net/mlx4_en: Disable vlan HW acceleration when in VF vlan protocol 802.1ad mode
[linux-2.6-block.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
cf85d08f 44#include <net/dsa.h>
7a6b6f51 45#ifdef CONFIG_DCB
2f90b865
AD
46#include <net/dcbnl.h>
47#endif
5bc1421e 48#include <net/netprio_cgroup.h>
a050c33f 49
a59e2ecb 50#include <linux/netdev_features.h>
77162022 51#include <linux/neighbour.h>
607ca46e 52#include <uapi/linux/netdevice.h>
61bd3857 53#include <uapi/linux/if_bonding.h>
e4c6734e 54#include <uapi/linux/pkt_cls.h>
59cc1f61 55#include <linux/hashtable.h>
a59e2ecb 56
115c1d6e 57struct netpoll_info;
313162d0 58struct device;
c1f19b51 59struct phy_device;
704232c2
JB
60/* 802.11 specific */
61struct wireless_dev;
98a18b6f
AA
62/* 802.15.4 specific */
63struct wpan_dev;
03c57747 64struct mpls_dev;
7c46a640
AD
65/* UDP Tunnel offloads */
66struct udp_tunnel_info;
a7862b45 67struct bpf_prog;
1da177e4 68
f629d208
JP
69void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
d07d7507 71
9a1654ba
JP
72/* Backlog congestion levels */
73#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74#define NET_RX_DROP 1 /* packet dropped */
75
572a9d7b
PM
76/*
77 * Transmit return codes: transmit return codes originate from three different
78 * namespaces:
79 *
80 * - qdisc return codes
81 * - driver transmit return codes
82 * - errno values
83 *
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
90 * others are propagated to higher layers.
91 */
92
93/* qdisc ->enqueue() return codes. */
94#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
95#define NET_XMIT_DROP 0x01 /* skb dropped */
96#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 97#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 98
b9df3cb8
GR
99/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 102#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
103#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
104
dc1f8bf6 105/* Driver transmit return codes */
9a1654ba 106#define NETDEV_TX_MASK 0xf0
572a9d7b 107
dc1f8bf6 108enum netdev_tx {
572a9d7b 109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
112};
113typedef enum netdev_tx netdev_tx_t;
114
9a1654ba
JP
115/*
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
118 */
119static inline bool dev_xmit_complete(int rc)
120{
121 /*
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
126 */
127 if (likely(rc < NET_XMIT_MASK))
128 return true;
129
130 return false;
131}
132
1da177e4 133/*
5e82b4b2 134 * Compute the worst-case header length according to the protocols
1da177e4
LT
135 * used.
136 */
fe2918b0 137
c0eb4540
KS
138#if defined(CONFIG_HYPERV_NET)
139# define LL_MAX_HEADER 128
140#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
141# if defined(CONFIG_MAC80211_MESH)
142# define LL_MAX_HEADER 128
143# else
144# define LL_MAX_HEADER 96
145# endif
1da177e4 146#else
8388e3da 147# define LL_MAX_HEADER 32
1da177e4
LT
148#endif
149
d11ead75
BH
150#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
152#define MAX_HEADER LL_MAX_HEADER
153#else
154#define MAX_HEADER (LL_MAX_HEADER + 48)
155#endif
156
157/*
be1f3c2c
BH
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
1da177e4 160 */
fe2918b0 161
d94d9fee 162struct net_device_stats {
3cfde79c
BH
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
1da177e4 172 unsigned long collisions;
1da177e4 173 unsigned long rx_length_errors;
3cfde79c
BH
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
1da177e4
LT
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
1da177e4
LT
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
186};
187
1da177e4
LT
188
189#include <linux/cache.h>
190#include <linux/skbuff.h>
191
adc9300e 192#ifdef CONFIG_RPS
c5905afb
IM
193#include <linux/static_key.h>
194extern struct static_key rps_needed;
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
319 spinlock_t poll_lock;
320 int poll_owner;
bea3348e 321#endif
5d38a079 322 struct net_device *dev;
d565b0a1 323 struct sk_buff *gro_list;
5d38a079 324 struct sk_buff *skb;
3b47d303 325 struct hrtimer timer;
404f7c9e 326 struct list_head dev_list;
af12fa6e
ET
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
bea3348e
SH
329};
330
d94d9fee 331enum {
bea3348e 332 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 333 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
bea3348e
SH
337};
338
5b252f0c 339enum gro_result {
d1c76af9
HX
340 GRO_MERGED,
341 GRO_MERGED_FREE,
342 GRO_HELD,
343 GRO_NORMAL,
344 GRO_DROP,
345};
5b252f0c 346typedef enum gro_result gro_result_t;
d1c76af9 347
8a4eb573
JP
348/*
349 * enum rx_handler_result - Possible return values for rx_handlers.
350 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
351 * further.
352 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
353 * case skb->dev was changed by rx_handler.
354 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 355 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
356 *
357 * rx_handlers are functions called from inside __netif_receive_skb(), to do
358 * special processing of the skb, prior to delivery to protocol handlers.
359 *
360 * Currently, a net_device can only have a single rx_handler registered. Trying
361 * to register a second rx_handler will return -EBUSY.
362 *
363 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
364 * To unregister a rx_handler on a net_device, use
365 * netdev_rx_handler_unregister().
366 *
367 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
368 * do with the skb.
369 *
5e82b4b2 370 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 371 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 372 * the skb to be delivered in some other way.
8a4eb573
JP
373 *
374 * If the rx_handler changed skb->dev, to divert the skb to another
375 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
376 * new device will be called if it exists.
377 *
5e82b4b2 378 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 379 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 380 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 381 *
5e82b4b2 382 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
383 * delivered, it should return RX_HANDLER_PASS.
384 *
385 * A device without a registered rx_handler will behave as if rx_handler
386 * returned RX_HANDLER_PASS.
387 */
388
389enum rx_handler_result {
390 RX_HANDLER_CONSUMED,
391 RX_HANDLER_ANOTHER,
392 RX_HANDLER_EXACT,
393 RX_HANDLER_PASS,
394};
395typedef enum rx_handler_result rx_handler_result_t;
396typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 397
f629d208 398void __napi_schedule(struct napi_struct *n);
bc9ad166 399void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 400
4d29515f 401static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
402{
403 return test_bit(NAPI_STATE_DISABLE, &n->state);
404}
405
bea3348e 406/**
5e82b4b2
BH
407 * napi_schedule_prep - check if NAPI can be scheduled
408 * @n: NAPI context
bea3348e
SH
409 *
410 * Test if NAPI routine is already running, and if not mark
5e82b4b2 411 * it as running. This is used as a condition variable to
a0a46196
DM
412 * insure only one NAPI poll instance runs. We also make
413 * sure there is no pending NAPI disable.
bea3348e 414 */
4d29515f 415static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 416{
a0a46196
DM
417 return !napi_disable_pending(n) &&
418 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
419}
420
421/**
422 * napi_schedule - schedule NAPI poll
5e82b4b2 423 * @n: NAPI context
bea3348e
SH
424 *
425 * Schedule NAPI poll routine to be called if it is not already
426 * running.
427 */
428static inline void napi_schedule(struct napi_struct *n)
429{
430 if (napi_schedule_prep(n))
431 __napi_schedule(n);
432}
433
bc9ad166
ED
434/**
435 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 436 * @n: NAPI context
bc9ad166
ED
437 *
438 * Variant of napi_schedule(), assuming hard irqs are masked.
439 */
440static inline void napi_schedule_irqoff(struct napi_struct *n)
441{
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
444}
445
bfe13f54 446/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 447static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
448{
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
4d29515f 451 return true;
bfe13f54 452 }
4d29515f 453 return false;
bfe13f54
RD
454}
455
3b47d303
ED
456void __napi_complete(struct napi_struct *n);
457void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
458/**
459 * napi_complete - NAPI processing complete
5e82b4b2 460 * @n: NAPI context
bea3348e
SH
461 *
462 * Mark NAPI processing as complete.
3b47d303 463 * Consider using napi_complete_done() instead.
bea3348e 464 */
3b47d303
ED
465static inline void napi_complete(struct napi_struct *n)
466{
467 return napi_complete_done(n, 0);
468}
bea3348e 469
af12fa6e
ET
470/**
471 * napi_hash_add - add a NAPI to global hashtable
5e82b4b2 472 * @napi: NAPI context
af12fa6e 473 *
5e82b4b2
BH
474 * Generate a new napi_id and store a @napi under it in napi_hash.
475 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
93d05d4a 476 * Note: This is normally automatically done from netif_napi_add(),
5e82b4b2 477 * so might disappear in a future Linux version.
af12fa6e 478 */
f629d208 479void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
480
481/**
482 * napi_hash_del - remove a NAPI from global table
5e82b4b2 483 * @napi: NAPI context
af12fa6e 484 *
5e82b4b2 485 * Warning: caller must observe RCU grace period
34cbe27e
ED
486 * before freeing memory containing @napi, if
487 * this function returns true.
93d05d4a 488 * Note: core networking stack automatically calls it
5e82b4b2 489 * from netif_napi_del().
93d05d4a 490 * Drivers might want to call this helper to combine all
5e82b4b2 491 * the needed RCU grace periods into a single one.
af12fa6e 492 */
34cbe27e 493bool napi_hash_del(struct napi_struct *napi);
af12fa6e 494
bea3348e
SH
495/**
496 * napi_disable - prevent NAPI from scheduling
5e82b4b2 497 * @n: NAPI context
bea3348e
SH
498 *
499 * Stop NAPI from being scheduled on this context.
500 * Waits till any outstanding processing completes.
501 */
3b47d303 502void napi_disable(struct napi_struct *n);
bea3348e
SH
503
504/**
505 * napi_enable - enable NAPI scheduling
5e82b4b2 506 * @n: NAPI context
bea3348e
SH
507 *
508 * Resume NAPI from being scheduled on this context.
509 * Must be paired with napi_disable.
510 */
511static inline void napi_enable(struct napi_struct *n)
512{
513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 514 smp_mb__before_atomic();
bea3348e 515 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 516 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
517}
518
c264c3de
SH
519/**
520 * napi_synchronize - wait until NAPI is not running
5e82b4b2 521 * @n: NAPI context
c264c3de
SH
522 *
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
526 */
527static inline void napi_synchronize(const struct napi_struct *n)
528{
facc432f
AB
529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
531 msleep(1);
532 else
533 barrier();
c264c3de 534}
c264c3de 535
d94d9fee 536enum netdev_queue_state_t {
73466498
TH
537 __QUEUE_STATE_DRV_XOFF,
538 __QUEUE_STATE_STACK_XOFF,
c3f26a26 539 __QUEUE_STATE_FROZEN,
79d16385 540};
8e2f1a63
DB
541
542#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
543#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
544#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
545
546#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 QUEUE_STATE_FROZEN)
549#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
550 QUEUE_STATE_FROZEN)
551
73466498
TH
552/*
553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
554 * netif_tx_* functions below are used to manipulate this flag. The
555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556 * queue independently. The netif_xmit_*stopped functions below are called
557 * to check if the queue has been stopped by the driver or stack (either
558 * of the XOFF bits are set in the state). Drivers should not need to call
559 * netif_xmit*stopped functions, they should only be using netif_tx_*.
560 */
79d16385 561
bb949fbd 562struct netdev_queue {
6a321cb3 563/*
5e82b4b2 564 * read-mostly part
6a321cb3 565 */
bb949fbd 566 struct net_device *dev;
46e5da40 567 struct Qdisc __rcu *qdisc;
b0e1e646 568 struct Qdisc *qdisc_sleeping;
ccf5ff69 569#ifdef CONFIG_SYSFS
1d24eb48
TH
570 struct kobject kobj;
571#endif
f2cd2d3e
ED
572#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 int numa_node;
574#endif
c0ef079c
FW
575 unsigned long tx_maxrate;
576 /*
577 * Number of TX timeouts for this queue
578 * (/sys/class/net/DEV/Q/trans_timeout)
579 */
580 unsigned long trans_timeout;
6a321cb3 581/*
5e82b4b2 582 * write-mostly part
6a321cb3
ED
583 */
584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
585 int xmit_lock_owner;
9d21493b 586 /*
9b36627a 587 * Time (in jiffies) of last Tx
9d21493b
ED
588 */
589 unsigned long trans_start;
ccf5ff69 590
114cf580
TH
591 unsigned long state;
592
593#ifdef CONFIG_BQL
594 struct dql dql;
595#endif
e8a0464c 596} ____cacheline_aligned_in_smp;
bb949fbd 597
f2cd2d3e
ED
598static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599{
600#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 return q->numa_node;
602#else
b236da69 603 return NUMA_NO_NODE;
f2cd2d3e
ED
604#endif
605}
606
607static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608{
609#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 q->numa_node = node;
611#endif
612}
613
df334545 614#ifdef CONFIG_RPS
0a9627f2
TH
615/*
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
618 */
619struct rps_map {
620 unsigned int len;
621 struct rcu_head rcu;
622 u16 cpus[0];
623};
60b778ce 624#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 625
fec5e652 626/*
c445477d
BH
627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628 * tail pointer for that CPU's input queue at the time of last enqueue, and
629 * a hardware filter index.
fec5e652
TH
630 */
631struct rps_dev_flow {
632 u16 cpu;
c445477d 633 u16 filter;
fec5e652
TH
634 unsigned int last_qtail;
635};
c445477d 636#define RPS_NO_FILTER 0xffff
fec5e652
TH
637
638/*
639 * The rps_dev_flow_table structure contains a table of flow mappings.
640 */
641struct rps_dev_flow_table {
642 unsigned int mask;
643 struct rcu_head rcu;
fec5e652
TH
644 struct rps_dev_flow flows[0];
645};
646#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 647 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
648
649/*
650 * The rps_sock_flow_table contains mappings of flows to the last CPU
651 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
652 * Each entry is a 32bit value. Upper part is the high-order bits
653 * of flow hash, lower part is CPU number.
567e4b79 654 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 657 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
658 */
659struct rps_sock_flow_table {
567e4b79 660 u32 mask;
93c1af6c
ED
661
662 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 663};
567e4b79 664#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
665
666#define RPS_NO_CPU 0xffff
667
567e4b79
ED
668extern u32 rps_cpu_mask;
669extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670
fec5e652
TH
671static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
672 u32 hash)
673{
674 if (table && hash) {
567e4b79
ED
675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
fec5e652 677
5e82b4b2 678 /* We only give a hint, preemption can change CPU under us */
567e4b79 679 val |= raw_smp_processor_id();
fec5e652 680
567e4b79
ED
681 if (table->ents[index] != val)
682 table->ents[index] = val;
fec5e652
TH
683 }
684}
685
c445477d 686#ifdef CONFIG_RFS_ACCEL
f629d208
JP
687bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
688 u16 filter_id);
c445477d 689#endif
a953be53 690#endif /* CONFIG_RPS */
c445477d 691
0a9627f2
TH
692/* This structure contains an instance of an RX queue. */
693struct netdev_rx_queue {
a953be53 694#ifdef CONFIG_RPS
6e3f7faf
ED
695 struct rps_map __rcu *rps_map;
696 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 697#endif
6e3f7faf 698 struct kobject kobj;
fe822240 699 struct net_device *dev;
0a9627f2 700} ____cacheline_aligned_in_smp;
a953be53
MD
701
702/*
703 * RX queue sysfs structures and functions.
