net: Add extack to ndo_add_slave
[linux-2.6-block.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
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>
a050c33f 47
a59e2ecb 48#include <linux/netdev_features.h>
77162022 49#include <linux/neighbour.h>
607ca46e 50#include <uapi/linux/netdevice.h>
61bd3857 51#include <uapi/linux/if_bonding.h>
e4c6734e 52#include <uapi/linux/pkt_cls.h>
59cc1f61 53#include <linux/hashtable.h>
a59e2ecb 54
115c1d6e 55struct netpoll_info;
313162d0 56struct device;
c1f19b51 57struct phy_device;
2f657a60 58struct dsa_port;
c6e970a0 59
704232c2
JB
60/* 802.11 specific */
61struct wireless_dev;
98a18b6f
AA
62/* 802.15.4 specific */
63struct wpan_dev;
03c57747 64struct mpls_dev;
7c46a640
AD
65/* UDP Tunnel offloads */
66struct udp_tunnel_info;
a7862b45 67struct bpf_prog;
814abfab 68struct xdp_buff;
1da177e4 69
f629d208
JP
70void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
d07d7507 72
9a1654ba
JP
73/* Backlog congestion levels */
74#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75#define NET_RX_DROP 1 /* packet dropped */
76
572a9d7b
PM
77/*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
91 * others are propagated to higher layers.
92 */
93
94/* qdisc ->enqueue() return codes. */
95#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
96#define NET_XMIT_DROP 0x01 /* skb dropped */
97#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 98#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 99
b9df3cb8
GR
100/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 103#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
104#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105
dc1f8bf6 106/* Driver transmit return codes */
9a1654ba 107#define NETDEV_TX_MASK 0xf0
572a9d7b 108
dc1f8bf6 109enum netdev_tx {
572a9d7b 110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
113};
114typedef enum netdev_tx netdev_tx_t;
115
9a1654ba
JP
116/*
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 */
120static inline bool dev_xmit_complete(int rc)
121{
122 /*
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 */
128 if (likely(rc < NET_XMIT_MASK))
129 return true;
130
131 return false;
132}
133
1da177e4 134/*
5e82b4b2 135 * Compute the worst-case header length according to the protocols
1da177e4
LT
136 * used.
137 */
fe2918b0 138
c0eb4540
KS
139#if defined(CONFIG_HYPERV_NET)
140# define LL_MAX_HEADER 128
141#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
142# if defined(CONFIG_MAC80211_MESH)
143# define LL_MAX_HEADER 128
144# else
145# define LL_MAX_HEADER 96
146# endif
1da177e4 147#else
8388e3da 148# define LL_MAX_HEADER 32
1da177e4
LT
149#endif
150
d11ead75
BH
151#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
153#define MAX_HEADER LL_MAX_HEADER
154#else
155#define MAX_HEADER (LL_MAX_HEADER + 48)
156#endif
157
158/*
be1f3c2c
BH
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
1da177e4 161 */
fe2918b0 162
d94d9fee 163struct net_device_stats {
3cfde79c
BH
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
1da177e4 173 unsigned long collisions;
1da177e4 174 unsigned long rx_length_errors;
3cfde79c
BH
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
1da177e4
LT
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
1da177e4
LT
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187};
188
1da177e4
LT
189
190#include <linux/cache.h>
191#include <linux/skbuff.h>
192
adc9300e 193#ifdef CONFIG_RPS
c5905afb
IM
194#include <linux/static_key.h>
195extern struct static_key rps_needed;
13bfff25 196extern struct static_key rfs_needed;
adc9300e
ED
197#endif
198
1da177e4
LT
199struct neighbour;
200struct neigh_parms;
201struct sk_buff;
202
f001fde5
JP
203struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
206 unsigned char type;
ccffad25
JP
207#define NETDEV_HW_ADDR_T_LAN 1
208#define NETDEV_HW_ADDR_T_SAN 2
209#define NETDEV_HW_ADDR_T_SLAVE 3
210#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 211#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 212 bool global_use;
4cd729b0 213 int sync_cnt;
8f8f103d 214 int refcount;
4543fbef 215 int synced;
f001fde5
JP
216 struct rcu_head rcu_head;
217};
218
31278e71
JP
219struct netdev_hw_addr_list {
220 struct list_head list;
221 int count;
222};
223
22bedad3
JP
224#define netdev_hw_addr_list_count(l) ((l)->count)
225#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226#define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 228
22bedad3
JP
229#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231#define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 233
22bedad3
JP
234#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 236#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 238
d94d9fee 239struct hh_cache {
5b3dc2f3 240 unsigned int hh_len;
3644f0ce 241 seqlock_t hh_lock;
1da177e4
LT
242
243 /* cached hardware header; allow for machine alignment needs. */
244#define HH_DATA_MOD 16
245#define HH_DATA_OFF(__len) \
5ba0eac6 246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
247#define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250};
251
5e82b4b2 252/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260#define LL_RESERVED_SPACE(dev) \
f5184d26 261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 262#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 264
3b04ddde
SH
265struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
95c96174 268 const void *saddr, unsigned int len);
3b04ddde 269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
2793a23a 274 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
275};
276
1da177e4 277/* These flag bits are private to the generic network queueing
5e82b4b2 278 * layer; they may not be explicitly referenced by any other
1da177e4
LT
279 * code.
280 */
281
d94d9fee 282enum netdev_state_t {
1da177e4
LT
283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
1da177e4 285 __LINK_STATE_NOCARRIER,
b00055aa
SR
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
1da177e4
LT
288};
289
290
291/*
5e82b4b2 292 * This structure holds boot-time configured netdevice settings. They
fe2918b0 293 * are then used in the device probing.
1da177e4
LT
294 */
295struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298};
299#define NETDEV_BOOT_SETUP_MAX 8
300
f629d208 301int __init netdev_boot_setup(char *str);
1da177e4 302
bea3348e
SH
303/*
304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 310 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
404f7c9e 317 unsigned int gro_count;
bea3348e
SH
318 int (*poll)(struct napi_struct *, int);
319#ifdef CONFIG_NETPOLL
bea3348e 320 int poll_owner;
bea3348e 321#endif
5d38a079 322 struct net_device *dev;
d565b0a1 323 struct sk_buff *gro_list;
5d38a079 324 struct sk_buff *skb;
3b47d303 325 struct hrtimer timer;
404f7c9e 326 struct list_head dev_list;
af12fa6e
ET
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
bea3348e
SH
329};
330
d94d9fee 331enum {
bea3348e 332 NAPI_STATE_SCHED, /* Poll is scheduled */
39e6c820 333 NAPI_STATE_MISSED, /* reschedule a napi */
a0a46196 334 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 335 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
336 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
337 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
338 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
339};
340
341enum {
39e6c820
ED
342 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
343 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
344 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
345 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
346 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
347 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
348 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
349};
350
5b252f0c 351enum gro_result {
d1c76af9
HX
352 GRO_MERGED,
353 GRO_MERGED_FREE,
354 GRO_HELD,
355 GRO_NORMAL,
356 GRO_DROP,
25393d3f 357 GRO_CONSUMED,
d1c76af9 358};
5b252f0c 359typedef enum gro_result gro_result_t;
d1c76af9 360
8a4eb573
JP
361/*
362 * enum rx_handler_result - Possible return values for rx_handlers.
363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
364 * further.
365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366 * case skb->dev was changed by rx_handler.
367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
369 *
370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
371 * special processing of the skb, prior to delivery to protocol handlers.
372 *
373 * Currently, a net_device can only have a single rx_handler registered. Trying
374 * to register a second rx_handler will return -EBUSY.
375 *
376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377 * To unregister a rx_handler on a net_device, use
378 * netdev_rx_handler_unregister().
379 *
380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
381 * do with the skb.
382 *
5e82b4b2 383 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 385 * the skb to be delivered in some other way.
8a4eb573
JP
386 *
387 * If the rx_handler changed skb->dev, to divert the skb to another
388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389 * new device will be called if it exists.
390 *
5e82b4b2 391 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 393 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 394 *
5e82b4b2 395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
396 * delivered, it should return RX_HANDLER_PASS.
397 *
398 * A device without a registered rx_handler will behave as if rx_handler
399 * returned RX_HANDLER_PASS.
400 */
401
402enum rx_handler_result {
403 RX_HANDLER_CONSUMED,
404 RX_HANDLER_ANOTHER,
405 RX_HANDLER_EXACT,
406 RX_HANDLER_PASS,
407};
408typedef enum rx_handler_result rx_handler_result_t;
409typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 410
f629d208 411void __napi_schedule(struct napi_struct *n);
bc9ad166 412void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 413
4d29515f 414static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
415{
416 return test_bit(NAPI_STATE_DISABLE, &n->state);
417}
418
39e6c820 419bool napi_schedule_prep(struct napi_struct *n);
bea3348e
SH
420
421/**
422 * napi_schedule - schedule NAPI poll
5e82b4b2 423 * @n: NAPI context
bea3348e
SH
424 *
425 * Schedule NAPI poll routine to be called if it is not already
426 * running.
427 */
428static inline void napi_schedule(struct napi_struct *n)
429{
430 if (napi_schedule_prep(n))
431 __napi_schedule(n);
432}
433
bc9ad166
ED
434/**
435 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 436 * @n: NAPI context
bc9ad166
ED
437 *
438 * Variant of napi_schedule(), assuming hard irqs are masked.
439 */
440static inline void napi_schedule_irqoff(struct napi_struct *n)
441{
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
444}
445
bfe13f54 446/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 447static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
448{
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
4d29515f 451 return true;
bfe13f54 452 }
4d29515f 453 return false;
bfe13f54
RD
454}
455
364b6055 456bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
457/**
458 * napi_complete - NAPI processing complete
5e82b4b2 459 * @n: NAPI context
bea3348e
SH
460 *
461 * Mark NAPI processing as complete.
3b47d303 462 * Consider using napi_complete_done() instead.
364b6055 463 * Return false if device should avoid rearming interrupts.
bea3348e 464 */
364b6055 465static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
466{
467 return napi_complete_done(n, 0);
468}
bea3348e 469
af12fa6e
ET
470/**
471 * napi_hash_del - remove a NAPI from global table
5e82b4b2 472 * @napi: NAPI context
af12fa6e 473 *
5e82b4b2 474 * Warning: caller must observe RCU grace period
34cbe27e
ED
475 * before freeing memory containing @napi, if
476 * this function returns true.
93d05d4a 477 * Note: core networking stack automatically calls it
5e82b4b2 478 * from netif_napi_del().
93d05d4a 479 * Drivers might want to call this helper to combine all
5e82b4b2 480 * the needed RCU grace periods into a single one.
af12fa6e 481 */
34cbe27e 482bool napi_hash_del(struct napi_struct *napi);
af12fa6e 483
bea3348e
SH
484/**
485 * napi_disable - prevent NAPI from scheduling
5e82b4b2 486 * @n: NAPI context
bea3348e
SH
487 *
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
490 */
3b47d303 491void napi_disable(struct napi_struct *n);
bea3348e
SH
492
493/**
494 * napi_enable - enable NAPI scheduling
5e82b4b2 495 * @n: NAPI context
bea3348e
SH
496 *
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
499 */
500static inline void napi_enable(struct napi_struct *n)
501{
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 503 smp_mb__before_atomic();
bea3348e 504 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 505 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
506}
507
c264c3de
SH
508/**
509 * napi_synchronize - wait until NAPI is not running
5e82b4b2 510 * @n: NAPI context
c264c3de
SH
511 *
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
515 */
516static inline void napi_synchronize(const struct napi_struct *n)
517{
facc432f
AB
518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
520 msleep(1);
521 else
522 barrier();
c264c3de 523}
c264c3de 524
d94d9fee 525enum netdev_queue_state_t {
73466498
TH
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
c3f26a26 528 __QUEUE_STATE_FROZEN,
79d16385 529};
8e2f1a63
DB
530
531#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
534
535#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
537 QUEUE_STATE_FROZEN)
538#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
539 QUEUE_STATE_FROZEN)
540
73466498
TH
541/*
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
549 */
79d16385 550
bb949fbd 551struct netdev_queue {
6a321cb3 552/*
5e82b4b2 553 * read-mostly part
6a321cb3 554 */
bb949fbd 555 struct net_device *dev;
46e5da40 556 struct Qdisc __rcu *qdisc;
b0e1e646 557 struct Qdisc *qdisc_sleeping;
ccf5ff69 558#ifdef CONFIG_SYSFS
1d24eb48
TH
559 struct kobject kobj;
560#endif
f2cd2d3e
ED
561#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 int numa_node;
563#endif
c0ef079c
FW
564 unsigned long tx_maxrate;
565 /*
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
568 */
569 unsigned long trans_timeout;
6a321cb3 570/*
5e82b4b2 571 * write-mostly part
6a321cb3
ED
572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
9d21493b 575 /*
9b36627a 576 * Time (in jiffies) of last Tx
9d21493b
ED
577 */
578 unsigned long trans_start;
ccf5ff69 579
114cf580
TH
580 unsigned long state;
581
582#ifdef CONFIG_BQL
583 struct dql dql;
584#endif
e8a0464c 585} ____cacheline_aligned_in_smp;
bb949fbd 586
f2cd2d3e
ED
587static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
588{
589#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 return q->numa_node;
591#else
b236da69 592 return NUMA_NO_NODE;
f2cd2d3e
ED
593#endif
594}
595
596static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
597{
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 q->numa_node = node;
600#endif
601}
602
df334545 603#ifdef CONFIG_RPS
0a9627f2
TH
604/*
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
607 */
608struct rps_map {
609 unsigned int len;
610 struct rcu_head rcu;
611 u16 cpus[0];
612};
60b778ce 613#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 614
fec5e652 615/*
c445477d
BH
616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
fec5e652
TH
619 */
620struct rps_dev_flow {
621 u16 cpu;
c445477d 622 u16 filter;
fec5e652
TH
623 unsigned int last_qtail;
624};
c445477d 625#define RPS_NO_FILTER 0xffff
fec5e652
TH
626
627/*
628 * The rps_dev_flow_table structure contains a table of flow mappings.
629 */
630struct rps_dev_flow_table {
631 unsigned int mask;
632 struct rcu_head rcu;
fec5e652
TH
633 struct rps_dev_flow flows[0];
634};
635#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 636 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
637
638/*
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
567e4b79 643 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 646 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
647 */
648struct rps_sock_flow_table {
567e4b79 649 u32 mask;
93c1af6c
ED
650
651 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 652};
567e4b79 653#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
654
655#define RPS_NO_CPU 0xffff
656
567e4b79
ED
657extern u32 rps_cpu_mask;
658extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
659
fec5e652
TH
660static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
661 u32 hash)
662{
663 if (table && hash) {
567e4b79
ED
664 unsigned int index = hash & table->mask;
665 u32 val = hash & ~rps_cpu_mask;
fec5e652 666
5e82b4b2 667 /* We only give a hint, preemption can change CPU under us */
567e4b79 668 val |= raw_smp_processor_id();
fec5e652 669
567e4b79
ED
670 if (table->ents[index] != val)
671 table->ents[index] = val;
fec5e652
TH
672 }
673}
674
c445477d 675#ifdef CONFIG_RFS_ACCEL
f629d208
JP
676bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
677 u16 filter_id);
c445477d 678#endif
a953be53 679#endif /* CONFIG_RPS */
c445477d 680
0a9627f2
TH
681/* This structure contains an instance of an RX queue. */
682struct netdev_rx_queue {
a953be53 683#ifdef CONFIG_RPS
6e3f7faf
ED
684 struct rps_map __rcu *rps_map;
685 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 686#endif
6e3f7faf 687 struct kobject kobj;
fe822240 688 struct net_device *dev;
0a9627f2 689} ____cacheline_aligned_in_smp;
a953be53
MD
690
691/*
692 * RX queue sysfs structures and functions.
693 */
694struct rx_queue_attribute {
695 struct attribute attr;
718ad681 696 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
a953be53 697 ssize_t (*store)(struct netdev_rx_queue *queue,
718ad681 698 const char *buf, size_t len);
a953be53 699};
d314774c 700
bf264145
TH
701#ifdef CONFIG_XPS
702/*
703 * This structure holds an XPS map which can be of variable length. The
704 * map is an array of queues.
705 */
706struct xps_map {
707 unsigned int len;
708 unsigned int alloc_len;
709 struct rcu_head rcu;
710 u16 queues[0];
711};
60b778ce 712#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
713#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
714 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
715
716/*
717 * This structure holds all XPS maps for device. Maps are indexed by CPU.
718 */
719struct xps_dev_maps {
720 struct rcu_head rcu;
a4177869 721 struct xps_map __rcu *cpu_map[0];
bf264145 722};
184c449f
AD
723#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
724 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
725#endif /* CONFIG_XPS */
726
4f57c087
JF
727#define TC_MAX_QUEUE 16
728#define TC_BITMASK 15
729/* HW offloaded queuing disciplines txq count and offset maps */
730struct netdev_tc_txq {
731 u16 count;
732 u16 offset;
733};
734
68bad94e
NP
735#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
736/*
737 * This structure is to hold information about the device
738 * configured to run FCoE protocol stack.
739 */
740struct netdev_fcoe_hbainfo {
741 char manufacturer[64];
742 char serial_number[64];
743 char hardware_version[64];
744 char driver_version[64];
745 char optionrom_version[64];
746 char firmware_version[64];
747 char model[256];
748 char model_description[256];
749};
750#endif
751
02637fce 752#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 753
02637fce
JP
754/* This structure holds a unique identifier to identify some
755 * physical item (port for example) used by a netdevice.
