net: use reciprocal_scale() helper
[deliverable/linux.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 59
f629d208
JP
60void netdev_set_default_ethtool_ops(struct net_device *dev,
61 const struct ethtool_ops *ops);
d07d7507 62
9a1654ba
JP
63/* Backlog congestion levels */
64#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
65#define NET_RX_DROP 1 /* packet dropped */
66
572a9d7b
PM
67/*
68 * Transmit return codes: transmit return codes originate from three different
69 * namespaces:
70 *
71 * - qdisc return codes
72 * - driver transmit return codes
73 * - errno values
74 *
75 * Drivers are allowed to return any one of those in their hard_start_xmit()
76 * function. Real network devices commonly used with qdiscs should only return
77 * the driver transmit return codes though - when qdiscs are used, the actual
78 * transmission happens asynchronously, so the value is not propagated to
79 * higher layers. Virtual network devices transmit synchronously, in this case
80 * the driver transmit return codes are consumed by dev_queue_xmit(), all
81 * others are propagated to higher layers.
82 */
83
84/* qdisc ->enqueue() return codes. */
85#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
86#define NET_XMIT_DROP 0x01 /* skb dropped */
87#define NET_XMIT_CN 0x02 /* congestion notification */
88#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
89#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 90
b9df3cb8
GR
91/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
92 * indicates that the device will soon be dropping packets, or already drops
93 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 94#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
95#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
96
dc1f8bf6 97/* Driver transmit return codes */
9a1654ba 98#define NETDEV_TX_MASK 0xf0
572a9d7b 99
dc1f8bf6 100enum netdev_tx {
572a9d7b 101 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
102 NETDEV_TX_OK = 0x00, /* driver took care of packet */
103 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
104 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
dc1f8bf6
SH
105};
106typedef enum netdev_tx netdev_tx_t;
107
9a1654ba
JP
108/*
109 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
110 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
111 */
112static inline bool dev_xmit_complete(int rc)
113{
114 /*
115 * Positive cases with an skb consumed by a driver:
116 * - successful transmission (rc == NETDEV_TX_OK)
117 * - error while transmitting (rc < 0)
118 * - error while queueing to a different device (rc & NET_XMIT_MASK)
119 */
120 if (likely(rc < NET_XMIT_MASK))
121 return true;
122
123 return false;
124}
125
1da177e4
LT
126/*
127 * Compute the worst case header length according to the protocols
128 * used.
129 */
fe2918b0 130
d11ead75 131#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
132# if defined(CONFIG_MAC80211_MESH)
133# define LL_MAX_HEADER 128
134# else
135# define LL_MAX_HEADER 96
136# endif
1da177e4 137#else
8388e3da 138# define LL_MAX_HEADER 32
1da177e4
LT
139#endif
140
d11ead75
BH
141#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
142 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
143#define MAX_HEADER LL_MAX_HEADER
144#else
145#define MAX_HEADER (LL_MAX_HEADER + 48)
146#endif
147
148/*
be1f3c2c
BH
149 * Old network device statistics. Fields are native words
150 * (unsigned long) so they can be read and written atomically.
1da177e4 151 */
fe2918b0 152
d94d9fee 153struct net_device_stats {
3cfde79c
BH
154 unsigned long rx_packets;
155 unsigned long tx_packets;
156 unsigned long rx_bytes;
157 unsigned long tx_bytes;
158 unsigned long rx_errors;
159 unsigned long tx_errors;
160 unsigned long rx_dropped;
161 unsigned long tx_dropped;
162 unsigned long multicast;
1da177e4 163 unsigned long collisions;
1da177e4 164 unsigned long rx_length_errors;
3cfde79c
BH
165 unsigned long rx_over_errors;
166 unsigned long rx_crc_errors;
167 unsigned long rx_frame_errors;
168 unsigned long rx_fifo_errors;
169 unsigned long rx_missed_errors;
1da177e4
LT
170 unsigned long tx_aborted_errors;
171 unsigned long tx_carrier_errors;
172 unsigned long tx_fifo_errors;
173 unsigned long tx_heartbeat_errors;
174 unsigned long tx_window_errors;
1da177e4
LT
175 unsigned long rx_compressed;
176 unsigned long tx_compressed;
177};
178
1da177e4
LT
179
180#include <linux/cache.h>
181#include <linux/skbuff.h>
182
adc9300e 183#ifdef CONFIG_RPS
c5905afb
IM
184#include <linux/static_key.h>
185extern struct static_key rps_needed;
adc9300e
ED
186#endif
187
1da177e4
LT
188struct neighbour;
189struct neigh_parms;
190struct sk_buff;
191
f001fde5
JP
192struct netdev_hw_addr {
193 struct list_head list;
194 unsigned char addr[MAX_ADDR_LEN];
195 unsigned char type;
ccffad25
JP
196#define NETDEV_HW_ADDR_T_LAN 1
197#define NETDEV_HW_ADDR_T_SAN 2
198#define NETDEV_HW_ADDR_T_SLAVE 3
199#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 200#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 201 bool global_use;
4cd729b0 202 int sync_cnt;
8f8f103d 203 int refcount;
4543fbef 204 int synced;
f001fde5
JP
205 struct rcu_head rcu_head;
206};
207
31278e71
JP
208struct netdev_hw_addr_list {
209 struct list_head list;
210 int count;
211};
212
22bedad3
JP
213#define netdev_hw_addr_list_count(l) ((l)->count)
214#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
215#define netdev_hw_addr_list_for_each(ha, l) \
216 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 217
22bedad3
JP
218#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
219#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
220#define netdev_for_each_uc_addr(ha, dev) \
221 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 222
22bedad3
JP
223#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
224#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 225#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 226 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 227
d94d9fee 228struct hh_cache {
f6b72b62 229 u16 hh_len;
5c25f686 230 u16 __pad;
3644f0ce 231 seqlock_t hh_lock;
1da177e4
LT
232
233 /* cached hardware header; allow for machine alignment needs. */
234#define HH_DATA_MOD 16
235#define HH_DATA_OFF(__len) \
5ba0eac6 236 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
237#define HH_DATA_ALIGN(__len) \
238 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
239 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
240};
241
242/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
243 * Alternative is:
244 * dev->hard_header_len ? (dev->hard_header_len +
245 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
246 *
247 * We could use other alignment values, but we must maintain the
248 * relationship HH alignment <= LL alignment.
249 */
250#define LL_RESERVED_SPACE(dev) \
f5184d26 251 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 252#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 253 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 254
3b04ddde
SH
255struct header_ops {
256 int (*create) (struct sk_buff *skb, struct net_device *dev,
257 unsigned short type, const void *daddr,
95c96174 258 const void *saddr, unsigned int len);
3b04ddde
SH
259 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
260 int (*rebuild)(struct sk_buff *skb);
e69dd336 261 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
262 void (*cache_update)(struct hh_cache *hh,
263 const struct net_device *dev,
264 const unsigned char *haddr);
265};
266
1da177e4
LT
267/* These flag bits are private to the generic network queueing
268 * layer, they may not be explicitly referenced by any other
269 * code.
270 */
271
d94d9fee 272enum netdev_state_t {
1da177e4
LT
273 __LINK_STATE_START,
274 __LINK_STATE_PRESENT,
1da177e4 275 __LINK_STATE_NOCARRIER,
b00055aa
SR
276 __LINK_STATE_LINKWATCH_PENDING,
277 __LINK_STATE_DORMANT,
1da177e4
LT
278};
279
280
281/*
282 * This structure holds at boot time configured netdevice settings. They
fe2918b0 283 * are then used in the device probing.
1da177e4
LT
284 */
285struct netdev_boot_setup {
286 char name[IFNAMSIZ];
287 struct ifmap map;
288};
289#define NETDEV_BOOT_SETUP_MAX 8
290
f629d208 291int __init netdev_boot_setup(char *str);
1da177e4 292
bea3348e
SH
293/*
294 * Structure for NAPI scheduling similar to tasklet but with weighting
295 */
296struct napi_struct {
297 /* The poll_list must only be managed by the entity which
298 * changes the state of the NAPI_STATE_SCHED bit. This means
299 * whoever atomically sets that bit can add this napi_struct
300 * to the per-cpu poll_list, and whoever clears that bit
301 * can remove from the list right before clearing the bit.
302 */
303 struct list_head poll_list;
304
305 unsigned long state;
306 int weight;
404f7c9e 307 unsigned int gro_count;
bea3348e
SH
308 int (*poll)(struct napi_struct *, int);
309#ifdef CONFIG_NETPOLL
310 spinlock_t poll_lock;
311 int poll_owner;
bea3348e 312#endif
5d38a079 313 struct net_device *dev;
d565b0a1 314 struct sk_buff *gro_list;
5d38a079 315 struct sk_buff *skb;
404f7c9e 316 struct list_head dev_list;
af12fa6e
ET
317 struct hlist_node napi_hash_node;
318 unsigned int napi_id;
bea3348e
SH
319};
320
d94d9fee 321enum {
bea3348e 322 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 323 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 324 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
af12fa6e 325 NAPI_STATE_HASHED, /* In NAPI hash */
bea3348e
SH
326};
327
5b252f0c 328enum gro_result {
d1c76af9
HX
329 GRO_MERGED,
330 GRO_MERGED_FREE,
331 GRO_HELD,
332 GRO_NORMAL,
333 GRO_DROP,
334};
5b252f0c 335typedef enum gro_result gro_result_t;
d1c76af9 336
8a4eb573
JP
337/*
338 * enum rx_handler_result - Possible return values for rx_handlers.
339 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
340 * further.
341 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
342 * case skb->dev was changed by rx_handler.
343 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
344 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
345 *
346 * rx_handlers are functions called from inside __netif_receive_skb(), to do
347 * special processing of the skb, prior to delivery to protocol handlers.
348 *
349 * Currently, a net_device can only have a single rx_handler registered. Trying
350 * to register a second rx_handler will return -EBUSY.
351 *
352 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
353 * To unregister a rx_handler on a net_device, use
354 * netdev_rx_handler_unregister().
355 *
356 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
357 * do with the skb.
358 *
359 * If the rx_handler consumed to skb in some way, it should return
360 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
361 * the skb to be delivered in some other ways.
362 *
363 * If the rx_handler changed skb->dev, to divert the skb to another
364 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
365 * new device will be called if it exists.
366 *
367 * If the rx_handler consider the skb should be ignored, it should return
368 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 369 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
370 *
371 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
372 * delivered, it should return RX_HANDLER_PASS.
373 *
374 * A device without a registered rx_handler will behave as if rx_handler
375 * returned RX_HANDLER_PASS.
376 */
377
378enum rx_handler_result {
379 RX_HANDLER_CONSUMED,
380 RX_HANDLER_ANOTHER,
381 RX_HANDLER_EXACT,
382 RX_HANDLER_PASS,
383};
384typedef enum rx_handler_result rx_handler_result_t;
385typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 386
f629d208 387void __napi_schedule(struct napi_struct *n);
bea3348e 388
4d29515f 389static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
390{
391 return test_bit(NAPI_STATE_DISABLE, &n->state);
392}
393
bea3348e
SH
394/**
395 * napi_schedule_prep - check if napi can be scheduled
396 * @n: napi context
397 *
398 * Test if NAPI routine is already running, and if not mark
399 * it as running. This is used as a condition variable
a0a46196
DM
400 * insure only one NAPI poll instance runs. We also make
401 * sure there is no pending NAPI disable.
bea3348e 402 */
4d29515f 403static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 404{
a0a46196
DM
405 return !napi_disable_pending(n) &&
406 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
407}
408
409/**
410 * napi_schedule - schedule NAPI poll
411 * @n: napi context
412 *
413 * Schedule NAPI poll routine to be called if it is not already
414 * running.
415 */
416static inline void napi_schedule(struct napi_struct *n)
417{
418 if (napi_schedule_prep(n))
419 __napi_schedule(n);
420}
421
bfe13f54 422/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 423static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
424{
425 if (napi_schedule_prep(napi)) {
426 __napi_schedule(napi);
4d29515f 427 return true;
bfe13f54 428 }
4d29515f 429 return false;
bfe13f54
RD
430}
431
bea3348e
SH
432/**
433 * napi_complete - NAPI processing complete
434 * @n: napi context
435 *
436 * Mark NAPI processing as complete.
437 */
f629d208
JP
438void __napi_complete(struct napi_struct *n);
439void napi_complete(struct napi_struct *n);
bea3348e 440
af12fa6e
ET
441/**
442 * napi_by_id - lookup a NAPI by napi_id
443 * @napi_id: hashed napi_id
444 *
445 * lookup @napi_id in napi_hash table
446 * must be called under rcu_read_lock()
447 */
f629d208 448struct napi_struct *napi_by_id(unsigned int napi_id);
af12fa6e
ET
449
450/**
451 * napi_hash_add - add a NAPI to global hashtable
452 * @napi: napi context
453 *
454 * generate a new napi_id and store a @napi under it in napi_hash
455 */
f629d208 456void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
457
458/**
459 * napi_hash_del - remove a NAPI from global table
460 * @napi: napi context
461 *
462 * Warning: caller must observe rcu grace period
463 * before freeing memory containing @napi
464 */
f629d208 465void napi_hash_del(struct napi_struct *napi);
af12fa6e 466
bea3348e
SH
467/**
468 * napi_disable - prevent NAPI from scheduling
469 * @n: napi context
470 *
471 * Stop NAPI from being scheduled on this context.
472 * Waits till any outstanding processing completes.
473 */
474static inline void napi_disable(struct napi_struct *n)
475{
80c33ddd 476 might_sleep();
a0a46196 477 set_bit(NAPI_STATE_DISABLE, &n->state);
bea3348e 478 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
43cc7380 479 msleep(1);
a0a46196 480 clear_bit(NAPI_STATE_DISABLE, &n->state);
bea3348e
SH
481}
482
483/**
484 * napi_enable - enable NAPI scheduling
485 * @n: napi context
486 *
487 * Resume NAPI from being scheduled on this context.
488 * Must be paired with napi_disable.
489 */
490static inline void napi_enable(struct napi_struct *n)
491{
492 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 493 smp_mb__before_atomic();
bea3348e
SH
494 clear_bit(NAPI_STATE_SCHED, &n->state);
495}
496
c264c3de
SH
497#ifdef CONFIG_SMP
498/**
499 * napi_synchronize - wait until NAPI is not running
500 * @n: napi context
501 *
502 * Wait until NAPI is done being scheduled on this context.
503 * Waits till any outstanding processing completes but
504 * does not disable future activations.
505 */
506static inline void napi_synchronize(const struct napi_struct *n)
507{
508 while (test_bit(NAPI_STATE_SCHED, &n->state))
509 msleep(1);
510}
511#else
512# define napi_synchronize(n) barrier()
513#endif
514
d94d9fee 515enum netdev_queue_state_t {
73466498
TH
516 __QUEUE_STATE_DRV_XOFF,
517 __QUEUE_STATE_STACK_XOFF,
c3f26a26 518 __QUEUE_STATE_FROZEN,
79d16385 519};
8e2f1a63
DB
520
521#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
522#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
523#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
524
525#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
526#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
527 QUEUE_STATE_FROZEN)
528#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
529 QUEUE_STATE_FROZEN)
530
73466498
TH
531/*
532 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
533 * netif_tx_* functions below are used to manipulate this flag. The
534 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
535 * queue independently. The netif_xmit_*stopped functions below are called
536 * to check if the queue has been stopped by the driver or stack (either
537 * of the XOFF bits are set in the state). Drivers should not need to call
538 * netif_xmit*stopped functions, they should only be using netif_tx_*.