704 */
705struct rx_queue_attribute {
706 struct attribute attr;
707 ssize_t (*show)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, char *buf);
709 ssize_t (*store)(struct netdev_rx_queue *queue,
710 struct rx_queue_attribute *attr, const char *buf, size_t len);
711};
d314774c 712
bf264145
TH
713#ifdef CONFIG_XPS
714/*
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
717 */
718struct xps_map {
719 unsigned int len;
720 unsigned int alloc_len;
721 struct rcu_head rcu;
722 u16 queues[0];
723};
60b778ce 724#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
725#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
727
728/*
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 */
731struct xps_dev_maps {
732 struct rcu_head rcu;
a4177869 733 struct xps_map __rcu *cpu_map[0];
bf264145
TH
734};
735#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
736 (nr_cpu_ids * sizeof(struct xps_map *)))
737#endif /* CONFIG_XPS */
738
4f57c087
JF
739#define TC_MAX_QUEUE 16
740#define TC_BITMASK 15
741/* HW offloaded queuing disciplines txq count and offset maps */
742struct netdev_tc_txq {
743 u16 count;
744 u16 offset;
745};
746
68bad94e
NP
747#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748/*
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
751 */
752struct netdev_fcoe_hbainfo {
753 char manufacturer[64];
754 char serial_number[64];
755 char hardware_version[64];
756 char driver_version[64];
757 char optionrom_version[64];
758 char firmware_version[64];
759 char model[256];
760 char model_description[256];
761};
762#endif
763
02637fce 764#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 765
02637fce
JP
766/* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
66b52b0d 768 */
02637fce
JP
769struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
771 unsigned char id_len;
772};
773
d754f98b
SF
774static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
776{
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
779}
780
99932d4f
DB
781typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
783
a1b7c5fd 784/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
785 * that are being passed to the netdevice through the setup_tc op.
786 */
787enum {
788 TC_SETUP_MQPRIO,
a1b7c5fd 789 TC_SETUP_CLSU32,
5b33f488 790 TC_SETUP_CLSFLOWER,
b87f7936 791 TC_SETUP_MATCHALL,
332ae8e2 792 TC_SETUP_CLSBPF,
16e5cc64
JF
793};
794
a1b7c5fd
JF
795struct tc_cls_u32_offload;
796
16e5cc64
JF
797struct tc_to_netdev {
798 unsigned int type;
799 union {
800 u8 tc;
a1b7c5fd 801 struct tc_cls_u32_offload *cls_u32;
5b33f488 802 struct tc_cls_flower_offload *cls_flower;
b87f7936 803 struct tc_cls_matchall_offload *cls_mall;
332ae8e2 804 struct tc_cls_bpf_offload *cls_bpf;
16e5cc64
JF
805 };
806};
807
a7862b45
BB
808/* These structures hold the attributes of xdp state that are being passed
809 * to the netdevice through the xdp op.
810 */
811enum xdp_netdev_command {
812 /* Set or clear a bpf program used in the earliest stages of packet
813 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
814 * is responsible for calling bpf_prog_put on any old progs that are
815 * stored. In case of error, the callee need not release the new prog
816 * reference, but on success it takes ownership and must bpf_prog_put
817 * when it is no longer used.
818 */
819 XDP_SETUP_PROG,
820 /* Check if a bpf program is set on the device. The callee should
821 * return true if a program is currently attached and running.
822 */
823 XDP_QUERY_PROG,
824};
825
826struct netdev_xdp {
827 enum xdp_netdev_command command;
828 union {
829 /* XDP_SETUP_PROG */
830 struct bpf_prog *prog;
831 /* XDP_QUERY_PROG */
832 bool prog_attached;
833 };
834};
16e5cc64 835
d314774c
SH
836/*
837 * This structure defines the management hooks for network devices.
00829823
SH
838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
d314774c
SH
840 *
841 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
d314774c
SH
846 *
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
850 *
851 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 852 * This function is called when a network device transitions to the up
d314774c
SH
853 * state.
854 *
855 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 856 * This function is called when a network device transitions to the down
d314774c
SH
857 * state.
858 *
dc1f8bf6
SH
859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
00829823 861 * Called when a packet needs to be transmitted.
e79d8429
RR
862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 866 * Required; cannot be NULL.
00829823 867 *
cdba756f
ED
868 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
869 * netdev_features_t features);
870 * Adjusts the requested feature flags according to device-specific
871 * constraints, and returns the resulting flags. Must not modify
872 * the device state.
873 *
f663dd9a 874 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 875 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 876 * Called to decide which queue to use when device supports multiple
00829823
SH
877 * transmit queues.
878 *
d314774c
SH
879 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
880 * This function is called to allow device receiver to make
5e82b4b2 881 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
882 *
883 * void (*ndo_set_rx_mode)(struct net_device *dev);
884 * This function is called device changes address list filtering.
01789349 885 * If driver handles unicast address filtering, it should set
5e82b4b2 886 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
887 *
888 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
889 * This function is called when the Media Access Control address
37b607c5 890 * needs to be changed. If this interface is not defined, the
5e82b4b2 891 * MAC address can not be changed.
d314774c
SH
892 *
893 * int (*ndo_validate_addr)(struct net_device *dev);
894 * Test if Media Access Control address is valid for the device.
895 *
896 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
897 * Called when a user requests an ioctl which can't be handled by
898 * the generic interface code. If not defined ioctls return
d314774c
SH
899 * not supported error code.
900 *
901 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
902 * Used to set network devices bus interface parameters. This interface
5e82b4b2 903 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
904 * interface (PCI) for low level management.
905 *
906 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
907 * Called when a user wants to change the Maximum Transfer Unit
908 * of a device. If not defined, any request to change MTU will
909 * will return an error.
910 *
00829823 911 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 912 * Callback used when the transmitter has not made any progress
d314774c
SH
913 * for dev->watchdog ticks.
914 *
3cfde79c 915 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 916 * struct rtnl_link_stats64 *storage);
d308e38f 917 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 918 * Called when a user wants to get the network device usage
be1f3c2c 919 * statistics. Drivers must do one of the following:
3cfde79c
BH
920 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
921 * rtnl_link_stats64 structure passed by the caller.
82695d9b 922 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
923 * (which should normally be dev->stats) and return a pointer to
924 * it. The structure may be changed asynchronously only if each
925 * field is written atomically.
926 * 3. Update dev->stats asynchronously and atomically, and define
927 * neither operation.
d314774c 928 *
2c9d85d4
NF
929 * bool (*ndo_has_offload_stats)(int attr_id)
930 * Return true if this device supports offload stats of this attr_id.
931 *
932 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
933 * void *attr_data)
934 * Get statistics for offload operations by attr_id. Write it into the
935 * attr_data pointer.
936 *
5d632cb7 937 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 938 * If device supports VLAN filtering this function is called when a
80d5c368 939 * VLAN id is registered.
d314774c 940 *
5d632cb7 941 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 942 * If device supports VLAN filtering this function is called when a
80d5c368 943 * VLAN id is unregistered.
d314774c
SH
944 *
945 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
946 *
947 * SR-IOV management functions.
948 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
949 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
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);
2c9d85d4
NF
1168 bool (*ndo_has_offload_stats)(int attr_id);
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,
1190 int queue, u16 vlan, u8 qos);
ed616689
SC
1191 int (*ndo_set_vf_rate)(struct net_device *dev,
1192 int vf, int min_tx_rate,
1193 int max_tx_rate);
5f8444a3
GR
1194 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1195 int vf, bool setting);
dd461d6a
HS
1196 int (*ndo_set_vf_trust)(struct net_device *dev,
1197 int vf, bool setting);
95c26df8
WM
1198 int (*ndo_get_vf_config)(struct net_device *dev,
1199 int vf,
1200 struct ifla_vf_info *ivf);
1d8faf48
RE
1201 int (*ndo_set_vf_link_state)(struct net_device *dev,
1202 int vf, int link_state);
3b766cd8
EBE
1203 int (*ndo_get_vf_stats)(struct net_device *dev,
1204 int vf,
1205 struct ifla_vf_stats
1206 *vf_stats);
57b61080
SF
1207 int (*ndo_set_vf_port)(struct net_device *dev,
1208 int vf,
1209 struct nlattr *port[]);
1210 int (*ndo_get_vf_port)(struct net_device *dev,
1211 int vf, struct sk_buff *skb);
cc8e27cc
EC
1212 int (*ndo_set_vf_guid)(struct net_device *dev,
1213 int vf, u64 guid,
1214 int guid_type);
01a3d796
VZ
1215 int (*ndo_set_vf_rss_query_en)(
1216 struct net_device *dev,
1217 int vf, bool setting);
16e5cc64
JF
1218 int (*ndo_setup_tc)(struct net_device *dev,
1219 u32 handle,
1220 __be16 protocol,
1221 struct tc_to_netdev *tc);
d11ead75 1222#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1223 int (*ndo_fcoe_enable)(struct net_device *dev);
1224 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1225 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1226 u16 xid,
1227 struct scatterlist *sgl,
1228 unsigned int sgc);
1229 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1230 u16 xid);
6247e086
YZ
1231 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1232 u16 xid,
1233 struct scatterlist *sgl,
1234 unsigned int sgc);
68bad94e
NP
1235 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1236 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1237#endif
1238
d11ead75 1239#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1240#define NETDEV_FCOE_WWNN 0
1241#define NETDEV_FCOE_WWPN 1
1242 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1243 u64 *wwn, int type);
4d288d57 1244#endif
3c9c36bc 1245
c445477d
BH
1246#ifdef CONFIG_RFS_ACCEL
1247 int (*ndo_rx_flow_steer)(struct net_device *dev,
1248 const struct sk_buff *skb,
1249 u16 rxq_index,
1250 u32 flow_id);
1251#endif
fbaec0ea
JP
1252 int (*ndo_add_slave)(struct net_device *dev,
1253 struct net_device *slave_dev);
1254 int (*ndo_del_slave)(struct net_device *dev,
1255 struct net_device *slave_dev);
c8f44aff
MM
1256 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1257 netdev_features_t features);
5455c699 1258 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1259 netdev_features_t features);
503eebc2
JP
1260 int (*ndo_neigh_construct)(struct net_device *dev,
1261 struct neighbour *n);
1262 void (*ndo_neigh_destroy)(struct net_device *dev,
1263 struct neighbour *n);
77162022
JF
1264
1265 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1266 struct nlattr *tb[],
77162022 1267 struct net_device *dev,
6b6e2725 1268 const unsigned char *addr,
f6f6424b 1269 u16 vid,
77162022
JF
1270 u16 flags);
1271 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1272 struct nlattr *tb[],
77162022 1273 struct net_device *dev,
f6f6424b
JP
1274 const unsigned char *addr,
1275 u16 vid);
77162022
JF
1276 int (*ndo_fdb_dump)(struct sk_buff *skb,
1277 struct netlink_callback *cb,
1278 struct net_device *dev,
5d5eacb3 1279 struct net_device *filter_dev,
d297653d 1280 int *idx);
e5a55a89
JF
1281
1282 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1283 struct nlmsghdr *nlh,
1284 u16 flags);
e5a55a89
JF
1285 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1286 u32 pid, u32 seq,
6cbdceeb 1287 struct net_device *dev,
46c264da
ND
1288 u32 filter_mask,
1289 int nlflags);
407af329 1290 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1291 struct nlmsghdr *nlh,
1292 u16 flags);
4bf84c35
JP
1293 int (*ndo_change_carrier)(struct net_device *dev,
1294 bool new_carrier);
66b52b0d 1295 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1296 struct netdev_phys_item_id *ppid);
db24a904
DA
1297 int (*ndo_get_phys_port_name)(struct net_device *dev,
1298 char *name, size_t len);
7c46a640
AD
1299 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1300 struct udp_tunnel_info *ti);
1301 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1302 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1303 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1304 struct net_device *dev);
1305 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1306 void *priv);
1307
1308 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1309 struct net_device *dev,
1310 void *priv);
25175ba5 1311 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1312 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1313 int queue_index,
1314 u32 maxrate);
a54acb3a 1315 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1316 int (*ndo_change_proto_down)(struct net_device *dev,
1317 bool proto_down);
fc4099f1
PS
1318 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1319 struct sk_buff *skb);
871b642a
PA
1320 void (*ndo_set_rx_headroom)(struct net_device *dev,
1321 int needed_headroom);
a7862b45
BB
1322 int (*ndo_xdp)(struct net_device *dev,
1323 struct netdev_xdp *xdp);
d314774c
SH
1324};
1325
7aa98047
LR
1326/**
1327 * enum net_device_priv_flags - &struct net_device priv_flags
1328 *
1329 * These are the &struct net_device, they are only set internally
1330 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1331 * userspace; this means that the order of these flags can change
7aa98047
LR
1332 * during any kernel release.
1333 *
1334 * You should have a pretty good reason to be extending these flags.
1335 *
1336 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1337 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1338 * @IFF_BONDING: bonding master or slave
7aa98047 1339 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1340 * @IFF_WAN_HDLC: WAN HDLC device
1341 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1342 * release skb->dst
1343 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1344 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1345 * @IFF_MACVLAN_PORT: device used as macvlan port
1346 * @IFF_BRIDGE_PORT: device used as bridge port
1347 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1348 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1349 * @IFF_UNICAST_FLT: Supports unicast filtering
1350 * @IFF_TEAM_PORT: device used as team port
1351 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1352 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1353 * change when it's running
1354 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1355 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1356 * underlying stacked devices
1357 * @IFF_IPVLAN_MASTER: IPvlan master device
1358 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1359 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1360 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1361 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1362 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1363 * @IFF_TEAM: device is a team device
d4ab4286 1364 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1365 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1366 * entity (i.e. the master device for bridged veth)
3c175784 1367 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1368 */
1369enum netdev_priv_flags {
1370 IFF_802_1Q_VLAN = 1<<0,
1371 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1372 IFF_BONDING = 1<<2,
1373 IFF_ISATAP = 1<<3,
1374 IFF_WAN_HDLC = 1<<4,
1375 IFF_XMIT_DST_RELEASE = 1<<5,
1376 IFF_DONT_BRIDGE = 1<<6,
1377 IFF_DISABLE_NETPOLL = 1<<7,
1378 IFF_MACVLAN_PORT = 1<<8,
1379 IFF_BRIDGE_PORT = 1<<9,
1380 IFF_OVS_DATAPATH = 1<<10,
1381 IFF_TX_SKB_SHARING = 1<<11,
1382 IFF_UNICAST_FLT = 1<<12,
1383 IFF_TEAM_PORT = 1<<13,
1384 IFF_SUPP_NOFCS = 1<<14,
1385 IFF_LIVE_ADDR_CHANGE = 1<<15,
1386 IFF_MACVLAN = 1<<16,
1387 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1388 IFF_IPVLAN_MASTER = 1<<18,
1389 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1390 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1391 IFF_NO_QUEUE = 1<<21,
1392 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1393 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1394 IFF_TEAM = 1<<24,
d4ab4286 1395 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1396 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1397 IFF_MACSEC = 1<<27,
7aa98047
LR
1398};
1399
1400#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1401#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1402#define IFF_BONDING IFF_BONDING
7aa98047 1403#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1404#define IFF_WAN_HDLC IFF_WAN_HDLC
1405#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1406#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1407#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1408#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1409#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1410#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1411#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1412#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1413#define IFF_TEAM_PORT IFF_TEAM_PORT
1414#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1415#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1416#define IFF_MACVLAN IFF_MACVLAN
02875878 1417#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1418#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1419#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1420#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1421#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1422#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1423#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1424#define IFF_TEAM IFF_TEAM
d4ab4286 1425#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1426#define IFF_MACSEC IFF_MACSEC
7aa98047 1427
536721b1
KK
1428/**
1429 * struct net_device - The DEVICE structure.
