036 typeof(type) *stat; \
ec65aafb 2037 stat = per_cpu_ptr(pcpu_stats, __cpu); \
1c213bd2
WC
2038 u64_stats_init(&stat->syncp); \
2039 } \
2040 } \
2041 pcpu_stats; \
2042})
2043
1da177e4
LT
2044#include <linux/notifier.h>
2045
dcfe1421
AW
2046/* netdevice notifier chain. Please remember to update the rtnetlink
2047 * notification exclusion list in rtnetlink_event() when adding new
2048 * types.
2049 */
2050#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2051#define NETDEV_DOWN 0x0002
2052#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2053 detected a hardware crash and restarted
2054 - we can use this eg to kick tcp sessions
2055 once done */
2056#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2057#define NETDEV_REGISTER 0x0005
2058#define NETDEV_UNREGISTER 0x0006
1d486bfb 2059#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2060#define NETDEV_CHANGEADDR 0x0008
2061#define NETDEV_GOING_DOWN 0x0009
2062#define NETDEV_CHANGENAME 0x000A
2063#define NETDEV_FEAT_CHANGE 0x000B
2064#define NETDEV_BONDING_FAILOVER 0x000C
2065#define NETDEV_PRE_UP 0x000D
2066#define NETDEV_PRE_TYPE_CHANGE 0x000E
2067#define NETDEV_POST_TYPE_CHANGE 0x000F
2068#define NETDEV_POST_INIT 0x0010
0115e8e3 2069#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2070#define NETDEV_RELEASE 0x0012
2071#define NETDEV_NOTIFY_PEERS 0x0013
2072#define NETDEV_JOIN 0x0014
42e52bf9 2073#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2074#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2075#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2076#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2077#define NETDEV_BONDING_INFO 0x0019
dcfe1421 2078
f629d208
JP
2079int register_netdevice_notifier(struct notifier_block *nb);
2080int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2081
2082struct netdev_notifier_info {
2083 struct net_device *dev;
2084};
2085
be9efd36
JP
2086struct netdev_notifier_change_info {
2087 struct netdev_notifier_info info; /* must be first */
2088 unsigned int flags_changed;
2089};
2090
75538c2b
CW
2091static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2092 struct net_device *dev)
2093{
2094 info->dev = dev;
2095}
2096
351638e7
JP
2097static inline struct net_device *
2098netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2099{
2100 return info->dev;
2101}
2102
f629d208 2103int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2104
2105
1da177e4
LT
2106extern rwlock_t dev_base_lock; /* Device list lock */
2107
881d966b
EB
2108#define for_each_netdev(net, d) \
2109 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2110#define for_each_netdev_reverse(net, d) \
2111 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2112#define for_each_netdev_rcu(net, d) \
2113 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2114#define for_each_netdev_safe(net, d, n) \
2115 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2116#define for_each_netdev_continue(net, d) \
2117 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2118#define for_each_netdev_continue_rcu(net, d) \
2119 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2120#define for_each_netdev_in_bond_rcu(bond, slave) \
2121 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2122 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2123#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2124
a050c33f
DL
2125static inline struct net_device *next_net_device(struct net_device *dev)
2126{
2127 struct list_head *lh;
2128 struct net *net;
2129
c346dca1 2130 net = dev_net(dev);
a050c33f
DL
2131 lh = dev->dev_list.next;
2132 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2133}
2134
ce81b76a
ED
2135static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2136{
2137 struct list_head *lh;
2138 struct net *net;
2139
2140 net = dev_net(dev);
ccf43438 2141 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2142 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2143}
2144
a050c33f
DL
2145static inline struct net_device *first_net_device(struct net *net)
2146{
2147 return list_empty(&net->dev_base_head) ? NULL :
2148 net_device_entry(net->dev_base_head.next);
2149}
7562f876 2150
ccf43438
ED
2151static inline struct net_device *first_net_device_rcu(struct net *net)
2152{
2153 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2154
2155 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2156}
2157
f629d208
JP
2158int netdev_boot_setup_check(struct net_device *dev);
2159unsigned long netdev_boot_base(const char *prefix, int unit);
2160struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2161 const char *hwaddr);
2162struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2163struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2164void dev_add_pack(struct packet_type *pt);
2165void dev_remove_pack(struct packet_type *pt);
2166void __dev_remove_pack(struct packet_type *pt);
2167void dev_add_offload(struct packet_offload *po);
2168void dev_remove_offload(struct packet_offload *po);
f629d208 2169
a54acb3a 2170int dev_get_iflink(const struct net_device *dev);
6c555490
WC
2171struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2172 unsigned short mask);
f629d208
JP
2173struct net_device *dev_get_by_name(struct net *net, const char *name);
2174struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2175struct net_device *__dev_get_by_name(struct net *net, const char *name);
2176int dev_alloc_name(struct net_device *dev, const char *name);
2177int dev_open(struct net_device *dev);
2178int dev_close(struct net_device *dev);
99c4a26a 2179int dev_close_many(struct list_head *head, bool unlink);
f629d208 2180void dev_disable_lro(struct net_device *dev);
7026b1dd
DM
2181int dev_loopback_xmit(struct sock *sk, struct sk_buff *newskb);
2182int dev_queue_xmit_sk(struct sock *sk, struct sk_buff *skb);
2183static inline int dev_queue_xmit(struct sk_buff *skb)
2184{
2185 return dev_queue_xmit_sk(skb->sk, skb);
2186}
f663dd9a 2187int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2188int register_netdevice(struct net_device *dev);
2189void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2190void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2191static inline void unregister_netdevice(struct net_device *dev)
2192{
2193 unregister_netdevice_queue(dev, NULL);
2194}
2195
f629d208
JP
2196int netdev_refcnt_read(const struct net_device *dev);
2197void free_netdev(struct net_device *dev);
74d332c1 2198void netdev_freemem(struct net_device *dev);
f629d208
JP
2199void synchronize_net(void);
2200int init_dummy_netdev(struct net_device *dev);
937f1ba5 2201
f60e5990 2202DECLARE_PER_CPU(int, xmit_recursion);
2203static inline int dev_recursion_level(void)
2204{
2205 return this_cpu_read(xmit_recursion);
2206}
2207
f629d208
JP
2208struct net_device *dev_get_by_index(struct net *net, int ifindex);
2209struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2210struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2211int netdev_get_name(struct net *net, char *name, int ifindex);
2212int dev_restart(struct net_device *dev);
f629d208 2213int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2214
2215static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2216{
2217 return NAPI_GRO_CB(skb)->data_offset;
2218}
2219
2220static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2221{
2222 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2223}
2224
2225static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2226{
2227 NAPI_GRO_CB(skb)->data_offset += len;
2228}
2229
a5b1cf28
HX
2230static inline void *skb_gro_header_fast(struct sk_buff *skb,
2231 unsigned int offset)
86911732 2232{
a5b1cf28
HX
2233 return NAPI_GRO_CB(skb)->frag0 + offset;
2234}
78a478d0 2235
a5b1cf28
HX
2236static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2237{
2238 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2239}
78a478d0 2240
a5b1cf28
HX
2241static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2242 unsigned int offset)
2243{
17dd759c
HX
2244 if (!pskb_may_pull(skb, hlen))
2245 return NULL;
2246
a5b1cf28
HX
2247 NAPI_GRO_CB(skb)->frag0 = NULL;
2248 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2249 return skb->data + offset;
86911732 2250}
1da177e4 2251
36e7b1b8
HX
2252static inline void *skb_gro_network_header(struct sk_buff *skb)
2253{
78d3fd0b
HX
2254 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2255 skb_network_offset(skb);
36e7b1b8
HX
2256}
2257
bf5a755f
JC
2258static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2259 const void *start, unsigned int len)
2260{
573e8fca 2261 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2262 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2263 csum_partial(start, len, 0));
2264}
2265
573e8fca
TH
2266/* GRO checksum functions. These are logical equivalents of the normal
2267 * checksum functions (in skbuff.h) except that they operate on the GRO
2268 * offsets and fields in sk_buff.
