]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/netdevice.h
regulator: ab8500: Remove AB8505 USB regulator
[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
8626a0c8 1542 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1543 * @dev_id: Used to differentiate devices that share
1544 * the same link layer address
1545 * @dev_port: Used to differentiate devices that share
1546 * the same function
1547 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1548 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1549 * has been enabled due to the need to listen to
1550 * additional unicast addresses in a device that
1551 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1552 * @uc: unicast mac addresses
1553 * @mc: multicast mac addresses
1554 * @dev_addrs: list of device hw addresses
1555 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1556 * @promiscuity: Number of times the NIC is told to work in
1557 * promiscuous mode; if it becomes 0 the NIC will
1558 * exit promiscuous mode
536721b1
KK
1559 * @allmulti: Counter, enables or disables allmulticast mode
1560 *
1561 * @vlan_info: VLAN info
1562 * @dsa_ptr: dsa specific data
1563 * @tipc_ptr: TIPC specific data
1564 * @atalk_ptr: AppleTalk link
1565 * @ip_ptr: IPv4 specific data
1566 * @dn_ptr: DECnet specific data
1567 * @ip6_ptr: IPv6 specific data
1568 * @ax25_ptr: AX.25 specific data
1569 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1570 *
536721b1
KK
1571 * @dev_addr: Hw address (before bcast,
1572 * because most packets are unicast)
1573 *
1574 * @_rx: Array of RX queues
1575 * @num_rx_queues: Number of RX queues
1576 * allocated at register_netdev() time
1577 * @real_num_rx_queues: Number of RX queues currently active in device
1578 *
1579 * @rx_handler: handler for received packets
1580 * @rx_handler_data: XXX: need comments on this one
46209401
JP
1581 * @miniq_ingress: ingress/clsact qdisc specific data for
1582 * ingress processing
536721b1
KK
1583 * @ingress_queue: XXX: need comments on this one
1584 * @broadcast: hw bcast address
1585 *
14ffbbb8
TG
1586 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1587 * indexed by RX queue number. Assigned by driver.
1588 * This must only be set if the ndo_rx_flow_steer
1589 * operation is defined
1590 * @index_hlist: Device index hash chain
1591 *
536721b1
KK
1592 * @_tx: Array of TX queues
1593 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1594 * @real_num_tx_queues: Number of TX queues currently active in device
1595 * @qdisc: Root qdisc from userspace point of view
1596 * @tx_queue_len: Max frames per queue allowed
1597 * @tx_global_lock: XXX: need comments on this one
1598 *
1599 * @xps_maps: XXX: need comments on this one
46209401
JP
1600 * @miniq_egress: clsact qdisc specific data for
1601 * egress processing
536721b1 1602 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1603 * the watchdog (see dev_watchdog())
536721b1
KK
1604 * @watchdog_timer: List of timers
1605 *
1606 * @pcpu_refcnt: Number of references to this device
1607 * @todo_list: Delayed register/unregister
536721b1
KK
1608 * @link_watch_list: XXX: need comments on this one
1609 *
1610 * @reg_state: Register/unregister state machine
1611 * @dismantle: Device is going to be freed
1612 * @rtnl_link_state: This enum represents the phases of creating
1613 * a new link
1614 *
cf124db5
DM
1615 * @needs_free_netdev: Should unregister perform free_netdev?
1616 * @priv_destructor: Called from unregister
536721b1
KK
1617 * @npinfo: XXX: need comments on this one
1618 * @nd_net: Network namespace this network device is inside
1619 *
1620 * @ml_priv: Mid-layer private
1621 * @lstats: Loopback statistics
1622 * @tstats: Tunnel statistics
1623 * @dstats: Dummy statistics
1624 * @vstats: Virtual ethernet statistics
1625 *
1626 * @garp_port: GARP
1627 * @mrp_port: MRP
1628 *
1629 * @dev: Class/net/name entry
1630 * @sysfs_groups: Space for optional device, statistics and wireless
1631 * sysfs groups
1632 *
1633 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1634 * @rtnl_link_ops: Rtnl_link_ops
1635 *
1636 * @gso_max_size: Maximum size of generic segmentation offload
1637 * @gso_max_segs: Maximum number of segments that can be passed to the
1638 * NIC for GSO
1639 *
1640 * @dcbnl_ops: Data Center Bridging netlink ops
1641 * @num_tc: Number of traffic classes in the net device
1642 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1643 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1644 *
1645 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1646 *
1647 * @priomap: XXX: need comments on this one
1648 * @phydev: Physical device may attach itself
1649 * for hardware timestamping
1650 *
123b3652
ED
1651 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1652 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1653 *
d746d707
AK
1654 * @proto_down: protocol port state information can be sent to the
1655 * switch driver and used to set the phys state of the
1656 * switch port.
1657 *
1da177e4
LT
1658 * FIXME: cleanup struct net_device such that network protocol info
1659 * moves out.
1660 */
1661
d94d9fee 1662struct net_device {
1da177e4 1663 char name[IFNAMSIZ];
9356b8fc 1664 struct hlist_node name_hlist;
6c557001 1665 struct dev_ifalias __rcu *ifalias;
1da177e4
LT
1666 /*
1667 * I/O specific fields
1668 * FIXME: Merge these and struct ifmap into one
1669 */
536721b1
KK
1670 unsigned long mem_end;
1671 unsigned long mem_start;
1672 unsigned long base_addr;
1673 int irq;
1da177e4
LT
1674
1675 /*
536721b1
KK
1676 * Some hardware also needs these fields (state,dev_list,
1677 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1678 * part of the usual set specified in Space.c.
1679 */
1680
1da177e4
LT
1681 unsigned long state;
1682
7562f876 1683 struct list_head dev_list;
bea3348e 1684 struct list_head napi_list;
44a0873d 1685 struct list_head unreg_list;
5cde2829 1686 struct list_head close_list;
7866a621
SN
1687 struct list_head ptype_all;
1688 struct list_head ptype_specific;
2f268f12 1689
2f268f12
VF
1690 struct {
1691 struct list_head upper;
1692 struct list_head lower;
1693 } adj_list;
1694
c8f44aff 1695 netdev_features_t features;
c8f44aff 1696 netdev_features_t hw_features;
c8f44aff 1697 netdev_features_t wanted_features;
c8f44aff 1698 netdev_features_t vlan_features;
6a674e9c 1699 netdev_features_t hw_enc_features;
0d89d203 1700 netdev_features_t mpls_features;
802ab55a 1701 netdev_features_t gso_partial_features;
04ed3e74 1702
1da177e4 1703 int ifindex;
7a66bbc9 1704 int group;
1da177e4 1705
c45d286e 1706 struct net_device_stats stats;
015f0688 1707
015f0688
ED
1708 atomic_long_t rx_dropped;
1709 atomic_long_t tx_dropped;
6e7333d3 1710 atomic_long_t rx_nohandler;
1da177e4 1711
5e7f6597
DD
1712 /* Stats to monitor link on/off, flapping */
1713 atomic_t carrier_up_count;
1714 atomic_t carrier_down_count;
1715
b86e0280 1716#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1717 const struct iw_handler_def *wireless_handlers;
1718 struct iw_public_data *wireless_data;
b86e0280 1719#endif
d314774c 1720 const struct net_device_ops *netdev_ops;
76fd8593 1721 const struct ethtool_ops *ethtool_ops;
4170604f 1722#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1723 const struct switchdev_ops *switchdev_ops;
4170604f 1724#endif
1b69c6d0
DA
1725#ifdef CONFIG_NET_L3_MASTER_DEV
1726 const struct l3mdev_ops *l3mdev_ops;
1727#endif
f997c55c
AA
1728#if IS_ENABLED(CONFIG_IPV6)
1729 const struct ndisc_ops *ndisc_ops;
1730#endif
1da177e4 1731
d77e38e6
SK
1732#ifdef CONFIG_XFRM
1733 const struct xfrmdev_ops *xfrmdev_ops;
1734#endif
1735
3b04ddde
SH
1736 const struct header_ops *header_ops;
1737
536721b1
KK
1738 unsigned int flags;
1739 unsigned int priv_flags;
1740
1da177e4 1741 unsigned short gflags;
536721b1 1742 unsigned short padded;
1da177e4 1743
536721b1
KK
1744 unsigned char operstate;
1745 unsigned char link_mode;
b00055aa 1746
536721b1
KK
1747 unsigned char if_port;
1748 unsigned char dma;
bdc220da 1749
df0f598f
ED
1750 /* Note : dev->mtu is often read without holding a lock.
1751 * Writers usually hold RTNL.
1752 * It is recommended to use READ_ONCE() to annotate the reads,
1753 * and to use WRITE_ONCE() to annotate the writes.
