]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
net, sched: add clsact qdisc
[mirror_ubuntu-artful-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
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>
61bd3857 53#include <uapi/linux/if_bonding.h>
a59e2ecb 54
115c1d6e 55struct netpoll_info;
313162d0 56struct device;
c1f19b51 57struct phy_device;
704232c2
JB
58/* 802.11 specific */
59struct wireless_dev;
98a18b6f
AA
60/* 802.15.4 specific */
61struct wpan_dev;
03c57747 62struct mpls_dev;
1da177e4 63
f629d208
JP
64void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
d07d7507 66
9a1654ba
JP
67/* Backlog congestion levels */
68#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69#define NET_RX_DROP 1 /* packet dropped */
70
572a9d7b
PM
71/*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88/* qdisc ->enqueue() return codes. */
89#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
90#define NET_XMIT_DROP 0x01 /* skb dropped */
91#define NET_XMIT_CN 0x02 /* congestion notification */
92#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 94
b9df3cb8
GR
95/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 98#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
99#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
dc1f8bf6 101/* Driver transmit return codes */
9a1654ba 102#define NETDEV_TX_MASK 0xf0
572a9d7b 103
dc1f8bf6 104enum netdev_tx {
572a9d7b 105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
dc1f8bf6
SH
109};
110typedef enum netdev_tx netdev_tx_t;
111
9a1654ba
JP
112/*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116static inline bool dev_xmit_complete(int rc)
117{
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128}
129
1da177e4
LT
130/*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
fe2918b0 134
c0eb4540
KS
135#if defined(CONFIG_HYPERV_NET)
136# define LL_MAX_HEADER 128
137#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
138# if defined(CONFIG_MAC80211_MESH)
139# define LL_MAX_HEADER 128
140# else
141# define LL_MAX_HEADER 96
142# endif
1da177e4 143#else
8388e3da 144# define LL_MAX_HEADER 32
1da177e4
LT
145#endif
146
d11ead75
BH
147#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
148 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
149#define MAX_HEADER LL_MAX_HEADER
150#else
151#define MAX_HEADER (LL_MAX_HEADER + 48)
152#endif
153
154/*
be1f3c2c
BH
155 * Old network device statistics. Fields are native words
156 * (unsigned long) so they can be read and written atomically.
1da177e4 157 */
fe2918b0 158
d94d9fee 159struct net_device_stats {
3cfde79c
BH
160 unsigned long rx_packets;
161 unsigned long tx_packets;
162 unsigned long rx_bytes;
163 unsigned long tx_bytes;
164 unsigned long rx_errors;
165 unsigned long tx_errors;
166 unsigned long rx_dropped;
167 unsigned long tx_dropped;
168 unsigned long multicast;
1da177e4 169 unsigned long collisions;
1da177e4 170 unsigned long rx_length_errors;
3cfde79c
BH
171 unsigned long rx_over_errors;
172 unsigned long rx_crc_errors;
173 unsigned long rx_frame_errors;
174 unsigned long rx_fifo_errors;
175 unsigned long rx_missed_errors;
1da177e4
LT
176 unsigned long tx_aborted_errors;
177 unsigned long tx_carrier_errors;
178 unsigned long tx_fifo_errors;
179 unsigned long tx_heartbeat_errors;
180 unsigned long tx_window_errors;
1da177e4
LT
181 unsigned long rx_compressed;
182 unsigned long tx_compressed;
183};
184
1da177e4
LT
185
186#include <linux/cache.h>
187#include <linux/skbuff.h>
188
adc9300e 189#ifdef CONFIG_RPS
c5905afb
IM
190#include <linux/static_key.h>
191extern struct static_key rps_needed;
adc9300e
ED
192#endif
193
1da177e4
LT
194struct neighbour;
195struct neigh_parms;
196struct sk_buff;
197
f001fde5
JP
198struct netdev_hw_addr {
199 struct list_head list;
200 unsigned char addr[MAX_ADDR_LEN];
201 unsigned char type;
ccffad25
JP
202#define NETDEV_HW_ADDR_T_LAN 1
203#define NETDEV_HW_ADDR_T_SAN 2
204#define NETDEV_HW_ADDR_T_SLAVE 3
205#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 206#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 207 bool global_use;
4cd729b0 208 int sync_cnt;
8f8f103d 209 int refcount;
4543fbef 210 int synced;
f001fde5
JP
211 struct rcu_head rcu_head;
212};
213
31278e71
JP
214struct netdev_hw_addr_list {
215 struct list_head list;
216 int count;
217};
218
22bedad3
JP
219#define netdev_hw_addr_list_count(l) ((l)->count)
220#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
221#define netdev_hw_addr_list_for_each(ha, l) \
222 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 223
22bedad3
JP
224#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
225#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
226#define netdev_for_each_uc_addr(ha, dev) \
227 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 228
22bedad3
JP
229#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
230#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 231#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 232 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 233
d94d9fee 234struct hh_cache {
f6b72b62 235 u16 hh_len;
5c25f686 236 u16 __pad;
3644f0ce 237 seqlock_t hh_lock;
1da177e4
LT
238
239 /* cached hardware header; allow for machine alignment needs. */
240#define HH_DATA_MOD 16
241#define HH_DATA_OFF(__len) \
5ba0eac6 242 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
243#define HH_DATA_ALIGN(__len) \
244 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
245 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
246};
247
248/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
249 * Alternative is:
250 * dev->hard_header_len ? (dev->hard_header_len +
251 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
252 *
253 * We could use other alignment values, but we must maintain the
254 * relationship HH alignment <= LL alignment.
255 */
256#define LL_RESERVED_SPACE(dev) \
f5184d26 257 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 258#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 259 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 260
3b04ddde
SH
261struct header_ops {
262 int (*create) (struct sk_buff *skb, struct net_device *dev,
263 unsigned short type, const void *daddr,
95c96174 264 const void *saddr, unsigned int len);
3b04ddde 265 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 266 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
267 void (*cache_update)(struct hh_cache *hh,
268 const struct net_device *dev,
269 const unsigned char *haddr);
270};
271
1da177e4
LT
272/* These flag bits are private to the generic network queueing
273 * layer, they may not be explicitly referenced by any other
274 * code.
275 */
276
d94d9fee 277enum netdev_state_t {
1da177e4
LT
278 __LINK_STATE_START,
279 __LINK_STATE_PRESENT,
1da177e4 280 __LINK_STATE_NOCARRIER,
b00055aa
SR
281 __LINK_STATE_LINKWATCH_PENDING,
282 __LINK_STATE_DORMANT,
1da177e4
LT
283};
284
285
286/*
287 * This structure holds at boot time configured netdevice settings. They
fe2918b0 288 * are then used in the device probing.
1da177e4
LT
289 */
290struct netdev_boot_setup {
291 char name[IFNAMSIZ];
292 struct ifmap map;
293};
294#define NETDEV_BOOT_SETUP_MAX 8
295
f629d208 296int __init netdev_boot_setup(char *str);
1da177e4 297
bea3348e
SH
298/*
299 * Structure for NAPI scheduling similar to tasklet but with weighting
300 */
301struct napi_struct {
302 /* The poll_list must only be managed by the entity which
303 * changes the state of the NAPI_STATE_SCHED bit. This means
304 * whoever atomically sets that bit can add this napi_struct
305 * to the per-cpu poll_list, and whoever clears that bit
306 * can remove from the list right before clearing the bit.
307 */
308 struct list_head poll_list;
309
310 unsigned long state;
311 int weight;
404f7c9e 312 unsigned int gro_count;
bea3348e
SH
313 int (*poll)(struct napi_struct *, int);
314#ifdef CONFIG_NETPOLL
315 spinlock_t poll_lock;
316 int poll_owner;
bea3348e 317#endif
5d38a079 318 struct net_device *dev;
d565b0a1 319 struct sk_buff *gro_list;
5d38a079 320 struct sk_buff *skb;
3b47d303 321 struct hrtimer timer;
404f7c9e 322 struct list_head dev_list;
af12fa6e
ET
323 struct hlist_node napi_hash_node;
324 unsigned int napi_id;
bea3348e
SH
325};
326
d94d9fee 327enum {
bea3348e 328 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 329 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 330 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
331 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
332 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
bea3348e
SH
333};
334
5b252f0c 335enum gro_result {
d1c76af9
HX
336 GRO_MERGED,
337 GRO_MERGED_FREE,
338 GRO_HELD,
339 GRO_NORMAL,
340 GRO_DROP,
341};
5b252f0c 342typedef enum gro_result gro_result_t;
d1c76af9 343
8a4eb573
JP
344/*
345 * enum rx_handler_result - Possible return values for rx_handlers.
346 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
347 * further.
348 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
349 * case skb->dev was changed by rx_handler.
350 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
351 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
352 *
353 * rx_handlers are functions called from inside __netif_receive_skb(), to do
354 * special processing of the skb, prior to delivery to protocol handlers.
355 *
356 * Currently, a net_device can only have a single rx_handler registered. Trying
357 * to register a second rx_handler will return -EBUSY.
358 *
359 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
360 * To unregister a rx_handler on a net_device, use
361 * netdev_rx_handler_unregister().
362 *
363 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
364 * do with the skb.
365 *
366 * If the rx_handler consumed to skb in some way, it should return
367 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
368 * the skb to be delivered in some other ways.
369 *
370 * If the rx_handler changed skb->dev, to divert the skb to another
371 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
372 * new device will be called if it exists.
373 *
374 * If the rx_handler consider the skb should be ignored, it should return
375 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 376 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
377 *
378 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
379 * delivered, it should return RX_HANDLER_PASS.
380 *
381 * A device without a registered rx_handler will behave as if rx_handler
382 * returned RX_HANDLER_PASS.
383 */
384
385enum rx_handler_result {
386 RX_HANDLER_CONSUMED,
387 RX_HANDLER_ANOTHER,
388 RX_HANDLER_EXACT,
389 RX_HANDLER_PASS,
390};
391typedef enum rx_handler_result rx_handler_result_t;
392typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 393
f629d208 394void __napi_schedule(struct napi_struct *n);
bc9ad166 395void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 396
4d29515f 397static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
398{
399 return test_bit(NAPI_STATE_DISABLE, &n->state);
400}
401
bea3348e
SH
402/**
403 * napi_schedule_prep - check if napi can be scheduled
404 * @n: napi context
405 *
406 * Test if NAPI routine is already running, and if not mark
407 * it as running. This is used as a condition variable
a0a46196
DM
408 * insure only one NAPI poll instance runs. We also make
409 * sure there is no pending NAPI disable.
bea3348e 410 */
4d29515f 411static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 412{
a0a46196
DM
413 return !napi_disable_pending(n) &&
414 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
415}
416
417/**
418 * napi_schedule - schedule NAPI poll
419 * @n: napi context
420 *
421 * Schedule NAPI poll routine to be called if it is not already
422 * running.
423 */
424static inline void napi_schedule(struct napi_struct *n)
425{
426 if (napi_schedule_prep(n))
427 __napi_schedule(n);
428}
429
bc9ad166
ED
430/**
431 * napi_schedule_irqoff - schedule NAPI poll
432 * @n: napi context
433 *
434 * Variant of napi_schedule(), assuming hard irqs are masked.
435 */
436static inline void napi_schedule_irqoff(struct napi_struct *n)
437{
438 if (napi_schedule_prep(n))
439 __napi_schedule_irqoff(n);
440}
441
bfe13f54 442/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 443static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
444{
445 if (napi_schedule_prep(napi)) {
446 __napi_schedule(napi);
4d29515f 447 return true;
bfe13f54 448 }
4d29515f 449 return false;
bfe13f54
RD
450}
451
3b47d303
ED
452void __napi_complete(struct napi_struct *n);
453void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
454/**
455 * napi_complete - NAPI processing complete
456 * @n: napi context
457 *
458 * Mark NAPI processing as complete.
3b47d303 459 * Consider using napi_complete_done() instead.
bea3348e 460 */
3b47d303
ED
461static inline void napi_complete(struct napi_struct *n)
462{
463 return napi_complete_done(n, 0);
464}
bea3348e 465
af12fa6e
ET
466/**
467 * napi_hash_add - add a NAPI to global hashtable
468 * @napi: napi context
469 *
470 * generate a new napi_id and store a @napi under it in napi_hash
93d05d4a
ED
471 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL)
472 * Note: This is normally automatically done from netif_napi_add(),
473 * so might disappear in a future linux version.
af12fa6e 474 */
f629d208 475void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
476
477/**
478 * napi_hash_del - remove a NAPI from global table
479 * @napi: napi context
480 *
481 * Warning: caller must observe rcu grace period
34cbe27e
ED
482 * before freeing memory containing @napi, if
483 * this function returns true.
93d05d4a
ED
484 * Note: core networking stack automatically calls it
485 * from netif_napi_del()
486 * Drivers might want to call this helper to combine all
487 * the needed rcu grace periods into a single one.
af12fa6e 488 */
34cbe27e 489bool napi_hash_del(struct napi_struct *napi);
af12fa6e 490
bea3348e
SH
491/**
492 * napi_disable - prevent NAPI from scheduling
493 * @n: napi context
494 *
495 * Stop NAPI from being scheduled on this context.
496 * Waits till any outstanding processing completes.
497 */
3b47d303 498void napi_disable(struct napi_struct *n);
bea3348e
SH
499
500/**
501 * napi_enable - enable NAPI scheduling
502 * @n: napi context
503 *
504 * Resume NAPI from being scheduled on this context.
505 * Must be paired with napi_disable.
506 */
507static inline void napi_enable(struct napi_struct *n)
508{
509 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 510 smp_mb__before_atomic();
bea3348e 511 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 512 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
513}
514
c264c3de
SH
515#ifdef CONFIG_SMP
516/**
517 * napi_synchronize - wait until NAPI is not running
518 * @n: napi context
519 *
520 * Wait until NAPI is done being scheduled on this context.
521 * Waits till any outstanding processing completes but
522 * does not disable future activations.
523 */
524static inline void napi_synchronize(const struct napi_struct *n)
525{
526 while (test_bit(NAPI_STATE_SCHED, &n->state))
527 msleep(1);
528}
529#else
530# define napi_synchronize(n) barrier()
531#endif
532
d94d9fee 533enum netdev_queue_state_t {
73466498
TH
534 __QUEUE_STATE_DRV_XOFF,
535 __QUEUE_STATE_STACK_XOFF,
c3f26a26 536 __QUEUE_STATE_FROZEN,
79d16385 537};
8e2f1a63
DB
538
539#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
540#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
541#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
542
543#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
544#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
545 QUEUE_STATE_FROZEN)
546#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
547 QUEUE_STATE_FROZEN)
548
73466498
TH
549/*
550 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
551 * netif_tx_* functions below are used to manipulate this flag. The
552 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
553 * queue independently. The netif_xmit_*stopped functions below are called
554 * to check if the queue has been stopped by the driver or stack (either
555 * of the XOFF bits are set in the state). Drivers should not need to call
556 * netif_xmit*stopped functions, they should only be using netif_tx_*.
557 */
79d16385 558
bb949fbd 559struct netdev_queue {
6a321cb3
ED
560/*
561 * read mostly part
562 */
bb949fbd 563 struct net_device *dev;
46e5da40 564 struct Qdisc __rcu *qdisc;
b0e1e646 565 struct Qdisc *qdisc_sleeping;
ccf5ff69 566#ifdef CONFIG_SYSFS
1d24eb48
TH
567 struct kobject kobj;
568#endif
f2cd2d3e
ED
569#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
570 int numa_node;
571#endif
6a321cb3
ED
572/*
573 * write mostly part
574 */
575 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
576 int xmit_lock_owner;
9d21493b
ED
577 /*
578 * please use this field instead of dev->trans_start
579 */
580 unsigned long trans_start;
ccf5ff69 581
582 /*
583 * Number of TX timeouts for this queue
584 * (/sys/class/net/DEV/Q/trans_timeout)
585 */
586 unsigned long trans_timeout;
114cf580
TH
587
588 unsigned long state;
589
590#ifdef CONFIG_BQL
591 struct dql dql;
592#endif
822b3b2e 593 unsigned long tx_maxrate;
e8a0464c 594} ____cacheline_aligned_in_smp;
bb949fbd 595
f2cd2d3e
ED
596static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
597{
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 return q->numa_node;
600#else
b236da69 601 return NUMA_NO_NODE;
f2cd2d3e
ED
602#endif
603}
604
605static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
606{
607#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 q->numa_node = node;
609#endif
610}
611
df334545 612#ifdef CONFIG_RPS
0a9627f2
TH
613/*
614 * This structure holds an RPS map which can be of variable length. The
615 * map is an array of CPUs.
