]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
Revert "UBUNTU: SAUCE: net: add recursion limit to GRO"
[mirror_ubuntu-artful-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
cf85d08f 44#include <net/dsa.h>
7a6b6f51 45#ifdef CONFIG_DCB
2f90b865
AD
46#include <net/dcbnl.h>
47#endif
5bc1421e 48#include <net/netprio_cgroup.h>
a050c33f 49
a59e2ecb 50#include <linux/netdev_features.h>
77162022 51#include <linux/neighbour.h>
607ca46e 52#include <uapi/linux/netdevice.h>
61bd3857 53#include <uapi/linux/if_bonding.h>
4e659d74 54#include <uapi/linux/pkt_cls.h>
a59e2ecb 55
115c1d6e 56struct netpoll_info;
313162d0 57struct device;
c1f19b51 58struct phy_device;
704232c2
JB
59/* 802.11 specific */
60struct wireless_dev;
98a18b6f
AA
61/* 802.15.4 specific */
62struct wpan_dev;
03c57747 63struct mpls_dev;
1da177e4 64
f629d208
JP
65void netdev_set_default_ethtool_ops(struct net_device *dev,
66 const struct ethtool_ops *ops);
d07d7507 67
9a1654ba
JP
68/* Backlog congestion levels */
69#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
70#define NET_RX_DROP 1 /* packet dropped */
71
572a9d7b
PM
72/*
73 * Transmit return codes: transmit return codes originate from three different
74 * namespaces:
75 *
76 * - qdisc return codes
77 * - driver transmit return codes
78 * - errno values
79 *
80 * Drivers are allowed to return any one of those in their hard_start_xmit()
81 * function. Real network devices commonly used with qdiscs should only return
82 * the driver transmit return codes though - when qdiscs are used, the actual
83 * transmission happens asynchronously, so the value is not propagated to
84 * higher layers. Virtual network devices transmit synchronously, in this case
85 * the driver transmit return codes are consumed by dev_queue_xmit(), all
86 * others are propagated to higher layers.
87 */
88
89/* qdisc ->enqueue() return codes. */
90#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
91#define NET_XMIT_DROP 0x01 /* skb dropped */
92#define NET_XMIT_CN 0x02 /* congestion notification */
93#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
94#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 95
b9df3cb8
GR
96/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
97 * indicates that the device will soon be dropping packets, or already drops
98 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 99#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
100#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101
dc1f8bf6 102/* Driver transmit return codes */
9a1654ba 103#define NETDEV_TX_MASK 0xf0
572a9d7b 104
dc1f8bf6 105enum netdev_tx {
572a9d7b 106 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
107 NETDEV_TX_OK = 0x00, /* driver took care of packet */
108 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
109 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
dc1f8bf6
SH
110};
111typedef enum netdev_tx netdev_tx_t;
112
9a1654ba
JP
113/*
114 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
115 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 */
117static inline bool dev_xmit_complete(int rc)
118{
119 /*
120 * Positive cases with an skb consumed by a driver:
121 * - successful transmission (rc == NETDEV_TX_OK)
122 * - error while transmitting (rc < 0)
123 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 */
125 if (likely(rc < NET_XMIT_MASK))
126 return true;
127
128 return false;
129}
130
1da177e4
LT
131/*
132 * Compute the worst case header length according to the protocols
133 * used.
134 */
fe2918b0 135
9a1c5125
KS
136#if defined(CONFIG_HYPERV_NET)
137# define LL_MAX_HEADER 128
138#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
139# if defined(CONFIG_MAC80211_MESH)
140# define LL_MAX_HEADER 128
141# else
142# define LL_MAX_HEADER 96
143# endif
1da177e4 144#else
8388e3da 145# define LL_MAX_HEADER 32
1da177e4
LT
146#endif
147
d11ead75
BH
148#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
149 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
150#define MAX_HEADER LL_MAX_HEADER
151#else
152#define MAX_HEADER (LL_MAX_HEADER + 48)
153#endif
154
155/*
be1f3c2c
BH
156 * Old network device statistics. Fields are native words
157 * (unsigned long) so they can be read and written atomically.
1da177e4 158 */
fe2918b0 159
d94d9fee 160struct net_device_stats {
3cfde79c
BH
161 unsigned long rx_packets;
162 unsigned long tx_packets;
163 unsigned long rx_bytes;
164 unsigned long tx_bytes;
165 unsigned long rx_errors;
166 unsigned long tx_errors;
167 unsigned long rx_dropped;
168 unsigned long tx_dropped;
169 unsigned long multicast;
1da177e4 170 unsigned long collisions;
1da177e4 171 unsigned long rx_length_errors;
3cfde79c
BH
172 unsigned long rx_over_errors;
173 unsigned long rx_crc_errors;
174 unsigned long rx_frame_errors;
175 unsigned long rx_fifo_errors;
176 unsigned long rx_missed_errors;
1da177e4
LT
177 unsigned long tx_aborted_errors;
178 unsigned long tx_carrier_errors;
179 unsigned long tx_fifo_errors;
180 unsigned long tx_heartbeat_errors;
181 unsigned long tx_window_errors;
1da177e4
LT
182 unsigned long rx_compressed;
183 unsigned long tx_compressed;
184};
185
1da177e4
LT
186
187#include <linux/cache.h>
188#include <linux/skbuff.h>
189
adc9300e 190#ifdef CONFIG_RPS
c5905afb
IM
191#include <linux/static_key.h>
192extern struct static_key rps_needed;
adc9300e
ED
193#endif
194
1da177e4
LT
195struct neighbour;
196struct neigh_parms;
197struct sk_buff;
198
f001fde5
JP
199struct netdev_hw_addr {
200 struct list_head list;
201 unsigned char addr[MAX_ADDR_LEN];
202 unsigned char type;
ccffad25
JP
203#define NETDEV_HW_ADDR_T_LAN 1
204#define NETDEV_HW_ADDR_T_SAN 2
205#define NETDEV_HW_ADDR_T_SLAVE 3
206#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 207#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 208 bool global_use;
4cd729b0 209 int sync_cnt;
8f8f103d 210 int refcount;
4543fbef 211 int synced;
f001fde5
JP
212 struct rcu_head rcu_head;
213};
214
31278e71
JP
215struct netdev_hw_addr_list {
216 struct list_head list;
217 int count;
218};
219
22bedad3
JP
220#define netdev_hw_addr_list_count(l) ((l)->count)
221#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
222#define netdev_hw_addr_list_for_each(ha, l) \
223 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 224
22bedad3
JP
225#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
226#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
227#define netdev_for_each_uc_addr(ha, dev) \
228 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 229
22bedad3
JP
230#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
231#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 232#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 233 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 234
d94d9fee 235struct hh_cache {
f6b72b62 236 u16 hh_len;
5c25f686 237 u16 __pad;
3644f0ce 238 seqlock_t hh_lock;
1da177e4
LT
239
240 /* cached hardware header; allow for machine alignment needs. */
241#define HH_DATA_MOD 16
242#define HH_DATA_OFF(__len) \
5ba0eac6 243 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
244#define HH_DATA_ALIGN(__len) \
245 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
246 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
247};
248
249/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
250 * Alternative is:
251 * dev->hard_header_len ? (dev->hard_header_len +
252 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
253 *
254 * We could use other alignment values, but we must maintain the
255 * relationship HH alignment <= LL alignment.
256 */
257#define LL_RESERVED_SPACE(dev) \
f5184d26 258 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 259#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 260 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 261
3b04ddde
SH
262struct header_ops {
263 int (*create) (struct sk_buff *skb, struct net_device *dev,
264 unsigned short type, const void *daddr,
95c96174 265 const void *saddr, unsigned int len);
3b04ddde 266 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 267 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
268 void (*cache_update)(struct hh_cache *hh,
269 const struct net_device *dev,
270 const unsigned char *haddr);
0780eb5e 271 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
272};
273
1da177e4
LT
274/* These flag bits are private to the generic network queueing
275 * layer, they may not be explicitly referenced by any other
276 * code.
277 */
278
d94d9fee 279enum netdev_state_t {
1da177e4
LT
280 __LINK_STATE_START,
281 __LINK_STATE_PRESENT,
1da177e4 282 __LINK_STATE_NOCARRIER,
b00055aa
SR
283 __LINK_STATE_LINKWATCH_PENDING,
284 __LINK_STATE_DORMANT,
1da177e4
LT
285};
286
287
288/*
289 * This structure holds at boot time configured netdevice settings. They
fe2918b0 290 * are then used in the device probing.
1da177e4
LT
291 */
292struct netdev_boot_setup {
293 char name[IFNAMSIZ];
294 struct ifmap map;
295};
296#define NETDEV_BOOT_SETUP_MAX 8
297
f629d208 298int __init netdev_boot_setup(char *str);
1da177e4 299
bea3348e
SH
300/*
301 * Structure for NAPI scheduling similar to tasklet but with weighting
302 */
303struct napi_struct {
304 /* The poll_list must only be managed by the entity which
305 * changes the state of the NAPI_STATE_SCHED bit. This means
306 * whoever atomically sets that bit can add this napi_struct
307 * to the per-cpu poll_list, and whoever clears that bit
308 * can remove from the list right before clearing the bit.
309 */
310 struct list_head poll_list;
311
312 unsigned long state;
313 int weight;
404f7c9e 314 unsigned int gro_count;
bea3348e
SH
315 int (*poll)(struct napi_struct *, int);
316#ifdef CONFIG_NETPOLL
317 spinlock_t poll_lock;
318 int poll_owner;
bea3348e 319#endif
5d38a079 320 struct net_device *dev;
d565b0a1 321 struct sk_buff *gro_list;
5d38a079 322 struct sk_buff *skb;
3b47d303 323 struct hrtimer timer;
404f7c9e 324 struct list_head dev_list;
af12fa6e
ET
325 struct hlist_node napi_hash_node;
326 unsigned int napi_id;
bea3348e
SH
327};
328
d94d9fee 329enum {
bea3348e 330 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 331 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 332 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
af12fa6e 333 NAPI_STATE_HASHED, /* In NAPI hash */
bea3348e
SH
334};
335
5b252f0c 336enum gro_result {
d1c76af9
HX
337 GRO_MERGED,
338 GRO_MERGED_FREE,
339 GRO_HELD,
340 GRO_NORMAL,
341 GRO_DROP,
342};
5b252f0c 343typedef enum gro_result gro_result_t;
d1c76af9 344
8a4eb573
JP
345/*
346 * enum rx_handler_result - Possible return values for rx_handlers.
347 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
348 * further.
349 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
350 * case skb->dev was changed by rx_handler.
351 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
352 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
353 *
354 * rx_handlers are functions called from inside __netif_receive_skb(), to do
355 * special processing of the skb, prior to delivery to protocol handlers.
356 *
357 * Currently, a net_device can only have a single rx_handler registered. Trying
358 * to register a second rx_handler will return -EBUSY.
359 *
360 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
361 * To unregister a rx_handler on a net_device, use
362 * netdev_rx_handler_unregister().
363 *
364 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
365 * do with the skb.
366 *
367 * If the rx_handler consumed to skb in some way, it should return
368 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
369 * the skb to be delivered in some other ways.
370 *
371 * If the rx_handler changed skb->dev, to divert the skb to another
372 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
373 * new device will be called if it exists.
374 *
375 * If the rx_handler consider the skb should be ignored, it should return
376 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 377 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
378 *
379 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
380 * delivered, it should return RX_HANDLER_PASS.
381 *
382 * A device without a registered rx_handler will behave as if rx_handler
383 * returned RX_HANDLER_PASS.
384 */
385
386enum rx_handler_result {
387 RX_HANDLER_CONSUMED,
388 RX_HANDLER_ANOTHER,
389 RX_HANDLER_EXACT,
390 RX_HANDLER_PASS,
391};
392typedef enum rx_handler_result rx_handler_result_t;
393typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 394
f629d208 395void __napi_schedule(struct napi_struct *n);
bc9ad166 396void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 397
4d29515f 398static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
399{
400 return test_bit(NAPI_STATE_DISABLE, &n->state);
401}
402
bea3348e
SH
403/**
404 * napi_schedule_prep - check if napi can be scheduled
405 * @n: napi context
406 *
407 * Test if NAPI routine is already running, and if not mark
408 * it as running. This is used as a condition variable
a0a46196
DM
409 * insure only one NAPI poll instance runs. We also make
410 * sure there is no pending NAPI disable.
bea3348e 411 */
4d29515f 412static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 413{
a0a46196
DM
414 return !napi_disable_pending(n) &&
415 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
416}
417
418/**
419 * napi_schedule - schedule NAPI poll
420 * @n: napi context
421 *
422 * Schedule NAPI poll routine to be called if it is not already
423 * running.
424 */
425static inline void napi_schedule(struct napi_struct *n)
426{
427 if (napi_schedule_prep(n))
428 __napi_schedule(n);
429}
430
bc9ad166
ED
431/**
432 * napi_schedule_irqoff - schedule NAPI poll
433 * @n: napi context
434 *
435 * Variant of napi_schedule(), assuming hard irqs are masked.
436 */
437static inline void napi_schedule_irqoff(struct napi_struct *n)
438{
439 if (napi_schedule_prep(n))
440 __napi_schedule_irqoff(n);
441}
442
bfe13f54 443/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 444static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
445{
446 if (napi_schedule_prep(napi)) {
447 __napi_schedule(napi);
4d29515f 448 return true;
bfe13f54 449 }
4d29515f 450 return false;
bfe13f54
RD
451}
452
3b47d303
ED
453void __napi_complete(struct napi_struct *n);
454void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
455/**
456 * napi_complete - NAPI processing complete
457 * @n: napi context
458 *
459 * Mark NAPI processing as complete.
3b47d303 460 * Consider using napi_complete_done() instead.
bea3348e 461 */
3b47d303
ED
462static inline void napi_complete(struct napi_struct *n)
463{
464 return napi_complete_done(n, 0);
465}
bea3348e 466
af12fa6e
ET
467/**
468 * napi_by_id - lookup a NAPI by napi_id
469 * @napi_id: hashed napi_id
470 *
471 * lookup @napi_id in napi_hash table
472 * must be called under rcu_read_lock()
473 */
f629d208 474struct napi_struct *napi_by_id(unsigned int napi_id);
af12fa6e
ET
475
476/**
477 * napi_hash_add - add a NAPI to global hashtable
478 * @napi: napi context
479 *
480 * generate a new napi_id and store a @napi under it in napi_hash
481 */
f629d208 482void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
483
484/**
485 * napi_hash_del - remove a NAPI from global table
486 * @napi: napi context
487 *
488 * Warning: caller must observe rcu grace period
489 * before freeing memory containing @napi
490 */
f629d208 491void napi_hash_del(struct napi_struct *napi);
af12fa6e 492
bea3348e
SH
493/**
494 * napi_disable - prevent NAPI from scheduling
495 * @n: napi context
496 *
497 * Stop NAPI from being scheduled on this context.
498 * Waits till any outstanding processing completes.
499 */
3b47d303 500void napi_disable(struct napi_struct *n);
bea3348e
SH
501
502/**
503 * napi_enable - enable NAPI scheduling
504 * @n: napi context
505 *
506 * Resume NAPI from being scheduled on this context.
507 * Must be paired with napi_disable.
508 */
509static inline void napi_enable(struct napi_struct *n)
510{
511 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 512 smp_mb__before_atomic();
bea3348e 513 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 514 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
515}
516
c264c3de
SH
517/**
518 * napi_synchronize - wait until NAPI is not running
519 * @n: napi context
520 *
521 * Wait until NAPI is done being scheduled on this context.
522 * Waits till any outstanding processing completes but
523 * does not disable future activations.
524 */
525static inline void napi_synchronize(const struct napi_struct *n)
526{
7da4387a
AB
527 if (IS_ENABLED(CONFIG_SMP))
528 while (test_bit(NAPI_STATE_SCHED, &n->state))
529 msleep(1);
530 else
531 barrier();
c264c3de 532}
c264c3de 533
d94d9fee 534enum netdev_queue_state_t {
73466498
TH
535 __QUEUE_STATE_DRV_XOFF,
536 __QUEUE_STATE_STACK_XOFF,
c3f26a26 537 __QUEUE_STATE_FROZEN,
79d16385 538};
8e2f1a63
DB
539
540#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
541#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
542#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
543
544#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
545#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
546 QUEUE_STATE_FROZEN)
547#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
548 QUEUE_STATE_FROZEN)
549
73466498
TH
550/*
551 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
552 * netif_tx_* functions below are used to manipulate this flag. The
553 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
554 * queue independently. The netif_xmit_*stopped functions below are called
555 * to check if the queue has been stopped by the driver or stack (either
556 * of the XOFF bits are set in the state). Drivers should not need to call
557 * netif_xmit*stopped functions, they should only be using netif_tx_*.
