]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
vxlan: Have the NIC drivers do less work for offloads
[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{
a0a46196 486 set_bit(NAPI_STATE_DISABLE, &n->state);
bea3348e 487 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
43cc7380 488 msleep(1);
a0a46196 489 clear_bit(NAPI_STATE_DISABLE, &n->state);
bea3348e
SH
490}
491
492/**
493 * napi_enable - enable NAPI scheduling
494 * @n: napi context
495 *
496 * Resume NAPI from being scheduled on this context.
497 * Must be paired with napi_disable.
498 */
499static inline void napi_enable(struct napi_struct *n)
500{
501 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
502 smp_mb__before_clear_bit();
503 clear_bit(NAPI_STATE_SCHED, &n->state);
504}
505
c264c3de
SH
506#ifdef CONFIG_SMP
507/**
508 * napi_synchronize - wait until NAPI is not running
509 * @n: napi context
510 *
511 * Wait until NAPI is done being scheduled on this context.
512 * Waits till any outstanding processing completes but
513 * does not disable future activations.
514 */
515static inline void napi_synchronize(const struct napi_struct *n)
516{
517 while (test_bit(NAPI_STATE_SCHED, &n->state))
518 msleep(1);
519}
520#else
521# define napi_synchronize(n) barrier()
522#endif
523
d94d9fee 524enum netdev_queue_state_t {
73466498
TH
525 __QUEUE_STATE_DRV_XOFF,
526 __QUEUE_STATE_STACK_XOFF,
c3f26a26 527 __QUEUE_STATE_FROZEN,
73466498
TH
528#define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
529 (1 << __QUEUE_STATE_STACK_XOFF))
530#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
531 (1 << __QUEUE_STATE_FROZEN))
79d16385 532};
73466498
TH
533/*
534 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
535 * netif_tx_* functions below are used to manipulate this flag. The
536 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
537 * queue independently. The netif_xmit_*stopped functions below are called
538 * to check if the queue has been stopped by the driver or stack (either
539 * of the XOFF bits are set in the state). Drivers should not need to call
540 * netif_xmit*stopped functions, they should only be using netif_tx_*.
541 */
79d16385 542
bb949fbd 543struct netdev_queue {
6a321cb3
ED
544/*
545 * read mostly part
546 */
bb949fbd 547 struct net_device *dev;
b0e1e646
DM
548 struct Qdisc *qdisc;
549 struct Qdisc *qdisc_sleeping;
ccf5ff69 550#ifdef CONFIG_SYSFS
1d24eb48
TH
551 struct kobject kobj;
552#endif
f2cd2d3e
ED
553#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
554 int numa_node;
555#endif
6a321cb3
ED
556/*
557 * write mostly part
558 */
559 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
560 int xmit_lock_owner;
9d21493b
ED
561 /*
562 * please use this field instead of dev->trans_start
563 */
564 unsigned long trans_start;
ccf5ff69 565
566 /*
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
569 */
570 unsigned long trans_timeout;
114cf580
TH
571
572 unsigned long state;
573
574#ifdef CONFIG_BQL
575 struct dql dql;
576#endif
e8a0464c 577} ____cacheline_aligned_in_smp;
bb949fbd 578
f2cd2d3e
ED
579static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
580{
581#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
582 return q->numa_node;
583#else
b236da69 584 return NUMA_NO_NODE;
f2cd2d3e
ED
585#endif
586}
587
588static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
589{
590#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
591 q->numa_node = node;
592#endif
593}
594
df334545 595#ifdef CONFIG_RPS
0a9627f2
TH
596/*
597 * This structure holds an RPS map which can be of variable length. The
598 * map is an array of CPUs.
599 */
600struct rps_map {
601 unsigned int len;
602 struct rcu_head rcu;
603 u16 cpus[0];
604};
60b778ce 605#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 606
fec5e652 607/*
c445477d
BH
608 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
609 * tail pointer for that CPU's input queue at the time of last enqueue, and
610 * a hardware filter index.
fec5e652
TH
611 */
612struct rps_dev_flow {
613 u16 cpu;
c445477d 614 u16 filter;
fec5e652
TH
615 unsigned int last_qtail;
616};
c445477d 617#define RPS_NO_FILTER 0xffff
fec5e652
TH
618
619/*
620 * The rps_dev_flow_table structure contains a table of flow mappings.
621 */
622struct rps_dev_flow_table {
623 unsigned int mask;
624 struct rcu_head rcu;
fec5e652
TH
625 struct rps_dev_flow flows[0];
626};
627#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 628 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
629
630/*
631 * The rps_sock_flow_table contains mappings of flows to the last CPU
632 * on which they were processed by the application (set in recvmsg).
633 */
634struct rps_sock_flow_table {
635 unsigned int mask;
636 u16 ents[0];
637};
638#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
60b778ce 639 ((_num) * sizeof(u16)))
fec5e652
TH
640
641#define RPS_NO_CPU 0xffff
642
643static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
644 u32 hash)
645{
646 if (table && hash) {
647 unsigned int cpu, index = hash & table->mask;
648
649 /* We only give a hint, preemption can change cpu under us */
650 cpu = raw_smp_processor_id();
651
652 if (table->ents[index] != cpu)
653 table->ents[index] = cpu;
654 }
655}
656
657static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
658 u32 hash)
659{
660 if (table && hash)
661 table->ents[hash & table->mask] = RPS_NO_CPU;
662}
663
6e3f7faf 664extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
fec5e652 665
c445477d 666#ifdef CONFIG_RFS_ACCEL
f629d208
JP
667bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
668 u16 filter_id);
c445477d
BH
669#endif
670
0a9627f2
TH
671/* This structure contains an instance of an RX queue. */
672struct netdev_rx_queue {
6e3f7faf
ED
673 struct rps_map __rcu *rps_map;
674 struct rps_dev_flow_table __rcu *rps_flow_table;
675 struct kobject kobj;
fe822240 676 struct net_device *dev;
0a9627f2 677} ____cacheline_aligned_in_smp;
fec5e652 678#endif /* CONFIG_RPS */
d314774c 679
bf264145
TH
680#ifdef CONFIG_XPS
681/*
682 * This structure holds an XPS map which can be of variable length. The
683 * map is an array of queues.
684 */
685struct xps_map {
686 unsigned int len;
687 unsigned int alloc_len;
688 struct rcu_head rcu;
689 u16 queues[0];
690};
60b778ce 691#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
bf264145
TH
692#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
693 / sizeof(u16))
694
695/*
696 * This structure holds all XPS maps for device. Maps are indexed by CPU.
697 */
698struct xps_dev_maps {
699 struct rcu_head rcu;
a4177869 700 struct xps_map __rcu *cpu_map[0];
bf264145
TH
701};
702#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
703 (nr_cpu_ids * sizeof(struct xps_map *)))
704#endif /* CONFIG_XPS */
705
4f57c087
JF
706#define TC_MAX_QUEUE 16
707#define TC_BITMASK 15
708/* HW offloaded queuing disciplines txq count and offset maps */
709struct netdev_tc_txq {
710 u16 count;
711 u16 offset;
712};
713
68bad94e
NP
714#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
715/*
716 * This structure is to hold information about the device
717 * configured to run FCoE protocol stack.
718 */
719struct netdev_fcoe_hbainfo {
720 char manufacturer[64];
721 char serial_number[64];
722 char hardware_version[64];
723 char driver_version[64];
724 char optionrom_version[64];
725 char firmware_version[64];
726 char model[256];
727 char model_description[256];
728};
729#endif
730
66b52b0d
JP
731#define MAX_PHYS_PORT_ID_LEN 32
732
733/* This structure holds a unique identifier to identify the
734 * physical port used by a netdevice.
735 */
736struct netdev_phys_port_id {
737 unsigned char id[MAX_PHYS_PORT_ID_LEN];
738 unsigned char id_len;
739};
740
d314774c
SH
741/*
742 * This structure defines the management hooks for network devices.
00829823
SH
743 * The following hooks can be defined; unless noted otherwise, they are
744 * optional and can be filled with a null pointer.
d314774c
SH
745 *
746 * int (*ndo_init)(struct net_device *dev);
747 * This function is called once when network device is registered.
748 * The network device can use this to any late stage initializaton
749 * or semantic validattion. It can fail with an error code which will
750 * be propogated back to register_netdev
751 *
752 * void (*ndo_uninit)(struct net_device *dev);
753 * This function is called when device is unregistered or when registration
754 * fails. It is not called if init fails.
755 *
756 * int (*ndo_open)(struct net_device *dev);
757 * This function is called when network device transistions to the up
758 * state.
759 *
760 * int (*ndo_stop)(struct net_device *dev);
761 * This function is called when network device transistions to the down
762 * state.
763 *
dc1f8bf6
SH
764 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
765 * struct net_device *dev);
00829823 766 * Called when a packet needs to be transmitted.
dc1f8bf6
SH
767 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
768 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
769 * Required can not be NULL.
770 *
771 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
772 * Called to decide which queue to when device supports multiple
773 * transmit queues.
774 *
d314774c
SH
775 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
776 * This function is called to allow device receiver to make
777 * changes to configuration when multicast or promiscious is enabled.
778 *
779 * void (*ndo_set_rx_mode)(struct net_device *dev);
780 * This function is called device changes address list filtering.
01789349
JP
781 * If driver handles unicast address filtering, it should set
782 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
783 *
784 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
785 * This function is called when the Media Access Control address
37b607c5 786 * needs to be changed. If this interface is not defined, the
d314774c
SH
787 * mac address can not be changed.
788 *
789 * int (*ndo_validate_addr)(struct net_device *dev);
790 * Test if Media Access Control address is valid for the device.
791 *
792 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
793 * Called when a user request an ioctl which can't be handled by
794 * the generic interface code. If not defined ioctl's return
795 * not supported error code.
796 *
797 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
798 * Used to set network devices bus interface parameters. This interface
799 * is retained for legacy reason, new devices should use the bus
800 * interface (PCI) for low level management.
801 *
802 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
803 * Called when a user wants to change the Maximum Transfer Unit
804 * of a device. If not defined, any request to change MTU will
805 * will return an error.
806 *
00829823 807 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
808 * Callback uses when the transmitter has not made any progress
809 * for dev->watchdog ticks.
810 *
3cfde79c 811 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 812 * struct rtnl_link_stats64 *storage);
d308e38f 813 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 814 * Called when a user wants to get the network device usage
be1f3c2c 815 * statistics. Drivers must do one of the following:
3cfde79c
BH
816 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
817 * rtnl_link_stats64 structure passed by the caller.
82695d9b 818 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
819 * (which should normally be dev->stats) and return a pointer to
820 * it. The structure may be changed asynchronously only if each
821 * field is written atomically.
822 * 3. Update dev->stats asynchronously and atomically, and define
823 * neither operation.
d314774c 824 *
80d5c368
PM
825 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
826 * If device support VLAN filtering this function is called when a
827 * VLAN id is registered.
d314774c 828 *
8e586137 829 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
80d5c368
PM
830 * If device support VLAN filtering this function is called when a
831 * VLAN id is unregistered.
d314774c
SH
832 *
833 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
834 *
835 * SR-IOV management functions.
836 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
837 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
838 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
5f8444a3 839 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
840 * int (*ndo_get_vf_config)(struct net_device *dev,
841 * int vf, struct ifla_vf_info *ivf);
1d8faf48 842 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
843 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
844 * struct nlattr *port[]);
845 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
846 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
847 * Called to setup 'tc' number of traffic classes in the net device. This
848 * is always called from the stack with the rtnl lock held and netif tx
849 * queues stopped. This allows the netdevice to perform queue management
850 * safely.
c445477d 851 *
e9bce845
YZ
852 * Fiber Channel over Ethernet (FCoE) offload functions.