616 */
617struct rps_map {
618 unsigned int len;
619 struct rcu_head rcu;
620 u16 cpus[0];
621};
60b778ce 622#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 623
fec5e652 624/*
c445477d
BH
625 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
626 * tail pointer for that CPU's input queue at the time of last enqueue, and
627 * a hardware filter index.
fec5e652
TH
628 */
629struct rps_dev_flow {
630 u16 cpu;
c445477d 631 u16 filter;
fec5e652
TH
632 unsigned int last_qtail;
633};
c445477d 634#define RPS_NO_FILTER 0xffff
fec5e652
TH
635
636/*
637 * The rps_dev_flow_table structure contains a table of flow mappings.
638 */
639struct rps_dev_flow_table {
640 unsigned int mask;
641 struct rcu_head rcu;
fec5e652
TH
642 struct rps_dev_flow flows[0];
643};
644#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 645 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
646
647/*
648 * The rps_sock_flow_table contains mappings of flows to the last CPU
649 * on which they were processed by the application (set in recvmsg).
567e4b79
ED
650 * Each entry is a 32bit value. Upper part is the high order bits
651 * of flow hash, lower part is cpu number.
652 * rps_cpu_mask is used to partition the space, depending on number of
653 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
654 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
655 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
656 */
657struct rps_sock_flow_table {
567e4b79 658 u32 mask;
93c1af6c
ED
659
660 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 661};
567e4b79 662#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
663
664#define RPS_NO_CPU 0xffff
665
567e4b79
ED
666extern u32 rps_cpu_mask;
667extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
668
fec5e652
TH
669static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
670 u32 hash)
671{
672 if (table && hash) {
567e4b79
ED
673 unsigned int index = hash & table->mask;
674 u32 val = hash & ~rps_cpu_mask;
fec5e652
TH
675
676 /* We only give a hint, preemption can change cpu under us */
567e4b79 677 val |= raw_smp_processor_id();
fec5e652 678
567e4b79
ED
679 if (table->ents[index] != val)
680 table->ents[index] = val;
fec5e652
TH
681 }
682}
683
c445477d 684#ifdef CONFIG_RFS_ACCEL
f629d208
JP
685bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
686 u16 filter_id);
c445477d 687#endif
a953be53 688#endif /* CONFIG_RPS */
c445477d 689
0a9627f2
TH
690/* This structure contains an instance of an RX queue. */
691struct netdev_rx_queue {
a953be53 692#ifdef CONFIG_RPS
6e3f7faf
ED
693 struct rps_map __rcu *rps_map;
694 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 695#endif
6e3f7faf 696 struct kobject kobj;
fe822240 697 struct net_device *dev;
0a9627f2 698} ____cacheline_aligned_in_smp;
a953be53
MD
699
700/*
701 * RX queue sysfs structures and functions.
702 */
703struct rx_queue_attribute {
704 struct attribute attr;
705 ssize_t (*show)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, char *buf);
707 ssize_t (*store)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, const char *buf, size_t len);
709};
d314774c 710
bf264145
TH
711#ifdef CONFIG_XPS
712/*
713 * This structure holds an XPS map which can be of variable length. The
714 * map is an array of queues.
715 */
716struct xps_map {
717 unsigned int len;
718 unsigned int alloc_len;
719 struct rcu_head rcu;
720 u16 queues[0];
721};
60b778ce 722#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
723#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
724 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
725
726/*
727 * This structure holds all XPS maps for device. Maps are indexed by CPU.
728 */
729struct xps_dev_maps {
730 struct rcu_head rcu;
a4177869 731 struct xps_map __rcu *cpu_map[0];
bf264145
TH
732};
733#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
734 (nr_cpu_ids * sizeof(struct xps_map *)))
735#endif /* CONFIG_XPS */
736
4f57c087
JF
737#define TC_MAX_QUEUE 16
738#define TC_BITMASK 15
739/* HW offloaded queuing disciplines txq count and offset maps */
740struct netdev_tc_txq {
741 u16 count;
742 u16 offset;
743};
744
68bad94e
NP
745#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
746/*
747 * This structure is to hold information about the device
748 * configured to run FCoE protocol stack.
749 */
750struct netdev_fcoe_hbainfo {
751 char manufacturer[64];
752 char serial_number[64];
753 char hardware_version[64];
754 char driver_version[64];
755 char optionrom_version[64];
756 char firmware_version[64];
757 char model[256];
758 char model_description[256];
759};
760#endif
761
02637fce 762#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 763
02637fce
JP
764/* This structure holds a unique identifier to identify some
765 * physical item (port for example) used by a netdevice.
66b52b0d 766 */
02637fce
JP
767struct netdev_phys_item_id {
768 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
769 unsigned char id_len;
770};
771
d754f98b
SF
772static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
773 struct netdev_phys_item_id *b)
774{
775 return a->id_len == b->id_len &&
776 memcmp(a->id, b->id, a->id_len) == 0;
777}
778
99932d4f
DB
779typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
780 struct sk_buff *skb);
781
d314774c
SH
782/*
783 * This structure defines the management hooks for network devices.
00829823
SH
784 * The following hooks can be defined; unless noted otherwise, they are
785 * optional and can be filled with a null pointer.
d314774c
SH
786 *
787 * int (*ndo_init)(struct net_device *dev);
788 * This function is called once when network device is registered.
789 * The network device can use this to any late stage initializaton
790 * or semantic validattion. It can fail with an error code which will
791 * be propogated back to register_netdev
792 *
793 * void (*ndo_uninit)(struct net_device *dev);
794 * This function is called when device is unregistered or when registration
795 * fails. It is not called if init fails.
796 *
797 * int (*ndo_open)(struct net_device *dev);
798 * This function is called when network device transistions to the up
799 * state.
800 *
801 * int (*ndo_stop)(struct net_device *dev);
802 * This function is called when network device transistions to the down
803 * state.
804 *
dc1f8bf6
SH
805 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
806 * struct net_device *dev);
00829823 807 * Called when a packet needs to be transmitted.
e79d8429
RR
808 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
809 * the queue before that can happen; it's for obsolete devices and weird
810 * corner cases, but the stack really does a non-trivial amount
811 * of useless work if you return NETDEV_TX_BUSY.
dc1f8bf6 812 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
813 * Required can not be NULL.
814 *
cdba756f
ED
815 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
816 * netdev_features_t features);
817 * Adjusts the requested feature flags according to device-specific
818 * constraints, and returns the resulting flags. Must not modify
819 * the device state.
820 *
f663dd9a 821 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 822 * void *accel_priv, select_queue_fallback_t fallback);
00829823
SH
823 * Called to decide which queue to when device supports multiple
824 * transmit queues.
825 *
d314774c
SH
826 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
827 * This function is called to allow device receiver to make
828 * changes to configuration when multicast or promiscious is enabled.
829 *
830 * void (*ndo_set_rx_mode)(struct net_device *dev);
831 * This function is called device changes address list filtering.
01789349
JP
832 * If driver handles unicast address filtering, it should set
833 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
834 *
835 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
836 * This function is called when the Media Access Control address
37b607c5 837 * needs to be changed. If this interface is not defined, the
d314774c
SH
838 * mac address can not be changed.
839 *
840 * int (*ndo_validate_addr)(struct net_device *dev);
841 * Test if Media Access Control address is valid for the device.
842 *
843 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
844 * Called when a user request an ioctl which can't be handled by
845 * the generic interface code. If not defined ioctl's return
846 * not supported error code.
847 *
848 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
849 * Used to set network devices bus interface parameters. This interface
850 * is retained for legacy reason, new devices should use the bus
851 * interface (PCI) for low level management.
852 *
853 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
854 * Called when a user wants to change the Maximum Transfer Unit
855 * of a device. If not defined, any request to change MTU will
856 * will return an error.
857 *
00829823 858 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
859 * Callback uses when the transmitter has not made any progress
860 * for dev->watchdog ticks.
861 *
3cfde79c 862 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 863 * struct rtnl_link_stats64 *storage);
d308e38f 864 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 865 * Called when a user wants to get the network device usage
be1f3c2c 866 * statistics. Drivers must do one of the following:
3cfde79c
BH
867 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
868 * rtnl_link_stats64 structure passed by the caller.
82695d9b 869 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
870 * (which should normally be dev->stats) and return a pointer to
871 * it. The structure may be changed asynchronously only if each
872 * field is written atomically.
873 * 3. Update dev->stats asynchronously and atomically, and define
874 * neither operation.
d314774c 875 *
5d632cb7 876 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
877 * If device support VLAN filtering this function is called when a
878 * VLAN id is registered.
d314774c 879 *
5d632cb7 880 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
881 * If device support VLAN filtering this function is called when a
882 * VLAN id is unregistered.
d314774c
SH
883 *
884 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
885 *
886 * SR-IOV management functions.
887 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
888 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
889 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
890 * int max_tx_rate);
5f8444a3 891 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 892 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
893 * int (*ndo_get_vf_config)(struct net_device *dev,
894 * int vf, struct ifla_vf_info *ivf);
1d8faf48 895 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
896 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
897 * struct nlattr *port[]);
01a3d796
VZ
898 *
899 * Enable or disable the VF ability to query its RSS Redirection Table and
900 * Hash Key. This is needed since on some devices VF share this information
901 * with PF and querying it may adduce a theoretical security risk.
902 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 903 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
904 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
905 * Called to setup 'tc' number of traffic classes in the net device. This
906 * is always called from the stack with the rtnl lock held and netif tx
907 * queues stopped. This allows the netdevice to perform queue management
908 * safely.
c445477d 909 *
e9bce845
YZ
910 * Fiber Channel over Ethernet (FCoE) offload functions.
911 * int (*ndo_fcoe_enable)(struct net_device *dev);
912 * Called when the FCoE protocol stack wants to start using LLD for FCoE
913 * so the underlying device can perform whatever needed configuration or
914 * initialization to support acceleration of FCoE traffic.
915 *
916 * int (*ndo_fcoe_disable)(struct net_device *dev);
917 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
918 * so the underlying device can perform whatever needed clean-ups to
919 * stop supporting acceleration of FCoE traffic.
920 *
921 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
922 * struct scatterlist *sgl, unsigned int sgc);
923 * Called when the FCoE Initiator wants to initialize an I/O that
924 * is a possible candidate for Direct Data Placement (DDP). The LLD can
925 * perform necessary setup and returns 1 to indicate the device is set up
926 * successfully to perform DDP on this I/O, otherwise this returns 0.
927 *
928 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
929 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
930 * indicated by the FC exchange id 'xid', so the underlying device can
931 * clean up and reuse resources for later DDP requests.
932 *
933 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
934 * struct scatterlist *sgl, unsigned int sgc);
935 * Called when the FCoE Target wants to initialize an I/O that
936 * is a possible candidate for Direct Data Placement (DDP). The LLD can
937 * perform necessary setup and returns 1 to indicate the device is set up
938 * successfully to perform DDP on this I/O, otherwise this returns 0.
939 *
68bad94e
NP
940 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
941 * struct netdev_fcoe_hbainfo *hbainfo);
942 * Called when the FCoE Protocol stack wants information on the underlying
943 * device. This information is utilized by the FCoE protocol stack to
944 * register attributes with Fiber Channel management service as per the
945 * FC-GS Fabric Device Management Information(FDMI) specification.
946 *
e9bce845
YZ
947 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
948 * Called when the underlying device wants to override default World Wide
949 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
950 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
951 * protocol stack to use.
952 *
c445477d
BH
953 * RFS acceleration.
954 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
955 * u16 rxq_index, u32 flow_id);
956 * Set hardware filter for RFS. rxq_index is the target queue index;
957 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
958 * Return the filter ID on success, or a negative error code.
fbaec0ea 959 *
8b98a70c 960 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
961 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
962 * Called to make another netdev an underling.
963 *
964 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
965 * Called to release previously enslaved netdev.
5455c699
MM
966 *
967 * Feature/offload setting functions.
c8f44aff 968 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
969 * Called to update device configuration to new features. Passed
970 * feature set might be less than what was returned by ndo_fix_features()).
971 * Must return >0 or -errno if it changed dev->features itself.
972 *
edc7d573 973 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
974 * struct net_device *dev,
f6f6424b 975 * const unsigned char *addr, u16 vid, u16 flags)
77162022 976 * Adds an FDB entry to dev for addr.
1690be63
VY
977 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
978 * struct net_device *dev,
f6f6424b 979 * const unsigned char *addr, u16 vid)
77162022
JF
980 * Deletes the FDB entry from dev coresponding to addr.
981 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
982 * struct net_device *dev, struct net_device *filter_dev,
983 * int idx)
77162022
JF
984 * Used to add FDB entries to dump requests. Implementers should add
985 * entries to skb and update idx with the number of entries.
e5a55a89 986 *
ad41faa8
ND
987 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
988 * u16 flags)
e5a55a89 989 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
990 * struct net_device *dev, u32 filter_mask,
991 * int nlflags)
ad41faa8
ND
992 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
993 * u16 flags);
4bf84c35
JP
994 *
995 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
996 * Called to change device carrier. Soft-devices (like dummy, team, etc)
997 * which do not represent real hardware may define this to allow their
998 * userspace components to manage their virtual carrier state. Devices
999 * that determine carrier state from physical hardware properties (eg
1000 * network cables) or protocol-dependent mechanisms (eg
1001 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1002 *
1003 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1004 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1005 * Called to get ID of physical port of this device. If driver does
1006 * not implement this, it is assumed that the hw is not able to have
1007 * multiple net devices on single physical port.
53cf5275
JG
1008 *
1009 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 1010 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1011 * Called by vxlan to notiy a driver about the UDP port and socket
1012 * address family that vxlan is listnening to. It is called only when
1013 * a new port starts listening. The operation is protected by the
1014 * vxlan_net->sock_lock.
1015 *
a8170d2b
SA
1016 * void (*ndo_add_geneve_port)(struct net_device *dev,
1017 * sa_family_t sa_family, __be16 port);
1018 * Called by geneve to notify a driver about the UDP port and socket
1019 * address family that geneve is listnening to. It is called only when
1020 * a new port starts listening. The operation is protected by the
1021 * geneve_net->sock_lock.
1022 *
1023 * void (*ndo_del_geneve_port)(struct net_device *dev,
1024 * sa_family_t sa_family, __be16 port);
1025 * Called by geneve to notify the driver about a UDP port and socket
1026 * address family that geneve is not listening to anymore. The operation
1027 * is protected by the geneve_net->sock_lock.
1028 *
53cf5275 1029 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 1030 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1031 * Called by vxlan to notify the driver about a UDP port and socket
1032 * address family that vxlan is not listening to anymore. The operation
1033 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
1034 *
1035 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1036 * struct net_device *dev)
1037 * Called by upper layer devices to accelerate switching or other
1038 * station functionality into hardware. 'pdev is the lowerdev
1039 * to use for the offload and 'dev' is the net device that will
1040 * back the offload. Returns a pointer to the private structure
1041 * the upper layer will maintain.
1042 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1043 * Called by upper layer device to delete the station created
1044 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1045 * the station and priv is the structure returned by the add
1046 * operation.
1047 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1048 * struct net_device *dev,
1049 * void *priv);
1050 * Callback to use for xmit over the accelerated station. This
1051 * is used in place of ndo_start_xmit on accelerated net
1052 * devices.
5f35227e
JG
1053 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1054 * struct net_device *dev
1055 * netdev_features_t features);
04ffcb25 1056 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1057 * performing offload operations on a given packet. This is to give
1058 * the device an opportunity to implement any restrictions that cannot
1059 * be otherwise expressed by feature flags. The check is called with
1060 * the set of features that the stack has calculated and it returns
1061 * those the driver believes to be appropriate.
822b3b2e
JF
1062 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1063 * int queue_index, u32 maxrate);
1064 * Called when a user wants to set a max-rate limitation of specific
1065 * TX queue.
a54acb3a
ND
1066 * int (*ndo_get_iflink)(const struct net_device *dev);
1067 * Called to get the iflink value of this device.
d746d707
AK
1068 * void (*ndo_change_proto_down)(struct net_device *dev,
1069 * bool proto_down);
1070 * This function is used to pass protocol port error state information
1071 * to the switch driver. The switch driver can react to the proto_down
1072 * by doing a phys down on the associated switch port.
fc4099f1
PS
1073 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1074 * This function is used to get egress tunnel information for given skb.