2269 */
2270
2271__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2272
15e2396d
TH
2273static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2274{
2275 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2276 skb_gro_offset(skb));
2277}
2278
573e8fca
TH
2279static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2280 bool zero_okay,
2281 __sum16 check)
2282{
6edec0e6
TH
2283 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2284 skb_checksum_start_offset(skb) <
2285 skb_gro_offset(skb)) &&
15e2396d 2286 !skb_at_gro_remcsum_start(skb) &&
662880f4 2287 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2288 (!zero_okay || check));
2289}
2290
2291static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2292 __wsum psum)
2293{
2294 if (NAPI_GRO_CB(skb)->csum_valid &&
2295 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2296 return 0;
2297
2298 NAPI_GRO_CB(skb)->csum = psum;
2299
2300 return __skb_gro_checksum_complete(skb);
2301}
2302
573e8fca
TH
2303static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2304{
662880f4
TH
2305 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2306 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2307 NAPI_GRO_CB(skb)->csum_cnt--;
2308 } else {
2309 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2310 * verified a new top level checksum or an encapsulated one
2311 * during GRO. This saves work if we fallback to normal path.
2312 */
2313 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2314 }
2315}
2316
2317#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2318 compute_pseudo) \
2319({ \
2320 __sum16 __ret = 0; \
2321 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2322 __ret = __skb_gro_checksum_validate_complete(skb, \
2323 compute_pseudo(skb, proto)); \
5a212329
TH
2324 if (__ret) \
2325 __skb_mark_checksum_bad(skb); \
2326 else \
573e8fca
TH
2327 skb_gro_incr_csum_unnecessary(skb); \
2328 __ret; \
2329})
2330
2331#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2332 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2333
2334#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2335 compute_pseudo) \
2336 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2337
2338#define skb_gro_checksum_simple_validate(skb) \
2339 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2340
d96535a1
TH
2341static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2342{
2343 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2344 !NAPI_GRO_CB(skb)->csum_valid);
2345}
2346
2347static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2348 __sum16 check, __wsum pseudo)
2349{
2350 NAPI_GRO_CB(skb)->csum = ~pseudo;
2351 NAPI_GRO_CB(skb)->csum_valid = 1;
2352}
2353
2354#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2355do { \
2356 if (__skb_gro_checksum_convert_check(skb)) \
2357 __skb_gro_checksum_convert(skb, check, \
2358 compute_pseudo(skb, proto)); \
2359} while (0)
2360
26c4f7da
TH
2361struct gro_remcsum {
2362 int offset;
2363 __wsum delta;
2364};
2365
2366static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2367{
846cd667 2368 grc->offset = 0;
26c4f7da
TH
2369 grc->delta = 0;
2370}
2371
dcdc8994 2372static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
26c4f7da 2373 int start, int offset,
15e2396d
TH
2374 struct gro_remcsum *grc,
2375 bool nopartial)
dcdc8994
TH
2376{
2377 __wsum delta;
2378
2379 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2380
15e2396d
TH
2381 if (!nopartial) {
2382 NAPI_GRO_CB(skb)->gro_remcsum_start =
2383 ((unsigned char *)ptr + start) - skb->head;
2384 return;
2385 }
2386
dcdc8994
TH
2387 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2388
2389 /* Adjust skb->csum since we changed the packet */
dcdc8994 2390 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da
TH
2391
2392 grc->offset = (ptr + offset) - (void *)skb->head;
2393 grc->delta = delta;
dcdc8994
TH
2394}
2395
26c4f7da
TH
2396static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2397 struct gro_remcsum *grc)
2398{
2399 if (!grc->delta)
2400 return;
2401
2402 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2403}
dcdc8994 2404
0c4e8581
SH
2405static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2406 unsigned short type,
3b04ddde 2407 const void *daddr, const void *saddr,
95c96174 2408 unsigned int len)
0c4e8581 2409{
f1ecfd5d 2410 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2411 return 0;
3b04ddde
SH
2412
2413 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2414}
2415
b95cce35
SH
2416static inline int dev_parse_header(const struct sk_buff *skb,
2417 unsigned char *haddr)
2418{
2419 const struct net_device *dev = skb->dev;
2420
1b83336b 2421 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2422 return 0;
3b04ddde 2423 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2424}
2425
1da177e4 2426typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2427int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2428static inline int unregister_gifconf(unsigned int family)
2429{
2430 return register_gifconf(family, NULL);
2431}
2432
99bbc707 2433#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2434#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2435struct sd_flow_limit {
2436 u64 count;
2437 unsigned int num_buckets;
2438 unsigned int history_head;
2439 u16 history[FLOW_LIMIT_HISTORY];
2440 u8 buckets[];
2441};
2442
2443extern int netdev_flow_limit_table_len;
2444#endif /* CONFIG_NET_FLOW_LIMIT */
2445
1da177e4 2446/*
88751275 2447 * Incoming packets are placed on per-cpu queues
1da177e4 2448 */
d94d9fee 2449struct softnet_data {
1da177e4 2450 struct list_head poll_list;
6e7676c1 2451 struct sk_buff_head process_queue;
1da177e4 2452
dee42870 2453 /* stats */
cd7b5396
DM
2454 unsigned int processed;
2455 unsigned int time_squeeze;
2456 unsigned int cpu_collision;
2457 unsigned int received_rps;
fd793d89 2458#ifdef CONFIG_RPS
88751275 2459 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2460#endif
2461#ifdef CONFIG_NET_FLOW_LIMIT
2462 struct sd_flow_limit __rcu *flow_limit;
2463#endif
2464 struct Qdisc *output_queue;
2465 struct Qdisc **output_queue_tailp;
2466 struct sk_buff *completion_queue;
88751275 2467
4cdb1e2e 2468#ifdef CONFIG_RPS
88751275 2469 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 2470 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2471 struct softnet_data *rps_ipi_next;
2472 unsigned int cpu;
fec5e652 2473 unsigned int input_queue_head;
76cc8b13 2474 unsigned int input_queue_tail;
1e94d72f 2475#endif
95c96174 2476 unsigned int dropped;
0a9627f2 2477 struct sk_buff_head input_pkt_queue;
bea3348e 2478 struct napi_struct backlog;
99bbc707 2479
1da177e4
LT
2480};
2481
76cc8b13 2482static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2483{
2484#ifdef CONFIG_RPS
76cc8b13
TH
2485 sd->input_queue_head++;
2486#endif
2487}
2488
2489static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2490 unsigned int *qtail)
2491{
2492#ifdef CONFIG_RPS
2493 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2494#endif
2495}
2496
0a9627f2 2497DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2498
f629d208 2499void __netif_schedule(struct Qdisc *q);
46e5da40 2500void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2501
fd2ea0a7
DM
2502static inline void netif_tx_schedule_all(struct net_device *dev)
2503{
2504 unsigned int i;
2505
2506 for (i = 0; i < dev->num_tx_queues; i++)
2507 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2508}
2509
d29f749e
DJ
2510static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2511{
73466498 2512 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2513}
2514
bea3348e
SH
2515/**
2516 * netif_start_queue - allow transmit
2517 * @dev: network device
2518 *
2519 * Allow upper layers to call the device hard_start_xmit routine.
