]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
skb_array: add wrappers for resizing
[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;
7c46a640
AD
64/* UDP Tunnel offloads */
65struct udp_tunnel_info;
1da177e4 66
f629d208
JP
67void netdev_set_default_ethtool_ops(struct net_device *dev,
68 const struct ethtool_ops *ops);
d07d7507 69
9a1654ba
JP
70/* Backlog congestion levels */
71#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
72#define NET_RX_DROP 1 /* packet dropped */
73
572a9d7b
PM
74/*
75 * Transmit return codes: transmit return codes originate from three different
76 * namespaces:
77 *
78 * - qdisc return codes
79 * - driver transmit return codes
80 * - errno values
81 *
82 * Drivers are allowed to return any one of those in their hard_start_xmit()
83 * function. Real network devices commonly used with qdiscs should only return
84 * the driver transmit return codes though - when qdiscs are used, the actual
85 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
86 * higher layers. Virtual network devices transmit synchronously; in this case
87 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
88 * others are propagated to higher layers.
89 */
90
91/* qdisc ->enqueue() return codes. */
92#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
93#define NET_XMIT_DROP 0x01 /* skb dropped */
94#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 95#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 96
b9df3cb8
GR
97/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
98 * indicates that the device will soon be dropping packets, or already drops
99 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 100#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
101#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
102
dc1f8bf6 103/* Driver transmit return codes */
9a1654ba 104#define NETDEV_TX_MASK 0xf0
572a9d7b 105
dc1f8bf6 106enum netdev_tx {
572a9d7b 107 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
108 NETDEV_TX_OK = 0x00, /* driver took care of packet */
109 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
110};
111typedef enum netdev_tx netdev_tx_t;
112
9a1654ba
JP
113/*
114 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
115 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 */
117static inline bool dev_xmit_complete(int rc)
118{
119 /*
120 * Positive cases with an skb consumed by a driver:
121 * - successful transmission (rc == NETDEV_TX_OK)
122 * - error while transmitting (rc < 0)
123 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 */
125 if (likely(rc < NET_XMIT_MASK))
126 return true;
127
128 return false;
129}
130
1da177e4 131/*
5e82b4b2 132 * Compute the worst-case header length according to the protocols
1da177e4
LT
133 * used.
134 */
fe2918b0 135
c0eb4540
KS
136#if defined(CONFIG_HYPERV_NET)
137# define LL_MAX_HEADER 128
138#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
139# if defined(CONFIG_MAC80211_MESH)
140# define LL_MAX_HEADER 128
141# else
142# define LL_MAX_HEADER 96
143# endif
1da177e4 144#else
8388e3da 145# define LL_MAX_HEADER 32
1da177e4
LT
146#endif
147
d11ead75
BH
148#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
149 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
150#define MAX_HEADER LL_MAX_HEADER
151#else
152#define MAX_HEADER (LL_MAX_HEADER + 48)
153#endif
154
155/*
be1f3c2c
BH
156 * Old network device statistics. Fields are native words
157 * (unsigned long) so they can be read and written atomically.
