sit: fix a double free on error path
[linux-2.6-block.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
cf85d08f 44#include <net/dsa.h>
7a6b6f51 45#ifdef CONFIG_DCB
2f90b865
AD
46#include <net/dcbnl.h>
47#endif
5bc1421e 48#include <net/netprio_cgroup.h>
a050c33f 49
a59e2ecb 50#include <linux/netdev_features.h>
77162022 51#include <linux/neighbour.h>
607ca46e 52#include <uapi/linux/netdevice.h>
61bd3857 53#include <uapi/linux/if_bonding.h>
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;
13bfff25 195extern struct static_key rfs_needed;
adc9300e
ED
196#endif
197
1da177e4
LT
198struct neighbour;
199struct neigh_parms;
200struct sk_buff;
201
f001fde5
JP
202struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
205 unsigned char type;
ccffad25
JP
206#define NETDEV_HW_ADDR_T_LAN 1
207#define NETDEV_HW_ADDR_T_SAN 2
208#define NETDEV_HW_ADDR_T_SLAVE 3
209#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 210#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 211 bool global_use;
4cd729b0 212 int sync_cnt;
8f8f103d 213 int refcount;
4543fbef 214 int synced;
f001fde5
JP
215 struct rcu_head rcu_head;
216};
217
31278e71
JP
218struct netdev_hw_addr_list {
219 struct list_head list;
220 int count;
221};
222
22bedad3
JP
223#define netdev_hw_addr_list_count(l) ((l)->count)
224#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225#define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 227
22bedad3
JP
228#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230#define netdev_for_each_uc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 232
22bedad3
JP
233#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 235#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 236 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 237
d94d9fee 238struct hh_cache {
f6b72b62 239 u16 hh_len;
5c25f686 240 u16 __pad;
3644f0ce 241 seqlock_t hh_lock;
1da177e4
LT
242
243 /* cached hardware header; allow for machine alignment needs. */
244#define HH_DATA_MOD 16
245#define HH_DATA_OFF(__len) \
5ba0eac6 246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
247#define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250};
251
5e82b4b2 252/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260#define LL_RESERVED_SPACE(dev) \
f5184d26 261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 262#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 264
3b04ddde
SH
265struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
95c96174 268 const void *saddr, unsigned int len);
3b04ddde 269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
2793a23a 274 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
275};
276
1da177e4 277/* These flag bits are private to the generic network queueing
5e82b4b2 278 * layer; they may not be explicitly referenced by any other
1da177e4
LT
279 * code.
280 */
281
d94d9fee 282enum netdev_state_t {
1da177e4
LT
283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
1da177e4 285 __LINK_STATE_NOCARRIER,
b00055aa
SR
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
1da177e4
LT
288};
289
290
291/*
5e82b4b2 292 * This structure holds boot-time configured netdevice settings. They
fe2918b0 293 * are then used in the device probing.
1da177e4
LT
294 */
295struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298};
299#define NETDEV_BOOT_SETUP_MAX 8
300
f629d208 301int __init netdev_boot_setup(char *str);
1da177e4 302
bea3348e
SH
303/*
304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 310 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
404f7c9e 317 unsigned int gro_count;
bea3348e
SH
318 int (*poll)(struct napi_struct *, int);
319#ifdef CONFIG_NETPOLL
bea3348e 320 int poll_owner;
bea3348e 321#endif
5d38a079 322 struct net_device *dev;
d565b0a1 323 struct sk_buff *gro_list;
5d38a079 324 struct sk_buff *skb;
3b47d303 325 struct hrtimer timer;
404f7c9e 326 struct list_head dev_list;
af12fa6e
ET
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
bea3348e
SH
329};
330
d94d9fee 331enum {
bea3348e 332 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 333 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
337 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
338};
339
340enum {
341 NAPIF_STATE_SCHED = (1UL << NAPI_STATE_SCHED),
342 NAPIF_STATE_DISABLE = (1UL << NAPI_STATE_DISABLE),
343 NAPIF_STATE_NPSVC = (1UL << NAPI_STATE_NPSVC),
344 NAPIF_STATE_HASHED = (1UL << NAPI_STATE_HASHED),
345 NAPIF_STATE_NO_BUSY_POLL = (1UL << NAPI_STATE_NO_BUSY_POLL),
346 NAPIF_STATE_IN_BUSY_POLL = (1UL << NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
347};
348
5b252f0c 349enum gro_result {
d1c76af9
HX
350 GRO_MERGED,
351 GRO_MERGED_FREE,
352 GRO_HELD,
353 GRO_NORMAL,
354 GRO_DROP,
355};
5b252f0c 356typedef enum gro_result gro_result_t;
d1c76af9 357
8a4eb573
JP
358/*
359 * enum rx_handler_result - Possible return values for rx_handlers.
360 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
361 * further.
362 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
363 * case skb->dev was changed by rx_handler.
364 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 365 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
366 *
367 * rx_handlers are functions called from inside __netif_receive_skb(), to do
368 * special processing of the skb, prior to delivery to protocol handlers.
369 *
370 * Currently, a net_device can only have a single rx_handler registered. Trying
371 * to register a second rx_handler will return -EBUSY.
372 *
373 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
374 * To unregister a rx_handler on a net_device, use
375 * netdev_rx_handler_unregister().
376 *
377 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
378 * do with the skb.
379 *
5e82b4b2 380 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 381 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 382 * the skb to be delivered in some other way.
8a4eb573
JP
383 *
384 * If the rx_handler changed skb->dev, to divert the skb to another
385 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
386 * new device will be called if it exists.
387 *
5e82b4b2 388 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 389 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 390 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 391 *
5e82b4b2 392 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
393 * delivered, it should return RX_HANDLER_PASS.
394 *
395 * A device without a registered rx_handler will behave as if rx_handler
396 * returned RX_HANDLER_PASS.
397 */
398
399enum rx_handler_result {
400 RX_HANDLER_CONSUMED,
401 RX_HANDLER_ANOTHER,
402 RX_HANDLER_EXACT,
403 RX_HANDLER_PASS,
404};
405typedef enum rx_handler_result rx_handler_result_t;
406typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 407
f629d208 408void __napi_schedule(struct napi_struct *n);
bc9ad166 409void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 410
4d29515f 411static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
412{
413 return test_bit(NAPI_STATE_DISABLE, &n->state);
414}
415
bea3348e 416/**
5e82b4b2
BH
417 * napi_schedule_prep - check if NAPI can be scheduled
418 * @n: NAPI context
bea3348e
SH
419 *
420 * Test if NAPI routine is already running, and if not mark
5e82b4b2 421 * it as running. This is used as a condition variable to
a0a46196
DM
422 * insure only one NAPI poll instance runs. We also make
423 * sure there is no pending NAPI disable.
bea3348e 424 */
4d29515f 425static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 426{
a0a46196
DM
427 return !napi_disable_pending(n) &&
428 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
429}
430
431/**
432 * napi_schedule - schedule NAPI poll
5e82b4b2 433 * @n: NAPI context
bea3348e
SH
434 *
435 * Schedule NAPI poll routine to be called if it is not already
436 * running.
437 */
438static inline void napi_schedule(struct napi_struct *n)
439{
440 if (napi_schedule_prep(n))
441 __napi_schedule(n);
442}
443
bc9ad166
ED
444/**
445 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 446 * @n: NAPI context
bc9ad166
ED
447 *
448 * Variant of napi_schedule(), assuming hard irqs are masked.
449 */
450static inline void napi_schedule_irqoff(struct napi_struct *n)
451{
452 if (napi_schedule_prep(n))
453 __napi_schedule_irqoff(n);
454}
455
bfe13f54 456/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 457static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
458{
459 if (napi_schedule_prep(napi)) {
460 __napi_schedule(napi);
4d29515f 461 return true;
bfe13f54 462 }
4d29515f 463 return false;
bfe13f54
RD
464}
465
364b6055
ED
466bool __napi_complete(struct napi_struct *n);
467bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
468/**
469 * napi_complete - NAPI processing complete
5e82b4b2 470 * @n: NAPI context
bea3348e
SH
471 *
472 * Mark NAPI processing as complete.
3b47d303 473 * Consider using napi_complete_done() instead.
364b6055 474 * Return false if device should avoid rearming interrupts.
bea3348e 475 */
364b6055 476static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
477{
478 return napi_complete_done(n, 0);
479}
bea3348e 480
af12fa6e
ET
481/**
482 * napi_hash_del - remove a NAPI from global table
5e82b4b2 483 * @napi: NAPI context
af12fa6e 484 *
5e82b4b2 485 * Warning: caller must observe RCU grace period
34cbe27e
ED
486 * before freeing memory containing @napi, if
487 * this function returns true.
93d05d4a 488 * Note: core networking stack automatically calls it
5e82b4b2 489 * from netif_napi_del().
93d05d4a 490 * Drivers might want to call this helper to combine all
5e82b4b2 491 * the needed RCU grace periods into a single one.
af12fa6e 492 */
34cbe27e 493bool napi_hash_del(struct napi_struct *napi);
af12fa6e 494
bea3348e
SH
495/**
496 * napi_disable - prevent NAPI from scheduling
5e82b4b2 497 * @n: NAPI context
bea3348e
SH
498 *
499 * Stop NAPI from being scheduled on this context.
500 * Waits till any outstanding processing completes.
501 */
3b47d303 502void napi_disable(struct napi_struct *n);
bea3348e
SH
503
504/**
505 * napi_enable - enable NAPI scheduling
5e82b4b2 506 * @n: NAPI context
bea3348e
SH
507 *
508 * Resume NAPI from being scheduled on this context.
509 * Must be paired with napi_disable.
510 */
511static inline void napi_enable(struct napi_struct *n)
512{
513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 514 smp_mb__before_atomic();
bea3348e 515 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 516 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
517}
518
c264c3de
SH
519/**
520 * napi_synchronize - wait until NAPI is not running
5e82b4b2 521 * @n: NAPI context
c264c3de
SH
522 *
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
526 */
527static inline void napi_synchronize(const struct napi_struct *n)
528{
facc432f
AB
529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
531 msleep(1);
532 else
533 barrier();
c264c3de 534}
c264c3de 535
d94d9fee 536enum netdev_queue_state_t {
73466498
TH
537 __QUEUE_STATE_DRV_XOFF,
538 __QUEUE_STATE_STACK_XOFF,
c3f26a26 539 __QUEUE_STATE_FROZEN,
79d16385 540};
8e2f1a63
DB
541
542#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
543#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
544#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
545
546#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 QUEUE_STATE_FROZEN)
549#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
550 QUEUE_STATE_FROZEN)
551
73466498
TH
552/*
553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
554 * netif_tx_* functions below are used to manipulate this flag. The
555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556 * queue independently. The netif_xmit_*stopped functions below are called
557 * to check if the queue has been stopped by the driver or stack (either
558 * of the XOFF bits are set in the state). Drivers should not need to call
559 * netif_xmit*stopped functions, they should only be using netif_tx_*.
560 */
79d16385 561
bb949fbd 562struct netdev_queue {
6a321cb3 563/*
5e82b4b2 564 * read-mostly part
6a321cb3 565 */
bb949fbd 566 struct net_device *dev;
46e5da40 567 struct Qdisc __rcu *qdisc;
b0e1e646 568 struct Qdisc *qdisc_sleeping;
ccf5ff69 569#ifdef CONFIG_SYSFS
1d24eb48
TH
570 struct kobject kobj;
571#endif
f2cd2d3e
ED
572#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 int numa_node;
574#endif
c0ef079c
FW
575 unsigned long tx_maxrate;
576 /*
577 * Number of TX timeouts for this queue
578 * (/sys/class/net/DEV/Q/trans_timeout)
579 */
580 unsigned long trans_timeout;
6a321cb3 581/*
5e82b4b2 582 * write-mostly part
6a321cb3
ED
583 */
584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
585 int xmit_lock_owner;
9d21493b 586 /*
9b36627a 587 * Time (in jiffies) of last Tx
9d21493b
ED
588 */
589 unsigned long trans_start;
ccf5ff69 590
114cf580
TH
591 unsigned long state;
592
593#ifdef CONFIG_BQL
594 struct dql dql;
595#endif
e8a0464c 596} ____cacheline_aligned_in_smp;
bb949fbd 597
f2cd2d3e
ED
598static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599{
600#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 return q->numa_node;
602#else
b236da69 603 return NUMA_NO_NODE;
f2cd2d3e
ED
604#endif
605}
606
607static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608{
609#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 q->numa_node = node;
611#endif
612}
613
df334545 614#ifdef CONFIG_RPS
0a9627f2
TH
615/*
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
618 */
619struct rps_map {
620 unsigned int len;
621 struct rcu_head rcu;
622 u16 cpus[0];
623};
60b778ce 624#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 625
fec5e652 626/*
c445477d
BH
627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628 * tail pointer for that CPU's input queue at the time of last enqueue, and
629 * a hardware filter index.
fec5e652
TH
630 */
631struct rps_dev_flow {
632 u16 cpu;
c445477d 633 u16 filter;
fec5e652
TH
634 unsigned int last_qtail;
635};
c445477d 636#define RPS_NO_FILTER 0xffff
fec5e652
TH
637
638/*
639 * The rps_dev_flow_table structure contains a table of flow mappings.
640 */
641struct rps_dev_flow_table {
642 unsigned int mask;
643 struct rcu_head rcu;
fec5e652
TH
644 struct rps_dev_flow flows[0];
645};
646#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 647 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
648
649/*
650 * The rps_sock_flow_table contains mappings of flows to the last CPU
651 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
652 * Each entry is a 32bit value. Upper part is the high-order bits
653 * of flow hash, lower part is CPU number.
567e4b79 654 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 657 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
658 */
659struct rps_sock_flow_table {
567e4b79 660 u32 mask;
93c1af6c
ED
661
662 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 663};
567e4b79 664#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
665
666#define RPS_NO_CPU 0xffff
667
567e4b79
ED
668extern u32 rps_cpu_mask;
669extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670
fec5e652
TH
671static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
672 u32 hash)
673{
674 if (table && hash) {
567e4b79
ED
675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
fec5e652 677
5e82b4b2 678 /* We only give a hint, preemption can change CPU under us */
567e4b79 679 val |= raw_smp_processor_id();
fec5e652 680
567e4b79
ED
681 if (table->ents[index] != val)
682 table->ents[index] = val;
fec5e652
TH
683 }
684}
685
c445477d 686#ifdef CONFIG_RFS_ACCEL
f629d208
JP
687bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
688 u16 filter_id);
c445477d 689#endif
a953be53 690#endif /* CONFIG_RPS */
c445477d 691
0a9627f2
TH
692/* This structure contains an instance of an RX queue. */
693struct netdev_rx_queue {
a953be53 694#ifdef CONFIG_RPS
6e3f7faf
ED
695 struct rps_map __rcu *rps_map;
696 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 697#endif
6e3f7faf 698 struct kobject kobj;
fe822240 699 struct net_device *dev;
0a9627f2 700} ____cacheline_aligned_in_smp;
a953be53
MD
701
702/*
703 * RX queue sysfs structures and functions.
704 */
705struct rx_queue_attribute {
706 struct attribute attr;
707 ssize_t (*show)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, char *buf);
709 ssize_t (*store)(struct netdev_rx_queue *queue,
710 struct rx_queue_attribute *attr, const char *buf, size_t len);
711};
d314774c 712
bf264145
TH
713#ifdef CONFIG_XPS
714/*
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
717 */
718struct xps_map {
719 unsigned int len;
720 unsigned int alloc_len;
721 struct rcu_head rcu;
722 u16 queues[0];
723};
60b778ce 724#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
725#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
727
728/*
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 */
731struct xps_dev_maps {
732 struct rcu_head rcu;
a4177869 733 struct xps_map __rcu *cpu_map[0];
bf264145 734};
184c449f
AD
735#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
736 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
737#endif /* CONFIG_XPS */
738
4f57c087
JF
739#define TC_MAX_QUEUE 16
740#define TC_BITMASK 15
741/* HW offloaded queuing disciplines txq count and offset maps */
742struct netdev_tc_txq {
743 u16 count;
744 u16 offset;
745};
746
68bad94e
NP
747#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748/*
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
751 */
752struct netdev_fcoe_hbainfo {
753 char manufacturer[64];
754 char serial_number[64];
755 char hardware_version[64];
756 char driver_version[64];
757 char optionrom_version[64];
758 char firmware_version[64];
759 char model[256];
760 char model_description[256];
761};
762#endif
763
02637fce 764#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 765
02637fce
JP
766/* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
66b52b0d 768 */
02637fce
JP
769struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
771 unsigned char id_len;
772};
773
d754f98b
SF
774static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
776{
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
779}
780
99932d4f
DB
781typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
783
a1b7c5fd 784/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
785 * that are being passed to the netdevice through the setup_tc op.
