]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
packet: respect devices with LLTX flag in direct xmit
[mirror_ubuntu-artful-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
e8db0be1 28#include <linux/pm_qos.h>
d7fe0f24 29#include <linux/timer.h>
187f1882 30#include <linux/bug.h>
bea3348e 31#include <linux/delay.h>
60063497 32#include <linux/atomic.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>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
cf85d08f 44#include <net/dsa.h>
7a6b6f51 45#ifdef CONFIG_DCB
2f90b865
AD
46#include <net/dcbnl.h>
47#endif
5bc1421e 48#include <net/netprio_cgroup.h>
a050c33f 49
a59e2ecb 50#include <linux/netdev_features.h>
77162022 51#include <linux/neighbour.h>
607ca46e 52#include <uapi/linux/netdevice.h>
a59e2ecb 53
115c1d6e 54struct netpoll_info;
313162d0 55struct device;
c1f19b51 56struct phy_device;
704232c2
JB
57/* 802.11 specific */
58struct wireless_dev;
1da177e4
LT
59 /* source back-compat hooks */
60#define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
62
f629d208
JP
63void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct ethtool_ops *ops);
d07d7507 65
c1f79426
SA
66/* hardware address assignment types */
67#define NET_ADDR_PERM 0 /* address is permanent (default) */
68#define NET_ADDR_RANDOM 1 /* address is generated randomly */
69#define NET_ADDR_STOLEN 2 /* address is stolen from other device */
fbdeca2d
JP
70#define NET_ADDR_SET 3 /* address is set using
71 * dev_set_mac_address() */
c1f79426 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
89 * higher layers. Virtual network devices transmit synchronously, in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), all
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 */
98#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
99#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 100
b9df3cb8
GR
101/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 104#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
105#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
dc1f8bf6 107/* Driver transmit return codes */
9a1654ba 108#define NETDEV_TX_MASK 0xf0
572a9d7b 109
dc1f8bf6 110enum netdev_tx {
572a9d7b 111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
dc1f8bf6
SH
115};
116typedef enum netdev_tx netdev_tx_t;
117
9a1654ba
JP
118/*
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121 */
122static inline bool dev_xmit_complete(int rc)
123{
124 /*
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 */
130 if (likely(rc < NET_XMIT_MASK))
131 return true;
132
133 return false;
134}
135
1da177e4
LT
136/*
137 * Compute the worst case header length according to the protocols
138 * used.
139 */
fe2918b0 140
d11ead75 141#if 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;
adc9300e
ED
196#endif
197
1da177e4
LT
198struct neighbour;
199struct neigh_parms;
200struct sk_buff;
201
f001fde5
JP
202struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
205 unsigned char type;
ccffad25
JP
206#define NETDEV_HW_ADDR_T_LAN 1
207#define NETDEV_HW_ADDR_T_SAN 2
208#define NETDEV_HW_ADDR_T_SLAVE 3
209#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 210#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 211 bool global_use;
4cd729b0 212 int sync_cnt;
8f8f103d 213 int refcount;
4543fbef 214 int synced;
f001fde5
JP
215 struct rcu_head rcu_head;
216};
217
31278e71
JP
218struct netdev_hw_addr_list {
219 struct list_head list;
220 int count;
221};
222
22bedad3
JP
223#define netdev_hw_addr_list_count(l) ((l)->count)
224#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225#define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 227
22bedad3
JP
228#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230#define netdev_for_each_uc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 232
22bedad3
JP
233#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 235#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 236 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 237
d94d9fee 238struct hh_cache {
f6b72b62 239 u16 hh_len;
5c25f686 240 u16 __pad;
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
252/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
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
SH
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*rebuild)(struct sk_buff *skb);
e69dd336 271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
275};
276
1da177e4
LT
277/* These flag bits are private to the generic network queueing
278 * layer, they may not be explicitly referenced by any other
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/*
292 * This structure holds at 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
310 * to the per-cpu poll_list, and whoever clears that bit
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
320 spinlock_t poll_lock;
321 int poll_owner;
bea3348e 322#endif
5d38a079 323 struct net_device *dev;
d565b0a1 324 struct sk_buff *gro_list;
5d38a079 325 struct sk_buff *skb;
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 */
a0a46196 333 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
af12fa6e 335 NAPI_STATE_HASHED, /* In NAPI hash */
bea3348e
SH
336};
337
5b252f0c 338enum gro_result {
d1c76af9
HX
339 GRO_MERGED,
340 GRO_MERGED_FREE,
341 GRO_HELD,
342 GRO_NORMAL,
343 GRO_DROP,
344};
5b252f0c 345typedef enum gro_result gro_result_t;
d1c76af9 346
8a4eb573
JP
347/*
348 * enum rx_handler_result - Possible return values for rx_handlers.
349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
350 * further.
351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
352 * case skb->dev was changed by rx_handler.
353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
354 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
355 *
356 * rx_handlers are functions called from inside __netif_receive_skb(), to do
357 * special processing of the skb, prior to delivery to protocol handlers.
358 *
359 * Currently, a net_device can only have a single rx_handler registered. Trying
360 * to register a second rx_handler will return -EBUSY.
361 *
362 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
363 * To unregister a rx_handler on a net_device, use
364 * netdev_rx_handler_unregister().
365 *
366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
367 * do with the skb.
368 *
369 * If the rx_handler consumed to skb in some way, it should return
370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
371 * the skb to be delivered in some other ways.
372 *
373 * If the rx_handler changed skb->dev, to divert the skb to another
374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
375 * new device will be called if it exists.
376 *
377 * If the rx_handler consider the skb should be ignored, it should return
378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 379 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
380 *
381 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
382 * delivered, it should return RX_HANDLER_PASS.
383 *
384 * A device without a registered rx_handler will behave as if rx_handler
385 * returned RX_HANDLER_PASS.
386 */
387
388enum rx_handler_result {
389 RX_HANDLER_CONSUMED,
390 RX_HANDLER_ANOTHER,
391 RX_HANDLER_EXACT,
392 RX_HANDLER_PASS,
393};
394typedef enum rx_handler_result rx_handler_result_t;
395typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 396
f629d208 397void __napi_schedule(struct napi_struct *n);
bea3348e 398
4d29515f 399static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
400{
401 return test_bit(NAPI_STATE_DISABLE, &n->state);
402}
403
bea3348e
SH
404/**
405 * napi_schedule_prep - check if napi can be scheduled
406 * @n: napi context
407 *
408 * Test if NAPI routine is already running, and if not mark
409 * it as running. This is used as a condition variable
a0a46196
DM
410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
bea3348e 412 */
4d29515f 413static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 414{
a0a46196
DM
415 return !napi_disable_pending(n) &&
416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
417}
418
419/**
420 * napi_schedule - schedule NAPI poll
421 * @n: napi context
422 *
423 * Schedule NAPI poll routine to be called if it is not already
424 * running.
425 */
426static inline void napi_schedule(struct napi_struct *n)
427{
428 if (napi_schedule_prep(n))
429 __napi_schedule(n);
430}
431
bfe13f54 432/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 433static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
434{
435 if (napi_schedule_prep(napi)) {
436 __napi_schedule(napi);
4d29515f 437 return true;
bfe13f54 438 }
4d29515f 439 return false;
bfe13f54
RD
440}
441
bea3348e
SH
442/**
443 * napi_complete - NAPI processing complete
444 * @n: napi context
445 *
446 * Mark NAPI processing as complete.
447 */
f629d208
JP
448void __napi_complete(struct napi_struct *n);
449void napi_complete(struct napi_struct *n);
bea3348e 450
af12fa6e
ET
451/**
452 * napi_by_id - lookup a NAPI by napi_id
453 * @napi_id: hashed napi_id
454 *
455 * lookup @napi_id in napi_hash table
456 * must be called under rcu_read_lock()
457 */
f629d208 458struct napi_struct *napi_by_id(unsigned int napi_id);
af12fa6e
ET
459
460/**
461 * napi_hash_add - add a NAPI to global hashtable
462 * @napi: napi context
463 *
464 * generate a new napi_id and store a @napi under it in napi_hash
465 */
f629d208 466void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
467
468/**
469 * napi_hash_del - remove a NAPI from global table
470 * @napi: napi context
471 *
472 * Warning: caller must observe rcu grace period
473 * before freeing memory containing @napi
474 */
f629d208 475void napi_hash_del(struct napi_struct *napi);
af12fa6e 476
bea3348e
SH
477/**
478 * napi_disable - prevent NAPI from scheduling
479 * @n: napi context
480 *
481 * Stop NAPI from being scheduled on this context.
482 * Waits till any outstanding processing completes.
483 */
484static inline void napi_disable(struct napi_struct *n)
485{
80c33ddd 486 might_sleep();
a0a46196 487 set_bit(NAPI_STATE_DISABLE, &n->state);
bea3348e 488 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
43cc7380 489 msleep(1);
a0a46196 490 clear_bit(NAPI_STATE_DISABLE, &n->state);
bea3348e
SH
491}
492
493/**
494 * napi_enable - enable NAPI scheduling
495 * @n: napi context
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));
503 smp_mb__before_clear_bit();
504 clear_bit(NAPI_STATE_SCHED, &n->state);
505}
506
c264c3de
SH
507#ifdef CONFIG_SMP
508/**
509 * napi_synchronize - wait until NAPI is not running
510 * @n: napi context
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{
518 while (test_bit(NAPI_STATE_SCHED, &n->state))
519 msleep(1);
520}
521#else
522# define napi_synchronize(n) barrier()
523#endif
524
d94d9fee 525enum netdev_queue_state_t {
73466498
TH
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
c3f26a26 528 __QUEUE_STATE_FROZEN,
73466498
TH
529#define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
530 (1 << __QUEUE_STATE_STACK_XOFF))
531#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
532 (1 << __QUEUE_STATE_FROZEN))
79d16385 533};
73466498
TH
534/*
535 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
536 * netif_tx_* functions below are used to manipulate this flag. The
537 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
538 * queue independently. The netif_xmit_*stopped functions below are called
539 * to check if the queue has been stopped by the driver or stack (either
540 * of the XOFF bits are set in the state). Drivers should not need to call
541 * netif_xmit*stopped functions, they should only be using netif_tx_*.
542 */
79d16385 543
bb949fbd 544struct netdev_queue {
6a321cb3
ED
545/*
546 * read mostly part
547 */
bb949fbd 548 struct net_device *dev;
b0e1e646
DM
549 struct Qdisc *qdisc;
550 struct Qdisc *qdisc_sleeping;
ccf5ff69 551#ifdef CONFIG_SYSFS
1d24eb48
TH
552 struct kobject kobj;
553#endif
f2cd2d3e
ED
554#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
555 int numa_node;
556#endif
6a321cb3
ED
557/*
558 * write mostly part
559 */
560 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
561 int xmit_lock_owner;
9d21493b
ED
562 /*
563 * please use this field instead of dev->trans_start
564 */
565 unsigned long trans_start;
ccf5ff69 566
567 /*
568 * Number of TX timeouts for this queue
569 * (/sys/class/net/DEV/Q/trans_timeout)
570 */
571 unsigned long trans_timeout;
114cf580
TH
572
573 unsigned long state;
574
575#ifdef CONFIG_BQL
576 struct dql dql;
577#endif
e8a0464c 578} ____cacheline_aligned_in_smp;
bb949fbd 579
f2cd2d3e
ED
580static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
581{
582#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
583 return q->numa_node;
584#else
b236da69 585 return NUMA_NO_NODE;
f2cd2d3e
ED
586#endif
587}
588
589static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
590{
591#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
592 q->numa_node = node;
593#endif
594}
595
df334545 596#ifdef CONFIG_RPS
0a9627f2
TH
597/*
598 * This structure holds an RPS map which can be of variable length. The
599 * map is an array of CPUs.
600 */
601struct rps_map {
602 unsigned int len;
603 struct rcu_head rcu;
604 u16 cpus[0];
605};
60b778ce 606#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 607
fec5e652 608/*
c445477d
BH
609 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
610 * tail pointer for that CPU's input queue at the time of last enqueue, and
611 * a hardware filter index.
fec5e652
TH
612 */
613struct rps_dev_flow {
614 u16 cpu;
c445477d 615 u16 filter;
fec5e652
TH
616 unsigned int last_qtail;
617};
c445477d 618#define RPS_NO_FILTER 0xffff
fec5e652
TH
619
620/*
621 * The rps_dev_flow_table structure contains a table of flow mappings.
622 */
623struct rps_dev_flow_table {
624 unsigned int mask;
625 struct rcu_head rcu;
fec5e652
TH
626 struct rps_dev_flow flows[0];
627};
628#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 629 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
630
631/*
632 * The rps_sock_flow_table contains mappings of flows to the last CPU
633 * on which they were processed by the application (set in recvmsg).
