]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/netdevice.h
net: rtnetlink: validate IFLA_MTU attribute in rtnl_create_link()
[mirror_ubuntu-bionic-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
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 {
575ed7d3 773 TC_SETUP_QDISC_MQPRIO,
a1b7c5fd 774 TC_SETUP_CLSU32,
5b33f488 775 TC_SETUP_CLSFLOWER,
ade9b658 776 TC_SETUP_CLSMATCHALL,
332ae8e2 777 TC_SETUP_CLSBPF,
8c4083b3 778 TC_SETUP_BLOCK,
8521db4c 779 TC_SETUP_QDISC_CBS,
602f3baf 780 TC_SETUP_QDISC_RED,
16e5cc64
JF
781};
782
f4e63525
JK
783/* These structures hold the attributes of bpf state that are being passed
784 * to the netdevice through the bpf op.
a7862b45 785 */
f4e63525 786enum bpf_netdev_command {
a7862b45
BB
787 /* Set or clear a bpf program used in the earliest stages of packet
788 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
789 * is responsible for calling bpf_prog_put on any old progs that are
790 * stored. In case of error, the callee need not release the new prog
791 * reference, but on success it takes ownership and must bpf_prog_put
792 * when it is no longer used.
793 */
794 XDP_SETUP_PROG,
ee5d032f 795 XDP_SETUP_PROG_HW,
a7862b45 796 /* Check if a bpf program is set on the device. The callee should
ce158e58
JK
797 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
798 * is equivalent to XDP_ATTACHED_DRV.
a7862b45
BB
799 */
800 XDP_QUERY_PROG,
ab3f0063
JK
801 /* BPF program for offload callbacks, invoked at program load time. */
802 BPF_OFFLOAD_VERIFIER_PREP,
803 BPF_OFFLOAD_TRANSLATE,
804 BPF_OFFLOAD_DESTROY,
a7862b45
BB
805};
806
ab3f0063 807struct bpf_ext_analyzer_ops;
ddf9f970
JK
808struct netlink_ext_ack;
809
f4e63525
JK
810struct netdev_bpf {
811 enum bpf_netdev_command command;
a7862b45
BB
812 union {
813 /* XDP_SETUP_PROG */
ddf9f970 814 struct {
32d60277 815 u32 flags;
ddf9f970
JK
816 struct bpf_prog *prog;
817 struct netlink_ext_ack *extack;
818 };
a7862b45 819 /* XDP_QUERY_PROG */
58038695 820 struct {
ce158e58 821 u8 prog_attached;
58038695
MKL
822 u32 prog_id;
823 };
ab3f0063
JK
824 /* BPF_OFFLOAD_VERIFIER_PREP */
825 struct {
826 struct bpf_prog *prog;
827 const struct bpf_ext_analyzer_ops *ops; /* callee set */
828 } verifier;
829 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
830 struct {
831 struct bpf_prog *prog;
832 } offload;
a7862b45
BB
833 };
834};
16e5cc64 835
d77e38e6
SK
836#ifdef CONFIG_XFRM_OFFLOAD
837struct xfrmdev_ops {
838 int (*xdo_dev_state_add) (struct xfrm_state *x);
839 void (*xdo_dev_state_delete) (struct xfrm_state *x);
840 void (*xdo_dev_state_free) (struct xfrm_state *x);
841 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
842 struct xfrm_state *x);
843};
844#endif
845
6c557001
FW
846struct dev_ifalias {
847 struct rcu_head rcuhead;
848 char ifalias[];
849};
850
d314774c
SH
851/*
852 * This structure defines the management hooks for network devices.
00829823
SH
853 * The following hooks can be defined; unless noted otherwise, they are
854 * optional and can be filled with a null pointer.
d314774c
SH
855 *
856 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
857 * This function is called once when a network device is registered.
858 * The network device can use this for any late stage initialization
859 * or semantic validation. It can fail with an error code which will
860 * be propagated back to register_netdev.
d314774c
SH
861 *
862 * void (*ndo_uninit)(struct net_device *dev);
863 * This function is called when device is unregistered or when registration
864 * fails. It is not called if init fails.
865 *
866 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 867 * This function is called when a network device transitions to the up
d314774c
SH
868 * state.
869 *
870 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 871 * This function is called when a network device transitions to the down
d314774c
SH
872 * state.
873 *
dc1f8bf6
SH
874 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
875 * struct net_device *dev);
00829823 876 * Called when a packet needs to be transmitted.
e79d8429
RR
877 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
878 * the queue before that can happen; it's for obsolete devices and weird
879 * corner cases, but the stack really does a non-trivial amount
880 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 881 * Required; cannot be NULL.
00829823 882 *
1a2a1444
DM
883 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
884 * struct net_device *dev
885 * netdev_features_t features);
886 * Called by core transmit path to determine if device is capable of
887 * performing offload operations on a given packet. This is to give
888 * the device an opportunity to implement any restrictions that cannot
889 * be otherwise expressed by feature flags. The check is called with
890 * the set of features that the stack has calculated and it returns
891 * those the driver believes to be appropriate.
cdba756f 892 *
f663dd9a 893 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 894 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 895 * Called to decide which queue to use when device supports multiple
00829823
SH
896 * transmit queues.
897 *
d314774c
SH
898 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
899 * This function is called to allow device receiver to make
5e82b4b2 900 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
901 *
902 * void (*ndo_set_rx_mode)(struct net_device *dev);
903 * This function is called device changes address list filtering.
01789349 904 * If driver handles unicast address filtering, it should set
5e82b4b2 905 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
906 *
907 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
908 * This function is called when the Media Access Control address
37b607c5 909 * needs to be changed. If this interface is not defined, the
5e82b4b2 910 * MAC address can not be changed.
d314774c
SH
911 *
912 * int (*ndo_validate_addr)(struct net_device *dev);
913 * Test if Media Access Control address is valid for the device.
914 *
915 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
916 * Called when a user requests an ioctl which can't be handled by
917 * the generic interface code. If not defined ioctls return
d314774c
SH
918 * not supported error code.
919 *
920 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
921 * Used to set network devices bus interface parameters. This interface
5e82b4b2 922 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
923 * interface (PCI) for low level management.
924 *
925 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
926 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 927 * of a device.
d314774c 928 *
00829823 929 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 930 * Callback used when the transmitter has not made any progress
d314774c
SH
931 * for dev->watchdog ticks.
932 *
bc1f4470 933 * void (*ndo_get_stats64)(struct net_device *dev,
934 * struct rtnl_link_stats64 *storage);
d308e38f 935 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 936 * Called when a user wants to get the network device usage
be1f3c2c 937 * statistics. Drivers must do one of the following:
3cfde79c
BH
938 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
939 * rtnl_link_stats64 structure passed by the caller.
82695d9b 940 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
941 * (which should normally be dev->stats) and return a pointer to
942 * it. The structure may be changed asynchronously only if each
943 * field is written atomically.
944 * 3. Update dev->stats asynchronously and atomically, and define
945 * neither operation.
d314774c 946 *
3df5b3c6 947 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
948 * Return true if this device supports offload stats of this attr_id.
949 *
950 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
951 * void *attr_data)
952 * Get statistics for offload operations by attr_id. Write it into the
953 * attr_data pointer.
954 *
5d632cb7 955 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 956 * If device supports VLAN filtering this function is called when a
80d5c368 957 * VLAN id is registered.
d314774c 958 *
5d632cb7 959 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 960 * If device supports VLAN filtering this function is called when a
80d5c368 961 * VLAN id is unregistered.
d314774c
SH
962 *
963 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
964 *
965 * SR-IOV management functions.
966 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
967 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
968 * u8 qos, __be16 proto);
ed616689
SC
969 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
970 * int max_tx_rate);
5f8444a3 971 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 972 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
973 * int (*ndo_get_vf_config)(struct net_device *dev,
974 * int vf, struct ifla_vf_info *ivf);
1d8faf48 975 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
976 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
977 * struct nlattr *port[]);
01a3d796
VZ
978 *
979 * Enable or disable the VF ability to query its RSS Redirection Table and
980 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 981 * with PF and querying it may introduce a theoretical security risk.
01a3d796 982 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 983 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
2572ac53 984 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
de4784ca 985 * void *type_data);
6a4bc2b4
FF
986 * Called to setup any 'tc' scheduler, classifier or action on @dev.
987 * This is always called from the stack with the rtnl lock held and netif
988 * tx queues stopped. This allows the netdevice to perform queue
989 * management safely.
c445477d 990 *
e9bce845
YZ
991 * Fiber Channel over Ethernet (FCoE) offload functions.
992 * int (*ndo_fcoe_enable)(struct net_device *dev);
993 * Called when the FCoE protocol stack wants to start using LLD for FCoE
994 * so the underlying device can perform whatever needed configuration or
995 * initialization to support acceleration of FCoE traffic.
996 *
997 * int (*ndo_fcoe_disable)(struct net_device *dev);
998 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
999 * so the underlying device can perform whatever needed clean-ups to
1000 * stop supporting acceleration of FCoE traffic.
1001 *
1002 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1003 * struct scatterlist *sgl, unsigned int sgc);
1004 * Called when the FCoE Initiator wants to initialize an I/O that
1005 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1006 * perform necessary setup and returns 1 to indicate the device is set up
1007 * successfully to perform DDP on this I/O, otherwise this returns 0.
1008 *
1009 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1010 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1011 * indicated by the FC exchange id 'xid', so the underlying device can
1012 * clean up and reuse resources for later DDP requests.
1013 *
1014 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1015 * struct scatterlist *sgl, unsigned int sgc);
1016 * Called when the FCoE Target wants to initialize an I/O that
1017 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1018 * perform necessary setup and returns 1 to indicate the device is set up
1019 * successfully to perform DDP on this I/O, otherwise this returns 0.
1020 *
68bad94e
NP
1021 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1022 * struct netdev_fcoe_hbainfo *hbainfo);
1023 * Called when the FCoE Protocol stack wants information on the underlying
1024 * device. This information is utilized by the FCoE protocol stack to
1025 * register attributes with Fiber Channel management service as per the
1026 * FC-GS Fabric Device Management Information(FDMI) specification.
1027 *
e9bce845
YZ
1028 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1029 * Called when the underlying device wants to override default World Wide
1030 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1031 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1032 * protocol stack to use.
1033 *
c445477d
BH
1034 * RFS acceleration.
1035 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1036 * u16 rxq_index, u32 flow_id);
1037 * Set hardware filter for RFS. rxq_index is the target queue index;
1038 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1039 * Return the filter ID on success, or a negative error code.
fbaec0ea 1040 *
8b98a70c 1041 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1042 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1043 * Called to make another netdev an underling.
1044 *
1045 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1046 * Called to release previously enslaved netdev.
5455c699
MM
1047 *
1048 * Feature/offload setting functions.
1a2a1444
DM
1049 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1050 * netdev_features_t features);
1051 * Adjusts the requested feature flags according to device-specific
1052 * constraints, and returns the resulting flags. Must not modify
1053 * the device state.
1054 *
c8f44aff 1055 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1056 * Called to update device configuration to new features. Passed
1057 * feature set might be less than what was returned by ndo_fix_features()).
1058 * Must return >0 or -errno if it changed dev->features itself.
1059 *
edc7d573 1060 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1061 * struct net_device *dev,
f6f6424b 1062 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1063 * Adds an FDB entry to dev for addr.
1690be63
VY
1064 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1065 * struct net_device *dev,
f6f6424b 1066 * const unsigned char *addr, u16 vid)
77162022
JF
1067 * Deletes the FDB entry from dev coresponding to addr.
1068 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1069 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1070 * int *idx)
77162022
JF
1071 * Used to add FDB entries to dump requests. Implementers should add
1072 * entries to skb and update idx with the number of entries.
e5a55a89 1073 *
ad41faa8
ND
1074 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1075 * u16 flags)
e5a55a89 1076 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1077 * struct net_device *dev, u32 filter_mask,
1078 * int nlflags)
ad41faa8
ND
1079 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1080 * u16 flags);
4bf84c35
JP
1081 *
1082 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1083 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1084 * which do not represent real hardware may define this to allow their
1085 * userspace components to manage their virtual carrier state. Devices
1086 * that determine carrier state from physical hardware properties (eg
1087 * network cables) or protocol-dependent mechanisms (eg
1088 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1089 *
1090 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1091 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1092 * Called to get ID of physical port of this device. If driver does
1093 * not implement this, it is assumed that the hw is not able to have
1094 * multiple net devices on single physical port.
53cf5275 1095 *
7c46a640
AD
1096 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1097 * struct udp_tunnel_info *ti);
1098 * Called by UDP tunnel to notify a driver about the UDP port and socket
1099 * address family that a UDP tunnel is listnening to. It is called only
1100 * when a new port starts listening. The operation is protected by the
1101 * RTNL.
1102 *
1103 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1104 * struct udp_tunnel_info *ti);
1105 * Called by UDP tunnel to notify the driver about a UDP port and socket
1106 * address family that the UDP tunnel is not listening to anymore. The
1107 * operation is protected by the RTNL.
1108 *
a6cc0cfa
JF
1109 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1110 * struct net_device *dev)
1111 * Called by upper layer devices to accelerate switching or other
1112 * station functionality into hardware. 'pdev is the lowerdev
1113 * to use for the offload and 'dev' is the net device that will
1114 * back the offload. Returns a pointer to the private structure
1115 * the upper layer will maintain.
1116 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1117 * Called by upper layer device to delete the station created
1118 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1119 * the station and priv is the structure returned by the add
1120 * operation.
822b3b2e
JF
1121 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1122 * int queue_index, u32 maxrate);
1123 * Called when a user wants to set a max-rate limitation of specific
1124 * TX queue.
a54acb3a
ND
1125 * int (*ndo_get_iflink)(const struct net_device *dev);
1126 * Called to get the iflink value of this device.
d746d707 1127 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1128 * bool proto_down);
d746d707
AK
1129 * This function is used to pass protocol port error state information
1130 * to the switch driver. The switch driver can react to the proto_down
1131 * by doing a phys down on the associated switch port.
fc4099f1
PS
1132 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1133 * This function is used to get egress tunnel information for given skb.
1134 * This is useful for retrieving outer tunnel header parameters while
1135 * sampling packet.
871b642a
PA
1136 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1137 * This function is used to specify the headroom that the skb must
1138 * consider when allocation skb during packet reception. Setting
1139 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1140 * forward. Setting a negative value resets the rx headroom to the
871b642a 1141 * default value.
f4e63525 1142 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
a7862b45 1143 * This function is used to set or query state related to XDP on the
f4e63525
JK
1144 * netdevice and manage BPF offload. See definition of
1145 * enum bpf_netdev_command for details.
814abfab
JF
1146 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1147 * This function is used to submit a XDP packet for transmit on a
1148 * netdevice.