786 */
787enum {
788 TC_SETUP_MQPRIO,
a1b7c5fd 789 TC_SETUP_CLSU32,
5b33f488 790 TC_SETUP_CLSFLOWER,
b87f7936 791 TC_SETUP_MATCHALL,
332ae8e2 792 TC_SETUP_CLSBPF,
16e5cc64
JF
793};
794
a1b7c5fd
JF
795struct tc_cls_u32_offload;
796
16e5cc64
JF
797struct tc_to_netdev {
798 unsigned int type;
799 union {
800 u8 tc;
a1b7c5fd 801 struct tc_cls_u32_offload *cls_u32;
5b33f488 802 struct tc_cls_flower_offload *cls_flower;
b87f7936 803 struct tc_cls_matchall_offload *cls_mall;
332ae8e2 804 struct tc_cls_bpf_offload *cls_bpf;
16e5cc64 805 };
7091d8c7 806 bool egress_dev;
16e5cc64
JF
807};
808
a7862b45
BB
809/* These structures hold the attributes of xdp state that are being passed
810 * to the netdevice through the xdp op.
811 */
812enum xdp_netdev_command {
813 /* Set or clear a bpf program used in the earliest stages of packet
814 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
815 * is responsible for calling bpf_prog_put on any old progs that are
816 * stored. In case of error, the callee need not release the new prog
817 * reference, but on success it takes ownership and must bpf_prog_put
818 * when it is no longer used.
819 */
820 XDP_SETUP_PROG,
821 /* Check if a bpf program is set on the device. The callee should
822 * return true if a program is currently attached and running.
823 */
824 XDP_QUERY_PROG,
825};
826
827struct netdev_xdp {
828 enum xdp_netdev_command command;
829 union {
830 /* XDP_SETUP_PROG */
831 struct bpf_prog *prog;
832 /* XDP_QUERY_PROG */
833 bool prog_attached;
834 };
835};
16e5cc64 836
d314774c
SH
837/*
838 * This structure defines the management hooks for network devices.
00829823
SH
839 * The following hooks can be defined; unless noted otherwise, they are
840 * optional and can be filled with a null pointer.
d314774c
SH
841 *
842 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
843 * This function is called once when a network device is registered.
844 * The network device can use this for any late stage initialization
845 * or semantic validation. It can fail with an error code which will
846 * be propagated back to register_netdev.
d314774c
SH
847 *
848 * void (*ndo_uninit)(struct net_device *dev);
849 * This function is called when device is unregistered or when registration
850 * fails. It is not called if init fails.
851 *
852 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 853 * This function is called when a network device transitions to the up
d314774c
SH
854 * state.
855 *
856 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 857 * This function is called when a network device transitions to the down
d314774c
SH
858 * state.
859 *
dc1f8bf6
SH
860 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
861 * struct net_device *dev);
00829823 862 * Called when a packet needs to be transmitted.
e79d8429
RR
863 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
864 * the queue before that can happen; it's for obsolete devices and weird
865 * corner cases, but the stack really does a non-trivial amount
866 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 867 * Required; cannot be NULL.
00829823 868 *
1a2a1444
DM
869 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
870 * struct net_device *dev
871 * netdev_features_t features);
872 * Called by core transmit path to determine if device is capable of
873 * performing offload operations on a given packet. This is to give
874 * the device an opportunity to implement any restrictions that cannot
875 * be otherwise expressed by feature flags. The check is called with
876 * the set of features that the stack has calculated and it returns
877 * those the driver believes to be appropriate.
cdba756f 878 *
f663dd9a 879 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 880 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 881 * Called to decide which queue to use when device supports multiple
00829823
SH
882 * transmit queues.
883 *
d314774c
SH
884 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
885 * This function is called to allow device receiver to make
5e82b4b2 886 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
887 *
888 * void (*ndo_set_rx_mode)(struct net_device *dev);
889 * This function is called device changes address list filtering.
01789349 890 * If driver handles unicast address filtering, it should set
5e82b4b2 891 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
892 *
893 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
894 * This function is called when the Media Access Control address
37b607c5 895 * needs to be changed. If this interface is not defined, the
5e82b4b2 896 * MAC address can not be changed.
d314774c
SH
897 *
898 * int (*ndo_validate_addr)(struct net_device *dev);
899 * Test if Media Access Control address is valid for the device.
900 *
901 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
902 * Called when a user requests an ioctl which can't be handled by
903 * the generic interface code. If not defined ioctls return
d314774c
SH
904 * not supported error code.
905 *
906 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
907 * Used to set network devices bus interface parameters. This interface
5e82b4b2 908 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
909 * interface (PCI) for low level management.
910 *
911 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
912 * Called when a user wants to change the Maximum Transfer Unit
913 * of a device. If not defined, any request to change MTU will
914 * will return an error.
915 *
00829823 916 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 917 * Callback used when the transmitter has not made any progress
d314774c
SH
918 * for dev->watchdog ticks.
919 *
3cfde79c 920 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 921 * struct rtnl_link_stats64 *storage);
d308e38f 922 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 923 * Called when a user wants to get the network device usage
be1f3c2c 924 * statistics. Drivers must do one of the following:
3cfde79c
BH
925 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
926 * rtnl_link_stats64 structure passed by the caller.
82695d9b 927 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
928 * (which should normally be dev->stats) and return a pointer to
929 * it. The structure may be changed asynchronously only if each
930 * field is written atomically.
931 * 3. Update dev->stats asynchronously and atomically, and define
932 * neither operation.
d314774c 933 *
3df5b3c6 934 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
935 * Return true if this device supports offload stats of this attr_id.
936 *
937 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
938 * void *attr_data)
939 * Get statistics for offload operations by attr_id. Write it into the
940 * attr_data pointer.
941 *
5d632cb7 942 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 943 * If device supports VLAN filtering this function is called when a
80d5c368 944 * VLAN id is registered.
d314774c 945 *
5d632cb7 946 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 947 * If device supports VLAN filtering this function is called when a
80d5c368 948 * VLAN id is unregistered.
d314774c
SH
949 *
950 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
951 *
952 * SR-IOV management functions.
953 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
954 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
955 * u8 qos, __be16 proto);
ed616689
SC
956 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
957 * int max_tx_rate);
5f8444a3 958 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 959 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
960 * int (*ndo_get_vf_config)(struct net_device *dev,
961 * int vf, struct ifla_vf_info *ivf);
1d8faf48 962 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
963 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
964 * struct nlattr *port[]);
01a3d796
VZ
965 *
966 * Enable or disable the VF ability to query its RSS Redirection Table and
967 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 968 * with PF and querying it may introduce a theoretical security risk.
01a3d796 969 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 970 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
971 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
972 * Called to setup 'tc' number of traffic classes in the net device. This
973 * is always called from the stack with the rtnl lock held and netif tx
974 * queues stopped. This allows the netdevice to perform queue management
975 * safely.
c445477d 976 *
e9bce845
YZ
977 * Fiber Channel over Ethernet (FCoE) offload functions.
978 * int (*ndo_fcoe_enable)(struct net_device *dev);
979 * Called when the FCoE protocol stack wants to start using LLD for FCoE
980 * so the underlying device can perform whatever needed configuration or
981 * initialization to support acceleration of FCoE traffic.
982 *
983 * int (*ndo_fcoe_disable)(struct net_device *dev);
984 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
985 * so the underlying device can perform whatever needed clean-ups to
986 * stop supporting acceleration of FCoE traffic.
987 *
988 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
989 * struct scatterlist *sgl, unsigned int sgc);
990 * Called when the FCoE Initiator wants to initialize an I/O that
991 * is a possible candidate for Direct Data Placement (DDP). The LLD can
992 * perform necessary setup and returns 1 to indicate the device is set up
993 * successfully to perform DDP on this I/O, otherwise this returns 0.
994 *
995 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
996 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
997 * indicated by the FC exchange id 'xid', so the underlying device can
998 * clean up and reuse resources for later DDP requests.
999 *
1000 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1001 * struct scatterlist *sgl, unsigned int sgc);
1002 * Called when the FCoE Target wants to initialize an I/O that
1003 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1004 * perform necessary setup and returns 1 to indicate the device is set up
1005 * successfully to perform DDP on this I/O, otherwise this returns 0.
1006 *
68bad94e
NP
1007 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1008 * struct netdev_fcoe_hbainfo *hbainfo);
1009 * Called when the FCoE Protocol stack wants information on the underlying
1010 * device. This information is utilized by the FCoE protocol stack to
1011 * register attributes with Fiber Channel management service as per the
1012 * FC-GS Fabric Device Management Information(FDMI) specification.
1013 *
e9bce845
YZ
1014 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1015 * Called when the underlying device wants to override default World Wide
1016 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1017 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1018 * protocol stack to use.
1019 *
c445477d
BH
1020 * RFS acceleration.
1021 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1022 * u16 rxq_index, u32 flow_id);
1023 * Set hardware filter for RFS. rxq_index is the target queue index;
1024 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1025 * Return the filter ID on success, or a negative error code.
fbaec0ea 1026 *
8b98a70c 1027 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1028 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1029 * Called to make another netdev an underling.
1030 *
1031 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1032 * Called to release previously enslaved netdev.
5455c699
MM
1033 *
1034 * Feature/offload setting functions.
1a2a1444
DM
1035 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1036 * netdev_features_t features);
1037 * Adjusts the requested feature flags according to device-specific
1038 * constraints, and returns the resulting flags. Must not modify
1039 * the device state.
1040 *
c8f44aff 1041 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1042 * Called to update device configuration to new features. Passed
1043 * feature set might be less than what was returned by ndo_fix_features()).
1044 * Must return >0 or -errno if it changed dev->features itself.
1045 *
edc7d573 1046 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1047 * struct net_device *dev,
f6f6424b 1048 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1049 * Adds an FDB entry to dev for addr.
1690be63
VY
1050 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1051 * struct net_device *dev,
f6f6424b 1052 * const unsigned char *addr, u16 vid)
77162022
JF
1053 * Deletes the FDB entry from dev coresponding to addr.
1054 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1055 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1056 * int *idx)
77162022
JF
1057 * Used to add FDB entries to dump requests. Implementers should add
1058 * entries to skb and update idx with the number of entries.
e5a55a89 1059 *
ad41faa8
ND
1060 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1061 * u16 flags)
e5a55a89 1062 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1063 * struct net_device *dev, u32 filter_mask,
1064 * int nlflags)
ad41faa8
ND
1065 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1066 * u16 flags);
4bf84c35
JP
1067 *
1068 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1069 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1070 * which do not represent real hardware may define this to allow their
1071 * userspace components to manage their virtual carrier state. Devices
1072 * that determine carrier state from physical hardware properties (eg
1073 * network cables) or protocol-dependent mechanisms (eg
1074 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1075 *
1076 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1077 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1078 * Called to get ID of physical port of this device. If driver does
1079 * not implement this, it is assumed that the hw is not able to have
1080 * multiple net devices on single physical port.
53cf5275 1081 *
7c46a640
AD
1082 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1083 * struct udp_tunnel_info *ti);
1084 * Called by UDP tunnel to notify a driver about the UDP port and socket
1085 * address family that a UDP tunnel is listnening to. It is called only
1086 * when a new port starts listening. The operation is protected by the
1087 * RTNL.
1088 *
1089 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1090 * struct udp_tunnel_info *ti);
1091 * Called by UDP tunnel to notify the driver about a UDP port and socket
1092 * address family that the UDP tunnel is not listening to anymore. The
1093 * operation is protected by the RTNL.
1094 *
a6cc0cfa
JF
1095 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1096 * struct net_device *dev)
1097 * Called by upper layer devices to accelerate switching or other
1098 * station functionality into hardware. 'pdev is the lowerdev
1099 * to use for the offload and 'dev' is the net device that will
1100 * back the offload. Returns a pointer to the private structure
1101 * the upper layer will maintain.
1102 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1103 * Called by upper layer device to delete the station created
1104 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1105 * the station and priv is the structure returned by the add
1106 * operation.
1107 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1108 * struct net_device *dev,
1109 * void *priv);
1110 * Callback to use for xmit over the accelerated station. This
1111 * is used in place of ndo_start_xmit on accelerated net
1112 * devices.
822b3b2e
JF
1113 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1114 * int queue_index, u32 maxrate);
1115 * Called when a user wants to set a max-rate limitation of specific
1116 * TX queue.
a54acb3a
ND
1117 * int (*ndo_get_iflink)(const struct net_device *dev);
1118 * Called to get the iflink value of this device.
d746d707 1119 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1120 * bool proto_down);
d746d707
AK
1121 * This function is used to pass protocol port error state information
1122 * to the switch driver. The switch driver can react to the proto_down
1123 * by doing a phys down on the associated switch port.
fc4099f1
PS
1124 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1125 * This function is used to get egress tunnel information for given skb.
1126 * This is useful for retrieving outer tunnel header parameters while
1127 * sampling packet.
871b642a
PA
1128 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1129 * This function is used to specify the headroom that the skb must
1130 * consider when allocation skb during packet reception. Setting
1131 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1132 * forward. Setting a negative value resets the rx headroom to the
871b642a 1133 * default value.
a7862b45
BB
1134 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1135 * This function is used to set or query state related to XDP on the
1136 * netdevice. See definition of enum xdp_netdev_command for details.
