]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/netdevice.h
bpf: Fix compilation warning in __bpf_lru_list_rotate_inactive
[mirror_ubuntu-bionic-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
cf85d08f 44#include <net/dsa.h>
7a6b6f51 45#ifdef CONFIG_DCB
2f90b865
AD
46#include <net/dcbnl.h>
47#endif
5bc1421e 48#include <net/netprio_cgroup.h>
a050c33f 49
a59e2ecb 50#include <linux/netdev_features.h>
77162022 51#include <linux/neighbour.h>
607ca46e 52#include <uapi/linux/netdevice.h>
61bd3857 53#include <uapi/linux/if_bonding.h>
e4c6734e 54#include <uapi/linux/pkt_cls.h>
59cc1f61 55#include <linux/hashtable.h>
a59e2ecb 56
115c1d6e 57struct netpoll_info;
313162d0 58struct device;
c1f19b51 59struct phy_device;
704232c2
JB
60/* 802.11 specific */
61struct wireless_dev;
98a18b6f
AA
62/* 802.15.4 specific */
63struct wpan_dev;
03c57747 64struct mpls_dev;
7c46a640
AD
65/* UDP Tunnel offloads */
66struct udp_tunnel_info;
a7862b45 67struct bpf_prog;
1da177e4 68
f629d208
JP
69void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
d07d7507 71
9a1654ba
JP
72/* Backlog congestion levels */
73#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74#define NET_RX_DROP 1 /* packet dropped */
75
572a9d7b
PM
76/*
77 * Transmit return codes: transmit return codes originate from three different
78 * namespaces:
79 *
80 * - qdisc return codes
81 * - driver transmit return codes
82 * - errno values
83 *
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
90 * others are propagated to higher layers.
91 */
92
93/* qdisc ->enqueue() return codes. */
94#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
95#define NET_XMIT_DROP 0x01 /* skb dropped */
96#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 97#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 98
b9df3cb8
GR
99/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 102#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
103#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
104
dc1f8bf6 105/* Driver transmit return codes */
9a1654ba 106#define NETDEV_TX_MASK 0xf0
572a9d7b 107
dc1f8bf6 108enum netdev_tx {
572a9d7b 109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
112};
113typedef enum netdev_tx netdev_tx_t;
114
9a1654ba
JP
115/*
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
118 */
119static inline bool dev_xmit_complete(int rc)
120{
121 /*
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
126 */
127 if (likely(rc < NET_XMIT_MASK))
128 return true;
129
130 return false;
131}
132
1da177e4 133/*
5e82b4b2 134 * Compute the worst-case header length according to the protocols
1da177e4
LT
135 * used.
136 */
fe2918b0 137
c0eb4540
KS
138#if defined(CONFIG_HYPERV_NET)
139# define LL_MAX_HEADER 128
140#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
141# if defined(CONFIG_MAC80211_MESH)
142# define LL_MAX_HEADER 128
143# else
144# define LL_MAX_HEADER 96
145# endif
1da177e4 146#else
8388e3da 147# define LL_MAX_HEADER 32
1da177e4
LT
148#endif
149
d11ead75
BH
150#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
152#define MAX_HEADER LL_MAX_HEADER
153#else
154#define MAX_HEADER (LL_MAX_HEADER + 48)
155#endif
156
157/*
be1f3c2c
BH
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
1da177e4 160 */
fe2918b0 161
d94d9fee 162struct net_device_stats {
3cfde79c
BH
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
1da177e4 172 unsigned long collisions;
1da177e4 173 unsigned long rx_length_errors;
3cfde79c
BH
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
1da177e4
LT
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
1da177e4
LT
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
186};
187
1da177e4
LT
188
189#include <linux/cache.h>
190#include <linux/skbuff.h>
191
adc9300e 192#ifdef CONFIG_RPS
c5905afb
IM
193#include <linux/static_key.h>
194extern struct static_key rps_needed;
adc9300e
ED
195#endif
196
1da177e4
LT
197struct neighbour;
198struct neigh_parms;
199struct sk_buff;
200
f001fde5
JP
201struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
ccffad25
JP
205#define NETDEV_HW_ADDR_T_LAN 1
206#define NETDEV_HW_ADDR_T_SAN 2
207#define NETDEV_HW_ADDR_T_SLAVE 3
208#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 209#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 210 bool global_use;
4cd729b0 211 int sync_cnt;
8f8f103d 212 int refcount;
4543fbef 213 int synced;
f001fde5
JP
214 struct rcu_head rcu_head;
215};
216
31278e71
JP
217struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220};
221
22bedad3
JP
222#define netdev_hw_addr_list_count(l) ((l)->count)
223#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224#define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 226
22bedad3
JP
227#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229#define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 231
22bedad3
JP
232#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 234#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 236
d94d9fee 237struct hh_cache {
f6b72b62 238 u16 hh_len;
5c25f686 239 u16 __pad;
3644f0ce 240 seqlock_t hh_lock;
1da177e4
LT
241
242 /* cached hardware header; allow for machine alignment needs. */
243#define HH_DATA_MOD 16
244#define HH_DATA_OFF(__len) \
5ba0eac6 245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
246#define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
249};
250
5e82b4b2 251/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
252 * Alternative is:
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
255 *
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
258 */
259#define LL_RESERVED_SPACE(dev) \
f5184d26 260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 261#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 263
3b04ddde
SH
264struct header_ops {
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
95c96174 267 const void *saddr, unsigned int len);
3b04ddde 268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
270 void (*cache_update)(struct hh_cache *hh,
271 const struct net_device *dev,
272 const unsigned char *haddr);
2793a23a 273 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
274};
275
1da177e4 276/* These flag bits are private to the generic network queueing
5e82b4b2 277 * layer; they may not be explicitly referenced by any other
1da177e4
LT
278 * code.
279 */
280
d94d9fee 281enum netdev_state_t {
1da177e4
LT
282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
1da177e4 284 __LINK_STATE_NOCARRIER,
b00055aa
SR
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
1da177e4
LT
287};
288
289
290/*
5e82b4b2 291 * This structure holds boot-time configured netdevice settings. They
fe2918b0 292 * are then used in the device probing.
1da177e4
LT
293 */
294struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297};
298#define NETDEV_BOOT_SETUP_MAX 8
299
f629d208 300int __init netdev_boot_setup(char *str);
1da177e4 301
bea3348e
SH
302/*
303 * Structure for NAPI scheduling similar to tasklet but with weighting
304 */
305struct napi_struct {
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 309 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
310 * can remove from the list right before clearing the bit.
311 */
312 struct list_head poll_list;
313
314 unsigned long state;
315 int weight;
404f7c9e 316 unsigned int gro_count;
bea3348e
SH
317 int (*poll)(struct napi_struct *, int);
318#ifdef CONFIG_NETPOLL
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_del - remove a NAPI from global table
5e82b4b2 472 * @napi: NAPI context
af12fa6e 473 *
5e82b4b2 474 * Warning: caller must observe RCU grace period
34cbe27e
ED
475 * before freeing memory containing @napi, if
476 * this function returns true.
93d05d4a 477 * Note: core networking stack automatically calls it
5e82b4b2 478 * from netif_napi_del().
93d05d4a 479 * Drivers might want to call this helper to combine all
5e82b4b2 480 * the needed RCU grace periods into a single one.
af12fa6e 481 */
34cbe27e 482bool napi_hash_del(struct napi_struct *napi);
af12fa6e 483
bea3348e
SH
484/**
485 * napi_disable - prevent NAPI from scheduling
5e82b4b2 486 * @n: NAPI context
bea3348e
SH
487 *
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
490 */
3b47d303 491void napi_disable(struct napi_struct *n);
bea3348e
SH
492
493/**
494 * napi_enable - enable NAPI scheduling
5e82b4b2 495 * @n: NAPI context
bea3348e
SH
496 *
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
499 */
500static inline void napi_enable(struct napi_struct *n)
501{
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 503 smp_mb__before_atomic();
bea3348e 504 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 505 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
506}
507
c264c3de
SH
508/**
509 * napi_synchronize - wait until NAPI is not running
5e82b4b2 510 * @n: NAPI context
c264c3de
SH
511 *
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
515 */
516static inline void napi_synchronize(const struct napi_struct *n)
517{
facc432f
AB
518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
520 msleep(1);
521 else
522 barrier();
c264c3de 523}
c264c3de 524
d94d9fee 525enum netdev_queue_state_t {
73466498
TH
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
c3f26a26 528 __QUEUE_STATE_FROZEN,
79d16385 529};
8e2f1a63
DB
530
531#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
534
535#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
537 QUEUE_STATE_FROZEN)
538#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
539 QUEUE_STATE_FROZEN)
540
73466498
TH
541/*
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
549 */
79d16385 550
bb949fbd 551struct netdev_queue {
6a321cb3 552/*
5e82b4b2 553 * read-mostly part
6a321cb3 554 */
bb949fbd 555 struct net_device *dev;
46e5da40 556 struct Qdisc __rcu *qdisc;
b0e1e646 557 struct Qdisc *qdisc_sleeping;
ccf5ff69 558#ifdef CONFIG_SYSFS
1d24eb48
TH
559 struct kobject kobj;
560#endif
f2cd2d3e
ED
561#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 int numa_node;
563#endif
c0ef079c
FW
564 unsigned long tx_maxrate;
565 /*
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
568 */
569 unsigned long trans_timeout;
6a321cb3 570/*
5e82b4b2 571 * write-mostly part
6a321cb3
ED
572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
9d21493b 575 /*
9b36627a 576 * Time (in jiffies) of last Tx
9d21493b
ED
577 */
578 unsigned long trans_start;
ccf5ff69 579
114cf580
TH
580 unsigned long state;
581
582#ifdef CONFIG_BQL
583 struct dql dql;
584#endif
e8a0464c 585} ____cacheline_aligned_in_smp;
bb949fbd 586
f2cd2d3e
ED
587static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
588{
589#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 return q->numa_node;
591#else
b236da69 592 return NUMA_NO_NODE;
f2cd2d3e
ED
593#endif
594}
595
596static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
597{
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 q->numa_node = node;
600#endif
601}
602
df334545 603#ifdef CONFIG_RPS
0a9627f2
TH
604/*
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
607 */
608struct rps_map {
609 unsigned int len;
610 struct rcu_head rcu;
611 u16 cpus[0];
612};
60b778ce 613#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 614
fec5e652 615/*
c445477d
BH
616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
fec5e652
TH
619 */
620struct rps_dev_flow {
621 u16 cpu;
c445477d 622 u16 filter;
fec5e652
TH
623 unsigned int last_qtail;
624};
c445477d 625#define RPS_NO_FILTER 0xffff
fec5e652
TH
626
627/*
628 * The rps_dev_flow_table structure contains a table of flow mappings.
629 */
630struct rps_dev_flow_table {
631 unsigned int mask;
632 struct rcu_head rcu;
fec5e652
TH
633 struct rps_dev_flow flows[0];
634};
635#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 636 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
637
638/*
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
567e4b79 643 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 646 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
647 */
648struct rps_sock_flow_table {
567e4b79 649 u32 mask;
93c1af6c
ED
650
651 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 652};
567e4b79 653#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
654
655#define RPS_NO_CPU 0xffff
656
567e4b79
ED
657extern u32 rps_cpu_mask;
658extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
659
fec5e652
TH
660static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
661 u32 hash)
662{
663 if (table && hash) {
567e4b79
ED
664 unsigned int index = hash & table->mask;
665 u32 val = hash & ~rps_cpu_mask;
fec5e652 666
5e82b4b2 667 /* We only give a hint, preemption can change CPU under us */
567e4b79 668 val |= raw_smp_processor_id();
fec5e652 669
567e4b79
ED
670 if (table->ents[index] != val)
671 table->ents[index] = val;
fec5e652
TH
672 }
673}
674
c445477d 675#ifdef CONFIG_RFS_ACCEL
f629d208
JP
676bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
677 u16 filter_id);
c445477d 678#endif
a953be53 679#endif /* CONFIG_RPS */
c445477d 680
0a9627f2
TH
681/* This structure contains an instance of an RX queue. */
682struct netdev_rx_queue {
a953be53 683#ifdef CONFIG_RPS
6e3f7faf
ED
684 struct rps_map __rcu *rps_map;
685 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 686#endif
6e3f7faf 687 struct kobject kobj;
fe822240 688 struct net_device *dev;
0a9627f2 689} ____cacheline_aligned_in_smp;
a953be53
MD
690
691/*
692 * RX queue sysfs structures and functions.
693 */
694struct rx_queue_attribute {
695 struct attribute attr;
696 ssize_t (*show)(struct netdev_rx_queue *queue,
697 struct rx_queue_attribute *attr, char *buf);
698 ssize_t (*store)(struct netdev_rx_queue *queue,
699 struct rx_queue_attribute *attr, const char *buf, size_t len);
700};
d314774c 701
bf264145
TH
702#ifdef CONFIG_XPS
703/*
704 * This structure holds an XPS map which can be of variable length. The
705 * map is an array of queues.
706 */
707struct xps_map {
708 unsigned int len;
709 unsigned int alloc_len;
710 struct rcu_head rcu;
711 u16 queues[0];
712};
60b778ce 713#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
714#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
715 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
716
717/*
718 * This structure holds all XPS maps for device. Maps are indexed by CPU.
719 */
720struct xps_dev_maps {
721 struct rcu_head rcu;
a4177869 722 struct xps_map __rcu *cpu_map[0];
bf264145 723};
184c449f
AD
724#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
725 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
726#endif /* CONFIG_XPS */
727
4f57c087
JF
728#define TC_MAX_QUEUE 16
729#define TC_BITMASK 15
730/* HW offloaded queuing disciplines txq count and offset maps */
731struct netdev_tc_txq {
732 u16 count;
733 u16 offset;
734};
735
68bad94e
NP
736#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
737/*
738 * This structure is to hold information about the device
739 * configured to run FCoE protocol stack.
740 */
741struct netdev_fcoe_hbainfo {
742 char manufacturer[64];
743 char serial_number[64];
744 char hardware_version[64];
745 char driver_version[64];
746 char optionrom_version[64];
747 char firmware_version[64];
748 char model[256];
749 char model_description[256];
750};
751#endif
752
02637fce 753#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 754
02637fce
JP
755/* This structure holds a unique identifier to identify some
756 * physical item (port for example) used by a netdevice.
66b52b0d 757 */
02637fce
JP
758struct netdev_phys_item_id {
759 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
760 unsigned char id_len;
761};
762
d754f98b
SF
763static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
764 struct netdev_phys_item_id *b)
765{
766 return a->id_len == b->id_len &&
767 memcmp(a->id, b->id, a->id_len) == 0;
768}
769
99932d4f
DB
770typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
771 struct sk_buff *skb);
772
a1b7c5fd 773/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
774 * that are being passed to the netdevice through the setup_tc op.
775 */
776enum {
777 TC_SETUP_MQPRIO,
a1b7c5fd 778 TC_SETUP_CLSU32,
5b33f488 779 TC_SETUP_CLSFLOWER,
b87f7936 780 TC_SETUP_MATCHALL,
332ae8e2 781 TC_SETUP_CLSBPF,
16e5cc64
JF
782};
783
a1b7c5fd
JF
784struct tc_cls_u32_offload;
785
16e5cc64
JF
786struct tc_to_netdev {
787 unsigned int type;
788 union {
789 u8 tc;
a1b7c5fd 790 struct tc_cls_u32_offload *cls_u32;
5b33f488 791 struct tc_cls_flower_offload *cls_flower;
b87f7936 792 struct tc_cls_matchall_offload *cls_mall;
332ae8e2 793 struct tc_cls_bpf_offload *cls_bpf;
16e5cc64
JF
794 };
795};
796
a7862b45
BB
797/* These structures hold the attributes of xdp state that are being passed
798 * to the netdevice through the xdp op.
799 */
800enum xdp_netdev_command {
801 /* Set or clear a bpf program used in the earliest stages of packet
802 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
803 * is responsible for calling bpf_prog_put on any old progs that are
804 * stored. In case of error, the callee need not release the new prog
805 * reference, but on success it takes ownership and must bpf_prog_put
806 * when it is no longer used.
807 */
808 XDP_SETUP_PROG,
809 /* Check if a bpf program is set on the device. The callee should
810 * return true if a program is currently attached and running.
