Merge branch 'net-ethernet-ti-k3-introduce-common-platform-time-sync-driver-cpts'
[linux-block.git] / include / linux / netdevice.h
CommitLineData
2874c5fd 1/* SPDX-License-Identifier: GPL-2.0-or-later */
1da177e4
LT
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
02c30a84 11 * Authors: Ross Biro
1da177e4
LT
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
1da177e4
LT
19 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
d7fe0f24 24#include <linux/timer.h>
187f1882 25#include <linux/bug.h>
bea3348e 26#include <linux/delay.h>
60063497 27#include <linux/atomic.h>
53511453 28#include <linux/prefetch.h>
1da177e4
LT
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31
1da177e4 32#include <linux/percpu.h>
4d5b78c0 33#include <linux/rculist.h>
bea3348e 34#include <linux/workqueue.h>
114cf580 35#include <linux/dynamic_queue_limits.h>
1da177e4 36
b1b67dd4 37#include <linux/ethtool.h>
a050c33f 38#include <net/net_namespace.h>
7a6b6f51 39#ifdef CONFIG_DCB
2f90b865
AD
40#include <net/dcbnl.h>
41#endif
5bc1421e 42#include <net/netprio_cgroup.h>
e817f856 43#include <net/xdp.h>
a050c33f 44
a59e2ecb 45#include <linux/netdev_features.h>
77162022 46#include <linux/neighbour.h>
607ca46e 47#include <uapi/linux/netdevice.h>
61bd3857 48#include <uapi/linux/if_bonding.h>
e4c6734e 49#include <uapi/linux/pkt_cls.h>
59cc1f61 50#include <linux/hashtable.h>
a59e2ecb 51
115c1d6e 52struct netpoll_info;
313162d0 53struct device;
c1f19b51 54struct phy_device;
2f657a60 55struct dsa_port;
30e9bb84
AT
56struct macsec_context;
57struct macsec_ops;
c6e970a0 58
e679c9c1 59struct sfp_bus;
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;
814abfab 68struct xdp_buff;
1da177e4 69
f629d208
JP
70void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
d07d7507 72
9a1654ba
JP
73/* Backlog congestion levels */
74#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75#define NET_RX_DROP 1 /* packet dropped */
76
7151affe
TY
77#define MAX_NEST_DEV 8
78
572a9d7b
PM
79/*
80 * Transmit return codes: transmit return codes originate from three different
81 * namespaces:
82 *
83 * - qdisc return codes
84 * - driver transmit return codes
85 * - errno values
86 *
87 * Drivers are allowed to return any one of those in their hard_start_xmit()
88 * function. Real network devices commonly used with qdiscs should only return
89 * the driver transmit return codes though - when qdiscs are used, the actual
90 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
91 * higher layers. Virtual network devices transmit synchronously; in this case
92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
93 * others are propagated to higher layers.
94 */
95
96/* qdisc ->enqueue() return codes. */
97#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
98#define NET_XMIT_DROP 0x01 /* skb dropped */
99#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 100#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 101
b9df3cb8
GR
102/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
103 * indicates that the device will soon be dropping packets, or already drops
104 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 105#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
106#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
107
dc1f8bf6 108/* Driver transmit return codes */
9a1654ba 109#define NETDEV_TX_MASK 0xf0
572a9d7b 110
dc1f8bf6 111enum netdev_tx {
572a9d7b 112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
113 NETDEV_TX_OK = 0x00, /* driver took care of packet */
114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
115};
116typedef enum netdev_tx netdev_tx_t;
117
9a1654ba
JP
118/*
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121 */
122static inline bool dev_xmit_complete(int rc)
123{
124 /*
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 */
130 if (likely(rc < NET_XMIT_MASK))
131 return true;
132
133 return false;
134}
135
1da177e4 136/*
5e82b4b2 137 * Compute the worst-case header length according to the protocols
1da177e4
LT
138 * used.
139 */
fe2918b0 140
c0eb4540
KS
141#if defined(CONFIG_HYPERV_NET)
142# define LL_MAX_HEADER 128
143#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
144# if defined(CONFIG_MAC80211_MESH)
145# define LL_MAX_HEADER 128
146# else
147# define LL_MAX_HEADER 96
148# endif
1da177e4 149#else
8388e3da 150# define LL_MAX_HEADER 32
1da177e4
LT
151#endif
152
d11ead75
BH
153#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
155#define MAX_HEADER LL_MAX_HEADER
156#else
157#define MAX_HEADER (LL_MAX_HEADER + 48)
158#endif
159
160/*
be1f3c2c
BH
161 * Old network device statistics. Fields are native words
162 * (unsigned long) so they can be read and written atomically.
1da177e4 163 */
fe2918b0 164
d94d9fee 165struct net_device_stats {
3cfde79c
BH
166 unsigned long rx_packets;
167 unsigned long tx_packets;
168 unsigned long rx_bytes;
169 unsigned long tx_bytes;
170 unsigned long rx_errors;
171 unsigned long tx_errors;
172 unsigned long rx_dropped;
173 unsigned long tx_dropped;
174 unsigned long multicast;
1da177e4 175 unsigned long collisions;
1da177e4 176 unsigned long rx_length_errors;
3cfde79c
BH
177 unsigned long rx_over_errors;
178 unsigned long rx_crc_errors;
179 unsigned long rx_frame_errors;
180 unsigned long rx_fifo_errors;
181 unsigned long rx_missed_errors;
1da177e4
LT
182 unsigned long tx_aborted_errors;
183 unsigned long tx_carrier_errors;
184 unsigned long tx_fifo_errors;
185 unsigned long tx_heartbeat_errors;
186 unsigned long tx_window_errors;
1da177e4
LT
187 unsigned long rx_compressed;
188 unsigned long tx_compressed;
189};
190
1da177e4
LT
191
192#include <linux/cache.h>
193#include <linux/skbuff.h>
194
adc9300e 195#ifdef CONFIG_RPS
c5905afb 196#include <linux/static_key.h>
dc05360f
ED
197extern struct static_key_false rps_needed;
198extern struct static_key_false rfs_needed;
adc9300e
ED
199#endif
200
1da177e4
LT
201struct neighbour;
202struct neigh_parms;
203struct sk_buff;
204
f001fde5
JP
205struct netdev_hw_addr {
206 struct list_head list;
207 unsigned char addr[MAX_ADDR_LEN];
208 unsigned char type;
ccffad25
JP
209#define NETDEV_HW_ADDR_T_LAN 1
210#define NETDEV_HW_ADDR_T_SAN 2
211#define NETDEV_HW_ADDR_T_SLAVE 3
212#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 213#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 214 bool global_use;
4cd729b0 215 int sync_cnt;
8f8f103d 216 int refcount;
4543fbef 217 int synced;
f001fde5
JP
218 struct rcu_head rcu_head;
219};
220
31278e71
JP
221struct netdev_hw_addr_list {
222 struct list_head list;
223 int count;
224};
225
22bedad3
JP
226#define netdev_hw_addr_list_count(l) ((l)->count)
227#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
228#define netdev_hw_addr_list_for_each(ha, l) \
229 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 230
22bedad3
JP
231#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
232#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
233#define netdev_for_each_uc_addr(ha, dev) \
234 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 235
22bedad3
JP
236#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
237#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 238#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 239 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 240
d94d9fee 241struct hh_cache {
5b3dc2f3 242 unsigned int hh_len;
3644f0ce 243 seqlock_t hh_lock;
1da177e4
LT
244
245 /* cached hardware header; allow for machine alignment needs. */
246#define HH_DATA_MOD 16
247#define HH_DATA_OFF(__len) \
5ba0eac6 248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
249#define HH_DATA_ALIGN(__len) \
250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
252};
253
5e82b4b2 254/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
255 * Alternative is:
256 * dev->hard_header_len ? (dev->hard_header_len +
257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
258 *
259 * We could use other alignment values, but we must maintain the
260 * relationship HH alignment <= LL alignment.
261 */
262#define LL_RESERVED_SPACE(dev) \
f5184d26 263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 264#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 266
3b04ddde
SH
267struct header_ops {
268 int (*create) (struct sk_buff *skb, struct net_device *dev,
269 unsigned short type, const void *daddr,
95c96174 270 const void *saddr, unsigned int len);
3b04ddde 271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
273 void (*cache_update)(struct hh_cache *hh,
274 const struct net_device *dev,
275 const unsigned char *haddr);
2793a23a 276 bool (*validate)(const char *ll_header, unsigned int len);
e78b2915 277 __be16 (*parse_protocol)(const struct sk_buff *skb);
3b04ddde
SH
278};
279
1da177e4 280/* These flag bits are private to the generic network queueing
5e82b4b2 281 * layer; they may not be explicitly referenced by any other
1da177e4
LT
282 * code.
283 */
284
d94d9fee 285enum netdev_state_t {
1da177e4
LT
286 __LINK_STATE_START,
287 __LINK_STATE_PRESENT,
1da177e4 288 __LINK_STATE_NOCARRIER,
b00055aa
SR
289 __LINK_STATE_LINKWATCH_PENDING,
290 __LINK_STATE_DORMANT,
eec517cd 291 __LINK_STATE_TESTING,
1da177e4
LT
292};
293
294
295/*
5e82b4b2 296 * This structure holds boot-time configured netdevice settings. They
fe2918b0 297 * are then used in the device probing.
1da177e4
LT
298 */
299struct netdev_boot_setup {
300 char name[IFNAMSIZ];
301 struct ifmap map;
302};
303#define NETDEV_BOOT_SETUP_MAX 8
304
f629d208 305int __init netdev_boot_setup(char *str);
1da177e4 306
6312fe77
LR
307struct gro_list {
308 struct list_head list;
309 int count;
310};
311
bea3348e 312/*
d9f37d01
LR
313 * size of gro hash buckets, must less than bit number of
314 * napi_struct::gro_bitmask
bea3348e 315 */
07d78363 316#define GRO_HASH_BUCKETS 8
d9f37d01
LR
317
318/*
319 * Structure for NAPI scheduling similar to tasklet but with weighting
320 */
bea3348e
SH
321struct napi_struct {
322 /* The poll_list must only be managed by the entity which
323 * changes the state of the NAPI_STATE_SCHED bit. This means
324 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 325 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
326 * can remove from the list right before clearing the bit.
327 */
328 struct list_head poll_list;
329
330 unsigned long state;
331 int weight;
6f8b12d6 332 int defer_hard_irqs_count;
d9f37d01 333 unsigned long gro_bitmask;
bea3348e
SH
334 int (*poll)(struct napi_struct *, int);
335#ifdef CONFIG_NETPOLL
bea3348e 336 int poll_owner;
bea3348e 337#endif
5d38a079 338 struct net_device *dev;
6312fe77 339 struct gro_list gro_hash[GRO_HASH_BUCKETS];
5d38a079 340 struct sk_buff *skb;
323ebb61
EC
341 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
342 int rx_count; /* length of rx_list */
3b47d303 343 struct hrtimer timer;
404f7c9e 344 struct list_head dev_list;
af12fa6e
ET
345 struct hlist_node napi_hash_node;
346 unsigned int napi_id;
bea3348e
SH
347};
348
d94d9fee 349enum {
bea3348e 350 NAPI_STATE_SCHED, /* Poll is scheduled */
39e6c820 351 NAPI_STATE_MISSED, /* reschedule a napi */
a0a46196 352 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 353 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
354 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
355 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
356 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
357};
358
359enum {
39e6c820
ED
360 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
361 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
362 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
363 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
364 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
365 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
366 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
367};
368
5b252f0c 369enum gro_result {
d1c76af9
HX
370 GRO_MERGED,
371 GRO_MERGED_FREE,
372 GRO_HELD,
373 GRO_NORMAL,
374 GRO_DROP,
25393d3f 375 GRO_CONSUMED,
d1c76af9 376};
5b252f0c 377typedef enum gro_result gro_result_t;
d1c76af9 378
8a4eb573
JP
379/*
380 * enum rx_handler_result - Possible return values for rx_handlers.
381 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
382 * further.
383 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
384 * case skb->dev was changed by rx_handler.
385 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 386 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
387 *
388 * rx_handlers are functions called from inside __netif_receive_skb(), to do
389 * special processing of the skb, prior to delivery to protocol handlers.
390 *
391 * Currently, a net_device can only have a single rx_handler registered. Trying
392 * to register a second rx_handler will return -EBUSY.
393 *
394 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
395 * To unregister a rx_handler on a net_device, use
396 * netdev_rx_handler_unregister().
397 *
398 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
399 * do with the skb.
400 *
5e82b4b2 401 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 402 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 403 * the skb to be delivered in some other way.
8a4eb573
JP
404 *
405 * If the rx_handler changed skb->dev, to divert the skb to another
406 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
407 * new device will be called if it exists.
408 *
5e82b4b2 409 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 410 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 411 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 412 *
5e82b4b2 413 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
414 * delivered, it should return RX_HANDLER_PASS.
415 *
416 * A device without a registered rx_handler will behave as if rx_handler
417 * returned RX_HANDLER_PASS.
418 */
419
420enum rx_handler_result {
421 RX_HANDLER_CONSUMED,
422 RX_HANDLER_ANOTHER,
423 RX_HANDLER_EXACT,
424 RX_HANDLER_PASS,
425};
426typedef enum rx_handler_result rx_handler_result_t;
427typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 428
f629d208 429void __napi_schedule(struct napi_struct *n);
bc9ad166 430void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 431
4d29515f 432static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
433{
434 return test_bit(NAPI_STATE_DISABLE, &n->state);
435}
436
39e6c820 437bool napi_schedule_prep(struct napi_struct *n);
bea3348e
SH
438
439/**
440 * napi_schedule - schedule NAPI poll
5e82b4b2 441 * @n: NAPI context
bea3348e
SH
442 *
443 * Schedule NAPI poll routine to be called if it is not already
444 * running.
445 */
446static inline void napi_schedule(struct napi_struct *n)
447{
448 if (napi_schedule_prep(n))
449 __napi_schedule(n);
450}
451
bc9ad166
ED
452/**
453 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 454 * @n: NAPI context
bc9ad166
ED
455 *
456 * Variant of napi_schedule(), assuming hard irqs are masked.
457 */
458static inline void napi_schedule_irqoff(struct napi_struct *n)
459{
460 if (napi_schedule_prep(n))
461 __napi_schedule_irqoff(n);
462}
463
bfe13f54 464/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 465static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
466{
467 if (napi_schedule_prep(napi)) {
468 __napi_schedule(napi);
4d29515f 469 return true;
bfe13f54 470 }
4d29515f 471 return false;
bfe13f54
RD
472}
473
364b6055 474bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
475/**
476 * napi_complete - NAPI processing complete
5e82b4b2 477 * @n: NAPI context
bea3348e
SH
478 *
479 * Mark NAPI processing as complete.
3b47d303 480 * Consider using napi_complete_done() instead.
364b6055 481 * Return false if device should avoid rearming interrupts.
bea3348e 482 */
364b6055 483static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
484{
485 return napi_complete_done(n, 0);
486}
bea3348e 487
af12fa6e
ET
488/**
489 * napi_hash_del - remove a NAPI from global table
5e82b4b2 490 * @napi: NAPI context
af12fa6e 491 *
5e82b4b2 492 * Warning: caller must observe RCU grace period
34cbe27e
ED
493 * before freeing memory containing @napi, if
494 * this function returns true.
93d05d4a 495 * Note: core networking stack automatically calls it
5e82b4b2 496 * from netif_napi_del().
93d05d4a 497 * Drivers might want to call this helper to combine all
5e82b4b2 498 * the needed RCU grace periods into a single one.
af12fa6e 499 */
34cbe27e 500bool napi_hash_del(struct napi_struct *napi);
af12fa6e 501
bea3348e
SH
502/**
503 * napi_disable - prevent NAPI from scheduling
5e82b4b2 504 * @n: NAPI context
bea3348e
SH
505 *
506 * Stop NAPI from being scheduled on this context.
507 * Waits till any outstanding processing completes.
508 */
3b47d303 509void napi_disable(struct napi_struct *n);
bea3348e
SH
510
511/**
512 * napi_enable - enable NAPI scheduling
5e82b4b2 513 * @n: NAPI context
bea3348e
SH
514 *
515 * Resume NAPI from being scheduled on this context.
516 * Must be paired with napi_disable.
517 */
518static inline void napi_enable(struct napi_struct *n)
519{
520 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 521 smp_mb__before_atomic();
bea3348e 522 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 523 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
524}
525
c264c3de
SH
526/**
527 * napi_synchronize - wait until NAPI is not running
5e82b4b2 528 * @n: NAPI context
c264c3de
SH
529 *
530 * Wait until NAPI is done being scheduled on this context.
531 * Waits till any outstanding processing completes but
532 * does not disable future activations.
533 */
534static inline void napi_synchronize(const struct napi_struct *n)
535{
facc432f
AB
536 if (IS_ENABLED(CONFIG_SMP))
537 while (test_bit(NAPI_STATE_SCHED, &n->state))
538 msleep(1);
539 else
540 barrier();
c264c3de 541}
c264c3de 542
6c5c9581
MK
543/**
544 * napi_if_scheduled_mark_missed - if napi is running, set the
545 * NAPIF_STATE_MISSED
546 * @n: NAPI context
547 *
548 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
549 * NAPI is scheduled.
550 **/
551static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
552{
553 unsigned long val, new;
554
555 do {
556 val = READ_ONCE(n->state);
557 if (val & NAPIF_STATE_DISABLE)
558 return true;
559
560 if (!(val & NAPIF_STATE_SCHED))
561 return false;
562
563 new = val | NAPIF_STATE_MISSED;
564 } while (cmpxchg(&n->state, val, new) != val);
565
566 return true;
567}
568
d94d9fee 569enum netdev_queue_state_t {
73466498
TH
570 __QUEUE_STATE_DRV_XOFF,
571 __QUEUE_STATE_STACK_XOFF,
c3f26a26 572 __QUEUE_STATE_FROZEN,
79d16385 573};
8e2f1a63
DB
574
575#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
576#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
577#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
578
579#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
580#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
581 QUEUE_STATE_FROZEN)
582#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
583 QUEUE_STATE_FROZEN)
584
73466498
TH
585/*
586 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
587 * netif_tx_* functions below are used to manipulate this flag. The
588 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
589 * queue independently. The netif_xmit_*stopped functions below are called
590 * to check if the queue has been stopped by the driver or stack (either
591 * of the XOFF bits are set in the state). Drivers should not need to call
592 * netif_xmit*stopped functions, they should only be using netif_tx_*.
593 */
79d16385 594
bb949fbd 595struct netdev_queue {
6a321cb3 596/*
5e82b4b2 597 * read-mostly part
6a321cb3 598 */
bb949fbd 599 struct net_device *dev;
46e5da40 600 struct Qdisc __rcu *qdisc;
b0e1e646 601 struct Qdisc *qdisc_sleeping;
ccf5ff69 602#ifdef CONFIG_SYSFS
1d24eb48
TH
603 struct kobject kobj;
604#endif
f2cd2d3e
ED
605#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 int numa_node;
607#endif
c0ef079c
FW
608 unsigned long tx_maxrate;
609 /*
610 * Number of TX timeouts for this queue
611 * (/sys/class/net/DEV/Q/trans_timeout)
612 */
613 unsigned long trans_timeout;
ffcfe25b
AD
614
615 /* Subordinate device that the queue has been assigned to */
616 struct net_device *sb_dev;
661b8d1b
MK
617#ifdef CONFIG_XDP_SOCKETS
618 struct xdp_umem *umem;
619#endif
6a321cb3 620/*
5e82b4b2 621 * write-mostly part
6a321cb3
ED
622 */
623 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
624 int xmit_lock_owner;
9d21493b 625 /*
9b36627a 626 * Time (in jiffies) of last Tx
9d21493b
ED
627 */
628 unsigned long trans_start;
ccf5ff69 629
114cf580
TH
630 unsigned long state;
631
632#ifdef CONFIG_BQL
633 struct dql dql;
634#endif
e8a0464c 635} ____cacheline_aligned_in_smp;
bb949fbd 636
79134e6c 637extern int sysctl_fb_tunnels_only_for_init_net;
856c395c 638extern int sysctl_devconf_inherit_init_net;
79134e6c
ED
639
640static inline bool net_has_fallback_tunnels(const struct net *net)
641{
be9fc097
AB
642 return net == &init_net ||
643 !IS_ENABLED(CONFIG_SYSCTL) ||
644 !sysctl_fb_tunnels_only_for_init_net;
79134e6c
ED
645}
646
f2cd2d3e
ED
647static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
648{
649#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
650 return q->numa_node;
651#else
b236da69 652 return NUMA_NO_NODE;
f2cd2d3e
ED
653#endif
654}
655
656static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
657{
658#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
659 q->numa_node = node;
660#endif
661}
662
df334545 663#ifdef CONFIG_RPS
0a9627f2
TH
664/*
665 * This structure holds an RPS map which can be of variable length. The
666 * map is an array of CPUs.
667 */
668struct rps_map {
669 unsigned int len;
670 struct rcu_head rcu;
bb4cf02d 671 u16 cpus[];
0a9627f2 672};
60b778ce 673#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 674
fec5e652 675/*
c445477d
BH
676 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
677 * tail pointer for that CPU's input queue at the time of last enqueue, and
678 * a hardware filter index.
fec5e652
TH
679 */
680struct rps_dev_flow {
681 u16 cpu;
c445477d 682 u16 filter;
fec5e652
TH
683 unsigned int last_qtail;
684};
c445477d 685#define RPS_NO_FILTER 0xffff
fec5e652
TH
686
687/*
688 * The rps_dev_flow_table structure contains a table of flow mappings.