66b52b0d 756 */
02637fce
JP
757struct netdev_phys_item_id {
758 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
759 unsigned char id_len;
760};
761
d754f98b
SF
762static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
763 struct netdev_phys_item_id *b)
764{
765 return a->id_len == b->id_len &&
766 memcmp(a->id, b->id, a->id_len) == 0;
767}
768
99932d4f
DB
769typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 struct sk_buff *skb);
771
2572ac53 772enum tc_setup_type {
16e5cc64 773 TC_SETUP_MQPRIO,
a1b7c5fd 774 TC_SETUP_CLSU32,
5b33f488 775 TC_SETUP_CLSFLOWER,
ade9b658 776 TC_SETUP_CLSMATCHALL,
332ae8e2 777 TC_SETUP_CLSBPF,
16e5cc64
JF
778};
779
a7862b45
BB
780/* These structures hold the attributes of xdp state that are being passed
781 * to the netdevice through the xdp op.
782 */
783enum xdp_netdev_command {
784 /* Set or clear a bpf program used in the earliest stages of packet
785 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
786 * is responsible for calling bpf_prog_put on any old progs that are
787 * stored. In case of error, the callee need not release the new prog
788 * reference, but on success it takes ownership and must bpf_prog_put
789 * when it is no longer used.
790 */
791 XDP_SETUP_PROG,
ee5d032f 792 XDP_SETUP_PROG_HW,
a7862b45 793 /* Check if a bpf program is set on the device. The callee should
ce158e58
JK
794 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
795 * is equivalent to XDP_ATTACHED_DRV.
a7862b45
BB
796 */
797 XDP_QUERY_PROG,
798};
799
ddf9f970
JK
800struct netlink_ext_ack;
801
a7862b45
BB
802struct netdev_xdp {
803 enum xdp_netdev_command command;
804 union {
805 /* XDP_SETUP_PROG */
ddf9f970 806 struct {
32d60277 807 u32 flags;
ddf9f970
JK
808 struct bpf_prog *prog;
809 struct netlink_ext_ack *extack;
810 };
a7862b45 811 /* XDP_QUERY_PROG */
58038695 812 struct {
ce158e58 813 u8 prog_attached;
58038695
MKL
814 u32 prog_id;
815 };
a7862b45
BB
816 };
817};
16e5cc64 818
d77e38e6
SK
819#ifdef CONFIG_XFRM_OFFLOAD
820struct xfrmdev_ops {
821 int (*xdo_dev_state_add) (struct xfrm_state *x);
822 void (*xdo_dev_state_delete) (struct xfrm_state *x);
823 void (*xdo_dev_state_free) (struct xfrm_state *x);
824 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
825 struct xfrm_state *x);
826};
827#endif
828
6c557001
FW
829struct dev_ifalias {
830 struct rcu_head rcuhead;
831 char ifalias[];
832};
833
d314774c
SH
834/*
835 * This structure defines the management hooks for network devices.
00829823
SH
836 * The following hooks can be defined; unless noted otherwise, they are
837 * optional and can be filled with a null pointer.
d314774c
SH
838 *
839 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
840 * This function is called once when a network device is registered.
841 * The network device can use this for any late stage initialization
842 * or semantic validation. It can fail with an error code which will
843 * be propagated back to register_netdev.
d314774c
SH
844 *
845 * void (*ndo_uninit)(struct net_device *dev);
846 * This function is called when device is unregistered or when registration
847 * fails. It is not called if init fails.
848 *
849 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 850 * This function is called when a network device transitions to the up
d314774c
SH
851 * state.
852 *
853 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 854 * This function is called when a network device transitions to the down
d314774c
SH
855 * state.
856 *
dc1f8bf6
SH
857 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
858 * struct net_device *dev);
00829823 859 * Called when a packet needs to be transmitted.
e79d8429
RR
860 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
861 * the queue before that can happen; it's for obsolete devices and weird
862 * corner cases, but the stack really does a non-trivial amount
863 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 864 * Required; cannot be NULL.
00829823 865 *
1a2a1444
DM
866 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
867 * struct net_device *dev
868 * netdev_features_t features);
869 * Called by core transmit path to determine if device is capable of
870 * performing offload operations on a given packet. This is to give
871 * the device an opportunity to implement any restrictions that cannot
872 * be otherwise expressed by feature flags. The check is called with
873 * the set of features that the stack has calculated and it returns
874 * those the driver believes to be appropriate.
cdba756f 875 *
f663dd9a 876 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 877 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 878 * Called to decide which queue to use when device supports multiple
00829823
SH
879 * transmit queues.
880 *
d314774c
SH
881 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
882 * This function is called to allow device receiver to make
5e82b4b2 883 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
884 *
885 * void (*ndo_set_rx_mode)(struct net_device *dev);
886 * This function is called device changes address list filtering.
01789349 887 * If driver handles unicast address filtering, it should set
5e82b4b2 888 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
889 *
890 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
891 * This function is called when the Media Access Control address
37b607c5 892 * needs to be changed. If this interface is not defined, the
5e82b4b2 893 * MAC address can not be changed.
d314774c
SH
894 *
895 * int (*ndo_validate_addr)(struct net_device *dev);
896 * Test if Media Access Control address is valid for the device.
897 *
898 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
899 * Called when a user requests an ioctl which can't be handled by
900 * the generic interface code. If not defined ioctls return
d314774c
SH
901 * not supported error code.
902 *
903 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
904 * Used to set network devices bus interface parameters. This interface
5e82b4b2 905 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
906 * interface (PCI) for low level management.
907 *
908 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
909 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 910 * of a device.
d314774c 911 *
00829823 912 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 913 * Callback used when the transmitter has not made any progress
d314774c
SH
914 * for dev->watchdog ticks.
915 *
bc1f4470 916 * void (*ndo_get_stats64)(struct net_device *dev,
917 * struct rtnl_link_stats64 *storage);
d308e38f 918 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 919 * Called when a user wants to get the network device usage
be1f3c2c 920 * statistics. Drivers must do one of the following:
3cfde79c
BH
921 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
922 * rtnl_link_stats64 structure passed by the caller.
82695d9b 923 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
924 * (which should normally be dev->stats) and return a pointer to
925 * it. The structure may be changed asynchronously only if each
926 * field is written atomically.
927 * 3. Update dev->stats asynchronously and atomically, and define
928 * neither operation.
d314774c 929 *
3df5b3c6 930 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
931 * Return true if this device supports offload stats of this attr_id.
932 *
933 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
934 * void *attr_data)
935 * Get statistics for offload operations by attr_id. Write it into the
936 * attr_data pointer.
937 *
5d632cb7 938 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 939 * If device supports VLAN filtering this function is called when a
80d5c368 940 * VLAN id is registered.
d314774c 941 *
5d632cb7 942 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 943 * If device supports VLAN filtering this function is called when a
80d5c368 944 * VLAN id is unregistered.
d314774c
SH
945 *
946 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
947 *
948 * SR-IOV management functions.
949 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
950 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
951 * u8 qos, __be16 proto);
ed616689
SC
952 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
953 * int max_tx_rate);
5f8444a3 954 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 955 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
956 * int (*ndo_get_vf_config)(struct net_device *dev,
957 * int vf, struct ifla_vf_info *ivf);
1d8faf48 958 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
959 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
960 * struct nlattr *port[]);
01a3d796
VZ
961 *
962 * Enable or disable the VF ability to query its RSS Redirection Table and
963 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 964 * with PF and querying it may introduce a theoretical security risk.
01a3d796 965 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 966 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
2572ac53 967 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
de4784ca 968 * void *type_data);
6a4bc2b4
FF
969 * Called to setup any 'tc' scheduler, classifier or action on @dev.
970 * This is always called from the stack with the rtnl lock held and netif
971 * tx queues stopped. This allows the netdevice to perform queue
972 * management safely.
c445477d 973 *
e9bce845
YZ
974 * Fiber Channel over Ethernet (FCoE) offload functions.
975 * int (*ndo_fcoe_enable)(struct net_device *dev);
976 * Called when the FCoE protocol stack wants to start using LLD for FCoE
977 * so the underlying device can perform whatever needed configuration or
978 * initialization to support acceleration of FCoE traffic.
979 *
980 * int (*ndo_fcoe_disable)(struct net_device *dev);
981 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
982 * so the underlying device can perform whatever needed clean-ups to
983 * stop supporting acceleration of FCoE traffic.
984 *
985 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
986 * struct scatterlist *sgl, unsigned int sgc);
987 * Called when the FCoE Initiator wants to initialize an I/O that
988 * is a possible candidate for Direct Data Placement (DDP). The LLD can
989 * perform necessary setup and returns 1 to indicate the device is set up
990 * successfully to perform DDP on this I/O, otherwise this returns 0.
991 *
992 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
993 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
994 * indicated by the FC exchange id 'xid', so the underlying device can
995 * clean up and reuse resources for later DDP requests.
996 *
997 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
998 * struct scatterlist *sgl, unsigned int sgc);
999 * Called when the FCoE Target wants to initialize an I/O that
1000 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1001 * perform necessary setup and returns 1 to indicate the device is set up
1002 * successfully to perform DDP on this I/O, otherwise this returns 0.
1003 *
68bad94e
NP
1004 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1005 * struct netdev_fcoe_hbainfo *hbainfo);
1006 * Called when the FCoE Protocol stack wants information on the underlying
1007 * device. This information is utilized by the FCoE protocol stack to
1008 * register attributes with Fiber Channel management service as per the
1009 * FC-GS Fabric Device Management Information(FDMI) specification.
1010 *
e9bce845
YZ
1011 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1012 * Called when the underlying device wants to override default World Wide
1013 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1014 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1015 * protocol stack to use.
1016 *
c445477d
BH
1017 * RFS acceleration.
1018 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1019 * u16 rxq_index, u32 flow_id);
1020 * Set hardware filter for RFS. rxq_index is the target queue index;
1021 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1022 * Return the filter ID on success, or a negative error code.
fbaec0ea 1023 *
8b98a70c 1024 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1025 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1026 * Called to make another netdev an underling.
1027 *
1028 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1029 * Called to release previously enslaved netdev.
5455c699
MM
1030 *
1031 * Feature/offload setting functions.
1a2a1444
DM
1032 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1033 * netdev_features_t features);
1034 * Adjusts the requested feature flags according to device-specific
1035 * constraints, and returns the resulting flags. Must not modify
1036 * the device state.
1037 *
c8f44aff 1038 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1039 * Called to update device configuration to new features. Passed
1040 * feature set might be less than what was returned by ndo_fix_features()).
1041 * Must return >0 or -errno if it changed dev->features itself.
1042 *
edc7d573 1043 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1044 * struct net_device *dev,
f6f6424b 1045 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1046 * Adds an FDB entry to dev for addr.
1690be63
VY
1047 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1048 * struct net_device *dev,
f6f6424b 1049 * const unsigned char *addr, u16 vid)
77162022
JF
1050 * Deletes the FDB entry from dev coresponding to addr.
1051 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1052 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1053 * int *idx)
77162022
JF
1054 * Used to add FDB entries to dump requests. Implementers should add
1055 * entries to skb and update idx with the number of entries.
e5a55a89 1056 *
ad41faa8
ND
1057 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1058 * u16 flags)
e5a55a89 1059 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1060 * struct net_device *dev, u32 filter_mask,
1061 * int nlflags)
ad41faa8
ND
1062 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1063 * u16 flags);
4bf84c35
JP
1064 *
1065 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1066 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1067 * which do not represent real hardware may define this to allow their
1068 * userspace components to manage their virtual carrier state. Devices
1069 * that determine carrier state from physical hardware properties (eg
1070 * network cables) or protocol-dependent mechanisms (eg
1071 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1072 *
1073 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1074 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1075 * Called to get ID of physical port of this device. If driver does
1076 * not implement this, it is assumed that the hw is not able to have
1077 * multiple net devices on single physical port.
53cf5275 1078 *
7c46a640
AD
1079 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1080 * struct udp_tunnel_info *ti);
1081 * Called by UDP tunnel to notify a driver about the UDP port and socket
1082 * address family that a UDP tunnel is listnening to. It is called only
1083 * when a new port starts listening. The operation is protected by the
1084 * RTNL.
1085 *
1086 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1087 * struct udp_tunnel_info *ti);
1088 * Called by UDP tunnel to notify the driver about a UDP port and socket
1089 * address family that the UDP tunnel is not listening to anymore. The
1090 * operation is protected by the RTNL.
1091 *
a6cc0cfa
JF
1092 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1093 * struct net_device *dev)
1094 * Called by upper layer devices to accelerate switching or other
1095 * station functionality into hardware. 'pdev is the lowerdev
1096 * to use for the offload and 'dev' is the net device that will
1097 * back the offload. Returns a pointer to the private structure
1098 * the upper layer will maintain.
1099 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1100 * Called by upper layer device to delete the station created
1101 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1102 * the station and priv is the structure returned by the add
1103 * operation.
822b3b2e
JF
1104 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1105 * int queue_index, u32 maxrate);
1106 * Called when a user wants to set a max-rate limitation of specific
1107 * TX queue.
a54acb3a
ND
1108 * int (*ndo_get_iflink)(const struct net_device *dev);
1109 * Called to get the iflink value of this device.
d746d707 1110 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1111 * bool proto_down);
d746d707
AK
1112 * This function is used to pass protocol port error state information
1113 * to the switch driver. The switch driver can react to the proto_down
1114 * by doing a phys down on the associated switch port.
fc4099f1
PS
1115 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1116 * This function is used to get egress tunnel information for given skb.
1117 * This is useful for retrieving outer tunnel header parameters while
1118 * sampling packet.
871b642a
PA
1119 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1120 * This function is used to specify the headroom that the skb must
1121 * consider when allocation skb during packet reception. Setting
1122 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1123 * forward. Setting a negative value resets the rx headroom to the
871b642a 1124 * default value.
a7862b45
BB
1125 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1126 * This function is used to set or query state related to XDP on the
1127 * netdevice. See definition of enum xdp_netdev_command for details.
814abfab
JF
1128 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1129 * This function is used to submit a XDP packet for transmit on a
1130 * netdevice.
11393cc9 1131 * void (*ndo_xdp_flush)(struct net_device *dev);
f16ded59
RR
1132 * This function is used to inform the driver to flush a particular
1133 * xdp tx queue. Must be called on same CPU as xdp_xmit.
