]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - include/linux/netdevice.h
net: push code from net notifier reg/unreg into helpers
[mirror_ubuntu-hirsute-kernel.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;
c6e970a0 56
e679c9c1 57struct sfp_bus;
704232c2
JB
58/* 802.11 specific */
59struct wireless_dev;
98a18b6f
AA
60/* 802.15.4 specific */
61struct wpan_dev;
03c57747 62struct mpls_dev;
7c46a640
AD
63/* UDP Tunnel offloads */
64struct udp_tunnel_info;
a7862b45 65struct bpf_prog;
814abfab 66struct xdp_buff;
1da177e4 67
f629d208
JP
68void netdev_set_default_ethtool_ops(struct net_device *dev,
69 const struct ethtool_ops *ops);
d07d7507 70
9a1654ba
JP
71/* Backlog congestion levels */
72#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
73#define NET_RX_DROP 1 /* packet dropped */
74
572a9d7b
PM
75/*
76 * Transmit return codes: transmit return codes originate from three different
77 * namespaces:
78 *
79 * - qdisc return codes
80 * - driver transmit return codes
81 * - errno values
82 *
83 * Drivers are allowed to return any one of those in their hard_start_xmit()
84 * function. Real network devices commonly used with qdiscs should only return
85 * the driver transmit return codes though - when qdiscs are used, the actual
86 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
87 * higher layers. Virtual network devices transmit synchronously; in this case
88 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
89 * others are propagated to higher layers.
90 */
91
92/* qdisc ->enqueue() return codes. */
93#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
94#define NET_XMIT_DROP 0x01 /* skb dropped */
95#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 96#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 97
b9df3cb8
GR
98/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
99 * indicates that the device will soon be dropping packets, or already drops
100 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 101#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
102#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
103
dc1f8bf6 104/* Driver transmit return codes */
9a1654ba 105#define NETDEV_TX_MASK 0xf0
572a9d7b 106
dc1f8bf6 107enum netdev_tx {
572a9d7b 108 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
109 NETDEV_TX_OK = 0x00, /* driver took care of packet */
110 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
111};
112typedef enum netdev_tx netdev_tx_t;
113
9a1654ba
JP
114/*
115 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
116 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
117 */
118static inline bool dev_xmit_complete(int rc)
119{
120 /*
121 * Positive cases with an skb consumed by a driver:
122 * - successful transmission (rc == NETDEV_TX_OK)
123 * - error while transmitting (rc < 0)
124 * - error while queueing to a different device (rc & NET_XMIT_MASK)
125 */
126 if (likely(rc < NET_XMIT_MASK))
127 return true;
128
129 return false;
130}
131
1da177e4 132/*
5e82b4b2 133 * Compute the worst-case header length according to the protocols
1da177e4
LT
134 * used.
135 */
fe2918b0 136
c0eb4540
KS
137#if defined(CONFIG_HYPERV_NET)
138# define LL_MAX_HEADER 128
139#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
140# if defined(CONFIG_MAC80211_MESH)
141# define LL_MAX_HEADER 128
142# else
143# define LL_MAX_HEADER 96
144# endif
1da177e4 145#else
8388e3da 146# define LL_MAX_HEADER 32
1da177e4
LT
147#endif
148
d11ead75
BH
149#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
150 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
151#define MAX_HEADER LL_MAX_HEADER
152#else
153#define MAX_HEADER (LL_MAX_HEADER + 48)
154#endif
155
156/*
be1f3c2c
BH
157 * Old network device statistics. Fields are native words
158 * (unsigned long) so they can be read and written atomically.
1da177e4 159 */
fe2918b0 160
d94d9fee 161struct net_device_stats {
3cfde79c
BH
162 unsigned long rx_packets;
163 unsigned long tx_packets;
164 unsigned long rx_bytes;
165 unsigned long tx_bytes;
166 unsigned long rx_errors;
167 unsigned long tx_errors;
168 unsigned long rx_dropped;
169 unsigned long tx_dropped;
170 unsigned long multicast;
1da177e4 171 unsigned long collisions;
1da177e4 172 unsigned long rx_length_errors;
3cfde79c
BH
173 unsigned long rx_over_errors;
174 unsigned long rx_crc_errors;
175 unsigned long rx_frame_errors;
176 unsigned long rx_fifo_errors;
177 unsigned long rx_missed_errors;
1da177e4
LT
178 unsigned long tx_aborted_errors;
179 unsigned long tx_carrier_errors;
180 unsigned long tx_fifo_errors;
181 unsigned long tx_heartbeat_errors;
182 unsigned long tx_window_errors;
1da177e4
LT
183 unsigned long rx_compressed;
184 unsigned long tx_compressed;
185};
186
1da177e4
LT
187
188#include <linux/cache.h>
189#include <linux/skbuff.h>
190
adc9300e 191#ifdef CONFIG_RPS
c5905afb 192#include <linux/static_key.h>
dc05360f
ED
193extern struct static_key_false rps_needed;
194extern struct static_key_false rfs_needed;
adc9300e
ED
195#endif
196
1da177e4
LT
197struct neighbour;
198struct neigh_parms;
199struct sk_buff;
200
f001fde5
JP
201struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
ccffad25
JP
205#define NETDEV_HW_ADDR_T_LAN 1
206#define NETDEV_HW_ADDR_T_SAN 2
207#define NETDEV_HW_ADDR_T_SLAVE 3
208#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 209#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 210 bool global_use;
4cd729b0 211 int sync_cnt;
8f8f103d 212 int refcount;
4543fbef 213 int synced;
f001fde5
JP
214 struct rcu_head rcu_head;
215};
216
31278e71
JP
217struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220};
221
22bedad3
JP
222#define netdev_hw_addr_list_count(l) ((l)->count)
223#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224#define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 226
22bedad3
JP
227#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229#define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 231
22bedad3
JP
232#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 234#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 236
d94d9fee 237struct hh_cache {
5b3dc2f3 238 unsigned int hh_len;
3644f0ce 239 seqlock_t hh_lock;
1da177e4
LT
240
241 /* cached hardware header; allow for machine alignment needs. */
242#define HH_DATA_MOD 16
243#define HH_DATA_OFF(__len) \
5ba0eac6 244 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
245#define HH_DATA_ALIGN(__len) \
246 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
247 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
248};
249
5e82b4b2 250/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
251 * Alternative is:
252 * dev->hard_header_len ? (dev->hard_header_len +
253 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
254 *
255 * We could use other alignment values, but we must maintain the
256 * relationship HH alignment <= LL alignment.
257 */
258#define LL_RESERVED_SPACE(dev) \
f5184d26 259 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 260#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 261 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 262
3b04ddde
SH
263struct header_ops {
264 int (*create) (struct sk_buff *skb, struct net_device *dev,
265 unsigned short type, const void *daddr,
95c96174 266 const void *saddr, unsigned int len);
3b04ddde 267 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 268 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
269 void (*cache_update)(struct hh_cache *hh,
270 const struct net_device *dev,
271 const unsigned char *haddr);
2793a23a 272 bool (*validate)(const char *ll_header, unsigned int len);
e78b2915 273 __be16 (*parse_protocol)(const struct sk_buff *skb);
3b04ddde
SH
274};
275
1da177e4 276/* These flag bits are private to the generic network queueing
5e82b4b2 277 * layer; they may not be explicitly referenced by any other
1da177e4
LT
278 * code.
279 */
280
d94d9fee 281enum netdev_state_t {
1da177e4
LT
282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
1da177e4 284 __LINK_STATE_NOCARRIER,
b00055aa
SR
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
1da177e4
LT
287};
288
289
290/*
5e82b4b2 291 * This structure holds boot-time configured netdevice settings. They
fe2918b0 292 * are then used in the device probing.
1da177e4
LT
293 */
294struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297};
298#define NETDEV_BOOT_SETUP_MAX 8
299
f629d208 300int __init netdev_boot_setup(char *str);
1da177e4 301
6312fe77
LR
302struct gro_list {
303 struct list_head list;
304 int count;
305};
306
bea3348e 307/*
d9f37d01
LR
308 * size of gro hash buckets, must less than bit number of
309 * napi_struct::gro_bitmask
bea3348e 310 */
07d78363 311#define GRO_HASH_BUCKETS 8
d9f37d01
LR
312
313/*
314 * Structure for NAPI scheduling similar to tasklet but with weighting
315 */
bea3348e
SH
316struct napi_struct {
317 /* The poll_list must only be managed by the entity which
318 * changes the state of the NAPI_STATE_SCHED bit. This means
319 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 320 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
321 * can remove from the list right before clearing the bit.
322 */
323 struct list_head poll_list;
324
325 unsigned long state;
326 int weight;
d9f37d01 327 unsigned long gro_bitmask;
bea3348e
SH
328 int (*poll)(struct napi_struct *, int);
329#ifdef CONFIG_NETPOLL
bea3348e 330 int poll_owner;
bea3348e 331#endif
5d38a079 332 struct net_device *dev;
6312fe77 333 struct gro_list gro_hash[GRO_HASH_BUCKETS];
5d38a079 334 struct sk_buff *skb;
323ebb61
EC
335 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
336 int rx_count; /* length of rx_list */
3b47d303 337 struct hrtimer timer;
404f7c9e 338 struct list_head dev_list;
af12fa6e
ET
339 struct hlist_node napi_hash_node;
340 unsigned int napi_id;
bea3348e
SH
341};
342
d94d9fee 343enum {
bea3348e 344 NAPI_STATE_SCHED, /* Poll is scheduled */
39e6c820 345 NAPI_STATE_MISSED, /* reschedule a napi */
a0a46196 346 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 347 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
348 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
349 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
350 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
351};
352
353enum {
39e6c820
ED
354 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
355 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
356 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
357 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
358 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
359 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
360 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
361};
362
5b252f0c 363enum gro_result {
d1c76af9
HX
364 GRO_MERGED,
365 GRO_MERGED_FREE,
366 GRO_HELD,
367 GRO_NORMAL,
368 GRO_DROP,
25393d3f 369 GRO_CONSUMED,
d1c76af9 370};
5b252f0c 371typedef enum gro_result gro_result_t;
d1c76af9 372
8a4eb573
JP
373/*
374 * enum rx_handler_result - Possible return values for rx_handlers.
375 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
376 * further.
377 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
378 * case skb->dev was changed by rx_handler.
379 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 380 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
381 *
382 * rx_handlers are functions called from inside __netif_receive_skb(), to do
383 * special processing of the skb, prior to delivery to protocol handlers.
384 *
385 * Currently, a net_device can only have a single rx_handler registered. Trying
386 * to register a second rx_handler will return -EBUSY.
387 *
388 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
389 * To unregister a rx_handler on a net_device, use
390 * netdev_rx_handler_unregister().
391 *
392 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
393 * do with the skb.
394 *
5e82b4b2 395 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 396 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 397 * the skb to be delivered in some other way.
8a4eb573
JP
398 *
399 * If the rx_handler changed skb->dev, to divert the skb to another
400 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
401 * new device will be called if it exists.
402 *
5e82b4b2 403 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 404 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 405 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 406 *
5e82b4b2 407 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
408 * delivered, it should return RX_HANDLER_PASS.
409 *
410 * A device without a registered rx_handler will behave as if rx_handler
411 * returned RX_HANDLER_PASS.
412 */
413
414enum rx_handler_result {
415 RX_HANDLER_CONSUMED,
416 RX_HANDLER_ANOTHER,
417 RX_HANDLER_EXACT,
418 RX_HANDLER_PASS,
419};
420typedef enum rx_handler_result rx_handler_result_t;
421typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 422
f629d208 423void __napi_schedule(struct napi_struct *n);
bc9ad166 424void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 425
4d29515f 426static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
427{
428 return test_bit(NAPI_STATE_DISABLE, &n->state);
429}
430
39e6c820 431bool napi_schedule_prep(struct napi_struct *n);
bea3348e
SH
432
433/**
434 * napi_schedule - schedule NAPI poll
5e82b4b2 435 * @n: NAPI context
bea3348e
SH
436 *
437 * Schedule NAPI poll routine to be called if it is not already
438 * running.
439 */
440static inline void napi_schedule(struct napi_struct *n)
441{
442 if (napi_schedule_prep(n))
443 __napi_schedule(n);
444}
445
bc9ad166
ED
446/**
447 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 448 * @n: NAPI context
bc9ad166
ED
449 *
450 * Variant of napi_schedule(), assuming hard irqs are masked.
451 */
452static inline void napi_schedule_irqoff(struct napi_struct *n)
453{
454 if (napi_schedule_prep(n))
455 __napi_schedule_irqoff(n);
456}
457
bfe13f54 458/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 459static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
460{
461 if (napi_schedule_prep(napi)) {
462 __napi_schedule(napi);
4d29515f 463 return true;
bfe13f54 464 }
4d29515f 465 return false;
bfe13f54
RD
466}
467
364b6055 468bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
469/**
470 * napi_complete - NAPI processing complete
5e82b4b2 471 * @n: NAPI context
bea3348e
SH
472 *
473 * Mark NAPI processing as complete.
3b47d303 474 * Consider using napi_complete_done() instead.
364b6055 475 * Return false if device should avoid rearming interrupts.
bea3348e 476 */
364b6055 477static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
478{
479 return napi_complete_done(n, 0);
480}
bea3348e 481
af12fa6e
ET
482/**
483 * napi_hash_del - remove a NAPI from global table
5e82b4b2 484 * @napi: NAPI context
af12fa6e 485 *
5e82b4b2 486 * Warning: caller must observe RCU grace period
34cbe27e
ED
487 * before freeing memory containing @napi, if
488 * this function returns true.
93d05d4a 489 * Note: core networking stack automatically calls it
5e82b4b2 490 * from netif_napi_del().
93d05d4a 491 * Drivers might want to call this helper to combine all
5e82b4b2 492 * the needed RCU grace periods into a single one.
af12fa6e 493 */
34cbe27e 494bool napi_hash_del(struct napi_struct *napi);
af12fa6e 495
bea3348e
SH
496/**
497 * napi_disable - prevent NAPI from scheduling
5e82b4b2 498 * @n: NAPI context
bea3348e
SH
499 *
500 * Stop NAPI from being scheduled on this context.
501 * Waits till any outstanding processing completes.
502 */
3b47d303 503void napi_disable(struct napi_struct *n);
bea3348e
SH
504
505/**
506 * napi_enable - enable NAPI scheduling
5e82b4b2 507 * @n: NAPI context
bea3348e
SH
508 *
509 * Resume NAPI from being scheduled on this context.
510 * Must be paired with napi_disable.
511 */
512static inline void napi_enable(struct napi_struct *n)
513{
514 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 515 smp_mb__before_atomic();
bea3348e 516 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 517 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
518}
519
c264c3de
SH
520/**
521 * napi_synchronize - wait until NAPI is not running
5e82b4b2 522 * @n: NAPI context
c264c3de
SH
523 *
524 * Wait until NAPI is done being scheduled on this context.
525 * Waits till any outstanding processing completes but
526 * does not disable future activations.
527 */
528static inline void napi_synchronize(const struct napi_struct *n)
529{
facc432f
AB
530 if (IS_ENABLED(CONFIG_SMP))
531 while (test_bit(NAPI_STATE_SCHED, &n->state))
532 msleep(1);
533 else
534 barrier();
c264c3de 535}
c264c3de 536
6c5c9581
MK
537/**
538 * napi_if_scheduled_mark_missed - if napi is running, set the
539 * NAPIF_STATE_MISSED
540 * @n: NAPI context
541 *
542 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
543 * NAPI is scheduled.
544 **/
545static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
546{
547 unsigned long val, new;
548
549 do {
550 val = READ_ONCE(n->state);
551 if (val & NAPIF_STATE_DISABLE)
552 return true;
553
554 if (!(val & NAPIF_STATE_SCHED))
555 return false;
556
557 new = val | NAPIF_STATE_MISSED;
558 } while (cmpxchg(&n->state, val, new) != val);
559
560 return true;
561}
562
d94d9fee 563enum netdev_queue_state_t {
73466498
TH
564 __QUEUE_STATE_DRV_XOFF,
565 __QUEUE_STATE_STACK_XOFF,
c3f26a26 566 __QUEUE_STATE_FROZEN,
79d16385 567};
8e2f1a63
DB
568
569#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
570#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
571#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
572
573#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
574#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
575 QUEUE_STATE_FROZEN)
576#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
577 QUEUE_STATE_FROZEN)
578
73466498
TH
579/*
580 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
581 * netif_tx_* functions below are used to manipulate this flag. The
582 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
583 * queue independently. The netif_xmit_*stopped functions below are called
584 * to check if the queue has been stopped by the driver or stack (either
585 * of the XOFF bits are set in the state). Drivers should not need to call
586 * netif_xmit*stopped functions, they should only be using netif_tx_*.
587 */
79d16385 588
bb949fbd 589struct netdev_queue {
6a321cb3 590/*
5e82b4b2 591 * read-mostly part
6a321cb3 592 */
bb949fbd 593 struct net_device *dev;
46e5da40 594 struct Qdisc __rcu *qdisc;
b0e1e646 595 struct Qdisc *qdisc_sleeping;
ccf5ff69 596#ifdef CONFIG_SYSFS
1d24eb48
TH
597 struct kobject kobj;
598#endif
f2cd2d3e
ED
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 int numa_node;
601#endif
c0ef079c
FW
602 unsigned long tx_maxrate;
603 /*
604 * Number of TX timeouts for this queue
605 * (/sys/class/net/DEV/Q/trans_timeout)
606 */
607 unsigned long trans_timeout;
ffcfe25b
AD
608
609 /* Subordinate device that the queue has been assigned to */
610 struct net_device *sb_dev;
661b8d1b
MK
611#ifdef CONFIG_XDP_SOCKETS
612 struct xdp_umem *umem;
613#endif
6a321cb3 614/*
5e82b4b2 615 * write-mostly part
6a321cb3
ED
616 */
617 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
618 int xmit_lock_owner;
9d21493b 619 /*
9b36627a 620 * Time (in jiffies) of last Tx
9d21493b
ED
621 */
622 unsigned long trans_start;
ccf5ff69 623
114cf580
TH
624 unsigned long state;
625
626#ifdef CONFIG_BQL
627 struct dql dql;
628#endif
e8a0464c 629} ____cacheline_aligned_in_smp;
bb949fbd 630
79134e6c 631extern int sysctl_fb_tunnels_only_for_init_net;
856c395c 632extern int sysctl_devconf_inherit_init_net;
79134e6c
ED
633
634static inline bool net_has_fallback_tunnels(const struct net *net)
635{
be9fc097
AB
636 return net == &init_net ||
637 !IS_ENABLED(CONFIG_SYSCTL) ||
638 !sysctl_fb_tunnels_only_for_init_net;
79134e6c
ED
639}
640
f2cd2d3e
ED
641static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
642{
643#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
644 return q->numa_node;
645#else
b236da69 646 return NUMA_NO_NODE;
f2cd2d3e
ED
647#endif
648}
649
650static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
651{
652#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
653 q->numa_node = node;
654#endif
655}
656
df334545 657#ifdef CONFIG_RPS
0a9627f2
TH
658/*
659 * This structure holds an RPS map which can be of variable length. The
660 * map is an array of CPUs.
661 */
662struct rps_map {
663 unsigned int len;
664 struct rcu_head rcu;
665 u16 cpus[0];
666};
60b778ce 667#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 668
fec5e652 669/*
c445477d
BH
670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
671 * tail pointer for that CPU's input queue at the time of last enqueue, and
672 * a hardware filter index.
fec5e652
TH
673 */
674struct rps_dev_flow {
675 u16 cpu;
c445477d 676 u16 filter;
fec5e652
TH
677 unsigned int last_qtail;
678};
c445477d 679#define RPS_NO_FILTER 0xffff
fec5e652
TH
680
681/*
682 * The rps_dev_flow_table structure contains a table of flow mappings.
683 */
684struct rps_dev_flow_table {
685 unsigned int mask;
686 struct rcu_head rcu;
fec5e652
TH
687 struct rps_dev_flow flows[0];
688};
689#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 690 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
691
692/*
693 * The rps_sock_flow_table contains mappings of flows to the last CPU
694 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
695 * Each entry is a 32bit value. Upper part is the high-order bits
696 * of flow hash, lower part is CPU number.
