]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/blame - include/linux/netdevice.h
ptp: gianfar: Use high resolution frequency method.
[mirror_ubuntu-eoan-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
cf85d08f 44#include <net/dsa.h>
7a6b6f51 45#ifdef CONFIG_DCB
2f90b865
AD
46#include <net/dcbnl.h>
47#endif
5bc1421e 48#include <net/netprio_cgroup.h>
a050c33f 49
a59e2ecb 50#include <linux/netdev_features.h>
77162022 51#include <linux/neighbour.h>
607ca46e 52#include <uapi/linux/netdevice.h>
61bd3857 53#include <uapi/linux/if_bonding.h>
e4c6734e 54#include <uapi/linux/pkt_cls.h>
59cc1f61 55#include <linux/hashtable.h>
a59e2ecb 56
115c1d6e 57struct netpoll_info;
313162d0 58struct device;
c1f19b51 59struct phy_device;
704232c2
JB
60/* 802.11 specific */
61struct wireless_dev;
98a18b6f
AA
62/* 802.15.4 specific */
63struct wpan_dev;
03c57747 64struct mpls_dev;
7c46a640
AD
65/* UDP Tunnel offloads */
66struct udp_tunnel_info;
a7862b45 67struct bpf_prog;
1da177e4 68
f629d208
JP
69void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
d07d7507 71
9a1654ba
JP
72/* Backlog congestion levels */
73#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74#define NET_RX_DROP 1 /* packet dropped */
75
572a9d7b
PM
76/*
77 * Transmit return codes: transmit return codes originate from three different
78 * namespaces:
79 *
80 * - qdisc return codes
81 * - driver transmit return codes
82 * - errno values
83 *
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
90 * others are propagated to higher layers.
91 */
92
93/* qdisc ->enqueue() return codes. */
94#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
95#define NET_XMIT_DROP 0x01 /* skb dropped */
96#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 97#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 98
b9df3cb8
GR
99/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 102#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
103#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
104
dc1f8bf6 105/* Driver transmit return codes */
9a1654ba 106#define NETDEV_TX_MASK 0xf0
572a9d7b 107
dc1f8bf6 108enum netdev_tx {
572a9d7b 109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
112};
113typedef enum netdev_tx netdev_tx_t;
114
9a1654ba
JP
115/*
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
118 */
119static inline bool dev_xmit_complete(int rc)
120{
121 /*
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
126 */
127 if (likely(rc < NET_XMIT_MASK))
128 return true;
129
130 return false;
131}
132
1da177e4 133/*
5e82b4b2 134 * Compute the worst-case header length according to the protocols
1da177e4
LT
135 * used.
136 */
fe2918b0 137
c0eb4540
KS
138#if defined(CONFIG_HYPERV_NET)
139# define LL_MAX_HEADER 128
140#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
141# if defined(CONFIG_MAC80211_MESH)
142# define LL_MAX_HEADER 128
143# else
144# define LL_MAX_HEADER 96
145# endif
1da177e4 146#else
8388e3da 147# define LL_MAX_HEADER 32
1da177e4
LT
148#endif
149
d11ead75
BH
150#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
152#define MAX_HEADER LL_MAX_HEADER
153#else
154#define MAX_HEADER (LL_MAX_HEADER + 48)
155#endif
156
157/*
be1f3c2c
BH
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
1da177e4 160 */
fe2918b0 161
d94d9fee 162struct net_device_stats {
3cfde79c
BH
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
1da177e4 172 unsigned long collisions;
1da177e4 173 unsigned long rx_length_errors;
3cfde79c
BH
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
1da177e4
LT
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
1da177e4
LT
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
186};
187
1da177e4
LT
188
189#include <linux/cache.h>
190#include <linux/skbuff.h>
191
adc9300e 192#ifdef CONFIG_RPS
c5905afb
IM
193#include <linux/static_key.h>
194extern struct static_key rps_needed;
adc9300e
ED
195#endif
196
1da177e4
LT
197struct neighbour;
198struct neigh_parms;
199struct sk_buff;
200
f001fde5
JP
201struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
ccffad25
JP
205#define NETDEV_HW_ADDR_T_LAN 1
206#define NETDEV_HW_ADDR_T_SAN 2
207#define NETDEV_HW_ADDR_T_SLAVE 3
208#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 209#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 210 bool global_use;
4cd729b0 211 int sync_cnt;
8f8f103d 212 int refcount;
4543fbef 213 int synced;
f001fde5
JP
214 struct rcu_head rcu_head;
215};
216
31278e71
JP
217struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220};
221
22bedad3
JP
222#define netdev_hw_addr_list_count(l) ((l)->count)
223#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224#define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 226
22bedad3
JP
227#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229#define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 231
22bedad3
JP
232#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 234#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 236
d94d9fee 237struct hh_cache {
f6b72b62 238 u16 hh_len;
5c25f686 239 u16 __pad;
3644f0ce 240 seqlock_t hh_lock;
1da177e4
LT
241
242 /* cached hardware header; allow for machine alignment needs. */
243#define HH_DATA_MOD 16
244#define HH_DATA_OFF(__len) \
5ba0eac6 245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
246#define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
249};
250
5e82b4b2 251/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
252 * Alternative is:
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
255 *
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
258 */
259#define LL_RESERVED_SPACE(dev) \
f5184d26 260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 261#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 263
3b04ddde
SH
264struct header_ops {
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
95c96174 267 const void *saddr, unsigned int len);
3b04ddde 268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
270 void (*cache_update)(struct hh_cache *hh,
271 const struct net_device *dev,
272 const unsigned char *haddr);
2793a23a 273 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
274};
275
1da177e4 276/* These flag bits are private to the generic network queueing
5e82b4b2 277 * layer; they may not be explicitly referenced by any other
1da177e4
LT
278 * code.
279 */
280
d94d9fee 281enum netdev_state_t {
1da177e4
LT
282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
1da177e4 284 __LINK_STATE_NOCARRIER,
b00055aa
SR
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
1da177e4
LT
287};
288
289
290/*
5e82b4b2 291 * This structure holds boot-time configured netdevice settings. They
fe2918b0 292 * are then used in the device probing.
1da177e4
LT
293 */
294struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297};
298#define NETDEV_BOOT_SETUP_MAX 8
299
f629d208 300int __init netdev_boot_setup(char *str);
1da177e4 301
bea3348e
SH
302/*
303 * Structure for NAPI scheduling similar to tasklet but with weighting
304 */
305struct napi_struct {
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 309 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
310 * can remove from the list right before clearing the bit.
311 */
312 struct list_head poll_list;
313
314 unsigned long state;
315 int weight;
404f7c9e 316 unsigned int gro_count;
bea3348e
SH
317 int (*poll)(struct napi_struct *, int);
318#ifdef CONFIG_NETPOLL
bea3348e 319 int poll_owner;
bea3348e 320#endif
5d38a079 321 struct net_device *dev;
d565b0a1 322 struct sk_buff *gro_list;
5d38a079 323 struct sk_buff *skb;
3b47d303 324 struct hrtimer timer;
404f7c9e 325 struct list_head dev_list;
af12fa6e
ET
326 struct hlist_node napi_hash_node;
327 unsigned int napi_id;
bea3348e
SH
328};
329
d94d9fee 330enum {
bea3348e 331 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 332 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 333 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
334 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
335 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
336 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
337};
338
339enum {
340 NAPIF_STATE_SCHED = (1UL << NAPI_STATE_SCHED),
341 NAPIF_STATE_DISABLE = (1UL << NAPI_STATE_DISABLE),
342 NAPIF_STATE_NPSVC = (1UL << NAPI_STATE_NPSVC),
343 NAPIF_STATE_HASHED = (1UL << NAPI_STATE_HASHED),
344 NAPIF_STATE_NO_BUSY_POLL = (1UL << NAPI_STATE_NO_BUSY_POLL),
345 NAPIF_STATE_IN_BUSY_POLL = (1UL << NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
346};
347
5b252f0c 348enum gro_result {
d1c76af9
HX
349 GRO_MERGED,
350 GRO_MERGED_FREE,
351 GRO_HELD,
352 GRO_NORMAL,
353 GRO_DROP,
354};
5b252f0c 355typedef enum gro_result gro_result_t;
d1c76af9 356
8a4eb573
JP
357/*
358 * enum rx_handler_result - Possible return values for rx_handlers.
359 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
360 * further.
361 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
362 * case skb->dev was changed by rx_handler.
363 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 364 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
365 *
366 * rx_handlers are functions called from inside __netif_receive_skb(), to do
367 * special processing of the skb, prior to delivery to protocol handlers.
368 *
369 * Currently, a net_device can only have a single rx_handler registered. Trying
370 * to register a second rx_handler will return -EBUSY.
371 *
372 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
373 * To unregister a rx_handler on a net_device, use
374 * netdev_rx_handler_unregister().
375 *
376 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
377 * do with the skb.
378 *
5e82b4b2 379 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 380 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 381 * the skb to be delivered in some other way.
8a4eb573
JP
382 *
383 * If the rx_handler changed skb->dev, to divert the skb to another
384 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
385 * new device will be called if it exists.
386 *
5e82b4b2 387 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 388 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 389 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 390 *
5e82b4b2 391 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
392 * delivered, it should return RX_HANDLER_PASS.
393 *
394 * A device without a registered rx_handler will behave as if rx_handler
395 * returned RX_HANDLER_PASS.
396 */
397
398enum rx_handler_result {
399 RX_HANDLER_CONSUMED,
400 RX_HANDLER_ANOTHER,
401 RX_HANDLER_EXACT,
402 RX_HANDLER_PASS,
403};
404typedef enum rx_handler_result rx_handler_result_t;
405typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 406
f629d208 407void __napi_schedule(struct napi_struct *n);
bc9ad166 408void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 409
4d29515f 410static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
411{
412 return test_bit(NAPI_STATE_DISABLE, &n->state);
413}
414
bea3348e 415/**
5e82b4b2
BH
416 * napi_schedule_prep - check if NAPI can be scheduled
417 * @n: NAPI context
bea3348e
SH
418 *
419 * Test if NAPI routine is already running, and if not mark
5e82b4b2 420 * it as running. This is used as a condition variable to
a0a46196
DM
421 * insure only one NAPI poll instance runs. We also make
422 * sure there is no pending NAPI disable.
bea3348e 423 */
4d29515f 424static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 425{
a0a46196
DM
426 return !napi_disable_pending(n) &&
427 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
428}
429
430/**
431 * napi_schedule - schedule NAPI poll
5e82b4b2 432 * @n: NAPI context
bea3348e
SH
433 *
434 * Schedule NAPI poll routine to be called if it is not already
435 * running.
436 */
437static inline void napi_schedule(struct napi_struct *n)
438{
439 if (napi_schedule_prep(n))
440 __napi_schedule(n);
441}
442
bc9ad166
ED
443/**
444 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 445 * @n: NAPI context
bc9ad166
ED
446 *
447 * Variant of napi_schedule(), assuming hard irqs are masked.
448 */
449static inline void napi_schedule_irqoff(struct napi_struct *n)
450{
451 if (napi_schedule_prep(n))
452 __napi_schedule_irqoff(n);
453}
454
bfe13f54 455/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 456static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
457{
458 if (napi_schedule_prep(napi)) {
459 __napi_schedule(napi);
4d29515f 460 return true;
bfe13f54 461 }
4d29515f 462 return false;
bfe13f54
RD
463}
464
364b6055
ED
465bool __napi_complete(struct napi_struct *n);
466bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
467/**
468 * napi_complete - NAPI processing complete
5e82b4b2 469 * @n: NAPI context
bea3348e
SH
470 *
471 * Mark NAPI processing as complete.
3b47d303 472 * Consider using napi_complete_done() instead.
364b6055 473 * Return false if device should avoid rearming interrupts.
bea3348e 474 */
364b6055 475static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
476{
477 return napi_complete_done(n, 0);
478}
bea3348e 479
af12fa6e
ET
480/**
481 * napi_hash_del - remove a NAPI from global table
5e82b4b2 482 * @napi: NAPI context
af12fa6e 483 *
5e82b4b2 484 * Warning: caller must observe RCU grace period
34cbe27e
ED
485 * before freeing memory containing @napi, if
486 * this function returns true.
93d05d4a 487 * Note: core networking stack automatically calls it
5e82b4b2 488 * from netif_napi_del().
93d05d4a 489 * Drivers might want to call this helper to combine all
5e82b4b2 490 * the needed RCU grace periods into a single one.
af12fa6e 491 */
34cbe27e 492bool napi_hash_del(struct napi_struct *napi);
af12fa6e 493
bea3348e
SH
494/**
495 * napi_disable - prevent NAPI from scheduling
5e82b4b2 496 * @n: NAPI context
bea3348e
SH
497 *
498 * Stop NAPI from being scheduled on this context.
499 * Waits till any outstanding processing completes.
500 */
3b47d303 501void napi_disable(struct napi_struct *n);
bea3348e
SH
502
503/**
504 * napi_enable - enable NAPI scheduling
5e82b4b2 505 * @n: NAPI context
bea3348e
SH
506 *
507 * Resume NAPI from being scheduled on this context.
508 * Must be paired with napi_disable.
509 */
510static inline void napi_enable(struct napi_struct *n)
511{
512 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 513 smp_mb__before_atomic();
bea3348e 514 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 515 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
516}
517
c264c3de
SH
518/**
519 * napi_synchronize - wait until NAPI is not running
5e82b4b2 520 * @n: NAPI context
c264c3de
SH
521 *
522 * Wait until NAPI is done being scheduled on this context.
523 * Waits till any outstanding processing completes but
524 * does not disable future activations.
525 */
526static inline void napi_synchronize(const struct napi_struct *n)
527{
facc432f
AB
528 if (IS_ENABLED(CONFIG_SMP))
529 while (test_bit(NAPI_STATE_SCHED, &n->state))
530 msleep(1);
531 else
532 barrier();
c264c3de 533}
c264c3de 534
d94d9fee 535enum netdev_queue_state_t {
73466498
TH
536 __QUEUE_STATE_DRV_XOFF,
537 __QUEUE_STATE_STACK_XOFF,
c3f26a26 538 __QUEUE_STATE_FROZEN,
79d16385 539};
8e2f1a63
DB
540
541#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
542#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
543#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
544
545#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
546#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
547 QUEUE_STATE_FROZEN)
548#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
549 QUEUE_STATE_FROZEN)
550
73466498
TH
551/*
552 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
553 * netif_tx_* functions below are used to manipulate this flag. The
554 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
555 * queue independently. The netif_xmit_*stopped functions below are called
556 * to check if the queue has been stopped by the driver or stack (either
557 * of the XOFF bits are set in the state). Drivers should not need to call
558 * netif_xmit*stopped functions, they should only be using netif_tx_*.
