macsec: remove first zero and add attribute name in comments
[linux-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>
e4c6734e 54#include <uapi/linux/pkt_cls.h>
59cc1f61 55#include <linux/hashtable.h>
a59e2ecb 56
115c1d6e 57struct netpoll_info;
313162d0 58struct device;
c1f19b51 59struct phy_device;
704232c2
JB
60/* 802.11 specific */
61struct wireless_dev;
98a18b6f
AA
62/* 802.15.4 specific */
63struct wpan_dev;
03c57747 64struct mpls_dev;
7c46a640
AD
65/* UDP Tunnel offloads */
66struct udp_tunnel_info;
a7862b45 67struct bpf_prog;
1da177e4 68
f629d208
JP
69void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
d07d7507 71
9a1654ba
JP
72/* Backlog congestion levels */
73#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74#define NET_RX_DROP 1 /* packet dropped */
75
572a9d7b
PM
76/*
77 * Transmit return codes: transmit return codes originate from three different
78 * namespaces:
79 *
80 * - qdisc return codes
81 * - driver transmit return codes
82 * - errno values
83 *
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
90 * others are propagated to higher layers.
91 */
92
93/* qdisc ->enqueue() return codes. */
94#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
95#define NET_XMIT_DROP 0x01 /* skb dropped */
96#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 97#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 98
b9df3cb8
GR
99/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 102#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
103#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
104
dc1f8bf6 105/* Driver transmit return codes */
9a1654ba 106#define NETDEV_TX_MASK 0xf0
572a9d7b 107
dc1f8bf6 108enum netdev_tx {
572a9d7b 109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
112};
113typedef enum netdev_tx netdev_tx_t;
114
9a1654ba
JP
115/*
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
118 */
119static inline bool dev_xmit_complete(int rc)
120{
121 /*
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
126 */
127 if (likely(rc < NET_XMIT_MASK))
128 return true;
129
130 return false;
131}
132
1da177e4 133/*
5e82b4b2 134 * Compute the worst-case header length according to the protocols
1da177e4
LT
135 * used.
136 */
fe2918b0 137
c0eb4540
KS
138#if defined(CONFIG_HYPERV_NET)
139# define LL_MAX_HEADER 128
140#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
141# if defined(CONFIG_MAC80211_MESH)
142# define LL_MAX_HEADER 128
143# else
144# define LL_MAX_HEADER 96
145# endif
1da177e4 146#else
8388e3da 147# define LL_MAX_HEADER 32
1da177e4
LT
148#endif
149
d11ead75
BH
150#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
152#define MAX_HEADER LL_MAX_HEADER
153#else
154#define MAX_HEADER (LL_MAX_HEADER + 48)
155#endif
156
157/*
be1f3c2c
BH
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
1da177e4 160 */
fe2918b0 161
d94d9fee 162struct net_device_stats {
3cfde79c
BH
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
1da177e4 172 unsigned long collisions;
1da177e4 173 unsigned long rx_length_errors;
3cfde79c
BH
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
1da177e4
LT
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
1da177e4
LT
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
186};
187
1da177e4
LT
188
189#include <linux/cache.h>
190#include <linux/skbuff.h>
191
adc9300e 192#ifdef CONFIG_RPS
c5905afb
IM
193#include <linux/static_key.h>
194extern struct static_key rps_needed;
adc9300e
ED
195#endif
196
1da177e4
LT
197struct neighbour;
198struct neigh_parms;
199struct sk_buff;
200
f001fde5
JP
201struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
ccffad25
JP
205#define NETDEV_HW_ADDR_T_LAN 1
206#define NETDEV_HW_ADDR_T_SAN 2
207#define NETDEV_HW_ADDR_T_SLAVE 3
208#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 209#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 210 bool global_use;
4cd729b0 211 int sync_cnt;
8f8f103d 212 int refcount;
4543fbef 213 int synced;
f001fde5
JP
214 struct rcu_head rcu_head;
215};
216
31278e71
JP
217struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220};
221
22bedad3
JP
222#define netdev_hw_addr_list_count(l) ((l)->count)
223#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224#define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 226
22bedad3
JP
227#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229#define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 231
22bedad3
JP
232#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 234#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 236
d94d9fee 237struct hh_cache {
f6b72b62 238 u16 hh_len;
5c25f686 239 u16 __pad;
3644f0ce 240 seqlock_t hh_lock;
1da177e4
LT
241
242 /* cached hardware header; allow for machine alignment needs. */
243#define HH_DATA_MOD 16
244#define HH_DATA_OFF(__len) \
5ba0eac6 245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
246#define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
249};
250
5e82b4b2 251/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
252 * Alternative is:
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
255 *
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
258 */
259#define LL_RESERVED_SPACE(dev) \
f5184d26 260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 261#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 263
3b04ddde
SH
264struct header_ops {
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
95c96174 267 const void *saddr, unsigned int len);
3b04ddde 268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
270 void (*cache_update)(struct hh_cache *hh,
271 const struct net_device *dev,
272 const unsigned char *haddr);
2793a23a 273 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
274};
275
1da177e4 276/* These flag bits are private to the generic network queueing
5e82b4b2 277 * layer; they may not be explicitly referenced by any other
1da177e4
LT
278 * code.
279 */
280
d94d9fee 281enum netdev_state_t {
1da177e4
LT
282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
1da177e4 284 __LINK_STATE_NOCARRIER,
b00055aa
SR
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
1da177e4
LT
287};
288
289
290/*
5e82b4b2 291 * This structure holds boot-time configured netdevice settings. They
fe2918b0 292 * are then used in the device probing.
1da177e4
LT
293 */
294struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297};
298#define NETDEV_BOOT_SETUP_MAX 8
299
f629d208 300int __init netdev_boot_setup(char *str);
1da177e4 301
bea3348e
SH
302/*
303 * Structure for NAPI scheduling similar to tasklet but with weighting
304 */
305struct napi_struct {
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 309 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
310 * can remove from the list right before clearing the bit.
311 */
312 struct list_head poll_list;
313
314 unsigned long state;
315 int weight;
404f7c9e 316 unsigned int gro_count;
bea3348e
SH
317 int (*poll)(struct napi_struct *, int);
318#ifdef CONFIG_NETPOLL
bea3348e 319 int poll_owner;
bea3348e 320#endif
5d38a079 321 struct net_device *dev;
d565b0a1 322 struct sk_buff *gro_list;
5d38a079 323 struct sk_buff *skb;
3b47d303 324 struct hrtimer timer;
404f7c9e 325 struct list_head dev_list;
af12fa6e
ET
326 struct hlist_node napi_hash_node;
327 unsigned int napi_id;
bea3348e
SH
328};
329
d94d9fee 330enum {
bea3348e 331 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 332 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 333 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
334 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
335 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
336 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
337};
338
339enum {
340 NAPIF_STATE_SCHED = (1UL << NAPI_STATE_SCHED),
341 NAPIF_STATE_DISABLE = (1UL << NAPI_STATE_DISABLE),
342 NAPIF_STATE_NPSVC = (1UL << NAPI_STATE_NPSVC),
343 NAPIF_STATE_HASHED = (1UL << NAPI_STATE_HASHED),
344 NAPIF_STATE_NO_BUSY_POLL = (1UL << NAPI_STATE_NO_BUSY_POLL),
345 NAPIF_STATE_IN_BUSY_POLL = (1UL << NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
346};
347
5b252f0c 348enum gro_result {
d1c76af9
HX
349 GRO_MERGED,
350 GRO_MERGED_FREE,
351 GRO_HELD,
352 GRO_NORMAL,
353 GRO_DROP,
354};
5b252f0c 355typedef enum gro_result gro_result_t;
d1c76af9 356
8a4eb573
JP
357/*
358 * enum rx_handler_result - Possible return values for rx_handlers.
359 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
360 * further.
361 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
362 * case skb->dev was changed by rx_handler.
363 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 364 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
365 *
366 * rx_handlers are functions called from inside __netif_receive_skb(), to do
367 * special processing of the skb, prior to delivery to protocol handlers.
368 *
369 * Currently, a net_device can only have a single rx_handler registered. Trying
370 * to register a second rx_handler will return -EBUSY.
371 *
372 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
373 * To unregister a rx_handler on a net_device, use
374 * netdev_rx_handler_unregister().
375 *
376 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
377 * do with the skb.
378 *
5e82b4b2 379 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 380 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 381 * the skb to be delivered in some other way.
8a4eb573
JP
382 *
383 * If the rx_handler changed skb->dev, to divert the skb to another
384 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
385 * new device will be called if it exists.
386 *
5e82b4b2 387 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 388 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 389 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 390 *
5e82b4b2 391 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
392 * delivered, it should return RX_HANDLER_PASS.
393 *
394 * A device without a registered rx_handler will behave as if rx_handler
395 * returned RX_HANDLER_PASS.
396 */
397
398enum rx_handler_result {
399 RX_HANDLER_CONSUMED,
400 RX_HANDLER_ANOTHER,
401 RX_HANDLER_EXACT,
402 RX_HANDLER_PASS,
403};
404typedef enum rx_handler_result rx_handler_result_t;
405typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 406
f629d208 407void __napi_schedule(struct napi_struct *n);
bc9ad166 408void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 409
4d29515f 410static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
411{
412 return test_bit(NAPI_STATE_DISABLE, &n->state);
413}
414
bea3348e 415/**
5e82b4b2
BH
416 * napi_schedule_prep - check if NAPI can be scheduled
417 * @n: NAPI context
bea3348e
SH
418 *
419 * Test if NAPI routine is already running, and if not mark
5e82b4b2 420 * it as running. This is used as a condition variable to
a0a46196
DM
421 * insure only one NAPI poll instance runs. We also make
422 * sure there is no pending NAPI disable.
bea3348e 423 */
4d29515f 424static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 425{
a0a46196
DM
426 return !napi_disable_pending(n) &&
427 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
428}
429
430/**
431 * napi_schedule - schedule NAPI poll
5e82b4b2 432 * @n: NAPI context
bea3348e
SH
433 *
434 * Schedule NAPI poll routine to be called if it is not already
435 * running.
436 */
437static inline void napi_schedule(struct napi_struct *n)
438{
439 if (napi_schedule_prep(n))
440 __napi_schedule(n);
441}
442
bc9ad166
ED
443/**
444 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 445 * @n: NAPI context
bc9ad166
ED
446 *
447 * Variant of napi_schedule(), assuming hard irqs are masked.
448 */
449static inline void napi_schedule_irqoff(struct napi_struct *n)
450{
451 if (napi_schedule_prep(n))
452 __napi_schedule_irqoff(n);
453}
454
bfe13f54 455/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 456static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
457{
458 if (napi_schedule_prep(napi)) {
459 __napi_schedule(napi);
4d29515f 460 return true;
bfe13f54 461 }
4d29515f 462 return false;
bfe13f54
RD
463}
464
364b6055
ED
465bool __napi_complete(struct napi_struct *n);
466bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
467/**
468 * napi_complete - NAPI processing complete
5e82b4b2 469 * @n: NAPI context
bea3348e
SH
470 *
471 * Mark NAPI processing as complete.
3b47d303 472 * Consider using napi_complete_done() instead.
364b6055 473 * Return false if device should avoid rearming interrupts.
bea3348e 474 */
364b6055 475static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
476{
477 return napi_complete_done(n, 0);
478}
bea3348e 479
af12fa6e
ET
480/**
481 * napi_hash_del - remove a NAPI from global table
5e82b4b2 482 * @napi: NAPI context
af12fa6e 483 *
5e82b4b2 484 * Warning: caller must observe RCU grace period
34cbe27e
ED
485 * before freeing memory containing @napi, if
486 * this function returns true.
93d05d4a 487 * Note: core networking stack automatically calls it
5e82b4b2 488 * from netif_napi_del().
93d05d4a 489 * Drivers might want to call this helper to combine all
5e82b4b2 490 * the needed RCU grace periods into a single one.
af12fa6e 491 */
34cbe27e 492bool napi_hash_del(struct napi_struct *napi);
af12fa6e 493
bea3348e
SH
494/**
495 * napi_disable - prevent NAPI from scheduling
5e82b4b2 496 * @n: NAPI context
bea3348e
SH
497 *
498 * Stop NAPI from being scheduled on this context.
499 * Waits till any outstanding processing completes.
500 */
3b47d303 501void napi_disable(struct napi_struct *n);
bea3348e
SH
502
503/**
504 * napi_enable - enable NAPI scheduling
5e82b4b2 505 * @n: NAPI context
bea3348e
SH
506 *
507 * Resume NAPI from being scheduled on this context.
508 * Must be paired with napi_disable.
509 */
510static inline void napi_enable(struct napi_struct *n)
511{
512 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 513 smp_mb__before_atomic();
bea3348e 514 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 515 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
516}
517
c264c3de
SH
518/**
519 * napi_synchronize - wait until NAPI is not running
5e82b4b2 520 * @n: NAPI context
c264c3de
SH
521 *
522 * Wait until NAPI is done being scheduled on this context.
523 * Waits till any outstanding processing completes but
524 * does not disable future activations.
525 */
526static inline void napi_synchronize(const struct napi_struct *n)
527{
facc432f
AB
528 if (IS_ENABLED(CONFIG_SMP))
529 while (test_bit(NAPI_STATE_SCHED, &n->state))
530 msleep(1);
531 else
532 barrier();
c264c3de 533}
c264c3de 534
d94d9fee 535enum netdev_queue_state_t {
73466498
TH
536 __QUEUE_STATE_DRV_XOFF,
537 __QUEUE_STATE_STACK_XOFF,
c3f26a26 538 __QUEUE_STATE_FROZEN,
79d16385 539};
8e2f1a63
DB
540
541#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
542#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
543#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
544
545#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
546#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
547 QUEUE_STATE_FROZEN)
548#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
549 QUEUE_STATE_FROZEN)
550
73466498
TH
551/*
552 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
553 * netif_tx_* functions below are used to manipulate this flag. The
554 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
555 * queue independently. The netif_xmit_*stopped functions below are called
556 * to check if the queue has been stopped by the driver or stack (either
557 * of the XOFF bits are set in the state). Drivers should not need to call
558 * netif_xmit*stopped functions, they should only be using netif_tx_*.
559 */
79d16385 560
bb949fbd 561struct netdev_queue {
6a321cb3 562/*
5e82b4b2 563 * read-mostly part
6a321cb3 564 */
bb949fbd 565 struct net_device *dev;
46e5da40 566 struct Qdisc __rcu *qdisc;
b0e1e646 567 struct Qdisc *qdisc_sleeping;
ccf5ff69 568#ifdef CONFIG_SYSFS
1d24eb48
TH
569 struct kobject kobj;
570#endif
f2cd2d3e
ED
571#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
572 int numa_node;
573#endif
c0ef079c
FW
574 unsigned long tx_maxrate;
575 /*
576 * Number of TX timeouts for this queue
577 * (/sys/class/net/DEV/Q/trans_timeout)
578 */
579 unsigned long trans_timeout;
6a321cb3 580/*
5e82b4b2 581 * write-mostly part
6a321cb3
ED
582 */
583 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
584 int xmit_lock_owner;
9d21493b 585 /*
9b36627a 586 * Time (in jiffies) of last Tx
9d21493b
ED
587 */
588 unsigned long trans_start;
ccf5ff69 589
114cf580
TH
590 unsigned long state;
591
592#ifdef CONFIG_BQL
593 struct dql dql;
594#endif
e8a0464c 595} ____cacheline_aligned_in_smp;
bb949fbd 596
f2cd2d3e
ED
597static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 return q->numa_node;
601#else
b236da69 602 return NUMA_NO_NODE;
f2cd2d3e
ED
603#endif
604}
605
606static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
607{
608#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
609 q->numa_node = node;
610#endif
611}
612
df334545 613#ifdef CONFIG_RPS
0a9627f2
TH
614/*
615 * This structure holds an RPS map which can be of variable length. The
616 * map is an array of CPUs.
617 */
618struct rps_map {
619 unsigned int len;
620 struct rcu_head rcu;
621 u16 cpus[0];
622};
60b778ce 623#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 624
fec5e652 625/*
c445477d
BH
626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627 * tail pointer for that CPU's input queue at the time of last enqueue, and
628 * a hardware filter index.
fec5e652
TH
629 */
630struct rps_dev_flow {
631 u16 cpu;
c445477d 632 u16 filter;
fec5e652
TH
633 unsigned int last_qtail;
634};
c445477d 635#define RPS_NO_FILTER 0xffff
fec5e652
TH
636
637/*
638 * The rps_dev_flow_table structure contains a table of flow mappings.
639 */
640struct rps_dev_flow_table {
641 unsigned int mask;
642 struct rcu_head rcu;
fec5e652
TH
643 struct rps_dev_flow flows[0];
644};
645#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 646 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
647
648/*
649 * The rps_sock_flow_table contains mappings of flows to the last CPU
650 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
651 * Each entry is a 32bit value. Upper part is the high-order bits
652 * of flow hash, lower part is CPU number.