689 */
690struct rps_dev_flow_table {
691 unsigned int mask;
692 struct rcu_head rcu;
bb4cf02d 693 struct rps_dev_flow flows[];
fec5e652
TH
694};
695#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 696 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
697
698/*
699 * The rps_sock_flow_table contains mappings of flows to the last CPU
700 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
701 * Each entry is a 32bit value. Upper part is the high-order bits
702 * of flow hash, lower part is CPU number.
567e4b79 703 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
704 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
705 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 706 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
707 */
708struct rps_sock_flow_table {
567e4b79 709 u32 mask;
93c1af6c 710
bb4cf02d 711 u32 ents[] ____cacheline_aligned_in_smp;
fec5e652 712};
567e4b79 713#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
714
715#define RPS_NO_CPU 0xffff
716
567e4b79
ED
717extern u32 rps_cpu_mask;
718extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
719
fec5e652
TH
720static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
721 u32 hash)
722{
723 if (table && hash) {
567e4b79
ED
724 unsigned int index = hash & table->mask;
725 u32 val = hash & ~rps_cpu_mask;
fec5e652 726
5e82b4b2 727 /* We only give a hint, preemption can change CPU under us */
567e4b79 728 val |= raw_smp_processor_id();
fec5e652 729
567e4b79
ED
730 if (table->ents[index] != val)
731 table->ents[index] = val;
fec5e652
TH
732 }
733}
734
c445477d 735#ifdef CONFIG_RFS_ACCEL
f629d208
JP
736bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
737 u16 filter_id);
c445477d 738#endif
a953be53 739#endif /* CONFIG_RPS */
c445477d 740
0a9627f2
TH
741/* This structure contains an instance of an RX queue. */
742struct netdev_rx_queue {
a953be53 743#ifdef CONFIG_RPS
6e3f7faf
ED
744 struct rps_map __rcu *rps_map;
745 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 746#endif
6e3f7faf 747 struct kobject kobj;
fe822240 748 struct net_device *dev;
e817f856 749 struct xdp_rxq_info xdp_rxq;
661b8d1b
MK
750#ifdef CONFIG_XDP_SOCKETS
751 struct xdp_umem *umem;
752#endif
0a9627f2 753} ____cacheline_aligned_in_smp;
a953be53
MD
754
755/*
756 * RX queue sysfs structures and functions.
757 */
758struct rx_queue_attribute {
759 struct attribute attr;
718ad681 760 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
a953be53 761 ssize_t (*store)(struct netdev_rx_queue *queue,
718ad681 762 const char *buf, size_t len);
a953be53 763};
d314774c 764
bf264145
TH
765#ifdef CONFIG_XPS
766/*
767 * This structure holds an XPS map which can be of variable length. The
768 * map is an array of queues.
769 */
770struct xps_map {
771 unsigned int len;
772 unsigned int alloc_len;
773 struct rcu_head rcu;
bb4cf02d 774 u16 queues[];
bf264145 775};
60b778ce 776#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
777#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
778 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
779
780/*
781 * This structure holds all XPS maps for device. Maps are indexed by CPU.
782 */
783struct xps_dev_maps {
784 struct rcu_head rcu;
bb4cf02d 785 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
bf264145 786};
80d19669
AN
787
788#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
184c449f 789 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
80d19669
AN
790
791#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
792 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
793
bf264145
TH
794#endif /* CONFIG_XPS */
795
4f57c087
JF
796#define TC_MAX_QUEUE 16
797#define TC_BITMASK 15
798/* HW offloaded queuing disciplines txq count and offset maps */
799struct netdev_tc_txq {
800 u16 count;
801 u16 offset;
802};
803
68bad94e
NP
804#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
805/*
806 * This structure is to hold information about the device
807 * configured to run FCoE protocol stack.
808 */
809struct netdev_fcoe_hbainfo {
810 char manufacturer[64];
811 char serial_number[64];
812 char hardware_version[64];
813 char driver_version[64];
814 char optionrom_version[64];
815 char firmware_version[64];
816 char model[256];
817 char model_description[256];
818};
819#endif
820
02637fce 821#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 822
02637fce
JP
823/* This structure holds a unique identifier to identify some
824 * physical item (port for example) used by a netdevice.
66b52b0d 825 */
02637fce
JP
826struct netdev_phys_item_id {
827 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
828 unsigned char id_len;
829};
830
d754f98b
SF
831static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
832 struct netdev_phys_item_id *b)
833{
834 return a->id_len == b->id_len &&
835 memcmp(a->id, b->id, a->id_len) == 0;
836}
837
99932d4f 838typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
8ec56fc3
AD
839 struct sk_buff *skb,
840 struct net_device *sb_dev);
99932d4f 841
2572ac53 842enum tc_setup_type {
575ed7d3 843 TC_SETUP_QDISC_MQPRIO,
a1b7c5fd 844 TC_SETUP_CLSU32,
5b33f488 845 TC_SETUP_CLSFLOWER,
ade9b658 846 TC_SETUP_CLSMATCHALL,
332ae8e2 847 TC_SETUP_CLSBPF,
8c4083b3 848 TC_SETUP_BLOCK,
8521db4c 849 TC_SETUP_QDISC_CBS,
602f3baf 850 TC_SETUP_QDISC_RED,
7fdb61b4 851 TC_SETUP_QDISC_PRIO,
f971b132 852 TC_SETUP_QDISC_MQ,
25db26a9 853 TC_SETUP_QDISC_ETF,
98b0e5f6 854 TC_SETUP_ROOT_QDISC,
890d8d23 855 TC_SETUP_QDISC_GRED,
9c66d156 856 TC_SETUP_QDISC_TAPRIO,
c29f74e0 857 TC_SETUP_FT,
d35eb52b 858 TC_SETUP_QDISC_ETS,
ef6aadcc 859 TC_SETUP_QDISC_TBF,
aaca9408 860 TC_SETUP_QDISC_FIFO,
16e5cc64
JF
861};
862
f4e63525
JK
863/* These structures hold the attributes of bpf state that are being passed
864 * to the netdevice through the bpf op.
a7862b45 865 */
f4e63525 866enum bpf_netdev_command {
a7862b45
BB
867 /* Set or clear a bpf program used in the earliest stages of packet
868 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
869 * is responsible for calling bpf_prog_put on any old progs that are
870 * stored. In case of error, the callee need not release the new prog
871 * reference, but on success it takes ownership and must bpf_prog_put
872 * when it is no longer used.
873 */
874 XDP_SETUP_PROG,
ee5d032f 875 XDP_SETUP_PROG_HW,
a7862b45 876 XDP_QUERY_PROG,
a25717d2 877 XDP_QUERY_PROG_HW,
ab3f0063 878 /* BPF program for offload callbacks, invoked at program load time. */
a3884572
JK
879 BPF_OFFLOAD_MAP_ALLOC,
880 BPF_OFFLOAD_MAP_FREE,
74515c57 881 XDP_SETUP_XSK_UMEM,
a7862b45
BB
882};
883
cae1927c 884struct bpf_prog_offload_ops;
ddf9f970 885struct netlink_ext_ack;
74515c57 886struct xdp_umem;
75ccae62 887struct xdp_dev_bulk_queue;
ddf9f970 888
f4e63525
JK
889struct netdev_bpf {
890 enum bpf_netdev_command command;
a7862b45
BB
891 union {
892 /* XDP_SETUP_PROG */
ddf9f970 893 struct {
32d60277 894 u32 flags;
ddf9f970
JK
895 struct bpf_prog *prog;
896 struct netlink_ext_ack *extack;
897 };
a25717d2 898 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
58038695 899 struct {
58038695 900 u32 prog_id;
92f0292b
JK
901 /* flags with which program was installed */
902 u32 prog_flags;
58038695 903 };
a3884572
JK
904 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
905 struct {
906 struct bpf_offloaded_map *offmap;
907 };
f8ebfaf6 908 /* XDP_SETUP_XSK_UMEM */
74515c57 909 struct {
f8ebfaf6
JS
910 struct xdp_umem *umem;
911 u16 queue_id;
74515c57 912 } xsk;
a7862b45
BB
913 };
914};
16e5cc64 915
9116e5e2
MK
916/* Flags for ndo_xsk_wakeup. */
917#define XDP_WAKEUP_RX (1 << 0)
918#define XDP_WAKEUP_TX (1 << 1)
919
d77e38e6
SK
920#ifdef CONFIG_XFRM_OFFLOAD
921struct xfrmdev_ops {
922 int (*xdo_dev_state_add) (struct xfrm_state *x);
923 void (*xdo_dev_state_delete) (struct xfrm_state *x);
924 void (*xdo_dev_state_free) (struct xfrm_state *x);
925 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
926 struct xfrm_state *x);
50bd870a 927 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
d77e38e6
SK
928};
929#endif
930
6c557001
FW
931struct dev_ifalias {
932 struct rcu_head rcuhead;
933 char ifalias[];
934};
935
b473b0d2 936struct devlink;
da68b4ad 937struct tlsdev_ops;
b473b0d2 938
ff927412
JP
939struct netdev_name_node {
940 struct hlist_node hlist;
36fbf1e5 941 struct list_head list;
ff927412
JP
942 struct net_device *dev;
943 const char *name;
944};
945
36fbf1e5
JP
946int netdev_name_node_alt_create(struct net_device *dev, const char *name);
947int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
948
93642e14
JP
949struct netdev_net_notifier {
950 struct list_head list;
951 struct notifier_block *nb;
952};
953
d314774c
SH
954/*
955 * This structure defines the management hooks for network devices.
00829823
SH
956 * The following hooks can be defined; unless noted otherwise, they are
957 * optional and can be filled with a null pointer.
d314774c
SH
958 *
959 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
960 * This function is called once when a network device is registered.
961 * The network device can use this for any late stage initialization
962 * or semantic validation. It can fail with an error code which will
963 * be propagated back to register_netdev.
d314774c
SH
964 *
965 * void (*ndo_uninit)(struct net_device *dev);
966 * This function is called when device is unregistered or when registration
967 * fails. It is not called if init fails.
968 *
969 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 970 * This function is called when a network device transitions to the up
d314774c
SH
971 * state.
972 *
973 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 974 * This function is called when a network device transitions to the down
d314774c
SH
975 * state.
976 *
dc1f8bf6
SH
977 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
978 * struct net_device *dev);
00829823 979 * Called when a packet needs to be transmitted.
e79d8429
RR
980 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
981 * the queue before that can happen; it's for obsolete devices and weird
982 * corner cases, but the stack really does a non-trivial amount
983 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 984 * Required; cannot be NULL.
00829823 985 *
1a2a1444
DM
986 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
987 * struct net_device *dev
988 * netdev_features_t features);
989 * Called by core transmit path to determine if device is capable of
990 * performing offload operations on a given packet. This is to give
991 * the device an opportunity to implement any restrictions that cannot
992 * be otherwise expressed by feature flags. The check is called with
993 * the set of features that the stack has calculated and it returns
994 * those the driver believes to be appropriate.
cdba756f 995 *
f663dd9a 996 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
a350ecce 997 * struct net_device *sb_dev);
5e82b4b2 998 * Called to decide which queue to use when device supports multiple
00829823
SH
999 * transmit queues.
1000 *
d314774c
SH
1001 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1002 * This function is called to allow device receiver to make
5e82b4b2 1003 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
1004 *
1005 * void (*ndo_set_rx_mode)(struct net_device *dev);
1006 * This function is called device changes address list filtering.
01789349 1007 * If driver handles unicast address filtering, it should set
5e82b4b2 1008 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
1009 *
1010 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1011 * This function is called when the Media Access Control address
37b607c5 1012 * needs to be changed. If this interface is not defined, the
5e82b4b2 1013 * MAC address can not be changed.
d314774c
SH
1014 *
1015 * int (*ndo_validate_addr)(struct net_device *dev);
1016 * Test if Media Access Control address is valid for the device.
1017 *
1018 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
1019 * Called when a user requests an ioctl which can't be handled by
1020 * the generic interface code. If not defined ioctls return
d314774c
SH
1021 * not supported error code.
1022 *
1023 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1024 * Used to set network devices bus interface parameters. This interface
5e82b4b2 1025 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
1026 * interface (PCI) for low level management.
1027 *
1028 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1029 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 1030 * of a device.
d314774c 1031 *
0290bd29 1032 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
5e82b4b2 1033 * Callback used when the transmitter has not made any progress
d314774c
SH
1034 * for dev->watchdog ticks.
1035 *
bc1f4470 1036 * void (*ndo_get_stats64)(struct net_device *dev,
1037 * struct rtnl_link_stats64 *storage);
d308e38f 1038 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 1039 * Called when a user wants to get the network device usage
be1f3c2c 1040 * statistics. Drivers must do one of the following:
3cfde79c
BH
1041 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1042 * rtnl_link_stats64 structure passed by the caller.
82695d9b 1043 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
1044 * (which should normally be dev->stats) and return a pointer to
1045 * it. The structure may be changed asynchronously only if each
1046 * field is written atomically.
1047 * 3. Update dev->stats asynchronously and atomically, and define
1048 * neither operation.
d314774c 1049 *
3df5b3c6 1050 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
1051 * Return true if this device supports offload stats of this attr_id.
1052 *
1053 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1054 * void *attr_data)
1055 * Get statistics for offload operations by attr_id. Write it into the
1056 * attr_data pointer.
1057 *
5d632cb7 1058 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 1059 * If device supports VLAN filtering this function is called when a
80d5c368 1060 * VLAN id is registered.
d314774c 1061 *
5d632cb7 1062 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 1063 * If device supports VLAN filtering this function is called when a
80d5c368 1064 * VLAN id is unregistered.
d314774c
SH
1065 *
1066 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
1067 *
1068 * SR-IOV management functions.
1069 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
1070 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1071 * u8 qos, __be16 proto);
ed616689
SC
1072 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1073 * int max_tx_rate);
5f8444a3 1074 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 1075 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
1076 * int (*ndo_get_vf_config)(struct net_device *dev,
1077 * int vf, struct ifla_vf_info *ivf);
1d8faf48 1078 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
1079 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1080 * struct nlattr *port[]);
01a3d796
VZ
1081 *
1082 * Enable or disable the VF ability to query its RSS Redirection Table and
1083 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 1084 * with PF and querying it may introduce a theoretical security risk.
01a3d796 1085 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 1086 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
2572ac53 1087 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
de4784ca 1088 * void *type_data);
6a4bc2b4
FF
1089 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1090 * This is always called from the stack with the rtnl lock held and netif
1091 * tx queues stopped. This allows the netdevice to perform queue
1092 * management safely.
c445477d 1093 *
e9bce845
YZ
1094 * Fiber Channel over Ethernet (FCoE) offload functions.
1095 * int (*ndo_fcoe_enable)(struct net_device *dev);
1096 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1097 * so the underlying device can perform whatever needed configuration or
1098 * initialization to support acceleration of FCoE traffic.
1099 *
1100 * int (*ndo_fcoe_disable)(struct net_device *dev);
1101 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1102 * so the underlying device can perform whatever needed clean-ups to
1103 * stop supporting acceleration of FCoE traffic.
1104 *
1105 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1106 * struct scatterlist *sgl, unsigned int sgc);
1107 * Called when the FCoE Initiator wants to initialize an I/O that
1108 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1109 * perform necessary setup and returns 1 to indicate the device is set up
1110 * successfully to perform DDP on this I/O, otherwise this returns 0.
1111 *
1112 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1113 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1114 * indicated by the FC exchange id 'xid', so the underlying device can
1115 * clean up and reuse resources for later DDP requests.
1116 *
1117 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1118 * struct scatterlist *sgl, unsigned int sgc);
1119 * Called when the FCoE Target wants to initialize an I/O that
1120 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1121 * perform necessary setup and returns 1 to indicate the device is set up
1122 * successfully to perform DDP on this I/O, otherwise this returns 0.
1123 *
68bad94e
NP
1124 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1125 * struct netdev_fcoe_hbainfo *hbainfo);
1126 * Called when the FCoE Protocol stack wants information on the underlying
1127 * device. This information is utilized by the FCoE protocol stack to
1128 * register attributes with Fiber Channel management service as per the
1129 * FC-GS Fabric Device Management Information(FDMI) specification.
1130 *
e9bce845
YZ
1131 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1132 * Called when the underlying device wants to override default World Wide
1133 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1134 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1135 * protocol stack to use.
1136 *
c445477d
BH
1137 * RFS acceleration.
1138 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1139 * u16 rxq_index, u32 flow_id);
1140 * Set hardware filter for RFS. rxq_index is the target queue index;
1141 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1142 * Return the filter ID on success, or a negative error code.
fbaec0ea 1143 *
8b98a70c 1144 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1145 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1146 * Called to make another netdev an underling.
1147 *
1148 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1149 * Called to release previously enslaved netdev.
5455c699
MM
1150 *
1151 * Feature/offload setting functions.
1a2a1444
DM
1152 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1153 * netdev_features_t features);
1154 * Adjusts the requested feature flags according to device-specific
1155 * constraints, and returns the resulting flags. Must not modify
1156 * the device state.
1157 *
c8f44aff 1158 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1159 * Called to update device configuration to new features. Passed
1160 * feature set might be less than what was returned by ndo_fix_features()).
1161 * Must return >0 or -errno if it changed dev->features itself.
1162 *
edc7d573 1163 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1164 * struct net_device *dev,
87b0984e
PM
1165 * const unsigned char *addr, u16 vid, u16 flags,
1166 * struct netlink_ext_ack *extack);
77162022 1167 * Adds an FDB entry to dev for addr.
1690be63
VY
1168 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1169 * struct net_device *dev,
f6f6424b 1170 * const unsigned char *addr, u16 vid)
77162022
JF
1171 * Deletes the FDB entry from dev coresponding to addr.
1172 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1173 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1174 * int *idx)
77162022
JF
1175 * Used to add FDB entries to dump requests. Implementers should add
1176 * entries to skb and update idx with the number of entries.
e5a55a89 1177 *
ad41faa8 1178 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
2fd527b7 1179 * u16 flags, struct netlink_ext_ack *extack)
e5a55a89 1180 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1181 * struct net_device *dev, u32 filter_mask,
1182 * int nlflags)
ad41faa8
ND
1183 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1184 * u16 flags);
4bf84c35
JP
1185 *
1186 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1187 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1188 * which do not represent real hardware may define this to allow their
1189 * userspace components to manage their virtual carrier state. Devices
1190 * that determine carrier state from physical hardware properties (eg
1191 * network cables) or protocol-dependent mechanisms (eg
1192 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1193 *
1194 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1195 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1196 * Called to get ID of physical port of this device. If driver does
1197 * not implement this, it is assumed that the hw is not able to have
1198 * multiple net devices on single physical port.
53cf5275 1199 *
d6abc596
FF
1200 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1201 * struct netdev_phys_item_id *ppid)
1202 * Called to get the parent ID of the physical port of this device.
1203 *
7c46a640
AD
1204 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1205 * struct udp_tunnel_info *ti);
1206 * Called by UDP tunnel to notify a driver about the UDP port and socket
1207 * address family that a UDP tunnel is listnening to. It is called only
1208 * when a new port starts listening. The operation is protected by the
1209 * RTNL.
1210 *
1211 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1212 * struct udp_tunnel_info *ti);
1213 * Called by UDP tunnel to notify the driver about a UDP port and socket
1214 * address family that the UDP tunnel is not listening to anymore. The
1215 * operation is protected by the RTNL.
1216 *
a6cc0cfa
JF
1217 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1218 * struct net_device *dev)
1219 * Called by upper layer devices to accelerate switching or other
1220 * station functionality into hardware. 'pdev is the lowerdev
1221 * to use for the offload and 'dev' is the net device that will
1222 * back the offload. Returns a pointer to the private structure
1223 * the upper layer will maintain.
1224 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1225 * Called by upper layer device to delete the station created
1226 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1227 * the station and priv is the structure returned by the add
1228 * operation.
822b3b2e
JF
1229 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1230 * int queue_index, u32 maxrate);
1231 * Called when a user wants to set a max-rate limitation of specific
1232 * TX queue.
a54acb3a
ND
1233 * int (*ndo_get_iflink)(const struct net_device *dev);
1234 * Called to get the iflink value of this device.
d746d707 1235 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1236 * bool proto_down);
d746d707
AK
1237 * This function is used to pass protocol port error state information
1238 * to the switch driver. The switch driver can react to the proto_down
1239 * by doing a phys down on the associated switch port.
fc4099f1
PS
1240 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1241 * This function is used to get egress tunnel information for given skb.
1242 * This is useful for retrieving outer tunnel header parameters while
1243 * sampling packet.
871b642a
PA
1244 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1245 * This function is used to specify the headroom that the skb must
1246 * consider when allocation skb during packet reception. Setting
1247 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1248 * forward. Setting a negative value resets the rx headroom to the
871b642a 1249 * default value.
f4e63525 1250 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
a7862b45 1251 * This function is used to set or query state related to XDP on the
f4e63525
JK
1252 * netdevice and manage BPF offload. See definition of
1253 * enum bpf_netdev_command for details.
42b33468
JDB
1254 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1255 * u32 flags);
735fc405
JDB
1256 * This function is used to submit @n XDP packets for transmit on a
1257 * netdevice. Returns number of frames successfully transmitted, frames
1258 * that got dropped are freed/returned via xdp_return_frame().
1259 * Returns negative number, means general error invoking ndo, meaning
1260 * no frames were xmit'ed and core-caller will free all frames.
