]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
tcp: Fix conditions to determine checksum offload
[mirror_ubuntu-artful-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
cf85d08f 44#include <net/dsa.h>
7a6b6f51 45#ifdef CONFIG_DCB
2f90b865
AD
46#include <net/dcbnl.h>
47#endif
5bc1421e 48#include <net/netprio_cgroup.h>
a050c33f 49
a59e2ecb 50#include <linux/netdev_features.h>
77162022 51#include <linux/neighbour.h>
607ca46e 52#include <uapi/linux/netdevice.h>
61bd3857 53#include <uapi/linux/if_bonding.h>
a59e2ecb 54
115c1d6e 55struct netpoll_info;
313162d0 56struct device;
c1f19b51 57struct phy_device;
704232c2
JB
58/* 802.11 specific */
59struct wireless_dev;
98a18b6f
AA
60/* 802.15.4 specific */
61struct wpan_dev;
03c57747 62struct mpls_dev;
1da177e4 63
f629d208
JP
64void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
d07d7507 66
9a1654ba
JP
67/* Backlog congestion levels */
68#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69#define NET_RX_DROP 1 /* packet dropped */
70
572a9d7b
PM
71/*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88/* qdisc ->enqueue() return codes. */
89#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
90#define NET_XMIT_DROP 0x01 /* skb dropped */
91#define NET_XMIT_CN 0x02 /* congestion notification */
92#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 94
b9df3cb8
GR
95/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 98#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
99#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
dc1f8bf6 101/* Driver transmit return codes */
9a1654ba 102#define NETDEV_TX_MASK 0xf0
572a9d7b 103
dc1f8bf6 104enum netdev_tx {
572a9d7b 105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
dc1f8bf6
SH
109};
110typedef enum netdev_tx netdev_tx_t;
111
9a1654ba
JP
112/*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116static inline bool dev_xmit_complete(int rc)
117{
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128}
129
1da177e4
LT
130/*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
fe2918b0 134
c0eb4540
KS
135#if defined(CONFIG_HYPERV_NET)
136# define LL_MAX_HEADER 128
137#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
138# if defined(CONFIG_MAC80211_MESH)
139# define LL_MAX_HEADER 128
140# else
141# define LL_MAX_HEADER 96
142# endif
1da177e4 143#else
8388e3da 144# define LL_MAX_HEADER 32
1da177e4
LT
145#endif
146
d11ead75
BH
147#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
148 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
149#define MAX_HEADER LL_MAX_HEADER
150#else
151#define MAX_HEADER (LL_MAX_HEADER + 48)
152#endif
153
154/*
be1f3c2c
BH
155 * Old network device statistics. Fields are native words
156 * (unsigned long) so they can be read and written atomically.
1da177e4 157 */
fe2918b0 158
d94d9fee 159struct net_device_stats {
3cfde79c
BH
160 unsigned long rx_packets;
161 unsigned long tx_packets;
162 unsigned long rx_bytes;
163 unsigned long tx_bytes;
164 unsigned long rx_errors;
165 unsigned long tx_errors;
166 unsigned long rx_dropped;
167 unsigned long tx_dropped;
168 unsigned long multicast;
1da177e4 169 unsigned long collisions;
1da177e4 170 unsigned long rx_length_errors;
3cfde79c
BH
171 unsigned long rx_over_errors;
172 unsigned long rx_crc_errors;
173 unsigned long rx_frame_errors;
174 unsigned long rx_fifo_errors;
175 unsigned long rx_missed_errors;
1da177e4
LT
176 unsigned long tx_aborted_errors;
177 unsigned long tx_carrier_errors;
178 unsigned long tx_fifo_errors;
179 unsigned long tx_heartbeat_errors;
180 unsigned long tx_window_errors;
1da177e4
LT
181 unsigned long rx_compressed;
182 unsigned long tx_compressed;
183};
184
1da177e4
LT
185
186#include <linux/cache.h>
187#include <linux/skbuff.h>
188
adc9300e 189#ifdef CONFIG_RPS
c5905afb
IM
190#include <linux/static_key.h>
191extern struct static_key rps_needed;
adc9300e
ED
192#endif
193
1da177e4
LT
194struct neighbour;
195struct neigh_parms;
196struct sk_buff;
197
f001fde5
JP
198struct netdev_hw_addr {
199 struct list_head list;
200 unsigned char addr[MAX_ADDR_LEN];
201 unsigned char type;
ccffad25
JP
202#define NETDEV_HW_ADDR_T_LAN 1
203#define NETDEV_HW_ADDR_T_SAN 2
204#define NETDEV_HW_ADDR_T_SLAVE 3
205#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 206#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 207 bool global_use;
4cd729b0 208 int sync_cnt;
8f8f103d 209 int refcount;
4543fbef 210 int synced;
f001fde5
JP
211 struct rcu_head rcu_head;
212};
213
31278e71
JP
214struct netdev_hw_addr_list {
215 struct list_head list;
216 int count;
217};
218
22bedad3
JP
219#define netdev_hw_addr_list_count(l) ((l)->count)
220#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
221#define netdev_hw_addr_list_for_each(ha, l) \
222 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 223
22bedad3
JP
224#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
225#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
226#define netdev_for_each_uc_addr(ha, dev) \
227 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 228
22bedad3
JP
229#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
230#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 231#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 232 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 233
d94d9fee 234struct hh_cache {
f6b72b62 235 u16 hh_len;
5c25f686 236 u16 __pad;
3644f0ce 237 seqlock_t hh_lock;
1da177e4
LT
238
239 /* cached hardware header; allow for machine alignment needs. */
240#define HH_DATA_MOD 16
241#define HH_DATA_OFF(__len) \
5ba0eac6 242 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
243#define HH_DATA_ALIGN(__len) \
244 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
245 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
246};
247
248/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
249 * Alternative is:
250 * dev->hard_header_len ? (dev->hard_header_len +
251 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
252 *
253 * We could use other alignment values, but we must maintain the
254 * relationship HH alignment <= LL alignment.
255 */
256#define LL_RESERVED_SPACE(dev) \
f5184d26 257 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 258#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 259 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 260
3b04ddde
SH
261struct header_ops {
262 int (*create) (struct sk_buff *skb, struct net_device *dev,
263 unsigned short type, const void *daddr,
95c96174 264 const void *saddr, unsigned int len);
3b04ddde 265 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 266 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
267 void (*cache_update)(struct hh_cache *hh,
268 const struct net_device *dev,
269 const unsigned char *haddr);
270};
271
1da177e4
LT
272/* These flag bits are private to the generic network queueing
273 * layer, they may not be explicitly referenced by any other
274 * code.
275 */
276
d94d9fee 277enum netdev_state_t {
1da177e4
LT
278 __LINK_STATE_START,
279 __LINK_STATE_PRESENT,
1da177e4 280 __LINK_STATE_NOCARRIER,
b00055aa
SR
281 __LINK_STATE_LINKWATCH_PENDING,
282 __LINK_STATE_DORMANT,
1da177e4
LT
283};
284
285
286/*
287 * This structure holds at boot time configured netdevice settings. They
fe2918b0 288 * are then used in the device probing.
1da177e4
LT
289 */
290struct netdev_boot_setup {
291 char name[IFNAMSIZ];
292 struct ifmap map;
293};
294#define NETDEV_BOOT_SETUP_MAX 8
295
f629d208 296int __init netdev_boot_setup(char *str);
1da177e4 297
bea3348e
SH
298/*
299 * Structure for NAPI scheduling similar to tasklet but with weighting
300 */
301struct napi_struct {
302 /* The poll_list must only be managed by the entity which
303 * changes the state of the NAPI_STATE_SCHED bit. This means
304 * whoever atomically sets that bit can add this napi_struct
305 * to the per-cpu poll_list, and whoever clears that bit
306 * can remove from the list right before clearing the bit.
307 */
308 struct list_head poll_list;
309
310 unsigned long state;
311 int weight;
404f7c9e 312 unsigned int gro_count;
bea3348e
SH
313 int (*poll)(struct napi_struct *, int);
314#ifdef CONFIG_NETPOLL
315 spinlock_t poll_lock;
316 int poll_owner;
bea3348e 317#endif
5d38a079 318 struct net_device *dev;
d565b0a1 319 struct sk_buff *gro_list;
5d38a079 320 struct sk_buff *skb;
3b47d303 321 struct hrtimer timer;
404f7c9e 322 struct list_head dev_list;
af12fa6e
ET
323 struct hlist_node napi_hash_node;
324 unsigned int napi_id;
bea3348e
SH
325};
326
d94d9fee 327enum {
bea3348e 328 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 329 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 330 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
331 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
332 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
bea3348e
SH
333};
334
5b252f0c 335enum gro_result {
d1c76af9
HX
336 GRO_MERGED,
337 GRO_MERGED_FREE,
338 GRO_HELD,
339 GRO_NORMAL,
340 GRO_DROP,
341};
5b252f0c 342typedef enum gro_result gro_result_t;
d1c76af9 343
8a4eb573
JP
344/*
345 * enum rx_handler_result - Possible return values for rx_handlers.
346 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
347 * further.
348 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
349 * case skb->dev was changed by rx_handler.
350 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
351 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
352 *
353 * rx_handlers are functions called from inside __netif_receive_skb(), to do
354 * special processing of the skb, prior to delivery to protocol handlers.
355 *
356 * Currently, a net_device can only have a single rx_handler registered. Trying
357 * to register a second rx_handler will return -EBUSY.
358 *
359 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
360 * To unregister a rx_handler on a net_device, use
361 * netdev_rx_handler_unregister().
362 *
363 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
364 * do with the skb.
365 *
366 * If the rx_handler consumed to skb in some way, it should return
367 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
368 * the skb to be delivered in some other ways.
369 *
370 * If the rx_handler changed skb->dev, to divert the skb to another
371 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
372 * new device will be called if it exists.
373 *
374 * If the rx_handler consider the skb should be ignored, it should return
375 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 376 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
377 *
378 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
379 * delivered, it should return RX_HANDLER_PASS.
380 *
381 * A device without a registered rx_handler will behave as if rx_handler
382 * returned RX_HANDLER_PASS.
383 */
384
385enum rx_handler_result {
386 RX_HANDLER_CONSUMED,
387 RX_HANDLER_ANOTHER,
388 RX_HANDLER_EXACT,
389 RX_HANDLER_PASS,
390};
391typedef enum rx_handler_result rx_handler_result_t;
392typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 393
f629d208 394void __napi_schedule(struct napi_struct *n);
bc9ad166 395void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 396
4d29515f 397static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
398{
399 return test_bit(NAPI_STATE_DISABLE, &n->state);
400}
401
bea3348e
SH
402/**
403 * napi_schedule_prep - check if napi can be scheduled
404 * @n: napi context
405 *
406 * Test if NAPI routine is already running, and if not mark
407 * it as running. This is used as a condition variable
a0a46196
DM
408 * insure only one NAPI poll instance runs. We also make
409 * sure there is no pending NAPI disable.
bea3348e 410 */
4d29515f 411static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 412{
a0a46196
DM
413 return !napi_disable_pending(n) &&
414 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
415}
416
417/**
418 * napi_schedule - schedule NAPI poll
419 * @n: napi context
420 *
421 * Schedule NAPI poll routine to be called if it is not already
422 * running.
423 */
424static inline void napi_schedule(struct napi_struct *n)
425{
426 if (napi_schedule_prep(n))
427 __napi_schedule(n);
428}
429
bc9ad166
ED
430/**
431 * napi_schedule_irqoff - schedule NAPI poll
432 * @n: napi context
433 *
434 * Variant of napi_schedule(), assuming hard irqs are masked.
435 */
436static inline void napi_schedule_irqoff(struct napi_struct *n)
437{
438 if (napi_schedule_prep(n))
439 __napi_schedule_irqoff(n);
440}
441
bfe13f54 442/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 443static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
444{
445 if (napi_schedule_prep(napi)) {
446 __napi_schedule(napi);
4d29515f 447 return true;
bfe13f54 448 }
4d29515f 449 return false;
bfe13f54
RD
450}
451
3b47d303
ED
452void __napi_complete(struct napi_struct *n);
453void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
454/**
455 * napi_complete - NAPI processing complete
456 * @n: napi context
457 *
458 * Mark NAPI processing as complete.
3b47d303 459 * Consider using napi_complete_done() instead.
bea3348e 460 */
3b47d303
ED
461static inline void napi_complete(struct napi_struct *n)
462{
463 return napi_complete_done(n, 0);
464}
bea3348e 465
af12fa6e
ET
466/**
467 * napi_hash_add - add a NAPI to global hashtable
468 * @napi: napi context
469 *
470 * generate a new napi_id and store a @napi under it in napi_hash
93d05d4a
ED
471 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL)
472 * Note: This is normally automatically done from netif_napi_add(),
473 * so might disappear in a future linux version.
af12fa6e 474 */
f629d208 475void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
476
477/**
478 * napi_hash_del - remove a NAPI from global table
479 * @napi: napi context
480 *
481 * Warning: caller must observe rcu grace period
34cbe27e
ED
482 * before freeing memory containing @napi, if
483 * this function returns true.
93d05d4a
ED
484 * Note: core networking stack automatically calls it
485 * from netif_napi_del()
486 * Drivers might want to call this helper to combine all
487 * the needed rcu grace periods into a single one.
af12fa6e 488 */
34cbe27e 489bool napi_hash_del(struct napi_struct *napi);
af12fa6e 490
bea3348e
SH
491/**
492 * napi_disable - prevent NAPI from scheduling
493 * @n: napi context
494 *
495 * Stop NAPI from being scheduled on this context.
496 * Waits till any outstanding processing completes.
497 */
3b47d303 498void napi_disable(struct napi_struct *n);
bea3348e
SH
499
500/**
501 * napi_enable - enable NAPI scheduling
502 * @n: napi context
503 *
504 * Resume NAPI from being scheduled on this context.
505 * Must be paired with napi_disable.
506 */
507static inline void napi_enable(struct napi_struct *n)
508{
509 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 510 smp_mb__before_atomic();
bea3348e 511 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 512 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
513}
514
c264c3de
SH
515#ifdef CONFIG_SMP
516/**
517 * napi_synchronize - wait until NAPI is not running
518 * @n: napi context
519 *
520 * Wait until NAPI is done being scheduled on this context.
521 * Waits till any outstanding processing completes but
522 * does not disable future activations.
523 */
524static inline void napi_synchronize(const struct napi_struct *n)
525{
526 while (test_bit(NAPI_STATE_SCHED, &n->state))
527 msleep(1);
528}
529#else
530# define napi_synchronize(n) barrier()
531#endif
532
d94d9fee 533enum netdev_queue_state_t {
73466498
TH
534 __QUEUE_STATE_DRV_XOFF,
535 __QUEUE_STATE_STACK_XOFF,
c3f26a26 536 __QUEUE_STATE_FROZEN,
79d16385 537};
8e2f1a63
DB
538
539#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
540#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
541#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
542
543#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
544#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
545 QUEUE_STATE_FROZEN)
546#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
547 QUEUE_STATE_FROZEN)
548
73466498
TH
549/*
550 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
551 * netif_tx_* functions below are used to manipulate this flag. The
552 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
553 * queue independently. The netif_xmit_*stopped functions below are called
554 * to check if the queue has been stopped by the driver or stack (either
555 * of the XOFF bits are set in the state). Drivers should not need to call
556 * netif_xmit*stopped functions, they should only be using netif_tx_*.
557 */
79d16385 558
bb949fbd 559struct netdev_queue {
6a321cb3
ED
560/*
561 * read mostly part
562 */
bb949fbd 563 struct net_device *dev;
46e5da40 564 struct Qdisc __rcu *qdisc;
b0e1e646 565 struct Qdisc *qdisc_sleeping;
ccf5ff69 566#ifdef CONFIG_SYSFS
1d24eb48
TH
567 struct kobject kobj;
568#endif
f2cd2d3e
ED
569#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
570 int numa_node;
571#endif
6a321cb3
ED
572/*
573 * write mostly part
574 */
575 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
576 int xmit_lock_owner;
9d21493b
ED
577 /*
578 * please use this field instead of dev->trans_start
579 */
580 unsigned long trans_start;
ccf5ff69 581
582 /*
583 * Number of TX timeouts for this queue
584 * (/sys/class/net/DEV/Q/trans_timeout)
585 */
586 unsigned long trans_timeout;
114cf580
TH
587
588 unsigned long state;
589
590#ifdef CONFIG_BQL
591 struct dql dql;
592#endif
822b3b2e 593 unsigned long tx_maxrate;
e8a0464c 594} ____cacheline_aligned_in_smp;
bb949fbd 595
f2cd2d3e
ED
596static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
597{
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 return q->numa_node;
600#else
b236da69 601 return NUMA_NO_NODE;
f2cd2d3e
ED
602#endif
603}
604
605static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
606{
607#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 q->numa_node = node;
609#endif
610}
611
df334545 612#ifdef CONFIG_RPS
0a9627f2
TH
613/*
614 * This structure holds an RPS map which can be of variable length. The
615 * map is an array of CPUs.