d746d707 1137 *
d314774c
SH
1138 */
1139struct net_device_ops {
1140 int (*ndo_init)(struct net_device *dev);
1141 void (*ndo_uninit)(struct net_device *dev);
1142 int (*ndo_open)(struct net_device *dev);
1143 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1144 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1145 struct net_device *dev);
1146 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1147 struct net_device *dev,
1148 netdev_features_t features);
00829823 1149 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1150 struct sk_buff *skb,
99932d4f
DB
1151 void *accel_priv,
1152 select_queue_fallback_t fallback);
d314774c
SH
1153 void (*ndo_change_rx_flags)(struct net_device *dev,
1154 int flags);
d314774c 1155 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1156 int (*ndo_set_mac_address)(struct net_device *dev,
1157 void *addr);
d314774c 1158 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1159 int (*ndo_do_ioctl)(struct net_device *dev,
1160 struct ifreq *ifr, int cmd);
d314774c
SH
1161 int (*ndo_set_config)(struct net_device *dev,
1162 struct ifmap *map);
00829823
SH
1163 int (*ndo_change_mtu)(struct net_device *dev,
1164 int new_mtu);
1165 int (*ndo_neigh_setup)(struct net_device *dev,
1166 struct neigh_parms *);
d314774c
SH
1167 void (*ndo_tx_timeout) (struct net_device *dev);
1168
28172739
ED
1169 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1170 struct rtnl_link_stats64 *storage);
3df5b3c6 1171 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1172 int (*ndo_get_offload_stats)(int attr_id,
1173 const struct net_device *dev,
1174 void *attr_data);
d314774c
SH
1175 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1176
8e586137 1177 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1178 __be16 proto, u16 vid);
8e586137 1179 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1180 __be16 proto, u16 vid);
d314774c 1181#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1182 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1183 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1184 struct netpoll_info *info);
0e34e931 1185 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1186#endif
e0d1095a 1187#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1188 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1189#endif
95c26df8
WM
1190 int (*ndo_set_vf_mac)(struct net_device *dev,
1191 int queue, u8 *mac);
1192 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1193 int queue, u16 vlan,
1194 u8 qos, __be16 proto);
ed616689
SC
1195 int (*ndo_set_vf_rate)(struct net_device *dev,
1196 int vf, int min_tx_rate,
1197 int max_tx_rate);
5f8444a3
GR
1198 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1199 int vf, bool setting);
dd461d6a
HS
1200 int (*ndo_set_vf_trust)(struct net_device *dev,
1201 int vf, bool setting);
95c26df8
WM
1202 int (*ndo_get_vf_config)(struct net_device *dev,
1203 int vf,
1204 struct ifla_vf_info *ivf);
1d8faf48
RE
1205 int (*ndo_set_vf_link_state)(struct net_device *dev,
1206 int vf, int link_state);
3b766cd8
EBE
1207 int (*ndo_get_vf_stats)(struct net_device *dev,
1208 int vf,
1209 struct ifla_vf_stats
1210 *vf_stats);
57b61080
SF
1211 int (*ndo_set_vf_port)(struct net_device *dev,
1212 int vf,
1213 struct nlattr *port[]);
1214 int (*ndo_get_vf_port)(struct net_device *dev,
1215 int vf, struct sk_buff *skb);
cc8e27cc
EC
1216 int (*ndo_set_vf_guid)(struct net_device *dev,
1217 int vf, u64 guid,
1218 int guid_type);
01a3d796
VZ
1219 int (*ndo_set_vf_rss_query_en)(
1220 struct net_device *dev,
1221 int vf, bool setting);
16e5cc64
JF
1222 int (*ndo_setup_tc)(struct net_device *dev,
1223 u32 handle,
1224 __be16 protocol,
1225 struct tc_to_netdev *tc);
d11ead75 1226#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1227 int (*ndo_fcoe_enable)(struct net_device *dev);
1228 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1229 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1230 u16 xid,
1231 struct scatterlist *sgl,
1232 unsigned int sgc);
1233 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1234 u16 xid);
6247e086
YZ
1235 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1236 u16 xid,
1237 struct scatterlist *sgl,
1238 unsigned int sgc);
68bad94e
NP
1239 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1240 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1241#endif
1242
d11ead75 1243#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1244#define NETDEV_FCOE_WWNN 0
1245#define NETDEV_FCOE_WWPN 1
1246 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1247 u64 *wwn, int type);
4d288d57 1248#endif
3c9c36bc 1249
c445477d
BH
1250#ifdef CONFIG_RFS_ACCEL
1251 int (*ndo_rx_flow_steer)(struct net_device *dev,
1252 const struct sk_buff *skb,
1253 u16 rxq_index,
1254 u32 flow_id);
1255#endif
fbaec0ea
JP
1256 int (*ndo_add_slave)(struct net_device *dev,
1257 struct net_device *slave_dev);
1258 int (*ndo_del_slave)(struct net_device *dev,
1259 struct net_device *slave_dev);
c8f44aff
MM
1260 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1261 netdev_features_t features);
5455c699 1262 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1263 netdev_features_t features);
503eebc2
JP
1264 int (*ndo_neigh_construct)(struct net_device *dev,
1265 struct neighbour *n);
1266 void (*ndo_neigh_destroy)(struct net_device *dev,
1267 struct neighbour *n);
77162022
JF
1268
1269 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1270 struct nlattr *tb[],
77162022 1271 struct net_device *dev,
6b6e2725 1272 const unsigned char *addr,
f6f6424b 1273 u16 vid,
77162022
JF
1274 u16 flags);
1275 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1276 struct nlattr *tb[],
77162022 1277 struct net_device *dev,
f6f6424b
JP
1278 const unsigned char *addr,
1279 u16 vid);
77162022
JF
1280 int (*ndo_fdb_dump)(struct sk_buff *skb,
1281 struct netlink_callback *cb,
1282 struct net_device *dev,
5d5eacb3 1283 struct net_device *filter_dev,
d297653d 1284 int *idx);
e5a55a89
JF
1285
1286 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1287 struct nlmsghdr *nlh,
1288 u16 flags);
e5a55a89
JF
1289 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1290 u32 pid, u32 seq,
6cbdceeb 1291 struct net_device *dev,
46c264da
ND
1292 u32 filter_mask,
1293 int nlflags);
407af329 1294 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1295 struct nlmsghdr *nlh,
1296 u16 flags);
4bf84c35
JP
1297 int (*ndo_change_carrier)(struct net_device *dev,
1298 bool new_carrier);
66b52b0d 1299 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1300 struct netdev_phys_item_id *ppid);
db24a904
DA
1301 int (*ndo_get_phys_port_name)(struct net_device *dev,
1302 char *name, size_t len);
7c46a640
AD
1303 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1304 struct udp_tunnel_info *ti);
1305 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1306 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1307 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1308 struct net_device *dev);
1309 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1310 void *priv);
1311
1312 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1313 struct net_device *dev,
1314 void *priv);
25175ba5 1315 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1316 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1317 int queue_index,
1318 u32 maxrate);
a54acb3a 1319 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1320 int (*ndo_change_proto_down)(struct net_device *dev,
1321 bool proto_down);
fc4099f1
PS
1322 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1323 struct sk_buff *skb);
871b642a
PA
1324 void (*ndo_set_rx_headroom)(struct net_device *dev,
1325 int needed_headroom);
a7862b45
BB
1326 int (*ndo_xdp)(struct net_device *dev,
1327 struct netdev_xdp *xdp);
d314774c
SH
1328};
1329
7aa98047
LR
1330/**
1331 * enum net_device_priv_flags - &struct net_device priv_flags
1332 *
1333 * These are the &struct net_device, they are only set internally
1334 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1335 * userspace; this means that the order of these flags can change
7aa98047
LR
1336 * during any kernel release.
1337 *
1338 * You should have a pretty good reason to be extending these flags.
1339 *
1340 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1341 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1342 * @IFF_BONDING: bonding master or slave
7aa98047 1343 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1344 * @IFF_WAN_HDLC: WAN HDLC device
1345 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1346 * release skb->dst
1347 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1348 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1349 * @IFF_MACVLAN_PORT: device used as macvlan port
1350 * @IFF_BRIDGE_PORT: device used as bridge port
1351 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1352 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1353 * @IFF_UNICAST_FLT: Supports unicast filtering
1354 * @IFF_TEAM_PORT: device used as team port
1355 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1356 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1357 * change when it's running
1358 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1359 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1360 * underlying stacked devices
1361 * @IFF_IPVLAN_MASTER: IPvlan master device
1362 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1363 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1364 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1365 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1366 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1367 * @IFF_TEAM: device is a team device
d4ab4286 1368 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1369 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1370 * entity (i.e. the master device for bridged veth)
3c175784 1371 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1372 */
1373enum netdev_priv_flags {
1374 IFF_802_1Q_VLAN = 1<<0,
1375 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1376 IFF_BONDING = 1<<2,
1377 IFF_ISATAP = 1<<3,
1378 IFF_WAN_HDLC = 1<<4,
1379 IFF_XMIT_DST_RELEASE = 1<<5,
1380 IFF_DONT_BRIDGE = 1<<6,
1381 IFF_DISABLE_NETPOLL = 1<<7,
1382 IFF_MACVLAN_PORT = 1<<8,
1383 IFF_BRIDGE_PORT = 1<<9,
1384 IFF_OVS_DATAPATH = 1<<10,
1385 IFF_TX_SKB_SHARING = 1<<11,
1386 IFF_UNICAST_FLT = 1<<12,
1387 IFF_TEAM_PORT = 1<<13,
1388 IFF_SUPP_NOFCS = 1<<14,
1389 IFF_LIVE_ADDR_CHANGE = 1<<15,
1390 IFF_MACVLAN = 1<<16,
1391 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1392 IFF_IPVLAN_MASTER = 1<<18,
1393 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1394 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1395 IFF_NO_QUEUE = 1<<21,
1396 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1397 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1398 IFF_TEAM = 1<<24,
d4ab4286 1399 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1400 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1401 IFF_MACSEC = 1<<27,
7aa98047
LR
1402};
1403
1404#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1405#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1406#define IFF_BONDING IFF_BONDING
7aa98047 1407#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1408#define IFF_WAN_HDLC IFF_WAN_HDLC
1409#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1410#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1411#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1412#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1413#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1414#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1415#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1416#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1417#define IFF_TEAM_PORT IFF_TEAM_PORT
1418#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1419#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1420#define IFF_MACVLAN IFF_MACVLAN
02875878 1421#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1422#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1423#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1424#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1425#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1426#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1427#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1428#define IFF_TEAM IFF_TEAM
d4ab4286 1429#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1430#define IFF_MACSEC IFF_MACSEC
7aa98047 1431
536721b1
KK
1432/**
1433 * struct net_device - The DEVICE structure.
1434 * Actually, this whole structure is a big mistake. It mixes I/O
1435 * data with strictly "high-level" data, and it has to know about
1436 * almost every data structure used in the INET module.
1437 *
1438 * @name: This is the first field of the "visible" part of this structure
1439 * (i.e. as seen by users in the "Space.c" file). It is the name
1440 * of the interface.
1441 *
1442 * @name_hlist: Device name hash chain, please keep it close to name[]
1443 * @ifalias: SNMP alias
1444 * @mem_end: Shared memory end
1445 * @mem_start: Shared memory start
1446 * @base_addr: Device I/O address
1447 * @irq: Device IRQ number
1448 *
14ffbbb8
TG
1449 * @carrier_changes: Stats to monitor carrier on<->off transitions
1450 *
536721b1
KK
1451 * @state: Generic network queuing layer state, see netdev_state_t
1452 * @dev_list: The global list of network devices
5e82b4b2
BH
1453 * @napi_list: List entry used for polling NAPI devices
1454 * @unreg_list: List entry when we are unregistering the
1455 * device; see the function unregister_netdev
1456 * @close_list: List entry used when we are closing the device
62d885fe
BP
1457 * @ptype_all: Device-specific packet handlers for all protocols
1458 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1459 *
1460 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1461 * @features: Currently active device features
1462 * @hw_features: User-changeable features
1463 *
1464 * @wanted_features: User-requested features
1465 * @vlan_features: Mask of features inheritable by VLAN devices
1466 *
1467 * @hw_enc_features: Mask of features inherited by encapsulating devices
1468 * This field indicates what encapsulation
1469 * offloads the hardware is capable of doing,
1470 * and drivers will need to set them appropriately.
1471 *
1472 * @mpls_features: Mask of features inheritable by MPLS
1473 *
1474 * @ifindex: interface index
5e82b4b2 1475 * @group: The group the device belongs to
536721b1
KK
1476 *
1477 * @stats: Statistics struct, which was left as a legacy, use
1478 * rtnl_link_stats64 instead
1479 *
1480 * @rx_dropped: Dropped packets by core network,
1481 * do not use this in drivers
1482 * @tx_dropped: Dropped packets by core network,
1483 * do not use this in drivers
6e7333d3
JW
1484 * @rx_nohandler: nohandler dropped packets by core network on
1485 * inactive devices, do not use this in drivers
536721b1 1486 *
536721b1
KK
1487 * @wireless_handlers: List of functions to handle Wireless Extensions,
1488 * instead of ioctl,
1489 * see <net/iw_handler.h> for details.
1490 * @wireless_data: Instance data managed by the core of wireless extensions
1491 *
1492 * @netdev_ops: Includes several pointers to callbacks,
1493 * if one wants to override the ndo_*() functions
1494 * @ethtool_ops: Management operations
f997c55c
AA
1495 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1496 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1497 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1498 * of Layer 2 headers.
1499 *
1500 * @flags: Interface flags (a la BSD)
1501 * @priv_flags: Like 'flags' but invisible to userspace,
1502 * see if.h for the definitions
1503 * @gflags: Global flags ( kept as legacy )
1504 * @padded: How much padding added by alloc_netdev()
1505 * @operstate: RFC2863 operstate
1506 * @link_mode: Mapping policy to operstate
1507 * @if_port: Selectable AUI, TP, ...
1508 * @dma: DMA channel
1509 * @mtu: Interface MTU value
61e84623
JW
1510 * @min_mtu: Interface Minimum MTU value
1511 * @max_mtu: Interface Maximum MTU value
536721b1 1512 * @type: Interface hardware type
2793a23a 1513 * @hard_header_len: Maximum hardware header length.
536721b1
KK
1514 *
1515 * @needed_headroom: Extra headroom the hardware may need, but not in all
1516 * cases can this be guaranteed
1517 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1518 * cases can this be guaranteed. Some cases also use
1519 * LL_MAX_HEADER instead to allocate the skb
1520 *
1521 * interface address info:
1522 *
1523 * @perm_addr: Permanent hw address
1524 * @addr_assign_type: Hw address assignment type
1525 * @addr_len: Hardware address length
8626a0c8 1526 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1527 * @dev_id: Used to differentiate devices that share
1528 * the same link layer address
1529 * @dev_port: Used to differentiate devices that share
1530 * the same function
1531 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1532 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1533 * has been enabled due to the need to listen to
1534 * additional unicast addresses in a device that
1535 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1536 * @uc: unicast mac addresses
1537 * @mc: multicast mac addresses
1538 * @dev_addrs: list of device hw addresses
1539 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1540 * @promiscuity: Number of times the NIC is told to work in
1541 * promiscuous mode; if it becomes 0 the NIC will
1542 * exit promiscuous mode
536721b1
KK
1543 * @allmulti: Counter, enables or disables allmulticast mode
1544 *
1545 * @vlan_info: VLAN info
1546 * @dsa_ptr: dsa specific data
1547 * @tipc_ptr: TIPC specific data
1548 * @atalk_ptr: AppleTalk link
1549 * @ip_ptr: IPv4 specific data
1550 * @dn_ptr: DECnet specific data
1551 * @ip6_ptr: IPv6 specific data
1552 * @ax25_ptr: AX.25 specific data
1553 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1554 *
1555 * @last_rx: Time of last Rx
1556 * @dev_addr: Hw address (before bcast,
1557 * because most packets are unicast)
1558 *
1559 * @_rx: Array of RX queues
1560 * @num_rx_queues: Number of RX queues
1561 * allocated at register_netdev() time
1562 * @real_num_rx_queues: Number of RX queues currently active in device
1563 *
1564 * @rx_handler: handler for received packets
1565 * @rx_handler_data: XXX: need comments on this one
1566 * @ingress_queue: XXX: need comments on this one
1567 * @broadcast: hw bcast address
1568 *
14ffbbb8
TG
1569 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1570 * indexed by RX queue number. Assigned by driver.
1571 * This must only be set if the ndo_rx_flow_steer
1572 * operation is defined
1573 * @index_hlist: Device index hash chain
1574 *
536721b1
KK
1575 * @_tx: Array of TX queues
1576 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1577 * @real_num_tx_queues: Number of TX queues currently active in device
1578 * @qdisc: Root qdisc from userspace point of view
1579 * @tx_queue_len: Max frames per queue allowed
1580 * @tx_global_lock: XXX: need comments on this one
1581 *
1582 * @xps_maps: XXX: need comments on this one
1583 *
536721b1 1584 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1585 * the watchdog (see dev_watchdog())
536721b1
KK
1586 * @watchdog_timer: List of timers
1587 *
1588 * @pcpu_refcnt: Number of references to this device
1589 * @todo_list: Delayed register/unregister
536721b1
KK
1590 * @link_watch_list: XXX: need comments on this one
1591 *
1592 * @reg_state: Register/unregister state machine
1593 * @dismantle: Device is going to be freed
1594 * @rtnl_link_state: This enum represents the phases of creating
1595 * a new link
1596 *
1597 * @destructor: Called from unregister,
1598 * can be used to call free_netdev
1599 * @npinfo: XXX: need comments on this one
1600 * @nd_net: Network namespace this network device is inside
1601 *
1602 * @ml_priv: Mid-layer private
1603 * @lstats: Loopback statistics
1604 * @tstats: Tunnel statistics
1605 * @dstats: Dummy statistics
1606 * @vstats: Virtual ethernet statistics
1607 *
1608 * @garp_port: GARP
1609 * @mrp_port: MRP
1610 *
1611 * @dev: Class/net/name entry
1612 * @sysfs_groups: Space for optional device, statistics and wireless
1613 * sysfs groups
1614 *
1615 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1616 * @rtnl_link_ops: Rtnl_link_ops
1617 *
1618 * @gso_max_size: Maximum size of generic segmentation offload
1619 * @gso_max_segs: Maximum number of segments that can be passed to the
1620 * NIC for GSO
1621 *
1622 * @dcbnl_ops: Data Center Bridging netlink ops
1623 * @num_tc: Number of traffic classes in the net device
1624 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1625 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1626 *
1627 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1628 *
1629 * @priomap: XXX: need comments on this one
1630 * @phydev: Physical device may attach itself
1631 * for hardware timestamping
1632 *
123b3652
ED
1633 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1634 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1635 *
d746d707
AK
1636 * @proto_down: protocol port state information can be sent to the
1637 * switch driver and used to set the phys state of the
1638 * switch port.
1639 *
1da177e4
LT
1640 * FIXME: cleanup struct net_device such that network protocol info
1641 * moves out.
1642 */
1643
d94d9fee 1644struct net_device {
1da177e4 1645 char name[IFNAMSIZ];
9356b8fc 1646 struct hlist_node name_hlist;
0b815a1a 1647 char *ifalias;
1da177e4
LT
1648 /*
1649 * I/O specific fields
1650 * FIXME: Merge these and struct ifmap into one
1651 */
536721b1
KK
1652 unsigned long mem_end;
1653 unsigned long mem_start;
1654 unsigned long base_addr;
1655 int irq;
1da177e4 1656
14ffbbb8
TG
1657 atomic_t carrier_changes;
1658
1da177e4 1659 /*
536721b1
KK
1660 * Some hardware also needs these fields (state,dev_list,
1661 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1662 * part of the usual set specified in Space.c.