1430 * Actually, this whole structure is a big mistake. It mixes I/O
1431 * data with strictly "high-level" data, and it has to know about
1432 * almost every data structure used in the INET module.
1433 *
1434 * @name: This is the first field of the "visible" part of this structure
1435 * (i.e. as seen by users in the "Space.c" file). It is the name
1436 * of the interface.
1437 *
1438 * @name_hlist: Device name hash chain, please keep it close to name[]
1439 * @ifalias: SNMP alias
1440 * @mem_end: Shared memory end
1441 * @mem_start: Shared memory start
1442 * @base_addr: Device I/O address
1443 * @irq: Device IRQ number
1444 *
14ffbbb8
TG
1445 * @carrier_changes: Stats to monitor carrier on<->off transitions
1446 *
536721b1
KK
1447 * @state: Generic network queuing layer state, see netdev_state_t
1448 * @dev_list: The global list of network devices
5e82b4b2
BH
1449 * @napi_list: List entry used for polling NAPI devices
1450 * @unreg_list: List entry when we are unregistering the
1451 * device; see the function unregister_netdev
1452 * @close_list: List entry used when we are closing the device
62d885fe
BP
1453 * @ptype_all: Device-specific packet handlers for all protocols
1454 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1455 *
1456 * @adj_list: Directly linked devices, like slaves for bonding
1457 * @all_adj_list: All linked devices, *including* neighbours
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
1507 * @type: Interface hardware type
2793a23a 1508 * @hard_header_len: Maximum hardware header length.
536721b1
KK
1509 *
1510 * @needed_headroom: Extra headroom the hardware may need, but not in all
1511 * cases can this be guaranteed
1512 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1513 * cases can this be guaranteed. Some cases also use
1514 * LL_MAX_HEADER instead to allocate the skb
1515 *
1516 * interface address info:
1517 *
1518 * @perm_addr: Permanent hw address
1519 * @addr_assign_type: Hw address assignment type
1520 * @addr_len: Hardware address length
8626a0c8 1521 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1522 * @dev_id: Used to differentiate devices that share
1523 * the same link layer address
1524 * @dev_port: Used to differentiate devices that share
1525 * the same function
1526 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1527 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1528 * has been enabled due to the need to listen to
1529 * additional unicast addresses in a device that
1530 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1531 * @uc: unicast mac addresses
1532 * @mc: multicast mac addresses
1533 * @dev_addrs: list of device hw addresses
1534 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1535 * @promiscuity: Number of times the NIC is told to work in
1536 * promiscuous mode; if it becomes 0 the NIC will
1537 * exit promiscuous mode
536721b1
KK
1538 * @allmulti: Counter, enables or disables allmulticast mode
1539 *
1540 * @vlan_info: VLAN info
1541 * @dsa_ptr: dsa specific data
1542 * @tipc_ptr: TIPC specific data
1543 * @atalk_ptr: AppleTalk link
1544 * @ip_ptr: IPv4 specific data
1545 * @dn_ptr: DECnet specific data
1546 * @ip6_ptr: IPv6 specific data
1547 * @ax25_ptr: AX.25 specific data
1548 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1549 *
1550 * @last_rx: Time of last Rx
1551 * @dev_addr: Hw address (before bcast,
1552 * because most packets are unicast)
1553 *
1554 * @_rx: Array of RX queues
1555 * @num_rx_queues: Number of RX queues
1556 * allocated at register_netdev() time
1557 * @real_num_rx_queues: Number of RX queues currently active in device
1558 *
1559 * @rx_handler: handler for received packets
1560 * @rx_handler_data: XXX: need comments on this one
1561 * @ingress_queue: XXX: need comments on this one
1562 * @broadcast: hw bcast address
1563 *
14ffbbb8
TG
1564 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1565 * indexed by RX queue number. Assigned by driver.
1566 * This must only be set if the ndo_rx_flow_steer
1567 * operation is defined
1568 * @index_hlist: Device index hash chain
1569 *
536721b1
KK
1570 * @_tx: Array of TX queues
1571 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1572 * @real_num_tx_queues: Number of TX queues currently active in device
1573 * @qdisc: Root qdisc from userspace point of view
1574 * @tx_queue_len: Max frames per queue allowed
1575 * @tx_global_lock: XXX: need comments on this one
1576 *
1577 * @xps_maps: XXX: need comments on this one
1578 *
536721b1 1579 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1580 * the watchdog (see dev_watchdog())
536721b1
KK
1581 * @watchdog_timer: List of timers
1582 *
1583 * @pcpu_refcnt: Number of references to this device
1584 * @todo_list: Delayed register/unregister
536721b1
KK
1585 * @link_watch_list: XXX: need comments on this one
1586 *
1587 * @reg_state: Register/unregister state machine
1588 * @dismantle: Device is going to be freed
1589 * @rtnl_link_state: This enum represents the phases of creating
1590 * a new link
1591 *
1592 * @destructor: Called from unregister,
1593 * can be used to call free_netdev
1594 * @npinfo: XXX: need comments on this one
1595 * @nd_net: Network namespace this network device is inside
1596 *
1597 * @ml_priv: Mid-layer private
1598 * @lstats: Loopback statistics
1599 * @tstats: Tunnel statistics
1600 * @dstats: Dummy statistics
1601 * @vstats: Virtual ethernet statistics
1602 *
1603 * @garp_port: GARP
1604 * @mrp_port: MRP
1605 *
1606 * @dev: Class/net/name entry
1607 * @sysfs_groups: Space for optional device, statistics and wireless
1608 * sysfs groups
1609 *
1610 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1611 * @rtnl_link_ops: Rtnl_link_ops
1612 *
1613 * @gso_max_size: Maximum size of generic segmentation offload
1614 * @gso_max_segs: Maximum number of segments that can be passed to the
1615 * NIC for GSO
1616 *
1617 * @dcbnl_ops: Data Center Bridging netlink ops
1618 * @num_tc: Number of traffic classes in the net device
1619 * @tc_to_txq: XXX: need comments on this one
1620 * @prio_tc_map XXX: need comments on this one
1621 *
1622 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1623 *
1624 * @priomap: XXX: need comments on this one
1625 * @phydev: Physical device may attach itself
1626 * for hardware timestamping
1627 *
123b3652
ED
1628 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1629 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1630 *
d746d707
AK
1631 * @proto_down: protocol port state information can be sent to the
1632 * switch driver and used to set the phys state of the
1633 * switch port.
1634 *
1da177e4
LT
1635 * FIXME: cleanup struct net_device such that network protocol info
1636 * moves out.
1637 */
1638
d94d9fee 1639struct net_device {
1da177e4 1640 char name[IFNAMSIZ];
9356b8fc 1641 struct hlist_node name_hlist;
0b815a1a 1642 char *ifalias;
1da177e4
LT
1643 /*
1644 * I/O specific fields
1645 * FIXME: Merge these and struct ifmap into one
1646 */
536721b1
KK
1647 unsigned long mem_end;
1648 unsigned long mem_start;
1649 unsigned long base_addr;
1650 int irq;
1da177e4 1651
14ffbbb8
TG
1652 atomic_t carrier_changes;
1653
1da177e4 1654 /*
536721b1
KK
1655 * Some hardware also needs these fields (state,dev_list,
1656 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1657 * part of the usual set specified in Space.c.
1658 */
1659
1da177e4
LT
1660 unsigned long state;
1661
7562f876 1662 struct list_head dev_list;
bea3348e 1663 struct list_head napi_list;
44a0873d 1664 struct list_head unreg_list;
5cde2829 1665 struct list_head close_list;
7866a621
SN
1666 struct list_head ptype_all;
1667 struct list_head ptype_specific;
2f268f12 1668
2f268f12
VF
1669 struct {
1670 struct list_head upper;
1671 struct list_head lower;
1672 } adj_list;
1673
2f268f12
VF
1674 struct {
1675 struct list_head upper;
1676 struct list_head lower;
1677 } all_adj_list;
4c3d5e7b 1678
c8f44aff 1679 netdev_features_t features;
c8f44aff 1680 netdev_features_t hw_features;
c8f44aff 1681 netdev_features_t wanted_features;
c8f44aff 1682 netdev_features_t vlan_features;
6a674e9c 1683 netdev_features_t hw_enc_features;
0d89d203 1684 netdev_features_t mpls_features;
802ab55a 1685 netdev_features_t gso_partial_features;
04ed3e74 1686
1da177e4 1687 int ifindex;
7a66bbc9 1688 int group;
1da177e4 1689
c45d286e 1690 struct net_device_stats stats;
015f0688 1691
015f0688
ED
1692 atomic_long_t rx_dropped;
1693 atomic_long_t tx_dropped;
6e7333d3 1694 atomic_long_t rx_nohandler;
1da177e4 1695
b86e0280 1696#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1697 const struct iw_handler_def *wireless_handlers;
1698 struct iw_public_data *wireless_data;
b86e0280 1699#endif
d314774c 1700 const struct net_device_ops *netdev_ops;
76fd8593 1701 const struct ethtool_ops *ethtool_ops;
4170604f 1702#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1703 const struct switchdev_ops *switchdev_ops;
4170604f 1704#endif
1b69c6d0
DA
1705#ifdef CONFIG_NET_L3_MASTER_DEV
1706 const struct l3mdev_ops *l3mdev_ops;
1707#endif
f997c55c
AA
1708#if IS_ENABLED(CONFIG_IPV6)
1709 const struct ndisc_ops *ndisc_ops;
1710#endif
1da177e4 1711
3b04ddde
SH
1712 const struct header_ops *header_ops;
1713
536721b1
KK
1714 unsigned int flags;
1715 unsigned int priv_flags;
1716
1da177e4 1717 unsigned short gflags;
536721b1 1718 unsigned short padded;
1da177e4 1719
536721b1
KK
1720 unsigned char operstate;
1721 unsigned char link_mode;
b00055aa 1722
536721b1
KK
1723 unsigned char if_port;
1724 unsigned char dma;
bdc220da 1725
536721b1
KK
1726 unsigned int mtu;
1727 unsigned short type;
1728 unsigned short hard_header_len;
1da177e4 1729
f5184d26
JB
1730 unsigned short needed_headroom;
1731 unsigned short needed_tailroom;
1732
1da177e4 1733 /* Interface address info. */
536721b1
KK
1734 unsigned char perm_addr[MAX_ADDR_LEN];
1735 unsigned char addr_assign_type;
1736 unsigned char addr_len;
a0a9663d 1737 unsigned short neigh_priv_len;
536721b1
KK
1738 unsigned short dev_id;
1739 unsigned short dev_port;
ccffad25 1740 spinlock_t addr_list_lock;
14ffbbb8
TG
1741 unsigned char name_assign_type;
1742 bool uc_promisc;
536721b1
KK
1743 struct netdev_hw_addr_list uc;
1744 struct netdev_hw_addr_list mc;
1745 struct netdev_hw_addr_list dev_addrs;
1746
4c3d5e7b
ED
1747#ifdef CONFIG_SYSFS
1748 struct kset *queues_kset;
1749#endif
9d45abe1
WC
1750 unsigned int promiscuity;
1751 unsigned int allmulti;
1da177e4 1752
1da177e4 1753
5e82b4b2 1754 /* Protocol-specific pointers */
65ac6a5f 1755
d11ead75 1756#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1757 struct vlan_info __rcu *vlan_info;
65ac6a5f 1758#endif
34a430d7 1759#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1760 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1761#endif
1762#if IS_ENABLED(CONFIG_TIPC)
536721b1 1763 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1764#endif
536721b1
KK
1765 void *atalk_ptr;
1766 struct in_device __rcu *ip_ptr;
1767 struct dn_dev __rcu *dn_ptr;
1768 struct inet6_dev __rcu *ip6_ptr;
1769 void *ax25_ptr;
1770 struct wireless_dev *ieee80211_ptr;
98a18b6f 1771 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1772#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1773 struct mpls_dev __rcu *mpls_ptr;
1774#endif
1da177e4 1775
9356b8fc 1776/*
cd13539b 1777 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1778 */
536721b1 1779 unsigned long last_rx;
4dc89133 1780
9356b8fc 1781 /* Interface address info used in eth_type_trans() */
536721b1 1782 unsigned char *dev_addr;
f001fde5 1783
a953be53 1784#ifdef CONFIG_SYSFS
0a9627f2
TH
1785 struct netdev_rx_queue *_rx;
1786
0a9627f2 1787 unsigned int num_rx_queues;
62fe0b40 1788 unsigned int real_num_rx_queues;
df334545 1789#endif
0a9627f2 1790
3b47d303 1791 unsigned long gro_flush_timeout;
61391cde 1792 rx_handler_func_t __rcu *rx_handler;
1793 void __rcu *rx_handler_data;
e8a0464c 1794
4cda01e8 1795#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1796 struct tcf_proto __rcu *ingress_cl_list;
1797#endif
24824a09 1798 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1799#ifdef CONFIG_NETFILTER_INGRESS
1800 struct list_head nf_hooks_ingress;
1801#endif
d2788d34 1802
536721b1 1803 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1804#ifdef CONFIG_RFS_ACCEL
1805 struct cpu_rmap *rx_cpu_rmap;
1806#endif
1807 struct hlist_node index_hlist;
cd13539b
ED
1808
1809/*
1810 * Cache lines mostly used on transmit path
1811 */
e8a0464c
DM
1812 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1813 unsigned int num_tx_queues;
fd2ea0a7 1814 unsigned int real_num_tx_queues;
af356afa 1815 struct Qdisc *qdisc;
59cc1f61
JK
1816#ifdef CONFIG_NET_SCHED
1817 DECLARE_HASHTABLE (qdisc_hash, 4);
1818#endif
536721b1 1819 unsigned long tx_queue_len;
c3f26a26 1820 spinlock_t tx_global_lock;
14ffbbb8 1821 int watchdog_timeo;
cd13539b 1822
bf264145 1823#ifdef CONFIG_XPS
a4177869 1824 struct xps_dev_maps __rcu *xps_maps;
bf264145 1825#endif
1f211a1b
DB
1826#ifdef CONFIG_NET_CLS_ACT
1827 struct tcf_proto __rcu *egress_cl_list;
1828#endif
0c4f691f 1829
9356b8fc 1830 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1831 struct timer_list watchdog_timer;
1832
29b4433d 1833 int __percpu *pcpu_refcnt;
1da177e4 1834 struct list_head todo_list;
1da177e4 1835
