net: busy-poll: remove LL_FLUSH_FAILED and LL_FLUSH_BUSY
[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 466bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
467/**
468 * napi_complete - NAPI processing complete
5e82b4b2 469 * @n: NAPI context
bea3348e
SH
470 *
471 * Mark NAPI processing as complete.
3b47d303 472 * Consider using napi_complete_done() instead.
364b6055 473 * Return false if device should avoid rearming interrupts.
bea3348e 474 */
364b6055 475static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
476{
477 return napi_complete_done(n, 0);
478}
bea3348e 479
af12fa6e
ET
480/**
481 * napi_hash_del - remove a NAPI from global table
5e82b4b2 482 * @napi: NAPI context
af12fa6e 483 *
5e82b4b2 484 * Warning: caller must observe RCU grace period
34cbe27e
ED
485 * before freeing memory containing @napi, if
486 * this function returns true.
93d05d4a 487 * Note: core networking stack automatically calls it
5e82b4b2 488 * from netif_napi_del().
93d05d4a 489 * Drivers might want to call this helper to combine all
5e82b4b2 490 * the needed RCU grace periods into a single one.
af12fa6e 491 */
34cbe27e 492bool napi_hash_del(struct napi_struct *napi);
af12fa6e 493
bea3348e
SH
494/**
495 * napi_disable - prevent NAPI from scheduling
5e82b4b2 496 * @n: NAPI context
bea3348e
SH
497 *
498 * Stop NAPI from being scheduled on this context.
499 * Waits till any outstanding processing completes.
500 */
3b47d303 501void napi_disable(struct napi_struct *n);
bea3348e
SH
502
503/**
504 * napi_enable - enable NAPI scheduling
5e82b4b2 505 * @n: NAPI context
bea3348e
SH
506 *
507 * Resume NAPI from being scheduled on this context.
508 * Must be paired with napi_disable.
509 */
510static inline void napi_enable(struct napi_struct *n)
511{
512 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 513 smp_mb__before_atomic();
bea3348e 514 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 515 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
516}
517
c264c3de
SH
518/**
519 * napi_synchronize - wait until NAPI is not running
5e82b4b2 520 * @n: NAPI context
c264c3de
SH
521 *
522 * Wait until NAPI is done being scheduled on this context.
523 * Waits till any outstanding processing completes but
524 * does not disable future activations.
525 */
526static inline void napi_synchronize(const struct napi_struct *n)
527{
facc432f
AB
528 if (IS_ENABLED(CONFIG_SMP))
529 while (test_bit(NAPI_STATE_SCHED, &n->state))
530 msleep(1);
531 else
532 barrier();
c264c3de 533}
c264c3de 534
d94d9fee 535enum netdev_queue_state_t {
73466498
TH
536 __QUEUE_STATE_DRV_XOFF,
537 __QUEUE_STATE_STACK_XOFF,
c3f26a26 538 __QUEUE_STATE_FROZEN,
79d16385 539};
8e2f1a63
DB
540
541#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
542#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
543#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
544
545#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
546#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
547 QUEUE_STATE_FROZEN)
548#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
549 QUEUE_STATE_FROZEN)
550
73466498
TH
551/*
552 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
553 * netif_tx_* functions below are used to manipulate this flag. The
554 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
555 * queue independently. The netif_xmit_*stopped functions below are called
556 * to check if the queue has been stopped by the driver or stack (either
557 * of the XOFF bits are set in the state). Drivers should not need to call
558 * netif_xmit*stopped functions, they should only be using netif_tx_*.
559 */
79d16385 560
bb949fbd 561struct netdev_queue {
6a321cb3 562/*
5e82b4b2 563 * read-mostly part
6a321cb3 564 */
bb949fbd 565 struct net_device *dev;
46e5da40 566 struct Qdisc __rcu *qdisc;
b0e1e646 567 struct Qdisc *qdisc_sleeping;
ccf5ff69 568#ifdef CONFIG_SYSFS
1d24eb48
TH
569 struct kobject kobj;
570#endif
f2cd2d3e
ED
571#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
572 int numa_node;
573#endif
c0ef079c
FW
574 unsigned long tx_maxrate;
575 /*
576 * Number of TX timeouts for this queue
577 * (/sys/class/net/DEV/Q/trans_timeout)
578 */
579 unsigned long trans_timeout;
6a321cb3 580/*
5e82b4b2 581 * write-mostly part
6a321cb3
ED
582 */
583 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
584 int xmit_lock_owner;
9d21493b 585 /*
9b36627a 586 * Time (in jiffies) of last Tx
9d21493b
ED
587 */
588 unsigned long trans_start;
ccf5ff69 589
114cf580
TH
590 unsigned long state;
591
592#ifdef CONFIG_BQL
593 struct dql dql;
594#endif
e8a0464c 595} ____cacheline_aligned_in_smp;
bb949fbd 596
f2cd2d3e
ED
597static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 return q->numa_node;
601#else
b236da69 602 return NUMA_NO_NODE;
f2cd2d3e
ED
603#endif
604}
605
606static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
607{
608#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
609 q->numa_node = node;
610#endif
611}
612
df334545 613#ifdef CONFIG_RPS
0a9627f2
TH
614/*
615 * This structure holds an RPS map which can be of variable length. The
616 * map is an array of CPUs.
617 */
618struct rps_map {
619 unsigned int len;
620 struct rcu_head rcu;
621 u16 cpus[0];
622};
60b778ce 623#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 624
fec5e652 625/*
c445477d
BH
626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627 * tail pointer for that CPU's input queue at the time of last enqueue, and
628 * a hardware filter index.
fec5e652
TH
629 */
630struct rps_dev_flow {
631 u16 cpu;
c445477d 632 u16 filter;
fec5e652
TH
633 unsigned int last_qtail;
634};
c445477d 635#define RPS_NO_FILTER 0xffff
fec5e652
TH
636
637/*
638 * The rps_dev_flow_table structure contains a table of flow mappings.
639 */
640struct rps_dev_flow_table {
641 unsigned int mask;
642 struct rcu_head rcu;
fec5e652
TH
643 struct rps_dev_flow flows[0];
644};
645#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 646 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
647
648/*
649 * The rps_sock_flow_table contains mappings of flows to the last CPU
650 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
651 * Each entry is a 32bit value. Upper part is the high-order bits
652 * of flow hash, lower part is CPU number.
567e4b79 653 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
654 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 656 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
657 */
658struct rps_sock_flow_table {
567e4b79 659 u32 mask;
93c1af6c
ED
660
661 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 662};
567e4b79 663#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
664
665#define RPS_NO_CPU 0xffff
666
567e4b79
ED
667extern u32 rps_cpu_mask;
668extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
669
fec5e652
TH
670static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
671 u32 hash)
672{
673 if (table && hash) {
567e4b79
ED
674 unsigned int index = hash & table->mask;
675 u32 val = hash & ~rps_cpu_mask;
fec5e652 676
5e82b4b2 677 /* We only give a hint, preemption can change CPU under us */
567e4b79 678 val |= raw_smp_processor_id();
fec5e652 679
567e4b79
ED
680 if (table->ents[index] != val)
681 table->ents[index] = val;
fec5e652
TH
682 }
683}
684
c445477d 685#ifdef CONFIG_RFS_ACCEL
f629d208
JP
686bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
687 u16 filter_id);
c445477d 688#endif
a953be53 689#endif /* CONFIG_RPS */
c445477d 690
0a9627f2
TH
691/* This structure contains an instance of an RX queue. */
692struct netdev_rx_queue {
a953be53 693#ifdef CONFIG_RPS
6e3f7faf
ED
694 struct rps_map __rcu *rps_map;
695 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 696#endif
6e3f7faf 697 struct kobject kobj;
fe822240 698 struct net_device *dev;
0a9627f2 699} ____cacheline_aligned_in_smp;
a953be53
MD
700
701/*
702 * RX queue sysfs structures and functions.
703 */
704struct rx_queue_attribute {
705 struct attribute attr;
706 ssize_t (*show)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, char *buf);
708 ssize_t (*store)(struct netdev_rx_queue *queue,
709 struct rx_queue_attribute *attr, const char *buf, size_t len);
710};
d314774c 711
bf264145
TH
712#ifdef CONFIG_XPS
713/*
714 * This structure holds an XPS map which can be of variable length. The
715 * map is an array of queues.
716 */
717struct xps_map {
718 unsigned int len;
719 unsigned int alloc_len;
720 struct rcu_head rcu;
721 u16 queues[0];
722};
60b778ce 723#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
724#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
726
727/*
728 * This structure holds all XPS maps for device. Maps are indexed by CPU.
729 */
730struct xps_dev_maps {
731 struct rcu_head rcu;
a4177869 732 struct xps_map __rcu *cpu_map[0];
bf264145 733};
184c449f
AD
734#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
735 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
736#endif /* CONFIG_XPS */
737
4f57c087
JF
738#define TC_MAX_QUEUE 16
739#define TC_BITMASK 15
740/* HW offloaded queuing disciplines txq count and offset maps */
741struct netdev_tc_txq {
742 u16 count;
743 u16 offset;
744};
745
68bad94e
NP
746#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
747/*
748 * This structure is to hold information about the device
749 * configured to run FCoE protocol stack.
750 */
751struct netdev_fcoe_hbainfo {
752 char manufacturer[64];
753 char serial_number[64];
754 char hardware_version[64];
755 char driver_version[64];
756 char optionrom_version[64];
757 char firmware_version[64];
758 char model[256];
759 char model_description[256];
760};
761#endif
762
02637fce 763#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 764
02637fce
JP
765/* This structure holds a unique identifier to identify some
766 * physical item (port for example) used by a netdevice.
66b52b0d 767 */
02637fce
JP
768struct netdev_phys_item_id {
769 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
770 unsigned char id_len;
771};
772
d754f98b
SF
773static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
774 struct netdev_phys_item_id *b)
775{
776 return a->id_len == b->id_len &&
777 memcmp(a->id, b->id, a->id_len) == 0;
778}
779
99932d4f
DB
780typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
781 struct sk_buff *skb);
782
a1b7c5fd 783/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
784 * that are being passed to the netdevice through the setup_tc op.
785 */
786enum {
787 TC_SETUP_MQPRIO,
a1b7c5fd 788 TC_SETUP_CLSU32,
5b33f488 789 TC_SETUP_CLSFLOWER,
b87f7936 790 TC_SETUP_MATCHALL,
332ae8e2 791 TC_SETUP_CLSBPF,
16e5cc64
JF
792};
793
a1b7c5fd
JF
794struct tc_cls_u32_offload;
795
16e5cc64
JF
796struct tc_to_netdev {
797 unsigned int type;
798 union {
799 u8 tc;
a1b7c5fd 800 struct tc_cls_u32_offload *cls_u32;
5b33f488 801 struct tc_cls_flower_offload *cls_flower;
b87f7936 802 struct tc_cls_matchall_offload *cls_mall;
332ae8e2 803 struct tc_cls_bpf_offload *cls_bpf;
16e5cc64 804 };
7091d8c7 805 bool egress_dev;
16e5cc64
JF
806};
807
a7862b45
BB
808/* These structures hold the attributes of xdp state that are being passed
809 * to the netdevice through the xdp op.
810 */
811enum xdp_netdev_command {
812 /* Set or clear a bpf program used in the earliest stages of packet
813 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
814 * is responsible for calling bpf_prog_put on any old progs that are
815 * stored. In case of error, the callee need not release the new prog
816 * reference, but on success it takes ownership and must bpf_prog_put
817 * when it is no longer used.
818 */
819 XDP_SETUP_PROG,
820 /* Check if a bpf program is set on the device. The callee should
821 * return true if a program is currently attached and running.
822 */
823 XDP_QUERY_PROG,
824};
825
826struct netdev_xdp {
827 enum xdp_netdev_command command;
828 union {
829 /* XDP_SETUP_PROG */
830 struct bpf_prog *prog;
831 /* XDP_QUERY_PROG */
832 bool prog_attached;
833 };
834};
16e5cc64 835
d314774c
SH
836/*
837 * This structure defines the management hooks for network devices.
00829823
SH
838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
d314774c
SH
840 *
841 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
d314774c
SH
846 *
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
850 *
851 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 852 * This function is called when a network device transitions to the up
d314774c
SH
853 * state.
854 *
855 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 856 * This function is called when a network device transitions to the down
d314774c
SH
857 * state.
858 *
dc1f8bf6
SH
859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
00829823 861 * Called when a packet needs to be transmitted.
e79d8429
RR
862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 866 * Required; cannot be NULL.
00829823 867 *
1a2a1444
DM
868 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
869 * struct net_device *dev
870 * netdev_features_t features);
871 * Called by core transmit path to determine if device is capable of
872 * performing offload operations on a given packet. This is to give
873 * the device an opportunity to implement any restrictions that cannot
874 * be otherwise expressed by feature flags. The check is called with
875 * the set of features that the stack has calculated and it returns
876 * those the driver believes to be appropriate.
cdba756f 877 *
f663dd9a 878 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 879 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 880 * Called to decide which queue to use when device supports multiple
00829823
SH
881 * transmit queues.
882 *
d314774c
SH
883 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
884 * This function is called to allow device receiver to make
5e82b4b2 885 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
886 *
887 * void (*ndo_set_rx_mode)(struct net_device *dev);
888 * This function is called device changes address list filtering.
01789349 889 * If driver handles unicast address filtering, it should set
5e82b4b2 890 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
891 *
892 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
893 * This function is called when the Media Access Control address
37b607c5 894 * needs to be changed. If this interface is not defined, the
5e82b4b2 895 * MAC address can not be changed.
d314774c
SH
896 *
897 * int (*ndo_validate_addr)(struct net_device *dev);
898 * Test if Media Access Control address is valid for the device.
899 *
900 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
901 * Called when a user requests an ioctl which can't be handled by
902 * the generic interface code. If not defined ioctls return
d314774c
SH
903 * not supported error code.
904 *
905 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
906 * Used to set network devices bus interface parameters. This interface
5e82b4b2 907 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
908 * interface (PCI) for low level management.
909 *
910 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
911 * Called when a user wants to change the Maximum Transfer Unit
912 * of a device. If not defined, any request to change MTU will
913 * will return an error.
914 *
00829823 915 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 916 * Callback used when the transmitter has not made any progress
d314774c
SH
917 * for dev->watchdog ticks.
918 *
bc1f4470 919 * void (*ndo_get_stats64)(struct net_device *dev,
920 * struct rtnl_link_stats64 *storage);
d308e38f 921 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 922 * Called when a user wants to get the network device usage
be1f3c2c 923 * statistics. Drivers must do one of the following:
3cfde79c
BH
924 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
925 * rtnl_link_stats64 structure passed by the caller.
82695d9b 926 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
927 * (which should normally be dev->stats) and return a pointer to
928 * it. The structure may be changed asynchronously only if each
929 * field is written atomically.