559 */
79d16385 560
bb949fbd 561struct netdev_queue {
6a321cb3 562/*
5e82b4b2 563 * read-mostly part
6a321cb3 564 */
bb949fbd 565 struct net_device *dev;
46e5da40 566 struct Qdisc __rcu *qdisc;
b0e1e646 567 struct Qdisc *qdisc_sleeping;
ccf5ff69 568#ifdef CONFIG_SYSFS
1d24eb48
TH
569 struct kobject kobj;
570#endif
f2cd2d3e
ED
571#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
572 int numa_node;
573#endif
c0ef079c
FW
574 unsigned long tx_maxrate;
575 /*
576 * Number of TX timeouts for this queue
577 * (/sys/class/net/DEV/Q/trans_timeout)
578 */
579 unsigned long trans_timeout;
6a321cb3 580/*
5e82b4b2 581 * write-mostly part
6a321cb3
ED
582 */
583 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
584 int xmit_lock_owner;
9d21493b 585 /*
9b36627a 586 * Time (in jiffies) of last Tx
9d21493b
ED
587 */
588 unsigned long trans_start;
ccf5ff69 589
114cf580
TH
590 unsigned long state;
591
592#ifdef CONFIG_BQL
593 struct dql dql;
594#endif
e8a0464c 595} ____cacheline_aligned_in_smp;
bb949fbd 596
f2cd2d3e
ED
597static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 return q->numa_node;
601#else
b236da69 602 return NUMA_NO_NODE;
f2cd2d3e
ED
603#endif
604}
605
606static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
607{
608#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
609 q->numa_node = node;
610#endif
611}
612
df334545 613#ifdef CONFIG_RPS
0a9627f2
TH
614/*
615 * This structure holds an RPS map which can be of variable length. The
616 * map is an array of CPUs.
617 */
618struct rps_map {
619 unsigned int len;
620 struct rcu_head rcu;
621 u16 cpus[0];
622};
60b778ce 623#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 624
fec5e652 625/*
c445477d
BH
626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627 * tail pointer for that CPU's input queue at the time of last enqueue, and
628 * a hardware filter index.
fec5e652
TH
629 */
630struct rps_dev_flow {
631 u16 cpu;
c445477d 632 u16 filter;
fec5e652
TH
633 unsigned int last_qtail;
634};
c445477d 635#define RPS_NO_FILTER 0xffff
fec5e652
TH
636
637/*
638 * The rps_dev_flow_table structure contains a table of flow mappings.
639 */
640struct rps_dev_flow_table {
641 unsigned int mask;
642 struct rcu_head rcu;
fec5e652
TH
643 struct rps_dev_flow flows[0];
644};
645#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 646 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
647
648/*
649 * The rps_sock_flow_table contains mappings of flows to the last CPU
650 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
651 * Each entry is a 32bit value. Upper part is the high-order bits
652 * of flow hash, lower part is CPU number.
567e4b79 653 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
654 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 656 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
657 */
658struct rps_sock_flow_table {
567e4b79 659 u32 mask;
93c1af6c
ED
660
661 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 662};
567e4b79 663#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
664
665#define RPS_NO_CPU 0xffff
666
567e4b79
ED
667extern u32 rps_cpu_mask;
668extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
669
fec5e652
TH
670static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
671 u32 hash)
672{
673 if (table && hash) {
567e4b79
ED
674 unsigned int index = hash & table->mask;
675 u32 val = hash & ~rps_cpu_mask;
fec5e652 676
5e82b4b2 677 /* We only give a hint, preemption can change CPU under us */
567e4b79 678 val |= raw_smp_processor_id();
fec5e652 679
567e4b79
ED
680 if (table->ents[index] != val)
681 table->ents[index] = val;
fec5e652
TH
682 }
683}
684
c445477d 685#ifdef CONFIG_RFS_ACCEL
f629d208
JP
686bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
687 u16 filter_id);
c445477d 688#endif
a953be53 689#endif /* CONFIG_RPS */
c445477d 690
0a9627f2
TH
691/* This structure contains an instance of an RX queue. */
692struct netdev_rx_queue {
a953be53 693#ifdef CONFIG_RPS
6e3f7faf
ED
694 struct rps_map __rcu *rps_map;
695 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 696#endif
6e3f7faf 697 struct kobject kobj;
fe822240 698 struct net_device *dev;
0a9627f2 699} ____cacheline_aligned_in_smp;
a953be53
MD
700
701/*
702 * RX queue sysfs structures and functions.
703 */
704struct rx_queue_attribute {
705 struct attribute attr;
706 ssize_t (*show)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, char *buf);
708 ssize_t (*store)(struct netdev_rx_queue *queue,
709 struct rx_queue_attribute *attr, const char *buf, size_t len);
710};
d314774c 711
bf264145
TH
712#ifdef CONFIG_XPS
713/*
714 * This structure holds an XPS map which can be of variable length. The
715 * map is an array of queues.
716 */
717struct xps_map {
718 unsigned int len;
719 unsigned int alloc_len;
720 struct rcu_head rcu;
721 u16 queues[0];
722};
60b778ce 723#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
724#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
726
727/*
728 * This structure holds all XPS maps for device. Maps are indexed by CPU.
729 */
730struct xps_dev_maps {
731 struct rcu_head rcu;
a4177869 732 struct xps_map __rcu *cpu_map[0];
bf264145 733};
184c449f
AD
734#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
735 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
736#endif /* CONFIG_XPS */
737
4f57c087
JF
738#define TC_MAX_QUEUE 16
739#define TC_BITMASK 15
740/* HW offloaded queuing disciplines txq count and offset maps */
741struct netdev_tc_txq {
742 u16 count;
743 u16 offset;
744};
745
68bad94e
NP
746#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
747/*
748 * This structure is to hold information about the device
749 * configured to run FCoE protocol stack.
750 */
751struct netdev_fcoe_hbainfo {
752 char manufacturer[64];
753 char serial_number[64];
754 char hardware_version[64];
755 char driver_version[64];
756 char optionrom_version[64];
757 char firmware_version[64];
758 char model[256];
759 char model_description[256];
760};
761#endif
762
02637fce 763#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 764
02637fce
JP
765/* This structure holds a unique identifier to identify some
766 * physical item (port for example) used by a netdevice.
66b52b0d 767 */
02637fce
JP
768struct netdev_phys_item_id {
769 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
770 unsigned char id_len;
771};
772
d754f98b
SF
773static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
774 struct netdev_phys_item_id *b)
775{
776 return a->id_len == b->id_len &&
777 memcmp(a->id, b->id, a->id_len) == 0;
778}
779
99932d4f
DB
780typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
781 struct sk_buff *skb);
782
a1b7c5fd 783/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
784 * that are being passed to the netdevice through the setup_tc op.
785 */
786enum {
787 TC_SETUP_MQPRIO,
a1b7c5fd 788 TC_SETUP_CLSU32,
5b33f488 789 TC_SETUP_CLSFLOWER,
b87f7936 790 TC_SETUP_MATCHALL,
332ae8e2 791 TC_SETUP_CLSBPF,
16e5cc64
JF
792};
793
a1b7c5fd
JF
794struct tc_cls_u32_offload;
795
16e5cc64
JF
796struct tc_to_netdev {
797 unsigned int type;
798 union {
799 u8 tc;
a1b7c5fd 800 struct tc_cls_u32_offload *cls_u32;
5b33f488 801 struct tc_cls_flower_offload *cls_flower;
b87f7936 802 struct tc_cls_matchall_offload *cls_mall;
332ae8e2 803 struct tc_cls_bpf_offload *cls_bpf;
16e5cc64
JF
804 };
805};
806
a7862b45
BB
807/* These structures hold the attributes of xdp state that are being passed
808 * to the netdevice through the xdp op.
809 */
810enum xdp_netdev_command {
811 /* Set or clear a bpf program used in the earliest stages of packet
812 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
813 * is responsible for calling bpf_prog_put on any old progs that are
814 * stored. In case of error, the callee need not release the new prog
815 * reference, but on success it takes ownership and must bpf_prog_put
816 * when it is no longer used.
817 */
818 XDP_SETUP_PROG,
819 /* Check if a bpf program is set on the device. The callee should
820 * return true if a program is currently attached and running.
821 */
822 XDP_QUERY_PROG,
823};
824
825struct netdev_xdp {
826 enum xdp_netdev_command command;
827 union {
828 /* XDP_SETUP_PROG */
829 struct bpf_prog *prog;
830 /* XDP_QUERY_PROG */
831 bool prog_attached;
832 };
833};
16e5cc64 834
d314774c
SH
835/*
836 * This structure defines the management hooks for network devices.
00829823
SH
837 * The following hooks can be defined; unless noted otherwise, they are
838 * optional and can be filled with a null pointer.
d314774c
SH
839 *
840 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
841 * This function is called once when a network device is registered.
842 * The network device can use this for any late stage initialization
843 * or semantic validation. It can fail with an error code which will
844 * be propagated back to register_netdev.
d314774c
SH
845 *
846 * void (*ndo_uninit)(struct net_device *dev);
847 * This function is called when device is unregistered or when registration
848 * fails. It is not called if init fails.
849 *
850 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 851 * This function is called when a network device transitions to the up
d314774c
SH
852 * state.
853 *
854 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 855 * This function is called when a network device transitions to the down
d314774c
SH
856 * state.
857 *
dc1f8bf6
SH
858 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
859 * struct net_device *dev);
00829823 860 * Called when a packet needs to be transmitted.
e79d8429
RR
861 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
862 * the queue before that can happen; it's for obsolete devices and weird
863 * corner cases, but the stack really does a non-trivial amount
864 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 865 * Required; cannot be NULL.
00829823 866 *
cdba756f
ED
867 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
868 * netdev_features_t features);
869 * Adjusts the requested feature flags according to device-specific
870 * constraints, and returns the resulting flags. Must not modify
871 * the device state.
872 *
f663dd9a 873 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 874 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 875 * Called to decide which queue to use when device supports multiple
00829823
SH
876 * transmit queues.
877 *
d314774c
SH
878 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
879 * This function is called to allow device receiver to make
5e82b4b2 880 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
881 *
882 * void (*ndo_set_rx_mode)(struct net_device *dev);
883 * This function is called device changes address list filtering.
01789349 884 * If driver handles unicast address filtering, it should set
5e82b4b2 885 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
886 *
887 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
888 * This function is called when the Media Access Control address
37b607c5 889 * needs to be changed. If this interface is not defined, the
5e82b4b2 890 * MAC address can not be changed.
d314774c
SH
891 *
892 * int (*ndo_validate_addr)(struct net_device *dev);
893 * Test if Media Access Control address is valid for the device.
894 *
895 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
896 * Called when a user requests an ioctl which can't be handled by
897 * the generic interface code. If not defined ioctls return
d314774c
SH
898 * not supported error code.
899 *
900 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
901 * Used to set network devices bus interface parameters. This interface
5e82b4b2 902 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
903 * interface (PCI) for low level management.
904 *
905 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
906 * Called when a user wants to change the Maximum Transfer Unit
907 * of a device. If not defined, any request to change MTU will
908 * will return an error.
909 *
00829823 910 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 911 * Callback used when the transmitter has not made any progress
d314774c
SH
912 * for dev->watchdog ticks.
913 *
3cfde79c 914 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 915 * struct rtnl_link_stats64 *storage);
d308e38f 916 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 917 * Called when a user wants to get the network device usage
be1f3c2c 918 * statistics. Drivers must do one of the following:
3cfde79c
BH
919 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
920 * rtnl_link_stats64 structure passed by the caller.
82695d9b 921 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
922 * (which should normally be dev->stats) and return a pointer to
923 * it. The structure may be changed asynchronously only if each
924 * field is written atomically.
925 * 3. Update dev->stats asynchronously and atomically, and define
926 * neither operation.
d314774c 927 *
3df5b3c6 928 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
929 * Return true if this device supports offload stats of this attr_id.
930 *
931 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
932 * void *attr_data)
933 * Get statistics for offload operations by attr_id. Write it into the
934 * attr_data pointer.
935 *
5d632cb7 936 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 937 * If device supports VLAN filtering this function is called when a
80d5c368 938 * VLAN id is registered.
d314774c 939 *
5d632cb7 940 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 941 * If device supports VLAN filtering this function is called when a
80d5c368 942 * VLAN id is unregistered.
d314774c
SH
943 *
944 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
945 *
946 * SR-IOV management functions.
947 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
948 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
949 * u8 qos, __be16 proto);
ed616689
SC
950 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
951 * int max_tx_rate);
5f8444a3 952 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 953 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
954 * int (*ndo_get_vf_config)(struct net_device *dev,
955 * int vf, struct ifla_vf_info *ivf);
1d8faf48 956 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
957 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
958 * struct nlattr *port[]);
01a3d796
VZ
959 *
960 * Enable or disable the VF ability to query its RSS Redirection Table and
961 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 962 * with PF and querying it may introduce a theoretical security risk.
01a3d796 963 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 964 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
965 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
966 * Called to setup 'tc' number of traffic classes in the net device. This
967 * is always called from the stack with the rtnl lock held and netif tx
968 * queues stopped. This allows the netdevice to perform queue management
969 * safely.
c445477d 970 *
e9bce845
YZ
971 * Fiber Channel over Ethernet (FCoE) offload functions.
972 * int (*ndo_fcoe_enable)(struct net_device *dev);
973 * Called when the FCoE protocol stack wants to start using LLD for FCoE
974 * so the underlying device can perform whatever needed configuration or
975 * initialization to support acceleration of FCoE traffic.
976 *
977 * int (*ndo_fcoe_disable)(struct net_device *dev);
978 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
979 * so the underlying device can perform whatever needed clean-ups to
980 * stop supporting acceleration of FCoE traffic.
981 *
982 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
983 * struct scatterlist *sgl, unsigned int sgc);
984 * Called when the FCoE Initiator wants to initialize an I/O that
985 * is a possible candidate for Direct Data Placement (DDP). The LLD can
986 * perform necessary setup and returns 1 to indicate the device is set up
987 * successfully to perform DDP on this I/O, otherwise this returns 0.
988 *
989 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
990 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
991 * indicated by the FC exchange id 'xid', so the underlying device can
992 * clean up and reuse resources for later DDP requests.
993 *
994 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
995 * struct scatterlist *sgl, unsigned int sgc);
996 * Called when the FCoE Target wants to initialize an I/O that
997 * is a possible candidate for Direct Data Placement (DDP). The LLD can
998 * perform necessary setup and returns 1 to indicate the device is set up
999 * successfully to perform DDP on this I/O, otherwise this returns 0.
1000 *
68bad94e
NP
1001 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1002 * struct netdev_fcoe_hbainfo *hbainfo);
1003 * Called when the FCoE Protocol stack wants information on the underlying
1004 * device. This information is utilized by the FCoE protocol stack to
1005 * register attributes with Fiber Channel management service as per the
1006 * FC-GS Fabric Device Management Information(FDMI) specification.
