]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/netdevice.h
net/xen-netfront: only clean up queues if present
[mirror_ubuntu-bionic-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
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
d11ead75 135#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
136# if defined(CONFIG_MAC80211_MESH)
137# define LL_MAX_HEADER 128
138# else
139# define LL_MAX_HEADER 96
140# endif
1da177e4 141#else
8388e3da 142# define LL_MAX_HEADER 32
1da177e4
LT
143#endif
144
d11ead75
BH
145#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
147#define MAX_HEADER LL_MAX_HEADER
148#else
149#define MAX_HEADER (LL_MAX_HEADER + 48)
150#endif
151
152/*
be1f3c2c
BH
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
1da177e4 155 */
fe2918b0 156
d94d9fee 157struct net_device_stats {
3cfde79c
BH
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
1da177e4 167 unsigned long collisions;
1da177e4 168 unsigned long rx_length_errors;
3cfde79c
BH
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
1da177e4
LT
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
1da177e4
LT
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
181};
182
1da177e4
LT
183
184#include <linux/cache.h>
185#include <linux/skbuff.h>
186
adc9300e 187#ifdef CONFIG_RPS
c5905afb
IM
188#include <linux/static_key.h>
189extern struct static_key rps_needed;
adc9300e
ED
190#endif
191
1da177e4
LT
192struct neighbour;
193struct neigh_parms;
194struct sk_buff;
195
f001fde5
JP
196struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
199 unsigned char type;
ccffad25
JP
200#define NETDEV_HW_ADDR_T_LAN 1
201#define NETDEV_HW_ADDR_T_SAN 2
202#define NETDEV_HW_ADDR_T_SLAVE 3
203#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 204#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 205 bool global_use;
4cd729b0 206 int sync_cnt;
8f8f103d 207 int refcount;
4543fbef 208 int synced;
f001fde5
JP
209 struct rcu_head rcu_head;
210};
211
31278e71
JP
212struct netdev_hw_addr_list {
213 struct list_head list;
214 int count;
215};
216
22bedad3
JP
217#define netdev_hw_addr_list_count(l) ((l)->count)
218#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219#define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 221
22bedad3
JP
222#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224#define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 226
22bedad3
JP
227#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 229#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 231
d94d9fee 232struct hh_cache {
f6b72b62 233 u16 hh_len;
5c25f686 234 u16 __pad;
3644f0ce 235 seqlock_t hh_lock;
1da177e4
LT
236
237 /* cached hardware header; allow for machine alignment needs. */
238#define HH_DATA_MOD 16
239#define HH_DATA_OFF(__len) \
5ba0eac6 240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
241#define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
244};
245
246/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * Alternative is:
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 *
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
253 */
254#define LL_RESERVED_SPACE(dev) \
f5184d26 255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 256#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 258
3b04ddde
SH
259struct header_ops {
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
95c96174 262 const void *saddr, unsigned int len);
3b04ddde 263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268};
269
1da177e4
LT
270/* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
272 * code.
273 */
274
d94d9fee 275enum netdev_state_t {
1da177e4
LT
276 __LINK_STATE_START,
277 __LINK_STATE_PRESENT,
1da177e4 278 __LINK_STATE_NOCARRIER,
b00055aa
SR
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
1da177e4
LT
281};
282
283
284/*
285 * This structure holds at boot time configured netdevice settings. They
fe2918b0 286 * are then used in the device probing.
1da177e4
LT
287 */
288struct netdev_boot_setup {
289 char name[IFNAMSIZ];
290 struct ifmap map;
291};
292#define NETDEV_BOOT_SETUP_MAX 8
293
f629d208 294int __init netdev_boot_setup(char *str);
1da177e4 295
bea3348e
SH
296/*
297 * Structure for NAPI scheduling similar to tasklet but with weighting
298 */
299struct napi_struct {
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
305 */
306 struct list_head poll_list;
307
308 unsigned long state;
309 int weight;
404f7c9e 310 unsigned int gro_count;
bea3348e
SH
311 int (*poll)(struct napi_struct *, int);
312#ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
314 int poll_owner;
bea3348e 315#endif
5d38a079 316 struct net_device *dev;
d565b0a1 317 struct sk_buff *gro_list;
5d38a079 318 struct sk_buff *skb;
3b47d303 319 struct hrtimer timer;
404f7c9e 320 struct list_head dev_list;
af12fa6e
ET
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
bea3348e
SH
323};
324
d94d9fee 325enum {
bea3348e 326 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 327 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
af12fa6e 329 NAPI_STATE_HASHED, /* In NAPI hash */
bea3348e
SH
330};
331
5b252f0c 332enum gro_result {
d1c76af9
HX
333 GRO_MERGED,
334 GRO_MERGED_FREE,
335 GRO_HELD,
336 GRO_NORMAL,
337 GRO_DROP,
338};
5b252f0c 339typedef enum gro_result gro_result_t;
d1c76af9 340
8a4eb573
JP
341/*
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * further.
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 *
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
352 *
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
355 *
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
359 *
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
361 * do with the skb.
362 *
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
366 *
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
370 *
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 373 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
374 *
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
377 *
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
380 */
381
382enum rx_handler_result {
383 RX_HANDLER_CONSUMED,
384 RX_HANDLER_ANOTHER,
385 RX_HANDLER_EXACT,
386 RX_HANDLER_PASS,
387};
388typedef enum rx_handler_result rx_handler_result_t;
389typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 390
f629d208 391void __napi_schedule(struct napi_struct *n);
bc9ad166 392void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 393
4d29515f 394static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
395{
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
397}
398
bea3348e
SH
399/**
400 * napi_schedule_prep - check if napi can be scheduled
401 * @n: napi context
402 *
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
a0a46196
DM
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
bea3348e 407 */
4d29515f 408static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 409{
a0a46196
DM
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
412}
413
414/**
415 * napi_schedule - schedule NAPI poll
416 * @n: napi context
417 *
418 * Schedule NAPI poll routine to be called if it is not already
419 * running.
420 */
421static inline void napi_schedule(struct napi_struct *n)
422{
423 if (napi_schedule_prep(n))
424 __napi_schedule(n);
425}
426
bc9ad166
ED
427/**
428 * napi_schedule_irqoff - schedule NAPI poll
429 * @n: napi context
430 *
431 * Variant of napi_schedule(), assuming hard irqs are masked.
432 */
433static inline void napi_schedule_irqoff(struct napi_struct *n)
434{
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
437}
438
bfe13f54 439/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 440static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
441{
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
4d29515f 444 return true;
bfe13f54 445 }
4d29515f 446 return false;
bfe13f54
RD
447}
448
3b47d303
ED
449void __napi_complete(struct napi_struct *n);
450void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
451/**
452 * napi_complete - NAPI processing complete
453 * @n: napi context
454 *
455 * Mark NAPI processing as complete.
3b47d303 456 * Consider using napi_complete_done() instead.
bea3348e 457 */
3b47d303
ED
458static inline void napi_complete(struct napi_struct *n)
459{
460 return napi_complete_done(n, 0);
461}
bea3348e 462
af12fa6e
ET
463/**
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
466 *
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
469 */
f629d208 470struct napi_struct *napi_by_id(unsigned int napi_id);
af12fa6e
ET
471
472/**
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
475 *
476 * generate a new napi_id and store a @napi under it in napi_hash
477 */
f629d208 478void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
479
480/**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
483 *
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
486 */
f629d208 487void napi_hash_del(struct napi_struct *napi);
af12fa6e 488
bea3348e
SH
489/**
490 * napi_disable - prevent NAPI from scheduling
491 * @n: napi context
492 *
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
495 */
3b47d303 496void napi_disable(struct napi_struct *n);
bea3348e
SH
497
498/**
499 * napi_enable - enable NAPI scheduling
500 * @n: napi context
501 *
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
504 */
505static inline void napi_enable(struct napi_struct *n)
506{
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 508 smp_mb__before_atomic();
bea3348e
SH
509 clear_bit(NAPI_STATE_SCHED, &n->state);
510}
511
c264c3de
SH
512#ifdef CONFIG_SMP
513/**
514 * napi_synchronize - wait until NAPI is not running
515 * @n: napi context
516 *
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
520 */
521static inline void napi_synchronize(const struct napi_struct *n)
522{
523 while (test_bit(NAPI_STATE_SCHED, &n->state))
524 msleep(1);
525}
526#else
527# define napi_synchronize(n) barrier()
528#endif
529
d94d9fee 530enum netdev_queue_state_t {
73466498
TH
531 __QUEUE_STATE_DRV_XOFF,
532 __QUEUE_STATE_STACK_XOFF,
c3f26a26 533 __QUEUE_STATE_FROZEN,
79d16385 534};
8e2f1a63
DB
535
536#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
537#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
538#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
539
540#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
541#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
542 QUEUE_STATE_FROZEN)
543#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
544 QUEUE_STATE_FROZEN)
545
73466498
TH
546/*
547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
548 * netif_tx_* functions below are used to manipulate this flag. The
549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
550 * queue independently. The netif_xmit_*stopped functions below are called
551 * to check if the queue has been stopped by the driver or stack (either
552 * of the XOFF bits are set in the state). Drivers should not need to call
553 * netif_xmit*stopped functions, they should only be using netif_tx_*.
554 */
79d16385 555
bb949fbd 556struct netdev_queue {
6a321cb3
ED
557/*
558 * read mostly part
559 */
bb949fbd 560 struct net_device *dev;
46e5da40 561 struct Qdisc __rcu *qdisc;
b0e1e646 562 struct Qdisc *qdisc_sleeping;
ccf5ff69 563#ifdef CONFIG_SYSFS
1d24eb48
TH
564 struct kobject kobj;
565#endif
f2cd2d3e
ED
566#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
567 int numa_node;
568#endif
6a321cb3
ED
569/*
570 * write mostly part
571 */
572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
573 int xmit_lock_owner;
9d21493b
ED
574 /*
575 * please use this field instead of dev->trans_start
576 */
577 unsigned long trans_start;
ccf5ff69 578
579 /*
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
582 */
583 unsigned long trans_timeout;
114cf580
TH
584
585 unsigned long state;
586
587#ifdef CONFIG_BQL
588 struct dql dql;
589#endif
822b3b2e 590 unsigned long tx_maxrate;
e8a0464c 591} ____cacheline_aligned_in_smp;
bb949fbd 592
f2cd2d3e
ED
593static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
594{
595#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 return q->numa_node;
597#else
b236da69 598 return NUMA_NO_NODE;
f2cd2d3e
ED
599#endif
600}
601
602static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
603{
604#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 q->numa_node = node;
606#endif
607}
608
df334545 609#ifdef CONFIG_RPS
0a9627f2
TH
610/*
611 * This structure holds an RPS map which can be of variable length. The
612 * map is an array of CPUs.
613 */
614struct rps_map {
615 unsigned int len;
616 struct rcu_head rcu;
617 u16 cpus[0];
618};
60b778ce 619#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 620
fec5e652 621/*
c445477d
BH
622 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
623 * tail pointer for that CPU's input queue at the time of last enqueue, and
624 * a hardware filter index.
fec5e652
TH
625 */
626struct rps_dev_flow {
627 u16 cpu;
c445477d 628 u16 filter;
fec5e652
TH
629 unsigned int last_qtail;
630};
c445477d 631#define RPS_NO_FILTER 0xffff
fec5e652
TH
632
633/*
634 * The rps_dev_flow_table structure contains a table of flow mappings.
635 */
636struct rps_dev_flow_table {
637 unsigned int mask;
638 struct rcu_head rcu;
fec5e652
TH
639 struct rps_dev_flow flows[0];
640};
641#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 642 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
643
644/*
645 * The rps_sock_flow_table contains mappings of flows to the last CPU
646 * on which they were processed by the application (set in recvmsg).
567e4b79
ED
647 * Each entry is a 32bit value. Upper part is the high order bits
648 * of flow hash, lower part is cpu number.
649 * rps_cpu_mask is used to partition the space, depending on number of
650 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
651 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
652 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
653 */
654struct rps_sock_flow_table {
567e4b79 655 u32 mask;
93c1af6c
ED
656
657 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 658};
567e4b79 659#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
660
661#define RPS_NO_CPU 0xffff
662
567e4b79
ED
663extern u32 rps_cpu_mask;
664extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665
fec5e652
TH
666static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
667 u32 hash)
668{
669 if (table && hash) {
567e4b79
ED
670 unsigned int index = hash & table->mask;
671 u32 val = hash & ~rps_cpu_mask;
fec5e652
TH
672
673 /* We only give a hint, preemption can change cpu under us */
567e4b79 674 val |= raw_smp_processor_id();
fec5e652 675
567e4b79
ED
676 if (table->ents[index] != val)
677 table->ents[index] = val;
fec5e652
TH
678 }
679}
680
c445477d 681#ifdef CONFIG_RFS_ACCEL
f629d208
JP
682bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
683 u16 filter_id);
c445477d 684#endif
a953be53 685#endif /* CONFIG_RPS */
c445477d 686
0a9627f2
TH
687/* This structure contains an instance of an RX queue. */
688struct netdev_rx_queue {
a953be53 689#ifdef CONFIG_RPS
6e3f7faf
ED
690 struct rps_map __rcu *rps_map;
691 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 692#endif
6e3f7faf 693 struct kobject kobj;
fe822240 694 struct net_device *dev;
0a9627f2 695} ____cacheline_aligned_in_smp;
a953be53
MD
696
697/*
698 * RX queue sysfs structures and functions.
699 */
700struct rx_queue_attribute {
701 struct attribute attr;
702 ssize_t (*show)(struct netdev_rx_queue *queue,
703 struct rx_queue_attribute *attr, char *buf);
704 ssize_t (*store)(struct netdev_rx_queue *queue,
705 struct rx_queue_attribute *attr, const char *buf, size_t len);
706};
d314774c 707
bf264145
TH
708#ifdef CONFIG_XPS
709/*
710 * This structure holds an XPS map which can be of variable length. The
711 * map is an array of queues.
712 */
713struct xps_map {
714 unsigned int len;
715 unsigned int alloc_len;
716 struct rcu_head rcu;
717 u16 queues[0];
718};
60b778ce 719#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
bf264145
TH
720#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
721 / sizeof(u16))
722
723/*
724 * This structure holds all XPS maps for device. Maps are indexed by CPU.
725 */
726struct xps_dev_maps {
727 struct rcu_head rcu;
a4177869 728 struct xps_map __rcu *cpu_map[0];
bf264145
TH
729};
730#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
731 (nr_cpu_ids * sizeof(struct xps_map *)))
732#endif /* CONFIG_XPS */
733
4f57c087
JF
734#define TC_MAX_QUEUE 16
735#define TC_BITMASK 15
736/* HW offloaded queuing disciplines txq count and offset maps */
737struct netdev_tc_txq {
738 u16 count;
739 u16 offset;
740};
741
68bad94e
NP
742#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
743/*
744 * This structure is to hold information about the device
745 * configured to run FCoE protocol stack.
746 */
747struct netdev_fcoe_hbainfo {
748 char manufacturer[64];
749 char serial_number[64];
750 char hardware_version[64];
751 char driver_version[64];
752 char optionrom_version[64];
753 char firmware_version[64];
754 char model[256];
755 char model_description[256];
756};
757#endif
758
02637fce 759#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 760
02637fce
JP
761/* This structure holds a unique identifier to identify some
762 * physical item (port for example) used by a netdevice.