616 */
617struct rps_map {
618 unsigned int len;
619 struct rcu_head rcu;
620 u16 cpus[0];
621};
60b778ce 622#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 623
fec5e652 624/*
c445477d
BH
625 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
626 * tail pointer for that CPU's input queue at the time of last enqueue, and
627 * a hardware filter index.
fec5e652
TH
628 */
629struct rps_dev_flow {
630 u16 cpu;
c445477d 631 u16 filter;
fec5e652
TH
632 unsigned int last_qtail;
633};
c445477d 634#define RPS_NO_FILTER 0xffff
fec5e652
TH
635
636/*
637 * The rps_dev_flow_table structure contains a table of flow mappings.
638 */
639struct rps_dev_flow_table {
640 unsigned int mask;
641 struct rcu_head rcu;
fec5e652
TH
642 struct rps_dev_flow flows[0];
643};
644#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 645 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
646
647/*
648 * The rps_sock_flow_table contains mappings of flows to the last CPU
649 * on which they were processed by the application (set in recvmsg).
567e4b79
ED
650 * Each entry is a 32bit value. Upper part is the high order bits
651 * of flow hash, lower part is cpu number.
652 * rps_cpu_mask is used to partition the space, depending on number of
653 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
654 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
655 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
656 */
657struct rps_sock_flow_table {
567e4b79 658 u32 mask;
93c1af6c
ED
659
660 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 661};
567e4b79 662#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
663
664#define RPS_NO_CPU 0xffff
665
567e4b79
ED
666extern u32 rps_cpu_mask;
667extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
668
fec5e652
TH
669static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
670 u32 hash)
671{
672 if (table && hash) {
567e4b79
ED
673 unsigned int index = hash & table->mask;
674 u32 val = hash & ~rps_cpu_mask;
fec5e652
TH
675
676 /* We only give a hint, preemption can change cpu under us */
567e4b79 677 val |= raw_smp_processor_id();
fec5e652 678
567e4b79
ED
679 if (table->ents[index] != val)
680 table->ents[index] = val;
fec5e652
TH
681 }
682}
683
c445477d 684#ifdef CONFIG_RFS_ACCEL
f629d208
JP
685bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
686 u16 filter_id);
c445477d 687#endif
a953be53 688#endif /* CONFIG_RPS */
c445477d 689
0a9627f2
TH
690/* This structure contains an instance of an RX queue. */
691struct netdev_rx_queue {
a953be53 692#ifdef CONFIG_RPS
6e3f7faf
ED
693 struct rps_map __rcu *rps_map;
694 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 695#endif
6e3f7faf 696 struct kobject kobj;
fe822240 697 struct net_device *dev;
0a9627f2 698} ____cacheline_aligned_in_smp;
a953be53
MD
699
700/*
701 * RX queue sysfs structures and functions.
702 */
703struct rx_queue_attribute {
704 struct attribute attr;
705 ssize_t (*show)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, char *buf);
707 ssize_t (*store)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, const char *buf, size_t len);
709};
d314774c 710
bf264145
TH
711#ifdef CONFIG_XPS
712/*
713 * This structure holds an XPS map which can be of variable length. The
714 * map is an array of queues.
715 */
716struct xps_map {
717 unsigned int len;
718 unsigned int alloc_len;
719 struct rcu_head rcu;
720 u16 queues[0];
721};
60b778ce 722#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
723#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
724 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
725
726/*
727 * This structure holds all XPS maps for device. Maps are indexed by CPU.
728 */
729struct xps_dev_maps {
730 struct rcu_head rcu;
a4177869 731 struct xps_map __rcu *cpu_map[0];
bf264145
TH
732};
733#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
734 (nr_cpu_ids * sizeof(struct xps_map *)))
735#endif /* CONFIG_XPS */
736
4f57c087
JF
737#define TC_MAX_QUEUE 16
738#define TC_BITMASK 15
739/* HW offloaded queuing disciplines txq count and offset maps */
740struct netdev_tc_txq {
741 u16 count;
742 u16 offset;
743};
744
68bad94e
NP
745#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
746/*
747 * This structure is to hold information about the device
748 * configured to run FCoE protocol stack.
749 */
750struct netdev_fcoe_hbainfo {
751 char manufacturer[64];
752 char serial_number[64];
753 char hardware_version[64];
754 char driver_version[64];
755 char optionrom_version[64];
756 char firmware_version[64];
757 char model[256];
758 char model_description[256];
759};
760#endif
761
02637fce 762#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 763
02637fce
JP
764/* This structure holds a unique identifier to identify some
765 * physical item (port for example) used by a netdevice.
66b52b0d 766 */
02637fce
JP
767struct netdev_phys_item_id {
768 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
769 unsigned char id_len;
770};
771
d754f98b
SF
772static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
773 struct netdev_phys_item_id *b)
774{
775 return a->id_len == b->id_len &&
776 memcmp(a->id, b->id, a->id_len) == 0;
777}
778
99932d4f
DB
779typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
780 struct sk_buff *skb);
781
d314774c
SH
782/*
783 * This structure defines the management hooks for network devices.
00829823
SH
784 * The following hooks can be defined; unless noted otherwise, they are
785 * optional and can be filled with a null pointer.
d314774c
SH
786 *
787 * int (*ndo_init)(struct net_device *dev);
788 * This function is called once when network device is registered.
789 * The network device can use this to any late stage initializaton
790 * or semantic validattion. It can fail with an error code which will
791 * be propogated back to register_netdev
792 *
793 * void (*ndo_uninit)(struct net_device *dev);
794 * This function is called when device is unregistered or when registration
795 * fails. It is not called if init fails.
796 *
797 * int (*ndo_open)(struct net_device *dev);
798 * This function is called when network device transistions to the up
799 * state.
800 *
801 * int (*ndo_stop)(struct net_device *dev);
802 * This function is called when network device transistions to the down
803 * state.
804 *
dc1f8bf6
SH
805 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
806 * struct net_device *dev);
00829823 807 * Called when a packet needs to be transmitted.
e79d8429
RR
808 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
809 * the queue before that can happen; it's for obsolete devices and weird
810 * corner cases, but the stack really does a non-trivial amount
811 * of useless work if you return NETDEV_TX_BUSY.
dc1f8bf6 812 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
813 * Required can not be NULL.
814 *
f663dd9a 815 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 816 * void *accel_priv, select_queue_fallback_t fallback);
00829823
SH
817 * Called to decide which queue to when device supports multiple
818 * transmit queues.
819 *
d314774c
SH
820 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
821 * This function is called to allow device receiver to make
822 * changes to configuration when multicast or promiscious is enabled.
823 *
824 * void (*ndo_set_rx_mode)(struct net_device *dev);
825 * This function is called device changes address list filtering.
01789349
JP
826 * If driver handles unicast address filtering, it should set
827 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
828 *
829 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
830 * This function is called when the Media Access Control address
37b607c5 831 * needs to be changed. If this interface is not defined, the
d314774c
SH
832 * mac address can not be changed.
833 *
834 * int (*ndo_validate_addr)(struct net_device *dev);
835 * Test if Media Access Control address is valid for the device.
836 *
837 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
838 * Called when a user request an ioctl which can't be handled by
839 * the generic interface code. If not defined ioctl's return
840 * not supported error code.
841 *
842 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
843 * Used to set network devices bus interface parameters. This interface
844 * is retained for legacy reason, new devices should use the bus
845 * interface (PCI) for low level management.
846 *
847 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
848 * Called when a user wants to change the Maximum Transfer Unit
849 * of a device. If not defined, any request to change MTU will
850 * will return an error.
851 *
00829823 852 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
853 * Callback uses when the transmitter has not made any progress
854 * for dev->watchdog ticks.
855 *
3cfde79c 856 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 857 * struct rtnl_link_stats64 *storage);
d308e38f 858 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 859 * Called when a user wants to get the network device usage
be1f3c2c 860 * statistics. Drivers must do one of the following:
3cfde79c
BH
861 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
862 * rtnl_link_stats64 structure passed by the caller.
82695d9b 863 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
864 * (which should normally be dev->stats) and return a pointer to
865 * it. The structure may be changed asynchronously only if each
866 * field is written atomically.
867 * 3. Update dev->stats asynchronously and atomically, and define
868 * neither operation.
d314774c 869 *
5d632cb7 870 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
871 * If device support VLAN filtering this function is called when a
872 * VLAN id is registered.
d314774c 873 *
5d632cb7 874 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
875 * If device support VLAN filtering this function is called when a
876 * VLAN id is unregistered.
d314774c
SH
877 *
878 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
879 *
880 * SR-IOV management functions.
881 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
882 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
883 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
884 * int max_tx_rate);
5f8444a3 885 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 886 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
887 * int (*ndo_get_vf_config)(struct net_device *dev,
888 * int vf, struct ifla_vf_info *ivf);
1d8faf48 889 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
890 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
891 * struct nlattr *port[]);
01a3d796
VZ
892 *
893 * Enable or disable the VF ability to query its RSS Redirection Table and
894 * Hash Key. This is needed since on some devices VF share this information
895 * with PF and querying it may adduce a theoretical security risk.
896 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 897 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
898 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
899 * Called to setup 'tc' number of traffic classes in the net device. This
900 * is always called from the stack with the rtnl lock held and netif tx
901 * queues stopped. This allows the netdevice to perform queue management
902 * safely.
c445477d 903 *
e9bce845
YZ
904 * Fiber Channel over Ethernet (FCoE) offload functions.
905 * int (*ndo_fcoe_enable)(struct net_device *dev);
906 * Called when the FCoE protocol stack wants to start using LLD for FCoE
907 * so the underlying device can perform whatever needed configuration or
908 * initialization to support acceleration of FCoE traffic.
909 *
910 * int (*ndo_fcoe_disable)(struct net_device *dev);
911 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
912 * so the underlying device can perform whatever needed clean-ups to
913 * stop supporting acceleration of FCoE traffic.
914 *
915 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
916 * struct scatterlist *sgl, unsigned int sgc);
917 * Called when the FCoE Initiator wants to initialize an I/O that
918 * is a possible candidate for Direct Data Placement (DDP). The LLD can
919 * perform necessary setup and returns 1 to indicate the device is set up
920 * successfully to perform DDP on this I/O, otherwise this returns 0.
921 *
922 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
923 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
924 * indicated by the FC exchange id 'xid', so the underlying device can
925 * clean up and reuse resources for later DDP requests.
926 *
927 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
928 * struct scatterlist *sgl, unsigned int sgc);
929 * Called when the FCoE Target wants to initialize an I/O that
930 * is a possible candidate for Direct Data Placement (DDP). The LLD can
931 * perform necessary setup and returns 1 to indicate the device is set up
932 * successfully to perform DDP on this I/O, otherwise this returns 0.
933 *
68bad94e
NP
934 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
935 * struct netdev_fcoe_hbainfo *hbainfo);
936 * Called when the FCoE Protocol stack wants information on the underlying
937 * device. This information is utilized by the FCoE protocol stack to
938 * register attributes with Fiber Channel management service as per the
939 * FC-GS Fabric Device Management Information(FDMI) specification.
940 *
e9bce845
YZ
941 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
942 * Called when the underlying device wants to override default World Wide
943 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
944 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
945 * protocol stack to use.
946 *
c445477d
BH
947 * RFS acceleration.
948 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
949 * u16 rxq_index, u32 flow_id);
950 * Set hardware filter for RFS. rxq_index is the target queue index;
951 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
952 * Return the filter ID on success, or a negative error code.
fbaec0ea 953 *
8b98a70c 954 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
955 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
956 * Called to make another netdev an underling.
957 *
958 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
959 * Called to release previously enslaved netdev.
5455c699
MM
960 *
961 * Feature/offload setting functions.
c8f44aff
MM
962 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
963 * netdev_features_t features);
5455c699
MM
964 * Adjusts the requested feature flags according to device-specific
965 * constraints, and returns the resulting flags. Must not modify
966 * the device state.
967 *
c8f44aff 968 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
969 * Called to update device configuration to new features. Passed
970 * feature set might be less than what was returned by ndo_fix_features()).
971 * Must return >0 or -errno if it changed dev->features itself.
972 *
edc7d573 973 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
974 * struct net_device *dev,
f6f6424b 975 * const unsigned char *addr, u16 vid, u16 flags)
77162022 976 * Adds an FDB entry to dev for addr.
1690be63
VY
977 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
978 * struct net_device *dev,
f6f6424b 979 * const unsigned char *addr, u16 vid)
77162022
JF
980 * Deletes the FDB entry from dev coresponding to addr.
981 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
982 * struct net_device *dev, struct net_device *filter_dev,
983 * int idx)
77162022
JF
984 * Used to add FDB entries to dump requests. Implementers should add
985 * entries to skb and update idx with the number of entries.
e5a55a89 986 *
ad41faa8
ND
987 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
988 * u16 flags)
e5a55a89 989 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
990 * struct net_device *dev, u32 filter_mask,
991 * int nlflags)
ad41faa8
ND
992 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
993 * u16 flags);
4bf84c35
JP
994 *
995 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
996 * Called to change device carrier. Soft-devices (like dummy, team, etc)
997 * which do not represent real hardware may define this to allow their
998 * userspace components to manage their virtual carrier state. Devices
999 * that determine carrier state from physical hardware properties (eg
1000 * network cables) or protocol-dependent mechanisms (eg
1001 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1002 *
1003 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1004 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1005 * Called to get ID of physical port of this device. If driver does
1006 * not implement this, it is assumed that the hw is not able to have
1007 * multiple net devices on single physical port.
53cf5275
JG
1008 *
1009 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 1010 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1011 * Called by vxlan to notiy a driver about the UDP port and socket
1012 * address family that vxlan is listnening to. It is called only when
1013 * a new port starts listening. The operation is protected by the
1014 * vxlan_net->sock_lock.
1015 *
1016 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 1017 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1018 * Called by vxlan to notify the driver about a UDP port and socket
1019 * address family that vxlan is not listening to anymore. The operation
1020 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
1021 *
1022 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1023 * struct net_device *dev)
1024 * Called by upper layer devices to accelerate switching or other
1025 * station functionality into hardware. 'pdev is the lowerdev
1026 * to use for the offload and 'dev' is the net device that will
1027 * back the offload. Returns a pointer to the private structure
1028 * the upper layer will maintain.
1029 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1030 * Called by upper layer device to delete the station created
1031 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1032 * the station and priv is the structure returned by the add
1033 * operation.
1034 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1035 * struct net_device *dev,
1036 * void *priv);
1037 * Callback to use for xmit over the accelerated station. This
1038 * is used in place of ndo_start_xmit on accelerated net
1039 * devices.
5f35227e
JG
1040 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1041 * struct net_device *dev
1042 * netdev_features_t features);
04ffcb25 1043 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1044 * performing offload operations on a given packet. This is to give
1045 * the device an opportunity to implement any restrictions that cannot
1046 * be otherwise expressed by feature flags. The check is called with
1047 * the set of features that the stack has calculated and it returns
1048 * those the driver believes to be appropriate.
822b3b2e
JF
1049 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1050 * int queue_index, u32 maxrate);
1051 * Called when a user wants to set a max-rate limitation of specific
1052 * TX queue.
a54acb3a
ND
1053 * int (*ndo_get_iflink)(const struct net_device *dev);
1054 * Called to get the iflink value of this device.
d746d707
AK
1055 * void (*ndo_change_proto_down)(struct net_device *dev,
1056 * bool proto_down);
1057 * This function is used to pass protocol port error state information
1058 * to the switch driver. The switch driver can react to the proto_down
1059 * by doing a phys down on the associated switch port.
fc4099f1
PS
1060 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1061 * This function is used to get egress tunnel information for given skb.