567e4b79 697 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 700 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
701 */
702struct rps_sock_flow_table {
567e4b79 703 u32 mask;
93c1af6c
ED
704
705 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 706};
567e4b79 707#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
708
709#define RPS_NO_CPU 0xffff
710
567e4b79
ED
711extern u32 rps_cpu_mask;
712extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
713
fec5e652
TH
714static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
715 u32 hash)
716{
717 if (table && hash) {
567e4b79
ED
718 unsigned int index = hash & table->mask;
719 u32 val = hash & ~rps_cpu_mask;
fec5e652 720
5e82b4b2 721 /* We only give a hint, preemption can change CPU under us */
567e4b79 722 val |= raw_smp_processor_id();
fec5e652 723
567e4b79
ED
724 if (table->ents[index] != val)
725 table->ents[index] = val;
fec5e652
TH
726 }
727}
728
c445477d 729#ifdef CONFIG_RFS_ACCEL
f629d208
JP
730bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
731 u16 filter_id);
c445477d 732#endif
a953be53 733#endif /* CONFIG_RPS */
c445477d 734
0a9627f2
TH
735/* This structure contains an instance of an RX queue. */
736struct netdev_rx_queue {
a953be53 737#ifdef CONFIG_RPS
6e3f7faf
ED
738 struct rps_map __rcu *rps_map;
739 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 740#endif
6e3f7faf 741 struct kobject kobj;
fe822240 742 struct net_device *dev;
e817f856 743 struct xdp_rxq_info xdp_rxq;
661b8d1b
MK
744#ifdef CONFIG_XDP_SOCKETS
745 struct xdp_umem *umem;
746#endif
0a9627f2 747} ____cacheline_aligned_in_smp;
a953be53
MD
748
749/*
750 * RX queue sysfs structures and functions.
751 */
752struct rx_queue_attribute {
753 struct attribute attr;
718ad681 754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
a953be53 755 ssize_t (*store)(struct netdev_rx_queue *queue,
718ad681 756 const char *buf, size_t len);
a953be53 757};
d314774c 758
bf264145
TH
759#ifdef CONFIG_XPS
760/*
761 * This structure holds an XPS map which can be of variable length. The
762 * map is an array of queues.
763 */
764struct xps_map {
765 unsigned int len;
766 unsigned int alloc_len;
767 struct rcu_head rcu;
768 u16 queues[0];
769};
60b778ce 770#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
771#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
772 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
773
774/*
775 * This structure holds all XPS maps for device. Maps are indexed by CPU.
776 */
777struct xps_dev_maps {
778 struct rcu_head rcu;
80d19669 779 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
bf264145 780};
80d19669
AN
781
782#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
184c449f 783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
80d19669
AN
784
785#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
786 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
787
bf264145
TH
788#endif /* CONFIG_XPS */
789
4f57c087
JF
790#define TC_MAX_QUEUE 16
791#define TC_BITMASK 15
792/* HW offloaded queuing disciplines txq count and offset maps */
793struct netdev_tc_txq {
794 u16 count;
795 u16 offset;
796};
797
68bad94e
NP
798#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
799/*
800 * This structure is to hold information about the device
801 * configured to run FCoE protocol stack.
802 */
803struct netdev_fcoe_hbainfo {
804 char manufacturer[64];
805 char serial_number[64];
806 char hardware_version[64];
807 char driver_version[64];
808 char optionrom_version[64];
809 char firmware_version[64];
810 char model[256];
811 char model_description[256];
812};
813#endif
814
02637fce 815#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 816
02637fce
JP
817/* This structure holds a unique identifier to identify some
818 * physical item (port for example) used by a netdevice.
66b52b0d 819 */
02637fce
JP
820struct netdev_phys_item_id {
821 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
822 unsigned char id_len;
823};
824
d754f98b
SF
825static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
826 struct netdev_phys_item_id *b)
827{
828 return a->id_len == b->id_len &&
829 memcmp(a->id, b->id, a->id_len) == 0;
830}
831
99932d4f 832typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
8ec56fc3
AD
833 struct sk_buff *skb,
834 struct net_device *sb_dev);
99932d4f 835
2572ac53 836enum tc_setup_type {
575ed7d3 837 TC_SETUP_QDISC_MQPRIO,
a1b7c5fd 838 TC_SETUP_CLSU32,
5b33f488 839 TC_SETUP_CLSFLOWER,
ade9b658 840 TC_SETUP_CLSMATCHALL,
332ae8e2 841 TC_SETUP_CLSBPF,
8c4083b3 842 TC_SETUP_BLOCK,
8521db4c 843 TC_SETUP_QDISC_CBS,
602f3baf 844 TC_SETUP_QDISC_RED,
7fdb61b4 845 TC_SETUP_QDISC_PRIO,
f971b132 846 TC_SETUP_QDISC_MQ,
25db26a9 847 TC_SETUP_QDISC_ETF,
98b0e5f6 848 TC_SETUP_ROOT_QDISC,
890d8d23 849 TC_SETUP_QDISC_GRED,
9c66d156 850 TC_SETUP_QDISC_TAPRIO,
c29f74e0 851 TC_SETUP_FT,
d35eb52b 852 TC_SETUP_QDISC_ETS,
ef6aadcc 853 TC_SETUP_QDISC_TBF,
16e5cc64
JF
854};
855
f4e63525
JK
856/* These structures hold the attributes of bpf state that are being passed
857 * to the netdevice through the bpf op.
a7862b45 858 */
f4e63525 859enum bpf_netdev_command {
a7862b45
BB
860 /* Set or clear a bpf program used in the earliest stages of packet
861 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
862 * is responsible for calling bpf_prog_put on any old progs that are
863 * stored. In case of error, the callee need not release the new prog
864 * reference, but on success it takes ownership and must bpf_prog_put
865 * when it is no longer used.
866 */
867 XDP_SETUP_PROG,
ee5d032f 868 XDP_SETUP_PROG_HW,
a7862b45 869 XDP_QUERY_PROG,
a25717d2 870 XDP_QUERY_PROG_HW,
ab3f0063 871 /* BPF program for offload callbacks, invoked at program load time. */
a3884572
JK
872 BPF_OFFLOAD_MAP_ALLOC,
873 BPF_OFFLOAD_MAP_FREE,
74515c57 874 XDP_SETUP_XSK_UMEM,
a7862b45
BB
875};
876
cae1927c 877struct bpf_prog_offload_ops;
ddf9f970 878struct netlink_ext_ack;
74515c57 879struct xdp_umem;
75ccae62 880struct xdp_dev_bulk_queue;
ddf9f970 881
f4e63525
JK
882struct netdev_bpf {
883 enum bpf_netdev_command command;
a7862b45
BB
884 union {
885 /* XDP_SETUP_PROG */
ddf9f970 886 struct {
32d60277 887 u32 flags;
ddf9f970
JK
888 struct bpf_prog *prog;
889 struct netlink_ext_ack *extack;
890 };
a25717d2 891 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
58038695 892 struct {
58038695 893 u32 prog_id;
92f0292b
JK
894 /* flags with which program was installed */
895 u32 prog_flags;
58038695 896 };
a3884572
JK
897 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
898 struct {
899 struct bpf_offloaded_map *offmap;
900 };
f8ebfaf6 901 /* XDP_SETUP_XSK_UMEM */
74515c57 902 struct {
f8ebfaf6
JS
903 struct xdp_umem *umem;
904 u16 queue_id;
74515c57 905 } xsk;
a7862b45
BB
906 };
907};
16e5cc64 908
9116e5e2
MK
909/* Flags for ndo_xsk_wakeup. */
910#define XDP_WAKEUP_RX (1 << 0)
911#define XDP_WAKEUP_TX (1 << 1)
912
d77e38e6
SK
913#ifdef CONFIG_XFRM_OFFLOAD
914struct xfrmdev_ops {
915 int (*xdo_dev_state_add) (struct xfrm_state *x);
916 void (*xdo_dev_state_delete) (struct xfrm_state *x);
917 void (*xdo_dev_state_free) (struct xfrm_state *x);
918 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
919 struct xfrm_state *x);
50bd870a 920 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
d77e38e6
SK
921};
922#endif
923
6c557001
FW
924struct dev_ifalias {
925 struct rcu_head rcuhead;
926 char ifalias[];
927};
928
b473b0d2 929struct devlink;
da68b4ad 930struct tlsdev_ops;
b473b0d2 931
ff927412
JP
932struct netdev_name_node {
933 struct hlist_node hlist;
36fbf1e5 934 struct list_head list;
ff927412
JP
935 struct net_device *dev;
936 const char *name;
937};
938
36fbf1e5
JP
939int netdev_name_node_alt_create(struct net_device *dev, const char *name);
940int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
941
d314774c
SH
942/*
943 * This structure defines the management hooks for network devices.
00829823
SH
944 * The following hooks can be defined; unless noted otherwise, they are
945 * optional and can be filled with a null pointer.
d314774c
SH
946 *
947 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
948 * This function is called once when a network device is registered.
949 * The network device can use this for any late stage initialization
950 * or semantic validation. It can fail with an error code which will
951 * be propagated back to register_netdev.
d314774c
SH
952 *
953 * void (*ndo_uninit)(struct net_device *dev);
954 * This function is called when device is unregistered or when registration
955 * fails. It is not called if init fails.
956 *
957 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 958 * This function is called when a network device transitions to the up
d314774c
SH
959 * state.
960 *
961 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 962 * This function is called when a network device transitions to the down
d314774c
SH
963 * state.
964 *
dc1f8bf6
SH
965 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
966 * struct net_device *dev);
00829823 967 * Called when a packet needs to be transmitted.
e79d8429
RR
968 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
969 * the queue before that can happen; it's for obsolete devices and weird
970 * corner cases, but the stack really does a non-trivial amount
971 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 972 * Required; cannot be NULL.
00829823 973 *
1a2a1444
DM
974 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
975 * struct net_device *dev
976 * netdev_features_t features);
977 * Called by core transmit path to determine if device is capable of
978 * performing offload operations on a given packet. This is to give
979 * the device an opportunity to implement any restrictions that cannot
980 * be otherwise expressed by feature flags. The check is called with
981 * the set of features that the stack has calculated and it returns
982 * those the driver believes to be appropriate.
cdba756f 983 *
f663dd9a 984 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
a350ecce 985 * struct net_device *sb_dev);
5e82b4b2 986 * Called to decide which queue to use when device supports multiple
00829823
SH
987 * transmit queues.
988 *
d314774c
SH
989 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
990 * This function is called to allow device receiver to make
5e82b4b2 991 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
992 *
993 * void (*ndo_set_rx_mode)(struct net_device *dev);
994 * This function is called device changes address list filtering.
01789349 995 * If driver handles unicast address filtering, it should set
5e82b4b2 996 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
997 *
998 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
999 * This function is called when the Media Access Control address
37b607c5 1000 * needs to be changed. If this interface is not defined, the
5e82b4b2 1001 * MAC address can not be changed.
d314774c
SH
1002 *
1003 * int (*ndo_validate_addr)(struct net_device *dev);
1004 * Test if Media Access Control address is valid for the device.
1005 *
1006 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
1007 * Called when a user requests an ioctl which can't be handled by
1008 * the generic interface code. If not defined ioctls return
d314774c
SH
1009 * not supported error code.
1010 *
1011 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1012 * Used to set network devices bus interface parameters. This interface
5e82b4b2 1013 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
1014 * interface (PCI) for low level management.
1015 *
1016 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1017 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 1018 * of a device.
d314774c 1019 *
0290bd29 1020 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
5e82b4b2 1021 * Callback used when the transmitter has not made any progress
d314774c
SH
1022 * for dev->watchdog ticks.
1023 *
bc1f4470 1024 * void (*ndo_get_stats64)(struct net_device *dev,
1025 * struct rtnl_link_stats64 *storage);
d308e38f 1026 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 1027 * Called when a user wants to get the network device usage
be1f3c2c 1028 * statistics. Drivers must do one of the following:
3cfde79c
BH
1029 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1030 * rtnl_link_stats64 structure passed by the caller.
82695d9b 1031 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
1032 * (which should normally be dev->stats) and return a pointer to
1033 * it. The structure may be changed asynchronously only if each
1034 * field is written atomically.
1035 * 3. Update dev->stats asynchronously and atomically, and define
1036 * neither operation.
d314774c 1037 *
3df5b3c6 1038 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
1039 * Return true if this device supports offload stats of this attr_id.
1040 *
1041 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1042 * void *attr_data)
1043 * Get statistics for offload operations by attr_id. Write it into the
1044 * attr_data pointer.
1045 *
5d632cb7 1046 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 1047 * If device supports VLAN filtering this function is called when a
80d5c368 1048 * VLAN id is registered.
d314774c 1049 *
5d632cb7 1050 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 1051 * If device supports VLAN filtering this function is called when a
80d5c368 1052 * VLAN id is unregistered.
d314774c
SH
1053 *
1054 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
1055 *
1056 * SR-IOV management functions.
1057 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
1058 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1059 * u8 qos, __be16 proto);
ed616689
SC
1060 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1061 * int max_tx_rate);
5f8444a3 1062 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 1063 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
1064 * int (*ndo_get_vf_config)(struct net_device *dev,
1065 * int vf, struct ifla_vf_info *ivf);
1d8faf48 1066 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
1067 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1068 * struct nlattr *port[]);
01a3d796
VZ
1069 *
1070 * Enable or disable the VF ability to query its RSS Redirection Table and
1071 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 1072 * with PF and querying it may introduce a theoretical security risk.
01a3d796 1073 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 1074 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
2572ac53 1075 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
de4784ca 1076 * void *type_data);
6a4bc2b4
FF
1077 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1078 * This is always called from the stack with the rtnl lock held and netif
1079 * tx queues stopped. This allows the netdevice to perform queue
1080 * management safely.
c445477d 1081 *
e9bce845
YZ
1082 * Fiber Channel over Ethernet (FCoE) offload functions.
1083 * int (*ndo_fcoe_enable)(struct net_device *dev);
1084 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1085 * so the underlying device can perform whatever needed configuration or
1086 * initialization to support acceleration of FCoE traffic.
1087 *
1088 * int (*ndo_fcoe_disable)(struct net_device *dev);
1089 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1090 * so the underlying device can perform whatever needed clean-ups to
1091 * stop supporting acceleration of FCoE traffic.
1092 *
1093 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1094 * struct scatterlist *sgl, unsigned int sgc);
1095 * Called when the FCoE Initiator wants to initialize an I/O that
1096 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1097 * perform necessary setup and returns 1 to indicate the device is set up
1098 * successfully to perform DDP on this I/O, otherwise this returns 0.
1099 *
1100 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1101 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1102 * indicated by the FC exchange id 'xid', so the underlying device can
1103 * clean up and reuse resources for later DDP requests.
1104 *
1105 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1106 * struct scatterlist *sgl, unsigned int sgc);
1107 * Called when the FCoE Target 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 *
68bad94e
NP
1112 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1113 * struct netdev_fcoe_hbainfo *hbainfo);
1114 * Called when the FCoE Protocol stack wants information on the underlying
1115 * device. This information is utilized by the FCoE protocol stack to
1116 * register attributes with Fiber Channel management service as per the
1117 * FC-GS Fabric Device Management Information(FDMI) specification.
1118 *
e9bce845
YZ
1119 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1120 * Called when the underlying device wants to override default World Wide
1121 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1122 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1123 * protocol stack to use.
1124 *
c445477d
BH
1125 * RFS acceleration.
1126 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1127 * u16 rxq_index, u32 flow_id);
1128 * Set hardware filter for RFS. rxq_index is the target queue index;
1129 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1130 * Return the filter ID on success, or a negative error code.
fbaec0ea 1131 *
8b98a70c 1132 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1133 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1134 * Called to make another netdev an underling.
1135 *
1136 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1137 * Called to release previously enslaved netdev.
5455c699
MM
1138 *
1139 * Feature/offload setting functions.
1a2a1444
DM
1140 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1141 * netdev_features_t features);
1142 * Adjusts the requested feature flags according to device-specific
1143 * constraints, and returns the resulting flags. Must not modify
1144 * the device state.
1145 *
c8f44aff 1146 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1147 * Called to update device configuration to new features. Passed
1148 * feature set might be less than what was returned by ndo_fix_features()).
1149 * Must return >0 or -errno if it changed dev->features itself.
1150 *
edc7d573 1151 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1152 * struct net_device *dev,
87b0984e
PM
1153 * const unsigned char *addr, u16 vid, u16 flags,
1154 * struct netlink_ext_ack *extack);
77162022 1155 * Adds an FDB entry to dev for addr.
1690be63
VY
1156 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1157 * struct net_device *dev,
f6f6424b 1158 * const unsigned char *addr, u16 vid)
77162022
JF
1159 * Deletes the FDB entry from dev coresponding to addr.
1160 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1161 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1162 * int *idx)
77162022
JF
1163 * Used to add FDB entries to dump requests. Implementers should add
1164 * entries to skb and update idx with the number of entries.
e5a55a89 1165 *
ad41faa8 1166 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
2fd527b7 1167 * u16 flags, struct netlink_ext_ack *extack)
e5a55a89 1168 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1169 * struct net_device *dev, u32 filter_mask,
1170 * int nlflags)
ad41faa8
ND
1171 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1172 * u16 flags);
4bf84c35
JP
1173 *
1174 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1175 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1176 * which do not represent real hardware may define this to allow their
1177 * userspace components to manage their virtual carrier state. Devices
1178 * that determine carrier state from physical hardware properties (eg
1179 * network cables) or protocol-dependent mechanisms (eg
1180 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1181 *
1182 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1183 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1184 * Called to get ID of physical port of this device. If driver does
1185 * not implement this, it is assumed that the hw is not able to have
1186 * multiple net devices on single physical port.
53cf5275 1187 *
d6abc596
FF
1188 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1189 * struct netdev_phys_item_id *ppid)
1190 * Called to get the parent ID of the physical port of this device.
1191 *
7c46a640
AD
1192 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1193 * struct udp_tunnel_info *ti);
1194 * Called by UDP tunnel to notify a driver about the UDP port and socket
1195 * address family that a UDP tunnel is listnening to. It is called only
1196 * when a new port starts listening. The operation is protected by the
1197 * RTNL.
1198 *
1199 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1200 * struct udp_tunnel_info *ti);
1201 * Called by UDP tunnel to notify the driver about a UDP port and socket
1202 * address family that the UDP tunnel is not listening to anymore. The
1203 * operation is protected by the RTNL.
1204 *
a6cc0cfa
JF
1205 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1206 * struct net_device *dev)
1207 * Called by upper layer devices to accelerate switching or other
1208 * station functionality into hardware. 'pdev is the lowerdev
1209 * to use for the offload and 'dev' is the net device that will
1210 * back the offload. Returns a pointer to the private structure
1211 * the upper layer will maintain.
1212 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1213 * Called by upper layer device to delete the station created
1214 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1215 * the station and priv is the structure returned by the add
1216 * operation.
822b3b2e
JF
1217 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1218 * int queue_index, u32 maxrate);
1219 * Called when a user wants to set a max-rate limitation of specific
1220 * TX queue.
a54acb3a
ND
1221 * int (*ndo_get_iflink)(const struct net_device *dev);
1222 * Called to get the iflink value of this device.
d746d707 1223 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1224 * bool proto_down);
d746d707
AK
1225 * This function is used to pass protocol port error state information
1226 * to the switch driver. The switch driver can react to the proto_down
1227 * by doing a phys down on the associated switch port.
fc4099f1
PS
1228 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1229 * This function is used to get egress tunnel information for given skb.
1230 * This is useful for retrieving outer tunnel header parameters while
1231 * sampling packet.
871b642a
PA
1232 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1233 * This function is used to specify the headroom that the skb must
1234 * consider when allocation skb during packet reception. Setting
1235 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1236 * forward. Setting a negative value resets the rx headroom to the
871b642a 1237 * default value.
f4e63525 1238 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
a7862b45 1239 * This function is used to set or query state related to XDP on the
f4e63525
JK
1240 * netdevice and manage BPF offload. See definition of
1241 * enum bpf_netdev_command for details.
42b33468
JDB
1242 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1243 * u32 flags);
735fc405
JDB
1244 * This function is used to submit @n XDP packets for transmit on a
1245 * netdevice. Returns number of frames successfully transmitted, frames
1246 * that got dropped are freed/returned via xdp_return_frame().
1247 * Returns negative number, means general error invoking ndo, meaning
1248 * no frames were xmit'ed and core-caller will free all frames.
