]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - include/linux/netdevice.h
vrf: add CRC32c offload to device features
[mirror_ubuntu-hirsute-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
bea3348e 38#include <linux/workqueue.h>
114cf580 39#include <linux/dynamic_queue_limits.h>
1da177e4 40
b1b67dd4 41#include <linux/ethtool.h>
a050c33f 42#include <net/net_namespace.h>
7a6b6f51 43#ifdef CONFIG_DCB
2f90b865
AD
44#include <net/dcbnl.h>
45#endif
5bc1421e 46#include <net/netprio_cgroup.h>
e817f856 47#include <net/xdp.h>
a050c33f 48
a59e2ecb 49#include <linux/netdev_features.h>
77162022 50#include <linux/neighbour.h>
607ca46e 51#include <uapi/linux/netdevice.h>
61bd3857 52#include <uapi/linux/if_bonding.h>
e4c6734e 53#include <uapi/linux/pkt_cls.h>
59cc1f61 54#include <linux/hashtable.h>
a59e2ecb 55
115c1d6e 56struct netpoll_info;
313162d0 57struct device;
c1f19b51 58struct phy_device;
2f657a60 59struct dsa_port;
c6e970a0 60
e679c9c1 61struct sfp_bus;
704232c2
JB
62/* 802.11 specific */
63struct wireless_dev;
98a18b6f
AA
64/* 802.15.4 specific */
65struct wpan_dev;
03c57747 66struct mpls_dev;
7c46a640
AD
67/* UDP Tunnel offloads */
68struct udp_tunnel_info;
a7862b45 69struct bpf_prog;
814abfab 70struct xdp_buff;
1da177e4 71
f629d208
JP
72void netdev_set_default_ethtool_ops(struct net_device *dev,
73 const struct ethtool_ops *ops);
d07d7507 74
9a1654ba
JP
75/* Backlog congestion levels */
76#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
77#define NET_RX_DROP 1 /* packet dropped */
78
572a9d7b
PM
79/*
80 * Transmit return codes: transmit return codes originate from three different
81 * namespaces:
82 *
83 * - qdisc return codes
84 * - driver transmit return codes
85 * - errno values
86 *
87 * Drivers are allowed to return any one of those in their hard_start_xmit()
88 * function. Real network devices commonly used with qdiscs should only return
89 * the driver transmit return codes though - when qdiscs are used, the actual
90 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
91 * higher layers. Virtual network devices transmit synchronously; in this case
92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
93 * others are propagated to higher layers.
94 */
95
96/* qdisc ->enqueue() return codes. */
97#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
98#define NET_XMIT_DROP 0x01 /* skb dropped */
99#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 100#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 101
b9df3cb8
GR
102/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
103 * indicates that the device will soon be dropping packets, or already drops
104 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 105#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
106#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
107
dc1f8bf6 108/* Driver transmit return codes */
9a1654ba 109#define NETDEV_TX_MASK 0xf0
572a9d7b 110
dc1f8bf6 111enum netdev_tx {
572a9d7b 112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
113 NETDEV_TX_OK = 0x00, /* driver took care of packet */
114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
115};
116typedef enum netdev_tx netdev_tx_t;
117
9a1654ba
JP
118/*
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121 */
122static inline bool dev_xmit_complete(int rc)
123{
124 /*
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 */
130 if (likely(rc < NET_XMIT_MASK))
131 return true;
132
133 return false;
134}
135
1da177e4 136/*
5e82b4b2 137 * Compute the worst-case header length according to the protocols
1da177e4
LT
138 * used.
139 */
fe2918b0 140
c0eb4540
KS
141#if defined(CONFIG_HYPERV_NET)
142# define LL_MAX_HEADER 128
143#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
144# if defined(CONFIG_MAC80211_MESH)
145# define LL_MAX_HEADER 128
146# else
147# define LL_MAX_HEADER 96
148# endif
1da177e4 149#else
8388e3da 150# define LL_MAX_HEADER 32
1da177e4
LT
151#endif
152
d11ead75
BH
153#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
155#define MAX_HEADER LL_MAX_HEADER
156#else
157#define MAX_HEADER (LL_MAX_HEADER + 48)
158#endif
159
160/*
be1f3c2c
BH
161 * Old network device statistics. Fields are native words
162 * (unsigned long) so they can be read and written atomically.
1da177e4 163 */
fe2918b0 164
d94d9fee 165struct net_device_stats {
3cfde79c
BH
166 unsigned long rx_packets;
167 unsigned long tx_packets;
168 unsigned long rx_bytes;
169 unsigned long tx_bytes;
170 unsigned long rx_errors;
171 unsigned long tx_errors;
172 unsigned long rx_dropped;
173 unsigned long tx_dropped;
174 unsigned long multicast;
1da177e4 175 unsigned long collisions;
1da177e4 176 unsigned long rx_length_errors;
3cfde79c
BH
177 unsigned long rx_over_errors;
178 unsigned long rx_crc_errors;
179 unsigned long rx_frame_errors;
180 unsigned long rx_fifo_errors;
181 unsigned long rx_missed_errors;
1da177e4
LT
182 unsigned long tx_aborted_errors;
183 unsigned long tx_carrier_errors;
184 unsigned long tx_fifo_errors;
185 unsigned long tx_heartbeat_errors;
186 unsigned long tx_window_errors;
1da177e4
LT
187 unsigned long rx_compressed;
188 unsigned long tx_compressed;
189};
190
1da177e4
LT
191
192#include <linux/cache.h>
193#include <linux/skbuff.h>
194
adc9300e 195#ifdef CONFIG_RPS
c5905afb
IM
196#include <linux/static_key.h>
197extern struct static_key rps_needed;
13bfff25 198extern struct static_key rfs_needed;
adc9300e
ED
199#endif
200
1da177e4
LT
201struct neighbour;
202struct neigh_parms;
203struct sk_buff;
204
f001fde5
JP
205struct netdev_hw_addr {
206 struct list_head list;
207 unsigned char addr[MAX_ADDR_LEN];
208 unsigned char type;
ccffad25
JP
209#define NETDEV_HW_ADDR_T_LAN 1
210#define NETDEV_HW_ADDR_T_SAN 2
211#define NETDEV_HW_ADDR_T_SLAVE 3
212#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 213#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 214 bool global_use;
4cd729b0 215 int sync_cnt;
8f8f103d 216 int refcount;
4543fbef 217 int synced;
f001fde5
JP
218 struct rcu_head rcu_head;
219};
220
31278e71
JP
221struct netdev_hw_addr_list {
222 struct list_head list;
223 int count;
224};
225
22bedad3
JP
226#define netdev_hw_addr_list_count(l) ((l)->count)
227#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
228#define netdev_hw_addr_list_for_each(ha, l) \
229 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 230
22bedad3
JP
231#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
232#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
233#define netdev_for_each_uc_addr(ha, dev) \
234 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 235
22bedad3
JP
236#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
237#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 238#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 239 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 240
d94d9fee 241struct hh_cache {
5b3dc2f3 242 unsigned int hh_len;
3644f0ce 243 seqlock_t hh_lock;
1da177e4
LT
244
245 /* cached hardware header; allow for machine alignment needs. */
246#define HH_DATA_MOD 16
247#define HH_DATA_OFF(__len) \
5ba0eac6 248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
249#define HH_DATA_ALIGN(__len) \
250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
252};
253
5e82b4b2 254/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
255 * Alternative is:
256 * dev->hard_header_len ? (dev->hard_header_len +
257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
258 *
259 * We could use other alignment values, but we must maintain the
260 * relationship HH alignment <= LL alignment.
261 */
262#define LL_RESERVED_SPACE(dev) \
f5184d26 263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 264#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 266
3b04ddde
SH
267struct header_ops {
268 int (*create) (struct sk_buff *skb, struct net_device *dev,
269 unsigned short type, const void *daddr,
95c96174 270 const void *saddr, unsigned int len);
3b04ddde 271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
273 void (*cache_update)(struct hh_cache *hh,
274 const struct net_device *dev,
275 const unsigned char *haddr);
2793a23a 276 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
277};
278
1da177e4 279/* These flag bits are private to the generic network queueing
5e82b4b2 280 * layer; they may not be explicitly referenced by any other
1da177e4
LT
281 * code.
282 */
283
d94d9fee 284enum netdev_state_t {
1da177e4
LT
285 __LINK_STATE_START,
286 __LINK_STATE_PRESENT,
1da177e4 287 __LINK_STATE_NOCARRIER,
b00055aa
SR
288 __LINK_STATE_LINKWATCH_PENDING,
289 __LINK_STATE_DORMANT,
1da177e4
LT
290};
291
292
293/*
5e82b4b2 294 * This structure holds boot-time configured netdevice settings. They
fe2918b0 295 * are then used in the device probing.
1da177e4
LT
296 */
297struct netdev_boot_setup {
298 char name[IFNAMSIZ];
299 struct ifmap map;
300};
301#define NETDEV_BOOT_SETUP_MAX 8
302
f629d208 303int __init netdev_boot_setup(char *str);
1da177e4 304
bea3348e
SH
305/*
306 * Structure for NAPI scheduling similar to tasklet but with weighting
307 */
308struct napi_struct {
309 /* The poll_list must only be managed by the entity which
310 * changes the state of the NAPI_STATE_SCHED bit. This means
311 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 312 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
313 * can remove from the list right before clearing the bit.
314 */
315 struct list_head poll_list;
316
317 unsigned long state;
318 int weight;
404f7c9e 319 unsigned int gro_count;
bea3348e
SH
320 int (*poll)(struct napi_struct *, int);
321#ifdef CONFIG_NETPOLL
bea3348e 322 int poll_owner;
bea3348e 323#endif
5d38a079 324 struct net_device *dev;
d565b0a1 325 struct sk_buff *gro_list;
5d38a079 326 struct sk_buff *skb;
3b47d303 327 struct hrtimer timer;
404f7c9e 328 struct list_head dev_list;
af12fa6e
ET
329 struct hlist_node napi_hash_node;
330 unsigned int napi_id;
bea3348e
SH
331};
332
d94d9fee 333enum {
bea3348e 334 NAPI_STATE_SCHED, /* Poll is scheduled */
39e6c820 335 NAPI_STATE_MISSED, /* reschedule a napi */
a0a46196 336 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 337 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
338 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
339 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
340 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
341};
342
343enum {
39e6c820
ED
344 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
345 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
346 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
347 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
348 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
349 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
350 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
351};
352
5b252f0c 353enum gro_result {
d1c76af9
HX
354 GRO_MERGED,
355 GRO_MERGED_FREE,
356 GRO_HELD,
357 GRO_NORMAL,
358 GRO_DROP,
25393d3f 359 GRO_CONSUMED,
d1c76af9 360};
5b252f0c 361typedef enum gro_result gro_result_t;
d1c76af9 362
8a4eb573
JP
363/*
364 * enum rx_handler_result - Possible return values for rx_handlers.
365 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
366 * further.
367 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
368 * case skb->dev was changed by rx_handler.
369 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 370 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
371 *
372 * rx_handlers are functions called from inside __netif_receive_skb(), to do
373 * special processing of the skb, prior to delivery to protocol handlers.
374 *
375 * Currently, a net_device can only have a single rx_handler registered. Trying
376 * to register a second rx_handler will return -EBUSY.
377 *
378 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
379 * To unregister a rx_handler on a net_device, use
380 * netdev_rx_handler_unregister().
381 *
382 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
383 * do with the skb.
384 *
5e82b4b2 385 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 386 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 387 * the skb to be delivered in some other way.
8a4eb573
JP
388 *
389 * If the rx_handler changed skb->dev, to divert the skb to another
390 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
391 * new device will be called if it exists.
392 *
5e82b4b2 393 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 394 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 395 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 396 *
5e82b4b2 397 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
398 * delivered, it should return RX_HANDLER_PASS.
399 *
400 * A device without a registered rx_handler will behave as if rx_handler
401 * returned RX_HANDLER_PASS.
402 */
403
404enum rx_handler_result {
405 RX_HANDLER_CONSUMED,
406 RX_HANDLER_ANOTHER,
407 RX_HANDLER_EXACT,
408 RX_HANDLER_PASS,
409};
410typedef enum rx_handler_result rx_handler_result_t;
411typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 412
f629d208 413void __napi_schedule(struct napi_struct *n);
bc9ad166 414void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 415
4d29515f 416static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
417{
418 return test_bit(NAPI_STATE_DISABLE, &n->state);
419}
420
39e6c820 421bool napi_schedule_prep(struct napi_struct *n);
bea3348e
SH
422
423/**
424 * napi_schedule - schedule NAPI poll
5e82b4b2 425 * @n: NAPI context
bea3348e
SH
426 *
427 * Schedule NAPI poll routine to be called if it is not already
428 * running.
429 */
430static inline void napi_schedule(struct napi_struct *n)
431{
432 if (napi_schedule_prep(n))
433 __napi_schedule(n);
434}
435
bc9ad166
ED
436/**
437 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 438 * @n: NAPI context
bc9ad166
ED
439 *
440 * Variant of napi_schedule(), assuming hard irqs are masked.
441 */
442static inline void napi_schedule_irqoff(struct napi_struct *n)
443{
444 if (napi_schedule_prep(n))
445 __napi_schedule_irqoff(n);
446}
447
bfe13f54 448/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 449static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
450{
451 if (napi_schedule_prep(napi)) {
452 __napi_schedule(napi);
4d29515f 453 return true;
bfe13f54 454 }
4d29515f 455 return false;
bfe13f54
RD
456}
457
364b6055 458bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
459/**
460 * napi_complete - NAPI processing complete
5e82b4b2 461 * @n: NAPI context
bea3348e
SH
462 *
463 * Mark NAPI processing as complete.
3b47d303 464 * Consider using napi_complete_done() instead.
364b6055 465 * Return false if device should avoid rearming interrupts.
bea3348e 466 */
364b6055 467static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
468{
469 return napi_complete_done(n, 0);
470}
bea3348e 471
af12fa6e
ET
472/**
473 * napi_hash_del - remove a NAPI from global table
5e82b4b2 474 * @napi: NAPI context
af12fa6e 475 *
5e82b4b2 476 * Warning: caller must observe RCU grace period
34cbe27e
ED
477 * before freeing memory containing @napi, if
478 * this function returns true.
93d05d4a 479 * Note: core networking stack automatically calls it
5e82b4b2 480 * from netif_napi_del().
93d05d4a 481 * Drivers might want to call this helper to combine all
5e82b4b2 482 * the needed RCU grace periods into a single one.
af12fa6e 483 */
34cbe27e 484bool napi_hash_del(struct napi_struct *napi);
af12fa6e 485
bea3348e
SH
486/**
487 * napi_disable - prevent NAPI from scheduling
5e82b4b2 488 * @n: NAPI context
bea3348e
SH
489 *
490 * Stop NAPI from being scheduled on this context.
491 * Waits till any outstanding processing completes.
492 */
3b47d303 493void napi_disable(struct napi_struct *n);
bea3348e
SH
494
495/**
496 * napi_enable - enable NAPI scheduling
5e82b4b2 497 * @n: NAPI context
bea3348e
SH
498 *
499 * Resume NAPI from being scheduled on this context.
500 * Must be paired with napi_disable.
501 */
502static inline void napi_enable(struct napi_struct *n)
503{
504 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 505 smp_mb__before_atomic();
bea3348e 506 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 507 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
508}
509
c264c3de
SH
510/**
511 * napi_synchronize - wait until NAPI is not running
5e82b4b2 512 * @n: NAPI context
c264c3de
SH
513 *
514 * Wait until NAPI is done being scheduled on this context.
515 * Waits till any outstanding processing completes but
516 * does not disable future activations.
517 */
518static inline void napi_synchronize(const struct napi_struct *n)
519{
facc432f
AB
520 if (IS_ENABLED(CONFIG_SMP))
521 while (test_bit(NAPI_STATE_SCHED, &n->state))
522 msleep(1);
523 else
524 barrier();
c264c3de 525}
c264c3de 526
d94d9fee 527enum netdev_queue_state_t {
73466498
TH
528 __QUEUE_STATE_DRV_XOFF,
529 __QUEUE_STATE_STACK_XOFF,
c3f26a26 530 __QUEUE_STATE_FROZEN,
79d16385 531};
8e2f1a63
DB
532
533#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
534#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
535#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
536
537#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
538#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
539 QUEUE_STATE_FROZEN)
540#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
541 QUEUE_STATE_FROZEN)
542
73466498
TH
543/*
544 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
545 * netif_tx_* functions below are used to manipulate this flag. The
546 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
547 * queue independently. The netif_xmit_*stopped functions below are called
548 * to check if the queue has been stopped by the driver or stack (either
549 * of the XOFF bits are set in the state). Drivers should not need to call
550 * netif_xmit*stopped functions, they should only be using netif_tx_*.
551 */
79d16385 552
bb949fbd 553struct netdev_queue {
6a321cb3 554/*
5e82b4b2 555 * read-mostly part
6a321cb3 556 */
bb949fbd 557 struct net_device *dev;
46e5da40 558 struct Qdisc __rcu *qdisc;
b0e1e646 559 struct Qdisc *qdisc_sleeping;
ccf5ff69 560#ifdef CONFIG_SYSFS
1d24eb48
TH
561 struct kobject kobj;
562#endif
f2cd2d3e
ED
563#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
564 int numa_node;
565#endif
c0ef079c
FW
566 unsigned long tx_maxrate;
567 /*
568 * Number of TX timeouts for this queue
569 * (/sys/class/net/DEV/Q/trans_timeout)
570 */
571 unsigned long trans_timeout;
6a321cb3 572/*
5e82b4b2 573 * write-mostly part
6a321cb3
ED
574 */
575 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
576 int xmit_lock_owner;
9d21493b 577 /*
9b36627a 578 * Time (in jiffies) of last Tx
9d21493b
ED
579 */
580 unsigned long trans_start;
ccf5ff69 581
114cf580
TH
582 unsigned long state;
583
584#ifdef CONFIG_BQL
585 struct dql dql;
586#endif
e8a0464c 587} ____cacheline_aligned_in_smp;
bb949fbd 588
79134e6c
ED
589extern int sysctl_fb_tunnels_only_for_init_net;
590
591static inline bool net_has_fallback_tunnels(const struct net *net)
592{
be9fc097
AB
593 return net == &init_net ||
594 !IS_ENABLED(CONFIG_SYSCTL) ||
595 !sysctl_fb_tunnels_only_for_init_net;
79134e6c
ED
596}
597
f2cd2d3e
ED
598static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599{
600#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 return q->numa_node;
602#else
b236da69 603 return NUMA_NO_NODE;
f2cd2d3e
ED
604#endif
605}
606
607static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608{
609#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 q->numa_node = node;
611#endif
612}
613
df334545 614#ifdef CONFIG_RPS
0a9627f2
TH
615/*
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
618 */
619struct rps_map {
620 unsigned int len;
621 struct rcu_head rcu;
622 u16 cpus[0];
623};
60b778ce 624#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 625
fec5e652 626/*
c445477d
BH
627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628 * tail pointer for that CPU's input queue at the time of last enqueue, and
629 * a hardware filter index.
fec5e652
TH
630 */
631struct rps_dev_flow {
632 u16 cpu;
c445477d 633 u16 filter;
fec5e652
TH
634 unsigned int last_qtail;
635};
c445477d 636#define RPS_NO_FILTER 0xffff
fec5e652
TH
637
638/*
639 * The rps_dev_flow_table structure contains a table of flow mappings.