d314774c
SH
1134 */
1135struct net_device_ops {
1136 int (*ndo_init)(struct net_device *dev);
1137 void (*ndo_uninit)(struct net_device *dev);
1138 int (*ndo_open)(struct net_device *dev);
1139 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1140 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1141 struct net_device *dev);
1142 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1143 struct net_device *dev,
1144 netdev_features_t features);
00829823 1145 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1146 struct sk_buff *skb,
99932d4f
DB
1147 void *accel_priv,
1148 select_queue_fallback_t fallback);
d314774c
SH
1149 void (*ndo_change_rx_flags)(struct net_device *dev,
1150 int flags);
d314774c 1151 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1152 int (*ndo_set_mac_address)(struct net_device *dev,
1153 void *addr);
d314774c 1154 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1155 int (*ndo_do_ioctl)(struct net_device *dev,
1156 struct ifreq *ifr, int cmd);
d314774c
SH
1157 int (*ndo_set_config)(struct net_device *dev,
1158 struct ifmap *map);
00829823
SH
1159 int (*ndo_change_mtu)(struct net_device *dev,
1160 int new_mtu);
1161 int (*ndo_neigh_setup)(struct net_device *dev,
1162 struct neigh_parms *);
d314774c
SH
1163 void (*ndo_tx_timeout) (struct net_device *dev);
1164
bc1f4470 1165 void (*ndo_get_stats64)(struct net_device *dev,
1166 struct rtnl_link_stats64 *storage);
3df5b3c6 1167 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1168 int (*ndo_get_offload_stats)(int attr_id,
1169 const struct net_device *dev,
1170 void *attr_data);
d314774c
SH
1171 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1172
8e586137 1173 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1174 __be16 proto, u16 vid);
8e586137 1175 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1176 __be16 proto, u16 vid);
d314774c 1177#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1178 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1179 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1180 struct netpoll_info *info);
0e34e931 1181 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1182#endif
95c26df8
WM
1183 int (*ndo_set_vf_mac)(struct net_device *dev,
1184 int queue, u8 *mac);
1185 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1186 int queue, u16 vlan,
1187 u8 qos, __be16 proto);
ed616689
SC
1188 int (*ndo_set_vf_rate)(struct net_device *dev,
1189 int vf, int min_tx_rate,
1190 int max_tx_rate);
5f8444a3
GR
1191 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1192 int vf, bool setting);
dd461d6a
HS
1193 int (*ndo_set_vf_trust)(struct net_device *dev,
1194 int vf, bool setting);
95c26df8
WM
1195 int (*ndo_get_vf_config)(struct net_device *dev,
1196 int vf,
1197 struct ifla_vf_info *ivf);
1d8faf48
RE
1198 int (*ndo_set_vf_link_state)(struct net_device *dev,
1199 int vf, int link_state);
3b766cd8
EBE
1200 int (*ndo_get_vf_stats)(struct net_device *dev,
1201 int vf,
1202 struct ifla_vf_stats
1203 *vf_stats);
57b61080
SF
1204 int (*ndo_set_vf_port)(struct net_device *dev,
1205 int vf,
1206 struct nlattr *port[]);
1207 int (*ndo_get_vf_port)(struct net_device *dev,
1208 int vf, struct sk_buff *skb);
cc8e27cc
EC
1209 int (*ndo_set_vf_guid)(struct net_device *dev,
1210 int vf, u64 guid,
1211 int guid_type);
01a3d796
VZ
1212 int (*ndo_set_vf_rss_query_en)(
1213 struct net_device *dev,
1214 int vf, bool setting);
16e5cc64 1215 int (*ndo_setup_tc)(struct net_device *dev,
2572ac53 1216 enum tc_setup_type type,
de4784ca 1217 void *type_data);
d11ead75 1218#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1219 int (*ndo_fcoe_enable)(struct net_device *dev);
1220 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1221 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1222 u16 xid,
1223 struct scatterlist *sgl,
1224 unsigned int sgc);
1225 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1226 u16 xid);
6247e086
YZ
1227 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1228 u16 xid,
1229 struct scatterlist *sgl,
1230 unsigned int sgc);
68bad94e
NP
1231 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1232 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1233#endif
1234
d11ead75 1235#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1236#define NETDEV_FCOE_WWNN 0
1237#define NETDEV_FCOE_WWPN 1
1238 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1239 u64 *wwn, int type);
4d288d57 1240#endif
3c9c36bc 1241
c445477d
BH
1242#ifdef CONFIG_RFS_ACCEL
1243 int (*ndo_rx_flow_steer)(struct net_device *dev,
1244 const struct sk_buff *skb,
1245 u16 rxq_index,
1246 u32 flow_id);
1247#endif
fbaec0ea 1248 int (*ndo_add_slave)(struct net_device *dev,
33eaf2a6
DA
1249 struct net_device *slave_dev,
1250 struct netlink_ext_ack *extack);
fbaec0ea
JP
1251 int (*ndo_del_slave)(struct net_device *dev,
1252 struct net_device *slave_dev);
c8f44aff
MM
1253 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1254 netdev_features_t features);
5455c699 1255 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1256 netdev_features_t features);
503eebc2
JP
1257 int (*ndo_neigh_construct)(struct net_device *dev,
1258 struct neighbour *n);
1259 void (*ndo_neigh_destroy)(struct net_device *dev,
1260 struct neighbour *n);
77162022
JF
1261
1262 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1263 struct nlattr *tb[],
77162022 1264 struct net_device *dev,
6b6e2725 1265 const unsigned char *addr,
f6f6424b 1266 u16 vid,
77162022
JF
1267 u16 flags);
1268 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1269 struct nlattr *tb[],
77162022 1270 struct net_device *dev,
f6f6424b
JP
1271 const unsigned char *addr,
1272 u16 vid);
77162022
JF
1273 int (*ndo_fdb_dump)(struct sk_buff *skb,
1274 struct netlink_callback *cb,
1275 struct net_device *dev,
5d5eacb3 1276 struct net_device *filter_dev,
d297653d 1277 int *idx);
e5a55a89
JF
1278
1279 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1280 struct nlmsghdr *nlh,
1281 u16 flags);
e5a55a89
JF
1282 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1283 u32 pid, u32 seq,
6cbdceeb 1284 struct net_device *dev,
46c264da
ND
1285 u32 filter_mask,
1286 int nlflags);
407af329 1287 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1288 struct nlmsghdr *nlh,
1289 u16 flags);
4bf84c35
JP
1290 int (*ndo_change_carrier)(struct net_device *dev,
1291 bool new_carrier);
66b52b0d 1292 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1293 struct netdev_phys_item_id *ppid);
db24a904
DA
1294 int (*ndo_get_phys_port_name)(struct net_device *dev,
1295 char *name, size_t len);
7c46a640
AD
1296 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1297 struct udp_tunnel_info *ti);
1298 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1299 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1300 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1301 struct net_device *dev);
1302 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1303 void *priv);
1304
25175ba5 1305 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1306 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1307 int queue_index,
1308 u32 maxrate);
a54acb3a 1309 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1310 int (*ndo_change_proto_down)(struct net_device *dev,
1311 bool proto_down);
fc4099f1
PS
1312 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1313 struct sk_buff *skb);
871b642a
PA
1314 void (*ndo_set_rx_headroom)(struct net_device *dev,
1315 int needed_headroom);
a7862b45
BB
1316 int (*ndo_xdp)(struct net_device *dev,
1317 struct netdev_xdp *xdp);
814abfab
JF
1318 int (*ndo_xdp_xmit)(struct net_device *dev,
1319 struct xdp_buff *xdp);
11393cc9 1320 void (*ndo_xdp_flush)(struct net_device *dev);
d314774c
SH
1321};
1322
7aa98047
LR
1323/**
1324 * enum net_device_priv_flags - &struct net_device priv_flags
1325 *
1326 * These are the &struct net_device, they are only set internally
1327 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1328 * userspace; this means that the order of these flags can change
7aa98047
LR
1329 * during any kernel release.
1330 *
1331 * You should have a pretty good reason to be extending these flags.
1332 *
1333 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1334 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1335 * @IFF_BONDING: bonding master or slave
7aa98047 1336 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1337 * @IFF_WAN_HDLC: WAN HDLC device
1338 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1339 * release skb->dst
1340 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1341 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1342 * @IFF_MACVLAN_PORT: device used as macvlan port
1343 * @IFF_BRIDGE_PORT: device used as bridge port
1344 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1345 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1346 * @IFF_UNICAST_FLT: Supports unicast filtering
1347 * @IFF_TEAM_PORT: device used as team port
1348 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1349 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1350 * change when it's running
1351 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1352 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1353 * underlying stacked devices
1354 * @IFF_IPVLAN_MASTER: IPvlan master device
1355 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1356 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1357 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1358 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1359 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1360 * @IFF_TEAM: device is a team device
d4ab4286 1361 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1362 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1363 * entity (i.e. the master device for bridged veth)
3c175784 1364 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1365 */
1366enum netdev_priv_flags {
1367 IFF_802_1Q_VLAN = 1<<0,
1368 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1369 IFF_BONDING = 1<<2,
1370 IFF_ISATAP = 1<<3,
1371 IFF_WAN_HDLC = 1<<4,
1372 IFF_XMIT_DST_RELEASE = 1<<5,
1373 IFF_DONT_BRIDGE = 1<<6,
1374 IFF_DISABLE_NETPOLL = 1<<7,
1375 IFF_MACVLAN_PORT = 1<<8,
1376 IFF_BRIDGE_PORT = 1<<9,
1377 IFF_OVS_DATAPATH = 1<<10,
1378 IFF_TX_SKB_SHARING = 1<<11,
1379 IFF_UNICAST_FLT = 1<<12,
1380 IFF_TEAM_PORT = 1<<13,
1381 IFF_SUPP_NOFCS = 1<<14,
1382 IFF_LIVE_ADDR_CHANGE = 1<<15,
1383 IFF_MACVLAN = 1<<16,
1384 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1385 IFF_IPVLAN_MASTER = 1<<18,
1386 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1387 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1388 IFF_NO_QUEUE = 1<<21,
1389 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1390 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1391 IFF_TEAM = 1<<24,
d4ab4286 1392 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1393 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1394 IFF_MACSEC = 1<<27,
7aa98047
LR
1395};
1396
1397#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1398#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1399#define IFF_BONDING IFF_BONDING
7aa98047 1400#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1401#define IFF_WAN_HDLC IFF_WAN_HDLC
1402#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1403#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1404#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1405#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1406#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1407#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1408#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1409#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1410#define IFF_TEAM_PORT IFF_TEAM_PORT
1411#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1412#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1413#define IFF_MACVLAN IFF_MACVLAN
02875878 1414#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1415#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1416#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1417#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1418#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1419#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1420#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1421#define IFF_TEAM IFF_TEAM
d4ab4286 1422#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1423#define IFF_MACSEC IFF_MACSEC
7aa98047 1424
536721b1
KK
1425/**
1426 * struct net_device - The DEVICE structure.
d651983d
MCC
1427 *
1428 * Actually, this whole structure is a big mistake. It mixes I/O
1429 * data with strictly "high-level" data, and it has to know about
1430 * almost every data structure used in the INET module.
536721b1
KK
1431 *
1432 * @name: This is the first field of the "visible" part of this structure
1433 * (i.e. as seen by users in the "Space.c" file). It is the name
d651983d 1434 * of the interface.
536721b1
KK
1435 *
1436 * @name_hlist: Device name hash chain, please keep it close to name[]
1437 * @ifalias: SNMP alias
1438 * @mem_end: Shared memory end
1439 * @mem_start: Shared memory start
1440 * @base_addr: Device I/O address
1441 * @irq: Device IRQ number
1442 *
14ffbbb8
TG
1443 * @carrier_changes: Stats to monitor carrier on<->off transitions
1444 *
536721b1
KK
1445 * @state: Generic network queuing layer state, see netdev_state_t
1446 * @dev_list: The global list of network devices
5e82b4b2
BH
1447 * @napi_list: List entry used for polling NAPI devices
1448 * @unreg_list: List entry when we are unregistering the
1449 * device; see the function unregister_netdev
1450 * @close_list: List entry used when we are closing the device
62d885fe
BP
1451 * @ptype_all: Device-specific packet handlers for all protocols
1452 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1453 *
1454 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1455 * @features: Currently active device features
1456 * @hw_features: User-changeable features
1457 *
1458 * @wanted_features: User-requested features
1459 * @vlan_features: Mask of features inheritable by VLAN devices
1460 *
1461 * @hw_enc_features: Mask of features inherited by encapsulating devices
1462 * This field indicates what encapsulation
1463 * offloads the hardware is capable of doing,
1464 * and drivers will need to set them appropriately.
1465 *
1466 * @mpls_features: Mask of features inheritable by MPLS
1467 *
1468 * @ifindex: interface index
5e82b4b2 1469 * @group: The group the device belongs to
536721b1
KK
1470 *
1471 * @stats: Statistics struct, which was left as a legacy, use
1472 * rtnl_link_stats64 instead
1473 *
1474 * @rx_dropped: Dropped packets by core network,
1475 * do not use this in drivers
1476 * @tx_dropped: Dropped packets by core network,
1477 * do not use this in drivers
6e7333d3
JW
1478 * @rx_nohandler: nohandler dropped packets by core network on
1479 * inactive devices, do not use this in drivers
536721b1 1480 *
536721b1
KK
1481 * @wireless_handlers: List of functions to handle Wireless Extensions,
1482 * instead of ioctl,
1483 * see <net/iw_handler.h> for details.
1484 * @wireless_data: Instance data managed by the core of wireless extensions
1485 *
1486 * @netdev_ops: Includes several pointers to callbacks,
1487 * if one wants to override the ndo_*() functions
1488 * @ethtool_ops: Management operations
f997c55c
AA
1489 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1490 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1491 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1492 * of Layer 2 headers.
1493 *
1494 * @flags: Interface flags (a la BSD)
1495 * @priv_flags: Like 'flags' but invisible to userspace,
1496 * see if.h for the definitions
1497 * @gflags: Global flags ( kept as legacy )
1498 * @padded: How much padding added by alloc_netdev()
1499 * @operstate: RFC2863 operstate
1500 * @link_mode: Mapping policy to operstate
1501 * @if_port: Selectable AUI, TP, ...
1502 * @dma: DMA channel
1503 * @mtu: Interface MTU value
61e84623
JW
1504 * @min_mtu: Interface Minimum MTU value
1505 * @max_mtu: Interface Maximum MTU value
536721b1 1506 * @type: Interface hardware type
2793a23a 1507 * @hard_header_len: Maximum hardware header length.
217e6fa2 1508 * @min_header_len: Minimum hardware header length
536721b1
KK
1509 *
1510 * @needed_headroom: Extra headroom the hardware may need, but not in all
1511 * cases can this be guaranteed
1512 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1513 * cases can this be guaranteed. Some cases also use
1514 * LL_MAX_HEADER instead to allocate the skb
1515 *
1516 * interface address info:
1517 *
1518 * @perm_addr: Permanent hw address
1519 * @addr_assign_type: Hw address assignment type
1520 * @addr_len: Hardware address length
8626a0c8 1521 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1522 * @dev_id: Used to differentiate devices that share
1523 * the same link layer address
1524 * @dev_port: Used to differentiate devices that share
1525 * the same function
1526 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1527 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1528 * has been enabled due to the need to listen to
1529 * additional unicast addresses in a device that
1530 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1531 * @uc: unicast mac addresses
1532 * @mc: multicast mac addresses
1533 * @dev_addrs: list of device hw addresses
1534 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1535 * @promiscuity: Number of times the NIC is told to work in
1536 * promiscuous mode; if it becomes 0 the NIC will
1537 * exit promiscuous mode
536721b1
KK
1538 * @allmulti: Counter, enables or disables allmulticast mode
1539 *
1540 * @vlan_info: VLAN info
1541 * @dsa_ptr: dsa specific data
1542 * @tipc_ptr: TIPC specific data
1543 * @atalk_ptr: AppleTalk link
1544 * @ip_ptr: IPv4 specific data
1545 * @dn_ptr: DECnet specific data
1546 * @ip6_ptr: IPv6 specific data
1547 * @ax25_ptr: AX.25 specific data
1548 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1549 *
536721b1
KK
1550 * @dev_addr: Hw address (before bcast,
1551 * because most packets are unicast)
1552 *
1553 * @_rx: Array of RX queues
1554 * @num_rx_queues: Number of RX queues
1555 * allocated at register_netdev() time
1556 * @real_num_rx_queues: Number of RX queues currently active in device
1557 *
1558 * @rx_handler: handler for received packets
1559 * @rx_handler_data: XXX: need comments on this one
1560 * @ingress_queue: XXX: need comments on this one
1561 * @broadcast: hw bcast address
1562 *
14ffbbb8
TG
1563 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1564 * indexed by RX queue number. Assigned by driver.
1565 * This must only be set if the ndo_rx_flow_steer
1566 * operation is defined
1567 * @index_hlist: Device index hash chain
1568 *
536721b1
KK
1569 * @_tx: Array of TX queues
1570 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1571 * @real_num_tx_queues: Number of TX queues currently active in device
1572 * @qdisc: Root qdisc from userspace point of view
1573 * @tx_queue_len: Max frames per queue allowed
1574 * @tx_global_lock: XXX: need comments on this one
1575 *
1576 * @xps_maps: XXX: need comments on this one
1577 *
536721b1 1578 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1579 * the watchdog (see dev_watchdog())
536721b1
KK
1580 * @watchdog_timer: List of timers
1581 *
1582 * @pcpu_refcnt: Number of references to this device
1583 * @todo_list: Delayed register/unregister
536721b1
KK
1584 * @link_watch_list: XXX: need comments on this one
1585 *
1586 * @reg_state: Register/unregister state machine
1587 * @dismantle: Device is going to be freed
1588 * @rtnl_link_state: This enum represents the phases of creating
1589 * a new link
1590 *
cf124db5
DM
1591 * @needs_free_netdev: Should unregister perform free_netdev?
1592 * @priv_destructor: Called from unregister
536721b1
KK
1593 * @npinfo: XXX: need comments on this one
1594 * @nd_net: Network namespace this network device is inside
1595 *
1596 * @ml_priv: Mid-layer private
1597 * @lstats: Loopback statistics
1598 * @tstats: Tunnel statistics
1599 * @dstats: Dummy statistics
1600 * @vstats: Virtual ethernet statistics
1601 *
1602 * @garp_port: GARP
1603 * @mrp_port: MRP
1604 *
1605 * @dev: Class/net/name entry
1606 * @sysfs_groups: Space for optional device, statistics and wireless
1607 * sysfs groups
1608 *
1609 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1610 * @rtnl_link_ops: Rtnl_link_ops
1611 *
1612 * @gso_max_size: Maximum size of generic segmentation offload
1613 * @gso_max_segs: Maximum number of segments that can be passed to the
1614 * NIC for GSO
1615 *
1616 * @dcbnl_ops: Data Center Bridging netlink ops
1617 * @num_tc: Number of traffic classes in the net device
1618 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1619 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1620 *
1621 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1622 *
1623 * @priomap: XXX: need comments on this one
1624 * @phydev: Physical device may attach itself
1625 * for hardware timestamping
1626 *
123b3652
ED
1627 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1628 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1629 *
d746d707
AK
1630 * @proto_down: protocol port state information can be sent to the
1631 * switch driver and used to set the phys state of the
1632 * switch port.
1633 *
1da177e4
LT
1634 * FIXME: cleanup struct net_device such that network protocol info
1635 * moves out.
1636 */
1637
d94d9fee 1638struct net_device {
1da177e4 1639 char name[IFNAMSIZ];
9356b8fc 1640 struct hlist_node name_hlist;
6c557001 1641 struct dev_ifalias __rcu *ifalias;
1da177e4
LT
1642 /*
1643 * I/O specific fields
1644 * FIXME: Merge these and struct ifmap into one
1645 */
536721b1
KK
1646 unsigned long mem_end;
1647 unsigned long mem_start;
1648 unsigned long base_addr;
1649 int irq;
1da177e4 1650
14ffbbb8
TG
1651 atomic_t carrier_changes;
1652
1da177e4 1653 /*
536721b1
KK
1654 * Some hardware also needs these fields (state,dev_list,
1655 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1656 * part of the usual set specified in Space.c.