66b52b0d 763 */
02637fce
JP
764struct netdev_phys_item_id {
765 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
766 unsigned char id_len;
767};
768
d754f98b
SF
769static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
770 struct netdev_phys_item_id *b)
771{
772 return a->id_len == b->id_len &&
773 memcmp(a->id, b->id, a->id_len) == 0;
774}
775
99932d4f
DB
776typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
777 struct sk_buff *skb);
778
d314774c
SH
779/*
780 * This structure defines the management hooks for network devices.
00829823
SH
781 * The following hooks can be defined; unless noted otherwise, they are
782 * optional and can be filled with a null pointer.
d314774c
SH
783 *
784 * int (*ndo_init)(struct net_device *dev);
785 * This function is called once when network device is registered.
786 * The network device can use this to any late stage initializaton
787 * or semantic validattion. It can fail with an error code which will
788 * be propogated back to register_netdev
789 *
790 * void (*ndo_uninit)(struct net_device *dev);
791 * This function is called when device is unregistered or when registration
792 * fails. It is not called if init fails.
793 *
794 * int (*ndo_open)(struct net_device *dev);
795 * This function is called when network device transistions to the up
796 * state.
797 *
798 * int (*ndo_stop)(struct net_device *dev);
799 * This function is called when network device transistions to the down
800 * state.
801 *
dc1f8bf6
SH
802 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
803 * struct net_device *dev);
00829823 804 * Called when a packet needs to be transmitted.
e79d8429
RR
805 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
806 * the queue before that can happen; it's for obsolete devices and weird
807 * corner cases, but the stack really does a non-trivial amount
808 * of useless work if you return NETDEV_TX_BUSY.
dc1f8bf6 809 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
810 * Required can not be NULL.
811 *
f663dd9a 812 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 813 * void *accel_priv, select_queue_fallback_t fallback);
00829823
SH
814 * Called to decide which queue to when device supports multiple
815 * transmit queues.
816 *
d314774c
SH
817 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
818 * This function is called to allow device receiver to make
819 * changes to configuration when multicast or promiscious is enabled.
820 *
821 * void (*ndo_set_rx_mode)(struct net_device *dev);
822 * This function is called device changes address list filtering.
01789349
JP
823 * If driver handles unicast address filtering, it should set
824 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
825 *
826 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
827 * This function is called when the Media Access Control address
37b607c5 828 * needs to be changed. If this interface is not defined, the
d314774c
SH
829 * mac address can not be changed.
830 *
831 * int (*ndo_validate_addr)(struct net_device *dev);
832 * Test if Media Access Control address is valid for the device.
833 *
834 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
835 * Called when a user request an ioctl which can't be handled by
836 * the generic interface code. If not defined ioctl's return
837 * not supported error code.
838 *
839 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
840 * Used to set network devices bus interface parameters. This interface
841 * is retained for legacy reason, new devices should use the bus
842 * interface (PCI) for low level management.
843 *
844 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
845 * Called when a user wants to change the Maximum Transfer Unit
846 * of a device. If not defined, any request to change MTU will
847 * will return an error.
848 *
00829823 849 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
850 * Callback uses when the transmitter has not made any progress
851 * for dev->watchdog ticks.
852 *
3cfde79c 853 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 854 * struct rtnl_link_stats64 *storage);
d308e38f 855 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 856 * Called when a user wants to get the network device usage
be1f3c2c 857 * statistics. Drivers must do one of the following:
3cfde79c
BH
858 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
859 * rtnl_link_stats64 structure passed by the caller.
82695d9b 860 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
861 * (which should normally be dev->stats) and return a pointer to
862 * it. The structure may be changed asynchronously only if each
863 * field is written atomically.
864 * 3. Update dev->stats asynchronously and atomically, and define
865 * neither operation.
d314774c 866 *
5d632cb7 867 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
868 * If device support VLAN filtering this function is called when a
869 * VLAN id is registered.
d314774c 870 *
5d632cb7 871 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
872 * If device support VLAN filtering this function is called when a
873 * VLAN id is unregistered.
d314774c
SH
874 *
875 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
876 *
877 * SR-IOV management functions.
878 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
879 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
880 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
881 * int max_tx_rate);
5f8444a3 882 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
883 * int (*ndo_get_vf_config)(struct net_device *dev,
884 * int vf, struct ifla_vf_info *ivf);
1d8faf48 885 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
886 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
887 * struct nlattr *port[]);
01a3d796
VZ
888 *
889 * Enable or disable the VF ability to query its RSS Redirection Table and
890 * Hash Key. This is needed since on some devices VF share this information
891 * with PF and querying it may adduce a theoretical security risk.
892 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 893 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
894 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
895 * Called to setup 'tc' number of traffic classes in the net device. This
896 * is always called from the stack with the rtnl lock held and netif tx
897 * queues stopped. This allows the netdevice to perform queue management
898 * safely.
c445477d 899 *
e9bce845
YZ
900 * Fiber Channel over Ethernet (FCoE) offload functions.
901 * int (*ndo_fcoe_enable)(struct net_device *dev);
902 * Called when the FCoE protocol stack wants to start using LLD for FCoE
903 * so the underlying device can perform whatever needed configuration or
904 * initialization to support acceleration of FCoE traffic.
905 *
906 * int (*ndo_fcoe_disable)(struct net_device *dev);
907 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
908 * so the underlying device can perform whatever needed clean-ups to
909 * stop supporting acceleration of FCoE traffic.
910 *
911 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
912 * struct scatterlist *sgl, unsigned int sgc);
913 * Called when the FCoE Initiator wants to initialize an I/O that
914 * is a possible candidate for Direct Data Placement (DDP). The LLD can
915 * perform necessary setup and returns 1 to indicate the device is set up
916 * successfully to perform DDP on this I/O, otherwise this returns 0.
917 *
918 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
919 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
920 * indicated by the FC exchange id 'xid', so the underlying device can
921 * clean up and reuse resources for later DDP requests.
922 *
923 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
924 * struct scatterlist *sgl, unsigned int sgc);
925 * Called when the FCoE Target wants to initialize an I/O that
926 * is a possible candidate for Direct Data Placement (DDP). The LLD can
927 * perform necessary setup and returns 1 to indicate the device is set up
928 * successfully to perform DDP on this I/O, otherwise this returns 0.
929 *
68bad94e
NP
930 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
931 * struct netdev_fcoe_hbainfo *hbainfo);
932 * Called when the FCoE Protocol stack wants information on the underlying
933 * device. This information is utilized by the FCoE protocol stack to
934 * register attributes with Fiber Channel management service as per the
935 * FC-GS Fabric Device Management Information(FDMI) specification.
936 *
e9bce845
YZ
937 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
938 * Called when the underlying device wants to override default World Wide
939 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
940 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
941 * protocol stack to use.
942 *
c445477d
BH
943 * RFS acceleration.
944 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
945 * u16 rxq_index, u32 flow_id);
946 * Set hardware filter for RFS. rxq_index is the target queue index;
947 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
948 * Return the filter ID on success, or a negative error code.
fbaec0ea 949 *
8b98a70c 950 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
951 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
952 * Called to make another netdev an underling.
953 *
954 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
955 * Called to release previously enslaved netdev.
5455c699
MM
956 *
957 * Feature/offload setting functions.
c8f44aff
MM
958 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
959 * netdev_features_t features);
5455c699
MM
960 * Adjusts the requested feature flags according to device-specific
961 * constraints, and returns the resulting flags. Must not modify
962 * the device state.
963 *
c8f44aff 964 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
965 * Called to update device configuration to new features. Passed
966 * feature set might be less than what was returned by ndo_fix_features()).
967 * Must return >0 or -errno if it changed dev->features itself.
968 *
edc7d573 969 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
970 * struct net_device *dev,
f6f6424b 971 * const unsigned char *addr, u16 vid, u16 flags)
77162022 972 * Adds an FDB entry to dev for addr.
1690be63
VY
973 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
974 * struct net_device *dev,
f6f6424b 975 * const unsigned char *addr, u16 vid)
77162022
JF
976 * Deletes the FDB entry from dev coresponding to addr.
977 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
978 * struct net_device *dev, struct net_device *filter_dev,
979 * int idx)
77162022
JF
980 * Used to add FDB entries to dump requests. Implementers should add
981 * entries to skb and update idx with the number of entries.
e5a55a89 982 *
ad41faa8
ND
983 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
984 * u16 flags)
e5a55a89 985 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
986 * struct net_device *dev, u32 filter_mask,
987 * int nlflags)
ad41faa8
ND
988 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
989 * u16 flags);
4bf84c35
JP
990 *
991 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
992 * Called to change device carrier. Soft-devices (like dummy, team, etc)
993 * which do not represent real hardware may define this to allow their
994 * userspace components to manage their virtual carrier state. Devices
995 * that determine carrier state from physical hardware properties (eg
996 * network cables) or protocol-dependent mechanisms (eg
997 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
998 *
999 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1000 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1001 * Called to get ID of physical port of this device. If driver does
1002 * not implement this, it is assumed that the hw is not able to have
1003 * multiple net devices on single physical port.
53cf5275
JG
1004 *
1005 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 1006 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1007 * Called by vxlan to notiy a driver about the UDP port and socket
1008 * address family that vxlan is listnening to. It is called only when
1009 * a new port starts listening. The operation is protected by the
1010 * vxlan_net->sock_lock.
1011 *
1012 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 1013 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1014 * Called by vxlan to notify the driver about a UDP port and socket
1015 * address family that vxlan is not listening to anymore. The operation
1016 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
1017 *
1018 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1019 * struct net_device *dev)
1020 * Called by upper layer devices to accelerate switching or other
1021 * station functionality into hardware. 'pdev is the lowerdev
1022 * to use for the offload and 'dev' is the net device that will
1023 * back the offload. Returns a pointer to the private structure
1024 * the upper layer will maintain.
1025 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1026 * Called by upper layer device to delete the station created
1027 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1028 * the station and priv is the structure returned by the add
1029 * operation.
1030 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1031 * struct net_device *dev,
1032 * void *priv);
1033 * Callback to use for xmit over the accelerated station. This
1034 * is used in place of ndo_start_xmit on accelerated net
1035 * devices.
5f35227e
JG
1036 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1037 * struct net_device *dev
1038 * netdev_features_t features);
04ffcb25 1039 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1040 * performing offload operations on a given packet. This is to give
1041 * the device an opportunity to implement any restrictions that cannot
1042 * be otherwise expressed by feature flags. The check is called with
1043 * the set of features that the stack has calculated and it returns
1044 * those the driver believes to be appropriate.
822b3b2e
JF
1045 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1046 * int queue_index, u32 maxrate);
1047 * Called when a user wants to set a max-rate limitation of specific
1048 * TX queue.
a54acb3a
ND
1049 * int (*ndo_get_iflink)(const struct net_device *dev);
1050 * Called to get the iflink value of this device.
d746d707
AK
1051 * void (*ndo_change_proto_down)(struct net_device *dev,
1052 * bool proto_down);
1053 * This function is used to pass protocol port error state information
1054 * to the switch driver. The switch driver can react to the proto_down
1055 * by doing a phys down on the associated switch port.
1056 *
d314774c
SH
1057 */
1058struct net_device_ops {
1059 int (*ndo_init)(struct net_device *dev);
1060 void (*ndo_uninit)(struct net_device *dev);
1061 int (*ndo_open)(struct net_device *dev);
1062 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 1063 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
1064 struct net_device *dev);
1065 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1066 struct sk_buff *skb,
99932d4f
DB
1067 void *accel_priv,
1068 select_queue_fallback_t fallback);
d314774c
SH
1069 void (*ndo_change_rx_flags)(struct net_device *dev,
1070 int flags);
d314774c 1071 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1072 int (*ndo_set_mac_address)(struct net_device *dev,
1073 void *addr);
d314774c 1074 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1075 int (*ndo_do_ioctl)(struct net_device *dev,
1076 struct ifreq *ifr, int cmd);
d314774c
SH
1077 int (*ndo_set_config)(struct net_device *dev,
1078 struct ifmap *map);
00829823
SH
1079 int (*ndo_change_mtu)(struct net_device *dev,
1080 int new_mtu);
1081 int (*ndo_neigh_setup)(struct net_device *dev,
1082 struct neigh_parms *);
d314774c
SH
1083 void (*ndo_tx_timeout) (struct net_device *dev);
1084
28172739
ED
1085 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1086 struct rtnl_link_stats64 *storage);
d314774c
SH
1087 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1088
8e586137 1089 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1090 __be16 proto, u16 vid);
8e586137 1091 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1092 __be16 proto, u16 vid);
d314774c 1093#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1094 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1095 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1096 struct netpoll_info *info);
0e34e931 1097 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1098#endif
e0d1095a 1099#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1100 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1101#endif
95c26df8
WM
1102 int (*ndo_set_vf_mac)(struct net_device *dev,
1103 int queue, u8 *mac);
1104 int (*ndo_set_vf_vlan)(struct net_device *dev,
1105 int queue, u16 vlan, u8 qos);
ed616689
SC
1106 int (*ndo_set_vf_rate)(struct net_device *dev,
1107 int vf, int min_tx_rate,
1108 int max_tx_rate);
5f8444a3
GR
1109 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1110 int vf, bool setting);
95c26df8
WM
1111 int (*ndo_get_vf_config)(struct net_device *dev,
1112 int vf,
1113 struct ifla_vf_info *ivf);
1d8faf48
RE
1114 int (*ndo_set_vf_link_state)(struct net_device *dev,
1115 int vf, int link_state);
3b766cd8
EBE
1116 int (*ndo_get_vf_stats)(struct net_device *dev,
1117 int vf,
1118 struct ifla_vf_stats
1119 *vf_stats);
57b61080
SF
1120 int (*ndo_set_vf_port)(struct net_device *dev,
1121 int vf,
1122 struct nlattr *port[]);
1123 int (*ndo_get_vf_port)(struct net_device *dev,
1124 int vf, struct sk_buff *skb);
01a3d796
VZ
1125 int (*ndo_set_vf_rss_query_en)(
1126 struct net_device *dev,
1127 int vf, bool setting);
4f57c087 1128 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
d11ead75 1129#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1130 int (*ndo_fcoe_enable)(struct net_device *dev);
1131 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1132 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1133 u16 xid,
1134 struct scatterlist *sgl,
1135 unsigned int sgc);
1136 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1137 u16 xid);
6247e086
YZ
1138 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1139 u16 xid,
1140 struct scatterlist *sgl,
1141 unsigned int sgc);
68bad94e
NP
1142 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1143 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1144#endif
1145
d11ead75 1146#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1147#define NETDEV_FCOE_WWNN 0
1148#define NETDEV_FCOE_WWPN 1
1149 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1150 u64 *wwn, int type);
4d288d57 1151#endif
3c9c36bc 1152
c445477d
BH
1153#ifdef CONFIG_RFS_ACCEL
1154 int (*ndo_rx_flow_steer)(struct net_device *dev,
1155 const struct sk_buff *skb,
1156 u16 rxq_index,
1157 u32 flow_id);
1158#endif
fbaec0ea
JP
1159 int (*ndo_add_slave)(struct net_device *dev,
1160 struct net_device *slave_dev);
1161 int (*ndo_del_slave)(struct net_device *dev,
1162 struct net_device *slave_dev);
c8f44aff
MM
1163 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1164 netdev_features_t features);
5455c699 1165 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1166 netdev_features_t features);
da6a8fa0 1167 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1168 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1169
1170 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1171 struct nlattr *tb[],
77162022 1172 struct net_device *dev,
6b6e2725 1173 const unsigned char *addr,
f6f6424b 1174 u16 vid,
77162022
JF
1175 u16 flags);
1176 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1177 struct nlattr *tb[],
77162022 1178 struct net_device *dev,
f6f6424b
JP
1179 const unsigned char *addr,
1180 u16 vid);
77162022
JF
1181 int (*ndo_fdb_dump)(struct sk_buff *skb,
1182 struct netlink_callback *cb,
1183 struct net_device *dev,
5d5eacb3 1184 struct net_device *filter_dev,
77162022 1185 int idx);
e5a55a89
JF
1186
1187 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1188 struct nlmsghdr *nlh,
1189 u16 flags);
e5a55a89
JF
1190 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1191 u32 pid, u32 seq,
6cbdceeb 1192 struct net_device *dev,
46c264da
ND
1193 u32 filter_mask,
1194 int nlflags);
407af329 1195 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1196 struct nlmsghdr *nlh,
1197 u16 flags);
4bf84c35
JP
1198 int (*ndo_change_carrier)(struct net_device *dev,
1199 bool new_carrier);
66b52b0d 1200 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1201 struct netdev_phys_item_id *ppid);
db24a904
DA
1202 int (*ndo_get_phys_port_name)(struct net_device *dev,
1203 char *name, size_t len);
53cf5275
JG
1204 void (*ndo_add_vxlan_port)(struct net_device *dev,
1205 sa_family_t sa_family,
35e42379 1206 __be16 port);
53cf5275
JG
1207 void (*ndo_del_vxlan_port)(struct net_device *dev,
1208 sa_family_t sa_family,
35e42379 1209 __be16 port);
a6cc0cfa
JF
1210
1211 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1212 struct net_device *dev);
1213 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1214 void *priv);
1215
1216 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1217 struct net_device *dev,
1218 void *priv);
25175ba5 1219 int (*ndo_get_lock_subclass)(struct net_device *dev);
5f35227e
JG
1220 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1221 struct net_device *dev,
1222 netdev_features_t features);
822b3b2e
JF
1223 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1224 int queue_index,
1225 u32 maxrate);
a54acb3a 1226 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1227 int (*ndo_change_proto_down)(struct net_device *dev,
1228 bool proto_down);
d314774c
SH
1229};
1230
7aa98047
LR
1231/**
1232 * enum net_device_priv_flags - &struct net_device priv_flags
1233 *
1234 * These are the &struct net_device, they are only set internally
1235 * by drivers and used in the kernel. These flags are invisible to
1236 * userspace, this means that the order of these flags can change
1237 * during any kernel release.
