]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/netdevice.h
tipc: improve naming and comment consistency in media layer
[mirror_ubuntu-bionic-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
e8db0be1 28#include <linux/pm_qos.h>
d7fe0f24 29#include <linux/timer.h>
187f1882 30#include <linux/bug.h>
bea3348e 31#include <linux/delay.h>
60063497 32#include <linux/atomic.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
cf85d08f 44#include <net/dsa.h>
7a6b6f51 45#ifdef CONFIG_DCB
2f90b865
AD
46#include <net/dcbnl.h>
47#endif
5bc1421e 48#include <net/netprio_cgroup.h>
a050c33f 49
a59e2ecb 50#include <linux/netdev_features.h>
77162022 51#include <linux/neighbour.h>
607ca46e 52#include <uapi/linux/netdevice.h>
a59e2ecb 53
115c1d6e 54struct netpoll_info;
313162d0 55struct device;
c1f19b51 56struct phy_device;
704232c2
JB
57/* 802.11 specific */
58struct wireless_dev;
1da177e4
LT
59 /* source back-compat hooks */
60#define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
62
f629d208
JP
63void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct ethtool_ops *ops);
d07d7507 65
c1f79426
SA
66/* hardware address assignment types */
67#define NET_ADDR_PERM 0 /* address is permanent (default) */
68#define NET_ADDR_RANDOM 1 /* address is generated randomly */
69#define NET_ADDR_STOLEN 2 /* address is stolen from other device */
fbdeca2d
JP
70#define NET_ADDR_SET 3 /* address is set using
71 * dev_set_mac_address() */
c1f79426 72
9a1654ba
JP
73/* Backlog congestion levels */
74#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75#define NET_RX_DROP 1 /* packet dropped */
76
572a9d7b
PM
77/*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously, in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), all
91 * others are propagated to higher layers.
92 */
93
94/* qdisc ->enqueue() return codes. */
95#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
96#define NET_XMIT_DROP 0x01 /* skb dropped */
97#define NET_XMIT_CN 0x02 /* congestion notification */
98#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
99#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 100
b9df3cb8
GR
101/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 104#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
105#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
dc1f8bf6 107/* Driver transmit return codes */
9a1654ba 108#define NETDEV_TX_MASK 0xf0
572a9d7b 109
dc1f8bf6 110enum netdev_tx {
572a9d7b 111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
dc1f8bf6
SH
115};
116typedef enum netdev_tx netdev_tx_t;
117
9a1654ba
JP
118/*
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121 */
122static inline bool dev_xmit_complete(int rc)
123{
124 /*
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 */
130 if (likely(rc < NET_XMIT_MASK))
131 return true;
132
133 return false;
134}
135
1da177e4
LT
136/*
137 * Compute the worst case header length according to the protocols
138 * used.
139 */
fe2918b0 140
d11ead75 141#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
142# if defined(CONFIG_MAC80211_MESH)
143# define LL_MAX_HEADER 128
144# else
145# define LL_MAX_HEADER 96
146# endif
1da177e4 147#else
8388e3da 148# define LL_MAX_HEADER 32
1da177e4
LT
149#endif
150
d11ead75
BH
151#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
153#define MAX_HEADER LL_MAX_HEADER
154#else
155#define MAX_HEADER (LL_MAX_HEADER + 48)
156#endif
157
158/*
be1f3c2c
BH
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
1da177e4 161 */
fe2918b0 162
d94d9fee 163struct net_device_stats {
3cfde79c
BH
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
1da177e4 173 unsigned long collisions;
1da177e4 174 unsigned long rx_length_errors;
3cfde79c
BH
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
1da177e4
LT
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
1da177e4
LT
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187};
188
1da177e4
LT
189
190#include <linux/cache.h>
191#include <linux/skbuff.h>
192
adc9300e 193#ifdef CONFIG_RPS
c5905afb
IM
194#include <linux/static_key.h>
195extern struct static_key rps_needed;
adc9300e
ED
196#endif
197
1da177e4
LT
198struct neighbour;
199struct neigh_parms;
200struct sk_buff;
201
f001fde5
JP
202struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
205 unsigned char type;
ccffad25
JP
206#define NETDEV_HW_ADDR_T_LAN 1
207#define NETDEV_HW_ADDR_T_SAN 2
208#define NETDEV_HW_ADDR_T_SLAVE 3
209#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 210#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 211 bool global_use;
4cd729b0 212 int sync_cnt;
8f8f103d 213 int refcount;
4543fbef 214 int synced;
f001fde5
JP
215 struct rcu_head rcu_head;
216};
217
31278e71
JP
218struct netdev_hw_addr_list {
219 struct list_head list;
220 int count;
221};
222
22bedad3
JP
223#define netdev_hw_addr_list_count(l) ((l)->count)
224#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225#define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 227
22bedad3
JP
228#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230#define netdev_for_each_uc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 232
22bedad3
JP
233#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 235#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 236 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 237
d94d9fee 238struct hh_cache {
f6b72b62 239 u16 hh_len;
5c25f686 240 u16 __pad;
3644f0ce 241 seqlock_t hh_lock;
1da177e4
LT
242
243 /* cached hardware header; allow for machine alignment needs. */
244#define HH_DATA_MOD 16
245#define HH_DATA_OFF(__len) \
5ba0eac6 246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
247#define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250};
251
252/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260#define LL_RESERVED_SPACE(dev) \
f5184d26 261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 262#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 264
3b04ddde
SH
265struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
95c96174 268 const void *saddr, unsigned int len);
3b04ddde
SH
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*rebuild)(struct sk_buff *skb);
e69dd336 271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
275};
276
1da177e4
LT
277/* These flag bits are private to the generic network queueing
278 * layer, they may not be explicitly referenced by any other
279 * code.
280 */
281
d94d9fee 282enum netdev_state_t {
1da177e4
LT
283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
1da177e4 285 __LINK_STATE_NOCARRIER,
b00055aa
SR
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
1da177e4
LT
288};
289
290
291/*
292 * This structure holds at boot time configured netdevice settings. They
fe2918b0 293 * are then used in the device probing.
1da177e4
LT
294 */
295struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298};
299#define NETDEV_BOOT_SETUP_MAX 8
300
f629d208 301int __init netdev_boot_setup(char *str);
1da177e4 302
bea3348e
SH
303/*
304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-cpu poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
404f7c9e 317 unsigned int gro_count;
bea3348e
SH
318 int (*poll)(struct napi_struct *, int);
319#ifdef CONFIG_NETPOLL
320 spinlock_t poll_lock;
321 int poll_owner;
bea3348e 322#endif
5d38a079 323 struct net_device *dev;
d565b0a1 324 struct sk_buff *gro_list;
5d38a079 325 struct sk_buff *skb;
404f7c9e 326 struct list_head dev_list;
af12fa6e
ET
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
bea3348e
SH
329};
330
d94d9fee 331enum {
bea3348e 332 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 333 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
af12fa6e 335 NAPI_STATE_HASHED, /* In NAPI hash */
bea3348e
SH
336};
337
5b252f0c 338enum gro_result {
d1c76af9
HX
339 GRO_MERGED,
340 GRO_MERGED_FREE,
341 GRO_HELD,
342 GRO_NORMAL,
343 GRO_DROP,
344};
5b252f0c 345typedef enum gro_result gro_result_t;
d1c76af9 346
8a4eb573
JP
347/*
348 * enum rx_handler_result - Possible return values for rx_handlers.
349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
350 * further.
351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
352 * case skb->dev was changed by rx_handler.
353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
354 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
355 *
356 * rx_handlers are functions called from inside __netif_receive_skb(), to do
357 * special processing of the skb, prior to delivery to protocol handlers.
358 *
359 * Currently, a net_device can only have a single rx_handler registered. Trying
360 * to register a second rx_handler will return -EBUSY.
361 *
362 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
363 * To unregister a rx_handler on a net_device, use
364 * netdev_rx_handler_unregister().
365 *
366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
367 * do with the skb.
368 *
369 * If the rx_handler consumed to skb in some way, it should return
370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
371 * the skb to be delivered in some other ways.
372 *
373 * If the rx_handler changed skb->dev, to divert the skb to another
374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
375 * new device will be called if it exists.
376 *
377 * If the rx_handler consider the skb should be ignored, it should return
378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 379 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
380 *
381 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
382 * delivered, it should return RX_HANDLER_PASS.
383 *
384 * A device without a registered rx_handler will behave as if rx_handler
385 * returned RX_HANDLER_PASS.
386 */
387
388enum rx_handler_result {
389 RX_HANDLER_CONSUMED,
390 RX_HANDLER_ANOTHER,
391 RX_HANDLER_EXACT,
392 RX_HANDLER_PASS,
393};
394typedef enum rx_handler_result rx_handler_result_t;
395typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 396
f629d208 397void __napi_schedule(struct napi_struct *n);
bea3348e 398
4d29515f 399static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
400{
401 return test_bit(NAPI_STATE_DISABLE, &n->state);
402}
403
bea3348e
SH
404/**
405 * napi_schedule_prep - check if napi can be scheduled
406 * @n: napi context
407 *
408 * Test if NAPI routine is already running, and if not mark
409 * it as running. This is used as a condition variable
a0a46196
DM
410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
bea3348e 412 */
4d29515f 413static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 414{
a0a46196
DM
415 return !napi_disable_pending(n) &&
416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
417}
418
419/**
420 * napi_schedule - schedule NAPI poll
421 * @n: napi context
422 *
423 * Schedule NAPI poll routine to be called if it is not already
424 * running.
425 */
426static inline void napi_schedule(struct napi_struct *n)
427{
428 if (napi_schedule_prep(n))
429 __napi_schedule(n);
430}
431
bfe13f54 432/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 433static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
434{
435 if (napi_schedule_prep(napi)) {
436 __napi_schedule(napi);
4d29515f 437 return true;
bfe13f54 438 }
4d29515f 439 return false;
bfe13f54
RD
440}
441
bea3348e
SH
442/**
443 * napi_complete - NAPI processing complete
444 * @n: napi context
445 *
446 * Mark NAPI processing as complete.
447 */
f629d208
JP
448void __napi_complete(struct napi_struct *n);
449void napi_complete(struct napi_struct *n);
bea3348e 450
af12fa6e
ET
451/**
452 * napi_by_id - lookup a NAPI by napi_id
453 * @napi_id: hashed napi_id
454 *
455 * lookup @napi_id in napi_hash table
456 * must be called under rcu_read_lock()
457 */
f629d208 458struct napi_struct *napi_by_id(unsigned int napi_id);
af12fa6e
ET
459
460/**
461 * napi_hash_add - add a NAPI to global hashtable
462 * @napi: napi context
463 *
464 * generate a new napi_id and store a @napi under it in napi_hash
465 */
f629d208 466void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
467
468/**
469 * napi_hash_del - remove a NAPI from global table
470 * @napi: napi context
471 *
472 * Warning: caller must observe rcu grace period
473 * before freeing memory containing @napi
474 */
f629d208 475void napi_hash_del(struct napi_struct *napi);
af12fa6e 476
bea3348e
SH
477/**
478 * napi_disable - prevent NAPI from scheduling
479 * @n: napi context
480 *
481 * Stop NAPI from being scheduled on this context.
482 * Waits till any outstanding processing completes.
483 */
484static inline void napi_disable(struct napi_struct *n)
485{
80c33ddd 486 might_sleep();
a0a46196 487 set_bit(NAPI_STATE_DISABLE, &n->state);
bea3348e 488 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
43cc7380 489 msleep(1);
a0a46196 490 clear_bit(NAPI_STATE_DISABLE, &n->state);
bea3348e
SH
491}
492
493/**
494 * napi_enable - enable NAPI scheduling
495 * @n: napi context
496 *
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
499 */
500static inline void napi_enable(struct napi_struct *n)
501{
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
503 smp_mb__before_clear_bit();
504 clear_bit(NAPI_STATE_SCHED, &n->state);
505}
506
c264c3de
SH
507#ifdef CONFIG_SMP
508/**
509 * napi_synchronize - wait until NAPI is not running
510 * @n: napi context
511 *
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
515 */
516static inline void napi_synchronize(const struct napi_struct *n)
517{
518 while (test_bit(NAPI_STATE_SCHED, &n->state))
519 msleep(1);
520}
521#else
522# define napi_synchronize(n) barrier()
523#endif
524
d94d9fee 525enum netdev_queue_state_t {
73466498
TH
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
c3f26a26 528 __QUEUE_STATE_FROZEN,
73466498
TH
529#define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
530 (1 << __QUEUE_STATE_STACK_XOFF))
531#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
532 (1 << __QUEUE_STATE_FROZEN))
79d16385 533};
73466498
TH
534/*
535 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
536 * netif_tx_* functions below are used to manipulate this flag. The
537 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
538 * queue independently. The netif_xmit_*stopped functions below are called
539 * to check if the queue has been stopped by the driver or stack (either
540 * of the XOFF bits are set in the state). Drivers should not need to call
541 * netif_xmit*stopped functions, they should only be using netif_tx_*.
