]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
net: sched: fix qdisc->running lockdep annotations
[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>
e4c6734e 54#include <uapi/linux/pkt_cls.h>
a59e2ecb 55
115c1d6e 56struct netpoll_info;
313162d0 57struct device;
c1f19b51 58struct phy_device;
704232c2
JB
59/* 802.11 specific */
60struct wireless_dev;
98a18b6f
AA
61/* 802.15.4 specific */
62struct wpan_dev;
03c57747 63struct mpls_dev;
1da177e4 64
f629d208
JP
65void netdev_set_default_ethtool_ops(struct net_device *dev,
66 const struct ethtool_ops *ops);
d07d7507 67
9a1654ba
JP
68/* Backlog congestion levels */
69#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
70#define NET_RX_DROP 1 /* packet dropped */
71
572a9d7b
PM
72/*
73 * Transmit return codes: transmit return codes originate from three different
74 * namespaces:
75 *
76 * - qdisc return codes
77 * - driver transmit return codes
78 * - errno values
79 *
80 * Drivers are allowed to return any one of those in their hard_start_xmit()
81 * function. Real network devices commonly used with qdiscs should only return
82 * the driver transmit return codes though - when qdiscs are used, the actual
83 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
84 * higher layers. Virtual network devices transmit synchronously; in this case
85 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
86 * others are propagated to higher layers.
87 */
88
89/* qdisc ->enqueue() return codes. */
90#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
91#define NET_XMIT_DROP 0x01 /* skb dropped */
92#define NET_XMIT_CN 0x02 /* congestion notification */
93#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
94#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 95
b9df3cb8
GR
96/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
97 * indicates that the device will soon be dropping packets, or already drops
98 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 99#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
100#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101
dc1f8bf6 102/* Driver transmit return codes */
9a1654ba 103#define NETDEV_TX_MASK 0xf0
572a9d7b 104
dc1f8bf6 105enum netdev_tx {
572a9d7b 106 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
107 NETDEV_TX_OK = 0x00, /* driver took care of packet */
108 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
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 130/*
5e82b4b2 131 * Compute the worst-case header length according to the protocols
1da177e4
LT
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
5e82b4b2 248/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
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);
2793a23a 270 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
271};
272
1da177e4 273/* These flag bits are private to the generic network queueing
5e82b4b2 274 * layer; they may not be explicitly referenced by any other
1da177e4
LT
275 * code.
276 */
277
d94d9fee 278enum netdev_state_t {
1da177e4
LT
279 __LINK_STATE_START,
280 __LINK_STATE_PRESENT,
1da177e4 281 __LINK_STATE_NOCARRIER,
b00055aa
SR
282 __LINK_STATE_LINKWATCH_PENDING,
283 __LINK_STATE_DORMANT,
1da177e4
LT
284};
285
286
287/*
5e82b4b2 288 * This structure holds boot-time configured netdevice settings. They
fe2918b0 289 * are then used in the device probing.
1da177e4
LT
290 */
291struct netdev_boot_setup {
292 char name[IFNAMSIZ];
293 struct ifmap map;
294};
295#define NETDEV_BOOT_SETUP_MAX 8
296
f629d208 297int __init netdev_boot_setup(char *str);
1da177e4 298
bea3348e
SH
299/*
300 * Structure for NAPI scheduling similar to tasklet but with weighting
301 */
302struct napi_struct {
303 /* The poll_list must only be managed by the entity which
304 * changes the state of the NAPI_STATE_SCHED bit. This means
305 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 306 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
307 * can remove from the list right before clearing the bit.
308 */
309 struct list_head poll_list;
310
311 unsigned long state;
312 int weight;
404f7c9e 313 unsigned int gro_count;
bea3348e
SH
314 int (*poll)(struct napi_struct *, int);
315#ifdef CONFIG_NETPOLL
316 spinlock_t poll_lock;
317 int poll_owner;
bea3348e 318#endif
5d38a079 319 struct net_device *dev;
d565b0a1 320 struct sk_buff *gro_list;
5d38a079 321 struct sk_buff *skb;
3b47d303 322 struct hrtimer timer;
404f7c9e 323 struct list_head dev_list;
af12fa6e
ET
324 struct hlist_node napi_hash_node;
325 unsigned int napi_id;
bea3348e
SH
326};
327
d94d9fee 328enum {
bea3348e 329 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 330 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 331 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
332 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
333 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
bea3348e
SH
334};
335
5b252f0c 336enum gro_result {
d1c76af9
HX
337 GRO_MERGED,
338 GRO_MERGED_FREE,
339 GRO_HELD,
340 GRO_NORMAL,
341 GRO_DROP,
342};
5b252f0c 343typedef enum gro_result gro_result_t;
d1c76af9 344
8a4eb573
JP
345/*
346 * enum rx_handler_result - Possible return values for rx_handlers.
347 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
348 * further.
349 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
350 * case skb->dev was changed by rx_handler.
351 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 352 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
353 *
354 * rx_handlers are functions called from inside __netif_receive_skb(), to do
355 * special processing of the skb, prior to delivery to protocol handlers.
356 *
357 * Currently, a net_device can only have a single rx_handler registered. Trying
358 * to register a second rx_handler will return -EBUSY.
359 *
360 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
361 * To unregister a rx_handler on a net_device, use
362 * netdev_rx_handler_unregister().
363 *
364 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
365 * do with the skb.
366 *
5e82b4b2 367 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 368 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 369 * the skb to be delivered in some other way.
8a4eb573
JP
370 *
371 * If the rx_handler changed skb->dev, to divert the skb to another
372 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
373 * new device will be called if it exists.
374 *
5e82b4b2 375 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 376 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 377 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 378 *
5e82b4b2 379 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
380 * delivered, it should return RX_HANDLER_PASS.
381 *
382 * A device without a registered rx_handler will behave as if rx_handler
383 * returned RX_HANDLER_PASS.
384 */
385
386enum rx_handler_result {
387 RX_HANDLER_CONSUMED,
388 RX_HANDLER_ANOTHER,
389 RX_HANDLER_EXACT,
390 RX_HANDLER_PASS,
391};
392typedef enum rx_handler_result rx_handler_result_t;
393typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 394
f629d208 395void __napi_schedule(struct napi_struct *n);
bc9ad166 396void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 397
4d29515f 398static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
399{
400 return test_bit(NAPI_STATE_DISABLE, &n->state);
401}
402
bea3348e 403/**
5e82b4b2
BH
404 * napi_schedule_prep - check if NAPI can be scheduled
405 * @n: NAPI context
bea3348e
SH
406 *
407 * Test if NAPI routine is already running, and if not mark
5e82b4b2 408 * it as running. This is used as a condition variable to
a0a46196
DM
409 * insure only one NAPI poll instance runs. We also make
410 * sure there is no pending NAPI disable.
bea3348e 411 */
4d29515f 412static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 413{
a0a46196
DM
414 return !napi_disable_pending(n) &&
415 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
416}
417
418/**
419 * napi_schedule - schedule NAPI poll
5e82b4b2 420 * @n: NAPI context
bea3348e
SH
421 *
422 * Schedule NAPI poll routine to be called if it is not already
423 * running.
424 */
425static inline void napi_schedule(struct napi_struct *n)
426{
427 if (napi_schedule_prep(n))
428 __napi_schedule(n);
429}
430
bc9ad166
ED
431/**
432 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 433 * @n: NAPI context
bc9ad166
ED
434 *
435 * Variant of napi_schedule(), assuming hard irqs are masked.
436 */
437static inline void napi_schedule_irqoff(struct napi_struct *n)
438{
439 if (napi_schedule_prep(n))
440 __napi_schedule_irqoff(n);
441}
442
bfe13f54 443/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 444static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
445{
446 if (napi_schedule_prep(napi)) {
447 __napi_schedule(napi);
4d29515f 448 return true;
bfe13f54 449 }
4d29515f 450 return false;
bfe13f54
RD
451}
452
3b47d303
ED
453void __napi_complete(struct napi_struct *n);
454void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
455/**
456 * napi_complete - NAPI processing complete
5e82b4b2 457 * @n: NAPI context
bea3348e
SH
458 *
459 * Mark NAPI processing as complete.
3b47d303 460 * Consider using napi_complete_done() instead.
bea3348e 461 */
3b47d303
ED
462static inline void napi_complete(struct napi_struct *n)
463{
464 return napi_complete_done(n, 0);
465}
bea3348e 466
af12fa6e
ET
467/**
468 * napi_hash_add - add a NAPI to global hashtable
5e82b4b2 469 * @napi: NAPI context
af12fa6e 470 *
5e82b4b2
BH
471 * Generate a new napi_id and store a @napi under it in napi_hash.
472 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
93d05d4a 473 * Note: This is normally automatically done from netif_napi_add(),
5e82b4b2 474 * so might disappear in a future Linux version.
af12fa6e 475 */
f629d208 476void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
477
478/**
479 * napi_hash_del - remove a NAPI from global table
5e82b4b2 480 * @napi: NAPI context
af12fa6e 481 *
5e82b4b2 482 * Warning: caller must observe RCU grace period
34cbe27e
ED
483 * before freeing memory containing @napi, if
484 * this function returns true.
93d05d4a 485 * Note: core networking stack automatically calls it
5e82b4b2 486 * from netif_napi_del().
93d05d4a 487 * Drivers might want to call this helper to combine all
5e82b4b2 488 * the needed RCU grace periods into a single one.
af12fa6e 489 */
34cbe27e 490bool napi_hash_del(struct napi_struct *napi);
af12fa6e 491
bea3348e
SH
492/**
493 * napi_disable - prevent NAPI from scheduling
5e82b4b2 494 * @n: NAPI context
bea3348e
SH
495 *
496 * Stop NAPI from being scheduled on this context.
497 * Waits till any outstanding processing completes.
498 */
3b47d303 499void napi_disable(struct napi_struct *n);
bea3348e
SH
500
501/**
502 * napi_enable - enable NAPI scheduling
5e82b4b2 503 * @n: NAPI context
bea3348e
SH
504 *
505 * Resume NAPI from being scheduled on this context.
506 * Must be paired with napi_disable.
507 */
508static inline void napi_enable(struct napi_struct *n)
509{
510 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 511 smp_mb__before_atomic();
bea3348e 512 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 513 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
514}
515
c264c3de
SH
516/**
517 * napi_synchronize - wait until NAPI is not running
5e82b4b2 518 * @n: NAPI context
c264c3de
SH
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{
facc432f
AB
526 if (IS_ENABLED(CONFIG_SMP))
527 while (test_bit(NAPI_STATE_SCHED, &n->state))
528 msleep(1);
529 else
530 barrier();
c264c3de 531}
c264c3de 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 560/*
5e82b4b2 561 * read-mostly part
6a321cb3 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
c0ef079c
FW
572 unsigned long tx_maxrate;
573 /*
574 * Number of TX timeouts for this queue
575 * (/sys/class/net/DEV/Q/trans_timeout)
576 */
577 unsigned long trans_timeout;
6a321cb3 578/*
5e82b4b2 579 * write-mostly part
6a321cb3
ED
580 */
581 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
582 int xmit_lock_owner;
9d21493b 583 /*
9b36627a 584 * Time (in jiffies) of last Tx
9d21493b
ED
585 */
586 unsigned long trans_start;
ccf5ff69 587
114cf580
TH
588 unsigned long state;
589
590#ifdef CONFIG_BQL
591 struct dql dql;
592#endif
e8a0464c 593} ____cacheline_aligned_in_smp;
bb949fbd 594
f2cd2d3e
ED
595static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
596{
597#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
598 return q->numa_node;
599#else
b236da69 600 return NUMA_NO_NODE;
f2cd2d3e
ED
601#endif
602}
603
604static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
605{
606#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
607 q->numa_node = node;
608#endif
609}
610
df334545 611#ifdef CONFIG_RPS
0a9627f2
TH
612/*
613 * This structure holds an RPS map which can be of variable length. The
614 * map is an array of CPUs.
615 */
616struct rps_map {
617 unsigned int len;
618 struct rcu_head rcu;
619 u16 cpus[0];
620};
60b778ce 621#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 622
fec5e652 623/*
c445477d
BH
624 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
625 * tail pointer for that CPU's input queue at the time of last enqueue, and
626 * a hardware filter index.
fec5e652
TH
627 */
628struct rps_dev_flow {
629 u16 cpu;
c445477d 630 u16 filter;
fec5e652
TH
631 unsigned int last_qtail;
632};
c445477d 633#define RPS_NO_FILTER 0xffff
fec5e652
TH
634
635/*
636 * The rps_dev_flow_table structure contains a table of flow mappings.
637 */
638struct rps_dev_flow_table {
639 unsigned int mask;
640 struct rcu_head rcu;
fec5e652
TH
641 struct rps_dev_flow flows[0];
642};
643#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 644 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
645
646/*
647 * The rps_sock_flow_table contains mappings of flows to the last CPU
648 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
649 * Each entry is a 32bit value. Upper part is the high-order bits
650 * of flow hash, lower part is CPU number.
567e4b79 651 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
652 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
653 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 654 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
655 */
656struct rps_sock_flow_table {
567e4b79 657 u32 mask;
93c1af6c
ED
658
659 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 660};
567e4b79 661#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
662
663#define RPS_NO_CPU 0xffff
664
567e4b79
ED
665extern u32 rps_cpu_mask;
666extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
667
fec5e652
TH
668static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
669 u32 hash)
670{
671 if (table && hash) {
567e4b79
ED
672 unsigned int index = hash & table->mask;
673 u32 val = hash & ~rps_cpu_mask;
fec5e652 674
5e82b4b2 675 /* We only give a hint, preemption can change CPU under us */
567e4b79 676 val |= raw_smp_processor_id();
fec5e652 677
567e4b79
ED
678 if (table->ents[index] != val)
679 table->ents[index] = val;
fec5e652
TH
680 }
681}
682
c445477d 683#ifdef CONFIG_RFS_ACCEL
f629d208
JP
684bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
685 u16 filter_id);
c445477d 686#endif
a953be53 687#endif /* CONFIG_RPS */
c445477d 688
0a9627f2
TH
689/* This structure contains an instance of an RX queue. */
690struct netdev_rx_queue {
a953be53 691#ifdef CONFIG_RPS
6e3f7faf
ED
692 struct rps_map __rcu *rps_map;
693 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 694#endif
6e3f7faf 695 struct kobject kobj;
fe822240 696 struct net_device *dev;
0a9627f2 697} ____cacheline_aligned_in_smp;
a953be53
MD
698
699/*
700 * RX queue sysfs structures and functions.
701 */
702struct rx_queue_attribute {
703 struct attribute attr;
704 ssize_t (*show)(struct netdev_rx_queue *queue,
705 struct rx_queue_attribute *attr, char *buf);
706 ssize_t (*store)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, const char *buf, size_t len);
708};
d314774c 709
bf264145
TH
710#ifdef CONFIG_XPS
711/*
712 * This structure holds an XPS map which can be of variable length. The
713 * map is an array of queues.
714 */
715struct xps_map {
716 unsigned int len;
717 unsigned int alloc_len;
718 struct rcu_head rcu;
719 u16 queues[0];
720};
60b778ce 721#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
722#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
723 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
724
725/*
726 * This structure holds all XPS maps for device. Maps are indexed by CPU.
727 */
728struct xps_dev_maps {
729 struct rcu_head rcu;
a4177869 730 struct xps_map __rcu *cpu_map[0];
bf264145
TH
731};
732#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
733 (nr_cpu_ids * sizeof(struct xps_map *)))
734#endif /* CONFIG_XPS */
735
4f57c087
JF
736#define TC_MAX_QUEUE 16
737#define TC_BITMASK 15
738/* HW offloaded queuing disciplines txq count and offset maps */
739struct netdev_tc_txq {
740 u16 count;
741 u16 offset;
742};
743
68bad94e
NP
744#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
745/*
746 * This structure is to hold information about the device
747 * configured to run FCoE protocol stack.
748 */
749struct netdev_fcoe_hbainfo {
750 char manufacturer[64];
751 char serial_number[64];
752 char hardware_version[64];
753 char driver_version[64];
754 char optionrom_version[64];
755 char firmware_version[64];
756 char model[256];
757 char model_description[256];
758};
759#endif
760
02637fce 761#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 762
02637fce
JP
763/* This structure holds a unique identifier to identify some
764 * physical item (port for example) used by a netdevice.