1657 */
1658
1da177e4
LT
1659 unsigned long state;
1660
7562f876 1661 struct list_head dev_list;
bea3348e 1662 struct list_head napi_list;
44a0873d 1663 struct list_head unreg_list;
5cde2829 1664 struct list_head close_list;
7866a621
SN
1665 struct list_head ptype_all;
1666 struct list_head ptype_specific;
2f268f12 1667
2f268f12
VF
1668 struct {
1669 struct list_head upper;
1670 struct list_head lower;
1671 } adj_list;
1672
c8f44aff 1673 netdev_features_t features;
c8f44aff 1674 netdev_features_t hw_features;
c8f44aff 1675 netdev_features_t wanted_features;
c8f44aff 1676 netdev_features_t vlan_features;
6a674e9c 1677 netdev_features_t hw_enc_features;
0d89d203 1678 netdev_features_t mpls_features;
802ab55a 1679 netdev_features_t gso_partial_features;
04ed3e74 1680
1da177e4 1681 int ifindex;
7a66bbc9 1682 int group;
1da177e4 1683
c45d286e 1684 struct net_device_stats stats;
015f0688 1685
015f0688
ED
1686 atomic_long_t rx_dropped;
1687 atomic_long_t tx_dropped;
6e7333d3 1688 atomic_long_t rx_nohandler;
1da177e4 1689
b86e0280 1690#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1691 const struct iw_handler_def *wireless_handlers;
1692 struct iw_public_data *wireless_data;
b86e0280 1693#endif
d314774c 1694 const struct net_device_ops *netdev_ops;
76fd8593 1695 const struct ethtool_ops *ethtool_ops;
4170604f 1696#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1697 const struct switchdev_ops *switchdev_ops;
4170604f 1698#endif
1b69c6d0
DA
1699#ifdef CONFIG_NET_L3_MASTER_DEV
1700 const struct l3mdev_ops *l3mdev_ops;
1701#endif
f997c55c
AA
1702#if IS_ENABLED(CONFIG_IPV6)
1703 const struct ndisc_ops *ndisc_ops;
1704#endif
1da177e4 1705
d77e38e6
SK
1706#ifdef CONFIG_XFRM
1707 const struct xfrmdev_ops *xfrmdev_ops;
1708#endif
1709
3b04ddde
SH
1710 const struct header_ops *header_ops;
1711
536721b1
KK
1712 unsigned int flags;
1713 unsigned int priv_flags;
1714
1da177e4 1715 unsigned short gflags;
536721b1 1716 unsigned short padded;
1da177e4 1717
536721b1
KK
1718 unsigned char operstate;
1719 unsigned char link_mode;
b00055aa 1720
536721b1
KK
1721 unsigned char if_port;
1722 unsigned char dma;
bdc220da 1723
536721b1 1724 unsigned int mtu;
61e84623
JW
1725 unsigned int min_mtu;
1726 unsigned int max_mtu;
536721b1
KK
1727 unsigned short type;
1728 unsigned short hard_header_len;
d92be7a4 1729 unsigned char min_header_len;
1da177e4 1730
f5184d26
JB
1731 unsigned short needed_headroom;
1732 unsigned short needed_tailroom;
1733
1da177e4 1734 /* Interface address info. */
536721b1
KK
1735 unsigned char perm_addr[MAX_ADDR_LEN];
1736 unsigned char addr_assign_type;
1737 unsigned char addr_len;
a0a9663d 1738 unsigned short neigh_priv_len;
536721b1
KK
1739 unsigned short dev_id;
1740 unsigned short dev_port;
ccffad25 1741 spinlock_t addr_list_lock;
14ffbbb8
TG
1742 unsigned char name_assign_type;
1743 bool uc_promisc;
536721b1
KK
1744 struct netdev_hw_addr_list uc;
1745 struct netdev_hw_addr_list mc;
1746 struct netdev_hw_addr_list dev_addrs;
1747
4c3d5e7b
ED
1748#ifdef CONFIG_SYSFS
1749 struct kset *queues_kset;
1750#endif
9d45abe1
WC
1751 unsigned int promiscuity;
1752 unsigned int allmulti;
1da177e4 1753
1da177e4 1754
5e82b4b2 1755 /* Protocol-specific pointers */
65ac6a5f 1756
d11ead75 1757#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1758 struct vlan_info __rcu *vlan_info;
65ac6a5f 1759#endif
34a430d7 1760#if IS_ENABLED(CONFIG_NET_DSA)
2f657a60 1761 struct dsa_port *dsa_ptr;
37cb0620
YX
1762#endif
1763#if IS_ENABLED(CONFIG_TIPC)
536721b1 1764 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1765#endif
536721b1
KK
1766 void *atalk_ptr;
1767 struct in_device __rcu *ip_ptr;
1768 struct dn_dev __rcu *dn_ptr;
1769 struct inet6_dev __rcu *ip6_ptr;
1770 void *ax25_ptr;
1771 struct wireless_dev *ieee80211_ptr;
98a18b6f 1772 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1773#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1774 struct mpls_dev __rcu *mpls_ptr;
1775#endif
1da177e4 1776
9356b8fc 1777/*
cd13539b 1778 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1779 */
9356b8fc 1780 /* Interface address info used in eth_type_trans() */
536721b1 1781 unsigned char *dev_addr;
f001fde5 1782
a953be53 1783#ifdef CONFIG_SYSFS
0a9627f2
TH
1784 struct netdev_rx_queue *_rx;
1785
0a9627f2 1786 unsigned int num_rx_queues;
62fe0b40 1787 unsigned int real_num_rx_queues;
df334545 1788#endif
0a9627f2 1789
7acedaf5 1790 struct bpf_prog __rcu *xdp_prog;
3b47d303 1791 unsigned long gro_flush_timeout;
61391cde 1792 rx_handler_func_t __rcu *rx_handler;
1793 void __rcu *rx_handler_data;
e8a0464c 1794
4cda01e8 1795#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1796 struct tcf_proto __rcu *ingress_cl_list;
1797#endif
24824a09 1798 struct netdev_queue __rcu *ingress_queue;
e687ad60 1799#ifdef CONFIG_NETFILTER_INGRESS
960632ec 1800 struct nf_hook_entries __rcu *nf_hooks_ingress;
e687ad60 1801#endif
d2788d34 1802
536721b1 1803 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1804#ifdef CONFIG_RFS_ACCEL
1805 struct cpu_rmap *rx_cpu_rmap;
1806#endif
1807 struct hlist_node index_hlist;
cd13539b
ED
1808
1809/*
1810 * Cache lines mostly used on transmit path
1811 */
e8a0464c
DM
1812 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1813 unsigned int num_tx_queues;
fd2ea0a7 1814 unsigned int real_num_tx_queues;
af356afa 1815 struct Qdisc *qdisc;
59cc1f61
JK
1816#ifdef CONFIG_NET_SCHED
1817 DECLARE_HASHTABLE (qdisc_hash, 4);
1818#endif
0cd29503 1819 unsigned int tx_queue_len;
c3f26a26 1820 spinlock_t tx_global_lock;
14ffbbb8 1821 int watchdog_timeo;
cd13539b 1822
bf264145 1823#ifdef CONFIG_XPS
a4177869 1824 struct xps_dev_maps __rcu *xps_maps;
bf264145 1825#endif
1f211a1b
DB
1826#ifdef CONFIG_NET_CLS_ACT
1827 struct tcf_proto __rcu *egress_cl_list;
1828#endif
0c4f691f 1829
9356b8fc 1830 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1831 struct timer_list watchdog_timer;
1832
29b4433d 1833 int __percpu *pcpu_refcnt;
1da177e4 1834 struct list_head todo_list;
1da177e4 1835
e014debe 1836 struct list_head link_watch_list;
572a103d 1837
1da177e4 1838 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1839 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1840 NETREG_UNREGISTERING, /* called unregister_netdevice */
1841 NETREG_UNREGISTERED, /* completed unregister todo */
1842 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1843 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1844 } reg_state:8;
1845
536721b1 1846 bool dismantle;
a2835763
PM
1847
1848 enum {
1849 RTNL_LINK_INITIALIZED,
1850 RTNL_LINK_INITIALIZING,
1851 } rtnl_link_state:16;
1da177e4 1852
cf124db5
DM
1853 bool needs_free_netdev;
1854 void (*priv_destructor)(struct net_device *dev);
1da177e4 1855
1da177e4 1856#ifdef CONFIG_NETPOLL
5fbee843 1857 struct netpoll_info __rcu *npinfo;
1da177e4 1858#endif
eae792b7 1859
0c5c9fb5 1860 possible_net_t nd_net;
4a1c5371 1861
4951704b 1862 /* mid-layer private */
a7855c78 1863 union {
536721b1
KK
1864 void *ml_priv;
1865 struct pcpu_lstats __percpu *lstats;
8f84985f 1866 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1867 struct pcpu_dstats __percpu *dstats;
1868 struct pcpu_vstats __percpu *vstats;
a7855c78 1869 };
536721b1 1870
fb585b44 1871#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1872 struct garp_port __rcu *garp_port;
fb585b44
TK
1873#endif
1874#if IS_ENABLED(CONFIG_MRP)
febf018d 1875 struct mrp_port __rcu *mrp_port;
fb585b44 1876#endif
1da177e4 1877
5e82b4b2 1878 struct device dev;
0c509a6c 1879 const struct attribute_group *sysfs_groups[4];
a953be53 1880 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1881
38f7b870 1882 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1883
82cc1a7a
PWJ
1884 /* for setting kernel sock attribute on TCP connection setup */
1885#define GSO_MAX_SIZE 65536
1886 unsigned int gso_max_size;
30b678d8
BH
1887#define GSO_MAX_SEGS 65535
1888 u16 gso_max_segs;
743b03a8 1889
7a6b6f51 1890#ifdef CONFIG_DCB
32953543 1891 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1892#endif
5e82b4b2
BH
1893 u8 num_tc;
1894 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1895 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1896
d11ead75 1897#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1898 unsigned int fcoe_ddp_xid;
5bc1421e 1899#endif
86f8515f 1900#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1901 struct netprio_map __rcu *priomap;
4d288d57 1902#endif
5e82b4b2
BH
1903 struct phy_device *phydev;
1904 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1905 struct lock_class_key *qdisc_running_key;
5e82b4b2 1906 bool proto_down;
1da177e4 1907};
43cb76d9 1908#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1909
b5cdae32
DM
1910static inline bool netif_elide_gro(const struct net_device *dev)
1911{
1912 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1913 return true;
1914 return false;
1915}
1916
1da177e4 1917#define NETDEV_ALIGN 32
1da177e4 1918
4f57c087
JF
1919static inline
1920int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1921{
1922 return dev->prio_tc_map[prio & TC_BITMASK];
1923}
1924
1925static inline
1926int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1927{
1928 if (tc >= dev->num_tc)
1929 return -EINVAL;
1930
1931 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1932 return 0;
1933}
1934
8d059b0f 1935int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1936void netdev_reset_tc(struct net_device *dev);
1937int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1938int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1939
1940static inline
1941int netdev_get_num_tc(struct net_device *dev)
1942{
1943 return dev->num_tc;
1944}
1945
e8a0464c
DM
1946static inline
1947struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1948 unsigned int index)
1949{
1950 return &dev->_tx[index];
1951}
1952
10c51b56
DB
1953static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1954 const struct sk_buff *skb)
1955{
1956 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1957}
1958
e8a0464c
DM
1959static inline void netdev_for_each_tx_queue(struct net_device *dev,
1960 void (*f)(struct net_device *,
1961 struct netdev_queue *,
1962 void *),
1963 void *arg)
1964{
1965 unsigned int i;
1966
1967 for (i = 0; i < dev->num_tx_queues; i++)
1968 f(dev, &dev->_tx[i], arg);
1969}
1970
d3fff6c4
ED
1971#define netdev_lockdep_set_classes(dev) \
1972{ \
1973 static struct lock_class_key qdisc_tx_busylock_key; \
1974 static struct lock_class_key qdisc_running_key; \
1975 static struct lock_class_key qdisc_xmit_lock_key; \
1976 static struct lock_class_key dev_addr_list_lock_key; \
1977 unsigned int i; \
1978 \
1979 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1980 (dev)->qdisc_running_key = &qdisc_running_key; \
1981 lockdep_set_class(&(dev)->addr_list_lock, \
1982 &dev_addr_list_lock_key); \
1983 for (i = 0; i < (dev)->num_tx_queues; i++) \
1984 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1985 &qdisc_xmit_lock_key); \
1986}
1987
f629d208 1988struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1989 struct sk_buff *skb,
1990 void *accel_priv);
8c4c49df 1991
871b642a
PA
1992/* returns the headroom that the master device needs to take in account
1993 * when forwarding to this dev
1994 */
1995static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1996{
1997 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1998}
1999
2000static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2001{
2002 if (dev->netdev_ops->ndo_set_rx_headroom)
2003 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2004}
2005
2006/* set the device rx headroom to the dev's default */
2007static inline void netdev_reset_rx_headroom(struct net_device *dev)
2008{
2009 netdev_set_rx_headroom(dev, -1);
2010}
2011
c346dca1
YH
2012/*
2013 * Net namespace inlines
2014 */
2015static inline
2016struct net *dev_net(const struct net_device *dev)
2017{
c2d9ba9b 2018 return read_pnet(&dev->nd_net);
c346dca1
YH
2019}
2020
2021static inline
f5aa23fd 2022void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2023{
0c5c9fb5 2024 write_pnet(&dev->nd_net, net);
c346dca1
YH
2025}
2026
bea3348e
SH
2027/**
2028 * netdev_priv - access network device private data
2029 * @dev: network device
2030 *
2031 * Get network device private data
2032 */
6472ce60 2033static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2034{
1ce8e7b5 2035 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2036}
2037
1da177e4
LT
2038/* Set the sysfs physical device reference for the network logical device
2039 * if set prior to registration will cause a symlink during initialization.
2040 */
43cb76d9 2041#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2042
384912ed 2043/* Set the sysfs device type for the network logical device to allow
3f79410c 2044 * fine-grained identification of different network device types. For
5e82b4b2 2045 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2046 */
2047#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2048
82dc3c63
ED
2049/* Default NAPI poll() weight
2050 * Device drivers are strongly advised to not use bigger value
2051 */
2052#define NAPI_POLL_WEIGHT 64
2053
3b582cc1 2054/**
5e82b4b2 2055 * netif_napi_add - initialize a NAPI context
3b582cc1 2056 * @dev: network device
5e82b4b2 2057 * @napi: NAPI context
3b582cc1
SH
2058 * @poll: polling function
2059 * @weight: default weight
2060 *
5e82b4b2
BH
2061 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2062 * *any* of the other NAPI-related functions.
3b582cc1 2063 */
d565b0a1
HX
2064void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2065 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2066
d64b5e85 2067/**
5e82b4b2 2068 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2069 * @dev: network device
5e82b4b2 2070 * @napi: NAPI context
d64b5e85
ED
2071 * @poll: polling function
2072 * @weight: default weight
2073 *
2074 * This variant of netif_napi_add() should be used from drivers using NAPI
2075 * to exclusively poll a TX queue.