567e4b79 653 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
654 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 656 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
657 */
658struct rps_sock_flow_table {
567e4b79 659 u32 mask;
93c1af6c
ED
660
661 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 662};
567e4b79 663#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
664
665#define RPS_NO_CPU 0xffff
666
567e4b79
ED
667extern u32 rps_cpu_mask;
668extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
669
fec5e652
TH
670static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
671 u32 hash)
672{
673 if (table && hash) {
567e4b79
ED
674 unsigned int index = hash & table->mask;
675 u32 val = hash & ~rps_cpu_mask;
fec5e652 676
5e82b4b2 677 /* We only give a hint, preemption can change CPU under us */
567e4b79 678 val |= raw_smp_processor_id();
fec5e652 679
567e4b79
ED
680 if (table->ents[index] != val)
681 table->ents[index] = val;
fec5e652
TH
682 }
683}
684
c445477d 685#ifdef CONFIG_RFS_ACCEL
f629d208
JP
686bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
687 u16 filter_id);
c445477d 688#endif
a953be53 689#endif /* CONFIG_RPS */
c445477d 690
0a9627f2
TH
691/* This structure contains an instance of an RX queue. */
692struct netdev_rx_queue {
a953be53 693#ifdef CONFIG_RPS
6e3f7faf
ED
694 struct rps_map __rcu *rps_map;
695 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 696#endif
6e3f7faf 697 struct kobject kobj;
fe822240 698 struct net_device *dev;
0a9627f2 699} ____cacheline_aligned_in_smp;
a953be53
MD
700
701/*
702 * RX queue sysfs structures and functions.
703 */
704struct rx_queue_attribute {
705 struct attribute attr;
706 ssize_t (*show)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, char *buf);
708 ssize_t (*store)(struct netdev_rx_queue *queue,
709 struct rx_queue_attribute *attr, const char *buf, size_t len);
710};
d314774c 711
bf264145
TH
712#ifdef CONFIG_XPS
713/*
714 * This structure holds an XPS map which can be of variable length. The
715 * map is an array of queues.
716 */
717struct xps_map {
718 unsigned int len;
719 unsigned int alloc_len;
720 struct rcu_head rcu;
721 u16 queues[0];
722};
60b778ce 723#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
724#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
726
727/*
728 * This structure holds all XPS maps for device. Maps are indexed by CPU.
729 */
730struct xps_dev_maps {
731 struct rcu_head rcu;
a4177869 732 struct xps_map __rcu *cpu_map[0];
bf264145 733};
184c449f
AD
734#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
735 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
736#endif /* CONFIG_XPS */
737
4f57c087
JF
738#define TC_MAX_QUEUE 16
739#define TC_BITMASK 15
740/* HW offloaded queuing disciplines txq count and offset maps */
741struct netdev_tc_txq {
742 u16 count;
743 u16 offset;
744};
745
68bad94e
NP
746#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
747/*
748 * This structure is to hold information about the device
749 * configured to run FCoE protocol stack.
750 */
751struct netdev_fcoe_hbainfo {
752 char manufacturer[64];
753 char serial_number[64];
754 char hardware_version[64];
755 char driver_version[64];
756 char optionrom_version[64];
757 char firmware_version[64];
758 char model[256];
759 char model_description[256];
760};
761#endif
762
02637fce 763#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 764
02637fce
JP
765/* This structure holds a unique identifier to identify some
766 * physical item (port for example) used by a netdevice.
66b52b0d 767 */
02637fce
JP
768struct netdev_phys_item_id {
769 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
770 unsigned char id_len;
771};
772
d754f98b
SF
773static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
774 struct netdev_phys_item_id *b)
775{
776 return a->id_len == b->id_len &&
777 memcmp(a->id, b->id, a->id_len) == 0;
778}
779
99932d4f
DB
780typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
781 struct sk_buff *skb);
782
a1b7c5fd 783/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
784 * that are being passed to the netdevice through the setup_tc op.
785 */
786enum {
787 TC_SETUP_MQPRIO,
a1b7c5fd 788 TC_SETUP_CLSU32,
5b33f488 789 TC_SETUP_CLSFLOWER,
b87f7936 790 TC_SETUP_MATCHALL,
332ae8e2 791 TC_SETUP_CLSBPF,
16e5cc64
JF
792};
793
a1b7c5fd
JF
794struct tc_cls_u32_offload;
795
16e5cc64
JF
796struct tc_to_netdev {
797 unsigned int type;
798 union {
799 u8 tc;
a1b7c5fd 800 struct tc_cls_u32_offload *cls_u32;
5b33f488 801 struct tc_cls_flower_offload *cls_flower;
b87f7936 802 struct tc_cls_matchall_offload *cls_mall;
332ae8e2 803 struct tc_cls_bpf_offload *cls_bpf;
16e5cc64 804 };
7091d8c7 805 bool egress_dev;
16e5cc64
JF
806};
807
a7862b45
BB
808/* These structures hold the attributes of xdp state that are being passed
809 * to the netdevice through the xdp op.
810 */
811enum xdp_netdev_command {
812 /* Set or clear a bpf program used in the earliest stages of packet
813 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
814 * is responsible for calling bpf_prog_put on any old progs that are
815 * stored. In case of error, the callee need not release the new prog
816 * reference, but on success it takes ownership and must bpf_prog_put
817 * when it is no longer used.
818 */
819 XDP_SETUP_PROG,
820 /* Check if a bpf program is set on the device. The callee should
821 * return true if a program is currently attached and running.
822 */
823 XDP_QUERY_PROG,
824};
825
826struct netdev_xdp {
827 enum xdp_netdev_command command;
828 union {
829 /* XDP_SETUP_PROG */
830 struct bpf_prog *prog;
831 /* XDP_QUERY_PROG */
832 bool prog_attached;
833 };
834};
16e5cc64 835
d314774c
SH
836/*
837 * This structure defines the management hooks for network devices.
00829823
SH
838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
d314774c
SH
840 *
841 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
d314774c
SH
846 *
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
850 *
851 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 852 * This function is called when a network device transitions to the up
d314774c
SH
853 * state.
854 *
855 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 856 * This function is called when a network device transitions to the down
d314774c
SH
857 * state.
858 *
dc1f8bf6
SH
859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
00829823 861 * Called when a packet needs to be transmitted.
e79d8429
RR
862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 866 * Required; cannot be NULL.
00829823 867 *
cdba756f
ED
868 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
869 * netdev_features_t features);
870 * Adjusts the requested feature flags according to device-specific
871 * constraints, and returns the resulting flags. Must not modify
872 * the device state.
873 *
f663dd9a 874 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 875 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 876 * Called to decide which queue to use when device supports multiple
00829823
SH
877 * transmit queues.
878 *
d314774c
SH
879 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
880 * This function is called to allow device receiver to make
5e82b4b2 881 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
882 *
883 * void (*ndo_set_rx_mode)(struct net_device *dev);
884 * This function is called device changes address list filtering.
01789349 885 * If driver handles unicast address filtering, it should set
5e82b4b2 886 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
887 *
888 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
889 * This function is called when the Media Access Control address
37b607c5 890 * needs to be changed. If this interface is not defined, the
5e82b4b2 891 * MAC address can not be changed.
d314774c
SH
892 *
893 * int (*ndo_validate_addr)(struct net_device *dev);
894 * Test if Media Access Control address is valid for the device.
895 *
896 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
897 * Called when a user requests an ioctl which can't be handled by
898 * the generic interface code. If not defined ioctls return
d314774c
SH
899 * not supported error code.
900 *
901 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
902 * Used to set network devices bus interface parameters. This interface
5e82b4b2 903 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
904 * interface (PCI) for low level management.
905 *
906 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
907 * Called when a user wants to change the Maximum Transfer Unit
908 * of a device. If not defined, any request to change MTU will
909 * will return an error.
910 *
00829823 911 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 912 * Callback used when the transmitter has not made any progress
d314774c
SH
913 * for dev->watchdog ticks.
914 *
3cfde79c 915 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 916 * struct rtnl_link_stats64 *storage);
d308e38f 917 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 918 * Called when a user wants to get the network device usage
be1f3c2c 919 * statistics. Drivers must do one of the following:
3cfde79c
BH
920 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
921 * rtnl_link_stats64 structure passed by the caller.
82695d9b 922 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
923 * (which should normally be dev->stats) and return a pointer to
924 * it. The structure may be changed asynchronously only if each
925 * field is written atomically.
926 * 3. Update dev->stats asynchronously and atomically, and define
927 * neither operation.
d314774c 928 *
3df5b3c6 929 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
930 * Return true if this device supports offload stats of this attr_id.
931 *
932 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
933 * void *attr_data)
934 * Get statistics for offload operations by attr_id. Write it into the
935 * attr_data pointer.
936 *
5d632cb7 937 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 938 * If device supports VLAN filtering this function is called when a
80d5c368 939 * VLAN id is registered.
d314774c 940 *
5d632cb7 941 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 942 * If device supports VLAN filtering this function is called when a
80d5c368 943 * VLAN id is unregistered.
d314774c
SH
944 *
945 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
946 *
947 * SR-IOV management functions.
948 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
949 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
950 * u8 qos, __be16 proto);
ed616689
SC
951 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
952 * int max_tx_rate);
5f8444a3 953 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 954 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
955 * int (*ndo_get_vf_config)(struct net_device *dev,
956 * int vf, struct ifla_vf_info *ivf);
1d8faf48 957 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
958 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
959 * struct nlattr *port[]);
01a3d796
VZ
960 *
961 * Enable or disable the VF ability to query its RSS Redirection Table and
962 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 963 * with PF and querying it may introduce a theoretical security risk.
01a3d796 964 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 965 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
966 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
967 * Called to setup 'tc' number of traffic classes in the net device. This
968 * is always called from the stack with the rtnl lock held and netif tx
969 * queues stopped. This allows the netdevice to perform queue management
970 * safely.
c445477d 971 *
e9bce845
YZ
972 * Fiber Channel over Ethernet (FCoE) offload functions.
973 * int (*ndo_fcoe_enable)(struct net_device *dev);
974 * Called when the FCoE protocol stack wants to start using LLD for FCoE
975 * so the underlying device can perform whatever needed configuration or
976 * initialization to support acceleration of FCoE traffic.
977 *
978 * int (*ndo_fcoe_disable)(struct net_device *dev);
979 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
980 * so the underlying device can perform whatever needed clean-ups to
981 * stop supporting acceleration of FCoE traffic.
982 *
983 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
984 * struct scatterlist *sgl, unsigned int sgc);
985 * Called when the FCoE Initiator wants to initialize an I/O that
986 * is a possible candidate for Direct Data Placement (DDP). The LLD can
987 * perform necessary setup and returns 1 to indicate the device is set up
988 * successfully to perform DDP on this I/O, otherwise this returns 0.
989 *
990 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
991 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
992 * indicated by the FC exchange id 'xid', so the underlying device can
993 * clean up and reuse resources for later DDP requests.
994 *
995 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
996 * struct scatterlist *sgl, unsigned int sgc);
997 * Called when the FCoE Target wants to initialize an I/O that
998 * is a possible candidate for Direct Data Placement (DDP). The LLD can
999 * perform necessary setup and returns 1 to indicate the device is set up
1000 * successfully to perform DDP on this I/O, otherwise this returns 0.
1001 *
68bad94e
NP
1002 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1003 * struct netdev_fcoe_hbainfo *hbainfo);
1004 * Called when the FCoE Protocol stack wants information on the underlying
1005 * device. This information is utilized by the FCoE protocol stack to
1006 * register attributes with Fiber Channel management service as per the
1007 * FC-GS Fabric Device Management Information(FDMI) specification.
1008 *
e9bce845
YZ
1009 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1010 * Called when the underlying device wants to override default World Wide
1011 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1012 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1013 * protocol stack to use.
1014 *
c445477d
BH
1015 * RFS acceleration.
1016 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1017 * u16 rxq_index, u32 flow_id);
1018 * Set hardware filter for RFS. rxq_index is the target queue index;
1019 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1020 * Return the filter ID on success, or a negative error code.
fbaec0ea 1021 *
8b98a70c 1022 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1023 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1024 * Called to make another netdev an underling.
1025 *
1026 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1027 * Called to release previously enslaved netdev.
5455c699
MM
1028 *
1029 * Feature/offload setting functions.
c8f44aff 1030 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1031 * Called to update device configuration to new features. Passed
1032 * feature set might be less than what was returned by ndo_fix_features()).
1033 * Must return >0 or -errno if it changed dev->features itself.
1034 *
edc7d573 1035 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1036 * struct net_device *dev,
f6f6424b 1037 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1038 * Adds an FDB entry to dev for addr.
1690be63
VY
1039 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1040 * struct net_device *dev,
f6f6424b 1041 * const unsigned char *addr, u16 vid)
77162022
JF
1042 * Deletes the FDB entry from dev coresponding to addr.
1043 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1044 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1045 * int *idx)
77162022
JF
1046 * Used to add FDB entries to dump requests. Implementers should add
1047 * entries to skb and update idx with the number of entries.
e5a55a89 1048 *
ad41faa8
ND
1049 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1050 * u16 flags)
e5a55a89 1051 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1052 * struct net_device *dev, u32 filter_mask,
1053 * int nlflags)
ad41faa8
ND
1054 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1055 * u16 flags);
4bf84c35
JP
1056 *
1057 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1058 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1059 * which do not represent real hardware may define this to allow their
1060 * userspace components to manage their virtual carrier state. Devices
1061 * that determine carrier state from physical hardware properties (eg
1062 * network cables) or protocol-dependent mechanisms (eg
1063 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1064 *
1065 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1066 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1067 * Called to get ID of physical port of this device. If driver does
1068 * not implement this, it is assumed that the hw is not able to have
1069 * multiple net devices on single physical port.
53cf5275 1070 *
7c46a640
AD
1071 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1072 * struct udp_tunnel_info *ti);
1073 * Called by UDP tunnel to notify a driver about the UDP port and socket
1074 * address family that a UDP tunnel is listnening to. It is called only
1075 * when a new port starts listening. The operation is protected by the
1076 * RTNL.
1077 *
1078 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1079 * struct udp_tunnel_info *ti);
1080 * Called by UDP tunnel to notify the driver about a UDP port and socket
1081 * address family that the UDP tunnel is not listening to anymore. The
1082 * operation is protected by the RTNL.
1083 *
a6cc0cfa
JF
1084 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1085 * struct net_device *dev)
1086 * Called by upper layer devices to accelerate switching or other
1087 * station functionality into hardware. 'pdev is the lowerdev
1088 * to use for the offload and 'dev' is the net device that will
1089 * back the offload. Returns a pointer to the private structure
1090 * the upper layer will maintain.
1091 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1092 * Called by upper layer device to delete the station created
1093 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1094 * the station and priv is the structure returned by the add
1095 * operation.
1096 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1097 * struct net_device *dev,
1098 * void *priv);
1099 * Callback to use for xmit over the accelerated station. This
1100 * is used in place of ndo_start_xmit on accelerated net
1101 * devices.
5e82b4b2
BH
1102 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1103 * struct net_device *dev
1104 * netdev_features_t features);
04ffcb25 1105 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1106 * performing offload operations on a given packet. This is to give
1107 * the device an opportunity to implement any restrictions that cannot
1108 * be otherwise expressed by feature flags. The check is called with
1109 * the set of features that the stack has calculated and it returns
1110 * those the driver believes to be appropriate.
822b3b2e
JF
1111 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1112 * int queue_index, u32 maxrate);
1113 * Called when a user wants to set a max-rate limitation of specific
1114 * TX queue.
a54acb3a
ND
1115 * int (*ndo_get_iflink)(const struct net_device *dev);
1116 * Called to get the iflink value of this device.
d746d707 1117 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1118 * bool proto_down);
d746d707
AK
1119 * This function is used to pass protocol port error state information
1120 * to the switch driver. The switch driver can react to the proto_down
1121 * by doing a phys down on the associated switch port.
fc4099f1
PS
1122 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1123 * This function is used to get egress tunnel information for given skb.
1124 * This is useful for retrieving outer tunnel header parameters while
1125 * sampling packet.