539 */
79d16385 540
bb949fbd 541struct netdev_queue {
6a321cb3
ED
542/*
543 * read mostly part
544 */
bb949fbd 545 struct net_device *dev;
b0e1e646
DM
546 struct Qdisc *qdisc;
547 struct Qdisc *qdisc_sleeping;
ccf5ff69 548#ifdef CONFIG_SYSFS
1d24eb48
TH
549 struct kobject kobj;
550#endif
f2cd2d3e
ED
551#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
552 int numa_node;
553#endif
6a321cb3
ED
554/*
555 * write mostly part
556 */
557 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
558 int xmit_lock_owner;
9d21493b
ED
559 /*
560 * please use this field instead of dev->trans_start
561 */
562 unsigned long trans_start;
ccf5ff69 563
564 /*
565 * Number of TX timeouts for this queue
566 * (/sys/class/net/DEV/Q/trans_timeout)
567 */
568 unsigned long trans_timeout;
114cf580
TH
569
570 unsigned long state;
571
572#ifdef CONFIG_BQL
573 struct dql dql;
574#endif
e8a0464c 575} ____cacheline_aligned_in_smp;
bb949fbd 576
f2cd2d3e
ED
577static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
578{
579#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
580 return q->numa_node;
581#else
b236da69 582 return NUMA_NO_NODE;
f2cd2d3e
ED
583#endif
584}
585
586static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
587{
588#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
589 q->numa_node = node;
590#endif
591}
592
df334545 593#ifdef CONFIG_RPS
0a9627f2
TH
594/*
595 * This structure holds an RPS map which can be of variable length. The
596 * map is an array of CPUs.
597 */
598struct rps_map {
599 unsigned int len;
600 struct rcu_head rcu;
601 u16 cpus[0];
602};
60b778ce 603#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 604
fec5e652 605/*
c445477d
BH
606 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
607 * tail pointer for that CPU's input queue at the time of last enqueue, and
608 * a hardware filter index.
fec5e652
TH
609 */
610struct rps_dev_flow {
611 u16 cpu;
c445477d 612 u16 filter;
fec5e652
TH
613 unsigned int last_qtail;
614};
c445477d 615#define RPS_NO_FILTER 0xffff
fec5e652
TH
616
617/*
618 * The rps_dev_flow_table structure contains a table of flow mappings.
619 */
620struct rps_dev_flow_table {
621 unsigned int mask;
622 struct rcu_head rcu;
fec5e652
TH
623 struct rps_dev_flow flows[0];
624};
625#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 626 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
627
628/*
629 * The rps_sock_flow_table contains mappings of flows to the last CPU
630 * on which they were processed by the application (set in recvmsg).
631 */
632struct rps_sock_flow_table {
633 unsigned int mask;
634 u16 ents[0];
635};
636#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
60b778ce 637 ((_num) * sizeof(u16)))
fec5e652
TH
638
639#define RPS_NO_CPU 0xffff
640
641static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
642 u32 hash)
643{
644 if (table && hash) {
645 unsigned int cpu, index = hash & table->mask;
646
647 /* We only give a hint, preemption can change cpu under us */
648 cpu = raw_smp_processor_id();
649
650 if (table->ents[index] != cpu)
651 table->ents[index] = cpu;
652 }
653}
654
655static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
656 u32 hash)
657{
658 if (table && hash)
659 table->ents[hash & table->mask] = RPS_NO_CPU;
660}
661
6e3f7faf 662extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
fec5e652 663
c445477d 664#ifdef CONFIG_RFS_ACCEL
f629d208
JP
665bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
666 u16 filter_id);
c445477d 667#endif
a953be53 668#endif /* CONFIG_RPS */
c445477d 669
0a9627f2
TH
670/* This structure contains an instance of an RX queue. */
671struct netdev_rx_queue {
a953be53 672#ifdef CONFIG_RPS
6e3f7faf
ED
673 struct rps_map __rcu *rps_map;
674 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 675#endif
6e3f7faf 676 struct kobject kobj;
fe822240 677 struct net_device *dev;
0a9627f2 678} ____cacheline_aligned_in_smp;
a953be53
MD
679
680/*
681 * RX queue sysfs structures and functions.
682 */
683struct rx_queue_attribute {
684 struct attribute attr;
685 ssize_t (*show)(struct netdev_rx_queue *queue,
686 struct rx_queue_attribute *attr, char *buf);
687 ssize_t (*store)(struct netdev_rx_queue *queue,
688 struct rx_queue_attribute *attr, const char *buf, size_t len);
689};
d314774c 690
bf264145
TH
691#ifdef CONFIG_XPS
692/*
693 * This structure holds an XPS map which can be of variable length. The
694 * map is an array of queues.
695 */
696struct xps_map {
697 unsigned int len;
698 unsigned int alloc_len;
699 struct rcu_head rcu;
700 u16 queues[0];
701};
60b778ce 702#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
bf264145
TH
703#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
704 / sizeof(u16))
705
706/*
707 * This structure holds all XPS maps for device. Maps are indexed by CPU.
708 */
709struct xps_dev_maps {
710 struct rcu_head rcu;
a4177869 711 struct xps_map __rcu *cpu_map[0];
bf264145
TH
712};
713#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
714 (nr_cpu_ids * sizeof(struct xps_map *)))
715#endif /* CONFIG_XPS */
716
4f57c087
JF
717#define TC_MAX_QUEUE 16
718#define TC_BITMASK 15
719/* HW offloaded queuing disciplines txq count and offset maps */
720struct netdev_tc_txq {
721 u16 count;
722 u16 offset;
723};
724
68bad94e
NP
725#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
726/*
727 * This structure is to hold information about the device
728 * configured to run FCoE protocol stack.
729 */
730struct netdev_fcoe_hbainfo {
731 char manufacturer[64];
732 char serial_number[64];
733 char hardware_version[64];
734 char driver_version[64];
735 char optionrom_version[64];
736 char firmware_version[64];
737 char model[256];
738 char model_description[256];
739};
740#endif
741
66b52b0d
JP
742#define MAX_PHYS_PORT_ID_LEN 32
743
744/* This structure holds a unique identifier to identify the
745 * physical port used by a netdevice.
746 */
747struct netdev_phys_port_id {
748 unsigned char id[MAX_PHYS_PORT_ID_LEN];
749 unsigned char id_len;
750};
751
99932d4f
DB
752typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
753 struct sk_buff *skb);
754
d314774c
SH
755/*
756 * This structure defines the management hooks for network devices.
00829823
SH
757 * The following hooks can be defined; unless noted otherwise, they are
758 * optional and can be filled with a null pointer.
d314774c
SH
759 *
760 * int (*ndo_init)(struct net_device *dev);
761 * This function is called once when network device is registered.
762 * The network device can use this to any late stage initializaton
763 * or semantic validattion. It can fail with an error code which will
764 * be propogated back to register_netdev
765 *
766 * void (*ndo_uninit)(struct net_device *dev);
767 * This function is called when device is unregistered or when registration
768 * fails. It is not called if init fails.
769 *
770 * int (*ndo_open)(struct net_device *dev);
771 * This function is called when network device transistions to the up
772 * state.
773 *
774 * int (*ndo_stop)(struct net_device *dev);
775 * This function is called when network device transistions to the down
776 * state.
777 *
dc1f8bf6
SH
778 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
779 * struct net_device *dev);
00829823 780 * Called when a packet needs to be transmitted.
dc1f8bf6
SH
781 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
782 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
783 * Required can not be NULL.
784 *
f663dd9a 785 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 786 * void *accel_priv, select_queue_fallback_t fallback);
00829823
SH
787 * Called to decide which queue to when device supports multiple
788 * transmit queues.
789 *
d314774c
SH
790 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
791 * This function is called to allow device receiver to make
792 * changes to configuration when multicast or promiscious is enabled.
793 *
794 * void (*ndo_set_rx_mode)(struct net_device *dev);
795 * This function is called device changes address list filtering.
01789349
JP
796 * If driver handles unicast address filtering, it should set
797 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
798 *
799 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
800 * This function is called when the Media Access Control address
37b607c5 801 * needs to be changed. If this interface is not defined, the
d314774c
SH
802 * mac address can not be changed.
803 *
804 * int (*ndo_validate_addr)(struct net_device *dev);
805 * Test if Media Access Control address is valid for the device.
806 *
807 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
808 * Called when a user request an ioctl which can't be handled by
809 * the generic interface code. If not defined ioctl's return
810 * not supported error code.
811 *
812 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
813 * Used to set network devices bus interface parameters. This interface
814 * is retained for legacy reason, new devices should use the bus
815 * interface (PCI) for low level management.
816 *
817 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
818 * Called when a user wants to change the Maximum Transfer Unit
819 * of a device. If not defined, any request to change MTU will
820 * will return an error.
821 *
00829823 822 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
823 * Callback uses when the transmitter has not made any progress
824 * for dev->watchdog ticks.
825 *
3cfde79c 826 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 827 * struct rtnl_link_stats64 *storage);
d308e38f 828 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 829 * Called when a user wants to get the network device usage
be1f3c2c 830 * statistics. Drivers must do one of the following:
3cfde79c
BH
831 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
832 * rtnl_link_stats64 structure passed by the caller.
82695d9b 833 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
834 * (which should normally be dev->stats) and return a pointer to
835 * it. The structure may be changed asynchronously only if each
836 * field is written atomically.
837 * 3. Update dev->stats asynchronously and atomically, and define
838 * neither operation.
d314774c 839 *
80d5c368
PM
840 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
841 * If device support VLAN filtering this function is called when a
842 * VLAN id is registered.
d314774c 843 *
8e586137 844 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
80d5c368
PM
845 * If device support VLAN filtering this function is called when a
846 * VLAN id is unregistered.
d314774c
SH
847 *
848 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
849 *
850 * SR-IOV management functions.
851 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
852 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
853 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
854 * int max_tx_rate);
5f8444a3 855 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
856 * int (*ndo_get_vf_config)(struct net_device *dev,
857 * int vf, struct ifla_vf_info *ivf);
1d8faf48 858 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
859 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
860 * struct nlattr *port[]);
861 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
862 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
863 * Called to setup 'tc' number of traffic classes in the net device. This
864 * is always called from the stack with the rtnl lock held and netif tx
865 * queues stopped. This allows the netdevice to perform queue management
866 * safely.
c445477d 867 *
e9bce845
YZ
868 * Fiber Channel over Ethernet (FCoE) offload functions.
869 * int (*ndo_fcoe_enable)(struct net_device *dev);
870 * Called when the FCoE protocol stack wants to start using LLD for FCoE
871 * so the underlying device can perform whatever needed configuration or
872 * initialization to support acceleration of FCoE traffic.
873 *
874 * int (*ndo_fcoe_disable)(struct net_device *dev);
875 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
876 * so the underlying device can perform whatever needed clean-ups to
877 * stop supporting acceleration of FCoE traffic.
878 *
879 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
880 * struct scatterlist *sgl, unsigned int sgc);
881 * Called when the FCoE Initiator wants to initialize an I/O that
882 * is a possible candidate for Direct Data Placement (DDP). The LLD can
883 * perform necessary setup and returns 1 to indicate the device is set up
884 * successfully to perform DDP on this I/O, otherwise this returns 0.
885 *
886 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
887 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
888 * indicated by the FC exchange id 'xid', so the underlying device can
889 * clean up and reuse resources for later DDP requests.
890 *
891 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
892 * struct scatterlist *sgl, unsigned int sgc);
893 * Called when the FCoE Target wants to initialize an I/O that
894 * is a possible candidate for Direct Data Placement (DDP). The LLD can
895 * perform necessary setup and returns 1 to indicate the device is set up
896 * successfully to perform DDP on this I/O, otherwise this returns 0.
897 *
68bad94e
NP
898 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
899 * struct netdev_fcoe_hbainfo *hbainfo);
900 * Called when the FCoE Protocol stack wants information on the underlying
901 * device. This information is utilized by the FCoE protocol stack to
902 * register attributes with Fiber Channel management service as per the
903 * FC-GS Fabric Device Management Information(FDMI) specification.
904 *
e9bce845
YZ
905 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
906 * Called when the underlying device wants to override default World Wide
907 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
908 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
909 * protocol stack to use.
910 *
c445477d
BH
911 * RFS acceleration.
912 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
913 * u16 rxq_index, u32 flow_id);
914 * Set hardware filter for RFS. rxq_index is the target queue index;
915 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
916 * Return the filter ID on success, or a negative error code.
fbaec0ea 917 *
8b98a70c 918 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
919 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
920 * Called to make another netdev an underling.
921 *
922 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
923 * Called to release previously enslaved netdev.
5455c699
MM
924 *
925 * Feature/offload setting functions.
c8f44aff
MM
926 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
927 * netdev_features_t features);
5455c699
MM
928 * Adjusts the requested feature flags according to device-specific
929 * constraints, and returns the resulting flags. Must not modify
930 * the device state.
931 *
c8f44aff 932 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
933 * Called to update device configuration to new features. Passed
934 * feature set might be less than what was returned by ndo_fix_features()).
935 * Must return >0 or -errno if it changed dev->features itself.
936 *
edc7d573 937 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
938 * struct net_device *dev,
6b6e2725 939 * const unsigned char *addr, u16 flags)
77162022 940 * Adds an FDB entry to dev for addr.
1690be63
VY
941 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
942 * struct net_device *dev,
6b6e2725 943 * const unsigned char *addr)
77162022
JF
944 * Deletes the FDB entry from dev coresponding to addr.
945 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
946 * struct net_device *dev, struct net_device *filter_dev,
947 * int idx)
77162022
JF
948 * Used to add FDB entries to dump requests. Implementers should add
949 * entries to skb and update idx with the number of entries.
e5a55a89
JF
950 *
951 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
952 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
24f11a5c 953 * struct net_device *dev, u32 filter_mask)
4bf84c35
JP
954 *
955 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
956 * Called to change device carrier. Soft-devices (like dummy, team, etc)
957 * which do not represent real hardware may define this to allow their
958 * userspace components to manage their virtual carrier state. Devices
959 * that determine carrier state from physical hardware properties (eg
960 * network cables) or protocol-dependent mechanisms (eg
961 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
962 *
963 * int (*ndo_get_phys_port_id)(struct net_device *dev,
964 * struct netdev_phys_port_id *ppid);
965 * Called to get ID of physical port of this device. If driver does
966 * not implement this, it is assumed that the hw is not able to have
967 * multiple net devices on single physical port.
53cf5275
JG
968 *
969 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 970 * sa_family_t sa_family, __be16 port);
53cf5275
JG
971 * Called by vxlan to notiy a driver about the UDP port and socket
972 * address family that vxlan is listnening to. It is called only when
973 * a new port starts listening. The operation is protected by the
974 * vxlan_net->sock_lock.
