openvswitch: Fix egress tunnel info.
[linux-2.6-block.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
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
cf85d08f 44#include <net/dsa.h>
7a6b6f51 45#ifdef CONFIG_DCB
2f90b865
AD
46#include <net/dcbnl.h>
47#endif
5bc1421e 48#include <net/netprio_cgroup.h>
a050c33f 49
a59e2ecb 50#include <linux/netdev_features.h>
77162022 51#include <linux/neighbour.h>
607ca46e 52#include <uapi/linux/netdevice.h>
61bd3857 53#include <uapi/linux/if_bonding.h>
a59e2ecb 54
115c1d6e 55struct netpoll_info;
313162d0 56struct device;
c1f19b51 57struct phy_device;
704232c2
JB
58/* 802.11 specific */
59struct wireless_dev;
98a18b6f
AA
60/* 802.15.4 specific */
61struct wpan_dev;
03c57747 62struct mpls_dev;
1da177e4 63
f629d208
JP
64void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
d07d7507 66
9a1654ba
JP
67/* Backlog congestion levels */
68#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69#define NET_RX_DROP 1 /* packet dropped */
70
572a9d7b
PM
71/*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88/* qdisc ->enqueue() return codes. */
89#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
90#define NET_XMIT_DROP 0x01 /* skb dropped */
91#define NET_XMIT_CN 0x02 /* congestion notification */
92#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 94
b9df3cb8
GR
95/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 98#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
99#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
dc1f8bf6 101/* Driver transmit return codes */
9a1654ba 102#define NETDEV_TX_MASK 0xf0
572a9d7b 103
dc1f8bf6 104enum netdev_tx {
572a9d7b 105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
dc1f8bf6
SH
109};
110typedef enum netdev_tx netdev_tx_t;
111
9a1654ba
JP
112/*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116static inline bool dev_xmit_complete(int rc)
117{
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128}
129
1da177e4
LT
130/*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
fe2918b0 134
d11ead75 135#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
136# if defined(CONFIG_MAC80211_MESH)
137# define LL_MAX_HEADER 128
138# else
139# define LL_MAX_HEADER 96
140# endif
1da177e4 141#else
8388e3da 142# define LL_MAX_HEADER 32
1da177e4
LT
143#endif
144
d11ead75
BH
145#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
147#define MAX_HEADER LL_MAX_HEADER
148#else
149#define MAX_HEADER (LL_MAX_HEADER + 48)
150#endif
151
152/*
be1f3c2c
BH
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
1da177e4 155 */
fe2918b0 156
d94d9fee 157struct net_device_stats {
3cfde79c
BH
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
1da177e4 167 unsigned long collisions;
1da177e4 168 unsigned long rx_length_errors;
3cfde79c
BH
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
1da177e4
LT
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
1da177e4
LT
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
181};
182
1da177e4
LT
183
184#include <linux/cache.h>
185#include <linux/skbuff.h>
186
adc9300e 187#ifdef CONFIG_RPS
c5905afb
IM
188#include <linux/static_key.h>
189extern struct static_key rps_needed;
adc9300e
ED
190#endif
191
1da177e4
LT
192struct neighbour;
193struct neigh_parms;
194struct sk_buff;
195
f001fde5
JP
196struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
199 unsigned char type;
ccffad25
JP
200#define NETDEV_HW_ADDR_T_LAN 1
201#define NETDEV_HW_ADDR_T_SAN 2
202#define NETDEV_HW_ADDR_T_SLAVE 3
203#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 204#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 205 bool global_use;
4cd729b0 206 int sync_cnt;
8f8f103d 207 int refcount;
4543fbef 208 int synced;
f001fde5
JP
209 struct rcu_head rcu_head;
210};
211
31278e71
JP
212struct netdev_hw_addr_list {
213 struct list_head list;
214 int count;
215};
216
22bedad3
JP
217#define netdev_hw_addr_list_count(l) ((l)->count)
218#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219#define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 221
22bedad3
JP
222#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224#define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 226
22bedad3
JP
227#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 229#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 231
d94d9fee 232struct hh_cache {
f6b72b62 233 u16 hh_len;
5c25f686 234 u16 __pad;
3644f0ce 235 seqlock_t hh_lock;
1da177e4
LT
236
237 /* cached hardware header; allow for machine alignment needs. */
238#define HH_DATA_MOD 16
239#define HH_DATA_OFF(__len) \
5ba0eac6 240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
241#define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
244};
245
246/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * Alternative is:
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 *
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
253 */
254#define LL_RESERVED_SPACE(dev) \
f5184d26 255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 256#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 258
3b04ddde
SH
259struct header_ops {
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
95c96174 262 const void *saddr, unsigned int len);
3b04ddde 263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268};
269
1da177e4
LT
270/* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
272 * code.
273 */
274
d94d9fee 275enum netdev_state_t {
1da177e4
LT
276 __LINK_STATE_START,
277 __LINK_STATE_PRESENT,
1da177e4 278 __LINK_STATE_NOCARRIER,
b00055aa
SR
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
1da177e4
LT
281};
282
283
284/*
285 * This structure holds at boot time configured netdevice settings. They
fe2918b0 286 * are then used in the device probing.
1da177e4
LT
287 */
288struct netdev_boot_setup {
289 char name[IFNAMSIZ];
290 struct ifmap map;
291};
292#define NETDEV_BOOT_SETUP_MAX 8
293
f629d208 294int __init netdev_boot_setup(char *str);
1da177e4 295
bea3348e
SH
296/*
297 * Structure for NAPI scheduling similar to tasklet but with weighting
298 */
299struct napi_struct {
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
305 */
306 struct list_head poll_list;
307
308 unsigned long state;
309 int weight;
404f7c9e 310 unsigned int gro_count;
bea3348e
SH
311 int (*poll)(struct napi_struct *, int);
312#ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
314 int poll_owner;
bea3348e 315#endif
5d38a079 316 struct net_device *dev;
d565b0a1 317 struct sk_buff *gro_list;
5d38a079 318 struct sk_buff *skb;
3b47d303 319 struct hrtimer timer;
404f7c9e 320 struct list_head dev_list;
af12fa6e
ET
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
bea3348e
SH
323};
324
d94d9fee 325enum {
bea3348e 326 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 327 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
af12fa6e 329 NAPI_STATE_HASHED, /* In NAPI hash */
bea3348e
SH
330};
331
5b252f0c 332enum gro_result {
d1c76af9
HX
333 GRO_MERGED,
334 GRO_MERGED_FREE,
335 GRO_HELD,
336 GRO_NORMAL,
337 GRO_DROP,
338};
5b252f0c 339typedef enum gro_result gro_result_t;
d1c76af9 340
8a4eb573
JP
341/*
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * further.
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 *
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
352 *
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
355 *
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
359 *
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
361 * do with the skb.
362 *
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
366 *
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
370 *
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 373 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
374 *
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
377 *
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
380 */
381
382enum rx_handler_result {
383 RX_HANDLER_CONSUMED,
384 RX_HANDLER_ANOTHER,
385 RX_HANDLER_EXACT,
386 RX_HANDLER_PASS,
387};
388typedef enum rx_handler_result rx_handler_result_t;
389typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 390
f629d208 391void __napi_schedule(struct napi_struct *n);
bc9ad166 392void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 393
4d29515f 394static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
395{
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
397}
398
bea3348e
SH
399/**
400 * napi_schedule_prep - check if napi can be scheduled
401 * @n: napi context
402 *
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
a0a46196
DM
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
bea3348e 407 */
4d29515f 408static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 409{
a0a46196
DM
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
412}
413
414/**
415 * napi_schedule - schedule NAPI poll
416 * @n: napi context
417 *
418 * Schedule NAPI poll routine to be called if it is not already
419 * running.
420 */
421static inline void napi_schedule(struct napi_struct *n)
422{
423 if (napi_schedule_prep(n))
424 __napi_schedule(n);
425}
426
bc9ad166
ED
427/**
428 * napi_schedule_irqoff - schedule NAPI poll
429 * @n: napi context
430 *
431 * Variant of napi_schedule(), assuming hard irqs are masked.
432 */
433static inline void napi_schedule_irqoff(struct napi_struct *n)
434{
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
437}
438
bfe13f54 439/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 440static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
441{
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
4d29515f 444 return true;
bfe13f54 445 }
4d29515f 446 return false;
bfe13f54
RD
447}
448
3b47d303
ED
449void __napi_complete(struct napi_struct *n);
450void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
451/**
452 * napi_complete - NAPI processing complete
453 * @n: napi context
454 *
455 * Mark NAPI processing as complete.
3b47d303 456 * Consider using napi_complete_done() instead.
bea3348e 457 */
3b47d303
ED
458static inline void napi_complete(struct napi_struct *n)
459{
460 return napi_complete_done(n, 0);
461}
bea3348e 462
af12fa6e
ET
463/**
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
466 *
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
469 */
f629d208 470struct napi_struct *napi_by_id(unsigned int napi_id);
af12fa6e
ET
471
472/**
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
475 *
476 * generate a new napi_id and store a @napi under it in napi_hash
477 */
f629d208 478void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
479
480/**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
483 *
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
486 */
f629d208 487void napi_hash_del(struct napi_struct *napi);
af12fa6e 488
bea3348e
SH
489/**
490 * napi_disable - prevent NAPI from scheduling
491 * @n: napi context
492 *
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
495 */
3b47d303 496void napi_disable(struct napi_struct *n);
bea3348e
SH
497
498/**
499 * napi_enable - enable NAPI scheduling
500 * @n: napi context
501 *
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
504 */
505static inline void napi_enable(struct napi_struct *n)
506{
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 508 smp_mb__before_atomic();
bea3348e 509 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 510 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
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
d754f98b
SF
770static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
771 struct netdev_phys_item_id *b)
772{
773 return a->id_len == b->id_len &&
774 memcmp(a->id, b->id, a->id_len) == 0;
775}
776
99932d4f
DB
777typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
778 struct sk_buff *skb);
779
d314774c
SH
780/*
781 * This structure defines the management hooks for network devices.
00829823
SH
782 * The following hooks can be defined; unless noted otherwise, they are
783 * optional and can be filled with a null pointer.
d314774c
SH
784 *
785 * int (*ndo_init)(struct net_device *dev);
786 * This function is called once when network device is registered.
787 * The network device can use this to any late stage initializaton
788 * or semantic validattion. It can fail with an error code which will
789 * be propogated back to register_netdev
790 *
791 * void (*ndo_uninit)(struct net_device *dev);
792 * This function is called when device is unregistered or when registration
793 * fails. It is not called if init fails.
794 *
795 * int (*ndo_open)(struct net_device *dev);
796 * This function is called when network device transistions to the up
797 * state.
798 *
799 * int (*ndo_stop)(struct net_device *dev);
800 * This function is called when network device transistions to the down
801 * state.
802 *
dc1f8bf6
SH
803 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
804 * struct net_device *dev);
00829823 805 * Called when a packet needs to be transmitted.
e79d8429
RR
806 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
807 * the queue before that can happen; it's for obsolete devices and weird
808 * corner cases, but the stack really does a non-trivial amount
809 * of useless work if you return NETDEV_TX_BUSY.
dc1f8bf6 810 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
811 * Required can not be NULL.
812 *
f663dd9a 813 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 814 * void *accel_priv, select_queue_fallback_t fallback);
00829823
SH
815 * Called to decide which queue to when device supports multiple
816 * transmit queues.
817 *
d314774c
SH
818 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
819 * This function is called to allow device receiver to make
820 * changes to configuration when multicast or promiscious is enabled.
821 *
822 * void (*ndo_set_rx_mode)(struct net_device *dev);
823 * This function is called device changes address list filtering.
01789349
JP
824 * If driver handles unicast address filtering, it should set
825 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
826 *
827 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
828 * This function is called when the Media Access Control address
37b607c5 829 * needs to be changed. If this interface is not defined, the
d314774c
SH
830 * mac address can not be changed.
831 *
832 * int (*ndo_validate_addr)(struct net_device *dev);
833 * Test if Media Access Control address is valid for the device.
834 *
835 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
836 * Called when a user request an ioctl which can't be handled by
837 * the generic interface code. If not defined ioctl's return
838 * not supported error code.
839 *
840 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
841 * Used to set network devices bus interface parameters. This interface
842 * is retained for legacy reason, new devices should use the bus
843 * interface (PCI) for low level management.
844 *
845 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
846 * Called when a user wants to change the Maximum Transfer Unit
847 * of a device. If not defined, any request to change MTU will
848 * will return an error.
849 *
00829823 850 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
851 * Callback uses when the transmitter has not made any progress
852 * for dev->watchdog ticks.
853 *
3cfde79c 854 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 855 * struct rtnl_link_stats64 *storage);
d308e38f 856 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 857 * Called when a user wants to get the network device usage
be1f3c2c 858 * statistics. Drivers must do one of the following:
3cfde79c
BH
859 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
860 * rtnl_link_stats64 structure passed by the caller.
82695d9b 861 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
862 * (which should normally be dev->stats) and return a pointer to
863 * it. The structure may be changed asynchronously only if each
864 * field is written atomically.
865 * 3. Update dev->stats asynchronously and atomically, and define
866 * neither operation.
d314774c 867 *
5d632cb7 868 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
869 * If device support VLAN filtering this function is called when a
870 * VLAN id is registered.
d314774c 871 *
5d632cb7 872 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
873 * If device support VLAN filtering this function is called when a
874 * VLAN id is unregistered.
d314774c
SH
875 *
876 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
877 *
878 * SR-IOV management functions.
879 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
880 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
881 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
882 * int max_tx_rate);
5f8444a3 883 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
884 * int (*ndo_get_vf_config)(struct net_device *dev,
885 * int vf, struct ifla_vf_info *ivf);
1d8faf48 886 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
887 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
888 * struct nlattr *port[]);
01a3d796
VZ
889 *
890 * Enable or disable the VF ability to query its RSS Redirection Table and
891 * Hash Key. This is needed since on some devices VF share this information
892 * with PF and querying it may adduce a theoretical security risk.
893 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 894 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
895 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
896 * Called to setup 'tc' number of traffic classes in the net device. This
897 * is always called from the stack with the rtnl lock held and netif tx
898 * queues stopped. This allows the netdevice to perform queue management
899 * safely.
c445477d 900 *
e9bce845
YZ
901 * Fiber Channel over Ethernet (FCoE) offload functions.
902 * int (*ndo_fcoe_enable)(struct net_device *dev);
903 * Called when the FCoE protocol stack wants to start using LLD for FCoE
904 * so the underlying device can perform whatever needed configuration or
905 * initialization to support acceleration of FCoE traffic.
906 *
907 * int (*ndo_fcoe_disable)(struct net_device *dev);
908 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
909 * so the underlying device can perform whatever needed clean-ups to
910 * stop supporting acceleration of FCoE traffic.