871b642a
PA
1126 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1127 * This function is used to specify the headroom that the skb must
1128 * consider when allocation skb during packet reception. Setting
1129 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1130 * forward. Setting a negative value resets the rx headroom to the
871b642a 1131 * default value.
a7862b45
BB
1132 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1133 * This function is used to set or query state related to XDP on the
1134 * netdevice. See definition of enum xdp_netdev_command for details.
d746d707 1135 *
d314774c
SH
1136 */
1137struct net_device_ops {
1138 int (*ndo_init)(struct net_device *dev);
1139 void (*ndo_uninit)(struct net_device *dev);
1140 int (*ndo_open)(struct net_device *dev);
1141 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1142 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1143 struct net_device *dev);
1144 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1145 struct net_device *dev,
1146 netdev_features_t features);
00829823 1147 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1148 struct sk_buff *skb,
99932d4f
DB
1149 void *accel_priv,
1150 select_queue_fallback_t fallback);
d314774c
SH
1151 void (*ndo_change_rx_flags)(struct net_device *dev,
1152 int flags);
d314774c 1153 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1154 int (*ndo_set_mac_address)(struct net_device *dev,
1155 void *addr);
d314774c 1156 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1157 int (*ndo_do_ioctl)(struct net_device *dev,
1158 struct ifreq *ifr, int cmd);
d314774c
SH
1159 int (*ndo_set_config)(struct net_device *dev,
1160 struct ifmap *map);
00829823
SH
1161 int (*ndo_change_mtu)(struct net_device *dev,
1162 int new_mtu);
1163 int (*ndo_neigh_setup)(struct net_device *dev,
1164 struct neigh_parms *);
d314774c
SH
1165 void (*ndo_tx_timeout) (struct net_device *dev);
1166
28172739
ED
1167 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1168 struct rtnl_link_stats64 *storage);
3df5b3c6 1169 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1170 int (*ndo_get_offload_stats)(int attr_id,
1171 const struct net_device *dev,
1172 void *attr_data);
d314774c
SH
1173 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1174
8e586137 1175 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1176 __be16 proto, u16 vid);
8e586137 1177 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1178 __be16 proto, u16 vid);
d314774c 1179#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1180 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1181 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1182 struct netpoll_info *info);
0e34e931 1183 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1184#endif
e0d1095a 1185#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1186 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1187#endif
95c26df8
WM
1188 int (*ndo_set_vf_mac)(struct net_device *dev,
1189 int queue, u8 *mac);
1190 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1191 int queue, u16 vlan,
1192 u8 qos, __be16 proto);
ed616689
SC
1193 int (*ndo_set_vf_rate)(struct net_device *dev,
1194 int vf, int min_tx_rate,
1195 int max_tx_rate);
5f8444a3
GR
1196 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1197 int vf, bool setting);
dd461d6a
HS
1198 int (*ndo_set_vf_trust)(struct net_device *dev,
1199 int vf, bool setting);
95c26df8
WM
1200 int (*ndo_get_vf_config)(struct net_device *dev,
1201 int vf,
1202 struct ifla_vf_info *ivf);
1d8faf48
RE
1203 int (*ndo_set_vf_link_state)(struct net_device *dev,
1204 int vf, int link_state);
3b766cd8
EBE
1205 int (*ndo_get_vf_stats)(struct net_device *dev,
1206 int vf,
1207 struct ifla_vf_stats
1208 *vf_stats);
57b61080
SF
1209 int (*ndo_set_vf_port)(struct net_device *dev,
1210 int vf,
1211 struct nlattr *port[]);
1212 int (*ndo_get_vf_port)(struct net_device *dev,
1213 int vf, struct sk_buff *skb);
cc8e27cc
EC
1214 int (*ndo_set_vf_guid)(struct net_device *dev,
1215 int vf, u64 guid,
1216 int guid_type);
01a3d796
VZ
1217 int (*ndo_set_vf_rss_query_en)(
1218 struct net_device *dev,
1219 int vf, bool setting);
16e5cc64
JF
1220 int (*ndo_setup_tc)(struct net_device *dev,
1221 u32 handle,
1222 __be16 protocol,
1223 struct tc_to_netdev *tc);
d11ead75 1224#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1225 int (*ndo_fcoe_enable)(struct net_device *dev);
1226 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1227 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1228 u16 xid,
1229 struct scatterlist *sgl,
1230 unsigned int sgc);
1231 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1232 u16 xid);
6247e086
YZ
1233 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1234 u16 xid,
1235 struct scatterlist *sgl,
1236 unsigned int sgc);
68bad94e
NP
1237 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1238 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1239#endif
1240
d11ead75 1241#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1242#define NETDEV_FCOE_WWNN 0
1243#define NETDEV_FCOE_WWPN 1
1244 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1245 u64 *wwn, int type);
4d288d57 1246#endif
3c9c36bc 1247
c445477d
BH
1248#ifdef CONFIG_RFS_ACCEL
1249 int (*ndo_rx_flow_steer)(struct net_device *dev,
1250 const struct sk_buff *skb,
1251 u16 rxq_index,
1252 u32 flow_id);
1253#endif
fbaec0ea
JP
1254 int (*ndo_add_slave)(struct net_device *dev,
1255 struct net_device *slave_dev);
1256 int (*ndo_del_slave)(struct net_device *dev,
1257 struct net_device *slave_dev);
c8f44aff
MM
1258 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1259 netdev_features_t features);
5455c699 1260 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1261 netdev_features_t features);
503eebc2
JP
1262 int (*ndo_neigh_construct)(struct net_device *dev,
1263 struct neighbour *n);
1264 void (*ndo_neigh_destroy)(struct net_device *dev,
1265 struct neighbour *n);
77162022
JF
1266
1267 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1268 struct nlattr *tb[],
77162022 1269 struct net_device *dev,
6b6e2725 1270 const unsigned char *addr,
f6f6424b 1271 u16 vid,
77162022
JF
1272 u16 flags);
1273 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1274 struct nlattr *tb[],
77162022 1275 struct net_device *dev,
f6f6424b
JP
1276 const unsigned char *addr,
1277 u16 vid);
77162022
JF
1278 int (*ndo_fdb_dump)(struct sk_buff *skb,
1279 struct netlink_callback *cb,
1280 struct net_device *dev,
5d5eacb3 1281 struct net_device *filter_dev,
d297653d 1282 int *idx);
e5a55a89
JF
1283
1284 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1285 struct nlmsghdr *nlh,
1286 u16 flags);
e5a55a89
JF
1287 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1288 u32 pid, u32 seq,
6cbdceeb 1289 struct net_device *dev,
46c264da
ND
1290 u32 filter_mask,
1291 int nlflags);
407af329 1292 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1293 struct nlmsghdr *nlh,
1294 u16 flags);
4bf84c35
JP
1295 int (*ndo_change_carrier)(struct net_device *dev,
1296 bool new_carrier);
66b52b0d 1297 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1298 struct netdev_phys_item_id *ppid);
db24a904
DA
1299 int (*ndo_get_phys_port_name)(struct net_device *dev,
1300 char *name, size_t len);
7c46a640
AD
1301 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1302 struct udp_tunnel_info *ti);
1303 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1304 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1305 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1306 struct net_device *dev);
1307 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1308 void *priv);
1309
1310 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1311 struct net_device *dev,
1312 void *priv);
25175ba5 1313 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1314 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1315 int queue_index,
1316 u32 maxrate);
a54acb3a 1317 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1318 int (*ndo_change_proto_down)(struct net_device *dev,
1319 bool proto_down);
fc4099f1
PS
1320 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1321 struct sk_buff *skb);
871b642a
PA
1322 void (*ndo_set_rx_headroom)(struct net_device *dev,
1323 int needed_headroom);
a7862b45
BB
1324 int (*ndo_xdp)(struct net_device *dev,
1325 struct netdev_xdp *xdp);
d314774c
SH
1326};
1327
7aa98047
LR
1328/**
1329 * enum net_device_priv_flags - &struct net_device priv_flags
1330 *
1331 * These are the &struct net_device, they are only set internally
1332 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1333 * userspace; this means that the order of these flags can change
7aa98047
LR
1334 * during any kernel release.
1335 *
1336 * You should have a pretty good reason to be extending these flags.
1337 *
1338 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1339 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1340 * @IFF_BONDING: bonding master or slave
7aa98047 1341 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1342 * @IFF_WAN_HDLC: WAN HDLC device
1343 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1344 * release skb->dst
1345 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1346 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1347 * @IFF_MACVLAN_PORT: device used as macvlan port
1348 * @IFF_BRIDGE_PORT: device used as bridge port
1349 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1350 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1351 * @IFF_UNICAST_FLT: Supports unicast filtering
1352 * @IFF_TEAM_PORT: device used as team port
1353 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1354 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1355 * change when it's running
1356 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1357 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1358 * underlying stacked devices
1359 * @IFF_IPVLAN_MASTER: IPvlan master device
1360 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1361 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1362 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1363 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1364 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1365 * @IFF_TEAM: device is a team device
d4ab4286 1366 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1367 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1368 * entity (i.e. the master device for bridged veth)
3c175784 1369 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1370 */
1371enum netdev_priv_flags {
1372 IFF_802_1Q_VLAN = 1<<0,
1373 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1374 IFF_BONDING = 1<<2,
1375 IFF_ISATAP = 1<<3,
1376 IFF_WAN_HDLC = 1<<4,
1377 IFF_XMIT_DST_RELEASE = 1<<5,
1378 IFF_DONT_BRIDGE = 1<<6,
1379 IFF_DISABLE_NETPOLL = 1<<7,
1380 IFF_MACVLAN_PORT = 1<<8,
1381 IFF_BRIDGE_PORT = 1<<9,
1382 IFF_OVS_DATAPATH = 1<<10,
1383 IFF_TX_SKB_SHARING = 1<<11,
1384 IFF_UNICAST_FLT = 1<<12,
1385 IFF_TEAM_PORT = 1<<13,
1386 IFF_SUPP_NOFCS = 1<<14,
1387 IFF_LIVE_ADDR_CHANGE = 1<<15,
1388 IFF_MACVLAN = 1<<16,
1389 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1390 IFF_IPVLAN_MASTER = 1<<18,
1391 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1392 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1393 IFF_NO_QUEUE = 1<<21,
1394 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1395 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1396 IFF_TEAM = 1<<24,
d4ab4286 1397 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1398 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1399 IFF_MACSEC = 1<<27,
7aa98047
LR
1400};
1401
1402#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1403#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1404#define IFF_BONDING IFF_BONDING
7aa98047 1405#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1406#define IFF_WAN_HDLC IFF_WAN_HDLC
1407#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1408#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1409#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1410#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1411#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1412#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1413#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1414#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1415#define IFF_TEAM_PORT IFF_TEAM_PORT
1416#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1417#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1418#define IFF_MACVLAN IFF_MACVLAN
02875878 1419#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1420#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1421#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1422#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1423#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1424#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1425#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1426#define IFF_TEAM IFF_TEAM
d4ab4286 1427#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1428#define IFF_MACSEC IFF_MACSEC
7aa98047 1429
536721b1
KK
1430/**
1431 * struct net_device - The DEVICE structure.
1432 * Actually, this whole structure is a big mistake. It mixes I/O
1433 * data with strictly "high-level" data, and it has to know about
1434 * almost every data structure used in the INET module.
1435 *
1436 * @name: This is the first field of the "visible" part of this structure
1437 * (i.e. as seen by users in the "Space.c" file). It is the name
1438 * of the interface.
1439 *
1440 * @name_hlist: Device name hash chain, please keep it close to name[]
1441 * @ifalias: SNMP alias
1442 * @mem_end: Shared memory end
1443 * @mem_start: Shared memory start
1444 * @base_addr: Device I/O address
1445 * @irq: Device IRQ number
1446 *
14ffbbb8
TG
1447 * @carrier_changes: Stats to monitor carrier on<->off transitions
1448 *
536721b1
KK
1449 * @state: Generic network queuing layer state, see netdev_state_t
1450 * @dev_list: The global list of network devices
5e82b4b2
BH
1451 * @napi_list: List entry used for polling NAPI devices
1452 * @unreg_list: List entry when we are unregistering the
1453 * device; see the function unregister_netdev
1454 * @close_list: List entry used when we are closing the device
62d885fe
BP
1455 * @ptype_all: Device-specific packet handlers for all protocols
1456 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1457 *
1458 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1459 * @features: Currently active device features
1460 * @hw_features: User-changeable features
1461 *
1462 * @wanted_features: User-requested features
1463 * @vlan_features: Mask of features inheritable by VLAN devices
1464 *
1465 * @hw_enc_features: Mask of features inherited by encapsulating devices
1466 * This field indicates what encapsulation
1467 * offloads the hardware is capable of doing,
1468 * and drivers will need to set them appropriately.
1469 *
1470 * @mpls_features: Mask of features inheritable by MPLS
1471 *
1472 * @ifindex: interface index
5e82b4b2 1473 * @group: The group the device belongs to
536721b1
KK
1474 *
1475 * @stats: Statistics struct, which was left as a legacy, use
1476 * rtnl_link_stats64 instead
1477 *
1478 * @rx_dropped: Dropped packets by core network,
1479 * do not use this in drivers
1480 * @tx_dropped: Dropped packets by core network,
1481 * do not use this in drivers
6e7333d3
JW
1482 * @rx_nohandler: nohandler dropped packets by core network on
1483 * inactive devices, do not use this in drivers
536721b1 1484 *
536721b1
KK
1485 * @wireless_handlers: List of functions to handle Wireless Extensions,
1486 * instead of ioctl,
1487 * see <net/iw_handler.h> for details.
1488 * @wireless_data: Instance data managed by the core of wireless extensions
1489 *
1490 * @netdev_ops: Includes several pointers to callbacks,
1491 * if one wants to override the ndo_*() functions
1492 * @ethtool_ops: Management operations
f997c55c
AA
1493 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1494 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1495 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1496 * of Layer 2 headers.
1497 *
1498 * @flags: Interface flags (a la BSD)
1499 * @priv_flags: Like 'flags' but invisible to userspace,
1500 * see if.h for the definitions
1501 * @gflags: Global flags ( kept as legacy )
1502 * @padded: How much padding added by alloc_netdev()
1503 * @operstate: RFC2863 operstate
1504 * @link_mode: Mapping policy to operstate
1505 * @if_port: Selectable AUI, TP, ...
1506 * @dma: DMA channel
1507 * @mtu: Interface MTU value
61e84623
JW
1508 * @min_mtu: Interface Minimum MTU value
1509 * @max_mtu: Interface Maximum MTU value
536721b1 1510 * @type: Interface hardware type
2793a23a 1511 * @hard_header_len: Maximum hardware header length.
536721b1
KK
1512 *
1513 * @needed_headroom: Extra headroom the hardware may need, but not in all
1514 * cases can this be guaranteed
1515 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1516 * cases can this be guaranteed. Some cases also use
1517 * LL_MAX_HEADER instead to allocate the skb
1518 *
1519 * interface address info:
1520 *
1521 * @perm_addr: Permanent hw address
1522 * @addr_assign_type: Hw address assignment type
1523 * @addr_len: Hardware address length
8626a0c8 1524 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1525 * @dev_id: Used to differentiate devices that share
1526 * the same link layer address
1527 * @dev_port: Used to differentiate devices that share
1528 * the same function
1529 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1530 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1531 * has been enabled due to the need to listen to
1532 * additional unicast addresses in a device that
1533 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1534 * @uc: unicast mac addresses
1535 * @mc: multicast mac addresses
1536 * @dev_addrs: list of device hw addresses
1537 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1538 * @promiscuity: Number of times the NIC is told to work in
1539 * promiscuous mode; if it becomes 0 the NIC will
1540 * exit promiscuous mode
536721b1
KK
1541 * @allmulti: Counter, enables or disables allmulticast mode
1542 *
1543 * @vlan_info: VLAN info
1544 * @dsa_ptr: dsa specific data
1545 * @tipc_ptr: TIPC specific data
1546 * @atalk_ptr: AppleTalk link
1547 * @ip_ptr: IPv4 specific data
1548 * @dn_ptr: DECnet specific data
1549 * @ip6_ptr: IPv6 specific data
1550 * @ax25_ptr: AX.25 specific data
1551 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1552 *
1553 * @last_rx: Time of last Rx
1554 * @dev_addr: Hw address (before bcast,
1555 * because most packets are unicast)
1556 *
1557 * @_rx: Array of RX queues
1558 * @num_rx_queues: Number of RX queues
1559 * allocated at register_netdev() time
1560 * @real_num_rx_queues: Number of RX queues currently active in device
1561 *
1562 * @rx_handler: handler for received packets
1563 * @rx_handler_data: XXX: need comments on this one
1564 * @ingress_queue: XXX: need comments on this one
1565 * @broadcast: hw bcast address
1566 *
14ffbbb8
TG
1567 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1568 * indexed by RX queue number. Assigned by driver.
1569 * This must only be set if the ndo_rx_flow_steer
1570 * operation is defined
1571 * @index_hlist: Device index hash chain
1572 *
536721b1
KK
1573 * @_tx: Array of TX queues
1574 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1575 * @real_num_tx_queues: Number of TX queues currently active in device
1576 * @qdisc: Root qdisc from userspace point of view
1577 * @tx_queue_len: Max frames per queue allowed
1578 * @tx_global_lock: XXX: need comments on this one
1579 *
1580 * @xps_maps: XXX: need comments on this one
1581 *
536721b1 1582 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1583 * the watchdog (see dev_watchdog())
536721b1
KK
1584 * @watchdog_timer: List of timers
1585 *
1586 * @pcpu_refcnt: Number of references to this device
1587 * @todo_list: Delayed register/unregister
536721b1
KK
1588 * @link_watch_list: XXX: need comments on this one
1589 *
1590 * @reg_state: Register/unregister state machine
1591 * @dismantle: Device is going to be freed
1592 * @rtnl_link_state: This enum represents the phases of creating
1593 * a new link
1594 *
1595 * @destructor: Called from unregister,
1596 * can be used to call free_netdev
1597 * @npinfo: XXX: need comments on this one
1598 * @nd_net: Network namespace this network device is inside
1599 *
1600 * @ml_priv: Mid-layer private
1601 * @lstats: Loopback statistics
1602 * @tstats: Tunnel statistics
1603 * @dstats: Dummy statistics
1604 * @vstats: Virtual ethernet statistics
1605 *
1606 * @garp_port: GARP
1607 * @mrp_port: MRP
1608 *
1609 * @dev: Class/net/name entry
1610 * @sysfs_groups: Space for optional device, statistics and wireless
1611 * sysfs groups
1612 *
1613 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1614 * @rtnl_link_ops: Rtnl_link_ops
1615 *
1616 * @gso_max_size: Maximum size of generic segmentation offload
1617 * @gso_max_segs: Maximum number of segments that can be passed to the
1618 * NIC for GSO
1619 *
1620 * @dcbnl_ops: Data Center Bridging netlink ops
1621 * @num_tc: Number of traffic classes in the net device
1622 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1623 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1624 *
1625 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1626 *
1627 * @priomap: XXX: need comments on this one
1628 * @phydev: Physical device may attach itself
1629 * for hardware timestamping
1630 *
123b3652
ED
1631 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1632 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1633 *
d746d707
AK
1634 * @proto_down: protocol port state information can be sent to the
1635 * switch driver and used to set the phys state of the
1636 * switch port.