853 * int (*ndo_fcoe_enable)(struct net_device *dev);
854 * Called when the FCoE protocol stack wants to start using LLD for FCoE
855 * so the underlying device can perform whatever needed configuration or
856 * initialization to support acceleration of FCoE traffic.
857 *
858 * int (*ndo_fcoe_disable)(struct net_device *dev);
859 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
860 * so the underlying device can perform whatever needed clean-ups to
861 * stop supporting acceleration of FCoE traffic.
862 *
863 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
864 * struct scatterlist *sgl, unsigned int sgc);
865 * Called when the FCoE Initiator wants to initialize an I/O that
866 * is a possible candidate for Direct Data Placement (DDP). The LLD can
867 * perform necessary setup and returns 1 to indicate the device is set up
868 * successfully to perform DDP on this I/O, otherwise this returns 0.
869 *
870 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
871 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
872 * indicated by the FC exchange id 'xid', so the underlying device can
873 * clean up and reuse resources for later DDP requests.
874 *
875 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
876 * struct scatterlist *sgl, unsigned int sgc);
877 * Called when the FCoE Target wants to initialize an I/O that
878 * is a possible candidate for Direct Data Placement (DDP). The LLD can
879 * perform necessary setup and returns 1 to indicate the device is set up
880 * successfully to perform DDP on this I/O, otherwise this returns 0.
881 *
68bad94e
NP
882 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
883 * struct netdev_fcoe_hbainfo *hbainfo);
884 * Called when the FCoE Protocol stack wants information on the underlying
885 * device. This information is utilized by the FCoE protocol stack to
886 * register attributes with Fiber Channel management service as per the
887 * FC-GS Fabric Device Management Information(FDMI) specification.
888 *
e9bce845
YZ
889 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
890 * Called when the underlying device wants to override default World Wide
891 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
892 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
893 * protocol stack to use.
894 *
c445477d
BH
895 * RFS acceleration.
896 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
897 * u16 rxq_index, u32 flow_id);
898 * Set hardware filter for RFS. rxq_index is the target queue index;
899 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
900 * Return the filter ID on success, or a negative error code.
fbaec0ea 901 *
8b98a70c 902 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
903 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
904 * Called to make another netdev an underling.
905 *
906 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
907 * Called to release previously enslaved netdev.
5455c699
MM
908 *
909 * Feature/offload setting functions.
c8f44aff
MM
910 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
911 * netdev_features_t features);
5455c699
MM
912 * Adjusts the requested feature flags according to device-specific
913 * constraints, and returns the resulting flags. Must not modify
914 * the device state.
915 *
c8f44aff 916 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
917 * Called to update device configuration to new features. Passed
918 * feature set might be less than what was returned by ndo_fix_features()).
919 * Must return >0 or -errno if it changed dev->features itself.
920 *
edc7d573 921 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
922 * struct net_device *dev,
6b6e2725 923 * const unsigned char *addr, u16 flags)
77162022 924 * Adds an FDB entry to dev for addr.
1690be63
VY
925 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
926 * struct net_device *dev,
6b6e2725 927 * const unsigned char *addr)
77162022
JF
928 * Deletes the FDB entry from dev coresponding to addr.
929 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
930 * struct net_device *dev, int idx)
931 * Used to add FDB entries to dump requests. Implementers should add
932 * entries to skb and update idx with the number of entries.
e5a55a89
JF
933 *
934 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
935 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
24f11a5c 936 * struct net_device *dev, u32 filter_mask)
4bf84c35
JP
937 *
938 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
939 * Called to change device carrier. Soft-devices (like dummy, team, etc)
940 * which do not represent real hardware may define this to allow their
941 * userspace components to manage their virtual carrier state. Devices
942 * that determine carrier state from physical hardware properties (eg
943 * network cables) or protocol-dependent mechanisms (eg
944 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
945 *
946 * int (*ndo_get_phys_port_id)(struct net_device *dev,
947 * struct netdev_phys_port_id *ppid);
948 * Called to get ID of physical port of this device. If driver does
949 * not implement this, it is assumed that the hw is not able to have
950 * multiple net devices on single physical port.
53cf5275
JG
951 *
952 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 953 * sa_family_t sa_family, __be16 port);
53cf5275
JG
954 * Called by vxlan to notiy a driver about the UDP port and socket
955 * address family that vxlan is listnening to. It is called only when
956 * a new port starts listening. The operation is protected by the
957 * vxlan_net->sock_lock.
958 *
959 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 960 * sa_family_t sa_family, __be16 port);
53cf5275
JG
961 * Called by vxlan to notify the driver about a UDP port and socket
962 * address family that vxlan is not listening to anymore. The operation
963 * is protected by the vxlan_net->sock_lock.
d314774c
SH
964 */
965struct net_device_ops {
966 int (*ndo_init)(struct net_device *dev);
967 void (*ndo_uninit)(struct net_device *dev);
968 int (*ndo_open)(struct net_device *dev);
969 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 970 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
971 struct net_device *dev);
972 u16 (*ndo_select_queue)(struct net_device *dev,
973 struct sk_buff *skb);
d314774c
SH
974 void (*ndo_change_rx_flags)(struct net_device *dev,
975 int flags);
d314774c 976 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
977 int (*ndo_set_mac_address)(struct net_device *dev,
978 void *addr);
d314774c 979 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
980 int (*ndo_do_ioctl)(struct net_device *dev,
981 struct ifreq *ifr, int cmd);
d314774c
SH
982 int (*ndo_set_config)(struct net_device *dev,
983 struct ifmap *map);
00829823
SH
984 int (*ndo_change_mtu)(struct net_device *dev,
985 int new_mtu);
986 int (*ndo_neigh_setup)(struct net_device *dev,
987 struct neigh_parms *);
d314774c
SH
988 void (*ndo_tx_timeout) (struct net_device *dev);
989
28172739
ED
990 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
991 struct rtnl_link_stats64 *storage);
d314774c
SH
992 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
993
8e586137 994 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 995 __be16 proto, u16 vid);
8e586137 996 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 997 __be16 proto, u16 vid);
d314774c 998#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 999 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1000 int (*ndo_netpoll_setup)(struct net_device *dev,
47be03a2
AW
1001 struct netpoll_info *info,
1002 gfp_t gfp);
0e34e931 1003 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1004#endif
e0d1095a 1005#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1006 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1007#endif
95c26df8
WM
1008 int (*ndo_set_vf_mac)(struct net_device *dev,
1009 int queue, u8 *mac);
1010 int (*ndo_set_vf_vlan)(struct net_device *dev,
1011 int queue, u16 vlan, u8 qos);
1012 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
1013 int vf, int rate);
5f8444a3
GR
1014 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1015 int vf, bool setting);
95c26df8
WM
1016 int (*ndo_get_vf_config)(struct net_device *dev,
1017 int vf,
1018 struct ifla_vf_info *ivf);
1d8faf48
RE
1019 int (*ndo_set_vf_link_state)(struct net_device *dev,
1020 int vf, int link_state);
57b61080
SF
1021 int (*ndo_set_vf_port)(struct net_device *dev,
1022 int vf,
1023 struct nlattr *port[]);
1024 int (*ndo_get_vf_port)(struct net_device *dev,
1025 int vf, struct sk_buff *skb);
4f57c087 1026 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
d11ead75 1027#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1028 int (*ndo_fcoe_enable)(struct net_device *dev);
1029 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1030 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1031 u16 xid,
1032 struct scatterlist *sgl,
1033 unsigned int sgc);
1034 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1035 u16 xid);
6247e086
YZ
1036 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1037 u16 xid,
1038 struct scatterlist *sgl,
1039 unsigned int sgc);
68bad94e
NP
1040 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1041 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1042#endif
1043
d11ead75 1044#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1045#define NETDEV_FCOE_WWNN 0
1046#define NETDEV_FCOE_WWPN 1
1047 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1048 u64 *wwn, int type);
4d288d57 1049#endif
3c9c36bc 1050
c445477d
BH
1051#ifdef CONFIG_RFS_ACCEL
1052 int (*ndo_rx_flow_steer)(struct net_device *dev,
1053 const struct sk_buff *skb,
1054 u16 rxq_index,
1055 u32 flow_id);
1056#endif
fbaec0ea
JP
1057 int (*ndo_add_slave)(struct net_device *dev,
1058 struct net_device *slave_dev);
1059 int (*ndo_del_slave)(struct net_device *dev,
1060 struct net_device *slave_dev);
c8f44aff
MM
1061 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1062 netdev_features_t features);
5455c699 1063 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1064 netdev_features_t features);
da6a8fa0 1065 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1066 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1067
1068 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1069 struct nlattr *tb[],
77162022 1070 struct net_device *dev,
6b6e2725 1071 const unsigned char *addr,
77162022
JF
1072 u16 flags);
1073 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1074 struct nlattr *tb[],
77162022 1075 struct net_device *dev,
6b6e2725 1076 const unsigned char *addr);
77162022
JF
1077 int (*ndo_fdb_dump)(struct sk_buff *skb,
1078 struct netlink_callback *cb,
1079 struct net_device *dev,
1080 int idx);
e5a55a89
JF
1081
1082 int (*ndo_bridge_setlink)(struct net_device *dev,
1083 struct nlmsghdr *nlh);
1084 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1085 u32 pid, u32 seq,
6cbdceeb
VY
1086 struct net_device *dev,
1087 u32 filter_mask);
407af329
VY
1088 int (*ndo_bridge_dellink)(struct net_device *dev,
1089 struct nlmsghdr *nlh);
4bf84c35
JP
1090 int (*ndo_change_carrier)(struct net_device *dev,
1091 bool new_carrier);
66b52b0d
JP
1092 int (*ndo_get_phys_port_id)(struct net_device *dev,
1093 struct netdev_phys_port_id *ppid);
53cf5275
JG
1094 void (*ndo_add_vxlan_port)(struct net_device *dev,
1095 sa_family_t sa_family,
35e42379 1096 __be16 port);
53cf5275
JG
1097 void (*ndo_del_vxlan_port)(struct net_device *dev,
1098 sa_family_t sa_family,
35e42379 1099 __be16 port);
d314774c
SH
1100};
1101
1da177e4
LT
1102/*
1103 * The DEVICE structure.
1104 * Actually, this whole structure is a big mistake. It mixes I/O
1105 * data with strictly "high-level" data, and it has to know about
1106 * almost every data structure used in the INET module.
1107 *
1108 * FIXME: cleanup struct net_device such that network protocol info
1109 * moves out.
1110 */
1111
d94d9fee 1112struct net_device {
1da177e4
LT
1113
1114 /*
1115 * This is the first field of the "visible" part of this structure
1116 * (i.e. as seen by users in the "Space.c" file). It is the name
724df615 1117 * of the interface.
1da177e4
LT
1118 */
1119 char name[IFNAMSIZ];
ed77134b 1120
9136461a 1121 /* device name hash chain, please keep it close to name[] */
9356b8fc 1122 struct hlist_node name_hlist;
9136461a 1123
0b815a1a
SH
1124 /* snmp alias */
1125 char *ifalias;
1da177e4
LT
1126
1127 /*
1128 * I/O specific fields
1129 * FIXME: Merge these and struct ifmap into one
1130 */
1131 unsigned long mem_end; /* shared mem end */
1132 unsigned long mem_start; /* shared mem start */
1133 unsigned long base_addr; /* device I/O address */
1134 unsigned int irq; /* device IRQ number */
1135
1136 /*
1137 * Some hardware also needs these fields, but they are not
1138 * part of the usual set specified in Space.c.