1663 */
1664
1da177e4
LT
1665 unsigned long state;
1666
7562f876 1667 struct list_head dev_list;
bea3348e 1668 struct list_head napi_list;
44a0873d 1669 struct list_head unreg_list;
5cde2829 1670 struct list_head close_list;
7866a621
SN
1671 struct list_head ptype_all;
1672 struct list_head ptype_specific;
2f268f12 1673
2f268f12
VF
1674 struct {
1675 struct list_head upper;
1676 struct list_head lower;
1677 } adj_list;
1678
c8f44aff 1679 netdev_features_t features;
c8f44aff 1680 netdev_features_t hw_features;
c8f44aff 1681 netdev_features_t wanted_features;
c8f44aff 1682 netdev_features_t vlan_features;
6a674e9c 1683 netdev_features_t hw_enc_features;
0d89d203 1684 netdev_features_t mpls_features;
802ab55a 1685 netdev_features_t gso_partial_features;
04ed3e74 1686
1da177e4 1687 int ifindex;
7a66bbc9 1688 int group;
1da177e4 1689
c45d286e 1690 struct net_device_stats stats;
015f0688 1691
015f0688
ED
1692 atomic_long_t rx_dropped;
1693 atomic_long_t tx_dropped;
6e7333d3 1694 atomic_long_t rx_nohandler;
1da177e4 1695
b86e0280 1696#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1697 const struct iw_handler_def *wireless_handlers;
1698 struct iw_public_data *wireless_data;
b86e0280 1699#endif
d314774c 1700 const struct net_device_ops *netdev_ops;
76fd8593 1701 const struct ethtool_ops *ethtool_ops;
4170604f 1702#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1703 const struct switchdev_ops *switchdev_ops;
4170604f 1704#endif
1b69c6d0
DA
1705#ifdef CONFIG_NET_L3_MASTER_DEV
1706 const struct l3mdev_ops *l3mdev_ops;
1707#endif
f997c55c
AA
1708#if IS_ENABLED(CONFIG_IPV6)
1709 const struct ndisc_ops *ndisc_ops;
1710#endif
1da177e4 1711
3b04ddde
SH
1712 const struct header_ops *header_ops;
1713
536721b1
KK
1714 unsigned int flags;
1715 unsigned int priv_flags;
1716
1da177e4 1717 unsigned short gflags;
536721b1 1718 unsigned short padded;
1da177e4 1719
536721b1
KK
1720 unsigned char operstate;
1721 unsigned char link_mode;
b00055aa 1722
536721b1
KK
1723 unsigned char if_port;
1724 unsigned char dma;
bdc220da 1725
536721b1 1726 unsigned int mtu;
61e84623
JW
1727 unsigned int min_mtu;
1728 unsigned int max_mtu;
536721b1
KK
1729 unsigned short type;
1730 unsigned short hard_header_len;
1da177e4 1731
f5184d26
JB
1732 unsigned short needed_headroom;
1733 unsigned short needed_tailroom;
1734
1da177e4 1735 /* Interface address info. */
536721b1
KK
1736 unsigned char perm_addr[MAX_ADDR_LEN];
1737 unsigned char addr_assign_type;
1738 unsigned char addr_len;
a0a9663d 1739 unsigned short neigh_priv_len;
536721b1
KK
1740 unsigned short dev_id;
1741 unsigned short dev_port;
ccffad25 1742 spinlock_t addr_list_lock;
14ffbbb8
TG
1743 unsigned char name_assign_type;
1744 bool uc_promisc;
536721b1
KK
1745 struct netdev_hw_addr_list uc;
1746 struct netdev_hw_addr_list mc;
1747 struct netdev_hw_addr_list dev_addrs;
1748
4c3d5e7b
ED
1749#ifdef CONFIG_SYSFS
1750 struct kset *queues_kset;
1751#endif
9d45abe1
WC
1752 unsigned int promiscuity;
1753 unsigned int allmulti;
1da177e4 1754
1da177e4 1755
5e82b4b2 1756 /* Protocol-specific pointers */
65ac6a5f 1757
d11ead75 1758#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1759 struct vlan_info __rcu *vlan_info;
65ac6a5f 1760#endif
34a430d7 1761#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1762 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1763#endif
1764#if IS_ENABLED(CONFIG_TIPC)
536721b1 1765 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1766#endif
536721b1
KK
1767 void *atalk_ptr;
1768 struct in_device __rcu *ip_ptr;
1769 struct dn_dev __rcu *dn_ptr;
1770 struct inet6_dev __rcu *ip6_ptr;
1771 void *ax25_ptr;
1772 struct wireless_dev *ieee80211_ptr;
98a18b6f 1773 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1774#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1775 struct mpls_dev __rcu *mpls_ptr;
1776#endif
1da177e4 1777
9356b8fc 1778/*
cd13539b 1779 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1780 */
536721b1 1781 unsigned long last_rx;
4dc89133 1782
9356b8fc 1783 /* Interface address info used in eth_type_trans() */
536721b1 1784 unsigned char *dev_addr;
f001fde5 1785
a953be53 1786#ifdef CONFIG_SYSFS
0a9627f2
TH
1787 struct netdev_rx_queue *_rx;
1788
0a9627f2 1789 unsigned int num_rx_queues;
62fe0b40 1790 unsigned int real_num_rx_queues;
df334545 1791#endif
0a9627f2 1792
3b47d303 1793 unsigned long gro_flush_timeout;
61391cde 1794 rx_handler_func_t __rcu *rx_handler;
1795 void __rcu *rx_handler_data;
e8a0464c 1796
4cda01e8 1797#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1798 struct tcf_proto __rcu *ingress_cl_list;
1799#endif
24824a09 1800 struct netdev_queue __rcu *ingress_queue;
e687ad60 1801#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1802 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1803#endif
d2788d34 1804
536721b1 1805 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1806#ifdef CONFIG_RFS_ACCEL
1807 struct cpu_rmap *rx_cpu_rmap;
1808#endif
1809 struct hlist_node index_hlist;
cd13539b
ED
1810
1811/*
1812 * Cache lines mostly used on transmit path
1813 */
e8a0464c
DM
1814 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1815 unsigned int num_tx_queues;
fd2ea0a7 1816 unsigned int real_num_tx_queues;
af356afa 1817 struct Qdisc *qdisc;
59cc1f61
JK
1818#ifdef CONFIG_NET_SCHED
1819 DECLARE_HASHTABLE (qdisc_hash, 4);
1820#endif
536721b1 1821 unsigned long tx_queue_len;
c3f26a26 1822 spinlock_t tx_global_lock;
14ffbbb8 1823 int watchdog_timeo;
cd13539b 1824
bf264145 1825#ifdef CONFIG_XPS
a4177869 1826 struct xps_dev_maps __rcu *xps_maps;
bf264145 1827#endif
1f211a1b
DB
1828#ifdef CONFIG_NET_CLS_ACT
1829 struct tcf_proto __rcu *egress_cl_list;
1830#endif
0c4f691f 1831
9356b8fc 1832 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1833 struct timer_list watchdog_timer;
1834
29b4433d 1835 int __percpu *pcpu_refcnt;
1da177e4 1836 struct list_head todo_list;
1da177e4 1837
e014debe 1838 struct list_head link_watch_list;
572a103d 1839
1da177e4 1840 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1841 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1842 NETREG_UNREGISTERING, /* called unregister_netdevice */
1843 NETREG_UNREGISTERED, /* completed unregister todo */
1844 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1845 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1846 } reg_state:8;
1847
536721b1 1848 bool dismantle;
a2835763
PM
1849
1850 enum {
1851 RTNL_LINK_INITIALIZED,
1852 RTNL_LINK_INITIALIZING,
1853 } rtnl_link_state:16;
1da177e4 1854
d314774c 1855 void (*destructor)(struct net_device *dev);
1da177e4 1856
1da177e4 1857#ifdef CONFIG_NETPOLL
5fbee843 1858 struct netpoll_info __rcu *npinfo;
1da177e4 1859#endif
eae792b7 1860
0c5c9fb5 1861 possible_net_t nd_net;
4a1c5371 1862
4951704b 1863 /* mid-layer private */
a7855c78 1864 union {
536721b1
KK
1865 void *ml_priv;
1866 struct pcpu_lstats __percpu *lstats;
8f84985f 1867 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1868 struct pcpu_dstats __percpu *dstats;
1869 struct pcpu_vstats __percpu *vstats;
a7855c78 1870 };
536721b1 1871
3cc77ec7 1872 struct garp_port __rcu *garp_port;
febf018d 1873 struct mrp_port __rcu *mrp_port;
1da177e4 1874
5e82b4b2 1875 struct device dev;
0c509a6c 1876 const struct attribute_group *sysfs_groups[4];
a953be53 1877 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1878
38f7b870 1879 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1880
82cc1a7a
PWJ
1881 /* for setting kernel sock attribute on TCP connection setup */
1882#define GSO_MAX_SIZE 65536
1883 unsigned int gso_max_size;
30b678d8
BH
1884#define GSO_MAX_SEGS 65535
1885 u16 gso_max_segs;
743b03a8 1886
7a6b6f51 1887#ifdef CONFIG_DCB
32953543 1888 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1889#endif
5e82b4b2
BH
1890 u8 num_tc;
1891 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1892 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1893
d11ead75 1894#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1895 unsigned int fcoe_ddp_xid;
5bc1421e 1896#endif
86f8515f 1897#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1898 struct netprio_map __rcu *priomap;
4d288d57 1899#endif
5e82b4b2
BH
1900 struct phy_device *phydev;
1901 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1902 struct lock_class_key *qdisc_running_key;
5e82b4b2 1903 bool proto_down;
1da177e4 1904};
43cb76d9 1905#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1906
1907#define NETDEV_ALIGN 32
1da177e4 1908
4f57c087
JF
1909static inline
1910int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1911{
1912 return dev->prio_tc_map[prio & TC_BITMASK];
1913}
1914
1915static inline
1916int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1917{
1918 if (tc >= dev->num_tc)
1919 return -EINVAL;
1920
1921 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1922 return 0;
1923}
1924
8d059b0f 1925int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1926void netdev_reset_tc(struct net_device *dev);
1927int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1928int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1929
1930static inline
1931int netdev_get_num_tc(struct net_device *dev)
1932{
1933 return dev->num_tc;
1934}
1935
e8a0464c
DM
1936static inline
1937struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1938 unsigned int index)
1939{
1940 return &dev->_tx[index];
1941}
1942
10c51b56
DB
1943static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1944 const struct sk_buff *skb)
1945{
1946 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1947}
1948
e8a0464c
DM
1949static inline void netdev_for_each_tx_queue(struct net_device *dev,
1950 void (*f)(struct net_device *,
1951 struct netdev_queue *,
1952 void *),
1953 void *arg)
1954{
1955 unsigned int i;
1956
1957 for (i = 0; i < dev->num_tx_queues; i++)
1958 f(dev, &dev->_tx[i], arg);
1959}
1960
d3fff6c4
ED
1961#define netdev_lockdep_set_classes(dev) \
1962{ \
1963 static struct lock_class_key qdisc_tx_busylock_key; \
1964 static struct lock_class_key qdisc_running_key; \
1965 static struct lock_class_key qdisc_xmit_lock_key; \
1966 static struct lock_class_key dev_addr_list_lock_key; \
1967 unsigned int i; \
1968 \
1969 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1970 (dev)->qdisc_running_key = &qdisc_running_key; \
1971 lockdep_set_class(&(dev)->addr_list_lock, \
1972 &dev_addr_list_lock_key); \
1973 for (i = 0; i < (dev)->num_tx_queues; i++) \
1974 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1975 &qdisc_xmit_lock_key); \
1976}
1977
f629d208 1978struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1979 struct sk_buff *skb,
1980 void *accel_priv);
8c4c49df 1981
871b642a
PA
1982/* returns the headroom that the master device needs to take in account
1983 * when forwarding to this dev
1984 */
1985static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1986{
1987 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1988}
1989
1990static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1991{
1992 if (dev->netdev_ops->ndo_set_rx_headroom)
1993 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1994}
1995
1996/* set the device rx headroom to the dev's default */
1997static inline void netdev_reset_rx_headroom(struct net_device *dev)
1998{
1999 netdev_set_rx_headroom(dev, -1);
2000}
2001
c346dca1
YH
2002/*
2003 * Net namespace inlines
2004 */
2005static inline
2006struct net *dev_net(const struct net_device *dev)
2007{
c2d9ba9b 2008 return read_pnet(&dev->nd_net);
c346dca1
YH
2009}
2010
2011static inline
f5aa23fd 2012void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2013{
0c5c9fb5 2014 write_pnet(&dev->nd_net, net);
c346dca1
YH
2015}
2016
3e8a72d1 2017static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 2018{
3fc88677 2019#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
2020 if (dev->dsa_ptr != NULL)
2021 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 2022#endif
5aed85ce 2023 return false;
396138f0
LB
2024}
2025
bea3348e
SH
2026/**
2027 * netdev_priv - access network device private data
2028 * @dev: network device
2029 *
2030 * Get network device private data
2031 */
6472ce60 2032static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2033{
1ce8e7b5 2034 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2035}
2036
1da177e4
LT
2037/* Set the sysfs physical device reference for the network logical device
2038 * if set prior to registration will cause a symlink during initialization.
2039 */
43cb76d9 2040#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2041
384912ed 2042/* Set the sysfs device type for the network logical device to allow
3f79410c 2043 * fine-grained identification of different network device types. For
5e82b4b2 2044 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2045 */
2046#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2047
82dc3c63
ED
2048/* Default NAPI poll() weight
2049 * Device drivers are strongly advised to not use bigger value
2050 */
2051#define NAPI_POLL_WEIGHT 64
2052
3b582cc1 2053/**
5e82b4b2 2054 * netif_napi_add - initialize a NAPI context
3b582cc1 2055 * @dev: network device
5e82b4b2 2056 * @napi: NAPI context
3b582cc1
SH
2057 * @poll: polling function
2058 * @weight: default weight
2059 *
5e82b4b2
BH
2060 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2061 * *any* of the other NAPI-related functions.
3b582cc1 2062 */
d565b0a1
HX
2063void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2064 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2065
d64b5e85 2066/**
5e82b4b2 2067 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2068 * @dev: network device
5e82b4b2 2069 * @napi: NAPI context
d64b5e85
ED
2070 * @poll: polling function
2071 * @weight: default weight
2072 *
2073 * This variant of netif_napi_add() should be used from drivers using NAPI
2074 * to exclusively poll a TX queue.