911 *
912 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
913 * struct scatterlist *sgl, unsigned int sgc);
914 * Called when the FCoE Initiator wants to initialize an I/O that
915 * is a possible candidate for Direct Data Placement (DDP). The LLD can
916 * perform necessary setup and returns 1 to indicate the device is set up
917 * successfully to perform DDP on this I/O, otherwise this returns 0.
918 *
919 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
920 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
921 * indicated by the FC exchange id 'xid', so the underlying device can
922 * clean up and reuse resources for later DDP requests.
923 *
924 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
925 * struct scatterlist *sgl, unsigned int sgc);
926 * Called when the FCoE Target wants to initialize an I/O that
927 * is a possible candidate for Direct Data Placement (DDP). The LLD can
928 * perform necessary setup and returns 1 to indicate the device is set up
929 * successfully to perform DDP on this I/O, otherwise this returns 0.
930 *
68bad94e
NP
931 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
932 * struct netdev_fcoe_hbainfo *hbainfo);
933 * Called when the FCoE Protocol stack wants information on the underlying
934 * device. This information is utilized by the FCoE protocol stack to
935 * register attributes with Fiber Channel management service as per the
936 * FC-GS Fabric Device Management Information(FDMI) specification.
937 *
e9bce845
YZ
938 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
939 * Called when the underlying device wants to override default World Wide
940 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
941 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
942 * protocol stack to use.
943 *
c445477d
BH
944 * RFS acceleration.
945 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
946 * u16 rxq_index, u32 flow_id);
947 * Set hardware filter for RFS. rxq_index is the target queue index;
948 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
949 * Return the filter ID on success, or a negative error code.
fbaec0ea 950 *
8b98a70c 951 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
952 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
953 * Called to make another netdev an underling.
954 *
955 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
956 * Called to release previously enslaved netdev.
5455c699
MM
957 *
958 * Feature/offload setting functions.
c8f44aff
MM
959 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
960 * netdev_features_t features);
5455c699
MM
961 * Adjusts the requested feature flags according to device-specific
962 * constraints, and returns the resulting flags. Must not modify
963 * the device state.
964 *
c8f44aff 965 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
966 * Called to update device configuration to new features. Passed
967 * feature set might be less than what was returned by ndo_fix_features()).
968 * Must return >0 or -errno if it changed dev->features itself.
969 *
edc7d573 970 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
971 * struct net_device *dev,
f6f6424b 972 * const unsigned char *addr, u16 vid, u16 flags)
77162022 973 * Adds an FDB entry to dev for addr.
1690be63
VY
974 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
975 * struct net_device *dev,
f6f6424b 976 * const unsigned char *addr, u16 vid)
77162022
JF
977 * Deletes the FDB entry from dev coresponding to addr.
978 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
979 * struct net_device *dev, struct net_device *filter_dev,
980 * int idx)
77162022
JF
981 * Used to add FDB entries to dump requests. Implementers should add
982 * entries to skb and update idx with the number of entries.
e5a55a89 983 *
ad41faa8
ND
984 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
985 * u16 flags)
e5a55a89 986 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
987 * struct net_device *dev, u32 filter_mask,
988 * int nlflags)
ad41faa8
ND
989 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
990 * u16 flags);
4bf84c35
JP
991 *
992 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
993 * Called to change device carrier. Soft-devices (like dummy, team, etc)
994 * which do not represent real hardware may define this to allow their
995 * userspace components to manage their virtual carrier state. Devices
996 * that determine carrier state from physical hardware properties (eg
997 * network cables) or protocol-dependent mechanisms (eg
998 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
999 *
1000 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1001 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1002 * Called to get ID of physical port of this device. If driver does
1003 * not implement this, it is assumed that the hw is not able to have
1004 * multiple net devices on single physical port.
53cf5275
JG
1005 *
1006 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 1007 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1008 * Called by vxlan to notiy a driver about the UDP port and socket
1009 * address family that vxlan is listnening to. It is called only when
1010 * a new port starts listening. The operation is protected by the
1011 * vxlan_net->sock_lock.
1012 *
1013 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 1014 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1015 * Called by vxlan to notify the driver about a UDP port and socket
1016 * address family that vxlan is not listening to anymore. The operation
1017 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
1018 *
1019 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1020 * struct net_device *dev)
1021 * Called by upper layer devices to accelerate switching or other
1022 * station functionality into hardware. 'pdev is the lowerdev
1023 * to use for the offload and 'dev' is the net device that will
1024 * back the offload. Returns a pointer to the private structure
1025 * the upper layer will maintain.
1026 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1027 * Called by upper layer device to delete the station created
1028 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1029 * the station and priv is the structure returned by the add
1030 * operation.
1031 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1032 * struct net_device *dev,
1033 * void *priv);
1034 * Callback to use for xmit over the accelerated station. This
1035 * is used in place of ndo_start_xmit on accelerated net
1036 * devices.
5f35227e
JG
1037 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1038 * struct net_device *dev
1039 * netdev_features_t features);
04ffcb25 1040 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1041 * performing offload operations on a given packet. This is to give
1042 * the device an opportunity to implement any restrictions that cannot
1043 * be otherwise expressed by feature flags. The check is called with
1044 * the set of features that the stack has calculated and it returns
1045 * those the driver believes to be appropriate.
822b3b2e
JF
1046 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1047 * int queue_index, u32 maxrate);
1048 * Called when a user wants to set a max-rate limitation of specific
1049 * TX queue.
a54acb3a
ND
1050 * int (*ndo_get_iflink)(const struct net_device *dev);
1051 * Called to get the iflink value of this device.
d746d707
AK
1052 * void (*ndo_change_proto_down)(struct net_device *dev,
1053 * bool proto_down);
1054 * This function is used to pass protocol port error state information
1055 * to the switch driver. The switch driver can react to the proto_down
1056 * by doing a phys down on the associated switch port.
fc4099f1
PS
1057 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1058 * This function is used to get egress tunnel information for given skb.
1059 * This is useful for retrieving outer tunnel header parameters while
1060 * sampling packet.
d746d707 1061 *
d314774c
SH
1062 */
1063struct net_device_ops {
1064 int (*ndo_init)(struct net_device *dev);
1065 void (*ndo_uninit)(struct net_device *dev);
1066 int (*ndo_open)(struct net_device *dev);
1067 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 1068 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
1069 struct net_device *dev);
1070 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1071 struct sk_buff *skb,
99932d4f
DB
1072 void *accel_priv,
1073 select_queue_fallback_t fallback);
d314774c
SH
1074 void (*ndo_change_rx_flags)(struct net_device *dev,
1075 int flags);
d314774c 1076 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1077 int (*ndo_set_mac_address)(struct net_device *dev,
1078 void *addr);
d314774c 1079 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1080 int (*ndo_do_ioctl)(struct net_device *dev,
1081 struct ifreq *ifr, int cmd);
d314774c
SH
1082 int (*ndo_set_config)(struct net_device *dev,
1083 struct ifmap *map);
00829823
SH
1084 int (*ndo_change_mtu)(struct net_device *dev,
1085 int new_mtu);
1086 int (*ndo_neigh_setup)(struct net_device *dev,
1087 struct neigh_parms *);
d314774c
SH
1088 void (*ndo_tx_timeout) (struct net_device *dev);
1089
28172739
ED
1090 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1091 struct rtnl_link_stats64 *storage);
d314774c
SH
1092 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1093
8e586137 1094 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1095 __be16 proto, u16 vid);
8e586137 1096 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1097 __be16 proto, u16 vid);
d314774c 1098#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1099 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1100 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1101 struct netpoll_info *info);
0e34e931 1102 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1103#endif
e0d1095a 1104#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1105 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1106#endif
95c26df8
WM
1107 int (*ndo_set_vf_mac)(struct net_device *dev,
1108 int queue, u8 *mac);
1109 int (*ndo_set_vf_vlan)(struct net_device *dev,
1110 int queue, u16 vlan, u8 qos);
ed616689
SC
1111 int (*ndo_set_vf_rate)(struct net_device *dev,
1112 int vf, int min_tx_rate,
1113 int max_tx_rate);
5f8444a3
GR
1114 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1115 int vf, bool setting);
95c26df8
WM
1116 int (*ndo_get_vf_config)(struct net_device *dev,
1117 int vf,
1118 struct ifla_vf_info *ivf);
1d8faf48
RE
1119 int (*ndo_set_vf_link_state)(struct net_device *dev,
1120 int vf, int link_state);
3b766cd8
EBE
1121 int (*ndo_get_vf_stats)(struct net_device *dev,
1122 int vf,
1123 struct ifla_vf_stats
1124 *vf_stats);
57b61080
SF
1125 int (*ndo_set_vf_port)(struct net_device *dev,
1126 int vf,
1127 struct nlattr *port[]);
1128 int (*ndo_get_vf_port)(struct net_device *dev,
1129 int vf, struct sk_buff *skb);
01a3d796
VZ
1130 int (*ndo_set_vf_rss_query_en)(
1131 struct net_device *dev,
1132 int vf, bool setting);
4f57c087 1133 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
d11ead75 1134#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1135 int (*ndo_fcoe_enable)(struct net_device *dev);
1136 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1137 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1138 u16 xid,
1139 struct scatterlist *sgl,
1140 unsigned int sgc);
1141 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1142 u16 xid);
6247e086
YZ
1143 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1144 u16 xid,
1145 struct scatterlist *sgl,
1146 unsigned int sgc);
68bad94e
NP
1147 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1148 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1149#endif
1150
d11ead75 1151#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1152#define NETDEV_FCOE_WWNN 0
1153#define NETDEV_FCOE_WWPN 1
1154 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1155 u64 *wwn, int type);
4d288d57 1156#endif
3c9c36bc 1157
c445477d
BH
1158#ifdef CONFIG_RFS_ACCEL
1159 int (*ndo_rx_flow_steer)(struct net_device *dev,
1160 const struct sk_buff *skb,
1161 u16 rxq_index,
1162 u32 flow_id);
1163#endif
fbaec0ea
JP
1164 int (*ndo_add_slave)(struct net_device *dev,
1165 struct net_device *slave_dev);
1166 int (*ndo_del_slave)(struct net_device *dev,
1167 struct net_device *slave_dev);
c8f44aff
MM
1168 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1169 netdev_features_t features);
5455c699 1170 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1171 netdev_features_t features);
da6a8fa0 1172 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1173 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1174
1175 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1176 struct nlattr *tb[],
77162022 1177 struct net_device *dev,
6b6e2725 1178 const unsigned char *addr,
f6f6424b 1179 u16 vid,
77162022
JF
1180 u16 flags);
1181 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1182 struct nlattr *tb[],
77162022 1183 struct net_device *dev,
f6f6424b
JP
1184 const unsigned char *addr,
1185 u16 vid);
77162022
JF
1186 int (*ndo_fdb_dump)(struct sk_buff *skb,
1187 struct netlink_callback *cb,
1188 struct net_device *dev,
5d5eacb3 1189 struct net_device *filter_dev,
77162022 1190 int idx);
e5a55a89
JF
1191
1192 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1193 struct nlmsghdr *nlh,
1194 u16 flags);
e5a55a89
JF
1195 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1196 u32 pid, u32 seq,
6cbdceeb 1197 struct net_device *dev,
46c264da
ND
1198 u32 filter_mask,
1199 int nlflags);
407af329 1200 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1201 struct nlmsghdr *nlh,
1202 u16 flags);
4bf84c35
JP
1203 int (*ndo_change_carrier)(struct net_device *dev,
1204 bool new_carrier);
66b52b0d 1205 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1206 struct netdev_phys_item_id *ppid);
db24a904
DA
1207 int (*ndo_get_phys_port_name)(struct net_device *dev,
1208 char *name, size_t len);
53cf5275
JG
1209 void (*ndo_add_vxlan_port)(struct net_device *dev,
1210 sa_family_t sa_family,
35e42379 1211 __be16 port);
53cf5275
JG
1212 void (*ndo_del_vxlan_port)(struct net_device *dev,
1213 sa_family_t sa_family,
35e42379 1214 __be16 port);
a6cc0cfa
JF
1215
1216 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1217 struct net_device *dev);
1218 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1219 void *priv);
1220
1221 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1222 struct net_device *dev,
1223 void *priv);
25175ba5 1224 int (*ndo_get_lock_subclass)(struct net_device *dev);
5f35227e
JG
1225 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1226 struct net_device *dev,
1227 netdev_features_t features);
822b3b2e
JF
1228 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1229 int queue_index,
1230 u32 maxrate);
a54acb3a 1231 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1232 int (*ndo_change_proto_down)(struct net_device *dev,
1233 bool proto_down);
fc4099f1
PS
1234 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1235 struct sk_buff *skb);
d314774c
SH
1236};
1237
7aa98047
LR
1238/**
1239 * enum net_device_priv_flags - &struct net_device priv_flags
1240 *
1241 * These are the &struct net_device, they are only set internally
1242 * by drivers and used in the kernel. These flags are invisible to
1243 * userspace, this means that the order of these flags can change
1244 * during any kernel release.
1245 *
1246 * You should have a pretty good reason to be extending these flags.
1247 *
1248 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1249 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1250 * @IFF_BONDING: bonding master or slave
7aa98047 1251 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1252 * @IFF_WAN_HDLC: WAN HDLC device
1253 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1254 * release skb->dst
1255 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1256 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1257 * @IFF_MACVLAN_PORT: device used as macvlan port
1258 * @IFF_BRIDGE_PORT: device used as bridge port
1259 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1260 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1261 * @IFF_UNICAST_FLT: Supports unicast filtering
1262 * @IFF_TEAM_PORT: device used as team port
1263 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1264 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1265 * change when it's running
1266 * @IFF_MACVLAN: Macvlan device
808d28c4 1267 * @IFF_VRF_MASTER: device is a VRF master
fa8187c9 1268 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1269 * @IFF_OPENVSWITCH: device is a Open vSwitch master
7aa98047
LR
1270 */
1271enum netdev_priv_flags {
1272 IFF_802_1Q_VLAN = 1<<0,
1273 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1274 IFF_BONDING = 1<<2,
1275 IFF_ISATAP = 1<<3,
1276 IFF_WAN_HDLC = 1<<4,
1277 IFF_XMIT_DST_RELEASE = 1<<5,
1278 IFF_DONT_BRIDGE = 1<<6,
1279 IFF_DISABLE_NETPOLL = 1<<7,
1280 IFF_MACVLAN_PORT = 1<<8,
1281 IFF_BRIDGE_PORT = 1<<9,
1282 IFF_OVS_DATAPATH = 1<<10,
1283 IFF_TX_SKB_SHARING = 1<<11,
1284 IFF_UNICAST_FLT = 1<<12,
1285 IFF_TEAM_PORT = 1<<13,
1286 IFF_SUPP_NOFCS = 1<<14,
1287 IFF_LIVE_ADDR_CHANGE = 1<<15,
1288 IFF_MACVLAN = 1<<16,
1289 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1290 IFF_IPVLAN_MASTER = 1<<18,
1291 IFF_IPVLAN_SLAVE = 1<<19,
1292 IFF_VRF_MASTER = 1<<20,
1293 IFF_NO_QUEUE = 1<<21,
1294 IFF_OPENVSWITCH = 1<<22,
7aa98047
LR
1295};
1296
1297#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1298#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1299#define IFF_BONDING IFF_BONDING
7aa98047 1300#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1301#define IFF_WAN_HDLC IFF_WAN_HDLC
1302#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1303#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1304#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1305#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1306#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1307#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1308#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1309#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1310#define IFF_TEAM_PORT IFF_TEAM_PORT
1311#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1312#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1313#define IFF_MACVLAN IFF_MACVLAN
02875878 1314#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1315#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1316#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
4e3c8992 1317#define IFF_VRF_MASTER IFF_VRF_MASTER
fa8187c9 1318#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1319#define IFF_OPENVSWITCH IFF_OPENVSWITCH
7aa98047 1320
536721b1
KK
1321/**
1322 * struct net_device - The DEVICE structure.