9116e5e2
MK
1249 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1250 * This function is used to wake up the softirq, ksoftirqd or kthread
1251 * responsible for sending and/or receiving packets on a specific
1252 * queue id bound to an AF_XDP socket. The flags field specifies if
1253 * only RX, only Tx, or both should be woken up using the flags
1254 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
5dc37bb9
JP
1255 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1256 * Get devlink port instance associated with a given netdev.
b473b0d2
JK
1257 * Called with a reference on the netdevice and devlink locks only,
1258 * rtnl_lock is not held.
d314774c
SH
1259 */
1260struct net_device_ops {
1261 int (*ndo_init)(struct net_device *dev);
1262 void (*ndo_uninit)(struct net_device *dev);
1263 int (*ndo_open)(struct net_device *dev);
1264 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1265 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1266 struct net_device *dev);
1267 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1268 struct net_device *dev,
1269 netdev_features_t features);
00829823 1270 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1271 struct sk_buff *skb,
a350ecce 1272 struct net_device *sb_dev);
d314774c
SH
1273 void (*ndo_change_rx_flags)(struct net_device *dev,
1274 int flags);
d314774c 1275 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1276 int (*ndo_set_mac_address)(struct net_device *dev,
1277 void *addr);
d314774c 1278 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1279 int (*ndo_do_ioctl)(struct net_device *dev,
1280 struct ifreq *ifr, int cmd);
d314774c
SH
1281 int (*ndo_set_config)(struct net_device *dev,
1282 struct ifmap *map);
00829823
SH
1283 int (*ndo_change_mtu)(struct net_device *dev,
1284 int new_mtu);
1285 int (*ndo_neigh_setup)(struct net_device *dev,
1286 struct neigh_parms *);
0290bd29
MT
1287 void (*ndo_tx_timeout) (struct net_device *dev,
1288 unsigned int txqueue);
d314774c 1289
bc1f4470 1290 void (*ndo_get_stats64)(struct net_device *dev,
1291 struct rtnl_link_stats64 *storage);
3df5b3c6 1292 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1293 int (*ndo_get_offload_stats)(int attr_id,
1294 const struct net_device *dev,
1295 void *attr_data);
d314774c
SH
1296 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1297
8e586137 1298 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1299 __be16 proto, u16 vid);
8e586137 1300 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1301 __be16 proto, u16 vid);
d314774c 1302#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1303 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1304 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1305 struct netpoll_info *info);
0e34e931 1306 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1307#endif
95c26df8
WM
1308 int (*ndo_set_vf_mac)(struct net_device *dev,
1309 int queue, u8 *mac);
1310 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1311 int queue, u16 vlan,
1312 u8 qos, __be16 proto);
ed616689
SC
1313 int (*ndo_set_vf_rate)(struct net_device *dev,
1314 int vf, int min_tx_rate,
1315 int max_tx_rate);
5f8444a3
GR
1316 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1317 int vf, bool setting);
dd461d6a
HS
1318 int (*ndo_set_vf_trust)(struct net_device *dev,
1319 int vf, bool setting);
95c26df8
WM
1320 int (*ndo_get_vf_config)(struct net_device *dev,
1321 int vf,
1322 struct ifla_vf_info *ivf);
1d8faf48
RE
1323 int (*ndo_set_vf_link_state)(struct net_device *dev,
1324 int vf, int link_state);
3b766cd8
EBE
1325 int (*ndo_get_vf_stats)(struct net_device *dev,
1326 int vf,
1327 struct ifla_vf_stats
1328 *vf_stats);
57b61080
SF
1329 int (*ndo_set_vf_port)(struct net_device *dev,
1330 int vf,
1331 struct nlattr *port[]);
1332 int (*ndo_get_vf_port)(struct net_device *dev,
1333 int vf, struct sk_buff *skb);
30aad417
DG
1334 int (*ndo_get_vf_guid)(struct net_device *dev,
1335 int vf,
1336 struct ifla_vf_guid *node_guid,
1337 struct ifla_vf_guid *port_guid);
cc8e27cc
EC
1338 int (*ndo_set_vf_guid)(struct net_device *dev,
1339 int vf, u64 guid,
1340 int guid_type);
01a3d796
VZ
1341 int (*ndo_set_vf_rss_query_en)(
1342 struct net_device *dev,
1343 int vf, bool setting);
16e5cc64 1344 int (*ndo_setup_tc)(struct net_device *dev,
2572ac53 1345 enum tc_setup_type type,
de4784ca 1346 void *type_data);
d11ead75 1347#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1348 int (*ndo_fcoe_enable)(struct net_device *dev);
1349 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1350 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1351 u16 xid,
1352 struct scatterlist *sgl,
1353 unsigned int sgc);
1354 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1355 u16 xid);
6247e086
YZ
1356 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1357 u16 xid,
1358 struct scatterlist *sgl,
1359 unsigned int sgc);
68bad94e
NP
1360 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1361 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1362#endif
1363
d11ead75 1364#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1365#define NETDEV_FCOE_WWNN 0
1366#define NETDEV_FCOE_WWPN 1
1367 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1368 u64 *wwn, int type);
4d288d57 1369#endif
3c9c36bc 1370
c445477d
BH
1371#ifdef CONFIG_RFS_ACCEL
1372 int (*ndo_rx_flow_steer)(struct net_device *dev,
1373 const struct sk_buff *skb,
1374 u16 rxq_index,
1375 u32 flow_id);
1376#endif
fbaec0ea 1377 int (*ndo_add_slave)(struct net_device *dev,
33eaf2a6
DA
1378 struct net_device *slave_dev,
1379 struct netlink_ext_ack *extack);
fbaec0ea
JP
1380 int (*ndo_del_slave)(struct net_device *dev,
1381 struct net_device *slave_dev);
c8f44aff
MM
1382 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1383 netdev_features_t features);
5455c699 1384 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1385 netdev_features_t features);
503eebc2
JP
1386 int (*ndo_neigh_construct)(struct net_device *dev,
1387 struct neighbour *n);
1388 void (*ndo_neigh_destroy)(struct net_device *dev,
1389 struct neighbour *n);
77162022
JF
1390
1391 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1392 struct nlattr *tb[],
77162022 1393 struct net_device *dev,
6b6e2725 1394 const unsigned char *addr,
f6f6424b 1395 u16 vid,
87b0984e
PM
1396 u16 flags,
1397 struct netlink_ext_ack *extack);
77162022 1398 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1399 struct nlattr *tb[],
77162022 1400 struct net_device *dev,
f6f6424b
JP
1401 const unsigned char *addr,
1402 u16 vid);
77162022
JF
1403 int (*ndo_fdb_dump)(struct sk_buff *skb,
1404 struct netlink_callback *cb,
1405 struct net_device *dev,
5d5eacb3 1406 struct net_device *filter_dev,
d297653d 1407 int *idx);
5b2f94b2
RP
1408 int (*ndo_fdb_get)(struct sk_buff *skb,
1409 struct nlattr *tb[],
1410 struct net_device *dev,
1411 const unsigned char *addr,
1412 u16 vid, u32 portid, u32 seq,
1413 struct netlink_ext_ack *extack);
e5a55a89 1414 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3 1415 struct nlmsghdr *nlh,
2fd527b7
PM
1416 u16 flags,
1417 struct netlink_ext_ack *extack);
e5a55a89
JF
1418 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1419 u32 pid, u32 seq,
6cbdceeb 1420 struct net_device *dev,
46c264da
ND
1421 u32 filter_mask,
1422 int nlflags);
407af329 1423 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1424 struct nlmsghdr *nlh,
1425 u16 flags);
4bf84c35
JP
1426 int (*ndo_change_carrier)(struct net_device *dev,
1427 bool new_carrier);
66b52b0d 1428 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1429 struct netdev_phys_item_id *ppid);
d6abc596
FF
1430 int (*ndo_get_port_parent_id)(struct net_device *dev,
1431 struct netdev_phys_item_id *ppid);
db24a904
DA
1432 int (*ndo_get_phys_port_name)(struct net_device *dev,
1433 char *name, size_t len);
7c46a640
AD
1434 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1435 struct udp_tunnel_info *ti);
1436 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1437 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1438 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1439 struct net_device *dev);
1440 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1441 void *priv);
1442
822b3b2e
JF
1443 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1444 int queue_index,
1445 u32 maxrate);
a54acb3a 1446 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1447 int (*ndo_change_proto_down)(struct net_device *dev,
1448 bool proto_down);
fc4099f1
PS
1449 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1450 struct sk_buff *skb);
871b642a
PA
1451 void (*ndo_set_rx_headroom)(struct net_device *dev,
1452 int needed_headroom);
f4e63525
JK
1453 int (*ndo_bpf)(struct net_device *dev,
1454 struct netdev_bpf *bpf);
735fc405 1455 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
42b33468
JDB
1456 struct xdp_frame **xdp,
1457 u32 flags);
9116e5e2
MK
1458 int (*ndo_xsk_wakeup)(struct net_device *dev,
1459 u32 queue_id, u32 flags);
5dc37bb9 1460 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
d314774c
SH
1461};
1462
7aa98047
LR
1463/**
1464 * enum net_device_priv_flags - &struct net_device priv_flags
1465 *
1466 * These are the &struct net_device, they are only set internally
1467 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1468 * userspace; this means that the order of these flags can change
7aa98047
LR
1469 * during any kernel release.
1470 *
1471 * You should have a pretty good reason to be extending these flags.
1472 *
1473 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1474 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1475 * @IFF_BONDING: bonding master or slave
7aa98047 1476 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1477 * @IFF_WAN_HDLC: WAN HDLC device
1478 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1479 * release skb->dst
1480 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1481 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1482 * @IFF_MACVLAN_PORT: device used as macvlan port
1483 * @IFF_BRIDGE_PORT: device used as bridge port
1484 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1485 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1486 * @IFF_UNICAST_FLT: Supports unicast filtering
1487 * @IFF_TEAM_PORT: device used as team port
1488 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1489 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1490 * change when it's running
1491 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1492 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1493 * underlying stacked devices
007979ea 1494 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1495 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1496 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1497 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1498 * @IFF_TEAM: device is a team device
d4ab4286 1499 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1500 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1501 * entity (i.e. the master device for bridged veth)
3c175784 1502 * @IFF_MACSEC: device is a MACsec device
f5426250 1503 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
30c8bd5a
SS
1504 * @IFF_FAILOVER: device is a failover master device
1505 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
d5256083 1506 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
8065a779 1507 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
7aa98047
LR
1508 */
1509enum netdev_priv_flags {
1510 IFF_802_1Q_VLAN = 1<<0,
1511 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1512 IFF_BONDING = 1<<2,
1513 IFF_ISATAP = 1<<3,
1514 IFF_WAN_HDLC = 1<<4,
1515 IFF_XMIT_DST_RELEASE = 1<<5,
1516 IFF_DONT_BRIDGE = 1<<6,
1517 IFF_DISABLE_NETPOLL = 1<<7,
1518 IFF_MACVLAN_PORT = 1<<8,
1519 IFF_BRIDGE_PORT = 1<<9,
1520 IFF_OVS_DATAPATH = 1<<10,
1521 IFF_TX_SKB_SHARING = 1<<11,
1522 IFF_UNICAST_FLT = 1<<12,
1523 IFF_TEAM_PORT = 1<<13,
1524 IFF_SUPP_NOFCS = 1<<14,
1525 IFF_LIVE_ADDR_CHANGE = 1<<15,
1526 IFF_MACVLAN = 1<<16,
1527 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1ec54cb4
PA
1528 IFF_L3MDEV_MASTER = 1<<18,
1529 IFF_NO_QUEUE = 1<<19,
1530 IFF_OPENVSWITCH = 1<<20,
1531 IFF_L3MDEV_SLAVE = 1<<21,
1532 IFF_TEAM = 1<<22,
1533 IFF_RXFH_CONFIGURED = 1<<23,
1534 IFF_PHONY_HEADROOM = 1<<24,
1535 IFF_MACSEC = 1<<25,
f5426250 1536 IFF_NO_RX_HANDLER = 1<<26,
30c8bd5a
SS
1537 IFF_FAILOVER = 1<<27,
1538 IFF_FAILOVER_SLAVE = 1<<28,
d5256083 1539 IFF_L3MDEV_RX_HANDLER = 1<<29,
8065a779 1540 IFF_LIVE_RENAME_OK = 1<<30,
7aa98047
LR
1541};
1542
1543#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1544#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1545#define IFF_BONDING IFF_BONDING
7aa98047 1546#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1547#define IFF_WAN_HDLC IFF_WAN_HDLC
1548#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1549#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1550#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1551#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1552#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1553#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1554#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1555#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1556#define IFF_TEAM_PORT IFF_TEAM_PORT
1557#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1558#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1559#define IFF_MACVLAN IFF_MACVLAN
02875878 1560#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
007979ea 1561#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1562#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1563#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1564#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1565#define IFF_TEAM IFF_TEAM
d4ab4286 1566#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1567#define IFF_MACSEC IFF_MACSEC
f5426250 1568#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
30c8bd5a
SS
1569#define IFF_FAILOVER IFF_FAILOVER
1570#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
d5256083 1571#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
8065a779 1572#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
7aa98047 1573
536721b1
KK
1574/**
1575 * struct net_device - The DEVICE structure.
d651983d
MCC
1576 *
1577 * Actually, this whole structure is a big mistake. It mixes I/O
1578 * data with strictly "high-level" data, and it has to know about
1579 * almost every data structure used in the INET module.
536721b1
KK
1580 *
1581 * @name: This is the first field of the "visible" part of this structure
1582 * (i.e. as seen by users in the "Space.c" file). It is the name
d651983d 1583 * of the interface.
536721b1 1584 *
ff927412 1585 * @name_node: Name hashlist node
536721b1
KK
1586 * @ifalias: SNMP alias
1587 * @mem_end: Shared memory end
1588 * @mem_start: Shared memory start
1589 * @base_addr: Device I/O address
1590 * @irq: Device IRQ number
1591 *
1592 * @state: Generic network queuing layer state, see netdev_state_t
1593 * @dev_list: The global list of network devices
5e82b4b2
BH
1594 * @napi_list: List entry used for polling NAPI devices
1595 * @unreg_list: List entry when we are unregistering the
1596 * device; see the function unregister_netdev
1597 * @close_list: List entry used when we are closing the device
62d885fe
BP
1598 * @ptype_all: Device-specific packet handlers for all protocols
1599 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1600 *
1601 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1602 * @features: Currently active device features
1603 * @hw_features: User-changeable features
1604 *
1605 * @wanted_features: User-requested features
1606 * @vlan_features: Mask of features inheritable by VLAN devices
1607 *
1608 * @hw_enc_features: Mask of features inherited by encapsulating devices
1609 * This field indicates what encapsulation
1610 * offloads the hardware is capable of doing,
1611 * and drivers will need to set them appropriately.
1612 *
1613 * @mpls_features: Mask of features inheritable by MPLS
1614 *
1615 * @ifindex: interface index
5e82b4b2 1616 * @group: The group the device belongs to
536721b1
KK
1617 *
1618 * @stats: Statistics struct, which was left as a legacy, use
1619 * rtnl_link_stats64 instead
1620 *
1621 * @rx_dropped: Dropped packets by core network,
1622 * do not use this in drivers
1623 * @tx_dropped: Dropped packets by core network,
1624 * do not use this in drivers
6e7333d3
JW
1625 * @rx_nohandler: nohandler dropped packets by core network on
1626 * inactive devices, do not use this in drivers
9e55e5d3
FF
1627 * @carrier_up_count: Number of times the carrier has been up
1628 * @carrier_down_count: Number of times the carrier has been down
536721b1 1629 *
536721b1
KK
1630 * @wireless_handlers: List of functions to handle Wireless Extensions,
1631 * instead of ioctl,
1632 * see <net/iw_handler.h> for details.
1633 * @wireless_data: Instance data managed by the core of wireless extensions
1634 *
1635 * @netdev_ops: Includes several pointers to callbacks,
1636 * if one wants to override the ndo_*() functions
1637 * @ethtool_ops: Management operations
f997c55c
AA
1638 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1639 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1640 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1641 * of Layer 2 headers.
1642 *
1643 * @flags: Interface flags (a la BSD)
1644 * @priv_flags: Like 'flags' but invisible to userspace,
1645 * see if.h for the definitions
1646 * @gflags: Global flags ( kept as legacy )
1647 * @padded: How much padding added by alloc_netdev()
1648 * @operstate: RFC2863 operstate
1649 * @link_mode: Mapping policy to operstate
1650 * @if_port: Selectable AUI, TP, ...
1651 * @dma: DMA channel
1652 * @mtu: Interface MTU value
61e84623
JW
1653 * @min_mtu: Interface Minimum MTU value
1654 * @max_mtu: Interface Maximum MTU value
536721b1 1655 * @type: Interface hardware type
2793a23a 1656 * @hard_header_len: Maximum hardware header length.
217e6fa2 1657 * @min_header_len: Minimum hardware header length
536721b1
KK
1658 *
1659 * @needed_headroom: Extra headroom the hardware may need, but not in all
1660 * cases can this be guaranteed
1661 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1662 * cases can this be guaranteed. Some cases also use
1663 * LL_MAX_HEADER instead to allocate the skb
1664 *
1665 * interface address info:
1666 *
1667 * @perm_addr: Permanent hw address
1668 * @addr_assign_type: Hw address assignment type
1669 * @addr_len: Hardware address length
5343da4c
TY
1670 * @upper_level: Maximum depth level of upper devices.
1671 * @lower_level: Maximum depth level of lower devices.
8626a0c8 1672 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1673 * @dev_id: Used to differentiate devices that share
1674 * the same link layer address
1675 * @dev_port: Used to differentiate devices that share
1676 * the same function
1677 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1678 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1679 * has been enabled due to the need to listen to
1680 * additional unicast addresses in a device that
1681 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1682 * @uc: unicast mac addresses
1683 * @mc: multicast mac addresses
1684 * @dev_addrs: list of device hw addresses
1685 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1686 * @promiscuity: Number of times the NIC is told to work in
1687 * promiscuous mode; if it becomes 0 the NIC will
1688 * exit promiscuous mode
536721b1
KK
1689 * @allmulti: Counter, enables or disables allmulticast mode
1690 *
1691 * @vlan_info: VLAN info
1692 * @dsa_ptr: dsa specific data
1693 * @tipc_ptr: TIPC specific data
1694 * @atalk_ptr: AppleTalk link
1695 * @ip_ptr: IPv4 specific data
1696 * @dn_ptr: DECnet specific data
1697 * @ip6_ptr: IPv6 specific data
1698 * @ax25_ptr: AX.25 specific data
1699 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1700 *
536721b1
KK
1701 * @dev_addr: Hw address (before bcast,
1702 * because most packets are unicast)
1703 *
1704 * @_rx: Array of RX queues
1705 * @num_rx_queues: Number of RX queues
1706 * allocated at register_netdev() time
1707 * @real_num_rx_queues: Number of RX queues currently active in device
1708 *
1709 * @rx_handler: handler for received packets
1710 * @rx_handler_data: XXX: need comments on this one
46209401
JP
1711 * @miniq_ingress: ingress/clsact qdisc specific data for
1712 * ingress processing
536721b1 1713 * @ingress_queue: XXX: need comments on this one
2f5e70c8 1714 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
536721b1
KK
1715 * @broadcast: hw bcast address
1716 *
14ffbbb8
TG
1717 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1718 * indexed by RX queue number. Assigned by driver.
1719 * This must only be set if the ndo_rx_flow_steer
1720 * operation is defined
1721 * @index_hlist: Device index hash chain
1722 *
536721b1
KK
1723 * @_tx: Array of TX queues
1724 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1725 * @real_num_tx_queues: Number of TX queues currently active in device
1726 * @qdisc: Root qdisc from userspace point of view
1727 * @tx_queue_len: Max frames per queue allowed
1728 * @tx_global_lock: XXX: need comments on this one
1729 *
1730 * @xps_maps: XXX: need comments on this one
46209401
JP
1731 * @miniq_egress: clsact qdisc specific data for
1732 * egress processing
536721b1 1733 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1734 * the watchdog (see dev_watchdog())
536721b1
KK
1735 * @watchdog_timer: List of timers
1736 *
1737 * @pcpu_refcnt: Number of references to this device
1738 * @todo_list: Delayed register/unregister
536721b1
KK
1739 * @link_watch_list: XXX: need comments on this one
1740 *
1741 * @reg_state: Register/unregister state machine
1742 * @dismantle: Device is going to be freed
1743 * @rtnl_link_state: This enum represents the phases of creating
1744 * a new link
1745 *
cf124db5
DM
1746 * @needs_free_netdev: Should unregister perform free_netdev?
1747 * @priv_destructor: Called from unregister
536721b1
KK
1748 * @npinfo: XXX: need comments on this one
1749 * @nd_net: Network namespace this network device is inside
1750 *
1751 * @ml_priv: Mid-layer private
1752 * @lstats: Loopback statistics
1753 * @tstats: Tunnel statistics
1754 * @dstats: Dummy statistics
1755 * @vstats: Virtual ethernet statistics
1756 *
1757 * @garp_port: GARP
1758 * @mrp_port: MRP
1759 *
1760 * @dev: Class/net/name entry
1761 * @sysfs_groups: Space for optional device, statistics and wireless
1762 * sysfs groups
1763 *
1764 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1765 * @rtnl_link_ops: Rtnl_link_ops
1766 *
1767 * @gso_max_size: Maximum size of generic segmentation offload
1768 * @gso_max_segs: Maximum number of segments that can be passed to the
1769 * NIC for GSO
1770 *
1771 * @dcbnl_ops: Data Center Bridging netlink ops
1772 * @num_tc: Number of traffic classes in the net device
1773 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1774 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1775 *
1776 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1777 *
1778 * @priomap: XXX: need comments on this one
1779 * @phydev: Physical device may attach itself
1780 * for hardware timestamping
e679c9c1 1781 * @sfp_bus: attached &struct sfp_bus structure.
ab92d68f 1782 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock
1f26c0d3 1783 * spinlock
ab92d68f
TY
1784 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1785 * @qdisc_xmit_lock_key: lockdep class annotating
1786 * netdev_queue->_xmit_lock spinlock
1787 * @addr_list_lock_key: lockdep class annotating
1788 * net_device->addr_list_lock spinlock
536721b1 1789 *
d746d707
AK
1790 * @proto_down: protocol port state information can be sent to the
1791 * switch driver and used to set the phys state of the
1792 * switch port.