542 */
79d16385 543
bb949fbd 544struct netdev_queue {
6a321cb3
ED
545/*
546 * read mostly part
547 */
bb949fbd 548 struct net_device *dev;
b0e1e646
DM
549 struct Qdisc *qdisc;
550 struct Qdisc *qdisc_sleeping;
ccf5ff69 551#ifdef CONFIG_SYSFS
1d24eb48
TH
552 struct kobject kobj;
553#endif
f2cd2d3e
ED
554#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
555 int numa_node;
556#endif
6a321cb3
ED
557/*
558 * write mostly part
559 */
560 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
561 int xmit_lock_owner;
9d21493b
ED
562 /*
563 * please use this field instead of dev->trans_start
564 */
565 unsigned long trans_start;
ccf5ff69 566
567 /*
568 * Number of TX timeouts for this queue
569 * (/sys/class/net/DEV/Q/trans_timeout)
570 */
571 unsigned long trans_timeout;
114cf580
TH
572
573 unsigned long state;
574
575#ifdef CONFIG_BQL
576 struct dql dql;
577#endif
e8a0464c 578} ____cacheline_aligned_in_smp;
bb949fbd 579
f2cd2d3e
ED
580static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
581{
582#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
583 return q->numa_node;
584#else
b236da69 585 return NUMA_NO_NODE;
f2cd2d3e
ED
586#endif
587}
588
589static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
590{
591#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
592 q->numa_node = node;
593#endif
594}
595
df334545 596#ifdef CONFIG_RPS
0a9627f2
TH
597/*
598 * This structure holds an RPS map which can be of variable length. The
599 * map is an array of CPUs.
600 */
601struct rps_map {
602 unsigned int len;
603 struct rcu_head rcu;
604 u16 cpus[0];
605};
60b778ce 606#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 607
fec5e652 608/*
c445477d
BH
609 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
610 * tail pointer for that CPU's input queue at the time of last enqueue, and
611 * a hardware filter index.
fec5e652
TH
612 */
613struct rps_dev_flow {
614 u16 cpu;
c445477d 615 u16 filter;
fec5e652
TH
616 unsigned int last_qtail;
617};
c445477d 618#define RPS_NO_FILTER 0xffff
fec5e652
TH
619
620/*
621 * The rps_dev_flow_table structure contains a table of flow mappings.
622 */
623struct rps_dev_flow_table {
624 unsigned int mask;
625 struct rcu_head rcu;
fec5e652
TH
626 struct rps_dev_flow flows[0];
627};
628#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 629 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
630
631/*
632 * The rps_sock_flow_table contains mappings of flows to the last CPU
633 * on which they were processed by the application (set in recvmsg).
634 */
635struct rps_sock_flow_table {
636 unsigned int mask;
637 u16 ents[0];
638};
639#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
60b778ce 640 ((_num) * sizeof(u16)))
fec5e652
TH
641
642#define RPS_NO_CPU 0xffff
643
644static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
645 u32 hash)
646{
647 if (table && hash) {
648 unsigned int cpu, index = hash & table->mask;
649
650 /* We only give a hint, preemption can change cpu under us */
651 cpu = raw_smp_processor_id();
652
653 if (table->ents[index] != cpu)
654 table->ents[index] = cpu;
655 }
656}
657
658static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
659 u32 hash)
660{
661 if (table && hash)
662 table->ents[hash & table->mask] = RPS_NO_CPU;
663}
664
6e3f7faf 665extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
fec5e652 666
c445477d 667#ifdef CONFIG_RFS_ACCEL
f629d208
JP
668bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
669 u16 filter_id);
c445477d
BH
670#endif
671
0a9627f2
TH
672/* This structure contains an instance of an RX queue. */
673struct netdev_rx_queue {
6e3f7faf
ED
674 struct rps_map __rcu *rps_map;
675 struct rps_dev_flow_table __rcu *rps_flow_table;
676 struct kobject kobj;
fe822240 677 struct net_device *dev;
0a9627f2 678} ____cacheline_aligned_in_smp;
fec5e652 679#endif /* CONFIG_RPS */
d314774c 680
bf264145
TH
681#ifdef CONFIG_XPS
682/*
683 * This structure holds an XPS map which can be of variable length. The
684 * map is an array of queues.
685 */
686struct xps_map {
687 unsigned int len;
688 unsigned int alloc_len;
689 struct rcu_head rcu;
690 u16 queues[0];
691};
60b778ce 692#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
bf264145
TH
693#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
694 / sizeof(u16))
695
696/*
697 * This structure holds all XPS maps for device. Maps are indexed by CPU.
698 */
699struct xps_dev_maps {
700 struct rcu_head rcu;
a4177869 701 struct xps_map __rcu *cpu_map[0];
bf264145
TH
702};
703#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
704 (nr_cpu_ids * sizeof(struct xps_map *)))
705#endif /* CONFIG_XPS */
706
4f57c087
JF
707#define TC_MAX_QUEUE 16
708#define TC_BITMASK 15
709/* HW offloaded queuing disciplines txq count and offset maps */
710struct netdev_tc_txq {
711 u16 count;
712 u16 offset;
713};
714
68bad94e
NP
715#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
716/*
717 * This structure is to hold information about the device
718 * configured to run FCoE protocol stack.
719 */
720struct netdev_fcoe_hbainfo {
721 char manufacturer[64];
722 char serial_number[64];
723 char hardware_version[64];
724 char driver_version[64];
725 char optionrom_version[64];
726 char firmware_version[64];
727 char model[256];
728 char model_description[256];
729};
730#endif
731
66b52b0d
JP
732#define MAX_PHYS_PORT_ID_LEN 32
733
734/* This structure holds a unique identifier to identify the
735 * physical port used by a netdevice.
736 */
737struct netdev_phys_port_id {
738 unsigned char id[MAX_PHYS_PORT_ID_LEN];
739 unsigned char id_len;
740};
741
d314774c
SH
742/*
743 * This structure defines the management hooks for network devices.
00829823
SH
744 * The following hooks can be defined; unless noted otherwise, they are
745 * optional and can be filled with a null pointer.
d314774c
SH
746 *
747 * int (*ndo_init)(struct net_device *dev);
748 * This function is called once when network device is registered.
749 * The network device can use this to any late stage initializaton
750 * or semantic validattion. It can fail with an error code which will
751 * be propogated back to register_netdev
752 *
753 * void (*ndo_uninit)(struct net_device *dev);
754 * This function is called when device is unregistered or when registration
755 * fails. It is not called if init fails.
756 *
757 * int (*ndo_open)(struct net_device *dev);
758 * This function is called when network device transistions to the up
759 * state.
760 *
761 * int (*ndo_stop)(struct net_device *dev);
762 * This function is called when network device transistions to the down
763 * state.
764 *
dc1f8bf6
SH
765 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
766 * struct net_device *dev);
00829823 767 * Called when a packet needs to be transmitted.
dc1f8bf6
SH
768 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
769 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
770 * Required can not be NULL.
771 *
772 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
773 * Called to decide which queue to when device supports multiple
774 * transmit queues.
775 *
d314774c
SH
776 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
777 * This function is called to allow device receiver to make
778 * changes to configuration when multicast or promiscious is enabled.
779 *
780 * void (*ndo_set_rx_mode)(struct net_device *dev);
781 * This function is called device changes address list filtering.
01789349
JP
782 * If driver handles unicast address filtering, it should set
783 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
784 *
785 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
786 * This function is called when the Media Access Control address
37b607c5 787 * needs to be changed. If this interface is not defined, the
d314774c
SH
788 * mac address can not be changed.
789 *
790 * int (*ndo_validate_addr)(struct net_device *dev);
791 * Test if Media Access Control address is valid for the device.
792 *
793 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
794 * Called when a user request an ioctl which can't be handled by
795 * the generic interface code. If not defined ioctl's return
796 * not supported error code.
797 *
798 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
799 * Used to set network devices bus interface parameters. This interface
800 * is retained for legacy reason, new devices should use the bus
801 * interface (PCI) for low level management.
802 *
803 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
804 * Called when a user wants to change the Maximum Transfer Unit
805 * of a device. If not defined, any request to change MTU will
806 * will return an error.
807 *
00829823 808 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
809 * Callback uses when the transmitter has not made any progress
810 * for dev->watchdog ticks.
811 *
3cfde79c 812 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 813 * struct rtnl_link_stats64 *storage);
d308e38f 814 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 815 * Called when a user wants to get the network device usage
be1f3c2c 816 * statistics. Drivers must do one of the following:
3cfde79c
BH
817 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
818 * rtnl_link_stats64 structure passed by the caller.
82695d9b 819 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
820 * (which should normally be dev->stats) and return a pointer to
821 * it. The structure may be changed asynchronously only if each
822 * field is written atomically.
823 * 3. Update dev->stats asynchronously and atomically, and define
824 * neither operation.
d314774c 825 *
80d5c368
PM
826 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
827 * If device support VLAN filtering this function is called when a
828 * VLAN id is registered.
d314774c 829 *
8e586137 830 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
80d5c368
PM
831 * If device support VLAN filtering this function is called when a
832 * VLAN id is unregistered.
d314774c
SH
833 *
834 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
835 *
836 * SR-IOV management functions.
837 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
838 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
839 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
5f8444a3 840 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
841 * int (*ndo_get_vf_config)(struct net_device *dev,
842 * int vf, struct ifla_vf_info *ivf);
1d8faf48 843 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
844 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
845 * struct nlattr *port[]);
846 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
847 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
848 * Called to setup 'tc' number of traffic classes in the net device. This
849 * is always called from the stack with the rtnl lock held and netif tx
850 * queues stopped. This allows the netdevice to perform queue management
851 * safely.
c445477d 852 *
e9bce845
YZ
853 * Fiber Channel over Ethernet (FCoE) offload functions.
854 * int (*ndo_fcoe_enable)(struct net_device *dev);
855 * Called when the FCoE protocol stack wants to start using LLD for FCoE
856 * so the underlying device can perform whatever needed configuration or
857 * initialization to support acceleration of FCoE traffic.
858 *
859 * int (*ndo_fcoe_disable)(struct net_device *dev);
860 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
861 * so the underlying device can perform whatever needed clean-ups to
862 * stop supporting acceleration of FCoE traffic.
863 *
864 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
865 * struct scatterlist *sgl, unsigned int sgc);
866 * Called when the FCoE Initiator wants to initialize an I/O that
867 * is a possible candidate for Direct Data Placement (DDP). The LLD can
868 * perform necessary setup and returns 1 to indicate the device is set up
869 * successfully to perform DDP on this I/O, otherwise this returns 0.
870 *
871 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
872 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
873 * indicated by the FC exchange id 'xid', so the underlying device can
874 * clean up and reuse resources for later DDP requests.
875 *
876 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
877 * struct scatterlist *sgl, unsigned int sgc);
878 * Called when the FCoE Target wants to initialize an I/O that
879 * is a possible candidate for Direct Data Placement (DDP). The LLD can
880 * perform necessary setup and returns 1 to indicate the device is set up
881 * successfully to perform DDP on this I/O, otherwise this returns 0.
882 *
68bad94e
NP
883 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
884 * struct netdev_fcoe_hbainfo *hbainfo);
885 * Called when the FCoE Protocol stack wants information on the underlying
886 * device. This information is utilized by the FCoE protocol stack to
887 * register attributes with Fiber Channel management service as per the
888 * FC-GS Fabric Device Management Information(FDMI) specification.
889 *
e9bce845
YZ
890 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
891 * Called when the underlying device wants to override default World Wide
892 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
893 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
894 * protocol stack to use.
895 *
c445477d
BH
896 * RFS acceleration.
897 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
898 * u16 rxq_index, u32 flow_id);
899 * Set hardware filter for RFS. rxq_index is the target queue index;
900 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
901 * Return the filter ID on success, or a negative error code.
fbaec0ea 902 *
8b98a70c 903 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
904 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
905 * Called to make another netdev an underling.
906 *
907 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
908 * Called to release previously enslaved netdev.
5455c699
MM
909 *
910 * Feature/offload setting functions.
c8f44aff
MM
911 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
912 * netdev_features_t features);
5455c699
MM
913 * Adjusts the requested feature flags according to device-specific
914 * constraints, and returns the resulting flags. Must not modify
915 * the device state.
916 *
c8f44aff 917 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
918 * Called to update device configuration to new features. Passed
919 * feature set might be less than what was returned by ndo_fix_features()).
920 * Must return >0 or -errno if it changed dev->features itself.
921 *
edc7d573 922 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
923 * struct net_device *dev,
6b6e2725 924 * const unsigned char *addr, u16 flags)
77162022 925 * Adds an FDB entry to dev for addr.
1690be63
VY
926 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
927 * struct net_device *dev,
6b6e2725 928 * const unsigned char *addr)
77162022
JF
929 * Deletes the FDB entry from dev coresponding to addr.