930 * 3. Update dev->stats asynchronously and atomically, and define
931 * neither operation.
d314774c 932 *
3df5b3c6 933 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
934 * Return true if this device supports offload stats of this attr_id.
935 *
936 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
937 * void *attr_data)
938 * Get statistics for offload operations by attr_id. Write it into the
939 * attr_data pointer.
940 *
5d632cb7 941 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 942 * If device supports VLAN filtering this function is called when a
80d5c368 943 * VLAN id is registered.
d314774c 944 *
5d632cb7 945 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 946 * If device supports VLAN filtering this function is called when a
80d5c368 947 * VLAN id is unregistered.
d314774c
SH
948 *
949 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
950 *
951 * SR-IOV management functions.
952 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
953 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
954 * u8 qos, __be16 proto);
ed616689
SC
955 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
956 * int max_tx_rate);
5f8444a3 957 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 958 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
959 * int (*ndo_get_vf_config)(struct net_device *dev,
960 * int vf, struct ifla_vf_info *ivf);
1d8faf48 961 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
962 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
963 * struct nlattr *port[]);
01a3d796
VZ
964 *
965 * Enable or disable the VF ability to query its RSS Redirection Table and
966 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 967 * with PF and querying it may introduce a theoretical security risk.
01a3d796 968 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 969 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
6a4bc2b4
FF
970 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
971 * __be16 protocol, struct tc_to_netdev *tc);
972 * Called to setup any 'tc' scheduler, classifier or action on @dev.
973 * This is always called from the stack with the rtnl lock held and netif
974 * tx queues stopped. This allows the netdevice to perform queue
975 * management 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
bc1f4470 1169 void (*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);
d314774c 1186#endif
95c26df8
WM
1187 int (*ndo_set_vf_mac)(struct net_device *dev,
1188 int queue, u8 *mac);
1189 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1190 int queue, u16 vlan,
1191 u8 qos, __be16 proto);
ed616689
SC
1192 int (*ndo_set_vf_rate)(struct net_device *dev,
1193 int vf, int min_tx_rate,
1194 int max_tx_rate);
5f8444a3
GR
1195 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1196 int vf, bool setting);
dd461d6a
HS
1197 int (*ndo_set_vf_trust)(struct net_device *dev,
1198 int vf, bool setting);
95c26df8
WM
1199 int (*ndo_get_vf_config)(struct net_device *dev,
1200 int vf,
1201 struct ifla_vf_info *ivf);
1d8faf48
RE
1202 int (*ndo_set_vf_link_state)(struct net_device *dev,
1203 int vf, int link_state);
3b766cd8
EBE
1204 int (*ndo_get_vf_stats)(struct net_device *dev,
1205 int vf,
1206 struct ifla_vf_stats
1207 *vf_stats);
57b61080
SF
1208 int (*ndo_set_vf_port)(struct net_device *dev,
1209 int vf,
1210 struct nlattr *port[]);
1211 int (*ndo_get_vf_port)(struct net_device *dev,
1212 int vf, struct sk_buff *skb);
cc8e27cc
EC
1213 int (*ndo_set_vf_guid)(struct net_device *dev,
1214 int vf, u64 guid,
1215 int guid_type);
01a3d796
VZ
1216 int (*ndo_set_vf_rss_query_en)(
1217 struct net_device *dev,
1218 int vf, bool setting);
16e5cc64
JF
1219 int (*ndo_setup_tc)(struct net_device *dev,
1220 u32 handle,
1221 __be16 protocol,
1222 struct tc_to_netdev *tc);
d11ead75 1223#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1224 int (*ndo_fcoe_enable)(struct net_device *dev);
1225 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1226 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1227 u16 xid,
1228 struct scatterlist *sgl,
1229 unsigned int sgc);
1230 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1231 u16 xid);
6247e086
YZ
1232 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1233 u16 xid,
1234 struct scatterlist *sgl,
1235 unsigned int sgc);
68bad94e
NP
1236 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1237 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1238#endif
1239
d11ead75 1240#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1241#define NETDEV_FCOE_WWNN 0
1242#define NETDEV_FCOE_WWPN 1
1243 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1244 u64 *wwn, int type);
4d288d57 1245#endif
3c9c36bc 1246
c445477d
BH
1247#ifdef CONFIG_RFS_ACCEL
1248 int (*ndo_rx_flow_steer)(struct net_device *dev,
1249 const struct sk_buff *skb,
1250 u16 rxq_index,
1251 u32 flow_id);
1252#endif
fbaec0ea
JP
1253 int (*ndo_add_slave)(struct net_device *dev,
1254 struct net_device *slave_dev);
1255 int (*ndo_del_slave)(struct net_device *dev,
1256 struct net_device *slave_dev);
c8f44aff
MM
1257 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1258 netdev_features_t features);
5455c699 1259 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1260 netdev_features_t features);
503eebc2
JP
1261 int (*ndo_neigh_construct)(struct net_device *dev,
1262 struct neighbour *n);
1263 void (*ndo_neigh_destroy)(struct net_device *dev,
1264 struct neighbour *n);
77162022
JF
1265
1266 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1267 struct nlattr *tb[],
77162022 1268 struct net_device *dev,
6b6e2725 1269 const unsigned char *addr,
f6f6424b 1270 u16 vid,
77162022
JF
1271 u16 flags);
1272 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1273 struct nlattr *tb[],
77162022 1274 struct net_device *dev,
f6f6424b
JP
1275 const unsigned char *addr,
1276 u16 vid);
77162022
JF
1277 int (*ndo_fdb_dump)(struct sk_buff *skb,
1278 struct netlink_callback *cb,
1279 struct net_device *dev,
5d5eacb3 1280 struct net_device *filter_dev,
d297653d 1281 int *idx);
e5a55a89
JF
1282
1283 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1284 struct nlmsghdr *nlh,
1285 u16 flags);
e5a55a89
JF
1286 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1287 u32 pid, u32 seq,
6cbdceeb 1288 struct net_device *dev,
46c264da
ND
1289 u32 filter_mask,
1290 int nlflags);
407af329 1291 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1292 struct nlmsghdr *nlh,
1293 u16 flags);
4bf84c35
JP
1294 int (*ndo_change_carrier)(struct net_device *dev,
1295 bool new_carrier);
66b52b0d 1296 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1297 struct netdev_phys_item_id *ppid);
db24a904
DA
1298 int (*ndo_get_phys_port_name)(struct net_device *dev,
1299 char *name, size_t len);
7c46a640
AD
1300 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1301 struct udp_tunnel_info *ti);
1302 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1303 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1304 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1305 struct net_device *dev);
1306 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1307 void *priv);
1308
1309 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1310 struct net_device *dev,
1311 void *priv);
25175ba5 1312 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1313 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1314 int queue_index,
1315 u32 maxrate);
a54acb3a 1316 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1317 int (*ndo_change_proto_down)(struct net_device *dev,
1318 bool proto_down);
fc4099f1
PS
1319 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1320 struct sk_buff *skb);
871b642a
PA
1321 void (*ndo_set_rx_headroom)(struct net_device *dev,
1322 int needed_headroom);
a7862b45
BB
1323 int (*ndo_xdp)(struct net_device *dev,
1324 struct netdev_xdp *xdp);
d314774c
SH
1325};
1326
7aa98047
LR
1327/**
1328 * enum net_device_priv_flags - &struct net_device priv_flags
1329 *
1330 * These are the &struct net_device, they are only set internally
1331 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1332 * userspace; this means that the order of these flags can change
7aa98047
LR
1333 * during any kernel release.
1334 *
1335 * You should have a pretty good reason to be extending these flags.
1336 *
1337 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1338 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1339 * @IFF_BONDING: bonding master or slave
7aa98047 1340 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1341 * @IFF_WAN_HDLC: WAN HDLC device
1342 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1343 * release skb->dst
1344 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1345 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1346 * @IFF_MACVLAN_PORT: device used as macvlan port
1347 * @IFF_BRIDGE_PORT: device used as bridge port
1348 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1349 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1350 * @IFF_UNICAST_FLT: Supports unicast filtering
1351 * @IFF_TEAM_PORT: device used as team port
1352 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1353 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1354 * change when it's running
1355 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1356 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1357 * underlying stacked devices
1358 * @IFF_IPVLAN_MASTER: IPvlan master device
1359 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1360 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1361 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1362 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1363 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1364 * @IFF_TEAM: device is a team device
d4ab4286 1365 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1366 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1367 * entity (i.e. the master device for bridged veth)
3c175784 1368 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1369 */
1370enum netdev_priv_flags {
1371 IFF_802_1Q_VLAN = 1<<0,
1372 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1373 IFF_BONDING = 1<<2,
1374 IFF_ISATAP = 1<<3,
1375 IFF_WAN_HDLC = 1<<4,
1376 IFF_XMIT_DST_RELEASE = 1<<5,
1377 IFF_DONT_BRIDGE = 1<<6,
1378 IFF_DISABLE_NETPOLL = 1<<7,
1379 IFF_MACVLAN_PORT = 1<<8,
1380 IFF_BRIDGE_PORT = 1<<9,
1381 IFF_OVS_DATAPATH = 1<<10,
1382 IFF_TX_SKB_SHARING = 1<<11,
1383 IFF_UNICAST_FLT = 1<<12,
1384 IFF_TEAM_PORT = 1<<13,
1385 IFF_SUPP_NOFCS = 1<<14,
1386 IFF_LIVE_ADDR_CHANGE = 1<<15,
1387 IFF_MACVLAN = 1<<16,
1388 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1389 IFF_IPVLAN_MASTER = 1<<18,
1390 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1391 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1392 IFF_NO_QUEUE = 1<<21,
1393 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1394 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1395 IFF_TEAM = 1<<24,
d4ab4286 1396 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1397 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1398 IFF_MACSEC = 1<<27,
7aa98047
LR
1399};
1400
1401#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1402#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1403#define IFF_BONDING IFF_BONDING
7aa98047 1404#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1405#define IFF_WAN_HDLC IFF_WAN_HDLC
1406#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1407#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1408#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1409#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1410#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1411#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1412#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1413#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1414#define IFF_TEAM_PORT IFF_TEAM_PORT
1415#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1416#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1417#define IFF_MACVLAN IFF_MACVLAN
02875878 1418#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1419#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1420#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1421#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1422#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1423#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1424#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1425#define IFF_TEAM IFF_TEAM
d4ab4286 1426#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1427#define IFF_MACSEC IFF_MACSEC
7aa98047 1428
536721b1
KK
1429/**
1430 * struct net_device - The DEVICE structure.
1431 * Actually, this whole structure is a big mistake. It mixes I/O
1432 * data with strictly "high-level" data, and it has to know about
1433 * almost every data structure used in the INET module.
1434 *
1435 * @name: This is the first field of the "visible" part of this structure
1436 * (i.e. as seen by users in the "Space.c" file). It is the name
1437 * of the interface.
1438 *
1439 * @name_hlist: Device name hash chain, please keep it close to name[]
1440 * @ifalias: SNMP alias
1441 * @mem_end: Shared memory end
1442 * @mem_start: Shared memory start
1443 * @base_addr: Device I/O address
1444 * @irq: Device IRQ number
1445 *
14ffbbb8
TG
1446 * @carrier_changes: Stats to monitor carrier on<->off transitions
1447 *
536721b1
KK
1448 * @state: Generic network queuing layer state, see netdev_state_t
1449 * @dev_list: The global list of network devices
5e82b4b2
BH
1450 * @napi_list: List entry used for polling NAPI devices
1451 * @unreg_list: List entry when we are unregistering the
1452 * device; see the function unregister_netdev
1453 * @close_list: List entry used when we are closing the device
62d885fe
BP
1454 * @ptype_all: Device-specific packet handlers for all protocols
1455 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1456 *
1457 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1458 * @features: Currently active device features
1459 * @hw_features: User-changeable features
1460 *
1461 * @wanted_features: User-requested features
1462 * @vlan_features: Mask of features inheritable by VLAN devices
1463 *
1464 * @hw_enc_features: Mask of features inherited by encapsulating devices
1465 * This field indicates what encapsulation
1466 * offloads the hardware is capable of doing,
1467 * and drivers will need to set them appropriately.
1468 *
1469 * @mpls_features: Mask of features inheritable by MPLS
1470 *
1471 * @ifindex: interface index
5e82b4b2 1472 * @group: The group the device belongs to
536721b1
KK
1473 *
1474 * @stats: Statistics struct, which was left as a legacy, use
1475 * rtnl_link_stats64 instead
1476 *
1477 * @rx_dropped: Dropped packets by core network,
1478 * do not use this in drivers
1479 * @tx_dropped: Dropped packets by core network,
1480 * do not use this in drivers
6e7333d3
JW
1481 * @rx_nohandler: nohandler dropped packets by core network on
1482 * inactive devices, do not use this in drivers
536721b1 1483 *
536721b1
KK
1484 * @wireless_handlers: List of functions to handle Wireless Extensions,
1485 * instead of ioctl,
1486 * see <net/iw_handler.h> for details.
1487 * @wireless_data: Instance data managed by the core of wireless extensions
1488 *
1489 * @netdev_ops: Includes several pointers to callbacks,
1490 * if one wants to override the ndo_*() functions
1491 * @ethtool_ops: Management operations
f997c55c
AA
1492 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1493 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1494 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1495 * of Layer 2 headers.
1496 *
1497 * @flags: Interface flags (a la BSD)
1498 * @priv_flags: Like 'flags' but invisible to userspace,
1499 * see if.h for the definitions
1500 * @gflags: Global flags ( kept as legacy )
1501 * @padded: How much padding added by alloc_netdev()
1502 * @operstate: RFC2863 operstate
1503 * @link_mode: Mapping policy to operstate
1504 * @if_port: Selectable AUI, TP, ...
1505 * @dma: DMA channel
1506 * @mtu: Interface MTU value
61e84623
JW
1507 * @min_mtu: Interface Minimum MTU value
1508 * @max_mtu: Interface Maximum MTU value
536721b1 1509 * @type: Interface hardware type
2793a23a 1510 * @hard_header_len: Maximum hardware header length.
