]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
net: Combine GENEVE and VXLAN port notifiers into single functions
[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 */
9a1654ba 93#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 94
b9df3cb8
GR
95/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 98#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
99#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
dc1f8bf6 101/* Driver transmit return codes */
9a1654ba 102#define NETDEV_TX_MASK 0xf0
572a9d7b 103
dc1f8bf6 104enum netdev_tx {
572a9d7b 105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
108};
109typedef enum netdev_tx netdev_tx_t;
110
9a1654ba
JP
111/*
112 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
113 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
114 */
115static inline bool dev_xmit_complete(int rc)
116{
117 /*
118 * Positive cases with an skb consumed by a driver:
119 * - successful transmission (rc == NETDEV_TX_OK)
120 * - error while transmitting (rc < 0)
121 * - error while queueing to a different device (rc & NET_XMIT_MASK)
122 */
123 if (likely(rc < NET_XMIT_MASK))
124 return true;
125
126 return false;
127}
128
1da177e4 129/*
5e82b4b2 130 * Compute the worst-case header length according to the protocols
1da177e4
LT
131 * used.
132 */
fe2918b0 133
c0eb4540
KS
134#if defined(CONFIG_HYPERV_NET)
135# define LL_MAX_HEADER 128
136#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
137# if defined(CONFIG_MAC80211_MESH)
138# define LL_MAX_HEADER 128
139# else
140# define LL_MAX_HEADER 96
141# endif
1da177e4 142#else
8388e3da 143# define LL_MAX_HEADER 32
1da177e4
LT
144#endif
145
d11ead75
BH
146#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
147 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
148#define MAX_HEADER LL_MAX_HEADER
149#else
150#define MAX_HEADER (LL_MAX_HEADER + 48)
151#endif
152
153/*
be1f3c2c
BH
154 * Old network device statistics. Fields are native words
155 * (unsigned long) so they can be read and written atomically.
1da177e4 156 */
fe2918b0 157
d94d9fee 158struct net_device_stats {
3cfde79c
BH
159 unsigned long rx_packets;
160 unsigned long tx_packets;
161 unsigned long rx_bytes;
162 unsigned long tx_bytes;
163 unsigned long rx_errors;
164 unsigned long tx_errors;
165 unsigned long rx_dropped;
166 unsigned long tx_dropped;
167 unsigned long multicast;
1da177e4 168 unsigned long collisions;
1da177e4 169 unsigned long rx_length_errors;
3cfde79c
BH
170 unsigned long rx_over_errors;
171 unsigned long rx_crc_errors;
172 unsigned long rx_frame_errors;
173 unsigned long rx_fifo_errors;
174 unsigned long rx_missed_errors;
1da177e4
LT
175 unsigned long tx_aborted_errors;
176 unsigned long tx_carrier_errors;
177 unsigned long tx_fifo_errors;
178 unsigned long tx_heartbeat_errors;
179 unsigned long tx_window_errors;
1da177e4
LT
180 unsigned long rx_compressed;
181 unsigned long tx_compressed;
182};
183
1da177e4
LT
184
185#include <linux/cache.h>
186#include <linux/skbuff.h>
187
adc9300e 188#ifdef CONFIG_RPS
c5905afb
IM
189#include <linux/static_key.h>
190extern struct static_key rps_needed;
adc9300e
ED
191#endif
192
1da177e4
LT
193struct neighbour;
194struct neigh_parms;
195struct sk_buff;
196
f001fde5
JP
197struct netdev_hw_addr {
198 struct list_head list;
199 unsigned char addr[MAX_ADDR_LEN];
200 unsigned char type;
ccffad25
JP
201#define NETDEV_HW_ADDR_T_LAN 1
202#define NETDEV_HW_ADDR_T_SAN 2
203#define NETDEV_HW_ADDR_T_SLAVE 3
204#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 205#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 206 bool global_use;
4cd729b0 207 int sync_cnt;
8f8f103d 208 int refcount;
4543fbef 209 int synced;
f001fde5
JP
210 struct rcu_head rcu_head;
211};
212
31278e71
JP
213struct netdev_hw_addr_list {
214 struct list_head list;
215 int count;
216};
217
22bedad3
JP
218#define netdev_hw_addr_list_count(l) ((l)->count)
219#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
220#define netdev_hw_addr_list_for_each(ha, l) \
221 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 222
22bedad3
JP
223#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
224#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
225#define netdev_for_each_uc_addr(ha, dev) \
226 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 227
22bedad3
JP
228#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
229#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 230#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 231 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 232
d94d9fee 233struct hh_cache {
f6b72b62 234 u16 hh_len;
5c25f686 235 u16 __pad;
3644f0ce 236 seqlock_t hh_lock;
1da177e4
LT
237
238 /* cached hardware header; allow for machine alignment needs. */
239#define HH_DATA_MOD 16
240#define HH_DATA_OFF(__len) \
5ba0eac6 241 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
242#define HH_DATA_ALIGN(__len) \
243 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
244 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
245};
246
5e82b4b2 247/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
248 * Alternative is:
249 * dev->hard_header_len ? (dev->hard_header_len +
250 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
251 *
252 * We could use other alignment values, but we must maintain the
253 * relationship HH alignment <= LL alignment.
254 */
255#define LL_RESERVED_SPACE(dev) \
f5184d26 256 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 257#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 258 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 259
3b04ddde
SH
260struct header_ops {
261 int (*create) (struct sk_buff *skb, struct net_device *dev,
262 unsigned short type, const void *daddr,
95c96174 263 const void *saddr, unsigned int len);
3b04ddde 264 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 265 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
266 void (*cache_update)(struct hh_cache *hh,
267 const struct net_device *dev,
268 const unsigned char *haddr);
2793a23a 269 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
270};
271
1da177e4 272/* These flag bits are private to the generic network queueing
5e82b4b2 273 * layer; they may not be explicitly referenced by any other
1da177e4
LT
274 * code.
275 */
276
d94d9fee 277enum netdev_state_t {
1da177e4
LT
278 __LINK_STATE_START,
279 __LINK_STATE_PRESENT,
1da177e4 280 __LINK_STATE_NOCARRIER,
b00055aa
SR
281 __LINK_STATE_LINKWATCH_PENDING,
282 __LINK_STATE_DORMANT,
1da177e4
LT
283};
284
285
286/*
5e82b4b2 287 * This structure holds boot-time configured netdevice settings. They
fe2918b0 288 * are then used in the device probing.
1da177e4
LT
289 */
290struct netdev_boot_setup {
291 char name[IFNAMSIZ];
292 struct ifmap map;
293};
294#define NETDEV_BOOT_SETUP_MAX 8
295
f629d208 296int __init netdev_boot_setup(char *str);
1da177e4 297
bea3348e
SH
298/*
299 * Structure for NAPI scheduling similar to tasklet but with weighting
300 */
301struct napi_struct {
302 /* The poll_list must only be managed by the entity which
303 * changes the state of the NAPI_STATE_SCHED bit. This means
304 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 305 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
306 * can remove from the list right before clearing the bit.
307 */
308 struct list_head poll_list;
309
310 unsigned long state;
311 int weight;
404f7c9e 312 unsigned int gro_count;
bea3348e
SH
313 int (*poll)(struct napi_struct *, int);
314#ifdef CONFIG_NETPOLL
315 spinlock_t poll_lock;
316 int poll_owner;
bea3348e 317#endif
5d38a079 318 struct net_device *dev;
d565b0a1 319 struct sk_buff *gro_list;
5d38a079 320 struct sk_buff *skb;
3b47d303 321 struct hrtimer timer;
404f7c9e 322 struct list_head dev_list;
af12fa6e
ET
323 struct hlist_node napi_hash_node;
324 unsigned int napi_id;
bea3348e
SH
325};
326
d94d9fee 327enum {
bea3348e 328 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 329 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 330 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
331 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
332 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
bea3348e
SH
333};
334
5b252f0c 335enum gro_result {
d1c76af9
HX
336 GRO_MERGED,
337 GRO_MERGED_FREE,
338 GRO_HELD,
339 GRO_NORMAL,
340 GRO_DROP,
341};
5b252f0c 342typedef enum gro_result gro_result_t;
d1c76af9 343
8a4eb573
JP
344/*
345 * enum rx_handler_result - Possible return values for rx_handlers.
346 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
347 * further.
348 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
349 * case skb->dev was changed by rx_handler.
350 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 351 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
352 *
353 * rx_handlers are functions called from inside __netif_receive_skb(), to do
354 * special processing of the skb, prior to delivery to protocol handlers.
355 *
356 * Currently, a net_device can only have a single rx_handler registered. Trying
357 * to register a second rx_handler will return -EBUSY.
358 *
359 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
360 * To unregister a rx_handler on a net_device, use
361 * netdev_rx_handler_unregister().
362 *
363 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
364 * do with the skb.
365 *
5e82b4b2 366 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 367 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 368 * the skb to be delivered in some other way.
8a4eb573
JP
369 *
370 * If the rx_handler changed skb->dev, to divert the skb to another
371 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
372 * new device will be called if it exists.
373 *
5e82b4b2 374 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 375 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 376 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 377 *
5e82b4b2 378 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
379 * delivered, it should return RX_HANDLER_PASS.
380 *
381 * A device without a registered rx_handler will behave as if rx_handler
382 * returned RX_HANDLER_PASS.
383 */
384
385enum rx_handler_result {
386 RX_HANDLER_CONSUMED,
387 RX_HANDLER_ANOTHER,
388 RX_HANDLER_EXACT,
389 RX_HANDLER_PASS,
390};
391typedef enum rx_handler_result rx_handler_result_t;
392typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 393
f629d208 394void __napi_schedule(struct napi_struct *n);
bc9ad166 395void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 396
4d29515f 397static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
398{
399 return test_bit(NAPI_STATE_DISABLE, &n->state);
400}
401
bea3348e 402/**
5e82b4b2
BH
403 * napi_schedule_prep - check if NAPI can be scheduled
404 * @n: NAPI context
bea3348e
SH
405 *
406 * Test if NAPI routine is already running, and if not mark
5e82b4b2 407 * it as running. This is used as a condition variable to
a0a46196
DM
408 * insure only one NAPI poll instance runs. We also make
409 * sure there is no pending NAPI disable.
bea3348e 410 */
4d29515f 411static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 412{
a0a46196
DM
413 return !napi_disable_pending(n) &&
414 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
415}
416
417/**
418 * napi_schedule - schedule NAPI poll
5e82b4b2 419 * @n: NAPI context
bea3348e
SH
420 *
421 * Schedule NAPI poll routine to be called if it is not already
422 * running.
423 */
424static inline void napi_schedule(struct napi_struct *n)
425{
426 if (napi_schedule_prep(n))
427 __napi_schedule(n);
428}
429
bc9ad166
ED
430/**
431 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 432 * @n: NAPI context
bc9ad166
ED
433 *
434 * Variant of napi_schedule(), assuming hard irqs are masked.
435 */
436static inline void napi_schedule_irqoff(struct napi_struct *n)
437{
438 if (napi_schedule_prep(n))
439 __napi_schedule_irqoff(n);
440}
441
bfe13f54 442/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 443static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
444{
445 if (napi_schedule_prep(napi)) {
446 __napi_schedule(napi);
4d29515f 447 return true;
bfe13f54 448 }
4d29515f 449 return false;
bfe13f54
RD
450}
451
3b47d303
ED
452void __napi_complete(struct napi_struct *n);
453void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
454/**
455 * napi_complete - NAPI processing complete
5e82b4b2 456 * @n: NAPI context
bea3348e
SH
457 *
458 * Mark NAPI processing as complete.
3b47d303 459 * Consider using napi_complete_done() instead.
bea3348e 460 */
3b47d303
ED
461static inline void napi_complete(struct napi_struct *n)
462{
463 return napi_complete_done(n, 0);
464}
bea3348e 465
af12fa6e
ET
466/**
467 * napi_hash_add - add a NAPI to global hashtable
5e82b4b2 468 * @napi: NAPI context
af12fa6e 469 *
5e82b4b2
BH
470 * Generate a new napi_id and store a @napi under it in napi_hash.
471 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
93d05d4a 472 * Note: This is normally automatically done from netif_napi_add(),
5e82b4b2 473 * so might disappear in a future Linux version.
af12fa6e 474 */
f629d208 475void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
476
477/**
478 * napi_hash_del - remove a NAPI from global table
5e82b4b2 479 * @napi: NAPI context
af12fa6e 480 *
5e82b4b2 481 * Warning: caller must observe RCU grace period
34cbe27e
ED
482 * before freeing memory containing @napi, if
483 * this function returns true.
93d05d4a 484 * Note: core networking stack automatically calls it
5e82b4b2 485 * from netif_napi_del().
93d05d4a 486 * Drivers might want to call this helper to combine all
5e82b4b2 487 * the needed RCU grace periods into a single one.
af12fa6e 488 */
34cbe27e 489bool napi_hash_del(struct napi_struct *napi);
af12fa6e 490
bea3348e
SH
491/**
492 * napi_disable - prevent NAPI from scheduling
5e82b4b2 493 * @n: NAPI context
bea3348e
SH
494 *
495 * Stop NAPI from being scheduled on this context.
496 * Waits till any outstanding processing completes.
497 */
3b47d303 498void napi_disable(struct napi_struct *n);
bea3348e
SH
499
500/**
501 * napi_enable - enable NAPI scheduling
5e82b4b2 502 * @n: NAPI context
bea3348e
SH
503 *
504 * Resume NAPI from being scheduled on this context.
505 * Must be paired with napi_disable.
506 */
507static inline void napi_enable(struct napi_struct *n)
508{
509 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 510 smp_mb__before_atomic();
bea3348e 511 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 512 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
513}
514
c264c3de
SH
515/**
516 * napi_synchronize - wait until NAPI is not running
5e82b4b2 517 * @n: NAPI context
c264c3de
SH
518 *
519 * Wait until NAPI is done being scheduled on this context.
520 * Waits till any outstanding processing completes but
521 * does not disable future activations.
522 */
523static inline void napi_synchronize(const struct napi_struct *n)
524{
facc432f
AB
525 if (IS_ENABLED(CONFIG_SMP))
526 while (test_bit(NAPI_STATE_SCHED, &n->state))
527 msleep(1);
528 else
529 barrier();
c264c3de 530}
c264c3de 531
d94d9fee 532enum netdev_queue_state_t {
73466498
TH
533 __QUEUE_STATE_DRV_XOFF,
534 __QUEUE_STATE_STACK_XOFF,
c3f26a26 535 __QUEUE_STATE_FROZEN,
79d16385 536};
8e2f1a63
DB
537
538#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
539#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
540#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
541
542#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
543#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
544 QUEUE_STATE_FROZEN)
545#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
546 QUEUE_STATE_FROZEN)
547
73466498
TH
548/*
549 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
550 * netif_tx_* functions below are used to manipulate this flag. The
551 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
552 * queue independently. The netif_xmit_*stopped functions below are called
553 * to check if the queue has been stopped by the driver or stack (either
554 * of the XOFF bits are set in the state). Drivers should not need to call
555 * netif_xmit*stopped functions, they should only be using netif_tx_*.
556 */
79d16385 557
bb949fbd 558struct netdev_queue {
6a321cb3 559/*
5e82b4b2 560 * read-mostly part
6a321cb3 561 */
bb949fbd 562 struct net_device *dev;
46e5da40 563 struct Qdisc __rcu *qdisc;
b0e1e646 564 struct Qdisc *qdisc_sleeping;
ccf5ff69 565#ifdef CONFIG_SYSFS
1d24eb48
TH
566 struct kobject kobj;
567#endif
f2cd2d3e
ED
568#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
569 int numa_node;
570#endif
c0ef079c
FW
571 unsigned long tx_maxrate;
572 /*
573 * Number of TX timeouts for this queue
574 * (/sys/class/net/DEV/Q/trans_timeout)
575 */
576 unsigned long trans_timeout;
6a321cb3 577/*
5e82b4b2 578 * write-mostly part
6a321cb3
ED
579 */
580 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
581 int xmit_lock_owner;
9d21493b 582 /*
9b36627a 583 * Time (in jiffies) of last Tx
9d21493b
ED
584 */
585 unsigned long trans_start;
ccf5ff69 586
114cf580
TH
587 unsigned long state;
588
589#ifdef CONFIG_BQL
590 struct dql dql;
591#endif
e8a0464c 592} ____cacheline_aligned_in_smp;
bb949fbd 593
f2cd2d3e
ED
594static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
595{
596#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
597 return q->numa_node;
598#else
b236da69 599 return NUMA_NO_NODE;
f2cd2d3e
ED
600#endif
601}
602
603static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
604{
605#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 q->numa_node = node;
607#endif
608}
609
df334545 610#ifdef CONFIG_RPS
0a9627f2
TH
611/*
612 * This structure holds an RPS map which can be of variable length. The
613 * map is an array of CPUs.