930 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
931 * struct net_device *dev, int idx)
932 * Used to add FDB entries to dump requests. Implementers should add
933 * entries to skb and update idx with the number of entries.
e5a55a89
JF
934 *
935 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
936 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
24f11a5c 937 * struct net_device *dev, u32 filter_mask)
4bf84c35
JP
938 *
939 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
940 * Called to change device carrier. Soft-devices (like dummy, team, etc)
941 * which do not represent real hardware may define this to allow their
942 * userspace components to manage their virtual carrier state. Devices
943 * that determine carrier state from physical hardware properties (eg
944 * network cables) or protocol-dependent mechanisms (eg
945 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
946 *
947 * int (*ndo_get_phys_port_id)(struct net_device *dev,
948 * struct netdev_phys_port_id *ppid);
949 * Called to get ID of physical port of this device. If driver does
950 * not implement this, it is assumed that the hw is not able to have
951 * multiple net devices on single physical port.
53cf5275
JG
952 *
953 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 954 * sa_family_t sa_family, __be16 port);
53cf5275
JG
955 * Called by vxlan to notiy a driver about the UDP port and socket
956 * address family that vxlan is listnening to. It is called only when
957 * a new port starts listening. The operation is protected by the
958 * vxlan_net->sock_lock.
959 *
960 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 961 * sa_family_t sa_family, __be16 port);
53cf5275
JG
962 * Called by vxlan to notify the driver about a UDP port and socket
963 * address family that vxlan is not listening to anymore. The operation
964 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
965 *
966 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
967 * struct net_device *dev)
968 * Called by upper layer devices to accelerate switching or other
969 * station functionality into hardware. 'pdev is the lowerdev
970 * to use for the offload and 'dev' is the net device that will
971 * back the offload. Returns a pointer to the private structure
972 * the upper layer will maintain.
973 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
974 * Called by upper layer device to delete the station created
975 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
976 * the station and priv is the structure returned by the add
977 * operation.
978 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
979 * struct net_device *dev,
980 * void *priv);
981 * Callback to use for xmit over the accelerated station. This
982 * is used in place of ndo_start_xmit on accelerated net
983 * devices.
d314774c
SH
984 */
985struct net_device_ops {
986 int (*ndo_init)(struct net_device *dev);
987 void (*ndo_uninit)(struct net_device *dev);
988 int (*ndo_open)(struct net_device *dev);
989 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 990 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
991 struct net_device *dev);
992 u16 (*ndo_select_queue)(struct net_device *dev,
993 struct sk_buff *skb);
d314774c
SH
994 void (*ndo_change_rx_flags)(struct net_device *dev,
995 int flags);
d314774c 996 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
997 int (*ndo_set_mac_address)(struct net_device *dev,
998 void *addr);
d314774c 999 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1000 int (*ndo_do_ioctl)(struct net_device *dev,
1001 struct ifreq *ifr, int cmd);
d314774c
SH
1002 int (*ndo_set_config)(struct net_device *dev,
1003 struct ifmap *map);
00829823
SH
1004 int (*ndo_change_mtu)(struct net_device *dev,
1005 int new_mtu);
1006 int (*ndo_neigh_setup)(struct net_device *dev,
1007 struct neigh_parms *);
d314774c
SH
1008 void (*ndo_tx_timeout) (struct net_device *dev);
1009
28172739
ED
1010 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1011 struct rtnl_link_stats64 *storage);
d314774c
SH
1012 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1013
8e586137 1014 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1015 __be16 proto, u16 vid);
8e586137 1016 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1017 __be16 proto, u16 vid);
d314774c 1018#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1019 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1020 int (*ndo_netpoll_setup)(struct net_device *dev,
47be03a2
AW
1021 struct netpoll_info *info,
1022 gfp_t gfp);
0e34e931 1023 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1024#endif
e0d1095a 1025#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1026 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1027#endif
95c26df8
WM
1028 int (*ndo_set_vf_mac)(struct net_device *dev,
1029 int queue, u8 *mac);
1030 int (*ndo_set_vf_vlan)(struct net_device *dev,
1031 int queue, u16 vlan, u8 qos);
1032 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
1033 int vf, int rate);
5f8444a3
GR
1034 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1035 int vf, bool setting);
95c26df8
WM
1036 int (*ndo_get_vf_config)(struct net_device *dev,
1037 int vf,
1038 struct ifla_vf_info *ivf);
1d8faf48
RE
1039 int (*ndo_set_vf_link_state)(struct net_device *dev,
1040 int vf, int link_state);
57b61080
SF
1041 int (*ndo_set_vf_port)(struct net_device *dev,
1042 int vf,
1043 struct nlattr *port[]);
1044 int (*ndo_get_vf_port)(struct net_device *dev,
1045 int vf, struct sk_buff *skb);
4f57c087 1046 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
d11ead75 1047#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1048 int (*ndo_fcoe_enable)(struct net_device *dev);
1049 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1050 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1051 u16 xid,
1052 struct scatterlist *sgl,
1053 unsigned int sgc);
1054 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1055 u16 xid);
6247e086
YZ
1056 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1057 u16 xid,
1058 struct scatterlist *sgl,
1059 unsigned int sgc);
68bad94e
NP
1060 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1061 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1062#endif
1063
d11ead75 1064#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1065#define NETDEV_FCOE_WWNN 0
1066#define NETDEV_FCOE_WWPN 1
1067 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1068 u64 *wwn, int type);
4d288d57 1069#endif
3c9c36bc 1070
c445477d
BH
1071#ifdef CONFIG_RFS_ACCEL
1072 int (*ndo_rx_flow_steer)(struct net_device *dev,
1073 const struct sk_buff *skb,
1074 u16 rxq_index,
1075 u32 flow_id);
1076#endif
fbaec0ea
JP
1077 int (*ndo_add_slave)(struct net_device *dev,
1078 struct net_device *slave_dev);
1079 int (*ndo_del_slave)(struct net_device *dev,
1080 struct net_device *slave_dev);
c8f44aff
MM
1081 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1082 netdev_features_t features);
5455c699 1083 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1084 netdev_features_t features);
da6a8fa0 1085 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1086 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1087
1088 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1089 struct nlattr *tb[],
77162022 1090 struct net_device *dev,
6b6e2725 1091 const unsigned char *addr,
77162022
JF
1092 u16 flags);
1093 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1094 struct nlattr *tb[],
77162022 1095 struct net_device *dev,
6b6e2725 1096 const unsigned char *addr);
77162022
JF
1097 int (*ndo_fdb_dump)(struct sk_buff *skb,
1098 struct netlink_callback *cb,
1099 struct net_device *dev,
1100 int idx);
e5a55a89
JF
1101
1102 int (*ndo_bridge_setlink)(struct net_device *dev,
1103 struct nlmsghdr *nlh);
1104 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1105 u32 pid, u32 seq,
6cbdceeb
VY
1106 struct net_device *dev,
1107 u32 filter_mask);
407af329
VY
1108 int (*ndo_bridge_dellink)(struct net_device *dev,
1109 struct nlmsghdr *nlh);
4bf84c35
JP
1110 int (*ndo_change_carrier)(struct net_device *dev,
1111 bool new_carrier);
66b52b0d
JP
1112 int (*ndo_get_phys_port_id)(struct net_device *dev,
1113 struct netdev_phys_port_id *ppid);
53cf5275
JG
1114 void (*ndo_add_vxlan_port)(struct net_device *dev,
1115 sa_family_t sa_family,
35e42379 1116 __be16 port);
53cf5275
JG
1117 void (*ndo_del_vxlan_port)(struct net_device *dev,
1118 sa_family_t sa_family,
35e42379 1119 __be16 port);
a6cc0cfa
JF
1120
1121 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1122 struct net_device *dev);
1123 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1124 void *priv);
1125
1126 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1127 struct net_device *dev,
1128 void *priv);
d314774c
SH
1129};
1130
1da177e4
LT
1131/*
1132 * The DEVICE structure.
1133 * Actually, this whole structure is a big mistake. It mixes I/O
1134 * data with strictly "high-level" data, and it has to know about
1135 * almost every data structure used in the INET module.
1136 *
1137 * FIXME: cleanup struct net_device such that network protocol info
1138 * moves out.
1139 */
1140
d94d9fee 1141struct net_device {
1da177e4
LT
1142
1143 /*
1144 * This is the first field of the "visible" part of this structure
1145 * (i.e. as seen by users in the "Space.c" file). It is the name
724df615 1146 * of the interface.
1da177e4
LT
1147 */
1148 char name[IFNAMSIZ];
ed77134b 1149
9136461a 1150 /* device name hash chain, please keep it close to name[] */
9356b8fc 1151 struct hlist_node name_hlist;
9136461a 1152
0b815a1a
SH
1153 /* snmp alias */
1154 char *ifalias;
1da177e4
LT
1155
1156 /*
1157 * I/O specific fields
1158 * FIXME: Merge these and struct ifmap into one
1159 */
1160 unsigned long mem_end; /* shared mem end */
1161 unsigned long mem_start; /* shared mem start */
1162 unsigned long base_addr; /* device I/O address */
df42153c 1163 int irq; /* device IRQ number */
1da177e4
LT
1164
1165 /*
1166 * Some hardware also needs these fields, but they are not
1167 * part of the usual set specified in Space.c.
1168 */
1169
1da177e4
LT
1170 unsigned long state;
1171
7562f876 1172 struct list_head dev_list;
bea3348e 1173 struct list_head napi_list;
44a0873d 1174 struct list_head unreg_list;
5cde2829 1175 struct list_head close_list;
2f268f12
VF
1176
1177 /* directly linked devices, like slaves for bonding */
1178 struct {
1179 struct list_head upper;
1180 struct list_head lower;
1181 } adj_list;
1182
1183 /* all linked devices, *including* neighbours */
1184 struct {
1185 struct list_head upper;
1186 struct list_head lower;
1187 } all_adj_list;
4c3d5e7b 1188
1da177e4 1189
5455c699 1190 /* currently active device features */
c8f44aff 1191 netdev_features_t features;
5455c699 1192 /* user-changeable features */
c8f44aff 1193 netdev_features_t hw_features;
5455c699 1194 /* user-requested features */
c8f44aff 1195 netdev_features_t wanted_features;
1aac6267 1196 /* mask of features inheritable by VLAN devices */
c8f44aff 1197 netdev_features_t vlan_features;
6a674e9c
JG
1198 /* mask of features inherited by encapsulating devices
1199 * This field indicates what encapsulation offloads
1200 * the hardware is capable of doing, and drivers will
1201 * need to set them appropriately.
1202 */
1203 netdev_features_t hw_enc_features;
0d89d203
SH
1204 /* mask of fetures inheritable by MPLS */
1205 netdev_features_t mpls_features;
04ed3e74 1206
1da177e4
LT
1207 /* Interface index. Unique device identifier */
1208 int ifindex;
1209 int iflink;
1210
c45d286e 1211 struct net_device_stats stats;
caf586e5
ED
1212 atomic_long_t rx_dropped; /* dropped packets by core network
1213 * Do not use this in drivers.
1214 */
1da177e4 1215
b86e0280 1216#ifdef CONFIG_WIRELESS_EXT
1da177e4
LT
1217 /* List of functions to handle Wireless Extensions (instead of ioctl).
1218 * See <net/iw_handler.h> for details. Jean II */
1219 const struct iw_handler_def * wireless_handlers;
1220 /* Instance data managed by the core of Wireless Extensions. */
1221 struct iw_public_data * wireless_data;
b86e0280 1222#endif
d314774c
SH
1223 /* Management operations */
1224 const struct net_device_ops *netdev_ops;
76fd8593 1225 const struct ethtool_ops *ethtool_ops;
a6cc0cfa 1226 const struct forwarding_accel_ops *fwd_ops;
1da177e4 1227
3b04ddde
SH
1228 /* Hardware header description */
1229 const struct header_ops *header_ops;
1230
b00055aa 1231 unsigned int flags; /* interface flags (a la BSD) */
3bdc0eba
BG
1232 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1233 * See if.h for definitions. */
1da177e4 1234 unsigned short gflags;
1da177e4
LT
1235 unsigned short padded; /* How much padding added by alloc_netdev() */
1236
b00055aa
SR
1237 unsigned char operstate; /* RFC2863 operstate */
1238 unsigned char link_mode; /* mapping policy to operstate */
1239
bdc220da
JP
1240 unsigned char if_port; /* Selectable AUI, TP,..*/
1241 unsigned char dma; /* DMA channel */
1242
cd7b5396 1243 unsigned int mtu; /* interface MTU value */
1da177e4
LT
1244 unsigned short type; /* interface hardware type */
1245 unsigned short hard_header_len; /* hardware hdr length */
1da177e4 1246
f5184d26
JB
1247 /* extra head- and tailroom the hardware may need, but not in all cases
1248 * can this be guaranteed, especially tailroom. Some cases also use
1249 * LL_MAX_HEADER instead to allocate the skb.