11393cc9 1149 * void (*ndo_xdp_flush)(struct net_device *dev);
f16ded59
RR
1150 * This function is used to inform the driver to flush a particular
1151 * xdp tx queue. Must be called on same CPU as xdp_xmit.
d314774c
SH
1152 */
1153struct net_device_ops {
1154 int (*ndo_init)(struct net_device *dev);
1155 void (*ndo_uninit)(struct net_device *dev);
1156 int (*ndo_open)(struct net_device *dev);
1157 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1158 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1159 struct net_device *dev);
1160 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1161 struct net_device *dev,
1162 netdev_features_t features);
00829823 1163 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1164 struct sk_buff *skb,
99932d4f
DB
1165 void *accel_priv,
1166 select_queue_fallback_t fallback);
d314774c
SH
1167 void (*ndo_change_rx_flags)(struct net_device *dev,
1168 int flags);
d314774c 1169 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1170 int (*ndo_set_mac_address)(struct net_device *dev,
1171 void *addr);
d314774c 1172 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1173 int (*ndo_do_ioctl)(struct net_device *dev,
1174 struct ifreq *ifr, int cmd);
d314774c
SH
1175 int (*ndo_set_config)(struct net_device *dev,
1176 struct ifmap *map);
00829823
SH
1177 int (*ndo_change_mtu)(struct net_device *dev,
1178 int new_mtu);
1179 int (*ndo_neigh_setup)(struct net_device *dev,
1180 struct neigh_parms *);
d314774c
SH
1181 void (*ndo_tx_timeout) (struct net_device *dev);
1182
bc1f4470 1183 void (*ndo_get_stats64)(struct net_device *dev,
1184 struct rtnl_link_stats64 *storage);
3df5b3c6 1185 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1186 int (*ndo_get_offload_stats)(int attr_id,
1187 const struct net_device *dev,
1188 void *attr_data);
d314774c
SH
1189 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1190
8e586137 1191 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1192 __be16 proto, u16 vid);
8e586137 1193 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1194 __be16 proto, u16 vid);
d314774c 1195#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1196 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1197 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1198 struct netpoll_info *info);
0e34e931 1199 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1200#endif
95c26df8
WM
1201 int (*ndo_set_vf_mac)(struct net_device *dev,
1202 int queue, u8 *mac);
1203 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1204 int queue, u16 vlan,
1205 u8 qos, __be16 proto);
ed616689
SC
1206 int (*ndo_set_vf_rate)(struct net_device *dev,
1207 int vf, int min_tx_rate,
1208 int max_tx_rate);
5f8444a3
GR
1209 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1210 int vf, bool setting);
dd461d6a
HS
1211 int (*ndo_set_vf_trust)(struct net_device *dev,
1212 int vf, bool setting);
95c26df8
WM
1213 int (*ndo_get_vf_config)(struct net_device *dev,
1214 int vf,
1215 struct ifla_vf_info *ivf);
1d8faf48
RE
1216 int (*ndo_set_vf_link_state)(struct net_device *dev,
1217 int vf, int link_state);
3b766cd8
EBE
1218 int (*ndo_get_vf_stats)(struct net_device *dev,
1219 int vf,
1220 struct ifla_vf_stats
1221 *vf_stats);
57b61080
SF
1222 int (*ndo_set_vf_port)(struct net_device *dev,
1223 int vf,
1224 struct nlattr *port[]);
1225 int (*ndo_get_vf_port)(struct net_device *dev,
1226 int vf, struct sk_buff *skb);
cc8e27cc
EC
1227 int (*ndo_set_vf_guid)(struct net_device *dev,
1228 int vf, u64 guid,
1229 int guid_type);
01a3d796
VZ
1230 int (*ndo_set_vf_rss_query_en)(
1231 struct net_device *dev,
1232 int vf, bool setting);
16e5cc64 1233 int (*ndo_setup_tc)(struct net_device *dev,
2572ac53 1234 enum tc_setup_type type,
de4784ca 1235 void *type_data);
d11ead75 1236#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1237 int (*ndo_fcoe_enable)(struct net_device *dev);
1238 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1239 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1240 u16 xid,
1241 struct scatterlist *sgl,
1242 unsigned int sgc);
1243 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1244 u16 xid);
6247e086
YZ
1245 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1246 u16 xid,
1247 struct scatterlist *sgl,
1248 unsigned int sgc);
68bad94e
NP
1249 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1250 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1251#endif
1252
d11ead75 1253#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1254#define NETDEV_FCOE_WWNN 0
1255#define NETDEV_FCOE_WWPN 1
1256 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1257 u64 *wwn, int type);
4d288d57 1258#endif
3c9c36bc 1259
c445477d
BH
1260#ifdef CONFIG_RFS_ACCEL
1261 int (*ndo_rx_flow_steer)(struct net_device *dev,
1262 const struct sk_buff *skb,
1263 u16 rxq_index,
1264 u32 flow_id);
1265#endif
fbaec0ea 1266 int (*ndo_add_slave)(struct net_device *dev,
33eaf2a6
DA
1267 struct net_device *slave_dev,
1268 struct netlink_ext_ack *extack);
fbaec0ea
JP
1269 int (*ndo_del_slave)(struct net_device *dev,
1270 struct net_device *slave_dev);
c8f44aff
MM
1271 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1272 netdev_features_t features);
5455c699 1273 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1274 netdev_features_t features);
503eebc2
JP
1275 int (*ndo_neigh_construct)(struct net_device *dev,
1276 struct neighbour *n);
1277 void (*ndo_neigh_destroy)(struct net_device *dev,
1278 struct neighbour *n);
77162022
JF
1279
1280 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1281 struct nlattr *tb[],
77162022 1282 struct net_device *dev,
6b6e2725 1283 const unsigned char *addr,
f6f6424b 1284 u16 vid,
77162022
JF
1285 u16 flags);
1286 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1287 struct nlattr *tb[],
77162022 1288 struct net_device *dev,
f6f6424b
JP
1289 const unsigned char *addr,
1290 u16 vid);
77162022
JF
1291 int (*ndo_fdb_dump)(struct sk_buff *skb,
1292 struct netlink_callback *cb,
1293 struct net_device *dev,
5d5eacb3 1294 struct net_device *filter_dev,
d297653d 1295 int *idx);
e5a55a89
JF
1296
1297 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1298 struct nlmsghdr *nlh,
1299 u16 flags);
e5a55a89
JF
1300 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1301 u32 pid, u32 seq,
6cbdceeb 1302 struct net_device *dev,
46c264da
ND
1303 u32 filter_mask,
1304 int nlflags);
407af329 1305 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1306 struct nlmsghdr *nlh,
1307 u16 flags);
4bf84c35
JP
1308 int (*ndo_change_carrier)(struct net_device *dev,
1309 bool new_carrier);
66b52b0d 1310 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1311 struct netdev_phys_item_id *ppid);
db24a904
DA
1312 int (*ndo_get_phys_port_name)(struct net_device *dev,
1313 char *name, size_t len);
7c46a640
AD
1314 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1315 struct udp_tunnel_info *ti);
1316 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1317 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1318 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1319 struct net_device *dev);
1320 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1321 void *priv);
1322
25175ba5 1323 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1324 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1325 int queue_index,
1326 u32 maxrate);
a54acb3a 1327 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1328 int (*ndo_change_proto_down)(struct net_device *dev,
1329 bool proto_down);
fc4099f1
PS
1330 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1331 struct sk_buff *skb);
871b642a
PA
1332 void (*ndo_set_rx_headroom)(struct net_device *dev,
1333 int needed_headroom);
f4e63525
JK
1334 int (*ndo_bpf)(struct net_device *dev,
1335 struct netdev_bpf *bpf);
814abfab
JF
1336 int (*ndo_xdp_xmit)(struct net_device *dev,
1337 struct xdp_buff *xdp);
11393cc9 1338 void (*ndo_xdp_flush)(struct net_device *dev);
d314774c
SH
1339};
1340
7aa98047
LR
1341/**
1342 * enum net_device_priv_flags - &struct net_device priv_flags
1343 *
1344 * These are the &struct net_device, they are only set internally
1345 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1346 * userspace; this means that the order of these flags can change
7aa98047
LR
1347 * during any kernel release.
1348 *
1349 * You should have a pretty good reason to be extending these flags.
1350 *
1351 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1352 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1353 * @IFF_BONDING: bonding master or slave
7aa98047 1354 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1355 * @IFF_WAN_HDLC: WAN HDLC device
1356 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1357 * release skb->dst
1358 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1359 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1360 * @IFF_MACVLAN_PORT: device used as macvlan port
1361 * @IFF_BRIDGE_PORT: device used as bridge port
1362 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1363 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1364 * @IFF_UNICAST_FLT: Supports unicast filtering
1365 * @IFF_TEAM_PORT: device used as team port
1366 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1367 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1368 * change when it's running
1369 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1370 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1371 * underlying stacked devices
1372 * @IFF_IPVLAN_MASTER: IPvlan master device
1373 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1374 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1375 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1376 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1377 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1378 * @IFF_TEAM: device is a team device
d4ab4286 1379 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1380 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1381 * entity (i.e. the master device for bridged veth)
3c175784 1382 * @IFF_MACSEC: device is a MACsec device
9d150e29 1383 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
7aa98047
LR
1384 */
1385enum netdev_priv_flags {
1386 IFF_802_1Q_VLAN = 1<<0,
1387 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1388 IFF_BONDING = 1<<2,
1389 IFF_ISATAP = 1<<3,
1390 IFF_WAN_HDLC = 1<<4,
1391 IFF_XMIT_DST_RELEASE = 1<<5,
1392 IFF_DONT_BRIDGE = 1<<6,
1393 IFF_DISABLE_NETPOLL = 1<<7,
1394 IFF_MACVLAN_PORT = 1<<8,
1395 IFF_BRIDGE_PORT = 1<<9,
1396 IFF_OVS_DATAPATH = 1<<10,
1397 IFF_TX_SKB_SHARING = 1<<11,
1398 IFF_UNICAST_FLT = 1<<12,
1399 IFF_TEAM_PORT = 1<<13,
1400 IFF_SUPP_NOFCS = 1<<14,
1401 IFF_LIVE_ADDR_CHANGE = 1<<15,
1402 IFF_MACVLAN = 1<<16,
1403 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1404 IFF_IPVLAN_MASTER = 1<<18,
1405 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1406 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1407 IFF_NO_QUEUE = 1<<21,
1408 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1409 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1410 IFF_TEAM = 1<<24,
d4ab4286 1411 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1412 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1413 IFF_MACSEC = 1<<27,
9d150e29 1414 IFF_L3MDEV_RX_HANDLER = 1<<28,
7aa98047
LR
1415};
1416
1417#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1418#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1419#define IFF_BONDING IFF_BONDING
7aa98047 1420#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1421#define IFF_WAN_HDLC IFF_WAN_HDLC
1422#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1423#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1424#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1425#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1426#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1427#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1428#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1429#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1430#define IFF_TEAM_PORT IFF_TEAM_PORT
1431#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1432#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1433#define IFF_MACVLAN IFF_MACVLAN
02875878 1434#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1435#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1436#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1437#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1438#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1439#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1440#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1441#define IFF_TEAM IFF_TEAM
d4ab4286 1442#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1443#define IFF_MACSEC IFF_MACSEC
9d150e29 1444#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
7aa98047 1445
536721b1
KK
1446/**
1447 * struct net_device - The DEVICE structure.
d651983d
MCC
1448 *
1449 * Actually, this whole structure is a big mistake. It mixes I/O
1450 * data with strictly "high-level" data, and it has to know about
1451 * almost every data structure used in the INET module.
536721b1
KK
1452 *
1453 * @name: This is the first field of the "visible" part of this structure
1454 * (i.e. as seen by users in the "Space.c" file). It is the name
d651983d 1455 * of the interface.
536721b1
KK
1456 *
1457 * @name_hlist: Device name hash chain, please keep it close to name[]
1458 * @ifalias: SNMP alias
1459 * @mem_end: Shared memory end
1460 * @mem_start: Shared memory start
1461 * @base_addr: Device I/O address
1462 * @irq: Device IRQ number
1463 *
14ffbbb8
TG
1464 * @carrier_changes: Stats to monitor carrier on<->off transitions
1465 *
536721b1
KK
1466 * @state: Generic network queuing layer state, see netdev_state_t
1467 * @dev_list: The global list of network devices
5e82b4b2
BH
1468 * @napi_list: List entry used for polling NAPI devices
1469 * @unreg_list: List entry when we are unregistering the
1470 * device; see the function unregister_netdev
1471 * @close_list: List entry used when we are closing the device
62d885fe
BP
1472 * @ptype_all: Device-specific packet handlers for all protocols
1473 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1474 *
1475 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1476 * @features: Currently active device features
1477 * @hw_features: User-changeable features
1478 *
1479 * @wanted_features: User-requested features
1480 * @vlan_features: Mask of features inheritable by VLAN devices
1481 *
1482 * @hw_enc_features: Mask of features inherited by encapsulating devices
1483 * This field indicates what encapsulation
1484 * offloads the hardware is capable of doing,
1485 * and drivers will need to set them appropriately.
1486 *
1487 * @mpls_features: Mask of features inheritable by MPLS
1488 *
1489 * @ifindex: interface index
5e82b4b2 1490 * @group: The group the device belongs to
536721b1
KK
1491 *
1492 * @stats: Statistics struct, which was left as a legacy, use
1493 * rtnl_link_stats64 instead
1494 *
1495 * @rx_dropped: Dropped packets by core network,
1496 * do not use this in drivers
1497 * @tx_dropped: Dropped packets by core network,
1498 * do not use this in drivers
6e7333d3
JW
1499 * @rx_nohandler: nohandler dropped packets by core network on
1500 * inactive devices, do not use this in drivers
536721b1 1501 *
536721b1
KK
1502 * @wireless_handlers: List of functions to handle Wireless Extensions,
1503 * instead of ioctl,
1504 * see <net/iw_handler.h> for details.
1505 * @wireless_data: Instance data managed by the core of wireless extensions
1506 *
1507 * @netdev_ops: Includes several pointers to callbacks,
1508 * if one wants to override the ndo_*() functions
1509 * @ethtool_ops: Management operations
f997c55c
AA
1510 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1511 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1512 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1513 * of Layer 2 headers.
1514 *
1515 * @flags: Interface flags (a la BSD)
1516 * @priv_flags: Like 'flags' but invisible to userspace,
1517 * see if.h for the definitions
1518 * @gflags: Global flags ( kept as legacy )
1519 * @padded: How much padding added by alloc_netdev()
1520 * @operstate: RFC2863 operstate
1521 * @link_mode: Mapping policy to operstate
1522 * @if_port: Selectable AUI, TP, ...
1523 * @dma: DMA channel
1524 * @mtu: Interface MTU value
61e84623
JW
1525 * @min_mtu: Interface Minimum MTU value
1526 * @max_mtu: Interface Maximum MTU value
536721b1 1527 * @type: Interface hardware type
2793a23a 1528 * @hard_header_len: Maximum hardware header length.
217e6fa2 1529 * @min_header_len: Minimum hardware header length
536721b1
KK
1530 *
1531 * @needed_headroom: Extra headroom the hardware may need, but not in all
1532 * cases can this be guaranteed
1533 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1534 * cases can this be guaranteed. Some cases also use
1535 * LL_MAX_HEADER instead to allocate the skb
1536 *
1537 * interface address info:
1538 *
1539 * @perm_addr: Permanent hw address
1540 * @addr_assign_type: Hw address assignment type
1541 * @addr_len: Hardware address length
b634e72d
TY
1542 * @upper_level: Maximum depth level of upper devices.
1543 * @lower_level: Maximum depth level of lower devices.
8626a0c8 1544 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1545 * @dev_id: Used to differentiate devices that share
1546 * the same link layer address
1547 * @dev_port: Used to differentiate devices that share
1548 * the same function
1549 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1550 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1551 * has been enabled due to the need to listen to
1552 * additional unicast addresses in a device that
1553 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1554 * @uc: unicast mac addresses
1555 * @mc: multicast mac addresses
1556 * @dev_addrs: list of device hw addresses
1557 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1558 * @promiscuity: Number of times the NIC is told to work in
1559 * promiscuous mode; if it becomes 0 the NIC will
1560 * exit promiscuous mode
536721b1
KK
1561 * @allmulti: Counter, enables or disables allmulticast mode
1562 *
1563 * @vlan_info: VLAN info
1564 * @dsa_ptr: dsa specific data
1565 * @tipc_ptr: TIPC specific data
1566 * @atalk_ptr: AppleTalk link
1567 * @ip_ptr: IPv4 specific data
1568 * @dn_ptr: DECnet specific data
1569 * @ip6_ptr: IPv6 specific data
1570 * @ax25_ptr: AX.25 specific data
1571 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1572 *
536721b1
KK
1573 * @dev_addr: Hw address (before bcast,
1574 * because most packets are unicast)
1575 *
1576 * @_rx: Array of RX queues
1577 * @num_rx_queues: Number of RX queues
1578 * allocated at register_netdev() time
1579 * @real_num_rx_queues: Number of RX queues currently active in device
1580 *
1581 * @rx_handler: handler for received packets
1582 * @rx_handler_data: XXX: need comments on this one
46209401
JP
1583 * @miniq_ingress: ingress/clsact qdisc specific data for
1584 * ingress processing
536721b1
KK
1585 * @ingress_queue: XXX: need comments on this one
1586 * @broadcast: hw bcast address
1587 *
14ffbbb8
TG
1588 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1589 * indexed by RX queue number. Assigned by driver.