1da177e4 158 */
fe2918b0 159
d94d9fee 160struct net_device_stats {
3cfde79c
BH
161 unsigned long rx_packets;
162 unsigned long tx_packets;
163 unsigned long rx_bytes;
164 unsigned long tx_bytes;
165 unsigned long rx_errors;
166 unsigned long tx_errors;
167 unsigned long rx_dropped;
168 unsigned long tx_dropped;
169 unsigned long multicast;
1da177e4 170 unsigned long collisions;
1da177e4 171 unsigned long rx_length_errors;
3cfde79c
BH
172 unsigned long rx_over_errors;
173 unsigned long rx_crc_errors;
174 unsigned long rx_frame_errors;
175 unsigned long rx_fifo_errors;
176 unsigned long rx_missed_errors;
1da177e4
LT
177 unsigned long tx_aborted_errors;
178 unsigned long tx_carrier_errors;
179 unsigned long tx_fifo_errors;
180 unsigned long tx_heartbeat_errors;
181 unsigned long tx_window_errors;
1da177e4
LT
182 unsigned long rx_compressed;
183 unsigned long tx_compressed;
184};
185
1da177e4
LT
186
187#include <linux/cache.h>
188#include <linux/skbuff.h>
189
adc9300e 190#ifdef CONFIG_RPS
c5905afb
IM
191#include <linux/static_key.h>
192extern struct static_key rps_needed;
adc9300e
ED
193#endif
194
1da177e4
LT
195struct neighbour;
196struct neigh_parms;
197struct sk_buff;
198
f001fde5
JP
199struct netdev_hw_addr {
200 struct list_head list;
201 unsigned char addr[MAX_ADDR_LEN];
202 unsigned char type;
ccffad25
JP
203#define NETDEV_HW_ADDR_T_LAN 1
204#define NETDEV_HW_ADDR_T_SAN 2
205#define NETDEV_HW_ADDR_T_SLAVE 3
206#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 207#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 208 bool global_use;
4cd729b0 209 int sync_cnt;
8f8f103d 210 int refcount;
4543fbef 211 int synced;
f001fde5
JP
212 struct rcu_head rcu_head;
213};
214
31278e71
JP
215struct netdev_hw_addr_list {
216 struct list_head list;
217 int count;
218};
219
22bedad3
JP
220#define netdev_hw_addr_list_count(l) ((l)->count)
221#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
222#define netdev_hw_addr_list_for_each(ha, l) \
223 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 224
22bedad3
JP
225#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
226#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
227#define netdev_for_each_uc_addr(ha, dev) \
228 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 229
22bedad3
JP
230#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
231#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 232#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 233 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 234
d94d9fee 235struct hh_cache {
f6b72b62 236 u16 hh_len;
5c25f686 237 u16 __pad;
3644f0ce 238 seqlock_t hh_lock;
1da177e4
LT
239
240 /* cached hardware header; allow for machine alignment needs. */
241#define HH_DATA_MOD 16
242#define HH_DATA_OFF(__len) \
5ba0eac6 243 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
244#define HH_DATA_ALIGN(__len) \
245 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
246 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
247};
248
5e82b4b2 249/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
250 * Alternative is:
251 * dev->hard_header_len ? (dev->hard_header_len +
252 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
253 *
254 * We could use other alignment values, but we must maintain the
255 * relationship HH alignment <= LL alignment.
256 */
257#define LL_RESERVED_SPACE(dev) \
f5184d26 258 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 259#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 260 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 261
3b04ddde
SH
262struct header_ops {
263 int (*create) (struct sk_buff *skb, struct net_device *dev,
264 unsigned short type, const void *daddr,
95c96174 265 const void *saddr, unsigned int len);
3b04ddde 266 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 267 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
268 void (*cache_update)(struct hh_cache *hh,
269 const struct net_device *dev,
270 const unsigned char *haddr);
2793a23a 271 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
272};
273
1da177e4 274/* These flag bits are private to the generic network queueing
5e82b4b2 275 * layer; they may not be explicitly referenced by any other
1da177e4
LT
276 * code.
277 */
278
d94d9fee 279enum netdev_state_t {
1da177e4
LT
280 __LINK_STATE_START,
281 __LINK_STATE_PRESENT,
1da177e4 282 __LINK_STATE_NOCARRIER,
b00055aa
SR
283 __LINK_STATE_LINKWATCH_PENDING,
284 __LINK_STATE_DORMANT,
1da177e4
LT
285};
286
287
288/*
5e82b4b2 289 * This structure holds boot-time configured netdevice settings. They
fe2918b0 290 * are then used in the device probing.