1007 *
e9bce845
YZ
1008 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1009 * Called when the underlying device wants to override default World Wide
1010 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1011 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1012 * protocol stack to use.
1013 *
c445477d
BH
1014 * RFS acceleration.
1015 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1016 * u16 rxq_index, u32 flow_id);
1017 * Set hardware filter for RFS. rxq_index is the target queue index;
1018 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1019 * Return the filter ID on success, or a negative error code.
fbaec0ea 1020 *
8b98a70c 1021 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1022 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1023 * Called to make another netdev an underling.
1024 *
1025 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1026 * Called to release previously enslaved netdev.
5455c699
MM
1027 *
1028 * Feature/offload setting functions.
c8f44aff 1029 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1030 * Called to update device configuration to new features. Passed
1031 * feature set might be less than what was returned by ndo_fix_features()).
1032 * Must return >0 or -errno if it changed dev->features itself.
1033 *
edc7d573 1034 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1035 * struct net_device *dev,
f6f6424b 1036 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1037 * Adds an FDB entry to dev for addr.
1690be63
VY
1038 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1039 * struct net_device *dev,
f6f6424b 1040 * const unsigned char *addr, u16 vid)
77162022
JF
1041 * Deletes the FDB entry from dev coresponding to addr.
1042 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1043 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1044 * int *idx)
77162022
JF
1045 * Used to add FDB entries to dump requests. Implementers should add
1046 * entries to skb and update idx with the number of entries.
e5a55a89 1047 *
ad41faa8
ND
1048 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1049 * u16 flags)
e5a55a89 1050 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1051 * struct net_device *dev, u32 filter_mask,
1052 * int nlflags)
ad41faa8
ND
1053 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1054 * u16 flags);
4bf84c35
JP
1055 *
1056 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1057 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1058 * which do not represent real hardware may define this to allow their
1059 * userspace components to manage their virtual carrier state. Devices
1060 * that determine carrier state from physical hardware properties (eg
1061 * network cables) or protocol-dependent mechanisms (eg
1062 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1063 *
1064 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1065 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1066 * Called to get ID of physical port of this device. If driver does
1067 * not implement this, it is assumed that the hw is not able to have
1068 * multiple net devices on single physical port.
53cf5275 1069 *
7c46a640
AD
1070 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1071 * struct udp_tunnel_info *ti);
1072 * Called by UDP tunnel to notify a driver about the UDP port and socket
1073 * address family that a UDP tunnel is listnening to. It is called only
1074 * when a new port starts listening. The operation is protected by the
1075 * RTNL.
1076 *
1077 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1078 * struct udp_tunnel_info *ti);
1079 * Called by UDP tunnel to notify the driver about a UDP port and socket
1080 * address family that the UDP tunnel is not listening to anymore. The
1081 * operation is protected by the RTNL.
1082 *
a6cc0cfa
JF
1083 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1084 * struct net_device *dev)
1085 * Called by upper layer devices to accelerate switching or other
1086 * station functionality into hardware. 'pdev is the lowerdev
1087 * to use for the offload and 'dev' is the net device that will
1088 * back the offload. Returns a pointer to the private structure
1089 * the upper layer will maintain.
1090 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1091 * Called by upper layer device to delete the station created
1092 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1093 * the station and priv is the structure returned by the add
1094 * operation.
1095 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1096 * struct net_device *dev,
1097 * void *priv);
1098 * Callback to use for xmit over the accelerated station. This
1099 * is used in place of ndo_start_xmit on accelerated net
1100 * devices.
5e82b4b2
BH
1101 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1102 * struct net_device *dev
1103 * netdev_features_t features);
04ffcb25 1104 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1105 * performing offload operations on a given packet. This is to give
1106 * the device an opportunity to implement any restrictions that cannot
1107 * be otherwise expressed by feature flags. The check is called with
1108 * the set of features that the stack has calculated and it returns
1109 * those the driver believes to be appropriate.
822b3b2e
JF
1110 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1111 * int queue_index, u32 maxrate);
1112 * Called when a user wants to set a max-rate limitation of specific
1113 * TX queue.
a54acb3a
ND
1114 * int (*ndo_get_iflink)(const struct net_device *dev);
1115 * Called to get the iflink value of this device.
d746d707 1116 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1117 * bool proto_down);
d746d707
AK
1118 * This function is used to pass protocol port error state information
1119 * to the switch driver. The switch driver can react to the proto_down
1120 * by doing a phys down on the associated switch port.
fc4099f1
PS
1121 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1122 * This function is used to get egress tunnel information for given skb.
1123 * This is useful for retrieving outer tunnel header parameters while
1124 * sampling packet.
871b642a
PA
1125 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1126 * This function is used to specify the headroom that the skb must
1127 * consider when allocation skb during packet reception. Setting
1128 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1129 * forward. Setting a negative value resets the rx headroom to the
871b642a 1130 * default value.
a7862b45
BB
1131 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1132 * This function is used to set or query state related to XDP on the
1133 * netdevice. See definition of enum xdp_netdev_command for details.
d746d707 1134 *
d314774c
SH
1135 */
1136struct net_device_ops {
1137 int (*ndo_init)(struct net_device *dev);
1138 void (*ndo_uninit)(struct net_device *dev);
1139 int (*ndo_open)(struct net_device *dev);
1140 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1141 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1142 struct net_device *dev);
1143 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1144 struct net_device *dev,
1145 netdev_features_t features);
00829823 1146 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1147 struct sk_buff *skb,
99932d4f
DB
1148 void *accel_priv,
1149 select_queue_fallback_t fallback);
d314774c
SH
1150 void (*ndo_change_rx_flags)(struct net_device *dev,
1151 int flags);
d314774c 1152 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1153 int (*ndo_set_mac_address)(struct net_device *dev,
1154 void *addr);
d314774c 1155 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1156 int (*ndo_do_ioctl)(struct net_device *dev,
1157 struct ifreq *ifr, int cmd);
d314774c
SH
1158 int (*ndo_set_config)(struct net_device *dev,
1159 struct ifmap *map);
00829823
SH
1160 int (*ndo_change_mtu)(struct net_device *dev,
1161 int new_mtu);
1162 int (*ndo_neigh_setup)(struct net_device *dev,
1163 struct neigh_parms *);
d314774c
SH
1164 void (*ndo_tx_timeout) (struct net_device *dev);
1165
28172739
ED
1166 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1167 struct rtnl_link_stats64 *storage);
3df5b3c6 1168 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1169 int (*ndo_get_offload_stats)(int attr_id,
1170 const struct net_device *dev,
1171 void *attr_data);
d314774c
SH
1172 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1173
8e586137 1174 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1175 __be16 proto, u16 vid);
8e586137 1176 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1177 __be16 proto, u16 vid);
d314774c 1178#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1179 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1180 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1181 struct netpoll_info *info);
0e34e931 1182 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1183#endif
e0d1095a 1184#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1185 int (*ndo_busy_poll)(struct napi_struct *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.
536721b1
KK
1511 *
1512 * @needed_headroom: Extra headroom the hardware may need, but not in all
1513 * cases can this be guaranteed
1514 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1515 * cases can this be guaranteed. Some cases also use
1516 * LL_MAX_HEADER instead to allocate the skb
1517 *
1518 * interface address info:
1519 *
1520 * @perm_addr: Permanent hw address
1521 * @addr_assign_type: Hw address assignment type
1522 * @addr_len: Hardware address length
8626a0c8 1523 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1524 * @dev_id: Used to differentiate devices that share
1525 * the same link layer address
1526 * @dev_port: Used to differentiate devices that share
1527 * the same function
1528 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1529 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1530 * has been enabled due to the need to listen to
1531 * additional unicast addresses in a device that
1532 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1533 * @uc: unicast mac addresses
1534 * @mc: multicast mac addresses
1535 * @dev_addrs: list of device hw addresses
1536 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1537 * @promiscuity: Number of times the NIC is told to work in
1538 * promiscuous mode; if it becomes 0 the NIC will
1539 * exit promiscuous mode
536721b1
KK
1540 * @allmulti: Counter, enables or disables allmulticast mode
1541 *
1542 * @vlan_info: VLAN info
1543 * @dsa_ptr: dsa specific data
1544 * @tipc_ptr: TIPC specific data
1545 * @atalk_ptr: AppleTalk link
1546 * @ip_ptr: IPv4 specific data
1547 * @dn_ptr: DECnet specific data
1548 * @ip6_ptr: IPv6 specific data
1549 * @ax25_ptr: AX.25 specific data
1550 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1551 *
1552 * @last_rx: Time of last Rx
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;
1da177e4 1728
f5184d26
JB
1729 unsigned short needed_headroom;
1730 unsigned short needed_tailroom;
1731
1da177e4 1732 /* Interface address info. */
536721b1
KK
1733 unsigned char perm_addr[MAX_ADDR_LEN];
1734 unsigned char addr_assign_type;
1735 unsigned char addr_len;
a0a9663d 1736 unsigned short neigh_priv_len;
536721b1
KK
1737 unsigned short dev_id;
1738 unsigned short dev_port;
ccffad25 1739 spinlock_t addr_list_lock;
14ffbbb8
TG
1740 unsigned char name_assign_type;
1741 bool uc_promisc;
536721b1
KK
1742 struct netdev_hw_addr_list uc;
1743 struct netdev_hw_addr_list mc;
1744 struct netdev_hw_addr_list dev_addrs;
1745
4c3d5e7b
ED
1746#ifdef CONFIG_SYSFS
1747 struct kset *queues_kset;
1748#endif
9d45abe1
WC
1749 unsigned int promiscuity;
1750 unsigned int allmulti;
1da177e4 1751
1da177e4 1752
5e82b4b2 1753 /* Protocol-specific pointers */
65ac6a5f 1754
d11ead75 1755#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1756 struct vlan_info __rcu *vlan_info;
65ac6a5f 1757#endif
34a430d7 1758#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1759 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1760#endif
1761#if IS_ENABLED(CONFIG_TIPC)
536721b1 1762 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1763#endif
536721b1
KK
1764 void *atalk_ptr;
1765 struct in_device __rcu *ip_ptr;
1766 struct dn_dev __rcu *dn_ptr;
1767 struct inet6_dev __rcu *ip6_ptr;
1768 void *ax25_ptr;
1769 struct wireless_dev *ieee80211_ptr;
98a18b6f 1770 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1771#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1772 struct mpls_dev __rcu *mpls_ptr;
1773#endif
1da177e4 1774
9356b8fc 1775/*
cd13539b 1776 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1777 */
536721b1 1778 unsigned long last_rx;
4dc89133 1779
9356b8fc 1780 /* Interface address info used in eth_type_trans() */
536721b1 1781 unsigned char *dev_addr;
f001fde5 1782
a953be53 1783#ifdef CONFIG_SYSFS
0a9627f2
TH
1784 struct netdev_rx_queue *_rx;
1785
0a9627f2 1786 unsigned int num_rx_queues;
62fe0b40 1787 unsigned int real_num_rx_queues;
df334545 1788#endif
0a9627f2 1789
3b47d303 1790 unsigned long gro_flush_timeout;
61391cde 1791 rx_handler_func_t __rcu *rx_handler;
1792 void __rcu *rx_handler_data;
e8a0464c 1793
4cda01e8 1794#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1795 struct tcf_proto __rcu *ingress_cl_list;
1796#endif
24824a09 1797 struct netdev_queue __rcu *ingress_queue;
e687ad60 1798#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1799 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1800#endif
d2788d34 1801
536721b1 1802 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1803#ifdef CONFIG_RFS_ACCEL
1804 struct cpu_rmap *rx_cpu_rmap;
1805#endif
1806 struct hlist_node index_hlist;
cd13539b
ED
1807
1808/*
1809 * Cache lines mostly used on transmit path
1810 */
e8a0464c
DM
1811 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1812 unsigned int num_tx_queues;
fd2ea0a7 1813 unsigned int real_num_tx_queues;
af356afa 1814 struct Qdisc *qdisc;
59cc1f61
JK
1815#ifdef CONFIG_NET_SCHED
1816 DECLARE_HASHTABLE (qdisc_hash, 4);
1817#endif
536721b1 1818 unsigned long tx_queue_len;
c3f26a26 1819 spinlock_t tx_global_lock;
14ffbbb8 1820 int watchdog_timeo;
cd13539b 1821
bf264145 1822#ifdef CONFIG_XPS
a4177869 1823 struct xps_dev_maps __rcu *xps_maps;
bf264145 1824#endif
1f211a1b
DB
1825#ifdef CONFIG_NET_CLS_ACT
1826 struct tcf_proto __rcu *egress_cl_list;
1827#endif
0c4f691f 1828
9356b8fc 1829 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1830 struct timer_list watchdog_timer;
1831
29b4433d 1832 int __percpu *pcpu_refcnt;
1da177e4 1833 struct list_head todo_list;
1da177e4 1834
e014debe 1835 struct list_head link_watch_list;
572a103d 1836
1da177e4 1837 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1838 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1839 NETREG_UNREGISTERING, /* called unregister_netdevice */
1840 NETREG_UNREGISTERED, /* completed unregister todo */
1841 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1842 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1843 } reg_state:8;
1844
536721b1 1845 bool dismantle;
a2835763
PM
1846
1847 enum {
1848 RTNL_LINK_INITIALIZED,
1849 RTNL_LINK_INITIALIZING,
1850 } rtnl_link_state:16;
1da177e4 1851
d314774c 1852 void (*destructor)(struct net_device *dev);
1da177e4 1853
1da177e4 1854#ifdef CONFIG_NETPOLL
5fbee843 1855 struct netpoll_info __rcu *npinfo;
1da177e4 1856#endif
eae792b7 1857
0c5c9fb5 1858 possible_net_t nd_net;
4a1c5371 1859
4951704b 1860 /* mid-layer private */
a7855c78 1861 union {
536721b1
KK
1862 void *ml_priv;
1863 struct pcpu_lstats __percpu *lstats;
8f84985f 1864 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1865 struct pcpu_dstats __percpu *dstats;
1866 struct pcpu_vstats __percpu *vstats;
a7855c78 1867 };
536721b1 1868
3cc77ec7 1869 struct garp_port __rcu *garp_port;
febf018d 1870 struct mrp_port __rcu *mrp_port;
1da177e4 1871
5e82b4b2 1872 struct device dev;
0c509a6c 1873 const struct attribute_group *sysfs_groups[4];
a953be53 1874 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1875
38f7b870 1876 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1877
82cc1a7a
PWJ
1878 /* for setting kernel sock attribute on TCP connection setup */
1879#define GSO_MAX_SIZE 65536
1880 unsigned int gso_max_size;
30b678d8
BH
1881#define GSO_MAX_SEGS 65535
1882 u16 gso_max_segs;
743b03a8 1883
7a6b6f51 1884#ifdef CONFIG_DCB
32953543 1885 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1886#endif
5e82b4b2
BH
1887 u8 num_tc;
1888 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1889 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1890
d11ead75 1891#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1892 unsigned int fcoe_ddp_xid;
5bc1421e 1893#endif
86f8515f 1894#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1895 struct netprio_map __rcu *priomap;
4d288d57 1896#endif
5e82b4b2
BH
1897 struct phy_device *phydev;
1898 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1899 struct lock_class_key *qdisc_running_key;
5e82b4b2 1900 bool proto_down;
1da177e4 1901};
43cb76d9 1902#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1903
1904#define NETDEV_ALIGN 32
1da177e4 1905
4f57c087
JF
1906static inline
1907int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1908{
1909 return dev->prio_tc_map[prio & TC_BITMASK];
1910}
1911
1912static inline
1913int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1914{
1915 if (tc >= dev->num_tc)
1916 return -EINVAL;
1917
1918 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1919 return 0;
1920}
1921
8d059b0f 1922int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1923void netdev_reset_tc(struct net_device *dev);
1924int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1925int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1926
1927static inline
1928int netdev_get_num_tc(struct net_device *dev)
1929{
1930 return dev->num_tc;
1931}
1932
e8a0464c
DM
1933static inline
1934struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1935 unsigned int index)
1936{
1937 return &dev->_tx[index];
1938}
1939
10c51b56
DB
1940static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1941 const struct sk_buff *skb)
1942{
1943 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1944}
1945
e8a0464c
DM
1946static inline void netdev_for_each_tx_queue(struct net_device *dev,
1947 void (*f)(struct net_device *,
1948 struct netdev_queue *,
1949 void *),
1950 void *arg)
1951{
1952 unsigned int i;
1953
1954 for (i = 0; i < dev->num_tx_queues; i++)
1955 f(dev, &dev->_tx[i], arg);
1956}
1957
d3fff6c4
ED
1958#define netdev_lockdep_set_classes(dev) \
1959{ \
1960 static struct lock_class_key qdisc_tx_busylock_key; \
1961 static struct lock_class_key qdisc_running_key; \
1962 static struct lock_class_key qdisc_xmit_lock_key; \
1963 static struct lock_class_key dev_addr_list_lock_key; \
1964 unsigned int i; \
1965 \
1966 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1967 (dev)->qdisc_running_key = &qdisc_running_key; \
1968 lockdep_set_class(&(dev)->addr_list_lock, \
1969 &dev_addr_list_lock_key); \
1970 for (i = 0; i < (dev)->num_tx_queues; i++) \
1971 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1972 &qdisc_xmit_lock_key); \
1973}
1974
f629d208 1975struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1976 struct sk_buff *skb,
1977 void *accel_priv);
8c4c49df 1978
871b642a
PA
1979/* returns the headroom that the master device needs to take in account
1980 * when forwarding to this dev
1981 */
1982static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1983{
1984 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1985}
1986
1987static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1988{
1989 if (dev->netdev_ops->ndo_set_rx_headroom)
1990 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1991}
1992
1993/* set the device rx headroom to the dev's default */
1994static inline void netdev_reset_rx_headroom(struct net_device *dev)
1995{
1996 netdev_set_rx_headroom(dev, -1);
1997}
1998
c346dca1
YH
1999/*
2000 * Net namespace inlines
2001 */
2002static inline
2003struct net *dev_net(const struct net_device *dev)
2004{
c2d9ba9b 2005 return read_pnet(&dev->nd_net);
c346dca1
YH
2006}
2007
2008static inline
f5aa23fd 2009void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2010{
0c5c9fb5 2011 write_pnet(&dev->nd_net, net);
c346dca1
YH
2012}
2013
3e8a72d1 2014static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 2015{
3fc88677 2016#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
2017 if (dev->dsa_ptr != NULL)
2018 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 2019#endif
5aed85ce 2020 return false;
396138f0
LB
2021}
2022
bea3348e
SH
2023/**
2024 * netdev_priv - access network device private data
2025 * @dev: network device
2026 *
2027 * Get network device private data
2028 */
6472ce60 2029static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2030{
1ce8e7b5 2031 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2032}
2033
1da177e4
LT
2034/* Set the sysfs physical device reference for the network logical device
2035 * if set prior to registration will cause a symlink during initialization.