1793 *
61941143
HK
1794 * @wol_enabled: Wake-on-LAN is enabled
1795 *
1da177e4
LT
1796 * FIXME: cleanup struct net_device such that network protocol info
1797 * moves out.
1798 */
1799
d94d9fee 1800struct net_device {
1da177e4 1801 char name[IFNAMSIZ];
ff927412 1802 struct netdev_name_node *name_node;
6c557001 1803 struct dev_ifalias __rcu *ifalias;
1da177e4
LT
1804 /*
1805 * I/O specific fields
1806 * FIXME: Merge these and struct ifmap into one
1807 */
536721b1
KK
1808 unsigned long mem_end;
1809 unsigned long mem_start;
1810 unsigned long base_addr;
1811 int irq;
1da177e4
LT
1812
1813 /*
536721b1
KK
1814 * Some hardware also needs these fields (state,dev_list,
1815 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1816 * part of the usual set specified in Space.c.
1817 */
1818
1da177e4
LT
1819 unsigned long state;
1820
7562f876 1821 struct list_head dev_list;
bea3348e 1822 struct list_head napi_list;
44a0873d 1823 struct list_head unreg_list;
5cde2829 1824 struct list_head close_list;
7866a621
SN
1825 struct list_head ptype_all;
1826 struct list_head ptype_specific;
2f268f12 1827
2f268f12
VF
1828 struct {
1829 struct list_head upper;
1830 struct list_head lower;
1831 } adj_list;
1832
c8f44aff 1833 netdev_features_t features;
c8f44aff 1834 netdev_features_t hw_features;
c8f44aff 1835 netdev_features_t wanted_features;
c8f44aff 1836 netdev_features_t vlan_features;
6a674e9c 1837 netdev_features_t hw_enc_features;
0d89d203 1838 netdev_features_t mpls_features;
802ab55a 1839 netdev_features_t gso_partial_features;
04ed3e74 1840
1da177e4 1841 int ifindex;
7a66bbc9 1842 int group;
1da177e4 1843
c45d286e 1844 struct net_device_stats stats;
015f0688 1845
015f0688
ED
1846 atomic_long_t rx_dropped;
1847 atomic_long_t tx_dropped;
6e7333d3 1848 atomic_long_t rx_nohandler;
1da177e4 1849
b2d3bcfa
DD
1850 /* Stats to monitor link on/off, flapping */
1851 atomic_t carrier_up_count;
1852 atomic_t carrier_down_count;
1853
b86e0280 1854#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1855 const struct iw_handler_def *wireless_handlers;
1856 struct iw_public_data *wireless_data;
b86e0280 1857#endif
d314774c 1858 const struct net_device_ops *netdev_ops;
76fd8593 1859 const struct ethtool_ops *ethtool_ops;
1b69c6d0
DA
1860#ifdef CONFIG_NET_L3_MASTER_DEV
1861 const struct l3mdev_ops *l3mdev_ops;
1862#endif
f997c55c
AA
1863#if IS_ENABLED(CONFIG_IPV6)
1864 const struct ndisc_ops *ndisc_ops;
1865#endif
1da177e4 1866
9cb0d21d 1867#ifdef CONFIG_XFRM_OFFLOAD
d77e38e6
SK
1868 const struct xfrmdev_ops *xfrmdev_ops;
1869#endif
1870
a5c37c63
IL
1871#if IS_ENABLED(CONFIG_TLS_DEVICE)
1872 const struct tlsdev_ops *tlsdev_ops;
1873#endif
1874
3b04ddde
SH
1875 const struct header_ops *header_ops;
1876
536721b1
KK
1877 unsigned int flags;
1878 unsigned int priv_flags;
1879
1da177e4 1880 unsigned short gflags;
536721b1 1881 unsigned short padded;
1da177e4 1882
536721b1
KK
1883 unsigned char operstate;
1884 unsigned char link_mode;
b00055aa 1885
536721b1
KK
1886 unsigned char if_port;
1887 unsigned char dma;
bdc220da 1888
501a90c9
ED
1889 /* Note : dev->mtu is often read without holding a lock.
1890 * Writers usually hold RTNL.
1891 * It is recommended to use READ_ONCE() to annotate the reads,
1892 * and to use WRITE_ONCE() to annotate the writes.
1893 */
536721b1 1894 unsigned int mtu;
61e84623
JW
1895 unsigned int min_mtu;
1896 unsigned int max_mtu;
536721b1
KK
1897 unsigned short type;
1898 unsigned short hard_header_len;
d92be7a4 1899 unsigned char min_header_len;
1da177e4 1900
f5184d26
JB
1901 unsigned short needed_headroom;
1902 unsigned short needed_tailroom;
1903
1da177e4 1904 /* Interface address info. */
536721b1
KK
1905 unsigned char perm_addr[MAX_ADDR_LEN];
1906 unsigned char addr_assign_type;
1907 unsigned char addr_len;
5343da4c
TY
1908 unsigned char upper_level;
1909 unsigned char lower_level;
a0a9663d 1910 unsigned short neigh_priv_len;
536721b1
KK
1911 unsigned short dev_id;
1912 unsigned short dev_port;
ccffad25 1913 spinlock_t addr_list_lock;
14ffbbb8
TG
1914 unsigned char name_assign_type;
1915 bool uc_promisc;
536721b1
KK
1916 struct netdev_hw_addr_list uc;
1917 struct netdev_hw_addr_list mc;
1918 struct netdev_hw_addr_list dev_addrs;
1919
4c3d5e7b
ED
1920#ifdef CONFIG_SYSFS
1921 struct kset *queues_kset;
1922#endif
9d45abe1
WC
1923 unsigned int promiscuity;
1924 unsigned int allmulti;
1da177e4 1925
1da177e4 1926
5e82b4b2 1927 /* Protocol-specific pointers */
65ac6a5f 1928
d11ead75 1929#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1930 struct vlan_info __rcu *vlan_info;
65ac6a5f 1931#endif
34a430d7 1932#if IS_ENABLED(CONFIG_NET_DSA)
2f657a60 1933 struct dsa_port *dsa_ptr;
37cb0620
YX
1934#endif
1935#if IS_ENABLED(CONFIG_TIPC)
536721b1 1936 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1937#endif
89e58148 1938#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
536721b1 1939 void *atalk_ptr;
89e58148 1940#endif
536721b1 1941 struct in_device __rcu *ip_ptr;
330c7272 1942#if IS_ENABLED(CONFIG_DECNET)
536721b1 1943 struct dn_dev __rcu *dn_ptr;
330c7272 1944#endif
536721b1 1945 struct inet6_dev __rcu *ip6_ptr;
19ff13f2 1946#if IS_ENABLED(CONFIG_AX25)
536721b1 1947 void *ax25_ptr;
19ff13f2 1948#endif
536721b1 1949 struct wireless_dev *ieee80211_ptr;
98a18b6f 1950 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1951#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1952 struct mpls_dev __rcu *mpls_ptr;
1953#endif
1da177e4 1954
9356b8fc 1955/*
cd13539b 1956 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1957 */
9356b8fc 1958 /* Interface address info used in eth_type_trans() */
536721b1 1959 unsigned char *dev_addr;
f001fde5 1960
0a9627f2 1961 struct netdev_rx_queue *_rx;
0a9627f2 1962 unsigned int num_rx_queues;
62fe0b40 1963 unsigned int real_num_rx_queues;
0a9627f2 1964
7acedaf5 1965 struct bpf_prog __rcu *xdp_prog;
3b47d303 1966 unsigned long gro_flush_timeout;
61391cde 1967 rx_handler_func_t __rcu *rx_handler;
1968 void __rcu *rx_handler_data;
e8a0464c 1969
4cda01e8 1970#ifdef CONFIG_NET_CLS_ACT
46209401 1971 struct mini_Qdisc __rcu *miniq_ingress;
d2788d34 1972#endif
24824a09 1973 struct netdev_queue __rcu *ingress_queue;
e687ad60 1974#ifdef CONFIG_NETFILTER_INGRESS
960632ec 1975 struct nf_hook_entries __rcu *nf_hooks_ingress;
e687ad60 1976#endif
d2788d34 1977
536721b1 1978 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1979#ifdef CONFIG_RFS_ACCEL
1980 struct cpu_rmap *rx_cpu_rmap;
1981#endif
1982 struct hlist_node index_hlist;
cd13539b
ED
1983
1984/*
1985 * Cache lines mostly used on transmit path
1986 */
e8a0464c
DM
1987 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1988 unsigned int num_tx_queues;
fd2ea0a7 1989 unsigned int real_num_tx_queues;
af356afa 1990 struct Qdisc *qdisc;
0cd29503 1991 unsigned int tx_queue_len;
c3f26a26 1992 spinlock_t tx_global_lock;
75ccae62
THJ
1993
1994 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
cd13539b 1995
bf264145 1996#ifdef CONFIG_XPS
80d19669
AN
1997 struct xps_dev_maps __rcu *xps_cpus_map;
1998 struct xps_dev_maps __rcu *xps_rxqs_map;
bf264145 1999#endif
1f211a1b 2000#ifdef CONFIG_NET_CLS_ACT
46209401 2001 struct mini_Qdisc __rcu *miniq_egress;
1f211a1b 2002#endif
0c4f691f 2003
75ccae62
THJ
2004#ifdef CONFIG_NET_SCHED
2005 DECLARE_HASHTABLE (qdisc_hash, 4);
2006#endif
9356b8fc 2007 /* These may be needed for future network-power-down code. */
9356b8fc 2008 struct timer_list watchdog_timer;
75ccae62 2009 int watchdog_timeo;
9356b8fc 2010
1da177e4 2011 struct list_head todo_list;
75ccae62 2012 int __percpu *pcpu_refcnt;
1da177e4 2013
e014debe 2014 struct list_head link_watch_list;
572a103d 2015
1da177e4 2016 enum { NETREG_UNINITIALIZED=0,
b17a7c17 2017 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
2018 NETREG_UNREGISTERING, /* called unregister_netdevice */
2019 NETREG_UNREGISTERED, /* completed unregister todo */
2020 NETREG_RELEASED, /* called free_netdev */
937f1ba5 2021 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
2022 } reg_state:8;
2023
536721b1 2024 bool dismantle;
a2835763
PM
2025
2026 enum {
2027 RTNL_LINK_INITIALIZED,
2028 RTNL_LINK_INITIALIZING,
2029 } rtnl_link_state:16;
1da177e4 2030
cf124db5
DM
2031 bool needs_free_netdev;
2032 void (*priv_destructor)(struct net_device *dev);
1da177e4 2033
1da177e4 2034#ifdef CONFIG_NETPOLL
5fbee843 2035 struct netpoll_info __rcu *npinfo;
1da177e4 2036#endif
eae792b7 2037
0c5c9fb5 2038 possible_net_t nd_net;
4a1c5371 2039
4951704b 2040 /* mid-layer private */
a7855c78 2041 union {
536721b1
KK
2042 void *ml_priv;
2043 struct pcpu_lstats __percpu *lstats;
8f84985f 2044 struct pcpu_sw_netstats __percpu *tstats;
536721b1 2045 struct pcpu_dstats __percpu *dstats;
a7855c78 2046 };
536721b1 2047
fb585b44 2048#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 2049 struct garp_port __rcu *garp_port;
fb585b44
TK
2050#endif
2051#if IS_ENABLED(CONFIG_MRP)
febf018d 2052 struct mrp_port __rcu *mrp_port;
fb585b44 2053#endif
1da177e4 2054
5e82b4b2 2055 struct device dev;
0c509a6c 2056 const struct attribute_group *sysfs_groups[4];
a953be53 2057 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 2058
38f7b870 2059 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 2060
82cc1a7a
PWJ
2061 /* for setting kernel sock attribute on TCP connection setup */
2062#define GSO_MAX_SIZE 65536
2063 unsigned int gso_max_size;
30b678d8
BH
2064#define GSO_MAX_SEGS 65535
2065 u16 gso_max_segs;
743b03a8 2066
7a6b6f51 2067#ifdef CONFIG_DCB
32953543 2068 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 2069#endif
ffcfe25b 2070 s16 num_tc;
5e82b4b2
BH
2071 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2072 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 2073
d11ead75 2074#if IS_ENABLED(CONFIG_FCOE)
4d288d57 2075 unsigned int fcoe_ddp_xid;
5bc1421e 2076#endif
86f8515f 2077#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 2078 struct netprio_map __rcu *priomap;
4d288d57 2079#endif
5e82b4b2 2080 struct phy_device *phydev;
e679c9c1 2081 struct sfp_bus *sfp_bus;
ab92d68f
TY
2082 struct lock_class_key qdisc_tx_busylock_key;
2083 struct lock_class_key qdisc_running_key;
2084 struct lock_class_key qdisc_xmit_lock_key;
2085 struct lock_class_key addr_list_lock_key;
5e82b4b2 2086 bool proto_down;
61941143 2087 unsigned wol_enabled:1;
1da177e4 2088};
43cb76d9 2089#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 2090
b5cdae32
DM
2091static inline bool netif_elide_gro(const struct net_device *dev)
2092{
2093 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2094 return true;
2095 return false;
2096}
2097
1da177e4 2098#define NETDEV_ALIGN 32
1da177e4 2099
4f57c087
JF
2100static inline
2101int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2102{
2103 return dev->prio_tc_map[prio & TC_BITMASK];
2104}
2105
2106static inline
2107int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2108{
2109 if (tc >= dev->num_tc)
2110 return -EINVAL;
2111
2112 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2113 return 0;
2114}
2115
8d059b0f 2116int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
2117void netdev_reset_tc(struct net_device *dev);
2118int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2119int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
2120
2121static inline
2122int netdev_get_num_tc(struct net_device *dev)
2123{
2124 return dev->num_tc;
2125}
2126
ffcfe25b
AD
2127void netdev_unbind_sb_channel(struct net_device *dev,
2128 struct net_device *sb_dev);
2129int netdev_bind_sb_channel_queue(struct net_device *dev,
2130 struct net_device *sb_dev,
2131 u8 tc, u16 count, u16 offset);
2132int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2133static inline int netdev_get_sb_channel(struct net_device *dev)
2134{
2135 return max_t(int, -dev->num_tc, 0);
2136}
2137
e8a0464c
DM
2138static inline
2139struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2140 unsigned int index)
2141{
2142 return &dev->_tx[index];
2143}
2144
10c51b56
DB
2145static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2146 const struct sk_buff *skb)
2147{
2148 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2149}
2150
e8a0464c
DM
2151static inline void netdev_for_each_tx_queue(struct net_device *dev,
2152 void (*f)(struct net_device *,
2153 struct netdev_queue *,
2154 void *),
2155 void *arg)
2156{
2157 unsigned int i;
2158
2159 for (i = 0; i < dev->num_tx_queues; i++)
2160 f(dev, &dev->_tx[i], arg);
2161}
2162
b71b5837
PA
2163u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2164 struct net_device *sb_dev);
4bd97d51
PA
2165struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2166 struct sk_buff *skb,
2167 struct net_device *sb_dev);
8c4c49df 2168
871b642a
PA
2169/* returns the headroom that the master device needs to take in account
2170 * when forwarding to this dev
2171 */
2172static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2173{
2174 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2175}
2176
2177static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2178{
2179 if (dev->netdev_ops->ndo_set_rx_headroom)
2180 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2181}
2182
2183/* set the device rx headroom to the dev's default */
2184static inline void netdev_reset_rx_headroom(struct net_device *dev)
2185{
2186 netdev_set_rx_headroom(dev, -1);
2187}
2188
c346dca1
YH
2189/*
2190 * Net namespace inlines
2191 */
2192static inline
2193struct net *dev_net(const struct net_device *dev)
2194{
c2d9ba9b 2195 return read_pnet(&dev->nd_net);
c346dca1
YH
2196}
2197
2198static inline
f5aa23fd 2199void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2200{
0c5c9fb5 2201 write_pnet(&dev->nd_net, net);
c346dca1
YH
2202}
2203
bea3348e
SH
2204/**
2205 * netdev_priv - access network device private data
2206 * @dev: network device
2207 *
2208 * Get network device private data
2209 */
6472ce60 2210static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2211{
1ce8e7b5 2212 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2213}
2214
1da177e4
LT
2215/* Set the sysfs physical device reference for the network logical device
2216 * if set prior to registration will cause a symlink during initialization.
2217 */
43cb76d9 2218#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2219
384912ed 2220/* Set the sysfs device type for the network logical device to allow
3f79410c 2221 * fine-grained identification of different network device types. For
5e82b4b2 2222 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2223 */
2224#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2225
82dc3c63
ED
2226/* Default NAPI poll() weight
2227 * Device drivers are strongly advised to not use bigger value
2228 */
2229#define NAPI_POLL_WEIGHT 64
2230
3b582cc1 2231/**
5e82b4b2 2232 * netif_napi_add - initialize a NAPI context
3b582cc1 2233 * @dev: network device
5e82b4b2 2234 * @napi: NAPI context
3b582cc1
SH
2235 * @poll: polling function
2236 * @weight: default weight
2237 *
5e82b4b2
BH
2238 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2239 * *any* of the other NAPI-related functions.
3b582cc1 2240 */
d565b0a1
HX
2241void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2242 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2243
d64b5e85 2244/**
5e82b4b2 2245 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2246 * @dev: network device
5e82b4b2 2247 * @napi: NAPI context
d64b5e85
ED
2248 * @poll: polling function
2249 * @weight: default weight
2250 *
2251 * This variant of netif_napi_add() should be used from drivers using NAPI
2252 * to exclusively poll a TX queue.