e014debe 1836 struct list_head link_watch_list;
572a103d 1837
1da177e4 1838 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1839 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1840 NETREG_UNREGISTERING, /* called unregister_netdevice */
1841 NETREG_UNREGISTERED, /* completed unregister todo */
1842 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1843 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1844 } reg_state:8;
1845
536721b1 1846 bool dismantle;
a2835763
PM
1847
1848 enum {
1849 RTNL_LINK_INITIALIZED,
1850 RTNL_LINK_INITIALIZING,
1851 } rtnl_link_state:16;
1da177e4 1852
d314774c 1853 void (*destructor)(struct net_device *dev);
1da177e4 1854
1da177e4 1855#ifdef CONFIG_NETPOLL
5fbee843 1856 struct netpoll_info __rcu *npinfo;
1da177e4 1857#endif
eae792b7 1858
0c5c9fb5 1859 possible_net_t nd_net;
4a1c5371 1860
4951704b 1861 /* mid-layer private */
a7855c78 1862 union {
536721b1
KK
1863 void *ml_priv;
1864 struct pcpu_lstats __percpu *lstats;
8f84985f 1865 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1866 struct pcpu_dstats __percpu *dstats;
1867 struct pcpu_vstats __percpu *vstats;
a7855c78 1868 };
536721b1 1869
3cc77ec7 1870 struct garp_port __rcu *garp_port;
febf018d 1871 struct mrp_port __rcu *mrp_port;
1da177e4 1872
5e82b4b2 1873 struct device dev;
0c509a6c 1874 const struct attribute_group *sysfs_groups[4];
a953be53 1875 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1876
38f7b870 1877 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1878
82cc1a7a
PWJ
1879 /* for setting kernel sock attribute on TCP connection setup */
1880#define GSO_MAX_SIZE 65536
1881 unsigned int gso_max_size;
30b678d8
BH
1882#define GSO_MAX_SEGS 65535
1883 u16 gso_max_segs;
743b03a8 1884
7a6b6f51 1885#ifdef CONFIG_DCB
32953543 1886 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1887#endif
5e82b4b2
BH
1888 u8 num_tc;
1889 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1890 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1891
d11ead75 1892#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1893 unsigned int fcoe_ddp_xid;
5bc1421e 1894#endif
86f8515f 1895#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1896 struct netprio_map __rcu *priomap;
4d288d57 1897#endif
5e82b4b2
BH
1898 struct phy_device *phydev;
1899 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1900 struct lock_class_key *qdisc_running_key;
5e82b4b2 1901 bool proto_down;
1da177e4 1902};
43cb76d9 1903#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1904
1905#define NETDEV_ALIGN 32
1da177e4 1906
4f57c087
JF
1907static inline
1908int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1909{
1910 return dev->prio_tc_map[prio & TC_BITMASK];
1911}
1912
1913static inline
1914int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1915{
1916 if (tc >= dev->num_tc)
1917 return -EINVAL;
1918
1919 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1920 return 0;
1921}
1922
1923static inline
1924void netdev_reset_tc(struct net_device *dev)
1925{
1926 dev->num_tc = 0;
1927 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1928 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1929}
1930
1931static inline
1932int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1933{
1934 if (tc >= dev->num_tc)
1935 return -EINVAL;
1936
1937 dev->tc_to_txq[tc].count = count;
1938 dev->tc_to_txq[tc].offset = offset;
1939 return 0;
1940}
1941
1942static inline
1943int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1944{
1945 if (num_tc > TC_MAX_QUEUE)
1946 return -EINVAL;
1947
1948 dev->num_tc = num_tc;
1949 return 0;
1950}
1951
1952static inline
1953int netdev_get_num_tc(struct net_device *dev)
1954{
1955 return dev->num_tc;
1956}
1957
e8a0464c
DM
1958static inline
1959struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1960 unsigned int index)
1961{
1962 return &dev->_tx[index];
1963}
1964
10c51b56
DB
1965static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1966 const struct sk_buff *skb)
1967{
1968 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1969}
1970
e8a0464c
DM
1971static inline void netdev_for_each_tx_queue(struct net_device *dev,
1972 void (*f)(struct net_device *,
1973 struct netdev_queue *,
1974 void *),
1975 void *arg)
1976{
1977 unsigned int i;
1978
1979 for (i = 0; i < dev->num_tx_queues; i++)
1980 f(dev, &dev->_tx[i], arg);
1981}
1982
d3fff6c4
ED
1983#define netdev_lockdep_set_classes(dev) \
1984{ \
1985 static struct lock_class_key qdisc_tx_busylock_key; \
1986 static struct lock_class_key qdisc_running_key; \
1987 static struct lock_class_key qdisc_xmit_lock_key; \
1988 static struct lock_class_key dev_addr_list_lock_key; \
1989 unsigned int i; \
1990 \
1991 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1992 (dev)->qdisc_running_key = &qdisc_running_key; \
1993 lockdep_set_class(&(dev)->addr_list_lock, \
1994 &dev_addr_list_lock_key); \
1995 for (i = 0; i < (dev)->num_tx_queues; i++) \
1996 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1997 &qdisc_xmit_lock_key); \
1998}
1999
f629d208 2000struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
2001 struct sk_buff *skb,
2002 void *accel_priv);
8c4c49df 2003
871b642a
PA
2004/* returns the headroom that the master device needs to take in account
2005 * when forwarding to this dev
2006 */
2007static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2008{
2009 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2010}
2011
2012static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2013{
2014 if (dev->netdev_ops->ndo_set_rx_headroom)
2015 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2016}
2017
2018/* set the device rx headroom to the dev's default */
2019static inline void netdev_reset_rx_headroom(struct net_device *dev)
2020{
2021 netdev_set_rx_headroom(dev, -1);
2022}
2023
c346dca1
YH
2024/*
2025 * Net namespace inlines
2026 */
2027static inline
2028struct net *dev_net(const struct net_device *dev)
2029{
c2d9ba9b 2030 return read_pnet(&dev->nd_net);
c346dca1
YH
2031}
2032
2033static inline
f5aa23fd 2034void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2035{
0c5c9fb5 2036 write_pnet(&dev->nd_net, net);
c346dca1
YH
2037}
2038
3e8a72d1 2039static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 2040{
3fc88677 2041#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
2042 if (dev->dsa_ptr != NULL)
2043 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 2044#endif
5aed85ce 2045 return false;
396138f0
LB
2046}
2047
bea3348e
SH
2048/**
2049 * netdev_priv - access network device private data
2050 * @dev: network device
2051 *
2052 * Get network device private data
2053 */
6472ce60 2054static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2055{
1ce8e7b5 2056 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2057}
2058
1da177e4
LT
2059/* Set the sysfs physical device reference for the network logical device
2060 * if set prior to registration will cause a symlink during initialization.
2061 */
43cb76d9 2062#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2063
384912ed 2064/* Set the sysfs device type for the network logical device to allow
3f79410c 2065 * fine-grained identification of different network device types. For
5e82b4b2 2066 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2067 */
2068#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2069
82dc3c63
ED
2070/* Default NAPI poll() weight
2071 * Device drivers are strongly advised to not use bigger value
2072 */
2073#define NAPI_POLL_WEIGHT 64
2074
3b582cc1 2075/**
5e82b4b2 2076 * netif_napi_add - initialize a NAPI context
3b582cc1 2077 * @dev: network device
5e82b4b2 2078 * @napi: NAPI context
3b582cc1
SH
2079 * @poll: polling function
2080 * @weight: default weight
2081 *
5e82b4b2
BH
2082 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2083 * *any* of the other NAPI-related functions.
3b582cc1 2084 */
d565b0a1
HX
2085void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2086 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2087
d64b5e85 2088/**
5e82b4b2 2089 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2090 * @dev: network device
5e82b4b2 2091 * @napi: NAPI context
d64b5e85
ED
2092 * @poll: polling function
2093 * @weight: default weight
2094 *
2095 * This variant of netif_napi_add() should be used from drivers using NAPI
2096 * to exclusively poll a TX queue.
2097 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2098 */
2099static inline void netif_tx_napi_add(struct net_device *dev,
2100 struct napi_struct *napi,
2101 int (*poll)(struct napi_struct *, int),
2102 int weight)
2103{
2104 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2105 netif_napi_add(dev, napi, poll, weight);
2106}
2107
d8156534 2108/**
5e82b4b2
BH
2109 * netif_napi_del - remove a NAPI context
2110 * @napi: NAPI context
d8156534 2111 *
5e82b4b2 2112 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2113 */
d565b0a1
HX
2114void netif_napi_del(struct napi_struct *napi);
2115
2116struct napi_gro_cb {
78a478d0 2117 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2118 void *frag0;
78a478d0 2119
7489594c
HX
2120 /* Length of frag0. */
2121 unsigned int frag0_len;
2122
86911732 2123 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2124 int data_offset;
86911732 2125
d565b0a1 2126 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2127 u16 flush;
2128
2129 /* Save the IP ID here and check when we get to the transport layer */
2130 u16 flush_id;
d565b0a1
HX
2131
2132 /* Number of segments aggregated. */
2e71a6f8
ED
2133 u16 count;
2134
15e2396d
TH
2135 /* Start offset for remote checksum offload */
2136 u16 gro_remcsum_start;
2137
2e71a6f8
ED
2138 /* jiffies when first packet was created/queued */
2139 unsigned long age;
86347245 2140
afe93325 2141 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2142 u16 proto;
2143
baa32ff4
TH
2144 /* This is non-zero if the packet may be of the same flow. */
2145 u8 same_flow:1;
2146
fac8e0f5
JG
2147 /* Used in tunnel GRO receive */
2148 u8 encap_mark:1;
573e8fca
TH
2149
2150 /* GRO checksum is valid */
2151 u8 csum_valid:1;
2152
662880f4
TH
2153 /* Number of checksums via CHECKSUM_UNNECESSARY */
2154 u8 csum_cnt:3;
c3c7c254 2155
baa32ff4
TH
2156 /* Free the skb? */
2157 u8 free:2;
2158#define NAPI_GRO_FREE 1
2159#define NAPI_GRO_FREE_STOLEN_HEAD 2
2160
efc98d08
TH
2161 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2162 u8 is_ipv6:1;
2163
a0ca153f
AD
2164 /* Used in GRE, set in fou/gue_gro_receive */
2165 u8 is_fou:1;
2166
1530545e
AD
2167 /* Used to determine if flush_id can be ignored */
2168 u8 is_atomic:1;
2169
2170 /* 5 bit hole */
baa32ff4 2171
bf5a755f
JC
2172 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2173 __wsum csum;
2174
c3c7c254
ED
2175 /* used in skb_gro_receive() slow path */
2176 struct sk_buff *last;
d565b0a1
HX
2177};
2178
2179#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2180
1da177e4 2181struct packet_type {
f2ccd8fa
DM
2182 __be16 type; /* This is really htons(ether_type). */
2183 struct net_device *dev; /* NULL is wildcarded here */
2184 int (*func) (struct sk_buff *,
2185 struct net_device *,
2186 struct packet_type *,
2187 struct net_device *);
c0de08d0
EL
2188 bool (*id_match)(struct packet_type *ptype,
2189 struct sock *sk);
1da177e4
LT
2190 void *af_packet_priv;
2191 struct list_head list;
2192};
2193
f191a1d1 2194struct offload_callbacks {
576a30eb 2195 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2196 netdev_features_t features);
d565b0a1 2197 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2198 struct sk_buff *skb);
299603e8 2199 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2200};
2201
2202struct packet_offload {
2203 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2204 u16 priority;
f191a1d1
VY
2205 struct offload_callbacks callbacks;
2206 struct list_head list;
1da177e4
LT
2207};
2208
5e82b4b2 2209/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2210struct pcpu_sw_netstats {
2211 u64 rx_packets;
2212 u64 rx_bytes;
2213 u64 tx_packets;
2214 u64 tx_bytes;
2215 struct u64_stats_sync syncp;
2216};
2217
aabc92bb
PNA
2218#define __netdev_alloc_pcpu_stats(type, gfp) \
2219({ \
2220 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2221 if (pcpu_stats) { \
2222 int __cpu; \
2223 for_each_possible_cpu(__cpu) { \
2224 typeof(type) *stat; \
2225 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2226 u64_stats_init(&stat->syncp); \
2227 } \
2228 } \
2229 pcpu_stats; \
1c213bd2
WC
2230})
2231
aabc92bb 2232#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2233 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2234
764f5e54
JP
2235enum netdev_lag_tx_type {
2236 NETDEV_LAG_TX_TYPE_UNKNOWN,
2237 NETDEV_LAG_TX_TYPE_RANDOM,
2238 NETDEV_LAG_TX_TYPE_BROADCAST,
2239 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2240 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2241 NETDEV_LAG_TX_TYPE_HASH,
2242};
2243
2244struct netdev_lag_upper_info {
2245 enum netdev_lag_tx_type tx_type;
2246};
2247
fb1b2e3c
JP
2248struct netdev_lag_lower_state_info {
2249 u8 link_up : 1,
2250 tx_enabled : 1;
2251};
2252
1da177e4
LT
2253#include <linux/notifier.h>
2254
dcfe1421
AW
2255/* netdevice notifier chain. Please remember to update the rtnetlink
2256 * notification exclusion list in rtnetlink_event() when adding new
2257 * types.