1238 *
1239 * You should have a pretty good reason to be extending these flags.
1240 *
1241 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1242 * @IFF_EBRIDGE: Ethernet bridging device
1243 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1244 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1245 * @IFF_MASTER_ALB: bonding master, balance-alb
1246 * @IFF_BONDING: bonding master or slave
1247 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1248 * @IFF_ISATAP: ISATAP interface (RFC4214)
1249 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1250 * @IFF_WAN_HDLC: WAN HDLC device
1251 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1252 * release skb->dst
1253 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1254 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1255 * @IFF_MACVLAN_PORT: device used as macvlan port
1256 * @IFF_BRIDGE_PORT: device used as bridge port
1257 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1258 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1259 * @IFF_UNICAST_FLT: Supports unicast filtering
1260 * @IFF_TEAM_PORT: device used as team port
1261 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1262 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1263 * change when it's running
1264 * @IFF_MACVLAN: Macvlan device
808d28c4 1265 * @IFF_VRF_MASTER: device is a VRF master
fa8187c9 1266 * @IFF_NO_QUEUE: device can run without qdisc attached
7aa98047
LR
1267 */
1268enum netdev_priv_flags {
1269 IFF_802_1Q_VLAN = 1<<0,
1270 IFF_EBRIDGE = 1<<1,
1271 IFF_SLAVE_INACTIVE = 1<<2,
1272 IFF_MASTER_8023AD = 1<<3,
1273 IFF_MASTER_ALB = 1<<4,
1274 IFF_BONDING = 1<<5,
1275 IFF_SLAVE_NEEDARP = 1<<6,
1276 IFF_ISATAP = 1<<7,
1277 IFF_MASTER_ARPMON = 1<<8,
1278 IFF_WAN_HDLC = 1<<9,
1279 IFF_XMIT_DST_RELEASE = 1<<10,
1280 IFF_DONT_BRIDGE = 1<<11,
1281 IFF_DISABLE_NETPOLL = 1<<12,
1282 IFF_MACVLAN_PORT = 1<<13,
1283 IFF_BRIDGE_PORT = 1<<14,
1284 IFF_OVS_DATAPATH = 1<<15,
1285 IFF_TX_SKB_SHARING = 1<<16,
1286 IFF_UNICAST_FLT = 1<<17,
1287 IFF_TEAM_PORT = 1<<18,
1288 IFF_SUPP_NOFCS = 1<<19,
1289 IFF_LIVE_ADDR_CHANGE = 1<<20,
1290 IFF_MACVLAN = 1<<21,
02875878 1291 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
2ad7bf36
MB
1292 IFF_IPVLAN_MASTER = 1<<23,
1293 IFF_IPVLAN_SLAVE = 1<<24,
4e3c8992 1294 IFF_VRF_MASTER = 1<<25,
fa8187c9 1295 IFF_NO_QUEUE = 1<<26,
7aa98047
LR
1296};
1297
1298#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1299#define IFF_EBRIDGE IFF_EBRIDGE
1300#define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1301#define IFF_MASTER_8023AD IFF_MASTER_8023AD
1302#define IFF_MASTER_ALB IFF_MASTER_ALB
1303#define IFF_BONDING IFF_BONDING
1304#define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1305#define IFF_ISATAP IFF_ISATAP
1306#define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1307#define IFF_WAN_HDLC IFF_WAN_HDLC
1308#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1309#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1310#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1311#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1312#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1313#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1314#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1315#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1316#define IFF_TEAM_PORT IFF_TEAM_PORT
1317#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1318#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1319#define IFF_MACVLAN IFF_MACVLAN
02875878 1320#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1321#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1322#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
4e3c8992 1323#define IFF_VRF_MASTER IFF_VRF_MASTER
fa8187c9 1324#define IFF_NO_QUEUE IFF_NO_QUEUE
7aa98047 1325
536721b1
KK
1326/**
1327 * struct net_device - The DEVICE structure.
1328 * Actually, this whole structure is a big mistake. It mixes I/O
1329 * data with strictly "high-level" data, and it has to know about
1330 * almost every data structure used in the INET module.
1331 *
1332 * @name: This is the first field of the "visible" part of this structure
1333 * (i.e. as seen by users in the "Space.c" file). It is the name
1334 * of the interface.
1335 *
1336 * @name_hlist: Device name hash chain, please keep it close to name[]
1337 * @ifalias: SNMP alias
1338 * @mem_end: Shared memory end
1339 * @mem_start: Shared memory start
1340 * @base_addr: Device I/O address
1341 * @irq: Device IRQ number
1342 *
14ffbbb8
TG
1343 * @carrier_changes: Stats to monitor carrier on<->off transitions
1344 *
536721b1
KK
1345 * @state: Generic network queuing layer state, see netdev_state_t
1346 * @dev_list: The global list of network devices
1347 * @napi_list: List entry, that is used for polling napi devices
1348 * @unreg_list: List entry, that is used, when we are unregistering the
1349 * device, see the function unregister_netdev
1350 * @close_list: List entry, that is used, when we are closing the device
1351 *
1352 * @adj_list: Directly linked devices, like slaves for bonding
1353 * @all_adj_list: All linked devices, *including* neighbours
1354 * @features: Currently active device features
1355 * @hw_features: User-changeable features
1356 *
1357 * @wanted_features: User-requested features
1358 * @vlan_features: Mask of features inheritable by VLAN devices
1359 *
1360 * @hw_enc_features: Mask of features inherited by encapsulating devices
1361 * This field indicates what encapsulation
1362 * offloads the hardware is capable of doing,
1363 * and drivers will need to set them appropriately.
1364 *
1365 * @mpls_features: Mask of features inheritable by MPLS
1366 *
1367 * @ifindex: interface index
388069d3 1368 * @group: The group, that the device belongs to
536721b1
KK
1369 *
1370 * @stats: Statistics struct, which was left as a legacy, use
1371 * rtnl_link_stats64 instead
1372 *
1373 * @rx_dropped: Dropped packets by core network,
1374 * do not use this in drivers
1375 * @tx_dropped: Dropped packets by core network,
1376 * do not use this in drivers
1377 *
536721b1
KK
1378 * @wireless_handlers: List of functions to handle Wireless Extensions,
1379 * instead of ioctl,
1380 * see <net/iw_handler.h> for details.
1381 * @wireless_data: Instance data managed by the core of wireless extensions
1382 *
1383 * @netdev_ops: Includes several pointers to callbacks,
1384 * if one wants to override the ndo_*() functions
1385 * @ethtool_ops: Management operations
d476059e 1386 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1387 * of Layer 2 headers.
1388 *
1389 * @flags: Interface flags (a la BSD)
1390 * @priv_flags: Like 'flags' but invisible to userspace,
1391 * see if.h for the definitions
1392 * @gflags: Global flags ( kept as legacy )
1393 * @padded: How much padding added by alloc_netdev()
1394 * @operstate: RFC2863 operstate
1395 * @link_mode: Mapping policy to operstate
1396 * @if_port: Selectable AUI, TP, ...
1397 * @dma: DMA channel
1398 * @mtu: Interface MTU value
1399 * @type: Interface hardware type
1400 * @hard_header_len: Hardware header length
1401 *
1402 * @needed_headroom: Extra headroom the hardware may need, but not in all
1403 * cases can this be guaranteed
1404 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1405 * cases can this be guaranteed. Some cases also use
1406 * LL_MAX_HEADER instead to allocate the skb
1407 *
1408 * interface address info:
1409 *
1410 * @perm_addr: Permanent hw address
1411 * @addr_assign_type: Hw address assignment type
1412 * @addr_len: Hardware address length
1413 * @neigh_priv_len; Used in neigh_alloc(),
1414 * initialized only in atm/clip.c
1415 * @dev_id: Used to differentiate devices that share
1416 * the same link layer address
1417 * @dev_port: Used to differentiate devices that share
1418 * the same function
1419 * @addr_list_lock: XXX: need comments on this one
536721b1
KK
1420 * @uc_promisc: Counter, that indicates, that promiscuous mode
1421 * has been enabled due to the need to listen to
1422 * additional unicast addresses in a device that
1423 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1424 * @uc: unicast mac addresses
1425 * @mc: multicast mac addresses
1426 * @dev_addrs: list of device hw addresses
1427 * @queues_kset: Group of all Kobjects in the Tx and RX queues
536721b1
KK
1428 * @promiscuity: Number of times, the NIC is told to work in
1429 * Promiscuous mode, if it becomes 0 the NIC will
1430 * exit from working in Promiscuous mode
1431 * @allmulti: Counter, enables or disables allmulticast mode
1432 *
1433 * @vlan_info: VLAN info
1434 * @dsa_ptr: dsa specific data
1435 * @tipc_ptr: TIPC specific data
1436 * @atalk_ptr: AppleTalk link
1437 * @ip_ptr: IPv4 specific data
1438 * @dn_ptr: DECnet specific data
1439 * @ip6_ptr: IPv6 specific data
1440 * @ax25_ptr: AX.25 specific data
4e3c8992 1441 * @vrf_ptr: VRF specific data
536721b1
KK
1442 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1443 *
1444 * @last_rx: Time of last Rx
1445 * @dev_addr: Hw address (before bcast,
1446 * because most packets are unicast)
1447 *
1448 * @_rx: Array of RX queues
1449 * @num_rx_queues: Number of RX queues
1450 * allocated at register_netdev() time
1451 * @real_num_rx_queues: Number of RX queues currently active in device
1452 *
1453 * @rx_handler: handler for received packets
1454 * @rx_handler_data: XXX: need comments on this one
1455 * @ingress_queue: XXX: need comments on this one
1456 * @broadcast: hw bcast address
1457 *
14ffbbb8
TG
1458 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1459 * indexed by RX queue number. Assigned by driver.
1460 * This must only be set if the ndo_rx_flow_steer
1461 * operation is defined
1462 * @index_hlist: Device index hash chain
1463 *
536721b1
KK
1464 * @_tx: Array of TX queues
1465 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1466 * @real_num_tx_queues: Number of TX queues currently active in device
1467 * @qdisc: Root qdisc from userspace point of view
1468 * @tx_queue_len: Max frames per queue allowed
1469 * @tx_global_lock: XXX: need comments on this one
1470 *
1471 * @xps_maps: XXX: need comments on this one
1472 *
0c4f691f
SF
1473 * @offload_fwd_mark: Offload device fwding mark
1474 *
536721b1
KK
1475 * @trans_start: Time (in jiffies) of last Tx
1476 * @watchdog_timeo: Represents the timeout that is used by
1477 * the watchdog ( see dev_watchdog() )
1478 * @watchdog_timer: List of timers
1479 *
1480 * @pcpu_refcnt: Number of references to this device
1481 * @todo_list: Delayed register/unregister
536721b1
KK
1482 * @link_watch_list: XXX: need comments on this one
1483 *
1484 * @reg_state: Register/unregister state machine
1485 * @dismantle: Device is going to be freed
1486 * @rtnl_link_state: This enum represents the phases of creating
1487 * a new link
1488 *
1489 * @destructor: Called from unregister,
1490 * can be used to call free_netdev
1491 * @npinfo: XXX: need comments on this one
1492 * @nd_net: Network namespace this network device is inside
1493 *
1494 * @ml_priv: Mid-layer private
1495 * @lstats: Loopback statistics
1496 * @tstats: Tunnel statistics
1497 * @dstats: Dummy statistics
1498 * @vstats: Virtual ethernet statistics
1499 *
1500 * @garp_port: GARP
1501 * @mrp_port: MRP
1502 *
1503 * @dev: Class/net/name entry
1504 * @sysfs_groups: Space for optional device, statistics and wireless
1505 * sysfs groups
1506 *
1507 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1508 * @rtnl_link_ops: Rtnl_link_ops
1509 *
1510 * @gso_max_size: Maximum size of generic segmentation offload
1511 * @gso_max_segs: Maximum number of segments that can be passed to the
1512 * NIC for GSO
fcbeb976
ED
1513 * @gso_min_segs: Minimum number of segments that can be passed to the
1514 * NIC for GSO
536721b1
KK
1515 *
1516 * @dcbnl_ops: Data Center Bridging netlink ops
1517 * @num_tc: Number of traffic classes in the net device
1518 * @tc_to_txq: XXX: need comments on this one
1519 * @prio_tc_map XXX: need comments on this one
1520 *
1521 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1522 *
1523 * @priomap: XXX: need comments on this one
1524 * @phydev: Physical device may attach itself
1525 * for hardware timestamping
1526 *
1527 * @qdisc_tx_busylock: XXX: need comments on this one
1528 *
d746d707
AK
1529 * @proto_down: protocol port state information can be sent to the
1530 * switch driver and used to set the phys state of the
1531 * switch port.
1532 *
1da177e4
LT
1533 * FIXME: cleanup struct net_device such that network protocol info
1534 * moves out.
1535 */
1536
d94d9fee 1537struct net_device {
1da177e4 1538 char name[IFNAMSIZ];
9356b8fc 1539 struct hlist_node name_hlist;
0b815a1a 1540 char *ifalias;
1da177e4
LT
1541 /*
1542 * I/O specific fields
1543 * FIXME: Merge these and struct ifmap into one
1544 */
536721b1
KK
1545 unsigned long mem_end;
1546 unsigned long mem_start;
1547 unsigned long base_addr;
1548 int irq;
1da177e4 1549
14ffbbb8
TG
1550 atomic_t carrier_changes;
1551
1da177e4 1552 /*
536721b1
KK
1553 * Some hardware also needs these fields (state,dev_list,
1554 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1555 * part of the usual set specified in Space.c.