2520 */
1da177e4
LT
2521static inline void netif_start_queue(struct net_device *dev)
2522{
e8a0464c 2523 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2524}
2525
fd2ea0a7
DM
2526static inline void netif_tx_start_all_queues(struct net_device *dev)
2527{
2528 unsigned int i;
2529
2530 for (i = 0; i < dev->num_tx_queues; i++) {
2531 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2532 netif_tx_start_queue(txq);
2533 }
2534}
2535
46e5da40 2536void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2537
d29f749e
DJ
2538/**
2539 * netif_wake_queue - restart transmit
2540 * @dev: network device
2541 *
2542 * Allow upper layers to call the device hard_start_xmit routine.
2543 * Used for flow control when transmit resources are available.
2544 */
79d16385
DM
2545static inline void netif_wake_queue(struct net_device *dev)
2546{
e8a0464c 2547 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2548}
2549
fd2ea0a7
DM
2550static inline void netif_tx_wake_all_queues(struct net_device *dev)
2551{
2552 unsigned int i;
2553
2554 for (i = 0; i < dev->num_tx_queues; i++) {
2555 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2556 netif_tx_wake_queue(txq);
2557 }
2558}
2559
d29f749e
DJ
2560static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2561{
73466498 2562 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2563}
2564
bea3348e
SH
2565/**
2566 * netif_stop_queue - stop transmitted packets
2567 * @dev: network device
2568 *
2569 * Stop upper layers calling the device hard_start_xmit routine.
2570 * Used for flow control when transmit resources are unavailable.
2571 */
1da177e4
LT
2572static inline void netif_stop_queue(struct net_device *dev)
2573{
e8a0464c 2574 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2575}
2576
a2029240 2577void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2578
4d29515f 2579static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2580{
73466498 2581 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2582}
2583
bea3348e
SH
2584/**
2585 * netif_queue_stopped - test if transmit queue is flowblocked
2586 * @dev: network device
2587 *
2588 * Test if transmit queue on device is currently unable to send.
2589 */
4d29515f 2590static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2591{
e8a0464c 2592 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2593}
2594
4d29515f 2595static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2596{
73466498
TH
2597 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2598}
2599
8e2f1a63
DB
2600static inline bool
2601netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2602{
2603 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2604}
2605
8e2f1a63
DB
2606static inline bool
2607netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2608{
2609 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2610}
2611
53511453
ED
2612/**
2613 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2614 * @dev_queue: pointer to transmit queue
2615 *
2616 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2617 * to give appropriate hint to the cpu.
2618 */
2619static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2620{
2621#ifdef CONFIG_BQL
2622 prefetchw(&dev_queue->dql.num_queued);
2623#endif
2624}
2625
2626/**
2627 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2628 * @dev_queue: pointer to transmit queue
2629 *
2630 * BQL enabled drivers might use this helper in their TX completion path,
2631 * to give appropriate hint to the cpu.
2632 */
2633static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2634{
2635#ifdef CONFIG_BQL
2636 prefetchw(&dev_queue->dql.limit);
2637#endif
2638}
2639
c5d67bd7
TH
2640static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2641 unsigned int bytes)
2642{
114cf580
TH
2643#ifdef CONFIG_BQL
2644 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2645
2646 if (likely(dql_avail(&dev_queue->dql) >= 0))
2647 return;
2648
2649 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2650
2651 /*
2652 * The XOFF flag must be set before checking the dql_avail below,
2653 * because in netdev_tx_completed_queue we update the dql_completed
2654 * before checking the XOFF flag.
2655 */
2656 smp_mb();
2657
2658 /* check again in case another CPU has just made room avail */
2659 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2660 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2661#endif
c5d67bd7
TH
2662}
2663
0042d0c8
FF
2664/**
2665 * netdev_sent_queue - report the number of bytes queued to hardware
2666 * @dev: network device
2667 * @bytes: number of bytes queued to the hardware device queue
2668 *
2669 * Report the number of bytes queued for sending/completion to the network
2670 * device hardware queue. @bytes should be a good approximation and should
2671 * exactly match netdev_completed_queue() @bytes
2672 */
c5d67bd7
TH
2673static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2674{
2675 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2676}
2677
2678static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2679 unsigned int pkts, unsigned int bytes)
c5d67bd7 2680{
114cf580 2681#ifdef CONFIG_BQL
b37c0fbe
AD
2682 if (unlikely(!bytes))
2683 return;
2684
2685 dql_completed(&dev_queue->dql, bytes);
2686
2687 /*
2688 * Without the memory barrier there is a small possiblity that
2689 * netdev_tx_sent_queue will miss the update and cause the queue to
2690 * be stopped forever
2691 */
2692 smp_mb();
2693
2694 if (dql_avail(&dev_queue->dql) < 0)
2695 return;
2696
2697 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2698 netif_schedule_queue(dev_queue);
114cf580 2699#endif
c5d67bd7
TH
2700}
2701
0042d0c8
FF
2702/**
2703 * netdev_completed_queue - report bytes and packets completed by device
2704 * @dev: network device
2705 * @pkts: actual number of packets sent over the medium
2706 * @bytes: actual number of bytes sent over the medium
2707 *
2708 * Report the number of bytes and packets transmitted by the network device
2709 * hardware queue over the physical medium, @bytes must exactly match the
2710 * @bytes amount passed to netdev_sent_queue()
2711 */
c5d67bd7 2712static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2713 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2714{
2715 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2716}
2717
2718static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2719{
114cf580 2720#ifdef CONFIG_BQL
5c490354 2721 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2722 dql_reset(&q->dql);
2723#endif
c5d67bd7
TH
2724}
2725
0042d0c8
FF
2726/**
2727 * netdev_reset_queue - reset the packets and bytes count of a network device
2728 * @dev_queue: network device
2729 *
2730 * Reset the bytes and packet count of a network device and clear the
2731 * software flow control OFF bit for this network device
2732 */
c5d67bd7
TH
2733static inline void netdev_reset_queue(struct net_device *dev_queue)
2734{
2735 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2736}
2737
b9507bda
DB
2738/**
2739 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2740 * @dev: network device
2741 * @queue_index: given tx queue index
2742 *
2743 * Returns 0 if given tx queue index >= number of device tx queues,
2744 * otherwise returns the originally passed tx queue index.
2745 */
2746static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2747{
2748 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2749 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2750 dev->name, queue_index,
2751 dev->real_num_tx_queues);
2752 return 0;
2753 }
2754
2755 return queue_index;
2756}
2757
bea3348e
SH
2758/**
2759 * netif_running - test if up
2760 * @dev: network device
2761 *
2762 * Test if the device has been brought up.
2763 */
4d29515f 2764static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2765{
2766 return test_bit(__LINK_STATE_START, &dev->state);
2767}
2768
f25f4e44
PWJ
2769/*
2770 * Routines to manage the subqueues on a device. We only need start
2771 * stop, and a check if it's stopped. All other device management is
2772 * done at the overall netdevice level.
2773 * Also test the device if we're multiqueue.
2774 */
bea3348e
SH
2775
2776/**
2777 * netif_start_subqueue - allow sending packets on subqueue
2778 * @dev: network device
2779 * @queue_index: sub queue index
2780 *
2781 * Start individual transmit queue of a device with multiple transmit queues.