614 */
615struct rps_map {
616 unsigned int len;
617 struct rcu_head rcu;
618 u16 cpus[0];
619};
60b778ce 620#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 621
fec5e652 622/*
c445477d
BH
623 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
624 * tail pointer for that CPU's input queue at the time of last enqueue, and
625 * a hardware filter index.
fec5e652
TH
626 */
627struct rps_dev_flow {
628 u16 cpu;
c445477d 629 u16 filter;
fec5e652
TH
630 unsigned int last_qtail;
631};
c445477d 632#define RPS_NO_FILTER 0xffff
fec5e652
TH
633
634/*
635 * The rps_dev_flow_table structure contains a table of flow mappings.
636 */
637struct rps_dev_flow_table {
638 unsigned int mask;
639 struct rcu_head rcu;
fec5e652
TH
640 struct rps_dev_flow flows[0];
641};
642#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 643 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
644
645/*
646 * The rps_sock_flow_table contains mappings of flows to the last CPU
647 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
648 * Each entry is a 32bit value. Upper part is the high-order bits
649 * of flow hash, lower part is CPU number.
567e4b79 650 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
651 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
652 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 653 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
654 */
655struct rps_sock_flow_table {
567e4b79 656 u32 mask;
93c1af6c
ED
657
658 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 659};
567e4b79 660#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
661
662#define RPS_NO_CPU 0xffff
663
567e4b79
ED
664extern u32 rps_cpu_mask;
665extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
666
fec5e652
TH
667static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
668 u32 hash)
669{
670 if (table && hash) {
567e4b79
ED
671 unsigned int index = hash & table->mask;
672 u32 val = hash & ~rps_cpu_mask;
fec5e652 673
5e82b4b2 674 /* We only give a hint, preemption can change CPU under us */
567e4b79 675 val |= raw_smp_processor_id();
fec5e652 676
567e4b79
ED
677 if (table->ents[index] != val)
678 table->ents[index] = val;
fec5e652
TH
679 }
680}
681
c445477d 682#ifdef CONFIG_RFS_ACCEL
f629d208
JP
683bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
684 u16 filter_id);
c445477d 685#endif
a953be53 686#endif /* CONFIG_RPS */
c445477d 687
0a9627f2
TH
688/* This structure contains an instance of an RX queue. */
689struct netdev_rx_queue {
a953be53 690#ifdef CONFIG_RPS
6e3f7faf
ED
691 struct rps_map __rcu *rps_map;
692 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 693#endif
6e3f7faf 694 struct kobject kobj;
fe822240 695 struct net_device *dev;
0a9627f2 696} ____cacheline_aligned_in_smp;
a953be53
MD
697
698/*
699 * RX queue sysfs structures and functions.
700 */
701struct rx_queue_attribute {
702 struct attribute attr;
703 ssize_t (*show)(struct netdev_rx_queue *queue,
704 struct rx_queue_attribute *attr, char *buf);
705 ssize_t (*store)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, const char *buf, size_t len);
707};
d314774c 708
bf264145
TH
709#ifdef CONFIG_XPS
710/*
711 * This structure holds an XPS map which can be of variable length. The
712 * map is an array of queues.
713 */
714struct xps_map {
715 unsigned int len;
716 unsigned int alloc_len;
717 struct rcu_head rcu;
718 u16 queues[0];
719};
60b778ce 720#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
721#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
722 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
723
724/*
725 * This structure holds all XPS maps for device. Maps are indexed by CPU.
726 */
727struct xps_dev_maps {
728 struct rcu_head rcu;
a4177869 729 struct xps_map __rcu *cpu_map[0];
bf264145
TH
730};
731#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
732 (nr_cpu_ids * sizeof(struct xps_map *)))
733#endif /* CONFIG_XPS */
734
4f57c087
JF
735#define TC_MAX_QUEUE 16
736#define TC_BITMASK 15
737/* HW offloaded queuing disciplines txq count and offset maps */
738struct netdev_tc_txq {
739 u16 count;
740 u16 offset;
741};
742
68bad94e
NP
743#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
744/*
745 * This structure is to hold information about the device
746 * configured to run FCoE protocol stack.
747 */
748struct netdev_fcoe_hbainfo {
749 char manufacturer[64];
750 char serial_number[64];
751 char hardware_version[64];
752 char driver_version[64];
753 char optionrom_version[64];
754 char firmware_version[64];
755 char model[256];
756 char model_description[256];
757};
758#endif
759
02637fce 760#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 761
02637fce
JP
762/* This structure holds a unique identifier to identify some
763 * physical item (port for example) used by a netdevice.
66b52b0d 764 */
02637fce
JP
765struct netdev_phys_item_id {
766 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
767 unsigned char id_len;
768};
769
d754f98b
SF
770static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
771 struct netdev_phys_item_id *b)
772{
773 return a->id_len == b->id_len &&
774 memcmp(a->id, b->id, a->id_len) == 0;
775}
776
99932d4f
DB
777typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
778 struct sk_buff *skb);
779
a1b7c5fd 780/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
781 * that are being passed to the netdevice through the setup_tc op.
782 */
783enum {
784 TC_SETUP_MQPRIO,
a1b7c5fd 785 TC_SETUP_CLSU32,
5b33f488 786 TC_SETUP_CLSFLOWER,
16e5cc64
JF
787};
788
a1b7c5fd
JF
789struct tc_cls_u32_offload;
790
16e5cc64
JF
791struct tc_to_netdev {
792 unsigned int type;
793 union {
794 u8 tc;
a1b7c5fd 795 struct tc_cls_u32_offload *cls_u32;
5b33f488 796 struct tc_cls_flower_offload *cls_flower;
16e5cc64
JF
797 };
798};
799
800
d314774c
SH
801/*
802 * This structure defines the management hooks for network devices.
00829823
SH
803 * The following hooks can be defined; unless noted otherwise, they are
804 * optional and can be filled with a null pointer.
d314774c
SH
805 *
806 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
807 * This function is called once when a network device is registered.
808 * The network device can use this for any late stage initialization
809 * or semantic validation. It can fail with an error code which will
810 * be propagated back to register_netdev.
d314774c
SH
811 *
812 * void (*ndo_uninit)(struct net_device *dev);
813 * This function is called when device is unregistered or when registration
814 * fails. It is not called if init fails.
815 *
816 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 817 * This function is called when a network device transitions to the up
d314774c
SH
818 * state.
819 *
820 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 821 * This function is called when a network device transitions to the down
d314774c
SH
822 * state.
823 *
dc1f8bf6
SH
824 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
825 * struct net_device *dev);
00829823 826 * Called when a packet needs to be transmitted.
e79d8429
RR
827 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
828 * the queue before that can happen; it's for obsolete devices and weird
829 * corner cases, but the stack really does a non-trivial amount
830 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 831 * Required; cannot be NULL.
00829823 832 *
cdba756f
ED
833 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
834 * netdev_features_t features);
835 * Adjusts the requested feature flags according to device-specific
836 * constraints, and returns the resulting flags. Must not modify
837 * the device state.
838 *
f663dd9a 839 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 840 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 841 * Called to decide which queue to use when device supports multiple
00829823
SH
842 * transmit queues.
843 *
d314774c
SH
844 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
845 * This function is called to allow device receiver to make
5e82b4b2 846 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
847 *
848 * void (*ndo_set_rx_mode)(struct net_device *dev);
849 * This function is called device changes address list filtering.
01789349 850 * If driver handles unicast address filtering, it should set
5e82b4b2 851 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
852 *
853 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
854 * This function is called when the Media Access Control address
37b607c5 855 * needs to be changed. If this interface is not defined, the
5e82b4b2 856 * MAC address can not be changed.
d314774c
SH
857 *
858 * int (*ndo_validate_addr)(struct net_device *dev);
859 * Test if Media Access Control address is valid for the device.
860 *
861 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
862 * Called when a user requests an ioctl which can't be handled by
863 * the generic interface code. If not defined ioctls return
d314774c
SH
864 * not supported error code.
865 *
866 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
867 * Used to set network devices bus interface parameters. This interface
5e82b4b2 868 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
869 * interface (PCI) for low level management.
870 *
871 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
872 * Called when a user wants to change the Maximum Transfer Unit
873 * of a device. If not defined, any request to change MTU will
874 * will return an error.
875 *
00829823 876 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 877 * Callback used when the transmitter has not made any progress
d314774c
SH
878 * for dev->watchdog ticks.
879 *
3cfde79c 880 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 881 * struct rtnl_link_stats64 *storage);
d308e38f 882 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 883 * Called when a user wants to get the network device usage
be1f3c2c 884 * statistics. Drivers must do one of the following:
3cfde79c
BH
885 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
886 * rtnl_link_stats64 structure passed by the caller.
82695d9b 887 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
888 * (which should normally be dev->stats) and return a pointer to
889 * it. The structure may be changed asynchronously only if each
890 * field is written atomically.
891 * 3. Update dev->stats asynchronously and atomically, and define
892 * neither operation.
d314774c 893 *
5d632cb7 894 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 895 * If device supports VLAN filtering this function is called when a
80d5c368 896 * VLAN id is registered.
d314774c 897 *
5d632cb7 898 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 899 * If device supports VLAN filtering this function is called when a
80d5c368 900 * VLAN id is unregistered.
d314774c
SH
901 *
902 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
903 *
904 * SR-IOV management functions.
905 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
906 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
907 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
908 * int max_tx_rate);
5f8444a3 909 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 910 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
911 * int (*ndo_get_vf_config)(struct net_device *dev,
912 * int vf, struct ifla_vf_info *ivf);
1d8faf48 913 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
914 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
915 * struct nlattr *port[]);
01a3d796
VZ
916 *
917 * Enable or disable the VF ability to query its RSS Redirection Table and
918 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 919 * with PF and querying it may introduce a theoretical security risk.
01a3d796 920 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 921 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
922 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
923 * Called to setup 'tc' number of traffic classes in the net device. This
924 * is always called from the stack with the rtnl lock held and netif tx
925 * queues stopped. This allows the netdevice to perform queue management
926 * safely.
c445477d 927 *
e9bce845
YZ
928 * Fiber Channel over Ethernet (FCoE) offload functions.
929 * int (*ndo_fcoe_enable)(struct net_device *dev);
930 * Called when the FCoE protocol stack wants to start using LLD for FCoE
931 * so the underlying device can perform whatever needed configuration or
932 * initialization to support acceleration of FCoE traffic.
933 *
934 * int (*ndo_fcoe_disable)(struct net_device *dev);
935 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
936 * so the underlying device can perform whatever needed clean-ups to
937 * stop supporting acceleration of FCoE traffic.
938 *
939 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
940 * struct scatterlist *sgl, unsigned int sgc);
941 * Called when the FCoE Initiator wants to initialize an I/O that
942 * is a possible candidate for Direct Data Placement (DDP). The LLD can
943 * perform necessary setup and returns 1 to indicate the device is set up
944 * successfully to perform DDP on this I/O, otherwise this returns 0.
945 *
946 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
947 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
948 * indicated by the FC exchange id 'xid', so the underlying device can
949 * clean up and reuse resources for later DDP requests.
950 *
951 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
952 * struct scatterlist *sgl, unsigned int sgc);
953 * Called when the FCoE Target wants to initialize an I/O that
954 * is a possible candidate for Direct Data Placement (DDP). The LLD can
955 * perform necessary setup and returns 1 to indicate the device is set up
956 * successfully to perform DDP on this I/O, otherwise this returns 0.
957 *
68bad94e
NP
958 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
959 * struct netdev_fcoe_hbainfo *hbainfo);
960 * Called when the FCoE Protocol stack wants information on the underlying
961 * device. This information is utilized by the FCoE protocol stack to
962 * register attributes with Fiber Channel management service as per the
963 * FC-GS Fabric Device Management Information(FDMI) specification.
964 *
e9bce845
YZ
965 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
966 * Called when the underlying device wants to override default World Wide
967 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
968 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
969 * protocol stack to use.
970 *
c445477d
BH
971 * RFS acceleration.
972 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
973 * u16 rxq_index, u32 flow_id);
974 * Set hardware filter for RFS. rxq_index is the target queue index;
975 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
976 * Return the filter ID on success, or a negative error code.
fbaec0ea 977 *
8b98a70c 978 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
979 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
980 * Called to make another netdev an underling.
981 *
982 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
983 * Called to release previously enslaved netdev.
5455c699
MM
984 *
985 * Feature/offload setting functions.
c8f44aff 986 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
987 * Called to update device configuration to new features. Passed
988 * feature set might be less than what was returned by ndo_fix_features()).
989 * Must return >0 or -errno if it changed dev->features itself.
990 *
edc7d573 991 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
992 * struct net_device *dev,
f6f6424b 993 * const unsigned char *addr, u16 vid, u16 flags)
77162022 994 * Adds an FDB entry to dev for addr.
1690be63
VY
995 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
996 * struct net_device *dev,
f6f6424b 997 * const unsigned char *addr, u16 vid)
77162022
JF
998 * Deletes the FDB entry from dev coresponding to addr.
999 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
1000 * struct net_device *dev, struct net_device *filter_dev,
1001 * int idx)
77162022
JF
1002 * Used to add FDB entries to dump requests. Implementers should add
1003 * entries to skb and update idx with the number of entries.
e5a55a89 1004 *
ad41faa8
ND
1005 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1006 * u16 flags)
e5a55a89 1007 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1008 * struct net_device *dev, u32 filter_mask,
1009 * int nlflags)
ad41faa8
ND
1010 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1011 * u16 flags);
4bf84c35
JP
1012 *
1013 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1014 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1015 * which do not represent real hardware may define this to allow their
1016 * userspace components to manage their virtual carrier state. Devices
1017 * that determine carrier state from physical hardware properties (eg
1018 * network cables) or protocol-dependent mechanisms (eg
1019 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1020 *
1021 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1022 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1023 * Called to get ID of physical port of this device. If driver does
1024 * not implement this, it is assumed that the hw is not able to have
1025 * multiple net devices on single physical port.
53cf5275
JG
1026 *
1027 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 1028 * sa_family_t sa_family, __be16 port);
5e82b4b2
BH
1029 * Called by vxlan to notify a driver about the UDP port and socket
1030 * address family that vxlan is listening to. It is called only when
53cf5275
JG
1031 * a new port starts listening. The operation is protected by the
1032 * vxlan_net->sock_lock.
1033 *
a8170d2b 1034 * void (*ndo_add_geneve_port)(struct net_device *dev,
5e82b4b2 1035 * sa_family_t sa_family, __be16 port);
a8170d2b
SA
1036 * Called by geneve to notify a driver about the UDP port and socket
1037 * address family that geneve is listnening to. It is called only when
1038 * a new port starts listening. The operation is protected by the
1039 * geneve_net->sock_lock.
1040 *
1041 * void (*ndo_del_geneve_port)(struct net_device *dev,
5e82b4b2 1042 * sa_family_t sa_family, __be16 port);
a8170d2b
SA
1043 * Called by geneve to notify the driver about a UDP port and socket
1044 * address family that geneve is not listening to anymore. The operation
1045 * is protected by the geneve_net->sock_lock.
1046 *
53cf5275 1047 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 1048 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1049 * Called by vxlan to notify the driver about a UDP port and socket
1050 * address family that vxlan is not listening to anymore. The operation
1051 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
1052 *
1053 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1054 * struct net_device *dev)
1055 * Called by upper layer devices to accelerate switching or other
1056 * station functionality into hardware. 'pdev is the lowerdev
1057 * to use for the offload and 'dev' is the net device that will
1058 * back the offload. Returns a pointer to the private structure
1059 * the upper layer will maintain.
1060 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1061 * Called by upper layer device to delete the station created
1062 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1063 * the station and priv is the structure returned by the add
1064 * operation.