558 */
79d16385 559
bb949fbd 560struct netdev_queue {
6a321cb3
ED
561/*
562 * read mostly part
563 */
bb949fbd 564 struct net_device *dev;
46e5da40 565 struct Qdisc __rcu *qdisc;
b0e1e646 566 struct Qdisc *qdisc_sleeping;
ccf5ff69 567#ifdef CONFIG_SYSFS
1d24eb48
TH
568 struct kobject kobj;
569#endif
f2cd2d3e
ED
570#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
571 int numa_node;
572#endif
6a321cb3
ED
573/*
574 * write mostly part
575 */
576 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
577 int xmit_lock_owner;
9d21493b
ED
578 /*
579 * please use this field instead of dev->trans_start
580 */
581 unsigned long trans_start;
ccf5ff69 582
583 /*
584 * Number of TX timeouts for this queue
585 * (/sys/class/net/DEV/Q/trans_timeout)
586 */
587 unsigned long trans_timeout;
114cf580
TH
588
589 unsigned long state;
590
591#ifdef CONFIG_BQL
592 struct dql dql;
593#endif
822b3b2e 594 unsigned long tx_maxrate;
e8a0464c 595} ____cacheline_aligned_in_smp;
bb949fbd 596
f2cd2d3e
ED
597static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 return q->numa_node;
601#else
b236da69 602 return NUMA_NO_NODE;
f2cd2d3e
ED
603#endif
604}
605
606static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
607{
608#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
609 q->numa_node = node;
610#endif
611}
612
df334545 613#ifdef CONFIG_RPS
0a9627f2
TH
614/*
615 * This structure holds an RPS map which can be of variable length. The
616 * map is an array of CPUs.
617 */
618struct rps_map {
619 unsigned int len;
620 struct rcu_head rcu;
621 u16 cpus[0];
622};
60b778ce 623#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 624
fec5e652 625/*
c445477d
BH
626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627 * tail pointer for that CPU's input queue at the time of last enqueue, and
628 * a hardware filter index.
fec5e652
TH
629 */
630struct rps_dev_flow {
631 u16 cpu;
c445477d 632 u16 filter;
fec5e652
TH
633 unsigned int last_qtail;
634};
c445477d 635#define RPS_NO_FILTER 0xffff
fec5e652
TH
636
637/*
638 * The rps_dev_flow_table structure contains a table of flow mappings.
639 */
640struct rps_dev_flow_table {
641 unsigned int mask;
642 struct rcu_head rcu;
fec5e652
TH
643 struct rps_dev_flow flows[0];
644};
645#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 646 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
647
648/*
649 * The rps_sock_flow_table contains mappings of flows to the last CPU
650 * on which they were processed by the application (set in recvmsg).
567e4b79
ED
651 * Each entry is a 32bit value. Upper part is the high order bits
652 * of flow hash, lower part is cpu number.
653 * rps_cpu_mask is used to partition the space, depending on number of
654 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
656 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
657 */
658struct rps_sock_flow_table {
567e4b79 659 u32 mask;
93c1af6c
ED
660
661 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 662};
567e4b79 663#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
664
665#define RPS_NO_CPU 0xffff
666
567e4b79
ED
667extern u32 rps_cpu_mask;
668extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
669
fec5e652
TH
670static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
671 u32 hash)
672{
673 if (table && hash) {
567e4b79
ED
674 unsigned int index = hash & table->mask;
675 u32 val = hash & ~rps_cpu_mask;
fec5e652
TH
676
677 /* We only give a hint, preemption can change cpu under us */
567e4b79 678 val |= raw_smp_processor_id();
fec5e652 679
567e4b79
ED
680 if (table->ents[index] != val)
681 table->ents[index] = val;
fec5e652
TH
682 }
683}
684
c445477d 685#ifdef CONFIG_RFS_ACCEL
f629d208
JP
686bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
687 u16 filter_id);
c445477d 688#endif
a953be53 689#endif /* CONFIG_RPS */
c445477d 690
0a9627f2
TH
691/* This structure contains an instance of an RX queue. */
692struct netdev_rx_queue {
a953be53 693#ifdef CONFIG_RPS
6e3f7faf
ED
694 struct rps_map __rcu *rps_map;
695 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 696#endif
6e3f7faf 697 struct kobject kobj;
fe822240 698 struct net_device *dev;
0a9627f2 699} ____cacheline_aligned_in_smp;
a953be53
MD
700
701/*
702 * RX queue sysfs structures and functions.
703 */
704struct rx_queue_attribute {
705 struct attribute attr;
706 ssize_t (*show)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, char *buf);
708 ssize_t (*store)(struct netdev_rx_queue *queue,
709 struct rx_queue_attribute *attr, const char *buf, size_t len);
710};
d314774c 711
bf264145
TH
712#ifdef CONFIG_XPS
713/*
714 * This structure holds an XPS map which can be of variable length. The
715 * map is an array of queues.
716 */
717struct xps_map {
718 unsigned int len;
719 unsigned int alloc_len;
720 struct rcu_head rcu;
721 u16 queues[0];
722};
60b778ce 723#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
724#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
726
727/*
728 * This structure holds all XPS maps for device. Maps are indexed by CPU.
729 */
730struct xps_dev_maps {
731 struct rcu_head rcu;
a4177869 732 struct xps_map __rcu *cpu_map[0];
bf264145
TH
733};
734#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
735 (nr_cpu_ids * sizeof(struct xps_map *)))
736#endif /* CONFIG_XPS */
737
4f57c087
JF
738#define TC_MAX_QUEUE 16
739#define TC_BITMASK 15
740/* HW offloaded queuing disciplines txq count and offset maps */
741struct netdev_tc_txq {
742 u16 count;
743 u16 offset;
744};
745
68bad94e
NP
746#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
747/*
748 * This structure is to hold information about the device
749 * configured to run FCoE protocol stack.
750 */
751struct netdev_fcoe_hbainfo {
752 char manufacturer[64];
753 char serial_number[64];
754 char hardware_version[64];
755 char driver_version[64];
756 char optionrom_version[64];
757 char firmware_version[64];
758 char model[256];
759 char model_description[256];
760};
761#endif
762
02637fce 763#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 764
02637fce
JP
765/* This structure holds a unique identifier to identify some
766 * physical item (port for example) used by a netdevice.
66b52b0d 767 */
02637fce
JP
768struct netdev_phys_item_id {
769 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
770 unsigned char id_len;
771};
772
d754f98b
SF
773static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
774 struct netdev_phys_item_id *b)
775{
776 return a->id_len == b->id_len &&
777 memcmp(a->id, b->id, a->id_len) == 0;
778}
779
99932d4f
DB
780typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
781 struct sk_buff *skb);
782
2f94659a 783/* These structures hold the attributes of qdisc and classifiers
cc16f69f
JF
784 * that are being passed to the netdevice through the setup_tc op.
785 */
786enum {
787 TC_SETUP_MQPRIO,
2f94659a 788 TC_SETUP_CLSU32,
cc16f69f
JF
789};
790
2f94659a
JF
791struct tc_cls_u32_offload;
792
cc16f69f
JF
793struct tc_to_netdev {
794 unsigned int type;
795 union {
796 u8 tc;
2f94659a 797 struct tc_cls_u32_offload *cls_u32;
cc16f69f
JF
798 };
799};
800
801
d314774c
SH
802/*
803 * This structure defines the management hooks for network devices.
00829823
SH
804 * The following hooks can be defined; unless noted otherwise, they are
805 * optional and can be filled with a null pointer.
d314774c
SH
806 *
807 * int (*ndo_init)(struct net_device *dev);
808 * This function is called once when network device is registered.
809 * The network device can use this to any late stage initializaton
810 * or semantic validattion. It can fail with an error code which will
811 * be propogated back to register_netdev
812 *
813 * void (*ndo_uninit)(struct net_device *dev);
814 * This function is called when device is unregistered or when registration
815 * fails. It is not called if init fails.
816 *
817 * int (*ndo_open)(struct net_device *dev);
818 * This function is called when network device transistions to the up
819 * state.
820 *
821 * int (*ndo_stop)(struct net_device *dev);
822 * This function is called when network device transistions to the down
823 * state.
824 *
dc1f8bf6
SH
825 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
826 * struct net_device *dev);
00829823 827 * Called when a packet needs to be transmitted.
e79d8429
RR
828 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
829 * the queue before that can happen; it's for obsolete devices and weird
830 * corner cases, but the stack really does a non-trivial amount
831 * of useless work if you return NETDEV_TX_BUSY.
dc1f8bf6 832 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
833 * Required can not be NULL.
834 *
f663dd9a 835 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 836 * void *accel_priv, select_queue_fallback_t fallback);
00829823
SH
837 * Called to decide which queue to when device supports multiple
838 * transmit queues.
839 *
d314774c
SH
840 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
841 * This function is called to allow device receiver to make
842 * changes to configuration when multicast or promiscious is enabled.
843 *
844 * void (*ndo_set_rx_mode)(struct net_device *dev);
845 * This function is called device changes address list filtering.
01789349
JP
846 * If driver handles unicast address filtering, it should set
847 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
848 *
849 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
850 * This function is called when the Media Access Control address
37b607c5 851 * needs to be changed. If this interface is not defined, the
d314774c
SH
852 * mac address can not be changed.
853 *
854 * int (*ndo_validate_addr)(struct net_device *dev);
855 * Test if Media Access Control address is valid for the device.
856 *
857 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
858 * Called when a user request an ioctl which can't be handled by
859 * the generic interface code. If not defined ioctl's return
860 * not supported error code.
861 *
862 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
863 * Used to set network devices bus interface parameters. This interface
864 * is retained for legacy reason, new devices should use the bus
865 * interface (PCI) for low level management.
866 *
867 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
868 * Called when a user wants to change the Maximum Transfer Unit
869 * of a device. If not defined, any request to change MTU will
870 * will return an error.
871 *
00829823 872 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
873 * Callback uses when the transmitter has not made any progress
874 * for dev->watchdog ticks.
875 *
3cfde79c 876 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 877 * struct rtnl_link_stats64 *storage);
d308e38f 878 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 879 * Called when a user wants to get the network device usage
be1f3c2c 880 * statistics. Drivers must do one of the following:
3cfde79c
BH
881 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
882 * rtnl_link_stats64 structure passed by the caller.
82695d9b 883 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
884 * (which should normally be dev->stats) and return a pointer to
885 * it. The structure may be changed asynchronously only if each
886 * field is written atomically.
887 * 3. Update dev->stats asynchronously and atomically, and define
888 * neither operation.
d314774c 889 *
5d632cb7 890 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
891 * If device support VLAN filtering this function is called when a
892 * VLAN id is registered.
d314774c 893 *
5d632cb7 894 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
895 * If device support VLAN filtering this function is called when a
896 * VLAN id is unregistered.
d314774c
SH
897 *
898 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
899 *
900 * SR-IOV management functions.
901 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
902 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
903 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
904 * int max_tx_rate);
5f8444a3 905 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 906 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
907 * int (*ndo_get_vf_config)(struct net_device *dev,
908 * int vf, struct ifla_vf_info *ivf);
1d8faf48 909 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
910 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
911 * struct nlattr *port[]);
01a3d796
VZ
912 *
913 * Enable or disable the VF ability to query its RSS Redirection Table and
914 * Hash Key. This is needed since on some devices VF share this information
915 * with PF and querying it may adduce a theoretical security risk.
916 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 917 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
918 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
919 * Called to setup 'tc' number of traffic classes in the net device. This
920 * is always called from the stack with the rtnl lock held and netif tx
921 * queues stopped. This allows the netdevice to perform queue management
922 * safely.
c445477d 923 *
e9bce845
YZ
924 * Fiber Channel over Ethernet (FCoE) offload functions.
925 * int (*ndo_fcoe_enable)(struct net_device *dev);
926 * Called when the FCoE protocol stack wants to start using LLD for FCoE
927 * so the underlying device can perform whatever needed configuration or
928 * initialization to support acceleration of FCoE traffic.
929 *
930 * int (*ndo_fcoe_disable)(struct net_device *dev);
931 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
932 * so the underlying device can perform whatever needed clean-ups to
933 * stop supporting acceleration of FCoE traffic.
934 *
935 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
936 * struct scatterlist *sgl, unsigned int sgc);
937 * Called when the FCoE Initiator wants to initialize an I/O that
938 * is a possible candidate for Direct Data Placement (DDP). The LLD can
939 * perform necessary setup and returns 1 to indicate the device is set up
940 * successfully to perform DDP on this I/O, otherwise this returns 0.
941 *
942 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
943 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
944 * indicated by the FC exchange id 'xid', so the underlying device can
945 * clean up and reuse resources for later DDP requests.
946 *
947 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
948 * struct scatterlist *sgl, unsigned int sgc);
949 * Called when the FCoE Target wants to initialize an I/O that
950 * is a possible candidate for Direct Data Placement (DDP). The LLD can
951 * perform necessary setup and returns 1 to indicate the device is set up
952 * successfully to perform DDP on this I/O, otherwise this returns 0.
953 *
68bad94e
NP
954 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
955 * struct netdev_fcoe_hbainfo *hbainfo);
956 * Called when the FCoE Protocol stack wants information on the underlying
957 * device. This information is utilized by the FCoE protocol stack to
958 * register attributes with Fiber Channel management service as per the
959 * FC-GS Fabric Device Management Information(FDMI) specification.
960 *
e9bce845
YZ
961 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
962 * Called when the underlying device wants to override default World Wide
963 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
964 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
965 * protocol stack to use.
966 *
c445477d
BH
967 * RFS acceleration.
968 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
969 * u16 rxq_index, u32 flow_id);
970 * Set hardware filter for RFS. rxq_index is the target queue index;
971 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
972 * Return the filter ID on success, or a negative error code.
fbaec0ea 973 *
8b98a70c 974 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
975 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
976 * Called to make another netdev an underling.
977 *
978 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
979 * Called to release previously enslaved netdev.
5455c699
MM
980 *
981 * Feature/offload setting functions.
c8f44aff
MM
982 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
983 * netdev_features_t features);
5455c699
MM
984 * Adjusts the requested feature flags according to device-specific
985 * constraints, and returns the resulting flags. Must not modify
986 * the device state.
987 *
c8f44aff 988 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
989 * Called to update device configuration to new features. Passed
990 * feature set might be less than what was returned by ndo_fix_features()).
991 * Must return >0 or -errno if it changed dev->features itself.
992 *
edc7d573 993 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
994 * struct net_device *dev,
f6f6424b 995 * const unsigned char *addr, u16 vid, u16 flags)
77162022 996 * Adds an FDB entry to dev for addr.
1690be63
VY
997 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
998 * struct net_device *dev,
f6f6424b 999 * const unsigned char *addr, u16 vid)
77162022
JF
1000 * Deletes the FDB entry from dev coresponding to addr.
1001 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
1002 * struct net_device *dev, struct net_device *filter_dev,
1003 * int idx)
77162022
JF
1004 * Used to add FDB entries to dump requests. Implementers should add
1005 * entries to skb and update idx with the number of entries.
e5a55a89 1006 *
ad41faa8
ND
1007 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1008 * u16 flags)
e5a55a89 1009 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1010 * struct net_device *dev, u32 filter_mask,
1011 * int nlflags)
ad41faa8
ND
1012 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1013 * u16 flags);
4bf84c35
JP
1014 *
1015 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1016 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1017 * which do not represent real hardware may define this to allow their
1018 * userspace components to manage their virtual carrier state. Devices
1019 * that determine carrier state from physical hardware properties (eg
1020 * network cables) or protocol-dependent mechanisms (eg
1021 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1022 *
1023 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1024 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1025 * Called to get ID of physical port of this device. If driver does
1026 * not implement this, it is assumed that the hw is not able to have
1027 * multiple net devices on single physical port.
53cf5275
JG
1028 *
1029 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 1030 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1031 * Called by vxlan to notiy a driver about the UDP port and socket
1032 * address family that vxlan is listnening to. It is called only when
1033 * a new port starts listening. The operation is protected by the
1034 * vxlan_net->sock_lock.
1035 *
e6e241a1
SA
1036 * void (*ndo_add_geneve_port)(struct net_device *dev,
1037 * sa_family_t sa_family, __be16 port);
1038 * Called by geneve to notify a driver about the UDP port and socket
1039 * address family that geneve is listnening to. It is called only when
1040 * a new port starts listening. The operation is protected by the
1041 * geneve_net->sock_lock.
1042 *
1043 * void (*ndo_del_geneve_port)(struct net_device *dev,
1044 * sa_family_t sa_family, __be16 port);
1045 * Called by geneve to notify the driver about a UDP port and socket
1046 * address family that geneve is not listening to anymore. The operation
1047 * is protected by the geneve_net->sock_lock.