640 */
641struct rps_dev_flow_table {
642 unsigned int mask;
643 struct rcu_head rcu;
fec5e652
TH
644 struct rps_dev_flow flows[0];
645};
646#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 647 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
648
649/*
650 * The rps_sock_flow_table contains mappings of flows to the last CPU
651 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
652 * Each entry is a 32bit value. Upper part is the high-order bits
653 * of flow hash, lower part is CPU number.
567e4b79 654 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 657 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
658 */
659struct rps_sock_flow_table {
567e4b79 660 u32 mask;
93c1af6c
ED
661
662 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 663};
567e4b79 664#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
665
666#define RPS_NO_CPU 0xffff
667
567e4b79
ED
668extern u32 rps_cpu_mask;
669extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670
fec5e652
TH
671static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
672 u32 hash)
673{
674 if (table && hash) {
567e4b79
ED
675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
fec5e652 677
5e82b4b2 678 /* We only give a hint, preemption can change CPU under us */
567e4b79 679 val |= raw_smp_processor_id();
fec5e652 680
567e4b79
ED
681 if (table->ents[index] != val)
682 table->ents[index] = val;
fec5e652
TH
683 }
684}
685
c445477d 686#ifdef CONFIG_RFS_ACCEL
f629d208
JP
687bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
688 u16 filter_id);
c445477d 689#endif
a953be53 690#endif /* CONFIG_RPS */
c445477d 691
0a9627f2
TH
692/* This structure contains an instance of an RX queue. */
693struct netdev_rx_queue {
a953be53 694#ifdef CONFIG_RPS
6e3f7faf
ED
695 struct rps_map __rcu *rps_map;
696 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 697#endif
6e3f7faf 698 struct kobject kobj;
fe822240 699 struct net_device *dev;
e817f856 700 struct xdp_rxq_info xdp_rxq;
0a9627f2 701} ____cacheline_aligned_in_smp;
a953be53
MD
702
703/*
704 * RX queue sysfs structures and functions.
705 */
706struct rx_queue_attribute {
707 struct attribute attr;
718ad681 708 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
a953be53 709 ssize_t (*store)(struct netdev_rx_queue *queue,
718ad681 710 const char *buf, size_t len);
a953be53 711};
d314774c 712
bf264145
TH
713#ifdef CONFIG_XPS
714/*
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
717 */
718struct xps_map {
719 unsigned int len;
720 unsigned int alloc_len;
721 struct rcu_head rcu;
722 u16 queues[0];
723};
60b778ce 724#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
725#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
727
728/*
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 */
731struct xps_dev_maps {
732 struct rcu_head rcu;
a4177869 733 struct xps_map __rcu *cpu_map[0];
bf264145 734};
184c449f
AD
735#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
736 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
737#endif /* CONFIG_XPS */
738
4f57c087
JF
739#define TC_MAX_QUEUE 16
740#define TC_BITMASK 15
741/* HW offloaded queuing disciplines txq count and offset maps */
742struct netdev_tc_txq {
743 u16 count;
744 u16 offset;
745};
746
68bad94e
NP
747#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748/*
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
751 */
752struct netdev_fcoe_hbainfo {
753 char manufacturer[64];
754 char serial_number[64];
755 char hardware_version[64];
756 char driver_version[64];
757 char optionrom_version[64];
758 char firmware_version[64];
759 char model[256];
760 char model_description[256];
761};
762#endif
763
02637fce 764#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 765
02637fce
JP
766/* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
66b52b0d 768 */
02637fce
JP
769struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
771 unsigned char id_len;
772};
773
d754f98b
SF
774static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
776{
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
779}
780
99932d4f
DB
781typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
783
2572ac53 784enum tc_setup_type {
575ed7d3 785 TC_SETUP_QDISC_MQPRIO,
a1b7c5fd 786 TC_SETUP_CLSU32,
5b33f488 787 TC_SETUP_CLSFLOWER,
ade9b658 788 TC_SETUP_CLSMATCHALL,
332ae8e2 789 TC_SETUP_CLSBPF,
8c4083b3 790 TC_SETUP_BLOCK,
8521db4c 791 TC_SETUP_QDISC_CBS,
602f3baf 792 TC_SETUP_QDISC_RED,
7fdb61b4 793 TC_SETUP_QDISC_PRIO,
16e5cc64
JF
794};
795
f4e63525
JK
796/* These structures hold the attributes of bpf state that are being passed
797 * to the netdevice through the bpf op.
a7862b45 798 */
f4e63525 799enum bpf_netdev_command {
a7862b45
BB
800 /* Set or clear a bpf program used in the earliest stages of packet
801 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
802 * is responsible for calling bpf_prog_put on any old progs that are
803 * stored. In case of error, the callee need not release the new prog
804 * reference, but on success it takes ownership and must bpf_prog_put
805 * when it is no longer used.
806 */
807 XDP_SETUP_PROG,
ee5d032f 808 XDP_SETUP_PROG_HW,
a7862b45 809 /* Check if a bpf program is set on the device. The callee should
ce158e58
JK
810 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
811 * is equivalent to XDP_ATTACHED_DRV.
a7862b45
BB
812 */
813 XDP_QUERY_PROG,
ab3f0063
JK
814 /* BPF program for offload callbacks, invoked at program load time. */
815 BPF_OFFLOAD_VERIFIER_PREP,
816 BPF_OFFLOAD_TRANSLATE,
817 BPF_OFFLOAD_DESTROY,
a3884572
JK
818 BPF_OFFLOAD_MAP_ALLOC,
819 BPF_OFFLOAD_MAP_FREE,
a7862b45
BB
820};
821
cae1927c 822struct bpf_prog_offload_ops;
ddf9f970
JK
823struct netlink_ext_ack;
824
f4e63525
JK
825struct netdev_bpf {
826 enum bpf_netdev_command command;
a7862b45
BB
827 union {
828 /* XDP_SETUP_PROG */
ddf9f970 829 struct {
32d60277 830 u32 flags;
ddf9f970
JK
831 struct bpf_prog *prog;
832 struct netlink_ext_ack *extack;
833 };
a7862b45 834 /* XDP_QUERY_PROG */
58038695 835 struct {
ce158e58 836 u8 prog_attached;
58038695 837 u32 prog_id;
92f0292b
JK
838 /* flags with which program was installed */
839 u32 prog_flags;
58038695 840 };
ab3f0063
JK
841 /* BPF_OFFLOAD_VERIFIER_PREP */
842 struct {
843 struct bpf_prog *prog;
cae1927c 844 const struct bpf_prog_offload_ops *ops; /* callee set */
ab3f0063
JK
845 } verifier;
846 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
847 struct {
848 struct bpf_prog *prog;
849 } offload;
a3884572
JK
850 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
851 struct {
852 struct bpf_offloaded_map *offmap;
853 };
a7862b45
BB
854 };
855};
16e5cc64 856
d77e38e6
SK
857#ifdef CONFIG_XFRM_OFFLOAD
858struct xfrmdev_ops {
859 int (*xdo_dev_state_add) (struct xfrm_state *x);
860 void (*xdo_dev_state_delete) (struct xfrm_state *x);
861 void (*xdo_dev_state_free) (struct xfrm_state *x);
862 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
863 struct xfrm_state *x);
50bd870a 864 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
d77e38e6
SK
865};
866#endif
867
a5c37c63
IL
868#if IS_ENABLED(CONFIG_TLS_DEVICE)
869enum tls_offload_ctx_dir {
870 TLS_OFFLOAD_CTX_DIR_RX,
871 TLS_OFFLOAD_CTX_DIR_TX,
872};
873
874struct tls_crypto_info;
875struct tls_context;
876
877struct tlsdev_ops {
878 int (*tls_dev_add)(struct net_device *netdev, struct sock *sk,
879 enum tls_offload_ctx_dir direction,
880 struct tls_crypto_info *crypto_info,
881 u32 start_offload_tcp_sn);
882 void (*tls_dev_del)(struct net_device *netdev,
883 struct tls_context *ctx,
884 enum tls_offload_ctx_dir direction);
885};
886#endif
887
6c557001
FW
888struct dev_ifalias {
889 struct rcu_head rcuhead;
890 char ifalias[];
891};
892
d314774c
SH
893/*
894 * This structure defines the management hooks for network devices.
00829823
SH
895 * The following hooks can be defined; unless noted otherwise, they are
896 * optional and can be filled with a null pointer.
d314774c
SH
897 *
898 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
899 * This function is called once when a network device is registered.
900 * The network device can use this for any late stage initialization
901 * or semantic validation. It can fail with an error code which will
902 * be propagated back to register_netdev.
d314774c
SH
903 *
904 * void (*ndo_uninit)(struct net_device *dev);
905 * This function is called when device is unregistered or when registration
906 * fails. It is not called if init fails.
907 *
908 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 909 * This function is called when a network device transitions to the up
d314774c
SH
910 * state.
911 *
912 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 913 * This function is called when a network device transitions to the down
d314774c
SH
914 * state.
915 *
dc1f8bf6
SH
916 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
917 * struct net_device *dev);
00829823 918 * Called when a packet needs to be transmitted.
e79d8429
RR
919 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
920 * the queue before that can happen; it's for obsolete devices and weird
921 * corner cases, but the stack really does a non-trivial amount
922 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 923 * Required; cannot be NULL.
00829823 924 *
1a2a1444
DM
925 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
926 * struct net_device *dev
927 * netdev_features_t features);
928 * Called by core transmit path to determine if device is capable of
929 * performing offload operations on a given packet. This is to give
930 * the device an opportunity to implement any restrictions that cannot
931 * be otherwise expressed by feature flags. The check is called with
932 * the set of features that the stack has calculated and it returns
933 * those the driver believes to be appropriate.
cdba756f 934 *
f663dd9a 935 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 936 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 937 * Called to decide which queue to use when device supports multiple
00829823
SH
938 * transmit queues.
939 *
d314774c
SH
940 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
941 * This function is called to allow device receiver to make
5e82b4b2 942 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
943 *
944 * void (*ndo_set_rx_mode)(struct net_device *dev);
945 * This function is called device changes address list filtering.
01789349 946 * If driver handles unicast address filtering, it should set
5e82b4b2 947 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
948 *
949 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
950 * This function is called when the Media Access Control address
37b607c5 951 * needs to be changed. If this interface is not defined, the
5e82b4b2 952 * MAC address can not be changed.
d314774c
SH
953 *
954 * int (*ndo_validate_addr)(struct net_device *dev);
955 * Test if Media Access Control address is valid for the device.
956 *
957 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
958 * Called when a user requests an ioctl which can't be handled by
959 * the generic interface code. If not defined ioctls return
d314774c
SH
960 * not supported error code.
961 *
962 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
963 * Used to set network devices bus interface parameters. This interface
5e82b4b2 964 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
965 * interface (PCI) for low level management.
966 *
967 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
968 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 969 * of a device.
d314774c 970 *
00829823 971 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 972 * Callback used when the transmitter has not made any progress
d314774c
SH
973 * for dev->watchdog ticks.
974 *
bc1f4470 975 * void (*ndo_get_stats64)(struct net_device *dev,
976 * struct rtnl_link_stats64 *storage);
d308e38f 977 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 978 * Called when a user wants to get the network device usage
be1f3c2c 979 * statistics. Drivers must do one of the following:
3cfde79c
BH
980 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
981 * rtnl_link_stats64 structure passed by the caller.
82695d9b 982 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
983 * (which should normally be dev->stats) and return a pointer to
984 * it. The structure may be changed asynchronously only if each
985 * field is written atomically.
986 * 3. Update dev->stats asynchronously and atomically, and define
987 * neither operation.
d314774c 988 *
3df5b3c6 989 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
990 * Return true if this device supports offload stats of this attr_id.
991 *
992 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
993 * void *attr_data)
994 * Get statistics for offload operations by attr_id. Write it into the
995 * attr_data pointer.
996 *
5d632cb7 997 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 998 * If device supports VLAN filtering this function is called when a
80d5c368 999 * VLAN id is registered.
d314774c 1000 *
5d632cb7 1001 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 1002 * If device supports VLAN filtering this function is called when a
80d5c368 1003 * VLAN id is unregistered.
d314774c
SH
1004 *
1005 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
1006 *
1007 * SR-IOV management functions.
1008 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
1009 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1010 * u8 qos, __be16 proto);
ed616689
SC
1011 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1012 * int max_tx_rate);
5f8444a3 1013 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 1014 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
1015 * int (*ndo_get_vf_config)(struct net_device *dev,
1016 * int vf, struct ifla_vf_info *ivf);
1d8faf48 1017 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
1018 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1019 * struct nlattr *port[]);
01a3d796
VZ
1020 *
1021 * Enable or disable the VF ability to query its RSS Redirection Table and
1022 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 1023 * with PF and querying it may introduce a theoretical security risk.
01a3d796 1024 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 1025 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
2572ac53 1026 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
de4784ca 1027 * void *type_data);
6a4bc2b4
FF
1028 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1029 * This is always called from the stack with the rtnl lock held and netif
1030 * tx queues stopped. This allows the netdevice to perform queue
1031 * management safely.
c445477d 1032 *
e9bce845
YZ
1033 * Fiber Channel over Ethernet (FCoE) offload functions.
1034 * int (*ndo_fcoe_enable)(struct net_device *dev);
1035 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1036 * so the underlying device can perform whatever needed configuration or
1037 * initialization to support acceleration of FCoE traffic.
1038 *
1039 * int (*ndo_fcoe_disable)(struct net_device *dev);
1040 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1041 * so the underlying device can perform whatever needed clean-ups to
1042 * stop supporting acceleration of FCoE traffic.
1043 *
1044 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1045 * struct scatterlist *sgl, unsigned int sgc);
1046 * Called when the FCoE Initiator wants to initialize an I/O that
1047 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1048 * perform necessary setup and returns 1 to indicate the device is set up
1049 * successfully to perform DDP on this I/O, otherwise this returns 0.
1050 *
1051 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1052 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1053 * indicated by the FC exchange id 'xid', so the underlying device can
1054 * clean up and reuse resources for later DDP requests.
1055 *
1056 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1057 * struct scatterlist *sgl, unsigned int sgc);
1058 * Called when the FCoE Target wants to initialize an I/O that
1059 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1060 * perform necessary setup and returns 1 to indicate the device is set up
1061 * successfully to perform DDP on this I/O, otherwise this returns 0.
1062 *
68bad94e
NP
1063 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1064 * struct netdev_fcoe_hbainfo *hbainfo);
1065 * Called when the FCoE Protocol stack wants information on the underlying
1066 * device. This information is utilized by the FCoE protocol stack to
1067 * register attributes with Fiber Channel management service as per the
1068 * FC-GS Fabric Device Management Information(FDMI) specification.
1069 *
e9bce845
YZ
1070 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1071 * Called when the underlying device wants to override default World Wide
1072 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1073 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1074 * protocol stack to use.
1075 *
c445477d
BH
1076 * RFS acceleration.
1077 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1078 * u16 rxq_index, u32 flow_id);
1079 * Set hardware filter for RFS. rxq_index is the target queue index;
1080 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1081 * Return the filter ID on success, or a negative error code.
fbaec0ea 1082 *
8b98a70c 1083 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1084 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1085 * Called to make another netdev an underling.
1086 *
1087 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1088 * Called to release previously enslaved netdev.
5455c699
MM
1089 *
1090 * Feature/offload setting functions.
1a2a1444
DM
1091 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1092 * netdev_features_t features);
1093 * Adjusts the requested feature flags according to device-specific
1094 * constraints, and returns the resulting flags. Must not modify
1095 * the device state.
1096 *
c8f44aff 1097 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1098 * Called to update device configuration to new features. Passed
1099 * feature set might be less than what was returned by ndo_fix_features()).
1100 * Must return >0 or -errno if it changed dev->features itself.
1101 *
edc7d573 1102 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1103 * struct net_device *dev,
f6f6424b 1104 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1105 * Adds an FDB entry to dev for addr.
1690be63
VY
1106 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1107 * struct net_device *dev,
f6f6424b 1108 * const unsigned char *addr, u16 vid)
77162022
JF
1109 * Deletes the FDB entry from dev coresponding to addr.
1110 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1111 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1112 * int *idx)
77162022
JF
1113 * Used to add FDB entries to dump requests. Implementers should add
1114 * entries to skb and update idx with the number of entries.
e5a55a89 1115 *
ad41faa8
ND
1116 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1117 * u16 flags)
e5a55a89 1118 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1119 * struct net_device *dev, u32 filter_mask,
1120 * int nlflags)
ad41faa8
ND
1121 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1122 * u16 flags);
4bf84c35
JP
1123 *
1124 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1125 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1126 * which do not represent real hardware may define this to allow their
1127 * userspace components to manage their virtual carrier state. Devices
1128 * that determine carrier state from physical hardware properties (eg
1129 * network cables) or protocol-dependent mechanisms (eg
1130 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1131 *
1132 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1133 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1134 * Called to get ID of physical port of this device. If driver does
1135 * not implement this, it is assumed that the hw is not able to have
1136 * multiple net devices on single physical port.
53cf5275 1137 *
7c46a640
AD
1138 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1139 * struct udp_tunnel_info *ti);
1140 * Called by UDP tunnel to notify a driver about the UDP port and socket
1141 * address family that a UDP tunnel is listnening to. It is called only
1142 * when a new port starts listening. The operation is protected by the
1143 * RTNL.
1144 *
1145 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1146 * struct udp_tunnel_info *ti);
1147 * Called by UDP tunnel to notify the driver about a UDP port and socket
1148 * address family that the UDP tunnel is not listening to anymore. The
1149 * operation is protected by the RTNL.
1150 *
a6cc0cfa
JF
1151 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1152 * struct net_device *dev)
1153 * Called by upper layer devices to accelerate switching or other
1154 * station functionality into hardware. 'pdev is the lowerdev
1155 * to use for the offload and 'dev' is the net device that will
1156 * back the offload. Returns a pointer to the private structure
1157 * the upper layer will maintain.
1158 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1159 * Called by upper layer device to delete the station created
1160 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1161 * the station and priv is the structure returned by the add
1162 * operation.