1590 * This must only be set if the ndo_rx_flow_steer
1591 * operation is defined
1592 * @index_hlist: Device index hash chain
1593 *
536721b1
KK
1594 * @_tx: Array of TX queues
1595 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1596 * @real_num_tx_queues: Number of TX queues currently active in device
1597 * @qdisc: Root qdisc from userspace point of view
1598 * @tx_queue_len: Max frames per queue allowed
1599 * @tx_global_lock: XXX: need comments on this one
1600 *
1601 * @xps_maps: XXX: need comments on this one
46209401
JP
1602 * @miniq_egress: clsact qdisc specific data for
1603 * egress processing
536721b1 1604 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1605 * the watchdog (see dev_watchdog())
536721b1
KK
1606 * @watchdog_timer: List of timers
1607 *
1608 * @pcpu_refcnt: Number of references to this device
1609 * @todo_list: Delayed register/unregister
536721b1
KK
1610 * @link_watch_list: XXX: need comments on this one
1611 *
1612 * @reg_state: Register/unregister state machine
1613 * @dismantle: Device is going to be freed
1614 * @rtnl_link_state: This enum represents the phases of creating
1615 * a new link
1616 *
cf124db5
DM
1617 * @needs_free_netdev: Should unregister perform free_netdev?
1618 * @priv_destructor: Called from unregister
536721b1
KK
1619 * @npinfo: XXX: need comments on this one
1620 * @nd_net: Network namespace this network device is inside
1621 *
1622 * @ml_priv: Mid-layer private
1623 * @lstats: Loopback statistics
1624 * @tstats: Tunnel statistics
1625 * @dstats: Dummy statistics
1626 * @vstats: Virtual ethernet statistics
1627 *
1628 * @garp_port: GARP
1629 * @mrp_port: MRP
1630 *
1631 * @dev: Class/net/name entry
1632 * @sysfs_groups: Space for optional device, statistics and wireless
1633 * sysfs groups
1634 *
1635 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1636 * @rtnl_link_ops: Rtnl_link_ops
1637 *
1638 * @gso_max_size: Maximum size of generic segmentation offload
1639 * @gso_max_segs: Maximum number of segments that can be passed to the
1640 * NIC for GSO
1641 *
1642 * @dcbnl_ops: Data Center Bridging netlink ops
1643 * @num_tc: Number of traffic classes in the net device
1644 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1645 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1646 *
1647 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1648 *
1649 * @priomap: XXX: need comments on this one
1650 * @phydev: Physical device may attach itself
1651 * for hardware timestamping
1652 *
123b3652
ED
1653 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1654 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1655 *
d746d707
AK
1656 * @proto_down: protocol port state information can be sent to the
1657 * switch driver and used to set the phys state of the
1658 * switch port.
1659 *
1da177e4
LT
1660 * FIXME: cleanup struct net_device such that network protocol info
1661 * moves out.
1662 */
1663
d94d9fee 1664struct net_device {
1da177e4 1665 char name[IFNAMSIZ];
9356b8fc 1666 struct hlist_node name_hlist;
6c557001 1667 struct dev_ifalias __rcu *ifalias;
1da177e4
LT
1668 /*
1669 * I/O specific fields
1670 * FIXME: Merge these and struct ifmap into one
1671 */
536721b1
KK
1672 unsigned long mem_end;
1673 unsigned long mem_start;
1674 unsigned long base_addr;
1675 int irq;
1da177e4
LT
1676
1677 /*
536721b1
KK
1678 * Some hardware also needs these fields (state,dev_list,
1679 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1680 * part of the usual set specified in Space.c.
1681 */
1682
1da177e4
LT
1683 unsigned long state;
1684
7562f876 1685 struct list_head dev_list;
bea3348e 1686 struct list_head napi_list;
44a0873d 1687 struct list_head unreg_list;
5cde2829 1688 struct list_head close_list;
7866a621
SN
1689 struct list_head ptype_all;
1690 struct list_head ptype_specific;
2f268f12 1691
2f268f12
VF
1692 struct {
1693 struct list_head upper;
1694 struct list_head lower;
1695 } adj_list;
1696
c8f44aff 1697 netdev_features_t features;
c8f44aff 1698 netdev_features_t hw_features;
c8f44aff 1699 netdev_features_t wanted_features;
c8f44aff 1700 netdev_features_t vlan_features;
6a674e9c 1701 netdev_features_t hw_enc_features;
0d89d203 1702 netdev_features_t mpls_features;
802ab55a 1703 netdev_features_t gso_partial_features;
04ed3e74 1704
1da177e4 1705 int ifindex;
7a66bbc9 1706 int group;
1da177e4 1707
c45d286e 1708 struct net_device_stats stats;
015f0688 1709
015f0688
ED
1710 atomic_long_t rx_dropped;
1711 atomic_long_t tx_dropped;
6e7333d3 1712 atomic_long_t rx_nohandler;
1da177e4 1713
5e7f6597
DD
1714 /* Stats to monitor link on/off, flapping */
1715 atomic_t carrier_up_count;
1716 atomic_t carrier_down_count;
1717
b86e0280 1718#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1719 const struct iw_handler_def *wireless_handlers;
1720 struct iw_public_data *wireless_data;
b86e0280 1721#endif
d314774c 1722 const struct net_device_ops *netdev_ops;
76fd8593 1723 const struct ethtool_ops *ethtool_ops;
4170604f 1724#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1725 const struct switchdev_ops *switchdev_ops;
4170604f 1726#endif
1b69c6d0
DA
1727#ifdef CONFIG_NET_L3_MASTER_DEV
1728 const struct l3mdev_ops *l3mdev_ops;
1729#endif
f997c55c
AA
1730#if IS_ENABLED(CONFIG_IPV6)
1731 const struct ndisc_ops *ndisc_ops;
1732#endif
1da177e4 1733
d77e38e6
SK
1734#ifdef CONFIG_XFRM
1735 const struct xfrmdev_ops *xfrmdev_ops;
1736#endif
1737
3b04ddde
SH
1738 const struct header_ops *header_ops;
1739
536721b1
KK
1740 unsigned int flags;
1741 unsigned int priv_flags;
1742
1da177e4 1743 unsigned short gflags;
536721b1 1744 unsigned short padded;
1da177e4 1745
536721b1
KK
1746 unsigned char operstate;
1747 unsigned char link_mode;
b00055aa 1748
536721b1
KK
1749 unsigned char if_port;
1750 unsigned char dma;
bdc220da 1751
df0f598f
ED
1752 /* Note : dev->mtu is often read without holding a lock.
1753 * Writers usually hold RTNL.
1754 * It is recommended to use READ_ONCE() to annotate the reads,
1755 * and to use WRITE_ONCE() to annotate the writes.
1756 */
536721b1 1757 unsigned int mtu;
61e84623
JW
1758 unsigned int min_mtu;
1759 unsigned int max_mtu;
536721b1
KK
1760 unsigned short type;
1761 unsigned short hard_header_len;
d92be7a4 1762 unsigned char min_header_len;
1da177e4 1763
f5184d26
JB
1764 unsigned short needed_headroom;
1765 unsigned short needed_tailroom;
1766
1da177e4 1767 /* Interface address info. */
536721b1
KK
1768 unsigned char perm_addr[MAX_ADDR_LEN];
1769 unsigned char addr_assign_type;
1770 unsigned char addr_len;
b634e72d
TY
1771 unsigned char upper_level;
1772 unsigned char lower_level;
a0a9663d 1773 unsigned short neigh_priv_len;
536721b1
KK
1774 unsigned short dev_id;
1775 unsigned short dev_port;
ccffad25 1776 spinlock_t addr_list_lock;
14ffbbb8
TG
1777 unsigned char name_assign_type;
1778 bool uc_promisc;
536721b1
KK
1779 struct netdev_hw_addr_list uc;
1780 struct netdev_hw_addr_list mc;
1781 struct netdev_hw_addr_list dev_addrs;
1782
4c3d5e7b
ED
1783#ifdef CONFIG_SYSFS
1784 struct kset *queues_kset;
1785#endif
9d45abe1
WC
1786 unsigned int promiscuity;
1787 unsigned int allmulti;
1da177e4 1788
1da177e4 1789
5e82b4b2 1790 /* Protocol-specific pointers */
65ac6a5f 1791
d11ead75 1792#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1793 struct vlan_info __rcu *vlan_info;
65ac6a5f 1794#endif
34a430d7 1795#if IS_ENABLED(CONFIG_NET_DSA)
2f657a60 1796 struct dsa_port *dsa_ptr;
37cb0620
YX
1797#endif
1798#if IS_ENABLED(CONFIG_TIPC)
536721b1 1799 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1800#endif
536721b1
KK
1801 void *atalk_ptr;
1802 struct in_device __rcu *ip_ptr;
1803 struct dn_dev __rcu *dn_ptr;
1804 struct inet6_dev __rcu *ip6_ptr;
1805 void *ax25_ptr;
1806 struct wireless_dev *ieee80211_ptr;
98a18b6f 1807 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1808#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1809 struct mpls_dev __rcu *mpls_ptr;
1810#endif
1da177e4 1811
9356b8fc 1812/*
cd13539b 1813 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1814 */
9356b8fc 1815 /* Interface address info used in eth_type_trans() */
536721b1 1816 unsigned char *dev_addr;
f001fde5 1817
a953be53 1818#ifdef CONFIG_SYSFS
0a9627f2
TH
1819 struct netdev_rx_queue *_rx;
1820
0a9627f2 1821 unsigned int num_rx_queues;
62fe0b40 1822 unsigned int real_num_rx_queues;
df334545 1823#endif
0a9627f2 1824
7acedaf5 1825 struct bpf_prog __rcu *xdp_prog;
3b47d303 1826 unsigned long gro_flush_timeout;
61391cde 1827 rx_handler_func_t __rcu *rx_handler;
1828 void __rcu *rx_handler_data;
e8a0464c 1829
4cda01e8 1830#ifdef CONFIG_NET_CLS_ACT
46209401 1831 struct mini_Qdisc __rcu *miniq_ingress;
d2788d34 1832#endif
24824a09 1833 struct netdev_queue __rcu *ingress_queue;
e687ad60 1834#ifdef CONFIG_NETFILTER_INGRESS
960632ec 1835 struct nf_hook_entries __rcu *nf_hooks_ingress;
e687ad60 1836#endif
d2788d34 1837
536721b1 1838 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1839#ifdef CONFIG_RFS_ACCEL
1840 struct cpu_rmap *rx_cpu_rmap;
1841#endif
1842 struct hlist_node index_hlist;
cd13539b
ED
1843
1844/*
1845 * Cache lines mostly used on transmit path
1846 */
e8a0464c
DM
1847 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1848 unsigned int num_tx_queues;
fd2ea0a7 1849 unsigned int real_num_tx_queues;
af356afa 1850 struct Qdisc *qdisc;
59cc1f61
JK
1851#ifdef CONFIG_NET_SCHED
1852 DECLARE_HASHTABLE (qdisc_hash, 4);
1853#endif
0cd29503 1854 unsigned int tx_queue_len;
c3f26a26 1855 spinlock_t tx_global_lock;
14ffbbb8 1856 int watchdog_timeo;
cd13539b 1857
bf264145 1858#ifdef CONFIG_XPS
a4177869 1859 struct xps_dev_maps __rcu *xps_maps;
bf264145 1860#endif
1f211a1b 1861#ifdef CONFIG_NET_CLS_ACT
46209401 1862 struct mini_Qdisc __rcu *miniq_egress;
1f211a1b 1863#endif
0c4f691f 1864
9356b8fc 1865 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1866 struct timer_list watchdog_timer;
1867
29b4433d 1868 int __percpu *pcpu_refcnt;
1da177e4 1869 struct list_head todo_list;
1da177e4 1870
e014debe 1871 struct list_head link_watch_list;
572a103d 1872
1da177e4 1873 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1874 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1875 NETREG_UNREGISTERING, /* called unregister_netdevice */
1876 NETREG_UNREGISTERED, /* completed unregister todo */
1877 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1878 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1879 } reg_state:8;
1880
536721b1 1881 bool dismantle;
a2835763
PM
1882
1883 enum {
1884 RTNL_LINK_INITIALIZED,
1885 RTNL_LINK_INITIALIZING,
1886 } rtnl_link_state:16;
1da177e4 1887
cf124db5
DM
1888 bool needs_free_netdev;
1889 void (*priv_destructor)(struct net_device *dev);
1da177e4 1890
1da177e4 1891#ifdef CONFIG_NETPOLL
5fbee843 1892 struct netpoll_info __rcu *npinfo;
1da177e4 1893#endif
eae792b7 1894
0c5c9fb5 1895 possible_net_t nd_net;
4a1c5371 1896
4951704b 1897 /* mid-layer private */
a7855c78 1898 union {
536721b1
KK
1899 void *ml_priv;
1900 struct pcpu_lstats __percpu *lstats;
8f84985f 1901 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1902 struct pcpu_dstats __percpu *dstats;
1903 struct pcpu_vstats __percpu *vstats;
a7855c78 1904 };
536721b1 1905
fb585b44 1906#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1907 struct garp_port __rcu *garp_port;
fb585b44
TK
1908#endif
1909#if IS_ENABLED(CONFIG_MRP)
febf018d 1910 struct mrp_port __rcu *mrp_port;
fb585b44 1911#endif
1da177e4 1912
5e82b4b2 1913 struct device dev;
0c509a6c 1914 const struct attribute_group *sysfs_groups[4];
a953be53 1915 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1916
38f7b870 1917 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1918
82cc1a7a
PWJ
1919 /* for setting kernel sock attribute on TCP connection setup */
1920#define GSO_MAX_SIZE 65536
1921 unsigned int gso_max_size;
30b678d8
BH
1922#define GSO_MAX_SEGS 65535
1923 u16 gso_max_segs;
743b03a8 1924
7a6b6f51 1925#ifdef CONFIG_DCB
32953543 1926 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1927#endif
5e82b4b2
BH
1928 u8 num_tc;
1929 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1930 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1931
d11ead75 1932#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1933 unsigned int fcoe_ddp_xid;
5bc1421e 1934#endif
86f8515f 1935#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1936 struct netprio_map __rcu *priomap;
4d288d57 1937#endif
5e82b4b2
BH
1938 struct phy_device *phydev;
1939 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1940 struct lock_class_key *qdisc_running_key;
5e82b4b2 1941 bool proto_down;
1da177e4 1942};
43cb76d9 1943#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1944
b5cdae32
DM
1945static inline bool netif_elide_gro(const struct net_device *dev)
1946{
1947 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1948 return true;
1949 return false;
1950}
1951
1da177e4 1952#define NETDEV_ALIGN 32
1da177e4 1953
4f57c087
JF
1954static inline
1955int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1956{
1957 return dev->prio_tc_map[prio & TC_BITMASK];
1958}
1959
1960static inline
1961int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1962{
1963 if (tc >= dev->num_tc)
1964 return -EINVAL;
1965
1966 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1967 return 0;
1968}
1969
8d059b0f 1970int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1971void netdev_reset_tc(struct net_device *dev);
1972int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1973int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1974
1975static inline
1976int netdev_get_num_tc(struct net_device *dev)
1977{
1978 return dev->num_tc;
1979}
1980
e8a0464c
DM
1981static inline
1982struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1983 unsigned int index)
1984{
1985 return &dev->_tx[index];
1986}
1987
10c51b56
DB
1988static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1989 const struct sk_buff *skb)
1990{
1991 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1992}
1993
e8a0464c
DM
1994static inline void netdev_for_each_tx_queue(struct net_device *dev,
1995 void (*f)(struct net_device *,
1996 struct netdev_queue *,
1997 void *),
1998 void *arg)
1999{
2000 unsigned int i;
2001
2002 for (i = 0; i < dev->num_tx_queues; i++)
2003 f(dev, &dev->_tx[i], arg);
2004}
2005
d3fff6c4
ED
2006#define netdev_lockdep_set_classes(dev) \
2007{ \
2008 static struct lock_class_key qdisc_tx_busylock_key; \
2009 static struct lock_class_key qdisc_running_key; \
2010 static struct lock_class_key qdisc_xmit_lock_key; \
2011 static struct lock_class_key dev_addr_list_lock_key; \
2012 unsigned int i; \
2013 \
2014 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2015 (dev)->qdisc_running_key = &qdisc_running_key; \
2016 lockdep_set_class(&(dev)->addr_list_lock, \
2017 &dev_addr_list_lock_key); \
2018 for (i = 0; i < (dev)->num_tx_queues; i++) \
2019 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2020 &qdisc_xmit_lock_key); \
2021}
2022
f629d208 2023struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
2024 struct sk_buff *skb,
2025 void *accel_priv);
8c4c49df 2026
871b642a
PA
2027/* returns the headroom that the master device needs to take in account
2028 * when forwarding to this dev
2029 */
2030static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2031{
2032 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2033}
2034
2035static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2036{
2037 if (dev->netdev_ops->ndo_set_rx_headroom)
2038 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2039}
2040
2041/* set the device rx headroom to the dev's default */
2042static inline void netdev_reset_rx_headroom(struct net_device *dev)
2043{
2044 netdev_set_rx_headroom(dev, -1);
2045}
2046
c346dca1
YH
2047/*
2048 * Net namespace inlines
2049 */
2050static inline
2051struct net *dev_net(const struct net_device *dev)
2052{
c2d9ba9b 2053 return read_pnet(&dev->nd_net);
c346dca1
YH
2054}
2055
2056static inline
f5aa23fd 2057void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2058{
0c5c9fb5 2059 write_pnet(&dev->nd_net, net);
c346dca1
YH
2060}
2061
bea3348e
SH
2062/**
2063 * netdev_priv - access network device private data
2064 * @dev: network device
2065 *
2066 * Get network device private data
2067 */
6472ce60 2068static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2069{
1ce8e7b5 2070 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2071}
2072
1da177e4
LT
2073/* Set the sysfs physical device reference for the network logical device
2074 * if set prior to registration will cause a symlink during initialization.