9116e5e2
MK
1261 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1262 * This function is used to wake up the softirq, ksoftirqd or kthread
1263 * responsible for sending and/or receiving packets on a specific
1264 * queue id bound to an AF_XDP socket. The flags field specifies if
1265 * only RX, only Tx, or both should be woken up using the flags
1266 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
5dc37bb9
JP
1267 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1268 * Get devlink port instance associated with a given netdev.
b473b0d2
JK
1269 * Called with a reference on the netdevice and devlink locks only,
1270 * rtnl_lock is not held.
d314774c
SH
1271 */
1272struct net_device_ops {
1273 int (*ndo_init)(struct net_device *dev);
1274 void (*ndo_uninit)(struct net_device *dev);
1275 int (*ndo_open)(struct net_device *dev);
1276 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1277 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1278 struct net_device *dev);
1279 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1280 struct net_device *dev,
1281 netdev_features_t features);
00829823 1282 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1283 struct sk_buff *skb,
a350ecce 1284 struct net_device *sb_dev);
d314774c
SH
1285 void (*ndo_change_rx_flags)(struct net_device *dev,
1286 int flags);
d314774c 1287 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1288 int (*ndo_set_mac_address)(struct net_device *dev,
1289 void *addr);
d314774c 1290 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1291 int (*ndo_do_ioctl)(struct net_device *dev,
1292 struct ifreq *ifr, int cmd);
d314774c
SH
1293 int (*ndo_set_config)(struct net_device *dev,
1294 struct ifmap *map);
00829823
SH
1295 int (*ndo_change_mtu)(struct net_device *dev,
1296 int new_mtu);
1297 int (*ndo_neigh_setup)(struct net_device *dev,
1298 struct neigh_parms *);
0290bd29
MT
1299 void (*ndo_tx_timeout) (struct net_device *dev,
1300 unsigned int txqueue);
d314774c 1301
bc1f4470 1302 void (*ndo_get_stats64)(struct net_device *dev,
1303 struct rtnl_link_stats64 *storage);
3df5b3c6 1304 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1305 int (*ndo_get_offload_stats)(int attr_id,
1306 const struct net_device *dev,
1307 void *attr_data);
d314774c
SH
1308 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1309
8e586137 1310 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1311 __be16 proto, u16 vid);
8e586137 1312 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1313 __be16 proto, u16 vid);
d314774c 1314#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1315 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1316 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1317 struct netpoll_info *info);
0e34e931 1318 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1319#endif
95c26df8
WM
1320 int (*ndo_set_vf_mac)(struct net_device *dev,
1321 int queue, u8 *mac);
1322 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1323 int queue, u16 vlan,
1324 u8 qos, __be16 proto);
ed616689
SC
1325 int (*ndo_set_vf_rate)(struct net_device *dev,
1326 int vf, int min_tx_rate,
1327 int max_tx_rate);
5f8444a3
GR
1328 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1329 int vf, bool setting);
dd461d6a
HS
1330 int (*ndo_set_vf_trust)(struct net_device *dev,
1331 int vf, bool setting);
95c26df8
WM
1332 int (*ndo_get_vf_config)(struct net_device *dev,
1333 int vf,
1334 struct ifla_vf_info *ivf);
1d8faf48
RE
1335 int (*ndo_set_vf_link_state)(struct net_device *dev,
1336 int vf, int link_state);
3b766cd8
EBE
1337 int (*ndo_get_vf_stats)(struct net_device *dev,
1338 int vf,
1339 struct ifla_vf_stats
1340 *vf_stats);
57b61080
SF
1341 int (*ndo_set_vf_port)(struct net_device *dev,
1342 int vf,
1343 struct nlattr *port[]);
1344 int (*ndo_get_vf_port)(struct net_device *dev,
1345 int vf, struct sk_buff *skb);
30aad417
DG
1346 int (*ndo_get_vf_guid)(struct net_device *dev,
1347 int vf,
1348 struct ifla_vf_guid *node_guid,
1349 struct ifla_vf_guid *port_guid);
cc8e27cc
EC
1350 int (*ndo_set_vf_guid)(struct net_device *dev,
1351 int vf, u64 guid,
1352 int guid_type);
01a3d796
VZ
1353 int (*ndo_set_vf_rss_query_en)(
1354 struct net_device *dev,
1355 int vf, bool setting);
16e5cc64 1356 int (*ndo_setup_tc)(struct net_device *dev,
2572ac53 1357 enum tc_setup_type type,
de4784ca 1358 void *type_data);
d11ead75 1359#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1360 int (*ndo_fcoe_enable)(struct net_device *dev);
1361 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1362 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1363 u16 xid,
1364 struct scatterlist *sgl,
1365 unsigned int sgc);
1366 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1367 u16 xid);
6247e086
YZ
1368 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1369 u16 xid,
1370 struct scatterlist *sgl,
1371 unsigned int sgc);
68bad94e
NP
1372 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1373 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1374#endif
1375
d11ead75 1376#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1377#define NETDEV_FCOE_WWNN 0
1378#define NETDEV_FCOE_WWPN 1
1379 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1380 u64 *wwn, int type);
4d288d57 1381#endif
3c9c36bc 1382
c445477d
BH
1383#ifdef CONFIG_RFS_ACCEL
1384 int (*ndo_rx_flow_steer)(struct net_device *dev,
1385 const struct sk_buff *skb,
1386 u16 rxq_index,
1387 u32 flow_id);
1388#endif
fbaec0ea 1389 int (*ndo_add_slave)(struct net_device *dev,
33eaf2a6
DA
1390 struct net_device *slave_dev,
1391 struct netlink_ext_ack *extack);
fbaec0ea
JP
1392 int (*ndo_del_slave)(struct net_device *dev,
1393 struct net_device *slave_dev);
c8f44aff
MM
1394 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1395 netdev_features_t features);
5455c699 1396 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1397 netdev_features_t features);
503eebc2
JP
1398 int (*ndo_neigh_construct)(struct net_device *dev,
1399 struct neighbour *n);
1400 void (*ndo_neigh_destroy)(struct net_device *dev,
1401 struct neighbour *n);
77162022
JF
1402
1403 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1404 struct nlattr *tb[],
77162022 1405 struct net_device *dev,
6b6e2725 1406 const unsigned char *addr,
f6f6424b 1407 u16 vid,
87b0984e
PM
1408 u16 flags,
1409 struct netlink_ext_ack *extack);
77162022 1410 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1411 struct nlattr *tb[],
77162022 1412 struct net_device *dev,
f6f6424b
JP
1413 const unsigned char *addr,
1414 u16 vid);
77162022
JF
1415 int (*ndo_fdb_dump)(struct sk_buff *skb,
1416 struct netlink_callback *cb,
1417 struct net_device *dev,
5d5eacb3 1418 struct net_device *filter_dev,
d297653d 1419 int *idx);
5b2f94b2
RP
1420 int (*ndo_fdb_get)(struct sk_buff *skb,
1421 struct nlattr *tb[],
1422 struct net_device *dev,
1423 const unsigned char *addr,
1424 u16 vid, u32 portid, u32 seq,
1425 struct netlink_ext_ack *extack);
e5a55a89 1426 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3 1427 struct nlmsghdr *nlh,
2fd527b7
PM
1428 u16 flags,
1429 struct netlink_ext_ack *extack);
e5a55a89
JF
1430 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1431 u32 pid, u32 seq,
6cbdceeb 1432 struct net_device *dev,
46c264da
ND
1433 u32 filter_mask,
1434 int nlflags);
407af329 1435 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1436 struct nlmsghdr *nlh,
1437 u16 flags);
4bf84c35
JP
1438 int (*ndo_change_carrier)(struct net_device *dev,
1439 bool new_carrier);
66b52b0d 1440 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1441 struct netdev_phys_item_id *ppid);
d6abc596
FF
1442 int (*ndo_get_port_parent_id)(struct net_device *dev,
1443 struct netdev_phys_item_id *ppid);
db24a904
DA
1444 int (*ndo_get_phys_port_name)(struct net_device *dev,
1445 char *name, size_t len);
7c46a640
AD
1446 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1447 struct udp_tunnel_info *ti);
1448 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1449 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1450 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1451 struct net_device *dev);
1452 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1453 void *priv);
1454
822b3b2e
JF
1455 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1456 int queue_index,
1457 u32 maxrate);
a54acb3a 1458 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1459 int (*ndo_change_proto_down)(struct net_device *dev,
1460 bool proto_down);
fc4099f1
PS
1461 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1462 struct sk_buff *skb);
871b642a
PA
1463 void (*ndo_set_rx_headroom)(struct net_device *dev,
1464 int needed_headroom);
f4e63525
JK
1465 int (*ndo_bpf)(struct net_device *dev,
1466 struct netdev_bpf *bpf);
735fc405 1467 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
42b33468
JDB
1468 struct xdp_frame **xdp,
1469 u32 flags);
9116e5e2
MK
1470 int (*ndo_xsk_wakeup)(struct net_device *dev,
1471 u32 queue_id, u32 flags);
5dc37bb9 1472 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
d314774c
SH
1473};
1474
7aa98047
LR
1475/**
1476 * enum net_device_priv_flags - &struct net_device priv_flags
1477 *
1478 * These are the &struct net_device, they are only set internally
1479 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1480 * userspace; this means that the order of these flags can change
7aa98047
LR
1481 * during any kernel release.
1482 *
1483 * You should have a pretty good reason to be extending these flags.
1484 *
1485 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1486 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1487 * @IFF_BONDING: bonding master or slave
7aa98047 1488 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1489 * @IFF_WAN_HDLC: WAN HDLC device
1490 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1491 * release skb->dst
1492 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1493 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1494 * @IFF_MACVLAN_PORT: device used as macvlan port
1495 * @IFF_BRIDGE_PORT: device used as bridge port
1496 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1497 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1498 * @IFF_UNICAST_FLT: Supports unicast filtering
1499 * @IFF_TEAM_PORT: device used as team port
1500 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1501 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1502 * change when it's running
1503 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1504 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1505 * underlying stacked devices
007979ea 1506 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1507 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1508 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1509 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1510 * @IFF_TEAM: device is a team device
d4ab4286 1511 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1512 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1513 * entity (i.e. the master device for bridged veth)
3c175784 1514 * @IFF_MACSEC: device is a MACsec device
f5426250 1515 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
30c8bd5a
SS
1516 * @IFF_FAILOVER: device is a failover master device
1517 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
d5256083 1518 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
8065a779 1519 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
7aa98047
LR
1520 */
1521enum netdev_priv_flags {
1522 IFF_802_1Q_VLAN = 1<<0,
1523 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1524 IFF_BONDING = 1<<2,
1525 IFF_ISATAP = 1<<3,
1526 IFF_WAN_HDLC = 1<<4,
1527 IFF_XMIT_DST_RELEASE = 1<<5,
1528 IFF_DONT_BRIDGE = 1<<6,
1529 IFF_DISABLE_NETPOLL = 1<<7,
1530 IFF_MACVLAN_PORT = 1<<8,
1531 IFF_BRIDGE_PORT = 1<<9,
1532 IFF_OVS_DATAPATH = 1<<10,
1533 IFF_TX_SKB_SHARING = 1<<11,
1534 IFF_UNICAST_FLT = 1<<12,
1535 IFF_TEAM_PORT = 1<<13,
1536 IFF_SUPP_NOFCS = 1<<14,
1537 IFF_LIVE_ADDR_CHANGE = 1<<15,
1538 IFF_MACVLAN = 1<<16,
1539 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1ec54cb4
PA
1540 IFF_L3MDEV_MASTER = 1<<18,
1541 IFF_NO_QUEUE = 1<<19,
1542 IFF_OPENVSWITCH = 1<<20,
1543 IFF_L3MDEV_SLAVE = 1<<21,
1544 IFF_TEAM = 1<<22,
1545 IFF_RXFH_CONFIGURED = 1<<23,
1546 IFF_PHONY_HEADROOM = 1<<24,
1547 IFF_MACSEC = 1<<25,
f5426250 1548 IFF_NO_RX_HANDLER = 1<<26,
30c8bd5a
SS
1549 IFF_FAILOVER = 1<<27,
1550 IFF_FAILOVER_SLAVE = 1<<28,
d5256083 1551 IFF_L3MDEV_RX_HANDLER = 1<<29,
8065a779 1552 IFF_LIVE_RENAME_OK = 1<<30,
7aa98047
LR
1553};
1554
1555#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1556#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1557#define IFF_BONDING IFF_BONDING
7aa98047 1558#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1559#define IFF_WAN_HDLC IFF_WAN_HDLC
1560#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1561#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1562#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1563#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1564#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1565#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1566#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1567#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1568#define IFF_TEAM_PORT IFF_TEAM_PORT
1569#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1570#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1571#define IFF_MACVLAN IFF_MACVLAN
02875878 1572#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
007979ea 1573#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1574#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1575#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1576#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1577#define IFF_TEAM IFF_TEAM
d4ab4286 1578#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1579#define IFF_MACSEC IFF_MACSEC
f5426250 1580#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
30c8bd5a
SS
1581#define IFF_FAILOVER IFF_FAILOVER
1582#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
d5256083 1583#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
8065a779 1584#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
7aa98047 1585
536721b1
KK
1586/**
1587 * struct net_device - The DEVICE structure.
d651983d
MCC
1588 *
1589 * Actually, this whole structure is a big mistake. It mixes I/O
1590 * data with strictly "high-level" data, and it has to know about
1591 * almost every data structure used in the INET module.
536721b1
KK
1592 *
1593 * @name: This is the first field of the "visible" part of this structure
1594 * (i.e. as seen by users in the "Space.c" file). It is the name
d651983d 1595 * of the interface.
536721b1 1596 *
ff927412 1597 * @name_node: Name hashlist node
536721b1
KK
1598 * @ifalias: SNMP alias
1599 * @mem_end: Shared memory end
1600 * @mem_start: Shared memory start
1601 * @base_addr: Device I/O address
1602 * @irq: Device IRQ number
1603 *
1604 * @state: Generic network queuing layer state, see netdev_state_t
1605 * @dev_list: The global list of network devices
5e82b4b2
BH
1606 * @napi_list: List entry used for polling NAPI devices
1607 * @unreg_list: List entry when we are unregistering the
1608 * device; see the function unregister_netdev
1609 * @close_list: List entry used when we are closing the device
62d885fe
BP
1610 * @ptype_all: Device-specific packet handlers for all protocols
1611 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1612 *
1613 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1614 * @features: Currently active device features
1615 * @hw_features: User-changeable features
1616 *
1617 * @wanted_features: User-requested features
1618 * @vlan_features: Mask of features inheritable by VLAN devices
1619 *
1620 * @hw_enc_features: Mask of features inherited by encapsulating devices
1621 * This field indicates what encapsulation
1622 * offloads the hardware is capable of doing,
1623 * and drivers will need to set them appropriately.
1624 *
1625 * @mpls_features: Mask of features inheritable by MPLS
a1fa83bd 1626 * @gso_partial_features: value(s) from NETIF_F_GSO\*
536721b1
KK
1627 *
1628 * @ifindex: interface index
5e82b4b2 1629 * @group: The group the device belongs to
536721b1
KK
1630 *
1631 * @stats: Statistics struct, which was left as a legacy, use
1632 * rtnl_link_stats64 instead
1633 *
1634 * @rx_dropped: Dropped packets by core network,
1635 * do not use this in drivers
1636 * @tx_dropped: Dropped packets by core network,
1637 * do not use this in drivers
6e7333d3
JW
1638 * @rx_nohandler: nohandler dropped packets by core network on
1639 * inactive devices, do not use this in drivers
9e55e5d3
FF
1640 * @carrier_up_count: Number of times the carrier has been up
1641 * @carrier_down_count: Number of times the carrier has been down
536721b1 1642 *
536721b1
KK
1643 * @wireless_handlers: List of functions to handle Wireless Extensions,
1644 * instead of ioctl,
1645 * see <net/iw_handler.h> for details.
1646 * @wireless_data: Instance data managed by the core of wireless extensions
1647 *
1648 * @netdev_ops: Includes several pointers to callbacks,
1649 * if one wants to override the ndo_*() functions
1650 * @ethtool_ops: Management operations
a1fa83bd 1651 * @l3mdev_ops: Layer 3 master device operations
f997c55c
AA
1652 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1653 * discovery handling. Necessary for e.g. 6LoWPAN.
a1fa83bd
RD
1654 * @xfrmdev_ops: Transformation offload operations
1655 * @tlsdev_ops: Transport Layer Security offload operations
d476059e 1656 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1657 * of Layer 2 headers.
1658 *
1659 * @flags: Interface flags (a la BSD)
1660 * @priv_flags: Like 'flags' but invisible to userspace,
1661 * see if.h for the definitions
1662 * @gflags: Global flags ( kept as legacy )
1663 * @padded: How much padding added by alloc_netdev()
1664 * @operstate: RFC2863 operstate
1665 * @link_mode: Mapping policy to operstate
1666 * @if_port: Selectable AUI, TP, ...
1667 * @dma: DMA channel
1668 * @mtu: Interface MTU value
61e84623
JW
1669 * @min_mtu: Interface Minimum MTU value
1670 * @max_mtu: Interface Maximum MTU value
536721b1 1671 * @type: Interface hardware type
2793a23a 1672 * @hard_header_len: Maximum hardware header length.
217e6fa2 1673 * @min_header_len: Minimum hardware header length
536721b1
KK
1674 *
1675 * @needed_headroom: Extra headroom the hardware may need, but not in all
1676 * cases can this be guaranteed
1677 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1678 * cases can this be guaranteed. Some cases also use
1679 * LL_MAX_HEADER instead to allocate the skb
1680 *
1681 * interface address info:
1682 *
1683 * @perm_addr: Permanent hw address
1684 * @addr_assign_type: Hw address assignment type
1685 * @addr_len: Hardware address length
5343da4c
TY
1686 * @upper_level: Maximum depth level of upper devices.
1687 * @lower_level: Maximum depth level of lower devices.
8626a0c8 1688 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1689 * @dev_id: Used to differentiate devices that share
1690 * the same link layer address
1691 * @dev_port: Used to differentiate devices that share
1692 * the same function
1693 * @addr_list_lock: XXX: need comments on this one
a1fa83bd 1694 * @name_assign_type: network interface name assignment type
5e82b4b2 1695 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1696 * has been enabled due to the need to listen to
1697 * additional unicast addresses in a device that
1698 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1699 * @uc: unicast mac addresses
1700 * @mc: multicast mac addresses
1701 * @dev_addrs: list of device hw addresses
1702 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1703 * @promiscuity: Number of times the NIC is told to work in
1704 * promiscuous mode; if it becomes 0 the NIC will
1705 * exit promiscuous mode
536721b1
KK
1706 * @allmulti: Counter, enables or disables allmulticast mode
1707 *
1708 * @vlan_info: VLAN info
1709 * @dsa_ptr: dsa specific data
1710 * @tipc_ptr: TIPC specific data
1711 * @atalk_ptr: AppleTalk link
1712 * @ip_ptr: IPv4 specific data
1713 * @dn_ptr: DECnet specific data
1714 * @ip6_ptr: IPv6 specific data
1715 * @ax25_ptr: AX.25 specific data
1716 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
a1fa83bd
RD
1717 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1718 * device struct
1719 * @mpls_ptr: mpls_dev struct pointer
536721b1 1720 *
536721b1
KK
1721 * @dev_addr: Hw address (before bcast,
1722 * because most packets are unicast)
1723 *
1724 * @_rx: Array of RX queues
1725 * @num_rx_queues: Number of RX queues
1726 * allocated at register_netdev() time
1727 * @real_num_rx_queues: Number of RX queues currently active in device
a1fa83bd
RD
1728 * @xdp_prog: XDP sockets filter program pointer
1729 * @gro_flush_timeout: timeout for GRO layer in NAPI
536721b1
KK
1730 *
1731 * @rx_handler: handler for received packets
1732 * @rx_handler_data: XXX: need comments on this one
46209401
JP
1733 * @miniq_ingress: ingress/clsact qdisc specific data for
1734 * ingress processing
536721b1 1735 * @ingress_queue: XXX: need comments on this one
2f5e70c8 1736 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
536721b1
KK
1737 * @broadcast: hw bcast address
1738 *
14ffbbb8
TG
1739 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1740 * indexed by RX queue number. Assigned by driver.
1741 * This must only be set if the ndo_rx_flow_steer
1742 * operation is defined
1743 * @index_hlist: Device index hash chain
1744 *
536721b1
KK
1745 * @_tx: Array of TX queues
1746 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1747 * @real_num_tx_queues: Number of TX queues currently active in device
1748 * @qdisc: Root qdisc from userspace point of view
1749 * @tx_queue_len: Max frames per queue allowed
1750 * @tx_global_lock: XXX: need comments on this one
a1fa83bd
RD
1751 * @xdp_bulkq: XDP device bulk queue
1752 * @xps_cpus_map: all CPUs map for XPS device
1753 * @xps_rxqs_map: all RXQs map for XPS device
536721b1
KK
1754 *
1755 * @xps_maps: XXX: need comments on this one
46209401
JP
1756 * @miniq_egress: clsact qdisc specific data for
1757 * egress processing
a1fa83bd 1758 * @qdisc_hash: qdisc hash table
536721b1 1759 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1760 * the watchdog (see dev_watchdog())
536721b1
KK
1761 * @watchdog_timer: List of timers
1762 *
1763 * @pcpu_refcnt: Number of references to this device
1764 * @todo_list: Delayed register/unregister
536721b1
KK
1765 * @link_watch_list: XXX: need comments on this one
1766 *
1767 * @reg_state: Register/unregister state machine
1768 * @dismantle: Device is going to be freed
1769 * @rtnl_link_state: This enum represents the phases of creating
1770 * a new link
1771 *
cf124db5
DM
1772 * @needs_free_netdev: Should unregister perform free_netdev?