1556 */
1557
1da177e4
LT
1558 unsigned long state;
1559
7562f876 1560 struct list_head dev_list;
bea3348e 1561 struct list_head napi_list;
44a0873d 1562 struct list_head unreg_list;
5cde2829 1563 struct list_head close_list;
7866a621
SN
1564 struct list_head ptype_all;
1565 struct list_head ptype_specific;
2f268f12 1566
2f268f12
VF
1567 struct {
1568 struct list_head upper;
1569 struct list_head lower;
1570 } adj_list;
1571
2f268f12
VF
1572 struct {
1573 struct list_head upper;
1574 struct list_head lower;
1575 } all_adj_list;
4c3d5e7b 1576
c8f44aff 1577 netdev_features_t features;
c8f44aff 1578 netdev_features_t hw_features;
c8f44aff 1579 netdev_features_t wanted_features;
c8f44aff 1580 netdev_features_t vlan_features;
6a674e9c 1581 netdev_features_t hw_enc_features;
0d89d203 1582 netdev_features_t mpls_features;
04ed3e74 1583
1da177e4 1584 int ifindex;
7a66bbc9 1585 int group;
1da177e4 1586
c45d286e 1587 struct net_device_stats stats;
015f0688 1588
015f0688
ED
1589 atomic_long_t rx_dropped;
1590 atomic_long_t tx_dropped;
1da177e4 1591
b86e0280 1592#ifdef CONFIG_WIRELESS_EXT
1da177e4 1593 const struct iw_handler_def * wireless_handlers;
1da177e4 1594 struct iw_public_data * wireless_data;
b86e0280 1595#endif
d314774c 1596 const struct net_device_ops *netdev_ops;
76fd8593 1597 const struct ethtool_ops *ethtool_ops;
4170604f 1598#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1599 const struct switchdev_ops *switchdev_ops;
4170604f 1600#endif
1da177e4 1601
3b04ddde
SH
1602 const struct header_ops *header_ops;
1603
536721b1
KK
1604 unsigned int flags;
1605 unsigned int priv_flags;
1606
1da177e4 1607 unsigned short gflags;
536721b1 1608 unsigned short padded;
1da177e4 1609
536721b1
KK
1610 unsigned char operstate;
1611 unsigned char link_mode;
b00055aa 1612
536721b1
KK
1613 unsigned char if_port;
1614 unsigned char dma;
bdc220da 1615
536721b1
KK
1616 unsigned int mtu;
1617 unsigned short type;
1618 unsigned short hard_header_len;
1da177e4 1619
f5184d26
JB
1620 unsigned short needed_headroom;
1621 unsigned short needed_tailroom;
1622
1da177e4 1623 /* Interface address info. */
536721b1
KK
1624 unsigned char perm_addr[MAX_ADDR_LEN];
1625 unsigned char addr_assign_type;
1626 unsigned char addr_len;
a0a9663d 1627 unsigned short neigh_priv_len;
536721b1
KK
1628 unsigned short dev_id;
1629 unsigned short dev_port;
ccffad25 1630 spinlock_t addr_list_lock;
14ffbbb8
TG
1631 unsigned char name_assign_type;
1632 bool uc_promisc;
536721b1
KK
1633 struct netdev_hw_addr_list uc;
1634 struct netdev_hw_addr_list mc;
1635 struct netdev_hw_addr_list dev_addrs;
1636
4c3d5e7b
ED
1637#ifdef CONFIG_SYSFS
1638 struct kset *queues_kset;
1639#endif
9d45abe1
WC
1640 unsigned int promiscuity;
1641 unsigned int allmulti;
1da177e4 1642
1da177e4
LT
1643
1644 /* Protocol specific pointers */
65ac6a5f 1645
d11ead75 1646#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1647 struct vlan_info __rcu *vlan_info;
65ac6a5f 1648#endif
34a430d7 1649#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1650 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1651#endif
1652#if IS_ENABLED(CONFIG_TIPC)
536721b1 1653 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1654#endif
536721b1
KK
1655 void *atalk_ptr;
1656 struct in_device __rcu *ip_ptr;
1657 struct dn_dev __rcu *dn_ptr;
1658 struct inet6_dev __rcu *ip6_ptr;
1659 void *ax25_ptr;
4e3c8992 1660 struct net_vrf_dev __rcu *vrf_ptr;
536721b1 1661 struct wireless_dev *ieee80211_ptr;
98a18b6f 1662 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1663#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1664 struct mpls_dev __rcu *mpls_ptr;
1665#endif
1da177e4 1666
9356b8fc 1667/*
cd13539b 1668 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1669 */
536721b1 1670 unsigned long last_rx;
4dc89133 1671
9356b8fc 1672 /* Interface address info used in eth_type_trans() */
536721b1 1673 unsigned char *dev_addr;
f001fde5 1674
0a9627f2 1675
a953be53 1676#ifdef CONFIG_SYSFS
0a9627f2
TH
1677 struct netdev_rx_queue *_rx;
1678
0a9627f2 1679 unsigned int num_rx_queues;
62fe0b40 1680 unsigned int real_num_rx_queues;
c445477d 1681
df334545 1682#endif
0a9627f2 1683
3b47d303 1684 unsigned long gro_flush_timeout;
61391cde 1685 rx_handler_func_t __rcu *rx_handler;
1686 void __rcu *rx_handler_data;
e8a0464c 1687
4cda01e8 1688#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1689 struct tcf_proto __rcu *ingress_cl_list;
1690#endif
24824a09 1691 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1692#ifdef CONFIG_NETFILTER_INGRESS
1693 struct list_head nf_hooks_ingress;
1694#endif
d2788d34 1695
536721b1 1696 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1697#ifdef CONFIG_RFS_ACCEL
1698 struct cpu_rmap *rx_cpu_rmap;
1699#endif
1700 struct hlist_node index_hlist;
cd13539b
ED
1701
1702/*
1703 * Cache lines mostly used on transmit path
1704 */
e8a0464c
DM
1705 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1706 unsigned int num_tx_queues;
fd2ea0a7 1707 unsigned int real_num_tx_queues;
af356afa 1708 struct Qdisc *qdisc;
536721b1 1709 unsigned long tx_queue_len;
c3f26a26 1710 spinlock_t tx_global_lock;
14ffbbb8 1711 int watchdog_timeo;
cd13539b 1712
bf264145 1713#ifdef CONFIG_XPS
a4177869 1714 struct xps_dev_maps __rcu *xps_maps;
bf264145 1715#endif
1d24eb48 1716
0c4f691f
SF
1717#ifdef CONFIG_NET_SWITCHDEV
1718 u32 offload_fwd_mark;
1719#endif
1720
9356b8fc 1721 /* These may be needed for future network-power-down code. */
9d21493b
ED
1722
1723 /*
1724 * trans_start here is expensive for high speed devices on SMP,
1725 * please use netdev_queue->trans_start instead.
1726 */
536721b1 1727 unsigned long trans_start;
9356b8fc 1728
9356b8fc
ED
1729 struct timer_list watchdog_timer;
1730
29b4433d 1731 int __percpu *pcpu_refcnt;
1da177e4 1732 struct list_head todo_list;
1da177e4 1733
e014debe 1734 struct list_head link_watch_list;
572a103d 1735
1da177e4 1736 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1737 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1738 NETREG_UNREGISTERING, /* called unregister_netdevice */
1739 NETREG_UNREGISTERED, /* completed unregister todo */
1740 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1741 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1742 } reg_state:8;
1743
536721b1 1744 bool dismantle;
a2835763
PM
1745
1746 enum {
1747 RTNL_LINK_INITIALIZED,
1748 RTNL_LINK_INITIALIZING,
1749 } rtnl_link_state:16;
1da177e4 1750
d314774c 1751 void (*destructor)(struct net_device *dev);
1da177e4 1752
1da177e4 1753#ifdef CONFIG_NETPOLL
5fbee843 1754 struct netpoll_info __rcu *npinfo;
1da177e4 1755#endif
eae792b7 1756
0c5c9fb5 1757 possible_net_t nd_net;
4a1c5371 1758
4951704b 1759 /* mid-layer private */
a7855c78 1760 union {
536721b1
KK
1761 void *ml_priv;
1762 struct pcpu_lstats __percpu *lstats;
8f84985f 1763 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1764 struct pcpu_dstats __percpu *dstats;
1765 struct pcpu_vstats __percpu *vstats;
a7855c78 1766 };
536721b1 1767
3cc77ec7 1768 struct garp_port __rcu *garp_port;
febf018d 1769 struct mrp_port __rcu *mrp_port;
1da177e4 1770
536721b1 1771 struct device dev;
0c509a6c 1772 const struct attribute_group *sysfs_groups[4];
a953be53 1773 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1774
38f7b870 1775 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1776
82cc1a7a
PWJ
1777 /* for setting kernel sock attribute on TCP connection setup */
1778#define GSO_MAX_SIZE 65536
1779 unsigned int gso_max_size;
30b678d8
BH
1780#define GSO_MAX_SEGS 65535
1781 u16 gso_max_segs;
fcbeb976 1782 u16 gso_min_segs;
7a6b6f51 1783#ifdef CONFIG_DCB
32953543 1784 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1785#endif
4f57c087
JF
1786 u8 num_tc;
1787 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1788 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1789
d11ead75 1790#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1791 unsigned int fcoe_ddp_xid;
5bc1421e 1792#endif
86f8515f 1793#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1794 struct netprio_map __rcu *priomap;
4d288d57 1795#endif
c1f19b51 1796 struct phy_device *phydev;
23d3b8bf 1797 struct lock_class_key *qdisc_tx_busylock;
d746d707 1798 bool proto_down;
1da177e4 1799};
43cb76d9 1800#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1801
1802#define NETDEV_ALIGN 32
1da177e4 1803
4f57c087
JF
1804static inline
1805int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1806{
1807 return dev->prio_tc_map[prio & TC_BITMASK];
1808}
1809
1810static inline
1811int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1812{
1813 if (tc >= dev->num_tc)
1814 return -EINVAL;
1815
1816 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1817 return 0;
1818}
1819
1820static inline
1821void netdev_reset_tc(struct net_device *dev)
1822{
1823 dev->num_tc = 0;
1824 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1825 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1826}
1827
1828static inline
1829int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1830{
1831 if (tc >= dev->num_tc)
1832 return -EINVAL;
1833
1834 dev->tc_to_txq[tc].count = count;
1835 dev->tc_to_txq[tc].offset = offset;
1836 return 0;
1837}
1838
1839static inline
1840int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1841{
1842 if (num_tc > TC_MAX_QUEUE)
1843 return -EINVAL;
1844
1845 dev->num_tc = num_tc;
1846 return 0;
1847}
1848
1849static inline
1850int netdev_get_num_tc(struct net_device *dev)
1851{
1852 return dev->num_tc;
1853}
1854
e8a0464c
DM
1855static inline
1856struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1857 unsigned int index)
1858{
1859 return &dev->_tx[index];
1860}
1861
10c51b56
DB
1862static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1863 const struct sk_buff *skb)
1864{
1865 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1866}
1867
e8a0464c
DM
1868static inline void netdev_for_each_tx_queue(struct net_device *dev,
1869 void (*f)(struct net_device *,
1870 struct netdev_queue *,
1871 void *),
1872 void *arg)
1873{
1874 unsigned int i;
1875
1876 for (i = 0; i < dev->num_tx_queues; i++)
1877 f(dev, &dev->_tx[i], arg);
1878}
1879
f629d208 1880struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1881 struct sk_buff *skb,
1882 void *accel_priv);
8c4c49df 1883
c346dca1
YH
1884/*
1885 * Net namespace inlines
1886 */
1887static inline
1888struct net *dev_net(const struct net_device *dev)
1889{
c2d9ba9b 1890 return read_pnet(&dev->nd_net);
c346dca1
YH
1891}
1892
1893static inline
f5aa23fd 1894void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 1895{
0c5c9fb5 1896 write_pnet(&dev->nd_net, net);
c346dca1
YH
1897}
1898
3e8a72d1 1899static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 1900{
3fc88677 1901#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
1902 if (dev->dsa_ptr != NULL)
1903 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 1904#endif
5aed85ce 1905 return false;
396138f0
LB
1906}
1907
bea3348e
SH
1908/**
1909 * netdev_priv - access network device private data
1910 * @dev: network device
1911 *
1912 * Get network device private data
1913 */
6472ce60 1914static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1915{
1ce8e7b5 1916 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1917}
1918
1da177e4
LT
1919/* Set the sysfs physical device reference for the network logical device
1920 * if set prior to registration will cause a symlink during initialization.
1921 */
43cb76d9 1922#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1923
384912ed 1924/* Set the sysfs device type for the network logical device to allow
3f79410c 1925 * fine-grained identification of different network device types. For
384912ed
MH
1926 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1927 */
1928#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1929
82dc3c63
ED
1930/* Default NAPI poll() weight
1931 * Device drivers are strongly advised to not use bigger value
1932 */
1933#define NAPI_POLL_WEIGHT 64
1934
3b582cc1
SH
1935/**
1936 * netif_napi_add - initialize a napi context
1937 * @dev: network device
1938 * @napi: napi context
1939 * @poll: polling function
1940 * @weight: default weight
1941 *
1942 * netif_napi_add() must be used to initialize a napi context prior to calling
1943 * *any* of the other napi related functions.