2782 */
f25f4e44
PWJ
2783static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2784{
fd2ea0a7 2785 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2786
2787 netif_tx_start_queue(txq);
f25f4e44
PWJ
2788}
2789
bea3348e
SH
2790/**
2791 * netif_stop_subqueue - stop sending packets on subqueue
2792 * @dev: network device
2793 * @queue_index: sub queue index
2794 *
2795 * Stop individual transmit queue of a device with multiple transmit queues.
2796 */
f25f4e44
PWJ
2797static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2798{
fd2ea0a7 2799 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 2800 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2801}
2802
bea3348e
SH
2803/**
2804 * netif_subqueue_stopped - test status of subqueue
2805 * @dev: network device
2806 * @queue_index: sub queue index
2807 *
2808 * Check individual transmit queue of a device with multiple transmit queues.
2809 */
4d29515f
DM
2810static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2811 u16 queue_index)
f25f4e44 2812{
fd2ea0a7 2813 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2814
2815 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2816}
2817
4d29515f
DM
2818static inline bool netif_subqueue_stopped(const struct net_device *dev,
2819 struct sk_buff *skb)
668f895a
PE
2820{
2821 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2822}
bea3348e 2823
46e5da40 2824void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 2825
537c00de 2826#ifdef CONFIG_XPS
53af53ae 2827int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2828 u16 index);
537c00de
AD
2829#else
2830static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2831 const struct cpumask *mask,
537c00de
AD
2832 u16 index)
2833{
2834 return 0;
2835}
2836#endif
2837
a3d22a68
VZ
2838/*
2839 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2840 * as a distribution range limit for the returned value.
2841 */
2842static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 2843 struct sk_buff *skb)
a3d22a68
VZ
2844{
2845 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2846}
2847
bea3348e
SH
2848/**
2849 * netif_is_multiqueue - test if device has multiple transmit queues
2850 * @dev: network device
2851 *
2852 * Check if device has multiple transmit queues
bea3348e 2853 */
4d29515f 2854static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2855{
a02cec21 2856 return dev->num_tx_queues > 1;
f25f4e44 2857}
1da177e4 2858
f629d208 2859int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2860
a953be53 2861#ifdef CONFIG_SYSFS
f629d208 2862int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
2863#else
2864static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2865 unsigned int rxq)
2866{
2867 return 0;
2868}
2869#endif
2870
a953be53
MD
2871#ifdef CONFIG_SYSFS
2872static inline unsigned int get_netdev_rx_queue_index(
2873 struct netdev_rx_queue *queue)
2874{
2875 struct net_device *dev = queue->dev;
2876 int index = queue - dev->_rx;
2877
2878 BUG_ON(index >= dev->num_rx_queues);
2879 return index;
2880}
2881#endif
2882
16917b87 2883#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 2884int netif_get_num_default_rss_queues(void);
16917b87 2885
e6247027
ED
2886enum skb_free_reason {
2887 SKB_REASON_CONSUMED,
2888 SKB_REASON_DROPPED,
2889};
2890
2891void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2892void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 2893
e6247027
ED
2894/*
2895 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2896 * interrupt context or with hardware interrupts being disabled.
2897 * (in_irq() || irqs_disabled())
2898 *
2899 * We provide four helpers that can be used in following contexts :
2900 *
2901 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2902 * replacing kfree_skb(skb)
2903 *
2904 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2905 * Typically used in place of consume_skb(skb) in TX completion path
2906 *
2907 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2908 * replacing kfree_skb(skb)
2909 *
2910 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2911 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 2912 */
e6247027
ED
2913static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2914{
2915 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2916}
2917
2918static inline void dev_consume_skb_irq(struct sk_buff *skb)
2919{
2920 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2921}
2922
2923static inline void dev_kfree_skb_any(struct sk_buff *skb)
2924{
2925 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2926}
2927
2928static inline void dev_consume_skb_any(struct sk_buff *skb)
2929{
2930 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2931}
1da177e4 2932
f629d208
JP
2933int netif_rx(struct sk_buff *skb);
2934int netif_rx_ni(struct sk_buff *skb);
7026b1dd
DM
2935int netif_receive_skb_sk(struct sock *sk, struct sk_buff *skb);
2936static inline int netif_receive_skb(struct sk_buff *skb)
2937{
2938 return netif_receive_skb_sk(skb->sk, skb);
2939}
f629d208
JP
2940gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2941void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2942struct sk_buff *napi_get_frags(struct napi_struct *napi);
2943gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
2944struct packet_offload *gro_find_receive_by_type(__be16 type);
2945struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
2946
2947static inline void napi_free_frags(struct napi_struct *napi)
2948{
2949 kfree_skb(napi->skb);
2950 napi->skb = NULL;
2951}
2952
f629d208
JP
2953int netdev_rx_handler_register(struct net_device *dev,
2954 rx_handler_func_t *rx_handler,
2955 void *rx_handler_data);
2956void netdev_rx_handler_unregister(struct net_device *dev);
2957
2958bool dev_valid_name(const char *name);
2959int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2960int dev_ethtool(struct net *net, struct ifreq *);
2961unsigned int dev_get_flags(const struct net_device *);
2962int __dev_change_flags(struct net_device *, unsigned int flags);
2963int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
2964void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2965 unsigned int gchanges);
f629d208
JP
2966int dev_change_name(struct net_device *, const char *);
2967int dev_set_alias(struct net_device *, const char *, size_t);
2968int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2969int dev_set_mtu(struct net_device *, int);
2970void dev_set_group(struct net_device *, int);
2971int dev_set_mac_address(struct net_device *, struct sockaddr *);
2972int dev_change_carrier(struct net_device *, bool new_carrier);
2973int dev_get_phys_port_id(struct net_device *dev,
02637fce 2974 struct netdev_phys_item_id *ppid);
db24a904
DA
2975int dev_get_phys_port_name(struct net_device *dev,
2976 char *name, size_t len);
55a93b3e 2977struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
2978struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2979 struct netdev_queue *txq, int *ret);
a0265d28 2980int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 2981int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1ee481fb 2982bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
1da177e4 2983
20380731 2984extern int netdev_budget;
1da177e4
LT
2985
2986/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 2987void netdev_run_todo(void);
1da177e4 2988
bea3348e
SH
2989/**
2990 * dev_put - release reference to device
2991 * @dev: network device
2992 *
9ef4429b 2993 * Release reference to device to allow it to be freed.
bea3348e 2994 */
1da177e4
LT
2995static inline void dev_put(struct net_device *dev)
2996{
933393f5 2997 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
2998}
2999
bea3348e
SH
3000/**
3001 * dev_hold - get reference to device
3002 * @dev: network device
3003 *
9ef4429b 3004 * Hold reference to device to keep it from being freed.
bea3348e 3005 */
15333061
SH
3006static inline void dev_hold(struct net_device *dev)
3007{
933393f5 3008 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3009}
1da177e4
LT
3010
3011/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3012 * and _off may be called from IRQ context, but it is caller
3013 * who is responsible for serialization of these calls.
b00055aa
SR
3014 *
3015 * The name carrier is inappropriate, these functions should really be
3016 * called netif_lowerlayer_*() because they represent the state of any
3017 * kind of lower layer not just hardware media.