1773 * @priv_destructor: Called from unregister
536721b1
KK
1774 * @npinfo: XXX: need comments on this one
1775 * @nd_net: Network namespace this network device is inside
1776 *
1777 * @ml_priv: Mid-layer private
1778 * @lstats: Loopback statistics
1779 * @tstats: Tunnel statistics
1780 * @dstats: Dummy statistics
1781 * @vstats: Virtual ethernet statistics
1782 *
1783 * @garp_port: GARP
1784 * @mrp_port: MRP
1785 *
1786 * @dev: Class/net/name entry
1787 * @sysfs_groups: Space for optional device, statistics and wireless
1788 * sysfs groups
1789 *
1790 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1791 * @rtnl_link_ops: Rtnl_link_ops
1792 *
1793 * @gso_max_size: Maximum size of generic segmentation offload
1794 * @gso_max_segs: Maximum number of segments that can be passed to the
1795 * NIC for GSO
1796 *
1797 * @dcbnl_ops: Data Center Bridging netlink ops
1798 * @num_tc: Number of traffic classes in the net device
1799 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1800 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1801 *
1802 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1803 *
1804 * @priomap: XXX: need comments on this one
1805 * @phydev: Physical device may attach itself
1806 * for hardware timestamping
e679c9c1 1807 * @sfp_bus: attached &struct sfp_bus structure.
ab92d68f 1808 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock
1f26c0d3 1809 * spinlock
ab92d68f
TY
1810 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1811 * @qdisc_xmit_lock_key: lockdep class annotating
1812 * netdev_queue->_xmit_lock spinlock
1813 * @addr_list_lock_key: lockdep class annotating
1814 * net_device->addr_list_lock spinlock
536721b1 1815 *
d746d707
AK
1816 * @proto_down: protocol port state information can be sent to the
1817 * switch driver and used to set the phys state of the
1818 * switch port.
1819 *
61941143
HK
1820 * @wol_enabled: Wake-on-LAN is enabled
1821 *
93642e14
JP
1822 * @net_notifier_list: List of per-net netdev notifier block
1823 * that follow this device when it is moved
1824 * to another network namespace.
1825 *
30e9bb84
AT
1826 * @macsec_ops: MACsec offloading ops
1827 *
1da177e4
LT
1828 * FIXME: cleanup struct net_device such that network protocol info
1829 * moves out.
1830 */
1831
d94d9fee 1832struct net_device {
1da177e4 1833 char name[IFNAMSIZ];
ff927412 1834 struct netdev_name_node *name_node;
6c557001 1835 struct dev_ifalias __rcu *ifalias;
1da177e4
LT
1836 /*
1837 * I/O specific fields
1838 * FIXME: Merge these and struct ifmap into one
1839 */
536721b1
KK
1840 unsigned long mem_end;
1841 unsigned long mem_start;
1842 unsigned long base_addr;
1843 int irq;
1da177e4
LT
1844
1845 /*
536721b1
KK
1846 * Some hardware also needs these fields (state,dev_list,
1847 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1848 * part of the usual set specified in Space.c.
1849 */
1850
1da177e4
LT
1851 unsigned long state;
1852
7562f876 1853 struct list_head dev_list;
bea3348e 1854 struct list_head napi_list;
44a0873d 1855 struct list_head unreg_list;
5cde2829 1856 struct list_head close_list;
7866a621
SN
1857 struct list_head ptype_all;
1858 struct list_head ptype_specific;
2f268f12 1859
2f268f12
VF
1860 struct {
1861 struct list_head upper;
1862 struct list_head lower;
1863 } adj_list;
1864
c8f44aff 1865 netdev_features_t features;
c8f44aff 1866 netdev_features_t hw_features;
c8f44aff 1867 netdev_features_t wanted_features;
c8f44aff 1868 netdev_features_t vlan_features;
6a674e9c 1869 netdev_features_t hw_enc_features;
0d89d203 1870 netdev_features_t mpls_features;
802ab55a 1871 netdev_features_t gso_partial_features;
04ed3e74 1872
1da177e4 1873 int ifindex;
7a66bbc9 1874 int group;
1da177e4 1875
c45d286e 1876 struct net_device_stats stats;
015f0688 1877
015f0688
ED
1878 atomic_long_t rx_dropped;
1879 atomic_long_t tx_dropped;
6e7333d3 1880 atomic_long_t rx_nohandler;
1da177e4 1881
b2d3bcfa
DD
1882 /* Stats to monitor link on/off, flapping */
1883 atomic_t carrier_up_count;
1884 atomic_t carrier_down_count;
1885
b86e0280 1886#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1887 const struct iw_handler_def *wireless_handlers;
1888 struct iw_public_data *wireless_data;
b86e0280 1889#endif
d314774c 1890 const struct net_device_ops *netdev_ops;
76fd8593 1891 const struct ethtool_ops *ethtool_ops;
1b69c6d0
DA
1892#ifdef CONFIG_NET_L3_MASTER_DEV
1893 const struct l3mdev_ops *l3mdev_ops;
1894#endif
f997c55c
AA
1895#if IS_ENABLED(CONFIG_IPV6)
1896 const struct ndisc_ops *ndisc_ops;
1897#endif
1da177e4 1898
9cb0d21d 1899#ifdef CONFIG_XFRM_OFFLOAD
d77e38e6
SK
1900 const struct xfrmdev_ops *xfrmdev_ops;
1901#endif
1902
a5c37c63
IL
1903#if IS_ENABLED(CONFIG_TLS_DEVICE)
1904 const struct tlsdev_ops *tlsdev_ops;
1905#endif
1906
3b04ddde
SH
1907 const struct header_ops *header_ops;
1908
536721b1
KK
1909 unsigned int flags;
1910 unsigned int priv_flags;
1911
1da177e4 1912 unsigned short gflags;
536721b1 1913 unsigned short padded;
1da177e4 1914
536721b1
KK
1915 unsigned char operstate;
1916 unsigned char link_mode;
b00055aa 1917
536721b1
KK
1918 unsigned char if_port;
1919 unsigned char dma;
bdc220da 1920
501a90c9
ED
1921 /* Note : dev->mtu is often read without holding a lock.
1922 * Writers usually hold RTNL.
1923 * It is recommended to use READ_ONCE() to annotate the reads,
1924 * and to use WRITE_ONCE() to annotate the writes.
1925 */
536721b1 1926 unsigned int mtu;
61e84623
JW
1927 unsigned int min_mtu;
1928 unsigned int max_mtu;
536721b1
KK
1929 unsigned short type;
1930 unsigned short hard_header_len;
d92be7a4 1931 unsigned char min_header_len;
1da177e4 1932
f5184d26
JB
1933 unsigned short needed_headroom;
1934 unsigned short needed_tailroom;
1935
1da177e4 1936 /* Interface address info. */
536721b1
KK
1937 unsigned char perm_addr[MAX_ADDR_LEN];
1938 unsigned char addr_assign_type;
1939 unsigned char addr_len;
5343da4c
TY
1940 unsigned char upper_level;
1941 unsigned char lower_level;
a0a9663d 1942 unsigned short neigh_priv_len;
536721b1
KK
1943 unsigned short dev_id;
1944 unsigned short dev_port;
ccffad25 1945 spinlock_t addr_list_lock;
14ffbbb8
TG
1946 unsigned char name_assign_type;
1947 bool uc_promisc;
536721b1
KK
1948 struct netdev_hw_addr_list uc;
1949 struct netdev_hw_addr_list mc;
1950 struct netdev_hw_addr_list dev_addrs;
1951
4c3d5e7b
ED
1952#ifdef CONFIG_SYSFS
1953 struct kset *queues_kset;
1954#endif
9d45abe1
WC
1955 unsigned int promiscuity;
1956 unsigned int allmulti;
1da177e4 1957
1da177e4 1958
5e82b4b2 1959 /* Protocol-specific pointers */
65ac6a5f 1960
d11ead75 1961#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1962 struct vlan_info __rcu *vlan_info;
65ac6a5f 1963#endif
34a430d7 1964#if IS_ENABLED(CONFIG_NET_DSA)
2f657a60 1965 struct dsa_port *dsa_ptr;
37cb0620
YX
1966#endif
1967#if IS_ENABLED(CONFIG_TIPC)
536721b1 1968 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1969#endif
89e58148 1970#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
536721b1 1971 void *atalk_ptr;
89e58148 1972#endif
536721b1 1973 struct in_device __rcu *ip_ptr;
330c7272 1974#if IS_ENABLED(CONFIG_DECNET)
536721b1 1975 struct dn_dev __rcu *dn_ptr;
330c7272 1976#endif
536721b1 1977 struct inet6_dev __rcu *ip6_ptr;
19ff13f2 1978#if IS_ENABLED(CONFIG_AX25)
536721b1 1979 void *ax25_ptr;
19ff13f2 1980#endif
536721b1 1981 struct wireless_dev *ieee80211_ptr;
98a18b6f 1982 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1983#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1984 struct mpls_dev __rcu *mpls_ptr;
1985#endif
1da177e4 1986
9356b8fc 1987/*
cd13539b 1988 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1989 */
9356b8fc 1990 /* Interface address info used in eth_type_trans() */
536721b1 1991 unsigned char *dev_addr;
f001fde5 1992
0a9627f2 1993 struct netdev_rx_queue *_rx;
0a9627f2 1994 unsigned int num_rx_queues;
62fe0b40 1995 unsigned int real_num_rx_queues;
0a9627f2 1996
7acedaf5 1997 struct bpf_prog __rcu *xdp_prog;
3b47d303 1998 unsigned long gro_flush_timeout;
6f8b12d6 1999 int napi_defer_hard_irqs;
61391cde 2000 rx_handler_func_t __rcu *rx_handler;
2001 void __rcu *rx_handler_data;
e8a0464c 2002
4cda01e8 2003#ifdef CONFIG_NET_CLS_ACT
46209401 2004 struct mini_Qdisc __rcu *miniq_ingress;
d2788d34 2005#endif
24824a09 2006 struct netdev_queue __rcu *ingress_queue;
e687ad60 2007#ifdef CONFIG_NETFILTER_INGRESS
960632ec 2008 struct nf_hook_entries __rcu *nf_hooks_ingress;
e687ad60 2009#endif
d2788d34 2010
536721b1 2011 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
2012#ifdef CONFIG_RFS_ACCEL
2013 struct cpu_rmap *rx_cpu_rmap;
2014#endif
2015 struct hlist_node index_hlist;
cd13539b
ED
2016
2017/*
2018 * Cache lines mostly used on transmit path
2019 */
e8a0464c
DM
2020 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2021 unsigned int num_tx_queues;
fd2ea0a7 2022 unsigned int real_num_tx_queues;
af356afa 2023 struct Qdisc *qdisc;
0cd29503 2024 unsigned int tx_queue_len;
c3f26a26 2025 spinlock_t tx_global_lock;
75ccae62
THJ
2026
2027 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
cd13539b 2028
bf264145 2029#ifdef CONFIG_XPS
80d19669
AN
2030 struct xps_dev_maps __rcu *xps_cpus_map;
2031 struct xps_dev_maps __rcu *xps_rxqs_map;
bf264145 2032#endif
1f211a1b 2033#ifdef CONFIG_NET_CLS_ACT
46209401 2034 struct mini_Qdisc __rcu *miniq_egress;
1f211a1b 2035#endif
0c4f691f 2036
75ccae62
THJ
2037#ifdef CONFIG_NET_SCHED
2038 DECLARE_HASHTABLE (qdisc_hash, 4);
2039#endif
9356b8fc 2040 /* These may be needed for future network-power-down code. */
9356b8fc 2041 struct timer_list watchdog_timer;
75ccae62 2042 int watchdog_timeo;
9356b8fc 2043
1da177e4 2044 struct list_head todo_list;
75ccae62 2045 int __percpu *pcpu_refcnt;
1da177e4 2046
e014debe 2047 struct list_head link_watch_list;
572a103d 2048
1da177e4 2049 enum { NETREG_UNINITIALIZED=0,
b17a7c17 2050 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
2051 NETREG_UNREGISTERING, /* called unregister_netdevice */
2052 NETREG_UNREGISTERED, /* completed unregister todo */
2053 NETREG_RELEASED, /* called free_netdev */
937f1ba5 2054 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
2055 } reg_state:8;
2056
536721b1 2057 bool dismantle;
a2835763
PM
2058
2059 enum {
2060 RTNL_LINK_INITIALIZED,
2061 RTNL_LINK_INITIALIZING,
2062 } rtnl_link_state:16;
1da177e4 2063
cf124db5
DM
2064 bool needs_free_netdev;
2065 void (*priv_destructor)(struct net_device *dev);
1da177e4 2066
1da177e4 2067#ifdef CONFIG_NETPOLL
5fbee843 2068 struct netpoll_info __rcu *npinfo;
1da177e4 2069#endif
eae792b7 2070
0c5c9fb5 2071 possible_net_t nd_net;
4a1c5371 2072
4951704b 2073 /* mid-layer private */
a7855c78 2074 union {
536721b1
KK
2075 void *ml_priv;
2076 struct pcpu_lstats __percpu *lstats;
8f84985f 2077 struct pcpu_sw_netstats __percpu *tstats;
536721b1 2078 struct pcpu_dstats __percpu *dstats;
a7855c78 2079 };
536721b1 2080
fb585b44 2081#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 2082 struct garp_port __rcu *garp_port;
fb585b44
TK
2083#endif
2084#if IS_ENABLED(CONFIG_MRP)
febf018d 2085 struct mrp_port __rcu *mrp_port;
fb585b44 2086#endif
1da177e4 2087
5e82b4b2 2088 struct device dev;
0c509a6c 2089 const struct attribute_group *sysfs_groups[4];
a953be53 2090 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 2091
38f7b870 2092 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 2093
82cc1a7a
PWJ
2094 /* for setting kernel sock attribute on TCP connection setup */
2095#define GSO_MAX_SIZE 65536
2096 unsigned int gso_max_size;
30b678d8
BH
2097#define GSO_MAX_SEGS 65535
2098 u16 gso_max_segs;
743b03a8 2099
7a6b6f51 2100#ifdef CONFIG_DCB
32953543 2101 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 2102#endif
ffcfe25b 2103 s16 num_tc;
5e82b4b2
BH
2104 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2105 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 2106
d11ead75 2107#if IS_ENABLED(CONFIG_FCOE)
4d288d57 2108 unsigned int fcoe_ddp_xid;
5bc1421e 2109#endif
86f8515f 2110#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 2111 struct netprio_map __rcu *priomap;
4d288d57 2112#endif
5e82b4b2 2113 struct phy_device *phydev;
e679c9c1 2114 struct sfp_bus *sfp_bus;
ab92d68f
TY
2115 struct lock_class_key qdisc_tx_busylock_key;
2116 struct lock_class_key qdisc_running_key;
2117 struct lock_class_key qdisc_xmit_lock_key;
2118 struct lock_class_key addr_list_lock_key;
5e82b4b2 2119 bool proto_down;
61941143 2120 unsigned wol_enabled:1;
93642e14
JP
2121
2122 struct list_head net_notifier_list;
30e9bb84
AT
2123
2124#if IS_ENABLED(CONFIG_MACSEC)
2125 /* MACsec management functions */
2126 const struct macsec_ops *macsec_ops;
2127#endif
1da177e4 2128};
43cb76d9 2129#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 2130
b5cdae32
DM
2131static inline bool netif_elide_gro(const struct net_device *dev)
2132{
2133 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2134 return true;
2135 return false;
2136}
2137
1da177e4 2138#define NETDEV_ALIGN 32
1da177e4 2139
4f57c087
JF
2140static inline
2141int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2142{
2143 return dev->prio_tc_map[prio & TC_BITMASK];
2144}
2145
2146static inline
2147int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2148{
2149 if (tc >= dev->num_tc)
2150 return -EINVAL;
2151
2152 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2153 return 0;
2154}
2155
8d059b0f 2156int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
2157void netdev_reset_tc(struct net_device *dev);
2158int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2159int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
2160
2161static inline
2162int netdev_get_num_tc(struct net_device *dev)
2163{
2164 return dev->num_tc;
2165}
2166
ffcfe25b
AD
2167void netdev_unbind_sb_channel(struct net_device *dev,
2168 struct net_device *sb_dev);
2169int netdev_bind_sb_channel_queue(struct net_device *dev,
2170 struct net_device *sb_dev,
2171 u8 tc, u16 count, u16 offset);
2172int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2173static inline int netdev_get_sb_channel(struct net_device *dev)
2174{
2175 return max_t(int, -dev->num_tc, 0);
2176}
2177
e8a0464c
DM
2178static inline
2179struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2180 unsigned int index)
2181{
2182 return &dev->_tx[index];
2183}
2184
10c51b56
DB
2185static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2186 const struct sk_buff *skb)
2187{
2188 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2189}
2190
e8a0464c
DM
2191static inline void netdev_for_each_tx_queue(struct net_device *dev,
2192 void (*f)(struct net_device *,
2193 struct netdev_queue *,
2194 void *),
2195 void *arg)
2196{
2197 unsigned int i;
2198
2199 for (i = 0; i < dev->num_tx_queues; i++)
2200 f(dev, &dev->_tx[i], arg);
2201}
2202
b71b5837
PA
2203u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2204 struct net_device *sb_dev);
4bd97d51
PA
2205struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2206 struct sk_buff *skb,
2207 struct net_device *sb_dev);
8c4c49df 2208
871b642a
PA
2209/* returns the headroom that the master device needs to take in account
2210 * when forwarding to this dev
2211 */
2212static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2213{
2214 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2215}
2216
2217static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2218{
2219 if (dev->netdev_ops->ndo_set_rx_headroom)
2220 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2221}
2222
2223/* set the device rx headroom to the dev's default */
2224static inline void netdev_reset_rx_headroom(struct net_device *dev)
2225{
2226 netdev_set_rx_headroom(dev, -1);
2227}
2228
c346dca1
YH
2229/*
2230 * Net namespace inlines
2231 */
2232static inline
2233struct net *dev_net(const struct net_device *dev)
2234{
c2d9ba9b 2235 return read_pnet(&dev->nd_net);
c346dca1
YH
2236}
2237
2238static inline
f5aa23fd 2239void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2240{
0c5c9fb5 2241 write_pnet(&dev->nd_net, net);
c346dca1
YH
2242}
2243
bea3348e
SH
2244/**
2245 * netdev_priv - access network device private data
2246 * @dev: network device
2247 *
2248 * Get network device private data
2249 */
6472ce60 2250static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2251{
1ce8e7b5 2252 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2253}
2254
1da177e4
LT
2255/* Set the sysfs physical device reference for the network logical device
2256 * if set prior to registration will cause a symlink during initialization.
2257 */
43cb76d9 2258#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2259
384912ed 2260/* Set the sysfs device type for the network logical device to allow
3f79410c 2261 * fine-grained identification of different network device types. For
5e82b4b2 2262 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2263 */
2264#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2265
82dc3c63
ED
2266/* Default NAPI poll() weight
2267 * Device drivers are strongly advised to not use bigger value
2268 */
2269#define NAPI_POLL_WEIGHT 64
2270
3b582cc1 2271/**
5e82b4b2 2272 * netif_napi_add - initialize a NAPI context
3b582cc1 2273 * @dev: network device
5e82b4b2 2274 * @napi: NAPI context
3b582cc1
SH
2275 * @poll: polling function
2276 * @weight: default weight
2277 *
5e82b4b2
BH
2278 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2279 * *any* of the other NAPI-related functions.
3b582cc1 2280 */
d565b0a1
HX
2281void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2282 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2283
d64b5e85 2284/**
5e82b4b2 2285 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2286 * @dev: network device
5e82b4b2 2287 * @napi: NAPI context
d64b5e85
ED
2288 * @poll: polling function
2289 * @weight: default weight
2290 *
2291 * This variant of netif_napi_add() should be used from drivers using NAPI
2292 * to exclusively poll a TX queue.