217e6fa2 1511 * @min_header_len: Minimum hardware header length
536721b1
KK
1512 *
1513 * @needed_headroom: Extra headroom the hardware may need, but not in all
1514 * cases can this be guaranteed
1515 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1516 * cases can this be guaranteed. Some cases also use
1517 * LL_MAX_HEADER instead to allocate the skb
1518 *
1519 * interface address info:
1520 *
1521 * @perm_addr: Permanent hw address
1522 * @addr_assign_type: Hw address assignment type
1523 * @addr_len: Hardware address length
8626a0c8 1524 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1525 * @dev_id: Used to differentiate devices that share
1526 * the same link layer address
1527 * @dev_port: Used to differentiate devices that share
1528 * the same function
1529 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1530 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1531 * has been enabled due to the need to listen to
1532 * additional unicast addresses in a device that
1533 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1534 * @uc: unicast mac addresses
1535 * @mc: multicast mac addresses
1536 * @dev_addrs: list of device hw addresses
1537 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1538 * @promiscuity: Number of times the NIC is told to work in
1539 * promiscuous mode; if it becomes 0 the NIC will
1540 * exit promiscuous mode
536721b1
KK
1541 * @allmulti: Counter, enables or disables allmulticast mode
1542 *
1543 * @vlan_info: VLAN info
1544 * @dsa_ptr: dsa specific data
1545 * @tipc_ptr: TIPC specific data
1546 * @atalk_ptr: AppleTalk link
1547 * @ip_ptr: IPv4 specific data
1548 * @dn_ptr: DECnet specific data
1549 * @ip6_ptr: IPv6 specific data
1550 * @ax25_ptr: AX.25 specific data
1551 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1552 *
536721b1
KK
1553 * @dev_addr: Hw address (before bcast,
1554 * because most packets are unicast)
1555 *
1556 * @_rx: Array of RX queues
1557 * @num_rx_queues: Number of RX queues
1558 * allocated at register_netdev() time
1559 * @real_num_rx_queues: Number of RX queues currently active in device
1560 *
1561 * @rx_handler: handler for received packets
1562 * @rx_handler_data: XXX: need comments on this one
1563 * @ingress_queue: XXX: need comments on this one
1564 * @broadcast: hw bcast address
1565 *
14ffbbb8
TG
1566 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1567 * indexed by RX queue number. Assigned by driver.
1568 * This must only be set if the ndo_rx_flow_steer
1569 * operation is defined
1570 * @index_hlist: Device index hash chain
1571 *
536721b1
KK
1572 * @_tx: Array of TX queues
1573 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1574 * @real_num_tx_queues: Number of TX queues currently active in device
1575 * @qdisc: Root qdisc from userspace point of view
1576 * @tx_queue_len: Max frames per queue allowed
1577 * @tx_global_lock: XXX: need comments on this one
1578 *
1579 * @xps_maps: XXX: need comments on this one
1580 *
536721b1 1581 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1582 * the watchdog (see dev_watchdog())
536721b1
KK
1583 * @watchdog_timer: List of timers
1584 *
1585 * @pcpu_refcnt: Number of references to this device
1586 * @todo_list: Delayed register/unregister
536721b1
KK
1587 * @link_watch_list: XXX: need comments on this one
1588 *
1589 * @reg_state: Register/unregister state machine
1590 * @dismantle: Device is going to be freed
1591 * @rtnl_link_state: This enum represents the phases of creating
1592 * a new link
1593 *
1594 * @destructor: Called from unregister,
1595 * can be used to call free_netdev
1596 * @npinfo: XXX: need comments on this one
1597 * @nd_net: Network namespace this network device is inside
1598 *
1599 * @ml_priv: Mid-layer private
1600 * @lstats: Loopback statistics
1601 * @tstats: Tunnel statistics
1602 * @dstats: Dummy statistics
1603 * @vstats: Virtual ethernet statistics
1604 *
1605 * @garp_port: GARP
1606 * @mrp_port: MRP
1607 *
1608 * @dev: Class/net/name entry
1609 * @sysfs_groups: Space for optional device, statistics and wireless
1610 * sysfs groups
1611 *
1612 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1613 * @rtnl_link_ops: Rtnl_link_ops
1614 *
1615 * @gso_max_size: Maximum size of generic segmentation offload
1616 * @gso_max_segs: Maximum number of segments that can be passed to the
1617 * NIC for GSO
1618 *
1619 * @dcbnl_ops: Data Center Bridging netlink ops
1620 * @num_tc: Number of traffic classes in the net device
1621 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1622 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1623 *
1624 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1625 *
1626 * @priomap: XXX: need comments on this one
1627 * @phydev: Physical device may attach itself
1628 * for hardware timestamping
1629 *
123b3652
ED
1630 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1631 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1632 *
d746d707
AK
1633 * @proto_down: protocol port state information can be sent to the
1634 * switch driver and used to set the phys state of the
1635 * switch port.
1636 *
1da177e4
LT
1637 * FIXME: cleanup struct net_device such that network protocol info
1638 * moves out.
1639 */
1640
d94d9fee 1641struct net_device {
1da177e4 1642 char name[IFNAMSIZ];
9356b8fc 1643 struct hlist_node name_hlist;
0b815a1a 1644 char *ifalias;
1da177e4
LT
1645 /*
1646 * I/O specific fields
1647 * FIXME: Merge these and struct ifmap into one
1648 */
536721b1
KK
1649 unsigned long mem_end;
1650 unsigned long mem_start;
1651 unsigned long base_addr;
1652 int irq;
1da177e4 1653
14ffbbb8
TG
1654 atomic_t carrier_changes;
1655
1da177e4 1656 /*
536721b1
KK
1657 * Some hardware also needs these fields (state,dev_list,
1658 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1659 * part of the usual set specified in Space.c.
1660 */
1661
1da177e4
LT
1662 unsigned long state;
1663
7562f876 1664 struct list_head dev_list;
bea3348e 1665 struct list_head napi_list;
44a0873d 1666 struct list_head unreg_list;
5cde2829 1667 struct list_head close_list;
7866a621
SN
1668 struct list_head ptype_all;
1669 struct list_head ptype_specific;
2f268f12 1670
2f268f12
VF
1671 struct {
1672 struct list_head upper;
1673 struct list_head lower;
1674 } adj_list;
1675
c8f44aff 1676 netdev_features_t features;
c8f44aff 1677 netdev_features_t hw_features;
c8f44aff 1678 netdev_features_t wanted_features;
c8f44aff 1679 netdev_features_t vlan_features;
6a674e9c 1680 netdev_features_t hw_enc_features;
0d89d203 1681 netdev_features_t mpls_features;
802ab55a 1682 netdev_features_t gso_partial_features;
04ed3e74 1683
1da177e4 1684 int ifindex;
7a66bbc9 1685 int group;
1da177e4 1686
c45d286e 1687 struct net_device_stats stats;
015f0688 1688
015f0688
ED
1689 atomic_long_t rx_dropped;
1690 atomic_long_t tx_dropped;
6e7333d3 1691 atomic_long_t rx_nohandler;
1da177e4 1692
b86e0280 1693#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1694 const struct iw_handler_def *wireless_handlers;
1695 struct iw_public_data *wireless_data;
b86e0280 1696#endif
d314774c 1697 const struct net_device_ops *netdev_ops;
76fd8593 1698 const struct ethtool_ops *ethtool_ops;
4170604f 1699#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1700 const struct switchdev_ops *switchdev_ops;
4170604f 1701#endif
1b69c6d0
DA
1702#ifdef CONFIG_NET_L3_MASTER_DEV
1703 const struct l3mdev_ops *l3mdev_ops;
1704#endif
f997c55c
AA
1705#if IS_ENABLED(CONFIG_IPV6)
1706 const struct ndisc_ops *ndisc_ops;
1707#endif
1da177e4 1708
3b04ddde
SH
1709 const struct header_ops *header_ops;
1710
536721b1
KK
1711 unsigned int flags;
1712 unsigned int priv_flags;
1713
1da177e4 1714 unsigned short gflags;
536721b1 1715 unsigned short padded;
1da177e4 1716
536721b1
KK
1717 unsigned char operstate;
1718 unsigned char link_mode;
b00055aa 1719
536721b1
KK
1720 unsigned char if_port;
1721 unsigned char dma;
bdc220da 1722
536721b1 1723 unsigned int mtu;
61e84623
JW
1724 unsigned int min_mtu;
1725 unsigned int max_mtu;
536721b1
KK
1726 unsigned short type;
1727 unsigned short hard_header_len;
217e6fa2 1728 unsigned short min_header_len;
1da177e4 1729
f5184d26
JB
1730 unsigned short needed_headroom;
1731 unsigned short needed_tailroom;
1732
1da177e4 1733 /* Interface address info. */
536721b1
KK
1734 unsigned char perm_addr[MAX_ADDR_LEN];
1735 unsigned char addr_assign_type;
1736 unsigned char addr_len;
a0a9663d 1737 unsigned short neigh_priv_len;
536721b1
KK
1738 unsigned short dev_id;
1739 unsigned short dev_port;
ccffad25 1740 spinlock_t addr_list_lock;
14ffbbb8
TG
1741 unsigned char name_assign_type;
1742 bool uc_promisc;
536721b1
KK
1743 struct netdev_hw_addr_list uc;
1744 struct netdev_hw_addr_list mc;
1745 struct netdev_hw_addr_list dev_addrs;
1746
4c3d5e7b
ED
1747#ifdef CONFIG_SYSFS
1748 struct kset *queues_kset;
1749#endif
9d45abe1
WC
1750 unsigned int promiscuity;
1751 unsigned int allmulti;
1da177e4 1752
1da177e4 1753
5e82b4b2 1754 /* Protocol-specific pointers */
65ac6a5f 1755
d11ead75 1756#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1757 struct vlan_info __rcu *vlan_info;
65ac6a5f 1758#endif
34a430d7 1759#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1760 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1761#endif
1762#if IS_ENABLED(CONFIG_TIPC)
536721b1 1763 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1764#endif
536721b1
KK
1765 void *atalk_ptr;
1766 struct in_device __rcu *ip_ptr;
1767 struct dn_dev __rcu *dn_ptr;
1768 struct inet6_dev __rcu *ip6_ptr;
1769 void *ax25_ptr;
1770 struct wireless_dev *ieee80211_ptr;
98a18b6f 1771 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1772#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1773 struct mpls_dev __rcu *mpls_ptr;
1774#endif
1da177e4 1775
9356b8fc 1776/*
cd13539b 1777 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1778 */
9356b8fc 1779 /* Interface address info used in eth_type_trans() */
536721b1 1780 unsigned char *dev_addr;
f001fde5 1781
a953be53 1782#ifdef CONFIG_SYSFS
0a9627f2
TH
1783 struct netdev_rx_queue *_rx;
1784
0a9627f2 1785 unsigned int num_rx_queues;
62fe0b40 1786 unsigned int real_num_rx_queues;
df334545 1787#endif
0a9627f2 1788
3b47d303 1789 unsigned long gro_flush_timeout;
61391cde 1790 rx_handler_func_t __rcu *rx_handler;
1791 void __rcu *rx_handler_data;
e8a0464c 1792
4cda01e8 1793#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1794 struct tcf_proto __rcu *ingress_cl_list;
1795#endif
24824a09 1796 struct netdev_queue __rcu *ingress_queue;
e687ad60 1797#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1798 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1799#endif
d2788d34 1800
536721b1 1801 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1802#ifdef CONFIG_RFS_ACCEL
1803 struct cpu_rmap *rx_cpu_rmap;
1804#endif
1805 struct hlist_node index_hlist;
cd13539b
ED
1806
1807/*
1808 * Cache lines mostly used on transmit path
1809 */
e8a0464c
DM
1810 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1811 unsigned int num_tx_queues;
fd2ea0a7 1812 unsigned int real_num_tx_queues;
af356afa 1813 struct Qdisc *qdisc;
59cc1f61
JK
1814#ifdef CONFIG_NET_SCHED
1815 DECLARE_HASHTABLE (qdisc_hash, 4);
1816#endif
536721b1 1817 unsigned long tx_queue_len;
c3f26a26 1818 spinlock_t tx_global_lock;
14ffbbb8 1819 int watchdog_timeo;
cd13539b 1820
bf264145 1821#ifdef CONFIG_XPS
a4177869 1822 struct xps_dev_maps __rcu *xps_maps;
bf264145 1823#endif
1f211a1b
DB
1824#ifdef CONFIG_NET_CLS_ACT
1825 struct tcf_proto __rcu *egress_cl_list;
1826#endif
0c4f691f 1827
9356b8fc 1828 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1829 struct timer_list watchdog_timer;
1830
29b4433d 1831 int __percpu *pcpu_refcnt;
1da177e4 1832 struct list_head todo_list;
1da177e4 1833
e014debe 1834 struct list_head link_watch_list;
572a103d 1835
1da177e4 1836 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1837 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1838 NETREG_UNREGISTERING, /* called unregister_netdevice */
1839 NETREG_UNREGISTERED, /* completed unregister todo */
1840 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1841 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1842 } reg_state:8;
1843
536721b1 1844 bool dismantle;
a2835763
PM
1845
1846 enum {
1847 RTNL_LINK_INITIALIZED,
1848 RTNL_LINK_INITIALIZING,
1849 } rtnl_link_state:16;
1da177e4 1850
d314774c 1851 void (*destructor)(struct net_device *dev);
1da177e4 1852
1da177e4 1853#ifdef CONFIG_NETPOLL
5fbee843 1854 struct netpoll_info __rcu *npinfo;
1da177e4 1855#endif
eae792b7 1856
0c5c9fb5 1857 possible_net_t nd_net;
4a1c5371 1858
4951704b 1859 /* mid-layer private */
a7855c78 1860 union {
536721b1
KK
1861 void *ml_priv;
1862 struct pcpu_lstats __percpu *lstats;
8f84985f 1863 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1864 struct pcpu_dstats __percpu *dstats;
1865 struct pcpu_vstats __percpu *vstats;
a7855c78 1866 };
536721b1 1867
3cc77ec7 1868 struct garp_port __rcu *garp_port;
febf018d 1869 struct mrp_port __rcu *mrp_port;
1da177e4 1870
5e82b4b2 1871 struct device dev;
0c509a6c 1872 const struct attribute_group *sysfs_groups[4];
a953be53 1873 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1874
38f7b870 1875 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1876
82cc1a7a
PWJ
1877 /* for setting kernel sock attribute on TCP connection setup */
1878#define GSO_MAX_SIZE 65536
1879 unsigned int gso_max_size;
30b678d8
BH
1880#define GSO_MAX_SEGS 65535
1881 u16 gso_max_segs;
743b03a8 1882
7a6b6f51 1883#ifdef CONFIG_DCB
32953543 1884 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1885#endif
5e82b4b2
BH
1886 u8 num_tc;
1887 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1888 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1889
d11ead75 1890#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1891 unsigned int fcoe_ddp_xid;
5bc1421e 1892#endif
86f8515f 1893#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1894 struct netprio_map __rcu *priomap;
4d288d57 1895#endif
5e82b4b2
BH
1896 struct phy_device *phydev;
1897 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1898 struct lock_class_key *qdisc_running_key;
5e82b4b2 1899 bool proto_down;
1da177e4 1900};
43cb76d9 1901#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1902
1903#define NETDEV_ALIGN 32
1da177e4 1904
4f57c087
JF
1905static inline
1906int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1907{
1908 return dev->prio_tc_map[prio & TC_BITMASK];
1909}
1910
1911static inline
1912int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1913{
1914 if (tc >= dev->num_tc)
1915 return -EINVAL;
1916
1917 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1918 return 0;
1919}
1920
8d059b0f 1921int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1922void netdev_reset_tc(struct net_device *dev);
1923int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1924int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1925
1926static inline
1927int netdev_get_num_tc(struct net_device *dev)
1928{
1929 return dev->num_tc;
1930}
1931
e8a0464c
DM
1932static inline
1933struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1934 unsigned int index)
1935{
1936 return &dev->_tx[index];
1937}
1938
10c51b56
DB
1939static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1940 const struct sk_buff *skb)
1941{
1942 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1943}
1944
e8a0464c
DM
1945static inline void netdev_for_each_tx_queue(struct net_device *dev,
1946 void (*f)(struct net_device *,
1947 struct netdev_queue *,
1948 void *),
1949 void *arg)
1950{
1951 unsigned int i;
1952
1953 for (i = 0; i < dev->num_tx_queues; i++)
1954 f(dev, &dev->_tx[i], arg);
1955}
1956
d3fff6c4
ED
1957#define netdev_lockdep_set_classes(dev) \
1958{ \
1959 static struct lock_class_key qdisc_tx_busylock_key; \
1960 static struct lock_class_key qdisc_running_key; \
1961 static struct lock_class_key qdisc_xmit_lock_key; \
1962 static struct lock_class_key dev_addr_list_lock_key; \
1963 unsigned int i; \
1964 \
1965 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1966 (dev)->qdisc_running_key = &qdisc_running_key; \
1967 lockdep_set_class(&(dev)->addr_list_lock, \
1968 &dev_addr_list_lock_key); \
1969 for (i = 0; i < (dev)->num_tx_queues; i++) \
1970 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1971 &qdisc_xmit_lock_key); \
1972}
1973
f629d208 1974struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1975 struct sk_buff *skb,
1976 void *accel_priv);
8c4c49df 1977
871b642a
PA
1978/* returns the headroom that the master device needs to take in account
1979 * when forwarding to this dev
1980 */
1981static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1982{
1983 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1984}
1985
1986static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1987{
1988 if (dev->netdev_ops->ndo_set_rx_headroom)
1989 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1990}
1991
1992/* set the device rx headroom to the dev's default */
1993static inline void netdev_reset_rx_headroom(struct net_device *dev)
1994{
1995 netdev_set_rx_headroom(dev, -1);
1996}
1997
c346dca1
YH
1998/*
1999 * Net namespace inlines
2000 */
2001static inline
2002struct net *dev_net(const struct net_device *dev)
2003{
c2d9ba9b 2004 return read_pnet(&dev->nd_net);
c346dca1
YH
2005}
2006
2007static inline
f5aa23fd 2008void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2009{
0c5c9fb5 2010 write_pnet(&dev->nd_net, net);
c346dca1
YH
2011}
2012
3e8a72d1 2013static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 2014{
3fc88677 2015#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
2016 if (dev->dsa_ptr != NULL)
2017 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 2018#endif
5aed85ce 2019 return false;
396138f0
LB
2020}
2021
bea3348e
SH
2022/**
2023 * netdev_priv - access network device private data
2024 * @dev: network device
2025 *
2026 * Get network device private data
2027 */
6472ce60 2028static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2029{
1ce8e7b5 2030 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2031}
2032
1da177e4
LT
2033/* Set the sysfs physical device reference for the network logical device
2034 * if set prior to registration will cause a symlink during initialization.