2036 */
43cb76d9 2037#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2038
384912ed 2039/* Set the sysfs device type for the network logical device to allow
3f79410c 2040 * fine-grained identification of different network device types. For
5e82b4b2 2041 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2042 */
2043#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2044
82dc3c63
ED
2045/* Default NAPI poll() weight
2046 * Device drivers are strongly advised to not use bigger value
2047 */
2048#define NAPI_POLL_WEIGHT 64
2049
3b582cc1 2050/**
5e82b4b2 2051 * netif_napi_add - initialize a NAPI context
3b582cc1 2052 * @dev: network device
5e82b4b2 2053 * @napi: NAPI context
3b582cc1
SH
2054 * @poll: polling function
2055 * @weight: default weight
2056 *
5e82b4b2
BH
2057 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2058 * *any* of the other NAPI-related functions.
3b582cc1 2059 */
d565b0a1
HX
2060void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2061 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2062
d64b5e85 2063/**
5e82b4b2 2064 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2065 * @dev: network device
5e82b4b2 2066 * @napi: NAPI context
d64b5e85
ED
2067 * @poll: polling function
2068 * @weight: default weight
2069 *
2070 * This variant of netif_napi_add() should be used from drivers using NAPI
2071 * to exclusively poll a TX queue.
2072 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2073 */
2074static inline void netif_tx_napi_add(struct net_device *dev,
2075 struct napi_struct *napi,
2076 int (*poll)(struct napi_struct *, int),
2077 int weight)
2078{
2079 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2080 netif_napi_add(dev, napi, poll, weight);
2081}
2082
d8156534 2083/**
5e82b4b2
BH
2084 * netif_napi_del - remove a NAPI context
2085 * @napi: NAPI context
d8156534 2086 *
5e82b4b2 2087 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2088 */
d565b0a1
HX
2089void netif_napi_del(struct napi_struct *napi);
2090
2091struct napi_gro_cb {
78a478d0 2092 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2093 void *frag0;
78a478d0 2094
7489594c
HX
2095 /* Length of frag0. */
2096 unsigned int frag0_len;
2097
86911732 2098 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2099 int data_offset;
86911732 2100
d565b0a1 2101 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2102 u16 flush;
2103
2104 /* Save the IP ID here and check when we get to the transport layer */
2105 u16 flush_id;
d565b0a1
HX
2106
2107 /* Number of segments aggregated. */
2e71a6f8
ED
2108 u16 count;
2109
15e2396d
TH
2110 /* Start offset for remote checksum offload */
2111 u16 gro_remcsum_start;
2112
2e71a6f8
ED
2113 /* jiffies when first packet was created/queued */
2114 unsigned long age;
86347245 2115
afe93325 2116 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2117 u16 proto;
2118
baa32ff4
TH
2119 /* This is non-zero if the packet may be of the same flow. */
2120 u8 same_flow:1;
2121
fac8e0f5
JG
2122 /* Used in tunnel GRO receive */
2123 u8 encap_mark:1;
573e8fca
TH
2124
2125 /* GRO checksum is valid */
2126 u8 csum_valid:1;
2127
662880f4
TH
2128 /* Number of checksums via CHECKSUM_UNNECESSARY */
2129 u8 csum_cnt:3;
c3c7c254 2130
baa32ff4
TH
2131 /* Free the skb? */
2132 u8 free:2;
2133#define NAPI_GRO_FREE 1
2134#define NAPI_GRO_FREE_STOLEN_HEAD 2
2135
efc98d08
TH
2136 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2137 u8 is_ipv6:1;
2138
a0ca153f
AD
2139 /* Used in GRE, set in fou/gue_gro_receive */
2140 u8 is_fou:1;
2141
1530545e
AD
2142 /* Used to determine if flush_id can be ignored */
2143 u8 is_atomic:1;
2144
fcd91dd4
SD
2145 /* Number of gro_receive callbacks this packet already went through */
2146 u8 recursion_counter:4;
2147
2148 /* 1 bit hole */
baa32ff4 2149
bf5a755f
JC
2150 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2151 __wsum csum;
2152
c3c7c254
ED
2153 /* used in skb_gro_receive() slow path */
2154 struct sk_buff *last;
d565b0a1
HX
2155};
2156
2157#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2158
fcd91dd4
SD
2159#define GRO_RECURSION_LIMIT 15
2160static inline int gro_recursion_inc_test(struct sk_buff *skb)
2161{
2162 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2163}
2164
2165typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2166static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2167 struct sk_buff **head,
2168 struct sk_buff *skb)
2169{
2170 if (unlikely(gro_recursion_inc_test(skb))) {
2171 NAPI_GRO_CB(skb)->flush |= 1;
2172 return NULL;
2173 }
2174
2175 return cb(head, skb);
2176}
2177
2178typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2179 struct sk_buff *);
2180static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2181 struct sock *sk,
2182 struct sk_buff **head,
2183 struct sk_buff *skb)
2184{
2185 if (unlikely(gro_recursion_inc_test(skb))) {
2186 NAPI_GRO_CB(skb)->flush |= 1;
2187 return NULL;
2188 }
2189
2190 return cb(sk, head, skb);
2191}
2192
1da177e4 2193struct packet_type {
f2ccd8fa
DM
2194 __be16 type; /* This is really htons(ether_type). */
2195 struct net_device *dev; /* NULL is wildcarded here */
2196 int (*func) (struct sk_buff *,
2197 struct net_device *,
2198 struct packet_type *,
2199 struct net_device *);
c0de08d0
EL
2200 bool (*id_match)(struct packet_type *ptype,
2201 struct sock *sk);
1da177e4
LT
2202 void *af_packet_priv;
2203 struct list_head list;
2204};
2205
f191a1d1 2206struct offload_callbacks {
576a30eb 2207 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2208 netdev_features_t features);
d565b0a1 2209 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2210 struct sk_buff *skb);
299603e8 2211 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2212};
2213
2214struct packet_offload {
2215 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2216 u16 priority;
f191a1d1
VY
2217 struct offload_callbacks callbacks;
2218 struct list_head list;
1da177e4
LT
2219};
2220
5e82b4b2 2221/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2222struct pcpu_sw_netstats {
2223 u64 rx_packets;
2224 u64 rx_bytes;
2225 u64 tx_packets;
2226 u64 tx_bytes;
2227 struct u64_stats_sync syncp;
2228};
2229
aabc92bb
PNA
2230#define __netdev_alloc_pcpu_stats(type, gfp) \
2231({ \
2232 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2233 if (pcpu_stats) { \
2234 int __cpu; \
2235 for_each_possible_cpu(__cpu) { \
2236 typeof(type) *stat; \
2237 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2238 u64_stats_init(&stat->syncp); \
2239 } \
2240 } \
2241 pcpu_stats; \
1c213bd2
WC
2242})
2243
aabc92bb 2244#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2245 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2246
764f5e54
JP
2247enum netdev_lag_tx_type {
2248 NETDEV_LAG_TX_TYPE_UNKNOWN,
2249 NETDEV_LAG_TX_TYPE_RANDOM,
2250 NETDEV_LAG_TX_TYPE_BROADCAST,
2251 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2252 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2253 NETDEV_LAG_TX_TYPE_HASH,
2254};
2255
2256struct netdev_lag_upper_info {
2257 enum netdev_lag_tx_type tx_type;
2258};
2259
fb1b2e3c
JP
2260struct netdev_lag_lower_state_info {
2261 u8 link_up : 1,
2262 tx_enabled : 1;
2263};
2264
1da177e4
LT
2265#include <linux/notifier.h>
2266
dcfe1421
AW
2267/* netdevice notifier chain. Please remember to update the rtnetlink
2268 * notification exclusion list in rtnetlink_event() when adding new
2269 * types.