1062 * This is useful for retrieving outer tunnel header parameters while
1063 * sampling packet.
d746d707 1064 *
d314774c
SH
1065 */
1066struct net_device_ops {
1067 int (*ndo_init)(struct net_device *dev);
1068 void (*ndo_uninit)(struct net_device *dev);
1069 int (*ndo_open)(struct net_device *dev);
1070 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 1071 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
1072 struct net_device *dev);
1073 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1074 struct sk_buff *skb,
99932d4f
DB
1075 void *accel_priv,
1076 select_queue_fallback_t fallback);
d314774c
SH
1077 void (*ndo_change_rx_flags)(struct net_device *dev,
1078 int flags);
d314774c 1079 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1080 int (*ndo_set_mac_address)(struct net_device *dev,
1081 void *addr);
d314774c 1082 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1083 int (*ndo_do_ioctl)(struct net_device *dev,
1084 struct ifreq *ifr, int cmd);
d314774c
SH
1085 int (*ndo_set_config)(struct net_device *dev,
1086 struct ifmap *map);
00829823
SH
1087 int (*ndo_change_mtu)(struct net_device *dev,
1088 int new_mtu);
1089 int (*ndo_neigh_setup)(struct net_device *dev,
1090 struct neigh_parms *);
d314774c
SH
1091 void (*ndo_tx_timeout) (struct net_device *dev);
1092
28172739
ED
1093 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1094 struct rtnl_link_stats64 *storage);
d314774c
SH
1095 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1096
8e586137 1097 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1098 __be16 proto, u16 vid);
8e586137 1099 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1100 __be16 proto, u16 vid);
d314774c 1101#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1102 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1103 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1104 struct netpoll_info *info);
0e34e931 1105 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1106#endif
e0d1095a 1107#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1108 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1109#endif
95c26df8
WM
1110 int (*ndo_set_vf_mac)(struct net_device *dev,
1111 int queue, u8 *mac);
1112 int (*ndo_set_vf_vlan)(struct net_device *dev,
1113 int queue, u16 vlan, u8 qos);
ed616689
SC
1114 int (*ndo_set_vf_rate)(struct net_device *dev,
1115 int vf, int min_tx_rate,
1116 int max_tx_rate);
5f8444a3
GR
1117 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1118 int vf, bool setting);
dd461d6a
HS
1119 int (*ndo_set_vf_trust)(struct net_device *dev,
1120 int vf, bool setting);
95c26df8
WM
1121 int (*ndo_get_vf_config)(struct net_device *dev,
1122 int vf,
1123 struct ifla_vf_info *ivf);
1d8faf48
RE
1124 int (*ndo_set_vf_link_state)(struct net_device *dev,
1125 int vf, int link_state);
3b766cd8
EBE
1126 int (*ndo_get_vf_stats)(struct net_device *dev,
1127 int vf,
1128 struct ifla_vf_stats
1129 *vf_stats);
57b61080
SF
1130 int (*ndo_set_vf_port)(struct net_device *dev,
1131 int vf,
1132 struct nlattr *port[]);
1133 int (*ndo_get_vf_port)(struct net_device *dev,
1134 int vf, struct sk_buff *skb);
01a3d796
VZ
1135 int (*ndo_set_vf_rss_query_en)(
1136 struct net_device *dev,
1137 int vf, bool setting);
4f57c087 1138 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
d11ead75 1139#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1140 int (*ndo_fcoe_enable)(struct net_device *dev);
1141 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1142 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1143 u16 xid,
1144 struct scatterlist *sgl,
1145 unsigned int sgc);
1146 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1147 u16 xid);
6247e086
YZ
1148 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1149 u16 xid,
1150 struct scatterlist *sgl,
1151 unsigned int sgc);
68bad94e
NP
1152 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1153 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1154#endif
1155
d11ead75 1156#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1157#define NETDEV_FCOE_WWNN 0
1158#define NETDEV_FCOE_WWPN 1
1159 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1160 u64 *wwn, int type);
4d288d57 1161#endif
3c9c36bc 1162
c445477d
BH
1163#ifdef CONFIG_RFS_ACCEL
1164 int (*ndo_rx_flow_steer)(struct net_device *dev,
1165 const struct sk_buff *skb,
1166 u16 rxq_index,
1167 u32 flow_id);
1168#endif
fbaec0ea
JP
1169 int (*ndo_add_slave)(struct net_device *dev,
1170 struct net_device *slave_dev);
1171 int (*ndo_del_slave)(struct net_device *dev,
1172 struct net_device *slave_dev);
c8f44aff
MM
1173 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1174 netdev_features_t features);
5455c699 1175 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1176 netdev_features_t features);
da6a8fa0 1177 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1178 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1179
1180 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1181 struct nlattr *tb[],
77162022 1182 struct net_device *dev,
6b6e2725 1183 const unsigned char *addr,
f6f6424b 1184 u16 vid,
77162022
JF
1185 u16 flags);
1186 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1187 struct nlattr *tb[],
77162022 1188 struct net_device *dev,
f6f6424b
JP
1189 const unsigned char *addr,
1190 u16 vid);
77162022
JF
1191 int (*ndo_fdb_dump)(struct sk_buff *skb,
1192 struct netlink_callback *cb,
1193 struct net_device *dev,
5d5eacb3 1194 struct net_device *filter_dev,
77162022 1195 int idx);
e5a55a89
JF
1196
1197 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1198 struct nlmsghdr *nlh,
1199 u16 flags);
e5a55a89
JF
1200 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1201 u32 pid, u32 seq,
6cbdceeb 1202 struct net_device *dev,
46c264da
ND
1203 u32 filter_mask,
1204 int nlflags);
407af329 1205 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1206 struct nlmsghdr *nlh,
1207 u16 flags);
4bf84c35
JP
1208 int (*ndo_change_carrier)(struct net_device *dev,
1209 bool new_carrier);
66b52b0d 1210 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1211 struct netdev_phys_item_id *ppid);
db24a904
DA
1212 int (*ndo_get_phys_port_name)(struct net_device *dev,
1213 char *name, size_t len);
53cf5275
JG
1214 void (*ndo_add_vxlan_port)(struct net_device *dev,
1215 sa_family_t sa_family,
35e42379 1216 __be16 port);
53cf5275
JG
1217 void (*ndo_del_vxlan_port)(struct net_device *dev,
1218 sa_family_t sa_family,
35e42379 1219 __be16 port);
a6cc0cfa
JF
1220
1221 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1222 struct net_device *dev);
1223 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1224 void *priv);
1225
1226 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1227 struct net_device *dev,
1228 void *priv);
25175ba5 1229 int (*ndo_get_lock_subclass)(struct net_device *dev);
5f35227e
JG
1230 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1231 struct net_device *dev,
1232 netdev_features_t features);
822b3b2e
JF
1233 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1234 int queue_index,
1235 u32 maxrate);
a54acb3a 1236 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1237 int (*ndo_change_proto_down)(struct net_device *dev,
1238 bool proto_down);
fc4099f1
PS
1239 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1240 struct sk_buff *skb);
d314774c
SH
1241};
1242
7aa98047
LR
1243/**
1244 * enum net_device_priv_flags - &struct net_device priv_flags
1245 *
1246 * These are the &struct net_device, they are only set internally
1247 * by drivers and used in the kernel. These flags are invisible to
1248 * userspace, this means that the order of these flags can change
1249 * during any kernel release.
1250 *
1251 * You should have a pretty good reason to be extending these flags.
1252 *
1253 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1254 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1255 * @IFF_BONDING: bonding master or slave
7aa98047 1256 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1257 * @IFF_WAN_HDLC: WAN HDLC device
1258 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1259 * release skb->dst
1260 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1261 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1262 * @IFF_MACVLAN_PORT: device used as macvlan port
1263 * @IFF_BRIDGE_PORT: device used as bridge port
1264 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1265 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1266 * @IFF_UNICAST_FLT: Supports unicast filtering
1267 * @IFF_TEAM_PORT: device used as team port
1268 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1269 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1270 * change when it's running
1271 * @IFF_MACVLAN: Macvlan device
007979ea 1272 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1273 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1274 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1275 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1276 * @IFF_TEAM: device is a team device
7aa98047
LR
1277 */
1278enum netdev_priv_flags {
1279 IFF_802_1Q_VLAN = 1<<0,
1280 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1281 IFF_BONDING = 1<<2,
1282 IFF_ISATAP = 1<<3,
1283 IFF_WAN_HDLC = 1<<4,
1284 IFF_XMIT_DST_RELEASE = 1<<5,
1285 IFF_DONT_BRIDGE = 1<<6,
1286 IFF_DISABLE_NETPOLL = 1<<7,
1287 IFF_MACVLAN_PORT = 1<<8,
1288 IFF_BRIDGE_PORT = 1<<9,
1289 IFF_OVS_DATAPATH = 1<<10,
1290 IFF_TX_SKB_SHARING = 1<<11,
1291 IFF_UNICAST_FLT = 1<<12,
1292 IFF_TEAM_PORT = 1<<13,
1293 IFF_SUPP_NOFCS = 1<<14,
1294 IFF_LIVE_ADDR_CHANGE = 1<<15,
1295 IFF_MACVLAN = 1<<16,
1296 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1297 IFF_IPVLAN_MASTER = 1<<18,
1298 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1299 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1300 IFF_NO_QUEUE = 1<<21,
1301 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1302 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1303 IFF_TEAM = 1<<24,
7aa98047
LR
1304};
1305
1306#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1307#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1308#define IFF_BONDING IFF_BONDING
7aa98047 1309#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1310#define IFF_WAN_HDLC IFF_WAN_HDLC
1311#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1312#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1313#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1314#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1315#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1316#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1317#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1318#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1319#define IFF_TEAM_PORT IFF_TEAM_PORT
1320#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1321#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1322#define IFF_MACVLAN IFF_MACVLAN
02875878 1323#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1324#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1325#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1326#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1327#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1328#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1329#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1330#define IFF_TEAM IFF_TEAM
7aa98047 1331
536721b1
KK
1332/**
1333 * struct net_device - The DEVICE structure.
1334 * Actually, this whole structure is a big mistake. It mixes I/O
1335 * data with strictly "high-level" data, and it has to know about
1336 * almost every data structure used in the INET module.
1337 *
1338 * @name: This is the first field of the "visible" part of this structure
1339 * (i.e. as seen by users in the "Space.c" file). It is the name
1340 * of the interface.
1341 *
1342 * @name_hlist: Device name hash chain, please keep it close to name[]
1343 * @ifalias: SNMP alias
1344 * @mem_end: Shared memory end
1345 * @mem_start: Shared memory start
1346 * @base_addr: Device I/O address
1347 * @irq: Device IRQ number
1348 *
14ffbbb8
TG
1349 * @carrier_changes: Stats to monitor carrier on<->off transitions
1350 *
536721b1
KK
1351 * @state: Generic network queuing layer state, see netdev_state_t
1352 * @dev_list: The global list of network devices
1353 * @napi_list: List entry, that is used for polling napi devices
1354 * @unreg_list: List entry, that is used, when we are unregistering the
1355 * device, see the function unregister_netdev
1356 * @close_list: List entry, that is used, when we are closing the device
1357 *
1358 * @adj_list: Directly linked devices, like slaves for bonding
1359 * @all_adj_list: All linked devices, *including* neighbours
1360 * @features: Currently active device features
1361 * @hw_features: User-changeable features
1362 *
1363 * @wanted_features: User-requested features
1364 * @vlan_features: Mask of features inheritable by VLAN devices
1365 *
1366 * @hw_enc_features: Mask of features inherited by encapsulating devices
1367 * This field indicates what encapsulation
1368 * offloads the hardware is capable of doing,
1369 * and drivers will need to set them appropriately.
1370 *
1371 * @mpls_features: Mask of features inheritable by MPLS
1372 *
1373 * @ifindex: interface index
388069d3 1374 * @group: The group, that the device belongs to
536721b1
KK
1375 *
1376 * @stats: Statistics struct, which was left as a legacy, use
1377 * rtnl_link_stats64 instead
1378 *
1379 * @rx_dropped: Dropped packets by core network,
1380 * do not use this in drivers
1381 * @tx_dropped: Dropped packets by core network,
1382 * do not use this in drivers
1383 *
536721b1
KK
1384 * @wireless_handlers: List of functions to handle Wireless Extensions,
1385 * instead of ioctl,
1386 * see <net/iw_handler.h> for details.
1387 * @wireless_data: Instance data managed by the core of wireless extensions
1388 *
1389 * @netdev_ops: Includes several pointers to callbacks,
1390 * if one wants to override the ndo_*() functions
1391 * @ethtool_ops: Management operations
d476059e 1392 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1393 * of Layer 2 headers.
1394 *
1395 * @flags: Interface flags (a la BSD)
1396 * @priv_flags: Like 'flags' but invisible to userspace,
1397 * see if.h for the definitions
1398 * @gflags: Global flags ( kept as legacy )
1399 * @padded: How much padding added by alloc_netdev()
1400 * @operstate: RFC2863 operstate
1401 * @link_mode: Mapping policy to operstate
1402 * @if_port: Selectable AUI, TP, ...
1403 * @dma: DMA channel
1404 * @mtu: Interface MTU value
1405 * @type: Interface hardware type
880621c2
MB
1406 * @hard_header_len: Hardware header length, which means that this is the
1407 * minimum size of a packet.
536721b1
KK
1408 *
1409 * @needed_headroom: Extra headroom the hardware may need, but not in all
1410 * cases can this be guaranteed
1411 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1412 * cases can this be guaranteed. Some cases also use
1413 * LL_MAX_HEADER instead to allocate the skb
1414 *
1415 * interface address info:
1416 *
1417 * @perm_addr: Permanent hw address
1418 * @addr_assign_type: Hw address assignment type
1419 * @addr_len: Hardware address length
1420 * @neigh_priv_len; Used in neigh_alloc(),
1421 * initialized only in atm/clip.c
1422 * @dev_id: Used to differentiate devices that share
1423 * the same link layer address
1424 * @dev_port: Used to differentiate devices that share
1425 * the same function
1426 * @addr_list_lock: XXX: need comments on this one
536721b1
KK
1427 * @uc_promisc: Counter, that indicates, that promiscuous mode
1428 * has been enabled due to the need to listen to
1429 * additional unicast addresses in a device that
1430 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1431 * @uc: unicast mac addresses
1432 * @mc: multicast mac addresses
1433 * @dev_addrs: list of device hw addresses
1434 * @queues_kset: Group of all Kobjects in the Tx and RX queues
536721b1
KK
1435 * @promiscuity: Number of times, the NIC is told to work in
1436 * Promiscuous mode, if it becomes 0 the NIC will
1437 * exit from working in Promiscuous mode
1438 * @allmulti: Counter, enables or disables allmulticast mode
1439 *
1440 * @vlan_info: VLAN info
1441 * @dsa_ptr: dsa specific data
1442 * @tipc_ptr: TIPC specific data
1443 * @atalk_ptr: AppleTalk link
1444 * @ip_ptr: IPv4 specific data
1445 * @dn_ptr: DECnet specific data
1446 * @ip6_ptr: IPv6 specific data
1447 * @ax25_ptr: AX.25 specific data
1448 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1449 *
1450 * @last_rx: Time of last Rx
1451 * @dev_addr: Hw address (before bcast,
1452 * because most packets are unicast)
1453 *
1454 * @_rx: Array of RX queues
1455 * @num_rx_queues: Number of RX queues
1456 * allocated at register_netdev() time
1457 * @real_num_rx_queues: Number of RX queues currently active in device
1458 *
1459 * @rx_handler: handler for received packets
1460 * @rx_handler_data: XXX: need comments on this one
1461 * @ingress_queue: XXX: need comments on this one
1462 * @broadcast: hw bcast address
1463 *
14ffbbb8
TG
1464 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1465 * indexed by RX queue number. Assigned by driver.