2035 */
43cb76d9 2036#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2037
384912ed 2038/* Set the sysfs device type for the network logical device to allow
3f79410c 2039 * fine-grained identification of different network device types. For
5e82b4b2 2040 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2041 */
2042#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2043
82dc3c63
ED
2044/* Default NAPI poll() weight
2045 * Device drivers are strongly advised to not use bigger value
2046 */
2047#define NAPI_POLL_WEIGHT 64
2048
3b582cc1 2049/**
5e82b4b2 2050 * netif_napi_add - initialize a NAPI context
3b582cc1 2051 * @dev: network device
5e82b4b2 2052 * @napi: NAPI context
3b582cc1
SH
2053 * @poll: polling function
2054 * @weight: default weight
2055 *
5e82b4b2
BH
2056 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2057 * *any* of the other NAPI-related functions.
3b582cc1 2058 */
d565b0a1
HX
2059void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2060 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2061
d64b5e85 2062/**
5e82b4b2 2063 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2064 * @dev: network device
5e82b4b2 2065 * @napi: NAPI context
d64b5e85
ED
2066 * @poll: polling function
2067 * @weight: default weight
2068 *
2069 * This variant of netif_napi_add() should be used from drivers using NAPI
2070 * to exclusively poll a TX queue.
2071 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2072 */
2073static inline void netif_tx_napi_add(struct net_device *dev,
2074 struct napi_struct *napi,
2075 int (*poll)(struct napi_struct *, int),
2076 int weight)
2077{
2078 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2079 netif_napi_add(dev, napi, poll, weight);
2080}
2081
d8156534 2082/**
5e82b4b2
BH
2083 * netif_napi_del - remove a NAPI context
2084 * @napi: NAPI context
d8156534 2085 *
5e82b4b2 2086 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2087 */
d565b0a1
HX
2088void netif_napi_del(struct napi_struct *napi);
2089
2090struct napi_gro_cb {
78a478d0 2091 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2092 void *frag0;
78a478d0 2093
7489594c
HX
2094 /* Length of frag0. */
2095 unsigned int frag0_len;
2096
86911732 2097 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2098 int data_offset;
86911732 2099
d565b0a1 2100 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2101 u16 flush;
2102
2103 /* Save the IP ID here and check when we get to the transport layer */
2104 u16 flush_id;
d565b0a1
HX
2105
2106 /* Number of segments aggregated. */
2e71a6f8
ED
2107 u16 count;
2108
15e2396d
TH
2109 /* Start offset for remote checksum offload */
2110 u16 gro_remcsum_start;
2111
2e71a6f8
ED
2112 /* jiffies when first packet was created/queued */
2113 unsigned long age;
86347245 2114
afe93325 2115 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2116 u16 proto;
2117
baa32ff4
TH
2118 /* This is non-zero if the packet may be of the same flow. */
2119 u8 same_flow:1;
2120
fac8e0f5
JG
2121 /* Used in tunnel GRO receive */
2122 u8 encap_mark:1;
573e8fca
TH
2123
2124 /* GRO checksum is valid */
2125 u8 csum_valid:1;
2126
662880f4
TH
2127 /* Number of checksums via CHECKSUM_UNNECESSARY */
2128 u8 csum_cnt:3;
c3c7c254 2129
baa32ff4
TH
2130 /* Free the skb? */
2131 u8 free:2;
2132#define NAPI_GRO_FREE 1
2133#define NAPI_GRO_FREE_STOLEN_HEAD 2
2134
efc98d08
TH
2135 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2136 u8 is_ipv6:1;
2137
a0ca153f
AD
2138 /* Used in GRE, set in fou/gue_gro_receive */
2139 u8 is_fou:1;
2140
1530545e
AD
2141 /* Used to determine if flush_id can be ignored */
2142 u8 is_atomic:1;
2143
fcd91dd4
SD
2144 /* Number of gro_receive callbacks this packet already went through */
2145 u8 recursion_counter:4;
2146
2147 /* 1 bit hole */
baa32ff4 2148
bf5a755f
JC
2149 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2150 __wsum csum;
2151
c3c7c254
ED
2152 /* used in skb_gro_receive() slow path */
2153 struct sk_buff *last;
d565b0a1
HX
2154};
2155
2156#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2157
fcd91dd4
SD
2158#define GRO_RECURSION_LIMIT 15
2159static inline int gro_recursion_inc_test(struct sk_buff *skb)
2160{
2161 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2162}
2163
2164typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2165static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2166 struct sk_buff **head,
2167 struct sk_buff *skb)
2168{
2169 if (unlikely(gro_recursion_inc_test(skb))) {
2170 NAPI_GRO_CB(skb)->flush |= 1;
2171 return NULL;
2172 }
2173
2174 return cb(head, skb);
2175}
2176
2177typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2178 struct sk_buff *);
2179static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2180 struct sock *sk,
2181 struct sk_buff **head,
2182 struct sk_buff *skb)
2183{
2184 if (unlikely(gro_recursion_inc_test(skb))) {
2185 NAPI_GRO_CB(skb)->flush |= 1;
2186 return NULL;
2187 }
2188
2189 return cb(sk, head, skb);
2190}
2191
1da177e4 2192struct packet_type {
f2ccd8fa
DM
2193 __be16 type; /* This is really htons(ether_type). */
2194 struct net_device *dev; /* NULL is wildcarded here */
2195 int (*func) (struct sk_buff *,
2196 struct net_device *,
2197 struct packet_type *,
2198 struct net_device *);
c0de08d0
EL
2199 bool (*id_match)(struct packet_type *ptype,
2200 struct sock *sk);
1da177e4
LT
2201 void *af_packet_priv;
2202 struct list_head list;
2203};
2204
f191a1d1 2205struct offload_callbacks {
576a30eb 2206 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2207 netdev_features_t features);
d565b0a1 2208 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2209 struct sk_buff *skb);
299603e8 2210 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2211};
2212
2213struct packet_offload {
2214 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2215 u16 priority;
f191a1d1
VY
2216 struct offload_callbacks callbacks;
2217 struct list_head list;
1da177e4
LT
2218};
2219
5e82b4b2 2220/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2221struct pcpu_sw_netstats {
2222 u64 rx_packets;
2223 u64 rx_bytes;
2224 u64 tx_packets;
2225 u64 tx_bytes;
2226 struct u64_stats_sync syncp;
2227};
2228
aabc92bb
PNA
2229#define __netdev_alloc_pcpu_stats(type, gfp) \
2230({ \
2231 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2232 if (pcpu_stats) { \
2233 int __cpu; \
2234 for_each_possible_cpu(__cpu) { \
2235 typeof(type) *stat; \
2236 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2237 u64_stats_init(&stat->syncp); \
2238 } \
2239 } \
2240 pcpu_stats; \
1c213bd2
WC
2241})
2242
aabc92bb 2243#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2244 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2245
764f5e54
JP
2246enum netdev_lag_tx_type {
2247 NETDEV_LAG_TX_TYPE_UNKNOWN,
2248 NETDEV_LAG_TX_TYPE_RANDOM,
2249 NETDEV_LAG_TX_TYPE_BROADCAST,
2250 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2251 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2252 NETDEV_LAG_TX_TYPE_HASH,
2253};
2254
2255struct netdev_lag_upper_info {
2256 enum netdev_lag_tx_type tx_type;
2257};
2258
fb1b2e3c
JP
2259struct netdev_lag_lower_state_info {
2260 u8 link_up : 1,
2261 tx_enabled : 1;
2262};
2263
1da177e4
LT
2264#include <linux/notifier.h>
2265
dcfe1421
AW
2266/* netdevice notifier chain. Please remember to update the rtnetlink
2267 * notification exclusion list in rtnetlink_event() when adding new
2268 * types.
2269 */
2270#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2271#define NETDEV_DOWN 0x0002
2272#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2273 detected a hardware crash and restarted
2274 - we can use this eg to kick tcp sessions
2275 once done */
2276#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2277#define NETDEV_REGISTER 0x0005
2278#define NETDEV_UNREGISTER 0x0006
1d486bfb 2279#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2280#define NETDEV_CHANGEADDR 0x0008
2281#define NETDEV_GOING_DOWN 0x0009
2282#define NETDEV_CHANGENAME 0x000A
2283#define NETDEV_FEAT_CHANGE 0x000B
2284#define NETDEV_BONDING_FAILOVER 0x000C
2285#define NETDEV_PRE_UP 0x000D
2286#define NETDEV_PRE_TYPE_CHANGE 0x000E
2287#define NETDEV_POST_TYPE_CHANGE 0x000F
2288#define NETDEV_POST_INIT 0x0010
0115e8e3 2289#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2290#define NETDEV_RELEASE 0x0012
2291#define NETDEV_NOTIFY_PEERS 0x0013
2292#define NETDEV_JOIN 0x0014
42e52bf9 2293#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2294#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2295#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2296#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2297#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2298#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2299#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2300#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2301#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2302
f629d208
JP
2303int register_netdevice_notifier(struct notifier_block *nb);
2304int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2305
2306struct netdev_notifier_info {
2307 struct net_device *dev;
2308};
2309
be9efd36
JP
2310struct netdev_notifier_change_info {
2311 struct netdev_notifier_info info; /* must be first */
2312 unsigned int flags_changed;
2313};
2314
0e4ead9d
JP
2315struct netdev_notifier_changeupper_info {
2316 struct netdev_notifier_info info; /* must be first */
2317 struct net_device *upper_dev; /* new upper dev */
2318 bool master; /* is upper dev master */
5e82b4b2 2319 bool linking; /* is the notification for link or unlink */
29bf24af 2320 void *upper_info; /* upper dev info */
0e4ead9d
JP
2321};
2322
04d48266
JP
2323struct netdev_notifier_changelowerstate_info {
2324 struct netdev_notifier_info info; /* must be first */
2325 void *lower_state_info; /* is lower dev state */
2326};
2327
75538c2b
CW
2328static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2329 struct net_device *dev)
2330{
2331 info->dev = dev;
2332}
2333
351638e7
JP
2334static inline struct net_device *
2335netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2336{
2337 return info->dev;
2338}
2339
f629d208 2340int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2341
2342
1da177e4
LT
2343extern rwlock_t dev_base_lock; /* Device list lock */
2344
881d966b
EB
2345#define for_each_netdev(net, d) \
2346 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2347#define for_each_netdev_reverse(net, d) \
2348 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2349#define for_each_netdev_rcu(net, d) \
2350 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2351#define for_each_netdev_safe(net, d, n) \
2352 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2353#define for_each_netdev_continue(net, d) \
2354 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2355#define for_each_netdev_continue_rcu(net, d) \
2356 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2357#define for_each_netdev_in_bond_rcu(bond, slave) \
2358 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2359 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2360#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2361
a050c33f
DL
2362static inline struct net_device *next_net_device(struct net_device *dev)
2363{
2364 struct list_head *lh;
2365 struct net *net;
2366
c346dca1 2367 net = dev_net(dev);
a050c33f
DL
2368 lh = dev->dev_list.next;
2369 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2370}
2371
ce81b76a
ED
2372static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2373{
2374 struct list_head *lh;
2375 struct net *net;
2376
2377 net = dev_net(dev);
ccf43438 2378 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2379 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2380}
2381
a050c33f
DL
2382static inline struct net_device *first_net_device(struct net *net)
2383{
2384 return list_empty(&net->dev_base_head) ? NULL :
2385 net_device_entry(net->dev_base_head.next);
2386}
7562f876 2387
ccf43438
ED
2388static inline struct net_device *first_net_device_rcu(struct net *net)
2389{
2390 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2391
2392 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2393}
2394
f629d208
JP
2395int netdev_boot_setup_check(struct net_device *dev);
2396unsigned long netdev_boot_base(const char *prefix, int unit);
2397struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2398 const char *hwaddr);
2399struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2400struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2401void dev_add_pack(struct packet_type *pt);
2402void dev_remove_pack(struct packet_type *pt);
2403void __dev_remove_pack(struct packet_type *pt);
2404void dev_add_offload(struct packet_offload *po);
2405void dev_remove_offload(struct packet_offload *po);
f629d208 2406
a54acb3a 2407int dev_get_iflink(const struct net_device *dev);
fc4099f1 2408int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2409struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2410 unsigned short mask);
f629d208
JP
2411struct net_device *dev_get_by_name(struct net *net, const char *name);
2412struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2413struct net_device *__dev_get_by_name(struct net *net, const char *name);
2414int dev_alloc_name(struct net_device *dev, const char *name);
2415int dev_open(struct net_device *dev);
2416int dev_close(struct net_device *dev);
99c4a26a 2417int dev_close_many(struct list_head *head, bool unlink);
f629d208 2418void dev_disable_lro(struct net_device *dev);
0c4b51f0 2419int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2420int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2421int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2422int register_netdevice(struct net_device *dev);
2423void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2424void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2425static inline void unregister_netdevice(struct net_device *dev)
2426{
2427 unregister_netdevice_queue(dev, NULL);
2428}
2429
f629d208
JP
2430int netdev_refcnt_read(const struct net_device *dev);
2431void free_netdev(struct net_device *dev);
74d332c1 2432void netdev_freemem(struct net_device *dev);
f629d208
JP
2433void synchronize_net(void);
2434int init_dummy_netdev(struct net_device *dev);
937f1ba5 2435
f60e5990 2436DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2437#define XMIT_RECURSION_LIMIT 10
2438
f60e5990 2439static inline int dev_recursion_level(void)
2440{
2441 return this_cpu_read(xmit_recursion);
2442}
2443
f629d208
JP
2444struct net_device *dev_get_by_index(struct net *net, int ifindex);
2445struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2446struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2447int netdev_get_name(struct net *net, char *name, int ifindex);
2448int dev_restart(struct net_device *dev);
f629d208 2449int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2450
2451static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2452{
2453 return NAPI_GRO_CB(skb)->data_offset;
2454}
2455
2456static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2457{
2458 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2459}
2460
2461static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2462{
2463 NAPI_GRO_CB(skb)->data_offset += len;
2464}
2465
a5b1cf28
HX
2466static inline void *skb_gro_header_fast(struct sk_buff *skb,
2467 unsigned int offset)
86911732 2468{
a5b1cf28
HX
2469 return NAPI_GRO_CB(skb)->frag0 + offset;
2470}
78a478d0 2471
a5b1cf28
HX
2472static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2473{
2474 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2475}
78a478d0 2476
57ea52a8
HX
2477static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2478{
2479 NAPI_GRO_CB(skb)->frag0 = NULL;
2480 NAPI_GRO_CB(skb)->frag0_len = 0;
2481}
2482
a5b1cf28
HX
2483static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2484 unsigned int offset)
2485{
17dd759c
HX
2486 if (!pskb_may_pull(skb, hlen))
2487 return NULL;
2488
57ea52a8 2489 skb_gro_frag0_invalidate(skb);
17dd759c 2490 return skb->data + offset;
86911732 2491}
1da177e4 2492
36e7b1b8
HX
2493static inline void *skb_gro_network_header(struct sk_buff *skb)
2494{
78d3fd0b
HX
2495 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2496 skb_network_offset(skb);
36e7b1b8
HX
2497}
2498
bf5a755f
JC
2499static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2500 const void *start, unsigned int len)
2501{
573e8fca 2502 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2503 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2504 csum_partial(start, len, 0));
2505}
2506
573e8fca
TH
2507/* GRO checksum functions. These are logical equivalents of the normal
2508 * checksum functions (in skbuff.h) except that they operate on the GRO
2509 * offsets and fields in sk_buff.