634 */
635struct rps_sock_flow_table {
636 unsigned int mask;
637 u16 ents[0];
638};
639#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
60b778ce 640 ((_num) * sizeof(u16)))
fec5e652
TH
641
642#define RPS_NO_CPU 0xffff
643
644static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
645 u32 hash)
646{
647 if (table && hash) {
648 unsigned int cpu, index = hash & table->mask;
649
650 /* We only give a hint, preemption can change cpu under us */
651 cpu = raw_smp_processor_id();
652
653 if (table->ents[index] != cpu)
654 table->ents[index] = cpu;
655 }
656}
657
658static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
659 u32 hash)
660{
661 if (table && hash)
662 table->ents[hash & table->mask] = RPS_NO_CPU;
663}
664
6e3f7faf 665extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
fec5e652 666
c445477d 667#ifdef CONFIG_RFS_ACCEL
f629d208
JP
668bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
669 u16 filter_id);
c445477d 670#endif
a953be53 671#endif /* CONFIG_RPS */
c445477d 672
0a9627f2
TH
673/* This structure contains an instance of an RX queue. */
674struct netdev_rx_queue {
a953be53 675#ifdef CONFIG_RPS
6e3f7faf
ED
676 struct rps_map __rcu *rps_map;
677 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 678#endif
6e3f7faf 679 struct kobject kobj;
fe822240 680 struct net_device *dev;
0a9627f2 681} ____cacheline_aligned_in_smp;
a953be53
MD
682
683/*
684 * RX queue sysfs structures and functions.
685 */
686struct rx_queue_attribute {
687 struct attribute attr;
688 ssize_t (*show)(struct netdev_rx_queue *queue,
689 struct rx_queue_attribute *attr, char *buf);
690 ssize_t (*store)(struct netdev_rx_queue *queue,
691 struct rx_queue_attribute *attr, const char *buf, size_t len);
692};
d314774c 693
bf264145
TH
694#ifdef CONFIG_XPS
695/*
696 * This structure holds an XPS map which can be of variable length. The
697 * map is an array of queues.
698 */
699struct xps_map {
700 unsigned int len;
701 unsigned int alloc_len;
702 struct rcu_head rcu;
703 u16 queues[0];
704};
60b778ce 705#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
bf264145
TH
706#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
707 / sizeof(u16))
708
709/*
710 * This structure holds all XPS maps for device. Maps are indexed by CPU.
711 */
712struct xps_dev_maps {
713 struct rcu_head rcu;
a4177869 714 struct xps_map __rcu *cpu_map[0];
bf264145
TH
715};
716#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
717 (nr_cpu_ids * sizeof(struct xps_map *)))
718#endif /* CONFIG_XPS */
719
4f57c087
JF
720#define TC_MAX_QUEUE 16
721#define TC_BITMASK 15
722/* HW offloaded queuing disciplines txq count and offset maps */
723struct netdev_tc_txq {
724 u16 count;
725 u16 offset;
726};
727
68bad94e
NP
728#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
729/*
730 * This structure is to hold information about the device
731 * configured to run FCoE protocol stack.
732 */
733struct netdev_fcoe_hbainfo {
734 char manufacturer[64];
735 char serial_number[64];
736 char hardware_version[64];
737 char driver_version[64];
738 char optionrom_version[64];
739 char firmware_version[64];
740 char model[256];
741 char model_description[256];
742};
743#endif
744
66b52b0d
JP
745#define MAX_PHYS_PORT_ID_LEN 32
746
747/* This structure holds a unique identifier to identify the
748 * physical port used by a netdevice.
749 */
750struct netdev_phys_port_id {
751 unsigned char id[MAX_PHYS_PORT_ID_LEN];
752 unsigned char id_len;
753};
754
99932d4f
DB
755typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
756 struct sk_buff *skb);
757
d314774c
SH
758/*
759 * This structure defines the management hooks for network devices.
00829823
SH
760 * The following hooks can be defined; unless noted otherwise, they are
761 * optional and can be filled with a null pointer.
d314774c
SH
762 *
763 * int (*ndo_init)(struct net_device *dev);
764 * This function is called once when network device is registered.
765 * The network device can use this to any late stage initializaton
766 * or semantic validattion. It can fail with an error code which will
767 * be propogated back to register_netdev
768 *
769 * void (*ndo_uninit)(struct net_device *dev);
770 * This function is called when device is unregistered or when registration
771 * fails. It is not called if init fails.
772 *
773 * int (*ndo_open)(struct net_device *dev);
774 * This function is called when network device transistions to the up
775 * state.
776 *
777 * int (*ndo_stop)(struct net_device *dev);
778 * This function is called when network device transistions to the down
779 * state.
780 *
dc1f8bf6
SH
781 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
782 * struct net_device *dev);
00829823 783 * Called when a packet needs to be transmitted.
dc1f8bf6
SH
784 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
785 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
786 * Required can not be NULL.
787 *
f663dd9a 788 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 789 * void *accel_priv, select_queue_fallback_t fallback);
00829823
SH
790 * Called to decide which queue to when device supports multiple
791 * transmit queues.
792 *
d314774c
SH
793 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
794 * This function is called to allow device receiver to make
795 * changes to configuration when multicast or promiscious is enabled.
796 *
797 * void (*ndo_set_rx_mode)(struct net_device *dev);
798 * This function is called device changes address list filtering.
01789349
JP
799 * If driver handles unicast address filtering, it should set
800 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
801 *
802 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
803 * This function is called when the Media Access Control address
37b607c5 804 * needs to be changed. If this interface is not defined, the
d314774c
SH
805 * mac address can not be changed.
806 *
807 * int (*ndo_validate_addr)(struct net_device *dev);
808 * Test if Media Access Control address is valid for the device.
809 *
810 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
811 * Called when a user request an ioctl which can't be handled by
812 * the generic interface code. If not defined ioctl's return
813 * not supported error code.
814 *
815 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
816 * Used to set network devices bus interface parameters. This interface
817 * is retained for legacy reason, new devices should use the bus
818 * interface (PCI) for low level management.
819 *
820 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
821 * Called when a user wants to change the Maximum Transfer Unit
822 * of a device. If not defined, any request to change MTU will
823 * will return an error.
824 *
00829823 825 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
826 * Callback uses when the transmitter has not made any progress
827 * for dev->watchdog ticks.
828 *
3cfde79c 829 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 830 * struct rtnl_link_stats64 *storage);
d308e38f 831 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 832 * Called when a user wants to get the network device usage
be1f3c2c 833 * statistics. Drivers must do one of the following:
3cfde79c
BH
834 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
835 * rtnl_link_stats64 structure passed by the caller.
82695d9b 836 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
837 * (which should normally be dev->stats) and return a pointer to
838 * it. The structure may be changed asynchronously only if each
839 * field is written atomically.
840 * 3. Update dev->stats asynchronously and atomically, and define
841 * neither operation.
d314774c 842 *
80d5c368
PM
843 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
844 * If device support VLAN filtering this function is called when a
845 * VLAN id is registered.
d314774c 846 *
8e586137 847 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
80d5c368
PM
848 * If device support VLAN filtering this function is called when a
849 * VLAN id is unregistered.
d314774c
SH
850 *
851 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
852 *
853 * SR-IOV management functions.
854 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
855 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
856 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
5f8444a3 857 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
858 * int (*ndo_get_vf_config)(struct net_device *dev,
859 * int vf, struct ifla_vf_info *ivf);
1d8faf48 860 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
861 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
862 * struct nlattr *port[]);
863 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
864 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
865 * Called to setup 'tc' number of traffic classes in the net device. This
866 * is always called from the stack with the rtnl lock held and netif tx
867 * queues stopped. This allows the netdevice to perform queue management
868 * safely.
c445477d 869 *
e9bce845
YZ
870 * Fiber Channel over Ethernet (FCoE) offload functions.
871 * int (*ndo_fcoe_enable)(struct net_device *dev);
872 * Called when the FCoE protocol stack wants to start using LLD for FCoE
873 * so the underlying device can perform whatever needed configuration or
874 * initialization to support acceleration of FCoE traffic.
875 *
876 * int (*ndo_fcoe_disable)(struct net_device *dev);
877 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
878 * so the underlying device can perform whatever needed clean-ups to
879 * stop supporting acceleration of FCoE traffic.
880 *
881 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
882 * struct scatterlist *sgl, unsigned int sgc);
883 * Called when the FCoE Initiator wants to initialize an I/O that
884 * is a possible candidate for Direct Data Placement (DDP). The LLD can
885 * perform necessary setup and returns 1 to indicate the device is set up
886 * successfully to perform DDP on this I/O, otherwise this returns 0.
887 *
888 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
889 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
890 * indicated by the FC exchange id 'xid', so the underlying device can
891 * clean up and reuse resources for later DDP requests.
892 *
893 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
894 * struct scatterlist *sgl, unsigned int sgc);
895 * Called when the FCoE Target wants to initialize an I/O that
896 * is a possible candidate for Direct Data Placement (DDP). The LLD can
897 * perform necessary setup and returns 1 to indicate the device is set up
898 * successfully to perform DDP on this I/O, otherwise this returns 0.
899 *
68bad94e
NP
900 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
901 * struct netdev_fcoe_hbainfo *hbainfo);
902 * Called when the FCoE Protocol stack wants information on the underlying
903 * device. This information is utilized by the FCoE protocol stack to
904 * register attributes with Fiber Channel management service as per the
905 * FC-GS Fabric Device Management Information(FDMI) specification.
906 *
e9bce845
YZ
907 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
908 * Called when the underlying device wants to override default World Wide
909 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
910 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
911 * protocol stack to use.
912 *
c445477d
BH
913 * RFS acceleration.
914 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
915 * u16 rxq_index, u32 flow_id);
916 * Set hardware filter for RFS. rxq_index is the target queue index;
917 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
918 * Return the filter ID on success, or a negative error code.
fbaec0ea 919 *
8b98a70c 920 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
921 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
922 * Called to make another netdev an underling.
923 *
924 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
925 * Called to release previously enslaved netdev.
5455c699
MM
926 *
927 * Feature/offload setting functions.
c8f44aff
MM
928 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
929 * netdev_features_t features);
5455c699
MM
930 * Adjusts the requested feature flags according to device-specific
931 * constraints, and returns the resulting flags. Must not modify
932 * the device state.
933 *
c8f44aff 934 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
935 * Called to update device configuration to new features. Passed
936 * feature set might be less than what was returned by ndo_fix_features()).
937 * Must return >0 or -errno if it changed dev->features itself.
938 *
edc7d573 939 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
940 * struct net_device *dev,
6b6e2725 941 * const unsigned char *addr, u16 flags)
77162022 942 * Adds an FDB entry to dev for addr.
1690be63
VY
943 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
944 * struct net_device *dev,
6b6e2725 945 * const unsigned char *addr)
77162022
JF
946 * Deletes the FDB entry from dev coresponding to addr.
947 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
948 * struct net_device *dev, int idx)
949 * Used to add FDB entries to dump requests. Implementers should add
950 * entries to skb and update idx with the number of entries.
e5a55a89
JF
951 *
952 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
953 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
24f11a5c 954 * struct net_device *dev, u32 filter_mask)
4bf84c35
JP
955 *
956 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
957 * Called to change device carrier. Soft-devices (like dummy, team, etc)
958 * which do not represent real hardware may define this to allow their
959 * userspace components to manage their virtual carrier state. Devices
960 * that determine carrier state from physical hardware properties (eg
961 * network cables) or protocol-dependent mechanisms (eg
962 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
963 *
964 * int (*ndo_get_phys_port_id)(struct net_device *dev,
965 * struct netdev_phys_port_id *ppid);
966 * Called to get ID of physical port of this device. If driver does
967 * not implement this, it is assumed that the hw is not able to have
968 * multiple net devices on single physical port.
53cf5275
JG
969 *
970 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 971 * sa_family_t sa_family, __be16 port);
53cf5275
JG
972 * Called by vxlan to notiy a driver about the UDP port and socket
973 * address family that vxlan is listnening to. It is called only when
974 * a new port starts listening. The operation is protected by the
975 * vxlan_net->sock_lock.
976 *
977 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 978 * sa_family_t sa_family, __be16 port);
53cf5275
JG
979 * Called by vxlan to notify the driver about a UDP port and socket
980 * address family that vxlan is not listening to anymore. The operation
981 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
982 *
983 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
984 * struct net_device *dev)
985 * Called by upper layer devices to accelerate switching or other
986 * station functionality into hardware. 'pdev is the lowerdev
987 * to use for the offload and 'dev' is the net device that will
988 * back the offload. Returns a pointer to the private structure
989 * the upper layer will maintain.
990 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
991 * Called by upper layer device to delete the station created
992 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
993 * the station and priv is the structure returned by the add
994 * operation.