1075 * This is useful for retrieving outer tunnel header parameters while
1076 * sampling packet.
d746d707 1077 *
d314774c
SH
1078 */
1079struct net_device_ops {
1080 int (*ndo_init)(struct net_device *dev);
1081 void (*ndo_uninit)(struct net_device *dev);
1082 int (*ndo_open)(struct net_device *dev);
1083 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1084 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1085 struct net_device *dev);
1086 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1087 struct net_device *dev,
1088 netdev_features_t features);
00829823 1089 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1090 struct sk_buff *skb,
99932d4f
DB
1091 void *accel_priv,
1092 select_queue_fallback_t fallback);
d314774c
SH
1093 void (*ndo_change_rx_flags)(struct net_device *dev,
1094 int flags);
d314774c 1095 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1096 int (*ndo_set_mac_address)(struct net_device *dev,
1097 void *addr);
d314774c 1098 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1099 int (*ndo_do_ioctl)(struct net_device *dev,
1100 struct ifreq *ifr, int cmd);
d314774c
SH
1101 int (*ndo_set_config)(struct net_device *dev,
1102 struct ifmap *map);
00829823
SH
1103 int (*ndo_change_mtu)(struct net_device *dev,
1104 int new_mtu);
1105 int (*ndo_neigh_setup)(struct net_device *dev,
1106 struct neigh_parms *);
d314774c
SH
1107 void (*ndo_tx_timeout) (struct net_device *dev);
1108
28172739
ED
1109 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1110 struct rtnl_link_stats64 *storage);
d314774c
SH
1111 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1112
8e586137 1113 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1114 __be16 proto, u16 vid);
8e586137 1115 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1116 __be16 proto, u16 vid);
d314774c 1117#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1118 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1119 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1120 struct netpoll_info *info);
0e34e931 1121 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1122#endif
e0d1095a 1123#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1124 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1125#endif
95c26df8
WM
1126 int (*ndo_set_vf_mac)(struct net_device *dev,
1127 int queue, u8 *mac);
1128 int (*ndo_set_vf_vlan)(struct net_device *dev,
1129 int queue, u16 vlan, u8 qos);
ed616689
SC
1130 int (*ndo_set_vf_rate)(struct net_device *dev,
1131 int vf, int min_tx_rate,
1132 int max_tx_rate);
5f8444a3
GR
1133 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1134 int vf, bool setting);
dd461d6a
HS
1135 int (*ndo_set_vf_trust)(struct net_device *dev,
1136 int vf, bool setting);
95c26df8
WM
1137 int (*ndo_get_vf_config)(struct net_device *dev,
1138 int vf,
1139 struct ifla_vf_info *ivf);
1d8faf48
RE
1140 int (*ndo_set_vf_link_state)(struct net_device *dev,
1141 int vf, int link_state);
3b766cd8
EBE
1142 int (*ndo_get_vf_stats)(struct net_device *dev,
1143 int vf,
1144 struct ifla_vf_stats
1145 *vf_stats);
57b61080
SF
1146 int (*ndo_set_vf_port)(struct net_device *dev,
1147 int vf,
1148 struct nlattr *port[]);
1149 int (*ndo_get_vf_port)(struct net_device *dev,
1150 int vf, struct sk_buff *skb);
01a3d796
VZ
1151 int (*ndo_set_vf_rss_query_en)(
1152 struct net_device *dev,
1153 int vf, bool setting);
4f57c087 1154 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
d11ead75 1155#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1156 int (*ndo_fcoe_enable)(struct net_device *dev);
1157 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1158 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1159 u16 xid,
1160 struct scatterlist *sgl,
1161 unsigned int sgc);
1162 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1163 u16 xid);
6247e086
YZ
1164 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1165 u16 xid,
1166 struct scatterlist *sgl,
1167 unsigned int sgc);
68bad94e
NP
1168 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1169 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1170#endif
1171
d11ead75 1172#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1173#define NETDEV_FCOE_WWNN 0
1174#define NETDEV_FCOE_WWPN 1
1175 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1176 u64 *wwn, int type);
4d288d57 1177#endif
3c9c36bc 1178
c445477d
BH
1179#ifdef CONFIG_RFS_ACCEL
1180 int (*ndo_rx_flow_steer)(struct net_device *dev,
1181 const struct sk_buff *skb,
1182 u16 rxq_index,
1183 u32 flow_id);
1184#endif
fbaec0ea
JP
1185 int (*ndo_add_slave)(struct net_device *dev,
1186 struct net_device *slave_dev);
1187 int (*ndo_del_slave)(struct net_device *dev,
1188 struct net_device *slave_dev);
c8f44aff
MM
1189 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1190 netdev_features_t features);
5455c699 1191 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1192 netdev_features_t features);
da6a8fa0 1193 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1194 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1195
1196 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1197 struct nlattr *tb[],
77162022 1198 struct net_device *dev,
6b6e2725 1199 const unsigned char *addr,
f6f6424b 1200 u16 vid,
77162022
JF
1201 u16 flags);
1202 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1203 struct nlattr *tb[],
77162022 1204 struct net_device *dev,
f6f6424b
JP
1205 const unsigned char *addr,
1206 u16 vid);
77162022
JF
1207 int (*ndo_fdb_dump)(struct sk_buff *skb,
1208 struct netlink_callback *cb,
1209 struct net_device *dev,
5d5eacb3 1210 struct net_device *filter_dev,
77162022 1211 int idx);
e5a55a89
JF
1212
1213 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1214 struct nlmsghdr *nlh,
1215 u16 flags);
e5a55a89
JF
1216 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1217 u32 pid, u32 seq,
6cbdceeb 1218 struct net_device *dev,
46c264da
ND
1219 u32 filter_mask,
1220 int nlflags);
407af329 1221 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1222 struct nlmsghdr *nlh,
1223 u16 flags);
4bf84c35
JP
1224 int (*ndo_change_carrier)(struct net_device *dev,
1225 bool new_carrier);
66b52b0d 1226 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1227 struct netdev_phys_item_id *ppid);
db24a904
DA
1228 int (*ndo_get_phys_port_name)(struct net_device *dev,
1229 char *name, size_t len);
53cf5275
JG
1230 void (*ndo_add_vxlan_port)(struct net_device *dev,
1231 sa_family_t sa_family,
35e42379 1232 __be16 port);
53cf5275
JG
1233 void (*ndo_del_vxlan_port)(struct net_device *dev,
1234 sa_family_t sa_family,
35e42379 1235 __be16 port);
a8170d2b
SA
1236 void (*ndo_add_geneve_port)(struct net_device *dev,
1237 sa_family_t sa_family,
1238 __be16 port);
1239 void (*ndo_del_geneve_port)(struct net_device *dev,
1240 sa_family_t sa_family,
1241 __be16 port);
a6cc0cfa
JF
1242 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1243 struct net_device *dev);
1244 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1245 void *priv);
1246
1247 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1248 struct net_device *dev,
1249 void *priv);
25175ba5 1250 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1251 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1252 int queue_index,
1253 u32 maxrate);
a54acb3a 1254 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1255 int (*ndo_change_proto_down)(struct net_device *dev,
1256 bool proto_down);
fc4099f1
PS
1257 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1258 struct sk_buff *skb);
d314774c
SH
1259};
1260
7aa98047
LR
1261/**
1262 * enum net_device_priv_flags - &struct net_device priv_flags
1263 *
1264 * These are the &struct net_device, they are only set internally
1265 * by drivers and used in the kernel. These flags are invisible to
1266 * userspace, this means that the order of these flags can change
1267 * during any kernel release.
1268 *
1269 * You should have a pretty good reason to be extending these flags.
1270 *
1271 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1272 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1273 * @IFF_BONDING: bonding master or slave
7aa98047 1274 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1275 * @IFF_WAN_HDLC: WAN HDLC device
1276 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1277 * release skb->dst
1278 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1279 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1280 * @IFF_MACVLAN_PORT: device used as macvlan port
1281 * @IFF_BRIDGE_PORT: device used as bridge port
1282 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1283 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1284 * @IFF_UNICAST_FLT: Supports unicast filtering
1285 * @IFF_TEAM_PORT: device used as team port
1286 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1287 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1288 * change when it's running
1289 * @IFF_MACVLAN: Macvlan device
007979ea 1290 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1291 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1292 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1293 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1294 * @IFF_TEAM: device is a team device
7aa98047
LR
1295 */
1296enum netdev_priv_flags {
1297 IFF_802_1Q_VLAN = 1<<0,
1298 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1299 IFF_BONDING = 1<<2,
1300 IFF_ISATAP = 1<<3,
1301 IFF_WAN_HDLC = 1<<4,
1302 IFF_XMIT_DST_RELEASE = 1<<5,
1303 IFF_DONT_BRIDGE = 1<<6,
1304 IFF_DISABLE_NETPOLL = 1<<7,
1305 IFF_MACVLAN_PORT = 1<<8,
1306 IFF_BRIDGE_PORT = 1<<9,
1307 IFF_OVS_DATAPATH = 1<<10,
1308 IFF_TX_SKB_SHARING = 1<<11,
1309 IFF_UNICAST_FLT = 1<<12,
1310 IFF_TEAM_PORT = 1<<13,
1311 IFF_SUPP_NOFCS = 1<<14,
1312 IFF_LIVE_ADDR_CHANGE = 1<<15,
1313 IFF_MACVLAN = 1<<16,
1314 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1315 IFF_IPVLAN_MASTER = 1<<18,
1316 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1317 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1318 IFF_NO_QUEUE = 1<<21,
1319 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1320 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1321 IFF_TEAM = 1<<24,
7aa98047
LR
1322};
1323
1324#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1325#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1326#define IFF_BONDING IFF_BONDING
7aa98047 1327#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1328#define IFF_WAN_HDLC IFF_WAN_HDLC
1329#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1330#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1331#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1332#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1333#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1334#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1335#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1336#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1337#define IFF_TEAM_PORT IFF_TEAM_PORT
1338#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1339#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1340#define IFF_MACVLAN IFF_MACVLAN
02875878 1341#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1342#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1343#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1344#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1345#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1346#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1347#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1348#define IFF_TEAM IFF_TEAM
7aa98047 1349
536721b1
KK
1350/**
1351 * struct net_device - The DEVICE structure.
1352 * Actually, this whole structure is a big mistake. It mixes I/O
1353 * data with strictly "high-level" data, and it has to know about
1354 * almost every data structure used in the INET module.
1355 *
1356 * @name: This is the first field of the "visible" part of this structure
1357 * (i.e. as seen by users in the "Space.c" file). It is the name
1358 * of the interface.
1359 *
1360 * @name_hlist: Device name hash chain, please keep it close to name[]
1361 * @ifalias: SNMP alias
1362 * @mem_end: Shared memory end
1363 * @mem_start: Shared memory start
1364 * @base_addr: Device I/O address
1365 * @irq: Device IRQ number
1366 *
14ffbbb8
TG
1367 * @carrier_changes: Stats to monitor carrier on<->off transitions
1368 *
536721b1
KK
1369 * @state: Generic network queuing layer state, see netdev_state_t
1370 * @dev_list: The global list of network devices
1371 * @napi_list: List entry, that is used for polling napi devices
1372 * @unreg_list: List entry, that is used, when we are unregistering the
1373 * device, see the function unregister_netdev
1374 * @close_list: List entry, that is used, when we are closing the device
1375 *
1376 * @adj_list: Directly linked devices, like slaves for bonding
1377 * @all_adj_list: All linked devices, *including* neighbours
1378 * @features: Currently active device features
1379 * @hw_features: User-changeable features
1380 *
1381 * @wanted_features: User-requested features
1382 * @vlan_features: Mask of features inheritable by VLAN devices
1383 *
1384 * @hw_enc_features: Mask of features inherited by encapsulating devices
1385 * This field indicates what encapsulation
1386 * offloads the hardware is capable of doing,
1387 * and drivers will need to set them appropriately.
1388 *
1389 * @mpls_features: Mask of features inheritable by MPLS
1390 *
1391 * @ifindex: interface index
388069d3 1392 * @group: The group, that the device belongs to
536721b1
KK
1393 *
1394 * @stats: Statistics struct, which was left as a legacy, use
1395 * rtnl_link_stats64 instead
1396 *
1397 * @rx_dropped: Dropped packets by core network,
1398 * do not use this in drivers
1399 * @tx_dropped: Dropped packets by core network,
1400 * do not use this in drivers
1401 *
536721b1
KK
1402 * @wireless_handlers: List of functions to handle Wireless Extensions,
1403 * instead of ioctl,
1404 * see <net/iw_handler.h> for details.
1405 * @wireless_data: Instance data managed by the core of wireless extensions
1406 *
1407 * @netdev_ops: Includes several pointers to callbacks,
1408 * if one wants to override the ndo_*() functions
1409 * @ethtool_ops: Management operations
d476059e 1410 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1411 * of Layer 2 headers.
1412 *
1413 * @flags: Interface flags (a la BSD)
1414 * @priv_flags: Like 'flags' but invisible to userspace,
1415 * see if.h for the definitions
1416 * @gflags: Global flags ( kept as legacy )
1417 * @padded: How much padding added by alloc_netdev()
1418 * @operstate: RFC2863 operstate
1419 * @link_mode: Mapping policy to operstate
1420 * @if_port: Selectable AUI, TP, ...
1421 * @dma: DMA channel
1422 * @mtu: Interface MTU value
1423 * @type: Interface hardware type
880621c2
MB
1424 * @hard_header_len: Hardware header length, which means that this is the
1425 * minimum size of a packet.
536721b1
KK
1426 *
1427 * @needed_headroom: Extra headroom the hardware may need, but not in all
1428 * cases can this be guaranteed
1429 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1430 * cases can this be guaranteed. Some cases also use
1431 * LL_MAX_HEADER instead to allocate the skb
1432 *
1433 * interface address info:
1434 *
1435 * @perm_addr: Permanent hw address
1436 * @addr_assign_type: Hw address assignment type
1437 * @addr_len: Hardware address length
1438 * @neigh_priv_len; Used in neigh_alloc(),
1439 * initialized only in atm/clip.c
1440 * @dev_id: Used to differentiate devices that share
1441 * the same link layer address
1442 * @dev_port: Used to differentiate devices that share
1443 * the same function
1444 * @addr_list_lock: XXX: need comments on this one
536721b1
KK
1445 * @uc_promisc: Counter, that indicates, that promiscuous mode
1446 * has been enabled due to the need to listen to
1447 * additional unicast addresses in a device that
1448 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1449 * @uc: unicast mac addresses
1450 * @mc: multicast mac addresses
1451 * @dev_addrs: list of device hw addresses
1452 * @queues_kset: Group of all Kobjects in the Tx and RX queues
536721b1
KK
1453 * @promiscuity: Number of times, the NIC is told to work in
1454 * Promiscuous mode, if it becomes 0 the NIC will
1455 * exit from working in Promiscuous mode
1456 * @allmulti: Counter, enables or disables allmulticast mode
1457 *
1458 * @vlan_info: VLAN info
1459 * @dsa_ptr: dsa specific data
1460 * @tipc_ptr: TIPC specific data
1461 * @atalk_ptr: AppleTalk link
1462 * @ip_ptr: IPv4 specific data
1463 * @dn_ptr: DECnet specific data
1464 * @ip6_ptr: IPv6 specific data
1465 * @ax25_ptr: AX.25 specific data
1466 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1467 *
1468 * @last_rx: Time of last Rx
1469 * @dev_addr: Hw address (before bcast,
1470 * because most packets are unicast)
1471 *
1472 * @_rx: Array of RX queues
1473 * @num_rx_queues: Number of RX queues
1474 * allocated at register_netdev() time
1475 * @real_num_rx_queues: Number of RX queues currently active in device
1476 *
1477 * @rx_handler: handler for received packets
1478 * @rx_handler_data: XXX: need comments on this one
1479 * @ingress_queue: XXX: need comments on this one
1480 * @broadcast: hw bcast address
1481 *
14ffbbb8
TG
1482 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1483 * indexed by RX queue number. Assigned by driver.
1484 * This must only be set if the ndo_rx_flow_steer
1485 * operation is defined
1486 * @index_hlist: Device index hash chain
1487 *
536721b1
KK
1488 * @_tx: Array of TX queues
1489 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1490 * @real_num_tx_queues: Number of TX queues currently active in device
1491 * @qdisc: Root qdisc from userspace point of view
1492 * @tx_queue_len: Max frames per queue allowed
1493 * @tx_global_lock: XXX: need comments on this one
1494 *
1495 * @xps_maps: XXX: need comments on this one
1496 *
0c4f691f
SF
1497 * @offload_fwd_mark: Offload device fwding mark
1498 *
536721b1
KK
1499 * @trans_start: Time (in jiffies) of last Tx
1500 * @watchdog_timeo: Represents the timeout that is used by
1501 * the watchdog ( see dev_watchdog() )
1502 * @watchdog_timer: List of timers
1503 *
1504 * @pcpu_refcnt: Number of references to this device
1505 * @todo_list: Delayed register/unregister
536721b1
KK
1506 * @link_watch_list: XXX: need comments on this one
1507 *
1508 * @reg_state: Register/unregister state machine
1509 * @dismantle: Device is going to be freed
1510 * @rtnl_link_state: This enum represents the phases of creating
1511 * a new link
1512 *
1513 * @destructor: Called from unregister,
1514 * can be used to call free_netdev
1515 * @npinfo: XXX: need comments on this one
1516 * @nd_net: Network namespace this network device is inside
1517 *
1518 * @ml_priv: Mid-layer private
1519 * @lstats: Loopback statistics
1520 * @tstats: Tunnel statistics
1521 * @dstats: Dummy statistics
1522 * @vstats: Virtual ethernet statistics
1523 *
1524 * @garp_port: GARP
1525 * @mrp_port: MRP
1526 *
1527 * @dev: Class/net/name entry
1528 * @sysfs_groups: Space for optional device, statistics and wireless
1529 * sysfs groups
1530 *
1531 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1532 * @rtnl_link_ops: Rtnl_link_ops
1533 *
1534 * @gso_max_size: Maximum size of generic segmentation offload
1535 * @gso_max_segs: Maximum number of segments that can be passed to the
1536 * NIC for GSO
fcbeb976
ED
1537 * @gso_min_segs: Minimum number of segments that can be passed to the
1538 * NIC for GSO
536721b1
KK
1539 *
1540 * @dcbnl_ops: Data Center Bridging netlink ops
1541 * @num_tc: Number of traffic classes in the net device
1542 * @tc_to_txq: XXX: need comments on this one
1543 * @prio_tc_map XXX: need comments on this one
1544 *
1545 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1546 *
1547 * @priomap: XXX: need comments on this one
1548 * @phydev: Physical device may attach itself
1549 * for hardware timestamping
1550 *
1551 * @qdisc_tx_busylock: XXX: need comments on this one
1552 *
d746d707
AK
1553 * @proto_down: protocol port state information can be sent to the
1554 * switch driver and used to set the phys state of the
1555 * switch port.