2293 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2294 */
2295static inline void netif_tx_napi_add(struct net_device *dev,
2296 struct napi_struct *napi,
2297 int (*poll)(struct napi_struct *, int),
2298 int weight)
2299{
2300 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2301 netif_napi_add(dev, napi, poll, weight);
2302}
2303
d8156534 2304/**
5e82b4b2
BH
2305 * netif_napi_del - remove a NAPI context
2306 * @napi: NAPI context
d8156534 2307 *
5e82b4b2 2308 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2309 */
d565b0a1
HX
2310void netif_napi_del(struct napi_struct *napi);
2311
2312struct napi_gro_cb {
78a478d0 2313 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2314 void *frag0;
78a478d0 2315
7489594c
HX
2316 /* Length of frag0. */
2317 unsigned int frag0_len;
2318
86911732 2319 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2320 int data_offset;
86911732 2321
d565b0a1 2322 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2323 u16 flush;
2324
2325 /* Save the IP ID here and check when we get to the transport layer */
2326 u16 flush_id;
d565b0a1
HX
2327
2328 /* Number of segments aggregated. */
2e71a6f8
ED
2329 u16 count;
2330
15e2396d
TH
2331 /* Start offset for remote checksum offload */
2332 u16 gro_remcsum_start;
2333
2e71a6f8
ED
2334 /* jiffies when first packet was created/queued */
2335 unsigned long age;
86347245 2336
afe93325 2337 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2338 u16 proto;
2339
baa32ff4
TH
2340 /* This is non-zero if the packet may be of the same flow. */
2341 u8 same_flow:1;
2342
fac8e0f5
JG
2343 /* Used in tunnel GRO receive */
2344 u8 encap_mark:1;
573e8fca
TH
2345
2346 /* GRO checksum is valid */
2347 u8 csum_valid:1;
2348
662880f4
TH
2349 /* Number of checksums via CHECKSUM_UNNECESSARY */
2350 u8 csum_cnt:3;
c3c7c254 2351
baa32ff4
TH
2352 /* Free the skb? */
2353 u8 free:2;
2354#define NAPI_GRO_FREE 1
2355#define NAPI_GRO_FREE_STOLEN_HEAD 2
2356
efc98d08
TH
2357 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2358 u8 is_ipv6:1;
2359
a0ca153f
AD
2360 /* Used in GRE, set in fou/gue_gro_receive */
2361 u8 is_fou:1;
2362
1530545e
AD
2363 /* Used to determine if flush_id can be ignored */
2364 u8 is_atomic:1;
2365
fcd91dd4
SD
2366 /* Number of gro_receive callbacks this packet already went through */
2367 u8 recursion_counter:4;
2368
3a1296a3
SK
2369 /* GRO is done by frag_list pointer chaining. */
2370 u8 is_flist:1;
baa32ff4 2371
bf5a755f
JC
2372 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2373 __wsum csum;
2374
c3c7c254
ED
2375 /* used in skb_gro_receive() slow path */
2376 struct sk_buff *last;
d565b0a1
HX
2377};
2378
2379#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2380
fcd91dd4
SD
2381#define GRO_RECURSION_LIMIT 15
2382static inline int gro_recursion_inc_test(struct sk_buff *skb)
2383{
2384 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2385}
2386
d4546c25
DM
2387typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2388static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2389 struct list_head *head,
2390 struct sk_buff *skb)
fcd91dd4
SD
2391{
2392 if (unlikely(gro_recursion_inc_test(skb))) {
2393 NAPI_GRO_CB(skb)->flush |= 1;
2394 return NULL;
2395 }
2396
2397 return cb(head, skb);
2398}
2399
d4546c25
DM
2400typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2401 struct sk_buff *);
2402static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2403 struct sock *sk,
2404 struct list_head *head,
2405 struct sk_buff *skb)
fcd91dd4
SD
2406{
2407 if (unlikely(gro_recursion_inc_test(skb))) {
2408 NAPI_GRO_CB(skb)->flush |= 1;
2409 return NULL;
2410 }
2411
2412 return cb(sk, head, skb);
2413}
2414
1da177e4 2415struct packet_type {
f2ccd8fa 2416 __be16 type; /* This is really htons(ether_type). */
fa788d98 2417 bool ignore_outgoing;
f2ccd8fa
DM
2418 struct net_device *dev; /* NULL is wildcarded here */
2419 int (*func) (struct sk_buff *,
2420 struct net_device *,
2421 struct packet_type *,
2422 struct net_device *);
17266ee9
EC
2423 void (*list_func) (struct list_head *,
2424 struct packet_type *,
2425 struct net_device *);
c0de08d0
EL
2426 bool (*id_match)(struct packet_type *ptype,
2427 struct sock *sk);
1da177e4
LT
2428 void *af_packet_priv;
2429 struct list_head list;
2430};
2431
f191a1d1 2432struct offload_callbacks {
576a30eb 2433 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2434 netdev_features_t features);
d4546c25
DM
2435 struct sk_buff *(*gro_receive)(struct list_head *head,
2436 struct sk_buff *skb);
299603e8 2437 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2438};
2439
2440struct packet_offload {
2441 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2442 u16 priority;
f191a1d1
VY
2443 struct offload_callbacks callbacks;
2444 struct list_head list;
1da177e4
LT
2445};
2446
5e82b4b2 2447/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2448struct pcpu_sw_netstats {
2449 u64 rx_packets;
2450 u64 rx_bytes;
2451 u64 tx_packets;
2452 u64 tx_bytes;
2453 struct u64_stats_sync syncp;
9a5ee462 2454} __aligned(4 * sizeof(u64));
52bb6677
LR
2455
2456struct pcpu_lstats {
fd2f4737
ED
2457 u64_stats_t packets;
2458 u64_stats_t bytes;
52bb6677 2459 struct u64_stats_sync syncp;
9a5ee462 2460} __aligned(2 * sizeof(u64));
8f84985f 2461
de7d5084
ED
2462void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2463
dd5382a0
ED
2464static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2465{
2466 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2467
2468 u64_stats_update_begin(&lstats->syncp);
fd2f4737
ED
2469 u64_stats_add(&lstats->bytes, len);
2470 u64_stats_inc(&lstats->packets);
dd5382a0
ED
2471 u64_stats_update_end(&lstats->syncp);
2472}
2473
aabc92bb
PNA
2474#define __netdev_alloc_pcpu_stats(type, gfp) \
2475({ \
2476 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2477 if (pcpu_stats) { \
2478 int __cpu; \
2479 for_each_possible_cpu(__cpu) { \
2480 typeof(type) *stat; \
2481 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2482 u64_stats_init(&stat->syncp); \
2483 } \
2484 } \
2485 pcpu_stats; \
1c213bd2
WC
2486})
2487
aabc92bb 2488#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2489 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2490
764f5e54
JP
2491enum netdev_lag_tx_type {
2492 NETDEV_LAG_TX_TYPE_UNKNOWN,
2493 NETDEV_LAG_TX_TYPE_RANDOM,
2494 NETDEV_LAG_TX_TYPE_BROADCAST,
2495 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2496 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2497 NETDEV_LAG_TX_TYPE_HASH,
2498};
2499
f44aa9ef
JH
2500enum netdev_lag_hash {
2501 NETDEV_LAG_HASH_NONE,
2502 NETDEV_LAG_HASH_L2,
2503 NETDEV_LAG_HASH_L34,
2504 NETDEV_LAG_HASH_L23,
2505 NETDEV_LAG_HASH_E23,
2506 NETDEV_LAG_HASH_E34,
2507 NETDEV_LAG_HASH_UNKNOWN,
2508};
2509
764f5e54
JP
2510struct netdev_lag_upper_info {
2511 enum netdev_lag_tx_type tx_type;
f44aa9ef 2512 enum netdev_lag_hash hash_type;
764f5e54
JP
2513};
2514
fb1b2e3c
JP
2515struct netdev_lag_lower_state_info {
2516 u8 link_up : 1,
2517 tx_enabled : 1;
2518};
2519
1da177e4
LT
2520#include <linux/notifier.h>
2521
ede2762d
KT
2522/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2523 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2524 * adding new types.
dcfe1421 2525 */
ede2762d
KT
2526enum netdev_cmd {
2527 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2528 NETDEV_DOWN,
2529 NETDEV_REBOOT, /* Tell a protocol stack a network interface
dcfe1421
AW
2530 detected a hardware crash and restarted
2531 - we can use this eg to kick tcp sessions
2532 once done */
ede2762d
KT
2533 NETDEV_CHANGE, /* Notify device state change */
2534 NETDEV_REGISTER,
2535 NETDEV_UNREGISTER,
2536 NETDEV_CHANGEMTU, /* notify after mtu change happened */
1570415f
PM
2537 NETDEV_CHANGEADDR, /* notify after the address change */
2538 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
ede2762d
KT
2539 NETDEV_GOING_DOWN,
2540 NETDEV_CHANGENAME,
2541 NETDEV_FEAT_CHANGE,
2542 NETDEV_BONDING_FAILOVER,
2543 NETDEV_PRE_UP,
2544 NETDEV_PRE_TYPE_CHANGE,
2545 NETDEV_POST_TYPE_CHANGE,
2546 NETDEV_POST_INIT,
ede2762d
KT
2547 NETDEV_RELEASE,
2548 NETDEV_NOTIFY_PEERS,
2549 NETDEV_JOIN,
2550 NETDEV_CHANGEUPPER,
2551 NETDEV_RESEND_IGMP,
2552 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2553 NETDEV_CHANGEINFODATA,
2554 NETDEV_BONDING_INFO,
2555 NETDEV_PRECHANGEUPPER,
2556 NETDEV_CHANGELOWERSTATE,
2557 NETDEV_UDP_TUNNEL_PUSH_INFO,
2558 NETDEV_UDP_TUNNEL_DROP_INFO,
2559 NETDEV_CHANGE_TX_QUEUE_LEN,
9daae9bd
GP
2560 NETDEV_CVLAN_FILTER_PUSH_INFO,
2561 NETDEV_CVLAN_FILTER_DROP_INFO,
2562 NETDEV_SVLAN_FILTER_PUSH_INFO,
2563 NETDEV_SVLAN_FILTER_DROP_INFO,
ede2762d
KT
2564};
2565const char *netdev_cmd_to_name(enum netdev_cmd cmd);
dcfe1421 2566
f629d208
JP
2567int register_netdevice_notifier(struct notifier_block *nb);
2568int unregister_netdevice_notifier(struct notifier_block *nb);
a30c7b42
JP
2569int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2570int unregister_netdevice_notifier_net(struct net *net,
2571 struct notifier_block *nb);
93642e14
JP
2572int register_netdevice_notifier_dev_net(struct net_device *dev,
2573 struct notifier_block *nb,
2574 struct netdev_net_notifier *nn);
2575int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2576 struct notifier_block *nb,
2577 struct netdev_net_notifier *nn);
351638e7
JP
2578
2579struct netdev_notifier_info {
51d0c047
DA
2580 struct net_device *dev;
2581 struct netlink_ext_ack *extack;
351638e7
JP
2582};
2583
af7d6cce
SD
2584struct netdev_notifier_info_ext {
2585 struct netdev_notifier_info info; /* must be first */
2586 union {
2587 u32 mtu;
2588 } ext;
2589};
2590
be9efd36
JP
2591struct netdev_notifier_change_info {
2592 struct netdev_notifier_info info; /* must be first */
2593 unsigned int flags_changed;
2594};
2595
0e4ead9d
JP
2596struct netdev_notifier_changeupper_info {
2597 struct netdev_notifier_info info; /* must be first */
2598 struct net_device *upper_dev; /* new upper dev */
2599 bool master; /* is upper dev master */
5e82b4b2 2600 bool linking; /* is the notification for link or unlink */
29bf24af 2601 void *upper_info; /* upper dev info */
0e4ead9d
JP
2602};
2603
04d48266
JP
2604struct netdev_notifier_changelowerstate_info {
2605 struct netdev_notifier_info info; /* must be first */
2606 void *lower_state_info; /* is lower dev state */
2607};
2608
1570415f
PM
2609struct netdev_notifier_pre_changeaddr_info {
2610 struct netdev_notifier_info info; /* must be first */
2611 const unsigned char *dev_addr;
2612};
2613
75538c2b
CW
2614static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2615 struct net_device *dev)
2616{
2617 info->dev = dev;
51d0c047 2618 info->extack = NULL;
75538c2b
CW
2619}
2620
351638e7
JP
2621static inline struct net_device *
2622netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2623{
2624 return info->dev;
2625}
2626
51d0c047
DA
2627static inline struct netlink_ext_ack *
2628netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2629{
2630 return info->extack;
2631}
2632
f629d208 2633int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2634
2635
1da177e4
LT
2636extern rwlock_t dev_base_lock; /* Device list lock */
2637
881d966b
EB
2638#define for_each_netdev(net, d) \
2639 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2640#define for_each_netdev_reverse(net, d) \
2641 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2642#define for_each_netdev_rcu(net, d) \
2643 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2644#define for_each_netdev_safe(net, d, n) \
2645 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2646#define for_each_netdev_continue(net, d) \
2647 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
afa0df59
JP
2648#define for_each_netdev_continue_reverse(net, d) \
2649 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2650 dev_list)
254245d2 2651#define for_each_netdev_continue_rcu(net, d) \
2652 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2653#define for_each_netdev_in_bond_rcu(bond, slave) \
2654 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2655 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2656#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2657
a050c33f
DL
2658static inline struct net_device *next_net_device(struct net_device *dev)
2659{
2660 struct list_head *lh;
2661 struct net *net;
2662
c346dca1 2663 net = dev_net(dev);
a050c33f
DL
2664 lh = dev->dev_list.next;
2665 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2666}
2667
ce81b76a
ED
2668static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2669{
2670 struct list_head *lh;
2671 struct net *net;
2672
2673 net = dev_net(dev);
ccf43438 2674 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2675 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2676}
2677
a050c33f
DL
2678static inline struct net_device *first_net_device(struct net *net)
2679{
2680 return list_empty(&net->dev_base_head) ? NULL :
2681 net_device_entry(net->dev_base_head.next);
2682}
7562f876 2683
ccf43438
ED
2684static inline struct net_device *first_net_device_rcu(struct net *net)
2685{
2686 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2687
2688 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2689}
2690
f629d208
JP
2691int netdev_boot_setup_check(struct net_device *dev);
2692unsigned long netdev_boot_base(const char *prefix, int unit);
2693struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2694 const char *hwaddr);
2695struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2696struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2697void dev_add_pack(struct packet_type *pt);
2698void dev_remove_pack(struct packet_type *pt);
2699void __dev_remove_pack(struct packet_type *pt);
2700void dev_add_offload(struct packet_offload *po);
2701void dev_remove_offload(struct packet_offload *po);
f629d208 2702
a54acb3a 2703int dev_get_iflink(const struct net_device *dev);
fc4099f1 2704int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2705struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2706 unsigned short mask);
f629d208
JP
2707struct net_device *dev_get_by_name(struct net *net, const char *name);
2708struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2709struct net_device *__dev_get_by_name(struct net *net, const char *name);
2710int dev_alloc_name(struct net_device *dev, const char *name);
00f54e68 2711int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
7051b88a 2712void dev_close(struct net_device *dev);
2713void dev_close_many(struct list_head *head, bool unlink);
f629d208 2714void dev_disable_lro(struct net_device *dev);
0c4b51f0 2715int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
a4ea8a3d 2716u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
a350ecce 2717 struct net_device *sb_dev);
a4ea8a3d 2718u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
a350ecce 2719 struct net_device *sb_dev);
2b4aa3ce 2720int dev_queue_xmit(struct sk_buff *skb);
eadec877 2721int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
865b03f2 2722int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
f629d208
JP
2723int register_netdevice(struct net_device *dev);
2724void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2725void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2726static inline void unregister_netdevice(struct net_device *dev)
2727{
2728 unregister_netdevice_queue(dev, NULL);
2729}
2730
f629d208
JP
2731int netdev_refcnt_read(const struct net_device *dev);
2732void free_netdev(struct net_device *dev);
74d332c1 2733void netdev_freemem(struct net_device *dev);
f629d208
JP
2734void synchronize_net(void);
2735int init_dummy_netdev(struct net_device *dev);
937f1ba5 2736
f629d208
JP
2737struct net_device *dev_get_by_index(struct net *net, int ifindex);
2738struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2739struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2740struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2741int netdev_get_name(struct net *net, char *name, int ifindex);
2742int dev_restart(struct net_device *dev);
d4546c25 2743int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
3a1296a3 2744int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
86911732
HX
2745
2746static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2747{
2748 return NAPI_GRO_CB(skb)->data_offset;
2749}
2750
2751static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2752{
2753 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2754}
2755
2756static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2757{
2758 NAPI_GRO_CB(skb)->data_offset += len;
2759}
2760
a5b1cf28
HX
2761static inline void *skb_gro_header_fast(struct sk_buff *skb,
2762 unsigned int offset)
86911732 2763{
a5b1cf28
HX
2764 return NAPI_GRO_CB(skb)->frag0 + offset;
2765}
78a478d0 2766
a5b1cf28
HX
2767static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2768{
2769 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2770}
78a478d0 2771
57ea52a8
HX
2772static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2773{
2774 NAPI_GRO_CB(skb)->frag0 = NULL;
2775 NAPI_GRO_CB(skb)->frag0_len = 0;
2776}
2777
a5b1cf28
HX
2778static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2779 unsigned int offset)
2780{
17dd759c
HX
2781 if (!pskb_may_pull(skb, hlen))
2782 return NULL;
2783
57ea52a8 2784 skb_gro_frag0_invalidate(skb);
17dd759c 2785 return skb->data + offset;
86911732 2786}
1da177e4 2787
36e7b1b8
HX
2788static inline void *skb_gro_network_header(struct sk_buff *skb)
2789{
78d3fd0b
HX
2790 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2791 skb_network_offset(skb);
36e7b1b8
HX
2792}
2793
bf5a755f
JC
2794static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2795 const void *start, unsigned int len)
2796{
573e8fca 2797 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2798 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2799 csum_partial(start, len, 0));
2800}
2801
573e8fca
TH
2802/* GRO checksum functions. These are logical equivalents of the normal
2803 * checksum functions (in skbuff.h) except that they operate on the GRO
2804 * offsets and fields in sk_buff.
2805 */
2806
2807__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2808
15e2396d
TH
2809static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2810{
b7fe10e5 2811 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2812}
2813
573e8fca
TH
2814static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2815 bool zero_okay,
2816 __sum16 check)
2817{
6edec0e6
TH
2818 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2819 skb_checksum_start_offset(skb) <
2820 skb_gro_offset(skb)) &&
15e2396d 2821 !skb_at_gro_remcsum_start(skb) &&
662880f4 2822 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2823 (!zero_okay || check));
2824}
2825
2826static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2827 __wsum psum)
2828{
2829 if (NAPI_GRO_CB(skb)->csum_valid &&
2830 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2831 return 0;
2832
2833 NAPI_GRO_CB(skb)->csum = psum;
2834
2835 return __skb_gro_checksum_complete(skb);
2836}
2837
573e8fca
TH
2838static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2839{
662880f4
TH
2840 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2841 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2842 NAPI_GRO_CB(skb)->csum_cnt--;
2843 } else {
2844 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2845 * verified a new top level checksum or an encapsulated one
2846 * during GRO. This saves work if we fallback to normal path.
2847 */
2848 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2849 }
2850}
2851
2852#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2853 compute_pseudo) \
2854({ \
2855 __sum16 __ret = 0; \
2856 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2857 __ret = __skb_gro_checksum_validate_complete(skb, \
2858 compute_pseudo(skb, proto)); \
219f1d79 2859 if (!__ret) \
573e8fca
TH
2860 skb_gro_incr_csum_unnecessary(skb); \
2861 __ret; \
2862})
2863
2864#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2865 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2866
2867#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2868 compute_pseudo) \
2869 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2870
2871#define skb_gro_checksum_simple_validate(skb) \
2872 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2873
d96535a1
TH
2874static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2875{
2876 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2877 !NAPI_GRO_CB(skb)->csum_valid);
2878}
2879
2880static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
b39c78b2 2881 __wsum pseudo)
d96535a1
TH
2882{
2883 NAPI_GRO_CB(skb)->csum = ~pseudo;
2884 NAPI_GRO_CB(skb)->csum_valid = 1;
2885}
2886
b39c78b2 2887#define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
2888do { \
2889 if (__skb_gro_checksum_convert_check(skb)) \
b39c78b2 2890 __skb_gro_checksum_convert(skb, \
d96535a1
TH
2891 compute_pseudo(skb, proto)); \
2892} while (0)
2893
26c4f7da
TH
2894struct gro_remcsum {
2895 int offset;
2896 __wsum delta;
2897};
2898
2899static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2900{
846cd667 2901 grc->offset = 0;
26c4f7da
TH
2902 grc->delta = 0;
2903}
2904
b7fe10e5
TH
2905static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2906 unsigned int off, size_t hdrlen,
2907 int start, int offset,
2908 struct gro_remcsum *grc,
2909 bool nopartial)
dcdc8994
TH
2910{
2911 __wsum delta;
b7fe10e5 2912 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2913
2914 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2915
15e2396d 2916 if (!nopartial) {
b7fe10e5
TH
2917 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2918 return ptr;
2919 }
2920
2921 ptr = skb_gro_header_fast(skb, off);
2922 if (skb_gro_header_hard(skb, off + plen)) {
2923 ptr = skb_gro_header_slow(skb, off + plen, off);
2924 if (!ptr)
2925 return NULL;
15e2396d
TH
2926 }
2927
b7fe10e5
TH
2928 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2929 start, offset);
dcdc8994
TH
2930
2931 /* Adjust skb->csum since we changed the packet */
dcdc8994 2932 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2933
b7fe10e5 2934 grc->offset = off + hdrlen + offset;
26c4f7da 2935 grc->delta = delta;
b7fe10e5
TH
2936
2937 return ptr;
dcdc8994
TH
2938}
2939
26c4f7da
TH
2940static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2941 struct gro_remcsum *grc)
2942{
b7fe10e5
TH
2943 void *ptr;
2944 size_t plen = grc->offset + sizeof(u16);
2945
26c4f7da
TH
2946 if (!grc->delta)
2947 return;
2948
b7fe10e5
TH
2949 ptr = skb_gro_header_fast(skb, grc->offset);
2950 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2951 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2952 if (!ptr)
2953 return;
2954 }
2955
2956 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2957}
dcdc8994 2958
25393d3f 2959#ifdef CONFIG_XFRM_OFFLOAD
d4546c25 2960static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
25393d3f
SK
2961{
2962 if (PTR_ERR(pp) != -EINPROGRESS)
2963 NAPI_GRO_CB(skb)->flush |= flush;
2964}
603d4cf8 2965static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
5cd3da4b 2966 struct sk_buff *pp,
603d4cf8
SD
2967 int flush,
2968 struct gro_remcsum *grc)
2969{
2970 if (PTR_ERR(pp) != -EINPROGRESS) {
2971 NAPI_GRO_CB(skb)->flush |= flush;
2972 skb_gro_remcsum_cleanup(skb, grc);
2973 skb->remcsum_offload = 0;
2974 }
2975}
25393d3f 2976#else
d4546c25 2977static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
5f114163
SK
2978{
2979 NAPI_GRO_CB(skb)->flush |= flush;
2980}
603d4cf8 2981static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
5cd3da4b 2982 struct sk_buff *pp,
603d4cf8
SD
2983 int flush,
2984 struct gro_remcsum *grc)
2985{
2986 NAPI_GRO_CB(skb)->flush |= flush;
2987 skb_gro_remcsum_cleanup(skb, grc);
2988 skb->remcsum_offload = 0;
2989}
25393d3f 2990#endif
5f114163 2991
0c4e8581
SH
2992static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2993 unsigned short type,
3b04ddde 2994 const void *daddr, const void *saddr,
95c96174 2995 unsigned int len)
0c4e8581 2996{
f1ecfd5d 2997 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2998 return 0;
3b04ddde
SH
2999
3000 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
3001}
3002
b95cce35
SH
3003static inline int dev_parse_header(const struct sk_buff *skb,
3004 unsigned char *haddr)
3005{
3006 const struct net_device *dev = skb->dev;
3007
1b83336b 3008 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 3009 return 0;
3b04ddde 3010 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
3011}
3012
e78b2915
MM
3013static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3014{
3015 const struct net_device *dev = skb->dev;
3016
3017 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3018 return 0;
3019 return dev->header_ops->parse_protocol(skb);
3020}
3021
2793a23a
WB
3022/* ll_header must have at least hard_header_len allocated */
3023static inline bool dev_validate_header(const struct net_device *dev,
3024 char *ll_header, int len)
3025{
3026 if (likely(len >= dev->hard_header_len))
3027 return true;
217e6fa2
WB
3028 if (len < dev->min_header_len)
3029 return false;
2793a23a
WB
3030
3031 if (capable(CAP_SYS_RAWIO)) {
3032 memset(ll_header + len, 0, dev->hard_header_len - len);
3033 return true;
3034 }
3035
3036 if (dev->header_ops && dev->header_ops->validate)
3037 return dev->header_ops->validate(ll_header, len);
3038
3039 return false;
3040}
3041
36fd633e
AV
3042typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3043 int len, int size);
f629d208 3044int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
3045static inline int unregister_gifconf(unsigned int family)
3046{
3047 return register_gifconf(family, NULL);
3048}
3049
99bbc707 3050#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 3051#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
3052struct sd_flow_limit {
3053 u64 count;
3054 unsigned int num_buckets;
3055 unsigned int history_head;
3056 u16 history[FLOW_LIMIT_HISTORY];
3057 u8 buckets[];
3058};
3059
3060extern int netdev_flow_limit_table_len;
3061#endif /* CONFIG_NET_FLOW_LIMIT */
3062
1da177e4 3063/*
5e82b4b2 3064 * Incoming packets are placed on per-CPU queues
1da177e4 3065 */
d94d9fee 3066struct softnet_data {
1da177e4 3067 struct list_head poll_list;
6e7676c1 3068 struct sk_buff_head process_queue;
1da177e4 3069
dee42870 3070 /* stats */
cd7b5396
DM
3071 unsigned int processed;
3072 unsigned int time_squeeze;
cd7b5396 3073 unsigned int received_rps;
fd793d89 3074#ifdef CONFIG_RPS
88751275 3075 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
3076#endif
3077#ifdef CONFIG_NET_FLOW_LIMIT
3078 struct sd_flow_limit __rcu *flow_limit;
3079#endif
3080 struct Qdisc *output_queue;
3081 struct Qdisc **output_queue_tailp;
3082 struct sk_buff *completion_queue;
f53c7239
SK
3083#ifdef CONFIG_XFRM_OFFLOAD
3084 struct sk_buff_head xfrm_backlog;
3085#endif
97cdcf37
FW
3086 /* written and read only by owning cpu: */
3087 struct {
3088 u16 recursion;
3089 u8 more;
3090 } xmit;
4cdb1e2e 3091#ifdef CONFIG_RPS
501e7ef5
ED
3092 /* input_queue_head should be written by cpu owning this struct,
3093 * and only read by other cpus. Worth using a cache line.