2075 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2076 */
2077static inline void netif_tx_napi_add(struct net_device *dev,
2078 struct napi_struct *napi,
2079 int (*poll)(struct napi_struct *, int),
2080 int weight)
2081{
2082 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2083 netif_napi_add(dev, napi, poll, weight);
2084}
2085
d8156534 2086/**
5e82b4b2
BH
2087 * netif_napi_del - remove a NAPI context
2088 * @napi: NAPI context
d8156534 2089 *
5e82b4b2 2090 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2091 */
d565b0a1
HX
2092void netif_napi_del(struct napi_struct *napi);
2093
2094struct napi_gro_cb {
78a478d0 2095 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2096 void *frag0;
78a478d0 2097
7489594c
HX
2098 /* Length of frag0. */
2099 unsigned int frag0_len;
2100
86911732 2101 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2102 int data_offset;
86911732 2103
d565b0a1 2104 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2105 u16 flush;
2106
2107 /* Save the IP ID here and check when we get to the transport layer */
2108 u16 flush_id;
d565b0a1
HX
2109
2110 /* Number of segments aggregated. */
2e71a6f8
ED
2111 u16 count;
2112
15e2396d
TH
2113 /* Start offset for remote checksum offload */
2114 u16 gro_remcsum_start;
2115
2e71a6f8
ED
2116 /* jiffies when first packet was created/queued */
2117 unsigned long age;
86347245 2118
afe93325 2119 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2120 u16 proto;
2121
baa32ff4
TH
2122 /* This is non-zero if the packet may be of the same flow. */
2123 u8 same_flow:1;
2124
fac8e0f5
JG
2125 /* Used in tunnel GRO receive */
2126 u8 encap_mark:1;
573e8fca
TH
2127
2128 /* GRO checksum is valid */
2129 u8 csum_valid:1;
2130
662880f4
TH
2131 /* Number of checksums via CHECKSUM_UNNECESSARY */
2132 u8 csum_cnt:3;
c3c7c254 2133
baa32ff4
TH
2134 /* Free the skb? */
2135 u8 free:2;
2136#define NAPI_GRO_FREE 1
2137#define NAPI_GRO_FREE_STOLEN_HEAD 2
2138
efc98d08
TH
2139 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2140 u8 is_ipv6:1;
2141
a0ca153f
AD
2142 /* Used in GRE, set in fou/gue_gro_receive */
2143 u8 is_fou:1;
2144
1530545e
AD
2145 /* Used to determine if flush_id can be ignored */
2146 u8 is_atomic:1;
2147
fcd91dd4
SD
2148 /* Number of gro_receive callbacks this packet already went through */
2149 u8 recursion_counter:4;
2150
2151 /* 1 bit hole */
baa32ff4 2152
bf5a755f
JC
2153 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2154 __wsum csum;
2155
c3c7c254
ED
2156 /* used in skb_gro_receive() slow path */
2157 struct sk_buff *last;
d565b0a1
HX
2158};
2159
2160#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2161
fcd91dd4
SD
2162#define GRO_RECURSION_LIMIT 15
2163static inline int gro_recursion_inc_test(struct sk_buff *skb)
2164{
2165 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2166}
2167
2168typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2169static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2170 struct sk_buff **head,
2171 struct sk_buff *skb)
2172{
2173 if (unlikely(gro_recursion_inc_test(skb))) {
2174 NAPI_GRO_CB(skb)->flush |= 1;
2175 return NULL;
2176 }
2177
2178 return cb(head, skb);
2179}
2180
2181typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2182 struct sk_buff *);
2183static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2184 struct sock *sk,
2185 struct sk_buff **head,
2186 struct sk_buff *skb)
2187{
2188 if (unlikely(gro_recursion_inc_test(skb))) {
2189 NAPI_GRO_CB(skb)->flush |= 1;
2190 return NULL;
2191 }
2192
2193 return cb(sk, head, skb);
2194}
2195
1da177e4 2196struct packet_type {
f2ccd8fa
DM
2197 __be16 type; /* This is really htons(ether_type). */
2198 struct net_device *dev; /* NULL is wildcarded here */
2199 int (*func) (struct sk_buff *,
2200 struct net_device *,
2201 struct packet_type *,
2202 struct net_device *);
c0de08d0
EL
2203 bool (*id_match)(struct packet_type *ptype,
2204 struct sock *sk);
1da177e4
LT
2205 void *af_packet_priv;
2206 struct list_head list;
2207};
2208
f191a1d1 2209struct offload_callbacks {
576a30eb 2210 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2211 netdev_features_t features);
d565b0a1 2212 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2213 struct sk_buff *skb);
299603e8 2214 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2215};
2216
2217struct packet_offload {
2218 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2219 u16 priority;
f191a1d1
VY
2220 struct offload_callbacks callbacks;
2221 struct list_head list;
1da177e4
LT
2222};
2223
5e82b4b2 2224/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2225struct pcpu_sw_netstats {
2226 u64 rx_packets;
2227 u64 rx_bytes;
2228 u64 tx_packets;
2229 u64 tx_bytes;
2230 struct u64_stats_sync syncp;
2231};
2232
aabc92bb
PNA
2233#define __netdev_alloc_pcpu_stats(type, gfp) \
2234({ \
2235 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2236 if (pcpu_stats) { \
2237 int __cpu; \
2238 for_each_possible_cpu(__cpu) { \
2239 typeof(type) *stat; \
2240 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2241 u64_stats_init(&stat->syncp); \
2242 } \
2243 } \
2244 pcpu_stats; \
1c213bd2
WC
2245})
2246
aabc92bb 2247#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2248 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2249
764f5e54
JP
2250enum netdev_lag_tx_type {
2251 NETDEV_LAG_TX_TYPE_UNKNOWN,
2252 NETDEV_LAG_TX_TYPE_RANDOM,
2253 NETDEV_LAG_TX_TYPE_BROADCAST,
2254 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2255 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2256 NETDEV_LAG_TX_TYPE_HASH,
2257};
2258
2259struct netdev_lag_upper_info {
2260 enum netdev_lag_tx_type tx_type;
2261};
2262
fb1b2e3c
JP
2263struct netdev_lag_lower_state_info {
2264 u8 link_up : 1,
2265 tx_enabled : 1;
2266};
2267
1da177e4
LT
2268#include <linux/notifier.h>
2269
dcfe1421
AW
2270/* netdevice notifier chain. Please remember to update the rtnetlink
2271 * notification exclusion list in rtnetlink_event() when adding new
2272 * types.
2273 */
2274#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2275#define NETDEV_DOWN 0x0002
2276#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2277 detected a hardware crash and restarted
2278 - we can use this eg to kick tcp sessions
2279 once done */
2280#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2281#define NETDEV_REGISTER 0x0005
2282#define NETDEV_UNREGISTER 0x0006
1d486bfb 2283#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2284#define NETDEV_CHANGEADDR 0x0008
2285#define NETDEV_GOING_DOWN 0x0009
2286#define NETDEV_CHANGENAME 0x000A
2287#define NETDEV_FEAT_CHANGE 0x000B
2288#define NETDEV_BONDING_FAILOVER 0x000C
2289#define NETDEV_PRE_UP 0x000D
2290#define NETDEV_PRE_TYPE_CHANGE 0x000E
2291#define NETDEV_POST_TYPE_CHANGE 0x000F
2292#define NETDEV_POST_INIT 0x0010
0115e8e3 2293#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2294#define NETDEV_RELEASE 0x0012
2295#define NETDEV_NOTIFY_PEERS 0x0013
2296#define NETDEV_JOIN 0x0014
42e52bf9 2297#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2298#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2299#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2300#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2301#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2302#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2303#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2304#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2305#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2306
f629d208
JP
2307int register_netdevice_notifier(struct notifier_block *nb);
2308int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2309
2310struct netdev_notifier_info {
2311 struct net_device *dev;
2312};
2313
be9efd36
JP
2314struct netdev_notifier_change_info {
2315 struct netdev_notifier_info info; /* must be first */
2316 unsigned int flags_changed;
2317};
2318
0e4ead9d
JP
2319struct netdev_notifier_changeupper_info {
2320 struct netdev_notifier_info info; /* must be first */
2321 struct net_device *upper_dev; /* new upper dev */
2322 bool master; /* is upper dev master */
5e82b4b2 2323 bool linking; /* is the notification for link or unlink */
29bf24af 2324 void *upper_info; /* upper dev info */
0e4ead9d
JP
2325};
2326
04d48266
JP
2327struct netdev_notifier_changelowerstate_info {
2328 struct netdev_notifier_info info; /* must be first */
2329 void *lower_state_info; /* is lower dev state */
2330};
2331
75538c2b
CW
2332static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2333 struct net_device *dev)
2334{
2335 info->dev = dev;
2336}
2337
351638e7
JP
2338static inline struct net_device *
2339netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2340{
2341 return info->dev;
2342}
2343
f629d208 2344int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2345
2346
1da177e4
LT
2347extern rwlock_t dev_base_lock; /* Device list lock */
2348
881d966b
EB
2349#define for_each_netdev(net, d) \
2350 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2351#define for_each_netdev_reverse(net, d) \
2352 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2353#define for_each_netdev_rcu(net, d) \
2354 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2355#define for_each_netdev_safe(net, d, n) \
2356 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2357#define for_each_netdev_continue(net, d) \
2358 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2359#define for_each_netdev_continue_rcu(net, d) \
2360 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2361#define for_each_netdev_in_bond_rcu(bond, slave) \
2362 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2363 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2364#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2365
a050c33f
DL
2366static inline struct net_device *next_net_device(struct net_device *dev)
2367{
2368 struct list_head *lh;
2369 struct net *net;
2370
c346dca1 2371 net = dev_net(dev);
a050c33f
DL
2372 lh = dev->dev_list.next;
2373 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2374}
2375
ce81b76a
ED
2376static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2377{
2378 struct list_head *lh;
2379 struct net *net;
2380
2381 net = dev_net(dev);
ccf43438 2382 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2383 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2384}
2385
a050c33f
DL
2386static inline struct net_device *first_net_device(struct net *net)
2387{
2388 return list_empty(&net->dev_base_head) ? NULL :
2389 net_device_entry(net->dev_base_head.next);
2390}
7562f876 2391
ccf43438
ED
2392static inline struct net_device *first_net_device_rcu(struct net *net)
2393{
2394 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2395
2396 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2397}
2398
f629d208
JP
2399int netdev_boot_setup_check(struct net_device *dev);
2400unsigned long netdev_boot_base(const char *prefix, int unit);
2401struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2402 const char *hwaddr);
2403struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2404struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2405void dev_add_pack(struct packet_type *pt);
2406void dev_remove_pack(struct packet_type *pt);
2407void __dev_remove_pack(struct packet_type *pt);
2408void dev_add_offload(struct packet_offload *po);
2409void dev_remove_offload(struct packet_offload *po);
f629d208 2410
a54acb3a 2411int dev_get_iflink(const struct net_device *dev);
fc4099f1 2412int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2413struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2414 unsigned short mask);
f629d208
JP
2415struct net_device *dev_get_by_name(struct net *net, const char *name);
2416struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2417struct net_device *__dev_get_by_name(struct net *net, const char *name);
2418int dev_alloc_name(struct net_device *dev, const char *name);
2419int dev_open(struct net_device *dev);
2420int dev_close(struct net_device *dev);
99c4a26a 2421int dev_close_many(struct list_head *head, bool unlink);
f629d208 2422void dev_disable_lro(struct net_device *dev);
0c4b51f0 2423int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2424int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2425int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2426int register_netdevice(struct net_device *dev);
2427void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2428void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2429static inline void unregister_netdevice(struct net_device *dev)
2430{
2431 unregister_netdevice_queue(dev, NULL);
2432}
2433
f629d208
JP
2434int netdev_refcnt_read(const struct net_device *dev);
2435void free_netdev(struct net_device *dev);
74d332c1 2436void netdev_freemem(struct net_device *dev);
f629d208
JP
2437void synchronize_net(void);
2438int init_dummy_netdev(struct net_device *dev);
937f1ba5 2439
f60e5990 2440DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2441#define XMIT_RECURSION_LIMIT 10
2442
f60e5990 2443static inline int dev_recursion_level(void)
2444{
2445 return this_cpu_read(xmit_recursion);
2446}
2447
f629d208
JP
2448struct net_device *dev_get_by_index(struct net *net, int ifindex);
2449struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2450struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2451int netdev_get_name(struct net *net, char *name, int ifindex);
2452int dev_restart(struct net_device *dev);
f629d208 2453int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2454
2455static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2456{
2457 return NAPI_GRO_CB(skb)->data_offset;
2458}
2459
2460static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2461{
2462 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2463}
2464
2465static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2466{
2467 NAPI_GRO_CB(skb)->data_offset += len;
2468}
2469
a5b1cf28
HX
2470static inline void *skb_gro_header_fast(struct sk_buff *skb,
2471 unsigned int offset)
86911732 2472{
a5b1cf28
HX
2473 return NAPI_GRO_CB(skb)->frag0 + offset;
2474}
78a478d0 2475
a5b1cf28
HX
2476static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2477{
2478 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2479}
78a478d0 2480
57ea52a8
HX
2481static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2482{
2483 NAPI_GRO_CB(skb)->frag0 = NULL;
2484 NAPI_GRO_CB(skb)->frag0_len = 0;
2485}
2486
a5b1cf28
HX
2487static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2488 unsigned int offset)
2489{
17dd759c
HX
2490 if (!pskb_may_pull(skb, hlen))
2491 return NULL;
2492
57ea52a8 2493 skb_gro_frag0_invalidate(skb);
17dd759c 2494 return skb->data + offset;
86911732 2495}
1da177e4 2496
36e7b1b8
HX
2497static inline void *skb_gro_network_header(struct sk_buff *skb)
2498{
78d3fd0b
HX
2499 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2500 skb_network_offset(skb);
36e7b1b8
HX
2501}
2502
bf5a755f
JC
2503static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2504 const void *start, unsigned int len)
2505{
573e8fca 2506 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2507 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2508 csum_partial(start, len, 0));
2509}
2510
573e8fca
TH
2511/* GRO checksum functions. These are logical equivalents of the normal
2512 * checksum functions (in skbuff.h) except that they operate on the GRO
2513 * offsets and fields in sk_buff.
2514 */
2515
2516__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2517
15e2396d
TH
2518static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2519{
b7fe10e5 2520 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2521}
2522
573e8fca
TH
2523static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2524 bool zero_okay,
2525 __sum16 check)
2526{
6edec0e6
TH
2527 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2528 skb_checksum_start_offset(skb) <
2529 skb_gro_offset(skb)) &&
15e2396d 2530 !skb_at_gro_remcsum_start(skb) &&
662880f4 2531 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2532 (!zero_okay || check));
2533}
2534
2535static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2536 __wsum psum)
2537{
2538 if (NAPI_GRO_CB(skb)->csum_valid &&
2539 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2540 return 0;
2541
2542 NAPI_GRO_CB(skb)->csum = psum;
2543
2544 return __skb_gro_checksum_complete(skb);
2545}
2546
573e8fca
TH
2547static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2548{
662880f4
TH
2549 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2550 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2551 NAPI_GRO_CB(skb)->csum_cnt--;
2552 } else {
2553 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2554 * verified a new top level checksum or an encapsulated one
2555 * during GRO. This saves work if we fallback to normal path.
2556 */
2557 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2558 }
2559}
2560
2561#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2562 compute_pseudo) \
2563({ \
2564 __sum16 __ret = 0; \
2565 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2566 __ret = __skb_gro_checksum_validate_complete(skb, \
2567 compute_pseudo(skb, proto)); \
5a212329
TH
2568 if (__ret) \
2569 __skb_mark_checksum_bad(skb); \
2570 else \
573e8fca
TH
2571 skb_gro_incr_csum_unnecessary(skb); \
2572 __ret; \
2573})
2574
2575#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2576 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2577
2578#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2579 compute_pseudo) \
2580 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2581
2582#define skb_gro_checksum_simple_validate(skb) \
2583 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2584
d96535a1
TH
2585static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2586{
2587 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2588 !NAPI_GRO_CB(skb)->csum_valid);
2589}
2590
2591static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2592 __sum16 check, __wsum pseudo)
2593{
2594 NAPI_GRO_CB(skb)->csum = ~pseudo;
2595 NAPI_GRO_CB(skb)->csum_valid = 1;
2596}
2597
2598#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2599do { \
2600 if (__skb_gro_checksum_convert_check(skb)) \
2601 __skb_gro_checksum_convert(skb, check, \
2602 compute_pseudo(skb, proto)); \
2603} while (0)
2604
26c4f7da
TH
2605struct gro_remcsum {
2606 int offset;
2607 __wsum delta;
2608};
2609
2610static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2611{
846cd667 2612 grc->offset = 0;
26c4f7da
TH
2613 grc->delta = 0;
2614}
2615
b7fe10e5
TH
2616static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2617 unsigned int off, size_t hdrlen,
2618 int start, int offset,
2619 struct gro_remcsum *grc,
2620 bool nopartial)
dcdc8994
TH
2621{
2622 __wsum delta;
b7fe10e5 2623 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2624
2625 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2626
15e2396d 2627 if (!nopartial) {
b7fe10e5
TH
2628 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2629 return ptr;
2630 }
2631
2632 ptr = skb_gro_header_fast(skb, off);
2633 if (skb_gro_header_hard(skb, off + plen)) {
2634 ptr = skb_gro_header_slow(skb, off + plen, off);
2635 if (!ptr)
2636 return NULL;
15e2396d
TH
2637 }
2638
b7fe10e5
TH
2639 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2640 start, offset);
dcdc8994
TH
2641
2642 /* Adjust skb->csum since we changed the packet */
dcdc8994 2643 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2644
b7fe10e5 2645 grc->offset = off + hdrlen + offset;
26c4f7da 2646 grc->delta = delta;
b7fe10e5
TH
2647
2648 return ptr;
dcdc8994
TH
2649}
2650
26c4f7da
TH
2651static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2652 struct gro_remcsum *grc)
2653{
b7fe10e5
TH
2654 void *ptr;
2655 size_t plen = grc->offset + sizeof(u16);
2656
26c4f7da
TH
2657 if (!grc->delta)
2658 return;
2659
b7fe10e5
TH
2660 ptr = skb_gro_header_fast(skb, grc->offset);
2661 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2662 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2663 if (!ptr)
2664 return;
2665 }
2666
2667 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2668}
dcdc8994 2669
0c4e8581
SH
2670static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2671 unsigned short type,
3b04ddde 2672 const void *daddr, const void *saddr,
95c96174 2673 unsigned int len)
0c4e8581 2674{
f1ecfd5d 2675 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2676 return 0;
3b04ddde
SH
2677
2678 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2679}
2680
b95cce35
SH
2681static inline int dev_parse_header(const struct sk_buff *skb,
2682 unsigned char *haddr)
2683{
2684 const struct net_device *dev = skb->dev;
2685
1b83336b 2686 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2687 return 0;
3b04ddde 2688 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2689}
2690
2793a23a
WB
2691/* ll_header must have at least hard_header_len allocated */
2692static inline bool dev_validate_header(const struct net_device *dev,
2693 char *ll_header, int len)
2694{
2695 if (likely(len >= dev->hard_header_len))
2696 return true;
2697
2698 if (capable(CAP_SYS_RAWIO)) {
2699 memset(ll_header + len, 0, dev->hard_header_len - len);
2700 return true;
2701 }
2702
2703 if (dev->header_ops && dev->header_ops->validate)
2704 return dev->header_ops->validate(ll_header, len);
2705
2706 return false;
2707}
2708
1da177e4 2709typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2710int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2711static inline int unregister_gifconf(unsigned int family)
2712{
2713 return register_gifconf(family, NULL);
2714}
2715
99bbc707 2716#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2717#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2718struct sd_flow_limit {
2719 u64 count;
2720 unsigned int num_buckets;
2721 unsigned int history_head;
2722 u16 history[FLOW_LIMIT_HISTORY];
2723 u8 buckets[];
2724};
2725
2726extern int netdev_flow_limit_table_len;
2727#endif /* CONFIG_NET_FLOW_LIMIT */
2728
1da177e4 2729/*
5e82b4b2 2730 * Incoming packets are placed on per-CPU queues
1da177e4 2731 */
d94d9fee 2732struct softnet_data {
1da177e4 2733 struct list_head poll_list;
6e7676c1 2734 struct sk_buff_head process_queue;
1da177e4 2735
dee42870 2736 /* stats */
cd7b5396
DM
2737 unsigned int processed;
2738 unsigned int time_squeeze;
cd7b5396 2739 unsigned int received_rps;
fd793d89 2740#ifdef CONFIG_RPS
88751275 2741 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2742#endif
2743#ifdef CONFIG_NET_FLOW_LIMIT
2744 struct sd_flow_limit __rcu *flow_limit;
2745#endif
2746 struct Qdisc *output_queue;
2747 struct Qdisc **output_queue_tailp;
2748 struct sk_buff *completion_queue;
88751275 2749
4cdb1e2e 2750#ifdef CONFIG_RPS
501e7ef5
ED
2751 /* input_queue_head should be written by cpu owning this struct,
2752 * and only read by other cpus. Worth using a cache line.