975 *
976 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 977 * sa_family_t sa_family, __be16 port);
53cf5275
JG
978 * Called by vxlan to notify the driver about a UDP port and socket
979 * address family that vxlan is not listening to anymore. The operation
980 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
981 *
982 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
983 * struct net_device *dev)
984 * Called by upper layer devices to accelerate switching or other
985 * station functionality into hardware. 'pdev is the lowerdev
986 * to use for the offload and 'dev' is the net device that will
987 * back the offload. Returns a pointer to the private structure
988 * the upper layer will maintain.
989 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
990 * Called by upper layer device to delete the station created
991 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
992 * the station and priv is the structure returned by the add
993 * operation.
994 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
995 * struct net_device *dev,
996 * void *priv);
997 * Callback to use for xmit over the accelerated station. This
998 * is used in place of ndo_start_xmit on accelerated net
999 * devices.
d314774c
SH
1000 */
1001struct net_device_ops {
1002 int (*ndo_init)(struct net_device *dev);
1003 void (*ndo_uninit)(struct net_device *dev);
1004 int (*ndo_open)(struct net_device *dev);
1005 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 1006 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
1007 struct net_device *dev);
1008 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1009 struct sk_buff *skb,
99932d4f
DB
1010 void *accel_priv,
1011 select_queue_fallback_t fallback);
d314774c
SH
1012 void (*ndo_change_rx_flags)(struct net_device *dev,
1013 int flags);
d314774c 1014 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1015 int (*ndo_set_mac_address)(struct net_device *dev,
1016 void *addr);
d314774c 1017 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1018 int (*ndo_do_ioctl)(struct net_device *dev,
1019 struct ifreq *ifr, int cmd);
d314774c
SH
1020 int (*ndo_set_config)(struct net_device *dev,
1021 struct ifmap *map);
00829823
SH
1022 int (*ndo_change_mtu)(struct net_device *dev,
1023 int new_mtu);
1024 int (*ndo_neigh_setup)(struct net_device *dev,
1025 struct neigh_parms *);
d314774c
SH
1026 void (*ndo_tx_timeout) (struct net_device *dev);
1027
28172739
ED
1028 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1029 struct rtnl_link_stats64 *storage);
d314774c
SH
1030 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1031
8e586137 1032 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1033 __be16 proto, u16 vid);
8e586137 1034 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1035 __be16 proto, u16 vid);
d314774c 1036#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1037 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1038 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1039 struct netpoll_info *info);
0e34e931 1040 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1041#endif
e0d1095a 1042#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1043 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1044#endif
95c26df8
WM
1045 int (*ndo_set_vf_mac)(struct net_device *dev,
1046 int queue, u8 *mac);
1047 int (*ndo_set_vf_vlan)(struct net_device *dev,
1048 int queue, u16 vlan, u8 qos);
ed616689
SC
1049 int (*ndo_set_vf_rate)(struct net_device *dev,
1050 int vf, int min_tx_rate,
1051 int max_tx_rate);
5f8444a3
GR
1052 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1053 int vf, bool setting);
95c26df8
WM
1054 int (*ndo_get_vf_config)(struct net_device *dev,
1055 int vf,
1056 struct ifla_vf_info *ivf);
1d8faf48
RE
1057 int (*ndo_set_vf_link_state)(struct net_device *dev,
1058 int vf, int link_state);
57b61080
SF
1059 int (*ndo_set_vf_port)(struct net_device *dev,
1060 int vf,
1061 struct nlattr *port[]);
1062 int (*ndo_get_vf_port)(struct net_device *dev,
1063 int vf, struct sk_buff *skb);
4f57c087 1064 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
d11ead75 1065#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1066 int (*ndo_fcoe_enable)(struct net_device *dev);
1067 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1068 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1069 u16 xid,
1070 struct scatterlist *sgl,
1071 unsigned int sgc);
1072 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1073 u16 xid);
6247e086
YZ
1074 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1075 u16 xid,
1076 struct scatterlist *sgl,
1077 unsigned int sgc);
68bad94e
NP
1078 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1079 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1080#endif
1081
d11ead75 1082#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1083#define NETDEV_FCOE_WWNN 0
1084#define NETDEV_FCOE_WWPN 1
1085 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1086 u64 *wwn, int type);
4d288d57 1087#endif
3c9c36bc 1088
c445477d
BH
1089#ifdef CONFIG_RFS_ACCEL
1090 int (*ndo_rx_flow_steer)(struct net_device *dev,
1091 const struct sk_buff *skb,
1092 u16 rxq_index,
1093 u32 flow_id);
1094#endif
fbaec0ea
JP
1095 int (*ndo_add_slave)(struct net_device *dev,
1096 struct net_device *slave_dev);
1097 int (*ndo_del_slave)(struct net_device *dev,
1098 struct net_device *slave_dev);
c8f44aff
MM
1099 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1100 netdev_features_t features);
5455c699 1101 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1102 netdev_features_t features);
da6a8fa0 1103 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1104 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1105
1106 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1107 struct nlattr *tb[],
77162022 1108 struct net_device *dev,
6b6e2725 1109 const unsigned char *addr,
77162022
JF
1110 u16 flags);
1111 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1112 struct nlattr *tb[],
77162022 1113 struct net_device *dev,
6b6e2725 1114 const unsigned char *addr);
77162022
JF
1115 int (*ndo_fdb_dump)(struct sk_buff *skb,
1116 struct netlink_callback *cb,
1117 struct net_device *dev,
5d5eacb3 1118 struct net_device *filter_dev,
77162022 1119 int idx);
e5a55a89
JF
1120
1121 int (*ndo_bridge_setlink)(struct net_device *dev,
1122 struct nlmsghdr *nlh);
1123 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1124 u32 pid, u32 seq,
6cbdceeb
VY
1125 struct net_device *dev,
1126 u32 filter_mask);
407af329
VY
1127 int (*ndo_bridge_dellink)(struct net_device *dev,
1128 struct nlmsghdr *nlh);
4bf84c35
JP
1129 int (*ndo_change_carrier)(struct net_device *dev,
1130 bool new_carrier);
66b52b0d
JP
1131 int (*ndo_get_phys_port_id)(struct net_device *dev,
1132 struct netdev_phys_port_id *ppid);
53cf5275
JG
1133 void (*ndo_add_vxlan_port)(struct net_device *dev,
1134 sa_family_t sa_family,
35e42379 1135 __be16 port);
53cf5275
JG
1136 void (*ndo_del_vxlan_port)(struct net_device *dev,
1137 sa_family_t sa_family,
35e42379 1138 __be16 port);
a6cc0cfa
JF
1139
1140 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1141 struct net_device *dev);
1142 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1143 void *priv);
1144
1145 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1146 struct net_device *dev,
1147 void *priv);
25175ba5 1148 int (*ndo_get_lock_subclass)(struct net_device *dev);
d314774c
SH
1149};
1150
7aa98047
LR
1151/**
1152 * enum net_device_priv_flags - &struct net_device priv_flags
1153 *
1154 * These are the &struct net_device, they are only set internally
1155 * by drivers and used in the kernel. These flags are invisible to
1156 * userspace, this means that the order of these flags can change
1157 * during any kernel release.
1158 *
1159 * You should have a pretty good reason to be extending these flags.
1160 *
1161 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1162 * @IFF_EBRIDGE: Ethernet bridging device
1163 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1164 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1165 * @IFF_MASTER_ALB: bonding master, balance-alb
1166 * @IFF_BONDING: bonding master or slave
1167 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1168 * @IFF_ISATAP: ISATAP interface (RFC4214)
1169 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1170 * @IFF_WAN_HDLC: WAN HDLC device
1171 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1172 * release skb->dst
1173 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1174 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1175 * @IFF_MACVLAN_PORT: device used as macvlan port
1176 * @IFF_BRIDGE_PORT: device used as bridge port
1177 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1178 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1179 * @IFF_UNICAST_FLT: Supports unicast filtering
1180 * @IFF_TEAM_PORT: device used as team port
1181 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1182 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1183 * change when it's running
1184 * @IFF_MACVLAN: Macvlan device
1185 */
1186enum netdev_priv_flags {
1187 IFF_802_1Q_VLAN = 1<<0,
1188 IFF_EBRIDGE = 1<<1,
1189 IFF_SLAVE_INACTIVE = 1<<2,
1190 IFF_MASTER_8023AD = 1<<3,
1191 IFF_MASTER_ALB = 1<<4,
1192 IFF_BONDING = 1<<5,
1193 IFF_SLAVE_NEEDARP = 1<<6,
1194 IFF_ISATAP = 1<<7,
1195 IFF_MASTER_ARPMON = 1<<8,
1196 IFF_WAN_HDLC = 1<<9,
1197 IFF_XMIT_DST_RELEASE = 1<<10,
1198 IFF_DONT_BRIDGE = 1<<11,
1199 IFF_DISABLE_NETPOLL = 1<<12,
1200 IFF_MACVLAN_PORT = 1<<13,
1201 IFF_BRIDGE_PORT = 1<<14,
1202 IFF_OVS_DATAPATH = 1<<15,
1203 IFF_TX_SKB_SHARING = 1<<16,
1204 IFF_UNICAST_FLT = 1<<17,
1205 IFF_TEAM_PORT = 1<<18,
1206 IFF_SUPP_NOFCS = 1<<19,
1207 IFF_LIVE_ADDR_CHANGE = 1<<20,
1208 IFF_MACVLAN = 1<<21,
1209};
1210
1211#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1212#define IFF_EBRIDGE IFF_EBRIDGE
1213#define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1214#define IFF_MASTER_8023AD IFF_MASTER_8023AD
1215#define IFF_MASTER_ALB IFF_MASTER_ALB
1216#define IFF_BONDING IFF_BONDING
1217#define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1218#define IFF_ISATAP IFF_ISATAP
1219#define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1220#define IFF_WAN_HDLC IFF_WAN_HDLC
1221#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1222#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1223#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1224#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1225#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1226#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1227#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1228#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1229#define IFF_TEAM_PORT IFF_TEAM_PORT
1230#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1231#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1232#define IFF_MACVLAN IFF_MACVLAN
1233
536721b1
KK
1234/**
1235 * struct net_device - The DEVICE structure.
1236 * Actually, this whole structure is a big mistake. It mixes I/O
1237 * data with strictly "high-level" data, and it has to know about
1238 * almost every data structure used in the INET module.
1239 *
1240 * @name: This is the first field of the "visible" part of this structure
1241 * (i.e. as seen by users in the "Space.c" file). It is the name
1242 * of the interface.
1243 *
1244 * @name_hlist: Device name hash chain, please keep it close to name[]
1245 * @ifalias: SNMP alias
1246 * @mem_end: Shared memory end
1247 * @mem_start: Shared memory start
1248 * @base_addr: Device I/O address
1249 * @irq: Device IRQ number
1250 *
1251 * @state: Generic network queuing layer state, see netdev_state_t
1252 * @dev_list: The global list of network devices
1253 * @napi_list: List entry, that is used for polling napi devices
1254 * @unreg_list: List entry, that is used, when we are unregistering the
1255 * device, see the function unregister_netdev
1256 * @close_list: List entry, that is used, when we are closing the device
1257 *
1258 * @adj_list: Directly linked devices, like slaves for bonding
1259 * @all_adj_list: All linked devices, *including* neighbours
1260 * @features: Currently active device features
1261 * @hw_features: User-changeable features
1262 *
1263 * @wanted_features: User-requested features
1264 * @vlan_features: Mask of features inheritable by VLAN devices
1265 *
1266 * @hw_enc_features: Mask of features inherited by encapsulating devices
1267 * This field indicates what encapsulation
1268 * offloads the hardware is capable of doing,
1269 * and drivers will need to set them appropriately.
1270 *
1271 * @mpls_features: Mask of features inheritable by MPLS
1272 *
1273 * @ifindex: interface index
1274 * @iflink: unique device identifier
1275 *
1276 * @stats: Statistics struct, which was left as a legacy, use
1277 * rtnl_link_stats64 instead
1278 *
1279 * @rx_dropped: Dropped packets by core network,
1280 * do not use this in drivers
1281 * @tx_dropped: Dropped packets by core network,
1282 * do not use this in drivers
1283 *
1284 * @carrier_changes: Stats to monitor carrier on<->off transitions
1285 *
1286 * @wireless_handlers: List of functions to handle Wireless Extensions,
1287 * instead of ioctl,
1288 * see <net/iw_handler.h> for details.
1289 * @wireless_data: Instance data managed by the core of wireless extensions
1290 *
1291 * @netdev_ops: Includes several pointers to callbacks,
1292 * if one wants to override the ndo_*() functions
1293 * @ethtool_ops: Management operations
1294 * @fwd_ops: Management operations
1295 * @header_ops: Includes callbacks for creating,parsing,rebuilding,etc
1296 * of Layer 2 headers.
1297 *
1298 * @flags: Interface flags (a la BSD)
1299 * @priv_flags: Like 'flags' but invisible to userspace,
1300 * see if.h for the definitions
1301 * @gflags: Global flags ( kept as legacy )
1302 * @padded: How much padding added by alloc_netdev()
1303 * @operstate: RFC2863 operstate
1304 * @link_mode: Mapping policy to operstate
1305 * @if_port: Selectable AUI, TP, ...
1306 * @dma: DMA channel
1307 * @mtu: Interface MTU value
1308 * @type: Interface hardware type
1309 * @hard_header_len: Hardware header length
1310 *
1311 * @needed_headroom: Extra headroom the hardware may need, but not in all
1312 * cases can this be guaranteed
1313 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1314 * cases can this be guaranteed. Some cases also use
1315 * LL_MAX_HEADER instead to allocate the skb
1316 *
1317 * interface address info:
1318 *
1319 * @perm_addr: Permanent hw address
1320 * @addr_assign_type: Hw address assignment type
1321 * @addr_len: Hardware address length
1322 * @neigh_priv_len; Used in neigh_alloc(),
1323 * initialized only in atm/clip.c
1324 * @dev_id: Used to differentiate devices that share
1325 * the same link layer address
1326 * @dev_port: Used to differentiate devices that share
1327 * the same function
1328 * @addr_list_lock: XXX: need comments on this one
1329 * @uc: unicast mac addresses
1330 * @mc: multicast mac addresses
1331 * @dev_addrs: list of device hw addresses
1332 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1333 * @uc_promisc: Counter, that indicates, that promiscuous mode
1334 * has been enabled due to the need to listen to
1335 * additional unicast addresses in a device that
1336 * does not implement ndo_set_rx_mode()
1337 * @promiscuity: Number of times, the NIC is told to work in
1338 * Promiscuous mode, if it becomes 0 the NIC will
1339 * exit from working in Promiscuous mode
1340 * @allmulti: Counter, enables or disables allmulticast mode
1341 *
1342 * @vlan_info: VLAN info
1343 * @dsa_ptr: dsa specific data
1344 * @tipc_ptr: TIPC specific data
1345 * @atalk_ptr: AppleTalk link
1346 * @ip_ptr: IPv4 specific data
1347 * @dn_ptr: DECnet specific data
1348 * @ip6_ptr: IPv6 specific data
1349 * @ax25_ptr: AX.25 specific data
1350 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1351 *
1352 * @last_rx: Time of last Rx
1353 * @dev_addr: Hw address (before bcast,
1354 * because most packets are unicast)
1355 *
1356 * @_rx: Array of RX queues
1357 * @num_rx_queues: Number of RX queues
1358 * allocated at register_netdev() time
1359 * @real_num_rx_queues: Number of RX queues currently active in device
1360 *
1361 * @rx_handler: handler for received packets
1362 * @rx_handler_data: XXX: need comments on this one
1363 * @ingress_queue: XXX: need comments on this one
1364 * @broadcast: hw bcast address
1365 *
1366 * @_tx: Array of TX queues
1367 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1368 * @real_num_tx_queues: Number of TX queues currently active in device
1369 * @qdisc: Root qdisc from userspace point of view
1370 * @tx_queue_len: Max frames per queue allowed
1371 * @tx_global_lock: XXX: need comments on this one
1372 *
1373 * @xps_maps: XXX: need comments on this one
1374 *
1375 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1376 * indexed by RX queue number. Assigned by driver.