1556 *
1da177e4
LT
1557 * FIXME: cleanup struct net_device such that network protocol info
1558 * moves out.
1559 */
1560
d94d9fee 1561struct net_device {
1da177e4 1562 char name[IFNAMSIZ];
9356b8fc 1563 struct hlist_node name_hlist;
0b815a1a 1564 char *ifalias;
1da177e4
LT
1565 /*
1566 * I/O specific fields
1567 * FIXME: Merge these and struct ifmap into one
1568 */
536721b1
KK
1569 unsigned long mem_end;
1570 unsigned long mem_start;
1571 unsigned long base_addr;
1572 int irq;
1da177e4 1573
14ffbbb8
TG
1574 atomic_t carrier_changes;
1575
1da177e4 1576 /*
536721b1
KK
1577 * Some hardware also needs these fields (state,dev_list,
1578 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1579 * part of the usual set specified in Space.c.
1580 */
1581
1da177e4
LT
1582 unsigned long state;
1583
7562f876 1584 struct list_head dev_list;
bea3348e 1585 struct list_head napi_list;
44a0873d 1586 struct list_head unreg_list;
5cde2829 1587 struct list_head close_list;
7866a621
SN
1588 struct list_head ptype_all;
1589 struct list_head ptype_specific;
2f268f12 1590
2f268f12
VF
1591 struct {
1592 struct list_head upper;
1593 struct list_head lower;
1594 } adj_list;
1595
2f268f12
VF
1596 struct {
1597 struct list_head upper;
1598 struct list_head lower;
1599 } all_adj_list;
4c3d5e7b 1600
c8f44aff 1601 netdev_features_t features;
c8f44aff 1602 netdev_features_t hw_features;
c8f44aff 1603 netdev_features_t wanted_features;
c8f44aff 1604 netdev_features_t vlan_features;
6a674e9c 1605 netdev_features_t hw_enc_features;
0d89d203 1606 netdev_features_t mpls_features;
04ed3e74 1607
1da177e4 1608 int ifindex;
7a66bbc9 1609 int group;
1da177e4 1610
c45d286e 1611 struct net_device_stats stats;
015f0688 1612
015f0688
ED
1613 atomic_long_t rx_dropped;
1614 atomic_long_t tx_dropped;
1da177e4 1615
b86e0280 1616#ifdef CONFIG_WIRELESS_EXT
1da177e4 1617 const struct iw_handler_def * wireless_handlers;
1da177e4 1618 struct iw_public_data * wireless_data;
b86e0280 1619#endif
d314774c 1620 const struct net_device_ops *netdev_ops;
76fd8593 1621 const struct ethtool_ops *ethtool_ops;
4170604f 1622#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1623 const struct switchdev_ops *switchdev_ops;
4170604f 1624#endif
1b69c6d0
DA
1625#ifdef CONFIG_NET_L3_MASTER_DEV
1626 const struct l3mdev_ops *l3mdev_ops;
1627#endif
1da177e4 1628
3b04ddde
SH
1629 const struct header_ops *header_ops;
1630
536721b1
KK
1631 unsigned int flags;
1632 unsigned int priv_flags;
1633
1da177e4 1634 unsigned short gflags;
536721b1 1635 unsigned short padded;
1da177e4 1636
536721b1
KK
1637 unsigned char operstate;
1638 unsigned char link_mode;
b00055aa 1639
536721b1
KK
1640 unsigned char if_port;
1641 unsigned char dma;
bdc220da 1642
536721b1
KK
1643 unsigned int mtu;
1644 unsigned short type;
1645 unsigned short hard_header_len;
1da177e4 1646
f5184d26
JB
1647 unsigned short needed_headroom;
1648 unsigned short needed_tailroom;
1649
1da177e4 1650 /* Interface address info. */
536721b1
KK
1651 unsigned char perm_addr[MAX_ADDR_LEN];
1652 unsigned char addr_assign_type;
1653 unsigned char addr_len;
a0a9663d 1654 unsigned short neigh_priv_len;
536721b1
KK
1655 unsigned short dev_id;
1656 unsigned short dev_port;
ccffad25 1657 spinlock_t addr_list_lock;
14ffbbb8
TG
1658 unsigned char name_assign_type;
1659 bool uc_promisc;
536721b1
KK
1660 struct netdev_hw_addr_list uc;
1661 struct netdev_hw_addr_list mc;
1662 struct netdev_hw_addr_list dev_addrs;
1663
4c3d5e7b
ED
1664#ifdef CONFIG_SYSFS
1665 struct kset *queues_kset;
1666#endif
9d45abe1
WC
1667 unsigned int promiscuity;
1668 unsigned int allmulti;
1da177e4 1669
1da177e4
LT
1670
1671 /* Protocol specific pointers */
65ac6a5f 1672
d11ead75 1673#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1674 struct vlan_info __rcu *vlan_info;
65ac6a5f 1675#endif
34a430d7 1676#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1677 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1678#endif
1679#if IS_ENABLED(CONFIG_TIPC)
536721b1 1680 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1681#endif
536721b1
KK
1682 void *atalk_ptr;
1683 struct in_device __rcu *ip_ptr;
1684 struct dn_dev __rcu *dn_ptr;
1685 struct inet6_dev __rcu *ip6_ptr;
1686 void *ax25_ptr;
1687 struct wireless_dev *ieee80211_ptr;
98a18b6f 1688 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1689#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1690 struct mpls_dev __rcu *mpls_ptr;
1691#endif
1da177e4 1692
9356b8fc 1693/*
cd13539b 1694 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1695 */
536721b1 1696 unsigned long last_rx;
4dc89133 1697
9356b8fc 1698 /* Interface address info used in eth_type_trans() */
536721b1 1699 unsigned char *dev_addr;
f001fde5 1700
0a9627f2 1701
a953be53 1702#ifdef CONFIG_SYSFS
0a9627f2
TH
1703 struct netdev_rx_queue *_rx;
1704
0a9627f2 1705 unsigned int num_rx_queues;
62fe0b40 1706 unsigned int real_num_rx_queues;
c445477d 1707
df334545 1708#endif
0a9627f2 1709
3b47d303 1710 unsigned long gro_flush_timeout;
61391cde 1711 rx_handler_func_t __rcu *rx_handler;
1712 void __rcu *rx_handler_data;
e8a0464c 1713
4cda01e8 1714#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1715 struct tcf_proto __rcu *ingress_cl_list;
1716#endif
24824a09 1717 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1718#ifdef CONFIG_NETFILTER_INGRESS
1719 struct list_head nf_hooks_ingress;
1720#endif
d2788d34 1721
536721b1 1722 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1723#ifdef CONFIG_RFS_ACCEL
1724 struct cpu_rmap *rx_cpu_rmap;
1725#endif
1726 struct hlist_node index_hlist;
cd13539b
ED
1727
1728/*
1729 * Cache lines mostly used on transmit path
1730 */
e8a0464c
DM
1731 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1732 unsigned int num_tx_queues;
fd2ea0a7 1733 unsigned int real_num_tx_queues;
af356afa 1734 struct Qdisc *qdisc;
536721b1 1735 unsigned long tx_queue_len;
c3f26a26 1736 spinlock_t tx_global_lock;
14ffbbb8 1737 int watchdog_timeo;
cd13539b 1738
bf264145 1739#ifdef CONFIG_XPS
a4177869 1740 struct xps_dev_maps __rcu *xps_maps;
bf264145 1741#endif
1f211a1b
DB
1742#ifdef CONFIG_NET_CLS_ACT
1743 struct tcf_proto __rcu *egress_cl_list;
1744#endif
0c4f691f
SF
1745#ifdef CONFIG_NET_SWITCHDEV
1746 u32 offload_fwd_mark;
1747#endif
1748
9356b8fc 1749 /* These may be needed for future network-power-down code. */
9d21493b
ED
1750
1751 /*
1752 * trans_start here is expensive for high speed devices on SMP,
1753 * please use netdev_queue->trans_start instead.
1754 */
536721b1 1755 unsigned long trans_start;
9356b8fc 1756
9356b8fc
ED
1757 struct timer_list watchdog_timer;
1758
29b4433d 1759 int __percpu *pcpu_refcnt;
1da177e4 1760 struct list_head todo_list;
1da177e4 1761
e014debe 1762 struct list_head link_watch_list;
572a103d 1763
1da177e4 1764 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1765 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1766 NETREG_UNREGISTERING, /* called unregister_netdevice */
1767 NETREG_UNREGISTERED, /* completed unregister todo */
1768 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1769 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1770 } reg_state:8;
1771
536721b1 1772 bool dismantle;
a2835763
PM
1773
1774 enum {
1775 RTNL_LINK_INITIALIZED,
1776 RTNL_LINK_INITIALIZING,
1777 } rtnl_link_state:16;
1da177e4 1778
d314774c 1779 void (*destructor)(struct net_device *dev);
1da177e4 1780
1da177e4 1781#ifdef CONFIG_NETPOLL
5fbee843 1782 struct netpoll_info __rcu *npinfo;
1da177e4 1783#endif
eae792b7 1784
0c5c9fb5 1785 possible_net_t nd_net;
4a1c5371 1786
4951704b 1787 /* mid-layer private */
a7855c78 1788 union {
536721b1
KK
1789 void *ml_priv;
1790 struct pcpu_lstats __percpu *lstats;
8f84985f 1791 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1792 struct pcpu_dstats __percpu *dstats;
1793 struct pcpu_vstats __percpu *vstats;
a7855c78 1794 };
536721b1 1795
3cc77ec7 1796 struct garp_port __rcu *garp_port;
febf018d 1797 struct mrp_port __rcu *mrp_port;
1da177e4 1798
536721b1 1799 struct device dev;
0c509a6c 1800 const struct attribute_group *sysfs_groups[4];
a953be53 1801 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1802
38f7b870 1803 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1804
82cc1a7a
PWJ
1805 /* for setting kernel sock attribute on TCP connection setup */
1806#define GSO_MAX_SIZE 65536
1807 unsigned int gso_max_size;
30b678d8
BH
1808#define GSO_MAX_SEGS 65535
1809 u16 gso_max_segs;
fcbeb976 1810 u16 gso_min_segs;
7a6b6f51 1811#ifdef CONFIG_DCB
32953543 1812 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1813#endif
4f57c087
JF
1814 u8 num_tc;
1815 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1816 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1817
d11ead75 1818#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1819 unsigned int fcoe_ddp_xid;
5bc1421e 1820#endif
86f8515f 1821#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1822 struct netprio_map __rcu *priomap;
4d288d57 1823#endif
c1f19b51 1824 struct phy_device *phydev;
23d3b8bf 1825 struct lock_class_key *qdisc_tx_busylock;
d746d707 1826 bool proto_down;
1da177e4 1827};
43cb76d9 1828#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1829
1830#define NETDEV_ALIGN 32
1da177e4 1831
4f57c087
JF
1832static inline
1833int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1834{
1835 return dev->prio_tc_map[prio & TC_BITMASK];
1836}
1837
1838static inline
1839int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1840{
1841 if (tc >= dev->num_tc)
1842 return -EINVAL;
1843
1844 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1845 return 0;
1846}
1847
1848static inline
1849void netdev_reset_tc(struct net_device *dev)
1850{
1851 dev->num_tc = 0;
1852 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1853 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1854}
1855
1856static inline
1857int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1858{
1859 if (tc >= dev->num_tc)
1860 return -EINVAL;
1861
1862 dev->tc_to_txq[tc].count = count;
1863 dev->tc_to_txq[tc].offset = offset;
1864 return 0;
1865}
1866
1867static inline
1868int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1869{
1870 if (num_tc > TC_MAX_QUEUE)
1871 return -EINVAL;
1872
1873 dev->num_tc = num_tc;
1874 return 0;
1875}
1876
1877static inline
1878int netdev_get_num_tc(struct net_device *dev)
1879{
1880 return dev->num_tc;
1881}
1882
e8a0464c
DM
1883static inline
1884struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1885 unsigned int index)
1886{
1887 return &dev->_tx[index];
1888}
1889
10c51b56
DB
1890static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1891 const struct sk_buff *skb)
1892{
1893 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1894}
1895
e8a0464c
DM
1896static inline void netdev_for_each_tx_queue(struct net_device *dev,
1897 void (*f)(struct net_device *,
1898 struct netdev_queue *,
1899 void *),
1900 void *arg)
1901{
1902 unsigned int i;
1903
1904 for (i = 0; i < dev->num_tx_queues; i++)
1905 f(dev, &dev->_tx[i], arg);
1906}
1907
f629d208 1908struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1909 struct sk_buff *skb,
1910 void *accel_priv);
8c4c49df 1911
c346dca1
YH
1912/*
1913 * Net namespace inlines
1914 */
1915static inline
1916struct net *dev_net(const struct net_device *dev)
1917{
c2d9ba9b 1918 return read_pnet(&dev->nd_net);
c346dca1
YH
1919}
1920
1921static inline
f5aa23fd 1922void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 1923{
0c5c9fb5 1924 write_pnet(&dev->nd_net, net);
c346dca1
YH
1925}
1926
3e8a72d1 1927static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 1928{
3fc88677 1929#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
1930 if (dev->dsa_ptr != NULL)
1931 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 1932#endif
5aed85ce 1933 return false;
396138f0
LB
1934}
1935
bea3348e
SH
1936/**
1937 * netdev_priv - access network device private data
1938 * @dev: network device
1939 *
1940 * Get network device private data
1941 */
6472ce60 1942static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1943{
1ce8e7b5 1944 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1945}
1946
1da177e4
LT
1947/* Set the sysfs physical device reference for the network logical device
1948 * if set prior to registration will cause a symlink during initialization.
1949 */
43cb76d9 1950#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1951
384912ed 1952/* Set the sysfs device type for the network logical device to allow
3f79410c 1953 * fine-grained identification of different network device types. For
384912ed
MH
1954 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1955 */
1956#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1957
82dc3c63
ED
1958/* Default NAPI poll() weight
1959 * Device drivers are strongly advised to not use bigger value
1960 */
1961#define NAPI_POLL_WEIGHT 64
1962
3b582cc1
SH
1963/**
1964 * netif_napi_add - initialize a napi context
1965 * @dev: network device
1966 * @napi: napi context
1967 * @poll: polling function
1968 * @weight: default weight
1969 *
1970 * netif_napi_add() must be used to initialize a napi context prior to calling
1971 * *any* of the other napi related functions.
1972 */
d565b0a1
HX
1973void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1974 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1975
d64b5e85
ED
1976/**
1977 * netif_tx_napi_add - initialize a napi context
1978 * @dev: network device
1979 * @napi: napi context
1980 * @poll: polling function
1981 * @weight: default weight
1982 *
1983 * This variant of netif_napi_add() should be used from drivers using NAPI
1984 * to exclusively poll a TX queue.
1985 * This will avoid we add it into napi_hash[], thus polluting this hash table.