1065 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1066 * struct net_device *dev,
1067 * void *priv);
1068 * Callback to use for xmit over the accelerated station. This
1069 * is used in place of ndo_start_xmit on accelerated net
1070 * devices.
5e82b4b2
BH
1071 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1072 * struct net_device *dev
1073 * netdev_features_t features);
04ffcb25 1074 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1075 * performing offload operations on a given packet. This is to give
1076 * the device an opportunity to implement any restrictions that cannot
1077 * be otherwise expressed by feature flags. The check is called with
1078 * the set of features that the stack has calculated and it returns
1079 * those the driver believes to be appropriate.
822b3b2e
JF
1080 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1081 * int queue_index, u32 maxrate);
1082 * Called when a user wants to set a max-rate limitation of specific
1083 * TX queue.
a54acb3a
ND
1084 * int (*ndo_get_iflink)(const struct net_device *dev);
1085 * Called to get the iflink value of this device.
d746d707 1086 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1087 * bool proto_down);
d746d707
AK
1088 * This function is used to pass protocol port error state information
1089 * to the switch driver. The switch driver can react to the proto_down
1090 * by doing a phys down on the associated switch port.
fc4099f1
PS
1091 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1092 * This function is used to get egress tunnel information for given skb.
1093 * This is useful for retrieving outer tunnel header parameters while
1094 * sampling packet.
871b642a
PA
1095 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1096 * This function is used to specify the headroom that the skb must
1097 * consider when allocation skb during packet reception. Setting
1098 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1099 * forward. Setting a negative value resets the rx headroom to the
871b642a 1100 * default value.
d746d707 1101 *
d314774c
SH
1102 */
1103struct net_device_ops {
1104 int (*ndo_init)(struct net_device *dev);
1105 void (*ndo_uninit)(struct net_device *dev);
1106 int (*ndo_open)(struct net_device *dev);
1107 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1108 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1109 struct net_device *dev);
1110 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1111 struct net_device *dev,
1112 netdev_features_t features);
00829823 1113 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1114 struct sk_buff *skb,
99932d4f
DB
1115 void *accel_priv,
1116 select_queue_fallback_t fallback);
d314774c
SH
1117 void (*ndo_change_rx_flags)(struct net_device *dev,
1118 int flags);
d314774c 1119 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1120 int (*ndo_set_mac_address)(struct net_device *dev,
1121 void *addr);
d314774c 1122 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1123 int (*ndo_do_ioctl)(struct net_device *dev,
1124 struct ifreq *ifr, int cmd);
d314774c
SH
1125 int (*ndo_set_config)(struct net_device *dev,
1126 struct ifmap *map);
00829823
SH
1127 int (*ndo_change_mtu)(struct net_device *dev,
1128 int new_mtu);
1129 int (*ndo_neigh_setup)(struct net_device *dev,
1130 struct neigh_parms *);
d314774c
SH
1131 void (*ndo_tx_timeout) (struct net_device *dev);
1132
28172739
ED
1133 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1134 struct rtnl_link_stats64 *storage);
d314774c
SH
1135 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1136
8e586137 1137 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1138 __be16 proto, u16 vid);
8e586137 1139 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1140 __be16 proto, u16 vid);
d314774c 1141#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1142 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1143 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1144 struct netpoll_info *info);
0e34e931 1145 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1146#endif
e0d1095a 1147#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1148 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1149#endif
95c26df8
WM
1150 int (*ndo_set_vf_mac)(struct net_device *dev,
1151 int queue, u8 *mac);
1152 int (*ndo_set_vf_vlan)(struct net_device *dev,
1153 int queue, u16 vlan, u8 qos);
ed616689
SC
1154 int (*ndo_set_vf_rate)(struct net_device *dev,
1155 int vf, int min_tx_rate,
1156 int max_tx_rate);
5f8444a3
GR
1157 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1158 int vf, bool setting);
dd461d6a
HS
1159 int (*ndo_set_vf_trust)(struct net_device *dev,
1160 int vf, bool setting);
95c26df8
WM
1161 int (*ndo_get_vf_config)(struct net_device *dev,
1162 int vf,
1163 struct ifla_vf_info *ivf);
1d8faf48
RE
1164 int (*ndo_set_vf_link_state)(struct net_device *dev,
1165 int vf, int link_state);
3b766cd8
EBE
1166 int (*ndo_get_vf_stats)(struct net_device *dev,
1167 int vf,
1168 struct ifla_vf_stats
1169 *vf_stats);
57b61080
SF
1170 int (*ndo_set_vf_port)(struct net_device *dev,
1171 int vf,
1172 struct nlattr *port[]);
1173 int (*ndo_get_vf_port)(struct net_device *dev,
1174 int vf, struct sk_buff *skb);
cc8e27cc
EC
1175 int (*ndo_set_vf_guid)(struct net_device *dev,
1176 int vf, u64 guid,
1177 int guid_type);
01a3d796
VZ
1178 int (*ndo_set_vf_rss_query_en)(
1179 struct net_device *dev,
1180 int vf, bool setting);
16e5cc64
JF
1181 int (*ndo_setup_tc)(struct net_device *dev,
1182 u32 handle,
1183 __be16 protocol,
1184 struct tc_to_netdev *tc);
d11ead75 1185#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1186 int (*ndo_fcoe_enable)(struct net_device *dev);
1187 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1188 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1189 u16 xid,
1190 struct scatterlist *sgl,
1191 unsigned int sgc);
1192 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1193 u16 xid);
6247e086
YZ
1194 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1195 u16 xid,
1196 struct scatterlist *sgl,
1197 unsigned int sgc);
68bad94e
NP
1198 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1199 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1200#endif
1201
d11ead75 1202#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1203#define NETDEV_FCOE_WWNN 0
1204#define NETDEV_FCOE_WWPN 1
1205 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1206 u64 *wwn, int type);
4d288d57 1207#endif
3c9c36bc 1208
c445477d
BH
1209#ifdef CONFIG_RFS_ACCEL
1210 int (*ndo_rx_flow_steer)(struct net_device *dev,
1211 const struct sk_buff *skb,
1212 u16 rxq_index,
1213 u32 flow_id);
1214#endif
fbaec0ea
JP
1215 int (*ndo_add_slave)(struct net_device *dev,
1216 struct net_device *slave_dev);
1217 int (*ndo_del_slave)(struct net_device *dev,
1218 struct net_device *slave_dev);
c8f44aff
MM
1219 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1220 netdev_features_t features);
5455c699 1221 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1222 netdev_features_t features);
da6a8fa0 1223 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1224 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1225
1226 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1227 struct nlattr *tb[],
77162022 1228 struct net_device *dev,
6b6e2725 1229 const unsigned char *addr,
f6f6424b 1230 u16 vid,
77162022
JF
1231 u16 flags);
1232 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1233 struct nlattr *tb[],
77162022 1234 struct net_device *dev,
f6f6424b
JP
1235 const unsigned char *addr,
1236 u16 vid);
77162022
JF
1237 int (*ndo_fdb_dump)(struct sk_buff *skb,
1238 struct netlink_callback *cb,
1239 struct net_device *dev,
5d5eacb3 1240 struct net_device *filter_dev,
77162022 1241 int idx);
e5a55a89
JF
1242
1243 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1244 struct nlmsghdr *nlh,
1245 u16 flags);
e5a55a89
JF
1246 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1247 u32 pid, u32 seq,
6cbdceeb 1248 struct net_device *dev,
46c264da
ND
1249 u32 filter_mask,
1250 int nlflags);
407af329 1251 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1252 struct nlmsghdr *nlh,
1253 u16 flags);
4bf84c35
JP
1254 int (*ndo_change_carrier)(struct net_device *dev,
1255 bool new_carrier);
66b52b0d 1256 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1257 struct netdev_phys_item_id *ppid);
db24a904
DA
1258 int (*ndo_get_phys_port_name)(struct net_device *dev,
1259 char *name, size_t len);
53cf5275
JG
1260 void (*ndo_add_vxlan_port)(struct net_device *dev,
1261 sa_family_t sa_family,
35e42379 1262 __be16 port);
53cf5275
JG
1263 void (*ndo_del_vxlan_port)(struct net_device *dev,
1264 sa_family_t sa_family,
35e42379 1265 __be16 port);
a8170d2b
SA
1266 void (*ndo_add_geneve_port)(struct net_device *dev,
1267 sa_family_t sa_family,
1268 __be16 port);
1269 void (*ndo_del_geneve_port)(struct net_device *dev,
1270 sa_family_t sa_family,
1271 __be16 port);
a6cc0cfa
JF
1272 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1273 struct net_device *dev);
1274 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1275 void *priv);
1276
1277 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1278 struct net_device *dev,
1279 void *priv);
25175ba5 1280 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1281 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1282 int queue_index,
1283 u32 maxrate);
a54acb3a 1284 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1285 int (*ndo_change_proto_down)(struct net_device *dev,
1286 bool proto_down);
fc4099f1
PS
1287 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1288 struct sk_buff *skb);
871b642a
PA
1289 void (*ndo_set_rx_headroom)(struct net_device *dev,
1290 int needed_headroom);
d314774c
SH
1291};
1292
7aa98047
LR
1293/**
1294 * enum net_device_priv_flags - &struct net_device priv_flags
1295 *
1296 * These are the &struct net_device, they are only set internally
1297 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1298 * userspace; this means that the order of these flags can change
7aa98047
LR
1299 * during any kernel release.
1300 *
1301 * You should have a pretty good reason to be extending these flags.
1302 *
1303 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1304 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1305 * @IFF_BONDING: bonding master or slave
7aa98047 1306 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1307 * @IFF_WAN_HDLC: WAN HDLC device
1308 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1309 * release skb->dst
1310 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1311 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1312 * @IFF_MACVLAN_PORT: device used as macvlan port
1313 * @IFF_BRIDGE_PORT: device used as bridge port
1314 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1315 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1316 * @IFF_UNICAST_FLT: Supports unicast filtering
1317 * @IFF_TEAM_PORT: device used as team port
1318 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1319 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1320 * change when it's running
1321 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1322 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1323 * underlying stacked devices
1324 * @IFF_IPVLAN_MASTER: IPvlan master device
1325 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1326 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1327 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1328 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1329 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1330 * @IFF_TEAM: device is a team device
d4ab4286 1331 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1332 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1333 * entity (i.e. the master device for bridged veth)
3c175784 1334 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1335 */
1336enum netdev_priv_flags {
1337 IFF_802_1Q_VLAN = 1<<0,
1338 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1339 IFF_BONDING = 1<<2,
1340 IFF_ISATAP = 1<<3,
1341 IFF_WAN_HDLC = 1<<4,
1342 IFF_XMIT_DST_RELEASE = 1<<5,
1343 IFF_DONT_BRIDGE = 1<<6,
1344 IFF_DISABLE_NETPOLL = 1<<7,
1345 IFF_MACVLAN_PORT = 1<<8,
1346 IFF_BRIDGE_PORT = 1<<9,
1347 IFF_OVS_DATAPATH = 1<<10,
1348 IFF_TX_SKB_SHARING = 1<<11,
1349 IFF_UNICAST_FLT = 1<<12,
1350 IFF_TEAM_PORT = 1<<13,
1351 IFF_SUPP_NOFCS = 1<<14,
1352 IFF_LIVE_ADDR_CHANGE = 1<<15,
1353 IFF_MACVLAN = 1<<16,
1354 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1355 IFF_IPVLAN_MASTER = 1<<18,
1356 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1357 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1358 IFF_NO_QUEUE = 1<<21,
1359 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1360 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1361 IFF_TEAM = 1<<24,
d4ab4286 1362 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1363 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1364 IFF_MACSEC = 1<<27,
7aa98047
LR
1365};
1366
1367#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1368#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1369#define IFF_BONDING IFF_BONDING
7aa98047 1370#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1371#define IFF_WAN_HDLC IFF_WAN_HDLC
1372#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1373#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1374#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1375#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1376#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1377#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1378#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1379#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1380#define IFF_TEAM_PORT IFF_TEAM_PORT
1381#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1382#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1383#define IFF_MACVLAN IFF_MACVLAN
02875878 1384#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1385#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1386#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1387#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1388#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1389#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1390#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1391#define IFF_TEAM IFF_TEAM
d4ab4286 1392#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1393#define IFF_MACSEC IFF_MACSEC
7aa98047 1394
536721b1
KK
1395/**
1396 * struct net_device - The DEVICE structure.
1397 * Actually, this whole structure is a big mistake. It mixes I/O
1398 * data with strictly "high-level" data, and it has to know about
1399 * almost every data structure used in the INET module.
1400 *
1401 * @name: This is the first field of the "visible" part of this structure
1402 * (i.e. as seen by users in the "Space.c" file). It is the name
1403 * of the interface.
1404 *
1405 * @name_hlist: Device name hash chain, please keep it close to name[]
1406 * @ifalias: SNMP alias
1407 * @mem_end: Shared memory end
1408 * @mem_start: Shared memory start
1409 * @base_addr: Device I/O address
1410 * @irq: Device IRQ number
1411 *
14ffbbb8
TG
1412 * @carrier_changes: Stats to monitor carrier on<->off transitions
1413 *
536721b1
KK
1414 * @state: Generic network queuing layer state, see netdev_state_t
1415 * @dev_list: The global list of network devices
5e82b4b2
BH
1416 * @napi_list: List entry used for polling NAPI devices
1417 * @unreg_list: List entry when we are unregistering the
1418 * device; see the function unregister_netdev
1419 * @close_list: List entry used when we are closing the device
62d885fe
BP
1420 * @ptype_all: Device-specific packet handlers for all protocols
1421 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1422 *
1423 * @adj_list: Directly linked devices, like slaves for bonding
1424 * @all_adj_list: All linked devices, *including* neighbours
1425 * @features: Currently active device features
1426 * @hw_features: User-changeable features
1427 *
1428 * @wanted_features: User-requested features
1429 * @vlan_features: Mask of features inheritable by VLAN devices
1430 *
1431 * @hw_enc_features: Mask of features inherited by encapsulating devices
1432 * This field indicates what encapsulation
1433 * offloads the hardware is capable of doing,
1434 * and drivers will need to set them appropriately.
1435 *
1436 * @mpls_features: Mask of features inheritable by MPLS
1437 *
1438 * @ifindex: interface index
5e82b4b2 1439 * @group: The group the device belongs to
536721b1
KK
1440 *
1441 * @stats: Statistics struct, which was left as a legacy, use
1442 * rtnl_link_stats64 instead
1443 *
1444 * @rx_dropped: Dropped packets by core network,
1445 * do not use this in drivers
1446 * @tx_dropped: Dropped packets by core network,
1447 * do not use this in drivers
6e7333d3
JW
1448 * @rx_nohandler: nohandler dropped packets by core network on
1449 * inactive devices, do not use this in drivers
536721b1 1450 *
536721b1
KK
1451 * @wireless_handlers: List of functions to handle Wireless Extensions,
1452 * instead of ioctl,
1453 * see <net/iw_handler.h> for details.
1454 * @wireless_data: Instance data managed by the core of wireless extensions
1455 *
1456 * @netdev_ops: Includes several pointers to callbacks,
1457 * if one wants to override the ndo_*() functions
1458 * @ethtool_ops: Management operations
f997c55c
AA
1459 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1460 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1461 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1462 * of Layer 2 headers.
1463 *
1464 * @flags: Interface flags (a la BSD)
1465 * @priv_flags: Like 'flags' but invisible to userspace,
1466 * see if.h for the definitions
1467 * @gflags: Global flags ( kept as legacy )
1468 * @padded: How much padding added by alloc_netdev()
1469 * @operstate: RFC2863 operstate
1470 * @link_mode: Mapping policy to operstate
1471 * @if_port: Selectable AUI, TP, ...
1472 * @dma: DMA channel
1473 * @mtu: Interface MTU value
1474 * @type: Interface hardware type
2793a23a 1475 * @hard_header_len: Maximum hardware header length.