2258 */
2259#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2260#define NETDEV_DOWN 0x0002
2261#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2262 detected a hardware crash and restarted
2263 - we can use this eg to kick tcp sessions
2264 once done */
2265#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2266#define NETDEV_REGISTER 0x0005
2267#define NETDEV_UNREGISTER 0x0006
1d486bfb 2268#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2269#define NETDEV_CHANGEADDR 0x0008
2270#define NETDEV_GOING_DOWN 0x0009
2271#define NETDEV_CHANGENAME 0x000A
2272#define NETDEV_FEAT_CHANGE 0x000B
2273#define NETDEV_BONDING_FAILOVER 0x000C
2274#define NETDEV_PRE_UP 0x000D
2275#define NETDEV_PRE_TYPE_CHANGE 0x000E
2276#define NETDEV_POST_TYPE_CHANGE 0x000F
2277#define NETDEV_POST_INIT 0x0010
0115e8e3 2278#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2279#define NETDEV_RELEASE 0x0012
2280#define NETDEV_NOTIFY_PEERS 0x0013
2281#define NETDEV_JOIN 0x0014
42e52bf9 2282#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2283#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2284#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2285#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2286#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2287#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2288#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2289#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2290#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2291
f629d208
JP
2292int register_netdevice_notifier(struct notifier_block *nb);
2293int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2294
2295struct netdev_notifier_info {
2296 struct net_device *dev;
2297};
2298
be9efd36
JP
2299struct netdev_notifier_change_info {
2300 struct netdev_notifier_info info; /* must be first */
2301 unsigned int flags_changed;
2302};
2303
0e4ead9d
JP
2304struct netdev_notifier_changeupper_info {
2305 struct netdev_notifier_info info; /* must be first */
2306 struct net_device *upper_dev; /* new upper dev */
2307 bool master; /* is upper dev master */
5e82b4b2 2308 bool linking; /* is the notification for link or unlink */
29bf24af 2309 void *upper_info; /* upper dev info */
0e4ead9d
JP
2310};
2311
04d48266
JP
2312struct netdev_notifier_changelowerstate_info {
2313 struct netdev_notifier_info info; /* must be first */
2314 void *lower_state_info; /* is lower dev state */
2315};
2316
75538c2b
CW
2317static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2318 struct net_device *dev)
2319{
2320 info->dev = dev;
2321}
2322
351638e7
JP
2323static inline struct net_device *
2324netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2325{
2326 return info->dev;
2327}
2328
f629d208 2329int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2330
2331
1da177e4
LT
2332extern rwlock_t dev_base_lock; /* Device list lock */
2333
881d966b
EB
2334#define for_each_netdev(net, d) \
2335 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2336#define for_each_netdev_reverse(net, d) \
2337 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2338#define for_each_netdev_rcu(net, d) \
2339 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2340#define for_each_netdev_safe(net, d, n) \
2341 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2342#define for_each_netdev_continue(net, d) \
2343 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2344#define for_each_netdev_continue_rcu(net, d) \
2345 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2346#define for_each_netdev_in_bond_rcu(bond, slave) \
2347 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2348 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2349#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2350
a050c33f
DL
2351static inline struct net_device *next_net_device(struct net_device *dev)
2352{
2353 struct list_head *lh;
2354 struct net *net;
2355
c346dca1 2356 net = dev_net(dev);
a050c33f
DL
2357 lh = dev->dev_list.next;
2358 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2359}
2360
ce81b76a
ED
2361static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2362{
2363 struct list_head *lh;
2364 struct net *net;
2365
2366 net = dev_net(dev);
ccf43438 2367 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2368 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2369}
2370
a050c33f
DL
2371static inline struct net_device *first_net_device(struct net *net)
2372{
2373 return list_empty(&net->dev_base_head) ? NULL :
2374 net_device_entry(net->dev_base_head.next);
2375}
7562f876 2376
ccf43438
ED
2377static inline struct net_device *first_net_device_rcu(struct net *net)
2378{
2379 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2380
2381 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2382}
2383
f629d208
JP
2384int netdev_boot_setup_check(struct net_device *dev);
2385unsigned long netdev_boot_base(const char *prefix, int unit);
2386struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2387 const char *hwaddr);
2388struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2389struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2390void dev_add_pack(struct packet_type *pt);
2391void dev_remove_pack(struct packet_type *pt);
2392void __dev_remove_pack(struct packet_type *pt);
2393void dev_add_offload(struct packet_offload *po);
2394void dev_remove_offload(struct packet_offload *po);
f629d208 2395
a54acb3a 2396int dev_get_iflink(const struct net_device *dev);
fc4099f1 2397int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2398struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2399 unsigned short mask);
f629d208
JP
2400struct net_device *dev_get_by_name(struct net *net, const char *name);
2401struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2402struct net_device *__dev_get_by_name(struct net *net, const char *name);
2403int dev_alloc_name(struct net_device *dev, const char *name);
2404int dev_open(struct net_device *dev);
2405int dev_close(struct net_device *dev);
99c4a26a 2406int dev_close_many(struct list_head *head, bool unlink);
f629d208 2407void dev_disable_lro(struct net_device *dev);
0c4b51f0 2408int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2409int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2410int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2411int register_netdevice(struct net_device *dev);
2412void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2413void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2414static inline void unregister_netdevice(struct net_device *dev)
2415{
2416 unregister_netdevice_queue(dev, NULL);
2417}
2418
f629d208
JP
2419int netdev_refcnt_read(const struct net_device *dev);
2420void free_netdev(struct net_device *dev);
74d332c1 2421void netdev_freemem(struct net_device *dev);
f629d208
JP
2422void synchronize_net(void);
2423int init_dummy_netdev(struct net_device *dev);
937f1ba5 2424
f60e5990 2425DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2426#define XMIT_RECURSION_LIMIT 10
2427
f60e5990 2428static inline int dev_recursion_level(void)
2429{
2430 return this_cpu_read(xmit_recursion);
2431}
2432
f629d208
JP
2433struct net_device *dev_get_by_index(struct net *net, int ifindex);
2434struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2435struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2436int netdev_get_name(struct net *net, char *name, int ifindex);
2437int dev_restart(struct net_device *dev);
f629d208 2438int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2439
2440static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2441{
2442 return NAPI_GRO_CB(skb)->data_offset;
2443}
2444
2445static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2446{
2447 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2448}
2449
2450static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2451{
2452 NAPI_GRO_CB(skb)->data_offset += len;
2453}
2454
a5b1cf28
HX
2455static inline void *skb_gro_header_fast(struct sk_buff *skb,
2456 unsigned int offset)
86911732 2457{
a5b1cf28
HX
2458 return NAPI_GRO_CB(skb)->frag0 + offset;
2459}
78a478d0 2460
a5b1cf28
HX
2461static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2462{
2463 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2464}
78a478d0 2465
a5b1cf28
HX
2466static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2467 unsigned int offset)
2468{
17dd759c
HX
2469 if (!pskb_may_pull(skb, hlen))
2470 return NULL;
2471
a5b1cf28
HX
2472 NAPI_GRO_CB(skb)->frag0 = NULL;
2473 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2474 return skb->data + offset;
86911732 2475}
1da177e4 2476
36e7b1b8
HX
2477static inline void *skb_gro_network_header(struct sk_buff *skb)
2478{
78d3fd0b
HX
2479 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2480 skb_network_offset(skb);
36e7b1b8
HX
2481}
2482
bf5a755f
JC
2483static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2484 const void *start, unsigned int len)
2485{
573e8fca 2486 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2487 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2488 csum_partial(start, len, 0));
2489}
2490
573e8fca
TH
2491/* GRO checksum functions. These are logical equivalents of the normal
2492 * checksum functions (in skbuff.h) except that they operate on the GRO
2493 * offsets and fields in sk_buff.
2494 */
2495
2496__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2497
15e2396d
TH
2498static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2499{
b7fe10e5 2500 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2501}
2502
573e8fca
TH
2503static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2504 bool zero_okay,
2505 __sum16 check)
2506{
6edec0e6
TH
2507 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2508 skb_checksum_start_offset(skb) <
2509 skb_gro_offset(skb)) &&
15e2396d 2510 !skb_at_gro_remcsum_start(skb) &&
662880f4 2511 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2512 (!zero_okay || check));
2513}
2514
2515static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2516 __wsum psum)
2517{
2518 if (NAPI_GRO_CB(skb)->csum_valid &&
2519 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2520 return 0;
2521
2522 NAPI_GRO_CB(skb)->csum = psum;
2523
2524 return __skb_gro_checksum_complete(skb);
2525}
2526
573e8fca
TH
2527static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2528{
662880f4
TH
2529 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2530 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2531 NAPI_GRO_CB(skb)->csum_cnt--;
2532 } else {
2533 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2534 * verified a new top level checksum or an encapsulated one
2535 * during GRO. This saves work if we fallback to normal path.
2536 */
2537 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2538 }
2539}
2540
2541#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2542 compute_pseudo) \
2543({ \
2544 __sum16 __ret = 0; \
2545 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2546 __ret = __skb_gro_checksum_validate_complete(skb, \
2547 compute_pseudo(skb, proto)); \
5a212329
TH
2548 if (__ret) \
2549 __skb_mark_checksum_bad(skb); \
2550 else \
573e8fca
TH
2551 skb_gro_incr_csum_unnecessary(skb); \
2552 __ret; \
2553})
2554
2555#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2556 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2557
2558#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2559 compute_pseudo) \
2560 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2561
2562#define skb_gro_checksum_simple_validate(skb) \
2563 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2564
d96535a1
TH
2565static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2566{
2567 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2568 !NAPI_GRO_CB(skb)->csum_valid);
2569}
2570
2571static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2572 __sum16 check, __wsum pseudo)
2573{
2574 NAPI_GRO_CB(skb)->csum = ~pseudo;
2575 NAPI_GRO_CB(skb)->csum_valid = 1;
2576}
2577
2578#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2579do { \
2580 if (__skb_gro_checksum_convert_check(skb)) \
2581 __skb_gro_checksum_convert(skb, check, \
2582 compute_pseudo(skb, proto)); \
2583} while (0)
2584
26c4f7da
TH
2585struct gro_remcsum {
2586 int offset;
2587 __wsum delta;
2588};
2589
2590static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2591{
846cd667 2592 grc->offset = 0;
26c4f7da
TH
2593 grc->delta = 0;
2594}
2595
b7fe10e5
TH
2596static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2597 unsigned int off, size_t hdrlen,
2598 int start, int offset,
2599 struct gro_remcsum *grc,
2600 bool nopartial)
dcdc8994
TH
2601{
2602 __wsum delta;
b7fe10e5 2603 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2604
2605 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2606
15e2396d 2607 if (!nopartial) {
b7fe10e5
TH
2608 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2609 return ptr;
2610 }
2611
2612 ptr = skb_gro_header_fast(skb, off);
2613 if (skb_gro_header_hard(skb, off + plen)) {
2614 ptr = skb_gro_header_slow(skb, off + plen, off);
2615 if (!ptr)
2616 return NULL;
15e2396d
TH
2617 }
2618
b7fe10e5
TH
2619 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2620 start, offset);
dcdc8994
TH
2621
2622 /* Adjust skb->csum since we changed the packet */
dcdc8994 2623 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2624
b7fe10e5 2625 grc->offset = off + hdrlen + offset;
26c4f7da 2626 grc->delta = delta;
b7fe10e5
TH
2627
2628 return ptr;
dcdc8994
TH
2629}
2630
26c4f7da
TH
2631static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2632 struct gro_remcsum *grc)
2633{
b7fe10e5
TH
2634 void *ptr;
2635 size_t plen = grc->offset + sizeof(u16);
2636
26c4f7da
TH
2637 if (!grc->delta)
2638 return;
2639
b7fe10e5
TH
2640 ptr = skb_gro_header_fast(skb, grc->offset);
2641 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2642 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2643 if (!ptr)
2644 return;
2645 }
2646
2647 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2648}
dcdc8994 2649
6ae23ad3
TH
2650struct skb_csum_offl_spec {
2651 __u16 ipv4_okay:1,
2652 ipv6_okay:1,
2653 encap_okay:1,
2654 ip_options_okay:1,
2655 ext_hdrs_okay:1,
2656 tcp_okay:1,
2657 udp_okay:1,
2658 sctp_okay:1,
2659 vlan_okay:1,
2660 no_encapped_ipv6:1,
2661 no_not_encapped:1;
2662};
2663
2664bool __skb_csum_offload_chk(struct sk_buff *skb,
2665 const struct skb_csum_offl_spec *spec,
2666 bool *csum_encapped,
2667 bool csum_help);
2668
2669static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2670 const struct skb_csum_offl_spec *spec,
2671 bool *csum_encapped,
2672 bool csum_help)
2673{
2674 if (skb->ip_summed != CHECKSUM_PARTIAL)
2675 return false;
2676
2677 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2678}
2679
2680static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2681 const struct skb_csum_offl_spec *spec)
2682{
2683 bool csum_encapped;
2684
2685 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2686}
2687
2688static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2689{
2690 static const struct skb_csum_offl_spec csum_offl_spec = {
2691 .ipv4_okay = 1,
2692 .ip_options_okay = 1,
2693 .ipv6_okay = 1,
2694 .vlan_okay = 1,
2695 .tcp_okay = 1,
2696 .udp_okay = 1,
2697 };
2698
2699 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2700}
2701
2702static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2703{
2704 static const struct skb_csum_offl_spec csum_offl_spec = {
2705 .ipv4_okay = 1,
2706 .ip_options_okay = 1,
2707 .tcp_okay = 1,
2708 .udp_okay = 1,
2709 .vlan_okay = 1,
2710 };
2711
2712 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2713}
2714
0c4e8581
SH
2715static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2716 unsigned short type,
3b04ddde 2717 const void *daddr, const void *saddr,
95c96174 2718 unsigned int len)
0c4e8581 2719{
f1ecfd5d 2720 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2721 return 0;
3b04ddde
SH
2722
2723 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2724}
2725
b95cce35
SH
2726static inline int dev_parse_header(const struct sk_buff *skb,
2727 unsigned char *haddr)
2728{
2729 const struct net_device *dev = skb->dev;
2730
1b83336b 2731 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2732 return 0;
3b04ddde 2733 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2734}
2735
2793a23a
WB
2736/* ll_header must have at least hard_header_len allocated */
2737static inline bool dev_validate_header(const struct net_device *dev,
2738 char *ll_header, int len)
2739{
2740 if (likely(len >= dev->hard_header_len))
2741 return true;
2742
2743 if (capable(CAP_SYS_RAWIO)) {
2744 memset(ll_header + len, 0, dev->hard_header_len - len);
2745 return true;
2746 }
2747
2748 if (dev->header_ops && dev->header_ops->validate)
2749 return dev->header_ops->validate(ll_header, len);
2750
2751 return false;
2752}
2753
1da177e4 2754typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2755int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2756static inline int unregister_gifconf(unsigned int family)
2757{
2758 return register_gifconf(family, NULL);
2759}
2760
99bbc707 2761#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2762#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2763struct sd_flow_limit {
2764 u64 count;
2765 unsigned int num_buckets;
2766 unsigned int history_head;
2767 u16 history[FLOW_LIMIT_HISTORY];
2768 u8 buckets[];
2769};
2770
2771extern int netdev_flow_limit_table_len;
2772#endif /* CONFIG_NET_FLOW_LIMIT */
2773
1da177e4 2774/*
5e82b4b2 2775 * Incoming packets are placed on per-CPU queues
1da177e4 2776 */
d94d9fee 2777struct softnet_data {
1da177e4 2778 struct list_head poll_list;
6e7676c1 2779 struct sk_buff_head process_queue;
1da177e4 2780
dee42870 2781 /* stats */
cd7b5396
DM
2782 unsigned int processed;
2783 unsigned int time_squeeze;
cd7b5396 2784 unsigned int received_rps;
fd793d89 2785#ifdef CONFIG_RPS
88751275 2786 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2787#endif
2788#ifdef CONFIG_NET_FLOW_LIMIT
2789 struct sd_flow_limit __rcu *flow_limit;
2790#endif
2791 struct Qdisc *output_queue;
2792 struct Qdisc **output_queue_tailp;
2793 struct sk_buff *completion_queue;
88751275 2794
4cdb1e2e 2795#ifdef CONFIG_RPS
501e7ef5
ED
2796 /* input_queue_head should be written by cpu owning this struct,
2797 * and only read by other cpus. Worth using a cache line.
2798 */
2799 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2800
2801 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2802 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2803 struct softnet_data *rps_ipi_next;
2804 unsigned int cpu;
76cc8b13 2805 unsigned int input_queue_tail;
1e94d72f 2806#endif
95c96174 2807 unsigned int dropped;
0a9627f2 2808 struct sk_buff_head input_pkt_queue;
bea3348e 2809 struct napi_struct backlog;
99bbc707 2810
1da177e4
LT
2811};
2812
76cc8b13 2813static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2814{
2815#ifdef CONFIG_RPS
76cc8b13
TH
2816 sd->input_queue_head++;
2817#endif
2818}
2819
2820static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2821 unsigned int *qtail)
2822{
2823#ifdef CONFIG_RPS
2824 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2825#endif
2826}
2827
0a9627f2 2828DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2829
f629d208 2830void __netif_schedule(struct Qdisc *q);
46e5da40 2831void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2832
fd2ea0a7
DM
2833static inline void netif_tx_schedule_all(struct net_device *dev)
2834{
2835 unsigned int i;
2836
2837 for (i = 0; i < dev->num_tx_queues; i++)
2838 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2839}
2840
f9a7cbbf 2841static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2842{
73466498 2843 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2844}
2845
bea3348e
SH
2846/**
2847 * netif_start_queue - allow transmit
2848 * @dev: network device
2849 *
2850 * Allow upper layers to call the device hard_start_xmit routine.