2076 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2077 */
2078static inline void netif_tx_napi_add(struct net_device *dev,
2079 struct napi_struct *napi,
2080 int (*poll)(struct napi_struct *, int),
2081 int weight)
2082{
2083 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2084 netif_napi_add(dev, napi, poll, weight);
2085}
2086
d8156534 2087/**
5e82b4b2
BH
2088 * netif_napi_del - remove a NAPI context
2089 * @napi: NAPI context
d8156534 2090 *
5e82b4b2 2091 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2092 */
d565b0a1
HX
2093void netif_napi_del(struct napi_struct *napi);
2094
2095struct napi_gro_cb {
78a478d0 2096 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2097 void *frag0;
78a478d0 2098
7489594c
HX
2099 /* Length of frag0. */
2100 unsigned int frag0_len;
2101
86911732 2102 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2103 int data_offset;
86911732 2104
d565b0a1 2105 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2106 u16 flush;
2107
2108 /* Save the IP ID here and check when we get to the transport layer */
2109 u16 flush_id;
d565b0a1
HX
2110
2111 /* Number of segments aggregated. */
2e71a6f8
ED
2112 u16 count;
2113
15e2396d
TH
2114 /* Start offset for remote checksum offload */
2115 u16 gro_remcsum_start;
2116
2e71a6f8
ED
2117 /* jiffies when first packet was created/queued */
2118 unsigned long age;
86347245 2119
afe93325 2120 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2121 u16 proto;
2122
baa32ff4
TH
2123 /* This is non-zero if the packet may be of the same flow. */
2124 u8 same_flow:1;
2125
fac8e0f5
JG
2126 /* Used in tunnel GRO receive */
2127 u8 encap_mark:1;
573e8fca
TH
2128
2129 /* GRO checksum is valid */
2130 u8 csum_valid:1;
2131
662880f4
TH
2132 /* Number of checksums via CHECKSUM_UNNECESSARY */
2133 u8 csum_cnt:3;
c3c7c254 2134
baa32ff4
TH
2135 /* Free the skb? */
2136 u8 free:2;
2137#define NAPI_GRO_FREE 1
2138#define NAPI_GRO_FREE_STOLEN_HEAD 2
2139
efc98d08
TH
2140 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2141 u8 is_ipv6:1;
2142
a0ca153f
AD
2143 /* Used in GRE, set in fou/gue_gro_receive */
2144 u8 is_fou:1;
2145
1530545e
AD
2146 /* Used to determine if flush_id can be ignored */
2147 u8 is_atomic:1;
2148
fcd91dd4
SD
2149 /* Number of gro_receive callbacks this packet already went through */
2150 u8 recursion_counter:4;
2151
2152 /* 1 bit hole */
baa32ff4 2153
bf5a755f
JC
2154 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2155 __wsum csum;
2156
c3c7c254
ED
2157 /* used in skb_gro_receive() slow path */
2158 struct sk_buff *last;
d565b0a1
HX
2159};
2160
2161#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2162
fcd91dd4
SD
2163#define GRO_RECURSION_LIMIT 15
2164static inline int gro_recursion_inc_test(struct sk_buff *skb)
2165{
2166 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2167}
2168
2169typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2170static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2171 struct sk_buff **head,
2172 struct sk_buff *skb)
2173{
2174 if (unlikely(gro_recursion_inc_test(skb))) {
2175 NAPI_GRO_CB(skb)->flush |= 1;
2176 return NULL;
2177 }
2178
2179 return cb(head, skb);
2180}
2181
2182typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2183 struct sk_buff *);
2184static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2185 struct sock *sk,
2186 struct sk_buff **head,
2187 struct sk_buff *skb)
2188{
2189 if (unlikely(gro_recursion_inc_test(skb))) {
2190 NAPI_GRO_CB(skb)->flush |= 1;
2191 return NULL;
2192 }
2193
2194 return cb(sk, head, skb);
2195}
2196
1da177e4 2197struct packet_type {
f2ccd8fa
DM
2198 __be16 type; /* This is really htons(ether_type). */
2199 struct net_device *dev; /* NULL is wildcarded here */
2200 int (*func) (struct sk_buff *,
2201 struct net_device *,
2202 struct packet_type *,
2203 struct net_device *);
c0de08d0
EL
2204 bool (*id_match)(struct packet_type *ptype,
2205 struct sock *sk);
1da177e4
LT
2206 void *af_packet_priv;
2207 struct list_head list;
2208};
2209
f191a1d1 2210struct offload_callbacks {
576a30eb 2211 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2212 netdev_features_t features);
d565b0a1 2213 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2214 struct sk_buff *skb);
299603e8 2215 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2216};
2217
2218struct packet_offload {
2219 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2220 u16 priority;
f191a1d1
VY
2221 struct offload_callbacks callbacks;
2222 struct list_head list;
1da177e4
LT
2223};
2224
5e82b4b2 2225/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2226struct pcpu_sw_netstats {
2227 u64 rx_packets;
2228 u64 rx_bytes;
2229 u64 tx_packets;
2230 u64 tx_bytes;
2231 struct u64_stats_sync syncp;
2232};
2233
aabc92bb
PNA
2234#define __netdev_alloc_pcpu_stats(type, gfp) \
2235({ \
2236 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2237 if (pcpu_stats) { \
2238 int __cpu; \
2239 for_each_possible_cpu(__cpu) { \
2240 typeof(type) *stat; \
2241 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2242 u64_stats_init(&stat->syncp); \
2243 } \
2244 } \
2245 pcpu_stats; \
1c213bd2
WC
2246})
2247
aabc92bb 2248#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2249 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2250
764f5e54
JP
2251enum netdev_lag_tx_type {
2252 NETDEV_LAG_TX_TYPE_UNKNOWN,
2253 NETDEV_LAG_TX_TYPE_RANDOM,
2254 NETDEV_LAG_TX_TYPE_BROADCAST,
2255 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2256 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2257 NETDEV_LAG_TX_TYPE_HASH,
2258};
2259
2260struct netdev_lag_upper_info {
2261 enum netdev_lag_tx_type tx_type;
2262};
2263
fb1b2e3c
JP
2264struct netdev_lag_lower_state_info {
2265 u8 link_up : 1,
2266 tx_enabled : 1;
2267};
2268
1da177e4
LT
2269#include <linux/notifier.h>
2270
dcfe1421
AW
2271/* netdevice notifier chain. Please remember to update the rtnetlink
2272 * notification exclusion list in rtnetlink_event() when adding new
2273 * types.
2274 */
2275#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2276#define NETDEV_DOWN 0x0002
2277#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2278 detected a hardware crash and restarted
2279 - we can use this eg to kick tcp sessions
2280 once done */
2281#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2282#define NETDEV_REGISTER 0x0005
2283#define NETDEV_UNREGISTER 0x0006
1d486bfb 2284#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2285#define NETDEV_CHANGEADDR 0x0008
2286#define NETDEV_GOING_DOWN 0x0009
2287#define NETDEV_CHANGENAME 0x000A
2288#define NETDEV_FEAT_CHANGE 0x000B
2289#define NETDEV_BONDING_FAILOVER 0x000C
2290#define NETDEV_PRE_UP 0x000D
2291#define NETDEV_PRE_TYPE_CHANGE 0x000E
2292#define NETDEV_POST_TYPE_CHANGE 0x000F
2293#define NETDEV_POST_INIT 0x0010
0115e8e3 2294#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2295#define NETDEV_RELEASE 0x0012
2296#define NETDEV_NOTIFY_PEERS 0x0013
2297#define NETDEV_JOIN 0x0014
42e52bf9 2298#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2299#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2300#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2301#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2302#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2303#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2304#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2305#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
296d8ee3 2306#define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
08294a26 2307#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2308
f629d208
JP
2309int register_netdevice_notifier(struct notifier_block *nb);
2310int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2311
2312struct netdev_notifier_info {
51d0c047
DA
2313 struct net_device *dev;
2314 struct netlink_ext_ack *extack;
351638e7
JP
2315};
2316
be9efd36
JP
2317struct netdev_notifier_change_info {
2318 struct netdev_notifier_info info; /* must be first */
2319 unsigned int flags_changed;
2320};
2321
0e4ead9d
JP
2322struct netdev_notifier_changeupper_info {
2323 struct netdev_notifier_info info; /* must be first */
2324 struct net_device *upper_dev; /* new upper dev */
2325 bool master; /* is upper dev master */
5e82b4b2 2326 bool linking; /* is the notification for link or unlink */
29bf24af 2327 void *upper_info; /* upper dev info */
0e4ead9d
JP
2328};
2329
04d48266
JP
2330struct netdev_notifier_changelowerstate_info {
2331 struct netdev_notifier_info info; /* must be first */
2332 void *lower_state_info; /* is lower dev state */
2333};
2334
75538c2b
CW
2335static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2336 struct net_device *dev)
2337{
2338 info->dev = dev;
51d0c047 2339 info->extack = NULL;
75538c2b
CW
2340}
2341
351638e7
JP
2342static inline struct net_device *
2343netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2344{
2345 return info->dev;
2346}
2347
51d0c047
DA
2348static inline struct netlink_ext_ack *
2349netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2350{
2351 return info->extack;
2352}
2353
f629d208 2354int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2355
2356
1da177e4
LT
2357extern rwlock_t dev_base_lock; /* Device list lock */
2358
881d966b
EB
2359#define for_each_netdev(net, d) \
2360 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2361#define for_each_netdev_reverse(net, d) \
2362 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2363#define for_each_netdev_rcu(net, d) \
2364 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2365#define for_each_netdev_safe(net, d, n) \
2366 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2367#define for_each_netdev_continue(net, d) \
2368 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2369#define for_each_netdev_continue_rcu(net, d) \
2370 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2371#define for_each_netdev_in_bond_rcu(bond, slave) \
2372 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2373 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2374#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2375
a050c33f
DL
2376static inline struct net_device *next_net_device(struct net_device *dev)
2377{
2378 struct list_head *lh;
2379 struct net *net;
2380
c346dca1 2381 net = dev_net(dev);
a050c33f
DL
2382 lh = dev->dev_list.next;
2383 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2384}
2385
ce81b76a
ED
2386static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2387{
2388 struct list_head *lh;
2389 struct net *net;
2390
2391 net = dev_net(dev);
ccf43438 2392 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2393 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2394}
2395
a050c33f
DL
2396static inline struct net_device *first_net_device(struct net *net)
2397{
2398 return list_empty(&net->dev_base_head) ? NULL :
2399 net_device_entry(net->dev_base_head.next);
2400}
7562f876 2401
ccf43438
ED
2402static inline struct net_device *first_net_device_rcu(struct net *net)
2403{
2404 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2405
2406 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2407}
2408
f629d208
JP
2409int netdev_boot_setup_check(struct net_device *dev);
2410unsigned long netdev_boot_base(const char *prefix, int unit);
2411struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2412 const char *hwaddr);
2413struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2414struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2415void dev_add_pack(struct packet_type *pt);
2416void dev_remove_pack(struct packet_type *pt);
2417void __dev_remove_pack(struct packet_type *pt);
2418void dev_add_offload(struct packet_offload *po);
2419void dev_remove_offload(struct packet_offload *po);
f629d208 2420
a54acb3a 2421int dev_get_iflink(const struct net_device *dev);
fc4099f1 2422int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2423struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2424 unsigned short mask);
f629d208
JP
2425struct net_device *dev_get_by_name(struct net *net, const char *name);
2426struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2427struct net_device *__dev_get_by_name(struct net *net, const char *name);
2428int dev_alloc_name(struct net_device *dev, const char *name);
2429int dev_open(struct net_device *dev);
7051b88a 2430void dev_close(struct net_device *dev);
2431void dev_close_many(struct list_head *head, bool unlink);
f629d208 2432void dev_disable_lro(struct net_device *dev);
0c4b51f0 2433int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2434int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2435int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2436int register_netdevice(struct net_device *dev);
2437void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2438void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2439static inline void unregister_netdevice(struct net_device *dev)
2440{
2441 unregister_netdevice_queue(dev, NULL);
2442}
2443
f629d208
JP
2444int netdev_refcnt_read(const struct net_device *dev);
2445void free_netdev(struct net_device *dev);
74d332c1 2446void netdev_freemem(struct net_device *dev);
f629d208
JP
2447void synchronize_net(void);
2448int init_dummy_netdev(struct net_device *dev);
937f1ba5 2449
f60e5990 2450DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2451#define XMIT_RECURSION_LIMIT 10
2452
f60e5990 2453static inline int dev_recursion_level(void)
2454{
2455 return this_cpu_read(xmit_recursion);
2456}
2457
f629d208
JP
2458struct net_device *dev_get_by_index(struct net *net, int ifindex);
2459struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2460struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2461struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2462int netdev_get_name(struct net *net, char *name, int ifindex);
2463int dev_restart(struct net_device *dev);
f629d208 2464int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2465
2466static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2467{
2468 return NAPI_GRO_CB(skb)->data_offset;
2469}
2470
2471static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2472{
2473 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2474}
2475
2476static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2477{
2478 NAPI_GRO_CB(skb)->data_offset += len;
2479}
2480
a5b1cf28
HX
2481static inline void *skb_gro_header_fast(struct sk_buff *skb,
2482 unsigned int offset)
86911732 2483{
a5b1cf28
HX
2484 return NAPI_GRO_CB(skb)->frag0 + offset;
2485}
78a478d0 2486
a5b1cf28
HX
2487static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2488{
2489 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2490}
78a478d0 2491
57ea52a8
HX
2492static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2493{
2494 NAPI_GRO_CB(skb)->frag0 = NULL;
2495 NAPI_GRO_CB(skb)->frag0_len = 0;
2496}
2497
a5b1cf28
HX
2498static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2499 unsigned int offset)
2500{
17dd759c
HX
2501 if (!pskb_may_pull(skb, hlen))
2502 return NULL;
2503
57ea52a8 2504 skb_gro_frag0_invalidate(skb);
17dd759c 2505 return skb->data + offset;
86911732 2506}
1da177e4 2507
36e7b1b8
HX
2508static inline void *skb_gro_network_header(struct sk_buff *skb)
2509{
78d3fd0b
HX
2510 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2511 skb_network_offset(skb);
36e7b1b8
HX
2512}
2513
bf5a755f
JC
2514static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2515 const void *start, unsigned int len)
2516{
573e8fca 2517 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2518 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2519 csum_partial(start, len, 0));
2520}
2521
573e8fca
TH
2522/* GRO checksum functions. These are logical equivalents of the normal
2523 * checksum functions (in skbuff.h) except that they operate on the GRO
2524 * offsets and fields in sk_buff.
2525 */
2526
2527__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2528
15e2396d
TH
2529static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2530{
b7fe10e5 2531 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2532}
2533
573e8fca
TH
2534static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2535 bool zero_okay,
2536 __sum16 check)
2537{
6edec0e6
TH
2538 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2539 skb_checksum_start_offset(skb) <
2540 skb_gro_offset(skb)) &&
15e2396d 2541 !skb_at_gro_remcsum_start(skb) &&
662880f4 2542 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2543 (!zero_okay || check));
2544}
2545
2546static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2547 __wsum psum)
2548{
2549 if (NAPI_GRO_CB(skb)->csum_valid &&
2550 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2551 return 0;
2552
2553 NAPI_GRO_CB(skb)->csum = psum;
2554
2555 return __skb_gro_checksum_complete(skb);
2556}
2557
573e8fca
TH
2558static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2559{
662880f4
TH
2560 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2561 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2562 NAPI_GRO_CB(skb)->csum_cnt--;
2563 } else {
2564 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2565 * verified a new top level checksum or an encapsulated one
2566 * during GRO. This saves work if we fallback to normal path.
2567 */
2568 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2569 }
2570}
2571
2572#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2573 compute_pseudo) \
2574({ \
2575 __sum16 __ret = 0; \
2576 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2577 __ret = __skb_gro_checksum_validate_complete(skb, \
2578 compute_pseudo(skb, proto)); \
219f1d79 2579 if (!__ret) \
573e8fca
TH
2580 skb_gro_incr_csum_unnecessary(skb); \
2581 __ret; \
2582})
2583
2584#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2585 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2586
2587#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2588 compute_pseudo) \
2589 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2590
2591#define skb_gro_checksum_simple_validate(skb) \
2592 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2593
d96535a1
TH
2594static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2595{
2596 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2597 !NAPI_GRO_CB(skb)->csum_valid);
2598}
2599
2600static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2601 __sum16 check, __wsum pseudo)
2602{
2603 NAPI_GRO_CB(skb)->csum = ~pseudo;
2604 NAPI_GRO_CB(skb)->csum_valid = 1;
2605}
2606
2607#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2608do { \
2609 if (__skb_gro_checksum_convert_check(skb)) \
2610 __skb_gro_checksum_convert(skb, check, \
2611 compute_pseudo(skb, proto)); \
2612} while (0)
2613
26c4f7da
TH
2614struct gro_remcsum {
2615 int offset;
2616 __wsum delta;
2617};
2618
2619static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2620{
846cd667 2621 grc->offset = 0;
26c4f7da
TH
2622 grc->delta = 0;
2623}
2624
b7fe10e5
TH
2625static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2626 unsigned int off, size_t hdrlen,
2627 int start, int offset,
2628 struct gro_remcsum *grc,
2629 bool nopartial)
dcdc8994
TH
2630{
2631 __wsum delta;
b7fe10e5 2632 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2633
2634 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2635
15e2396d 2636 if (!nopartial) {
b7fe10e5
TH
2637 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2638 return ptr;
2639 }
2640
2641 ptr = skb_gro_header_fast(skb, off);
2642 if (skb_gro_header_hard(skb, off + plen)) {
2643 ptr = skb_gro_header_slow(skb, off + plen, off);
2644 if (!ptr)
2645 return NULL;
15e2396d
TH
2646 }
2647
b7fe10e5
TH
2648 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2649 start, offset);
dcdc8994
TH
2650
2651 /* Adjust skb->csum since we changed the packet */
dcdc8994 2652 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2653
b7fe10e5 2654 grc->offset = off + hdrlen + offset;
26c4f7da 2655 grc->delta = delta;
b7fe10e5
TH
2656
2657 return ptr;
dcdc8994
TH
2658}
2659
26c4f7da
TH
2660static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2661 struct gro_remcsum *grc)
2662{
b7fe10e5
TH
2663 void *ptr;
2664 size_t plen = grc->offset + sizeof(u16);
2665
26c4f7da
TH
2666 if (!grc->delta)
2667 return;
2668
b7fe10e5
TH
2669 ptr = skb_gro_header_fast(skb, grc->offset);
2670 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2671 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2672 if (!ptr)
2673 return;
2674 }
2675
2676 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2677}
dcdc8994 2678
25393d3f
SK
2679#ifdef CONFIG_XFRM_OFFLOAD
2680static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2681{
2682 if (PTR_ERR(pp) != -EINPROGRESS)
2683 NAPI_GRO_CB(skb)->flush |= flush;
2684}
2685#else
5f114163
SK
2686static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2687{
2688 NAPI_GRO_CB(skb)->flush |= flush;
2689}
25393d3f 2690#endif
5f114163 2691
0c4e8581
SH
2692static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2693 unsigned short type,
3b04ddde 2694 const void *daddr, const void *saddr,
95c96174 2695 unsigned int len)
0c4e8581 2696{
f1ecfd5d 2697 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2698 return 0;
3b04ddde
SH
2699
2700 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2701}
2702
b95cce35
SH
2703static inline int dev_parse_header(const struct sk_buff *skb,
2704 unsigned char *haddr)
2705{
2706 const struct net_device *dev = skb->dev;
2707
1b83336b 2708 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2709 return 0;
3b04ddde 2710 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2711}
2712
2793a23a
WB
2713/* ll_header must have at least hard_header_len allocated */
2714static inline bool dev_validate_header(const struct net_device *dev,
2715 char *ll_header, int len)
2716{
2717 if (likely(len >= dev->hard_header_len))
2718 return true;
217e6fa2
WB
2719 if (len < dev->min_header_len)
2720 return false;
2793a23a
WB
2721
2722 if (capable(CAP_SYS_RAWIO)) {
2723 memset(ll_header + len, 0, dev->hard_header_len - len);
2724 return true;
2725 }
2726
2727 if (dev->header_ops && dev->header_ops->validate)
2728 return dev->header_ops->validate(ll_header, len);
2729
2730 return false;
2731}
2732
1da177e4 2733typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2734int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2735static inline int unregister_gifconf(unsigned int family)
2736{
2737 return register_gifconf(family, NULL);
2738}
2739
99bbc707 2740#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2741#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2742struct sd_flow_limit {
2743 u64 count;
2744 unsigned int num_buckets;
2745 unsigned int history_head;
2746 u16 history[FLOW_LIMIT_HISTORY];
2747 u8 buckets[];
2748};
2749
2750extern int netdev_flow_limit_table_len;
2751#endif /* CONFIG_NET_FLOW_LIMIT */
2752
1da177e4 2753/*
5e82b4b2 2754 * Incoming packets are placed on per-CPU queues
1da177e4 2755 */
d94d9fee 2756struct softnet_data {
1da177e4 2757 struct list_head poll_list;
6e7676c1 2758 struct sk_buff_head process_queue;
1da177e4 2759
dee42870 2760 /* stats */
cd7b5396
DM
2761 unsigned int processed;
2762 unsigned int time_squeeze;
cd7b5396 2763 unsigned int received_rps;
fd793d89 2764#ifdef CONFIG_RPS
88751275 2765 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2766#endif
2767#ifdef CONFIG_NET_FLOW_LIMIT
2768 struct sd_flow_limit __rcu *flow_limit;
2769#endif
2770 struct Qdisc *output_queue;
2771 struct Qdisc **output_queue_tailp;
2772 struct sk_buff *completion_queue;
88751275 2773
4cdb1e2e 2774#ifdef CONFIG_RPS
501e7ef5
ED
2775 /* input_queue_head should be written by cpu owning this struct,
2776 * and only read by other cpus. Worth using a cache line.