536721b1
KK
1476 *
1477 * @needed_headroom: Extra headroom the hardware may need, but not in all
1478 * cases can this be guaranteed
1479 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1480 * cases can this be guaranteed. Some cases also use
1481 * LL_MAX_HEADER instead to allocate the skb
1482 *
1483 * interface address info:
1484 *
1485 * @perm_addr: Permanent hw address
1486 * @addr_assign_type: Hw address assignment type
1487 * @addr_len: Hardware address length
8626a0c8 1488 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
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
f997c55c
AA
1677#if IS_ENABLED(CONFIG_IPV6)
1678 const struct ndisc_ops *ndisc_ops;
1679#endif
1da177e4 1680
3b04ddde
SH
1681 const struct header_ops *header_ops;
1682
536721b1
KK
1683 unsigned int flags;
1684 unsigned int priv_flags;
1685
1da177e4 1686 unsigned short gflags;
536721b1 1687 unsigned short padded;
1da177e4 1688
536721b1
KK
1689 unsigned char operstate;
1690 unsigned char link_mode;
b00055aa 1691
536721b1
KK
1692 unsigned char if_port;
1693 unsigned char dma;
bdc220da 1694
536721b1
KK
1695 unsigned int mtu;
1696 unsigned short type;
1697 unsigned short hard_header_len;
1da177e4 1698
f5184d26
JB
1699 unsigned short needed_headroom;
1700 unsigned short needed_tailroom;
1701
1da177e4 1702 /* Interface address info. */
536721b1
KK
1703 unsigned char perm_addr[MAX_ADDR_LEN];
1704 unsigned char addr_assign_type;
1705 unsigned char addr_len;
a0a9663d 1706 unsigned short neigh_priv_len;
536721b1
KK
1707 unsigned short dev_id;
1708 unsigned short dev_port;
ccffad25 1709 spinlock_t addr_list_lock;
14ffbbb8
TG
1710 unsigned char name_assign_type;
1711 bool uc_promisc;
536721b1
KK
1712 struct netdev_hw_addr_list uc;
1713 struct netdev_hw_addr_list mc;
1714 struct netdev_hw_addr_list dev_addrs;
1715
4c3d5e7b
ED
1716#ifdef CONFIG_SYSFS
1717 struct kset *queues_kset;
1718#endif
9d45abe1
WC
1719 unsigned int promiscuity;
1720 unsigned int allmulti;
1da177e4 1721
1da177e4 1722
5e82b4b2 1723 /* Protocol-specific pointers */
65ac6a5f 1724
d11ead75 1725#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1726 struct vlan_info __rcu *vlan_info;
65ac6a5f 1727#endif
34a430d7 1728#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1729 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1730#endif
1731#if IS_ENABLED(CONFIG_TIPC)
536721b1 1732 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1733#endif
536721b1
KK
1734 void *atalk_ptr;
1735 struct in_device __rcu *ip_ptr;
1736 struct dn_dev __rcu *dn_ptr;
1737 struct inet6_dev __rcu *ip6_ptr;
1738 void *ax25_ptr;
1739 struct wireless_dev *ieee80211_ptr;
98a18b6f 1740 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1741#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1742 struct mpls_dev __rcu *mpls_ptr;
1743#endif
1da177e4 1744
9356b8fc 1745/*
cd13539b 1746 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1747 */
536721b1 1748 unsigned long last_rx;
4dc89133 1749
9356b8fc 1750 /* Interface address info used in eth_type_trans() */
536721b1 1751 unsigned char *dev_addr;
f001fde5 1752
a953be53 1753#ifdef CONFIG_SYSFS
0a9627f2
TH
1754 struct netdev_rx_queue *_rx;
1755
0a9627f2 1756 unsigned int num_rx_queues;
62fe0b40 1757 unsigned int real_num_rx_queues;
df334545 1758#endif
0a9627f2 1759
3b47d303 1760 unsigned long gro_flush_timeout;
61391cde 1761 rx_handler_func_t __rcu *rx_handler;
1762 void __rcu *rx_handler_data;
e8a0464c 1763
4cda01e8 1764#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1765 struct tcf_proto __rcu *ingress_cl_list;
1766#endif
24824a09 1767 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1768#ifdef CONFIG_NETFILTER_INGRESS
1769 struct list_head nf_hooks_ingress;
1770#endif
d2788d34 1771
536721b1 1772 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1773#ifdef CONFIG_RFS_ACCEL
1774 struct cpu_rmap *rx_cpu_rmap;
1775#endif
1776 struct hlist_node index_hlist;
cd13539b
ED
1777
1778/*
1779 * Cache lines mostly used on transmit path
1780 */
e8a0464c
DM
1781 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1782 unsigned int num_tx_queues;
fd2ea0a7 1783 unsigned int real_num_tx_queues;
af356afa 1784 struct Qdisc *qdisc;
536721b1 1785 unsigned long tx_queue_len;
c3f26a26 1786 spinlock_t tx_global_lock;
14ffbbb8 1787 int watchdog_timeo;
cd13539b 1788
bf264145 1789#ifdef CONFIG_XPS
a4177869 1790 struct xps_dev_maps __rcu *xps_maps;
bf264145 1791#endif
1f211a1b
DB
1792#ifdef CONFIG_NET_CLS_ACT
1793 struct tcf_proto __rcu *egress_cl_list;
1794#endif
0c4f691f
SF
1795#ifdef CONFIG_NET_SWITCHDEV
1796 u32 offload_fwd_mark;
1797#endif
1798
9356b8fc 1799 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1800 struct timer_list watchdog_timer;
1801
29b4433d 1802 int __percpu *pcpu_refcnt;
1da177e4 1803 struct list_head todo_list;
1da177e4 1804
e014debe 1805 struct list_head link_watch_list;
572a103d 1806
1da177e4 1807 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1808 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1809 NETREG_UNREGISTERING, /* called unregister_netdevice */
1810 NETREG_UNREGISTERED, /* completed unregister todo */
1811 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1812 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1813 } reg_state:8;
1814
536721b1 1815 bool dismantle;
a2835763
PM
1816
1817 enum {
1818 RTNL_LINK_INITIALIZED,
1819 RTNL_LINK_INITIALIZING,
1820 } rtnl_link_state:16;
1da177e4 1821
d314774c 1822 void (*destructor)(struct net_device *dev);
1da177e4 1823
1da177e4 1824#ifdef CONFIG_NETPOLL
5fbee843 1825 struct netpoll_info __rcu *npinfo;
1da177e4 1826#endif
eae792b7 1827
0c5c9fb5 1828 possible_net_t nd_net;
4a1c5371 1829
4951704b 1830 /* mid-layer private */
a7855c78 1831 union {
536721b1
KK
1832 void *ml_priv;
1833 struct pcpu_lstats __percpu *lstats;
8f84985f 1834 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1835 struct pcpu_dstats __percpu *dstats;
1836 struct pcpu_vstats __percpu *vstats;
a7855c78 1837 };
536721b1 1838
3cc77ec7 1839 struct garp_port __rcu *garp_port;
febf018d 1840 struct mrp_port __rcu *mrp_port;
1da177e4 1841
5e82b4b2 1842 struct device dev;
0c509a6c 1843 const struct attribute_group *sysfs_groups[4];
a953be53 1844 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1845
38f7b870 1846 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1847
82cc1a7a
PWJ
1848 /* for setting kernel sock attribute on TCP connection setup */
1849#define GSO_MAX_SIZE 65536
1850 unsigned int gso_max_size;
30b678d8
BH
1851#define GSO_MAX_SEGS 65535
1852 u16 gso_max_segs;
743b03a8 1853
7a6b6f51 1854#ifdef CONFIG_DCB
32953543 1855 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1856#endif
5e82b4b2
BH
1857 u8 num_tc;
1858 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1859 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1860
d11ead75 1861#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1862 unsigned int fcoe_ddp_xid;
5bc1421e 1863#endif
86f8515f 1864#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1865 struct netprio_map __rcu *priomap;
4d288d57 1866#endif
5e82b4b2
BH
1867 struct phy_device *phydev;
1868 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1869 struct lock_class_key *qdisc_running_key;
5e82b4b2 1870 bool proto_down;
1da177e4 1871};
43cb76d9 1872#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1873
1874#define NETDEV_ALIGN 32
1da177e4 1875
4f57c087
JF
1876static inline
1877int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1878{
1879 return dev->prio_tc_map[prio & TC_BITMASK];
1880}
1881
1882static inline
1883int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1884{
1885 if (tc >= dev->num_tc)
1886 return -EINVAL;
1887
1888 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1889 return 0;
1890}
1891
1892static inline
1893void netdev_reset_tc(struct net_device *dev)
1894{
1895 dev->num_tc = 0;
1896 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1897 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1898}
1899
1900static inline
1901int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1902{
1903 if (tc >= dev->num_tc)
1904 return -EINVAL;
1905
1906 dev->tc_to_txq[tc].count = count;
1907 dev->tc_to_txq[tc].offset = offset;
1908 return 0;
1909}
1910
1911static inline
1912int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1913{
1914 if (num_tc > TC_MAX_QUEUE)
1915 return -EINVAL;
1916
1917 dev->num_tc = num_tc;
1918 return 0;
1919}
1920
1921static inline
1922int netdev_get_num_tc(struct net_device *dev)
1923{
1924 return dev->num_tc;
1925}
1926
e8a0464c
DM
1927static inline
1928struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1929 unsigned int index)
1930{
1931 return &dev->_tx[index];
1932}
1933
10c51b56
DB
1934static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1935 const struct sk_buff *skb)
1936{
1937 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1938}
1939
e8a0464c
DM
1940static inline void netdev_for_each_tx_queue(struct net_device *dev,
1941 void (*f)(struct net_device *,
1942 struct netdev_queue *,
1943 void *),
1944 void *arg)
1945{
1946 unsigned int i;
1947
1948 for (i = 0; i < dev->num_tx_queues; i++)
1949 f(dev, &dev->_tx[i], arg);
1950}
1951
d3fff6c4
ED
1952#define netdev_lockdep_set_classes(dev) \
1953{ \
1954 static struct lock_class_key qdisc_tx_busylock_key; \
1955 static struct lock_class_key qdisc_running_key; \
1956 static struct lock_class_key qdisc_xmit_lock_key; \
1957 static struct lock_class_key dev_addr_list_lock_key; \
1958 unsigned int i; \
1959 \
1960 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1961 (dev)->qdisc_running_key = &qdisc_running_key; \
1962 lockdep_set_class(&(dev)->addr_list_lock, \
1963 &dev_addr_list_lock_key); \
1964 for (i = 0; i < (dev)->num_tx_queues; i++) \
1965 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1966 &qdisc_xmit_lock_key); \
1967}
1968
f629d208 1969struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1970 struct sk_buff *skb,
1971 void *accel_priv);
8c4c49df 1972
871b642a
PA
1973/* returns the headroom that the master device needs to take in account
1974 * when forwarding to this dev
1975 */
1976static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1977{
1978 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1979}
1980
1981static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1982{
1983 if (dev->netdev_ops->ndo_set_rx_headroom)
1984 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1985}
1986
1987/* set the device rx headroom to the dev's default */
1988static inline void netdev_reset_rx_headroom(struct net_device *dev)
1989{
1990 netdev_set_rx_headroom(dev, -1);
1991}
1992
c346dca1
YH
1993/*
1994 * Net namespace inlines
1995 */
1996static inline
1997struct net *dev_net(const struct net_device *dev)
1998{
c2d9ba9b 1999 return read_pnet(&dev->nd_net);
c346dca1
YH
2000}
2001
2002static inline
f5aa23fd 2003void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2004{
0c5c9fb5 2005 write_pnet(&dev->nd_net, net);
c346dca1
YH
2006}
2007
3e8a72d1 2008static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 2009{
3fc88677 2010#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
2011 if (dev->dsa_ptr != NULL)
2012 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 2013#endif
5aed85ce 2014 return false;
396138f0
LB
2015}
2016
bea3348e
SH
2017/**
2018 * netdev_priv - access network device private data
2019 * @dev: network device
2020 *
2021 * Get network device private data
2022 */
6472ce60 2023static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2024{
1ce8e7b5 2025 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2026}
2027
1da177e4
LT
2028/* Set the sysfs physical device reference for the network logical device
2029 * if set prior to registration will cause a symlink during initialization.
2030 */
43cb76d9 2031#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2032
384912ed 2033/* Set the sysfs device type for the network logical device to allow
3f79410c 2034 * fine-grained identification of different network device types. For
5e82b4b2 2035 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2036 */
2037#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2038
82dc3c63
ED
2039/* Default NAPI poll() weight
2040 * Device drivers are strongly advised to not use bigger value
2041 */
2042#define NAPI_POLL_WEIGHT 64
2043
3b582cc1 2044/**
5e82b4b2 2045 * netif_napi_add - initialize a NAPI context
3b582cc1 2046 * @dev: network device
5e82b4b2 2047 * @napi: NAPI context
3b582cc1
SH
2048 * @poll: polling function
2049 * @weight: default weight
2050 *
5e82b4b2
BH
2051 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2052 * *any* of the other NAPI-related functions.
3b582cc1 2053 */
d565b0a1
HX
2054void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2055 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2056
d64b5e85 2057/**
5e82b4b2 2058 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2059 * @dev: network device
5e82b4b2 2060 * @napi: NAPI context
d64b5e85
ED
2061 * @poll: polling function
2062 * @weight: default weight
2063 *
2064 * This variant of netif_napi_add() should be used from drivers using NAPI
2065 * to exclusively poll a TX queue.
2066 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2067 */
2068static inline void netif_tx_napi_add(struct net_device *dev,
2069 struct napi_struct *napi,
2070 int (*poll)(struct napi_struct *, int),
2071 int weight)
2072{
2073 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2074 netif_napi_add(dev, napi, poll, weight);
2075}
2076
d8156534 2077/**
5e82b4b2
BH
2078 * netif_napi_del - remove a NAPI context
2079 * @napi: NAPI context
d8156534 2080 *
5e82b4b2 2081 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2082 */
d565b0a1
HX
2083void netif_napi_del(struct napi_struct *napi);
2084
2085struct napi_gro_cb {
78a478d0 2086 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2087 void *frag0;
78a478d0 2088
7489594c
HX
2089 /* Length of frag0. */
2090 unsigned int frag0_len;
2091
86911732 2092 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2093 int data_offset;
86911732 2094
d565b0a1 2095 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2096 u16 flush;
2097
2098 /* Save the IP ID here and check when we get to the transport layer */
2099 u16 flush_id;
d565b0a1
HX
2100
2101 /* Number of segments aggregated. */
2e71a6f8
ED
2102 u16 count;
2103
15e2396d
TH
2104 /* Start offset for remote checksum offload */
2105 u16 gro_remcsum_start;
2106
2e71a6f8
ED
2107 /* jiffies when first packet was created/queued */
2108 unsigned long age;
86347245 2109
afe93325 2110 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2111 u16 proto;
2112
baa32ff4
TH
2113 /* This is non-zero if the packet may be of the same flow. */
2114 u8 same_flow:1;
2115
fac8e0f5
JG
2116 /* Used in tunnel GRO receive */
2117 u8 encap_mark:1;
573e8fca
TH
2118
2119 /* GRO checksum is valid */
2120 u8 csum_valid:1;
2121
662880f4
TH
2122 /* Number of checksums via CHECKSUM_UNNECESSARY */
2123 u8 csum_cnt:3;
c3c7c254 2124
baa32ff4
TH
2125 /* Free the skb? */
2126 u8 free:2;
2127#define NAPI_GRO_FREE 1
2128#define NAPI_GRO_FREE_STOLEN_HEAD 2
2129
efc98d08
TH
2130 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2131 u8 is_ipv6:1;
2132
a0ca153f
AD
2133 /* Used in GRE, set in fou/gue_gro_receive */
2134 u8 is_fou:1;
2135
1530545e
AD
2136 /* Used to determine if flush_id can be ignored */
2137 u8 is_atomic:1;
2138
2139 /* 5 bit hole */
baa32ff4 2140
bf5a755f
JC
2141 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2142 __wsum csum;
2143
c3c7c254
ED
2144 /* used in skb_gro_receive() slow path */
2145 struct sk_buff *last;
d565b0a1
HX
2146};
2147
2148#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2149
1da177e4 2150struct packet_type {
f2ccd8fa
DM
2151 __be16 type; /* This is really htons(ether_type). */
2152 struct net_device *dev; /* NULL is wildcarded here */
2153 int (*func) (struct sk_buff *,
2154 struct net_device *,
2155 struct packet_type *,
2156 struct net_device *);
c0de08d0
EL
2157 bool (*id_match)(struct packet_type *ptype,
2158 struct sock *sk);
1da177e4
LT
2159 void *af_packet_priv;
2160 struct list_head list;
2161};
2162
f191a1d1 2163struct offload_callbacks {
576a30eb 2164 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2165 netdev_features_t features);
d565b0a1 2166 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2167 struct sk_buff *skb);
299603e8 2168 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2169};
2170
2171struct packet_offload {
2172 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2173 u16 priority;
f191a1d1
VY
2174 struct offload_callbacks callbacks;
2175 struct list_head list;
1da177e4
LT
2176};
2177
5e82b4b2 2178/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2179struct pcpu_sw_netstats {
2180 u64 rx_packets;
2181 u64 rx_bytes;
2182 u64 tx_packets;
2183 u64 tx_bytes;
2184 struct u64_stats_sync syncp;
2185};
2186
aabc92bb
PNA
2187#define __netdev_alloc_pcpu_stats(type, gfp) \
2188({ \
2189 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2190 if (pcpu_stats) { \
2191 int __cpu; \
2192 for_each_possible_cpu(__cpu) { \
2193 typeof(type) *stat; \
2194 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2195 u64_stats_init(&stat->syncp); \
2196 } \
2197 } \
2198 pcpu_stats; \
1c213bd2
WC
2199})
2200
aabc92bb 2201#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2202 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2203
764f5e54
JP
2204enum netdev_lag_tx_type {
2205 NETDEV_LAG_TX_TYPE_UNKNOWN,
2206 NETDEV_LAG_TX_TYPE_RANDOM,
2207 NETDEV_LAG_TX_TYPE_BROADCAST,
2208 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2209 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2210 NETDEV_LAG_TX_TYPE_HASH,
2211};
2212
2213struct netdev_lag_upper_info {
2214 enum netdev_lag_tx_type tx_type;
2215};
2216
fb1b2e3c
JP
2217struct netdev_lag_lower_state_info {
2218 u8 link_up : 1,
2219 tx_enabled : 1;
2220};
2221
1da177e4
LT
2222#include <linux/notifier.h>
2223
dcfe1421
AW
2224/* netdevice notifier chain. Please remember to update the rtnetlink
2225 * notification exclusion list in rtnetlink_event() when adding new
2226 * types.