2753 */
2754 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2755
2756 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2757 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2758 struct softnet_data *rps_ipi_next;
2759 unsigned int cpu;
76cc8b13 2760 unsigned int input_queue_tail;
1e94d72f 2761#endif
95c96174 2762 unsigned int dropped;
0a9627f2 2763 struct sk_buff_head input_pkt_queue;
bea3348e 2764 struct napi_struct backlog;
99bbc707 2765
1da177e4
LT
2766};
2767
76cc8b13 2768static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2769{
2770#ifdef CONFIG_RPS
76cc8b13
TH
2771 sd->input_queue_head++;
2772#endif
2773}
2774
2775static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2776 unsigned int *qtail)
2777{
2778#ifdef CONFIG_RPS
2779 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2780#endif
2781}
2782
0a9627f2 2783DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2784
f629d208 2785void __netif_schedule(struct Qdisc *q);
46e5da40 2786void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2787
fd2ea0a7
DM
2788static inline void netif_tx_schedule_all(struct net_device *dev)
2789{
2790 unsigned int i;
2791
2792 for (i = 0; i < dev->num_tx_queues; i++)
2793 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2794}
2795
f9a7cbbf 2796static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2797{
73466498 2798 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2799}
2800
bea3348e
SH
2801/**
2802 * netif_start_queue - allow transmit
2803 * @dev: network device
2804 *
2805 * Allow upper layers to call the device hard_start_xmit routine.
2806 */
1da177e4
LT
2807static inline void netif_start_queue(struct net_device *dev)
2808{
e8a0464c 2809 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2810}
2811
fd2ea0a7
DM
2812static inline void netif_tx_start_all_queues(struct net_device *dev)
2813{
2814 unsigned int i;
2815
2816 for (i = 0; i < dev->num_tx_queues; i++) {
2817 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2818 netif_tx_start_queue(txq);
2819 }
2820}
2821
46e5da40 2822void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2823
d29f749e
DJ
2824/**
2825 * netif_wake_queue - restart transmit
2826 * @dev: network device
2827 *
2828 * Allow upper layers to call the device hard_start_xmit routine.
2829 * Used for flow control when transmit resources are available.
2830 */
79d16385
DM
2831static inline void netif_wake_queue(struct net_device *dev)
2832{
e8a0464c 2833 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2834}
2835
fd2ea0a7
DM
2836static inline void netif_tx_wake_all_queues(struct net_device *dev)
2837{
2838 unsigned int i;
2839
2840 for (i = 0; i < dev->num_tx_queues; i++) {
2841 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2842 netif_tx_wake_queue(txq);
2843 }
2844}
2845
f9a7cbbf 2846static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2847{
73466498 2848 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2849}
2850
bea3348e
SH
2851/**
2852 * netif_stop_queue - stop transmitted packets
2853 * @dev: network device
2854 *
2855 * Stop upper layers calling the device hard_start_xmit routine.
2856 * Used for flow control when transmit resources are unavailable.
2857 */
1da177e4
LT
2858static inline void netif_stop_queue(struct net_device *dev)
2859{
e8a0464c 2860 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2861}
2862
a2029240 2863void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2864
4d29515f 2865static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2866{
73466498 2867 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2868}
2869
bea3348e
SH
2870/**
2871 * netif_queue_stopped - test if transmit queue is flowblocked
2872 * @dev: network device
2873 *
2874 * Test if transmit queue on device is currently unable to send.
2875 */
4d29515f 2876static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2877{
e8a0464c 2878 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2879}
2880
4d29515f 2881static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2882{
73466498
TH
2883 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2884}
2885
8e2f1a63
DB
2886static inline bool
2887netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2888{
2889 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2890}
2891
8e2f1a63
DB
2892static inline bool
2893netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2894{
2895 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2896}
2897
53511453
ED
2898/**
2899 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2900 * @dev_queue: pointer to transmit queue
2901 *
2902 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2903 * to give appropriate hint to the CPU.
53511453
ED
2904 */
2905static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2906{
2907#ifdef CONFIG_BQL
2908 prefetchw(&dev_queue->dql.num_queued);
2909#endif
2910}
2911
2912/**
2913 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2914 * @dev_queue: pointer to transmit queue
2915 *
2916 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2917 * to give appropriate hint to the CPU.
53511453
ED
2918 */
2919static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2920{
2921#ifdef CONFIG_BQL
2922 prefetchw(&dev_queue->dql.limit);
2923#endif
2924}
2925
c5d67bd7
TH
2926static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2927 unsigned int bytes)
2928{
114cf580
TH
2929#ifdef CONFIG_BQL
2930 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2931
2932 if (likely(dql_avail(&dev_queue->dql) >= 0))
2933 return;
2934
2935 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2936
2937 /*
2938 * The XOFF flag must be set before checking the dql_avail below,
2939 * because in netdev_tx_completed_queue we update the dql_completed
2940 * before checking the XOFF flag.
2941 */
2942 smp_mb();
2943
2944 /* check again in case another CPU has just made room avail */
2945 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2946 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2947#endif
c5d67bd7
TH
2948}
2949
0042d0c8
FF
2950/**
2951 * netdev_sent_queue - report the number of bytes queued to hardware
2952 * @dev: network device
2953 * @bytes: number of bytes queued to the hardware device queue
2954 *
2955 * Report the number of bytes queued for sending/completion to the network
2956 * device hardware queue. @bytes should be a good approximation and should
2957 * exactly match netdev_completed_queue() @bytes
2958 */
c5d67bd7
TH
2959static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2960{
2961 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2962}
2963
2964static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2965 unsigned int pkts, unsigned int bytes)
c5d67bd7 2966{
114cf580 2967#ifdef CONFIG_BQL
b37c0fbe
AD
2968 if (unlikely(!bytes))
2969 return;
2970
2971 dql_completed(&dev_queue->dql, bytes);
2972
2973 /*
2974 * Without the memory barrier there is a small possiblity that
2975 * netdev_tx_sent_queue will miss the update and cause the queue to
2976 * be stopped forever
2977 */
2978 smp_mb();
2979
2980 if (dql_avail(&dev_queue->dql) < 0)
2981 return;
2982
2983 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2984 netif_schedule_queue(dev_queue);
114cf580 2985#endif
c5d67bd7
TH
2986}
2987
0042d0c8
FF
2988/**
2989 * netdev_completed_queue - report bytes and packets completed by device
2990 * @dev: network device
2991 * @pkts: actual number of packets sent over the medium
2992 * @bytes: actual number of bytes sent over the medium
2993 *
2994 * Report the number of bytes and packets transmitted by the network device
2995 * hardware queue over the physical medium, @bytes must exactly match the
2996 * @bytes amount passed to netdev_sent_queue()
2997 */
c5d67bd7 2998static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2999 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3000{
3001 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3002}
3003
3004static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3005{
114cf580 3006#ifdef CONFIG_BQL
5c490354 3007 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3008 dql_reset(&q->dql);
3009#endif
c5d67bd7
TH
3010}
3011
0042d0c8
FF
3012/**
3013 * netdev_reset_queue - reset the packets and bytes count of a network device
3014 * @dev_queue: network device
3015 *
3016 * Reset the bytes and packet count of a network device and clear the
3017 * software flow control OFF bit for this network device
3018 */
c5d67bd7
TH
3019static inline void netdev_reset_queue(struct net_device *dev_queue)
3020{
3021 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3022}
3023
b9507bda
DB
3024/**
3025 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3026 * @dev: network device
3027 * @queue_index: given tx queue index
3028 *
3029 * Returns 0 if given tx queue index >= number of device tx queues,
3030 * otherwise returns the originally passed tx queue index.
3031 */
3032static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3033{
3034 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3035 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3036 dev->name, queue_index,
3037 dev->real_num_tx_queues);
3038 return 0;
3039 }
3040
3041 return queue_index;
3042}
3043
bea3348e
SH
3044/**
3045 * netif_running - test if up
3046 * @dev: network device
3047 *
3048 * Test if the device has been brought up.
3049 */
4d29515f 3050static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3051{
3052 return test_bit(__LINK_STATE_START, &dev->state);
3053}
3054
f25f4e44 3055/*
5e82b4b2 3056 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3057 * stop, and a check if it's stopped. All other device management is
3058 * done at the overall netdevice level.
3059 * Also test the device if we're multiqueue.
3060 */
bea3348e
SH
3061
3062/**
3063 * netif_start_subqueue - allow sending packets on subqueue
3064 * @dev: network device
3065 * @queue_index: sub queue index
3066 *
3067 * Start individual transmit queue of a device with multiple transmit queues.
3068 */
f25f4e44
PWJ
3069static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3070{
fd2ea0a7 3071 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3072
3073 netif_tx_start_queue(txq);
f25f4e44
PWJ
3074}
3075
bea3348e
SH
3076/**
3077 * netif_stop_subqueue - stop sending packets on subqueue
3078 * @dev: network device
3079 * @queue_index: sub queue index
3080 *
3081 * Stop individual transmit queue of a device with multiple transmit queues.
3082 */
f25f4e44
PWJ
3083static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3084{
fd2ea0a7 3085 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3086 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3087}
3088
bea3348e
SH
3089/**
3090 * netif_subqueue_stopped - test status of subqueue
3091 * @dev: network device
3092 * @queue_index: sub queue index
3093 *
3094 * Check individual transmit queue of a device with multiple transmit queues.
3095 */
4d29515f
DM
3096static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3097 u16 queue_index)
f25f4e44 3098{
fd2ea0a7 3099 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3100
3101 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3102}
3103
4d29515f
DM
3104static inline bool netif_subqueue_stopped(const struct net_device *dev,
3105 struct sk_buff *skb)
668f895a
PE
3106{
3107 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3108}
bea3348e 3109
46e5da40 3110void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 3111
537c00de 3112#ifdef CONFIG_XPS
53af53ae 3113int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3114 u16 index);
537c00de
AD
3115#else
3116static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3117 const struct cpumask *mask,
537c00de
AD
3118 u16 index)
3119{
3120 return 0;
3121}
3122#endif
3123
5605c762
JP
3124u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3125 unsigned int num_tx_queues);
3126
a3d22a68
VZ
3127/*
3128 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3129 * as a distribution range limit for the returned value.
3130 */
3131static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3132 struct sk_buff *skb)
a3d22a68
VZ
3133{
3134 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3135}
3136
bea3348e
SH
3137/**
3138 * netif_is_multiqueue - test if device has multiple transmit queues
3139 * @dev: network device
3140 *
3141 * Check if device has multiple transmit queues
bea3348e 3142 */
4d29515f 3143static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3144{
a02cec21 3145 return dev->num_tx_queues > 1;
f25f4e44 3146}
1da177e4 3147
f629d208 3148int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3149
a953be53 3150#ifdef CONFIG_SYSFS
f629d208 3151int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3152#else
3153static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3154 unsigned int rxq)
3155{
3156 return 0;
3157}
3158#endif
3159
a953be53
MD
3160#ifdef CONFIG_SYSFS
3161static inline unsigned int get_netdev_rx_queue_index(
3162 struct netdev_rx_queue *queue)
3163{
3164 struct net_device *dev = queue->dev;
3165 int index = queue - dev->_rx;
3166
3167 BUG_ON(index >= dev->num_rx_queues);
3168 return index;
3169}
3170#endif
3171
16917b87 3172#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3173int netif_get_num_default_rss_queues(void);
16917b87 3174
e6247027
ED
3175enum skb_free_reason {
3176 SKB_REASON_CONSUMED,
3177 SKB_REASON_DROPPED,
3178};
3179
3180void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3181void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3182
e6247027
ED
3183/*
3184 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3185 * interrupt context or with hardware interrupts being disabled.
3186 * (in_irq() || irqs_disabled())
3187 *
3188 * We provide four helpers that can be used in following contexts :
3189 *
3190 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3191 * replacing kfree_skb(skb)
3192 *
3193 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3194 * Typically used in place of consume_skb(skb) in TX completion path
3195 *
3196 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3197 * replacing kfree_skb(skb)
3198 *
3199 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3200 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3201 */
e6247027
ED
3202static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3203{
3204 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3205}
3206
3207static inline void dev_consume_skb_irq(struct sk_buff *skb)
3208{
3209 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3210}
3211
3212static inline void dev_kfree_skb_any(struct sk_buff *skb)
3213{
3214 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3215}
3216
3217static inline void dev_consume_skb_any(struct sk_buff *skb)
3218{
3219 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3220}
1da177e4 3221
f629d208
JP
3222int netif_rx(struct sk_buff *skb);
3223int netif_rx_ni(struct sk_buff *skb);
04eb4489 3224int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3225gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3226void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3227struct sk_buff *napi_get_frags(struct napi_struct *napi);
3228gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3229struct packet_offload *gro_find_receive_by_type(__be16 type);
3230struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3231
3232static inline void napi_free_frags(struct napi_struct *napi)
3233{
3234 kfree_skb(napi->skb);
3235 napi->skb = NULL;
3236}
3237
24b27fc4 3238bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3239int netdev_rx_handler_register(struct net_device *dev,
3240 rx_handler_func_t *rx_handler,
3241 void *rx_handler_data);
3242void netdev_rx_handler_unregister(struct net_device *dev);
3243
3244bool dev_valid_name(const char *name);
3245int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3246int dev_ethtool(struct net *net, struct ifreq *);
3247unsigned int dev_get_flags(const struct net_device *);
3248int __dev_change_flags(struct net_device *, unsigned int flags);
3249int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3250void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3251 unsigned int gchanges);
f629d208
JP
3252int dev_change_name(struct net_device *, const char *);
3253int dev_set_alias(struct net_device *, const char *, size_t);
3254int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3255int dev_set_mtu(struct net_device *, int);
3256void dev_set_group(struct net_device *, int);
3257int dev_set_mac_address(struct net_device *, struct sockaddr *);
3258int dev_change_carrier(struct net_device *, bool new_carrier);
3259int dev_get_phys_port_id(struct net_device *dev,
02637fce 3260 struct netdev_phys_item_id *ppid);
db24a904
DA
3261int dev_get_phys_port_name(struct net_device *dev,
3262 char *name, size_t len);
d746d707 3263int dev_change_proto_down(struct net_device *dev, bool proto_down);
85de8576 3264int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
55a93b3e 3265struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3266struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3267 struct netdev_queue *txq, int *ret);
a0265d28 3268int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3269int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3270bool is_skb_forwardable(const struct net_device *dev,
3271 const struct sk_buff *skb);
1da177e4 3272
4e3264d2
MKL
3273static __always_inline int ____dev_forward_skb(struct net_device *dev,
3274 struct sk_buff *skb)
3275{
3276 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3277 unlikely(!is_skb_forwardable(dev, skb))) {
3278 atomic_long_inc(&dev->rx_dropped);
3279 kfree_skb(skb);
3280 return NET_RX_DROP;
3281 }
3282
3283 skb_scrub_packet(skb, true);
3284 skb->priority = 0;
3285 return 0;
3286}
3287
74b20582
DA
3288void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3289
20380731 3290extern int netdev_budget;
1da177e4
LT
3291
3292/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3293void netdev_run_todo(void);
1da177e4 3294
bea3348e
SH
3295/**
3296 * dev_put - release reference to device
3297 * @dev: network device
3298 *
9ef4429b 3299 * Release reference to device to allow it to be freed.
bea3348e 3300 */
1da177e4
LT
3301static inline void dev_put(struct net_device *dev)
3302{
933393f5 3303 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3304}
3305
bea3348e
SH
3306/**
3307 * dev_hold - get reference to device
3308 * @dev: network device
3309 *
9ef4429b 3310 * Hold reference to device to keep it from being freed.
bea3348e 3311 */
15333061
SH
3312static inline void dev_hold(struct net_device *dev)
3313{
933393f5 3314 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3315}
1da177e4
LT
3316
3317/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3318 * and _off may be called from IRQ context, but it is caller
3319 * who is responsible for serialization of these calls.
b00055aa
SR
3320 *
3321 * The name carrier is inappropriate, these functions should really be
3322 * called netif_lowerlayer_*() because they represent the state of any
3323 * kind of lower layer not just hardware media.
