net: bpf: rename ndo_xdp to ndo_bpf
[linux-2.6-block.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
bea3348e 38#include <linux/workqueue.h>
114cf580 39#include <linux/dynamic_queue_limits.h>
1da177e4 40
b1b67dd4 41#include <linux/ethtool.h>
a050c33f 42#include <net/net_namespace.h>
7a6b6f51 43#ifdef CONFIG_DCB
2f90b865
AD
44#include <net/dcbnl.h>
45#endif
5bc1421e 46#include <net/netprio_cgroup.h>
a050c33f 47
a59e2ecb 48#include <linux/netdev_features.h>
77162022 49#include <linux/neighbour.h>
607ca46e 50#include <uapi/linux/netdevice.h>
61bd3857 51#include <uapi/linux/if_bonding.h>
e4c6734e 52#include <uapi/linux/pkt_cls.h>
59cc1f61 53#include <linux/hashtable.h>
a59e2ecb 54
115c1d6e 55struct netpoll_info;
313162d0 56struct device;
c1f19b51 57struct phy_device;
2f657a60 58struct dsa_port;
c6e970a0 59
704232c2
JB
60/* 802.11 specific */
61struct wireless_dev;
98a18b6f
AA
62/* 802.15.4 specific */
63struct wpan_dev;
03c57747 64struct mpls_dev;
7c46a640
AD
65/* UDP Tunnel offloads */
66struct udp_tunnel_info;
a7862b45 67struct bpf_prog;
814abfab 68struct xdp_buff;
1da177e4 69
f629d208
JP
70void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
d07d7507 72
9a1654ba
JP
73/* Backlog congestion levels */
74#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75#define NET_RX_DROP 1 /* packet dropped */
76
572a9d7b
PM
77/*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
91 * others are propagated to higher layers.
92 */
93
94/* qdisc ->enqueue() return codes. */
95#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
96#define NET_XMIT_DROP 0x01 /* skb dropped */
97#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 98#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 99
b9df3cb8
GR
100/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 103#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
104#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105
dc1f8bf6 106/* Driver transmit return codes */
9a1654ba 107#define NETDEV_TX_MASK 0xf0
572a9d7b 108
dc1f8bf6 109enum netdev_tx {
572a9d7b 110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
113};
114typedef enum netdev_tx netdev_tx_t;
115
9a1654ba
JP
116/*
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 */
120static inline bool dev_xmit_complete(int rc)
121{
122 /*
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 */
128 if (likely(rc < NET_XMIT_MASK))
129 return true;
130
131 return false;
132}
133
1da177e4 134/*
5e82b4b2 135 * Compute the worst-case header length according to the protocols
1da177e4
LT
136 * used.
137 */
fe2918b0 138
c0eb4540
KS
139#if defined(CONFIG_HYPERV_NET)
140# define LL_MAX_HEADER 128
141#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
142# if defined(CONFIG_MAC80211_MESH)
143# define LL_MAX_HEADER 128
144# else
145# define LL_MAX_HEADER 96
146# endif
1da177e4 147#else
8388e3da 148# define LL_MAX_HEADER 32
1da177e4
LT
149#endif
150
d11ead75
BH
151#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
153#define MAX_HEADER LL_MAX_HEADER
154#else
155#define MAX_HEADER (LL_MAX_HEADER + 48)
156#endif
157
158/*
be1f3c2c
BH
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
1da177e4 161 */
fe2918b0 162
d94d9fee 163struct net_device_stats {
3cfde79c
BH
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
1da177e4 173 unsigned long collisions;
1da177e4 174 unsigned long rx_length_errors;
3cfde79c
BH
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
1da177e4
LT
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
1da177e4
LT
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187};
188
1da177e4
LT
189
190#include <linux/cache.h>
191#include <linux/skbuff.h>
192
adc9300e 193#ifdef CONFIG_RPS
c5905afb
IM
194#include <linux/static_key.h>
195extern struct static_key rps_needed;
13bfff25 196extern struct static_key rfs_needed;
adc9300e
ED
197#endif
198
1da177e4
LT
199struct neighbour;
200struct neigh_parms;
201struct sk_buff;
202
f001fde5
JP
203struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
206 unsigned char type;
ccffad25
JP
207#define NETDEV_HW_ADDR_T_LAN 1
208#define NETDEV_HW_ADDR_T_SAN 2
209#define NETDEV_HW_ADDR_T_SLAVE 3
210#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 211#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 212 bool global_use;
4cd729b0 213 int sync_cnt;
8f8f103d 214 int refcount;
4543fbef 215 int synced;
f001fde5
JP
216 struct rcu_head rcu_head;
217};
218
31278e71
JP
219struct netdev_hw_addr_list {
220 struct list_head list;
221 int count;
222};
223
22bedad3
JP
224#define netdev_hw_addr_list_count(l) ((l)->count)
225#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226#define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 228
22bedad3
JP
229#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231#define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 233
22bedad3
JP
234#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 236#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 238
d94d9fee 239struct hh_cache {
5b3dc2f3 240 unsigned int hh_len;
3644f0ce 241 seqlock_t hh_lock;
1da177e4
LT
242
243 /* cached hardware header; allow for machine alignment needs. */
244#define HH_DATA_MOD 16
245#define HH_DATA_OFF(__len) \
5ba0eac6 246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
247#define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250};
251
5e82b4b2 252/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260#define LL_RESERVED_SPACE(dev) \
f5184d26 261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 262#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 264
3b04ddde
SH
265struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
95c96174 268 const void *saddr, unsigned int len);
3b04ddde 269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
2793a23a 274 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
275};
276
1da177e4 277/* These flag bits are private to the generic network queueing
5e82b4b2 278 * layer; they may not be explicitly referenced by any other
1da177e4
LT
279 * code.
280 */
281
d94d9fee 282enum netdev_state_t {
1da177e4
LT
283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
1da177e4 285 __LINK_STATE_NOCARRIER,
b00055aa
SR
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
1da177e4
LT
288};
289
290
291/*
5e82b4b2 292 * This structure holds boot-time configured netdevice settings. They
fe2918b0 293 * are then used in the device probing.
1da177e4
LT
294 */
295struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298};
299#define NETDEV_BOOT_SETUP_MAX 8
300
f629d208 301int __init netdev_boot_setup(char *str);
1da177e4 302
bea3348e
SH
303/*
304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 310 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
404f7c9e 317 unsigned int gro_count;
bea3348e
SH
318 int (*poll)(struct napi_struct *, int);
319#ifdef CONFIG_NETPOLL
bea3348e 320 int poll_owner;
bea3348e 321#endif
5d38a079 322 struct net_device *dev;
d565b0a1 323 struct sk_buff *gro_list;
5d38a079 324 struct sk_buff *skb;
3b47d303 325 struct hrtimer timer;
404f7c9e 326 struct list_head dev_list;
af12fa6e
ET
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
bea3348e
SH
329};
330
d94d9fee 331enum {
bea3348e 332 NAPI_STATE_SCHED, /* Poll is scheduled */
39e6c820 333 NAPI_STATE_MISSED, /* reschedule a napi */
a0a46196 334 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 335 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
336 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
337 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
338 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
339};
340
341enum {
39e6c820
ED
342 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
343 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
344 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
345 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
346 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
347 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
348 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
349};
350
5b252f0c 351enum gro_result {
d1c76af9
HX
352 GRO_MERGED,
353 GRO_MERGED_FREE,
354 GRO_HELD,
355 GRO_NORMAL,
356 GRO_DROP,
25393d3f 357 GRO_CONSUMED,
d1c76af9 358};
5b252f0c 359typedef enum gro_result gro_result_t;
d1c76af9 360
8a4eb573
JP
361/*
362 * enum rx_handler_result - Possible return values for rx_handlers.
363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
364 * further.
365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366 * case skb->dev was changed by rx_handler.
367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
369 *
370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
371 * special processing of the skb, prior to delivery to protocol handlers.
372 *
373 * Currently, a net_device can only have a single rx_handler registered. Trying
374 * to register a second rx_handler will return -EBUSY.
375 *
376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377 * To unregister a rx_handler on a net_device, use
378 * netdev_rx_handler_unregister().
379 *
380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
381 * do with the skb.
382 *
5e82b4b2 383 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 385 * the skb to be delivered in some other way.
8a4eb573
JP
386 *
387 * If the rx_handler changed skb->dev, to divert the skb to another
388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389 * new device will be called if it exists.
390 *
5e82b4b2 391 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 393 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 394 *
5e82b4b2 395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
396 * delivered, it should return RX_HANDLER_PASS.
397 *
398 * A device without a registered rx_handler will behave as if rx_handler
399 * returned RX_HANDLER_PASS.
400 */
401
402enum rx_handler_result {
403 RX_HANDLER_CONSUMED,
404 RX_HANDLER_ANOTHER,
405 RX_HANDLER_EXACT,
406 RX_HANDLER_PASS,
407};
408typedef enum rx_handler_result rx_handler_result_t;
409typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 410
f629d208 411void __napi_schedule(struct napi_struct *n);
bc9ad166 412void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 413
4d29515f 414static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
415{
416 return test_bit(NAPI_STATE_DISABLE, &n->state);
417}
418
39e6c820 419bool napi_schedule_prep(struct napi_struct *n);
bea3348e
SH
420
421/**
422 * napi_schedule - schedule NAPI poll
5e82b4b2 423 * @n: NAPI context
bea3348e
SH
424 *
425 * Schedule NAPI poll routine to be called if it is not already
426 * running.
427 */
428static inline void napi_schedule(struct napi_struct *n)
429{
430 if (napi_schedule_prep(n))
431 __napi_schedule(n);
432}
433
bc9ad166
ED
434/**
435 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 436 * @n: NAPI context
bc9ad166
ED
437 *
438 * Variant of napi_schedule(), assuming hard irqs are masked.
439 */
440static inline void napi_schedule_irqoff(struct napi_struct *n)
441{
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
444}
445
bfe13f54 446/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 447static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
448{
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
4d29515f 451 return true;
bfe13f54 452 }
4d29515f 453 return false;
bfe13f54
RD
454}
455
364b6055 456bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
457/**
458 * napi_complete - NAPI processing complete
5e82b4b2 459 * @n: NAPI context
bea3348e
SH
460 *
461 * Mark NAPI processing as complete.
3b47d303 462 * Consider using napi_complete_done() instead.
364b6055 463 * Return false if device should avoid rearming interrupts.
bea3348e 464 */
364b6055 465static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
466{
467 return napi_complete_done(n, 0);
468}
bea3348e 469
af12fa6e
ET
470/**
471 * napi_hash_del - remove a NAPI from global table
5e82b4b2 472 * @napi: NAPI context
af12fa6e 473 *
5e82b4b2 474 * Warning: caller must observe RCU grace period
34cbe27e
ED
475 * before freeing memory containing @napi, if
476 * this function returns true.
93d05d4a 477 * Note: core networking stack automatically calls it
5e82b4b2 478 * from netif_napi_del().
93d05d4a 479 * Drivers might want to call this helper to combine all
5e82b4b2 480 * the needed RCU grace periods into a single one.
af12fa6e 481 */
34cbe27e 482bool napi_hash_del(struct napi_struct *napi);
af12fa6e 483
bea3348e
SH
484/**
485 * napi_disable - prevent NAPI from scheduling
5e82b4b2 486 * @n: NAPI context
bea3348e
SH
487 *
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
490 */
3b47d303 491void napi_disable(struct napi_struct *n);
bea3348e
SH
492
493/**
494 * napi_enable - enable NAPI scheduling
5e82b4b2 495 * @n: NAPI context
bea3348e
SH
496 *
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
499 */
500static inline void napi_enable(struct napi_struct *n)
501{
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 503 smp_mb__before_atomic();
bea3348e 504 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 505 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
506}
507
c264c3de
SH
508/**
509 * napi_synchronize - wait until NAPI is not running
5e82b4b2 510 * @n: NAPI context
c264c3de
SH
511 *
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
515 */
516static inline void napi_synchronize(const struct napi_struct *n)
517{
facc432f
AB
518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
520 msleep(1);
521 else
522 barrier();
c264c3de 523}
c264c3de 524
d94d9fee 525enum netdev_queue_state_t {
73466498
TH
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
c3f26a26 528 __QUEUE_STATE_FROZEN,
79d16385 529};
8e2f1a63
DB
530
531#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
534
535#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
537 QUEUE_STATE_FROZEN)
538#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
539 QUEUE_STATE_FROZEN)
540
73466498
TH
541/*
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
549 */
79d16385 550
bb949fbd 551struct netdev_queue {
6a321cb3 552/*
5e82b4b2 553 * read-mostly part
6a321cb3 554 */
bb949fbd 555 struct net_device *dev;
46e5da40 556 struct Qdisc __rcu *qdisc;
b0e1e646 557 struct Qdisc *qdisc_sleeping;
ccf5ff69 558#ifdef CONFIG_SYSFS
1d24eb48
TH
559 struct kobject kobj;
560#endif
f2cd2d3e
ED
561#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 int numa_node;
563#endif
c0ef079c
FW
564 unsigned long tx_maxrate;
565 /*
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
568 */
569 unsigned long trans_timeout;
6a321cb3 570/*
5e82b4b2 571 * write-mostly part
6a321cb3
ED
572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
9d21493b 575 /*
9b36627a 576 * Time (in jiffies) of last Tx
9d21493b
ED
577 */
578 unsigned long trans_start;
ccf5ff69 579
114cf580
TH
580 unsigned long state;
581
582#ifdef CONFIG_BQL
583 struct dql dql;
584#endif
e8a0464c 585} ____cacheline_aligned_in_smp;
bb949fbd 586
f2cd2d3e
ED
587static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
588{
589#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 return q->numa_node;
591#else
b236da69 592 return NUMA_NO_NODE;
f2cd2d3e
ED
593#endif
594}
595
596static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
597{
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 q->numa_node = node;
600#endif
601}
602
df334545 603#ifdef CONFIG_RPS
0a9627f2
TH
604/*
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
607 */
608struct rps_map {
609 unsigned int len;
610 struct rcu_head rcu;
611 u16 cpus[0];
612};
60b778ce 613#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 614
fec5e652 615/*
c445477d
BH
616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
fec5e652
TH
619 */
620struct rps_dev_flow {
621 u16 cpu;
c445477d 622 u16 filter;
fec5e652
TH
623 unsigned int last_qtail;
624};
c445477d 625#define RPS_NO_FILTER 0xffff
fec5e652
TH
626
627/*
628 * The rps_dev_flow_table structure contains a table of flow mappings.
629 */
630struct rps_dev_flow_table {
631 unsigned int mask;
632 struct rcu_head rcu;
fec5e652
TH
633 struct rps_dev_flow flows[0];
634};
635#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 636 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
637
638/*
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
567e4b79 643 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 646 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
647 */
648struct rps_sock_flow_table {
567e4b79 649 u32 mask;
93c1af6c
ED
650
651 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 652};
567e4b79 653#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
654
655#define RPS_NO_CPU 0xffff
656
567e4b79
ED
657extern u32 rps_cpu_mask;
658extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
659
fec5e652
TH
660static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
661 u32 hash)
662{
663 if (table && hash) {
567e4b79
ED
664 unsigned int index = hash & table->mask;
665 u32 val = hash & ~rps_cpu_mask;
fec5e652 666
5e82b4b2 667 /* We only give a hint, preemption can change CPU under us */
567e4b79 668 val |= raw_smp_processor_id();
fec5e652 669
567e4b79
ED
670 if (table->ents[index] != val)
671 table->ents[index] = val;
fec5e652
TH
672 }
673}
674
c445477d 675#ifdef CONFIG_RFS_ACCEL
f629d208
JP
676bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
677 u16 filter_id);
c445477d 678#endif
a953be53 679#endif /* CONFIG_RPS */
c445477d 680
0a9627f2
TH
681/* This structure contains an instance of an RX queue. */
682struct netdev_rx_queue {
a953be53 683#ifdef CONFIG_RPS
6e3f7faf
ED
684 struct rps_map __rcu *rps_map;
685 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 686#endif
6e3f7faf 687 struct kobject kobj;
fe822240 688 struct net_device *dev;
0a9627f2 689} ____cacheline_aligned_in_smp;
a953be53
MD
690
691/*
692 * RX queue sysfs structures and functions.
693 */
694struct rx_queue_attribute {
695 struct attribute attr;
718ad681 696 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
a953be53 697 ssize_t (*store)(struct netdev_rx_queue *queue,
718ad681 698 const char *buf, size_t len);
a953be53 699};
d314774c 700
bf264145
TH
701#ifdef CONFIG_XPS
702/*
703 * This structure holds an XPS map which can be of variable length. The
704 * map is an array of queues.
705 */
706struct xps_map {
707 unsigned int len;
708 unsigned int alloc_len;
709 struct rcu_head rcu;
710 u16 queues[0];
711};
60b778ce 712#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
713#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
714 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
715
716/*
717 * This structure holds all XPS maps for device. Maps are indexed by CPU.
718 */
719struct xps_dev_maps {
720 struct rcu_head rcu;
a4177869 721 struct xps_map __rcu *cpu_map[0];
bf264145 722};
184c449f
AD
723#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
724 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
725#endif /* CONFIG_XPS */
726
4f57c087
JF
727#define TC_MAX_QUEUE 16
728#define TC_BITMASK 15
729/* HW offloaded queuing disciplines txq count and offset maps */
730struct netdev_tc_txq {
731 u16 count;
732 u16 offset;
733};
734
68bad94e
NP
735#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
736/*
737 * This structure is to hold information about the device
738 * configured to run FCoE protocol stack.