1da177e4
LT
3018 */
3019
f629d208
JP
3020void linkwatch_init_dev(struct net_device *dev);
3021void linkwatch_fire_event(struct net_device *dev);
3022void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3023
bea3348e
SH
3024/**
3025 * netif_carrier_ok - test if carrier present
3026 * @dev: network device
3027 *
3028 * Check if carrier is present on device
3029 */
4d29515f 3030static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3031{
3032 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3033}
3034
f629d208 3035unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3036
f629d208 3037void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3038
f629d208 3039void netif_carrier_on(struct net_device *dev);
1da177e4 3040
f629d208 3041void netif_carrier_off(struct net_device *dev);
1da177e4 3042
bea3348e
SH
3043/**
3044 * netif_dormant_on - mark device as dormant.
3045 * @dev: network device
3046 *
3047 * Mark device as dormant (as per RFC2863).
3048 *
3049 * The dormant state indicates that the relevant interface is not
3050 * actually in a condition to pass packets (i.e., it is not 'up') but is
3051 * in a "pending" state, waiting for some external event. For "on-
3052 * demand" interfaces, this new state identifies the situation where the
3053 * interface is waiting for events to place it in the up state.
3054 *
3055 */
b00055aa
SR
3056static inline void netif_dormant_on(struct net_device *dev)
3057{
3058 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3059 linkwatch_fire_event(dev);
3060}
3061
bea3348e
SH
3062/**
3063 * netif_dormant_off - set device as not dormant.
3064 * @dev: network device
3065 *
3066 * Device is not in dormant state.
3067 */
b00055aa
SR
3068static inline void netif_dormant_off(struct net_device *dev)
3069{
3070 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3071 linkwatch_fire_event(dev);
3072}
3073
bea3348e
SH
3074/**
3075 * netif_dormant - test if carrier present
3076 * @dev: network device
3077 *
3078 * Check if carrier is present on device
3079 */
4d29515f 3080static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3081{
3082 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3083}
3084
3085
bea3348e
SH
3086/**
3087 * netif_oper_up - test if device is operational
3088 * @dev: network device
3089 *
3090 * Check if carrier is operational
3091 */
4d29515f 3092static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3093{
b00055aa
SR
3094 return (dev->operstate == IF_OPER_UP ||
3095 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3096}
3097
bea3348e
SH
3098/**
3099 * netif_device_present - is device available or removed
3100 * @dev: network device
3101 *
3102 * Check if device has not been removed from system.
3103 */
4d29515f 3104static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3105{
3106 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3107}
3108
f629d208 3109void netif_device_detach(struct net_device *dev);
1da177e4 3110
f629d208 3111void netif_device_attach(struct net_device *dev);
1da177e4
LT
3112
3113/*
3114 * Network interface message level settings
3115 */
1da177e4
LT
3116
3117enum {
3118 NETIF_MSG_DRV = 0x0001,
3119 NETIF_MSG_PROBE = 0x0002,
3120 NETIF_MSG_LINK = 0x0004,
3121 NETIF_MSG_TIMER = 0x0008,
3122 NETIF_MSG_IFDOWN = 0x0010,
3123 NETIF_MSG_IFUP = 0x0020,
3124 NETIF_MSG_RX_ERR = 0x0040,
3125 NETIF_MSG_TX_ERR = 0x0080,
3126 NETIF_MSG_TX_QUEUED = 0x0100,
3127 NETIF_MSG_INTR = 0x0200,
3128 NETIF_MSG_TX_DONE = 0x0400,
3129 NETIF_MSG_RX_STATUS = 0x0800,
3130 NETIF_MSG_PKTDATA = 0x1000,
3131 NETIF_MSG_HW = 0x2000,
3132 NETIF_MSG_WOL = 0x4000,
3133};
3134
3135#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3136#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3137#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3138#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3139#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3140#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3141#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3142#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3143#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3144#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3145#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3146#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3147#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3148#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3149#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3150
3151static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3152{
3153 /* use default */
3154 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3155 return default_msg_enable_bits;
3156 if (debug_value == 0) /* no output */
3157 return 0;
3158 /* set low N bits */
3159 return (1 << debug_value) - 1;
3160}
3161
c773e847 3162static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3163{
c773e847
DM
3164 spin_lock(&txq->_xmit_lock);
3165 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3166}
3167
fd2ea0a7
DM
3168static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3169{
3170 spin_lock_bh(&txq->_xmit_lock);
3171 txq->xmit_lock_owner = smp_processor_id();
3172}
3173
4d29515f 3174static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3175{
4d29515f 3176 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3177 if (likely(ok))
3178 txq->xmit_lock_owner = smp_processor_id();
3179 return ok;
3180}
3181
3182static inline void __netif_tx_unlock(struct netdev_queue *txq)
3183{
3184 txq->xmit_lock_owner = -1;
3185 spin_unlock(&txq->_xmit_lock);
3186}
3187
3188static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3189{
3190 txq->xmit_lock_owner = -1;
3191 spin_unlock_bh(&txq->_xmit_lock);
3192}
3193
08baf561
ED
3194static inline void txq_trans_update(struct netdev_queue *txq)
3195{
3196 if (txq->xmit_lock_owner != -1)
3197 txq->trans_start = jiffies;
3198}
3199
d29f749e
DJ
3200/**
3201 * netif_tx_lock - grab network device transmit lock
3202 * @dev: network device
d29f749e
DJ
3203 *
3204 * Get network device transmit lock
3205 */
22dd7495
JHS
3206static inline void netif_tx_lock(struct net_device *dev)
3207{
e8a0464c 3208 unsigned int i;
c3f26a26 3209 int cpu;
c773e847 3210
c3f26a26
DM
3211 spin_lock(&dev->tx_global_lock);
3212 cpu = smp_processor_id();
e8a0464c
DM
3213 for (i = 0; i < dev->num_tx_queues; i++) {
3214 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3215
3216 /* We are the only thread of execution doing a
3217 * freeze, but we have to grab the _xmit_lock in
3218 * order to synchronize with threads which are in
3219 * the ->hard_start_xmit() handler and already
3220 * checked the frozen bit.
3221 */
e8a0464c 3222 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3223 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3224 __netif_tx_unlock(txq);
e8a0464c 3225 }
932ff279
HX
3226}
3227
3228static inline void netif_tx_lock_bh(struct net_device *dev)
3229{
e8a0464c
DM
3230 local_bh_disable();
3231 netif_tx_lock(dev);
932ff279
HX
3232}
3233
932ff279
HX
3234static inline void netif_tx_unlock(struct net_device *dev)
3235{
e8a0464c
DM
3236 unsigned int i;
3237
3238 for (i = 0; i < dev->num_tx_queues; i++) {
3239 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3240
c3f26a26
DM
3241 /* No need to grab the _xmit_lock here. If the
3242 * queue is not stopped for another reason, we
3243 * force a schedule.