1754 */
536721b1 1755 unsigned int mtu;
61e84623
JW
1756 unsigned int min_mtu;
1757 unsigned int max_mtu;
536721b1
KK
1758 unsigned short type;
1759 unsigned short hard_header_len;
d92be7a4 1760 unsigned char min_header_len;
1da177e4 1761
f5184d26
JB
1762 unsigned short needed_headroom;
1763 unsigned short needed_tailroom;
1764
1da177e4 1765 /* Interface address info. */
536721b1
KK
1766 unsigned char perm_addr[MAX_ADDR_LEN];
1767 unsigned char addr_assign_type;
1768 unsigned char addr_len;
a0a9663d 1769 unsigned short neigh_priv_len;
536721b1
KK
1770 unsigned short dev_id;
1771 unsigned short dev_port;
ccffad25 1772 spinlock_t addr_list_lock;
14ffbbb8
TG
1773 unsigned char name_assign_type;
1774 bool uc_promisc;
536721b1
KK
1775 struct netdev_hw_addr_list uc;
1776 struct netdev_hw_addr_list mc;
1777 struct netdev_hw_addr_list dev_addrs;
1778
4c3d5e7b
ED
1779#ifdef CONFIG_SYSFS
1780 struct kset *queues_kset;
1781#endif
9d45abe1
WC
1782 unsigned int promiscuity;
1783 unsigned int allmulti;
1da177e4 1784
1da177e4 1785
5e82b4b2 1786 /* Protocol-specific pointers */
65ac6a5f 1787
d11ead75 1788#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1789 struct vlan_info __rcu *vlan_info;
65ac6a5f 1790#endif
34a430d7 1791#if IS_ENABLED(CONFIG_NET_DSA)
2f657a60 1792 struct dsa_port *dsa_ptr;
37cb0620
YX
1793#endif
1794#if IS_ENABLED(CONFIG_TIPC)
536721b1 1795 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1796#endif
536721b1
KK
1797 void *atalk_ptr;
1798 struct in_device __rcu *ip_ptr;
1799 struct dn_dev __rcu *dn_ptr;
1800 struct inet6_dev __rcu *ip6_ptr;
1801 void *ax25_ptr;
1802 struct wireless_dev *ieee80211_ptr;
98a18b6f 1803 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1804#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1805 struct mpls_dev __rcu *mpls_ptr;
1806#endif
1da177e4 1807
9356b8fc 1808/*
cd13539b 1809 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1810 */
9356b8fc 1811 /* Interface address info used in eth_type_trans() */
536721b1 1812 unsigned char *dev_addr;
f001fde5 1813
a953be53 1814#ifdef CONFIG_SYSFS
0a9627f2
TH
1815 struct netdev_rx_queue *_rx;
1816
0a9627f2 1817 unsigned int num_rx_queues;
62fe0b40 1818 unsigned int real_num_rx_queues;
df334545 1819#endif
0a9627f2 1820
7acedaf5 1821 struct bpf_prog __rcu *xdp_prog;
3b47d303 1822 unsigned long gro_flush_timeout;
61391cde 1823 rx_handler_func_t __rcu *rx_handler;
1824 void __rcu *rx_handler_data;
e8a0464c 1825
4cda01e8 1826#ifdef CONFIG_NET_CLS_ACT
46209401 1827 struct mini_Qdisc __rcu *miniq_ingress;
d2788d34 1828#endif
24824a09 1829 struct netdev_queue __rcu *ingress_queue;
e687ad60 1830#ifdef CONFIG_NETFILTER_INGRESS
960632ec 1831 struct nf_hook_entries __rcu *nf_hooks_ingress;
e687ad60 1832#endif
d2788d34 1833
536721b1 1834 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1835#ifdef CONFIG_RFS_ACCEL
1836 struct cpu_rmap *rx_cpu_rmap;
1837#endif
1838 struct hlist_node index_hlist;
cd13539b
ED
1839
1840/*
1841 * Cache lines mostly used on transmit path
1842 */
e8a0464c
DM
1843 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1844 unsigned int num_tx_queues;
fd2ea0a7 1845 unsigned int real_num_tx_queues;
af356afa 1846 struct Qdisc *qdisc;
59cc1f61
JK
1847#ifdef CONFIG_NET_SCHED
1848 DECLARE_HASHTABLE (qdisc_hash, 4);
1849#endif
0cd29503 1850 unsigned int tx_queue_len;
c3f26a26 1851 spinlock_t tx_global_lock;
14ffbbb8 1852 int watchdog_timeo;
cd13539b 1853
bf264145 1854#ifdef CONFIG_XPS
a4177869 1855 struct xps_dev_maps __rcu *xps_maps;
bf264145 1856#endif
1f211a1b 1857#ifdef CONFIG_NET_CLS_ACT
46209401 1858 struct mini_Qdisc __rcu *miniq_egress;
1f211a1b 1859#endif
0c4f691f 1860
9356b8fc 1861 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1862 struct timer_list watchdog_timer;
1863
29b4433d 1864 int __percpu *pcpu_refcnt;
1da177e4 1865 struct list_head todo_list;
1da177e4 1866
e014debe 1867 struct list_head link_watch_list;
572a103d 1868
1da177e4 1869 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1870 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1871 NETREG_UNREGISTERING, /* called unregister_netdevice */
1872 NETREG_UNREGISTERED, /* completed unregister todo */
1873 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1874 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1875 } reg_state:8;
1876
536721b1 1877 bool dismantle;
a2835763
PM
1878
1879 enum {
1880 RTNL_LINK_INITIALIZED,
1881 RTNL_LINK_INITIALIZING,
1882 } rtnl_link_state:16;
1da177e4 1883
cf124db5
DM
1884 bool needs_free_netdev;
1885 void (*priv_destructor)(struct net_device *dev);
1da177e4 1886
1da177e4 1887#ifdef CONFIG_NETPOLL
5fbee843 1888 struct netpoll_info __rcu *npinfo;
1da177e4 1889#endif
eae792b7 1890
0c5c9fb5 1891 possible_net_t nd_net;
4a1c5371 1892
4951704b 1893 /* mid-layer private */
a7855c78 1894 union {
536721b1
KK
1895 void *ml_priv;
1896 struct pcpu_lstats __percpu *lstats;
8f84985f 1897 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1898 struct pcpu_dstats __percpu *dstats;
1899 struct pcpu_vstats __percpu *vstats;
a7855c78 1900 };
536721b1 1901
fb585b44 1902#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1903 struct garp_port __rcu *garp_port;
fb585b44
TK
1904#endif
1905#if IS_ENABLED(CONFIG_MRP)
febf018d 1906 struct mrp_port __rcu *mrp_port;
fb585b44 1907#endif
1da177e4 1908
5e82b4b2 1909 struct device dev;
0c509a6c 1910 const struct attribute_group *sysfs_groups[4];
a953be53 1911 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1912
38f7b870 1913 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1914
82cc1a7a
PWJ
1915 /* for setting kernel sock attribute on TCP connection setup */
1916#define GSO_MAX_SIZE 65536
1917 unsigned int gso_max_size;
30b678d8
BH
1918#define GSO_MAX_SEGS 65535
1919 u16 gso_max_segs;
743b03a8 1920
7a6b6f51 1921#ifdef CONFIG_DCB
32953543 1922 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1923#endif
5e82b4b2
BH
1924 u8 num_tc;
1925 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1926 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1927
d11ead75 1928#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1929 unsigned int fcoe_ddp_xid;
5bc1421e 1930#endif
86f8515f 1931#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1932 struct netprio_map __rcu *priomap;
4d288d57 1933#endif
5e82b4b2
BH
1934 struct phy_device *phydev;
1935 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1936 struct lock_class_key *qdisc_running_key;
5e82b4b2 1937 bool proto_down;
1da177e4 1938};
43cb76d9 1939#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1940
b5cdae32
DM
1941static inline bool netif_elide_gro(const struct net_device *dev)
1942{
1943 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1944 return true;
1945 return false;
1946}
1947
1da177e4 1948#define NETDEV_ALIGN 32
1da177e4 1949
4f57c087
JF
1950static inline
1951int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1952{
1953 return dev->prio_tc_map[prio & TC_BITMASK];
1954}
1955
1956static inline
1957int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1958{
1959 if (tc >= dev->num_tc)
1960 return -EINVAL;
1961
1962 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1963 return 0;
1964}
1965
8d059b0f 1966int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1967void netdev_reset_tc(struct net_device *dev);
1968int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1969int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1970
1971static inline
1972int netdev_get_num_tc(struct net_device *dev)
1973{
1974 return dev->num_tc;
1975}
1976
e8a0464c
DM
1977static inline
1978struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1979 unsigned int index)
1980{
1981 return &dev->_tx[index];
1982}
1983
10c51b56
DB
1984static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1985 const struct sk_buff *skb)
1986{
1987 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1988}
1989
e8a0464c
DM
1990static inline void netdev_for_each_tx_queue(struct net_device *dev,
1991 void (*f)(struct net_device *,
1992 struct netdev_queue *,
1993 void *),
1994 void *arg)
1995{
1996 unsigned int i;
1997
1998 for (i = 0; i < dev->num_tx_queues; i++)
1999 f(dev, &dev->_tx[i], arg);
2000}
2001
d3fff6c4
ED
2002#define netdev_lockdep_set_classes(dev) \
2003{ \
2004 static struct lock_class_key qdisc_tx_busylock_key; \
2005 static struct lock_class_key qdisc_running_key; \
2006 static struct lock_class_key qdisc_xmit_lock_key; \
2007 static struct lock_class_key dev_addr_list_lock_key; \
2008 unsigned int i; \
2009 \
2010 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2011 (dev)->qdisc_running_key = &qdisc_running_key; \
2012 lockdep_set_class(&(dev)->addr_list_lock, \
2013 &dev_addr_list_lock_key); \
2014 for (i = 0; i < (dev)->num_tx_queues; i++) \
2015 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2016 &qdisc_xmit_lock_key); \
2017}
2018
f629d208 2019struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
2020 struct sk_buff *skb,
2021 void *accel_priv);
8c4c49df 2022
871b642a
PA
2023/* returns the headroom that the master device needs to take in account
2024 * when forwarding to this dev
2025 */
2026static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2027{
2028 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2029}
2030
2031static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2032{
2033 if (dev->netdev_ops->ndo_set_rx_headroom)
2034 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2035}
2036
2037/* set the device rx headroom to the dev's default */
2038static inline void netdev_reset_rx_headroom(struct net_device *dev)
2039{
2040 netdev_set_rx_headroom(dev, -1);
2041}
2042
c346dca1
YH
2043/*
2044 * Net namespace inlines
2045 */
2046static inline
2047struct net *dev_net(const struct net_device *dev)
2048{
c2d9ba9b 2049 return read_pnet(&dev->nd_net);
c346dca1
YH
2050}
2051
2052static inline
f5aa23fd 2053void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2054{
0c5c9fb5 2055 write_pnet(&dev->nd_net, net);
c346dca1
YH
2056}
2057
bea3348e
SH
2058/**
2059 * netdev_priv - access network device private data
2060 * @dev: network device
2061 *
2062 * Get network device private data
2063 */
6472ce60 2064static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2065{
1ce8e7b5 2066 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2067}
2068
1da177e4
LT
2069/* Set the sysfs physical device reference for the network logical device
2070 * if set prior to registration will cause a symlink during initialization.
2071 */
43cb76d9 2072#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2073
384912ed 2074/* Set the sysfs device type for the network logical device to allow
3f79410c 2075 * fine-grained identification of different network device types. For
5e82b4b2 2076 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2077 */
2078#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2079
82dc3c63
ED
2080/* Default NAPI poll() weight
2081 * Device drivers are strongly advised to not use bigger value
2082 */
2083#define NAPI_POLL_WEIGHT 64
2084
3b582cc1 2085/**
5e82b4b2 2086 * netif_napi_add - initialize a NAPI context
3b582cc1 2087 * @dev: network device
5e82b4b2 2088 * @napi: NAPI context
3b582cc1
SH
2089 * @poll: polling function
2090 * @weight: default weight
2091 *
5e82b4b2
BH
2092 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2093 * *any* of the other NAPI-related functions.
3b582cc1 2094 */
d565b0a1
HX
2095void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2096 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2097
d64b5e85 2098/**
5e82b4b2 2099 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2100 * @dev: network device
5e82b4b2 2101 * @napi: NAPI context
d64b5e85
ED
2102 * @poll: polling function
2103 * @weight: default weight
2104 *
2105 * This variant of netif_napi_add() should be used from drivers using NAPI
2106 * to exclusively poll a TX queue.