66b52b0d 765 */
02637fce
JP
766struct netdev_phys_item_id {
767 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
768 unsigned char id_len;
769};
770
d754f98b
SF
771static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
772 struct netdev_phys_item_id *b)
773{
774 return a->id_len == b->id_len &&
775 memcmp(a->id, b->id, a->id_len) == 0;
776}
777
99932d4f
DB
778typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
779 struct sk_buff *skb);
780
a1b7c5fd 781/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
782 * that are being passed to the netdevice through the setup_tc op.
783 */
784enum {
785 TC_SETUP_MQPRIO,
a1b7c5fd 786 TC_SETUP_CLSU32,
5b33f488 787 TC_SETUP_CLSFLOWER,
16e5cc64
JF
788};
789
a1b7c5fd
JF
790struct tc_cls_u32_offload;
791
16e5cc64
JF
792struct tc_to_netdev {
793 unsigned int type;
794 union {
795 u8 tc;
a1b7c5fd 796 struct tc_cls_u32_offload *cls_u32;
5b33f488 797 struct tc_cls_flower_offload *cls_flower;
16e5cc64
JF
798 };
799};
800
801
d314774c
SH
802/*
803 * This structure defines the management hooks for network devices.
00829823
SH
804 * The following hooks can be defined; unless noted otherwise, they are
805 * optional and can be filled with a null pointer.
d314774c
SH
806 *
807 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
808 * This function is called once when a network device is registered.
809 * The network device can use this for any late stage initialization
810 * or semantic validation. It can fail with an error code which will
811 * be propagated back to register_netdev.
d314774c
SH
812 *
813 * void (*ndo_uninit)(struct net_device *dev);
814 * This function is called when device is unregistered or when registration
815 * fails. It is not called if init fails.
816 *
817 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 818 * This function is called when a network device transitions to the up
d314774c
SH
819 * state.
820 *
821 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 822 * This function is called when a network device transitions to the down
d314774c
SH
823 * state.
824 *
dc1f8bf6
SH
825 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
826 * struct net_device *dev);
00829823 827 * Called when a packet needs to be transmitted.
e79d8429
RR
828 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
829 * the queue before that can happen; it's for obsolete devices and weird
830 * corner cases, but the stack really does a non-trivial amount
831 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 832 * Required; cannot be NULL.
00829823 833 *
cdba756f
ED
834 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
835 * netdev_features_t features);
836 * Adjusts the requested feature flags according to device-specific
837 * constraints, and returns the resulting flags. Must not modify
838 * the device state.
839 *
f663dd9a 840 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 841 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 842 * Called to decide which queue to use when device supports multiple
00829823
SH
843 * transmit queues.
844 *
d314774c
SH
845 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
846 * This function is called to allow device receiver to make
5e82b4b2 847 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
848 *
849 * void (*ndo_set_rx_mode)(struct net_device *dev);
850 * This function is called device changes address list filtering.
01789349 851 * If driver handles unicast address filtering, it should set
5e82b4b2 852 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
853 *
854 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
855 * This function is called when the Media Access Control address
37b607c5 856 * needs to be changed. If this interface is not defined, the
5e82b4b2 857 * MAC address can not be changed.
d314774c
SH
858 *
859 * int (*ndo_validate_addr)(struct net_device *dev);
860 * Test if Media Access Control address is valid for the device.
861 *
862 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
863 * Called when a user requests an ioctl which can't be handled by
864 * the generic interface code. If not defined ioctls return
d314774c
SH
865 * not supported error code.
866 *
867 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
868 * Used to set network devices bus interface parameters. This interface
5e82b4b2 869 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
870 * interface (PCI) for low level management.
871 *
872 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
873 * Called when a user wants to change the Maximum Transfer Unit
874 * of a device. If not defined, any request to change MTU will
875 * will return an error.
876 *
00829823 877 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 878 * Callback used when the transmitter has not made any progress
d314774c
SH
879 * for dev->watchdog ticks.
880 *
3cfde79c 881 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 882 * struct rtnl_link_stats64 *storage);
d308e38f 883 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 884 * Called when a user wants to get the network device usage
be1f3c2c 885 * statistics. Drivers must do one of the following:
3cfde79c
BH
886 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
887 * rtnl_link_stats64 structure passed by the caller.
82695d9b 888 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
889 * (which should normally be dev->stats) and return a pointer to
890 * it. The structure may be changed asynchronously only if each
891 * field is written atomically.
892 * 3. Update dev->stats asynchronously and atomically, and define
893 * neither operation.
d314774c 894 *
5d632cb7 895 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 896 * If device supports VLAN filtering this function is called when a
80d5c368 897 * VLAN id is registered.
d314774c 898 *
5d632cb7 899 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 900 * If device supports VLAN filtering this function is called when a
80d5c368 901 * VLAN id is unregistered.
d314774c
SH
902 *
903 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
904 *
905 * SR-IOV management functions.
906 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
907 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
908 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
909 * int max_tx_rate);
5f8444a3 910 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 911 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
912 * int (*ndo_get_vf_config)(struct net_device *dev,
913 * int vf, struct ifla_vf_info *ivf);
1d8faf48 914 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
915 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
916 * struct nlattr *port[]);
01a3d796
VZ
917 *
918 * Enable or disable the VF ability to query its RSS Redirection Table and
919 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 920 * with PF and querying it may introduce a theoretical security risk.
01a3d796 921 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 922 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
923 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
924 * Called to setup 'tc' number of traffic classes in the net device. This
925 * is always called from the stack with the rtnl lock held and netif tx
926 * queues stopped. This allows the netdevice to perform queue management
927 * safely.
c445477d 928 *
e9bce845
YZ
929 * Fiber Channel over Ethernet (FCoE) offload functions.
930 * int (*ndo_fcoe_enable)(struct net_device *dev);
931 * Called when the FCoE protocol stack wants to start using LLD for FCoE
932 * so the underlying device can perform whatever needed configuration or
933 * initialization to support acceleration of FCoE traffic.
934 *
935 * int (*ndo_fcoe_disable)(struct net_device *dev);
936 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
937 * so the underlying device can perform whatever needed clean-ups to
938 * stop supporting acceleration of FCoE traffic.
939 *
940 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
941 * struct scatterlist *sgl, unsigned int sgc);
942 * Called when the FCoE Initiator wants to initialize an I/O that
943 * is a possible candidate for Direct Data Placement (DDP). The LLD can
944 * perform necessary setup and returns 1 to indicate the device is set up
945 * successfully to perform DDP on this I/O, otherwise this returns 0.
946 *
947 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
948 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
949 * indicated by the FC exchange id 'xid', so the underlying device can
950 * clean up and reuse resources for later DDP requests.
951 *
952 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
953 * struct scatterlist *sgl, unsigned int sgc);
954 * Called when the FCoE Target wants to initialize an I/O that
955 * is a possible candidate for Direct Data Placement (DDP). The LLD can
956 * perform necessary setup and returns 1 to indicate the device is set up
957 * successfully to perform DDP on this I/O, otherwise this returns 0.
958 *
68bad94e
NP
959 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
960 * struct netdev_fcoe_hbainfo *hbainfo);
961 * Called when the FCoE Protocol stack wants information on the underlying
962 * device. This information is utilized by the FCoE protocol stack to
963 * register attributes with Fiber Channel management service as per the
964 * FC-GS Fabric Device Management Information(FDMI) specification.
965 *
e9bce845
YZ
966 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
967 * Called when the underlying device wants to override default World Wide
968 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
969 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
970 * protocol stack to use.
971 *
c445477d
BH
972 * RFS acceleration.
973 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
974 * u16 rxq_index, u32 flow_id);
975 * Set hardware filter for RFS. rxq_index is the target queue index;
976 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
977 * Return the filter ID on success, or a negative error code.
fbaec0ea 978 *
8b98a70c 979 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
980 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
981 * Called to make another netdev an underling.
982 *
983 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
984 * Called to release previously enslaved netdev.
5455c699
MM
985 *
986 * Feature/offload setting functions.
c8f44aff 987 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
988 * Called to update device configuration to new features. Passed
989 * feature set might be less than what was returned by ndo_fix_features()).
990 * Must return >0 or -errno if it changed dev->features itself.
991 *
edc7d573 992 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
993 * struct net_device *dev,
f6f6424b 994 * const unsigned char *addr, u16 vid, u16 flags)
77162022 995 * Adds an FDB entry to dev for addr.
1690be63
VY
996 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
997 * struct net_device *dev,
f6f6424b 998 * const unsigned char *addr, u16 vid)
77162022
JF
999 * Deletes the FDB entry from dev coresponding to addr.
1000 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
1001 * struct net_device *dev, struct net_device *filter_dev,
1002 * int idx)
77162022
JF
1003 * Used to add FDB entries to dump requests. Implementers should add
1004 * entries to skb and update idx with the number of entries.
e5a55a89 1005 *
ad41faa8
ND
1006 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1007 * u16 flags)
e5a55a89 1008 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1009 * struct net_device *dev, u32 filter_mask,
1010 * int nlflags)
ad41faa8
ND
1011 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1012 * u16 flags);
4bf84c35
JP
1013 *
1014 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1015 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1016 * which do not represent real hardware may define this to allow their
1017 * userspace components to manage their virtual carrier state. Devices
1018 * that determine carrier state from physical hardware properties (eg
1019 * network cables) or protocol-dependent mechanisms (eg
1020 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1021 *
1022 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1023 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1024 * Called to get ID of physical port of this device. If driver does
1025 * not implement this, it is assumed that the hw is not able to have
1026 * multiple net devices on single physical port.
53cf5275
JG
1027 *
1028 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 1029 * sa_family_t sa_family, __be16 port);
5e82b4b2
BH
1030 * Called by vxlan to notify a driver about the UDP port and socket
1031 * address family that vxlan is listening to. It is called only when
53cf5275
JG
1032 * a new port starts listening. The operation is protected by the
1033 * vxlan_net->sock_lock.
1034 *
a8170d2b 1035 * void (*ndo_add_geneve_port)(struct net_device *dev,
5e82b4b2 1036 * sa_family_t sa_family, __be16 port);
a8170d2b
SA
1037 * Called by geneve to notify a driver about the UDP port and socket
1038 * address family that geneve is listnening to. It is called only when
1039 * a new port starts listening. The operation is protected by the
1040 * geneve_net->sock_lock.
1041 *
1042 * void (*ndo_del_geneve_port)(struct net_device *dev,
5e82b4b2 1043 * sa_family_t sa_family, __be16 port);
a8170d2b
SA
1044 * Called by geneve to notify the driver about a UDP port and socket
1045 * address family that geneve is not listening to anymore. The operation
1046 * is protected by the geneve_net->sock_lock.
1047 *
53cf5275 1048 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 1049 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1050 * Called by vxlan to notify the driver about a UDP port and socket
1051 * address family that vxlan is not listening to anymore. The operation
1052 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
1053 *
1054 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1055 * struct net_device *dev)
1056 * Called by upper layer devices to accelerate switching or other
1057 * station functionality into hardware. 'pdev is the lowerdev
1058 * to use for the offload and 'dev' is the net device that will
1059 * back the offload. Returns a pointer to the private structure
1060 * the upper layer will maintain.
1061 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1062 * Called by upper layer device to delete the station created
1063 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1064 * the station and priv is the structure returned by the add
1065 * operation.
1066 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1067 * struct net_device *dev,
1068 * void *priv);
1069 * Callback to use for xmit over the accelerated station. This
1070 * is used in place of ndo_start_xmit on accelerated net
1071 * devices.
5e82b4b2
BH
1072 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1073 * struct net_device *dev
1074 * netdev_features_t features);
04ffcb25 1075 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1076 * performing offload operations on a given packet. This is to give
1077 * the device an opportunity to implement any restrictions that cannot
1078 * be otherwise expressed by feature flags. The check is called with
1079 * the set of features that the stack has calculated and it returns
1080 * those the driver believes to be appropriate.
822b3b2e
JF
1081 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1082 * int queue_index, u32 maxrate);
1083 * Called when a user wants to set a max-rate limitation of specific
1084 * TX queue.
a54acb3a
ND
1085 * int (*ndo_get_iflink)(const struct net_device *dev);
1086 * Called to get the iflink value of this device.
d746d707 1087 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1088 * bool proto_down);
d746d707
AK
1089 * This function is used to pass protocol port error state information
1090 * to the switch driver. The switch driver can react to the proto_down
1091 * by doing a phys down on the associated switch port.
fc4099f1
PS
1092 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1093 * This function is used to get egress tunnel information for given skb.
1094 * This is useful for retrieving outer tunnel header parameters while
1095 * sampling packet.
871b642a
PA
1096 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1097 * This function is used to specify the headroom that the skb must
1098 * consider when allocation skb during packet reception. Setting
1099 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1100 * forward. Setting a negative value resets the rx headroom to the
871b642a 1101 * default value.
d746d707 1102 *
d314774c
SH
1103 */
1104struct net_device_ops {
1105 int (*ndo_init)(struct net_device *dev);
1106 void (*ndo_uninit)(struct net_device *dev);
1107 int (*ndo_open)(struct net_device *dev);
1108 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1109 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1110 struct net_device *dev);
1111 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1112 struct net_device *dev,
1113 netdev_features_t features);
00829823 1114 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1115 struct sk_buff *skb,
99932d4f
DB
1116 void *accel_priv,
1117 select_queue_fallback_t fallback);
d314774c
SH
1118 void (*ndo_change_rx_flags)(struct net_device *dev,
1119 int flags);
d314774c 1120 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1121 int (*ndo_set_mac_address)(struct net_device *dev,
1122 void *addr);
d314774c 1123 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1124 int (*ndo_do_ioctl)(struct net_device *dev,
1125 struct ifreq *ifr, int cmd);
d314774c
SH
1126 int (*ndo_set_config)(struct net_device *dev,
1127 struct ifmap *map);
00829823
SH
1128 int (*ndo_change_mtu)(struct net_device *dev,
1129 int new_mtu);
1130 int (*ndo_neigh_setup)(struct net_device *dev,
1131 struct neigh_parms *);
d314774c
SH
1132 void (*ndo_tx_timeout) (struct net_device *dev);
1133
28172739
ED
1134 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1135 struct rtnl_link_stats64 *storage);
d314774c
SH
1136 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1137
8e586137 1138 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1139 __be16 proto, u16 vid);
8e586137 1140 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1141 __be16 proto, u16 vid);
d314774c 1142#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1143 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1144 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1145 struct netpoll_info *info);
0e34e931 1146 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1147#endif
e0d1095a 1148#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1149 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1150#endif
95c26df8
WM
1151 int (*ndo_set_vf_mac)(struct net_device *dev,
1152 int queue, u8 *mac);
1153 int (*ndo_set_vf_vlan)(struct net_device *dev,
1154 int queue, u16 vlan, u8 qos);
ed616689
SC
1155 int (*ndo_set_vf_rate)(struct net_device *dev,
1156 int vf, int min_tx_rate,
1157 int max_tx_rate);
5f8444a3
GR
1158 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1159 int vf, bool setting);
dd461d6a
HS
1160 int (*ndo_set_vf_trust)(struct net_device *dev,
1161 int vf, bool setting);
95c26df8
WM
1162 int (*ndo_get_vf_config)(struct net_device *dev,
1163 int vf,
1164 struct ifla_vf_info *ivf);
1d8faf48
RE
1165 int (*ndo_set_vf_link_state)(struct net_device *dev,
1166 int vf, int link_state);
3b766cd8
EBE
1167 int (*ndo_get_vf_stats)(struct net_device *dev,
1168 int vf,
1169 struct ifla_vf_stats
1170 *vf_stats);
57b61080
SF
1171 int (*ndo_set_vf_port)(struct net_device *dev,
1172 int vf,
1173 struct nlattr *port[]);
1174 int (*ndo_get_vf_port)(struct net_device *dev,
1175 int vf, struct sk_buff *skb);
cc8e27cc
EC
1176 int (*ndo_set_vf_guid)(struct net_device *dev,
1177 int vf, u64 guid,
1178 int guid_type);
01a3d796
VZ
1179 int (*ndo_set_vf_rss_query_en)(
1180 struct net_device *dev,
1181 int vf, bool setting);
16e5cc64
JF
1182 int (*ndo_setup_tc)(struct net_device *dev,
1183 u32 handle,
1184 __be16 protocol,
1185 struct tc_to_netdev *tc);
d11ead75 1186#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1187 int (*ndo_fcoe_enable)(struct net_device *dev);
1188 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1189 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1190 u16 xid,
1191 struct scatterlist *sgl,
1192 unsigned int sgc);
1193 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1194 u16 xid);
6247e086
YZ
1195 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1196 u16 xid,
1197 struct scatterlist *sgl,
1198 unsigned int sgc);
68bad94e
NP
1199 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1200 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1201#endif
1202
d11ead75 1203#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1204#define NETDEV_FCOE_WWNN 0
1205#define NETDEV_FCOE_WWPN 1
1206 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1207 u64 *wwn, int type);
4d288d57 1208#endif
3c9c36bc 1209
c445477d
BH
1210#ifdef CONFIG_RFS_ACCEL
1211 int (*ndo_rx_flow_steer)(struct net_device *dev,
1212 const struct sk_buff *skb,
1213 u16 rxq_index,
1214 u32 flow_id);
1215#endif
fbaec0ea
JP
1216 int (*ndo_add_slave)(struct net_device *dev,
1217 struct net_device *slave_dev);
1218 int (*ndo_del_slave)(struct net_device *dev,
1219 struct net_device *slave_dev);
c8f44aff
MM
1220 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1221 netdev_features_t features);
5455c699 1222 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1223 netdev_features_t features);
da6a8fa0 1224 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1225 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1226
1227 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1228 struct nlattr *tb[],
77162022 1229 struct net_device *dev,
6b6e2725 1230 const unsigned char *addr,
f6f6424b 1231 u16 vid,
77162022
JF
1232 u16 flags);
1233 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1234 struct nlattr *tb[],
77162022 1235 struct net_device *dev,
f6f6424b
JP
1236 const unsigned char *addr,
1237 u16 vid);
77162022
JF
1238 int (*ndo_fdb_dump)(struct sk_buff *skb,
1239 struct netlink_callback *cb,
1240 struct net_device *dev,
5d5eacb3 1241 struct net_device *filter_dev,
77162022 1242 int idx);
e5a55a89
JF
1243
1244 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1245 struct nlmsghdr *nlh,
1246 u16 flags);
e5a55a89
JF
1247 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1248 u32 pid, u32 seq,
6cbdceeb 1249 struct net_device *dev,
46c264da
ND
1250 u32 filter_mask,
1251 int nlflags);
407af329 1252 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1253 struct nlmsghdr *nlh,
1254 u16 flags);
4bf84c35
JP
1255 int (*ndo_change_carrier)(struct net_device *dev,
1256 bool new_carrier);
66b52b0d 1257 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1258 struct netdev_phys_item_id *ppid);
db24a904
DA
1259 int (*ndo_get_phys_port_name)(struct net_device *dev,
1260 char *name, size_t len);
53cf5275
JG
1261 void (*ndo_add_vxlan_port)(struct net_device *dev,
1262 sa_family_t sa_family,
35e42379 1263 __be16 port);
53cf5275
JG
1264 void (*ndo_del_vxlan_port)(struct net_device *dev,
1265 sa_family_t sa_family,
35e42379 1266 __be16 port);
a8170d2b
SA
1267 void (*ndo_add_geneve_port)(struct net_device *dev,
1268 sa_family_t sa_family,
1269 __be16 port);
1270 void (*ndo_del_geneve_port)(struct net_device *dev,
1271 sa_family_t sa_family,
1272 __be16 port);
a6cc0cfa
JF
1273 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1274 struct net_device *dev);
1275 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1276 void *priv);
1277
1278 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1279 struct net_device *dev,
1280 void *priv);
25175ba5 1281 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1282 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1283 int queue_index,
1284 u32 maxrate);
a54acb3a 1285 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1286 int (*ndo_change_proto_down)(struct net_device *dev,
1287 bool proto_down);
fc4099f1
PS
1288 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1289 struct sk_buff *skb);
871b642a
PA
1290 void (*ndo_set_rx_headroom)(struct net_device *dev,
1291 int needed_headroom);
d314774c
SH
1292};
1293
7aa98047
LR
1294/**
1295 * enum net_device_priv_flags - &struct net_device priv_flags
1296 *
1297 * These are the &struct net_device, they are only set internally
1298 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1299 * userspace; this means that the order of these flags can change
7aa98047
LR
1300 * during any kernel release.