2270 */
2271#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2272#define NETDEV_DOWN 0x0002
2273#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2274 detected a hardware crash and restarted
2275 - we can use this eg to kick tcp sessions
2276 once done */
2277#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2278#define NETDEV_REGISTER 0x0005
2279#define NETDEV_UNREGISTER 0x0006
1d486bfb 2280#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2281#define NETDEV_CHANGEADDR 0x0008
2282#define NETDEV_GOING_DOWN 0x0009
2283#define NETDEV_CHANGENAME 0x000A
2284#define NETDEV_FEAT_CHANGE 0x000B
2285#define NETDEV_BONDING_FAILOVER 0x000C
2286#define NETDEV_PRE_UP 0x000D
2287#define NETDEV_PRE_TYPE_CHANGE 0x000E
2288#define NETDEV_POST_TYPE_CHANGE 0x000F
2289#define NETDEV_POST_INIT 0x0010
0115e8e3 2290#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2291#define NETDEV_RELEASE 0x0012
2292#define NETDEV_NOTIFY_PEERS 0x0013
2293#define NETDEV_JOIN 0x0014
42e52bf9 2294#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2295#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2296#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2297#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2298#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2299#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2300#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2301#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2302#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2303
f629d208
JP
2304int register_netdevice_notifier(struct notifier_block *nb);
2305int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2306
2307struct netdev_notifier_info {
2308 struct net_device *dev;
2309};
2310
be9efd36
JP
2311struct netdev_notifier_change_info {
2312 struct netdev_notifier_info info; /* must be first */
2313 unsigned int flags_changed;
2314};
2315
0e4ead9d
JP
2316struct netdev_notifier_changeupper_info {
2317 struct netdev_notifier_info info; /* must be first */
2318 struct net_device *upper_dev; /* new upper dev */
2319 bool master; /* is upper dev master */
5e82b4b2 2320 bool linking; /* is the notification for link or unlink */
29bf24af 2321 void *upper_info; /* upper dev info */
0e4ead9d
JP
2322};
2323
04d48266
JP
2324struct netdev_notifier_changelowerstate_info {
2325 struct netdev_notifier_info info; /* must be first */
2326 void *lower_state_info; /* is lower dev state */
2327};
2328
75538c2b
CW
2329static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2330 struct net_device *dev)
2331{
2332 info->dev = dev;
2333}
2334
351638e7
JP
2335static inline struct net_device *
2336netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2337{
2338 return info->dev;
2339}
2340
f629d208 2341int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2342
2343
1da177e4
LT
2344extern rwlock_t dev_base_lock; /* Device list lock */
2345
881d966b
EB
2346#define for_each_netdev(net, d) \
2347 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2348#define for_each_netdev_reverse(net, d) \
2349 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2350#define for_each_netdev_rcu(net, d) \
2351 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2352#define for_each_netdev_safe(net, d, n) \
2353 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2354#define for_each_netdev_continue(net, d) \
2355 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2356#define for_each_netdev_continue_rcu(net, d) \
2357 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2358#define for_each_netdev_in_bond_rcu(bond, slave) \
2359 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2360 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2361#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2362
a050c33f
DL
2363static inline struct net_device *next_net_device(struct net_device *dev)
2364{
2365 struct list_head *lh;
2366 struct net *net;
2367
c346dca1 2368 net = dev_net(dev);
a050c33f
DL
2369 lh = dev->dev_list.next;
2370 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2371}
2372
ce81b76a
ED
2373static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2374{
2375 struct list_head *lh;
2376 struct net *net;
2377
2378 net = dev_net(dev);
ccf43438 2379 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2380 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2381}
2382
a050c33f
DL
2383static inline struct net_device *first_net_device(struct net *net)
2384{
2385 return list_empty(&net->dev_base_head) ? NULL :
2386 net_device_entry(net->dev_base_head.next);
2387}
7562f876 2388
ccf43438
ED
2389static inline struct net_device *first_net_device_rcu(struct net *net)
2390{
2391 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2392
2393 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2394}
2395
f629d208
JP
2396int netdev_boot_setup_check(struct net_device *dev);
2397unsigned long netdev_boot_base(const char *prefix, int unit);
2398struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2399 const char *hwaddr);
2400struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2401struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2402void dev_add_pack(struct packet_type *pt);
2403void dev_remove_pack(struct packet_type *pt);
2404void __dev_remove_pack(struct packet_type *pt);
2405void dev_add_offload(struct packet_offload *po);
2406void dev_remove_offload(struct packet_offload *po);
f629d208 2407
a54acb3a 2408int dev_get_iflink(const struct net_device *dev);
fc4099f1 2409int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2410struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2411 unsigned short mask);
f629d208
JP
2412struct net_device *dev_get_by_name(struct net *net, const char *name);
2413struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2414struct net_device *__dev_get_by_name(struct net *net, const char *name);
2415int dev_alloc_name(struct net_device *dev, const char *name);
2416int dev_open(struct net_device *dev);
2417int dev_close(struct net_device *dev);
99c4a26a 2418int dev_close_many(struct list_head *head, bool unlink);
f629d208 2419void dev_disable_lro(struct net_device *dev);
0c4b51f0 2420int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2421int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2422int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2423int register_netdevice(struct net_device *dev);
2424void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2425void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2426static inline void unregister_netdevice(struct net_device *dev)
2427{
2428 unregister_netdevice_queue(dev, NULL);
2429}
2430
f629d208
JP
2431int netdev_refcnt_read(const struct net_device *dev);
2432void free_netdev(struct net_device *dev);
74d332c1 2433void netdev_freemem(struct net_device *dev);
f629d208
JP
2434void synchronize_net(void);
2435int init_dummy_netdev(struct net_device *dev);
937f1ba5 2436
f60e5990 2437DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2438#define XMIT_RECURSION_LIMIT 10
2439
f60e5990 2440static inline int dev_recursion_level(void)
2441{
2442 return this_cpu_read(xmit_recursion);
2443}
2444
f629d208
JP
2445struct net_device *dev_get_by_index(struct net *net, int ifindex);
2446struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2447struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2448int netdev_get_name(struct net *net, char *name, int ifindex);
2449int dev_restart(struct net_device *dev);
f629d208 2450int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2451
2452static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2453{
2454 return NAPI_GRO_CB(skb)->data_offset;
2455}
2456
2457static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2458{
2459 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2460}
2461
2462static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2463{
2464 NAPI_GRO_CB(skb)->data_offset += len;
2465}
2466
a5b1cf28
HX
2467static inline void *skb_gro_header_fast(struct sk_buff *skb,
2468 unsigned int offset)
86911732 2469{
a5b1cf28
HX
2470 return NAPI_GRO_CB(skb)->frag0 + offset;
2471}
78a478d0 2472
a5b1cf28
HX
2473static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2474{
2475 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2476}
78a478d0 2477
a5b1cf28
HX
2478static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2479 unsigned int offset)
2480{
17dd759c
HX
2481 if (!pskb_may_pull(skb, hlen))
2482 return NULL;
2483
a5b1cf28
HX
2484 NAPI_GRO_CB(skb)->frag0 = NULL;
2485 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2486 return skb->data + offset;
86911732 2487}
1da177e4 2488
36e7b1b8
HX
2489static inline void *skb_gro_network_header(struct sk_buff *skb)
2490{
78d3fd0b
HX
2491 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2492 skb_network_offset(skb);
36e7b1b8
HX
2493}
2494
bf5a755f
JC
2495static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2496 const void *start, unsigned int len)
2497{
573e8fca 2498 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2499 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2500 csum_partial(start, len, 0));
2501}
2502
573e8fca
TH
2503/* GRO checksum functions. These are logical equivalents of the normal
2504 * checksum functions (in skbuff.h) except that they operate on the GRO
2505 * offsets and fields in sk_buff.
2506 */
2507
2508__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2509
15e2396d
TH
2510static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2511{
b7fe10e5 2512 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2513}
2514
573e8fca
TH
2515static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2516 bool zero_okay,
2517 __sum16 check)
2518{
6edec0e6
TH
2519 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2520 skb_checksum_start_offset(skb) <
2521 skb_gro_offset(skb)) &&
15e2396d 2522 !skb_at_gro_remcsum_start(skb) &&
662880f4 2523 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2524 (!zero_okay || check));
2525}
2526
2527static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2528 __wsum psum)
2529{
2530 if (NAPI_GRO_CB(skb)->csum_valid &&
2531 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2532 return 0;
2533
2534 NAPI_GRO_CB(skb)->csum = psum;
2535
2536 return __skb_gro_checksum_complete(skb);
2537}
2538
573e8fca
TH
2539static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2540{
662880f4
TH
2541 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2542 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2543 NAPI_GRO_CB(skb)->csum_cnt--;
2544 } else {
2545 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2546 * verified a new top level checksum or an encapsulated one
2547 * during GRO. This saves work if we fallback to normal path.
2548 */
2549 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2550 }
2551}
2552
2553#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2554 compute_pseudo) \
2555({ \
2556 __sum16 __ret = 0; \
2557 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2558 __ret = __skb_gro_checksum_validate_complete(skb, \
2559 compute_pseudo(skb, proto)); \
5a212329
TH
2560 if (__ret) \
2561 __skb_mark_checksum_bad(skb); \
2562 else \
573e8fca
TH
2563 skb_gro_incr_csum_unnecessary(skb); \
2564 __ret; \
2565})
2566
2567#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2568 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2569
2570#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2571 compute_pseudo) \
2572 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2573
2574#define skb_gro_checksum_simple_validate(skb) \
2575 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2576
d96535a1
TH
2577static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2578{
2579 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2580 !NAPI_GRO_CB(skb)->csum_valid);
2581}
2582
2583static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2584 __sum16 check, __wsum pseudo)
2585{
2586 NAPI_GRO_CB(skb)->csum = ~pseudo;
2587 NAPI_GRO_CB(skb)->csum_valid = 1;
2588}
2589
2590#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2591do { \
2592 if (__skb_gro_checksum_convert_check(skb)) \
2593 __skb_gro_checksum_convert(skb, check, \
2594 compute_pseudo(skb, proto)); \
2595} while (0)
2596
26c4f7da
TH
2597struct gro_remcsum {
2598 int offset;
2599 __wsum delta;
2600};
2601
2602static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2603{
846cd667 2604 grc->offset = 0;
26c4f7da
TH
2605 grc->delta = 0;
2606}
2607
b7fe10e5
TH
2608static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2609 unsigned int off, size_t hdrlen,
2610 int start, int offset,
2611 struct gro_remcsum *grc,
2612 bool nopartial)
dcdc8994
TH
2613{
2614 __wsum delta;
b7fe10e5 2615 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2616
2617 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2618
15e2396d 2619 if (!nopartial) {
b7fe10e5
TH
2620 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2621 return ptr;
2622 }
2623
2624 ptr = skb_gro_header_fast(skb, off);
2625 if (skb_gro_header_hard(skb, off + plen)) {
2626 ptr = skb_gro_header_slow(skb, off + plen, off);
2627 if (!ptr)
2628 return NULL;
15e2396d
TH
2629 }
2630
b7fe10e5
TH
2631 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2632 start, offset);
dcdc8994
TH
2633
2634 /* Adjust skb->csum since we changed the packet */
dcdc8994 2635 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2636
b7fe10e5 2637 grc->offset = off + hdrlen + offset;
26c4f7da 2638 grc->delta = delta;
b7fe10e5
TH
2639
2640 return ptr;
dcdc8994
TH
2641}
2642
26c4f7da
TH
2643static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2644 struct gro_remcsum *grc)
2645{
b7fe10e5
TH
2646 void *ptr;
2647 size_t plen = grc->offset + sizeof(u16);
2648
26c4f7da
TH
2649 if (!grc->delta)
2650 return;
2651
b7fe10e5
TH
2652 ptr = skb_gro_header_fast(skb, grc->offset);
2653 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2654 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2655 if (!ptr)
2656 return;
2657 }
2658
2659 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2660}
dcdc8994 2661
0c4e8581
SH
2662static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2663 unsigned short type,
3b04ddde 2664 const void *daddr, const void *saddr,
95c96174 2665 unsigned int len)
0c4e8581 2666{
f1ecfd5d 2667 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2668 return 0;
3b04ddde
SH
2669
2670 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2671}
2672
b95cce35
SH
2673static inline int dev_parse_header(const struct sk_buff *skb,
2674 unsigned char *haddr)
2675{
2676 const struct net_device *dev = skb->dev;
2677
1b83336b 2678 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2679 return 0;
3b04ddde 2680 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2681}
2682
2793a23a
WB
2683/* ll_header must have at least hard_header_len allocated */
2684static inline bool dev_validate_header(const struct net_device *dev,
2685 char *ll_header, int len)
2686{
2687 if (likely(len >= dev->hard_header_len))
2688 return true;
2689
2690 if (capable(CAP_SYS_RAWIO)) {
2691 memset(ll_header + len, 0, dev->hard_header_len - len);
2692 return true;
2693 }
2694
2695 if (dev->header_ops && dev->header_ops->validate)
2696 return dev->header_ops->validate(ll_header, len);
2697
2698 return false;
2699}
2700
1da177e4 2701typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2702int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2703static inline int unregister_gifconf(unsigned int family)
2704{
2705 return register_gifconf(family, NULL);
2706}
2707
99bbc707 2708#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2709#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2710struct sd_flow_limit {
2711 u64 count;
2712 unsigned int num_buckets;
2713 unsigned int history_head;
2714 u16 history[FLOW_LIMIT_HISTORY];
2715 u8 buckets[];
2716};
2717
2718extern int netdev_flow_limit_table_len;
2719#endif /* CONFIG_NET_FLOW_LIMIT */
2720
1da177e4 2721/*
5e82b4b2 2722 * Incoming packets are placed on per-CPU queues
1da177e4 2723 */
d94d9fee 2724struct softnet_data {
1da177e4 2725 struct list_head poll_list;
6e7676c1 2726 struct sk_buff_head process_queue;
1da177e4 2727
dee42870 2728 /* stats */
cd7b5396
DM
2729 unsigned int processed;
2730 unsigned int time_squeeze;
cd7b5396 2731 unsigned int received_rps;
fd793d89 2732#ifdef CONFIG_RPS
88751275 2733 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2734#endif
2735#ifdef CONFIG_NET_FLOW_LIMIT
2736 struct sd_flow_limit __rcu *flow_limit;
2737#endif
2738 struct Qdisc *output_queue;
2739 struct Qdisc **output_queue_tailp;
2740 struct sk_buff *completion_queue;
88751275 2741
4cdb1e2e 2742#ifdef CONFIG_RPS
501e7ef5
ED
2743 /* input_queue_head should be written by cpu owning this struct,
2744 * and only read by other cpus. Worth using a cache line.
2745 */
2746 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2747
2748 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2749 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2750 struct softnet_data *rps_ipi_next;
2751 unsigned int cpu;
76cc8b13 2752 unsigned int input_queue_tail;
1e94d72f 2753#endif
95c96174 2754 unsigned int dropped;
0a9627f2 2755 struct sk_buff_head input_pkt_queue;
bea3348e 2756 struct napi_struct backlog;
99bbc707 2757
1da177e4
LT
2758};
2759
76cc8b13 2760static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2761{
2762#ifdef CONFIG_RPS
76cc8b13
TH
2763 sd->input_queue_head++;
2764#endif
2765}
2766
2767static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2768 unsigned int *qtail)
2769{
2770#ifdef CONFIG_RPS
2771 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2772#endif
2773}
2774
0a9627f2 2775DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2776
f629d208 2777void __netif_schedule(struct Qdisc *q);
46e5da40 2778void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2779
fd2ea0a7
DM
2780static inline void netif_tx_schedule_all(struct net_device *dev)
2781{
2782 unsigned int i;
2783
2784 for (i = 0; i < dev->num_tx_queues; i++)
2785 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2786}
2787
f9a7cbbf 2788static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2789{
73466498 2790 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2791}
2792
bea3348e
SH
2793/**
2794 * netif_start_queue - allow transmit
2795 * @dev: network device
2796 *
2797 * Allow upper layers to call the device hard_start_xmit routine.
2798 */
1da177e4
LT
2799static inline void netif_start_queue(struct net_device *dev)
2800{
e8a0464c 2801 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2802}
2803
fd2ea0a7
DM
2804static inline void netif_tx_start_all_queues(struct net_device *dev)
2805{
2806 unsigned int i;
2807
2808 for (i = 0; i < dev->num_tx_queues; i++) {
2809 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2810 netif_tx_start_queue(txq);
2811 }
2812}
2813
46e5da40 2814void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2815
d29f749e
DJ
2816/**
2817 * netif_wake_queue - restart transmit
2818 * @dev: network device
2819 *
2820 * Allow upper layers to call the device hard_start_xmit routine.
2821 * Used for flow control when transmit resources are available.
2822 */
79d16385
DM
2823static inline void netif_wake_queue(struct net_device *dev)
2824{
e8a0464c 2825 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2826}
2827
fd2ea0a7
DM
2828static inline void netif_tx_wake_all_queues(struct net_device *dev)
2829{
2830 unsigned int i;
2831
2832 for (i = 0; i < dev->num_tx_queues; i++) {
2833 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2834 netif_tx_wake_queue(txq);
2835 }
2836}
2837
f9a7cbbf 2838static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2839{
73466498 2840 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2841}
2842
bea3348e
SH
2843/**
2844 * netif_stop_queue - stop transmitted packets
2845 * @dev: network device
2846 *
2847 * Stop upper layers calling the device hard_start_xmit routine.
2848 * Used for flow control when transmit resources are unavailable.
2849 */
1da177e4
LT
2850static inline void netif_stop_queue(struct net_device *dev)
2851{
e8a0464c 2852 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2853}
2854
a2029240 2855void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2856
4d29515f 2857static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2858{
73466498 2859 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2860}
2861
bea3348e
SH
2862/**
2863 * netif_queue_stopped - test if transmit queue is flowblocked
2864 * @dev: network device
2865 *
2866 * Test if transmit queue on device is currently unable to send.