2253 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2254 */
2255static inline void netif_tx_napi_add(struct net_device *dev,
2256 struct napi_struct *napi,
2257 int (*poll)(struct napi_struct *, int),
2258 int weight)
2259{
2260 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2261 netif_napi_add(dev, napi, poll, weight);
2262}
2263
d8156534 2264/**
5e82b4b2
BH
2265 * netif_napi_del - remove a NAPI context
2266 * @napi: NAPI context
d8156534 2267 *
5e82b4b2 2268 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2269 */
d565b0a1
HX
2270void netif_napi_del(struct napi_struct *napi);
2271
2272struct napi_gro_cb {
78a478d0 2273 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2274 void *frag0;
78a478d0 2275
7489594c
HX
2276 /* Length of frag0. */
2277 unsigned int frag0_len;
2278
86911732 2279 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2280 int data_offset;
86911732 2281
d565b0a1 2282 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2283 u16 flush;
2284
2285 /* Save the IP ID here and check when we get to the transport layer */
2286 u16 flush_id;
d565b0a1
HX
2287
2288 /* Number of segments aggregated. */
2e71a6f8
ED
2289 u16 count;
2290
15e2396d
TH
2291 /* Start offset for remote checksum offload */
2292 u16 gro_remcsum_start;
2293
2e71a6f8
ED
2294 /* jiffies when first packet was created/queued */
2295 unsigned long age;
86347245 2296
afe93325 2297 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2298 u16 proto;
2299
baa32ff4
TH
2300 /* This is non-zero if the packet may be of the same flow. */
2301 u8 same_flow:1;
2302
fac8e0f5
JG
2303 /* Used in tunnel GRO receive */
2304 u8 encap_mark:1;
573e8fca
TH
2305
2306 /* GRO checksum is valid */
2307 u8 csum_valid:1;
2308
662880f4
TH
2309 /* Number of checksums via CHECKSUM_UNNECESSARY */
2310 u8 csum_cnt:3;
c3c7c254 2311
baa32ff4
TH
2312 /* Free the skb? */
2313 u8 free:2;
2314#define NAPI_GRO_FREE 1
2315#define NAPI_GRO_FREE_STOLEN_HEAD 2
2316
efc98d08
TH
2317 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2318 u8 is_ipv6:1;
2319
a0ca153f
AD
2320 /* Used in GRE, set in fou/gue_gro_receive */
2321 u8 is_fou:1;
2322
1530545e
AD
2323 /* Used to determine if flush_id can be ignored */
2324 u8 is_atomic:1;
2325
fcd91dd4
SD
2326 /* Number of gro_receive callbacks this packet already went through */
2327 u8 recursion_counter:4;
2328
3a1296a3
SK
2329 /* GRO is done by frag_list pointer chaining. */
2330 u8 is_flist:1;
baa32ff4 2331
bf5a755f
JC
2332 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2333 __wsum csum;
2334
c3c7c254
ED
2335 /* used in skb_gro_receive() slow path */
2336 struct sk_buff *last;
d565b0a1
HX
2337};
2338
2339#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2340
fcd91dd4
SD
2341#define GRO_RECURSION_LIMIT 15
2342static inline int gro_recursion_inc_test(struct sk_buff *skb)
2343{
2344 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2345}
2346
d4546c25
DM
2347typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2348static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2349 struct list_head *head,
2350 struct sk_buff *skb)
fcd91dd4
SD
2351{
2352 if (unlikely(gro_recursion_inc_test(skb))) {
2353 NAPI_GRO_CB(skb)->flush |= 1;
2354 return NULL;
2355 }
2356
2357 return cb(head, skb);
2358}
2359
d4546c25
DM
2360typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2361 struct sk_buff *);
2362static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2363 struct sock *sk,
2364 struct list_head *head,
2365 struct sk_buff *skb)
fcd91dd4
SD
2366{
2367 if (unlikely(gro_recursion_inc_test(skb))) {
2368 NAPI_GRO_CB(skb)->flush |= 1;
2369 return NULL;
2370 }
2371
2372 return cb(sk, head, skb);
2373}
2374
1da177e4 2375struct packet_type {
f2ccd8fa 2376 __be16 type; /* This is really htons(ether_type). */
fa788d98 2377 bool ignore_outgoing;
f2ccd8fa
DM
2378 struct net_device *dev; /* NULL is wildcarded here */
2379 int (*func) (struct sk_buff *,
2380 struct net_device *,
2381 struct packet_type *,
2382 struct net_device *);
17266ee9
EC
2383 void (*list_func) (struct list_head *,
2384 struct packet_type *,
2385 struct net_device *);
c0de08d0
EL
2386 bool (*id_match)(struct packet_type *ptype,
2387 struct sock *sk);
1da177e4
LT
2388 void *af_packet_priv;
2389 struct list_head list;
2390};
2391
f191a1d1 2392struct offload_callbacks {
576a30eb 2393 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2394 netdev_features_t features);
d4546c25
DM
2395 struct sk_buff *(*gro_receive)(struct list_head *head,
2396 struct sk_buff *skb);
299603e8 2397 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2398};
2399
2400struct packet_offload {
2401 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2402 u16 priority;
f191a1d1
VY
2403 struct offload_callbacks callbacks;
2404 struct list_head list;
1da177e4
LT
2405};
2406
5e82b4b2 2407/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2408struct pcpu_sw_netstats {
2409 u64 rx_packets;
2410 u64 rx_bytes;
2411 u64 tx_packets;
2412 u64 tx_bytes;
2413 struct u64_stats_sync syncp;
9a5ee462 2414} __aligned(4 * sizeof(u64));
52bb6677
LR
2415
2416struct pcpu_lstats {
fd2f4737
ED
2417 u64_stats_t packets;
2418 u64_stats_t bytes;
52bb6677 2419 struct u64_stats_sync syncp;
9a5ee462 2420} __aligned(2 * sizeof(u64));
8f84985f 2421
de7d5084
ED
2422void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2423
dd5382a0
ED
2424static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2425{
2426 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2427
2428 u64_stats_update_begin(&lstats->syncp);
fd2f4737
ED
2429 u64_stats_add(&lstats->bytes, len);
2430 u64_stats_inc(&lstats->packets);
dd5382a0
ED
2431 u64_stats_update_end(&lstats->syncp);
2432}
2433
aabc92bb
PNA
2434#define __netdev_alloc_pcpu_stats(type, gfp) \
2435({ \
2436 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2437 if (pcpu_stats) { \
2438 int __cpu; \
2439 for_each_possible_cpu(__cpu) { \
2440 typeof(type) *stat; \
2441 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2442 u64_stats_init(&stat->syncp); \
2443 } \
2444 } \
2445 pcpu_stats; \
1c213bd2
WC
2446})
2447
aabc92bb 2448#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2449 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2450
764f5e54
JP
2451enum netdev_lag_tx_type {
2452 NETDEV_LAG_TX_TYPE_UNKNOWN,
2453 NETDEV_LAG_TX_TYPE_RANDOM,
2454 NETDEV_LAG_TX_TYPE_BROADCAST,
2455 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2456 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2457 NETDEV_LAG_TX_TYPE_HASH,
2458};
2459
f44aa9ef
JH
2460enum netdev_lag_hash {
2461 NETDEV_LAG_HASH_NONE,
2462 NETDEV_LAG_HASH_L2,
2463 NETDEV_LAG_HASH_L34,
2464 NETDEV_LAG_HASH_L23,
2465 NETDEV_LAG_HASH_E23,
2466 NETDEV_LAG_HASH_E34,
2467 NETDEV_LAG_HASH_UNKNOWN,
2468};
2469
764f5e54
JP
2470struct netdev_lag_upper_info {
2471 enum netdev_lag_tx_type tx_type;
f44aa9ef 2472 enum netdev_lag_hash hash_type;
764f5e54
JP
2473};
2474
fb1b2e3c
JP
2475struct netdev_lag_lower_state_info {
2476 u8 link_up : 1,
2477 tx_enabled : 1;
2478};
2479
1da177e4
LT
2480#include <linux/notifier.h>
2481
ede2762d
KT
2482/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2483 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2484 * adding new types.
dcfe1421 2485 */
ede2762d
KT
2486enum netdev_cmd {
2487 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2488 NETDEV_DOWN,
2489 NETDEV_REBOOT, /* Tell a protocol stack a network interface
dcfe1421
AW
2490 detected a hardware crash and restarted
2491 - we can use this eg to kick tcp sessions
2492 once done */
ede2762d
KT
2493 NETDEV_CHANGE, /* Notify device state change */
2494 NETDEV_REGISTER,
2495 NETDEV_UNREGISTER,
2496 NETDEV_CHANGEMTU, /* notify after mtu change happened */
1570415f
PM
2497 NETDEV_CHANGEADDR, /* notify after the address change */
2498 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
ede2762d
KT
2499 NETDEV_GOING_DOWN,
2500 NETDEV_CHANGENAME,
2501 NETDEV_FEAT_CHANGE,
2502 NETDEV_BONDING_FAILOVER,
2503 NETDEV_PRE_UP,
2504 NETDEV_PRE_TYPE_CHANGE,
2505 NETDEV_POST_TYPE_CHANGE,
2506 NETDEV_POST_INIT,
ede2762d
KT
2507 NETDEV_RELEASE,
2508 NETDEV_NOTIFY_PEERS,
2509 NETDEV_JOIN,
2510 NETDEV_CHANGEUPPER,
2511 NETDEV_RESEND_IGMP,
2512 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2513 NETDEV_CHANGEINFODATA,
2514 NETDEV_BONDING_INFO,
2515 NETDEV_PRECHANGEUPPER,
2516 NETDEV_CHANGELOWERSTATE,
2517 NETDEV_UDP_TUNNEL_PUSH_INFO,
2518 NETDEV_UDP_TUNNEL_DROP_INFO,
2519 NETDEV_CHANGE_TX_QUEUE_LEN,
9daae9bd
GP
2520 NETDEV_CVLAN_FILTER_PUSH_INFO,
2521 NETDEV_CVLAN_FILTER_DROP_INFO,
2522 NETDEV_SVLAN_FILTER_PUSH_INFO,
2523 NETDEV_SVLAN_FILTER_DROP_INFO,
ede2762d
KT
2524};
2525const char *netdev_cmd_to_name(enum netdev_cmd cmd);
dcfe1421 2526
f629d208
JP
2527int register_netdevice_notifier(struct notifier_block *nb);
2528int unregister_netdevice_notifier(struct notifier_block *nb);
a30c7b42
JP
2529int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2530int unregister_netdevice_notifier_net(struct net *net,
2531 struct notifier_block *nb);
351638e7
JP
2532
2533struct netdev_notifier_info {
51d0c047
DA
2534 struct net_device *dev;
2535 struct netlink_ext_ack *extack;
351638e7
JP
2536};
2537
af7d6cce
SD
2538struct netdev_notifier_info_ext {
2539 struct netdev_notifier_info info; /* must be first */
2540 union {
2541 u32 mtu;
2542 } ext;
2543};
2544
be9efd36
JP
2545struct netdev_notifier_change_info {
2546 struct netdev_notifier_info info; /* must be first */
2547 unsigned int flags_changed;
2548};
2549
0e4ead9d
JP
2550struct netdev_notifier_changeupper_info {
2551 struct netdev_notifier_info info; /* must be first */
2552 struct net_device *upper_dev; /* new upper dev */
2553 bool master; /* is upper dev master */
5e82b4b2 2554 bool linking; /* is the notification for link or unlink */
29bf24af 2555 void *upper_info; /* upper dev info */
0e4ead9d
JP
2556};
2557
04d48266
JP
2558struct netdev_notifier_changelowerstate_info {
2559 struct netdev_notifier_info info; /* must be first */
2560 void *lower_state_info; /* is lower dev state */
2561};
2562
1570415f
PM
2563struct netdev_notifier_pre_changeaddr_info {
2564 struct netdev_notifier_info info; /* must be first */
2565 const unsigned char *dev_addr;
2566};
2567
75538c2b
CW
2568static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2569 struct net_device *dev)
2570{
2571 info->dev = dev;
51d0c047 2572 info->extack = NULL;
75538c2b
CW
2573}
2574
351638e7
JP
2575static inline struct net_device *
2576netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2577{
2578 return info->dev;
2579}
2580
51d0c047
DA
2581static inline struct netlink_ext_ack *
2582netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2583{
2584 return info->extack;
2585}
2586
f629d208 2587int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2588
2589
1da177e4
LT
2590extern rwlock_t dev_base_lock; /* Device list lock */
2591
881d966b
EB
2592#define for_each_netdev(net, d) \
2593 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2594#define for_each_netdev_reverse(net, d) \
2595 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2596#define for_each_netdev_rcu(net, d) \
2597 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2598#define for_each_netdev_safe(net, d, n) \
2599 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2600#define for_each_netdev_continue(net, d) \
2601 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
afa0df59
JP
2602#define for_each_netdev_continue_reverse(net, d) \
2603 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2604 dev_list)
254245d2 2605#define for_each_netdev_continue_rcu(net, d) \
2606 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2607#define for_each_netdev_in_bond_rcu(bond, slave) \
2608 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2609 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2610#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2611
a050c33f
DL
2612static inline struct net_device *next_net_device(struct net_device *dev)
2613{
2614 struct list_head *lh;
2615 struct net *net;
2616
c346dca1 2617 net = dev_net(dev);
a050c33f
DL
2618 lh = dev->dev_list.next;
2619 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2620}
2621
ce81b76a
ED
2622static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2623{
2624 struct list_head *lh;
2625 struct net *net;
2626
2627 net = dev_net(dev);
ccf43438 2628 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2629 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2630}
2631
a050c33f
DL
2632static inline struct net_device *first_net_device(struct net *net)
2633{
2634 return list_empty(&net->dev_base_head) ? NULL :
2635 net_device_entry(net->dev_base_head.next);
2636}
7562f876 2637
ccf43438
ED
2638static inline struct net_device *first_net_device_rcu(struct net *net)
2639{
2640 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2641
2642 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2643}
2644
f629d208
JP
2645int netdev_boot_setup_check(struct net_device *dev);
2646unsigned long netdev_boot_base(const char *prefix, int unit);
2647struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2648 const char *hwaddr);
2649struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2650struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2651void dev_add_pack(struct packet_type *pt);
2652void dev_remove_pack(struct packet_type *pt);
2653void __dev_remove_pack(struct packet_type *pt);
2654void dev_add_offload(struct packet_offload *po);
2655void dev_remove_offload(struct packet_offload *po);
f629d208 2656
a54acb3a 2657int dev_get_iflink(const struct net_device *dev);
fc4099f1 2658int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2659struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2660 unsigned short mask);
f629d208
JP
2661struct net_device *dev_get_by_name(struct net *net, const char *name);
2662struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2663struct net_device *__dev_get_by_name(struct net *net, const char *name);
2664int dev_alloc_name(struct net_device *dev, const char *name);
00f54e68 2665int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
7051b88a 2666void dev_close(struct net_device *dev);
2667void dev_close_many(struct list_head *head, bool unlink);
f629d208 2668void dev_disable_lro(struct net_device *dev);
0c4b51f0 2669int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
a4ea8a3d 2670u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
a350ecce 2671 struct net_device *sb_dev);
a4ea8a3d 2672u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
a350ecce 2673 struct net_device *sb_dev);
2b4aa3ce 2674int dev_queue_xmit(struct sk_buff *skb);
eadec877 2675int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
865b03f2 2676int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
f629d208
JP
2677int register_netdevice(struct net_device *dev);
2678void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2679void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2680static inline void unregister_netdevice(struct net_device *dev)
2681{
2682 unregister_netdevice_queue(dev, NULL);
2683}
2684
f629d208
JP
2685int netdev_refcnt_read(const struct net_device *dev);
2686void free_netdev(struct net_device *dev);
74d332c1 2687void netdev_freemem(struct net_device *dev);
f629d208
JP
2688void synchronize_net(void);
2689int init_dummy_netdev(struct net_device *dev);
937f1ba5 2690
f629d208
JP
2691struct net_device *dev_get_by_index(struct net *net, int ifindex);
2692struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2693struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2694struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2695int netdev_get_name(struct net *net, char *name, int ifindex);
2696int dev_restart(struct net_device *dev);
d4546c25 2697int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
3a1296a3 2698int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
86911732
HX
2699
2700static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2701{
2702 return NAPI_GRO_CB(skb)->data_offset;
2703}
2704
2705static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2706{
2707 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2708}
2709
2710static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2711{
2712 NAPI_GRO_CB(skb)->data_offset += len;
2713}
2714
a5b1cf28
HX
2715static inline void *skb_gro_header_fast(struct sk_buff *skb,
2716 unsigned int offset)
86911732 2717{
a5b1cf28
HX
2718 return NAPI_GRO_CB(skb)->frag0 + offset;
2719}
78a478d0 2720
a5b1cf28
HX
2721static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2722{
2723 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2724}
78a478d0 2725
57ea52a8
HX
2726static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2727{
2728 NAPI_GRO_CB(skb)->frag0 = NULL;
2729 NAPI_GRO_CB(skb)->frag0_len = 0;
2730}
2731
a5b1cf28
HX
2732static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2733 unsigned int offset)
2734{
17dd759c
HX
2735 if (!pskb_may_pull(skb, hlen))
2736 return NULL;
2737
57ea52a8 2738 skb_gro_frag0_invalidate(skb);
17dd759c 2739 return skb->data + offset;
86911732 2740}
1da177e4 2741
36e7b1b8
HX
2742static inline void *skb_gro_network_header(struct sk_buff *skb)
2743{
78d3fd0b
HX
2744 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2745 skb_network_offset(skb);
36e7b1b8
HX
2746}
2747
bf5a755f
JC
2748static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2749 const void *start, unsigned int len)
2750{
573e8fca 2751 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2752 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2753 csum_partial(start, len, 0));
2754}
2755
573e8fca
TH
2756/* GRO checksum functions. These are logical equivalents of the normal
2757 * checksum functions (in skbuff.h) except that they operate on the GRO
2758 * offsets and fields in sk_buff.
2759 */
2760
2761__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2762
15e2396d
TH
2763static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2764{
b7fe10e5 2765 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2766}
2767
573e8fca
TH
2768static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2769 bool zero_okay,
2770 __sum16 check)
2771{
6edec0e6
TH
2772 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2773 skb_checksum_start_offset(skb) <
2774 skb_gro_offset(skb)) &&
15e2396d 2775 !skb_at_gro_remcsum_start(skb) &&
662880f4 2776 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2777 (!zero_okay || check));
2778}
2779
2780static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2781 __wsum psum)
2782{
2783 if (NAPI_GRO_CB(skb)->csum_valid &&
2784 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2785 return 0;
2786
2787 NAPI_GRO_CB(skb)->csum = psum;
2788
2789 return __skb_gro_checksum_complete(skb);
2790}
2791
573e8fca
TH
2792static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2793{
662880f4
TH
2794 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2795 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2796 NAPI_GRO_CB(skb)->csum_cnt--;
2797 } else {
2798 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2799 * verified a new top level checksum or an encapsulated one
2800 * during GRO. This saves work if we fallback to normal path.
2801 */
2802 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2803 }
2804}
2805
2806#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2807 compute_pseudo) \
2808({ \
2809 __sum16 __ret = 0; \
2810 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2811 __ret = __skb_gro_checksum_validate_complete(skb, \
2812 compute_pseudo(skb, proto)); \
219f1d79 2813 if (!__ret) \
573e8fca
TH
2814 skb_gro_incr_csum_unnecessary(skb); \
2815 __ret; \
2816})
2817
2818#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2819 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2820
2821#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2822 compute_pseudo) \
2823 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2824
2825#define skb_gro_checksum_simple_validate(skb) \
2826 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2827
d96535a1
TH
2828static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2829{
2830 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2831 !NAPI_GRO_CB(skb)->csum_valid);
2832}
2833
2834static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
b39c78b2 2835 __wsum pseudo)
d96535a1
TH
2836{
2837 NAPI_GRO_CB(skb)->csum = ~pseudo;
2838 NAPI_GRO_CB(skb)->csum_valid = 1;
2839}
2840
b39c78b2 2841#define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
2842do { \
2843 if (__skb_gro_checksum_convert_check(skb)) \
b39c78b2 2844 __skb_gro_checksum_convert(skb, \
d96535a1
TH
2845 compute_pseudo(skb, proto)); \
2846} while (0)
2847
26c4f7da
TH
2848struct gro_remcsum {
2849 int offset;
2850 __wsum delta;
2851};
2852
2853static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2854{
846cd667 2855 grc->offset = 0;
26c4f7da
TH
2856 grc->delta = 0;
2857}
2858
b7fe10e5
TH
2859static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2860 unsigned int off, size_t hdrlen,
2861 int start, int offset,
2862 struct gro_remcsum *grc,
2863 bool nopartial)
dcdc8994
TH
2864{
2865 __wsum delta;
b7fe10e5 2866 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2867
2868 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2869
15e2396d 2870 if (!nopartial) {
b7fe10e5
TH
2871 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2872 return ptr;
2873 }
2874
2875 ptr = skb_gro_header_fast(skb, off);
2876 if (skb_gro_header_hard(skb, off + plen)) {
2877 ptr = skb_gro_header_slow(skb, off + plen, off);
2878 if (!ptr)
2879 return NULL;
15e2396d
TH
2880 }
2881
b7fe10e5
TH
2882 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2883 start, offset);
dcdc8994
TH
2884
2885 /* Adjust skb->csum since we changed the packet */
dcdc8994 2886 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2887
b7fe10e5 2888 grc->offset = off + hdrlen + offset;
26c4f7da 2889 grc->delta = delta;
b7fe10e5
TH
2890
2891 return ptr;
dcdc8994
TH
2892}
2893
26c4f7da
TH
2894static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2895 struct gro_remcsum *grc)
2896{
b7fe10e5
TH
2897 void *ptr;
2898 size_t plen = grc->offset + sizeof(u16);
2899
26c4f7da
TH
2900 if (!grc->delta)
2901 return;
2902
b7fe10e5
TH
2903 ptr = skb_gro_header_fast(skb, grc->offset);
2904 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2905 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2906 if (!ptr)
2907 return;
2908 }
2909
2910 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2911}
dcdc8994 2912
25393d3f 2913#ifdef CONFIG_XFRM_OFFLOAD
d4546c25 2914static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
25393d3f
SK
2915{
2916 if (PTR_ERR(pp) != -EINPROGRESS)
2917 NAPI_GRO_CB(skb)->flush |= flush;
2918}
603d4cf8 2919static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
5cd3da4b 2920 struct sk_buff *pp,
603d4cf8
SD
2921 int flush,
2922 struct gro_remcsum *grc)
2923{
2924 if (PTR_ERR(pp) != -EINPROGRESS) {
2925 NAPI_GRO_CB(skb)->flush |= flush;
2926 skb_gro_remcsum_cleanup(skb, grc);
2927 skb->remcsum_offload = 0;
2928 }
2929}
25393d3f 2930#else
d4546c25 2931static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
5f114163
SK
2932{
2933 NAPI_GRO_CB(skb)->flush |= flush;
2934}
603d4cf8 2935static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
5cd3da4b 2936 struct sk_buff *pp,
603d4cf8
SD
2937 int flush,
2938 struct gro_remcsum *grc)
2939{
2940 NAPI_GRO_CB(skb)->flush |= flush;
2941 skb_gro_remcsum_cleanup(skb, grc);
2942 skb->remcsum_offload = 0;
2943}
25393d3f 2944#endif
5f114163 2945
0c4e8581
SH
2946static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2947 unsigned short type,
3b04ddde 2948 const void *daddr, const void *saddr,
95c96174 2949 unsigned int len)
0c4e8581 2950{
f1ecfd5d 2951 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2952 return 0;
3b04ddde
SH
2953
2954 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2955}
2956
b95cce35
SH
2957static inline int dev_parse_header(const struct sk_buff *skb,
2958 unsigned char *haddr)
2959{
2960 const struct net_device *dev = skb->dev;
2961
1b83336b 2962 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2963 return 0;
3b04ddde 2964 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2965}
2966
e78b2915
MM
2967static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2968{
2969 const struct net_device *dev = skb->dev;
2970
2971 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2972 return 0;
2973 return dev->header_ops->parse_protocol(skb);
2974}
2975
2793a23a
WB
2976/* ll_header must have at least hard_header_len allocated */
2977static inline bool dev_validate_header(const struct net_device *dev,
2978 char *ll_header, int len)
2979{
2980 if (likely(len >= dev->hard_header_len))
2981 return true;
217e6fa2
WB
2982 if (len < dev->min_header_len)
2983 return false;
2793a23a
WB
2984
2985 if (capable(CAP_SYS_RAWIO)) {
2986 memset(ll_header + len, 0, dev->hard_header_len - len);
2987 return true;
2988 }
2989
2990 if (dev->header_ops && dev->header_ops->validate)
2991 return dev->header_ops->validate(ll_header, len);
2992
2993 return false;
2994}
2995
36fd633e
AV
2996typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2997 int len, int size);
f629d208 2998int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2999static inline int unregister_gifconf(unsigned int family)
3000{
3001 return register_gifconf(family, NULL);
3002}
3003
99bbc707 3004#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 3005#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
3006struct sd_flow_limit {
3007 u64 count;
3008 unsigned int num_buckets;
3009 unsigned int history_head;
3010 u16 history[FLOW_LIMIT_HISTORY];
3011 u8 buckets[];
3012};
3013
3014extern int netdev_flow_limit_table_len;
3015#endif /* CONFIG_NET_FLOW_LIMIT */
3016
1da177e4 3017/*
5e82b4b2 3018 * Incoming packets are placed on per-CPU queues
1da177e4 3019 */
d94d9fee 3020struct softnet_data {
1da177e4 3021 struct list_head poll_list;
6e7676c1 3022 struct sk_buff_head process_queue;
1da177e4 3023
dee42870 3024 /* stats */
cd7b5396
DM
3025 unsigned int processed;
3026 unsigned int time_squeeze;
cd7b5396 3027 unsigned int received_rps;
fd793d89 3028#ifdef CONFIG_RPS
88751275 3029 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
3030#endif
3031#ifdef CONFIG_NET_FLOW_LIMIT
3032 struct sd_flow_limit __rcu *flow_limit;
3033#endif
3034 struct Qdisc *output_queue;
3035 struct Qdisc **output_queue_tailp;
3036 struct sk_buff *completion_queue;
f53c7239
SK
3037#ifdef CONFIG_XFRM_OFFLOAD
3038 struct sk_buff_head xfrm_backlog;
3039#endif
97cdcf37
FW
3040 /* written and read only by owning cpu: */
3041 struct {
3042 u16 recursion;
3043 u8 more;
3044 } xmit;
4cdb1e2e 3045#ifdef CONFIG_RPS
501e7ef5
ED
3046 /* input_queue_head should be written by cpu owning this struct,
3047 * and only read by other cpus. Worth using a cache line.