995 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
996 * struct net_device *dev,
997 * void *priv);
998 * Callback to use for xmit over the accelerated station. This
999 * is used in place of ndo_start_xmit on accelerated net
1000 * devices.
d314774c
SH
1001 */
1002struct net_device_ops {
1003 int (*ndo_init)(struct net_device *dev);
1004 void (*ndo_uninit)(struct net_device *dev);
1005 int (*ndo_open)(struct net_device *dev);
1006 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 1007 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
1008 struct net_device *dev);
1009 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1010 struct sk_buff *skb,
99932d4f
DB
1011 void *accel_priv,
1012 select_queue_fallback_t fallback);
d314774c
SH
1013 void (*ndo_change_rx_flags)(struct net_device *dev,
1014 int flags);
d314774c 1015 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1016 int (*ndo_set_mac_address)(struct net_device *dev,
1017 void *addr);
d314774c 1018 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1019 int (*ndo_do_ioctl)(struct net_device *dev,
1020 struct ifreq *ifr, int cmd);
d314774c
SH
1021 int (*ndo_set_config)(struct net_device *dev,
1022 struct ifmap *map);
00829823
SH
1023 int (*ndo_change_mtu)(struct net_device *dev,
1024 int new_mtu);
1025 int (*ndo_neigh_setup)(struct net_device *dev,
1026 struct neigh_parms *);
d314774c
SH
1027 void (*ndo_tx_timeout) (struct net_device *dev);
1028
28172739
ED
1029 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1030 struct rtnl_link_stats64 *storage);
d314774c
SH
1031 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1032
8e586137 1033 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1034 __be16 proto, u16 vid);
8e586137 1035 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1036 __be16 proto, u16 vid);
d314774c 1037#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1038 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1039 int (*ndo_netpoll_setup)(struct net_device *dev,
47be03a2
AW
1040 struct netpoll_info *info,
1041 gfp_t gfp);
0e34e931 1042 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1043#endif
e0d1095a 1044#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1045 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1046#endif
95c26df8
WM
1047 int (*ndo_set_vf_mac)(struct net_device *dev,
1048 int queue, u8 *mac);
1049 int (*ndo_set_vf_vlan)(struct net_device *dev,
1050 int queue, u16 vlan, u8 qos);
1051 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
1052 int vf, int rate);
5f8444a3
GR
1053 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1054 int vf, bool setting);
95c26df8
WM
1055 int (*ndo_get_vf_config)(struct net_device *dev,
1056 int vf,
1057 struct ifla_vf_info *ivf);
1d8faf48
RE
1058 int (*ndo_set_vf_link_state)(struct net_device *dev,
1059 int vf, int link_state);
57b61080
SF
1060 int (*ndo_set_vf_port)(struct net_device *dev,
1061 int vf,
1062 struct nlattr *port[]);
1063 int (*ndo_get_vf_port)(struct net_device *dev,
1064 int vf, struct sk_buff *skb);
4f57c087 1065 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
d11ead75 1066#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1067 int (*ndo_fcoe_enable)(struct net_device *dev);
1068 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1069 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1070 u16 xid,
1071 struct scatterlist *sgl,
1072 unsigned int sgc);
1073 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1074 u16 xid);
6247e086
YZ
1075 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1076 u16 xid,
1077 struct scatterlist *sgl,
1078 unsigned int sgc);
68bad94e
NP
1079 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1080 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1081#endif
1082
d11ead75 1083#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1084#define NETDEV_FCOE_WWNN 0
1085#define NETDEV_FCOE_WWPN 1
1086 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1087 u64 *wwn, int type);
4d288d57 1088#endif
3c9c36bc 1089
c445477d
BH
1090#ifdef CONFIG_RFS_ACCEL
1091 int (*ndo_rx_flow_steer)(struct net_device *dev,
1092 const struct sk_buff *skb,
1093 u16 rxq_index,
1094 u32 flow_id);
1095#endif
fbaec0ea
JP
1096 int (*ndo_add_slave)(struct net_device *dev,
1097 struct net_device *slave_dev);
1098 int (*ndo_del_slave)(struct net_device *dev,
1099 struct net_device *slave_dev);
c8f44aff
MM
1100 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1101 netdev_features_t features);
5455c699 1102 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1103 netdev_features_t features);
da6a8fa0 1104 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1105 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1106
1107 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1108 struct nlattr *tb[],
77162022 1109 struct net_device *dev,
6b6e2725 1110 const unsigned char *addr,
77162022
JF
1111 u16 flags);
1112 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1113 struct nlattr *tb[],
77162022 1114 struct net_device *dev,
6b6e2725 1115 const unsigned char *addr);
77162022
JF
1116 int (*ndo_fdb_dump)(struct sk_buff *skb,
1117 struct netlink_callback *cb,
1118 struct net_device *dev,
1119 int idx);
e5a55a89
JF
1120
1121 int (*ndo_bridge_setlink)(struct net_device *dev,
1122 struct nlmsghdr *nlh);
1123 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1124 u32 pid, u32 seq,
6cbdceeb
VY
1125 struct net_device *dev,
1126 u32 filter_mask);
407af329
VY
1127 int (*ndo_bridge_dellink)(struct net_device *dev,
1128 struct nlmsghdr *nlh);
4bf84c35
JP
1129 int (*ndo_change_carrier)(struct net_device *dev,
1130 bool new_carrier);
66b52b0d
JP
1131 int (*ndo_get_phys_port_id)(struct net_device *dev,
1132 struct netdev_phys_port_id *ppid);
53cf5275
JG
1133 void (*ndo_add_vxlan_port)(struct net_device *dev,
1134 sa_family_t sa_family,
35e42379 1135 __be16 port);
53cf5275
JG
1136 void (*ndo_del_vxlan_port)(struct net_device *dev,
1137 sa_family_t sa_family,
35e42379 1138 __be16 port);
a6cc0cfa
JF
1139
1140 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1141 struct net_device *dev);
1142 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1143 void *priv);
1144
1145 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1146 struct net_device *dev,
1147 void *priv);
d314774c
SH
1148};
1149
7aa98047
LR
1150/**
1151 * enum net_device_priv_flags - &struct net_device priv_flags
1152 *
1153 * These are the &struct net_device, they are only set internally
1154 * by drivers and used in the kernel. These flags are invisible to
1155 * userspace, this means that the order of these flags can change
1156 * during any kernel release.
1157 *
1158 * You should have a pretty good reason to be extending these flags.
1159 *
1160 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1161 * @IFF_EBRIDGE: Ethernet bridging device
1162 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1163 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1164 * @IFF_MASTER_ALB: bonding master, balance-alb
1165 * @IFF_BONDING: bonding master or slave
1166 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1167 * @IFF_ISATAP: ISATAP interface (RFC4214)
1168 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1169 * @IFF_WAN_HDLC: WAN HDLC device
1170 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1171 * release skb->dst
1172 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1173 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1174 * @IFF_MACVLAN_PORT: device used as macvlan port
1175 * @IFF_BRIDGE_PORT: device used as bridge port
1176 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1177 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1178 * @IFF_UNICAST_FLT: Supports unicast filtering
1179 * @IFF_TEAM_PORT: device used as team port
1180 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1181 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1182 * change when it's running
1183 * @IFF_MACVLAN: Macvlan device
1184 */
1185enum netdev_priv_flags {
1186 IFF_802_1Q_VLAN = 1<<0,
1187 IFF_EBRIDGE = 1<<1,
1188 IFF_SLAVE_INACTIVE = 1<<2,
1189 IFF_MASTER_8023AD = 1<<3,
1190 IFF_MASTER_ALB = 1<<4,
1191 IFF_BONDING = 1<<5,
1192 IFF_SLAVE_NEEDARP = 1<<6,
1193 IFF_ISATAP = 1<<7,
1194 IFF_MASTER_ARPMON = 1<<8,
1195 IFF_WAN_HDLC = 1<<9,
1196 IFF_XMIT_DST_RELEASE = 1<<10,
1197 IFF_DONT_BRIDGE = 1<<11,
1198 IFF_DISABLE_NETPOLL = 1<<12,
1199 IFF_MACVLAN_PORT = 1<<13,
1200 IFF_BRIDGE_PORT = 1<<14,
1201 IFF_OVS_DATAPATH = 1<<15,
1202 IFF_TX_SKB_SHARING = 1<<16,
1203 IFF_UNICAST_FLT = 1<<17,
1204 IFF_TEAM_PORT = 1<<18,
1205 IFF_SUPP_NOFCS = 1<<19,
1206 IFF_LIVE_ADDR_CHANGE = 1<<20,
1207 IFF_MACVLAN = 1<<21,
1208};
1209
1210#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1211#define IFF_EBRIDGE IFF_EBRIDGE
1212#define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1213#define IFF_MASTER_8023AD IFF_MASTER_8023AD
1214#define IFF_MASTER_ALB IFF_MASTER_ALB
1215#define IFF_BONDING IFF_BONDING
1216#define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1217#define IFF_ISATAP IFF_ISATAP
1218#define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1219#define IFF_WAN_HDLC IFF_WAN_HDLC
1220#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1221#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1222#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1223#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1224#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1225#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1226#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1227#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1228#define IFF_TEAM_PORT IFF_TEAM_PORT
1229#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1230#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1231#define IFF_MACVLAN IFF_MACVLAN
1232
1da177e4
LT
1233/*
1234 * The DEVICE structure.
1235 * Actually, this whole structure is a big mistake. It mixes I/O
1236 * data with strictly "high-level" data, and it has to know about
1237 * almost every data structure used in the INET module.
1238 *
1239 * FIXME: cleanup struct net_device such that network protocol info
1240 * moves out.
1241 */
1242
d94d9fee 1243struct net_device {
1da177e4
LT
1244
1245 /*
1246 * This is the first field of the "visible" part of this structure
1247 * (i.e. as seen by users in the "Space.c" file). It is the name
724df615 1248 * of the interface.
1da177e4
LT
1249 */
1250 char name[IFNAMSIZ];
ed77134b 1251
9136461a 1252 /* device name hash chain, please keep it close to name[] */
9356b8fc 1253 struct hlist_node name_hlist;
9136461a 1254
0b815a1a
SH
1255 /* snmp alias */
1256 char *ifalias;
1da177e4
LT
1257
1258 /*
1259 * I/O specific fields
1260 * FIXME: Merge these and struct ifmap into one
1261 */
1262 unsigned long mem_end; /* shared mem end */
1263 unsigned long mem_start; /* shared mem start */
1264 unsigned long base_addr; /* device I/O address */
df42153c 1265 int irq; /* device IRQ number */
1da177e4
LT
1266
1267 /*
1268 * Some hardware also needs these fields, but they are not
1269 * part of the usual set specified in Space.c.
1270 */
1271
1da177e4
LT
1272 unsigned long state;
1273
7562f876 1274 struct list_head dev_list;
bea3348e 1275 struct list_head napi_list;
44a0873d 1276 struct list_head unreg_list;
5cde2829 1277 struct list_head close_list;
2f268f12
VF
1278
1279 /* directly linked devices, like slaves for bonding */
1280 struct {
1281 struct list_head upper;
1282 struct list_head lower;
1283 } adj_list;
1284
1285 /* all linked devices, *including* neighbours */
1286 struct {
1287 struct list_head upper;
1288 struct list_head lower;
1289 } all_adj_list;
4c3d5e7b 1290
1da177e4 1291
5455c699 1292 /* currently active device features */
c8f44aff 1293 netdev_features_t features;
5455c699 1294 /* user-changeable features */
c8f44aff 1295 netdev_features_t hw_features;
5455c699 1296 /* user-requested features */
c8f44aff 1297 netdev_features_t wanted_features;
1aac6267 1298 /* mask of features inheritable by VLAN devices */
c8f44aff 1299 netdev_features_t vlan_features;
6a674e9c
JG
1300 /* mask of features inherited by encapsulating devices
1301 * This field indicates what encapsulation offloads
1302 * the hardware is capable of doing, and drivers will
1303 * need to set them appropriately.
1304 */
1305 netdev_features_t hw_enc_features;
0d89d203
SH
1306 /* mask of fetures inheritable by MPLS */
1307 netdev_features_t mpls_features;
04ed3e74 1308
1da177e4
LT
1309 /* Interface index. Unique device identifier */
1310 int ifindex;
1311 int iflink;
1312
c45d286e 1313 struct net_device_stats stats;
caf586e5
ED
1314 atomic_long_t rx_dropped; /* dropped packets by core network
1315 * Do not use this in drivers.
1316 */
1da177e4 1317
b86e0280 1318#ifdef CONFIG_WIRELESS_EXT
1da177e4
LT
1319 /* List of functions to handle Wireless Extensions (instead of ioctl).
1320 * See <net/iw_handler.h> for details. Jean II */
1321 const struct iw_handler_def * wireless_handlers;
1322 /* Instance data managed by the core of Wireless Extensions. */
1323 struct iw_public_data * wireless_data;
b86e0280 1324#endif
d314774c
SH
1325 /* Management operations */
1326 const struct net_device_ops *netdev_ops;
76fd8593 1327 const struct ethtool_ops *ethtool_ops;
a6cc0cfa 1328 const struct forwarding_accel_ops *fwd_ops;
1da177e4 1329
3b04ddde
SH
1330 /* Hardware header description */
1331 const struct header_ops *header_ops;
1332
b00055aa 1333 unsigned int flags; /* interface flags (a la BSD) */
3bdc0eba
BG
1334 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1335 * See if.h for definitions. */
1da177e4 1336 unsigned short gflags;
1da177e4
LT
1337 unsigned short padded; /* How much padding added by alloc_netdev() */
1338
b00055aa
SR
1339 unsigned char operstate; /* RFC2863 operstate */
1340 unsigned char link_mode; /* mapping policy to operstate */
1341
bdc220da
JP
1342 unsigned char if_port; /* Selectable AUI, TP,..*/
1343 unsigned char dma; /* DMA channel */
1344
cd7b5396 1345 unsigned int mtu; /* interface MTU value */
1da177e4
LT
1346 unsigned short type; /* interface hardware type */
1347 unsigned short hard_header_len; /* hardware hdr length */
1da177e4 1348
f5184d26
JB
1349 /* extra head- and tailroom the hardware may need, but not in all cases
1350 * can this be guaranteed, especially tailroom. Some cases also use
1351 * LL_MAX_HEADER instead to allocate the skb.