811 */
812 XDP_QUERY_PROG,
813};
814
815struct netdev_xdp {
816 enum xdp_netdev_command command;
817 union {
818 /* XDP_SETUP_PROG */
819 struct bpf_prog *prog;
820 /* XDP_QUERY_PROG */
821 bool prog_attached;
822 };
823};
16e5cc64 824
d314774c
SH
825/*
826 * This structure defines the management hooks for network devices.
00829823
SH
827 * The following hooks can be defined; unless noted otherwise, they are
828 * optional and can be filled with a null pointer.
d314774c
SH
829 *
830 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
831 * This function is called once when a network device is registered.
832 * The network device can use this for any late stage initialization
833 * or semantic validation. It can fail with an error code which will
834 * be propagated back to register_netdev.
d314774c
SH
835 *
836 * void (*ndo_uninit)(struct net_device *dev);
837 * This function is called when device is unregistered or when registration
838 * fails. It is not called if init fails.
839 *
840 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 841 * This function is called when a network device transitions to the up
d314774c
SH
842 * state.
843 *
844 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 845 * This function is called when a network device transitions to the down
d314774c
SH
846 * state.
847 *
dc1f8bf6
SH
848 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
849 * struct net_device *dev);
00829823 850 * Called when a packet needs to be transmitted.
e79d8429
RR
851 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
852 * the queue before that can happen; it's for obsolete devices and weird
853 * corner cases, but the stack really does a non-trivial amount
854 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 855 * Required; cannot be NULL.
00829823 856 *
cdba756f
ED
857 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
858 * netdev_features_t features);
859 * Adjusts the requested feature flags according to device-specific
860 * constraints, and returns the resulting flags. Must not modify
861 * the device state.
862 *
f663dd9a 863 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 864 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 865 * Called to decide which queue to use when device supports multiple
00829823
SH
866 * transmit queues.
867 *
d314774c
SH
868 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
869 * This function is called to allow device receiver to make
5e82b4b2 870 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
871 *
872 * void (*ndo_set_rx_mode)(struct net_device *dev);
873 * This function is called device changes address list filtering.
01789349 874 * If driver handles unicast address filtering, it should set
5e82b4b2 875 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
876 *
877 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
878 * This function is called when the Media Access Control address
37b607c5 879 * needs to be changed. If this interface is not defined, the
5e82b4b2 880 * MAC address can not be changed.
d314774c
SH
881 *
882 * int (*ndo_validate_addr)(struct net_device *dev);
883 * Test if Media Access Control address is valid for the device.
884 *
885 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
886 * Called when a user requests an ioctl which can't be handled by
887 * the generic interface code. If not defined ioctls return
d314774c
SH
888 * not supported error code.
889 *
890 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
891 * Used to set network devices bus interface parameters. This interface
5e82b4b2 892 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
893 * interface (PCI) for low level management.
894 *
895 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
896 * Called when a user wants to change the Maximum Transfer Unit
897 * of a device. If not defined, any request to change MTU will
898 * will return an error.
899 *
00829823 900 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 901 * Callback used when the transmitter has not made any progress
d314774c
SH
902 * for dev->watchdog ticks.
903 *
3cfde79c 904 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 905 * struct rtnl_link_stats64 *storage);
d308e38f 906 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 907 * Called when a user wants to get the network device usage
be1f3c2c 908 * statistics. Drivers must do one of the following:
3cfde79c
BH
909 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
910 * rtnl_link_stats64 structure passed by the caller.
82695d9b 911 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
912 * (which should normally be dev->stats) and return a pointer to
913 * it. The structure may be changed asynchronously only if each
914 * field is written atomically.
915 * 3. Update dev->stats asynchronously and atomically, and define
916 * neither operation.
d314774c 917 *
2c9d85d4
NF
918 * bool (*ndo_has_offload_stats)(int attr_id)
919 * Return true if this device supports offload stats of this attr_id.
920 *
921 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
922 * void *attr_data)
923 * Get statistics for offload operations by attr_id. Write it into the
924 * attr_data pointer.
925 *
5d632cb7 926 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 927 * If device supports VLAN filtering this function is called when a
80d5c368 928 * VLAN id is registered.
d314774c 929 *
5d632cb7 930 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 931 * If device supports VLAN filtering this function is called when a
80d5c368 932 * VLAN id is unregistered.
d314774c
SH
933 *
934 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
935 *
936 * SR-IOV management functions.
937 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
938 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
939 * u8 qos, __be16 proto);
ed616689
SC
940 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
941 * int max_tx_rate);
5f8444a3 942 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 943 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
944 * int (*ndo_get_vf_config)(struct net_device *dev,
945 * int vf, struct ifla_vf_info *ivf);
1d8faf48 946 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
947 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
948 * struct nlattr *port[]);
01a3d796
VZ
949 *
950 * Enable or disable the VF ability to query its RSS Redirection Table and
951 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 952 * with PF and querying it may introduce a theoretical security risk.
01a3d796 953 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 954 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
955 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
956 * Called to setup 'tc' number of traffic classes in the net device. This
957 * is always called from the stack with the rtnl lock held and netif tx
958 * queues stopped. This allows the netdevice to perform queue management
959 * safely.
c445477d 960 *
e9bce845
YZ
961 * Fiber Channel over Ethernet (FCoE) offload functions.
962 * int (*ndo_fcoe_enable)(struct net_device *dev);
963 * Called when the FCoE protocol stack wants to start using LLD for FCoE
964 * so the underlying device can perform whatever needed configuration or
965 * initialization to support acceleration of FCoE traffic.
966 *
967 * int (*ndo_fcoe_disable)(struct net_device *dev);
968 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
969 * so the underlying device can perform whatever needed clean-ups to
970 * stop supporting acceleration of FCoE traffic.
971 *
972 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
973 * struct scatterlist *sgl, unsigned int sgc);
974 * Called when the FCoE Initiator wants to initialize an I/O that
975 * is a possible candidate for Direct Data Placement (DDP). The LLD can
976 * perform necessary setup and returns 1 to indicate the device is set up
977 * successfully to perform DDP on this I/O, otherwise this returns 0.
978 *
979 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
980 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
981 * indicated by the FC exchange id 'xid', so the underlying device can
982 * clean up and reuse resources for later DDP requests.
983 *
984 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
985 * struct scatterlist *sgl, unsigned int sgc);
986 * Called when the FCoE Target wants to initialize an I/O that
987 * is a possible candidate for Direct Data Placement (DDP). The LLD can
988 * perform necessary setup and returns 1 to indicate the device is set up
989 * successfully to perform DDP on this I/O, otherwise this returns 0.
990 *
68bad94e
NP
991 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
992 * struct netdev_fcoe_hbainfo *hbainfo);
993 * Called when the FCoE Protocol stack wants information on the underlying
994 * device. This information is utilized by the FCoE protocol stack to
995 * register attributes with Fiber Channel management service as per the
996 * FC-GS Fabric Device Management Information(FDMI) specification.
997 *
e9bce845
YZ
998 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
999 * Called when the underlying device wants to override default World Wide
1000 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1001 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1002 * protocol stack to use.
1003 *
c445477d
BH
1004 * RFS acceleration.
1005 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1006 * u16 rxq_index, u32 flow_id);
1007 * Set hardware filter for RFS. rxq_index is the target queue index;
1008 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1009 * Return the filter ID on success, or a negative error code.
fbaec0ea 1010 *
8b98a70c 1011 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1012 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1013 * Called to make another netdev an underling.
1014 *
1015 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1016 * Called to release previously enslaved netdev.
5455c699
MM
1017 *
1018 * Feature/offload setting functions.
c8f44aff 1019 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1020 * Called to update device configuration to new features. Passed
1021 * feature set might be less than what was returned by ndo_fix_features()).
1022 * Must return >0 or -errno if it changed dev->features itself.
1023 *
edc7d573 1024 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1025 * struct net_device *dev,
f6f6424b 1026 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1027 * Adds an FDB entry to dev for addr.
1690be63
VY
1028 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1029 * struct net_device *dev,
f6f6424b 1030 * const unsigned char *addr, u16 vid)
77162022
JF
1031 * Deletes the FDB entry from dev coresponding to addr.
1032 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1033 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1034 * int *idx)
77162022
JF
1035 * Used to add FDB entries to dump requests. Implementers should add
1036 * entries to skb and update idx with the number of entries.
e5a55a89 1037 *
ad41faa8
ND
1038 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1039 * u16 flags)
e5a55a89 1040 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1041 * struct net_device *dev, u32 filter_mask,
1042 * int nlflags)
ad41faa8
ND
1043 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1044 * u16 flags);
4bf84c35
JP
1045 *
1046 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1047 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1048 * which do not represent real hardware may define this to allow their
1049 * userspace components to manage their virtual carrier state. Devices
1050 * that determine carrier state from physical hardware properties (eg
1051 * network cables) or protocol-dependent mechanisms (eg
1052 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1053 *
1054 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1055 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1056 * Called to get ID of physical port of this device. If driver does
1057 * not implement this, it is assumed that the hw is not able to have
1058 * multiple net devices on single physical port.
53cf5275 1059 *
7c46a640
AD
1060 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1061 * struct udp_tunnel_info *ti);
1062 * Called by UDP tunnel to notify a driver about the UDP port and socket
1063 * address family that a UDP tunnel is listnening to. It is called only
1064 * when a new port starts listening. The operation is protected by the
1065 * RTNL.
1066 *
1067 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1068 * struct udp_tunnel_info *ti);
1069 * Called by UDP tunnel to notify the driver about a UDP port and socket
1070 * address family that the UDP tunnel is not listening to anymore. The
1071 * operation is protected by the RTNL.
1072 *
a6cc0cfa
JF
1073 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1074 * struct net_device *dev)
1075 * Called by upper layer devices to accelerate switching or other
1076 * station functionality into hardware. 'pdev is the lowerdev
1077 * to use for the offload and 'dev' is the net device that will
1078 * back the offload. Returns a pointer to the private structure
1079 * the upper layer will maintain.
1080 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1081 * Called by upper layer device to delete the station created
1082 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1083 * the station and priv is the structure returned by the add
1084 * operation.
1085 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1086 * struct net_device *dev,
1087 * void *priv);
1088 * Callback to use for xmit over the accelerated station. This
1089 * is used in place of ndo_start_xmit on accelerated net
1090 * devices.
5e82b4b2
BH
1091 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1092 * struct net_device *dev
1093 * netdev_features_t features);
04ffcb25 1094 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1095 * performing offload operations on a given packet. This is to give
1096 * the device an opportunity to implement any restrictions that cannot
1097 * be otherwise expressed by feature flags. The check is called with
1098 * the set of features that the stack has calculated and it returns
1099 * those the driver believes to be appropriate.
822b3b2e
JF
1100 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1101 * int queue_index, u32 maxrate);
1102 * Called when a user wants to set a max-rate limitation of specific
1103 * TX queue.
a54acb3a
ND
1104 * int (*ndo_get_iflink)(const struct net_device *dev);
1105 * Called to get the iflink value of this device.
d746d707 1106 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1107 * bool proto_down);
d746d707
AK
1108 * This function is used to pass protocol port error state information
1109 * to the switch driver. The switch driver can react to the proto_down
1110 * by doing a phys down on the associated switch port.
fc4099f1
PS
1111 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1112 * This function is used to get egress tunnel information for given skb.
1113 * This is useful for retrieving outer tunnel header parameters while
1114 * sampling packet.
871b642a
PA
1115 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1116 * This function is used to specify the headroom that the skb must
1117 * consider when allocation skb during packet reception. Setting
1118 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1119 * forward. Setting a negative value resets the rx headroom to the
871b642a 1120 * default value.
a7862b45
BB
1121 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1122 * This function is used to set or query state related to XDP on the
1123 * netdevice. See definition of enum xdp_netdev_command for details.
d746d707 1124 *
d314774c
SH
1125 */
1126struct net_device_ops {
1127 int (*ndo_init)(struct net_device *dev);
1128 void (*ndo_uninit)(struct net_device *dev);
1129 int (*ndo_open)(struct net_device *dev);
1130 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1131 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1132 struct net_device *dev);
1133 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1134 struct net_device *dev,
1135 netdev_features_t features);
00829823 1136 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1137 struct sk_buff *skb,
99932d4f
DB
1138 void *accel_priv,
1139 select_queue_fallback_t fallback);
d314774c
SH
1140 void (*ndo_change_rx_flags)(struct net_device *dev,
1141 int flags);
d314774c 1142 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1143 int (*ndo_set_mac_address)(struct net_device *dev,
1144 void *addr);
d314774c 1145 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1146 int (*ndo_do_ioctl)(struct net_device *dev,
1147 struct ifreq *ifr, int cmd);
d314774c
SH
1148 int (*ndo_set_config)(struct net_device *dev,
1149 struct ifmap *map);
00829823
SH
1150 int (*ndo_change_mtu)(struct net_device *dev,
1151 int new_mtu);
1152 int (*ndo_neigh_setup)(struct net_device *dev,
1153 struct neigh_parms *);
d314774c
SH
1154 void (*ndo_tx_timeout) (struct net_device *dev);
1155
28172739
ED
1156 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1157 struct rtnl_link_stats64 *storage);
2c9d85d4
NF
1158 bool (*ndo_has_offload_stats)(int attr_id);
1159 int (*ndo_get_offload_stats)(int attr_id,
1160 const struct net_device *dev,
1161 void *attr_data);
d314774c
SH
1162 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1163
8e586137 1164 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1165 __be16 proto, u16 vid);
8e586137 1166 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1167 __be16 proto, u16 vid);
d314774c 1168#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1169 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1170 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1171 struct netpoll_info *info);
0e34e931 1172 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1173#endif
e0d1095a 1174#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1175 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1176#endif
95c26df8
WM
1177 int (*ndo_set_vf_mac)(struct net_device *dev,
1178 int queue, u8 *mac);
1179 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1180 int queue, u16 vlan,
1181 u8 qos, __be16 proto);
ed616689
SC
1182 int (*ndo_set_vf_rate)(struct net_device *dev,
1183 int vf, int min_tx_rate,
1184 int max_tx_rate);
5f8444a3
GR
1185 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1186 int vf, bool setting);
dd461d6a
HS
1187 int (*ndo_set_vf_trust)(struct net_device *dev,
1188 int vf, bool setting);
95c26df8
WM
1189 int (*ndo_get_vf_config)(struct net_device *dev,
1190 int vf,
1191 struct ifla_vf_info *ivf);
1d8faf48
RE
1192 int (*ndo_set_vf_link_state)(struct net_device *dev,
1193 int vf, int link_state);
3b766cd8
EBE
1194 int (*ndo_get_vf_stats)(struct net_device *dev,
1195 int vf,
1196 struct ifla_vf_stats
1197 *vf_stats);
57b61080
SF
1198 int (*ndo_set_vf_port)(struct net_device *dev,
1199 int vf,
1200 struct nlattr *port[]);
1201 int (*ndo_get_vf_port)(struct net_device *dev,
1202 int vf, struct sk_buff *skb);
cc8e27cc
EC
1203 int (*ndo_set_vf_guid)(struct net_device *dev,
1204 int vf, u64 guid,
1205 int guid_type);
01a3d796
VZ
1206 int (*ndo_set_vf_rss_query_en)(
1207 struct net_device *dev,
1208 int vf, bool setting);
16e5cc64
JF
1209 int (*ndo_setup_tc)(struct net_device *dev,
1210 u32 handle,
1211 __be16 protocol,
1212 struct tc_to_netdev *tc);
d11ead75 1213#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1214 int (*ndo_fcoe_enable)(struct net_device *dev);
1215 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1216 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1217 u16 xid,
1218 struct scatterlist *sgl,
1219 unsigned int sgc);
1220 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1221 u16 xid);
6247e086
YZ
1222 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1223 u16 xid,
1224 struct scatterlist *sgl,
1225 unsigned int sgc);
68bad94e
NP
1226 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1227 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1228#endif
1229
d11ead75 1230#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1231#define NETDEV_FCOE_WWNN 0
1232#define NETDEV_FCOE_WWPN 1
1233 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1234 u64 *wwn, int type);
4d288d57 1235#endif
3c9c36bc 1236
c445477d
BH
1237#ifdef CONFIG_RFS_ACCEL
1238 int (*ndo_rx_flow_steer)(struct net_device *dev,
1239 const struct sk_buff *skb,
1240 u16 rxq_index,
1241 u32 flow_id);
1242#endif
fbaec0ea
JP
1243 int (*ndo_add_slave)(struct net_device *dev,
1244 struct net_device *slave_dev);
1245 int (*ndo_del_slave)(struct net_device *dev,
1246 struct net_device *slave_dev);
c8f44aff
MM
1247 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1248 netdev_features_t features);
5455c699 1249 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1250 netdev_features_t features);
503eebc2
JP
1251 int (*ndo_neigh_construct)(struct net_device *dev,
1252 struct neighbour *n);
1253 void (*ndo_neigh_destroy)(struct net_device *dev,
1254 struct neighbour *n);
77162022
JF
1255
1256 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1257 struct nlattr *tb[],
77162022 1258 struct net_device *dev,
6b6e2725 1259 const unsigned char *addr,
f6f6424b 1260 u16 vid,
77162022
JF
1261 u16 flags);
1262 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1263 struct nlattr *tb[],
77162022 1264 struct net_device *dev,
f6f6424b
JP
1265 const unsigned char *addr,
1266 u16 vid);
77162022
JF
1267 int (*ndo_fdb_dump)(struct sk_buff *skb,
1268 struct netlink_callback *cb,
1269 struct net_device *dev,
5d5eacb3 1270 struct net_device *filter_dev,
d297653d 1271 int *idx);
e5a55a89
JF
1272
1273 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1274 struct nlmsghdr *nlh,
1275 u16 flags);
e5a55a89
JF
1276 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1277 u32 pid, u32 seq,
6cbdceeb 1278 struct net_device *dev,
46c264da
ND
1279 u32 filter_mask,
1280 int nlflags);
407af329 1281 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1282 struct nlmsghdr *nlh,
1283 u16 flags);
4bf84c35
JP
1284 int (*ndo_change_carrier)(struct net_device *dev,
1285 bool new_carrier);
66b52b0d 1286 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1287 struct netdev_phys_item_id *ppid);
db24a904
DA
1288 int (*ndo_get_phys_port_name)(struct net_device *dev,
1289 char *name, size_t len);
7c46a640
AD
1290 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1291 struct udp_tunnel_info *ti);
1292 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1293 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1294 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1295 struct net_device *dev);
1296 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1297 void *priv);
1298
1299 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1300 struct net_device *dev,
1301 void *priv);
25175ba5 1302 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1303 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1304 int queue_index,
1305 u32 maxrate);
a54acb3a 1306 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1307 int (*ndo_change_proto_down)(struct net_device *dev,
1308 bool proto_down);
fc4099f1
PS
1309 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1310 struct sk_buff *skb);
871b642a
PA
1311 void (*ndo_set_rx_headroom)(struct net_device *dev,
1312 int needed_headroom);
a7862b45
BB
1313 int (*ndo_xdp)(struct net_device *dev,
1314 struct netdev_xdp *xdp);
d314774c
SH
1315};
1316
7aa98047
LR
1317/**
1318 * enum net_device_priv_flags - &struct net_device priv_flags
1319 *
1320 * These are the &struct net_device, they are only set internally
1321 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1322 * userspace; this means that the order of these flags can change
7aa98047
LR
1323 * during any kernel release.