1301 *
1302 * You should have a pretty good reason to be extending these flags.
1303 *
1304 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1305 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1306 * @IFF_BONDING: bonding master or slave
7aa98047 1307 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1308 * @IFF_WAN_HDLC: WAN HDLC device
1309 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1310 * release skb->dst
1311 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1312 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1313 * @IFF_MACVLAN_PORT: device used as macvlan port
1314 * @IFF_BRIDGE_PORT: device used as bridge port
1315 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1316 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1317 * @IFF_UNICAST_FLT: Supports unicast filtering
1318 * @IFF_TEAM_PORT: device used as team port
1319 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1320 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1321 * change when it's running
1322 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1323 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1324 * underlying stacked devices
1325 * @IFF_IPVLAN_MASTER: IPvlan master device
1326 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1327 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1328 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1329 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1330 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1331 * @IFF_TEAM: device is a team device
d4ab4286 1332 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1333 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1334 * entity (i.e. the master device for bridged veth)
3c175784 1335 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1336 */
1337enum netdev_priv_flags {
1338 IFF_802_1Q_VLAN = 1<<0,
1339 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1340 IFF_BONDING = 1<<2,
1341 IFF_ISATAP = 1<<3,
1342 IFF_WAN_HDLC = 1<<4,
1343 IFF_XMIT_DST_RELEASE = 1<<5,
1344 IFF_DONT_BRIDGE = 1<<6,
1345 IFF_DISABLE_NETPOLL = 1<<7,
1346 IFF_MACVLAN_PORT = 1<<8,
1347 IFF_BRIDGE_PORT = 1<<9,
1348 IFF_OVS_DATAPATH = 1<<10,
1349 IFF_TX_SKB_SHARING = 1<<11,
1350 IFF_UNICAST_FLT = 1<<12,
1351 IFF_TEAM_PORT = 1<<13,
1352 IFF_SUPP_NOFCS = 1<<14,
1353 IFF_LIVE_ADDR_CHANGE = 1<<15,
1354 IFF_MACVLAN = 1<<16,
1355 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1356 IFF_IPVLAN_MASTER = 1<<18,
1357 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1358 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1359 IFF_NO_QUEUE = 1<<21,
1360 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1361 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1362 IFF_TEAM = 1<<24,
d4ab4286 1363 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1364 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1365 IFF_MACSEC = 1<<27,
7aa98047
LR
1366};
1367
1368#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1369#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1370#define IFF_BONDING IFF_BONDING
7aa98047 1371#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1372#define IFF_WAN_HDLC IFF_WAN_HDLC
1373#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1374#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1375#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1376#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1377#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1378#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1379#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1380#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1381#define IFF_TEAM_PORT IFF_TEAM_PORT
1382#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1383#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1384#define IFF_MACVLAN IFF_MACVLAN
02875878 1385#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1386#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1387#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1388#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1389#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1390#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1391#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1392#define IFF_TEAM IFF_TEAM
d4ab4286 1393#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1394#define IFF_MACSEC IFF_MACSEC
7aa98047 1395
536721b1
KK
1396/**
1397 * struct net_device - The DEVICE structure.
1398 * Actually, this whole structure is a big mistake. It mixes I/O
1399 * data with strictly "high-level" data, and it has to know about
1400 * almost every data structure used in the INET module.
1401 *
1402 * @name: This is the first field of the "visible" part of this structure
1403 * (i.e. as seen by users in the "Space.c" file). It is the name
1404 * of the interface.
1405 *
1406 * @name_hlist: Device name hash chain, please keep it close to name[]
1407 * @ifalias: SNMP alias
1408 * @mem_end: Shared memory end
1409 * @mem_start: Shared memory start
1410 * @base_addr: Device I/O address
1411 * @irq: Device IRQ number
1412 *
14ffbbb8
TG
1413 * @carrier_changes: Stats to monitor carrier on<->off transitions
1414 *
536721b1
KK
1415 * @state: Generic network queuing layer state, see netdev_state_t
1416 * @dev_list: The global list of network devices
5e82b4b2
BH
1417 * @napi_list: List entry used for polling NAPI devices
1418 * @unreg_list: List entry when we are unregistering the
1419 * device; see the function unregister_netdev
1420 * @close_list: List entry used when we are closing the device
62d885fe
BP
1421 * @ptype_all: Device-specific packet handlers for all protocols
1422 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1423 *
1424 * @adj_list: Directly linked devices, like slaves for bonding
1425 * @all_adj_list: All linked devices, *including* neighbours
1426 * @features: Currently active device features
1427 * @hw_features: User-changeable features
1428 *
1429 * @wanted_features: User-requested features
1430 * @vlan_features: Mask of features inheritable by VLAN devices
1431 *
1432 * @hw_enc_features: Mask of features inherited by encapsulating devices
1433 * This field indicates what encapsulation
1434 * offloads the hardware is capable of doing,
1435 * and drivers will need to set them appropriately.
1436 *
1437 * @mpls_features: Mask of features inheritable by MPLS
1438 *
1439 * @ifindex: interface index
5e82b4b2 1440 * @group: The group the device belongs to
536721b1
KK
1441 *
1442 * @stats: Statistics struct, which was left as a legacy, use
1443 * rtnl_link_stats64 instead
1444 *
1445 * @rx_dropped: Dropped packets by core network,
1446 * do not use this in drivers
1447 * @tx_dropped: Dropped packets by core network,
1448 * do not use this in drivers
6e7333d3
JW
1449 * @rx_nohandler: nohandler dropped packets by core network on
1450 * inactive devices, do not use this in drivers
536721b1 1451 *
536721b1
KK
1452 * @wireless_handlers: List of functions to handle Wireless Extensions,
1453 * instead of ioctl,
1454 * see <net/iw_handler.h> for details.
1455 * @wireless_data: Instance data managed by the core of wireless extensions
1456 *
1457 * @netdev_ops: Includes several pointers to callbacks,
1458 * if one wants to override the ndo_*() functions
1459 * @ethtool_ops: Management operations
d476059e 1460 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1461 * of Layer 2 headers.
1462 *
1463 * @flags: Interface flags (a la BSD)
1464 * @priv_flags: Like 'flags' but invisible to userspace,
1465 * see if.h for the definitions
1466 * @gflags: Global flags ( kept as legacy )
1467 * @padded: How much padding added by alloc_netdev()
1468 * @operstate: RFC2863 operstate
1469 * @link_mode: Mapping policy to operstate
1470 * @if_port: Selectable AUI, TP, ...
1471 * @dma: DMA channel
1472 * @mtu: Interface MTU value
1473 * @type: Interface hardware type
2793a23a 1474 * @hard_header_len: Maximum hardware header length.
536721b1
KK
1475 *
1476 * @needed_headroom: Extra headroom the hardware may need, but not in all
1477 * cases can this be guaranteed
1478 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1479 * cases can this be guaranteed. Some cases also use
1480 * LL_MAX_HEADER instead to allocate the skb
1481 *
1482 * interface address info:
1483 *
1484 * @perm_addr: Permanent hw address
1485 * @addr_assign_type: Hw address assignment type
1486 * @addr_len: Hardware address length
1487 * @neigh_priv_len; Used in neigh_alloc(),
1488 * initialized only in atm/clip.c
1489 * @dev_id: Used to differentiate devices that share
1490 * the same link layer address
1491 * @dev_port: Used to differentiate devices that share
1492 * the same function
1493 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1494 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1495 * has been enabled due to the need to listen to
1496 * additional unicast addresses in a device that
1497 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1498 * @uc: unicast mac addresses
1499 * @mc: multicast mac addresses
1500 * @dev_addrs: list of device hw addresses
1501 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1502 * @promiscuity: Number of times the NIC is told to work in
1503 * promiscuous mode; if it becomes 0 the NIC will
1504 * exit promiscuous mode
536721b1
KK
1505 * @allmulti: Counter, enables or disables allmulticast mode
1506 *
1507 * @vlan_info: VLAN info
1508 * @dsa_ptr: dsa specific data
1509 * @tipc_ptr: TIPC specific data
1510 * @atalk_ptr: AppleTalk link
1511 * @ip_ptr: IPv4 specific data
1512 * @dn_ptr: DECnet specific data
1513 * @ip6_ptr: IPv6 specific data
1514 * @ax25_ptr: AX.25 specific data
1515 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1516 *
1517 * @last_rx: Time of last Rx
1518 * @dev_addr: Hw address (before bcast,
1519 * because most packets are unicast)
1520 *
1521 * @_rx: Array of RX queues
1522 * @num_rx_queues: Number of RX queues
1523 * allocated at register_netdev() time
1524 * @real_num_rx_queues: Number of RX queues currently active in device
1525 *
1526 * @rx_handler: handler for received packets
1527 * @rx_handler_data: XXX: need comments on this one
1528 * @ingress_queue: XXX: need comments on this one
1529 * @broadcast: hw bcast address
1530 *
14ffbbb8
TG
1531 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1532 * indexed by RX queue number. Assigned by driver.
1533 * This must only be set if the ndo_rx_flow_steer
1534 * operation is defined
1535 * @index_hlist: Device index hash chain
1536 *
536721b1
KK
1537 * @_tx: Array of TX queues
1538 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1539 * @real_num_tx_queues: Number of TX queues currently active in device
1540 * @qdisc: Root qdisc from userspace point of view
1541 * @tx_queue_len: Max frames per queue allowed
1542 * @tx_global_lock: XXX: need comments on this one
1543 *
1544 * @xps_maps: XXX: need comments on this one
1545 *
0c4f691f
SF
1546 * @offload_fwd_mark: Offload device fwding mark
1547 *
536721b1 1548 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1549 * the watchdog (see dev_watchdog())
536721b1
KK
1550 * @watchdog_timer: List of timers
1551 *
1552 * @pcpu_refcnt: Number of references to this device
1553 * @todo_list: Delayed register/unregister
536721b1
KK
1554 * @link_watch_list: XXX: need comments on this one
1555 *
1556 * @reg_state: Register/unregister state machine
1557 * @dismantle: Device is going to be freed
1558 * @rtnl_link_state: This enum represents the phases of creating
1559 * a new link
1560 *
1561 * @destructor: Called from unregister,
1562 * can be used to call free_netdev
1563 * @npinfo: XXX: need comments on this one
1564 * @nd_net: Network namespace this network device is inside
1565 *
1566 * @ml_priv: Mid-layer private
1567 * @lstats: Loopback statistics
1568 * @tstats: Tunnel statistics
1569 * @dstats: Dummy statistics
1570 * @vstats: Virtual ethernet statistics
1571 *
1572 * @garp_port: GARP
1573 * @mrp_port: MRP
1574 *
1575 * @dev: Class/net/name entry
1576 * @sysfs_groups: Space for optional device, statistics and wireless
1577 * sysfs groups
1578 *
1579 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1580 * @rtnl_link_ops: Rtnl_link_ops
1581 *
1582 * @gso_max_size: Maximum size of generic segmentation offload
1583 * @gso_max_segs: Maximum number of segments that can be passed to the
1584 * NIC for GSO
1585 *
1586 * @dcbnl_ops: Data Center Bridging netlink ops
1587 * @num_tc: Number of traffic classes in the net device
1588 * @tc_to_txq: XXX: need comments on this one
1589 * @prio_tc_map XXX: need comments on this one
1590 *
1591 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1592 *
1593 * @priomap: XXX: need comments on this one
1594 * @phydev: Physical device may attach itself
1595 * for hardware timestamping
1596 *
123b3652
ED
1597 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1598 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1599 *
d746d707
AK
1600 * @proto_down: protocol port state information can be sent to the
1601 * switch driver and used to set the phys state of the
1602 * switch port.
1603 *
1da177e4
LT
1604 * FIXME: cleanup struct net_device such that network protocol info
1605 * moves out.
1606 */
1607
d94d9fee 1608struct net_device {
1da177e4 1609 char name[IFNAMSIZ];
9356b8fc 1610 struct hlist_node name_hlist;
0b815a1a 1611 char *ifalias;
1da177e4
LT
1612 /*
1613 * I/O specific fields
1614 * FIXME: Merge these and struct ifmap into one
1615 */
536721b1
KK
1616 unsigned long mem_end;
1617 unsigned long mem_start;
1618 unsigned long base_addr;
1619 int irq;
1da177e4 1620
14ffbbb8
TG
1621 atomic_t carrier_changes;
1622
1da177e4 1623 /*
536721b1
KK
1624 * Some hardware also needs these fields (state,dev_list,
1625 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1626 * part of the usual set specified in Space.c.