2227 */
2228#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2229#define NETDEV_DOWN 0x0002
2230#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2231 detected a hardware crash and restarted
2232 - we can use this eg to kick tcp sessions
2233 once done */
2234#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2235#define NETDEV_REGISTER 0x0005
2236#define NETDEV_UNREGISTER 0x0006
1d486bfb 2237#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2238#define NETDEV_CHANGEADDR 0x0008
2239#define NETDEV_GOING_DOWN 0x0009
2240#define NETDEV_CHANGENAME 0x000A
2241#define NETDEV_FEAT_CHANGE 0x000B
2242#define NETDEV_BONDING_FAILOVER 0x000C
2243#define NETDEV_PRE_UP 0x000D
2244#define NETDEV_PRE_TYPE_CHANGE 0x000E
2245#define NETDEV_POST_TYPE_CHANGE 0x000F
2246#define NETDEV_POST_INIT 0x0010
0115e8e3 2247#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2248#define NETDEV_RELEASE 0x0012
2249#define NETDEV_NOTIFY_PEERS 0x0013
2250#define NETDEV_JOIN 0x0014
42e52bf9 2251#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2252#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2253#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2254#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2255#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2256#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2257#define NETDEV_CHANGELOWERSTATE 0x001B
b7aade15 2258#define NETDEV_OFFLOAD_PUSH_VXLAN 0x001C
681e683f 2259#define NETDEV_OFFLOAD_PUSH_GENEVE 0x001D
dcfe1421 2260
f629d208
JP
2261int register_netdevice_notifier(struct notifier_block *nb);
2262int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2263
2264struct netdev_notifier_info {
2265 struct net_device *dev;
2266};
2267
be9efd36
JP
2268struct netdev_notifier_change_info {
2269 struct netdev_notifier_info info; /* must be first */
2270 unsigned int flags_changed;
2271};
2272
0e4ead9d
JP
2273struct netdev_notifier_changeupper_info {
2274 struct netdev_notifier_info info; /* must be first */
2275 struct net_device *upper_dev; /* new upper dev */
2276 bool master; /* is upper dev master */
5e82b4b2 2277 bool linking; /* is the notification for link or unlink */
29bf24af 2278 void *upper_info; /* upper dev info */
0e4ead9d
JP
2279};
2280
04d48266
JP
2281struct netdev_notifier_changelowerstate_info {
2282 struct netdev_notifier_info info; /* must be first */
2283 void *lower_state_info; /* is lower dev state */
2284};
2285
75538c2b
CW
2286static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2287 struct net_device *dev)
2288{
2289 info->dev = dev;
2290}
2291
351638e7
JP
2292static inline struct net_device *
2293netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2294{
2295 return info->dev;
2296}
2297
f629d208 2298int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2299
2300
1da177e4
LT
2301extern rwlock_t dev_base_lock; /* Device list lock */
2302
881d966b
EB
2303#define for_each_netdev(net, d) \
2304 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2305#define for_each_netdev_reverse(net, d) \
2306 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2307#define for_each_netdev_rcu(net, d) \
2308 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2309#define for_each_netdev_safe(net, d, n) \
2310 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2311#define for_each_netdev_continue(net, d) \
2312 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2313#define for_each_netdev_continue_rcu(net, d) \
2314 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2315#define for_each_netdev_in_bond_rcu(bond, slave) \
2316 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2317 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2318#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2319
a050c33f
DL
2320static inline struct net_device *next_net_device(struct net_device *dev)
2321{
2322 struct list_head *lh;
2323 struct net *net;
2324
c346dca1 2325 net = dev_net(dev);
a050c33f
DL
2326 lh = dev->dev_list.next;
2327 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2328}
2329
ce81b76a
ED
2330static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2331{
2332 struct list_head *lh;
2333 struct net *net;
2334
2335 net = dev_net(dev);
ccf43438 2336 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2337 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2338}
2339
a050c33f
DL
2340static inline struct net_device *first_net_device(struct net *net)
2341{
2342 return list_empty(&net->dev_base_head) ? NULL :
2343 net_device_entry(net->dev_base_head.next);
2344}
7562f876 2345
ccf43438
ED
2346static inline struct net_device *first_net_device_rcu(struct net *net)
2347{
2348 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2349
2350 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2351}
2352
f629d208
JP
2353int netdev_boot_setup_check(struct net_device *dev);
2354unsigned long netdev_boot_base(const char *prefix, int unit);
2355struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2356 const char *hwaddr);
2357struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2358struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2359void dev_add_pack(struct packet_type *pt);
2360void dev_remove_pack(struct packet_type *pt);
2361void __dev_remove_pack(struct packet_type *pt);
2362void dev_add_offload(struct packet_offload *po);
2363void dev_remove_offload(struct packet_offload *po);
f629d208 2364
a54acb3a 2365int dev_get_iflink(const struct net_device *dev);
fc4099f1 2366int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2367struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2368 unsigned short mask);
f629d208
JP
2369struct net_device *dev_get_by_name(struct net *net, const char *name);
2370struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2371struct net_device *__dev_get_by_name(struct net *net, const char *name);
2372int dev_alloc_name(struct net_device *dev, const char *name);
2373int dev_open(struct net_device *dev);
2374int dev_close(struct net_device *dev);
99c4a26a 2375int dev_close_many(struct list_head *head, bool unlink);
f629d208 2376void dev_disable_lro(struct net_device *dev);
0c4b51f0 2377int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2378int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2379int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2380int register_netdevice(struct net_device *dev);
2381void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2382void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2383static inline void unregister_netdevice(struct net_device *dev)
2384{
2385 unregister_netdevice_queue(dev, NULL);
2386}
2387
f629d208
JP
2388int netdev_refcnt_read(const struct net_device *dev);
2389void free_netdev(struct net_device *dev);
74d332c1 2390void netdev_freemem(struct net_device *dev);
f629d208
JP
2391void synchronize_net(void);
2392int init_dummy_netdev(struct net_device *dev);
937f1ba5 2393
f60e5990 2394DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2395#define XMIT_RECURSION_LIMIT 10
2396
f60e5990 2397static inline int dev_recursion_level(void)
2398{
2399 return this_cpu_read(xmit_recursion);
2400}
2401
f629d208
JP
2402struct net_device *dev_get_by_index(struct net *net, int ifindex);
2403struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2404struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2405int netdev_get_name(struct net *net, char *name, int ifindex);
2406int dev_restart(struct net_device *dev);
f629d208 2407int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2408
2409static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2410{
2411 return NAPI_GRO_CB(skb)->data_offset;
2412}
2413
2414static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2415{
2416 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2417}
2418
2419static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2420{
2421 NAPI_GRO_CB(skb)->data_offset += len;
2422}
2423
a5b1cf28
HX
2424static inline void *skb_gro_header_fast(struct sk_buff *skb,
2425 unsigned int offset)
86911732 2426{
a5b1cf28
HX
2427 return NAPI_GRO_CB(skb)->frag0 + offset;
2428}
78a478d0 2429
a5b1cf28
HX
2430static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2431{
2432 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2433}
78a478d0 2434
a5b1cf28
HX
2435static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2436 unsigned int offset)
2437{
17dd759c
HX
2438 if (!pskb_may_pull(skb, hlen))
2439 return NULL;
2440
a5b1cf28
HX
2441 NAPI_GRO_CB(skb)->frag0 = NULL;
2442 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2443 return skb->data + offset;
86911732 2444}
1da177e4 2445
36e7b1b8
HX
2446static inline void *skb_gro_network_header(struct sk_buff *skb)
2447{
78d3fd0b
HX
2448 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2449 skb_network_offset(skb);
36e7b1b8
HX
2450}
2451
bf5a755f
JC
2452static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2453 const void *start, unsigned int len)
2454{
573e8fca 2455 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2456 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2457 csum_partial(start, len, 0));
2458}
2459
573e8fca
TH
2460/* GRO checksum functions. These are logical equivalents of the normal
2461 * checksum functions (in skbuff.h) except that they operate on the GRO
2462 * offsets and fields in sk_buff.
2463 */
2464
2465__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2466
15e2396d
TH
2467static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2468{
b7fe10e5 2469 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2470}
2471
573e8fca
TH
2472static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2473 bool zero_okay,
2474 __sum16 check)
2475{
6edec0e6
TH
2476 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2477 skb_checksum_start_offset(skb) <
2478 skb_gro_offset(skb)) &&
15e2396d 2479 !skb_at_gro_remcsum_start(skb) &&
662880f4 2480 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2481 (!zero_okay || check));
2482}
2483
2484static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2485 __wsum psum)
2486{
2487 if (NAPI_GRO_CB(skb)->csum_valid &&
2488 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2489 return 0;
2490
2491 NAPI_GRO_CB(skb)->csum = psum;
2492
2493 return __skb_gro_checksum_complete(skb);
2494}
2495
573e8fca
TH
2496static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2497{
662880f4
TH
2498 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2499 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2500 NAPI_GRO_CB(skb)->csum_cnt--;
2501 } else {
2502 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2503 * verified a new top level checksum or an encapsulated one
2504 * during GRO. This saves work if we fallback to normal path.
2505 */
2506 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2507 }
2508}
2509
2510#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2511 compute_pseudo) \
2512({ \
2513 __sum16 __ret = 0; \
2514 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2515 __ret = __skb_gro_checksum_validate_complete(skb, \
2516 compute_pseudo(skb, proto)); \
5a212329
TH
2517 if (__ret) \
2518 __skb_mark_checksum_bad(skb); \
2519 else \
573e8fca
TH
2520 skb_gro_incr_csum_unnecessary(skb); \
2521 __ret; \
2522})
2523
2524#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2525 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2526
2527#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2528 compute_pseudo) \
2529 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2530
2531#define skb_gro_checksum_simple_validate(skb) \
2532 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2533
d96535a1
TH
2534static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2535{
2536 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2537 !NAPI_GRO_CB(skb)->csum_valid);
2538}
2539
2540static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2541 __sum16 check, __wsum pseudo)
2542{
2543 NAPI_GRO_CB(skb)->csum = ~pseudo;
2544 NAPI_GRO_CB(skb)->csum_valid = 1;
2545}
2546
2547#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2548do { \
2549 if (__skb_gro_checksum_convert_check(skb)) \
2550 __skb_gro_checksum_convert(skb, check, \
2551 compute_pseudo(skb, proto)); \
2552} while (0)
2553
26c4f7da
TH
2554struct gro_remcsum {
2555 int offset;
2556 __wsum delta;
2557};
2558
2559static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2560{
846cd667 2561 grc->offset = 0;
26c4f7da
TH
2562 grc->delta = 0;
2563}
2564
b7fe10e5
TH
2565static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2566 unsigned int off, size_t hdrlen,
2567 int start, int offset,
2568 struct gro_remcsum *grc,
2569 bool nopartial)
dcdc8994
TH
2570{
2571 __wsum delta;
b7fe10e5 2572 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2573
2574 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2575
15e2396d 2576 if (!nopartial) {
b7fe10e5
TH
2577 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2578 return ptr;
2579 }
2580
2581 ptr = skb_gro_header_fast(skb, off);
2582 if (skb_gro_header_hard(skb, off + plen)) {
2583 ptr = skb_gro_header_slow(skb, off + plen, off);
2584 if (!ptr)
2585 return NULL;
15e2396d
TH
2586 }
2587
b7fe10e5
TH
2588 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2589 start, offset);
dcdc8994
TH
2590
2591 /* Adjust skb->csum since we changed the packet */
dcdc8994 2592 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2593
b7fe10e5 2594 grc->offset = off + hdrlen + offset;
26c4f7da 2595 grc->delta = delta;
b7fe10e5
TH
2596
2597 return ptr;
dcdc8994
TH
2598}
2599
26c4f7da
TH
2600static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2601 struct gro_remcsum *grc)
2602{
b7fe10e5
TH
2603 void *ptr;
2604 size_t plen = grc->offset + sizeof(u16);
2605
26c4f7da
TH
2606 if (!grc->delta)
2607 return;
2608
b7fe10e5
TH
2609 ptr = skb_gro_header_fast(skb, grc->offset);
2610 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2611 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2612 if (!ptr)
2613 return;
2614 }
2615
2616 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2617}
dcdc8994 2618
6ae23ad3
TH
2619struct skb_csum_offl_spec {
2620 __u16 ipv4_okay:1,
2621 ipv6_okay:1,
2622 encap_okay:1,
2623 ip_options_okay:1,
2624 ext_hdrs_okay:1,
2625 tcp_okay:1,
2626 udp_okay:1,
2627 sctp_okay:1,
2628 vlan_okay:1,
2629 no_encapped_ipv6:1,
2630 no_not_encapped:1;
2631};
2632
2633bool __skb_csum_offload_chk(struct sk_buff *skb,
2634 const struct skb_csum_offl_spec *spec,
2635 bool *csum_encapped,
2636 bool csum_help);
2637
2638static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2639 const struct skb_csum_offl_spec *spec,
2640 bool *csum_encapped,
2641 bool csum_help)
2642{
2643 if (skb->ip_summed != CHECKSUM_PARTIAL)
2644 return false;
2645
2646 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2647}
2648
2649static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2650 const struct skb_csum_offl_spec *spec)
2651{
2652 bool csum_encapped;
2653
2654 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2655}
2656
2657static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2658{
2659 static const struct skb_csum_offl_spec csum_offl_spec = {
2660 .ipv4_okay = 1,
2661 .ip_options_okay = 1,
2662 .ipv6_okay = 1,
2663 .vlan_okay = 1,
2664 .tcp_okay = 1,
2665 .udp_okay = 1,
2666 };
2667
2668 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2669}
2670
2671static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2672{
2673 static const struct skb_csum_offl_spec csum_offl_spec = {
2674 .ipv4_okay = 1,
2675 .ip_options_okay = 1,
2676 .tcp_okay = 1,
2677 .udp_okay = 1,
2678 .vlan_okay = 1,
2679 };
2680
2681 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2682}
2683
0c4e8581
SH
2684static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2685 unsigned short type,
3b04ddde 2686 const void *daddr, const void *saddr,
95c96174 2687 unsigned int len)
0c4e8581 2688{
f1ecfd5d 2689 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2690 return 0;
3b04ddde
SH
2691
2692 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2693}
2694
b95cce35
SH
2695static inline int dev_parse_header(const struct sk_buff *skb,
2696 unsigned char *haddr)
2697{
2698 const struct net_device *dev = skb->dev;
2699
1b83336b 2700 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2701 return 0;
3b04ddde 2702 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2703}
2704
2793a23a
WB
2705/* ll_header must have at least hard_header_len allocated */
2706static inline bool dev_validate_header(const struct net_device *dev,
2707 char *ll_header, int len)
2708{
2709 if (likely(len >= dev->hard_header_len))
2710 return true;
2711
2712 if (capable(CAP_SYS_RAWIO)) {
2713 memset(ll_header + len, 0, dev->hard_header_len - len);
2714 return true;
2715 }
2716
2717 if (dev->header_ops && dev->header_ops->validate)
2718 return dev->header_ops->validate(ll_header, len);
2719
2720 return false;
2721}
2722
1da177e4 2723typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2724int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2725static inline int unregister_gifconf(unsigned int family)
2726{
2727 return register_gifconf(family, NULL);
2728}
2729
99bbc707 2730#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2731#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2732struct sd_flow_limit {
2733 u64 count;
2734 unsigned int num_buckets;
2735 unsigned int history_head;
2736 u16 history[FLOW_LIMIT_HISTORY];
2737 u8 buckets[];
2738};
2739
2740extern int netdev_flow_limit_table_len;
2741#endif /* CONFIG_NET_FLOW_LIMIT */
2742
1da177e4 2743/*
5e82b4b2 2744 * Incoming packets are placed on per-CPU queues
1da177e4 2745 */
d94d9fee 2746struct softnet_data {
1da177e4 2747 struct list_head poll_list;
6e7676c1 2748 struct sk_buff_head process_queue;
1da177e4 2749
dee42870 2750 /* stats */
cd7b5396
DM
2751 unsigned int processed;
2752 unsigned int time_squeeze;
cd7b5396 2753 unsigned int received_rps;
fd793d89 2754#ifdef CONFIG_RPS
88751275 2755 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2756#endif
2757#ifdef CONFIG_NET_FLOW_LIMIT
2758 struct sd_flow_limit __rcu *flow_limit;
2759#endif
2760 struct Qdisc *output_queue;
2761 struct Qdisc **output_queue_tailp;
2762 struct sk_buff *completion_queue;
88751275 2763
4cdb1e2e 2764#ifdef CONFIG_RPS
501e7ef5
ED
2765 /* input_queue_head should be written by cpu owning this struct,
2766 * and only read by other cpus. Worth using a cache line.