1048 *
53cf5275 1049 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 1050 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1051 * Called by vxlan to notify the driver about a UDP port and socket
1052 * address family that vxlan is not listening to anymore. The operation
1053 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
1054 *
1055 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1056 * struct net_device *dev)
1057 * Called by upper layer devices to accelerate switching or other
1058 * station functionality into hardware. 'pdev is the lowerdev
1059 * to use for the offload and 'dev' is the net device that will
1060 * back the offload. Returns a pointer to the private structure
1061 * the upper layer will maintain.
1062 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1063 * Called by upper layer device to delete the station created
1064 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1065 * the station and priv is the structure returned by the add
1066 * operation.
1067 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1068 * struct net_device *dev,
1069 * void *priv);
1070 * Callback to use for xmit over the accelerated station. This
1071 * is used in place of ndo_start_xmit on accelerated net
1072 * devices.
5f35227e
JG
1073 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1074 * struct net_device *dev
1075 * netdev_features_t features);
04ffcb25 1076 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1077 * performing offload operations on a given packet. This is to give
1078 * the device an opportunity to implement any restrictions that cannot
1079 * be otherwise expressed by feature flags. The check is called with
1080 * the set of features that the stack has calculated and it returns
1081 * those the driver believes to be appropriate.
822b3b2e
JF
1082 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1083 * int queue_index, u32 maxrate);
1084 * Called when a user wants to set a max-rate limitation of specific
1085 * TX queue.
a54acb3a
ND
1086 * int (*ndo_get_iflink)(const struct net_device *dev);
1087 * Called to get the iflink value of this device.
d746d707
AK
1088 * void (*ndo_change_proto_down)(struct net_device *dev,
1089 * bool proto_down);
1090 * This function is used to pass protocol port error state information
1091 * to the switch driver. The switch driver can react to the proto_down
1092 * by doing a phys down on the associated switch port.
fc4099f1
PS
1093 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1094 * This function is used to get egress tunnel information for given skb.
1095 * This is useful for retrieving outer tunnel header parameters while
1096 * sampling packet.
d746d707 1097 *
d314774c
SH
1098 */
1099struct net_device_ops {
1100 int (*ndo_init)(struct net_device *dev);
1101 void (*ndo_uninit)(struct net_device *dev);
1102 int (*ndo_open)(struct net_device *dev);
1103 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 1104 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
1105 struct net_device *dev);
1106 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1107 struct sk_buff *skb,
99932d4f
DB
1108 void *accel_priv,
1109 select_queue_fallback_t fallback);
d314774c
SH
1110 void (*ndo_change_rx_flags)(struct net_device *dev,
1111 int flags);
d314774c 1112 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1113 int (*ndo_set_mac_address)(struct net_device *dev,
1114 void *addr);
d314774c 1115 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1116 int (*ndo_do_ioctl)(struct net_device *dev,
1117 struct ifreq *ifr, int cmd);
d314774c
SH
1118 int (*ndo_set_config)(struct net_device *dev,
1119 struct ifmap *map);
00829823
SH
1120 int (*ndo_change_mtu)(struct net_device *dev,
1121 int new_mtu);
1122 int (*ndo_neigh_setup)(struct net_device *dev,
1123 struct neigh_parms *);
d314774c
SH
1124 void (*ndo_tx_timeout) (struct net_device *dev);
1125
28172739
ED
1126 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1127 struct rtnl_link_stats64 *storage);
d314774c
SH
1128 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1129
8e586137 1130 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1131 __be16 proto, u16 vid);
8e586137 1132 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1133 __be16 proto, u16 vid);
d314774c 1134#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1135 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1136 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1137 struct netpoll_info *info);
0e34e931 1138 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1139#endif
e0d1095a 1140#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1141 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1142#endif
95c26df8
WM
1143 int (*ndo_set_vf_mac)(struct net_device *dev,
1144 int queue, u8 *mac);
1145 int (*ndo_set_vf_vlan)(struct net_device *dev,
1146 int queue, u16 vlan, u8 qos);
ed616689
SC
1147 int (*ndo_set_vf_rate)(struct net_device *dev,
1148 int vf, int min_tx_rate,
1149 int max_tx_rate);
5f8444a3
GR
1150 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1151 int vf, bool setting);
dd461d6a
HS
1152 int (*ndo_set_vf_trust)(struct net_device *dev,
1153 int vf, bool setting);
95c26df8
WM
1154 int (*ndo_get_vf_config)(struct net_device *dev,
1155 int vf,
1156 struct ifla_vf_info *ivf);
1d8faf48
RE
1157 int (*ndo_set_vf_link_state)(struct net_device *dev,
1158 int vf, int link_state);
3b766cd8
EBE
1159 int (*ndo_get_vf_stats)(struct net_device *dev,
1160 int vf,
1161 struct ifla_vf_stats
1162 *vf_stats);
57b61080
SF
1163 int (*ndo_set_vf_port)(struct net_device *dev,
1164 int vf,
1165 struct nlattr *port[]);
1166 int (*ndo_get_vf_port)(struct net_device *dev,
1167 int vf, struct sk_buff *skb);
01a3d796
VZ
1168 int (*ndo_set_vf_rss_query_en)(
1169 struct net_device *dev,
1170 int vf, bool setting);
cc16f69f
JF
1171 int (*ndo_setup_tc)(struct net_device *dev,
1172 u32 handle,
1173 __be16 protocol,
1174 struct tc_to_netdev *tc);
d11ead75 1175#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1176 int (*ndo_fcoe_enable)(struct net_device *dev);
1177 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1178 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1179 u16 xid,
1180 struct scatterlist *sgl,
1181 unsigned int sgc);
1182 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1183 u16 xid);
6247e086
YZ
1184 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1185 u16 xid,
1186 struct scatterlist *sgl,
1187 unsigned int sgc);
68bad94e
NP
1188 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1189 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1190#endif
1191
d11ead75 1192#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1193#define NETDEV_FCOE_WWNN 0
1194#define NETDEV_FCOE_WWPN 1
1195 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1196 u64 *wwn, int type);
4d288d57 1197#endif
3c9c36bc 1198
c445477d
BH
1199#ifdef CONFIG_RFS_ACCEL
1200 int (*ndo_rx_flow_steer)(struct net_device *dev,
1201 const struct sk_buff *skb,
1202 u16 rxq_index,
1203 u32 flow_id);
1204#endif
fbaec0ea
JP
1205 int (*ndo_add_slave)(struct net_device *dev,
1206 struct net_device *slave_dev);
1207 int (*ndo_del_slave)(struct net_device *dev,
1208 struct net_device *slave_dev);
c8f44aff
MM
1209 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1210 netdev_features_t features);
5455c699 1211 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1212 netdev_features_t features);
da6a8fa0 1213 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1214 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1215
1216 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1217 struct nlattr *tb[],
77162022 1218 struct net_device *dev,
6b6e2725 1219 const unsigned char *addr,
f6f6424b 1220 u16 vid,
77162022
JF
1221 u16 flags);
1222 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1223 struct nlattr *tb[],
77162022 1224 struct net_device *dev,
f6f6424b
JP
1225 const unsigned char *addr,
1226 u16 vid);
77162022
JF
1227 int (*ndo_fdb_dump)(struct sk_buff *skb,
1228 struct netlink_callback *cb,
1229 struct net_device *dev,
5d5eacb3 1230 struct net_device *filter_dev,
77162022 1231 int idx);
e5a55a89
JF
1232
1233 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1234 struct nlmsghdr *nlh,
1235 u16 flags);
e5a55a89
JF
1236 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1237 u32 pid, u32 seq,
6cbdceeb 1238 struct net_device *dev,
46c264da
ND
1239 u32 filter_mask,
1240 int nlflags);
407af329 1241 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1242 struct nlmsghdr *nlh,
1243 u16 flags);
4bf84c35
JP
1244 int (*ndo_change_carrier)(struct net_device *dev,
1245 bool new_carrier);
66b52b0d 1246 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1247 struct netdev_phys_item_id *ppid);
db24a904
DA
1248 int (*ndo_get_phys_port_name)(struct net_device *dev,
1249 char *name, size_t len);
53cf5275
JG
1250 void (*ndo_add_vxlan_port)(struct net_device *dev,
1251 sa_family_t sa_family,
35e42379 1252 __be16 port);
53cf5275
JG
1253 void (*ndo_del_vxlan_port)(struct net_device *dev,
1254 sa_family_t sa_family,
35e42379 1255 __be16 port);
e6e241a1
SA
1256 void (*ndo_add_geneve_port)(struct net_device *dev,
1257 sa_family_t sa_family,
1258 __be16 port);
1259 void (*ndo_del_geneve_port)(struct net_device *dev,
1260 sa_family_t sa_family,
1261 __be16 port);
a6cc0cfa
JF
1262 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1263 struct net_device *dev);
1264 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1265 void *priv);
1266
1267 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1268 struct net_device *dev,
1269 void *priv);
25175ba5 1270 int (*ndo_get_lock_subclass)(struct net_device *dev);
5f35227e
JG
1271 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1272 struct net_device *dev,
1273 netdev_features_t features);
822b3b2e
JF
1274 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1275 int queue_index,
1276 u32 maxrate);
a54acb3a 1277 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1278 int (*ndo_change_proto_down)(struct net_device *dev,
1279 bool proto_down);
fc4099f1
PS
1280 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1281 struct sk_buff *skb);
d314774c
SH
1282};
1283
7aa98047
LR
1284/**
1285 * enum net_device_priv_flags - &struct net_device priv_flags
1286 *
1287 * These are the &struct net_device, they are only set internally
1288 * by drivers and used in the kernel. These flags are invisible to
1289 * userspace, this means that the order of these flags can change
1290 * during any kernel release.
1291 *
1292 * You should have a pretty good reason to be extending these flags.
1293 *
1294 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1295 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1296 * @IFF_BONDING: bonding master or slave
7aa98047 1297 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1298 * @IFF_WAN_HDLC: WAN HDLC device
1299 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1300 * release skb->dst
1301 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1302 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1303 * @IFF_MACVLAN_PORT: device used as macvlan port
1304 * @IFF_BRIDGE_PORT: device used as bridge port
1305 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1306 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1307 * @IFF_UNICAST_FLT: Supports unicast filtering
1308 * @IFF_TEAM_PORT: device used as team port
1309 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1310 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1311 * change when it's running
1312 * @IFF_MACVLAN: Macvlan device
007979ea 1313 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1314 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1315 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1316 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
85c5f09b 1317 * @IFF_TEAM: device is a team device
c023158d 1318 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
7aa98047
LR
1319 */
1320enum netdev_priv_flags {
1321 IFF_802_1Q_VLAN = 1<<0,
1322 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1323 IFF_BONDING = 1<<2,
1324 IFF_ISATAP = 1<<3,
1325 IFF_WAN_HDLC = 1<<4,
1326 IFF_XMIT_DST_RELEASE = 1<<5,
1327 IFF_DONT_BRIDGE = 1<<6,
1328 IFF_DISABLE_NETPOLL = 1<<7,
1329 IFF_MACVLAN_PORT = 1<<8,
1330 IFF_BRIDGE_PORT = 1<<9,
1331 IFF_OVS_DATAPATH = 1<<10,
1332 IFF_TX_SKB_SHARING = 1<<11,
1333 IFF_UNICAST_FLT = 1<<12,
1334 IFF_TEAM_PORT = 1<<13,
1335 IFF_SUPP_NOFCS = 1<<14,
1336 IFF_LIVE_ADDR_CHANGE = 1<<15,
1337 IFF_MACVLAN = 1<<16,
1338 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1339 IFF_IPVLAN_MASTER = 1<<18,
1340 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1341 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1342 IFF_NO_QUEUE = 1<<21,
1343 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1344 IFF_L3MDEV_SLAVE = 1<<23,
85c5f09b 1345 IFF_TEAM = 1<<24,
c023158d 1346 IFF_RXFH_CONFIGURED = 1<<25,
7aa98047
LR
1347};
1348
1349#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1350#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1351#define IFF_BONDING IFF_BONDING
7aa98047 1352#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1353#define IFF_WAN_HDLC IFF_WAN_HDLC
1354#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1355#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1356#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1357#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1358#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1359#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1360#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1361#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1362#define IFF_TEAM_PORT IFF_TEAM_PORT
1363#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1364#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1365#define IFF_MACVLAN IFF_MACVLAN
02875878 1366#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1367#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1368#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1369#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1370#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1371#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1372#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
85c5f09b 1373#define IFF_TEAM IFF_TEAM
c023158d 1374#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
7aa98047 1375
536721b1
KK
1376/**
1377 * struct net_device - The DEVICE structure.
1378 * Actually, this whole structure is a big mistake. It mixes I/O
1379 * data with strictly "high-level" data, and it has to know about
1380 * almost every data structure used in the INET module.
1381 *
1382 * @name: This is the first field of the "visible" part of this structure
1383 * (i.e. as seen by users in the "Space.c" file). It is the name
1384 * of the interface.
1385 *
1386 * @name_hlist: Device name hash chain, please keep it close to name[]
1387 * @ifalias: SNMP alias
1388 * @mem_end: Shared memory end
1389 * @mem_start: Shared memory start
1390 * @base_addr: Device I/O address
1391 * @irq: Device IRQ number
1392 *
14ffbbb8
TG
1393 * @carrier_changes: Stats to monitor carrier on<->off transitions
1394 *
536721b1
KK
1395 * @state: Generic network queuing layer state, see netdev_state_t
1396 * @dev_list: The global list of network devices
1397 * @napi_list: List entry, that is used for polling napi devices
1398 * @unreg_list: List entry, that is used, when we are unregistering the
1399 * device, see the function unregister_netdev
1400 * @close_list: List entry, that is used, when we are closing the device
1401 *
1402 * @adj_list: Directly linked devices, like slaves for bonding
1403 * @all_adj_list: All linked devices, *including* neighbours
1404 * @features: Currently active device features
1405 * @hw_features: User-changeable features
1406 *
1407 * @wanted_features: User-requested features
1408 * @vlan_features: Mask of features inheritable by VLAN devices
1409 *
1410 * @hw_enc_features: Mask of features inherited by encapsulating devices
1411 * This field indicates what encapsulation
1412 * offloads the hardware is capable of doing,
1413 * and drivers will need to set them appropriately.
1414 *
1415 * @mpls_features: Mask of features inheritable by MPLS
1416 *
1417 * @ifindex: interface index
388069d3 1418 * @group: The group, that the device belongs to
536721b1
KK
1419 *
1420 * @stats: Statistics struct, which was left as a legacy, use
1421 * rtnl_link_stats64 instead
1422 *
1423 * @rx_dropped: Dropped packets by core network,
1424 * do not use this in drivers
1425 * @tx_dropped: Dropped packets by core network,
1426 * do not use this in drivers
1427 *
536721b1
KK
1428 * @wireless_handlers: List of functions to handle Wireless Extensions,
1429 * instead of ioctl,
1430 * see <net/iw_handler.h> for details.
1431 * @wireless_data: Instance data managed by the core of wireless extensions
1432 *
1433 * @netdev_ops: Includes several pointers to callbacks,
1434 * if one wants to override the ndo_*() functions
1435 * @ethtool_ops: Management operations
d476059e 1436 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1437 * of Layer 2 headers.
1438 *
1439 * @flags: Interface flags (a la BSD)
1440 * @priv_flags: Like 'flags' but invisible to userspace,
1441 * see if.h for the definitions
1442 * @gflags: Global flags ( kept as legacy )
1443 * @padded: How much padding added by alloc_netdev()
1444 * @operstate: RFC2863 operstate
1445 * @link_mode: Mapping policy to operstate
1446 * @if_port: Selectable AUI, TP, ...
1447 * @dma: DMA channel
1448 * @mtu: Interface MTU value
1449 * @type: Interface hardware type
0780eb5e 1450 * @hard_header_len: Maximum hardware header length.