1944 */
d565b0a1
HX
1945void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1946 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1947
d8156534
AD
1948/**
1949 * netif_napi_del - remove a napi context
1950 * @napi: napi context
1951 *
1952 * netif_napi_del() removes a napi context from the network device napi list
1953 */
d565b0a1
HX
1954void netif_napi_del(struct napi_struct *napi);
1955
1956struct napi_gro_cb {
78a478d0
HX
1957 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1958 void *frag0;
1959
7489594c
HX
1960 /* Length of frag0. */
1961 unsigned int frag0_len;
1962
86911732
HX
1963 /* This indicates where we are processing relative to skb->data. */
1964 int data_offset;
1965
d565b0a1 1966 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
1967 u16 flush;
1968
1969 /* Save the IP ID here and check when we get to the transport layer */
1970 u16 flush_id;
d565b0a1
HX
1971
1972 /* Number of segments aggregated. */
2e71a6f8
ED
1973 u16 count;
1974
15e2396d
TH
1975 /* Start offset for remote checksum offload */
1976 u16 gro_remcsum_start;
1977
2e71a6f8
ED
1978 /* jiffies when first packet was created/queued */
1979 unsigned long age;
86347245 1980
afe93325 1981 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
1982 u16 proto;
1983
baa32ff4
TH
1984 /* This is non-zero if the packet may be of the same flow. */
1985 u8 same_flow:1;
1986
b582ef09 1987 /* Used in udp_gro_receive */
573e8fca
TH
1988 u8 udp_mark:1;
1989
1990 /* GRO checksum is valid */
1991 u8 csum_valid:1;
1992
662880f4
TH
1993 /* Number of checksums via CHECKSUM_UNNECESSARY */
1994 u8 csum_cnt:3;
c3c7c254 1995
baa32ff4
TH
1996 /* Free the skb? */
1997 u8 free:2;
1998#define NAPI_GRO_FREE 1
1999#define NAPI_GRO_FREE_STOLEN_HEAD 2
2000
efc98d08
TH
2001 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2002 u8 is_ipv6:1;
2003
baa32ff4
TH
2004 /* 7 bit hole */
2005
bf5a755f
JC
2006 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2007 __wsum csum;
2008
c3c7c254
ED
2009 /* used in skb_gro_receive() slow path */
2010 struct sk_buff *last;
d565b0a1
HX
2011};
2012
2013#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2014
1da177e4 2015struct packet_type {
f2ccd8fa
DM
2016 __be16 type; /* This is really htons(ether_type). */
2017 struct net_device *dev; /* NULL is wildcarded here */
2018 int (*func) (struct sk_buff *,
2019 struct net_device *,
2020 struct packet_type *,
2021 struct net_device *);
c0de08d0
EL
2022 bool (*id_match)(struct packet_type *ptype,
2023 struct sock *sk);
1da177e4
LT
2024 void *af_packet_priv;
2025 struct list_head list;
2026};
2027
f191a1d1 2028struct offload_callbacks {
576a30eb 2029 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2030 netdev_features_t features);
d565b0a1 2031 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2032 struct sk_buff *skb);
299603e8 2033 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2034};
2035
2036struct packet_offload {
2037 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2038 u16 priority;
f191a1d1
VY
2039 struct offload_callbacks callbacks;
2040 struct list_head list;
1da177e4
LT
2041};
2042
a2b12f3c
TH
2043struct udp_offload;
2044
2045struct udp_offload_callbacks {
2046 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2047 struct sk_buff *skb,
2048 struct udp_offload *uoff);
2049 int (*gro_complete)(struct sk_buff *skb,
2050 int nhoff,
2051 struct udp_offload *uoff);
2052};
2053
b582ef09
OG
2054struct udp_offload {
2055 __be16 port;
afe93325 2056 u8 ipproto;
a2b12f3c 2057 struct udp_offload_callbacks callbacks;
b582ef09
OG
2058};
2059
8f84985f
LR
2060/* often modified stats are per cpu, other are shared (netdev->stats) */
2061struct pcpu_sw_netstats {
2062 u64 rx_packets;
2063 u64 rx_bytes;
2064 u64 tx_packets;
2065 u64 tx_bytes;
2066 struct u64_stats_sync syncp;
2067};
2068
1c213bd2
WC
2069#define netdev_alloc_pcpu_stats(type) \
2070({ \
693350c2 2071 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1c213bd2 2072 if (pcpu_stats) { \
ec65aafb
JB
2073 int __cpu; \
2074 for_each_possible_cpu(__cpu) { \
1c213bd2 2075 typeof(type) *stat; \
ec65aafb 2076 stat = per_cpu_ptr(pcpu_stats, __cpu); \
1c213bd2
WC
2077 u64_stats_init(&stat->syncp); \
2078 } \
2079 } \
2080 pcpu_stats; \
2081})
2082
1da177e4
LT
2083#include <linux/notifier.h>
2084
dcfe1421
AW
2085/* netdevice notifier chain. Please remember to update the rtnetlink
2086 * notification exclusion list in rtnetlink_event() when adding new
2087 * types.
2088 */
2089#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2090#define NETDEV_DOWN 0x0002
2091#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2092 detected a hardware crash and restarted
2093 - we can use this eg to kick tcp sessions
2094 once done */
2095#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2096#define NETDEV_REGISTER 0x0005
2097#define NETDEV_UNREGISTER 0x0006
1d486bfb 2098#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2099#define NETDEV_CHANGEADDR 0x0008
2100#define NETDEV_GOING_DOWN 0x0009
2101#define NETDEV_CHANGENAME 0x000A
2102#define NETDEV_FEAT_CHANGE 0x000B
2103#define NETDEV_BONDING_FAILOVER 0x000C
2104#define NETDEV_PRE_UP 0x000D
2105#define NETDEV_PRE_TYPE_CHANGE 0x000E
2106#define NETDEV_POST_TYPE_CHANGE 0x000F
2107#define NETDEV_POST_INIT 0x0010
0115e8e3 2108#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2109#define NETDEV_RELEASE 0x0012
2110#define NETDEV_NOTIFY_PEERS 0x0013
2111#define NETDEV_JOIN 0x0014
42e52bf9 2112#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2113#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2114#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2115#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2116#define NETDEV_BONDING_INFO 0x0019
dcfe1421 2117
f629d208
JP
2118int register_netdevice_notifier(struct notifier_block *nb);
2119int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2120
2121struct netdev_notifier_info {
2122 struct net_device *dev;
2123};
2124
be9efd36
JP
2125struct netdev_notifier_change_info {
2126 struct netdev_notifier_info info; /* must be first */
2127 unsigned int flags_changed;
2128};
2129
75538c2b
CW
2130static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2131 struct net_device *dev)
2132{
2133 info->dev = dev;
2134}
2135
351638e7
JP
2136static inline struct net_device *
2137netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2138{
2139 return info->dev;
2140}
2141
f629d208 2142int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2143
2144
1da177e4
LT
2145extern rwlock_t dev_base_lock; /* Device list lock */
2146
881d966b
EB
2147#define for_each_netdev(net, d) \
2148 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2149#define for_each_netdev_reverse(net, d) \
2150 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2151#define for_each_netdev_rcu(net, d) \
2152 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2153#define for_each_netdev_safe(net, d, n) \
2154 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2155#define for_each_netdev_continue(net, d) \
2156 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2157#define for_each_netdev_continue_rcu(net, d) \
2158 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2159#define for_each_netdev_in_bond_rcu(bond, slave) \
2160 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2161 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2162#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2163
a050c33f
DL
2164static inline struct net_device *next_net_device(struct net_device *dev)
2165{
2166 struct list_head *lh;
2167 struct net *net;
2168
c346dca1 2169 net = dev_net(dev);
a050c33f
DL
2170 lh = dev->dev_list.next;
2171 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2172}
2173
ce81b76a
ED
2174static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2175{
2176 struct list_head *lh;
2177 struct net *net;
2178
2179 net = dev_net(dev);
ccf43438 2180 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2181 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2182}
2183
a050c33f
DL
2184static inline struct net_device *first_net_device(struct net *net)
2185{
2186 return list_empty(&net->dev_base_head) ? NULL :
2187 net_device_entry(net->dev_base_head.next);
2188}
7562f876 2189
ccf43438
ED
2190static inline struct net_device *first_net_device_rcu(struct net *net)
2191{
2192 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2193
2194 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2195}
2196
f629d208
JP
2197int netdev_boot_setup_check(struct net_device *dev);
2198unsigned long netdev_boot_base(const char *prefix, int unit);
2199struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2200 const char *hwaddr);
2201struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2202struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2203void dev_add_pack(struct packet_type *pt);
2204void dev_remove_pack(struct packet_type *pt);
2205void __dev_remove_pack(struct packet_type *pt);
2206void dev_add_offload(struct packet_offload *po);
2207void dev_remove_offload(struct packet_offload *po);
f629d208 2208
a54acb3a 2209int dev_get_iflink(const struct net_device *dev);
6c555490
WC
2210struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2211 unsigned short mask);
f629d208
JP
2212struct net_device *dev_get_by_name(struct net *net, const char *name);
2213struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2214struct net_device *__dev_get_by_name(struct net *net, const char *name);
2215int dev_alloc_name(struct net_device *dev, const char *name);
2216int dev_open(struct net_device *dev);
2217int dev_close(struct net_device *dev);
99c4a26a 2218int dev_close_many(struct list_head *head, bool unlink);
f629d208 2219void dev_disable_lro(struct net_device *dev);
7026b1dd
DM
2220int dev_loopback_xmit(struct sock *sk, struct sk_buff *newskb);
2221int dev_queue_xmit_sk(struct sock *sk, struct sk_buff *skb);
2222static inline int dev_queue_xmit(struct sk_buff *skb)
2223{
2224 return dev_queue_xmit_sk(skb->sk, skb);
2225}
f663dd9a 2226int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2227int register_netdevice(struct net_device *dev);
2228void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2229void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2230static inline void unregister_netdevice(struct net_device *dev)
2231{
2232 unregister_netdevice_queue(dev, NULL);
2233}
2234
f629d208
JP
2235int netdev_refcnt_read(const struct net_device *dev);
2236void free_netdev(struct net_device *dev);
74d332c1 2237void netdev_freemem(struct net_device *dev);
f629d208
JP
2238void synchronize_net(void);
2239int init_dummy_netdev(struct net_device *dev);
937f1ba5 2240
f60e5990 2241DECLARE_PER_CPU(int, xmit_recursion);
2242static inline int dev_recursion_level(void)
2243{
2244 return this_cpu_read(xmit_recursion);
2245}
2246
f629d208
JP
2247struct net_device *dev_get_by_index(struct net *net, int ifindex);
2248struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2249struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2250int netdev_get_name(struct net *net, char *name, int ifindex);
2251int dev_restart(struct net_device *dev);
f629d208 2252int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2253
2254static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2255{
2256 return NAPI_GRO_CB(skb)->data_offset;
2257}
2258
2259static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2260{
2261 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2262}
2263
2264static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2265{
2266 NAPI_GRO_CB(skb)->data_offset += len;
2267}
2268
a5b1cf28
HX
2269static inline void *skb_gro_header_fast(struct sk_buff *skb,
2270 unsigned int offset)
86911732 2271{
a5b1cf28
HX
2272 return NAPI_GRO_CB(skb)->frag0 + offset;
2273}
78a478d0 2274
a5b1cf28
HX
2275static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2276{
2277 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2278}
78a478d0 2279
a5b1cf28
HX
2280static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2281 unsigned int offset)
2282{
17dd759c
HX
2283 if (!pskb_may_pull(skb, hlen))
2284 return NULL;
2285
a5b1cf28
HX
2286 NAPI_GRO_CB(skb)->frag0 = NULL;
2287 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2288 return skb->data + offset;
86911732 2289}
1da177e4 2290
36e7b1b8
HX
2291static inline void *skb_gro_network_header(struct sk_buff *skb)
2292{
78d3fd0b
HX
2293 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2294 skb_network_offset(skb);
36e7b1b8
HX
2295}
2296
bf5a755f
JC
2297static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2298 const void *start, unsigned int len)
2299{
573e8fca 2300 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2301 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2302 csum_partial(start, len, 0));
2303}
2304
573e8fca
TH
2305/* GRO checksum functions. These are logical equivalents of the normal
2306 * checksum functions (in skbuff.h) except that they operate on the GRO
2307 * offsets and fields in sk_buff.
2308 */
2309
2310__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2311
15e2396d
TH
2312static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2313{
2314 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2315 skb_gro_offset(skb));
2316}
2317
573e8fca
TH
2318static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2319 bool zero_okay,
2320 __sum16 check)
2321{
6edec0e6
TH
2322 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2323 skb_checksum_start_offset(skb) <
2324 skb_gro_offset(skb)) &&
15e2396d 2325 !skb_at_gro_remcsum_start(skb) &&
662880f4 2326 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2327 (!zero_okay || check));
2328}
2329
2330static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2331 __wsum psum)
2332{
2333 if (NAPI_GRO_CB(skb)->csum_valid &&
2334 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2335 return 0;
2336
2337 NAPI_GRO_CB(skb)->csum = psum;
2338
2339 return __skb_gro_checksum_complete(skb);
2340}
2341
573e8fca
TH
2342static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2343{
662880f4
TH
2344 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2345 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2346 NAPI_GRO_CB(skb)->csum_cnt--;
2347 } else {
2348 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2349 * verified a new top level checksum or an encapsulated one
2350 * during GRO. This saves work if we fallback to normal path.
2351 */
2352 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2353 }
2354}
2355
2356#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2357 compute_pseudo) \
2358({ \
2359 __sum16 __ret = 0; \
2360 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2361 __ret = __skb_gro_checksum_validate_complete(skb, \
2362 compute_pseudo(skb, proto)); \
5a212329
TH
2363 if (__ret) \
2364 __skb_mark_checksum_bad(skb); \
2365 else \
573e8fca
TH
2366 skb_gro_incr_csum_unnecessary(skb); \
2367 __ret; \
2368})
2369
2370#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2371 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2372
2373#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2374 compute_pseudo) \
2375 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2376
2377#define skb_gro_checksum_simple_validate(skb) \
2378 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2379
d96535a1
TH
2380static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2381{
2382 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2383 !NAPI_GRO_CB(skb)->csum_valid);
2384}
2385
2386static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2387 __sum16 check, __wsum pseudo)
2388{
2389 NAPI_GRO_CB(skb)->csum = ~pseudo;
2390 NAPI_GRO_CB(skb)->csum_valid = 1;
2391}
2392
2393#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2394do { \
2395 if (__skb_gro_checksum_convert_check(skb)) \
2396 __skb_gro_checksum_convert(skb, check, \
2397 compute_pseudo(skb, proto)); \
2398} while (0)
2399
26c4f7da
TH
2400struct gro_remcsum {
2401 int offset;
2402 __wsum delta;
2403};
2404
2405static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2406{
846cd667 2407 grc->offset = 0;
26c4f7da
TH
2408 grc->delta = 0;
2409}
2410
dcdc8994 2411static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
26c4f7da 2412 int start, int offset,
15e2396d
TH
2413 struct gro_remcsum *grc,
2414 bool nopartial)
dcdc8994
TH
2415{
2416 __wsum delta;
2417
2418 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2419
15e2396d
TH
2420 if (!nopartial) {
2421 NAPI_GRO_CB(skb)->gro_remcsum_start =
2422 ((unsigned char *)ptr + start) - skb->head;
2423 return;
2424 }
2425
dcdc8994
TH
2426 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2427
2428 /* Adjust skb->csum since we changed the packet */
dcdc8994 2429 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da
TH
2430
2431 grc->offset = (ptr + offset) - (void *)skb->head;
2432 grc->delta = delta;
dcdc8994
TH
2433}
2434
26c4f7da
TH
2435static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2436 struct gro_remcsum *grc)
2437{
2438 if (!grc->delta)
2439 return;
2440
2441 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2442}
dcdc8994 2443
0c4e8581
SH
2444static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2445 unsigned short type,
3b04ddde 2446 const void *daddr, const void *saddr,
95c96174 2447 unsigned int len)
0c4e8581 2448{
f1ecfd5d 2449 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2450 return 0;
3b04ddde
SH
2451
2452 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2453}
2454
b95cce35
SH
2455static inline int dev_parse_header(const struct sk_buff *skb,
2456 unsigned char *haddr)
2457{
2458 const struct net_device *dev = skb->dev;
2459
1b83336b 2460 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2461 return 0;
3b04ddde 2462 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2463}
2464
1da177e4 2465typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2466int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2467static inline int unregister_gifconf(unsigned int family)
2468{
2469 return register_gifconf(family, NULL);
2470}
2471
99bbc707 2472#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2473#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2474struct sd_flow_limit {
2475 u64 count;
2476 unsigned int num_buckets;
2477 unsigned int history_head;
2478 u16 history[FLOW_LIMIT_HISTORY];
2479 u8 buckets[];
2480};
2481
2482extern int netdev_flow_limit_table_len;
2483#endif /* CONFIG_NET_FLOW_LIMIT */
2484
1da177e4 2485/*
88751275 2486 * Incoming packets are placed on per-cpu queues
1da177e4 2487 */
d94d9fee 2488struct softnet_data {
1da177e4 2489 struct list_head poll_list;
6e7676c1 2490 struct sk_buff_head process_queue;
1da177e4 2491
dee42870 2492 /* stats */
cd7b5396
DM
2493 unsigned int processed;
2494 unsigned int time_squeeze;
2495 unsigned int cpu_collision;
2496 unsigned int received_rps;
fd793d89 2497#ifdef CONFIG_RPS
88751275 2498 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2499#endif
2500#ifdef CONFIG_NET_FLOW_LIMIT
2501 struct sd_flow_limit __rcu *flow_limit;
2502#endif
2503 struct Qdisc *output_queue;
2504 struct Qdisc **output_queue_tailp;
2505 struct sk_buff *completion_queue;
88751275 2506
4cdb1e2e 2507#ifdef CONFIG_RPS
88751275 2508 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 2509 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2510 struct softnet_data *rps_ipi_next;
2511 unsigned int cpu;
fec5e652 2512 unsigned int input_queue_head;
76cc8b13 2513 unsigned int input_queue_tail;
1e94d72f 2514#endif
95c96174 2515 unsigned int dropped;
0a9627f2 2516 struct sk_buff_head input_pkt_queue;
bea3348e 2517 struct napi_struct backlog;
99bbc707 2518
1da177e4
LT
2519};
2520
76cc8b13 2521static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2522{
2523#ifdef CONFIG_RPS
76cc8b13
TH
2524 sd->input_queue_head++;
2525#endif
2526}
2527
2528static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2529 unsigned int *qtail)
2530{
2531#ifdef CONFIG_RPS
2532 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2533#endif
2534}
2535
0a9627f2 2536DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2537
f629d208 2538void __netif_schedule(struct Qdisc *q);
46e5da40 2539void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2540
fd2ea0a7
DM
2541static inline void netif_tx_schedule_all(struct net_device *dev)
2542{
2543 unsigned int i;
2544
2545 for (i = 0; i < dev->num_tx_queues; i++)
2546 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2547}
2548
d29f749e
DJ
2549static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2550{
73466498 2551 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2552}
2553
bea3348e
SH
2554/**
2555 * netif_start_queue - allow transmit
2556 * @dev: network device
2557 *
2558 * Allow upper layers to call the device hard_start_xmit routine.