3048 */
3049 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3050
3051 /* Elements below can be accessed between CPUs for RPS/RFS */
966a9671 3052 call_single_data_t csd ____cacheline_aligned_in_smp;
88751275
ED
3053 struct softnet_data *rps_ipi_next;
3054 unsigned int cpu;
76cc8b13 3055 unsigned int input_queue_tail;
1e94d72f 3056#endif
95c96174 3057 unsigned int dropped;
0a9627f2 3058 struct sk_buff_head input_pkt_queue;
bea3348e 3059 struct napi_struct backlog;
99bbc707 3060
1da177e4
LT
3061};
3062
76cc8b13 3063static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
3064{
3065#ifdef CONFIG_RPS
76cc8b13
TH
3066 sd->input_queue_head++;
3067#endif
3068}
3069
3070static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3071 unsigned int *qtail)
3072{
3073#ifdef CONFIG_RPS
3074 *qtail = ++sd->input_queue_tail;
fec5e652
TH
3075#endif
3076}
3077
0a9627f2 3078DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 3079
97cdcf37
FW
3080static inline int dev_recursion_level(void)
3081{
28b05b92 3082 return this_cpu_read(softnet_data.xmit.recursion);
97cdcf37
FW
3083}
3084
3085#define XMIT_RECURSION_LIMIT 10
3086static inline bool dev_xmit_recursion(void)
3087{
3088 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3089 XMIT_RECURSION_LIMIT);
3090}
3091
3092static inline void dev_xmit_recursion_inc(void)
3093{
3094 __this_cpu_inc(softnet_data.xmit.recursion);
3095}
3096
3097static inline void dev_xmit_recursion_dec(void)
3098{
3099 __this_cpu_dec(softnet_data.xmit.recursion);
3100}
3101
f629d208 3102void __netif_schedule(struct Qdisc *q);
46e5da40 3103void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 3104
fd2ea0a7
DM
3105static inline void netif_tx_schedule_all(struct net_device *dev)
3106{
3107 unsigned int i;
3108
3109 for (i = 0; i < dev->num_tx_queues; i++)
3110 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3111}
3112
f9a7cbbf 3113static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 3114{
73466498 3115 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
3116}
3117
bea3348e
SH
3118/**
3119 * netif_start_queue - allow transmit
3120 * @dev: network device
3121 *
3122 * Allow upper layers to call the device hard_start_xmit routine.
3123 */
1da177e4
LT
3124static inline void netif_start_queue(struct net_device *dev)
3125{
e8a0464c 3126 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3127}
3128
fd2ea0a7
DM
3129static inline void netif_tx_start_all_queues(struct net_device *dev)
3130{
3131 unsigned int i;
3132
3133 for (i = 0; i < dev->num_tx_queues; i++) {
3134 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3135 netif_tx_start_queue(txq);
3136 }
3137}
3138
46e5da40 3139void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 3140
d29f749e
DJ
3141/**
3142 * netif_wake_queue - restart transmit
3143 * @dev: network device
3144 *
3145 * Allow upper layers to call the device hard_start_xmit routine.
3146 * Used for flow control when transmit resources are available.
3147 */
79d16385
DM
3148static inline void netif_wake_queue(struct net_device *dev)
3149{
e8a0464c 3150 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3151}
3152
fd2ea0a7
DM
3153static inline void netif_tx_wake_all_queues(struct net_device *dev)
3154{
3155 unsigned int i;
3156
3157 for (i = 0; i < dev->num_tx_queues; i++) {
3158 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3159 netif_tx_wake_queue(txq);
3160 }
3161}
3162
f9a7cbbf 3163static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 3164{
73466498 3165 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
3166}
3167
bea3348e
SH
3168/**
3169 * netif_stop_queue - stop transmitted packets
3170 * @dev: network device
3171 *
3172 * Stop upper layers calling the device hard_start_xmit routine.
3173 * Used for flow control when transmit resources are unavailable.
3174 */
1da177e4
LT
3175static inline void netif_stop_queue(struct net_device *dev)
3176{
e8a0464c 3177 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3178}
3179
a2029240 3180void netif_tx_stop_all_queues(struct net_device *dev);
ab92d68f 3181void netdev_update_lockdep_key(struct net_device *dev);
fd2ea0a7 3182
4d29515f 3183static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 3184{
73466498 3185 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
3186}
3187
bea3348e
SH
3188/**
3189 * netif_queue_stopped - test if transmit queue is flowblocked
3190 * @dev: network device
3191 *
3192 * Test if transmit queue on device is currently unable to send.
3193 */
4d29515f 3194static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 3195{
e8a0464c 3196 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3197}
3198
4d29515f 3199static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 3200{
73466498
TH
3201 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3202}
3203
8e2f1a63
DB
3204static inline bool
3205netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
3206{
3207 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3208}
3209
8e2f1a63
DB
3210static inline bool
3211netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3212{
3213 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3214}
3215
53511453
ED
3216/**
3217 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3218 * @dev_queue: pointer to transmit queue
3219 *
3220 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 3221 * to give appropriate hint to the CPU.
53511453
ED
3222 */
3223static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3224{
3225#ifdef CONFIG_BQL
3226 prefetchw(&dev_queue->dql.num_queued);
3227#endif
3228}
3229
3230/**
3231 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3232 * @dev_queue: pointer to transmit queue
3233 *
3234 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 3235 * to give appropriate hint to the CPU.
53511453
ED
3236 */
3237static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3238{
3239#ifdef CONFIG_BQL
3240 prefetchw(&dev_queue->dql.limit);
3241#endif
3242}
3243
c5d67bd7
TH
3244static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3245 unsigned int bytes)
3246{
114cf580
TH
3247#ifdef CONFIG_BQL
3248 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
3249
3250 if (likely(dql_avail(&dev_queue->dql) >= 0))
3251 return;
3252
3253 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3254
3255 /*
3256 * The XOFF flag must be set before checking the dql_avail below,
3257 * because in netdev_tx_completed_queue we update the dql_completed
3258 * before checking the XOFF flag.
3259 */
3260 smp_mb();
3261
3262 /* check again in case another CPU has just made room avail */
3263 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3264 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 3265#endif
c5d67bd7
TH
3266}
3267
3e59020a
ED
3268/* Variant of netdev_tx_sent_queue() for drivers that are aware
3269 * that they should not test BQL status themselves.
3270 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3271 * skb of a batch.
3272 * Returns true if the doorbell must be used to kick the NIC.
3273 */
3274static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3275 unsigned int bytes,
3276 bool xmit_more)
3277{
3278 if (xmit_more) {
3279#ifdef CONFIG_BQL
3280 dql_queued(&dev_queue->dql, bytes);
3281#endif
3282 return netif_tx_queue_stopped(dev_queue);
3283 }
3284 netdev_tx_sent_queue(dev_queue, bytes);
3285 return true;
3286}
3287
0042d0c8
FF
3288/**
3289 * netdev_sent_queue - report the number of bytes queued to hardware
3290 * @dev: network device
3291 * @bytes: number of bytes queued to the hardware device queue
3292 *
3293 * Report the number of bytes queued for sending/completion to the network
3294 * device hardware queue. @bytes should be a good approximation and should
3295 * exactly match netdev_completed_queue() @bytes
3296 */
c5d67bd7
TH
3297static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3298{
3299 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3300}
3301
620344c4
HK
3302static inline bool __netdev_sent_queue(struct net_device *dev,
3303 unsigned int bytes,
3304 bool xmit_more)
3305{
3306 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3307 xmit_more);
3308}
3309
c5d67bd7 3310static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 3311 unsigned int pkts, unsigned int bytes)
c5d67bd7 3312{
114cf580 3313#ifdef CONFIG_BQL
b37c0fbe
AD
3314 if (unlikely(!bytes))
3315 return;
3316
3317 dql_completed(&dev_queue->dql, bytes);
3318
3319 /*
3320 * Without the memory barrier there is a small possiblity that
3321 * netdev_tx_sent_queue will miss the update and cause the queue to
3322 * be stopped forever
3323 */
3324 smp_mb();
3325
f3acd33d 3326 if (unlikely(dql_avail(&dev_queue->dql) < 0))
b37c0fbe
AD
3327 return;
3328
3329 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3330 netif_schedule_queue(dev_queue);
114cf580 3331#endif
c5d67bd7
TH
3332}
3333
0042d0c8
FF
3334/**
3335 * netdev_completed_queue - report bytes and packets completed by device
3336 * @dev: network device
3337 * @pkts: actual number of packets sent over the medium
3338 * @bytes: actual number of bytes sent over the medium
3339 *
3340 * Report the number of bytes and packets transmitted by the network device
3341 * hardware queue over the physical medium, @bytes must exactly match the
3342 * @bytes amount passed to netdev_sent_queue()
3343 */
c5d67bd7 3344static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3345 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3346{
3347 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3348}
3349
3350static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3351{
114cf580 3352#ifdef CONFIG_BQL
5c490354 3353 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3354 dql_reset(&q->dql);
3355#endif
c5d67bd7
TH
3356}
3357
0042d0c8
FF
3358/**
3359 * netdev_reset_queue - reset the packets and bytes count of a network device
3360 * @dev_queue: network device
3361 *
3362 * Reset the bytes and packet count of a network device and clear the
3363 * software flow control OFF bit for this network device
3364 */
c5d67bd7
TH
3365static inline void netdev_reset_queue(struct net_device *dev_queue)
3366{
3367 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3368}
3369
b9507bda
DB
3370/**
3371 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3372 * @dev: network device
3373 * @queue_index: given tx queue index
3374 *
3375 * Returns 0 if given tx queue index >= number of device tx queues,
3376 * otherwise returns the originally passed tx queue index.
3377 */
3378static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3379{
3380 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3381 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3382 dev->name, queue_index,
3383 dev->real_num_tx_queues);
3384 return 0;
3385 }
3386
3387 return queue_index;
3388}
3389
bea3348e
SH
3390/**
3391 * netif_running - test if up
3392 * @dev: network device
3393 *
3394 * Test if the device has been brought up.
3395 */
4d29515f 3396static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3397{
3398 return test_bit(__LINK_STATE_START, &dev->state);
3399}
3400
f25f4e44 3401/*
5e82b4b2 3402 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3403 * stop, and a check if it's stopped. All other device management is
3404 * done at the overall netdevice level.
3405 * Also test the device if we're multiqueue.
3406 */
bea3348e
SH
3407
3408/**
3409 * netif_start_subqueue - allow sending packets on subqueue
3410 * @dev: network device
3411 * @queue_index: sub queue index
3412 *
3413 * Start individual transmit queue of a device with multiple transmit queues.
3414 */
f25f4e44
PWJ
3415static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3416{
fd2ea0a7 3417 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3418
3419 netif_tx_start_queue(txq);
f25f4e44
PWJ
3420}
3421
bea3348e
SH
3422/**
3423 * netif_stop_subqueue - stop sending packets on subqueue
3424 * @dev: network device
3425 * @queue_index: sub queue index
3426 *
3427 * Stop individual transmit queue of a device with multiple transmit queues.
3428 */
f25f4e44
PWJ
3429static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3430{
fd2ea0a7 3431 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3432 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3433}
3434
bea3348e
SH
3435/**
3436 * netif_subqueue_stopped - test status of subqueue
3437 * @dev: network device
3438 * @queue_index: sub queue index
3439 *
3440 * Check individual transmit queue of a device with multiple transmit queues.
3441 */
4d29515f
DM
3442static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3443 u16 queue_index)
f25f4e44 3444{
fd2ea0a7 3445 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3446
3447 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3448}
3449
4d29515f
DM
3450static inline bool netif_subqueue_stopped(const struct net_device *dev,
3451 struct sk_buff *skb)
668f895a
PE
3452{
3453 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3454}
bea3348e 3455
738b35cc
FF
3456/**
3457 * netif_wake_subqueue - allow sending packets on subqueue
3458 * @dev: network device
3459 * @queue_index: sub queue index
3460 *
3461 * Resume individual transmit queue of a device with multiple transmit queues.
3462 */
3463static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3464{
3465 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3466
3467 netif_tx_wake_queue(txq);
3468}
f25f4e44 3469
537c00de 3470#ifdef CONFIG_XPS
53af53ae 3471int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3472 u16 index);
80d19669
AN
3473int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3474 u16 index, bool is_rxqs_map);
3475
3476/**
3477 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3478 * @j: CPU/Rx queue index
3479 * @mask: bitmask of all cpus/rx queues
3480 * @nr_bits: number of bits in the bitmask
3481 *
3482 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3483 */
3484static inline bool netif_attr_test_mask(unsigned long j,
3485 const unsigned long *mask,
3486 unsigned int nr_bits)
3487{
3488 cpu_max_bits_warn(j, nr_bits);
3489 return test_bit(j, mask);
3490}
3491
3492/**
3493 * netif_attr_test_online - Test for online CPU/Rx queue
3494 * @j: CPU/Rx queue index
3495 * @online_mask: bitmask for CPUs/Rx queues that are online
3496 * @nr_bits: number of bits in the bitmask
3497 *
3498 * Returns true if a CPU/Rx queue is online.
3499 */
3500static inline bool netif_attr_test_online(unsigned long j,
3501 const unsigned long *online_mask,
3502 unsigned int nr_bits)
3503{
3504 cpu_max_bits_warn(j, nr_bits);
3505
3506 if (online_mask)
3507 return test_bit(j, online_mask);
3508
3509 return (j < nr_bits);
3510}
3511
3512/**
3513 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3514 * @n: CPU/Rx queue index
3515 * @srcp: the cpumask/Rx queue mask pointer
3516 * @nr_bits: number of bits in the bitmask
3517 *
3518 * Returns >= nr_bits if no further CPUs/Rx queues set.
3519 */
3520static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3521 unsigned int nr_bits)
3522{
3523 /* -1 is a legal arg here. */
3524 if (n != -1)
3525 cpu_max_bits_warn(n, nr_bits);
3526
3527 if (srcp)
3528 return find_next_bit(srcp, nr_bits, n + 1);
3529
3530 return n + 1;
3531}
3532
3533/**
3534 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3535 * @n: CPU/Rx queue index
3536 * @src1p: the first CPUs/Rx queues mask pointer
3537 * @src2p: the second CPUs/Rx queues mask pointer
3538 * @nr_bits: number of bits in the bitmask
3539 *
3540 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3541 */
3542static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3543 const unsigned long *src2p,
3544 unsigned int nr_bits)
3545{
3546 /* -1 is a legal arg here. */
3547 if (n != -1)
3548 cpu_max_bits_warn(n, nr_bits);
3549
3550 if (src1p && src2p)
3551 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3552 else if (src1p)
3553 return find_next_bit(src1p, nr_bits, n + 1);
3554 else if (src2p)
3555 return find_next_bit(src2p, nr_bits, n + 1);
3556
3557 return n + 1;
3558}
537c00de
AD
3559#else
3560static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3561 const struct cpumask *mask,
537c00de
AD
3562 u16 index)
3563{
3564 return 0;
3565}
c9fbb2d2
KK
3566
3567static inline int __netif_set_xps_queue(struct net_device *dev,
3568 const unsigned long *mask,
3569 u16 index, bool is_rxqs_map)
3570{
3571 return 0;
3572}
537c00de
AD
3573#endif
3574
bea3348e
SH
3575/**
3576 * netif_is_multiqueue - test if device has multiple transmit queues
3577 * @dev: network device
3578 *
3579 * Check if device has multiple transmit queues
bea3348e 3580 */
4d29515f 3581static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3582{
a02cec21 3583 return dev->num_tx_queues > 1;
f25f4e44 3584}
1da177e4 3585
f629d208 3586int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3587
a953be53 3588#ifdef CONFIG_SYSFS
f629d208 3589int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3590#else
3591static inline int netif_set_real_num_rx_queues(struct net_device *dev,
c29c2ebd 3592 unsigned int rxqs)
62fe0b40 3593{
c29c2ebd 3594 dev->real_num_rx_queues = rxqs;
62fe0b40
BH
3595 return 0;
3596}
3597#endif
3598
65073a67
DB
3599static inline struct netdev_rx_queue *
3600__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3601{
3602 return dev->_rx + rxq;
3603}
3604
a953be53
MD
3605#ifdef CONFIG_SYSFS
3606static inline unsigned int get_netdev_rx_queue_index(
3607 struct netdev_rx_queue *queue)
3608{
3609 struct net_device *dev = queue->dev;
3610 int index = queue - dev->_rx;
3611
3612 BUG_ON(index >= dev->num_rx_queues);
3613 return index;
3614}
3615#endif
3616
16917b87 3617#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3618int netif_get_num_default_rss_queues(void);
16917b87 3619
e6247027
ED
3620enum skb_free_reason {
3621 SKB_REASON_CONSUMED,
3622 SKB_REASON_DROPPED,
3623};
3624
3625void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3626void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3627
e6247027
ED
3628/*
3629 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3630 * interrupt context or with hardware interrupts being disabled.