2107 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2108 */
2109static inline void netif_tx_napi_add(struct net_device *dev,
2110 struct napi_struct *napi,
2111 int (*poll)(struct napi_struct *, int),
2112 int weight)
2113{
2114 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2115 netif_napi_add(dev, napi, poll, weight);
2116}
2117
d8156534 2118/**
5e82b4b2
BH
2119 * netif_napi_del - remove a NAPI context
2120 * @napi: NAPI context
d8156534 2121 *
5e82b4b2 2122 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2123 */
d565b0a1
HX
2124void netif_napi_del(struct napi_struct *napi);
2125
2126struct napi_gro_cb {
78a478d0 2127 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2128 void *frag0;
78a478d0 2129
7489594c
HX
2130 /* Length of frag0. */
2131 unsigned int frag0_len;
2132
86911732 2133 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2134 int data_offset;
86911732 2135
d565b0a1 2136 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2137 u16 flush;
2138
2139 /* Save the IP ID here and check when we get to the transport layer */
2140 u16 flush_id;
d565b0a1
HX
2141
2142 /* Number of segments aggregated. */
2e71a6f8
ED
2143 u16 count;
2144
15e2396d
TH
2145 /* Start offset for remote checksum offload */
2146 u16 gro_remcsum_start;
2147
2e71a6f8
ED
2148 /* jiffies when first packet was created/queued */
2149 unsigned long age;
86347245 2150
afe93325 2151 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2152 u16 proto;
2153
baa32ff4
TH
2154 /* This is non-zero if the packet may be of the same flow. */
2155 u8 same_flow:1;
2156
fac8e0f5
JG
2157 /* Used in tunnel GRO receive */
2158 u8 encap_mark:1;
573e8fca
TH
2159
2160 /* GRO checksum is valid */
2161 u8 csum_valid:1;
2162
662880f4
TH
2163 /* Number of checksums via CHECKSUM_UNNECESSARY */
2164 u8 csum_cnt:3;
c3c7c254 2165
baa32ff4
TH
2166 /* Free the skb? */
2167 u8 free:2;
2168#define NAPI_GRO_FREE 1
2169#define NAPI_GRO_FREE_STOLEN_HEAD 2
2170
efc98d08
TH
2171 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2172 u8 is_ipv6:1;
2173
a0ca153f
AD
2174 /* Used in GRE, set in fou/gue_gro_receive */
2175 u8 is_fou:1;
2176
1530545e
AD
2177 /* Used to determine if flush_id can be ignored */
2178 u8 is_atomic:1;
2179
fcd91dd4
SD
2180 /* Number of gro_receive callbacks this packet already went through */
2181 u8 recursion_counter:4;
2182
2183 /* 1 bit hole */
baa32ff4 2184
bf5a755f
JC
2185 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2186 __wsum csum;
2187
c3c7c254
ED
2188 /* used in skb_gro_receive() slow path */
2189 struct sk_buff *last;
d565b0a1
HX
2190};
2191
2192#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2193
fcd91dd4
SD
2194#define GRO_RECURSION_LIMIT 15
2195static inline int gro_recursion_inc_test(struct sk_buff *skb)
2196{
2197 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2198}
2199
2200typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2201static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2202 struct sk_buff **head,
2203 struct sk_buff *skb)
2204{
2205 if (unlikely(gro_recursion_inc_test(skb))) {
2206 NAPI_GRO_CB(skb)->flush |= 1;
2207 return NULL;
2208 }
2209
2210 return cb(head, skb);
2211}
2212
2213typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2214 struct sk_buff *);
2215static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2216 struct sock *sk,
2217 struct sk_buff **head,
2218 struct sk_buff *skb)
2219{
2220 if (unlikely(gro_recursion_inc_test(skb))) {
2221 NAPI_GRO_CB(skb)->flush |= 1;
2222 return NULL;
2223 }
2224
2225 return cb(sk, head, skb);
2226}
2227
1da177e4 2228struct packet_type {
f2ccd8fa
DM
2229 __be16 type; /* This is really htons(ether_type). */
2230 struct net_device *dev; /* NULL is wildcarded here */
2231 int (*func) (struct sk_buff *,
2232 struct net_device *,
2233 struct packet_type *,
2234 struct net_device *);
c0de08d0
EL
2235 bool (*id_match)(struct packet_type *ptype,
2236 struct sock *sk);
1da177e4
LT
2237 void *af_packet_priv;
2238 struct list_head list;
2239};
2240
f191a1d1 2241struct offload_callbacks {
576a30eb 2242 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2243 netdev_features_t features);
d565b0a1 2244 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2245 struct sk_buff *skb);
299603e8 2246 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2247};
2248
2249struct packet_offload {
2250 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2251 u16 priority;
f191a1d1
VY
2252 struct offload_callbacks callbacks;
2253 struct list_head list;
1da177e4
LT
2254};
2255
5e82b4b2 2256/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2257struct pcpu_sw_netstats {
2258 u64 rx_packets;
2259 u64 rx_bytes;
2260 u64 tx_packets;
2261 u64 tx_bytes;
2262 struct u64_stats_sync syncp;
2263};
2264
aabc92bb
PNA
2265#define __netdev_alloc_pcpu_stats(type, gfp) \
2266({ \
2267 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2268 if (pcpu_stats) { \
2269 int __cpu; \
2270 for_each_possible_cpu(__cpu) { \
2271 typeof(type) *stat; \
2272 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2273 u64_stats_init(&stat->syncp); \
2274 } \
2275 } \
2276 pcpu_stats; \
1c213bd2
WC
2277})
2278
aabc92bb 2279#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2280 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2281
764f5e54
JP
2282enum netdev_lag_tx_type {
2283 NETDEV_LAG_TX_TYPE_UNKNOWN,
2284 NETDEV_LAG_TX_TYPE_RANDOM,
2285 NETDEV_LAG_TX_TYPE_BROADCAST,
2286 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2287 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2288 NETDEV_LAG_TX_TYPE_HASH,
2289};
2290
2291struct netdev_lag_upper_info {
2292 enum netdev_lag_tx_type tx_type;
2293};
2294
fb1b2e3c
JP
2295struct netdev_lag_lower_state_info {
2296 u8 link_up : 1,
2297 tx_enabled : 1;
2298};
2299
1da177e4
LT
2300#include <linux/notifier.h>
2301
dcfe1421
AW
2302/* netdevice notifier chain. Please remember to update the rtnetlink
2303 * notification exclusion list in rtnetlink_event() when adding new
2304 * types.
2305 */
2306#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2307#define NETDEV_DOWN 0x0002
2308#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2309 detected a hardware crash and restarted
2310 - we can use this eg to kick tcp sessions
2311 once done */
2312#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2313#define NETDEV_REGISTER 0x0005
2314#define NETDEV_UNREGISTER 0x0006
1d486bfb 2315#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2316#define NETDEV_CHANGEADDR 0x0008
2317#define NETDEV_GOING_DOWN 0x0009
2318#define NETDEV_CHANGENAME 0x000A
2319#define NETDEV_FEAT_CHANGE 0x000B
2320#define NETDEV_BONDING_FAILOVER 0x000C
2321#define NETDEV_PRE_UP 0x000D
2322#define NETDEV_PRE_TYPE_CHANGE 0x000E
2323#define NETDEV_POST_TYPE_CHANGE 0x000F
2324#define NETDEV_POST_INIT 0x0010
0115e8e3 2325#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2326#define NETDEV_RELEASE 0x0012
2327#define NETDEV_NOTIFY_PEERS 0x0013
2328#define NETDEV_JOIN 0x0014
42e52bf9 2329#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2330#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2331#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2332#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2333#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2334#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2335#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2336#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
296d8ee3 2337#define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
08294a26 2338#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2339
f629d208
JP
2340int register_netdevice_notifier(struct notifier_block *nb);
2341int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2342
2343struct netdev_notifier_info {
51d0c047
DA
2344 struct net_device *dev;
2345 struct netlink_ext_ack *extack;
351638e7
JP
2346};
2347
63419c91
SD
2348struct netdev_notifier_info_ext {
2349 struct netdev_notifier_info info; /* must be first */
2350 union {
2351 u32 mtu;
2352 } ext;
2353};
2354
be9efd36
JP
2355struct netdev_notifier_change_info {
2356 struct netdev_notifier_info info; /* must be first */
2357 unsigned int flags_changed;
2358};
2359
0e4ead9d
JP
2360struct netdev_notifier_changeupper_info {
2361 struct netdev_notifier_info info; /* must be first */
2362 struct net_device *upper_dev; /* new upper dev */
2363 bool master; /* is upper dev master */
5e82b4b2 2364 bool linking; /* is the notification for link or unlink */
29bf24af 2365 void *upper_info; /* upper dev info */
0e4ead9d
JP
2366};
2367
04d48266
JP
2368struct netdev_notifier_changelowerstate_info {
2369 struct netdev_notifier_info info; /* must be first */
2370 void *lower_state_info; /* is lower dev state */
2371};
2372
75538c2b
CW
2373static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2374 struct net_device *dev)
2375{
2376 info->dev = dev;
51d0c047 2377 info->extack = NULL;
75538c2b
CW
2378}
2379
351638e7
JP
2380static inline struct net_device *
2381netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2382{
2383 return info->dev;
2384}
2385
51d0c047
DA
2386static inline struct netlink_ext_ack *
2387netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2388{
2389 return info->extack;
2390}
2391
f629d208 2392int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2393
2394
1da177e4
LT
2395extern rwlock_t dev_base_lock; /* Device list lock */
2396
881d966b
EB
2397#define for_each_netdev(net, d) \
2398 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2399#define for_each_netdev_reverse(net, d) \
2400 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2401#define for_each_netdev_rcu(net, d) \
2402 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2403#define for_each_netdev_safe(net, d, n) \
2404 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2405#define for_each_netdev_continue(net, d) \
2406 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2407#define for_each_netdev_continue_rcu(net, d) \
2408 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2409#define for_each_netdev_in_bond_rcu(bond, slave) \
2410 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2411 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2412#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2413
a050c33f
DL
2414static inline struct net_device *next_net_device(struct net_device *dev)
2415{
2416 struct list_head *lh;
2417 struct net *net;
2418
c346dca1 2419 net = dev_net(dev);
a050c33f
DL
2420 lh = dev->dev_list.next;
2421 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2422}
2423
ce81b76a
ED
2424static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2425{
2426 struct list_head *lh;
2427 struct net *net;
2428
2429 net = dev_net(dev);
ccf43438 2430 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2431 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2432}
2433
a050c33f
DL
2434static inline struct net_device *first_net_device(struct net *net)
2435{
2436 return list_empty(&net->dev_base_head) ? NULL :
2437 net_device_entry(net->dev_base_head.next);
2438}
7562f876 2439
ccf43438
ED
2440static inline struct net_device *first_net_device_rcu(struct net *net)
2441{
2442 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2443
2444 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2445}
2446
f629d208
JP
2447int netdev_boot_setup_check(struct net_device *dev);
2448unsigned long netdev_boot_base(const char *prefix, int unit);
2449struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2450 const char *hwaddr);
2451struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2452struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2453void dev_add_pack(struct packet_type *pt);
2454void dev_remove_pack(struct packet_type *pt);
2455void __dev_remove_pack(struct packet_type *pt);
2456void dev_add_offload(struct packet_offload *po);
2457void dev_remove_offload(struct packet_offload *po);
f629d208 2458
a54acb3a 2459int dev_get_iflink(const struct net_device *dev);
fc4099f1 2460int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2461struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2462 unsigned short mask);
f629d208
JP
2463struct net_device *dev_get_by_name(struct net *net, const char *name);
2464struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2465struct net_device *__dev_get_by_name(struct net *net, const char *name);
2466int dev_alloc_name(struct net_device *dev, const char *name);
2467int dev_open(struct net_device *dev);
7051b88a 2468void dev_close(struct net_device *dev);
2469void dev_close_many(struct list_head *head, bool unlink);
f629d208 2470void dev_disable_lro(struct net_device *dev);
0c4b51f0 2471int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2472int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2473int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2474int register_netdevice(struct net_device *dev);
2475void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2476void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2477static inline void unregister_netdevice(struct net_device *dev)
2478{
2479 unregister_netdevice_queue(dev, NULL);
2480}
2481
f629d208
JP
2482int netdev_refcnt_read(const struct net_device *dev);
2483void free_netdev(struct net_device *dev);
74d332c1 2484void netdev_freemem(struct net_device *dev);
f629d208
JP
2485void synchronize_net(void);
2486int init_dummy_netdev(struct net_device *dev);
937f1ba5 2487
f60e5990 2488DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2489#define XMIT_RECURSION_LIMIT 10
2490
f60e5990 2491static inline int dev_recursion_level(void)
2492{
2493 return this_cpu_read(xmit_recursion);
2494}
2495
f629d208
JP
2496struct net_device *dev_get_by_index(struct net *net, int ifindex);
2497struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2498struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2499struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2500int netdev_get_name(struct net *net, char *name, int ifindex);
2501int dev_restart(struct net_device *dev);
f629d208 2502int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2503
2504static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2505{
2506 return NAPI_GRO_CB(skb)->data_offset;
2507}
2508
2509static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2510{
2511 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2512}
2513
2514static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2515{
2516 NAPI_GRO_CB(skb)->data_offset += len;
2517}
2518
a5b1cf28
HX
2519static inline void *skb_gro_header_fast(struct sk_buff *skb,
2520 unsigned int offset)
86911732 2521{
a5b1cf28
HX
2522 return NAPI_GRO_CB(skb)->frag0 + offset;
2523}
78a478d0 2524
a5b1cf28
HX
2525static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2526{
2527 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2528}
78a478d0 2529
57ea52a8
HX
2530static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2531{
2532 NAPI_GRO_CB(skb)->frag0 = NULL;
2533 NAPI_GRO_CB(skb)->frag0_len = 0;
2534}
2535
a5b1cf28
HX
2536static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2537 unsigned int offset)
2538{
17dd759c
HX
2539 if (!pskb_may_pull(skb, hlen))
2540 return NULL;
2541
57ea52a8 2542 skb_gro_frag0_invalidate(skb);
17dd759c 2543 return skb->data + offset;
86911732 2544}
1da177e4 2545
36e7b1b8
HX
2546static inline void *skb_gro_network_header(struct sk_buff *skb)
2547{
78d3fd0b
HX
2548 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2549 skb_network_offset(skb);
36e7b1b8
HX
2550}
2551
bf5a755f
JC
2552static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2553 const void *start, unsigned int len)
2554{
573e8fca 2555 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2556 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2557 csum_partial(start, len, 0));
2558}
2559
573e8fca
TH
2560/* GRO checksum functions. These are logical equivalents of the normal
2561 * checksum functions (in skbuff.h) except that they operate on the GRO
2562 * offsets and fields in sk_buff.