1986 */
1987static inline void netif_tx_napi_add(struct net_device *dev,
1988 struct napi_struct *napi,
1989 int (*poll)(struct napi_struct *, int),
1990 int weight)
1991{
1992 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
1993 netif_napi_add(dev, napi, poll, weight);
1994}
1995
d8156534
AD
1996/**
1997 * netif_napi_del - remove a napi context
1998 * @napi: napi context
1999 *
2000 * netif_napi_del() removes a napi context from the network device napi list
2001 */
d565b0a1
HX
2002void netif_napi_del(struct napi_struct *napi);
2003
2004struct napi_gro_cb {
78a478d0
HX
2005 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2006 void *frag0;
2007
7489594c
HX
2008 /* Length of frag0. */
2009 unsigned int frag0_len;
2010
86911732
HX
2011 /* This indicates where we are processing relative to skb->data. */
2012 int data_offset;
2013
d565b0a1 2014 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2015 u16 flush;
2016
2017 /* Save the IP ID here and check when we get to the transport layer */
2018 u16 flush_id;
d565b0a1
HX
2019
2020 /* Number of segments aggregated. */
2e71a6f8
ED
2021 u16 count;
2022
15e2396d
TH
2023 /* Start offset for remote checksum offload */
2024 u16 gro_remcsum_start;
2025
2e71a6f8
ED
2026 /* jiffies when first packet was created/queued */
2027 unsigned long age;
86347245 2028
afe93325 2029 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2030 u16 proto;
2031
baa32ff4
TH
2032 /* This is non-zero if the packet may be of the same flow. */
2033 u8 same_flow:1;
2034
b582ef09 2035 /* Used in udp_gro_receive */
573e8fca
TH
2036 u8 udp_mark:1;
2037
2038 /* GRO checksum is valid */
2039 u8 csum_valid:1;
2040
662880f4
TH
2041 /* Number of checksums via CHECKSUM_UNNECESSARY */
2042 u8 csum_cnt:3;
c3c7c254 2043
baa32ff4
TH
2044 /* Free the skb? */
2045 u8 free:2;
2046#define NAPI_GRO_FREE 1
2047#define NAPI_GRO_FREE_STOLEN_HEAD 2
2048
efc98d08
TH
2049 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2050 u8 is_ipv6:1;
2051
baa32ff4
TH
2052 /* 7 bit hole */
2053
bf5a755f
JC
2054 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2055 __wsum csum;
2056
c3c7c254
ED
2057 /* used in skb_gro_receive() slow path */
2058 struct sk_buff *last;
d565b0a1
HX
2059};
2060
2061#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2062
1da177e4 2063struct packet_type {
f2ccd8fa
DM
2064 __be16 type; /* This is really htons(ether_type). */
2065 struct net_device *dev; /* NULL is wildcarded here */
2066 int (*func) (struct sk_buff *,
2067 struct net_device *,
2068 struct packet_type *,
2069 struct net_device *);
c0de08d0
EL
2070 bool (*id_match)(struct packet_type *ptype,
2071 struct sock *sk);
1da177e4
LT
2072 void *af_packet_priv;
2073 struct list_head list;
2074};
2075
f191a1d1 2076struct offload_callbacks {
576a30eb 2077 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2078 netdev_features_t features);
d565b0a1 2079 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2080 struct sk_buff *skb);
299603e8 2081 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2082};
2083
2084struct packet_offload {
2085 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2086 u16 priority;
f191a1d1
VY
2087 struct offload_callbacks callbacks;
2088 struct list_head list;
1da177e4
LT
2089};
2090
a2b12f3c
TH
2091struct udp_offload;
2092
2093struct udp_offload_callbacks {
2094 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2095 struct sk_buff *skb,
2096 struct udp_offload *uoff);
2097 int (*gro_complete)(struct sk_buff *skb,
2098 int nhoff,
2099 struct udp_offload *uoff);
2100};
2101
b582ef09
OG
2102struct udp_offload {
2103 __be16 port;
afe93325 2104 u8 ipproto;
a2b12f3c 2105 struct udp_offload_callbacks callbacks;
b582ef09
OG
2106};
2107
8f84985f
LR
2108/* often modified stats are per cpu, other are shared (netdev->stats) */
2109struct pcpu_sw_netstats {
2110 u64 rx_packets;
2111 u64 rx_bytes;
2112 u64 tx_packets;
2113 u64 tx_bytes;
2114 struct u64_stats_sync syncp;
2115};
2116
aabc92bb
PNA
2117#define __netdev_alloc_pcpu_stats(type, gfp) \
2118({ \
2119 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2120 if (pcpu_stats) { \
2121 int __cpu; \
2122 for_each_possible_cpu(__cpu) { \
2123 typeof(type) *stat; \
2124 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2125 u64_stats_init(&stat->syncp); \
2126 } \
2127 } \
2128 pcpu_stats; \
1c213bd2
WC
2129})
2130
aabc92bb 2131#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2132 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2133
764f5e54
JP
2134enum netdev_lag_tx_type {
2135 NETDEV_LAG_TX_TYPE_UNKNOWN,
2136 NETDEV_LAG_TX_TYPE_RANDOM,
2137 NETDEV_LAG_TX_TYPE_BROADCAST,
2138 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2139 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2140 NETDEV_LAG_TX_TYPE_HASH,
2141};
2142
2143struct netdev_lag_upper_info {
2144 enum netdev_lag_tx_type tx_type;
2145};
2146
fb1b2e3c
JP
2147struct netdev_lag_lower_state_info {
2148 u8 link_up : 1,
2149 tx_enabled : 1;
2150};
2151
1da177e4
LT
2152#include <linux/notifier.h>
2153
dcfe1421
AW
2154/* netdevice notifier chain. Please remember to update the rtnetlink
2155 * notification exclusion list in rtnetlink_event() when adding new
2156 * types.
2157 */
2158#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2159#define NETDEV_DOWN 0x0002
2160#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2161 detected a hardware crash and restarted
2162 - we can use this eg to kick tcp sessions
2163 once done */
2164#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2165#define NETDEV_REGISTER 0x0005
2166#define NETDEV_UNREGISTER 0x0006
1d486bfb 2167#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2168#define NETDEV_CHANGEADDR 0x0008
2169#define NETDEV_GOING_DOWN 0x0009
2170#define NETDEV_CHANGENAME 0x000A
2171#define NETDEV_FEAT_CHANGE 0x000B
2172#define NETDEV_BONDING_FAILOVER 0x000C
2173#define NETDEV_PRE_UP 0x000D
2174#define NETDEV_PRE_TYPE_CHANGE 0x000E
2175#define NETDEV_POST_TYPE_CHANGE 0x000F
2176#define NETDEV_POST_INIT 0x0010
0115e8e3 2177#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2178#define NETDEV_RELEASE 0x0012
2179#define NETDEV_NOTIFY_PEERS 0x0013
2180#define NETDEV_JOIN 0x0014
42e52bf9 2181#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2182#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2183#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2184#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2185#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2186#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2187#define NETDEV_CHANGELOWERSTATE 0x001B
dcfe1421 2188
f629d208
JP
2189int register_netdevice_notifier(struct notifier_block *nb);
2190int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2191
2192struct netdev_notifier_info {
2193 struct net_device *dev;
2194};
2195
be9efd36
JP
2196struct netdev_notifier_change_info {
2197 struct netdev_notifier_info info; /* must be first */
2198 unsigned int flags_changed;
2199};
2200
0e4ead9d
JP
2201struct netdev_notifier_changeupper_info {
2202 struct netdev_notifier_info info; /* must be first */
2203 struct net_device *upper_dev; /* new upper dev */
2204 bool master; /* is upper dev master */
2205 bool linking; /* is the nofication for link or unlink */
29bf24af 2206 void *upper_info; /* upper dev info */
0e4ead9d
JP
2207};
2208
04d48266
JP
2209struct netdev_notifier_changelowerstate_info {
2210 struct netdev_notifier_info info; /* must be first */
2211 void *lower_state_info; /* is lower dev state */
2212};
2213
75538c2b
CW
2214static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2215 struct net_device *dev)
2216{
2217 info->dev = dev;
2218}
2219
351638e7
JP
2220static inline struct net_device *
2221netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2222{
2223 return info->dev;
2224}
2225
f629d208 2226int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2227
2228
1da177e4
LT
2229extern rwlock_t dev_base_lock; /* Device list lock */
2230
881d966b
EB
2231#define for_each_netdev(net, d) \
2232 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2233#define for_each_netdev_reverse(net, d) \
2234 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2235#define for_each_netdev_rcu(net, d) \
2236 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2237#define for_each_netdev_safe(net, d, n) \
2238 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2239#define for_each_netdev_continue(net, d) \
2240 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2241#define for_each_netdev_continue_rcu(net, d) \
2242 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2243#define for_each_netdev_in_bond_rcu(bond, slave) \
2244 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2245 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2246#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2247
a050c33f
DL
2248static inline struct net_device *next_net_device(struct net_device *dev)
2249{
2250 struct list_head *lh;
2251 struct net *net;
2252
c346dca1 2253 net = dev_net(dev);
a050c33f
DL
2254 lh = dev->dev_list.next;
2255 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2256}
2257
ce81b76a
ED
2258static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2259{
2260 struct list_head *lh;
2261 struct net *net;
2262
2263 net = dev_net(dev);
ccf43438 2264 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2265 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2266}
2267
a050c33f
DL
2268static inline struct net_device *first_net_device(struct net *net)
2269{
2270 return list_empty(&net->dev_base_head) ? NULL :
2271 net_device_entry(net->dev_base_head.next);
2272}
7562f876 2273
ccf43438
ED
2274static inline struct net_device *first_net_device_rcu(struct net *net)
2275{
2276 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2277
2278 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2279}
2280
f629d208
JP
2281int netdev_boot_setup_check(struct net_device *dev);
2282unsigned long netdev_boot_base(const char *prefix, int unit);
2283struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2284 const char *hwaddr);
2285struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2286struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2287void dev_add_pack(struct packet_type *pt);
2288void dev_remove_pack(struct packet_type *pt);
2289void __dev_remove_pack(struct packet_type *pt);
2290void dev_add_offload(struct packet_offload *po);
2291void dev_remove_offload(struct packet_offload *po);
f629d208 2292
a54acb3a 2293int dev_get_iflink(const struct net_device *dev);
fc4099f1 2294int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2295struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2296 unsigned short mask);
f629d208
JP
2297struct net_device *dev_get_by_name(struct net *net, const char *name);
2298struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2299struct net_device *__dev_get_by_name(struct net *net, const char *name);
2300int dev_alloc_name(struct net_device *dev, const char *name);
2301int dev_open(struct net_device *dev);
2302int dev_close(struct net_device *dev);
99c4a26a 2303int dev_close_many(struct list_head *head, bool unlink);
f629d208 2304void dev_disable_lro(struct net_device *dev);
0c4b51f0 2305int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2306int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2307int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2308int register_netdevice(struct net_device *dev);
2309void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2310void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2311static inline void unregister_netdevice(struct net_device *dev)
2312{
2313 unregister_netdevice_queue(dev, NULL);
2314}
2315
f629d208
JP
2316int netdev_refcnt_read(const struct net_device *dev);
2317void free_netdev(struct net_device *dev);
74d332c1 2318void netdev_freemem(struct net_device *dev);
f629d208
JP
2319void synchronize_net(void);
2320int init_dummy_netdev(struct net_device *dev);
937f1ba5 2321
f60e5990 2322DECLARE_PER_CPU(int, xmit_recursion);
2323static inline int dev_recursion_level(void)
2324{
2325 return this_cpu_read(xmit_recursion);
2326}
2327
f629d208
JP
2328struct net_device *dev_get_by_index(struct net *net, int ifindex);
2329struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2330struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2331int netdev_get_name(struct net *net, char *name, int ifindex);
2332int dev_restart(struct net_device *dev);
f629d208 2333int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2334
2335static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2336{
2337 return NAPI_GRO_CB(skb)->data_offset;
2338}
2339
2340static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2341{
2342 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2343}
2344
2345static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2346{
2347 NAPI_GRO_CB(skb)->data_offset += len;
2348}
2349
a5b1cf28
HX
2350static inline void *skb_gro_header_fast(struct sk_buff *skb,
2351 unsigned int offset)
86911732 2352{
a5b1cf28
HX
2353 return NAPI_GRO_CB(skb)->frag0 + offset;
2354}
78a478d0 2355
a5b1cf28
HX
2356static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2357{
2358 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2359}
78a478d0 2360
a5b1cf28
HX
2361static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2362 unsigned int offset)
2363{
17dd759c
HX
2364 if (!pskb_may_pull(skb, hlen))
2365 return NULL;
2366
a5b1cf28
HX
2367 NAPI_GRO_CB(skb)->frag0 = NULL;
2368 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2369 return skb->data + offset;
86911732 2370}
1da177e4 2371
36e7b1b8
HX
2372static inline void *skb_gro_network_header(struct sk_buff *skb)
2373{
78d3fd0b
HX
2374 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2375 skb_network_offset(skb);
36e7b1b8
HX
2376}
2377
bf5a755f
JC
2378static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2379 const void *start, unsigned int len)
2380{
573e8fca 2381 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2382 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2383 csum_partial(start, len, 0));
2384}
2385
573e8fca
TH
2386/* GRO checksum functions. These are logical equivalents of the normal
2387 * checksum functions (in skbuff.h) except that they operate on the GRO
2388 * offsets and fields in sk_buff.
2389 */
2390
2391__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2392
15e2396d
TH
2393static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2394{
b7fe10e5 2395 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2396}
2397
573e8fca
TH
2398static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2399 bool zero_okay,
2400 __sum16 check)
2401{
6edec0e6
TH
2402 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2403 skb_checksum_start_offset(skb) <
2404 skb_gro_offset(skb)) &&
15e2396d 2405 !skb_at_gro_remcsum_start(skb) &&
662880f4 2406 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2407 (!zero_okay || check));
2408}
2409
2410static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2411 __wsum psum)
2412{
2413 if (NAPI_GRO_CB(skb)->csum_valid &&
2414 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2415 return 0;
2416
2417 NAPI_GRO_CB(skb)->csum = psum;
2418
2419 return __skb_gro_checksum_complete(skb);
2420}
2421
573e8fca
TH
2422static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2423{
662880f4
TH
2424 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2425 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2426 NAPI_GRO_CB(skb)->csum_cnt--;
2427 } else {
2428 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2429 * verified a new top level checksum or an encapsulated one
2430 * during GRO. This saves work if we fallback to normal path.