2867 */
4d29515f 2868static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2869{
e8a0464c 2870 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2871}
2872
4d29515f 2873static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2874{
73466498
TH
2875 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2876}
2877
8e2f1a63
DB
2878static inline bool
2879netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2880{
2881 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2882}
2883
8e2f1a63
DB
2884static inline bool
2885netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2886{
2887 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2888}
2889
53511453
ED
2890/**
2891 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2892 * @dev_queue: pointer to transmit queue
2893 *
2894 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2895 * to give appropriate hint to the CPU.
53511453
ED
2896 */
2897static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2898{
2899#ifdef CONFIG_BQL
2900 prefetchw(&dev_queue->dql.num_queued);
2901#endif
2902}
2903
2904/**
2905 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2906 * @dev_queue: pointer to transmit queue
2907 *
2908 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2909 * to give appropriate hint to the CPU.
53511453
ED
2910 */
2911static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2912{
2913#ifdef CONFIG_BQL
2914 prefetchw(&dev_queue->dql.limit);
2915#endif
2916}
2917
c5d67bd7
TH
2918static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2919 unsigned int bytes)
2920{
114cf580
TH
2921#ifdef CONFIG_BQL
2922 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2923
2924 if (likely(dql_avail(&dev_queue->dql) >= 0))
2925 return;
2926
2927 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2928
2929 /*
2930 * The XOFF flag must be set before checking the dql_avail below,
2931 * because in netdev_tx_completed_queue we update the dql_completed
2932 * before checking the XOFF flag.
2933 */
2934 smp_mb();
2935
2936 /* check again in case another CPU has just made room avail */
2937 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2938 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2939#endif
c5d67bd7
TH
2940}
2941
0042d0c8
FF
2942/**
2943 * netdev_sent_queue - report the number of bytes queued to hardware
2944 * @dev: network device
2945 * @bytes: number of bytes queued to the hardware device queue
2946 *
2947 * Report the number of bytes queued for sending/completion to the network
2948 * device hardware queue. @bytes should be a good approximation and should
2949 * exactly match netdev_completed_queue() @bytes
2950 */
c5d67bd7
TH
2951static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2952{
2953 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2954}
2955
2956static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2957 unsigned int pkts, unsigned int bytes)
c5d67bd7 2958{
114cf580 2959#ifdef CONFIG_BQL
b37c0fbe
AD
2960 if (unlikely(!bytes))
2961 return;
2962
2963 dql_completed(&dev_queue->dql, bytes);
2964
2965 /*
2966 * Without the memory barrier there is a small possiblity that
2967 * netdev_tx_sent_queue will miss the update and cause the queue to
2968 * be stopped forever
2969 */
2970 smp_mb();
2971
2972 if (dql_avail(&dev_queue->dql) < 0)
2973 return;
2974
2975 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2976 netif_schedule_queue(dev_queue);
114cf580 2977#endif
c5d67bd7
TH
2978}
2979
0042d0c8
FF
2980/**
2981 * netdev_completed_queue - report bytes and packets completed by device
2982 * @dev: network device
2983 * @pkts: actual number of packets sent over the medium
2984 * @bytes: actual number of bytes sent over the medium
2985 *
2986 * Report the number of bytes and packets transmitted by the network device
2987 * hardware queue over the physical medium, @bytes must exactly match the
2988 * @bytes amount passed to netdev_sent_queue()
2989 */
c5d67bd7 2990static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2991 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2992{
2993 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2994}
2995
2996static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2997{
114cf580 2998#ifdef CONFIG_BQL
5c490354 2999 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3000 dql_reset(&q->dql);
3001#endif
c5d67bd7
TH
3002}
3003
0042d0c8
FF
3004/**
3005 * netdev_reset_queue - reset the packets and bytes count of a network device
3006 * @dev_queue: network device
3007 *
3008 * Reset the bytes and packet count of a network device and clear the
3009 * software flow control OFF bit for this network device
3010 */
c5d67bd7
TH
3011static inline void netdev_reset_queue(struct net_device *dev_queue)
3012{
3013 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3014}
3015
b9507bda
DB
3016/**
3017 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3018 * @dev: network device
3019 * @queue_index: given tx queue index
3020 *
3021 * Returns 0 if given tx queue index >= number of device tx queues,
3022 * otherwise returns the originally passed tx queue index.
3023 */
3024static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3025{
3026 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3027 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3028 dev->name, queue_index,
3029 dev->real_num_tx_queues);
3030 return 0;
3031 }
3032
3033 return queue_index;
3034}
3035
bea3348e
SH
3036/**
3037 * netif_running - test if up
3038 * @dev: network device
3039 *
3040 * Test if the device has been brought up.
3041 */
4d29515f 3042static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3043{
3044 return test_bit(__LINK_STATE_START, &dev->state);
3045}
3046
f25f4e44 3047/*
5e82b4b2 3048 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3049 * stop, and a check if it's stopped. All other device management is
3050 * done at the overall netdevice level.
3051 * Also test the device if we're multiqueue.
3052 */
bea3348e
SH
3053
3054/**
3055 * netif_start_subqueue - allow sending packets on subqueue
3056 * @dev: network device
3057 * @queue_index: sub queue index
3058 *
3059 * Start individual transmit queue of a device with multiple transmit queues.
3060 */
f25f4e44
PWJ
3061static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3062{
fd2ea0a7 3063 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3064
3065 netif_tx_start_queue(txq);
f25f4e44
PWJ
3066}
3067
bea3348e
SH
3068/**
3069 * netif_stop_subqueue - stop sending packets on subqueue
3070 * @dev: network device
3071 * @queue_index: sub queue index
3072 *
3073 * Stop individual transmit queue of a device with multiple transmit queues.
3074 */
f25f4e44
PWJ
3075static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3076{
fd2ea0a7 3077 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3078 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3079}
3080
bea3348e
SH
3081/**
3082 * netif_subqueue_stopped - test status of subqueue
3083 * @dev: network device
3084 * @queue_index: sub queue index
3085 *
3086 * Check individual transmit queue of a device with multiple transmit queues.
3087 */
4d29515f
DM
3088static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3089 u16 queue_index)
f25f4e44 3090{
fd2ea0a7 3091 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3092
3093 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3094}
3095
4d29515f
DM
3096static inline bool netif_subqueue_stopped(const struct net_device *dev,
3097 struct sk_buff *skb)
668f895a
PE
3098{
3099 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3100}
bea3348e 3101
46e5da40 3102void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 3103
537c00de 3104#ifdef CONFIG_XPS
53af53ae 3105int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3106 u16 index);
537c00de
AD
3107#else
3108static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3109 const struct cpumask *mask,
537c00de
AD
3110 u16 index)
3111{
3112 return 0;
3113}
3114#endif
3115
5605c762
JP
3116u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3117 unsigned int num_tx_queues);
3118
a3d22a68
VZ
3119/*
3120 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3121 * as a distribution range limit for the returned value.
3122 */
3123static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3124 struct sk_buff *skb)
a3d22a68
VZ
3125{
3126 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3127}
3128
bea3348e
SH
3129/**
3130 * netif_is_multiqueue - test if device has multiple transmit queues
3131 * @dev: network device
3132 *
3133 * Check if device has multiple transmit queues
bea3348e 3134 */
4d29515f 3135static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3136{
a02cec21 3137 return dev->num_tx_queues > 1;
f25f4e44 3138}
1da177e4 3139
f629d208 3140int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3141
a953be53 3142#ifdef CONFIG_SYSFS
f629d208 3143int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3144#else
3145static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3146 unsigned int rxq)
3147{
3148 return 0;
3149}
3150#endif
3151
a953be53
MD
3152#ifdef CONFIG_SYSFS
3153static inline unsigned int get_netdev_rx_queue_index(
3154 struct netdev_rx_queue *queue)
3155{
3156 struct net_device *dev = queue->dev;
3157 int index = queue - dev->_rx;
3158
3159 BUG_ON(index >= dev->num_rx_queues);
3160 return index;
3161}
3162#endif
3163
16917b87 3164#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3165int netif_get_num_default_rss_queues(void);
16917b87 3166
e6247027
ED
3167enum skb_free_reason {
3168 SKB_REASON_CONSUMED,
3169 SKB_REASON_DROPPED,
3170};
3171
3172void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3173void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3174
e6247027
ED
3175/*
3176 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3177 * interrupt context or with hardware interrupts being disabled.
3178 * (in_irq() || irqs_disabled())
3179 *
3180 * We provide four helpers that can be used in following contexts :
3181 *
3182 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3183 * replacing kfree_skb(skb)
3184 *
3185 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3186 * Typically used in place of consume_skb(skb) in TX completion path
3187 *
3188 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3189 * replacing kfree_skb(skb)
3190 *
3191 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3192 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3193 */
e6247027
ED
3194static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3195{
3196 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3197}
3198
3199static inline void dev_consume_skb_irq(struct sk_buff *skb)
3200{
3201 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3202}
3203
3204static inline void dev_kfree_skb_any(struct sk_buff *skb)
3205{
3206 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3207}
3208
3209static inline void dev_consume_skb_any(struct sk_buff *skb)
3210{
3211 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3212}
1da177e4 3213
f629d208
JP
3214int netif_rx(struct sk_buff *skb);
3215int netif_rx_ni(struct sk_buff *skb);
04eb4489 3216int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3217gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3218void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3219struct sk_buff *napi_get_frags(struct napi_struct *napi);
3220gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3221struct packet_offload *gro_find_receive_by_type(__be16 type);
3222struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3223
3224static inline void napi_free_frags(struct napi_struct *napi)
3225{
3226 kfree_skb(napi->skb);
3227 napi->skb = NULL;
3228}
3229
24b27fc4 3230bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3231int netdev_rx_handler_register(struct net_device *dev,
3232 rx_handler_func_t *rx_handler,
3233 void *rx_handler_data);
3234void netdev_rx_handler_unregister(struct net_device *dev);
3235
3236bool dev_valid_name(const char *name);
3237int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3238int dev_ethtool(struct net *net, struct ifreq *);
3239unsigned int dev_get_flags(const struct net_device *);
3240int __dev_change_flags(struct net_device *, unsigned int flags);
3241int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3242void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3243 unsigned int gchanges);
f629d208
JP
3244int dev_change_name(struct net_device *, const char *);
3245int dev_set_alias(struct net_device *, const char *, size_t);
3246int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3247int dev_set_mtu(struct net_device *, int);
3248void dev_set_group(struct net_device *, int);
3249int dev_set_mac_address(struct net_device *, struct sockaddr *);
3250int dev_change_carrier(struct net_device *, bool new_carrier);
3251int dev_get_phys_port_id(struct net_device *dev,
02637fce 3252 struct netdev_phys_item_id *ppid);
db24a904
DA
3253int dev_get_phys_port_name(struct net_device *dev,
3254 char *name, size_t len);
d746d707 3255int dev_change_proto_down(struct net_device *dev, bool proto_down);
a7862b45 3256int dev_change_xdp_fd(struct net_device *dev, int fd);
55a93b3e 3257struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3258struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3259 struct netdev_queue *txq, int *ret);
a0265d28 3260int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3261int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3262bool is_skb_forwardable(const struct net_device *dev,
3263 const struct sk_buff *skb);
1da177e4 3264
4e3264d2
MKL
3265static __always_inline int ____dev_forward_skb(struct net_device *dev,
3266 struct sk_buff *skb)
3267{
3268 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3269 unlikely(!is_skb_forwardable(dev, skb))) {
3270 atomic_long_inc(&dev->rx_dropped);
3271 kfree_skb(skb);
3272 return NET_RX_DROP;
3273 }
3274
3275 skb_scrub_packet(skb, true);
3276 skb->priority = 0;
3277 return 0;
3278}
3279
74b20582
DA
3280void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3281
20380731 3282extern int netdev_budget;
1da177e4
LT
3283
3284/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3285void netdev_run_todo(void);
1da177e4 3286
bea3348e
SH
3287/**
3288 * dev_put - release reference to device
3289 * @dev: network device
3290 *
9ef4429b 3291 * Release reference to device to allow it to be freed.
bea3348e 3292 */
1da177e4
LT
3293static inline void dev_put(struct net_device *dev)
3294{
933393f5 3295 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3296}
3297
bea3348e
SH
3298/**
3299 * dev_hold - get reference to device
3300 * @dev: network device
3301 *
9ef4429b 3302 * Hold reference to device to keep it from being freed.
bea3348e 3303 */
15333061
SH
3304static inline void dev_hold(struct net_device *dev)
3305{
933393f5 3306 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3307}
1da177e4
LT
3308
3309/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3310 * and _off may be called from IRQ context, but it is caller
3311 * who is responsible for serialization of these calls.
b00055aa
SR
3312 *
3313 * The name carrier is inappropriate, these functions should really be
3314 * called netif_lowerlayer_*() because they represent the state of any
3315 * kind of lower layer not just hardware media.
1da177e4
LT
3316 */
3317
f629d208
JP
3318void linkwatch_init_dev(struct net_device *dev);
3319void linkwatch_fire_event(struct net_device *dev);
3320void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3321
bea3348e
SH
3322/**
3323 * netif_carrier_ok - test if carrier present
3324 * @dev: network device
3325 *
3326 * Check if carrier is present on device
3327 */
4d29515f 3328static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3329{
3330 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3331}
3332
f629d208 3333unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3334
f629d208 3335void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3336
f629d208 3337void netif_carrier_on(struct net_device *dev);
1da177e4 3338
f629d208 3339void netif_carrier_off(struct net_device *dev);
1da177e4 3340
bea3348e
SH
3341/**
3342 * netif_dormant_on - mark device as dormant.
3343 * @dev: network device
3344 *
3345 * Mark device as dormant (as per RFC2863).
3346 *
3347 * The dormant state indicates that the relevant interface is not
3348 * actually in a condition to pass packets (i.e., it is not 'up') but is
3349 * in a "pending" state, waiting for some external event. For "on-
3350 * demand" interfaces, this new state identifies the situation where the
3351 * interface is waiting for events to place it in the up state.
bea3348e 3352 */
b00055aa
SR
3353static inline void netif_dormant_on(struct net_device *dev)
3354{
3355 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3356 linkwatch_fire_event(dev);
3357}
3358
bea3348e
SH
3359/**
3360 * netif_dormant_off - set device as not dormant.
3361 * @dev: network device
3362 *
3363 * Device is not in dormant state.