2563 */
2564
2565__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2566
15e2396d
TH
2567static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2568{
b7fe10e5 2569 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2570}
2571
573e8fca
TH
2572static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2573 bool zero_okay,
2574 __sum16 check)
2575{
6edec0e6
TH
2576 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2577 skb_checksum_start_offset(skb) <
2578 skb_gro_offset(skb)) &&
15e2396d 2579 !skb_at_gro_remcsum_start(skb) &&
662880f4 2580 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2581 (!zero_okay || check));
2582}
2583
2584static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2585 __wsum psum)
2586{
2587 if (NAPI_GRO_CB(skb)->csum_valid &&
2588 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2589 return 0;
2590
2591 NAPI_GRO_CB(skb)->csum = psum;
2592
2593 return __skb_gro_checksum_complete(skb);
2594}
2595
573e8fca
TH
2596static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2597{
662880f4
TH
2598 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2599 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2600 NAPI_GRO_CB(skb)->csum_cnt--;
2601 } else {
2602 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2603 * verified a new top level checksum or an encapsulated one
2604 * during GRO. This saves work if we fallback to normal path.
2605 */
2606 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2607 }
2608}
2609
2610#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2611 compute_pseudo) \
2612({ \
2613 __sum16 __ret = 0; \
2614 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2615 __ret = __skb_gro_checksum_validate_complete(skb, \
2616 compute_pseudo(skb, proto)); \
219f1d79 2617 if (!__ret) \
573e8fca
TH
2618 skb_gro_incr_csum_unnecessary(skb); \
2619 __ret; \
2620})
2621
2622#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2623 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2624
2625#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2626 compute_pseudo) \
2627 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2628
2629#define skb_gro_checksum_simple_validate(skb) \
2630 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2631
d96535a1
TH
2632static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2633{
2634 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2635 !NAPI_GRO_CB(skb)->csum_valid);
2636}
2637
2638static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2639 __sum16 check, __wsum pseudo)
2640{
2641 NAPI_GRO_CB(skb)->csum = ~pseudo;
2642 NAPI_GRO_CB(skb)->csum_valid = 1;
2643}
2644
2645#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2646do { \
2647 if (__skb_gro_checksum_convert_check(skb)) \
2648 __skb_gro_checksum_convert(skb, check, \
2649 compute_pseudo(skb, proto)); \
2650} while (0)
2651
26c4f7da
TH
2652struct gro_remcsum {
2653 int offset;
2654 __wsum delta;
2655};
2656
2657static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2658{
846cd667 2659 grc->offset = 0;
26c4f7da
TH
2660 grc->delta = 0;
2661}
2662
b7fe10e5
TH
2663static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2664 unsigned int off, size_t hdrlen,
2665 int start, int offset,
2666 struct gro_remcsum *grc,
2667 bool nopartial)
dcdc8994
TH
2668{
2669 __wsum delta;
b7fe10e5 2670 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2671
2672 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2673
15e2396d 2674 if (!nopartial) {
b7fe10e5
TH
2675 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2676 return ptr;
2677 }
2678
2679 ptr = skb_gro_header_fast(skb, off);
2680 if (skb_gro_header_hard(skb, off + plen)) {
2681 ptr = skb_gro_header_slow(skb, off + plen, off);
2682 if (!ptr)
2683 return NULL;
15e2396d
TH
2684 }
2685
b7fe10e5
TH
2686 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2687 start, offset);
dcdc8994
TH
2688
2689 /* Adjust skb->csum since we changed the packet */
dcdc8994 2690 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2691
b7fe10e5 2692 grc->offset = off + hdrlen + offset;
26c4f7da 2693 grc->delta = delta;
b7fe10e5
TH
2694
2695 return ptr;
dcdc8994
TH
2696}
2697
26c4f7da
TH
2698static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2699 struct gro_remcsum *grc)
2700{
b7fe10e5
TH
2701 void *ptr;
2702 size_t plen = grc->offset + sizeof(u16);
2703
26c4f7da
TH
2704 if (!grc->delta)
2705 return;
2706
b7fe10e5
TH
2707 ptr = skb_gro_header_fast(skb, grc->offset);
2708 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2709 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2710 if (!ptr)
2711 return;
2712 }
2713
2714 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2715}
dcdc8994 2716
25393d3f
SK
2717#ifdef CONFIG_XFRM_OFFLOAD
2718static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2719{
2720 if (PTR_ERR(pp) != -EINPROGRESS)
2721 NAPI_GRO_CB(skb)->flush |= flush;
2722}
12df8689
SD
2723static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2724 struct sk_buff **pp,
2725 int flush,
2726 struct gro_remcsum *grc)
2727{
2728 if (PTR_ERR(pp) != -EINPROGRESS) {
2729 NAPI_GRO_CB(skb)->flush |= flush;
2730 skb_gro_remcsum_cleanup(skb, grc);
2731 skb->remcsum_offload = 0;
2732 }
2733}
25393d3f 2734#else
5f114163
SK
2735static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2736{
2737 NAPI_GRO_CB(skb)->flush |= flush;
2738}
12df8689
SD
2739static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2740 struct sk_buff **pp,
2741 int flush,
2742 struct gro_remcsum *grc)
2743{
2744 NAPI_GRO_CB(skb)->flush |= flush;
2745 skb_gro_remcsum_cleanup(skb, grc);
2746 skb->remcsum_offload = 0;
2747}
25393d3f 2748#endif
5f114163 2749
0c4e8581
SH
2750static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2751 unsigned short type,
3b04ddde 2752 const void *daddr, const void *saddr,
95c96174 2753 unsigned int len)
0c4e8581 2754{
f1ecfd5d 2755 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2756 return 0;
3b04ddde
SH
2757
2758 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2759}
2760
b95cce35
SH
2761static inline int dev_parse_header(const struct sk_buff *skb,
2762 unsigned char *haddr)
2763{
2764 const struct net_device *dev = skb->dev;
2765
1b83336b 2766 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2767 return 0;
3b04ddde 2768 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2769}
2770
2793a23a
WB
2771/* ll_header must have at least hard_header_len allocated */
2772static inline bool dev_validate_header(const struct net_device *dev,
2773 char *ll_header, int len)
2774{
2775 if (likely(len >= dev->hard_header_len))
2776 return true;
217e6fa2
WB
2777 if (len < dev->min_header_len)
2778 return false;
2793a23a
WB
2779
2780 if (capable(CAP_SYS_RAWIO)) {
2781 memset(ll_header + len, 0, dev->hard_header_len - len);
2782 return true;
2783 }
2784
2785 if (dev->header_ops && dev->header_ops->validate)
2786 return dev->header_ops->validate(ll_header, len);
2787
2788 return false;
2789}
2790
1da177e4 2791typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2792int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2793static inline int unregister_gifconf(unsigned int family)
2794{
2795 return register_gifconf(family, NULL);
2796}
2797
99bbc707 2798#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2799#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2800struct sd_flow_limit {
2801 u64 count;
2802 unsigned int num_buckets;
2803 unsigned int history_head;
2804 u16 history[FLOW_LIMIT_HISTORY];
2805 u8 buckets[];
2806};
2807
2808extern int netdev_flow_limit_table_len;
2809#endif /* CONFIG_NET_FLOW_LIMIT */
2810
1da177e4 2811/*
5e82b4b2 2812 * Incoming packets are placed on per-CPU queues
1da177e4 2813 */
d94d9fee 2814struct softnet_data {
1da177e4 2815 struct list_head poll_list;
6e7676c1 2816 struct sk_buff_head process_queue;
1da177e4 2817
dee42870 2818 /* stats */
cd7b5396
DM
2819 unsigned int processed;
2820 unsigned int time_squeeze;
cd7b5396 2821 unsigned int received_rps;
fd793d89 2822#ifdef CONFIG_RPS
88751275 2823 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2824#endif
2825#ifdef CONFIG_NET_FLOW_LIMIT
2826 struct sd_flow_limit __rcu *flow_limit;
2827#endif
2828 struct Qdisc *output_queue;
2829 struct Qdisc **output_queue_tailp;
2830 struct sk_buff *completion_queue;
88751275 2831
4cdb1e2e 2832#ifdef CONFIG_RPS
501e7ef5
ED
2833 /* input_queue_head should be written by cpu owning this struct,
2834 * and only read by other cpus. Worth using a cache line.
2835 */
2836 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2837
2838 /* Elements below can be accessed between CPUs for RPS/RFS */
966a9671 2839 call_single_data_t csd ____cacheline_aligned_in_smp;
88751275
ED
2840 struct softnet_data *rps_ipi_next;
2841 unsigned int cpu;
76cc8b13 2842 unsigned int input_queue_tail;
1e94d72f 2843#endif
95c96174 2844 unsigned int dropped;
0a9627f2 2845 struct sk_buff_head input_pkt_queue;
bea3348e 2846 struct napi_struct backlog;
99bbc707 2847
1da177e4
LT
2848};
2849
76cc8b13 2850static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2851{
2852#ifdef CONFIG_RPS
76cc8b13
TH
2853 sd->input_queue_head++;
2854#endif
2855}
2856
2857static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2858 unsigned int *qtail)
2859{
2860#ifdef CONFIG_RPS
2861 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2862#endif
2863}
2864
0a9627f2 2865DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2866
f629d208 2867void __netif_schedule(struct Qdisc *q);
46e5da40 2868void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2869
fd2ea0a7
DM
2870static inline void netif_tx_schedule_all(struct net_device *dev)
2871{
2872 unsigned int i;
2873
2874 for (i = 0; i < dev->num_tx_queues; i++)
2875 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2876}
2877
f9a7cbbf 2878static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2879{
73466498 2880 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2881}
2882
bea3348e
SH
2883/**
2884 * netif_start_queue - allow transmit
2885 * @dev: network device
2886 *
2887 * Allow upper layers to call the device hard_start_xmit routine.
2888 */
1da177e4
LT
2889static inline void netif_start_queue(struct net_device *dev)
2890{
e8a0464c 2891 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2892}
2893
fd2ea0a7
DM
2894static inline void netif_tx_start_all_queues(struct net_device *dev)
2895{
2896 unsigned int i;
2897
2898 for (i = 0; i < dev->num_tx_queues; i++) {
2899 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2900 netif_tx_start_queue(txq);
2901 }
2902}
2903
46e5da40 2904void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2905
d29f749e
DJ
2906/**
2907 * netif_wake_queue - restart transmit
2908 * @dev: network device
2909 *
2910 * Allow upper layers to call the device hard_start_xmit routine.
2911 * Used for flow control when transmit resources are available.
2912 */
79d16385
DM
2913static inline void netif_wake_queue(struct net_device *dev)
2914{
e8a0464c 2915 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2916}
2917
fd2ea0a7
DM
2918static inline void netif_tx_wake_all_queues(struct net_device *dev)
2919{
2920 unsigned int i;
2921
2922 for (i = 0; i < dev->num_tx_queues; i++) {
2923 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2924 netif_tx_wake_queue(txq);
2925 }
2926}
2927
f9a7cbbf 2928static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2929{
73466498 2930 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2931}
2932
bea3348e
SH
2933/**
2934 * netif_stop_queue - stop transmitted packets
2935 * @dev: network device
2936 *
2937 * Stop upper layers calling the device hard_start_xmit routine.
2938 * Used for flow control when transmit resources are unavailable.
2939 */
1da177e4
LT
2940static inline void netif_stop_queue(struct net_device *dev)
2941{
e8a0464c 2942 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2943}
2944
a2029240 2945void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2946
4d29515f 2947static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2948{
73466498 2949 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2950}
2951
bea3348e
SH
2952/**
2953 * netif_queue_stopped - test if transmit queue is flowblocked
2954 * @dev: network device
2955 *
2956 * Test if transmit queue on device is currently unable to send.
2957 */
4d29515f 2958static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2959{
e8a0464c 2960 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2961}
2962
4d29515f 2963static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2964{
73466498
TH
2965 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2966}
2967
8e2f1a63
DB
2968static inline bool
2969netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2970{
2971 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2972}
2973
8e2f1a63
DB
2974static inline bool
2975netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2976{
2977 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2978}
2979
53511453
ED
2980/**
2981 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2982 * @dev_queue: pointer to transmit queue
2983 *
2984 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2985 * to give appropriate hint to the CPU.