2510 */
2511
2512__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2513
15e2396d
TH
2514static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2515{
b7fe10e5 2516 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2517}
2518
573e8fca
TH
2519static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2520 bool zero_okay,
2521 __sum16 check)
2522{
6edec0e6
TH
2523 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2524 skb_checksum_start_offset(skb) <
2525 skb_gro_offset(skb)) &&
15e2396d 2526 !skb_at_gro_remcsum_start(skb) &&
662880f4 2527 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2528 (!zero_okay || check));
2529}
2530
2531static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2532 __wsum psum)
2533{
2534 if (NAPI_GRO_CB(skb)->csum_valid &&
2535 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2536 return 0;
2537
2538 NAPI_GRO_CB(skb)->csum = psum;
2539
2540 return __skb_gro_checksum_complete(skb);
2541}
2542
573e8fca
TH
2543static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2544{
662880f4
TH
2545 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2546 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2547 NAPI_GRO_CB(skb)->csum_cnt--;
2548 } else {
2549 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2550 * verified a new top level checksum or an encapsulated one
2551 * during GRO. This saves work if we fallback to normal path.
2552 */
2553 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2554 }
2555}
2556
2557#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2558 compute_pseudo) \
2559({ \
2560 __sum16 __ret = 0; \
2561 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2562 __ret = __skb_gro_checksum_validate_complete(skb, \
2563 compute_pseudo(skb, proto)); \
5a212329
TH
2564 if (__ret) \
2565 __skb_mark_checksum_bad(skb); \
2566 else \
573e8fca
TH
2567 skb_gro_incr_csum_unnecessary(skb); \
2568 __ret; \
2569})
2570
2571#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2572 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2573
2574#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2575 compute_pseudo) \
2576 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2577
2578#define skb_gro_checksum_simple_validate(skb) \
2579 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2580
d96535a1
TH
2581static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2582{
2583 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2584 !NAPI_GRO_CB(skb)->csum_valid);
2585}
2586
2587static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2588 __sum16 check, __wsum pseudo)
2589{
2590 NAPI_GRO_CB(skb)->csum = ~pseudo;
2591 NAPI_GRO_CB(skb)->csum_valid = 1;
2592}
2593
2594#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2595do { \
2596 if (__skb_gro_checksum_convert_check(skb)) \
2597 __skb_gro_checksum_convert(skb, check, \
2598 compute_pseudo(skb, proto)); \
2599} while (0)
2600
26c4f7da
TH
2601struct gro_remcsum {
2602 int offset;
2603 __wsum delta;
2604};
2605
2606static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2607{
846cd667 2608 grc->offset = 0;
26c4f7da
TH
2609 grc->delta = 0;
2610}
2611
b7fe10e5
TH
2612static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2613 unsigned int off, size_t hdrlen,
2614 int start, int offset,
2615 struct gro_remcsum *grc,
2616 bool nopartial)
dcdc8994
TH
2617{
2618 __wsum delta;
b7fe10e5 2619 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2620
2621 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2622
15e2396d 2623 if (!nopartial) {
b7fe10e5
TH
2624 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2625 return ptr;
2626 }
2627
2628 ptr = skb_gro_header_fast(skb, off);
2629 if (skb_gro_header_hard(skb, off + plen)) {
2630 ptr = skb_gro_header_slow(skb, off + plen, off);
2631 if (!ptr)
2632 return NULL;
15e2396d
TH
2633 }
2634
b7fe10e5
TH
2635 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2636 start, offset);
dcdc8994
TH
2637
2638 /* Adjust skb->csum since we changed the packet */
dcdc8994 2639 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2640
b7fe10e5 2641 grc->offset = off + hdrlen + offset;
26c4f7da 2642 grc->delta = delta;
b7fe10e5
TH
2643
2644 return ptr;
dcdc8994
TH
2645}
2646
26c4f7da
TH
2647static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2648 struct gro_remcsum *grc)
2649{
b7fe10e5
TH
2650 void *ptr;
2651 size_t plen = grc->offset + sizeof(u16);
2652
26c4f7da
TH
2653 if (!grc->delta)
2654 return;
2655
b7fe10e5
TH
2656 ptr = skb_gro_header_fast(skb, grc->offset);
2657 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2658 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2659 if (!ptr)
2660 return;
2661 }
2662
2663 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2664}
dcdc8994 2665
0c4e8581
SH
2666static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2667 unsigned short type,
3b04ddde 2668 const void *daddr, const void *saddr,
95c96174 2669 unsigned int len)
0c4e8581 2670{
f1ecfd5d 2671 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2672 return 0;
3b04ddde
SH
2673
2674 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2675}
2676
b95cce35
SH
2677static inline int dev_parse_header(const struct sk_buff *skb,
2678 unsigned char *haddr)
2679{
2680 const struct net_device *dev = skb->dev;
2681
1b83336b 2682 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2683 return 0;
3b04ddde 2684 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2685}
2686
2793a23a
WB
2687/* ll_header must have at least hard_header_len allocated */
2688static inline bool dev_validate_header(const struct net_device *dev,
2689 char *ll_header, int len)
2690{
2691 if (likely(len >= dev->hard_header_len))
2692 return true;
217e6fa2
WB
2693 if (len < dev->min_header_len)
2694 return false;
2793a23a
WB
2695
2696 if (capable(CAP_SYS_RAWIO)) {
2697 memset(ll_header + len, 0, dev->hard_header_len - len);
2698 return true;
2699 }
2700
2701 if (dev->header_ops && dev->header_ops->validate)
2702 return dev->header_ops->validate(ll_header, len);
2703
2704 return false;
2705}
2706
1da177e4 2707typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2708int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2709static inline int unregister_gifconf(unsigned int family)
2710{
2711 return register_gifconf(family, NULL);
2712}
2713
99bbc707 2714#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2715#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2716struct sd_flow_limit {
2717 u64 count;
2718 unsigned int num_buckets;
2719 unsigned int history_head;
2720 u16 history[FLOW_LIMIT_HISTORY];
2721 u8 buckets[];
2722};
2723
2724extern int netdev_flow_limit_table_len;
2725#endif /* CONFIG_NET_FLOW_LIMIT */
2726
1da177e4 2727/*
5e82b4b2 2728 * Incoming packets are placed on per-CPU queues
1da177e4 2729 */
d94d9fee 2730struct softnet_data {
1da177e4 2731 struct list_head poll_list;
6e7676c1 2732 struct sk_buff_head process_queue;
1da177e4 2733
dee42870 2734 /* stats */
cd7b5396
DM
2735 unsigned int processed;
2736 unsigned int time_squeeze;
cd7b5396 2737 unsigned int received_rps;
fd793d89 2738#ifdef CONFIG_RPS
88751275 2739 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2740#endif
2741#ifdef CONFIG_NET_FLOW_LIMIT
2742 struct sd_flow_limit __rcu *flow_limit;
2743#endif
2744 struct Qdisc *output_queue;
2745 struct Qdisc **output_queue_tailp;
2746 struct sk_buff *completion_queue;
88751275 2747
4cdb1e2e 2748#ifdef CONFIG_RPS
501e7ef5
ED
2749 /* input_queue_head should be written by cpu owning this struct,
2750 * and only read by other cpus. Worth using a cache line.
2751 */
2752 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2753
2754 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2755 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2756 struct softnet_data *rps_ipi_next;
2757 unsigned int cpu;
76cc8b13 2758 unsigned int input_queue_tail;
1e94d72f 2759#endif
95c96174 2760 unsigned int dropped;
0a9627f2 2761 struct sk_buff_head input_pkt_queue;
bea3348e 2762 struct napi_struct backlog;
99bbc707 2763
1da177e4
LT
2764};
2765
76cc8b13 2766static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2767{
2768#ifdef CONFIG_RPS
76cc8b13
TH
2769 sd->input_queue_head++;
2770#endif
2771}
2772
2773static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2774 unsigned int *qtail)
2775{
2776#ifdef CONFIG_RPS
2777 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2778#endif
2779}
2780
0a9627f2 2781DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2782
f629d208 2783void __netif_schedule(struct Qdisc *q);
46e5da40 2784void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2785
fd2ea0a7
DM
2786static inline void netif_tx_schedule_all(struct net_device *dev)
2787{
2788 unsigned int i;
2789
2790 for (i = 0; i < dev->num_tx_queues; i++)
2791 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2792}
2793
f9a7cbbf 2794static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2795{
73466498 2796 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2797}
2798
bea3348e
SH
2799/**
2800 * netif_start_queue - allow transmit
2801 * @dev: network device
2802 *
2803 * Allow upper layers to call the device hard_start_xmit routine.
2804 */
1da177e4
LT
2805static inline void netif_start_queue(struct net_device *dev)
2806{
e8a0464c 2807 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2808}
2809
fd2ea0a7
DM
2810static inline void netif_tx_start_all_queues(struct net_device *dev)
2811{
2812 unsigned int i;
2813
2814 for (i = 0; i < dev->num_tx_queues; i++) {
2815 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2816 netif_tx_start_queue(txq);
2817 }
2818}
2819
46e5da40 2820void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2821
d29f749e
DJ
2822/**
2823 * netif_wake_queue - restart transmit
2824 * @dev: network device
2825 *
2826 * Allow upper layers to call the device hard_start_xmit routine.
2827 * Used for flow control when transmit resources are available.
2828 */
79d16385
DM
2829static inline void netif_wake_queue(struct net_device *dev)
2830{
e8a0464c 2831 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2832}
2833
fd2ea0a7
DM
2834static inline void netif_tx_wake_all_queues(struct net_device *dev)
2835{
2836 unsigned int i;
2837
2838 for (i = 0; i < dev->num_tx_queues; i++) {
2839 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2840 netif_tx_wake_queue(txq);
2841 }
2842}
2843
f9a7cbbf 2844static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2845{
73466498 2846 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2847}
2848
bea3348e
SH
2849/**
2850 * netif_stop_queue - stop transmitted packets
2851 * @dev: network device
2852 *
2853 * Stop upper layers calling the device hard_start_xmit routine.
2854 * Used for flow control when transmit resources are unavailable.
2855 */
1da177e4
LT
2856static inline void netif_stop_queue(struct net_device *dev)
2857{
e8a0464c 2858 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2859}
2860
a2029240 2861void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2862
4d29515f 2863static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2864{
73466498 2865 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2866}
2867
bea3348e
SH
2868/**
2869 * netif_queue_stopped - test if transmit queue is flowblocked
2870 * @dev: network device
2871 *
2872 * Test if transmit queue on device is currently unable to send.
2873 */
4d29515f 2874static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2875{
e8a0464c 2876 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2877}
2878
4d29515f 2879static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2880{
73466498
TH
2881 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2882}
2883
8e2f1a63
DB
2884static inline bool
2885netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2886{
2887 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2888}
2889
8e2f1a63
DB
2890static inline bool
2891netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2892{
2893 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2894}
2895
53511453
ED
2896/**
2897 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2898 * @dev_queue: pointer to transmit queue
2899 *
2900 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2901 * to give appropriate hint to the CPU.
53511453
ED
2902 */
2903static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2904{
2905#ifdef CONFIG_BQL
2906 prefetchw(&dev_queue->dql.num_queued);
2907#endif
2908}
2909
2910/**
2911 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2912 * @dev_queue: pointer to transmit queue
2913 *
2914 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2915 * to give appropriate hint to the CPU.