2777 */
2778 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2779
2780 /* Elements below can be accessed between CPUs for RPS/RFS */
966a9671 2781 call_single_data_t csd ____cacheline_aligned_in_smp;
88751275
ED
2782 struct softnet_data *rps_ipi_next;
2783 unsigned int cpu;
76cc8b13 2784 unsigned int input_queue_tail;
1e94d72f 2785#endif
95c96174 2786 unsigned int dropped;
0a9627f2 2787 struct sk_buff_head input_pkt_queue;
bea3348e 2788 struct napi_struct backlog;
99bbc707 2789
1da177e4
LT
2790};
2791
76cc8b13 2792static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2793{
2794#ifdef CONFIG_RPS
76cc8b13
TH
2795 sd->input_queue_head++;
2796#endif
2797}
2798
2799static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2800 unsigned int *qtail)
2801{
2802#ifdef CONFIG_RPS
2803 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2804#endif
2805}
2806
0a9627f2 2807DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2808
f629d208 2809void __netif_schedule(struct Qdisc *q);
46e5da40 2810void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2811
fd2ea0a7
DM
2812static inline void netif_tx_schedule_all(struct net_device *dev)
2813{
2814 unsigned int i;
2815
2816 for (i = 0; i < dev->num_tx_queues; i++)
2817 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2818}
2819
f9a7cbbf 2820static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2821{
73466498 2822 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2823}
2824
bea3348e
SH
2825/**
2826 * netif_start_queue - allow transmit
2827 * @dev: network device
2828 *
2829 * Allow upper layers to call the device hard_start_xmit routine.
2830 */
1da177e4
LT
2831static inline void netif_start_queue(struct net_device *dev)
2832{
e8a0464c 2833 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2834}
2835
fd2ea0a7
DM
2836static inline void netif_tx_start_all_queues(struct net_device *dev)
2837{
2838 unsigned int i;
2839
2840 for (i = 0; i < dev->num_tx_queues; i++) {
2841 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2842 netif_tx_start_queue(txq);
2843 }
2844}
2845
46e5da40 2846void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2847
d29f749e
DJ
2848/**
2849 * netif_wake_queue - restart transmit
2850 * @dev: network device
2851 *
2852 * Allow upper layers to call the device hard_start_xmit routine.
2853 * Used for flow control when transmit resources are available.
2854 */
79d16385
DM
2855static inline void netif_wake_queue(struct net_device *dev)
2856{
e8a0464c 2857 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2858}
2859
fd2ea0a7
DM
2860static inline void netif_tx_wake_all_queues(struct net_device *dev)
2861{
2862 unsigned int i;
2863
2864 for (i = 0; i < dev->num_tx_queues; i++) {
2865 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2866 netif_tx_wake_queue(txq);
2867 }
2868}
2869
f9a7cbbf 2870static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2871{
73466498 2872 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2873}
2874
bea3348e
SH
2875/**
2876 * netif_stop_queue - stop transmitted packets
2877 * @dev: network device
2878 *
2879 * Stop upper layers calling the device hard_start_xmit routine.
2880 * Used for flow control when transmit resources are unavailable.
2881 */
1da177e4
LT
2882static inline void netif_stop_queue(struct net_device *dev)
2883{
e8a0464c 2884 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2885}
2886
a2029240 2887void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2888
4d29515f 2889static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2890{
73466498 2891 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2892}
2893
bea3348e
SH
2894/**
2895 * netif_queue_stopped - test if transmit queue is flowblocked
2896 * @dev: network device
2897 *
2898 * Test if transmit queue on device is currently unable to send.
2899 */
4d29515f 2900static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2901{
e8a0464c 2902 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2903}
2904
4d29515f 2905static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2906{
73466498
TH
2907 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2908}
2909
8e2f1a63
DB
2910static inline bool
2911netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2912{
2913 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2914}
2915
8e2f1a63
DB
2916static inline bool
2917netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2918{
2919 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2920}
2921
53511453
ED
2922/**
2923 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2924 * @dev_queue: pointer to transmit queue
2925 *
2926 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2927 * to give appropriate hint to the CPU.
53511453
ED
2928 */
2929static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2930{
2931#ifdef CONFIG_BQL
2932 prefetchw(&dev_queue->dql.num_queued);
2933#endif
2934}
2935
2936/**
2937 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2938 * @dev_queue: pointer to transmit queue
2939 *
2940 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2941 * to give appropriate hint to the CPU.
53511453
ED
2942 */
2943static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2944{
2945#ifdef CONFIG_BQL
2946 prefetchw(&dev_queue->dql.limit);
2947#endif
2948}
2949
c5d67bd7
TH
2950static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2951 unsigned int bytes)
2952{
114cf580
TH
2953#ifdef CONFIG_BQL
2954 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2955
2956 if (likely(dql_avail(&dev_queue->dql) >= 0))
2957 return;
2958
2959 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2960
2961 /*
2962 * The XOFF flag must be set before checking the dql_avail below,
2963 * because in netdev_tx_completed_queue we update the dql_completed
2964 * before checking the XOFF flag.
2965 */
2966 smp_mb();
2967
2968 /* check again in case another CPU has just made room avail */
2969 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2970 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2971#endif
c5d67bd7
TH
2972}
2973
0042d0c8
FF
2974/**
2975 * netdev_sent_queue - report the number of bytes queued to hardware
2976 * @dev: network device
2977 * @bytes: number of bytes queued to the hardware device queue
2978 *
2979 * Report the number of bytes queued for sending/completion to the network
2980 * device hardware queue. @bytes should be a good approximation and should
2981 * exactly match netdev_completed_queue() @bytes
2982 */
c5d67bd7
TH
2983static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2984{
2985 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2986}
2987
2988static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2989 unsigned int pkts, unsigned int bytes)
c5d67bd7 2990{
114cf580 2991#ifdef CONFIG_BQL
b37c0fbe
AD
2992 if (unlikely(!bytes))
2993 return;
2994
2995 dql_completed(&dev_queue->dql, bytes);
2996
2997 /*
2998 * Without the memory barrier there is a small possiblity that
2999 * netdev_tx_sent_queue will miss the update and cause the queue to
3000 * be stopped forever
3001 */
3002 smp_mb();
3003
3004 if (dql_avail(&dev_queue->dql) < 0)
3005 return;
3006
3007 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3008 netif_schedule_queue(dev_queue);
114cf580 3009#endif
c5d67bd7
TH
3010}
3011
0042d0c8
FF
3012/**
3013 * netdev_completed_queue - report bytes and packets completed by device
3014 * @dev: network device
3015 * @pkts: actual number of packets sent over the medium
3016 * @bytes: actual number of bytes sent over the medium
3017 *
3018 * Report the number of bytes and packets transmitted by the network device
3019 * hardware queue over the physical medium, @bytes must exactly match the
3020 * @bytes amount passed to netdev_sent_queue()
3021 */
c5d67bd7 3022static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3023 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3024{
3025 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3026}
3027
3028static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3029{
114cf580 3030#ifdef CONFIG_BQL
5c490354 3031 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3032 dql_reset(&q->dql);
3033#endif
c5d67bd7
TH
3034}
3035
0042d0c8
FF
3036/**
3037 * netdev_reset_queue - reset the packets and bytes count of a network device
3038 * @dev_queue: network device
3039 *
3040 * Reset the bytes and packet count of a network device and clear the
3041 * software flow control OFF bit for this network device
3042 */
c5d67bd7
TH
3043static inline void netdev_reset_queue(struct net_device *dev_queue)
3044{
3045 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3046}
3047
b9507bda
DB
3048/**
3049 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3050 * @dev: network device
3051 * @queue_index: given tx queue index
3052 *
3053 * Returns 0 if given tx queue index >= number of device tx queues,
3054 * otherwise returns the originally passed tx queue index.
3055 */
3056static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3057{
3058 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3059 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3060 dev->name, queue_index,
3061 dev->real_num_tx_queues);
3062 return 0;
3063 }
3064
3065 return queue_index;
3066}
3067
bea3348e
SH
3068/**
3069 * netif_running - test if up
3070 * @dev: network device
3071 *
3072 * Test if the device has been brought up.
3073 */
4d29515f 3074static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3075{
3076 return test_bit(__LINK_STATE_START, &dev->state);
3077}
3078
f25f4e44 3079/*
5e82b4b2 3080 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3081 * stop, and a check if it's stopped. All other device management is
3082 * done at the overall netdevice level.
3083 * Also test the device if we're multiqueue.
3084 */
bea3348e
SH
3085
3086/**
3087 * netif_start_subqueue - allow sending packets on subqueue
3088 * @dev: network device
3089 * @queue_index: sub queue index
3090 *
3091 * Start individual transmit queue of a device with multiple transmit queues.
3092 */
f25f4e44
PWJ
3093static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3094{
fd2ea0a7 3095 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3096
3097 netif_tx_start_queue(txq);
f25f4e44
PWJ
3098}
3099
bea3348e
SH
3100/**
3101 * netif_stop_subqueue - stop sending packets on subqueue
3102 * @dev: network device
3103 * @queue_index: sub queue index
3104 *
3105 * Stop individual transmit queue of a device with multiple transmit queues.
3106 */
f25f4e44
PWJ
3107static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3108{
fd2ea0a7 3109 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3110 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3111}
3112
bea3348e
SH
3113/**
3114 * netif_subqueue_stopped - test status of subqueue
3115 * @dev: network device
3116 * @queue_index: sub queue index
3117 *
3118 * Check individual transmit queue of a device with multiple transmit queues.
3119 */
4d29515f
DM
3120static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3121 u16 queue_index)
f25f4e44 3122{
fd2ea0a7 3123 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3124
3125 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3126}
3127
4d29515f
DM
3128static inline bool netif_subqueue_stopped(const struct net_device *dev,
3129 struct sk_buff *skb)
668f895a
PE
3130{
3131 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3132}
bea3348e 3133
738b35cc
FF
3134/**
3135 * netif_wake_subqueue - allow sending packets on subqueue
3136 * @dev: network device
3137 * @queue_index: sub queue index
3138 *
3139 * Resume individual transmit queue of a device with multiple transmit queues.
3140 */
3141static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3142{
3143 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3144
3145 netif_tx_wake_queue(txq);
3146}
f25f4e44 3147
537c00de 3148#ifdef CONFIG_XPS
53af53ae 3149int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3150 u16 index);
537c00de
AD
3151#else
3152static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3153 const struct cpumask *mask,
537c00de
AD
3154 u16 index)
3155{
3156 return 0;
3157}
3158#endif
3159
5605c762
JP
3160u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3161 unsigned int num_tx_queues);
3162
a3d22a68
VZ
3163/*
3164 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3165 * as a distribution range limit for the returned value.
3166 */
3167static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3168 struct sk_buff *skb)
a3d22a68
VZ
3169{
3170 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3171}
3172
bea3348e
SH
3173/**
3174 * netif_is_multiqueue - test if device has multiple transmit queues
3175 * @dev: network device
3176 *
3177 * Check if device has multiple transmit queues
bea3348e 3178 */
4d29515f 3179static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3180{
a02cec21 3181 return dev->num_tx_queues > 1;
f25f4e44 3182}
1da177e4 3183
f629d208 3184int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3185
a953be53 3186#ifdef CONFIG_SYSFS
f629d208 3187int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3188#else
3189static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3190 unsigned int rxq)
3191{
3192 return 0;
3193}
3194#endif
3195
a953be53
MD
3196#ifdef CONFIG_SYSFS
3197static inline unsigned int get_netdev_rx_queue_index(
3198 struct netdev_rx_queue *queue)
3199{
3200 struct net_device *dev = queue->dev;
3201 int index = queue - dev->_rx;
3202
3203 BUG_ON(index >= dev->num_rx_queues);
3204 return index;
3205}
3206#endif
3207
16917b87 3208#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3209int netif_get_num_default_rss_queues(void);
16917b87 3210
e6247027
ED
3211enum skb_free_reason {
3212 SKB_REASON_CONSUMED,
3213 SKB_REASON_DROPPED,
3214};
3215
3216void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3217void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3218
e6247027
ED
3219/*
3220 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3221 * interrupt context or with hardware interrupts being disabled.
3222 * (in_irq() || irqs_disabled())
3223 *
3224 * We provide four helpers that can be used in following contexts :
3225 *
3226 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3227 * replacing kfree_skb(skb)
3228 *
3229 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3230 * Typically used in place of consume_skb(skb) in TX completion path
3231 *
3232 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3233 * replacing kfree_skb(skb)
3234 *
3235 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3236 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3237 */
e6247027
ED
3238static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3239{
3240 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3241}
3242
3243static inline void dev_consume_skb_irq(struct sk_buff *skb)
3244{
3245 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3246}
3247
3248static inline void dev_kfree_skb_any(struct sk_buff *skb)
3249{
3250 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3251}
3252
3253static inline void dev_consume_skb_any(struct sk_buff *skb)
3254{
3255 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3256}
1da177e4 3257
7c497478
JW
3258void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3259int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
f629d208
JP
3260int netif_rx(struct sk_buff *skb);
3261int netif_rx_ni(struct sk_buff *skb);
04eb4489 3262int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3263gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3264void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3265struct sk_buff *napi_get_frags(struct napi_struct *napi);
3266gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3267struct packet_offload *gro_find_receive_by_type(__be16 type);
3268struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3269
3270static inline void napi_free_frags(struct napi_struct *napi)
3271{
3272 kfree_skb(napi->skb);
3273 napi->skb = NULL;
3274}
3275
24b27fc4 3276bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3277int netdev_rx_handler_register(struct net_device *dev,
3278 rx_handler_func_t *rx_handler,
3279 void *rx_handler_data);
3280void netdev_rx_handler_unregister(struct net_device *dev);
3281
3282bool dev_valid_name(const char *name);
3283int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3284int dev_ethtool(struct net *net, struct ifreq *);
3285unsigned int dev_get_flags(const struct net_device *);
3286int __dev_change_flags(struct net_device *, unsigned int flags);
3287int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3288void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3289 unsigned int gchanges);
f629d208
JP
3290int dev_change_name(struct net_device *, const char *);
3291int dev_set_alias(struct net_device *, const char *, size_t);
6c557001 3292int dev_get_alias(const struct net_device *, char *, size_t);
f629d208 3293int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3294int __dev_set_mtu(struct net_device *, int);
f629d208
JP
3295int dev_set_mtu(struct net_device *, int);
3296void dev_set_group(struct net_device *, int);
3297int dev_set_mac_address(struct net_device *, struct sockaddr *);
3298int dev_change_carrier(struct net_device *, bool new_carrier);
3299int dev_get_phys_port_id(struct net_device *dev,
02637fce 3300 struct netdev_phys_item_id *ppid);
db24a904
DA
3301int dev_get_phys_port_name(struct net_device *dev,
3302 char *name, size_t len);
d746d707 3303int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3304struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3305struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3306 struct netdev_queue *txq, int *ret);
d67b9cd2
DB
3307
3308typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3309int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3310 int fd, u32 flags);
ce158e58 3311u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
d67b9cd2 3312
a0265d28 3313int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3314int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3315bool is_skb_forwardable(const struct net_device *dev,
3316 const struct sk_buff *skb);
1da177e4 3317
4e3264d2
MKL
3318static __always_inline int ____dev_forward_skb(struct net_device *dev,
3319 struct sk_buff *skb)
3320{
3321 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3322 unlikely(!is_skb_forwardable(dev, skb))) {
3323 atomic_long_inc(&dev->rx_dropped);
3324 kfree_skb(skb);
3325 return NET_RX_DROP;
3326 }
3327
3328 skb_scrub_packet(skb, true);
3329 skb->priority = 0;
3330 return 0;
3331}
3332
74b20582
DA
3333void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3334
20380731 3335extern int netdev_budget;
7acf8a1e 3336extern unsigned int netdev_budget_usecs;
1da177e4
LT
3337
3338/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3339void netdev_run_todo(void);
1da177e4 3340
bea3348e
SH
3341/**
3342 * dev_put - release reference to device
3343 * @dev: network device
3344 *
9ef4429b 3345 * Release reference to device to allow it to be freed.
bea3348e 3346 */
1da177e4
LT
3347static inline void dev_put(struct net_device *dev)
3348{
933393f5 3349 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3350}
3351
bea3348e
SH
3352/**
3353 * dev_hold - get reference to device
3354 * @dev: network device
3355 *
9ef4429b 3356 * Hold reference to device to keep it from being freed.
bea3348e 3357 */
15333061
SH
3358static inline void dev_hold(struct net_device *dev)
3359{
933393f5 3360 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3361}
1da177e4
LT
3362
3363/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3364 * and _off may be called from IRQ context, but it is caller
3365 * who is responsible for serialization of these calls.
b00055aa
SR
3366 *
3367 * The name carrier is inappropriate, these functions should really be
3368 * called netif_lowerlayer_*() because they represent the state of any
3369 * kind of lower layer not just hardware media.