1324 *
1325 * You should have a pretty good reason to be extending these flags.
1326 *
1327 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1328 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1329 * @IFF_BONDING: bonding master or slave
7aa98047 1330 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1331 * @IFF_WAN_HDLC: WAN HDLC device
1332 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1333 * release skb->dst
1334 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1335 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1336 * @IFF_MACVLAN_PORT: device used as macvlan port
1337 * @IFF_BRIDGE_PORT: device used as bridge port
1338 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1339 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1340 * @IFF_UNICAST_FLT: Supports unicast filtering
1341 * @IFF_TEAM_PORT: device used as team port
1342 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1343 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1344 * change when it's running
1345 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1346 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1347 * underlying stacked devices
1348 * @IFF_IPVLAN_MASTER: IPvlan master device
1349 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1350 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1351 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1352 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1353 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1354 * @IFF_TEAM: device is a team device
d4ab4286 1355 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1356 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1357 * entity (i.e. the master device for bridged veth)
3c175784 1358 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1359 */
1360enum netdev_priv_flags {
1361 IFF_802_1Q_VLAN = 1<<0,
1362 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1363 IFF_BONDING = 1<<2,
1364 IFF_ISATAP = 1<<3,
1365 IFF_WAN_HDLC = 1<<4,
1366 IFF_XMIT_DST_RELEASE = 1<<5,
1367 IFF_DONT_BRIDGE = 1<<6,
1368 IFF_DISABLE_NETPOLL = 1<<7,
1369 IFF_MACVLAN_PORT = 1<<8,
1370 IFF_BRIDGE_PORT = 1<<9,
1371 IFF_OVS_DATAPATH = 1<<10,
1372 IFF_TX_SKB_SHARING = 1<<11,
1373 IFF_UNICAST_FLT = 1<<12,
1374 IFF_TEAM_PORT = 1<<13,
1375 IFF_SUPP_NOFCS = 1<<14,
1376 IFF_LIVE_ADDR_CHANGE = 1<<15,
1377 IFF_MACVLAN = 1<<16,
1378 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1379 IFF_IPVLAN_MASTER = 1<<18,
1380 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1381 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1382 IFF_NO_QUEUE = 1<<21,
1383 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1384 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1385 IFF_TEAM = 1<<24,
d4ab4286 1386 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1387 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1388 IFF_MACSEC = 1<<27,
7aa98047
LR
1389};
1390
1391#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1392#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1393#define IFF_BONDING IFF_BONDING
7aa98047 1394#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1395#define IFF_WAN_HDLC IFF_WAN_HDLC
1396#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1397#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1398#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1399#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1400#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1401#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1402#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1403#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1404#define IFF_TEAM_PORT IFF_TEAM_PORT
1405#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1406#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1407#define IFF_MACVLAN IFF_MACVLAN
02875878 1408#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1409#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1410#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1411#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1412#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1413#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1414#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1415#define IFF_TEAM IFF_TEAM
d4ab4286 1416#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1417#define IFF_MACSEC IFF_MACSEC
7aa98047 1418
536721b1
KK
1419/**
1420 * struct net_device - The DEVICE structure.
1421 * Actually, this whole structure is a big mistake. It mixes I/O
1422 * data with strictly "high-level" data, and it has to know about
1423 * almost every data structure used in the INET module.
1424 *
1425 * @name: This is the first field of the "visible" part of this structure
1426 * (i.e. as seen by users in the "Space.c" file). It is the name
1427 * of the interface.
1428 *
1429 * @name_hlist: Device name hash chain, please keep it close to name[]
1430 * @ifalias: SNMP alias
1431 * @mem_end: Shared memory end
1432 * @mem_start: Shared memory start
1433 * @base_addr: Device I/O address
1434 * @irq: Device IRQ number
1435 *
14ffbbb8
TG
1436 * @carrier_changes: Stats to monitor carrier on<->off transitions
1437 *
536721b1
KK
1438 * @state: Generic network queuing layer state, see netdev_state_t
1439 * @dev_list: The global list of network devices
5e82b4b2
BH
1440 * @napi_list: List entry used for polling NAPI devices
1441 * @unreg_list: List entry when we are unregistering the
1442 * device; see the function unregister_netdev
1443 * @close_list: List entry used when we are closing the device
62d885fe
BP
1444 * @ptype_all: Device-specific packet handlers for all protocols
1445 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1446 *
1447 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1448 * @features: Currently active device features
1449 * @hw_features: User-changeable features
1450 *
1451 * @wanted_features: User-requested features
1452 * @vlan_features: Mask of features inheritable by VLAN devices
1453 *
1454 * @hw_enc_features: Mask of features inherited by encapsulating devices
1455 * This field indicates what encapsulation
1456 * offloads the hardware is capable of doing,
1457 * and drivers will need to set them appropriately.
1458 *
1459 * @mpls_features: Mask of features inheritable by MPLS
1460 *
1461 * @ifindex: interface index
5e82b4b2 1462 * @group: The group the device belongs to
536721b1
KK
1463 *
1464 * @stats: Statistics struct, which was left as a legacy, use
1465 * rtnl_link_stats64 instead
1466 *
1467 * @rx_dropped: Dropped packets by core network,
1468 * do not use this in drivers
1469 * @tx_dropped: Dropped packets by core network,
1470 * do not use this in drivers
6e7333d3
JW
1471 * @rx_nohandler: nohandler dropped packets by core network on
1472 * inactive devices, do not use this in drivers
536721b1 1473 *
536721b1
KK
1474 * @wireless_handlers: List of functions to handle Wireless Extensions,
1475 * instead of ioctl,
1476 * see <net/iw_handler.h> for details.
1477 * @wireless_data: Instance data managed by the core of wireless extensions
1478 *
1479 * @netdev_ops: Includes several pointers to callbacks,
1480 * if one wants to override the ndo_*() functions
1481 * @ethtool_ops: Management operations
f997c55c
AA
1482 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1483 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1484 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1485 * of Layer 2 headers.
1486 *
1487 * @flags: Interface flags (a la BSD)
1488 * @priv_flags: Like 'flags' but invisible to userspace,
1489 * see if.h for the definitions
1490 * @gflags: Global flags ( kept as legacy )
1491 * @padded: How much padding added by alloc_netdev()
1492 * @operstate: RFC2863 operstate
1493 * @link_mode: Mapping policy to operstate
1494 * @if_port: Selectable AUI, TP, ...
1495 * @dma: DMA channel
1496 * @mtu: Interface MTU value
61e84623
JW
1497 * @min_mtu: Interface Minimum MTU value
1498 * @max_mtu: Interface Maximum MTU value
536721b1 1499 * @type: Interface hardware type
2793a23a 1500 * @hard_header_len: Maximum hardware header length.
536721b1
KK
1501 *
1502 * @needed_headroom: Extra headroom the hardware may need, but not in all
1503 * cases can this be guaranteed
1504 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1505 * cases can this be guaranteed. Some cases also use
1506 * LL_MAX_HEADER instead to allocate the skb
1507 *
1508 * interface address info:
1509 *
1510 * @perm_addr: Permanent hw address
1511 * @addr_assign_type: Hw address assignment type
1512 * @addr_len: Hardware address length
8626a0c8 1513 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1514 * @dev_id: Used to differentiate devices that share
1515 * the same link layer address
1516 * @dev_port: Used to differentiate devices that share
1517 * the same function
1518 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1519 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1520 * has been enabled due to the need to listen to
1521 * additional unicast addresses in a device that
1522 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1523 * @uc: unicast mac addresses
1524 * @mc: multicast mac addresses
1525 * @dev_addrs: list of device hw addresses
1526 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1527 * @promiscuity: Number of times the NIC is told to work in
1528 * promiscuous mode; if it becomes 0 the NIC will
1529 * exit promiscuous mode
536721b1
KK
1530 * @allmulti: Counter, enables or disables allmulticast mode
1531 *
1532 * @vlan_info: VLAN info
1533 * @dsa_ptr: dsa specific data
1534 * @tipc_ptr: TIPC specific data
1535 * @atalk_ptr: AppleTalk link
1536 * @ip_ptr: IPv4 specific data
1537 * @dn_ptr: DECnet specific data
1538 * @ip6_ptr: IPv6 specific data
1539 * @ax25_ptr: AX.25 specific data
1540 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1541 *
1542 * @last_rx: Time of last Rx
1543 * @dev_addr: Hw address (before bcast,
1544 * because most packets are unicast)
1545 *
1546 * @_rx: Array of RX queues
1547 * @num_rx_queues: Number of RX queues
1548 * allocated at register_netdev() time
1549 * @real_num_rx_queues: Number of RX queues currently active in device
1550 *
1551 * @rx_handler: handler for received packets
1552 * @rx_handler_data: XXX: need comments on this one
1553 * @ingress_queue: XXX: need comments on this one
1554 * @broadcast: hw bcast address
1555 *
14ffbbb8
TG
1556 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1557 * indexed by RX queue number. Assigned by driver.
1558 * This must only be set if the ndo_rx_flow_steer
1559 * operation is defined
1560 * @index_hlist: Device index hash chain
1561 *
536721b1
KK
1562 * @_tx: Array of TX queues
1563 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1564 * @real_num_tx_queues: Number of TX queues currently active in device
1565 * @qdisc: Root qdisc from userspace point of view
1566 * @tx_queue_len: Max frames per queue allowed
1567 * @tx_global_lock: XXX: need comments on this one
1568 *
1569 * @xps_maps: XXX: need comments on this one
1570 *
536721b1 1571 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1572 * the watchdog (see dev_watchdog())
536721b1
KK
1573 * @watchdog_timer: List of timers
1574 *
1575 * @pcpu_refcnt: Number of references to this device
1576 * @todo_list: Delayed register/unregister
536721b1
KK
1577 * @link_watch_list: XXX: need comments on this one
1578 *
1579 * @reg_state: Register/unregister state machine
1580 * @dismantle: Device is going to be freed
1581 * @rtnl_link_state: This enum represents the phases of creating
1582 * a new link
1583 *
1584 * @destructor: Called from unregister,
1585 * can be used to call free_netdev
1586 * @npinfo: XXX: need comments on this one
1587 * @nd_net: Network namespace this network device is inside
1588 *
1589 * @ml_priv: Mid-layer private
1590 * @lstats: Loopback statistics
1591 * @tstats: Tunnel statistics
1592 * @dstats: Dummy statistics
1593 * @vstats: Virtual ethernet statistics
1594 *
1595 * @garp_port: GARP
1596 * @mrp_port: MRP
1597 *
1598 * @dev: Class/net/name entry
1599 * @sysfs_groups: Space for optional device, statistics and wireless
1600 * sysfs groups
1601 *
1602 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1603 * @rtnl_link_ops: Rtnl_link_ops
1604 *
1605 * @gso_max_size: Maximum size of generic segmentation offload
1606 * @gso_max_segs: Maximum number of segments that can be passed to the
1607 * NIC for GSO
1608 *
1609 * @dcbnl_ops: Data Center Bridging netlink ops
1610 * @num_tc: Number of traffic classes in the net device
1611 * @tc_to_txq: XXX: need comments on this one
1612 * @prio_tc_map XXX: need comments on this one
1613 *
1614 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1615 *
1616 * @priomap: XXX: need comments on this one
1617 * @phydev: Physical device may attach itself
1618 * for hardware timestamping
1619 *
123b3652
ED
1620 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1621 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1622 *
d746d707
AK
1623 * @proto_down: protocol port state information can be sent to the
1624 * switch driver and used to set the phys state of the
1625 * switch port.
1626 *
1da177e4
LT
1627 * FIXME: cleanup struct net_device such that network protocol info
1628 * moves out.
1629 */
1630
d94d9fee 1631struct net_device {
1da177e4 1632 char name[IFNAMSIZ];
9356b8fc 1633 struct hlist_node name_hlist;
0b815a1a 1634 char *ifalias;
1da177e4
LT
1635 /*
1636 * I/O specific fields
1637 * FIXME: Merge these and struct ifmap into one
1638 */
536721b1
KK
1639 unsigned long mem_end;
1640 unsigned long mem_start;
1641 unsigned long base_addr;
1642 int irq;
1da177e4 1643
14ffbbb8
TG
1644 atomic_t carrier_changes;
1645
1da177e4 1646 /*
536721b1
KK
1647 * Some hardware also needs these fields (state,dev_list,
1648 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1649 * part of the usual set specified in Space.c.