1377 * This must only be set if the ndo_rx_flow_steer
1378 * operation is defined
1379 *
1380 * @trans_start: Time (in jiffies) of last Tx
1381 * @watchdog_timeo: Represents the timeout that is used by
1382 * the watchdog ( see dev_watchdog() )
1383 * @watchdog_timer: List of timers
1384 *
1385 * @pcpu_refcnt: Number of references to this device
1386 * @todo_list: Delayed register/unregister
1387 * @index_hlist: Device index hash chain
1388 * @link_watch_list: XXX: need comments on this one
1389 *
1390 * @reg_state: Register/unregister state machine
1391 * @dismantle: Device is going to be freed
1392 * @rtnl_link_state: This enum represents the phases of creating
1393 * a new link
1394 *
1395 * @destructor: Called from unregister,
1396 * can be used to call free_netdev
1397 * @npinfo: XXX: need comments on this one
1398 * @nd_net: Network namespace this network device is inside
1399 *
1400 * @ml_priv: Mid-layer private
1401 * @lstats: Loopback statistics
1402 * @tstats: Tunnel statistics
1403 * @dstats: Dummy statistics
1404 * @vstats: Virtual ethernet statistics
1405 *
1406 * @garp_port: GARP
1407 * @mrp_port: MRP
1408 *
1409 * @dev: Class/net/name entry
1410 * @sysfs_groups: Space for optional device, statistics and wireless
1411 * sysfs groups
1412 *
1413 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1414 * @rtnl_link_ops: Rtnl_link_ops
1415 *
1416 * @gso_max_size: Maximum size of generic segmentation offload
1417 * @gso_max_segs: Maximum number of segments that can be passed to the
1418 * NIC for GSO
1419 *
1420 * @dcbnl_ops: Data Center Bridging netlink ops
1421 * @num_tc: Number of traffic classes in the net device
1422 * @tc_to_txq: XXX: need comments on this one
1423 * @prio_tc_map XXX: need comments on this one
1424 *
1425 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1426 *
1427 * @priomap: XXX: need comments on this one
1428 * @phydev: Physical device may attach itself
1429 * for hardware timestamping
1430 *
1431 * @qdisc_tx_busylock: XXX: need comments on this one
1432 *
1433 * @group: The group, that the device belongs to
1434 * @pm_qos_req: Power Management QoS object
1da177e4
LT
1435 *
1436 * FIXME: cleanup struct net_device such that network protocol info
1437 * moves out.
1438 */
1439
d94d9fee 1440struct net_device {
1da177e4 1441 char name[IFNAMSIZ];
9356b8fc 1442 struct hlist_node name_hlist;
0b815a1a 1443 char *ifalias;
1da177e4
LT
1444 /*
1445 * I/O specific fields
1446 * FIXME: Merge these and struct ifmap into one
1447 */
536721b1
KK
1448 unsigned long mem_end;
1449 unsigned long mem_start;
1450 unsigned long base_addr;
1451 int irq;
1da177e4
LT
1452
1453 /*
536721b1
KK
1454 * Some hardware also needs these fields (state,dev_list,
1455 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1456 * part of the usual set specified in Space.c.
1457 */
1458
1da177e4
LT
1459 unsigned long state;
1460
7562f876 1461 struct list_head dev_list;
bea3348e 1462 struct list_head napi_list;
44a0873d 1463 struct list_head unreg_list;
5cde2829 1464 struct list_head close_list;
2f268f12 1465
2f268f12
VF
1466 struct {
1467 struct list_head upper;
1468 struct list_head lower;
1469 } adj_list;
1470
2f268f12
VF
1471 struct {
1472 struct list_head upper;
1473 struct list_head lower;
1474 } all_adj_list;
4c3d5e7b 1475
c8f44aff 1476 netdev_features_t features;
c8f44aff 1477 netdev_features_t hw_features;
c8f44aff 1478 netdev_features_t wanted_features;
c8f44aff 1479 netdev_features_t vlan_features;
6a674e9c 1480 netdev_features_t hw_enc_features;
0d89d203 1481 netdev_features_t mpls_features;
04ed3e74 1482
1da177e4
LT
1483 int ifindex;
1484 int iflink;
1485
c45d286e 1486 struct net_device_stats stats;
015f0688 1487
015f0688
ED
1488 atomic_long_t rx_dropped;
1489 atomic_long_t tx_dropped;
1da177e4 1490
2d3b479d 1491 atomic_t carrier_changes;
1492
b86e0280 1493#ifdef CONFIG_WIRELESS_EXT
1da177e4 1494 const struct iw_handler_def * wireless_handlers;
1da177e4 1495 struct iw_public_data * wireless_data;
b86e0280 1496#endif
d314774c 1497 const struct net_device_ops *netdev_ops;
76fd8593 1498 const struct ethtool_ops *ethtool_ops;
a6cc0cfa 1499 const struct forwarding_accel_ops *fwd_ops;
1da177e4 1500
3b04ddde
SH
1501 const struct header_ops *header_ops;
1502
536721b1
KK
1503 unsigned int flags;
1504 unsigned int priv_flags;
1505
1da177e4 1506 unsigned short gflags;
536721b1 1507 unsigned short padded;
1da177e4 1508
536721b1
KK
1509 unsigned char operstate;
1510 unsigned char link_mode;
b00055aa 1511
536721b1
KK
1512 unsigned char if_port;
1513 unsigned char dma;
bdc220da 1514
536721b1
KK
1515 unsigned int mtu;
1516 unsigned short type;
1517 unsigned short hard_header_len;
1da177e4 1518
f5184d26
JB
1519 unsigned short needed_headroom;
1520 unsigned short needed_tailroom;
1521
1da177e4 1522 /* Interface address info. */
536721b1
KK
1523 unsigned char perm_addr[MAX_ADDR_LEN];
1524 unsigned char addr_assign_type;
1525 unsigned char addr_len;
a0a9663d 1526 unsigned short neigh_priv_len;
536721b1
KK
1527 unsigned short dev_id;
1528 unsigned short dev_port;
ccffad25 1529 spinlock_t addr_list_lock;
536721b1
KK
1530 struct netdev_hw_addr_list uc;
1531 struct netdev_hw_addr_list mc;
1532 struct netdev_hw_addr_list dev_addrs;
1533
4c3d5e7b
ED
1534#ifdef CONFIG_SYSFS
1535 struct kset *queues_kset;
1536#endif
1537
685343fc
TG
1538 unsigned char name_assign_type;
1539
2d348d1f 1540 bool uc_promisc;
9d45abe1
WC
1541 unsigned int promiscuity;
1542 unsigned int allmulti;
1da177e4 1543
1da177e4
LT
1544
1545 /* Protocol specific pointers */
65ac6a5f 1546
d11ead75 1547#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1548 struct vlan_info __rcu *vlan_info;
65ac6a5f 1549#endif
34a430d7 1550#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1551 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1552#endif
1553#if IS_ENABLED(CONFIG_TIPC)
536721b1 1554 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1555#endif
536721b1
KK
1556 void *atalk_ptr;
1557 struct in_device __rcu *ip_ptr;
1558 struct dn_dev __rcu *dn_ptr;
1559 struct inet6_dev __rcu *ip6_ptr;
1560 void *ax25_ptr;
1561 struct wireless_dev *ieee80211_ptr;
1da177e4 1562
9356b8fc 1563/*
cd13539b 1564 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1565 */
536721b1 1566 unsigned long last_rx;
4dc89133 1567
9356b8fc 1568 /* Interface address info used in eth_type_trans() */
536721b1 1569 unsigned char *dev_addr;
f001fde5 1570
0a9627f2 1571
a953be53 1572#ifdef CONFIG_SYSFS
0a9627f2
TH
1573 struct netdev_rx_queue *_rx;
1574
0a9627f2 1575 unsigned int num_rx_queues;
62fe0b40 1576 unsigned int real_num_rx_queues;
c445477d 1577
df334545 1578#endif
0a9627f2 1579
61391cde 1580 rx_handler_func_t __rcu *rx_handler;
1581 void __rcu *rx_handler_data;
e8a0464c 1582
24824a09 1583 struct netdev_queue __rcu *ingress_queue;
536721b1 1584 unsigned char broadcast[MAX_ADDR_LEN];
4c3d5e7b 1585
cd13539b
ED
1586
1587/*
1588 * Cache lines mostly used on transmit path
1589 */
e8a0464c
DM
1590 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1591 unsigned int num_tx_queues;
fd2ea0a7 1592 unsigned int real_num_tx_queues;
af356afa 1593 struct Qdisc *qdisc;
536721b1 1594 unsigned long tx_queue_len;
c3f26a26 1595 spinlock_t tx_global_lock;
cd13539b 1596
bf264145 1597#ifdef CONFIG_XPS
a4177869 1598 struct xps_dev_maps __rcu *xps_maps;
bf264145 1599#endif
4c3d5e7b 1600#ifdef CONFIG_RFS_ACCEL
4c3d5e7b
ED
1601 struct cpu_rmap *rx_cpu_rmap;
1602#endif
1d24eb48 1603
9356b8fc 1604 /* These may be needed for future network-power-down code. */
9d21493b
ED
1605
1606 /*
1607 * trans_start here is expensive for high speed devices on SMP,
1608 * please use netdev_queue->trans_start instead.
1609 */
536721b1 1610 unsigned long trans_start;
9356b8fc 1611
536721b1 1612 int watchdog_timeo;
9356b8fc
ED
1613 struct timer_list watchdog_timer;
1614
29b4433d 1615 int __percpu *pcpu_refcnt;
1da177e4 1616 struct list_head todo_list;
1da177e4 1617
536721b1 1618 struct hlist_node index_hlist;
e014debe 1619 struct list_head link_watch_list;
572a103d 1620
1da177e4 1621 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1622 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1623 NETREG_UNREGISTERING, /* called unregister_netdevice */
1624 NETREG_UNREGISTERED, /* completed unregister todo */
1625 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1626 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1627 } reg_state:8;
1628
536721b1 1629 bool dismantle;
a2835763
PM
1630
1631 enum {
1632 RTNL_LINK_INITIALIZED,
1633 RTNL_LINK_INITIALIZING,
1634 } rtnl_link_state:16;
1da177e4 1635
d314774c 1636 void (*destructor)(struct net_device *dev);
1da177e4 1637
1da177e4 1638#ifdef CONFIG_NETPOLL
5fbee843 1639 struct netpoll_info __rcu *npinfo;
1da177e4 1640#endif
eae792b7 1641
c346dca1 1642#ifdef CONFIG_NET_NS
4a1c5371 1643 struct net *nd_net;
c346dca1 1644#endif
4a1c5371 1645
4951704b 1646 /* mid-layer private */
a7855c78 1647 union {
536721b1
KK
1648 void *ml_priv;
1649 struct pcpu_lstats __percpu *lstats;
8f84985f 1650 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1651 struct pcpu_dstats __percpu *dstats;
1652 struct pcpu_vstats __percpu *vstats;
a7855c78 1653 };
536721b1 1654
3cc77ec7 1655 struct garp_port __rcu *garp_port;
febf018d 1656 struct mrp_port __rcu *mrp_port;
1da177e4 1657
536721b1 1658 struct device dev;
0c509a6c 1659 const struct attribute_group *sysfs_groups[4];
a953be53 1660 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1661
38f7b870 1662 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1663
82cc1a7a
PWJ
1664 /* for setting kernel sock attribute on TCP connection setup */
1665#define GSO_MAX_SIZE 65536
1666 unsigned int gso_max_size;
30b678d8
BH
1667#define GSO_MAX_SEGS 65535
1668 u16 gso_max_segs;
d314774c 1669
7a6b6f51 1670#ifdef CONFIG_DCB
32953543 1671 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1672#endif
4f57c087
JF
1673 u8 num_tc;
1674 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1675 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1676
d11ead75 1677#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1678 unsigned int fcoe_ddp_xid;
5bc1421e 1679#endif
86f8515f 1680#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1681 struct netprio_map __rcu *priomap;
4d288d57 1682#endif
c1f19b51 1683 struct phy_device *phydev;
23d3b8bf 1684 struct lock_class_key *qdisc_tx_busylock;
cbda10fa 1685 int group;
9136461a 1686 struct pm_qos_request pm_qos_req;
1da177e4 1687};
43cb76d9 1688#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1689
1690#define NETDEV_ALIGN 32
1da177e4 1691
4f57c087
JF
1692static inline
1693int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1694{
1695 return dev->prio_tc_map[prio & TC_BITMASK];
1696}
1697
1698static inline
1699int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1700{
1701 if (tc >= dev->num_tc)
1702 return -EINVAL;
1703
1704 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1705 return 0;
1706}
1707
1708static inline
1709void netdev_reset_tc(struct net_device *dev)
1710{
1711 dev->num_tc = 0;
1712 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1713 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1714}
1715
1716static inline
1717int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1718{
1719 if (tc >= dev->num_tc)
1720 return -EINVAL;
1721
1722 dev->tc_to_txq[tc].count = count;
1723 dev->tc_to_txq[tc].offset = offset;
1724 return 0;
1725}
1726
1727static inline
1728int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1729{
1730 if (num_tc > TC_MAX_QUEUE)
1731 return -EINVAL;
1732
1733 dev->num_tc = num_tc;
1734 return 0;
1735}
1736
1737static inline
1738int netdev_get_num_tc(struct net_device *dev)
1739{
1740 return dev->num_tc;
1741}
1742
e8a0464c
DM
1743static inline
1744struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1745 unsigned int index)
1746{
1747 return &dev->_tx[index];
1748}
1749
1750static inline void netdev_for_each_tx_queue(struct net_device *dev,
1751 void (*f)(struct net_device *,
1752 struct netdev_queue *,
1753 void *),
1754 void *arg)
1755{
1756 unsigned int i;
1757
1758 for (i = 0; i < dev->num_tx_queues; i++)
1759 f(dev, &dev->_tx[i], arg);
1760}
1761
f629d208 1762struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1763 struct sk_buff *skb,
1764 void *accel_priv);
8c4c49df 1765
c346dca1
YH
1766/*
1767 * Net namespace inlines
1768 */
1769static inline
1770struct net *dev_net(const struct net_device *dev)
1771{
c2d9ba9b 1772 return read_pnet(&dev->nd_net);
c346dca1
YH
1773}
1774
1775static inline
f5aa23fd 1776void dev_net_set(struct net_device *dev, struct net *net)
c346dca1
YH
1777{
1778#ifdef CONFIG_NET_NS
f3005d7f
DL
1779 release_net(dev->nd_net);
1780 dev->nd_net = hold_net(net);
c346dca1
YH
1781#endif
1782}
1783
cf85d08f
LB
1784static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1785{
1786#ifdef CONFIG_NET_DSA_TAG_DSA
1787 if (dev->dsa_ptr != NULL)
1788 return dsa_uses_dsa_tags(dev->dsa_ptr);
1789#endif
1790
1791 return 0;
1792}
1793
396138f0
LB
1794static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1795{
1796#ifdef CONFIG_NET_DSA_TAG_TRAILER
1797 if (dev->dsa_ptr != NULL)
1798 return dsa_uses_trailer_tags(dev->dsa_ptr);
1799#endif
1800
1801 return 0;
1802}
1803
bea3348e
SH
1804/**
1805 * netdev_priv - access network device private data
1806 * @dev: network device
1807 *
1808 * Get network device private data
1809 */
6472ce60 1810static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1811{
1ce8e7b5 1812 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1813}
1814
1da177e4
LT
1815/* Set the sysfs physical device reference for the network logical device
1816 * if set prior to registration will cause a symlink during initialization.