536721b1
KK
1451 *
1452 * @needed_headroom: Extra headroom the hardware may need, but not in all
1453 * cases can this be guaranteed
1454 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1455 * cases can this be guaranteed. Some cases also use
1456 * LL_MAX_HEADER instead to allocate the skb
1457 *
1458 * interface address info:
1459 *
1460 * @perm_addr: Permanent hw address
1461 * @addr_assign_type: Hw address assignment type
1462 * @addr_len: Hardware address length
1463 * @neigh_priv_len; Used in neigh_alloc(),
1464 * initialized only in atm/clip.c
1465 * @dev_id: Used to differentiate devices that share
1466 * the same link layer address
1467 * @dev_port: Used to differentiate devices that share
1468 * the same function
1469 * @addr_list_lock: XXX: need comments on this one
536721b1
KK
1470 * @uc_promisc: Counter, that indicates, that promiscuous mode
1471 * has been enabled due to the need to listen to
1472 * additional unicast addresses in a device that
1473 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1474 * @uc: unicast mac addresses
1475 * @mc: multicast mac addresses
1476 * @dev_addrs: list of device hw addresses
1477 * @queues_kset: Group of all Kobjects in the Tx and RX queues
536721b1
KK
1478 * @promiscuity: Number of times, the NIC is told to work in
1479 * Promiscuous mode, if it becomes 0 the NIC will
1480 * exit from working in Promiscuous mode
1481 * @allmulti: Counter, enables or disables allmulticast mode
1482 *
1483 * @vlan_info: VLAN info
1484 * @dsa_ptr: dsa specific data
1485 * @tipc_ptr: TIPC specific data
1486 * @atalk_ptr: AppleTalk link
1487 * @ip_ptr: IPv4 specific data
1488 * @dn_ptr: DECnet specific data
1489 * @ip6_ptr: IPv6 specific data
1490 * @ax25_ptr: AX.25 specific data
1491 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1492 *
1493 * @last_rx: Time of last Rx
1494 * @dev_addr: Hw address (before bcast,
1495 * because most packets are unicast)
1496 *
1497 * @_rx: Array of RX queues
1498 * @num_rx_queues: Number of RX queues
1499 * allocated at register_netdev() time
1500 * @real_num_rx_queues: Number of RX queues currently active in device
1501 *
1502 * @rx_handler: handler for received packets
1503 * @rx_handler_data: XXX: need comments on this one
1504 * @ingress_queue: XXX: need comments on this one
1505 * @broadcast: hw bcast address
1506 *
14ffbbb8
TG
1507 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1508 * indexed by RX queue number. Assigned by driver.
1509 * This must only be set if the ndo_rx_flow_steer
1510 * operation is defined
1511 * @index_hlist: Device index hash chain
1512 *
536721b1
KK
1513 * @_tx: Array of TX queues
1514 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1515 * @real_num_tx_queues: Number of TX queues currently active in device
1516 * @qdisc: Root qdisc from userspace point of view
1517 * @tx_queue_len: Max frames per queue allowed
1518 * @tx_global_lock: XXX: need comments on this one
1519 *
1520 * @xps_maps: XXX: need comments on this one
1521 *
0c4f691f
SF
1522 * @offload_fwd_mark: Offload device fwding mark
1523 *
536721b1
KK
1524 * @trans_start: Time (in jiffies) of last Tx
1525 * @watchdog_timeo: Represents the timeout that is used by
1526 * the watchdog ( see dev_watchdog() )
1527 * @watchdog_timer: List of timers
1528 *
1529 * @pcpu_refcnt: Number of references to this device
1530 * @todo_list: Delayed register/unregister
536721b1
KK
1531 * @link_watch_list: XXX: need comments on this one
1532 *
1533 * @reg_state: Register/unregister state machine
1534 * @dismantle: Device is going to be freed
1535 * @rtnl_link_state: This enum represents the phases of creating
1536 * a new link
1537 *
1538 * @destructor: Called from unregister,
1539 * can be used to call free_netdev
1540 * @npinfo: XXX: need comments on this one
1541 * @nd_net: Network namespace this network device is inside
1542 *
1543 * @ml_priv: Mid-layer private
1544 * @lstats: Loopback statistics
1545 * @tstats: Tunnel statistics
1546 * @dstats: Dummy statistics
1547 * @vstats: Virtual ethernet statistics
1548 *
1549 * @garp_port: GARP
1550 * @mrp_port: MRP
1551 *
1552 * @dev: Class/net/name entry
1553 * @sysfs_groups: Space for optional device, statistics and wireless
1554 * sysfs groups
1555 *
1556 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1557 * @rtnl_link_ops: Rtnl_link_ops
1558 *
1559 * @gso_max_size: Maximum size of generic segmentation offload
1560 * @gso_max_segs: Maximum number of segments that can be passed to the
1561 * NIC for GSO
fcbeb976
ED
1562 * @gso_min_segs: Minimum number of segments that can be passed to the
1563 * NIC for GSO
536721b1
KK
1564 *
1565 * @dcbnl_ops: Data Center Bridging netlink ops
1566 * @num_tc: Number of traffic classes in the net device
1567 * @tc_to_txq: XXX: need comments on this one
1568 * @prio_tc_map XXX: need comments on this one
1569 *
1570 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1571 *
1572 * @priomap: XXX: need comments on this one
1573 * @phydev: Physical device may attach itself
1574 * for hardware timestamping
1575 *
1576 * @qdisc_tx_busylock: XXX: need comments on this one
1577 *
d746d707
AK
1578 * @proto_down: protocol port state information can be sent to the
1579 * switch driver and used to set the phys state of the
1580 * switch port.
1581 *
1da177e4
LT
1582 * FIXME: cleanup struct net_device such that network protocol info
1583 * moves out.
1584 */
1585
d94d9fee 1586struct net_device {
1da177e4 1587 char name[IFNAMSIZ];
9356b8fc 1588 struct hlist_node name_hlist;
0b815a1a 1589 char *ifalias;
1da177e4
LT
1590 /*
1591 * I/O specific fields
1592 * FIXME: Merge these and struct ifmap into one
1593 */
536721b1
KK
1594 unsigned long mem_end;
1595 unsigned long mem_start;
1596 unsigned long base_addr;
1597 int irq;
1da177e4 1598
14ffbbb8
TG
1599 atomic_t carrier_changes;
1600
1da177e4 1601 /*
536721b1
KK
1602 * Some hardware also needs these fields (state,dev_list,
1603 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1604 * part of the usual set specified in Space.c.
1605 */
1606
1da177e4
LT
1607 unsigned long state;
1608
7562f876 1609 struct list_head dev_list;
bea3348e 1610 struct list_head napi_list;
44a0873d 1611 struct list_head unreg_list;
5cde2829 1612 struct list_head close_list;
7866a621
SN
1613 struct list_head ptype_all;
1614 struct list_head ptype_specific;
2f268f12 1615
2f268f12
VF
1616 struct {
1617 struct list_head upper;
1618 struct list_head lower;
1619 } adj_list;
1620
2f268f12
VF
1621 struct {
1622 struct list_head upper;
1623 struct list_head lower;
1624 } all_adj_list;
4c3d5e7b 1625
c8f44aff 1626 netdev_features_t features;
c8f44aff 1627 netdev_features_t hw_features;
c8f44aff 1628 netdev_features_t wanted_features;
c8f44aff 1629 netdev_features_t vlan_features;
6a674e9c 1630 netdev_features_t hw_enc_features;
0d89d203 1631 netdev_features_t mpls_features;
04ed3e74 1632
1da177e4 1633 int ifindex;
7a66bbc9 1634 int group;
1da177e4 1635
c45d286e 1636 struct net_device_stats stats;
015f0688 1637
015f0688
ED
1638 atomic_long_t rx_dropped;
1639 atomic_long_t tx_dropped;
1da177e4 1640
b86e0280 1641#ifdef CONFIG_WIRELESS_EXT
1da177e4 1642 const struct iw_handler_def * wireless_handlers;
1da177e4 1643 struct iw_public_data * wireless_data;
b86e0280 1644#endif
d314774c 1645 const struct net_device_ops *netdev_ops;
76fd8593 1646 const struct ethtool_ops *ethtool_ops;
4170604f 1647#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1648 const struct switchdev_ops *switchdev_ops;
4170604f 1649#endif
1b69c6d0
DA
1650#ifdef CONFIG_NET_L3_MASTER_DEV
1651 const struct l3mdev_ops *l3mdev_ops;
1652#endif
1da177e4 1653
3b04ddde
SH
1654 const struct header_ops *header_ops;
1655
536721b1
KK
1656 unsigned int flags;
1657 unsigned int priv_flags;
1658
1da177e4 1659 unsigned short gflags;
536721b1 1660 unsigned short padded;
1da177e4 1661
536721b1
KK
1662 unsigned char operstate;
1663 unsigned char link_mode;
b00055aa 1664
536721b1
KK
1665 unsigned char if_port;
1666 unsigned char dma;
bdc220da 1667
536721b1
KK
1668 unsigned int mtu;
1669 unsigned short type;
1670 unsigned short hard_header_len;
1da177e4 1671
f5184d26
JB
1672 unsigned short needed_headroom;
1673 unsigned short needed_tailroom;
1674
1da177e4 1675 /* Interface address info. */
536721b1
KK
1676 unsigned char perm_addr[MAX_ADDR_LEN];
1677 unsigned char addr_assign_type;
1678 unsigned char addr_len;
a0a9663d 1679 unsigned short neigh_priv_len;
536721b1
KK
1680 unsigned short dev_id;
1681 unsigned short dev_port;
ccffad25 1682 spinlock_t addr_list_lock;
14ffbbb8
TG
1683 unsigned char name_assign_type;
1684 bool uc_promisc;
536721b1
KK
1685 struct netdev_hw_addr_list uc;
1686 struct netdev_hw_addr_list mc;
1687 struct netdev_hw_addr_list dev_addrs;
1688
4c3d5e7b
ED
1689#ifdef CONFIG_SYSFS
1690 struct kset *queues_kset;
1691#endif
9d45abe1
WC
1692 unsigned int promiscuity;
1693 unsigned int allmulti;
1da177e4 1694
1da177e4
LT
1695
1696 /* Protocol specific pointers */
65ac6a5f 1697
d11ead75 1698#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1699 struct vlan_info __rcu *vlan_info;
65ac6a5f 1700#endif
34a430d7 1701#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1702 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1703#endif
1704#if IS_ENABLED(CONFIG_TIPC)
536721b1 1705 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1706#endif
536721b1
KK
1707 void *atalk_ptr;
1708 struct in_device __rcu *ip_ptr;
1709 struct dn_dev __rcu *dn_ptr;
1710 struct inet6_dev __rcu *ip6_ptr;
1711 void *ax25_ptr;
1712 struct wireless_dev *ieee80211_ptr;
98a18b6f 1713 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1714#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1715 struct mpls_dev __rcu *mpls_ptr;
1716#endif
1da177e4 1717
9356b8fc 1718/*
cd13539b 1719 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1720 */
536721b1 1721 unsigned long last_rx;
4dc89133 1722
9356b8fc 1723 /* Interface address info used in eth_type_trans() */
536721b1 1724 unsigned char *dev_addr;
f001fde5 1725
0a9627f2 1726
a953be53 1727#ifdef CONFIG_SYSFS
0a9627f2
TH
1728 struct netdev_rx_queue *_rx;
1729
0a9627f2 1730 unsigned int num_rx_queues;
62fe0b40 1731 unsigned int real_num_rx_queues;
c445477d 1732
df334545 1733#endif
0a9627f2 1734
3b47d303 1735 unsigned long gro_flush_timeout;
61391cde 1736 rx_handler_func_t __rcu *rx_handler;
1737 void __rcu *rx_handler_data;
e8a0464c 1738
4cda01e8 1739#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1740 struct tcf_proto __rcu *ingress_cl_list;
1741#endif
24824a09 1742 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1743#ifdef CONFIG_NETFILTER_INGRESS
1744 struct list_head nf_hooks_ingress;
1745#endif
d2788d34 1746
536721b1 1747 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1748#ifdef CONFIG_RFS_ACCEL
1749 struct cpu_rmap *rx_cpu_rmap;
1750#endif
1751 struct hlist_node index_hlist;
cd13539b
ED
1752
1753/*
1754 * Cache lines mostly used on transmit path
1755 */
e8a0464c
DM
1756 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1757 unsigned int num_tx_queues;
fd2ea0a7 1758 unsigned int real_num_tx_queues;
af356afa 1759 struct Qdisc *qdisc;
536721b1 1760 unsigned long tx_queue_len;
c3f26a26 1761 spinlock_t tx_global_lock;
14ffbbb8 1762 int watchdog_timeo;
cd13539b 1763
bf264145 1764#ifdef CONFIG_XPS
a4177869 1765 struct xps_dev_maps __rcu *xps_maps;
bf264145 1766#endif
1d24eb48 1767
0c4f691f
SF
1768#ifdef CONFIG_NET_SWITCHDEV
1769 u32 offload_fwd_mark;
1770#endif
1771
9356b8fc 1772 /* These may be needed for future network-power-down code. */
9d21493b
ED
1773
1774 /*
1775 * trans_start here is expensive for high speed devices on SMP,
1776 * please use netdev_queue->trans_start instead.
1777 */
536721b1 1778 unsigned long trans_start;
9356b8fc 1779
9356b8fc
ED
1780 struct timer_list watchdog_timer;
1781
29b4433d 1782 int __percpu *pcpu_refcnt;
1da177e4 1783 struct list_head todo_list;
1da177e4 1784
e014debe 1785 struct list_head link_watch_list;
572a103d 1786
1da177e4 1787 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1788 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1789 NETREG_UNREGISTERING, /* called unregister_netdevice */
1790 NETREG_UNREGISTERED, /* completed unregister todo */
1791 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1792 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1793 } reg_state:8;
1794
536721b1 1795 bool dismantle;
a2835763
PM
1796
1797 enum {
1798 RTNL_LINK_INITIALIZED,
1799 RTNL_LINK_INITIALIZING,
1800 } rtnl_link_state:16;
1da177e4 1801
d314774c 1802 void (*destructor)(struct net_device *dev);
1da177e4 1803
1da177e4 1804#ifdef CONFIG_NETPOLL
5fbee843 1805 struct netpoll_info __rcu *npinfo;
1da177e4 1806#endif
eae792b7 1807
0c5c9fb5 1808 possible_net_t nd_net;
4a1c5371 1809
4951704b 1810 /* mid-layer private */
a7855c78 1811 union {
536721b1
KK
1812 void *ml_priv;
1813 struct pcpu_lstats __percpu *lstats;
8f84985f 1814 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1815 struct pcpu_dstats __percpu *dstats;
1816 struct pcpu_vstats __percpu *vstats;
a7855c78 1817 };
536721b1 1818
3cc77ec7 1819 struct garp_port __rcu *garp_port;
febf018d 1820 struct mrp_port __rcu *mrp_port;
1da177e4 1821
536721b1 1822 struct device dev;
0c509a6c 1823 const struct attribute_group *sysfs_groups[4];
a953be53 1824 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1825
38f7b870 1826 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1827
82cc1a7a
PWJ
1828 /* for setting kernel sock attribute on TCP connection setup */
1829#define GSO_MAX_SIZE 65536
1830 unsigned int gso_max_size;
30b678d8
BH
1831#define GSO_MAX_SEGS 65535
1832 u16 gso_max_segs;
fcbeb976 1833 u16 gso_min_segs;
7a6b6f51 1834#ifdef CONFIG_DCB
32953543 1835 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1836#endif
4f57c087
JF
1837 u8 num_tc;
1838 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1839 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1840
d11ead75 1841#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1842 unsigned int fcoe_ddp_xid;
5bc1421e 1843#endif
86f8515f 1844#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1845 struct netprio_map __rcu *priomap;
4d288d57 1846#endif
c1f19b51 1847 struct phy_device *phydev;
23d3b8bf 1848 struct lock_class_key *qdisc_tx_busylock;
d746d707 1849 bool proto_down;
1da177e4 1850};
43cb76d9 1851#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1852
1853#define NETDEV_ALIGN 32
1da177e4 1854
4f57c087
JF
1855static inline
1856int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1857{
1858 return dev->prio_tc_map[prio & TC_BITMASK];
1859}
1860
1861static inline
1862int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1863{
1864 if (tc >= dev->num_tc)
1865 return -EINVAL;
1866
1867 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1868 return 0;
1869}
1870
1871static inline
1872void netdev_reset_tc(struct net_device *dev)
1873{
1874 dev->num_tc = 0;
1875 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1876 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1877}
1878
1879static inline
1880int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1881{
1882 if (tc >= dev->num_tc)
1883 return -EINVAL;
1884
1885 dev->tc_to_txq[tc].count = count;
1886 dev->tc_to_txq[tc].offset = offset;
1887 return 0;
1888}
1889
1890static inline
1891int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1892{
1893 if (num_tc > TC_MAX_QUEUE)
1894 return -EINVAL;
1895
1896 dev->num_tc = num_tc;
1897 return 0;
1898}
1899
1900static inline
1901int netdev_get_num_tc(struct net_device *dev)
1902{
1903 return dev->num_tc;
1904}
1905
e8a0464c
DM
1906static inline
1907struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1908 unsigned int index)
1909{
1910 return &dev->_tx[index];
1911}
1912
10c51b56
DB
1913static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1914 const struct sk_buff *skb)
1915{
1916 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1917}
1918
e8a0464c
DM
1919static inline void netdev_for_each_tx_queue(struct net_device *dev,
1920 void (*f)(struct net_device *,
1921 struct netdev_queue *,
1922 void *),
1923 void *arg)
1924{
1925 unsigned int i;
1926
1927 for (i = 0; i < dev->num_tx_queues; i++)
1928 f(dev, &dev->_tx[i], arg);
1929}
1930
f629d208 1931struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1932 struct sk_buff *skb,
1933 void *accel_priv);
8c4c49df 1934
c346dca1
YH
1935/*
1936 * Net namespace inlines
1937 */
1938static inline
1939struct net *dev_net(const struct net_device *dev)
1940{
c2d9ba9b 1941 return read_pnet(&dev->nd_net);
c346dca1
YH
1942}
1943
1944static inline
f5aa23fd 1945void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 1946{
0c5c9fb5 1947 write_pnet(&dev->nd_net, net);
c346dca1
YH
1948}
1949
3e8a72d1 1950static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 1951{
3fc88677 1952#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
1953 if (dev->dsa_ptr != NULL)
1954 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 1955#endif
5aed85ce 1956 return false;
396138f0
LB
1957}
1958
bea3348e
SH
1959/**
1960 * netdev_priv - access network device private data
1961 * @dev: network device
1962 *
1963 * Get network device private data
1964 */
6472ce60 1965static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1966{
1ce8e7b5 1967 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1968}
1969
1da177e4
LT
1970/* Set the sysfs physical device reference for the network logical device
1971 * if set prior to registration will cause a symlink during initialization.