1352 */
1353 unsigned short needed_headroom;
1354 unsigned short needed_tailroom;
1355
1da177e4 1356 /* Interface address info. */
a6f9a705 1357 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
c1f79426 1358 unsigned char addr_assign_type; /* hw address assignment type */
1da177e4 1359 unsigned char addr_len; /* hardware address length */
a0a9663d 1360 unsigned short neigh_priv_len;
dffebd2c
N
1361 unsigned short dev_id; /* Used to differentiate devices
1362 * that share the same link
1363 * layer address
1364 */
3f85944f
AV
1365 unsigned short dev_port; /* Used to differentiate
1366 * devices that share the same
1367 * function
1368 */
ccffad25 1369 spinlock_t addr_list_lock;
22bedad3
JP
1370 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1371 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
4c3d5e7b
ED
1372 struct netdev_hw_addr_list dev_addrs; /* list of device
1373 * hw addresses
1374 */
1375#ifdef CONFIG_SYSFS
1376 struct kset *queues_kset;
1377#endif
1378
2d348d1f 1379 bool uc_promisc;
9d45abe1
WC
1380 unsigned int promiscuity;
1381 unsigned int allmulti;
1da177e4 1382
1da177e4
LT
1383
1384 /* Protocol specific pointers */
65ac6a5f 1385
d11ead75 1386#if IS_ENABLED(CONFIG_VLAN_8021Q)
5b9ea6e0 1387 struct vlan_info __rcu *vlan_info; /* VLAN info */
65ac6a5f 1388#endif
34a430d7 1389#if IS_ENABLED(CONFIG_NET_DSA)
cf50dcc2 1390 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
37cb0620
YX
1391#endif
1392#if IS_ENABLED(CONFIG_TIPC)
1393 struct tipc_bearer __rcu *tipc_ptr; /* TIPC specific data */
91da11f8 1394#endif
1da177e4 1395 void *atalk_ptr; /* AppleTalk link */
95ae6b22 1396 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
fc766e4c 1397 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
198caeca 1398 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1da177e4 1399 void *ax25_ptr; /* AX.25 specific data */
704232c2
JB
1400 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1401 assign before registering */
1da177e4 1402
9356b8fc 1403/*
cd13539b 1404 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1405 */
f8ff080d 1406 unsigned long last_rx; /* Time of last Rx */
4dc89133 1407
9356b8fc 1408 /* Interface address info used in eth_type_trans() */
f001fde5
JP
1409 unsigned char *dev_addr; /* hw address, (before bcast
1410 because most packets are
1411 unicast) */
1412
0a9627f2 1413
a953be53 1414#ifdef CONFIG_SYSFS
0a9627f2
TH
1415 struct netdev_rx_queue *_rx;
1416
62fe0b40 1417 /* Number of RX queues allocated at register_netdev() time */
0a9627f2 1418 unsigned int num_rx_queues;
62fe0b40
BH
1419
1420 /* Number of RX queues currently active in device */
1421 unsigned int real_num_rx_queues;
c445477d 1422
df334545 1423#endif
0a9627f2 1424
61391cde 1425 rx_handler_func_t __rcu *rx_handler;
1426 void __rcu *rx_handler_data;
e8a0464c 1427
24824a09 1428 struct netdev_queue __rcu *ingress_queue;
4c3d5e7b
ED
1429 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1430
cd13539b
ED
1431
1432/*
1433 * Cache lines mostly used on transmit path
1434 */
e8a0464c 1435 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
fd2ea0a7
DM
1436
1437 /* Number of TX queues allocated at alloc_netdev_mq() time */
e8a0464c 1438 unsigned int num_tx_queues;
fd2ea0a7
DM
1439
1440 /* Number of TX queues currently active in device */
1441 unsigned int real_num_tx_queues;
1442
af356afa
PM
1443 /* root qdisc from userspace point of view */
1444 struct Qdisc *qdisc;
1445
1da177e4 1446 unsigned long tx_queue_len; /* Max frames per queue allowed */
c3f26a26 1447 spinlock_t tx_global_lock;
cd13539b 1448
bf264145 1449#ifdef CONFIG_XPS
a4177869 1450 struct xps_dev_maps __rcu *xps_maps;
bf264145 1451#endif
4c3d5e7b
ED
1452#ifdef CONFIG_RFS_ACCEL
1453 /* CPU reverse-mapping for RX completion interrupts, indexed
1454 * by RX queue number. Assigned by driver. This must only be
1455 * set if the ndo_rx_flow_steer operation is defined. */
1456 struct cpu_rmap *rx_cpu_rmap;
1457#endif
1d24eb48 1458
9356b8fc 1459 /* These may be needed for future network-power-down code. */
9d21493b
ED
1460
1461 /*
1462 * trans_start here is expensive for high speed devices on SMP,
1463 * please use netdev_queue->trans_start instead.
1464 */
9356b8fc
ED
1465 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1466
1467 int watchdog_timeo; /* used by dev_watchdog() */
1468 struct timer_list watchdog_timer;
1469
1da177e4 1470 /* Number of references to this device */
29b4433d 1471 int __percpu *pcpu_refcnt;
9356b8fc 1472
1da177e4
LT
1473 /* delayed register/unregister */
1474 struct list_head todo_list;
1da177e4
LT
1475 /* device index hash chain */
1476 struct hlist_node index_hlist;
1477
e014debe 1478 struct list_head link_watch_list;
572a103d 1479
1da177e4
LT
1480 /* register/unregister state machine */
1481 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1482 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1483 NETREG_UNREGISTERING, /* called unregister_netdevice */
1484 NETREG_UNREGISTERED, /* completed unregister todo */
1485 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1486 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1487 } reg_state:8;
1488
1489 bool dismantle; /* device is going do be freed */
a2835763
PM
1490
1491 enum {
1492 RTNL_LINK_INITIALIZED,
1493 RTNL_LINK_INITIALIZING,
1494 } rtnl_link_state:16;
1da177e4 1495
d314774c
SH
1496 /* Called from unregister, can be used to call free_netdev */
1497 void (*destructor)(struct net_device *dev);
1da177e4 1498
1da177e4 1499#ifdef CONFIG_NETPOLL
5fbee843 1500 struct netpoll_info __rcu *npinfo;
1da177e4 1501#endif
eae792b7 1502
c346dca1 1503#ifdef CONFIG_NET_NS
4a1c5371
EB
1504 /* Network namespace this network device is inside */
1505 struct net *nd_net;
c346dca1 1506#endif
4a1c5371 1507
4951704b 1508 /* mid-layer private */
a7855c78
ED
1509 union {
1510 void *ml_priv;
1511 struct pcpu_lstats __percpu *lstats; /* loopback stats */
8f84985f 1512 struct pcpu_sw_netstats __percpu *tstats;
6d81f41c 1513 struct pcpu_dstats __percpu *dstats; /* dummy stats */
2681128f 1514 struct pcpu_vstats __percpu *vstats; /* veth stats */
a7855c78 1515 };
eca9ebac 1516 /* GARP */
3cc77ec7 1517 struct garp_port __rcu *garp_port;
febf018d
DW
1518 /* MRP */
1519 struct mrp_port __rcu *mrp_port;
1da177e4 1520
1da177e4 1521 /* class/net/name entry */
43cb76d9 1522 struct device dev;
0c509a6c
EB
1523 /* space for optional device, statistics, and wireless sysfs groups */
1524 const struct attribute_group *sysfs_groups[4];
a953be53
MD
1525 /* space for optional per-rx queue attributes */
1526 const struct attribute_group *sysfs_rx_queue_group;
38f7b870
PM
1527
1528 /* rtnetlink link ops */
1529 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1530
82cc1a7a
PWJ
1531 /* for setting kernel sock attribute on TCP connection setup */
1532#define GSO_MAX_SIZE 65536
1533 unsigned int gso_max_size;
30b678d8
BH
1534#define GSO_MAX_SEGS 65535
1535 u16 gso_max_segs;
d314774c 1536
7a6b6f51 1537#ifdef CONFIG_DCB
2f90b865 1538 /* Data Center Bridging netlink ops */
32953543 1539 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1540#endif
4f57c087
JF
1541 u8 num_tc;
1542 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1543 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1544
d11ead75 1545#if IS_ENABLED(CONFIG_FCOE)
4d288d57
YZ
1546 /* max exchange id for FCoE LRO by ddp */
1547 unsigned int fcoe_ddp_xid;
5bc1421e 1548#endif
86f8515f 1549#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1550 struct netprio_map __rcu *priomap;
4d288d57 1551#endif
c1f19b51
RC
1552 /* phy device may attach itself for hardware timestamping */
1553 struct phy_device *phydev;
cbda10fa 1554
23d3b8bf
ED
1555 struct lock_class_key *qdisc_tx_busylock;
1556
cbda10fa
VD
1557 /* group the device belongs to */
1558 int group;
9136461a
ED
1559
1560 struct pm_qos_request pm_qos_req;
1da177e4 1561};
43cb76d9 1562#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1563
1564#define NETDEV_ALIGN 32
1da177e4 1565
4f57c087
JF
1566static inline
1567int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1568{
1569 return dev->prio_tc_map[prio & TC_BITMASK];
1570}
1571
1572static inline
1573int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1574{
1575 if (tc >= dev->num_tc)
1576 return -EINVAL;
1577
1578 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1579 return 0;
1580}
1581
1582static inline
1583void netdev_reset_tc(struct net_device *dev)
1584{
1585 dev->num_tc = 0;
1586 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1587 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1588}
1589
1590static inline
1591int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1592{
1593 if (tc >= dev->num_tc)
1594 return -EINVAL;
1595
1596 dev->tc_to_txq[tc].count = count;
1597 dev->tc_to_txq[tc].offset = offset;
1598 return 0;
1599}
1600
1601static inline
1602int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1603{
1604 if (num_tc > TC_MAX_QUEUE)
1605 return -EINVAL;
1606
1607 dev->num_tc = num_tc;
1608 return 0;
1609}
1610
1611static inline
1612int netdev_get_num_tc(struct net_device *dev)
1613{
1614 return dev->num_tc;
1615}
1616
e8a0464c
DM
1617static inline
1618struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1619 unsigned int index)
1620{
1621 return &dev->_tx[index];
1622}
1623
1624static inline void netdev_for_each_tx_queue(struct net_device *dev,
1625 void (*f)(struct net_device *,
1626 struct netdev_queue *,
1627 void *),
1628 void *arg)
1629{
1630 unsigned int i;
1631
1632 for (i = 0; i < dev->num_tx_queues; i++)
1633 f(dev, &dev->_tx[i], arg);
1634}
1635
f629d208 1636struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1637 struct sk_buff *skb,
1638 void *accel_priv);
8c4c49df 1639
c346dca1
YH
1640/*
1641 * Net namespace inlines
1642 */
1643static inline
1644struct net *dev_net(const struct net_device *dev)
1645{
c2d9ba9b 1646 return read_pnet(&dev->nd_net);
c346dca1
YH
1647}
1648
1649static inline
f5aa23fd 1650void dev_net_set(struct net_device *dev, struct net *net)
c346dca1
YH
1651{
1652#ifdef CONFIG_NET_NS
f3005d7f
DL
1653 release_net(dev->nd_net);
1654 dev->nd_net = hold_net(net);
c346dca1
YH
1655#endif
1656}
1657
cf85d08f
LB
1658static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1659{
1660#ifdef CONFIG_NET_DSA_TAG_DSA
1661 if (dev->dsa_ptr != NULL)
1662 return dsa_uses_dsa_tags(dev->dsa_ptr);
1663#endif
1664
1665 return 0;
1666}
1667
396138f0
LB
1668static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1669{
1670#ifdef CONFIG_NET_DSA_TAG_TRAILER
1671 if (dev->dsa_ptr != NULL)
1672 return dsa_uses_trailer_tags(dev->dsa_ptr);
1673#endif
1674
1675 return 0;
1676}
1677
bea3348e
SH
1678/**
1679 * netdev_priv - access network device private data
1680 * @dev: network device
1681 *
1682 * Get network device private data
1683 */
6472ce60 1684static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1685{
1ce8e7b5 1686 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1687}
1688
1da177e4
LT
1689/* Set the sysfs physical device reference for the network logical device
1690 * if set prior to registration will cause a symlink during initialization.
1691 */
43cb76d9 1692#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1693
384912ed 1694/* Set the sysfs device type for the network logical device to allow
3f79410c 1695 * fine-grained identification of different network device types. For
384912ed
MH
1696 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1697 */
1698#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1699
82dc3c63
ED
1700/* Default NAPI poll() weight
1701 * Device drivers are strongly advised to not use bigger value
1702 */
1703#define NAPI_POLL_WEIGHT 64
1704
3b582cc1
SH
1705/**
1706 * netif_napi_add - initialize a napi context
1707 * @dev: network device
1708 * @napi: napi context
1709 * @poll: polling function
1710 * @weight: default weight
1711 *
1712 * netif_napi_add() must be used to initialize a napi context prior to calling
1713 * *any* of the other napi related functions.