3631 * (in_irq() || irqs_disabled())
3632 *
3633 * We provide four helpers that can be used in following contexts :
3634 *
3635 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3636 * replacing kfree_skb(skb)
3637 *
3638 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3639 * Typically used in place of consume_skb(skb) in TX completion path
3640 *
3641 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3642 * replacing kfree_skb(skb)
3643 *
3644 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3645 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3646 */
e6247027
ED
3647static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3648{
3649 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3650}
3651
3652static inline void dev_consume_skb_irq(struct sk_buff *skb)
3653{
3654 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3655}
3656
3657static inline void dev_kfree_skb_any(struct sk_buff *skb)
3658{
3659 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3660}
3661
3662static inline void dev_consume_skb_any(struct sk_buff *skb)
3663{
3664 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3665}
1da177e4 3666
7c497478
JW
3667void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3668int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
f629d208
JP
3669int netif_rx(struct sk_buff *skb);
3670int netif_rx_ni(struct sk_buff *skb);
04eb4489 3671int netif_receive_skb(struct sk_buff *skb);
1c601d82 3672int netif_receive_skb_core(struct sk_buff *skb);
f6ad8c1b 3673void netif_receive_skb_list(struct list_head *head);
f629d208
JP
3674gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3675void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3676struct sk_buff *napi_get_frags(struct napi_struct *napi);
3677gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3678struct packet_offload *gro_find_receive_by_type(__be16 type);
3679struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3680
3681static inline void napi_free_frags(struct napi_struct *napi)
3682{
3683 kfree_skb(napi->skb);
3684 napi->skb = NULL;
3685}
3686
24b27fc4 3687bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3688int netdev_rx_handler_register(struct net_device *dev,
3689 rx_handler_func_t *rx_handler,
3690 void *rx_handler_data);
3691void netdev_rx_handler_unregister(struct net_device *dev);
3692
3693bool dev_valid_name(const char *name);
44c02a2c
AV
3694int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3695 bool *need_copyout);
36fd633e 3696int dev_ifconf(struct net *net, struct ifconf *, int);
f629d208
JP
3697int dev_ethtool(struct net *net, struct ifreq *);
3698unsigned int dev_get_flags(const struct net_device *);
6d040321
PM
3699int __dev_change_flags(struct net_device *dev, unsigned int flags,
3700 struct netlink_ext_ack *extack);
567c5e13
PM
3701int dev_change_flags(struct net_device *dev, unsigned int flags,
3702 struct netlink_ext_ack *extack);
cb178190
DM
3703void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3704 unsigned int gchanges);
f629d208
JP
3705int dev_change_name(struct net_device *, const char *);
3706int dev_set_alias(struct net_device *, const char *, size_t);
6c557001 3707int dev_get_alias(const struct net_device *, char *, size_t);
f629d208 3708int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3709int __dev_set_mtu(struct net_device *, int);
d836f5c6
ED
3710int dev_validate_mtu(struct net_device *dev, int mtu,
3711 struct netlink_ext_ack *extack);
7a4c53be
SH
3712int dev_set_mtu_ext(struct net_device *dev, int mtu,
3713 struct netlink_ext_ack *extack);
f629d208 3714int dev_set_mtu(struct net_device *, int);
6a643ddb 3715int dev_change_tx_queue_len(struct net_device *, unsigned long);
f629d208 3716void dev_set_group(struct net_device *, int);
d59cdf94
PM
3717int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3718 struct netlink_ext_ack *extack);
3a37a963
PM
3719int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3720 struct netlink_ext_ack *extack);
f629d208
JP
3721int dev_change_carrier(struct net_device *, bool new_carrier);
3722int dev_get_phys_port_id(struct net_device *dev,
02637fce 3723 struct netdev_phys_item_id *ppid);
db24a904
DA
3724int dev_get_phys_port_name(struct net_device *dev,
3725 char *name, size_t len);
d6abc596
FF
3726int dev_get_port_parent_id(struct net_device *dev,
3727 struct netdev_phys_item_id *ppid, bool recurse);
3728bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
d746d707 3729int dev_change_proto_down(struct net_device *dev, bool proto_down);
b5899679 3730int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
f53c7239 3731struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
ce93718f
DM
3732struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3733 struct netdev_queue *txq, int *ret);
d67b9cd2 3734
f4e63525 3735typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
d67b9cd2
DB
3736int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3737 int fd, u32 flags);
a25717d2
JK
3738u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3739 enum bpf_netdev_command cmd);
84c6b868 3740int xdp_umem_query(struct net_device *dev, u16 queue_id);
d67b9cd2 3741
a0265d28 3742int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3743int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3744bool is_skb_forwardable(const struct net_device *dev,
3745 const struct sk_buff *skb);
1da177e4 3746
4e3264d2
MKL
3747static __always_inline int ____dev_forward_skb(struct net_device *dev,
3748 struct sk_buff *skb)
3749{
3750 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3751 unlikely(!is_skb_forwardable(dev, skb))) {
3752 atomic_long_inc(&dev->rx_dropped);
3753 kfree_skb(skb);
3754 return NET_RX_DROP;
3755 }
3756
3757 skb_scrub_packet(skb, true);
3758 skb->priority = 0;
3759 return 0;
3760}
3761
9f9a742d 3762bool dev_nit_active(struct net_device *dev);
74b20582
DA
3763void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3764
20380731 3765extern int netdev_budget;
7acf8a1e 3766extern unsigned int netdev_budget_usecs;
1da177e4
LT
3767
3768/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3769void netdev_run_todo(void);
1da177e4 3770
bea3348e
SH
3771/**
3772 * dev_put - release reference to device
3773 * @dev: network device
3774 *
9ef4429b 3775 * Release reference to device to allow it to be freed.
bea3348e 3776 */
1da177e4
LT
3777static inline void dev_put(struct net_device *dev)
3778{
933393f5 3779 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3780}
3781
bea3348e
SH
3782/**
3783 * dev_hold - get reference to device
3784 * @dev: network device
3785 *
9ef4429b 3786 * Hold reference to device to keep it from being freed.
bea3348e 3787 */
15333061
SH
3788static inline void dev_hold(struct net_device *dev)
3789{
933393f5 3790 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3791}
1da177e4
LT
3792
3793/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3794 * and _off may be called from IRQ context, but it is caller
3795 * who is responsible for serialization of these calls.
b00055aa
SR
3796 *
3797 * The name carrier is inappropriate, these functions should really be
3798 * called netif_lowerlayer_*() because they represent the state of any
3799 * kind of lower layer not just hardware media.
1da177e4
LT
3800 */
3801
f629d208
JP
3802void linkwatch_init_dev(struct net_device *dev);
3803void linkwatch_fire_event(struct net_device *dev);
3804void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3805
bea3348e
SH
3806/**
3807 * netif_carrier_ok - test if carrier present
3808 * @dev: network device
3809 *
3810 * Check if carrier is present on device
3811 */
4d29515f 3812static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3813{
3814 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3815}
3816
f629d208 3817unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3818
f629d208 3819void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3820
f629d208 3821void netif_carrier_on(struct net_device *dev);
1da177e4 3822
f629d208 3823void netif_carrier_off(struct net_device *dev);
1da177e4 3824
bea3348e
SH
3825/**
3826 * netif_dormant_on - mark device as dormant.
3827 * @dev: network device
3828 *
3829 * Mark device as dormant (as per RFC2863).
3830 *
3831 * The dormant state indicates that the relevant interface is not
3832 * actually in a condition to pass packets (i.e., it is not 'up') but is
3833 * in a "pending" state, waiting for some external event. For "on-
3834 * demand" interfaces, this new state identifies the situation where the
3835 * interface is waiting for events to place it in the up state.
bea3348e 3836 */
b00055aa
SR
3837static inline void netif_dormant_on(struct net_device *dev)
3838{
3839 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3840 linkwatch_fire_event(dev);
3841}
3842
bea3348e
SH
3843/**
3844 * netif_dormant_off - set device as not dormant.
3845 * @dev: network device
3846 *
3847 * Device is not in dormant state.
3848 */
b00055aa
SR
3849static inline void netif_dormant_off(struct net_device *dev)
3850{
3851 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3852 linkwatch_fire_event(dev);
3853}
3854
bea3348e 3855/**
8ecbc40a 3856 * netif_dormant - test if device is dormant
bea3348e
SH
3857 * @dev: network device
3858 *
8ecbc40a 3859 * Check if device is dormant.
bea3348e 3860 */
4d29515f 3861static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3862{
3863 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3864}
3865
3866
bea3348e
SH
3867/**
3868 * netif_oper_up - test if device is operational
3869 * @dev: network device
3870 *
3871 * Check if carrier is operational
3872 */
4d29515f 3873static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3874{
b00055aa
SR
3875 return (dev->operstate == IF_OPER_UP ||
3876 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3877}
3878
bea3348e
SH
3879/**
3880 * netif_device_present - is device available or removed
3881 * @dev: network device
3882 *
3883 * Check if device has not been removed from system.
3884 */
4d29515f 3885static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3886{
3887 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3888}
3889
f629d208 3890void netif_device_detach(struct net_device *dev);
1da177e4 3891
f629d208 3892void netif_device_attach(struct net_device *dev);
1da177e4
LT
3893
3894/*
3895 * Network interface message level settings
3896 */
1da177e4
LT
3897
3898enum {
3899 NETIF_MSG_DRV = 0x0001,
3900 NETIF_MSG_PROBE = 0x0002,
3901 NETIF_MSG_LINK = 0x0004,
3902 NETIF_MSG_TIMER = 0x0008,
3903 NETIF_MSG_IFDOWN = 0x0010,
3904 NETIF_MSG_IFUP = 0x0020,
3905 NETIF_MSG_RX_ERR = 0x0040,
3906 NETIF_MSG_TX_ERR = 0x0080,
3907 NETIF_MSG_TX_QUEUED = 0x0100,
3908 NETIF_MSG_INTR = 0x0200,
3909 NETIF_MSG_TX_DONE = 0x0400,
3910 NETIF_MSG_RX_STATUS = 0x0800,
3911 NETIF_MSG_PKTDATA = 0x1000,
3912 NETIF_MSG_HW = 0x2000,
3913 NETIF_MSG_WOL = 0x4000,
3914};
3915
3916#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3917#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3918#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3919#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3920#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3921#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3922#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3923#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3924#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3925#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3926#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3927#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3928#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3929#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3930#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3931
3932static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3933{
3934 /* use default */
3935 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3936 return default_msg_enable_bits;
3937 if (debug_value == 0) /* no output */
3938 return 0;
3939 /* set low N bits */
f4d7b3e2 3940 return (1U << debug_value) - 1;
1da177e4
LT
3941}
3942
c773e847 3943static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3944{
c773e847
DM
3945 spin_lock(&txq->_xmit_lock);
3946 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3947}
3948
5a717f4f
MT
3949static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3950{
3951 __acquire(&txq->_xmit_lock);
3952 return true;
3953}
3954
3955static inline void __netif_tx_release(struct netdev_queue *txq)
3956{
3957 __release(&txq->_xmit_lock);
3958}
3959
fd2ea0a7
DM
3960static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3961{
3962 spin_lock_bh(&txq->_xmit_lock);
3963 txq->xmit_lock_owner = smp_processor_id();
3964}
3965
4d29515f 3966static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3967{
4d29515f 3968 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3969 if (likely(ok))
3970 txq->xmit_lock_owner = smp_processor_id();
3971 return ok;
3972}
3973
3974static inline void __netif_tx_unlock(struct netdev_queue *txq)
3975{
3976 txq->xmit_lock_owner = -1;
3977 spin_unlock(&txq->_xmit_lock);
3978}
3979
3980static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3981{
3982 txq->xmit_lock_owner = -1;
3983 spin_unlock_bh(&txq->_xmit_lock);
3984}
3985
08baf561
ED
3986static inline void txq_trans_update(struct netdev_queue *txq)
3987{
3988 if (txq->xmit_lock_owner != -1)
3989 txq->trans_start = jiffies;
3990}
3991
ba162f8e
FW
3992/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3993static inline void netif_trans_update(struct net_device *dev)
3994{
9b36627a
FW
3995 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3996
3997 if (txq->trans_start != jiffies)
3998 txq->trans_start = jiffies;
ba162f8e
FW
3999}
4000
d29f749e
DJ
4001/**
4002 * netif_tx_lock - grab network device transmit lock
4003 * @dev: network device
d29f749e
DJ
4004 *
4005 * Get network device transmit lock
4006 */
22dd7495
JHS
4007static inline void netif_tx_lock(struct net_device *dev)
4008{
e8a0464c 4009 unsigned int i;
c3f26a26 4010 int cpu;
c773e847 4011
c3f26a26
DM
4012 spin_lock(&dev->tx_global_lock);
4013 cpu = smp_processor_id();
e8a0464c
DM
4014 for (i = 0; i < dev->num_tx_queues; i++) {
4015 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
4016
4017 /* We are the only thread of execution doing a
4018 * freeze, but we have to grab the _xmit_lock in
4019 * order to synchronize with threads which are in
4020 * the ->hard_start_xmit() handler and already
4021 * checked the frozen bit.
4022 */
e8a0464c 4023 __netif_tx_lock(txq, cpu);
c3f26a26
DM
4024 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4025 __netif_tx_unlock(txq);
e8a0464c 4026 }
932ff279
HX
4027}
4028
4029static inline void netif_tx_lock_bh(struct net_device *dev)
4030{
e8a0464c
DM
4031 local_bh_disable();
4032 netif_tx_lock(dev);
932ff279
HX
4033}
4034
932ff279
HX
4035static inline void netif_tx_unlock(struct net_device *dev)
4036{
e8a0464c
DM
4037 unsigned int i;
4038
4039 for (i = 0; i < dev->num_tx_queues; i++) {
4040 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 4041
c3f26a26
DM
4042 /* No need to grab the _xmit_lock here. If the
4043 * queue is not stopped for another reason, we
4044 * force a schedule.
4045 */
4046 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 4047 netif_schedule_queue(txq);
c3f26a26
DM
4048 }
4049 spin_unlock(&dev->tx_global_lock);
932ff279
HX
4050}
4051
4052static inline void netif_tx_unlock_bh(struct net_device *dev)
4053{
e8a0464c
DM
4054 netif_tx_unlock(dev);
4055 local_bh_enable();
932ff279
HX
4056}
4057
c773e847 4058#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 4059 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 4060 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
4061 } else { \
4062 __netif_tx_acquire(txq); \
22dd7495
JHS
4063 } \
4064}
4065
5efeac44
EB
4066#define HARD_TX_TRYLOCK(dev, txq) \
4067 (((dev->features & NETIF_F_LLTX) == 0) ? \
4068 __netif_tx_trylock(txq) : \
5a717f4f 4069 __netif_tx_acquire(txq))
5efeac44 4070
c773e847 4071#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 4072 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 4073 __netif_tx_unlock(txq); \
5a717f4f
MT
4074 } else { \
4075 __netif_tx_release(txq); \
22dd7495
JHS
4076 } \
4077}
4078
1da177e4
LT
4079static inline void netif_tx_disable(struct net_device *dev)
4080{
fd2ea0a7 4081 unsigned int i;
c3f26a26 4082 int cpu;
fd2ea0a7 4083
c3f26a26
DM
4084 local_bh_disable();
4085 cpu = smp_processor_id();
fd2ea0a7
DM
4086 for (i = 0; i < dev->num_tx_queues; i++) {
4087 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
4088
4089 __netif_tx_lock(txq, cpu);
fd2ea0a7 4090 netif_tx_stop_queue(txq);
c3f26a26 4091 __netif_tx_unlock(txq);
fd2ea0a7 4092 }
c3f26a26 4093 local_bh_enable();
1da177e4
LT
4094}
4095
e308a5d8
DM
4096static inline void netif_addr_lock(struct net_device *dev)
4097{
4098 spin_lock(&dev->addr_list_lock);
4099}
4100
4101static inline void netif_addr_lock_bh(struct net_device *dev)
4102{
4103 spin_lock_bh(&dev->addr_list_lock);
4104}
4105
4106static inline void netif_addr_unlock(struct net_device *dev)
4107{
4108 spin_unlock(&dev->addr_list_lock);
4109}
4110
4111static inline void netif_addr_unlock_bh(struct net_device *dev)
4112{
4113 spin_unlock_bh(&dev->addr_list_lock);
4114}
4115
f001fde5 4116/*
31278e71 4117 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
4118 * rcu_read_lock held.
4119 */
4120#define for_each_dev_addr(dev, ha) \
31278e71 4121 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 4122
1da177e4
LT
4123/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4124
f629d208 4125void ether_setup(struct net_device *dev);
1da177e4
LT
4126
4127/* Support for loadable net-drivers */
f629d208 4128struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 4129 unsigned char name_assign_type,
f629d208
JP
4130 void (*setup)(struct net_device *),
4131 unsigned int txqs, unsigned int rxqs);
c835a677
TG
4132#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4133 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 4134
c835a677
TG
4135#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4136 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4137 count)
36909ea4 4138
f629d208
JP
4139int register_netdev(struct net_device *dev);
4140void unregister_netdev(struct net_device *dev);
f001fde5 4141
22bedad3 4142/* General hardware address lists handling functions */
f629d208
JP
4143int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4144 struct netdev_hw_addr_list *from_list, int addr_len);
4145void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4146 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
4147int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4148 struct net_device *dev,
4149 int (*sync)(struct net_device *, const unsigned char *),
4150 int (*unsync)(struct net_device *,
4151 const unsigned char *));
e7946760
IK
4152int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4153 struct net_device *dev,
4154 int (*sync)(struct net_device *,
4155 const unsigned char *, int),
4156 int (*unsync)(struct net_device *,
4157 const unsigned char *, int));
4158void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4159 struct net_device *dev,
4160 int (*unsync)(struct net_device *,
4161 const unsigned char *, int));
670e5b8e
AD
4162void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4163 struct net_device *dev,
4164 int (*unsync)(struct net_device *,
4165 const unsigned char *));
f629d208 4166void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 4167
f001fde5 4168/* Functions used for device addresses handling */
f629d208
JP
4169int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4170 unsigned char addr_type);
4171int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4172 unsigned char addr_type);
f629d208
JP
4173void dev_addr_flush(struct net_device *dev);
4174int dev_addr_init(struct net_device *dev);
a748ee24
JP
4175
4176/* Functions used for unicast addresses handling */
f629d208
JP
4177int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4178int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4179int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4180int dev_uc_sync(struct net_device *to, struct net_device *from);
4181int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4182void dev_uc_unsync(struct net_device *to, struct net_device *from);
4183void dev_uc_flush(struct net_device *dev);
4184void dev_uc_init(struct net_device *dev);
f001fde5 4185
670e5b8e
AD
4186/**
4187 * __dev_uc_sync - Synchonize device's unicast list
4188 * @dev: device to sync
4189 * @sync: function to call if address should be added
4190 * @unsync: function to call if address should be removed
4191 *
4192 * Add newly added addresses to the interface, and release
4193 * addresses that have been deleted.
5e82b4b2 4194 */
670e5b8e
AD
4195static inline int __dev_uc_sync(struct net_device *dev,
4196 int (*sync)(struct net_device *,
4197 const unsigned char *),
4198 int (*unsync)(struct net_device *,
4199 const unsigned char *))
4200{
4201 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4202}
4203
4204/**
e793c0f7 4205 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
4206 * @dev: device to sync
4207 * @unsync: function to call if address should be removed
4208 *
4209 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 4210 */
670e5b8e
AD
4211static inline void __dev_uc_unsync(struct net_device *dev,
4212 int (*unsync)(struct net_device *,
4213 const unsigned char *))
4214{
4215 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4216}
4217
22bedad3 4218/* Functions used for multicast addresses handling */
f629d208
JP
4219int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4220int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4221int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4222int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4223int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4224int dev_mc_sync(struct net_device *to, struct net_device *from);
4225int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4226void dev_mc_unsync(struct net_device *to, struct net_device *from);
4227void dev_mc_flush(struct net_device *dev);
4228void dev_mc_init(struct net_device *dev);
f001fde5 4229
670e5b8e
AD
4230/**
4231 * __dev_mc_sync - Synchonize device's multicast list
4232 * @dev: device to sync
4233 * @sync: function to call if address should be added
4234 * @unsync: function to call if address should be removed
4235 *
4236 * Add newly added addresses to the interface, and release
4237 * addresses that have been deleted.