1da177e4
LT
3370 */
3371
f629d208
JP
3372void linkwatch_init_dev(struct net_device *dev);
3373void linkwatch_fire_event(struct net_device *dev);
3374void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3375
bea3348e
SH
3376/**
3377 * netif_carrier_ok - test if carrier present
3378 * @dev: network device
3379 *
3380 * Check if carrier is present on device
3381 */
4d29515f 3382static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3383{
3384 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3385}
3386
f629d208 3387unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3388
f629d208 3389void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3390
f629d208 3391void netif_carrier_on(struct net_device *dev);
1da177e4 3392
f629d208 3393void netif_carrier_off(struct net_device *dev);
1da177e4 3394
bea3348e
SH
3395/**
3396 * netif_dormant_on - mark device as dormant.
3397 * @dev: network device
3398 *
3399 * Mark device as dormant (as per RFC2863).
3400 *
3401 * The dormant state indicates that the relevant interface is not
3402 * actually in a condition to pass packets (i.e., it is not 'up') but is
3403 * in a "pending" state, waiting for some external event. For "on-
3404 * demand" interfaces, this new state identifies the situation where the
3405 * interface is waiting for events to place it in the up state.
bea3348e 3406 */
b00055aa
SR
3407static inline void netif_dormant_on(struct net_device *dev)
3408{
3409 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3410 linkwatch_fire_event(dev);
3411}
3412
bea3348e
SH
3413/**
3414 * netif_dormant_off - set device as not dormant.
3415 * @dev: network device
3416 *
3417 * Device is not in dormant state.
3418 */
b00055aa
SR
3419static inline void netif_dormant_off(struct net_device *dev)
3420{
3421 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3422 linkwatch_fire_event(dev);
3423}
3424
bea3348e 3425/**
8ecbc40a 3426 * netif_dormant - test if device is dormant
bea3348e
SH
3427 * @dev: network device
3428 *
8ecbc40a 3429 * Check if device is dormant.
bea3348e 3430 */
4d29515f 3431static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3432{
3433 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3434}
3435
3436
bea3348e
SH
3437/**
3438 * netif_oper_up - test if device is operational
3439 * @dev: network device
3440 *
3441 * Check if carrier is operational
3442 */
4d29515f 3443static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3444{
b00055aa
SR
3445 return (dev->operstate == IF_OPER_UP ||
3446 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3447}
3448
bea3348e
SH
3449/**
3450 * netif_device_present - is device available or removed
3451 * @dev: network device
3452 *
3453 * Check if device has not been removed from system.
3454 */
4d29515f 3455static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3456{
3457 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3458}
3459
f629d208 3460void netif_device_detach(struct net_device *dev);
1da177e4 3461
f629d208 3462void netif_device_attach(struct net_device *dev);
1da177e4
LT
3463
3464/*
3465 * Network interface message level settings
3466 */
1da177e4
LT
3467
3468enum {
3469 NETIF_MSG_DRV = 0x0001,
3470 NETIF_MSG_PROBE = 0x0002,
3471 NETIF_MSG_LINK = 0x0004,
3472 NETIF_MSG_TIMER = 0x0008,
3473 NETIF_MSG_IFDOWN = 0x0010,
3474 NETIF_MSG_IFUP = 0x0020,
3475 NETIF_MSG_RX_ERR = 0x0040,
3476 NETIF_MSG_TX_ERR = 0x0080,
3477 NETIF_MSG_TX_QUEUED = 0x0100,
3478 NETIF_MSG_INTR = 0x0200,
3479 NETIF_MSG_TX_DONE = 0x0400,
3480 NETIF_MSG_RX_STATUS = 0x0800,
3481 NETIF_MSG_PKTDATA = 0x1000,
3482 NETIF_MSG_HW = 0x2000,
3483 NETIF_MSG_WOL = 0x4000,
3484};
3485
3486#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3487#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3488#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3489#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3490#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3491#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3492#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3493#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3494#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3495#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3496#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3497#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3498#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3499#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3500#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3501
3502static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3503{
3504 /* use default */
3505 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3506 return default_msg_enable_bits;
3507 if (debug_value == 0) /* no output */
3508 return 0;
3509 /* set low N bits */
3510 return (1 << debug_value) - 1;
3511}
3512
c773e847 3513static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3514{
c773e847
DM
3515 spin_lock(&txq->_xmit_lock);
3516 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3517}
3518
5a717f4f
MT
3519static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3520{
3521 __acquire(&txq->_xmit_lock);
3522 return true;
3523}
3524
3525static inline void __netif_tx_release(struct netdev_queue *txq)
3526{
3527 __release(&txq->_xmit_lock);
3528}
3529
fd2ea0a7
DM
3530static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3531{
3532 spin_lock_bh(&txq->_xmit_lock);
3533 txq->xmit_lock_owner = smp_processor_id();
3534}
3535
4d29515f 3536static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3537{
4d29515f 3538 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3539 if (likely(ok))
3540 txq->xmit_lock_owner = smp_processor_id();
3541 return ok;
3542}
3543
3544static inline void __netif_tx_unlock(struct netdev_queue *txq)
3545{
3546 txq->xmit_lock_owner = -1;
3547 spin_unlock(&txq->_xmit_lock);
3548}
3549
3550static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3551{
3552 txq->xmit_lock_owner = -1;
3553 spin_unlock_bh(&txq->_xmit_lock);
3554}
3555
08baf561
ED
3556static inline void txq_trans_update(struct netdev_queue *txq)
3557{
3558 if (txq->xmit_lock_owner != -1)
3559 txq->trans_start = jiffies;
3560}
3561
ba162f8e
FW
3562/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3563static inline void netif_trans_update(struct net_device *dev)
3564{
9b36627a
FW
3565 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3566
3567 if (txq->trans_start != jiffies)
3568 txq->trans_start = jiffies;
ba162f8e
FW
3569}
3570
d29f749e
DJ
3571/**
3572 * netif_tx_lock - grab network device transmit lock
3573 * @dev: network device
d29f749e
DJ
3574 *
3575 * Get network device transmit lock
3576 */
22dd7495
JHS
3577static inline void netif_tx_lock(struct net_device *dev)
3578{
e8a0464c 3579 unsigned int i;
c3f26a26 3580 int cpu;
c773e847 3581
c3f26a26
DM
3582 spin_lock(&dev->tx_global_lock);
3583 cpu = smp_processor_id();
e8a0464c
DM
3584 for (i = 0; i < dev->num_tx_queues; i++) {
3585 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3586
3587 /* We are the only thread of execution doing a
3588 * freeze, but we have to grab the _xmit_lock in
3589 * order to synchronize with threads which are in
3590 * the ->hard_start_xmit() handler and already
3591 * checked the frozen bit.
3592 */
e8a0464c 3593 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3594 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3595 __netif_tx_unlock(txq);
e8a0464c 3596 }
932ff279
HX
3597}
3598
3599static inline void netif_tx_lock_bh(struct net_device *dev)
3600{
e8a0464c
DM
3601 local_bh_disable();
3602 netif_tx_lock(dev);
932ff279
HX
3603}
3604
932ff279
HX
3605static inline void netif_tx_unlock(struct net_device *dev)
3606{
e8a0464c
DM
3607 unsigned int i;
3608
3609 for (i = 0; i < dev->num_tx_queues; i++) {
3610 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3611
c3f26a26
DM
3612 /* No need to grab the _xmit_lock here. If the
3613 * queue is not stopped for another reason, we
3614 * force a schedule.
3615 */
3616 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3617 netif_schedule_queue(txq);
c3f26a26
DM
3618 }
3619 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3620}
3621
3622static inline void netif_tx_unlock_bh(struct net_device *dev)
3623{
e8a0464c
DM
3624 netif_tx_unlock(dev);
3625 local_bh_enable();
932ff279
HX
3626}
3627
c773e847 3628#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3629 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3630 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3631 } else { \
3632 __netif_tx_acquire(txq); \
22dd7495
JHS
3633 } \
3634}
3635
5efeac44
EB
3636#define HARD_TX_TRYLOCK(dev, txq) \
3637 (((dev->features & NETIF_F_LLTX) == 0) ? \
3638 __netif_tx_trylock(txq) : \
5a717f4f 3639 __netif_tx_acquire(txq))
5efeac44 3640
c773e847 3641#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3642 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3643 __netif_tx_unlock(txq); \
5a717f4f
MT
3644 } else { \
3645 __netif_tx_release(txq); \
22dd7495
JHS
3646 } \
3647}
3648
1da177e4
LT
3649static inline void netif_tx_disable(struct net_device *dev)
3650{
fd2ea0a7 3651 unsigned int i;
c3f26a26 3652 int cpu;
fd2ea0a7 3653
c3f26a26
DM
3654 local_bh_disable();
3655 cpu = smp_processor_id();
fd2ea0a7
DM
3656 for (i = 0; i < dev->num_tx_queues; i++) {
3657 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3658
3659 __netif_tx_lock(txq, cpu);
fd2ea0a7 3660 netif_tx_stop_queue(txq);
c3f26a26 3661 __netif_tx_unlock(txq);
fd2ea0a7 3662 }
c3f26a26 3663 local_bh_enable();
1da177e4
LT
3664}
3665
e308a5d8
DM
3666static inline void netif_addr_lock(struct net_device *dev)
3667{
3668 spin_lock(&dev->addr_list_lock);
3669}
3670
2429f7ac
JP
3671static inline void netif_addr_lock_nested(struct net_device *dev)
3672{
25175ba5
VY
3673 int subclass = SINGLE_DEPTH_NESTING;
3674
3675 if (dev->netdev_ops->ndo_get_lock_subclass)
3676 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3677
3678 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3679}
3680
e308a5d8
DM
3681static inline void netif_addr_lock_bh(struct net_device *dev)
3682{
3683 spin_lock_bh(&dev->addr_list_lock);
3684}
3685
3686static inline void netif_addr_unlock(struct net_device *dev)
3687{
3688 spin_unlock(&dev->addr_list_lock);
3689}
3690
3691static inline void netif_addr_unlock_bh(struct net_device *dev)
3692{
3693 spin_unlock_bh(&dev->addr_list_lock);
3694}
3695
f001fde5 3696/*
31278e71 3697 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3698 * rcu_read_lock held.
3699 */
3700#define for_each_dev_addr(dev, ha) \
31278e71 3701 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3702
1da177e4
LT
3703/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3704
f629d208 3705void ether_setup(struct net_device *dev);
1da177e4
LT
3706
3707/* Support for loadable net-drivers */
f629d208 3708struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3709 unsigned char name_assign_type,
f629d208
JP
3710 void (*setup)(struct net_device *),
3711 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3712#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3713 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3714
c835a677
TG
3715#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3716 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3717 count)
36909ea4 3718
f629d208
JP
3719int register_netdev(struct net_device *dev);
3720void unregister_netdev(struct net_device *dev);
f001fde5 3721
22bedad3 3722/* General hardware address lists handling functions */
f629d208
JP
3723int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3724 struct netdev_hw_addr_list *from_list, int addr_len);
3725void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3726 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3727int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3728 struct net_device *dev,
3729 int (*sync)(struct net_device *, const unsigned char *),
3730 int (*unsync)(struct net_device *,
3731 const unsigned char *));
3732void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3733 struct net_device *dev,
3734 int (*unsync)(struct net_device *,
3735 const unsigned char *));
f629d208 3736void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3737
f001fde5 3738/* Functions used for device addresses handling */
f629d208
JP
3739int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3740 unsigned char addr_type);
3741int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3742 unsigned char addr_type);
f629d208
JP
3743void dev_addr_flush(struct net_device *dev);
3744int dev_addr_init(struct net_device *dev);
a748ee24
JP
3745
3746/* Functions used for unicast addresses handling */
f629d208
JP
3747int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3748int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3749int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3750int dev_uc_sync(struct net_device *to, struct net_device *from);
3751int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3752void dev_uc_unsync(struct net_device *to, struct net_device *from);
3753void dev_uc_flush(struct net_device *dev);
3754void dev_uc_init(struct net_device *dev);
f001fde5 3755
670e5b8e
AD
3756/**
3757 * __dev_uc_sync - Synchonize device's unicast list
3758 * @dev: device to sync
3759 * @sync: function to call if address should be added
3760 * @unsync: function to call if address should be removed
3761 *
3762 * Add newly added addresses to the interface, and release
3763 * addresses that have been deleted.
5e82b4b2 3764 */
670e5b8e
AD
3765static inline int __dev_uc_sync(struct net_device *dev,
3766 int (*sync)(struct net_device *,
3767 const unsigned char *),
3768 int (*unsync)(struct net_device *,
3769 const unsigned char *))
3770{
3771 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3772}
3773
3774/**
e793c0f7 3775 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3776 * @dev: device to sync
3777 * @unsync: function to call if address should be removed
3778 *
3779 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3780 */
670e5b8e
AD
3781static inline void __dev_uc_unsync(struct net_device *dev,
3782 int (*unsync)(struct net_device *,
3783 const unsigned char *))
3784{
3785 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3786}
3787
22bedad3 3788/* Functions used for multicast addresses handling */
f629d208
JP
3789int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3790int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3791int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3792int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3793int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3794int dev_mc_sync(struct net_device *to, struct net_device *from);
3795int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3796void dev_mc_unsync(struct net_device *to, struct net_device *from);
3797void dev_mc_flush(struct net_device *dev);
3798void dev_mc_init(struct net_device *dev);
f001fde5 3799
670e5b8e
AD
3800/**
3801 * __dev_mc_sync - Synchonize device's multicast list
3802 * @dev: device to sync
3803 * @sync: function to call if address should be added
3804 * @unsync: function to call if address should be removed
3805 *
3806 * Add newly added addresses to the interface, and release
3807 * addresses that have been deleted.