2767 */
2768 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2769
2770 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2771 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2772 struct softnet_data *rps_ipi_next;
2773 unsigned int cpu;
76cc8b13 2774 unsigned int input_queue_tail;
1e94d72f 2775#endif
95c96174 2776 unsigned int dropped;
0a9627f2 2777 struct sk_buff_head input_pkt_queue;
bea3348e 2778 struct napi_struct backlog;
99bbc707 2779
1da177e4
LT
2780};
2781
76cc8b13 2782static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2783{
2784#ifdef CONFIG_RPS
76cc8b13
TH
2785 sd->input_queue_head++;
2786#endif
2787}
2788
2789static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2790 unsigned int *qtail)
2791{
2792#ifdef CONFIG_RPS
2793 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2794#endif
2795}
2796
0a9627f2 2797DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2798
f629d208 2799void __netif_schedule(struct Qdisc *q);
46e5da40 2800void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2801
fd2ea0a7
DM
2802static inline void netif_tx_schedule_all(struct net_device *dev)
2803{
2804 unsigned int i;
2805
2806 for (i = 0; i < dev->num_tx_queues; i++)
2807 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2808}
2809
f9a7cbbf 2810static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2811{
73466498 2812 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2813}
2814
bea3348e
SH
2815/**
2816 * netif_start_queue - allow transmit
2817 * @dev: network device
2818 *
2819 * Allow upper layers to call the device hard_start_xmit routine.
2820 */
1da177e4
LT
2821static inline void netif_start_queue(struct net_device *dev)
2822{
e8a0464c 2823 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2824}
2825
fd2ea0a7
DM
2826static inline void netif_tx_start_all_queues(struct net_device *dev)
2827{
2828 unsigned int i;
2829
2830 for (i = 0; i < dev->num_tx_queues; i++) {
2831 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2832 netif_tx_start_queue(txq);
2833 }
2834}
2835
46e5da40 2836void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2837
d29f749e
DJ
2838/**
2839 * netif_wake_queue - restart transmit
2840 * @dev: network device
2841 *
2842 * Allow upper layers to call the device hard_start_xmit routine.
2843 * Used for flow control when transmit resources are available.
2844 */
79d16385
DM
2845static inline void netif_wake_queue(struct net_device *dev)
2846{
e8a0464c 2847 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2848}
2849
fd2ea0a7
DM
2850static inline void netif_tx_wake_all_queues(struct net_device *dev)
2851{
2852 unsigned int i;
2853
2854 for (i = 0; i < dev->num_tx_queues; i++) {
2855 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2856 netif_tx_wake_queue(txq);
2857 }
2858}
2859
f9a7cbbf 2860static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2861{
73466498 2862 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2863}
2864
bea3348e
SH
2865/**
2866 * netif_stop_queue - stop transmitted packets
2867 * @dev: network device
2868 *
2869 * Stop upper layers calling the device hard_start_xmit routine.
2870 * Used for flow control when transmit resources are unavailable.
2871 */
1da177e4
LT
2872static inline void netif_stop_queue(struct net_device *dev)
2873{
e8a0464c 2874 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2875}
2876
a2029240 2877void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2878
4d29515f 2879static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2880{
73466498 2881 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2882}
2883
bea3348e
SH
2884/**
2885 * netif_queue_stopped - test if transmit queue is flowblocked
2886 * @dev: network device
2887 *
2888 * Test if transmit queue on device is currently unable to send.
2889 */
4d29515f 2890static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2891{
e8a0464c 2892 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2893}
2894
4d29515f 2895static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2896{
73466498
TH
2897 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2898}
2899
8e2f1a63
DB
2900static inline bool
2901netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2902{
2903 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2904}
2905
8e2f1a63
DB
2906static inline bool
2907netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2908{
2909 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2910}
2911
53511453
ED
2912/**
2913 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2914 * @dev_queue: pointer to transmit queue
2915 *
2916 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2917 * to give appropriate hint to the CPU.
53511453
ED
2918 */
2919static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2920{
2921#ifdef CONFIG_BQL
2922 prefetchw(&dev_queue->dql.num_queued);
2923#endif
2924}
2925
2926/**
2927 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2928 * @dev_queue: pointer to transmit queue
2929 *
2930 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2931 * to give appropriate hint to the CPU.
53511453
ED
2932 */
2933static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2934{
2935#ifdef CONFIG_BQL
2936 prefetchw(&dev_queue->dql.limit);
2937#endif
2938}
2939
c5d67bd7
TH
2940static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2941 unsigned int bytes)
2942{
114cf580
TH
2943#ifdef CONFIG_BQL
2944 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2945
2946 if (likely(dql_avail(&dev_queue->dql) >= 0))
2947 return;
2948
2949 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2950
2951 /*
2952 * The XOFF flag must be set before checking the dql_avail below,
2953 * because in netdev_tx_completed_queue we update the dql_completed
2954 * before checking the XOFF flag.
2955 */
2956 smp_mb();
2957
2958 /* check again in case another CPU has just made room avail */
2959 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2960 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2961#endif
c5d67bd7
TH
2962}
2963
0042d0c8
FF
2964/**
2965 * netdev_sent_queue - report the number of bytes queued to hardware
2966 * @dev: network device
2967 * @bytes: number of bytes queued to the hardware device queue
2968 *
2969 * Report the number of bytes queued for sending/completion to the network
2970 * device hardware queue. @bytes should be a good approximation and should
2971 * exactly match netdev_completed_queue() @bytes
2972 */
c5d67bd7
TH
2973static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2974{
2975 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2976}
2977
2978static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2979 unsigned int pkts, unsigned int bytes)
c5d67bd7 2980{
114cf580 2981#ifdef CONFIG_BQL
b37c0fbe
AD
2982 if (unlikely(!bytes))
2983 return;
2984
2985 dql_completed(&dev_queue->dql, bytes);
2986
2987 /*
2988 * Without the memory barrier there is a small possiblity that
2989 * netdev_tx_sent_queue will miss the update and cause the queue to
2990 * be stopped forever
2991 */
2992 smp_mb();
2993
2994 if (dql_avail(&dev_queue->dql) < 0)
2995 return;
2996
2997 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2998 netif_schedule_queue(dev_queue);
114cf580 2999#endif
c5d67bd7
TH
3000}
3001
0042d0c8
FF
3002/**
3003 * netdev_completed_queue - report bytes and packets completed by device
3004 * @dev: network device
3005 * @pkts: actual number of packets sent over the medium
3006 * @bytes: actual number of bytes sent over the medium
3007 *
3008 * Report the number of bytes and packets transmitted by the network device
3009 * hardware queue over the physical medium, @bytes must exactly match the
3010 * @bytes amount passed to netdev_sent_queue()
3011 */
c5d67bd7 3012static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3013 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3014{
3015 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3016}
3017
3018static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3019{
114cf580 3020#ifdef CONFIG_BQL
5c490354 3021 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3022 dql_reset(&q->dql);
3023#endif
c5d67bd7
TH
3024}
3025
0042d0c8
FF
3026/**
3027 * netdev_reset_queue - reset the packets and bytes count of a network device
3028 * @dev_queue: network device
3029 *
3030 * Reset the bytes and packet count of a network device and clear the
3031 * software flow control OFF bit for this network device
3032 */
c5d67bd7
TH
3033static inline void netdev_reset_queue(struct net_device *dev_queue)
3034{
3035 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3036}
3037
b9507bda
DB
3038/**
3039 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3040 * @dev: network device
3041 * @queue_index: given tx queue index
3042 *
3043 * Returns 0 if given tx queue index >= number of device tx queues,
3044 * otherwise returns the originally passed tx queue index.
3045 */
3046static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3047{
3048 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3049 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3050 dev->name, queue_index,
3051 dev->real_num_tx_queues);
3052 return 0;
3053 }
3054
3055 return queue_index;
3056}
3057
bea3348e
SH
3058/**
3059 * netif_running - test if up
3060 * @dev: network device
3061 *
3062 * Test if the device has been brought up.
3063 */
4d29515f 3064static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3065{
3066 return test_bit(__LINK_STATE_START, &dev->state);
3067}
3068
f25f4e44 3069/*
5e82b4b2 3070 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3071 * stop, and a check if it's stopped. All other device management is
3072 * done at the overall netdevice level.
3073 * Also test the device if we're multiqueue.
3074 */
bea3348e
SH
3075
3076/**
3077 * netif_start_subqueue - allow sending packets on subqueue
3078 * @dev: network device
3079 * @queue_index: sub queue index
3080 *
3081 * Start individual transmit queue of a device with multiple transmit queues.
3082 */
f25f4e44
PWJ
3083static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3084{
fd2ea0a7 3085 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3086
3087 netif_tx_start_queue(txq);
f25f4e44
PWJ
3088}
3089
bea3348e
SH
3090/**
3091 * netif_stop_subqueue - stop sending packets on subqueue
3092 * @dev: network device
3093 * @queue_index: sub queue index
3094 *
3095 * Stop individual transmit queue of a device with multiple transmit queues.
3096 */
f25f4e44
PWJ
3097static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3098{
fd2ea0a7 3099 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3100 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3101}
3102
bea3348e
SH
3103/**
3104 * netif_subqueue_stopped - test status of subqueue
3105 * @dev: network device
3106 * @queue_index: sub queue index
3107 *
3108 * Check individual transmit queue of a device with multiple transmit queues.
3109 */
4d29515f
DM
3110static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3111 u16 queue_index)
f25f4e44 3112{
fd2ea0a7 3113 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3114
3115 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3116}
3117
4d29515f
DM
3118static inline bool netif_subqueue_stopped(const struct net_device *dev,
3119 struct sk_buff *skb)
668f895a
PE
3120{
3121 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3122}
bea3348e 3123
46e5da40 3124void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 3125
537c00de 3126#ifdef CONFIG_XPS
53af53ae 3127int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3128 u16 index);
537c00de
AD
3129#else
3130static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3131 const struct cpumask *mask,
537c00de
AD
3132 u16 index)
3133{
3134 return 0;
3135}
3136#endif
3137
5605c762
JP
3138u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3139 unsigned int num_tx_queues);
3140
a3d22a68
VZ
3141/*
3142 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3143 * as a distribution range limit for the returned value.
3144 */
3145static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3146 struct sk_buff *skb)
a3d22a68
VZ
3147{
3148 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3149}
3150
bea3348e
SH
3151/**
3152 * netif_is_multiqueue - test if device has multiple transmit queues
3153 * @dev: network device
3154 *
3155 * Check if device has multiple transmit queues
bea3348e 3156 */
4d29515f 3157static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3158{
a02cec21 3159 return dev->num_tx_queues > 1;
f25f4e44 3160}
1da177e4 3161
f629d208 3162int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3163
a953be53 3164#ifdef CONFIG_SYSFS
f629d208 3165int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3166#else
3167static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3168 unsigned int rxq)
3169{
3170 return 0;
3171}
3172#endif
3173
a953be53
MD
3174#ifdef CONFIG_SYSFS
3175static inline unsigned int get_netdev_rx_queue_index(
3176 struct netdev_rx_queue *queue)
3177{
3178 struct net_device *dev = queue->dev;
3179 int index = queue - dev->_rx;
3180
3181 BUG_ON(index >= dev->num_rx_queues);
3182 return index;
3183}
3184#endif
3185
16917b87 3186#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3187int netif_get_num_default_rss_queues(void);
16917b87 3188
e6247027
ED
3189enum skb_free_reason {
3190 SKB_REASON_CONSUMED,
3191 SKB_REASON_DROPPED,
3192};
3193
3194void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3195void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3196
e6247027
ED
3197/*
3198 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3199 * interrupt context or with hardware interrupts being disabled.
3200 * (in_irq() || irqs_disabled())
3201 *
3202 * We provide four helpers that can be used in following contexts :
3203 *
3204 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3205 * replacing kfree_skb(skb)
3206 *
3207 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3208 * Typically used in place of consume_skb(skb) in TX completion path
3209 *
3210 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3211 * replacing kfree_skb(skb)
3212 *
3213 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3214 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3215 */
e6247027
ED
3216static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3217{
3218 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3219}
3220
3221static inline void dev_consume_skb_irq(struct sk_buff *skb)
3222{
3223 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3224}
3225
3226static inline void dev_kfree_skb_any(struct sk_buff *skb)
3227{
3228 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3229}
3230
3231static inline void dev_consume_skb_any(struct sk_buff *skb)
3232{
3233 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3234}
1da177e4 3235
f629d208
JP
3236int netif_rx(struct sk_buff *skb);
3237int netif_rx_ni(struct sk_buff *skb);
04eb4489 3238int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3239gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3240void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3241struct sk_buff *napi_get_frags(struct napi_struct *napi);
3242gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3243struct packet_offload *gro_find_receive_by_type(__be16 type);
3244struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3245
3246static inline void napi_free_frags(struct napi_struct *napi)
3247{
3248 kfree_skb(napi->skb);
3249 napi->skb = NULL;
3250}
3251
f629d208
JP
3252int netdev_rx_handler_register(struct net_device *dev,
3253 rx_handler_func_t *rx_handler,
3254 void *rx_handler_data);
3255void netdev_rx_handler_unregister(struct net_device *dev);
3256
3257bool dev_valid_name(const char *name);
3258int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3259int dev_ethtool(struct net *net, struct ifreq *);
3260unsigned int dev_get_flags(const struct net_device *);
3261int __dev_change_flags(struct net_device *, unsigned int flags);
3262int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3263void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3264 unsigned int gchanges);
f629d208
JP
3265int dev_change_name(struct net_device *, const char *);
3266int dev_set_alias(struct net_device *, const char *, size_t);
3267int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3268int dev_set_mtu(struct net_device *, int);
3269void dev_set_group(struct net_device *, int);
3270int dev_set_mac_address(struct net_device *, struct sockaddr *);
3271int dev_change_carrier(struct net_device *, bool new_carrier);
3272int dev_get_phys_port_id(struct net_device *dev,
02637fce 3273 struct netdev_phys_item_id *ppid);
db24a904
DA
3274int dev_get_phys_port_name(struct net_device *dev,
3275 char *name, size_t len);
d746d707 3276int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3277struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3278struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3279 struct netdev_queue *txq, int *ret);
a0265d28 3280int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3281int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3282bool is_skb_forwardable(const struct net_device *dev,
3283 const struct sk_buff *skb);
1da177e4 3284
74b20582
DA
3285void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3286
20380731 3287extern int netdev_budget;
1da177e4
LT
3288
3289/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3290void netdev_run_todo(void);
1da177e4 3291
bea3348e
SH
3292/**
3293 * dev_put - release reference to device
3294 * @dev: network device
3295 *
9ef4429b 3296 * Release reference to device to allow it to be freed.
bea3348e 3297 */
1da177e4
LT
3298static inline void dev_put(struct net_device *dev)
3299{
933393f5 3300 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3301}
3302
bea3348e
SH
3303/**
3304 * dev_hold - get reference to device
3305 * @dev: network device
3306 *
9ef4429b 3307 * Hold reference to device to keep it from being freed.
bea3348e 3308 */
15333061
SH
3309static inline void dev_hold(struct net_device *dev)
3310{
933393f5 3311 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3312}
1da177e4
LT
3313
3314/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3315 * and _off may be called from IRQ context, but it is caller
3316 * who is responsible for serialization of these calls.
b00055aa
SR
3317 *
3318 * The name carrier is inappropriate, these functions should really be
3319 * called netif_lowerlayer_*() because they represent the state of any
3320 * kind of lower layer not just hardware media.