1637 *
1da177e4
LT
1638 * FIXME: cleanup struct net_device such that network protocol info
1639 * moves out.
1640 */
1641
d94d9fee 1642struct net_device {
1da177e4 1643 char name[IFNAMSIZ];
9356b8fc 1644 struct hlist_node name_hlist;
0b815a1a 1645 char *ifalias;
1da177e4
LT
1646 /*
1647 * I/O specific fields
1648 * FIXME: Merge these and struct ifmap into one
1649 */
536721b1
KK
1650 unsigned long mem_end;
1651 unsigned long mem_start;
1652 unsigned long base_addr;
1653 int irq;
1da177e4 1654
14ffbbb8
TG
1655 atomic_t carrier_changes;
1656
1da177e4 1657 /*
536721b1
KK
1658 * Some hardware also needs these fields (state,dev_list,
1659 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1660 * part of the usual set specified in Space.c.
1661 */
1662
1da177e4
LT
1663 unsigned long state;
1664
7562f876 1665 struct list_head dev_list;
bea3348e 1666 struct list_head napi_list;
44a0873d 1667 struct list_head unreg_list;
5cde2829 1668 struct list_head close_list;
7866a621
SN
1669 struct list_head ptype_all;
1670 struct list_head ptype_specific;
2f268f12 1671
2f268f12
VF
1672 struct {
1673 struct list_head upper;
1674 struct list_head lower;
1675 } adj_list;
1676
c8f44aff 1677 netdev_features_t features;
c8f44aff 1678 netdev_features_t hw_features;
c8f44aff 1679 netdev_features_t wanted_features;
c8f44aff 1680 netdev_features_t vlan_features;
6a674e9c 1681 netdev_features_t hw_enc_features;
0d89d203 1682 netdev_features_t mpls_features;
802ab55a 1683 netdev_features_t gso_partial_features;
04ed3e74 1684
1da177e4 1685 int ifindex;
7a66bbc9 1686 int group;
1da177e4 1687
c45d286e 1688 struct net_device_stats stats;
015f0688 1689
015f0688
ED
1690 atomic_long_t rx_dropped;
1691 atomic_long_t tx_dropped;
6e7333d3 1692 atomic_long_t rx_nohandler;
1da177e4 1693
b86e0280 1694#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1695 const struct iw_handler_def *wireless_handlers;
1696 struct iw_public_data *wireless_data;
b86e0280 1697#endif
d314774c 1698 const struct net_device_ops *netdev_ops;
76fd8593 1699 const struct ethtool_ops *ethtool_ops;
4170604f 1700#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1701 const struct switchdev_ops *switchdev_ops;
4170604f 1702#endif
1b69c6d0
DA
1703#ifdef CONFIG_NET_L3_MASTER_DEV
1704 const struct l3mdev_ops *l3mdev_ops;
1705#endif
f997c55c
AA
1706#if IS_ENABLED(CONFIG_IPV6)
1707 const struct ndisc_ops *ndisc_ops;
1708#endif
1da177e4 1709
3b04ddde
SH
1710 const struct header_ops *header_ops;
1711
536721b1
KK
1712 unsigned int flags;
1713 unsigned int priv_flags;
1714
1da177e4 1715 unsigned short gflags;
536721b1 1716 unsigned short padded;
1da177e4 1717
536721b1
KK
1718 unsigned char operstate;
1719 unsigned char link_mode;
b00055aa 1720
536721b1
KK
1721 unsigned char if_port;
1722 unsigned char dma;
bdc220da 1723
536721b1 1724 unsigned int mtu;
61e84623
JW
1725 unsigned int min_mtu;
1726 unsigned int max_mtu;
536721b1
KK
1727 unsigned short type;
1728 unsigned short hard_header_len;
1da177e4 1729
f5184d26
JB
1730 unsigned short needed_headroom;
1731 unsigned short needed_tailroom;
1732
1da177e4 1733 /* Interface address info. */
536721b1
KK
1734 unsigned char perm_addr[MAX_ADDR_LEN];
1735 unsigned char addr_assign_type;
1736 unsigned char addr_len;
a0a9663d 1737 unsigned short neigh_priv_len;
536721b1
KK
1738 unsigned short dev_id;
1739 unsigned short dev_port;
ccffad25 1740 spinlock_t addr_list_lock;
14ffbbb8
TG
1741 unsigned char name_assign_type;
1742 bool uc_promisc;
536721b1
KK
1743 struct netdev_hw_addr_list uc;
1744 struct netdev_hw_addr_list mc;
1745 struct netdev_hw_addr_list dev_addrs;
1746
4c3d5e7b
ED
1747#ifdef CONFIG_SYSFS
1748 struct kset *queues_kset;
1749#endif
9d45abe1
WC
1750 unsigned int promiscuity;
1751 unsigned int allmulti;
1da177e4 1752
1da177e4 1753
5e82b4b2 1754 /* Protocol-specific pointers */
65ac6a5f 1755
d11ead75 1756#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1757 struct vlan_info __rcu *vlan_info;
65ac6a5f 1758#endif
34a430d7 1759#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1760 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1761#endif
1762#if IS_ENABLED(CONFIG_TIPC)
536721b1 1763 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1764#endif
536721b1
KK
1765 void *atalk_ptr;
1766 struct in_device __rcu *ip_ptr;
1767 struct dn_dev __rcu *dn_ptr;
1768 struct inet6_dev __rcu *ip6_ptr;
1769 void *ax25_ptr;
1770 struct wireless_dev *ieee80211_ptr;
98a18b6f 1771 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1772#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1773 struct mpls_dev __rcu *mpls_ptr;
1774#endif
1da177e4 1775
9356b8fc 1776/*
cd13539b 1777 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1778 */
536721b1 1779 unsigned long last_rx;
4dc89133 1780
9356b8fc 1781 /* Interface address info used in eth_type_trans() */
536721b1 1782 unsigned char *dev_addr;
f001fde5 1783
a953be53 1784#ifdef CONFIG_SYSFS
0a9627f2
TH
1785 struct netdev_rx_queue *_rx;
1786
0a9627f2 1787 unsigned int num_rx_queues;
62fe0b40 1788 unsigned int real_num_rx_queues;
df334545 1789#endif
0a9627f2 1790
3b47d303 1791 unsigned long gro_flush_timeout;
61391cde 1792 rx_handler_func_t __rcu *rx_handler;
1793 void __rcu *rx_handler_data;
e8a0464c 1794
4cda01e8 1795#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1796 struct tcf_proto __rcu *ingress_cl_list;
1797#endif
24824a09 1798 struct netdev_queue __rcu *ingress_queue;
e687ad60 1799#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1800 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1801#endif
d2788d34 1802
536721b1 1803 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1804#ifdef CONFIG_RFS_ACCEL
1805 struct cpu_rmap *rx_cpu_rmap;
1806#endif
1807 struct hlist_node index_hlist;
cd13539b
ED
1808
1809/*
1810 * Cache lines mostly used on transmit path
1811 */
e8a0464c
DM
1812 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1813 unsigned int num_tx_queues;
fd2ea0a7 1814 unsigned int real_num_tx_queues;
af356afa 1815 struct Qdisc *qdisc;
59cc1f61
JK
1816#ifdef CONFIG_NET_SCHED
1817 DECLARE_HASHTABLE (qdisc_hash, 4);
1818#endif
536721b1 1819 unsigned long tx_queue_len;
c3f26a26 1820 spinlock_t tx_global_lock;
14ffbbb8 1821 int watchdog_timeo;
cd13539b 1822
bf264145 1823#ifdef CONFIG_XPS
a4177869 1824 struct xps_dev_maps __rcu *xps_maps;
bf264145 1825#endif
1f211a1b
DB
1826#ifdef CONFIG_NET_CLS_ACT
1827 struct tcf_proto __rcu *egress_cl_list;
1828#endif
0c4f691f 1829
9356b8fc 1830 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1831 struct timer_list watchdog_timer;
1832
29b4433d 1833 int __percpu *pcpu_refcnt;
1da177e4 1834 struct list_head todo_list;
1da177e4 1835
e014debe 1836 struct list_head link_watch_list;
572a103d 1837
1da177e4 1838 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1839 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1840 NETREG_UNREGISTERING, /* called unregister_netdevice */
1841 NETREG_UNREGISTERED, /* completed unregister todo */
1842 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1843 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1844 } reg_state:8;
1845
536721b1 1846 bool dismantle;
a2835763
PM
1847
1848 enum {
1849 RTNL_LINK_INITIALIZED,
1850 RTNL_LINK_INITIALIZING,
1851 } rtnl_link_state:16;
1da177e4 1852
d314774c 1853 void (*destructor)(struct net_device *dev);
1da177e4 1854
1da177e4 1855#ifdef CONFIG_NETPOLL
5fbee843 1856 struct netpoll_info __rcu *npinfo;
1da177e4 1857#endif
eae792b7 1858
0c5c9fb5 1859 possible_net_t nd_net;
4a1c5371 1860
4951704b 1861 /* mid-layer private */
a7855c78 1862 union {
536721b1
KK
1863 void *ml_priv;
1864 struct pcpu_lstats __percpu *lstats;
8f84985f 1865 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1866 struct pcpu_dstats __percpu *dstats;
1867 struct pcpu_vstats __percpu *vstats;
a7855c78 1868 };
536721b1 1869
3cc77ec7 1870 struct garp_port __rcu *garp_port;
febf018d 1871 struct mrp_port __rcu *mrp_port;
1da177e4 1872
5e82b4b2 1873 struct device dev;
0c509a6c 1874 const struct attribute_group *sysfs_groups[4];
a953be53 1875 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1876
38f7b870 1877 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1878
82cc1a7a
PWJ
1879 /* for setting kernel sock attribute on TCP connection setup */
1880#define GSO_MAX_SIZE 65536
1881 unsigned int gso_max_size;
30b678d8
BH
1882#define GSO_MAX_SEGS 65535
1883 u16 gso_max_segs;
743b03a8 1884
7a6b6f51 1885#ifdef CONFIG_DCB
32953543 1886 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1887#endif
5e82b4b2
BH
1888 u8 num_tc;
1889 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1890 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1891
d11ead75 1892#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1893 unsigned int fcoe_ddp_xid;
5bc1421e 1894#endif
86f8515f 1895#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1896 struct netprio_map __rcu *priomap;
4d288d57 1897#endif
5e82b4b2
BH
1898 struct phy_device *phydev;
1899 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1900 struct lock_class_key *qdisc_running_key;
5e82b4b2 1901 bool proto_down;
1da177e4 1902};
43cb76d9 1903#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1904
1905#define NETDEV_ALIGN 32
1da177e4 1906
4f57c087
JF
1907static inline
1908int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1909{
1910 return dev->prio_tc_map[prio & TC_BITMASK];
1911}
1912
1913static inline
1914int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1915{
1916 if (tc >= dev->num_tc)
1917 return -EINVAL;
1918
1919 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1920 return 0;
1921}
1922
8d059b0f 1923int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1924void netdev_reset_tc(struct net_device *dev);
1925int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1926int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1927
1928static inline
1929int netdev_get_num_tc(struct net_device *dev)
1930{
1931 return dev->num_tc;
1932}
1933
e8a0464c
DM
1934static inline
1935struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1936 unsigned int index)
1937{
1938 return &dev->_tx[index];
1939}
1940
10c51b56
DB
1941static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1942 const struct sk_buff *skb)
1943{
1944 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1945}
1946
e8a0464c
DM
1947static inline void netdev_for_each_tx_queue(struct net_device *dev,
1948 void (*f)(struct net_device *,
1949 struct netdev_queue *,
1950 void *),
1951 void *arg)
1952{
1953 unsigned int i;
1954
1955 for (i = 0; i < dev->num_tx_queues; i++)
1956 f(dev, &dev->_tx[i], arg);
1957}
1958
d3fff6c4
ED
1959#define netdev_lockdep_set_classes(dev) \
1960{ \
1961 static struct lock_class_key qdisc_tx_busylock_key; \
1962 static struct lock_class_key qdisc_running_key; \
1963 static struct lock_class_key qdisc_xmit_lock_key; \
1964 static struct lock_class_key dev_addr_list_lock_key; \
1965 unsigned int i; \
1966 \
1967 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1968 (dev)->qdisc_running_key = &qdisc_running_key; \
1969 lockdep_set_class(&(dev)->addr_list_lock, \
1970 &dev_addr_list_lock_key); \
1971 for (i = 0; i < (dev)->num_tx_queues; i++) \
1972 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1973 &qdisc_xmit_lock_key); \
1974}
1975
f629d208 1976struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1977 struct sk_buff *skb,
1978 void *accel_priv);
8c4c49df 1979
871b642a
PA
1980/* returns the headroom that the master device needs to take in account
1981 * when forwarding to this dev
1982 */
1983static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1984{
1985 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1986}
1987
1988static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1989{
1990 if (dev->netdev_ops->ndo_set_rx_headroom)
1991 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1992}
1993
1994/* set the device rx headroom to the dev's default */
1995static inline void netdev_reset_rx_headroom(struct net_device *dev)
1996{
1997 netdev_set_rx_headroom(dev, -1);
1998}
1999
c346dca1
YH
2000/*
2001 * Net namespace inlines
2002 */
2003static inline
2004struct net *dev_net(const struct net_device *dev)
2005{
c2d9ba9b 2006 return read_pnet(&dev->nd_net);
c346dca1
YH
2007}
2008
2009static inline
f5aa23fd 2010void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2011{
0c5c9fb5 2012 write_pnet(&dev->nd_net, net);
c346dca1
YH
2013}
2014
3e8a72d1 2015static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 2016{
3fc88677 2017#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
2018 if (dev->dsa_ptr != NULL)
2019 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 2020#endif
5aed85ce 2021 return false;
396138f0
LB
2022}
2023
bea3348e
SH
2024/**
2025 * netdev_priv - access network device private data
2026 * @dev: network device
2027 *
2028 * Get network device private data
2029 */
6472ce60 2030static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2031{
1ce8e7b5 2032 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2033}
2034
1da177e4
LT
2035/* Set the sysfs physical device reference for the network logical device
2036 * if set prior to registration will cause a symlink during initialization.
2037 */
43cb76d9 2038#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2039
384912ed 2040/* Set the sysfs device type for the network logical device to allow
3f79410c 2041 * fine-grained identification of different network device types. For
5e82b4b2 2042 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2043 */
2044#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2045
82dc3c63
ED
2046/* Default NAPI poll() weight
2047 * Device drivers are strongly advised to not use bigger value
2048 */
2049#define NAPI_POLL_WEIGHT 64
2050
3b582cc1 2051/**
5e82b4b2 2052 * netif_napi_add - initialize a NAPI context
3b582cc1 2053 * @dev: network device
5e82b4b2 2054 * @napi: NAPI context
3b582cc1
SH
2055 * @poll: polling function
2056 * @weight: default weight
2057 *
5e82b4b2
BH
2058 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2059 * *any* of the other NAPI-related functions.
3b582cc1 2060 */
d565b0a1
HX
2061void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2062 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2063
d64b5e85 2064/**
5e82b4b2 2065 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2066 * @dev: network device
5e82b4b2 2067 * @napi: NAPI context
d64b5e85
ED
2068 * @poll: polling function
2069 * @weight: default weight
2070 *
2071 * This variant of netif_napi_add() should be used from drivers using NAPI
2072 * to exclusively poll a TX queue.