53511453
ED
2916 */
2917static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2918{
2919#ifdef CONFIG_BQL
2920 prefetchw(&dev_queue->dql.limit);
2921#endif
2922}
2923
c5d67bd7
TH
2924static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2925 unsigned int bytes)
2926{
114cf580
TH
2927#ifdef CONFIG_BQL
2928 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2929
2930 if (likely(dql_avail(&dev_queue->dql) >= 0))
2931 return;
2932
2933 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2934
2935 /*
2936 * The XOFF flag must be set before checking the dql_avail below,
2937 * because in netdev_tx_completed_queue we update the dql_completed
2938 * before checking the XOFF flag.
2939 */
2940 smp_mb();
2941
2942 /* check again in case another CPU has just made room avail */
2943 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2944 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2945#endif
c5d67bd7
TH
2946}
2947
0042d0c8
FF
2948/**
2949 * netdev_sent_queue - report the number of bytes queued to hardware
2950 * @dev: network device
2951 * @bytes: number of bytes queued to the hardware device queue
2952 *
2953 * Report the number of bytes queued for sending/completion to the network
2954 * device hardware queue. @bytes should be a good approximation and should
2955 * exactly match netdev_completed_queue() @bytes
2956 */
c5d67bd7
TH
2957static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2958{
2959 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2960}
2961
2962static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2963 unsigned int pkts, unsigned int bytes)
c5d67bd7 2964{
114cf580 2965#ifdef CONFIG_BQL
b37c0fbe
AD
2966 if (unlikely(!bytes))
2967 return;
2968
2969 dql_completed(&dev_queue->dql, bytes);
2970
2971 /*
2972 * Without the memory barrier there is a small possiblity that
2973 * netdev_tx_sent_queue will miss the update and cause the queue to
2974 * be stopped forever
2975 */
2976 smp_mb();
2977
2978 if (dql_avail(&dev_queue->dql) < 0)
2979 return;
2980
2981 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2982 netif_schedule_queue(dev_queue);
114cf580 2983#endif
c5d67bd7
TH
2984}
2985
0042d0c8
FF
2986/**
2987 * netdev_completed_queue - report bytes and packets completed by device
2988 * @dev: network device
2989 * @pkts: actual number of packets sent over the medium
2990 * @bytes: actual number of bytes sent over the medium
2991 *
2992 * Report the number of bytes and packets transmitted by the network device
2993 * hardware queue over the physical medium, @bytes must exactly match the
2994 * @bytes amount passed to netdev_sent_queue()
2995 */
c5d67bd7 2996static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2997 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2998{
2999 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3000}
3001
3002static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3003{
114cf580 3004#ifdef CONFIG_BQL
5c490354 3005 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3006 dql_reset(&q->dql);
3007#endif
c5d67bd7
TH
3008}
3009
0042d0c8
FF
3010/**
3011 * netdev_reset_queue - reset the packets and bytes count of a network device
3012 * @dev_queue: network device
3013 *
3014 * Reset the bytes and packet count of a network device and clear the
3015 * software flow control OFF bit for this network device
3016 */
c5d67bd7
TH
3017static inline void netdev_reset_queue(struct net_device *dev_queue)
3018{
3019 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3020}
3021
b9507bda
DB
3022/**
3023 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3024 * @dev: network device
3025 * @queue_index: given tx queue index
3026 *
3027 * Returns 0 if given tx queue index >= number of device tx queues,
3028 * otherwise returns the originally passed tx queue index.
3029 */
3030static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3031{
3032 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3033 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3034 dev->name, queue_index,
3035 dev->real_num_tx_queues);
3036 return 0;
3037 }
3038
3039 return queue_index;
3040}
3041
bea3348e
SH
3042/**
3043 * netif_running - test if up
3044 * @dev: network device
3045 *
3046 * Test if the device has been brought up.
3047 */
4d29515f 3048static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3049{
3050 return test_bit(__LINK_STATE_START, &dev->state);
3051}
3052
f25f4e44 3053/*
5e82b4b2 3054 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3055 * stop, and a check if it's stopped. All other device management is
3056 * done at the overall netdevice level.
3057 * Also test the device if we're multiqueue.
3058 */
bea3348e
SH
3059
3060/**
3061 * netif_start_subqueue - allow sending packets on subqueue
3062 * @dev: network device
3063 * @queue_index: sub queue index
3064 *
3065 * Start individual transmit queue of a device with multiple transmit queues.
3066 */
f25f4e44
PWJ
3067static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3068{
fd2ea0a7 3069 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3070
3071 netif_tx_start_queue(txq);
f25f4e44
PWJ
3072}
3073
bea3348e
SH
3074/**
3075 * netif_stop_subqueue - stop sending packets on subqueue
3076 * @dev: network device
3077 * @queue_index: sub queue index
3078 *
3079 * Stop individual transmit queue of a device with multiple transmit queues.
3080 */
f25f4e44
PWJ
3081static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3082{
fd2ea0a7 3083 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3084 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3085}
3086
bea3348e
SH
3087/**
3088 * netif_subqueue_stopped - test status of subqueue
3089 * @dev: network device
3090 * @queue_index: sub queue index
3091 *
3092 * Check individual transmit queue of a device with multiple transmit queues.
3093 */
4d29515f
DM
3094static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3095 u16 queue_index)
f25f4e44 3096{
fd2ea0a7 3097 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3098
3099 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3100}
3101
4d29515f
DM
3102static inline bool netif_subqueue_stopped(const struct net_device *dev,
3103 struct sk_buff *skb)
668f895a
PE
3104{
3105 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3106}
bea3348e 3107
738b35cc
FF
3108/**
3109 * netif_wake_subqueue - allow sending packets on subqueue
3110 * @dev: network device
3111 * @queue_index: sub queue index
3112 *
3113 * Resume individual transmit queue of a device with multiple transmit queues.
3114 */
3115static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3116{
3117 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3118
3119 netif_tx_wake_queue(txq);
3120}
f25f4e44 3121
537c00de 3122#ifdef CONFIG_XPS
53af53ae 3123int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3124 u16 index);
537c00de
AD
3125#else
3126static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3127 const struct cpumask *mask,
537c00de
AD
3128 u16 index)
3129{
3130 return 0;
3131}
3132#endif
3133
5605c762
JP
3134u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3135 unsigned int num_tx_queues);
3136
a3d22a68
VZ
3137/*
3138 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3139 * as a distribution range limit for the returned value.
3140 */
3141static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3142 struct sk_buff *skb)
a3d22a68
VZ
3143{
3144 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3145}
3146
bea3348e
SH
3147/**
3148 * netif_is_multiqueue - test if device has multiple transmit queues
3149 * @dev: network device
3150 *
3151 * Check if device has multiple transmit queues
bea3348e 3152 */
4d29515f 3153static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3154{
a02cec21 3155 return dev->num_tx_queues > 1;
f25f4e44 3156}
1da177e4 3157
f629d208 3158int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3159
a953be53 3160#ifdef CONFIG_SYSFS
f629d208 3161int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3162#else
3163static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3164 unsigned int rxq)
3165{
3166 return 0;
3167}
3168#endif
3169
a953be53
MD
3170#ifdef CONFIG_SYSFS
3171static inline unsigned int get_netdev_rx_queue_index(
3172 struct netdev_rx_queue *queue)
3173{
3174 struct net_device *dev = queue->dev;
3175 int index = queue - dev->_rx;
3176
3177 BUG_ON(index >= dev->num_rx_queues);
3178 return index;
3179}
3180#endif
3181
16917b87 3182#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3183int netif_get_num_default_rss_queues(void);
16917b87 3184
e6247027
ED
3185enum skb_free_reason {
3186 SKB_REASON_CONSUMED,
3187 SKB_REASON_DROPPED,
3188};
3189
3190void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3191void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3192
e6247027
ED
3193/*
3194 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3195 * interrupt context or with hardware interrupts being disabled.
3196 * (in_irq() || irqs_disabled())
3197 *
3198 * We provide four helpers that can be used in following contexts :
3199 *
3200 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3201 * replacing kfree_skb(skb)
3202 *
3203 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3204 * Typically used in place of consume_skb(skb) in TX completion path
3205 *
3206 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3207 * replacing kfree_skb(skb)
3208 *
3209 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3210 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3211 */
e6247027
ED
3212static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3213{
3214 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3215}
3216
3217static inline void dev_consume_skb_irq(struct sk_buff *skb)
3218{
3219 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3220}
3221
3222static inline void dev_kfree_skb_any(struct sk_buff *skb)
3223{
3224 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3225}
3226
3227static inline void dev_consume_skb_any(struct sk_buff *skb)
3228{
3229 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3230}
1da177e4 3231
f629d208
JP
3232int netif_rx(struct sk_buff *skb);
3233int netif_rx_ni(struct sk_buff *skb);
04eb4489 3234int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3235gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3236void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3237struct sk_buff *napi_get_frags(struct napi_struct *napi);
3238gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3239struct packet_offload *gro_find_receive_by_type(__be16 type);
3240struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3241
3242static inline void napi_free_frags(struct napi_struct *napi)
3243{
3244 kfree_skb(napi->skb);
3245 napi->skb = NULL;
3246}
3247
24b27fc4 3248bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3249int netdev_rx_handler_register(struct net_device *dev,
3250 rx_handler_func_t *rx_handler,
3251 void *rx_handler_data);
3252void netdev_rx_handler_unregister(struct net_device *dev);
3253
3254bool dev_valid_name(const char *name);
3255int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3256int dev_ethtool(struct net *net, struct ifreq *);
3257unsigned int dev_get_flags(const struct net_device *);
3258int __dev_change_flags(struct net_device *, unsigned int flags);
3259int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3260void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3261 unsigned int gchanges);
f629d208
JP
3262int dev_change_name(struct net_device *, const char *);
3263int dev_set_alias(struct net_device *, const char *, size_t);
3264int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3265int dev_set_mtu(struct net_device *, int);
3266void dev_set_group(struct net_device *, int);
3267int dev_set_mac_address(struct net_device *, struct sockaddr *);
3268int dev_change_carrier(struct net_device *, bool new_carrier);
3269int dev_get_phys_port_id(struct net_device *dev,
02637fce 3270 struct netdev_phys_item_id *ppid);
db24a904
DA
3271int dev_get_phys_port_name(struct net_device *dev,
3272 char *name, size_t len);
d746d707 3273int dev_change_proto_down(struct net_device *dev, bool proto_down);
85de8576 3274int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
55a93b3e 3275struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3276struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3277 struct netdev_queue *txq, int *ret);
a0265d28 3278int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3279int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3280bool is_skb_forwardable(const struct net_device *dev,
3281 const struct sk_buff *skb);
1da177e4 3282
4e3264d2
MKL
3283static __always_inline int ____dev_forward_skb(struct net_device *dev,
3284 struct sk_buff *skb)
3285{
3286 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3287 unlikely(!is_skb_forwardable(dev, skb))) {
3288 atomic_long_inc(&dev->rx_dropped);
3289 kfree_skb(skb);
3290 return NET_RX_DROP;
3291 }
3292
3293 skb_scrub_packet(skb, true);
3294 skb->priority = 0;
3295 return 0;
3296}
3297
74b20582
DA
3298void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3299
20380731 3300extern int netdev_budget;
1da177e4
LT
3301
3302/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3303void netdev_run_todo(void);
1da177e4 3304
bea3348e
SH
3305/**
3306 * dev_put - release reference to device
3307 * @dev: network device
3308 *
9ef4429b 3309 * Release reference to device to allow it to be freed.
bea3348e 3310 */
1da177e4
LT
3311static inline void dev_put(struct net_device *dev)
3312{
933393f5 3313 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3314}
3315
bea3348e
SH
3316/**
3317 * dev_hold - get reference to device
3318 * @dev: network device
3319 *
9ef4429b 3320 * Hold reference to device to keep it from being freed.
bea3348e 3321 */
15333061
SH
3322static inline void dev_hold(struct net_device *dev)
3323{
933393f5 3324 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3325}
1da177e4
LT
3326
3327/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3328 * and _off may be called from IRQ context, but it is caller
3329 * who is responsible for serialization of these calls.
b00055aa
SR
3330 *
3331 * The name carrier is inappropriate, these functions should really be
3332 * called netif_lowerlayer_*() because they represent the state of any
3333 * kind of lower layer not just hardware media.
1da177e4
LT
3334 */
3335
f629d208
JP
3336void linkwatch_init_dev(struct net_device *dev);
3337void linkwatch_fire_event(struct net_device *dev);
3338void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3339
bea3348e
SH
3340/**
3341 * netif_carrier_ok - test if carrier present
3342 * @dev: network device
3343 *
3344 * Check if carrier is present on device
3345 */
4d29515f 3346static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3347{
3348 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3349}
3350
f629d208 3351unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3352
f629d208 3353void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3354
f629d208 3355void netif_carrier_on(struct net_device *dev);
1da177e4 3356
f629d208 3357void netif_carrier_off(struct net_device *dev);
1da177e4 3358
bea3348e
SH
3359/**
3360 * netif_dormant_on - mark device as dormant.
3361 * @dev: network device
3362 *
3363 * Mark device as dormant (as per RFC2863).
3364 *
3365 * The dormant state indicates that the relevant interface is not
3366 * actually in a condition to pass packets (i.e., it is not 'up') but is
3367 * in a "pending" state, waiting for some external event. For "on-
3368 * demand" interfaces, this new state identifies the situation where the
3369 * interface is waiting for events to place it in the up state.
bea3348e 3370 */
b00055aa
SR
3371static inline void netif_dormant_on(struct net_device *dev)
3372{
3373 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3374 linkwatch_fire_event(dev);
3375}
3376
bea3348e
SH
3377/**
3378 * netif_dormant_off - set device as not dormant.
3379 * @dev: network device
3380 *
3381 * Device is not in dormant state.
3382 */
b00055aa
SR
3383static inline void netif_dormant_off(struct net_device *dev)
3384{
3385 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3386 linkwatch_fire_event(dev);
3387}
3388
bea3348e
SH
3389/**
3390 * netif_dormant - test if carrier present
3391 * @dev: network device
3392 *
3393 * Check if carrier is present on device
3394 */
4d29515f 3395static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3396{
3397 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3398}
3399
3400
bea3348e
SH
3401/**
3402 * netif_oper_up - test if device is operational
3403 * @dev: network device
3404 *
3405 * Check if carrier is operational
3406 */
4d29515f 3407static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3408{
b00055aa
SR
3409 return (dev->operstate == IF_OPER_UP ||
3410 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3411}
3412
bea3348e
SH
3413/**
3414 * netif_device_present - is device available or removed
3415 * @dev: network device
3416 *
3417 * Check if device has not been removed from system.