1627 */
1628
1da177e4
LT
1629 unsigned long state;
1630
7562f876 1631 struct list_head dev_list;
bea3348e 1632 struct list_head napi_list;
44a0873d 1633 struct list_head unreg_list;
5cde2829 1634 struct list_head close_list;
7866a621
SN
1635 struct list_head ptype_all;
1636 struct list_head ptype_specific;
2f268f12 1637
2f268f12
VF
1638 struct {
1639 struct list_head upper;
1640 struct list_head lower;
1641 } adj_list;
1642
2f268f12
VF
1643 struct {
1644 struct list_head upper;
1645 struct list_head lower;
1646 } all_adj_list;
4c3d5e7b 1647
c8f44aff 1648 netdev_features_t features;
c8f44aff 1649 netdev_features_t hw_features;
c8f44aff 1650 netdev_features_t wanted_features;
c8f44aff 1651 netdev_features_t vlan_features;
6a674e9c 1652 netdev_features_t hw_enc_features;
0d89d203 1653 netdev_features_t mpls_features;
802ab55a 1654 netdev_features_t gso_partial_features;
04ed3e74 1655
1da177e4 1656 int ifindex;
7a66bbc9 1657 int group;
1da177e4 1658
c45d286e 1659 struct net_device_stats stats;
015f0688 1660
015f0688
ED
1661 atomic_long_t rx_dropped;
1662 atomic_long_t tx_dropped;
6e7333d3 1663 atomic_long_t rx_nohandler;
1da177e4 1664
b86e0280 1665#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1666 const struct iw_handler_def *wireless_handlers;
1667 struct iw_public_data *wireless_data;
b86e0280 1668#endif
d314774c 1669 const struct net_device_ops *netdev_ops;
76fd8593 1670 const struct ethtool_ops *ethtool_ops;
4170604f 1671#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1672 const struct switchdev_ops *switchdev_ops;
4170604f 1673#endif
1b69c6d0
DA
1674#ifdef CONFIG_NET_L3_MASTER_DEV
1675 const struct l3mdev_ops *l3mdev_ops;
1676#endif
1da177e4 1677
3b04ddde
SH
1678 const struct header_ops *header_ops;
1679
536721b1
KK
1680 unsigned int flags;
1681 unsigned int priv_flags;
1682
1da177e4 1683 unsigned short gflags;
536721b1 1684 unsigned short padded;
1da177e4 1685
536721b1
KK
1686 unsigned char operstate;
1687 unsigned char link_mode;
b00055aa 1688
536721b1
KK
1689 unsigned char if_port;
1690 unsigned char dma;
bdc220da 1691
536721b1
KK
1692 unsigned int mtu;
1693 unsigned short type;
1694 unsigned short hard_header_len;
1da177e4 1695
f5184d26
JB
1696 unsigned short needed_headroom;
1697 unsigned short needed_tailroom;
1698
1da177e4 1699 /* Interface address info. */
536721b1
KK
1700 unsigned char perm_addr[MAX_ADDR_LEN];
1701 unsigned char addr_assign_type;
1702 unsigned char addr_len;
a0a9663d 1703 unsigned short neigh_priv_len;
536721b1
KK
1704 unsigned short dev_id;
1705 unsigned short dev_port;
ccffad25 1706 spinlock_t addr_list_lock;
14ffbbb8
TG
1707 unsigned char name_assign_type;
1708 bool uc_promisc;
536721b1
KK
1709 struct netdev_hw_addr_list uc;
1710 struct netdev_hw_addr_list mc;
1711 struct netdev_hw_addr_list dev_addrs;
1712
4c3d5e7b
ED
1713#ifdef CONFIG_SYSFS
1714 struct kset *queues_kset;
1715#endif
9d45abe1
WC
1716 unsigned int promiscuity;
1717 unsigned int allmulti;
1da177e4 1718
1da177e4 1719
5e82b4b2 1720 /* Protocol-specific pointers */
65ac6a5f 1721
d11ead75 1722#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1723 struct vlan_info __rcu *vlan_info;
65ac6a5f 1724#endif
34a430d7 1725#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1726 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1727#endif
1728#if IS_ENABLED(CONFIG_TIPC)
536721b1 1729 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1730#endif
536721b1
KK
1731 void *atalk_ptr;
1732 struct in_device __rcu *ip_ptr;
1733 struct dn_dev __rcu *dn_ptr;
1734 struct inet6_dev __rcu *ip6_ptr;
1735 void *ax25_ptr;
1736 struct wireless_dev *ieee80211_ptr;
98a18b6f 1737 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1738#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1739 struct mpls_dev __rcu *mpls_ptr;
1740#endif
1da177e4 1741
9356b8fc 1742/*
cd13539b 1743 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1744 */
536721b1 1745 unsigned long last_rx;
4dc89133 1746
9356b8fc 1747 /* Interface address info used in eth_type_trans() */
536721b1 1748 unsigned char *dev_addr;
f001fde5 1749
a953be53 1750#ifdef CONFIG_SYSFS
0a9627f2
TH
1751 struct netdev_rx_queue *_rx;
1752
0a9627f2 1753 unsigned int num_rx_queues;
62fe0b40 1754 unsigned int real_num_rx_queues;
df334545 1755#endif
0a9627f2 1756
3b47d303 1757 unsigned long gro_flush_timeout;
61391cde 1758 rx_handler_func_t __rcu *rx_handler;
1759 void __rcu *rx_handler_data;
e8a0464c 1760
4cda01e8 1761#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1762 struct tcf_proto __rcu *ingress_cl_list;
1763#endif
24824a09 1764 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1765#ifdef CONFIG_NETFILTER_INGRESS
1766 struct list_head nf_hooks_ingress;
1767#endif
d2788d34 1768
536721b1 1769 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1770#ifdef CONFIG_RFS_ACCEL
1771 struct cpu_rmap *rx_cpu_rmap;
1772#endif
1773 struct hlist_node index_hlist;
cd13539b
ED
1774
1775/*
1776 * Cache lines mostly used on transmit path
1777 */
e8a0464c
DM
1778 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1779 unsigned int num_tx_queues;
fd2ea0a7 1780 unsigned int real_num_tx_queues;
af356afa 1781 struct Qdisc *qdisc;
536721b1 1782 unsigned long tx_queue_len;
c3f26a26 1783 spinlock_t tx_global_lock;
14ffbbb8 1784 int watchdog_timeo;
cd13539b 1785
bf264145 1786#ifdef CONFIG_XPS
a4177869 1787 struct xps_dev_maps __rcu *xps_maps;
bf264145 1788#endif
1f211a1b
DB
1789#ifdef CONFIG_NET_CLS_ACT
1790 struct tcf_proto __rcu *egress_cl_list;
1791#endif
0c4f691f
SF
1792#ifdef CONFIG_NET_SWITCHDEV
1793 u32 offload_fwd_mark;
1794#endif
1795
9356b8fc 1796 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1797 struct timer_list watchdog_timer;
1798
29b4433d 1799 int __percpu *pcpu_refcnt;
1da177e4 1800 struct list_head todo_list;
1da177e4 1801
e014debe 1802 struct list_head link_watch_list;
572a103d 1803
1da177e4 1804 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1805 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1806 NETREG_UNREGISTERING, /* called unregister_netdevice */
1807 NETREG_UNREGISTERED, /* completed unregister todo */
1808 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1809 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1810 } reg_state:8;
1811
536721b1 1812 bool dismantle;
a2835763
PM
1813
1814 enum {
1815 RTNL_LINK_INITIALIZED,
1816 RTNL_LINK_INITIALIZING,
1817 } rtnl_link_state:16;
1da177e4 1818
d314774c 1819 void (*destructor)(struct net_device *dev);
1da177e4 1820
1da177e4 1821#ifdef CONFIG_NETPOLL
5fbee843 1822 struct netpoll_info __rcu *npinfo;
1da177e4 1823#endif
eae792b7 1824
0c5c9fb5 1825 possible_net_t nd_net;
4a1c5371 1826
4951704b 1827 /* mid-layer private */
a7855c78 1828 union {
536721b1
KK
1829 void *ml_priv;
1830 struct pcpu_lstats __percpu *lstats;
8f84985f 1831 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1832 struct pcpu_dstats __percpu *dstats;
1833 struct pcpu_vstats __percpu *vstats;
a7855c78 1834 };
536721b1 1835
3cc77ec7 1836 struct garp_port __rcu *garp_port;
febf018d 1837 struct mrp_port __rcu *mrp_port;
1da177e4 1838
5e82b4b2 1839 struct device dev;
0c509a6c 1840 const struct attribute_group *sysfs_groups[4];
a953be53 1841 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1842
38f7b870 1843 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1844
82cc1a7a
PWJ
1845 /* for setting kernel sock attribute on TCP connection setup */
1846#define GSO_MAX_SIZE 65536
1847 unsigned int gso_max_size;
30b678d8
BH
1848#define GSO_MAX_SEGS 65535
1849 u16 gso_max_segs;
743b03a8 1850
7a6b6f51 1851#ifdef CONFIG_DCB
32953543 1852 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1853#endif
5e82b4b2
BH
1854 u8 num_tc;
1855 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1856 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1857
d11ead75 1858#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1859 unsigned int fcoe_ddp_xid;
5bc1421e 1860#endif
86f8515f 1861#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1862 struct netprio_map __rcu *priomap;
4d288d57 1863#endif
5e82b4b2
BH
1864 struct phy_device *phydev;
1865 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1866 struct lock_class_key *qdisc_running_key;
5e82b4b2 1867 bool proto_down;
1da177e4 1868};
43cb76d9 1869#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1870
1871#define NETDEV_ALIGN 32
1da177e4 1872
4f57c087
JF
1873static inline
1874int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1875{
1876 return dev->prio_tc_map[prio & TC_BITMASK];
1877}
1878
1879static inline
1880int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1881{
1882 if (tc >= dev->num_tc)
1883 return -EINVAL;
1884
1885 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1886 return 0;
1887}
1888
1889static inline
1890void netdev_reset_tc(struct net_device *dev)
1891{
1892 dev->num_tc = 0;
1893 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1894 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1895}
1896
1897static inline
1898int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1899{
1900 if (tc >= dev->num_tc)
1901 return -EINVAL;
1902
1903 dev->tc_to_txq[tc].count = count;
1904 dev->tc_to_txq[tc].offset = offset;
1905 return 0;
1906}
1907
1908static inline
1909int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1910{
1911 if (num_tc > TC_MAX_QUEUE)
1912 return -EINVAL;
1913
1914 dev->num_tc = num_tc;
1915 return 0;
1916}
1917
1918static inline
1919int netdev_get_num_tc(struct net_device *dev)
1920{
1921 return dev->num_tc;
1922}
1923
e8a0464c
DM
1924static inline
1925struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1926 unsigned int index)
1927{
1928 return &dev->_tx[index];
1929}
1930
10c51b56
DB
1931static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1932 const struct sk_buff *skb)
1933{
1934 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1935}
1936
e8a0464c
DM
1937static inline void netdev_for_each_tx_queue(struct net_device *dev,
1938 void (*f)(struct net_device *,
1939 struct netdev_queue *,
1940 void *),
1941 void *arg)
1942{
1943 unsigned int i;
1944
1945 for (i = 0; i < dev->num_tx_queues; i++)
1946 f(dev, &dev->_tx[i], arg);
1947}
1948
f629d208 1949struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1950 struct sk_buff *skb,
1951 void *accel_priv);
8c4c49df 1952
871b642a
PA
1953/* returns the headroom that the master device needs to take in account
1954 * when forwarding to this dev
1955 */
1956static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1957{
1958 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1959}
1960
1961static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1962{
1963 if (dev->netdev_ops->ndo_set_rx_headroom)
1964 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1965}
1966
1967/* set the device rx headroom to the dev's default */
1968static inline void netdev_reset_rx_headroom(struct net_device *dev)
1969{
1970 netdev_set_rx_headroom(dev, -1);
1971}
1972
c346dca1
YH
1973/*
1974 * Net namespace inlines
1975 */
1976static inline
1977struct net *dev_net(const struct net_device *dev)
1978{
c2d9ba9b 1979 return read_pnet(&dev->nd_net);
c346dca1
YH
1980}
1981
1982static inline
f5aa23fd 1983void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 1984{
0c5c9fb5 1985 write_pnet(&dev->nd_net, net);
c346dca1
YH
1986}
1987
3e8a72d1 1988static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 1989{
3fc88677 1990#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
1991 if (dev->dsa_ptr != NULL)
1992 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 1993#endif
5aed85ce 1994 return false;
396138f0
LB
1995}
1996
bea3348e
SH
1997/**
1998 * netdev_priv - access network device private data
1999 * @dev: network device
2000 *
2001 * Get network device private data
2002 */
6472ce60 2003static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2004{
1ce8e7b5 2005 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2006}
2007
1da177e4
LT
2008/* Set the sysfs physical device reference for the network logical device
2009 * if set prior to registration will cause a symlink during initialization.
2010 */
43cb76d9 2011#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2012
384912ed 2013/* Set the sysfs device type for the network logical device to allow
3f79410c 2014 * fine-grained identification of different network device types. For
5e82b4b2 2015 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2016 */
2017#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2018
82dc3c63
ED
2019/* Default NAPI poll() weight
2020 * Device drivers are strongly advised to not use bigger value
2021 */
2022#define NAPI_POLL_WEIGHT 64
2023
3b582cc1 2024/**
5e82b4b2 2025 * netif_napi_add - initialize a NAPI context
3b582cc1 2026 * @dev: network device
5e82b4b2 2027 * @napi: NAPI context
3b582cc1
SH
2028 * @poll: polling function
2029 * @weight: default weight
2030 *
5e82b4b2
BH
2031 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2032 * *any* of the other NAPI-related functions.
3b582cc1 2033 */
d565b0a1
HX
2034void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2035 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2036
d64b5e85 2037/**
5e82b4b2 2038 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2039 * @dev: network device
5e82b4b2 2040 * @napi: NAPI context
d64b5e85
ED
2041 * @poll: polling function
2042 * @weight: default weight
2043 *
2044 * This variant of netif_napi_add() should be used from drivers using NAPI
2045 * to exclusively poll a TX queue.