3364 */
b00055aa
SR
3365static inline void netif_dormant_off(struct net_device *dev)
3366{
3367 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3368 linkwatch_fire_event(dev);
3369}
3370
bea3348e
SH
3371/**
3372 * netif_dormant - test if carrier present
3373 * @dev: network device
3374 *
3375 * Check if carrier is present on device
3376 */
4d29515f 3377static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3378{
3379 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3380}
3381
3382
bea3348e
SH
3383/**
3384 * netif_oper_up - test if device is operational
3385 * @dev: network device
3386 *
3387 * Check if carrier is operational
3388 */
4d29515f 3389static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3390{
b00055aa
SR
3391 return (dev->operstate == IF_OPER_UP ||
3392 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3393}
3394
bea3348e
SH
3395/**
3396 * netif_device_present - is device available or removed
3397 * @dev: network device
3398 *
3399 * Check if device has not been removed from system.
3400 */
4d29515f 3401static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3402{
3403 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3404}
3405
f629d208 3406void netif_device_detach(struct net_device *dev);
1da177e4 3407
f629d208 3408void netif_device_attach(struct net_device *dev);
1da177e4
LT
3409
3410/*
3411 * Network interface message level settings
3412 */
1da177e4
LT
3413
3414enum {
3415 NETIF_MSG_DRV = 0x0001,
3416 NETIF_MSG_PROBE = 0x0002,
3417 NETIF_MSG_LINK = 0x0004,
3418 NETIF_MSG_TIMER = 0x0008,
3419 NETIF_MSG_IFDOWN = 0x0010,
3420 NETIF_MSG_IFUP = 0x0020,
3421 NETIF_MSG_RX_ERR = 0x0040,
3422 NETIF_MSG_TX_ERR = 0x0080,
3423 NETIF_MSG_TX_QUEUED = 0x0100,
3424 NETIF_MSG_INTR = 0x0200,
3425 NETIF_MSG_TX_DONE = 0x0400,
3426 NETIF_MSG_RX_STATUS = 0x0800,
3427 NETIF_MSG_PKTDATA = 0x1000,
3428 NETIF_MSG_HW = 0x2000,
3429 NETIF_MSG_WOL = 0x4000,
3430};
3431
3432#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3433#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3434#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3435#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3436#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3437#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3438#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3439#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3440#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3441#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3442#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3443#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3444#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3445#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3446#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3447
3448static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3449{
3450 /* use default */
3451 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3452 return default_msg_enable_bits;
3453 if (debug_value == 0) /* no output */
3454 return 0;
3455 /* set low N bits */
3456 return (1 << debug_value) - 1;
3457}
3458
c773e847 3459static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3460{
c773e847
DM
3461 spin_lock(&txq->_xmit_lock);
3462 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3463}
3464
fd2ea0a7
DM
3465static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3466{
3467 spin_lock_bh(&txq->_xmit_lock);
3468 txq->xmit_lock_owner = smp_processor_id();
3469}
3470
4d29515f 3471static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3472{
4d29515f 3473 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3474 if (likely(ok))
3475 txq->xmit_lock_owner = smp_processor_id();
3476 return ok;
3477}
3478
3479static inline void __netif_tx_unlock(struct netdev_queue *txq)
3480{
3481 txq->xmit_lock_owner = -1;
3482 spin_unlock(&txq->_xmit_lock);
3483}
3484
3485static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3486{
3487 txq->xmit_lock_owner = -1;
3488 spin_unlock_bh(&txq->_xmit_lock);
3489}
3490
08baf561
ED
3491static inline void txq_trans_update(struct netdev_queue *txq)
3492{
3493 if (txq->xmit_lock_owner != -1)
3494 txq->trans_start = jiffies;
3495}
3496
ba162f8e
FW
3497/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3498static inline void netif_trans_update(struct net_device *dev)
3499{
9b36627a
FW
3500 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3501
3502 if (txq->trans_start != jiffies)
3503 txq->trans_start = jiffies;
ba162f8e
FW
3504}
3505
d29f749e
DJ
3506/**
3507 * netif_tx_lock - grab network device transmit lock
3508 * @dev: network device
d29f749e
DJ
3509 *
3510 * Get network device transmit lock
3511 */
22dd7495
JHS
3512static inline void netif_tx_lock(struct net_device *dev)
3513{
e8a0464c 3514 unsigned int i;
c3f26a26 3515 int cpu;
c773e847 3516
c3f26a26
DM
3517 spin_lock(&dev->tx_global_lock);
3518 cpu = smp_processor_id();
e8a0464c
DM
3519 for (i = 0; i < dev->num_tx_queues; i++) {
3520 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3521
3522 /* We are the only thread of execution doing a
3523 * freeze, but we have to grab the _xmit_lock in
3524 * order to synchronize with threads which are in
3525 * the ->hard_start_xmit() handler and already
3526 * checked the frozen bit.
3527 */
e8a0464c 3528 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3529 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3530 __netif_tx_unlock(txq);
e8a0464c 3531 }
932ff279
HX
3532}
3533
3534static inline void netif_tx_lock_bh(struct net_device *dev)
3535{
e8a0464c
DM
3536 local_bh_disable();
3537 netif_tx_lock(dev);
932ff279
HX
3538}
3539
932ff279
HX
3540static inline void netif_tx_unlock(struct net_device *dev)
3541{
e8a0464c
DM
3542 unsigned int i;
3543
3544 for (i = 0; i < dev->num_tx_queues; i++) {
3545 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3546
c3f26a26
DM
3547 /* No need to grab the _xmit_lock here. If the
3548 * queue is not stopped for another reason, we
3549 * force a schedule.
3550 */
3551 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3552 netif_schedule_queue(txq);
c3f26a26
DM
3553 }
3554 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3555}
3556
3557static inline void netif_tx_unlock_bh(struct net_device *dev)
3558{
e8a0464c
DM
3559 netif_tx_unlock(dev);
3560 local_bh_enable();
932ff279
HX
3561}
3562
c773e847 3563#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3564 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3565 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3566 } \
3567}
3568
5efeac44
EB
3569#define HARD_TX_TRYLOCK(dev, txq) \
3570 (((dev->features & NETIF_F_LLTX) == 0) ? \
3571 __netif_tx_trylock(txq) : \
3572 true )
3573
c773e847 3574#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3575 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3576 __netif_tx_unlock(txq); \
22dd7495
JHS
3577 } \
3578}
3579
1da177e4
LT
3580static inline void netif_tx_disable(struct net_device *dev)
3581{
fd2ea0a7 3582 unsigned int i;
c3f26a26 3583 int cpu;
fd2ea0a7 3584
c3f26a26
DM
3585 local_bh_disable();
3586 cpu = smp_processor_id();
fd2ea0a7
DM
3587 for (i = 0; i < dev->num_tx_queues; i++) {
3588 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3589
3590 __netif_tx_lock(txq, cpu);
fd2ea0a7 3591 netif_tx_stop_queue(txq);
c3f26a26 3592 __netif_tx_unlock(txq);
fd2ea0a7 3593 }
c3f26a26 3594 local_bh_enable();
1da177e4
LT
3595}
3596
e308a5d8
DM
3597static inline void netif_addr_lock(struct net_device *dev)
3598{
3599 spin_lock(&dev->addr_list_lock);
3600}
3601
2429f7ac
JP
3602static inline void netif_addr_lock_nested(struct net_device *dev)
3603{
25175ba5
VY
3604 int subclass = SINGLE_DEPTH_NESTING;
3605
3606 if (dev->netdev_ops->ndo_get_lock_subclass)
3607 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3608
3609 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3610}
3611
e308a5d8
DM
3612static inline void netif_addr_lock_bh(struct net_device *dev)
3613{
3614 spin_lock_bh(&dev->addr_list_lock);
3615}
3616
3617static inline void netif_addr_unlock(struct net_device *dev)
3618{
3619 spin_unlock(&dev->addr_list_lock);
3620}
3621
3622static inline void netif_addr_unlock_bh(struct net_device *dev)
3623{
3624 spin_unlock_bh(&dev->addr_list_lock);
3625}
3626
f001fde5 3627/*
31278e71 3628 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3629 * rcu_read_lock held.
3630 */
3631#define for_each_dev_addr(dev, ha) \
31278e71 3632 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3633
1da177e4
LT
3634/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3635
f629d208 3636void ether_setup(struct net_device *dev);
1da177e4
LT
3637
3638/* Support for loadable net-drivers */
f629d208 3639struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3640 unsigned char name_assign_type,
f629d208
JP
3641 void (*setup)(struct net_device *),
3642 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3643#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3644 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3645
c835a677
TG
3646#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3647 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3648 count)
36909ea4 3649
f629d208
JP
3650int register_netdev(struct net_device *dev);
3651void unregister_netdev(struct net_device *dev);
f001fde5 3652
22bedad3 3653/* General hardware address lists handling functions */
f629d208
JP
3654int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3655 struct netdev_hw_addr_list *from_list, int addr_len);
3656void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3657 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3658int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3659 struct net_device *dev,
3660 int (*sync)(struct net_device *, const unsigned char *),
3661 int (*unsync)(struct net_device *,
3662 const unsigned char *));
3663void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3664 struct net_device *dev,
3665 int (*unsync)(struct net_device *,
3666 const unsigned char *));
f629d208 3667void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3668
f001fde5 3669/* Functions used for device addresses handling */
f629d208
JP
3670int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3671 unsigned char addr_type);
3672int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3673 unsigned char addr_type);
f629d208
JP
3674void dev_addr_flush(struct net_device *dev);
3675int dev_addr_init(struct net_device *dev);
a748ee24
JP
3676
3677/* Functions used for unicast addresses handling */
f629d208
JP
3678int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3679int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3680int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3681int dev_uc_sync(struct net_device *to, struct net_device *from);
3682int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3683void dev_uc_unsync(struct net_device *to, struct net_device *from);
3684void dev_uc_flush(struct net_device *dev);
3685void dev_uc_init(struct net_device *dev);
f001fde5 3686
670e5b8e
AD
3687/**
3688 * __dev_uc_sync - Synchonize device's unicast list
3689 * @dev: device to sync
3690 * @sync: function to call if address should be added
3691 * @unsync: function to call if address should be removed
3692 *
3693 * Add newly added addresses to the interface, and release
3694 * addresses that have been deleted.
5e82b4b2 3695 */
670e5b8e
AD
3696static inline int __dev_uc_sync(struct net_device *dev,
3697 int (*sync)(struct net_device *,
3698 const unsigned char *),
3699 int (*unsync)(struct net_device *,
3700 const unsigned char *))
3701{
3702 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3703}
3704
3705/**
e793c0f7 3706 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3707 * @dev: device to sync
3708 * @unsync: function to call if address should be removed
3709 *
3710 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3711 */
670e5b8e
AD
3712static inline void __dev_uc_unsync(struct net_device *dev,
3713 int (*unsync)(struct net_device *,
3714 const unsigned char *))
3715{
3716 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3717}
3718
22bedad3 3719/* Functions used for multicast addresses handling */
f629d208
JP
3720int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3721int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3722int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3723int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3724int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3725int dev_mc_sync(struct net_device *to, struct net_device *from);
3726int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3727void dev_mc_unsync(struct net_device *to, struct net_device *from);
3728void dev_mc_flush(struct net_device *dev);
3729void dev_mc_init(struct net_device *dev);
f001fde5 3730
670e5b8e
AD
3731/**
3732 * __dev_mc_sync - Synchonize device's multicast list
3733 * @dev: device to sync
3734 * @sync: function to call if address should be added
3735 * @unsync: function to call if address should be removed
3736 *
3737 * Add newly added addresses to the interface, and release
3738 * addresses that have been deleted.