1466 * This must only be set if the ndo_rx_flow_steer
1467 * operation is defined
1468 * @index_hlist: Device index hash chain
1469 *
536721b1
KK
1470 * @_tx: Array of TX queues
1471 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1472 * @real_num_tx_queues: Number of TX queues currently active in device
1473 * @qdisc: Root qdisc from userspace point of view
1474 * @tx_queue_len: Max frames per queue allowed
1475 * @tx_global_lock: XXX: need comments on this one
1476 *
1477 * @xps_maps: XXX: need comments on this one
1478 *
0c4f691f
SF
1479 * @offload_fwd_mark: Offload device fwding mark
1480 *
536721b1
KK
1481 * @trans_start: Time (in jiffies) of last Tx
1482 * @watchdog_timeo: Represents the timeout that is used by
1483 * the watchdog ( see dev_watchdog() )
1484 * @watchdog_timer: List of timers
1485 *
1486 * @pcpu_refcnt: Number of references to this device
1487 * @todo_list: Delayed register/unregister
536721b1
KK
1488 * @link_watch_list: XXX: need comments on this one
1489 *
1490 * @reg_state: Register/unregister state machine
1491 * @dismantle: Device is going to be freed
1492 * @rtnl_link_state: This enum represents the phases of creating
1493 * a new link
1494 *
1495 * @destructor: Called from unregister,
1496 * can be used to call free_netdev
1497 * @npinfo: XXX: need comments on this one
1498 * @nd_net: Network namespace this network device is inside
1499 *
1500 * @ml_priv: Mid-layer private
1501 * @lstats: Loopback statistics
1502 * @tstats: Tunnel statistics
1503 * @dstats: Dummy statistics
1504 * @vstats: Virtual ethernet statistics
1505 *
1506 * @garp_port: GARP
1507 * @mrp_port: MRP
1508 *
1509 * @dev: Class/net/name entry
1510 * @sysfs_groups: Space for optional device, statistics and wireless
1511 * sysfs groups
1512 *
1513 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1514 * @rtnl_link_ops: Rtnl_link_ops
1515 *
1516 * @gso_max_size: Maximum size of generic segmentation offload
1517 * @gso_max_segs: Maximum number of segments that can be passed to the
1518 * NIC for GSO
fcbeb976
ED
1519 * @gso_min_segs: Minimum number of segments that can be passed to the
1520 * NIC for GSO
536721b1
KK
1521 *
1522 * @dcbnl_ops: Data Center Bridging netlink ops
1523 * @num_tc: Number of traffic classes in the net device
1524 * @tc_to_txq: XXX: need comments on this one
1525 * @prio_tc_map XXX: need comments on this one
1526 *
1527 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1528 *
1529 * @priomap: XXX: need comments on this one
1530 * @phydev: Physical device may attach itself
1531 * for hardware timestamping
1532 *
1533 * @qdisc_tx_busylock: XXX: need comments on this one
1534 *
d746d707
AK
1535 * @proto_down: protocol port state information can be sent to the
1536 * switch driver and used to set the phys state of the
1537 * switch port.
1538 *
1da177e4
LT
1539 * FIXME: cleanup struct net_device such that network protocol info
1540 * moves out.
1541 */
1542
d94d9fee 1543struct net_device {
1da177e4 1544 char name[IFNAMSIZ];
9356b8fc 1545 struct hlist_node name_hlist;
0b815a1a 1546 char *ifalias;
1da177e4
LT
1547 /*
1548 * I/O specific fields
1549 * FIXME: Merge these and struct ifmap into one
1550 */
536721b1
KK
1551 unsigned long mem_end;
1552 unsigned long mem_start;
1553 unsigned long base_addr;
1554 int irq;
1da177e4 1555
14ffbbb8
TG
1556 atomic_t carrier_changes;
1557
1da177e4 1558 /*
536721b1
KK
1559 * Some hardware also needs these fields (state,dev_list,
1560 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1561 * part of the usual set specified in Space.c.
1562 */
1563
1da177e4
LT
1564 unsigned long state;
1565
7562f876 1566 struct list_head dev_list;
bea3348e 1567 struct list_head napi_list;
44a0873d 1568 struct list_head unreg_list;
5cde2829 1569 struct list_head close_list;
7866a621
SN
1570 struct list_head ptype_all;
1571 struct list_head ptype_specific;
2f268f12 1572
2f268f12
VF
1573 struct {
1574 struct list_head upper;
1575 struct list_head lower;
1576 } adj_list;
1577
2f268f12
VF
1578 struct {
1579 struct list_head upper;
1580 struct list_head lower;
1581 } all_adj_list;
4c3d5e7b 1582
c8f44aff 1583 netdev_features_t features;
c8f44aff 1584 netdev_features_t hw_features;
c8f44aff 1585 netdev_features_t wanted_features;
c8f44aff 1586 netdev_features_t vlan_features;
6a674e9c 1587 netdev_features_t hw_enc_features;
0d89d203 1588 netdev_features_t mpls_features;
04ed3e74 1589
1da177e4 1590 int ifindex;
7a66bbc9 1591 int group;
1da177e4 1592
c45d286e 1593 struct net_device_stats stats;
015f0688 1594
015f0688
ED
1595 atomic_long_t rx_dropped;
1596 atomic_long_t tx_dropped;
1da177e4 1597
b86e0280 1598#ifdef CONFIG_WIRELESS_EXT
1da177e4 1599 const struct iw_handler_def * wireless_handlers;
1da177e4 1600 struct iw_public_data * wireless_data;
b86e0280 1601#endif
d314774c 1602 const struct net_device_ops *netdev_ops;
76fd8593 1603 const struct ethtool_ops *ethtool_ops;
4170604f 1604#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1605 const struct switchdev_ops *switchdev_ops;
4170604f 1606#endif
1b69c6d0
DA
1607#ifdef CONFIG_NET_L3_MASTER_DEV
1608 const struct l3mdev_ops *l3mdev_ops;
1609#endif
1da177e4 1610
3b04ddde
SH
1611 const struct header_ops *header_ops;
1612
536721b1
KK
1613 unsigned int flags;
1614 unsigned int priv_flags;
1615
1da177e4 1616 unsigned short gflags;
536721b1 1617 unsigned short padded;
1da177e4 1618
536721b1
KK
1619 unsigned char operstate;
1620 unsigned char link_mode;
b00055aa 1621
536721b1
KK
1622 unsigned char if_port;
1623 unsigned char dma;
bdc220da 1624
536721b1
KK
1625 unsigned int mtu;
1626 unsigned short type;
1627 unsigned short hard_header_len;
1da177e4 1628
f5184d26
JB
1629 unsigned short needed_headroom;
1630 unsigned short needed_tailroom;
1631
1da177e4 1632 /* Interface address info. */
536721b1
KK
1633 unsigned char perm_addr[MAX_ADDR_LEN];
1634 unsigned char addr_assign_type;
1635 unsigned char addr_len;
a0a9663d 1636 unsigned short neigh_priv_len;
536721b1
KK
1637 unsigned short dev_id;
1638 unsigned short dev_port;
ccffad25 1639 spinlock_t addr_list_lock;
14ffbbb8
TG
1640 unsigned char name_assign_type;
1641 bool uc_promisc;
536721b1
KK
1642 struct netdev_hw_addr_list uc;
1643 struct netdev_hw_addr_list mc;
1644 struct netdev_hw_addr_list dev_addrs;
1645
4c3d5e7b
ED
1646#ifdef CONFIG_SYSFS
1647 struct kset *queues_kset;
1648#endif
9d45abe1
WC
1649 unsigned int promiscuity;
1650 unsigned int allmulti;
1da177e4 1651
1da177e4
LT
1652
1653 /* Protocol specific pointers */
65ac6a5f 1654
d11ead75 1655#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1656 struct vlan_info __rcu *vlan_info;
65ac6a5f 1657#endif
34a430d7 1658#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1659 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1660#endif
1661#if IS_ENABLED(CONFIG_TIPC)
536721b1 1662 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1663#endif
536721b1
KK
1664 void *atalk_ptr;
1665 struct in_device __rcu *ip_ptr;
1666 struct dn_dev __rcu *dn_ptr;
1667 struct inet6_dev __rcu *ip6_ptr;
1668 void *ax25_ptr;
1669 struct wireless_dev *ieee80211_ptr;
98a18b6f 1670 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1671#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1672 struct mpls_dev __rcu *mpls_ptr;
1673#endif
1da177e4 1674
9356b8fc 1675/*
cd13539b 1676 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1677 */
536721b1 1678 unsigned long last_rx;
4dc89133 1679
9356b8fc 1680 /* Interface address info used in eth_type_trans() */
536721b1 1681 unsigned char *dev_addr;
f001fde5 1682
0a9627f2 1683
a953be53 1684#ifdef CONFIG_SYSFS
0a9627f2
TH
1685 struct netdev_rx_queue *_rx;
1686
0a9627f2 1687 unsigned int num_rx_queues;
62fe0b40 1688 unsigned int real_num_rx_queues;
c445477d 1689
df334545 1690#endif
0a9627f2 1691
3b47d303 1692 unsigned long gro_flush_timeout;
61391cde 1693 rx_handler_func_t __rcu *rx_handler;
1694 void __rcu *rx_handler_data;
e8a0464c 1695
4cda01e8 1696#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1697 struct tcf_proto __rcu *ingress_cl_list;
1698#endif
24824a09 1699 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1700#ifdef CONFIG_NETFILTER_INGRESS
1701 struct list_head nf_hooks_ingress;
1702#endif
d2788d34 1703
536721b1 1704 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1705#ifdef CONFIG_RFS_ACCEL
1706 struct cpu_rmap *rx_cpu_rmap;
1707#endif
1708 struct hlist_node index_hlist;
cd13539b
ED
1709
1710/*
1711 * Cache lines mostly used on transmit path
1712 */
e8a0464c
DM
1713 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1714 unsigned int num_tx_queues;
fd2ea0a7 1715 unsigned int real_num_tx_queues;
af356afa 1716 struct Qdisc *qdisc;
536721b1 1717 unsigned long tx_queue_len;
c3f26a26 1718 spinlock_t tx_global_lock;
14ffbbb8 1719 int watchdog_timeo;
cd13539b 1720
bf264145 1721#ifdef CONFIG_XPS
a4177869 1722 struct xps_dev_maps __rcu *xps_maps;
bf264145 1723#endif
1d24eb48 1724
0c4f691f
SF
1725#ifdef CONFIG_NET_SWITCHDEV
1726 u32 offload_fwd_mark;
1727#endif
1728
9356b8fc 1729 /* These may be needed for future network-power-down code. */
9d21493b
ED
1730
1731 /*
1732 * trans_start here is expensive for high speed devices on SMP,
1733 * please use netdev_queue->trans_start instead.
1734 */
536721b1 1735 unsigned long trans_start;
9356b8fc 1736
9356b8fc
ED
1737 struct timer_list watchdog_timer;
1738
29b4433d 1739 int __percpu *pcpu_refcnt;
1da177e4 1740 struct list_head todo_list;
1da177e4 1741
e014debe 1742 struct list_head link_watch_list;
572a103d 1743
1da177e4 1744 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1745 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1746 NETREG_UNREGISTERING, /* called unregister_netdevice */
1747 NETREG_UNREGISTERED, /* completed unregister todo */
1748 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1749 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1750 } reg_state:8;
1751
536721b1 1752 bool dismantle;
a2835763
PM
1753
1754 enum {
1755 RTNL_LINK_INITIALIZED,
1756 RTNL_LINK_INITIALIZING,
1757 } rtnl_link_state:16;
1da177e4 1758
d314774c 1759 void (*destructor)(struct net_device *dev);
1da177e4 1760
1da177e4 1761#ifdef CONFIG_NETPOLL
5fbee843 1762 struct netpoll_info __rcu *npinfo;
1da177e4 1763#endif
eae792b7 1764
0c5c9fb5 1765 possible_net_t nd_net;
4a1c5371 1766
4951704b 1767 /* mid-layer private */
a7855c78 1768 union {
536721b1
KK
1769 void *ml_priv;
1770 struct pcpu_lstats __percpu *lstats;
8f84985f 1771 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1772 struct pcpu_dstats __percpu *dstats;
1773 struct pcpu_vstats __percpu *vstats;
a7855c78 1774 };
536721b1 1775
3cc77ec7 1776 struct garp_port __rcu *garp_port;
febf018d 1777 struct mrp_port __rcu *mrp_port;
1da177e4 1778
536721b1 1779 struct device dev;
0c509a6c 1780 const struct attribute_group *sysfs_groups[4];
a953be53 1781 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1782
38f7b870 1783 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1784
82cc1a7a
PWJ
1785 /* for setting kernel sock attribute on TCP connection setup */
1786#define GSO_MAX_SIZE 65536
1787 unsigned int gso_max_size;
30b678d8
BH
1788#define GSO_MAX_SEGS 65535
1789 u16 gso_max_segs;
fcbeb976 1790 u16 gso_min_segs;
7a6b6f51 1791#ifdef CONFIG_DCB
32953543 1792 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1793#endif
4f57c087
JF
1794 u8 num_tc;
1795 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1796 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1797
d11ead75 1798#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1799 unsigned int fcoe_ddp_xid;
5bc1421e 1800#endif
86f8515f 1801#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1802 struct netprio_map __rcu *priomap;
4d288d57 1803#endif
c1f19b51 1804 struct phy_device *phydev;
23d3b8bf 1805 struct lock_class_key *qdisc_tx_busylock;
d746d707 1806 bool proto_down;
1da177e4 1807};
43cb76d9 1808#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1809
1810#define NETDEV_ALIGN 32
1da177e4 1811
4f57c087
JF
1812static inline
1813int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1814{
1815 return dev->prio_tc_map[prio & TC_BITMASK];
1816}
1817
1818static inline
1819int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1820{
1821 if (tc >= dev->num_tc)
1822 return -EINVAL;
1823
1824 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1825 return 0;
1826}
1827
1828static inline
1829void netdev_reset_tc(struct net_device *dev)
1830{
1831 dev->num_tc = 0;
1832 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1833 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1834}
1835
1836static inline
1837int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1838{
1839 if (tc >= dev->num_tc)
1840 return -EINVAL;
1841
1842 dev->tc_to_txq[tc].count = count;
1843 dev->tc_to_txq[tc].offset = offset;
1844 return 0;
1845}
1846
1847static inline
1848int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1849{
1850 if (num_tc > TC_MAX_QUEUE)
1851 return -EINVAL;
1852
1853 dev->num_tc = num_tc;
1854 return 0;
1855}
1856
1857static inline
1858int netdev_get_num_tc(struct net_device *dev)
1859{
1860 return dev->num_tc;
1861}
1862
e8a0464c
DM
1863static inline
1864struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1865 unsigned int index)
1866{
1867 return &dev->_tx[index];
1868}
1869
10c51b56
DB
1870static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1871 const struct sk_buff *skb)
1872{
1873 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1874}
1875
e8a0464c
DM
1876static inline void netdev_for_each_tx_queue(struct net_device *dev,
1877 void (*f)(struct net_device *,
1878 struct netdev_queue *,
1879 void *),
1880 void *arg)
1881{
1882 unsigned int i;
1883
1884 for (i = 0; i < dev->num_tx_queues; i++)
1885 f(dev, &dev->_tx[i], arg);
1886}
1887
f629d208 1888struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1889 struct sk_buff *skb,
1890 void *accel_priv);
8c4c49df 1891
c346dca1
YH
1892/*
1893 * Net namespace inlines
1894 */
1895static inline
1896struct net *dev_net(const struct net_device *dev)
1897{
c2d9ba9b 1898 return read_pnet(&dev->nd_net);
c346dca1
YH
1899}
1900
1901static inline
f5aa23fd 1902void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 1903{
0c5c9fb5 1904 write_pnet(&dev->nd_net, net);
c346dca1
YH
1905}
1906
3e8a72d1 1907static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 1908{
3fc88677 1909#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
1910 if (dev->dsa_ptr != NULL)
1911 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 1912#endif
5aed85ce 1913 return false;
396138f0
LB
1914}
1915
bea3348e
SH
1916/**
1917 * netdev_priv - access network device private data
1918 * @dev: network device
1919 *
1920 * Get network device private data
1921 */
6472ce60 1922static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1923{
1ce8e7b5 1924 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1925}
1926
1da177e4
LT
1927/* Set the sysfs physical device reference for the network logical device
1928 * if set prior to registration will cause a symlink during initialization.
1929 */
43cb76d9 1930#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1931
384912ed 1932/* Set the sysfs device type for the network logical device to allow
3f79410c 1933 * fine-grained identification of different network device types. For
384912ed
MH
1934 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1935 */
1936#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1937
82dc3c63
ED
1938/* Default NAPI poll() weight
1939 * Device drivers are strongly advised to not use bigger value
1940 */
1941#define NAPI_POLL_WEIGHT 64
1942
3b582cc1
SH
1943/**
1944 * netif_napi_add - initialize a napi context
1945 * @dev: network device
1946 * @napi: napi context
1947 * @poll: polling function
1948 * @weight: default weight
1949 *
1950 * netif_napi_add() must be used to initialize a napi context prior to calling
1951 * *any* of the other napi related functions.
1952 */
d565b0a1
HX
1953void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1954 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1955
d64b5e85
ED
1956/**
1957 * netif_tx_napi_add - initialize a napi context
1958 * @dev: network device
1959 * @napi: napi context
1960 * @poll: polling function
1961 * @weight: default weight
1962 *
1963 * This variant of netif_napi_add() should be used from drivers using NAPI
1964 * to exclusively poll a TX queue.