1139 */
1140
1da177e4
LT
1141 unsigned long state;
1142
7562f876 1143 struct list_head dev_list;
bea3348e 1144 struct list_head napi_list;
44a0873d 1145 struct list_head unreg_list;
5cde2829 1146 struct list_head close_list;
2f268f12
VF
1147
1148 /* directly linked devices, like slaves for bonding */
1149 struct {
1150 struct list_head upper;
1151 struct list_head lower;
1152 } adj_list;
1153
1154 /* all linked devices, *including* neighbours */
1155 struct {
1156 struct list_head upper;
1157 struct list_head lower;
1158 } all_adj_list;
4c3d5e7b 1159
1da177e4 1160
5455c699 1161 /* currently active device features */
c8f44aff 1162 netdev_features_t features;
5455c699 1163 /* user-changeable features */
c8f44aff 1164 netdev_features_t hw_features;
5455c699 1165 /* user-requested features */
c8f44aff 1166 netdev_features_t wanted_features;
1aac6267 1167 /* mask of features inheritable by VLAN devices */
c8f44aff 1168 netdev_features_t vlan_features;
6a674e9c
JG
1169 /* mask of features inherited by encapsulating devices
1170 * This field indicates what encapsulation offloads
1171 * the hardware is capable of doing, and drivers will
1172 * need to set them appropriately.
1173 */
1174 netdev_features_t hw_enc_features;
0d89d203
SH
1175 /* mask of fetures inheritable by MPLS */
1176 netdev_features_t mpls_features;
04ed3e74 1177
1da177e4
LT
1178 /* Interface index. Unique device identifier */
1179 int ifindex;
1180 int iflink;
1181
c45d286e 1182 struct net_device_stats stats;
caf586e5
ED
1183 atomic_long_t rx_dropped; /* dropped packets by core network
1184 * Do not use this in drivers.
1185 */
1da177e4 1186
b86e0280 1187#ifdef CONFIG_WIRELESS_EXT
1da177e4
LT
1188 /* List of functions to handle Wireless Extensions (instead of ioctl).
1189 * See <net/iw_handler.h> for details. Jean II */
1190 const struct iw_handler_def * wireless_handlers;
1191 /* Instance data managed by the core of Wireless Extensions. */
1192 struct iw_public_data * wireless_data;
b86e0280 1193#endif
d314774c
SH
1194 /* Management operations */
1195 const struct net_device_ops *netdev_ops;
76fd8593 1196 const struct ethtool_ops *ethtool_ops;
1da177e4 1197
3b04ddde
SH
1198 /* Hardware header description */
1199 const struct header_ops *header_ops;
1200
b00055aa 1201 unsigned int flags; /* interface flags (a la BSD) */
3bdc0eba
BG
1202 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1203 * See if.h for definitions. */
1da177e4 1204 unsigned short gflags;
1da177e4
LT
1205 unsigned short padded; /* How much padding added by alloc_netdev() */
1206
b00055aa
SR
1207 unsigned char operstate; /* RFC2863 operstate */
1208 unsigned char link_mode; /* mapping policy to operstate */
1209
bdc220da
JP
1210 unsigned char if_port; /* Selectable AUI, TP,..*/
1211 unsigned char dma; /* DMA channel */
1212
cd7b5396 1213 unsigned int mtu; /* interface MTU value */
1da177e4
LT
1214 unsigned short type; /* interface hardware type */
1215 unsigned short hard_header_len; /* hardware hdr length */
1da177e4 1216
f5184d26
JB
1217 /* extra head- and tailroom the hardware may need, but not in all cases
1218 * can this be guaranteed, especially tailroom. Some cases also use
1219 * LL_MAX_HEADER instead to allocate the skb.
1220 */
1221 unsigned short needed_headroom;
1222 unsigned short needed_tailroom;
1223
1da177e4 1224 /* Interface address info. */
a6f9a705 1225 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
c1f79426 1226 unsigned char addr_assign_type; /* hw address assignment type */
1da177e4 1227 unsigned char addr_len; /* hardware address length */
596b9b68 1228 unsigned char neigh_priv_len;
dffebd2c
N
1229 unsigned short dev_id; /* Used to differentiate devices
1230 * that share the same link
1231 * layer address
1232 */
ccffad25 1233 spinlock_t addr_list_lock;
22bedad3
JP
1234 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1235 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
4c3d5e7b
ED
1236 struct netdev_hw_addr_list dev_addrs; /* list of device
1237 * hw addresses
1238 */
1239#ifdef CONFIG_SYSFS
1240 struct kset *queues_kset;
1241#endif
1242
2d348d1f 1243 bool uc_promisc;
9d45abe1
WC
1244 unsigned int promiscuity;
1245 unsigned int allmulti;
1da177e4 1246
1da177e4
LT
1247
1248 /* Protocol specific pointers */
65ac6a5f 1249
d11ead75 1250#if IS_ENABLED(CONFIG_VLAN_8021Q)
5b9ea6e0 1251 struct vlan_info __rcu *vlan_info; /* VLAN info */
65ac6a5f 1252#endif
34a430d7 1253#if IS_ENABLED(CONFIG_NET_DSA)
cf50dcc2 1254 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
91da11f8 1255#endif
1da177e4 1256 void *atalk_ptr; /* AppleTalk link */
95ae6b22 1257 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
fc766e4c 1258 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
198caeca 1259 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1da177e4 1260 void *ax25_ptr; /* AX.25 specific data */
704232c2
JB
1261 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1262 assign before registering */
1da177e4 1263
9356b8fc 1264/*
cd13539b 1265 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1266 */
4dc89133
ED
1267 unsigned long last_rx; /* Time of last Rx
1268 * This should not be set in
1269 * drivers, unless really needed,
1270 * because network stack (bonding)
1271 * use it if/when necessary, to
1272 * avoid dirtying this cache line.
1273 */
1274
9356b8fc 1275 /* Interface address info used in eth_type_trans() */
f001fde5
JP
1276 unsigned char *dev_addr; /* hw address, (before bcast
1277 because most packets are
1278 unicast) */
1279
0a9627f2 1280
ccf5ff69 1281#ifdef CONFIG_RPS
0a9627f2
TH
1282 struct netdev_rx_queue *_rx;
1283
62fe0b40 1284 /* Number of RX queues allocated at register_netdev() time */
0a9627f2 1285 unsigned int num_rx_queues;
62fe0b40
BH
1286
1287 /* Number of RX queues currently active in device */
1288 unsigned int real_num_rx_queues;
c445477d 1289
df334545 1290#endif
0a9627f2 1291
61391cde 1292 rx_handler_func_t __rcu *rx_handler;
1293 void __rcu *rx_handler_data;
e8a0464c 1294
24824a09 1295 struct netdev_queue __rcu *ingress_queue;
4c3d5e7b
ED
1296 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1297
cd13539b
ED
1298
1299/*
1300 * Cache lines mostly used on transmit path
1301 */
e8a0464c 1302 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
fd2ea0a7
DM
1303
1304 /* Number of TX queues allocated at alloc_netdev_mq() time */
e8a0464c 1305 unsigned int num_tx_queues;
fd2ea0a7
DM
1306
1307 /* Number of TX queues currently active in device */
1308 unsigned int real_num_tx_queues;
1309
af356afa
PM
1310 /* root qdisc from userspace point of view */
1311 struct Qdisc *qdisc;
1312
1da177e4 1313 unsigned long tx_queue_len; /* Max frames per queue allowed */
c3f26a26 1314 spinlock_t tx_global_lock;
cd13539b 1315
bf264145 1316#ifdef CONFIG_XPS
a4177869 1317 struct xps_dev_maps __rcu *xps_maps;
bf264145 1318#endif
4c3d5e7b
ED
1319#ifdef CONFIG_RFS_ACCEL
1320 /* CPU reverse-mapping for RX completion interrupts, indexed
1321 * by RX queue number. Assigned by driver. This must only be
1322 * set if the ndo_rx_flow_steer operation is defined. */
1323 struct cpu_rmap *rx_cpu_rmap;
1324#endif
1d24eb48 1325
9356b8fc 1326 /* These may be needed for future network-power-down code. */
9d21493b
ED
1327
1328 /*
1329 * trans_start here is expensive for high speed devices on SMP,
1330 * please use netdev_queue->trans_start instead.