1817 */
43cb76d9 1818#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1819
384912ed 1820/* Set the sysfs device type for the network logical device to allow
3f79410c 1821 * fine-grained identification of different network device types. For
384912ed
MH
1822 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1823 */
1824#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1825
82dc3c63
ED
1826/* Default NAPI poll() weight
1827 * Device drivers are strongly advised to not use bigger value
1828 */
1829#define NAPI_POLL_WEIGHT 64
1830
3b582cc1
SH
1831/**
1832 * netif_napi_add - initialize a napi context
1833 * @dev: network device
1834 * @napi: napi context
1835 * @poll: polling function
1836 * @weight: default weight
1837 *
1838 * netif_napi_add() must be used to initialize a napi context prior to calling
1839 * *any* of the other napi related functions.
1840 */
d565b0a1
HX
1841void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1842 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1843
d8156534
AD
1844/**
1845 * netif_napi_del - remove a napi context
1846 * @napi: napi context
1847 *
1848 * netif_napi_del() removes a napi context from the network device napi list
1849 */
d565b0a1
HX
1850void netif_napi_del(struct napi_struct *napi);
1851
1852struct napi_gro_cb {
78a478d0
HX
1853 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1854 void *frag0;
1855
7489594c
HX
1856 /* Length of frag0. */
1857 unsigned int frag0_len;
1858
86911732
HX
1859 /* This indicates where we are processing relative to skb->data. */
1860 int data_offset;
1861
d565b0a1 1862 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
1863 u16 flush;
1864
1865 /* Save the IP ID here and check when we get to the transport layer */
1866 u16 flush_id;
d565b0a1
HX
1867
1868 /* Number of segments aggregated. */
2e71a6f8
ED
1869 u16 count;
1870
1871 /* This is non-zero if the packet may be of the same flow. */
1872 u8 same_flow;
5d38a079
HX
1873
1874 /* Free the skb? */
2e71a6f8 1875 u8 free;
d7e8883c
ED
1876#define NAPI_GRO_FREE 1
1877#define NAPI_GRO_FREE_STOLEN_HEAD 2
2e71a6f8
ED
1878
1879 /* jiffies when first packet was created/queued */
1880 unsigned long age;
86347245
ED
1881
1882 /* Used in ipv6_gro_receive() */
b582ef09
OG
1883 u16 proto;
1884
1885 /* Used in udp_gro_receive */
1886 u16 udp_mark;
c3c7c254 1887
bf5a755f
JC
1888 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1889 __wsum csum;
1890
c3c7c254
ED
1891 /* used in skb_gro_receive() slow path */
1892 struct sk_buff *last;
d565b0a1
HX
1893};
1894
1895#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 1896
1da177e4 1897struct packet_type {
f2ccd8fa
DM
1898 __be16 type; /* This is really htons(ether_type). */
1899 struct net_device *dev; /* NULL is wildcarded here */
1900 int (*func) (struct sk_buff *,
1901 struct net_device *,
1902 struct packet_type *,
1903 struct net_device *);
c0de08d0
EL
1904 bool (*id_match)(struct packet_type *ptype,
1905 struct sock *sk);
1da177e4
LT
1906 void *af_packet_priv;
1907 struct list_head list;
1908};
1909
f191a1d1 1910struct offload_callbacks {
576a30eb 1911 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 1912 netdev_features_t features);
a430a43d 1913 int (*gso_send_check)(struct sk_buff *skb);
d565b0a1
HX
1914 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1915 struct sk_buff *skb);
299603e8 1916 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
1917};
1918
1919struct packet_offload {
1920 __be16 type; /* This is really htons(ether_type). */
1921 struct offload_callbacks callbacks;
1922 struct list_head list;
1da177e4
LT
1923};
1924
b582ef09
OG
1925struct udp_offload {
1926 __be16 port;
1927 struct offload_callbacks callbacks;
1928};
1929
8f84985f
LR
1930/* often modified stats are per cpu, other are shared (netdev->stats) */
1931struct pcpu_sw_netstats {
1932 u64 rx_packets;
1933 u64 rx_bytes;
1934 u64 tx_packets;
1935 u64 tx_bytes;
1936 struct u64_stats_sync syncp;
1937};
1938
1c213bd2
WC
1939#define netdev_alloc_pcpu_stats(type) \
1940({ \
693350c2 1941 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1c213bd2
WC
1942 if (pcpu_stats) { \
1943 int i; \
1944 for_each_possible_cpu(i) { \
1945 typeof(type) *stat; \
1946 stat = per_cpu_ptr(pcpu_stats, i); \
1947 u64_stats_init(&stat->syncp); \
1948 } \
1949 } \
1950 pcpu_stats; \
1951})
1952
1da177e4
LT
1953#include <linux/notifier.h>
1954
dcfe1421
AW
1955/* netdevice notifier chain. Please remember to update the rtnetlink
1956 * notification exclusion list in rtnetlink_event() when adding new
1957 * types.
1958 */
1959#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1960#define NETDEV_DOWN 0x0002
1961#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1962 detected a hardware crash and restarted
1963 - we can use this eg to kick tcp sessions
1964 once done */
1965#define NETDEV_CHANGE 0x0004 /* Notify device state change */
1966#define NETDEV_REGISTER 0x0005
1967#define NETDEV_UNREGISTER 0x0006
1d486bfb 1968#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
1969#define NETDEV_CHANGEADDR 0x0008
1970#define NETDEV_GOING_DOWN 0x0009
1971#define NETDEV_CHANGENAME 0x000A
1972#define NETDEV_FEAT_CHANGE 0x000B
1973#define NETDEV_BONDING_FAILOVER 0x000C
1974#define NETDEV_PRE_UP 0x000D
1975#define NETDEV_PRE_TYPE_CHANGE 0x000E
1976#define NETDEV_POST_TYPE_CHANGE 0x000F
1977#define NETDEV_POST_INIT 0x0010
0115e8e3 1978#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
1979#define NETDEV_RELEASE 0x0012
1980#define NETDEV_NOTIFY_PEERS 0x0013
1981#define NETDEV_JOIN 0x0014
42e52bf9 1982#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 1983#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 1984#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 1985#define NETDEV_CHANGEINFODATA 0x0018
dcfe1421 1986
f629d208
JP
1987int register_netdevice_notifier(struct notifier_block *nb);
1988int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
1989
1990struct netdev_notifier_info {
1991 struct net_device *dev;
1992};
1993
be9efd36
JP
1994struct netdev_notifier_change_info {
1995 struct netdev_notifier_info info; /* must be first */
1996 unsigned int flags_changed;
1997};
1998
75538c2b
CW
1999static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2000 struct net_device *dev)
2001{
2002 info->dev = dev;
2003}
2004
351638e7
JP
2005static inline struct net_device *
2006netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2007{
2008 return info->dev;
2009}
2010
f629d208 2011int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2012
2013
1da177e4
LT
2014extern rwlock_t dev_base_lock; /* Device list lock */
2015
881d966b
EB
2016#define for_each_netdev(net, d) \
2017 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2018#define for_each_netdev_reverse(net, d) \
2019 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2020#define for_each_netdev_rcu(net, d) \
2021 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2022#define for_each_netdev_safe(net, d, n) \
2023 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2024#define for_each_netdev_continue(net, d) \
2025 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2026#define for_each_netdev_continue_rcu(net, d) \
2027 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2028#define for_each_netdev_in_bond_rcu(bond, slave) \
2029 for_each_netdev_rcu(&init_net, slave) \
2030 if (netdev_master_upper_dev_get_rcu(slave) == bond)
881d966b 2031#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2032
a050c33f
DL
2033static inline struct net_device *next_net_device(struct net_device *dev)
2034{
2035 struct list_head *lh;
2036 struct net *net;
2037
c346dca1 2038 net = dev_net(dev);
a050c33f
DL
2039 lh = dev->dev_list.next;
2040 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2041}
2042
ce81b76a
ED
2043static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2044{
2045 struct list_head *lh;
2046 struct net *net;
2047
2048 net = dev_net(dev);
ccf43438 2049 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2050 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2051}
2052
a050c33f
DL
2053static inline struct net_device *first_net_device(struct net *net)
2054{
2055 return list_empty(&net->dev_base_head) ? NULL :
2056 net_device_entry(net->dev_base_head.next);
2057}
7562f876 2058
ccf43438
ED
2059static inline struct net_device *first_net_device_rcu(struct net *net)
2060{
2061 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2062
2063 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2064}
2065
f629d208
JP
2066int netdev_boot_setup_check(struct net_device *dev);
2067unsigned long netdev_boot_base(const char *prefix, int unit);
2068struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2069 const char *hwaddr);
2070struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2071struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2072void dev_add_pack(struct packet_type *pt);
2073void dev_remove_pack(struct packet_type *pt);
2074void __dev_remove_pack(struct packet_type *pt);
2075void dev_add_offload(struct packet_offload *po);
2076void dev_remove_offload(struct packet_offload *po);
f629d208
JP
2077
2078struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
2079 unsigned short mask);
2080struct net_device *dev_get_by_name(struct net *net, const char *name);
2081struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2082struct net_device *__dev_get_by_name(struct net *net, const char *name);
2083int dev_alloc_name(struct net_device *dev, const char *name);
2084int dev_open(struct net_device *dev);
2085int dev_close(struct net_device *dev);
2086void dev_disable_lro(struct net_device *dev);
2087int dev_loopback_xmit(struct sk_buff *newskb);
2088int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2089int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2090int register_netdevice(struct net_device *dev);
2091void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2092void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2093static inline void unregister_netdevice(struct net_device *dev)
2094{
2095 unregister_netdevice_queue(dev, NULL);
2096}
2097
f629d208
JP
2098int netdev_refcnt_read(const struct net_device *dev);
2099void free_netdev(struct net_device *dev);
74d332c1 2100void netdev_freemem(struct net_device *dev);
f629d208
JP
2101void synchronize_net(void);
2102int init_dummy_netdev(struct net_device *dev);
937f1ba5 2103
f629d208
JP
2104struct net_device *dev_get_by_index(struct net *net, int ifindex);
2105struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2106struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2107int netdev_get_name(struct net *net, char *name, int ifindex);
2108int dev_restart(struct net_device *dev);
f629d208 2109int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2110
2111static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2112{
2113 return NAPI_GRO_CB(skb)->data_offset;
2114}
2115
2116static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2117{
2118 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2119}
2120
2121static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2122{
2123 NAPI_GRO_CB(skb)->data_offset += len;
2124}
2125
a5b1cf28
HX
2126static inline void *skb_gro_header_fast(struct sk_buff *skb,
2127 unsigned int offset)
86911732 2128{
a5b1cf28
HX
2129 return NAPI_GRO_CB(skb)->frag0 + offset;
2130}
78a478d0 2131
a5b1cf28
HX
2132static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2133{
2134 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2135}
78a478d0 2136
a5b1cf28
HX
2137static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2138 unsigned int offset)
2139{
17dd759c
HX
2140 if (!pskb_may_pull(skb, hlen))
2141 return NULL;
2142
a5b1cf28
HX
2143 NAPI_GRO_CB(skb)->frag0 = NULL;
2144 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2145 return skb->data + offset;
86911732 2146}
1da177e4 2147
36e7b1b8
HX
2148static inline void *skb_gro_network_header(struct sk_buff *skb)
2149{
78d3fd0b
HX
2150 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2151 skb_network_offset(skb);
36e7b1b8
HX
2152}
2153
bf5a755f
JC
2154static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2155 const void *start, unsigned int len)
2156{
2157 if (skb->ip_summed == CHECKSUM_COMPLETE)
2158 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2159 csum_partial(start, len, 0));
2160}
2161
0c4e8581
SH
2162static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2163 unsigned short type,
3b04ddde 2164 const void *daddr, const void *saddr,
95c96174 2165 unsigned int len)
0c4e8581 2166{
f1ecfd5d 2167 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2168 return 0;
3b04ddde
SH
2169
2170 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2171}
2172
b95cce35
SH
2173static inline int dev_parse_header(const struct sk_buff *skb,
2174 unsigned char *haddr)
2175{
2176 const struct net_device *dev = skb->dev;
2177
1b83336b 2178 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2179 return 0;
3b04ddde 2180 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2181}
2182
2205369a
DM
2183static inline int dev_rebuild_header(struct sk_buff *skb)
2184{
2185 const struct net_device *dev = skb->dev;
2186
2187 if (!dev->header_ops || !dev->header_ops->rebuild)
2188 return 0;
2189 return dev->header_ops->rebuild(skb);
2190}
2191
1da177e4 2192typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2193int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2194static inline int unregister_gifconf(unsigned int family)
2195{
2196 return register_gifconf(family, NULL);
2197}
2198
99bbc707 2199#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2200#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2201struct sd_flow_limit {
2202 u64 count;
2203 unsigned int num_buckets;
2204 unsigned int history_head;
2205 u16 history[FLOW_LIMIT_HISTORY];
2206 u8 buckets[];
2207};
2208
2209extern int netdev_flow_limit_table_len;
2210#endif /* CONFIG_NET_FLOW_LIMIT */
2211
1da177e4 2212/*
88751275 2213 * Incoming packets are placed on per-cpu queues
1da177e4 2214 */
d94d9fee 2215struct softnet_data {
37437bb2 2216 struct Qdisc *output_queue;
a9cbd588 2217 struct Qdisc **output_queue_tailp;
1da177e4 2218 struct list_head poll_list;
1da177e4 2219 struct sk_buff *completion_queue;
6e7676c1 2220 struct sk_buff_head process_queue;
1da177e4 2221
dee42870 2222 /* stats */
cd7b5396
DM
2223 unsigned int processed;
2224 unsigned int time_squeeze;
2225 unsigned int cpu_collision;
2226 unsigned int received_rps;
dee42870 2227
fd793d89 2228#ifdef CONFIG_RPS
88751275
ED
2229 struct softnet_data *rps_ipi_list;
2230
2231 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 2232 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2233 struct softnet_data *rps_ipi_next;
2234 unsigned int cpu;
fec5e652 2235 unsigned int input_queue_head;
76cc8b13 2236 unsigned int input_queue_tail;
1e94d72f 2237#endif
95c96174 2238 unsigned int dropped;
0a9627f2 2239 struct sk_buff_head input_pkt_queue;
bea3348e 2240 struct napi_struct backlog;
99bbc707
WB
2241
2242#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2243 struct sd_flow_limit __rcu *flow_limit;
99bbc707 2244#endif
1da177e4
LT
2245};
2246
76cc8b13 2247static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2248{
2249#ifdef CONFIG_RPS
76cc8b13
TH
2250 sd->input_queue_head++;
2251#endif
2252}
2253
2254static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2255 unsigned int *qtail)
2256{
2257#ifdef CONFIG_RPS
2258 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2259#endif
2260}
2261
0a9627f2 2262DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2263
f629d208 2264void __netif_schedule(struct Qdisc *q);
1da177e4 2265
86d804e1 2266static inline void netif_schedule_queue(struct netdev_queue *txq)
1da177e4 2267{
73466498 2268 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
37437bb2 2269 __netif_schedule(txq->qdisc);
86d804e1
DM
2270}
2271
fd2ea0a7
DM
2272static inline void netif_tx_schedule_all(struct net_device *dev)
2273{
2274 unsigned int i;
2275
2276 for (i = 0; i < dev->num_tx_queues; i++)
2277 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2278}
2279
d29f749e
DJ
2280static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2281{
73466498 2282 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2283}
2284
bea3348e
SH
2285/**
2286 * netif_start_queue - allow transmit
2287 * @dev: network device
2288 *
2289 * Allow upper layers to call the device hard_start_xmit routine.