1972 */
43cb76d9 1973#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1974
384912ed 1975/* Set the sysfs device type for the network logical device to allow
3f79410c 1976 * fine-grained identification of different network device types. For
384912ed
MH
1977 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1978 */
1979#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1980
82dc3c63
ED
1981/* Default NAPI poll() weight
1982 * Device drivers are strongly advised to not use bigger value
1983 */
1984#define NAPI_POLL_WEIGHT 64
1985
3b582cc1
SH
1986/**
1987 * netif_napi_add - initialize a napi context
1988 * @dev: network device
1989 * @napi: napi context
1990 * @poll: polling function
1991 * @weight: default weight
1992 *
1993 * netif_napi_add() must be used to initialize a napi context prior to calling
1994 * *any* of the other napi related functions.
1995 */
d565b0a1
HX
1996void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1997 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1998
d8156534
AD
1999/**
2000 * netif_napi_del - remove a napi context
2001 * @napi: napi context
2002 *
2003 * netif_napi_del() removes a napi context from the network device napi list
2004 */
d565b0a1
HX
2005void netif_napi_del(struct napi_struct *napi);
2006
2007struct napi_gro_cb {
78a478d0
HX
2008 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2009 void *frag0;
2010
7489594c
HX
2011 /* Length of frag0. */
2012 unsigned int frag0_len;
2013
86911732
HX
2014 /* This indicates where we are processing relative to skb->data. */
2015 int data_offset;
2016
d565b0a1 2017 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2018 u16 flush;
2019
2020 /* Save the IP ID here and check when we get to the transport layer */
2021 u16 flush_id;
d565b0a1
HX
2022
2023 /* Number of segments aggregated. */
2e71a6f8
ED
2024 u16 count;
2025
15e2396d
TH
2026 /* Start offset for remote checksum offload */
2027 u16 gro_remcsum_start;
2028
2e71a6f8
ED
2029 /* jiffies when first packet was created/queued */
2030 unsigned long age;
86347245 2031
afe93325 2032 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2033 u16 proto;
2034
baa32ff4
TH
2035 /* This is non-zero if the packet may be of the same flow. */
2036 u8 same_flow:1;
2037
f062e9b9
JG
2038 /* Used in tunnel GRO receive */
2039 u8 encap_mark:1;
573e8fca
TH
2040
2041 /* GRO checksum is valid */
2042 u8 csum_valid:1;
2043
662880f4
TH
2044 /* Number of checksums via CHECKSUM_UNNECESSARY */
2045 u8 csum_cnt:3;
c3c7c254 2046
baa32ff4
TH
2047 /* Free the skb? */
2048 u8 free:2;
2049#define NAPI_GRO_FREE 1
2050#define NAPI_GRO_FREE_STOLEN_HEAD 2
2051
efc98d08
TH
2052 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2053 u8 is_ipv6:1;
2054
8ec6feef 2055 /* 7 bit hole */
baa32ff4 2056
bf5a755f
JC
2057 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2058 __wsum csum;
2059
c3c7c254
ED
2060 /* used in skb_gro_receive() slow path */
2061 struct sk_buff *last;
d565b0a1
HX
2062};
2063
2064#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2065
1da177e4 2066struct packet_type {
f2ccd8fa
DM
2067 __be16 type; /* This is really htons(ether_type). */
2068 struct net_device *dev; /* NULL is wildcarded here */
2069 int (*func) (struct sk_buff *,
2070 struct net_device *,
2071 struct packet_type *,
2072 struct net_device *);
c0de08d0
EL
2073 bool (*id_match)(struct packet_type *ptype,
2074 struct sock *sk);
1da177e4
LT
2075 void *af_packet_priv;
2076 struct list_head list;
2077};
2078
f191a1d1 2079struct offload_callbacks {
576a30eb 2080 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2081 netdev_features_t features);
d565b0a1 2082 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2083 struct sk_buff *skb);
299603e8 2084 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2085};
2086
2087struct packet_offload {
2088 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2089 u16 priority;
f191a1d1
VY
2090 struct offload_callbacks callbacks;
2091 struct list_head list;
1da177e4
LT
2092};
2093
a2b12f3c
TH
2094struct udp_offload;
2095
2096struct udp_offload_callbacks {
2097 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2098 struct sk_buff *skb,
2099 struct udp_offload *uoff);
2100 int (*gro_complete)(struct sk_buff *skb,
2101 int nhoff,
2102 struct udp_offload *uoff);
2103};
2104
b582ef09
OG
2105struct udp_offload {
2106 __be16 port;
afe93325 2107 u8 ipproto;
a2b12f3c 2108 struct udp_offload_callbacks callbacks;
b582ef09
OG
2109};
2110
8f84985f
LR
2111/* often modified stats are per cpu, other are shared (netdev->stats) */
2112struct pcpu_sw_netstats {
2113 u64 rx_packets;
2114 u64 rx_bytes;
2115 u64 tx_packets;
2116 u64 tx_bytes;
2117 struct u64_stats_sync syncp;
2118};
2119
aabc92bb
PNA
2120#define __netdev_alloc_pcpu_stats(type, gfp) \
2121({ \
2122 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2123 if (pcpu_stats) { \
2124 int __cpu; \
2125 for_each_possible_cpu(__cpu) { \
2126 typeof(type) *stat; \
2127 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2128 u64_stats_init(&stat->syncp); \
2129 } \
2130 } \
2131 pcpu_stats; \
1c213bd2
WC
2132})
2133
aabc92bb 2134#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2135 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2136
1da177e4
LT
2137#include <linux/notifier.h>
2138
dcfe1421
AW
2139/* netdevice notifier chain. Please remember to update the rtnetlink
2140 * notification exclusion list in rtnetlink_event() when adding new
2141 * types.
2142 */
2143#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2144#define NETDEV_DOWN 0x0002
2145#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2146 detected a hardware crash and restarted
2147 - we can use this eg to kick tcp sessions
2148 once done */
2149#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2150#define NETDEV_REGISTER 0x0005
2151#define NETDEV_UNREGISTER 0x0006
1d486bfb 2152#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2153#define NETDEV_CHANGEADDR 0x0008
2154#define NETDEV_GOING_DOWN 0x0009
2155#define NETDEV_CHANGENAME 0x000A
2156#define NETDEV_FEAT_CHANGE 0x000B
2157#define NETDEV_BONDING_FAILOVER 0x000C
2158#define NETDEV_PRE_UP 0x000D
2159#define NETDEV_PRE_TYPE_CHANGE 0x000E
2160#define NETDEV_POST_TYPE_CHANGE 0x000F
2161#define NETDEV_POST_INIT 0x0010
0115e8e3 2162#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2163#define NETDEV_RELEASE 0x0012
2164#define NETDEV_NOTIFY_PEERS 0x0013
2165#define NETDEV_JOIN 0x0014
42e52bf9 2166#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2167#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2168#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2169#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2170#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2171#define NETDEV_PRECHANGEUPPER 0x001A
dcfe1421 2172
f629d208
JP
2173int register_netdevice_notifier(struct notifier_block *nb);
2174int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2175
2176struct netdev_notifier_info {
2177 struct net_device *dev;
2178};
2179
be9efd36
JP
2180struct netdev_notifier_change_info {
2181 struct netdev_notifier_info info; /* must be first */
2182 unsigned int flags_changed;
2183};
2184
0e4ead9d
JP
2185struct netdev_notifier_changeupper_info {
2186 struct netdev_notifier_info info; /* must be first */
2187 struct net_device *upper_dev; /* new upper dev */
2188 bool master; /* is upper dev master */
2189 bool linking; /* is the nofication for link or unlink */
2190};
2191
75538c2b
CW
2192static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2193 struct net_device *dev)
2194{
2195 info->dev = dev;
2196}
2197
351638e7
JP
2198static inline struct net_device *
2199netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2200{
2201 return info->dev;
2202}
2203
f629d208 2204int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2205
2206
1da177e4
LT
2207extern rwlock_t dev_base_lock; /* Device list lock */
2208
881d966b
EB
2209#define for_each_netdev(net, d) \
2210 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2211#define for_each_netdev_reverse(net, d) \
2212 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2213#define for_each_netdev_rcu(net, d) \
2214 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2215#define for_each_netdev_safe(net, d, n) \
2216 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2217#define for_each_netdev_continue(net, d) \
2218 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2219#define for_each_netdev_continue_rcu(net, d) \
2220 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2221#define for_each_netdev_in_bond_rcu(bond, slave) \
2222 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2223 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2224#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2225
a050c33f
DL
2226static inline struct net_device *next_net_device(struct net_device *dev)
2227{
2228 struct list_head *lh;
2229 struct net *net;
2230
c346dca1 2231 net = dev_net(dev);
a050c33f
DL
2232 lh = dev->dev_list.next;
2233 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2234}
2235
ce81b76a
ED
2236static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2237{
2238 struct list_head *lh;
2239 struct net *net;
2240
2241 net = dev_net(dev);
ccf43438 2242 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2243 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2244}
2245
a050c33f
DL
2246static inline struct net_device *first_net_device(struct net *net)
2247{
2248 return list_empty(&net->dev_base_head) ? NULL :
2249 net_device_entry(net->dev_base_head.next);
2250}
7562f876 2251
ccf43438
ED
2252static inline struct net_device *first_net_device_rcu(struct net *net)
2253{
2254 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2255
2256 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2257}
2258
f629d208
JP
2259int netdev_boot_setup_check(struct net_device *dev);
2260unsigned long netdev_boot_base(const char *prefix, int unit);
2261struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2262 const char *hwaddr);
2263struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2264struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2265void dev_add_pack(struct packet_type *pt);
2266void dev_remove_pack(struct packet_type *pt);
2267void __dev_remove_pack(struct packet_type *pt);
2268void dev_add_offload(struct packet_offload *po);
2269void dev_remove_offload(struct packet_offload *po);
f629d208 2270
a54acb3a 2271int dev_get_iflink(const struct net_device *dev);
fc4099f1 2272int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2273struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2274 unsigned short mask);
f629d208
JP
2275struct net_device *dev_get_by_name(struct net *net, const char *name);
2276struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2277struct net_device *__dev_get_by_name(struct net *net, const char *name);
2278int dev_alloc_name(struct net_device *dev, const char *name);
2279int dev_open(struct net_device *dev);
2280int dev_close(struct net_device *dev);
99c4a26a 2281int dev_close_many(struct list_head *head, bool unlink);
f629d208 2282void dev_disable_lro(struct net_device *dev);
0c4b51f0 2283int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2284int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2285int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2286int register_netdevice(struct net_device *dev);
2287void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2288void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2289static inline void unregister_netdevice(struct net_device *dev)
2290{
2291 unregister_netdevice_queue(dev, NULL);
2292}
2293
f629d208
JP
2294int netdev_refcnt_read(const struct net_device *dev);
2295void free_netdev(struct net_device *dev);
74d332c1 2296void netdev_freemem(struct net_device *dev);
f629d208
JP
2297void synchronize_net(void);
2298int init_dummy_netdev(struct net_device *dev);
937f1ba5 2299
f60e5990 2300DECLARE_PER_CPU(int, xmit_recursion);
2301static inline int dev_recursion_level(void)
2302{
2303 return this_cpu_read(xmit_recursion);
2304}
2305
f629d208
JP
2306struct net_device *dev_get_by_index(struct net *net, int ifindex);
2307struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2308struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2309int netdev_get_name(struct net *net, char *name, int ifindex);
2310int dev_restart(struct net_device *dev);
f629d208 2311int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2312
2313static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2314{
2315 return NAPI_GRO_CB(skb)->data_offset;
2316}
2317
2318static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2319{
2320 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2321}
2322
2323static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2324{
2325 NAPI_GRO_CB(skb)->data_offset += len;
2326}
2327
a5b1cf28
HX
2328static inline void *skb_gro_header_fast(struct sk_buff *skb,
2329 unsigned int offset)
86911732 2330{
a5b1cf28
HX
2331 return NAPI_GRO_CB(skb)->frag0 + offset;
2332}
78a478d0 2333
a5b1cf28
HX
2334static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2335{
2336 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2337}
78a478d0 2338
a5b1cf28
HX
2339static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2340 unsigned int offset)
2341{
17dd759c
HX
2342 if (!pskb_may_pull(skb, hlen))
2343 return NULL;
2344
a5b1cf28
HX
2345 NAPI_GRO_CB(skb)->frag0 = NULL;
2346 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2347 return skb->data + offset;
86911732 2348}
1da177e4 2349
36e7b1b8
HX
2350static inline void *skb_gro_network_header(struct sk_buff *skb)
2351{
78d3fd0b
HX
2352 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2353 skb_network_offset(skb);
36e7b1b8
HX
2354}
2355
bf5a755f
JC
2356static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2357 const void *start, unsigned int len)
2358{
573e8fca 2359 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2360 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2361 csum_partial(start, len, 0));
2362}
2363
573e8fca
TH
2364/* GRO checksum functions. These are logical equivalents of the normal
2365 * checksum functions (in skbuff.h) except that they operate on the GRO
2366 * offsets and fields in sk_buff.
2367 */
2368
2369__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2370
15e2396d
TH
2371static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2372{
b7fe10e5 2373 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2374}
2375
573e8fca
TH
2376static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2377 bool zero_okay,
2378 __sum16 check)
2379{
6edec0e6
TH
2380 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2381 skb_checksum_start_offset(skb) <
2382 skb_gro_offset(skb)) &&
15e2396d 2383 !skb_at_gro_remcsum_start(skb) &&
662880f4 2384 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2385 (!zero_okay || check));
2386}
2387
2388static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2389 __wsum psum)
2390{
2391 if (NAPI_GRO_CB(skb)->csum_valid &&
2392 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2393 return 0;
2394
2395 NAPI_GRO_CB(skb)->csum = psum;
2396
2397 return __skb_gro_checksum_complete(skb);
2398}
2399
573e8fca
TH
2400static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2401{
662880f4
TH
2402 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2403 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2404 NAPI_GRO_CB(skb)->csum_cnt--;
2405 } else {
2406 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2407 * verified a new top level checksum or an encapsulated one
2408 * during GRO. This saves work if we fallback to normal path.