1331 */
9356b8fc
ED
1332 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1333
1334 int watchdog_timeo; /* used by dev_watchdog() */
1335 struct timer_list watchdog_timer;
1336
1da177e4 1337 /* Number of references to this device */
29b4433d 1338 int __percpu *pcpu_refcnt;
9356b8fc 1339
1da177e4
LT
1340 /* delayed register/unregister */
1341 struct list_head todo_list;
1da177e4
LT
1342 /* device index hash chain */
1343 struct hlist_node index_hlist;
1344
e014debe 1345 struct list_head link_watch_list;
572a103d 1346
1da177e4
LT
1347 /* register/unregister state machine */
1348 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1349 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1350 NETREG_UNREGISTERING, /* called unregister_netdevice */
1351 NETREG_UNREGISTERED, /* completed unregister todo */
1352 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1353 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1354 } reg_state:8;
1355
1356 bool dismantle; /* device is going do be freed */
a2835763
PM
1357
1358 enum {
1359 RTNL_LINK_INITIALIZED,
1360 RTNL_LINK_INITIALIZING,
1361 } rtnl_link_state:16;
1da177e4 1362
d314774c
SH
1363 /* Called from unregister, can be used to call free_netdev */
1364 void (*destructor)(struct net_device *dev);
1da177e4 1365
1da177e4 1366#ifdef CONFIG_NETPOLL
5fbee843 1367 struct netpoll_info __rcu *npinfo;
1da177e4 1368#endif
eae792b7 1369
c346dca1 1370#ifdef CONFIG_NET_NS
4a1c5371
EB
1371 /* Network namespace this network device is inside */
1372 struct net *nd_net;
c346dca1 1373#endif
4a1c5371 1374
4951704b 1375 /* mid-layer private */
a7855c78
ED
1376 union {
1377 void *ml_priv;
1378 struct pcpu_lstats __percpu *lstats; /* loopback stats */
290b895e 1379 struct pcpu_tstats __percpu *tstats; /* tunnel stats */
6d81f41c 1380 struct pcpu_dstats __percpu *dstats; /* dummy stats */
2681128f 1381 struct pcpu_vstats __percpu *vstats; /* veth stats */
a7855c78 1382 };
eca9ebac 1383 /* GARP */
3cc77ec7 1384 struct garp_port __rcu *garp_port;
febf018d
DW
1385 /* MRP */
1386 struct mrp_port __rcu *mrp_port;
1da177e4 1387
1da177e4 1388 /* class/net/name entry */
43cb76d9 1389 struct device dev;
0c509a6c
EB
1390 /* space for optional device, statistics, and wireless sysfs groups */
1391 const struct attribute_group *sysfs_groups[4];
38f7b870
PM
1392
1393 /* rtnetlink link ops */
1394 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1395
82cc1a7a
PWJ
1396 /* for setting kernel sock attribute on TCP connection setup */
1397#define GSO_MAX_SIZE 65536
1398 unsigned int gso_max_size;
30b678d8
BH
1399#define GSO_MAX_SEGS 65535
1400 u16 gso_max_segs;
d314774c 1401
7a6b6f51 1402#ifdef CONFIG_DCB
2f90b865 1403 /* Data Center Bridging netlink ops */
32953543 1404 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1405#endif
4f57c087
JF
1406 u8 num_tc;
1407 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1408 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1409
d11ead75 1410#if IS_ENABLED(CONFIG_FCOE)
4d288d57
YZ
1411 /* max exchange id for FCoE LRO by ddp */
1412 unsigned int fcoe_ddp_xid;
5bc1421e
NH
1413#endif
1414#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
1415 struct netprio_map __rcu *priomap;
4d288d57 1416#endif
c1f19b51
RC
1417 /* phy device may attach itself for hardware timestamping */
1418 struct phy_device *phydev;
cbda10fa 1419
23d3b8bf
ED
1420 struct lock_class_key *qdisc_tx_busylock;
1421
cbda10fa
VD
1422 /* group the device belongs to */
1423 int group;
9136461a
ED
1424
1425 struct pm_qos_request pm_qos_req;
1da177e4 1426};
43cb76d9 1427#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1428
1429#define NETDEV_ALIGN 32
1da177e4 1430
4f57c087
JF
1431static inline
1432int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1433{
1434 return dev->prio_tc_map[prio & TC_BITMASK];
1435}
1436
1437static inline
1438int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1439{
1440 if (tc >= dev->num_tc)
1441 return -EINVAL;
1442
1443 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1444 return 0;
1445}
1446
1447static inline
1448void netdev_reset_tc(struct net_device *dev)
1449{
1450 dev->num_tc = 0;
1451 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1452 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1453}
1454
1455static inline
1456int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1457{
1458 if (tc >= dev->num_tc)
1459 return -EINVAL;
1460
1461 dev->tc_to_txq[tc].count = count;
1462 dev->tc_to_txq[tc].offset = offset;
1463 return 0;
1464}
1465
1466static inline
1467int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1468{
1469 if (num_tc > TC_MAX_QUEUE)
1470 return -EINVAL;
1471
1472 dev->num_tc = num_tc;
1473 return 0;
1474}
1475
1476static inline
1477int netdev_get_num_tc(struct net_device *dev)
1478{
1479 return dev->num_tc;
1480}
1481
e8a0464c
DM
1482static inline
1483struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1484 unsigned int index)
1485{
1486 return &dev->_tx[index];
1487}
1488
1489static inline void netdev_for_each_tx_queue(struct net_device *dev,
1490 void (*f)(struct net_device *,
1491 struct netdev_queue *,
1492 void *),
1493 void *arg)
1494{
1495 unsigned int i;
1496
1497 for (i = 0; i < dev->num_tx_queues; i++)
1498 f(dev, &dev->_tx[i], arg);
1499}
1500
f629d208
JP
1501struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1502 struct sk_buff *skb);
1503u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb);
8c4c49df 1504
c346dca1
YH
1505/*
1506 * Net namespace inlines
1507 */
1508static inline
1509struct net *dev_net(const struct net_device *dev)
1510{
c2d9ba9b 1511 return read_pnet(&dev->nd_net);
c346dca1
YH
1512}
1513
1514static inline
f5aa23fd 1515void dev_net_set(struct net_device *dev, struct net *net)
c346dca1
YH
1516{
1517#ifdef CONFIG_NET_NS
f3005d7f
DL
1518 release_net(dev->nd_net);
1519 dev->nd_net = hold_net(net);
c346dca1
YH
1520#endif
1521}
1522
cf85d08f
LB
1523static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1524{
1525#ifdef CONFIG_NET_DSA_TAG_DSA
1526 if (dev->dsa_ptr != NULL)
1527 return dsa_uses_dsa_tags(dev->dsa_ptr);
1528#endif
1529
1530 return 0;
1531}
1532
396138f0
LB
1533static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1534{
1535#ifdef CONFIG_NET_DSA_TAG_TRAILER
1536 if (dev->dsa_ptr != NULL)
1537 return dsa_uses_trailer_tags(dev->dsa_ptr);
1538#endif
1539
1540 return 0;
1541}
1542
bea3348e
SH
1543/**
1544 * netdev_priv - access network device private data
1545 * @dev: network device
1546 *
1547 * Get network device private data
1548 */
6472ce60 1549static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1550{
1ce8e7b5 1551 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1552}
1553
1da177e4
LT
1554/* Set the sysfs physical device reference for the network logical device
1555 * if set prior to registration will cause a symlink during initialization.
1556 */
43cb76d9 1557#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1558
384912ed
MH
1559/* Set the sysfs device type for the network logical device to allow
1560 * fin grained indentification of different network device types. For
1561 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1562 */
1563#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1564
82dc3c63
ED
1565/* Default NAPI poll() weight
1566 * Device drivers are strongly advised to not use bigger value
1567 */
1568#define NAPI_POLL_WEIGHT 64
1569
3b582cc1
SH
1570/**
1571 * netif_napi_add - initialize a napi context
1572 * @dev: network device
1573 * @napi: napi context
1574 * @poll: polling function
1575 * @weight: default weight
1576 *
1577 * netif_napi_add() must be used to initialize a napi context prior to calling
1578 * *any* of the other napi related functions.
1579 */
d565b0a1
HX
1580void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1581 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1582
d8156534
AD
1583/**
1584 * netif_napi_del - remove a napi context
1585 * @napi: napi context
1586 *
1587 * netif_napi_del() removes a napi context from the network device napi list
1588 */
d565b0a1
HX
1589void netif_napi_del(struct napi_struct *napi);
1590
1591struct napi_gro_cb {
78a478d0
HX
1592 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1593 void *frag0;
1594
7489594c
HX
1595 /* Length of frag0. */
1596 unsigned int frag0_len;
1597
86911732
HX
1598 /* This indicates where we are processing relative to skb->data. */
1599 int data_offset;
1600
d565b0a1
HX
1601 /* This is non-zero if the packet cannot be merged with the new skb. */
1602 int flush;
1603
1604 /* Number of segments aggregated. */
2e71a6f8
ED
1605 u16 count;
1606
1607 /* This is non-zero if the packet may be of the same flow. */
1608 u8 same_flow;
5d38a079
HX
1609
1610 /* Free the skb? */
2e71a6f8 1611 u8 free;
d7e8883c
ED
1612#define NAPI_GRO_FREE 1
1613#define NAPI_GRO_FREE_STOLEN_HEAD 2
2e71a6f8
ED
1614
1615 /* jiffies when first packet was created/queued */
1616 unsigned long age;
86347245
ED
1617
1618 /* Used in ipv6_gro_receive() */
1619 int proto;
c3c7c254
ED
1620
1621 /* used in skb_gro_receive() slow path */
1622 struct sk_buff *last;
d565b0a1
HX
1623};
1624
1625#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 1626
1da177e4 1627struct packet_type {
f2ccd8fa
DM
1628 __be16 type; /* This is really htons(ether_type). */
1629 struct net_device *dev; /* NULL is wildcarded here */
1630 int (*func) (struct sk_buff *,
1631 struct net_device *,
1632 struct packet_type *,
1633 struct net_device *);
c0de08d0
EL
1634 bool (*id_match)(struct packet_type *ptype,
1635 struct sock *sk);
1da177e4
LT
1636 void *af_packet_priv;
1637 struct list_head list;
1638};
1639
f191a1d1 1640struct offload_callbacks {
576a30eb 1641 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 1642 netdev_features_t features);
a430a43d 1643 int (*gso_send_check)(struct sk_buff *skb);
d565b0a1
HX
1644 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1645 struct sk_buff *skb);
1646 int (*gro_complete)(struct sk_buff *skb);
f191a1d1
VY
1647};
1648
1649struct packet_offload {
1650 __be16 type; /* This is really htons(ether_type). */
1651 struct offload_callbacks callbacks;
1652 struct list_head list;
1da177e4
LT
1653};
1654
1da177e4
LT
1655#include <linux/notifier.h>
1656
dcfe1421
AW
1657/* netdevice notifier chain. Please remember to update the rtnetlink
1658 * notification exclusion list in rtnetlink_event() when adding new
1659 * types.