2046 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2047 */
2048static inline void netif_tx_napi_add(struct net_device *dev,
2049 struct napi_struct *napi,
2050 int (*poll)(struct napi_struct *, int),
2051 int weight)
2052{
2053 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2054 netif_napi_add(dev, napi, poll, weight);
2055}
2056
d8156534 2057/**
5e82b4b2
BH
2058 * netif_napi_del - remove a NAPI context
2059 * @napi: NAPI context
d8156534 2060 *
5e82b4b2 2061 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2062 */
d565b0a1
HX
2063void netif_napi_del(struct napi_struct *napi);
2064
2065struct napi_gro_cb {
78a478d0 2066 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2067 void *frag0;
78a478d0 2068
7489594c
HX
2069 /* Length of frag0. */
2070 unsigned int frag0_len;
2071
86911732 2072 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2073 int data_offset;
86911732 2074
d565b0a1 2075 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2076 u16 flush;
2077
2078 /* Save the IP ID here and check when we get to the transport layer */
2079 u16 flush_id;
d565b0a1
HX
2080
2081 /* Number of segments aggregated. */
2e71a6f8
ED
2082 u16 count;
2083
15e2396d
TH
2084 /* Start offset for remote checksum offload */
2085 u16 gro_remcsum_start;
2086
2e71a6f8
ED
2087 /* jiffies when first packet was created/queued */
2088 unsigned long age;
86347245 2089
afe93325 2090 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2091 u16 proto;
2092
baa32ff4
TH
2093 /* This is non-zero if the packet may be of the same flow. */
2094 u8 same_flow:1;
2095
fac8e0f5
JG
2096 /* Used in tunnel GRO receive */
2097 u8 encap_mark:1;
573e8fca
TH
2098
2099 /* GRO checksum is valid */
2100 u8 csum_valid:1;
2101
662880f4
TH
2102 /* Number of checksums via CHECKSUM_UNNECESSARY */
2103 u8 csum_cnt:3;
c3c7c254 2104
baa32ff4
TH
2105 /* Free the skb? */
2106 u8 free:2;
2107#define NAPI_GRO_FREE 1
2108#define NAPI_GRO_FREE_STOLEN_HEAD 2
2109
efc98d08
TH
2110 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2111 u8 is_ipv6:1;
2112
a0ca153f
AD
2113 /* Used in GRE, set in fou/gue_gro_receive */
2114 u8 is_fou:1;
2115
1530545e
AD
2116 /* Used to determine if flush_id can be ignored */
2117 u8 is_atomic:1;
2118
2119 /* 5 bit hole */
baa32ff4 2120
bf5a755f
JC
2121 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2122 __wsum csum;
2123
c3c7c254
ED
2124 /* used in skb_gro_receive() slow path */
2125 struct sk_buff *last;
d565b0a1
HX
2126};
2127
2128#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2129
1da177e4 2130struct packet_type {
f2ccd8fa
DM
2131 __be16 type; /* This is really htons(ether_type). */
2132 struct net_device *dev; /* NULL is wildcarded here */
2133 int (*func) (struct sk_buff *,
2134 struct net_device *,
2135 struct packet_type *,
2136 struct net_device *);
c0de08d0
EL
2137 bool (*id_match)(struct packet_type *ptype,
2138 struct sock *sk);
1da177e4
LT
2139 void *af_packet_priv;
2140 struct list_head list;
2141};
2142
f191a1d1 2143struct offload_callbacks {
576a30eb 2144 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2145 netdev_features_t features);
d565b0a1 2146 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2147 struct sk_buff *skb);
299603e8 2148 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2149};
2150
2151struct packet_offload {
2152 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2153 u16 priority;
f191a1d1
VY
2154 struct offload_callbacks callbacks;
2155 struct list_head list;
1da177e4
LT
2156};
2157
5e82b4b2 2158/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2159struct pcpu_sw_netstats {
2160 u64 rx_packets;
2161 u64 rx_bytes;
2162 u64 tx_packets;
2163 u64 tx_bytes;
2164 struct u64_stats_sync syncp;
2165};
2166
aabc92bb
PNA
2167#define __netdev_alloc_pcpu_stats(type, gfp) \
2168({ \
2169 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2170 if (pcpu_stats) { \
2171 int __cpu; \
2172 for_each_possible_cpu(__cpu) { \
2173 typeof(type) *stat; \
2174 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2175 u64_stats_init(&stat->syncp); \
2176 } \
2177 } \
2178 pcpu_stats; \
1c213bd2
WC
2179})
2180
aabc92bb 2181#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2182 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2183
764f5e54
JP
2184enum netdev_lag_tx_type {
2185 NETDEV_LAG_TX_TYPE_UNKNOWN,
2186 NETDEV_LAG_TX_TYPE_RANDOM,
2187 NETDEV_LAG_TX_TYPE_BROADCAST,
2188 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2189 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2190 NETDEV_LAG_TX_TYPE_HASH,
2191};
2192
2193struct netdev_lag_upper_info {
2194 enum netdev_lag_tx_type tx_type;
2195};
2196
fb1b2e3c
JP
2197struct netdev_lag_lower_state_info {
2198 u8 link_up : 1,
2199 tx_enabled : 1;
2200};
2201
1da177e4
LT
2202#include <linux/notifier.h>
2203
dcfe1421
AW
2204/* netdevice notifier chain. Please remember to update the rtnetlink
2205 * notification exclusion list in rtnetlink_event() when adding new
2206 * types.
2207 */
2208#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2209#define NETDEV_DOWN 0x0002
2210#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2211 detected a hardware crash and restarted
2212 - we can use this eg to kick tcp sessions
2213 once done */
2214#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2215#define NETDEV_REGISTER 0x0005
2216#define NETDEV_UNREGISTER 0x0006
1d486bfb 2217#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2218#define NETDEV_CHANGEADDR 0x0008
2219#define NETDEV_GOING_DOWN 0x0009
2220#define NETDEV_CHANGENAME 0x000A
2221#define NETDEV_FEAT_CHANGE 0x000B
2222#define NETDEV_BONDING_FAILOVER 0x000C
2223#define NETDEV_PRE_UP 0x000D
2224#define NETDEV_PRE_TYPE_CHANGE 0x000E
2225#define NETDEV_POST_TYPE_CHANGE 0x000F
2226#define NETDEV_POST_INIT 0x0010
0115e8e3 2227#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2228#define NETDEV_RELEASE 0x0012
2229#define NETDEV_NOTIFY_PEERS 0x0013
2230#define NETDEV_JOIN 0x0014
42e52bf9 2231#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2232#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2233#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2234#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2235#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2236#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2237#define NETDEV_CHANGELOWERSTATE 0x001B
b7aade15 2238#define NETDEV_OFFLOAD_PUSH_VXLAN 0x001C
681e683f 2239#define NETDEV_OFFLOAD_PUSH_GENEVE 0x001D
dcfe1421 2240
f629d208
JP
2241int register_netdevice_notifier(struct notifier_block *nb);
2242int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2243
2244struct netdev_notifier_info {
2245 struct net_device *dev;
2246};
2247
be9efd36
JP
2248struct netdev_notifier_change_info {
2249 struct netdev_notifier_info info; /* must be first */
2250 unsigned int flags_changed;
2251};
2252
0e4ead9d
JP
2253struct netdev_notifier_changeupper_info {
2254 struct netdev_notifier_info info; /* must be first */
2255 struct net_device *upper_dev; /* new upper dev */
2256 bool master; /* is upper dev master */
5e82b4b2 2257 bool linking; /* is the notification for link or unlink */
29bf24af 2258 void *upper_info; /* upper dev info */
0e4ead9d
JP
2259};
2260
04d48266
JP
2261struct netdev_notifier_changelowerstate_info {
2262 struct netdev_notifier_info info; /* must be first */
2263 void *lower_state_info; /* is lower dev state */
2264};
2265
75538c2b
CW
2266static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2267 struct net_device *dev)
2268{
2269 info->dev = dev;
2270}
2271
351638e7
JP
2272static inline struct net_device *
2273netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2274{
2275 return info->dev;
2276}
2277
f629d208 2278int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2279
2280
1da177e4
LT
2281extern rwlock_t dev_base_lock; /* Device list lock */
2282
881d966b
EB
2283#define for_each_netdev(net, d) \
2284 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2285#define for_each_netdev_reverse(net, d) \
2286 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2287#define for_each_netdev_rcu(net, d) \
2288 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2289#define for_each_netdev_safe(net, d, n) \
2290 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2291#define for_each_netdev_continue(net, d) \
2292 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2293#define for_each_netdev_continue_rcu(net, d) \
2294 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2295#define for_each_netdev_in_bond_rcu(bond, slave) \
2296 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2297 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2298#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2299
a050c33f
DL
2300static inline struct net_device *next_net_device(struct net_device *dev)
2301{
2302 struct list_head *lh;
2303 struct net *net;
2304
c346dca1 2305 net = dev_net(dev);
a050c33f
DL
2306 lh = dev->dev_list.next;
2307 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2308}
2309
ce81b76a
ED
2310static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2311{
2312 struct list_head *lh;
2313 struct net *net;
2314
2315 net = dev_net(dev);
ccf43438 2316 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2317 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2318}
2319
a050c33f
DL
2320static inline struct net_device *first_net_device(struct net *net)
2321{
2322 return list_empty(&net->dev_base_head) ? NULL :
2323 net_device_entry(net->dev_base_head.next);
2324}
7562f876 2325
ccf43438
ED
2326static inline struct net_device *first_net_device_rcu(struct net *net)
2327{
2328 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2329
2330 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2331}
2332
f629d208
JP
2333int netdev_boot_setup_check(struct net_device *dev);
2334unsigned long netdev_boot_base(const char *prefix, int unit);
2335struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2336 const char *hwaddr);
2337struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2338struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2339void dev_add_pack(struct packet_type *pt);
2340void dev_remove_pack(struct packet_type *pt);
2341void __dev_remove_pack(struct packet_type *pt);
2342void dev_add_offload(struct packet_offload *po);
2343void dev_remove_offload(struct packet_offload *po);
f629d208 2344
a54acb3a 2345int dev_get_iflink(const struct net_device *dev);
fc4099f1 2346int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2347struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2348 unsigned short mask);
f629d208
JP
2349struct net_device *dev_get_by_name(struct net *net, const char *name);
2350struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2351struct net_device *__dev_get_by_name(struct net *net, const char *name);
2352int dev_alloc_name(struct net_device *dev, const char *name);
2353int dev_open(struct net_device *dev);
2354int dev_close(struct net_device *dev);
99c4a26a 2355int dev_close_many(struct list_head *head, bool unlink);
f629d208 2356void dev_disable_lro(struct net_device *dev);
0c4b51f0 2357int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2358int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2359int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2360int register_netdevice(struct net_device *dev);
2361void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2362void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2363static inline void unregister_netdevice(struct net_device *dev)
2364{
2365 unregister_netdevice_queue(dev, NULL);
2366}
2367
f629d208
JP
2368int netdev_refcnt_read(const struct net_device *dev);
2369void free_netdev(struct net_device *dev);
74d332c1 2370void netdev_freemem(struct net_device *dev);
f629d208
JP
2371void synchronize_net(void);
2372int init_dummy_netdev(struct net_device *dev);
937f1ba5 2373
f60e5990 2374DECLARE_PER_CPU(int, xmit_recursion);
2375static inline int dev_recursion_level(void)
2376{
2377 return this_cpu_read(xmit_recursion);
2378}
2379
f629d208
JP
2380struct net_device *dev_get_by_index(struct net *net, int ifindex);
2381struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2382struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2383int netdev_get_name(struct net *net, char *name, int ifindex);
2384int dev_restart(struct net_device *dev);
f629d208 2385int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2386
2387static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2388{
2389 return NAPI_GRO_CB(skb)->data_offset;
2390}
2391
2392static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2393{
2394 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2395}
2396
2397static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2398{
2399 NAPI_GRO_CB(skb)->data_offset += len;
2400}
2401
a5b1cf28
HX
2402static inline void *skb_gro_header_fast(struct sk_buff *skb,
2403 unsigned int offset)
86911732 2404{
a5b1cf28
HX
2405 return NAPI_GRO_CB(skb)->frag0 + offset;
2406}
78a478d0 2407
a5b1cf28
HX
2408static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2409{
2410 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2411}
78a478d0 2412
a5b1cf28
HX
2413static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2414 unsigned int offset)
2415{
17dd759c
HX
2416 if (!pskb_may_pull(skb, hlen))
2417 return NULL;
2418
a5b1cf28
HX
2419 NAPI_GRO_CB(skb)->frag0 = NULL;
2420 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2421 return skb->data + offset;
86911732 2422}
1da177e4 2423
36e7b1b8
HX
2424static inline void *skb_gro_network_header(struct sk_buff *skb)
2425{
78d3fd0b
HX
2426 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2427 skb_network_offset(skb);
36e7b1b8
HX
2428}
2429
bf5a755f
JC
2430static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2431 const void *start, unsigned int len)
2432{
573e8fca 2433 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2434 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2435 csum_partial(start, len, 0));
2436}
2437
573e8fca
TH
2438/* GRO checksum functions. These are logical equivalents of the normal
2439 * checksum functions (in skbuff.h) except that they operate on the GRO
2440 * offsets and fields in sk_buff.
2441 */
2442
2443__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2444
15e2396d
TH
2445static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2446{
b7fe10e5 2447 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2448}
2449
573e8fca
TH
2450static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2451 bool zero_okay,
2452 __sum16 check)
2453{
6edec0e6
TH
2454 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2455 skb_checksum_start_offset(skb) <
2456 skb_gro_offset(skb)) &&
15e2396d 2457 !skb_at_gro_remcsum_start(skb) &&
662880f4 2458 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2459 (!zero_okay || check));
2460}
2461
2462static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2463 __wsum psum)
2464{
2465 if (NAPI_GRO_CB(skb)->csum_valid &&
2466 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2467 return 0;
2468
2469 NAPI_GRO_CB(skb)->csum = psum;
2470
2471 return __skb_gro_checksum_complete(skb);
2472}
2473
573e8fca
TH
2474static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2475{
662880f4
TH
2476 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2477 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2478 NAPI_GRO_CB(skb)->csum_cnt--;
2479 } else {
2480 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2481 * verified a new top level checksum or an encapsulated one
2482 * during GRO. This saves work if we fallback to normal path.
2483 */
2484 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2485 }
2486}
2487
2488#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2489 compute_pseudo) \
2490({ \
2491 __sum16 __ret = 0; \
2492 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2493 __ret = __skb_gro_checksum_validate_complete(skb, \
2494 compute_pseudo(skb, proto)); \
5a212329
TH
2495 if (__ret) \
2496 __skb_mark_checksum_bad(skb); \
2497 else \
573e8fca
TH
2498 skb_gro_incr_csum_unnecessary(skb); \
2499 __ret; \
2500})
2501
2502#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2503 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2504
2505#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2506 compute_pseudo) \
2507 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2508
2509#define skb_gro_checksum_simple_validate(skb) \
2510 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2511
d96535a1
TH
2512static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2513{
2514 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2515 !NAPI_GRO_CB(skb)->csum_valid);
2516}
2517
2518static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2519 __sum16 check, __wsum pseudo)
2520{
2521 NAPI_GRO_CB(skb)->csum = ~pseudo;
2522 NAPI_GRO_CB(skb)->csum_valid = 1;
2523}
2524
2525#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2526do { \
2527 if (__skb_gro_checksum_convert_check(skb)) \
2528 __skb_gro_checksum_convert(skb, check, \
2529 compute_pseudo(skb, proto)); \
2530} while (0)
2531
26c4f7da
TH
2532struct gro_remcsum {
2533 int offset;
2534 __wsum delta;
2535};
2536
2537static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2538{
846cd667 2539 grc->offset = 0;
26c4f7da
TH
2540 grc->delta = 0;
2541}
2542
b7fe10e5
TH
2543static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2544 unsigned int off, size_t hdrlen,
2545 int start, int offset,
2546 struct gro_remcsum *grc,
2547 bool nopartial)
dcdc8994
TH
2548{
2549 __wsum delta;
b7fe10e5 2550 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2551
2552 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2553
15e2396d 2554 if (!nopartial) {
b7fe10e5
TH
2555 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2556 return ptr;
2557 }
2558
2559 ptr = skb_gro_header_fast(skb, off);
2560 if (skb_gro_header_hard(skb, off + plen)) {
2561 ptr = skb_gro_header_slow(skb, off + plen, off);
2562 if (!ptr)
2563 return NULL;
15e2396d
TH
2564 }
2565
b7fe10e5
TH
2566 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2567 start, offset);
dcdc8994
TH
2568
2569 /* Adjust skb->csum since we changed the packet */
dcdc8994 2570 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2571
b7fe10e5 2572 grc->offset = off + hdrlen + offset;
26c4f7da 2573 grc->delta = delta;
b7fe10e5
TH
2574
2575 return ptr;
dcdc8994
TH
2576}
2577
26c4f7da
TH
2578static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2579 struct gro_remcsum *grc)
2580{
b7fe10e5
TH
2581 void *ptr;
2582 size_t plen = grc->offset + sizeof(u16);
2583
26c4f7da
TH
2584 if (!grc->delta)
2585 return;
2586
b7fe10e5
TH
2587 ptr = skb_gro_header_fast(skb, grc->offset);
2588 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2589 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2590 if (!ptr)
2591 return;
2592 }
2593
2594 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2595}
dcdc8994 2596
6ae23ad3
TH
2597struct skb_csum_offl_spec {
2598 __u16 ipv4_okay:1,
2599 ipv6_okay:1,
2600 encap_okay:1,
2601 ip_options_okay:1,
2602 ext_hdrs_okay:1,
2603 tcp_okay:1,
2604 udp_okay:1,
2605 sctp_okay:1,
2606 vlan_okay:1,
2607 no_encapped_ipv6:1,
2608 no_not_encapped:1;
2609};
2610
2611bool __skb_csum_offload_chk(struct sk_buff *skb,
2612 const struct skb_csum_offl_spec *spec,
2613 bool *csum_encapped,
2614 bool csum_help);
2615
2616static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2617 const struct skb_csum_offl_spec *spec,
2618 bool *csum_encapped,
2619 bool csum_help)
2620{
2621 if (skb->ip_summed != CHECKSUM_PARTIAL)
2622 return false;
2623
2624 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2625}
2626
2627static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2628 const struct skb_csum_offl_spec *spec)
2629{
2630 bool csum_encapped;
2631
2632 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2633}
2634
2635static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2636{
2637 static const struct skb_csum_offl_spec csum_offl_spec = {
2638 .ipv4_okay = 1,
2639 .ip_options_okay = 1,
2640 .ipv6_okay = 1,
2641 .vlan_okay = 1,
2642 .tcp_okay = 1,
2643 .udp_okay = 1,
2644 };
2645
2646 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2647}
2648
2649static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2650{
2651 static const struct skb_csum_offl_spec csum_offl_spec = {
2652 .ipv4_okay = 1,
2653 .ip_options_okay = 1,
2654 .tcp_okay = 1,
2655 .udp_okay = 1,
2656 .vlan_okay = 1,
2657 };
2658
2659 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2660}
2661
0c4e8581
SH
2662static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2663 unsigned short type,
3b04ddde 2664 const void *daddr, const void *saddr,
95c96174 2665 unsigned int len)
0c4e8581 2666{
f1ecfd5d 2667 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2668 return 0;
3b04ddde
SH
2669
2670 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2671}
2672
b95cce35
SH
2673static inline int dev_parse_header(const struct sk_buff *skb,
2674 unsigned char *haddr)
2675{
2676 const struct net_device *dev = skb->dev;
2677
1b83336b 2678 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2679 return 0;
3b04ddde 2680 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2681}
2682
2793a23a
WB
2683/* ll_header must have at least hard_header_len allocated */
2684static inline bool dev_validate_header(const struct net_device *dev,
2685 char *ll_header, int len)
2686{
2687 if (likely(len >= dev->hard_header_len))
2688 return true;
2689
2690 if (capable(CAP_SYS_RAWIO)) {
2691 memset(ll_header + len, 0, dev->hard_header_len - len);
2692 return true;
2693 }
2694
2695 if (dev->header_ops && dev->header_ops->validate)
2696 return dev->header_ops->validate(ll_header, len);
2697
2698 return false;
2699}
2700
1da177e4 2701typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2702int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2703static inline int unregister_gifconf(unsigned int family)
2704{
2705 return register_gifconf(family, NULL);
2706}
2707
99bbc707 2708#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2709#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2710struct sd_flow_limit {
2711 u64 count;
2712 unsigned int num_buckets;
2713 unsigned int history_head;
2714 u16 history[FLOW_LIMIT_HISTORY];
2715 u8 buckets[];
2716};
2717
2718extern int netdev_flow_limit_table_len;
2719#endif /* CONFIG_NET_FLOW_LIMIT */
2720
1da177e4 2721/*
5e82b4b2 2722 * Incoming packets are placed on per-CPU queues
1da177e4 2723 */
d94d9fee 2724struct softnet_data {
1da177e4 2725 struct list_head poll_list;
6e7676c1 2726 struct sk_buff_head process_queue;
1da177e4 2727
dee42870 2728 /* stats */
cd7b5396
DM
2729 unsigned int processed;
2730 unsigned int time_squeeze;
cd7b5396 2731 unsigned int received_rps;
fd793d89 2732#ifdef CONFIG_RPS
88751275 2733 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2734#endif
2735#ifdef CONFIG_NET_FLOW_LIMIT
2736 struct sd_flow_limit __rcu *flow_limit;
2737#endif
2738 struct Qdisc *output_queue;
2739 struct Qdisc **output_queue_tailp;
2740 struct sk_buff *completion_queue;
88751275 2741
4cdb1e2e 2742#ifdef CONFIG_RPS
501e7ef5
ED
2743 /* input_queue_head should be written by cpu owning this struct,
2744 * and only read by other cpus. Worth using a cache line.