1da177e4
LT
3321 */
3322
f629d208
JP
3323void linkwatch_init_dev(struct net_device *dev);
3324void linkwatch_fire_event(struct net_device *dev);
3325void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3326
bea3348e
SH
3327/**
3328 * netif_carrier_ok - test if carrier present
3329 * @dev: network device
3330 *
3331 * Check if carrier is present on device
3332 */
4d29515f 3333static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3334{
3335 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3336}
3337
f629d208 3338unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3339
f629d208 3340void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3341
f629d208 3342void netif_carrier_on(struct net_device *dev);
1da177e4 3343
f629d208 3344void netif_carrier_off(struct net_device *dev);
1da177e4 3345
bea3348e
SH
3346/**
3347 * netif_dormant_on - mark device as dormant.
3348 * @dev: network device
3349 *
3350 * Mark device as dormant (as per RFC2863).
3351 *
3352 * The dormant state indicates that the relevant interface is not
3353 * actually in a condition to pass packets (i.e., it is not 'up') but is
3354 * in a "pending" state, waiting for some external event. For "on-
3355 * demand" interfaces, this new state identifies the situation where the
3356 * interface is waiting for events to place it in the up state.
bea3348e 3357 */
b00055aa
SR
3358static inline void netif_dormant_on(struct net_device *dev)
3359{
3360 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3361 linkwatch_fire_event(dev);
3362}
3363
bea3348e
SH
3364/**
3365 * netif_dormant_off - set device as not dormant.
3366 * @dev: network device
3367 *
3368 * Device is not in dormant state.
3369 */
b00055aa
SR
3370static inline void netif_dormant_off(struct net_device *dev)
3371{
3372 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3373 linkwatch_fire_event(dev);
3374}
3375
bea3348e
SH
3376/**
3377 * netif_dormant - test if carrier present
3378 * @dev: network device
3379 *
3380 * Check if carrier is present on device
3381 */
4d29515f 3382static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3383{
3384 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3385}
3386
3387
bea3348e
SH
3388/**
3389 * netif_oper_up - test if device is operational
3390 * @dev: network device
3391 *
3392 * Check if carrier is operational
3393 */
4d29515f 3394static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3395{
b00055aa
SR
3396 return (dev->operstate == IF_OPER_UP ||
3397 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3398}
3399
bea3348e
SH
3400/**
3401 * netif_device_present - is device available or removed
3402 * @dev: network device
3403 *
3404 * Check if device has not been removed from system.
3405 */
4d29515f 3406static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3407{
3408 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3409}
3410
f629d208 3411void netif_device_detach(struct net_device *dev);
1da177e4 3412
f629d208 3413void netif_device_attach(struct net_device *dev);
1da177e4
LT
3414
3415/*
3416 * Network interface message level settings
3417 */
1da177e4
LT
3418
3419enum {
3420 NETIF_MSG_DRV = 0x0001,
3421 NETIF_MSG_PROBE = 0x0002,
3422 NETIF_MSG_LINK = 0x0004,
3423 NETIF_MSG_TIMER = 0x0008,
3424 NETIF_MSG_IFDOWN = 0x0010,
3425 NETIF_MSG_IFUP = 0x0020,
3426 NETIF_MSG_RX_ERR = 0x0040,
3427 NETIF_MSG_TX_ERR = 0x0080,
3428 NETIF_MSG_TX_QUEUED = 0x0100,
3429 NETIF_MSG_INTR = 0x0200,
3430 NETIF_MSG_TX_DONE = 0x0400,
3431 NETIF_MSG_RX_STATUS = 0x0800,
3432 NETIF_MSG_PKTDATA = 0x1000,
3433 NETIF_MSG_HW = 0x2000,
3434 NETIF_MSG_WOL = 0x4000,
3435};
3436
3437#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3438#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3439#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3440#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3441#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3442#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3443#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3444#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3445#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3446#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3447#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3448#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3449#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3450#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3451#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3452
3453static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3454{
3455 /* use default */
3456 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3457 return default_msg_enable_bits;
3458 if (debug_value == 0) /* no output */
3459 return 0;
3460 /* set low N bits */
3461 return (1 << debug_value) - 1;
3462}
3463
c773e847 3464static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3465{
c773e847
DM
3466 spin_lock(&txq->_xmit_lock);
3467 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3468}
3469
fd2ea0a7
DM
3470static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3471{
3472 spin_lock_bh(&txq->_xmit_lock);
3473 txq->xmit_lock_owner = smp_processor_id();
3474}
3475
4d29515f 3476static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3477{
4d29515f 3478 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3479 if (likely(ok))
3480 txq->xmit_lock_owner = smp_processor_id();
3481 return ok;
3482}
3483
3484static inline void __netif_tx_unlock(struct netdev_queue *txq)
3485{
3486 txq->xmit_lock_owner = -1;
3487 spin_unlock(&txq->_xmit_lock);
3488}
3489
3490static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3491{
3492 txq->xmit_lock_owner = -1;
3493 spin_unlock_bh(&txq->_xmit_lock);
3494}
3495
08baf561
ED
3496static inline void txq_trans_update(struct netdev_queue *txq)
3497{
3498 if (txq->xmit_lock_owner != -1)
3499 txq->trans_start = jiffies;
3500}
3501
ba162f8e
FW
3502/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3503static inline void netif_trans_update(struct net_device *dev)
3504{
9b36627a
FW
3505 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3506
3507 if (txq->trans_start != jiffies)
3508 txq->trans_start = jiffies;
ba162f8e
FW
3509}
3510
d29f749e
DJ
3511/**
3512 * netif_tx_lock - grab network device transmit lock
3513 * @dev: network device
d29f749e
DJ
3514 *
3515 * Get network device transmit lock
3516 */
22dd7495
JHS
3517static inline void netif_tx_lock(struct net_device *dev)
3518{
e8a0464c 3519 unsigned int i;
c3f26a26 3520 int cpu;
c773e847 3521
c3f26a26
DM
3522 spin_lock(&dev->tx_global_lock);
3523 cpu = smp_processor_id();
e8a0464c
DM
3524 for (i = 0; i < dev->num_tx_queues; i++) {
3525 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3526
3527 /* We are the only thread of execution doing a
3528 * freeze, but we have to grab the _xmit_lock in
3529 * order to synchronize with threads which are in
3530 * the ->hard_start_xmit() handler and already
3531 * checked the frozen bit.
3532 */
e8a0464c 3533 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3534 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3535 __netif_tx_unlock(txq);
e8a0464c 3536 }
932ff279
HX
3537}
3538
3539static inline void netif_tx_lock_bh(struct net_device *dev)
3540{
e8a0464c
DM
3541 local_bh_disable();
3542 netif_tx_lock(dev);
932ff279
HX
3543}
3544
932ff279
HX
3545static inline void netif_tx_unlock(struct net_device *dev)
3546{
e8a0464c
DM
3547 unsigned int i;
3548
3549 for (i = 0; i < dev->num_tx_queues; i++) {
3550 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3551
c3f26a26
DM
3552 /* No need to grab the _xmit_lock here. If the
3553 * queue is not stopped for another reason, we
3554 * force a schedule.
3555 */
3556 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3557 netif_schedule_queue(txq);
c3f26a26
DM
3558 }
3559 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3560}
3561
3562static inline void netif_tx_unlock_bh(struct net_device *dev)
3563{
e8a0464c
DM
3564 netif_tx_unlock(dev);
3565 local_bh_enable();
932ff279
HX
3566}
3567
c773e847 3568#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3569 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3570 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3571 } \
3572}
3573
5efeac44
EB
3574#define HARD_TX_TRYLOCK(dev, txq) \
3575 (((dev->features & NETIF_F_LLTX) == 0) ? \
3576 __netif_tx_trylock(txq) : \
3577 true )
3578
c773e847 3579#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3580 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3581 __netif_tx_unlock(txq); \
22dd7495
JHS
3582 } \
3583}
3584
1da177e4
LT
3585static inline void netif_tx_disable(struct net_device *dev)
3586{
fd2ea0a7 3587 unsigned int i;
c3f26a26 3588 int cpu;
fd2ea0a7 3589
c3f26a26
DM
3590 local_bh_disable();
3591 cpu = smp_processor_id();
fd2ea0a7
DM
3592 for (i = 0; i < dev->num_tx_queues; i++) {
3593 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3594
3595 __netif_tx_lock(txq, cpu);
fd2ea0a7 3596 netif_tx_stop_queue(txq);
c3f26a26 3597 __netif_tx_unlock(txq);
fd2ea0a7 3598 }
c3f26a26 3599 local_bh_enable();
1da177e4
LT
3600}
3601
e308a5d8
DM
3602static inline void netif_addr_lock(struct net_device *dev)
3603{
3604 spin_lock(&dev->addr_list_lock);
3605}
3606
2429f7ac
JP
3607static inline void netif_addr_lock_nested(struct net_device *dev)
3608{
25175ba5
VY
3609 int subclass = SINGLE_DEPTH_NESTING;
3610
3611 if (dev->netdev_ops->ndo_get_lock_subclass)
3612 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3613
3614 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3615}
3616
e308a5d8
DM
3617static inline void netif_addr_lock_bh(struct net_device *dev)
3618{
3619 spin_lock_bh(&dev->addr_list_lock);
3620}
3621
3622static inline void netif_addr_unlock(struct net_device *dev)
3623{
3624 spin_unlock(&dev->addr_list_lock);
3625}
3626
3627static inline void netif_addr_unlock_bh(struct net_device *dev)
3628{
3629 spin_unlock_bh(&dev->addr_list_lock);
3630}
3631
f001fde5 3632/*
31278e71 3633 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3634 * rcu_read_lock held.
3635 */
3636#define for_each_dev_addr(dev, ha) \
31278e71 3637 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3638
1da177e4
LT
3639/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3640
f629d208 3641void ether_setup(struct net_device *dev);
1da177e4
LT
3642
3643/* Support for loadable net-drivers */
f629d208 3644struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3645 unsigned char name_assign_type,
f629d208
JP
3646 void (*setup)(struct net_device *),
3647 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3648#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3649 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3650
c835a677
TG
3651#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3652 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3653 count)
36909ea4 3654
f629d208
JP
3655int register_netdev(struct net_device *dev);
3656void unregister_netdev(struct net_device *dev);
f001fde5 3657
22bedad3 3658/* General hardware address lists handling functions */
f629d208
JP
3659int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3660 struct netdev_hw_addr_list *from_list, int addr_len);
3661void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3662 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3663int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3664 struct net_device *dev,
3665 int (*sync)(struct net_device *, const unsigned char *),
3666 int (*unsync)(struct net_device *,
3667 const unsigned char *));
3668void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3669 struct net_device *dev,
3670 int (*unsync)(struct net_device *,
3671 const unsigned char *));
f629d208 3672void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3673
f001fde5 3674/* Functions used for device addresses handling */
f629d208
JP
3675int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3676 unsigned char addr_type);
3677int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3678 unsigned char addr_type);
f629d208
JP
3679void dev_addr_flush(struct net_device *dev);
3680int dev_addr_init(struct net_device *dev);
a748ee24
JP
3681
3682/* Functions used for unicast addresses handling */
f629d208
JP
3683int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3684int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3685int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3686int dev_uc_sync(struct net_device *to, struct net_device *from);
3687int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3688void dev_uc_unsync(struct net_device *to, struct net_device *from);
3689void dev_uc_flush(struct net_device *dev);
3690void dev_uc_init(struct net_device *dev);
f001fde5 3691
670e5b8e
AD
3692/**
3693 * __dev_uc_sync - Synchonize device's unicast list
3694 * @dev: device to sync
3695 * @sync: function to call if address should be added
3696 * @unsync: function to call if address should be removed
3697 *
3698 * Add newly added addresses to the interface, and release
3699 * addresses that have been deleted.
5e82b4b2 3700 */
670e5b8e
AD
3701static inline int __dev_uc_sync(struct net_device *dev,
3702 int (*sync)(struct net_device *,
3703 const unsigned char *),
3704 int (*unsync)(struct net_device *,
3705 const unsigned char *))
3706{
3707 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3708}
3709
3710/**
e793c0f7 3711 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3712 * @dev: device to sync
3713 * @unsync: function to call if address should be removed
3714 *
3715 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3716 */
670e5b8e
AD
3717static inline void __dev_uc_unsync(struct net_device *dev,
3718 int (*unsync)(struct net_device *,
3719 const unsigned char *))
3720{
3721 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3722}
3723
22bedad3 3724/* Functions used for multicast addresses handling */
f629d208
JP
3725int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3726int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3727int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3728int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3729int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3730int dev_mc_sync(struct net_device *to, struct net_device *from);
3731int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3732void dev_mc_unsync(struct net_device *to, struct net_device *from);
3733void dev_mc_flush(struct net_device *dev);
3734void dev_mc_init(struct net_device *dev);
f001fde5 3735
670e5b8e
AD
3736/**
3737 * __dev_mc_sync - Synchonize device's multicast list
3738 * @dev: device to sync
3739 * @sync: function to call if address should be added
3740 * @unsync: function to call if address should be removed
3741 *
3742 * Add newly added addresses to the interface, and release
3743 * addresses that have been deleted.