1660 */
1661#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1662#define NETDEV_DOWN 0x0002
1663#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1664 detected a hardware crash and restarted
1665 - we can use this eg to kick tcp sessions
1666 once done */
1667#define NETDEV_CHANGE 0x0004 /* Notify device state change */
1668#define NETDEV_REGISTER 0x0005
1669#define NETDEV_UNREGISTER 0x0006
1670#define NETDEV_CHANGEMTU 0x0007
1671#define NETDEV_CHANGEADDR 0x0008
1672#define NETDEV_GOING_DOWN 0x0009
1673#define NETDEV_CHANGENAME 0x000A
1674#define NETDEV_FEAT_CHANGE 0x000B
1675#define NETDEV_BONDING_FAILOVER 0x000C
1676#define NETDEV_PRE_UP 0x000D
1677#define NETDEV_PRE_TYPE_CHANGE 0x000E
1678#define NETDEV_POST_TYPE_CHANGE 0x000F
1679#define NETDEV_POST_INIT 0x0010
0115e8e3 1680#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
1681#define NETDEV_RELEASE 0x0012
1682#define NETDEV_NOTIFY_PEERS 0x0013
1683#define NETDEV_JOIN 0x0014
42e52bf9 1684#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 1685#define NETDEV_RESEND_IGMP 0x0016
dcfe1421 1686
f629d208
JP
1687int register_netdevice_notifier(struct notifier_block *nb);
1688int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
1689
1690struct netdev_notifier_info {
1691 struct net_device *dev;
1692};
1693
be9efd36
JP
1694struct netdev_notifier_change_info {
1695 struct netdev_notifier_info info; /* must be first */
1696 unsigned int flags_changed;
1697};
1698
75538c2b
CW
1699static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1700 struct net_device *dev)
1701{
1702 info->dev = dev;
1703}
1704
351638e7
JP
1705static inline struct net_device *
1706netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1707{
1708 return info->dev;
1709}
1710
f629d208
JP
1711int call_netdevice_notifiers_info(unsigned long val, struct net_device *dev,
1712 struct netdev_notifier_info *info);
1713int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
1714
1715
1da177e4
LT
1716extern rwlock_t dev_base_lock; /* Device list lock */
1717
881d966b
EB
1718#define for_each_netdev(net, d) \
1719 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
1720#define for_each_netdev_reverse(net, d) \
1721 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
1722#define for_each_netdev_rcu(net, d) \
1723 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
1724#define for_each_netdev_safe(net, d, n) \
1725 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1726#define for_each_netdev_continue(net, d) \
1727 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 1728#define for_each_netdev_continue_rcu(net, d) \
1729 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 1730#define for_each_netdev_in_bond_rcu(bond, slave) \
1731 for_each_netdev_rcu(&init_net, slave) \
1732 if (netdev_master_upper_dev_get_rcu(slave) == bond)
881d966b 1733#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 1734
a050c33f
DL
1735static inline struct net_device *next_net_device(struct net_device *dev)
1736{
1737 struct list_head *lh;
1738 struct net *net;
1739
c346dca1 1740 net = dev_net(dev);
a050c33f
DL
1741 lh = dev->dev_list.next;
1742 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1743}
1744
ce81b76a
ED
1745static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1746{
1747 struct list_head *lh;
1748 struct net *net;
1749
1750 net = dev_net(dev);
ccf43438 1751 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
1752 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1753}
1754
a050c33f
DL
1755static inline struct net_device *first_net_device(struct net *net)
1756{
1757 return list_empty(&net->dev_base_head) ? NULL :
1758 net_device_entry(net->dev_base_head.next);
1759}
7562f876 1760
ccf43438
ED
1761static inline struct net_device *first_net_device_rcu(struct net *net)
1762{
1763 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1764
1765 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1766}
1767
f629d208
JP
1768int netdev_boot_setup_check(struct net_device *dev);
1769unsigned long netdev_boot_base(const char *prefix, int unit);
1770struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1771 const char *hwaddr);
1772struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1773struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1774void dev_add_pack(struct packet_type *pt);
1775void dev_remove_pack(struct packet_type *pt);
1776void __dev_remove_pack(struct packet_type *pt);
1777void dev_add_offload(struct packet_offload *po);
1778void dev_remove_offload(struct packet_offload *po);
1779void __dev_remove_offload(struct packet_offload *po);
1780
1781struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1782 unsigned short mask);
1783struct net_device *dev_get_by_name(struct net *net, const char *name);
1784struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1785struct net_device *__dev_get_by_name(struct net *net, const char *name);
1786int dev_alloc_name(struct net_device *dev, const char *name);
1787int dev_open(struct net_device *dev);
1788int dev_close(struct net_device *dev);
1789void dev_disable_lro(struct net_device *dev);
1790int dev_loopback_xmit(struct sk_buff *newskb);
1791int dev_queue_xmit(struct sk_buff *skb);
1792int register_netdevice(struct net_device *dev);
1793void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
1794void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
1795static inline void unregister_netdevice(struct net_device *dev)
1796{
1797 unregister_netdevice_queue(dev, NULL);
1798}
1799
f629d208
JP
1800int netdev_refcnt_read(const struct net_device *dev);
1801void free_netdev(struct net_device *dev);
1802void synchronize_net(void);
1803int init_dummy_netdev(struct net_device *dev);
937f1ba5 1804
f629d208
JP
1805struct net_device *dev_get_by_index(struct net *net, int ifindex);
1806struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1807struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1808int netdev_get_name(struct net *net, char *name, int ifindex);
1809int dev_restart(struct net_device *dev);
1da177e4 1810#ifdef CONFIG_NETPOLL_TRAP
f629d208 1811int netpoll_trap(void);
1da177e4 1812#endif
f629d208 1813int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
1814
1815static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1816{
1817 return NAPI_GRO_CB(skb)->data_offset;
1818}
1819
1820static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1821{
1822 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1823}
1824
1825static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1826{
1827 NAPI_GRO_CB(skb)->data_offset += len;
1828}
1829
a5b1cf28
HX
1830static inline void *skb_gro_header_fast(struct sk_buff *skb,
1831 unsigned int offset)
86911732 1832{
a5b1cf28
HX
1833 return NAPI_GRO_CB(skb)->frag0 + offset;
1834}
78a478d0 1835
a5b1cf28
HX
1836static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1837{
1838 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1839}
78a478d0 1840
a5b1cf28
HX
1841static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1842 unsigned int offset)
1843{
17dd759c
HX
1844 if (!pskb_may_pull(skb, hlen))
1845 return NULL;
1846
a5b1cf28
HX
1847 NAPI_GRO_CB(skb)->frag0 = NULL;
1848 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 1849 return skb->data + offset;
86911732 1850}
1da177e4 1851
aa4b9f53
HX
1852static inline void *skb_gro_mac_header(struct sk_buff *skb)
1853{
78d3fd0b 1854 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
aa4b9f53
HX
1855}
1856
36e7b1b8
HX
1857static inline void *skb_gro_network_header(struct sk_buff *skb)
1858{
78d3fd0b
HX
1859 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1860 skb_network_offset(skb);
36e7b1b8
HX
1861}
1862
0c4e8581
SH
1863static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1864 unsigned short type,
3b04ddde 1865 const void *daddr, const void *saddr,
95c96174 1866 unsigned int len)
0c4e8581 1867{
f1ecfd5d 1868 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 1869 return 0;
3b04ddde
SH
1870
1871 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
1872}
1873
b95cce35
SH
1874static inline int dev_parse_header(const struct sk_buff *skb,
1875 unsigned char *haddr)
1876{
1877 const struct net_device *dev = skb->dev;
1878
1b83336b 1879 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 1880 return 0;
3b04ddde 1881 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
1882}
1883
1da177e4 1884typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 1885int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
1886static inline int unregister_gifconf(unsigned int family)
1887{
1888 return register_gifconf(family, NULL);
1889}
1890
99bbc707 1891#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 1892#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
1893struct sd_flow_limit {
1894 u64 count;
1895 unsigned int num_buckets;
1896 unsigned int history_head;
1897 u16 history[FLOW_LIMIT_HISTORY];
1898 u8 buckets[];
1899};
1900
1901extern int netdev_flow_limit_table_len;
1902#endif /* CONFIG_NET_FLOW_LIMIT */
1903
1da177e4 1904/*
88751275 1905 * Incoming packets are placed on per-cpu queues
1da177e4 1906 */
d94d9fee 1907struct softnet_data {
37437bb2 1908 struct Qdisc *output_queue;
a9cbd588 1909 struct Qdisc **output_queue_tailp;
1da177e4 1910 struct list_head poll_list;
1da177e4 1911 struct sk_buff *completion_queue;
6e7676c1 1912 struct sk_buff_head process_queue;
1da177e4 1913
dee42870 1914 /* stats */
cd7b5396
DM
1915 unsigned int processed;
1916 unsigned int time_squeeze;
1917 unsigned int cpu_collision;
1918 unsigned int received_rps;
dee42870 1919
fd793d89 1920#ifdef CONFIG_RPS
88751275
ED
1921 struct softnet_data *rps_ipi_list;
1922
1923 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 1924 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
1925 struct softnet_data *rps_ipi_next;
1926 unsigned int cpu;
fec5e652 1927 unsigned int input_queue_head;
76cc8b13 1928 unsigned int input_queue_tail;
1e94d72f 1929#endif
95c96174 1930 unsigned int dropped;
0a9627f2 1931 struct sk_buff_head input_pkt_queue;
bea3348e 1932 struct napi_struct backlog;
99bbc707
WB
1933
1934#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 1935 struct sd_flow_limit __rcu *flow_limit;
99bbc707 1936#endif
1da177e4
LT
1937};
1938
76cc8b13 1939static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
1940{
1941#ifdef CONFIG_RPS
76cc8b13
TH
1942 sd->input_queue_head++;
1943#endif
1944}
1945
1946static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1947 unsigned int *qtail)
1948{
1949#ifdef CONFIG_RPS
1950 *qtail = ++sd->input_queue_tail;
fec5e652
TH
1951#endif
1952}
1953
0a9627f2 1954DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 1955
f629d208 1956void __netif_schedule(struct Qdisc *q);
1da177e4 1957
86d804e1 1958static inline void netif_schedule_queue(struct netdev_queue *txq)
1da177e4 1959{
73466498 1960 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
37437bb2 1961 __netif_schedule(txq->qdisc);
86d804e1
DM
1962}
1963
fd2ea0a7
DM
1964static inline void netif_tx_schedule_all(struct net_device *dev)
1965{
1966 unsigned int i;
1967
1968 for (i = 0; i < dev->num_tx_queues; i++)
1969 netif_schedule_queue(netdev_get_tx_queue(dev, i));
1970}
1971
d29f749e
DJ
1972static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
1973{
73466498 1974 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
1975}
1976
bea3348e
SH
1977/**
1978 * netif_start_queue - allow transmit
1979 * @dev: network device
1980 *
1981 * Allow upper layers to call the device hard_start_xmit routine.
1982 */
1da177e4
LT
1983static inline void netif_start_queue(struct net_device *dev)
1984{
e8a0464c 1985 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
1986}
1987
fd2ea0a7
DM
1988static inline void netif_tx_start_all_queues(struct net_device *dev)
1989{
1990 unsigned int i;
1991
1992 for (i = 0; i < dev->num_tx_queues; i++) {
1993 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1994 netif_tx_start_queue(txq);
1995 }
1996}
1997
79d16385 1998static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1da177e4
LT
1999{
2000#ifdef CONFIG_NETPOLL_TRAP
5f286e11 2001 if (netpoll_trap()) {
7b3d3e4f 2002 netif_tx_start_queue(dev_queue);
1da177e4 2003 return;
5f286e11 2004 }
1da177e4 2005#endif
73466498 2006 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
37437bb2 2007 __netif_schedule(dev_queue->qdisc);
79d16385
DM
2008}
2009
d29f749e
DJ
2010/**
2011 * netif_wake_queue - restart transmit
2012 * @dev: network device
2013 *
2014 * Allow upper layers to call the device hard_start_xmit routine.
2015 * Used for flow control when transmit resources are available.
2016 */
79d16385
DM
2017static inline void netif_wake_queue(struct net_device *dev)
2018{
e8a0464c 2019 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2020}
2021
fd2ea0a7
DM
2022static inline void netif_tx_wake_all_queues(struct net_device *dev)
2023{
2024 unsigned int i;
2025
2026 for (i = 0; i < dev->num_tx_queues; i++) {
2027 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2028 netif_tx_wake_queue(txq);
2029 }
2030}
2031
d29f749e
DJ
2032static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2033{
18543a64 2034 if (WARN_ON(!dev_queue)) {
256ee435 2035 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
18543a64
GC
2036 return;
2037 }
73466498 2038 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2039}
2040
bea3348e
SH
2041/**
2042 * netif_stop_queue - stop transmitted packets
2043 * @dev: network device
2044 *
2045 * Stop upper layers calling the device hard_start_xmit routine.
2046 * Used for flow control when transmit resources are unavailable.
2047 */
1da177e4
LT
2048static inline void netif_stop_queue(struct net_device *dev)
2049{
e8a0464c 2050 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2051}
2052
fd2ea0a7
DM
2053static inline void netif_tx_stop_all_queues(struct net_device *dev)
2054{
2055 unsigned int i;
2056
2057 for (i = 0; i < dev->num_tx_queues; i++) {
2058 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2059 netif_tx_stop_queue(txq);
2060 }
2061}
2062
4d29515f 2063static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2064{
73466498 2065 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2066}
2067
bea3348e
SH
2068/**
2069 * netif_queue_stopped - test if transmit queue is flowblocked
2070 * @dev: network device
2071 *
2072 * Test if transmit queue on device is currently unable to send.
2073 */
4d29515f 2074static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2075{
e8a0464c 2076 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2077}
2078
4d29515f 2079static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2080{
73466498
TH
2081 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2082}
2083
4d29515f 2084static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2085{
2086 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2087}
2088
c5d67bd7
TH
2089static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2090 unsigned int bytes)
2091{
114cf580
TH
2092#ifdef CONFIG_BQL
2093 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2094
2095 if (likely(dql_avail(&dev_queue->dql) >= 0))
2096 return;
2097
2098 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2099
2100 /*
2101 * The XOFF flag must be set before checking the dql_avail below,
2102 * because in netdev_tx_completed_queue we update the dql_completed
2103 * before checking the XOFF flag.