2075 */
43cb76d9 2076#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2077
384912ed 2078/* Set the sysfs device type for the network logical device to allow
3f79410c 2079 * fine-grained identification of different network device types. For
5e82b4b2 2080 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2081 */
2082#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2083
82dc3c63
ED
2084/* Default NAPI poll() weight
2085 * Device drivers are strongly advised to not use bigger value
2086 */
2087#define NAPI_POLL_WEIGHT 64
2088
3b582cc1 2089/**
5e82b4b2 2090 * netif_napi_add - initialize a NAPI context
3b582cc1 2091 * @dev: network device
5e82b4b2 2092 * @napi: NAPI context
3b582cc1
SH
2093 * @poll: polling function
2094 * @weight: default weight
2095 *
5e82b4b2
BH
2096 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2097 * *any* of the other NAPI-related functions.
3b582cc1 2098 */
d565b0a1
HX
2099void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2100 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2101
d64b5e85 2102/**
5e82b4b2 2103 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2104 * @dev: network device
5e82b4b2 2105 * @napi: NAPI context
d64b5e85
ED
2106 * @poll: polling function
2107 * @weight: default weight
2108 *
2109 * This variant of netif_napi_add() should be used from drivers using NAPI
2110 * to exclusively poll a TX queue.
2111 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2112 */
2113static inline void netif_tx_napi_add(struct net_device *dev,
2114 struct napi_struct *napi,
2115 int (*poll)(struct napi_struct *, int),
2116 int weight)
2117{
2118 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2119 netif_napi_add(dev, napi, poll, weight);
2120}
2121
d8156534 2122/**
5e82b4b2
BH
2123 * netif_napi_del - remove a NAPI context
2124 * @napi: NAPI context
d8156534 2125 *
5e82b4b2 2126 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2127 */
d565b0a1
HX
2128void netif_napi_del(struct napi_struct *napi);
2129
2130struct napi_gro_cb {
78a478d0 2131 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2132 void *frag0;
78a478d0 2133
7489594c
HX
2134 /* Length of frag0. */
2135 unsigned int frag0_len;
2136
86911732 2137 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2138 int data_offset;
86911732 2139
d565b0a1 2140 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2141 u16 flush;
2142
2143 /* Save the IP ID here and check when we get to the transport layer */
2144 u16 flush_id;
d565b0a1
HX
2145
2146 /* Number of segments aggregated. */
2e71a6f8
ED
2147 u16 count;
2148
15e2396d
TH
2149 /* Start offset for remote checksum offload */
2150 u16 gro_remcsum_start;
2151
2e71a6f8
ED
2152 /* jiffies when first packet was created/queued */
2153 unsigned long age;
86347245 2154
afe93325 2155 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2156 u16 proto;
2157
baa32ff4
TH
2158 /* This is non-zero if the packet may be of the same flow. */
2159 u8 same_flow:1;
2160
fac8e0f5
JG
2161 /* Used in tunnel GRO receive */
2162 u8 encap_mark:1;
573e8fca
TH
2163
2164 /* GRO checksum is valid */
2165 u8 csum_valid:1;
2166
662880f4
TH
2167 /* Number of checksums via CHECKSUM_UNNECESSARY */
2168 u8 csum_cnt:3;
c3c7c254 2169
baa32ff4
TH
2170 /* Free the skb? */
2171 u8 free:2;
2172#define NAPI_GRO_FREE 1
2173#define NAPI_GRO_FREE_STOLEN_HEAD 2
2174
efc98d08
TH
2175 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2176 u8 is_ipv6:1;
2177
a0ca153f
AD
2178 /* Used in GRE, set in fou/gue_gro_receive */
2179 u8 is_fou:1;
2180
1530545e
AD
2181 /* Used to determine if flush_id can be ignored */
2182 u8 is_atomic:1;
2183
fcd91dd4
SD
2184 /* Number of gro_receive callbacks this packet already went through */
2185 u8 recursion_counter:4;
2186
2187 /* 1 bit hole */
baa32ff4 2188
bf5a755f
JC
2189 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2190 __wsum csum;
2191
c3c7c254
ED
2192 /* used in skb_gro_receive() slow path */
2193 struct sk_buff *last;
d565b0a1
HX
2194};
2195
2196#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2197
fcd91dd4
SD
2198#define GRO_RECURSION_LIMIT 15
2199static inline int gro_recursion_inc_test(struct sk_buff *skb)
2200{
2201 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2202}
2203
2204typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2205static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2206 struct sk_buff **head,
2207 struct sk_buff *skb)
2208{
2209 if (unlikely(gro_recursion_inc_test(skb))) {
2210 NAPI_GRO_CB(skb)->flush |= 1;
2211 return NULL;
2212 }
2213
2214 return cb(head, skb);
2215}
2216
2217typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2218 struct sk_buff *);
2219static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2220 struct sock *sk,
2221 struct sk_buff **head,
2222 struct sk_buff *skb)
2223{
2224 if (unlikely(gro_recursion_inc_test(skb))) {
2225 NAPI_GRO_CB(skb)->flush |= 1;
2226 return NULL;
2227 }
2228
2229 return cb(sk, head, skb);
2230}
2231
1da177e4 2232struct packet_type {
f2ccd8fa
DM
2233 __be16 type; /* This is really htons(ether_type). */
2234 struct net_device *dev; /* NULL is wildcarded here */
2235 int (*func) (struct sk_buff *,
2236 struct net_device *,
2237 struct packet_type *,
2238 struct net_device *);
c0de08d0
EL
2239 bool (*id_match)(struct packet_type *ptype,
2240 struct sock *sk);
1da177e4
LT
2241 void *af_packet_priv;
2242 struct list_head list;
2243};
2244
f191a1d1 2245struct offload_callbacks {
576a30eb 2246 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2247 netdev_features_t features);
d565b0a1 2248 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2249 struct sk_buff *skb);
299603e8 2250 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2251};
2252
2253struct packet_offload {
2254 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2255 u16 priority;
f191a1d1
VY
2256 struct offload_callbacks callbacks;
2257 struct list_head list;
1da177e4
LT
2258};
2259
5e82b4b2 2260/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2261struct pcpu_sw_netstats {
2262 u64 rx_packets;
2263 u64 rx_bytes;
2264 u64 tx_packets;
2265 u64 tx_bytes;
2266 struct u64_stats_sync syncp;
2267};
2268
aabc92bb
PNA
2269#define __netdev_alloc_pcpu_stats(type, gfp) \
2270({ \
2271 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2272 if (pcpu_stats) { \
2273 int __cpu; \
2274 for_each_possible_cpu(__cpu) { \
2275 typeof(type) *stat; \
2276 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2277 u64_stats_init(&stat->syncp); \
2278 } \
2279 } \
2280 pcpu_stats; \
1c213bd2
WC
2281})
2282
aabc92bb 2283#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2284 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2285
764f5e54
JP
2286enum netdev_lag_tx_type {
2287 NETDEV_LAG_TX_TYPE_UNKNOWN,
2288 NETDEV_LAG_TX_TYPE_RANDOM,
2289 NETDEV_LAG_TX_TYPE_BROADCAST,
2290 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2291 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2292 NETDEV_LAG_TX_TYPE_HASH,
2293};
2294
2295struct netdev_lag_upper_info {
2296 enum netdev_lag_tx_type tx_type;
2297};
2298
fb1b2e3c
JP
2299struct netdev_lag_lower_state_info {
2300 u8 link_up : 1,
2301 tx_enabled : 1;
2302};
2303
1da177e4
LT
2304#include <linux/notifier.h>
2305
dcfe1421
AW
2306/* netdevice notifier chain. Please remember to update the rtnetlink
2307 * notification exclusion list in rtnetlink_event() when adding new
2308 * types.
2309 */
2310#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2311#define NETDEV_DOWN 0x0002
2312#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2313 detected a hardware crash and restarted
2314 - we can use this eg to kick tcp sessions
2315 once done */
2316#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2317#define NETDEV_REGISTER 0x0005
2318#define NETDEV_UNREGISTER 0x0006
1d486bfb 2319#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2320#define NETDEV_CHANGEADDR 0x0008
2321#define NETDEV_GOING_DOWN 0x0009
2322#define NETDEV_CHANGENAME 0x000A
2323#define NETDEV_FEAT_CHANGE 0x000B
2324#define NETDEV_BONDING_FAILOVER 0x000C
2325#define NETDEV_PRE_UP 0x000D
2326#define NETDEV_PRE_TYPE_CHANGE 0x000E
2327#define NETDEV_POST_TYPE_CHANGE 0x000F
2328#define NETDEV_POST_INIT 0x0010
0115e8e3 2329#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2330#define NETDEV_RELEASE 0x0012
2331#define NETDEV_NOTIFY_PEERS 0x0013
2332#define NETDEV_JOIN 0x0014
42e52bf9 2333#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2334#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2335#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2336#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2337#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2338#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2339#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2340#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
296d8ee3 2341#define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
08294a26 2342#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2343
f629d208
JP
2344int register_netdevice_notifier(struct notifier_block *nb);
2345int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2346
2347struct netdev_notifier_info {
51d0c047
DA
2348 struct net_device *dev;
2349 struct netlink_ext_ack *extack;
351638e7
JP
2350};
2351
63419c91
SD
2352struct netdev_notifier_info_ext {
2353 struct netdev_notifier_info info; /* must be first */
2354 union {
2355 u32 mtu;
2356 } ext;
2357};
2358
be9efd36
JP
2359struct netdev_notifier_change_info {
2360 struct netdev_notifier_info info; /* must be first */
2361 unsigned int flags_changed;
2362};
2363
0e4ead9d
JP
2364struct netdev_notifier_changeupper_info {
2365 struct netdev_notifier_info info; /* must be first */
2366 struct net_device *upper_dev; /* new upper dev */
2367 bool master; /* is upper dev master */
5e82b4b2 2368 bool linking; /* is the notification for link or unlink */
29bf24af 2369 void *upper_info; /* upper dev info */
0e4ead9d
JP
2370};
2371
04d48266
JP
2372struct netdev_notifier_changelowerstate_info {
2373 struct netdev_notifier_info info; /* must be first */
2374 void *lower_state_info; /* is lower dev state */
2375};
2376
75538c2b
CW
2377static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2378 struct net_device *dev)
2379{
2380 info->dev = dev;
51d0c047 2381 info->extack = NULL;
75538c2b
CW
2382}
2383
351638e7
JP
2384static inline struct net_device *
2385netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2386{
2387 return info->dev;
2388}
2389
51d0c047
DA
2390static inline struct netlink_ext_ack *
2391netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2392{
2393 return info->extack;
2394}
2395
f629d208 2396int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2397
2398
1da177e4
LT
2399extern rwlock_t dev_base_lock; /* Device list lock */
2400
881d966b
EB
2401#define for_each_netdev(net, d) \
2402 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2403#define for_each_netdev_reverse(net, d) \
2404 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2405#define for_each_netdev_rcu(net, d) \
2406 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2407#define for_each_netdev_safe(net, d, n) \
2408 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2409#define for_each_netdev_continue(net, d) \
2410 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2411#define for_each_netdev_continue_rcu(net, d) \
2412 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2413#define for_each_netdev_in_bond_rcu(bond, slave) \
2414 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2415 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2416#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2417
a050c33f
DL
2418static inline struct net_device *next_net_device(struct net_device *dev)
2419{
2420 struct list_head *lh;
2421 struct net *net;
2422
c346dca1 2423 net = dev_net(dev);
a050c33f
DL
2424 lh = dev->dev_list.next;
2425 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2426}
2427
ce81b76a
ED
2428static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2429{
2430 struct list_head *lh;
2431 struct net *net;
2432
2433 net = dev_net(dev);
ccf43438 2434 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2435 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2436}
2437
a050c33f
DL
2438static inline struct net_device *first_net_device(struct net *net)
2439{
2440 return list_empty(&net->dev_base_head) ? NULL :
2441 net_device_entry(net->dev_base_head.next);
2442}
7562f876 2443
ccf43438
ED
2444static inline struct net_device *first_net_device_rcu(struct net *net)
2445{
2446 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2447
2448 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2449}
2450
f629d208
JP
2451int netdev_boot_setup_check(struct net_device *dev);
2452unsigned long netdev_boot_base(const char *prefix, int unit);
2453struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2454 const char *hwaddr);
2455struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2456struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2457void dev_add_pack(struct packet_type *pt);
2458void dev_remove_pack(struct packet_type *pt);
2459void __dev_remove_pack(struct packet_type *pt);
2460void dev_add_offload(struct packet_offload *po);
2461void dev_remove_offload(struct packet_offload *po);
f629d208 2462
a54acb3a 2463int dev_get_iflink(const struct net_device *dev);
fc4099f1 2464int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2465struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2466 unsigned short mask);
f629d208
JP
2467struct net_device *dev_get_by_name(struct net *net, const char *name);
2468struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2469struct net_device *__dev_get_by_name(struct net *net, const char *name);
2470int dev_alloc_name(struct net_device *dev, const char *name);
2471int dev_open(struct net_device *dev);
7051b88a 2472void dev_close(struct net_device *dev);
2473void dev_close_many(struct list_head *head, bool unlink);
f629d208 2474void dev_disable_lro(struct net_device *dev);
0c4b51f0 2475int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2476int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2477int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2478int register_netdevice(struct net_device *dev);
2479void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2480void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2481static inline void unregister_netdevice(struct net_device *dev)
2482{
2483 unregister_netdevice_queue(dev, NULL);
2484}
2485
f629d208
JP
2486int netdev_refcnt_read(const struct net_device *dev);
2487void free_netdev(struct net_device *dev);
74d332c1 2488void netdev_freemem(struct net_device *dev);
f629d208
JP
2489void synchronize_net(void);
2490int init_dummy_netdev(struct net_device *dev);
937f1ba5 2491
f60e5990 2492DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2493#define XMIT_RECURSION_LIMIT 10
2494
f60e5990 2495static inline int dev_recursion_level(void)
2496{
2497 return this_cpu_read(xmit_recursion);
2498}
2499
f629d208
JP
2500struct net_device *dev_get_by_index(struct net *net, int ifindex);
2501struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2502struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2503struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2504int netdev_get_name(struct net *net, char *name, int ifindex);
2505int dev_restart(struct net_device *dev);
f629d208 2506int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2507
2508static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2509{
2510 return NAPI_GRO_CB(skb)->data_offset;
2511}
2512
2513static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2514{
2515 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2516}
2517
2518static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2519{
2520 NAPI_GRO_CB(skb)->data_offset += len;
2521}
2522
a5b1cf28
HX
2523static inline void *skb_gro_header_fast(struct sk_buff *skb,
2524 unsigned int offset)
86911732 2525{
a5b1cf28
HX
2526 return NAPI_GRO_CB(skb)->frag0 + offset;
2527}
78a478d0 2528
a5b1cf28
HX
2529static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2530{
2531 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2532}
78a478d0 2533
57ea52a8
HX
2534static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2535{
2536 NAPI_GRO_CB(skb)->frag0 = NULL;
2537 NAPI_GRO_CB(skb)->frag0_len = 0;
2538}
2539
a5b1cf28
HX
2540static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2541 unsigned int offset)
2542{
17dd759c
HX
2543 if (!pskb_may_pull(skb, hlen))
2544 return NULL;
2545
57ea52a8 2546 skb_gro_frag0_invalidate(skb);
17dd759c 2547 return skb->data + offset;
86911732 2548}
1da177e4 2549
36e7b1b8
HX
2550static inline void *skb_gro_network_header(struct sk_buff *skb)
2551{
78d3fd0b
HX
2552 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2553 skb_network_offset(skb);
36e7b1b8
HX
2554}
2555
bf5a755f
JC
2556static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2557 const void *start, unsigned int len)
2558{
573e8fca 2559 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2560 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2561 csum_partial(start, len, 0));
2562}
2563
573e8fca
TH
2564/* GRO checksum functions. These are logical equivalents of the normal
2565 * checksum functions (in skbuff.h) except that they operate on the GRO
2566 * offsets and fields in sk_buff.