1da177e4
LT
291 */
292struct netdev_boot_setup {
293 char name[IFNAMSIZ];
294 struct ifmap map;
295};
296#define NETDEV_BOOT_SETUP_MAX 8
297
f629d208 298int __init netdev_boot_setup(char *str);
1da177e4 299
bea3348e
SH
300/*
301 * Structure for NAPI scheduling similar to tasklet but with weighting
302 */
303struct napi_struct {
304 /* The poll_list must only be managed by the entity which
305 * changes the state of the NAPI_STATE_SCHED bit. This means
306 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 307 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
308 * can remove from the list right before clearing the bit.
309 */
310 struct list_head poll_list;
311
312 unsigned long state;
313 int weight;
404f7c9e 314 unsigned int gro_count;
bea3348e
SH
315 int (*poll)(struct napi_struct *, int);
316#ifdef CONFIG_NETPOLL
317 spinlock_t poll_lock;
318 int poll_owner;
bea3348e 319#endif
5d38a079 320 struct net_device *dev;
d565b0a1 321 struct sk_buff *gro_list;
5d38a079 322 struct sk_buff *skb;
3b47d303 323 struct hrtimer timer;
404f7c9e 324 struct list_head dev_list;
af12fa6e
ET
325 struct hlist_node napi_hash_node;
326 unsigned int napi_id;
bea3348e
SH
327};
328
d94d9fee 329enum {
bea3348e 330 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 331 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 332 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
333 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
334 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
bea3348e
SH
335};
336
5b252f0c 337enum gro_result {
d1c76af9
HX
338 GRO_MERGED,
339 GRO_MERGED_FREE,
340 GRO_HELD,
341 GRO_NORMAL,
342 GRO_DROP,
343};
5b252f0c 344typedef enum gro_result gro_result_t;
d1c76af9 345
8a4eb573
JP
346/*
347 * enum rx_handler_result - Possible return values for rx_handlers.
348 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
349 * further.
350 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
351 * case skb->dev was changed by rx_handler.
352 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 353 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
354 *
355 * rx_handlers are functions called from inside __netif_receive_skb(), to do
356 * special processing of the skb, prior to delivery to protocol handlers.
357 *
358 * Currently, a net_device can only have a single rx_handler registered. Trying
359 * to register a second rx_handler will return -EBUSY.
360 *
361 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
362 * To unregister a rx_handler on a net_device, use
363 * netdev_rx_handler_unregister().
364 *
365 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
366 * do with the skb.
367 *
5e82b4b2 368 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 369 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 370 * the skb to be delivered in some other way.
8a4eb573
JP
371 *
372 * If the rx_handler changed skb->dev, to divert the skb to another
373 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
374 * new device will be called if it exists.
375 *
5e82b4b2 376 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 377 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 378 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 379 *
5e82b4b2 380 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
381 * delivered, it should return RX_HANDLER_PASS.
382 *
383 * A device without a registered rx_handler will behave as if rx_handler
384 * returned RX_HANDLER_PASS.
385 */
386
387enum rx_handler_result {
388 RX_HANDLER_CONSUMED,
389 RX_HANDLER_ANOTHER,
390 RX_HANDLER_EXACT,
391 RX_HANDLER_PASS,
392};
393typedef enum rx_handler_result rx_handler_result_t;
394typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 395
f629d208 396void __napi_schedule(struct napi_struct *n);
bc9ad166 397void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 398
4d29515f 399static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
400{
401 return test_bit(NAPI_STATE_DISABLE, &n->state);
402}
403
bea3348e 404/**
5e82b4b2
BH
405 * napi_schedule_prep - check if NAPI can be scheduled
406 * @n: NAPI context
bea3348e
SH
407 *
408 * Test if NAPI routine is already running, and if not mark
5e82b4b2 409 * it as running. This is used as a condition variable to
a0a46196
DM
410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
bea3348e 412 */
4d29515f 413static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 414{
a0a46196
DM
415 return !napi_disable_pending(n) &&
416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
417}
418
419/**
420 * napi_schedule - schedule NAPI poll
5e82b4b2 421 * @n: NAPI context
bea3348e
SH
422 *
423 * Schedule NAPI poll routine to be called if it is not already
424 * running.