53511453
ED
2986 */
2987static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2988{
2989#ifdef CONFIG_BQL
2990 prefetchw(&dev_queue->dql.num_queued);
2991#endif
2992}
2993
2994/**
2995 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2996 * @dev_queue: pointer to transmit queue
2997 *
2998 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2999 * to give appropriate hint to the CPU.
53511453
ED
3000 */
3001static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3002{
3003#ifdef CONFIG_BQL
3004 prefetchw(&dev_queue->dql.limit);
3005#endif
3006}
3007
c5d67bd7
TH
3008static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3009 unsigned int bytes)
3010{
114cf580
TH
3011#ifdef CONFIG_BQL
3012 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
3013
3014 if (likely(dql_avail(&dev_queue->dql) >= 0))
3015 return;
3016
3017 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3018
3019 /*
3020 * The XOFF flag must be set before checking the dql_avail below,
3021 * because in netdev_tx_completed_queue we update the dql_completed
3022 * before checking the XOFF flag.
3023 */
3024 smp_mb();
3025
3026 /* check again in case another CPU has just made room avail */
3027 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3028 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 3029#endif
c5d67bd7
TH
3030}
3031
0042d0c8
FF
3032/**
3033 * netdev_sent_queue - report the number of bytes queued to hardware
3034 * @dev: network device
3035 * @bytes: number of bytes queued to the hardware device queue
3036 *
3037 * Report the number of bytes queued for sending/completion to the network
3038 * device hardware queue. @bytes should be a good approximation and should
3039 * exactly match netdev_completed_queue() @bytes
3040 */
c5d67bd7
TH
3041static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3042{
3043 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3044}
3045
3046static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 3047 unsigned int pkts, unsigned int bytes)
c5d67bd7 3048{
114cf580 3049#ifdef CONFIG_BQL
b37c0fbe
AD
3050 if (unlikely(!bytes))
3051 return;
3052
3053 dql_completed(&dev_queue->dql, bytes);
3054
3055 /*
3056 * Without the memory barrier there is a small possiblity that
3057 * netdev_tx_sent_queue will miss the update and cause the queue to
3058 * be stopped forever
3059 */
3060 smp_mb();
3061
3062 if (dql_avail(&dev_queue->dql) < 0)
3063 return;
3064
3065 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3066 netif_schedule_queue(dev_queue);
114cf580 3067#endif
c5d67bd7
TH
3068}
3069
0042d0c8
FF
3070/**
3071 * netdev_completed_queue - report bytes and packets completed by device
3072 * @dev: network device
3073 * @pkts: actual number of packets sent over the medium
3074 * @bytes: actual number of bytes sent over the medium
3075 *
3076 * Report the number of bytes and packets transmitted by the network device
3077 * hardware queue over the physical medium, @bytes must exactly match the
3078 * @bytes amount passed to netdev_sent_queue()
3079 */
c5d67bd7 3080static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3081 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3082{
3083 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3084}
3085
3086static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3087{
114cf580 3088#ifdef CONFIG_BQL
5c490354 3089 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3090 dql_reset(&q->dql);
3091#endif
c5d67bd7
TH
3092}
3093
0042d0c8
FF
3094/**
3095 * netdev_reset_queue - reset the packets and bytes count of a network device
3096 * @dev_queue: network device
3097 *
3098 * Reset the bytes and packet count of a network device and clear the
3099 * software flow control OFF bit for this network device
3100 */
c5d67bd7
TH
3101static inline void netdev_reset_queue(struct net_device *dev_queue)
3102{
3103 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3104}
3105
b9507bda
DB
3106/**
3107 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3108 * @dev: network device
3109 * @queue_index: given tx queue index
3110 *
3111 * Returns 0 if given tx queue index >= number of device tx queues,
3112 * otherwise returns the originally passed tx queue index.
3113 */
3114static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3115{
3116 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3117 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3118 dev->name, queue_index,
3119 dev->real_num_tx_queues);
3120 return 0;
3121 }
3122
3123 return queue_index;
3124}
3125
bea3348e
SH
3126/**
3127 * netif_running - test if up
3128 * @dev: network device
3129 *
3130 * Test if the device has been brought up.
3131 */
4d29515f 3132static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3133{
3134 return test_bit(__LINK_STATE_START, &dev->state);
3135}
3136
f25f4e44 3137/*
5e82b4b2 3138 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3139 * stop, and a check if it's stopped. All other device management is
3140 * done at the overall netdevice level.
3141 * Also test the device if we're multiqueue.
3142 */
bea3348e
SH
3143
3144/**
3145 * netif_start_subqueue - allow sending packets on subqueue
3146 * @dev: network device
3147 * @queue_index: sub queue index
3148 *
3149 * Start individual transmit queue of a device with multiple transmit queues.
3150 */
f25f4e44
PWJ
3151static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3152{
fd2ea0a7 3153 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3154
3155 netif_tx_start_queue(txq);
f25f4e44
PWJ
3156}
3157
bea3348e
SH
3158/**
3159 * netif_stop_subqueue - stop sending packets on subqueue
3160 * @dev: network device
3161 * @queue_index: sub queue index
3162 *
3163 * Stop individual transmit queue of a device with multiple transmit queues.
3164 */
f25f4e44
PWJ
3165static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3166{
fd2ea0a7 3167 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3168 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3169}
3170
bea3348e
SH
3171/**
3172 * netif_subqueue_stopped - test status of subqueue
3173 * @dev: network device
3174 * @queue_index: sub queue index
3175 *
3176 * Check individual transmit queue of a device with multiple transmit queues.
3177 */
4d29515f
DM
3178static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3179 u16 queue_index)
f25f4e44 3180{
fd2ea0a7 3181 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3182
3183 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3184}
3185
4d29515f
DM
3186static inline bool netif_subqueue_stopped(const struct net_device *dev,
3187 struct sk_buff *skb)
668f895a
PE
3188{
3189 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3190}
bea3348e 3191
738b35cc
FF
3192/**
3193 * netif_wake_subqueue - allow sending packets on subqueue
3194 * @dev: network device
3195 * @queue_index: sub queue index
3196 *
3197 * Resume individual transmit queue of a device with multiple transmit queues.
3198 */
3199static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3200{
3201 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3202
3203 netif_tx_wake_queue(txq);
3204}
f25f4e44 3205
537c00de 3206#ifdef CONFIG_XPS
53af53ae 3207int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3208 u16 index);
537c00de
AD
3209#else
3210static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3211 const struct cpumask *mask,
537c00de
AD
3212 u16 index)
3213{
3214 return 0;
3215}
3216#endif
3217
5605c762
JP
3218u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3219 unsigned int num_tx_queues);
3220
a3d22a68
VZ
3221/*
3222 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3223 * as a distribution range limit for the returned value.
3224 */
3225static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3226 struct sk_buff *skb)
a3d22a68
VZ
3227{
3228 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3229}
3230
bea3348e
SH
3231/**
3232 * netif_is_multiqueue - test if device has multiple transmit queues
3233 * @dev: network device
3234 *
3235 * Check if device has multiple transmit queues
bea3348e 3236 */
4d29515f 3237static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3238{
a02cec21 3239 return dev->num_tx_queues > 1;
f25f4e44 3240}
1da177e4 3241
f629d208 3242int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3243
a953be53 3244#ifdef CONFIG_SYSFS
f629d208 3245int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3246#else
3247static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3248 unsigned int rxq)
3249{
3250 return 0;
3251}
3252#endif
3253
a953be53
MD
3254#ifdef CONFIG_SYSFS
3255static inline unsigned int get_netdev_rx_queue_index(
3256 struct netdev_rx_queue *queue)
3257{
3258 struct net_device *dev = queue->dev;
3259 int index = queue - dev->_rx;
3260
3261 BUG_ON(index >= dev->num_rx_queues);
3262 return index;
3263}
3264#endif
3265
16917b87 3266#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3267int netif_get_num_default_rss_queues(void);
16917b87 3268
e6247027
ED
3269enum skb_free_reason {
3270 SKB_REASON_CONSUMED,
3271 SKB_REASON_DROPPED,
3272};
3273
3274void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3275void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3276
e6247027
ED
3277/*
3278 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3279 * interrupt context or with hardware interrupts being disabled.
3280 * (in_irq() || irqs_disabled())
3281 *
3282 * We provide four helpers that can be used in following contexts :
3283 *
3284 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3285 * replacing kfree_skb(skb)
3286 *
3287 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3288 * Typically used in place of consume_skb(skb) in TX completion path
3289 *
3290 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3291 * replacing kfree_skb(skb)
3292 *
3293 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3294 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3295 */
e6247027
ED
3296static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3297{
3298 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3299}
3300
3301static inline void dev_consume_skb_irq(struct sk_buff *skb)
3302{
3303 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3304}
3305
3306static inline void dev_kfree_skb_any(struct sk_buff *skb)
3307{
3308 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3309}
3310
3311static inline void dev_consume_skb_any(struct sk_buff *skb)
3312{
3313 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3314}
1da177e4 3315
7c497478
JW
3316void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3317int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
f629d208
JP
3318int netif_rx(struct sk_buff *skb);
3319int netif_rx_ni(struct sk_buff *skb);
04eb4489 3320int netif_receive_skb(struct sk_buff *skb);
1c601d82 3321int netif_receive_skb_core(struct sk_buff *skb);
f629d208
JP
3322gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3323void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3324struct sk_buff *napi_get_frags(struct napi_struct *napi);
3325gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3326struct packet_offload *gro_find_receive_by_type(__be16 type);
3327struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3328
3329static inline void napi_free_frags(struct napi_struct *napi)
3330{
3331 kfree_skb(napi->skb);
3332 napi->skb = NULL;
3333}
3334
24b27fc4 3335bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3336int netdev_rx_handler_register(struct net_device *dev,
3337 rx_handler_func_t *rx_handler,
3338 void *rx_handler_data);
3339void netdev_rx_handler_unregister(struct net_device *dev);
3340
3341bool dev_valid_name(const char *name);
3342int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3343int dev_ethtool(struct net *net, struct ifreq *);
3344unsigned int dev_get_flags(const struct net_device *);
3345int __dev_change_flags(struct net_device *, unsigned int flags);
3346int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3347void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3348 unsigned int gchanges);
f629d208
JP
3349int dev_change_name(struct net_device *, const char *);
3350int dev_set_alias(struct net_device *, const char *, size_t);
6c557001 3351int dev_get_alias(const struct net_device *, char *, size_t);
f629d208 3352int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3353int __dev_set_mtu(struct net_device *, int);
f629d208
JP
3354int dev_set_mtu(struct net_device *, int);
3355void dev_set_group(struct net_device *, int);
3356int dev_set_mac_address(struct net_device *, struct sockaddr *);
3357int dev_change_carrier(struct net_device *, bool new_carrier);
3358int dev_get_phys_port_id(struct net_device *dev,
02637fce 3359 struct netdev_phys_item_id *ppid);
db24a904
DA
3360int dev_get_phys_port_name(struct net_device *dev,
3361 char *name, size_t len);
d746d707 3362int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3363struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3364struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3365 struct netdev_queue *txq, int *ret);
d67b9cd2 3366
f4e63525 3367typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
d67b9cd2
DB
3368int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3369 int fd, u32 flags);
f4e63525 3370u8 __dev_xdp_attached(struct net_device *dev, bpf_op_t xdp_op, u32 *prog_id);
d67b9cd2 3371
a0265d28 3372int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3373int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3374bool is_skb_forwardable(const struct net_device *dev,
3375 const struct sk_buff *skb);
1da177e4 3376
4e3264d2
MKL
3377static __always_inline int ____dev_forward_skb(struct net_device *dev,
3378 struct sk_buff *skb)
3379{
3380 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3381 unlikely(!is_skb_forwardable(dev, skb))) {
3382 atomic_long_inc(&dev->rx_dropped);
3383 kfree_skb(skb);
3384 return NET_RX_DROP;
3385 }
3386
3387 skb_scrub_packet(skb, true);
3388 skb->priority = 0;
3389 return 0;
3390}
3391
74b20582
DA
3392void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3393
20380731 3394extern int netdev_budget;
7acf8a1e 3395extern unsigned int netdev_budget_usecs;
1da177e4
LT
3396
3397/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3398void netdev_run_todo(void);
1da177e4 3399
bea3348e
SH
3400/**
3401 * dev_put - release reference to device
3402 * @dev: network device
3403 *
9ef4429b 3404 * Release reference to device to allow it to be freed.
bea3348e 3405 */
1da177e4
LT
3406static inline void dev_put(struct net_device *dev)
3407{
933393f5 3408 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3409}
3410
bea3348e
SH
3411/**
3412 * dev_hold - get reference to device
3413 * @dev: network device
3414 *
9ef4429b 3415 * Hold reference to device to keep it from being freed.
bea3348e 3416 */
15333061
SH
3417static inline void dev_hold(struct net_device *dev)
3418{
933393f5 3419 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3420}
1da177e4
LT
3421
3422/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3423 * and _off may be called from IRQ context, but it is caller
3424 * who is responsible for serialization of these calls.
b00055aa
SR
3425 *
3426 * The name carrier is inappropriate, these functions should really be
3427 * called netif_lowerlayer_*() because they represent the state of any
3428 * kind of lower layer not just hardware media.