5e82b4b2 3808 */
670e5b8e
AD
3809static inline int __dev_mc_sync(struct net_device *dev,
3810 int (*sync)(struct net_device *,
3811 const unsigned char *),
3812 int (*unsync)(struct net_device *,
3813 const unsigned char *))
3814{
3815 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3816}
3817
3818/**
e793c0f7 3819 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3820 * @dev: device to sync
3821 * @unsync: function to call if address should be removed
3822 *
3823 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3824 */
670e5b8e
AD
3825static inline void __dev_mc_unsync(struct net_device *dev,
3826 int (*unsync)(struct net_device *,
3827 const unsigned char *))
3828{
3829 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3830}
3831
4417da66 3832/* Functions used for secondary unicast and multicast support */
f629d208
JP
3833void dev_set_rx_mode(struct net_device *dev);
3834void __dev_set_rx_mode(struct net_device *dev);
3835int dev_set_promiscuity(struct net_device *dev, int inc);
3836int dev_set_allmulti(struct net_device *dev, int inc);
3837void netdev_state_change(struct net_device *dev);
3838void netdev_notify_peers(struct net_device *dev);
3839void netdev_features_change(struct net_device *dev);
1da177e4 3840/* Load a device via the kmod */
f629d208
JP
3841void dev_load(struct net *net, const char *name);
3842struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3843 struct rtnl_link_stats64 *storage);
3844void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3845 const struct net_device_stats *netdev_stats);
eeda3fd6 3846
1da177e4 3847extern int netdev_max_backlog;
3b098e2d 3848extern int netdev_tstamp_prequeue;
1da177e4 3849extern int weight_p;
3d48b53f
MT
3850extern int dev_weight_rx_bias;
3851extern int dev_weight_tx_bias;
3852extern int dev_rx_weight;
3853extern int dev_tx_weight;
9ff162a8 3854
f629d208 3855bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3856struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3857 struct list_head **iter);
f629d208
JP
3858struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3859 struct list_head **iter);
8b5be856 3860
44a40855
VY
3861/* iterate through upper list, must be called under RCU read lock */
3862#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3863 for (iter = &(dev)->adj_list.upper, \
3864 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3865 updev; \
3866 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3867
1a3f060c
DA
3868int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3869 int (*fn)(struct net_device *upper_dev,
3870 void *data),
3871 void *data);
3872
3873bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3874 struct net_device *upper_dev);
3875
25cc72a3
IS
3876bool netdev_has_any_upper_dev(struct net_device *dev);
3877
f629d208
JP
3878void *netdev_lower_get_next_private(struct net_device *dev,
3879 struct list_head **iter);
3880void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3881 struct list_head **iter);
31088a11
VF
3882
3883#define netdev_for_each_lower_private(dev, priv, iter) \
3884 for (iter = (dev)->adj_list.lower.next, \
3885 priv = netdev_lower_get_next_private(dev, &(iter)); \
3886 priv; \
3887 priv = netdev_lower_get_next_private(dev, &(iter)))
3888
3889#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3890 for (iter = &(dev)->adj_list.lower, \
3891 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3892 priv; \
3893 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3894
4085ebe8
VY
3895void *netdev_lower_get_next(struct net_device *dev,
3896 struct list_head **iter);
7ce856aa 3897
4085ebe8 3898#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3899 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3900 ldev = netdev_lower_get_next(dev, &(iter)); \
3901 ldev; \
3902 ldev = netdev_lower_get_next(dev, &(iter)))
3903
7ce856aa
JP
3904struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3905 struct list_head **iter);
3906struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3907 struct list_head **iter);
3908
1a3f060c
DA
3909int netdev_walk_all_lower_dev(struct net_device *dev,
3910 int (*fn)(struct net_device *lower_dev,
3911 void *data),
3912 void *data);
3913int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3914 int (*fn)(struct net_device *lower_dev,
3915 void *data),
3916 void *data);
3917
f629d208 3918void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3919void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3920struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3921struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3922int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3923int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3924 struct net_device *upper_dev,
29bf24af 3925 void *upper_priv, void *upper_info);
f629d208
JP
3926void netdev_upper_dev_unlink(struct net_device *dev,
3927 struct net_device *upper_dev);
5bb025fa 3928void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3929void *netdev_lower_dev_get_private(struct net_device *dev,
3930 struct net_device *lower_dev);
04d48266
JP
3931void netdev_lower_state_changed(struct net_device *lower_dev,
3932 void *lower_state_info);
960fb622
ED
3933
3934/* RSS keys are 40 or 52 bytes long */
3935#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3936extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3937void netdev_rss_key_fill(void *buffer, size_t len);
3938
952fcfd0 3939int dev_get_nest_level(struct net_device *dev);
f629d208 3940int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 3941int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
3942int skb_csum_hwoffload_help(struct sk_buff *skb,
3943 const netdev_features_t features);
3944
f629d208
JP
3945struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3946 netdev_features_t features, bool tx_path);
3947struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3948 netdev_features_t features);
12b0004d 3949
61bd3857
MS
3950struct netdev_bonding_info {
3951 ifslave slave;
3952 ifbond master;
3953};
3954
3955struct netdev_notifier_bonding_info {
3956 struct netdev_notifier_info info; /* must be first */
3957 struct netdev_bonding_info bonding_info;
3958};
3959
3960void netdev_bonding_info_change(struct net_device *dev,
3961 struct netdev_bonding_info *bonding_info);
3962
12b0004d
CW
3963static inline
3964struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3965{
3966 return __skb_gso_segment(skb, features, true);
3967}
53d6471c 3968__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3969
3970static inline bool can_checksum_protocol(netdev_features_t features,
3971 __be16 protocol)
3972{
c8cd0989
TH
3973 if (protocol == htons(ETH_P_FCOE))
3974 return !!(features & NETIF_F_FCOE_CRC);
3975
3976 /* Assume this is an IP checksum (not SCTP CRC) */
3977
3978 if (features & NETIF_F_HW_CSUM) {
3979 /* Can checksum everything */
3980 return true;
3981 }
3982
3983 switch (protocol) {
3984 case htons(ETH_P_IP):
3985 return !!(features & NETIF_F_IP_CSUM);
3986 case htons(ETH_P_IPV6):
3987 return !!(features & NETIF_F_IPV6_CSUM);
3988 default:
3989 return false;
3990 }
ec5f0615 3991}
12b0004d 3992
fb286bb2 3993#ifdef CONFIG_BUG
f629d208 3994void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3995#else
3996static inline void netdev_rx_csum_fault(struct net_device *dev)
3997{
3998}
3999#endif
1da177e4 4000/* rx skb timestamps */
f629d208
JP
4001void net_enable_timestamp(void);
4002void net_disable_timestamp(void);
1da177e4 4003
20380731 4004#ifdef CONFIG_PROC_FS
f629d208 4005int __init dev_proc_init(void);
900ff8c6
CW
4006#else
4007#define dev_proc_init() 0
20380731
ACM
4008#endif
4009
4798248e 4010static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4011 struct sk_buff *skb, struct net_device *dev,
4012 bool more)
4798248e 4013{
fa2dbdc2 4014 skb->xmit_more = more ? 1 : 0;
0b725a2c 4015 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4016}
4017
10b3ad8c 4018static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4019 struct netdev_queue *txq, bool more)
4798248e
DM
4020{
4021 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4022 int rc;
4798248e 4023
fa2dbdc2 4024 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4025 if (rc == NETDEV_TX_OK)
4026 txq_trans_update(txq);
4027
4028 return rc;
4798248e
DM
4029}
4030
b793dc5c 4031int netdev_class_create_file_ns(const struct class_attribute *class_attr,
42a2d923 4032 const void *ns);
b793dc5c 4033void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
42a2d923 4034 const void *ns);
58292cbe 4035
b793dc5c 4036static inline int netdev_class_create_file(const struct class_attribute *class_attr)
58292cbe
TH
4037{
4038 return netdev_class_create_file_ns(class_attr, NULL);
4039}
4040
b793dc5c 4041static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
58292cbe
TH
4042{
4043 netdev_class_remove_file_ns(class_attr, NULL);
4044}
b8a9787e 4045
737aec57 4046extern const struct kobj_ns_type_operations net_ns_type_operations;
04600794 4047
f629d208 4048const char *netdev_drivername(const struct net_device *dev);
6579e57b 4049
f629d208 4050void linkwatch_run_queue(void);
20380731 4051
da08143b
MK
4052static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4053 netdev_features_t f2)
4054{
c8cd0989
TH
4055 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4056 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4057 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4058 else
b6a0e72a 4059 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4060 }
da08143b 4061
c8cd0989 4062 return f1 & f2;
da08143b
MK
4063}
4064
c8f44aff
MM
4065static inline netdev_features_t netdev_get_wanted_features(
4066 struct net_device *dev)
5455c699
MM
4067{
4068 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4069}
c8f44aff
MM
4070netdev_features_t netdev_increment_features(netdev_features_t all,
4071 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4072
4073/* Allow TSO being used on stacked device :
4074 * Performing the GSO segmentation before last device
4075 * is a performance improvement.
4076 */
4077static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4078 netdev_features_t mask)
4079{
4080 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4081}
4082
6cb6a27c 4083int __netdev_update_features(struct net_device *dev);
5455c699 4084void netdev_update_features(struct net_device *dev);
afe12cc8 4085void netdev_change_features(struct net_device *dev);
7f353bf2 4086
fc4a7489
PM
4087void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4088 struct net_device *dev);
4089
e38f3025
TM
4090netdev_features_t passthru_features_check(struct sk_buff *skb,
4091 struct net_device *dev,
4092 netdev_features_t features);
c1e756bf 4093netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4094
4d29515f 4095static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4096{
7b748340 4097 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4098
4099 /* check flags correspondence */
4100 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4101 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4102 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4103 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4104 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4105 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4106 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4107 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4108 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4109 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4110 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4111 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4112 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4113 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4114 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4115 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0345e186 4116
d6b4991a 4117 return (features & feature) == feature;
576a30eb
HX
4118}
4119
4d29515f 4120static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4121{
278b2513 4122 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4123 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4124}
4125
8b86a61d 4126static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4127 netdev_features_t features)
7967168c 4128{
fc741216 4129 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4130 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4131 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4132}
4133
82cc1a7a
PWJ
4134static inline void netif_set_gso_max_size(struct net_device *dev,
4135 unsigned int size)
4136{
4137 dev->gso_max_size = size;
4138}
4139
7a7ffbab
WCC
4140static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4141 int pulled_hlen, u16 mac_offset,
4142 int mac_len)
4143{
4144 skb->protocol = protocol;
4145 skb->encapsulation = 1;
4146 skb_push(skb, pulled_hlen);
4147 skb_reset_transport_header(skb);
4148 skb->mac_header = mac_offset;
4149 skb->network_header = skb->mac_header + mac_len;
4150 skb->mac_len = mac_len;
4151}
4152
3c175784
SD
4153static inline bool netif_is_macsec(const struct net_device *dev)
4154{
4155 return dev->priv_flags & IFF_MACSEC;
4156}
4157
b618aaa9 4158static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4159{
4160 return dev->priv_flags & IFF_MACVLAN;
4161}
4162
b618aaa9 4163static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4164{
4165 return dev->priv_flags & IFF_MACVLAN_PORT;
4166}
4167
b618aaa9 4168static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4169{
4170 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4171}
4172
b618aaa9 4173static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4174{
4175 return dev->priv_flags & IFF_IPVLAN_MASTER;
4176}
4177
b618aaa9 4178static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4179{
4180 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4181}
4182
b618aaa9 4183static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4184{
4185 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4186}
4187
3bdc0eba
BG
4188static inline bool netif_supports_nofcs(struct net_device *dev)
4189{
4190 return dev->priv_flags & IFF_SUPP_NOFCS;
4191}
4192
007979ea 4193static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4194{
007979ea 4195 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4196}
4197
fee6d4c7
DA
4198static inline bool netif_is_l3_slave(const struct net_device *dev)
4199{
4200 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4201}
4202
0894ae3f
JP
4203static inline bool netif_is_bridge_master(const struct net_device *dev)
4204{
4205 return dev->priv_flags & IFF_EBRIDGE;
4206}
4207
28f9ee22
VY
4208static inline bool netif_is_bridge_port(const struct net_device *dev)
4209{
4210 return dev->priv_flags & IFF_BRIDGE_PORT;
4211}
4212
35d4e172
JP
4213static inline bool netif_is_ovs_master(const struct net_device *dev)
4214{
4215 return dev->priv_flags & IFF_OPENVSWITCH;
4216}
4217
5be66141
JP
4218static inline bool netif_is_ovs_port(const struct net_device *dev)
4219{
4220 return dev->priv_flags & IFF_OVS_DATAPATH;
4221}
4222
b618aaa9 4223static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4224{
4225 return dev->priv_flags & IFF_TEAM;
4226}
4227
b618aaa9 4228static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4229{
4230 return dev->priv_flags & IFF_TEAM_PORT;
4231}
4232
b618aaa9 4233static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4234{
4235 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4236}
4237
b618aaa9 4238static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4239{
4240 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4241}
4242
d4ab4286
KJ
4243static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4244{
4245 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4246}
4247
02875878
ED
4248/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4249static inline void netif_keep_dst(struct net_device *dev)
4250{
4251 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4252}
4253
18d3df3e
PA
4254/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4255static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4256{
4257 /* TODO: reserve and use an additional IFF bit, if we get more users */
4258 return dev->priv_flags & IFF_MACSEC;
4259}
4260
505d4f73 4261extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4262
571ba423
JP
4263/* Logging, debugging and troubleshooting/diagnostic helpers. */
4264
4265/* netdev_printk helpers, similar to dev_printk */
4266
4267static inline const char *netdev_name(const struct net_device *dev)
4268{
c6f854d5
VF
4269 if (!dev->name[0] || strchr(dev->name, '%'))
4270 return "(unnamed net_device)";
571ba423
JP
4271 return dev->name;
4272}
4273
8397ed36
DA
4274static inline bool netdev_unregistering(const struct net_device *dev)
4275{
4276 return dev->reg_state == NETREG_UNREGISTERING;
4277}
4278
ccc7f496
VF
4279static inline const char *netdev_reg_state(const struct net_device *dev)
4280{
4281 switch (dev->reg_state) {
4282 case NETREG_UNINITIALIZED: return " (uninitialized)";
4283 case NETREG_REGISTERED: return "";
4284 case NETREG_UNREGISTERING: return " (unregistering)";
4285 case NETREG_UNREGISTERED: return " (unregistered)";
4286 case NETREG_RELEASED: return " (released)";
4287 case NETREG_DUMMY: return " (dummy)";
4288 }
4289
4290 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4291 return " (unknown)";
4292}
4293
f629d208 4294__printf(3, 4)
6ea754eb
JP
4295void netdev_printk(const char *level, const struct net_device *dev,
4296 const char *format, ...);
f629d208 4297__printf(2, 3)
6ea754eb 4298void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4299__printf(2, 3)
6ea754eb 4300void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4301__printf(2, 3)
6ea754eb 4302void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4303__printf(2, 3)
6ea754eb 4304void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4305__printf(2, 3)
6ea754eb 4306void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4307__printf(2, 3)
6ea754eb 4308void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4309__printf(2, 3)
6ea754eb 4310void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4311
8909c9ad
VK
4312#define MODULE_ALIAS_NETDEV(device) \
4313 MODULE_ALIAS("netdev-" device)
4314
b558c96f 4315#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4316#define netdev_dbg(__dev, format, args...) \
4317do { \
ffa10cb4 4318 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4319} while (0)
b558c96f
JC
4320#elif defined(DEBUG)
4321#define netdev_dbg(__dev, format, args...) \
4322 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4323#else
4324#define netdev_dbg(__dev, format, args...) \
4325({ \
4326 if (0) \
4327 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4328})
4329#endif
4330
4331#if defined(VERBOSE_DEBUG)
4332#define netdev_vdbg netdev_dbg
4333#else
4334
4335#define netdev_vdbg(dev, format, args...) \
4336({ \
4337 if (0) \
4338 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4339 0; \
4340})
4341#endif
4342
4343/*
4344 * netdev_WARN() acts like dev_printk(), but with the key difference
4345 * of using a WARN/WARN_ON to get the message out, including the
4346 * file/line information and a backtrace.
4347 */
4348#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4349 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4350 netdev_reg_state(dev), ##args)
571ba423 4351
b3d95c5c
JP
4352/* netif printk helpers, similar to netdev_printk */
4353
4354#define netif_printk(priv, type, level, dev, fmt, args...) \
4355do { \
4356 if (netif_msg_##type(priv)) \
4357 netdev_printk(level, (dev), fmt, ##args); \
4358} while (0)
4359
f45f4321
JP
4360#define netif_level(level, priv, type, dev, fmt, args...) \
4361do { \
4362 if (netif_msg_##type(priv)) \
4363 netdev_##level(dev, fmt, ##args); \
4364} while (0)
4365
b3d95c5c 4366#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4367 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4368#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4369 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4370#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4371 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4372#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4373 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4374#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4375 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4376#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4377 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4378#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4379 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4380
0053ea9c 4381#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4382#define netif_dbg(priv, type, netdev, format, args...) \
4383do { \
4384 if (netif_msg_##type(priv)) \
b5fb0a03 4385 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4386} while (0)
0053ea9c
JP
4387#elif defined(DEBUG)
4388#define netif_dbg(priv, type, dev, format, args...) \
4389 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4390#else
4391#define netif_dbg(priv, type, dev, format, args...) \
4392({ \
4393 if (0) \
4394 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4395 0; \
4396})
4397#endif
4398
f617f276
EC
4399/* if @cond then downgrade to debug, else print at @level */
4400#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4401 do { \
4402 if (cond) \
4403 netif_dbg(priv, type, netdev, fmt, ##args); \
4404 else \
4405 netif_ ## level(priv, type, netdev, fmt, ##args); \
4406 } while (0)
4407
b3d95c5c 4408#if defined(VERBOSE_DEBUG)
bcfcc450 4409#define netif_vdbg netif_dbg
b3d95c5c
JP
4410#else
4411#define netif_vdbg(priv, type, dev, format, args...) \
4412({ \
4413 if (0) \
a4ed89cb 4414 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4415 0; \
4416})
4417#endif
571ba423 4418
900ff8c6
CW
4419/*
4420 * The list of packet types we will receive (as opposed to discard)
4421 * and the routines to invoke.
4422 *
4423 * Why 16. Because with 16 the only overlap we get on a hash of the
4424 * low nibble of the protocol value is RARP/SNAP/X.25.
4425 *
4426 * NOTE: That is no longer true with the addition of VLAN tags. Not
4427 * sure which should go first, but I bet it won't make much
4428 * difference if we are running VLANs. The good news is that
4429 * this protocol won't be in the list unless compiled in, so
4430 * the average user (w/out VLANs) will not be adversely affected.
4431 * --BLG
4432 *
4433 * 0800 IP
4434 * 8100 802.1Q VLAN
4435 * 0001 802.3
4436 * 0002 AX.25
4437 * 0004 802.2
4438 * 8035 RARP
4439 * 0005 SNAP
4440 * 0805 X.25
4441 * 0806 ARP
4442 * 8137 IPX
4443 * 0009 Localtalk
4444 * 86DD IPv6
4445 */
4446#define PTYPE_HASH_SIZE (16)
4447#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4448
385a154c 4449#endif /* _LINUX_NETDEVICE_H */