2409 */
2410 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2411 }
2412}
2413
2414#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2415 compute_pseudo) \
2416({ \
2417 __sum16 __ret = 0; \
2418 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2419 __ret = __skb_gro_checksum_validate_complete(skb, \
2420 compute_pseudo(skb, proto)); \
5a212329
TH
2421 if (__ret) \
2422 __skb_mark_checksum_bad(skb); \
2423 else \
573e8fca
TH
2424 skb_gro_incr_csum_unnecessary(skb); \
2425 __ret; \
2426})
2427
2428#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2429 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2430
2431#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2432 compute_pseudo) \
2433 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2434
2435#define skb_gro_checksum_simple_validate(skb) \
2436 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2437
d96535a1
TH
2438static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2439{
2440 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2441 !NAPI_GRO_CB(skb)->csum_valid);
2442}
2443
2444static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2445 __sum16 check, __wsum pseudo)
2446{
2447 NAPI_GRO_CB(skb)->csum = ~pseudo;
2448 NAPI_GRO_CB(skb)->csum_valid = 1;
2449}
2450
2451#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2452do { \
2453 if (__skb_gro_checksum_convert_check(skb)) \
2454 __skb_gro_checksum_convert(skb, check, \
2455 compute_pseudo(skb, proto)); \
2456} while (0)
2457
26c4f7da
TH
2458struct gro_remcsum {
2459 int offset;
2460 __wsum delta;
2461};
2462
2463static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2464{
846cd667 2465 grc->offset = 0;
26c4f7da
TH
2466 grc->delta = 0;
2467}
2468
b7fe10e5
TH
2469static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2470 unsigned int off, size_t hdrlen,
2471 int start, int offset,
2472 struct gro_remcsum *grc,
2473 bool nopartial)
dcdc8994
TH
2474{
2475 __wsum delta;
b7fe10e5 2476 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2477
2478 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2479
15e2396d 2480 if (!nopartial) {
b7fe10e5
TH
2481 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2482 return ptr;
2483 }
2484
2485 ptr = skb_gro_header_fast(skb, off);
2486 if (skb_gro_header_hard(skb, off + plen)) {
2487 ptr = skb_gro_header_slow(skb, off + plen, off);
2488 if (!ptr)
2489 return NULL;
15e2396d
TH
2490 }
2491
b7fe10e5
TH
2492 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2493 start, offset);
dcdc8994
TH
2494
2495 /* Adjust skb->csum since we changed the packet */
dcdc8994 2496 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2497
b7fe10e5 2498 grc->offset = off + hdrlen + offset;
26c4f7da 2499 grc->delta = delta;
b7fe10e5
TH
2500
2501 return ptr;
dcdc8994
TH
2502}
2503
26c4f7da
TH
2504static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2505 struct gro_remcsum *grc)
2506{
b7fe10e5
TH
2507 void *ptr;
2508 size_t plen = grc->offset + sizeof(u16);
2509
26c4f7da
TH
2510 if (!grc->delta)
2511 return;
2512
b7fe10e5
TH
2513 ptr = skb_gro_header_fast(skb, grc->offset);
2514 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2515 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2516 if (!ptr)
2517 return;
2518 }
2519
2520 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2521}
dcdc8994 2522
0c4e8581
SH
2523static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2524 unsigned short type,
3b04ddde 2525 const void *daddr, const void *saddr,
95c96174 2526 unsigned int len)
0c4e8581 2527{
f1ecfd5d 2528 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2529 return 0;
3b04ddde
SH
2530
2531 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2532}
2533
b95cce35
SH
2534static inline int dev_parse_header(const struct sk_buff *skb,
2535 unsigned char *haddr)
2536{
2537 const struct net_device *dev = skb->dev;
2538
1b83336b 2539 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2540 return 0;
3b04ddde 2541 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2542}
2543
0780eb5e
WB
2544/* ll_header must have at least hard_header_len allocated */
2545static inline bool dev_validate_header(const struct net_device *dev,
2546 char *ll_header, int len)
2547{
2548 if (likely(len >= dev->hard_header_len))
2549 return true;
2550
2551 if (capable(CAP_SYS_RAWIO)) {
2552 memset(ll_header + len, 0, dev->hard_header_len - len);
2553 return true;
2554 }
2555
2556 if (dev->header_ops && dev->header_ops->validate)
2557 return dev->header_ops->validate(ll_header, len);
2558
2559 return false;
2560}
2561
1da177e4 2562typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2563int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2564static inline int unregister_gifconf(unsigned int family)
2565{
2566 return register_gifconf(family, NULL);
2567}
2568
99bbc707 2569#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2570#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2571struct sd_flow_limit {
2572 u64 count;
2573 unsigned int num_buckets;
2574 unsigned int history_head;
2575 u16 history[FLOW_LIMIT_HISTORY];
2576 u8 buckets[];
2577};
2578
2579extern int netdev_flow_limit_table_len;
2580#endif /* CONFIG_NET_FLOW_LIMIT */
2581
1da177e4 2582/*
88751275 2583 * Incoming packets are placed on per-cpu queues
1da177e4 2584 */
d94d9fee 2585struct softnet_data {
1da177e4 2586 struct list_head poll_list;
6e7676c1 2587 struct sk_buff_head process_queue;
1da177e4 2588
dee42870 2589 /* stats */
cd7b5396
DM
2590 unsigned int processed;
2591 unsigned int time_squeeze;
2592 unsigned int cpu_collision;
2593 unsigned int received_rps;
fd793d89 2594#ifdef CONFIG_RPS
88751275 2595 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2596#endif
2597#ifdef CONFIG_NET_FLOW_LIMIT
2598 struct sd_flow_limit __rcu *flow_limit;
2599#endif
2600 struct Qdisc *output_queue;
2601 struct Qdisc **output_queue_tailp;
2602 struct sk_buff *completion_queue;
88751275 2603
4cdb1e2e 2604#ifdef CONFIG_RPS
88751275 2605 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 2606 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2607 struct softnet_data *rps_ipi_next;
2608 unsigned int cpu;
fec5e652 2609 unsigned int input_queue_head;
76cc8b13 2610 unsigned int input_queue_tail;
1e94d72f 2611#endif
95c96174 2612 unsigned int dropped;
0a9627f2 2613 struct sk_buff_head input_pkt_queue;
bea3348e 2614 struct napi_struct backlog;
99bbc707 2615
1da177e4
LT
2616};
2617
76cc8b13 2618static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2619{
2620#ifdef CONFIG_RPS
76cc8b13
TH
2621 sd->input_queue_head++;
2622#endif
2623}
2624
2625static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2626 unsigned int *qtail)
2627{
2628#ifdef CONFIG_RPS
2629 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2630#endif
2631}
2632
0a9627f2 2633DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2634
f629d208 2635void __netif_schedule(struct Qdisc *q);
46e5da40 2636void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2637
fd2ea0a7
DM
2638static inline void netif_tx_schedule_all(struct net_device *dev)
2639{
2640 unsigned int i;
2641
2642 for (i = 0; i < dev->num_tx_queues; i++)
2643 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2644}
2645
d29f749e
DJ
2646static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2647{
73466498 2648 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2649}
2650
bea3348e
SH
2651/**
2652 * netif_start_queue - allow transmit
2653 * @dev: network device
2654 *
2655 * Allow upper layers to call the device hard_start_xmit routine.
2656 */
1da177e4
LT
2657static inline void netif_start_queue(struct net_device *dev)
2658{
e8a0464c 2659 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2660}
2661
fd2ea0a7
DM
2662static inline void netif_tx_start_all_queues(struct net_device *dev)
2663{
2664 unsigned int i;
2665
2666 for (i = 0; i < dev->num_tx_queues; i++) {
2667 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2668 netif_tx_start_queue(txq);
2669 }
2670}
2671
46e5da40 2672void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2673
d29f749e
DJ
2674/**
2675 * netif_wake_queue - restart transmit
2676 * @dev: network device
2677 *
2678 * Allow upper layers to call the device hard_start_xmit routine.
2679 * Used for flow control when transmit resources are available.
2680 */
79d16385
DM
2681static inline void netif_wake_queue(struct net_device *dev)
2682{
e8a0464c 2683 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2684}
2685
fd2ea0a7
DM
2686static inline void netif_tx_wake_all_queues(struct net_device *dev)
2687{
2688 unsigned int i;
2689
2690 for (i = 0; i < dev->num_tx_queues; i++) {
2691 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2692 netif_tx_wake_queue(txq);
2693 }
2694}
2695
d29f749e
DJ
2696static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2697{
73466498 2698 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2699}
2700
bea3348e
SH
2701/**
2702 * netif_stop_queue - stop transmitted packets
2703 * @dev: network device
2704 *
2705 * Stop upper layers calling the device hard_start_xmit routine.
2706 * Used for flow control when transmit resources are unavailable.
2707 */
1da177e4
LT
2708static inline void netif_stop_queue(struct net_device *dev)
2709{
e8a0464c 2710 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2711}
2712
a2029240 2713void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2714
4d29515f 2715static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2716{
73466498 2717 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2718}
2719
bea3348e
SH
2720/**
2721 * netif_queue_stopped - test if transmit queue is flowblocked
2722 * @dev: network device
2723 *
2724 * Test if transmit queue on device is currently unable to send.
2725 */
4d29515f 2726static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2727{
e8a0464c 2728 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2729}
2730
4d29515f 2731static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2732{
73466498
TH
2733 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2734}
2735
8e2f1a63
DB
2736static inline bool
2737netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2738{
2739 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2740}
2741
8e2f1a63
DB
2742static inline bool
2743netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2744{
2745 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2746}
2747
53511453
ED
2748/**
2749 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2750 * @dev_queue: pointer to transmit queue
2751 *
2752 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2753 * to give appropriate hint to the cpu.
2754 */
2755static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2756{
2757#ifdef CONFIG_BQL
2758 prefetchw(&dev_queue->dql.num_queued);
2759#endif
2760}
2761
2762/**
2763 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2764 * @dev_queue: pointer to transmit queue
2765 *
2766 * BQL enabled drivers might use this helper in their TX completion path,
2767 * to give appropriate hint to the cpu.
2768 */
2769static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2770{
2771#ifdef CONFIG_BQL
2772 prefetchw(&dev_queue->dql.limit);
2773#endif
2774}
2775
c5d67bd7
TH
2776static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2777 unsigned int bytes)
2778{
114cf580
TH
2779#ifdef CONFIG_BQL
2780 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2781
2782 if (likely(dql_avail(&dev_queue->dql) >= 0))
2783 return;
2784
2785 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2786
2787 /*
2788 * The XOFF flag must be set before checking the dql_avail below,
2789 * because in netdev_tx_completed_queue we update the dql_completed
2790 * before checking the XOFF flag.
2791 */
2792 smp_mb();
2793
2794 /* check again in case another CPU has just made room avail */
2795 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2796 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2797#endif
c5d67bd7
TH
2798}
2799
0042d0c8
FF
2800/**
2801 * netdev_sent_queue - report the number of bytes queued to hardware
2802 * @dev: network device
2803 * @bytes: number of bytes queued to the hardware device queue
2804 *
2805 * Report the number of bytes queued for sending/completion to the network
2806 * device hardware queue. @bytes should be a good approximation and should
2807 * exactly match netdev_completed_queue() @bytes
2808 */
c5d67bd7
TH
2809static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2810{
2811 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2812}
2813
2814static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2815 unsigned int pkts, unsigned int bytes)
c5d67bd7 2816{
114cf580 2817#ifdef CONFIG_BQL
b37c0fbe
AD
2818 if (unlikely(!bytes))
2819 return;
2820
2821 dql_completed(&dev_queue->dql, bytes);
2822
2823 /*
2824 * Without the memory barrier there is a small possiblity that
2825 * netdev_tx_sent_queue will miss the update and cause the queue to
2826 * be stopped forever
2827 */
2828 smp_mb();
2829
2830 if (dql_avail(&dev_queue->dql) < 0)
2831 return;
2832
2833 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2834 netif_schedule_queue(dev_queue);
114cf580 2835#endif
c5d67bd7
TH
2836}
2837
0042d0c8
FF
2838/**
2839 * netdev_completed_queue - report bytes and packets completed by device
2840 * @dev: network device
2841 * @pkts: actual number of packets sent over the medium
2842 * @bytes: actual number of bytes sent over the medium
2843 *
2844 * Report the number of bytes and packets transmitted by the network device
2845 * hardware queue over the physical medium, @bytes must exactly match the
2846 * @bytes amount passed to netdev_sent_queue()
2847 */
c5d67bd7 2848static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2849 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2850{
2851 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2852}
2853
2854static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2855{
114cf580 2856#ifdef CONFIG_BQL
5c490354 2857 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2858 dql_reset(&q->dql);
2859#endif
c5d67bd7
TH
2860}
2861
0042d0c8
FF
2862/**
2863 * netdev_reset_queue - reset the packets and bytes count of a network device
2864 * @dev_queue: network device
2865 *
2866 * Reset the bytes and packet count of a network device and clear the
2867 * software flow control OFF bit for this network device
2868 */
c5d67bd7
TH
2869static inline void netdev_reset_queue(struct net_device *dev_queue)
2870{
2871 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2872}
2873
b9507bda
DB
2874/**
2875 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2876 * @dev: network device
2877 * @queue_index: given tx queue index
2878 *
2879 * Returns 0 if given tx queue index >= number of device tx queues,
2880 * otherwise returns the originally passed tx queue index.
2881 */
2882static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2883{
2884 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2885 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2886 dev->name, queue_index,
2887 dev->real_num_tx_queues);
2888 return 0;
2889 }
2890
2891 return queue_index;
2892}
2893
bea3348e
SH
2894/**
2895 * netif_running - test if up
2896 * @dev: network device
2897 *
2898 * Test if the device has been brought up.
2899 */
4d29515f 2900static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2901{
2902 return test_bit(__LINK_STATE_START, &dev->state);
2903}
2904
f25f4e44
PWJ
2905/*
2906 * Routines to manage the subqueues on a device. We only need start
2907 * stop, and a check if it's stopped. All other device management is
2908 * done at the overall netdevice level.
2909 * Also test the device if we're multiqueue.
2910 */
bea3348e
SH
2911
2912/**
2913 * netif_start_subqueue - allow sending packets on subqueue
2914 * @dev: network device
2915 * @queue_index: sub queue index
2916 *
2917 * Start individual transmit queue of a device with multiple transmit queues.
2918 */
f25f4e44
PWJ
2919static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2920{
fd2ea0a7 2921 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2922
2923 netif_tx_start_queue(txq);
f25f4e44
PWJ
2924}
2925
bea3348e
SH
2926/**
2927 * netif_stop_subqueue - stop sending packets on subqueue
2928 * @dev: network device
2929 * @queue_index: sub queue index
2930 *
2931 * Stop individual transmit queue of a device with multiple transmit queues.
2932 */
f25f4e44
PWJ
2933static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2934{
fd2ea0a7 2935 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 2936 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2937}
2938
bea3348e
SH
2939/**
2940 * netif_subqueue_stopped - test status of subqueue
2941 * @dev: network device
2942 * @queue_index: sub queue index
2943 *
2944 * Check individual transmit queue of a device with multiple transmit queues.
2945 */
4d29515f
DM
2946static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2947 u16 queue_index)
f25f4e44 2948{
fd2ea0a7 2949 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2950
2951 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2952}
2953
4d29515f
DM
2954static inline bool netif_subqueue_stopped(const struct net_device *dev,
2955 struct sk_buff *skb)
668f895a
PE
2956{
2957 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2958}
bea3348e 2959
46e5da40 2960void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 2961
537c00de 2962#ifdef CONFIG_XPS
53af53ae 2963int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2964 u16 index);
537c00de
AD
2965#else
2966static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2967 const struct cpumask *mask,
537c00de
AD
2968 u16 index)
2969{
2970 return 0;
2971}
2972#endif
2973
5605c762
JP
2974u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2975 unsigned int num_tx_queues);
2976
a3d22a68
VZ
2977/*
2978 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2979 * as a distribution range limit for the returned value.
2980 */
2981static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 2982 struct sk_buff *skb)
a3d22a68
VZ
2983{
2984 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2985}
2986
bea3348e
SH
2987/**
2988 * netif_is_multiqueue - test if device has multiple transmit queues
2989 * @dev: network device
2990 *
2991 * Check if device has multiple transmit queues
bea3348e 2992 */
4d29515f 2993static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2994{
a02cec21 2995 return dev->num_tx_queues > 1;
f25f4e44 2996}
1da177e4 2997
f629d208 2998int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2999
a953be53 3000#ifdef CONFIG_SYSFS
f629d208 3001int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3002#else
3003static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3004 unsigned int rxq)
3005{
3006 return 0;
3007}
3008#endif
3009
a953be53
MD
3010#ifdef CONFIG_SYSFS
3011static inline unsigned int get_netdev_rx_queue_index(
3012 struct netdev_rx_queue *queue)
3013{
3014 struct net_device *dev = queue->dev;
3015 int index = queue - dev->_rx;
3016
3017 BUG_ON(index >= dev->num_rx_queues);
3018 return index;
3019}
3020#endif
3021
16917b87 3022#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3023int netif_get_num_default_rss_queues(void);
16917b87 3024
e6247027
ED
3025enum skb_free_reason {
3026 SKB_REASON_CONSUMED,
3027 SKB_REASON_DROPPED,
3028};
3029
3030void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3031void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3032
e6247027
ED
3033/*
3034 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3035 * interrupt context or with hardware interrupts being disabled.