1714 */
d565b0a1
HX
1715void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1716 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1717
d8156534
AD
1718/**
1719 * netif_napi_del - remove a napi context
1720 * @napi: napi context
1721 *
1722 * netif_napi_del() removes a napi context from the network device napi list
1723 */
d565b0a1
HX
1724void netif_napi_del(struct napi_struct *napi);
1725
1726struct napi_gro_cb {
78a478d0
HX
1727 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1728 void *frag0;
1729
7489594c
HX
1730 /* Length of frag0. */
1731 unsigned int frag0_len;
1732
86911732
HX
1733 /* This indicates where we are processing relative to skb->data. */
1734 int data_offset;
1735
d565b0a1 1736 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
1737 u16 flush;
1738
1739 /* Save the IP ID here and check when we get to the transport layer */
1740 u16 flush_id;
d565b0a1
HX
1741
1742 /* Number of segments aggregated. */
2e71a6f8
ED
1743 u16 count;
1744
1745 /* This is non-zero if the packet may be of the same flow. */
1746 u8 same_flow;
5d38a079
HX
1747
1748 /* Free the skb? */
2e71a6f8 1749 u8 free;
d7e8883c
ED
1750#define NAPI_GRO_FREE 1
1751#define NAPI_GRO_FREE_STOLEN_HEAD 2
2e71a6f8
ED
1752
1753 /* jiffies when first packet was created/queued */
1754 unsigned long age;
86347245
ED
1755
1756 /* Used in ipv6_gro_receive() */
b582ef09
OG
1757 u16 proto;
1758
1759 /* Used in udp_gro_receive */
1760 u16 udp_mark;
c3c7c254 1761
bf5a755f
JC
1762 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1763 __wsum csum;
1764
c3c7c254
ED
1765 /* used in skb_gro_receive() slow path */
1766 struct sk_buff *last;
d565b0a1
HX
1767};
1768
1769#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 1770
1da177e4 1771struct packet_type {
f2ccd8fa
DM
1772 __be16 type; /* This is really htons(ether_type). */
1773 struct net_device *dev; /* NULL is wildcarded here */
1774 int (*func) (struct sk_buff *,
1775 struct net_device *,
1776 struct packet_type *,
1777 struct net_device *);
c0de08d0
EL
1778 bool (*id_match)(struct packet_type *ptype,
1779 struct sock *sk);
1da177e4
LT
1780 void *af_packet_priv;
1781 struct list_head list;
1782};
1783
f191a1d1 1784struct offload_callbacks {
576a30eb 1785 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 1786 netdev_features_t features);
a430a43d 1787 int (*gso_send_check)(struct sk_buff *skb);
d565b0a1
HX
1788 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1789 struct sk_buff *skb);
299603e8 1790 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
1791};
1792
1793struct packet_offload {
1794 __be16 type; /* This is really htons(ether_type). */
1795 struct offload_callbacks callbacks;
1796 struct list_head list;
1da177e4
LT
1797};
1798
b582ef09
OG
1799struct udp_offload {
1800 __be16 port;
1801 struct offload_callbacks callbacks;
1802};
1803
8f84985f
LR
1804/* often modified stats are per cpu, other are shared (netdev->stats) */
1805struct pcpu_sw_netstats {
1806 u64 rx_packets;
1807 u64 rx_bytes;
1808 u64 tx_packets;
1809 u64 tx_bytes;
1810 struct u64_stats_sync syncp;
1811};
1812
1c213bd2
WC
1813#define netdev_alloc_pcpu_stats(type) \
1814({ \
693350c2 1815 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1c213bd2
WC
1816 if (pcpu_stats) { \
1817 int i; \
1818 for_each_possible_cpu(i) { \
1819 typeof(type) *stat; \
1820 stat = per_cpu_ptr(pcpu_stats, i); \
1821 u64_stats_init(&stat->syncp); \
1822 } \
1823 } \
1824 pcpu_stats; \
1825})
1826
1da177e4
LT
1827#include <linux/notifier.h>
1828
dcfe1421
AW
1829/* netdevice notifier chain. Please remember to update the rtnetlink
1830 * notification exclusion list in rtnetlink_event() when adding new
1831 * types.
1832 */
1833#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1834#define NETDEV_DOWN 0x0002
1835#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1836 detected a hardware crash and restarted
1837 - we can use this eg to kick tcp sessions
1838 once done */
1839#define NETDEV_CHANGE 0x0004 /* Notify device state change */
1840#define NETDEV_REGISTER 0x0005
1841#define NETDEV_UNREGISTER 0x0006
1d486bfb 1842#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
1843#define NETDEV_CHANGEADDR 0x0008
1844#define NETDEV_GOING_DOWN 0x0009
1845#define NETDEV_CHANGENAME 0x000A
1846#define NETDEV_FEAT_CHANGE 0x000B
1847#define NETDEV_BONDING_FAILOVER 0x000C
1848#define NETDEV_PRE_UP 0x000D
1849#define NETDEV_PRE_TYPE_CHANGE 0x000E
1850#define NETDEV_POST_TYPE_CHANGE 0x000F
1851#define NETDEV_POST_INIT 0x0010
0115e8e3 1852#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
1853#define NETDEV_RELEASE 0x0012
1854#define NETDEV_NOTIFY_PEERS 0x0013
1855#define NETDEV_JOIN 0x0014
42e52bf9 1856#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 1857#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 1858#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
dcfe1421 1859
f629d208
JP
1860int register_netdevice_notifier(struct notifier_block *nb);
1861int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
1862
1863struct netdev_notifier_info {
1864 struct net_device *dev;
1865};
1866
be9efd36
JP
1867struct netdev_notifier_change_info {
1868 struct netdev_notifier_info info; /* must be first */
1869 unsigned int flags_changed;
1870};
1871
75538c2b
CW
1872static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1873 struct net_device *dev)
1874{
1875 info->dev = dev;
1876}
1877
351638e7
JP
1878static inline struct net_device *
1879netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1880{
1881 return info->dev;
1882}
1883
f629d208 1884int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
1885
1886
1da177e4
LT
1887extern rwlock_t dev_base_lock; /* Device list lock */
1888
881d966b
EB
1889#define for_each_netdev(net, d) \
1890 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
1891#define for_each_netdev_reverse(net, d) \
1892 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
1893#define for_each_netdev_rcu(net, d) \
1894 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
1895#define for_each_netdev_safe(net, d, n) \
1896 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1897#define for_each_netdev_continue(net, d) \
1898 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 1899#define for_each_netdev_continue_rcu(net, d) \
1900 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 1901#define for_each_netdev_in_bond_rcu(bond, slave) \
1902 for_each_netdev_rcu(&init_net, slave) \
1903 if (netdev_master_upper_dev_get_rcu(slave) == bond)
881d966b 1904#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 1905
a050c33f
DL
1906static inline struct net_device *next_net_device(struct net_device *dev)
1907{
1908 struct list_head *lh;
1909 struct net *net;
1910
c346dca1 1911 net = dev_net(dev);
a050c33f
DL
1912 lh = dev->dev_list.next;
1913 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1914}
1915
ce81b76a
ED
1916static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1917{
1918 struct list_head *lh;
1919 struct net *net;
1920
1921 net = dev_net(dev);
ccf43438 1922 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
1923 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1924}
1925
a050c33f
DL
1926static inline struct net_device *first_net_device(struct net *net)
1927{
1928 return list_empty(&net->dev_base_head) ? NULL :
1929 net_device_entry(net->dev_base_head.next);
1930}
7562f876 1931
ccf43438
ED
1932static inline struct net_device *first_net_device_rcu(struct net *net)
1933{
1934 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1935
1936 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1937}
1938
f629d208
JP
1939int netdev_boot_setup_check(struct net_device *dev);
1940unsigned long netdev_boot_base(const char *prefix, int unit);
1941struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1942 const char *hwaddr);
1943struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1944struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1945void dev_add_pack(struct packet_type *pt);
1946void dev_remove_pack(struct packet_type *pt);
1947void __dev_remove_pack(struct packet_type *pt);
1948void dev_add_offload(struct packet_offload *po);
1949void dev_remove_offload(struct packet_offload *po);
f629d208
JP
1950
1951struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1952 unsigned short mask);
1953struct net_device *dev_get_by_name(struct net *net, const char *name);
1954struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1955struct net_device *__dev_get_by_name(struct net *net, const char *name);
1956int dev_alloc_name(struct net_device *dev, const char *name);
1957int dev_open(struct net_device *dev);
1958int dev_close(struct net_device *dev);
1959void dev_disable_lro(struct net_device *dev);
1960int dev_loopback_xmit(struct sk_buff *newskb);
1961int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 1962int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
1963int register_netdevice(struct net_device *dev);
1964void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
1965void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
1966static inline void unregister_netdevice(struct net_device *dev)
1967{
1968 unregister_netdevice_queue(dev, NULL);
1969}
1970
f629d208
JP
1971int netdev_refcnt_read(const struct net_device *dev);
1972void free_netdev(struct net_device *dev);
74d332c1 1973void netdev_freemem(struct net_device *dev);
f629d208
JP
1974void synchronize_net(void);
1975int init_dummy_netdev(struct net_device *dev);
937f1ba5 1976
f629d208
JP
1977struct net_device *dev_get_by_index(struct net *net, int ifindex);
1978struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1979struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1980int netdev_get_name(struct net *net, char *name, int ifindex);
1981int dev_restart(struct net_device *dev);
f629d208 1982int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
1983
1984static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1985{
1986 return NAPI_GRO_CB(skb)->data_offset;
1987}
1988
1989static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1990{
1991 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1992}
1993
1994static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1995{
1996 NAPI_GRO_CB(skb)->data_offset += len;
1997}
1998
a5b1cf28
HX
1999static inline void *skb_gro_header_fast(struct sk_buff *skb,
2000 unsigned int offset)
86911732 2001{
a5b1cf28
HX
2002 return NAPI_GRO_CB(skb)->frag0 + offset;
2003}
78a478d0 2004
a5b1cf28
HX
2005static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2006{
2007 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2008}
78a478d0 2009
a5b1cf28
HX
2010static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2011 unsigned int offset)
2012{
17dd759c
HX
2013 if (!pskb_may_pull(skb, hlen))
2014 return NULL;
2015
a5b1cf28
HX
2016 NAPI_GRO_CB(skb)->frag0 = NULL;
2017 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2018 return skb->data + offset;
86911732 2019}
1da177e4 2020
aa4b9f53
HX
2021static inline void *skb_gro_mac_header(struct sk_buff *skb)
2022{
78d3fd0b 2023 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
aa4b9f53
HX
2024}
2025
36e7b1b8
HX
2026static inline void *skb_gro_network_header(struct sk_buff *skb)
2027{
78d3fd0b
HX
2028 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2029 skb_network_offset(skb);
36e7b1b8
HX
2030}
2031
bf5a755f
JC
2032static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2033 const void *start, unsigned int len)
2034{
2035 if (skb->ip_summed == CHECKSUM_COMPLETE)
2036 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2037 csum_partial(start, len, 0));
2038}
2039
0c4e8581
SH
2040static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2041 unsigned short type,
3b04ddde 2042 const void *daddr, const void *saddr,
95c96174 2043 unsigned int len)
0c4e8581 2044{
f1ecfd5d 2045 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2046 return 0;
3b04ddde
SH
2047
2048 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2049}
2050
b95cce35
SH
2051static inline int dev_parse_header(const struct sk_buff *skb,
2052 unsigned char *haddr)
2053{
2054 const struct net_device *dev = skb->dev;
2055
1b83336b 2056 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2057 return 0;
3b04ddde 2058 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2059}
2060
2205369a
DM
2061static inline int dev_rebuild_header(struct sk_buff *skb)
2062{
2063 const struct net_device *dev = skb->dev;
2064
2065 if (!dev->header_ops || !dev->header_ops->rebuild)
2066 return 0;
2067 return dev->header_ops->rebuild(skb);
2068}
2069
1da177e4 2070typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2071int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2072static inline int unregister_gifconf(unsigned int family)
2073{
2074 return register_gifconf(family, NULL);
2075}
2076
99bbc707 2077#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2078#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2079struct sd_flow_limit {
2080 u64 count;
2081 unsigned int num_buckets;
2082 unsigned int history_head;
2083 u16 history[FLOW_LIMIT_HISTORY];
2084 u8 buckets[];
2085};
2086
2087extern int netdev_flow_limit_table_len;
2088#endif /* CONFIG_NET_FLOW_LIMIT */
2089
1da177e4 2090/*
88751275 2091 * Incoming packets are placed on per-cpu queues
1da177e4 2092 */
d94d9fee 2093struct softnet_data {
37437bb2 2094 struct Qdisc *output_queue;
a9cbd588 2095 struct Qdisc **output_queue_tailp;
1da177e4 2096 struct list_head poll_list;
1da177e4 2097 struct sk_buff *completion_queue;
6e7676c1 2098 struct sk_buff_head process_queue;
1da177e4 2099
dee42870 2100 /* stats */
cd7b5396
DM
2101 unsigned int processed;
2102 unsigned int time_squeeze;
2103 unsigned int cpu_collision;
2104 unsigned int received_rps;
dee42870 2105
fd793d89 2106#ifdef CONFIG_RPS
88751275
ED
2107 struct softnet_data *rps_ipi_list;
2108
2109 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 2110 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2111 struct softnet_data *rps_ipi_next;
2112 unsigned int cpu;
fec5e652 2113 unsigned int input_queue_head;
76cc8b13 2114 unsigned int input_queue_tail;
1e94d72f 2115#endif
95c96174 2116 unsigned int dropped;
0a9627f2 2117 struct sk_buff_head input_pkt_queue;
bea3348e 2118 struct napi_struct backlog;
99bbc707
WB
2119
2120#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2121 struct sd_flow_limit __rcu *flow_limit;
99bbc707 2122#endif
1da177e4
LT
2123};
2124
76cc8b13 2125static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2126{
2127#ifdef CONFIG_RPS
76cc8b13
TH
2128 sd->input_queue_head++;
2129#endif
2130}
2131
2132static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2133 unsigned int *qtail)
2134{
2135#ifdef CONFIG_RPS
2136 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2137#endif
2138}
2139
0a9627f2 2140DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2141
f629d208 2142void __netif_schedule(struct Qdisc *q);
1da177e4 2143
86d804e1 2144static inline void netif_schedule_queue(struct netdev_queue *txq)
1da177e4 2145{
73466498 2146 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
37437bb2 2147 __netif_schedule(txq->qdisc);
86d804e1
DM
2148}
2149
fd2ea0a7
DM
2150static inline void netif_tx_schedule_all(struct net_device *dev)
2151{
2152 unsigned int i;
2153
2154 for (i = 0; i < dev->num_tx_queues; i++)
2155 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2156}
2157
d29f749e
DJ
2158static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2159{
73466498 2160 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2161}
2162
bea3348e
SH
2163/**
2164 * netif_start_queue - allow transmit
2165 * @dev: network device
2166 *
2167 * Allow upper layers to call the device hard_start_xmit routine.