1650 */
1651
1da177e4
LT
1652 unsigned long state;
1653
7562f876 1654 struct list_head dev_list;
bea3348e 1655 struct list_head napi_list;
44a0873d 1656 struct list_head unreg_list;
5cde2829 1657 struct list_head close_list;
7866a621
SN
1658 struct list_head ptype_all;
1659 struct list_head ptype_specific;
2f268f12 1660
2f268f12
VF
1661 struct {
1662 struct list_head upper;
1663 struct list_head lower;
1664 } adj_list;
1665
c8f44aff 1666 netdev_features_t features;
c8f44aff 1667 netdev_features_t hw_features;
c8f44aff 1668 netdev_features_t wanted_features;
c8f44aff 1669 netdev_features_t vlan_features;
6a674e9c 1670 netdev_features_t hw_enc_features;
0d89d203 1671 netdev_features_t mpls_features;
802ab55a 1672 netdev_features_t gso_partial_features;
04ed3e74 1673
1da177e4 1674 int ifindex;
7a66bbc9 1675 int group;
1da177e4 1676
c45d286e 1677 struct net_device_stats stats;
015f0688 1678
015f0688
ED
1679 atomic_long_t rx_dropped;
1680 atomic_long_t tx_dropped;
6e7333d3 1681 atomic_long_t rx_nohandler;
1da177e4 1682
b86e0280 1683#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1684 const struct iw_handler_def *wireless_handlers;
1685 struct iw_public_data *wireless_data;
b86e0280 1686#endif
d314774c 1687 const struct net_device_ops *netdev_ops;
76fd8593 1688 const struct ethtool_ops *ethtool_ops;
4170604f 1689#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1690 const struct switchdev_ops *switchdev_ops;
4170604f 1691#endif
1b69c6d0
DA
1692#ifdef CONFIG_NET_L3_MASTER_DEV
1693 const struct l3mdev_ops *l3mdev_ops;
1694#endif
f997c55c
AA
1695#if IS_ENABLED(CONFIG_IPV6)
1696 const struct ndisc_ops *ndisc_ops;
1697#endif
1da177e4 1698
3b04ddde
SH
1699 const struct header_ops *header_ops;
1700
536721b1
KK
1701 unsigned int flags;
1702 unsigned int priv_flags;
1703
1da177e4 1704 unsigned short gflags;
536721b1 1705 unsigned short padded;
1da177e4 1706
536721b1
KK
1707 unsigned char operstate;
1708 unsigned char link_mode;
b00055aa 1709
536721b1
KK
1710 unsigned char if_port;
1711 unsigned char dma;
bdc220da 1712
536721b1 1713 unsigned int mtu;
61e84623
JW
1714 unsigned int min_mtu;
1715 unsigned int max_mtu;
536721b1
KK
1716 unsigned short type;
1717 unsigned short hard_header_len;
1da177e4 1718
f5184d26
JB
1719 unsigned short needed_headroom;
1720 unsigned short needed_tailroom;
1721
1da177e4 1722 /* Interface address info. */
536721b1
KK
1723 unsigned char perm_addr[MAX_ADDR_LEN];
1724 unsigned char addr_assign_type;
1725 unsigned char addr_len;
a0a9663d 1726 unsigned short neigh_priv_len;
536721b1
KK
1727 unsigned short dev_id;
1728 unsigned short dev_port;
ccffad25 1729 spinlock_t addr_list_lock;
14ffbbb8
TG
1730 unsigned char name_assign_type;
1731 bool uc_promisc;
536721b1
KK
1732 struct netdev_hw_addr_list uc;
1733 struct netdev_hw_addr_list mc;
1734 struct netdev_hw_addr_list dev_addrs;
1735
4c3d5e7b
ED
1736#ifdef CONFIG_SYSFS
1737 struct kset *queues_kset;
1738#endif
9d45abe1
WC
1739 unsigned int promiscuity;
1740 unsigned int allmulti;
1da177e4 1741
1da177e4 1742
5e82b4b2 1743 /* Protocol-specific pointers */
65ac6a5f 1744
d11ead75 1745#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1746 struct vlan_info __rcu *vlan_info;
65ac6a5f 1747#endif
34a430d7 1748#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1749 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1750#endif
1751#if IS_ENABLED(CONFIG_TIPC)
536721b1 1752 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1753#endif
536721b1
KK
1754 void *atalk_ptr;
1755 struct in_device __rcu *ip_ptr;
1756 struct dn_dev __rcu *dn_ptr;
1757 struct inet6_dev __rcu *ip6_ptr;
1758 void *ax25_ptr;
1759 struct wireless_dev *ieee80211_ptr;
98a18b6f 1760 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1761#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1762 struct mpls_dev __rcu *mpls_ptr;
1763#endif
1da177e4 1764
9356b8fc 1765/*
cd13539b 1766 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1767 */
536721b1 1768 unsigned long last_rx;
4dc89133 1769
9356b8fc 1770 /* Interface address info used in eth_type_trans() */
536721b1 1771 unsigned char *dev_addr;
f001fde5 1772
a953be53 1773#ifdef CONFIG_SYSFS
0a9627f2
TH
1774 struct netdev_rx_queue *_rx;
1775
0a9627f2 1776 unsigned int num_rx_queues;
62fe0b40 1777 unsigned int real_num_rx_queues;
df334545 1778#endif
0a9627f2 1779
3b47d303 1780 unsigned long gro_flush_timeout;
61391cde 1781 rx_handler_func_t __rcu *rx_handler;
1782 void __rcu *rx_handler_data;
e8a0464c 1783
4cda01e8 1784#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1785 struct tcf_proto __rcu *ingress_cl_list;
1786#endif
24824a09 1787 struct netdev_queue __rcu *ingress_queue;
e687ad60 1788#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1789 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1790#endif
d2788d34 1791
536721b1 1792 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1793#ifdef CONFIG_RFS_ACCEL
1794 struct cpu_rmap *rx_cpu_rmap;
1795#endif
1796 struct hlist_node index_hlist;
cd13539b
ED
1797
1798/*
1799 * Cache lines mostly used on transmit path
1800 */
e8a0464c
DM
1801 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1802 unsigned int num_tx_queues;
fd2ea0a7 1803 unsigned int real_num_tx_queues;
af356afa 1804 struct Qdisc *qdisc;
59cc1f61
JK
1805#ifdef CONFIG_NET_SCHED
1806 DECLARE_HASHTABLE (qdisc_hash, 4);
1807#endif
536721b1 1808 unsigned long tx_queue_len;
c3f26a26 1809 spinlock_t tx_global_lock;
14ffbbb8 1810 int watchdog_timeo;
cd13539b 1811
bf264145 1812#ifdef CONFIG_XPS
a4177869 1813 struct xps_dev_maps __rcu *xps_maps;
bf264145 1814#endif
1f211a1b
DB
1815#ifdef CONFIG_NET_CLS_ACT
1816 struct tcf_proto __rcu *egress_cl_list;
1817#endif
0c4f691f 1818
9356b8fc 1819 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1820 struct timer_list watchdog_timer;
1821
29b4433d 1822 int __percpu *pcpu_refcnt;
1da177e4 1823 struct list_head todo_list;
1da177e4 1824
e014debe 1825 struct list_head link_watch_list;
572a103d 1826
1da177e4 1827 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1828 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1829 NETREG_UNREGISTERING, /* called unregister_netdevice */
1830 NETREG_UNREGISTERED, /* completed unregister todo */
1831 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1832 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1833 } reg_state:8;
1834
536721b1 1835 bool dismantle;
a2835763
PM
1836
1837 enum {
1838 RTNL_LINK_INITIALIZED,
1839 RTNL_LINK_INITIALIZING,
1840 } rtnl_link_state:16;
1da177e4 1841
d314774c 1842 void (*destructor)(struct net_device *dev);
1da177e4 1843
1da177e4 1844#ifdef CONFIG_NETPOLL
5fbee843 1845 struct netpoll_info __rcu *npinfo;
1da177e4 1846#endif
eae792b7 1847
0c5c9fb5 1848 possible_net_t nd_net;
4a1c5371 1849
4951704b 1850 /* mid-layer private */
a7855c78 1851 union {
536721b1
KK
1852 void *ml_priv;
1853 struct pcpu_lstats __percpu *lstats;
8f84985f 1854 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1855 struct pcpu_dstats __percpu *dstats;
1856 struct pcpu_vstats __percpu *vstats;
a7855c78 1857 };
536721b1 1858
3cc77ec7 1859 struct garp_port __rcu *garp_port;
febf018d 1860 struct mrp_port __rcu *mrp_port;
1da177e4 1861
5e82b4b2 1862 struct device dev;
0c509a6c 1863 const struct attribute_group *sysfs_groups[4];
a953be53 1864 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1865
38f7b870 1866 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1867
82cc1a7a
PWJ
1868 /* for setting kernel sock attribute on TCP connection setup */
1869#define GSO_MAX_SIZE 65536
1870 unsigned int gso_max_size;
30b678d8
BH
1871#define GSO_MAX_SEGS 65535
1872 u16 gso_max_segs;
743b03a8 1873
7a6b6f51 1874#ifdef CONFIG_DCB
32953543 1875 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1876#endif
5e82b4b2
BH
1877 u8 num_tc;
1878 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1879 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1880
d11ead75 1881#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1882 unsigned int fcoe_ddp_xid;
5bc1421e 1883#endif
86f8515f 1884#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1885 struct netprio_map __rcu *priomap;
4d288d57 1886#endif
5e82b4b2
BH
1887 struct phy_device *phydev;
1888 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1889 struct lock_class_key *qdisc_running_key;
5e82b4b2 1890 bool proto_down;
1da177e4 1891};
43cb76d9 1892#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1893
1894#define NETDEV_ALIGN 32
1da177e4 1895
4f57c087
JF
1896static inline
1897int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1898{
1899 return dev->prio_tc_map[prio & TC_BITMASK];
1900}
1901
1902static inline
1903int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1904{
1905 if (tc >= dev->num_tc)
1906 return -EINVAL;
1907
1908 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1909 return 0;
1910}
1911
8d059b0f 1912int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1913void netdev_reset_tc(struct net_device *dev);
1914int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1915int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1916
1917static inline
1918int netdev_get_num_tc(struct net_device *dev)
1919{
1920 return dev->num_tc;
1921}
1922
e8a0464c
DM
1923static inline
1924struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1925 unsigned int index)
1926{
1927 return &dev->_tx[index];
1928}
1929
10c51b56
DB
1930static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1931 const struct sk_buff *skb)
1932{
1933 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1934}
1935
e8a0464c
DM
1936static inline void netdev_for_each_tx_queue(struct net_device *dev,
1937 void (*f)(struct net_device *,
1938 struct netdev_queue *,
1939 void *),
1940 void *arg)
1941{
1942 unsigned int i;
1943
1944 for (i = 0; i < dev->num_tx_queues; i++)
1945 f(dev, &dev->_tx[i], arg);
1946}
1947
d3fff6c4
ED
1948#define netdev_lockdep_set_classes(dev) \
1949{ \
1950 static struct lock_class_key qdisc_tx_busylock_key; \
1951 static struct lock_class_key qdisc_running_key; \
1952 static struct lock_class_key qdisc_xmit_lock_key; \
1953 static struct lock_class_key dev_addr_list_lock_key; \
1954 unsigned int i; \
1955 \
1956 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1957 (dev)->qdisc_running_key = &qdisc_running_key; \
1958 lockdep_set_class(&(dev)->addr_list_lock, \
1959 &dev_addr_list_lock_key); \
1960 for (i = 0; i < (dev)->num_tx_queues; i++) \
1961 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1962 &qdisc_xmit_lock_key); \
1963}
1964
f629d208 1965struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1966 struct sk_buff *skb,
1967 void *accel_priv);
8c4c49df 1968
871b642a
PA
1969/* returns the headroom that the master device needs to take in account
1970 * when forwarding to this dev
1971 */
1972static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1973{
1974 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1975}
1976
1977static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1978{
1979 if (dev->netdev_ops->ndo_set_rx_headroom)
1980 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1981}
1982
1983/* set the device rx headroom to the dev's default */
1984static inline void netdev_reset_rx_headroom(struct net_device *dev)
1985{
1986 netdev_set_rx_headroom(dev, -1);
1987}
1988
c346dca1
YH
1989/*
1990 * Net namespace inlines
1991 */
1992static inline
1993struct net *dev_net(const struct net_device *dev)
1994{
c2d9ba9b 1995 return read_pnet(&dev->nd_net);
c346dca1
YH
1996}
1997
1998static inline
f5aa23fd 1999void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2000{
0c5c9fb5 2001 write_pnet(&dev->nd_net, net);
c346dca1
YH
2002}
2003
3e8a72d1 2004static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 2005{
3fc88677 2006#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
2007 if (dev->dsa_ptr != NULL)
2008 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 2009#endif
5aed85ce 2010 return false;
396138f0
LB
2011}
2012
bea3348e
SH
2013/**
2014 * netdev_priv - access network device private data
2015 * @dev: network device
2016 *
2017 * Get network device private data
2018 */
6472ce60 2019static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2020{
1ce8e7b5 2021 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2022}
2023
1da177e4
LT
2024/* Set the sysfs physical device reference for the network logical device
2025 * if set prior to registration will cause a symlink during initialization.
2026 */
43cb76d9 2027#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2028
384912ed 2029/* Set the sysfs device type for the network logical device to allow
3f79410c 2030 * fine-grained identification of different network device types. For
5e82b4b2 2031 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2032 */
2033#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2034
82dc3c63
ED
2035/* Default NAPI poll() weight
2036 * Device drivers are strongly advised to not use bigger value
2037 */
2038#define NAPI_POLL_WEIGHT 64
2039
3b582cc1 2040/**
5e82b4b2 2041 * netif_napi_add - initialize a NAPI context
3b582cc1 2042 * @dev: network device
5e82b4b2 2043 * @napi: NAPI context
3b582cc1
SH
2044 * @poll: polling function
2045 * @weight: default weight
2046 *
5e82b4b2
BH
2047 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2048 * *any* of the other NAPI-related functions.
3b582cc1 2049 */
d565b0a1
HX
2050void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2051 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2052
d64b5e85 2053/**
5e82b4b2 2054 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2055 * @dev: network device
5e82b4b2 2056 * @napi: NAPI context
d64b5e85
ED
2057 * @poll: polling function
2058 * @weight: default weight
2059 *
2060 * This variant of netif_napi_add() should be used from drivers using NAPI
2061 * to exclusively poll a TX queue.