1323 * Actually, this whole structure is a big mistake. It mixes I/O
1324 * data with strictly "high-level" data, and it has to know about
1325 * almost every data structure used in the INET module.
1326 *
1327 * @name: This is the first field of the "visible" part of this structure
1328 * (i.e. as seen by users in the "Space.c" file). It is the name
1329 * of the interface.
1330 *
1331 * @name_hlist: Device name hash chain, please keep it close to name[]
1332 * @ifalias: SNMP alias
1333 * @mem_end: Shared memory end
1334 * @mem_start: Shared memory start
1335 * @base_addr: Device I/O address
1336 * @irq: Device IRQ number
1337 *
14ffbbb8
TG
1338 * @carrier_changes: Stats to monitor carrier on<->off transitions
1339 *
536721b1
KK
1340 * @state: Generic network queuing layer state, see netdev_state_t
1341 * @dev_list: The global list of network devices
1342 * @napi_list: List entry, that is used for polling napi devices
1343 * @unreg_list: List entry, that is used, when we are unregistering the
1344 * device, see the function unregister_netdev
1345 * @close_list: List entry, that is used, when we are closing the device
1346 *
1347 * @adj_list: Directly linked devices, like slaves for bonding
1348 * @all_adj_list: All linked devices, *including* neighbours
1349 * @features: Currently active device features
1350 * @hw_features: User-changeable features
1351 *
1352 * @wanted_features: User-requested features
1353 * @vlan_features: Mask of features inheritable by VLAN devices
1354 *
1355 * @hw_enc_features: Mask of features inherited by encapsulating devices
1356 * This field indicates what encapsulation
1357 * offloads the hardware is capable of doing,
1358 * and drivers will need to set them appropriately.
1359 *
1360 * @mpls_features: Mask of features inheritable by MPLS
1361 *
1362 * @ifindex: interface index
388069d3 1363 * @group: The group, that the device belongs to
536721b1
KK
1364 *
1365 * @stats: Statistics struct, which was left as a legacy, use
1366 * rtnl_link_stats64 instead
1367 *
1368 * @rx_dropped: Dropped packets by core network,
1369 * do not use this in drivers
1370 * @tx_dropped: Dropped packets by core network,
1371 * do not use this in drivers
1372 *
536721b1
KK
1373 * @wireless_handlers: List of functions to handle Wireless Extensions,
1374 * instead of ioctl,
1375 * see <net/iw_handler.h> for details.
1376 * @wireless_data: Instance data managed by the core of wireless extensions
1377 *
1378 * @netdev_ops: Includes several pointers to callbacks,
1379 * if one wants to override the ndo_*() functions
1380 * @ethtool_ops: Management operations
d476059e 1381 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1382 * of Layer 2 headers.
1383 *
1384 * @flags: Interface flags (a la BSD)
1385 * @priv_flags: Like 'flags' but invisible to userspace,
1386 * see if.h for the definitions
1387 * @gflags: Global flags ( kept as legacy )
1388 * @padded: How much padding added by alloc_netdev()
1389 * @operstate: RFC2863 operstate
1390 * @link_mode: Mapping policy to operstate
1391 * @if_port: Selectable AUI, TP, ...
1392 * @dma: DMA channel
1393 * @mtu: Interface MTU value
1394 * @type: Interface hardware type
1395 * @hard_header_len: Hardware header length
1396 *
1397 * @needed_headroom: Extra headroom the hardware may need, but not in all
1398 * cases can this be guaranteed
1399 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1400 * cases can this be guaranteed. Some cases also use
1401 * LL_MAX_HEADER instead to allocate the skb
1402 *
1403 * interface address info:
1404 *
1405 * @perm_addr: Permanent hw address
1406 * @addr_assign_type: Hw address assignment type
1407 * @addr_len: Hardware address length
1408 * @neigh_priv_len; Used in neigh_alloc(),
1409 * initialized only in atm/clip.c
1410 * @dev_id: Used to differentiate devices that share
1411 * the same link layer address
1412 * @dev_port: Used to differentiate devices that share
1413 * the same function
1414 * @addr_list_lock: XXX: need comments on this one
536721b1
KK
1415 * @uc_promisc: Counter, that indicates, that promiscuous mode
1416 * has been enabled due to the need to listen to
1417 * additional unicast addresses in a device that
1418 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1419 * @uc: unicast mac addresses
1420 * @mc: multicast mac addresses
1421 * @dev_addrs: list of device hw addresses
1422 * @queues_kset: Group of all Kobjects in the Tx and RX queues
536721b1
KK
1423 * @promiscuity: Number of times, the NIC is told to work in
1424 * Promiscuous mode, if it becomes 0 the NIC will
1425 * exit from working in Promiscuous mode
1426 * @allmulti: Counter, enables or disables allmulticast mode
1427 *
1428 * @vlan_info: VLAN info
1429 * @dsa_ptr: dsa specific data
1430 * @tipc_ptr: TIPC specific data
1431 * @atalk_ptr: AppleTalk link
1432 * @ip_ptr: IPv4 specific data
1433 * @dn_ptr: DECnet specific data
1434 * @ip6_ptr: IPv6 specific data
1435 * @ax25_ptr: AX.25 specific data
4e3c8992 1436 * @vrf_ptr: VRF specific data
536721b1
KK
1437 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1438 *
1439 * @last_rx: Time of last Rx
1440 * @dev_addr: Hw address (before bcast,
1441 * because most packets are unicast)
1442 *
1443 * @_rx: Array of RX queues
1444 * @num_rx_queues: Number of RX queues
1445 * allocated at register_netdev() time
1446 * @real_num_rx_queues: Number of RX queues currently active in device
1447 *
1448 * @rx_handler: handler for received packets
1449 * @rx_handler_data: XXX: need comments on this one
1450 * @ingress_queue: XXX: need comments on this one
1451 * @broadcast: hw bcast address
1452 *
14ffbbb8
TG
1453 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1454 * indexed by RX queue number. Assigned by driver.
1455 * This must only be set if the ndo_rx_flow_steer
1456 * operation is defined
1457 * @index_hlist: Device index hash chain
1458 *
536721b1
KK
1459 * @_tx: Array of TX queues
1460 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1461 * @real_num_tx_queues: Number of TX queues currently active in device
1462 * @qdisc: Root qdisc from userspace point of view
1463 * @tx_queue_len: Max frames per queue allowed
1464 * @tx_global_lock: XXX: need comments on this one
1465 *
1466 * @xps_maps: XXX: need comments on this one
1467 *
0c4f691f
SF
1468 * @offload_fwd_mark: Offload device fwding mark
1469 *
536721b1
KK
1470 * @trans_start: Time (in jiffies) of last Tx
1471 * @watchdog_timeo: Represents the timeout that is used by
1472 * the watchdog ( see dev_watchdog() )
1473 * @watchdog_timer: List of timers
1474 *
1475 * @pcpu_refcnt: Number of references to this device
1476 * @todo_list: Delayed register/unregister
536721b1
KK
1477 * @link_watch_list: XXX: need comments on this one
1478 *
1479 * @reg_state: Register/unregister state machine
1480 * @dismantle: Device is going to be freed
1481 * @rtnl_link_state: This enum represents the phases of creating
1482 * a new link
1483 *
1484 * @destructor: Called from unregister,
1485 * can be used to call free_netdev
1486 * @npinfo: XXX: need comments on this one
1487 * @nd_net: Network namespace this network device is inside
1488 *
1489 * @ml_priv: Mid-layer private
1490 * @lstats: Loopback statistics
1491 * @tstats: Tunnel statistics
1492 * @dstats: Dummy statistics
1493 * @vstats: Virtual ethernet statistics
1494 *
1495 * @garp_port: GARP
1496 * @mrp_port: MRP
1497 *
1498 * @dev: Class/net/name entry
1499 * @sysfs_groups: Space for optional device, statistics and wireless
1500 * sysfs groups
1501 *
1502 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1503 * @rtnl_link_ops: Rtnl_link_ops
1504 *
1505 * @gso_max_size: Maximum size of generic segmentation offload
1506 * @gso_max_segs: Maximum number of segments that can be passed to the
1507 * NIC for GSO
fcbeb976
ED
1508 * @gso_min_segs: Minimum number of segments that can be passed to the
1509 * NIC for GSO
536721b1
KK
1510 *
1511 * @dcbnl_ops: Data Center Bridging netlink ops
1512 * @num_tc: Number of traffic classes in the net device
1513 * @tc_to_txq: XXX: need comments on this one
1514 * @prio_tc_map XXX: need comments on this one
1515 *
1516 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1517 *
1518 * @priomap: XXX: need comments on this one
1519 * @phydev: Physical device may attach itself
1520 * for hardware timestamping
1521 *
1522 * @qdisc_tx_busylock: XXX: need comments on this one
1523 *
d746d707
AK
1524 * @proto_down: protocol port state information can be sent to the
1525 * switch driver and used to set the phys state of the
1526 * switch port.
1527 *
1da177e4
LT
1528 * FIXME: cleanup struct net_device such that network protocol info
1529 * moves out.
1530 */
1531
d94d9fee 1532struct net_device {
1da177e4 1533 char name[IFNAMSIZ];
9356b8fc 1534 struct hlist_node name_hlist;
0b815a1a 1535 char *ifalias;
1da177e4
LT
1536 /*
1537 * I/O specific fields
1538 * FIXME: Merge these and struct ifmap into one
1539 */
536721b1
KK
1540 unsigned long mem_end;
1541 unsigned long mem_start;
1542 unsigned long base_addr;
1543 int irq;
1da177e4 1544
14ffbbb8
TG
1545 atomic_t carrier_changes;
1546
1da177e4 1547 /*
536721b1
KK
1548 * Some hardware also needs these fields (state,dev_list,
1549 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1550 * part of the usual set specified in Space.c.
1551 */
1552
1da177e4
LT
1553 unsigned long state;
1554
7562f876 1555 struct list_head dev_list;
bea3348e 1556 struct list_head napi_list;
44a0873d 1557 struct list_head unreg_list;
5cde2829 1558 struct list_head close_list;
7866a621
SN
1559 struct list_head ptype_all;
1560 struct list_head ptype_specific;
2f268f12 1561
2f268f12
VF
1562 struct {
1563 struct list_head upper;
1564 struct list_head lower;
1565 } adj_list;
1566
2f268f12
VF
1567 struct {
1568 struct list_head upper;
1569 struct list_head lower;
1570 } all_adj_list;
4c3d5e7b 1571
c8f44aff 1572 netdev_features_t features;
c8f44aff 1573 netdev_features_t hw_features;
c8f44aff 1574 netdev_features_t wanted_features;
c8f44aff 1575 netdev_features_t vlan_features;
6a674e9c 1576 netdev_features_t hw_enc_features;
0d89d203 1577 netdev_features_t mpls_features;
04ed3e74 1578
1da177e4 1579 int ifindex;
7a66bbc9 1580 int group;
1da177e4 1581
c45d286e 1582 struct net_device_stats stats;
015f0688 1583
015f0688
ED
1584 atomic_long_t rx_dropped;
1585 atomic_long_t tx_dropped;
1da177e4 1586
b86e0280 1587#ifdef CONFIG_WIRELESS_EXT
1da177e4 1588 const struct iw_handler_def * wireless_handlers;
1da177e4 1589 struct iw_public_data * wireless_data;
b86e0280 1590#endif
d314774c 1591 const struct net_device_ops *netdev_ops;
76fd8593 1592 const struct ethtool_ops *ethtool_ops;
4170604f 1593#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1594 const struct switchdev_ops *switchdev_ops;
4170604f 1595#endif
1da177e4 1596
3b04ddde
SH
1597 const struct header_ops *header_ops;
1598
536721b1
KK
1599 unsigned int flags;
1600 unsigned int priv_flags;
1601
1da177e4 1602 unsigned short gflags;
536721b1 1603 unsigned short padded;
1da177e4 1604
536721b1
KK
1605 unsigned char operstate;
1606 unsigned char link_mode;
b00055aa 1607
536721b1
KK
1608 unsigned char if_port;
1609 unsigned char dma;
bdc220da 1610
536721b1
KK
1611 unsigned int mtu;
1612 unsigned short type;
1613 unsigned short hard_header_len;
1da177e4 1614
f5184d26
JB
1615 unsigned short needed_headroom;
1616 unsigned short needed_tailroom;
1617
1da177e4 1618 /* Interface address info. */
536721b1
KK
1619 unsigned char perm_addr[MAX_ADDR_LEN];
1620 unsigned char addr_assign_type;
1621 unsigned char addr_len;
a0a9663d 1622 unsigned short neigh_priv_len;
536721b1
KK
1623 unsigned short dev_id;
1624 unsigned short dev_port;
ccffad25 1625 spinlock_t addr_list_lock;
14ffbbb8
TG
1626 unsigned char name_assign_type;
1627 bool uc_promisc;
536721b1
KK
1628 struct netdev_hw_addr_list uc;
1629 struct netdev_hw_addr_list mc;
1630 struct netdev_hw_addr_list dev_addrs;
1631
4c3d5e7b
ED
1632#ifdef CONFIG_SYSFS
1633 struct kset *queues_kset;
1634#endif
9d45abe1
WC
1635 unsigned int promiscuity;
1636 unsigned int allmulti;
1da177e4 1637
1da177e4
LT
1638
1639 /* Protocol specific pointers */
65ac6a5f 1640
d11ead75 1641#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1642 struct vlan_info __rcu *vlan_info;
65ac6a5f 1643#endif
34a430d7 1644#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1645 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1646#endif
1647#if IS_ENABLED(CONFIG_TIPC)
536721b1 1648 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1649#endif
536721b1
KK
1650 void *atalk_ptr;
1651 struct in_device __rcu *ip_ptr;
1652 struct dn_dev __rcu *dn_ptr;
1653 struct inet6_dev __rcu *ip6_ptr;
1654 void *ax25_ptr;
4e3c8992 1655 struct net_vrf_dev __rcu *vrf_ptr;
536721b1 1656 struct wireless_dev *ieee80211_ptr;
98a18b6f 1657 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1658#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1659 struct mpls_dev __rcu *mpls_ptr;
1660#endif
1da177e4 1661
9356b8fc 1662/*
cd13539b 1663 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1664 */
536721b1 1665 unsigned long last_rx;
4dc89133 1666
9356b8fc 1667 /* Interface address info used in eth_type_trans() */
536721b1 1668 unsigned char *dev_addr;
f001fde5 1669
0a9627f2 1670
a953be53 1671#ifdef CONFIG_SYSFS
0a9627f2
TH
1672 struct netdev_rx_queue *_rx;
1673
0a9627f2 1674 unsigned int num_rx_queues;
62fe0b40 1675 unsigned int real_num_rx_queues;
c445477d 1676
df334545 1677#endif
0a9627f2 1678
3b47d303 1679 unsigned long gro_flush_timeout;
61391cde 1680 rx_handler_func_t __rcu *rx_handler;
1681 void __rcu *rx_handler_data;
e8a0464c 1682
4cda01e8 1683#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1684 struct tcf_proto __rcu *ingress_cl_list;
1685#endif
24824a09 1686 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1687#ifdef CONFIG_NETFILTER_INGRESS
1688 struct list_head nf_hooks_ingress;
1689#endif
d2788d34 1690
536721b1 1691 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1692#ifdef CONFIG_RFS_ACCEL
1693 struct cpu_rmap *rx_cpu_rmap;
1694#endif
1695 struct hlist_node index_hlist;
cd13539b
ED
1696
1697/*
1698 * Cache lines mostly used on transmit path
1699 */
e8a0464c
DM
1700 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1701 unsigned int num_tx_queues;
fd2ea0a7 1702 unsigned int real_num_tx_queues;
af356afa 1703 struct Qdisc *qdisc;
536721b1 1704 unsigned long tx_queue_len;
c3f26a26 1705 spinlock_t tx_global_lock;
14ffbbb8 1706 int watchdog_timeo;
cd13539b 1707
bf264145 1708#ifdef CONFIG_XPS
a4177869 1709 struct xps_dev_maps __rcu *xps_maps;
bf264145 1710#endif
1d24eb48 1711
0c4f691f
SF
1712#ifdef CONFIG_NET_SWITCHDEV
1713 u32 offload_fwd_mark;
1714#endif
1715
9356b8fc 1716 /* These may be needed for future network-power-down code. */
9d21493b
ED
1717
1718 /*
1719 * trans_start here is expensive for high speed devices on SMP,
1720 * please use netdev_queue->trans_start instead.