739 */
740struct netdev_fcoe_hbainfo {
741 char manufacturer[64];
742 char serial_number[64];
743 char hardware_version[64];
744 char driver_version[64];
745 char optionrom_version[64];
746 char firmware_version[64];
747 char model[256];
748 char model_description[256];
749};
750#endif
751
02637fce 752#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 753
02637fce
JP
754/* This structure holds a unique identifier to identify some
755 * physical item (port for example) used by a netdevice.
66b52b0d 756 */
02637fce
JP
757struct netdev_phys_item_id {
758 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
759 unsigned char id_len;
760};
761
d754f98b
SF
762static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
763 struct netdev_phys_item_id *b)
764{
765 return a->id_len == b->id_len &&
766 memcmp(a->id, b->id, a->id_len) == 0;
767}
768
99932d4f
DB
769typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 struct sk_buff *skb);
771
2572ac53 772enum tc_setup_type {
16e5cc64 773 TC_SETUP_MQPRIO,
a1b7c5fd 774 TC_SETUP_CLSU32,
5b33f488 775 TC_SETUP_CLSFLOWER,
ade9b658 776 TC_SETUP_CLSMATCHALL,
332ae8e2 777 TC_SETUP_CLSBPF,
8c4083b3 778 TC_SETUP_BLOCK,
3d0bd028 779 TC_SETUP_CBS,
16e5cc64
JF
780};
781
f4e63525
JK
782/* These structures hold the attributes of bpf state that are being passed
783 * to the netdevice through the bpf op.
a7862b45 784 */
f4e63525 785enum bpf_netdev_command {
a7862b45
BB
786 /* Set or clear a bpf program used in the earliest stages of packet
787 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
788 * is responsible for calling bpf_prog_put on any old progs that are
789 * stored. In case of error, the callee need not release the new prog
790 * reference, but on success it takes ownership and must bpf_prog_put
791 * when it is no longer used.
792 */
793 XDP_SETUP_PROG,
ee5d032f 794 XDP_SETUP_PROG_HW,
a7862b45 795 /* Check if a bpf program is set on the device. The callee should
ce158e58
JK
796 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
797 * is equivalent to XDP_ATTACHED_DRV.
a7862b45
BB
798 */
799 XDP_QUERY_PROG,
800};
801
ddf9f970
JK
802struct netlink_ext_ack;
803
f4e63525
JK
804struct netdev_bpf {
805 enum bpf_netdev_command command;
a7862b45
BB
806 union {
807 /* XDP_SETUP_PROG */
ddf9f970 808 struct {
32d60277 809 u32 flags;
ddf9f970
JK
810 struct bpf_prog *prog;
811 struct netlink_ext_ack *extack;
812 };
a7862b45 813 /* XDP_QUERY_PROG */
58038695 814 struct {
ce158e58 815 u8 prog_attached;
58038695
MKL
816 u32 prog_id;
817 };
a7862b45
BB
818 };
819};
16e5cc64 820
d77e38e6
SK
821#ifdef CONFIG_XFRM_OFFLOAD
822struct xfrmdev_ops {
823 int (*xdo_dev_state_add) (struct xfrm_state *x);
824 void (*xdo_dev_state_delete) (struct xfrm_state *x);
825 void (*xdo_dev_state_free) (struct xfrm_state *x);
826 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
827 struct xfrm_state *x);
828};
829#endif
830
6c557001
FW
831struct dev_ifalias {
832 struct rcu_head rcuhead;
833 char ifalias[];
834};
835
d314774c
SH
836/*
837 * This structure defines the management hooks for network devices.
00829823
SH
838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
d314774c
SH
840 *
841 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
d314774c
SH
846 *
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
850 *
851 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 852 * This function is called when a network device transitions to the up
d314774c
SH
853 * state.
854 *
855 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 856 * This function is called when a network device transitions to the down
d314774c
SH
857 * state.
858 *
dc1f8bf6
SH
859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
00829823 861 * Called when a packet needs to be transmitted.
e79d8429
RR
862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 866 * Required; cannot be NULL.
00829823 867 *
1a2a1444
DM
868 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
869 * struct net_device *dev
870 * netdev_features_t features);
871 * Called by core transmit path to determine if device is capable of
872 * performing offload operations on a given packet. This is to give
873 * the device an opportunity to implement any restrictions that cannot
874 * be otherwise expressed by feature flags. The check is called with
875 * the set of features that the stack has calculated and it returns
876 * those the driver believes to be appropriate.
cdba756f 877 *
f663dd9a 878 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 879 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 880 * Called to decide which queue to use when device supports multiple
00829823
SH
881 * transmit queues.
882 *
d314774c
SH
883 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
884 * This function is called to allow device receiver to make
5e82b4b2 885 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
886 *
887 * void (*ndo_set_rx_mode)(struct net_device *dev);
888 * This function is called device changes address list filtering.
01789349 889 * If driver handles unicast address filtering, it should set
5e82b4b2 890 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
891 *
892 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
893 * This function is called when the Media Access Control address
37b607c5 894 * needs to be changed. If this interface is not defined, the
5e82b4b2 895 * MAC address can not be changed.
d314774c
SH
896 *
897 * int (*ndo_validate_addr)(struct net_device *dev);
898 * Test if Media Access Control address is valid for the device.
899 *
900 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
901 * Called when a user requests an ioctl which can't be handled by
902 * the generic interface code. If not defined ioctls return
d314774c
SH
903 * not supported error code.
904 *
905 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
906 * Used to set network devices bus interface parameters. This interface
5e82b4b2 907 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
908 * interface (PCI) for low level management.
909 *
910 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
911 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 912 * of a device.
d314774c 913 *
00829823 914 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 915 * Callback used when the transmitter has not made any progress
d314774c
SH
916 * for dev->watchdog ticks.
917 *
bc1f4470 918 * void (*ndo_get_stats64)(struct net_device *dev,
919 * struct rtnl_link_stats64 *storage);
d308e38f 920 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 921 * Called when a user wants to get the network device usage
be1f3c2c 922 * statistics. Drivers must do one of the following:
3cfde79c
BH
923 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
924 * rtnl_link_stats64 structure passed by the caller.
82695d9b 925 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
926 * (which should normally be dev->stats) and return a pointer to
927 * it. The structure may be changed asynchronously only if each
928 * field is written atomically.
929 * 3. Update dev->stats asynchronously and atomically, and define
930 * neither operation.
d314774c 931 *
3df5b3c6 932 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
933 * Return true if this device supports offload stats of this attr_id.
934 *
935 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
936 * void *attr_data)
937 * Get statistics for offload operations by attr_id. Write it into the
938 * attr_data pointer.
939 *
5d632cb7 940 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 941 * If device supports VLAN filtering this function is called when a
80d5c368 942 * VLAN id is registered.
d314774c 943 *
5d632cb7 944 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 945 * If device supports VLAN filtering this function is called when a
80d5c368 946 * VLAN id is unregistered.
d314774c
SH
947 *
948 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
949 *
950 * SR-IOV management functions.
951 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
952 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
953 * u8 qos, __be16 proto);
ed616689
SC
954 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
955 * int max_tx_rate);
5f8444a3 956 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 957 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
958 * int (*ndo_get_vf_config)(struct net_device *dev,
959 * int vf, struct ifla_vf_info *ivf);
1d8faf48 960 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
961 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
962 * struct nlattr *port[]);
01a3d796
VZ
963 *
964 * Enable or disable the VF ability to query its RSS Redirection Table and
965 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 966 * with PF and querying it may introduce a theoretical security risk.
01a3d796 967 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 968 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
2572ac53 969 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
de4784ca 970 * void *type_data);
6a4bc2b4
FF
971 * Called to setup any 'tc' scheduler, classifier or action on @dev.
972 * This is always called from the stack with the rtnl lock held and netif
973 * tx queues stopped. This allows the netdevice to perform queue
974 * management safely.
c445477d 975 *
e9bce845
YZ
976 * Fiber Channel over Ethernet (FCoE) offload functions.
977 * int (*ndo_fcoe_enable)(struct net_device *dev);
978 * Called when the FCoE protocol stack wants to start using LLD for FCoE
979 * so the underlying device can perform whatever needed configuration or
980 * initialization to support acceleration of FCoE traffic.
981 *
982 * int (*ndo_fcoe_disable)(struct net_device *dev);
983 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
984 * so the underlying device can perform whatever needed clean-ups to
985 * stop supporting acceleration of FCoE traffic.
986 *
987 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
988 * struct scatterlist *sgl, unsigned int sgc);
989 * Called when the FCoE Initiator wants to initialize an I/O that
990 * is a possible candidate for Direct Data Placement (DDP). The LLD can
991 * perform necessary setup and returns 1 to indicate the device is set up
992 * successfully to perform DDP on this I/O, otherwise this returns 0.
993 *
994 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
995 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
996 * indicated by the FC exchange id 'xid', so the underlying device can
997 * clean up and reuse resources for later DDP requests.
998 *
999 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1000 * struct scatterlist *sgl, unsigned int sgc);
1001 * Called when the FCoE Target wants to initialize an I/O that
1002 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1003 * perform necessary setup and returns 1 to indicate the device is set up
1004 * successfully to perform DDP on this I/O, otherwise this returns 0.
1005 *
68bad94e
NP
1006 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1007 * struct netdev_fcoe_hbainfo *hbainfo);
1008 * Called when the FCoE Protocol stack wants information on the underlying
1009 * device. This information is utilized by the FCoE protocol stack to
1010 * register attributes with Fiber Channel management service as per the
1011 * FC-GS Fabric Device Management Information(FDMI) specification.
1012 *
e9bce845
YZ
1013 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1014 * Called when the underlying device wants to override default World Wide
1015 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1016 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1017 * protocol stack to use.
1018 *
c445477d
BH
1019 * RFS acceleration.
1020 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1021 * u16 rxq_index, u32 flow_id);
1022 * Set hardware filter for RFS. rxq_index is the target queue index;
1023 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1024 * Return the filter ID on success, or a negative error code.
fbaec0ea 1025 *
8b98a70c 1026 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1027 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1028 * Called to make another netdev an underling.
1029 *
1030 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1031 * Called to release previously enslaved netdev.
5455c699
MM
1032 *
1033 * Feature/offload setting functions.
1a2a1444
DM
1034 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1035 * netdev_features_t features);
1036 * Adjusts the requested feature flags according to device-specific
1037 * constraints, and returns the resulting flags. Must not modify
1038 * the device state.
1039 *
c8f44aff 1040 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1041 * Called to update device configuration to new features. Passed
1042 * feature set might be less than what was returned by ndo_fix_features()).
1043 * Must return >0 or -errno if it changed dev->features itself.
1044 *
edc7d573 1045 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1046 * struct net_device *dev,
f6f6424b 1047 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1048 * Adds an FDB entry to dev for addr.
1690be63
VY
1049 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1050 * struct net_device *dev,
f6f6424b 1051 * const unsigned char *addr, u16 vid)
77162022
JF
1052 * Deletes the FDB entry from dev coresponding to addr.
1053 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1054 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1055 * int *idx)
77162022
JF
1056 * Used to add FDB entries to dump requests. Implementers should add
1057 * entries to skb and update idx with the number of entries.
e5a55a89 1058 *
ad41faa8
ND
1059 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1060 * u16 flags)
e5a55a89 1061 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1062 * struct net_device *dev, u32 filter_mask,
1063 * int nlflags)
ad41faa8
ND
1064 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1065 * u16 flags);
4bf84c35
JP
1066 *
1067 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1068 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1069 * which do not represent real hardware may define this to allow their
1070 * userspace components to manage their virtual carrier state. Devices
1071 * that determine carrier state from physical hardware properties (eg
1072 * network cables) or protocol-dependent mechanisms (eg
1073 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1074 *
1075 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1076 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1077 * Called to get ID of physical port of this device. If driver does
1078 * not implement this, it is assumed that the hw is not able to have
1079 * multiple net devices on single physical port.
53cf5275 1080 *
7c46a640
AD
1081 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1082 * struct udp_tunnel_info *ti);
1083 * Called by UDP tunnel to notify a driver about the UDP port and socket
1084 * address family that a UDP tunnel is listnening to. It is called only
1085 * when a new port starts listening. The operation is protected by the
1086 * RTNL.
1087 *
1088 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1089 * struct udp_tunnel_info *ti);
1090 * Called by UDP tunnel to notify the driver about a UDP port and socket
1091 * address family that the UDP tunnel is not listening to anymore. The
1092 * operation is protected by the RTNL.
1093 *
a6cc0cfa
JF
1094 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1095 * struct net_device *dev)
1096 * Called by upper layer devices to accelerate switching or other
1097 * station functionality into hardware. 'pdev is the lowerdev
1098 * to use for the offload and 'dev' is the net device that will
1099 * back the offload. Returns a pointer to the private structure
1100 * the upper layer will maintain.
1101 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1102 * Called by upper layer device to delete the station created
1103 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1104 * the station and priv is the structure returned by the add
1105 * operation.
822b3b2e
JF
1106 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1107 * int queue_index, u32 maxrate);
1108 * Called when a user wants to set a max-rate limitation of specific
1109 * TX queue.
a54acb3a
ND
1110 * int (*ndo_get_iflink)(const struct net_device *dev);
1111 * Called to get the iflink value of this device.
d746d707 1112 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1113 * bool proto_down);
d746d707
AK
1114 * This function is used to pass protocol port error state information
1115 * to the switch driver. The switch driver can react to the proto_down
1116 * by doing a phys down on the associated switch port.
fc4099f1
PS
1117 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1118 * This function is used to get egress tunnel information for given skb.
1119 * This is useful for retrieving outer tunnel header parameters while
1120 * sampling packet.
871b642a
PA
1121 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1122 * This function is used to specify the headroom that the skb must
1123 * consider when allocation skb during packet reception. Setting
1124 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1125 * forward. Setting a negative value resets the rx headroom to the
871b642a 1126 * default value.
f4e63525 1127 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
a7862b45 1128 * This function is used to set or query state related to XDP on the
f4e63525
JK
1129 * netdevice and manage BPF offload. See definition of
1130 * enum bpf_netdev_command for details.
814abfab
JF
1131 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1132 * This function is used to submit a XDP packet for transmit on a
1133 * netdevice.
11393cc9 1134 * void (*ndo_xdp_flush)(struct net_device *dev);
f16ded59
RR
1135 * This function is used to inform the driver to flush a particular
1136 * xdp tx queue. Must be called on same CPU as xdp_xmit.