1250 */
1251 unsigned short needed_headroom;
1252 unsigned short needed_tailroom;
1253
1da177e4 1254 /* Interface address info. */
a6f9a705 1255 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
c1f79426 1256 unsigned char addr_assign_type; /* hw address assignment type */
1da177e4 1257 unsigned char addr_len; /* hardware address length */
596b9b68 1258 unsigned char neigh_priv_len;
dffebd2c
N
1259 unsigned short dev_id; /* Used to differentiate devices
1260 * that share the same link
1261 * layer address
1262 */
ccffad25 1263 spinlock_t addr_list_lock;
22bedad3
JP
1264 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1265 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
4c3d5e7b
ED
1266 struct netdev_hw_addr_list dev_addrs; /* list of device
1267 * hw addresses
1268 */
1269#ifdef CONFIG_SYSFS
1270 struct kset *queues_kset;
1271#endif
1272
2d348d1f 1273 bool uc_promisc;
9d45abe1
WC
1274 unsigned int promiscuity;
1275 unsigned int allmulti;
1da177e4 1276
1da177e4
LT
1277
1278 /* Protocol specific pointers */
65ac6a5f 1279
d11ead75 1280#if IS_ENABLED(CONFIG_VLAN_8021Q)
5b9ea6e0 1281 struct vlan_info __rcu *vlan_info; /* VLAN info */
65ac6a5f 1282#endif
34a430d7 1283#if IS_ENABLED(CONFIG_NET_DSA)
cf50dcc2 1284 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
91da11f8 1285#endif
1da177e4 1286 void *atalk_ptr; /* AppleTalk link */
95ae6b22 1287 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
fc766e4c 1288 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
198caeca 1289 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1da177e4 1290 void *ax25_ptr; /* AX.25 specific data */
704232c2
JB
1291 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1292 assign before registering */
1da177e4 1293
9356b8fc 1294/*
cd13539b 1295 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1296 */
4dc89133
ED
1297 unsigned long last_rx; /* Time of last Rx
1298 * This should not be set in
1299 * drivers, unless really needed,
1300 * because network stack (bonding)
1301 * use it if/when necessary, to
1302 * avoid dirtying this cache line.
1303 */
1304
9356b8fc 1305 /* Interface address info used in eth_type_trans() */
f001fde5
JP
1306 unsigned char *dev_addr; /* hw address, (before bcast
1307 because most packets are
1308 unicast) */
1309
0a9627f2 1310
ccf5ff69 1311#ifdef CONFIG_RPS
0a9627f2
TH
1312 struct netdev_rx_queue *_rx;
1313
62fe0b40 1314 /* Number of RX queues allocated at register_netdev() time */
0a9627f2 1315 unsigned int num_rx_queues;
62fe0b40
BH
1316
1317 /* Number of RX queues currently active in device */
1318 unsigned int real_num_rx_queues;
c445477d 1319
df334545 1320#endif
0a9627f2 1321
61391cde 1322 rx_handler_func_t __rcu *rx_handler;
1323 void __rcu *rx_handler_data;
e8a0464c 1324
24824a09 1325 struct netdev_queue __rcu *ingress_queue;
4c3d5e7b
ED
1326 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1327
cd13539b
ED
1328
1329/*
1330 * Cache lines mostly used on transmit path
1331 */
e8a0464c 1332 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
fd2ea0a7
DM
1333
1334 /* Number of TX queues allocated at alloc_netdev_mq() time */
e8a0464c 1335 unsigned int num_tx_queues;
fd2ea0a7
DM
1336
1337 /* Number of TX queues currently active in device */
1338 unsigned int real_num_tx_queues;
1339
af356afa
PM
1340 /* root qdisc from userspace point of view */
1341 struct Qdisc *qdisc;
1342
1da177e4 1343 unsigned long tx_queue_len; /* Max frames per queue allowed */
c3f26a26 1344 spinlock_t tx_global_lock;
cd13539b 1345
bf264145 1346#ifdef CONFIG_XPS
a4177869 1347 struct xps_dev_maps __rcu *xps_maps;
bf264145 1348#endif
4c3d5e7b
ED
1349#ifdef CONFIG_RFS_ACCEL
1350 /* CPU reverse-mapping for RX completion interrupts, indexed
1351 * by RX queue number. Assigned by driver. This must only be
1352 * set if the ndo_rx_flow_steer operation is defined. */
1353 struct cpu_rmap *rx_cpu_rmap;
1354#endif
1d24eb48 1355
9356b8fc 1356 /* These may be needed for future network-power-down code. */
9d21493b
ED
1357
1358 /*
1359 * trans_start here is expensive for high speed devices on SMP,
1360 * please use netdev_queue->trans_start instead.
1361 */
9356b8fc
ED
1362 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1363
1364 int watchdog_timeo; /* used by dev_watchdog() */
1365 struct timer_list watchdog_timer;
1366
1da177e4 1367 /* Number of references to this device */
29b4433d 1368 int __percpu *pcpu_refcnt;
9356b8fc 1369
1da177e4
LT
1370 /* delayed register/unregister */
1371 struct list_head todo_list;
1da177e4
LT
1372 /* device index hash chain */
1373 struct hlist_node index_hlist;
1374
e014debe 1375 struct list_head link_watch_list;
572a103d 1376
1da177e4
LT
1377 /* register/unregister state machine */
1378 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1379 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1380 NETREG_UNREGISTERING, /* called unregister_netdevice */
1381 NETREG_UNREGISTERED, /* completed unregister todo */
1382 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1383 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1384 } reg_state:8;
1385
1386 bool dismantle; /* device is going do be freed */
a2835763
PM
1387
1388 enum {
1389 RTNL_LINK_INITIALIZED,
1390 RTNL_LINK_INITIALIZING,
1391 } rtnl_link_state:16;
1da177e4 1392
d314774c
SH
1393 /* Called from unregister, can be used to call free_netdev */
1394 void (*destructor)(struct net_device *dev);
1da177e4 1395
1da177e4 1396#ifdef CONFIG_NETPOLL
5fbee843 1397 struct netpoll_info __rcu *npinfo;
1da177e4 1398#endif
eae792b7 1399
c346dca1 1400#ifdef CONFIG_NET_NS
4a1c5371
EB
1401 /* Network namespace this network device is inside */
1402 struct net *nd_net;
c346dca1 1403#endif
4a1c5371 1404
4951704b 1405 /* mid-layer private */
a7855c78
ED
1406 union {
1407 void *ml_priv;
1408 struct pcpu_lstats __percpu *lstats; /* loopback stats */
290b895e 1409 struct pcpu_tstats __percpu *tstats; /* tunnel stats */
6d81f41c 1410 struct pcpu_dstats __percpu *dstats; /* dummy stats */
2681128f 1411 struct pcpu_vstats __percpu *vstats; /* veth stats */
a7855c78 1412 };
eca9ebac 1413 /* GARP */
3cc77ec7 1414 struct garp_port __rcu *garp_port;
febf018d
DW
1415 /* MRP */
1416 struct mrp_port __rcu *mrp_port;
1da177e4 1417
1da177e4 1418 /* class/net/name entry */
43cb76d9 1419 struct device dev;
0c509a6c
EB
1420 /* space for optional device, statistics, and wireless sysfs groups */
1421 const struct attribute_group *sysfs_groups[4];
38f7b870
PM
1422
1423 /* rtnetlink link ops */
1424 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1425
82cc1a7a
PWJ
1426 /* for setting kernel sock attribute on TCP connection setup */
1427#define GSO_MAX_SIZE 65536
1428 unsigned int gso_max_size;
30b678d8
BH
1429#define GSO_MAX_SEGS 65535
1430 u16 gso_max_segs;
d314774c 1431
7a6b6f51 1432#ifdef CONFIG_DCB
2f90b865 1433 /* Data Center Bridging netlink ops */
32953543 1434 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1435#endif
4f57c087
JF
1436 u8 num_tc;
1437 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1438 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1439
d11ead75 1440#if IS_ENABLED(CONFIG_FCOE)
4d288d57
YZ
1441 /* max exchange id for FCoE LRO by ddp */
1442 unsigned int fcoe_ddp_xid;
5bc1421e
NH
1443#endif
1444#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
1445 struct netprio_map __rcu *priomap;
4d288d57 1446#endif
c1f19b51
RC
1447 /* phy device may attach itself for hardware timestamping */
1448 struct phy_device *phydev;
cbda10fa 1449
23d3b8bf
ED
1450 struct lock_class_key *qdisc_tx_busylock;
1451
cbda10fa
VD
1452 /* group the device belongs to */
1453 int group;
9136461a
ED
1454
1455 struct pm_qos_request pm_qos_req;
1da177e4 1456};
43cb76d9 1457#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1458
1459#define NETDEV_ALIGN 32
1da177e4 1460
4f57c087
JF
1461static inline
1462int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1463{
1464 return dev->prio_tc_map[prio & TC_BITMASK];
1465}
1466
1467static inline
1468int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1469{
1470 if (tc >= dev->num_tc)
1471 return -EINVAL;
1472
1473 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1474 return 0;
1475}
1476
1477static inline
1478void netdev_reset_tc(struct net_device *dev)
1479{
1480 dev->num_tc = 0;
1481 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1482 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1483}
1484
1485static inline
1486int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1487{
1488 if (tc >= dev->num_tc)
1489 return -EINVAL;
1490
1491 dev->tc_to_txq[tc].count = count;
1492 dev->tc_to_txq[tc].offset = offset;
1493 return 0;
1494}
1495
1496static inline
1497int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1498{
1499 if (num_tc > TC_MAX_QUEUE)
1500 return -EINVAL;
1501
1502 dev->num_tc = num_tc;
1503 return 0;
1504}
1505
1506static inline
1507int netdev_get_num_tc(struct net_device *dev)
1508{
1509 return dev->num_tc;
1510}
1511
e8a0464c
DM
1512static inline
1513struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1514 unsigned int index)
1515{
1516 return &dev->_tx[index];
1517}
1518
1519static inline void netdev_for_each_tx_queue(struct net_device *dev,
1520 void (*f)(struct net_device *,
1521 struct netdev_queue *,
1522 void *),
1523 void *arg)
1524{
1525 unsigned int i;
1526
1527 for (i = 0; i < dev->num_tx_queues; i++)
1528 f(dev, &dev->_tx[i], arg);
1529}
1530
f629d208
JP
1531struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1532 struct sk_buff *skb);
1533u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb);
8c4c49df 1534
c346dca1
YH
1535/*
1536 * Net namespace inlines
1537 */
1538static inline
1539struct net *dev_net(const struct net_device *dev)
1540{
c2d9ba9b 1541 return read_pnet(&dev->nd_net);
c346dca1
YH
1542}
1543
1544static inline
f5aa23fd 1545void dev_net_set(struct net_device *dev, struct net *net)
c346dca1
YH
1546{
1547#ifdef CONFIG_NET_NS
f3005d7f
DL
1548 release_net(dev->nd_net);
1549 dev->nd_net = hold_net(net);
c346dca1
YH
1550#endif
1551}
1552
cf85d08f
LB
1553static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1554{
1555#ifdef CONFIG_NET_DSA_TAG_DSA
1556 if (dev->dsa_ptr != NULL)
1557 return dsa_uses_dsa_tags(dev->dsa_ptr);
1558#endif
1559
1560 return 0;
1561}
1562
396138f0
LB
1563static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1564{
1565#ifdef CONFIG_NET_DSA_TAG_TRAILER
1566 if (dev->dsa_ptr != NULL)
1567 return dsa_uses_trailer_tags(dev->dsa_ptr);
1568#endif
1569
1570 return 0;
1571}
1572
bea3348e
SH
1573/**
1574 * netdev_priv - access network device private data
1575 * @dev: network device
1576 *
1577 * Get network device private data
1578 */
6472ce60 1579static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1580{
1ce8e7b5 1581 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1582}
1583
1da177e4
LT
1584/* Set the sysfs physical device reference for the network logical device
1585 * if set prior to registration will cause a symlink during initialization.
1586 */
43cb76d9 1587#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1588
384912ed 1589/* Set the sysfs device type for the network logical device to allow
3f79410c 1590 * fine-grained identification of different network device types. For
384912ed
MH
1591 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1592 */
1593#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1594
82dc3c63
ED
1595/* Default NAPI poll() weight
1596 * Device drivers are strongly advised to not use bigger value
1597 */
1598#define NAPI_POLL_WEIGHT 64
1599
3b582cc1
SH
1600/**
1601 * netif_napi_add - initialize a napi context
1602 * @dev: network device
1603 * @napi: napi context
1604 * @poll: polling function
1605 * @weight: default weight
1606 *
1607 * netif_napi_add() must be used to initialize a napi context prior to calling
1608 * *any* of the other napi related functions.