2567 */
2568
2569__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2570
15e2396d
TH
2571static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2572{
b7fe10e5 2573 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2574}
2575
573e8fca
TH
2576static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2577 bool zero_okay,
2578 __sum16 check)
2579{
6edec0e6
TH
2580 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2581 skb_checksum_start_offset(skb) <
2582 skb_gro_offset(skb)) &&
15e2396d 2583 !skb_at_gro_remcsum_start(skb) &&
662880f4 2584 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2585 (!zero_okay || check));
2586}
2587
2588static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2589 __wsum psum)
2590{
2591 if (NAPI_GRO_CB(skb)->csum_valid &&
2592 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2593 return 0;
2594
2595 NAPI_GRO_CB(skb)->csum = psum;
2596
2597 return __skb_gro_checksum_complete(skb);
2598}
2599
573e8fca
TH
2600static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2601{
662880f4
TH
2602 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2603 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2604 NAPI_GRO_CB(skb)->csum_cnt--;
2605 } else {
2606 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2607 * verified a new top level checksum or an encapsulated one
2608 * during GRO. This saves work if we fallback to normal path.
2609 */
2610 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2611 }
2612}
2613
2614#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2615 compute_pseudo) \
2616({ \
2617 __sum16 __ret = 0; \
2618 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2619 __ret = __skb_gro_checksum_validate_complete(skb, \
2620 compute_pseudo(skb, proto)); \
219f1d79 2621 if (!__ret) \
573e8fca
TH
2622 skb_gro_incr_csum_unnecessary(skb); \
2623 __ret; \
2624})
2625
2626#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2627 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2628
2629#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2630 compute_pseudo) \
2631 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2632
2633#define skb_gro_checksum_simple_validate(skb) \
2634 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2635
d96535a1
TH
2636static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2637{
2638 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2639 !NAPI_GRO_CB(skb)->csum_valid);
2640}
2641
2642static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2643 __sum16 check, __wsum pseudo)
2644{
2645 NAPI_GRO_CB(skb)->csum = ~pseudo;
2646 NAPI_GRO_CB(skb)->csum_valid = 1;
2647}
2648
2649#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2650do { \
2651 if (__skb_gro_checksum_convert_check(skb)) \
2652 __skb_gro_checksum_convert(skb, check, \
2653 compute_pseudo(skb, proto)); \
2654} while (0)
2655
26c4f7da
TH
2656struct gro_remcsum {
2657 int offset;
2658 __wsum delta;
2659};
2660
2661static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2662{
846cd667 2663 grc->offset = 0;
26c4f7da
TH
2664 grc->delta = 0;
2665}
2666
b7fe10e5
TH
2667static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2668 unsigned int off, size_t hdrlen,
2669 int start, int offset,
2670 struct gro_remcsum *grc,
2671 bool nopartial)
dcdc8994
TH
2672{
2673 __wsum delta;
b7fe10e5 2674 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2675
2676 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2677
15e2396d 2678 if (!nopartial) {
b7fe10e5
TH
2679 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2680 return ptr;
2681 }
2682
2683 ptr = skb_gro_header_fast(skb, off);
2684 if (skb_gro_header_hard(skb, off + plen)) {
2685 ptr = skb_gro_header_slow(skb, off + plen, off);
2686 if (!ptr)
2687 return NULL;
15e2396d
TH
2688 }
2689
b7fe10e5
TH
2690 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2691 start, offset);
dcdc8994
TH
2692
2693 /* Adjust skb->csum since we changed the packet */
dcdc8994 2694 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2695
b7fe10e5 2696 grc->offset = off + hdrlen + offset;
26c4f7da 2697 grc->delta = delta;
b7fe10e5
TH
2698
2699 return ptr;
dcdc8994
TH
2700}
2701
26c4f7da
TH
2702static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2703 struct gro_remcsum *grc)
2704{
b7fe10e5
TH
2705 void *ptr;
2706 size_t plen = grc->offset + sizeof(u16);
2707
26c4f7da
TH
2708 if (!grc->delta)
2709 return;
2710
b7fe10e5
TH
2711 ptr = skb_gro_header_fast(skb, grc->offset);
2712 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2713 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2714 if (!ptr)
2715 return;
2716 }
2717
2718 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2719}
dcdc8994 2720
25393d3f
SK
2721#ifdef CONFIG_XFRM_OFFLOAD
2722static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2723{
2724 if (PTR_ERR(pp) != -EINPROGRESS)
2725 NAPI_GRO_CB(skb)->flush |= flush;
2726}
12df8689
SD
2727static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2728 struct sk_buff **pp,
2729 int flush,
2730 struct gro_remcsum *grc)
2731{
2732 if (PTR_ERR(pp) != -EINPROGRESS) {
2733 NAPI_GRO_CB(skb)->flush |= flush;
2734 skb_gro_remcsum_cleanup(skb, grc);
2735 skb->remcsum_offload = 0;
2736 }
2737}
25393d3f 2738#else
5f114163
SK
2739static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2740{
2741 NAPI_GRO_CB(skb)->flush |= flush;
2742}
12df8689
SD
2743static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2744 struct sk_buff **pp,
2745 int flush,
2746 struct gro_remcsum *grc)
2747{
2748 NAPI_GRO_CB(skb)->flush |= flush;
2749 skb_gro_remcsum_cleanup(skb, grc);
2750 skb->remcsum_offload = 0;
2751}
25393d3f 2752#endif
5f114163 2753
0c4e8581
SH
2754static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2755 unsigned short type,
3b04ddde 2756 const void *daddr, const void *saddr,
95c96174 2757 unsigned int len)
0c4e8581 2758{
f1ecfd5d 2759 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2760 return 0;
3b04ddde
SH
2761
2762 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2763}
2764
b95cce35
SH
2765static inline int dev_parse_header(const struct sk_buff *skb,
2766 unsigned char *haddr)
2767{
2768 const struct net_device *dev = skb->dev;
2769
1b83336b 2770 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2771 return 0;
3b04ddde 2772 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2773}
2774
2793a23a
WB
2775/* ll_header must have at least hard_header_len allocated */
2776static inline bool dev_validate_header(const struct net_device *dev,
2777 char *ll_header, int len)
2778{
2779 if (likely(len >= dev->hard_header_len))
2780 return true;
217e6fa2
WB
2781 if (len < dev->min_header_len)
2782 return false;
2793a23a
WB
2783
2784 if (capable(CAP_SYS_RAWIO)) {
2785 memset(ll_header + len, 0, dev->hard_header_len - len);
2786 return true;
2787 }
2788
2789 if (dev->header_ops && dev->header_ops->validate)
2790 return dev->header_ops->validate(ll_header, len);
2791
2792 return false;
2793}
2794
1da177e4 2795typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2796int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2797static inline int unregister_gifconf(unsigned int family)
2798{
2799 return register_gifconf(family, NULL);
2800}
2801
99bbc707 2802#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2803#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2804struct sd_flow_limit {
2805 u64 count;
2806 unsigned int num_buckets;
2807 unsigned int history_head;
2808 u16 history[FLOW_LIMIT_HISTORY];
2809 u8 buckets[];
2810};
2811
2812extern int netdev_flow_limit_table_len;
2813#endif /* CONFIG_NET_FLOW_LIMIT */
2814
1da177e4 2815/*
5e82b4b2 2816 * Incoming packets are placed on per-CPU queues
1da177e4 2817 */
d94d9fee 2818struct softnet_data {
1da177e4 2819 struct list_head poll_list;
6e7676c1 2820 struct sk_buff_head process_queue;
1da177e4 2821
dee42870 2822 /* stats */
cd7b5396
DM
2823 unsigned int processed;
2824 unsigned int time_squeeze;
cd7b5396 2825 unsigned int received_rps;
fd793d89 2826#ifdef CONFIG_RPS
88751275 2827 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2828#endif
2829#ifdef CONFIG_NET_FLOW_LIMIT
2830 struct sd_flow_limit __rcu *flow_limit;
2831#endif
2832 struct Qdisc *output_queue;
2833 struct Qdisc **output_queue_tailp;
2834 struct sk_buff *completion_queue;
88751275 2835
4cdb1e2e 2836#ifdef CONFIG_RPS
501e7ef5
ED
2837 /* input_queue_head should be written by cpu owning this struct,
2838 * and only read by other cpus. Worth using a cache line.
2839 */
2840 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2841
2842 /* Elements below can be accessed between CPUs for RPS/RFS */
966a9671 2843 call_single_data_t csd ____cacheline_aligned_in_smp;
88751275
ED
2844 struct softnet_data *rps_ipi_next;
2845 unsigned int cpu;
76cc8b13 2846 unsigned int input_queue_tail;
1e94d72f 2847#endif
95c96174 2848 unsigned int dropped;
0a9627f2 2849 struct sk_buff_head input_pkt_queue;
bea3348e 2850 struct napi_struct backlog;
99bbc707 2851
1da177e4
LT
2852};
2853
76cc8b13 2854static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2855{
2856#ifdef CONFIG_RPS
76cc8b13
TH
2857 sd->input_queue_head++;
2858#endif
2859}
2860
2861static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2862 unsigned int *qtail)
2863{
2864#ifdef CONFIG_RPS
2865 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2866#endif
2867}
2868
0a9627f2 2869DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2870
f629d208 2871void __netif_schedule(struct Qdisc *q);
46e5da40 2872void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2873
fd2ea0a7
DM
2874static inline void netif_tx_schedule_all(struct net_device *dev)
2875{
2876 unsigned int i;
2877
2878 for (i = 0; i < dev->num_tx_queues; i++)
2879 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2880}
2881
f9a7cbbf 2882static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2883{
73466498 2884 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2885}
2886
bea3348e
SH
2887/**
2888 * netif_start_queue - allow transmit
2889 * @dev: network device
2890 *
2891 * Allow upper layers to call the device hard_start_xmit routine.
2892 */
1da177e4
LT
2893static inline void netif_start_queue(struct net_device *dev)
2894{
e8a0464c 2895 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2896}
2897
fd2ea0a7
DM
2898static inline void netif_tx_start_all_queues(struct net_device *dev)
2899{
2900 unsigned int i;
2901
2902 for (i = 0; i < dev->num_tx_queues; i++) {
2903 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2904 netif_tx_start_queue(txq);
2905 }
2906}
2907
46e5da40 2908void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2909
d29f749e
DJ
2910/**
2911 * netif_wake_queue - restart transmit
2912 * @dev: network device
2913 *
2914 * Allow upper layers to call the device hard_start_xmit routine.
2915 * Used for flow control when transmit resources are available.
2916 */
79d16385
DM
2917static inline void netif_wake_queue(struct net_device *dev)
2918{
e8a0464c 2919 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2920}
2921
fd2ea0a7
DM
2922static inline void netif_tx_wake_all_queues(struct net_device *dev)
2923{
2924 unsigned int i;
2925
2926 for (i = 0; i < dev->num_tx_queues; i++) {
2927 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2928 netif_tx_wake_queue(txq);
2929 }
2930}
2931
f9a7cbbf 2932static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2933{
73466498 2934 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2935}
2936
bea3348e
SH
2937/**
2938 * netif_stop_queue - stop transmitted packets
2939 * @dev: network device
2940 *
2941 * Stop upper layers calling the device hard_start_xmit routine.
2942 * Used for flow control when transmit resources are unavailable.
2943 */
1da177e4
LT
2944static inline void netif_stop_queue(struct net_device *dev)
2945{
e8a0464c 2946 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2947}
2948
a2029240 2949void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2950
4d29515f 2951static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2952{
73466498 2953 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2954}
2955
bea3348e
SH
2956/**
2957 * netif_queue_stopped - test if transmit queue is flowblocked
2958 * @dev: network device
2959 *
2960 * Test if transmit queue on device is currently unable to send.
2961 */
4d29515f 2962static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2963{
e8a0464c 2964 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2965}
2966
4d29515f 2967static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2968{
73466498
TH
2969 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2970}
2971
8e2f1a63
DB
2972static inline bool
2973netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2974{
2975 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2976}
2977
8e2f1a63
DB
2978static inline bool
2979netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2980{
2981 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2982}
2983
53511453
ED
2984/**
2985 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2986 * @dev_queue: pointer to transmit queue
2987 *
2988 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2989 * to give appropriate hint to the CPU.
53511453
ED
2990 */
2991static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2992{
2993#ifdef CONFIG_BQL
2994 prefetchw(&dev_queue->dql.num_queued);
2995#endif
2996}
2997
2998/**
2999 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3000 * @dev_queue: pointer to transmit queue
3001 *
3002 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 3003 * to give appropriate hint to the CPU.