822b3b2e
JF
1163 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1164 * int queue_index, u32 maxrate);
1165 * Called when a user wants to set a max-rate limitation of specific
1166 * TX queue.
a54acb3a
ND
1167 * int (*ndo_get_iflink)(const struct net_device *dev);
1168 * Called to get the iflink value of this device.
d746d707 1169 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1170 * bool proto_down);
d746d707
AK
1171 * This function is used to pass protocol port error state information
1172 * to the switch driver. The switch driver can react to the proto_down
1173 * by doing a phys down on the associated switch port.
fc4099f1
PS
1174 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1175 * This function is used to get egress tunnel information for given skb.
1176 * This is useful for retrieving outer tunnel header parameters while
1177 * sampling packet.
871b642a
PA
1178 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1179 * This function is used to specify the headroom that the skb must
1180 * consider when allocation skb during packet reception. Setting
1181 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1182 * forward. Setting a negative value resets the rx headroom to the
871b642a 1183 * default value.
f4e63525 1184 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
a7862b45 1185 * This function is used to set or query state related to XDP on the
f4e63525
JK
1186 * netdevice and manage BPF offload. See definition of
1187 * enum bpf_netdev_command for details.
735fc405
JDB
1188 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp);
1189 * This function is used to submit @n XDP packets for transmit on a
1190 * netdevice. Returns number of frames successfully transmitted, frames
1191 * that got dropped are freed/returned via xdp_return_frame().
1192 * Returns negative number, means general error invoking ndo, meaning
1193 * no frames were xmit'ed and core-caller will free all frames.
1194 * TODO: Consider add flag to allow sending flush operation.
11393cc9 1195 * void (*ndo_xdp_flush)(struct net_device *dev);
f16ded59
RR
1196 * This function is used to inform the driver to flush a particular
1197 * xdp tx queue. Must be called on same CPU as xdp_xmit.
d314774c
SH
1198 */
1199struct net_device_ops {
1200 int (*ndo_init)(struct net_device *dev);
1201 void (*ndo_uninit)(struct net_device *dev);
1202 int (*ndo_open)(struct net_device *dev);
1203 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1204 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1205 struct net_device *dev);
1206 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1207 struct net_device *dev,
1208 netdev_features_t features);
00829823 1209 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1210 struct sk_buff *skb,
99932d4f
DB
1211 void *accel_priv,
1212 select_queue_fallback_t fallback);
d314774c
SH
1213 void (*ndo_change_rx_flags)(struct net_device *dev,
1214 int flags);
d314774c 1215 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1216 int (*ndo_set_mac_address)(struct net_device *dev,
1217 void *addr);
d314774c 1218 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1219 int (*ndo_do_ioctl)(struct net_device *dev,
1220 struct ifreq *ifr, int cmd);
d314774c
SH
1221 int (*ndo_set_config)(struct net_device *dev,
1222 struct ifmap *map);
00829823
SH
1223 int (*ndo_change_mtu)(struct net_device *dev,
1224 int new_mtu);
1225 int (*ndo_neigh_setup)(struct net_device *dev,
1226 struct neigh_parms *);
d314774c
SH
1227 void (*ndo_tx_timeout) (struct net_device *dev);
1228
bc1f4470 1229 void (*ndo_get_stats64)(struct net_device *dev,
1230 struct rtnl_link_stats64 *storage);
3df5b3c6 1231 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1232 int (*ndo_get_offload_stats)(int attr_id,
1233 const struct net_device *dev,
1234 void *attr_data);
d314774c
SH
1235 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1236
8e586137 1237 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1238 __be16 proto, u16 vid);
8e586137 1239 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1240 __be16 proto, u16 vid);
d314774c 1241#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1242 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1243 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1244 struct netpoll_info *info);
0e34e931 1245 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1246#endif
95c26df8
WM
1247 int (*ndo_set_vf_mac)(struct net_device *dev,
1248 int queue, u8 *mac);
1249 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1250 int queue, u16 vlan,
1251 u8 qos, __be16 proto);
ed616689
SC
1252 int (*ndo_set_vf_rate)(struct net_device *dev,
1253 int vf, int min_tx_rate,
1254 int max_tx_rate);
5f8444a3
GR
1255 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1256 int vf, bool setting);
dd461d6a
HS
1257 int (*ndo_set_vf_trust)(struct net_device *dev,
1258 int vf, bool setting);
95c26df8
WM
1259 int (*ndo_get_vf_config)(struct net_device *dev,
1260 int vf,
1261 struct ifla_vf_info *ivf);
1d8faf48
RE
1262 int (*ndo_set_vf_link_state)(struct net_device *dev,
1263 int vf, int link_state);
3b766cd8
EBE
1264 int (*ndo_get_vf_stats)(struct net_device *dev,
1265 int vf,
1266 struct ifla_vf_stats
1267 *vf_stats);
57b61080
SF
1268 int (*ndo_set_vf_port)(struct net_device *dev,
1269 int vf,
1270 struct nlattr *port[]);
1271 int (*ndo_get_vf_port)(struct net_device *dev,
1272 int vf, struct sk_buff *skb);
cc8e27cc
EC
1273 int (*ndo_set_vf_guid)(struct net_device *dev,
1274 int vf, u64 guid,
1275 int guid_type);
01a3d796
VZ
1276 int (*ndo_set_vf_rss_query_en)(
1277 struct net_device *dev,
1278 int vf, bool setting);
16e5cc64 1279 int (*ndo_setup_tc)(struct net_device *dev,
2572ac53 1280 enum tc_setup_type type,
de4784ca 1281 void *type_data);
d11ead75 1282#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1283 int (*ndo_fcoe_enable)(struct net_device *dev);
1284 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1285 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1286 u16 xid,
1287 struct scatterlist *sgl,
1288 unsigned int sgc);
1289 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1290 u16 xid);
6247e086
YZ
1291 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1292 u16 xid,
1293 struct scatterlist *sgl,
1294 unsigned int sgc);
68bad94e
NP
1295 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1296 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1297#endif
1298
d11ead75 1299#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1300#define NETDEV_FCOE_WWNN 0
1301#define NETDEV_FCOE_WWPN 1
1302 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1303 u64 *wwn, int type);
4d288d57 1304#endif
3c9c36bc 1305
c445477d
BH
1306#ifdef CONFIG_RFS_ACCEL
1307 int (*ndo_rx_flow_steer)(struct net_device *dev,
1308 const struct sk_buff *skb,
1309 u16 rxq_index,
1310 u32 flow_id);
1311#endif
fbaec0ea 1312 int (*ndo_add_slave)(struct net_device *dev,
33eaf2a6
DA
1313 struct net_device *slave_dev,
1314 struct netlink_ext_ack *extack);
fbaec0ea
JP
1315 int (*ndo_del_slave)(struct net_device *dev,
1316 struct net_device *slave_dev);
c8f44aff
MM
1317 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1318 netdev_features_t features);
5455c699 1319 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1320 netdev_features_t features);
503eebc2
JP
1321 int (*ndo_neigh_construct)(struct net_device *dev,
1322 struct neighbour *n);
1323 void (*ndo_neigh_destroy)(struct net_device *dev,
1324 struct neighbour *n);
77162022
JF
1325
1326 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1327 struct nlattr *tb[],
77162022 1328 struct net_device *dev,
6b6e2725 1329 const unsigned char *addr,
f6f6424b 1330 u16 vid,
77162022
JF
1331 u16 flags);
1332 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1333 struct nlattr *tb[],
77162022 1334 struct net_device *dev,
f6f6424b
JP
1335 const unsigned char *addr,
1336 u16 vid);
77162022
JF
1337 int (*ndo_fdb_dump)(struct sk_buff *skb,
1338 struct netlink_callback *cb,
1339 struct net_device *dev,
5d5eacb3 1340 struct net_device *filter_dev,
d297653d 1341 int *idx);
e5a55a89
JF
1342
1343 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1344 struct nlmsghdr *nlh,
1345 u16 flags);
e5a55a89
JF
1346 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1347 u32 pid, u32 seq,
6cbdceeb 1348 struct net_device *dev,
46c264da
ND
1349 u32 filter_mask,
1350 int nlflags);
407af329 1351 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1352 struct nlmsghdr *nlh,
1353 u16 flags);
4bf84c35
JP
1354 int (*ndo_change_carrier)(struct net_device *dev,
1355 bool new_carrier);
66b52b0d 1356 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1357 struct netdev_phys_item_id *ppid);
db24a904
DA
1358 int (*ndo_get_phys_port_name)(struct net_device *dev,
1359 char *name, size_t len);
7c46a640
AD
1360 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1361 struct udp_tunnel_info *ti);
1362 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1363 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1364 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1365 struct net_device *dev);
1366 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1367 void *priv);
1368
25175ba5 1369 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1370 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1371 int queue_index,
1372 u32 maxrate);
a54acb3a 1373 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1374 int (*ndo_change_proto_down)(struct net_device *dev,
1375 bool proto_down);
fc4099f1
PS
1376 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1377 struct sk_buff *skb);
871b642a
PA
1378 void (*ndo_set_rx_headroom)(struct net_device *dev,
1379 int needed_headroom);
f4e63525
JK
1380 int (*ndo_bpf)(struct net_device *dev,
1381 struct netdev_bpf *bpf);
735fc405
JDB
1382 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1383 struct xdp_frame **xdp);
11393cc9 1384 void (*ndo_xdp_flush)(struct net_device *dev);
d314774c
SH
1385};
1386
7aa98047
LR
1387/**
1388 * enum net_device_priv_flags - &struct net_device priv_flags
1389 *
1390 * These are the &struct net_device, they are only set internally
1391 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1392 * userspace; this means that the order of these flags can change
7aa98047
LR
1393 * during any kernel release.
1394 *
1395 * You should have a pretty good reason to be extending these flags.
1396 *
1397 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1398 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1399 * @IFF_BONDING: bonding master or slave
7aa98047 1400 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1401 * @IFF_WAN_HDLC: WAN HDLC device
1402 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1403 * release skb->dst
1404 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1405 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1406 * @IFF_MACVLAN_PORT: device used as macvlan port
1407 * @IFF_BRIDGE_PORT: device used as bridge port
1408 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1409 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1410 * @IFF_UNICAST_FLT: Supports unicast filtering
1411 * @IFF_TEAM_PORT: device used as team port
1412 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1413 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1414 * change when it's running
1415 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1416 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1417 * underlying stacked devices
007979ea 1418 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1419 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1420 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1421 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1422 * @IFF_TEAM: device is a team device
d4ab4286 1423 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1424 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1425 * entity (i.e. the master device for bridged veth)
3c175784 1426 * @IFF_MACSEC: device is a MACsec device
f5426250 1427 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
7aa98047
LR
1428 */
1429enum netdev_priv_flags {
1430 IFF_802_1Q_VLAN = 1<<0,
1431 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1432 IFF_BONDING = 1<<2,
1433 IFF_ISATAP = 1<<3,
1434 IFF_WAN_HDLC = 1<<4,
1435 IFF_XMIT_DST_RELEASE = 1<<5,
1436 IFF_DONT_BRIDGE = 1<<6,
1437 IFF_DISABLE_NETPOLL = 1<<7,
1438 IFF_MACVLAN_PORT = 1<<8,
1439 IFF_BRIDGE_PORT = 1<<9,
1440 IFF_OVS_DATAPATH = 1<<10,
1441 IFF_TX_SKB_SHARING = 1<<11,
1442 IFF_UNICAST_FLT = 1<<12,
1443 IFF_TEAM_PORT = 1<<13,
1444 IFF_SUPP_NOFCS = 1<<14,
1445 IFF_LIVE_ADDR_CHANGE = 1<<15,
1446 IFF_MACVLAN = 1<<16,
1447 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1ec54cb4
PA
1448 IFF_L3MDEV_MASTER = 1<<18,
1449 IFF_NO_QUEUE = 1<<19,
1450 IFF_OPENVSWITCH = 1<<20,
1451 IFF_L3MDEV_SLAVE = 1<<21,
1452 IFF_TEAM = 1<<22,
1453 IFF_RXFH_CONFIGURED = 1<<23,
1454 IFF_PHONY_HEADROOM = 1<<24,
1455 IFF_MACSEC = 1<<25,
f5426250 1456 IFF_NO_RX_HANDLER = 1<<26,
7aa98047
LR
1457};
1458
1459#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1460#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1461#define IFF_BONDING IFF_BONDING
7aa98047 1462#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1463#define IFF_WAN_HDLC IFF_WAN_HDLC
1464#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1465#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1466#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1467#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1468#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1469#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1470#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1471#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1472#define IFF_TEAM_PORT IFF_TEAM_PORT
1473#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1474#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1475#define IFF_MACVLAN IFF_MACVLAN
02875878 1476#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
007979ea 1477#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1478#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1479#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1480#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1481#define IFF_TEAM IFF_TEAM
d4ab4286 1482#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1483#define IFF_MACSEC IFF_MACSEC
f5426250 1484#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
7aa98047 1485
536721b1
KK
1486/**
1487 * struct net_device - The DEVICE structure.
d651983d
MCC
1488 *
1489 * Actually, this whole structure is a big mistake. It mixes I/O
1490 * data with strictly "high-level" data, and it has to know about
1491 * almost every data structure used in the INET module.
536721b1
KK
1492 *
1493 * @name: This is the first field of the "visible" part of this structure
1494 * (i.e. as seen by users in the "Space.c" file). It is the name
d651983d 1495 * of the interface.
536721b1
KK
1496 *
1497 * @name_hlist: Device name hash chain, please keep it close to name[]
1498 * @ifalias: SNMP alias
1499 * @mem_end: Shared memory end
1500 * @mem_start: Shared memory start
1501 * @base_addr: Device I/O address
1502 * @irq: Device IRQ number
1503 *
1504 * @state: Generic network queuing layer state, see netdev_state_t
1505 * @dev_list: The global list of network devices
5e82b4b2
BH
1506 * @napi_list: List entry used for polling NAPI devices
1507 * @unreg_list: List entry when we are unregistering the
1508 * device; see the function unregister_netdev
1509 * @close_list: List entry used when we are closing the device
62d885fe
BP
1510 * @ptype_all: Device-specific packet handlers for all protocols
1511 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1512 *
1513 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1514 * @features: Currently active device features
1515 * @hw_features: User-changeable features
1516 *
1517 * @wanted_features: User-requested features
1518 * @vlan_features: Mask of features inheritable by VLAN devices
1519 *
1520 * @hw_enc_features: Mask of features inherited by encapsulating devices
1521 * This field indicates what encapsulation
1522 * offloads the hardware is capable of doing,
1523 * and drivers will need to set them appropriately.
1524 *
1525 * @mpls_features: Mask of features inheritable by MPLS
1526 *
1527 * @ifindex: interface index
5e82b4b2 1528 * @group: The group the device belongs to
536721b1
KK
1529 *
1530 * @stats: Statistics struct, which was left as a legacy, use
1531 * rtnl_link_stats64 instead
1532 *
1533 * @rx_dropped: Dropped packets by core network,
1534 * do not use this in drivers
1535 * @tx_dropped: Dropped packets by core network,
1536 * do not use this in drivers
6e7333d3
JW
1537 * @rx_nohandler: nohandler dropped packets by core network on
1538 * inactive devices, do not use this in drivers
9e55e5d3
FF
1539 * @carrier_up_count: Number of times the carrier has been up
1540 * @carrier_down_count: Number of times the carrier has been down
536721b1 1541 *
536721b1
KK
1542 * @wireless_handlers: List of functions to handle Wireless Extensions,
1543 * instead of ioctl,
1544 * see <net/iw_handler.h> for details.
1545 * @wireless_data: Instance data managed by the core of wireless extensions
1546 *
1547 * @netdev_ops: Includes several pointers to callbacks,
1548 * if one wants to override the ndo_*() functions
1549 * @ethtool_ops: Management operations
f997c55c
AA
1550 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1551 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1552 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1553 * of Layer 2 headers.
1554 *
1555 * @flags: Interface flags (a la BSD)
1556 * @priv_flags: Like 'flags' but invisible to userspace,
1557 * see if.h for the definitions
1558 * @gflags: Global flags ( kept as legacy )
1559 * @padded: How much padding added by alloc_netdev()
1560 * @operstate: RFC2863 operstate
1561 * @link_mode: Mapping policy to operstate
1562 * @if_port: Selectable AUI, TP, ...
1563 * @dma: DMA channel
1564 * @mtu: Interface MTU value
61e84623
JW
1565 * @min_mtu: Interface Minimum MTU value
1566 * @max_mtu: Interface Maximum MTU value
536721b1 1567 * @type: Interface hardware type
2793a23a 1568 * @hard_header_len: Maximum hardware header length.
217e6fa2 1569 * @min_header_len: Minimum hardware header length
536721b1
KK
1570 *
1571 * @needed_headroom: Extra headroom the hardware may need, but not in all
1572 * cases can this be guaranteed
1573 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1574 * cases can this be guaranteed. Some cases also use
1575 * LL_MAX_HEADER instead to allocate the skb
1576 *
1577 * interface address info:
1578 *
1579 * @perm_addr: Permanent hw address
1580 * @addr_assign_type: Hw address assignment type
1581 * @addr_len: Hardware address length
8626a0c8 1582 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1583 * @dev_id: Used to differentiate devices that share
1584 * the same link layer address
1585 * @dev_port: Used to differentiate devices that share
1586 * the same function
1587 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1588 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1589 * has been enabled due to the need to listen to
1590 * additional unicast addresses in a device that
1591 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1592 * @uc: unicast mac addresses
1593 * @mc: multicast mac addresses
1594 * @dev_addrs: list of device hw addresses
1595 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1596 * @promiscuity: Number of times the NIC is told to work in
1597 * promiscuous mode; if it becomes 0 the NIC will
1598 * exit promiscuous mode
536721b1
KK
1599 * @allmulti: Counter, enables or disables allmulticast mode
1600 *
1601 * @vlan_info: VLAN info
1602 * @dsa_ptr: dsa specific data
1603 * @tipc_ptr: TIPC specific data
1604 * @atalk_ptr: AppleTalk link
1605 * @ip_ptr: IPv4 specific data
1606 * @dn_ptr: DECnet specific data
1607 * @ip6_ptr: IPv6 specific data
1608 * @ax25_ptr: AX.25 specific data
1609 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1610 *
536721b1
KK
1611 * @dev_addr: Hw address (before bcast,
1612 * because most packets are unicast)
1613 *
1614 * @_rx: Array of RX queues
1615 * @num_rx_queues: Number of RX queues
1616 * allocated at register_netdev() time
1617 * @real_num_rx_queues: Number of RX queues currently active in device
1618 *
1619 * @rx_handler: handler for received packets
1620 * @rx_handler_data: XXX: need comments on this one
46209401
JP
1621 * @miniq_ingress: ingress/clsact qdisc specific data for
1622 * ingress processing
536721b1
KK
1623 * @ingress_queue: XXX: need comments on this one
1624 * @broadcast: hw bcast address
1625 *
14ffbbb8
TG
1626 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1627 * indexed by RX queue number. Assigned by driver.