2851 */
1da177e4
LT
2852static inline void netif_start_queue(struct net_device *dev)
2853{
e8a0464c 2854 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2855}
2856
fd2ea0a7
DM
2857static inline void netif_tx_start_all_queues(struct net_device *dev)
2858{
2859 unsigned int i;
2860
2861 for (i = 0; i < dev->num_tx_queues; i++) {
2862 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2863 netif_tx_start_queue(txq);
2864 }
2865}
2866
46e5da40 2867void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2868
d29f749e
DJ
2869/**
2870 * netif_wake_queue - restart transmit
2871 * @dev: network device
2872 *
2873 * Allow upper layers to call the device hard_start_xmit routine.
2874 * Used for flow control when transmit resources are available.
2875 */
79d16385
DM
2876static inline void netif_wake_queue(struct net_device *dev)
2877{
e8a0464c 2878 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2879}
2880
fd2ea0a7
DM
2881static inline void netif_tx_wake_all_queues(struct net_device *dev)
2882{
2883 unsigned int i;
2884
2885 for (i = 0; i < dev->num_tx_queues; i++) {
2886 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2887 netif_tx_wake_queue(txq);
2888 }
2889}
2890
f9a7cbbf 2891static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2892{
73466498 2893 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2894}
2895
bea3348e
SH
2896/**
2897 * netif_stop_queue - stop transmitted packets
2898 * @dev: network device
2899 *
2900 * Stop upper layers calling the device hard_start_xmit routine.
2901 * Used for flow control when transmit resources are unavailable.
2902 */
1da177e4
LT
2903static inline void netif_stop_queue(struct net_device *dev)
2904{
e8a0464c 2905 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2906}
2907
a2029240 2908void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2909
4d29515f 2910static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2911{
73466498 2912 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2913}
2914
bea3348e
SH
2915/**
2916 * netif_queue_stopped - test if transmit queue is flowblocked
2917 * @dev: network device
2918 *
2919 * Test if transmit queue on device is currently unable to send.
2920 */
4d29515f 2921static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2922{
e8a0464c 2923 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2924}
2925
4d29515f 2926static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2927{
73466498
TH
2928 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2929}
2930
8e2f1a63
DB
2931static inline bool
2932netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2933{
2934 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2935}
2936
8e2f1a63
DB
2937static inline bool
2938netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2939{
2940 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2941}
2942
53511453
ED
2943/**
2944 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2945 * @dev_queue: pointer to transmit queue
2946 *
2947 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2948 * to give appropriate hint to the CPU.
53511453
ED
2949 */
2950static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2951{
2952#ifdef CONFIG_BQL
2953 prefetchw(&dev_queue->dql.num_queued);
2954#endif
2955}
2956
2957/**
2958 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2959 * @dev_queue: pointer to transmit queue
2960 *
2961 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2962 * to give appropriate hint to the CPU.
53511453
ED
2963 */
2964static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2965{
2966#ifdef CONFIG_BQL
2967 prefetchw(&dev_queue->dql.limit);
2968#endif
2969}
2970
c5d67bd7
TH
2971static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2972 unsigned int bytes)
2973{
114cf580
TH
2974#ifdef CONFIG_BQL
2975 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2976
2977 if (likely(dql_avail(&dev_queue->dql) >= 0))
2978 return;
2979
2980 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2981
2982 /*
2983 * The XOFF flag must be set before checking the dql_avail below,
2984 * because in netdev_tx_completed_queue we update the dql_completed
2985 * before checking the XOFF flag.
2986 */
2987 smp_mb();
2988
2989 /* check again in case another CPU has just made room avail */
2990 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2991 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2992#endif
c5d67bd7
TH
2993}
2994
0042d0c8
FF
2995/**
2996 * netdev_sent_queue - report the number of bytes queued to hardware
2997 * @dev: network device
2998 * @bytes: number of bytes queued to the hardware device queue
2999 *
3000 * Report the number of bytes queued for sending/completion to the network
3001 * device hardware queue. @bytes should be a good approximation and should
3002 * exactly match netdev_completed_queue() @bytes
3003 */
c5d67bd7
TH
3004static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3005{
3006 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3007}
3008
3009static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 3010 unsigned int pkts, unsigned int bytes)
c5d67bd7 3011{
114cf580 3012#ifdef CONFIG_BQL
b37c0fbe
AD
3013 if (unlikely(!bytes))
3014 return;
3015
3016 dql_completed(&dev_queue->dql, bytes);
3017
3018 /*
3019 * Without the memory barrier there is a small possiblity that
3020 * netdev_tx_sent_queue will miss the update and cause the queue to
3021 * be stopped forever
3022 */
3023 smp_mb();
3024
3025 if (dql_avail(&dev_queue->dql) < 0)
3026 return;
3027
3028 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3029 netif_schedule_queue(dev_queue);
114cf580 3030#endif
c5d67bd7
TH
3031}
3032
0042d0c8
FF
3033/**
3034 * netdev_completed_queue - report bytes and packets completed by device
3035 * @dev: network device
3036 * @pkts: actual number of packets sent over the medium
3037 * @bytes: actual number of bytes sent over the medium
3038 *
3039 * Report the number of bytes and packets transmitted by the network device
3040 * hardware queue over the physical medium, @bytes must exactly match the
3041 * @bytes amount passed to netdev_sent_queue()
3042 */
c5d67bd7 3043static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3044 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3045{
3046 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3047}
3048
3049static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3050{
114cf580 3051#ifdef CONFIG_BQL
5c490354 3052 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3053 dql_reset(&q->dql);
3054#endif
c5d67bd7
TH
3055}
3056
0042d0c8
FF
3057/**
3058 * netdev_reset_queue - reset the packets and bytes count of a network device
3059 * @dev_queue: network device
3060 *
3061 * Reset the bytes and packet count of a network device and clear the
3062 * software flow control OFF bit for this network device
3063 */
c5d67bd7
TH
3064static inline void netdev_reset_queue(struct net_device *dev_queue)
3065{
3066 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3067}
3068
b9507bda
DB
3069/**
3070 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3071 * @dev: network device
3072 * @queue_index: given tx queue index
3073 *
3074 * Returns 0 if given tx queue index >= number of device tx queues,
3075 * otherwise returns the originally passed tx queue index.
3076 */
3077static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3078{
3079 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3080 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3081 dev->name, queue_index,
3082 dev->real_num_tx_queues);
3083 return 0;
3084 }
3085
3086 return queue_index;
3087}
3088
bea3348e
SH
3089/**
3090 * netif_running - test if up
3091 * @dev: network device
3092 *
3093 * Test if the device has been brought up.
3094 */
4d29515f 3095static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3096{
3097 return test_bit(__LINK_STATE_START, &dev->state);
3098}
3099
f25f4e44 3100/*
5e82b4b2 3101 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3102 * stop, and a check if it's stopped. All other device management is
3103 * done at the overall netdevice level.
3104 * Also test the device if we're multiqueue.
3105 */
bea3348e
SH
3106
3107/**
3108 * netif_start_subqueue - allow sending packets on subqueue
3109 * @dev: network device
3110 * @queue_index: sub queue index
3111 *
3112 * Start individual transmit queue of a device with multiple transmit queues.
3113 */
f25f4e44
PWJ
3114static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3115{
fd2ea0a7 3116 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3117
3118 netif_tx_start_queue(txq);
f25f4e44
PWJ
3119}
3120
bea3348e
SH
3121/**
3122 * netif_stop_subqueue - stop sending packets on subqueue
3123 * @dev: network device
3124 * @queue_index: sub queue index
3125 *
3126 * Stop individual transmit queue of a device with multiple transmit queues.
3127 */
f25f4e44
PWJ
3128static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3129{
fd2ea0a7 3130 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3131 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3132}
3133
bea3348e
SH
3134/**
3135 * netif_subqueue_stopped - test status of subqueue
3136 * @dev: network device
3137 * @queue_index: sub queue index
3138 *
3139 * Check individual transmit queue of a device with multiple transmit queues.
3140 */
4d29515f
DM
3141static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3142 u16 queue_index)
f25f4e44 3143{
fd2ea0a7 3144 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3145
3146 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3147}
3148
4d29515f
DM
3149static inline bool netif_subqueue_stopped(const struct net_device *dev,
3150 struct sk_buff *skb)
668f895a
PE
3151{
3152 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3153}
bea3348e 3154
46e5da40 3155void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 3156
537c00de 3157#ifdef CONFIG_XPS
53af53ae 3158int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3159 u16 index);
537c00de
AD
3160#else
3161static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3162 const struct cpumask *mask,
537c00de
AD
3163 u16 index)
3164{
3165 return 0;
3166}
3167#endif
3168
5605c762
JP
3169u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3170 unsigned int num_tx_queues);
3171
a3d22a68
VZ
3172/*
3173 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3174 * as a distribution range limit for the returned value.
3175 */
3176static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3177 struct sk_buff *skb)
a3d22a68
VZ
3178{
3179 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3180}
3181
bea3348e
SH
3182/**
3183 * netif_is_multiqueue - test if device has multiple transmit queues
3184 * @dev: network device
3185 *
3186 * Check if device has multiple transmit queues
bea3348e 3187 */
4d29515f 3188static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3189{
a02cec21 3190 return dev->num_tx_queues > 1;
f25f4e44 3191}
1da177e4 3192
f629d208 3193int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3194
a953be53 3195#ifdef CONFIG_SYSFS
f629d208 3196int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3197#else
3198static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3199 unsigned int rxq)
3200{
3201 return 0;
3202}
3203#endif
3204
a953be53
MD
3205#ifdef CONFIG_SYSFS
3206static inline unsigned int get_netdev_rx_queue_index(
3207 struct netdev_rx_queue *queue)
3208{
3209 struct net_device *dev = queue->dev;
3210 int index = queue - dev->_rx;
3211
3212 BUG_ON(index >= dev->num_rx_queues);
3213 return index;
3214}
3215#endif
3216
16917b87 3217#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3218int netif_get_num_default_rss_queues(void);
16917b87 3219
e6247027
ED
3220enum skb_free_reason {
3221 SKB_REASON_CONSUMED,
3222 SKB_REASON_DROPPED,
3223};
3224
3225void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3226void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3227
e6247027
ED
3228/*
3229 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3230 * interrupt context or with hardware interrupts being disabled.
3231 * (in_irq() || irqs_disabled())
3232 *
3233 * We provide four helpers that can be used in following contexts :
3234 *
3235 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3236 * replacing kfree_skb(skb)
3237 *
3238 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3239 * Typically used in place of consume_skb(skb) in TX completion path
3240 *
3241 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3242 * replacing kfree_skb(skb)
3243 *
3244 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3245 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3246 */
e6247027
ED
3247static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3248{
3249 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3250}
3251
3252static inline void dev_consume_skb_irq(struct sk_buff *skb)
3253{
3254 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3255}
3256
3257static inline void dev_kfree_skb_any(struct sk_buff *skb)
3258{
3259 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3260}
3261
3262static inline void dev_consume_skb_any(struct sk_buff *skb)
3263{
3264 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3265}
1da177e4 3266
f629d208
JP
3267int netif_rx(struct sk_buff *skb);
3268int netif_rx_ni(struct sk_buff *skb);
04eb4489 3269int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3270gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3271void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3272struct sk_buff *napi_get_frags(struct napi_struct *napi);
3273gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3274struct packet_offload *gro_find_receive_by_type(__be16 type);
3275struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3276
3277static inline void napi_free_frags(struct napi_struct *napi)
3278{
3279 kfree_skb(napi->skb);
3280 napi->skb = NULL;
3281}
3282
24b27fc4 3283bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3284int netdev_rx_handler_register(struct net_device *dev,
3285 rx_handler_func_t *rx_handler,
3286 void *rx_handler_data);
3287void netdev_rx_handler_unregister(struct net_device *dev);
3288
3289bool dev_valid_name(const char *name);
3290int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3291int dev_ethtool(struct net *net, struct ifreq *);
3292unsigned int dev_get_flags(const struct net_device *);
3293int __dev_change_flags(struct net_device *, unsigned int flags);
3294int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3295void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3296 unsigned int gchanges);
f629d208
JP
3297int dev_change_name(struct net_device *, const char *);
3298int dev_set_alias(struct net_device *, const char *, size_t);
3299int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3300int dev_set_mtu(struct net_device *, int);
3301void dev_set_group(struct net_device *, int);
3302int dev_set_mac_address(struct net_device *, struct sockaddr *);
3303int dev_change_carrier(struct net_device *, bool new_carrier);
3304int dev_get_phys_port_id(struct net_device *dev,
02637fce 3305 struct netdev_phys_item_id *ppid);
db24a904
DA
3306int dev_get_phys_port_name(struct net_device *dev,
3307 char *name, size_t len);
d746d707 3308int dev_change_proto_down(struct net_device *dev, bool proto_down);
a7862b45 3309int dev_change_xdp_fd(struct net_device *dev, int fd);
55a93b3e 3310struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3311struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3312 struct netdev_queue *txq, int *ret);
a0265d28 3313int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3314int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3315bool is_skb_forwardable(const struct net_device *dev,
3316 const struct sk_buff *skb);
1da177e4 3317
74b20582
DA
3318void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3319
20380731 3320extern int netdev_budget;
1da177e4
LT
3321
3322/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3323void netdev_run_todo(void);
1da177e4 3324
bea3348e
SH
3325/**
3326 * dev_put - release reference to device
3327 * @dev: network device
3328 *
9ef4429b 3329 * Release reference to device to allow it to be freed.
bea3348e 3330 */
1da177e4
LT
3331static inline void dev_put(struct net_device *dev)
3332{
933393f5 3333 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3334}
3335
bea3348e
SH
3336/**
3337 * dev_hold - get reference to device
3338 * @dev: network device
3339 *
9ef4429b 3340 * Hold reference to device to keep it from being freed.
bea3348e 3341 */
15333061
SH
3342static inline void dev_hold(struct net_device *dev)
3343{
933393f5 3344 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3345}
1da177e4
LT
3346
3347/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3348 * and _off may be called from IRQ context, but it is caller
3349 * who is responsible for serialization of these calls.
b00055aa
SR
3350 *
3351 * The name carrier is inappropriate, these functions should really be
3352 * called netif_lowerlayer_*() because they represent the state of any
3353 * kind of lower layer not just hardware media.
1da177e4
LT
3354 */
3355
f629d208
JP
3356void linkwatch_init_dev(struct net_device *dev);
3357void linkwatch_fire_event(struct net_device *dev);
3358void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3359
bea3348e
SH
3360/**
3361 * netif_carrier_ok - test if carrier present
3362 * @dev: network device
3363 *
3364 * Check if carrier is present on device
3365 */
4d29515f 3366static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3367{
3368 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3369}
3370
f629d208 3371unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3372
f629d208 3373void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3374
f629d208 3375void netif_carrier_on(struct net_device *dev);
1da177e4 3376
f629d208 3377void netif_carrier_off(struct net_device *dev);
1da177e4 3378
bea3348e
SH
3379/**
3380 * netif_dormant_on - mark device as dormant.
3381 * @dev: network device
3382 *
3383 * Mark device as dormant (as per RFC2863).
3384 *
3385 * The dormant state indicates that the relevant interface is not
3386 * actually in a condition to pass packets (i.e., it is not 'up') but is
3387 * in a "pending" state, waiting for some external event. For "on-
3388 * demand" interfaces, this new state identifies the situation where the
3389 * interface is waiting for events to place it in the up state.
bea3348e 3390 */
b00055aa
SR
3391static inline void netif_dormant_on(struct net_device *dev)
3392{
3393 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3394 linkwatch_fire_event(dev);
3395}
3396
bea3348e
SH
3397/**
3398 * netif_dormant_off - set device as not dormant.