2559 */
1da177e4
LT
2560static inline void netif_start_queue(struct net_device *dev)
2561{
e8a0464c 2562 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2563}
2564
fd2ea0a7
DM
2565static inline void netif_tx_start_all_queues(struct net_device *dev)
2566{
2567 unsigned int i;
2568
2569 for (i = 0; i < dev->num_tx_queues; i++) {
2570 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2571 netif_tx_start_queue(txq);
2572 }
2573}
2574
46e5da40 2575void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2576
d29f749e
DJ
2577/**
2578 * netif_wake_queue - restart transmit
2579 * @dev: network device
2580 *
2581 * Allow upper layers to call the device hard_start_xmit routine.
2582 * Used for flow control when transmit resources are available.
2583 */
79d16385
DM
2584static inline void netif_wake_queue(struct net_device *dev)
2585{
e8a0464c 2586 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2587}
2588
fd2ea0a7
DM
2589static inline void netif_tx_wake_all_queues(struct net_device *dev)
2590{
2591 unsigned int i;
2592
2593 for (i = 0; i < dev->num_tx_queues; i++) {
2594 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2595 netif_tx_wake_queue(txq);
2596 }
2597}
2598
d29f749e
DJ
2599static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2600{
73466498 2601 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2602}
2603
bea3348e
SH
2604/**
2605 * netif_stop_queue - stop transmitted packets
2606 * @dev: network device
2607 *
2608 * Stop upper layers calling the device hard_start_xmit routine.
2609 * Used for flow control when transmit resources are unavailable.
2610 */
1da177e4
LT
2611static inline void netif_stop_queue(struct net_device *dev)
2612{
e8a0464c 2613 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2614}
2615
a2029240 2616void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2617
4d29515f 2618static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2619{
73466498 2620 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2621}
2622
bea3348e
SH
2623/**
2624 * netif_queue_stopped - test if transmit queue is flowblocked
2625 * @dev: network device
2626 *
2627 * Test if transmit queue on device is currently unable to send.
2628 */
4d29515f 2629static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2630{
e8a0464c 2631 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2632}
2633
4d29515f 2634static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2635{
73466498
TH
2636 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2637}
2638
8e2f1a63
DB
2639static inline bool
2640netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2641{
2642 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2643}
2644
8e2f1a63
DB
2645static inline bool
2646netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2647{
2648 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2649}
2650
53511453
ED
2651/**
2652 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2653 * @dev_queue: pointer to transmit queue
2654 *
2655 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2656 * to give appropriate hint to the cpu.
2657 */
2658static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2659{
2660#ifdef CONFIG_BQL
2661 prefetchw(&dev_queue->dql.num_queued);
2662#endif
2663}
2664
2665/**
2666 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2667 * @dev_queue: pointer to transmit queue
2668 *
2669 * BQL enabled drivers might use this helper in their TX completion path,
2670 * to give appropriate hint to the cpu.
2671 */
2672static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2673{
2674#ifdef CONFIG_BQL
2675 prefetchw(&dev_queue->dql.limit);
2676#endif
2677}
2678
c5d67bd7
TH
2679static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2680 unsigned int bytes)
2681{
114cf580
TH
2682#ifdef CONFIG_BQL
2683 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2684
2685 if (likely(dql_avail(&dev_queue->dql) >= 0))
2686 return;
2687
2688 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2689
2690 /*
2691 * The XOFF flag must be set before checking the dql_avail below,
2692 * because in netdev_tx_completed_queue we update the dql_completed
2693 * before checking the XOFF flag.
2694 */
2695 smp_mb();
2696
2697 /* check again in case another CPU has just made room avail */
2698 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2699 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2700#endif
c5d67bd7
TH
2701}
2702
0042d0c8
FF
2703/**
2704 * netdev_sent_queue - report the number of bytes queued to hardware
2705 * @dev: network device
2706 * @bytes: number of bytes queued to the hardware device queue
2707 *
2708 * Report the number of bytes queued for sending/completion to the network
2709 * device hardware queue. @bytes should be a good approximation and should
2710 * exactly match netdev_completed_queue() @bytes
2711 */
c5d67bd7
TH
2712static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2713{
2714 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2715}
2716
2717static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2718 unsigned int pkts, unsigned int bytes)
c5d67bd7 2719{
114cf580 2720#ifdef CONFIG_BQL
b37c0fbe
AD
2721 if (unlikely(!bytes))
2722 return;
2723
2724 dql_completed(&dev_queue->dql, bytes);
2725
2726 /*
2727 * Without the memory barrier there is a small possiblity that
2728 * netdev_tx_sent_queue will miss the update and cause the queue to
2729 * be stopped forever
2730 */
2731 smp_mb();
2732
2733 if (dql_avail(&dev_queue->dql) < 0)
2734 return;
2735
2736 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2737 netif_schedule_queue(dev_queue);
114cf580 2738#endif
c5d67bd7
TH
2739}
2740
0042d0c8
FF
2741/**
2742 * netdev_completed_queue - report bytes and packets completed by device
2743 * @dev: network device
2744 * @pkts: actual number of packets sent over the medium
2745 * @bytes: actual number of bytes sent over the medium
2746 *
2747 * Report the number of bytes and packets transmitted by the network device
2748 * hardware queue over the physical medium, @bytes must exactly match the
2749 * @bytes amount passed to netdev_sent_queue()
2750 */
c5d67bd7 2751static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2752 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2753{
2754 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2755}
2756
2757static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2758{
114cf580 2759#ifdef CONFIG_BQL
5c490354 2760 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2761 dql_reset(&q->dql);
2762#endif
c5d67bd7
TH
2763}
2764
0042d0c8
FF
2765/**
2766 * netdev_reset_queue - reset the packets and bytes count of a network device
2767 * @dev_queue: network device
2768 *
2769 * Reset the bytes and packet count of a network device and clear the
2770 * software flow control OFF bit for this network device
2771 */
c5d67bd7
TH
2772static inline void netdev_reset_queue(struct net_device *dev_queue)
2773{
2774 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2775}
2776
b9507bda
DB
2777/**
2778 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2779 * @dev: network device
2780 * @queue_index: given tx queue index
2781 *
2782 * Returns 0 if given tx queue index >= number of device tx queues,
2783 * otherwise returns the originally passed tx queue index.
2784 */
2785static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2786{
2787 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2788 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2789 dev->name, queue_index,
2790 dev->real_num_tx_queues);
2791 return 0;
2792 }
2793
2794 return queue_index;
2795}
2796
bea3348e
SH
2797/**
2798 * netif_running - test if up
2799 * @dev: network device
2800 *
2801 * Test if the device has been brought up.
2802 */
4d29515f 2803static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2804{
2805 return test_bit(__LINK_STATE_START, &dev->state);
2806}
2807
f25f4e44
PWJ
2808/*
2809 * Routines to manage the subqueues on a device. We only need start
2810 * stop, and a check if it's stopped. All other device management is
2811 * done at the overall netdevice level.
2812 * Also test the device if we're multiqueue.
2813 */
bea3348e
SH
2814
2815/**
2816 * netif_start_subqueue - allow sending packets on subqueue
2817 * @dev: network device
2818 * @queue_index: sub queue index
2819 *
2820 * Start individual transmit queue of a device with multiple transmit queues.
2821 */
f25f4e44
PWJ
2822static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2823{
fd2ea0a7 2824 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2825
2826 netif_tx_start_queue(txq);
f25f4e44
PWJ
2827}
2828
bea3348e
SH
2829/**
2830 * netif_stop_subqueue - stop sending packets on subqueue
2831 * @dev: network device
2832 * @queue_index: sub queue index
2833 *
2834 * Stop individual transmit queue of a device with multiple transmit queues.
2835 */
f25f4e44
PWJ
2836static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2837{
fd2ea0a7 2838 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 2839 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2840}
2841
bea3348e
SH
2842/**
2843 * netif_subqueue_stopped - test status of subqueue
2844 * @dev: network device
2845 * @queue_index: sub queue index
2846 *
2847 * Check individual transmit queue of a device with multiple transmit queues.
2848 */
4d29515f
DM
2849static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2850 u16 queue_index)
f25f4e44 2851{
fd2ea0a7 2852 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2853
2854 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2855}
2856
4d29515f
DM
2857static inline bool netif_subqueue_stopped(const struct net_device *dev,
2858 struct sk_buff *skb)
668f895a
PE
2859{
2860 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2861}
bea3348e 2862
46e5da40 2863void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 2864
537c00de 2865#ifdef CONFIG_XPS
53af53ae 2866int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2867 u16 index);
537c00de
AD
2868#else
2869static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2870 const struct cpumask *mask,
537c00de
AD
2871 u16 index)
2872{
2873 return 0;
2874}
2875#endif
2876
5605c762
JP
2877u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2878 unsigned int num_tx_queues);
2879
a3d22a68
VZ
2880/*
2881 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2882 * as a distribution range limit for the returned value.
2883 */
2884static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 2885 struct sk_buff *skb)
a3d22a68
VZ
2886{
2887 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2888}
2889
bea3348e
SH
2890/**
2891 * netif_is_multiqueue - test if device has multiple transmit queues
2892 * @dev: network device
2893 *
2894 * Check if device has multiple transmit queues
bea3348e 2895 */
4d29515f 2896static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2897{
a02cec21 2898 return dev->num_tx_queues > 1;
f25f4e44 2899}
1da177e4 2900
f629d208 2901int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2902
a953be53 2903#ifdef CONFIG_SYSFS
f629d208 2904int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
2905#else
2906static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2907 unsigned int rxq)
2908{
2909 return 0;
2910}
2911#endif
2912
a953be53
MD
2913#ifdef CONFIG_SYSFS
2914static inline unsigned int get_netdev_rx_queue_index(
2915 struct netdev_rx_queue *queue)
2916{
2917 struct net_device *dev = queue->dev;
2918 int index = queue - dev->_rx;
2919
2920 BUG_ON(index >= dev->num_rx_queues);
2921 return index;
2922}
2923#endif
2924
16917b87 2925#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 2926int netif_get_num_default_rss_queues(void);
16917b87 2927
e6247027
ED
2928enum skb_free_reason {
2929 SKB_REASON_CONSUMED,
2930 SKB_REASON_DROPPED,
2931};
2932
2933void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2934void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 2935
e6247027
ED
2936/*
2937 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2938 * interrupt context or with hardware interrupts being disabled.
2939 * (in_irq() || irqs_disabled())
2940 *
2941 * We provide four helpers that can be used in following contexts :
2942 *
2943 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2944 * replacing kfree_skb(skb)
2945 *
2946 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2947 * Typically used in place of consume_skb(skb) in TX completion path
2948 *
2949 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2950 * replacing kfree_skb(skb)
2951 *
2952 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2953 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 2954 */
e6247027
ED
2955static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2956{
2957 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2958}
2959
2960static inline void dev_consume_skb_irq(struct sk_buff *skb)
2961{
2962 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2963}
2964
2965static inline void dev_kfree_skb_any(struct sk_buff *skb)
2966{
2967 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2968}
2969
2970static inline void dev_consume_skb_any(struct sk_buff *skb)
2971{
2972 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2973}
1da177e4 2974
f629d208
JP
2975int netif_rx(struct sk_buff *skb);
2976int netif_rx_ni(struct sk_buff *skb);
7026b1dd
DM
2977int netif_receive_skb_sk(struct sock *sk, struct sk_buff *skb);
2978static inline int netif_receive_skb(struct sk_buff *skb)
2979{
2980 return netif_receive_skb_sk(skb->sk, skb);
2981}
f629d208
JP
2982gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2983void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2984struct sk_buff *napi_get_frags(struct napi_struct *napi);
2985gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
2986struct packet_offload *gro_find_receive_by_type(__be16 type);
2987struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
2988
2989static inline void napi_free_frags(struct napi_struct *napi)
2990{
2991 kfree_skb(napi->skb);
2992 napi->skb = NULL;
2993}
2994
f629d208
JP
2995int netdev_rx_handler_register(struct net_device *dev,
2996 rx_handler_func_t *rx_handler,
2997 void *rx_handler_data);
2998void netdev_rx_handler_unregister(struct net_device *dev);
2999
3000bool dev_valid_name(const char *name);
3001int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3002int dev_ethtool(struct net *net, struct ifreq *);
3003unsigned int dev_get_flags(const struct net_device *);
3004int __dev_change_flags(struct net_device *, unsigned int flags);
3005int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3006void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3007 unsigned int gchanges);
f629d208
JP
3008int dev_change_name(struct net_device *, const char *);
3009int dev_set_alias(struct net_device *, const char *, size_t);
3010int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3011int dev_set_mtu(struct net_device *, int);
3012void dev_set_group(struct net_device *, int);
3013int dev_set_mac_address(struct net_device *, struct sockaddr *);
3014int dev_change_carrier(struct net_device *, bool new_carrier);
3015int dev_get_phys_port_id(struct net_device *dev,
02637fce 3016 struct netdev_phys_item_id *ppid);
db24a904
DA
3017int dev_get_phys_port_name(struct net_device *dev,
3018 char *name, size_t len);
d746d707 3019int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3020struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3021struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3022 struct netdev_queue *txq, int *ret);
a0265d28 3023int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3024int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1ee481fb 3025bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
1da177e4 3026
20380731 3027extern int netdev_budget;
1da177e4
LT
3028
3029/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3030void netdev_run_todo(void);
1da177e4 3031
bea3348e
SH
3032/**
3033 * dev_put - release reference to device
3034 * @dev: network device
3035 *
9ef4429b 3036 * Release reference to device to allow it to be freed.
bea3348e 3037 */
1da177e4
LT
3038static inline void dev_put(struct net_device *dev)
3039{
933393f5 3040 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3041}
3042
bea3348e
SH
3043/**
3044 * dev_hold - get reference to device
3045 * @dev: network device
3046 *
9ef4429b 3047 * Hold reference to device to keep it from being freed.
bea3348e 3048 */
15333061
SH
3049static inline void dev_hold(struct net_device *dev)
3050{
933393f5 3051 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3052}
1da177e4
LT
3053
3054/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3055 * and _off may be called from IRQ context, but it is caller
3056 * who is responsible for serialization of these calls.
b00055aa
SR
3057 *
3058 * The name carrier is inappropriate, these functions should really be
3059 * called netif_lowerlayer_*() because they represent the state of any
3060 * kind of lower layer not just hardware media.