1da177e4
LT
3429 */
3430
f629d208
JP
3431void linkwatch_init_dev(struct net_device *dev);
3432void linkwatch_fire_event(struct net_device *dev);
3433void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3434
bea3348e
SH
3435/**
3436 * netif_carrier_ok - test if carrier present
3437 * @dev: network device
3438 *
3439 * Check if carrier is present on device
3440 */
4d29515f 3441static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3442{
3443 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3444}
3445
f629d208 3446unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3447
f629d208 3448void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3449
f629d208 3450void netif_carrier_on(struct net_device *dev);
1da177e4 3451
f629d208 3452void netif_carrier_off(struct net_device *dev);
1da177e4 3453
bea3348e
SH
3454/**
3455 * netif_dormant_on - mark device as dormant.
3456 * @dev: network device
3457 *
3458 * Mark device as dormant (as per RFC2863).
3459 *
3460 * The dormant state indicates that the relevant interface is not
3461 * actually in a condition to pass packets (i.e., it is not 'up') but is
3462 * in a "pending" state, waiting for some external event. For "on-
3463 * demand" interfaces, this new state identifies the situation where the
3464 * interface is waiting for events to place it in the up state.
bea3348e 3465 */
b00055aa
SR
3466static inline void netif_dormant_on(struct net_device *dev)
3467{
3468 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3469 linkwatch_fire_event(dev);
3470}
3471
bea3348e
SH
3472/**
3473 * netif_dormant_off - set device as not dormant.
3474 * @dev: network device
3475 *
3476 * Device is not in dormant state.
3477 */
b00055aa
SR
3478static inline void netif_dormant_off(struct net_device *dev)
3479{
3480 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3481 linkwatch_fire_event(dev);
3482}
3483
bea3348e 3484/**
8ecbc40a 3485 * netif_dormant - test if device is dormant
bea3348e
SH
3486 * @dev: network device
3487 *
8ecbc40a 3488 * Check if device is dormant.
bea3348e 3489 */
4d29515f 3490static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3491{
3492 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3493}
3494
3495
bea3348e
SH
3496/**
3497 * netif_oper_up - test if device is operational
3498 * @dev: network device
3499 *
3500 * Check if carrier is operational
3501 */
4d29515f 3502static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3503{
b00055aa
SR
3504 return (dev->operstate == IF_OPER_UP ||
3505 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3506}
3507
bea3348e
SH
3508/**
3509 * netif_device_present - is device available or removed
3510 * @dev: network device
3511 *
3512 * Check if device has not been removed from system.
3513 */
4d29515f 3514static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3515{
3516 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3517}
3518
f629d208 3519void netif_device_detach(struct net_device *dev);
1da177e4 3520
f629d208 3521void netif_device_attach(struct net_device *dev);
1da177e4
LT
3522
3523/*
3524 * Network interface message level settings
3525 */
1da177e4
LT
3526
3527enum {
3528 NETIF_MSG_DRV = 0x0001,
3529 NETIF_MSG_PROBE = 0x0002,
3530 NETIF_MSG_LINK = 0x0004,
3531 NETIF_MSG_TIMER = 0x0008,
3532 NETIF_MSG_IFDOWN = 0x0010,
3533 NETIF_MSG_IFUP = 0x0020,
3534 NETIF_MSG_RX_ERR = 0x0040,
3535 NETIF_MSG_TX_ERR = 0x0080,
3536 NETIF_MSG_TX_QUEUED = 0x0100,
3537 NETIF_MSG_INTR = 0x0200,
3538 NETIF_MSG_TX_DONE = 0x0400,
3539 NETIF_MSG_RX_STATUS = 0x0800,
3540 NETIF_MSG_PKTDATA = 0x1000,
3541 NETIF_MSG_HW = 0x2000,
3542 NETIF_MSG_WOL = 0x4000,
3543};
3544
3545#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3546#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3547#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3548#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3549#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3550#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3551#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3552#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3553#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3554#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3555#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3556#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3557#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3558#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3559#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3560
3561static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3562{
3563 /* use default */
3564 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3565 return default_msg_enable_bits;
3566 if (debug_value == 0) /* no output */
3567 return 0;
3568 /* set low N bits */
5c199eb7 3569 return (1U << debug_value) - 1;
1da177e4
LT
3570}
3571
c773e847 3572static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3573{
c773e847
DM
3574 spin_lock(&txq->_xmit_lock);
3575 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3576}
3577
5a717f4f
MT
3578static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3579{
3580 __acquire(&txq->_xmit_lock);
3581 return true;
3582}
3583
3584static inline void __netif_tx_release(struct netdev_queue *txq)
3585{
3586 __release(&txq->_xmit_lock);
3587}
3588
fd2ea0a7
DM
3589static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3590{
3591 spin_lock_bh(&txq->_xmit_lock);
3592 txq->xmit_lock_owner = smp_processor_id();
3593}
3594
4d29515f 3595static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3596{
4d29515f 3597 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3598 if (likely(ok))
3599 txq->xmit_lock_owner = smp_processor_id();
3600 return ok;
3601}
3602
3603static inline void __netif_tx_unlock(struct netdev_queue *txq)
3604{
3605 txq->xmit_lock_owner = -1;
3606 spin_unlock(&txq->_xmit_lock);
3607}
3608
3609static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3610{
3611 txq->xmit_lock_owner = -1;
3612 spin_unlock_bh(&txq->_xmit_lock);
3613}
3614
08baf561
ED
3615static inline void txq_trans_update(struct netdev_queue *txq)
3616{
3617 if (txq->xmit_lock_owner != -1)
3618 txq->trans_start = jiffies;
3619}
3620
ba162f8e
FW
3621/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3622static inline void netif_trans_update(struct net_device *dev)
3623{
9b36627a
FW
3624 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3625
3626 if (txq->trans_start != jiffies)
3627 txq->trans_start = jiffies;
ba162f8e
FW
3628}
3629
d29f749e
DJ
3630/**
3631 * netif_tx_lock - grab network device transmit lock
3632 * @dev: network device
d29f749e
DJ
3633 *
3634 * Get network device transmit lock
3635 */
22dd7495
JHS
3636static inline void netif_tx_lock(struct net_device *dev)
3637{
e8a0464c 3638 unsigned int i;
c3f26a26 3639 int cpu;
c773e847 3640
c3f26a26
DM
3641 spin_lock(&dev->tx_global_lock);
3642 cpu = smp_processor_id();
e8a0464c
DM
3643 for (i = 0; i < dev->num_tx_queues; i++) {
3644 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3645
3646 /* We are the only thread of execution doing a
3647 * freeze, but we have to grab the _xmit_lock in
3648 * order to synchronize with threads which are in
3649 * the ->hard_start_xmit() handler and already
3650 * checked the frozen bit.
3651 */
e8a0464c 3652 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3653 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3654 __netif_tx_unlock(txq);
e8a0464c 3655 }
932ff279
HX
3656}
3657
3658static inline void netif_tx_lock_bh(struct net_device *dev)
3659{
e8a0464c
DM
3660 local_bh_disable();
3661 netif_tx_lock(dev);
932ff279
HX
3662}
3663
932ff279
HX
3664static inline void netif_tx_unlock(struct net_device *dev)
3665{
e8a0464c
DM
3666 unsigned int i;
3667
3668 for (i = 0; i < dev->num_tx_queues; i++) {
3669 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3670
c3f26a26
DM
3671 /* No need to grab the _xmit_lock here. If the
3672 * queue is not stopped for another reason, we
3673 * force a schedule.
3674 */
3675 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3676 netif_schedule_queue(txq);
c3f26a26
DM
3677 }
3678 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3679}
3680
3681static inline void netif_tx_unlock_bh(struct net_device *dev)
3682{
e8a0464c
DM
3683 netif_tx_unlock(dev);
3684 local_bh_enable();
932ff279
HX
3685}
3686
c773e847 3687#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3688 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3689 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3690 } else { \
3691 __netif_tx_acquire(txq); \
22dd7495
JHS
3692 } \
3693}
3694
5efeac44
EB
3695#define HARD_TX_TRYLOCK(dev, txq) \
3696 (((dev->features & NETIF_F_LLTX) == 0) ? \
3697 __netif_tx_trylock(txq) : \
5a717f4f 3698 __netif_tx_acquire(txq))
5efeac44 3699
c773e847 3700#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3701 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3702 __netif_tx_unlock(txq); \
5a717f4f
MT
3703 } else { \
3704 __netif_tx_release(txq); \
22dd7495
JHS
3705 } \
3706}
3707
1da177e4
LT
3708static inline void netif_tx_disable(struct net_device *dev)
3709{
fd2ea0a7 3710 unsigned int i;
c3f26a26 3711 int cpu;
fd2ea0a7 3712
c3f26a26
DM
3713 local_bh_disable();
3714 cpu = smp_processor_id();
fd2ea0a7
DM
3715 for (i = 0; i < dev->num_tx_queues; i++) {
3716 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3717
3718 __netif_tx_lock(txq, cpu);
fd2ea0a7 3719 netif_tx_stop_queue(txq);
c3f26a26 3720 __netif_tx_unlock(txq);
fd2ea0a7 3721 }
c3f26a26 3722 local_bh_enable();
1da177e4
LT
3723}
3724
e308a5d8
DM
3725static inline void netif_addr_lock(struct net_device *dev)
3726{
3727 spin_lock(&dev->addr_list_lock);
3728}
3729
2429f7ac
JP
3730static inline void netif_addr_lock_nested(struct net_device *dev)
3731{
25175ba5
VY
3732 int subclass = SINGLE_DEPTH_NESTING;
3733
3734 if (dev->netdev_ops->ndo_get_lock_subclass)
3735 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3736
3737 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3738}
3739
e308a5d8
DM
3740static inline void netif_addr_lock_bh(struct net_device *dev)
3741{
3742 spin_lock_bh(&dev->addr_list_lock);
3743}
3744
3745static inline void netif_addr_unlock(struct net_device *dev)
3746{
3747 spin_unlock(&dev->addr_list_lock);
3748}
3749
3750static inline void netif_addr_unlock_bh(struct net_device *dev)
3751{
3752 spin_unlock_bh(&dev->addr_list_lock);
3753}
3754
f001fde5 3755/*
31278e71 3756 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3757 * rcu_read_lock held.