2073 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2074 */
2075static inline void netif_tx_napi_add(struct net_device *dev,
2076 struct napi_struct *napi,
2077 int (*poll)(struct napi_struct *, int),
2078 int weight)
2079{
2080 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2081 netif_napi_add(dev, napi, poll, weight);
2082}
2083
d8156534 2084/**
5e82b4b2
BH
2085 * netif_napi_del - remove a NAPI context
2086 * @napi: NAPI context
d8156534 2087 *
5e82b4b2 2088 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2089 */
d565b0a1
HX
2090void netif_napi_del(struct napi_struct *napi);
2091
2092struct napi_gro_cb {
78a478d0 2093 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2094 void *frag0;
78a478d0 2095
7489594c
HX
2096 /* Length of frag0. */
2097 unsigned int frag0_len;
2098
86911732 2099 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2100 int data_offset;
86911732 2101
d565b0a1 2102 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2103 u16 flush;
2104
2105 /* Save the IP ID here and check when we get to the transport layer */
2106 u16 flush_id;
d565b0a1
HX
2107
2108 /* Number of segments aggregated. */
2e71a6f8
ED
2109 u16 count;
2110
15e2396d
TH
2111 /* Start offset for remote checksum offload */
2112 u16 gro_remcsum_start;
2113
2e71a6f8
ED
2114 /* jiffies when first packet was created/queued */
2115 unsigned long age;
86347245 2116
afe93325 2117 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2118 u16 proto;
2119
baa32ff4
TH
2120 /* This is non-zero if the packet may be of the same flow. */
2121 u8 same_flow:1;
2122
fac8e0f5
JG
2123 /* Used in tunnel GRO receive */
2124 u8 encap_mark:1;
573e8fca
TH
2125
2126 /* GRO checksum is valid */
2127 u8 csum_valid:1;
2128
662880f4
TH
2129 /* Number of checksums via CHECKSUM_UNNECESSARY */
2130 u8 csum_cnt:3;
c3c7c254 2131
baa32ff4
TH
2132 /* Free the skb? */
2133 u8 free:2;
2134#define NAPI_GRO_FREE 1
2135#define NAPI_GRO_FREE_STOLEN_HEAD 2
2136
efc98d08
TH
2137 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2138 u8 is_ipv6:1;
2139
a0ca153f
AD
2140 /* Used in GRE, set in fou/gue_gro_receive */
2141 u8 is_fou:1;
2142
1530545e
AD
2143 /* Used to determine if flush_id can be ignored */
2144 u8 is_atomic:1;
2145
fcd91dd4
SD
2146 /* Number of gro_receive callbacks this packet already went through */
2147 u8 recursion_counter:4;
2148
2149 /* 1 bit hole */
baa32ff4 2150
bf5a755f
JC
2151 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2152 __wsum csum;
2153
c3c7c254
ED
2154 /* used in skb_gro_receive() slow path */
2155 struct sk_buff *last;
d565b0a1
HX
2156};
2157
2158#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2159
fcd91dd4
SD
2160#define GRO_RECURSION_LIMIT 15
2161static inline int gro_recursion_inc_test(struct sk_buff *skb)
2162{
2163 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2164}
2165
2166typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2167static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2168 struct sk_buff **head,
2169 struct sk_buff *skb)
2170{
2171 if (unlikely(gro_recursion_inc_test(skb))) {
2172 NAPI_GRO_CB(skb)->flush |= 1;
2173 return NULL;
2174 }
2175
2176 return cb(head, skb);
2177}
2178
2179typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2180 struct sk_buff *);
2181static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2182 struct sock *sk,
2183 struct sk_buff **head,
2184 struct sk_buff *skb)
2185{
2186 if (unlikely(gro_recursion_inc_test(skb))) {
2187 NAPI_GRO_CB(skb)->flush |= 1;
2188 return NULL;
2189 }
2190
2191 return cb(sk, head, skb);
2192}
2193
1da177e4 2194struct packet_type {
f2ccd8fa
DM
2195 __be16 type; /* This is really htons(ether_type). */
2196 struct net_device *dev; /* NULL is wildcarded here */
2197 int (*func) (struct sk_buff *,
2198 struct net_device *,
2199 struct packet_type *,
2200 struct net_device *);
c0de08d0
EL
2201 bool (*id_match)(struct packet_type *ptype,
2202 struct sock *sk);
1da177e4
LT
2203 void *af_packet_priv;
2204 struct list_head list;
2205};
2206
f191a1d1 2207struct offload_callbacks {
576a30eb 2208 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2209 netdev_features_t features);
d565b0a1 2210 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2211 struct sk_buff *skb);
299603e8 2212 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2213};
2214
2215struct packet_offload {
2216 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2217 u16 priority;
f191a1d1
VY
2218 struct offload_callbacks callbacks;
2219 struct list_head list;
1da177e4
LT
2220};
2221
5e82b4b2 2222/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2223struct pcpu_sw_netstats {
2224 u64 rx_packets;
2225 u64 rx_bytes;
2226 u64 tx_packets;
2227 u64 tx_bytes;
2228 struct u64_stats_sync syncp;
2229};
2230
aabc92bb
PNA
2231#define __netdev_alloc_pcpu_stats(type, gfp) \
2232({ \
2233 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2234 if (pcpu_stats) { \
2235 int __cpu; \
2236 for_each_possible_cpu(__cpu) { \
2237 typeof(type) *stat; \
2238 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2239 u64_stats_init(&stat->syncp); \
2240 } \
2241 } \
2242 pcpu_stats; \
1c213bd2
WC
2243})
2244
aabc92bb 2245#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2246 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2247
764f5e54
JP
2248enum netdev_lag_tx_type {
2249 NETDEV_LAG_TX_TYPE_UNKNOWN,
2250 NETDEV_LAG_TX_TYPE_RANDOM,
2251 NETDEV_LAG_TX_TYPE_BROADCAST,
2252 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2253 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2254 NETDEV_LAG_TX_TYPE_HASH,
2255};
2256
2257struct netdev_lag_upper_info {
2258 enum netdev_lag_tx_type tx_type;
2259};
2260
fb1b2e3c
JP
2261struct netdev_lag_lower_state_info {
2262 u8 link_up : 1,
2263 tx_enabled : 1;
2264};
2265
1da177e4
LT
2266#include <linux/notifier.h>
2267
dcfe1421
AW
2268/* netdevice notifier chain. Please remember to update the rtnetlink
2269 * notification exclusion list in rtnetlink_event() when adding new
2270 * types.
2271 */
2272#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2273#define NETDEV_DOWN 0x0002
2274#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2275 detected a hardware crash and restarted
2276 - we can use this eg to kick tcp sessions
2277 once done */
2278#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2279#define NETDEV_REGISTER 0x0005
2280#define NETDEV_UNREGISTER 0x0006
1d486bfb 2281#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2282#define NETDEV_CHANGEADDR 0x0008
2283#define NETDEV_GOING_DOWN 0x0009
2284#define NETDEV_CHANGENAME 0x000A
2285#define NETDEV_FEAT_CHANGE 0x000B
2286#define NETDEV_BONDING_FAILOVER 0x000C
2287#define NETDEV_PRE_UP 0x000D
2288#define NETDEV_PRE_TYPE_CHANGE 0x000E
2289#define NETDEV_POST_TYPE_CHANGE 0x000F
2290#define NETDEV_POST_INIT 0x0010
0115e8e3 2291#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2292#define NETDEV_RELEASE 0x0012
2293#define NETDEV_NOTIFY_PEERS 0x0013
2294#define NETDEV_JOIN 0x0014
42e52bf9 2295#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2296#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2297#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2298#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2299#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2300#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2301#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2302#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2303#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2304
f629d208
JP
2305int register_netdevice_notifier(struct notifier_block *nb);
2306int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2307
2308struct netdev_notifier_info {
2309 struct net_device *dev;
2310};
2311
be9efd36
JP
2312struct netdev_notifier_change_info {
2313 struct netdev_notifier_info info; /* must be first */
2314 unsigned int flags_changed;
2315};
2316
0e4ead9d
JP
2317struct netdev_notifier_changeupper_info {
2318 struct netdev_notifier_info info; /* must be first */
2319 struct net_device *upper_dev; /* new upper dev */
2320 bool master; /* is upper dev master */
5e82b4b2 2321 bool linking; /* is the notification for link or unlink */
29bf24af 2322 void *upper_info; /* upper dev info */
0e4ead9d
JP
2323};
2324
04d48266
JP
2325struct netdev_notifier_changelowerstate_info {
2326 struct netdev_notifier_info info; /* must be first */
2327 void *lower_state_info; /* is lower dev state */
2328};
2329
75538c2b
CW
2330static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2331 struct net_device *dev)
2332{
2333 info->dev = dev;
2334}
2335
351638e7
JP
2336static inline struct net_device *
2337netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2338{
2339 return info->dev;
2340}
2341
f629d208 2342int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2343
2344
1da177e4
LT
2345extern rwlock_t dev_base_lock; /* Device list lock */
2346
881d966b
EB
2347#define for_each_netdev(net, d) \
2348 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2349#define for_each_netdev_reverse(net, d) \
2350 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2351#define for_each_netdev_rcu(net, d) \
2352 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2353#define for_each_netdev_safe(net, d, n) \
2354 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2355#define for_each_netdev_continue(net, d) \
2356 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2357#define for_each_netdev_continue_rcu(net, d) \
2358 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2359#define for_each_netdev_in_bond_rcu(bond, slave) \
2360 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2361 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2362#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2363
a050c33f
DL
2364static inline struct net_device *next_net_device(struct net_device *dev)
2365{
2366 struct list_head *lh;
2367 struct net *net;
2368
c346dca1 2369 net = dev_net(dev);
a050c33f
DL
2370 lh = dev->dev_list.next;
2371 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2372}
2373
ce81b76a
ED
2374static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2375{
2376 struct list_head *lh;
2377 struct net *net;
2378
2379 net = dev_net(dev);
ccf43438 2380 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2381 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2382}
2383
a050c33f
DL
2384static inline struct net_device *first_net_device(struct net *net)
2385{
2386 return list_empty(&net->dev_base_head) ? NULL :
2387 net_device_entry(net->dev_base_head.next);
2388}
7562f876 2389
ccf43438
ED
2390static inline struct net_device *first_net_device_rcu(struct net *net)
2391{
2392 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2393
2394 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2395}
2396
f629d208
JP
2397int netdev_boot_setup_check(struct net_device *dev);
2398unsigned long netdev_boot_base(const char *prefix, int unit);
2399struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2400 const char *hwaddr);
2401struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2402struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2403void dev_add_pack(struct packet_type *pt);
2404void dev_remove_pack(struct packet_type *pt);
2405void __dev_remove_pack(struct packet_type *pt);
2406void dev_add_offload(struct packet_offload *po);
2407void dev_remove_offload(struct packet_offload *po);
f629d208 2408
a54acb3a 2409int dev_get_iflink(const struct net_device *dev);
fc4099f1 2410int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2411struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2412 unsigned short mask);
f629d208
JP
2413struct net_device *dev_get_by_name(struct net *net, const char *name);
2414struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2415struct net_device *__dev_get_by_name(struct net *net, const char *name);
2416int dev_alloc_name(struct net_device *dev, const char *name);
2417int dev_open(struct net_device *dev);
2418int dev_close(struct net_device *dev);
99c4a26a 2419int dev_close_many(struct list_head *head, bool unlink);
f629d208 2420void dev_disable_lro(struct net_device *dev);
0c4b51f0 2421int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2422int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2423int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2424int register_netdevice(struct net_device *dev);
2425void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2426void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2427static inline void unregister_netdevice(struct net_device *dev)
2428{
2429 unregister_netdevice_queue(dev, NULL);
2430}
2431
f629d208
JP
2432int netdev_refcnt_read(const struct net_device *dev);
2433void free_netdev(struct net_device *dev);
74d332c1 2434void netdev_freemem(struct net_device *dev);
f629d208
JP
2435void synchronize_net(void);
2436int init_dummy_netdev(struct net_device *dev);
937f1ba5 2437
f60e5990 2438DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2439#define XMIT_RECURSION_LIMIT 10
2440
f60e5990 2441static inline int dev_recursion_level(void)
2442{
2443 return this_cpu_read(xmit_recursion);
2444}
2445
f629d208
JP
2446struct net_device *dev_get_by_index(struct net *net, int ifindex);
2447struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2448struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2449int netdev_get_name(struct net *net, char *name, int ifindex);
2450int dev_restart(struct net_device *dev);
f629d208 2451int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2452
2453static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2454{
2455 return NAPI_GRO_CB(skb)->data_offset;
2456}
2457
2458static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2459{
2460 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2461}
2462
2463static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2464{
2465 NAPI_GRO_CB(skb)->data_offset += len;
2466}
2467
a5b1cf28
HX
2468static inline void *skb_gro_header_fast(struct sk_buff *skb,
2469 unsigned int offset)
86911732 2470{
a5b1cf28
HX
2471 return NAPI_GRO_CB(skb)->frag0 + offset;
2472}
78a478d0 2473
a5b1cf28
HX
2474static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2475{
2476 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2477}
78a478d0 2478
a5b1cf28
HX
2479static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2480 unsigned int offset)
2481{
17dd759c
HX
2482 if (!pskb_may_pull(skb, hlen))
2483 return NULL;
2484
a5b1cf28
HX
2485 NAPI_GRO_CB(skb)->frag0 = NULL;
2486 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2487 return skb->data + offset;
86911732 2488}
1da177e4 2489
36e7b1b8
HX
2490static inline void *skb_gro_network_header(struct sk_buff *skb)
2491{
78d3fd0b
HX
2492 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2493 skb_network_offset(skb);
36e7b1b8
HX
2494}
2495
bf5a755f
JC
2496static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2497 const void *start, unsigned int len)
2498{
573e8fca 2499 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2500 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2501 csum_partial(start, len, 0));
2502}
2503
573e8fca
TH
2504/* GRO checksum functions. These are logical equivalents of the normal
2505 * checksum functions (in skbuff.h) except that they operate on the GRO
2506 * offsets and fields in sk_buff.
2507 */
2508
2509__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2510
15e2396d
TH
2511static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2512{
b7fe10e5 2513 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2514}
2515
573e8fca
TH
2516static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2517 bool zero_okay,
2518 __sum16 check)
2519{
6edec0e6
TH
2520 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2521 skb_checksum_start_offset(skb) <
2522 skb_gro_offset(skb)) &&
15e2396d 2523 !skb_at_gro_remcsum_start(skb) &&
662880f4 2524 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2525 (!zero_okay || check));
2526}
2527
2528static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2529 __wsum psum)
2530{
2531 if (NAPI_GRO_CB(skb)->csum_valid &&
2532 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2533 return 0;
2534
2535 NAPI_GRO_CB(skb)->csum = psum;
2536
2537 return __skb_gro_checksum_complete(skb);
2538}
2539
573e8fca
TH
2540static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2541{
662880f4
TH
2542 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2543 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2544 NAPI_GRO_CB(skb)->csum_cnt--;
2545 } else {
2546 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2547 * verified a new top level checksum or an encapsulated one
2548 * during GRO. This saves work if we fallback to normal path.