2431 */
2432 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2433 }
2434}
2435
2436#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2437 compute_pseudo) \
2438({ \
2439 __sum16 __ret = 0; \
2440 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2441 __ret = __skb_gro_checksum_validate_complete(skb, \
2442 compute_pseudo(skb, proto)); \
5a212329
TH
2443 if (__ret) \
2444 __skb_mark_checksum_bad(skb); \
2445 else \
573e8fca
TH
2446 skb_gro_incr_csum_unnecessary(skb); \
2447 __ret; \
2448})
2449
2450#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2451 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2452
2453#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2454 compute_pseudo) \
2455 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2456
2457#define skb_gro_checksum_simple_validate(skb) \
2458 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2459
d96535a1
TH
2460static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2461{
2462 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2463 !NAPI_GRO_CB(skb)->csum_valid);
2464}
2465
2466static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2467 __sum16 check, __wsum pseudo)
2468{
2469 NAPI_GRO_CB(skb)->csum = ~pseudo;
2470 NAPI_GRO_CB(skb)->csum_valid = 1;
2471}
2472
2473#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2474do { \
2475 if (__skb_gro_checksum_convert_check(skb)) \
2476 __skb_gro_checksum_convert(skb, check, \
2477 compute_pseudo(skb, proto)); \
2478} while (0)
2479
26c4f7da
TH
2480struct gro_remcsum {
2481 int offset;
2482 __wsum delta;
2483};
2484
2485static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2486{
846cd667 2487 grc->offset = 0;
26c4f7da
TH
2488 grc->delta = 0;
2489}
2490
b7fe10e5
TH
2491static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2492 unsigned int off, size_t hdrlen,
2493 int start, int offset,
2494 struct gro_remcsum *grc,
2495 bool nopartial)
dcdc8994
TH
2496{
2497 __wsum delta;
b7fe10e5 2498 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2499
2500 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2501
15e2396d 2502 if (!nopartial) {
b7fe10e5
TH
2503 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2504 return ptr;
2505 }
2506
2507 ptr = skb_gro_header_fast(skb, off);
2508 if (skb_gro_header_hard(skb, off + plen)) {
2509 ptr = skb_gro_header_slow(skb, off + plen, off);
2510 if (!ptr)
2511 return NULL;
15e2396d
TH
2512 }
2513
b7fe10e5
TH
2514 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2515 start, offset);
dcdc8994
TH
2516
2517 /* Adjust skb->csum since we changed the packet */
dcdc8994 2518 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2519
b7fe10e5 2520 grc->offset = off + hdrlen + offset;
26c4f7da 2521 grc->delta = delta;
b7fe10e5
TH
2522
2523 return ptr;
dcdc8994
TH
2524}
2525
26c4f7da
TH
2526static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2527 struct gro_remcsum *grc)
2528{
b7fe10e5
TH
2529 void *ptr;
2530 size_t plen = grc->offset + sizeof(u16);
2531
26c4f7da
TH
2532 if (!grc->delta)
2533 return;
2534
b7fe10e5
TH
2535 ptr = skb_gro_header_fast(skb, grc->offset);
2536 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2537 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2538 if (!ptr)
2539 return;
2540 }
2541
2542 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2543}
dcdc8994 2544
6ae23ad3
TH
2545struct skb_csum_offl_spec {
2546 __u16 ipv4_okay:1,
2547 ipv6_okay:1,
2548 encap_okay:1,
2549 ip_options_okay:1,
2550 ext_hdrs_okay:1,
2551 tcp_okay:1,
2552 udp_okay:1,
2553 sctp_okay:1,
2554 vlan_okay:1,
2555 no_encapped_ipv6:1,
2556 no_not_encapped:1;
2557};
2558
2559bool __skb_csum_offload_chk(struct sk_buff *skb,
2560 const struct skb_csum_offl_spec *spec,
2561 bool *csum_encapped,
2562 bool csum_help);
2563
2564static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2565 const struct skb_csum_offl_spec *spec,
2566 bool *csum_encapped,
2567 bool csum_help)
2568{
2569 if (skb->ip_summed != CHECKSUM_PARTIAL)
2570 return false;
2571
2572 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2573}
2574
2575static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2576 const struct skb_csum_offl_spec *spec)
2577{
2578 bool csum_encapped;
2579
2580 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2581}
2582
2583static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2584{
2585 static const struct skb_csum_offl_spec csum_offl_spec = {
2586 .ipv4_okay = 1,
2587 .ip_options_okay = 1,
2588 .ipv6_okay = 1,
2589 .vlan_okay = 1,
2590 .tcp_okay = 1,
2591 .udp_okay = 1,
2592 };
2593
2594 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2595}
2596
2597static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2598{
2599 static const struct skb_csum_offl_spec csum_offl_spec = {
2600 .ipv4_okay = 1,
2601 .ip_options_okay = 1,
2602 .tcp_okay = 1,
2603 .udp_okay = 1,
2604 .vlan_okay = 1,
2605 };
2606
2607 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2608}
2609
0c4e8581
SH
2610static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2611 unsigned short type,
3b04ddde 2612 const void *daddr, const void *saddr,
95c96174 2613 unsigned int len)
0c4e8581 2614{
f1ecfd5d 2615 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2616 return 0;
3b04ddde
SH
2617
2618 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2619}
2620
b95cce35
SH
2621static inline int dev_parse_header(const struct sk_buff *skb,
2622 unsigned char *haddr)
2623{
2624 const struct net_device *dev = skb->dev;
2625
1b83336b 2626 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2627 return 0;
3b04ddde 2628 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2629}
2630
1da177e4 2631typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2632int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2633static inline int unregister_gifconf(unsigned int family)
2634{
2635 return register_gifconf(family, NULL);
2636}
2637
99bbc707 2638#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2639#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2640struct sd_flow_limit {
2641 u64 count;
2642 unsigned int num_buckets;
2643 unsigned int history_head;
2644 u16 history[FLOW_LIMIT_HISTORY];
2645 u8 buckets[];
2646};
2647
2648extern int netdev_flow_limit_table_len;
2649#endif /* CONFIG_NET_FLOW_LIMIT */
2650
1da177e4 2651/*
88751275 2652 * Incoming packets are placed on per-cpu queues
1da177e4 2653 */
d94d9fee 2654struct softnet_data {
1da177e4 2655 struct list_head poll_list;
6e7676c1 2656 struct sk_buff_head process_queue;
1da177e4 2657
dee42870 2658 /* stats */
cd7b5396
DM
2659 unsigned int processed;
2660 unsigned int time_squeeze;
2661 unsigned int cpu_collision;
2662 unsigned int received_rps;
fd793d89 2663#ifdef CONFIG_RPS
88751275 2664 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2665#endif
2666#ifdef CONFIG_NET_FLOW_LIMIT
2667 struct sd_flow_limit __rcu *flow_limit;
2668#endif
2669 struct Qdisc *output_queue;
2670 struct Qdisc **output_queue_tailp;
2671 struct sk_buff *completion_queue;
88751275 2672
4cdb1e2e 2673#ifdef CONFIG_RPS
88751275 2674 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 2675 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2676 struct softnet_data *rps_ipi_next;
2677 unsigned int cpu;
fec5e652 2678 unsigned int input_queue_head;
76cc8b13 2679 unsigned int input_queue_tail;
1e94d72f 2680#endif
95c96174 2681 unsigned int dropped;
0a9627f2 2682 struct sk_buff_head input_pkt_queue;
bea3348e 2683 struct napi_struct backlog;
99bbc707 2684
1da177e4
LT
2685};
2686
76cc8b13 2687static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2688{
2689#ifdef CONFIG_RPS
76cc8b13
TH
2690 sd->input_queue_head++;
2691#endif
2692}
2693
2694static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2695 unsigned int *qtail)
2696{
2697#ifdef CONFIG_RPS
2698 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2699#endif
2700}
2701
0a9627f2 2702DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2703
f629d208 2704void __netif_schedule(struct Qdisc *q);
46e5da40 2705void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2706
fd2ea0a7
DM
2707static inline void netif_tx_schedule_all(struct net_device *dev)
2708{
2709 unsigned int i;
2710
2711 for (i = 0; i < dev->num_tx_queues; i++)
2712 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2713}
2714
d29f749e
DJ
2715static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2716{
73466498 2717 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2718}
2719
bea3348e
SH
2720/**
2721 * netif_start_queue - allow transmit
2722 * @dev: network device
2723 *
2724 * Allow upper layers to call the device hard_start_xmit routine.
2725 */
1da177e4
LT
2726static inline void netif_start_queue(struct net_device *dev)
2727{
e8a0464c 2728 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2729}
2730
fd2ea0a7
DM
2731static inline void netif_tx_start_all_queues(struct net_device *dev)
2732{
2733 unsigned int i;
2734
2735 for (i = 0; i < dev->num_tx_queues; i++) {
2736 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2737 netif_tx_start_queue(txq);
2738 }
2739}
2740
46e5da40 2741void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2742
d29f749e
DJ
2743/**
2744 * netif_wake_queue - restart transmit
2745 * @dev: network device
2746 *
2747 * Allow upper layers to call the device hard_start_xmit routine.
2748 * Used for flow control when transmit resources are available.
2749 */
79d16385
DM
2750static inline void netif_wake_queue(struct net_device *dev)
2751{
e8a0464c 2752 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2753}
2754
fd2ea0a7
DM
2755static inline void netif_tx_wake_all_queues(struct net_device *dev)
2756{
2757 unsigned int i;
2758
2759 for (i = 0; i < dev->num_tx_queues; i++) {
2760 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2761 netif_tx_wake_queue(txq);
2762 }
2763}
2764
d29f749e
DJ
2765static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2766{
73466498 2767 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2768}
2769
bea3348e
SH
2770/**
2771 * netif_stop_queue - stop transmitted packets
2772 * @dev: network device
2773 *
2774 * Stop upper layers calling the device hard_start_xmit routine.
2775 * Used for flow control when transmit resources are unavailable.
2776 */
1da177e4
LT
2777static inline void netif_stop_queue(struct net_device *dev)
2778{
e8a0464c 2779 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2780}
2781
a2029240 2782void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2783
4d29515f 2784static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2785{
73466498 2786 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2787}
2788
bea3348e
SH
2789/**
2790 * netif_queue_stopped - test if transmit queue is flowblocked
2791 * @dev: network device
2792 *
2793 * Test if transmit queue on device is currently unable to send.
2794 */
4d29515f 2795static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2796{
e8a0464c 2797 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2798}
2799
4d29515f 2800static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2801{
73466498
TH
2802 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2803}
2804
8e2f1a63
DB
2805static inline bool
2806netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2807{
2808 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2809}
2810
8e2f1a63
DB
2811static inline bool
2812netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2813{
2814 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2815}
2816
53511453
ED
2817/**
2818 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2819 * @dev_queue: pointer to transmit queue
2820 *
2821 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2822 * to give appropriate hint to the cpu.
2823 */
2824static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2825{
2826#ifdef CONFIG_BQL
2827 prefetchw(&dev_queue->dql.num_queued);
2828#endif
2829}
2830
2831/**
2832 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2833 * @dev_queue: pointer to transmit queue
2834 *
2835 * BQL enabled drivers might use this helper in their TX completion path,
2836 * to give appropriate hint to the cpu.
2837 */
2838static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2839{
2840#ifdef CONFIG_BQL
2841 prefetchw(&dev_queue->dql.limit);
2842#endif
2843}
2844
c5d67bd7
TH
2845static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2846 unsigned int bytes)
2847{
114cf580
TH
2848#ifdef CONFIG_BQL
2849 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2850
2851 if (likely(dql_avail(&dev_queue->dql) >= 0))
2852 return;
2853
2854 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2855
2856 /*
2857 * The XOFF flag must be set before checking the dql_avail below,
2858 * because in netdev_tx_completed_queue we update the dql_completed
2859 * before checking the XOFF flag.
2860 */
2861 smp_mb();
2862
2863 /* check again in case another CPU has just made room avail */
2864 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2865 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2866#endif
c5d67bd7
TH
2867}
2868
0042d0c8
FF
2869/**
2870 * netdev_sent_queue - report the number of bytes queued to hardware
2871 * @dev: network device
2872 * @bytes: number of bytes queued to the hardware device queue
2873 *
2874 * Report the number of bytes queued for sending/completion to the network
2875 * device hardware queue. @bytes should be a good approximation and should
2876 * exactly match netdev_completed_queue() @bytes
2877 */
c5d67bd7
TH
2878static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2879{
2880 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2881}
2882
2883static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2884 unsigned int pkts, unsigned int bytes)
c5d67bd7 2885{
114cf580 2886#ifdef CONFIG_BQL
b37c0fbe
AD
2887 if (unlikely(!bytes))
2888 return;
2889
2890 dql_completed(&dev_queue->dql, bytes);
2891
2892 /*
2893 * Without the memory barrier there is a small possiblity that
2894 * netdev_tx_sent_queue will miss the update and cause the queue to
2895 * be stopped forever
2896 */
2897 smp_mb();
2898
2899 if (dql_avail(&dev_queue->dql) < 0)
2900 return;
2901
2902 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2903 netif_schedule_queue(dev_queue);
114cf580 2904#endif
c5d67bd7
TH
2905}
2906
0042d0c8
FF
2907/**
2908 * netdev_completed_queue - report bytes and packets completed by device
2909 * @dev: network device
2910 * @pkts: actual number of packets sent over the medium
2911 * @bytes: actual number of bytes sent over the medium
2912 *
2913 * Report the number of bytes and packets transmitted by the network device
2914 * hardware queue over the physical medium, @bytes must exactly match the
2915 * @bytes amount passed to netdev_sent_queue()
2916 */
c5d67bd7 2917static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2918 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2919{
2920 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2921}
2922
2923static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2924{
114cf580 2925#ifdef CONFIG_BQL
5c490354 2926 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2927 dql_reset(&q->dql);
2928#endif
c5d67bd7
TH
2929}
2930
0042d0c8
FF
2931/**
2932 * netdev_reset_queue - reset the packets and bytes count of a network device
2933 * @dev_queue: network device
2934 *
2935 * Reset the bytes and packet count of a network device and clear the
2936 * software flow control OFF bit for this network device
2937 */
c5d67bd7
TH
2938static inline void netdev_reset_queue(struct net_device *dev_queue)
2939{
2940 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2941}
2942
b9507bda
DB
2943/**
2944 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2945 * @dev: network device
2946 * @queue_index: given tx queue index
2947 *
2948 * Returns 0 if given tx queue index >= number of device tx queues,
2949 * otherwise returns the originally passed tx queue index.
2950 */
2951static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2952{
2953 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2954 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2955 dev->name, queue_index,
2956 dev->real_num_tx_queues);
2957 return 0;
2958 }
2959
2960 return queue_index;
2961}
2962
bea3348e
SH
2963/**
2964 * netif_running - test if up
2965 * @dev: network device
2966 *
2967 * Test if the device has been brought up.
2968 */
4d29515f 2969static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2970{
2971 return test_bit(__LINK_STATE_START, &dev->state);
2972}
2973
f25f4e44
PWJ
2974/*
2975 * Routines to manage the subqueues on a device. We only need start
2976 * stop, and a check if it's stopped. All other device management is
2977 * done at the overall netdevice level.
2978 * Also test the device if we're multiqueue.
2979 */
bea3348e
SH
2980
2981/**
2982 * netif_start_subqueue - allow sending packets on subqueue
2983 * @dev: network device
2984 * @queue_index: sub queue index
2985 *
2986 * Start individual transmit queue of a device with multiple transmit queues.
2987 */
f25f4e44
PWJ
2988static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2989{
fd2ea0a7 2990 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2991
2992 netif_tx_start_queue(txq);
f25f4e44
PWJ
2993}
2994
bea3348e
SH
2995/**
2996 * netif_stop_subqueue - stop sending packets on subqueue
2997 * @dev: network device
2998 * @queue_index: sub queue index
2999 *
3000 * Stop individual transmit queue of a device with multiple transmit queues.
3001 */
f25f4e44
PWJ
3002static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3003{
fd2ea0a7 3004 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3005 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3006}
3007
bea3348e
SH
3008/**
3009 * netif_subqueue_stopped - test status of subqueue
3010 * @dev: network device
3011 * @queue_index: sub queue index
3012 *
3013 * Check individual transmit queue of a device with multiple transmit queues.
3014 */
4d29515f
DM
3015static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3016 u16 queue_index)
f25f4e44 3017{
fd2ea0a7 3018 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3019
3020 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3021}
3022
4d29515f
DM
3023static inline bool netif_subqueue_stopped(const struct net_device *dev,
3024 struct sk_buff *skb)
668f895a
PE
3025{
3026 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3027}
bea3348e 3028
46e5da40 3029void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 3030
537c00de 3031#ifdef CONFIG_XPS
53af53ae 3032int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3033 u16 index);
537c00de
AD
3034#else
3035static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3036 const struct cpumask *mask,
537c00de
AD
3037 u16 index)
3038{
3039 return 0;
3040}
3041#endif
3042
5605c762
JP
3043u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3044 unsigned int num_tx_queues);
3045
a3d22a68
VZ
3046/*
3047 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3048 * as a distribution range limit for the returned value.
3049 */
3050static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3051 struct sk_buff *skb)
a3d22a68
VZ
3052{
3053 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3054}
3055
bea3348e
SH
3056/**
3057 * netif_is_multiqueue - test if device has multiple transmit queues
3058 * @dev: network device
3059 *
3060 * Check if device has multiple transmit queues
bea3348e 3061 */
4d29515f 3062static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3063{
a02cec21 3064 return dev->num_tx_queues > 1;
f25f4e44 3065}
1da177e4 3066
f629d208 3067int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3068
a953be53 3069#ifdef CONFIG_SYSFS
f629d208 3070int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3071#else
3072static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3073 unsigned int rxq)
3074{
3075 return 0;
3076}
3077#endif
3078
a953be53
MD
3079#ifdef CONFIG_SYSFS
3080static inline unsigned int get_netdev_rx_queue_index(
3081 struct netdev_rx_queue *queue)
3082{
3083 struct net_device *dev = queue->dev;
3084 int index = queue - dev->_rx;
3085
3086 BUG_ON(index >= dev->num_rx_queues);
3087 return index;
3088}
3089#endif
3090
16917b87 3091#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3092int netif_get_num_default_rss_queues(void);
16917b87 3093
e6247027
ED
3094enum skb_free_reason {
3095 SKB_REASON_CONSUMED,
3096 SKB_REASON_DROPPED,
3097};
3098
3099void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3100void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3101
e6247027
ED
3102/*
3103 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3104 * interrupt context or with hardware interrupts being disabled.