2104 */
2105 smp_mb();
2106
2107 /* check again in case another CPU has just made room avail */
2108 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2109 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2110#endif
c5d67bd7
TH
2111}
2112
0042d0c8
FF
2113/**
2114 * netdev_sent_queue - report the number of bytes queued to hardware
2115 * @dev: network device
2116 * @bytes: number of bytes queued to the hardware device queue
2117 *
2118 * Report the number of bytes queued for sending/completion to the network
2119 * device hardware queue. @bytes should be a good approximation and should
2120 * exactly match netdev_completed_queue() @bytes
2121 */
c5d67bd7
TH
2122static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2123{
2124 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2125}
2126
2127static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2128 unsigned int pkts, unsigned int bytes)
c5d67bd7 2129{
114cf580 2130#ifdef CONFIG_BQL
b37c0fbe
AD
2131 if (unlikely(!bytes))
2132 return;
2133
2134 dql_completed(&dev_queue->dql, bytes);
2135
2136 /*
2137 * Without the memory barrier there is a small possiblity that
2138 * netdev_tx_sent_queue will miss the update and cause the queue to
2139 * be stopped forever
2140 */
2141 smp_mb();
2142
2143 if (dql_avail(&dev_queue->dql) < 0)
2144 return;
2145
2146 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2147 netif_schedule_queue(dev_queue);
114cf580 2148#endif
c5d67bd7
TH
2149}
2150
0042d0c8
FF
2151/**
2152 * netdev_completed_queue - report bytes and packets completed by device
2153 * @dev: network device
2154 * @pkts: actual number of packets sent over the medium
2155 * @bytes: actual number of bytes sent over the medium
2156 *
2157 * Report the number of bytes and packets transmitted by the network device
2158 * hardware queue over the physical medium, @bytes must exactly match the
2159 * @bytes amount passed to netdev_sent_queue()
2160 */
c5d67bd7 2161static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2162 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2163{
2164 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2165}
2166
2167static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2168{
114cf580 2169#ifdef CONFIG_BQL
5c490354 2170 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2171 dql_reset(&q->dql);
2172#endif
c5d67bd7
TH
2173}
2174
0042d0c8
FF
2175/**
2176 * netdev_reset_queue - reset the packets and bytes count of a network device
2177 * @dev_queue: network device
2178 *
2179 * Reset the bytes and packet count of a network device and clear the
2180 * software flow control OFF bit for this network device
2181 */
c5d67bd7
TH
2182static inline void netdev_reset_queue(struct net_device *dev_queue)
2183{
2184 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2185}
2186
bea3348e
SH
2187/**
2188 * netif_running - test if up
2189 * @dev: network device
2190 *
2191 * Test if the device has been brought up.
2192 */
4d29515f 2193static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2194{
2195 return test_bit(__LINK_STATE_START, &dev->state);
2196}
2197
f25f4e44
PWJ
2198/*
2199 * Routines to manage the subqueues on a device. We only need start
2200 * stop, and a check if it's stopped. All other device management is
2201 * done at the overall netdevice level.
2202 * Also test the device if we're multiqueue.
2203 */
bea3348e
SH
2204
2205/**
2206 * netif_start_subqueue - allow sending packets on subqueue
2207 * @dev: network device
2208 * @queue_index: sub queue index
2209 *
2210 * Start individual transmit queue of a device with multiple transmit queues.
2211 */
f25f4e44
PWJ
2212static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2213{
fd2ea0a7 2214 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2215
2216 netif_tx_start_queue(txq);
f25f4e44
PWJ
2217}
2218
bea3348e
SH
2219/**
2220 * netif_stop_subqueue - stop sending packets on subqueue
2221 * @dev: network device
2222 * @queue_index: sub queue index
2223 *
2224 * Stop individual transmit queue of a device with multiple transmit queues.
2225 */
f25f4e44
PWJ
2226static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2227{
fd2ea0a7 2228 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
f25f4e44
PWJ
2229#ifdef CONFIG_NETPOLL_TRAP
2230 if (netpoll_trap())
2231 return;
2232#endif
7b3d3e4f 2233 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2234}
2235
bea3348e
SH
2236/**
2237 * netif_subqueue_stopped - test status of subqueue
2238 * @dev: network device
2239 * @queue_index: sub queue index
2240 *
2241 * Check individual transmit queue of a device with multiple transmit queues.
2242 */
4d29515f
DM
2243static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2244 u16 queue_index)
f25f4e44 2245{
fd2ea0a7 2246 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2247
2248 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2249}
2250
4d29515f
DM
2251static inline bool netif_subqueue_stopped(const struct net_device *dev,
2252 struct sk_buff *skb)
668f895a
PE
2253{
2254 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2255}
bea3348e
SH
2256
2257/**
2258 * netif_wake_subqueue - allow sending packets on subqueue
2259 * @dev: network device
2260 * @queue_index: sub queue index
2261 *
2262 * Resume individual transmit queue of a device with multiple transmit queues.
2263 */
f25f4e44
PWJ
2264static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2265{
fd2ea0a7 2266 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
f25f4e44
PWJ
2267#ifdef CONFIG_NETPOLL_TRAP
2268 if (netpoll_trap())
2269 return;
2270#endif
73466498 2271 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
37437bb2 2272 __netif_schedule(txq->qdisc);
f25f4e44
PWJ
2273}
2274
537c00de 2275#ifdef CONFIG_XPS
53af53ae 2276int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2277 u16 index);
537c00de
AD
2278#else
2279static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2280 const struct cpumask *mask,
537c00de
AD
2281 u16 index)
2282{
2283 return 0;
2284}
2285#endif
2286
a3d22a68
VZ
2287/*
2288 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2289 * as a distribution range limit for the returned value.
2290 */
2291static inline u16 skb_tx_hash(const struct net_device *dev,
2292 const struct sk_buff *skb)
2293{
2294 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2295}
2296
bea3348e
SH
2297/**
2298 * netif_is_multiqueue - test if device has multiple transmit queues
2299 * @dev: network device
2300 *
2301 * Check if device has multiple transmit queues
bea3348e 2302 */
4d29515f 2303static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2304{
a02cec21 2305 return dev->num_tx_queues > 1;
f25f4e44 2306}
1da177e4 2307
f629d208 2308int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2309
62fe0b40 2310#ifdef CONFIG_RPS
f629d208 2311int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
2312#else
2313static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2314 unsigned int rxq)
2315{
2316 return 0;
2317}
2318#endif
2319
3171d026
BH
2320static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2321 const struct net_device *from_dev)
2322{
ee6ae1a1
JP
2323 int err;
2324
2325 err = netif_set_real_num_tx_queues(to_dev,
2326 from_dev->real_num_tx_queues);
2327 if (err)
2328 return err;
3171d026
BH
2329#ifdef CONFIG_RPS
2330 return netif_set_real_num_rx_queues(to_dev,
2331 from_dev->real_num_rx_queues);
2332#else
2333 return 0;
2334#endif
2335}
2336
16917b87 2337#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 2338int netif_get_num_default_rss_queues(void);
16917b87 2339
1da177e4 2340/* Use this variant when it is known for sure that it
0ef47309
ML
2341 * is executing from hardware interrupt context or with hardware interrupts
2342 * disabled.
1da177e4 2343 */
f629d208 2344void dev_kfree_skb_irq(struct sk_buff *skb);
1da177e4
LT
2345
2346/* Use this variant in places where it could be invoked
0ef47309
ML
2347 * from either hardware interrupt or other context, with hardware interrupts
2348 * either disabled or enabled.
1da177e4 2349 */
f629d208 2350void dev_kfree_skb_any(struct sk_buff *skb);
1da177e4 2351
f629d208
JP
2352int netif_rx(struct sk_buff *skb);
2353int netif_rx_ni(struct sk_buff *skb);
2354int netif_receive_skb(struct sk_buff *skb);
2355gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2356void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2357struct sk_buff *napi_get_frags(struct napi_struct *napi);
2358gro_result_t napi_gro_frags(struct napi_struct *napi);
76620aaf
HX
2359
2360static inline void napi_free_frags(struct napi_struct *napi)
2361{
2362 kfree_skb(napi->skb);
2363 napi->skb = NULL;
2364}
2365
f629d208
JP
2366int netdev_rx_handler_register(struct net_device *dev,
2367 rx_handler_func_t *rx_handler,
2368 void *rx_handler_data);
2369void netdev_rx_handler_unregister(struct net_device *dev);
2370
2371bool dev_valid_name(const char *name);
2372int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2373int dev_ethtool(struct net *net, struct ifreq *);
2374unsigned int dev_get_flags(const struct net_device *);
2375int __dev_change_flags(struct net_device *, unsigned int flags);
2376int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
2377void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2378 unsigned int gchanges);
f629d208
JP
2379int dev_change_name(struct net_device *, const char *);
2380int dev_set_alias(struct net_device *, const char *, size_t);
2381int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2382int dev_set_mtu(struct net_device *, int);
2383void dev_set_group(struct net_device *, int);
2384int dev_set_mac_address(struct net_device *, struct sockaddr *);
2385int dev_change_carrier(struct net_device *, bool new_carrier);
2386int dev_get_phys_port_id(struct net_device *dev,
2387 struct netdev_phys_port_id *ppid);
2388int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2389 struct netdev_queue *txq);
2390int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1da177e4 2391
20380731 2392extern int netdev_budget;
1da177e4
LT
2393
2394/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 2395void netdev_run_todo(void);
1da177e4 2396
bea3348e
SH
2397/**
2398 * dev_put - release reference to device
2399 * @dev: network device
2400 *
9ef4429b 2401 * Release reference to device to allow it to be freed.
bea3348e 2402 */
1da177e4
LT
2403static inline void dev_put(struct net_device *dev)
2404{
933393f5 2405 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
2406}
2407
bea3348e
SH
2408/**
2409 * dev_hold - get reference to device
2410 * @dev: network device
2411 *
9ef4429b 2412 * Hold reference to device to keep it from being freed.
bea3348e 2413 */
15333061
SH
2414static inline void dev_hold(struct net_device *dev)
2415{
933393f5 2416 this_cpu_inc(*dev->pcpu_refcnt);
15333061 2417}
1da177e4
LT
2418
2419/* Carrier loss detection, dial on demand. The functions netif_carrier_on
2420 * and _off may be called from IRQ context, but it is caller
2421 * who is responsible for serialization of these calls.
b00055aa
SR
2422 *
2423 * The name carrier is inappropriate, these functions should really be
2424 * called netif_lowerlayer_*() because they represent the state of any
2425 * kind of lower layer not just hardware media.
1da177e4
LT
2426 */
2427
f629d208
JP
2428void linkwatch_init_dev(struct net_device *dev);
2429void linkwatch_fire_event(struct net_device *dev);
2430void linkwatch_forget_dev(struct net_device *dev);
1da177e4 2431
bea3348e
SH
2432/**
2433 * netif_carrier_ok - test if carrier present
2434 * @dev: network device
2435 *
2436 * Check if carrier is present on device
2437 */
4d29515f 2438static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
2439{
2440 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2441}
2442
f629d208 2443unsigned long dev_trans_start(struct net_device *dev);
9d21493b 2444
f629d208 2445void __netdev_watchdog_up(struct net_device *dev);
1da177e4 2446
f629d208 2447void netif_carrier_on(struct net_device *dev);
1da177e4 2448
f629d208 2449void netif_carrier_off(struct net_device *dev);
1da177e4 2450
bea3348e
SH
2451/**
2452 * netif_dormant_on - mark device as dormant.
2453 * @dev: network device
2454 *
2455 * Mark device as dormant (as per RFC2863).
2456 *
2457 * The dormant state indicates that the relevant interface is not
2458 * actually in a condition to pass packets (i.e., it is not 'up') but is
2459 * in a "pending" state, waiting for some external event. For "on-
2460 * demand" interfaces, this new state identifies the situation where the
2461 * interface is waiting for events to place it in the up state.