1965 * This will avoid we add it into napi_hash[], thus polluting this hash table.
1966 */
1967static inline void netif_tx_napi_add(struct net_device *dev,
1968 struct napi_struct *napi,
1969 int (*poll)(struct napi_struct *, int),
1970 int weight)
1971{
1972 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
1973 netif_napi_add(dev, napi, poll, weight);
1974}
1975
d8156534
AD
1976/**
1977 * netif_napi_del - remove a napi context
1978 * @napi: napi context
1979 *
1980 * netif_napi_del() removes a napi context from the network device napi list
1981 */
d565b0a1
HX
1982void netif_napi_del(struct napi_struct *napi);
1983
1984struct napi_gro_cb {
78a478d0
HX
1985 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1986 void *frag0;
1987
7489594c
HX
1988 /* Length of frag0. */
1989 unsigned int frag0_len;
1990
86911732
HX
1991 /* This indicates where we are processing relative to skb->data. */
1992 int data_offset;
1993
d565b0a1 1994 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
1995 u16 flush;
1996
1997 /* Save the IP ID here and check when we get to the transport layer */
1998 u16 flush_id;
d565b0a1
HX
1999
2000 /* Number of segments aggregated. */
2e71a6f8
ED
2001 u16 count;
2002
15e2396d
TH
2003 /* Start offset for remote checksum offload */
2004 u16 gro_remcsum_start;
2005
2e71a6f8
ED
2006 /* jiffies when first packet was created/queued */
2007 unsigned long age;
86347245 2008
afe93325 2009 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2010 u16 proto;
2011
baa32ff4
TH
2012 /* This is non-zero if the packet may be of the same flow. */
2013 u8 same_flow:1;
2014
b582ef09 2015 /* Used in udp_gro_receive */
573e8fca
TH
2016 u8 udp_mark:1;
2017
2018 /* GRO checksum is valid */
2019 u8 csum_valid:1;
2020
662880f4
TH
2021 /* Number of checksums via CHECKSUM_UNNECESSARY */
2022 u8 csum_cnt:3;
c3c7c254 2023
baa32ff4
TH
2024 /* Free the skb? */
2025 u8 free:2;
2026#define NAPI_GRO_FREE 1
2027#define NAPI_GRO_FREE_STOLEN_HEAD 2
2028
efc98d08
TH
2029 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2030 u8 is_ipv6:1;
2031
baa32ff4
TH
2032 /* 7 bit hole */
2033
bf5a755f
JC
2034 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2035 __wsum csum;
2036
c3c7c254
ED
2037 /* used in skb_gro_receive() slow path */
2038 struct sk_buff *last;
d565b0a1
HX
2039};
2040
2041#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2042
1da177e4 2043struct packet_type {
f2ccd8fa
DM
2044 __be16 type; /* This is really htons(ether_type). */
2045 struct net_device *dev; /* NULL is wildcarded here */
2046 int (*func) (struct sk_buff *,
2047 struct net_device *,
2048 struct packet_type *,
2049 struct net_device *);
c0de08d0
EL
2050 bool (*id_match)(struct packet_type *ptype,
2051 struct sock *sk);
1da177e4
LT
2052 void *af_packet_priv;
2053 struct list_head list;
2054};
2055
f191a1d1 2056struct offload_callbacks {
576a30eb 2057 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2058 netdev_features_t features);
d565b0a1 2059 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2060 struct sk_buff *skb);
299603e8 2061 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2062};
2063
2064struct packet_offload {
2065 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2066 u16 priority;
f191a1d1
VY
2067 struct offload_callbacks callbacks;
2068 struct list_head list;
1da177e4
LT
2069};
2070
a2b12f3c
TH
2071struct udp_offload;
2072
2073struct udp_offload_callbacks {
2074 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2075 struct sk_buff *skb,
2076 struct udp_offload *uoff);
2077 int (*gro_complete)(struct sk_buff *skb,
2078 int nhoff,
2079 struct udp_offload *uoff);
2080};
2081
b582ef09
OG
2082struct udp_offload {
2083 __be16 port;
afe93325 2084 u8 ipproto;
a2b12f3c 2085 struct udp_offload_callbacks callbacks;
b582ef09
OG
2086};
2087
8f84985f
LR
2088/* often modified stats are per cpu, other are shared (netdev->stats) */
2089struct pcpu_sw_netstats {
2090 u64 rx_packets;
2091 u64 rx_bytes;
2092 u64 tx_packets;
2093 u64 tx_bytes;
2094 struct u64_stats_sync syncp;
2095};
2096
aabc92bb
PNA
2097#define __netdev_alloc_pcpu_stats(type, gfp) \
2098({ \
2099 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2100 if (pcpu_stats) { \
2101 int __cpu; \
2102 for_each_possible_cpu(__cpu) { \
2103 typeof(type) *stat; \
2104 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2105 u64_stats_init(&stat->syncp); \
2106 } \
2107 } \
2108 pcpu_stats; \
1c213bd2
WC
2109})
2110
aabc92bb
PNA
2111#define netdev_alloc_pcpu_stats(type) \
2112 __netdev_alloc_pcpu_stats(type, GFP_KERNEL);
2113
764f5e54
JP
2114enum netdev_lag_tx_type {
2115 NETDEV_LAG_TX_TYPE_UNKNOWN,
2116 NETDEV_LAG_TX_TYPE_RANDOM,
2117 NETDEV_LAG_TX_TYPE_BROADCAST,
2118 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2119 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2120 NETDEV_LAG_TX_TYPE_HASH,
2121};
2122
2123struct netdev_lag_upper_info {
2124 enum netdev_lag_tx_type tx_type;
2125};
2126
fb1b2e3c
JP
2127struct netdev_lag_lower_state_info {
2128 u8 link_up : 1,
2129 tx_enabled : 1;
2130};
2131
1da177e4
LT
2132#include <linux/notifier.h>
2133
dcfe1421
AW
2134/* netdevice notifier chain. Please remember to update the rtnetlink
2135 * notification exclusion list in rtnetlink_event() when adding new
2136 * types.
2137 */
2138#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2139#define NETDEV_DOWN 0x0002
2140#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2141 detected a hardware crash and restarted
2142 - we can use this eg to kick tcp sessions
2143 once done */
2144#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2145#define NETDEV_REGISTER 0x0005
2146#define NETDEV_UNREGISTER 0x0006
1d486bfb 2147#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2148#define NETDEV_CHANGEADDR 0x0008
2149#define NETDEV_GOING_DOWN 0x0009
2150#define NETDEV_CHANGENAME 0x000A
2151#define NETDEV_FEAT_CHANGE 0x000B
2152#define NETDEV_BONDING_FAILOVER 0x000C
2153#define NETDEV_PRE_UP 0x000D
2154#define NETDEV_PRE_TYPE_CHANGE 0x000E
2155#define NETDEV_POST_TYPE_CHANGE 0x000F
2156#define NETDEV_POST_INIT 0x0010
0115e8e3 2157#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2158#define NETDEV_RELEASE 0x0012
2159#define NETDEV_NOTIFY_PEERS 0x0013
2160#define NETDEV_JOIN 0x0014
42e52bf9 2161#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2162#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2163#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2164#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2165#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2166#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2167#define NETDEV_CHANGELOWERSTATE 0x001B
dcfe1421 2168
f629d208
JP
2169int register_netdevice_notifier(struct notifier_block *nb);
2170int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2171
2172struct netdev_notifier_info {
2173 struct net_device *dev;
2174};
2175
be9efd36
JP
2176struct netdev_notifier_change_info {
2177 struct netdev_notifier_info info; /* must be first */
2178 unsigned int flags_changed;
2179};
2180
0e4ead9d
JP
2181struct netdev_notifier_changeupper_info {
2182 struct netdev_notifier_info info; /* must be first */
2183 struct net_device *upper_dev; /* new upper dev */
2184 bool master; /* is upper dev master */
2185 bool linking; /* is the nofication for link or unlink */
29bf24af 2186 void *upper_info; /* upper dev info */
0e4ead9d
JP
2187};
2188
04d48266
JP
2189struct netdev_notifier_changelowerstate_info {
2190 struct netdev_notifier_info info; /* must be first */
2191 void *lower_state_info; /* is lower dev state */
2192};
2193
75538c2b
CW
2194static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2195 struct net_device *dev)
2196{
2197 info->dev = dev;
2198}
2199
351638e7
JP
2200static inline struct net_device *
2201netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2202{
2203 return info->dev;
2204}
2205
f629d208 2206int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2207
2208
1da177e4
LT
2209extern rwlock_t dev_base_lock; /* Device list lock */
2210
881d966b
EB
2211#define for_each_netdev(net, d) \
2212 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2213#define for_each_netdev_reverse(net, d) \
2214 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2215#define for_each_netdev_rcu(net, d) \
2216 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2217#define for_each_netdev_safe(net, d, n) \
2218 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2219#define for_each_netdev_continue(net, d) \
2220 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2221#define for_each_netdev_continue_rcu(net, d) \
2222 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2223#define for_each_netdev_in_bond_rcu(bond, slave) \
2224 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2225 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2226#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2227
a050c33f
DL
2228static inline struct net_device *next_net_device(struct net_device *dev)
2229{
2230 struct list_head *lh;
2231 struct net *net;
2232
c346dca1 2233 net = dev_net(dev);
a050c33f
DL
2234 lh = dev->dev_list.next;
2235 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2236}
2237
ce81b76a
ED
2238static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2239{
2240 struct list_head *lh;
2241 struct net *net;
2242
2243 net = dev_net(dev);
ccf43438 2244 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2245 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2246}
2247
a050c33f
DL
2248static inline struct net_device *first_net_device(struct net *net)
2249{
2250 return list_empty(&net->dev_base_head) ? NULL :
2251 net_device_entry(net->dev_base_head.next);
2252}
7562f876 2253
ccf43438
ED
2254static inline struct net_device *first_net_device_rcu(struct net *net)
2255{
2256 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2257
2258 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2259}
2260
f629d208
JP
2261int netdev_boot_setup_check(struct net_device *dev);
2262unsigned long netdev_boot_base(const char *prefix, int unit);
2263struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2264 const char *hwaddr);
2265struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2266struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2267void dev_add_pack(struct packet_type *pt);
2268void dev_remove_pack(struct packet_type *pt);
2269void __dev_remove_pack(struct packet_type *pt);
2270void dev_add_offload(struct packet_offload *po);
2271void dev_remove_offload(struct packet_offload *po);
f629d208 2272
a54acb3a 2273int dev_get_iflink(const struct net_device *dev);
fc4099f1 2274int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2275struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2276 unsigned short mask);
f629d208
JP
2277struct net_device *dev_get_by_name(struct net *net, const char *name);
2278struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2279struct net_device *__dev_get_by_name(struct net *net, const char *name);
2280int dev_alloc_name(struct net_device *dev, const char *name);
2281int dev_open(struct net_device *dev);
2282int dev_close(struct net_device *dev);
99c4a26a 2283int dev_close_many(struct list_head *head, bool unlink);
f629d208 2284void dev_disable_lro(struct net_device *dev);
0c4b51f0 2285int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2286int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2287int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2288int register_netdevice(struct net_device *dev);
2289void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2290void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2291static inline void unregister_netdevice(struct net_device *dev)
2292{
2293 unregister_netdevice_queue(dev, NULL);
2294}
2295
f629d208
JP
2296int netdev_refcnt_read(const struct net_device *dev);
2297void free_netdev(struct net_device *dev);
74d332c1 2298void netdev_freemem(struct net_device *dev);
f629d208
JP
2299void synchronize_net(void);
2300int init_dummy_netdev(struct net_device *dev);
937f1ba5 2301
f60e5990 2302DECLARE_PER_CPU(int, xmit_recursion);
2303static inline int dev_recursion_level(void)
2304{
2305 return this_cpu_read(xmit_recursion);
2306}
2307
f629d208
JP
2308struct net_device *dev_get_by_index(struct net *net, int ifindex);
2309struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2310struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2311int netdev_get_name(struct net *net, char *name, int ifindex);
2312int dev_restart(struct net_device *dev);
f629d208 2313int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2314
2315static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2316{
2317 return NAPI_GRO_CB(skb)->data_offset;
2318}
2319
2320static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2321{
2322 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2323}
2324
2325static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2326{
2327 NAPI_GRO_CB(skb)->data_offset += len;
2328}
2329
a5b1cf28
HX
2330static inline void *skb_gro_header_fast(struct sk_buff *skb,
2331 unsigned int offset)
86911732 2332{
a5b1cf28
HX
2333 return NAPI_GRO_CB(skb)->frag0 + offset;
2334}
78a478d0 2335
a5b1cf28
HX
2336static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2337{
2338 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2339}
78a478d0 2340
a5b1cf28
HX
2341static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2342 unsigned int offset)
2343{
17dd759c
HX
2344 if (!pskb_may_pull(skb, hlen))
2345 return NULL;
2346
a5b1cf28
HX
2347 NAPI_GRO_CB(skb)->frag0 = NULL;
2348 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2349 return skb->data + offset;
86911732 2350}
1da177e4 2351
36e7b1b8
HX
2352static inline void *skb_gro_network_header(struct sk_buff *skb)
2353{
78d3fd0b
HX
2354 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2355 skb_network_offset(skb);
36e7b1b8
HX
2356}
2357
bf5a755f
JC
2358static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2359 const void *start, unsigned int len)
2360{
573e8fca 2361 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2362 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2363 csum_partial(start, len, 0));
2364}
2365
573e8fca
TH
2366/* GRO checksum functions. These are logical equivalents of the normal
2367 * checksum functions (in skbuff.h) except that they operate on the GRO
2368 * offsets and fields in sk_buff.
2369 */
2370
2371__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2372
15e2396d
TH
2373static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2374{
b7fe10e5 2375 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2376}
2377
573e8fca
TH
2378static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2379 bool zero_okay,
2380 __sum16 check)
2381{
6edec0e6
TH
2382 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2383 skb_checksum_start_offset(skb) <
2384 skb_gro_offset(skb)) &&
15e2396d 2385 !skb_at_gro_remcsum_start(skb) &&
662880f4 2386 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2387 (!zero_okay || check));
2388}
2389
2390static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2391 __wsum psum)
2392{
2393 if (NAPI_GRO_CB(skb)->csum_valid &&
2394 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2395 return 0;
2396
2397 NAPI_GRO_CB(skb)->csum = psum;
2398
2399 return __skb_gro_checksum_complete(skb);
2400}
2401
573e8fca
TH
2402static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2403{
662880f4
TH
2404 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2405 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2406 NAPI_GRO_CB(skb)->csum_cnt--;
2407 } else {
2408 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2409 * verified a new top level checksum or an encapsulated one
2410 * during GRO. This saves work if we fallback to normal path.