3036 * (in_irq() || irqs_disabled())
3037 *
3038 * We provide four helpers that can be used in following contexts :
3039 *
3040 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3041 * replacing kfree_skb(skb)
3042 *
3043 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3044 * Typically used in place of consume_skb(skb) in TX completion path
3045 *
3046 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3047 * replacing kfree_skb(skb)
3048 *
3049 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3050 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3051 */
e6247027
ED
3052static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3053{
3054 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3055}
3056
3057static inline void dev_consume_skb_irq(struct sk_buff *skb)
3058{
3059 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3060}
3061
3062static inline void dev_kfree_skb_any(struct sk_buff *skb)
3063{
3064 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3065}
3066
3067static inline void dev_consume_skb_any(struct sk_buff *skb)
3068{
3069 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3070}
1da177e4 3071
f629d208
JP
3072int netif_rx(struct sk_buff *skb);
3073int netif_rx_ni(struct sk_buff *skb);
04eb4489 3074int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3075gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3076void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3077struct sk_buff *napi_get_frags(struct napi_struct *napi);
3078gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3079struct packet_offload *gro_find_receive_by_type(__be16 type);
3080struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3081
3082static inline void napi_free_frags(struct napi_struct *napi)
3083{
3084 kfree_skb(napi->skb);
3085 napi->skb = NULL;
3086}
3087
4b90139c 3088bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3089int netdev_rx_handler_register(struct net_device *dev,
3090 rx_handler_func_t *rx_handler,
3091 void *rx_handler_data);
3092void netdev_rx_handler_unregister(struct net_device *dev);
3093
3094bool dev_valid_name(const char *name);
3095int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3096int dev_ethtool(struct net *net, struct ifreq *);
3097unsigned int dev_get_flags(const struct net_device *);
3098int __dev_change_flags(struct net_device *, unsigned int flags);
3099int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3100void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3101 unsigned int gchanges);
f629d208
JP
3102int dev_change_name(struct net_device *, const char *);
3103int dev_set_alias(struct net_device *, const char *, size_t);
3104int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3105int dev_set_mtu(struct net_device *, int);
3106void dev_set_group(struct net_device *, int);
3107int dev_set_mac_address(struct net_device *, struct sockaddr *);
3108int dev_change_carrier(struct net_device *, bool new_carrier);
3109int dev_get_phys_port_id(struct net_device *dev,
02637fce 3110 struct netdev_phys_item_id *ppid);
db24a904
DA
3111int dev_get_phys_port_name(struct net_device *dev,
3112 char *name, size_t len);
d746d707 3113int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3114struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3115struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3116 struct netdev_queue *txq, int *ret);
a0265d28 3117int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3118int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1ee481fb 3119bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
1da177e4 3120
20380731 3121extern int netdev_budget;
1da177e4
LT
3122
3123/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3124void netdev_run_todo(void);
1da177e4 3125
bea3348e
SH
3126/**
3127 * dev_put - release reference to device
3128 * @dev: network device
3129 *
9ef4429b 3130 * Release reference to device to allow it to be freed.
bea3348e 3131 */
1da177e4
LT
3132static inline void dev_put(struct net_device *dev)
3133{
933393f5 3134 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3135}
3136
bea3348e
SH
3137/**
3138 * dev_hold - get reference to device
3139 * @dev: network device
3140 *
9ef4429b 3141 * Hold reference to device to keep it from being freed.
bea3348e 3142 */
15333061
SH
3143static inline void dev_hold(struct net_device *dev)
3144{
933393f5 3145 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3146}
1da177e4
LT
3147
3148/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3149 * and _off may be called from IRQ context, but it is caller
3150 * who is responsible for serialization of these calls.
b00055aa
SR
3151 *
3152 * The name carrier is inappropriate, these functions should really be
3153 * called netif_lowerlayer_*() because they represent the state of any
3154 * kind of lower layer not just hardware media.
1da177e4
LT
3155 */
3156
f629d208
JP
3157void linkwatch_init_dev(struct net_device *dev);
3158void linkwatch_fire_event(struct net_device *dev);
3159void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3160
bea3348e
SH
3161/**
3162 * netif_carrier_ok - test if carrier present
3163 * @dev: network device
3164 *
3165 * Check if carrier is present on device
3166 */
4d29515f 3167static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3168{
3169 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3170}
3171
f629d208 3172unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3173
f629d208 3174void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3175
f629d208 3176void netif_carrier_on(struct net_device *dev);
1da177e4 3177
f629d208 3178void netif_carrier_off(struct net_device *dev);
1da177e4 3179
bea3348e
SH
3180/**
3181 * netif_dormant_on - mark device as dormant.
3182 * @dev: network device
3183 *
3184 * Mark device as dormant (as per RFC2863).
3185 *
3186 * The dormant state indicates that the relevant interface is not
3187 * actually in a condition to pass packets (i.e., it is not 'up') but is
3188 * in a "pending" state, waiting for some external event. For "on-
3189 * demand" interfaces, this new state identifies the situation where the
3190 * interface is waiting for events to place it in the up state.
3191 *
3192 */
b00055aa
SR
3193static inline void netif_dormant_on(struct net_device *dev)
3194{
3195 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3196 linkwatch_fire_event(dev);
3197}
3198
bea3348e
SH
3199/**
3200 * netif_dormant_off - set device as not dormant.
3201 * @dev: network device
3202 *
3203 * Device is not in dormant state.
3204 */
b00055aa
SR
3205static inline void netif_dormant_off(struct net_device *dev)
3206{
3207 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3208 linkwatch_fire_event(dev);
3209}
3210
bea3348e
SH
3211/**
3212 * netif_dormant - test if carrier present
3213 * @dev: network device
3214 *
3215 * Check if carrier is present on device
3216 */
4d29515f 3217static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3218{
3219 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3220}
3221
3222
bea3348e
SH
3223/**
3224 * netif_oper_up - test if device is operational
3225 * @dev: network device
3226 *
3227 * Check if carrier is operational
3228 */
4d29515f 3229static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3230{
b00055aa
SR
3231 return (dev->operstate == IF_OPER_UP ||
3232 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3233}
3234
bea3348e
SH
3235/**
3236 * netif_device_present - is device available or removed
3237 * @dev: network device
3238 *
3239 * Check if device has not been removed from system.
3240 */
4d29515f 3241static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3242{
3243 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3244}
3245
f629d208 3246void netif_device_detach(struct net_device *dev);
1da177e4 3247
f629d208 3248void netif_device_attach(struct net_device *dev);
1da177e4
LT
3249
3250/*
3251 * Network interface message level settings
3252 */
1da177e4
LT
3253
3254enum {
3255 NETIF_MSG_DRV = 0x0001,
3256 NETIF_MSG_PROBE = 0x0002,
3257 NETIF_MSG_LINK = 0x0004,
3258 NETIF_MSG_TIMER = 0x0008,
3259 NETIF_MSG_IFDOWN = 0x0010,
3260 NETIF_MSG_IFUP = 0x0020,
3261 NETIF_MSG_RX_ERR = 0x0040,
3262 NETIF_MSG_TX_ERR = 0x0080,
3263 NETIF_MSG_TX_QUEUED = 0x0100,
3264 NETIF_MSG_INTR = 0x0200,
3265 NETIF_MSG_TX_DONE = 0x0400,
3266 NETIF_MSG_RX_STATUS = 0x0800,
3267 NETIF_MSG_PKTDATA = 0x1000,
3268 NETIF_MSG_HW = 0x2000,
3269 NETIF_MSG_WOL = 0x4000,
3270};
3271
3272#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3273#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3274#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3275#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3276#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3277#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3278#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3279#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3280#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3281#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3282#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3283#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3284#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3285#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3286#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3287
3288static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3289{
3290 /* use default */
3291 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3292 return default_msg_enable_bits;
3293 if (debug_value == 0) /* no output */
3294 return 0;
3295 /* set low N bits */
3296 return (1 << debug_value) - 1;
3297}
3298
c773e847 3299static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3300{
c773e847
DM
3301 spin_lock(&txq->_xmit_lock);
3302 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3303}
3304
fd2ea0a7
DM
3305static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3306{
3307 spin_lock_bh(&txq->_xmit_lock);
3308 txq->xmit_lock_owner = smp_processor_id();
3309}
3310
4d29515f 3311static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3312{
4d29515f 3313 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3314 if (likely(ok))
3315 txq->xmit_lock_owner = smp_processor_id();
3316 return ok;
3317}
3318
3319static inline void __netif_tx_unlock(struct netdev_queue *txq)
3320{
3321 txq->xmit_lock_owner = -1;
3322 spin_unlock(&txq->_xmit_lock);
3323}
3324
3325static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3326{
3327 txq->xmit_lock_owner = -1;
3328 spin_unlock_bh(&txq->_xmit_lock);
3329}
3330
08baf561
ED
3331static inline void txq_trans_update(struct netdev_queue *txq)
3332{
3333 if (txq->xmit_lock_owner != -1)
3334 txq->trans_start = jiffies;
3335}
3336
d29f749e
DJ
3337/**
3338 * netif_tx_lock - grab network device transmit lock
3339 * @dev: network device
d29f749e
DJ
3340 *
3341 * Get network device transmit lock
3342 */
22dd7495
JHS
3343static inline void netif_tx_lock(struct net_device *dev)
3344{
e8a0464c 3345 unsigned int i;
c3f26a26 3346 int cpu;
c773e847 3347
c3f26a26
DM
3348 spin_lock(&dev->tx_global_lock);
3349 cpu = smp_processor_id();
e8a0464c
DM
3350 for (i = 0; i < dev->num_tx_queues; i++) {
3351 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3352
3353 /* We are the only thread of execution doing a
3354 * freeze, but we have to grab the _xmit_lock in
3355 * order to synchronize with threads which are in
3356 * the ->hard_start_xmit() handler and already
3357 * checked the frozen bit.
3358 */
e8a0464c 3359 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3360 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3361 __netif_tx_unlock(txq);
e8a0464c 3362 }
932ff279
HX
3363}
3364
3365static inline void netif_tx_lock_bh(struct net_device *dev)
3366{
e8a0464c
DM
3367 local_bh_disable();
3368 netif_tx_lock(dev);
932ff279
HX
3369}
3370
932ff279
HX
3371static inline void netif_tx_unlock(struct net_device *dev)
3372{
e8a0464c
DM
3373 unsigned int i;
3374
3375 for (i = 0; i < dev->num_tx_queues; i++) {
3376 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3377
c3f26a26
DM
3378 /* No need to grab the _xmit_lock here. If the
3379 * queue is not stopped for another reason, we
3380 * force a schedule.
3381 */
3382 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3383 netif_schedule_queue(txq);
c3f26a26
DM
3384 }
3385 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3386}
3387
3388static inline void netif_tx_unlock_bh(struct net_device *dev)
3389{
e8a0464c
DM
3390 netif_tx_unlock(dev);
3391 local_bh_enable();
932ff279
HX
3392}
3393
c773e847 3394#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3395 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3396 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3397 } \
3398}
3399
5efeac44
EB
3400#define HARD_TX_TRYLOCK(dev, txq) \
3401 (((dev->features & NETIF_F_LLTX) == 0) ? \
3402 __netif_tx_trylock(txq) : \
3403 true )
3404
c773e847 3405#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3406 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3407 __netif_tx_unlock(txq); \
22dd7495
JHS
3408 } \
3409}
3410
1da177e4
LT
3411static inline void netif_tx_disable(struct net_device *dev)
3412{
fd2ea0a7 3413 unsigned int i;
c3f26a26 3414 int cpu;
fd2ea0a7 3415
c3f26a26
DM
3416 local_bh_disable();
3417 cpu = smp_processor_id();
fd2ea0a7
DM
3418 for (i = 0; i < dev->num_tx_queues; i++) {
3419 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3420
3421 __netif_tx_lock(txq, cpu);
fd2ea0a7 3422 netif_tx_stop_queue(txq);
c3f26a26 3423 __netif_tx_unlock(txq);
fd2ea0a7 3424 }
c3f26a26 3425 local_bh_enable();
1da177e4
LT
3426}
3427
e308a5d8
DM
3428static inline void netif_addr_lock(struct net_device *dev)
3429{
3430 spin_lock(&dev->addr_list_lock);
3431}
3432
2429f7ac
JP
3433static inline void netif_addr_lock_nested(struct net_device *dev)
3434{
25175ba5
VY
3435 int subclass = SINGLE_DEPTH_NESTING;
3436
3437 if (dev->netdev_ops->ndo_get_lock_subclass)
3438 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3439
3440 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3441}
3442
e308a5d8
DM
3443static inline void netif_addr_lock_bh(struct net_device *dev)
3444{
3445 spin_lock_bh(&dev->addr_list_lock);
3446}
3447
3448static inline void netif_addr_unlock(struct net_device *dev)
3449{
3450 spin_unlock(&dev->addr_list_lock);
3451}
3452
3453static inline void netif_addr_unlock_bh(struct net_device *dev)
3454{
3455 spin_unlock_bh(&dev->addr_list_lock);
3456}
3457
f001fde5 3458/*
31278e71 3459 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3460 * rcu_read_lock held.
3461 */
3462#define for_each_dev_addr(dev, ha) \
31278e71 3463 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3464
1da177e4
LT
3465/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3466
f629d208 3467void ether_setup(struct net_device *dev);
1da177e4
LT
3468
3469/* Support for loadable net-drivers */
f629d208 3470struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3471 unsigned char name_assign_type,
f629d208
JP
3472 void (*setup)(struct net_device *),
3473 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3474#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3475 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3476
c835a677
TG
3477#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3478 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3479 count)
36909ea4 3480
f629d208
JP
3481int register_netdev(struct net_device *dev);
3482void unregister_netdev(struct net_device *dev);
f001fde5 3483
22bedad3 3484/* General hardware address lists handling functions */
f629d208
JP
3485int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3486 struct netdev_hw_addr_list *from_list, int addr_len);
3487void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3488 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3489int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3490 struct net_device *dev,
3491 int (*sync)(struct net_device *, const unsigned char *),
3492 int (*unsync)(struct net_device *,
3493 const unsigned char *));
3494void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3495 struct net_device *dev,
3496 int (*unsync)(struct net_device *,
3497 const unsigned char *));
f629d208 3498void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3499
f001fde5 3500/* Functions used for device addresses handling */
f629d208
JP
3501int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3502 unsigned char addr_type);
3503int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3504 unsigned char addr_type);
f629d208
JP
3505void dev_addr_flush(struct net_device *dev);
3506int dev_addr_init(struct net_device *dev);
a748ee24
JP
3507
3508/* Functions used for unicast addresses handling */
f629d208
JP
3509int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3510int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3511int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3512int dev_uc_sync(struct net_device *to, struct net_device *from);
3513int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3514void dev_uc_unsync(struct net_device *to, struct net_device *from);
3515void dev_uc_flush(struct net_device *dev);
3516void dev_uc_init(struct net_device *dev);
f001fde5 3517
670e5b8e
AD
3518/**
3519 * __dev_uc_sync - Synchonize device's unicast list
3520 * @dev: device to sync
3521 * @sync: function to call if address should be added
3522 * @unsync: function to call if address should be removed
3523 *
3524 * Add newly added addresses to the interface, and release
3525 * addresses that have been deleted.
3526 **/
3527static inline int __dev_uc_sync(struct net_device *dev,
3528 int (*sync)(struct net_device *,
3529 const unsigned char *),
3530 int (*unsync)(struct net_device *,
3531 const unsigned char *))
3532{
3533 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3534}
3535
3536/**
e793c0f7 3537 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3538 * @dev: device to sync
3539 * @unsync: function to call if address should be removed
3540 *
3541 * Remove all addresses that were added to the device by dev_uc_sync().
3542 **/
3543static inline void __dev_uc_unsync(struct net_device *dev,
3544 int (*unsync)(struct net_device *,
3545 const unsigned char *))
3546{
3547 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3548}
3549
22bedad3 3550/* Functions used for multicast addresses handling */
f629d208
JP
3551int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3552int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3553int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3554int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3555int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3556int dev_mc_sync(struct net_device *to, struct net_device *from);
3557int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3558void dev_mc_unsync(struct net_device *to, struct net_device *from);
3559void dev_mc_flush(struct net_device *dev);
3560void dev_mc_init(struct net_device *dev);
f001fde5 3561
670e5b8e
AD
3562/**
3563 * __dev_mc_sync - Synchonize device's multicast list
3564 * @dev: device to sync
3565 * @sync: function to call if address should be added
3566 * @unsync: function to call if address should be removed
3567 *
3568 * Add newly added addresses to the interface, and release
3569 * addresses that have been deleted.
3570 **/
3571static inline int __dev_mc_sync(struct net_device *dev,
3572 int (*sync)(struct net_device *,
3573 const unsigned char *),
3574 int (*unsync)(struct net_device *,
3575 const unsigned char *))
3576{
3577 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3578}
3579
3580/**
e793c0f7 3581 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3582 * @dev: device to sync
3583 * @unsync: function to call if address should be removed
3584 *
3585 * Remove all addresses that were added to the device by dev_mc_sync().