53511453
ED
3004 */
3005static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3006{
3007#ifdef CONFIG_BQL
3008 prefetchw(&dev_queue->dql.limit);
3009#endif
3010}
3011
c5d67bd7
TH
3012static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3013 unsigned int bytes)
3014{
114cf580
TH
3015#ifdef CONFIG_BQL
3016 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
3017
3018 if (likely(dql_avail(&dev_queue->dql) >= 0))
3019 return;
3020
3021 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3022
3023 /*
3024 * The XOFF flag must be set before checking the dql_avail below,
3025 * because in netdev_tx_completed_queue we update the dql_completed
3026 * before checking the XOFF flag.
3027 */
3028 smp_mb();
3029
3030 /* check again in case another CPU has just made room avail */
3031 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3032 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 3033#endif
c5d67bd7
TH
3034}
3035
0042d0c8
FF
3036/**
3037 * netdev_sent_queue - report the number of bytes queued to hardware
3038 * @dev: network device
3039 * @bytes: number of bytes queued to the hardware device queue
3040 *
3041 * Report the number of bytes queued for sending/completion to the network
3042 * device hardware queue. @bytes should be a good approximation and should
3043 * exactly match netdev_completed_queue() @bytes
3044 */
c5d67bd7
TH
3045static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3046{
3047 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3048}
3049
3050static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 3051 unsigned int pkts, unsigned int bytes)
c5d67bd7 3052{
114cf580 3053#ifdef CONFIG_BQL
b37c0fbe
AD
3054 if (unlikely(!bytes))
3055 return;
3056
3057 dql_completed(&dev_queue->dql, bytes);
3058
3059 /*
3060 * Without the memory barrier there is a small possiblity that
3061 * netdev_tx_sent_queue will miss the update and cause the queue to
3062 * be stopped forever
3063 */
3064 smp_mb();
3065
3066 if (dql_avail(&dev_queue->dql) < 0)
3067 return;
3068
3069 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3070 netif_schedule_queue(dev_queue);
114cf580 3071#endif
c5d67bd7
TH
3072}
3073
0042d0c8
FF
3074/**
3075 * netdev_completed_queue - report bytes and packets completed by device
3076 * @dev: network device
3077 * @pkts: actual number of packets sent over the medium
3078 * @bytes: actual number of bytes sent over the medium
3079 *
3080 * Report the number of bytes and packets transmitted by the network device
3081 * hardware queue over the physical medium, @bytes must exactly match the
3082 * @bytes amount passed to netdev_sent_queue()
3083 */
c5d67bd7 3084static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3085 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3086{
3087 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3088}
3089
3090static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3091{
114cf580 3092#ifdef CONFIG_BQL
5c490354 3093 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3094 dql_reset(&q->dql);
3095#endif
c5d67bd7
TH
3096}
3097
0042d0c8
FF
3098/**
3099 * netdev_reset_queue - reset the packets and bytes count of a network device
3100 * @dev_queue: network device
3101 *
3102 * Reset the bytes and packet count of a network device and clear the
3103 * software flow control OFF bit for this network device
3104 */
c5d67bd7
TH
3105static inline void netdev_reset_queue(struct net_device *dev_queue)
3106{
3107 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3108}
3109
b9507bda
DB
3110/**
3111 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3112 * @dev: network device
3113 * @queue_index: given tx queue index
3114 *
3115 * Returns 0 if given tx queue index >= number of device tx queues,
3116 * otherwise returns the originally passed tx queue index.
3117 */
3118static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3119{
3120 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3121 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3122 dev->name, queue_index,
3123 dev->real_num_tx_queues);
3124 return 0;
3125 }
3126
3127 return queue_index;
3128}
3129
bea3348e
SH
3130/**
3131 * netif_running - test if up
3132 * @dev: network device
3133 *
3134 * Test if the device has been brought up.
3135 */
4d29515f 3136static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3137{
3138 return test_bit(__LINK_STATE_START, &dev->state);
3139}
3140
f25f4e44 3141/*
5e82b4b2 3142 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3143 * stop, and a check if it's stopped. All other device management is
3144 * done at the overall netdevice level.
3145 * Also test the device if we're multiqueue.
3146 */
bea3348e
SH
3147
3148/**
3149 * netif_start_subqueue - allow sending packets on subqueue
3150 * @dev: network device
3151 * @queue_index: sub queue index
3152 *
3153 * Start individual transmit queue of a device with multiple transmit queues.
3154 */
f25f4e44
PWJ
3155static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3156{
fd2ea0a7 3157 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3158
3159 netif_tx_start_queue(txq);
f25f4e44
PWJ
3160}
3161
bea3348e
SH
3162/**
3163 * netif_stop_subqueue - stop sending packets on subqueue
3164 * @dev: network device
3165 * @queue_index: sub queue index
3166 *
3167 * Stop individual transmit queue of a device with multiple transmit queues.
3168 */
f25f4e44
PWJ
3169static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3170{
fd2ea0a7 3171 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3172 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3173}
3174
bea3348e
SH
3175/**
3176 * netif_subqueue_stopped - test status of subqueue
3177 * @dev: network device
3178 * @queue_index: sub queue index
3179 *
3180 * Check individual transmit queue of a device with multiple transmit queues.
3181 */
4d29515f
DM
3182static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3183 u16 queue_index)
f25f4e44 3184{
fd2ea0a7 3185 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3186
3187 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3188}
3189
4d29515f
DM
3190static inline bool netif_subqueue_stopped(const struct net_device *dev,
3191 struct sk_buff *skb)
668f895a
PE
3192{
3193 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3194}
bea3348e 3195
738b35cc
FF
3196/**
3197 * netif_wake_subqueue - allow sending packets on subqueue
3198 * @dev: network device
3199 * @queue_index: sub queue index
3200 *
3201 * Resume individual transmit queue of a device with multiple transmit queues.
3202 */
3203static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3204{
3205 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3206
3207 netif_tx_wake_queue(txq);
3208}
f25f4e44 3209
537c00de 3210#ifdef CONFIG_XPS
53af53ae 3211int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3212 u16 index);
537c00de
AD
3213#else
3214static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3215 const struct cpumask *mask,
537c00de
AD
3216 u16 index)
3217{
3218 return 0;
3219}
3220#endif
3221
5605c762
JP
3222u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3223 unsigned int num_tx_queues);
3224
a3d22a68
VZ
3225/*
3226 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3227 * as a distribution range limit for the returned value.
3228 */
3229static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3230 struct sk_buff *skb)
a3d22a68
VZ
3231{
3232 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3233}
3234
bea3348e
SH
3235/**
3236 * netif_is_multiqueue - test if device has multiple transmit queues
3237 * @dev: network device
3238 *
3239 * Check if device has multiple transmit queues
bea3348e 3240 */
4d29515f 3241static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3242{
a02cec21 3243 return dev->num_tx_queues > 1;
f25f4e44 3244}
1da177e4 3245
f629d208 3246int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3247
a953be53 3248#ifdef CONFIG_SYSFS
f629d208 3249int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3250#else
3251static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3252 unsigned int rxq)
3253{
3254 return 0;
3255}
3256#endif
3257
a953be53
MD
3258#ifdef CONFIG_SYSFS
3259static inline unsigned int get_netdev_rx_queue_index(
3260 struct netdev_rx_queue *queue)
3261{
3262 struct net_device *dev = queue->dev;
3263 int index = queue - dev->_rx;
3264
3265 BUG_ON(index >= dev->num_rx_queues);
3266 return index;
3267}
3268#endif
3269
16917b87 3270#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3271int netif_get_num_default_rss_queues(void);
16917b87 3272
e6247027
ED
3273enum skb_free_reason {
3274 SKB_REASON_CONSUMED,
3275 SKB_REASON_DROPPED,
3276};
3277
3278void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3279void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3280
e6247027
ED
3281/*
3282 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3283 * interrupt context or with hardware interrupts being disabled.
3284 * (in_irq() || irqs_disabled())
3285 *
3286 * We provide four helpers that can be used in following contexts :
3287 *
3288 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3289 * replacing kfree_skb(skb)
3290 *
3291 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3292 * Typically used in place of consume_skb(skb) in TX completion path
3293 *
3294 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3295 * replacing kfree_skb(skb)
3296 *
3297 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3298 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3299 */
e6247027
ED
3300static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3301{
3302 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3303}
3304
3305static inline void dev_consume_skb_irq(struct sk_buff *skb)
3306{
3307 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3308}
3309
3310static inline void dev_kfree_skb_any(struct sk_buff *skb)
3311{
3312 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3313}
3314
3315static inline void dev_consume_skb_any(struct sk_buff *skb)
3316{
3317 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3318}
1da177e4 3319
7c497478
JW
3320void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3321int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
f629d208
JP
3322int netif_rx(struct sk_buff *skb);
3323int netif_rx_ni(struct sk_buff *skb);
04eb4489 3324int netif_receive_skb(struct sk_buff *skb);
1c601d82 3325int netif_receive_skb_core(struct sk_buff *skb);
f629d208
JP
3326gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3327void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3328struct sk_buff *napi_get_frags(struct napi_struct *napi);
3329gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3330struct packet_offload *gro_find_receive_by_type(__be16 type);
3331struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3332
3333static inline void napi_free_frags(struct napi_struct *napi)
3334{
3335 kfree_skb(napi->skb);
3336 napi->skb = NULL;
3337}
3338
24b27fc4 3339bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3340int netdev_rx_handler_register(struct net_device *dev,
3341 rx_handler_func_t *rx_handler,
3342 void *rx_handler_data);
3343void netdev_rx_handler_unregister(struct net_device *dev);
3344
3345bool dev_valid_name(const char *name);
3346int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3347int dev_ethtool(struct net *net, struct ifreq *);
3348unsigned int dev_get_flags(const struct net_device *);
3349int __dev_change_flags(struct net_device *, unsigned int flags);
3350int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3351void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3352 unsigned int gchanges);
f629d208
JP
3353int dev_change_name(struct net_device *, const char *);
3354int dev_set_alias(struct net_device *, const char *, size_t);
6c557001 3355int dev_get_alias(const struct net_device *, char *, size_t);
f629d208 3356int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3357int __dev_set_mtu(struct net_device *, int);
f629d208 3358int dev_set_mtu(struct net_device *, int);
19cdeabd 3359int dev_validate_mtu(struct net_device *dev, int mtu);
f629d208
JP
3360void dev_set_group(struct net_device *, int);
3361int dev_set_mac_address(struct net_device *, struct sockaddr *);
3362int dev_change_carrier(struct net_device *, bool new_carrier);
3363int dev_get_phys_port_id(struct net_device *dev,
02637fce 3364 struct netdev_phys_item_id *ppid);
db24a904
DA
3365int dev_get_phys_port_name(struct net_device *dev,
3366 char *name, size_t len);
d746d707 3367int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3368struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3369struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3370 struct netdev_queue *txq, int *ret);
d67b9cd2 3371
f4e63525 3372typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
d67b9cd2
DB
3373int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3374 int fd, u32 flags);
f4e63525 3375u8 __dev_xdp_attached(struct net_device *dev, bpf_op_t xdp_op, u32 *prog_id);
d67b9cd2 3376
a0265d28 3377int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3378int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3379bool is_skb_forwardable(const struct net_device *dev,
3380 const struct sk_buff *skb);
1da177e4 3381
4e3264d2
MKL
3382static __always_inline int ____dev_forward_skb(struct net_device *dev,
3383 struct sk_buff *skb)
3384{
3385 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3386 unlikely(!is_skb_forwardable(dev, skb))) {
3387 atomic_long_inc(&dev->rx_dropped);
3388 kfree_skb(skb);
3389 return NET_RX_DROP;
3390 }
3391
3392 skb_scrub_packet(skb, true);
3393 skb->priority = 0;
3394 return 0;
3395}
3396
74b20582
DA
3397void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3398
20380731 3399extern int netdev_budget;
7acf8a1e 3400extern unsigned int netdev_budget_usecs;
1da177e4
LT
3401
3402/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3403void netdev_run_todo(void);
1da177e4 3404
bea3348e
SH
3405/**
3406 * dev_put - release reference to device
3407 * @dev: network device
3408 *
9ef4429b 3409 * Release reference to device to allow it to be freed.
bea3348e 3410 */
1da177e4
LT
3411static inline void dev_put(struct net_device *dev)
3412{
933393f5 3413 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3414}
3415
bea3348e
SH
3416/**
3417 * dev_hold - get reference to device
3418 * @dev: network device
3419 *
9ef4429b 3420 * Hold reference to device to keep it from being freed.
bea3348e 3421 */
15333061
SH
3422static inline void dev_hold(struct net_device *dev)
3423{
933393f5 3424 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3425}
1da177e4
LT
3426
3427/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3428 * and _off may be called from IRQ context, but it is caller
3429 * who is responsible for serialization of these calls.
b00055aa
SR
3430 *
3431 * The name carrier is inappropriate, these functions should really be
3432 * called netif_lowerlayer_*() because they represent the state of any
3433 * kind of lower layer not just hardware media.
1da177e4
LT
3434 */
3435
f629d208
JP
3436void linkwatch_init_dev(struct net_device *dev);
3437void linkwatch_fire_event(struct net_device *dev);
3438void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3439
bea3348e
SH
3440/**
3441 * netif_carrier_ok - test if carrier present
3442 * @dev: network device
3443 *
3444 * Check if carrier is present on device
3445 */
4d29515f 3446static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3447{
3448 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3449}
3450
f629d208 3451unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3452
f629d208 3453void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3454
f629d208 3455void netif_carrier_on(struct net_device *dev);
1da177e4 3456
f629d208 3457void netif_carrier_off(struct net_device *dev);
1da177e4 3458
bea3348e
SH
3459/**
3460 * netif_dormant_on - mark device as dormant.
3461 * @dev: network device
3462 *
3463 * Mark device as dormant (as per RFC2863).
3464 *
3465 * The dormant state indicates that the relevant interface is not
3466 * actually in a condition to pass packets (i.e., it is not 'up') but is
3467 * in a "pending" state, waiting for some external event. For "on-
3468 * demand" interfaces, this new state identifies the situation where the
3469 * interface is waiting for events to place it in the up state.
bea3348e 3470 */
b00055aa
SR
3471static inline void netif_dormant_on(struct net_device *dev)
3472{
3473 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3474 linkwatch_fire_event(dev);
3475}
3476
bea3348e
SH
3477/**
3478 * netif_dormant_off - set device as not dormant.
3479 * @dev: network device
3480 *
3481 * Device is not in dormant state.
3482 */
b00055aa
SR
3483static inline void netif_dormant_off(struct net_device *dev)
3484{
3485 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3486 linkwatch_fire_event(dev);
3487}
3488
bea3348e 3489/**
8ecbc40a 3490 * netif_dormant - test if device is dormant
bea3348e
SH
3491 * @dev: network device
3492 *
8ecbc40a 3493 * Check if device is dormant.
bea3348e 3494 */
4d29515f 3495static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3496{
3497 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3498}
3499
3500
bea3348e
SH
3501/**
3502 * netif_oper_up - test if device is operational
3503 * @dev: network device
3504 *
3505 * Check if carrier is operational
3506 */
4d29515f 3507static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3508{
b00055aa
SR
3509 return (dev->operstate == IF_OPER_UP ||
3510 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3511}
3512
bea3348e
SH
3513/**
3514 * netif_device_present - is device available or removed
3515 * @dev: network device
3516 *
3517 * Check if device has not been removed from system.