2411 */
2412 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2413 }
2414}
2415
2416#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2417 compute_pseudo) \
2418({ \
2419 __sum16 __ret = 0; \
2420 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2421 __ret = __skb_gro_checksum_validate_complete(skb, \
2422 compute_pseudo(skb, proto)); \
5a212329
TH
2423 if (__ret) \
2424 __skb_mark_checksum_bad(skb); \
2425 else \
573e8fca
TH
2426 skb_gro_incr_csum_unnecessary(skb); \
2427 __ret; \
2428})
2429
2430#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2431 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2432
2433#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2434 compute_pseudo) \
2435 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2436
2437#define skb_gro_checksum_simple_validate(skb) \
2438 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2439
d96535a1
TH
2440static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2441{
2442 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2443 !NAPI_GRO_CB(skb)->csum_valid);
2444}
2445
2446static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2447 __sum16 check, __wsum pseudo)
2448{
2449 NAPI_GRO_CB(skb)->csum = ~pseudo;
2450 NAPI_GRO_CB(skb)->csum_valid = 1;
2451}
2452
2453#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2454do { \
2455 if (__skb_gro_checksum_convert_check(skb)) \
2456 __skb_gro_checksum_convert(skb, check, \
2457 compute_pseudo(skb, proto)); \
2458} while (0)
2459
26c4f7da
TH
2460struct gro_remcsum {
2461 int offset;
2462 __wsum delta;
2463};
2464
2465static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2466{
846cd667 2467 grc->offset = 0;
26c4f7da
TH
2468 grc->delta = 0;
2469}
2470
b7fe10e5
TH
2471static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2472 unsigned int off, size_t hdrlen,
2473 int start, int offset,
2474 struct gro_remcsum *grc,
2475 bool nopartial)
dcdc8994
TH
2476{
2477 __wsum delta;
b7fe10e5 2478 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2479
2480 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2481
15e2396d 2482 if (!nopartial) {
b7fe10e5
TH
2483 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2484 return ptr;
2485 }
2486
2487 ptr = skb_gro_header_fast(skb, off);
2488 if (skb_gro_header_hard(skb, off + plen)) {
2489 ptr = skb_gro_header_slow(skb, off + plen, off);
2490 if (!ptr)
2491 return NULL;
15e2396d
TH
2492 }
2493
b7fe10e5
TH
2494 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2495 start, offset);
dcdc8994
TH
2496
2497 /* Adjust skb->csum since we changed the packet */
dcdc8994 2498 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2499
b7fe10e5 2500 grc->offset = off + hdrlen + offset;
26c4f7da 2501 grc->delta = delta;
b7fe10e5
TH
2502
2503 return ptr;
dcdc8994
TH
2504}
2505
26c4f7da
TH
2506static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2507 struct gro_remcsum *grc)
2508{
b7fe10e5
TH
2509 void *ptr;
2510 size_t plen = grc->offset + sizeof(u16);
2511
26c4f7da
TH
2512 if (!grc->delta)
2513 return;
2514
b7fe10e5
TH
2515 ptr = skb_gro_header_fast(skb, grc->offset);
2516 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2517 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2518 if (!ptr)
2519 return;
2520 }
2521
2522 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2523}
dcdc8994 2524
0c4e8581
SH
2525static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2526 unsigned short type,
3b04ddde 2527 const void *daddr, const void *saddr,
95c96174 2528 unsigned int len)
0c4e8581 2529{
f1ecfd5d 2530 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2531 return 0;
3b04ddde
SH
2532
2533 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2534}
2535
b95cce35
SH
2536static inline int dev_parse_header(const struct sk_buff *skb,
2537 unsigned char *haddr)
2538{
2539 const struct net_device *dev = skb->dev;
2540
1b83336b 2541 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2542 return 0;
3b04ddde 2543 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2544}
2545
1da177e4 2546typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2547int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2548static inline int unregister_gifconf(unsigned int family)
2549{
2550 return register_gifconf(family, NULL);
2551}
2552
99bbc707 2553#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2554#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2555struct sd_flow_limit {
2556 u64 count;
2557 unsigned int num_buckets;
2558 unsigned int history_head;
2559 u16 history[FLOW_LIMIT_HISTORY];
2560 u8 buckets[];
2561};
2562
2563extern int netdev_flow_limit_table_len;
2564#endif /* CONFIG_NET_FLOW_LIMIT */
2565
1da177e4 2566/*
88751275 2567 * Incoming packets are placed on per-cpu queues
1da177e4 2568 */
d94d9fee 2569struct softnet_data {
1da177e4 2570 struct list_head poll_list;
6e7676c1 2571 struct sk_buff_head process_queue;
1da177e4 2572
dee42870 2573 /* stats */
cd7b5396
DM
2574 unsigned int processed;
2575 unsigned int time_squeeze;
2576 unsigned int cpu_collision;
2577 unsigned int received_rps;
fd793d89 2578#ifdef CONFIG_RPS
88751275 2579 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2580#endif
2581#ifdef CONFIG_NET_FLOW_LIMIT
2582 struct sd_flow_limit __rcu *flow_limit;
2583#endif
2584 struct Qdisc *output_queue;
2585 struct Qdisc **output_queue_tailp;
2586 struct sk_buff *completion_queue;
88751275 2587
4cdb1e2e 2588#ifdef CONFIG_RPS
88751275 2589 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 2590 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2591 struct softnet_data *rps_ipi_next;
2592 unsigned int cpu;
fec5e652 2593 unsigned int input_queue_head;
76cc8b13 2594 unsigned int input_queue_tail;
1e94d72f 2595#endif
95c96174 2596 unsigned int dropped;
0a9627f2 2597 struct sk_buff_head input_pkt_queue;
bea3348e 2598 struct napi_struct backlog;
99bbc707 2599
1da177e4
LT
2600};
2601
76cc8b13 2602static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2603{
2604#ifdef CONFIG_RPS
76cc8b13
TH
2605 sd->input_queue_head++;
2606#endif
2607}
2608
2609static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2610 unsigned int *qtail)
2611{
2612#ifdef CONFIG_RPS
2613 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2614#endif
2615}
2616
0a9627f2 2617DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2618
f629d208 2619void __netif_schedule(struct Qdisc *q);
46e5da40 2620void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2621
fd2ea0a7
DM
2622static inline void netif_tx_schedule_all(struct net_device *dev)
2623{
2624 unsigned int i;
2625
2626 for (i = 0; i < dev->num_tx_queues; i++)
2627 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2628}
2629
d29f749e
DJ
2630static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2631{
73466498 2632 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2633}
2634
bea3348e
SH
2635/**
2636 * netif_start_queue - allow transmit
2637 * @dev: network device
2638 *
2639 * Allow upper layers to call the device hard_start_xmit routine.
2640 */
1da177e4
LT
2641static inline void netif_start_queue(struct net_device *dev)
2642{
e8a0464c 2643 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2644}
2645
fd2ea0a7
DM
2646static inline void netif_tx_start_all_queues(struct net_device *dev)
2647{
2648 unsigned int i;
2649
2650 for (i = 0; i < dev->num_tx_queues; i++) {
2651 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2652 netif_tx_start_queue(txq);
2653 }
2654}
2655
46e5da40 2656void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2657
d29f749e
DJ
2658/**
2659 * netif_wake_queue - restart transmit
2660 * @dev: network device
2661 *
2662 * Allow upper layers to call the device hard_start_xmit routine.
2663 * Used for flow control when transmit resources are available.
2664 */
79d16385
DM
2665static inline void netif_wake_queue(struct net_device *dev)
2666{
e8a0464c 2667 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2668}
2669
fd2ea0a7
DM
2670static inline void netif_tx_wake_all_queues(struct net_device *dev)
2671{
2672 unsigned int i;
2673
2674 for (i = 0; i < dev->num_tx_queues; i++) {
2675 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2676 netif_tx_wake_queue(txq);
2677 }
2678}
2679
d29f749e
DJ
2680static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2681{
73466498 2682 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2683}
2684
bea3348e
SH
2685/**
2686 * netif_stop_queue - stop transmitted packets
2687 * @dev: network device
2688 *
2689 * Stop upper layers calling the device hard_start_xmit routine.
2690 * Used for flow control when transmit resources are unavailable.
2691 */
1da177e4
LT
2692static inline void netif_stop_queue(struct net_device *dev)
2693{
e8a0464c 2694 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2695}
2696
a2029240 2697void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2698
4d29515f 2699static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2700{
73466498 2701 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2702}
2703
bea3348e
SH
2704/**
2705 * netif_queue_stopped - test if transmit queue is flowblocked
2706 * @dev: network device
2707 *
2708 * Test if transmit queue on device is currently unable to send.
2709 */
4d29515f 2710static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2711{
e8a0464c 2712 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2713}
2714
4d29515f 2715static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2716{
73466498
TH
2717 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2718}
2719
8e2f1a63
DB
2720static inline bool
2721netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2722{
2723 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2724}
2725
8e2f1a63
DB
2726static inline bool
2727netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2728{
2729 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2730}
2731
53511453
ED
2732/**
2733 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2734 * @dev_queue: pointer to transmit queue
2735 *
2736 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2737 * to give appropriate hint to the cpu.
2738 */
2739static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2740{
2741#ifdef CONFIG_BQL
2742 prefetchw(&dev_queue->dql.num_queued);
2743#endif
2744}
2745
2746/**
2747 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2748 * @dev_queue: pointer to transmit queue
2749 *
2750 * BQL enabled drivers might use this helper in their TX completion path,
2751 * to give appropriate hint to the cpu.
2752 */
2753static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2754{
2755#ifdef CONFIG_BQL
2756 prefetchw(&dev_queue->dql.limit);
2757#endif
2758}
2759
c5d67bd7
TH
2760static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2761 unsigned int bytes)
2762{
114cf580
TH
2763#ifdef CONFIG_BQL
2764 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2765
2766 if (likely(dql_avail(&dev_queue->dql) >= 0))
2767 return;
2768
2769 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2770
2771 /*
2772 * The XOFF flag must be set before checking the dql_avail below,
2773 * because in netdev_tx_completed_queue we update the dql_completed
2774 * before checking the XOFF flag.
2775 */
2776 smp_mb();
2777
2778 /* check again in case another CPU has just made room avail */
2779 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2780 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2781#endif
c5d67bd7
TH
2782}
2783
0042d0c8
FF
2784/**
2785 * netdev_sent_queue - report the number of bytes queued to hardware
2786 * @dev: network device
2787 * @bytes: number of bytes queued to the hardware device queue
2788 *
2789 * Report the number of bytes queued for sending/completion to the network
2790 * device hardware queue. @bytes should be a good approximation and should
2791 * exactly match netdev_completed_queue() @bytes
2792 */
c5d67bd7
TH
2793static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2794{
2795 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2796}
2797
2798static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2799 unsigned int pkts, unsigned int bytes)
c5d67bd7 2800{
114cf580 2801#ifdef CONFIG_BQL
b37c0fbe
AD
2802 if (unlikely(!bytes))
2803 return;
2804
2805 dql_completed(&dev_queue->dql, bytes);
2806
2807 /*
2808 * Without the memory barrier there is a small possiblity that
2809 * netdev_tx_sent_queue will miss the update and cause the queue to
2810 * be stopped forever
2811 */
2812 smp_mb();
2813
2814 if (dql_avail(&dev_queue->dql) < 0)
2815 return;
2816
2817 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2818 netif_schedule_queue(dev_queue);
114cf580 2819#endif
c5d67bd7
TH
2820}
2821
0042d0c8
FF
2822/**
2823 * netdev_completed_queue - report bytes and packets completed by device
2824 * @dev: network device
2825 * @pkts: actual number of packets sent over the medium
2826 * @bytes: actual number of bytes sent over the medium
2827 *
2828 * Report the number of bytes and packets transmitted by the network device
2829 * hardware queue over the physical medium, @bytes must exactly match the
2830 * @bytes amount passed to netdev_sent_queue()
2831 */
c5d67bd7 2832static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2833 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2834{
2835 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2836}
2837
2838static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2839{
114cf580 2840#ifdef CONFIG_BQL
5c490354 2841 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2842 dql_reset(&q->dql);
2843#endif
c5d67bd7
TH
2844}
2845
0042d0c8
FF
2846/**
2847 * netdev_reset_queue - reset the packets and bytes count of a network device
2848 * @dev_queue: network device
2849 *
2850 * Reset the bytes and packet count of a network device and clear the
2851 * software flow control OFF bit for this network device
2852 */
c5d67bd7
TH
2853static inline void netdev_reset_queue(struct net_device *dev_queue)
2854{
2855 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2856}
2857
b9507bda
DB
2858/**
2859 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2860 * @dev: network device
2861 * @queue_index: given tx queue index
2862 *
2863 * Returns 0 if given tx queue index >= number of device tx queues,
2864 * otherwise returns the originally passed tx queue index.
2865 */
2866static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2867{
2868 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2869 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2870 dev->name, queue_index,
2871 dev->real_num_tx_queues);
2872 return 0;
2873 }
2874
2875 return queue_index;
2876}
2877
bea3348e
SH
2878/**
2879 * netif_running - test if up
2880 * @dev: network device
2881 *
2882 * Test if the device has been brought up.
2883 */
4d29515f 2884static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2885{
2886 return test_bit(__LINK_STATE_START, &dev->state);
2887}
2888
f25f4e44
PWJ
2889/*
2890 * Routines to manage the subqueues on a device. We only need start
2891 * stop, and a check if it's stopped. All other device management is
2892 * done at the overall netdevice level.
2893 * Also test the device if we're multiqueue.
2894 */
bea3348e
SH
2895
2896/**
2897 * netif_start_subqueue - allow sending packets on subqueue
2898 * @dev: network device
2899 * @queue_index: sub queue index
2900 *
2901 * Start individual transmit queue of a device with multiple transmit queues.
2902 */
f25f4e44
PWJ
2903static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2904{
fd2ea0a7 2905 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2906
2907 netif_tx_start_queue(txq);
f25f4e44
PWJ
2908}
2909
bea3348e
SH
2910/**
2911 * netif_stop_subqueue - stop sending packets on subqueue
2912 * @dev: network device
2913 * @queue_index: sub queue index
2914 *
2915 * Stop individual transmit queue of a device with multiple transmit queues.
2916 */
f25f4e44
PWJ
2917static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2918{
fd2ea0a7 2919 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 2920 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2921}
2922
bea3348e
SH
2923/**
2924 * netif_subqueue_stopped - test status of subqueue
2925 * @dev: network device
2926 * @queue_index: sub queue index
2927 *
2928 * Check individual transmit queue of a device with multiple transmit queues.
2929 */
4d29515f
DM
2930static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2931 u16 queue_index)
f25f4e44 2932{
fd2ea0a7 2933 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2934
2935 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2936}
2937
4d29515f
DM
2938static inline bool netif_subqueue_stopped(const struct net_device *dev,
2939 struct sk_buff *skb)
668f895a
PE
2940{
2941 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2942}
bea3348e 2943
46e5da40 2944void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 2945
537c00de 2946#ifdef CONFIG_XPS
53af53ae 2947int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2948 u16 index);
537c00de
AD
2949#else
2950static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2951 const struct cpumask *mask,
537c00de
AD
2952 u16 index)
2953{
2954 return 0;
2955}
2956#endif
2957
5605c762
JP
2958u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2959 unsigned int num_tx_queues);
2960
a3d22a68
VZ
2961/*
2962 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2963 * as a distribution range limit for the returned value.
2964 */
2965static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 2966 struct sk_buff *skb)
a3d22a68
VZ
2967{
2968 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2969}
2970
bea3348e
SH
2971/**
2972 * netif_is_multiqueue - test if device has multiple transmit queues
2973 * @dev: network device
2974 *
2975 * Check if device has multiple transmit queues
bea3348e 2976 */
4d29515f 2977static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2978{
a02cec21 2979 return dev->num_tx_queues > 1;
f25f4e44 2980}
1da177e4 2981
f629d208 2982int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2983
a953be53 2984#ifdef CONFIG_SYSFS
f629d208 2985int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
2986#else
2987static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2988 unsigned int rxq)
2989{
2990 return 0;
2991}
2992#endif
2993
a953be53
MD
2994#ifdef CONFIG_SYSFS
2995static inline unsigned int get_netdev_rx_queue_index(
2996 struct netdev_rx_queue *queue)
2997{
2998 struct net_device *dev = queue->dev;
2999 int index = queue - dev->_rx;
3000
3001 BUG_ON(index >= dev->num_rx_queues);
3002 return index;
3003}
3004#endif
3005
16917b87 3006#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3007int netif_get_num_default_rss_queues(void);
16917b87 3008
e6247027
ED
3009enum skb_free_reason {
3010 SKB_REASON_CONSUMED,
3011 SKB_REASON_DROPPED,
3012};
3013
3014void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3015void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3016
e6247027
ED
3017/*
3018 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3019 * interrupt context or with hardware interrupts being disabled.