d314774c
SH
1137 */
1138struct net_device_ops {
1139 int (*ndo_init)(struct net_device *dev);
1140 void (*ndo_uninit)(struct net_device *dev);
1141 int (*ndo_open)(struct net_device *dev);
1142 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1143 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1144 struct net_device *dev);
1145 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1146 struct net_device *dev,
1147 netdev_features_t features);
00829823 1148 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1149 struct sk_buff *skb,
99932d4f
DB
1150 void *accel_priv,
1151 select_queue_fallback_t fallback);
d314774c
SH
1152 void (*ndo_change_rx_flags)(struct net_device *dev,
1153 int flags);
d314774c 1154 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1155 int (*ndo_set_mac_address)(struct net_device *dev,
1156 void *addr);
d314774c 1157 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1158 int (*ndo_do_ioctl)(struct net_device *dev,
1159 struct ifreq *ifr, int cmd);
d314774c
SH
1160 int (*ndo_set_config)(struct net_device *dev,
1161 struct ifmap *map);
00829823
SH
1162 int (*ndo_change_mtu)(struct net_device *dev,
1163 int new_mtu);
1164 int (*ndo_neigh_setup)(struct net_device *dev,
1165 struct neigh_parms *);
d314774c
SH
1166 void (*ndo_tx_timeout) (struct net_device *dev);
1167
bc1f4470 1168 void (*ndo_get_stats64)(struct net_device *dev,
1169 struct rtnl_link_stats64 *storage);
3df5b3c6 1170 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1171 int (*ndo_get_offload_stats)(int attr_id,
1172 const struct net_device *dev,
1173 void *attr_data);
d314774c
SH
1174 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1175
8e586137 1176 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1177 __be16 proto, u16 vid);
8e586137 1178 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1179 __be16 proto, u16 vid);
d314774c 1180#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1181 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1182 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1183 struct netpoll_info *info);
0e34e931 1184 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1185#endif
95c26df8
WM
1186 int (*ndo_set_vf_mac)(struct net_device *dev,
1187 int queue, u8 *mac);
1188 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1189 int queue, u16 vlan,
1190 u8 qos, __be16 proto);
ed616689
SC
1191 int (*ndo_set_vf_rate)(struct net_device *dev,
1192 int vf, int min_tx_rate,
1193 int max_tx_rate);
5f8444a3
GR
1194 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1195 int vf, bool setting);
dd461d6a
HS
1196 int (*ndo_set_vf_trust)(struct net_device *dev,
1197 int vf, bool setting);
95c26df8
WM
1198 int (*ndo_get_vf_config)(struct net_device *dev,
1199 int vf,
1200 struct ifla_vf_info *ivf);
1d8faf48
RE
1201 int (*ndo_set_vf_link_state)(struct net_device *dev,
1202 int vf, int link_state);
3b766cd8
EBE
1203 int (*ndo_get_vf_stats)(struct net_device *dev,
1204 int vf,
1205 struct ifla_vf_stats
1206 *vf_stats);
57b61080
SF
1207 int (*ndo_set_vf_port)(struct net_device *dev,
1208 int vf,
1209 struct nlattr *port[]);
1210 int (*ndo_get_vf_port)(struct net_device *dev,
1211 int vf, struct sk_buff *skb);
cc8e27cc
EC
1212 int (*ndo_set_vf_guid)(struct net_device *dev,
1213 int vf, u64 guid,
1214 int guid_type);
01a3d796
VZ
1215 int (*ndo_set_vf_rss_query_en)(
1216 struct net_device *dev,
1217 int vf, bool setting);
16e5cc64 1218 int (*ndo_setup_tc)(struct net_device *dev,
2572ac53 1219 enum tc_setup_type type,
de4784ca 1220 void *type_data);
d11ead75 1221#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1222 int (*ndo_fcoe_enable)(struct net_device *dev);
1223 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1224 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1225 u16 xid,
1226 struct scatterlist *sgl,
1227 unsigned int sgc);
1228 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1229 u16 xid);
6247e086
YZ
1230 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1231 u16 xid,
1232 struct scatterlist *sgl,
1233 unsigned int sgc);
68bad94e
NP
1234 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1235 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1236#endif
1237
d11ead75 1238#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1239#define NETDEV_FCOE_WWNN 0
1240#define NETDEV_FCOE_WWPN 1
1241 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1242 u64 *wwn, int type);
4d288d57 1243#endif
3c9c36bc 1244
c445477d
BH
1245#ifdef CONFIG_RFS_ACCEL
1246 int (*ndo_rx_flow_steer)(struct net_device *dev,
1247 const struct sk_buff *skb,
1248 u16 rxq_index,
1249 u32 flow_id);
1250#endif
fbaec0ea 1251 int (*ndo_add_slave)(struct net_device *dev,
33eaf2a6
DA
1252 struct net_device *slave_dev,
1253 struct netlink_ext_ack *extack);
fbaec0ea
JP
1254 int (*ndo_del_slave)(struct net_device *dev,
1255 struct net_device *slave_dev);
c8f44aff
MM
1256 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1257 netdev_features_t features);
5455c699 1258 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1259 netdev_features_t features);
503eebc2
JP
1260 int (*ndo_neigh_construct)(struct net_device *dev,
1261 struct neighbour *n);
1262 void (*ndo_neigh_destroy)(struct net_device *dev,
1263 struct neighbour *n);
77162022
JF
1264
1265 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1266 struct nlattr *tb[],
77162022 1267 struct net_device *dev,
6b6e2725 1268 const unsigned char *addr,
f6f6424b 1269 u16 vid,
77162022
JF
1270 u16 flags);
1271 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1272 struct nlattr *tb[],
77162022 1273 struct net_device *dev,
f6f6424b
JP
1274 const unsigned char *addr,
1275 u16 vid);
77162022
JF
1276 int (*ndo_fdb_dump)(struct sk_buff *skb,
1277 struct netlink_callback *cb,
1278 struct net_device *dev,
5d5eacb3 1279 struct net_device *filter_dev,
d297653d 1280 int *idx);
e5a55a89
JF
1281
1282 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1283 struct nlmsghdr *nlh,
1284 u16 flags);
e5a55a89
JF
1285 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1286 u32 pid, u32 seq,
6cbdceeb 1287 struct net_device *dev,
46c264da
ND
1288 u32 filter_mask,
1289 int nlflags);
407af329 1290 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1291 struct nlmsghdr *nlh,
1292 u16 flags);
4bf84c35
JP
1293 int (*ndo_change_carrier)(struct net_device *dev,
1294 bool new_carrier);
66b52b0d 1295 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1296 struct netdev_phys_item_id *ppid);
db24a904
DA
1297 int (*ndo_get_phys_port_name)(struct net_device *dev,
1298 char *name, size_t len);
7c46a640
AD
1299 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1300 struct udp_tunnel_info *ti);
1301 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1302 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1303 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1304 struct net_device *dev);
1305 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1306 void *priv);
1307
25175ba5 1308 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1309 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1310 int queue_index,
1311 u32 maxrate);
a54acb3a 1312 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1313 int (*ndo_change_proto_down)(struct net_device *dev,
1314 bool proto_down);
fc4099f1
PS
1315 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1316 struct sk_buff *skb);
871b642a
PA
1317 void (*ndo_set_rx_headroom)(struct net_device *dev,
1318 int needed_headroom);
f4e63525
JK
1319 int (*ndo_bpf)(struct net_device *dev,
1320 struct netdev_bpf *bpf);
814abfab
JF
1321 int (*ndo_xdp_xmit)(struct net_device *dev,
1322 struct xdp_buff *xdp);
11393cc9 1323 void (*ndo_xdp_flush)(struct net_device *dev);
d314774c
SH
1324};
1325
7aa98047
LR
1326/**
1327 * enum net_device_priv_flags - &struct net_device priv_flags
1328 *
1329 * These are the &struct net_device, they are only set internally
1330 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1331 * userspace; this means that the order of these flags can change
7aa98047
LR
1332 * during any kernel release.
1333 *
1334 * You should have a pretty good reason to be extending these flags.
1335 *
1336 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1337 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1338 * @IFF_BONDING: bonding master or slave
7aa98047 1339 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1340 * @IFF_WAN_HDLC: WAN HDLC device
1341 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1342 * release skb->dst
1343 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1344 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1345 * @IFF_MACVLAN_PORT: device used as macvlan port
1346 * @IFF_BRIDGE_PORT: device used as bridge port
1347 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1348 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1349 * @IFF_UNICAST_FLT: Supports unicast filtering
1350 * @IFF_TEAM_PORT: device used as team port
1351 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1352 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1353 * change when it's running
1354 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1355 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1356 * underlying stacked devices
1357 * @IFF_IPVLAN_MASTER: IPvlan master device
1358 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1359 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1360 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1361 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1362 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1363 * @IFF_TEAM: device is a team device
d4ab4286 1364 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1365 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1366 * entity (i.e. the master device for bridged veth)
3c175784 1367 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1368 */
1369enum netdev_priv_flags {
1370 IFF_802_1Q_VLAN = 1<<0,
1371 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1372 IFF_BONDING = 1<<2,
1373 IFF_ISATAP = 1<<3,
1374 IFF_WAN_HDLC = 1<<4,
1375 IFF_XMIT_DST_RELEASE = 1<<5,
1376 IFF_DONT_BRIDGE = 1<<6,
1377 IFF_DISABLE_NETPOLL = 1<<7,
1378 IFF_MACVLAN_PORT = 1<<8,
1379 IFF_BRIDGE_PORT = 1<<9,
1380 IFF_OVS_DATAPATH = 1<<10,
1381 IFF_TX_SKB_SHARING = 1<<11,
1382 IFF_UNICAST_FLT = 1<<12,
1383 IFF_TEAM_PORT = 1<<13,
1384 IFF_SUPP_NOFCS = 1<<14,
1385 IFF_LIVE_ADDR_CHANGE = 1<<15,
1386 IFF_MACVLAN = 1<<16,
1387 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1388 IFF_IPVLAN_MASTER = 1<<18,
1389 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1390 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1391 IFF_NO_QUEUE = 1<<21,
1392 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1393 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1394 IFF_TEAM = 1<<24,
d4ab4286 1395 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1396 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1397 IFF_MACSEC = 1<<27,
7aa98047
LR
1398};
1399
1400#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1401#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1402#define IFF_BONDING IFF_BONDING
7aa98047 1403#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1404#define IFF_WAN_HDLC IFF_WAN_HDLC
1405#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1406#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1407#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1408#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1409#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1410#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1411#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1412#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1413#define IFF_TEAM_PORT IFF_TEAM_PORT
1414#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1415#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1416#define IFF_MACVLAN IFF_MACVLAN
02875878 1417#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1418#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1419#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1420#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1421#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1422#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1423#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1424#define IFF_TEAM IFF_TEAM
d4ab4286 1425#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1426#define IFF_MACSEC IFF_MACSEC
7aa98047 1427
536721b1
KK
1428/**
1429 * struct net_device - The DEVICE structure.
d651983d
MCC
1430 *
1431 * Actually, this whole structure is a big mistake. It mixes I/O
1432 * data with strictly "high-level" data, and it has to know about
1433 * almost every data structure used in the INET module.
536721b1
KK
1434 *
1435 * @name: This is the first field of the "visible" part of this structure
1436 * (i.e. as seen by users in the "Space.c" file). It is the name
d651983d 1437 * of the interface.
536721b1
KK
1438 *
1439 * @name_hlist: Device name hash chain, please keep it close to name[]
1440 * @ifalias: SNMP alias
1441 * @mem_end: Shared memory end
1442 * @mem_start: Shared memory start
1443 * @base_addr: Device I/O address
1444 * @irq: Device IRQ number
1445 *
14ffbbb8
TG
1446 * @carrier_changes: Stats to monitor carrier on<->off transitions
1447 *
536721b1
KK
1448 * @state: Generic network queuing layer state, see netdev_state_t
1449 * @dev_list: The global list of network devices
5e82b4b2
BH
1450 * @napi_list: List entry used for polling NAPI devices
1451 * @unreg_list: List entry when we are unregistering the
1452 * device; see the function unregister_netdev
1453 * @close_list: List entry used when we are closing the device
62d885fe
BP
1454 * @ptype_all: Device-specific packet handlers for all protocols
1455 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1456 *
1457 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1458 * @features: Currently active device features
1459 * @hw_features: User-changeable features
1460 *
1461 * @wanted_features: User-requested features
1462 * @vlan_features: Mask of features inheritable by VLAN devices
1463 *
1464 * @hw_enc_features: Mask of features inherited by encapsulating devices
1465 * This field indicates what encapsulation
1466 * offloads the hardware is capable of doing,
1467 * and drivers will need to set them appropriately.
1468 *
1469 * @mpls_features: Mask of features inheritable by MPLS
1470 *
1471 * @ifindex: interface index
5e82b4b2 1472 * @group: The group the device belongs to
536721b1
KK
1473 *
1474 * @stats: Statistics struct, which was left as a legacy, use
1475 * rtnl_link_stats64 instead
1476 *
1477 * @rx_dropped: Dropped packets by core network,
1478 * do not use this in drivers
1479 * @tx_dropped: Dropped packets by core network,
1480 * do not use this in drivers
6e7333d3
JW
1481 * @rx_nohandler: nohandler dropped packets by core network on
1482 * inactive devices, do not use this in drivers
536721b1 1483 *
536721b1
KK
1484 * @wireless_handlers: List of functions to handle Wireless Extensions,
1485 * instead of ioctl,
1486 * see <net/iw_handler.h> for details.
1487 * @wireless_data: Instance data managed by the core of wireless extensions
1488 *
1489 * @netdev_ops: Includes several pointers to callbacks,
1490 * if one wants to override the ndo_*() functions
1491 * @ethtool_ops: Management operations
f997c55c
AA
1492 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1493 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1494 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1495 * of Layer 2 headers.
1496 *
1497 * @flags: Interface flags (a la BSD)
1498 * @priv_flags: Like 'flags' but invisible to userspace,
1499 * see if.h for the definitions
1500 * @gflags: Global flags ( kept as legacy )
1501 * @padded: How much padding added by alloc_netdev()
1502 * @operstate: RFC2863 operstate
1503 * @link_mode: Mapping policy to operstate
1504 * @if_port: Selectable AUI, TP, ...
1505 * @dma: DMA channel
1506 * @mtu: Interface MTU value
61e84623
JW
1507 * @min_mtu: Interface Minimum MTU value
1508 * @max_mtu: Interface Maximum MTU value
536721b1 1509 * @type: Interface hardware type
2793a23a 1510 * @hard_header_len: Maximum hardware header length.
217e6fa2 1511 * @min_header_len: Minimum hardware header length
536721b1
KK
1512 *
1513 * @needed_headroom: Extra headroom the hardware may need, but not in all
1514 * cases can this be guaranteed
1515 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1516 * cases can this be guaranteed. Some cases also use
1517 * LL_MAX_HEADER instead to allocate the skb
1518 *
1519 * interface address info:
1520 *
1521 * @perm_addr: Permanent hw address
1522 * @addr_assign_type: Hw address assignment type
1523 * @addr_len: Hardware address length
8626a0c8 1524 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1525 * @dev_id: Used to differentiate devices that share
1526 * the same link layer address
1527 * @dev_port: Used to differentiate devices that share
1528 * the same function
1529 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1530 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1531 * has been enabled due to the need to listen to
1532 * additional unicast addresses in a device that
1533 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1534 * @uc: unicast mac addresses
1535 * @mc: multicast mac addresses
1536 * @dev_addrs: list of device hw addresses
1537 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1538 * @promiscuity: Number of times the NIC is told to work in
1539 * promiscuous mode; if it becomes 0 the NIC will
1540 * exit promiscuous mode
536721b1
KK
1541 * @allmulti: Counter, enables or disables allmulticast mode
1542 *
1543 * @vlan_info: VLAN info
1544 * @dsa_ptr: dsa specific data
1545 * @tipc_ptr: TIPC specific data
1546 * @atalk_ptr: AppleTalk link
1547 * @ip_ptr: IPv4 specific data
1548 * @dn_ptr: DECnet specific data
1549 * @ip6_ptr: IPv6 specific data
1550 * @ax25_ptr: AX.25 specific data
1551 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1552 *
536721b1
KK
1553 * @dev_addr: Hw address (before bcast,
1554 * because most packets are unicast)
1555 *
1556 * @_rx: Array of RX queues
1557 * @num_rx_queues: Number of RX queues
1558 * allocated at register_netdev() time
1559 * @real_num_rx_queues: Number of RX queues currently active in device
1560 *
1561 * @rx_handler: handler for received packets
1562 * @rx_handler_data: XXX: need comments on this one
46209401
JP
1563 * @miniq_ingress: ingress/clsact qdisc specific data for
1564 * ingress processing
536721b1
KK
1565 * @ingress_queue: XXX: need comments on this one
1566 * @broadcast: hw bcast address
1567 *
14ffbbb8
TG
1568 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1569 * indexed by RX queue number. Assigned by driver.
1570 * This must only be set if the ndo_rx_flow_steer
1571 * operation is defined
1572 * @index_hlist: Device index hash chain
1573 *
536721b1
KK
1574 * @_tx: Array of TX queues
1575 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1576 * @real_num_tx_queues: Number of TX queues currently active in device
1577 * @qdisc: Root qdisc from userspace point of view
1578 * @tx_queue_len: Max frames per queue allowed
1579 * @tx_global_lock: XXX: need comments on this one
1580 *
1581 * @xps_maps: XXX: need comments on this one
46209401
JP
1582 * @miniq_egress: clsact qdisc specific data for
1583 * egress processing
536721b1 1584 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1585 * the watchdog (see dev_watchdog())
536721b1
KK
1586 * @watchdog_timer: List of timers
1587 *
1588 * @pcpu_refcnt: Number of references to this device
1589 * @todo_list: Delayed register/unregister
536721b1
KK
1590 * @link_watch_list: XXX: need comments on this one
1591 *
1592 * @reg_state: Register/unregister state machine
1593 * @dismantle: Device is going to be freed
1594 * @rtnl_link_state: This enum represents the phases of creating
1595 * a new link
1596 *
cf124db5
DM
1597 * @needs_free_netdev: Should unregister perform free_netdev?
1598 * @priv_destructor: Called from unregister
536721b1
KK
1599 * @npinfo: XXX: need comments on this one
1600 * @nd_net: Network namespace this network device is inside
1601 *
1602 * @ml_priv: Mid-layer private
1603 * @lstats: Loopback statistics
1604 * @tstats: Tunnel statistics
1605 * @dstats: Dummy statistics
1606 * @vstats: Virtual ethernet statistics
1607 *
1608 * @garp_port: GARP
1609 * @mrp_port: MRP
1610 *
1611 * @dev: Class/net/name entry
1612 * @sysfs_groups: Space for optional device, statistics and wireless
1613 * sysfs groups
1614 *
1615 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1616 * @rtnl_link_ops: Rtnl_link_ops
1617 *
1618 * @gso_max_size: Maximum size of generic segmentation offload
1619 * @gso_max_segs: Maximum number of segments that can be passed to the
1620 * NIC for GSO
1621 *
1622 * @dcbnl_ops: Data Center Bridging netlink ops
1623 * @num_tc: Number of traffic classes in the net device
1624 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1625 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1626 *
1627 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1628 *
1629 * @priomap: XXX: need comments on this one
1630 * @phydev: Physical device may attach itself
1631 * for hardware timestamping
1632 *
123b3652
ED
1633 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1634 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1635 *
d746d707
AK
1636 * @proto_down: protocol port state information can be sent to the
1637 * switch driver and used to set the phys state of the
1638 * switch port.
1639 *
1da177e4
LT
1640 * FIXME: cleanup struct net_device such that network protocol info
1641 * moves out.
1642 */
1643
d94d9fee 1644struct net_device {
1da177e4 1645 char name[IFNAMSIZ];
9356b8fc 1646 struct hlist_node name_hlist;
6c557001 1647 struct dev_ifalias __rcu *ifalias;
1da177e4
LT
1648 /*
1649 * I/O specific fields
1650 * FIXME: Merge these and struct ifmap into one
1651 */
536721b1
KK
1652 unsigned long mem_end;
1653 unsigned long mem_start;
1654 unsigned long base_addr;
1655 int irq;
1da177e4 1656
14ffbbb8
TG
1657 atomic_t carrier_changes;
1658
1da177e4 1659 /*
536721b1
KK
1660 * Some hardware also needs these fields (state,dev_list,
1661 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1662 * part of the usual set specified in Space.c.