3418 */
4d29515f 3419static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3420{
3421 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3422}
3423
f629d208 3424void netif_device_detach(struct net_device *dev);
1da177e4 3425
f629d208 3426void netif_device_attach(struct net_device *dev);
1da177e4
LT
3427
3428/*
3429 * Network interface message level settings
3430 */
1da177e4
LT
3431
3432enum {
3433 NETIF_MSG_DRV = 0x0001,
3434 NETIF_MSG_PROBE = 0x0002,
3435 NETIF_MSG_LINK = 0x0004,
3436 NETIF_MSG_TIMER = 0x0008,
3437 NETIF_MSG_IFDOWN = 0x0010,
3438 NETIF_MSG_IFUP = 0x0020,
3439 NETIF_MSG_RX_ERR = 0x0040,
3440 NETIF_MSG_TX_ERR = 0x0080,
3441 NETIF_MSG_TX_QUEUED = 0x0100,
3442 NETIF_MSG_INTR = 0x0200,
3443 NETIF_MSG_TX_DONE = 0x0400,
3444 NETIF_MSG_RX_STATUS = 0x0800,
3445 NETIF_MSG_PKTDATA = 0x1000,
3446 NETIF_MSG_HW = 0x2000,
3447 NETIF_MSG_WOL = 0x4000,
3448};
3449
3450#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3451#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3452#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3453#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3454#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3455#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3456#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3457#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3458#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3459#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3460#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3461#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3462#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3463#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3464#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3465
3466static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3467{
3468 /* use default */
3469 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3470 return default_msg_enable_bits;
3471 if (debug_value == 0) /* no output */
3472 return 0;
3473 /* set low N bits */
3474 return (1 << debug_value) - 1;
3475}
3476
c773e847 3477static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3478{
c773e847
DM
3479 spin_lock(&txq->_xmit_lock);
3480 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3481}
3482
5a717f4f
MT
3483static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3484{
3485 __acquire(&txq->_xmit_lock);
3486 return true;
3487}
3488
3489static inline void __netif_tx_release(struct netdev_queue *txq)
3490{
3491 __release(&txq->_xmit_lock);
3492}
3493
fd2ea0a7
DM
3494static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3495{
3496 spin_lock_bh(&txq->_xmit_lock);
3497 txq->xmit_lock_owner = smp_processor_id();
3498}
3499
4d29515f 3500static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3501{
4d29515f 3502 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3503 if (likely(ok))
3504 txq->xmit_lock_owner = smp_processor_id();
3505 return ok;
3506}
3507
3508static inline void __netif_tx_unlock(struct netdev_queue *txq)
3509{
3510 txq->xmit_lock_owner = -1;
3511 spin_unlock(&txq->_xmit_lock);
3512}
3513
3514static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3515{
3516 txq->xmit_lock_owner = -1;
3517 spin_unlock_bh(&txq->_xmit_lock);
3518}
3519
08baf561
ED
3520static inline void txq_trans_update(struct netdev_queue *txq)
3521{
3522 if (txq->xmit_lock_owner != -1)
3523 txq->trans_start = jiffies;
3524}
3525
ba162f8e
FW
3526/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3527static inline void netif_trans_update(struct net_device *dev)
3528{
9b36627a
FW
3529 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3530
3531 if (txq->trans_start != jiffies)
3532 txq->trans_start = jiffies;
ba162f8e
FW
3533}
3534
d29f749e
DJ
3535/**
3536 * netif_tx_lock - grab network device transmit lock
3537 * @dev: network device
d29f749e
DJ
3538 *
3539 * Get network device transmit lock
3540 */
22dd7495
JHS
3541static inline void netif_tx_lock(struct net_device *dev)
3542{
e8a0464c 3543 unsigned int i;
c3f26a26 3544 int cpu;
c773e847 3545
c3f26a26
DM
3546 spin_lock(&dev->tx_global_lock);
3547 cpu = smp_processor_id();
e8a0464c
DM
3548 for (i = 0; i < dev->num_tx_queues; i++) {
3549 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3550
3551 /* We are the only thread of execution doing a
3552 * freeze, but we have to grab the _xmit_lock in
3553 * order to synchronize with threads which are in
3554 * the ->hard_start_xmit() handler and already
3555 * checked the frozen bit.
3556 */
e8a0464c 3557 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3558 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3559 __netif_tx_unlock(txq);
e8a0464c 3560 }
932ff279
HX
3561}
3562
3563static inline void netif_tx_lock_bh(struct net_device *dev)
3564{
e8a0464c
DM
3565 local_bh_disable();
3566 netif_tx_lock(dev);
932ff279
HX
3567}
3568
932ff279
HX
3569static inline void netif_tx_unlock(struct net_device *dev)
3570{
e8a0464c
DM
3571 unsigned int i;
3572
3573 for (i = 0; i < dev->num_tx_queues; i++) {
3574 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3575
c3f26a26
DM
3576 /* No need to grab the _xmit_lock here. If the
3577 * queue is not stopped for another reason, we
3578 * force a schedule.
3579 */
3580 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3581 netif_schedule_queue(txq);
c3f26a26
DM
3582 }
3583 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3584}
3585
3586static inline void netif_tx_unlock_bh(struct net_device *dev)
3587{
e8a0464c
DM
3588 netif_tx_unlock(dev);
3589 local_bh_enable();
932ff279
HX
3590}
3591
c773e847 3592#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3593 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3594 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3595 } else { \
3596 __netif_tx_acquire(txq); \
22dd7495
JHS
3597 } \
3598}
3599
5efeac44
EB
3600#define HARD_TX_TRYLOCK(dev, txq) \
3601 (((dev->features & NETIF_F_LLTX) == 0) ? \
3602 __netif_tx_trylock(txq) : \
5a717f4f 3603 __netif_tx_acquire(txq))
5efeac44 3604
c773e847 3605#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3606 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3607 __netif_tx_unlock(txq); \
5a717f4f
MT
3608 } else { \
3609 __netif_tx_release(txq); \
22dd7495
JHS
3610 } \
3611}
3612
1da177e4
LT
3613static inline void netif_tx_disable(struct net_device *dev)
3614{
fd2ea0a7 3615 unsigned int i;
c3f26a26 3616 int cpu;
fd2ea0a7 3617
c3f26a26
DM
3618 local_bh_disable();
3619 cpu = smp_processor_id();
fd2ea0a7
DM
3620 for (i = 0; i < dev->num_tx_queues; i++) {
3621 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3622
3623 __netif_tx_lock(txq, cpu);
fd2ea0a7 3624 netif_tx_stop_queue(txq);
c3f26a26 3625 __netif_tx_unlock(txq);
fd2ea0a7 3626 }
c3f26a26 3627 local_bh_enable();
1da177e4
LT
3628}
3629
e308a5d8
DM
3630static inline void netif_addr_lock(struct net_device *dev)
3631{
3632 spin_lock(&dev->addr_list_lock);
3633}
3634
2429f7ac
JP
3635static inline void netif_addr_lock_nested(struct net_device *dev)
3636{
25175ba5
VY
3637 int subclass = SINGLE_DEPTH_NESTING;
3638
3639 if (dev->netdev_ops->ndo_get_lock_subclass)
3640 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3641
3642 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3643}
3644
e308a5d8
DM
3645static inline void netif_addr_lock_bh(struct net_device *dev)
3646{
3647 spin_lock_bh(&dev->addr_list_lock);
3648}
3649
3650static inline void netif_addr_unlock(struct net_device *dev)
3651{
3652 spin_unlock(&dev->addr_list_lock);
3653}
3654
3655static inline void netif_addr_unlock_bh(struct net_device *dev)
3656{
3657 spin_unlock_bh(&dev->addr_list_lock);
3658}
3659
f001fde5 3660/*
31278e71 3661 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3662 * rcu_read_lock held.
3663 */
3664#define for_each_dev_addr(dev, ha) \
31278e71 3665 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3666
1da177e4
LT
3667/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3668
f629d208 3669void ether_setup(struct net_device *dev);
1da177e4
LT
3670
3671/* Support for loadable net-drivers */
f629d208 3672struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3673 unsigned char name_assign_type,
f629d208
JP
3674 void (*setup)(struct net_device *),
3675 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3676#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3677 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3678
c835a677
TG
3679#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3680 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3681 count)
36909ea4 3682
f629d208
JP
3683int register_netdev(struct net_device *dev);
3684void unregister_netdev(struct net_device *dev);
f001fde5 3685
22bedad3 3686/* General hardware address lists handling functions */
f629d208
JP
3687int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3688 struct netdev_hw_addr_list *from_list, int addr_len);
3689void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3690 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3691int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3692 struct net_device *dev,
3693 int (*sync)(struct net_device *, const unsigned char *),
3694 int (*unsync)(struct net_device *,
3695 const unsigned char *));
3696void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3697 struct net_device *dev,
3698 int (*unsync)(struct net_device *,
3699 const unsigned char *));
f629d208 3700void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3701
f001fde5 3702/* Functions used for device addresses handling */
f629d208
JP
3703int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3704 unsigned char addr_type);
3705int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3706 unsigned char addr_type);
f629d208
JP
3707void dev_addr_flush(struct net_device *dev);
3708int dev_addr_init(struct net_device *dev);
a748ee24
JP
3709
3710/* Functions used for unicast addresses handling */
f629d208
JP
3711int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3712int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3713int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3714int dev_uc_sync(struct net_device *to, struct net_device *from);
3715int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3716void dev_uc_unsync(struct net_device *to, struct net_device *from);
3717void dev_uc_flush(struct net_device *dev);
3718void dev_uc_init(struct net_device *dev);
f001fde5 3719
670e5b8e
AD
3720/**
3721 * __dev_uc_sync - Synchonize device's unicast list
3722 * @dev: device to sync
3723 * @sync: function to call if address should be added
3724 * @unsync: function to call if address should be removed
3725 *
3726 * Add newly added addresses to the interface, and release
3727 * addresses that have been deleted.
5e82b4b2 3728 */
670e5b8e
AD
3729static inline int __dev_uc_sync(struct net_device *dev,
3730 int (*sync)(struct net_device *,
3731 const unsigned char *),
3732 int (*unsync)(struct net_device *,
3733 const unsigned char *))
3734{
3735 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3736}
3737
3738/**
e793c0f7 3739 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3740 * @dev: device to sync
3741 * @unsync: function to call if address should be removed
3742 *
3743 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3744 */
670e5b8e
AD
3745static inline void __dev_uc_unsync(struct net_device *dev,
3746 int (*unsync)(struct net_device *,
3747 const unsigned char *))
3748{
3749 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3750}
3751
22bedad3 3752/* Functions used for multicast addresses handling */
f629d208
JP
3753int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3754int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3755int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3756int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3757int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3758int dev_mc_sync(struct net_device *to, struct net_device *from);
3759int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3760void dev_mc_unsync(struct net_device *to, struct net_device *from);
3761void dev_mc_flush(struct net_device *dev);
3762void dev_mc_init(struct net_device *dev);
f001fde5 3763
670e5b8e
AD
3764/**
3765 * __dev_mc_sync - Synchonize device's multicast list
3766 * @dev: device to sync
3767 * @sync: function to call if address should be added
3768 * @unsync: function to call if address should be removed
3769 *
3770 * Add newly added addresses to the interface, and release
3771 * addresses that have been deleted.