2062 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2063 */
2064static inline void netif_tx_napi_add(struct net_device *dev,
2065 struct napi_struct *napi,
2066 int (*poll)(struct napi_struct *, int),
2067 int weight)
2068{
2069 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2070 netif_napi_add(dev, napi, poll, weight);
2071}
2072
d8156534 2073/**
5e82b4b2
BH
2074 * netif_napi_del - remove a NAPI context
2075 * @napi: NAPI context
d8156534 2076 *
5e82b4b2 2077 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2078 */
d565b0a1
HX
2079void netif_napi_del(struct napi_struct *napi);
2080
2081struct napi_gro_cb {
78a478d0 2082 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2083 void *frag0;
78a478d0 2084
7489594c
HX
2085 /* Length of frag0. */
2086 unsigned int frag0_len;
2087
86911732 2088 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2089 int data_offset;
86911732 2090
d565b0a1 2091 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2092 u16 flush;
2093
2094 /* Save the IP ID here and check when we get to the transport layer */
2095 u16 flush_id;
d565b0a1
HX
2096
2097 /* Number of segments aggregated. */
2e71a6f8
ED
2098 u16 count;
2099
15e2396d
TH
2100 /* Start offset for remote checksum offload */
2101 u16 gro_remcsum_start;
2102
2e71a6f8
ED
2103 /* jiffies when first packet was created/queued */
2104 unsigned long age;
86347245 2105
afe93325 2106 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2107 u16 proto;
2108
baa32ff4
TH
2109 /* This is non-zero if the packet may be of the same flow. */
2110 u8 same_flow:1;
2111
fac8e0f5
JG
2112 /* Used in tunnel GRO receive */
2113 u8 encap_mark:1;
573e8fca
TH
2114
2115 /* GRO checksum is valid */
2116 u8 csum_valid:1;
2117
662880f4
TH
2118 /* Number of checksums via CHECKSUM_UNNECESSARY */
2119 u8 csum_cnt:3;
c3c7c254 2120
baa32ff4
TH
2121 /* Free the skb? */
2122 u8 free:2;
2123#define NAPI_GRO_FREE 1
2124#define NAPI_GRO_FREE_STOLEN_HEAD 2
2125
efc98d08
TH
2126 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2127 u8 is_ipv6:1;
2128
a0ca153f
AD
2129 /* Used in GRE, set in fou/gue_gro_receive */
2130 u8 is_fou:1;
2131
1530545e
AD
2132 /* Used to determine if flush_id can be ignored */
2133 u8 is_atomic:1;
2134
fcd91dd4
SD
2135 /* Number of gro_receive callbacks this packet already went through */
2136 u8 recursion_counter:4;
2137
2138 /* 1 bit hole */
baa32ff4 2139
bf5a755f
JC
2140 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2141 __wsum csum;
2142
c3c7c254
ED
2143 /* used in skb_gro_receive() slow path */
2144 struct sk_buff *last;
d565b0a1
HX
2145};
2146
2147#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2148
fcd91dd4
SD
2149#define GRO_RECURSION_LIMIT 15
2150static inline int gro_recursion_inc_test(struct sk_buff *skb)
2151{
2152 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2153}
2154
2155typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2156static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2157 struct sk_buff **head,
2158 struct sk_buff *skb)
2159{
2160 if (unlikely(gro_recursion_inc_test(skb))) {
2161 NAPI_GRO_CB(skb)->flush |= 1;
2162 return NULL;
2163 }
2164
2165 return cb(head, skb);
2166}
2167
2168typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2169 struct sk_buff *);
2170static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2171 struct sock *sk,
2172 struct sk_buff **head,
2173 struct sk_buff *skb)
2174{
2175 if (unlikely(gro_recursion_inc_test(skb))) {
2176 NAPI_GRO_CB(skb)->flush |= 1;
2177 return NULL;
2178 }
2179
2180 return cb(sk, head, skb);
2181}
2182
1da177e4 2183struct packet_type {
f2ccd8fa
DM
2184 __be16 type; /* This is really htons(ether_type). */
2185 struct net_device *dev; /* NULL is wildcarded here */
2186 int (*func) (struct sk_buff *,
2187 struct net_device *,
2188 struct packet_type *,
2189 struct net_device *);
c0de08d0
EL
2190 bool (*id_match)(struct packet_type *ptype,
2191 struct sock *sk);
1da177e4
LT
2192 void *af_packet_priv;
2193 struct list_head list;
2194};
2195
f191a1d1 2196struct offload_callbacks {
576a30eb 2197 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2198 netdev_features_t features);
d565b0a1 2199 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2200 struct sk_buff *skb);
299603e8 2201 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2202};
2203
2204struct packet_offload {
2205 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2206 u16 priority;
f191a1d1
VY
2207 struct offload_callbacks callbacks;
2208 struct list_head list;
1da177e4
LT
2209};
2210
5e82b4b2 2211/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2212struct pcpu_sw_netstats {
2213 u64 rx_packets;
2214 u64 rx_bytes;
2215 u64 tx_packets;
2216 u64 tx_bytes;
2217 struct u64_stats_sync syncp;
2218};
2219
aabc92bb
PNA
2220#define __netdev_alloc_pcpu_stats(type, gfp) \
2221({ \
2222 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2223 if (pcpu_stats) { \
2224 int __cpu; \
2225 for_each_possible_cpu(__cpu) { \
2226 typeof(type) *stat; \
2227 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2228 u64_stats_init(&stat->syncp); \
2229 } \
2230 } \
2231 pcpu_stats; \
1c213bd2
WC
2232})
2233
aabc92bb 2234#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2235 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2236
764f5e54
JP
2237enum netdev_lag_tx_type {
2238 NETDEV_LAG_TX_TYPE_UNKNOWN,
2239 NETDEV_LAG_TX_TYPE_RANDOM,
2240 NETDEV_LAG_TX_TYPE_BROADCAST,
2241 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2242 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2243 NETDEV_LAG_TX_TYPE_HASH,
2244};
2245
2246struct netdev_lag_upper_info {
2247 enum netdev_lag_tx_type tx_type;
2248};
2249
fb1b2e3c
JP
2250struct netdev_lag_lower_state_info {
2251 u8 link_up : 1,
2252 tx_enabled : 1;
2253};
2254
1da177e4
LT
2255#include <linux/notifier.h>
2256
dcfe1421
AW
2257/* netdevice notifier chain. Please remember to update the rtnetlink
2258 * notification exclusion list in rtnetlink_event() when adding new
2259 * types.
2260 */
2261#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2262#define NETDEV_DOWN 0x0002
2263#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2264 detected a hardware crash and restarted
2265 - we can use this eg to kick tcp sessions
2266 once done */
2267#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2268#define NETDEV_REGISTER 0x0005
2269#define NETDEV_UNREGISTER 0x0006
1d486bfb 2270#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2271#define NETDEV_CHANGEADDR 0x0008
2272#define NETDEV_GOING_DOWN 0x0009
2273#define NETDEV_CHANGENAME 0x000A
2274#define NETDEV_FEAT_CHANGE 0x000B
2275#define NETDEV_BONDING_FAILOVER 0x000C
2276#define NETDEV_PRE_UP 0x000D
2277#define NETDEV_PRE_TYPE_CHANGE 0x000E
2278#define NETDEV_POST_TYPE_CHANGE 0x000F
2279#define NETDEV_POST_INIT 0x0010
0115e8e3 2280#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2281#define NETDEV_RELEASE 0x0012
2282#define NETDEV_NOTIFY_PEERS 0x0013
2283#define NETDEV_JOIN 0x0014
42e52bf9 2284#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2285#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2286#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2287#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2288#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2289#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2290#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2291#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2292#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2293
f629d208
JP
2294int register_netdevice_notifier(struct notifier_block *nb);
2295int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2296
2297struct netdev_notifier_info {
2298 struct net_device *dev;
2299};
2300
be9efd36
JP
2301struct netdev_notifier_change_info {
2302 struct netdev_notifier_info info; /* must be first */
2303 unsigned int flags_changed;
2304};
2305
0e4ead9d
JP
2306struct netdev_notifier_changeupper_info {
2307 struct netdev_notifier_info info; /* must be first */
2308 struct net_device *upper_dev; /* new upper dev */
2309 bool master; /* is upper dev master */
5e82b4b2 2310 bool linking; /* is the notification for link or unlink */
29bf24af 2311 void *upper_info; /* upper dev info */
0e4ead9d
JP
2312};
2313
04d48266
JP
2314struct netdev_notifier_changelowerstate_info {
2315 struct netdev_notifier_info info; /* must be first */
2316 void *lower_state_info; /* is lower dev state */
2317};
2318
75538c2b
CW
2319static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2320 struct net_device *dev)
2321{
2322 info->dev = dev;
2323}
2324
351638e7
JP
2325static inline struct net_device *
2326netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2327{
2328 return info->dev;
2329}
2330
f629d208 2331int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2332
2333
1da177e4
LT
2334extern rwlock_t dev_base_lock; /* Device list lock */
2335
881d966b
EB
2336#define for_each_netdev(net, d) \
2337 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2338#define for_each_netdev_reverse(net, d) \
2339 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2340#define for_each_netdev_rcu(net, d) \
2341 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2342#define for_each_netdev_safe(net, d, n) \
2343 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2344#define for_each_netdev_continue(net, d) \
2345 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2346#define for_each_netdev_continue_rcu(net, d) \
2347 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2348#define for_each_netdev_in_bond_rcu(bond, slave) \
2349 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2350 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2351#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2352
a050c33f
DL
2353static inline struct net_device *next_net_device(struct net_device *dev)
2354{
2355 struct list_head *lh;
2356 struct net *net;
2357
c346dca1 2358 net = dev_net(dev);
a050c33f
DL
2359 lh = dev->dev_list.next;
2360 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2361}
2362
ce81b76a
ED
2363static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2364{
2365 struct list_head *lh;
2366 struct net *net;
2367
2368 net = dev_net(dev);
ccf43438 2369 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2370 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2371}
2372
a050c33f
DL
2373static inline struct net_device *first_net_device(struct net *net)
2374{
2375 return list_empty(&net->dev_base_head) ? NULL :
2376 net_device_entry(net->dev_base_head.next);
2377}
7562f876 2378
ccf43438
ED
2379static inline struct net_device *first_net_device_rcu(struct net *net)
2380{
2381 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2382
2383 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2384}
2385
f629d208
JP
2386int netdev_boot_setup_check(struct net_device *dev);
2387unsigned long netdev_boot_base(const char *prefix, int unit);
2388struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2389 const char *hwaddr);
2390struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2391struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2392void dev_add_pack(struct packet_type *pt);
2393void dev_remove_pack(struct packet_type *pt);
2394void __dev_remove_pack(struct packet_type *pt);
2395void dev_add_offload(struct packet_offload *po);
2396void dev_remove_offload(struct packet_offload *po);
f629d208 2397
a54acb3a 2398int dev_get_iflink(const struct net_device *dev);
fc4099f1 2399int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2400struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2401 unsigned short mask);
f629d208
JP
2402struct net_device *dev_get_by_name(struct net *net, const char *name);
2403struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2404struct net_device *__dev_get_by_name(struct net *net, const char *name);
2405int dev_alloc_name(struct net_device *dev, const char *name);
2406int dev_open(struct net_device *dev);
2407int dev_close(struct net_device *dev);
99c4a26a 2408int dev_close_many(struct list_head *head, bool unlink);
f629d208 2409void dev_disable_lro(struct net_device *dev);
0c4b51f0 2410int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2411int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2412int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2413int register_netdevice(struct net_device *dev);
2414void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2415void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2416static inline void unregister_netdevice(struct net_device *dev)
2417{
2418 unregister_netdevice_queue(dev, NULL);
2419}
2420
f629d208
JP
2421int netdev_refcnt_read(const struct net_device *dev);
2422void free_netdev(struct net_device *dev);
74d332c1 2423void netdev_freemem(struct net_device *dev);
f629d208
JP
2424void synchronize_net(void);
2425int init_dummy_netdev(struct net_device *dev);
937f1ba5 2426
f60e5990 2427DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2428#define XMIT_RECURSION_LIMIT 10
2429
f60e5990 2430static inline int dev_recursion_level(void)
2431{
2432 return this_cpu_read(xmit_recursion);
2433}
2434
f629d208
JP
2435struct net_device *dev_get_by_index(struct net *net, int ifindex);
2436struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2437struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2438int netdev_get_name(struct net *net, char *name, int ifindex);
2439int dev_restart(struct net_device *dev);
f629d208 2440int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2441
2442static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2443{
2444 return NAPI_GRO_CB(skb)->data_offset;
2445}
2446
2447static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2448{
2449 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2450}
2451
2452static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2453{
2454 NAPI_GRO_CB(skb)->data_offset += len;
2455}
2456
a5b1cf28
HX
2457static inline void *skb_gro_header_fast(struct sk_buff *skb,
2458 unsigned int offset)
86911732 2459{
a5b1cf28
HX
2460 return NAPI_GRO_CB(skb)->frag0 + offset;
2461}
78a478d0 2462
a5b1cf28
HX
2463static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2464{
2465 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2466}
78a478d0 2467
a5b1cf28
HX
2468static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2469 unsigned int offset)
2470{
17dd759c
HX
2471 if (!pskb_may_pull(skb, hlen))
2472 return NULL;
2473
a5b1cf28
HX
2474 NAPI_GRO_CB(skb)->frag0 = NULL;
2475 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2476 return skb->data + offset;
86911732 2477}
1da177e4 2478
36e7b1b8
HX
2479static inline void *skb_gro_network_header(struct sk_buff *skb)
2480{
78d3fd0b
HX
2481 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2482 skb_network_offset(skb);
36e7b1b8
HX
2483}
2484
bf5a755f
JC
2485static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2486 const void *start, unsigned int len)
2487{
573e8fca 2488 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2489 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2490 csum_partial(start, len, 0));
2491}
2492
573e8fca
TH
2493/* GRO checksum functions. These are logical equivalents of the normal
2494 * checksum functions (in skbuff.h) except that they operate on the GRO
2495 * offsets and fields in sk_buff.
2496 */
2497
2498__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2499
15e2396d
TH
2500static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2501{
b7fe10e5 2502 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2503}
2504
573e8fca
TH
2505static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2506 bool zero_okay,
2507 __sum16 check)
2508{
6edec0e6
TH
2509 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2510 skb_checksum_start_offset(skb) <
2511 skb_gro_offset(skb)) &&
15e2396d 2512 !skb_at_gro_remcsum_start(skb) &&
662880f4 2513 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2514 (!zero_okay || check));
2515}
2516
2517static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2518 __wsum psum)
2519{
2520 if (NAPI_GRO_CB(skb)->csum_valid &&
2521 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2522 return 0;
2523
2524 NAPI_GRO_CB(skb)->csum = psum;
2525
2526 return __skb_gro_checksum_complete(skb);
2527}
2528
573e8fca
TH
2529static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2530{
662880f4
TH
2531 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2532 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2533 NAPI_GRO_CB(skb)->csum_cnt--;
2534 } else {
2535 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2536 * verified a new top level checksum or an encapsulated one
2537 * during GRO. This saves work if we fallback to normal path.
2538 */
2539 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2540 }
2541}
2542
2543#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2544 compute_pseudo) \
2545({ \
2546 __sum16 __ret = 0; \
2547 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2548 __ret = __skb_gro_checksum_validate_complete(skb, \
2549 compute_pseudo(skb, proto)); \
5a212329
TH
2550 if (__ret) \
2551 __skb_mark_checksum_bad(skb); \
2552 else \
573e8fca
TH
2553 skb_gro_incr_csum_unnecessary(skb); \
2554 __ret; \
2555})
2556
2557#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2558 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2559
2560#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2561 compute_pseudo) \
2562 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2563
2564#define skb_gro_checksum_simple_validate(skb) \
2565 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2566
d96535a1
TH
2567static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2568{
2569 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2570 !NAPI_GRO_CB(skb)->csum_valid);
2571}
2572
2573static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2574 __sum16 check, __wsum pseudo)
2575{
2576 NAPI_GRO_CB(skb)->csum = ~pseudo;
2577 NAPI_GRO_CB(skb)->csum_valid = 1;
2578}
2579
2580#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2581do { \
2582 if (__skb_gro_checksum_convert_check(skb)) \
2583 __skb_gro_checksum_convert(skb, check, \
2584 compute_pseudo(skb, proto)); \
2585} while (0)
2586
26c4f7da
TH
2587struct gro_remcsum {
2588 int offset;
2589 __wsum delta;
2590};
2591
2592static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2593{
846cd667 2594 grc->offset = 0;
26c4f7da
TH
2595 grc->delta = 0;
2596}
2597
b7fe10e5
TH
2598static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2599 unsigned int off, size_t hdrlen,
2600 int start, int offset,
2601 struct gro_remcsum *grc,
2602 bool nopartial)
dcdc8994
TH
2603{
2604 __wsum delta;
b7fe10e5 2605 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2606
2607 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2608
15e2396d 2609 if (!nopartial) {
b7fe10e5
TH
2610 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2611 return ptr;
2612 }
2613
2614 ptr = skb_gro_header_fast(skb, off);
2615 if (skb_gro_header_hard(skb, off + plen)) {
2616 ptr = skb_gro_header_slow(skb, off + plen, off);
2617 if (!ptr)
2618 return NULL;
15e2396d
TH
2619 }
2620
b7fe10e5
TH
2621 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2622 start, offset);
dcdc8994
TH
2623
2624 /* Adjust skb->csum since we changed the packet */
dcdc8994 2625 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2626
b7fe10e5 2627 grc->offset = off + hdrlen + offset;
26c4f7da 2628 grc->delta = delta;
b7fe10e5
TH
2629
2630 return ptr;
dcdc8994
TH
2631}
2632
26c4f7da
TH
2633static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2634 struct gro_remcsum *grc)
2635{
b7fe10e5
TH
2636 void *ptr;
2637 size_t plen = grc->offset + sizeof(u16);
2638
26c4f7da
TH
2639 if (!grc->delta)
2640 return;
2641
b7fe10e5
TH
2642 ptr = skb_gro_header_fast(skb, grc->offset);
2643 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2644 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2645 if (!ptr)
2646 return;
2647 }
2648
2649 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2650}
dcdc8994 2651
0c4e8581
SH
2652static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2653 unsigned short type,
3b04ddde 2654 const void *daddr, const void *saddr,
95c96174 2655 unsigned int len)
0c4e8581 2656{
f1ecfd5d 2657 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2658 return 0;
3b04ddde
SH
2659
2660 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2661}
2662
b95cce35
SH
2663static inline int dev_parse_header(const struct sk_buff *skb,
2664 unsigned char *haddr)
2665{
2666 const struct net_device *dev = skb->dev;
2667
1b83336b 2668 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2669 return 0;
3b04ddde 2670 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2671}
2672
2793a23a
WB
2673/* ll_header must have at least hard_header_len allocated */
2674static inline bool dev_validate_header(const struct net_device *dev,
2675 char *ll_header, int len)
2676{
2677 if (likely(len >= dev->hard_header_len))
2678 return true;
2679
2680 if (capable(CAP_SYS_RAWIO)) {
2681 memset(ll_header + len, 0, dev->hard_header_len - len);
2682 return true;
2683 }
2684
2685 if (dev->header_ops && dev->header_ops->validate)
2686 return dev->header_ops->validate(ll_header, len);
2687
2688 return false;
2689}
2690
1da177e4 2691typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2692int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2693static inline int unregister_gifconf(unsigned int family)
2694{
2695 return register_gifconf(family, NULL);
2696}
2697
99bbc707 2698#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2699#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2700struct sd_flow_limit {
2701 u64 count;
2702 unsigned int num_buckets;
2703 unsigned int history_head;
2704 u16 history[FLOW_LIMIT_HISTORY];
2705 u8 buckets[];
2706};
2707
2708extern int netdev_flow_limit_table_len;
2709#endif /* CONFIG_NET_FLOW_LIMIT */
2710
1da177e4 2711/*
5e82b4b2 2712 * Incoming packets are placed on per-CPU queues
1da177e4 2713 */
d94d9fee 2714struct softnet_data {
1da177e4 2715 struct list_head poll_list;
6e7676c1 2716 struct sk_buff_head process_queue;
1da177e4 2717
dee42870 2718 /* stats */
cd7b5396
DM
2719 unsigned int processed;
2720 unsigned int time_squeeze;
cd7b5396 2721 unsigned int received_rps;
fd793d89 2722#ifdef CONFIG_RPS
88751275 2723 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2724#endif
2725#ifdef CONFIG_NET_FLOW_LIMIT
2726 struct sd_flow_limit __rcu *flow_limit;
2727#endif
2728 struct Qdisc *output_queue;
2729 struct Qdisc **output_queue_tailp;
2730 struct sk_buff *completion_queue;
88751275 2731
4cdb1e2e 2732#ifdef CONFIG_RPS
501e7ef5
ED
2733 /* input_queue_head should be written by cpu owning this struct,
2734 * and only read by other cpus. Worth using a cache line.