1663 */
1664
1da177e4
LT
1665 unsigned long state;
1666
7562f876 1667 struct list_head dev_list;
bea3348e 1668 struct list_head napi_list;
44a0873d 1669 struct list_head unreg_list;
5cde2829 1670 struct list_head close_list;
7866a621
SN
1671 struct list_head ptype_all;
1672 struct list_head ptype_specific;
2f268f12 1673
2f268f12
VF
1674 struct {
1675 struct list_head upper;
1676 struct list_head lower;
1677 } adj_list;
1678
c8f44aff 1679 netdev_features_t features;
c8f44aff 1680 netdev_features_t hw_features;
c8f44aff 1681 netdev_features_t wanted_features;
c8f44aff 1682 netdev_features_t vlan_features;
6a674e9c 1683 netdev_features_t hw_enc_features;
0d89d203 1684 netdev_features_t mpls_features;
802ab55a 1685 netdev_features_t gso_partial_features;
04ed3e74 1686
1da177e4 1687 int ifindex;
7a66bbc9 1688 int group;
1da177e4 1689
c45d286e 1690 struct net_device_stats stats;
015f0688 1691
015f0688
ED
1692 atomic_long_t rx_dropped;
1693 atomic_long_t tx_dropped;
6e7333d3 1694 atomic_long_t rx_nohandler;
1da177e4 1695
b86e0280 1696#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1697 const struct iw_handler_def *wireless_handlers;
1698 struct iw_public_data *wireless_data;
b86e0280 1699#endif
d314774c 1700 const struct net_device_ops *netdev_ops;
76fd8593 1701 const struct ethtool_ops *ethtool_ops;
4170604f 1702#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1703 const struct switchdev_ops *switchdev_ops;
4170604f 1704#endif
1b69c6d0
DA
1705#ifdef CONFIG_NET_L3_MASTER_DEV
1706 const struct l3mdev_ops *l3mdev_ops;
1707#endif
f997c55c
AA
1708#if IS_ENABLED(CONFIG_IPV6)
1709 const struct ndisc_ops *ndisc_ops;
1710#endif
1da177e4 1711
d77e38e6
SK
1712#ifdef CONFIG_XFRM
1713 const struct xfrmdev_ops *xfrmdev_ops;
1714#endif
1715
3b04ddde
SH
1716 const struct header_ops *header_ops;
1717
536721b1
KK
1718 unsigned int flags;
1719 unsigned int priv_flags;
1720
1da177e4 1721 unsigned short gflags;
536721b1 1722 unsigned short padded;
1da177e4 1723
536721b1
KK
1724 unsigned char operstate;
1725 unsigned char link_mode;
b00055aa 1726
536721b1
KK
1727 unsigned char if_port;
1728 unsigned char dma;
bdc220da 1729
536721b1 1730 unsigned int mtu;
61e84623
JW
1731 unsigned int min_mtu;
1732 unsigned int max_mtu;
536721b1
KK
1733 unsigned short type;
1734 unsigned short hard_header_len;
d92be7a4 1735 unsigned char min_header_len;
1da177e4 1736
f5184d26
JB
1737 unsigned short needed_headroom;
1738 unsigned short needed_tailroom;
1739
1da177e4 1740 /* Interface address info. */
536721b1
KK
1741 unsigned char perm_addr[MAX_ADDR_LEN];
1742 unsigned char addr_assign_type;
1743 unsigned char addr_len;
a0a9663d 1744 unsigned short neigh_priv_len;
536721b1
KK
1745 unsigned short dev_id;
1746 unsigned short dev_port;
ccffad25 1747 spinlock_t addr_list_lock;
14ffbbb8
TG
1748 unsigned char name_assign_type;
1749 bool uc_promisc;
536721b1
KK
1750 struct netdev_hw_addr_list uc;
1751 struct netdev_hw_addr_list mc;
1752 struct netdev_hw_addr_list dev_addrs;
1753
4c3d5e7b
ED
1754#ifdef CONFIG_SYSFS
1755 struct kset *queues_kset;
1756#endif
9d45abe1
WC
1757 unsigned int promiscuity;
1758 unsigned int allmulti;
1da177e4 1759
1da177e4 1760
5e82b4b2 1761 /* Protocol-specific pointers */
65ac6a5f 1762
d11ead75 1763#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1764 struct vlan_info __rcu *vlan_info;
65ac6a5f 1765#endif
34a430d7 1766#if IS_ENABLED(CONFIG_NET_DSA)
2f657a60 1767 struct dsa_port *dsa_ptr;
37cb0620
YX
1768#endif
1769#if IS_ENABLED(CONFIG_TIPC)
536721b1 1770 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1771#endif
536721b1
KK
1772 void *atalk_ptr;
1773 struct in_device __rcu *ip_ptr;
1774 struct dn_dev __rcu *dn_ptr;
1775 struct inet6_dev __rcu *ip6_ptr;
1776 void *ax25_ptr;
1777 struct wireless_dev *ieee80211_ptr;
98a18b6f 1778 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1779#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1780 struct mpls_dev __rcu *mpls_ptr;
1781#endif
1da177e4 1782
9356b8fc 1783/*
cd13539b 1784 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1785 */
9356b8fc 1786 /* Interface address info used in eth_type_trans() */
536721b1 1787 unsigned char *dev_addr;
f001fde5 1788
a953be53 1789#ifdef CONFIG_SYSFS
0a9627f2
TH
1790 struct netdev_rx_queue *_rx;
1791
0a9627f2 1792 unsigned int num_rx_queues;
62fe0b40 1793 unsigned int real_num_rx_queues;
df334545 1794#endif
0a9627f2 1795
7acedaf5 1796 struct bpf_prog __rcu *xdp_prog;
3b47d303 1797 unsigned long gro_flush_timeout;
61391cde 1798 rx_handler_func_t __rcu *rx_handler;
1799 void __rcu *rx_handler_data;
e8a0464c 1800
4cda01e8 1801#ifdef CONFIG_NET_CLS_ACT
46209401 1802 struct mini_Qdisc __rcu *miniq_ingress;
d2788d34 1803#endif
24824a09 1804 struct netdev_queue __rcu *ingress_queue;
e687ad60 1805#ifdef CONFIG_NETFILTER_INGRESS
960632ec 1806 struct nf_hook_entries __rcu *nf_hooks_ingress;
e687ad60 1807#endif
d2788d34 1808
536721b1 1809 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1810#ifdef CONFIG_RFS_ACCEL
1811 struct cpu_rmap *rx_cpu_rmap;
1812#endif
1813 struct hlist_node index_hlist;
cd13539b
ED
1814
1815/*
1816 * Cache lines mostly used on transmit path
1817 */
e8a0464c
DM
1818 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1819 unsigned int num_tx_queues;
fd2ea0a7 1820 unsigned int real_num_tx_queues;
af356afa 1821 struct Qdisc *qdisc;
59cc1f61
JK
1822#ifdef CONFIG_NET_SCHED
1823 DECLARE_HASHTABLE (qdisc_hash, 4);
1824#endif
0cd29503 1825 unsigned int tx_queue_len;
c3f26a26 1826 spinlock_t tx_global_lock;
14ffbbb8 1827 int watchdog_timeo;
cd13539b 1828
bf264145 1829#ifdef CONFIG_XPS
a4177869 1830 struct xps_dev_maps __rcu *xps_maps;
bf264145 1831#endif
1f211a1b 1832#ifdef CONFIG_NET_CLS_ACT
46209401 1833 struct mini_Qdisc __rcu *miniq_egress;
1f211a1b 1834#endif
0c4f691f 1835
9356b8fc 1836 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1837 struct timer_list watchdog_timer;
1838
29b4433d 1839 int __percpu *pcpu_refcnt;
1da177e4 1840 struct list_head todo_list;
1da177e4 1841
e014debe 1842 struct list_head link_watch_list;
572a103d 1843
1da177e4 1844 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1845 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1846 NETREG_UNREGISTERING, /* called unregister_netdevice */
1847 NETREG_UNREGISTERED, /* completed unregister todo */
1848 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1849 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1850 } reg_state:8;
1851
536721b1 1852 bool dismantle;
a2835763
PM
1853
1854 enum {
1855 RTNL_LINK_INITIALIZED,
1856 RTNL_LINK_INITIALIZING,
1857 } rtnl_link_state:16;
1da177e4 1858
cf124db5
DM
1859 bool needs_free_netdev;
1860 void (*priv_destructor)(struct net_device *dev);
1da177e4 1861
1da177e4 1862#ifdef CONFIG_NETPOLL
5fbee843 1863 struct netpoll_info __rcu *npinfo;
1da177e4 1864#endif
eae792b7 1865
0c5c9fb5 1866 possible_net_t nd_net;
4a1c5371 1867
4951704b 1868 /* mid-layer private */
a7855c78 1869 union {
536721b1
KK
1870 void *ml_priv;
1871 struct pcpu_lstats __percpu *lstats;
8f84985f 1872 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1873 struct pcpu_dstats __percpu *dstats;
1874 struct pcpu_vstats __percpu *vstats;
a7855c78 1875 };
536721b1 1876
fb585b44 1877#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1878 struct garp_port __rcu *garp_port;
fb585b44
TK
1879#endif
1880#if IS_ENABLED(CONFIG_MRP)
febf018d 1881 struct mrp_port __rcu *mrp_port;
fb585b44 1882#endif
1da177e4 1883
5e82b4b2 1884 struct device dev;
0c509a6c 1885 const struct attribute_group *sysfs_groups[4];
a953be53 1886 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1887
38f7b870 1888 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1889
82cc1a7a
PWJ
1890 /* for setting kernel sock attribute on TCP connection setup */
1891#define GSO_MAX_SIZE 65536
1892 unsigned int gso_max_size;
30b678d8
BH
1893#define GSO_MAX_SEGS 65535
1894 u16 gso_max_segs;
743b03a8 1895
7a6b6f51 1896#ifdef CONFIG_DCB
32953543 1897 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1898#endif
5e82b4b2
BH
1899 u8 num_tc;
1900 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1901 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1902
d11ead75 1903#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1904 unsigned int fcoe_ddp_xid;
5bc1421e 1905#endif
86f8515f 1906#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1907 struct netprio_map __rcu *priomap;
4d288d57 1908#endif
5e82b4b2
BH
1909 struct phy_device *phydev;
1910 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1911 struct lock_class_key *qdisc_running_key;
5e82b4b2 1912 bool proto_down;
1da177e4 1913};
43cb76d9 1914#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1915
b5cdae32
DM
1916static inline bool netif_elide_gro(const struct net_device *dev)
1917{
1918 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1919 return true;
1920 return false;
1921}
1922
1da177e4 1923#define NETDEV_ALIGN 32
1da177e4 1924
4f57c087
JF
1925static inline
1926int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1927{
1928 return dev->prio_tc_map[prio & TC_BITMASK];
1929}
1930
1931static inline
1932int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1933{
1934 if (tc >= dev->num_tc)
1935 return -EINVAL;
1936
1937 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1938 return 0;
1939}
1940
8d059b0f 1941int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1942void netdev_reset_tc(struct net_device *dev);
1943int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1944int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1945
1946static inline
1947int netdev_get_num_tc(struct net_device *dev)
1948{
1949 return dev->num_tc;
1950}
1951
e8a0464c
DM
1952static inline
1953struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1954 unsigned int index)
1955{
1956 return &dev->_tx[index];
1957}
1958
10c51b56
DB
1959static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1960 const struct sk_buff *skb)
1961{
1962 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1963}
1964
e8a0464c
DM
1965static inline void netdev_for_each_tx_queue(struct net_device *dev,
1966 void (*f)(struct net_device *,
1967 struct netdev_queue *,
1968 void *),
1969 void *arg)
1970{
1971 unsigned int i;
1972
1973 for (i = 0; i < dev->num_tx_queues; i++)
1974 f(dev, &dev->_tx[i], arg);
1975}
1976
d3fff6c4
ED
1977#define netdev_lockdep_set_classes(dev) \
1978{ \
1979 static struct lock_class_key qdisc_tx_busylock_key; \
1980 static struct lock_class_key qdisc_running_key; \
1981 static struct lock_class_key qdisc_xmit_lock_key; \
1982 static struct lock_class_key dev_addr_list_lock_key; \
1983 unsigned int i; \
1984 \
1985 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1986 (dev)->qdisc_running_key = &qdisc_running_key; \
1987 lockdep_set_class(&(dev)->addr_list_lock, \
1988 &dev_addr_list_lock_key); \
1989 for (i = 0; i < (dev)->num_tx_queues; i++) \
1990 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1991 &qdisc_xmit_lock_key); \
1992}
1993
f629d208 1994struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1995 struct sk_buff *skb,
1996 void *accel_priv);
8c4c49df 1997
871b642a
PA
1998/* returns the headroom that the master device needs to take in account
1999 * when forwarding to this dev
2000 */
2001static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2002{
2003 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2004}
2005
2006static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2007{
2008 if (dev->netdev_ops->ndo_set_rx_headroom)
2009 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2010}
2011
2012/* set the device rx headroom to the dev's default */
2013static inline void netdev_reset_rx_headroom(struct net_device *dev)
2014{
2015 netdev_set_rx_headroom(dev, -1);
2016}
2017
c346dca1
YH
2018/*
2019 * Net namespace inlines
2020 */
2021static inline
2022struct net *dev_net(const struct net_device *dev)
2023{
c2d9ba9b 2024 return read_pnet(&dev->nd_net);
c346dca1
YH
2025}
2026
2027static inline
f5aa23fd 2028void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2029{
0c5c9fb5 2030 write_pnet(&dev->nd_net, net);
c346dca1
YH
2031}
2032
bea3348e
SH
2033/**
2034 * netdev_priv - access network device private data
2035 * @dev: network device
2036 *
2037 * Get network device private data
2038 */
6472ce60 2039static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2040{
1ce8e7b5 2041 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2042}
2043
1da177e4
LT
2044/* Set the sysfs physical device reference for the network logical device
2045 * if set prior to registration will cause a symlink during initialization.
2046 */
43cb76d9 2047#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2048
384912ed 2049/* Set the sysfs device type for the network logical device to allow
3f79410c 2050 * fine-grained identification of different network device types. For
5e82b4b2 2051 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2052 */
2053#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2054
82dc3c63
ED
2055/* Default NAPI poll() weight
2056 * Device drivers are strongly advised to not use bigger value
2057 */
2058#define NAPI_POLL_WEIGHT 64
2059
3b582cc1 2060/**
5e82b4b2 2061 * netif_napi_add - initialize a NAPI context
3b582cc1 2062 * @dev: network device
5e82b4b2 2063 * @napi: NAPI context
3b582cc1
SH
2064 * @poll: polling function
2065 * @weight: default weight
2066 *
5e82b4b2
BH
2067 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2068 * *any* of the other NAPI-related functions.
3b582cc1 2069 */
d565b0a1
HX
2070void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2071 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2072
d64b5e85 2073/**
5e82b4b2 2074 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2075 * @dev: network device
5e82b4b2 2076 * @napi: NAPI context
d64b5e85
ED
2077 * @poll: polling function
2078 * @weight: default weight
2079 *
2080 * This variant of netif_napi_add() should be used from drivers using NAPI
2081 * to exclusively poll a TX queue.