3094 */
3095 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3096
3097 /* Elements below can be accessed between CPUs for RPS/RFS */
966a9671 3098 call_single_data_t csd ____cacheline_aligned_in_smp;
88751275
ED
3099 struct softnet_data *rps_ipi_next;
3100 unsigned int cpu;
76cc8b13 3101 unsigned int input_queue_tail;
1e94d72f 3102#endif
95c96174 3103 unsigned int dropped;
0a9627f2 3104 struct sk_buff_head input_pkt_queue;
bea3348e 3105 struct napi_struct backlog;
99bbc707 3106
1da177e4
LT
3107};
3108
76cc8b13 3109static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
3110{
3111#ifdef CONFIG_RPS
76cc8b13
TH
3112 sd->input_queue_head++;
3113#endif
3114}
3115
3116static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3117 unsigned int *qtail)
3118{
3119#ifdef CONFIG_RPS
3120 *qtail = ++sd->input_queue_tail;
fec5e652
TH
3121#endif
3122}
3123
0a9627f2 3124DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 3125
97cdcf37
FW
3126static inline int dev_recursion_level(void)
3127{
28b05b92 3128 return this_cpu_read(softnet_data.xmit.recursion);
97cdcf37
FW
3129}
3130
3131#define XMIT_RECURSION_LIMIT 10
3132static inline bool dev_xmit_recursion(void)
3133{
3134 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3135 XMIT_RECURSION_LIMIT);
3136}
3137
3138static inline void dev_xmit_recursion_inc(void)
3139{
3140 __this_cpu_inc(softnet_data.xmit.recursion);
3141}
3142
3143static inline void dev_xmit_recursion_dec(void)
3144{
3145 __this_cpu_dec(softnet_data.xmit.recursion);
3146}
3147
f629d208 3148void __netif_schedule(struct Qdisc *q);
46e5da40 3149void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 3150
fd2ea0a7
DM
3151static inline void netif_tx_schedule_all(struct net_device *dev)
3152{
3153 unsigned int i;
3154
3155 for (i = 0; i < dev->num_tx_queues; i++)
3156 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3157}
3158
f9a7cbbf 3159static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 3160{
73466498 3161 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
3162}
3163
bea3348e
SH
3164/**
3165 * netif_start_queue - allow transmit
3166 * @dev: network device
3167 *
3168 * Allow upper layers to call the device hard_start_xmit routine.
3169 */
1da177e4
LT
3170static inline void netif_start_queue(struct net_device *dev)
3171{
e8a0464c 3172 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3173}
3174
fd2ea0a7
DM
3175static inline void netif_tx_start_all_queues(struct net_device *dev)
3176{
3177 unsigned int i;
3178
3179 for (i = 0; i < dev->num_tx_queues; i++) {
3180 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3181 netif_tx_start_queue(txq);
3182 }
3183}
3184
46e5da40 3185void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 3186
d29f749e
DJ
3187/**
3188 * netif_wake_queue - restart transmit
3189 * @dev: network device
3190 *
3191 * Allow upper layers to call the device hard_start_xmit routine.
3192 * Used for flow control when transmit resources are available.
3193 */
79d16385
DM
3194static inline void netif_wake_queue(struct net_device *dev)
3195{
e8a0464c 3196 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3197}
3198
fd2ea0a7
DM
3199static inline void netif_tx_wake_all_queues(struct net_device *dev)
3200{
3201 unsigned int i;
3202
3203 for (i = 0; i < dev->num_tx_queues; i++) {
3204 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3205 netif_tx_wake_queue(txq);
3206 }
3207}
3208
f9a7cbbf 3209static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 3210{
73466498 3211 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
3212}
3213
bea3348e
SH
3214/**
3215 * netif_stop_queue - stop transmitted packets
3216 * @dev: network device
3217 *
3218 * Stop upper layers calling the device hard_start_xmit routine.
3219 * Used for flow control when transmit resources are unavailable.
3220 */
1da177e4
LT
3221static inline void netif_stop_queue(struct net_device *dev)
3222{
e8a0464c 3223 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3224}
3225
a2029240 3226void netif_tx_stop_all_queues(struct net_device *dev);
ab92d68f 3227void netdev_update_lockdep_key(struct net_device *dev);
fd2ea0a7 3228
4d29515f 3229static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 3230{
73466498 3231 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
3232}
3233
bea3348e
SH
3234/**
3235 * netif_queue_stopped - test if transmit queue is flowblocked
3236 * @dev: network device
3237 *
3238 * Test if transmit queue on device is currently unable to send.
3239 */
4d29515f 3240static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 3241{
e8a0464c 3242 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3243}
3244
4d29515f 3245static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 3246{
73466498
TH
3247 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3248}
3249
8e2f1a63
DB
3250static inline bool
3251netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
3252{
3253 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3254}
3255
8e2f1a63
DB
3256static inline bool
3257netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3258{
3259 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3260}
3261
53511453
ED
3262/**
3263 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3264 * @dev_queue: pointer to transmit queue
3265 *
3266 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 3267 * to give appropriate hint to the CPU.
53511453
ED
3268 */
3269static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3270{
3271#ifdef CONFIG_BQL
3272 prefetchw(&dev_queue->dql.num_queued);
3273#endif
3274}
3275
3276/**
3277 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3278 * @dev_queue: pointer to transmit queue
3279 *
3280 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 3281 * to give appropriate hint to the CPU.
53511453
ED
3282 */
3283static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3284{
3285#ifdef CONFIG_BQL
3286 prefetchw(&dev_queue->dql.limit);
3287#endif
3288}
3289
c5d67bd7
TH
3290static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3291 unsigned int bytes)
3292{
114cf580
TH
3293#ifdef CONFIG_BQL
3294 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
3295
3296 if (likely(dql_avail(&dev_queue->dql) >= 0))
3297 return;
3298
3299 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3300
3301 /*
3302 * The XOFF flag must be set before checking the dql_avail below,
3303 * because in netdev_tx_completed_queue we update the dql_completed
3304 * before checking the XOFF flag.
3305 */
3306 smp_mb();
3307
3308 /* check again in case another CPU has just made room avail */
3309 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3310 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 3311#endif
c5d67bd7
TH
3312}
3313
3e59020a
ED
3314/* Variant of netdev_tx_sent_queue() for drivers that are aware
3315 * that they should not test BQL status themselves.
3316 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3317 * skb of a batch.
3318 * Returns true if the doorbell must be used to kick the NIC.
3319 */
3320static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3321 unsigned int bytes,
3322 bool xmit_more)
3323{
3324 if (xmit_more) {
3325#ifdef CONFIG_BQL
3326 dql_queued(&dev_queue->dql, bytes);
3327#endif
3328 return netif_tx_queue_stopped(dev_queue);
3329 }
3330 netdev_tx_sent_queue(dev_queue, bytes);
3331 return true;
3332}
3333
0042d0c8
FF
3334/**
3335 * netdev_sent_queue - report the number of bytes queued to hardware
3336 * @dev: network device
3337 * @bytes: number of bytes queued to the hardware device queue
3338 *
3339 * Report the number of bytes queued for sending/completion to the network
3340 * device hardware queue. @bytes should be a good approximation and should
3341 * exactly match netdev_completed_queue() @bytes
3342 */
c5d67bd7
TH
3343static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3344{
3345 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3346}
3347
620344c4
HK
3348static inline bool __netdev_sent_queue(struct net_device *dev,
3349 unsigned int bytes,
3350 bool xmit_more)
3351{
3352 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3353 xmit_more);
3354}
3355
c5d67bd7 3356static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 3357 unsigned int pkts, unsigned int bytes)
c5d67bd7 3358{
114cf580 3359#ifdef CONFIG_BQL
b37c0fbe
AD
3360 if (unlikely(!bytes))
3361 return;
3362
3363 dql_completed(&dev_queue->dql, bytes);
3364
3365 /*
3366 * Without the memory barrier there is a small possiblity that
3367 * netdev_tx_sent_queue will miss the update and cause the queue to
3368 * be stopped forever
3369 */
3370 smp_mb();
3371
f3acd33d 3372 if (unlikely(dql_avail(&dev_queue->dql) < 0))
b37c0fbe
AD
3373 return;
3374
3375 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3376 netif_schedule_queue(dev_queue);
114cf580 3377#endif
c5d67bd7
TH
3378}
3379
0042d0c8
FF
3380/**
3381 * netdev_completed_queue - report bytes and packets completed by device
3382 * @dev: network device
3383 * @pkts: actual number of packets sent over the medium
3384 * @bytes: actual number of bytes sent over the medium
3385 *
3386 * Report the number of bytes and packets transmitted by the network device
3387 * hardware queue over the physical medium, @bytes must exactly match the
3388 * @bytes amount passed to netdev_sent_queue()
3389 */
c5d67bd7 3390static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3391 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3392{
3393 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3394}
3395
3396static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3397{
114cf580 3398#ifdef CONFIG_BQL
5c490354 3399 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3400 dql_reset(&q->dql);
3401#endif
c5d67bd7
TH
3402}
3403
0042d0c8
FF
3404/**
3405 * netdev_reset_queue - reset the packets and bytes count of a network device
3406 * @dev_queue: network device
3407 *
3408 * Reset the bytes and packet count of a network device and clear the
3409 * software flow control OFF bit for this network device
3410 */
c5d67bd7
TH
3411static inline void netdev_reset_queue(struct net_device *dev_queue)
3412{
3413 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3414}
3415
b9507bda
DB
3416/**
3417 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3418 * @dev: network device
3419 * @queue_index: given tx queue index
3420 *
3421 * Returns 0 if given tx queue index >= number of device tx queues,
3422 * otherwise returns the originally passed tx queue index.
3423 */
3424static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3425{
3426 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3427 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3428 dev->name, queue_index,
3429 dev->real_num_tx_queues);
3430 return 0;
3431 }
3432
3433 return queue_index;
3434}
3435
bea3348e
SH
3436/**
3437 * netif_running - test if up
3438 * @dev: network device
3439 *
3440 * Test if the device has been brought up.
3441 */
4d29515f 3442static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3443{
3444 return test_bit(__LINK_STATE_START, &dev->state);
3445}
3446
f25f4e44 3447/*
5e82b4b2 3448 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3449 * stop, and a check if it's stopped. All other device management is
3450 * done at the overall netdevice level.
3451 * Also test the device if we're multiqueue.
3452 */
bea3348e
SH
3453
3454/**
3455 * netif_start_subqueue - allow sending packets on subqueue
3456 * @dev: network device
3457 * @queue_index: sub queue index
3458 *
3459 * Start individual transmit queue of a device with multiple transmit queues.
3460 */
f25f4e44
PWJ
3461static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3462{
fd2ea0a7 3463 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3464
3465 netif_tx_start_queue(txq);
f25f4e44
PWJ
3466}
3467
bea3348e
SH
3468/**
3469 * netif_stop_subqueue - stop sending packets on subqueue
3470 * @dev: network device
3471 * @queue_index: sub queue index
3472 *
3473 * Stop individual transmit queue of a device with multiple transmit queues.
3474 */
f25f4e44
PWJ
3475static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3476{
fd2ea0a7 3477 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3478 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3479}
3480
bea3348e
SH
3481/**
3482 * netif_subqueue_stopped - test status of subqueue
3483 * @dev: network device
3484 * @queue_index: sub queue index
3485 *
3486 * Check individual transmit queue of a device with multiple transmit queues.
3487 */
4d29515f
DM
3488static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3489 u16 queue_index)
f25f4e44 3490{
fd2ea0a7 3491 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3492
3493 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3494}
3495
4d29515f
DM
3496static inline bool netif_subqueue_stopped(const struct net_device *dev,
3497 struct sk_buff *skb)
668f895a
PE
3498{
3499 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3500}
bea3348e 3501
738b35cc
FF
3502/**
3503 * netif_wake_subqueue - allow sending packets on subqueue
3504 * @dev: network device
3505 * @queue_index: sub queue index
3506 *
3507 * Resume individual transmit queue of a device with multiple transmit queues.
3508 */
3509static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3510{
3511 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3512
3513 netif_tx_wake_queue(txq);
3514}
f25f4e44 3515
537c00de 3516#ifdef CONFIG_XPS
53af53ae 3517int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3518 u16 index);
80d19669
AN
3519int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3520 u16 index, bool is_rxqs_map);
3521
3522/**
3523 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3524 * @j: CPU/Rx queue index
3525 * @mask: bitmask of all cpus/rx queues
3526 * @nr_bits: number of bits in the bitmask
3527 *
3528 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3529 */
3530static inline bool netif_attr_test_mask(unsigned long j,
3531 const unsigned long *mask,
3532 unsigned int nr_bits)
3533{
3534 cpu_max_bits_warn(j, nr_bits);
3535 return test_bit(j, mask);
3536}
3537
3538/**
3539 * netif_attr_test_online - Test for online CPU/Rx queue
3540 * @j: CPU/Rx queue index
3541 * @online_mask: bitmask for CPUs/Rx queues that are online
3542 * @nr_bits: number of bits in the bitmask
3543 *
3544 * Returns true if a CPU/Rx queue is online.
3545 */
3546static inline bool netif_attr_test_online(unsigned long j,
3547 const unsigned long *online_mask,
3548 unsigned int nr_bits)
3549{
3550 cpu_max_bits_warn(j, nr_bits);
3551
3552 if (online_mask)
3553 return test_bit(j, online_mask);
3554
3555 return (j < nr_bits);
3556}
3557
3558/**
3559 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3560 * @n: CPU/Rx queue index
3561 * @srcp: the cpumask/Rx queue mask pointer
3562 * @nr_bits: number of bits in the bitmask
3563 *
3564 * Returns >= nr_bits if no further CPUs/Rx queues set.
3565 */
3566static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3567 unsigned int nr_bits)
3568{
3569 /* -1 is a legal arg here. */
3570 if (n != -1)
3571 cpu_max_bits_warn(n, nr_bits);
3572
3573 if (srcp)
3574 return find_next_bit(srcp, nr_bits, n + 1);
3575
3576 return n + 1;
3577}
3578
3579/**
a1fa83bd 3580 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
80d19669
AN
3581 * @n: CPU/Rx queue index
3582 * @src1p: the first CPUs/Rx queues mask pointer
3583 * @src2p: the second CPUs/Rx queues mask pointer
3584 * @nr_bits: number of bits in the bitmask
3585 *
3586 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3587 */
3588static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3589 const unsigned long *src2p,
3590 unsigned int nr_bits)
3591{
3592 /* -1 is a legal arg here. */
3593 if (n != -1)
3594 cpu_max_bits_warn(n, nr_bits);
3595
3596 if (src1p && src2p)
3597 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3598 else if (src1p)
3599 return find_next_bit(src1p, nr_bits, n + 1);
3600 else if (src2p)
3601 return find_next_bit(src2p, nr_bits, n + 1);
3602
3603 return n + 1;
3604}
537c00de
AD
3605#else
3606static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3607 const struct cpumask *mask,
537c00de
AD
3608 u16 index)
3609{
3610 return 0;
3611}
c9fbb2d2
KK
3612
3613static inline int __netif_set_xps_queue(struct net_device *dev,
3614 const unsigned long *mask,
3615 u16 index, bool is_rxqs_map)
3616{
3617 return 0;
3618}
537c00de
AD
3619#endif
3620
bea3348e
SH
3621/**
3622 * netif_is_multiqueue - test if device has multiple transmit queues
3623 * @dev: network device
3624 *
3625 * Check if device has multiple transmit queues
bea3348e 3626 */
4d29515f 3627static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3628{
a02cec21 3629 return dev->num_tx_queues > 1;
f25f4e44 3630}
1da177e4 3631
f629d208 3632int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3633
a953be53 3634#ifdef CONFIG_SYSFS
f629d208 3635int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3636#else
3637static inline int netif_set_real_num_rx_queues(struct net_device *dev,
c29c2ebd 3638 unsigned int rxqs)
62fe0b40 3639{
c29c2ebd 3640 dev->real_num_rx_queues = rxqs;
62fe0b40
BH
3641 return 0;
3642}
3643#endif
3644
65073a67
DB
3645static inline struct netdev_rx_queue *
3646__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3647{
3648 return dev->_rx + rxq;
3649}
3650
a953be53
MD
3651#ifdef CONFIG_SYSFS
3652static inline unsigned int get_netdev_rx_queue_index(
3653 struct netdev_rx_queue *queue)
3654{
3655 struct net_device *dev = queue->dev;
3656 int index = queue - dev->_rx;
3657
3658 BUG_ON(index >= dev->num_rx_queues);
3659 return index;
3660}
3661#endif
3662
16917b87 3663#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3664int netif_get_num_default_rss_queues(void);
16917b87 3665
e6247027
ED
3666enum skb_free_reason {
3667 SKB_REASON_CONSUMED,
3668 SKB_REASON_DROPPED,
3669};
3670
3671void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3672void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3673
e6247027
ED
3674/*
3675 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3676 * interrupt context or with hardware interrupts being disabled.