3244 */
3245 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3246 netif_schedule_queue(txq);
c3f26a26
DM
3247 }
3248 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3249}
3250
3251static inline void netif_tx_unlock_bh(struct net_device *dev)
3252{
e8a0464c
DM
3253 netif_tx_unlock(dev);
3254 local_bh_enable();
932ff279
HX
3255}
3256
c773e847 3257#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3258 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3259 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3260 } \
3261}
3262
5efeac44
EB
3263#define HARD_TX_TRYLOCK(dev, txq) \
3264 (((dev->features & NETIF_F_LLTX) == 0) ? \
3265 __netif_tx_trylock(txq) : \
3266 true )
3267
c773e847 3268#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3269 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3270 __netif_tx_unlock(txq); \
22dd7495
JHS
3271 } \
3272}
3273
1da177e4
LT
3274static inline void netif_tx_disable(struct net_device *dev)
3275{
fd2ea0a7 3276 unsigned int i;
c3f26a26 3277 int cpu;
fd2ea0a7 3278
c3f26a26
DM
3279 local_bh_disable();
3280 cpu = smp_processor_id();
fd2ea0a7
DM
3281 for (i = 0; i < dev->num_tx_queues; i++) {
3282 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3283
3284 __netif_tx_lock(txq, cpu);
fd2ea0a7 3285 netif_tx_stop_queue(txq);
c3f26a26 3286 __netif_tx_unlock(txq);
fd2ea0a7 3287 }
c3f26a26 3288 local_bh_enable();
1da177e4
LT
3289}
3290
e308a5d8
DM
3291static inline void netif_addr_lock(struct net_device *dev)
3292{
3293 spin_lock(&dev->addr_list_lock);
3294}
3295
2429f7ac
JP
3296static inline void netif_addr_lock_nested(struct net_device *dev)
3297{
25175ba5
VY
3298 int subclass = SINGLE_DEPTH_NESTING;
3299
3300 if (dev->netdev_ops->ndo_get_lock_subclass)
3301 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3302
3303 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3304}
3305
e308a5d8
DM
3306static inline void netif_addr_lock_bh(struct net_device *dev)
3307{
3308 spin_lock_bh(&dev->addr_list_lock);
3309}
3310
3311static inline void netif_addr_unlock(struct net_device *dev)
3312{
3313 spin_unlock(&dev->addr_list_lock);
3314}
3315
3316static inline void netif_addr_unlock_bh(struct net_device *dev)
3317{
3318 spin_unlock_bh(&dev->addr_list_lock);
3319}
3320
f001fde5 3321/*
31278e71 3322 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3323 * rcu_read_lock held.
3324 */
3325#define for_each_dev_addr(dev, ha) \
31278e71 3326 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3327
1da177e4
LT
3328/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3329
f629d208 3330void ether_setup(struct net_device *dev);
1da177e4
LT
3331
3332/* Support for loadable net-drivers */
f629d208 3333struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3334 unsigned char name_assign_type,
f629d208
JP
3335 void (*setup)(struct net_device *),
3336 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3337#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3338 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3339
c835a677
TG
3340#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3341 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3342 count)
36909ea4 3343
f629d208
JP
3344int register_netdev(struct net_device *dev);
3345void unregister_netdev(struct net_device *dev);
f001fde5 3346
22bedad3 3347/* General hardware address lists handling functions */
f629d208
JP
3348int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3349 struct netdev_hw_addr_list *from_list, int addr_len);
3350void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3351 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3352int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3353 struct net_device *dev,
3354 int (*sync)(struct net_device *, const unsigned char *),
3355 int (*unsync)(struct net_device *,
3356 const unsigned char *));
3357void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3358 struct net_device *dev,
3359 int (*unsync)(struct net_device *,
3360 const unsigned char *));
f629d208 3361void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3362
f001fde5 3363/* Functions used for device addresses handling */
f629d208
JP
3364int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3365 unsigned char addr_type);
3366int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3367 unsigned char addr_type);
f629d208
JP
3368void dev_addr_flush(struct net_device *dev);
3369int dev_addr_init(struct net_device *dev);
a748ee24
JP
3370
3371/* Functions used for unicast addresses handling */
f629d208
JP
3372int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3373int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3374int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3375int dev_uc_sync(struct net_device *to, struct net_device *from);
3376int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3377void dev_uc_unsync(struct net_device *to, struct net_device *from);
3378void dev_uc_flush(struct net_device *dev);
3379void dev_uc_init(struct net_device *dev);
f001fde5 3380
670e5b8e
AD
3381/**
3382 * __dev_uc_sync - Synchonize device's unicast list
3383 * @dev: device to sync
3384 * @sync: function to call if address should be added
3385 * @unsync: function to call if address should be removed
3386 *
3387 * Add newly added addresses to the interface, and release
3388 * addresses that have been deleted.
3389 **/
3390static inline int __dev_uc_sync(struct net_device *dev,
3391 int (*sync)(struct net_device *,
3392 const unsigned char *),
3393 int (*unsync)(struct net_device *,
3394 const unsigned char *))
3395{
3396 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3397}
3398
3399/**
e793c0f7 3400 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3401 * @dev: device to sync
3402 * @unsync: function to call if address should be removed
3403 *
3404 * Remove all addresses that were added to the device by dev_uc_sync().
3405 **/
3406static inline void __dev_uc_unsync(struct net_device *dev,
3407 int (*unsync)(struct net_device *,
3408 const unsigned char *))
3409{
3410 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3411}
3412
22bedad3 3413/* Functions used for multicast addresses handling */
f629d208
JP
3414int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3415int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3416int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3417int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3418int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3419int dev_mc_sync(struct net_device *to, struct net_device *from);
3420int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3421void dev_mc_unsync(struct net_device *to, struct net_device *from);
3422void dev_mc_flush(struct net_device *dev);
3423void dev_mc_init(struct net_device *dev);
f001fde5 3424
670e5b8e
AD
3425/**
3426 * __dev_mc_sync - Synchonize device's multicast list
3427 * @dev: device to sync
3428 * @sync: function to call if address should be added
3429 * @unsync: function to call if address should be removed
3430 *
3431 * Add newly added addresses to the interface, and release
3432 * addresses that have been deleted.
3433 **/
3434static inline int __dev_mc_sync(struct net_device *dev,
3435 int (*sync)(struct net_device *,
3436 const unsigned char *),
3437 int (*unsync)(struct net_device *,
3438 const unsigned char *))
3439{
3440 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3441}
3442
3443/**
e793c0f7 3444 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3445 * @dev: device to sync
3446 * @unsync: function to call if address should be removed
3447 *
3448 * Remove all addresses that were added to the device by dev_mc_sync().