2082 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2083 */
2084static inline void netif_tx_napi_add(struct net_device *dev,
2085 struct napi_struct *napi,
2086 int (*poll)(struct napi_struct *, int),
2087 int weight)
2088{
2089 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2090 netif_napi_add(dev, napi, poll, weight);
2091}
2092
d8156534 2093/**
5e82b4b2
BH
2094 * netif_napi_del - remove a NAPI context
2095 * @napi: NAPI context
d8156534 2096 *
5e82b4b2 2097 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2098 */
d565b0a1
HX
2099void netif_napi_del(struct napi_struct *napi);
2100
2101struct napi_gro_cb {
78a478d0 2102 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2103 void *frag0;
78a478d0 2104
7489594c
HX
2105 /* Length of frag0. */
2106 unsigned int frag0_len;
2107
86911732 2108 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2109 int data_offset;
86911732 2110
d565b0a1 2111 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2112 u16 flush;
2113
2114 /* Save the IP ID here and check when we get to the transport layer */
2115 u16 flush_id;
d565b0a1
HX
2116
2117 /* Number of segments aggregated. */
2e71a6f8
ED
2118 u16 count;
2119
15e2396d
TH
2120 /* Start offset for remote checksum offload */
2121 u16 gro_remcsum_start;
2122
2e71a6f8
ED
2123 /* jiffies when first packet was created/queued */
2124 unsigned long age;
86347245 2125
afe93325 2126 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2127 u16 proto;
2128
baa32ff4
TH
2129 /* This is non-zero if the packet may be of the same flow. */
2130 u8 same_flow:1;
2131
fac8e0f5
JG
2132 /* Used in tunnel GRO receive */
2133 u8 encap_mark:1;
573e8fca
TH
2134
2135 /* GRO checksum is valid */
2136 u8 csum_valid:1;
2137
662880f4
TH
2138 /* Number of checksums via CHECKSUM_UNNECESSARY */
2139 u8 csum_cnt:3;
c3c7c254 2140
baa32ff4
TH
2141 /* Free the skb? */
2142 u8 free:2;
2143#define NAPI_GRO_FREE 1
2144#define NAPI_GRO_FREE_STOLEN_HEAD 2
2145
efc98d08
TH
2146 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2147 u8 is_ipv6:1;
2148
a0ca153f
AD
2149 /* Used in GRE, set in fou/gue_gro_receive */
2150 u8 is_fou:1;
2151
1530545e
AD
2152 /* Used to determine if flush_id can be ignored */
2153 u8 is_atomic:1;
2154
fcd91dd4
SD
2155 /* Number of gro_receive callbacks this packet already went through */
2156 u8 recursion_counter:4;
2157
2158 /* 1 bit hole */
baa32ff4 2159
bf5a755f
JC
2160 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2161 __wsum csum;
2162
c3c7c254
ED
2163 /* used in skb_gro_receive() slow path */
2164 struct sk_buff *last;
d565b0a1
HX
2165};
2166
2167#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2168
fcd91dd4
SD
2169#define GRO_RECURSION_LIMIT 15
2170static inline int gro_recursion_inc_test(struct sk_buff *skb)
2171{
2172 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2173}
2174
2175typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2176static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2177 struct sk_buff **head,
2178 struct sk_buff *skb)
2179{
2180 if (unlikely(gro_recursion_inc_test(skb))) {
2181 NAPI_GRO_CB(skb)->flush |= 1;
2182 return NULL;
2183 }
2184
2185 return cb(head, skb);
2186}
2187
2188typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2189 struct sk_buff *);
2190static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2191 struct sock *sk,
2192 struct sk_buff **head,
2193 struct sk_buff *skb)
2194{
2195 if (unlikely(gro_recursion_inc_test(skb))) {
2196 NAPI_GRO_CB(skb)->flush |= 1;
2197 return NULL;
2198 }
2199
2200 return cb(sk, head, skb);
2201}
2202
1da177e4 2203struct packet_type {
f2ccd8fa
DM
2204 __be16 type; /* This is really htons(ether_type). */
2205 struct net_device *dev; /* NULL is wildcarded here */
2206 int (*func) (struct sk_buff *,
2207 struct net_device *,
2208 struct packet_type *,
2209 struct net_device *);
c0de08d0
EL
2210 bool (*id_match)(struct packet_type *ptype,
2211 struct sock *sk);
1da177e4
LT
2212 void *af_packet_priv;
2213 struct list_head list;
2214};
2215
f191a1d1 2216struct offload_callbacks {
576a30eb 2217 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2218 netdev_features_t features);
d565b0a1 2219 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2220 struct sk_buff *skb);
299603e8 2221 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2222};
2223
2224struct packet_offload {
2225 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2226 u16 priority;
f191a1d1
VY
2227 struct offload_callbacks callbacks;
2228 struct list_head list;
1da177e4
LT
2229};
2230
5e82b4b2 2231/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2232struct pcpu_sw_netstats {
2233 u64 rx_packets;
2234 u64 rx_bytes;
2235 u64 tx_packets;
2236 u64 tx_bytes;
2237 struct u64_stats_sync syncp;
2238};
2239
aabc92bb
PNA
2240#define __netdev_alloc_pcpu_stats(type, gfp) \
2241({ \
2242 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2243 if (pcpu_stats) { \
2244 int __cpu; \
2245 for_each_possible_cpu(__cpu) { \
2246 typeof(type) *stat; \
2247 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2248 u64_stats_init(&stat->syncp); \
2249 } \
2250 } \
2251 pcpu_stats; \
1c213bd2
WC
2252})
2253
aabc92bb 2254#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2255 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2256
764f5e54
JP
2257enum netdev_lag_tx_type {
2258 NETDEV_LAG_TX_TYPE_UNKNOWN,
2259 NETDEV_LAG_TX_TYPE_RANDOM,
2260 NETDEV_LAG_TX_TYPE_BROADCAST,
2261 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2262 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2263 NETDEV_LAG_TX_TYPE_HASH,
2264};
2265
2266struct netdev_lag_upper_info {
2267 enum netdev_lag_tx_type tx_type;
2268};
2269
fb1b2e3c
JP
2270struct netdev_lag_lower_state_info {
2271 u8 link_up : 1,
2272 tx_enabled : 1;
2273};
2274
1da177e4
LT
2275#include <linux/notifier.h>
2276
dcfe1421
AW
2277/* netdevice notifier chain. Please remember to update the rtnetlink
2278 * notification exclusion list in rtnetlink_event() when adding new
2279 * types.
2280 */
2281#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2282#define NETDEV_DOWN 0x0002
2283#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2284 detected a hardware crash and restarted
2285 - we can use this eg to kick tcp sessions
2286 once done */
2287#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2288#define NETDEV_REGISTER 0x0005
2289#define NETDEV_UNREGISTER 0x0006
1d486bfb 2290#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2291#define NETDEV_CHANGEADDR 0x0008
2292#define NETDEV_GOING_DOWN 0x0009
2293#define NETDEV_CHANGENAME 0x000A
2294#define NETDEV_FEAT_CHANGE 0x000B
2295#define NETDEV_BONDING_FAILOVER 0x000C
2296#define NETDEV_PRE_UP 0x000D
2297#define NETDEV_PRE_TYPE_CHANGE 0x000E
2298#define NETDEV_POST_TYPE_CHANGE 0x000F
2299#define NETDEV_POST_INIT 0x0010
0115e8e3 2300#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2301#define NETDEV_RELEASE 0x0012
2302#define NETDEV_NOTIFY_PEERS 0x0013
2303#define NETDEV_JOIN 0x0014
42e52bf9 2304#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2305#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2306#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2307#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2308#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2309#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2310#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2311#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
296d8ee3 2312#define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
08294a26 2313#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2314
f629d208
JP
2315int register_netdevice_notifier(struct notifier_block *nb);
2316int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2317
2318struct netdev_notifier_info {
51d0c047
DA
2319 struct net_device *dev;
2320 struct netlink_ext_ack *extack;
351638e7
JP
2321};
2322
be9efd36
JP
2323struct netdev_notifier_change_info {
2324 struct netdev_notifier_info info; /* must be first */
2325 unsigned int flags_changed;
2326};
2327
0e4ead9d
JP
2328struct netdev_notifier_changeupper_info {
2329 struct netdev_notifier_info info; /* must be first */
2330 struct net_device *upper_dev; /* new upper dev */
2331 bool master; /* is upper dev master */
5e82b4b2 2332 bool linking; /* is the notification for link or unlink */
29bf24af 2333 void *upper_info; /* upper dev info */
0e4ead9d
JP
2334};
2335
04d48266
JP
2336struct netdev_notifier_changelowerstate_info {
2337 struct netdev_notifier_info info; /* must be first */
2338 void *lower_state_info; /* is lower dev state */
2339};
2340
75538c2b
CW
2341static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2342 struct net_device *dev)
2343{
2344 info->dev = dev;
51d0c047 2345 info->extack = NULL;
75538c2b
CW
2346}
2347
351638e7
JP
2348static inline struct net_device *
2349netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2350{
2351 return info->dev;
2352}
2353
51d0c047
DA
2354static inline struct netlink_ext_ack *
2355netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2356{
2357 return info->extack;
2358}
2359
f629d208 2360int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2361
2362
1da177e4
LT
2363extern rwlock_t dev_base_lock; /* Device list lock */
2364
881d966b
EB
2365#define for_each_netdev(net, d) \
2366 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2367#define for_each_netdev_reverse(net, d) \
2368 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2369#define for_each_netdev_rcu(net, d) \
2370 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2371#define for_each_netdev_safe(net, d, n) \
2372 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2373#define for_each_netdev_continue(net, d) \
2374 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2375#define for_each_netdev_continue_rcu(net, d) \
2376 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2377#define for_each_netdev_in_bond_rcu(bond, slave) \
2378 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2379 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2380#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2381
a050c33f
DL
2382static inline struct net_device *next_net_device(struct net_device *dev)
2383{
2384 struct list_head *lh;
2385 struct net *net;
2386
c346dca1 2387 net = dev_net(dev);
a050c33f
DL
2388 lh = dev->dev_list.next;
2389 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2390}
2391
ce81b76a
ED
2392static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2393{
2394 struct list_head *lh;
2395 struct net *net;
2396
2397 net = dev_net(dev);
ccf43438 2398 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2399 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2400}
2401
a050c33f
DL
2402static inline struct net_device *first_net_device(struct net *net)
2403{
2404 return list_empty(&net->dev_base_head) ? NULL :
2405 net_device_entry(net->dev_base_head.next);
2406}
7562f876 2407
ccf43438
ED
2408static inline struct net_device *first_net_device_rcu(struct net *net)
2409{
2410 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2411
2412 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2413}
2414
f629d208
JP
2415int netdev_boot_setup_check(struct net_device *dev);
2416unsigned long netdev_boot_base(const char *prefix, int unit);
2417struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2418 const char *hwaddr);
2419struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2420struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2421void dev_add_pack(struct packet_type *pt);
2422void dev_remove_pack(struct packet_type *pt);
2423void __dev_remove_pack(struct packet_type *pt);
2424void dev_add_offload(struct packet_offload *po);
2425void dev_remove_offload(struct packet_offload *po);
f629d208 2426
a54acb3a 2427int dev_get_iflink(const struct net_device *dev);
fc4099f1 2428int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2429struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2430 unsigned short mask);
f629d208
JP
2431struct net_device *dev_get_by_name(struct net *net, const char *name);
2432struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2433struct net_device *__dev_get_by_name(struct net *net, const char *name);
2434int dev_alloc_name(struct net_device *dev, const char *name);
2435int dev_open(struct net_device *dev);
7051b88a 2436void dev_close(struct net_device *dev);
2437void dev_close_many(struct list_head *head, bool unlink);
f629d208 2438void dev_disable_lro(struct net_device *dev);
0c4b51f0 2439int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2440int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2441int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2442int register_netdevice(struct net_device *dev);
2443void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2444void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2445static inline void unregister_netdevice(struct net_device *dev)
2446{
2447 unregister_netdevice_queue(dev, NULL);
2448}
2449
f629d208
JP
2450int netdev_refcnt_read(const struct net_device *dev);
2451void free_netdev(struct net_device *dev);
74d332c1 2452void netdev_freemem(struct net_device *dev);
f629d208
JP
2453void synchronize_net(void);
2454int init_dummy_netdev(struct net_device *dev);
937f1ba5 2455
f60e5990 2456DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2457#define XMIT_RECURSION_LIMIT 10
2458
f60e5990 2459static inline int dev_recursion_level(void)
2460{
2461 return this_cpu_read(xmit_recursion);
2462}
2463
f629d208
JP
2464struct net_device *dev_get_by_index(struct net *net, int ifindex);
2465struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2466struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2467struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2468int netdev_get_name(struct net *net, char *name, int ifindex);
2469int dev_restart(struct net_device *dev);
f629d208 2470int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2471
2472static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2473{
2474 return NAPI_GRO_CB(skb)->data_offset;
2475}
2476
2477static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2478{
2479 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2480}
2481
2482static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2483{
2484 NAPI_GRO_CB(skb)->data_offset += len;
2485}
2486
a5b1cf28
HX
2487static inline void *skb_gro_header_fast(struct sk_buff *skb,
2488 unsigned int offset)
86911732 2489{
a5b1cf28
HX
2490 return NAPI_GRO_CB(skb)->frag0 + offset;
2491}
78a478d0 2492
a5b1cf28
HX
2493static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2494{
2495 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2496}
78a478d0 2497
57ea52a8
HX
2498static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2499{
2500 NAPI_GRO_CB(skb)->frag0 = NULL;
2501 NAPI_GRO_CB(skb)->frag0_len = 0;
2502}
2503
a5b1cf28
HX
2504static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2505 unsigned int offset)
2506{
17dd759c
HX
2507 if (!pskb_may_pull(skb, hlen))
2508 return NULL;
2509
57ea52a8 2510 skb_gro_frag0_invalidate(skb);
17dd759c 2511 return skb->data + offset;
86911732 2512}
1da177e4 2513
36e7b1b8
HX
2514static inline void *skb_gro_network_header(struct sk_buff *skb)
2515{
78d3fd0b
HX
2516 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2517 skb_network_offset(skb);
36e7b1b8
HX
2518}
2519
bf5a755f
JC
2520static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2521 const void *start, unsigned int len)
2522{
573e8fca 2523 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2524 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2525 csum_partial(start, len, 0));
2526}
2527
573e8fca
TH
2528/* GRO checksum functions. These are logical equivalents of the normal
2529 * checksum functions (in skbuff.h) except that they operate on the GRO
2530 * offsets and fields in sk_buff.
2531 */
2532
2533__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2534
15e2396d
TH
2535static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2536{
b7fe10e5 2537 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2538}
2539
573e8fca
TH
2540static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2541 bool zero_okay,
2542 __sum16 check)
2543{
6edec0e6
TH
2544 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2545 skb_checksum_start_offset(skb) <
2546 skb_gro_offset(skb)) &&
15e2396d 2547 !skb_at_gro_remcsum_start(skb) &&
662880f4 2548 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2549 (!zero_okay || check));
2550}
2551
2552static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2553 __wsum psum)
2554{
2555 if (NAPI_GRO_CB(skb)->csum_valid &&
2556 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2557 return 0;
2558
2559 NAPI_GRO_CB(skb)->csum = psum;
2560
2561 return __skb_gro_checksum_complete(skb);
2562}
2563
573e8fca
TH
2564static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2565{
662880f4
TH
2566 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2567 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2568 NAPI_GRO_CB(skb)->csum_cnt--;
2569 } else {
2570 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2571 * verified a new top level checksum or an encapsulated one
2572 * during GRO. This saves work if we fallback to normal path.
2573 */
2574 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2575 }
2576}
2577
2578#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2579 compute_pseudo) \
2580({ \
2581 __sum16 __ret = 0; \
2582 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2583 __ret = __skb_gro_checksum_validate_complete(skb, \
2584 compute_pseudo(skb, proto)); \
219f1d79 2585 if (!__ret) \
573e8fca
TH
2586 skb_gro_incr_csum_unnecessary(skb); \
2587 __ret; \
2588})
2589
2590#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2591 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2592
2593#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2594 compute_pseudo) \
2595 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2596
2597#define skb_gro_checksum_simple_validate(skb) \
2598 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2599
d96535a1
TH
2600static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2601{
2602 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2603 !NAPI_GRO_CB(skb)->csum_valid);
2604}
2605
2606static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2607 __sum16 check, __wsum pseudo)
2608{
2609 NAPI_GRO_CB(skb)->csum = ~pseudo;
2610 NAPI_GRO_CB(skb)->csum_valid = 1;
2611}
2612
2613#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2614do { \
2615 if (__skb_gro_checksum_convert_check(skb)) \
2616 __skb_gro_checksum_convert(skb, check, \
2617 compute_pseudo(skb, proto)); \
2618} while (0)
2619
26c4f7da
TH
2620struct gro_remcsum {
2621 int offset;
2622 __wsum delta;
2623};
2624
2625static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2626{
846cd667 2627 grc->offset = 0;
26c4f7da
TH
2628 grc->delta = 0;
2629}
2630
b7fe10e5
TH
2631static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2632 unsigned int off, size_t hdrlen,
2633 int start, int offset,
2634 struct gro_remcsum *grc,
2635 bool nopartial)
dcdc8994
TH
2636{
2637 __wsum delta;
b7fe10e5 2638 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2639
2640 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2641
15e2396d 2642 if (!nopartial) {
b7fe10e5
TH
2643 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2644 return ptr;
2645 }
2646
2647 ptr = skb_gro_header_fast(skb, off);
2648 if (skb_gro_header_hard(skb, off + plen)) {
2649 ptr = skb_gro_header_slow(skb, off + plen, off);
2650 if (!ptr)
2651 return NULL;
15e2396d
TH
2652 }
2653
b7fe10e5
TH
2654 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2655 start, offset);
dcdc8994
TH
2656
2657 /* Adjust skb->csum since we changed the packet */
dcdc8994 2658 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2659
b7fe10e5 2660 grc->offset = off + hdrlen + offset;
26c4f7da 2661 grc->delta = delta;
b7fe10e5
TH
2662
2663 return ptr;
dcdc8994
TH
2664}
2665
26c4f7da
TH
2666static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2667 struct gro_remcsum *grc)
2668{
b7fe10e5
TH
2669 void *ptr;
2670 size_t plen = grc->offset + sizeof(u16);
2671
26c4f7da
TH
2672 if (!grc->delta)
2673 return;
2674
b7fe10e5
TH
2675 ptr = skb_gro_header_fast(skb, grc->offset);
2676 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2677 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2678 if (!ptr)
2679 return;
2680 }
2681
2682 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2683}
dcdc8994 2684
25393d3f
SK
2685#ifdef CONFIG_XFRM_OFFLOAD
2686static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2687{
2688 if (PTR_ERR(pp) != -EINPROGRESS)
2689 NAPI_GRO_CB(skb)->flush |= flush;
2690}
2691#else
5f114163
SK
2692static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2693{
2694 NAPI_GRO_CB(skb)->flush |= flush;
2695}
25393d3f 2696#endif
5f114163 2697
0c4e8581
SH
2698static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2699 unsigned short type,
3b04ddde 2700 const void *daddr, const void *saddr,
95c96174 2701 unsigned int len)
0c4e8581 2702{
f1ecfd5d 2703 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2704 return 0;
3b04ddde
SH
2705
2706 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2707}
2708
b95cce35
SH
2709static inline int dev_parse_header(const struct sk_buff *skb,
2710 unsigned char *haddr)
2711{
2712 const struct net_device *dev = skb->dev;
2713
1b83336b 2714 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2715 return 0;
3b04ddde 2716 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2717}
2718
2793a23a
WB
2719/* ll_header must have at least hard_header_len allocated */
2720static inline bool dev_validate_header(const struct net_device *dev,
2721 char *ll_header, int len)
2722{
2723 if (likely(len >= dev->hard_header_len))
2724 return true;
217e6fa2
WB
2725 if (len < dev->min_header_len)
2726 return false;
2793a23a
WB
2727
2728 if (capable(CAP_SYS_RAWIO)) {
2729 memset(ll_header + len, 0, dev->hard_header_len - len);
2730 return true;
2731 }
2732
2733 if (dev->header_ops && dev->header_ops->validate)
2734 return dev->header_ops->validate(ll_header, len);
2735
2736 return false;
2737}
2738
1da177e4 2739typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2740int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2741static inline int unregister_gifconf(unsigned int family)
2742{
2743 return register_gifconf(family, NULL);
2744}
2745
99bbc707 2746#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2747#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2748struct sd_flow_limit {
2749 u64 count;
2750 unsigned int num_buckets;
2751 unsigned int history_head;
2752 u16 history[FLOW_LIMIT_HISTORY];
2753 u8 buckets[];
2754};
2755
2756extern int netdev_flow_limit_table_len;
2757#endif /* CONFIG_NET_FLOW_LIMIT */
2758
1da177e4 2759/*
5e82b4b2 2760 * Incoming packets are placed on per-CPU queues
1da177e4 2761 */
d94d9fee 2762struct softnet_data {
1da177e4 2763 struct list_head poll_list;
6e7676c1 2764 struct sk_buff_head process_queue;
1da177e4 2765
dee42870 2766 /* stats */
cd7b5396
DM
2767 unsigned int processed;
2768 unsigned int time_squeeze;
cd7b5396 2769 unsigned int received_rps;
fd793d89 2770#ifdef CONFIG_RPS
88751275 2771 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2772#endif
2773#ifdef CONFIG_NET_FLOW_LIMIT
2774 struct sd_flow_limit __rcu *flow_limit;
2775#endif
2776 struct Qdisc *output_queue;
2777 struct Qdisc **output_queue_tailp;
2778 struct sk_buff *completion_queue;
88751275 2779
4cdb1e2e 2780#ifdef CONFIG_RPS
501e7ef5
ED
2781 /* input_queue_head should be written by cpu owning this struct,
2782 * and only read by other cpus. Worth using a cache line.