1628 * This must only be set if the ndo_rx_flow_steer
1629 * operation is defined
1630 * @index_hlist: Device index hash chain
1631 *
536721b1
KK
1632 * @_tx: Array of TX queues
1633 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1634 * @real_num_tx_queues: Number of TX queues currently active in device
1635 * @qdisc: Root qdisc from userspace point of view
1636 * @tx_queue_len: Max frames per queue allowed
1637 * @tx_global_lock: XXX: need comments on this one
1638 *
1639 * @xps_maps: XXX: need comments on this one
46209401
JP
1640 * @miniq_egress: clsact qdisc specific data for
1641 * egress processing
536721b1 1642 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1643 * the watchdog (see dev_watchdog())
536721b1
KK
1644 * @watchdog_timer: List of timers
1645 *
1646 * @pcpu_refcnt: Number of references to this device
1647 * @todo_list: Delayed register/unregister
536721b1
KK
1648 * @link_watch_list: XXX: need comments on this one
1649 *
1650 * @reg_state: Register/unregister state machine
1651 * @dismantle: Device is going to be freed
1652 * @rtnl_link_state: This enum represents the phases of creating
1653 * a new link
1654 *
cf124db5
DM
1655 * @needs_free_netdev: Should unregister perform free_netdev?
1656 * @priv_destructor: Called from unregister
536721b1
KK
1657 * @npinfo: XXX: need comments on this one
1658 * @nd_net: Network namespace this network device is inside
1659 *
1660 * @ml_priv: Mid-layer private
1661 * @lstats: Loopback statistics
1662 * @tstats: Tunnel statistics
1663 * @dstats: Dummy statistics
1664 * @vstats: Virtual ethernet statistics
1665 *
1666 * @garp_port: GARP
1667 * @mrp_port: MRP
1668 *
1669 * @dev: Class/net/name entry
1670 * @sysfs_groups: Space for optional device, statistics and wireless
1671 * sysfs groups
1672 *
1673 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1674 * @rtnl_link_ops: Rtnl_link_ops
1675 *
1676 * @gso_max_size: Maximum size of generic segmentation offload
1677 * @gso_max_segs: Maximum number of segments that can be passed to the
1678 * NIC for GSO
1679 *
1680 * @dcbnl_ops: Data Center Bridging netlink ops
1681 * @num_tc: Number of traffic classes in the net device
1682 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1683 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1684 *
1685 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1686 *
1687 * @priomap: XXX: need comments on this one
1688 * @phydev: Physical device may attach itself
1689 * for hardware timestamping
e679c9c1 1690 * @sfp_bus: attached &struct sfp_bus structure.
536721b1 1691 *
123b3652
ED
1692 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1693 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1694 *
d746d707
AK
1695 * @proto_down: protocol port state information can be sent to the
1696 * switch driver and used to set the phys state of the
1697 * switch port.
1698 *
1da177e4
LT
1699 * FIXME: cleanup struct net_device such that network protocol info
1700 * moves out.
1701 */
1702
d94d9fee 1703struct net_device {
1da177e4 1704 char name[IFNAMSIZ];
9356b8fc 1705 struct hlist_node name_hlist;
6c557001 1706 struct dev_ifalias __rcu *ifalias;
1da177e4
LT
1707 /*
1708 * I/O specific fields
1709 * FIXME: Merge these and struct ifmap into one
1710 */
536721b1
KK
1711 unsigned long mem_end;
1712 unsigned long mem_start;
1713 unsigned long base_addr;
1714 int irq;
1da177e4
LT
1715
1716 /*
536721b1
KK
1717 * Some hardware also needs these fields (state,dev_list,
1718 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1719 * part of the usual set specified in Space.c.
1720 */
1721
1da177e4
LT
1722 unsigned long state;
1723
7562f876 1724 struct list_head dev_list;
bea3348e 1725 struct list_head napi_list;
44a0873d 1726 struct list_head unreg_list;
5cde2829 1727 struct list_head close_list;
7866a621
SN
1728 struct list_head ptype_all;
1729 struct list_head ptype_specific;
2f268f12 1730
2f268f12
VF
1731 struct {
1732 struct list_head upper;
1733 struct list_head lower;
1734 } adj_list;
1735
c8f44aff 1736 netdev_features_t features;
c8f44aff 1737 netdev_features_t hw_features;
c8f44aff 1738 netdev_features_t wanted_features;
c8f44aff 1739 netdev_features_t vlan_features;
6a674e9c 1740 netdev_features_t hw_enc_features;
0d89d203 1741 netdev_features_t mpls_features;
802ab55a 1742 netdev_features_t gso_partial_features;
04ed3e74 1743
1da177e4 1744 int ifindex;
7a66bbc9 1745 int group;
1da177e4 1746
c45d286e 1747 struct net_device_stats stats;
015f0688 1748
015f0688
ED
1749 atomic_long_t rx_dropped;
1750 atomic_long_t tx_dropped;
6e7333d3 1751 atomic_long_t rx_nohandler;
1da177e4 1752
b2d3bcfa
DD
1753 /* Stats to monitor link on/off, flapping */
1754 atomic_t carrier_up_count;
1755 atomic_t carrier_down_count;
1756
b86e0280 1757#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1758 const struct iw_handler_def *wireless_handlers;
1759 struct iw_public_data *wireless_data;
b86e0280 1760#endif
d314774c 1761 const struct net_device_ops *netdev_ops;
76fd8593 1762 const struct ethtool_ops *ethtool_ops;
4170604f 1763#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1764 const struct switchdev_ops *switchdev_ops;
4170604f 1765#endif
1b69c6d0
DA
1766#ifdef CONFIG_NET_L3_MASTER_DEV
1767 const struct l3mdev_ops *l3mdev_ops;
1768#endif
f997c55c
AA
1769#if IS_ENABLED(CONFIG_IPV6)
1770 const struct ndisc_ops *ndisc_ops;
1771#endif
1da177e4 1772
9cb0d21d 1773#ifdef CONFIG_XFRM_OFFLOAD
d77e38e6
SK
1774 const struct xfrmdev_ops *xfrmdev_ops;
1775#endif
1776
a5c37c63
IL
1777#if IS_ENABLED(CONFIG_TLS_DEVICE)
1778 const struct tlsdev_ops *tlsdev_ops;
1779#endif
1780
3b04ddde
SH
1781 const struct header_ops *header_ops;
1782
536721b1
KK
1783 unsigned int flags;
1784 unsigned int priv_flags;
1785
1da177e4 1786 unsigned short gflags;
536721b1 1787 unsigned short padded;
1da177e4 1788
536721b1
KK
1789 unsigned char operstate;
1790 unsigned char link_mode;
b00055aa 1791
536721b1
KK
1792 unsigned char if_port;
1793 unsigned char dma;
bdc220da 1794
536721b1 1795 unsigned int mtu;
61e84623
JW
1796 unsigned int min_mtu;
1797 unsigned int max_mtu;
536721b1
KK
1798 unsigned short type;
1799 unsigned short hard_header_len;
d92be7a4 1800 unsigned char min_header_len;
1da177e4 1801
f5184d26
JB
1802 unsigned short needed_headroom;
1803 unsigned short needed_tailroom;
1804
1da177e4 1805 /* Interface address info. */
536721b1
KK
1806 unsigned char perm_addr[MAX_ADDR_LEN];
1807 unsigned char addr_assign_type;
1808 unsigned char addr_len;
a0a9663d 1809 unsigned short neigh_priv_len;
536721b1
KK
1810 unsigned short dev_id;
1811 unsigned short dev_port;
ccffad25 1812 spinlock_t addr_list_lock;
14ffbbb8
TG
1813 unsigned char name_assign_type;
1814 bool uc_promisc;
536721b1
KK
1815 struct netdev_hw_addr_list uc;
1816 struct netdev_hw_addr_list mc;
1817 struct netdev_hw_addr_list dev_addrs;
1818
4c3d5e7b
ED
1819#ifdef CONFIG_SYSFS
1820 struct kset *queues_kset;
1821#endif
9d45abe1
WC
1822 unsigned int promiscuity;
1823 unsigned int allmulti;
1da177e4 1824
1da177e4 1825
5e82b4b2 1826 /* Protocol-specific pointers */
65ac6a5f 1827
d11ead75 1828#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1829 struct vlan_info __rcu *vlan_info;
65ac6a5f 1830#endif
34a430d7 1831#if IS_ENABLED(CONFIG_NET_DSA)
2f657a60 1832 struct dsa_port *dsa_ptr;
37cb0620
YX
1833#endif
1834#if IS_ENABLED(CONFIG_TIPC)
536721b1 1835 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1836#endif
89e58148 1837#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
536721b1 1838 void *atalk_ptr;
89e58148 1839#endif
536721b1 1840 struct in_device __rcu *ip_ptr;
330c7272 1841#if IS_ENABLED(CONFIG_DECNET)
536721b1 1842 struct dn_dev __rcu *dn_ptr;
330c7272 1843#endif
536721b1 1844 struct inet6_dev __rcu *ip6_ptr;
19ff13f2 1845#if IS_ENABLED(CONFIG_AX25)
536721b1 1846 void *ax25_ptr;
19ff13f2 1847#endif
536721b1 1848 struct wireless_dev *ieee80211_ptr;
98a18b6f 1849 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1850#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1851 struct mpls_dev __rcu *mpls_ptr;
1852#endif
1da177e4 1853
9356b8fc 1854/*
cd13539b 1855 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1856 */
9356b8fc 1857 /* Interface address info used in eth_type_trans() */
536721b1 1858 unsigned char *dev_addr;
f001fde5 1859
0a9627f2 1860 struct netdev_rx_queue *_rx;
0a9627f2 1861 unsigned int num_rx_queues;
62fe0b40 1862 unsigned int real_num_rx_queues;
0a9627f2 1863
7acedaf5 1864 struct bpf_prog __rcu *xdp_prog;
3b47d303 1865 unsigned long gro_flush_timeout;
61391cde 1866 rx_handler_func_t __rcu *rx_handler;
1867 void __rcu *rx_handler_data;
e8a0464c 1868
4cda01e8 1869#ifdef CONFIG_NET_CLS_ACT
46209401 1870 struct mini_Qdisc __rcu *miniq_ingress;
d2788d34 1871#endif
24824a09 1872 struct netdev_queue __rcu *ingress_queue;
e687ad60 1873#ifdef CONFIG_NETFILTER_INGRESS
960632ec 1874 struct nf_hook_entries __rcu *nf_hooks_ingress;
e687ad60 1875#endif
d2788d34 1876
536721b1 1877 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1878#ifdef CONFIG_RFS_ACCEL
1879 struct cpu_rmap *rx_cpu_rmap;
1880#endif
1881 struct hlist_node index_hlist;
cd13539b
ED
1882
1883/*
1884 * Cache lines mostly used on transmit path
1885 */
e8a0464c
DM
1886 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1887 unsigned int num_tx_queues;
fd2ea0a7 1888 unsigned int real_num_tx_queues;
af356afa 1889 struct Qdisc *qdisc;
59cc1f61
JK
1890#ifdef CONFIG_NET_SCHED
1891 DECLARE_HASHTABLE (qdisc_hash, 4);
1892#endif
0cd29503 1893 unsigned int tx_queue_len;
c3f26a26 1894 spinlock_t tx_global_lock;
14ffbbb8 1895 int watchdog_timeo;
cd13539b 1896
bf264145 1897#ifdef CONFIG_XPS
a4177869 1898 struct xps_dev_maps __rcu *xps_maps;
bf264145 1899#endif
1f211a1b 1900#ifdef CONFIG_NET_CLS_ACT
46209401 1901 struct mini_Qdisc __rcu *miniq_egress;
1f211a1b 1902#endif
0c4f691f 1903
9356b8fc 1904 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1905 struct timer_list watchdog_timer;
1906
29b4433d 1907 int __percpu *pcpu_refcnt;
1da177e4 1908 struct list_head todo_list;
1da177e4 1909
e014debe 1910 struct list_head link_watch_list;
572a103d 1911
1da177e4 1912 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1913 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1914 NETREG_UNREGISTERING, /* called unregister_netdevice */
1915 NETREG_UNREGISTERED, /* completed unregister todo */
1916 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1917 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1918 } reg_state:8;
1919
536721b1 1920 bool dismantle;
a2835763
PM
1921
1922 enum {
1923 RTNL_LINK_INITIALIZED,
1924 RTNL_LINK_INITIALIZING,
1925 } rtnl_link_state:16;
1da177e4 1926
cf124db5
DM
1927 bool needs_free_netdev;
1928 void (*priv_destructor)(struct net_device *dev);
1da177e4 1929
1da177e4 1930#ifdef CONFIG_NETPOLL
5fbee843 1931 struct netpoll_info __rcu *npinfo;
1da177e4 1932#endif
eae792b7 1933
0c5c9fb5 1934 possible_net_t nd_net;
4a1c5371 1935
4951704b 1936 /* mid-layer private */
a7855c78 1937 union {
536721b1
KK
1938 void *ml_priv;
1939 struct pcpu_lstats __percpu *lstats;
8f84985f 1940 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1941 struct pcpu_dstats __percpu *dstats;
1942 struct pcpu_vstats __percpu *vstats;
a7855c78 1943 };
536721b1 1944
fb585b44 1945#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1946 struct garp_port __rcu *garp_port;
fb585b44
TK
1947#endif
1948#if IS_ENABLED(CONFIG_MRP)
febf018d 1949 struct mrp_port __rcu *mrp_port;
fb585b44 1950#endif
1da177e4 1951
5e82b4b2 1952 struct device dev;
0c509a6c 1953 const struct attribute_group *sysfs_groups[4];
a953be53 1954 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1955
38f7b870 1956 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1957
82cc1a7a
PWJ
1958 /* for setting kernel sock attribute on TCP connection setup */
1959#define GSO_MAX_SIZE 65536
1960 unsigned int gso_max_size;
30b678d8
BH
1961#define GSO_MAX_SEGS 65535
1962 u16 gso_max_segs;
743b03a8 1963
7a6b6f51 1964#ifdef CONFIG_DCB
32953543 1965 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1966#endif
5e82b4b2
BH
1967 u8 num_tc;
1968 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1969 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1970
d11ead75 1971#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1972 unsigned int fcoe_ddp_xid;
5bc1421e 1973#endif
86f8515f 1974#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1975 struct netprio_map __rcu *priomap;
4d288d57 1976#endif
5e82b4b2 1977 struct phy_device *phydev;
e679c9c1 1978 struct sfp_bus *sfp_bus;
5e82b4b2 1979 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1980 struct lock_class_key *qdisc_running_key;
5e82b4b2 1981 bool proto_down;
1da177e4 1982};
43cb76d9 1983#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1984
b5cdae32
DM
1985static inline bool netif_elide_gro(const struct net_device *dev)
1986{
1987 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1988 return true;
1989 return false;
1990}
1991
1da177e4 1992#define NETDEV_ALIGN 32
1da177e4 1993
4f57c087
JF
1994static inline
1995int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1996{
1997 return dev->prio_tc_map[prio & TC_BITMASK];
1998}
1999
2000static inline
2001int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2002{
2003 if (tc >= dev->num_tc)
2004 return -EINVAL;
2005
2006 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2007 return 0;
2008}
2009
8d059b0f 2010int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
2011void netdev_reset_tc(struct net_device *dev);
2012int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2013int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
2014
2015static inline
2016int netdev_get_num_tc(struct net_device *dev)
2017{
2018 return dev->num_tc;
2019}
2020
e8a0464c
DM
2021static inline
2022struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2023 unsigned int index)
2024{
2025 return &dev->_tx[index];
2026}
2027
10c51b56
DB
2028static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2029 const struct sk_buff *skb)
2030{
2031 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2032}
2033
e8a0464c
DM
2034static inline void netdev_for_each_tx_queue(struct net_device *dev,
2035 void (*f)(struct net_device *,
2036 struct netdev_queue *,
2037 void *),
2038 void *arg)
2039{
2040 unsigned int i;
2041
2042 for (i = 0; i < dev->num_tx_queues; i++)
2043 f(dev, &dev->_tx[i], arg);
2044}
2045
d3fff6c4
ED
2046#define netdev_lockdep_set_classes(dev) \
2047{ \
2048 static struct lock_class_key qdisc_tx_busylock_key; \
2049 static struct lock_class_key qdisc_running_key; \
2050 static struct lock_class_key qdisc_xmit_lock_key; \
2051 static struct lock_class_key dev_addr_list_lock_key; \
2052 unsigned int i; \
2053 \
2054 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2055 (dev)->qdisc_running_key = &qdisc_running_key; \
2056 lockdep_set_class(&(dev)->addr_list_lock, \
2057 &dev_addr_list_lock_key); \
2058 for (i = 0; i < (dev)->num_tx_queues; i++) \
2059 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2060 &qdisc_xmit_lock_key); \
2061}
2062
f629d208 2063struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
2064 struct sk_buff *skb,
2065 void *accel_priv);
8c4c49df 2066
871b642a
PA
2067/* returns the headroom that the master device needs to take in account
2068 * when forwarding to this dev
2069 */
2070static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2071{
2072 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2073}
2074
2075static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2076{
2077 if (dev->netdev_ops->ndo_set_rx_headroom)
2078 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2079}
2080
2081/* set the device rx headroom to the dev's default */
2082static inline void netdev_reset_rx_headroom(struct net_device *dev)
2083{
2084 netdev_set_rx_headroom(dev, -1);
2085}
2086
c346dca1
YH
2087/*
2088 * Net namespace inlines
2089 */
2090static inline
2091struct net *dev_net(const struct net_device *dev)
2092{
c2d9ba9b 2093 return read_pnet(&dev->nd_net);
c346dca1
YH
2094}
2095
2096static inline
f5aa23fd 2097void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2098{
0c5c9fb5 2099 write_pnet(&dev->nd_net, net);
c346dca1
YH
2100}
2101
bea3348e
SH
2102/**
2103 * netdev_priv - access network device private data
2104 * @dev: network device
2105 *
2106 * Get network device private data
2107 */
6472ce60 2108static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2109{
1ce8e7b5 2110 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2111}
2112
1da177e4
LT
2113/* Set the sysfs physical device reference for the network logical device
2114 * if set prior to registration will cause a symlink during initialization.
2115 */
43cb76d9 2116#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2117
384912ed 2118/* Set the sysfs device type for the network logical device to allow
3f79410c 2119 * fine-grained identification of different network device types. For
5e82b4b2 2120 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2121 */
2122#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2123
82dc3c63
ED
2124/* Default NAPI poll() weight
2125 * Device drivers are strongly advised to not use bigger value
2126 */
2127#define NAPI_POLL_WEIGHT 64
2128
3b582cc1 2129/**
5e82b4b2 2130 * netif_napi_add - initialize a NAPI context
3b582cc1 2131 * @dev: network device
5e82b4b2 2132 * @napi: NAPI context
3b582cc1
SH
2133 * @poll: polling function
2134 * @weight: default weight
2135 *
5e82b4b2
BH
2136 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2137 * *any* of the other NAPI-related functions.