1721 */
536721b1 1722 unsigned long trans_start;
9356b8fc 1723
9356b8fc
ED
1724 struct timer_list watchdog_timer;
1725
29b4433d 1726 int __percpu *pcpu_refcnt;
1da177e4 1727 struct list_head todo_list;
1da177e4 1728
e014debe 1729 struct list_head link_watch_list;
572a103d 1730
1da177e4 1731 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1732 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1733 NETREG_UNREGISTERING, /* called unregister_netdevice */
1734 NETREG_UNREGISTERED, /* completed unregister todo */
1735 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1736 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1737 } reg_state:8;
1738
536721b1 1739 bool dismantle;
a2835763
PM
1740
1741 enum {
1742 RTNL_LINK_INITIALIZED,
1743 RTNL_LINK_INITIALIZING,
1744 } rtnl_link_state:16;
1da177e4 1745
d314774c 1746 void (*destructor)(struct net_device *dev);
1da177e4 1747
1da177e4 1748#ifdef CONFIG_NETPOLL
5fbee843 1749 struct netpoll_info __rcu *npinfo;
1da177e4 1750#endif
eae792b7 1751
0c5c9fb5 1752 possible_net_t nd_net;
4a1c5371 1753
4951704b 1754 /* mid-layer private */
a7855c78 1755 union {
536721b1
KK
1756 void *ml_priv;
1757 struct pcpu_lstats __percpu *lstats;
8f84985f 1758 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1759 struct pcpu_dstats __percpu *dstats;
1760 struct pcpu_vstats __percpu *vstats;
a7855c78 1761 };
536721b1 1762
3cc77ec7 1763 struct garp_port __rcu *garp_port;
febf018d 1764 struct mrp_port __rcu *mrp_port;
1da177e4 1765
536721b1 1766 struct device dev;
0c509a6c 1767 const struct attribute_group *sysfs_groups[4];
a953be53 1768 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1769
38f7b870 1770 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1771
82cc1a7a
PWJ
1772 /* for setting kernel sock attribute on TCP connection setup */
1773#define GSO_MAX_SIZE 65536
1774 unsigned int gso_max_size;
30b678d8
BH
1775#define GSO_MAX_SEGS 65535
1776 u16 gso_max_segs;
fcbeb976 1777 u16 gso_min_segs;
7a6b6f51 1778#ifdef CONFIG_DCB
32953543 1779 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1780#endif
4f57c087
JF
1781 u8 num_tc;
1782 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1783 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1784
d11ead75 1785#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1786 unsigned int fcoe_ddp_xid;
5bc1421e 1787#endif
86f8515f 1788#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1789 struct netprio_map __rcu *priomap;
4d288d57 1790#endif
c1f19b51 1791 struct phy_device *phydev;
23d3b8bf 1792 struct lock_class_key *qdisc_tx_busylock;
d746d707 1793 bool proto_down;
1da177e4 1794};
43cb76d9 1795#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1796
1797#define NETDEV_ALIGN 32
1da177e4 1798
4f57c087
JF
1799static inline
1800int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1801{
1802 return dev->prio_tc_map[prio & TC_BITMASK];
1803}
1804
1805static inline
1806int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1807{
1808 if (tc >= dev->num_tc)
1809 return -EINVAL;
1810
1811 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1812 return 0;
1813}
1814
1815static inline
1816void netdev_reset_tc(struct net_device *dev)
1817{
1818 dev->num_tc = 0;
1819 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1820 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1821}
1822
1823static inline
1824int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1825{
1826 if (tc >= dev->num_tc)
1827 return -EINVAL;
1828
1829 dev->tc_to_txq[tc].count = count;
1830 dev->tc_to_txq[tc].offset = offset;
1831 return 0;
1832}
1833
1834static inline
1835int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1836{
1837 if (num_tc > TC_MAX_QUEUE)
1838 return -EINVAL;
1839
1840 dev->num_tc = num_tc;
1841 return 0;
1842}
1843
1844static inline
1845int netdev_get_num_tc(struct net_device *dev)
1846{
1847 return dev->num_tc;
1848}
1849
e8a0464c
DM
1850static inline
1851struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1852 unsigned int index)
1853{
1854 return &dev->_tx[index];
1855}
1856
10c51b56
DB
1857static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1858 const struct sk_buff *skb)
1859{
1860 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1861}
1862
e8a0464c
DM
1863static inline void netdev_for_each_tx_queue(struct net_device *dev,
1864 void (*f)(struct net_device *,
1865 struct netdev_queue *,
1866 void *),
1867 void *arg)
1868{
1869 unsigned int i;
1870
1871 for (i = 0; i < dev->num_tx_queues; i++)
1872 f(dev, &dev->_tx[i], arg);
1873}
1874
f629d208 1875struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1876 struct sk_buff *skb,
1877 void *accel_priv);
8c4c49df 1878
c346dca1
YH
1879/*
1880 * Net namespace inlines
1881 */
1882static inline
1883struct net *dev_net(const struct net_device *dev)
1884{
c2d9ba9b 1885 return read_pnet(&dev->nd_net);
c346dca1
YH
1886}
1887
1888static inline
f5aa23fd 1889void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 1890{
0c5c9fb5 1891 write_pnet(&dev->nd_net, net);
c346dca1
YH
1892}
1893
3e8a72d1 1894static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 1895{
3fc88677 1896#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
1897 if (dev->dsa_ptr != NULL)
1898 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 1899#endif
5aed85ce 1900 return false;
396138f0
LB
1901}
1902
bea3348e
SH
1903/**
1904 * netdev_priv - access network device private data
1905 * @dev: network device
1906 *
1907 * Get network device private data
1908 */
6472ce60 1909static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1910{
1ce8e7b5 1911 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1912}
1913
1da177e4
LT
1914/* Set the sysfs physical device reference for the network logical device
1915 * if set prior to registration will cause a symlink during initialization.
1916 */
43cb76d9 1917#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1918
384912ed 1919/* Set the sysfs device type for the network logical device to allow
3f79410c 1920 * fine-grained identification of different network device types. For
384912ed
MH
1921 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1922 */
1923#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1924
82dc3c63
ED
1925/* Default NAPI poll() weight
1926 * Device drivers are strongly advised to not use bigger value
1927 */
1928#define NAPI_POLL_WEIGHT 64
1929
3b582cc1
SH
1930/**
1931 * netif_napi_add - initialize a napi context
1932 * @dev: network device
1933 * @napi: napi context
1934 * @poll: polling function
1935 * @weight: default weight
1936 *
1937 * netif_napi_add() must be used to initialize a napi context prior to calling
1938 * *any* of the other napi related functions.
1939 */
d565b0a1
HX
1940void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1941 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1942
d8156534
AD
1943/**
1944 * netif_napi_del - remove a napi context
1945 * @napi: napi context
1946 *
1947 * netif_napi_del() removes a napi context from the network device napi list
1948 */
d565b0a1
HX
1949void netif_napi_del(struct napi_struct *napi);
1950
1951struct napi_gro_cb {
78a478d0
HX
1952 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1953 void *frag0;
1954
7489594c
HX
1955 /* Length of frag0. */
1956 unsigned int frag0_len;
1957
86911732
HX
1958 /* This indicates where we are processing relative to skb->data. */
1959 int data_offset;
1960
d565b0a1 1961 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
1962 u16 flush;
1963
1964 /* Save the IP ID here and check when we get to the transport layer */
1965 u16 flush_id;
d565b0a1
HX
1966
1967 /* Number of segments aggregated. */
2e71a6f8
ED
1968 u16 count;
1969
15e2396d
TH
1970 /* Start offset for remote checksum offload */
1971 u16 gro_remcsum_start;
1972
2e71a6f8
ED
1973 /* jiffies when first packet was created/queued */
1974 unsigned long age;
86347245 1975
afe93325 1976 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
1977 u16 proto;
1978
baa32ff4
TH
1979 /* This is non-zero if the packet may be of the same flow. */
1980 u8 same_flow:1;
1981
b582ef09 1982 /* Used in udp_gro_receive */
573e8fca
TH
1983 u8 udp_mark:1;
1984
1985 /* GRO checksum is valid */
1986 u8 csum_valid:1;
1987
662880f4
TH
1988 /* Number of checksums via CHECKSUM_UNNECESSARY */
1989 u8 csum_cnt:3;
c3c7c254 1990
baa32ff4
TH
1991 /* Free the skb? */
1992 u8 free:2;
1993#define NAPI_GRO_FREE 1
1994#define NAPI_GRO_FREE_STOLEN_HEAD 2
1995
efc98d08
TH
1996 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1997 u8 is_ipv6:1;
1998
baa32ff4
TH
1999 /* 7 bit hole */
2000
bf5a755f
JC
2001 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2002 __wsum csum;
2003
c3c7c254
ED
2004 /* used in skb_gro_receive() slow path */
2005 struct sk_buff *last;
d565b0a1
HX
2006};
2007
2008#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2009
1da177e4 2010struct packet_type {
f2ccd8fa
DM
2011 __be16 type; /* This is really htons(ether_type). */
2012 struct net_device *dev; /* NULL is wildcarded here */
2013 int (*func) (struct sk_buff *,
2014 struct net_device *,
2015 struct packet_type *,
2016 struct net_device *);
c0de08d0
EL
2017 bool (*id_match)(struct packet_type *ptype,
2018 struct sock *sk);
1da177e4
LT
2019 void *af_packet_priv;
2020 struct list_head list;
2021};
2022
f191a1d1 2023struct offload_callbacks {
576a30eb 2024 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2025 netdev_features_t features);
d565b0a1 2026 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2027 struct sk_buff *skb);
299603e8 2028 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2029};
2030
2031struct packet_offload {
2032 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2033 u16 priority;
f191a1d1
VY
2034 struct offload_callbacks callbacks;
2035 struct list_head list;
1da177e4
LT
2036};
2037
a2b12f3c
TH
2038struct udp_offload;
2039
2040struct udp_offload_callbacks {
2041 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2042 struct sk_buff *skb,
2043 struct udp_offload *uoff);
2044 int (*gro_complete)(struct sk_buff *skb,
2045 int nhoff,
2046 struct udp_offload *uoff);
2047};
2048
b582ef09
OG
2049struct udp_offload {
2050 __be16 port;
afe93325 2051 u8 ipproto;
a2b12f3c 2052 struct udp_offload_callbacks callbacks;
b582ef09
OG
2053};
2054
8f84985f
LR
2055/* often modified stats are per cpu, other are shared (netdev->stats) */
2056struct pcpu_sw_netstats {
2057 u64 rx_packets;
2058 u64 rx_bytes;
2059 u64 tx_packets;
2060 u64 tx_bytes;
2061 struct u64_stats_sync syncp;
2062};
2063
1c213bd2
WC
2064#define netdev_alloc_pcpu_stats(type) \
2065({ \
693350c2 2066 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1c213bd2 2067 if (pcpu_stats) { \
ec65aafb
JB
2068 int __cpu; \
2069 for_each_possible_cpu(__cpu) { \
1c213bd2 2070 typeof(type) *stat; \
ec65aafb 2071 stat = per_cpu_ptr(pcpu_stats, __cpu); \
1c213bd2
WC
2072 u64_stats_init(&stat->syncp); \
2073 } \
2074 } \
2075 pcpu_stats; \
2076})
2077
1da177e4
LT
2078#include <linux/notifier.h>
2079
dcfe1421
AW
2080/* netdevice notifier chain. Please remember to update the rtnetlink
2081 * notification exclusion list in rtnetlink_event() when adding new
2082 * types.