1609 */
d565b0a1
HX
1610void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1611 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1612
d8156534
AD
1613/**
1614 * netif_napi_del - remove a napi context
1615 * @napi: napi context
1616 *
1617 * netif_napi_del() removes a napi context from the network device napi list
1618 */
d565b0a1
HX
1619void netif_napi_del(struct napi_struct *napi);
1620
1621struct napi_gro_cb {
78a478d0
HX
1622 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1623 void *frag0;
1624
7489594c
HX
1625 /* Length of frag0. */
1626 unsigned int frag0_len;
1627
86911732
HX
1628 /* This indicates where we are processing relative to skb->data. */
1629 int data_offset;
1630
d565b0a1
HX
1631 /* This is non-zero if the packet cannot be merged with the new skb. */
1632 int flush;
1633
1634 /* Number of segments aggregated. */
2e71a6f8
ED
1635 u16 count;
1636
1637 /* This is non-zero if the packet may be of the same flow. */
1638 u8 same_flow;
5d38a079
HX
1639
1640 /* Free the skb? */
2e71a6f8 1641 u8 free;
d7e8883c
ED
1642#define NAPI_GRO_FREE 1
1643#define NAPI_GRO_FREE_STOLEN_HEAD 2
2e71a6f8
ED
1644
1645 /* jiffies when first packet was created/queued */
1646 unsigned long age;
86347245
ED
1647
1648 /* Used in ipv6_gro_receive() */
1649 int proto;
c3c7c254
ED
1650
1651 /* used in skb_gro_receive() slow path */
1652 struct sk_buff *last;
d565b0a1
HX
1653};
1654
1655#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 1656
1da177e4 1657struct packet_type {
f2ccd8fa
DM
1658 __be16 type; /* This is really htons(ether_type). */
1659 struct net_device *dev; /* NULL is wildcarded here */
1660 int (*func) (struct sk_buff *,
1661 struct net_device *,
1662 struct packet_type *,
1663 struct net_device *);
c0de08d0
EL
1664 bool (*id_match)(struct packet_type *ptype,
1665 struct sock *sk);
1da177e4
LT
1666 void *af_packet_priv;
1667 struct list_head list;
1668};
1669
f191a1d1 1670struct offload_callbacks {
576a30eb 1671 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 1672 netdev_features_t features);
a430a43d 1673 int (*gso_send_check)(struct sk_buff *skb);
d565b0a1
HX
1674 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1675 struct sk_buff *skb);
1676 int (*gro_complete)(struct sk_buff *skb);
f191a1d1
VY
1677};
1678
1679struct packet_offload {
1680 __be16 type; /* This is really htons(ether_type). */
1681 struct offload_callbacks callbacks;
1682 struct list_head list;
1da177e4
LT
1683};
1684
1da177e4
LT
1685#include <linux/notifier.h>
1686
dcfe1421
AW
1687/* netdevice notifier chain. Please remember to update the rtnetlink
1688 * notification exclusion list in rtnetlink_event() when adding new
1689 * types.
1690 */
1691#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1692#define NETDEV_DOWN 0x0002
1693#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1694 detected a hardware crash and restarted
1695 - we can use this eg to kick tcp sessions
1696 once done */
1697#define NETDEV_CHANGE 0x0004 /* Notify device state change */
1698#define NETDEV_REGISTER 0x0005
1699#define NETDEV_UNREGISTER 0x0006
1700#define NETDEV_CHANGEMTU 0x0007
1701#define NETDEV_CHANGEADDR 0x0008
1702#define NETDEV_GOING_DOWN 0x0009
1703#define NETDEV_CHANGENAME 0x000A
1704#define NETDEV_FEAT_CHANGE 0x000B
1705#define NETDEV_BONDING_FAILOVER 0x000C
1706#define NETDEV_PRE_UP 0x000D
1707#define NETDEV_PRE_TYPE_CHANGE 0x000E
1708#define NETDEV_POST_TYPE_CHANGE 0x000F
1709#define NETDEV_POST_INIT 0x0010
0115e8e3 1710#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
1711#define NETDEV_RELEASE 0x0012
1712#define NETDEV_NOTIFY_PEERS 0x0013
1713#define NETDEV_JOIN 0x0014
42e52bf9 1714#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 1715#define NETDEV_RESEND_IGMP 0x0016
dcfe1421 1716
f629d208
JP
1717int register_netdevice_notifier(struct notifier_block *nb);
1718int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
1719
1720struct netdev_notifier_info {
1721 struct net_device *dev;
1722};
1723
be9efd36
JP
1724struct netdev_notifier_change_info {
1725 struct netdev_notifier_info info; /* must be first */
1726 unsigned int flags_changed;
1727};
1728
75538c2b
CW
1729static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1730 struct net_device *dev)
1731{
1732 info->dev = dev;
1733}
1734
351638e7
JP
1735static inline struct net_device *
1736netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1737{
1738 return info->dev;
1739}
1740
f629d208
JP
1741int call_netdevice_notifiers_info(unsigned long val, struct net_device *dev,
1742 struct netdev_notifier_info *info);
1743int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
1744
1745
1da177e4
LT
1746extern rwlock_t dev_base_lock; /* Device list lock */
1747
881d966b
EB
1748#define for_each_netdev(net, d) \
1749 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
1750#define for_each_netdev_reverse(net, d) \
1751 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
1752#define for_each_netdev_rcu(net, d) \
1753 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
1754#define for_each_netdev_safe(net, d, n) \
1755 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1756#define for_each_netdev_continue(net, d) \
1757 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 1758#define for_each_netdev_continue_rcu(net, d) \
1759 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 1760#define for_each_netdev_in_bond_rcu(bond, slave) \
1761 for_each_netdev_rcu(&init_net, slave) \
1762 if (netdev_master_upper_dev_get_rcu(slave) == bond)
881d966b 1763#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 1764
a050c33f
DL
1765static inline struct net_device *next_net_device(struct net_device *dev)
1766{
1767 struct list_head *lh;
1768 struct net *net;
1769
c346dca1 1770 net = dev_net(dev);
a050c33f
DL
1771 lh = dev->dev_list.next;
1772 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1773}
1774
ce81b76a
ED
1775static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1776{
1777 struct list_head *lh;
1778 struct net *net;
1779
1780 net = dev_net(dev);
ccf43438 1781 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
1782 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1783}
1784
a050c33f
DL
1785static inline struct net_device *first_net_device(struct net *net)
1786{
1787 return list_empty(&net->dev_base_head) ? NULL :
1788 net_device_entry(net->dev_base_head.next);
1789}
7562f876 1790
ccf43438
ED
1791static inline struct net_device *first_net_device_rcu(struct net *net)
1792{
1793 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1794
1795 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1796}
1797
f629d208
JP
1798int netdev_boot_setup_check(struct net_device *dev);
1799unsigned long netdev_boot_base(const char *prefix, int unit);
1800struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1801 const char *hwaddr);
1802struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1803struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1804void dev_add_pack(struct packet_type *pt);
1805void dev_remove_pack(struct packet_type *pt);
1806void __dev_remove_pack(struct packet_type *pt);
1807void dev_add_offload(struct packet_offload *po);
1808void dev_remove_offload(struct packet_offload *po);
1809void __dev_remove_offload(struct packet_offload *po);
1810
1811struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1812 unsigned short mask);
1813struct net_device *dev_get_by_name(struct net *net, const char *name);
1814struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1815struct net_device *__dev_get_by_name(struct net *net, const char *name);
1816int dev_alloc_name(struct net_device *dev, const char *name);
1817int dev_open(struct net_device *dev);
1818int dev_close(struct net_device *dev);
1819void dev_disable_lro(struct net_device *dev);
1820int dev_loopback_xmit(struct sk_buff *newskb);
1821int dev_queue_xmit(struct sk_buff *skb);
1822int register_netdevice(struct net_device *dev);
1823void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
1824void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
1825static inline void unregister_netdevice(struct net_device *dev)
1826{
1827 unregister_netdevice_queue(dev, NULL);
1828}
1829
f629d208
JP
1830int netdev_refcnt_read(const struct net_device *dev);
1831void free_netdev(struct net_device *dev);
74d332c1 1832void netdev_freemem(struct net_device *dev);
f629d208
JP
1833void synchronize_net(void);
1834int init_dummy_netdev(struct net_device *dev);
937f1ba5 1835
f629d208
JP
1836struct net_device *dev_get_by_index(struct net *net, int ifindex);
1837struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1838struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1839int netdev_get_name(struct net *net, char *name, int ifindex);
1840int dev_restart(struct net_device *dev);
1da177e4 1841#ifdef CONFIG_NETPOLL_TRAP
f629d208 1842int netpoll_trap(void);
1da177e4 1843#endif
f629d208 1844int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
1845
1846static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1847{
1848 return NAPI_GRO_CB(skb)->data_offset;
1849}
1850
1851static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1852{
1853 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1854}
1855
1856static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1857{
1858 NAPI_GRO_CB(skb)->data_offset += len;
1859}
1860
a5b1cf28
HX
1861static inline void *skb_gro_header_fast(struct sk_buff *skb,
1862 unsigned int offset)
86911732 1863{
a5b1cf28
HX
1864 return NAPI_GRO_CB(skb)->frag0 + offset;
1865}
78a478d0 1866
a5b1cf28
HX
1867static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1868{
1869 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1870}
78a478d0 1871
a5b1cf28
HX
1872static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1873 unsigned int offset)
1874{
17dd759c
HX
1875 if (!pskb_may_pull(skb, hlen))
1876 return NULL;
1877
a5b1cf28
HX
1878 NAPI_GRO_CB(skb)->frag0 = NULL;
1879 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 1880 return skb->data + offset;
86911732 1881}
1da177e4 1882
aa4b9f53
HX
1883static inline void *skb_gro_mac_header(struct sk_buff *skb)
1884{
78d3fd0b 1885 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
aa4b9f53
HX
1886}
1887
36e7b1b8
HX
1888static inline void *skb_gro_network_header(struct sk_buff *skb)
1889{
78d3fd0b
HX
1890 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1891 skb_network_offset(skb);
36e7b1b8
HX
1892}
1893
0c4e8581
SH
1894static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1895 unsigned short type,
3b04ddde 1896 const void *daddr, const void *saddr,
95c96174 1897 unsigned int len)
0c4e8581 1898{
f1ecfd5d 1899 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 1900 return 0;
3b04ddde
SH
1901
1902 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
1903}
1904
b95cce35
SH
1905static inline int dev_parse_header(const struct sk_buff *skb,
1906 unsigned char *haddr)
1907{
1908 const struct net_device *dev = skb->dev;
1909
1b83336b 1910 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 1911 return 0;
3b04ddde 1912 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
1913}
1914
1da177e4 1915typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 1916int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
1917static inline int unregister_gifconf(unsigned int family)
1918{
1919 return register_gifconf(family, NULL);
1920}
1921
99bbc707 1922#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 1923#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
1924struct sd_flow_limit {
1925 u64 count;
1926 unsigned int num_buckets;
1927 unsigned int history_head;
1928 u16 history[FLOW_LIMIT_HISTORY];
1929 u8 buckets[];
1930};
1931
1932extern int netdev_flow_limit_table_len;
1933#endif /* CONFIG_NET_FLOW_LIMIT */
1934
1da177e4 1935/*
88751275 1936 * Incoming packets are placed on per-cpu queues
1da177e4 1937 */
d94d9fee 1938struct softnet_data {
37437bb2 1939 struct Qdisc *output_queue;
a9cbd588 1940 struct Qdisc **output_queue_tailp;
1da177e4 1941 struct list_head poll_list;
1da177e4 1942 struct sk_buff *completion_queue;
6e7676c1 1943 struct sk_buff_head process_queue;
1da177e4 1944
dee42870 1945 /* stats */
cd7b5396
DM
1946 unsigned int processed;
1947 unsigned int time_squeeze;
1948 unsigned int cpu_collision;
1949 unsigned int received_rps;
dee42870 1950
fd793d89 1951#ifdef CONFIG_RPS
88751275
ED
1952 struct softnet_data *rps_ipi_list;
1953
1954 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 1955 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
1956 struct softnet_data *rps_ipi_next;
1957 unsigned int cpu;
fec5e652 1958 unsigned int input_queue_head;
76cc8b13 1959 unsigned int input_queue_tail;
1e94d72f 1960#endif
95c96174 1961 unsigned int dropped;
0a9627f2 1962 struct sk_buff_head input_pkt_queue;
bea3348e 1963 struct napi_struct backlog;
99bbc707
WB
1964
1965#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 1966 struct sd_flow_limit __rcu *flow_limit;
99bbc707 1967#endif
1da177e4
LT
1968};
1969
76cc8b13 1970static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
1971{
1972#ifdef CONFIG_RPS
76cc8b13
TH
1973 sd->input_queue_head++;
1974#endif
1975}
1976
1977static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1978 unsigned int *qtail)
1979{
1980#ifdef CONFIG_RPS
1981 *qtail = ++sd->input_queue_tail;
fec5e652
TH
1982#endif
1983}
1984
0a9627f2 1985DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 1986
f629d208 1987void __netif_schedule(struct Qdisc *q);
1da177e4 1988
86d804e1 1989static inline void netif_schedule_queue(struct netdev_queue *txq)
1da177e4 1990{
73466498 1991 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
37437bb2 1992 __netif_schedule(txq->qdisc);
86d804e1
DM
1993}
1994
fd2ea0a7
DM
1995static inline void netif_tx_schedule_all(struct net_device *dev)
1996{
1997 unsigned int i;
1998
1999 for (i = 0; i < dev->num_tx_queues; i++)
2000 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2001}
2002
d29f749e
DJ
2003static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2004{
73466498 2005 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2006}
2007
bea3348e
SH
2008/**
2009 * netif_start_queue - allow transmit
2010 * @dev: network device
2011 *
2012 * Allow upper layers to call the device hard_start_xmit routine.