3b582cc1 2138 */
d565b0a1
HX
2139void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2140 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2141
d64b5e85 2142/**
5e82b4b2 2143 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2144 * @dev: network device
5e82b4b2 2145 * @napi: NAPI context
d64b5e85
ED
2146 * @poll: polling function
2147 * @weight: default weight
2148 *
2149 * This variant of netif_napi_add() should be used from drivers using NAPI
2150 * to exclusively poll a TX queue.
2151 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2152 */
2153static inline void netif_tx_napi_add(struct net_device *dev,
2154 struct napi_struct *napi,
2155 int (*poll)(struct napi_struct *, int),
2156 int weight)
2157{
2158 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2159 netif_napi_add(dev, napi, poll, weight);
2160}
2161
d8156534 2162/**
5e82b4b2
BH
2163 * netif_napi_del - remove a NAPI context
2164 * @napi: NAPI context
d8156534 2165 *
5e82b4b2 2166 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2167 */
d565b0a1
HX
2168void netif_napi_del(struct napi_struct *napi);
2169
2170struct napi_gro_cb {
78a478d0 2171 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2172 void *frag0;
78a478d0 2173
7489594c
HX
2174 /* Length of frag0. */
2175 unsigned int frag0_len;
2176
86911732 2177 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2178 int data_offset;
86911732 2179
d565b0a1 2180 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2181 u16 flush;
2182
2183 /* Save the IP ID here and check when we get to the transport layer */
2184 u16 flush_id;
d565b0a1
HX
2185
2186 /* Number of segments aggregated. */
2e71a6f8
ED
2187 u16 count;
2188
15e2396d
TH
2189 /* Start offset for remote checksum offload */
2190 u16 gro_remcsum_start;
2191
2e71a6f8
ED
2192 /* jiffies when first packet was created/queued */
2193 unsigned long age;
86347245 2194
afe93325 2195 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2196 u16 proto;
2197
baa32ff4
TH
2198 /* This is non-zero if the packet may be of the same flow. */
2199 u8 same_flow:1;
2200
fac8e0f5
JG
2201 /* Used in tunnel GRO receive */
2202 u8 encap_mark:1;
573e8fca
TH
2203
2204 /* GRO checksum is valid */
2205 u8 csum_valid:1;
2206
662880f4
TH
2207 /* Number of checksums via CHECKSUM_UNNECESSARY */
2208 u8 csum_cnt:3;
c3c7c254 2209
baa32ff4
TH
2210 /* Free the skb? */
2211 u8 free:2;
2212#define NAPI_GRO_FREE 1
2213#define NAPI_GRO_FREE_STOLEN_HEAD 2
2214
efc98d08
TH
2215 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2216 u8 is_ipv6:1;
2217
a0ca153f
AD
2218 /* Used in GRE, set in fou/gue_gro_receive */
2219 u8 is_fou:1;
2220
1530545e
AD
2221 /* Used to determine if flush_id can be ignored */
2222 u8 is_atomic:1;
2223
fcd91dd4
SD
2224 /* Number of gro_receive callbacks this packet already went through */
2225 u8 recursion_counter:4;
2226
2227 /* 1 bit hole */
baa32ff4 2228
bf5a755f
JC
2229 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2230 __wsum csum;
2231
c3c7c254
ED
2232 /* used in skb_gro_receive() slow path */
2233 struct sk_buff *last;
d565b0a1
HX
2234};
2235
2236#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2237
fcd91dd4
SD
2238#define GRO_RECURSION_LIMIT 15
2239static inline int gro_recursion_inc_test(struct sk_buff *skb)
2240{
2241 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2242}
2243
2244typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2245static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2246 struct sk_buff **head,
2247 struct sk_buff *skb)
2248{
2249 if (unlikely(gro_recursion_inc_test(skb))) {
2250 NAPI_GRO_CB(skb)->flush |= 1;
2251 return NULL;
2252 }
2253
2254 return cb(head, skb);
2255}
2256
2257typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2258 struct sk_buff *);
2259static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2260 struct sock *sk,
2261 struct sk_buff **head,
2262 struct sk_buff *skb)
2263{
2264 if (unlikely(gro_recursion_inc_test(skb))) {
2265 NAPI_GRO_CB(skb)->flush |= 1;
2266 return NULL;
2267 }
2268
2269 return cb(sk, head, skb);
2270}
2271
1da177e4 2272struct packet_type {
f2ccd8fa
DM
2273 __be16 type; /* This is really htons(ether_type). */
2274 struct net_device *dev; /* NULL is wildcarded here */
2275 int (*func) (struct sk_buff *,
2276 struct net_device *,
2277 struct packet_type *,
2278 struct net_device *);
c0de08d0
EL
2279 bool (*id_match)(struct packet_type *ptype,
2280 struct sock *sk);
1da177e4
LT
2281 void *af_packet_priv;
2282 struct list_head list;
2283};
2284
f191a1d1 2285struct offload_callbacks {
576a30eb 2286 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2287 netdev_features_t features);
d565b0a1 2288 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2289 struct sk_buff *skb);
299603e8 2290 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2291};
2292
2293struct packet_offload {
2294 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2295 u16 priority;
f191a1d1
VY
2296 struct offload_callbacks callbacks;
2297 struct list_head list;
1da177e4
LT
2298};
2299
5e82b4b2 2300/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2301struct pcpu_sw_netstats {
2302 u64 rx_packets;
2303 u64 rx_bytes;
2304 u64 tx_packets;
2305 u64 tx_bytes;
2306 struct u64_stats_sync syncp;
2307};
2308
aabc92bb
PNA
2309#define __netdev_alloc_pcpu_stats(type, gfp) \
2310({ \
2311 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2312 if (pcpu_stats) { \
2313 int __cpu; \
2314 for_each_possible_cpu(__cpu) { \
2315 typeof(type) *stat; \
2316 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2317 u64_stats_init(&stat->syncp); \
2318 } \
2319 } \
2320 pcpu_stats; \
1c213bd2
WC
2321})
2322
aabc92bb 2323#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2324 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2325
764f5e54
JP
2326enum netdev_lag_tx_type {
2327 NETDEV_LAG_TX_TYPE_UNKNOWN,
2328 NETDEV_LAG_TX_TYPE_RANDOM,
2329 NETDEV_LAG_TX_TYPE_BROADCAST,
2330 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2331 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2332 NETDEV_LAG_TX_TYPE_HASH,
2333};
2334
f44aa9ef
JH
2335enum netdev_lag_hash {
2336 NETDEV_LAG_HASH_NONE,
2337 NETDEV_LAG_HASH_L2,
2338 NETDEV_LAG_HASH_L34,
2339 NETDEV_LAG_HASH_L23,
2340 NETDEV_LAG_HASH_E23,
2341 NETDEV_LAG_HASH_E34,
2342 NETDEV_LAG_HASH_UNKNOWN,
2343};
2344
764f5e54
JP
2345struct netdev_lag_upper_info {
2346 enum netdev_lag_tx_type tx_type;
f44aa9ef 2347 enum netdev_lag_hash hash_type;
764f5e54
JP
2348};
2349
fb1b2e3c
JP
2350struct netdev_lag_lower_state_info {
2351 u8 link_up : 1,
2352 tx_enabled : 1;
2353};
2354
1da177e4
LT
2355#include <linux/notifier.h>
2356
ede2762d
KT
2357/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2358 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2359 * adding new types.
dcfe1421 2360 */
ede2762d
KT
2361enum netdev_cmd {
2362 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2363 NETDEV_DOWN,
2364 NETDEV_REBOOT, /* Tell a protocol stack a network interface
dcfe1421
AW
2365 detected a hardware crash and restarted
2366 - we can use this eg to kick tcp sessions
2367 once done */
ede2762d
KT
2368 NETDEV_CHANGE, /* Notify device state change */
2369 NETDEV_REGISTER,
2370 NETDEV_UNREGISTER,
2371 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2372 NETDEV_CHANGEADDR,
2373 NETDEV_GOING_DOWN,
2374 NETDEV_CHANGENAME,
2375 NETDEV_FEAT_CHANGE,
2376 NETDEV_BONDING_FAILOVER,
2377 NETDEV_PRE_UP,
2378 NETDEV_PRE_TYPE_CHANGE,
2379 NETDEV_POST_TYPE_CHANGE,
2380 NETDEV_POST_INIT,
ede2762d
KT
2381 NETDEV_RELEASE,
2382 NETDEV_NOTIFY_PEERS,
2383 NETDEV_JOIN,
2384 NETDEV_CHANGEUPPER,
2385 NETDEV_RESEND_IGMP,
2386 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2387 NETDEV_CHANGEINFODATA,
2388 NETDEV_BONDING_INFO,
2389 NETDEV_PRECHANGEUPPER,
2390 NETDEV_CHANGELOWERSTATE,
2391 NETDEV_UDP_TUNNEL_PUSH_INFO,
2392 NETDEV_UDP_TUNNEL_DROP_INFO,
2393 NETDEV_CHANGE_TX_QUEUE_LEN,
9daae9bd
GP
2394 NETDEV_CVLAN_FILTER_PUSH_INFO,
2395 NETDEV_CVLAN_FILTER_DROP_INFO,
2396 NETDEV_SVLAN_FILTER_PUSH_INFO,
2397 NETDEV_SVLAN_FILTER_DROP_INFO,
ede2762d
KT
2398};
2399const char *netdev_cmd_to_name(enum netdev_cmd cmd);
dcfe1421 2400
f629d208
JP
2401int register_netdevice_notifier(struct notifier_block *nb);
2402int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2403
2404struct netdev_notifier_info {
51d0c047
DA
2405 struct net_device *dev;
2406 struct netlink_ext_ack *extack;
351638e7
JP
2407};
2408
be9efd36
JP
2409struct netdev_notifier_change_info {
2410 struct netdev_notifier_info info; /* must be first */
2411 unsigned int flags_changed;
2412};
2413
0e4ead9d
JP
2414struct netdev_notifier_changeupper_info {
2415 struct netdev_notifier_info info; /* must be first */
2416 struct net_device *upper_dev; /* new upper dev */
2417 bool master; /* is upper dev master */
5e82b4b2 2418 bool linking; /* is the notification for link or unlink */
29bf24af 2419 void *upper_info; /* upper dev info */
0e4ead9d
JP
2420};
2421
04d48266
JP
2422struct netdev_notifier_changelowerstate_info {
2423 struct netdev_notifier_info info; /* must be first */
2424 void *lower_state_info; /* is lower dev state */
2425};
2426
75538c2b
CW
2427static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2428 struct net_device *dev)
2429{
2430 info->dev = dev;
51d0c047 2431 info->extack = NULL;
75538c2b
CW
2432}
2433
351638e7
JP
2434static inline struct net_device *
2435netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2436{
2437 return info->dev;
2438}
2439
51d0c047
DA
2440static inline struct netlink_ext_ack *
2441netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2442{
2443 return info->extack;
2444}
2445
f629d208 2446int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2447
2448
1da177e4
LT
2449extern rwlock_t dev_base_lock; /* Device list lock */
2450
881d966b
EB
2451#define for_each_netdev(net, d) \
2452 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2453#define for_each_netdev_reverse(net, d) \
2454 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2455#define for_each_netdev_rcu(net, d) \
2456 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2457#define for_each_netdev_safe(net, d, n) \
2458 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2459#define for_each_netdev_continue(net, d) \
2460 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2461#define for_each_netdev_continue_rcu(net, d) \
2462 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2463#define for_each_netdev_in_bond_rcu(bond, slave) \
2464 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2465 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2466#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2467
a050c33f
DL
2468static inline struct net_device *next_net_device(struct net_device *dev)
2469{
2470 struct list_head *lh;
2471 struct net *net;
2472
c346dca1 2473 net = dev_net(dev);
a050c33f
DL
2474 lh = dev->dev_list.next;
2475 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2476}
2477
ce81b76a
ED
2478static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2479{
2480 struct list_head *lh;
2481 struct net *net;
2482
2483 net = dev_net(dev);
ccf43438 2484 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2485 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2486}
2487
a050c33f
DL
2488static inline struct net_device *first_net_device(struct net *net)
2489{
2490 return list_empty(&net->dev_base_head) ? NULL :
2491 net_device_entry(net->dev_base_head.next);
2492}
7562f876 2493
ccf43438
ED
2494static inline struct net_device *first_net_device_rcu(struct net *net)
2495{
2496 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2497
2498 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2499}
2500
f629d208
JP
2501int netdev_boot_setup_check(struct net_device *dev);
2502unsigned long netdev_boot_base(const char *prefix, int unit);
2503struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2504 const char *hwaddr);
2505struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2506struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2507void dev_add_pack(struct packet_type *pt);
2508void dev_remove_pack(struct packet_type *pt);
2509void __dev_remove_pack(struct packet_type *pt);
2510void dev_add_offload(struct packet_offload *po);
2511void dev_remove_offload(struct packet_offload *po);
f629d208 2512
a54acb3a 2513int dev_get_iflink(const struct net_device *dev);
fc4099f1 2514int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2515struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2516 unsigned short mask);
f629d208
JP
2517struct net_device *dev_get_by_name(struct net *net, const char *name);
2518struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2519struct net_device *__dev_get_by_name(struct net *net, const char *name);
2520int dev_alloc_name(struct net_device *dev, const char *name);
2521int dev_open(struct net_device *dev);
7051b88a 2522void dev_close(struct net_device *dev);
2523void dev_close_many(struct list_head *head, bool unlink);
f629d208 2524void dev_disable_lro(struct net_device *dev);
0c4b51f0 2525int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2526int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2527int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
865b03f2 2528int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
f629d208
JP
2529int register_netdevice(struct net_device *dev);
2530void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2531void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2532static inline void unregister_netdevice(struct net_device *dev)
2533{
2534 unregister_netdevice_queue(dev, NULL);
2535}
2536
f629d208
JP
2537int netdev_refcnt_read(const struct net_device *dev);
2538void free_netdev(struct net_device *dev);
74d332c1 2539void netdev_freemem(struct net_device *dev);
f629d208
JP
2540void synchronize_net(void);
2541int init_dummy_netdev(struct net_device *dev);
937f1ba5 2542
f60e5990 2543DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2544#define XMIT_RECURSION_LIMIT 10
2545
f60e5990 2546static inline int dev_recursion_level(void)
2547{
2548 return this_cpu_read(xmit_recursion);
2549}
2550
f629d208
JP
2551struct net_device *dev_get_by_index(struct net *net, int ifindex);
2552struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2553struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2554struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2555int netdev_get_name(struct net *net, char *name, int ifindex);
2556int dev_restart(struct net_device *dev);
f629d208 2557int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2558
2559static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2560{
2561 return NAPI_GRO_CB(skb)->data_offset;
2562}
2563
2564static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2565{
2566 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2567}
2568
2569static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2570{
2571 NAPI_GRO_CB(skb)->data_offset += len;
2572}
2573
a5b1cf28
HX
2574static inline void *skb_gro_header_fast(struct sk_buff *skb,
2575 unsigned int offset)
86911732 2576{
a5b1cf28
HX
2577 return NAPI_GRO_CB(skb)->frag0 + offset;
2578}
78a478d0 2579
a5b1cf28
HX
2580static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2581{
2582 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2583}
78a478d0 2584
57ea52a8
HX
2585static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2586{
2587 NAPI_GRO_CB(skb)->frag0 = NULL;
2588 NAPI_GRO_CB(skb)->frag0_len = 0;
2589}
2590
a5b1cf28
HX
2591static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2592 unsigned int offset)
2593{
17dd759c
HX
2594 if (!pskb_may_pull(skb, hlen))
2595 return NULL;
2596
57ea52a8 2597 skb_gro_frag0_invalidate(skb);
17dd759c 2598 return skb->data + offset;
86911732 2599}
1da177e4 2600
36e7b1b8
HX
2601static inline void *skb_gro_network_header(struct sk_buff *skb)
2602{
78d3fd0b
HX
2603 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2604 skb_network_offset(skb);
36e7b1b8
HX
2605}
2606
bf5a755f
JC
2607static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2608 const void *start, unsigned int len)
2609{
573e8fca 2610 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2611 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2612 csum_partial(start, len, 0));
2613}
2614
573e8fca
TH
2615/* GRO checksum functions. These are logical equivalents of the normal
2616 * checksum functions (in skbuff.h) except that they operate on the GRO
2617 * offsets and fields in sk_buff.
2618 */
2619
2620__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2621
15e2396d
TH
2622static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2623{
b7fe10e5 2624 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2625}
2626
573e8fca
TH
2627static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2628 bool zero_okay,
2629 __sum16 check)
2630{
6edec0e6
TH
2631 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2632 skb_checksum_start_offset(skb) <
2633 skb_gro_offset(skb)) &&
15e2396d 2634 !skb_at_gro_remcsum_start(skb) &&
662880f4 2635 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2636 (!zero_okay || check));
2637}
2638
2639static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2640 __wsum psum)
2641{
2642 if (NAPI_GRO_CB(skb)->csum_valid &&
2643 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2644 return 0;
2645
2646 NAPI_GRO_CB(skb)->csum = psum;
2647
2648 return __skb_gro_checksum_complete(skb);
2649}
2650
573e8fca
TH
2651static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2652{
662880f4
TH
2653 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2654 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2655 NAPI_GRO_CB(skb)->csum_cnt--;
2656 } else {
2657 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2658 * verified a new top level checksum or an encapsulated one
2659 * during GRO. This saves work if we fallback to normal path.