5e82b4b2 3772 */
670e5b8e
AD
3773static inline int __dev_mc_sync(struct net_device *dev,
3774 int (*sync)(struct net_device *,
3775 const unsigned char *),
3776 int (*unsync)(struct net_device *,
3777 const unsigned char *))
3778{
3779 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3780}
3781
3782/**
e793c0f7 3783 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3784 * @dev: device to sync
3785 * @unsync: function to call if address should be removed
3786 *
3787 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3788 */
670e5b8e
AD
3789static inline void __dev_mc_unsync(struct net_device *dev,
3790 int (*unsync)(struct net_device *,
3791 const unsigned char *))
3792{
3793 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3794}
3795
4417da66 3796/* Functions used for secondary unicast and multicast support */
f629d208
JP
3797void dev_set_rx_mode(struct net_device *dev);
3798void __dev_set_rx_mode(struct net_device *dev);
3799int dev_set_promiscuity(struct net_device *dev, int inc);
3800int dev_set_allmulti(struct net_device *dev, int inc);
3801void netdev_state_change(struct net_device *dev);
3802void netdev_notify_peers(struct net_device *dev);
3803void netdev_features_change(struct net_device *dev);
1da177e4 3804/* Load a device via the kmod */
f629d208
JP
3805void dev_load(struct net *net, const char *name);
3806struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3807 struct rtnl_link_stats64 *storage);
3808void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3809 const struct net_device_stats *netdev_stats);
eeda3fd6 3810
1da177e4 3811extern int netdev_max_backlog;
3b098e2d 3812extern int netdev_tstamp_prequeue;
1da177e4 3813extern int weight_p;
3d48b53f
MT
3814extern int dev_weight_rx_bias;
3815extern int dev_weight_tx_bias;
3816extern int dev_rx_weight;
3817extern int dev_tx_weight;
9ff162a8 3818
f629d208 3819bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3820struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3821 struct list_head **iter);
f629d208
JP
3822struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3823 struct list_head **iter);
8b5be856 3824
44a40855
VY
3825/* iterate through upper list, must be called under RCU read lock */
3826#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3827 for (iter = &(dev)->adj_list.upper, \
3828 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3829 updev; \
3830 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3831
1a3f060c
DA
3832int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3833 int (*fn)(struct net_device *upper_dev,
3834 void *data),
3835 void *data);
3836
3837bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3838 struct net_device *upper_dev);
3839
f629d208
JP
3840void *netdev_lower_get_next_private(struct net_device *dev,
3841 struct list_head **iter);
3842void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3843 struct list_head **iter);
31088a11
VF
3844
3845#define netdev_for_each_lower_private(dev, priv, iter) \
3846 for (iter = (dev)->adj_list.lower.next, \
3847 priv = netdev_lower_get_next_private(dev, &(iter)); \
3848 priv; \
3849 priv = netdev_lower_get_next_private(dev, &(iter)))
3850
3851#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3852 for (iter = &(dev)->adj_list.lower, \
3853 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3854 priv; \
3855 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3856
4085ebe8
VY
3857void *netdev_lower_get_next(struct net_device *dev,
3858 struct list_head **iter);
7ce856aa 3859
4085ebe8 3860#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3861 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3862 ldev = netdev_lower_get_next(dev, &(iter)); \
3863 ldev; \
3864 ldev = netdev_lower_get_next(dev, &(iter)))
3865
7ce856aa
JP
3866struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3867 struct list_head **iter);
3868struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3869 struct list_head **iter);
3870
1a3f060c
DA
3871int netdev_walk_all_lower_dev(struct net_device *dev,
3872 int (*fn)(struct net_device *lower_dev,
3873 void *data),
3874 void *data);
3875int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3876 int (*fn)(struct net_device *lower_dev,
3877 void *data),
3878 void *data);
3879
f629d208 3880void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3881void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3882struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3883struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3884int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3885int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3886 struct net_device *upper_dev,
29bf24af 3887 void *upper_priv, void *upper_info);
f629d208
JP
3888void netdev_upper_dev_unlink(struct net_device *dev,
3889 struct net_device *upper_dev);
5bb025fa 3890void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3891void *netdev_lower_dev_get_private(struct net_device *dev,
3892 struct net_device *lower_dev);
04d48266
JP
3893void netdev_lower_state_changed(struct net_device *lower_dev,
3894 void *lower_state_info);
960fb622
ED
3895
3896/* RSS keys are 40 or 52 bytes long */
3897#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3898extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3899void netdev_rss_key_fill(void *buffer, size_t len);
3900
952fcfd0 3901int dev_get_nest_level(struct net_device *dev);
f629d208
JP
3902int skb_checksum_help(struct sk_buff *skb);
3903struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3904 netdev_features_t features, bool tx_path);
3905struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3906 netdev_features_t features);
12b0004d 3907
61bd3857
MS
3908struct netdev_bonding_info {
3909 ifslave slave;
3910 ifbond master;
3911};
3912
3913struct netdev_notifier_bonding_info {
3914 struct netdev_notifier_info info; /* must be first */
3915 struct netdev_bonding_info bonding_info;
3916};
3917
3918void netdev_bonding_info_change(struct net_device *dev,
3919 struct netdev_bonding_info *bonding_info);
3920
12b0004d
CW
3921static inline
3922struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3923{
3924 return __skb_gso_segment(skb, features, true);
3925}
53d6471c 3926__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3927
3928static inline bool can_checksum_protocol(netdev_features_t features,
3929 __be16 protocol)
3930{
c8cd0989
TH
3931 if (protocol == htons(ETH_P_FCOE))
3932 return !!(features & NETIF_F_FCOE_CRC);
3933
3934 /* Assume this is an IP checksum (not SCTP CRC) */
3935
3936 if (features & NETIF_F_HW_CSUM) {
3937 /* Can checksum everything */
3938 return true;
3939 }
3940
3941 switch (protocol) {
3942 case htons(ETH_P_IP):
3943 return !!(features & NETIF_F_IP_CSUM);
3944 case htons(ETH_P_IPV6):
3945 return !!(features & NETIF_F_IPV6_CSUM);
3946 default:
3947 return false;
3948 }
ec5f0615 3949}
12b0004d 3950
fb286bb2 3951#ifdef CONFIG_BUG
f629d208 3952void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3953#else
3954static inline void netdev_rx_csum_fault(struct net_device *dev)
3955{
3956}
3957#endif
1da177e4 3958/* rx skb timestamps */
f629d208
JP
3959void net_enable_timestamp(void);
3960void net_disable_timestamp(void);
1da177e4 3961
20380731 3962#ifdef CONFIG_PROC_FS
f629d208 3963int __init dev_proc_init(void);
900ff8c6
CW
3964#else
3965#define dev_proc_init() 0
20380731
ACM
3966#endif
3967
4798248e 3968static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3969 struct sk_buff *skb, struct net_device *dev,
3970 bool more)
4798248e 3971{
fa2dbdc2 3972 skb->xmit_more = more ? 1 : 0;
0b725a2c 3973 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3974}
3975
10b3ad8c 3976static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3977 struct netdev_queue *txq, bool more)
4798248e
DM
3978{
3979 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3980 int rc;
4798248e 3981
fa2dbdc2 3982 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3983 if (rc == NETDEV_TX_OK)
3984 txq_trans_update(txq);
3985
3986 return rc;
4798248e
DM
3987}
3988
42a2d923
LT
3989int netdev_class_create_file_ns(struct class_attribute *class_attr,
3990 const void *ns);
3991void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3992 const void *ns);
58292cbe
TH
3993
3994static inline int netdev_class_create_file(struct class_attribute *class_attr)
3995{
3996 return netdev_class_create_file_ns(class_attr, NULL);
3997}
3998
3999static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4000{
4001 netdev_class_remove_file_ns(class_attr, NULL);
4002}
b8a9787e 4003
04600794
JB
4004extern struct kobj_ns_type_operations net_ns_type_operations;
4005
f629d208 4006const char *netdev_drivername(const struct net_device *dev);
6579e57b 4007
f629d208 4008void linkwatch_run_queue(void);
20380731 4009
da08143b
MK
4010static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4011 netdev_features_t f2)
4012{
c8cd0989
TH
4013 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4014 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4015 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4016 else
b6a0e72a 4017 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4018 }
da08143b 4019
c8cd0989 4020 return f1 & f2;
da08143b
MK
4021}
4022
c8f44aff
MM
4023static inline netdev_features_t netdev_get_wanted_features(
4024 struct net_device *dev)
5455c699
MM
4025{
4026 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4027}
c8f44aff
MM
4028netdev_features_t netdev_increment_features(netdev_features_t all,
4029 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4030
4031/* Allow TSO being used on stacked device :
4032 * Performing the GSO segmentation before last device
4033 * is a performance improvement.
4034 */
4035static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4036 netdev_features_t mask)
4037{
4038 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4039}
4040
6cb6a27c 4041int __netdev_update_features(struct net_device *dev);
5455c699 4042void netdev_update_features(struct net_device *dev);
afe12cc8 4043void netdev_change_features(struct net_device *dev);
7f353bf2 4044
fc4a7489
PM
4045void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4046 struct net_device *dev);
4047
e38f3025
TM
4048netdev_features_t passthru_features_check(struct sk_buff *skb,
4049 struct net_device *dev,
4050 netdev_features_t features);
c1e756bf 4051netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4052
4d29515f 4053static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4054{
7b748340 4055 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4056
4057 /* check flags correspondence */
4058 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4059 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4060 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4061 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4062 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4063 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4064 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4065 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4066 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4067 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4068 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4069 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4070 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4071 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4072 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4073 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
0345e186 4074
d6b4991a 4075 return (features & feature) == feature;
576a30eb
HX
4076}
4077
4d29515f 4078static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4079{
278b2513 4080 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4081 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4082}
4083
8b86a61d 4084static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4085 netdev_features_t features)
7967168c 4086{
fc741216 4087 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4088 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4089 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4090}
4091
82cc1a7a
PWJ
4092static inline void netif_set_gso_max_size(struct net_device *dev,
4093 unsigned int size)
4094{
4095 dev->gso_max_size = size;
4096}
4097
7a7ffbab
WCC
4098static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4099 int pulled_hlen, u16 mac_offset,
4100 int mac_len)
4101{
4102 skb->protocol = protocol;
4103 skb->encapsulation = 1;
4104 skb_push(skb, pulled_hlen);
4105 skb_reset_transport_header(skb);
4106 skb->mac_header = mac_offset;
4107 skb->network_header = skb->mac_header + mac_len;
4108 skb->mac_len = mac_len;
4109}
4110
3c175784
SD
4111static inline bool netif_is_macsec(const struct net_device *dev)
4112{
4113 return dev->priv_flags & IFF_MACSEC;
4114}
4115
b618aaa9 4116static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4117{
4118 return dev->priv_flags & IFF_MACVLAN;
4119}
4120
b618aaa9 4121static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4122{
4123 return dev->priv_flags & IFF_MACVLAN_PORT;
4124}
4125
b618aaa9 4126static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4127{
4128 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4129}
4130
b618aaa9 4131static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4132{
4133 return dev->priv_flags & IFF_IPVLAN_MASTER;
4134}
4135
b618aaa9 4136static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4137{
4138 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4139}
4140
b618aaa9 4141static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4142{
4143 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4144}
4145
3bdc0eba
BG
4146static inline bool netif_supports_nofcs(struct net_device *dev)
4147{
4148 return dev->priv_flags & IFF_SUPP_NOFCS;
4149}
4150
007979ea 4151static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4152{
007979ea 4153 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4154}
4155
fee6d4c7
DA
4156static inline bool netif_is_l3_slave(const struct net_device *dev)
4157{
4158 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4159}
4160
0894ae3f
JP
4161static inline bool netif_is_bridge_master(const struct net_device *dev)
4162{
4163 return dev->priv_flags & IFF_EBRIDGE;
4164}
4165
28f9ee22
VY
4166static inline bool netif_is_bridge_port(const struct net_device *dev)
4167{
4168 return dev->priv_flags & IFF_BRIDGE_PORT;
4169}
4170
35d4e172
JP
4171static inline bool netif_is_ovs_master(const struct net_device *dev)
4172{
4173 return dev->priv_flags & IFF_OPENVSWITCH;
4174}
4175
b618aaa9 4176static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4177{
4178 return dev->priv_flags & IFF_TEAM;
4179}
4180
b618aaa9 4181static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4182{
4183 return dev->priv_flags & IFF_TEAM_PORT;
4184}
4185
b618aaa9 4186static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4187{
4188 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4189}
4190
b618aaa9 4191static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4192{
4193 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4194}
4195
d4ab4286
KJ
4196static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4197{
4198 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4199}
4200
02875878
ED
4201/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4202static inline void netif_keep_dst(struct net_device *dev)
4203{
4204 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4205}
4206
18d3df3e
PA
4207/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4208static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4209{
4210 /* TODO: reserve and use an additional IFF bit, if we get more users */
4211 return dev->priv_flags & IFF_MACSEC;
4212}
4213
505d4f73 4214extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4215
571ba423
JP
4216/* Logging, debugging and troubleshooting/diagnostic helpers. */
4217
4218/* netdev_printk helpers, similar to dev_printk */
4219
4220static inline const char *netdev_name(const struct net_device *dev)
4221{
c6f854d5
VF
4222 if (!dev->name[0] || strchr(dev->name, '%'))
4223 return "(unnamed net_device)";
571ba423
JP
4224 return dev->name;
4225}
4226
ccc7f496
VF
4227static inline const char *netdev_reg_state(const struct net_device *dev)
4228{
4229 switch (dev->reg_state) {
4230 case NETREG_UNINITIALIZED: return " (uninitialized)";
4231 case NETREG_REGISTERED: return "";
4232 case NETREG_UNREGISTERING: return " (unregistering)";
4233 case NETREG_UNREGISTERED: return " (unregistered)";
4234 case NETREG_RELEASED: return " (released)";
4235 case NETREG_DUMMY: return " (dummy)";
4236 }
4237
4238 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4239 return " (unknown)";
4240}
4241
f629d208 4242__printf(3, 4)
6ea754eb
JP
4243void netdev_printk(const char *level, const struct net_device *dev,
4244 const char *format, ...);
f629d208 4245__printf(2, 3)
6ea754eb 4246void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4247__printf(2, 3)
6ea754eb 4248void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4249__printf(2, 3)
6ea754eb 4250void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4251__printf(2, 3)
6ea754eb 4252void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4253__printf(2, 3)
6ea754eb 4254void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4255__printf(2, 3)
6ea754eb 4256void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4257__printf(2, 3)
6ea754eb 4258void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4259
8909c9ad
VK
4260#define MODULE_ALIAS_NETDEV(device) \
4261 MODULE_ALIAS("netdev-" device)
4262
b558c96f 4263#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4264#define netdev_dbg(__dev, format, args...) \
4265do { \
ffa10cb4 4266 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4267} while (0)
b558c96f
JC
4268#elif defined(DEBUG)
4269#define netdev_dbg(__dev, format, args...) \
4270 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4271#else
4272#define netdev_dbg(__dev, format, args...) \
4273({ \
4274 if (0) \
4275 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4276})
4277#endif
4278
4279#if defined(VERBOSE_DEBUG)
4280#define netdev_vdbg netdev_dbg
4281#else
4282
4283#define netdev_vdbg(dev, format, args...) \
4284({ \
4285 if (0) \
4286 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4287 0; \
4288})
4289#endif
4290
4291/*
4292 * netdev_WARN() acts like dev_printk(), but with the key difference
4293 * of using a WARN/WARN_ON to get the message out, including the
4294 * file/line information and a backtrace.
4295 */
4296#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4297 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4298 netdev_reg_state(dev), ##args)
571ba423 4299
b3d95c5c
JP
4300/* netif printk helpers, similar to netdev_printk */
4301
4302#define netif_printk(priv, type, level, dev, fmt, args...) \
4303do { \
4304 if (netif_msg_##type(priv)) \
4305 netdev_printk(level, (dev), fmt, ##args); \
4306} while (0)
4307
f45f4321
JP
4308#define netif_level(level, priv, type, dev, fmt, args...) \
4309do { \
4310 if (netif_msg_##type(priv)) \
4311 netdev_##level(dev, fmt, ##args); \
4312} while (0)
4313
b3d95c5c 4314#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4315 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4316#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4317 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4318#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4319 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4320#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4321 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4322#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4323 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4324#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4325 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4326#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4327 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4328
0053ea9c 4329#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4330#define netif_dbg(priv, type, netdev, format, args...) \
4331do { \
4332 if (netif_msg_##type(priv)) \
b5fb0a03 4333 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4334} while (0)
0053ea9c
JP
4335#elif defined(DEBUG)
4336#define netif_dbg(priv, type, dev, format, args...) \
4337 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4338#else
4339#define netif_dbg(priv, type, dev, format, args...) \
4340({ \
4341 if (0) \
4342 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4343 0; \
4344})
4345#endif
4346
f617f276
EC
4347/* if @cond then downgrade to debug, else print at @level */
4348#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4349 do { \
4350 if (cond) \
4351 netif_dbg(priv, type, netdev, fmt, ##args); \
4352 else \
4353 netif_ ## level(priv, type, netdev, fmt, ##args); \
4354 } while (0)
4355
b3d95c5c 4356#if defined(VERBOSE_DEBUG)
bcfcc450 4357#define netif_vdbg netif_dbg
b3d95c5c
JP
4358#else
4359#define netif_vdbg(priv, type, dev, format, args...) \
4360({ \
4361 if (0) \
a4ed89cb 4362 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4363 0; \
4364})
4365#endif
571ba423 4366
900ff8c6
CW
4367/*
4368 * The list of packet types we will receive (as opposed to discard)
4369 * and the routines to invoke.
4370 *
4371 * Why 16. Because with 16 the only overlap we get on a hash of the
4372 * low nibble of the protocol value is RARP/SNAP/X.25.
4373 *
4374 * NOTE: That is no longer true with the addition of VLAN tags. Not
4375 * sure which should go first, but I bet it won't make much
4376 * difference if we are running VLANs. The good news is that
4377 * this protocol won't be in the list unless compiled in, so
4378 * the average user (w/out VLANs) will not be adversely affected.
4379 * --BLG
4380 *
4381 * 0800 IP
4382 * 8100 802.1Q VLAN
4383 * 0001 802.3
4384 * 0002 AX.25
4385 * 0004 802.2
4386 * 8035 RARP
4387 * 0005 SNAP
4388 * 0805 X.25
4389 * 0806 ARP
4390 * 8137 IPX
4391 * 0009 Localtalk
4392 * 86DD IPv6
4393 */
4394#define PTYPE_HASH_SIZE (16)
4395#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4396
385a154c 4397#endif /* _LINUX_NETDEVICE_H */