3677 * (in_irq() || irqs_disabled())
3678 *
3679 * We provide four helpers that can be used in following contexts :
3680 *
3681 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3682 * replacing kfree_skb(skb)
3683 *
3684 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3685 * Typically used in place of consume_skb(skb) in TX completion path
3686 *
3687 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3688 * replacing kfree_skb(skb)
3689 *
3690 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3691 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3692 */
e6247027
ED
3693static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3694{
3695 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3696}
3697
3698static inline void dev_consume_skb_irq(struct sk_buff *skb)
3699{
3700 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3701}
3702
3703static inline void dev_kfree_skb_any(struct sk_buff *skb)
3704{
3705 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3706}
3707
3708static inline void dev_consume_skb_any(struct sk_buff *skb)
3709{
3710 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3711}
1da177e4 3712
7c497478
JW
3713void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3714int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
f629d208
JP
3715int netif_rx(struct sk_buff *skb);
3716int netif_rx_ni(struct sk_buff *skb);
04eb4489 3717int netif_receive_skb(struct sk_buff *skb);
1c601d82 3718int netif_receive_skb_core(struct sk_buff *skb);
f6ad8c1b 3719void netif_receive_skb_list(struct list_head *head);
f629d208
JP
3720gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3721void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3722struct sk_buff *napi_get_frags(struct napi_struct *napi);
3723gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3724struct packet_offload *gro_find_receive_by_type(__be16 type);
3725struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3726
3727static inline void napi_free_frags(struct napi_struct *napi)
3728{
3729 kfree_skb(napi->skb);
3730 napi->skb = NULL;
3731}
3732
24b27fc4 3733bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3734int netdev_rx_handler_register(struct net_device *dev,
3735 rx_handler_func_t *rx_handler,
3736 void *rx_handler_data);
3737void netdev_rx_handler_unregister(struct net_device *dev);
3738
3739bool dev_valid_name(const char *name);
44c02a2c
AV
3740int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3741 bool *need_copyout);
36fd633e 3742int dev_ifconf(struct net *net, struct ifconf *, int);
f629d208
JP
3743int dev_ethtool(struct net *net, struct ifreq *);
3744unsigned int dev_get_flags(const struct net_device *);
6d040321
PM
3745int __dev_change_flags(struct net_device *dev, unsigned int flags,
3746 struct netlink_ext_ack *extack);
567c5e13
PM
3747int dev_change_flags(struct net_device *dev, unsigned int flags,
3748 struct netlink_ext_ack *extack);
cb178190
DM
3749void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3750 unsigned int gchanges);
f629d208
JP
3751int dev_change_name(struct net_device *, const char *);
3752int dev_set_alias(struct net_device *, const char *, size_t);
6c557001 3753int dev_get_alias(const struct net_device *, char *, size_t);
f629d208 3754int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3755int __dev_set_mtu(struct net_device *, int);
d836f5c6
ED
3756int dev_validate_mtu(struct net_device *dev, int mtu,
3757 struct netlink_ext_ack *extack);
7a4c53be
SH
3758int dev_set_mtu_ext(struct net_device *dev, int mtu,
3759 struct netlink_ext_ack *extack);
f629d208 3760int dev_set_mtu(struct net_device *, int);
6a643ddb 3761int dev_change_tx_queue_len(struct net_device *, unsigned long);
f629d208 3762void dev_set_group(struct net_device *, int);
d59cdf94
PM
3763int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3764 struct netlink_ext_ack *extack);
3a37a963
PM
3765int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3766 struct netlink_ext_ack *extack);
f629d208
JP
3767int dev_change_carrier(struct net_device *, bool new_carrier);
3768int dev_get_phys_port_id(struct net_device *dev,
02637fce 3769 struct netdev_phys_item_id *ppid);
db24a904
DA
3770int dev_get_phys_port_name(struct net_device *dev,
3771 char *name, size_t len);
d6abc596
FF
3772int dev_get_port_parent_id(struct net_device *dev,
3773 struct netdev_phys_item_id *ppid, bool recurse);
3774bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
d746d707 3775int dev_change_proto_down(struct net_device *dev, bool proto_down);
b5899679 3776int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
f53c7239 3777struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
ce93718f
DM
3778struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3779 struct netdev_queue *txq, int *ret);
d67b9cd2 3780
f4e63525 3781typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
d67b9cd2 3782int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
92234c8f 3783 int fd, int expected_fd, u32 flags);
a25717d2
JK
3784u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3785 enum bpf_netdev_command cmd);
84c6b868 3786int xdp_umem_query(struct net_device *dev, u16 queue_id);
d67b9cd2 3787
a0265d28 3788int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3789int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3790bool is_skb_forwardable(const struct net_device *dev,
3791 const struct sk_buff *skb);
1da177e4 3792
4e3264d2
MKL
3793static __always_inline int ____dev_forward_skb(struct net_device *dev,
3794 struct sk_buff *skb)
3795{
3796 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3797 unlikely(!is_skb_forwardable(dev, skb))) {
3798 atomic_long_inc(&dev->rx_dropped);
3799 kfree_skb(skb);
3800 return NET_RX_DROP;
3801 }
3802
3803 skb_scrub_packet(skb, true);
3804 skb->priority = 0;
3805 return 0;
3806}
3807
9f9a742d 3808bool dev_nit_active(struct net_device *dev);
74b20582
DA
3809void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3810
20380731 3811extern int netdev_budget;
7acf8a1e 3812extern unsigned int netdev_budget_usecs;
1da177e4
LT
3813
3814/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3815void netdev_run_todo(void);
1da177e4 3816
bea3348e
SH
3817/**
3818 * dev_put - release reference to device
3819 * @dev: network device
3820 *
9ef4429b 3821 * Release reference to device to allow it to be freed.
bea3348e 3822 */
1da177e4
LT
3823static inline void dev_put(struct net_device *dev)
3824{
933393f5 3825 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3826}
3827
bea3348e
SH
3828/**
3829 * dev_hold - get reference to device
3830 * @dev: network device
3831 *
9ef4429b 3832 * Hold reference to device to keep it from being freed.
bea3348e 3833 */
15333061
SH
3834static inline void dev_hold(struct net_device *dev)
3835{
933393f5 3836 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3837}
1da177e4
LT
3838
3839/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3840 * and _off may be called from IRQ context, but it is caller
3841 * who is responsible for serialization of these calls.
b00055aa
SR
3842 *
3843 * The name carrier is inappropriate, these functions should really be
3844 * called netif_lowerlayer_*() because they represent the state of any
3845 * kind of lower layer not just hardware media.
1da177e4
LT
3846 */
3847
f629d208
JP
3848void linkwatch_init_dev(struct net_device *dev);
3849void linkwatch_fire_event(struct net_device *dev);
3850void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3851
bea3348e
SH
3852/**
3853 * netif_carrier_ok - test if carrier present
3854 * @dev: network device
3855 *
3856 * Check if carrier is present on device
3857 */
4d29515f 3858static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3859{
3860 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3861}
3862
f629d208 3863unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3864
f629d208 3865void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3866
f629d208 3867void netif_carrier_on(struct net_device *dev);
1da177e4 3868
f629d208 3869void netif_carrier_off(struct net_device *dev);
1da177e4 3870
bea3348e
SH
3871/**
3872 * netif_dormant_on - mark device as dormant.
3873 * @dev: network device
3874 *
3875 * Mark device as dormant (as per RFC2863).
3876 *
3877 * The dormant state indicates that the relevant interface is not
3878 * actually in a condition to pass packets (i.e., it is not 'up') but is
3879 * in a "pending" state, waiting for some external event. For "on-
3880 * demand" interfaces, this new state identifies the situation where the
3881 * interface is waiting for events to place it in the up state.
bea3348e 3882 */
b00055aa
SR
3883static inline void netif_dormant_on(struct net_device *dev)
3884{
3885 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3886 linkwatch_fire_event(dev);
3887}
3888
bea3348e
SH
3889/**
3890 * netif_dormant_off - set device as not dormant.
3891 * @dev: network device
3892 *
3893 * Device is not in dormant state.
3894 */
b00055aa
SR
3895static inline void netif_dormant_off(struct net_device *dev)
3896{
3897 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3898 linkwatch_fire_event(dev);
3899}
3900
bea3348e 3901/**
8ecbc40a 3902 * netif_dormant - test if device is dormant
bea3348e
SH
3903 * @dev: network device
3904 *
8ecbc40a 3905 * Check if device is dormant.
bea3348e 3906 */
4d29515f 3907static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3908{
3909 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3910}
3911
3912
eec517cd
AL
3913/**
3914 * netif_testing_on - mark device as under test.
3915 * @dev: network device
3916 *
3917 * Mark device as under test (as per RFC2863).
3918 *
3919 * The testing state indicates that some test(s) must be performed on
3920 * the interface. After completion, of the test, the interface state
3921 * will change to up, dormant, or down, as appropriate.
3922 */
3923static inline void netif_testing_on(struct net_device *dev)
3924{
3925 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
3926 linkwatch_fire_event(dev);
3927}
3928
3929/**
3930 * netif_testing_off - set device as not under test.
3931 * @dev: network device
3932 *
3933 * Device is not in testing state.
3934 */
3935static inline void netif_testing_off(struct net_device *dev)
3936{
3937 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
3938 linkwatch_fire_event(dev);
3939}
3940
3941/**
3942 * netif_testing - test if device is under test
3943 * @dev: network device
3944 *
3945 * Check if device is under test
3946 */
3947static inline bool netif_testing(const struct net_device *dev)
3948{
3949 return test_bit(__LINK_STATE_TESTING, &dev->state);
3950}
3951
3952
bea3348e
SH
3953/**
3954 * netif_oper_up - test if device is operational
3955 * @dev: network device
3956 *
3957 * Check if carrier is operational
3958 */
4d29515f 3959static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3960{
b00055aa
SR
3961 return (dev->operstate == IF_OPER_UP ||
3962 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3963}
3964
bea3348e
SH
3965/**
3966 * netif_device_present - is device available or removed
3967 * @dev: network device
3968 *
3969 * Check if device has not been removed from system.
3970 */
4d29515f 3971static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3972{
3973 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3974}
3975
f629d208 3976void netif_device_detach(struct net_device *dev);
1da177e4 3977
f629d208 3978void netif_device_attach(struct net_device *dev);
1da177e4
LT
3979
3980/*
3981 * Network interface message level settings
3982 */
1da177e4
LT
3983
3984enum {
6a94b8cc
MK
3985 NETIF_MSG_DRV_BIT,
3986 NETIF_MSG_PROBE_BIT,
3987 NETIF_MSG_LINK_BIT,
3988 NETIF_MSG_TIMER_BIT,
3989 NETIF_MSG_IFDOWN_BIT,
3990 NETIF_MSG_IFUP_BIT,
3991 NETIF_MSG_RX_ERR_BIT,
3992 NETIF_MSG_TX_ERR_BIT,
3993 NETIF_MSG_TX_QUEUED_BIT,
3994 NETIF_MSG_INTR_BIT,
3995 NETIF_MSG_TX_DONE_BIT,
3996 NETIF_MSG_RX_STATUS_BIT,
3997 NETIF_MSG_PKTDATA_BIT,
3998 NETIF_MSG_HW_BIT,
3999 NETIF_MSG_WOL_BIT,
4000
4001 /* When you add a new bit above, update netif_msg_class_names array
4002 * in net/ethtool/common.c
4003 */
4004 NETIF_MSG_CLASS_COUNT,
1da177e4 4005};
6a94b8cc
MK
4006/* Both ethtool_ops interface and internal driver implementation use u32 */
4007static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4008
4009#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4010#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4011
4012#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4013#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4014#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4015#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4016#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4017#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4018#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4019#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4020#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4021#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4022#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4023#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4024#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4025#define NETIF_MSG_HW __NETIF_MSG(HW)
4026#define NETIF_MSG_WOL __NETIF_MSG(WOL)
1da177e4
LT
4027
4028#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4029#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4030#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4031#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4032#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4033#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4034#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4035#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4036#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4037#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4038#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4039#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4040#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4041#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4042#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4043
4044static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4045{
4046 /* use default */
4047 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4048 return default_msg_enable_bits;
4049 if (debug_value == 0) /* no output */
4050 return 0;
4051 /* set low N bits */
f4d7b3e2 4052 return (1U << debug_value) - 1;
1da177e4
LT
4053}
4054
c773e847 4055static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 4056{
c773e847
DM
4057 spin_lock(&txq->_xmit_lock);
4058 txq->xmit_lock_owner = cpu;
22dd7495
JHS
4059}
4060
5a717f4f
MT
4061static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4062{
4063 __acquire(&txq->_xmit_lock);
4064 return true;
4065}
4066
4067static inline void __netif_tx_release(struct netdev_queue *txq)
4068{
4069 __release(&txq->_xmit_lock);
4070}
4071
fd2ea0a7
DM
4072static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4073{
4074 spin_lock_bh(&txq->_xmit_lock);
4075 txq->xmit_lock_owner = smp_processor_id();
4076}
4077
4d29515f 4078static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 4079{
4d29515f 4080 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
4081 if (likely(ok))
4082 txq->xmit_lock_owner = smp_processor_id();
4083 return ok;
4084}
4085
4086static inline void __netif_tx_unlock(struct netdev_queue *txq)
4087{
4088 txq->xmit_lock_owner = -1;
4089 spin_unlock(&txq->_xmit_lock);
4090}
4091
4092static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4093{
4094 txq->xmit_lock_owner = -1;
4095 spin_unlock_bh(&txq->_xmit_lock);
4096}
4097
08baf561
ED
4098static inline void txq_trans_update(struct netdev_queue *txq)
4099{
4100 if (txq->xmit_lock_owner != -1)
4101 txq->trans_start = jiffies;
4102}
4103
ba162f8e
FW
4104/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4105static inline void netif_trans_update(struct net_device *dev)
4106{
9b36627a
FW
4107 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4108
4109 if (txq->trans_start != jiffies)
4110 txq->trans_start = jiffies;
ba162f8e
FW
4111}
4112
d29f749e
DJ
4113/**
4114 * netif_tx_lock - grab network device transmit lock
4115 * @dev: network device
d29f749e
DJ
4116 *
4117 * Get network device transmit lock
4118 */
22dd7495
JHS
4119static inline void netif_tx_lock(struct net_device *dev)
4120{
e8a0464c 4121 unsigned int i;
c3f26a26 4122 int cpu;
c773e847 4123
c3f26a26
DM
4124 spin_lock(&dev->tx_global_lock);
4125 cpu = smp_processor_id();
e8a0464c
DM
4126 for (i = 0; i < dev->num_tx_queues; i++) {
4127 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
4128
4129 /* We are the only thread of execution doing a
4130 * freeze, but we have to grab the _xmit_lock in
4131 * order to synchronize with threads which are in
4132 * the ->hard_start_xmit() handler and already
4133 * checked the frozen bit.
4134 */
e8a0464c 4135 __netif_tx_lock(txq, cpu);
c3f26a26
DM
4136 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4137 __netif_tx_unlock(txq);
e8a0464c 4138 }
932ff279
HX
4139}
4140
4141static inline void netif_tx_lock_bh(struct net_device *dev)
4142{
e8a0464c
DM
4143 local_bh_disable();
4144 netif_tx_lock(dev);
932ff279
HX
4145}
4146
932ff279
HX
4147static inline void netif_tx_unlock(struct net_device *dev)
4148{
e8a0464c
DM
4149 unsigned int i;
4150
4151 for (i = 0; i < dev->num_tx_queues; i++) {
4152 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 4153
c3f26a26
DM
4154 /* No need to grab the _xmit_lock here. If the
4155 * queue is not stopped for another reason, we
4156 * force a schedule.
4157 */
4158 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 4159 netif_schedule_queue(txq);
c3f26a26
DM
4160 }
4161 spin_unlock(&dev->tx_global_lock);
932ff279
HX
4162}
4163
4164static inline void netif_tx_unlock_bh(struct net_device *dev)
4165{
e8a0464c
DM
4166 netif_tx_unlock(dev);
4167 local_bh_enable();
932ff279
HX
4168}
4169
c773e847 4170#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 4171 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 4172 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
4173 } else { \
4174 __netif_tx_acquire(txq); \
22dd7495
JHS
4175 } \
4176}
4177
5efeac44
EB
4178#define HARD_TX_TRYLOCK(dev, txq) \
4179 (((dev->features & NETIF_F_LLTX) == 0) ? \
4180 __netif_tx_trylock(txq) : \
5a717f4f 4181 __netif_tx_acquire(txq))
5efeac44 4182
c773e847 4183#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 4184 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 4185 __netif_tx_unlock(txq); \
5a717f4f
MT
4186 } else { \
4187 __netif_tx_release(txq); \
22dd7495
JHS
4188 } \
4189}
4190
1da177e4
LT
4191static inline void netif_tx_disable(struct net_device *dev)
4192{
fd2ea0a7 4193 unsigned int i;
c3f26a26 4194 int cpu;
fd2ea0a7 4195
c3f26a26
DM
4196 local_bh_disable();
4197 cpu = smp_processor_id();
fd2ea0a7
DM
4198 for (i = 0; i < dev->num_tx_queues; i++) {
4199 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
4200
4201 __netif_tx_lock(txq, cpu);
fd2ea0a7 4202 netif_tx_stop_queue(txq);
c3f26a26 4203 __netif_tx_unlock(txq);
fd2ea0a7 4204 }
c3f26a26 4205 local_bh_enable();
1da177e4
LT
4206}
4207
e308a5d8
DM
4208static inline void netif_addr_lock(struct net_device *dev)
4209{
4210 spin_lock(&dev->addr_list_lock);
4211}
4212
4213static inline void netif_addr_lock_bh(struct net_device *dev)
4214{
4215 spin_lock_bh(&dev->addr_list_lock);
4216}
4217
4218static inline void netif_addr_unlock(struct net_device *dev)
4219{
4220 spin_unlock(&dev->addr_list_lock);
4221}
4222
4223static inline void netif_addr_unlock_bh(struct net_device *dev)
4224{
4225 spin_unlock_bh(&dev->addr_list_lock);
4226}
4227
f001fde5 4228/*
31278e71 4229 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
4230 * rcu_read_lock held.
4231 */
4232#define for_each_dev_addr(dev, ha) \
31278e71 4233 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 4234
1da177e4
LT
4235/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4236
f629d208 4237void ether_setup(struct net_device *dev);
1da177e4
LT
4238
4239/* Support for loadable net-drivers */
f629d208 4240struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 4241 unsigned char name_assign_type,
f629d208
JP
4242 void (*setup)(struct net_device *),
4243 unsigned int txqs, unsigned int rxqs);
c835a677
TG
4244#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4245 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 4246
c835a677
TG
4247#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4248 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4249 count)
36909ea4 4250
f629d208
JP
4251int register_netdev(struct net_device *dev);
4252void unregister_netdev(struct net_device *dev);
f001fde5 4253
22bedad3 4254/* General hardware address lists handling functions */
f629d208
JP
4255int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4256 struct netdev_hw_addr_list *from_list, int addr_len);
4257void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4258 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
4259int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4260 struct net_device *dev,
4261 int (*sync)(struct net_device *, const unsigned char *),
4262 int (*unsync)(struct net_device *,
4263 const unsigned char *));
e7946760
IK
4264int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4265 struct net_device *dev,
4266 int (*sync)(struct net_device *,
4267 const unsigned char *, int),
4268 int (*unsync)(struct net_device *,
4269 const unsigned char *, int));
4270void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4271 struct net_device *dev,
4272 int (*unsync)(struct net_device *,
4273 const unsigned char *, int));
670e5b8e
AD
4274void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4275 struct net_device *dev,
4276 int (*unsync)(struct net_device *,
4277 const unsigned char *));
f629d208 4278void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 4279
f001fde5 4280/* Functions used for device addresses handling */
f629d208
JP
4281int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4282 unsigned char addr_type);
4283int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4284 unsigned char addr_type);
f629d208
JP
4285void dev_addr_flush(struct net_device *dev);
4286int dev_addr_init(struct net_device *dev);
a748ee24
JP
4287
4288/* Functions used for unicast addresses handling */
f629d208
JP
4289int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4290int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4291int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4292int dev_uc_sync(struct net_device *to, struct net_device *from);
4293int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4294void dev_uc_unsync(struct net_device *to, struct net_device *from);
4295void dev_uc_flush(struct net_device *dev);
4296void dev_uc_init(struct net_device *dev);
f001fde5 4297
670e5b8e
AD
4298/**
4299 * __dev_uc_sync - Synchonize device's unicast list
4300 * @dev: device to sync
4301 * @sync: function to call if address should be added
4302 * @unsync: function to call if address should be removed
4303 *
4304 * Add newly added addresses to the interface, and release
4305 * addresses that have been deleted.
5e82b4b2 4306 */
670e5b8e
AD
4307static inline int __dev_uc_sync(struct net_device *dev,
4308 int (*sync)(struct net_device *,
4309 const unsigned char *),
4310 int (*unsync)(struct net_device *,
4311 const unsigned char *))
4312{
4313 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4314}
4315
4316/**
e793c0f7 4317 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
4318 * @dev: device to sync
4319 * @unsync: function to call if address should be removed
4320 *
4321 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 4322 */
670e5b8e
AD
4323static inline void __dev_uc_unsync(struct net_device *dev,
4324 int (*unsync)(struct net_device *,
4325 const unsigned char *))
4326{
4327 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4328}
4329
22bedad3 4330/* Functions used for multicast addresses handling */
f629d208
JP
4331int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4332int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4333int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4334int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4335int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4336int dev_mc_sync(struct net_device *to, struct net_device *from);
4337int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4338void dev_mc_unsync(struct net_device *to, struct net_device *from);
4339void dev_mc_flush(struct net_device *dev);
4340void dev_mc_init(struct net_device *dev);
f001fde5 4341
670e5b8e
AD
4342/**
4343 * __dev_mc_sync - Synchonize device's multicast list
4344 * @dev: device to sync
4345 * @sync: function to call if address should be added
4346 * @unsync: function to call if address should be removed
4347 *
4348 * Add newly added addresses to the interface, and release
4349 * addresses that have been deleted.