2660 */
2661 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2662 }
2663}
2664
2665#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2666 compute_pseudo) \
2667({ \
2668 __sum16 __ret = 0; \
2669 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2670 __ret = __skb_gro_checksum_validate_complete(skb, \
2671 compute_pseudo(skb, proto)); \
219f1d79 2672 if (!__ret) \
573e8fca
TH
2673 skb_gro_incr_csum_unnecessary(skb); \
2674 __ret; \
2675})
2676
2677#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2678 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2679
2680#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2681 compute_pseudo) \
2682 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2683
2684#define skb_gro_checksum_simple_validate(skb) \
2685 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2686
d96535a1
TH
2687static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2688{
2689 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2690 !NAPI_GRO_CB(skb)->csum_valid);
2691}
2692
2693static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2694 __sum16 check, __wsum pseudo)
2695{
2696 NAPI_GRO_CB(skb)->csum = ~pseudo;
2697 NAPI_GRO_CB(skb)->csum_valid = 1;
2698}
2699
2700#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2701do { \
2702 if (__skb_gro_checksum_convert_check(skb)) \
2703 __skb_gro_checksum_convert(skb, check, \
2704 compute_pseudo(skb, proto)); \
2705} while (0)
2706
26c4f7da
TH
2707struct gro_remcsum {
2708 int offset;
2709 __wsum delta;
2710};
2711
2712static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2713{
846cd667 2714 grc->offset = 0;
26c4f7da
TH
2715 grc->delta = 0;
2716}
2717
b7fe10e5
TH
2718static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2719 unsigned int off, size_t hdrlen,
2720 int start, int offset,
2721 struct gro_remcsum *grc,
2722 bool nopartial)
dcdc8994
TH
2723{
2724 __wsum delta;
b7fe10e5 2725 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2726
2727 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2728
15e2396d 2729 if (!nopartial) {
b7fe10e5
TH
2730 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2731 return ptr;
2732 }
2733
2734 ptr = skb_gro_header_fast(skb, off);
2735 if (skb_gro_header_hard(skb, off + plen)) {
2736 ptr = skb_gro_header_slow(skb, off + plen, off);
2737 if (!ptr)
2738 return NULL;
15e2396d
TH
2739 }
2740
b7fe10e5
TH
2741 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2742 start, offset);
dcdc8994
TH
2743
2744 /* Adjust skb->csum since we changed the packet */
dcdc8994 2745 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2746
b7fe10e5 2747 grc->offset = off + hdrlen + offset;
26c4f7da 2748 grc->delta = delta;
b7fe10e5
TH
2749
2750 return ptr;
dcdc8994
TH
2751}
2752
26c4f7da
TH
2753static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2754 struct gro_remcsum *grc)
2755{
b7fe10e5
TH
2756 void *ptr;
2757 size_t plen = grc->offset + sizeof(u16);
2758
26c4f7da
TH
2759 if (!grc->delta)
2760 return;
2761
b7fe10e5
TH
2762 ptr = skb_gro_header_fast(skb, grc->offset);
2763 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2764 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2765 if (!ptr)
2766 return;
2767 }
2768
2769 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2770}
dcdc8994 2771
25393d3f
SK
2772#ifdef CONFIG_XFRM_OFFLOAD
2773static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2774{
2775 if (PTR_ERR(pp) != -EINPROGRESS)
2776 NAPI_GRO_CB(skb)->flush |= flush;
2777}
2778#else
5f114163
SK
2779static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2780{
2781 NAPI_GRO_CB(skb)->flush |= flush;
2782}
25393d3f 2783#endif
5f114163 2784
0c4e8581
SH
2785static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2786 unsigned short type,
3b04ddde 2787 const void *daddr, const void *saddr,
95c96174 2788 unsigned int len)
0c4e8581 2789{
f1ecfd5d 2790 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2791 return 0;
3b04ddde
SH
2792
2793 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2794}
2795
b95cce35
SH
2796static inline int dev_parse_header(const struct sk_buff *skb,
2797 unsigned char *haddr)
2798{
2799 const struct net_device *dev = skb->dev;
2800
1b83336b 2801 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2802 return 0;
3b04ddde 2803 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2804}
2805
2793a23a
WB
2806/* ll_header must have at least hard_header_len allocated */
2807static inline bool dev_validate_header(const struct net_device *dev,
2808 char *ll_header, int len)
2809{
2810 if (likely(len >= dev->hard_header_len))
2811 return true;
217e6fa2
WB
2812 if (len < dev->min_header_len)
2813 return false;
2793a23a
WB
2814
2815 if (capable(CAP_SYS_RAWIO)) {
2816 memset(ll_header + len, 0, dev->hard_header_len - len);
2817 return true;
2818 }
2819
2820 if (dev->header_ops && dev->header_ops->validate)
2821 return dev->header_ops->validate(ll_header, len);
2822
2823 return false;
2824}
2825
36fd633e
AV
2826typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2827 int len, int size);
f629d208 2828int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2829static inline int unregister_gifconf(unsigned int family)
2830{
2831 return register_gifconf(family, NULL);
2832}
2833
99bbc707 2834#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2835#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2836struct sd_flow_limit {
2837 u64 count;
2838 unsigned int num_buckets;
2839 unsigned int history_head;
2840 u16 history[FLOW_LIMIT_HISTORY];
2841 u8 buckets[];
2842};
2843
2844extern int netdev_flow_limit_table_len;
2845#endif /* CONFIG_NET_FLOW_LIMIT */
2846
1da177e4 2847/*
5e82b4b2 2848 * Incoming packets are placed on per-CPU queues
1da177e4 2849 */
d94d9fee 2850struct softnet_data {
1da177e4 2851 struct list_head poll_list;
6e7676c1 2852 struct sk_buff_head process_queue;
1da177e4 2853
dee42870 2854 /* stats */
cd7b5396
DM
2855 unsigned int processed;
2856 unsigned int time_squeeze;
cd7b5396 2857 unsigned int received_rps;
fd793d89 2858#ifdef CONFIG_RPS
88751275 2859 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2860#endif
2861#ifdef CONFIG_NET_FLOW_LIMIT
2862 struct sd_flow_limit __rcu *flow_limit;
2863#endif
2864 struct Qdisc *output_queue;
2865 struct Qdisc **output_queue_tailp;
2866 struct sk_buff *completion_queue;
f53c7239
SK
2867#ifdef CONFIG_XFRM_OFFLOAD
2868 struct sk_buff_head xfrm_backlog;
2869#endif
4cdb1e2e 2870#ifdef CONFIG_RPS
501e7ef5
ED
2871 /* input_queue_head should be written by cpu owning this struct,
2872 * and only read by other cpus. Worth using a cache line.
2873 */
2874 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2875
2876 /* Elements below can be accessed between CPUs for RPS/RFS */
966a9671 2877 call_single_data_t csd ____cacheline_aligned_in_smp;
88751275
ED
2878 struct softnet_data *rps_ipi_next;
2879 unsigned int cpu;
76cc8b13 2880 unsigned int input_queue_tail;
1e94d72f 2881#endif
95c96174 2882 unsigned int dropped;
0a9627f2 2883 struct sk_buff_head input_pkt_queue;
bea3348e 2884 struct napi_struct backlog;
99bbc707 2885
1da177e4
LT
2886};
2887
76cc8b13 2888static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2889{
2890#ifdef CONFIG_RPS
76cc8b13
TH
2891 sd->input_queue_head++;
2892#endif
2893}
2894
2895static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2896 unsigned int *qtail)
2897{
2898#ifdef CONFIG_RPS
2899 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2900#endif
2901}
2902
0a9627f2 2903DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2904
f629d208 2905void __netif_schedule(struct Qdisc *q);
46e5da40 2906void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2907
fd2ea0a7
DM
2908static inline void netif_tx_schedule_all(struct net_device *dev)
2909{
2910 unsigned int i;
2911
2912 for (i = 0; i < dev->num_tx_queues; i++)
2913 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2914}
2915
f9a7cbbf 2916static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2917{
73466498 2918 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2919}
2920
bea3348e
SH
2921/**
2922 * netif_start_queue - allow transmit
2923 * @dev: network device
2924 *
2925 * Allow upper layers to call the device hard_start_xmit routine.
2926 */
1da177e4
LT
2927static inline void netif_start_queue(struct net_device *dev)
2928{
e8a0464c 2929 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2930}
2931
fd2ea0a7
DM
2932static inline void netif_tx_start_all_queues(struct net_device *dev)
2933{
2934 unsigned int i;
2935
2936 for (i = 0; i < dev->num_tx_queues; i++) {
2937 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2938 netif_tx_start_queue(txq);
2939 }
2940}
2941
46e5da40 2942void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2943
d29f749e
DJ
2944/**
2945 * netif_wake_queue - restart transmit
2946 * @dev: network device
2947 *
2948 * Allow upper layers to call the device hard_start_xmit routine.
2949 * Used for flow control when transmit resources are available.
2950 */
79d16385
DM
2951static inline void netif_wake_queue(struct net_device *dev)
2952{
e8a0464c 2953 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2954}
2955
fd2ea0a7
DM
2956static inline void netif_tx_wake_all_queues(struct net_device *dev)
2957{
2958 unsigned int i;
2959
2960 for (i = 0; i < dev->num_tx_queues; i++) {
2961 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2962 netif_tx_wake_queue(txq);
2963 }
2964}
2965
f9a7cbbf 2966static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2967{
73466498 2968 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2969}
2970
bea3348e
SH
2971/**
2972 * netif_stop_queue - stop transmitted packets
2973 * @dev: network device
2974 *
2975 * Stop upper layers calling the device hard_start_xmit routine.
2976 * Used for flow control when transmit resources are unavailable.
2977 */
1da177e4
LT
2978static inline void netif_stop_queue(struct net_device *dev)
2979{
e8a0464c 2980 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2981}
2982
a2029240 2983void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2984
4d29515f 2985static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2986{
73466498 2987 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2988}
2989
bea3348e
SH
2990/**
2991 * netif_queue_stopped - test if transmit queue is flowblocked
2992 * @dev: network device
2993 *
2994 * Test if transmit queue on device is currently unable to send.
2995 */
4d29515f 2996static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2997{
e8a0464c 2998 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2999}
3000
4d29515f 3001static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 3002{
73466498
TH
3003 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3004}
3005
8e2f1a63
DB
3006static inline bool
3007netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
3008{
3009 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3010}
3011
8e2f1a63
DB
3012static inline bool
3013netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3014{
3015 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3016}
3017
53511453
ED
3018/**
3019 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3020 * @dev_queue: pointer to transmit queue
3021 *
3022 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 3023 * to give appropriate hint to the CPU.
53511453
ED
3024 */
3025static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3026{
3027#ifdef CONFIG_BQL
3028 prefetchw(&dev_queue->dql.num_queued);
3029#endif
3030}
3031
3032/**
3033 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3034 * @dev_queue: pointer to transmit queue
3035 *
3036 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 3037 * to give appropriate hint to the CPU.
53511453
ED
3038 */
3039static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3040{
3041#ifdef CONFIG_BQL
3042 prefetchw(&dev_queue->dql.limit);
3043#endif
3044}
3045
c5d67bd7
TH
3046static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3047 unsigned int bytes)
3048{
114cf580
TH
3049#ifdef CONFIG_BQL
3050 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
3051
3052 if (likely(dql_avail(&dev_queue->dql) >= 0))
3053 return;
3054
3055 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3056
3057 /*
3058 * The XOFF flag must be set before checking the dql_avail below,
3059 * because in netdev_tx_completed_queue we update the dql_completed
3060 * before checking the XOFF flag.
3061 */
3062 smp_mb();
3063
3064 /* check again in case another CPU has just made room avail */
3065 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3066 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 3067#endif
c5d67bd7
TH
3068}
3069
0042d0c8
FF
3070/**
3071 * netdev_sent_queue - report the number of bytes queued to hardware
3072 * @dev: network device
3073 * @bytes: number of bytes queued to the hardware device queue
3074 *
3075 * Report the number of bytes queued for sending/completion to the network
3076 * device hardware queue. @bytes should be a good approximation and should
3077 * exactly match netdev_completed_queue() @bytes
3078 */
c5d67bd7
TH
3079static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3080{
3081 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3082}
3083
3084static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 3085 unsigned int pkts, unsigned int bytes)
c5d67bd7 3086{
114cf580 3087#ifdef CONFIG_BQL
b37c0fbe
AD
3088 if (unlikely(!bytes))
3089 return;
3090
3091 dql_completed(&dev_queue->dql, bytes);
3092
3093 /*
3094 * Without the memory barrier there is a small possiblity that
3095 * netdev_tx_sent_queue will miss the update and cause the queue to
3096 * be stopped forever
3097 */
3098 smp_mb();
3099
3100 if (dql_avail(&dev_queue->dql) < 0)
3101 return;
3102
3103 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3104 netif_schedule_queue(dev_queue);
114cf580 3105#endif
c5d67bd7
TH
3106}
3107
0042d0c8
FF
3108/**
3109 * netdev_completed_queue - report bytes and packets completed by device
3110 * @dev: network device
3111 * @pkts: actual number of packets sent over the medium
3112 * @bytes: actual number of bytes sent over the medium
3113 *
3114 * Report the number of bytes and packets transmitted by the network device
3115 * hardware queue over the physical medium, @bytes must exactly match the
3116 * @bytes amount passed to netdev_sent_queue()
3117 */
c5d67bd7 3118static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3119 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3120{
3121 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3122}
3123
3124static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3125{
114cf580 3126#ifdef CONFIG_BQL
5c490354 3127 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3128 dql_reset(&q->dql);
3129#endif
c5d67bd7
TH
3130}
3131
0042d0c8
FF
3132/**
3133 * netdev_reset_queue - reset the packets and bytes count of a network device
3134 * @dev_queue: network device
3135 *
3136 * Reset the bytes and packet count of a network device and clear the
3137 * software flow control OFF bit for this network device
3138 */
c5d67bd7
TH
3139static inline void netdev_reset_queue(struct net_device *dev_queue)
3140{
3141 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3142}
3143
b9507bda
DB
3144/**
3145 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3146 * @dev: network device
3147 * @queue_index: given tx queue index
3148 *
3149 * Returns 0 if given tx queue index >= number of device tx queues,
3150 * otherwise returns the originally passed tx queue index.
3151 */
3152static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3153{
3154 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3155 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3156 dev->name, queue_index,
3157 dev->real_num_tx_queues);
3158 return 0;
3159 }
3160
3161 return queue_index;
3162}
3163
bea3348e
SH
3164/**
3165 * netif_running - test if up
3166 * @dev: network device
3167 *
3168 * Test if the device has been brought up.
3169 */
4d29515f 3170static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3171{
3172 return test_bit(__LINK_STATE_START, &dev->state);
3173}
3174
f25f4e44 3175/*
5e82b4b2 3176 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3177 * stop, and a check if it's stopped. All other device management is
3178 * done at the overall netdevice level.
3179 * Also test the device if we're multiqueue.
3180 */
bea3348e
SH
3181
3182/**
3183 * netif_start_subqueue - allow sending packets on subqueue
3184 * @dev: network device
3185 * @queue_index: sub queue index
3186 *
3187 * Start individual transmit queue of a device with multiple transmit queues.
3188 */
f25f4e44
PWJ
3189static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3190{
fd2ea0a7 3191 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3192
3193 netif_tx_start_queue(txq);
f25f4e44
PWJ
3194}
3195
bea3348e
SH
3196/**
3197 * netif_stop_subqueue - stop sending packets on subqueue
3198 * @dev: network device
3199 * @queue_index: sub queue index
3200 *
3201 * Stop individual transmit queue of a device with multiple transmit queues.
3202 */
f25f4e44
PWJ
3203static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3204{
fd2ea0a7 3205 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3206 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3207}
3208
bea3348e
SH
3209/**
3210 * netif_subqueue_stopped - test status of subqueue
3211 * @dev: network device
3212 * @queue_index: sub queue index
3213 *
3214 * Check individual transmit queue of a device with multiple transmit queues.
3215 */
4d29515f
DM
3216static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3217 u16 queue_index)
f25f4e44 3218{
fd2ea0a7 3219 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3220
3221 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3222}
3223
4d29515f
DM
3224static inline bool netif_subqueue_stopped(const struct net_device *dev,
3225 struct sk_buff *skb)
668f895a
PE
3226{
3227 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3228}
bea3348e 3229
738b35cc
FF
3230/**
3231 * netif_wake_subqueue - allow sending packets on subqueue
3232 * @dev: network device
3233 * @queue_index: sub queue index
3234 *
3235 * Resume individual transmit queue of a device with multiple transmit queues.
3236 */
3237static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3238{
3239 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3240
3241 netif_tx_wake_queue(txq);
3242}
f25f4e44 3243
537c00de 3244#ifdef CONFIG_XPS
53af53ae 3245int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3246 u16 index);
537c00de
AD
3247#else
3248static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3249 const struct cpumask *mask,
537c00de
AD
3250 u16 index)
3251{
3252 return 0;
3253}
3254#endif
3255
bea3348e
SH
3256/**
3257 * netif_is_multiqueue - test if device has multiple transmit queues
3258 * @dev: network device
3259 *
3260 * Check if device has multiple transmit queues
bea3348e 3261 */
4d29515f 3262static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3263{
a02cec21 3264 return dev->num_tx_queues > 1;
f25f4e44 3265}
1da177e4 3266
f629d208 3267int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3268
a953be53 3269#ifdef CONFIG_SYSFS
f629d208 3270int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3271#else
3272static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3273 unsigned int rxq)
3274{
3275 return 0;
3276}
3277#endif
3278
65073a67
DB
3279static inline struct netdev_rx_queue *
3280__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3281{
3282 return dev->_rx + rxq;
3283}
3284
a953be53
MD
3285#ifdef CONFIG_SYSFS
3286static inline unsigned int get_netdev_rx_queue_index(
3287 struct netdev_rx_queue *queue)
3288{
3289 struct net_device *dev = queue->dev;
3290 int index = queue - dev->_rx;
3291
3292 BUG_ON(index >= dev->num_rx_queues);
3293 return index;
3294}
3295#endif
3296
16917b87 3297#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3298int netif_get_num_default_rss_queues(void);
16917b87 3299
e6247027
ED
3300enum skb_free_reason {
3301 SKB_REASON_CONSUMED,
3302 SKB_REASON_DROPPED,
3303};
3304
3305void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3306void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3307
e6247027
ED
3308/*
3309 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3310 * interrupt context or with hardware interrupts being disabled.