3758 */
3759#define for_each_dev_addr(dev, ha) \
31278e71 3760 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3761
1da177e4
LT
3762/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3763
f629d208 3764void ether_setup(struct net_device *dev);
1da177e4
LT
3765
3766/* Support for loadable net-drivers */
f629d208 3767struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3768 unsigned char name_assign_type,
f629d208
JP
3769 void (*setup)(struct net_device *),
3770 unsigned int txqs, unsigned int rxqs);
0ad646c8
CW
3771int dev_get_valid_name(struct net *net, struct net_device *dev,
3772 const char *name);
3773
c835a677
TG
3774#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3775 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3776
c835a677
TG
3777#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3778 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3779 count)
36909ea4 3780
f629d208
JP
3781int register_netdev(struct net_device *dev);
3782void unregister_netdev(struct net_device *dev);
f001fde5 3783
22bedad3 3784/* General hardware address lists handling functions */
f629d208
JP
3785int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3786 struct netdev_hw_addr_list *from_list, int addr_len);
3787void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3788 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3789int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3790 struct net_device *dev,
3791 int (*sync)(struct net_device *, const unsigned char *),
3792 int (*unsync)(struct net_device *,
3793 const unsigned char *));
3794void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3795 struct net_device *dev,
3796 int (*unsync)(struct net_device *,
3797 const unsigned char *));
f629d208 3798void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3799
f001fde5 3800/* Functions used for device addresses handling */
f629d208
JP
3801int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3802 unsigned char addr_type);
3803int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3804 unsigned char addr_type);
f629d208
JP
3805void dev_addr_flush(struct net_device *dev);
3806int dev_addr_init(struct net_device *dev);
a748ee24
JP
3807
3808/* Functions used for unicast addresses handling */
f629d208
JP
3809int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3810int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3811int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3812int dev_uc_sync(struct net_device *to, struct net_device *from);
3813int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3814void dev_uc_unsync(struct net_device *to, struct net_device *from);
3815void dev_uc_flush(struct net_device *dev);
3816void dev_uc_init(struct net_device *dev);
f001fde5 3817
670e5b8e
AD
3818/**
3819 * __dev_uc_sync - Synchonize device's unicast list
3820 * @dev: device to sync
3821 * @sync: function to call if address should be added
3822 * @unsync: function to call if address should be removed
3823 *
3824 * Add newly added addresses to the interface, and release
3825 * addresses that have been deleted.
5e82b4b2 3826 */
670e5b8e
AD
3827static inline int __dev_uc_sync(struct net_device *dev,
3828 int (*sync)(struct net_device *,
3829 const unsigned char *),
3830 int (*unsync)(struct net_device *,
3831 const unsigned char *))
3832{
3833 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3834}
3835
3836/**
e793c0f7 3837 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3838 * @dev: device to sync
3839 * @unsync: function to call if address should be removed
3840 *
3841 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3842 */
670e5b8e
AD
3843static inline void __dev_uc_unsync(struct net_device *dev,
3844 int (*unsync)(struct net_device *,
3845 const unsigned char *))
3846{
3847 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3848}
3849
22bedad3 3850/* Functions used for multicast addresses handling */
f629d208
JP
3851int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3852int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3853int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3854int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3855int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3856int dev_mc_sync(struct net_device *to, struct net_device *from);
3857int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3858void dev_mc_unsync(struct net_device *to, struct net_device *from);
3859void dev_mc_flush(struct net_device *dev);
3860void dev_mc_init(struct net_device *dev);
f001fde5 3861
670e5b8e
AD
3862/**
3863 * __dev_mc_sync - Synchonize device's multicast list
3864 * @dev: device to sync
3865 * @sync: function to call if address should be added
3866 * @unsync: function to call if address should be removed
3867 *
3868 * Add newly added addresses to the interface, and release
3869 * addresses that have been deleted.
5e82b4b2 3870 */
670e5b8e
AD
3871static inline int __dev_mc_sync(struct net_device *dev,
3872 int (*sync)(struct net_device *,
3873 const unsigned char *),
3874 int (*unsync)(struct net_device *,
3875 const unsigned char *))
3876{
3877 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3878}
3879
3880/**
e793c0f7 3881 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3882 * @dev: device to sync
3883 * @unsync: function to call if address should be removed
3884 *
3885 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3886 */
670e5b8e
AD
3887static inline void __dev_mc_unsync(struct net_device *dev,
3888 int (*unsync)(struct net_device *,
3889 const unsigned char *))
3890{
3891 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3892}
3893
4417da66 3894/* Functions used for secondary unicast and multicast support */
f629d208
JP
3895void dev_set_rx_mode(struct net_device *dev);
3896void __dev_set_rx_mode(struct net_device *dev);
3897int dev_set_promiscuity(struct net_device *dev, int inc);
3898int dev_set_allmulti(struct net_device *dev, int inc);
3899void netdev_state_change(struct net_device *dev);
3900void netdev_notify_peers(struct net_device *dev);
3901void netdev_features_change(struct net_device *dev);
1da177e4 3902/* Load a device via the kmod */
f629d208
JP
3903void dev_load(struct net *net, const char *name);
3904struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3905 struct rtnl_link_stats64 *storage);
3906void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3907 const struct net_device_stats *netdev_stats);
eeda3fd6 3908
1da177e4 3909extern int netdev_max_backlog;
3b098e2d 3910extern int netdev_tstamp_prequeue;
1da177e4 3911extern int weight_p;
3d48b53f
MT
3912extern int dev_weight_rx_bias;
3913extern int dev_weight_tx_bias;
3914extern int dev_rx_weight;
3915extern int dev_tx_weight;
9ff162a8 3916
f629d208 3917bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3918struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3919 struct list_head **iter);
f629d208
JP
3920struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3921 struct list_head **iter);
8b5be856 3922
44a40855
VY
3923/* iterate through upper list, must be called under RCU read lock */
3924#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3925 for (iter = &(dev)->adj_list.upper, \
3926 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3927 updev; \
3928 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3929
1a3f060c
DA
3930int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3931 int (*fn)(struct net_device *upper_dev,
3932 void *data),
3933 void *data);
3934
3935bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3936 struct net_device *upper_dev);
3937
25cc72a3
IS
3938bool netdev_has_any_upper_dev(struct net_device *dev);
3939
f629d208
JP
3940void *netdev_lower_get_next_private(struct net_device *dev,
3941 struct list_head **iter);
3942void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3943 struct list_head **iter);
31088a11
VF
3944
3945#define netdev_for_each_lower_private(dev, priv, iter) \
3946 for (iter = (dev)->adj_list.lower.next, \
3947 priv = netdev_lower_get_next_private(dev, &(iter)); \
3948 priv; \
3949 priv = netdev_lower_get_next_private(dev, &(iter)))
3950
3951#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3952 for (iter = &(dev)->adj_list.lower, \
3953 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3954 priv; \
3955 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3956
4085ebe8
VY
3957void *netdev_lower_get_next(struct net_device *dev,
3958 struct list_head **iter);
7ce856aa 3959
4085ebe8 3960#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3961 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3962 ldev = netdev_lower_get_next(dev, &(iter)); \
3963 ldev; \
3964 ldev = netdev_lower_get_next(dev, &(iter)))
3965
7ce856aa
JP
3966struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3967 struct list_head **iter);
3968struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3969 struct list_head **iter);
3970
1a3f060c
DA
3971int netdev_walk_all_lower_dev(struct net_device *dev,
3972 int (*fn)(struct net_device *lower_dev,
3973 void *data),
3974 void *data);
3975int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3976 int (*fn)(struct net_device *lower_dev,
3977 void *data),
3978 void *data);
3979
f629d208 3980void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3981void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3982struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3983struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
42ab19ee
DA
3984int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
3985 struct netlink_ext_ack *extack);
f629d208 3986int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3987 struct net_device *upper_dev,
42ab19ee
DA
3988 void *upper_priv, void *upper_info,
3989 struct netlink_ext_ack *extack);
f629d208
JP
3990void netdev_upper_dev_unlink(struct net_device *dev,
3991 struct net_device *upper_dev);
5bb025fa 3992void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3993void *netdev_lower_dev_get_private(struct net_device *dev,
3994 struct net_device *lower_dev);
04d48266
JP
3995void netdev_lower_state_changed(struct net_device *lower_dev,
3996 void *lower_state_info);
960fb622
ED
3997
3998/* RSS keys are 40 or 52 bytes long */
3999#define NETDEV_RSS_KEY_LEN 52
ba905f5e 4000extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
4001void netdev_rss_key_fill(void *buffer, size_t len);
4002
952fcfd0 4003int dev_get_nest_level(struct net_device *dev);
f629d208 4004int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 4005int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
4006int skb_csum_hwoffload_help(struct sk_buff *skb,
4007 const netdev_features_t features);
4008
f629d208
JP
4009struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4010 netdev_features_t features, bool tx_path);
4011struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4012 netdev_features_t features);
12b0004d 4013
61bd3857
MS
4014struct netdev_bonding_info {
4015 ifslave slave;
4016 ifbond master;
4017};
4018
4019struct netdev_notifier_bonding_info {
4020 struct netdev_notifier_info info; /* must be first */
4021 struct netdev_bonding_info bonding_info;
4022};
4023
4024void netdev_bonding_info_change(struct net_device *dev,
4025 struct netdev_bonding_info *bonding_info);
4026
12b0004d
CW
4027static inline
4028struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4029{
4030 return __skb_gso_segment(skb, features, true);
4031}
53d6471c 4032__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
4033
4034static inline bool can_checksum_protocol(netdev_features_t features,
4035 __be16 protocol)
4036{
c8cd0989
TH
4037 if (protocol == htons(ETH_P_FCOE))
4038 return !!(features & NETIF_F_FCOE_CRC);
4039
4040 /* Assume this is an IP checksum (not SCTP CRC) */
4041
4042 if (features & NETIF_F_HW_CSUM) {
4043 /* Can checksum everything */
4044 return true;
4045 }
4046
4047 switch (protocol) {
4048 case htons(ETH_P_IP):
4049 return !!(features & NETIF_F_IP_CSUM);
4050 case htons(ETH_P_IPV6):
4051 return !!(features & NETIF_F_IPV6_CSUM);
4052 default:
4053 return false;
4054 }
ec5f0615 4055}
12b0004d 4056
fb286bb2 4057#ifdef CONFIG_BUG
f629d208 4058void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
4059#else
4060static inline void netdev_rx_csum_fault(struct net_device *dev)
4061{
4062}
4063#endif
1da177e4 4064/* rx skb timestamps */
f629d208
JP
4065void net_enable_timestamp(void);
4066void net_disable_timestamp(void);
1da177e4 4067
20380731 4068#ifdef CONFIG_PROC_FS
f629d208 4069int __init dev_proc_init(void);
900ff8c6
CW
4070#else
4071#define dev_proc_init() 0
20380731
ACM
4072#endif
4073
4798248e 4074static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4075 struct sk_buff *skb, struct net_device *dev,
4076 bool more)
4798248e 4077{
fa2dbdc2 4078 skb->xmit_more = more ? 1 : 0;
0b725a2c 4079 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4080}
4081
10b3ad8c 4082static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4083 struct netdev_queue *txq, bool more)
4798248e
DM
4084{
4085 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4086 int rc;
4798248e 4087
fa2dbdc2 4088 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4089 if (rc == NETDEV_TX_OK)
4090 txq_trans_update(txq);
4091
4092 return rc;
4798248e
DM
4093}
4094
b793dc5c 4095int netdev_class_create_file_ns(const struct class_attribute *class_attr,
42a2d923 4096 const void *ns);
b793dc5c 4097void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
42a2d923 4098 const void *ns);
58292cbe 4099
b793dc5c 4100static inline int netdev_class_create_file(const struct class_attribute *class_attr)
58292cbe
TH
4101{
4102 return netdev_class_create_file_ns(class_attr, NULL);
4103}
4104
b793dc5c 4105static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
58292cbe
TH
4106{
4107 netdev_class_remove_file_ns(class_attr, NULL);
4108}
b8a9787e 4109
737aec57 4110extern const struct kobj_ns_type_operations net_ns_type_operations;
04600794 4111
f629d208 4112const char *netdev_drivername(const struct net_device *dev);
6579e57b 4113
f629d208 4114void linkwatch_run_queue(void);
20380731 4115
da08143b
MK
4116static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4117 netdev_features_t f2)
4118{
c8cd0989
TH
4119 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4120 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4121 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4122 else
b6a0e72a 4123 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4124 }
da08143b 4125
c8cd0989 4126 return f1 & f2;
da08143b
MK
4127}
4128
c8f44aff
MM
4129static inline netdev_features_t netdev_get_wanted_features(
4130 struct net_device *dev)
5455c699
MM
4131{
4132 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4133}
c8f44aff
MM
4134netdev_features_t netdev_increment_features(netdev_features_t all,
4135 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4136
4137/* Allow TSO being used on stacked device :
4138 * Performing the GSO segmentation before last device
4139 * is a performance improvement.