2083 */
2084#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2085#define NETDEV_DOWN 0x0002
2086#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2087 detected a hardware crash and restarted
2088 - we can use this eg to kick tcp sessions
2089 once done */
2090#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2091#define NETDEV_REGISTER 0x0005
2092#define NETDEV_UNREGISTER 0x0006
1d486bfb 2093#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2094#define NETDEV_CHANGEADDR 0x0008
2095#define NETDEV_GOING_DOWN 0x0009
2096#define NETDEV_CHANGENAME 0x000A
2097#define NETDEV_FEAT_CHANGE 0x000B
2098#define NETDEV_BONDING_FAILOVER 0x000C
2099#define NETDEV_PRE_UP 0x000D
2100#define NETDEV_PRE_TYPE_CHANGE 0x000E
2101#define NETDEV_POST_TYPE_CHANGE 0x000F
2102#define NETDEV_POST_INIT 0x0010
0115e8e3 2103#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2104#define NETDEV_RELEASE 0x0012
2105#define NETDEV_NOTIFY_PEERS 0x0013
2106#define NETDEV_JOIN 0x0014
42e52bf9 2107#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2108#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2109#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2110#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2111#define NETDEV_BONDING_INFO 0x0019
dcfe1421 2112
f629d208
JP
2113int register_netdevice_notifier(struct notifier_block *nb);
2114int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2115
2116struct netdev_notifier_info {
2117 struct net_device *dev;
2118};
2119
be9efd36
JP
2120struct netdev_notifier_change_info {
2121 struct netdev_notifier_info info; /* must be first */
2122 unsigned int flags_changed;
2123};
2124
0e4ead9d
JP
2125struct netdev_notifier_changeupper_info {
2126 struct netdev_notifier_info info; /* must be first */
2127 struct net_device *upper_dev; /* new upper dev */
2128 bool master; /* is upper dev master */
2129 bool linking; /* is the nofication for link or unlink */
2130};
2131
75538c2b
CW
2132static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2133 struct net_device *dev)
2134{
2135 info->dev = dev;
2136}
2137
351638e7
JP
2138static inline struct net_device *
2139netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2140{
2141 return info->dev;
2142}
2143
f629d208 2144int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2145
2146
1da177e4
LT
2147extern rwlock_t dev_base_lock; /* Device list lock */
2148
881d966b
EB
2149#define for_each_netdev(net, d) \
2150 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2151#define for_each_netdev_reverse(net, d) \
2152 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2153#define for_each_netdev_rcu(net, d) \
2154 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2155#define for_each_netdev_safe(net, d, n) \
2156 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2157#define for_each_netdev_continue(net, d) \
2158 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2159#define for_each_netdev_continue_rcu(net, d) \
2160 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2161#define for_each_netdev_in_bond_rcu(bond, slave) \
2162 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2163 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2164#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2165
a050c33f
DL
2166static inline struct net_device *next_net_device(struct net_device *dev)
2167{
2168 struct list_head *lh;
2169 struct net *net;
2170
c346dca1 2171 net = dev_net(dev);
a050c33f
DL
2172 lh = dev->dev_list.next;
2173 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2174}
2175
ce81b76a
ED
2176static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2177{
2178 struct list_head *lh;
2179 struct net *net;
2180
2181 net = dev_net(dev);
ccf43438 2182 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2183 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2184}
2185
a050c33f
DL
2186static inline struct net_device *first_net_device(struct net *net)
2187{
2188 return list_empty(&net->dev_base_head) ? NULL :
2189 net_device_entry(net->dev_base_head.next);
2190}
7562f876 2191
ccf43438
ED
2192static inline struct net_device *first_net_device_rcu(struct net *net)
2193{
2194 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2195
2196 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2197}
2198
f629d208
JP
2199int netdev_boot_setup_check(struct net_device *dev);
2200unsigned long netdev_boot_base(const char *prefix, int unit);
2201struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2202 const char *hwaddr);
2203struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2204struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2205void dev_add_pack(struct packet_type *pt);
2206void dev_remove_pack(struct packet_type *pt);
2207void __dev_remove_pack(struct packet_type *pt);
2208void dev_add_offload(struct packet_offload *po);
2209void dev_remove_offload(struct packet_offload *po);
f629d208 2210
a54acb3a 2211int dev_get_iflink(const struct net_device *dev);
fc4099f1 2212int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2213struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2214 unsigned short mask);
f629d208
JP
2215struct net_device *dev_get_by_name(struct net *net, const char *name);
2216struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2217struct net_device *__dev_get_by_name(struct net *net, const char *name);
2218int dev_alloc_name(struct net_device *dev, const char *name);
2219int dev_open(struct net_device *dev);
2220int dev_close(struct net_device *dev);
99c4a26a 2221int dev_close_many(struct list_head *head, bool unlink);
f629d208 2222void dev_disable_lro(struct net_device *dev);
7026b1dd
DM
2223int dev_loopback_xmit(struct sock *sk, struct sk_buff *newskb);
2224int dev_queue_xmit_sk(struct sock *sk, struct sk_buff *skb);
2225static inline int dev_queue_xmit(struct sk_buff *skb)
2226{
2227 return dev_queue_xmit_sk(skb->sk, skb);
2228}
f663dd9a 2229int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2230int register_netdevice(struct net_device *dev);
2231void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2232void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2233static inline void unregister_netdevice(struct net_device *dev)
2234{
2235 unregister_netdevice_queue(dev, NULL);
2236}
2237
f629d208
JP
2238int netdev_refcnt_read(const struct net_device *dev);
2239void free_netdev(struct net_device *dev);
74d332c1 2240void netdev_freemem(struct net_device *dev);
f629d208
JP
2241void synchronize_net(void);
2242int init_dummy_netdev(struct net_device *dev);
937f1ba5 2243
f60e5990 2244DECLARE_PER_CPU(int, xmit_recursion);
2245static inline int dev_recursion_level(void)
2246{
2247 return this_cpu_read(xmit_recursion);
2248}
2249
f629d208
JP
2250struct net_device *dev_get_by_index(struct net *net, int ifindex);
2251struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2252struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2253int netdev_get_name(struct net *net, char *name, int ifindex);
2254int dev_restart(struct net_device *dev);
f629d208 2255int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2256
2257static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2258{
2259 return NAPI_GRO_CB(skb)->data_offset;
2260}
2261
2262static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2263{
2264 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2265}
2266
2267static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2268{
2269 NAPI_GRO_CB(skb)->data_offset += len;
2270}
2271
a5b1cf28
HX
2272static inline void *skb_gro_header_fast(struct sk_buff *skb,
2273 unsigned int offset)
86911732 2274{
a5b1cf28
HX
2275 return NAPI_GRO_CB(skb)->frag0 + offset;
2276}
78a478d0 2277
a5b1cf28
HX
2278static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2279{
2280 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2281}
78a478d0 2282
a5b1cf28
HX
2283static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2284 unsigned int offset)
2285{
17dd759c
HX
2286 if (!pskb_may_pull(skb, hlen))
2287 return NULL;
2288
a5b1cf28
HX
2289 NAPI_GRO_CB(skb)->frag0 = NULL;
2290 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2291 return skb->data + offset;
86911732 2292}
1da177e4 2293
36e7b1b8
HX
2294static inline void *skb_gro_network_header(struct sk_buff *skb)
2295{
78d3fd0b
HX
2296 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2297 skb_network_offset(skb);
36e7b1b8
HX
2298}
2299
bf5a755f
JC
2300static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2301 const void *start, unsigned int len)
2302{
573e8fca 2303 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2304 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2305 csum_partial(start, len, 0));
2306}
2307
573e8fca
TH
2308/* GRO checksum functions. These are logical equivalents of the normal
2309 * checksum functions (in skbuff.h) except that they operate on the GRO
2310 * offsets and fields in sk_buff.
2311 */
2312
2313__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2314
15e2396d
TH
2315static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2316{
b7fe10e5 2317 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2318}
2319
573e8fca
TH
2320static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2321 bool zero_okay,
2322 __sum16 check)
2323{
6edec0e6
TH
2324 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2325 skb_checksum_start_offset(skb) <
2326 skb_gro_offset(skb)) &&
15e2396d 2327 !skb_at_gro_remcsum_start(skb) &&
662880f4 2328 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2329 (!zero_okay || check));
2330}
2331
2332static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2333 __wsum psum)
2334{
2335 if (NAPI_GRO_CB(skb)->csum_valid &&
2336 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2337 return 0;
2338
2339 NAPI_GRO_CB(skb)->csum = psum;
2340
2341 return __skb_gro_checksum_complete(skb);
2342}
2343
573e8fca
TH
2344static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2345{
662880f4
TH
2346 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2347 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2348 NAPI_GRO_CB(skb)->csum_cnt--;
2349 } else {
2350 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2351 * verified a new top level checksum or an encapsulated one
2352 * during GRO. This saves work if we fallback to normal path.
2353 */
2354 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2355 }
2356}
2357
2358#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2359 compute_pseudo) \
2360({ \
2361 __sum16 __ret = 0; \
2362 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2363 __ret = __skb_gro_checksum_validate_complete(skb, \
2364 compute_pseudo(skb, proto)); \
5a212329
TH
2365 if (__ret) \
2366 __skb_mark_checksum_bad(skb); \
2367 else \
573e8fca
TH
2368 skb_gro_incr_csum_unnecessary(skb); \
2369 __ret; \
2370})
2371
2372#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2373 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2374
2375#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2376 compute_pseudo) \
2377 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2378
2379#define skb_gro_checksum_simple_validate(skb) \
2380 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2381
d96535a1
TH
2382static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2383{
2384 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2385 !NAPI_GRO_CB(skb)->csum_valid);
2386}
2387
2388static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2389 __sum16 check, __wsum pseudo)
2390{
2391 NAPI_GRO_CB(skb)->csum = ~pseudo;
2392 NAPI_GRO_CB(skb)->csum_valid = 1;
2393}
2394
2395#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2396do { \
2397 if (__skb_gro_checksum_convert_check(skb)) \
2398 __skb_gro_checksum_convert(skb, check, \
2399 compute_pseudo(skb, proto)); \
2400} while (0)
2401
26c4f7da
TH
2402struct gro_remcsum {
2403 int offset;
2404 __wsum delta;
2405};
2406
2407static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2408{
846cd667 2409 grc->offset = 0;
26c4f7da
TH
2410 grc->delta = 0;
2411}
2412
b7fe10e5
TH
2413static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2414 unsigned int off, size_t hdrlen,
2415 int start, int offset,
2416 struct gro_remcsum *grc,
2417 bool nopartial)
dcdc8994
TH
2418{
2419 __wsum delta;
b7fe10e5 2420 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2421
2422 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2423
15e2396d 2424 if (!nopartial) {
b7fe10e5
TH
2425 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2426 return ptr;
2427 }
2428
2429 ptr = skb_gro_header_fast(skb, off);
2430 if (skb_gro_header_hard(skb, off + plen)) {
2431 ptr = skb_gro_header_slow(skb, off + plen, off);
2432 if (!ptr)
2433 return NULL;
15e2396d
TH
2434 }
2435
b7fe10e5
TH
2436 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2437 start, offset);
dcdc8994
TH
2438
2439 /* Adjust skb->csum since we changed the packet */
dcdc8994 2440 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2441
b7fe10e5 2442 grc->offset = off + hdrlen + offset;
26c4f7da 2443 grc->delta = delta;
b7fe10e5
TH
2444
2445 return ptr;
dcdc8994
TH
2446}
2447
26c4f7da
TH
2448static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2449 struct gro_remcsum *grc)
2450{
b7fe10e5
TH
2451 void *ptr;
2452 size_t plen = grc->offset + sizeof(u16);
2453
26c4f7da
TH
2454 if (!grc->delta)
2455 return;
2456
b7fe10e5
TH
2457 ptr = skb_gro_header_fast(skb, grc->offset);
2458 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2459 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2460 if (!ptr)
2461 return;
2462 }
2463
2464 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2465}
dcdc8994 2466
0c4e8581
SH
2467static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2468 unsigned short type,
3b04ddde 2469 const void *daddr, const void *saddr,
95c96174 2470 unsigned int len)
0c4e8581 2471{
f1ecfd5d 2472 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2473 return 0;
3b04ddde
SH
2474
2475 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2476}
2477
b95cce35
SH
2478static inline int dev_parse_header(const struct sk_buff *skb,
2479 unsigned char *haddr)
2480{
2481 const struct net_device *dev = skb->dev;
2482
1b83336b 2483 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2484 return 0;
3b04ddde 2485 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2486}
2487
1da177e4 2488typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2489int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2490static inline int unregister_gifconf(unsigned int family)
2491{
2492 return register_gifconf(family, NULL);
2493}
2494
99bbc707 2495#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2496#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2497struct sd_flow_limit {
2498 u64 count;
2499 unsigned int num_buckets;
2500 unsigned int history_head;
2501 u16 history[FLOW_LIMIT_HISTORY];
2502 u8 buckets[];
2503};
2504
2505extern int netdev_flow_limit_table_len;
2506#endif /* CONFIG_NET_FLOW_LIMIT */
2507
1da177e4 2508/*
88751275 2509 * Incoming packets are placed on per-cpu queues
1da177e4 2510 */
d94d9fee 2511struct softnet_data {
1da177e4 2512 struct list_head poll_list;
6e7676c1 2513 struct sk_buff_head process_queue;
1da177e4 2514
dee42870 2515 /* stats */
cd7b5396
DM
2516 unsigned int processed;
2517 unsigned int time_squeeze;
2518 unsigned int cpu_collision;
2519 unsigned int received_rps;
fd793d89 2520#ifdef CONFIG_RPS
88751275 2521 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2522#endif
2523#ifdef CONFIG_NET_FLOW_LIMIT
2524 struct sd_flow_limit __rcu *flow_limit;
2525#endif
2526 struct Qdisc *output_queue;
2527 struct Qdisc **output_queue_tailp;
2528 struct sk_buff *completion_queue;
88751275 2529
4cdb1e2e 2530#ifdef CONFIG_RPS
88751275 2531 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 2532 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2533 struct softnet_data *rps_ipi_next;
2534 unsigned int cpu;
fec5e652 2535 unsigned int input_queue_head;
76cc8b13 2536 unsigned int input_queue_tail;
1e94d72f 2537#endif
95c96174 2538 unsigned int dropped;
0a9627f2 2539 struct sk_buff_head input_pkt_queue;
bea3348e 2540 struct napi_struct backlog;
99bbc707 2541
1da177e4
LT
2542};
2543
76cc8b13 2544static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2545{
2546#ifdef CONFIG_RPS
76cc8b13
TH
2547 sd->input_queue_head++;
2548#endif
2549}
2550
2551static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2552 unsigned int *qtail)
2553{
2554#ifdef CONFIG_RPS
2555 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2556#endif
2557}
2558
0a9627f2 2559DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2560
f629d208 2561void __netif_schedule(struct Qdisc *q);
46e5da40 2562void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2563
fd2ea0a7
DM
2564static inline void netif_tx_schedule_all(struct net_device *dev)
2565{
2566 unsigned int i;
2567
2568 for (i = 0; i < dev->num_tx_queues; i++)
2569 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2570}
2571
d29f749e
DJ
2572static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2573{
73466498 2574 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2575}
2576
bea3348e
SH
2577/**
2578 * netif_start_queue - allow transmit
2579 * @dev: network device
2580 *
2581 * Allow upper layers to call the device hard_start_xmit routine.
2582 */
1da177e4
LT
2583static inline void netif_start_queue(struct net_device *dev)
2584{
e8a0464c 2585 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2586}
2587
fd2ea0a7
DM
2588static inline void netif_tx_start_all_queues(struct net_device *dev)
2589{
2590 unsigned int i;
2591
2592 for (i = 0; i < dev->num_tx_queues; i++) {
2593 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2594 netif_tx_start_queue(txq);
2595 }
2596}
2597
46e5da40 2598void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2599
d29f749e
DJ
2600/**
2601 * netif_wake_queue - restart transmit
2602 * @dev: network device
2603 *
2604 * Allow upper layers to call the device hard_start_xmit routine.