2735 */
2736 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2737
2738 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2739 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2740 struct softnet_data *rps_ipi_next;
2741 unsigned int cpu;
76cc8b13 2742 unsigned int input_queue_tail;
1e94d72f 2743#endif
95c96174 2744 unsigned int dropped;
0a9627f2 2745 struct sk_buff_head input_pkt_queue;
bea3348e 2746 struct napi_struct backlog;
99bbc707 2747
1da177e4
LT
2748};
2749
76cc8b13 2750static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2751{
2752#ifdef CONFIG_RPS
76cc8b13
TH
2753 sd->input_queue_head++;
2754#endif
2755}
2756
2757static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2758 unsigned int *qtail)
2759{
2760#ifdef CONFIG_RPS
2761 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2762#endif
2763}
2764
0a9627f2 2765DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2766
f629d208 2767void __netif_schedule(struct Qdisc *q);
46e5da40 2768void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2769
fd2ea0a7
DM
2770static inline void netif_tx_schedule_all(struct net_device *dev)
2771{
2772 unsigned int i;
2773
2774 for (i = 0; i < dev->num_tx_queues; i++)
2775 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2776}
2777
f9a7cbbf 2778static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2779{
73466498 2780 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2781}
2782
bea3348e
SH
2783/**
2784 * netif_start_queue - allow transmit
2785 * @dev: network device
2786 *
2787 * Allow upper layers to call the device hard_start_xmit routine.
2788 */
1da177e4
LT
2789static inline void netif_start_queue(struct net_device *dev)
2790{
e8a0464c 2791 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2792}
2793
fd2ea0a7
DM
2794static inline void netif_tx_start_all_queues(struct net_device *dev)
2795{
2796 unsigned int i;
2797
2798 for (i = 0; i < dev->num_tx_queues; i++) {
2799 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2800 netif_tx_start_queue(txq);
2801 }
2802}
2803
46e5da40 2804void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2805
d29f749e
DJ
2806/**
2807 * netif_wake_queue - restart transmit
2808 * @dev: network device
2809 *
2810 * Allow upper layers to call the device hard_start_xmit routine.
2811 * Used for flow control when transmit resources are available.
2812 */
79d16385
DM
2813static inline void netif_wake_queue(struct net_device *dev)
2814{
e8a0464c 2815 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2816}
2817
fd2ea0a7
DM
2818static inline void netif_tx_wake_all_queues(struct net_device *dev)
2819{
2820 unsigned int i;
2821
2822 for (i = 0; i < dev->num_tx_queues; i++) {
2823 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2824 netif_tx_wake_queue(txq);
2825 }
2826}
2827
f9a7cbbf 2828static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2829{
73466498 2830 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2831}
2832
bea3348e
SH
2833/**
2834 * netif_stop_queue - stop transmitted packets
2835 * @dev: network device
2836 *
2837 * Stop upper layers calling the device hard_start_xmit routine.
2838 * Used for flow control when transmit resources are unavailable.
2839 */
1da177e4
LT
2840static inline void netif_stop_queue(struct net_device *dev)
2841{
e8a0464c 2842 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2843}
2844
a2029240 2845void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2846
4d29515f 2847static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2848{
73466498 2849 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2850}
2851
bea3348e
SH
2852/**
2853 * netif_queue_stopped - test if transmit queue is flowblocked
2854 * @dev: network device
2855 *
2856 * Test if transmit queue on device is currently unable to send.
2857 */
4d29515f 2858static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2859{
e8a0464c 2860 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2861}
2862
4d29515f 2863static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2864{
73466498
TH
2865 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2866}
2867
8e2f1a63
DB
2868static inline bool
2869netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2870{
2871 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2872}
2873
8e2f1a63
DB
2874static inline bool
2875netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2876{
2877 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2878}
2879
53511453
ED
2880/**
2881 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2882 * @dev_queue: pointer to transmit queue
2883 *
2884 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2885 * to give appropriate hint to the CPU.
53511453
ED
2886 */
2887static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2888{
2889#ifdef CONFIG_BQL
2890 prefetchw(&dev_queue->dql.num_queued);
2891#endif
2892}
2893
2894/**
2895 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2896 * @dev_queue: pointer to transmit queue
2897 *
2898 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2899 * to give appropriate hint to the CPU.
53511453
ED
2900 */
2901static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2902{
2903#ifdef CONFIG_BQL
2904 prefetchw(&dev_queue->dql.limit);
2905#endif
2906}
2907
c5d67bd7
TH
2908static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2909 unsigned int bytes)
2910{
114cf580
TH
2911#ifdef CONFIG_BQL
2912 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2913
2914 if (likely(dql_avail(&dev_queue->dql) >= 0))
2915 return;
2916
2917 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2918
2919 /*
2920 * The XOFF flag must be set before checking the dql_avail below,
2921 * because in netdev_tx_completed_queue we update the dql_completed
2922 * before checking the XOFF flag.
2923 */
2924 smp_mb();
2925
2926 /* check again in case another CPU has just made room avail */
2927 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2928 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2929#endif
c5d67bd7
TH
2930}
2931
0042d0c8
FF
2932/**
2933 * netdev_sent_queue - report the number of bytes queued to hardware
2934 * @dev: network device
2935 * @bytes: number of bytes queued to the hardware device queue
2936 *
2937 * Report the number of bytes queued for sending/completion to the network
2938 * device hardware queue. @bytes should be a good approximation and should
2939 * exactly match netdev_completed_queue() @bytes
2940 */
c5d67bd7
TH
2941static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2942{
2943 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2944}
2945
2946static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2947 unsigned int pkts, unsigned int bytes)
c5d67bd7 2948{
114cf580 2949#ifdef CONFIG_BQL
b37c0fbe
AD
2950 if (unlikely(!bytes))
2951 return;
2952
2953 dql_completed(&dev_queue->dql, bytes);
2954
2955 /*
2956 * Without the memory barrier there is a small possiblity that
2957 * netdev_tx_sent_queue will miss the update and cause the queue to
2958 * be stopped forever
2959 */
2960 smp_mb();
2961
2962 if (dql_avail(&dev_queue->dql) < 0)
2963 return;
2964
2965 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2966 netif_schedule_queue(dev_queue);
114cf580 2967#endif
c5d67bd7
TH
2968}
2969
0042d0c8
FF
2970/**
2971 * netdev_completed_queue - report bytes and packets completed by device
2972 * @dev: network device
2973 * @pkts: actual number of packets sent over the medium
2974 * @bytes: actual number of bytes sent over the medium
2975 *
2976 * Report the number of bytes and packets transmitted by the network device
2977 * hardware queue over the physical medium, @bytes must exactly match the
2978 * @bytes amount passed to netdev_sent_queue()
2979 */
c5d67bd7 2980static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2981 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2982{
2983 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2984}
2985
2986static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2987{
114cf580 2988#ifdef CONFIG_BQL
5c490354 2989 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2990 dql_reset(&q->dql);
2991#endif
c5d67bd7
TH
2992}
2993
0042d0c8
FF
2994/**
2995 * netdev_reset_queue - reset the packets and bytes count of a network device
2996 * @dev_queue: network device
2997 *
2998 * Reset the bytes and packet count of a network device and clear the
2999 * software flow control OFF bit for this network device
3000 */
c5d67bd7
TH
3001static inline void netdev_reset_queue(struct net_device *dev_queue)
3002{
3003 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3004}
3005
b9507bda
DB
3006/**
3007 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3008 * @dev: network device
3009 * @queue_index: given tx queue index
3010 *
3011 * Returns 0 if given tx queue index >= number of device tx queues,
3012 * otherwise returns the originally passed tx queue index.
3013 */
3014static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3015{
3016 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3017 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3018 dev->name, queue_index,
3019 dev->real_num_tx_queues);
3020 return 0;
3021 }
3022
3023 return queue_index;
3024}
3025
bea3348e
SH
3026/**
3027 * netif_running - test if up
3028 * @dev: network device
3029 *
3030 * Test if the device has been brought up.
3031 */
4d29515f 3032static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3033{
3034 return test_bit(__LINK_STATE_START, &dev->state);
3035}
3036
f25f4e44 3037/*
5e82b4b2 3038 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3039 * stop, and a check if it's stopped. All other device management is
3040 * done at the overall netdevice level.
3041 * Also test the device if we're multiqueue.
3042 */
bea3348e
SH
3043
3044/**
3045 * netif_start_subqueue - allow sending packets on subqueue
3046 * @dev: network device
3047 * @queue_index: sub queue index
3048 *
3049 * Start individual transmit queue of a device with multiple transmit queues.
3050 */
f25f4e44
PWJ
3051static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3052{
fd2ea0a7 3053 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3054
3055 netif_tx_start_queue(txq);
f25f4e44
PWJ
3056}
3057
bea3348e
SH
3058/**
3059 * netif_stop_subqueue - stop sending packets on subqueue
3060 * @dev: network device
3061 * @queue_index: sub queue index
3062 *
3063 * Stop individual transmit queue of a device with multiple transmit queues.
3064 */
f25f4e44
PWJ
3065static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3066{
fd2ea0a7 3067 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3068 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3069}
3070
bea3348e
SH
3071/**
3072 * netif_subqueue_stopped - test status of subqueue
3073 * @dev: network device
3074 * @queue_index: sub queue index
3075 *
3076 * Check individual transmit queue of a device with multiple transmit queues.
3077 */
4d29515f
DM
3078static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3079 u16 queue_index)
f25f4e44 3080{
fd2ea0a7 3081 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3082
3083 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3084}
3085
4d29515f
DM
3086static inline bool netif_subqueue_stopped(const struct net_device *dev,
3087 struct sk_buff *skb)
668f895a
PE
3088{
3089 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3090}
bea3348e 3091
46e5da40 3092void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 3093
537c00de 3094#ifdef CONFIG_XPS
53af53ae 3095int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3096 u16 index);
537c00de
AD
3097#else
3098static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3099 const struct cpumask *mask,
537c00de
AD
3100 u16 index)
3101{
3102 return 0;
3103}
3104#endif
3105
5605c762
JP
3106u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3107 unsigned int num_tx_queues);
3108
a3d22a68
VZ
3109/*
3110 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3111 * as a distribution range limit for the returned value.
3112 */
3113static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3114 struct sk_buff *skb)
a3d22a68
VZ
3115{
3116 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3117}
3118
bea3348e
SH
3119/**
3120 * netif_is_multiqueue - test if device has multiple transmit queues
3121 * @dev: network device
3122 *
3123 * Check if device has multiple transmit queues
bea3348e 3124 */
4d29515f 3125static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3126{
a02cec21 3127 return dev->num_tx_queues > 1;
f25f4e44 3128}
1da177e4 3129
f629d208 3130int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3131
a953be53 3132#ifdef CONFIG_SYSFS
f629d208 3133int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3134#else
3135static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3136 unsigned int rxq)
3137{
3138 return 0;
3139}
3140#endif
3141
a953be53
MD
3142#ifdef CONFIG_SYSFS
3143static inline unsigned int get_netdev_rx_queue_index(
3144 struct netdev_rx_queue *queue)
3145{
3146 struct net_device *dev = queue->dev;
3147 int index = queue - dev->_rx;
3148
3149 BUG_ON(index >= dev->num_rx_queues);
3150 return index;
3151}
3152#endif
3153
16917b87 3154#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3155int netif_get_num_default_rss_queues(void);
16917b87 3156
e6247027
ED
3157enum skb_free_reason {
3158 SKB_REASON_CONSUMED,
3159 SKB_REASON_DROPPED,
3160};
3161
3162void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3163void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3164
e6247027
ED
3165/*
3166 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3167 * interrupt context or with hardware interrupts being disabled.
3168 * (in_irq() || irqs_disabled())
3169 *
3170 * We provide four helpers that can be used in following contexts :
3171 *
3172 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3173 * replacing kfree_skb(skb)
3174 *
3175 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3176 * Typically used in place of consume_skb(skb) in TX completion path
3177 *
3178 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3179 * replacing kfree_skb(skb)
3180 *
3181 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3182 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3183 */
e6247027
ED
3184static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3185{
3186 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3187}
3188
3189static inline void dev_consume_skb_irq(struct sk_buff *skb)
3190{
3191 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3192}
3193
3194static inline void dev_kfree_skb_any(struct sk_buff *skb)
3195{
3196 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3197}
3198
3199static inline void dev_consume_skb_any(struct sk_buff *skb)
3200{
3201 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3202}
1da177e4 3203
f629d208
JP
3204int netif_rx(struct sk_buff *skb);
3205int netif_rx_ni(struct sk_buff *skb);
04eb4489 3206int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3207gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3208void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3209struct sk_buff *napi_get_frags(struct napi_struct *napi);
3210gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3211struct packet_offload *gro_find_receive_by_type(__be16 type);
3212struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3213
3214static inline void napi_free_frags(struct napi_struct *napi)
3215{
3216 kfree_skb(napi->skb);
3217 napi->skb = NULL;
3218}
3219
24b27fc4 3220bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3221int netdev_rx_handler_register(struct net_device *dev,
3222 rx_handler_func_t *rx_handler,
3223 void *rx_handler_data);
3224void netdev_rx_handler_unregister(struct net_device *dev);
3225
3226bool dev_valid_name(const char *name);
3227int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3228int dev_ethtool(struct net *net, struct ifreq *);
3229unsigned int dev_get_flags(const struct net_device *);
3230int __dev_change_flags(struct net_device *, unsigned int flags);
3231int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3232void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3233 unsigned int gchanges);
f629d208
JP
3234int dev_change_name(struct net_device *, const char *);
3235int dev_set_alias(struct net_device *, const char *, size_t);
3236int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3237int dev_set_mtu(struct net_device *, int);
3238void dev_set_group(struct net_device *, int);
3239int dev_set_mac_address(struct net_device *, struct sockaddr *);
3240int dev_change_carrier(struct net_device *, bool new_carrier);
3241int dev_get_phys_port_id(struct net_device *dev,
02637fce 3242 struct netdev_phys_item_id *ppid);
db24a904
DA
3243int dev_get_phys_port_name(struct net_device *dev,
3244 char *name, size_t len);
d746d707 3245int dev_change_proto_down(struct net_device *dev, bool proto_down);
a7862b45 3246int dev_change_xdp_fd(struct net_device *dev, int fd);
55a93b3e 3247struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3248struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3249 struct netdev_queue *txq, int *ret);
a0265d28 3250int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3251int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3252bool is_skb_forwardable(const struct net_device *dev,
3253 const struct sk_buff *skb);
1da177e4 3254
4e3264d2
MKL
3255static __always_inline int ____dev_forward_skb(struct net_device *dev,
3256 struct sk_buff *skb)
3257{
3258 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3259 unlikely(!is_skb_forwardable(dev, skb))) {
3260 atomic_long_inc(&dev->rx_dropped);
3261 kfree_skb(skb);
3262 return NET_RX_DROP;
3263 }
3264
3265 skb_scrub_packet(skb, true);
3266 skb->priority = 0;
3267 return 0;
3268}
3269
74b20582
DA
3270void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3271
20380731 3272extern int netdev_budget;
1da177e4
LT
3273
3274/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3275void netdev_run_todo(void);
1da177e4 3276
bea3348e
SH
3277/**
3278 * dev_put - release reference to device
3279 * @dev: network device
3280 *
9ef4429b 3281 * Release reference to device to allow it to be freed.
bea3348e 3282 */
1da177e4
LT
3283static inline void dev_put(struct net_device *dev)
3284{
933393f5 3285 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3286}
3287
bea3348e
SH
3288/**
3289 * dev_hold - get reference to device
3290 * @dev: network device
3291 *
9ef4429b 3292 * Hold reference to device to keep it from being freed.
bea3348e 3293 */
15333061
SH
3294static inline void dev_hold(struct net_device *dev)
3295{
933393f5 3296 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3297}
1da177e4
LT
3298
3299/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3300 * and _off may be called from IRQ context, but it is caller
3301 * who is responsible for serialization of these calls.
b00055aa
SR
3302 *
3303 * The name carrier is inappropriate, these functions should really be
3304 * called netif_lowerlayer_*() because they represent the state of any
3305 * kind of lower layer not just hardware media.