2549 */
2550 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2551 }
2552}
2553
2554#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2555 compute_pseudo) \
2556({ \
2557 __sum16 __ret = 0; \
2558 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2559 __ret = __skb_gro_checksum_validate_complete(skb, \
2560 compute_pseudo(skb, proto)); \
5a212329
TH
2561 if (__ret) \
2562 __skb_mark_checksum_bad(skb); \
2563 else \
573e8fca
TH
2564 skb_gro_incr_csum_unnecessary(skb); \
2565 __ret; \
2566})
2567
2568#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2569 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2570
2571#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2572 compute_pseudo) \
2573 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2574
2575#define skb_gro_checksum_simple_validate(skb) \
2576 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2577
d96535a1
TH
2578static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2579{
2580 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2581 !NAPI_GRO_CB(skb)->csum_valid);
2582}
2583
2584static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2585 __sum16 check, __wsum pseudo)
2586{
2587 NAPI_GRO_CB(skb)->csum = ~pseudo;
2588 NAPI_GRO_CB(skb)->csum_valid = 1;
2589}
2590
2591#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2592do { \
2593 if (__skb_gro_checksum_convert_check(skb)) \
2594 __skb_gro_checksum_convert(skb, check, \
2595 compute_pseudo(skb, proto)); \
2596} while (0)
2597
26c4f7da
TH
2598struct gro_remcsum {
2599 int offset;
2600 __wsum delta;
2601};
2602
2603static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2604{
846cd667 2605 grc->offset = 0;
26c4f7da
TH
2606 grc->delta = 0;
2607}
2608
b7fe10e5
TH
2609static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2610 unsigned int off, size_t hdrlen,
2611 int start, int offset,
2612 struct gro_remcsum *grc,
2613 bool nopartial)
dcdc8994
TH
2614{
2615 __wsum delta;
b7fe10e5 2616 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2617
2618 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2619
15e2396d 2620 if (!nopartial) {
b7fe10e5
TH
2621 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2622 return ptr;
2623 }
2624
2625 ptr = skb_gro_header_fast(skb, off);
2626 if (skb_gro_header_hard(skb, off + plen)) {
2627 ptr = skb_gro_header_slow(skb, off + plen, off);
2628 if (!ptr)
2629 return NULL;
15e2396d
TH
2630 }
2631
b7fe10e5
TH
2632 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2633 start, offset);
dcdc8994
TH
2634
2635 /* Adjust skb->csum since we changed the packet */
dcdc8994 2636 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2637
b7fe10e5 2638 grc->offset = off + hdrlen + offset;
26c4f7da 2639 grc->delta = delta;
b7fe10e5
TH
2640
2641 return ptr;
dcdc8994
TH
2642}
2643
26c4f7da
TH
2644static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2645 struct gro_remcsum *grc)
2646{
b7fe10e5
TH
2647 void *ptr;
2648 size_t plen = grc->offset + sizeof(u16);
2649
26c4f7da
TH
2650 if (!grc->delta)
2651 return;
2652
b7fe10e5
TH
2653 ptr = skb_gro_header_fast(skb, grc->offset);
2654 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2655 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2656 if (!ptr)
2657 return;
2658 }
2659
2660 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2661}
dcdc8994 2662
0c4e8581
SH
2663static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2664 unsigned short type,
3b04ddde 2665 const void *daddr, const void *saddr,
95c96174 2666 unsigned int len)
0c4e8581 2667{
f1ecfd5d 2668 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2669 return 0;
3b04ddde
SH
2670
2671 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2672}
2673
b95cce35
SH
2674static inline int dev_parse_header(const struct sk_buff *skb,
2675 unsigned char *haddr)
2676{
2677 const struct net_device *dev = skb->dev;
2678
1b83336b 2679 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2680 return 0;
3b04ddde 2681 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2682}
2683
2793a23a
WB
2684/* ll_header must have at least hard_header_len allocated */
2685static inline bool dev_validate_header(const struct net_device *dev,
2686 char *ll_header, int len)
2687{
2688 if (likely(len >= dev->hard_header_len))
2689 return true;
2690
2691 if (capable(CAP_SYS_RAWIO)) {
2692 memset(ll_header + len, 0, dev->hard_header_len - len);
2693 return true;
2694 }
2695
2696 if (dev->header_ops && dev->header_ops->validate)
2697 return dev->header_ops->validate(ll_header, len);
2698
2699 return false;
2700}
2701
1da177e4 2702typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2703int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2704static inline int unregister_gifconf(unsigned int family)
2705{
2706 return register_gifconf(family, NULL);
2707}
2708
99bbc707 2709#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2710#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2711struct sd_flow_limit {
2712 u64 count;
2713 unsigned int num_buckets;
2714 unsigned int history_head;
2715 u16 history[FLOW_LIMIT_HISTORY];
2716 u8 buckets[];
2717};
2718
2719extern int netdev_flow_limit_table_len;
2720#endif /* CONFIG_NET_FLOW_LIMIT */
2721
1da177e4 2722/*
5e82b4b2 2723 * Incoming packets are placed on per-CPU queues
1da177e4 2724 */
d94d9fee 2725struct softnet_data {
1da177e4 2726 struct list_head poll_list;
6e7676c1 2727 struct sk_buff_head process_queue;
1da177e4 2728
dee42870 2729 /* stats */
cd7b5396
DM
2730 unsigned int processed;
2731 unsigned int time_squeeze;
cd7b5396 2732 unsigned int received_rps;
fd793d89 2733#ifdef CONFIG_RPS
88751275 2734 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2735#endif
2736#ifdef CONFIG_NET_FLOW_LIMIT
2737 struct sd_flow_limit __rcu *flow_limit;
2738#endif
2739 struct Qdisc *output_queue;
2740 struct Qdisc **output_queue_tailp;
2741 struct sk_buff *completion_queue;
88751275 2742
4cdb1e2e 2743#ifdef CONFIG_RPS
501e7ef5
ED
2744 /* input_queue_head should be written by cpu owning this struct,
2745 * and only read by other cpus. Worth using a cache line.
2746 */
2747 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2748
2749 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2750 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2751 struct softnet_data *rps_ipi_next;
2752 unsigned int cpu;
76cc8b13 2753 unsigned int input_queue_tail;
1e94d72f 2754#endif
95c96174 2755 unsigned int dropped;
0a9627f2 2756 struct sk_buff_head input_pkt_queue;
bea3348e 2757 struct napi_struct backlog;
99bbc707 2758
1da177e4
LT
2759};
2760
76cc8b13 2761static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2762{
2763#ifdef CONFIG_RPS
76cc8b13
TH
2764 sd->input_queue_head++;
2765#endif
2766}
2767
2768static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2769 unsigned int *qtail)
2770{
2771#ifdef CONFIG_RPS
2772 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2773#endif
2774}
2775
0a9627f2 2776DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2777
f629d208 2778void __netif_schedule(struct Qdisc *q);
46e5da40 2779void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2780
fd2ea0a7
DM
2781static inline void netif_tx_schedule_all(struct net_device *dev)
2782{
2783 unsigned int i;
2784
2785 for (i = 0; i < dev->num_tx_queues; i++)
2786 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2787}
2788
f9a7cbbf 2789static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2790{
73466498 2791 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2792}
2793
bea3348e
SH
2794/**
2795 * netif_start_queue - allow transmit
2796 * @dev: network device
2797 *
2798 * Allow upper layers to call the device hard_start_xmit routine.
2799 */
1da177e4
LT
2800static inline void netif_start_queue(struct net_device *dev)
2801{
e8a0464c 2802 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2803}
2804
fd2ea0a7
DM
2805static inline void netif_tx_start_all_queues(struct net_device *dev)
2806{
2807 unsigned int i;
2808
2809 for (i = 0; i < dev->num_tx_queues; i++) {
2810 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2811 netif_tx_start_queue(txq);
2812 }
2813}
2814
46e5da40 2815void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2816
d29f749e
DJ
2817/**
2818 * netif_wake_queue - restart transmit
2819 * @dev: network device
2820 *
2821 * Allow upper layers to call the device hard_start_xmit routine.
2822 * Used for flow control when transmit resources are available.
2823 */
79d16385
DM
2824static inline void netif_wake_queue(struct net_device *dev)
2825{
e8a0464c 2826 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2827}
2828
fd2ea0a7
DM
2829static inline void netif_tx_wake_all_queues(struct net_device *dev)
2830{
2831 unsigned int i;
2832
2833 for (i = 0; i < dev->num_tx_queues; i++) {
2834 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2835 netif_tx_wake_queue(txq);
2836 }
2837}
2838
f9a7cbbf 2839static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2840{
73466498 2841 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2842}
2843
bea3348e
SH
2844/**
2845 * netif_stop_queue - stop transmitted packets
2846 * @dev: network device
2847 *
2848 * Stop upper layers calling the device hard_start_xmit routine.
2849 * Used for flow control when transmit resources are unavailable.
2850 */
1da177e4
LT
2851static inline void netif_stop_queue(struct net_device *dev)
2852{
e8a0464c 2853 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2854}
2855
a2029240 2856void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2857
4d29515f 2858static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2859{
73466498 2860 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2861}
2862
bea3348e
SH
2863/**
2864 * netif_queue_stopped - test if transmit queue is flowblocked
2865 * @dev: network device
2866 *
2867 * Test if transmit queue on device is currently unable to send.
2868 */
4d29515f 2869static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2870{
e8a0464c 2871 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2872}
2873
4d29515f 2874static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2875{
73466498
TH
2876 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2877}
2878
8e2f1a63
DB
2879static inline bool
2880netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2881{
2882 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2883}
2884
8e2f1a63
DB
2885static inline bool
2886netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2887{
2888 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2889}
2890
53511453
ED
2891/**
2892 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2893 * @dev_queue: pointer to transmit queue
2894 *
2895 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2896 * to give appropriate hint to the CPU.
53511453
ED
2897 */
2898static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2899{
2900#ifdef CONFIG_BQL
2901 prefetchw(&dev_queue->dql.num_queued);
2902#endif
2903}
2904
2905/**
2906 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2907 * @dev_queue: pointer to transmit queue
2908 *
2909 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2910 * to give appropriate hint to the CPU.
53511453
ED
2911 */
2912static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2913{
2914#ifdef CONFIG_BQL
2915 prefetchw(&dev_queue->dql.limit);
2916#endif
2917}
2918
c5d67bd7
TH
2919static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2920 unsigned int bytes)
2921{
114cf580
TH
2922#ifdef CONFIG_BQL
2923 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2924
2925 if (likely(dql_avail(&dev_queue->dql) >= 0))
2926 return;
2927
2928 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2929
2930 /*
2931 * The XOFF flag must be set before checking the dql_avail below,
2932 * because in netdev_tx_completed_queue we update the dql_completed
2933 * before checking the XOFF flag.
2934 */
2935 smp_mb();
2936
2937 /* check again in case another CPU has just made room avail */
2938 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2939 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2940#endif
c5d67bd7
TH
2941}
2942
0042d0c8
FF
2943/**
2944 * netdev_sent_queue - report the number of bytes queued to hardware
2945 * @dev: network device
2946 * @bytes: number of bytes queued to the hardware device queue
2947 *
2948 * Report the number of bytes queued for sending/completion to the network
2949 * device hardware queue. @bytes should be a good approximation and should
2950 * exactly match netdev_completed_queue() @bytes
2951 */
c5d67bd7
TH
2952static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2953{
2954 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2955}
2956
2957static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2958 unsigned int pkts, unsigned int bytes)
c5d67bd7 2959{
114cf580 2960#ifdef CONFIG_BQL
b37c0fbe
AD
2961 if (unlikely(!bytes))
2962 return;
2963
2964 dql_completed(&dev_queue->dql, bytes);
2965
2966 /*
2967 * Without the memory barrier there is a small possiblity that
2968 * netdev_tx_sent_queue will miss the update and cause the queue to
2969 * be stopped forever
2970 */
2971 smp_mb();
2972
2973 if (dql_avail(&dev_queue->dql) < 0)
2974 return;
2975
2976 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2977 netif_schedule_queue(dev_queue);
114cf580 2978#endif
c5d67bd7
TH
2979}
2980
0042d0c8
FF
2981/**
2982 * netdev_completed_queue - report bytes and packets completed by device
2983 * @dev: network device
2984 * @pkts: actual number of packets sent over the medium
2985 * @bytes: actual number of bytes sent over the medium
2986 *
2987 * Report the number of bytes and packets transmitted by the network device
2988 * hardware queue over the physical medium, @bytes must exactly match the
2989 * @bytes amount passed to netdev_sent_queue()
2990 */
c5d67bd7 2991static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2992 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2993{
2994 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2995}
2996
2997static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2998{
114cf580 2999#ifdef CONFIG_BQL
5c490354 3000 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3001 dql_reset(&q->dql);
3002#endif
c5d67bd7
TH
3003}
3004
0042d0c8
FF
3005/**
3006 * netdev_reset_queue - reset the packets and bytes count of a network device
3007 * @dev_queue: network device
3008 *
3009 * Reset the bytes and packet count of a network device and clear the
3010 * software flow control OFF bit for this network device
3011 */
c5d67bd7
TH
3012static inline void netdev_reset_queue(struct net_device *dev_queue)
3013{
3014 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3015}
3016
b9507bda
DB
3017/**
3018 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3019 * @dev: network device
3020 * @queue_index: given tx queue index
3021 *
3022 * Returns 0 if given tx queue index >= number of device tx queues,
3023 * otherwise returns the originally passed tx queue index.
3024 */
3025static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3026{
3027 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3028 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3029 dev->name, queue_index,
3030 dev->real_num_tx_queues);
3031 return 0;
3032 }
3033
3034 return queue_index;
3035}
3036
bea3348e
SH
3037/**
3038 * netif_running - test if up
3039 * @dev: network device
3040 *
3041 * Test if the device has been brought up.
3042 */
4d29515f 3043static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3044{
3045 return test_bit(__LINK_STATE_START, &dev->state);
3046}
3047
f25f4e44 3048/*
5e82b4b2 3049 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3050 * stop, and a check if it's stopped. All other device management is
3051 * done at the overall netdevice level.
3052 * Also test the device if we're multiqueue.
3053 */
bea3348e
SH
3054
3055/**
3056 * netif_start_subqueue - allow sending packets on subqueue
3057 * @dev: network device
3058 * @queue_index: sub queue index
3059 *
3060 * Start individual transmit queue of a device with multiple transmit queues.
3061 */
f25f4e44
PWJ
3062static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3063{
fd2ea0a7 3064 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3065
3066 netif_tx_start_queue(txq);
f25f4e44
PWJ
3067}
3068
bea3348e
SH
3069/**
3070 * netif_stop_subqueue - stop sending packets on subqueue
3071 * @dev: network device
3072 * @queue_index: sub queue index
3073 *
3074 * Stop individual transmit queue of a device with multiple transmit queues.
3075 */
f25f4e44
PWJ
3076static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3077{
fd2ea0a7 3078 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3079 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3080}
3081
bea3348e
SH
3082/**
3083 * netif_subqueue_stopped - test status of subqueue
3084 * @dev: network device
3085 * @queue_index: sub queue index
3086 *
3087 * Check individual transmit queue of a device with multiple transmit queues.
3088 */
4d29515f
DM
3089static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3090 u16 queue_index)
f25f4e44 3091{
fd2ea0a7 3092 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3093
3094 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3095}
3096
4d29515f
DM
3097static inline bool netif_subqueue_stopped(const struct net_device *dev,
3098 struct sk_buff *skb)
668f895a
PE
3099{
3100 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3101}
bea3348e 3102
46e5da40 3103void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 3104
537c00de 3105#ifdef CONFIG_XPS
53af53ae 3106int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3107 u16 index);
537c00de
AD
3108#else
3109static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3110 const struct cpumask *mask,
537c00de
AD
3111 u16 index)
3112{
3113 return 0;
3114}
3115#endif
3116
5605c762
JP
3117u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3118 unsigned int num_tx_queues);
3119
a3d22a68
VZ
3120/*
3121 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3122 * as a distribution range limit for the returned value.
3123 */
3124static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3125 struct sk_buff *skb)
a3d22a68
VZ
3126{
3127 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3128}
3129
bea3348e
SH
3130/**
3131 * netif_is_multiqueue - test if device has multiple transmit queues
3132 * @dev: network device
3133 *
3134 * Check if device has multiple transmit queues
bea3348e 3135 */
4d29515f 3136static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3137{
a02cec21 3138 return dev->num_tx_queues > 1;
f25f4e44 3139}
1da177e4 3140
f629d208 3141int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3142
a953be53 3143#ifdef CONFIG_SYSFS
f629d208 3144int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3145#else
3146static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3147 unsigned int rxq)
3148{
3149 return 0;
3150}
3151#endif
3152
a953be53
MD
3153#ifdef CONFIG_SYSFS
3154static inline unsigned int get_netdev_rx_queue_index(
3155 struct netdev_rx_queue *queue)
3156{
3157 struct net_device *dev = queue->dev;
3158 int index = queue - dev->_rx;
3159
3160 BUG_ON(index >= dev->num_rx_queues);
3161 return index;
3162}
3163#endif
3164
16917b87 3165#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3166int netif_get_num_default_rss_queues(void);
16917b87 3167
e6247027
ED
3168enum skb_free_reason {
3169 SKB_REASON_CONSUMED,
3170 SKB_REASON_DROPPED,
3171};
3172
3173void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3174void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3175
e6247027
ED
3176/*
3177 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3178 * interrupt context or with hardware interrupts being disabled.