425 */
426static inline void napi_schedule(struct napi_struct *n)
427{
428 if (napi_schedule_prep(n))
429 __napi_schedule(n);
430}
431
bc9ad166
ED
432/**
433 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 434 * @n: NAPI context
bc9ad166
ED
435 *
436 * Variant of napi_schedule(), assuming hard irqs are masked.
437 */
438static inline void napi_schedule_irqoff(struct napi_struct *n)
439{
440 if (napi_schedule_prep(n))
441 __napi_schedule_irqoff(n);
442}
443
bfe13f54 444/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 445static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
446{
447 if (napi_schedule_prep(napi)) {
448 __napi_schedule(napi);
4d29515f 449 return true;
bfe13f54 450 }
4d29515f 451 return false;
bfe13f54
RD
452}
453
3b47d303
ED
454void __napi_complete(struct napi_struct *n);
455void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
456/**
457 * napi_complete - NAPI processing complete
5e82b4b2 458 * @n: NAPI context
bea3348e
SH
459 *
460 * Mark NAPI processing as complete.
3b47d303 461 * Consider using napi_complete_done() instead.
bea3348e 462 */
3b47d303
ED
463static inline void napi_complete(struct napi_struct *n)
464{
465 return napi_complete_done(n, 0);
466}
bea3348e 467
af12fa6e
ET
468/**
469 * napi_hash_add - add a NAPI to global hashtable
5e82b4b2 470 * @napi: NAPI context
af12fa6e 471 *
5e82b4b2
BH
472 * Generate a new napi_id and store a @napi under it in napi_hash.
473 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
93d05d4a 474 * Note: This is normally automatically done from netif_napi_add(),
5e82b4b2 475 * so might disappear in a future Linux version.
af12fa6e 476 */
f629d208 477void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
478
479/**
480 * napi_hash_del - remove a NAPI from global table
5e82b4b2 481 * @napi: NAPI context
af12fa6e 482 *
5e82b4b2 483 * Warning: caller must observe RCU grace period
34cbe27e
ED
484 * before freeing memory containing @napi, if
485 * this function returns true.
93d05d4a 486 * Note: core networking stack automatically calls it
5e82b4b2 487 * from netif_napi_del().
93d05d4a 488 * Drivers might want to call this helper to combine all
5e82b4b2 489 * the needed RCU grace periods into a single one.
af12fa6e 490 */
34cbe27e 491bool napi_hash_del(struct napi_struct *napi);
af12fa6e 492
bea3348e
SH
493/**
494 * napi_disable - prevent NAPI from scheduling
5e82b4b2 495 * @n: NAPI context
bea3348e
SH
496 *
497 * Stop NAPI from being scheduled on this context.
498 * Waits till any outstanding processing completes.
499 */
3b47d303 500void napi_disable(struct napi_struct *n);
bea3348e
SH
501
502/**
503 * napi_enable - enable NAPI scheduling
5e82b4b2 504 * @n: NAPI context
bea3348e
SH
505 *
506 * Resume NAPI from being scheduled on this context.
507 * Must be paired with napi_disable.
508 */
509static inline void napi_enable(struct napi_struct *n)
510{
511 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 512 smp_mb__before_atomic();
bea3348e 513 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 514 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
515}
516
c264c3de
SH
517/**
518 * napi_synchronize - wait until NAPI is not running
5e82b4b2 519 * @n: NAPI context
c264c3de
SH
520 *
521 * Wait until NAPI is done being scheduled on this context.
522 * Waits till any outstanding processing completes but
523 * does not disable future activations.