2605 * Used for flow control when transmit resources are available.
2606 */
79d16385
DM
2607static inline void netif_wake_queue(struct net_device *dev)
2608{
e8a0464c 2609 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2610}
2611
fd2ea0a7
DM
2612static inline void netif_tx_wake_all_queues(struct net_device *dev)
2613{
2614 unsigned int i;
2615
2616 for (i = 0; i < dev->num_tx_queues; i++) {
2617 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2618 netif_tx_wake_queue(txq);
2619 }
2620}
2621
d29f749e
DJ
2622static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2623{
73466498 2624 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2625}
2626
bea3348e
SH
2627/**
2628 * netif_stop_queue - stop transmitted packets
2629 * @dev: network device
2630 *
2631 * Stop upper layers calling the device hard_start_xmit routine.
2632 * Used for flow control when transmit resources are unavailable.
2633 */
1da177e4
LT
2634static inline void netif_stop_queue(struct net_device *dev)
2635{
e8a0464c 2636 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2637}
2638
a2029240 2639void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2640
4d29515f 2641static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2642{
73466498 2643 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2644}
2645
bea3348e
SH
2646/**
2647 * netif_queue_stopped - test if transmit queue is flowblocked
2648 * @dev: network device
2649 *
2650 * Test if transmit queue on device is currently unable to send.
2651 */
4d29515f 2652static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2653{
e8a0464c 2654 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2655}
2656
4d29515f 2657static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2658{
73466498
TH
2659 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2660}
2661
8e2f1a63
DB
2662static inline bool
2663netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2664{
2665 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2666}
2667
8e2f1a63
DB
2668static inline bool
2669netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2670{
2671 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2672}
2673
53511453
ED
2674/**
2675 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2676 * @dev_queue: pointer to transmit queue
2677 *
2678 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2679 * to give appropriate hint to the cpu.
2680 */
2681static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2682{
2683#ifdef CONFIG_BQL
2684 prefetchw(&dev_queue->dql.num_queued);
2685#endif
2686}
2687
2688/**
2689 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2690 * @dev_queue: pointer to transmit queue
2691 *
2692 * BQL enabled drivers might use this helper in their TX completion path,
2693 * to give appropriate hint to the cpu.
2694 */
2695static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2696{
2697#ifdef CONFIG_BQL
2698 prefetchw(&dev_queue->dql.limit);
2699#endif
2700}
2701
c5d67bd7
TH
2702static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2703 unsigned int bytes)
2704{
114cf580
TH
2705#ifdef CONFIG_BQL
2706 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2707
2708 if (likely(dql_avail(&dev_queue->dql) >= 0))
2709 return;
2710
2711 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2712
2713 /*
2714 * The XOFF flag must be set before checking the dql_avail below,
2715 * because in netdev_tx_completed_queue we update the dql_completed
2716 * before checking the XOFF flag.
2717 */
2718 smp_mb();
2719
2720 /* check again in case another CPU has just made room avail */
2721 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2722 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2723#endif
c5d67bd7
TH
2724}
2725
0042d0c8
FF
2726/**
2727 * netdev_sent_queue - report the number of bytes queued to hardware
2728 * @dev: network device
2729 * @bytes: number of bytes queued to the hardware device queue
2730 *
2731 * Report the number of bytes queued for sending/completion to the network
2732 * device hardware queue. @bytes should be a good approximation and should
2733 * exactly match netdev_completed_queue() @bytes
2734 */
c5d67bd7
TH
2735static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2736{
2737 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2738}
2739
2740static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2741 unsigned int pkts, unsigned int bytes)
c5d67bd7 2742{
114cf580 2743#ifdef CONFIG_BQL
b37c0fbe
AD
2744 if (unlikely(!bytes))
2745 return;
2746
2747 dql_completed(&dev_queue->dql, bytes);
2748
2749 /*
2750 * Without the memory barrier there is a small possiblity that
2751 * netdev_tx_sent_queue will miss the update and cause the queue to
2752 * be stopped forever
2753 */
2754 smp_mb();
2755
2756 if (dql_avail(&dev_queue->dql) < 0)
2757 return;
2758
2759 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2760 netif_schedule_queue(dev_queue);
114cf580 2761#endif
c5d67bd7
TH
2762}
2763
0042d0c8
FF
2764/**
2765 * netdev_completed_queue - report bytes and packets completed by device
2766 * @dev: network device
2767 * @pkts: actual number of packets sent over the medium
2768 * @bytes: actual number of bytes sent over the medium
2769 *
2770 * Report the number of bytes and packets transmitted by the network device
2771 * hardware queue over the physical medium, @bytes must exactly match the
2772 * @bytes amount passed to netdev_sent_queue()
2773 */
c5d67bd7 2774static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2775 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2776{
2777 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2778}
2779
2780static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2781{
114cf580 2782#ifdef CONFIG_BQL
5c490354 2783 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2784 dql_reset(&q->dql);
2785#endif
c5d67bd7
TH
2786}
2787
0042d0c8
FF
2788/**
2789 * netdev_reset_queue - reset the packets and bytes count of a network device
2790 * @dev_queue: network device
2791 *
2792 * Reset the bytes and packet count of a network device and clear the
2793 * software flow control OFF bit for this network device
2794 */
c5d67bd7
TH
2795static inline void netdev_reset_queue(struct net_device *dev_queue)
2796{
2797 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2798}
2799
b9507bda
DB
2800/**
2801 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2802 * @dev: network device
2803 * @queue_index: given tx queue index
2804 *
2805 * Returns 0 if given tx queue index >= number of device tx queues,
2806 * otherwise returns the originally passed tx queue index.
2807 */
2808static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2809{
2810 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2811 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2812 dev->name, queue_index,
2813 dev->real_num_tx_queues);
2814 return 0;
2815 }
2816
2817 return queue_index;
2818}
2819
bea3348e
SH
2820/**
2821 * netif_running - test if up
2822 * @dev: network device
2823 *
2824 * Test if the device has been brought up.
2825 */
4d29515f 2826static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2827{
2828 return test_bit(__LINK_STATE_START, &dev->state);
2829}
2830
f25f4e44
PWJ
2831/*
2832 * Routines to manage the subqueues on a device. We only need start
2833 * stop, and a check if it's stopped. All other device management is
2834 * done at the overall netdevice level.
2835 * Also test the device if we're multiqueue.
2836 */
bea3348e
SH
2837
2838/**
2839 * netif_start_subqueue - allow sending packets on subqueue
2840 * @dev: network device
2841 * @queue_index: sub queue index
2842 *
2843 * Start individual transmit queue of a device with multiple transmit queues.
2844 */
f25f4e44
PWJ
2845static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2846{
fd2ea0a7 2847 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2848
2849 netif_tx_start_queue(txq);
f25f4e44
PWJ
2850}
2851
bea3348e
SH
2852/**
2853 * netif_stop_subqueue - stop sending packets on subqueue
2854 * @dev: network device
2855 * @queue_index: sub queue index
2856 *
2857 * Stop individual transmit queue of a device with multiple transmit queues.
2858 */
f25f4e44
PWJ
2859static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2860{
fd2ea0a7 2861 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 2862 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2863}
2864
bea3348e
SH
2865/**
2866 * netif_subqueue_stopped - test status of subqueue
2867 * @dev: network device
2868 * @queue_index: sub queue index
2869 *
2870 * Check individual transmit queue of a device with multiple transmit queues.
2871 */
4d29515f
DM
2872static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2873 u16 queue_index)
f25f4e44 2874{
fd2ea0a7 2875 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2876
2877 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2878}
2879
4d29515f
DM
2880static inline bool netif_subqueue_stopped(const struct net_device *dev,
2881 struct sk_buff *skb)
668f895a
PE
2882{
2883 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2884}
bea3348e 2885
46e5da40 2886void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 2887
537c00de 2888#ifdef CONFIG_XPS
53af53ae 2889int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2890 u16 index);
537c00de
AD
2891#else
2892static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2893 const struct cpumask *mask,
537c00de
AD
2894 u16 index)
2895{
2896 return 0;
2897}
2898#endif
2899
5605c762
JP
2900u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2901 unsigned int num_tx_queues);
2902
a3d22a68
VZ
2903/*
2904 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2905 * as a distribution range limit for the returned value.
2906 */
2907static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 2908 struct sk_buff *skb)
a3d22a68
VZ
2909{
2910 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2911}
2912
bea3348e
SH
2913/**
2914 * netif_is_multiqueue - test if device has multiple transmit queues
2915 * @dev: network device
2916 *
2917 * Check if device has multiple transmit queues
bea3348e 2918 */
4d29515f 2919static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2920{
a02cec21 2921 return dev->num_tx_queues > 1;
f25f4e44 2922}
1da177e4 2923
f629d208 2924int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2925
a953be53 2926#ifdef CONFIG_SYSFS
f629d208 2927int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
2928#else
2929static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2930 unsigned int rxq)
2931{
2932 return 0;
2933}
2934#endif
2935
a953be53
MD
2936#ifdef CONFIG_SYSFS
2937static inline unsigned int get_netdev_rx_queue_index(
2938 struct netdev_rx_queue *queue)
2939{
2940 struct net_device *dev = queue->dev;
2941 int index = queue - dev->_rx;
2942
2943 BUG_ON(index >= dev->num_rx_queues);
2944 return index;
2945}
2946#endif
2947
16917b87 2948#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 2949int netif_get_num_default_rss_queues(void);
16917b87 2950
e6247027
ED
2951enum skb_free_reason {
2952 SKB_REASON_CONSUMED,
2953 SKB_REASON_DROPPED,
2954};
2955
2956void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2957void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 2958
e6247027
ED
2959/*
2960 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2961 * interrupt context or with hardware interrupts being disabled.
2962 * (in_irq() || irqs_disabled())
2963 *
2964 * We provide four helpers that can be used in following contexts :
2965 *
2966 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2967 * replacing kfree_skb(skb)
2968 *
2969 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2970 * Typically used in place of consume_skb(skb) in TX completion path
2971 *
2972 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2973 * replacing kfree_skb(skb)
2974 *
2975 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2976 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 2977 */
e6247027
ED
2978static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2979{
2980 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2981}
2982
2983static inline void dev_consume_skb_irq(struct sk_buff *skb)
2984{
2985 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2986}
2987
2988static inline void dev_kfree_skb_any(struct sk_buff *skb)
2989{
2990 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2991}
2992
2993static inline void dev_consume_skb_any(struct sk_buff *skb)
2994{
2995 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2996}
1da177e4 2997
f629d208
JP
2998int netif_rx(struct sk_buff *skb);
2999int netif_rx_ni(struct sk_buff *skb);
7026b1dd
DM
3000int netif_receive_skb_sk(struct sock *sk, struct sk_buff *skb);
3001static inline int netif_receive_skb(struct sk_buff *skb)
3002{
3003 return netif_receive_skb_sk(skb->sk, skb);
3004}
f629d208
JP
3005gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3006void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3007struct sk_buff *napi_get_frags(struct napi_struct *napi);
3008gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3009struct packet_offload *gro_find_receive_by_type(__be16 type);
3010struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3011
3012static inline void napi_free_frags(struct napi_struct *napi)
3013{
3014 kfree_skb(napi->skb);
3015 napi->skb = NULL;
3016}
3017
f629d208
JP
3018int netdev_rx_handler_register(struct net_device *dev,
3019 rx_handler_func_t *rx_handler,
3020 void *rx_handler_data);
3021void netdev_rx_handler_unregister(struct net_device *dev);
3022
3023bool dev_valid_name(const char *name);
3024int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3025int dev_ethtool(struct net *net, struct ifreq *);
3026unsigned int dev_get_flags(const struct net_device *);
3027int __dev_change_flags(struct net_device *, unsigned int flags);
3028int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3029void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3030 unsigned int gchanges);
f629d208
JP
3031int dev_change_name(struct net_device *, const char *);
3032int dev_set_alias(struct net_device *, const char *, size_t);
3033int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3034int dev_set_mtu(struct net_device *, int);
3035void dev_set_group(struct net_device *, int);
3036int dev_set_mac_address(struct net_device *, struct sockaddr *);
3037int dev_change_carrier(struct net_device *, bool new_carrier);
3038int dev_get_phys_port_id(struct net_device *dev,
02637fce 3039 struct netdev_phys_item_id *ppid);
db24a904
DA
3040int dev_get_phys_port_name(struct net_device *dev,
3041 char *name, size_t len);
d746d707 3042int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3043struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3044struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3045 struct netdev_queue *txq, int *ret);
a0265d28 3046int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3047int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1ee481fb 3048bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
1da177e4 3049
20380731 3050extern int netdev_budget;
1da177e4
LT
3051
3052/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3053void netdev_run_todo(void);
1da177e4 3054
bea3348e
SH
3055/**
3056 * dev_put - release reference to device
3057 * @dev: network device
3058 *
9ef4429b 3059 * Release reference to device to allow it to be freed.
bea3348e 3060 */
1da177e4
LT
3061static inline void dev_put(struct net_device *dev)
3062{
933393f5 3063 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3064}
3065
bea3348e
SH
3066/**
3067 * dev_hold - get reference to device
3068 * @dev: network device
3069 *
9ef4429b 3070 * Hold reference to device to keep it from being freed.
bea3348e 3071 */
15333061
SH
3072static inline void dev_hold(struct net_device *dev)
3073{
933393f5 3074 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3075}
1da177e4
LT
3076
3077/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3078 * and _off may be called from IRQ context, but it is caller
3079 * who is responsible for serialization of these calls.
b00055aa
SR
3080 *
3081 * The name carrier is inappropriate, these functions should really be
3082 * called netif_lowerlayer_*() because they represent the state of any
3083 * kind of lower layer not just hardware media.
1da177e4
LT
3084 */
3085
f629d208
JP
3086void linkwatch_init_dev(struct net_device *dev);
3087void linkwatch_fire_event(struct net_device *dev);
3088void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3089
bea3348e
SH
3090/**
3091 * netif_carrier_ok - test if carrier present
3092 * @dev: network device
3093 *
3094 * Check if carrier is present on device
3095 */
4d29515f 3096static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3097{
3098 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3099}
3100
f629d208 3101unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3102
f629d208 3103void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3104
f629d208 3105void netif_carrier_on(struct net_device *dev);
1da177e4 3106
f629d208 3107void netif_carrier_off(struct net_device *dev);
1da177e4 3108
bea3348e
SH
3109/**
3110 * netif_dormant_on - mark device as dormant.
3111 * @dev: network device
3112 *
3113 * Mark device as dormant (as per RFC2863).
3114 *
3115 * The dormant state indicates that the relevant interface is not
3116 * actually in a condition to pass packets (i.e., it is not 'up') but is
3117 * in a "pending" state, waiting for some external event. For "on-
3118 * demand" interfaces, this new state identifies the situation where the
3119 * interface is waiting for events to place it in the up state.