1da177e4
LT
3061 */
3062
f629d208
JP
3063void linkwatch_init_dev(struct net_device *dev);
3064void linkwatch_fire_event(struct net_device *dev);
3065void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3066
bea3348e
SH
3067/**
3068 * netif_carrier_ok - test if carrier present
3069 * @dev: network device
3070 *
3071 * Check if carrier is present on device
3072 */
4d29515f 3073static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3074{
3075 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3076}
3077
f629d208 3078unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3079
f629d208 3080void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3081
f629d208 3082void netif_carrier_on(struct net_device *dev);
1da177e4 3083
f629d208 3084void netif_carrier_off(struct net_device *dev);
1da177e4 3085
bea3348e
SH
3086/**
3087 * netif_dormant_on - mark device as dormant.
3088 * @dev: network device
3089 *
3090 * Mark device as dormant (as per RFC2863).
3091 *
3092 * The dormant state indicates that the relevant interface is not
3093 * actually in a condition to pass packets (i.e., it is not 'up') but is
3094 * in a "pending" state, waiting for some external event. For "on-
3095 * demand" interfaces, this new state identifies the situation where the
3096 * interface is waiting for events to place it in the up state.
3097 *
3098 */
b00055aa
SR
3099static inline void netif_dormant_on(struct net_device *dev)
3100{
3101 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3102 linkwatch_fire_event(dev);
3103}
3104
bea3348e
SH
3105/**
3106 * netif_dormant_off - set device as not dormant.
3107 * @dev: network device
3108 *
3109 * Device is not in dormant state.
3110 */
b00055aa
SR
3111static inline void netif_dormant_off(struct net_device *dev)
3112{
3113 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3114 linkwatch_fire_event(dev);
3115}
3116
bea3348e
SH
3117/**
3118 * netif_dormant - test if carrier present
3119 * @dev: network device
3120 *
3121 * Check if carrier is present on device
3122 */
4d29515f 3123static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3124{
3125 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3126}
3127
3128
bea3348e
SH
3129/**
3130 * netif_oper_up - test if device is operational
3131 * @dev: network device
3132 *
3133 * Check if carrier is operational
3134 */
4d29515f 3135static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3136{
b00055aa
SR
3137 return (dev->operstate == IF_OPER_UP ||
3138 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3139}
3140
bea3348e
SH
3141/**
3142 * netif_device_present - is device available or removed
3143 * @dev: network device
3144 *
3145 * Check if device has not been removed from system.
3146 */
4d29515f 3147static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3148{
3149 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3150}
3151
f629d208 3152void netif_device_detach(struct net_device *dev);
1da177e4 3153
f629d208 3154void netif_device_attach(struct net_device *dev);
1da177e4
LT
3155
3156/*
3157 * Network interface message level settings
3158 */
1da177e4
LT
3159
3160enum {
3161 NETIF_MSG_DRV = 0x0001,
3162 NETIF_MSG_PROBE = 0x0002,
3163 NETIF_MSG_LINK = 0x0004,
3164 NETIF_MSG_TIMER = 0x0008,
3165 NETIF_MSG_IFDOWN = 0x0010,
3166 NETIF_MSG_IFUP = 0x0020,
3167 NETIF_MSG_RX_ERR = 0x0040,
3168 NETIF_MSG_TX_ERR = 0x0080,
3169 NETIF_MSG_TX_QUEUED = 0x0100,
3170 NETIF_MSG_INTR = 0x0200,
3171 NETIF_MSG_TX_DONE = 0x0400,
3172 NETIF_MSG_RX_STATUS = 0x0800,
3173 NETIF_MSG_PKTDATA = 0x1000,
3174 NETIF_MSG_HW = 0x2000,
3175 NETIF_MSG_WOL = 0x4000,
3176};
3177
3178#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3179#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3180#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3181#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3182#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3183#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3184#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3185#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3186#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3187#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3188#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3189#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3190#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3191#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3192#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3193
3194static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3195{
3196 /* use default */
3197 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3198 return default_msg_enable_bits;
3199 if (debug_value == 0) /* no output */
3200 return 0;
3201 /* set low N bits */
3202 return (1 << debug_value) - 1;
3203}
3204
c773e847 3205static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3206{
c773e847
DM
3207 spin_lock(&txq->_xmit_lock);
3208 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3209}
3210
fd2ea0a7
DM
3211static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3212{
3213 spin_lock_bh(&txq->_xmit_lock);
3214 txq->xmit_lock_owner = smp_processor_id();
3215}
3216
4d29515f 3217static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3218{
4d29515f 3219 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3220 if (likely(ok))
3221 txq->xmit_lock_owner = smp_processor_id();
3222 return ok;
3223}
3224
3225static inline void __netif_tx_unlock(struct netdev_queue *txq)
3226{
3227 txq->xmit_lock_owner = -1;
3228 spin_unlock(&txq->_xmit_lock);
3229}
3230
3231static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3232{
3233 txq->xmit_lock_owner = -1;
3234 spin_unlock_bh(&txq->_xmit_lock);
3235}
3236
08baf561
ED
3237static inline void txq_trans_update(struct netdev_queue *txq)
3238{
3239 if (txq->xmit_lock_owner != -1)
3240 txq->trans_start = jiffies;
3241}
3242
d29f749e
DJ
3243/**
3244 * netif_tx_lock - grab network device transmit lock
3245 * @dev: network device
d29f749e
DJ
3246 *
3247 * Get network device transmit lock
3248 */
22dd7495
JHS
3249static inline void netif_tx_lock(struct net_device *dev)
3250{
e8a0464c 3251 unsigned int i;
c3f26a26 3252 int cpu;
c773e847 3253
c3f26a26
DM
3254 spin_lock(&dev->tx_global_lock);
3255 cpu = smp_processor_id();
e8a0464c
DM
3256 for (i = 0; i < dev->num_tx_queues; i++) {
3257 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3258
3259 /* We are the only thread of execution doing a
3260 * freeze, but we have to grab the _xmit_lock in
3261 * order to synchronize with threads which are in
3262 * the ->hard_start_xmit() handler and already
3263 * checked the frozen bit.
3264 */
e8a0464c 3265 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3266 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3267 __netif_tx_unlock(txq);
e8a0464c 3268 }
932ff279
HX
3269}
3270
3271static inline void netif_tx_lock_bh(struct net_device *dev)
3272{
e8a0464c
DM
3273 local_bh_disable();
3274 netif_tx_lock(dev);
932ff279
HX
3275}
3276
932ff279
HX
3277static inline void netif_tx_unlock(struct net_device *dev)
3278{
e8a0464c
DM
3279 unsigned int i;
3280
3281 for (i = 0; i < dev->num_tx_queues; i++) {
3282 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3283
c3f26a26
DM
3284 /* No need to grab the _xmit_lock here. If the
3285 * queue is not stopped for another reason, we
3286 * force a schedule.
3287 */
3288 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3289 netif_schedule_queue(txq);
c3f26a26
DM
3290 }
3291 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3292}
3293
3294static inline void netif_tx_unlock_bh(struct net_device *dev)
3295{
e8a0464c
DM
3296 netif_tx_unlock(dev);
3297 local_bh_enable();
932ff279
HX
3298}
3299
c773e847 3300#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3301 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3302 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3303 } \
3304}
3305
5efeac44
EB
3306#define HARD_TX_TRYLOCK(dev, txq) \
3307 (((dev->features & NETIF_F_LLTX) == 0) ? \
3308 __netif_tx_trylock(txq) : \
3309 true )
3310
c773e847 3311#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3312 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3313 __netif_tx_unlock(txq); \
22dd7495
JHS
3314 } \
3315}
3316
1da177e4
LT
3317static inline void netif_tx_disable(struct net_device *dev)
3318{
fd2ea0a7 3319 unsigned int i;
c3f26a26 3320 int cpu;
fd2ea0a7 3321
c3f26a26
DM
3322 local_bh_disable();
3323 cpu = smp_processor_id();
fd2ea0a7
DM
3324 for (i = 0; i < dev->num_tx_queues; i++) {
3325 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3326
3327 __netif_tx_lock(txq, cpu);
fd2ea0a7 3328 netif_tx_stop_queue(txq);
c3f26a26 3329 __netif_tx_unlock(txq);
fd2ea0a7 3330 }
c3f26a26 3331 local_bh_enable();
1da177e4
LT
3332}
3333
e308a5d8
DM
3334static inline void netif_addr_lock(struct net_device *dev)
3335{
3336 spin_lock(&dev->addr_list_lock);
3337}
3338
2429f7ac
JP
3339static inline void netif_addr_lock_nested(struct net_device *dev)
3340{
25175ba5
VY
3341 int subclass = SINGLE_DEPTH_NESTING;
3342
3343 if (dev->netdev_ops->ndo_get_lock_subclass)
3344 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3345
3346 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3347}
3348
e308a5d8
DM
3349static inline void netif_addr_lock_bh(struct net_device *dev)
3350{
3351 spin_lock_bh(&dev->addr_list_lock);
3352}
3353
3354static inline void netif_addr_unlock(struct net_device *dev)
3355{
3356 spin_unlock(&dev->addr_list_lock);
3357}
3358
3359static inline void netif_addr_unlock_bh(struct net_device *dev)
3360{
3361 spin_unlock_bh(&dev->addr_list_lock);
3362}
3363
f001fde5 3364/*
31278e71 3365 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3366 * rcu_read_lock held.
3367 */
3368#define for_each_dev_addr(dev, ha) \
31278e71 3369 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3370
1da177e4
LT
3371/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3372
f629d208 3373void ether_setup(struct net_device *dev);
1da177e4
LT
3374
3375/* Support for loadable net-drivers */
f629d208 3376struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3377 unsigned char name_assign_type,
f629d208
JP
3378 void (*setup)(struct net_device *),
3379 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3380#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3381 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3382
c835a677
TG
3383#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3384 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3385 count)
36909ea4 3386
f629d208
JP
3387int register_netdev(struct net_device *dev);
3388void unregister_netdev(struct net_device *dev);
f001fde5 3389
22bedad3 3390/* General hardware address lists handling functions */
f629d208
JP
3391int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3392 struct netdev_hw_addr_list *from_list, int addr_len);
3393void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3394 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3395int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3396 struct net_device *dev,
3397 int (*sync)(struct net_device *, const unsigned char *),
3398 int (*unsync)(struct net_device *,
3399 const unsigned char *));
3400void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3401 struct net_device *dev,
3402 int (*unsync)(struct net_device *,
3403 const unsigned char *));
f629d208 3404void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3405
f001fde5 3406/* Functions used for device addresses handling */
f629d208
JP
3407int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3408 unsigned char addr_type);
3409int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3410 unsigned char addr_type);
f629d208
JP
3411void dev_addr_flush(struct net_device *dev);
3412int dev_addr_init(struct net_device *dev);
a748ee24
JP
3413
3414/* Functions used for unicast addresses handling */
f629d208
JP
3415int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3416int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3417int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3418int dev_uc_sync(struct net_device *to, struct net_device *from);
3419int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3420void dev_uc_unsync(struct net_device *to, struct net_device *from);
3421void dev_uc_flush(struct net_device *dev);
3422void dev_uc_init(struct net_device *dev);
f001fde5 3423
670e5b8e
AD
3424/**
3425 * __dev_uc_sync - Synchonize device's unicast list
3426 * @dev: device to sync
3427 * @sync: function to call if address should be added
3428 * @unsync: function to call if address should be removed
3429 *
3430 * Add newly added addresses to the interface, and release
3431 * addresses that have been deleted.
3432 **/
3433static inline int __dev_uc_sync(struct net_device *dev,
3434 int (*sync)(struct net_device *,
3435 const unsigned char *),
3436 int (*unsync)(struct net_device *,
3437 const unsigned char *))
3438{
3439 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3440}
3441
3442/**
e793c0f7 3443 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3444 * @dev: device to sync
3445 * @unsync: function to call if address should be removed
3446 *
3447 * Remove all addresses that were added to the device by dev_uc_sync().
3448 **/
3449static inline void __dev_uc_unsync(struct net_device *dev,
3450 int (*unsync)(struct net_device *,
3451 const unsigned char *))
3452{
3453 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3454}
3455
22bedad3 3456/* Functions used for multicast addresses handling */
f629d208
JP
3457int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3458int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3459int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3460int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3461int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3462int dev_mc_sync(struct net_device *to, struct net_device *from);
3463int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3464void dev_mc_unsync(struct net_device *to, struct net_device *from);
3465void dev_mc_flush(struct net_device *dev);
3466void dev_mc_init(struct net_device *dev);
f001fde5 3467
670e5b8e
AD
3468/**
3469 * __dev_mc_sync - Synchonize device's multicast list
3470 * @dev: device to sync
3471 * @sync: function to call if address should be added
3472 * @unsync: function to call if address should be removed
3473 *
3474 * Add newly added addresses to the interface, and release
3475 * addresses that have been deleted.
3476 **/
3477static inline int __dev_mc_sync(struct net_device *dev,
3478 int (*sync)(struct net_device *,
3479 const unsigned char *),
3480 int (*unsync)(struct net_device *,
3481 const unsigned char *))
3482{
3483 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3484}
3485
3486/**
e793c0f7 3487 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3488 * @dev: device to sync
3489 * @unsync: function to call if address should be removed
3490 *
3491 * Remove all addresses that were added to the device by dev_mc_sync().