3020 * (in_irq() || irqs_disabled())
3021 *
3022 * We provide four helpers that can be used in following contexts :
3023 *
3024 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3025 * replacing kfree_skb(skb)
3026 *
3027 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3028 * Typically used in place of consume_skb(skb) in TX completion path
3029 *
3030 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3031 * replacing kfree_skb(skb)
3032 *
3033 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3034 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3035 */
e6247027
ED
3036static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3037{
3038 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3039}
3040
3041static inline void dev_consume_skb_irq(struct sk_buff *skb)
3042{
3043 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3044}
3045
3046static inline void dev_kfree_skb_any(struct sk_buff *skb)
3047{
3048 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3049}
3050
3051static inline void dev_consume_skb_any(struct sk_buff *skb)
3052{
3053 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3054}
1da177e4 3055
f629d208
JP
3056int netif_rx(struct sk_buff *skb);
3057int netif_rx_ni(struct sk_buff *skb);
04eb4489 3058int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3059gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3060void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3061struct sk_buff *napi_get_frags(struct napi_struct *napi);
3062gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3063struct packet_offload *gro_find_receive_by_type(__be16 type);
3064struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3065
3066static inline void napi_free_frags(struct napi_struct *napi)
3067{
3068 kfree_skb(napi->skb);
3069 napi->skb = NULL;
3070}
3071
f629d208
JP
3072int netdev_rx_handler_register(struct net_device *dev,
3073 rx_handler_func_t *rx_handler,
3074 void *rx_handler_data);
3075void netdev_rx_handler_unregister(struct net_device *dev);
3076
3077bool dev_valid_name(const char *name);
3078int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3079int dev_ethtool(struct net *net, struct ifreq *);
3080unsigned int dev_get_flags(const struct net_device *);
3081int __dev_change_flags(struct net_device *, unsigned int flags);
3082int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3083void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3084 unsigned int gchanges);
f629d208
JP
3085int dev_change_name(struct net_device *, const char *);
3086int dev_set_alias(struct net_device *, const char *, size_t);
3087int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3088int dev_set_mtu(struct net_device *, int);
3089void dev_set_group(struct net_device *, int);
3090int dev_set_mac_address(struct net_device *, struct sockaddr *);
3091int dev_change_carrier(struct net_device *, bool new_carrier);
3092int dev_get_phys_port_id(struct net_device *dev,
02637fce 3093 struct netdev_phys_item_id *ppid);
db24a904
DA
3094int dev_get_phys_port_name(struct net_device *dev,
3095 char *name, size_t len);
d746d707 3096int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3097struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3098struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3099 struct netdev_queue *txq, int *ret);
a0265d28 3100int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3101int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1ee481fb 3102bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
1da177e4 3103
20380731 3104extern int netdev_budget;
1da177e4
LT
3105
3106/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3107void netdev_run_todo(void);
1da177e4 3108
bea3348e
SH
3109/**
3110 * dev_put - release reference to device
3111 * @dev: network device
3112 *
9ef4429b 3113 * Release reference to device to allow it to be freed.
bea3348e 3114 */
1da177e4
LT
3115static inline void dev_put(struct net_device *dev)
3116{
933393f5 3117 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3118}
3119
bea3348e
SH
3120/**
3121 * dev_hold - get reference to device
3122 * @dev: network device
3123 *
9ef4429b 3124 * Hold reference to device to keep it from being freed.
bea3348e 3125 */
15333061
SH
3126static inline void dev_hold(struct net_device *dev)
3127{
933393f5 3128 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3129}
1da177e4
LT
3130
3131/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3132 * and _off may be called from IRQ context, but it is caller
3133 * who is responsible for serialization of these calls.
b00055aa
SR
3134 *
3135 * The name carrier is inappropriate, these functions should really be
3136 * called netif_lowerlayer_*() because they represent the state of any
3137 * kind of lower layer not just hardware media.
1da177e4
LT
3138 */
3139
f629d208
JP
3140void linkwatch_init_dev(struct net_device *dev);
3141void linkwatch_fire_event(struct net_device *dev);
3142void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3143
bea3348e
SH
3144/**
3145 * netif_carrier_ok - test if carrier present
3146 * @dev: network device
3147 *
3148 * Check if carrier is present on device
3149 */
4d29515f 3150static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3151{
3152 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3153}
3154
f629d208 3155unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3156
f629d208 3157void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3158
f629d208 3159void netif_carrier_on(struct net_device *dev);
1da177e4 3160
f629d208 3161void netif_carrier_off(struct net_device *dev);
1da177e4 3162
bea3348e
SH
3163/**
3164 * netif_dormant_on - mark device as dormant.
3165 * @dev: network device
3166 *
3167 * Mark device as dormant (as per RFC2863).
3168 *
3169 * The dormant state indicates that the relevant interface is not
3170 * actually in a condition to pass packets (i.e., it is not 'up') but is
3171 * in a "pending" state, waiting for some external event. For "on-
3172 * demand" interfaces, this new state identifies the situation where the
3173 * interface is waiting for events to place it in the up state.
3174 *
3175 */
b00055aa
SR
3176static inline void netif_dormant_on(struct net_device *dev)
3177{
3178 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3179 linkwatch_fire_event(dev);
3180}
3181
bea3348e
SH
3182/**
3183 * netif_dormant_off - set device as not dormant.
3184 * @dev: network device
3185 *
3186 * Device is not in dormant state.
3187 */
b00055aa
SR
3188static inline void netif_dormant_off(struct net_device *dev)
3189{
3190 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3191 linkwatch_fire_event(dev);
3192}
3193
bea3348e
SH
3194/**
3195 * netif_dormant - test if carrier present
3196 * @dev: network device
3197 *
3198 * Check if carrier is present on device
3199 */
4d29515f 3200static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3201{
3202 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3203}
3204
3205
bea3348e
SH
3206/**
3207 * netif_oper_up - test if device is operational
3208 * @dev: network device
3209 *
3210 * Check if carrier is operational
3211 */
4d29515f 3212static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3213{
b00055aa
SR
3214 return (dev->operstate == IF_OPER_UP ||
3215 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3216}
3217
bea3348e
SH
3218/**
3219 * netif_device_present - is device available or removed
3220 * @dev: network device
3221 *
3222 * Check if device has not been removed from system.
3223 */
4d29515f 3224static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3225{
3226 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3227}
3228
f629d208 3229void netif_device_detach(struct net_device *dev);
1da177e4 3230
f629d208 3231void netif_device_attach(struct net_device *dev);
1da177e4
LT
3232
3233/*
3234 * Network interface message level settings
3235 */
1da177e4
LT
3236
3237enum {
3238 NETIF_MSG_DRV = 0x0001,
3239 NETIF_MSG_PROBE = 0x0002,
3240 NETIF_MSG_LINK = 0x0004,
3241 NETIF_MSG_TIMER = 0x0008,
3242 NETIF_MSG_IFDOWN = 0x0010,
3243 NETIF_MSG_IFUP = 0x0020,
3244 NETIF_MSG_RX_ERR = 0x0040,
3245 NETIF_MSG_TX_ERR = 0x0080,
3246 NETIF_MSG_TX_QUEUED = 0x0100,
3247 NETIF_MSG_INTR = 0x0200,
3248 NETIF_MSG_TX_DONE = 0x0400,
3249 NETIF_MSG_RX_STATUS = 0x0800,
3250 NETIF_MSG_PKTDATA = 0x1000,
3251 NETIF_MSG_HW = 0x2000,
3252 NETIF_MSG_WOL = 0x4000,
3253};
3254
3255#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3256#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3257#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3258#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3259#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3260#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3261#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3262#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3263#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3264#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3265#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3266#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3267#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3268#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3269#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3270
3271static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3272{
3273 /* use default */
3274 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3275 return default_msg_enable_bits;
3276 if (debug_value == 0) /* no output */
3277 return 0;
3278 /* set low N bits */
3279 return (1 << debug_value) - 1;
3280}
3281
c773e847 3282static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3283{
c773e847
DM
3284 spin_lock(&txq->_xmit_lock);
3285 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3286}
3287
fd2ea0a7
DM
3288static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3289{
3290 spin_lock_bh(&txq->_xmit_lock);
3291 txq->xmit_lock_owner = smp_processor_id();
3292}
3293
4d29515f 3294static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3295{
4d29515f 3296 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3297 if (likely(ok))
3298 txq->xmit_lock_owner = smp_processor_id();
3299 return ok;
3300}
3301
3302static inline void __netif_tx_unlock(struct netdev_queue *txq)
3303{
3304 txq->xmit_lock_owner = -1;
3305 spin_unlock(&txq->_xmit_lock);
3306}
3307
3308static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3309{
3310 txq->xmit_lock_owner = -1;
3311 spin_unlock_bh(&txq->_xmit_lock);
3312}
3313
08baf561
ED
3314static inline void txq_trans_update(struct netdev_queue *txq)
3315{
3316 if (txq->xmit_lock_owner != -1)
3317 txq->trans_start = jiffies;
3318}
3319
d29f749e
DJ
3320/**
3321 * netif_tx_lock - grab network device transmit lock
3322 * @dev: network device
d29f749e
DJ
3323 *
3324 * Get network device transmit lock
3325 */
22dd7495
JHS
3326static inline void netif_tx_lock(struct net_device *dev)
3327{
e8a0464c 3328 unsigned int i;
c3f26a26 3329 int cpu;
c773e847 3330
c3f26a26
DM
3331 spin_lock(&dev->tx_global_lock);
3332 cpu = smp_processor_id();
e8a0464c
DM
3333 for (i = 0; i < dev->num_tx_queues; i++) {
3334 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3335
3336 /* We are the only thread of execution doing a
3337 * freeze, but we have to grab the _xmit_lock in
3338 * order to synchronize with threads which are in
3339 * the ->hard_start_xmit() handler and already
3340 * checked the frozen bit.
3341 */
e8a0464c 3342 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3343 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3344 __netif_tx_unlock(txq);
e8a0464c 3345 }
932ff279
HX
3346}
3347
3348static inline void netif_tx_lock_bh(struct net_device *dev)
3349{
e8a0464c
DM
3350 local_bh_disable();
3351 netif_tx_lock(dev);
932ff279
HX
3352}
3353
932ff279
HX
3354static inline void netif_tx_unlock(struct net_device *dev)
3355{
e8a0464c
DM
3356 unsigned int i;
3357
3358 for (i = 0; i < dev->num_tx_queues; i++) {
3359 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3360
c3f26a26
DM
3361 /* No need to grab the _xmit_lock here. If the
3362 * queue is not stopped for another reason, we
3363 * force a schedule.
3364 */
3365 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3366 netif_schedule_queue(txq);
c3f26a26
DM
3367 }
3368 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3369}
3370
3371static inline void netif_tx_unlock_bh(struct net_device *dev)
3372{
e8a0464c
DM
3373 netif_tx_unlock(dev);
3374 local_bh_enable();
932ff279
HX
3375}
3376
c773e847 3377#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3378 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3379 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3380 } \
3381}
3382
5efeac44
EB
3383#define HARD_TX_TRYLOCK(dev, txq) \
3384 (((dev->features & NETIF_F_LLTX) == 0) ? \
3385 __netif_tx_trylock(txq) : \
3386 true )
3387
c773e847 3388#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3389 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3390 __netif_tx_unlock(txq); \
22dd7495
JHS
3391 } \
3392}
3393
1da177e4
LT
3394static inline void netif_tx_disable(struct net_device *dev)
3395{
fd2ea0a7 3396 unsigned int i;
c3f26a26 3397 int cpu;
fd2ea0a7 3398
c3f26a26
DM
3399 local_bh_disable();
3400 cpu = smp_processor_id();
fd2ea0a7
DM
3401 for (i = 0; i < dev->num_tx_queues; i++) {
3402 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3403
3404 __netif_tx_lock(txq, cpu);
fd2ea0a7 3405 netif_tx_stop_queue(txq);
c3f26a26 3406 __netif_tx_unlock(txq);
fd2ea0a7 3407 }
c3f26a26 3408 local_bh_enable();
1da177e4
LT
3409}
3410
e308a5d8
DM
3411static inline void netif_addr_lock(struct net_device *dev)
3412{
3413 spin_lock(&dev->addr_list_lock);
3414}
3415
2429f7ac
JP
3416static inline void netif_addr_lock_nested(struct net_device *dev)
3417{
25175ba5
VY
3418 int subclass = SINGLE_DEPTH_NESTING;
3419
3420 if (dev->netdev_ops->ndo_get_lock_subclass)
3421 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3422
3423 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3424}
3425
e308a5d8
DM
3426static inline void netif_addr_lock_bh(struct net_device *dev)
3427{
3428 spin_lock_bh(&dev->addr_list_lock);
3429}
3430
3431static inline void netif_addr_unlock(struct net_device *dev)
3432{
3433 spin_unlock(&dev->addr_list_lock);
3434}
3435
3436static inline void netif_addr_unlock_bh(struct net_device *dev)
3437{
3438 spin_unlock_bh(&dev->addr_list_lock);
3439}
3440
f001fde5 3441/*
31278e71 3442 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3443 * rcu_read_lock held.
3444 */
3445#define for_each_dev_addr(dev, ha) \
31278e71 3446 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3447
1da177e4
LT
3448/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3449
f629d208 3450void ether_setup(struct net_device *dev);
1da177e4
LT
3451
3452/* Support for loadable net-drivers */
f629d208 3453struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3454 unsigned char name_assign_type,
f629d208
JP
3455 void (*setup)(struct net_device *),
3456 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3457#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3458 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3459
c835a677
TG
3460#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3461 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3462 count)
36909ea4 3463
f629d208
JP
3464int register_netdev(struct net_device *dev);
3465void unregister_netdev(struct net_device *dev);
f001fde5 3466
22bedad3 3467/* General hardware address lists handling functions */
f629d208
JP
3468int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3469 struct netdev_hw_addr_list *from_list, int addr_len);
3470void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3471 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3472int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3473 struct net_device *dev,
3474 int (*sync)(struct net_device *, const unsigned char *),
3475 int (*unsync)(struct net_device *,
3476 const unsigned char *));
3477void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3478 struct net_device *dev,
3479 int (*unsync)(struct net_device *,
3480 const unsigned char *));
f629d208 3481void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3482
f001fde5 3483/* Functions used for device addresses handling */
f629d208
JP
3484int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3485 unsigned char addr_type);
3486int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3487 unsigned char addr_type);
f629d208
JP
3488void dev_addr_flush(struct net_device *dev);
3489int dev_addr_init(struct net_device *dev);
a748ee24
JP
3490
3491/* Functions used for unicast addresses handling */
f629d208
JP
3492int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3493int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3494int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3495int dev_uc_sync(struct net_device *to, struct net_device *from);
3496int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3497void dev_uc_unsync(struct net_device *to, struct net_device *from);
3498void dev_uc_flush(struct net_device *dev);
3499void dev_uc_init(struct net_device *dev);
f001fde5 3500
670e5b8e
AD
3501/**
3502 * __dev_uc_sync - Synchonize device's unicast list
3503 * @dev: device to sync
3504 * @sync: function to call if address should be added
3505 * @unsync: function to call if address should be removed
3506 *
3507 * Add newly added addresses to the interface, and release
3508 * addresses that have been deleted.
3509 **/
3510static inline int __dev_uc_sync(struct net_device *dev,
3511 int (*sync)(struct net_device *,
3512 const unsigned char *),
3513 int (*unsync)(struct net_device *,
3514 const unsigned char *))
3515{
3516 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3517}
3518
3519/**
e793c0f7 3520 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3521 * @dev: device to sync
3522 * @unsync: function to call if address should be removed
3523 *
3524 * Remove all addresses that were added to the device by dev_uc_sync().
3525 **/
3526static inline void __dev_uc_unsync(struct net_device *dev,
3527 int (*unsync)(struct net_device *,
3528 const unsigned char *))
3529{
3530 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3531}
3532
22bedad3 3533/* Functions used for multicast addresses handling */
f629d208
JP
3534int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3535int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3536int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3537int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3538int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3539int dev_mc_sync(struct net_device *to, struct net_device *from);
3540int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3541void dev_mc_unsync(struct net_device *to, struct net_device *from);
3542void dev_mc_flush(struct net_device *dev);
3543void dev_mc_init(struct net_device *dev);
f001fde5 3544
670e5b8e
AD
3545/**
3546 * __dev_mc_sync - Synchonize device's multicast list
3547 * @dev: device to sync
3548 * @sync: function to call if address should be added
3549 * @unsync: function to call if address should be removed
3550 *
3551 * Add newly added addresses to the interface, and release
3552 * addresses that have been deleted.
3553 **/
3554static inline int __dev_mc_sync(struct net_device *dev,
3555 int (*sync)(struct net_device *,
3556 const unsigned char *),
3557 int (*unsync)(struct net_device *,
3558 const unsigned char *))
3559{
3560 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3561}
3562
3563/**
e793c0f7 3564 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3565 * @dev: device to sync
3566 * @unsync: function to call if address should be removed
3567 *
3568 * Remove all addresses that were added to the device by dev_mc_sync().