3449 **/
3450static inline void __dev_mc_unsync(struct net_device *dev,
3451 int (*unsync)(struct net_device *,
3452 const unsigned char *))
3453{
3454 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3455}
3456
4417da66 3457/* Functions used for secondary unicast and multicast support */
f629d208
JP
3458void dev_set_rx_mode(struct net_device *dev);
3459void __dev_set_rx_mode(struct net_device *dev);
3460int dev_set_promiscuity(struct net_device *dev, int inc);
3461int dev_set_allmulti(struct net_device *dev, int inc);
3462void netdev_state_change(struct net_device *dev);
3463void netdev_notify_peers(struct net_device *dev);
3464void netdev_features_change(struct net_device *dev);
1da177e4 3465/* Load a device via the kmod */
f629d208
JP
3466void dev_load(struct net *net, const char *name);
3467struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3468 struct rtnl_link_stats64 *storage);
3469void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3470 const struct net_device_stats *netdev_stats);
eeda3fd6 3471
1da177e4 3472extern int netdev_max_backlog;
3b098e2d 3473extern int netdev_tstamp_prequeue;
1da177e4 3474extern int weight_p;
0a14842f 3475extern int bpf_jit_enable;
9ff162a8 3476
f629d208 3477bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3478struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3479 struct list_head **iter);
f629d208
JP
3480struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3481 struct list_head **iter);
8b5be856 3482
44a40855
VY
3483/* iterate through upper list, must be called under RCU read lock */
3484#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3485 for (iter = &(dev)->adj_list.upper, \
3486 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3487 updev; \
3488 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3489
8b5be856 3490/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3491#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3492 for (iter = &(dev)->all_adj_list.upper, \
3493 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3494 updev; \
3495 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3496
f629d208
JP
3497void *netdev_lower_get_next_private(struct net_device *dev,
3498 struct list_head **iter);
3499void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3500 struct list_head **iter);
31088a11
VF
3501
3502#define netdev_for_each_lower_private(dev, priv, iter) \
3503 for (iter = (dev)->adj_list.lower.next, \
3504 priv = netdev_lower_get_next_private(dev, &(iter)); \
3505 priv; \
3506 priv = netdev_lower_get_next_private(dev, &(iter)))
3507
3508#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3509 for (iter = &(dev)->adj_list.lower, \
3510 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3511 priv; \
3512 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3513
4085ebe8
VY
3514void *netdev_lower_get_next(struct net_device *dev,
3515 struct list_head **iter);
3516#define netdev_for_each_lower_dev(dev, ldev, iter) \
3517 for (iter = &(dev)->adj_list.lower, \
3518 ldev = netdev_lower_get_next(dev, &(iter)); \
3519 ldev; \
3520 ldev = netdev_lower_get_next(dev, &(iter)))
3521
f629d208 3522void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3523void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3524struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3525struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3526int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3527int netdev_master_upper_dev_link(struct net_device *dev,
9ff162a8 3528 struct net_device *upper_dev);
f629d208
JP
3529int netdev_master_upper_dev_link_private(struct net_device *dev,
3530 struct net_device *upper_dev,
3531 void *private);
3532void netdev_upper_dev_unlink(struct net_device *dev,
3533 struct net_device *upper_dev);
5bb025fa 3534void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3535void *netdev_lower_dev_get_private(struct net_device *dev,
3536 struct net_device *lower_dev);
960fb622
ED
3537
3538/* RSS keys are 40 or 52 bytes long */
3539#define NETDEV_RSS_KEY_LEN 52
3540extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3541void netdev_rss_key_fill(void *buffer, size_t len);
3542
4085ebe8
VY
3543int dev_get_nest_level(struct net_device *dev,
3544 bool (*type_check)(struct net_device *dev));
f629d208
JP
3545int skb_checksum_help(struct sk_buff *skb);
3546struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3547 netdev_features_t features, bool tx_path);
3548struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3549 netdev_features_t features);
12b0004d 3550
61bd3857
MS
3551struct netdev_bonding_info {
3552 ifslave slave;
3553 ifbond master;
3554};
3555
3556struct netdev_notifier_bonding_info {
3557 struct netdev_notifier_info info; /* must be first */
3558 struct netdev_bonding_info bonding_info;
3559};
3560
3561void netdev_bonding_info_change(struct net_device *dev,
3562 struct netdev_bonding_info *bonding_info);
3563
12b0004d
CW
3564static inline
3565struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3566{
3567 return __skb_gso_segment(skb, features, true);
3568}
53d6471c 3569__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3570
3571static inline bool can_checksum_protocol(netdev_features_t features,
3572 __be16 protocol)
3573{
3574 return ((features & NETIF_F_GEN_CSUM) ||
3575 ((features & NETIF_F_V4_CSUM) &&
3576 protocol == htons(ETH_P_IP)) ||
3577 ((features & NETIF_F_V6_CSUM) &&
3578 protocol == htons(ETH_P_IPV6)) ||
3579 ((features & NETIF_F_FCOE_CRC) &&
3580 protocol == htons(ETH_P_FCOE)));
3581}
12b0004d 3582
fb286bb2 3583#ifdef CONFIG_BUG
f629d208 3584void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3585#else
3586static inline void netdev_rx_csum_fault(struct net_device *dev)
3587{
3588}
3589#endif
1da177e4 3590/* rx skb timestamps */
f629d208
JP
3591void net_enable_timestamp(void);
3592void net_disable_timestamp(void);
1da177e4 3593
20380731 3594#ifdef CONFIG_PROC_FS
f629d208 3595int __init dev_proc_init(void);
900ff8c6
CW
3596#else
3597#define dev_proc_init() 0
20380731
ACM
3598#endif
3599
4798248e 3600static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3601 struct sk_buff *skb, struct net_device *dev,
3602 bool more)
4798248e 3603{
fa2dbdc2 3604 skb->xmit_more = more ? 1 : 0;
0b725a2c 3605 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3606}
3607
10b3ad8c 3608static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3609 struct netdev_queue *txq, bool more)
4798248e
DM
3610{
3611 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3612 int rc;
4798248e 3613
fa2dbdc2 3614 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3615 if (rc == NETDEV_TX_OK)
3616 txq_trans_update(txq);
3617
3618 return rc;
4798248e
DM
3619}
3620
42a2d923
LT
3621int netdev_class_create_file_ns(struct class_attribute *class_attr,
3622 const void *ns);
3623void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3624 const void *ns);
58292cbe
TH
3625
3626static inline int netdev_class_create_file(struct class_attribute *class_attr)
3627{
3628 return netdev_class_create_file_ns(class_attr, NULL);
3629}
3630
3631static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3632{
3633 netdev_class_remove_file_ns(class_attr, NULL);
3634}
b8a9787e 3635
04600794
JB
3636extern struct kobj_ns_type_operations net_ns_type_operations;
3637
f629d208 3638const char *netdev_drivername(const struct net_device *dev);
6579e57b 3639
f629d208 3640void linkwatch_run_queue(void);
20380731 3641
da08143b
MK
3642static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3643 netdev_features_t f2)
3644{
3645 if (f1 & NETIF_F_GEN_CSUM)
3646 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3647 if (f2 & NETIF_F_GEN_CSUM)
3648 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3649 f1 &= f2;
3650 if (f1 & NETIF_F_GEN_CSUM)
3651 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3652
3653 return f1;
3654}
3655
c8f44aff
MM
3656static inline netdev_features_t netdev_get_wanted_features(
3657 struct net_device *dev)
5455c699
MM
3658{
3659 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3660}
c8f44aff
MM
3661netdev_features_t netdev_increment_features(netdev_features_t all,
3662 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3663
3664/* Allow TSO being used on stacked device :
3665 * Performing the GSO segmentation before last device
3666 * is a performance improvement.