3105 * (in_irq() || irqs_disabled())
3106 *
3107 * We provide four helpers that can be used in following contexts :
3108 *
3109 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3110 * replacing kfree_skb(skb)
3111 *
3112 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3113 * Typically used in place of consume_skb(skb) in TX completion path
3114 *
3115 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3116 * replacing kfree_skb(skb)
3117 *
3118 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3119 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3120 */
e6247027
ED
3121static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3122{
3123 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3124}
3125
3126static inline void dev_consume_skb_irq(struct sk_buff *skb)
3127{
3128 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3129}
3130
3131static inline void dev_kfree_skb_any(struct sk_buff *skb)
3132{
3133 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3134}
3135
3136static inline void dev_consume_skb_any(struct sk_buff *skb)
3137{
3138 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3139}
1da177e4 3140
f629d208
JP
3141int netif_rx(struct sk_buff *skb);
3142int netif_rx_ni(struct sk_buff *skb);
04eb4489 3143int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3144gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3145void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3146struct sk_buff *napi_get_frags(struct napi_struct *napi);
3147gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3148struct packet_offload *gro_find_receive_by_type(__be16 type);
3149struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3150
3151static inline void napi_free_frags(struct napi_struct *napi)
3152{
3153 kfree_skb(napi->skb);
3154 napi->skb = NULL;
3155}
3156
f629d208
JP
3157int netdev_rx_handler_register(struct net_device *dev,
3158 rx_handler_func_t *rx_handler,
3159 void *rx_handler_data);
3160void netdev_rx_handler_unregister(struct net_device *dev);
3161
3162bool dev_valid_name(const char *name);
3163int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3164int dev_ethtool(struct net *net, struct ifreq *);
3165unsigned int dev_get_flags(const struct net_device *);
3166int __dev_change_flags(struct net_device *, unsigned int flags);
3167int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3168void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3169 unsigned int gchanges);
f629d208
JP
3170int dev_change_name(struct net_device *, const char *);
3171int dev_set_alias(struct net_device *, const char *, size_t);
3172int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3173int dev_set_mtu(struct net_device *, int);
3174void dev_set_group(struct net_device *, int);
3175int dev_set_mac_address(struct net_device *, struct sockaddr *);
3176int dev_change_carrier(struct net_device *, bool new_carrier);
3177int dev_get_phys_port_id(struct net_device *dev,
02637fce 3178 struct netdev_phys_item_id *ppid);
db24a904
DA
3179int dev_get_phys_port_name(struct net_device *dev,
3180 char *name, size_t len);
d746d707 3181int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3182struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3183struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3184 struct netdev_queue *txq, int *ret);
a0265d28 3185int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3186int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1ee481fb 3187bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
1da177e4 3188
20380731 3189extern int netdev_budget;
1da177e4
LT
3190
3191/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3192void netdev_run_todo(void);
1da177e4 3193
bea3348e
SH
3194/**
3195 * dev_put - release reference to device
3196 * @dev: network device
3197 *
9ef4429b 3198 * Release reference to device to allow it to be freed.
bea3348e 3199 */
1da177e4
LT
3200static inline void dev_put(struct net_device *dev)
3201{
933393f5 3202 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3203}
3204
bea3348e
SH
3205/**
3206 * dev_hold - get reference to device
3207 * @dev: network device
3208 *
9ef4429b 3209 * Hold reference to device to keep it from being freed.
bea3348e 3210 */
15333061
SH
3211static inline void dev_hold(struct net_device *dev)
3212{
933393f5 3213 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3214}
1da177e4
LT
3215
3216/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3217 * and _off may be called from IRQ context, but it is caller
3218 * who is responsible for serialization of these calls.
b00055aa
SR
3219 *
3220 * The name carrier is inappropriate, these functions should really be
3221 * called netif_lowerlayer_*() because they represent the state of any
3222 * kind of lower layer not just hardware media.
1da177e4
LT
3223 */
3224
f629d208
JP
3225void linkwatch_init_dev(struct net_device *dev);
3226void linkwatch_fire_event(struct net_device *dev);
3227void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3228
bea3348e
SH
3229/**
3230 * netif_carrier_ok - test if carrier present
3231 * @dev: network device
3232 *
3233 * Check if carrier is present on device
3234 */
4d29515f 3235static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3236{
3237 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3238}
3239
f629d208 3240unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3241
f629d208 3242void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3243
f629d208 3244void netif_carrier_on(struct net_device *dev);
1da177e4 3245
f629d208 3246void netif_carrier_off(struct net_device *dev);
1da177e4 3247
bea3348e
SH
3248/**
3249 * netif_dormant_on - mark device as dormant.
3250 * @dev: network device
3251 *
3252 * Mark device as dormant (as per RFC2863).
3253 *
3254 * The dormant state indicates that the relevant interface is not
3255 * actually in a condition to pass packets (i.e., it is not 'up') but is
3256 * in a "pending" state, waiting for some external event. For "on-
3257 * demand" interfaces, this new state identifies the situation where the
3258 * interface is waiting for events to place it in the up state.
3259 *
3260 */
b00055aa
SR
3261static inline void netif_dormant_on(struct net_device *dev)
3262{
3263 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3264 linkwatch_fire_event(dev);
3265}
3266
bea3348e
SH
3267/**
3268 * netif_dormant_off - set device as not dormant.
3269 * @dev: network device
3270 *
3271 * Device is not in dormant state.
3272 */
b00055aa
SR
3273static inline void netif_dormant_off(struct net_device *dev)
3274{
3275 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3276 linkwatch_fire_event(dev);
3277}
3278
bea3348e
SH
3279/**
3280 * netif_dormant - test if carrier present
3281 * @dev: network device
3282 *
3283 * Check if carrier is present on device
3284 */
4d29515f 3285static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3286{
3287 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3288}
3289
3290
bea3348e
SH
3291/**
3292 * netif_oper_up - test if device is operational
3293 * @dev: network device
3294 *
3295 * Check if carrier is operational
3296 */
4d29515f 3297static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3298{
b00055aa
SR
3299 return (dev->operstate == IF_OPER_UP ||
3300 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3301}
3302
bea3348e
SH
3303/**
3304 * netif_device_present - is device available or removed
3305 * @dev: network device
3306 *
3307 * Check if device has not been removed from system.
3308 */
4d29515f 3309static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3310{
3311 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3312}
3313
f629d208 3314void netif_device_detach(struct net_device *dev);
1da177e4 3315
f629d208 3316void netif_device_attach(struct net_device *dev);
1da177e4
LT
3317
3318/*
3319 * Network interface message level settings
3320 */
1da177e4
LT
3321
3322enum {
3323 NETIF_MSG_DRV = 0x0001,
3324 NETIF_MSG_PROBE = 0x0002,
3325 NETIF_MSG_LINK = 0x0004,
3326 NETIF_MSG_TIMER = 0x0008,
3327 NETIF_MSG_IFDOWN = 0x0010,
3328 NETIF_MSG_IFUP = 0x0020,
3329 NETIF_MSG_RX_ERR = 0x0040,
3330 NETIF_MSG_TX_ERR = 0x0080,
3331 NETIF_MSG_TX_QUEUED = 0x0100,
3332 NETIF_MSG_INTR = 0x0200,
3333 NETIF_MSG_TX_DONE = 0x0400,
3334 NETIF_MSG_RX_STATUS = 0x0800,
3335 NETIF_MSG_PKTDATA = 0x1000,
3336 NETIF_MSG_HW = 0x2000,
3337 NETIF_MSG_WOL = 0x4000,
3338};
3339
3340#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3341#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3342#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3343#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3344#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3345#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3346#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3347#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3348#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3349#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3350#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3351#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3352#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3353#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3354#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3355
3356static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3357{
3358 /* use default */
3359 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3360 return default_msg_enable_bits;
3361 if (debug_value == 0) /* no output */
3362 return 0;
3363 /* set low N bits */
3364 return (1 << debug_value) - 1;
3365}
3366
c773e847 3367static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3368{
c773e847
DM
3369 spin_lock(&txq->_xmit_lock);
3370 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3371}
3372
fd2ea0a7
DM
3373static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3374{
3375 spin_lock_bh(&txq->_xmit_lock);
3376 txq->xmit_lock_owner = smp_processor_id();
3377}
3378
4d29515f 3379static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3380{
4d29515f 3381 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3382 if (likely(ok))
3383 txq->xmit_lock_owner = smp_processor_id();
3384 return ok;
3385}
3386
3387static inline void __netif_tx_unlock(struct netdev_queue *txq)
3388{
3389 txq->xmit_lock_owner = -1;
3390 spin_unlock(&txq->_xmit_lock);
3391}
3392
3393static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3394{
3395 txq->xmit_lock_owner = -1;
3396 spin_unlock_bh(&txq->_xmit_lock);
3397}
3398
08baf561
ED
3399static inline void txq_trans_update(struct netdev_queue *txq)
3400{
3401 if (txq->xmit_lock_owner != -1)
3402 txq->trans_start = jiffies;
3403}
3404
d29f749e
DJ
3405/**
3406 * netif_tx_lock - grab network device transmit lock
3407 * @dev: network device
d29f749e
DJ
3408 *
3409 * Get network device transmit lock
3410 */
22dd7495
JHS
3411static inline void netif_tx_lock(struct net_device *dev)
3412{
e8a0464c 3413 unsigned int i;
c3f26a26 3414 int cpu;
c773e847 3415
c3f26a26
DM
3416 spin_lock(&dev->tx_global_lock);
3417 cpu = smp_processor_id();
e8a0464c
DM
3418 for (i = 0; i < dev->num_tx_queues; i++) {
3419 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3420
3421 /* We are the only thread of execution doing a
3422 * freeze, but we have to grab the _xmit_lock in
3423 * order to synchronize with threads which are in
3424 * the ->hard_start_xmit() handler and already
3425 * checked the frozen bit.
3426 */
e8a0464c 3427 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3428 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3429 __netif_tx_unlock(txq);
e8a0464c 3430 }
932ff279
HX
3431}
3432
3433static inline void netif_tx_lock_bh(struct net_device *dev)
3434{
e8a0464c
DM
3435 local_bh_disable();
3436 netif_tx_lock(dev);
932ff279
HX
3437}
3438
932ff279
HX
3439static inline void netif_tx_unlock(struct net_device *dev)
3440{
e8a0464c
DM
3441 unsigned int i;
3442
3443 for (i = 0; i < dev->num_tx_queues; i++) {
3444 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3445
c3f26a26
DM
3446 /* No need to grab the _xmit_lock here. If the
3447 * queue is not stopped for another reason, we
3448 * force a schedule.
3449 */
3450 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3451 netif_schedule_queue(txq);
c3f26a26
DM
3452 }
3453 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3454}
3455
3456static inline void netif_tx_unlock_bh(struct net_device *dev)
3457{
e8a0464c
DM
3458 netif_tx_unlock(dev);
3459 local_bh_enable();
932ff279
HX
3460}
3461
c773e847 3462#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3463 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3464 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3465 } \
3466}
3467
5efeac44
EB
3468#define HARD_TX_TRYLOCK(dev, txq) \
3469 (((dev->features & NETIF_F_LLTX) == 0) ? \
3470 __netif_tx_trylock(txq) : \
3471 true )
3472
c773e847 3473#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3474 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3475 __netif_tx_unlock(txq); \
22dd7495
JHS
3476 } \
3477}
3478
1da177e4
LT
3479static inline void netif_tx_disable(struct net_device *dev)
3480{
fd2ea0a7 3481 unsigned int i;
c3f26a26 3482 int cpu;
fd2ea0a7 3483
c3f26a26
DM
3484 local_bh_disable();
3485 cpu = smp_processor_id();
fd2ea0a7
DM
3486 for (i = 0; i < dev->num_tx_queues; i++) {
3487 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3488
3489 __netif_tx_lock(txq, cpu);
fd2ea0a7 3490 netif_tx_stop_queue(txq);
c3f26a26 3491 __netif_tx_unlock(txq);
fd2ea0a7 3492 }
c3f26a26 3493 local_bh_enable();
1da177e4
LT
3494}
3495
e308a5d8
DM
3496static inline void netif_addr_lock(struct net_device *dev)
3497{
3498 spin_lock(&dev->addr_list_lock);
3499}
3500
2429f7ac
JP
3501static inline void netif_addr_lock_nested(struct net_device *dev)
3502{
25175ba5
VY
3503 int subclass = SINGLE_DEPTH_NESTING;
3504
3505 if (dev->netdev_ops->ndo_get_lock_subclass)
3506 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3507
3508 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3509}
3510
e308a5d8
DM
3511static inline void netif_addr_lock_bh(struct net_device *dev)
3512{
3513 spin_lock_bh(&dev->addr_list_lock);
3514}
3515
3516static inline void netif_addr_unlock(struct net_device *dev)
3517{
3518 spin_unlock(&dev->addr_list_lock);
3519}
3520
3521static inline void netif_addr_unlock_bh(struct net_device *dev)
3522{
3523 spin_unlock_bh(&dev->addr_list_lock);
3524}
3525
f001fde5 3526/*
31278e71 3527 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3528 * rcu_read_lock held.
3529 */
3530#define for_each_dev_addr(dev, ha) \
31278e71 3531 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3532
1da177e4
LT
3533/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3534
f629d208 3535void ether_setup(struct net_device *dev);
1da177e4
LT
3536
3537/* Support for loadable net-drivers */
f629d208 3538struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3539 unsigned char name_assign_type,
f629d208
JP
3540 void (*setup)(struct net_device *),
3541 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3542#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3543 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3544
c835a677
TG
3545#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3546 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3547 count)
36909ea4 3548
f629d208
JP
3549int register_netdev(struct net_device *dev);
3550void unregister_netdev(struct net_device *dev);
f001fde5 3551
22bedad3 3552/* General hardware address lists handling functions */
f629d208
JP
3553int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3554 struct netdev_hw_addr_list *from_list, int addr_len);
3555void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3556 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3557int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3558 struct net_device *dev,
3559 int (*sync)(struct net_device *, const unsigned char *),
3560 int (*unsync)(struct net_device *,
3561 const unsigned char *));
3562void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3563 struct net_device *dev,
3564 int (*unsync)(struct net_device *,
3565 const unsigned char *));
f629d208 3566void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3567
f001fde5 3568/* Functions used for device addresses handling */
f629d208
JP
3569int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3570 unsigned char addr_type);
3571int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3572 unsigned char addr_type);
f629d208
JP
3573void dev_addr_flush(struct net_device *dev);
3574int dev_addr_init(struct net_device *dev);
a748ee24
JP
3575
3576/* Functions used for unicast addresses handling */
f629d208
JP
3577int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3578int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3579int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3580int dev_uc_sync(struct net_device *to, struct net_device *from);
3581int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3582void dev_uc_unsync(struct net_device *to, struct net_device *from);
3583void dev_uc_flush(struct net_device *dev);
3584void dev_uc_init(struct net_device *dev);
f001fde5 3585
670e5b8e
AD
3586/**
3587 * __dev_uc_sync - Synchonize device's unicast list
3588 * @dev: device to sync
3589 * @sync: function to call if address should be added
3590 * @unsync: function to call if address should be removed
3591 *
3592 * Add newly added addresses to the interface, and release
3593 * addresses that have been deleted.
3594 **/
3595static inline int __dev_uc_sync(struct net_device *dev,
3596 int (*sync)(struct net_device *,
3597 const unsigned char *),
3598 int (*unsync)(struct net_device *,
3599 const unsigned char *))
3600{
3601 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3602}
3603
3604/**
e793c0f7 3605 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3606 * @dev: device to sync
3607 * @unsync: function to call if address should be removed
3608 *
3609 * Remove all addresses that were added to the device by dev_uc_sync().
3610 **/
3611static inline void __dev_uc_unsync(struct net_device *dev,
3612 int (*unsync)(struct net_device *,
3613 const unsigned char *))
3614{
3615 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3616}
3617
22bedad3 3618/* Functions used for multicast addresses handling */
f629d208
JP
3619int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3620int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3621int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3622int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3623int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3624int dev_mc_sync(struct net_device *to, struct net_device *from);
3625int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3626void dev_mc_unsync(struct net_device *to, struct net_device *from);
3627void dev_mc_flush(struct net_device *dev);
3628void dev_mc_init(struct net_device *dev);
f001fde5 3629
670e5b8e
AD
3630/**
3631 * __dev_mc_sync - Synchonize device's multicast list
3632 * @dev: device to sync
3633 * @sync: function to call if address should be added
3634 * @unsync: function to call if address should be removed
3635 *
3636 * Add newly added addresses to the interface, and release
3637 * addresses that have been deleted.
3638 **/
3639static inline int __dev_mc_sync(struct net_device *dev,
3640 int (*sync)(struct net_device *,
3641 const unsigned char *),
3642 int (*unsync)(struct net_device *,
3643 const unsigned char *))
3644{
3645 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3646}
3647
3648/**
e793c0f7 3649 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3650 * @dev: device to sync
3651 * @unsync: function to call if address should be removed
3652 *
3653 * Remove all addresses that were added to the device by dev_mc_sync().