2290 */
1da177e4
LT
2291static inline void netif_start_queue(struct net_device *dev)
2292{
e8a0464c 2293 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2294}
2295
fd2ea0a7
DM
2296static inline void netif_tx_start_all_queues(struct net_device *dev)
2297{
2298 unsigned int i;
2299
2300 for (i = 0; i < dev->num_tx_queues; i++) {
2301 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2302 netif_tx_start_queue(txq);
2303 }
2304}
2305
79d16385 2306static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1da177e4 2307{
73466498 2308 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
37437bb2 2309 __netif_schedule(dev_queue->qdisc);
79d16385
DM
2310}
2311
d29f749e
DJ
2312/**
2313 * netif_wake_queue - restart transmit
2314 * @dev: network device
2315 *
2316 * Allow upper layers to call the device hard_start_xmit routine.
2317 * Used for flow control when transmit resources are available.
2318 */
79d16385
DM
2319static inline void netif_wake_queue(struct net_device *dev)
2320{
e8a0464c 2321 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2322}
2323
fd2ea0a7
DM
2324static inline void netif_tx_wake_all_queues(struct net_device *dev)
2325{
2326 unsigned int i;
2327
2328 for (i = 0; i < dev->num_tx_queues; i++) {
2329 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2330 netif_tx_wake_queue(txq);
2331 }
2332}
2333
d29f749e
DJ
2334static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2335{
18543a64 2336 if (WARN_ON(!dev_queue)) {
256ee435 2337 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
18543a64
GC
2338 return;
2339 }
73466498 2340 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2341}
2342
bea3348e
SH
2343/**
2344 * netif_stop_queue - stop transmitted packets
2345 * @dev: network device
2346 *
2347 * Stop upper layers calling the device hard_start_xmit routine.
2348 * Used for flow control when transmit resources are unavailable.
2349 */
1da177e4
LT
2350static inline void netif_stop_queue(struct net_device *dev)
2351{
e8a0464c 2352 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2353}
2354
fd2ea0a7
DM
2355static inline void netif_tx_stop_all_queues(struct net_device *dev)
2356{
2357 unsigned int i;
2358
2359 for (i = 0; i < dev->num_tx_queues; i++) {
2360 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2361 netif_tx_stop_queue(txq);
2362 }
2363}
2364
4d29515f 2365static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2366{
73466498 2367 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2368}
2369
bea3348e
SH
2370/**
2371 * netif_queue_stopped - test if transmit queue is flowblocked
2372 * @dev: network device
2373 *
2374 * Test if transmit queue on device is currently unable to send.
2375 */
4d29515f 2376static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2377{
e8a0464c 2378 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2379}
2380
4d29515f 2381static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2382{
73466498
TH
2383 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2384}
2385
8e2f1a63
DB
2386static inline bool
2387netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2388{
2389 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2390}
2391
8e2f1a63
DB
2392static inline bool
2393netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2394{
2395 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2396}
2397
c5d67bd7
TH
2398static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2399 unsigned int bytes)
2400{
114cf580
TH
2401#ifdef CONFIG_BQL
2402 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2403
2404 if (likely(dql_avail(&dev_queue->dql) >= 0))
2405 return;
2406
2407 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2408
2409 /*
2410 * The XOFF flag must be set before checking the dql_avail below,
2411 * because in netdev_tx_completed_queue we update the dql_completed
2412 * before checking the XOFF flag.
2413 */
2414 smp_mb();
2415
2416 /* check again in case another CPU has just made room avail */
2417 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2418 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2419#endif
c5d67bd7
TH
2420}
2421
0042d0c8
FF
2422/**
2423 * netdev_sent_queue - report the number of bytes queued to hardware
2424 * @dev: network device
2425 * @bytes: number of bytes queued to the hardware device queue
2426 *
2427 * Report the number of bytes queued for sending/completion to the network
2428 * device hardware queue. @bytes should be a good approximation and should
2429 * exactly match netdev_completed_queue() @bytes
2430 */
c5d67bd7
TH
2431static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2432{
2433 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2434}
2435
2436static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2437 unsigned int pkts, unsigned int bytes)
c5d67bd7 2438{
114cf580 2439#ifdef CONFIG_BQL
b37c0fbe
AD
2440 if (unlikely(!bytes))
2441 return;
2442
2443 dql_completed(&dev_queue->dql, bytes);
2444
2445 /*
2446 * Without the memory barrier there is a small possiblity that
2447 * netdev_tx_sent_queue will miss the update and cause the queue to
2448 * be stopped forever
2449 */
2450 smp_mb();
2451
2452 if (dql_avail(&dev_queue->dql) < 0)
2453 return;
2454
2455 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2456 netif_schedule_queue(dev_queue);
114cf580 2457#endif
c5d67bd7
TH
2458}
2459
0042d0c8
FF
2460/**
2461 * netdev_completed_queue - report bytes and packets completed by device
2462 * @dev: network device
2463 * @pkts: actual number of packets sent over the medium
2464 * @bytes: actual number of bytes sent over the medium
2465 *
2466 * Report the number of bytes and packets transmitted by the network device
2467 * hardware queue over the physical medium, @bytes must exactly match the
2468 * @bytes amount passed to netdev_sent_queue()
2469 */
c5d67bd7 2470static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2471 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2472{
2473 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2474}
2475
2476static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2477{
114cf580 2478#ifdef CONFIG_BQL
5c490354 2479 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2480 dql_reset(&q->dql);
2481#endif
c5d67bd7
TH
2482}
2483
0042d0c8
FF
2484/**
2485 * netdev_reset_queue - reset the packets and bytes count of a network device
2486 * @dev_queue: network device
2487 *
2488 * Reset the bytes and packet count of a network device and clear the
2489 * software flow control OFF bit for this network device
2490 */
c5d67bd7
TH
2491static inline void netdev_reset_queue(struct net_device *dev_queue)
2492{
2493 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2494}
2495
b9507bda
DB
2496/**
2497 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2498 * @dev: network device
2499 * @queue_index: given tx queue index
2500 *
2501 * Returns 0 if given tx queue index >= number of device tx queues,
2502 * otherwise returns the originally passed tx queue index.
2503 */
2504static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2505{
2506 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2507 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2508 dev->name, queue_index,
2509 dev->real_num_tx_queues);
2510 return 0;
2511 }
2512
2513 return queue_index;
2514}
2515
bea3348e
SH
2516/**
2517 * netif_running - test if up
2518 * @dev: network device
2519 *
2520 * Test if the device has been brought up.
2521 */
4d29515f 2522static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2523{
2524 return test_bit(__LINK_STATE_START, &dev->state);
2525}
2526
f25f4e44
PWJ
2527/*
2528 * Routines to manage the subqueues on a device. We only need start
2529 * stop, and a check if it's stopped. All other device management is
2530 * done at the overall netdevice level.
2531 * Also test the device if we're multiqueue.
2532 */
bea3348e
SH
2533
2534/**
2535 * netif_start_subqueue - allow sending packets on subqueue
2536 * @dev: network device
2537 * @queue_index: sub queue index
2538 *
2539 * Start individual transmit queue of a device with multiple transmit queues.
2540 */
f25f4e44
PWJ
2541static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2542{
fd2ea0a7 2543 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2544
2545 netif_tx_start_queue(txq);
f25f4e44
PWJ
2546}
2547
bea3348e
SH
2548/**
2549 * netif_stop_subqueue - stop sending packets on subqueue
2550 * @dev: network device
2551 * @queue_index: sub queue index
2552 *
2553 * Stop individual transmit queue of a device with multiple transmit queues.
2554 */
f25f4e44
PWJ
2555static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2556{
fd2ea0a7 2557 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 2558 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2559}
2560
bea3348e
SH
2561/**
2562 * netif_subqueue_stopped - test status of subqueue
2563 * @dev: network device
2564 * @queue_index: sub queue index
2565 *
2566 * Check individual transmit queue of a device with multiple transmit queues.
2567 */
4d29515f
DM
2568static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2569 u16 queue_index)
f25f4e44 2570{
fd2ea0a7 2571 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2572
2573 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2574}
2575
4d29515f
DM
2576static inline bool netif_subqueue_stopped(const struct net_device *dev,
2577 struct sk_buff *skb)
668f895a
PE
2578{
2579 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2580}
bea3348e
SH
2581
2582/**
2583 * netif_wake_subqueue - allow sending packets on subqueue
2584 * @dev: network device
2585 * @queue_index: sub queue index
2586 *
2587 * Resume individual transmit queue of a device with multiple transmit queues.
2588 */
f25f4e44
PWJ
2589static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2590{
fd2ea0a7 2591 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
73466498 2592 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
37437bb2 2593 __netif_schedule(txq->qdisc);
f25f4e44
PWJ
2594}
2595
537c00de 2596#ifdef CONFIG_XPS
53af53ae 2597int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2598 u16 index);
537c00de
AD
2599#else
2600static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2601 const struct cpumask *mask,
537c00de
AD
2602 u16 index)
2603{
2604 return 0;
2605}
2606#endif
2607
a3d22a68
VZ
2608/*
2609 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2610 * as a distribution range limit for the returned value.
2611 */
2612static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 2613 struct sk_buff *skb)
a3d22a68
VZ
2614{
2615 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2616}
2617
bea3348e
SH
2618/**
2619 * netif_is_multiqueue - test if device has multiple transmit queues
2620 * @dev: network device
2621 *
2622 * Check if device has multiple transmit queues
bea3348e 2623 */
4d29515f 2624static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2625{
a02cec21 2626 return dev->num_tx_queues > 1;
f25f4e44 2627}
1da177e4 2628
f629d208 2629int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2630
a953be53 2631#ifdef CONFIG_SYSFS
f629d208 2632int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
2633#else
2634static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2635 unsigned int rxq)
2636{
2637 return 0;
2638}
2639#endif
2640
3171d026
BH
2641static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2642 const struct net_device *from_dev)
2643{
ee6ae1a1
JP
2644 int err;
2645
2646 err = netif_set_real_num_tx_queues(to_dev,
2647 from_dev->real_num_tx_queues);
2648 if (err)
2649 return err;
a953be53 2650#ifdef CONFIG_SYSFS
3171d026
BH
2651 return netif_set_real_num_rx_queues(to_dev,
2652 from_dev->real_num_rx_queues);
2653#else
2654 return 0;
2655#endif
2656}
2657
a953be53
MD
2658#ifdef CONFIG_SYSFS
2659static inline unsigned int get_netdev_rx_queue_index(
2660 struct netdev_rx_queue *queue)
2661{
2662 struct net_device *dev = queue->dev;
2663 int index = queue - dev->_rx;
2664
2665 BUG_ON(index >= dev->num_rx_queues);
2666 return index;
2667}
2668#endif
2669
16917b87 2670#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 2671int netif_get_num_default_rss_queues(void);
16917b87 2672
e6247027
ED
2673enum skb_free_reason {
2674 SKB_REASON_CONSUMED,
2675 SKB_REASON_DROPPED,
2676};
2677
2678void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2679void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 2680
e6247027
ED
2681/*
2682 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2683 * interrupt context or with hardware interrupts being disabled.