3518 */
4d29515f 3519static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3520{
3521 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3522}
3523
f629d208 3524void netif_device_detach(struct net_device *dev);
1da177e4 3525
f629d208 3526void netif_device_attach(struct net_device *dev);
1da177e4
LT
3527
3528/*
3529 * Network interface message level settings
3530 */
1da177e4
LT
3531
3532enum {
3533 NETIF_MSG_DRV = 0x0001,
3534 NETIF_MSG_PROBE = 0x0002,
3535 NETIF_MSG_LINK = 0x0004,
3536 NETIF_MSG_TIMER = 0x0008,
3537 NETIF_MSG_IFDOWN = 0x0010,
3538 NETIF_MSG_IFUP = 0x0020,
3539 NETIF_MSG_RX_ERR = 0x0040,
3540 NETIF_MSG_TX_ERR = 0x0080,
3541 NETIF_MSG_TX_QUEUED = 0x0100,
3542 NETIF_MSG_INTR = 0x0200,
3543 NETIF_MSG_TX_DONE = 0x0400,
3544 NETIF_MSG_RX_STATUS = 0x0800,
3545 NETIF_MSG_PKTDATA = 0x1000,
3546 NETIF_MSG_HW = 0x2000,
3547 NETIF_MSG_WOL = 0x4000,
3548};
3549
3550#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3551#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3552#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3553#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3554#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3555#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3556#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3557#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3558#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3559#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3560#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3561#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3562#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3563#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3564#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3565
3566static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3567{
3568 /* use default */
3569 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3570 return default_msg_enable_bits;
3571 if (debug_value == 0) /* no output */
3572 return 0;
3573 /* set low N bits */
5c199eb7 3574 return (1U << debug_value) - 1;
1da177e4
LT
3575}
3576
c773e847 3577static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3578{
c773e847
DM
3579 spin_lock(&txq->_xmit_lock);
3580 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3581}
3582
5a717f4f
MT
3583static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3584{
3585 __acquire(&txq->_xmit_lock);
3586 return true;
3587}
3588
3589static inline void __netif_tx_release(struct netdev_queue *txq)
3590{
3591 __release(&txq->_xmit_lock);
3592}
3593
fd2ea0a7
DM
3594static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3595{
3596 spin_lock_bh(&txq->_xmit_lock);
3597 txq->xmit_lock_owner = smp_processor_id();
3598}
3599
4d29515f 3600static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3601{
4d29515f 3602 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3603 if (likely(ok))
3604 txq->xmit_lock_owner = smp_processor_id();
3605 return ok;
3606}
3607
3608static inline void __netif_tx_unlock(struct netdev_queue *txq)
3609{
3610 txq->xmit_lock_owner = -1;
3611 spin_unlock(&txq->_xmit_lock);
3612}
3613
3614static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3615{
3616 txq->xmit_lock_owner = -1;
3617 spin_unlock_bh(&txq->_xmit_lock);
3618}
3619
08baf561
ED
3620static inline void txq_trans_update(struct netdev_queue *txq)
3621{
3622 if (txq->xmit_lock_owner != -1)
3623 txq->trans_start = jiffies;
3624}
3625
ba162f8e
FW
3626/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3627static inline void netif_trans_update(struct net_device *dev)
3628{
9b36627a
FW
3629 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3630
3631 if (txq->trans_start != jiffies)
3632 txq->trans_start = jiffies;
ba162f8e
FW
3633}
3634
d29f749e
DJ
3635/**
3636 * netif_tx_lock - grab network device transmit lock
3637 * @dev: network device
d29f749e
DJ
3638 *
3639 * Get network device transmit lock
3640 */
22dd7495
JHS
3641static inline void netif_tx_lock(struct net_device *dev)
3642{
e8a0464c 3643 unsigned int i;
c3f26a26 3644 int cpu;
c773e847 3645
c3f26a26
DM
3646 spin_lock(&dev->tx_global_lock);
3647 cpu = smp_processor_id();
e8a0464c
DM
3648 for (i = 0; i < dev->num_tx_queues; i++) {
3649 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3650
3651 /* We are the only thread of execution doing a
3652 * freeze, but we have to grab the _xmit_lock in
3653 * order to synchronize with threads which are in
3654 * the ->hard_start_xmit() handler and already
3655 * checked the frozen bit.
3656 */
e8a0464c 3657 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3658 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3659 __netif_tx_unlock(txq);
e8a0464c 3660 }
932ff279
HX
3661}
3662
3663static inline void netif_tx_lock_bh(struct net_device *dev)
3664{
e8a0464c
DM
3665 local_bh_disable();
3666 netif_tx_lock(dev);
932ff279
HX
3667}
3668
932ff279
HX
3669static inline void netif_tx_unlock(struct net_device *dev)
3670{
e8a0464c
DM
3671 unsigned int i;
3672
3673 for (i = 0; i < dev->num_tx_queues; i++) {
3674 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3675
c3f26a26
DM
3676 /* No need to grab the _xmit_lock here. If the
3677 * queue is not stopped for another reason, we
3678 * force a schedule.
3679 */
3680 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3681 netif_schedule_queue(txq);
c3f26a26
DM
3682 }
3683 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3684}
3685
3686static inline void netif_tx_unlock_bh(struct net_device *dev)
3687{
e8a0464c
DM
3688 netif_tx_unlock(dev);
3689 local_bh_enable();
932ff279
HX
3690}
3691
c773e847 3692#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3693 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3694 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3695 } else { \
3696 __netif_tx_acquire(txq); \
22dd7495
JHS
3697 } \
3698}
3699
5efeac44
EB
3700#define HARD_TX_TRYLOCK(dev, txq) \
3701 (((dev->features & NETIF_F_LLTX) == 0) ? \
3702 __netif_tx_trylock(txq) : \
5a717f4f 3703 __netif_tx_acquire(txq))
5efeac44 3704
c773e847 3705#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3706 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3707 __netif_tx_unlock(txq); \
5a717f4f
MT
3708 } else { \
3709 __netif_tx_release(txq); \
22dd7495
JHS
3710 } \
3711}
3712
1da177e4
LT
3713static inline void netif_tx_disable(struct net_device *dev)
3714{
fd2ea0a7 3715 unsigned int i;
c3f26a26 3716 int cpu;
fd2ea0a7 3717
c3f26a26
DM
3718 local_bh_disable();
3719 cpu = smp_processor_id();
fd2ea0a7
DM
3720 for (i = 0; i < dev->num_tx_queues; i++) {
3721 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3722
3723 __netif_tx_lock(txq, cpu);
fd2ea0a7 3724 netif_tx_stop_queue(txq);
c3f26a26 3725 __netif_tx_unlock(txq);
fd2ea0a7 3726 }
c3f26a26 3727 local_bh_enable();
1da177e4
LT
3728}
3729
e308a5d8
DM
3730static inline void netif_addr_lock(struct net_device *dev)
3731{
3732 spin_lock(&dev->addr_list_lock);
3733}
3734
2429f7ac
JP
3735static inline void netif_addr_lock_nested(struct net_device *dev)
3736{
25175ba5
VY
3737 int subclass = SINGLE_DEPTH_NESTING;
3738
3739 if (dev->netdev_ops->ndo_get_lock_subclass)
3740 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3741
3742 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3743}
3744
e308a5d8
DM
3745static inline void netif_addr_lock_bh(struct net_device *dev)
3746{
3747 spin_lock_bh(&dev->addr_list_lock);
3748}
3749
3750static inline void netif_addr_unlock(struct net_device *dev)
3751{
3752 spin_unlock(&dev->addr_list_lock);
3753}
3754
3755static inline void netif_addr_unlock_bh(struct net_device *dev)
3756{
3757 spin_unlock_bh(&dev->addr_list_lock);
3758}
3759
f001fde5 3760/*
31278e71 3761 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3762 * rcu_read_lock held.
3763 */
3764#define for_each_dev_addr(dev, ha) \
31278e71 3765 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3766
1da177e4
LT
3767/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3768
f629d208 3769void ether_setup(struct net_device *dev);
1da177e4
LT
3770
3771/* Support for loadable net-drivers */
f629d208 3772struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3773 unsigned char name_assign_type,
f629d208
JP
3774 void (*setup)(struct net_device *),
3775 unsigned int txqs, unsigned int rxqs);
0ad646c8
CW
3776int dev_get_valid_name(struct net *net, struct net_device *dev,
3777 const char *name);
3778
c835a677
TG
3779#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3780 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3781
c835a677
TG
3782#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3783 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3784 count)
36909ea4 3785
f629d208
JP
3786int register_netdev(struct net_device *dev);
3787void unregister_netdev(struct net_device *dev);
f001fde5 3788
22bedad3 3789/* General hardware address lists handling functions */
f629d208
JP
3790int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3791 struct netdev_hw_addr_list *from_list, int addr_len);
3792void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3793 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3794int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3795 struct net_device *dev,
3796 int (*sync)(struct net_device *, const unsigned char *),
3797 int (*unsync)(struct net_device *,
3798 const unsigned char *));
3799void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3800 struct net_device *dev,
3801 int (*unsync)(struct net_device *,
3802 const unsigned char *));
f629d208 3803void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3804
f001fde5 3805/* Functions used for device addresses handling */
f629d208
JP
3806int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3807 unsigned char addr_type);
3808int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3809 unsigned char addr_type);
f629d208
JP
3810void dev_addr_flush(struct net_device *dev);
3811int dev_addr_init(struct net_device *dev);
a748ee24
JP
3812
3813/* Functions used for unicast addresses handling */
f629d208
JP
3814int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3815int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3816int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3817int dev_uc_sync(struct net_device *to, struct net_device *from);
3818int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3819void dev_uc_unsync(struct net_device *to, struct net_device *from);
3820void dev_uc_flush(struct net_device *dev);
3821void dev_uc_init(struct net_device *dev);
f001fde5 3822
670e5b8e
AD
3823/**
3824 * __dev_uc_sync - Synchonize device's unicast list
3825 * @dev: device to sync
3826 * @sync: function to call if address should be added
3827 * @unsync: function to call if address should be removed
3828 *
3829 * Add newly added addresses to the interface, and release
3830 * addresses that have been deleted.
5e82b4b2 3831 */
670e5b8e
AD
3832static inline int __dev_uc_sync(struct net_device *dev,
3833 int (*sync)(struct net_device *,
3834 const unsigned char *),
3835 int (*unsync)(struct net_device *,
3836 const unsigned char *))
3837{
3838 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3839}
3840
3841/**
e793c0f7 3842 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3843 * @dev: device to sync
3844 * @unsync: function to call if address should be removed
3845 *
3846 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3847 */
670e5b8e
AD
3848static inline void __dev_uc_unsync(struct net_device *dev,
3849 int (*unsync)(struct net_device *,
3850 const unsigned char *))
3851{
3852 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3853}
3854
22bedad3 3855/* Functions used for multicast addresses handling */
f629d208
JP
3856int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3857int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3858int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3859int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3860int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3861int dev_mc_sync(struct net_device *to, struct net_device *from);
3862int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3863void dev_mc_unsync(struct net_device *to, struct net_device *from);
3864void dev_mc_flush(struct net_device *dev);
3865void dev_mc_init(struct net_device *dev);
f001fde5 3866
670e5b8e
AD
3867/**
3868 * __dev_mc_sync - Synchonize device's multicast list
3869 * @dev: device to sync
3870 * @sync: function to call if address should be added
3871 * @unsync: function to call if address should be removed
3872 *
3873 * Add newly added addresses to the interface, and release
3874 * addresses that have been deleted.
5e82b4b2 3875 */
670e5b8e
AD
3876static inline int __dev_mc_sync(struct net_device *dev,
3877 int (*sync)(struct net_device *,
3878 const unsigned char *),
3879 int (*unsync)(struct net_device *,
3880 const unsigned char *))
3881{
3882 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3883}
3884
3885/**
e793c0f7 3886 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3887 * @dev: device to sync
3888 * @unsync: function to call if address should be removed
3889 *
3890 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3891 */
670e5b8e
AD
3892static inline void __dev_mc_unsync(struct net_device *dev,
3893 int (*unsync)(struct net_device *,
3894 const unsigned char *))
3895{
3896 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3897}
3898
4417da66 3899/* Functions used for secondary unicast and multicast support */
f629d208
JP
3900void dev_set_rx_mode(struct net_device *dev);
3901void __dev_set_rx_mode(struct net_device *dev);
3902int dev_set_promiscuity(struct net_device *dev, int inc);
3903int dev_set_allmulti(struct net_device *dev, int inc);
3904void netdev_state_change(struct net_device *dev);
3905void netdev_notify_peers(struct net_device *dev);
3906void netdev_features_change(struct net_device *dev);
1da177e4 3907/* Load a device via the kmod */
f629d208
JP
3908void dev_load(struct net *net, const char *name);
3909struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3910 struct rtnl_link_stats64 *storage);
3911void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3912 const struct net_device_stats *netdev_stats);
eeda3fd6 3913
1da177e4 3914extern int netdev_max_backlog;
3b098e2d 3915extern int netdev_tstamp_prequeue;
1da177e4 3916extern int weight_p;
3d48b53f
MT
3917extern int dev_weight_rx_bias;
3918extern int dev_weight_tx_bias;
3919extern int dev_rx_weight;
3920extern int dev_tx_weight;
9ff162a8 3921
f629d208 3922bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3923struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3924 struct list_head **iter);
f629d208
JP
3925struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3926 struct list_head **iter);
8b5be856 3927
44a40855
VY
3928/* iterate through upper list, must be called under RCU read lock */
3929#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3930 for (iter = &(dev)->adj_list.upper, \
3931 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3932 updev; \
3933 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3934
1a3f060c
DA
3935int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3936 int (*fn)(struct net_device *upper_dev,
3937 void *data),
3938 void *data);
3939
3940bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3941 struct net_device *upper_dev);
3942
25cc72a3
IS
3943bool netdev_has_any_upper_dev(struct net_device *dev);
3944
f629d208
JP
3945void *netdev_lower_get_next_private(struct net_device *dev,
3946 struct list_head **iter);
3947void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3948 struct list_head **iter);
31088a11
VF
3949
3950#define netdev_for_each_lower_private(dev, priv, iter) \
3951 for (iter = (dev)->adj_list.lower.next, \
3952 priv = netdev_lower_get_next_private(dev, &(iter)); \
3953 priv; \
3954 priv = netdev_lower_get_next_private(dev, &(iter)))
3955
3956#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3957 for (iter = &(dev)->adj_list.lower, \
3958 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3959 priv; \
3960 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3961
4085ebe8
VY
3962void *netdev_lower_get_next(struct net_device *dev,
3963 struct list_head **iter);
7ce856aa 3964
4085ebe8 3965#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3966 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3967 ldev = netdev_lower_get_next(dev, &(iter)); \
3968 ldev; \
3969 ldev = netdev_lower_get_next(dev, &(iter)))
3970
7ce856aa
JP
3971struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3972 struct list_head **iter);
3973struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3974 struct list_head **iter);
3975
1a3f060c
DA
3976int netdev_walk_all_lower_dev(struct net_device *dev,
3977 int (*fn)(struct net_device *lower_dev,
3978 void *data),
3979 void *data);
3980int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3981 int (*fn)(struct net_device *lower_dev,
3982 void *data),
3983 void *data);
3984
f629d208 3985void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3986void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3987struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3988struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
42ab19ee
DA
3989int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
3990 struct netlink_ext_ack *extack);
f629d208 3991int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3992 struct net_device *upper_dev,
42ab19ee
DA
3993 void *upper_priv, void *upper_info,
3994 struct netlink_ext_ack *extack);
f629d208
JP
3995void netdev_upper_dev_unlink(struct net_device *dev,
3996 struct net_device *upper_dev);
5bb025fa 3997void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3998void *netdev_lower_dev_get_private(struct net_device *dev,
3999 struct net_device *lower_dev);
04d48266
JP
4000void netdev_lower_state_changed(struct net_device *lower_dev,
4001 void *lower_state_info);
960fb622
ED
4002
4003/* RSS keys are 40 or 52 bytes long */
4004#define NETDEV_RSS_KEY_LEN 52
ba905f5e 4005extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
4006void netdev_rss_key_fill(void *buffer, size_t len);
4007
952fcfd0 4008int dev_get_nest_level(struct net_device *dev);
f629d208 4009int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 4010int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
4011int skb_csum_hwoffload_help(struct sk_buff *skb,
4012 const netdev_features_t features);
4013
f629d208
JP
4014struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4015 netdev_features_t features, bool tx_path);
4016struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4017 netdev_features_t features);
12b0004d 4018
61bd3857
MS
4019struct netdev_bonding_info {
4020 ifslave slave;
4021 ifbond master;
4022};
4023
4024struct netdev_notifier_bonding_info {
4025 struct netdev_notifier_info info; /* must be first */
4026 struct netdev_bonding_info bonding_info;
4027};
4028
4029void netdev_bonding_info_change(struct net_device *dev,
4030 struct netdev_bonding_info *bonding_info);
4031
12b0004d
CW
4032static inline
4033struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4034{
4035 return __skb_gso_segment(skb, features, true);
4036}
53d6471c 4037__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
4038
4039static inline bool can_checksum_protocol(netdev_features_t features,
4040 __be16 protocol)
4041{
c8cd0989
TH
4042 if (protocol == htons(ETH_P_FCOE))
4043 return !!(features & NETIF_F_FCOE_CRC);
4044
4045 /* Assume this is an IP checksum (not SCTP CRC) */
4046
4047 if (features & NETIF_F_HW_CSUM) {
4048 /* Can checksum everything */
4049 return true;
4050 }
4051
4052 switch (protocol) {
4053 case htons(ETH_P_IP):
4054 return !!(features & NETIF_F_IP_CSUM);
4055 case htons(ETH_P_IPV6):
4056 return !!(features & NETIF_F_IPV6_CSUM);
4057 default:
4058 return false;
4059 }
ec5f0615 4060}
12b0004d 4061
fb286bb2 4062#ifdef CONFIG_BUG
f629d208 4063void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
4064#else
4065static inline void netdev_rx_csum_fault(struct net_device *dev)
4066{
4067}
4068#endif
1da177e4 4069/* rx skb timestamps */
f629d208
JP
4070void net_enable_timestamp(void);
4071void net_disable_timestamp(void);
1da177e4 4072
20380731 4073#ifdef CONFIG_PROC_FS
f629d208 4074int __init dev_proc_init(void);
900ff8c6
CW
4075#else
4076#define dev_proc_init() 0
20380731
ACM
4077#endif
4078
4798248e 4079static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4080 struct sk_buff *skb, struct net_device *dev,
4081 bool more)
4798248e 4082{
fa2dbdc2 4083 skb->xmit_more = more ? 1 : 0;
0b725a2c 4084 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4085}
4086
10b3ad8c 4087static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4088 struct netdev_queue *txq, bool more)
4798248e
DM
4089{
4090 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4091 int rc;
4798248e 4092
fa2dbdc2 4093 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4094 if (rc == NETDEV_TX_OK)
4095 txq_trans_update(txq);
4096
4097 return rc;
4798248e
DM
4098}
4099
b793dc5c 4100int netdev_class_create_file_ns(const struct class_attribute *class_attr,
42a2d923 4101 const void *ns);
b793dc5c 4102void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
42a2d923 4103 const void *ns);
58292cbe 4104
b793dc5c 4105static inline int netdev_class_create_file(const struct class_attribute *class_attr)
58292cbe
TH
4106{
4107 return netdev_class_create_file_ns(class_attr, NULL);
4108}
4109
b793dc5c 4110static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
58292cbe
TH
4111{
4112 netdev_class_remove_file_ns(class_attr, NULL);
4113}
b8a9787e 4114
737aec57 4115extern const struct kobj_ns_type_operations net_ns_type_operations;
04600794 4116
f629d208 4117const char *netdev_drivername(const struct net_device *dev);
6579e57b 4118
f629d208 4119void linkwatch_run_queue(void);
20380731 4120
da08143b
MK
4121static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4122 netdev_features_t f2)
4123{
c8cd0989
TH
4124 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4125 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4126 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4127 else
b6a0e72a 4128 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4129 }
da08143b 4130
c8cd0989 4131 return f1 & f2;
da08143b
MK
4132}
4133
c8f44aff
MM
4134static inline netdev_features_t netdev_get_wanted_features(
4135 struct net_device *dev)
5455c699
MM
4136{
4137 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4138}
c8f44aff
MM
4139netdev_features_t netdev_increment_features(netdev_features_t all,
4140 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4141
4142/* Allow TSO being used on stacked device :
4143 * Performing the GSO segmentation before last device
4144 * is a performance improvement.