2013 */
1da177e4
LT
2014static inline void netif_start_queue(struct net_device *dev)
2015{
e8a0464c 2016 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2017}
2018
fd2ea0a7
DM
2019static inline void netif_tx_start_all_queues(struct net_device *dev)
2020{
2021 unsigned int i;
2022
2023 for (i = 0; i < dev->num_tx_queues; i++) {
2024 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2025 netif_tx_start_queue(txq);
2026 }
2027}
2028
79d16385 2029static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1da177e4
LT
2030{
2031#ifdef CONFIG_NETPOLL_TRAP
5f286e11 2032 if (netpoll_trap()) {
7b3d3e4f 2033 netif_tx_start_queue(dev_queue);
1da177e4 2034 return;
5f286e11 2035 }
1da177e4 2036#endif
73466498 2037 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
37437bb2 2038 __netif_schedule(dev_queue->qdisc);
79d16385
DM
2039}
2040
d29f749e
DJ
2041/**
2042 * netif_wake_queue - restart transmit
2043 * @dev: network device
2044 *
2045 * Allow upper layers to call the device hard_start_xmit routine.
2046 * Used for flow control when transmit resources are available.
2047 */
79d16385
DM
2048static inline void netif_wake_queue(struct net_device *dev)
2049{
e8a0464c 2050 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2051}
2052
fd2ea0a7
DM
2053static inline void netif_tx_wake_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_wake_queue(txq);
2060 }
2061}
2062
d29f749e
DJ
2063static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2064{
18543a64 2065 if (WARN_ON(!dev_queue)) {
256ee435 2066 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
18543a64
GC
2067 return;
2068 }
73466498 2069 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2070}
2071
bea3348e
SH
2072/**
2073 * netif_stop_queue - stop transmitted packets
2074 * @dev: network device
2075 *
2076 * Stop upper layers calling the device hard_start_xmit routine.
2077 * Used for flow control when transmit resources are unavailable.
2078 */
1da177e4
LT
2079static inline void netif_stop_queue(struct net_device *dev)
2080{
e8a0464c 2081 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2082}
2083
fd2ea0a7
DM
2084static inline void netif_tx_stop_all_queues(struct net_device *dev)
2085{
2086 unsigned int i;
2087
2088 for (i = 0; i < dev->num_tx_queues; i++) {
2089 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2090 netif_tx_stop_queue(txq);
2091 }
2092}
2093
4d29515f 2094static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2095{
73466498 2096 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2097}
2098
bea3348e
SH
2099/**
2100 * netif_queue_stopped - test if transmit queue is flowblocked
2101 * @dev: network device
2102 *
2103 * Test if transmit queue on device is currently unable to send.
2104 */
4d29515f 2105static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2106{
e8a0464c 2107 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2108}
2109
4d29515f 2110static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2111{
73466498
TH
2112 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2113}
2114
4d29515f 2115static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2116{
2117 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2118}
2119
c5d67bd7
TH
2120static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2121 unsigned int bytes)
2122{
114cf580
TH
2123#ifdef CONFIG_BQL
2124 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2125
2126 if (likely(dql_avail(&dev_queue->dql) >= 0))
2127 return;
2128
2129 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2130
2131 /*
2132 * The XOFF flag must be set before checking the dql_avail below,
2133 * because in netdev_tx_completed_queue we update the dql_completed
2134 * before checking the XOFF flag.
2135 */
2136 smp_mb();
2137
2138 /* check again in case another CPU has just made room avail */
2139 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2140 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2141#endif
c5d67bd7
TH
2142}
2143
0042d0c8
FF
2144/**
2145 * netdev_sent_queue - report the number of bytes queued to hardware
2146 * @dev: network device
2147 * @bytes: number of bytes queued to the hardware device queue
2148 *
2149 * Report the number of bytes queued for sending/completion to the network
2150 * device hardware queue. @bytes should be a good approximation and should
2151 * exactly match netdev_completed_queue() @bytes
2152 */
c5d67bd7
TH
2153static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2154{
2155 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2156}
2157
2158static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2159 unsigned int pkts, unsigned int bytes)
c5d67bd7 2160{
114cf580 2161#ifdef CONFIG_BQL
b37c0fbe
AD
2162 if (unlikely(!bytes))
2163 return;
2164
2165 dql_completed(&dev_queue->dql, bytes);
2166
2167 /*
2168 * Without the memory barrier there is a small possiblity that
2169 * netdev_tx_sent_queue will miss the update and cause the queue to
2170 * be stopped forever
2171 */
2172 smp_mb();
2173
2174 if (dql_avail(&dev_queue->dql) < 0)
2175 return;
2176
2177 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2178 netif_schedule_queue(dev_queue);
114cf580 2179#endif
c5d67bd7
TH
2180}
2181
0042d0c8
FF
2182/**
2183 * netdev_completed_queue - report bytes and packets completed by device
2184 * @dev: network device
2185 * @pkts: actual number of packets sent over the medium
2186 * @bytes: actual number of bytes sent over the medium
2187 *
2188 * Report the number of bytes and packets transmitted by the network device
2189 * hardware queue over the physical medium, @bytes must exactly match the
2190 * @bytes amount passed to netdev_sent_queue()
2191 */
c5d67bd7 2192static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2193 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2194{
2195 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2196}
2197
2198static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2199{
114cf580 2200#ifdef CONFIG_BQL
5c490354 2201 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2202 dql_reset(&q->dql);
2203#endif
c5d67bd7
TH
2204}
2205
0042d0c8
FF
2206/**
2207 * netdev_reset_queue - reset the packets and bytes count of a network device
2208 * @dev_queue: network device
2209 *
2210 * Reset the bytes and packet count of a network device and clear the
2211 * software flow control OFF bit for this network device
2212 */
c5d67bd7
TH
2213static inline void netdev_reset_queue(struct net_device *dev_queue)
2214{
2215 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2216}
2217
bea3348e
SH
2218/**
2219 * netif_running - test if up
2220 * @dev: network device
2221 *
2222 * Test if the device has been brought up.
2223 */
4d29515f 2224static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2225{
2226 return test_bit(__LINK_STATE_START, &dev->state);
2227}
2228
f25f4e44
PWJ
2229/*
2230 * Routines to manage the subqueues on a device. We only need start
2231 * stop, and a check if it's stopped. All other device management is
2232 * done at the overall netdevice level.
2233 * Also test the device if we're multiqueue.
2234 */
bea3348e
SH
2235
2236/**
2237 * netif_start_subqueue - allow sending packets on subqueue
2238 * @dev: network device
2239 * @queue_index: sub queue index
2240 *
2241 * Start individual transmit queue of a device with multiple transmit queues.
2242 */
f25f4e44
PWJ
2243static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2244{
fd2ea0a7 2245 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2246
2247 netif_tx_start_queue(txq);
f25f4e44
PWJ
2248}
2249
bea3348e
SH
2250/**
2251 * netif_stop_subqueue - stop sending packets on subqueue
2252 * @dev: network device
2253 * @queue_index: sub queue index
2254 *
2255 * Stop individual transmit queue of a device with multiple transmit queues.
2256 */
f25f4e44
PWJ
2257static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2258{
fd2ea0a7 2259 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
f25f4e44
PWJ
2260#ifdef CONFIG_NETPOLL_TRAP
2261 if (netpoll_trap())
2262 return;
2263#endif
7b3d3e4f 2264 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2265}
2266
bea3348e
SH
2267/**
2268 * netif_subqueue_stopped - test status of subqueue
2269 * @dev: network device
2270 * @queue_index: sub queue index
2271 *
2272 * Check individual transmit queue of a device with multiple transmit queues.
2273 */
4d29515f
DM
2274static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2275 u16 queue_index)
f25f4e44 2276{
fd2ea0a7 2277 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2278
2279 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2280}
2281
4d29515f
DM
2282static inline bool netif_subqueue_stopped(const struct net_device *dev,
2283 struct sk_buff *skb)
668f895a
PE
2284{
2285 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2286}
bea3348e
SH
2287
2288/**
2289 * netif_wake_subqueue - allow sending packets on subqueue
2290 * @dev: network device
2291 * @queue_index: sub queue index
2292 *
2293 * Resume individual transmit queue of a device with multiple transmit queues.
2294 */
f25f4e44
PWJ
2295static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2296{
fd2ea0a7 2297 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
f25f4e44
PWJ
2298#ifdef CONFIG_NETPOLL_TRAP
2299 if (netpoll_trap())
2300 return;
2301#endif
73466498 2302 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
37437bb2 2303 __netif_schedule(txq->qdisc);
f25f4e44
PWJ
2304}
2305
537c00de 2306#ifdef CONFIG_XPS
53af53ae 2307int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2308 u16 index);
537c00de
AD
2309#else
2310static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2311 const struct cpumask *mask,
537c00de
AD
2312 u16 index)
2313{
2314 return 0;
2315}
2316#endif
2317
a3d22a68
VZ
2318/*
2319 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2320 * as a distribution range limit for the returned value.
2321 */
2322static inline u16 skb_tx_hash(const struct net_device *dev,
2323 const struct sk_buff *skb)
2324{
2325 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2326}
2327
bea3348e
SH
2328/**
2329 * netif_is_multiqueue - test if device has multiple transmit queues
2330 * @dev: network device
2331 *
2332 * Check if device has multiple transmit queues
bea3348e 2333 */
4d29515f 2334static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2335{
a02cec21 2336 return dev->num_tx_queues > 1;
f25f4e44 2337}
1da177e4 2338
f629d208 2339int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2340
62fe0b40 2341#ifdef CONFIG_RPS
f629d208 2342int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
2343#else
2344static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2345 unsigned int rxq)
2346{
2347 return 0;
2348}
2349#endif
2350
3171d026
BH
2351static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2352 const struct net_device *from_dev)
2353{
ee6ae1a1
JP
2354 int err;
2355
2356 err = netif_set_real_num_tx_queues(to_dev,
2357 from_dev->real_num_tx_queues);
2358 if (err)
2359 return err;
3171d026
BH
2360#ifdef CONFIG_RPS
2361 return netif_set_real_num_rx_queues(to_dev,
2362 from_dev->real_num_rx_queues);
2363#else
2364 return 0;
2365#endif
2366}
2367
16917b87 2368#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 2369int netif_get_num_default_rss_queues(void);
16917b87 2370
e6247027
ED
2371enum skb_free_reason {
2372 SKB_REASON_CONSUMED,
2373 SKB_REASON_DROPPED,
2374};
2375
2376void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2377void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 2378
e6247027
ED
2379/*
2380 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2381 * interrupt context or with hardware interrupts being disabled.
2382 * (in_irq() || irqs_disabled())
2383 *
2384 * We provide four helpers that can be used in following contexts :
2385 *
2386 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2387 * replacing kfree_skb(skb)
2388 *
2389 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2390 * Typically used in place of consume_skb(skb) in TX completion path
2391 *
2392 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2393 * replacing kfree_skb(skb)
2394 *
2395 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2396 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 2397 */
e6247027
ED
2398static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2399{
2400 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2401}
2402
2403static inline void dev_consume_skb_irq(struct sk_buff *skb)
2404{
2405 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2406}
2407
2408static inline void dev_kfree_skb_any(struct sk_buff *skb)
2409{
2410 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2411}
2412
2413static inline void dev_consume_skb_any(struct sk_buff *skb)
2414{
2415 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2416}
1da177e4 2417
f629d208
JP
2418int netif_rx(struct sk_buff *skb);
2419int netif_rx_ni(struct sk_buff *skb);
2420int netif_receive_skb(struct sk_buff *skb);
2421gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2422void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2423struct sk_buff *napi_get_frags(struct napi_struct *napi);
2424gro_result_t napi_gro_frags(struct napi_struct *napi);
76620aaf
HX
2425
2426static inline void napi_free_frags(struct napi_struct *napi)
2427{
2428 kfree_skb(napi->skb);
2429 napi->skb = NULL;
2430}
2431
f629d208
JP
2432int netdev_rx_handler_register(struct net_device *dev,
2433 rx_handler_func_t *rx_handler,
2434 void *rx_handler_data);
2435void netdev_rx_handler_unregister(struct net_device *dev);
2436
2437bool dev_valid_name(const char *name);
2438int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2439int dev_ethtool(struct net *net, struct ifreq *);
2440unsigned int dev_get_flags(const struct net_device *);
2441int __dev_change_flags(struct net_device *, unsigned int flags);
2442int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
2443void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2444 unsigned int gchanges);
f629d208
JP
2445int dev_change_name(struct net_device *, const char *);
2446int dev_set_alias(struct net_device *, const char *, size_t);
2447int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2448int dev_set_mtu(struct net_device *, int);
2449void dev_set_group(struct net_device *, int);
2450int dev_set_mac_address(struct net_device *, struct sockaddr *);
2451int dev_change_carrier(struct net_device *, bool new_carrier);
2452int dev_get_phys_port_id(struct net_device *dev,
2453 struct netdev_phys_port_id *ppid);
2454int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
a6cc0cfa 2455 struct netdev_queue *txq, void *accel_priv);
f629d208 2456int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1da177e4 2457
20380731 2458extern int netdev_budget;
1da177e4
LT
2459
2460/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 2461void netdev_run_todo(void);
1da177e4 2462
bea3348e
SH
2463/**
2464 * dev_put - release reference to device
2465 * @dev: network device
2466 *
9ef4429b 2467 * Release reference to device to allow it to be freed.
bea3348e 2468 */
1da177e4
LT
2469static inline void dev_put(struct net_device *dev)
2470{
933393f5 2471 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
2472}
2473
bea3348e
SH
2474/**
2475 * dev_hold - get reference to device
2476 * @dev: network device
2477 *
9ef4429b 2478 * Hold reference to device to keep it from being freed.
bea3348e 2479 */
15333061
SH
2480static inline void dev_hold(struct net_device *dev)
2481{
933393f5 2482 this_cpu_inc(*dev->pcpu_refcnt);
15333061 2483}
1da177e4
LT
2484
2485/* Carrier loss detection, dial on demand. The functions netif_carrier_on
2486 * and _off may be called from IRQ context, but it is caller
2487 * who is responsible for serialization of these calls.
b00055aa
SR
2488 *
2489 * The name carrier is inappropriate, these functions should really be
2490 * called netif_lowerlayer_*() because they represent the state of any
2491 * kind of lower layer not just hardware media.