2745 */
2746 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2747
2748 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2749 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2750 struct softnet_data *rps_ipi_next;
2751 unsigned int cpu;
76cc8b13 2752 unsigned int input_queue_tail;
1e94d72f 2753#endif
95c96174 2754 unsigned int dropped;
0a9627f2 2755 struct sk_buff_head input_pkt_queue;
bea3348e 2756 struct napi_struct backlog;
99bbc707 2757
1da177e4
LT
2758};
2759
76cc8b13 2760static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2761{
2762#ifdef CONFIG_RPS
76cc8b13
TH
2763 sd->input_queue_head++;
2764#endif
2765}
2766
2767static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2768 unsigned int *qtail)
2769{
2770#ifdef CONFIG_RPS
2771 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2772#endif
2773}
2774
0a9627f2 2775DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2776
f629d208 2777void __netif_schedule(struct Qdisc *q);
46e5da40 2778void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2779
fd2ea0a7
DM
2780static inline void netif_tx_schedule_all(struct net_device *dev)
2781{
2782 unsigned int i;
2783
2784 for (i = 0; i < dev->num_tx_queues; i++)
2785 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2786}
2787
f9a7cbbf 2788static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2789{
73466498 2790 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2791}
2792
bea3348e
SH
2793/**
2794 * netif_start_queue - allow transmit
2795 * @dev: network device
2796 *
2797 * Allow upper layers to call the device hard_start_xmit routine.
2798 */
1da177e4
LT
2799static inline void netif_start_queue(struct net_device *dev)
2800{
e8a0464c 2801 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2802}
2803
fd2ea0a7
DM
2804static inline void netif_tx_start_all_queues(struct net_device *dev)
2805{
2806 unsigned int i;
2807
2808 for (i = 0; i < dev->num_tx_queues; i++) {
2809 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2810 netif_tx_start_queue(txq);
2811 }
2812}
2813
46e5da40 2814void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2815
d29f749e
DJ
2816/**
2817 * netif_wake_queue - restart transmit
2818 * @dev: network device
2819 *
2820 * Allow upper layers to call the device hard_start_xmit routine.
2821 * Used for flow control when transmit resources are available.
2822 */
79d16385
DM
2823static inline void netif_wake_queue(struct net_device *dev)
2824{
e8a0464c 2825 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2826}
2827
fd2ea0a7
DM
2828static inline void netif_tx_wake_all_queues(struct net_device *dev)
2829{
2830 unsigned int i;
2831
2832 for (i = 0; i < dev->num_tx_queues; i++) {
2833 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2834 netif_tx_wake_queue(txq);
2835 }
2836}
2837
f9a7cbbf 2838static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2839{
73466498 2840 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2841}
2842
bea3348e
SH
2843/**
2844 * netif_stop_queue - stop transmitted packets
2845 * @dev: network device
2846 *
2847 * Stop upper layers calling the device hard_start_xmit routine.
2848 * Used for flow control when transmit resources are unavailable.
2849 */
1da177e4
LT
2850static inline void netif_stop_queue(struct net_device *dev)
2851{
e8a0464c 2852 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2853}
2854
a2029240 2855void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2856
4d29515f 2857static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2858{
73466498 2859 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2860}
2861
bea3348e
SH
2862/**
2863 * netif_queue_stopped - test if transmit queue is flowblocked
2864 * @dev: network device
2865 *
2866 * Test if transmit queue on device is currently unable to send.
2867 */
4d29515f 2868static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2869{
e8a0464c 2870 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2871}
2872
4d29515f 2873static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2874{
73466498
TH
2875 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2876}
2877
8e2f1a63
DB
2878static inline bool
2879netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2880{
2881 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2882}
2883
8e2f1a63
DB
2884static inline bool
2885netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2886{
2887 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2888}
2889
53511453
ED
2890/**
2891 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2892 * @dev_queue: pointer to transmit queue
2893 *
2894 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2895 * to give appropriate hint to the CPU.
53511453
ED
2896 */
2897static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2898{
2899#ifdef CONFIG_BQL
2900 prefetchw(&dev_queue->dql.num_queued);
2901#endif
2902}
2903
2904/**
2905 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2906 * @dev_queue: pointer to transmit queue
2907 *
2908 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2909 * to give appropriate hint to the CPU.
53511453
ED
2910 */
2911static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2912{
2913#ifdef CONFIG_BQL
2914 prefetchw(&dev_queue->dql.limit);
2915#endif
2916}
2917
c5d67bd7
TH
2918static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2919 unsigned int bytes)
2920{
114cf580
TH
2921#ifdef CONFIG_BQL
2922 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2923
2924 if (likely(dql_avail(&dev_queue->dql) >= 0))
2925 return;
2926
2927 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2928
2929 /*
2930 * The XOFF flag must be set before checking the dql_avail below,
2931 * because in netdev_tx_completed_queue we update the dql_completed
2932 * before checking the XOFF flag.
2933 */
2934 smp_mb();
2935
2936 /* check again in case another CPU has just made room avail */
2937 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2938 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2939#endif
c5d67bd7
TH
2940}
2941
0042d0c8
FF
2942/**
2943 * netdev_sent_queue - report the number of bytes queued to hardware
2944 * @dev: network device
2945 * @bytes: number of bytes queued to the hardware device queue
2946 *
2947 * Report the number of bytes queued for sending/completion to the network
2948 * device hardware queue. @bytes should be a good approximation and should
2949 * exactly match netdev_completed_queue() @bytes
2950 */
c5d67bd7
TH
2951static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2952{
2953 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2954}
2955
2956static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2957 unsigned int pkts, unsigned int bytes)
c5d67bd7 2958{
114cf580 2959#ifdef CONFIG_BQL
b37c0fbe
AD
2960 if (unlikely(!bytes))
2961 return;
2962
2963 dql_completed(&dev_queue->dql, bytes);
2964
2965 /*
2966 * Without the memory barrier there is a small possiblity that
2967 * netdev_tx_sent_queue will miss the update and cause the queue to
2968 * be stopped forever
2969 */
2970 smp_mb();
2971
2972 if (dql_avail(&dev_queue->dql) < 0)
2973 return;
2974
2975 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2976 netif_schedule_queue(dev_queue);
114cf580 2977#endif
c5d67bd7
TH
2978}
2979
0042d0c8
FF
2980/**
2981 * netdev_completed_queue - report bytes and packets completed by device
2982 * @dev: network device
2983 * @pkts: actual number of packets sent over the medium
2984 * @bytes: actual number of bytes sent over the medium
2985 *
2986 * Report the number of bytes and packets transmitted by the network device
2987 * hardware queue over the physical medium, @bytes must exactly match the
2988 * @bytes amount passed to netdev_sent_queue()
2989 */
c5d67bd7 2990static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2991 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2992{
2993 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2994}
2995
2996static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2997{
114cf580 2998#ifdef CONFIG_BQL
5c490354 2999 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3000 dql_reset(&q->dql);
3001#endif
c5d67bd7
TH
3002}
3003
0042d0c8
FF
3004/**
3005 * netdev_reset_queue - reset the packets and bytes count of a network device
3006 * @dev_queue: network device
3007 *
3008 * Reset the bytes and packet count of a network device and clear the
3009 * software flow control OFF bit for this network device
3010 */
c5d67bd7
TH
3011static inline void netdev_reset_queue(struct net_device *dev_queue)
3012{
3013 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3014}
3015
b9507bda
DB
3016/**
3017 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3018 * @dev: network device
3019 * @queue_index: given tx queue index
3020 *
3021 * Returns 0 if given tx queue index >= number of device tx queues,
3022 * otherwise returns the originally passed tx queue index.
3023 */
3024static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3025{
3026 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3027 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3028 dev->name, queue_index,
3029 dev->real_num_tx_queues);
3030 return 0;
3031 }
3032
3033 return queue_index;
3034}
3035
bea3348e
SH
3036/**
3037 * netif_running - test if up
3038 * @dev: network device
3039 *
3040 * Test if the device has been brought up.
3041 */
4d29515f 3042static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3043{
3044 return test_bit(__LINK_STATE_START, &dev->state);
3045}
3046
f25f4e44 3047/*
5e82b4b2 3048 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3049 * stop, and a check if it's stopped. All other device management is
3050 * done at the overall netdevice level.
3051 * Also test the device if we're multiqueue.
3052 */
bea3348e
SH
3053
3054/**
3055 * netif_start_subqueue - allow sending packets on subqueue
3056 * @dev: network device
3057 * @queue_index: sub queue index
3058 *
3059 * Start individual transmit queue of a device with multiple transmit queues.
3060 */
f25f4e44
PWJ
3061static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3062{
fd2ea0a7 3063 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3064
3065 netif_tx_start_queue(txq);
f25f4e44
PWJ
3066}
3067
bea3348e
SH
3068/**
3069 * netif_stop_subqueue - stop sending packets on subqueue
3070 * @dev: network device
3071 * @queue_index: sub queue index
3072 *
3073 * Stop individual transmit queue of a device with multiple transmit queues.
3074 */
f25f4e44
PWJ
3075static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3076{
fd2ea0a7 3077 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3078 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3079}
3080
bea3348e
SH
3081/**
3082 * netif_subqueue_stopped - test status of subqueue
3083 * @dev: network device
3084 * @queue_index: sub queue index
3085 *
3086 * Check individual transmit queue of a device with multiple transmit queues.
3087 */
4d29515f
DM
3088static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3089 u16 queue_index)
f25f4e44 3090{
fd2ea0a7 3091 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3092
3093 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3094}
3095
4d29515f
DM
3096static inline bool netif_subqueue_stopped(const struct net_device *dev,
3097 struct sk_buff *skb)
668f895a
PE
3098{
3099 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3100}
bea3348e 3101
46e5da40 3102void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 3103
537c00de 3104#ifdef CONFIG_XPS
53af53ae 3105int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3106 u16 index);
537c00de
AD
3107#else
3108static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3109 const struct cpumask *mask,
537c00de
AD
3110 u16 index)
3111{
3112 return 0;
3113}
3114#endif
3115
5605c762
JP
3116u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3117 unsigned int num_tx_queues);
3118
a3d22a68
VZ
3119/*
3120 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3121 * as a distribution range limit for the returned value.
3122 */
3123static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3124 struct sk_buff *skb)
a3d22a68
VZ
3125{
3126 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3127}
3128
bea3348e
SH
3129/**
3130 * netif_is_multiqueue - test if device has multiple transmit queues
3131 * @dev: network device
3132 *
3133 * Check if device has multiple transmit queues
bea3348e 3134 */
4d29515f 3135static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3136{
a02cec21 3137 return dev->num_tx_queues > 1;
f25f4e44 3138}
1da177e4 3139
f629d208 3140int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3141
a953be53 3142#ifdef CONFIG_SYSFS
f629d208 3143int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3144#else
3145static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3146 unsigned int rxq)
3147{
3148 return 0;
3149}
3150#endif
3151
a953be53
MD
3152#ifdef CONFIG_SYSFS
3153static inline unsigned int get_netdev_rx_queue_index(
3154 struct netdev_rx_queue *queue)
3155{
3156 struct net_device *dev = queue->dev;
3157 int index = queue - dev->_rx;
3158
3159 BUG_ON(index >= dev->num_rx_queues);
3160 return index;
3161}
3162#endif
3163
16917b87 3164#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3165int netif_get_num_default_rss_queues(void);
16917b87 3166
e6247027
ED
3167enum skb_free_reason {
3168 SKB_REASON_CONSUMED,
3169 SKB_REASON_DROPPED,
3170};
3171
3172void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3173void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3174
e6247027
ED
3175/*
3176 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3177 * interrupt context or with hardware interrupts being disabled.
3178 * (in_irq() || irqs_disabled())
3179 *
3180 * We provide four helpers that can be used in following contexts :
3181 *
3182 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3183 * replacing kfree_skb(skb)
3184 *
3185 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3186 * Typically used in place of consume_skb(skb) in TX completion path
3187 *
3188 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3189 * replacing kfree_skb(skb)
3190 *
3191 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3192 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3193 */
e6247027
ED
3194static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3195{
3196 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3197}
3198
3199static inline void dev_consume_skb_irq(struct sk_buff *skb)
3200{
3201 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3202}
3203
3204static inline void dev_kfree_skb_any(struct sk_buff *skb)
3205{
3206 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3207}
3208
3209static inline void dev_consume_skb_any(struct sk_buff *skb)
3210{
3211 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3212}
1da177e4 3213
f629d208
JP
3214int netif_rx(struct sk_buff *skb);
3215int netif_rx_ni(struct sk_buff *skb);
04eb4489 3216int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3217gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3218void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3219struct sk_buff *napi_get_frags(struct napi_struct *napi);
3220gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3221struct packet_offload *gro_find_receive_by_type(__be16 type);
3222struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3223
3224static inline void napi_free_frags(struct napi_struct *napi)
3225{
3226 kfree_skb(napi->skb);
3227 napi->skb = NULL;
3228}
3229
f629d208
JP
3230int netdev_rx_handler_register(struct net_device *dev,
3231 rx_handler_func_t *rx_handler,
3232 void *rx_handler_data);
3233void netdev_rx_handler_unregister(struct net_device *dev);
3234
3235bool dev_valid_name(const char *name);
3236int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3237int dev_ethtool(struct net *net, struct ifreq *);
3238unsigned int dev_get_flags(const struct net_device *);
3239int __dev_change_flags(struct net_device *, unsigned int flags);
3240int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3241void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3242 unsigned int gchanges);
f629d208
JP
3243int dev_change_name(struct net_device *, const char *);
3244int dev_set_alias(struct net_device *, const char *, size_t);
3245int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3246int dev_set_mtu(struct net_device *, int);
3247void dev_set_group(struct net_device *, int);
3248int dev_set_mac_address(struct net_device *, struct sockaddr *);
3249int dev_change_carrier(struct net_device *, bool new_carrier);
3250int dev_get_phys_port_id(struct net_device *dev,
02637fce 3251 struct netdev_phys_item_id *ppid);
db24a904
DA
3252int dev_get_phys_port_name(struct net_device *dev,
3253 char *name, size_t len);
d746d707 3254int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3255struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3256struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3257 struct netdev_queue *txq, int *ret);
a0265d28 3258int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3259int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3260bool is_skb_forwardable(const struct net_device *dev,
3261 const struct sk_buff *skb);
1da177e4 3262
74b20582
DA
3263void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3264
20380731 3265extern int netdev_budget;
1da177e4
LT
3266
3267/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3268void netdev_run_todo(void);
1da177e4 3269
bea3348e
SH
3270/**
3271 * dev_put - release reference to device
3272 * @dev: network device
3273 *
9ef4429b 3274 * Release reference to device to allow it to be freed.
bea3348e 3275 */
1da177e4
LT
3276static inline void dev_put(struct net_device *dev)
3277{
933393f5 3278 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3279}
3280
bea3348e
SH
3281/**
3282 * dev_hold - get reference to device
3283 * @dev: network device
3284 *
9ef4429b 3285 * Hold reference to device to keep it from being freed.
bea3348e 3286 */
15333061
SH
3287static inline void dev_hold(struct net_device *dev)
3288{
933393f5 3289 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3290}
1da177e4
LT
3291
3292/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3293 * and _off may be called from IRQ context, but it is caller
3294 * who is responsible for serialization of these calls.
b00055aa
SR
3295 *
3296 * The name carrier is inappropriate, these functions should really be
3297 * called netif_lowerlayer_*() because they represent the state of any
3298 * kind of lower layer not just hardware media.