3492 **/
3493static inline void __dev_mc_unsync(struct net_device *dev,
3494 int (*unsync)(struct net_device *,
3495 const unsigned char *))
3496{
3497 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3498}
3499
4417da66 3500/* Functions used for secondary unicast and multicast support */
f629d208
JP
3501void dev_set_rx_mode(struct net_device *dev);
3502void __dev_set_rx_mode(struct net_device *dev);
3503int dev_set_promiscuity(struct net_device *dev, int inc);
3504int dev_set_allmulti(struct net_device *dev, int inc);
3505void netdev_state_change(struct net_device *dev);
3506void netdev_notify_peers(struct net_device *dev);
3507void netdev_features_change(struct net_device *dev);
1da177e4 3508/* Load a device via the kmod */
f629d208
JP
3509void dev_load(struct net *net, const char *name);
3510struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3511 struct rtnl_link_stats64 *storage);
3512void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3513 const struct net_device_stats *netdev_stats);
eeda3fd6 3514
1da177e4 3515extern int netdev_max_backlog;
3b098e2d 3516extern int netdev_tstamp_prequeue;
1da177e4 3517extern int weight_p;
0a14842f 3518extern int bpf_jit_enable;
9ff162a8 3519
f629d208 3520bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3521struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3522 struct list_head **iter);
f629d208
JP
3523struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3524 struct list_head **iter);
8b5be856 3525
44a40855
VY
3526/* iterate through upper list, must be called under RCU read lock */
3527#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3528 for (iter = &(dev)->adj_list.upper, \
3529 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3530 updev; \
3531 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3532
8b5be856 3533/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3534#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3535 for (iter = &(dev)->all_adj_list.upper, \
3536 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3537 updev; \
3538 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3539
f629d208
JP
3540void *netdev_lower_get_next_private(struct net_device *dev,
3541 struct list_head **iter);
3542void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3543 struct list_head **iter);
31088a11
VF
3544
3545#define netdev_for_each_lower_private(dev, priv, iter) \
3546 for (iter = (dev)->adj_list.lower.next, \
3547 priv = netdev_lower_get_next_private(dev, &(iter)); \
3548 priv; \
3549 priv = netdev_lower_get_next_private(dev, &(iter)))
3550
3551#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3552 for (iter = &(dev)->adj_list.lower, \
3553 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3554 priv; \
3555 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3556
4085ebe8
VY
3557void *netdev_lower_get_next(struct net_device *dev,
3558 struct list_head **iter);
3559#define netdev_for_each_lower_dev(dev, ldev, iter) \
3560 for (iter = &(dev)->adj_list.lower, \
3561 ldev = netdev_lower_get_next(dev, &(iter)); \
3562 ldev; \
3563 ldev = netdev_lower_get_next(dev, &(iter)))
3564
f629d208 3565void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3566void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3567struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3568struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3569int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3570int netdev_master_upper_dev_link(struct net_device *dev,
9ff162a8 3571 struct net_device *upper_dev);
f629d208
JP
3572int netdev_master_upper_dev_link_private(struct net_device *dev,
3573 struct net_device *upper_dev,
3574 void *private);
3575void netdev_upper_dev_unlink(struct net_device *dev,
3576 struct net_device *upper_dev);
5bb025fa 3577void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3578void *netdev_lower_dev_get_private(struct net_device *dev,
3579 struct net_device *lower_dev);
960fb622
ED
3580
3581/* RSS keys are 40 or 52 bytes long */
3582#define NETDEV_RSS_KEY_LEN 52
3583extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3584void netdev_rss_key_fill(void *buffer, size_t len);
3585
4085ebe8
VY
3586int dev_get_nest_level(struct net_device *dev,
3587 bool (*type_check)(struct net_device *dev));
f629d208
JP
3588int skb_checksum_help(struct sk_buff *skb);
3589struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3590 netdev_features_t features, bool tx_path);
3591struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3592 netdev_features_t features);
12b0004d 3593
61bd3857
MS
3594struct netdev_bonding_info {
3595 ifslave slave;
3596 ifbond master;
3597};
3598
3599struct netdev_notifier_bonding_info {
3600 struct netdev_notifier_info info; /* must be first */
3601 struct netdev_bonding_info bonding_info;
3602};
3603
3604void netdev_bonding_info_change(struct net_device *dev,
3605 struct netdev_bonding_info *bonding_info);
3606
12b0004d
CW
3607static inline
3608struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3609{
3610 return __skb_gso_segment(skb, features, true);
3611}
53d6471c 3612__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3613
3614static inline bool can_checksum_protocol(netdev_features_t features,
3615 __be16 protocol)
3616{
3617 return ((features & NETIF_F_GEN_CSUM) ||
3618 ((features & NETIF_F_V4_CSUM) &&
3619 protocol == htons(ETH_P_IP)) ||
3620 ((features & NETIF_F_V6_CSUM) &&
3621 protocol == htons(ETH_P_IPV6)) ||
3622 ((features & NETIF_F_FCOE_CRC) &&
3623 protocol == htons(ETH_P_FCOE)));
3624}
12b0004d 3625
fb286bb2 3626#ifdef CONFIG_BUG
f629d208 3627void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3628#else
3629static inline void netdev_rx_csum_fault(struct net_device *dev)
3630{
3631}
3632#endif
1da177e4 3633/* rx skb timestamps */
f629d208
JP
3634void net_enable_timestamp(void);
3635void net_disable_timestamp(void);
1da177e4 3636
20380731 3637#ifdef CONFIG_PROC_FS
f629d208 3638int __init dev_proc_init(void);
900ff8c6
CW
3639#else
3640#define dev_proc_init() 0
20380731
ACM
3641#endif
3642
4798248e 3643static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3644 struct sk_buff *skb, struct net_device *dev,
3645 bool more)
4798248e 3646{
fa2dbdc2 3647 skb->xmit_more = more ? 1 : 0;
0b725a2c 3648 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3649}
3650
10b3ad8c 3651static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3652 struct netdev_queue *txq, bool more)
4798248e
DM
3653{
3654 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3655 int rc;
4798248e 3656
fa2dbdc2 3657 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3658 if (rc == NETDEV_TX_OK)
3659 txq_trans_update(txq);
3660
3661 return rc;
4798248e
DM
3662}
3663
42a2d923
LT
3664int netdev_class_create_file_ns(struct class_attribute *class_attr,
3665 const void *ns);
3666void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3667 const void *ns);
58292cbe
TH
3668
3669static inline int netdev_class_create_file(struct class_attribute *class_attr)
3670{
3671 return netdev_class_create_file_ns(class_attr, NULL);
3672}
3673
3674static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3675{
3676 netdev_class_remove_file_ns(class_attr, NULL);
3677}
b8a9787e 3678
04600794
JB
3679extern struct kobj_ns_type_operations net_ns_type_operations;
3680
f629d208 3681const char *netdev_drivername(const struct net_device *dev);
6579e57b 3682
f629d208 3683void linkwatch_run_queue(void);
20380731 3684
da08143b
MK
3685static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3686 netdev_features_t f2)
3687{
3688 if (f1 & NETIF_F_GEN_CSUM)
3689 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3690 if (f2 & NETIF_F_GEN_CSUM)
3691 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3692 f1 &= f2;
3693 if (f1 & NETIF_F_GEN_CSUM)
3694 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3695
3696 return f1;
3697}
3698
c8f44aff
MM
3699static inline netdev_features_t netdev_get_wanted_features(
3700 struct net_device *dev)
5455c699
MM
3701{
3702 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3703}
c8f44aff
MM
3704netdev_features_t netdev_increment_features(netdev_features_t all,
3705 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3706
3707/* Allow TSO being used on stacked device :
3708 * Performing the GSO segmentation before last device
3709 * is a performance improvement.
3710 */
3711static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3712 netdev_features_t mask)
3713{
3714 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3715}
3716
6cb6a27c 3717int __netdev_update_features(struct net_device *dev);
5455c699 3718void netdev_update_features(struct net_device *dev);
afe12cc8 3719void netdev_change_features(struct net_device *dev);
7f353bf2 3720
fc4a7489
PM
3721void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3722 struct net_device *dev);
3723
e38f3025
TM
3724netdev_features_t passthru_features_check(struct sk_buff *skb,
3725 struct net_device *dev,
3726 netdev_features_t features);
c1e756bf 3727netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3728
4d29515f 3729static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3730{
c8f44aff 3731 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3732
3733 /* check flags correspondence */
3734 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3735 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3736 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3737 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3738 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3739 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
3740 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3741 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3742 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3743 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3744 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3745 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
e585f236 3746 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
0345e186 3747
d6b4991a 3748 return (features & feature) == feature;
576a30eb
HX
3749}
3750
4d29515f 3751static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3752{
278b2513 3753 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3754 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3755}
3756
8b86a61d 3757static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 3758 netdev_features_t features)
7967168c 3759{
fc741216 3760 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3761 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3762 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3763}
3764
82cc1a7a
PWJ
3765static inline void netif_set_gso_max_size(struct net_device *dev,
3766 unsigned int size)
3767{
3768 dev->gso_max_size = size;
3769}
3770
7a7ffbab
WCC
3771static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3772 int pulled_hlen, u16 mac_offset,
3773 int mac_len)
3774{
3775 skb->protocol = protocol;
3776 skb->encapsulation = 1;
3777 skb_push(skb, pulled_hlen);
3778 skb_reset_transport_header(skb);
3779 skb->mac_header = mac_offset;
3780 skb->network_header = skb->mac_header + mac_len;
3781 skb->mac_len = mac_len;
3782}
3783
a6cc0cfa
JF
3784static inline bool netif_is_macvlan(struct net_device *dev)
3785{
3786 return dev->priv_flags & IFF_MACVLAN;
3787}
3788
2f33e7d5
MB
3789static inline bool netif_is_macvlan_port(struct net_device *dev)
3790{
3791 return dev->priv_flags & IFF_MACVLAN_PORT;
3792}
3793
5933fea7
MB
3794static inline bool netif_is_ipvlan(struct net_device *dev)
3795{
3796 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3797}
3798
3799static inline bool netif_is_ipvlan_port(struct net_device *dev)
3800{
3801 return dev->priv_flags & IFF_IPVLAN_MASTER;
3802}
3803
8a7fbfab 3804static inline bool netif_is_bond_master(struct net_device *dev)
3805{
3806 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3807}
3808
4d29515f 3809static inline bool netif_is_bond_slave(struct net_device *dev)
1765a575
JP
3810{
3811 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3812}
3813
3bdc0eba
BG
3814static inline bool netif_supports_nofcs(struct net_device *dev)
3815{
3816 return dev->priv_flags & IFF_SUPP_NOFCS;
3817}
3818
4e3c8992
DA
3819static inline bool netif_is_vrf(const struct net_device *dev)
3820{
3821 return dev->priv_flags & IFF_VRF_MASTER;
3822}
3823
3824static inline bool netif_index_is_vrf(struct net *net, int ifindex)
3825{
4e3c8992
DA
3826 bool rc = false;
3827
2f52bdcf
DA
3828#if IS_ENABLED(CONFIG_NET_VRF)
3829 struct net_device *dev;
3830
3831 if (ifindex == 0)
3832 return false;
3833
3834 rcu_read_lock();
3835
3836 dev = dev_get_by_index_rcu(net, ifindex);
4e3c8992
DA
3837 if (dev)
3838 rc = netif_is_vrf(dev);
3839
2f52bdcf
DA
3840 rcu_read_unlock();
3841#endif
4e3c8992
DA
3842 return rc;
3843}
3844
02875878
ED
3845/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3846static inline void netif_keep_dst(struct net_device *dev)
3847{
3848 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3849}
3850
505d4f73 3851extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 3852
571ba423
JP
3853/* Logging, debugging and troubleshooting/diagnostic helpers. */
3854
3855/* netdev_printk helpers, similar to dev_printk */
3856
3857static inline const char *netdev_name(const struct net_device *dev)
3858{
c6f854d5
VF
3859 if (!dev->name[0] || strchr(dev->name, '%'))
3860 return "(unnamed net_device)";
571ba423
JP
3861 return dev->name;
3862}
3863
ccc7f496
VF
3864static inline const char *netdev_reg_state(const struct net_device *dev)
3865{
3866 switch (dev->reg_state) {
3867 case NETREG_UNINITIALIZED: return " (uninitialized)";
3868 case NETREG_REGISTERED: return "";
3869 case NETREG_UNREGISTERING: return " (unregistering)";
3870 case NETREG_UNREGISTERED: return " (unregistered)";
3871 case NETREG_RELEASED: return " (released)";
3872 case NETREG_DUMMY: return " (dummy)";
3873 }
3874
3875 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3876 return " (unknown)";
3877}
3878
f629d208 3879__printf(3, 4)
6ea754eb
JP
3880void netdev_printk(const char *level, const struct net_device *dev,
3881 const char *format, ...);
f629d208 3882__printf(2, 3)
6ea754eb 3883void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 3884__printf(2, 3)
6ea754eb 3885void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 3886__printf(2, 3)
6ea754eb 3887void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 3888__printf(2, 3)
6ea754eb 3889void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 3890__printf(2, 3)
6ea754eb 3891void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 3892__printf(2, 3)
6ea754eb 3893void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 3894__printf(2, 3)
6ea754eb 3895void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 3896
8909c9ad
VK
3897#define MODULE_ALIAS_NETDEV(device) \
3898 MODULE_ALIAS("netdev-" device)
3899
b558c96f 3900#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
3901#define netdev_dbg(__dev, format, args...) \
3902do { \
ffa10cb4 3903 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 3904} while (0)
b558c96f
JC
3905#elif defined(DEBUG)
3906#define netdev_dbg(__dev, format, args...) \
3907 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
3908#else
3909#define netdev_dbg(__dev, format, args...) \
3910({ \
3911 if (0) \
3912 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
3913})
3914#endif
3915
3916#if defined(VERBOSE_DEBUG)
3917#define netdev_vdbg netdev_dbg
3918#else
3919
3920#define netdev_vdbg(dev, format, args...) \
3921({ \
3922 if (0) \
3923 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3924 0; \
3925})
3926#endif
3927
3928/*
3929 * netdev_WARN() acts like dev_printk(), but with the key difference
3930 * of using a WARN/WARN_ON to get the message out, including the
3931 * file/line information and a backtrace.
3932 */
3933#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
3934 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3935 netdev_reg_state(dev), ##args)
571ba423 3936
b3d95c5c
JP
3937/* netif printk helpers, similar to netdev_printk */
3938
3939#define netif_printk(priv, type, level, dev, fmt, args...) \
3940do { \
3941 if (netif_msg_##type(priv)) \
3942 netdev_printk(level, (dev), fmt, ##args); \
3943} while (0)
3944
f45f4321
JP
3945#define netif_level(level, priv, type, dev, fmt, args...) \
3946do { \
3947 if (netif_msg_##type(priv)) \
3948 netdev_##level(dev, fmt, ##args); \
3949} while (0)
3950
b3d95c5c 3951#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 3952 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 3953#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 3954 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 3955#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 3956 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 3957#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 3958 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 3959#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 3960 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 3961#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 3962 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 3963#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 3964 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 3965
0053ea9c 3966#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
3967#define netif_dbg(priv, type, netdev, format, args...) \
3968do { \
3969 if (netif_msg_##type(priv)) \
b5fb0a03 3970 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 3971} while (0)
0053ea9c
JP
3972#elif defined(DEBUG)
3973#define netif_dbg(priv, type, dev, format, args...) \
3974 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
3975#else
3976#define netif_dbg(priv, type, dev, format, args...) \
3977({ \
3978 if (0) \
3979 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3980 0; \
3981})
3982#endif
3983
3984#if defined(VERBOSE_DEBUG)
bcfcc450 3985#define netif_vdbg netif_dbg
b3d95c5c
JP
3986#else
3987#define netif_vdbg(priv, type, dev, format, args...) \
3988({ \
3989 if (0) \
a4ed89cb 3990 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
3991 0; \
3992})
3993#endif
571ba423 3994
900ff8c6
CW
3995/*
3996 * The list of packet types we will receive (as opposed to discard)
3997 * and the routines to invoke.
3998 *
3999 * Why 16. Because with 16 the only overlap we get on a hash of the
4000 * low nibble of the protocol value is RARP/SNAP/X.25.
4001 *
4002 * NOTE: That is no longer true with the addition of VLAN tags. Not
4003 * sure which should go first, but I bet it won't make much
4004 * difference if we are running VLANs. The good news is that
4005 * this protocol won't be in the list unless compiled in, so
4006 * the average user (w/out VLANs) will not be adversely affected.
4007 * --BLG
4008 *
4009 * 0800 IP
4010 * 8100 802.1Q VLAN
4011 * 0001 802.3
4012 * 0002 AX.25
4013 * 0004 802.2
4014 * 8035 RARP
4015 * 0005 SNAP
4016 * 0805 X.25
4017 * 0806 ARP
4018 * 8137 IPX
4019 * 0009 Localtalk
4020 * 86DD IPv6
4021 */
4022#define PTYPE_HASH_SIZE (16)
4023#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4024
385a154c 4025#endif /* _LINUX_NETDEVICE_H */