5e82b4b2 4350 */
670e5b8e
AD
4351static inline int __dev_mc_sync(struct net_device *dev,
4352 int (*sync)(struct net_device *,
4353 const unsigned char *),
4354 int (*unsync)(struct net_device *,
4355 const unsigned char *))
4356{
4357 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4358}
4359
4360/**
e793c0f7 4361 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
4362 * @dev: device to sync
4363 * @unsync: function to call if address should be removed
4364 *
4365 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 4366 */
670e5b8e
AD
4367static inline void __dev_mc_unsync(struct net_device *dev,
4368 int (*unsync)(struct net_device *,
4369 const unsigned char *))
4370{
4371 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4372}
4373
4417da66 4374/* Functions used for secondary unicast and multicast support */
f629d208
JP
4375void dev_set_rx_mode(struct net_device *dev);
4376void __dev_set_rx_mode(struct net_device *dev);
4377int dev_set_promiscuity(struct net_device *dev, int inc);
4378int dev_set_allmulti(struct net_device *dev, int inc);
4379void netdev_state_change(struct net_device *dev);
4380void netdev_notify_peers(struct net_device *dev);
4381void netdev_features_change(struct net_device *dev);
1da177e4 4382/* Load a device via the kmod */
f629d208
JP
4383void dev_load(struct net *net, const char *name);
4384struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4385 struct rtnl_link_stats64 *storage);
4386void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4387 const struct net_device_stats *netdev_stats);
eeda3fd6 4388
1da177e4 4389extern int netdev_max_backlog;
3b098e2d 4390extern int netdev_tstamp_prequeue;
1da177e4 4391extern int weight_p;
3d48b53f
MT
4392extern int dev_weight_rx_bias;
4393extern int dev_weight_tx_bias;
4394extern int dev_rx_weight;
4395extern int dev_tx_weight;
323ebb61 4396extern int gro_normal_batch;
9ff162a8 4397
f629d208 4398bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
4399struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4400 struct list_head **iter);
f629d208
JP
4401struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4402 struct list_head **iter);
8b5be856 4403
44a40855
VY
4404/* iterate through upper list, must be called under RCU read lock */
4405#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4406 for (iter = &(dev)->adj_list.upper, \
4407 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4408 updev; \
4409 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4410
1a3f060c
DA
4411int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4412 int (*fn)(struct net_device *upper_dev,
4413 void *data),
4414 void *data);
4415
4416bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4417 struct net_device *upper_dev);
4418
25cc72a3
IS
4419bool netdev_has_any_upper_dev(struct net_device *dev);
4420
f629d208
JP
4421void *netdev_lower_get_next_private(struct net_device *dev,
4422 struct list_head **iter);
4423void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4424 struct list_head **iter);
31088a11
VF
4425
4426#define netdev_for_each_lower_private(dev, priv, iter) \
4427 for (iter = (dev)->adj_list.lower.next, \
4428 priv = netdev_lower_get_next_private(dev, &(iter)); \
4429 priv; \
4430 priv = netdev_lower_get_next_private(dev, &(iter)))
4431
4432#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4433 for (iter = &(dev)->adj_list.lower, \
4434 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4435 priv; \
4436 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4437
4085ebe8
VY
4438void *netdev_lower_get_next(struct net_device *dev,
4439 struct list_head **iter);
7ce856aa 4440
4085ebe8 4441#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 4442 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
4443 ldev = netdev_lower_get_next(dev, &(iter)); \
4444 ldev; \
4445 ldev = netdev_lower_get_next(dev, &(iter)))
4446
7151affe 4447struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
7ce856aa 4448 struct list_head **iter);
1a3f060c
DA
4449int netdev_walk_all_lower_dev(struct net_device *dev,
4450 int (*fn)(struct net_device *lower_dev,
4451 void *data),
4452 void *data);
4453int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4454 int (*fn)(struct net_device *lower_dev,
4455 void *data),
4456 void *data);
4457
f629d208 4458void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 4459void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
4460struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4461struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
42ab19ee
DA
4462int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4463 struct netlink_ext_ack *extack);
f629d208 4464int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 4465 struct net_device *upper_dev,
42ab19ee
DA
4466 void *upper_priv, void *upper_info,
4467 struct netlink_ext_ack *extack);
f629d208
JP
4468void netdev_upper_dev_unlink(struct net_device *dev,
4469 struct net_device *upper_dev);
32b6d34f
TY
4470int netdev_adjacent_change_prepare(struct net_device *old_dev,
4471 struct net_device *new_dev,
4472 struct net_device *dev,
4473 struct netlink_ext_ack *extack);
4474void netdev_adjacent_change_commit(struct net_device *old_dev,
4475 struct net_device *new_dev,
4476 struct net_device *dev);
4477void netdev_adjacent_change_abort(struct net_device *old_dev,
4478 struct net_device *new_dev,
4479 struct net_device *dev);
5bb025fa 4480void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
4481void *netdev_lower_dev_get_private(struct net_device *dev,
4482 struct net_device *lower_dev);
04d48266
JP
4483void netdev_lower_state_changed(struct net_device *lower_dev,
4484 void *lower_state_info);
960fb622
ED
4485
4486/* RSS keys are 40 or 52 bytes long */
4487#define NETDEV_RSS_KEY_LEN 52
ba905f5e 4488extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
4489void netdev_rss_key_fill(void *buffer, size_t len);
4490
f629d208 4491int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 4492int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
4493int skb_csum_hwoffload_help(struct sk_buff *skb,
4494 const netdev_features_t features);
4495
f629d208
JP
4496struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4497 netdev_features_t features, bool tx_path);
4498struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4499 netdev_features_t features);
12b0004d 4500
61bd3857
MS
4501struct netdev_bonding_info {
4502 ifslave slave;
4503 ifbond master;
4504};
4505
4506struct netdev_notifier_bonding_info {
4507 struct netdev_notifier_info info; /* must be first */
4508 struct netdev_bonding_info bonding_info;
4509};
4510
4511void netdev_bonding_info_change(struct net_device *dev,
4512 struct netdev_bonding_info *bonding_info);
4513
6b08d6c1
MK
4514#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4515void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4516#else
4517static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4518 const void *data)
4519{
4520}
4521#endif
4522
12b0004d
CW
4523static inline
4524struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4525{
4526 return __skb_gso_segment(skb, features, true);
4527}
53d6471c 4528__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
4529
4530static inline bool can_checksum_protocol(netdev_features_t features,
4531 __be16 protocol)
4532{
c8cd0989
TH
4533 if (protocol == htons(ETH_P_FCOE))
4534 return !!(features & NETIF_F_FCOE_CRC);
4535
4536 /* Assume this is an IP checksum (not SCTP CRC) */
4537
4538 if (features & NETIF_F_HW_CSUM) {
4539 /* Can checksum everything */
4540 return true;
4541 }
4542
4543 switch (protocol) {
4544 case htons(ETH_P_IP):
4545 return !!(features & NETIF_F_IP_CSUM);
4546 case htons(ETH_P_IPV6):
4547 return !!(features & NETIF_F_IPV6_CSUM);
4548 default:
4549 return false;
4550 }
ec5f0615 4551}
12b0004d 4552
fb286bb2 4553#ifdef CONFIG_BUG
7fe50ac8 4554void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
fb286bb2 4555#else
7fe50ac8
CW
4556static inline void netdev_rx_csum_fault(struct net_device *dev,
4557 struct sk_buff *skb)
fb286bb2
HX
4558{
4559}
4560#endif
1da177e4 4561/* rx skb timestamps */
f629d208
JP
4562void net_enable_timestamp(void);
4563void net_disable_timestamp(void);
1da177e4 4564
20380731 4565#ifdef CONFIG_PROC_FS
f629d208 4566int __init dev_proc_init(void);
900ff8c6
CW
4567#else
4568#define dev_proc_init() 0
20380731
ACM
4569#endif
4570
4798248e 4571static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4572 struct sk_buff *skb, struct net_device *dev,
4573 bool more)
4798248e 4574{
6b16f9ee 4575 __this_cpu_write(softnet_data.xmit.more, more);
0b725a2c 4576 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4577}
4578
97cdcf37
FW
4579static inline bool netdev_xmit_more(void)
4580{
4581 return __this_cpu_read(softnet_data.xmit.more);
4582}
4583
10b3ad8c 4584static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4585 struct netdev_queue *txq, bool more)
4798248e
DM
4586{
4587 const struct net_device_ops *ops = dev->netdev_ops;
2183435c 4588 netdev_tx_t rc;
4798248e 4589
fa2dbdc2 4590 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4591 if (rc == NETDEV_TX_OK)
4592 txq_trans_update(txq);
4593
4594 return rc;
4798248e
DM
4595}
4596
b793dc5c 4597int netdev_class_create_file_ns(const struct class_attribute *class_attr,
42a2d923 4598 const void *ns);
b793dc5c 4599void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
42a2d923 4600 const void *ns);
58292cbe 4601
b793dc5c 4602static inline int netdev_class_create_file(const struct class_attribute *class_attr)
58292cbe
TH
4603{
4604 return netdev_class_create_file_ns(class_attr, NULL);
4605}
4606
b793dc5c 4607static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
58292cbe
TH
4608{
4609 netdev_class_remove_file_ns(class_attr, NULL);
4610}
b8a9787e 4611
737aec57 4612extern const struct kobj_ns_type_operations net_ns_type_operations;
04600794 4613
f629d208 4614const char *netdev_drivername(const struct net_device *dev);
6579e57b 4615
f629d208 4616void linkwatch_run_queue(void);
20380731 4617
da08143b
MK
4618static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4619 netdev_features_t f2)
4620{
c8cd0989
TH
4621 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4622 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4623 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4624 else
b6a0e72a 4625 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4626 }
da08143b 4627
c8cd0989 4628 return f1 & f2;
da08143b
MK
4629}
4630
c8f44aff
MM
4631static inline netdev_features_t netdev_get_wanted_features(
4632 struct net_device *dev)
5455c699
MM
4633{
4634 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4635}
c8f44aff
MM
4636netdev_features_t netdev_increment_features(netdev_features_t all,
4637 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4638
4639/* Allow TSO being used on stacked device :
4640 * Performing the GSO segmentation before last device
4641 * is a performance improvement.
4642 */
4643static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4644 netdev_features_t mask)
4645{
4646 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4647}
4648
6cb6a27c 4649int __netdev_update_features(struct net_device *dev);
5455c699 4650void netdev_update_features(struct net_device *dev);
afe12cc8 4651void netdev_change_features(struct net_device *dev);
7f353bf2 4652
fc4a7489
PM
4653void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4654 struct net_device *dev);
4655
e38f3025
TM
4656netdev_features_t passthru_features_check(struct sk_buff *skb,
4657 struct net_device *dev,
4658 netdev_features_t features);
c1e756bf 4659netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4660
4d29515f 4661static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4662{
7b748340 4663 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4664
4665 /* check flags correspondence */
4666 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4667 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4668 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4669 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4670 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4671 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4672 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4673 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4674 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4675 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4676 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4677 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4678 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4679 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4680 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4681 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0c19f846 4682 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
83aa025f 4683 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
3b335832 4684 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
0345e186 4685
d6b4991a 4686 return (features & feature) == feature;
576a30eb
HX
4687}
4688
4d29515f 4689static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4690{
278b2513 4691 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4692 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4693}
4694
8b86a61d 4695static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4696 netdev_features_t features)
7967168c 4697{
fc741216 4698 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4699 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4700 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4701}
4702
82cc1a7a
PWJ
4703static inline void netif_set_gso_max_size(struct net_device *dev,
4704 unsigned int size)
4705{
4706 dev->gso_max_size = size;
4707}
4708
7a7ffbab
WCC
4709static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4710 int pulled_hlen, u16 mac_offset,
4711 int mac_len)
4712{
4713 skb->protocol = protocol;
4714 skb->encapsulation = 1;
4715 skb_push(skb, pulled_hlen);
4716 skb_reset_transport_header(skb);
4717 skb->mac_header = mac_offset;
4718 skb->network_header = skb->mac_header + mac_len;
4719 skb->mac_len = mac_len;
4720}
4721
3c175784
SD
4722static inline bool netif_is_macsec(const struct net_device *dev)
4723{
4724 return dev->priv_flags & IFF_MACSEC;
4725}
4726
b618aaa9 4727static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4728{
4729 return dev->priv_flags & IFF_MACVLAN;
4730}
4731
b618aaa9 4732static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4733{
4734 return dev->priv_flags & IFF_MACVLAN_PORT;
4735}
4736
b618aaa9 4737static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4738{
4739 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4740}
4741
b618aaa9 4742static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4743{
4744 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4745}
4746
3bdc0eba
BG
4747static inline bool netif_supports_nofcs(struct net_device *dev)
4748{
4749 return dev->priv_flags & IFF_SUPP_NOFCS;
4750}
4751
d5256083
DB
4752static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4753{
4754 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4755}
4756
007979ea 4757static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4758{
007979ea 4759 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4760}
4761
fee6d4c7
DA
4762static inline bool netif_is_l3_slave(const struct net_device *dev)
4763{
4764 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4765}
4766
0894ae3f
JP
4767static inline bool netif_is_bridge_master(const struct net_device *dev)
4768{
4769 return dev->priv_flags & IFF_EBRIDGE;
4770}
4771
28f9ee22
VY
4772static inline bool netif_is_bridge_port(const struct net_device *dev)
4773{
4774 return dev->priv_flags & IFF_BRIDGE_PORT;
4775}
4776
35d4e172
JP
4777static inline bool netif_is_ovs_master(const struct net_device *dev)
4778{
4779 return dev->priv_flags & IFF_OPENVSWITCH;
4780}
4781
5be66141
JP
4782static inline bool netif_is_ovs_port(const struct net_device *dev)
4783{
4784 return dev->priv_flags & IFF_OVS_DATAPATH;
4785}
4786
b618aaa9 4787static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4788{
4789 return dev->priv_flags & IFF_TEAM;
4790}
4791
b618aaa9 4792static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4793{
4794 return dev->priv_flags & IFF_TEAM_PORT;
4795}
4796
b618aaa9 4797static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4798{
4799 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4800}
4801
b618aaa9 4802static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4803{
4804 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4805}
4806
d4ab4286
KJ
4807static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4808{
4809 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4810}
4811
30c8bd5a
SS
4812static inline bool netif_is_failover(const struct net_device *dev)
4813{
4814 return dev->priv_flags & IFF_FAILOVER;
4815}
4816
4817static inline bool netif_is_failover_slave(const struct net_device *dev)
4818{
4819 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4820}
4821
02875878
ED
4822/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4823static inline void netif_keep_dst(struct net_device *dev)
4824{
4825 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4826}
4827
18d3df3e
PA
4828/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4829static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4830{
4831 /* TODO: reserve and use an additional IFF bit, if we get more users */
4832 return dev->priv_flags & IFF_MACSEC;
4833}
4834
505d4f73 4835extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4836
571ba423
JP
4837/* Logging, debugging and troubleshooting/diagnostic helpers. */
4838
4839/* netdev_printk helpers, similar to dev_printk */
4840
4841static inline const char *netdev_name(const struct net_device *dev)
4842{
c6f854d5
VF
4843 if (!dev->name[0] || strchr(dev->name, '%'))
4844 return "(unnamed net_device)";
571ba423
JP
4845 return dev->name;
4846}
4847
8397ed36
DA
4848static inline bool netdev_unregistering(const struct net_device *dev)
4849{
4850 return dev->reg_state == NETREG_UNREGISTERING;
4851}
4852
ccc7f496
VF
4853static inline const char *netdev_reg_state(const struct net_device *dev)
4854{
4855 switch (dev->reg_state) {
4856 case NETREG_UNINITIALIZED: return " (uninitialized)";
4857 case NETREG_REGISTERED: return "";
4858 case NETREG_UNREGISTERING: return " (unregistering)";
4859 case NETREG_UNREGISTERED: return " (unregistered)";
4860 case NETREG_RELEASED: return " (released)";
4861 case NETREG_DUMMY: return " (dummy)";
4862 }
4863
4864 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4865 return " (unknown)";
4866}
4867
ce3fdb69 4868__printf(3, 4) __cold
6ea754eb
JP
4869void netdev_printk(const char *level, const struct net_device *dev,
4870 const char *format, ...);
ce3fdb69 4871__printf(2, 3) __cold
6ea754eb 4872void netdev_emerg(const struct net_device *dev, const char *format, ...);
ce3fdb69 4873__printf(2, 3) __cold
6ea754eb 4874void netdev_alert(const struct net_device *dev, const char *format, ...);
ce3fdb69 4875__printf(2, 3) __cold
6ea754eb 4876void netdev_crit(const struct net_device *dev, const char *format, ...);
ce3fdb69 4877__printf(2, 3) __cold
6ea754eb 4878void netdev_err(const struct net_device *dev, const char *format, ...);
ce3fdb69 4879__printf(2, 3) __cold
6ea754eb 4880void netdev_warn(const struct net_device *dev, const char *format, ...);
ce3fdb69 4881__printf(2, 3) __cold
6ea754eb 4882void netdev_notice(const struct net_device *dev, const char *format, ...);
ce3fdb69 4883__printf(2, 3) __cold
6ea754eb 4884void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4885
375ef2b1
GP
4886#define netdev_level_once(level, dev, fmt, ...) \
4887do { \
4888 static bool __print_once __read_mostly; \
4889 \
4890 if (!__print_once) { \
4891 __print_once = true; \
4892 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4893 } \
4894} while (0)
4895
4896#define netdev_emerg_once(dev, fmt, ...) \
4897 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4898#define netdev_alert_once(dev, fmt, ...) \
4899 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4900#define netdev_crit_once(dev, fmt, ...) \
4901 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4902#define netdev_err_once(dev, fmt, ...) \
4903 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4904#define netdev_warn_once(dev, fmt, ...) \
4905 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4906#define netdev_notice_once(dev, fmt, ...) \
4907 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4908#define netdev_info_once(dev, fmt, ...) \
4909 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4910
8909c9ad
VK
4911#define MODULE_ALIAS_NETDEV(device) \
4912 MODULE_ALIAS("netdev-" device)
4913
b558c96f 4914#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4915#define netdev_dbg(__dev, format, args...) \
4916do { \
ffa10cb4 4917 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4918} while (0)
b558c96f
JC
4919#elif defined(DEBUG)
4920#define netdev_dbg(__dev, format, args...) \
4921 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4922#else
4923#define netdev_dbg(__dev, format, args...) \
4924({ \
4925 if (0) \
4926 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4927})
4928#endif
4929
4930#if defined(VERBOSE_DEBUG)
4931#define netdev_vdbg netdev_dbg
4932#else
4933
4934#define netdev_vdbg(dev, format, args...) \
4935({ \
4936 if (0) \
4937 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4938 0; \
4939})
4940#endif
4941
4942/*
4943 * netdev_WARN() acts like dev_printk(), but with the key difference
4944 * of using a WARN/WARN_ON to get the message out, including the
4945 * file/line information and a backtrace.
4946 */
4947#define netdev_WARN(dev, format, args...) \
e1cfe3d0 4948 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
ccc7f496 4949 netdev_reg_state(dev), ##args)
571ba423 4950
72dd831e 4951#define netdev_WARN_ONCE(dev, format, args...) \
e1cfe3d0 4952 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
375ef2b1
GP
4953 netdev_reg_state(dev), ##args)
4954
b3d95c5c
JP
4955/* netif printk helpers, similar to netdev_printk */
4956
4957#define netif_printk(priv, type, level, dev, fmt, args...) \
4958do { \
4959 if (netif_msg_##type(priv)) \
4960 netdev_printk(level, (dev), fmt, ##args); \
4961} while (0)
4962
f45f4321
JP
4963#define netif_level(level, priv, type, dev, fmt, args...) \
4964do { \
4965 if (netif_msg_##type(priv)) \
4966 netdev_##level(dev, fmt, ##args); \
4967} while (0)
4968
b3d95c5c 4969#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4970 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4971#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4972 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4973#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4974 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4975#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4976 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4977#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4978 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4979#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4980 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4981#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4982 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4983
0053ea9c 4984#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4985#define netif_dbg(priv, type, netdev, format, args...) \
4986do { \
4987 if (netif_msg_##type(priv)) \
b5fb0a03 4988 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4989} while (0)
0053ea9c
JP
4990#elif defined(DEBUG)
4991#define netif_dbg(priv, type, dev, format, args...) \
4992 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4993#else
4994#define netif_dbg(priv, type, dev, format, args...) \
4995({ \
4996 if (0) \
4997 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4998 0; \
4999})
5000#endif
5001
f617f276
EC
5002/* if @cond then downgrade to debug, else print at @level */
5003#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5004 do { \
5005 if (cond) \
5006 netif_dbg(priv, type, netdev, fmt, ##args); \
5007 else \
5008 netif_ ## level(priv, type, netdev, fmt, ##args); \
5009 } while (0)
5010
b3d95c5c 5011#if defined(VERBOSE_DEBUG)
bcfcc450 5012#define netif_vdbg netif_dbg
b3d95c5c
JP
5013#else
5014#define netif_vdbg(priv, type, dev, format, args...) \
5015({ \
5016 if (0) \
a4ed89cb 5017 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
5018 0; \
5019})
5020#endif
571ba423 5021
900ff8c6
CW
5022/*
5023 * The list of packet types we will receive (as opposed to discard)
5024 * and the routines to invoke.
5025 *
5026 * Why 16. Because with 16 the only overlap we get on a hash of the
5027 * low nibble of the protocol value is RARP/SNAP/X.25.
5028 *
900ff8c6 5029 * 0800 IP
900ff8c6
CW
5030 * 0001 802.3
5031 * 0002 AX.25
5032 * 0004 802.2
5033 * 8035 RARP
5034 * 0005 SNAP
5035 * 0805 X.25
5036 * 0806 ARP
5037 * 8137 IPX
5038 * 0009 Localtalk
5039 * 86DD IPv6
5040 */
5041#define PTYPE_HASH_SIZE (16)
5042#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5043
4de83b88
MB
5044extern struct net_device *blackhole_netdev;
5045
385a154c 5046#endif /* _LINUX_NETDEVICE_H */