2462 *
2463 */
b00055aa
SR
2464static inline void netif_dormant_on(struct net_device *dev)
2465{
2466 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2467 linkwatch_fire_event(dev);
2468}
2469
bea3348e
SH
2470/**
2471 * netif_dormant_off - set device as not dormant.
2472 * @dev: network device
2473 *
2474 * Device is not in dormant state.
2475 */
b00055aa
SR
2476static inline void netif_dormant_off(struct net_device *dev)
2477{
2478 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2479 linkwatch_fire_event(dev);
2480}
2481
bea3348e
SH
2482/**
2483 * netif_dormant - test if carrier present
2484 * @dev: network device
2485 *
2486 * Check if carrier is present on device
2487 */
4d29515f 2488static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
2489{
2490 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2491}
2492
2493
bea3348e
SH
2494/**
2495 * netif_oper_up - test if device is operational
2496 * @dev: network device
2497 *
2498 * Check if carrier is operational
2499 */
4d29515f 2500static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 2501{
b00055aa
SR
2502 return (dev->operstate == IF_OPER_UP ||
2503 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2504}
2505
bea3348e
SH
2506/**
2507 * netif_device_present - is device available or removed
2508 * @dev: network device
2509 *
2510 * Check if device has not been removed from system.
2511 */
4d29515f 2512static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
2513{
2514 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2515}
2516
f629d208 2517void netif_device_detach(struct net_device *dev);
1da177e4 2518
f629d208 2519void netif_device_attach(struct net_device *dev);
1da177e4
LT
2520
2521/*
2522 * Network interface message level settings
2523 */
1da177e4
LT
2524
2525enum {
2526 NETIF_MSG_DRV = 0x0001,
2527 NETIF_MSG_PROBE = 0x0002,
2528 NETIF_MSG_LINK = 0x0004,
2529 NETIF_MSG_TIMER = 0x0008,
2530 NETIF_MSG_IFDOWN = 0x0010,
2531 NETIF_MSG_IFUP = 0x0020,
2532 NETIF_MSG_RX_ERR = 0x0040,
2533 NETIF_MSG_TX_ERR = 0x0080,
2534 NETIF_MSG_TX_QUEUED = 0x0100,
2535 NETIF_MSG_INTR = 0x0200,
2536 NETIF_MSG_TX_DONE = 0x0400,
2537 NETIF_MSG_RX_STATUS = 0x0800,
2538 NETIF_MSG_PKTDATA = 0x1000,
2539 NETIF_MSG_HW = 0x2000,
2540 NETIF_MSG_WOL = 0x4000,
2541};
2542
2543#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2544#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2545#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2546#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2547#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2548#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2549#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2550#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2551#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2552#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2553#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2554#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2555#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2556#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2557#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2558
2559static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2560{
2561 /* use default */
2562 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2563 return default_msg_enable_bits;
2564 if (debug_value == 0) /* no output */
2565 return 0;
2566 /* set low N bits */
2567 return (1 << debug_value) - 1;
2568}
2569
c773e847 2570static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 2571{
c773e847
DM
2572 spin_lock(&txq->_xmit_lock);
2573 txq->xmit_lock_owner = cpu;
22dd7495
JHS
2574}
2575
fd2ea0a7
DM
2576static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2577{
2578 spin_lock_bh(&txq->_xmit_lock);
2579 txq->xmit_lock_owner = smp_processor_id();
2580}
2581
4d29515f 2582static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 2583{
4d29515f 2584 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
2585 if (likely(ok))
2586 txq->xmit_lock_owner = smp_processor_id();
2587 return ok;
2588}
2589
2590static inline void __netif_tx_unlock(struct netdev_queue *txq)
2591{
2592 txq->xmit_lock_owner = -1;
2593 spin_unlock(&txq->_xmit_lock);
2594}
2595
2596static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2597{
2598 txq->xmit_lock_owner = -1;
2599 spin_unlock_bh(&txq->_xmit_lock);
2600}
2601
08baf561
ED
2602static inline void txq_trans_update(struct netdev_queue *txq)
2603{
2604 if (txq->xmit_lock_owner != -1)
2605 txq->trans_start = jiffies;
2606}
2607
d29f749e
DJ
2608/**
2609 * netif_tx_lock - grab network device transmit lock
2610 * @dev: network device
d29f749e
DJ
2611 *
2612 * Get network device transmit lock
2613 */
22dd7495
JHS
2614static inline void netif_tx_lock(struct net_device *dev)
2615{
e8a0464c 2616 unsigned int i;
c3f26a26 2617 int cpu;
c773e847 2618
c3f26a26
DM
2619 spin_lock(&dev->tx_global_lock);
2620 cpu = smp_processor_id();
e8a0464c
DM
2621 for (i = 0; i < dev->num_tx_queues; i++) {
2622 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
2623
2624 /* We are the only thread of execution doing a
2625 * freeze, but we have to grab the _xmit_lock in
2626 * order to synchronize with threads which are in
2627 * the ->hard_start_xmit() handler and already
2628 * checked the frozen bit.
2629 */
e8a0464c 2630 __netif_tx_lock(txq, cpu);
c3f26a26
DM
2631 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2632 __netif_tx_unlock(txq);
e8a0464c 2633 }
932ff279
HX
2634}
2635
2636static inline void netif_tx_lock_bh(struct net_device *dev)
2637{
e8a0464c
DM
2638 local_bh_disable();
2639 netif_tx_lock(dev);
932ff279
HX
2640}
2641
932ff279
HX
2642static inline void netif_tx_unlock(struct net_device *dev)
2643{
e8a0464c
DM
2644 unsigned int i;
2645
2646 for (i = 0; i < dev->num_tx_queues; i++) {
2647 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 2648
c3f26a26
DM
2649 /* No need to grab the _xmit_lock here. If the
2650 * queue is not stopped for another reason, we
2651 * force a schedule.
2652 */
2653 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 2654 netif_schedule_queue(txq);
c3f26a26
DM
2655 }
2656 spin_unlock(&dev->tx_global_lock);
932ff279
HX
2657}
2658
2659static inline void netif_tx_unlock_bh(struct net_device *dev)
2660{
e8a0464c
DM
2661 netif_tx_unlock(dev);
2662 local_bh_enable();
932ff279
HX
2663}
2664
c773e847 2665#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 2666 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 2667 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
2668 } \
2669}
2670
c773e847 2671#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 2672 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 2673 __netif_tx_unlock(txq); \
22dd7495
JHS
2674 } \
2675}
2676
1da177e4
LT
2677static inline void netif_tx_disable(struct net_device *dev)
2678{
fd2ea0a7 2679 unsigned int i;
c3f26a26 2680 int cpu;
fd2ea0a7 2681
c3f26a26
DM
2682 local_bh_disable();
2683 cpu = smp_processor_id();
fd2ea0a7
DM
2684 for (i = 0; i < dev->num_tx_queues; i++) {
2685 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
2686
2687 __netif_tx_lock(txq, cpu);
fd2ea0a7 2688 netif_tx_stop_queue(txq);
c3f26a26 2689 __netif_tx_unlock(txq);
fd2ea0a7 2690 }
c3f26a26 2691 local_bh_enable();
1da177e4
LT
2692}
2693
e308a5d8
DM
2694static inline void netif_addr_lock(struct net_device *dev)
2695{
2696 spin_lock(&dev->addr_list_lock);
2697}
2698
2429f7ac
JP
2699static inline void netif_addr_lock_nested(struct net_device *dev)
2700{
2701 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2702}
2703
e308a5d8
DM
2704static inline void netif_addr_lock_bh(struct net_device *dev)
2705{
2706 spin_lock_bh(&dev->addr_list_lock);
2707}
2708
2709static inline void netif_addr_unlock(struct net_device *dev)
2710{
2711 spin_unlock(&dev->addr_list_lock);
2712}
2713
2714static inline void netif_addr_unlock_bh(struct net_device *dev)
2715{
2716 spin_unlock_bh(&dev->addr_list_lock);
2717}
2718
f001fde5 2719/*
31278e71 2720 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
2721 * rcu_read_lock held.