1da177e4
LT
3306 */
3307
f629d208
JP
3308void linkwatch_init_dev(struct net_device *dev);
3309void linkwatch_fire_event(struct net_device *dev);
3310void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3311
bea3348e
SH
3312/**
3313 * netif_carrier_ok - test if carrier present
3314 * @dev: network device
3315 *
3316 * Check if carrier is present on device
3317 */
4d29515f 3318static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3319{
3320 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3321}
3322
f629d208 3323unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3324
f629d208 3325void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3326
f629d208 3327void netif_carrier_on(struct net_device *dev);
1da177e4 3328
f629d208 3329void netif_carrier_off(struct net_device *dev);
1da177e4 3330
bea3348e
SH
3331/**
3332 * netif_dormant_on - mark device as dormant.
3333 * @dev: network device
3334 *
3335 * Mark device as dormant (as per RFC2863).
3336 *
3337 * The dormant state indicates that the relevant interface is not
3338 * actually in a condition to pass packets (i.e., it is not 'up') but is
3339 * in a "pending" state, waiting for some external event. For "on-
3340 * demand" interfaces, this new state identifies the situation where the
3341 * interface is waiting for events to place it in the up state.
bea3348e 3342 */
b00055aa
SR
3343static inline void netif_dormant_on(struct net_device *dev)
3344{
3345 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3346 linkwatch_fire_event(dev);
3347}
3348
bea3348e
SH
3349/**
3350 * netif_dormant_off - set device as not dormant.
3351 * @dev: network device
3352 *
3353 * Device is not in dormant state.
3354 */
b00055aa
SR
3355static inline void netif_dormant_off(struct net_device *dev)
3356{
3357 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3358 linkwatch_fire_event(dev);
3359}
3360
bea3348e
SH
3361/**
3362 * netif_dormant - test if carrier present
3363 * @dev: network device
3364 *
3365 * Check if carrier is present on device
3366 */
4d29515f 3367static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3368{
3369 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3370}
3371
3372
bea3348e
SH
3373/**
3374 * netif_oper_up - test if device is operational
3375 * @dev: network device
3376 *
3377 * Check if carrier is operational
3378 */
4d29515f 3379static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3380{
b00055aa
SR
3381 return (dev->operstate == IF_OPER_UP ||
3382 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3383}
3384
bea3348e
SH
3385/**
3386 * netif_device_present - is device available or removed
3387 * @dev: network device
3388 *
3389 * Check if device has not been removed from system.
3390 */
4d29515f 3391static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3392{
3393 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3394}
3395
f629d208 3396void netif_device_detach(struct net_device *dev);
1da177e4 3397
f629d208 3398void netif_device_attach(struct net_device *dev);
1da177e4
LT
3399
3400/*
3401 * Network interface message level settings
3402 */
1da177e4
LT
3403
3404enum {
3405 NETIF_MSG_DRV = 0x0001,
3406 NETIF_MSG_PROBE = 0x0002,
3407 NETIF_MSG_LINK = 0x0004,
3408 NETIF_MSG_TIMER = 0x0008,
3409 NETIF_MSG_IFDOWN = 0x0010,
3410 NETIF_MSG_IFUP = 0x0020,
3411 NETIF_MSG_RX_ERR = 0x0040,
3412 NETIF_MSG_TX_ERR = 0x0080,
3413 NETIF_MSG_TX_QUEUED = 0x0100,
3414 NETIF_MSG_INTR = 0x0200,
3415 NETIF_MSG_TX_DONE = 0x0400,
3416 NETIF_MSG_RX_STATUS = 0x0800,
3417 NETIF_MSG_PKTDATA = 0x1000,
3418 NETIF_MSG_HW = 0x2000,
3419 NETIF_MSG_WOL = 0x4000,
3420};
3421
3422#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3423#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3424#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3425#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3426#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3427#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3428#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3429#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3430#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3431#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3432#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3433#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3434#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3435#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3436#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3437
3438static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3439{
3440 /* use default */
3441 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3442 return default_msg_enable_bits;
3443 if (debug_value == 0) /* no output */
3444 return 0;
3445 /* set low N bits */
3446 return (1 << debug_value) - 1;
3447}
3448
c773e847 3449static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3450{
c773e847
DM
3451 spin_lock(&txq->_xmit_lock);
3452 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3453}
3454
fd2ea0a7
DM
3455static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3456{
3457 spin_lock_bh(&txq->_xmit_lock);
3458 txq->xmit_lock_owner = smp_processor_id();
3459}
3460
4d29515f 3461static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3462{
4d29515f 3463 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3464 if (likely(ok))
3465 txq->xmit_lock_owner = smp_processor_id();
3466 return ok;
3467}
3468
3469static inline void __netif_tx_unlock(struct netdev_queue *txq)
3470{
3471 txq->xmit_lock_owner = -1;
3472 spin_unlock(&txq->_xmit_lock);
3473}
3474
3475static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3476{
3477 txq->xmit_lock_owner = -1;
3478 spin_unlock_bh(&txq->_xmit_lock);
3479}
3480
08baf561
ED
3481static inline void txq_trans_update(struct netdev_queue *txq)
3482{
3483 if (txq->xmit_lock_owner != -1)
3484 txq->trans_start = jiffies;
3485}
3486
ba162f8e
FW
3487/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3488static inline void netif_trans_update(struct net_device *dev)
3489{
9b36627a
FW
3490 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3491
3492 if (txq->trans_start != jiffies)
3493 txq->trans_start = jiffies;
ba162f8e
FW
3494}
3495
d29f749e
DJ
3496/**
3497 * netif_tx_lock - grab network device transmit lock
3498 * @dev: network device
d29f749e
DJ
3499 *
3500 * Get network device transmit lock
3501 */
22dd7495
JHS
3502static inline void netif_tx_lock(struct net_device *dev)
3503{
e8a0464c 3504 unsigned int i;
c3f26a26 3505 int cpu;
c773e847 3506
c3f26a26
DM
3507 spin_lock(&dev->tx_global_lock);
3508 cpu = smp_processor_id();
e8a0464c
DM
3509 for (i = 0; i < dev->num_tx_queues; i++) {
3510 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3511
3512 /* We are the only thread of execution doing a
3513 * freeze, but we have to grab the _xmit_lock in
3514 * order to synchronize with threads which are in
3515 * the ->hard_start_xmit() handler and already
3516 * checked the frozen bit.
3517 */
e8a0464c 3518 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3519 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3520 __netif_tx_unlock(txq);
e8a0464c 3521 }
932ff279
HX
3522}
3523
3524static inline void netif_tx_lock_bh(struct net_device *dev)
3525{
e8a0464c
DM
3526 local_bh_disable();
3527 netif_tx_lock(dev);
932ff279
HX
3528}
3529
932ff279
HX
3530static inline void netif_tx_unlock(struct net_device *dev)
3531{
e8a0464c
DM
3532 unsigned int i;
3533
3534 for (i = 0; i < dev->num_tx_queues; i++) {
3535 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3536
c3f26a26
DM
3537 /* No need to grab the _xmit_lock here. If the
3538 * queue is not stopped for another reason, we
3539 * force a schedule.
3540 */
3541 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3542 netif_schedule_queue(txq);
c3f26a26
DM
3543 }
3544 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3545}
3546
3547static inline void netif_tx_unlock_bh(struct net_device *dev)
3548{
e8a0464c
DM
3549 netif_tx_unlock(dev);
3550 local_bh_enable();
932ff279
HX
3551}
3552
c773e847 3553#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3554 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3555 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3556 } \
3557}
3558
5efeac44
EB
3559#define HARD_TX_TRYLOCK(dev, txq) \
3560 (((dev->features & NETIF_F_LLTX) == 0) ? \
3561 __netif_tx_trylock(txq) : \
3562 true )
3563
c773e847 3564#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3565 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3566 __netif_tx_unlock(txq); \
22dd7495
JHS
3567 } \
3568}
3569
1da177e4
LT
3570static inline void netif_tx_disable(struct net_device *dev)
3571{
fd2ea0a7 3572 unsigned int i;
c3f26a26 3573 int cpu;
fd2ea0a7 3574
c3f26a26
DM
3575 local_bh_disable();
3576 cpu = smp_processor_id();
fd2ea0a7
DM
3577 for (i = 0; i < dev->num_tx_queues; i++) {
3578 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3579
3580 __netif_tx_lock(txq, cpu);
fd2ea0a7 3581 netif_tx_stop_queue(txq);
c3f26a26 3582 __netif_tx_unlock(txq);
fd2ea0a7 3583 }
c3f26a26 3584 local_bh_enable();
1da177e4
LT
3585}
3586
e308a5d8
DM
3587static inline void netif_addr_lock(struct net_device *dev)
3588{
3589 spin_lock(&dev->addr_list_lock);
3590}
3591
2429f7ac
JP
3592static inline void netif_addr_lock_nested(struct net_device *dev)
3593{
25175ba5
VY
3594 int subclass = SINGLE_DEPTH_NESTING;
3595
3596 if (dev->netdev_ops->ndo_get_lock_subclass)
3597 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3598
3599 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3600}
3601
e308a5d8
DM
3602static inline void netif_addr_lock_bh(struct net_device *dev)
3603{
3604 spin_lock_bh(&dev->addr_list_lock);
3605}
3606
3607static inline void netif_addr_unlock(struct net_device *dev)
3608{
3609 spin_unlock(&dev->addr_list_lock);
3610}
3611
3612static inline void netif_addr_unlock_bh(struct net_device *dev)
3613{
3614 spin_unlock_bh(&dev->addr_list_lock);
3615}
3616
f001fde5 3617/*
31278e71 3618 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3619 * rcu_read_lock held.
3620 */
3621#define for_each_dev_addr(dev, ha) \
31278e71 3622 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3623
1da177e4
LT
3624/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3625
f629d208 3626void ether_setup(struct net_device *dev);
1da177e4
LT
3627
3628/* Support for loadable net-drivers */
f629d208 3629struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3630 unsigned char name_assign_type,
f629d208
JP
3631 void (*setup)(struct net_device *),
3632 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3633#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3634 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3635
c835a677
TG
3636#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3637 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3638 count)
36909ea4 3639
f629d208
JP
3640int register_netdev(struct net_device *dev);
3641void unregister_netdev(struct net_device *dev);
f001fde5 3642
22bedad3 3643/* General hardware address lists handling functions */
f629d208
JP
3644int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3645 struct netdev_hw_addr_list *from_list, int addr_len);
3646void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3647 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3648int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3649 struct net_device *dev,
3650 int (*sync)(struct net_device *, const unsigned char *),
3651 int (*unsync)(struct net_device *,
3652 const unsigned char *));
3653void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3654 struct net_device *dev,
3655 int (*unsync)(struct net_device *,
3656 const unsigned char *));
f629d208 3657void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3658
f001fde5 3659/* Functions used for device addresses handling */
f629d208
JP
3660int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3661 unsigned char addr_type);
3662int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3663 unsigned char addr_type);
f629d208
JP
3664void dev_addr_flush(struct net_device *dev);
3665int dev_addr_init(struct net_device *dev);
a748ee24
JP
3666
3667/* Functions used for unicast addresses handling */
f629d208
JP
3668int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3669int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3670int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3671int dev_uc_sync(struct net_device *to, struct net_device *from);
3672int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3673void dev_uc_unsync(struct net_device *to, struct net_device *from);
3674void dev_uc_flush(struct net_device *dev);
3675void dev_uc_init(struct net_device *dev);
f001fde5 3676
670e5b8e
AD
3677/**
3678 * __dev_uc_sync - Synchonize device's unicast list
3679 * @dev: device to sync
3680 * @sync: function to call if address should be added
3681 * @unsync: function to call if address should be removed
3682 *
3683 * Add newly added addresses to the interface, and release
3684 * addresses that have been deleted.
5e82b4b2 3685 */
670e5b8e
AD
3686static inline int __dev_uc_sync(struct net_device *dev,
3687 int (*sync)(struct net_device *,
3688 const unsigned char *),
3689 int (*unsync)(struct net_device *,
3690 const unsigned char *))
3691{
3692 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3693}
3694
3695/**
e793c0f7 3696 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3697 * @dev: device to sync
3698 * @unsync: function to call if address should be removed
3699 *
3700 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3701 */
670e5b8e
AD
3702static inline void __dev_uc_unsync(struct net_device *dev,
3703 int (*unsync)(struct net_device *,
3704 const unsigned char *))
3705{
3706 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3707}
3708
22bedad3 3709/* Functions used for multicast addresses handling */
f629d208
JP
3710int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3711int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3712int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3713int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3714int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3715int dev_mc_sync(struct net_device *to, struct net_device *from);
3716int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3717void dev_mc_unsync(struct net_device *to, struct net_device *from);
3718void dev_mc_flush(struct net_device *dev);
3719void dev_mc_init(struct net_device *dev);
f001fde5 3720
670e5b8e
AD
3721/**
3722 * __dev_mc_sync - Synchonize device's multicast list
3723 * @dev: device to sync
3724 * @sync: function to call if address should be added
3725 * @unsync: function to call if address should be removed
3726 *
3727 * Add newly added addresses to the interface, and release
3728 * addresses that have been deleted.