2783 */
2784 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2785
2786 /* Elements below can be accessed between CPUs for RPS/RFS */
966a9671 2787 call_single_data_t csd ____cacheline_aligned_in_smp;
88751275
ED
2788 struct softnet_data *rps_ipi_next;
2789 unsigned int cpu;
76cc8b13 2790 unsigned int input_queue_tail;
1e94d72f 2791#endif
95c96174 2792 unsigned int dropped;
0a9627f2 2793 struct sk_buff_head input_pkt_queue;
bea3348e 2794 struct napi_struct backlog;
99bbc707 2795
1da177e4
LT
2796};
2797
76cc8b13 2798static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2799{
2800#ifdef CONFIG_RPS
76cc8b13
TH
2801 sd->input_queue_head++;
2802#endif
2803}
2804
2805static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2806 unsigned int *qtail)
2807{
2808#ifdef CONFIG_RPS
2809 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2810#endif
2811}
2812
0a9627f2 2813DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2814
f629d208 2815void __netif_schedule(struct Qdisc *q);
46e5da40 2816void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2817
fd2ea0a7
DM
2818static inline void netif_tx_schedule_all(struct net_device *dev)
2819{
2820 unsigned int i;
2821
2822 for (i = 0; i < dev->num_tx_queues; i++)
2823 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2824}
2825
f9a7cbbf 2826static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2827{
73466498 2828 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2829}
2830
bea3348e
SH
2831/**
2832 * netif_start_queue - allow transmit
2833 * @dev: network device
2834 *
2835 * Allow upper layers to call the device hard_start_xmit routine.
2836 */
1da177e4
LT
2837static inline void netif_start_queue(struct net_device *dev)
2838{
e8a0464c 2839 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2840}
2841
fd2ea0a7
DM
2842static inline void netif_tx_start_all_queues(struct net_device *dev)
2843{
2844 unsigned int i;
2845
2846 for (i = 0; i < dev->num_tx_queues; i++) {
2847 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2848 netif_tx_start_queue(txq);
2849 }
2850}
2851
46e5da40 2852void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2853
d29f749e
DJ
2854/**
2855 * netif_wake_queue - restart transmit
2856 * @dev: network device
2857 *
2858 * Allow upper layers to call the device hard_start_xmit routine.
2859 * Used for flow control when transmit resources are available.
2860 */
79d16385
DM
2861static inline void netif_wake_queue(struct net_device *dev)
2862{
e8a0464c 2863 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2864}
2865
fd2ea0a7
DM
2866static inline void netif_tx_wake_all_queues(struct net_device *dev)
2867{
2868 unsigned int i;
2869
2870 for (i = 0; i < dev->num_tx_queues; i++) {
2871 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2872 netif_tx_wake_queue(txq);
2873 }
2874}
2875
f9a7cbbf 2876static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2877{
73466498 2878 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2879}
2880
bea3348e
SH
2881/**
2882 * netif_stop_queue - stop transmitted packets
2883 * @dev: network device
2884 *
2885 * Stop upper layers calling the device hard_start_xmit routine.
2886 * Used for flow control when transmit resources are unavailable.
2887 */
1da177e4
LT
2888static inline void netif_stop_queue(struct net_device *dev)
2889{
e8a0464c 2890 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2891}
2892
a2029240 2893void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2894
4d29515f 2895static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2896{
73466498 2897 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2898}
2899
bea3348e
SH
2900/**
2901 * netif_queue_stopped - test if transmit queue is flowblocked
2902 * @dev: network device
2903 *
2904 * Test if transmit queue on device is currently unable to send.
2905 */
4d29515f 2906static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2907{
e8a0464c 2908 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2909}
2910
4d29515f 2911static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2912{
73466498
TH
2913 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2914}
2915
8e2f1a63
DB
2916static inline bool
2917netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2918{
2919 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2920}
2921
8e2f1a63
DB
2922static inline bool
2923netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2924{
2925 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2926}
2927
53511453
ED
2928/**
2929 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2930 * @dev_queue: pointer to transmit queue
2931 *
2932 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2933 * to give appropriate hint to the CPU.
53511453
ED
2934 */
2935static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2936{
2937#ifdef CONFIG_BQL
2938 prefetchw(&dev_queue->dql.num_queued);
2939#endif
2940}
2941
2942/**
2943 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2944 * @dev_queue: pointer to transmit queue
2945 *
2946 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2947 * to give appropriate hint to the CPU.
53511453
ED
2948 */
2949static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2950{
2951#ifdef CONFIG_BQL
2952 prefetchw(&dev_queue->dql.limit);
2953#endif
2954}
2955
c5d67bd7
TH
2956static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2957 unsigned int bytes)
2958{
114cf580
TH
2959#ifdef CONFIG_BQL
2960 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2961
2962 if (likely(dql_avail(&dev_queue->dql) >= 0))
2963 return;
2964
2965 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2966
2967 /*
2968 * The XOFF flag must be set before checking the dql_avail below,
2969 * because in netdev_tx_completed_queue we update the dql_completed
2970 * before checking the XOFF flag.
2971 */
2972 smp_mb();
2973
2974 /* check again in case another CPU has just made room avail */
2975 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2976 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2977#endif
c5d67bd7
TH
2978}
2979
0042d0c8
FF
2980/**
2981 * netdev_sent_queue - report the number of bytes queued to hardware
2982 * @dev: network device
2983 * @bytes: number of bytes queued to the hardware device queue
2984 *
2985 * Report the number of bytes queued for sending/completion to the network
2986 * device hardware queue. @bytes should be a good approximation and should
2987 * exactly match netdev_completed_queue() @bytes
2988 */
c5d67bd7
TH
2989static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2990{
2991 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2992}
2993
2994static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2995 unsigned int pkts, unsigned int bytes)
c5d67bd7 2996{
114cf580 2997#ifdef CONFIG_BQL
b37c0fbe
AD
2998 if (unlikely(!bytes))
2999 return;
3000
3001 dql_completed(&dev_queue->dql, bytes);
3002
3003 /*
3004 * Without the memory barrier there is a small possiblity that
3005 * netdev_tx_sent_queue will miss the update and cause the queue to
3006 * be stopped forever
3007 */
3008 smp_mb();
3009
3010 if (dql_avail(&dev_queue->dql) < 0)
3011 return;
3012
3013 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3014 netif_schedule_queue(dev_queue);
114cf580 3015#endif
c5d67bd7
TH
3016}
3017
0042d0c8
FF
3018/**
3019 * netdev_completed_queue - report bytes and packets completed by device
3020 * @dev: network device
3021 * @pkts: actual number of packets sent over the medium
3022 * @bytes: actual number of bytes sent over the medium
3023 *
3024 * Report the number of bytes and packets transmitted by the network device
3025 * hardware queue over the physical medium, @bytes must exactly match the
3026 * @bytes amount passed to netdev_sent_queue()
3027 */
c5d67bd7 3028static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3029 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3030{
3031 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3032}
3033
3034static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3035{
114cf580 3036#ifdef CONFIG_BQL
5c490354 3037 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3038 dql_reset(&q->dql);
3039#endif
c5d67bd7
TH
3040}
3041
0042d0c8
FF
3042/**
3043 * netdev_reset_queue - reset the packets and bytes count of a network device
3044 * @dev_queue: network device
3045 *
3046 * Reset the bytes and packet count of a network device and clear the
3047 * software flow control OFF bit for this network device
3048 */
c5d67bd7
TH
3049static inline void netdev_reset_queue(struct net_device *dev_queue)
3050{
3051 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3052}
3053
b9507bda
DB
3054/**
3055 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3056 * @dev: network device
3057 * @queue_index: given tx queue index
3058 *
3059 * Returns 0 if given tx queue index >= number of device tx queues,
3060 * otherwise returns the originally passed tx queue index.
3061 */
3062static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3063{
3064 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3065 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3066 dev->name, queue_index,
3067 dev->real_num_tx_queues);
3068 return 0;
3069 }
3070
3071 return queue_index;
3072}
3073
bea3348e
SH
3074/**
3075 * netif_running - test if up
3076 * @dev: network device
3077 *
3078 * Test if the device has been brought up.
3079 */
4d29515f 3080static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3081{
3082 return test_bit(__LINK_STATE_START, &dev->state);
3083}
3084
f25f4e44 3085/*
5e82b4b2 3086 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3087 * stop, and a check if it's stopped. All other device management is
3088 * done at the overall netdevice level.
3089 * Also test the device if we're multiqueue.
3090 */
bea3348e
SH
3091
3092/**
3093 * netif_start_subqueue - allow sending packets on subqueue
3094 * @dev: network device
3095 * @queue_index: sub queue index
3096 *
3097 * Start individual transmit queue of a device with multiple transmit queues.
3098 */
f25f4e44
PWJ
3099static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3100{
fd2ea0a7 3101 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3102
3103 netif_tx_start_queue(txq);
f25f4e44
PWJ
3104}
3105
bea3348e
SH
3106/**
3107 * netif_stop_subqueue - stop sending packets on subqueue
3108 * @dev: network device
3109 * @queue_index: sub queue index
3110 *
3111 * Stop individual transmit queue of a device with multiple transmit queues.
3112 */
f25f4e44
PWJ
3113static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3114{
fd2ea0a7 3115 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3116 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3117}
3118
bea3348e
SH
3119/**
3120 * netif_subqueue_stopped - test status of subqueue
3121 * @dev: network device
3122 * @queue_index: sub queue index
3123 *
3124 * Check individual transmit queue of a device with multiple transmit queues.
3125 */
4d29515f
DM
3126static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3127 u16 queue_index)
f25f4e44 3128{
fd2ea0a7 3129 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3130
3131 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3132}
3133
4d29515f
DM
3134static inline bool netif_subqueue_stopped(const struct net_device *dev,
3135 struct sk_buff *skb)
668f895a
PE
3136{
3137 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3138}
bea3348e 3139
738b35cc
FF
3140/**
3141 * netif_wake_subqueue - allow sending packets on subqueue
3142 * @dev: network device
3143 * @queue_index: sub queue index
3144 *
3145 * Resume individual transmit queue of a device with multiple transmit queues.
3146 */
3147static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3148{
3149 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3150
3151 netif_tx_wake_queue(txq);
3152}
f25f4e44 3153
537c00de 3154#ifdef CONFIG_XPS
53af53ae 3155int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3156 u16 index);
537c00de
AD
3157#else
3158static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3159 const struct cpumask *mask,
537c00de
AD
3160 u16 index)
3161{
3162 return 0;
3163}
3164#endif
3165
5605c762
JP
3166u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3167 unsigned int num_tx_queues);
3168
a3d22a68
VZ
3169/*
3170 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3171 * as a distribution range limit for the returned value.
3172 */
3173static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3174 struct sk_buff *skb)
a3d22a68
VZ
3175{
3176 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3177}
3178
bea3348e
SH
3179/**
3180 * netif_is_multiqueue - test if device has multiple transmit queues
3181 * @dev: network device
3182 *
3183 * Check if device has multiple transmit queues
bea3348e 3184 */
4d29515f 3185static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3186{
a02cec21 3187 return dev->num_tx_queues > 1;
f25f4e44 3188}
1da177e4 3189
f629d208 3190int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3191
a953be53 3192#ifdef CONFIG_SYSFS
f629d208 3193int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3194#else
3195static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3196 unsigned int rxq)
3197{
3198 return 0;
3199}
3200#endif
3201
a953be53
MD
3202#ifdef CONFIG_SYSFS
3203static inline unsigned int get_netdev_rx_queue_index(
3204 struct netdev_rx_queue *queue)
3205{
3206 struct net_device *dev = queue->dev;
3207 int index = queue - dev->_rx;
3208
3209 BUG_ON(index >= dev->num_rx_queues);
3210 return index;
3211}
3212#endif
3213
16917b87 3214#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3215int netif_get_num_default_rss_queues(void);
16917b87 3216
e6247027
ED
3217enum skb_free_reason {
3218 SKB_REASON_CONSUMED,
3219 SKB_REASON_DROPPED,
3220};
3221
3222void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3223void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3224
e6247027
ED
3225/*
3226 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3227 * interrupt context or with hardware interrupts being disabled.
3228 * (in_irq() || irqs_disabled())
3229 *
3230 * We provide four helpers that can be used in following contexts :
3231 *
3232 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3233 * replacing kfree_skb(skb)
3234 *
3235 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3236 * Typically used in place of consume_skb(skb) in TX completion path
3237 *
3238 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3239 * replacing kfree_skb(skb)
3240 *
3241 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3242 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3243 */
e6247027
ED
3244static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3245{
3246 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3247}
3248
3249static inline void dev_consume_skb_irq(struct sk_buff *skb)
3250{
3251 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3252}
3253
3254static inline void dev_kfree_skb_any(struct sk_buff *skb)
3255{
3256 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3257}
3258
3259static inline void dev_consume_skb_any(struct sk_buff *skb)
3260{
3261 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3262}
1da177e4 3263
7c497478
JW
3264void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3265int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
f629d208
JP
3266int netif_rx(struct sk_buff *skb);
3267int netif_rx_ni(struct sk_buff *skb);
04eb4489 3268int netif_receive_skb(struct sk_buff *skb);
1c601d82 3269int netif_receive_skb_core(struct sk_buff *skb);
f629d208
JP
3270gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3271void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3272struct sk_buff *napi_get_frags(struct napi_struct *napi);
3273gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3274struct packet_offload *gro_find_receive_by_type(__be16 type);
3275struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3276
3277static inline void napi_free_frags(struct napi_struct *napi)
3278{
3279 kfree_skb(napi->skb);
3280 napi->skb = NULL;
3281}
3282
24b27fc4 3283bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3284int netdev_rx_handler_register(struct net_device *dev,
3285 rx_handler_func_t *rx_handler,
3286 void *rx_handler_data);
3287void netdev_rx_handler_unregister(struct net_device *dev);
3288
3289bool dev_valid_name(const char *name);
3290int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3291int dev_ethtool(struct net *net, struct ifreq *);
3292unsigned int dev_get_flags(const struct net_device *);
3293int __dev_change_flags(struct net_device *, unsigned int flags);
3294int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3295void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3296 unsigned int gchanges);
f629d208
JP
3297int dev_change_name(struct net_device *, const char *);
3298int dev_set_alias(struct net_device *, const char *, size_t);
6c557001 3299int dev_get_alias(const struct net_device *, char *, size_t);
f629d208 3300int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3301int __dev_set_mtu(struct net_device *, int);
f629d208
JP
3302int dev_set_mtu(struct net_device *, int);
3303void dev_set_group(struct net_device *, int);
3304int dev_set_mac_address(struct net_device *, struct sockaddr *);
3305int dev_change_carrier(struct net_device *, bool new_carrier);
3306int dev_get_phys_port_id(struct net_device *dev,
02637fce 3307 struct netdev_phys_item_id *ppid);
db24a904
DA
3308int dev_get_phys_port_name(struct net_device *dev,
3309 char *name, size_t len);
d746d707 3310int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3311struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3312struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3313 struct netdev_queue *txq, int *ret);
d67b9cd2 3314
f4e63525 3315typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
d67b9cd2
DB
3316int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3317 int fd, u32 flags);
f4e63525 3318u8 __dev_xdp_attached(struct net_device *dev, bpf_op_t xdp_op, u32 *prog_id);
d67b9cd2 3319
a0265d28 3320int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3321int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3322bool is_skb_forwardable(const struct net_device *dev,
3323 const struct sk_buff *skb);
1da177e4 3324
4e3264d2
MKL
3325static __always_inline int ____dev_forward_skb(struct net_device *dev,
3326 struct sk_buff *skb)
3327{
3328 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3329 unlikely(!is_skb_forwardable(dev, skb))) {
3330 atomic_long_inc(&dev->rx_dropped);
3331 kfree_skb(skb);
3332 return NET_RX_DROP;
3333 }
3334
3335 skb_scrub_packet(skb, true);
3336 skb->priority = 0;
3337 return 0;
3338}
3339
74b20582
DA
3340void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3341
20380731 3342extern int netdev_budget;
7acf8a1e 3343extern unsigned int netdev_budget_usecs;
1da177e4
LT
3344
3345/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3346void netdev_run_todo(void);
1da177e4 3347
bea3348e
SH
3348/**
3349 * dev_put - release reference to device
3350 * @dev: network device
3351 *
9ef4429b 3352 * Release reference to device to allow it to be freed.
bea3348e 3353 */
1da177e4
LT
3354static inline void dev_put(struct net_device *dev)
3355{
933393f5 3356 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3357}
3358
bea3348e
SH
3359/**
3360 * dev_hold - get reference to device
3361 * @dev: network device
3362 *
9ef4429b 3363 * Hold reference to device to keep it from being freed.
bea3348e 3364 */
15333061
SH
3365static inline void dev_hold(struct net_device *dev)
3366{
933393f5 3367 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3368}
1da177e4
LT
3369
3370/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3371 * and _off may be called from IRQ context, but it is caller
3372 * who is responsible for serialization of these calls.
b00055aa
SR
3373 *
3374 * The name carrier is inappropriate, these functions should really be
3375 * called netif_lowerlayer_*() because they represent the state of any
3376 * kind of lower layer not just hardware media.