3399 * @dev: network device
3400 *
3401 * Device is not in dormant state.
3402 */
b00055aa
SR
3403static inline void netif_dormant_off(struct net_device *dev)
3404{
3405 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3406 linkwatch_fire_event(dev);
3407}
3408
bea3348e
SH
3409/**
3410 * netif_dormant - test if carrier present
3411 * @dev: network device
3412 *
3413 * Check if carrier is present on device
3414 */
4d29515f 3415static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3416{
3417 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3418}
3419
3420
bea3348e
SH
3421/**
3422 * netif_oper_up - test if device is operational
3423 * @dev: network device
3424 *
3425 * Check if carrier is operational
3426 */
4d29515f 3427static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3428{
b00055aa
SR
3429 return (dev->operstate == IF_OPER_UP ||
3430 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3431}
3432
bea3348e
SH
3433/**
3434 * netif_device_present - is device available or removed
3435 * @dev: network device
3436 *
3437 * Check if device has not been removed from system.
3438 */
4d29515f 3439static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3440{
3441 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3442}
3443
f629d208 3444void netif_device_detach(struct net_device *dev);
1da177e4 3445
f629d208 3446void netif_device_attach(struct net_device *dev);
1da177e4
LT
3447
3448/*
3449 * Network interface message level settings
3450 */
1da177e4
LT
3451
3452enum {
3453 NETIF_MSG_DRV = 0x0001,
3454 NETIF_MSG_PROBE = 0x0002,
3455 NETIF_MSG_LINK = 0x0004,
3456 NETIF_MSG_TIMER = 0x0008,
3457 NETIF_MSG_IFDOWN = 0x0010,
3458 NETIF_MSG_IFUP = 0x0020,
3459 NETIF_MSG_RX_ERR = 0x0040,
3460 NETIF_MSG_TX_ERR = 0x0080,
3461 NETIF_MSG_TX_QUEUED = 0x0100,
3462 NETIF_MSG_INTR = 0x0200,
3463 NETIF_MSG_TX_DONE = 0x0400,
3464 NETIF_MSG_RX_STATUS = 0x0800,
3465 NETIF_MSG_PKTDATA = 0x1000,
3466 NETIF_MSG_HW = 0x2000,
3467 NETIF_MSG_WOL = 0x4000,
3468};
3469
3470#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3471#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3472#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3473#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3474#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3475#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3476#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3477#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3478#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3479#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3480#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3481#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3482#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3483#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3484#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3485
3486static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3487{
3488 /* use default */
3489 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3490 return default_msg_enable_bits;
3491 if (debug_value == 0) /* no output */
3492 return 0;
3493 /* set low N bits */
3494 return (1 << debug_value) - 1;
3495}
3496
c773e847 3497static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3498{
c773e847
DM
3499 spin_lock(&txq->_xmit_lock);
3500 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3501}
3502
fd2ea0a7
DM
3503static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3504{
3505 spin_lock_bh(&txq->_xmit_lock);
3506 txq->xmit_lock_owner = smp_processor_id();
3507}
3508
4d29515f 3509static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3510{
4d29515f 3511 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3512 if (likely(ok))
3513 txq->xmit_lock_owner = smp_processor_id();
3514 return ok;
3515}
3516
3517static inline void __netif_tx_unlock(struct netdev_queue *txq)
3518{
3519 txq->xmit_lock_owner = -1;
3520 spin_unlock(&txq->_xmit_lock);
3521}
3522
3523static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3524{
3525 txq->xmit_lock_owner = -1;
3526 spin_unlock_bh(&txq->_xmit_lock);
3527}
3528
08baf561
ED
3529static inline void txq_trans_update(struct netdev_queue *txq)
3530{
3531 if (txq->xmit_lock_owner != -1)
3532 txq->trans_start = jiffies;
3533}
3534
ba162f8e
FW
3535/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3536static inline void netif_trans_update(struct net_device *dev)
3537{
9b36627a
FW
3538 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3539
3540 if (txq->trans_start != jiffies)
3541 txq->trans_start = jiffies;
ba162f8e
FW
3542}
3543
d29f749e
DJ
3544/**
3545 * netif_tx_lock - grab network device transmit lock
3546 * @dev: network device
d29f749e
DJ
3547 *
3548 * Get network device transmit lock
3549 */
22dd7495
JHS
3550static inline void netif_tx_lock(struct net_device *dev)
3551{
e8a0464c 3552 unsigned int i;
c3f26a26 3553 int cpu;
c773e847 3554
c3f26a26
DM
3555 spin_lock(&dev->tx_global_lock);
3556 cpu = smp_processor_id();
e8a0464c
DM
3557 for (i = 0; i < dev->num_tx_queues; i++) {
3558 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3559
3560 /* We are the only thread of execution doing a
3561 * freeze, but we have to grab the _xmit_lock in
3562 * order to synchronize with threads which are in
3563 * the ->hard_start_xmit() handler and already
3564 * checked the frozen bit.
3565 */
e8a0464c 3566 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3567 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3568 __netif_tx_unlock(txq);
e8a0464c 3569 }
932ff279
HX
3570}
3571
3572static inline void netif_tx_lock_bh(struct net_device *dev)
3573{
e8a0464c
DM
3574 local_bh_disable();
3575 netif_tx_lock(dev);
932ff279
HX
3576}
3577
932ff279
HX
3578static inline void netif_tx_unlock(struct net_device *dev)
3579{
e8a0464c
DM
3580 unsigned int i;
3581
3582 for (i = 0; i < dev->num_tx_queues; i++) {
3583 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3584
c3f26a26
DM
3585 /* No need to grab the _xmit_lock here. If the
3586 * queue is not stopped for another reason, we
3587 * force a schedule.
3588 */
3589 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3590 netif_schedule_queue(txq);
c3f26a26
DM
3591 }
3592 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3593}
3594
3595static inline void netif_tx_unlock_bh(struct net_device *dev)
3596{
e8a0464c
DM
3597 netif_tx_unlock(dev);
3598 local_bh_enable();
932ff279
HX
3599}
3600
c773e847 3601#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3602 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3603 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3604 } \
3605}
3606
5efeac44
EB
3607#define HARD_TX_TRYLOCK(dev, txq) \
3608 (((dev->features & NETIF_F_LLTX) == 0) ? \
3609 __netif_tx_trylock(txq) : \
3610 true )
3611
c773e847 3612#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3613 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3614 __netif_tx_unlock(txq); \
22dd7495
JHS
3615 } \
3616}
3617
1da177e4
LT
3618static inline void netif_tx_disable(struct net_device *dev)
3619{
fd2ea0a7 3620 unsigned int i;
c3f26a26 3621 int cpu;
fd2ea0a7 3622
c3f26a26
DM
3623 local_bh_disable();
3624 cpu = smp_processor_id();
fd2ea0a7
DM
3625 for (i = 0; i < dev->num_tx_queues; i++) {
3626 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3627
3628 __netif_tx_lock(txq, cpu);
fd2ea0a7 3629 netif_tx_stop_queue(txq);
c3f26a26 3630 __netif_tx_unlock(txq);
fd2ea0a7 3631 }
c3f26a26 3632 local_bh_enable();
1da177e4
LT
3633}
3634
e308a5d8
DM
3635static inline void netif_addr_lock(struct net_device *dev)
3636{
3637 spin_lock(&dev->addr_list_lock);
3638}
3639
2429f7ac
JP
3640static inline void netif_addr_lock_nested(struct net_device *dev)
3641{
25175ba5
VY
3642 int subclass = SINGLE_DEPTH_NESTING;
3643
3644 if (dev->netdev_ops->ndo_get_lock_subclass)
3645 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3646
3647 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3648}
3649
e308a5d8
DM
3650static inline void netif_addr_lock_bh(struct net_device *dev)
3651{
3652 spin_lock_bh(&dev->addr_list_lock);
3653}
3654
3655static inline void netif_addr_unlock(struct net_device *dev)
3656{
3657 spin_unlock(&dev->addr_list_lock);
3658}
3659
3660static inline void netif_addr_unlock_bh(struct net_device *dev)
3661{
3662 spin_unlock_bh(&dev->addr_list_lock);
3663}
3664
f001fde5 3665/*
31278e71 3666 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3667 * rcu_read_lock held.
3668 */
3669#define for_each_dev_addr(dev, ha) \
31278e71 3670 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3671
1da177e4
LT
3672/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3673
f629d208 3674void ether_setup(struct net_device *dev);
1da177e4
LT
3675
3676/* Support for loadable net-drivers */
f629d208 3677struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3678 unsigned char name_assign_type,
f629d208
JP
3679 void (*setup)(struct net_device *),
3680 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3681#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3682 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3683
c835a677
TG
3684#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3685 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3686 count)
36909ea4 3687
f629d208
JP
3688int register_netdev(struct net_device *dev);
3689void unregister_netdev(struct net_device *dev);
f001fde5 3690
22bedad3 3691/* General hardware address lists handling functions */
f629d208
JP
3692int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3693 struct netdev_hw_addr_list *from_list, int addr_len);
3694void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3695 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3696int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3697 struct net_device *dev,
3698 int (*sync)(struct net_device *, const unsigned char *),
3699 int (*unsync)(struct net_device *,
3700 const unsigned char *));
3701void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3702 struct net_device *dev,
3703 int (*unsync)(struct net_device *,
3704 const unsigned char *));
f629d208 3705void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3706
f001fde5 3707/* Functions used for device addresses handling */
f629d208
JP
3708int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3709 unsigned char addr_type);
3710int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3711 unsigned char addr_type);
f629d208
JP
3712void dev_addr_flush(struct net_device *dev);
3713int dev_addr_init(struct net_device *dev);
a748ee24
JP
3714
3715/* Functions used for unicast addresses handling */
f629d208
JP
3716int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3717int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3718int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3719int dev_uc_sync(struct net_device *to, struct net_device *from);
3720int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3721void dev_uc_unsync(struct net_device *to, struct net_device *from);
3722void dev_uc_flush(struct net_device *dev);
3723void dev_uc_init(struct net_device *dev);
f001fde5 3724
670e5b8e
AD
3725/**
3726 * __dev_uc_sync - Synchonize device's unicast list
3727 * @dev: device to sync
3728 * @sync: function to call if address should be added
3729 * @unsync: function to call if address should be removed
3730 *
3731 * Add newly added addresses to the interface, and release
3732 * addresses that have been deleted.
5e82b4b2 3733 */
670e5b8e
AD
3734static inline int __dev_uc_sync(struct net_device *dev,
3735 int (*sync)(struct net_device *,
3736 const unsigned char *),
3737 int (*unsync)(struct net_device *,
3738 const unsigned char *))
3739{
3740 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3741}
3742
3743/**
e793c0f7 3744 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3745 * @dev: device to sync
3746 * @unsync: function to call if address should be removed
3747 *
3748 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3749 */
670e5b8e
AD
3750static inline void __dev_uc_unsync(struct net_device *dev,
3751 int (*unsync)(struct net_device *,
3752 const unsigned char *))
3753{
3754 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3755}
3756
22bedad3 3757/* Functions used for multicast addresses handling */
f629d208
JP
3758int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3759int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3760int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3761int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3762int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3763int dev_mc_sync(struct net_device *to, struct net_device *from);
3764int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3765void dev_mc_unsync(struct net_device *to, struct net_device *from);
3766void dev_mc_flush(struct net_device *dev);
3767void dev_mc_init(struct net_device *dev);
f001fde5 3768
670e5b8e
AD
3769/**
3770 * __dev_mc_sync - Synchonize device's multicast list
3771 * @dev: device to sync
3772 * @sync: function to call if address should be added
3773 * @unsync: function to call if address should be removed
3774 *
3775 * Add newly added addresses to the interface, and release
3776 * addresses that have been deleted.