5e82b4b2 3739 */
670e5b8e
AD
3740static inline int __dev_mc_sync(struct net_device *dev,
3741 int (*sync)(struct net_device *,
3742 const unsigned char *),
3743 int (*unsync)(struct net_device *,
3744 const unsigned char *))
3745{
3746 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3747}
3748
3749/**
e793c0f7 3750 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3751 * @dev: device to sync
3752 * @unsync: function to call if address should be removed
3753 *
3754 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3755 */
670e5b8e
AD
3756static inline void __dev_mc_unsync(struct net_device *dev,
3757 int (*unsync)(struct net_device *,
3758 const unsigned char *))
3759{
3760 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3761}
3762
4417da66 3763/* Functions used for secondary unicast and multicast support */
f629d208
JP
3764void dev_set_rx_mode(struct net_device *dev);
3765void __dev_set_rx_mode(struct net_device *dev);
3766int dev_set_promiscuity(struct net_device *dev, int inc);
3767int dev_set_allmulti(struct net_device *dev, int inc);
3768void netdev_state_change(struct net_device *dev);
3769void netdev_notify_peers(struct net_device *dev);
3770void netdev_features_change(struct net_device *dev);
1da177e4 3771/* Load a device via the kmod */
f629d208
JP
3772void dev_load(struct net *net, const char *name);
3773struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3774 struct rtnl_link_stats64 *storage);
3775void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3776 const struct net_device_stats *netdev_stats);
eeda3fd6 3777
1da177e4 3778extern int netdev_max_backlog;
3b098e2d 3779extern int netdev_tstamp_prequeue;
1da177e4 3780extern int weight_p;
9ff162a8 3781
f629d208 3782bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3783struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3784 struct list_head **iter);
f629d208
JP
3785struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3786 struct list_head **iter);
8b5be856 3787
44a40855
VY
3788/* iterate through upper list, must be called under RCU read lock */
3789#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3790 for (iter = &(dev)->adj_list.upper, \
3791 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3792 updev; \
3793 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3794
1a3f060c
DA
3795int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3796 int (*fn)(struct net_device *upper_dev,
3797 void *data),
3798 void *data);
3799
3800bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3801 struct net_device *upper_dev);
3802
f629d208
JP
3803void *netdev_lower_get_next_private(struct net_device *dev,
3804 struct list_head **iter);
3805void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3806 struct list_head **iter);
31088a11
VF
3807
3808#define netdev_for_each_lower_private(dev, priv, iter) \
3809 for (iter = (dev)->adj_list.lower.next, \
3810 priv = netdev_lower_get_next_private(dev, &(iter)); \
3811 priv; \
3812 priv = netdev_lower_get_next_private(dev, &(iter)))
3813
3814#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3815 for (iter = &(dev)->adj_list.lower, \
3816 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3817 priv; \
3818 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3819
4085ebe8
VY
3820void *netdev_lower_get_next(struct net_device *dev,
3821 struct list_head **iter);
7ce856aa 3822
4085ebe8 3823#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3824 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3825 ldev = netdev_lower_get_next(dev, &(iter)); \
3826 ldev; \
3827 ldev = netdev_lower_get_next(dev, &(iter)))
3828
7ce856aa
JP
3829struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3830 struct list_head **iter);
3831struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3832 struct list_head **iter);
3833
1a3f060c
DA
3834int netdev_walk_all_lower_dev(struct net_device *dev,
3835 int (*fn)(struct net_device *lower_dev,
3836 void *data),
3837 void *data);
3838int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3839 int (*fn)(struct net_device *lower_dev,
3840 void *data),
3841 void *data);
3842
f629d208 3843void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3844void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3845struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3846struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3847int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3848int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3849 struct net_device *upper_dev,
29bf24af 3850 void *upper_priv, void *upper_info);
f629d208
JP
3851void netdev_upper_dev_unlink(struct net_device *dev,
3852 struct net_device *upper_dev);
5bb025fa 3853void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3854void *netdev_lower_dev_get_private(struct net_device *dev,
3855 struct net_device *lower_dev);
04d48266
JP
3856void netdev_lower_state_changed(struct net_device *lower_dev,
3857 void *lower_state_info);
18bfb924
JP
3858int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3859 struct neighbour *n);
3860void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3861 struct neighbour *n);
960fb622
ED
3862
3863/* RSS keys are 40 or 52 bytes long */
3864#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3865extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3866void netdev_rss_key_fill(void *buffer, size_t len);
3867
952fcfd0 3868int dev_get_nest_level(struct net_device *dev);
f629d208
JP
3869int skb_checksum_help(struct sk_buff *skb);
3870struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3871 netdev_features_t features, bool tx_path);
3872struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3873 netdev_features_t features);
12b0004d 3874
61bd3857
MS
3875struct netdev_bonding_info {
3876 ifslave slave;
3877 ifbond master;
3878};
3879
3880struct netdev_notifier_bonding_info {
3881 struct netdev_notifier_info info; /* must be first */
3882 struct netdev_bonding_info bonding_info;
3883};
3884
3885void netdev_bonding_info_change(struct net_device *dev,
3886 struct netdev_bonding_info *bonding_info);
3887
12b0004d
CW
3888static inline
3889struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3890{
3891 return __skb_gso_segment(skb, features, true);
3892}
53d6471c 3893__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3894
3895static inline bool can_checksum_protocol(netdev_features_t features,
3896 __be16 protocol)
3897{
c8cd0989
TH
3898 if (protocol == htons(ETH_P_FCOE))
3899 return !!(features & NETIF_F_FCOE_CRC);
3900
3901 /* Assume this is an IP checksum (not SCTP CRC) */
3902
3903 if (features & NETIF_F_HW_CSUM) {
3904 /* Can checksum everything */
3905 return true;
3906 }
3907
3908 switch (protocol) {
3909 case htons(ETH_P_IP):
3910 return !!(features & NETIF_F_IP_CSUM);
3911 case htons(ETH_P_IPV6):
3912 return !!(features & NETIF_F_IPV6_CSUM);
3913 default:
3914 return false;
3915 }
ec5f0615 3916}
12b0004d 3917
fb286bb2 3918#ifdef CONFIG_BUG
f629d208 3919void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3920#else
3921static inline void netdev_rx_csum_fault(struct net_device *dev)
3922{
3923}
3924#endif
1da177e4 3925/* rx skb timestamps */
f629d208
JP
3926void net_enable_timestamp(void);
3927void net_disable_timestamp(void);
1da177e4 3928
20380731 3929#ifdef CONFIG_PROC_FS
f629d208 3930int __init dev_proc_init(void);
900ff8c6
CW
3931#else
3932#define dev_proc_init() 0
20380731
ACM
3933#endif
3934
4798248e 3935static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3936 struct sk_buff *skb, struct net_device *dev,
3937 bool more)
4798248e 3938{
fa2dbdc2 3939 skb->xmit_more = more ? 1 : 0;
0b725a2c 3940 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3941}
3942
10b3ad8c 3943static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3944 struct netdev_queue *txq, bool more)
4798248e
DM
3945{
3946 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3947 int rc;
4798248e 3948
fa2dbdc2 3949 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3950 if (rc == NETDEV_TX_OK)
3951 txq_trans_update(txq);
3952
3953 return rc;
4798248e
DM
3954}
3955
42a2d923
LT
3956int netdev_class_create_file_ns(struct class_attribute *class_attr,
3957 const void *ns);
3958void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3959 const void *ns);
58292cbe
TH
3960
3961static inline int netdev_class_create_file(struct class_attribute *class_attr)
3962{
3963 return netdev_class_create_file_ns(class_attr, NULL);
3964}
3965
3966static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3967{
3968 netdev_class_remove_file_ns(class_attr, NULL);
3969}
b8a9787e 3970
04600794
JB
3971extern struct kobj_ns_type_operations net_ns_type_operations;
3972
f629d208 3973const char *netdev_drivername(const struct net_device *dev);
6579e57b 3974
f629d208 3975void linkwatch_run_queue(void);
20380731 3976
da08143b
MK
3977static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3978 netdev_features_t f2)
3979{
c8cd0989
TH
3980 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3981 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 3982 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 3983 else
b6a0e72a 3984 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 3985 }
da08143b 3986
c8cd0989 3987 return f1 & f2;
da08143b
MK
3988}
3989
c8f44aff
MM
3990static inline netdev_features_t netdev_get_wanted_features(
3991 struct net_device *dev)
5455c699
MM
3992{
3993 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3994}
c8f44aff
MM
3995netdev_features_t netdev_increment_features(netdev_features_t all,
3996 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3997
3998/* Allow TSO being used on stacked device :
3999 * Performing the GSO segmentation before last device
4000 * is a performance improvement.
4001 */
4002static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4003 netdev_features_t mask)
4004{
4005 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4006}
4007
6cb6a27c 4008int __netdev_update_features(struct net_device *dev);
5455c699 4009void netdev_update_features(struct net_device *dev);
afe12cc8 4010void netdev_change_features(struct net_device *dev);
7f353bf2 4011
fc4a7489
PM
4012void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4013 struct net_device *dev);
4014
e38f3025
TM
4015netdev_features_t passthru_features_check(struct sk_buff *skb,
4016 struct net_device *dev,
4017 netdev_features_t features);
c1e756bf 4018netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4019
4d29515f 4020static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4021{
7b748340 4022 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4023
4024 /* check flags correspondence */
4025 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4026 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4027 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4028 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4029 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4030 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4031 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4032 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4033 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4034 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4035 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4036 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4037 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4038 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4039 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4040 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
0345e186 4041
d6b4991a 4042 return (features & feature) == feature;
576a30eb
HX
4043}
4044
4d29515f 4045static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4046{
278b2513 4047 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4048 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4049}
4050
8b86a61d 4051static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4052 netdev_features_t features)
7967168c 4053{
fc741216 4054 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4055 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4056 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4057}
4058
82cc1a7a
PWJ
4059static inline void netif_set_gso_max_size(struct net_device *dev,
4060 unsigned int size)
4061{
4062 dev->gso_max_size = size;
4063}
4064
7a7ffbab
WCC
4065static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4066 int pulled_hlen, u16 mac_offset,
4067 int mac_len)
4068{
4069 skb->protocol = protocol;
4070 skb->encapsulation = 1;
4071 skb_push(skb, pulled_hlen);
4072 skb_reset_transport_header(skb);
4073 skb->mac_header = mac_offset;
4074 skb->network_header = skb->mac_header + mac_len;
4075 skb->mac_len = mac_len;
4076}
4077
3c175784
SD
4078static inline bool netif_is_macsec(const struct net_device *dev)
4079{
4080 return dev->priv_flags & IFF_MACSEC;
4081}
4082
b618aaa9 4083static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4084{
4085 return dev->priv_flags & IFF_MACVLAN;
4086}
4087
b618aaa9 4088static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4089{
4090 return dev->priv_flags & IFF_MACVLAN_PORT;
4091}
4092
b618aaa9 4093static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4094{
4095 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4096}
4097
b618aaa9 4098static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4099{
4100 return dev->priv_flags & IFF_IPVLAN_MASTER;
4101}
4102
b618aaa9 4103static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4104{
4105 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4106}
4107
b618aaa9 4108static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4109{
4110 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4111}
4112
3bdc0eba
BG
4113static inline bool netif_supports_nofcs(struct net_device *dev)
4114{
4115 return dev->priv_flags & IFF_SUPP_NOFCS;
4116}
4117
007979ea 4118static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4119{
007979ea 4120 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4121}
4122
fee6d4c7
DA
4123static inline bool netif_is_l3_slave(const struct net_device *dev)
4124{
4125 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4126}
4127
0894ae3f
JP
4128static inline bool netif_is_bridge_master(const struct net_device *dev)
4129{
4130 return dev->priv_flags & IFF_EBRIDGE;
4131}
4132
28f9ee22
VY
4133static inline bool netif_is_bridge_port(const struct net_device *dev)
4134{
4135 return dev->priv_flags & IFF_BRIDGE_PORT;
4136}
4137
35d4e172
JP
4138static inline bool netif_is_ovs_master(const struct net_device *dev)
4139{
4140 return dev->priv_flags & IFF_OPENVSWITCH;
4141}
4142
b618aaa9 4143static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4144{
4145 return dev->priv_flags & IFF_TEAM;
4146}
4147
b618aaa9 4148static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4149{
4150 return dev->priv_flags & IFF_TEAM_PORT;
4151}
4152
b618aaa9 4153static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4154{
4155 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4156}
4157
b618aaa9 4158static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4159{
4160 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4161}
4162
d4ab4286
KJ
4163static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4164{
4165 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4166}
4167
02875878
ED
4168/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4169static inline void netif_keep_dst(struct net_device *dev)
4170{
4171 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4172}
4173
18d3df3e
PA
4174/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4175static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4176{
4177 /* TODO: reserve and use an additional IFF bit, if we get more users */
4178 return dev->priv_flags & IFF_MACSEC;
4179}
4180
505d4f73 4181extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4182
571ba423
JP
4183/* Logging, debugging and troubleshooting/diagnostic helpers. */
4184
4185/* netdev_printk helpers, similar to dev_printk */
4186
4187static inline const char *netdev_name(const struct net_device *dev)
4188{
c6f854d5
VF
4189 if (!dev->name[0] || strchr(dev->name, '%'))
4190 return "(unnamed net_device)";
571ba423
JP
4191 return dev->name;
4192}
4193
ccc7f496
VF
4194static inline const char *netdev_reg_state(const struct net_device *dev)
4195{
4196 switch (dev->reg_state) {
4197 case NETREG_UNINITIALIZED: return " (uninitialized)";
4198 case NETREG_REGISTERED: return "";
4199 case NETREG_UNREGISTERING: return " (unregistering)";
4200 case NETREG_UNREGISTERED: return " (unregistered)";
4201 case NETREG_RELEASED: return " (released)";
4202 case NETREG_DUMMY: return " (dummy)";
4203 }
4204
4205 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4206 return " (unknown)";
4207}
4208
f629d208 4209__printf(3, 4)
6ea754eb
JP
4210void netdev_printk(const char *level, const struct net_device *dev,
4211 const char *format, ...);
f629d208 4212__printf(2, 3)
6ea754eb 4213void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4214__printf(2, 3)
6ea754eb 4215void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4216__printf(2, 3)
6ea754eb 4217void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4218__printf(2, 3)
6ea754eb 4219void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4220__printf(2, 3)
6ea754eb 4221void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4222__printf(2, 3)
6ea754eb 4223void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4224__printf(2, 3)
6ea754eb 4225void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4226
8909c9ad
VK
4227#define MODULE_ALIAS_NETDEV(device) \
4228 MODULE_ALIAS("netdev-" device)
4229
b558c96f 4230#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4231#define netdev_dbg(__dev, format, args...) \
4232do { \
ffa10cb4 4233 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4234} while (0)
b558c96f
JC
4235#elif defined(DEBUG)
4236#define netdev_dbg(__dev, format, args...) \
4237 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4238#else
4239#define netdev_dbg(__dev, format, args...) \
4240({ \
4241 if (0) \
4242 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4243})
4244#endif
4245
4246#if defined(VERBOSE_DEBUG)
4247#define netdev_vdbg netdev_dbg
4248#else
4249
4250#define netdev_vdbg(dev, format, args...) \
4251({ \
4252 if (0) \
4253 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4254 0; \
4255})
4256#endif
4257
4258/*
4259 * netdev_WARN() acts like dev_printk(), but with the key difference
4260 * of using a WARN/WARN_ON to get the message out, including the
4261 * file/line information and a backtrace.
4262 */
4263#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4264 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4265 netdev_reg_state(dev), ##args)
571ba423 4266
b3d95c5c
JP
4267/* netif printk helpers, similar to netdev_printk */
4268
4269#define netif_printk(priv, type, level, dev, fmt, args...) \
4270do { \
4271 if (netif_msg_##type(priv)) \
4272 netdev_printk(level, (dev), fmt, ##args); \
4273} while (0)
4274
f45f4321
JP
4275#define netif_level(level, priv, type, dev, fmt, args...) \
4276do { \
4277 if (netif_msg_##type(priv)) \
4278 netdev_##level(dev, fmt, ##args); \
4279} while (0)
4280
b3d95c5c 4281#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4282 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4283#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4284 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4285#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4286 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4287#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4288 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4289#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4290 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4291#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4292 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4293#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4294 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4295
0053ea9c 4296#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4297#define netif_dbg(priv, type, netdev, format, args...) \
4298do { \
4299 if (netif_msg_##type(priv)) \
b5fb0a03 4300 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4301} while (0)
0053ea9c
JP
4302#elif defined(DEBUG)
4303#define netif_dbg(priv, type, dev, format, args...) \
4304 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4305#else
4306#define netif_dbg(priv, type, dev, format, args...) \
4307({ \
4308 if (0) \
4309 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4310 0; \
4311})
4312#endif
4313
4314#if defined(VERBOSE_DEBUG)
bcfcc450 4315#define netif_vdbg netif_dbg
b3d95c5c
JP
4316#else
4317#define netif_vdbg(priv, type, dev, format, args...) \
4318({ \
4319 if (0) \
a4ed89cb 4320 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4321 0; \
4322})
4323#endif
571ba423 4324
900ff8c6
CW
4325/*
4326 * The list of packet types we will receive (as opposed to discard)
4327 * and the routines to invoke.
4328 *
4329 * Why 16. Because with 16 the only overlap we get on a hash of the
4330 * low nibble of the protocol value is RARP/SNAP/X.25.
4331 *
4332 * NOTE: That is no longer true with the addition of VLAN tags. Not
4333 * sure which should go first, but I bet it won't make much
4334 * difference if we are running VLANs. The good news is that
4335 * this protocol won't be in the list unless compiled in, so
4336 * the average user (w/out VLANs) will not be adversely affected.
4337 * --BLG
4338 *
4339 * 0800 IP
4340 * 8100 802.1Q VLAN
4341 * 0001 802.3
4342 * 0002 AX.25
4343 * 0004 802.2
4344 * 8035 RARP
4345 * 0005 SNAP
4346 * 0805 X.25
4347 * 0806 ARP
4348 * 8137 IPX
4349 * 0009 Localtalk
4350 * 86DD IPv6
4351 */
4352#define PTYPE_HASH_SIZE (16)
4353#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4354
385a154c 4355#endif /* _LINUX_NETDEVICE_H */