5e82b4b2 3729 */
670e5b8e
AD
3730static inline int __dev_mc_sync(struct net_device *dev,
3731 int (*sync)(struct net_device *,
3732 const unsigned char *),
3733 int (*unsync)(struct net_device *,
3734 const unsigned char *))
3735{
3736 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3737}
3738
3739/**
e793c0f7 3740 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3741 * @dev: device to sync
3742 * @unsync: function to call if address should be removed
3743 *
3744 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3745 */
670e5b8e
AD
3746static inline void __dev_mc_unsync(struct net_device *dev,
3747 int (*unsync)(struct net_device *,
3748 const unsigned char *))
3749{
3750 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3751}
3752
4417da66 3753/* Functions used for secondary unicast and multicast support */
f629d208
JP
3754void dev_set_rx_mode(struct net_device *dev);
3755void __dev_set_rx_mode(struct net_device *dev);
3756int dev_set_promiscuity(struct net_device *dev, int inc);
3757int dev_set_allmulti(struct net_device *dev, int inc);
3758void netdev_state_change(struct net_device *dev);
3759void netdev_notify_peers(struct net_device *dev);
3760void netdev_features_change(struct net_device *dev);
1da177e4 3761/* Load a device via the kmod */
f629d208
JP
3762void dev_load(struct net *net, const char *name);
3763struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3764 struct rtnl_link_stats64 *storage);
3765void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3766 const struct net_device_stats *netdev_stats);
eeda3fd6 3767
1da177e4 3768extern int netdev_max_backlog;
3b098e2d 3769extern int netdev_tstamp_prequeue;
1da177e4 3770extern int weight_p;
9ff162a8 3771
f629d208 3772bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3773struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3774 struct list_head **iter);
f629d208
JP
3775struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3776 struct list_head **iter);
8b5be856 3777
44a40855
VY
3778/* iterate through upper list, must be called under RCU read lock */
3779#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3780 for (iter = &(dev)->adj_list.upper, \
3781 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3782 updev; \
3783 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3784
1a3f060c
DA
3785int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3786 int (*fn)(struct net_device *upper_dev,
3787 void *data),
3788 void *data);
3789
3790bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3791 struct net_device *upper_dev);
3792
f629d208
JP
3793void *netdev_lower_get_next_private(struct net_device *dev,
3794 struct list_head **iter);
3795void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3796 struct list_head **iter);
31088a11
VF
3797
3798#define netdev_for_each_lower_private(dev, priv, iter) \
3799 for (iter = (dev)->adj_list.lower.next, \
3800 priv = netdev_lower_get_next_private(dev, &(iter)); \
3801 priv; \
3802 priv = netdev_lower_get_next_private(dev, &(iter)))
3803
3804#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3805 for (iter = &(dev)->adj_list.lower, \
3806 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3807 priv; \
3808 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3809
4085ebe8
VY
3810void *netdev_lower_get_next(struct net_device *dev,
3811 struct list_head **iter);
7ce856aa 3812
4085ebe8 3813#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3814 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3815 ldev = netdev_lower_get_next(dev, &(iter)); \
3816 ldev; \
3817 ldev = netdev_lower_get_next(dev, &(iter)))
3818
7ce856aa
JP
3819struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3820 struct list_head **iter);
3821struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3822 struct list_head **iter);
3823
1a3f060c
DA
3824int netdev_walk_all_lower_dev(struct net_device *dev,
3825 int (*fn)(struct net_device *lower_dev,
3826 void *data),
3827 void *data);
3828int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3829 int (*fn)(struct net_device *lower_dev,
3830 void *data),
3831 void *data);
3832
f629d208 3833void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3834void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3835struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3836struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3837int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3838int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3839 struct net_device *upper_dev,
29bf24af 3840 void *upper_priv, void *upper_info);
f629d208
JP
3841void netdev_upper_dev_unlink(struct net_device *dev,
3842 struct net_device *upper_dev);
5bb025fa 3843void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3844void *netdev_lower_dev_get_private(struct net_device *dev,
3845 struct net_device *lower_dev);
04d48266
JP
3846void netdev_lower_state_changed(struct net_device *lower_dev,
3847 void *lower_state_info);
18bfb924
JP
3848int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3849 struct neighbour *n);
3850void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3851 struct neighbour *n);
960fb622
ED
3852
3853/* RSS keys are 40 or 52 bytes long */
3854#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3855extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3856void netdev_rss_key_fill(void *buffer, size_t len);
3857
952fcfd0 3858int dev_get_nest_level(struct net_device *dev);
f629d208
JP
3859int skb_checksum_help(struct sk_buff *skb);
3860struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3861 netdev_features_t features, bool tx_path);
3862struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3863 netdev_features_t features);
12b0004d 3864
61bd3857
MS
3865struct netdev_bonding_info {
3866 ifslave slave;
3867 ifbond master;
3868};
3869
3870struct netdev_notifier_bonding_info {
3871 struct netdev_notifier_info info; /* must be first */
3872 struct netdev_bonding_info bonding_info;
3873};
3874
3875void netdev_bonding_info_change(struct net_device *dev,
3876 struct netdev_bonding_info *bonding_info);
3877
12b0004d
CW
3878static inline
3879struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3880{
3881 return __skb_gso_segment(skb, features, true);
3882}
53d6471c 3883__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3884
3885static inline bool can_checksum_protocol(netdev_features_t features,
3886 __be16 protocol)
3887{
c8cd0989
TH
3888 if (protocol == htons(ETH_P_FCOE))
3889 return !!(features & NETIF_F_FCOE_CRC);
3890
3891 /* Assume this is an IP checksum (not SCTP CRC) */
3892
3893 if (features & NETIF_F_HW_CSUM) {
3894 /* Can checksum everything */
3895 return true;
3896 }
3897
3898 switch (protocol) {
3899 case htons(ETH_P_IP):
3900 return !!(features & NETIF_F_IP_CSUM);
3901 case htons(ETH_P_IPV6):
3902 return !!(features & NETIF_F_IPV6_CSUM);
3903 default:
3904 return false;
3905 }
ec5f0615 3906}
12b0004d 3907
fb286bb2 3908#ifdef CONFIG_BUG
f629d208 3909void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3910#else
3911static inline void netdev_rx_csum_fault(struct net_device *dev)
3912{
3913}
3914#endif
1da177e4 3915/* rx skb timestamps */
f629d208
JP
3916void net_enable_timestamp(void);
3917void net_disable_timestamp(void);
1da177e4 3918
20380731 3919#ifdef CONFIG_PROC_FS
f629d208 3920int __init dev_proc_init(void);
900ff8c6
CW
3921#else
3922#define dev_proc_init() 0
20380731
ACM
3923#endif
3924
4798248e 3925static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3926 struct sk_buff *skb, struct net_device *dev,
3927 bool more)
4798248e 3928{
fa2dbdc2 3929 skb->xmit_more = more ? 1 : 0;
0b725a2c 3930 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3931}
3932
10b3ad8c 3933static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3934 struct netdev_queue *txq, bool more)
4798248e
DM
3935{
3936 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3937 int rc;
4798248e 3938
fa2dbdc2 3939 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3940 if (rc == NETDEV_TX_OK)
3941 txq_trans_update(txq);
3942
3943 return rc;
4798248e
DM
3944}
3945
42a2d923
LT
3946int netdev_class_create_file_ns(struct class_attribute *class_attr,
3947 const void *ns);
3948void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3949 const void *ns);
58292cbe
TH
3950
3951static inline int netdev_class_create_file(struct class_attribute *class_attr)
3952{
3953 return netdev_class_create_file_ns(class_attr, NULL);
3954}
3955
3956static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3957{
3958 netdev_class_remove_file_ns(class_attr, NULL);
3959}
b8a9787e 3960
04600794
JB
3961extern struct kobj_ns_type_operations net_ns_type_operations;
3962
f629d208 3963const char *netdev_drivername(const struct net_device *dev);
6579e57b 3964
f629d208 3965void linkwatch_run_queue(void);
20380731 3966
da08143b
MK
3967static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3968 netdev_features_t f2)
3969{
c8cd0989
TH
3970 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3971 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 3972 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 3973 else
b6a0e72a 3974 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 3975 }
da08143b 3976
c8cd0989 3977 return f1 & f2;
da08143b
MK
3978}
3979
c8f44aff
MM
3980static inline netdev_features_t netdev_get_wanted_features(
3981 struct net_device *dev)
5455c699
MM
3982{
3983 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3984}
c8f44aff
MM
3985netdev_features_t netdev_increment_features(netdev_features_t all,
3986 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3987
3988/* Allow TSO being used on stacked device :
3989 * Performing the GSO segmentation before last device
3990 * is a performance improvement.
3991 */
3992static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3993 netdev_features_t mask)
3994{
3995 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3996}
3997
6cb6a27c 3998int __netdev_update_features(struct net_device *dev);
5455c699 3999void netdev_update_features(struct net_device *dev);
afe12cc8 4000void netdev_change_features(struct net_device *dev);
7f353bf2 4001
fc4a7489
PM
4002void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4003 struct net_device *dev);
4004
e38f3025
TM
4005netdev_features_t passthru_features_check(struct sk_buff *skb,
4006 struct net_device *dev,
4007 netdev_features_t features);
c1e756bf 4008netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4009
4d29515f 4010static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4011{
7b748340 4012 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4013
4014 /* check flags correspondence */
4015 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4016 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4017 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4018 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4019 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4020 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4021 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4022 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4023 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4024 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4025 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4026 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4027 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4028 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4029 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4030 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
0345e186 4031
d6b4991a 4032 return (features & feature) == feature;
576a30eb
HX
4033}
4034
4d29515f 4035static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4036{
278b2513 4037 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4038 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4039}
4040
8b86a61d 4041static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4042 netdev_features_t features)
7967168c 4043{
fc741216 4044 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4045 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4046 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4047}
4048
82cc1a7a
PWJ
4049static inline void netif_set_gso_max_size(struct net_device *dev,
4050 unsigned int size)
4051{
4052 dev->gso_max_size = size;
4053}
4054
7a7ffbab
WCC
4055static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4056 int pulled_hlen, u16 mac_offset,
4057 int mac_len)
4058{
4059 skb->protocol = protocol;
4060 skb->encapsulation = 1;
4061 skb_push(skb, pulled_hlen);
4062 skb_reset_transport_header(skb);
4063 skb->mac_header = mac_offset;
4064 skb->network_header = skb->mac_header + mac_len;
4065 skb->mac_len = mac_len;
4066}
4067
3c175784
SD
4068static inline bool netif_is_macsec(const struct net_device *dev)
4069{
4070 return dev->priv_flags & IFF_MACSEC;
4071}
4072
b618aaa9 4073static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4074{
4075 return dev->priv_flags & IFF_MACVLAN;
4076}
4077
b618aaa9 4078static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4079{
4080 return dev->priv_flags & IFF_MACVLAN_PORT;
4081}
4082
b618aaa9 4083static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4084{
4085 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4086}
4087
b618aaa9 4088static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4089{
4090 return dev->priv_flags & IFF_IPVLAN_MASTER;
4091}
4092
b618aaa9 4093static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4094{
4095 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4096}
4097
b618aaa9 4098static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4099{
4100 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4101}
4102
3bdc0eba
BG
4103static inline bool netif_supports_nofcs(struct net_device *dev)
4104{
4105 return dev->priv_flags & IFF_SUPP_NOFCS;
4106}
4107
007979ea 4108static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4109{
007979ea 4110 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4111}
4112
fee6d4c7
DA
4113static inline bool netif_is_l3_slave(const struct net_device *dev)
4114{
4115 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4116}
4117
0894ae3f
JP
4118static inline bool netif_is_bridge_master(const struct net_device *dev)
4119{
4120 return dev->priv_flags & IFF_EBRIDGE;
4121}
4122
28f9ee22
VY
4123static inline bool netif_is_bridge_port(const struct net_device *dev)
4124{
4125 return dev->priv_flags & IFF_BRIDGE_PORT;
4126}
4127
35d4e172
JP
4128static inline bool netif_is_ovs_master(const struct net_device *dev)
4129{
4130 return dev->priv_flags & IFF_OPENVSWITCH;
4131}
4132
b618aaa9 4133static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4134{
4135 return dev->priv_flags & IFF_TEAM;
4136}
4137
b618aaa9 4138static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4139{
4140 return dev->priv_flags & IFF_TEAM_PORT;
4141}
4142
b618aaa9 4143static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4144{
4145 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4146}
4147
b618aaa9 4148static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4149{
4150 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4151}
4152
d4ab4286
KJ
4153static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4154{
4155 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4156}
4157
02875878
ED
4158/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4159static inline void netif_keep_dst(struct net_device *dev)
4160{
4161 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4162}
4163
18d3df3e
PA
4164/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4165static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4166{
4167 /* TODO: reserve and use an additional IFF bit, if we get more users */
4168 return dev->priv_flags & IFF_MACSEC;
4169}
4170
505d4f73 4171extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4172
571ba423
JP
4173/* Logging, debugging and troubleshooting/diagnostic helpers. */
4174
4175/* netdev_printk helpers, similar to dev_printk */
4176
4177static inline const char *netdev_name(const struct net_device *dev)
4178{
c6f854d5
VF
4179 if (!dev->name[0] || strchr(dev->name, '%'))
4180 return "(unnamed net_device)";
571ba423
JP
4181 return dev->name;
4182}
4183
ccc7f496
VF
4184static inline const char *netdev_reg_state(const struct net_device *dev)
4185{
4186 switch (dev->reg_state) {
4187 case NETREG_UNINITIALIZED: return " (uninitialized)";
4188 case NETREG_REGISTERED: return "";
4189 case NETREG_UNREGISTERING: return " (unregistering)";
4190 case NETREG_UNREGISTERED: return " (unregistered)";
4191 case NETREG_RELEASED: return " (released)";
4192 case NETREG_DUMMY: return " (dummy)";
4193 }
4194
4195 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4196 return " (unknown)";
4197}
4198
f629d208 4199__printf(3, 4)
6ea754eb
JP
4200void netdev_printk(const char *level, const struct net_device *dev,
4201 const char *format, ...);
f629d208 4202__printf(2, 3)
6ea754eb 4203void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4204__printf(2, 3)
6ea754eb 4205void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4206__printf(2, 3)
6ea754eb 4207void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4208__printf(2, 3)
6ea754eb 4209void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4210__printf(2, 3)
6ea754eb 4211void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4212__printf(2, 3)
6ea754eb 4213void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4214__printf(2, 3)
6ea754eb 4215void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4216
8909c9ad
VK
4217#define MODULE_ALIAS_NETDEV(device) \
4218 MODULE_ALIAS("netdev-" device)
4219
b558c96f 4220#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4221#define netdev_dbg(__dev, format, args...) \
4222do { \
ffa10cb4 4223 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4224} while (0)
b558c96f
JC
4225#elif defined(DEBUG)
4226#define netdev_dbg(__dev, format, args...) \
4227 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4228#else
4229#define netdev_dbg(__dev, format, args...) \
4230({ \
4231 if (0) \
4232 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4233})
4234#endif
4235
4236#if defined(VERBOSE_DEBUG)
4237#define netdev_vdbg netdev_dbg
4238#else
4239
4240#define netdev_vdbg(dev, format, args...) \
4241({ \
4242 if (0) \
4243 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4244 0; \
4245})
4246#endif
4247
4248/*
4249 * netdev_WARN() acts like dev_printk(), but with the key difference
4250 * of using a WARN/WARN_ON to get the message out, including the
4251 * file/line information and a backtrace.
4252 */
4253#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4254 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4255 netdev_reg_state(dev), ##args)
571ba423 4256
b3d95c5c
JP
4257/* netif printk helpers, similar to netdev_printk */
4258
4259#define netif_printk(priv, type, level, dev, fmt, args...) \
4260do { \
4261 if (netif_msg_##type(priv)) \
4262 netdev_printk(level, (dev), fmt, ##args); \
4263} while (0)
4264
f45f4321
JP
4265#define netif_level(level, priv, type, dev, fmt, args...) \
4266do { \
4267 if (netif_msg_##type(priv)) \
4268 netdev_##level(dev, fmt, ##args); \
4269} while (0)
4270
b3d95c5c 4271#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4272 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4273#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4274 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4275#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4276 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4277#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4278 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4279#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4280 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4281#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4282 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4283#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4284 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4285
0053ea9c 4286#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4287#define netif_dbg(priv, type, netdev, format, args...) \
4288do { \
4289 if (netif_msg_##type(priv)) \
b5fb0a03 4290 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4291} while (0)
0053ea9c
JP
4292#elif defined(DEBUG)
4293#define netif_dbg(priv, type, dev, format, args...) \
4294 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4295#else
4296#define netif_dbg(priv, type, dev, format, args...) \
4297({ \
4298 if (0) \
4299 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4300 0; \
4301})
4302#endif
4303
4304#if defined(VERBOSE_DEBUG)
bcfcc450 4305#define netif_vdbg netif_dbg
b3d95c5c
JP
4306#else
4307#define netif_vdbg(priv, type, dev, format, args...) \
4308({ \
4309 if (0) \
a4ed89cb 4310 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4311 0; \
4312})
4313#endif
571ba423 4314
900ff8c6
CW
4315/*
4316 * The list of packet types we will receive (as opposed to discard)
4317 * and the routines to invoke.
4318 *
4319 * Why 16. Because with 16 the only overlap we get on a hash of the
4320 * low nibble of the protocol value is RARP/SNAP/X.25.
4321 *
4322 * NOTE: That is no longer true with the addition of VLAN tags. Not
4323 * sure which should go first, but I bet it won't make much
4324 * difference if we are running VLANs. The good news is that
4325 * this protocol won't be in the list unless compiled in, so
4326 * the average user (w/out VLANs) will not be adversely affected.
4327 * --BLG
4328 *
4329 * 0800 IP
4330 * 8100 802.1Q VLAN
4331 * 0001 802.3
4332 * 0002 AX.25
4333 * 0004 802.2
4334 * 8035 RARP
4335 * 0005 SNAP
4336 * 0805 X.25
4337 * 0806 ARP
4338 * 8137 IPX
4339 * 0009 Localtalk
4340 * 86DD IPv6
4341 */
4342#define PTYPE_HASH_SIZE (16)
4343#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4344
385a154c 4345#endif /* _LINUX_NETDEVICE_H */