5e82b4b2 3777 */
670e5b8e
AD
3778static inline int __dev_mc_sync(struct net_device *dev,
3779 int (*sync)(struct net_device *,
3780 const unsigned char *),
3781 int (*unsync)(struct net_device *,
3782 const unsigned char *))
3783{
3784 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3785}
3786
3787/**
e793c0f7 3788 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3789 * @dev: device to sync
3790 * @unsync: function to call if address should be removed
3791 *
3792 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3793 */
670e5b8e
AD
3794static inline void __dev_mc_unsync(struct net_device *dev,
3795 int (*unsync)(struct net_device *,
3796 const unsigned char *))
3797{
3798 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3799}
3800
4417da66 3801/* Functions used for secondary unicast and multicast support */
f629d208
JP
3802void dev_set_rx_mode(struct net_device *dev);
3803void __dev_set_rx_mode(struct net_device *dev);
3804int dev_set_promiscuity(struct net_device *dev, int inc);
3805int dev_set_allmulti(struct net_device *dev, int inc);
3806void netdev_state_change(struct net_device *dev);
3807void netdev_notify_peers(struct net_device *dev);
3808void netdev_features_change(struct net_device *dev);
1da177e4 3809/* Load a device via the kmod */
f629d208
JP
3810void dev_load(struct net *net, const char *name);
3811struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3812 struct rtnl_link_stats64 *storage);
3813void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3814 const struct net_device_stats *netdev_stats);
eeda3fd6 3815
1da177e4 3816extern int netdev_max_backlog;
3b098e2d 3817extern int netdev_tstamp_prequeue;
1da177e4 3818extern int weight_p;
9ff162a8 3819
f629d208 3820bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3821struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3822 struct list_head **iter);
f629d208
JP
3823struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3824 struct list_head **iter);
8b5be856 3825
44a40855
VY
3826/* iterate through upper list, must be called under RCU read lock */
3827#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3828 for (iter = &(dev)->adj_list.upper, \
3829 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3830 updev; \
3831 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3832
8b5be856 3833/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3834#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3835 for (iter = &(dev)->all_adj_list.upper, \
3836 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3837 updev; \
3838 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3839
f629d208
JP
3840void *netdev_lower_get_next_private(struct net_device *dev,
3841 struct list_head **iter);
3842void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3843 struct list_head **iter);
31088a11
VF
3844
3845#define netdev_for_each_lower_private(dev, priv, iter) \
3846 for (iter = (dev)->adj_list.lower.next, \
3847 priv = netdev_lower_get_next_private(dev, &(iter)); \
3848 priv; \
3849 priv = netdev_lower_get_next_private(dev, &(iter)))
3850
3851#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3852 for (iter = &(dev)->adj_list.lower, \
3853 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3854 priv; \
3855 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3856
4085ebe8
VY
3857void *netdev_lower_get_next(struct net_device *dev,
3858 struct list_head **iter);
7ce856aa 3859
4085ebe8 3860#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3861 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3862 ldev = netdev_lower_get_next(dev, &(iter)); \
3863 ldev; \
3864 ldev = netdev_lower_get_next(dev, &(iter)))
3865
7ce856aa
JP
3866struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3867 struct list_head **iter);
3868struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3869 struct list_head **iter);
3870
3871#define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3872 for (iter = (dev)->all_adj_list.lower.next, \
3873 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3874 ldev; \
3875 ldev = netdev_all_lower_get_next(dev, &(iter)))
3876
3877#define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
3878 for (iter = (dev)->all_adj_list.lower.next, \
3879 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3880 ldev; \
3881 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3882
f629d208 3883void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3884void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3885struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3886struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3887int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3888int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3889 struct net_device *upper_dev,
29bf24af 3890 void *upper_priv, void *upper_info);
f629d208
JP
3891void netdev_upper_dev_unlink(struct net_device *dev,
3892 struct net_device *upper_dev);
5bb025fa 3893void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3894void *netdev_lower_dev_get_private(struct net_device *dev,
3895 struct net_device *lower_dev);
04d48266
JP
3896void netdev_lower_state_changed(struct net_device *lower_dev,
3897 void *lower_state_info);
18bfb924
JP
3898int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3899 struct neighbour *n);
3900void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3901 struct neighbour *n);
960fb622
ED
3902
3903/* RSS keys are 40 or 52 bytes long */
3904#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3905extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3906void netdev_rss_key_fill(void *buffer, size_t len);
3907
952fcfd0 3908int dev_get_nest_level(struct net_device *dev);
f629d208
JP
3909int skb_checksum_help(struct sk_buff *skb);
3910struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3911 netdev_features_t features, bool tx_path);
3912struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3913 netdev_features_t features);
12b0004d 3914
61bd3857
MS
3915struct netdev_bonding_info {
3916 ifslave slave;
3917 ifbond master;
3918};
3919
3920struct netdev_notifier_bonding_info {
3921 struct netdev_notifier_info info; /* must be first */
3922 struct netdev_bonding_info bonding_info;
3923};
3924
3925void netdev_bonding_info_change(struct net_device *dev,
3926 struct netdev_bonding_info *bonding_info);
3927
12b0004d
CW
3928static inline
3929struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3930{
3931 return __skb_gso_segment(skb, features, true);
3932}
53d6471c 3933__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3934
3935static inline bool can_checksum_protocol(netdev_features_t features,
3936 __be16 protocol)
3937{
c8cd0989
TH
3938 if (protocol == htons(ETH_P_FCOE))
3939 return !!(features & NETIF_F_FCOE_CRC);
3940
3941 /* Assume this is an IP checksum (not SCTP CRC) */
3942
3943 if (features & NETIF_F_HW_CSUM) {
3944 /* Can checksum everything */
3945 return true;
3946 }
3947
3948 switch (protocol) {
3949 case htons(ETH_P_IP):
3950 return !!(features & NETIF_F_IP_CSUM);
3951 case htons(ETH_P_IPV6):
3952 return !!(features & NETIF_F_IPV6_CSUM);
3953 default:
3954 return false;
3955 }
ec5f0615 3956}
12b0004d 3957
6ae23ad3
TH
3958/* Map an ethertype into IP protocol if possible */
3959static inline int eproto_to_ipproto(int eproto)
3960{
3961 switch (eproto) {
3962 case htons(ETH_P_IP):
3963 return IPPROTO_IP;
3964 case htons(ETH_P_IPV6):
3965 return IPPROTO_IPV6;
3966 default:
3967 return -1;
3968 }
3969}
3970
fb286bb2 3971#ifdef CONFIG_BUG
f629d208 3972void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3973#else
3974static inline void netdev_rx_csum_fault(struct net_device *dev)
3975{
3976}
3977#endif
1da177e4 3978/* rx skb timestamps */
f629d208
JP
3979void net_enable_timestamp(void);
3980void net_disable_timestamp(void);
1da177e4 3981
20380731 3982#ifdef CONFIG_PROC_FS
f629d208 3983int __init dev_proc_init(void);
900ff8c6
CW
3984#else
3985#define dev_proc_init() 0
20380731
ACM
3986#endif
3987
4798248e 3988static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3989 struct sk_buff *skb, struct net_device *dev,
3990 bool more)
4798248e 3991{
fa2dbdc2 3992 skb->xmit_more = more ? 1 : 0;
0b725a2c 3993 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3994}
3995
10b3ad8c 3996static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3997 struct netdev_queue *txq, bool more)
4798248e
DM
3998{
3999 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4000 int rc;
4798248e 4001
fa2dbdc2 4002 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4003 if (rc == NETDEV_TX_OK)
4004 txq_trans_update(txq);
4005
4006 return rc;
4798248e
DM
4007}
4008
42a2d923
LT
4009int netdev_class_create_file_ns(struct class_attribute *class_attr,
4010 const void *ns);
4011void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4012 const void *ns);
58292cbe
TH
4013
4014static inline int netdev_class_create_file(struct class_attribute *class_attr)
4015{
4016 return netdev_class_create_file_ns(class_attr, NULL);
4017}
4018
4019static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4020{
4021 netdev_class_remove_file_ns(class_attr, NULL);
4022}
b8a9787e 4023
04600794
JB
4024extern struct kobj_ns_type_operations net_ns_type_operations;
4025
f629d208 4026const char *netdev_drivername(const struct net_device *dev);
6579e57b 4027
f629d208 4028void linkwatch_run_queue(void);
20380731 4029
da08143b
MK
4030static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4031 netdev_features_t f2)
4032{
c8cd0989
TH
4033 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4034 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4035 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4036 else
b6a0e72a 4037 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4038 }
da08143b 4039
c8cd0989 4040 return f1 & f2;
da08143b
MK
4041}
4042
c8f44aff
MM
4043static inline netdev_features_t netdev_get_wanted_features(
4044 struct net_device *dev)
5455c699
MM
4045{
4046 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4047}
c8f44aff
MM
4048netdev_features_t netdev_increment_features(netdev_features_t all,
4049 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4050
4051/* Allow TSO being used on stacked device :
4052 * Performing the GSO segmentation before last device
4053 * is a performance improvement.
4054 */
4055static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4056 netdev_features_t mask)
4057{
4058 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4059}
4060
6cb6a27c 4061int __netdev_update_features(struct net_device *dev);
5455c699 4062void netdev_update_features(struct net_device *dev);
afe12cc8 4063void netdev_change_features(struct net_device *dev);
7f353bf2 4064
fc4a7489
PM
4065void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4066 struct net_device *dev);
4067
e38f3025
TM
4068netdev_features_t passthru_features_check(struct sk_buff *skb,
4069 struct net_device *dev,
4070 netdev_features_t features);
c1e756bf 4071netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4072
4d29515f 4073static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4074{
7b748340 4075 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4076
4077 /* check flags correspondence */
4078 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4079 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4080 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4081 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4082 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4083 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4084 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4085 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4086 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4087 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4088 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4089 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4090 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4091 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4092 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4093 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
0345e186 4094
d6b4991a 4095 return (features & feature) == feature;
576a30eb
HX
4096}
4097
4d29515f 4098static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4099{
278b2513 4100 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4101 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4102}
4103
8b86a61d 4104static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4105 netdev_features_t features)
7967168c 4106{
fc741216 4107 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4108 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4109 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4110}
4111
82cc1a7a
PWJ
4112static inline void netif_set_gso_max_size(struct net_device *dev,
4113 unsigned int size)
4114{
4115 dev->gso_max_size = size;
4116}
4117
7a7ffbab
WCC
4118static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4119 int pulled_hlen, u16 mac_offset,
4120 int mac_len)
4121{
4122 skb->protocol = protocol;
4123 skb->encapsulation = 1;
4124 skb_push(skb, pulled_hlen);
4125 skb_reset_transport_header(skb);
4126 skb->mac_header = mac_offset;
4127 skb->network_header = skb->mac_header + mac_len;
4128 skb->mac_len = mac_len;
4129}
4130
3c175784
SD
4131static inline bool netif_is_macsec(const struct net_device *dev)
4132{
4133 return dev->priv_flags & IFF_MACSEC;
4134}
4135
b618aaa9 4136static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4137{
4138 return dev->priv_flags & IFF_MACVLAN;
4139}
4140
b618aaa9 4141static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4142{
4143 return dev->priv_flags & IFF_MACVLAN_PORT;
4144}
4145
b618aaa9 4146static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4147{
4148 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4149}
4150
b618aaa9 4151static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4152{
4153 return dev->priv_flags & IFF_IPVLAN_MASTER;
4154}
4155
b618aaa9 4156static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4157{
4158 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4159}
4160
b618aaa9 4161static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4162{
4163 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4164}
4165
3bdc0eba
BG
4166static inline bool netif_supports_nofcs(struct net_device *dev)
4167{
4168 return dev->priv_flags & IFF_SUPP_NOFCS;
4169}
4170
007979ea 4171static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4172{
007979ea 4173 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4174}
4175
fee6d4c7
DA
4176static inline bool netif_is_l3_slave(const struct net_device *dev)
4177{
4178 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4179}
4180
0894ae3f
JP
4181static inline bool netif_is_bridge_master(const struct net_device *dev)
4182{
4183 return dev->priv_flags & IFF_EBRIDGE;
4184}
4185
28f9ee22
VY
4186static inline bool netif_is_bridge_port(const struct net_device *dev)
4187{
4188 return dev->priv_flags & IFF_BRIDGE_PORT;
4189}
4190
35d4e172
JP
4191static inline bool netif_is_ovs_master(const struct net_device *dev)
4192{
4193 return dev->priv_flags & IFF_OPENVSWITCH;
4194}
4195
b618aaa9 4196static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4197{
4198 return dev->priv_flags & IFF_TEAM;
4199}
4200
b618aaa9 4201static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4202{
4203 return dev->priv_flags & IFF_TEAM_PORT;
4204}
4205
b618aaa9 4206static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4207{
4208 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4209}
4210
b618aaa9 4211static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4212{
4213 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4214}
4215
d4ab4286
KJ
4216static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4217{
4218 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4219}
4220
02875878
ED
4221/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4222static inline void netif_keep_dst(struct net_device *dev)
4223{
4224 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4225}
4226
18d3df3e
PA
4227/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4228static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4229{
4230 /* TODO: reserve and use an additional IFF bit, if we get more users */
4231 return dev->priv_flags & IFF_MACSEC;
4232}
4233
505d4f73 4234extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4235
571ba423
JP
4236/* Logging, debugging and troubleshooting/diagnostic helpers. */
4237
4238/* netdev_printk helpers, similar to dev_printk */
4239
4240static inline const char *netdev_name(const struct net_device *dev)
4241{
c6f854d5
VF
4242 if (!dev->name[0] || strchr(dev->name, '%'))
4243 return "(unnamed net_device)";
571ba423
JP
4244 return dev->name;
4245}
4246
ccc7f496
VF
4247static inline const char *netdev_reg_state(const struct net_device *dev)
4248{
4249 switch (dev->reg_state) {
4250 case NETREG_UNINITIALIZED: return " (uninitialized)";
4251 case NETREG_REGISTERED: return "";
4252 case NETREG_UNREGISTERING: return " (unregistering)";
4253 case NETREG_UNREGISTERED: return " (unregistered)";
4254 case NETREG_RELEASED: return " (released)";
4255 case NETREG_DUMMY: return " (dummy)";
4256 }
4257
4258 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4259 return " (unknown)";
4260}
4261
f629d208 4262__printf(3, 4)
6ea754eb
JP
4263void netdev_printk(const char *level, const struct net_device *dev,
4264 const char *format, ...);
f629d208 4265__printf(2, 3)
6ea754eb 4266void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4267__printf(2, 3)
6ea754eb 4268void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4269__printf(2, 3)
6ea754eb 4270void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4271__printf(2, 3)
6ea754eb 4272void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4273__printf(2, 3)
6ea754eb 4274void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4275__printf(2, 3)
6ea754eb 4276void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4277__printf(2, 3)
6ea754eb 4278void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4279
8909c9ad
VK
4280#define MODULE_ALIAS_NETDEV(device) \
4281 MODULE_ALIAS("netdev-" device)
4282
b558c96f 4283#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4284#define netdev_dbg(__dev, format, args...) \
4285do { \
ffa10cb4 4286 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4287} while (0)
b558c96f
JC
4288#elif defined(DEBUG)
4289#define netdev_dbg(__dev, format, args...) \
4290 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4291#else
4292#define netdev_dbg(__dev, format, args...) \
4293({ \
4294 if (0) \
4295 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4296})
4297#endif
4298
4299#if defined(VERBOSE_DEBUG)
4300#define netdev_vdbg netdev_dbg
4301#else
4302
4303#define netdev_vdbg(dev, format, args...) \
4304({ \
4305 if (0) \
4306 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4307 0; \
4308})
4309#endif
4310
4311/*
4312 * netdev_WARN() acts like dev_printk(), but with the key difference
4313 * of using a WARN/WARN_ON to get the message out, including the
4314 * file/line information and a backtrace.
4315 */
4316#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4317 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4318 netdev_reg_state(dev), ##args)
571ba423 4319
b3d95c5c
JP
4320/* netif printk helpers, similar to netdev_printk */
4321
4322#define netif_printk(priv, type, level, dev, fmt, args...) \
4323do { \
4324 if (netif_msg_##type(priv)) \
4325 netdev_printk(level, (dev), fmt, ##args); \
4326} while (0)
4327
f45f4321
JP
4328#define netif_level(level, priv, type, dev, fmt, args...) \
4329do { \
4330 if (netif_msg_##type(priv)) \
4331 netdev_##level(dev, fmt, ##args); \
4332} while (0)
4333
b3d95c5c 4334#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4335 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4336#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4337 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4338#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4339 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4340#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4341 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4342#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4343 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4344#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4345 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4346#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4347 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4348
0053ea9c 4349#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4350#define netif_dbg(priv, type, netdev, format, args...) \
4351do { \
4352 if (netif_msg_##type(priv)) \
b5fb0a03 4353 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4354} while (0)
0053ea9c
JP
4355#elif defined(DEBUG)
4356#define netif_dbg(priv, type, dev, format, args...) \
4357 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4358#else
4359#define netif_dbg(priv, type, dev, format, args...) \
4360({ \
4361 if (0) \
4362 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4363 0; \
4364})
4365#endif
4366
4367#if defined(VERBOSE_DEBUG)
bcfcc450 4368#define netif_vdbg netif_dbg
b3d95c5c
JP
4369#else
4370#define netif_vdbg(priv, type, dev, format, args...) \
4371({ \
4372 if (0) \
a4ed89cb 4373 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4374 0; \
4375})
4376#endif
571ba423 4377
900ff8c6
CW
4378/*
4379 * The list of packet types we will receive (as opposed to discard)
4380 * and the routines to invoke.
4381 *
4382 * Why 16. Because with 16 the only overlap we get on a hash of the
4383 * low nibble of the protocol value is RARP/SNAP/X.25.
4384 *
4385 * NOTE: That is no longer true with the addition of VLAN tags. Not
4386 * sure which should go first, but I bet it won't make much
4387 * difference if we are running VLANs. The good news is that
4388 * this protocol won't be in the list unless compiled in, so
4389 * the average user (w/out VLANs) will not be adversely affected.
4390 * --BLG
4391 *
4392 * 0800 IP
4393 * 8100 802.1Q VLAN
4394 * 0001 802.3
4395 * 0002 AX.25
4396 * 0004 802.2
4397 * 8035 RARP
4398 * 0005 SNAP
4399 * 0805 X.25
4400 * 0806 ARP
4401 * 8137 IPX
4402 * 0009 Localtalk
4403 * 86DD IPv6
4404 */
4405#define PTYPE_HASH_SIZE (16)
4406#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4407
385a154c 4408#endif /* _LINUX_NETDEVICE_H */