1da177e4
LT
3377 */
3378
f629d208
JP
3379void linkwatch_init_dev(struct net_device *dev);
3380void linkwatch_fire_event(struct net_device *dev);
3381void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3382
bea3348e
SH
3383/**
3384 * netif_carrier_ok - test if carrier present
3385 * @dev: network device
3386 *
3387 * Check if carrier is present on device
3388 */
4d29515f 3389static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3390{
3391 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3392}
3393
f629d208 3394unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3395
f629d208 3396void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3397
f629d208 3398void netif_carrier_on(struct net_device *dev);
1da177e4 3399
f629d208 3400void netif_carrier_off(struct net_device *dev);
1da177e4 3401
bea3348e
SH
3402/**
3403 * netif_dormant_on - mark device as dormant.
3404 * @dev: network device
3405 *
3406 * Mark device as dormant (as per RFC2863).
3407 *
3408 * The dormant state indicates that the relevant interface is not
3409 * actually in a condition to pass packets (i.e., it is not 'up') but is
3410 * in a "pending" state, waiting for some external event. For "on-
3411 * demand" interfaces, this new state identifies the situation where the
3412 * interface is waiting for events to place it in the up state.
bea3348e 3413 */
b00055aa
SR
3414static inline void netif_dormant_on(struct net_device *dev)
3415{
3416 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3417 linkwatch_fire_event(dev);
3418}
3419
bea3348e
SH
3420/**
3421 * netif_dormant_off - set device as not dormant.
3422 * @dev: network device
3423 *
3424 * Device is not in dormant state.
3425 */
b00055aa
SR
3426static inline void netif_dormant_off(struct net_device *dev)
3427{
3428 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3429 linkwatch_fire_event(dev);
3430}
3431
bea3348e 3432/**
8ecbc40a 3433 * netif_dormant - test if device is dormant
bea3348e
SH
3434 * @dev: network device
3435 *
8ecbc40a 3436 * Check if device is dormant.
bea3348e 3437 */
4d29515f 3438static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3439{
3440 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3441}
3442
3443
bea3348e
SH
3444/**
3445 * netif_oper_up - test if device is operational
3446 * @dev: network device
3447 *
3448 * Check if carrier is operational
3449 */
4d29515f 3450static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3451{
b00055aa
SR
3452 return (dev->operstate == IF_OPER_UP ||
3453 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3454}
3455
bea3348e
SH
3456/**
3457 * netif_device_present - is device available or removed
3458 * @dev: network device
3459 *
3460 * Check if device has not been removed from system.
3461 */
4d29515f 3462static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3463{
3464 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3465}
3466
f629d208 3467void netif_device_detach(struct net_device *dev);
1da177e4 3468
f629d208 3469void netif_device_attach(struct net_device *dev);
1da177e4
LT
3470
3471/*
3472 * Network interface message level settings
3473 */
1da177e4
LT
3474
3475enum {
3476 NETIF_MSG_DRV = 0x0001,
3477 NETIF_MSG_PROBE = 0x0002,
3478 NETIF_MSG_LINK = 0x0004,
3479 NETIF_MSG_TIMER = 0x0008,
3480 NETIF_MSG_IFDOWN = 0x0010,
3481 NETIF_MSG_IFUP = 0x0020,
3482 NETIF_MSG_RX_ERR = 0x0040,
3483 NETIF_MSG_TX_ERR = 0x0080,
3484 NETIF_MSG_TX_QUEUED = 0x0100,
3485 NETIF_MSG_INTR = 0x0200,
3486 NETIF_MSG_TX_DONE = 0x0400,
3487 NETIF_MSG_RX_STATUS = 0x0800,
3488 NETIF_MSG_PKTDATA = 0x1000,
3489 NETIF_MSG_HW = 0x2000,
3490 NETIF_MSG_WOL = 0x4000,
3491};
3492
3493#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3494#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3495#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3496#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3497#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3498#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3499#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3500#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3501#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3502#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3503#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3504#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3505#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3506#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3507#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3508
3509static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3510{
3511 /* use default */
3512 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3513 return default_msg_enable_bits;
3514 if (debug_value == 0) /* no output */
3515 return 0;
3516 /* set low N bits */
3517 return (1 << debug_value) - 1;
3518}
3519
c773e847 3520static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3521{
c773e847
DM
3522 spin_lock(&txq->_xmit_lock);
3523 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3524}
3525
5a717f4f
MT
3526static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3527{
3528 __acquire(&txq->_xmit_lock);
3529 return true;
3530}
3531
3532static inline void __netif_tx_release(struct netdev_queue *txq)
3533{
3534 __release(&txq->_xmit_lock);
3535}
3536
fd2ea0a7
DM
3537static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3538{
3539 spin_lock_bh(&txq->_xmit_lock);
3540 txq->xmit_lock_owner = smp_processor_id();
3541}
3542
4d29515f 3543static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3544{
4d29515f 3545 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3546 if (likely(ok))
3547 txq->xmit_lock_owner = smp_processor_id();
3548 return ok;
3549}
3550
3551static inline void __netif_tx_unlock(struct netdev_queue *txq)
3552{
3553 txq->xmit_lock_owner = -1;
3554 spin_unlock(&txq->_xmit_lock);
3555}
3556
3557static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3558{
3559 txq->xmit_lock_owner = -1;
3560 spin_unlock_bh(&txq->_xmit_lock);
3561}
3562
08baf561
ED
3563static inline void txq_trans_update(struct netdev_queue *txq)
3564{
3565 if (txq->xmit_lock_owner != -1)
3566 txq->trans_start = jiffies;
3567}
3568
ba162f8e
FW
3569/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3570static inline void netif_trans_update(struct net_device *dev)
3571{
9b36627a
FW
3572 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3573
3574 if (txq->trans_start != jiffies)
3575 txq->trans_start = jiffies;
ba162f8e
FW
3576}
3577
d29f749e
DJ
3578/**
3579 * netif_tx_lock - grab network device transmit lock
3580 * @dev: network device
d29f749e
DJ
3581 *
3582 * Get network device transmit lock
3583 */
22dd7495
JHS
3584static inline void netif_tx_lock(struct net_device *dev)
3585{
e8a0464c 3586 unsigned int i;
c3f26a26 3587 int cpu;
c773e847 3588
c3f26a26
DM
3589 spin_lock(&dev->tx_global_lock);
3590 cpu = smp_processor_id();
e8a0464c
DM
3591 for (i = 0; i < dev->num_tx_queues; i++) {
3592 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3593
3594 /* We are the only thread of execution doing a
3595 * freeze, but we have to grab the _xmit_lock in
3596 * order to synchronize with threads which are in
3597 * the ->hard_start_xmit() handler and already
3598 * checked the frozen bit.
3599 */
e8a0464c 3600 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3601 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3602 __netif_tx_unlock(txq);
e8a0464c 3603 }
932ff279
HX
3604}
3605
3606static inline void netif_tx_lock_bh(struct net_device *dev)
3607{
e8a0464c
DM
3608 local_bh_disable();
3609 netif_tx_lock(dev);
932ff279
HX
3610}
3611
932ff279
HX
3612static inline void netif_tx_unlock(struct net_device *dev)
3613{
e8a0464c
DM
3614 unsigned int i;
3615
3616 for (i = 0; i < dev->num_tx_queues; i++) {
3617 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3618
c3f26a26
DM
3619 /* No need to grab the _xmit_lock here. If the
3620 * queue is not stopped for another reason, we
3621 * force a schedule.
3622 */
3623 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3624 netif_schedule_queue(txq);
c3f26a26
DM
3625 }
3626 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3627}
3628
3629static inline void netif_tx_unlock_bh(struct net_device *dev)
3630{
e8a0464c
DM
3631 netif_tx_unlock(dev);
3632 local_bh_enable();
932ff279
HX
3633}
3634
c773e847 3635#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3636 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3637 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3638 } else { \
3639 __netif_tx_acquire(txq); \
22dd7495
JHS
3640 } \
3641}
3642
5efeac44
EB
3643#define HARD_TX_TRYLOCK(dev, txq) \
3644 (((dev->features & NETIF_F_LLTX) == 0) ? \
3645 __netif_tx_trylock(txq) : \
5a717f4f 3646 __netif_tx_acquire(txq))
5efeac44 3647
c773e847 3648#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3649 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3650 __netif_tx_unlock(txq); \
5a717f4f
MT
3651 } else { \
3652 __netif_tx_release(txq); \
22dd7495
JHS
3653 } \
3654}
3655
1da177e4
LT
3656static inline void netif_tx_disable(struct net_device *dev)
3657{
fd2ea0a7 3658 unsigned int i;
c3f26a26 3659 int cpu;
fd2ea0a7 3660
c3f26a26
DM
3661 local_bh_disable();
3662 cpu = smp_processor_id();
fd2ea0a7
DM
3663 for (i = 0; i < dev->num_tx_queues; i++) {
3664 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3665
3666 __netif_tx_lock(txq, cpu);
fd2ea0a7 3667 netif_tx_stop_queue(txq);
c3f26a26 3668 __netif_tx_unlock(txq);
fd2ea0a7 3669 }
c3f26a26 3670 local_bh_enable();
1da177e4
LT
3671}
3672
e308a5d8
DM
3673static inline void netif_addr_lock(struct net_device *dev)
3674{
3675 spin_lock(&dev->addr_list_lock);
3676}
3677
2429f7ac
JP
3678static inline void netif_addr_lock_nested(struct net_device *dev)
3679{
25175ba5
VY
3680 int subclass = SINGLE_DEPTH_NESTING;
3681
3682 if (dev->netdev_ops->ndo_get_lock_subclass)
3683 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3684
3685 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3686}
3687
e308a5d8
DM
3688static inline void netif_addr_lock_bh(struct net_device *dev)
3689{
3690 spin_lock_bh(&dev->addr_list_lock);
3691}
3692
3693static inline void netif_addr_unlock(struct net_device *dev)
3694{
3695 spin_unlock(&dev->addr_list_lock);
3696}
3697
3698static inline void netif_addr_unlock_bh(struct net_device *dev)
3699{
3700 spin_unlock_bh(&dev->addr_list_lock);
3701}
3702
f001fde5 3703/*
31278e71 3704 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3705 * rcu_read_lock held.
3706 */
3707#define for_each_dev_addr(dev, ha) \
31278e71 3708 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3709
1da177e4
LT
3710/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3711
f629d208 3712void ether_setup(struct net_device *dev);
1da177e4
LT
3713
3714/* Support for loadable net-drivers */
f629d208 3715struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3716 unsigned char name_assign_type,
f629d208
JP
3717 void (*setup)(struct net_device *),
3718 unsigned int txqs, unsigned int rxqs);
0ad646c8
CW
3719int dev_get_valid_name(struct net *net, struct net_device *dev,
3720 const char *name);
3721
c835a677
TG
3722#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3723 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3724
c835a677
TG
3725#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3726 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3727 count)
36909ea4 3728
f629d208
JP
3729int register_netdev(struct net_device *dev);
3730void unregister_netdev(struct net_device *dev);
f001fde5 3731
22bedad3 3732/* General hardware address lists handling functions */
f629d208
JP
3733int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3734 struct netdev_hw_addr_list *from_list, int addr_len);
3735void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3736 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3737int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3738 struct net_device *dev,
3739 int (*sync)(struct net_device *, const unsigned char *),
3740 int (*unsync)(struct net_device *,
3741 const unsigned char *));
3742void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3743 struct net_device *dev,
3744 int (*unsync)(struct net_device *,
3745 const unsigned char *));
f629d208 3746void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3747
f001fde5 3748/* Functions used for device addresses handling */
f629d208
JP
3749int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3750 unsigned char addr_type);
3751int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3752 unsigned char addr_type);
f629d208
JP
3753void dev_addr_flush(struct net_device *dev);
3754int dev_addr_init(struct net_device *dev);
a748ee24
JP
3755
3756/* Functions used for unicast addresses handling */
f629d208
JP
3757int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3758int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3759int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3760int dev_uc_sync(struct net_device *to, struct net_device *from);
3761int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3762void dev_uc_unsync(struct net_device *to, struct net_device *from);
3763void dev_uc_flush(struct net_device *dev);
3764void dev_uc_init(struct net_device *dev);
f001fde5 3765
670e5b8e
AD
3766/**
3767 * __dev_uc_sync - Synchonize device's unicast list
3768 * @dev: device to sync
3769 * @sync: function to call if address should be added
3770 * @unsync: function to call if address should be removed
3771 *
3772 * Add newly added addresses to the interface, and release
3773 * addresses that have been deleted.
5e82b4b2 3774 */
670e5b8e
AD
3775static inline int __dev_uc_sync(struct net_device *dev,
3776 int (*sync)(struct net_device *,
3777 const unsigned char *),
3778 int (*unsync)(struct net_device *,
3779 const unsigned char *))
3780{
3781 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3782}
3783
3784/**
e793c0f7 3785 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3786 * @dev: device to sync
3787 * @unsync: function to call if address should be removed
3788 *
3789 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3790 */
670e5b8e
AD
3791static inline void __dev_uc_unsync(struct net_device *dev,
3792 int (*unsync)(struct net_device *,
3793 const unsigned char *))
3794{
3795 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3796}
3797
22bedad3 3798/* Functions used for multicast addresses handling */
f629d208
JP
3799int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3800int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3801int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3802int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3803int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3804int dev_mc_sync(struct net_device *to, struct net_device *from);
3805int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3806void dev_mc_unsync(struct net_device *to, struct net_device *from);
3807void dev_mc_flush(struct net_device *dev);
3808void dev_mc_init(struct net_device *dev);
f001fde5 3809
670e5b8e
AD
3810/**
3811 * __dev_mc_sync - Synchonize device's multicast list
3812 * @dev: device to sync
3813 * @sync: function to call if address should be added
3814 * @unsync: function to call if address should be removed
3815 *
3816 * Add newly added addresses to the interface, and release
3817 * addresses that have been deleted.