4140 */
4141static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4142 netdev_features_t mask)
4143{
4144 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4145}
4146
6cb6a27c 4147int __netdev_update_features(struct net_device *dev);
5455c699 4148void netdev_update_features(struct net_device *dev);
afe12cc8 4149void netdev_change_features(struct net_device *dev);
7f353bf2 4150
fc4a7489
PM
4151void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4152 struct net_device *dev);
4153
e38f3025
TM
4154netdev_features_t passthru_features_check(struct sk_buff *skb,
4155 struct net_device *dev,
4156 netdev_features_t features);
c1e756bf 4157netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4158
4d29515f 4159static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4160{
7b748340 4161 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4162
4163 /* check flags correspondence */
4164 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4165 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4166 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4167 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4168 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4169 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4170 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4171 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4172 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4173 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4174 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4175 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4176 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4177 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4178 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4179 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0c19f846 4180 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
0345e186 4181
d6b4991a 4182 return (features & feature) == feature;
576a30eb
HX
4183}
4184
4d29515f 4185static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4186{
278b2513 4187 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4188 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4189}
4190
8b86a61d 4191static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4192 netdev_features_t features)
7967168c 4193{
fc741216 4194 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4195 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4196 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4197}
4198
82cc1a7a
PWJ
4199static inline void netif_set_gso_max_size(struct net_device *dev,
4200 unsigned int size)
4201{
4202 dev->gso_max_size = size;
4203}
4204
7a7ffbab
WCC
4205static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4206 int pulled_hlen, u16 mac_offset,
4207 int mac_len)
4208{
4209 skb->protocol = protocol;
4210 skb->encapsulation = 1;
4211 skb_push(skb, pulled_hlen);
4212 skb_reset_transport_header(skb);
4213 skb->mac_header = mac_offset;
4214 skb->network_header = skb->mac_header + mac_len;
4215 skb->mac_len = mac_len;
4216}
4217
3c175784
SD
4218static inline bool netif_is_macsec(const struct net_device *dev)
4219{
4220 return dev->priv_flags & IFF_MACSEC;
4221}
4222
b618aaa9 4223static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4224{
4225 return dev->priv_flags & IFF_MACVLAN;
4226}
4227
b618aaa9 4228static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4229{
4230 return dev->priv_flags & IFF_MACVLAN_PORT;
4231}
4232
b618aaa9 4233static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4234{
4235 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4236}
4237
b618aaa9 4238static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4239{
4240 return dev->priv_flags & IFF_IPVLAN_MASTER;
4241}
4242
b618aaa9 4243static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4244{
4245 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4246}
4247
b618aaa9 4248static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4249{
4250 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4251}
4252
3bdc0eba
BG
4253static inline bool netif_supports_nofcs(struct net_device *dev)
4254{
4255 return dev->priv_flags & IFF_SUPP_NOFCS;
4256}
4257
9d150e29
DB
4258static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4259{
4260 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4261}
4262
007979ea 4263static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4264{
007979ea 4265 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4266}
4267
fee6d4c7
DA
4268static inline bool netif_is_l3_slave(const struct net_device *dev)
4269{
4270 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4271}
4272
0894ae3f
JP
4273static inline bool netif_is_bridge_master(const struct net_device *dev)
4274{
4275 return dev->priv_flags & IFF_EBRIDGE;
4276}
4277
28f9ee22
VY
4278static inline bool netif_is_bridge_port(const struct net_device *dev)
4279{
4280 return dev->priv_flags & IFF_BRIDGE_PORT;
4281}
4282
35d4e172
JP
4283static inline bool netif_is_ovs_master(const struct net_device *dev)
4284{
4285 return dev->priv_flags & IFF_OPENVSWITCH;
4286}
4287
5be66141
JP
4288static inline bool netif_is_ovs_port(const struct net_device *dev)
4289{
4290 return dev->priv_flags & IFF_OVS_DATAPATH;
4291}
4292
b618aaa9 4293static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4294{
4295 return dev->priv_flags & IFF_TEAM;
4296}
4297
b618aaa9 4298static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4299{
4300 return dev->priv_flags & IFF_TEAM_PORT;
4301}
4302
b618aaa9 4303static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4304{
4305 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4306}
4307
b618aaa9 4308static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4309{
4310 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4311}
4312
d4ab4286
KJ
4313static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4314{
4315 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4316}
4317
02875878
ED
4318/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4319static inline void netif_keep_dst(struct net_device *dev)
4320{
4321 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4322}
4323
18d3df3e
PA
4324/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4325static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4326{
4327 /* TODO: reserve and use an additional IFF bit, if we get more users */
4328 return dev->priv_flags & IFF_MACSEC;
4329}
4330
505d4f73 4331extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4332
571ba423
JP
4333/* Logging, debugging and troubleshooting/diagnostic helpers. */
4334
4335/* netdev_printk helpers, similar to dev_printk */
4336
4337static inline const char *netdev_name(const struct net_device *dev)
4338{
c6f854d5
VF
4339 if (!dev->name[0] || strchr(dev->name, '%'))
4340 return "(unnamed net_device)";
571ba423
JP
4341 return dev->name;
4342}
4343
8397ed36
DA
4344static inline bool netdev_unregistering(const struct net_device *dev)
4345{
4346 return dev->reg_state == NETREG_UNREGISTERING;
4347}
4348
ccc7f496
VF
4349static inline const char *netdev_reg_state(const struct net_device *dev)
4350{
4351 switch (dev->reg_state) {
4352 case NETREG_UNINITIALIZED: return " (uninitialized)";
4353 case NETREG_REGISTERED: return "";
4354 case NETREG_UNREGISTERING: return " (unregistering)";
4355 case NETREG_UNREGISTERED: return " (unregistered)";
4356 case NETREG_RELEASED: return " (released)";
4357 case NETREG_DUMMY: return " (dummy)";
4358 }
4359
4360 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4361 return " (unknown)";
4362}
4363
f629d208 4364__printf(3, 4)
6ea754eb
JP
4365void netdev_printk(const char *level, const struct net_device *dev,
4366 const char *format, ...);
f629d208 4367__printf(2, 3)
6ea754eb 4368void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4369__printf(2, 3)
6ea754eb 4370void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4371__printf(2, 3)
6ea754eb 4372void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4373__printf(2, 3)
6ea754eb 4374void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4375__printf(2, 3)
6ea754eb 4376void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4377__printf(2, 3)
6ea754eb 4378void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4379__printf(2, 3)
6ea754eb 4380void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4381
375ef2b1
GP
4382#define netdev_level_once(level, dev, fmt, ...) \
4383do { \
4384 static bool __print_once __read_mostly; \
4385 \
4386 if (!__print_once) { \
4387 __print_once = true; \
4388 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4389 } \
4390} while (0)
4391
4392#define netdev_emerg_once(dev, fmt, ...) \
4393 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4394#define netdev_alert_once(dev, fmt, ...) \
4395 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4396#define netdev_crit_once(dev, fmt, ...) \
4397 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4398#define netdev_err_once(dev, fmt, ...) \
4399 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4400#define netdev_warn_once(dev, fmt, ...) \
4401 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4402#define netdev_notice_once(dev, fmt, ...) \
4403 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4404#define netdev_info_once(dev, fmt, ...) \
4405 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4406
8909c9ad
VK
4407#define MODULE_ALIAS_NETDEV(device) \
4408 MODULE_ALIAS("netdev-" device)
4409
b558c96f 4410#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4411#define netdev_dbg(__dev, format, args...) \
4412do { \
ffa10cb4 4413 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4414} while (0)
b558c96f
JC
4415#elif defined(DEBUG)
4416#define netdev_dbg(__dev, format, args...) \
4417 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4418#else
4419#define netdev_dbg(__dev, format, args...) \
4420({ \
4421 if (0) \
4422 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4423})
4424#endif
4425
4426#if defined(VERBOSE_DEBUG)
4427#define netdev_vdbg netdev_dbg
4428#else
4429
4430#define netdev_vdbg(dev, format, args...) \
4431({ \
4432 if (0) \
4433 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4434 0; \
4435})
4436#endif
4437
4438/*
4439 * netdev_WARN() acts like dev_printk(), but with the key difference
4440 * of using a WARN/WARN_ON to get the message out, including the
4441 * file/line information and a backtrace.
4442 */
4443#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4444 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4445 netdev_reg_state(dev), ##args)
571ba423 4446
1e4a8e87
GP
4447#define netdev_WARN_ONCE(dev, format, args...) \
4448 WARN_ONCE(1, "netdevice: %s%s\n" format, netdev_name(dev), \
375ef2b1
GP
4449 netdev_reg_state(dev), ##args)
4450
b3d95c5c
JP
4451/* netif printk helpers, similar to netdev_printk */
4452
4453#define netif_printk(priv, type, level, dev, fmt, args...) \
4454do { \
4455 if (netif_msg_##type(priv)) \
4456 netdev_printk(level, (dev), fmt, ##args); \
4457} while (0)
4458
f45f4321
JP
4459#define netif_level(level, priv, type, dev, fmt, args...) \
4460do { \
4461 if (netif_msg_##type(priv)) \
4462 netdev_##level(dev, fmt, ##args); \
4463} while (0)
4464
b3d95c5c 4465#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4466 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4467#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4468 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4469#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4470 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4471#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4472 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4473#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4474 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4475#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4476 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4477#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4478 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4479
0053ea9c 4480#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4481#define netif_dbg(priv, type, netdev, format, args...) \
4482do { \
4483 if (netif_msg_##type(priv)) \
b5fb0a03 4484 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4485} while (0)
0053ea9c
JP
4486#elif defined(DEBUG)
4487#define netif_dbg(priv, type, dev, format, args...) \
4488 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4489#else
4490#define netif_dbg(priv, type, dev, format, args...) \
4491({ \
4492 if (0) \
4493 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4494 0; \
4495})
4496#endif
4497
f617f276
EC
4498/* if @cond then downgrade to debug, else print at @level */
4499#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4500 do { \
4501 if (cond) \
4502 netif_dbg(priv, type, netdev, fmt, ##args); \
4503 else \
4504 netif_ ## level(priv, type, netdev, fmt, ##args); \
4505 } while (0)
4506
b3d95c5c 4507#if defined(VERBOSE_DEBUG)
bcfcc450 4508#define netif_vdbg netif_dbg
b3d95c5c
JP
4509#else
4510#define netif_vdbg(priv, type, dev, format, args...) \
4511({ \
4512 if (0) \
a4ed89cb 4513 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4514 0; \
4515})
4516#endif
571ba423 4517
900ff8c6
CW
4518/*
4519 * The list of packet types we will receive (as opposed to discard)
4520 * and the routines to invoke.
4521 *
4522 * Why 16. Because with 16 the only overlap we get on a hash of the
4523 * low nibble of the protocol value is RARP/SNAP/X.25.
4524 *
900ff8c6 4525 * 0800 IP
900ff8c6
CW
4526 * 0001 802.3
4527 * 0002 AX.25
4528 * 0004 802.2
4529 * 8035 RARP
4530 * 0005 SNAP
4531 * 0805 X.25
4532 * 0806 ARP
4533 * 8137 IPX
4534 * 0009 Localtalk
4535 * 86DD IPv6
4536 */
4537#define PTYPE_HASH_SIZE (16)
4538#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4539
385a154c 4540#endif /* _LINUX_NETDEVICE_H */