3569 **/
3570static inline void __dev_mc_unsync(struct net_device *dev,
3571 int (*unsync)(struct net_device *,
3572 const unsigned char *))
3573{
3574 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3575}
3576
4417da66 3577/* Functions used for secondary unicast and multicast support */
f629d208
JP
3578void dev_set_rx_mode(struct net_device *dev);
3579void __dev_set_rx_mode(struct net_device *dev);
3580int dev_set_promiscuity(struct net_device *dev, int inc);
3581int dev_set_allmulti(struct net_device *dev, int inc);
3582void netdev_state_change(struct net_device *dev);
3583void netdev_notify_peers(struct net_device *dev);
3584void netdev_features_change(struct net_device *dev);
1da177e4 3585/* Load a device via the kmod */
f629d208
JP
3586void dev_load(struct net *net, const char *name);
3587struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3588 struct rtnl_link_stats64 *storage);
3589void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3590 const struct net_device_stats *netdev_stats);
eeda3fd6 3591
1da177e4 3592extern int netdev_max_backlog;
3b098e2d 3593extern int netdev_tstamp_prequeue;
1da177e4 3594extern int weight_p;
0a14842f 3595extern int bpf_jit_enable;
9ff162a8 3596
f629d208 3597bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3598struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3599 struct list_head **iter);
f629d208
JP
3600struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3601 struct list_head **iter);
8b5be856 3602
44a40855
VY
3603/* iterate through upper list, must be called under RCU read lock */
3604#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3605 for (iter = &(dev)->adj_list.upper, \
3606 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3607 updev; \
3608 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3609
8b5be856 3610/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3611#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3612 for (iter = &(dev)->all_adj_list.upper, \
3613 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3614 updev; \
3615 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3616
f629d208
JP
3617void *netdev_lower_get_next_private(struct net_device *dev,
3618 struct list_head **iter);
3619void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3620 struct list_head **iter);
31088a11
VF
3621
3622#define netdev_for_each_lower_private(dev, priv, iter) \
3623 for (iter = (dev)->adj_list.lower.next, \
3624 priv = netdev_lower_get_next_private(dev, &(iter)); \
3625 priv; \
3626 priv = netdev_lower_get_next_private(dev, &(iter)))
3627
3628#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3629 for (iter = &(dev)->adj_list.lower, \
3630 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3631 priv; \
3632 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3633
4085ebe8
VY
3634void *netdev_lower_get_next(struct net_device *dev,
3635 struct list_head **iter);
3636#define netdev_for_each_lower_dev(dev, ldev, iter) \
3637 for (iter = &(dev)->adj_list.lower, \
3638 ldev = netdev_lower_get_next(dev, &(iter)); \
3639 ldev; \
3640 ldev = netdev_lower_get_next(dev, &(iter)))
3641
f629d208 3642void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3643void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3644struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3645struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3646int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3647int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3648 struct net_device *upper_dev,
29bf24af 3649 void *upper_priv, void *upper_info);
f629d208
JP
3650void netdev_upper_dev_unlink(struct net_device *dev,
3651 struct net_device *upper_dev);
5bb025fa 3652void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3653void *netdev_lower_dev_get_private(struct net_device *dev,
3654 struct net_device *lower_dev);
04d48266
JP
3655void netdev_lower_state_changed(struct net_device *lower_dev,
3656 void *lower_state_info);
960fb622
ED
3657
3658/* RSS keys are 40 or 52 bytes long */
3659#define NETDEV_RSS_KEY_LEN 52
3660extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3661void netdev_rss_key_fill(void *buffer, size_t len);
3662
4085ebe8 3663int dev_get_nest_level(struct net_device *dev,
b618aaa9 3664 bool (*type_check)(const struct net_device *dev));
f629d208
JP
3665int skb_checksum_help(struct sk_buff *skb);
3666struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3667 netdev_features_t features, bool tx_path);
3668struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3669 netdev_features_t features);
12b0004d 3670
61bd3857
MS
3671struct netdev_bonding_info {
3672 ifslave slave;
3673 ifbond master;
3674};
3675
3676struct netdev_notifier_bonding_info {
3677 struct netdev_notifier_info info; /* must be first */
3678 struct netdev_bonding_info bonding_info;
3679};
3680
3681void netdev_bonding_info_change(struct net_device *dev,
3682 struct netdev_bonding_info *bonding_info);
3683
12b0004d
CW
3684static inline
3685struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3686{
3687 return __skb_gso_segment(skb, features, true);
3688}
53d6471c 3689__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3690
3691static inline bool can_checksum_protocol(netdev_features_t features,
3692 __be16 protocol)
3693{
c8cd0989
TH
3694 if (protocol == htons(ETH_P_FCOE))
3695 return !!(features & NETIF_F_FCOE_CRC);
3696
3697 /* Assume this is an IP checksum (not SCTP CRC) */
3698
3699 if (features & NETIF_F_HW_CSUM) {
3700 /* Can checksum everything */
3701 return true;
3702 }
3703
3704 switch (protocol) {
3705 case htons(ETH_P_IP):
3706 return !!(features & NETIF_F_IP_CSUM);
3707 case htons(ETH_P_IPV6):
3708 return !!(features & NETIF_F_IPV6_CSUM);
3709 default:
3710 return false;
3711 }
ec5f0615 3712}
12b0004d 3713
fb286bb2 3714#ifdef CONFIG_BUG
f629d208 3715void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3716#else
3717static inline void netdev_rx_csum_fault(struct net_device *dev)
3718{
3719}
3720#endif
1da177e4 3721/* rx skb timestamps */
f629d208
JP
3722void net_enable_timestamp(void);
3723void net_disable_timestamp(void);
1da177e4 3724
20380731 3725#ifdef CONFIG_PROC_FS
f629d208 3726int __init dev_proc_init(void);
900ff8c6
CW
3727#else
3728#define dev_proc_init() 0
20380731
ACM
3729#endif
3730
4798248e 3731static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3732 struct sk_buff *skb, struct net_device *dev,
3733 bool more)
4798248e 3734{
fa2dbdc2 3735 skb->xmit_more = more ? 1 : 0;
0b725a2c 3736 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3737}
3738
10b3ad8c 3739static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3740 struct netdev_queue *txq, bool more)
4798248e
DM
3741{
3742 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3743 int rc;
4798248e 3744
fa2dbdc2 3745 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3746 if (rc == NETDEV_TX_OK)
3747 txq_trans_update(txq);
3748
3749 return rc;
4798248e
DM
3750}
3751
42a2d923
LT
3752int netdev_class_create_file_ns(struct class_attribute *class_attr,
3753 const void *ns);
3754void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3755 const void *ns);
58292cbe
TH
3756
3757static inline int netdev_class_create_file(struct class_attribute *class_attr)
3758{
3759 return netdev_class_create_file_ns(class_attr, NULL);
3760}
3761
3762static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3763{
3764 netdev_class_remove_file_ns(class_attr, NULL);
3765}
b8a9787e 3766
04600794
JB
3767extern struct kobj_ns_type_operations net_ns_type_operations;
3768
f629d208 3769const char *netdev_drivername(const struct net_device *dev);
6579e57b 3770
f629d208 3771void linkwatch_run_queue(void);
20380731 3772
da08143b
MK
3773static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3774 netdev_features_t f2)
3775{
c8cd0989
TH
3776 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3777 if (f1 & NETIF_F_HW_CSUM)
3778 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IP_CSUM);
3779 else
3780 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IP_CSUM);
3781 }
da08143b 3782
c8cd0989 3783 return f1 & f2;
da08143b
MK
3784}
3785
c8f44aff
MM
3786static inline netdev_features_t netdev_get_wanted_features(
3787 struct net_device *dev)
5455c699
MM
3788{
3789 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3790}
c8f44aff
MM
3791netdev_features_t netdev_increment_features(netdev_features_t all,
3792 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3793
3794/* Allow TSO being used on stacked device :
3795 * Performing the GSO segmentation before last device
3796 * is a performance improvement.
3797 */
3798static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3799 netdev_features_t mask)
3800{
3801 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3802}
3803
6cb6a27c 3804int __netdev_update_features(struct net_device *dev);
5455c699 3805void netdev_update_features(struct net_device *dev);
afe12cc8 3806void netdev_change_features(struct net_device *dev);
7f353bf2 3807
fc4a7489
PM
3808void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3809 struct net_device *dev);
3810
e38f3025
TM
3811netdev_features_t passthru_features_check(struct sk_buff *skb,
3812 struct net_device *dev,
3813 netdev_features_t features);
c1e756bf 3814netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3815
4d29515f 3816static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3817{
c8f44aff 3818 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3819
3820 /* check flags correspondence */
3821 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3822 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3823 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3824 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3825 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3826 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
3827 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3828 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3829 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3830 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3831 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3832 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
e585f236 3833 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
0345e186 3834
d6b4991a 3835 return (features & feature) == feature;
576a30eb
HX
3836}
3837
4d29515f 3838static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3839{
278b2513 3840 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3841 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3842}
3843
8b86a61d 3844static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 3845 netdev_features_t features)
7967168c 3846{
fc741216 3847 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3848 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3849 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3850}
3851
82cc1a7a
PWJ
3852static inline void netif_set_gso_max_size(struct net_device *dev,
3853 unsigned int size)
3854{
3855 dev->gso_max_size = size;
3856}
3857
7a7ffbab
WCC
3858static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3859 int pulled_hlen, u16 mac_offset,
3860 int mac_len)
3861{
3862 skb->protocol = protocol;
3863 skb->encapsulation = 1;
3864 skb_push(skb, pulled_hlen);
3865 skb_reset_transport_header(skb);
3866 skb->mac_header = mac_offset;
3867 skb->network_header = skb->mac_header + mac_len;
3868 skb->mac_len = mac_len;
3869}
3870
b618aaa9 3871static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
3872{
3873 return dev->priv_flags & IFF_MACVLAN;
3874}
3875
b618aaa9 3876static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
3877{
3878 return dev->priv_flags & IFF_MACVLAN_PORT;
3879}
3880
b618aaa9 3881static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
3882{
3883 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3884}
3885
b618aaa9 3886static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
3887{
3888 return dev->priv_flags & IFF_IPVLAN_MASTER;
3889}
3890
b618aaa9 3891static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 3892{
3893 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3894}
3895
b618aaa9 3896static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
3897{
3898 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3899}
3900
3bdc0eba
BG
3901static inline bool netif_supports_nofcs(struct net_device *dev)
3902{
3903 return dev->priv_flags & IFF_SUPP_NOFCS;
3904}
3905
007979ea 3906static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 3907{
007979ea 3908 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
3909}
3910
fee6d4c7
DA
3911static inline bool netif_is_l3_slave(const struct net_device *dev)
3912{
3913 return dev->priv_flags & IFF_L3MDEV_SLAVE;
3914}
3915
0894ae3f
JP
3916static inline bool netif_is_bridge_master(const struct net_device *dev)
3917{
3918 return dev->priv_flags & IFF_EBRIDGE;
3919}
3920
28f9ee22
VY
3921static inline bool netif_is_bridge_port(const struct net_device *dev)
3922{
3923 return dev->priv_flags & IFF_BRIDGE_PORT;
3924}
3925
35d4e172
JP
3926static inline bool netif_is_ovs_master(const struct net_device *dev)
3927{
3928 return dev->priv_flags & IFF_OPENVSWITCH;
3929}
3930
b618aaa9 3931static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
3932{
3933 return dev->priv_flags & IFF_TEAM;
3934}
3935
b618aaa9 3936static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
3937{
3938 return dev->priv_flags & IFF_TEAM_PORT;
3939}
3940
b618aaa9 3941static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
3942{
3943 return netif_is_bond_master(dev) || netif_is_team_master(dev);
3944}
3945
b618aaa9 3946static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
3947{
3948 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
3949}
3950
02875878
ED
3951/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3952static inline void netif_keep_dst(struct net_device *dev)
3953{
3954 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3955}
3956
505d4f73 3957extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 3958
571ba423
JP
3959/* Logging, debugging and troubleshooting/diagnostic helpers. */
3960
3961/* netdev_printk helpers, similar to dev_printk */
3962
3963static inline const char *netdev_name(const struct net_device *dev)
3964{
c6f854d5
VF
3965 if (!dev->name[0] || strchr(dev->name, '%'))
3966 return "(unnamed net_device)";
571ba423
JP
3967 return dev->name;
3968}
3969
ccc7f496
VF
3970static inline const char *netdev_reg_state(const struct net_device *dev)
3971{
3972 switch (dev->reg_state) {
3973 case NETREG_UNINITIALIZED: return " (uninitialized)";
3974 case NETREG_REGISTERED: return "";
3975 case NETREG_UNREGISTERING: return " (unregistering)";
3976 case NETREG_UNREGISTERED: return " (unregistered)";
3977 case NETREG_RELEASED: return " (released)";
3978 case NETREG_DUMMY: return " (dummy)";
3979 }
3980
3981 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3982 return " (unknown)";
3983}
3984
f629d208 3985__printf(3, 4)
6ea754eb
JP
3986void netdev_printk(const char *level, const struct net_device *dev,
3987 const char *format, ...);
f629d208 3988__printf(2, 3)
6ea754eb 3989void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 3990__printf(2, 3)
6ea754eb 3991void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 3992__printf(2, 3)
6ea754eb 3993void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 3994__printf(2, 3)
6ea754eb 3995void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 3996__printf(2, 3)
6ea754eb 3997void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 3998__printf(2, 3)
6ea754eb 3999void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4000__printf(2, 3)
6ea754eb 4001void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4002
8909c9ad
VK
4003#define MODULE_ALIAS_NETDEV(device) \
4004 MODULE_ALIAS("netdev-" device)
4005
b558c96f 4006#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4007#define netdev_dbg(__dev, format, args...) \
4008do { \
ffa10cb4 4009 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4010} while (0)
b558c96f
JC
4011#elif defined(DEBUG)
4012#define netdev_dbg(__dev, format, args...) \
4013 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4014#else
4015#define netdev_dbg(__dev, format, args...) \
4016({ \
4017 if (0) \
4018 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4019})
4020#endif
4021
4022#if defined(VERBOSE_DEBUG)
4023#define netdev_vdbg netdev_dbg
4024#else
4025
4026#define netdev_vdbg(dev, format, args...) \
4027({ \
4028 if (0) \
4029 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4030 0; \
4031})
4032#endif
4033
4034/*
4035 * netdev_WARN() acts like dev_printk(), but with the key difference
4036 * of using a WARN/WARN_ON to get the message out, including the
4037 * file/line information and a backtrace.
4038 */
4039#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4040 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4041 netdev_reg_state(dev), ##args)
571ba423 4042
b3d95c5c
JP
4043/* netif printk helpers, similar to netdev_printk */
4044
4045#define netif_printk(priv, type, level, dev, fmt, args...) \
4046do { \
4047 if (netif_msg_##type(priv)) \
4048 netdev_printk(level, (dev), fmt, ##args); \
4049} while (0)
4050
f45f4321
JP
4051#define netif_level(level, priv, type, dev, fmt, args...) \
4052do { \
4053 if (netif_msg_##type(priv)) \
4054 netdev_##level(dev, fmt, ##args); \
4055} while (0)
4056
b3d95c5c 4057#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4058 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4059#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4060 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4061#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4062 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4063#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4064 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4065#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4066 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4067#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4068 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4069#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4070 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4071
0053ea9c 4072#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4073#define netif_dbg(priv, type, netdev, format, args...) \
4074do { \
4075 if (netif_msg_##type(priv)) \
b5fb0a03 4076 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4077} while (0)
0053ea9c
JP
4078#elif defined(DEBUG)
4079#define netif_dbg(priv, type, dev, format, args...) \
4080 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4081#else
4082#define netif_dbg(priv, type, dev, format, args...) \
4083({ \
4084 if (0) \
4085 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4086 0; \
4087})
4088#endif
4089
4090#if defined(VERBOSE_DEBUG)
bcfcc450 4091#define netif_vdbg netif_dbg
b3d95c5c
JP
4092#else
4093#define netif_vdbg(priv, type, dev, format, args...) \
4094({ \
4095 if (0) \
a4ed89cb 4096 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4097 0; \
4098})
4099#endif
571ba423 4100
900ff8c6
CW
4101/*
4102 * The list of packet types we will receive (as opposed to discard)
4103 * and the routines to invoke.
4104 *
4105 * Why 16. Because with 16 the only overlap we get on a hash of the
4106 * low nibble of the protocol value is RARP/SNAP/X.25.
4107 *
4108 * NOTE: That is no longer true with the addition of VLAN tags. Not
4109 * sure which should go first, but I bet it won't make much
4110 * difference if we are running VLANs. The good news is that
4111 * this protocol won't be in the list unless compiled in, so
4112 * the average user (w/out VLANs) will not be adversely affected.
4113 * --BLG
4114 *
4115 * 0800 IP
4116 * 8100 802.1Q VLAN
4117 * 0001 802.3
4118 * 0002 AX.25
4119 * 0004 802.2
4120 * 8035 RARP
4121 * 0005 SNAP
4122 * 0805 X.25
4123 * 0806 ARP
4124 * 8137 IPX
4125 * 0009 Localtalk
4126 * 86DD IPv6
4127 */
4128#define PTYPE_HASH_SIZE (16)
4129#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4130
385a154c 4131#endif /* _LINUX_NETDEVICE_H */