1da177e4
LT
2492 */
2493
f629d208
JP
2494void linkwatch_init_dev(struct net_device *dev);
2495void linkwatch_fire_event(struct net_device *dev);
2496void linkwatch_forget_dev(struct net_device *dev);
1da177e4 2497
bea3348e
SH
2498/**
2499 * netif_carrier_ok - test if carrier present
2500 * @dev: network device
2501 *
2502 * Check if carrier is present on device
2503 */
4d29515f 2504static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
2505{
2506 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2507}
2508
f629d208 2509unsigned long dev_trans_start(struct net_device *dev);
9d21493b 2510
f629d208 2511void __netdev_watchdog_up(struct net_device *dev);
1da177e4 2512
f629d208 2513void netif_carrier_on(struct net_device *dev);
1da177e4 2514
f629d208 2515void netif_carrier_off(struct net_device *dev);
1da177e4 2516
bea3348e
SH
2517/**
2518 * netif_dormant_on - mark device as dormant.
2519 * @dev: network device
2520 *
2521 * Mark device as dormant (as per RFC2863).
2522 *
2523 * The dormant state indicates that the relevant interface is not
2524 * actually in a condition to pass packets (i.e., it is not 'up') but is
2525 * in a "pending" state, waiting for some external event. For "on-
2526 * demand" interfaces, this new state identifies the situation where the
2527 * interface is waiting for events to place it in the up state.
2528 *
2529 */
b00055aa
SR
2530static inline void netif_dormant_on(struct net_device *dev)
2531{
2532 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2533 linkwatch_fire_event(dev);
2534}
2535
bea3348e
SH
2536/**
2537 * netif_dormant_off - set device as not dormant.
2538 * @dev: network device
2539 *
2540 * Device is not in dormant state.
2541 */
b00055aa
SR
2542static inline void netif_dormant_off(struct net_device *dev)
2543{
2544 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2545 linkwatch_fire_event(dev);
2546}
2547
bea3348e
SH
2548/**
2549 * netif_dormant - test if carrier present
2550 * @dev: network device
2551 *
2552 * Check if carrier is present on device
2553 */
4d29515f 2554static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
2555{
2556 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2557}
2558
2559
bea3348e
SH
2560/**
2561 * netif_oper_up - test if device is operational
2562 * @dev: network device
2563 *
2564 * Check if carrier is operational
2565 */
4d29515f 2566static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 2567{
b00055aa
SR
2568 return (dev->operstate == IF_OPER_UP ||
2569 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2570}
2571
bea3348e
SH
2572/**
2573 * netif_device_present - is device available or removed
2574 * @dev: network device
2575 *
2576 * Check if device has not been removed from system.
2577 */
4d29515f 2578static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
2579{
2580 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2581}
2582
f629d208 2583void netif_device_detach(struct net_device *dev);
1da177e4 2584
f629d208 2585void netif_device_attach(struct net_device *dev);
1da177e4
LT
2586
2587/*
2588 * Network interface message level settings
2589 */
1da177e4
LT
2590
2591enum {
2592 NETIF_MSG_DRV = 0x0001,
2593 NETIF_MSG_PROBE = 0x0002,
2594 NETIF_MSG_LINK = 0x0004,
2595 NETIF_MSG_TIMER = 0x0008,
2596 NETIF_MSG_IFDOWN = 0x0010,
2597 NETIF_MSG_IFUP = 0x0020,
2598 NETIF_MSG_RX_ERR = 0x0040,
2599 NETIF_MSG_TX_ERR = 0x0080,
2600 NETIF_MSG_TX_QUEUED = 0x0100,
2601 NETIF_MSG_INTR = 0x0200,
2602 NETIF_MSG_TX_DONE = 0x0400,
2603 NETIF_MSG_RX_STATUS = 0x0800,
2604 NETIF_MSG_PKTDATA = 0x1000,
2605 NETIF_MSG_HW = 0x2000,
2606 NETIF_MSG_WOL = 0x4000,
2607};
2608
2609#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2610#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2611#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2612#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2613#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2614#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2615#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2616#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2617#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2618#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2619#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2620#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2621#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2622#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2623#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2624
2625static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2626{
2627 /* use default */
2628 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2629 return default_msg_enable_bits;
2630 if (debug_value == 0) /* no output */
2631 return 0;
2632 /* set low N bits */
2633 return (1 << debug_value) - 1;
2634}
2635
c773e847 2636static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 2637{
c773e847
DM
2638 spin_lock(&txq->_xmit_lock);
2639 txq->xmit_lock_owner = cpu;
22dd7495
JHS
2640}
2641
fd2ea0a7
DM
2642static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2643{
2644 spin_lock_bh(&txq->_xmit_lock);
2645 txq->xmit_lock_owner = smp_processor_id();
2646}
2647
4d29515f 2648static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 2649{
4d29515f 2650 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
2651 if (likely(ok))
2652 txq->xmit_lock_owner = smp_processor_id();
2653 return ok;
2654}
2655
2656static inline void __netif_tx_unlock(struct netdev_queue *txq)
2657{
2658 txq->xmit_lock_owner = -1;
2659 spin_unlock(&txq->_xmit_lock);
2660}
2661
2662static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2663{
2664 txq->xmit_lock_owner = -1;
2665 spin_unlock_bh(&txq->_xmit_lock);
2666}
2667
08baf561
ED
2668static inline void txq_trans_update(struct netdev_queue *txq)
2669{
2670 if (txq->xmit_lock_owner != -1)
2671 txq->trans_start = jiffies;
2672}
2673
d29f749e
DJ
2674/**
2675 * netif_tx_lock - grab network device transmit lock
2676 * @dev: network device
d29f749e
DJ
2677 *
2678 * Get network device transmit lock
2679 */
22dd7495
JHS
2680static inline void netif_tx_lock(struct net_device *dev)
2681{
e8a0464c 2682 unsigned int i;
c3f26a26 2683 int cpu;
c773e847 2684
c3f26a26
DM
2685 spin_lock(&dev->tx_global_lock);
2686 cpu = smp_processor_id();
e8a0464c
DM
2687 for (i = 0; i < dev->num_tx_queues; i++) {
2688 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
2689
2690 /* We are the only thread of execution doing a
2691 * freeze, but we have to grab the _xmit_lock in
2692 * order to synchronize with threads which are in
2693 * the ->hard_start_xmit() handler and already
2694 * checked the frozen bit.
2695 */
e8a0464c 2696 __netif_tx_lock(txq, cpu);
c3f26a26
DM
2697 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2698 __netif_tx_unlock(txq);
e8a0464c 2699 }
932ff279
HX
2700}
2701
2702static inline void netif_tx_lock_bh(struct net_device *dev)
2703{
e8a0464c
DM
2704 local_bh_disable();
2705 netif_tx_lock(dev);
932ff279
HX
2706}
2707
932ff279
HX
2708static inline void netif_tx_unlock(struct net_device *dev)
2709{
e8a0464c
DM
2710 unsigned int i;
2711
2712 for (i = 0; i < dev->num_tx_queues; i++) {
2713 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 2714
c3f26a26
DM
2715 /* No need to grab the _xmit_lock here. If the
2716 * queue is not stopped for another reason, we
2717 * force a schedule.
2718 */
2719 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 2720 netif_schedule_queue(txq);
c3f26a26
DM
2721 }
2722 spin_unlock(&dev->tx_global_lock);
932ff279
HX
2723}
2724
2725static inline void netif_tx_unlock_bh(struct net_device *dev)
2726{
e8a0464c
DM
2727 netif_tx_unlock(dev);
2728 local_bh_enable();
932ff279
HX
2729}
2730
c773e847 2731#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 2732 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 2733 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
2734 } \
2735}
2736
c773e847 2737#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 2738 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 2739 __netif_tx_unlock(txq); \
22dd7495
JHS
2740 } \
2741}
2742
1da177e4
LT
2743static inline void netif_tx_disable(struct net_device *dev)
2744{
fd2ea0a7 2745 unsigned int i;
c3f26a26 2746 int cpu;
fd2ea0a7 2747
c3f26a26
DM
2748 local_bh_disable();
2749 cpu = smp_processor_id();
fd2ea0a7
DM
2750 for (i = 0; i < dev->num_tx_queues; i++) {
2751 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
2752
2753 __netif_tx_lock(txq, cpu);
fd2ea0a7 2754 netif_tx_stop_queue(txq);
c3f26a26 2755 __netif_tx_unlock(txq);
fd2ea0a7 2756 }
c3f26a26 2757 local_bh_enable();
1da177e4
LT
2758}
2759
e308a5d8
DM
2760static inline void netif_addr_lock(struct net_device *dev)
2761{
2762 spin_lock(&dev->addr_list_lock);
2763}
2764
2429f7ac
JP
2765static inline void netif_addr_lock_nested(struct net_device *dev)
2766{
2767 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2768}
2769
e308a5d8
DM
2770static inline void netif_addr_lock_bh(struct net_device *dev)
2771{
2772 spin_lock_bh(&dev->addr_list_lock);
2773}
2774
2775static inline void netif_addr_unlock(struct net_device *dev)
2776{
2777 spin_unlock(&dev->addr_list_lock);
2778}
2779
2780static inline void netif_addr_unlock_bh(struct net_device *dev)
2781{
2782 spin_unlock_bh(&dev->addr_list_lock);
2783}
2784
f001fde5 2785/*
31278e71 2786 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
2787 * rcu_read_lock held.