2684 * (in_irq() || irqs_disabled())
2685 *
2686 * We provide four helpers that can be used in following contexts :
2687 *
2688 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2689 * replacing kfree_skb(skb)
2690 *
2691 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2692 * Typically used in place of consume_skb(skb) in TX completion path
2693 *
2694 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2695 * replacing kfree_skb(skb)
2696 *
2697 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2698 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 2699 */
e6247027
ED
2700static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2701{
2702 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2703}
2704
2705static inline void dev_consume_skb_irq(struct sk_buff *skb)
2706{
2707 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2708}
2709
2710static inline void dev_kfree_skb_any(struct sk_buff *skb)
2711{
2712 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2713}
2714
2715static inline void dev_consume_skb_any(struct sk_buff *skb)
2716{
2717 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2718}
1da177e4 2719
f629d208
JP
2720int netif_rx(struct sk_buff *skb);
2721int netif_rx_ni(struct sk_buff *skb);
2722int netif_receive_skb(struct sk_buff *skb);
2723gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2724void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2725struct sk_buff *napi_get_frags(struct napi_struct *napi);
2726gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
2727struct packet_offload *gro_find_receive_by_type(__be16 type);
2728struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
2729
2730static inline void napi_free_frags(struct napi_struct *napi)
2731{
2732 kfree_skb(napi->skb);
2733 napi->skb = NULL;
2734}
2735
f629d208
JP
2736int netdev_rx_handler_register(struct net_device *dev,
2737 rx_handler_func_t *rx_handler,
2738 void *rx_handler_data);
2739void netdev_rx_handler_unregister(struct net_device *dev);
2740
2741bool dev_valid_name(const char *name);
2742int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2743int dev_ethtool(struct net *net, struct ifreq *);
2744unsigned int dev_get_flags(const struct net_device *);
2745int __dev_change_flags(struct net_device *, unsigned int flags);
2746int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
2747void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2748 unsigned int gchanges);
f629d208
JP
2749int dev_change_name(struct net_device *, const char *);
2750int dev_set_alias(struct net_device *, const char *, size_t);
2751int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2752int dev_set_mtu(struct net_device *, int);
2753void dev_set_group(struct net_device *, int);
2754int dev_set_mac_address(struct net_device *, struct sockaddr *);
2755int dev_change_carrier(struct net_device *, bool new_carrier);
2756int dev_get_phys_port_id(struct net_device *dev,
2757 struct netdev_phys_port_id *ppid);
2758int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
f663dd9a 2759 struct netdev_queue *txq);
a0265d28 2760int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 2761int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1ee481fb 2762bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
1da177e4 2763
20380731 2764extern int netdev_budget;
1da177e4
LT
2765
2766/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 2767void netdev_run_todo(void);
1da177e4 2768
bea3348e
SH
2769/**
2770 * dev_put - release reference to device
2771 * @dev: network device
2772 *
9ef4429b 2773 * Release reference to device to allow it to be freed.
bea3348e 2774 */
1da177e4
LT
2775static inline void dev_put(struct net_device *dev)
2776{
933393f5 2777 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
2778}
2779
bea3348e
SH
2780/**
2781 * dev_hold - get reference to device
2782 * @dev: network device
2783 *
9ef4429b 2784 * Hold reference to device to keep it from being freed.
bea3348e 2785 */
15333061
SH
2786static inline void dev_hold(struct net_device *dev)
2787{
933393f5 2788 this_cpu_inc(*dev->pcpu_refcnt);
15333061 2789}
1da177e4
LT
2790
2791/* Carrier loss detection, dial on demand. The functions netif_carrier_on
2792 * and _off may be called from IRQ context, but it is caller
2793 * who is responsible for serialization of these calls.
b00055aa
SR
2794 *
2795 * The name carrier is inappropriate, these functions should really be
2796 * called netif_lowerlayer_*() because they represent the state of any
2797 * kind of lower layer not just hardware media.
1da177e4
LT
2798 */
2799
f629d208
JP
2800void linkwatch_init_dev(struct net_device *dev);
2801void linkwatch_fire_event(struct net_device *dev);
2802void linkwatch_forget_dev(struct net_device *dev);
1da177e4 2803
bea3348e
SH
2804/**
2805 * netif_carrier_ok - test if carrier present
2806 * @dev: network device
2807 *
2808 * Check if carrier is present on device
2809 */
4d29515f 2810static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
2811{
2812 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2813}
2814
f629d208 2815unsigned long dev_trans_start(struct net_device *dev);
9d21493b 2816
f629d208 2817void __netdev_watchdog_up(struct net_device *dev);
1da177e4 2818
f629d208 2819void netif_carrier_on(struct net_device *dev);
1da177e4 2820
f629d208 2821void netif_carrier_off(struct net_device *dev);
1da177e4 2822
bea3348e
SH
2823/**
2824 * netif_dormant_on - mark device as dormant.
2825 * @dev: network device
2826 *
2827 * Mark device as dormant (as per RFC2863).
2828 *
2829 * The dormant state indicates that the relevant interface is not
2830 * actually in a condition to pass packets (i.e., it is not 'up') but is
2831 * in a "pending" state, waiting for some external event. For "on-
2832 * demand" interfaces, this new state identifies the situation where the
2833 * interface is waiting for events to place it in the up state.
2834 *
2835 */
b00055aa
SR
2836static inline void netif_dormant_on(struct net_device *dev)
2837{
2838 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2839 linkwatch_fire_event(dev);
2840}
2841
bea3348e
SH
2842/**
2843 * netif_dormant_off - set device as not dormant.
2844 * @dev: network device
2845 *
2846 * Device is not in dormant state.
2847 */
b00055aa
SR
2848static inline void netif_dormant_off(struct net_device *dev)
2849{
2850 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2851 linkwatch_fire_event(dev);
2852}
2853
bea3348e
SH
2854/**
2855 * netif_dormant - test if carrier present
2856 * @dev: network device
2857 *
2858 * Check if carrier is present on device
2859 */
4d29515f 2860static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
2861{
2862 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2863}
2864
2865
bea3348e
SH
2866/**
2867 * netif_oper_up - test if device is operational
2868 * @dev: network device
2869 *
2870 * Check if carrier is operational
2871 */
4d29515f 2872static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 2873{
b00055aa
SR
2874 return (dev->operstate == IF_OPER_UP ||
2875 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2876}
2877
bea3348e
SH
2878/**
2879 * netif_device_present - is device available or removed
2880 * @dev: network device
2881 *
2882 * Check if device has not been removed from system.
2883 */
4d29515f 2884static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
2885{
2886 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2887}
2888
f629d208 2889void netif_device_detach(struct net_device *dev);
1da177e4 2890
f629d208 2891void netif_device_attach(struct net_device *dev);
1da177e4
LT
2892
2893/*
2894 * Network interface message level settings
2895 */
1da177e4
LT
2896
2897enum {
2898 NETIF_MSG_DRV = 0x0001,
2899 NETIF_MSG_PROBE = 0x0002,
2900 NETIF_MSG_LINK = 0x0004,
2901 NETIF_MSG_TIMER = 0x0008,
2902 NETIF_MSG_IFDOWN = 0x0010,
2903 NETIF_MSG_IFUP = 0x0020,
2904 NETIF_MSG_RX_ERR = 0x0040,
2905 NETIF_MSG_TX_ERR = 0x0080,
2906 NETIF_MSG_TX_QUEUED = 0x0100,
2907 NETIF_MSG_INTR = 0x0200,
2908 NETIF_MSG_TX_DONE = 0x0400,
2909 NETIF_MSG_RX_STATUS = 0x0800,
2910 NETIF_MSG_PKTDATA = 0x1000,
2911 NETIF_MSG_HW = 0x2000,
2912 NETIF_MSG_WOL = 0x4000,
2913};
2914
2915#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2916#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2917#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2918#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2919#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2920#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2921#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2922#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2923#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2924#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2925#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2926#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2927#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2928#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2929#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2930
2931static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2932{
2933 /* use default */
2934 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2935 return default_msg_enable_bits;
2936 if (debug_value == 0) /* no output */
2937 return 0;
2938 /* set low N bits */
2939 return (1 << debug_value) - 1;
2940}
2941
c773e847 2942static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 2943{
c773e847
DM
2944 spin_lock(&txq->_xmit_lock);
2945 txq->xmit_lock_owner = cpu;
22dd7495
JHS
2946}
2947
fd2ea0a7
DM
2948static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2949{
2950 spin_lock_bh(&txq->_xmit_lock);
2951 txq->xmit_lock_owner = smp_processor_id();
2952}
2953
4d29515f 2954static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 2955{
4d29515f 2956 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
2957 if (likely(ok))
2958 txq->xmit_lock_owner = smp_processor_id();
2959 return ok;
2960}
2961
2962static inline void __netif_tx_unlock(struct netdev_queue *txq)
2963{
2964 txq->xmit_lock_owner = -1;
2965 spin_unlock(&txq->_xmit_lock);
2966}
2967
2968static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2969{
2970 txq->xmit_lock_owner = -1;
2971 spin_unlock_bh(&txq->_xmit_lock);
2972}
2973
08baf561
ED
2974static inline void txq_trans_update(struct netdev_queue *txq)
2975{
2976 if (txq->xmit_lock_owner != -1)
2977 txq->trans_start = jiffies;
2978}
2979
d29f749e
DJ
2980/**
2981 * netif_tx_lock - grab network device transmit lock
2982 * @dev: network device
d29f749e
DJ
2983 *
2984 * Get network device transmit lock
2985 */
22dd7495
JHS
2986static inline void netif_tx_lock(struct net_device *dev)
2987{
e8a0464c 2988 unsigned int i;
c3f26a26 2989 int cpu;
c773e847 2990
c3f26a26
DM
2991 spin_lock(&dev->tx_global_lock);
2992 cpu = smp_processor_id();
e8a0464c
DM
2993 for (i = 0; i < dev->num_tx_queues; i++) {
2994 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
2995
2996 /* We are the only thread of execution doing a
2997 * freeze, but we have to grab the _xmit_lock in
2998 * order to synchronize with threads which are in
2999 * the ->hard_start_xmit() handler and already
3000 * checked the frozen bit.
3001 */
e8a0464c 3002 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3003 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3004 __netif_tx_unlock(txq);
e8a0464c 3005 }
932ff279
HX
3006}
3007
3008static inline void netif_tx_lock_bh(struct net_device *dev)
3009{
e8a0464c
DM
3010 local_bh_disable();
3011 netif_tx_lock(dev);
932ff279
HX
3012}
3013
932ff279
HX
3014static inline void netif_tx_unlock(struct net_device *dev)
3015{
e8a0464c
DM
3016 unsigned int i;
3017
3018 for (i = 0; i < dev->num_tx_queues; i++) {
3019 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3020
c3f26a26
DM
3021 /* No need to grab the _xmit_lock here. If the
3022 * queue is not stopped for another reason, we
3023 * force a schedule.
3024 */
3025 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3026 netif_schedule_queue(txq);
c3f26a26
DM
3027 }
3028 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3029}
3030
3031static inline void netif_tx_unlock_bh(struct net_device *dev)
3032{
e8a0464c
DM
3033 netif_tx_unlock(dev);
3034 local_bh_enable();
932ff279
HX
3035}
3036
c773e847 3037#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3038 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3039 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3040 } \
3041}
3042
5efeac44
EB
3043#define HARD_TX_TRYLOCK(dev, txq) \
3044 (((dev->features & NETIF_F_LLTX) == 0) ? \
3045 __netif_tx_trylock(txq) : \
3046 true )
3047
c773e847 3048#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3049 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3050 __netif_tx_unlock(txq); \
22dd7495
JHS
3051 } \
3052}
3053
1da177e4
LT
3054static inline void netif_tx_disable(struct net_device *dev)
3055{
fd2ea0a7 3056 unsigned int i;
c3f26a26 3057 int cpu;
fd2ea0a7 3058
c3f26a26
DM
3059 local_bh_disable();
3060 cpu = smp_processor_id();
fd2ea0a7
DM
3061 for (i = 0; i < dev->num_tx_queues; i++) {
3062 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3063
3064 __netif_tx_lock(txq, cpu);
fd2ea0a7 3065 netif_tx_stop_queue(txq);
c3f26a26 3066 __netif_tx_unlock(txq);
fd2ea0a7 3067 }
c3f26a26 3068 local_bh_enable();
1da177e4
LT
3069}
3070
e308a5d8
DM
3071static inline void netif_addr_lock(struct net_device *dev)
3072{
3073 spin_lock(&dev->addr_list_lock);
3074}
3075
2429f7ac
JP
3076static inline void netif_addr_lock_nested(struct net_device *dev)
3077{
25175ba5
VY
3078 int subclass = SINGLE_DEPTH_NESTING;
3079
3080 if (dev->netdev_ops->ndo_get_lock_subclass)
3081 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3082
3083 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3084}
3085
e308a5d8
DM
3086static inline void netif_addr_lock_bh(struct net_device *dev)
3087{
3088 spin_lock_bh(&dev->addr_list_lock);
3089}
3090
3091static inline void netif_addr_unlock(struct net_device *dev)
3092{
3093 spin_unlock(&dev->addr_list_lock);
3094}
3095
3096static inline void netif_addr_unlock_bh(struct net_device *dev)
3097{
3098 spin_unlock_bh(&dev->addr_list_lock);
3099}
3100
f001fde5 3101/*
31278e71 3102 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3103 * rcu_read_lock held.
3104 */
3105#define for_each_dev_addr(dev, ha) \
31278e71 3106 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3107
1da177e4
LT
3108/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3109
f629d208 3110void ether_setup(struct net_device *dev);
1da177e4
LT
3111
3112/* Support for loadable net-drivers */
f629d208 3113struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3114 unsigned char name_assign_type,
f629d208
JP
3115 void (*setup)(struct net_device *),
3116 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3117#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3118 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3119
c835a677
TG
3120#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3121 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3122 count)
36909ea4 3123
f629d208
JP
3124int register_netdev(struct net_device *dev);
3125void unregister_netdev(struct net_device *dev);
f001fde5 3126
22bedad3 3127/* General hardware address lists handling functions */
f629d208
JP
3128int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3129 struct netdev_hw_addr_list *from_list, int addr_len);
3130void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3131 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3132int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3133 struct net_device *dev,
3134 int (*sync)(struct net_device *, const unsigned char *),
3135 int (*unsync)(struct net_device *,
3136 const unsigned char *));
3137void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3138 struct net_device *dev,
3139 int (*unsync)(struct net_device *,
3140 const unsigned char *));
f629d208 3141void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3142
f001fde5 3143/* Functions used for device addresses handling */
f629d208
JP
3144int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3145 unsigned char addr_type);
3146int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3147 unsigned char addr_type);
f629d208
JP
3148void dev_addr_flush(struct net_device *dev);
3149int dev_addr_init(struct net_device *dev);
a748ee24
JP
3150
3151/* Functions used for unicast addresses handling */
f629d208
JP
3152int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3153int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3154int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3155int dev_uc_sync(struct net_device *to, struct net_device *from);
3156int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3157void dev_uc_unsync(struct net_device *to, struct net_device *from);
3158void dev_uc_flush(struct net_device *dev);
3159void dev_uc_init(struct net_device *dev);
f001fde5 3160
670e5b8e
AD
3161/**
3162 * __dev_uc_sync - Synchonize device's unicast list
3163 * @dev: device to sync
3164 * @sync: function to call if address should be added
3165 * @unsync: function to call if address should be removed
3166 *
3167 * Add newly added addresses to the interface, and release
3168 * addresses that have been deleted.
3169 **/
3170static inline int __dev_uc_sync(struct net_device *dev,
3171 int (*sync)(struct net_device *,
3172 const unsigned char *),
3173 int (*unsync)(struct net_device *,
3174 const unsigned char *))
3175{
3176 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3177}
3178
3179/**
3180 * __dev_uc_unsync - Remove synchonized addresses from device
3181 * @dev: device to sync
3182 * @unsync: function to call if address should be removed
3183 *
3184 * Remove all addresses that were added to the device by dev_uc_sync().
3185 **/
3186static inline void __dev_uc_unsync(struct net_device *dev,
3187 int (*unsync)(struct net_device *,
3188 const unsigned char *))
3189{
3190 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3191}
3192
22bedad3 3193/* Functions used for multicast addresses handling */
f629d208
JP
3194int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3195int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3196int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3197int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3198int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3199int dev_mc_sync(struct net_device *to, struct net_device *from);
3200int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3201void dev_mc_unsync(struct net_device *to, struct net_device *from);
3202void dev_mc_flush(struct net_device *dev);
3203void dev_mc_init(struct net_device *dev);
f001fde5 3204
670e5b8e
AD
3205/**
3206 * __dev_mc_sync - Synchonize device's multicast list
3207 * @dev: device to sync
3208 * @sync: function to call if address should be added
3209 * @unsync: function to call if address should be removed
3210 *
3211 * Add newly added addresses to the interface, and release
3212 * addresses that have been deleted.