5e82b4b2 3744 */
670e5b8e
AD
3745static inline int __dev_mc_sync(struct net_device *dev,
3746 int (*sync)(struct net_device *,
3747 const unsigned char *),
3748 int (*unsync)(struct net_device *,
3749 const unsigned char *))
3750{
3751 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3752}
3753
3754/**
e793c0f7 3755 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3756 * @dev: device to sync
3757 * @unsync: function to call if address should be removed
3758 *
3759 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3760 */
670e5b8e
AD
3761static inline void __dev_mc_unsync(struct net_device *dev,
3762 int (*unsync)(struct net_device *,
3763 const unsigned char *))
3764{
3765 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3766}
3767
4417da66 3768/* Functions used for secondary unicast and multicast support */
f629d208
JP
3769void dev_set_rx_mode(struct net_device *dev);
3770void __dev_set_rx_mode(struct net_device *dev);
3771int dev_set_promiscuity(struct net_device *dev, int inc);
3772int dev_set_allmulti(struct net_device *dev, int inc);
3773void netdev_state_change(struct net_device *dev);
3774void netdev_notify_peers(struct net_device *dev);
3775void netdev_features_change(struct net_device *dev);
1da177e4 3776/* Load a device via the kmod */
f629d208
JP
3777void dev_load(struct net *net, const char *name);
3778struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3779 struct rtnl_link_stats64 *storage);
3780void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3781 const struct net_device_stats *netdev_stats);
eeda3fd6 3782
1da177e4 3783extern int netdev_max_backlog;
3b098e2d 3784extern int netdev_tstamp_prequeue;
1da177e4 3785extern int weight_p;
9ff162a8 3786
f629d208 3787bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3788struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3789 struct list_head **iter);
f629d208
JP
3790struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3791 struct list_head **iter);
8b5be856 3792
44a40855
VY
3793/* iterate through upper list, must be called under RCU read lock */
3794#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3795 for (iter = &(dev)->adj_list.upper, \
3796 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3797 updev; \
3798 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3799
8b5be856 3800/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3801#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3802 for (iter = &(dev)->all_adj_list.upper, \
3803 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3804 updev; \
3805 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3806
f629d208
JP
3807void *netdev_lower_get_next_private(struct net_device *dev,
3808 struct list_head **iter);
3809void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3810 struct list_head **iter);
31088a11
VF
3811
3812#define netdev_for_each_lower_private(dev, priv, iter) \
3813 for (iter = (dev)->adj_list.lower.next, \
3814 priv = netdev_lower_get_next_private(dev, &(iter)); \
3815 priv; \
3816 priv = netdev_lower_get_next_private(dev, &(iter)))
3817
3818#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3819 for (iter = &(dev)->adj_list.lower, \
3820 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3821 priv; \
3822 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3823
4085ebe8
VY
3824void *netdev_lower_get_next(struct net_device *dev,
3825 struct list_head **iter);
3826#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3827 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3828 ldev = netdev_lower_get_next(dev, &(iter)); \
3829 ldev; \
3830 ldev = netdev_lower_get_next(dev, &(iter)))
3831
f629d208 3832void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3833void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3834struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3835struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3836int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3837int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3838 struct net_device *upper_dev,
29bf24af 3839 void *upper_priv, void *upper_info);
f629d208
JP
3840void netdev_upper_dev_unlink(struct net_device *dev,
3841 struct net_device *upper_dev);
5bb025fa 3842void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3843void *netdev_lower_dev_get_private(struct net_device *dev,
3844 struct net_device *lower_dev);
04d48266
JP
3845void netdev_lower_state_changed(struct net_device *lower_dev,
3846 void *lower_state_info);
960fb622
ED
3847
3848/* RSS keys are 40 or 52 bytes long */
3849#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3850extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3851void netdev_rss_key_fill(void *buffer, size_t len);
3852
4085ebe8 3853int dev_get_nest_level(struct net_device *dev,
b618aaa9 3854 bool (*type_check)(const struct net_device *dev));
f629d208
JP
3855int skb_checksum_help(struct sk_buff *skb);
3856struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3857 netdev_features_t features, bool tx_path);
3858struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3859 netdev_features_t features);
12b0004d 3860
61bd3857
MS
3861struct netdev_bonding_info {
3862 ifslave slave;
3863 ifbond master;
3864};
3865
3866struct netdev_notifier_bonding_info {
3867 struct netdev_notifier_info info; /* must be first */
3868 struct netdev_bonding_info bonding_info;
3869};
3870
3871void netdev_bonding_info_change(struct net_device *dev,
3872 struct netdev_bonding_info *bonding_info);
3873
12b0004d
CW
3874static inline
3875struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3876{
3877 return __skb_gso_segment(skb, features, true);
3878}
53d6471c 3879__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3880
3881static inline bool can_checksum_protocol(netdev_features_t features,
3882 __be16 protocol)
3883{
c8cd0989
TH
3884 if (protocol == htons(ETH_P_FCOE))
3885 return !!(features & NETIF_F_FCOE_CRC);
3886
3887 /* Assume this is an IP checksum (not SCTP CRC) */
3888
3889 if (features & NETIF_F_HW_CSUM) {
3890 /* Can checksum everything */
3891 return true;
3892 }
3893
3894 switch (protocol) {
3895 case htons(ETH_P_IP):
3896 return !!(features & NETIF_F_IP_CSUM);
3897 case htons(ETH_P_IPV6):
3898 return !!(features & NETIF_F_IPV6_CSUM);
3899 default:
3900 return false;
3901 }
ec5f0615 3902}
12b0004d 3903
6ae23ad3
TH
3904/* Map an ethertype into IP protocol if possible */
3905static inline int eproto_to_ipproto(int eproto)
3906{
3907 switch (eproto) {
3908 case htons(ETH_P_IP):
3909 return IPPROTO_IP;
3910 case htons(ETH_P_IPV6):
3911 return IPPROTO_IPV6;
3912 default:
3913 return -1;
3914 }
3915}
3916
fb286bb2 3917#ifdef CONFIG_BUG
f629d208 3918void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3919#else
3920static inline void netdev_rx_csum_fault(struct net_device *dev)
3921{
3922}
3923#endif
1da177e4 3924/* rx skb timestamps */
f629d208
JP
3925void net_enable_timestamp(void);
3926void net_disable_timestamp(void);
1da177e4 3927
20380731 3928#ifdef CONFIG_PROC_FS
f629d208 3929int __init dev_proc_init(void);
900ff8c6
CW
3930#else
3931#define dev_proc_init() 0
20380731
ACM
3932#endif
3933
4798248e 3934static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3935 struct sk_buff *skb, struct net_device *dev,
3936 bool more)
4798248e 3937{
fa2dbdc2 3938 skb->xmit_more = more ? 1 : 0;
0b725a2c 3939 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3940}
3941
10b3ad8c 3942static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3943 struct netdev_queue *txq, bool more)
4798248e
DM
3944{
3945 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3946 int rc;
4798248e 3947
fa2dbdc2 3948 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3949 if (rc == NETDEV_TX_OK)
3950 txq_trans_update(txq);
3951
3952 return rc;
4798248e
DM
3953}
3954
42a2d923
LT
3955int netdev_class_create_file_ns(struct class_attribute *class_attr,
3956 const void *ns);
3957void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3958 const void *ns);
58292cbe
TH
3959
3960static inline int netdev_class_create_file(struct class_attribute *class_attr)
3961{
3962 return netdev_class_create_file_ns(class_attr, NULL);
3963}
3964
3965static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3966{
3967 netdev_class_remove_file_ns(class_attr, NULL);
3968}
b8a9787e 3969
04600794
JB
3970extern struct kobj_ns_type_operations net_ns_type_operations;
3971
f629d208 3972const char *netdev_drivername(const struct net_device *dev);
6579e57b 3973
f629d208 3974void linkwatch_run_queue(void);
20380731 3975
da08143b
MK
3976static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3977 netdev_features_t f2)
3978{
c8cd0989
TH
3979 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3980 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 3981 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 3982 else
b6a0e72a 3983 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 3984 }
da08143b 3985
c8cd0989 3986 return f1 & f2;
da08143b
MK
3987}
3988
c8f44aff
MM
3989static inline netdev_features_t netdev_get_wanted_features(
3990 struct net_device *dev)
5455c699
MM
3991{
3992 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3993}
c8f44aff
MM
3994netdev_features_t netdev_increment_features(netdev_features_t all,
3995 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3996
3997/* Allow TSO being used on stacked device :
3998 * Performing the GSO segmentation before last device
3999 * is a performance improvement.
4000 */
4001static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4002 netdev_features_t mask)
4003{
4004 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4005}
4006
6cb6a27c 4007int __netdev_update_features(struct net_device *dev);
5455c699 4008void netdev_update_features(struct net_device *dev);
afe12cc8 4009void netdev_change_features(struct net_device *dev);
7f353bf2 4010
fc4a7489
PM
4011void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4012 struct net_device *dev);
4013
e38f3025
TM
4014netdev_features_t passthru_features_check(struct sk_buff *skb,
4015 struct net_device *dev,
4016 netdev_features_t features);
c1e756bf 4017netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4018
4d29515f 4019static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4020{
7b748340 4021 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4022
4023 /* check flags correspondence */
4024 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4025 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4026 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4027 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4028 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4029 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4030 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4031 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4032 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4033 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4034 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4035 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4036 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4037 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4038 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4039 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
0345e186 4040
d6b4991a 4041 return (features & feature) == feature;
576a30eb
HX
4042}
4043
4d29515f 4044static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4045{
278b2513 4046 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4047 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4048}
4049
8b86a61d 4050static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4051 netdev_features_t features)
7967168c 4052{
fc741216 4053 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4054 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4055 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4056}
4057
82cc1a7a
PWJ
4058static inline void netif_set_gso_max_size(struct net_device *dev,
4059 unsigned int size)
4060{
4061 dev->gso_max_size = size;
4062}
4063
7a7ffbab
WCC
4064static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4065 int pulled_hlen, u16 mac_offset,
4066 int mac_len)
4067{
4068 skb->protocol = protocol;
4069 skb->encapsulation = 1;
4070 skb_push(skb, pulled_hlen);
4071 skb_reset_transport_header(skb);
4072 skb->mac_header = mac_offset;
4073 skb->network_header = skb->mac_header + mac_len;
4074 skb->mac_len = mac_len;
4075}
4076
3c175784
SD
4077static inline bool netif_is_macsec(const struct net_device *dev)
4078{
4079 return dev->priv_flags & IFF_MACSEC;
4080}
4081
b618aaa9 4082static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4083{
4084 return dev->priv_flags & IFF_MACVLAN;
4085}
4086
b618aaa9 4087static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4088{
4089 return dev->priv_flags & IFF_MACVLAN_PORT;
4090}
4091
b618aaa9 4092static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4093{
4094 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4095}
4096
b618aaa9 4097static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4098{
4099 return dev->priv_flags & IFF_IPVLAN_MASTER;
4100}
4101
b618aaa9 4102static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4103{
4104 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4105}
4106
b618aaa9 4107static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4108{
4109 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4110}
4111
3bdc0eba
BG
4112static inline bool netif_supports_nofcs(struct net_device *dev)
4113{
4114 return dev->priv_flags & IFF_SUPP_NOFCS;
4115}
4116
007979ea 4117static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4118{
007979ea 4119 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4120}
4121
fee6d4c7
DA
4122static inline bool netif_is_l3_slave(const struct net_device *dev)
4123{
4124 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4125}
4126
0894ae3f
JP
4127static inline bool netif_is_bridge_master(const struct net_device *dev)
4128{
4129 return dev->priv_flags & IFF_EBRIDGE;
4130}
4131
28f9ee22
VY
4132static inline bool netif_is_bridge_port(const struct net_device *dev)
4133{
4134 return dev->priv_flags & IFF_BRIDGE_PORT;
4135}
4136
35d4e172
JP
4137static inline bool netif_is_ovs_master(const struct net_device *dev)
4138{
4139 return dev->priv_flags & IFF_OPENVSWITCH;
4140}
4141
b618aaa9 4142static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4143{
4144 return dev->priv_flags & IFF_TEAM;
4145}
4146
b618aaa9 4147static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4148{
4149 return dev->priv_flags & IFF_TEAM_PORT;
4150}
4151
b618aaa9 4152static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4153{
4154 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4155}
4156
b618aaa9 4157static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4158{
4159 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4160}
4161
d4ab4286
KJ
4162static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4163{
4164 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4165}
4166
02875878
ED
4167/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4168static inline void netif_keep_dst(struct net_device *dev)
4169{
4170 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4171}
4172
505d4f73 4173extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4174
571ba423
JP
4175/* Logging, debugging and troubleshooting/diagnostic helpers. */
4176
4177/* netdev_printk helpers, similar to dev_printk */
4178
4179static inline const char *netdev_name(const struct net_device *dev)
4180{
c6f854d5
VF
4181 if (!dev->name[0] || strchr(dev->name, '%'))
4182 return "(unnamed net_device)";
571ba423
JP
4183 return dev->name;
4184}
4185
ccc7f496
VF
4186static inline const char *netdev_reg_state(const struct net_device *dev)
4187{
4188 switch (dev->reg_state) {
4189 case NETREG_UNINITIALIZED: return " (uninitialized)";
4190 case NETREG_REGISTERED: return "";
4191 case NETREG_UNREGISTERING: return " (unregistering)";
4192 case NETREG_UNREGISTERED: return " (unregistered)";
4193 case NETREG_RELEASED: return " (released)";
4194 case NETREG_DUMMY: return " (dummy)";
4195 }
4196
4197 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4198 return " (unknown)";
4199}
4200
f629d208 4201__printf(3, 4)
6ea754eb
JP
4202void netdev_printk(const char *level, const struct net_device *dev,
4203 const char *format, ...);
f629d208 4204__printf(2, 3)
6ea754eb 4205void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4206__printf(2, 3)
6ea754eb 4207void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4208__printf(2, 3)
6ea754eb 4209void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4210__printf(2, 3)
6ea754eb 4211void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4212__printf(2, 3)
6ea754eb 4213void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4214__printf(2, 3)
6ea754eb 4215void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4216__printf(2, 3)
6ea754eb 4217void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4218
8909c9ad
VK
4219#define MODULE_ALIAS_NETDEV(device) \
4220 MODULE_ALIAS("netdev-" device)
4221
b558c96f 4222#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4223#define netdev_dbg(__dev, format, args...) \
4224do { \
ffa10cb4 4225 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4226} while (0)
b558c96f
JC
4227#elif defined(DEBUG)
4228#define netdev_dbg(__dev, format, args...) \
4229 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4230#else
4231#define netdev_dbg(__dev, format, args...) \
4232({ \
4233 if (0) \
4234 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4235})
4236#endif
4237
4238#if defined(VERBOSE_DEBUG)
4239#define netdev_vdbg netdev_dbg
4240#else
4241
4242#define netdev_vdbg(dev, format, args...) \
4243({ \
4244 if (0) \
4245 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4246 0; \
4247})
4248#endif
4249
4250/*
4251 * netdev_WARN() acts like dev_printk(), but with the key difference
4252 * of using a WARN/WARN_ON to get the message out, including the
4253 * file/line information and a backtrace.
4254 */
4255#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4256 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4257 netdev_reg_state(dev), ##args)
571ba423 4258
b3d95c5c
JP
4259/* netif printk helpers, similar to netdev_printk */
4260
4261#define netif_printk(priv, type, level, dev, fmt, args...) \
4262do { \
4263 if (netif_msg_##type(priv)) \
4264 netdev_printk(level, (dev), fmt, ##args); \
4265} while (0)
4266
f45f4321
JP
4267#define netif_level(level, priv, type, dev, fmt, args...) \
4268do { \
4269 if (netif_msg_##type(priv)) \
4270 netdev_##level(dev, fmt, ##args); \
4271} while (0)
4272
b3d95c5c 4273#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4274 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4275#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4276 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4277#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4278 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4279#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4280 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4281#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4282 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4283#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4284 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4285#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4286 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4287
0053ea9c 4288#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4289#define netif_dbg(priv, type, netdev, format, args...) \
4290do { \
4291 if (netif_msg_##type(priv)) \
b5fb0a03 4292 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4293} while (0)
0053ea9c
JP
4294#elif defined(DEBUG)
4295#define netif_dbg(priv, type, dev, format, args...) \
4296 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4297#else
4298#define netif_dbg(priv, type, dev, format, args...) \
4299({ \
4300 if (0) \
4301 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4302 0; \
4303})
4304#endif
4305
4306#if defined(VERBOSE_DEBUG)
bcfcc450 4307#define netif_vdbg netif_dbg
b3d95c5c
JP
4308#else
4309#define netif_vdbg(priv, type, dev, format, args...) \
4310({ \
4311 if (0) \
a4ed89cb 4312 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4313 0; \
4314})
4315#endif
571ba423 4316
900ff8c6
CW
4317/*
4318 * The list of packet types we will receive (as opposed to discard)
4319 * and the routines to invoke.
4320 *
4321 * Why 16. Because with 16 the only overlap we get on a hash of the
4322 * low nibble of the protocol value is RARP/SNAP/X.25.
4323 *
4324 * NOTE: That is no longer true with the addition of VLAN tags. Not
4325 * sure which should go first, but I bet it won't make much
4326 * difference if we are running VLANs. The good news is that
4327 * this protocol won't be in the list unless compiled in, so
4328 * the average user (w/out VLANs) will not be adversely affected.
4329 * --BLG
4330 *
4331 * 0800 IP
4332 * 8100 802.1Q VLAN
4333 * 0001 802.3
4334 * 0002 AX.25
4335 * 0004 802.2
4336 * 8035 RARP
4337 * 0005 SNAP
4338 * 0805 X.25
4339 * 0806 ARP
4340 * 8137 IPX
4341 * 0009 Localtalk
4342 * 86DD IPv6
4343 */
4344#define PTYPE_HASH_SIZE (16)
4345#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4346
385a154c 4347#endif /* _LINUX_NETDEVICE_H */