4145 */
4146static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4147 netdev_features_t mask)
4148{
4149 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4150}
4151
6cb6a27c 4152int __netdev_update_features(struct net_device *dev);
5455c699 4153void netdev_update_features(struct net_device *dev);
afe12cc8 4154void netdev_change_features(struct net_device *dev);
7f353bf2 4155
fc4a7489
PM
4156void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4157 struct net_device *dev);
4158
e38f3025
TM
4159netdev_features_t passthru_features_check(struct sk_buff *skb,
4160 struct net_device *dev,
4161 netdev_features_t features);
c1e756bf 4162netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4163
4d29515f 4164static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4165{
7b748340 4166 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4167
4168 /* check flags correspondence */
4169 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4170 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4171 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4172 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4173 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4174 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4175 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4176 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4177 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4178 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4179 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4180 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4181 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4182 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4183 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4184 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0c19f846 4185 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
0345e186 4186
d6b4991a 4187 return (features & feature) == feature;
576a30eb
HX
4188}
4189
4d29515f 4190static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4191{
278b2513 4192 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4193 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4194}
4195
8b86a61d 4196static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4197 netdev_features_t features)
7967168c 4198{
fc741216 4199 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4200 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4201 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4202}
4203
82cc1a7a
PWJ
4204static inline void netif_set_gso_max_size(struct net_device *dev,
4205 unsigned int size)
4206{
4207 dev->gso_max_size = size;
4208}
4209
7a7ffbab
WCC
4210static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4211 int pulled_hlen, u16 mac_offset,
4212 int mac_len)
4213{
4214 skb->protocol = protocol;
4215 skb->encapsulation = 1;
4216 skb_push(skb, pulled_hlen);
4217 skb_reset_transport_header(skb);
4218 skb->mac_header = mac_offset;
4219 skb->network_header = skb->mac_header + mac_len;
4220 skb->mac_len = mac_len;
4221}
4222
3c175784
SD
4223static inline bool netif_is_macsec(const struct net_device *dev)
4224{
4225 return dev->priv_flags & IFF_MACSEC;
4226}
4227
b618aaa9 4228static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4229{
4230 return dev->priv_flags & IFF_MACVLAN;
4231}
4232
b618aaa9 4233static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4234{
4235 return dev->priv_flags & IFF_MACVLAN_PORT;
4236}
4237
b618aaa9 4238static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4239{
4240 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4241}
4242
b618aaa9 4243static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4244{
4245 return dev->priv_flags & IFF_IPVLAN_MASTER;
4246}
4247
b618aaa9 4248static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4249{
4250 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4251}
4252
b618aaa9 4253static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4254{
4255 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4256}
4257
3bdc0eba
BG
4258static inline bool netif_supports_nofcs(struct net_device *dev)
4259{
4260 return dev->priv_flags & IFF_SUPP_NOFCS;
4261}
4262
9d150e29
DB
4263static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4264{
4265 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4266}
4267
007979ea 4268static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4269{
007979ea 4270 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4271}
4272
fee6d4c7
DA
4273static inline bool netif_is_l3_slave(const struct net_device *dev)
4274{
4275 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4276}
4277
0894ae3f
JP
4278static inline bool netif_is_bridge_master(const struct net_device *dev)
4279{
4280 return dev->priv_flags & IFF_EBRIDGE;
4281}
4282
28f9ee22
VY
4283static inline bool netif_is_bridge_port(const struct net_device *dev)
4284{
4285 return dev->priv_flags & IFF_BRIDGE_PORT;
4286}
4287
35d4e172
JP
4288static inline bool netif_is_ovs_master(const struct net_device *dev)
4289{
4290 return dev->priv_flags & IFF_OPENVSWITCH;
4291}
4292
5be66141
JP
4293static inline bool netif_is_ovs_port(const struct net_device *dev)
4294{
4295 return dev->priv_flags & IFF_OVS_DATAPATH;
4296}
4297
b618aaa9 4298static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4299{
4300 return dev->priv_flags & IFF_TEAM;
4301}
4302
b618aaa9 4303static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4304{
4305 return dev->priv_flags & IFF_TEAM_PORT;
4306}
4307
b618aaa9 4308static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4309{
4310 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4311}
4312
b618aaa9 4313static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4314{
4315 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4316}
4317
d4ab4286
KJ
4318static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4319{
4320 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4321}
4322
02875878
ED
4323/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4324static inline void netif_keep_dst(struct net_device *dev)
4325{
4326 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4327}
4328
18d3df3e
PA
4329/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4330static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4331{
4332 /* TODO: reserve and use an additional IFF bit, if we get more users */
4333 return dev->priv_flags & IFF_MACSEC;
4334}
4335
505d4f73 4336extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4337
571ba423
JP
4338/* Logging, debugging and troubleshooting/diagnostic helpers. */
4339
4340/* netdev_printk helpers, similar to dev_printk */
4341
4342static inline const char *netdev_name(const struct net_device *dev)
4343{
c6f854d5
VF
4344 if (!dev->name[0] || strchr(dev->name, '%'))
4345 return "(unnamed net_device)";
571ba423
JP
4346 return dev->name;
4347}
4348
8397ed36
DA
4349static inline bool netdev_unregistering(const struct net_device *dev)
4350{
4351 return dev->reg_state == NETREG_UNREGISTERING;
4352}
4353
ccc7f496
VF
4354static inline const char *netdev_reg_state(const struct net_device *dev)
4355{
4356 switch (dev->reg_state) {
4357 case NETREG_UNINITIALIZED: return " (uninitialized)";
4358 case NETREG_REGISTERED: return "";
4359 case NETREG_UNREGISTERING: return " (unregistering)";
4360 case NETREG_UNREGISTERED: return " (unregistered)";
4361 case NETREG_RELEASED: return " (released)";
4362 case NETREG_DUMMY: return " (dummy)";
4363 }
4364
4365 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4366 return " (unknown)";
4367}
4368
f629d208 4369__printf(3, 4)
6ea754eb
JP
4370void netdev_printk(const char *level, const struct net_device *dev,
4371 const char *format, ...);
f629d208 4372__printf(2, 3)
6ea754eb 4373void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4374__printf(2, 3)
6ea754eb 4375void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4376__printf(2, 3)
6ea754eb 4377void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4378__printf(2, 3)
6ea754eb 4379void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4380__printf(2, 3)
6ea754eb 4381void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4382__printf(2, 3)
6ea754eb 4383void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4384__printf(2, 3)
6ea754eb 4385void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4386
375ef2b1
GP
4387#define netdev_level_once(level, dev, fmt, ...) \
4388do { \
4389 static bool __print_once __read_mostly; \
4390 \
4391 if (!__print_once) { \
4392 __print_once = true; \
4393 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4394 } \
4395} while (0)
4396
4397#define netdev_emerg_once(dev, fmt, ...) \
4398 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4399#define netdev_alert_once(dev, fmt, ...) \
4400 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4401#define netdev_crit_once(dev, fmt, ...) \
4402 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4403#define netdev_err_once(dev, fmt, ...) \
4404 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4405#define netdev_warn_once(dev, fmt, ...) \
4406 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4407#define netdev_notice_once(dev, fmt, ...) \
4408 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4409#define netdev_info_once(dev, fmt, ...) \
4410 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4411
8909c9ad
VK
4412#define MODULE_ALIAS_NETDEV(device) \
4413 MODULE_ALIAS("netdev-" device)
4414
b558c96f 4415#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4416#define netdev_dbg(__dev, format, args...) \
4417do { \
ffa10cb4 4418 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4419} while (0)
b558c96f
JC
4420#elif defined(DEBUG)
4421#define netdev_dbg(__dev, format, args...) \
4422 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4423#else
4424#define netdev_dbg(__dev, format, args...) \
4425({ \
4426 if (0) \
4427 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4428})
4429#endif
4430
4431#if defined(VERBOSE_DEBUG)
4432#define netdev_vdbg netdev_dbg
4433#else
4434
4435#define netdev_vdbg(dev, format, args...) \
4436({ \
4437 if (0) \
4438 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4439 0; \
4440})
4441#endif
4442
4443/*
4444 * netdev_WARN() acts like dev_printk(), but with the key difference
4445 * of using a WARN/WARN_ON to get the message out, including the
4446 * file/line information and a backtrace.
4447 */
4448#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4449 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4450 netdev_reg_state(dev), ##args)
571ba423 4451
1e4a8e87
GP
4452#define netdev_WARN_ONCE(dev, format, args...) \
4453 WARN_ONCE(1, "netdevice: %s%s\n" format, netdev_name(dev), \
375ef2b1
GP
4454 netdev_reg_state(dev), ##args)
4455
b3d95c5c
JP
4456/* netif printk helpers, similar to netdev_printk */
4457
4458#define netif_printk(priv, type, level, dev, fmt, args...) \
4459do { \
4460 if (netif_msg_##type(priv)) \
4461 netdev_printk(level, (dev), fmt, ##args); \
4462} while (0)
4463
f45f4321
JP
4464#define netif_level(level, priv, type, dev, fmt, args...) \
4465do { \
4466 if (netif_msg_##type(priv)) \
4467 netdev_##level(dev, fmt, ##args); \
4468} while (0)
4469
b3d95c5c 4470#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4471 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4472#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4473 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4474#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4475 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4476#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4477 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4478#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4479 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4480#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4481 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4482#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4483 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4484
0053ea9c 4485#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4486#define netif_dbg(priv, type, netdev, format, args...) \
4487do { \
4488 if (netif_msg_##type(priv)) \
b5fb0a03 4489 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4490} while (0)
0053ea9c
JP
4491#elif defined(DEBUG)
4492#define netif_dbg(priv, type, dev, format, args...) \
4493 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4494#else
4495#define netif_dbg(priv, type, dev, format, args...) \
4496({ \
4497 if (0) \
4498 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4499 0; \
4500})
4501#endif
4502
f617f276
EC
4503/* if @cond then downgrade to debug, else print at @level */
4504#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4505 do { \
4506 if (cond) \
4507 netif_dbg(priv, type, netdev, fmt, ##args); \
4508 else \
4509 netif_ ## level(priv, type, netdev, fmt, ##args); \
4510 } while (0)
4511
b3d95c5c 4512#if defined(VERBOSE_DEBUG)
bcfcc450 4513#define netif_vdbg netif_dbg
b3d95c5c
JP
4514#else
4515#define netif_vdbg(priv, type, dev, format, args...) \
4516({ \
4517 if (0) \
a4ed89cb 4518 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4519 0; \
4520})
4521#endif
571ba423 4522
900ff8c6
CW
4523/*
4524 * The list of packet types we will receive (as opposed to discard)
4525 * and the routines to invoke.
4526 *
4527 * Why 16. Because with 16 the only overlap we get on a hash of the
4528 * low nibble of the protocol value is RARP/SNAP/X.25.
4529 *
900ff8c6 4530 * 0800 IP
900ff8c6
CW
4531 * 0001 802.3
4532 * 0002 AX.25
4533 * 0004 802.2
4534 * 8035 RARP
4535 * 0005 SNAP
4536 * 0805 X.25
4537 * 0806 ARP
4538 * 8137 IPX
4539 * 0009 Localtalk
4540 * 86DD IPv6
4541 */
4542#define PTYPE_HASH_SIZE (16)
4543#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4544
385a154c 4545#endif /* _LINUX_NETDEVICE_H */