5e82b4b2 3818 */
670e5b8e
AD
3819static inline int __dev_mc_sync(struct net_device *dev,
3820 int (*sync)(struct net_device *,
3821 const unsigned char *),
3822 int (*unsync)(struct net_device *,
3823 const unsigned char *))
3824{
3825 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3826}
3827
3828/**
e793c0f7 3829 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3830 * @dev: device to sync
3831 * @unsync: function to call if address should be removed
3832 *
3833 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3834 */
670e5b8e
AD
3835static inline void __dev_mc_unsync(struct net_device *dev,
3836 int (*unsync)(struct net_device *,
3837 const unsigned char *))
3838{
3839 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3840}
3841
4417da66 3842/* Functions used for secondary unicast and multicast support */
f629d208
JP
3843void dev_set_rx_mode(struct net_device *dev);
3844void __dev_set_rx_mode(struct net_device *dev);
3845int dev_set_promiscuity(struct net_device *dev, int inc);
3846int dev_set_allmulti(struct net_device *dev, int inc);
3847void netdev_state_change(struct net_device *dev);
3848void netdev_notify_peers(struct net_device *dev);
3849void netdev_features_change(struct net_device *dev);
1da177e4 3850/* Load a device via the kmod */
f629d208
JP
3851void dev_load(struct net *net, const char *name);
3852struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3853 struct rtnl_link_stats64 *storage);
3854void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3855 const struct net_device_stats *netdev_stats);
eeda3fd6 3856
1da177e4 3857extern int netdev_max_backlog;
3b098e2d 3858extern int netdev_tstamp_prequeue;
1da177e4 3859extern int weight_p;
3d48b53f
MT
3860extern int dev_weight_rx_bias;
3861extern int dev_weight_tx_bias;
3862extern int dev_rx_weight;
3863extern int dev_tx_weight;
9ff162a8 3864
f629d208 3865bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3866struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3867 struct list_head **iter);
f629d208
JP
3868struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3869 struct list_head **iter);
8b5be856 3870
44a40855
VY
3871/* iterate through upper list, must be called under RCU read lock */
3872#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3873 for (iter = &(dev)->adj_list.upper, \
3874 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3875 updev; \
3876 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3877
1a3f060c
DA
3878int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3879 int (*fn)(struct net_device *upper_dev,
3880 void *data),
3881 void *data);
3882
3883bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3884 struct net_device *upper_dev);
3885
25cc72a3
IS
3886bool netdev_has_any_upper_dev(struct net_device *dev);
3887
f629d208
JP
3888void *netdev_lower_get_next_private(struct net_device *dev,
3889 struct list_head **iter);
3890void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3891 struct list_head **iter);
31088a11
VF
3892
3893#define netdev_for_each_lower_private(dev, priv, iter) \
3894 for (iter = (dev)->adj_list.lower.next, \
3895 priv = netdev_lower_get_next_private(dev, &(iter)); \
3896 priv; \
3897 priv = netdev_lower_get_next_private(dev, &(iter)))
3898
3899#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3900 for (iter = &(dev)->adj_list.lower, \
3901 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3902 priv; \
3903 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3904
4085ebe8
VY
3905void *netdev_lower_get_next(struct net_device *dev,
3906 struct list_head **iter);
7ce856aa 3907
4085ebe8 3908#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3909 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3910 ldev = netdev_lower_get_next(dev, &(iter)); \
3911 ldev; \
3912 ldev = netdev_lower_get_next(dev, &(iter)))
3913
7ce856aa
JP
3914struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3915 struct list_head **iter);
3916struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3917 struct list_head **iter);
3918
1a3f060c
DA
3919int netdev_walk_all_lower_dev(struct net_device *dev,
3920 int (*fn)(struct net_device *lower_dev,
3921 void *data),
3922 void *data);
3923int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3924 int (*fn)(struct net_device *lower_dev,
3925 void *data),
3926 void *data);
3927
f629d208 3928void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3929void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3930struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3931struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
42ab19ee
DA
3932int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
3933 struct netlink_ext_ack *extack);
f629d208 3934int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3935 struct net_device *upper_dev,
42ab19ee
DA
3936 void *upper_priv, void *upper_info,
3937 struct netlink_ext_ack *extack);
f629d208
JP
3938void netdev_upper_dev_unlink(struct net_device *dev,
3939 struct net_device *upper_dev);
5bb025fa 3940void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3941void *netdev_lower_dev_get_private(struct net_device *dev,
3942 struct net_device *lower_dev);
04d48266
JP
3943void netdev_lower_state_changed(struct net_device *lower_dev,
3944 void *lower_state_info);
960fb622
ED
3945
3946/* RSS keys are 40 or 52 bytes long */
3947#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3948extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3949void netdev_rss_key_fill(void *buffer, size_t len);
3950
952fcfd0 3951int dev_get_nest_level(struct net_device *dev);
f629d208 3952int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 3953int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
3954int skb_csum_hwoffload_help(struct sk_buff *skb,
3955 const netdev_features_t features);
3956
f629d208
JP
3957struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3958 netdev_features_t features, bool tx_path);
3959struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3960 netdev_features_t features);
12b0004d 3961
61bd3857
MS
3962struct netdev_bonding_info {
3963 ifslave slave;
3964 ifbond master;
3965};
3966
3967struct netdev_notifier_bonding_info {
3968 struct netdev_notifier_info info; /* must be first */
3969 struct netdev_bonding_info bonding_info;
3970};
3971
3972void netdev_bonding_info_change(struct net_device *dev,
3973 struct netdev_bonding_info *bonding_info);
3974
12b0004d
CW
3975static inline
3976struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3977{
3978 return __skb_gso_segment(skb, features, true);
3979}
53d6471c 3980__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3981
3982static inline bool can_checksum_protocol(netdev_features_t features,
3983 __be16 protocol)
3984{
c8cd0989
TH
3985 if (protocol == htons(ETH_P_FCOE))
3986 return !!(features & NETIF_F_FCOE_CRC);
3987
3988 /* Assume this is an IP checksum (not SCTP CRC) */
3989
3990 if (features & NETIF_F_HW_CSUM) {
3991 /* Can checksum everything */
3992 return true;
3993 }
3994
3995 switch (protocol) {
3996 case htons(ETH_P_IP):
3997 return !!(features & NETIF_F_IP_CSUM);
3998 case htons(ETH_P_IPV6):
3999 return !!(features & NETIF_F_IPV6_CSUM);
4000 default:
4001 return false;
4002 }
ec5f0615 4003}
12b0004d 4004
fb286bb2 4005#ifdef CONFIG_BUG
f629d208 4006void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
4007#else
4008static inline void netdev_rx_csum_fault(struct net_device *dev)
4009{
4010}
4011#endif
1da177e4 4012/* rx skb timestamps */
f629d208
JP
4013void net_enable_timestamp(void);
4014void net_disable_timestamp(void);
1da177e4 4015
20380731 4016#ifdef CONFIG_PROC_FS
f629d208 4017int __init dev_proc_init(void);
900ff8c6
CW
4018#else
4019#define dev_proc_init() 0
20380731
ACM
4020#endif
4021
4798248e 4022static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4023 struct sk_buff *skb, struct net_device *dev,
4024 bool more)
4798248e 4025{
fa2dbdc2 4026 skb->xmit_more = more ? 1 : 0;
0b725a2c 4027 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4028}
4029
10b3ad8c 4030static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4031 struct netdev_queue *txq, bool more)
4798248e
DM
4032{
4033 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4034 int rc;
4798248e 4035
fa2dbdc2 4036 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4037 if (rc == NETDEV_TX_OK)
4038 txq_trans_update(txq);
4039
4040 return rc;
4798248e
DM
4041}
4042
b793dc5c 4043int netdev_class_create_file_ns(const struct class_attribute *class_attr,
42a2d923 4044 const void *ns);
b793dc5c 4045void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
42a2d923 4046 const void *ns);
58292cbe 4047
b793dc5c 4048static inline int netdev_class_create_file(const struct class_attribute *class_attr)
58292cbe
TH
4049{
4050 return netdev_class_create_file_ns(class_attr, NULL);
4051}
4052
b793dc5c 4053static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
58292cbe
TH
4054{
4055 netdev_class_remove_file_ns(class_attr, NULL);
4056}
b8a9787e 4057
737aec57 4058extern const struct kobj_ns_type_operations net_ns_type_operations;
04600794 4059
f629d208 4060const char *netdev_drivername(const struct net_device *dev);
6579e57b 4061
f629d208 4062void linkwatch_run_queue(void);
20380731 4063
da08143b
MK
4064static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4065 netdev_features_t f2)
4066{
c8cd0989
TH
4067 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4068 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4069 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4070 else
b6a0e72a 4071 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4072 }
da08143b 4073
c8cd0989 4074 return f1 & f2;
da08143b
MK
4075}
4076
c8f44aff
MM
4077static inline netdev_features_t netdev_get_wanted_features(
4078 struct net_device *dev)
5455c699
MM
4079{
4080 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4081}
c8f44aff
MM
4082netdev_features_t netdev_increment_features(netdev_features_t all,
4083 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4084
4085/* Allow TSO being used on stacked device :
4086 * Performing the GSO segmentation before last device
4087 * is a performance improvement.
4088 */
4089static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4090 netdev_features_t mask)
4091{
4092 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4093}
4094
6cb6a27c 4095int __netdev_update_features(struct net_device *dev);
5455c699 4096void netdev_update_features(struct net_device *dev);
afe12cc8 4097void netdev_change_features(struct net_device *dev);
7f353bf2 4098
fc4a7489
PM
4099void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4100 struct net_device *dev);
4101
e38f3025
TM
4102netdev_features_t passthru_features_check(struct sk_buff *skb,
4103 struct net_device *dev,
4104 netdev_features_t features);
c1e756bf 4105netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4106
4d29515f 4107static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4108{
7b748340 4109 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4110
4111 /* check flags correspondence */
4112 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4113 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4114 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4115 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4116 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4117 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4118 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4119 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4120 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4121 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4122 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4123 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4124 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4125 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4126 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4127 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0345e186 4128
d6b4991a 4129 return (features & feature) == feature;
576a30eb
HX
4130}
4131
4d29515f 4132static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4133{
278b2513 4134 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4135 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4136}
4137
8b86a61d 4138static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4139 netdev_features_t features)
7967168c 4140{
fc741216 4141 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4142 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4143 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4144}
4145
82cc1a7a
PWJ
4146static inline void netif_set_gso_max_size(struct net_device *dev,
4147 unsigned int size)
4148{
4149 dev->gso_max_size = size;
4150}
4151
7a7ffbab
WCC
4152static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4153 int pulled_hlen, u16 mac_offset,
4154 int mac_len)
4155{
4156 skb->protocol = protocol;
4157 skb->encapsulation = 1;
4158 skb_push(skb, pulled_hlen);
4159 skb_reset_transport_header(skb);
4160 skb->mac_header = mac_offset;
4161 skb->network_header = skb->mac_header + mac_len;
4162 skb->mac_len = mac_len;
4163}
4164
3c175784
SD
4165static inline bool netif_is_macsec(const struct net_device *dev)
4166{
4167 return dev->priv_flags & IFF_MACSEC;
4168}
4169
b618aaa9 4170static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4171{
4172 return dev->priv_flags & IFF_MACVLAN;
4173}
4174
b618aaa9 4175static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4176{
4177 return dev->priv_flags & IFF_MACVLAN_PORT;
4178}
4179
b618aaa9 4180static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4181{
4182 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4183}
4184
b618aaa9 4185static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4186{
4187 return dev->priv_flags & IFF_IPVLAN_MASTER;
4188}
4189
b618aaa9 4190static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4191{
4192 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4193}
4194
b618aaa9 4195static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4196{
4197 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4198}
4199
3bdc0eba
BG
4200static inline bool netif_supports_nofcs(struct net_device *dev)
4201{
4202 return dev->priv_flags & IFF_SUPP_NOFCS;
4203}
4204
007979ea 4205static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4206{
007979ea 4207 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4208}
4209
fee6d4c7
DA
4210static inline bool netif_is_l3_slave(const struct net_device *dev)
4211{
4212 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4213}
4214
0894ae3f
JP
4215static inline bool netif_is_bridge_master(const struct net_device *dev)
4216{
4217 return dev->priv_flags & IFF_EBRIDGE;
4218}
4219
28f9ee22
VY
4220static inline bool netif_is_bridge_port(const struct net_device *dev)
4221{
4222 return dev->priv_flags & IFF_BRIDGE_PORT;
4223}
4224
35d4e172
JP
4225static inline bool netif_is_ovs_master(const struct net_device *dev)
4226{
4227 return dev->priv_flags & IFF_OPENVSWITCH;
4228}
4229
5be66141
JP
4230static inline bool netif_is_ovs_port(const struct net_device *dev)
4231{
4232 return dev->priv_flags & IFF_OVS_DATAPATH;
4233}
4234
b618aaa9 4235static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4236{
4237 return dev->priv_flags & IFF_TEAM;
4238}
4239
b618aaa9 4240static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4241{
4242 return dev->priv_flags & IFF_TEAM_PORT;
4243}
4244
b618aaa9 4245static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4246{
4247 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4248}
4249
b618aaa9 4250static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4251{
4252 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4253}
4254
d4ab4286
KJ
4255static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4256{
4257 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4258}
4259
02875878
ED
4260/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4261static inline void netif_keep_dst(struct net_device *dev)
4262{
4263 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4264}
4265
18d3df3e
PA
4266/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4267static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4268{
4269 /* TODO: reserve and use an additional IFF bit, if we get more users */
4270 return dev->priv_flags & IFF_MACSEC;
4271}
4272
505d4f73 4273extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4274
571ba423
JP
4275/* Logging, debugging and troubleshooting/diagnostic helpers. */
4276
4277/* netdev_printk helpers, similar to dev_printk */
4278
4279static inline const char *netdev_name(const struct net_device *dev)
4280{
c6f854d5
VF
4281 if (!dev->name[0] || strchr(dev->name, '%'))
4282 return "(unnamed net_device)";
571ba423
JP
4283 return dev->name;
4284}
4285
8397ed36
DA
4286static inline bool netdev_unregistering(const struct net_device *dev)
4287{
4288 return dev->reg_state == NETREG_UNREGISTERING;
4289}
4290
ccc7f496
VF
4291static inline const char *netdev_reg_state(const struct net_device *dev)
4292{
4293 switch (dev->reg_state) {
4294 case NETREG_UNINITIALIZED: return " (uninitialized)";
4295 case NETREG_REGISTERED: return "";
4296 case NETREG_UNREGISTERING: return " (unregistering)";
4297 case NETREG_UNREGISTERED: return " (unregistered)";
4298 case NETREG_RELEASED: return " (released)";
4299 case NETREG_DUMMY: return " (dummy)";
4300 }
4301
4302 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4303 return " (unknown)";
4304}
4305
f629d208 4306__printf(3, 4)
6ea754eb
JP
4307void netdev_printk(const char *level, const struct net_device *dev,
4308 const char *format, ...);
f629d208 4309__printf(2, 3)
6ea754eb 4310void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4311__printf(2, 3)
6ea754eb 4312void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4313__printf(2, 3)
6ea754eb 4314void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4315__printf(2, 3)
6ea754eb 4316void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4317__printf(2, 3)
6ea754eb 4318void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4319__printf(2, 3)
6ea754eb 4320void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4321__printf(2, 3)
6ea754eb 4322void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4323
8909c9ad
VK
4324#define MODULE_ALIAS_NETDEV(device) \
4325 MODULE_ALIAS("netdev-" device)
4326
b558c96f 4327#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4328#define netdev_dbg(__dev, format, args...) \
4329do { \
ffa10cb4 4330 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4331} while (0)
b558c96f
JC
4332#elif defined(DEBUG)
4333#define netdev_dbg(__dev, format, args...) \
4334 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4335#else
4336#define netdev_dbg(__dev, format, args...) \
4337({ \
4338 if (0) \
4339 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4340})
4341#endif
4342
4343#if defined(VERBOSE_DEBUG)
4344#define netdev_vdbg netdev_dbg
4345#else
4346
4347#define netdev_vdbg(dev, format, args...) \
4348({ \
4349 if (0) \
4350 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4351 0; \
4352})
4353#endif
4354
4355/*
4356 * netdev_WARN() acts like dev_printk(), but with the key difference
4357 * of using a WARN/WARN_ON to get the message out, including the
4358 * file/line information and a backtrace.
4359 */
4360#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4361 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4362 netdev_reg_state(dev), ##args)
571ba423 4363
b3d95c5c
JP
4364/* netif printk helpers, similar to netdev_printk */
4365
4366#define netif_printk(priv, type, level, dev, fmt, args...) \
4367do { \
4368 if (netif_msg_##type(priv)) \
4369 netdev_printk(level, (dev), fmt, ##args); \
4370} while (0)
4371
f45f4321
JP
4372#define netif_level(level, priv, type, dev, fmt, args...) \
4373do { \
4374 if (netif_msg_##type(priv)) \
4375 netdev_##level(dev, fmt, ##args); \
4376} while (0)
4377
b3d95c5c 4378#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4379 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4380#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4381 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4382#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4383 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4384#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4385 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4386#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4387 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4388#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4389 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4390#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4391 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4392
0053ea9c 4393#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4394#define netif_dbg(priv, type, netdev, format, args...) \
4395do { \
4396 if (netif_msg_##type(priv)) \
b5fb0a03 4397 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4398} while (0)
0053ea9c
JP
4399#elif defined(DEBUG)
4400#define netif_dbg(priv, type, dev, format, args...) \
4401 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4402#else
4403#define netif_dbg(priv, type, dev, format, args...) \
4404({ \
4405 if (0) \
4406 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4407 0; \
4408})
4409#endif
4410
f617f276
EC
4411/* if @cond then downgrade to debug, else print at @level */
4412#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4413 do { \
4414 if (cond) \
4415 netif_dbg(priv, type, netdev, fmt, ##args); \
4416 else \
4417 netif_ ## level(priv, type, netdev, fmt, ##args); \
4418 } while (0)
4419
b3d95c5c 4420#if defined(VERBOSE_DEBUG)
bcfcc450 4421#define netif_vdbg netif_dbg
b3d95c5c
JP
4422#else
4423#define netif_vdbg(priv, type, dev, format, args...) \
4424({ \
4425 if (0) \
a4ed89cb 4426 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4427 0; \
4428})
4429#endif
571ba423 4430
900ff8c6
CW
4431/*
4432 * The list of packet types we will receive (as opposed to discard)
4433 * and the routines to invoke.
4434 *
4435 * Why 16. Because with 16 the only overlap we get on a hash of the
4436 * low nibble of the protocol value is RARP/SNAP/X.25.
4437 *
4438 * NOTE: That is no longer true with the addition of VLAN tags. Not
4439 * sure which should go first, but I bet it won't make much
4440 * difference if we are running VLANs. The good news is that
4441 * this protocol won't be in the list unless compiled in, so
4442 * the average user (w/out VLANs) will not be adversely affected.
4443 * --BLG
4444 *
4445 * 0800 IP
4446 * 8100 802.1Q VLAN
4447 * 0001 802.3
4448 * 0002 AX.25
4449 * 0004 802.2
4450 * 8035 RARP
4451 * 0005 SNAP
4452 * 0805 X.25
4453 * 0806 ARP
4454 * 8137 IPX
4455 * 0009 Localtalk
4456 * 86DD IPv6
4457 */
4458#define PTYPE_HASH_SIZE (16)
4459#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4460
385a154c 4461#endif /* _LINUX_NETDEVICE_H */