1da177e4
LT
3299 */
3300
f629d208
JP
3301void linkwatch_init_dev(struct net_device *dev);
3302void linkwatch_fire_event(struct net_device *dev);
3303void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3304
bea3348e
SH
3305/**
3306 * netif_carrier_ok - test if carrier present
3307 * @dev: network device
3308 *
3309 * Check if carrier is present on device
3310 */
4d29515f 3311static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3312{
3313 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3314}
3315
f629d208 3316unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3317
f629d208 3318void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3319
f629d208 3320void netif_carrier_on(struct net_device *dev);
1da177e4 3321
f629d208 3322void netif_carrier_off(struct net_device *dev);
1da177e4 3323
bea3348e
SH
3324/**
3325 * netif_dormant_on - mark device as dormant.
3326 * @dev: network device
3327 *
3328 * Mark device as dormant (as per RFC2863).
3329 *
3330 * The dormant state indicates that the relevant interface is not
3331 * actually in a condition to pass packets (i.e., it is not 'up') but is
3332 * in a "pending" state, waiting for some external event. For "on-
3333 * demand" interfaces, this new state identifies the situation where the
3334 * interface is waiting for events to place it in the up state.
bea3348e 3335 */
b00055aa
SR
3336static inline void netif_dormant_on(struct net_device *dev)
3337{
3338 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3339 linkwatch_fire_event(dev);
3340}
3341
bea3348e
SH
3342/**
3343 * netif_dormant_off - set device as not dormant.
3344 * @dev: network device
3345 *
3346 * Device is not in dormant state.
3347 */
b00055aa
SR
3348static inline void netif_dormant_off(struct net_device *dev)
3349{
3350 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3351 linkwatch_fire_event(dev);
3352}
3353
bea3348e
SH
3354/**
3355 * netif_dormant - test if carrier present
3356 * @dev: network device
3357 *
3358 * Check if carrier is present on device
3359 */
4d29515f 3360static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3361{
3362 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3363}
3364
3365
bea3348e
SH
3366/**
3367 * netif_oper_up - test if device is operational
3368 * @dev: network device
3369 *
3370 * Check if carrier is operational
3371 */
4d29515f 3372static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3373{
b00055aa
SR
3374 return (dev->operstate == IF_OPER_UP ||
3375 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3376}
3377
bea3348e
SH
3378/**
3379 * netif_device_present - is device available or removed
3380 * @dev: network device
3381 *
3382 * Check if device has not been removed from system.
3383 */
4d29515f 3384static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3385{
3386 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3387}
3388
f629d208 3389void netif_device_detach(struct net_device *dev);
1da177e4 3390
f629d208 3391void netif_device_attach(struct net_device *dev);
1da177e4
LT
3392
3393/*
3394 * Network interface message level settings
3395 */
1da177e4
LT
3396
3397enum {
3398 NETIF_MSG_DRV = 0x0001,
3399 NETIF_MSG_PROBE = 0x0002,
3400 NETIF_MSG_LINK = 0x0004,
3401 NETIF_MSG_TIMER = 0x0008,
3402 NETIF_MSG_IFDOWN = 0x0010,
3403 NETIF_MSG_IFUP = 0x0020,
3404 NETIF_MSG_RX_ERR = 0x0040,
3405 NETIF_MSG_TX_ERR = 0x0080,
3406 NETIF_MSG_TX_QUEUED = 0x0100,
3407 NETIF_MSG_INTR = 0x0200,
3408 NETIF_MSG_TX_DONE = 0x0400,
3409 NETIF_MSG_RX_STATUS = 0x0800,
3410 NETIF_MSG_PKTDATA = 0x1000,
3411 NETIF_MSG_HW = 0x2000,
3412 NETIF_MSG_WOL = 0x4000,
3413};
3414
3415#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3416#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3417#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3418#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3419#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3420#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3421#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3422#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3423#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3424#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3425#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3426#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3427#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3428#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3429#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3430
3431static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3432{
3433 /* use default */
3434 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3435 return default_msg_enable_bits;
3436 if (debug_value == 0) /* no output */
3437 return 0;
3438 /* set low N bits */
3439 return (1 << debug_value) - 1;
3440}
3441
c773e847 3442static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3443{
c773e847
DM
3444 spin_lock(&txq->_xmit_lock);
3445 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3446}
3447
fd2ea0a7
DM
3448static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3449{
3450 spin_lock_bh(&txq->_xmit_lock);
3451 txq->xmit_lock_owner = smp_processor_id();
3452}
3453
4d29515f 3454static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3455{
4d29515f 3456 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3457 if (likely(ok))
3458 txq->xmit_lock_owner = smp_processor_id();
3459 return ok;
3460}
3461
3462static inline void __netif_tx_unlock(struct netdev_queue *txq)
3463{
3464 txq->xmit_lock_owner = -1;
3465 spin_unlock(&txq->_xmit_lock);
3466}
3467
3468static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3469{
3470 txq->xmit_lock_owner = -1;
3471 spin_unlock_bh(&txq->_xmit_lock);
3472}
3473
08baf561
ED
3474static inline void txq_trans_update(struct netdev_queue *txq)
3475{
3476 if (txq->xmit_lock_owner != -1)
3477 txq->trans_start = jiffies;
3478}
3479
ba162f8e
FW
3480/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3481static inline void netif_trans_update(struct net_device *dev)
3482{
9b36627a
FW
3483 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3484
3485 if (txq->trans_start != jiffies)
3486 txq->trans_start = jiffies;
ba162f8e
FW
3487}
3488
d29f749e
DJ
3489/**
3490 * netif_tx_lock - grab network device transmit lock
3491 * @dev: network device
d29f749e
DJ
3492 *
3493 * Get network device transmit lock
3494 */
22dd7495
JHS
3495static inline void netif_tx_lock(struct net_device *dev)
3496{
e8a0464c 3497 unsigned int i;
c3f26a26 3498 int cpu;
c773e847 3499
c3f26a26
DM
3500 spin_lock(&dev->tx_global_lock);
3501 cpu = smp_processor_id();
e8a0464c
DM
3502 for (i = 0; i < dev->num_tx_queues; i++) {
3503 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3504
3505 /* We are the only thread of execution doing a
3506 * freeze, but we have to grab the _xmit_lock in
3507 * order to synchronize with threads which are in
3508 * the ->hard_start_xmit() handler and already
3509 * checked the frozen bit.
3510 */
e8a0464c 3511 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3512 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3513 __netif_tx_unlock(txq);
e8a0464c 3514 }
932ff279
HX
3515}
3516
3517static inline void netif_tx_lock_bh(struct net_device *dev)
3518{
e8a0464c
DM
3519 local_bh_disable();
3520 netif_tx_lock(dev);
932ff279
HX
3521}
3522
932ff279
HX
3523static inline void netif_tx_unlock(struct net_device *dev)
3524{
e8a0464c
DM
3525 unsigned int i;
3526
3527 for (i = 0; i < dev->num_tx_queues; i++) {
3528 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3529
c3f26a26
DM
3530 /* No need to grab the _xmit_lock here. If the
3531 * queue is not stopped for another reason, we
3532 * force a schedule.
3533 */
3534 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3535 netif_schedule_queue(txq);
c3f26a26
DM
3536 }
3537 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3538}
3539
3540static inline void netif_tx_unlock_bh(struct net_device *dev)
3541{
e8a0464c
DM
3542 netif_tx_unlock(dev);
3543 local_bh_enable();
932ff279
HX
3544}
3545
c773e847 3546#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3547 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3548 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3549 } \
3550}
3551
5efeac44
EB
3552#define HARD_TX_TRYLOCK(dev, txq) \
3553 (((dev->features & NETIF_F_LLTX) == 0) ? \
3554 __netif_tx_trylock(txq) : \
3555 true )
3556
c773e847 3557#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3558 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3559 __netif_tx_unlock(txq); \
22dd7495
JHS
3560 } \
3561}
3562
1da177e4
LT
3563static inline void netif_tx_disable(struct net_device *dev)
3564{
fd2ea0a7 3565 unsigned int i;
c3f26a26 3566 int cpu;
fd2ea0a7 3567
c3f26a26
DM
3568 local_bh_disable();
3569 cpu = smp_processor_id();
fd2ea0a7
DM
3570 for (i = 0; i < dev->num_tx_queues; i++) {
3571 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3572
3573 __netif_tx_lock(txq, cpu);
fd2ea0a7 3574 netif_tx_stop_queue(txq);
c3f26a26 3575 __netif_tx_unlock(txq);
fd2ea0a7 3576 }
c3f26a26 3577 local_bh_enable();
1da177e4
LT
3578}
3579
e308a5d8
DM
3580static inline void netif_addr_lock(struct net_device *dev)
3581{
3582 spin_lock(&dev->addr_list_lock);
3583}
3584
2429f7ac
JP
3585static inline void netif_addr_lock_nested(struct net_device *dev)
3586{
25175ba5
VY
3587 int subclass = SINGLE_DEPTH_NESTING;
3588
3589 if (dev->netdev_ops->ndo_get_lock_subclass)
3590 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3591
3592 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3593}
3594
e308a5d8
DM
3595static inline void netif_addr_lock_bh(struct net_device *dev)
3596{
3597 spin_lock_bh(&dev->addr_list_lock);
3598}
3599
3600static inline void netif_addr_unlock(struct net_device *dev)
3601{
3602 spin_unlock(&dev->addr_list_lock);
3603}
3604
3605static inline void netif_addr_unlock_bh(struct net_device *dev)
3606{
3607 spin_unlock_bh(&dev->addr_list_lock);
3608}
3609
f001fde5 3610/*
31278e71 3611 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3612 * rcu_read_lock held.
3613 */
3614#define for_each_dev_addr(dev, ha) \
31278e71 3615 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3616
1da177e4
LT
3617/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3618
f629d208 3619void ether_setup(struct net_device *dev);
1da177e4
LT
3620
3621/* Support for loadable net-drivers */
f629d208 3622struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3623 unsigned char name_assign_type,
f629d208
JP
3624 void (*setup)(struct net_device *),
3625 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3626#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3627 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3628
c835a677
TG
3629#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3630 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3631 count)
36909ea4 3632
f629d208
JP
3633int register_netdev(struct net_device *dev);
3634void unregister_netdev(struct net_device *dev);
f001fde5 3635
22bedad3 3636/* General hardware address lists handling functions */
f629d208
JP
3637int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3638 struct netdev_hw_addr_list *from_list, int addr_len);
3639void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3640 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3641int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3642 struct net_device *dev,
3643 int (*sync)(struct net_device *, const unsigned char *),
3644 int (*unsync)(struct net_device *,
3645 const unsigned char *));
3646void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3647 struct net_device *dev,
3648 int (*unsync)(struct net_device *,
3649 const unsigned char *));
f629d208 3650void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3651
f001fde5 3652/* Functions used for device addresses handling */
f629d208
JP
3653int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3654 unsigned char addr_type);
3655int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3656 unsigned char addr_type);
f629d208
JP
3657void dev_addr_flush(struct net_device *dev);
3658int dev_addr_init(struct net_device *dev);
a748ee24
JP
3659
3660/* Functions used for unicast addresses handling */
f629d208
JP
3661int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3662int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3663int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3664int dev_uc_sync(struct net_device *to, struct net_device *from);
3665int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3666void dev_uc_unsync(struct net_device *to, struct net_device *from);
3667void dev_uc_flush(struct net_device *dev);
3668void dev_uc_init(struct net_device *dev);
f001fde5 3669
670e5b8e
AD
3670/**
3671 * __dev_uc_sync - Synchonize device's unicast list
3672 * @dev: device to sync
3673 * @sync: function to call if address should be added
3674 * @unsync: function to call if address should be removed
3675 *
3676 * Add newly added addresses to the interface, and release
3677 * addresses that have been deleted.
5e82b4b2 3678 */
670e5b8e
AD
3679static inline int __dev_uc_sync(struct net_device *dev,
3680 int (*sync)(struct net_device *,
3681 const unsigned char *),
3682 int (*unsync)(struct net_device *,
3683 const unsigned char *))
3684{
3685 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3686}
3687
3688/**
e793c0f7 3689 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3690 * @dev: device to sync
3691 * @unsync: function to call if address should be removed
3692 *
3693 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3694 */
670e5b8e
AD
3695static inline void __dev_uc_unsync(struct net_device *dev,
3696 int (*unsync)(struct net_device *,
3697 const unsigned char *))
3698{
3699 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3700}
3701
22bedad3 3702/* Functions used for multicast addresses handling */
f629d208
JP
3703int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3704int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3705int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3706int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3707int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3708int dev_mc_sync(struct net_device *to, struct net_device *from);
3709int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3710void dev_mc_unsync(struct net_device *to, struct net_device *from);
3711void dev_mc_flush(struct net_device *dev);
3712void dev_mc_init(struct net_device *dev);
f001fde5 3713
670e5b8e
AD
3714/**
3715 * __dev_mc_sync - Synchonize device's multicast list
3716 * @dev: device to sync
3717 * @sync: function to call if address should be added
3718 * @unsync: function to call if address should be removed
3719 *
3720 * Add newly added addresses to the interface, and release
3721 * addresses that have been deleted.