3586 **/
3587static inline void __dev_mc_unsync(struct net_device *dev,
3588 int (*unsync)(struct net_device *,
3589 const unsigned char *))
3590{
3591 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3592}
3593
4417da66 3594/* Functions used for secondary unicast and multicast support */
f629d208
JP
3595void dev_set_rx_mode(struct net_device *dev);
3596void __dev_set_rx_mode(struct net_device *dev);
3597int dev_set_promiscuity(struct net_device *dev, int inc);
3598int dev_set_allmulti(struct net_device *dev, int inc);
3599void netdev_state_change(struct net_device *dev);
3600void netdev_notify_peers(struct net_device *dev);
3601void netdev_features_change(struct net_device *dev);
1da177e4 3602/* Load a device via the kmod */
f629d208
JP
3603void dev_load(struct net *net, const char *name);
3604struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3605 struct rtnl_link_stats64 *storage);
3606void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3607 const struct net_device_stats *netdev_stats);
eeda3fd6 3608
1da177e4 3609extern int netdev_max_backlog;
3b098e2d 3610extern int netdev_tstamp_prequeue;
1da177e4 3611extern int weight_p;
0a14842f 3612extern int bpf_jit_enable;
9ff162a8 3613
f629d208 3614bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3615struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3616 struct list_head **iter);
f629d208
JP
3617struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3618 struct list_head **iter);
8b5be856 3619
44a40855
VY
3620/* iterate through upper list, must be called under RCU read lock */
3621#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3622 for (iter = &(dev)->adj_list.upper, \
3623 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3624 updev; \
3625 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3626
8b5be856 3627/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3628#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3629 for (iter = &(dev)->all_adj_list.upper, \
3630 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3631 updev; \
3632 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3633
f629d208
JP
3634void *netdev_lower_get_next_private(struct net_device *dev,
3635 struct list_head **iter);
3636void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3637 struct list_head **iter);
31088a11
VF
3638
3639#define netdev_for_each_lower_private(dev, priv, iter) \
3640 for (iter = (dev)->adj_list.lower.next, \
3641 priv = netdev_lower_get_next_private(dev, &(iter)); \
3642 priv; \
3643 priv = netdev_lower_get_next_private(dev, &(iter)))
3644
3645#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3646 for (iter = &(dev)->adj_list.lower, \
3647 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3648 priv; \
3649 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3650
4085ebe8
VY
3651void *netdev_lower_get_next(struct net_device *dev,
3652 struct list_head **iter);
3653#define netdev_for_each_lower_dev(dev, ldev, iter) \
3654 for (iter = &(dev)->adj_list.lower, \
3655 ldev = netdev_lower_get_next(dev, &(iter)); \
3656 ldev; \
3657 ldev = netdev_lower_get_next(dev, &(iter)))
3658
f629d208 3659void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3660void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3661struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3662struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3663int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3664int netdev_master_upper_dev_link(struct net_device *dev,
9ff162a8 3665 struct net_device *upper_dev);
f629d208
JP
3666int netdev_master_upper_dev_link_private(struct net_device *dev,
3667 struct net_device *upper_dev,
3668 void *private);
3669void netdev_upper_dev_unlink(struct net_device *dev,
3670 struct net_device *upper_dev);
5bb025fa 3671void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3672void *netdev_lower_dev_get_private(struct net_device *dev,
3673 struct net_device *lower_dev);
960fb622
ED
3674
3675/* RSS keys are 40 or 52 bytes long */
3676#define NETDEV_RSS_KEY_LEN 52
3677extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3678void netdev_rss_key_fill(void *buffer, size_t len);
3679
4085ebe8
VY
3680int dev_get_nest_level(struct net_device *dev,
3681 bool (*type_check)(struct net_device *dev));
f629d208
JP
3682int skb_checksum_help(struct sk_buff *skb);
3683struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3684 netdev_features_t features, bool tx_path);
3685struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3686 netdev_features_t features);
12b0004d 3687
61bd3857
MS
3688struct netdev_bonding_info {
3689 ifslave slave;
3690 ifbond master;
3691};
3692
3693struct netdev_notifier_bonding_info {
3694 struct netdev_notifier_info info; /* must be first */
3695 struct netdev_bonding_info bonding_info;
3696};
3697
3698void netdev_bonding_info_change(struct net_device *dev,
3699 struct netdev_bonding_info *bonding_info);
3700
12b0004d
CW
3701static inline
3702struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3703{
3704 return __skb_gso_segment(skb, features, true);
3705}
53d6471c 3706__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3707
3708static inline bool can_checksum_protocol(netdev_features_t features,
3709 __be16 protocol)
3710{
3711 return ((features & NETIF_F_GEN_CSUM) ||
3712 ((features & NETIF_F_V4_CSUM) &&
3713 protocol == htons(ETH_P_IP)) ||
3714 ((features & NETIF_F_V6_CSUM) &&
3715 protocol == htons(ETH_P_IPV6)) ||
3716 ((features & NETIF_F_FCOE_CRC) &&
3717 protocol == htons(ETH_P_FCOE)));
3718}
12b0004d 3719
fb286bb2 3720#ifdef CONFIG_BUG
f629d208 3721void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3722#else
3723static inline void netdev_rx_csum_fault(struct net_device *dev)
3724{
3725}
3726#endif
1da177e4 3727/* rx skb timestamps */
f629d208
JP
3728void net_enable_timestamp(void);
3729void net_disable_timestamp(void);
1da177e4 3730
20380731 3731#ifdef CONFIG_PROC_FS
f629d208 3732int __init dev_proc_init(void);
900ff8c6
CW
3733#else
3734#define dev_proc_init() 0
20380731
ACM
3735#endif
3736
4798248e 3737static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3738 struct sk_buff *skb, struct net_device *dev,
3739 bool more)
4798248e 3740{
fa2dbdc2 3741 skb->xmit_more = more ? 1 : 0;
0b725a2c 3742 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3743}
3744
10b3ad8c 3745static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3746 struct netdev_queue *txq, bool more)
4798248e
DM
3747{
3748 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3749 int rc;
4798248e 3750
fa2dbdc2 3751 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3752 if (rc == NETDEV_TX_OK)
3753 txq_trans_update(txq);
3754
3755 return rc;
4798248e
DM
3756}
3757
42a2d923
LT
3758int netdev_class_create_file_ns(struct class_attribute *class_attr,
3759 const void *ns);
3760void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3761 const void *ns);
58292cbe
TH
3762
3763static inline int netdev_class_create_file(struct class_attribute *class_attr)
3764{
3765 return netdev_class_create_file_ns(class_attr, NULL);
3766}
3767
3768static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3769{
3770 netdev_class_remove_file_ns(class_attr, NULL);
3771}
b8a9787e 3772
04600794
JB
3773extern struct kobj_ns_type_operations net_ns_type_operations;
3774
f629d208 3775const char *netdev_drivername(const struct net_device *dev);
6579e57b 3776
f629d208 3777void linkwatch_run_queue(void);
20380731 3778
da08143b
MK
3779static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3780 netdev_features_t f2)
3781{
3782 if (f1 & NETIF_F_GEN_CSUM)
3783 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3784 if (f2 & NETIF_F_GEN_CSUM)
3785 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3786 f1 &= f2;
3787 if (f1 & NETIF_F_GEN_CSUM)
3788 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3789
3790 return f1;
3791}
3792
c8f44aff
MM
3793static inline netdev_features_t netdev_get_wanted_features(
3794 struct net_device *dev)
5455c699
MM
3795{
3796 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3797}
c8f44aff
MM
3798netdev_features_t netdev_increment_features(netdev_features_t all,
3799 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3800
3801/* Allow TSO being used on stacked device :
3802 * Performing the GSO segmentation before last device
3803 * is a performance improvement.
3804 */
3805static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3806 netdev_features_t mask)
3807{
3808 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3809}
3810
6cb6a27c 3811int __netdev_update_features(struct net_device *dev);
5455c699 3812void netdev_update_features(struct net_device *dev);
afe12cc8 3813void netdev_change_features(struct net_device *dev);
7f353bf2 3814
fc4a7489
PM
3815void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3816 struct net_device *dev);
3817
e38f3025
TM
3818netdev_features_t passthru_features_check(struct sk_buff *skb,
3819 struct net_device *dev,
3820 netdev_features_t features);
c1e756bf 3821netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3822
4d29515f 3823static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3824{
c8f44aff 3825 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3826
3827 /* check flags correspondence */
3828 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3829 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3830 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3831 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3832 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3833 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
3834 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3835 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3836 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3837 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3838 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3839 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
e585f236 3840 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
0345e186 3841
d6b4991a 3842 return (features & feature) == feature;
576a30eb
HX
3843}
3844
4d29515f 3845static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3846{
278b2513 3847 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3848 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3849}
3850
8b86a61d 3851static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 3852 netdev_features_t features)
7967168c 3853{
fc741216 3854 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3855 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3856 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3857}
3858
82cc1a7a
PWJ
3859static inline void netif_set_gso_max_size(struct net_device *dev,
3860 unsigned int size)
3861{
3862 dev->gso_max_size = size;
3863}
3864
7a7ffbab
WCC
3865static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3866 int pulled_hlen, u16 mac_offset,
3867 int mac_len)
3868{
3869 skb->protocol = protocol;
3870 skb->encapsulation = 1;
3871 skb_push(skb, pulled_hlen);
3872 skb_reset_transport_header(skb);
3873 skb->mac_header = mac_offset;
3874 skb->network_header = skb->mac_header + mac_len;
3875 skb->mac_len = mac_len;
3876}
3877
a6cc0cfa
JF
3878static inline bool netif_is_macvlan(struct net_device *dev)
3879{
3880 return dev->priv_flags & IFF_MACVLAN;
3881}
3882
2f33e7d5
MB
3883static inline bool netif_is_macvlan_port(struct net_device *dev)
3884{
3885 return dev->priv_flags & IFF_MACVLAN_PORT;
3886}
3887
5933fea7
MB
3888static inline bool netif_is_ipvlan(struct net_device *dev)
3889{
3890 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3891}
3892
3893static inline bool netif_is_ipvlan_port(struct net_device *dev)
3894{
3895 return dev->priv_flags & IFF_IPVLAN_MASTER;
3896}
3897
8a7fbfab 3898static inline bool netif_is_bond_master(struct net_device *dev)
3899{
3900 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3901}
3902
4d29515f 3903static inline bool netif_is_bond_slave(struct net_device *dev)
1765a575
JP
3904{
3905 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3906}
3907
3bdc0eba
BG
3908static inline bool netif_supports_nofcs(struct net_device *dev)
3909{
3910 return dev->priv_flags & IFF_SUPP_NOFCS;
3911}
3912
007979ea 3913static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 3914{
007979ea 3915 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
3916}
3917
fee6d4c7
DA
3918static inline bool netif_is_l3_slave(const struct net_device *dev)
3919{
3920 return dev->priv_flags & IFF_L3MDEV_SLAVE;
3921}
3922
0894ae3f
JP
3923static inline bool netif_is_bridge_master(const struct net_device *dev)
3924{
3925 return dev->priv_flags & IFF_EBRIDGE;
3926}
3927
28f9ee22
VY
3928static inline bool netif_is_bridge_port(const struct net_device *dev)
3929{
3930 return dev->priv_flags & IFF_BRIDGE_PORT;
3931}
3932
35d4e172
JP
3933static inline bool netif_is_ovs_master(const struct net_device *dev)
3934{
3935 return dev->priv_flags & IFF_OPENVSWITCH;
3936}
3937
85c5f09b
JP
3938static inline bool netif_is_team_master(struct net_device *dev)
3939{
3940 return dev->priv_flags & IFF_TEAM;
3941}
3942
84783568
JP
3943static inline bool netif_is_team_port(struct net_device *dev)
3944{
3945 return dev->priv_flags & IFF_TEAM_PORT;
3946}
3947
394833da
JP
3948static inline bool netif_is_lag_master(struct net_device *dev)
3949{
3950 return netif_is_bond_master(dev) || netif_is_team_master(dev);
3951}
3952
ad4c7498
JP
3953static inline bool netif_is_lag_port(struct net_device *dev)
3954{
3955 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
3956}
3957
c023158d
KJ
3958static inline bool netif_is_rxfh_configured(const struct net_device *dev)
3959{
3960 return dev->priv_flags & IFF_RXFH_CONFIGURED;
3961}
3962
02875878
ED
3963/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3964static inline void netif_keep_dst(struct net_device *dev)
3965{
3966 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3967}
3968
505d4f73 3969extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 3970
571ba423
JP
3971/* Logging, debugging and troubleshooting/diagnostic helpers. */
3972
3973/* netdev_printk helpers, similar to dev_printk */
3974
3975static inline const char *netdev_name(const struct net_device *dev)
3976{
c6f854d5
VF
3977 if (!dev->name[0] || strchr(dev->name, '%'))
3978 return "(unnamed net_device)";
571ba423
JP
3979 return dev->name;
3980}
3981
ccc7f496
VF
3982static inline const char *netdev_reg_state(const struct net_device *dev)
3983{
3984 switch (dev->reg_state) {
3985 case NETREG_UNINITIALIZED: return " (uninitialized)";
3986 case NETREG_REGISTERED: return "";
3987 case NETREG_UNREGISTERING: return " (unregistering)";
3988 case NETREG_UNREGISTERED: return " (unregistered)";
3989 case NETREG_RELEASED: return " (released)";
3990 case NETREG_DUMMY: return " (dummy)";
3991 }
3992
3993 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3994 return " (unknown)";
3995}
3996
f629d208 3997__printf(3, 4)
6ea754eb
JP
3998void netdev_printk(const char *level, const struct net_device *dev,
3999 const char *format, ...);
f629d208 4000__printf(2, 3)
6ea754eb 4001void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4002__printf(2, 3)
6ea754eb 4003void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4004__printf(2, 3)
6ea754eb 4005void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4006__printf(2, 3)
6ea754eb 4007void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4008__printf(2, 3)
6ea754eb 4009void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4010__printf(2, 3)
6ea754eb 4011void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4012__printf(2, 3)
6ea754eb 4013void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4014
8909c9ad
VK
4015#define MODULE_ALIAS_NETDEV(device) \
4016 MODULE_ALIAS("netdev-" device)
4017
b558c96f 4018#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4019#define netdev_dbg(__dev, format, args...) \
4020do { \
ffa10cb4 4021 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4022} while (0)
b558c96f
JC
4023#elif defined(DEBUG)
4024#define netdev_dbg(__dev, format, args...) \
4025 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4026#else
4027#define netdev_dbg(__dev, format, args...) \
4028({ \
4029 if (0) \
4030 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4031})
4032#endif
4033
4034#if defined(VERBOSE_DEBUG)
4035#define netdev_vdbg netdev_dbg
4036#else
4037
4038#define netdev_vdbg(dev, format, args...) \
4039({ \
4040 if (0) \
4041 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4042 0; \
4043})
4044#endif
4045
4046/*
4047 * netdev_WARN() acts like dev_printk(), but with the key difference
4048 * of using a WARN/WARN_ON to get the message out, including the
4049 * file/line information and a backtrace.
4050 */
4051#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4052 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4053 netdev_reg_state(dev), ##args)
571ba423 4054
b3d95c5c
JP
4055/* netif printk helpers, similar to netdev_printk */
4056
4057#define netif_printk(priv, type, level, dev, fmt, args...) \
4058do { \
4059 if (netif_msg_##type(priv)) \
4060 netdev_printk(level, (dev), fmt, ##args); \
4061} while (0)
4062
f45f4321
JP
4063#define netif_level(level, priv, type, dev, fmt, args...) \
4064do { \
4065 if (netif_msg_##type(priv)) \
4066 netdev_##level(dev, fmt, ##args); \
4067} while (0)
4068
b3d95c5c 4069#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4070 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4071#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4072 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4073#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4074 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4075#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4076 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4077#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4078 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4079#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4080 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4081#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4082 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4083
0053ea9c 4084#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4085#define netif_dbg(priv, type, netdev, format, args...) \
4086do { \
4087 if (netif_msg_##type(priv)) \
b5fb0a03 4088 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4089} while (0)
0053ea9c
JP
4090#elif defined(DEBUG)
4091#define netif_dbg(priv, type, dev, format, args...) \
4092 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4093#else
4094#define netif_dbg(priv, type, dev, format, args...) \
4095({ \
4096 if (0) \
4097 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4098 0; \
4099})
4100#endif
4101
4102#if defined(VERBOSE_DEBUG)
bcfcc450 4103#define netif_vdbg netif_dbg
b3d95c5c
JP
4104#else
4105#define netif_vdbg(priv, type, dev, format, args...) \
4106({ \
4107 if (0) \
a4ed89cb 4108 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4109 0; \
4110})
4111#endif
571ba423 4112
900ff8c6
CW
4113/*
4114 * The list of packet types we will receive (as opposed to discard)
4115 * and the routines to invoke.
4116 *
4117 * Why 16. Because with 16 the only overlap we get on a hash of the
4118 * low nibble of the protocol value is RARP/SNAP/X.25.
4119 *
4120 * NOTE: That is no longer true with the addition of VLAN tags. Not
4121 * sure which should go first, but I bet it won't make much
4122 * difference if we are running VLANs. The good news is that
4123 * this protocol won't be in the list unless compiled in, so
4124 * the average user (w/out VLANs) will not be adversely affected.
4125 * --BLG
4126 *
4127 * 0800 IP
4128 * 8100 802.1Q VLAN
4129 * 0001 802.3
4130 * 0002 AX.25
4131 * 0004 802.2
4132 * 8035 RARP
4133 * 0005 SNAP
4134 * 0805 X.25
4135 * 0806 ARP
4136 * 8137 IPX
4137 * 0009 Localtalk
4138 * 86DD IPv6
4139 */
4140#define PTYPE_HASH_SIZE (16)
4141#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4142
385a154c 4143#endif /* _LINUX_NETDEVICE_H */