2722 */
2723#define for_each_dev_addr(dev, ha) \
31278e71 2724 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 2725
1da177e4
LT
2726/* These functions live elsewhere (drivers/net/net_init.c, but related) */
2727
f629d208 2728void ether_setup(struct net_device *dev);
1da177e4
LT
2729
2730/* Support for loadable net-drivers */
f629d208
JP
2731struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2732 void (*setup)(struct net_device *),
2733 unsigned int txqs, unsigned int rxqs);
f25f4e44 2734#define alloc_netdev(sizeof_priv, name, setup) \
36909ea4
TH
2735 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2736
2737#define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2738 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2739
f629d208
JP
2740int register_netdev(struct net_device *dev);
2741void unregister_netdev(struct net_device *dev);
f001fde5 2742
22bedad3 2743/* General hardware address lists handling functions */
f629d208
JP
2744int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2745 struct netdev_hw_addr_list *from_list,
2746 int addr_len, unsigned char addr_type);
2747void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2748 struct netdev_hw_addr_list *from_list,
2749 int addr_len, unsigned char addr_type);
2750int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2751 struct netdev_hw_addr_list *from_list, int addr_len);
2752void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2753 struct netdev_hw_addr_list *from_list, int addr_len);
2754void __hw_addr_flush(struct netdev_hw_addr_list *list);
2755void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 2756
f001fde5 2757/* Functions used for device addresses handling */
f629d208
JP
2758int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2759 unsigned char addr_type);
2760int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2761 unsigned char addr_type);
2762int dev_addr_add_multiple(struct net_device *to_dev,
2763 struct net_device *from_dev, unsigned char addr_type);
2764int dev_addr_del_multiple(struct net_device *to_dev,
2765 struct net_device *from_dev, unsigned char addr_type);
2766void dev_addr_flush(struct net_device *dev);
2767int dev_addr_init(struct net_device *dev);
a748ee24
JP
2768
2769/* Functions used for unicast addresses handling */
f629d208
JP
2770int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2771int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2772int dev_uc_del(struct net_device *dev, const unsigned char *addr);
2773int dev_uc_sync(struct net_device *to, struct net_device *from);
2774int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
2775void dev_uc_unsync(struct net_device *to, struct net_device *from);
2776void dev_uc_flush(struct net_device *dev);
2777void dev_uc_init(struct net_device *dev);
f001fde5 2778
22bedad3 2779/* Functions used for multicast addresses handling */
f629d208
JP
2780int dev_mc_add(struct net_device *dev, const unsigned char *addr);
2781int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
2782int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
2783int dev_mc_del(struct net_device *dev, const unsigned char *addr);
2784int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
2785int dev_mc_sync(struct net_device *to, struct net_device *from);
2786int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
2787void dev_mc_unsync(struct net_device *to, struct net_device *from);
2788void dev_mc_flush(struct net_device *dev);
2789void dev_mc_init(struct net_device *dev);
f001fde5 2790
4417da66 2791/* Functions used for secondary unicast and multicast support */
f629d208
JP
2792void dev_set_rx_mode(struct net_device *dev);
2793void __dev_set_rx_mode(struct net_device *dev);
2794int dev_set_promiscuity(struct net_device *dev, int inc);
2795int dev_set_allmulti(struct net_device *dev, int inc);
2796void netdev_state_change(struct net_device *dev);
2797void netdev_notify_peers(struct net_device *dev);
2798void netdev_features_change(struct net_device *dev);
1da177e4 2799/* Load a device via the kmod */
f629d208
JP
2800void dev_load(struct net *net, const char *name);
2801struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2802 struct rtnl_link_stats64 *storage);
2803void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
2804 const struct net_device_stats *netdev_stats);
eeda3fd6 2805
1da177e4 2806extern int netdev_max_backlog;
3b098e2d 2807extern int netdev_tstamp_prequeue;
1da177e4 2808extern int weight_p;
0a14842f 2809extern int bpf_jit_enable;
9ff162a8 2810
f629d208
JP
2811bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
2812bool netdev_has_any_upper_dev(struct net_device *dev);
2813struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
2814 struct list_head **iter);
8b5be856
VF
2815
2816/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
2817#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
2818 for (iter = &(dev)->all_adj_list.upper, \
2819 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
2820 updev; \
2821 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 2822
f629d208
JP
2823void *netdev_lower_get_next_private(struct net_device *dev,
2824 struct list_head **iter);
2825void *netdev_lower_get_next_private_rcu(struct net_device *dev,
2826 struct list_head **iter);
31088a11
VF
2827
2828#define netdev_for_each_lower_private(dev, priv, iter) \
2829 for (iter = (dev)->adj_list.lower.next, \
2830 priv = netdev_lower_get_next_private(dev, &(iter)); \
2831 priv; \
2832 priv = netdev_lower_get_next_private(dev, &(iter)))
2833
2834#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
2835 for (iter = &(dev)->adj_list.lower, \
2836 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
2837 priv; \
2838 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
2839
f629d208
JP
2840void *netdev_adjacent_get_private(struct list_head *adj_list);
2841struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
2842struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
2843int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
2844int netdev_master_upper_dev_link(struct net_device *dev,
9ff162a8 2845 struct net_device *upper_dev);
f629d208
JP
2846int netdev_master_upper_dev_link_private(struct net_device *dev,
2847 struct net_device *upper_dev,
2848 void *private);
2849void netdev_upper_dev_unlink(struct net_device *dev,
2850 struct net_device *upper_dev);
2851void *netdev_lower_dev_get_private_rcu(struct net_device *dev,
2852 struct net_device *lower_dev);
2853void *netdev_lower_dev_get_private(struct net_device *dev,
2854 struct net_device *lower_dev);
2855int skb_checksum_help(struct sk_buff *skb);
2856struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
2857 netdev_features_t features, bool tx_path);
2858struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
2859 netdev_features_t features);
12b0004d
CW
2860
2861static inline
2862struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
2863{
2864 return __skb_gso_segment(skb, features, true);
2865}
ec5f0615
PS
2866__be16 skb_network_protocol(struct sk_buff *skb);
2867
2868static inline bool can_checksum_protocol(netdev_features_t features,
2869 __be16 protocol)
2870{
2871 return ((features & NETIF_F_GEN_CSUM) ||
2872 ((features & NETIF_F_V4_CSUM) &&
2873 protocol == htons(ETH_P_IP)) ||
2874 ((features & NETIF_F_V6_CSUM) &&
2875 protocol == htons(ETH_P_IPV6)) ||
2876 ((features & NETIF_F_FCOE_CRC) &&
2877 protocol == htons(ETH_P_FCOE)));
2878}
12b0004d 2879
fb286bb2 2880#ifdef CONFIG_BUG
f629d208 2881void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
2882#else
2883static inline void netdev_rx_csum_fault(struct net_device *dev)
2884{
2885}
2886#endif
1da177e4 2887/* rx skb timestamps */
f629d208
JP
2888void net_enable_timestamp(void);
2889void net_disable_timestamp(void);
1da177e4 2890
20380731 2891#ifdef CONFIG_PROC_FS
f629d208 2892int __init dev_proc_init(void);
900ff8c6
CW
2893#else
2894#define dev_proc_init() 0
20380731
ACM
2895#endif
2896
f629d208
JP
2897int netdev_class_create_file(struct class_attribute *class_attr);
2898void netdev_class_remove_file(struct class_attribute *class_attr);
b8a9787e 2899
04600794
JB
2900extern struct kobj_ns_type_operations net_ns_type_operations;
2901
f629d208 2902const char *netdev_drivername(const struct net_device *dev);
6579e57b 2903
f629d208 2904void linkwatch_run_queue(void);
20380731 2905
c8f44aff
MM
2906static inline netdev_features_t netdev_get_wanted_features(
2907 struct net_device *dev)
5455c699
MM
2908{
2909 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2910}
c8f44aff
MM
2911netdev_features_t netdev_increment_features(netdev_features_t all,
2912 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
2913
2914/* Allow TSO being used on stacked device :
2915 * Performing the GSO segmentation before last device
2916 * is a performance improvement.
2917 */
2918static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
2919 netdev_features_t mask)
2920{
2921 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
2922}
2923
6cb6a27c 2924int __netdev_update_features(struct net_device *dev);
5455c699 2925void netdev_update_features(struct net_device *dev);
afe12cc8 2926void netdev_change_features(struct net_device *dev);
7f353bf2 2927
fc4a7489
PM
2928void netif_stacked_transfer_operstate(const struct net_device *rootdev,
2929 struct net_device *dev);
2930
c8f44aff 2931netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 2932
4d29515f 2933static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 2934{
c8f44aff 2935 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
2936
2937 /* check flags correspondence */
2938 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
2939 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
2940 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
2941 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
2942 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
2943 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
2944
d6b4991a 2945 return (features & feature) == feature;
576a30eb
HX
2946}
2947
4d29515f 2948static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 2949{
278b2513 2950 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 2951 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
2952}
2953
4d29515f
DM
2954static inline bool netif_needs_gso(struct sk_buff *skb,
2955 netdev_features_t features)
7967168c 2956{
fc741216 2957 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
2958 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
2959 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
2960}
2961
82cc1a7a
PWJ
2962static inline void netif_set_gso_max_size(struct net_device *dev,
2963 unsigned int size)
2964{
2965 dev->gso_max_size = size;
2966}
2967
8a7fbfab 2968static inline bool netif_is_bond_master(struct net_device *dev)
2969{
2970 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
2971}
2972
4d29515f 2973static inline bool netif_is_bond_slave(struct net_device *dev)
1765a575
JP
2974{
2975 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
2976}
2977
3bdc0eba
BG
2978static inline bool netif_supports_nofcs(struct net_device *dev)
2979{
2980 return dev->priv_flags & IFF_SUPP_NOFCS;
2981}
2982
505d4f73 2983extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 2984
571ba423
JP
2985/* Logging, debugging and troubleshooting/diagnostic helpers. */
2986
2987/* netdev_printk helpers, similar to dev_printk */
2988
2989static inline const char *netdev_name(const struct net_device *dev)
2990{
2991 if (dev->reg_state != NETREG_REGISTERED)
2992 return "(unregistered net_device)";
2993 return dev->name;
2994}
2995
f629d208 2996__printf(3, 4)
b9075fa9
JP
2997int netdev_printk(const char *level, const struct net_device *dev,
2998 const char *format, ...);
f629d208 2999__printf(2, 3)
b9075fa9 3000int netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 3001__printf(2, 3)
b9075fa9 3002int netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 3003__printf(2, 3)
b9075fa9 3004int netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 3005__printf(2, 3)
b9075fa9 3006int netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 3007__printf(2, 3)
b9075fa9 3008int netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 3009__printf(2, 3)
b9075fa9 3010int netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 3011__printf(2, 3)
b9075fa9 3012int netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 3013
8909c9ad
VK
3014#define MODULE_ALIAS_NETDEV(device) \
3015 MODULE_ALIAS("netdev-" device)
3016
b558c96f 3017#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
3018#define netdev_dbg(__dev, format, args...) \
3019do { \
ffa10cb4 3020 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 3021} while (0)
b558c96f
JC
3022#elif defined(DEBUG)
3023#define netdev_dbg(__dev, format, args...) \
3024 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
3025#else
3026#define netdev_dbg(__dev, format, args...) \
3027({ \
3028 if (0) \
3029 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3030 0; \
3031})
3032#endif
3033
3034#if defined(VERBOSE_DEBUG)
3035#define netdev_vdbg netdev_dbg
3036#else
3037
3038#define netdev_vdbg(dev, format, args...) \
3039({ \
3040 if (0) \
3041 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3042 0; \
3043})
3044#endif
3045
3046/*
3047 * netdev_WARN() acts like dev_printk(), but with the key difference
3048 * of using a WARN/WARN_ON to get the message out, including the
3049 * file/line information and a backtrace.
3050 */
3051#define netdev_WARN(dev, format, args...) \
7cc7c5e5 3052 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
571ba423 3053
b3d95c5c
JP
3054/* netif printk helpers, similar to netdev_printk */
3055
3056#define netif_printk(priv, type, level, dev, fmt, args...) \
3057do { \
3058 if (netif_msg_##type(priv)) \
3059 netdev_printk(level, (dev), fmt, ##args); \
3060} while (0)
3061
f45f4321
JP
3062#define netif_level(level, priv, type, dev, fmt, args...) \
3063do { \
3064 if (netif_msg_##type(priv)) \
3065 netdev_##level(dev, fmt, ##args); \
3066} while (0)
3067
b3d95c5c 3068#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 3069 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 3070#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 3071 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 3072#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 3073 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 3074#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 3075 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 3076#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 3077 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 3078#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 3079 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 3080#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 3081 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 3082
0053ea9c 3083#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
3084#define netif_dbg(priv, type, netdev, format, args...) \
3085do { \
3086 if (netif_msg_##type(priv)) \
b5fb0a03 3087 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 3088} while (0)
0053ea9c
JP
3089#elif defined(DEBUG)
3090#define netif_dbg(priv, type, dev, format, args...) \
3091 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
3092#else
3093#define netif_dbg(priv, type, dev, format, args...) \
3094({ \
3095 if (0) \
3096 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3097 0; \
3098})
3099#endif
3100
3101#if defined(VERBOSE_DEBUG)
bcfcc450 3102#define netif_vdbg netif_dbg
b3d95c5c
JP
3103#else
3104#define netif_vdbg(priv, type, dev, format, args...) \
3105({ \
3106 if (0) \
a4ed89cb 3107 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
3108 0; \
3109})
3110#endif
571ba423 3111
900ff8c6
CW
3112/*
3113 * The list of packet types we will receive (as opposed to discard)
3114 * and the routines to invoke.
3115 *
3116 * Why 16. Because with 16 the only overlap we get on a hash of the
3117 * low nibble of the protocol value is RARP/SNAP/X.25.
3118 *
3119 * NOTE: That is no longer true with the addition of VLAN tags. Not
3120 * sure which should go first, but I bet it won't make much
3121 * difference if we are running VLANs. The good news is that
3122 * this protocol won't be in the list unless compiled in, so
3123 * the average user (w/out VLANs) will not be adversely affected.
3124 * --BLG
3125 *
3126 * 0800 IP
3127 * 8100 802.1Q VLAN
3128 * 0001 802.3
3129 * 0002 AX.25
3130 * 0004 802.2
3131 * 8035 RARP
3132 * 0005 SNAP
3133 * 0805 X.25
3134 * 0806 ARP
3135 * 8137 IPX
3136 * 0009 Localtalk
3137 * 86DD IPv6
3138 */
3139#define PTYPE_HASH_SIZE (16)
3140#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3141
385a154c 3142#endif /* _LINUX_NETDEVICE_H */