3120 *
3121 */
b00055aa
SR
3122static inline void netif_dormant_on(struct net_device *dev)
3123{
3124 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3125 linkwatch_fire_event(dev);
3126}
3127
bea3348e
SH
3128/**
3129 * netif_dormant_off - set device as not dormant.
3130 * @dev: network device
3131 *
3132 * Device is not in dormant state.
3133 */
b00055aa
SR
3134static inline void netif_dormant_off(struct net_device *dev)
3135{
3136 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3137 linkwatch_fire_event(dev);
3138}
3139
bea3348e
SH
3140/**
3141 * netif_dormant - test if carrier present
3142 * @dev: network device
3143 *
3144 * Check if carrier is present on device
3145 */
4d29515f 3146static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3147{
3148 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3149}
3150
3151
bea3348e
SH
3152/**
3153 * netif_oper_up - test if device is operational
3154 * @dev: network device
3155 *
3156 * Check if carrier is operational
3157 */
4d29515f 3158static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3159{
b00055aa
SR
3160 return (dev->operstate == IF_OPER_UP ||
3161 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3162}
3163
bea3348e
SH
3164/**
3165 * netif_device_present - is device available or removed
3166 * @dev: network device
3167 *
3168 * Check if device has not been removed from system.
3169 */
4d29515f 3170static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3171{
3172 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3173}
3174
f629d208 3175void netif_device_detach(struct net_device *dev);
1da177e4 3176
f629d208 3177void netif_device_attach(struct net_device *dev);
1da177e4
LT
3178
3179/*
3180 * Network interface message level settings
3181 */
1da177e4
LT
3182
3183enum {
3184 NETIF_MSG_DRV = 0x0001,
3185 NETIF_MSG_PROBE = 0x0002,
3186 NETIF_MSG_LINK = 0x0004,
3187 NETIF_MSG_TIMER = 0x0008,
3188 NETIF_MSG_IFDOWN = 0x0010,
3189 NETIF_MSG_IFUP = 0x0020,
3190 NETIF_MSG_RX_ERR = 0x0040,
3191 NETIF_MSG_TX_ERR = 0x0080,
3192 NETIF_MSG_TX_QUEUED = 0x0100,
3193 NETIF_MSG_INTR = 0x0200,
3194 NETIF_MSG_TX_DONE = 0x0400,
3195 NETIF_MSG_RX_STATUS = 0x0800,
3196 NETIF_MSG_PKTDATA = 0x1000,
3197 NETIF_MSG_HW = 0x2000,
3198 NETIF_MSG_WOL = 0x4000,
3199};
3200
3201#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3202#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3203#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3204#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3205#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3206#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3207#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3208#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3209#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3210#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3211#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3212#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3213#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3214#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3215#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3216
3217static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3218{
3219 /* use default */
3220 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3221 return default_msg_enable_bits;
3222 if (debug_value == 0) /* no output */
3223 return 0;
3224 /* set low N bits */
3225 return (1 << debug_value) - 1;
3226}
3227
c773e847 3228static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3229{
c773e847
DM
3230 spin_lock(&txq->_xmit_lock);
3231 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3232}
3233
fd2ea0a7
DM
3234static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3235{
3236 spin_lock_bh(&txq->_xmit_lock);
3237 txq->xmit_lock_owner = smp_processor_id();
3238}
3239
4d29515f 3240static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3241{
4d29515f 3242 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3243 if (likely(ok))
3244 txq->xmit_lock_owner = smp_processor_id();
3245 return ok;
3246}
3247
3248static inline void __netif_tx_unlock(struct netdev_queue *txq)
3249{
3250 txq->xmit_lock_owner = -1;
3251 spin_unlock(&txq->_xmit_lock);
3252}
3253
3254static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3255{
3256 txq->xmit_lock_owner = -1;
3257 spin_unlock_bh(&txq->_xmit_lock);
3258}
3259
08baf561
ED
3260static inline void txq_trans_update(struct netdev_queue *txq)
3261{
3262 if (txq->xmit_lock_owner != -1)
3263 txq->trans_start = jiffies;
3264}
3265
d29f749e
DJ
3266/**
3267 * netif_tx_lock - grab network device transmit lock
3268 * @dev: network device
d29f749e
DJ
3269 *
3270 * Get network device transmit lock
3271 */
22dd7495
JHS
3272static inline void netif_tx_lock(struct net_device *dev)
3273{
e8a0464c 3274 unsigned int i;
c3f26a26 3275 int cpu;
c773e847 3276
c3f26a26
DM
3277 spin_lock(&dev->tx_global_lock);
3278 cpu = smp_processor_id();
e8a0464c
DM
3279 for (i = 0; i < dev->num_tx_queues; i++) {
3280 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3281
3282 /* We are the only thread of execution doing a
3283 * freeze, but we have to grab the _xmit_lock in
3284 * order to synchronize with threads which are in
3285 * the ->hard_start_xmit() handler and already
3286 * checked the frozen bit.
3287 */
e8a0464c 3288 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3289 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3290 __netif_tx_unlock(txq);
e8a0464c 3291 }
932ff279
HX
3292}
3293
3294static inline void netif_tx_lock_bh(struct net_device *dev)
3295{
e8a0464c
DM
3296 local_bh_disable();
3297 netif_tx_lock(dev);
932ff279
HX
3298}
3299
932ff279
HX
3300static inline void netif_tx_unlock(struct net_device *dev)
3301{
e8a0464c
DM
3302 unsigned int i;
3303
3304 for (i = 0; i < dev->num_tx_queues; i++) {
3305 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3306
c3f26a26
DM
3307 /* No need to grab the _xmit_lock here. If the
3308 * queue is not stopped for another reason, we
3309 * force a schedule.
3310 */
3311 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3312 netif_schedule_queue(txq);
c3f26a26
DM
3313 }
3314 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3315}
3316
3317static inline void netif_tx_unlock_bh(struct net_device *dev)
3318{
e8a0464c
DM
3319 netif_tx_unlock(dev);
3320 local_bh_enable();
932ff279
HX
3321}
3322
c773e847 3323#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3324 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3325 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3326 } \
3327}
3328
5efeac44
EB
3329#define HARD_TX_TRYLOCK(dev, txq) \
3330 (((dev->features & NETIF_F_LLTX) == 0) ? \
3331 __netif_tx_trylock(txq) : \
3332 true )
3333
c773e847 3334#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3335 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3336 __netif_tx_unlock(txq); \
22dd7495
JHS
3337 } \
3338}
3339
1da177e4
LT
3340static inline void netif_tx_disable(struct net_device *dev)
3341{
fd2ea0a7 3342 unsigned int i;
c3f26a26 3343 int cpu;
fd2ea0a7 3344
c3f26a26
DM
3345 local_bh_disable();
3346 cpu = smp_processor_id();
fd2ea0a7
DM
3347 for (i = 0; i < dev->num_tx_queues; i++) {
3348 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3349
3350 __netif_tx_lock(txq, cpu);
fd2ea0a7 3351 netif_tx_stop_queue(txq);
c3f26a26 3352 __netif_tx_unlock(txq);
fd2ea0a7 3353 }
c3f26a26 3354 local_bh_enable();
1da177e4
LT
3355}
3356
e308a5d8
DM
3357static inline void netif_addr_lock(struct net_device *dev)
3358{
3359 spin_lock(&dev->addr_list_lock);
3360}
3361
2429f7ac
JP
3362static inline void netif_addr_lock_nested(struct net_device *dev)
3363{
25175ba5
VY
3364 int subclass = SINGLE_DEPTH_NESTING;
3365
3366 if (dev->netdev_ops->ndo_get_lock_subclass)
3367 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3368
3369 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3370}
3371
e308a5d8
DM
3372static inline void netif_addr_lock_bh(struct net_device *dev)
3373{
3374 spin_lock_bh(&dev->addr_list_lock);
3375}
3376
3377static inline void netif_addr_unlock(struct net_device *dev)
3378{
3379 spin_unlock(&dev->addr_list_lock);
3380}
3381
3382static inline void netif_addr_unlock_bh(struct net_device *dev)
3383{
3384 spin_unlock_bh(&dev->addr_list_lock);
3385}
3386
f001fde5 3387/*
31278e71 3388 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3389 * rcu_read_lock held.
3390 */
3391#define for_each_dev_addr(dev, ha) \
31278e71 3392 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3393
1da177e4
LT
3394/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3395
f629d208 3396void ether_setup(struct net_device *dev);
1da177e4
LT
3397
3398/* Support for loadable net-drivers */
f629d208 3399struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3400 unsigned char name_assign_type,
f629d208
JP
3401 void (*setup)(struct net_device *),
3402 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3403#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3404 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3405
c835a677
TG
3406#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3407 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3408 count)
36909ea4 3409
f629d208
JP
3410int register_netdev(struct net_device *dev);
3411void unregister_netdev(struct net_device *dev);
f001fde5 3412
22bedad3 3413/* General hardware address lists handling functions */
f629d208
JP
3414int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3415 struct netdev_hw_addr_list *from_list, int addr_len);
3416void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3417 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3418int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3419 struct net_device *dev,
3420 int (*sync)(struct net_device *, const unsigned char *),
3421 int (*unsync)(struct net_device *,
3422 const unsigned char *));
3423void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3424 struct net_device *dev,
3425 int (*unsync)(struct net_device *,
3426 const unsigned char *));
f629d208 3427void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3428
f001fde5 3429/* Functions used for device addresses handling */
f629d208
JP
3430int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3431 unsigned char addr_type);
3432int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3433 unsigned char addr_type);
f629d208
JP
3434void dev_addr_flush(struct net_device *dev);
3435int dev_addr_init(struct net_device *dev);
a748ee24
JP
3436
3437/* Functions used for unicast addresses handling */
f629d208
JP
3438int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3439int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3440int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3441int dev_uc_sync(struct net_device *to, struct net_device *from);
3442int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3443void dev_uc_unsync(struct net_device *to, struct net_device *from);
3444void dev_uc_flush(struct net_device *dev);
3445void dev_uc_init(struct net_device *dev);
f001fde5 3446
670e5b8e
AD
3447/**
3448 * __dev_uc_sync - Synchonize device's unicast list
3449 * @dev: device to sync
3450 * @sync: function to call if address should be added
3451 * @unsync: function to call if address should be removed
3452 *
3453 * Add newly added addresses to the interface, and release
3454 * addresses that have been deleted.
3455 **/
3456static inline int __dev_uc_sync(struct net_device *dev,
3457 int (*sync)(struct net_device *,
3458 const unsigned char *),
3459 int (*unsync)(struct net_device *,
3460 const unsigned char *))
3461{
3462 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3463}
3464
3465/**
e793c0f7 3466 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3467 * @dev: device to sync
3468 * @unsync: function to call if address should be removed
3469 *
3470 * Remove all addresses that were added to the device by dev_uc_sync().
3471 **/
3472static inline void __dev_uc_unsync(struct net_device *dev,
3473 int (*unsync)(struct net_device *,
3474 const unsigned char *))
3475{
3476 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3477}
3478
22bedad3 3479/* Functions used for multicast addresses handling */
f629d208
JP
3480int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3481int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3482int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3483int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3484int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3485int dev_mc_sync(struct net_device *to, struct net_device *from);
3486int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3487void dev_mc_unsync(struct net_device *to, struct net_device *from);
3488void dev_mc_flush(struct net_device *dev);
3489void dev_mc_init(struct net_device *dev);
f001fde5 3490
670e5b8e
AD
3491/**
3492 * __dev_mc_sync - Synchonize device's multicast list
3493 * @dev: device to sync
3494 * @sync: function to call if address should be added
3495 * @unsync: function to call if address should be removed
3496 *
3497 * Add newly added addresses to the interface, and release
3498 * addresses that have been deleted.
3499 **/
3500static inline int __dev_mc_sync(struct net_device *dev,
3501 int (*sync)(struct net_device *,
3502 const unsigned char *),
3503 int (*unsync)(struct net_device *,
3504 const unsigned char *))
3505{
3506 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3507}
3508
3509/**
e793c0f7 3510 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3511 * @dev: device to sync
3512 * @unsync: function to call if address should be removed
3513 *
3514 * Remove all addresses that were added to the device by dev_mc_sync().