3654 **/
3655static inline void __dev_mc_unsync(struct net_device *dev,
3656 int (*unsync)(struct net_device *,
3657 const unsigned char *))
3658{
3659 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3660}
3661
4417da66 3662/* Functions used for secondary unicast and multicast support */
f629d208
JP
3663void dev_set_rx_mode(struct net_device *dev);
3664void __dev_set_rx_mode(struct net_device *dev);
3665int dev_set_promiscuity(struct net_device *dev, int inc);
3666int dev_set_allmulti(struct net_device *dev, int inc);
3667void netdev_state_change(struct net_device *dev);
3668void netdev_notify_peers(struct net_device *dev);
3669void netdev_features_change(struct net_device *dev);
1da177e4 3670/* Load a device via the kmod */
f629d208
JP
3671void dev_load(struct net *net, const char *name);
3672struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3673 struct rtnl_link_stats64 *storage);
3674void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3675 const struct net_device_stats *netdev_stats);
eeda3fd6 3676
1da177e4 3677extern int netdev_max_backlog;
3b098e2d 3678extern int netdev_tstamp_prequeue;
1da177e4 3679extern int weight_p;
0a14842f 3680extern int bpf_jit_enable;
9ff162a8 3681
f629d208 3682bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3683struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3684 struct list_head **iter);
f629d208
JP
3685struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3686 struct list_head **iter);
8b5be856 3687
44a40855
VY
3688/* iterate through upper list, must be called under RCU read lock */
3689#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3690 for (iter = &(dev)->adj_list.upper, \
3691 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3692 updev; \
3693 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3694
8b5be856 3695/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3696#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3697 for (iter = &(dev)->all_adj_list.upper, \
3698 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3699 updev; \
3700 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3701
f629d208
JP
3702void *netdev_lower_get_next_private(struct net_device *dev,
3703 struct list_head **iter);
3704void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3705 struct list_head **iter);
31088a11
VF
3706
3707#define netdev_for_each_lower_private(dev, priv, iter) \
3708 for (iter = (dev)->adj_list.lower.next, \
3709 priv = netdev_lower_get_next_private(dev, &(iter)); \
3710 priv; \
3711 priv = netdev_lower_get_next_private(dev, &(iter)))
3712
3713#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3714 for (iter = &(dev)->adj_list.lower, \
3715 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3716 priv; \
3717 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3718
4085ebe8
VY
3719void *netdev_lower_get_next(struct net_device *dev,
3720 struct list_head **iter);
3721#define netdev_for_each_lower_dev(dev, ldev, iter) \
3722 for (iter = &(dev)->adj_list.lower, \
3723 ldev = netdev_lower_get_next(dev, &(iter)); \
3724 ldev; \
3725 ldev = netdev_lower_get_next(dev, &(iter)))
3726
f629d208 3727void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3728void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3729struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3730struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3731int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3732int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3733 struct net_device *upper_dev,
29bf24af 3734 void *upper_priv, void *upper_info);
f629d208
JP
3735void netdev_upper_dev_unlink(struct net_device *dev,
3736 struct net_device *upper_dev);
5bb025fa 3737void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3738void *netdev_lower_dev_get_private(struct net_device *dev,
3739 struct net_device *lower_dev);
04d48266
JP
3740void netdev_lower_state_changed(struct net_device *lower_dev,
3741 void *lower_state_info);
960fb622
ED
3742
3743/* RSS keys are 40 or 52 bytes long */
3744#define NETDEV_RSS_KEY_LEN 52
3745extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3746void netdev_rss_key_fill(void *buffer, size_t len);
3747
4085ebe8 3748int dev_get_nest_level(struct net_device *dev,
b618aaa9 3749 bool (*type_check)(const struct net_device *dev));
f629d208
JP
3750int skb_checksum_help(struct sk_buff *skb);
3751struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3752 netdev_features_t features, bool tx_path);
3753struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3754 netdev_features_t features);
12b0004d 3755
61bd3857
MS
3756struct netdev_bonding_info {
3757 ifslave slave;
3758 ifbond master;
3759};
3760
3761struct netdev_notifier_bonding_info {
3762 struct netdev_notifier_info info; /* must be first */
3763 struct netdev_bonding_info bonding_info;
3764};
3765
3766void netdev_bonding_info_change(struct net_device *dev,
3767 struct netdev_bonding_info *bonding_info);
3768
12b0004d
CW
3769static inline
3770struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3771{
3772 return __skb_gso_segment(skb, features, true);
3773}
53d6471c 3774__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3775
3776static inline bool can_checksum_protocol(netdev_features_t features,
3777 __be16 protocol)
3778{
c8cd0989
TH
3779 if (protocol == htons(ETH_P_FCOE))
3780 return !!(features & NETIF_F_FCOE_CRC);
3781
3782 /* Assume this is an IP checksum (not SCTP CRC) */
3783
3784 if (features & NETIF_F_HW_CSUM) {
3785 /* Can checksum everything */
3786 return true;
3787 }
3788
3789 switch (protocol) {
3790 case htons(ETH_P_IP):
3791 return !!(features & NETIF_F_IP_CSUM);
3792 case htons(ETH_P_IPV6):
3793 return !!(features & NETIF_F_IPV6_CSUM);
3794 default:
3795 return false;
3796 }
ec5f0615 3797}
12b0004d 3798
6ae23ad3
TH
3799/* Map an ethertype into IP protocol if possible */
3800static inline int eproto_to_ipproto(int eproto)
3801{
3802 switch (eproto) {
3803 case htons(ETH_P_IP):
3804 return IPPROTO_IP;
3805 case htons(ETH_P_IPV6):
3806 return IPPROTO_IPV6;
3807 default:
3808 return -1;
3809 }
3810}
3811
fb286bb2 3812#ifdef CONFIG_BUG
f629d208 3813void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3814#else
3815static inline void netdev_rx_csum_fault(struct net_device *dev)
3816{
3817}
3818#endif
1da177e4 3819/* rx skb timestamps */
f629d208
JP
3820void net_enable_timestamp(void);
3821void net_disable_timestamp(void);
1da177e4 3822
20380731 3823#ifdef CONFIG_PROC_FS
f629d208 3824int __init dev_proc_init(void);
900ff8c6
CW
3825#else
3826#define dev_proc_init() 0
20380731
ACM
3827#endif
3828
4798248e 3829static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3830 struct sk_buff *skb, struct net_device *dev,
3831 bool more)
4798248e 3832{
fa2dbdc2 3833 skb->xmit_more = more ? 1 : 0;
0b725a2c 3834 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3835}
3836
10b3ad8c 3837static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3838 struct netdev_queue *txq, bool more)
4798248e
DM
3839{
3840 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3841 int rc;
4798248e 3842
fa2dbdc2 3843 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3844 if (rc == NETDEV_TX_OK)
3845 txq_trans_update(txq);
3846
3847 return rc;
4798248e
DM
3848}
3849
42a2d923
LT
3850int netdev_class_create_file_ns(struct class_attribute *class_attr,
3851 const void *ns);
3852void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3853 const void *ns);
58292cbe
TH
3854
3855static inline int netdev_class_create_file(struct class_attribute *class_attr)
3856{
3857 return netdev_class_create_file_ns(class_attr, NULL);
3858}
3859
3860static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3861{
3862 netdev_class_remove_file_ns(class_attr, NULL);
3863}
b8a9787e 3864
04600794
JB
3865extern struct kobj_ns_type_operations net_ns_type_operations;
3866
f629d208 3867const char *netdev_drivername(const struct net_device *dev);
6579e57b 3868
f629d208 3869void linkwatch_run_queue(void);
20380731 3870
da08143b
MK
3871static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3872 netdev_features_t f2)
3873{
c8cd0989
TH
3874 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3875 if (f1 & NETIF_F_HW_CSUM)
3876 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IP_CSUM);
3877 else
3878 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IP_CSUM);
3879 }
da08143b 3880
c8cd0989 3881 return f1 & f2;
da08143b
MK
3882}
3883
c8f44aff
MM
3884static inline netdev_features_t netdev_get_wanted_features(
3885 struct net_device *dev)
5455c699
MM
3886{
3887 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3888}
c8f44aff
MM
3889netdev_features_t netdev_increment_features(netdev_features_t all,
3890 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3891
3892/* Allow TSO being used on stacked device :
3893 * Performing the GSO segmentation before last device
3894 * is a performance improvement.
3895 */
3896static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3897 netdev_features_t mask)
3898{
3899 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3900}
3901
6cb6a27c 3902int __netdev_update_features(struct net_device *dev);
5455c699 3903void netdev_update_features(struct net_device *dev);
afe12cc8 3904void netdev_change_features(struct net_device *dev);
7f353bf2 3905
fc4a7489
PM
3906void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3907 struct net_device *dev);
3908
e38f3025
TM
3909netdev_features_t passthru_features_check(struct sk_buff *skb,
3910 struct net_device *dev,
3911 netdev_features_t features);
c1e756bf 3912netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3913
4d29515f 3914static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3915{
c8f44aff 3916 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3917
3918 /* check flags correspondence */
3919 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3920 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3921 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3922 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3923 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3924 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
3925 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3926 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3927 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3928 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3929 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3930 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
e585f236 3931 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
0345e186 3932
d6b4991a 3933 return (features & feature) == feature;
576a30eb
HX
3934}
3935
4d29515f 3936static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3937{
278b2513 3938 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3939 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3940}
3941
8b86a61d 3942static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 3943 netdev_features_t features)
7967168c 3944{
fc741216 3945 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3946 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3947 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3948}
3949
82cc1a7a
PWJ
3950static inline void netif_set_gso_max_size(struct net_device *dev,
3951 unsigned int size)
3952{
3953 dev->gso_max_size = size;
3954}
3955
7a7ffbab
WCC
3956static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3957 int pulled_hlen, u16 mac_offset,
3958 int mac_len)
3959{
3960 skb->protocol = protocol;
3961 skb->encapsulation = 1;
3962 skb_push(skb, pulled_hlen);
3963 skb_reset_transport_header(skb);
3964 skb->mac_header = mac_offset;
3965 skb->network_header = skb->mac_header + mac_len;
3966 skb->mac_len = mac_len;
3967}
3968
b618aaa9 3969static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
3970{
3971 return dev->priv_flags & IFF_MACVLAN;
3972}
3973
b618aaa9 3974static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
3975{
3976 return dev->priv_flags & IFF_MACVLAN_PORT;
3977}
3978
b618aaa9 3979static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
3980{
3981 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3982}
3983
b618aaa9 3984static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
3985{
3986 return dev->priv_flags & IFF_IPVLAN_MASTER;
3987}
3988
b618aaa9 3989static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 3990{
3991 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3992}
3993
b618aaa9 3994static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
3995{
3996 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3997}
3998
3bdc0eba
BG
3999static inline bool netif_supports_nofcs(struct net_device *dev)
4000{
4001 return dev->priv_flags & IFF_SUPP_NOFCS;
4002}
4003
007979ea 4004static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4005{
007979ea 4006 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4007}
4008
fee6d4c7
DA
4009static inline bool netif_is_l3_slave(const struct net_device *dev)
4010{
4011 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4012}
4013
0894ae3f
JP
4014static inline bool netif_is_bridge_master(const struct net_device *dev)
4015{
4016 return dev->priv_flags & IFF_EBRIDGE;
4017}
4018
28f9ee22
VY
4019static inline bool netif_is_bridge_port(const struct net_device *dev)
4020{
4021 return dev->priv_flags & IFF_BRIDGE_PORT;
4022}
4023
35d4e172
JP
4024static inline bool netif_is_ovs_master(const struct net_device *dev)
4025{
4026 return dev->priv_flags & IFF_OPENVSWITCH;
4027}
4028
b618aaa9 4029static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4030{
4031 return dev->priv_flags & IFF_TEAM;
4032}
4033
b618aaa9 4034static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4035{
4036 return dev->priv_flags & IFF_TEAM_PORT;
4037}
4038
b618aaa9 4039static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4040{
4041 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4042}
4043
b618aaa9 4044static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4045{
4046 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4047}
4048
02875878
ED
4049/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4050static inline void netif_keep_dst(struct net_device *dev)
4051{
4052 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4053}
4054
505d4f73 4055extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4056
571ba423
JP
4057/* Logging, debugging and troubleshooting/diagnostic helpers. */
4058
4059/* netdev_printk helpers, similar to dev_printk */
4060
4061static inline const char *netdev_name(const struct net_device *dev)
4062{
c6f854d5
VF
4063 if (!dev->name[0] || strchr(dev->name, '%'))
4064 return "(unnamed net_device)";
571ba423
JP
4065 return dev->name;
4066}
4067
ccc7f496
VF
4068static inline const char *netdev_reg_state(const struct net_device *dev)
4069{
4070 switch (dev->reg_state) {
4071 case NETREG_UNINITIALIZED: return " (uninitialized)";
4072 case NETREG_REGISTERED: return "";
4073 case NETREG_UNREGISTERING: return " (unregistering)";
4074 case NETREG_UNREGISTERED: return " (unregistered)";
4075 case NETREG_RELEASED: return " (released)";
4076 case NETREG_DUMMY: return " (dummy)";
4077 }
4078
4079 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4080 return " (unknown)";
4081}
4082
f629d208 4083__printf(3, 4)
6ea754eb
JP
4084void netdev_printk(const char *level, const struct net_device *dev,
4085 const char *format, ...);
f629d208 4086__printf(2, 3)
6ea754eb 4087void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4088__printf(2, 3)
6ea754eb 4089void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4090__printf(2, 3)
6ea754eb 4091void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4092__printf(2, 3)
6ea754eb 4093void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4094__printf(2, 3)
6ea754eb 4095void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4096__printf(2, 3)
6ea754eb 4097void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4098__printf(2, 3)
6ea754eb 4099void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4100
8909c9ad
VK
4101#define MODULE_ALIAS_NETDEV(device) \
4102 MODULE_ALIAS("netdev-" device)
4103
b558c96f 4104#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4105#define netdev_dbg(__dev, format, args...) \
4106do { \
ffa10cb4 4107 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4108} while (0)
b558c96f
JC
4109#elif defined(DEBUG)
4110#define netdev_dbg(__dev, format, args...) \
4111 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4112#else
4113#define netdev_dbg(__dev, format, args...) \
4114({ \
4115 if (0) \
4116 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4117})
4118#endif
4119
4120#if defined(VERBOSE_DEBUG)
4121#define netdev_vdbg netdev_dbg
4122#else
4123
4124#define netdev_vdbg(dev, format, args...) \
4125({ \
4126 if (0) \
4127 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4128 0; \
4129})
4130#endif
4131
4132/*
4133 * netdev_WARN() acts like dev_printk(), but with the key difference
4134 * of using a WARN/WARN_ON to get the message out, including the
4135 * file/line information and a backtrace.
4136 */
4137#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4138 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4139 netdev_reg_state(dev), ##args)
571ba423 4140
b3d95c5c
JP
4141/* netif printk helpers, similar to netdev_printk */
4142
4143#define netif_printk(priv, type, level, dev, fmt, args...) \
4144do { \
4145 if (netif_msg_##type(priv)) \
4146 netdev_printk(level, (dev), fmt, ##args); \
4147} while (0)
4148
f45f4321
JP
4149#define netif_level(level, priv, type, dev, fmt, args...) \
4150do { \
4151 if (netif_msg_##type(priv)) \
4152 netdev_##level(dev, fmt, ##args); \
4153} while (0)
4154
b3d95c5c 4155#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4156 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4157#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4158 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4159#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4160 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4161#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4162 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4163#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4164 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4165#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4166 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4167#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4168 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4169
0053ea9c 4170#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4171#define netif_dbg(priv, type, netdev, format, args...) \
4172do { \
4173 if (netif_msg_##type(priv)) \
b5fb0a03 4174 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4175} while (0)
0053ea9c
JP
4176#elif defined(DEBUG)
4177#define netif_dbg(priv, type, dev, format, args...) \
4178 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4179#else
4180#define netif_dbg(priv, type, dev, format, args...) \
4181({ \
4182 if (0) \
4183 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4184 0; \
4185})
4186#endif
4187
4188#if defined(VERBOSE_DEBUG)
bcfcc450 4189#define netif_vdbg netif_dbg
b3d95c5c
JP
4190#else
4191#define netif_vdbg(priv, type, dev, format, args...) \
4192({ \
4193 if (0) \
a4ed89cb 4194 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4195 0; \
4196})
4197#endif
571ba423 4198
900ff8c6
CW
4199/*
4200 * The list of packet types we will receive (as opposed to discard)
4201 * and the routines to invoke.
4202 *
4203 * Why 16. Because with 16 the only overlap we get on a hash of the
4204 * low nibble of the protocol value is RARP/SNAP/X.25.
4205 *
4206 * NOTE: That is no longer true with the addition of VLAN tags. Not
4207 * sure which should go first, but I bet it won't make much
4208 * difference if we are running VLANs. The good news is that
4209 * this protocol won't be in the list unless compiled in, so
4210 * the average user (w/out VLANs) will not be adversely affected.
4211 * --BLG
4212 *
4213 * 0800 IP
4214 * 8100 802.1Q VLAN
4215 * 0001 802.3
4216 * 0002 AX.25
4217 * 0004 802.2
4218 * 8035 RARP
4219 * 0005 SNAP
4220 * 0805 X.25
4221 * 0806 ARP
4222 * 8137 IPX
4223 * 0009 Localtalk
4224 * 86DD IPv6
4225 */
4226#define PTYPE_HASH_SIZE (16)
4227#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4228
385a154c 4229#endif /* _LINUX_NETDEVICE_H */