5e82b4b2 3722 */
670e5b8e
AD
3723static inline int __dev_mc_sync(struct net_device *dev,
3724 int (*sync)(struct net_device *,
3725 const unsigned char *),
3726 int (*unsync)(struct net_device *,
3727 const unsigned char *))
3728{
3729 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3730}
3731
3732/**
e793c0f7 3733 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3734 * @dev: device to sync
3735 * @unsync: function to call if address should be removed
3736 *
3737 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3738 */
670e5b8e
AD
3739static inline void __dev_mc_unsync(struct net_device *dev,
3740 int (*unsync)(struct net_device *,
3741 const unsigned char *))
3742{
3743 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3744}
3745
4417da66 3746/* Functions used for secondary unicast and multicast support */
f629d208
JP
3747void dev_set_rx_mode(struct net_device *dev);
3748void __dev_set_rx_mode(struct net_device *dev);
3749int dev_set_promiscuity(struct net_device *dev, int inc);
3750int dev_set_allmulti(struct net_device *dev, int inc);
3751void netdev_state_change(struct net_device *dev);
3752void netdev_notify_peers(struct net_device *dev);
3753void netdev_features_change(struct net_device *dev);
1da177e4 3754/* Load a device via the kmod */
f629d208
JP
3755void dev_load(struct net *net, const char *name);
3756struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3757 struct rtnl_link_stats64 *storage);
3758void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3759 const struct net_device_stats *netdev_stats);
eeda3fd6 3760
1da177e4 3761extern int netdev_max_backlog;
3b098e2d 3762extern int netdev_tstamp_prequeue;
1da177e4 3763extern int weight_p;
9ff162a8 3764
f629d208 3765bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3766struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3767 struct list_head **iter);
f629d208
JP
3768struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3769 struct list_head **iter);
8b5be856 3770
44a40855
VY
3771/* iterate through upper list, must be called under RCU read lock */
3772#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3773 for (iter = &(dev)->adj_list.upper, \
3774 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3775 updev; \
3776 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3777
8b5be856 3778/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3779#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3780 for (iter = &(dev)->all_adj_list.upper, \
3781 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3782 updev; \
3783 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3784
f629d208
JP
3785void *netdev_lower_get_next_private(struct net_device *dev,
3786 struct list_head **iter);
3787void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3788 struct list_head **iter);
31088a11
VF
3789
3790#define netdev_for_each_lower_private(dev, priv, iter) \
3791 for (iter = (dev)->adj_list.lower.next, \
3792 priv = netdev_lower_get_next_private(dev, &(iter)); \
3793 priv; \
3794 priv = netdev_lower_get_next_private(dev, &(iter)))
3795
3796#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3797 for (iter = &(dev)->adj_list.lower, \
3798 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3799 priv; \
3800 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3801
4085ebe8
VY
3802void *netdev_lower_get_next(struct net_device *dev,
3803 struct list_head **iter);
3804#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3805 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3806 ldev = netdev_lower_get_next(dev, &(iter)); \
3807 ldev; \
3808 ldev = netdev_lower_get_next(dev, &(iter)))
3809
f629d208 3810void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3811void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3812struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3813struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3814int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3815int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3816 struct net_device *upper_dev,
29bf24af 3817 void *upper_priv, void *upper_info);
f629d208
JP
3818void netdev_upper_dev_unlink(struct net_device *dev,
3819 struct net_device *upper_dev);
5bb025fa 3820void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3821void *netdev_lower_dev_get_private(struct net_device *dev,
3822 struct net_device *lower_dev);
04d48266
JP
3823void netdev_lower_state_changed(struct net_device *lower_dev,
3824 void *lower_state_info);
960fb622
ED
3825
3826/* RSS keys are 40 or 52 bytes long */
3827#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3828extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3829void netdev_rss_key_fill(void *buffer, size_t len);
3830
4085ebe8 3831int dev_get_nest_level(struct net_device *dev,
b618aaa9 3832 bool (*type_check)(const struct net_device *dev));
f629d208
JP
3833int skb_checksum_help(struct sk_buff *skb);
3834struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3835 netdev_features_t features, bool tx_path);
3836struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3837 netdev_features_t features);
12b0004d 3838
61bd3857
MS
3839struct netdev_bonding_info {
3840 ifslave slave;
3841 ifbond master;
3842};
3843
3844struct netdev_notifier_bonding_info {
3845 struct netdev_notifier_info info; /* must be first */
3846 struct netdev_bonding_info bonding_info;
3847};
3848
3849void netdev_bonding_info_change(struct net_device *dev,
3850 struct netdev_bonding_info *bonding_info);
3851
12b0004d
CW
3852static inline
3853struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3854{
3855 return __skb_gso_segment(skb, features, true);
3856}
53d6471c 3857__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3858
3859static inline bool can_checksum_protocol(netdev_features_t features,
3860 __be16 protocol)
3861{
c8cd0989
TH
3862 if (protocol == htons(ETH_P_FCOE))
3863 return !!(features & NETIF_F_FCOE_CRC);
3864
3865 /* Assume this is an IP checksum (not SCTP CRC) */
3866
3867 if (features & NETIF_F_HW_CSUM) {
3868 /* Can checksum everything */
3869 return true;
3870 }
3871
3872 switch (protocol) {
3873 case htons(ETH_P_IP):
3874 return !!(features & NETIF_F_IP_CSUM);
3875 case htons(ETH_P_IPV6):
3876 return !!(features & NETIF_F_IPV6_CSUM);
3877 default:
3878 return false;
3879 }
ec5f0615 3880}
12b0004d 3881
6ae23ad3
TH
3882/* Map an ethertype into IP protocol if possible */
3883static inline int eproto_to_ipproto(int eproto)
3884{
3885 switch (eproto) {
3886 case htons(ETH_P_IP):
3887 return IPPROTO_IP;
3888 case htons(ETH_P_IPV6):
3889 return IPPROTO_IPV6;
3890 default:
3891 return -1;
3892 }
3893}
3894
fb286bb2 3895#ifdef CONFIG_BUG
f629d208 3896void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3897#else
3898static inline void netdev_rx_csum_fault(struct net_device *dev)
3899{
3900}
3901#endif
1da177e4 3902/* rx skb timestamps */
f629d208
JP
3903void net_enable_timestamp(void);
3904void net_disable_timestamp(void);
1da177e4 3905
20380731 3906#ifdef CONFIG_PROC_FS
f629d208 3907int __init dev_proc_init(void);
900ff8c6
CW
3908#else
3909#define dev_proc_init() 0
20380731
ACM
3910#endif
3911
4798248e 3912static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3913 struct sk_buff *skb, struct net_device *dev,
3914 bool more)
4798248e 3915{
fa2dbdc2 3916 skb->xmit_more = more ? 1 : 0;
0b725a2c 3917 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3918}
3919
10b3ad8c 3920static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3921 struct netdev_queue *txq, bool more)
4798248e
DM
3922{
3923 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3924 int rc;
4798248e 3925
fa2dbdc2 3926 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3927 if (rc == NETDEV_TX_OK)
3928 txq_trans_update(txq);
3929
3930 return rc;
4798248e
DM
3931}
3932
42a2d923
LT
3933int netdev_class_create_file_ns(struct class_attribute *class_attr,
3934 const void *ns);
3935void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3936 const void *ns);
58292cbe
TH
3937
3938static inline int netdev_class_create_file(struct class_attribute *class_attr)
3939{
3940 return netdev_class_create_file_ns(class_attr, NULL);
3941}
3942
3943static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3944{
3945 netdev_class_remove_file_ns(class_attr, NULL);
3946}
b8a9787e 3947
04600794
JB
3948extern struct kobj_ns_type_operations net_ns_type_operations;
3949
f629d208 3950const char *netdev_drivername(const struct net_device *dev);
6579e57b 3951
f629d208 3952void linkwatch_run_queue(void);
20380731 3953
da08143b
MK
3954static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3955 netdev_features_t f2)
3956{
c8cd0989
TH
3957 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3958 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 3959 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 3960 else
b6a0e72a 3961 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 3962 }
da08143b 3963
c8cd0989 3964 return f1 & f2;
da08143b
MK
3965}
3966
c8f44aff
MM
3967static inline netdev_features_t netdev_get_wanted_features(
3968 struct net_device *dev)
5455c699
MM
3969{
3970 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3971}
c8f44aff
MM
3972netdev_features_t netdev_increment_features(netdev_features_t all,
3973 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3974
3975/* Allow TSO being used on stacked device :
3976 * Performing the GSO segmentation before last device
3977 * is a performance improvement.
3978 */
3979static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3980 netdev_features_t mask)
3981{
3982 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3983}
3984
6cb6a27c 3985int __netdev_update_features(struct net_device *dev);
5455c699 3986void netdev_update_features(struct net_device *dev);
afe12cc8 3987void netdev_change_features(struct net_device *dev);
7f353bf2 3988
fc4a7489
PM
3989void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3990 struct net_device *dev);
3991
e38f3025
TM
3992netdev_features_t passthru_features_check(struct sk_buff *skb,
3993 struct net_device *dev,
3994 netdev_features_t features);
c1e756bf 3995netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3996
4d29515f 3997static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3998{
7b748340 3999 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4000
4001 /* check flags correspondence */
4002 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4003 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4004 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4005 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4006 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4007 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4008 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4009 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4010 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4011 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4012 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4013 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4014 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4015 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4016 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4017 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
0345e186 4018
d6b4991a 4019 return (features & feature) == feature;
576a30eb
HX
4020}
4021
4d29515f 4022static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4023{
278b2513 4024 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4025 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4026}
4027
8b86a61d 4028static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4029 netdev_features_t features)
7967168c 4030{
fc741216 4031 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4032 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4033 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4034}
4035
82cc1a7a
PWJ
4036static inline void netif_set_gso_max_size(struct net_device *dev,
4037 unsigned int size)
4038{
4039 dev->gso_max_size = size;
4040}
4041
7a7ffbab
WCC
4042static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4043 int pulled_hlen, u16 mac_offset,
4044 int mac_len)
4045{
4046 skb->protocol = protocol;
4047 skb->encapsulation = 1;
4048 skb_push(skb, pulled_hlen);
4049 skb_reset_transport_header(skb);
4050 skb->mac_header = mac_offset;
4051 skb->network_header = skb->mac_header + mac_len;
4052 skb->mac_len = mac_len;
4053}
4054
3c175784
SD
4055static inline bool netif_is_macsec(const struct net_device *dev)
4056{
4057 return dev->priv_flags & IFF_MACSEC;
4058}
4059
b618aaa9 4060static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4061{
4062 return dev->priv_flags & IFF_MACVLAN;
4063}
4064
b618aaa9 4065static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4066{
4067 return dev->priv_flags & IFF_MACVLAN_PORT;
4068}
4069
b618aaa9 4070static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4071{
4072 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4073}
4074
b618aaa9 4075static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4076{
4077 return dev->priv_flags & IFF_IPVLAN_MASTER;
4078}
4079
b618aaa9 4080static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4081{
4082 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4083}
4084
b618aaa9 4085static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4086{
4087 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4088}
4089
3bdc0eba
BG
4090static inline bool netif_supports_nofcs(struct net_device *dev)
4091{
4092 return dev->priv_flags & IFF_SUPP_NOFCS;
4093}
4094
007979ea 4095static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4096{
007979ea 4097 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4098}
4099
fee6d4c7
DA
4100static inline bool netif_is_l3_slave(const struct net_device *dev)
4101{
4102 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4103}
4104
0894ae3f
JP
4105static inline bool netif_is_bridge_master(const struct net_device *dev)
4106{
4107 return dev->priv_flags & IFF_EBRIDGE;
4108}
4109
28f9ee22
VY
4110static inline bool netif_is_bridge_port(const struct net_device *dev)
4111{
4112 return dev->priv_flags & IFF_BRIDGE_PORT;
4113}
4114
35d4e172
JP
4115static inline bool netif_is_ovs_master(const struct net_device *dev)
4116{
4117 return dev->priv_flags & IFF_OPENVSWITCH;
4118}
4119
b618aaa9 4120static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4121{
4122 return dev->priv_flags & IFF_TEAM;
4123}
4124
b618aaa9 4125static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4126{
4127 return dev->priv_flags & IFF_TEAM_PORT;
4128}
4129
b618aaa9 4130static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4131{
4132 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4133}
4134
b618aaa9 4135static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4136{
4137 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4138}
4139
d4ab4286
KJ
4140static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4141{
4142 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4143}
4144
02875878
ED
4145/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4146static inline void netif_keep_dst(struct net_device *dev)
4147{
4148 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4149}
4150
505d4f73 4151extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4152
571ba423
JP
4153/* Logging, debugging and troubleshooting/diagnostic helpers. */
4154
4155/* netdev_printk helpers, similar to dev_printk */
4156
4157static inline const char *netdev_name(const struct net_device *dev)
4158{
c6f854d5
VF
4159 if (!dev->name[0] || strchr(dev->name, '%'))
4160 return "(unnamed net_device)";
571ba423
JP
4161 return dev->name;
4162}
4163
ccc7f496
VF
4164static inline const char *netdev_reg_state(const struct net_device *dev)
4165{
4166 switch (dev->reg_state) {
4167 case NETREG_UNINITIALIZED: return " (uninitialized)";
4168 case NETREG_REGISTERED: return "";
4169 case NETREG_UNREGISTERING: return " (unregistering)";
4170 case NETREG_UNREGISTERED: return " (unregistered)";
4171 case NETREG_RELEASED: return " (released)";
4172 case NETREG_DUMMY: return " (dummy)";
4173 }
4174
4175 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4176 return " (unknown)";
4177}
4178
f629d208 4179__printf(3, 4)
6ea754eb
JP
4180void netdev_printk(const char *level, const struct net_device *dev,
4181 const char *format, ...);
f629d208 4182__printf(2, 3)
6ea754eb 4183void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4184__printf(2, 3)
6ea754eb 4185void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4186__printf(2, 3)
6ea754eb 4187void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4188__printf(2, 3)
6ea754eb 4189void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4190__printf(2, 3)
6ea754eb 4191void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4192__printf(2, 3)
6ea754eb 4193void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4194__printf(2, 3)
6ea754eb 4195void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4196
8909c9ad
VK
4197#define MODULE_ALIAS_NETDEV(device) \
4198 MODULE_ALIAS("netdev-" device)
4199
b558c96f 4200#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4201#define netdev_dbg(__dev, format, args...) \
4202do { \
ffa10cb4 4203 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4204} while (0)
b558c96f
JC
4205#elif defined(DEBUG)
4206#define netdev_dbg(__dev, format, args...) \
4207 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4208#else
4209#define netdev_dbg(__dev, format, args...) \
4210({ \
4211 if (0) \
4212 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4213})
4214#endif
4215
4216#if defined(VERBOSE_DEBUG)
4217#define netdev_vdbg netdev_dbg
4218#else
4219
4220#define netdev_vdbg(dev, format, args...) \
4221({ \
4222 if (0) \
4223 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4224 0; \
4225})
4226#endif
4227
4228/*
4229 * netdev_WARN() acts like dev_printk(), but with the key difference
4230 * of using a WARN/WARN_ON to get the message out, including the
4231 * file/line information and a backtrace.
4232 */
4233#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4234 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4235 netdev_reg_state(dev), ##args)
571ba423 4236
b3d95c5c
JP
4237/* netif printk helpers, similar to netdev_printk */
4238
4239#define netif_printk(priv, type, level, dev, fmt, args...) \
4240do { \
4241 if (netif_msg_##type(priv)) \
4242 netdev_printk(level, (dev), fmt, ##args); \
4243} while (0)
4244
f45f4321
JP
4245#define netif_level(level, priv, type, dev, fmt, args...) \
4246do { \
4247 if (netif_msg_##type(priv)) \
4248 netdev_##level(dev, fmt, ##args); \
4249} while (0)
4250
b3d95c5c 4251#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4252 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4253#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4254 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4255#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4256 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4257#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4258 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4259#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4260 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4261#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4262 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4263#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4264 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4265
0053ea9c 4266#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4267#define netif_dbg(priv, type, netdev, format, args...) \
4268do { \
4269 if (netif_msg_##type(priv)) \
b5fb0a03 4270 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4271} while (0)
0053ea9c
JP
4272#elif defined(DEBUG)
4273#define netif_dbg(priv, type, dev, format, args...) \
4274 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4275#else
4276#define netif_dbg(priv, type, dev, format, args...) \
4277({ \
4278 if (0) \
4279 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4280 0; \
4281})
4282#endif
4283
4284#if defined(VERBOSE_DEBUG)
bcfcc450 4285#define netif_vdbg netif_dbg
b3d95c5c
JP
4286#else
4287#define netif_vdbg(priv, type, dev, format, args...) \
4288({ \
4289 if (0) \
a4ed89cb 4290 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4291 0; \
4292})
4293#endif
571ba423 4294
900ff8c6
CW
4295/*
4296 * The list of packet types we will receive (as opposed to discard)
4297 * and the routines to invoke.
4298 *
4299 * Why 16. Because with 16 the only overlap we get on a hash of the
4300 * low nibble of the protocol value is RARP/SNAP/X.25.
4301 *
4302 * NOTE: That is no longer true with the addition of VLAN tags. Not
4303 * sure which should go first, but I bet it won't make much
4304 * difference if we are running VLANs. The good news is that
4305 * this protocol won't be in the list unless compiled in, so
4306 * the average user (w/out VLANs) will not be adversely affected.
4307 * --BLG
4308 *
4309 * 0800 IP
4310 * 8100 802.1Q VLAN
4311 * 0001 802.3
4312 * 0002 AX.25
4313 * 0004 802.2
4314 * 8035 RARP
4315 * 0005 SNAP
4316 * 0805 X.25
4317 * 0806 ARP
4318 * 8137 IPX
4319 * 0009 Localtalk
4320 * 86DD IPv6
4321 */
4322#define PTYPE_HASH_SIZE (16)
4323#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4324
385a154c 4325#endif /* _LINUX_NETDEVICE_H */