3311 * (in_irq() || irqs_disabled())
3312 *
3313 * We provide four helpers that can be used in following contexts :
3314 *
3315 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3316 * replacing kfree_skb(skb)
3317 *
3318 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3319 * Typically used in place of consume_skb(skb) in TX completion path
3320 *
3321 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3322 * replacing kfree_skb(skb)
3323 *
3324 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3325 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3326 */
e6247027
ED
3327static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3328{
3329 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3330}
3331
3332static inline void dev_consume_skb_irq(struct sk_buff *skb)
3333{
3334 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3335}
3336
3337static inline void dev_kfree_skb_any(struct sk_buff *skb)
3338{
3339 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3340}
3341
3342static inline void dev_consume_skb_any(struct sk_buff *skb)
3343{
3344 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3345}
1da177e4 3346
7c497478
JW
3347void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3348int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
f629d208
JP
3349int netif_rx(struct sk_buff *skb);
3350int netif_rx_ni(struct sk_buff *skb);
04eb4489 3351int netif_receive_skb(struct sk_buff *skb);
1c601d82 3352int netif_receive_skb_core(struct sk_buff *skb);
f629d208
JP
3353gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3354void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3355struct sk_buff *napi_get_frags(struct napi_struct *napi);
3356gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3357struct packet_offload *gro_find_receive_by_type(__be16 type);
3358struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3359
3360static inline void napi_free_frags(struct napi_struct *napi)
3361{
3362 kfree_skb(napi->skb);
3363 napi->skb = NULL;
3364}
3365
24b27fc4 3366bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3367int netdev_rx_handler_register(struct net_device *dev,
3368 rx_handler_func_t *rx_handler,
3369 void *rx_handler_data);
3370void netdev_rx_handler_unregister(struct net_device *dev);
3371
3372bool dev_valid_name(const char *name);
44c02a2c
AV
3373int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3374 bool *need_copyout);
36fd633e 3375int dev_ifconf(struct net *net, struct ifconf *, int);
f629d208
JP
3376int dev_ethtool(struct net *net, struct ifreq *);
3377unsigned int dev_get_flags(const struct net_device *);
3378int __dev_change_flags(struct net_device *, unsigned int flags);
3379int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3380void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3381 unsigned int gchanges);
f629d208
JP
3382int dev_change_name(struct net_device *, const char *);
3383int dev_set_alias(struct net_device *, const char *, size_t);
6c557001 3384int dev_get_alias(const struct net_device *, char *, size_t);
f629d208 3385int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3386int __dev_set_mtu(struct net_device *, int);
f629d208 3387int dev_set_mtu(struct net_device *, int);
6a643ddb 3388int dev_change_tx_queue_len(struct net_device *, unsigned long);
f629d208
JP
3389void dev_set_group(struct net_device *, int);
3390int dev_set_mac_address(struct net_device *, struct sockaddr *);
3391int dev_change_carrier(struct net_device *, bool new_carrier);
3392int dev_get_phys_port_id(struct net_device *dev,
02637fce 3393 struct netdev_phys_item_id *ppid);
db24a904
DA
3394int dev_get_phys_port_name(struct net_device *dev,
3395 char *name, size_t len);
d746d707 3396int dev_change_proto_down(struct net_device *dev, bool proto_down);
f53c7239 3397struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
ce93718f
DM
3398struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3399 struct netdev_queue *txq, int *ret);
d67b9cd2 3400
f4e63525 3401typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
d67b9cd2
DB
3402int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3403 int fd, u32 flags);
118b4aa2
JK
3404void __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3405 struct netdev_bpf *xdp);
d67b9cd2 3406
a0265d28 3407int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3408int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3409bool is_skb_forwardable(const struct net_device *dev,
3410 const struct sk_buff *skb);
1da177e4 3411
4e3264d2
MKL
3412static __always_inline int ____dev_forward_skb(struct net_device *dev,
3413 struct sk_buff *skb)
3414{
3415 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3416 unlikely(!is_skb_forwardable(dev, skb))) {
3417 atomic_long_inc(&dev->rx_dropped);
3418 kfree_skb(skb);
3419 return NET_RX_DROP;
3420 }
3421
3422 skb_scrub_packet(skb, true);
3423 skb->priority = 0;
3424 return 0;
3425}
3426
74b20582
DA
3427void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3428
20380731 3429extern int netdev_budget;
7acf8a1e 3430extern unsigned int netdev_budget_usecs;
1da177e4
LT
3431
3432/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3433void netdev_run_todo(void);
1da177e4 3434
bea3348e
SH
3435/**
3436 * dev_put - release reference to device
3437 * @dev: network device
3438 *
9ef4429b 3439 * Release reference to device to allow it to be freed.
bea3348e 3440 */
1da177e4
LT
3441static inline void dev_put(struct net_device *dev)
3442{
933393f5 3443 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3444}
3445
bea3348e
SH
3446/**
3447 * dev_hold - get reference to device
3448 * @dev: network device
3449 *
9ef4429b 3450 * Hold reference to device to keep it from being freed.
bea3348e 3451 */
15333061
SH
3452static inline void dev_hold(struct net_device *dev)
3453{
933393f5 3454 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3455}
1da177e4
LT
3456
3457/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3458 * and _off may be called from IRQ context, but it is caller
3459 * who is responsible for serialization of these calls.
b00055aa
SR
3460 *
3461 * The name carrier is inappropriate, these functions should really be
3462 * called netif_lowerlayer_*() because they represent the state of any
3463 * kind of lower layer not just hardware media.
1da177e4
LT
3464 */
3465
f629d208
JP
3466void linkwatch_init_dev(struct net_device *dev);
3467void linkwatch_fire_event(struct net_device *dev);
3468void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3469
bea3348e
SH
3470/**
3471 * netif_carrier_ok - test if carrier present
3472 * @dev: network device
3473 *
3474 * Check if carrier is present on device
3475 */
4d29515f 3476static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3477{
3478 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3479}
3480
f629d208 3481unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3482
f629d208 3483void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3484
f629d208 3485void netif_carrier_on(struct net_device *dev);
1da177e4 3486
f629d208 3487void netif_carrier_off(struct net_device *dev);
1da177e4 3488
bea3348e
SH
3489/**
3490 * netif_dormant_on - mark device as dormant.
3491 * @dev: network device
3492 *
3493 * Mark device as dormant (as per RFC2863).
3494 *
3495 * The dormant state indicates that the relevant interface is not
3496 * actually in a condition to pass packets (i.e., it is not 'up') but is
3497 * in a "pending" state, waiting for some external event. For "on-
3498 * demand" interfaces, this new state identifies the situation where the
3499 * interface is waiting for events to place it in the up state.
bea3348e 3500 */
b00055aa
SR
3501static inline void netif_dormant_on(struct net_device *dev)
3502{
3503 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3504 linkwatch_fire_event(dev);
3505}
3506
bea3348e
SH
3507/**
3508 * netif_dormant_off - set device as not dormant.
3509 * @dev: network device
3510 *
3511 * Device is not in dormant state.
3512 */
b00055aa
SR
3513static inline void netif_dormant_off(struct net_device *dev)
3514{
3515 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3516 linkwatch_fire_event(dev);
3517}
3518
bea3348e 3519/**
8ecbc40a 3520 * netif_dormant - test if device is dormant
bea3348e
SH
3521 * @dev: network device
3522 *
8ecbc40a 3523 * Check if device is dormant.
bea3348e 3524 */
4d29515f 3525static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3526{
3527 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3528}
3529
3530
bea3348e
SH
3531/**
3532 * netif_oper_up - test if device is operational
3533 * @dev: network device
3534 *
3535 * Check if carrier is operational
3536 */
4d29515f 3537static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3538{
b00055aa
SR
3539 return (dev->operstate == IF_OPER_UP ||
3540 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3541}
3542
bea3348e
SH
3543/**
3544 * netif_device_present - is device available or removed
3545 * @dev: network device
3546 *
3547 * Check if device has not been removed from system.
3548 */
4d29515f 3549static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3550{
3551 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3552}
3553
f629d208 3554void netif_device_detach(struct net_device *dev);
1da177e4 3555
f629d208 3556void netif_device_attach(struct net_device *dev);
1da177e4
LT
3557
3558/*
3559 * Network interface message level settings
3560 */
1da177e4
LT
3561
3562enum {
3563 NETIF_MSG_DRV = 0x0001,
3564 NETIF_MSG_PROBE = 0x0002,
3565 NETIF_MSG_LINK = 0x0004,
3566 NETIF_MSG_TIMER = 0x0008,
3567 NETIF_MSG_IFDOWN = 0x0010,
3568 NETIF_MSG_IFUP = 0x0020,
3569 NETIF_MSG_RX_ERR = 0x0040,
3570 NETIF_MSG_TX_ERR = 0x0080,
3571 NETIF_MSG_TX_QUEUED = 0x0100,
3572 NETIF_MSG_INTR = 0x0200,
3573 NETIF_MSG_TX_DONE = 0x0400,
3574 NETIF_MSG_RX_STATUS = 0x0800,
3575 NETIF_MSG_PKTDATA = 0x1000,
3576 NETIF_MSG_HW = 0x2000,
3577 NETIF_MSG_WOL = 0x4000,
3578};
3579
3580#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3581#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3582#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3583#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3584#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3585#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3586#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3587#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3588#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3589#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3590#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3591#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3592#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3593#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3594#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3595
3596static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3597{
3598 /* use default */
3599 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3600 return default_msg_enable_bits;
3601 if (debug_value == 0) /* no output */
3602 return 0;
3603 /* set low N bits */
3604 return (1 << debug_value) - 1;
3605}
3606
c773e847 3607static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3608{
c773e847
DM
3609 spin_lock(&txq->_xmit_lock);
3610 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3611}
3612
5a717f4f
MT
3613static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3614{
3615 __acquire(&txq->_xmit_lock);
3616 return true;
3617}
3618
3619static inline void __netif_tx_release(struct netdev_queue *txq)
3620{
3621 __release(&txq->_xmit_lock);
3622}
3623
fd2ea0a7
DM
3624static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3625{
3626 spin_lock_bh(&txq->_xmit_lock);
3627 txq->xmit_lock_owner = smp_processor_id();
3628}
3629
4d29515f 3630static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3631{
4d29515f 3632 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3633 if (likely(ok))
3634 txq->xmit_lock_owner = smp_processor_id();
3635 return ok;
3636}
3637
3638static inline void __netif_tx_unlock(struct netdev_queue *txq)
3639{
3640 txq->xmit_lock_owner = -1;
3641 spin_unlock(&txq->_xmit_lock);
3642}
3643
3644static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3645{
3646 txq->xmit_lock_owner = -1;
3647 spin_unlock_bh(&txq->_xmit_lock);
3648}
3649
08baf561
ED
3650static inline void txq_trans_update(struct netdev_queue *txq)
3651{
3652 if (txq->xmit_lock_owner != -1)
3653 txq->trans_start = jiffies;
3654}
3655
ba162f8e
FW
3656/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3657static inline void netif_trans_update(struct net_device *dev)
3658{
9b36627a
FW
3659 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3660
3661 if (txq->trans_start != jiffies)
3662 txq->trans_start = jiffies;
ba162f8e
FW
3663}
3664
d29f749e
DJ
3665/**
3666 * netif_tx_lock - grab network device transmit lock
3667 * @dev: network device
d29f749e
DJ
3668 *
3669 * Get network device transmit lock
3670 */
22dd7495
JHS
3671static inline void netif_tx_lock(struct net_device *dev)
3672{
e8a0464c 3673 unsigned int i;
c3f26a26 3674 int cpu;
c773e847 3675
c3f26a26
DM
3676 spin_lock(&dev->tx_global_lock);
3677 cpu = smp_processor_id();
e8a0464c
DM
3678 for (i = 0; i < dev->num_tx_queues; i++) {
3679 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3680
3681 /* We are the only thread of execution doing a
3682 * freeze, but we have to grab the _xmit_lock in
3683 * order to synchronize with threads which are in
3684 * the ->hard_start_xmit() handler and already
3685 * checked the frozen bit.
3686 */
e8a0464c 3687 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3688 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3689 __netif_tx_unlock(txq);
e8a0464c 3690 }
932ff279
HX
3691}
3692
3693static inline void netif_tx_lock_bh(struct net_device *dev)
3694{
e8a0464c
DM
3695 local_bh_disable();
3696 netif_tx_lock(dev);
932ff279
HX
3697}
3698
932ff279
HX
3699static inline void netif_tx_unlock(struct net_device *dev)
3700{
e8a0464c
DM
3701 unsigned int i;
3702
3703 for (i = 0; i < dev->num_tx_queues; i++) {
3704 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3705
c3f26a26
DM
3706 /* No need to grab the _xmit_lock here. If the
3707 * queue is not stopped for another reason, we
3708 * force a schedule.
3709 */
3710 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3711 netif_schedule_queue(txq);
c3f26a26
DM
3712 }
3713 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3714}
3715
3716static inline void netif_tx_unlock_bh(struct net_device *dev)
3717{
e8a0464c
DM
3718 netif_tx_unlock(dev);
3719 local_bh_enable();
932ff279
HX
3720}
3721
c773e847 3722#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3723 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3724 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3725 } else { \
3726 __netif_tx_acquire(txq); \
22dd7495
JHS
3727 } \
3728}
3729
5efeac44
EB
3730#define HARD_TX_TRYLOCK(dev, txq) \
3731 (((dev->features & NETIF_F_LLTX) == 0) ? \
3732 __netif_tx_trylock(txq) : \
5a717f4f 3733 __netif_tx_acquire(txq))
5efeac44 3734
c773e847 3735#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3736 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3737 __netif_tx_unlock(txq); \
5a717f4f
MT
3738 } else { \
3739 __netif_tx_release(txq); \
22dd7495
JHS
3740 } \
3741}
3742
1da177e4
LT
3743static inline void netif_tx_disable(struct net_device *dev)
3744{
fd2ea0a7 3745 unsigned int i;
c3f26a26 3746 int cpu;
fd2ea0a7 3747
c3f26a26
DM
3748 local_bh_disable();
3749 cpu = smp_processor_id();
fd2ea0a7
DM
3750 for (i = 0; i < dev->num_tx_queues; i++) {
3751 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3752
3753 __netif_tx_lock(txq, cpu);
fd2ea0a7 3754 netif_tx_stop_queue(txq);
c3f26a26 3755 __netif_tx_unlock(txq);
fd2ea0a7 3756 }
c3f26a26 3757 local_bh_enable();
1da177e4
LT
3758}
3759
e308a5d8
DM
3760static inline void netif_addr_lock(struct net_device *dev)
3761{
3762 spin_lock(&dev->addr_list_lock);
3763}
3764
2429f7ac
JP
3765static inline void netif_addr_lock_nested(struct net_device *dev)
3766{
25175ba5
VY
3767 int subclass = SINGLE_DEPTH_NESTING;
3768
3769 if (dev->netdev_ops->ndo_get_lock_subclass)
3770 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3771
3772 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3773}
3774
e308a5d8
DM
3775static inline void netif_addr_lock_bh(struct net_device *dev)
3776{
3777 spin_lock_bh(&dev->addr_list_lock);
3778}
3779
3780static inline void netif_addr_unlock(struct net_device *dev)
3781{
3782 spin_unlock(&dev->addr_list_lock);
3783}
3784
3785static inline void netif_addr_unlock_bh(struct net_device *dev)
3786{
3787 spin_unlock_bh(&dev->addr_list_lock);
3788}
3789
f001fde5 3790/*
31278e71 3791 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3792 * rcu_read_lock held.
3793 */
3794#define for_each_dev_addr(dev, ha) \
31278e71 3795 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3796
1da177e4
LT
3797/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3798
f629d208 3799void ether_setup(struct net_device *dev);
1da177e4
LT
3800
3801/* Support for loadable net-drivers */
f629d208 3802struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3803 unsigned char name_assign_type,
f629d208
JP
3804 void (*setup)(struct net_device *),
3805 unsigned int txqs, unsigned int rxqs);
0ad646c8
CW
3806int dev_get_valid_name(struct net *net, struct net_device *dev,
3807 const char *name);
3808
c835a677
TG
3809#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3810 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3811
c835a677
TG
3812#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3813 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3814 count)
36909ea4 3815
f629d208
JP
3816int register_netdev(struct net_device *dev);
3817void unregister_netdev(struct net_device *dev);
f001fde5 3818
22bedad3 3819/* General hardware address lists handling functions */
f629d208
JP
3820int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3821 struct netdev_hw_addr_list *from_list, int addr_len);
3822void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3823 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3824int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3825 struct net_device *dev,
3826 int (*sync)(struct net_device *, const unsigned char *),
3827 int (*unsync)(struct net_device *,
3828 const unsigned char *));
3829void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3830 struct net_device *dev,
3831 int (*unsync)(struct net_device *,
3832 const unsigned char *));
f629d208 3833void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3834
f001fde5 3835/* Functions used for device addresses handling */
f629d208
JP
3836int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3837 unsigned char addr_type);
3838int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3839 unsigned char addr_type);
f629d208
JP
3840void dev_addr_flush(struct net_device *dev);
3841int dev_addr_init(struct net_device *dev);
a748ee24
JP
3842
3843/* Functions used for unicast addresses handling */
f629d208
JP
3844int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3845int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3846int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3847int dev_uc_sync(struct net_device *to, struct net_device *from);
3848int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3849void dev_uc_unsync(struct net_device *to, struct net_device *from);
3850void dev_uc_flush(struct net_device *dev);
3851void dev_uc_init(struct net_device *dev);
f001fde5 3852
670e5b8e
AD
3853/**
3854 * __dev_uc_sync - Synchonize device's unicast list
3855 * @dev: device to sync
3856 * @sync: function to call if address should be added
3857 * @unsync: function to call if address should be removed
3858 *
3859 * Add newly added addresses to the interface, and release
3860 * addresses that have been deleted.
5e82b4b2 3861 */
670e5b8e
AD
3862static inline int __dev_uc_sync(struct net_device *dev,
3863 int (*sync)(struct net_device *,
3864 const unsigned char *),
3865 int (*unsync)(struct net_device *,
3866 const unsigned char *))
3867{
3868 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3869}
3870
3871/**
e793c0f7 3872 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3873 * @dev: device to sync
3874 * @unsync: function to call if address should be removed
3875 *
3876 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3877 */
670e5b8e
AD
3878static inline void __dev_uc_unsync(struct net_device *dev,
3879 int (*unsync)(struct net_device *,
3880 const unsigned char *))
3881{
3882 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3883}
3884
22bedad3 3885/* Functions used for multicast addresses handling */
f629d208
JP
3886int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3887int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3888int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3889int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3890int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3891int dev_mc_sync(struct net_device *to, struct net_device *from);
3892int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3893void dev_mc_unsync(struct net_device *to, struct net_device *from);
3894void dev_mc_flush(struct net_device *dev);
3895void dev_mc_init(struct net_device *dev);
f001fde5 3896
670e5b8e
AD
3897/**
3898 * __dev_mc_sync - Synchonize device's multicast list
3899 * @dev: device to sync
3900 * @sync: function to call if address should be added
3901 * @unsync: function to call if address should be removed
3902 *
3903 * Add newly added addresses to the interface, and release
3904 * addresses that have been deleted.