3179 * (in_irq() || irqs_disabled())
3180 *
3181 * We provide four helpers that can be used in following contexts :
3182 *
3183 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3184 * replacing kfree_skb(skb)
3185 *
3186 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3187 * Typically used in place of consume_skb(skb) in TX completion path
3188 *
3189 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3190 * replacing kfree_skb(skb)
3191 *
3192 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3193 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3194 */
e6247027
ED
3195static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3196{
3197 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3198}
3199
3200static inline void dev_consume_skb_irq(struct sk_buff *skb)
3201{
3202 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3203}
3204
3205static inline void dev_kfree_skb_any(struct sk_buff *skb)
3206{
3207 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3208}
3209
3210static inline void dev_consume_skb_any(struct sk_buff *skb)
3211{
3212 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3213}
1da177e4 3214
f629d208
JP
3215int netif_rx(struct sk_buff *skb);
3216int netif_rx_ni(struct sk_buff *skb);
04eb4489 3217int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3218gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3219void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3220struct sk_buff *napi_get_frags(struct napi_struct *napi);
3221gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3222struct packet_offload *gro_find_receive_by_type(__be16 type);
3223struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3224
3225static inline void napi_free_frags(struct napi_struct *napi)
3226{
3227 kfree_skb(napi->skb);
3228 napi->skb = NULL;
3229}
3230
24b27fc4 3231bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3232int netdev_rx_handler_register(struct net_device *dev,
3233 rx_handler_func_t *rx_handler,
3234 void *rx_handler_data);
3235void netdev_rx_handler_unregister(struct net_device *dev);
3236
3237bool dev_valid_name(const char *name);
3238int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3239int dev_ethtool(struct net *net, struct ifreq *);
3240unsigned int dev_get_flags(const struct net_device *);
3241int __dev_change_flags(struct net_device *, unsigned int flags);
3242int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3243void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3244 unsigned int gchanges);
f629d208
JP
3245int dev_change_name(struct net_device *, const char *);
3246int dev_set_alias(struct net_device *, const char *, size_t);
3247int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3248int dev_set_mtu(struct net_device *, int);
3249void dev_set_group(struct net_device *, int);
3250int dev_set_mac_address(struct net_device *, struct sockaddr *);
3251int dev_change_carrier(struct net_device *, bool new_carrier);
3252int dev_get_phys_port_id(struct net_device *dev,
02637fce 3253 struct netdev_phys_item_id *ppid);
db24a904
DA
3254int dev_get_phys_port_name(struct net_device *dev,
3255 char *name, size_t len);
d746d707 3256int dev_change_proto_down(struct net_device *dev, bool proto_down);
85de8576 3257int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
55a93b3e 3258struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3259struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3260 struct netdev_queue *txq, int *ret);
a0265d28 3261int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3262int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3263bool is_skb_forwardable(const struct net_device *dev,
3264 const struct sk_buff *skb);
1da177e4 3265
4e3264d2
MKL
3266static __always_inline int ____dev_forward_skb(struct net_device *dev,
3267 struct sk_buff *skb)
3268{
3269 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3270 unlikely(!is_skb_forwardable(dev, skb))) {
3271 atomic_long_inc(&dev->rx_dropped);
3272 kfree_skb(skb);
3273 return NET_RX_DROP;
3274 }
3275
3276 skb_scrub_packet(skb, true);
3277 skb->priority = 0;
3278 return 0;
3279}
3280
74b20582
DA
3281void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3282
20380731 3283extern int netdev_budget;
1da177e4
LT
3284
3285/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3286void netdev_run_todo(void);
1da177e4 3287
bea3348e
SH
3288/**
3289 * dev_put - release reference to device
3290 * @dev: network device
3291 *
9ef4429b 3292 * Release reference to device to allow it to be freed.
bea3348e 3293 */
1da177e4
LT
3294static inline void dev_put(struct net_device *dev)
3295{
933393f5 3296 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3297}
3298
bea3348e
SH
3299/**
3300 * dev_hold - get reference to device
3301 * @dev: network device
3302 *
9ef4429b 3303 * Hold reference to device to keep it from being freed.
bea3348e 3304 */
15333061
SH
3305static inline void dev_hold(struct net_device *dev)
3306{
933393f5 3307 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3308}
1da177e4
LT
3309
3310/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3311 * and _off may be called from IRQ context, but it is caller
3312 * who is responsible for serialization of these calls.
b00055aa
SR
3313 *
3314 * The name carrier is inappropriate, these functions should really be
3315 * called netif_lowerlayer_*() because they represent the state of any
3316 * kind of lower layer not just hardware media.
1da177e4
LT
3317 */
3318
f629d208
JP
3319void linkwatch_init_dev(struct net_device *dev);
3320void linkwatch_fire_event(struct net_device *dev);
3321void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3322
bea3348e
SH
3323/**
3324 * netif_carrier_ok - test if carrier present
3325 * @dev: network device
3326 *
3327 * Check if carrier is present on device
3328 */
4d29515f 3329static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3330{
3331 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3332}
3333
f629d208 3334unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3335
f629d208 3336void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3337
f629d208 3338void netif_carrier_on(struct net_device *dev);
1da177e4 3339
f629d208 3340void netif_carrier_off(struct net_device *dev);
1da177e4 3341
bea3348e
SH
3342/**
3343 * netif_dormant_on - mark device as dormant.
3344 * @dev: network device
3345 *
3346 * Mark device as dormant (as per RFC2863).
3347 *
3348 * The dormant state indicates that the relevant interface is not
3349 * actually in a condition to pass packets (i.e., it is not 'up') but is
3350 * in a "pending" state, waiting for some external event. For "on-
3351 * demand" interfaces, this new state identifies the situation where the
3352 * interface is waiting for events to place it in the up state.
bea3348e 3353 */
b00055aa
SR
3354static inline void netif_dormant_on(struct net_device *dev)
3355{
3356 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3357 linkwatch_fire_event(dev);
3358}
3359
bea3348e
SH
3360/**
3361 * netif_dormant_off - set device as not dormant.
3362 * @dev: network device
3363 *
3364 * Device is not in dormant state.
3365 */
b00055aa
SR
3366static inline void netif_dormant_off(struct net_device *dev)
3367{
3368 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3369 linkwatch_fire_event(dev);
3370}
3371
bea3348e
SH
3372/**
3373 * netif_dormant - test if carrier present
3374 * @dev: network device
3375 *
3376 * Check if carrier is present on device
3377 */
4d29515f 3378static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3379{
3380 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3381}
3382
3383
bea3348e
SH
3384/**
3385 * netif_oper_up - test if device is operational
3386 * @dev: network device
3387 *
3388 * Check if carrier is operational
3389 */
4d29515f 3390static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3391{
b00055aa
SR
3392 return (dev->operstate == IF_OPER_UP ||
3393 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3394}
3395
bea3348e
SH
3396/**
3397 * netif_device_present - is device available or removed
3398 * @dev: network device
3399 *
3400 * Check if device has not been removed from system.
3401 */
4d29515f 3402static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3403{
3404 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3405}
3406
f629d208 3407void netif_device_detach(struct net_device *dev);
1da177e4 3408
f629d208 3409void netif_device_attach(struct net_device *dev);
1da177e4
LT
3410
3411/*
3412 * Network interface message level settings
3413 */
1da177e4
LT
3414
3415enum {
3416 NETIF_MSG_DRV = 0x0001,
3417 NETIF_MSG_PROBE = 0x0002,
3418 NETIF_MSG_LINK = 0x0004,
3419 NETIF_MSG_TIMER = 0x0008,
3420 NETIF_MSG_IFDOWN = 0x0010,
3421 NETIF_MSG_IFUP = 0x0020,
3422 NETIF_MSG_RX_ERR = 0x0040,
3423 NETIF_MSG_TX_ERR = 0x0080,
3424 NETIF_MSG_TX_QUEUED = 0x0100,
3425 NETIF_MSG_INTR = 0x0200,
3426 NETIF_MSG_TX_DONE = 0x0400,
3427 NETIF_MSG_RX_STATUS = 0x0800,
3428 NETIF_MSG_PKTDATA = 0x1000,
3429 NETIF_MSG_HW = 0x2000,
3430 NETIF_MSG_WOL = 0x4000,
3431};
3432
3433#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3434#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3435#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3436#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3437#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3438#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3439#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3440#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3441#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3442#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3443#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3444#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3445#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3446#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3447#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3448
3449static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3450{
3451 /* use default */
3452 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3453 return default_msg_enable_bits;
3454 if (debug_value == 0) /* no output */
3455 return 0;
3456 /* set low N bits */
3457 return (1 << debug_value) - 1;
3458}
3459
c773e847 3460static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3461{
c773e847
DM
3462 spin_lock(&txq->_xmit_lock);
3463 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3464}
3465
5a717f4f
MT
3466static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3467{
3468 __acquire(&txq->_xmit_lock);
3469 return true;
3470}
3471
3472static inline void __netif_tx_release(struct netdev_queue *txq)
3473{
3474 __release(&txq->_xmit_lock);
3475}
3476
fd2ea0a7
DM
3477static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3478{
3479 spin_lock_bh(&txq->_xmit_lock);
3480 txq->xmit_lock_owner = smp_processor_id();
3481}
3482
4d29515f 3483static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3484{
4d29515f 3485 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3486 if (likely(ok))
3487 txq->xmit_lock_owner = smp_processor_id();
3488 return ok;
3489}
3490
3491static inline void __netif_tx_unlock(struct netdev_queue *txq)
3492{
3493 txq->xmit_lock_owner = -1;
3494 spin_unlock(&txq->_xmit_lock);
3495}
3496
3497static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3498{
3499 txq->xmit_lock_owner = -1;
3500 spin_unlock_bh(&txq->_xmit_lock);
3501}
3502
08baf561
ED
3503static inline void txq_trans_update(struct netdev_queue *txq)
3504{
3505 if (txq->xmit_lock_owner != -1)
3506 txq->trans_start = jiffies;
3507}
3508
ba162f8e
FW
3509/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3510static inline void netif_trans_update(struct net_device *dev)
3511{
9b36627a
FW
3512 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3513
3514 if (txq->trans_start != jiffies)
3515 txq->trans_start = jiffies;
ba162f8e
FW
3516}
3517
d29f749e
DJ
3518/**
3519 * netif_tx_lock - grab network device transmit lock
3520 * @dev: network device
d29f749e
DJ
3521 *
3522 * Get network device transmit lock
3523 */
22dd7495
JHS
3524static inline void netif_tx_lock(struct net_device *dev)
3525{
e8a0464c 3526 unsigned int i;
c3f26a26 3527 int cpu;
c773e847 3528
c3f26a26
DM
3529 spin_lock(&dev->tx_global_lock);
3530 cpu = smp_processor_id();
e8a0464c
DM
3531 for (i = 0; i < dev->num_tx_queues; i++) {
3532 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3533
3534 /* We are the only thread of execution doing a
3535 * freeze, but we have to grab the _xmit_lock in
3536 * order to synchronize with threads which are in
3537 * the ->hard_start_xmit() handler and already
3538 * checked the frozen bit.
3539 */
e8a0464c 3540 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3541 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3542 __netif_tx_unlock(txq);
e8a0464c 3543 }
932ff279
HX
3544}
3545
3546static inline void netif_tx_lock_bh(struct net_device *dev)
3547{
e8a0464c
DM
3548 local_bh_disable();
3549 netif_tx_lock(dev);
932ff279
HX
3550}
3551
932ff279
HX
3552static inline void netif_tx_unlock(struct net_device *dev)
3553{
e8a0464c
DM
3554 unsigned int i;
3555
3556 for (i = 0; i < dev->num_tx_queues; i++) {
3557 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3558
c3f26a26
DM
3559 /* No need to grab the _xmit_lock here. If the
3560 * queue is not stopped for another reason, we
3561 * force a schedule.
3562 */
3563 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3564 netif_schedule_queue(txq);
c3f26a26
DM
3565 }
3566 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3567}
3568
3569static inline void netif_tx_unlock_bh(struct net_device *dev)
3570{
e8a0464c
DM
3571 netif_tx_unlock(dev);
3572 local_bh_enable();
932ff279
HX
3573}
3574
c773e847 3575#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3576 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3577 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3578 } else { \
3579 __netif_tx_acquire(txq); \
22dd7495
JHS
3580 } \
3581}
3582
5efeac44
EB
3583#define HARD_TX_TRYLOCK(dev, txq) \
3584 (((dev->features & NETIF_F_LLTX) == 0) ? \
3585 __netif_tx_trylock(txq) : \
5a717f4f 3586 __netif_tx_acquire(txq))
5efeac44 3587
c773e847 3588#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3589 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3590 __netif_tx_unlock(txq); \
5a717f4f
MT
3591 } else { \
3592 __netif_tx_release(txq); \
22dd7495
JHS
3593 } \
3594}
3595
1da177e4
LT
3596static inline void netif_tx_disable(struct net_device *dev)
3597{
fd2ea0a7 3598 unsigned int i;
c3f26a26 3599 int cpu;
fd2ea0a7 3600
c3f26a26
DM
3601 local_bh_disable();
3602 cpu = smp_processor_id();
fd2ea0a7
DM
3603 for (i = 0; i < dev->num_tx_queues; i++) {
3604 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3605
3606 __netif_tx_lock(txq, cpu);
fd2ea0a7 3607 netif_tx_stop_queue(txq);
c3f26a26 3608 __netif_tx_unlock(txq);
fd2ea0a7 3609 }
c3f26a26 3610 local_bh_enable();
1da177e4
LT
3611}
3612
e308a5d8
DM
3613static inline void netif_addr_lock(struct net_device *dev)
3614{
3615 spin_lock(&dev->addr_list_lock);
3616}
3617
2429f7ac
JP
3618static inline void netif_addr_lock_nested(struct net_device *dev)
3619{
25175ba5
VY
3620 int subclass = SINGLE_DEPTH_NESTING;
3621
3622 if (dev->netdev_ops->ndo_get_lock_subclass)
3623 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3624
3625 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3626}
3627
e308a5d8
DM
3628static inline void netif_addr_lock_bh(struct net_device *dev)
3629{
3630 spin_lock_bh(&dev->addr_list_lock);
3631}
3632
3633static inline void netif_addr_unlock(struct net_device *dev)
3634{
3635 spin_unlock(&dev->addr_list_lock);
3636}
3637
3638static inline void netif_addr_unlock_bh(struct net_device *dev)
3639{
3640 spin_unlock_bh(&dev->addr_list_lock);
3641}
3642
f001fde5 3643/*
31278e71 3644 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3645 * rcu_read_lock held.
3646 */
3647#define for_each_dev_addr(dev, ha) \
31278e71 3648 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3649
1da177e4
LT
3650/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3651
f629d208 3652void ether_setup(struct net_device *dev);
1da177e4
LT
3653
3654/* Support for loadable net-drivers */
f629d208 3655struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3656 unsigned char name_assign_type,
f629d208
JP
3657 void (*setup)(struct net_device *),
3658 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3659#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3660 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3661
c835a677
TG
3662#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3663 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3664 count)
36909ea4 3665
f629d208
JP
3666int register_netdev(struct net_device *dev);
3667void unregister_netdev(struct net_device *dev);
f001fde5 3668
22bedad3 3669/* General hardware address lists handling functions */
f629d208
JP
3670int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3671 struct netdev_hw_addr_list *from_list, int addr_len);
3672void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3673 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3674int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3675 struct net_device *dev,
3676 int (*sync)(struct net_device *, const unsigned char *),
3677 int (*unsync)(struct net_device *,
3678 const unsigned char *));
3679void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3680 struct net_device *dev,
3681 int (*unsync)(struct net_device *,
3682 const unsigned char *));
f629d208 3683void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3684
f001fde5 3685/* Functions used for device addresses handling */
f629d208
JP
3686int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3687 unsigned char addr_type);
3688int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3689 unsigned char addr_type);
f629d208
JP
3690void dev_addr_flush(struct net_device *dev);
3691int dev_addr_init(struct net_device *dev);
a748ee24
JP
3692
3693/* Functions used for unicast addresses handling */
f629d208
JP
3694int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3695int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3696int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3697int dev_uc_sync(struct net_device *to, struct net_device *from);
3698int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3699void dev_uc_unsync(struct net_device *to, struct net_device *from);
3700void dev_uc_flush(struct net_device *dev);
3701void dev_uc_init(struct net_device *dev);
f001fde5 3702
670e5b8e
AD
3703/**
3704 * __dev_uc_sync - Synchonize device's unicast list
3705 * @dev: device to sync
3706 * @sync: function to call if address should be added
3707 * @unsync: function to call if address should be removed
3708 *
3709 * Add newly added addresses to the interface, and release
3710 * addresses that have been deleted.
5e82b4b2 3711 */
670e5b8e
AD
3712static inline int __dev_uc_sync(struct net_device *dev,
3713 int (*sync)(struct net_device *,
3714 const unsigned char *),
3715 int (*unsync)(struct net_device *,
3716 const unsigned char *))
3717{
3718 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3719}
3720
3721/**
e793c0f7 3722 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3723 * @dev: device to sync
3724 * @unsync: function to call if address should be removed
3725 *
3726 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3727 */
670e5b8e
AD
3728static inline void __dev_uc_unsync(struct net_device *dev,
3729 int (*unsync)(struct net_device *,
3730 const unsigned char *))
3731{
3732 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3733}
3734
22bedad3 3735/* Functions used for multicast addresses handling */
f629d208
JP
3736int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3737int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3738int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3739int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3740int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3741int dev_mc_sync(struct net_device *to, struct net_device *from);
3742int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3743void dev_mc_unsync(struct net_device *to, struct net_device *from);
3744void dev_mc_flush(struct net_device *dev);
3745void dev_mc_init(struct net_device *dev);
f001fde5 3746
670e5b8e
AD
3747/**
3748 * __dev_mc_sync - Synchonize device's multicast list
3749 * @dev: device to sync
3750 * @sync: function to call if address should be added
3751 * @unsync: function to call if address should be removed
3752 *
3753 * Add newly added addresses to the interface, and release
3754 * addresses that have been deleted.