1da177e4
LT
3324 */
3325
f629d208
JP
3326void linkwatch_init_dev(struct net_device *dev);
3327void linkwatch_fire_event(struct net_device *dev);
3328void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3329
bea3348e
SH
3330/**
3331 * netif_carrier_ok - test if carrier present
3332 * @dev: network device
3333 *
3334 * Check if carrier is present on device
3335 */
4d29515f 3336static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3337{
3338 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3339}
3340
f629d208 3341unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3342
f629d208 3343void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3344
f629d208 3345void netif_carrier_on(struct net_device *dev);
1da177e4 3346
f629d208 3347void netif_carrier_off(struct net_device *dev);
1da177e4 3348
bea3348e
SH
3349/**
3350 * netif_dormant_on - mark device as dormant.
3351 * @dev: network device
3352 *
3353 * Mark device as dormant (as per RFC2863).
3354 *
3355 * The dormant state indicates that the relevant interface is not
3356 * actually in a condition to pass packets (i.e., it is not 'up') but is
3357 * in a "pending" state, waiting for some external event. For "on-
3358 * demand" interfaces, this new state identifies the situation where the
3359 * interface is waiting for events to place it in the up state.
bea3348e 3360 */
b00055aa
SR
3361static inline void netif_dormant_on(struct net_device *dev)
3362{
3363 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3364 linkwatch_fire_event(dev);
3365}
3366
bea3348e
SH
3367/**
3368 * netif_dormant_off - set device as not dormant.
3369 * @dev: network device
3370 *
3371 * Device is not in dormant state.
3372 */
b00055aa
SR
3373static inline void netif_dormant_off(struct net_device *dev)
3374{
3375 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3376 linkwatch_fire_event(dev);
3377}
3378
bea3348e
SH
3379/**
3380 * netif_dormant - test if carrier present
3381 * @dev: network device
3382 *
3383 * Check if carrier is present on device
3384 */
4d29515f 3385static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3386{
3387 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3388}
3389
3390
bea3348e
SH
3391/**
3392 * netif_oper_up - test if device is operational
3393 * @dev: network device
3394 *
3395 * Check if carrier is operational
3396 */
4d29515f 3397static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3398{
b00055aa
SR
3399 return (dev->operstate == IF_OPER_UP ||
3400 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3401}
3402
bea3348e
SH
3403/**
3404 * netif_device_present - is device available or removed
3405 * @dev: network device
3406 *
3407 * Check if device has not been removed from system.
3408 */
4d29515f 3409static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3410{
3411 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3412}
3413
f629d208 3414void netif_device_detach(struct net_device *dev);
1da177e4 3415
f629d208 3416void netif_device_attach(struct net_device *dev);
1da177e4
LT
3417
3418/*
3419 * Network interface message level settings
3420 */
1da177e4
LT
3421
3422enum {
3423 NETIF_MSG_DRV = 0x0001,
3424 NETIF_MSG_PROBE = 0x0002,
3425 NETIF_MSG_LINK = 0x0004,
3426 NETIF_MSG_TIMER = 0x0008,
3427 NETIF_MSG_IFDOWN = 0x0010,
3428 NETIF_MSG_IFUP = 0x0020,
3429 NETIF_MSG_RX_ERR = 0x0040,
3430 NETIF_MSG_TX_ERR = 0x0080,
3431 NETIF_MSG_TX_QUEUED = 0x0100,
3432 NETIF_MSG_INTR = 0x0200,
3433 NETIF_MSG_TX_DONE = 0x0400,
3434 NETIF_MSG_RX_STATUS = 0x0800,
3435 NETIF_MSG_PKTDATA = 0x1000,
3436 NETIF_MSG_HW = 0x2000,
3437 NETIF_MSG_WOL = 0x4000,
3438};
3439
3440#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3441#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3442#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3443#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3444#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3445#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3446#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3447#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3448#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3449#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3450#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3451#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3452#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3453#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3454#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3455
3456static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3457{
3458 /* use default */
3459 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3460 return default_msg_enable_bits;
3461 if (debug_value == 0) /* no output */
3462 return 0;
3463 /* set low N bits */
3464 return (1 << debug_value) - 1;
3465}
3466
c773e847 3467static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3468{
c773e847
DM
3469 spin_lock(&txq->_xmit_lock);
3470 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3471}
3472
5a717f4f
MT
3473static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3474{
3475 __acquire(&txq->_xmit_lock);
3476 return true;
3477}
3478
3479static inline void __netif_tx_release(struct netdev_queue *txq)
3480{
3481 __release(&txq->_xmit_lock);
3482}
3483
fd2ea0a7
DM
3484static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3485{
3486 spin_lock_bh(&txq->_xmit_lock);
3487 txq->xmit_lock_owner = smp_processor_id();
3488}
3489
4d29515f 3490static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3491{
4d29515f 3492 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3493 if (likely(ok))
3494 txq->xmit_lock_owner = smp_processor_id();
3495 return ok;
3496}
3497
3498static inline void __netif_tx_unlock(struct netdev_queue *txq)
3499{
3500 txq->xmit_lock_owner = -1;
3501 spin_unlock(&txq->_xmit_lock);
3502}
3503
3504static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3505{
3506 txq->xmit_lock_owner = -1;
3507 spin_unlock_bh(&txq->_xmit_lock);
3508}
3509
08baf561
ED
3510static inline void txq_trans_update(struct netdev_queue *txq)
3511{
3512 if (txq->xmit_lock_owner != -1)
3513 txq->trans_start = jiffies;
3514}
3515
ba162f8e
FW
3516/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3517static inline void netif_trans_update(struct net_device *dev)
3518{
9b36627a
FW
3519 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3520
3521 if (txq->trans_start != jiffies)
3522 txq->trans_start = jiffies;
ba162f8e
FW
3523}
3524
d29f749e
DJ
3525/**
3526 * netif_tx_lock - grab network device transmit lock
3527 * @dev: network device
d29f749e
DJ
3528 *
3529 * Get network device transmit lock
3530 */
22dd7495
JHS
3531static inline void netif_tx_lock(struct net_device *dev)
3532{
e8a0464c 3533 unsigned int i;
c3f26a26 3534 int cpu;
c773e847 3535
c3f26a26
DM
3536 spin_lock(&dev->tx_global_lock);
3537 cpu = smp_processor_id();
e8a0464c
DM
3538 for (i = 0; i < dev->num_tx_queues; i++) {
3539 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3540
3541 /* We are the only thread of execution doing a
3542 * freeze, but we have to grab the _xmit_lock in
3543 * order to synchronize with threads which are in
3544 * the ->hard_start_xmit() handler and already
3545 * checked the frozen bit.
3546 */
e8a0464c 3547 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3548 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3549 __netif_tx_unlock(txq);
e8a0464c 3550 }
932ff279
HX
3551}
3552
3553static inline void netif_tx_lock_bh(struct net_device *dev)
3554{
e8a0464c
DM
3555 local_bh_disable();
3556 netif_tx_lock(dev);
932ff279
HX
3557}
3558
932ff279
HX
3559static inline void netif_tx_unlock(struct net_device *dev)
3560{
e8a0464c
DM
3561 unsigned int i;
3562
3563 for (i = 0; i < dev->num_tx_queues; i++) {
3564 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3565
c3f26a26
DM
3566 /* No need to grab the _xmit_lock here. If the
3567 * queue is not stopped for another reason, we
3568 * force a schedule.
3569 */
3570 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3571 netif_schedule_queue(txq);
c3f26a26
DM
3572 }
3573 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3574}
3575
3576static inline void netif_tx_unlock_bh(struct net_device *dev)
3577{
e8a0464c
DM
3578 netif_tx_unlock(dev);
3579 local_bh_enable();
932ff279
HX
3580}
3581
c773e847 3582#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3583 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3584 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3585 } else { \
3586 __netif_tx_acquire(txq); \
22dd7495
JHS
3587 } \
3588}
3589
5efeac44
EB
3590#define HARD_TX_TRYLOCK(dev, txq) \
3591 (((dev->features & NETIF_F_LLTX) == 0) ? \
3592 __netif_tx_trylock(txq) : \
5a717f4f 3593 __netif_tx_acquire(txq))
5efeac44 3594
c773e847 3595#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3596 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3597 __netif_tx_unlock(txq); \
5a717f4f
MT
3598 } else { \
3599 __netif_tx_release(txq); \
22dd7495
JHS
3600 } \
3601}
3602
1da177e4
LT
3603static inline void netif_tx_disable(struct net_device *dev)
3604{
fd2ea0a7 3605 unsigned int i;
c3f26a26 3606 int cpu;
fd2ea0a7 3607
c3f26a26
DM
3608 local_bh_disable();
3609 cpu = smp_processor_id();
fd2ea0a7
DM
3610 for (i = 0; i < dev->num_tx_queues; i++) {
3611 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3612
3613 __netif_tx_lock(txq, cpu);
fd2ea0a7 3614 netif_tx_stop_queue(txq);
c3f26a26 3615 __netif_tx_unlock(txq);
fd2ea0a7 3616 }
c3f26a26 3617 local_bh_enable();
1da177e4
LT
3618}
3619
e308a5d8
DM
3620static inline void netif_addr_lock(struct net_device *dev)
3621{
3622 spin_lock(&dev->addr_list_lock);
3623}
3624
2429f7ac
JP
3625static inline void netif_addr_lock_nested(struct net_device *dev)
3626{
25175ba5
VY
3627 int subclass = SINGLE_DEPTH_NESTING;
3628
3629 if (dev->netdev_ops->ndo_get_lock_subclass)
3630 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3631
3632 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3633}
3634
e308a5d8
DM
3635static inline void netif_addr_lock_bh(struct net_device *dev)
3636{
3637 spin_lock_bh(&dev->addr_list_lock);
3638}
3639
3640static inline void netif_addr_unlock(struct net_device *dev)
3641{
3642 spin_unlock(&dev->addr_list_lock);
3643}
3644
3645static inline void netif_addr_unlock_bh(struct net_device *dev)
3646{
3647 spin_unlock_bh(&dev->addr_list_lock);
3648}
3649
f001fde5 3650/*
31278e71 3651 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3652 * rcu_read_lock held.
3653 */
3654#define for_each_dev_addr(dev, ha) \
31278e71 3655 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3656
1da177e4
LT
3657/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3658
f629d208 3659void ether_setup(struct net_device *dev);
1da177e4
LT
3660
3661/* Support for loadable net-drivers */
f629d208 3662struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3663 unsigned char name_assign_type,
f629d208
JP
3664 void (*setup)(struct net_device *),
3665 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3666#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3667 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3668
c835a677
TG
3669#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3670 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3671 count)
36909ea4 3672
f629d208
JP
3673int register_netdev(struct net_device *dev);
3674void unregister_netdev(struct net_device *dev);
f001fde5 3675
22bedad3 3676/* General hardware address lists handling functions */
f629d208
JP
3677int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3678 struct netdev_hw_addr_list *from_list, int addr_len);
3679void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3680 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3681int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3682 struct net_device *dev,
3683 int (*sync)(struct net_device *, const unsigned char *),
3684 int (*unsync)(struct net_device *,
3685 const unsigned char *));
3686void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3687 struct net_device *dev,
3688 int (*unsync)(struct net_device *,
3689 const unsigned char *));
f629d208 3690void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3691
f001fde5 3692/* Functions used for device addresses handling */
f629d208
JP
3693int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3694 unsigned char addr_type);
3695int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3696 unsigned char addr_type);
f629d208
JP
3697void dev_addr_flush(struct net_device *dev);
3698int dev_addr_init(struct net_device *dev);
a748ee24
JP
3699
3700/* Functions used for unicast addresses handling */
f629d208
JP
3701int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3702int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3703int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3704int dev_uc_sync(struct net_device *to, struct net_device *from);
3705int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3706void dev_uc_unsync(struct net_device *to, struct net_device *from);
3707void dev_uc_flush(struct net_device *dev);
3708void dev_uc_init(struct net_device *dev);
f001fde5 3709
670e5b8e
AD
3710/**
3711 * __dev_uc_sync - Synchonize device's unicast list
3712 * @dev: device to sync
3713 * @sync: function to call if address should be added
3714 * @unsync: function to call if address should be removed
3715 *
3716 * Add newly added addresses to the interface, and release
3717 * addresses that have been deleted.
5e82b4b2 3718 */
670e5b8e
AD
3719static inline int __dev_uc_sync(struct net_device *dev,
3720 int (*sync)(struct net_device *,
3721 const unsigned char *),
3722 int (*unsync)(struct net_device *,
3723 const unsigned char *))
3724{
3725 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3726}
3727
3728/**
e793c0f7 3729 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3730 * @dev: device to sync
3731 * @unsync: function to call if address should be removed
3732 *
3733 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3734 */
670e5b8e
AD
3735static inline void __dev_uc_unsync(struct net_device *dev,
3736 int (*unsync)(struct net_device *,
3737 const unsigned char *))
3738{
3739 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3740}
3741
22bedad3 3742/* Functions used for multicast addresses handling */
f629d208
JP
3743int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3744int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3745int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3746int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3747int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3748int dev_mc_sync(struct net_device *to, struct net_device *from);
3749int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3750void dev_mc_unsync(struct net_device *to, struct net_device *from);
3751void dev_mc_flush(struct net_device *dev);
3752void dev_mc_init(struct net_device *dev);
f001fde5 3753
670e5b8e
AD
3754/**
3755 * __dev_mc_sync - Synchonize device's multicast list
3756 * @dev: device to sync
3757 * @sync: function to call if address should be added
3758 * @unsync: function to call if address should be removed
3759 *
3760 * Add newly added addresses to the interface, and release
3761 * addresses that have been deleted.