2168 */
1da177e4
LT
2169static inline void netif_start_queue(struct net_device *dev)
2170{
e8a0464c 2171 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2172}
2173
fd2ea0a7
DM
2174static inline void netif_tx_start_all_queues(struct net_device *dev)
2175{
2176 unsigned int i;
2177
2178 for (i = 0; i < dev->num_tx_queues; i++) {
2179 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2180 netif_tx_start_queue(txq);
2181 }
2182}
2183
79d16385 2184static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1da177e4 2185{
73466498 2186 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
37437bb2 2187 __netif_schedule(dev_queue->qdisc);
79d16385
DM
2188}
2189
d29f749e
DJ
2190/**
2191 * netif_wake_queue - restart transmit
2192 * @dev: network device
2193 *
2194 * Allow upper layers to call the device hard_start_xmit routine.
2195 * Used for flow control when transmit resources are available.
2196 */
79d16385
DM
2197static inline void netif_wake_queue(struct net_device *dev)
2198{
e8a0464c 2199 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2200}
2201
fd2ea0a7
DM
2202static inline void netif_tx_wake_all_queues(struct net_device *dev)
2203{
2204 unsigned int i;
2205
2206 for (i = 0; i < dev->num_tx_queues; i++) {
2207 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2208 netif_tx_wake_queue(txq);
2209 }
2210}
2211
d29f749e
DJ
2212static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2213{
18543a64 2214 if (WARN_ON(!dev_queue)) {
256ee435 2215 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
18543a64
GC
2216 return;
2217 }
73466498 2218 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2219}
2220
bea3348e
SH
2221/**
2222 * netif_stop_queue - stop transmitted packets
2223 * @dev: network device
2224 *
2225 * Stop upper layers calling the device hard_start_xmit routine.
2226 * Used for flow control when transmit resources are unavailable.
2227 */
1da177e4
LT
2228static inline void netif_stop_queue(struct net_device *dev)
2229{
e8a0464c 2230 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2231}
2232
fd2ea0a7
DM
2233static inline void netif_tx_stop_all_queues(struct net_device *dev)
2234{
2235 unsigned int i;
2236
2237 for (i = 0; i < dev->num_tx_queues; i++) {
2238 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2239 netif_tx_stop_queue(txq);
2240 }
2241}
2242
4d29515f 2243static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2244{
73466498 2245 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2246}
2247
bea3348e
SH
2248/**
2249 * netif_queue_stopped - test if transmit queue is flowblocked
2250 * @dev: network device
2251 *
2252 * Test if transmit queue on device is currently unable to send.
2253 */
4d29515f 2254static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2255{
e8a0464c 2256 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2257}
2258
4d29515f 2259static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2260{
73466498
TH
2261 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2262}
2263
4d29515f 2264static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2265{
2266 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2267}
2268
c5d67bd7
TH
2269static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2270 unsigned int bytes)
2271{
114cf580
TH
2272#ifdef CONFIG_BQL
2273 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2274
2275 if (likely(dql_avail(&dev_queue->dql) >= 0))
2276 return;
2277
2278 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2279
2280 /*
2281 * The XOFF flag must be set before checking the dql_avail below,
2282 * because in netdev_tx_completed_queue we update the dql_completed
2283 * before checking the XOFF flag.
2284 */
2285 smp_mb();
2286
2287 /* check again in case another CPU has just made room avail */
2288 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2289 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2290#endif
c5d67bd7
TH
2291}
2292
0042d0c8
FF
2293/**
2294 * netdev_sent_queue - report the number of bytes queued to hardware
2295 * @dev: network device
2296 * @bytes: number of bytes queued to the hardware device queue
2297 *
2298 * Report the number of bytes queued for sending/completion to the network
2299 * device hardware queue. @bytes should be a good approximation and should
2300 * exactly match netdev_completed_queue() @bytes
2301 */
c5d67bd7
TH
2302static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2303{
2304 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2305}
2306
2307static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2308 unsigned int pkts, unsigned int bytes)
c5d67bd7 2309{
114cf580 2310#ifdef CONFIG_BQL
b37c0fbe
AD
2311 if (unlikely(!bytes))
2312 return;
2313
2314 dql_completed(&dev_queue->dql, bytes);
2315
2316 /*
2317 * Without the memory barrier there is a small possiblity that
2318 * netdev_tx_sent_queue will miss the update and cause the queue to
2319 * be stopped forever
2320 */
2321 smp_mb();
2322
2323 if (dql_avail(&dev_queue->dql) < 0)
2324 return;
2325
2326 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2327 netif_schedule_queue(dev_queue);
114cf580 2328#endif
c5d67bd7
TH
2329}
2330
0042d0c8
FF
2331/**
2332 * netdev_completed_queue - report bytes and packets completed by device
2333 * @dev: network device
2334 * @pkts: actual number of packets sent over the medium
2335 * @bytes: actual number of bytes sent over the medium
2336 *
2337 * Report the number of bytes and packets transmitted by the network device
2338 * hardware queue over the physical medium, @bytes must exactly match the
2339 * @bytes amount passed to netdev_sent_queue()
2340 */
c5d67bd7 2341static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2342 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2343{
2344 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2345}
2346
2347static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2348{
114cf580 2349#ifdef CONFIG_BQL
5c490354 2350 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2351 dql_reset(&q->dql);
2352#endif
c5d67bd7
TH
2353}
2354
0042d0c8
FF
2355/**
2356 * netdev_reset_queue - reset the packets and bytes count of a network device
2357 * @dev_queue: network device
2358 *
2359 * Reset the bytes and packet count of a network device and clear the
2360 * software flow control OFF bit for this network device
2361 */
c5d67bd7
TH
2362static inline void netdev_reset_queue(struct net_device *dev_queue)
2363{
2364 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2365}
2366
b9507bda
DB
2367/**
2368 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2369 * @dev: network device
2370 * @queue_index: given tx queue index
2371 *
2372 * Returns 0 if given tx queue index >= number of device tx queues,
2373 * otherwise returns the originally passed tx queue index.
2374 */
2375static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2376{
2377 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2378 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2379 dev->name, queue_index,
2380 dev->real_num_tx_queues);
2381 return 0;
2382 }
2383
2384 return queue_index;
2385}
2386
bea3348e
SH
2387/**
2388 * netif_running - test if up
2389 * @dev: network device
2390 *
2391 * Test if the device has been brought up.
2392 */
4d29515f 2393static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2394{
2395 return test_bit(__LINK_STATE_START, &dev->state);
2396}
2397
f25f4e44
PWJ
2398/*
2399 * Routines to manage the subqueues on a device. We only need start
2400 * stop, and a check if it's stopped. All other device management is
2401 * done at the overall netdevice level.
2402 * Also test the device if we're multiqueue.
2403 */
bea3348e
SH
2404
2405/**
2406 * netif_start_subqueue - allow sending packets on subqueue
2407 * @dev: network device
2408 * @queue_index: sub queue index
2409 *
2410 * Start individual transmit queue of a device with multiple transmit queues.
2411 */
f25f4e44
PWJ
2412static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2413{
fd2ea0a7 2414 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2415
2416 netif_tx_start_queue(txq);
f25f4e44
PWJ
2417}
2418
bea3348e
SH
2419/**
2420 * netif_stop_subqueue - stop sending packets on subqueue
2421 * @dev: network device
2422 * @queue_index: sub queue index
2423 *
2424 * Stop individual transmit queue of a device with multiple transmit queues.
2425 */
f25f4e44
PWJ
2426static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2427{
fd2ea0a7 2428 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 2429 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2430}
2431
bea3348e
SH
2432/**
2433 * netif_subqueue_stopped - test status of subqueue
2434 * @dev: network device
2435 * @queue_index: sub queue index
2436 *
2437 * Check individual transmit queue of a device with multiple transmit queues.
2438 */
4d29515f
DM
2439static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2440 u16 queue_index)
f25f4e44 2441{
fd2ea0a7 2442 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2443
2444 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2445}
2446
4d29515f
DM
2447static inline bool netif_subqueue_stopped(const struct net_device *dev,
2448 struct sk_buff *skb)
668f895a
PE
2449{
2450 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2451}
bea3348e
SH
2452
2453/**
2454 * netif_wake_subqueue - allow sending packets on subqueue
2455 * @dev: network device
2456 * @queue_index: sub queue index
2457 *
2458 * Resume individual transmit queue of a device with multiple transmit queues.
2459 */
f25f4e44
PWJ
2460static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2461{
fd2ea0a7 2462 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
73466498 2463 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
37437bb2 2464 __netif_schedule(txq->qdisc);
f25f4e44
PWJ
2465}
2466
537c00de 2467#ifdef CONFIG_XPS
53af53ae 2468int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2469 u16 index);
537c00de
AD
2470#else
2471static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2472 const struct cpumask *mask,
537c00de
AD
2473 u16 index)
2474{
2475 return 0;
2476}
2477#endif
2478
a3d22a68
VZ
2479/*
2480 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2481 * as a distribution range limit for the returned value.
2482 */
2483static inline u16 skb_tx_hash(const struct net_device *dev,
2484 const struct sk_buff *skb)
2485{
2486 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2487}
2488
bea3348e
SH
2489/**
2490 * netif_is_multiqueue - test if device has multiple transmit queues
2491 * @dev: network device
2492 *
2493 * Check if device has multiple transmit queues
bea3348e 2494 */
4d29515f 2495static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2496{
a02cec21 2497 return dev->num_tx_queues > 1;
f25f4e44 2498}
1da177e4 2499
f629d208 2500int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2501
a953be53 2502#ifdef CONFIG_SYSFS
f629d208 2503int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
2504#else
2505static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2506 unsigned int rxq)
2507{
2508 return 0;
2509}
2510#endif
2511
3171d026
BH
2512static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2513 const struct net_device *from_dev)
2514{
ee6ae1a1
JP
2515 int err;
2516
2517 err = netif_set_real_num_tx_queues(to_dev,
2518 from_dev->real_num_tx_queues);
2519 if (err)
2520 return err;
a953be53 2521#ifdef CONFIG_SYSFS
3171d026
BH
2522 return netif_set_real_num_rx_queues(to_dev,
2523 from_dev->real_num_rx_queues);
2524#else
2525 return 0;
2526#endif
2527}
2528
a953be53
MD
2529#ifdef CONFIG_SYSFS
2530static inline unsigned int get_netdev_rx_queue_index(
2531 struct netdev_rx_queue *queue)
2532{
2533 struct net_device *dev = queue->dev;
2534 int index = queue - dev->_rx;
2535
2536 BUG_ON(index >= dev->num_rx_queues);
2537 return index;
2538}
2539#endif
2540
16917b87 2541#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 2542int netif_get_num_default_rss_queues(void);
16917b87 2543
e6247027
ED
2544enum skb_free_reason {
2545 SKB_REASON_CONSUMED,
2546 SKB_REASON_DROPPED,
2547};
2548
2549void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2550void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 2551
e6247027
ED
2552/*
2553 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2554 * interrupt context or with hardware interrupts being disabled.