5e82b4b2 4238 */
670e5b8e
AD
4239static inline int __dev_mc_sync(struct net_device *dev,
4240 int (*sync)(struct net_device *,
4241 const unsigned char *),
4242 int (*unsync)(struct net_device *,
4243 const unsigned char *))
4244{
4245 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4246}
4247
4248/**
e793c0f7 4249 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
4250 * @dev: device to sync
4251 * @unsync: function to call if address should be removed
4252 *
4253 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 4254 */
670e5b8e
AD
4255static inline void __dev_mc_unsync(struct net_device *dev,
4256 int (*unsync)(struct net_device *,
4257 const unsigned char *))
4258{
4259 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4260}
4261
4417da66 4262/* Functions used for secondary unicast and multicast support */
f629d208
JP
4263void dev_set_rx_mode(struct net_device *dev);
4264void __dev_set_rx_mode(struct net_device *dev);
4265int dev_set_promiscuity(struct net_device *dev, int inc);
4266int dev_set_allmulti(struct net_device *dev, int inc);
4267void netdev_state_change(struct net_device *dev);
4268void netdev_notify_peers(struct net_device *dev);
4269void netdev_features_change(struct net_device *dev);
1da177e4 4270/* Load a device via the kmod */
f629d208
JP
4271void dev_load(struct net *net, const char *name);
4272struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4273 struct rtnl_link_stats64 *storage);
4274void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4275 const struct net_device_stats *netdev_stats);
eeda3fd6 4276
1da177e4 4277extern int netdev_max_backlog;
3b098e2d 4278extern int netdev_tstamp_prequeue;
1da177e4 4279extern int weight_p;
3d48b53f
MT
4280extern int dev_weight_rx_bias;
4281extern int dev_weight_tx_bias;
4282extern int dev_rx_weight;
4283extern int dev_tx_weight;
323ebb61 4284extern int gro_normal_batch;
9ff162a8 4285
f629d208 4286bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
4287struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4288 struct list_head **iter);
f629d208
JP
4289struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4290 struct list_head **iter);
8b5be856 4291
44a40855
VY
4292/* iterate through upper list, must be called under RCU read lock */
4293#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4294 for (iter = &(dev)->adj_list.upper, \
4295 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4296 updev; \
4297 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4298
1a3f060c
DA
4299int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4300 int (*fn)(struct net_device *upper_dev,
4301 void *data),
4302 void *data);
4303
4304bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4305 struct net_device *upper_dev);
4306
25cc72a3
IS
4307bool netdev_has_any_upper_dev(struct net_device *dev);
4308
f629d208
JP
4309void *netdev_lower_get_next_private(struct net_device *dev,
4310 struct list_head **iter);
4311void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4312 struct list_head **iter);
31088a11
VF
4313
4314#define netdev_for_each_lower_private(dev, priv, iter) \
4315 for (iter = (dev)->adj_list.lower.next, \
4316 priv = netdev_lower_get_next_private(dev, &(iter)); \
4317 priv; \
4318 priv = netdev_lower_get_next_private(dev, &(iter)))
4319
4320#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4321 for (iter = &(dev)->adj_list.lower, \
4322 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4323 priv; \
4324 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4325
4085ebe8
VY
4326void *netdev_lower_get_next(struct net_device *dev,
4327 struct list_head **iter);
7ce856aa 4328
4085ebe8 4329#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 4330 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
4331 ldev = netdev_lower_get_next(dev, &(iter)); \
4332 ldev; \
4333 ldev = netdev_lower_get_next(dev, &(iter)))
4334
7ce856aa
JP
4335struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4336 struct list_head **iter);
4337struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4338 struct list_head **iter);
4339
1a3f060c
DA
4340int netdev_walk_all_lower_dev(struct net_device *dev,
4341 int (*fn)(struct net_device *lower_dev,
4342 void *data),
4343 void *data);
4344int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4345 int (*fn)(struct net_device *lower_dev,
4346 void *data),
4347 void *data);
4348
f629d208 4349void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 4350void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
4351struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4352struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
42ab19ee
DA
4353int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4354 struct netlink_ext_ack *extack);
f629d208 4355int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 4356 struct net_device *upper_dev,
42ab19ee
DA
4357 void *upper_priv, void *upper_info,
4358 struct netlink_ext_ack *extack);
f629d208
JP
4359void netdev_upper_dev_unlink(struct net_device *dev,
4360 struct net_device *upper_dev);
32b6d34f
TY
4361int netdev_adjacent_change_prepare(struct net_device *old_dev,
4362 struct net_device *new_dev,
4363 struct net_device *dev,
4364 struct netlink_ext_ack *extack);
4365void netdev_adjacent_change_commit(struct net_device *old_dev,
4366 struct net_device *new_dev,
4367 struct net_device *dev);
4368void netdev_adjacent_change_abort(struct net_device *old_dev,
4369 struct net_device *new_dev,
4370 struct net_device *dev);
5bb025fa 4371void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
4372void *netdev_lower_dev_get_private(struct net_device *dev,
4373 struct net_device *lower_dev);
04d48266
JP
4374void netdev_lower_state_changed(struct net_device *lower_dev,
4375 void *lower_state_info);
960fb622
ED
4376
4377/* RSS keys are 40 or 52 bytes long */
4378#define NETDEV_RSS_KEY_LEN 52
ba905f5e 4379extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
4380void netdev_rss_key_fill(void *buffer, size_t len);
4381
f629d208 4382int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 4383int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
4384int skb_csum_hwoffload_help(struct sk_buff *skb,
4385 const netdev_features_t features);
4386
f629d208
JP
4387struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4388 netdev_features_t features, bool tx_path);
4389struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4390 netdev_features_t features);
12b0004d 4391
61bd3857
MS
4392struct netdev_bonding_info {
4393 ifslave slave;
4394 ifbond master;
4395};
4396
4397struct netdev_notifier_bonding_info {
4398 struct netdev_notifier_info info; /* must be first */
4399 struct netdev_bonding_info bonding_info;
4400};
4401
4402void netdev_bonding_info_change(struct net_device *dev,
4403 struct netdev_bonding_info *bonding_info);
4404
6b08d6c1
MK
4405#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4406void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4407#else
4408static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4409 const void *data)
4410{
4411}
4412#endif
4413
12b0004d
CW
4414static inline
4415struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4416{
4417 return __skb_gso_segment(skb, features, true);
4418}
53d6471c 4419__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
4420
4421static inline bool can_checksum_protocol(netdev_features_t features,
4422 __be16 protocol)
4423{
c8cd0989
TH
4424 if (protocol == htons(ETH_P_FCOE))
4425 return !!(features & NETIF_F_FCOE_CRC);
4426
4427 /* Assume this is an IP checksum (not SCTP CRC) */
4428
4429 if (features & NETIF_F_HW_CSUM) {
4430 /* Can checksum everything */
4431 return true;
4432 }
4433
4434 switch (protocol) {
4435 case htons(ETH_P_IP):
4436 return !!(features & NETIF_F_IP_CSUM);
4437 case htons(ETH_P_IPV6):
4438 return !!(features & NETIF_F_IPV6_CSUM);
4439 default:
4440 return false;
4441 }
ec5f0615 4442}
12b0004d 4443
fb286bb2 4444#ifdef CONFIG_BUG
7fe50ac8 4445void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
fb286bb2 4446#else
7fe50ac8
CW
4447static inline void netdev_rx_csum_fault(struct net_device *dev,
4448 struct sk_buff *skb)
fb286bb2
HX
4449{
4450}
4451#endif
1da177e4 4452/* rx skb timestamps */
f629d208
JP
4453void net_enable_timestamp(void);
4454void net_disable_timestamp(void);
1da177e4 4455
20380731 4456#ifdef CONFIG_PROC_FS
f629d208 4457int __init dev_proc_init(void);
900ff8c6
CW
4458#else
4459#define dev_proc_init() 0
20380731
ACM
4460#endif
4461
4798248e 4462static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4463 struct sk_buff *skb, struct net_device *dev,
4464 bool more)
4798248e 4465{
6b16f9ee 4466 __this_cpu_write(softnet_data.xmit.more, more);
0b725a2c 4467 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4468}
4469
97cdcf37
FW
4470static inline bool netdev_xmit_more(void)
4471{
4472 return __this_cpu_read(softnet_data.xmit.more);
4473}
4474
10b3ad8c 4475static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4476 struct netdev_queue *txq, bool more)
4798248e
DM
4477{
4478 const struct net_device_ops *ops = dev->netdev_ops;
2183435c 4479 netdev_tx_t rc;
4798248e 4480
fa2dbdc2 4481 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4482 if (rc == NETDEV_TX_OK)
4483 txq_trans_update(txq);
4484
4485 return rc;
4798248e
DM
4486}
4487
b793dc5c 4488int netdev_class_create_file_ns(const struct class_attribute *class_attr,
42a2d923 4489 const void *ns);
b793dc5c 4490void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
42a2d923 4491 const void *ns);
58292cbe 4492
b793dc5c 4493static inline int netdev_class_create_file(const struct class_attribute *class_attr)
58292cbe
TH
4494{
4495 return netdev_class_create_file_ns(class_attr, NULL);
4496}
4497
b793dc5c 4498static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
58292cbe
TH
4499{
4500 netdev_class_remove_file_ns(class_attr, NULL);
4501}
b8a9787e 4502
737aec57 4503extern const struct kobj_ns_type_operations net_ns_type_operations;
04600794 4504
f629d208 4505const char *netdev_drivername(const struct net_device *dev);
6579e57b 4506
f629d208 4507void linkwatch_run_queue(void);
20380731 4508
da08143b
MK
4509static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4510 netdev_features_t f2)
4511{
c8cd0989
TH
4512 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4513 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4514 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4515 else
b6a0e72a 4516 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4517 }
da08143b 4518
c8cd0989 4519 return f1 & f2;
da08143b
MK
4520}
4521
c8f44aff
MM
4522static inline netdev_features_t netdev_get_wanted_features(
4523 struct net_device *dev)
5455c699
MM
4524{
4525 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4526}
c8f44aff
MM
4527netdev_features_t netdev_increment_features(netdev_features_t all,
4528 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4529
4530/* Allow TSO being used on stacked device :
4531 * Performing the GSO segmentation before last device
4532 * is a performance improvement.
4533 */
4534static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4535 netdev_features_t mask)
4536{
4537 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4538}
4539
6cb6a27c 4540int __netdev_update_features(struct net_device *dev);
5455c699 4541void netdev_update_features(struct net_device *dev);
afe12cc8 4542void netdev_change_features(struct net_device *dev);
7f353bf2 4543
fc4a7489
PM
4544void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4545 struct net_device *dev);
4546
e38f3025
TM
4547netdev_features_t passthru_features_check(struct sk_buff *skb,
4548 struct net_device *dev,
4549 netdev_features_t features);
c1e756bf 4550netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4551
4d29515f 4552static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4553{
7b748340 4554 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4555
4556 /* check flags correspondence */
4557 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4558 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4559 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4560 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4561 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4562 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4563 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4564 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4565 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4566 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4567 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4568 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4569 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4570 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4571 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4572 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0c19f846 4573 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
83aa025f 4574 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
3b335832 4575 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
0345e186 4576
d6b4991a 4577 return (features & feature) == feature;
576a30eb
HX
4578}
4579
4d29515f 4580static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4581{
278b2513 4582 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4583 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4584}
4585
8b86a61d 4586static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4587 netdev_features_t features)
7967168c 4588{
fc741216 4589 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4590 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4591 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4592}
4593
82cc1a7a
PWJ
4594static inline void netif_set_gso_max_size(struct net_device *dev,
4595 unsigned int size)
4596{
4597 dev->gso_max_size = size;
4598}
4599
7a7ffbab
WCC
4600static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4601 int pulled_hlen, u16 mac_offset,
4602 int mac_len)
4603{
4604 skb->protocol = protocol;
4605 skb->encapsulation = 1;
4606 skb_push(skb, pulled_hlen);
4607 skb_reset_transport_header(skb);
4608 skb->mac_header = mac_offset;
4609 skb->network_header = skb->mac_header + mac_len;
4610 skb->mac_len = mac_len;
4611}
4612
3c175784
SD
4613static inline bool netif_is_macsec(const struct net_device *dev)
4614{
4615 return dev->priv_flags & IFF_MACSEC;
4616}
4617
b618aaa9 4618static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4619{
4620 return dev->priv_flags & IFF_MACVLAN;
4621}
4622
b618aaa9 4623static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4624{
4625 return dev->priv_flags & IFF_MACVLAN_PORT;
4626}
4627
b618aaa9 4628static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4629{
4630 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4631}
4632
b618aaa9 4633static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4634{
4635 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4636}
4637
3bdc0eba
BG
4638static inline bool netif_supports_nofcs(struct net_device *dev)
4639{
4640 return dev->priv_flags & IFF_SUPP_NOFCS;
4641}
4642
d5256083
DB
4643static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4644{
4645 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4646}
4647
007979ea 4648static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4649{
007979ea 4650 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4651}
4652
fee6d4c7
DA
4653static inline bool netif_is_l3_slave(const struct net_device *dev)
4654{
4655 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4656}
4657
0894ae3f
JP
4658static inline bool netif_is_bridge_master(const struct net_device *dev)
4659{
4660 return dev->priv_flags & IFF_EBRIDGE;
4661}
4662
28f9ee22
VY
4663static inline bool netif_is_bridge_port(const struct net_device *dev)
4664{
4665 return dev->priv_flags & IFF_BRIDGE_PORT;
4666}
4667
35d4e172
JP
4668static inline bool netif_is_ovs_master(const struct net_device *dev)
4669{
4670 return dev->priv_flags & IFF_OPENVSWITCH;
4671}
4672
5be66141
JP
4673static inline bool netif_is_ovs_port(const struct net_device *dev)
4674{
4675 return dev->priv_flags & IFF_OVS_DATAPATH;
4676}
4677
b618aaa9 4678static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4679{
4680 return dev->priv_flags & IFF_TEAM;
4681}
4682
b618aaa9 4683static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4684{
4685 return dev->priv_flags & IFF_TEAM_PORT;
4686}
4687
b618aaa9 4688static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4689{
4690 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4691}
4692
b618aaa9 4693static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4694{
4695 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4696}
4697
d4ab4286
KJ
4698static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4699{
4700 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4701}
4702
30c8bd5a
SS
4703static inline bool netif_is_failover(const struct net_device *dev)
4704{
4705 return dev->priv_flags & IFF_FAILOVER;
4706}
4707
4708static inline bool netif_is_failover_slave(const struct net_device *dev)
4709{
4710 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4711}
4712
02875878
ED
4713/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4714static inline void netif_keep_dst(struct net_device *dev)
4715{
4716 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4717}
4718
18d3df3e
PA
4719/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4720static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4721{
4722 /* TODO: reserve and use an additional IFF bit, if we get more users */
4723 return dev->priv_flags & IFF_MACSEC;
4724}
4725
505d4f73 4726extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4727
571ba423
JP
4728/* Logging, debugging and troubleshooting/diagnostic helpers. */
4729
4730/* netdev_printk helpers, similar to dev_printk */
4731
4732static inline const char *netdev_name(const struct net_device *dev)
4733{
c6f854d5
VF
4734 if (!dev->name[0] || strchr(dev->name, '%'))
4735 return "(unnamed net_device)";
571ba423
JP
4736 return dev->name;
4737}
4738
8397ed36
DA
4739static inline bool netdev_unregistering(const struct net_device *dev)
4740{
4741 return dev->reg_state == NETREG_UNREGISTERING;
4742}
4743
ccc7f496
VF
4744static inline const char *netdev_reg_state(const struct net_device *dev)
4745{
4746 switch (dev->reg_state) {
4747 case NETREG_UNINITIALIZED: return " (uninitialized)";
4748 case NETREG_REGISTERED: return "";
4749 case NETREG_UNREGISTERING: return " (unregistering)";
4750 case NETREG_UNREGISTERED: return " (unregistered)";
4751 case NETREG_RELEASED: return " (released)";
4752 case NETREG_DUMMY: return " (dummy)";
4753 }
4754
4755 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4756 return " (unknown)";
4757}
4758
ce3fdb69 4759__printf(3, 4) __cold
6ea754eb
JP
4760void netdev_printk(const char *level, const struct net_device *dev,
4761 const char *format, ...);
ce3fdb69 4762__printf(2, 3) __cold
6ea754eb 4763void netdev_emerg(const struct net_device *dev, const char *format, ...);
ce3fdb69 4764__printf(2, 3) __cold
6ea754eb 4765void netdev_alert(const struct net_device *dev, const char *format, ...);
ce3fdb69 4766__printf(2, 3) __cold
6ea754eb 4767void netdev_crit(const struct net_device *dev, const char *format, ...);
ce3fdb69 4768__printf(2, 3) __cold
6ea754eb 4769void netdev_err(const struct net_device *dev, const char *format, ...);
ce3fdb69 4770__printf(2, 3) __cold
6ea754eb 4771void netdev_warn(const struct net_device *dev, const char *format, ...);
ce3fdb69 4772__printf(2, 3) __cold
6ea754eb 4773void netdev_notice(const struct net_device *dev, const char *format, ...);
ce3fdb69 4774__printf(2, 3) __cold
6ea754eb 4775void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4776
375ef2b1
GP
4777#define netdev_level_once(level, dev, fmt, ...) \
4778do { \
4779 static bool __print_once __read_mostly; \
4780 \
4781 if (!__print_once) { \
4782 __print_once = true; \
4783 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4784 } \
4785} while (0)
4786
4787#define netdev_emerg_once(dev, fmt, ...) \
4788 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4789#define netdev_alert_once(dev, fmt, ...) \
4790 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4791#define netdev_crit_once(dev, fmt, ...) \
4792 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4793#define netdev_err_once(dev, fmt, ...) \
4794 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4795#define netdev_warn_once(dev, fmt, ...) \
4796 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4797#define netdev_notice_once(dev, fmt, ...) \
4798 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4799#define netdev_info_once(dev, fmt, ...) \
4800 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4801
8909c9ad
VK
4802#define MODULE_ALIAS_NETDEV(device) \
4803 MODULE_ALIAS("netdev-" device)
4804
b558c96f 4805#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4806#define netdev_dbg(__dev, format, args...) \
4807do { \
ffa10cb4 4808 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4809} while (0)
b558c96f
JC
4810#elif defined(DEBUG)
4811#define netdev_dbg(__dev, format, args...) \
4812 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4813#else
4814#define netdev_dbg(__dev, format, args...) \
4815({ \
4816 if (0) \
4817 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4818})
4819#endif
4820
4821#if defined(VERBOSE_DEBUG)
4822#define netdev_vdbg netdev_dbg
4823#else
4824
4825#define netdev_vdbg(dev, format, args...) \
4826({ \
4827 if (0) \
4828 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4829 0; \
4830})
4831#endif
4832
4833/*
4834 * netdev_WARN() acts like dev_printk(), but with the key difference
4835 * of using a WARN/WARN_ON to get the message out, including the
4836 * file/line information and a backtrace.
4837 */
4838#define netdev_WARN(dev, format, args...) \
e1cfe3d0 4839 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
ccc7f496 4840 netdev_reg_state(dev), ##args)
571ba423 4841
72dd831e 4842#define netdev_WARN_ONCE(dev, format, args...) \
e1cfe3d0 4843 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
375ef2b1
GP
4844 netdev_reg_state(dev), ##args)
4845
b3d95c5c
JP
4846/* netif printk helpers, similar to netdev_printk */
4847
4848#define netif_printk(priv, type, level, dev, fmt, args...) \
4849do { \
4850 if (netif_msg_##type(priv)) \
4851 netdev_printk(level, (dev), fmt, ##args); \
4852} while (0)
4853
f45f4321
JP
4854#define netif_level(level, priv, type, dev, fmt, args...) \
4855do { \
4856 if (netif_msg_##type(priv)) \
4857 netdev_##level(dev, fmt, ##args); \
4858} while (0)
4859
b3d95c5c 4860#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4861 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4862#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4863 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4864#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4865 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4866#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4867 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4868#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4869 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4870#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4871 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4872#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4873 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4874
0053ea9c 4875#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4876#define netif_dbg(priv, type, netdev, format, args...) \
4877do { \
4878 if (netif_msg_##type(priv)) \
b5fb0a03 4879 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4880} while (0)
0053ea9c
JP
4881#elif defined(DEBUG)
4882#define netif_dbg(priv, type, dev, format, args...) \
4883 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4884#else
4885#define netif_dbg(priv, type, dev, format, args...) \
4886({ \
4887 if (0) \
4888 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4889 0; \
4890})
4891#endif
4892
f617f276
EC
4893/* if @cond then downgrade to debug, else print at @level */
4894#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4895 do { \
4896 if (cond) \
4897 netif_dbg(priv, type, netdev, fmt, ##args); \
4898 else \
4899 netif_ ## level(priv, type, netdev, fmt, ##args); \
4900 } while (0)
4901
b3d95c5c 4902#if defined(VERBOSE_DEBUG)
bcfcc450 4903#define netif_vdbg netif_dbg
b3d95c5c
JP
4904#else
4905#define netif_vdbg(priv, type, dev, format, args...) \
4906({ \
4907 if (0) \
a4ed89cb 4908 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4909 0; \
4910})
4911#endif
571ba423 4912
900ff8c6
CW
4913/*
4914 * The list of packet types we will receive (as opposed to discard)
4915 * and the routines to invoke.
4916 *
4917 * Why 16. Because with 16 the only overlap we get on a hash of the
4918 * low nibble of the protocol value is RARP/SNAP/X.25.
4919 *
900ff8c6 4920 * 0800 IP
900ff8c6
CW
4921 * 0001 802.3
4922 * 0002 AX.25
4923 * 0004 802.2
4924 * 8035 RARP
4925 * 0005 SNAP
4926 * 0805 X.25
4927 * 0806 ARP
4928 * 8137 IPX
4929 * 0009 Localtalk
4930 * 86DD IPv6
4931 */
4932#define PTYPE_HASH_SIZE (16)
4933#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4934
4de83b88
MB
4935extern struct net_device *blackhole_netdev;
4936
385a154c 4937#endif /* _LINUX_NETDEVICE_H */