2788 */
2789#define for_each_dev_addr(dev, ha) \
31278e71 2790 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 2791
1da177e4
LT
2792/* These functions live elsewhere (drivers/net/net_init.c, but related) */
2793
f629d208 2794void ether_setup(struct net_device *dev);
1da177e4
LT
2795
2796/* Support for loadable net-drivers */
f629d208
JP
2797struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2798 void (*setup)(struct net_device *),
2799 unsigned int txqs, unsigned int rxqs);
f25f4e44 2800#define alloc_netdev(sizeof_priv, name, setup) \
36909ea4
TH
2801 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2802
2803#define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2804 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2805
f629d208
JP
2806int register_netdev(struct net_device *dev);
2807void unregister_netdev(struct net_device *dev);
f001fde5 2808
22bedad3 2809/* General hardware address lists handling functions */
f629d208
JP
2810int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2811 struct netdev_hw_addr_list *from_list,
2812 int addr_len, unsigned char addr_type);
2813void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2814 struct netdev_hw_addr_list *from_list,
2815 int addr_len, unsigned char addr_type);
2816int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2817 struct netdev_hw_addr_list *from_list, int addr_len);
2818void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2819 struct netdev_hw_addr_list *from_list, int addr_len);
2820void __hw_addr_flush(struct netdev_hw_addr_list *list);
2821void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 2822
f001fde5 2823/* Functions used for device addresses handling */
f629d208
JP
2824int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2825 unsigned char addr_type);
2826int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2827 unsigned char addr_type);
2828int dev_addr_add_multiple(struct net_device *to_dev,
2829 struct net_device *from_dev, unsigned char addr_type);
2830int dev_addr_del_multiple(struct net_device *to_dev,
2831 struct net_device *from_dev, unsigned char addr_type);
2832void dev_addr_flush(struct net_device *dev);
2833int dev_addr_init(struct net_device *dev);
a748ee24
JP
2834
2835/* Functions used for unicast addresses handling */
f629d208
JP
2836int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2837int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2838int dev_uc_del(struct net_device *dev, const unsigned char *addr);
2839int dev_uc_sync(struct net_device *to, struct net_device *from);
2840int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
2841void dev_uc_unsync(struct net_device *to, struct net_device *from);
2842void dev_uc_flush(struct net_device *dev);
2843void dev_uc_init(struct net_device *dev);
f001fde5 2844
22bedad3 2845/* Functions used for multicast addresses handling */
f629d208
JP
2846int dev_mc_add(struct net_device *dev, const unsigned char *addr);
2847int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
2848int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
2849int dev_mc_del(struct net_device *dev, const unsigned char *addr);
2850int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
2851int dev_mc_sync(struct net_device *to, struct net_device *from);
2852int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
2853void dev_mc_unsync(struct net_device *to, struct net_device *from);
2854void dev_mc_flush(struct net_device *dev);
2855void dev_mc_init(struct net_device *dev);
f001fde5 2856
4417da66 2857/* Functions used for secondary unicast and multicast support */
f629d208
JP
2858void dev_set_rx_mode(struct net_device *dev);
2859void __dev_set_rx_mode(struct net_device *dev);
2860int dev_set_promiscuity(struct net_device *dev, int inc);
2861int dev_set_allmulti(struct net_device *dev, int inc);
2862void netdev_state_change(struct net_device *dev);
2863void netdev_notify_peers(struct net_device *dev);
2864void netdev_features_change(struct net_device *dev);
1da177e4 2865/* Load a device via the kmod */
f629d208
JP
2866void dev_load(struct net *net, const char *name);
2867struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2868 struct rtnl_link_stats64 *storage);
2869void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
2870 const struct net_device_stats *netdev_stats);
eeda3fd6 2871
1da177e4 2872extern int netdev_max_backlog;
3b098e2d 2873extern int netdev_tstamp_prequeue;
1da177e4 2874extern int weight_p;
0a14842f 2875extern int bpf_jit_enable;
9ff162a8 2876
f629d208
JP
2877bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
2878bool netdev_has_any_upper_dev(struct net_device *dev);
2879struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
2880 struct list_head **iter);
8b5be856
VF
2881
2882/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
2883#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
2884 for (iter = &(dev)->all_adj_list.upper, \
2885 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
2886 updev; \
2887 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 2888
f629d208
JP
2889void *netdev_lower_get_next_private(struct net_device *dev,
2890 struct list_head **iter);
2891void *netdev_lower_get_next_private_rcu(struct net_device *dev,
2892 struct list_head **iter);
31088a11
VF
2893
2894#define netdev_for_each_lower_private(dev, priv, iter) \
2895 for (iter = (dev)->adj_list.lower.next, \
2896 priv = netdev_lower_get_next_private(dev, &(iter)); \
2897 priv; \
2898 priv = netdev_lower_get_next_private(dev, &(iter)))
2899
2900#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
2901 for (iter = &(dev)->adj_list.lower, \
2902 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
2903 priv; \
2904 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
2905
f629d208
JP
2906void *netdev_adjacent_get_private(struct list_head *adj_list);
2907struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
2908struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
2909int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
2910int netdev_master_upper_dev_link(struct net_device *dev,
9ff162a8 2911 struct net_device *upper_dev);
f629d208
JP
2912int netdev_master_upper_dev_link_private(struct net_device *dev,
2913 struct net_device *upper_dev,
2914 void *private);
2915void netdev_upper_dev_unlink(struct net_device *dev,
2916 struct net_device *upper_dev);
2917void *netdev_lower_dev_get_private_rcu(struct net_device *dev,
2918 struct net_device *lower_dev);
2919void *netdev_lower_dev_get_private(struct net_device *dev,
2920 struct net_device *lower_dev);
2921int skb_checksum_help(struct sk_buff *skb);
2922struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
2923 netdev_features_t features, bool tx_path);
2924struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
2925 netdev_features_t features);
12b0004d
CW
2926
2927static inline
2928struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
2929{
2930 return __skb_gso_segment(skb, features, true);
2931}
ec5f0615
PS
2932__be16 skb_network_protocol(struct sk_buff *skb);
2933
2934static inline bool can_checksum_protocol(netdev_features_t features,
2935 __be16 protocol)
2936{
2937 return ((features & NETIF_F_GEN_CSUM) ||
2938 ((features & NETIF_F_V4_CSUM) &&
2939 protocol == htons(ETH_P_IP)) ||
2940 ((features & NETIF_F_V6_CSUM) &&
2941 protocol == htons(ETH_P_IPV6)) ||
2942 ((features & NETIF_F_FCOE_CRC) &&
2943 protocol == htons(ETH_P_FCOE)));
2944}
12b0004d 2945
fb286bb2 2946#ifdef CONFIG_BUG
f629d208 2947void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
2948#else
2949static inline void netdev_rx_csum_fault(struct net_device *dev)
2950{
2951}
2952#endif
1da177e4 2953/* rx skb timestamps */
f629d208
JP
2954void net_enable_timestamp(void);
2955void net_disable_timestamp(void);
1da177e4 2956
20380731 2957#ifdef CONFIG_PROC_FS
f629d208 2958int __init dev_proc_init(void);
900ff8c6
CW
2959#else
2960#define dev_proc_init() 0
20380731
ACM
2961#endif
2962
42a2d923
LT
2963int netdev_class_create_file_ns(struct class_attribute *class_attr,
2964 const void *ns);
2965void netdev_class_remove_file_ns(struct class_attribute *class_attr,
2966 const void *ns);
58292cbe
TH
2967
2968static inline int netdev_class_create_file(struct class_attribute *class_attr)
2969{
2970 return netdev_class_create_file_ns(class_attr, NULL);
2971}
2972
2973static inline void netdev_class_remove_file(struct class_attribute *class_attr)
2974{
2975 netdev_class_remove_file_ns(class_attr, NULL);
2976}
b8a9787e 2977
04600794
JB
2978extern struct kobj_ns_type_operations net_ns_type_operations;
2979
f629d208 2980const char *netdev_drivername(const struct net_device *dev);
6579e57b 2981
f629d208 2982void linkwatch_run_queue(void);
20380731 2983
c8f44aff
MM
2984static inline netdev_features_t netdev_get_wanted_features(
2985 struct net_device *dev)
5455c699
MM
2986{
2987 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2988}
c8f44aff
MM
2989netdev_features_t netdev_increment_features(netdev_features_t all,
2990 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
2991
2992/* Allow TSO being used on stacked device :
2993 * Performing the GSO segmentation before last device
2994 * is a performance improvement.
2995 */
2996static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
2997 netdev_features_t mask)
2998{
2999 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3000}
3001
6cb6a27c 3002int __netdev_update_features(struct net_device *dev);
5455c699 3003void netdev_update_features(struct net_device *dev);
afe12cc8 3004void netdev_change_features(struct net_device *dev);
7f353bf2 3005
fc4a7489
PM
3006void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3007 struct net_device *dev);
3008
c8f44aff 3009netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3010
4d29515f 3011static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3012{
c8f44aff 3013 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3014
3015 /* check flags correspondence */
3016 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3017 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3018 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3019 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3020 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3021 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3022
d6b4991a 3023 return (features & feature) == feature;
576a30eb
HX
3024}
3025
4d29515f 3026static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3027{
278b2513 3028 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3029 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3030}
3031
4d29515f
DM
3032static inline bool netif_needs_gso(struct sk_buff *skb,
3033 netdev_features_t features)
7967168c 3034{
fc741216 3035 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3036 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3037 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3038}
3039
82cc1a7a
PWJ
3040static inline void netif_set_gso_max_size(struct net_device *dev,
3041 unsigned int size)
3042{
3043 dev->gso_max_size = size;
3044}
3045
a6cc0cfa
JF
3046static inline bool netif_is_macvlan(struct net_device *dev)
3047{
3048 return dev->priv_flags & IFF_MACVLAN;
3049}
3050
8a7fbfab 3051static inline bool netif_is_bond_master(struct net_device *dev)
3052{
3053 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3054}
3055
4d29515f 3056static inline bool netif_is_bond_slave(struct net_device *dev)
1765a575
JP
3057{
3058 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3059}
3060
3bdc0eba
BG
3061static inline bool netif_supports_nofcs(struct net_device *dev)
3062{
3063 return dev->priv_flags & IFF_SUPP_NOFCS;
3064}
3065
505d4f73 3066extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 3067
571ba423
JP
3068/* Logging, debugging and troubleshooting/diagnostic helpers. */
3069
3070/* netdev_printk helpers, similar to dev_printk */
3071
3072static inline const char *netdev_name(const struct net_device *dev)
3073{
3074 if (dev->reg_state != NETREG_REGISTERED)
3075 return "(unregistered net_device)";
3076 return dev->name;
3077}
3078
f629d208 3079__printf(3, 4)
b9075fa9
JP
3080int netdev_printk(const char *level, const struct net_device *dev,
3081 const char *format, ...);
f629d208 3082__printf(2, 3)
b9075fa9 3083int netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 3084__printf(2, 3)
b9075fa9 3085int netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 3086__printf(2, 3)
b9075fa9 3087int netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 3088__printf(2, 3)
b9075fa9 3089int netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 3090__printf(2, 3)
b9075fa9 3091int netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 3092__printf(2, 3)
b9075fa9 3093int netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 3094__printf(2, 3)
b9075fa9 3095int netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 3096
8909c9ad
VK
3097#define MODULE_ALIAS_NETDEV(device) \
3098 MODULE_ALIAS("netdev-" device)
3099
b558c96f 3100#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
3101#define netdev_dbg(__dev, format, args...) \
3102do { \
ffa10cb4 3103 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 3104} while (0)
b558c96f
JC
3105#elif defined(DEBUG)
3106#define netdev_dbg(__dev, format, args...) \
3107 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
3108#else
3109#define netdev_dbg(__dev, format, args...) \
3110({ \
3111 if (0) \
3112 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3113 0; \
3114})
3115#endif
3116
3117#if defined(VERBOSE_DEBUG)
3118#define netdev_vdbg netdev_dbg
3119#else
3120
3121#define netdev_vdbg(dev, format, args...) \
3122({ \
3123 if (0) \
3124 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3125 0; \
3126})
3127#endif
3128
3129/*
3130 * netdev_WARN() acts like dev_printk(), but with the key difference
3131 * of using a WARN/WARN_ON to get the message out, including the
3132 * file/line information and a backtrace.
3133 */
3134#define netdev_WARN(dev, format, args...) \
7cc7c5e5 3135 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
571ba423 3136
b3d95c5c
JP
3137/* netif printk helpers, similar to netdev_printk */
3138
3139#define netif_printk(priv, type, level, dev, fmt, args...) \
3140do { \
3141 if (netif_msg_##type(priv)) \
3142 netdev_printk(level, (dev), fmt, ##args); \
3143} while (0)
3144
f45f4321
JP
3145#define netif_level(level, priv, type, dev, fmt, args...) \
3146do { \
3147 if (netif_msg_##type(priv)) \
3148 netdev_##level(dev, fmt, ##args); \
3149} while (0)
3150
b3d95c5c 3151#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 3152 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 3153#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 3154 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 3155#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 3156 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 3157#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 3158 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 3159#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 3160 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 3161#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 3162 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 3163#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 3164 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 3165
0053ea9c 3166#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
3167#define netif_dbg(priv, type, netdev, format, args...) \
3168do { \
3169 if (netif_msg_##type(priv)) \
b5fb0a03 3170 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 3171} while (0)
0053ea9c
JP
3172#elif defined(DEBUG)
3173#define netif_dbg(priv, type, dev, format, args...) \
3174 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
3175#else
3176#define netif_dbg(priv, type, dev, format, args...) \
3177({ \
3178 if (0) \
3179 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3180 0; \
3181})
3182#endif
3183
3184#if defined(VERBOSE_DEBUG)
bcfcc450 3185#define netif_vdbg netif_dbg
b3d95c5c
JP
3186#else
3187#define netif_vdbg(priv, type, dev, format, args...) \
3188({ \
3189 if (0) \
a4ed89cb 3190 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
3191 0; \
3192})
3193#endif
571ba423 3194
900ff8c6
CW
3195/*
3196 * The list of packet types we will receive (as opposed to discard)
3197 * and the routines to invoke.
3198 *
3199 * Why 16. Because with 16 the only overlap we get on a hash of the
3200 * low nibble of the protocol value is RARP/SNAP/X.25.
3201 *
3202 * NOTE: That is no longer true with the addition of VLAN tags. Not
3203 * sure which should go first, but I bet it won't make much
3204 * difference if we are running VLANs. The good news is that
3205 * this protocol won't be in the list unless compiled in, so
3206 * the average user (w/out VLANs) will not be adversely affected.
3207 * --BLG
3208 *
3209 * 0800 IP
3210 * 8100 802.1Q VLAN
3211 * 0001 802.3
3212 * 0002 AX.25
3213 * 0004 802.2
3214 * 8035 RARP
3215 * 0005 SNAP
3216 * 0805 X.25
3217 * 0806 ARP
3218 * 8137 IPX
3219 * 0009 Localtalk
3220 * 86DD IPv6
3221 */
3222#define PTYPE_HASH_SIZE (16)
3223#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3224
385a154c 3225#endif /* _LINUX_NETDEVICE_H */