2555 * (in_irq() || irqs_disabled())
2556 *
2557 * We provide four helpers that can be used in following contexts :
2558 *
2559 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2560 * replacing kfree_skb(skb)
2561 *
2562 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2563 * Typically used in place of consume_skb(skb) in TX completion path
2564 *
2565 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2566 * replacing kfree_skb(skb)
2567 *
2568 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2569 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 2570 */
e6247027
ED
2571static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2572{
2573 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2574}
2575
2576static inline void dev_consume_skb_irq(struct sk_buff *skb)
2577{
2578 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2579}
2580
2581static inline void dev_kfree_skb_any(struct sk_buff *skb)
2582{
2583 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2584}
2585
2586static inline void dev_consume_skb_any(struct sk_buff *skb)
2587{
2588 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2589}
1da177e4 2590
f629d208
JP
2591int netif_rx(struct sk_buff *skb);
2592int netif_rx_ni(struct sk_buff *skb);
2593int netif_receive_skb(struct sk_buff *skb);
2594gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2595void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2596struct sk_buff *napi_get_frags(struct napi_struct *napi);
2597gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
2598struct packet_offload *gro_find_receive_by_type(__be16 type);
2599struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
2600
2601static inline void napi_free_frags(struct napi_struct *napi)
2602{
2603 kfree_skb(napi->skb);
2604 napi->skb = NULL;
2605}
2606
f629d208
JP
2607int netdev_rx_handler_register(struct net_device *dev,
2608 rx_handler_func_t *rx_handler,
2609 void *rx_handler_data);
2610void netdev_rx_handler_unregister(struct net_device *dev);
2611
2612bool dev_valid_name(const char *name);
2613int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2614int dev_ethtool(struct net *net, struct ifreq *);
2615unsigned int dev_get_flags(const struct net_device *);
2616int __dev_change_flags(struct net_device *, unsigned int flags);
2617int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
2618void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2619 unsigned int gchanges);
f629d208
JP
2620int dev_change_name(struct net_device *, const char *);
2621int dev_set_alias(struct net_device *, const char *, size_t);
2622int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2623int dev_set_mtu(struct net_device *, int);
2624void dev_set_group(struct net_device *, int);
2625int dev_set_mac_address(struct net_device *, struct sockaddr *);
2626int dev_change_carrier(struct net_device *, bool new_carrier);
2627int dev_get_phys_port_id(struct net_device *dev,
2628 struct netdev_phys_port_id *ppid);
2629int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
f663dd9a 2630 struct netdev_queue *txq);
f629d208 2631int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1da177e4 2632
20380731 2633extern int netdev_budget;
1da177e4
LT
2634
2635/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 2636void netdev_run_todo(void);
1da177e4 2637
bea3348e
SH
2638/**
2639 * dev_put - release reference to device
2640 * @dev: network device
2641 *
9ef4429b 2642 * Release reference to device to allow it to be freed.
bea3348e 2643 */
1da177e4
LT
2644static inline void dev_put(struct net_device *dev)
2645{
933393f5 2646 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
2647}
2648
bea3348e
SH
2649/**
2650 * dev_hold - get reference to device
2651 * @dev: network device
2652 *
9ef4429b 2653 * Hold reference to device to keep it from being freed.
bea3348e 2654 */
15333061
SH
2655static inline void dev_hold(struct net_device *dev)
2656{
933393f5 2657 this_cpu_inc(*dev->pcpu_refcnt);
15333061 2658}
1da177e4
LT
2659
2660/* Carrier loss detection, dial on demand. The functions netif_carrier_on
2661 * and _off may be called from IRQ context, but it is caller
2662 * who is responsible for serialization of these calls.
b00055aa
SR
2663 *
2664 * The name carrier is inappropriate, these functions should really be
2665 * called netif_lowerlayer_*() because they represent the state of any
2666 * kind of lower layer not just hardware media.
1da177e4
LT
2667 */
2668
f629d208
JP
2669void linkwatch_init_dev(struct net_device *dev);
2670void linkwatch_fire_event(struct net_device *dev);
2671void linkwatch_forget_dev(struct net_device *dev);
1da177e4 2672
bea3348e
SH
2673/**
2674 * netif_carrier_ok - test if carrier present
2675 * @dev: network device
2676 *
2677 * Check if carrier is present on device
2678 */
4d29515f 2679static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
2680{
2681 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2682}
2683
f629d208 2684unsigned long dev_trans_start(struct net_device *dev);
9d21493b 2685
f629d208 2686void __netdev_watchdog_up(struct net_device *dev);
1da177e4 2687
f629d208 2688void netif_carrier_on(struct net_device *dev);
1da177e4 2689
f629d208 2690void netif_carrier_off(struct net_device *dev);
1da177e4 2691
bea3348e
SH
2692/**
2693 * netif_dormant_on - mark device as dormant.
2694 * @dev: network device
2695 *
2696 * Mark device as dormant (as per RFC2863).
2697 *
2698 * The dormant state indicates that the relevant interface is not
2699 * actually in a condition to pass packets (i.e., it is not 'up') but is
2700 * in a "pending" state, waiting for some external event. For "on-
2701 * demand" interfaces, this new state identifies the situation where the
2702 * interface is waiting for events to place it in the up state.
2703 *
2704 */
b00055aa
SR
2705static inline void netif_dormant_on(struct net_device *dev)
2706{
2707 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2708 linkwatch_fire_event(dev);
2709}
2710
bea3348e
SH
2711/**
2712 * netif_dormant_off - set device as not dormant.
2713 * @dev: network device
2714 *
2715 * Device is not in dormant state.
2716 */
b00055aa
SR
2717static inline void netif_dormant_off(struct net_device *dev)
2718{
2719 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2720 linkwatch_fire_event(dev);
2721}
2722
bea3348e
SH
2723/**
2724 * netif_dormant - test if carrier present
2725 * @dev: network device
2726 *
2727 * Check if carrier is present on device
2728 */
4d29515f 2729static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
2730{
2731 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2732}
2733
2734
bea3348e
SH
2735/**
2736 * netif_oper_up - test if device is operational
2737 * @dev: network device
2738 *
2739 * Check if carrier is operational
2740 */
4d29515f 2741static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 2742{
b00055aa
SR
2743 return (dev->operstate == IF_OPER_UP ||
2744 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2745}
2746
bea3348e
SH
2747/**
2748 * netif_device_present - is device available or removed
2749 * @dev: network device
2750 *
2751 * Check if device has not been removed from system.
2752 */
4d29515f 2753static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
2754{
2755 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2756}
2757
f629d208 2758void netif_device_detach(struct net_device *dev);
1da177e4 2759
f629d208 2760void netif_device_attach(struct net_device *dev);
1da177e4
LT
2761
2762/*
2763 * Network interface message level settings
2764 */
1da177e4
LT
2765
2766enum {
2767 NETIF_MSG_DRV = 0x0001,
2768 NETIF_MSG_PROBE = 0x0002,
2769 NETIF_MSG_LINK = 0x0004,
2770 NETIF_MSG_TIMER = 0x0008,
2771 NETIF_MSG_IFDOWN = 0x0010,
2772 NETIF_MSG_IFUP = 0x0020,
2773 NETIF_MSG_RX_ERR = 0x0040,
2774 NETIF_MSG_TX_ERR = 0x0080,
2775 NETIF_MSG_TX_QUEUED = 0x0100,
2776 NETIF_MSG_INTR = 0x0200,
2777 NETIF_MSG_TX_DONE = 0x0400,
2778 NETIF_MSG_RX_STATUS = 0x0800,
2779 NETIF_MSG_PKTDATA = 0x1000,
2780 NETIF_MSG_HW = 0x2000,
2781 NETIF_MSG_WOL = 0x4000,
2782};
2783
2784#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2785#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2786#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2787#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2788#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2789#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2790#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2791#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2792#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2793#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2794#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2795#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2796#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2797#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2798#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2799
2800static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2801{
2802 /* use default */
2803 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2804 return default_msg_enable_bits;
2805 if (debug_value == 0) /* no output */
2806 return 0;
2807 /* set low N bits */
2808 return (1 << debug_value) - 1;
2809}
2810
c773e847 2811static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 2812{
c773e847
DM
2813 spin_lock(&txq->_xmit_lock);
2814 txq->xmit_lock_owner = cpu;
22dd7495
JHS
2815}
2816
fd2ea0a7
DM
2817static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2818{
2819 spin_lock_bh(&txq->_xmit_lock);
2820 txq->xmit_lock_owner = smp_processor_id();
2821}
2822
4d29515f 2823static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 2824{
4d29515f 2825 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
2826 if (likely(ok))
2827 txq->xmit_lock_owner = smp_processor_id();
2828 return ok;
2829}
2830
2831static inline void __netif_tx_unlock(struct netdev_queue *txq)
2832{
2833 txq->xmit_lock_owner = -1;
2834 spin_unlock(&txq->_xmit_lock);
2835}
2836
2837static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2838{
2839 txq->xmit_lock_owner = -1;
2840 spin_unlock_bh(&txq->_xmit_lock);
2841}
2842
08baf561
ED
2843static inline void txq_trans_update(struct netdev_queue *txq)
2844{
2845 if (txq->xmit_lock_owner != -1)
2846 txq->trans_start = jiffies;
2847}
2848
d29f749e
DJ
2849/**
2850 * netif_tx_lock - grab network device transmit lock
2851 * @dev: network device
d29f749e
DJ
2852 *
2853 * Get network device transmit lock
2854 */
22dd7495
JHS
2855static inline void netif_tx_lock(struct net_device *dev)
2856{
e8a0464c 2857 unsigned int i;
c3f26a26 2858 int cpu;
c773e847 2859
c3f26a26
DM
2860 spin_lock(&dev->tx_global_lock);
2861 cpu = smp_processor_id();
e8a0464c
DM
2862 for (i = 0; i < dev->num_tx_queues; i++) {
2863 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
2864
2865 /* We are the only thread of execution doing a
2866 * freeze, but we have to grab the _xmit_lock in
2867 * order to synchronize with threads which are in
2868 * the ->hard_start_xmit() handler and already
2869 * checked the frozen bit.
2870 */
e8a0464c 2871 __netif_tx_lock(txq, cpu);
c3f26a26
DM
2872 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2873 __netif_tx_unlock(txq);
e8a0464c 2874 }
932ff279
HX
2875}
2876
2877static inline void netif_tx_lock_bh(struct net_device *dev)
2878{
e8a0464c
DM
2879 local_bh_disable();
2880 netif_tx_lock(dev);
932ff279
HX
2881}
2882
932ff279
HX
2883static inline void netif_tx_unlock(struct net_device *dev)
2884{
e8a0464c
DM
2885 unsigned int i;
2886
2887 for (i = 0; i < dev->num_tx_queues; i++) {
2888 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 2889
c3f26a26
DM
2890 /* No need to grab the _xmit_lock here. If the
2891 * queue is not stopped for another reason, we
2892 * force a schedule.
2893 */
2894 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 2895 netif_schedule_queue(txq);
c3f26a26
DM
2896 }
2897 spin_unlock(&dev->tx_global_lock);
932ff279
HX
2898}
2899
2900static inline void netif_tx_unlock_bh(struct net_device *dev)
2901{
e8a0464c
DM
2902 netif_tx_unlock(dev);
2903 local_bh_enable();
932ff279
HX
2904}
2905
c773e847 2906#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 2907 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 2908 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
2909 } \
2910}
2911
c773e847 2912#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 2913 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 2914 __netif_tx_unlock(txq); \
22dd7495
JHS
2915 } \
2916}
2917
1da177e4
LT
2918static inline void netif_tx_disable(struct net_device *dev)
2919{
fd2ea0a7 2920 unsigned int i;
c3f26a26 2921 int cpu;
fd2ea0a7 2922
c3f26a26
DM
2923 local_bh_disable();
2924 cpu = smp_processor_id();
fd2ea0a7
DM
2925 for (i = 0; i < dev->num_tx_queues; i++) {
2926 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
2927
2928 __netif_tx_lock(txq, cpu);
fd2ea0a7 2929 netif_tx_stop_queue(txq);
c3f26a26 2930 __netif_tx_unlock(txq);
fd2ea0a7 2931 }
c3f26a26 2932 local_bh_enable();
1da177e4
LT
2933}
2934
e308a5d8
DM
2935static inline void netif_addr_lock(struct net_device *dev)
2936{
2937 spin_lock(&dev->addr_list_lock);
2938}
2939
2429f7ac
JP
2940static inline void netif_addr_lock_nested(struct net_device *dev)
2941{
2942 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2943}
2944
e308a5d8
DM
2945static inline void netif_addr_lock_bh(struct net_device *dev)
2946{
2947 spin_lock_bh(&dev->addr_list_lock);
2948}
2949
2950static inline void netif_addr_unlock(struct net_device *dev)
2951{
2952 spin_unlock(&dev->addr_list_lock);
2953}
2954
2955static inline void netif_addr_unlock_bh(struct net_device *dev)
2956{
2957 spin_unlock_bh(&dev->addr_list_lock);
2958}
2959
f001fde5 2960/*
31278e71 2961 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
2962 * rcu_read_lock held.