5e82b4b2 3755 */
670e5b8e
AD
3756static inline int __dev_mc_sync(struct net_device *dev,
3757 int (*sync)(struct net_device *,
3758 const unsigned char *),
3759 int (*unsync)(struct net_device *,
3760 const unsigned char *))
3761{
3762 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3763}
3764
3765/**
e793c0f7 3766 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3767 * @dev: device to sync
3768 * @unsync: function to call if address should be removed
3769 *
3770 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3771 */
670e5b8e
AD
3772static inline void __dev_mc_unsync(struct net_device *dev,
3773 int (*unsync)(struct net_device *,
3774 const unsigned char *))
3775{
3776 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3777}
3778
4417da66 3779/* Functions used for secondary unicast and multicast support */
f629d208
JP
3780void dev_set_rx_mode(struct net_device *dev);
3781void __dev_set_rx_mode(struct net_device *dev);
3782int dev_set_promiscuity(struct net_device *dev, int inc);
3783int dev_set_allmulti(struct net_device *dev, int inc);
3784void netdev_state_change(struct net_device *dev);
3785void netdev_notify_peers(struct net_device *dev);
3786void netdev_features_change(struct net_device *dev);
1da177e4 3787/* Load a device via the kmod */
f629d208
JP
3788void dev_load(struct net *net, const char *name);
3789struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3790 struct rtnl_link_stats64 *storage);
3791void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3792 const struct net_device_stats *netdev_stats);
eeda3fd6 3793
1da177e4 3794extern int netdev_max_backlog;
3b098e2d 3795extern int netdev_tstamp_prequeue;
1da177e4 3796extern int weight_p;
9ff162a8 3797
f629d208 3798bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3799struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3800 struct list_head **iter);
f629d208
JP
3801struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3802 struct list_head **iter);
8b5be856 3803
44a40855
VY
3804/* iterate through upper list, must be called under RCU read lock */
3805#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3806 for (iter = &(dev)->adj_list.upper, \
3807 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3808 updev; \
3809 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3810
1a3f060c
DA
3811int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3812 int (*fn)(struct net_device *upper_dev,
3813 void *data),
3814 void *data);
3815
3816bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3817 struct net_device *upper_dev);
3818
f629d208
JP
3819void *netdev_lower_get_next_private(struct net_device *dev,
3820 struct list_head **iter);
3821void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3822 struct list_head **iter);
31088a11
VF
3823
3824#define netdev_for_each_lower_private(dev, priv, iter) \
3825 for (iter = (dev)->adj_list.lower.next, \
3826 priv = netdev_lower_get_next_private(dev, &(iter)); \
3827 priv; \
3828 priv = netdev_lower_get_next_private(dev, &(iter)))
3829
3830#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3831 for (iter = &(dev)->adj_list.lower, \
3832 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3833 priv; \
3834 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3835
4085ebe8
VY
3836void *netdev_lower_get_next(struct net_device *dev,
3837 struct list_head **iter);
7ce856aa 3838
4085ebe8 3839#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3840 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3841 ldev = netdev_lower_get_next(dev, &(iter)); \
3842 ldev; \
3843 ldev = netdev_lower_get_next(dev, &(iter)))
3844
7ce856aa
JP
3845struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3846 struct list_head **iter);
3847struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3848 struct list_head **iter);
3849
1a3f060c
DA
3850int netdev_walk_all_lower_dev(struct net_device *dev,
3851 int (*fn)(struct net_device *lower_dev,
3852 void *data),
3853 void *data);
3854int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3855 int (*fn)(struct net_device *lower_dev,
3856 void *data),
3857 void *data);
3858
f629d208 3859void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3860void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3861struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3862struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3863int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3864int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3865 struct net_device *upper_dev,
29bf24af 3866 void *upper_priv, void *upper_info);
f629d208
JP
3867void netdev_upper_dev_unlink(struct net_device *dev,
3868 struct net_device *upper_dev);
5bb025fa 3869void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3870void *netdev_lower_dev_get_private(struct net_device *dev,
3871 struct net_device *lower_dev);
04d48266
JP
3872void netdev_lower_state_changed(struct net_device *lower_dev,
3873 void *lower_state_info);
18bfb924
JP
3874int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3875 struct neighbour *n);
3876void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3877 struct neighbour *n);
960fb622
ED
3878
3879/* RSS keys are 40 or 52 bytes long */
3880#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3881extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3882void netdev_rss_key_fill(void *buffer, size_t len);
3883
952fcfd0 3884int dev_get_nest_level(struct net_device *dev);
f629d208
JP
3885int skb_checksum_help(struct sk_buff *skb);
3886struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3887 netdev_features_t features, bool tx_path);
3888struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3889 netdev_features_t features);
12b0004d 3890
61bd3857
MS
3891struct netdev_bonding_info {
3892 ifslave slave;
3893 ifbond master;
3894};
3895
3896struct netdev_notifier_bonding_info {
3897 struct netdev_notifier_info info; /* must be first */
3898 struct netdev_bonding_info bonding_info;
3899};
3900
3901void netdev_bonding_info_change(struct net_device *dev,
3902 struct netdev_bonding_info *bonding_info);
3903
12b0004d
CW
3904static inline
3905struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3906{
3907 return __skb_gso_segment(skb, features, true);
3908}
53d6471c 3909__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3910
3911static inline bool can_checksum_protocol(netdev_features_t features,
3912 __be16 protocol)
3913{
c8cd0989
TH
3914 if (protocol == htons(ETH_P_FCOE))
3915 return !!(features & NETIF_F_FCOE_CRC);
3916
3917 /* Assume this is an IP checksum (not SCTP CRC) */
3918
3919 if (features & NETIF_F_HW_CSUM) {
3920 /* Can checksum everything */
3921 return true;
3922 }
3923
3924 switch (protocol) {
3925 case htons(ETH_P_IP):
3926 return !!(features & NETIF_F_IP_CSUM);
3927 case htons(ETH_P_IPV6):
3928 return !!(features & NETIF_F_IPV6_CSUM);
3929 default:
3930 return false;
3931 }
ec5f0615 3932}
12b0004d 3933
fb286bb2 3934#ifdef CONFIG_BUG
f629d208 3935void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3936#else
3937static inline void netdev_rx_csum_fault(struct net_device *dev)
3938{
3939}
3940#endif
1da177e4 3941/* rx skb timestamps */
f629d208
JP
3942void net_enable_timestamp(void);
3943void net_disable_timestamp(void);
1da177e4 3944
20380731 3945#ifdef CONFIG_PROC_FS
f629d208 3946int __init dev_proc_init(void);
900ff8c6
CW
3947#else
3948#define dev_proc_init() 0
20380731
ACM
3949#endif
3950
4798248e 3951static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3952 struct sk_buff *skb, struct net_device *dev,
3953 bool more)
4798248e 3954{
fa2dbdc2 3955 skb->xmit_more = more ? 1 : 0;
0b725a2c 3956 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3957}
3958
10b3ad8c 3959static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3960 struct netdev_queue *txq, bool more)
4798248e
DM
3961{
3962 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3963 int rc;
4798248e 3964
fa2dbdc2 3965 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3966 if (rc == NETDEV_TX_OK)
3967 txq_trans_update(txq);
3968
3969 return rc;
4798248e
DM
3970}
3971
42a2d923
LT
3972int netdev_class_create_file_ns(struct class_attribute *class_attr,
3973 const void *ns);
3974void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3975 const void *ns);
58292cbe
TH
3976
3977static inline int netdev_class_create_file(struct class_attribute *class_attr)
3978{
3979 return netdev_class_create_file_ns(class_attr, NULL);
3980}
3981
3982static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3983{
3984 netdev_class_remove_file_ns(class_attr, NULL);
3985}
b8a9787e 3986
04600794
JB
3987extern struct kobj_ns_type_operations net_ns_type_operations;
3988
f629d208 3989const char *netdev_drivername(const struct net_device *dev);
6579e57b 3990
f629d208 3991void linkwatch_run_queue(void);
20380731 3992
da08143b
MK
3993static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3994 netdev_features_t f2)
3995{
c8cd0989
TH
3996 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3997 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 3998 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 3999 else
b6a0e72a 4000 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4001 }
da08143b 4002
c8cd0989 4003 return f1 & f2;
da08143b
MK
4004}
4005
c8f44aff
MM
4006static inline netdev_features_t netdev_get_wanted_features(
4007 struct net_device *dev)
5455c699
MM
4008{
4009 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4010}
c8f44aff
MM
4011netdev_features_t netdev_increment_features(netdev_features_t all,
4012 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4013
4014/* Allow TSO being used on stacked device :
4015 * Performing the GSO segmentation before last device
4016 * is a performance improvement.
4017 */
4018static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4019 netdev_features_t mask)
4020{
4021 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4022}
4023
6cb6a27c 4024int __netdev_update_features(struct net_device *dev);
5455c699 4025void netdev_update_features(struct net_device *dev);
afe12cc8 4026void netdev_change_features(struct net_device *dev);
7f353bf2 4027
fc4a7489
PM
4028void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4029 struct net_device *dev);
4030
e38f3025
TM
4031netdev_features_t passthru_features_check(struct sk_buff *skb,
4032 struct net_device *dev,
4033 netdev_features_t features);
c1e756bf 4034netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4035
4d29515f 4036static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4037{
7b748340 4038 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4039
4040 /* check flags correspondence */
4041 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4042 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4043 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4044 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4045 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4046 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4047 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4048 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4049 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4050 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4051 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4052 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4053 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4054 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4055 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4056 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
0345e186 4057
d6b4991a 4058 return (features & feature) == feature;
576a30eb
HX
4059}
4060
4d29515f 4061static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4062{
278b2513 4063 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4064 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4065}
4066
8b86a61d 4067static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4068 netdev_features_t features)
7967168c 4069{
fc741216 4070 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4071 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4072 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4073}
4074
82cc1a7a
PWJ
4075static inline void netif_set_gso_max_size(struct net_device *dev,
4076 unsigned int size)
4077{
4078 dev->gso_max_size = size;
4079}
4080
7a7ffbab
WCC
4081static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4082 int pulled_hlen, u16 mac_offset,
4083 int mac_len)
4084{
4085 skb->protocol = protocol;
4086 skb->encapsulation = 1;
4087 skb_push(skb, pulled_hlen);
4088 skb_reset_transport_header(skb);
4089 skb->mac_header = mac_offset;
4090 skb->network_header = skb->mac_header + mac_len;
4091 skb->mac_len = mac_len;
4092}
4093
3c175784
SD
4094static inline bool netif_is_macsec(const struct net_device *dev)
4095{
4096 return dev->priv_flags & IFF_MACSEC;
4097}
4098
b618aaa9 4099static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4100{
4101 return dev->priv_flags & IFF_MACVLAN;
4102}
4103
b618aaa9 4104static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4105{
4106 return dev->priv_flags & IFF_MACVLAN_PORT;
4107}
4108
b618aaa9 4109static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4110{
4111 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4112}
4113
b618aaa9 4114static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4115{
4116 return dev->priv_flags & IFF_IPVLAN_MASTER;
4117}
4118
b618aaa9 4119static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4120{
4121 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4122}
4123
b618aaa9 4124static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4125{
4126 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4127}
4128
3bdc0eba
BG
4129static inline bool netif_supports_nofcs(struct net_device *dev)
4130{
4131 return dev->priv_flags & IFF_SUPP_NOFCS;
4132}
4133
007979ea 4134static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4135{
007979ea 4136 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4137}
4138
fee6d4c7
DA
4139static inline bool netif_is_l3_slave(const struct net_device *dev)
4140{
4141 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4142}
4143
0894ae3f
JP
4144static inline bool netif_is_bridge_master(const struct net_device *dev)
4145{
4146 return dev->priv_flags & IFF_EBRIDGE;
4147}
4148
28f9ee22
VY
4149static inline bool netif_is_bridge_port(const struct net_device *dev)
4150{
4151 return dev->priv_flags & IFF_BRIDGE_PORT;
4152}
4153
35d4e172
JP
4154static inline bool netif_is_ovs_master(const struct net_device *dev)
4155{
4156 return dev->priv_flags & IFF_OPENVSWITCH;
4157}
4158
b618aaa9 4159static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4160{
4161 return dev->priv_flags & IFF_TEAM;
4162}
4163
b618aaa9 4164static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4165{
4166 return dev->priv_flags & IFF_TEAM_PORT;
4167}
4168
b618aaa9 4169static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4170{
4171 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4172}
4173
b618aaa9 4174static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4175{
4176 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4177}
4178
d4ab4286
KJ
4179static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4180{
4181 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4182}
4183
02875878
ED
4184/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4185static inline void netif_keep_dst(struct net_device *dev)
4186{
4187 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4188}
4189
18d3df3e
PA
4190/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4191static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4192{
4193 /* TODO: reserve and use an additional IFF bit, if we get more users */
4194 return dev->priv_flags & IFF_MACSEC;
4195}
4196
505d4f73 4197extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4198
571ba423
JP
4199/* Logging, debugging and troubleshooting/diagnostic helpers. */
4200
4201/* netdev_printk helpers, similar to dev_printk */
4202
4203static inline const char *netdev_name(const struct net_device *dev)
4204{
c6f854d5
VF
4205 if (!dev->name[0] || strchr(dev->name, '%'))
4206 return "(unnamed net_device)";
571ba423
JP
4207 return dev->name;
4208}
4209
ccc7f496
VF
4210static inline const char *netdev_reg_state(const struct net_device *dev)
4211{
4212 switch (dev->reg_state) {
4213 case NETREG_UNINITIALIZED: return " (uninitialized)";
4214 case NETREG_REGISTERED: return "";
4215 case NETREG_UNREGISTERING: return " (unregistering)";
4216 case NETREG_UNREGISTERED: return " (unregistered)";
4217 case NETREG_RELEASED: return " (released)";
4218 case NETREG_DUMMY: return " (dummy)";
4219 }
4220
4221 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4222 return " (unknown)";
4223}
4224
f629d208 4225__printf(3, 4)
6ea754eb
JP
4226void netdev_printk(const char *level, const struct net_device *dev,
4227 const char *format, ...);
f629d208 4228__printf(2, 3)
6ea754eb 4229void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4230__printf(2, 3)
6ea754eb 4231void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4232__printf(2, 3)
6ea754eb 4233void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4234__printf(2, 3)
6ea754eb 4235void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4236__printf(2, 3)
6ea754eb 4237void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4238__printf(2, 3)
6ea754eb 4239void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4240__printf(2, 3)
6ea754eb 4241void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4242
8909c9ad
VK
4243#define MODULE_ALIAS_NETDEV(device) \
4244 MODULE_ALIAS("netdev-" device)
4245
b558c96f 4246#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4247#define netdev_dbg(__dev, format, args...) \
4248do { \
ffa10cb4 4249 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4250} while (0)
b558c96f
JC
4251#elif defined(DEBUG)
4252#define netdev_dbg(__dev, format, args...) \
4253 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4254#else
4255#define netdev_dbg(__dev, format, args...) \
4256({ \
4257 if (0) \
4258 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4259})
4260#endif
4261
4262#if defined(VERBOSE_DEBUG)
4263#define netdev_vdbg netdev_dbg
4264#else
4265
4266#define netdev_vdbg(dev, format, args...) \
4267({ \
4268 if (0) \
4269 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4270 0; \
4271})
4272#endif
4273
4274/*
4275 * netdev_WARN() acts like dev_printk(), but with the key difference
4276 * of using a WARN/WARN_ON to get the message out, including the
4277 * file/line information and a backtrace.
4278 */
4279#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4280 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4281 netdev_reg_state(dev), ##args)
571ba423 4282
b3d95c5c
JP
4283/* netif printk helpers, similar to netdev_printk */
4284
4285#define netif_printk(priv, type, level, dev, fmt, args...) \
4286do { \
4287 if (netif_msg_##type(priv)) \
4288 netdev_printk(level, (dev), fmt, ##args); \
4289} while (0)
4290
f45f4321
JP
4291#define netif_level(level, priv, type, dev, fmt, args...) \
4292do { \
4293 if (netif_msg_##type(priv)) \
4294 netdev_##level(dev, fmt, ##args); \
4295} while (0)
4296
b3d95c5c 4297#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4298 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4299#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4300 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4301#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4302 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4303#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4304 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4305#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4306 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4307#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4308 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4309#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4310 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4311
0053ea9c 4312#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4313#define netif_dbg(priv, type, netdev, format, args...) \
4314do { \
4315 if (netif_msg_##type(priv)) \
b5fb0a03 4316 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4317} while (0)
0053ea9c
JP
4318#elif defined(DEBUG)
4319#define netif_dbg(priv, type, dev, format, args...) \
4320 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4321#else
4322#define netif_dbg(priv, type, dev, format, args...) \
4323({ \
4324 if (0) \
4325 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4326 0; \
4327})
4328#endif
4329
4330#if defined(VERBOSE_DEBUG)
bcfcc450 4331#define netif_vdbg netif_dbg
b3d95c5c
JP
4332#else
4333#define netif_vdbg(priv, type, dev, format, args...) \
4334({ \
4335 if (0) \
a4ed89cb 4336 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4337 0; \
4338})
4339#endif
571ba423 4340
900ff8c6
CW
4341/*
4342 * The list of packet types we will receive (as opposed to discard)
4343 * and the routines to invoke.
4344 *
4345 * Why 16. Because with 16 the only overlap we get on a hash of the
4346 * low nibble of the protocol value is RARP/SNAP/X.25.
4347 *
4348 * NOTE: That is no longer true with the addition of VLAN tags. Not
4349 * sure which should go first, but I bet it won't make much
4350 * difference if we are running VLANs. The good news is that
4351 * this protocol won't be in the list unless compiled in, so
4352 * the average user (w/out VLANs) will not be adversely affected.
4353 * --BLG
4354 *
4355 * 0800 IP
4356 * 8100 802.1Q VLAN
4357 * 0001 802.3
4358 * 0002 AX.25
4359 * 0004 802.2
4360 * 8035 RARP
4361 * 0005 SNAP
4362 * 0805 X.25
4363 * 0806 ARP
4364 * 8137 IPX
4365 * 0009 Localtalk
4366 * 86DD IPv6
4367 */
4368#define PTYPE_HASH_SIZE (16)
4369#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4370
385a154c 4371#endif /* _LINUX_NETDEVICE_H */