5e82b4b2 3762 */
670e5b8e
AD
3763static inline int __dev_mc_sync(struct net_device *dev,
3764 int (*sync)(struct net_device *,
3765 const unsigned char *),
3766 int (*unsync)(struct net_device *,
3767 const unsigned char *))
3768{
3769 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3770}
3771
3772/**
e793c0f7 3773 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3774 * @dev: device to sync
3775 * @unsync: function to call if address should be removed
3776 *
3777 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3778 */
670e5b8e
AD
3779static inline void __dev_mc_unsync(struct net_device *dev,
3780 int (*unsync)(struct net_device *,
3781 const unsigned char *))
3782{
3783 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3784}
3785
4417da66 3786/* Functions used for secondary unicast and multicast support */
f629d208
JP
3787void dev_set_rx_mode(struct net_device *dev);
3788void __dev_set_rx_mode(struct net_device *dev);
3789int dev_set_promiscuity(struct net_device *dev, int inc);
3790int dev_set_allmulti(struct net_device *dev, int inc);
3791void netdev_state_change(struct net_device *dev);
3792void netdev_notify_peers(struct net_device *dev);
3793void netdev_features_change(struct net_device *dev);
1da177e4 3794/* Load a device via the kmod */
f629d208
JP
3795void dev_load(struct net *net, const char *name);
3796struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3797 struct rtnl_link_stats64 *storage);
3798void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3799 const struct net_device_stats *netdev_stats);
eeda3fd6 3800
1da177e4 3801extern int netdev_max_backlog;
3b098e2d 3802extern int netdev_tstamp_prequeue;
1da177e4 3803extern int weight_p;
9ff162a8 3804
f629d208 3805bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3806struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3807 struct list_head **iter);
f629d208
JP
3808struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3809 struct list_head **iter);
8b5be856 3810
44a40855
VY
3811/* iterate through upper list, must be called under RCU read lock */
3812#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3813 for (iter = &(dev)->adj_list.upper, \
3814 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3815 updev; \
3816 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3817
1a3f060c
DA
3818int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3819 int (*fn)(struct net_device *upper_dev,
3820 void *data),
3821 void *data);
3822
3823bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3824 struct net_device *upper_dev);
3825
f629d208
JP
3826void *netdev_lower_get_next_private(struct net_device *dev,
3827 struct list_head **iter);
3828void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3829 struct list_head **iter);
31088a11
VF
3830
3831#define netdev_for_each_lower_private(dev, priv, iter) \
3832 for (iter = (dev)->adj_list.lower.next, \
3833 priv = netdev_lower_get_next_private(dev, &(iter)); \
3834 priv; \
3835 priv = netdev_lower_get_next_private(dev, &(iter)))
3836
3837#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3838 for (iter = &(dev)->adj_list.lower, \
3839 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3840 priv; \
3841 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3842
4085ebe8
VY
3843void *netdev_lower_get_next(struct net_device *dev,
3844 struct list_head **iter);
7ce856aa 3845
4085ebe8 3846#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3847 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3848 ldev = netdev_lower_get_next(dev, &(iter)); \
3849 ldev; \
3850 ldev = netdev_lower_get_next(dev, &(iter)))
3851
7ce856aa
JP
3852struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3853 struct list_head **iter);
3854struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3855 struct list_head **iter);
3856
1a3f060c
DA
3857int netdev_walk_all_lower_dev(struct net_device *dev,
3858 int (*fn)(struct net_device *lower_dev,
3859 void *data),
3860 void *data);
3861int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3862 int (*fn)(struct net_device *lower_dev,
3863 void *data),
3864 void *data);
3865
f629d208 3866void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3867void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3868struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3869struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3870int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3871int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3872 struct net_device *upper_dev,
29bf24af 3873 void *upper_priv, void *upper_info);
f629d208
JP
3874void netdev_upper_dev_unlink(struct net_device *dev,
3875 struct net_device *upper_dev);
5bb025fa 3876void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3877void *netdev_lower_dev_get_private(struct net_device *dev,
3878 struct net_device *lower_dev);
04d48266
JP
3879void netdev_lower_state_changed(struct net_device *lower_dev,
3880 void *lower_state_info);
18bfb924
JP
3881int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3882 struct neighbour *n);
3883void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3884 struct neighbour *n);
960fb622
ED
3885
3886/* RSS keys are 40 or 52 bytes long */
3887#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3888extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3889void netdev_rss_key_fill(void *buffer, size_t len);
3890
952fcfd0 3891int dev_get_nest_level(struct net_device *dev);
f629d208
JP
3892int skb_checksum_help(struct sk_buff *skb);
3893struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3894 netdev_features_t features, bool tx_path);
3895struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3896 netdev_features_t features);
12b0004d 3897
61bd3857
MS
3898struct netdev_bonding_info {
3899 ifslave slave;
3900 ifbond master;
3901};
3902
3903struct netdev_notifier_bonding_info {
3904 struct netdev_notifier_info info; /* must be first */
3905 struct netdev_bonding_info bonding_info;
3906};
3907
3908void netdev_bonding_info_change(struct net_device *dev,
3909 struct netdev_bonding_info *bonding_info);
3910
12b0004d
CW
3911static inline
3912struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3913{
3914 return __skb_gso_segment(skb, features, true);
3915}
53d6471c 3916__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3917
3918static inline bool can_checksum_protocol(netdev_features_t features,
3919 __be16 protocol)
3920{
c8cd0989
TH
3921 if (protocol == htons(ETH_P_FCOE))
3922 return !!(features & NETIF_F_FCOE_CRC);
3923
3924 /* Assume this is an IP checksum (not SCTP CRC) */
3925
3926 if (features & NETIF_F_HW_CSUM) {
3927 /* Can checksum everything */
3928 return true;
3929 }
3930
3931 switch (protocol) {
3932 case htons(ETH_P_IP):
3933 return !!(features & NETIF_F_IP_CSUM);
3934 case htons(ETH_P_IPV6):
3935 return !!(features & NETIF_F_IPV6_CSUM);
3936 default:
3937 return false;
3938 }
ec5f0615 3939}
12b0004d 3940
fb286bb2 3941#ifdef CONFIG_BUG
f629d208 3942void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3943#else
3944static inline void netdev_rx_csum_fault(struct net_device *dev)
3945{
3946}
3947#endif
1da177e4 3948/* rx skb timestamps */
f629d208
JP
3949void net_enable_timestamp(void);
3950void net_disable_timestamp(void);
1da177e4 3951
20380731 3952#ifdef CONFIG_PROC_FS
f629d208 3953int __init dev_proc_init(void);
900ff8c6
CW
3954#else
3955#define dev_proc_init() 0
20380731
ACM
3956#endif
3957
4798248e 3958static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3959 struct sk_buff *skb, struct net_device *dev,
3960 bool more)
4798248e 3961{
fa2dbdc2 3962 skb->xmit_more = more ? 1 : 0;
0b725a2c 3963 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3964}
3965
10b3ad8c 3966static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3967 struct netdev_queue *txq, bool more)
4798248e
DM
3968{
3969 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3970 int rc;
4798248e 3971
fa2dbdc2 3972 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3973 if (rc == NETDEV_TX_OK)
3974 txq_trans_update(txq);
3975
3976 return rc;
4798248e
DM
3977}
3978
42a2d923
LT
3979int netdev_class_create_file_ns(struct class_attribute *class_attr,
3980 const void *ns);
3981void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3982 const void *ns);
58292cbe
TH
3983
3984static inline int netdev_class_create_file(struct class_attribute *class_attr)
3985{
3986 return netdev_class_create_file_ns(class_attr, NULL);
3987}
3988
3989static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3990{
3991 netdev_class_remove_file_ns(class_attr, NULL);
3992}
b8a9787e 3993
04600794
JB
3994extern struct kobj_ns_type_operations net_ns_type_operations;
3995
f629d208 3996const char *netdev_drivername(const struct net_device *dev);
6579e57b 3997
f629d208 3998void linkwatch_run_queue(void);
20380731 3999
da08143b
MK
4000static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4001 netdev_features_t f2)
4002{
c8cd0989
TH
4003 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4004 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4005 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4006 else
b6a0e72a 4007 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4008 }
da08143b 4009
c8cd0989 4010 return f1 & f2;
da08143b
MK
4011}
4012
c8f44aff
MM
4013static inline netdev_features_t netdev_get_wanted_features(
4014 struct net_device *dev)
5455c699
MM
4015{
4016 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4017}
c8f44aff
MM
4018netdev_features_t netdev_increment_features(netdev_features_t all,
4019 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4020
4021/* Allow TSO being used on stacked device :
4022 * Performing the GSO segmentation before last device
4023 * is a performance improvement.
4024 */
4025static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4026 netdev_features_t mask)
4027{
4028 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4029}
4030
6cb6a27c 4031int __netdev_update_features(struct net_device *dev);
5455c699 4032void netdev_update_features(struct net_device *dev);
afe12cc8 4033void netdev_change_features(struct net_device *dev);
7f353bf2 4034
fc4a7489
PM
4035void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4036 struct net_device *dev);
4037
e38f3025
TM
4038netdev_features_t passthru_features_check(struct sk_buff *skb,
4039 struct net_device *dev,
4040 netdev_features_t features);
c1e756bf 4041netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4042
4d29515f 4043static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4044{
7b748340 4045 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4046
4047 /* check flags correspondence */
4048 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4049 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4050 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4051 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4052 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4053 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4054 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4055 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4056 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4057 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4058 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4059 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4060 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4061 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4062 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4063 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
0345e186 4064
d6b4991a 4065 return (features & feature) == feature;
576a30eb
HX
4066}
4067
4d29515f 4068static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4069{
278b2513 4070 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4071 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4072}
4073
8b86a61d 4074static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4075 netdev_features_t features)
7967168c 4076{
fc741216 4077 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4078 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4079 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4080}
4081
82cc1a7a
PWJ
4082static inline void netif_set_gso_max_size(struct net_device *dev,
4083 unsigned int size)
4084{
4085 dev->gso_max_size = size;
4086}
4087
7a7ffbab
WCC
4088static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4089 int pulled_hlen, u16 mac_offset,
4090 int mac_len)
4091{
4092 skb->protocol = protocol;
4093 skb->encapsulation = 1;
4094 skb_push(skb, pulled_hlen);
4095 skb_reset_transport_header(skb);
4096 skb->mac_header = mac_offset;
4097 skb->network_header = skb->mac_header + mac_len;
4098 skb->mac_len = mac_len;
4099}
4100
3c175784
SD
4101static inline bool netif_is_macsec(const struct net_device *dev)
4102{
4103 return dev->priv_flags & IFF_MACSEC;
4104}
4105
b618aaa9 4106static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4107{
4108 return dev->priv_flags & IFF_MACVLAN;
4109}
4110
b618aaa9 4111static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4112{
4113 return dev->priv_flags & IFF_MACVLAN_PORT;
4114}
4115
b618aaa9 4116static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4117{
4118 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4119}
4120
b618aaa9 4121static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4122{
4123 return dev->priv_flags & IFF_IPVLAN_MASTER;
4124}
4125
b618aaa9 4126static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4127{
4128 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4129}
4130
b618aaa9 4131static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4132{
4133 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4134}
4135
3bdc0eba
BG
4136static inline bool netif_supports_nofcs(struct net_device *dev)
4137{
4138 return dev->priv_flags & IFF_SUPP_NOFCS;
4139}
4140
007979ea 4141static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4142{
007979ea 4143 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4144}
4145
fee6d4c7
DA
4146static inline bool netif_is_l3_slave(const struct net_device *dev)
4147{
4148 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4149}
4150
0894ae3f
JP
4151static inline bool netif_is_bridge_master(const struct net_device *dev)
4152{
4153 return dev->priv_flags & IFF_EBRIDGE;
4154}
4155
28f9ee22
VY
4156static inline bool netif_is_bridge_port(const struct net_device *dev)
4157{
4158 return dev->priv_flags & IFF_BRIDGE_PORT;
4159}
4160
35d4e172
JP
4161static inline bool netif_is_ovs_master(const struct net_device *dev)
4162{
4163 return dev->priv_flags & IFF_OPENVSWITCH;
4164}
4165
b618aaa9 4166static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4167{
4168 return dev->priv_flags & IFF_TEAM;
4169}
4170
b618aaa9 4171static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4172{
4173 return dev->priv_flags & IFF_TEAM_PORT;
4174}
4175
b618aaa9 4176static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4177{
4178 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4179}
4180
b618aaa9 4181static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4182{
4183 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4184}
4185
d4ab4286
KJ
4186static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4187{
4188 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4189}
4190
02875878
ED
4191/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4192static inline void netif_keep_dst(struct net_device *dev)
4193{
4194 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4195}
4196
18d3df3e
PA
4197/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4198static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4199{
4200 /* TODO: reserve and use an additional IFF bit, if we get more users */
4201 return dev->priv_flags & IFF_MACSEC;
4202}
4203
505d4f73 4204extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4205
571ba423
JP
4206/* Logging, debugging and troubleshooting/diagnostic helpers. */
4207
4208/* netdev_printk helpers, similar to dev_printk */
4209
4210static inline const char *netdev_name(const struct net_device *dev)
4211{
c6f854d5
VF
4212 if (!dev->name[0] || strchr(dev->name, '%'))
4213 return "(unnamed net_device)";
571ba423
JP
4214 return dev->name;
4215}
4216
ccc7f496
VF
4217static inline const char *netdev_reg_state(const struct net_device *dev)
4218{
4219 switch (dev->reg_state) {
4220 case NETREG_UNINITIALIZED: return " (uninitialized)";
4221 case NETREG_REGISTERED: return "";
4222 case NETREG_UNREGISTERING: return " (unregistering)";
4223 case NETREG_UNREGISTERED: return " (unregistered)";
4224 case NETREG_RELEASED: return " (released)";
4225 case NETREG_DUMMY: return " (dummy)";
4226 }
4227
4228 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4229 return " (unknown)";
4230}
4231
f629d208 4232__printf(3, 4)
6ea754eb
JP
4233void netdev_printk(const char *level, const struct net_device *dev,
4234 const char *format, ...);
f629d208 4235__printf(2, 3)
6ea754eb 4236void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4237__printf(2, 3)
6ea754eb 4238void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4239__printf(2, 3)
6ea754eb 4240void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4241__printf(2, 3)
6ea754eb 4242void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4243__printf(2, 3)
6ea754eb 4244void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4245__printf(2, 3)
6ea754eb 4246void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4247__printf(2, 3)
6ea754eb 4248void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4249
8909c9ad
VK
4250#define MODULE_ALIAS_NETDEV(device) \
4251 MODULE_ALIAS("netdev-" device)
4252
b558c96f 4253#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4254#define netdev_dbg(__dev, format, args...) \
4255do { \
ffa10cb4 4256 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4257} while (0)
b558c96f
JC
4258#elif defined(DEBUG)
4259#define netdev_dbg(__dev, format, args...) \
4260 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4261#else
4262#define netdev_dbg(__dev, format, args...) \
4263({ \
4264 if (0) \
4265 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4266})
4267#endif
4268
4269#if defined(VERBOSE_DEBUG)
4270#define netdev_vdbg netdev_dbg
4271#else
4272
4273#define netdev_vdbg(dev, format, args...) \
4274({ \
4275 if (0) \
4276 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4277 0; \
4278})
4279#endif
4280
4281/*
4282 * netdev_WARN() acts like dev_printk(), but with the key difference
4283 * of using a WARN/WARN_ON to get the message out, including the
4284 * file/line information and a backtrace.
4285 */
4286#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4287 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4288 netdev_reg_state(dev), ##args)
571ba423 4289
b3d95c5c
JP
4290/* netif printk helpers, similar to netdev_printk */
4291
4292#define netif_printk(priv, type, level, dev, fmt, args...) \
4293do { \
4294 if (netif_msg_##type(priv)) \
4295 netdev_printk(level, (dev), fmt, ##args); \
4296} while (0)
4297
f45f4321
JP
4298#define netif_level(level, priv, type, dev, fmt, args...) \
4299do { \
4300 if (netif_msg_##type(priv)) \
4301 netdev_##level(dev, fmt, ##args); \
4302} while (0)
4303
b3d95c5c 4304#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4305 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4306#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4307 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4308#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4309 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4310#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4311 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4312#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4313 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4314#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4315 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4316#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4317 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4318
0053ea9c 4319#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4320#define netif_dbg(priv, type, netdev, format, args...) \
4321do { \
4322 if (netif_msg_##type(priv)) \
b5fb0a03 4323 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4324} while (0)
0053ea9c
JP
4325#elif defined(DEBUG)
4326#define netif_dbg(priv, type, dev, format, args...) \
4327 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4328#else
4329#define netif_dbg(priv, type, dev, format, args...) \
4330({ \
4331 if (0) \
4332 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4333 0; \
4334})
4335#endif
4336
4337#if defined(VERBOSE_DEBUG)
bcfcc450 4338#define netif_vdbg netif_dbg
b3d95c5c
JP
4339#else
4340#define netif_vdbg(priv, type, dev, format, args...) \
4341({ \
4342 if (0) \
a4ed89cb 4343 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4344 0; \
4345})
4346#endif
571ba423 4347
900ff8c6
CW
4348/*
4349 * The list of packet types we will receive (as opposed to discard)
4350 * and the routines to invoke.
4351 *
4352 * Why 16. Because with 16 the only overlap we get on a hash of the
4353 * low nibble of the protocol value is RARP/SNAP/X.25.
4354 *
4355 * NOTE: That is no longer true with the addition of VLAN tags. Not
4356 * sure which should go first, but I bet it won't make much
4357 * difference if we are running VLANs. The good news is that
4358 * this protocol won't be in the list unless compiled in, so
4359 * the average user (w/out VLANs) will not be adversely affected.
4360 * --BLG
4361 *
4362 * 0800 IP
4363 * 8100 802.1Q VLAN
4364 * 0001 802.3
4365 * 0002 AX.25
4366 * 0004 802.2
4367 * 8035 RARP
4368 * 0005 SNAP
4369 * 0805 X.25
4370 * 0806 ARP
4371 * 8137 IPX
4372 * 0009 Localtalk
4373 * 86DD IPv6
4374 */
4375#define PTYPE_HASH_SIZE (16)
4376#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4377
385a154c 4378#endif /* _LINUX_NETDEVICE_H */