2963 */
2964#define for_each_dev_addr(dev, ha) \
31278e71 2965 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 2966
1da177e4
LT
2967/* These functions live elsewhere (drivers/net/net_init.c, but related) */
2968
f629d208 2969void ether_setup(struct net_device *dev);
1da177e4
LT
2970
2971/* Support for loadable net-drivers */
f629d208
JP
2972struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2973 void (*setup)(struct net_device *),
2974 unsigned int txqs, unsigned int rxqs);
f25f4e44 2975#define alloc_netdev(sizeof_priv, name, setup) \
36909ea4
TH
2976 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2977
2978#define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2979 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2980
f629d208
JP
2981int register_netdev(struct net_device *dev);
2982void unregister_netdev(struct net_device *dev);
f001fde5 2983
22bedad3 2984/* General hardware address lists handling functions */
f629d208
JP
2985int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2986 struct netdev_hw_addr_list *from_list, int addr_len);
2987void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2988 struct netdev_hw_addr_list *from_list, int addr_len);
f629d208 2989void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 2990
f001fde5 2991/* Functions used for device addresses handling */
f629d208
JP
2992int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2993 unsigned char addr_type);
2994int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2995 unsigned char addr_type);
f629d208
JP
2996void dev_addr_flush(struct net_device *dev);
2997int dev_addr_init(struct net_device *dev);
a748ee24
JP
2998
2999/* Functions used for unicast addresses handling */
f629d208
JP
3000int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3001int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3002int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3003int dev_uc_sync(struct net_device *to, struct net_device *from);
3004int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3005void dev_uc_unsync(struct net_device *to, struct net_device *from);
3006void dev_uc_flush(struct net_device *dev);
3007void dev_uc_init(struct net_device *dev);
f001fde5 3008
22bedad3 3009/* Functions used for multicast addresses handling */
f629d208
JP
3010int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3011int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3012int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3013int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3014int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3015int dev_mc_sync(struct net_device *to, struct net_device *from);
3016int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3017void dev_mc_unsync(struct net_device *to, struct net_device *from);
3018void dev_mc_flush(struct net_device *dev);
3019void dev_mc_init(struct net_device *dev);
f001fde5 3020
4417da66 3021/* Functions used for secondary unicast and multicast support */
f629d208
JP
3022void dev_set_rx_mode(struct net_device *dev);
3023void __dev_set_rx_mode(struct net_device *dev);
3024int dev_set_promiscuity(struct net_device *dev, int inc);
3025int dev_set_allmulti(struct net_device *dev, int inc);
3026void netdev_state_change(struct net_device *dev);
3027void netdev_notify_peers(struct net_device *dev);
3028void netdev_features_change(struct net_device *dev);
1da177e4 3029/* Load a device via the kmod */
f629d208
JP
3030void dev_load(struct net *net, const char *name);
3031struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3032 struct rtnl_link_stats64 *storage);
3033void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3034 const struct net_device_stats *netdev_stats);
eeda3fd6 3035
1da177e4 3036extern int netdev_max_backlog;
3b098e2d 3037extern int netdev_tstamp_prequeue;
1da177e4 3038extern int weight_p;
0a14842f 3039extern int bpf_jit_enable;
9ff162a8 3040
f629d208 3041bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
f629d208
JP
3042struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3043 struct list_head **iter);
8b5be856
VF
3044
3045/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3046#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3047 for (iter = &(dev)->all_adj_list.upper, \
3048 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3049 updev; \
3050 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3051
f629d208
JP
3052void *netdev_lower_get_next_private(struct net_device *dev,
3053 struct list_head **iter);
3054void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3055 struct list_head **iter);
31088a11
VF
3056
3057#define netdev_for_each_lower_private(dev, priv, iter) \
3058 for (iter = (dev)->adj_list.lower.next, \
3059 priv = netdev_lower_get_next_private(dev, &(iter)); \
3060 priv; \
3061 priv = netdev_lower_get_next_private(dev, &(iter)))
3062
3063#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3064 for (iter = &(dev)->adj_list.lower, \
3065 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3066 priv; \
3067 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3068
f629d208 3069void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3070void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3071struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3072struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3073int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3074int netdev_master_upper_dev_link(struct net_device *dev,
9ff162a8 3075 struct net_device *upper_dev);
f629d208
JP
3076int netdev_master_upper_dev_link_private(struct net_device *dev,
3077 struct net_device *upper_dev,
3078 void *private);
3079void netdev_upper_dev_unlink(struct net_device *dev,
3080 struct net_device *upper_dev);
5bb025fa 3081void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3082void *netdev_lower_dev_get_private(struct net_device *dev,
3083 struct net_device *lower_dev);
3084int skb_checksum_help(struct sk_buff *skb);
3085struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3086 netdev_features_t features, bool tx_path);
3087struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3088 netdev_features_t features);
12b0004d
CW
3089
3090static inline
3091struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3092{
3093 return __skb_gso_segment(skb, features, true);
3094}
ec5f0615
PS
3095__be16 skb_network_protocol(struct sk_buff *skb);
3096
3097static inline bool can_checksum_protocol(netdev_features_t features,
3098 __be16 protocol)
3099{
3100 return ((features & NETIF_F_GEN_CSUM) ||
3101 ((features & NETIF_F_V4_CSUM) &&
3102 protocol == htons(ETH_P_IP)) ||
3103 ((features & NETIF_F_V6_CSUM) &&
3104 protocol == htons(ETH_P_IPV6)) ||
3105 ((features & NETIF_F_FCOE_CRC) &&
3106 protocol == htons(ETH_P_FCOE)));
3107}
12b0004d 3108
fb286bb2 3109#ifdef CONFIG_BUG
f629d208 3110void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3111#else
3112static inline void netdev_rx_csum_fault(struct net_device *dev)
3113{
3114}
3115#endif
1da177e4 3116/* rx skb timestamps */
f629d208
JP
3117void net_enable_timestamp(void);
3118void net_disable_timestamp(void);
1da177e4 3119
20380731 3120#ifdef CONFIG_PROC_FS
f629d208 3121int __init dev_proc_init(void);
900ff8c6
CW
3122#else
3123#define dev_proc_init() 0
20380731
ACM
3124#endif
3125
42a2d923
LT
3126int netdev_class_create_file_ns(struct class_attribute *class_attr,
3127 const void *ns);
3128void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3129 const void *ns);
58292cbe
TH
3130
3131static inline int netdev_class_create_file(struct class_attribute *class_attr)
3132{
3133 return netdev_class_create_file_ns(class_attr, NULL);
3134}
3135
3136static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3137{
3138 netdev_class_remove_file_ns(class_attr, NULL);
3139}
b8a9787e 3140
04600794
JB
3141extern struct kobj_ns_type_operations net_ns_type_operations;
3142
f629d208 3143const char *netdev_drivername(const struct net_device *dev);
6579e57b 3144
f629d208 3145void linkwatch_run_queue(void);
20380731 3146
c8f44aff
MM
3147static inline netdev_features_t netdev_get_wanted_features(
3148 struct net_device *dev)
5455c699
MM
3149{
3150 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3151}
c8f44aff
MM
3152netdev_features_t netdev_increment_features(netdev_features_t all,
3153 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3154
3155/* Allow TSO being used on stacked device :
3156 * Performing the GSO segmentation before last device
3157 * is a performance improvement.
3158 */
3159static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3160 netdev_features_t mask)
3161{
3162 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3163}
3164
6cb6a27c 3165int __netdev_update_features(struct net_device *dev);
5455c699 3166void netdev_update_features(struct net_device *dev);
afe12cc8 3167void netdev_change_features(struct net_device *dev);
7f353bf2 3168
fc4a7489
PM
3169void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3170 struct net_device *dev);
3171
d2069403
FW
3172netdev_features_t netif_skb_dev_features(struct sk_buff *skb,
3173 const struct net_device *dev);
3174static inline netdev_features_t netif_skb_features(struct sk_buff *skb)
3175{
3176 return netif_skb_dev_features(skb, skb->dev);
3177}
58e998c6 3178
4d29515f 3179static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3180{
c8f44aff 3181 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3182
3183 /* check flags correspondence */
3184 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3185 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3186 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3187 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3188 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3189 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3190
d6b4991a 3191 return (features & feature) == feature;
576a30eb
HX
3192}
3193
4d29515f 3194static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3195{
278b2513 3196 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3197 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3198}
3199
4d29515f
DM
3200static inline bool netif_needs_gso(struct sk_buff *skb,
3201 netdev_features_t features)
7967168c 3202{
fc741216 3203 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3204 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3205 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3206}
3207
82cc1a7a
PWJ
3208static inline void netif_set_gso_max_size(struct net_device *dev,
3209 unsigned int size)
3210{
3211 dev->gso_max_size = size;
3212}
3213
7a7ffbab
WCC
3214static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3215 int pulled_hlen, u16 mac_offset,
3216 int mac_len)
3217{
3218 skb->protocol = protocol;
3219 skb->encapsulation = 1;
3220 skb_push(skb, pulled_hlen);
3221 skb_reset_transport_header(skb);
3222 skb->mac_header = mac_offset;
3223 skb->network_header = skb->mac_header + mac_len;
3224 skb->mac_len = mac_len;
3225}
3226
a6cc0cfa
JF
3227static inline bool netif_is_macvlan(struct net_device *dev)
3228{
3229 return dev->priv_flags & IFF_MACVLAN;
3230}
3231
8a7fbfab 3232static inline bool netif_is_bond_master(struct net_device *dev)
3233{
3234 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3235}
3236
4d29515f 3237static inline bool netif_is_bond_slave(struct net_device *dev)
1765a575
JP
3238{
3239 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3240}
3241
3bdc0eba
BG
3242static inline bool netif_supports_nofcs(struct net_device *dev)
3243{
3244 return dev->priv_flags & IFF_SUPP_NOFCS;
3245}
3246
505d4f73 3247extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 3248
571ba423
JP
3249/* Logging, debugging and troubleshooting/diagnostic helpers. */
3250
3251/* netdev_printk helpers, similar to dev_printk */
3252
3253static inline const char *netdev_name(const struct net_device *dev)
3254{
3255 if (dev->reg_state != NETREG_REGISTERED)
3256 return "(unregistered net_device)";
3257 return dev->name;
3258}
3259
f629d208 3260__printf(3, 4)
b9075fa9
JP
3261int netdev_printk(const char *level, const struct net_device *dev,
3262 const char *format, ...);
f629d208 3263__printf(2, 3)
b9075fa9 3264int netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 3265__printf(2, 3)
b9075fa9 3266int netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 3267__printf(2, 3)
b9075fa9 3268int netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 3269__printf(2, 3)
b9075fa9 3270int netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 3271__printf(2, 3)
b9075fa9 3272int netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 3273__printf(2, 3)
b9075fa9 3274int netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 3275__printf(2, 3)
b9075fa9 3276int netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 3277
8909c9ad
VK
3278#define MODULE_ALIAS_NETDEV(device) \
3279 MODULE_ALIAS("netdev-" device)
3280
b558c96f 3281#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
3282#define netdev_dbg(__dev, format, args...) \
3283do { \
ffa10cb4 3284 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 3285} while (0)
b558c96f
JC
3286#elif defined(DEBUG)
3287#define netdev_dbg(__dev, format, args...) \
3288 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
3289#else
3290#define netdev_dbg(__dev, format, args...) \
3291({ \
3292 if (0) \
3293 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3294 0; \
3295})
3296#endif
3297
3298#if defined(VERBOSE_DEBUG)
3299#define netdev_vdbg netdev_dbg
3300#else
3301
3302#define netdev_vdbg(dev, format, args...) \
3303({ \
3304 if (0) \
3305 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3306 0; \
3307})
3308#endif
3309
3310/*
3311 * netdev_WARN() acts like dev_printk(), but with the key difference
3312 * of using a WARN/WARN_ON to get the message out, including the
3313 * file/line information and a backtrace.
3314 */
3315#define netdev_WARN(dev, format, args...) \
7cc7c5e5 3316 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
571ba423 3317
b3d95c5c
JP
3318/* netif printk helpers, similar to netdev_printk */
3319
3320#define netif_printk(priv, type, level, dev, fmt, args...) \
3321do { \
3322 if (netif_msg_##type(priv)) \
3323 netdev_printk(level, (dev), fmt, ##args); \
3324} while (0)
3325
f45f4321
JP
3326#define netif_level(level, priv, type, dev, fmt, args...) \
3327do { \
3328 if (netif_msg_##type(priv)) \
3329 netdev_##level(dev, fmt, ##args); \
3330} while (0)
3331
b3d95c5c 3332#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 3333 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 3334#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 3335 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 3336#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 3337 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 3338#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 3339 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 3340#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 3341 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 3342#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 3343 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 3344#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 3345 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 3346
0053ea9c 3347#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
3348#define netif_dbg(priv, type, netdev, format, args...) \
3349do { \
3350 if (netif_msg_##type(priv)) \
b5fb0a03 3351 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 3352} while (0)
0053ea9c
JP
3353#elif defined(DEBUG)
3354#define netif_dbg(priv, type, dev, format, args...) \
3355 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
3356#else
3357#define netif_dbg(priv, type, dev, format, args...) \
3358({ \
3359 if (0) \
3360 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3361 0; \
3362})
3363#endif
3364
3365#if defined(VERBOSE_DEBUG)
bcfcc450 3366#define netif_vdbg netif_dbg
b3d95c5c
JP
3367#else
3368#define netif_vdbg(priv, type, dev, format, args...) \
3369({ \
3370 if (0) \
a4ed89cb 3371 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
3372 0; \
3373})
3374#endif
571ba423 3375
900ff8c6
CW
3376/*
3377 * The list of packet types we will receive (as opposed to discard)
3378 * and the routines to invoke.
3379 *
3380 * Why 16. Because with 16 the only overlap we get on a hash of the
3381 * low nibble of the protocol value is RARP/SNAP/X.25.
3382 *
3383 * NOTE: That is no longer true with the addition of VLAN tags. Not
3384 * sure which should go first, but I bet it won't make much
3385 * difference if we are running VLANs. The good news is that
3386 * this protocol won't be in the list unless compiled in, so
3387 * the average user (w/out VLANs) will not be adversely affected.
3388 * --BLG
3389 *
3390 * 0800 IP
3391 * 8100 802.1Q VLAN
3392 * 0001 802.3
3393 * 0002 AX.25
3394 * 0004 802.2
3395 * 8035 RARP
3396 * 0005 SNAP
3397 * 0805 X.25
3398 * 0806 ARP
3399 * 8137 IPX
3400 * 0009 Localtalk
3401 * 86DD IPv6
3402 */
3403#define PTYPE_HASH_SIZE (16)
3404#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3405
385a154c 3406#endif /* _LINUX_NETDEVICE_H */