3213 **/
3214static inline int __dev_mc_sync(struct net_device *dev,
3215 int (*sync)(struct net_device *,
3216 const unsigned char *),
3217 int (*unsync)(struct net_device *,
3218 const unsigned char *))
3219{
3220 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3221}
3222
3223/**
3224 * __dev_mc_unsync - Remove synchonized addresses from device
3225 * @dev: device to sync
3226 * @unsync: function to call if address should be removed
3227 *
3228 * Remove all addresses that were added to the device by dev_mc_sync().
3229 **/
3230static inline void __dev_mc_unsync(struct net_device *dev,
3231 int (*unsync)(struct net_device *,
3232 const unsigned char *))
3233{
3234 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3235}
3236
4417da66 3237/* Functions used for secondary unicast and multicast support */
f629d208
JP
3238void dev_set_rx_mode(struct net_device *dev);
3239void __dev_set_rx_mode(struct net_device *dev);
3240int dev_set_promiscuity(struct net_device *dev, int inc);
3241int dev_set_allmulti(struct net_device *dev, int inc);
3242void netdev_state_change(struct net_device *dev);
3243void netdev_notify_peers(struct net_device *dev);
3244void netdev_features_change(struct net_device *dev);
1da177e4 3245/* Load a device via the kmod */
f629d208
JP
3246void dev_load(struct net *net, const char *name);
3247struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3248 struct rtnl_link_stats64 *storage);
3249void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3250 const struct net_device_stats *netdev_stats);
eeda3fd6 3251
1da177e4 3252extern int netdev_max_backlog;
3b098e2d 3253extern int netdev_tstamp_prequeue;
1da177e4 3254extern int weight_p;
0a14842f 3255extern int bpf_jit_enable;
9ff162a8 3256
f629d208 3257bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3258struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3259 struct list_head **iter);
f629d208
JP
3260struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3261 struct list_head **iter);
8b5be856 3262
44a40855
VY
3263/* iterate through upper list, must be called under RCU read lock */
3264#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3265 for (iter = &(dev)->adj_list.upper, \
3266 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3267 updev; \
3268 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3269
8b5be856 3270/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3271#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3272 for (iter = &(dev)->all_adj_list.upper, \
3273 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3274 updev; \
3275 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3276
f629d208
JP
3277void *netdev_lower_get_next_private(struct net_device *dev,
3278 struct list_head **iter);
3279void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3280 struct list_head **iter);
31088a11
VF
3281
3282#define netdev_for_each_lower_private(dev, priv, iter) \
3283 for (iter = (dev)->adj_list.lower.next, \
3284 priv = netdev_lower_get_next_private(dev, &(iter)); \
3285 priv; \
3286 priv = netdev_lower_get_next_private(dev, &(iter)))
3287
3288#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3289 for (iter = &(dev)->adj_list.lower, \
3290 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3291 priv; \
3292 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3293
4085ebe8
VY
3294void *netdev_lower_get_next(struct net_device *dev,
3295 struct list_head **iter);
3296#define netdev_for_each_lower_dev(dev, ldev, iter) \
3297 for (iter = &(dev)->adj_list.lower, \
3298 ldev = netdev_lower_get_next(dev, &(iter)); \
3299 ldev; \
3300 ldev = netdev_lower_get_next(dev, &(iter)))
3301
f629d208 3302void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3303void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3304struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3305struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3306int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3307int netdev_master_upper_dev_link(struct net_device *dev,
9ff162a8 3308 struct net_device *upper_dev);
f629d208
JP
3309int netdev_master_upper_dev_link_private(struct net_device *dev,
3310 struct net_device *upper_dev,
3311 void *private);
3312void netdev_upper_dev_unlink(struct net_device *dev,
3313 struct net_device *upper_dev);
5bb025fa 3314void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3315void *netdev_lower_dev_get_private(struct net_device *dev,
3316 struct net_device *lower_dev);
4085ebe8
VY
3317int dev_get_nest_level(struct net_device *dev,
3318 bool (*type_check)(struct net_device *dev));
f629d208
JP
3319int skb_checksum_help(struct sk_buff *skb);
3320struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3321 netdev_features_t features, bool tx_path);
3322struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3323 netdev_features_t features);
12b0004d
CW
3324
3325static inline
3326struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3327{
3328 return __skb_gso_segment(skb, features, true);
3329}
53d6471c 3330__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3331
3332static inline bool can_checksum_protocol(netdev_features_t features,
3333 __be16 protocol)
3334{
3335 return ((features & NETIF_F_GEN_CSUM) ||
3336 ((features & NETIF_F_V4_CSUM) &&
3337 protocol == htons(ETH_P_IP)) ||
3338 ((features & NETIF_F_V6_CSUM) &&
3339 protocol == htons(ETH_P_IPV6)) ||
3340 ((features & NETIF_F_FCOE_CRC) &&
3341 protocol == htons(ETH_P_FCOE)));
3342}
12b0004d 3343
fb286bb2 3344#ifdef CONFIG_BUG
f629d208 3345void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3346#else
3347static inline void netdev_rx_csum_fault(struct net_device *dev)
3348{
3349}
3350#endif
1da177e4 3351/* rx skb timestamps */
f629d208
JP
3352void net_enable_timestamp(void);
3353void net_disable_timestamp(void);
1da177e4 3354
20380731 3355#ifdef CONFIG_PROC_FS
f629d208 3356int __init dev_proc_init(void);
900ff8c6
CW
3357#else
3358#define dev_proc_init() 0
20380731
ACM
3359#endif
3360
42a2d923
LT
3361int netdev_class_create_file_ns(struct class_attribute *class_attr,
3362 const void *ns);
3363void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3364 const void *ns);
58292cbe
TH
3365
3366static inline int netdev_class_create_file(struct class_attribute *class_attr)
3367{
3368 return netdev_class_create_file_ns(class_attr, NULL);
3369}
3370
3371static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3372{
3373 netdev_class_remove_file_ns(class_attr, NULL);
3374}
b8a9787e 3375
04600794
JB
3376extern struct kobj_ns_type_operations net_ns_type_operations;
3377
f629d208 3378const char *netdev_drivername(const struct net_device *dev);
6579e57b 3379
f629d208 3380void linkwatch_run_queue(void);
20380731 3381
da08143b
MK
3382static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3383 netdev_features_t f2)
3384{
3385 if (f1 & NETIF_F_GEN_CSUM)
3386 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3387 if (f2 & NETIF_F_GEN_CSUM)
3388 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3389 f1 &= f2;
3390 if (f1 & NETIF_F_GEN_CSUM)
3391 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3392
3393 return f1;
3394}
3395
c8f44aff
MM
3396static inline netdev_features_t netdev_get_wanted_features(
3397 struct net_device *dev)
5455c699
MM
3398{
3399 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3400}
c8f44aff
MM
3401netdev_features_t netdev_increment_features(netdev_features_t all,
3402 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3403
3404/* Allow TSO being used on stacked device :
3405 * Performing the GSO segmentation before last device
3406 * is a performance improvement.
3407 */
3408static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3409 netdev_features_t mask)
3410{
3411 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3412}
3413
6cb6a27c 3414int __netdev_update_features(struct net_device *dev);
5455c699 3415void netdev_update_features(struct net_device *dev);
afe12cc8 3416void netdev_change_features(struct net_device *dev);
7f353bf2 3417
fc4a7489
PM
3418void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3419 struct net_device *dev);
3420
c1e756bf 3421netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3422
4d29515f 3423static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3424{
c8f44aff 3425 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3426
3427 /* check flags correspondence */
3428 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3429 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3430 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3431 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3432 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3433 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
3434 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3435 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3436 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3437 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3438 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3439 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3440 BUILD_BUG_ON(SKB_GSO_MPLS != (NETIF_F_GSO_MPLS >> NETIF_F_GSO_SHIFT));
0345e186 3441
d6b4991a 3442 return (features & feature) == feature;
576a30eb
HX
3443}
3444
4d29515f 3445static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3446{
278b2513 3447 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3448 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3449}
3450
4d29515f
DM
3451static inline bool netif_needs_gso(struct sk_buff *skb,
3452 netdev_features_t features)
7967168c 3453{
fc741216 3454 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3455 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3456 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3457}
3458
82cc1a7a
PWJ
3459static inline void netif_set_gso_max_size(struct net_device *dev,
3460 unsigned int size)
3461{
3462 dev->gso_max_size = size;
3463}
3464
7a7ffbab
WCC
3465static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3466 int pulled_hlen, u16 mac_offset,
3467 int mac_len)
3468{
3469 skb->protocol = protocol;
3470 skb->encapsulation = 1;
3471 skb_push(skb, pulled_hlen);
3472 skb_reset_transport_header(skb);
3473 skb->mac_header = mac_offset;
3474 skb->network_header = skb->mac_header + mac_len;
3475 skb->mac_len = mac_len;
3476}
3477
a6cc0cfa
JF
3478static inline bool netif_is_macvlan(struct net_device *dev)
3479{
3480 return dev->priv_flags & IFF_MACVLAN;
3481}
3482
8a7fbfab 3483static inline bool netif_is_bond_master(struct net_device *dev)
3484{
3485 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3486}
3487
4d29515f 3488static inline bool netif_is_bond_slave(struct net_device *dev)
1765a575
JP
3489{
3490 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3491}
3492
3bdc0eba
BG
3493static inline bool netif_supports_nofcs(struct net_device *dev)
3494{
3495 return dev->priv_flags & IFF_SUPP_NOFCS;
3496}
3497
505d4f73 3498extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 3499
571ba423
JP
3500/* Logging, debugging and troubleshooting/diagnostic helpers. */
3501
3502/* netdev_printk helpers, similar to dev_printk */
3503
3504static inline const char *netdev_name(const struct net_device *dev)
3505{
c6f854d5
VF
3506 if (!dev->name[0] || strchr(dev->name, '%'))
3507 return "(unnamed net_device)";
571ba423
JP
3508 return dev->name;
3509}
3510
ccc7f496
VF
3511static inline const char *netdev_reg_state(const struct net_device *dev)
3512{
3513 switch (dev->reg_state) {
3514 case NETREG_UNINITIALIZED: return " (uninitialized)";
3515 case NETREG_REGISTERED: return "";
3516 case NETREG_UNREGISTERING: return " (unregistering)";
3517 case NETREG_UNREGISTERED: return " (unregistered)";
3518 case NETREG_RELEASED: return " (released)";
3519 case NETREG_DUMMY: return " (dummy)";
3520 }
3521
3522 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3523 return " (unknown)";
3524}
3525
f629d208 3526__printf(3, 4)
b9075fa9
JP
3527int netdev_printk(const char *level, const struct net_device *dev,
3528 const char *format, ...);
f629d208 3529__printf(2, 3)
b9075fa9 3530int netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 3531__printf(2, 3)
b9075fa9 3532int netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 3533__printf(2, 3)
b9075fa9 3534int netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 3535__printf(2, 3)
b9075fa9 3536int netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 3537__printf(2, 3)
b9075fa9 3538int netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 3539__printf(2, 3)
b9075fa9 3540int netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 3541__printf(2, 3)
b9075fa9 3542int netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 3543
8909c9ad
VK
3544#define MODULE_ALIAS_NETDEV(device) \
3545 MODULE_ALIAS("netdev-" device)
3546
b558c96f 3547#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
3548#define netdev_dbg(__dev, format, args...) \
3549do { \
ffa10cb4 3550 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 3551} while (0)
b558c96f
JC
3552#elif defined(DEBUG)
3553#define netdev_dbg(__dev, format, args...) \
3554 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
3555#else
3556#define netdev_dbg(__dev, format, args...) \
3557({ \
3558 if (0) \
3559 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3560 0; \
3561})
3562#endif
3563
3564#if defined(VERBOSE_DEBUG)
3565#define netdev_vdbg netdev_dbg
3566#else
3567
3568#define netdev_vdbg(dev, format, args...) \
3569({ \
3570 if (0) \
3571 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3572 0; \
3573})
3574#endif
3575
3576/*
3577 * netdev_WARN() acts like dev_printk(), but with the key difference
3578 * of using a WARN/WARN_ON to get the message out, including the
3579 * file/line information and a backtrace.
3580 */
3581#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
3582 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3583 netdev_reg_state(dev), ##args)
571ba423 3584
b3d95c5c
JP
3585/* netif printk helpers, similar to netdev_printk */
3586
3587#define netif_printk(priv, type, level, dev, fmt, args...) \
3588do { \
3589 if (netif_msg_##type(priv)) \
3590 netdev_printk(level, (dev), fmt, ##args); \
3591} while (0)
3592
f45f4321
JP
3593#define netif_level(level, priv, type, dev, fmt, args...) \
3594do { \
3595 if (netif_msg_##type(priv)) \
3596 netdev_##level(dev, fmt, ##args); \
3597} while (0)
3598
b3d95c5c 3599#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 3600 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 3601#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 3602 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 3603#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 3604 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 3605#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 3606 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 3607#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 3608 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 3609#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 3610 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 3611#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 3612 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 3613
0053ea9c 3614#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
3615#define netif_dbg(priv, type, netdev, format, args...) \
3616do { \
3617 if (netif_msg_##type(priv)) \
b5fb0a03 3618 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 3619} while (0)
0053ea9c
JP
3620#elif defined(DEBUG)
3621#define netif_dbg(priv, type, dev, format, args...) \
3622 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
3623#else
3624#define netif_dbg(priv, type, dev, format, args...) \
3625({ \
3626 if (0) \
3627 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3628 0; \
3629})
3630#endif
3631
3632#if defined(VERBOSE_DEBUG)
bcfcc450 3633#define netif_vdbg netif_dbg
b3d95c5c
JP
3634#else
3635#define netif_vdbg(priv, type, dev, format, args...) \
3636({ \
3637 if (0) \
a4ed89cb 3638 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
3639 0; \
3640})
3641#endif
571ba423 3642
900ff8c6
CW
3643/*
3644 * The list of packet types we will receive (as opposed to discard)
3645 * and the routines to invoke.
3646 *
3647 * Why 16. Because with 16 the only overlap we get on a hash of the
3648 * low nibble of the protocol value is RARP/SNAP/X.25.
3649 *
3650 * NOTE: That is no longer true with the addition of VLAN tags. Not
3651 * sure which should go first, but I bet it won't make much
3652 * difference if we are running VLANs. The good news is that
3653 * this protocol won't be in the list unless compiled in, so
3654 * the average user (w/out VLANs) will not be adversely affected.
3655 * --BLG
3656 *
3657 * 0800 IP
3658 * 8100 802.1Q VLAN
3659 * 0001 802.3
3660 * 0002 AX.25
3661 * 0004 802.2
3662 * 8035 RARP
3663 * 0005 SNAP
3664 * 0805 X.25
3665 * 0806 ARP
3666 * 8137 IPX
3667 * 0009 Localtalk
3668 * 86DD IPv6
3669 */
3670#define PTYPE_HASH_SIZE (16)
3671#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3672
385a154c 3673#endif /* _LINUX_NETDEVICE_H */
This page took 1.892307 seconds and 5 git commands to generate.