5e82b4b2 3905 */
670e5b8e
AD
3906static inline int __dev_mc_sync(struct net_device *dev,
3907 int (*sync)(struct net_device *,
3908 const unsigned char *),
3909 int (*unsync)(struct net_device *,
3910 const unsigned char *))
3911{
3912 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3913}
3914
3915/**
e793c0f7 3916 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3917 * @dev: device to sync
3918 * @unsync: function to call if address should be removed
3919 *
3920 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3921 */
670e5b8e
AD
3922static inline void __dev_mc_unsync(struct net_device *dev,
3923 int (*unsync)(struct net_device *,
3924 const unsigned char *))
3925{
3926 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3927}
3928
4417da66 3929/* Functions used for secondary unicast and multicast support */
f629d208
JP
3930void dev_set_rx_mode(struct net_device *dev);
3931void __dev_set_rx_mode(struct net_device *dev);
3932int dev_set_promiscuity(struct net_device *dev, int inc);
3933int dev_set_allmulti(struct net_device *dev, int inc);
3934void netdev_state_change(struct net_device *dev);
3935void netdev_notify_peers(struct net_device *dev);
3936void netdev_features_change(struct net_device *dev);
1da177e4 3937/* Load a device via the kmod */
f629d208
JP
3938void dev_load(struct net *net, const char *name);
3939struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3940 struct rtnl_link_stats64 *storage);
3941void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3942 const struct net_device_stats *netdev_stats);
eeda3fd6 3943
1da177e4 3944extern int netdev_max_backlog;
3b098e2d 3945extern int netdev_tstamp_prequeue;
1da177e4 3946extern int weight_p;
3d48b53f
MT
3947extern int dev_weight_rx_bias;
3948extern int dev_weight_tx_bias;
3949extern int dev_rx_weight;
3950extern int dev_tx_weight;
9ff162a8 3951
f629d208 3952bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3953struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3954 struct list_head **iter);
f629d208
JP
3955struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3956 struct list_head **iter);
8b5be856 3957
44a40855
VY
3958/* iterate through upper list, must be called under RCU read lock */
3959#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3960 for (iter = &(dev)->adj_list.upper, \
3961 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3962 updev; \
3963 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3964
1a3f060c
DA
3965int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3966 int (*fn)(struct net_device *upper_dev,
3967 void *data),
3968 void *data);
3969
3970bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3971 struct net_device *upper_dev);
3972
25cc72a3
IS
3973bool netdev_has_any_upper_dev(struct net_device *dev);
3974
f629d208
JP
3975void *netdev_lower_get_next_private(struct net_device *dev,
3976 struct list_head **iter);
3977void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3978 struct list_head **iter);
31088a11
VF
3979
3980#define netdev_for_each_lower_private(dev, priv, iter) \
3981 for (iter = (dev)->adj_list.lower.next, \
3982 priv = netdev_lower_get_next_private(dev, &(iter)); \
3983 priv; \
3984 priv = netdev_lower_get_next_private(dev, &(iter)))
3985
3986#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3987 for (iter = &(dev)->adj_list.lower, \
3988 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3989 priv; \
3990 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3991
4085ebe8
VY
3992void *netdev_lower_get_next(struct net_device *dev,
3993 struct list_head **iter);
7ce856aa 3994
4085ebe8 3995#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3996 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3997 ldev = netdev_lower_get_next(dev, &(iter)); \
3998 ldev; \
3999 ldev = netdev_lower_get_next(dev, &(iter)))
4000
7ce856aa
JP
4001struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4002 struct list_head **iter);
4003struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4004 struct list_head **iter);
4005
1a3f060c
DA
4006int netdev_walk_all_lower_dev(struct net_device *dev,
4007 int (*fn)(struct net_device *lower_dev,
4008 void *data),
4009 void *data);
4010int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4011 int (*fn)(struct net_device *lower_dev,
4012 void *data),
4013 void *data);
4014
f629d208 4015void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 4016void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
4017struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4018struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
42ab19ee
DA
4019int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4020 struct netlink_ext_ack *extack);
f629d208 4021int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 4022 struct net_device *upper_dev,
42ab19ee
DA
4023 void *upper_priv, void *upper_info,
4024 struct netlink_ext_ack *extack);
f629d208
JP
4025void netdev_upper_dev_unlink(struct net_device *dev,
4026 struct net_device *upper_dev);
5bb025fa 4027void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
4028void *netdev_lower_dev_get_private(struct net_device *dev,
4029 struct net_device *lower_dev);
04d48266
JP
4030void netdev_lower_state_changed(struct net_device *lower_dev,
4031 void *lower_state_info);
960fb622
ED
4032
4033/* RSS keys are 40 or 52 bytes long */
4034#define NETDEV_RSS_KEY_LEN 52
ba905f5e 4035extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
4036void netdev_rss_key_fill(void *buffer, size_t len);
4037
952fcfd0 4038int dev_get_nest_level(struct net_device *dev);
f629d208 4039int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 4040int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
4041int skb_csum_hwoffload_help(struct sk_buff *skb,
4042 const netdev_features_t features);
4043
f629d208
JP
4044struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4045 netdev_features_t features, bool tx_path);
4046struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4047 netdev_features_t features);
12b0004d 4048
61bd3857
MS
4049struct netdev_bonding_info {
4050 ifslave slave;
4051 ifbond master;
4052};
4053
4054struct netdev_notifier_bonding_info {
4055 struct netdev_notifier_info info; /* must be first */
4056 struct netdev_bonding_info bonding_info;
4057};
4058
4059void netdev_bonding_info_change(struct net_device *dev,
4060 struct netdev_bonding_info *bonding_info);
4061
12b0004d
CW
4062static inline
4063struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4064{
4065 return __skb_gso_segment(skb, features, true);
4066}
53d6471c 4067__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
4068
4069static inline bool can_checksum_protocol(netdev_features_t features,
4070 __be16 protocol)
4071{
c8cd0989
TH
4072 if (protocol == htons(ETH_P_FCOE))
4073 return !!(features & NETIF_F_FCOE_CRC);
4074
4075 /* Assume this is an IP checksum (not SCTP CRC) */
4076
4077 if (features & NETIF_F_HW_CSUM) {
4078 /* Can checksum everything */
4079 return true;
4080 }
4081
4082 switch (protocol) {
4083 case htons(ETH_P_IP):
4084 return !!(features & NETIF_F_IP_CSUM);
4085 case htons(ETH_P_IPV6):
4086 return !!(features & NETIF_F_IPV6_CSUM);
4087 default:
4088 return false;
4089 }
ec5f0615 4090}
12b0004d 4091
fb286bb2 4092#ifdef CONFIG_BUG
f629d208 4093void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
4094#else
4095static inline void netdev_rx_csum_fault(struct net_device *dev)
4096{
4097}
4098#endif
1da177e4 4099/* rx skb timestamps */
f629d208
JP
4100void net_enable_timestamp(void);
4101void net_disable_timestamp(void);
1da177e4 4102
20380731 4103#ifdef CONFIG_PROC_FS
f629d208 4104int __init dev_proc_init(void);
900ff8c6
CW
4105#else
4106#define dev_proc_init() 0
20380731
ACM
4107#endif
4108
4798248e 4109static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4110 struct sk_buff *skb, struct net_device *dev,
4111 bool more)
4798248e 4112{
fa2dbdc2 4113 skb->xmit_more = more ? 1 : 0;
0b725a2c 4114 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4115}
4116
10b3ad8c 4117static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4118 struct netdev_queue *txq, bool more)
4798248e
DM
4119{
4120 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4121 int rc;
4798248e 4122
fa2dbdc2 4123 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4124 if (rc == NETDEV_TX_OK)
4125 txq_trans_update(txq);
4126
4127 return rc;
4798248e
DM
4128}
4129
b793dc5c 4130int netdev_class_create_file_ns(const struct class_attribute *class_attr,
42a2d923 4131 const void *ns);
b793dc5c 4132void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
42a2d923 4133 const void *ns);
58292cbe 4134
b793dc5c 4135static inline int netdev_class_create_file(const struct class_attribute *class_attr)
58292cbe
TH
4136{
4137 return netdev_class_create_file_ns(class_attr, NULL);
4138}
4139
b793dc5c 4140static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
58292cbe
TH
4141{
4142 netdev_class_remove_file_ns(class_attr, NULL);
4143}
b8a9787e 4144
737aec57 4145extern const struct kobj_ns_type_operations net_ns_type_operations;
04600794 4146
f629d208 4147const char *netdev_drivername(const struct net_device *dev);
6579e57b 4148
f629d208 4149void linkwatch_run_queue(void);
20380731 4150
da08143b
MK
4151static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4152 netdev_features_t f2)
4153{
c8cd0989
TH
4154 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4155 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4156 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4157 else
b6a0e72a 4158 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4159 }
da08143b 4160
c8cd0989 4161 return f1 & f2;
da08143b
MK
4162}
4163
c8f44aff
MM
4164static inline netdev_features_t netdev_get_wanted_features(
4165 struct net_device *dev)
5455c699
MM
4166{
4167 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4168}
c8f44aff
MM
4169netdev_features_t netdev_increment_features(netdev_features_t all,
4170 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4171
4172/* Allow TSO being used on stacked device :
4173 * Performing the GSO segmentation before last device
4174 * is a performance improvement.
4175 */
4176static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4177 netdev_features_t mask)
4178{
4179 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4180}
4181
6cb6a27c 4182int __netdev_update_features(struct net_device *dev);
5455c699 4183void netdev_update_features(struct net_device *dev);
afe12cc8 4184void netdev_change_features(struct net_device *dev);
7f353bf2 4185
fc4a7489
PM
4186void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4187 struct net_device *dev);
4188
e38f3025
TM
4189netdev_features_t passthru_features_check(struct sk_buff *skb,
4190 struct net_device *dev,
4191 netdev_features_t features);
c1e756bf 4192netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4193
4d29515f 4194static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4195{
7b748340 4196 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4197
4198 /* check flags correspondence */
4199 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4200 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4201 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4202 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4203 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4204 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4205 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4206 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4207 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4208 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4209 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4210 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4211 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4212 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4213 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4214 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0c19f846 4215 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
83aa025f 4216 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
0345e186 4217
d6b4991a 4218 return (features & feature) == feature;
576a30eb
HX
4219}
4220
4d29515f 4221static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4222{
278b2513 4223 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4224 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4225}
4226
8b86a61d 4227static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4228 netdev_features_t features)
7967168c 4229{
fc741216 4230 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4231 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4232 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4233}
4234
82cc1a7a
PWJ
4235static inline void netif_set_gso_max_size(struct net_device *dev,
4236 unsigned int size)
4237{
4238 dev->gso_max_size = size;
4239}
4240
7a7ffbab
WCC
4241static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4242 int pulled_hlen, u16 mac_offset,
4243 int mac_len)
4244{
4245 skb->protocol = protocol;
4246 skb->encapsulation = 1;
4247 skb_push(skb, pulled_hlen);
4248 skb_reset_transport_header(skb);
4249 skb->mac_header = mac_offset;
4250 skb->network_header = skb->mac_header + mac_len;
4251 skb->mac_len = mac_len;
4252}
4253
3c175784
SD
4254static inline bool netif_is_macsec(const struct net_device *dev)
4255{
4256 return dev->priv_flags & IFF_MACSEC;
4257}
4258
b618aaa9 4259static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4260{
4261 return dev->priv_flags & IFF_MACVLAN;
4262}
4263
b618aaa9 4264static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4265{
4266 return dev->priv_flags & IFF_MACVLAN_PORT;
4267}
4268
b618aaa9 4269static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4270{
4271 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4272}
4273
b618aaa9 4274static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4275{
4276 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4277}
4278
3bdc0eba
BG
4279static inline bool netif_supports_nofcs(struct net_device *dev)
4280{
4281 return dev->priv_flags & IFF_SUPP_NOFCS;
4282}
4283
007979ea 4284static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4285{
007979ea 4286 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4287}
4288
fee6d4c7
DA
4289static inline bool netif_is_l3_slave(const struct net_device *dev)
4290{
4291 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4292}
4293
0894ae3f
JP
4294static inline bool netif_is_bridge_master(const struct net_device *dev)
4295{
4296 return dev->priv_flags & IFF_EBRIDGE;
4297}
4298
28f9ee22
VY
4299static inline bool netif_is_bridge_port(const struct net_device *dev)
4300{
4301 return dev->priv_flags & IFF_BRIDGE_PORT;
4302}
4303
35d4e172
JP
4304static inline bool netif_is_ovs_master(const struct net_device *dev)
4305{
4306 return dev->priv_flags & IFF_OPENVSWITCH;
4307}
4308
5be66141
JP
4309static inline bool netif_is_ovs_port(const struct net_device *dev)
4310{
4311 return dev->priv_flags & IFF_OVS_DATAPATH;
4312}
4313
b618aaa9 4314static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4315{
4316 return dev->priv_flags & IFF_TEAM;
4317}
4318
b618aaa9 4319static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4320{
4321 return dev->priv_flags & IFF_TEAM_PORT;
4322}
4323
b618aaa9 4324static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4325{
4326 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4327}
4328
b618aaa9 4329static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4330{
4331 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4332}
4333
d4ab4286
KJ
4334static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4335{
4336 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4337}
4338
02875878
ED
4339/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4340static inline void netif_keep_dst(struct net_device *dev)
4341{
4342 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4343}
4344
18d3df3e
PA
4345/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4346static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4347{
4348 /* TODO: reserve and use an additional IFF bit, if we get more users */
4349 return dev->priv_flags & IFF_MACSEC;
4350}
4351
505d4f73 4352extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4353
571ba423
JP
4354/* Logging, debugging and troubleshooting/diagnostic helpers. */
4355
4356/* netdev_printk helpers, similar to dev_printk */
4357
4358static inline const char *netdev_name(const struct net_device *dev)
4359{
c6f854d5
VF
4360 if (!dev->name[0] || strchr(dev->name, '%'))
4361 return "(unnamed net_device)";
571ba423
JP
4362 return dev->name;
4363}
4364
8397ed36
DA
4365static inline bool netdev_unregistering(const struct net_device *dev)
4366{
4367 return dev->reg_state == NETREG_UNREGISTERING;
4368}
4369
ccc7f496
VF
4370static inline const char *netdev_reg_state(const struct net_device *dev)
4371{
4372 switch (dev->reg_state) {
4373 case NETREG_UNINITIALIZED: return " (uninitialized)";
4374 case NETREG_REGISTERED: return "";
4375 case NETREG_UNREGISTERING: return " (unregistering)";
4376 case NETREG_UNREGISTERED: return " (unregistered)";
4377 case NETREG_RELEASED: return " (released)";
4378 case NETREG_DUMMY: return " (dummy)";
4379 }
4380
4381 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4382 return " (unknown)";
4383}
4384
f629d208 4385__printf(3, 4)
6ea754eb
JP
4386void netdev_printk(const char *level, const struct net_device *dev,
4387 const char *format, ...);
f629d208 4388__printf(2, 3)
6ea754eb 4389void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4390__printf(2, 3)
6ea754eb 4391void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4392__printf(2, 3)
6ea754eb 4393void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4394__printf(2, 3)
6ea754eb 4395void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4396__printf(2, 3)
6ea754eb 4397void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4398__printf(2, 3)
6ea754eb 4399void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4400__printf(2, 3)
6ea754eb 4401void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4402
375ef2b1
GP
4403#define netdev_level_once(level, dev, fmt, ...) \
4404do { \
4405 static bool __print_once __read_mostly; \
4406 \
4407 if (!__print_once) { \
4408 __print_once = true; \
4409 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4410 } \
4411} while (0)
4412
4413#define netdev_emerg_once(dev, fmt, ...) \
4414 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4415#define netdev_alert_once(dev, fmt, ...) \
4416 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4417#define netdev_crit_once(dev, fmt, ...) \
4418 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4419#define netdev_err_once(dev, fmt, ...) \
4420 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4421#define netdev_warn_once(dev, fmt, ...) \
4422 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4423#define netdev_notice_once(dev, fmt, ...) \
4424 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4425#define netdev_info_once(dev, fmt, ...) \
4426 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4427
8909c9ad
VK
4428#define MODULE_ALIAS_NETDEV(device) \
4429 MODULE_ALIAS("netdev-" device)
4430
b558c96f 4431#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4432#define netdev_dbg(__dev, format, args...) \
4433do { \
ffa10cb4 4434 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4435} while (0)
b558c96f
JC
4436#elif defined(DEBUG)
4437#define netdev_dbg(__dev, format, args...) \
4438 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4439#else
4440#define netdev_dbg(__dev, format, args...) \
4441({ \
4442 if (0) \
4443 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4444})
4445#endif
4446
4447#if defined(VERBOSE_DEBUG)
4448#define netdev_vdbg netdev_dbg
4449#else
4450
4451#define netdev_vdbg(dev, format, args...) \
4452({ \
4453 if (0) \
4454 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4455 0; \
4456})
4457#endif
4458
4459/*
4460 * netdev_WARN() acts like dev_printk(), but with the key difference
4461 * of using a WARN/WARN_ON to get the message out, including the
4462 * file/line information and a backtrace.
4463 */
4464#define netdev_WARN(dev, format, args...) \
e1cfe3d0 4465 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
ccc7f496 4466 netdev_reg_state(dev), ##args)
571ba423 4467
72dd831e 4468#define netdev_WARN_ONCE(dev, format, args...) \
e1cfe3d0 4469 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
375ef2b1
GP
4470 netdev_reg_state(dev), ##args)
4471
b3d95c5c
JP
4472/* netif printk helpers, similar to netdev_printk */
4473
4474#define netif_printk(priv, type, level, dev, fmt, args...) \
4475do { \
4476 if (netif_msg_##type(priv)) \
4477 netdev_printk(level, (dev), fmt, ##args); \
4478} while (0)
4479
f45f4321
JP
4480#define netif_level(level, priv, type, dev, fmt, args...) \
4481do { \
4482 if (netif_msg_##type(priv)) \
4483 netdev_##level(dev, fmt, ##args); \
4484} while (0)
4485
b3d95c5c 4486#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4487 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4488#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4489 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4490#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4491 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4492#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4493 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4494#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4495 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4496#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4497 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4498#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4499 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4500
0053ea9c 4501#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4502#define netif_dbg(priv, type, netdev, format, args...) \
4503do { \
4504 if (netif_msg_##type(priv)) \
b5fb0a03 4505 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4506} while (0)
0053ea9c
JP
4507#elif defined(DEBUG)
4508#define netif_dbg(priv, type, dev, format, args...) \
4509 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4510#else
4511#define netif_dbg(priv, type, dev, format, args...) \
4512({ \
4513 if (0) \
4514 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4515 0; \
4516})
4517#endif
4518
f617f276
EC
4519/* if @cond then downgrade to debug, else print at @level */
4520#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4521 do { \
4522 if (cond) \
4523 netif_dbg(priv, type, netdev, fmt, ##args); \
4524 else \
4525 netif_ ## level(priv, type, netdev, fmt, ##args); \
4526 } while (0)
4527
b3d95c5c 4528#if defined(VERBOSE_DEBUG)
bcfcc450 4529#define netif_vdbg netif_dbg
b3d95c5c
JP
4530#else
4531#define netif_vdbg(priv, type, dev, format, args...) \
4532({ \
4533 if (0) \
a4ed89cb 4534 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4535 0; \
4536})
4537#endif
571ba423 4538
900ff8c6
CW
4539/*
4540 * The list of packet types we will receive (as opposed to discard)
4541 * and the routines to invoke.
4542 *
4543 * Why 16. Because with 16 the only overlap we get on a hash of the
4544 * low nibble of the protocol value is RARP/SNAP/X.25.
4545 *
900ff8c6 4546 * 0800 IP
900ff8c6
CW
4547 * 0001 802.3
4548 * 0002 AX.25
4549 * 0004 802.2
4550 * 8035 RARP
4551 * 0005 SNAP
4552 * 0805 X.25
4553 * 0806 ARP
4554 * 8137 IPX
4555 * 0009 Localtalk
4556 * 86DD IPv6
4557 */
4558#define PTYPE_HASH_SIZE (16)
4559#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4560
385a154c 4561#endif /* _LINUX_NETDEVICE_H */