3515 **/
3516static inline void __dev_mc_unsync(struct net_device *dev,
3517 int (*unsync)(struct net_device *,
3518 const unsigned char *))
3519{
3520 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3521}
3522
4417da66 3523/* Functions used for secondary unicast and multicast support */
f629d208
JP
3524void dev_set_rx_mode(struct net_device *dev);
3525void __dev_set_rx_mode(struct net_device *dev);
3526int dev_set_promiscuity(struct net_device *dev, int inc);
3527int dev_set_allmulti(struct net_device *dev, int inc);
3528void netdev_state_change(struct net_device *dev);
3529void netdev_notify_peers(struct net_device *dev);
3530void netdev_features_change(struct net_device *dev);
1da177e4 3531/* Load a device via the kmod */
f629d208
JP
3532void dev_load(struct net *net, const char *name);
3533struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3534 struct rtnl_link_stats64 *storage);
3535void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3536 const struct net_device_stats *netdev_stats);
eeda3fd6 3537
1da177e4 3538extern int netdev_max_backlog;
3b098e2d 3539extern int netdev_tstamp_prequeue;
1da177e4 3540extern int weight_p;
0a14842f 3541extern int bpf_jit_enable;
9ff162a8 3542
f629d208 3543bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3544struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3545 struct list_head **iter);
f629d208
JP
3546struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3547 struct list_head **iter);
8b5be856 3548
44a40855
VY
3549/* iterate through upper list, must be called under RCU read lock */
3550#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3551 for (iter = &(dev)->adj_list.upper, \
3552 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3553 updev; \
3554 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3555
8b5be856 3556/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3557#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3558 for (iter = &(dev)->all_adj_list.upper, \
3559 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3560 updev; \
3561 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3562
f629d208
JP
3563void *netdev_lower_get_next_private(struct net_device *dev,
3564 struct list_head **iter);
3565void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3566 struct list_head **iter);
31088a11
VF
3567
3568#define netdev_for_each_lower_private(dev, priv, iter) \
3569 for (iter = (dev)->adj_list.lower.next, \
3570 priv = netdev_lower_get_next_private(dev, &(iter)); \
3571 priv; \
3572 priv = netdev_lower_get_next_private(dev, &(iter)))
3573
3574#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3575 for (iter = &(dev)->adj_list.lower, \
3576 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3577 priv; \
3578 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3579
4085ebe8
VY
3580void *netdev_lower_get_next(struct net_device *dev,
3581 struct list_head **iter);
3582#define netdev_for_each_lower_dev(dev, ldev, iter) \
3583 for (iter = &(dev)->adj_list.lower, \
3584 ldev = netdev_lower_get_next(dev, &(iter)); \
3585 ldev; \
3586 ldev = netdev_lower_get_next(dev, &(iter)))
3587
f629d208 3588void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3589void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3590struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3591struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3592int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3593int netdev_master_upper_dev_link(struct net_device *dev,
9ff162a8 3594 struct net_device *upper_dev);
f629d208
JP
3595int netdev_master_upper_dev_link_private(struct net_device *dev,
3596 struct net_device *upper_dev,
3597 void *private);
3598void netdev_upper_dev_unlink(struct net_device *dev,
3599 struct net_device *upper_dev);
5bb025fa 3600void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3601void *netdev_lower_dev_get_private(struct net_device *dev,
3602 struct net_device *lower_dev);
960fb622
ED
3603
3604/* RSS keys are 40 or 52 bytes long */
3605#define NETDEV_RSS_KEY_LEN 52
3606extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3607void netdev_rss_key_fill(void *buffer, size_t len);
3608
4085ebe8
VY
3609int dev_get_nest_level(struct net_device *dev,
3610 bool (*type_check)(struct net_device *dev));
f629d208
JP
3611int skb_checksum_help(struct sk_buff *skb);
3612struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3613 netdev_features_t features, bool tx_path);
3614struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3615 netdev_features_t features);
12b0004d 3616
61bd3857
MS
3617struct netdev_bonding_info {
3618 ifslave slave;
3619 ifbond master;
3620};
3621
3622struct netdev_notifier_bonding_info {
3623 struct netdev_notifier_info info; /* must be first */
3624 struct netdev_bonding_info bonding_info;
3625};
3626
3627void netdev_bonding_info_change(struct net_device *dev,
3628 struct netdev_bonding_info *bonding_info);
3629
12b0004d
CW
3630static inline
3631struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3632{
3633 return __skb_gso_segment(skb, features, true);
3634}
53d6471c 3635__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3636
3637static inline bool can_checksum_protocol(netdev_features_t features,
3638 __be16 protocol)
3639{
3640 return ((features & NETIF_F_GEN_CSUM) ||
3641 ((features & NETIF_F_V4_CSUM) &&
3642 protocol == htons(ETH_P_IP)) ||
3643 ((features & NETIF_F_V6_CSUM) &&
3644 protocol == htons(ETH_P_IPV6)) ||
3645 ((features & NETIF_F_FCOE_CRC) &&
3646 protocol == htons(ETH_P_FCOE)));
3647}
12b0004d 3648
fb286bb2 3649#ifdef CONFIG_BUG
f629d208 3650void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3651#else
3652static inline void netdev_rx_csum_fault(struct net_device *dev)
3653{
3654}
3655#endif
1da177e4 3656/* rx skb timestamps */
f629d208
JP
3657void net_enable_timestamp(void);
3658void net_disable_timestamp(void);
1da177e4 3659
20380731 3660#ifdef CONFIG_PROC_FS
f629d208 3661int __init dev_proc_init(void);
900ff8c6
CW
3662#else
3663#define dev_proc_init() 0
20380731
ACM
3664#endif
3665
4798248e 3666static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3667 struct sk_buff *skb, struct net_device *dev,
3668 bool more)
4798248e 3669{
fa2dbdc2 3670 skb->xmit_more = more ? 1 : 0;
0b725a2c 3671 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3672}
3673
10b3ad8c 3674static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3675 struct netdev_queue *txq, bool more)
4798248e
DM
3676{
3677 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3678 int rc;
4798248e 3679
fa2dbdc2 3680 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3681 if (rc == NETDEV_TX_OK)
3682 txq_trans_update(txq);
3683
3684 return rc;
4798248e
DM
3685}
3686
42a2d923
LT
3687int netdev_class_create_file_ns(struct class_attribute *class_attr,
3688 const void *ns);
3689void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3690 const void *ns);
58292cbe
TH
3691
3692static inline int netdev_class_create_file(struct class_attribute *class_attr)
3693{
3694 return netdev_class_create_file_ns(class_attr, NULL);
3695}
3696
3697static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3698{
3699 netdev_class_remove_file_ns(class_attr, NULL);
3700}
b8a9787e 3701
04600794
JB
3702extern struct kobj_ns_type_operations net_ns_type_operations;
3703
f629d208 3704const char *netdev_drivername(const struct net_device *dev);
6579e57b 3705
f629d208 3706void linkwatch_run_queue(void);
20380731 3707
da08143b
MK
3708static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3709 netdev_features_t f2)
3710{
3711 if (f1 & NETIF_F_GEN_CSUM)
3712 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3713 if (f2 & NETIF_F_GEN_CSUM)
3714 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3715 f1 &= f2;
3716 if (f1 & NETIF_F_GEN_CSUM)
3717 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3718
3719 return f1;
3720}
3721
c8f44aff
MM
3722static inline netdev_features_t netdev_get_wanted_features(
3723 struct net_device *dev)
5455c699
MM
3724{
3725 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3726}
c8f44aff
MM
3727netdev_features_t netdev_increment_features(netdev_features_t all,
3728 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3729
3730/* Allow TSO being used on stacked device :
3731 * Performing the GSO segmentation before last device
3732 * is a performance improvement.
3733 */
3734static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3735 netdev_features_t mask)
3736{
3737 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3738}
3739
6cb6a27c 3740int __netdev_update_features(struct net_device *dev);
5455c699 3741void netdev_update_features(struct net_device *dev);
afe12cc8 3742void netdev_change_features(struct net_device *dev);
7f353bf2 3743
fc4a7489
PM
3744void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3745 struct net_device *dev);
3746
e38f3025
TM
3747netdev_features_t passthru_features_check(struct sk_buff *skb,
3748 struct net_device *dev,
3749 netdev_features_t features);
c1e756bf 3750netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3751
4d29515f 3752static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3753{
c8f44aff 3754 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3755
3756 /* check flags correspondence */
3757 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3758 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3759 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3760 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3761 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3762 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
3763 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3764 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3765 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3766 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3767 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3768 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
e585f236 3769 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
0345e186 3770
d6b4991a 3771 return (features & feature) == feature;
576a30eb
HX
3772}
3773
4d29515f 3774static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3775{
278b2513 3776 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3777 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3778}
3779
8b86a61d 3780static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 3781 netdev_features_t features)
7967168c 3782{
fc741216 3783 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3784 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3785 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3786}
3787
82cc1a7a
PWJ
3788static inline void netif_set_gso_max_size(struct net_device *dev,
3789 unsigned int size)
3790{
3791 dev->gso_max_size = size;
3792}
3793
7a7ffbab
WCC
3794static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3795 int pulled_hlen, u16 mac_offset,
3796 int mac_len)
3797{
3798 skb->protocol = protocol;
3799 skb->encapsulation = 1;
3800 skb_push(skb, pulled_hlen);
3801 skb_reset_transport_header(skb);
3802 skb->mac_header = mac_offset;
3803 skb->network_header = skb->mac_header + mac_len;
3804 skb->mac_len = mac_len;
3805}
3806
a6cc0cfa
JF
3807static inline bool netif_is_macvlan(struct net_device *dev)
3808{
3809 return dev->priv_flags & IFF_MACVLAN;
3810}
3811
2f33e7d5
MB
3812static inline bool netif_is_macvlan_port(struct net_device *dev)
3813{
3814 return dev->priv_flags & IFF_MACVLAN_PORT;
3815}
3816
5933fea7
MB
3817static inline bool netif_is_ipvlan(struct net_device *dev)
3818{
3819 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3820}
3821
3822static inline bool netif_is_ipvlan_port(struct net_device *dev)
3823{
3824 return dev->priv_flags & IFF_IPVLAN_MASTER;
3825}
3826
8a7fbfab 3827static inline bool netif_is_bond_master(struct net_device *dev)
3828{
3829 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3830}
3831
4d29515f 3832static inline bool netif_is_bond_slave(struct net_device *dev)
1765a575
JP
3833{
3834 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3835}
3836
3bdc0eba
BG
3837static inline bool netif_supports_nofcs(struct net_device *dev)
3838{
3839 return dev->priv_flags & IFF_SUPP_NOFCS;
3840}
3841
4e3c8992
DA
3842static inline bool netif_is_vrf(const struct net_device *dev)
3843{
3844 return dev->priv_flags & IFF_VRF_MASTER;
3845}
3846
0894ae3f
JP
3847static inline bool netif_is_bridge_master(const struct net_device *dev)
3848{
3849 return dev->priv_flags & IFF_EBRIDGE;
3850}
3851
35d4e172
JP
3852static inline bool netif_is_ovs_master(const struct net_device *dev)
3853{
3854 return dev->priv_flags & IFF_OPENVSWITCH;
3855}
3856
4e3c8992
DA
3857static inline bool netif_index_is_vrf(struct net *net, int ifindex)
3858{
4e3c8992
DA
3859 bool rc = false;
3860
2f52bdcf
DA
3861#if IS_ENABLED(CONFIG_NET_VRF)
3862 struct net_device *dev;
3863
3864 if (ifindex == 0)
3865 return false;
3866
3867 rcu_read_lock();
3868
3869 dev = dev_get_by_index_rcu(net, ifindex);
4e3c8992
DA
3870 if (dev)
3871 rc = netif_is_vrf(dev);
3872
2f52bdcf
DA
3873 rcu_read_unlock();
3874#endif
4e3c8992
DA
3875 return rc;
3876}
3877
02875878
ED
3878/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3879static inline void netif_keep_dst(struct net_device *dev)
3880{
3881 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3882}
3883
505d4f73 3884extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 3885
571ba423
JP
3886/* Logging, debugging and troubleshooting/diagnostic helpers. */
3887
3888/* netdev_printk helpers, similar to dev_printk */
3889
3890static inline const char *netdev_name(const struct net_device *dev)
3891{
c6f854d5
VF
3892 if (!dev->name[0] || strchr(dev->name, '%'))
3893 return "(unnamed net_device)";
571ba423
JP
3894 return dev->name;
3895}
3896
ccc7f496
VF
3897static inline const char *netdev_reg_state(const struct net_device *dev)
3898{
3899 switch (dev->reg_state) {
3900 case NETREG_UNINITIALIZED: return " (uninitialized)";
3901 case NETREG_REGISTERED: return "";
3902 case NETREG_UNREGISTERING: return " (unregistering)";
3903 case NETREG_UNREGISTERED: return " (unregistered)";
3904 case NETREG_RELEASED: return " (released)";
3905 case NETREG_DUMMY: return " (dummy)";
3906 }
3907
3908 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3909 return " (unknown)";
3910}
3911
f629d208 3912__printf(3, 4)
6ea754eb
JP
3913void netdev_printk(const char *level, const struct net_device *dev,
3914 const char *format, ...);
f629d208 3915__printf(2, 3)
6ea754eb 3916void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 3917__printf(2, 3)
6ea754eb 3918void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 3919__printf(2, 3)
6ea754eb 3920void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 3921__printf(2, 3)
6ea754eb 3922void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 3923__printf(2, 3)
6ea754eb 3924void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 3925__printf(2, 3)
6ea754eb 3926void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 3927__printf(2, 3)
6ea754eb 3928void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 3929
8909c9ad
VK
3930#define MODULE_ALIAS_NETDEV(device) \
3931 MODULE_ALIAS("netdev-" device)
3932
b558c96f 3933#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
3934#define netdev_dbg(__dev, format, args...) \
3935do { \
ffa10cb4 3936 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 3937} while (0)
b558c96f
JC
3938#elif defined(DEBUG)
3939#define netdev_dbg(__dev, format, args...) \
3940 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
3941#else
3942#define netdev_dbg(__dev, format, args...) \
3943({ \
3944 if (0) \
3945 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
3946})
3947#endif
3948
3949#if defined(VERBOSE_DEBUG)
3950#define netdev_vdbg netdev_dbg
3951#else
3952
3953#define netdev_vdbg(dev, format, args...) \
3954({ \
3955 if (0) \
3956 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3957 0; \
3958})
3959#endif
3960
3961/*
3962 * netdev_WARN() acts like dev_printk(), but with the key difference
3963 * of using a WARN/WARN_ON to get the message out, including the
3964 * file/line information and a backtrace.
3965 */
3966#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
3967 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3968 netdev_reg_state(dev), ##args)
571ba423 3969
b3d95c5c
JP
3970/* netif printk helpers, similar to netdev_printk */
3971
3972#define netif_printk(priv, type, level, dev, fmt, args...) \
3973do { \
3974 if (netif_msg_##type(priv)) \
3975 netdev_printk(level, (dev), fmt, ##args); \
3976} while (0)
3977
f45f4321
JP
3978#define netif_level(level, priv, type, dev, fmt, args...) \
3979do { \
3980 if (netif_msg_##type(priv)) \
3981 netdev_##level(dev, fmt, ##args); \
3982} while (0)
3983
b3d95c5c 3984#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 3985 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 3986#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 3987 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 3988#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 3989 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 3990#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 3991 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 3992#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 3993 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 3994#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 3995 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 3996#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 3997 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 3998
0053ea9c 3999#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4000#define netif_dbg(priv, type, netdev, format, args...) \
4001do { \
4002 if (netif_msg_##type(priv)) \
b5fb0a03 4003 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4004} while (0)
0053ea9c
JP
4005#elif defined(DEBUG)
4006#define netif_dbg(priv, type, dev, format, args...) \
4007 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4008#else
4009#define netif_dbg(priv, type, dev, format, args...) \
4010({ \
4011 if (0) \
4012 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4013 0; \
4014})
4015#endif
4016
4017#if defined(VERBOSE_DEBUG)
bcfcc450 4018#define netif_vdbg netif_dbg
b3d95c5c
JP
4019#else
4020#define netif_vdbg(priv, type, dev, format, args...) \
4021({ \
4022 if (0) \
a4ed89cb 4023 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4024 0; \
4025})
4026#endif
571ba423 4027
900ff8c6
CW
4028/*
4029 * The list of packet types we will receive (as opposed to discard)
4030 * and the routines to invoke.
4031 *
4032 * Why 16. Because with 16 the only overlap we get on a hash of the
4033 * low nibble of the protocol value is RARP/SNAP/X.25.
4034 *
4035 * NOTE: That is no longer true with the addition of VLAN tags. Not
4036 * sure which should go first, but I bet it won't make much
4037 * difference if we are running VLANs. The good news is that
4038 * this protocol won't be in the list unless compiled in, so
4039 * the average user (w/out VLANs) will not be adversely affected.
4040 * --BLG
4041 *
4042 * 0800 IP
4043 * 8100 802.1Q VLAN
4044 * 0001 802.3
4045 * 0002 AX.25
4046 * 0004 802.2
4047 * 8035 RARP
4048 * 0005 SNAP
4049 * 0805 X.25
4050 * 0806 ARP
4051 * 8137 IPX
4052 * 0009 Localtalk
4053 * 86DD IPv6
4054 */
4055#define PTYPE_HASH_SIZE (16)
4056#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4057
385a154c 4058#endif /* _LINUX_NETDEVICE_H */