3667 */
3668static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3669 netdev_features_t mask)
3670{
3671 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3672}
3673
6cb6a27c 3674int __netdev_update_features(struct net_device *dev);
5455c699 3675void netdev_update_features(struct net_device *dev);
afe12cc8 3676void netdev_change_features(struct net_device *dev);
7f353bf2 3677
fc4a7489
PM
3678void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3679 struct net_device *dev);
3680
e38f3025
TM
3681netdev_features_t passthru_features_check(struct sk_buff *skb,
3682 struct net_device *dev,
3683 netdev_features_t features);
c1e756bf 3684netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3685
4d29515f 3686static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3687{
c8f44aff 3688 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3689
3690 /* check flags correspondence */
3691 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3692 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3693 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3694 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3695 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3696 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
3697 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3698 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3699 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3700 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3701 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3702 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
e585f236 3703 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
0345e186 3704
d6b4991a 3705 return (features & feature) == feature;
576a30eb
HX
3706}
3707
4d29515f 3708static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3709{
278b2513 3710 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3711 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3712}
3713
8b86a61d 3714static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 3715 netdev_features_t features)
7967168c 3716{
fc741216 3717 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3718 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3719 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3720}
3721
82cc1a7a
PWJ
3722static inline void netif_set_gso_max_size(struct net_device *dev,
3723 unsigned int size)
3724{
3725 dev->gso_max_size = size;
3726}
3727
7a7ffbab
WCC
3728static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3729 int pulled_hlen, u16 mac_offset,
3730 int mac_len)
3731{
3732 skb->protocol = protocol;
3733 skb->encapsulation = 1;
3734 skb_push(skb, pulled_hlen);
3735 skb_reset_transport_header(skb);
3736 skb->mac_header = mac_offset;
3737 skb->network_header = skb->mac_header + mac_len;
3738 skb->mac_len = mac_len;
3739}
3740
a6cc0cfa
JF
3741static inline bool netif_is_macvlan(struct net_device *dev)
3742{
3743 return dev->priv_flags & IFF_MACVLAN;
3744}
3745
2f33e7d5
MB
3746static inline bool netif_is_macvlan_port(struct net_device *dev)
3747{
3748 return dev->priv_flags & IFF_MACVLAN_PORT;
3749}
3750
5933fea7
MB
3751static inline bool netif_is_ipvlan(struct net_device *dev)
3752{
3753 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3754}
3755
3756static inline bool netif_is_ipvlan_port(struct net_device *dev)
3757{
3758 return dev->priv_flags & IFF_IPVLAN_MASTER;
3759}
3760
8a7fbfab 3761static inline bool netif_is_bond_master(struct net_device *dev)
3762{
3763 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3764}
3765
4d29515f 3766static inline bool netif_is_bond_slave(struct net_device *dev)
1765a575
JP
3767{
3768 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3769}
3770
3bdc0eba
BG
3771static inline bool netif_supports_nofcs(struct net_device *dev)
3772{
3773 return dev->priv_flags & IFF_SUPP_NOFCS;
3774}
3775
02875878
ED
3776/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3777static inline void netif_keep_dst(struct net_device *dev)
3778{
3779 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3780}
3781
505d4f73 3782extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 3783
571ba423
JP
3784/* Logging, debugging and troubleshooting/diagnostic helpers. */
3785
3786/* netdev_printk helpers, similar to dev_printk */
3787
3788static inline const char *netdev_name(const struct net_device *dev)
3789{
c6f854d5
VF
3790 if (!dev->name[0] || strchr(dev->name, '%'))
3791 return "(unnamed net_device)";
571ba423
JP
3792 return dev->name;
3793}
3794
ccc7f496
VF
3795static inline const char *netdev_reg_state(const struct net_device *dev)
3796{
3797 switch (dev->reg_state) {
3798 case NETREG_UNINITIALIZED: return " (uninitialized)";
3799 case NETREG_REGISTERED: return "";
3800 case NETREG_UNREGISTERING: return " (unregistering)";
3801 case NETREG_UNREGISTERED: return " (unregistered)";
3802 case NETREG_RELEASED: return " (released)";
3803 case NETREG_DUMMY: return " (dummy)";
3804 }
3805
3806 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3807 return " (unknown)";
3808}
3809
f629d208 3810__printf(3, 4)
6ea754eb
JP
3811void netdev_printk(const char *level, const struct net_device *dev,
3812 const char *format, ...);
f629d208 3813__printf(2, 3)
6ea754eb 3814void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 3815__printf(2, 3)
6ea754eb 3816void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 3817__printf(2, 3)
6ea754eb 3818void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 3819__printf(2, 3)
6ea754eb 3820void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 3821__printf(2, 3)
6ea754eb 3822void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 3823__printf(2, 3)
6ea754eb 3824void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 3825__printf(2, 3)
6ea754eb 3826void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 3827
8909c9ad
VK
3828#define MODULE_ALIAS_NETDEV(device) \
3829 MODULE_ALIAS("netdev-" device)
3830
b558c96f 3831#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
3832#define netdev_dbg(__dev, format, args...) \
3833do { \
ffa10cb4 3834 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 3835} while (0)
b558c96f
JC
3836#elif defined(DEBUG)
3837#define netdev_dbg(__dev, format, args...) \
3838 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
3839#else
3840#define netdev_dbg(__dev, format, args...) \
3841({ \
3842 if (0) \
3843 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
3844})
3845#endif
3846
3847#if defined(VERBOSE_DEBUG)
3848#define netdev_vdbg netdev_dbg
3849#else
3850
3851#define netdev_vdbg(dev, format, args...) \
3852({ \
3853 if (0) \
3854 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3855 0; \
3856})
3857#endif
3858
3859/*
3860 * netdev_WARN() acts like dev_printk(), but with the key difference
3861 * of using a WARN/WARN_ON to get the message out, including the
3862 * file/line information and a backtrace.
3863 */
3864#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
3865 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3866 netdev_reg_state(dev), ##args)
571ba423 3867
b3d95c5c
JP
3868/* netif printk helpers, similar to netdev_printk */
3869
3870#define netif_printk(priv, type, level, dev, fmt, args...) \
3871do { \
3872 if (netif_msg_##type(priv)) \
3873 netdev_printk(level, (dev), fmt, ##args); \
3874} while (0)
3875
f45f4321
JP
3876#define netif_level(level, priv, type, dev, fmt, args...) \
3877do { \
3878 if (netif_msg_##type(priv)) \
3879 netdev_##level(dev, fmt, ##args); \
3880} while (0)
3881
b3d95c5c 3882#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 3883 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 3884#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 3885 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 3886#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 3887 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 3888#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 3889 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 3890#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 3891 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 3892#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 3893 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 3894#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 3895 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 3896
0053ea9c 3897#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
3898#define netif_dbg(priv, type, netdev, format, args...) \
3899do { \
3900 if (netif_msg_##type(priv)) \
b5fb0a03 3901 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 3902} while (0)
0053ea9c
JP
3903#elif defined(DEBUG)
3904#define netif_dbg(priv, type, dev, format, args...) \
3905 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
3906#else
3907#define netif_dbg(priv, type, dev, format, args...) \
3908({ \
3909 if (0) \
3910 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3911 0; \
3912})
3913#endif
3914
3915#if defined(VERBOSE_DEBUG)
bcfcc450 3916#define netif_vdbg netif_dbg
b3d95c5c
JP
3917#else
3918#define netif_vdbg(priv, type, dev, format, args...) \
3919({ \
3920 if (0) \
a4ed89cb 3921 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
3922 0; \
3923})
3924#endif
571ba423 3925
900ff8c6
CW
3926/*
3927 * The list of packet types we will receive (as opposed to discard)
3928 * and the routines to invoke.
3929 *
3930 * Why 16. Because with 16 the only overlap we get on a hash of the
3931 * low nibble of the protocol value is RARP/SNAP/X.25.
3932 *
3933 * NOTE: That is no longer true with the addition of VLAN tags. Not
3934 * sure which should go first, but I bet it won't make much
3935 * difference if we are running VLANs. The good news is that
3936 * this protocol won't be in the list unless compiled in, so
3937 * the average user (w/out VLANs) will not be adversely affected.
3938 * --BLG
3939 *
3940 * 0800 IP
3941 * 8100 802.1Q VLAN
3942 * 0001 802.3
3943 * 0002 AX.25
3944 * 0004 802.2
3945 * 8035 RARP
3946 * 0005 SNAP
3947 * 0805 X.25
3948 * 0806 ARP
3949 * 8137 IPX
3950 * 0009 Localtalk
3951 * 86DD IPv6
3952 */
3953#define PTYPE_HASH_SIZE (16)
3954#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3955
385a154c 3956#endif /* _LINUX_NETDEVICE_H */
This page took 1.67037 seconds and 5 git commands to generate.