524 */
525static inline void napi_synchronize(const struct napi_struct *n)
526{
facc432f
AB
527 if (IS_ENABLED(CONFIG_SMP))
528 while (test_bit(NAPI_STATE_SCHED, &n->state))
529 msleep(1);
530 else
531 barrier();
c264c3de 532}
c264c3de 533
d94d9fee 534enum netdev_queue_state_t {
73466498
TH
535 __QUEUE_STATE_DRV_XOFF,
536 __QUEUE_STATE_STACK_XOFF,
c3f26a26 537 __QUEUE_STATE_FROZEN,
79d16385 538};
8e2f1a63
DB
539
540#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
541#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
542#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
543
544#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
545#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
546 QUEUE_STATE_FROZEN)
547#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
548 QUEUE_STATE_FROZEN)
549
73466498
TH
550/*
551 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
552 * netif_tx_* functions below are used to manipulate this flag. The
553 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
554 * queue independently. The netif_xmit_*stopped functions below are called
555 * to check if the queue has been stopped by the driver or stack (either
556 * of the XOFF bits are set in the state). Drivers should not need to call
557 * netif_xmit*stopped functions, they should only be using netif_tx_*.
558 */
79d16385 559
bb949fbd 560struct netdev_queue {
6a321cb3 561/*
5e82b4b2 562 * read-mostly part
6a321cb3 563 */
bb949fbd 564 struct net_device *dev;
46e5da40 565 struct Qdisc __rcu *qdisc;
b0e1e646 566 struct Qdisc *qdisc_sleeping;
ccf5ff69 567#ifdef CONFIG_SYSFS
1d24eb48
TH
568 struct kobject kobj;
569#endif
f2cd2d3e
ED
570#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
571 int numa_node;
572#endif
c0ef079c
FW
573 unsigned long tx_maxrate;
574 /*
575 * Number of TX timeouts for this queue
576 * (/sys/class/net/DEV/Q/trans_timeout)
577 */
578 unsigned long trans_timeout;
6a321cb3 579/*
5e82b4b2 580 * write-mostly part
6a321cb3
ED
581 */
582 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
583 int xmit_lock_owner;
9d21493b 584 /*
9b36627a 585 * Time (in jiffies) of last Tx
9d21493b
ED
586 */
587 unsigned long trans_start;
ccf5ff69 588
114cf580
TH
589 unsigned long state;
590
591#ifdef CONFIG_BQL
592 struct dql dql;
593#endif
e8a0464c 594} ____cacheline_aligned_in_smp;
bb949fbd 595
f2cd2d3e
ED
596static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
597{
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 return q->numa_node;
600#else
b236da69 601 return NUMA_NO_NODE;
f2cd2d3e
ED
602#endif
603}
604
605static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
606{
607#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 q->numa_node = node;
609#endif
610}
611
df334545 612#ifdef CONFIG_RPS
0a9627f2
TH
613/*
614 * This structure holds an RPS map which can be of variable length. The
615 * map is an array of CPUs.
616 */
617struct rps_map {
618 unsigned int len;
619 struct rcu_head rcu;
620 u16 cpus[0];
621};
60b778ce 622#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 623
fec5e652 624/*
c445477d
BH
625 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
626 * tail pointer for that CPU's input queue at the time of last enqueue, and
627 * a hardware filter index.
fec5e652
TH
628 */
629struct rps_dev_flow {
630 u16 cpu;
c445477d 631 u16 filter;
fec5e652
TH
632 unsigned int last_qtail;
633};
c445477d 634#define RPS_NO_FILTER 0xffff
fec5e652
TH
635
636/*
637 * The rps_dev_flow_table structure contains a table of flow mappings.
638 */
639struct rps_dev_flow_table {
640 unsigned int mask;
641 struct rcu_head rcu;
fec5e652
TH
642 struct rps_dev_flow flows[0];
643};
644#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 645 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
646
647/*
648 * The rps_sock_flow_table contains mappings of flows to the last CPU
649 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
650 * Each entry is a 32bit value. Upper part is the high-order bits
651 * of flow hash, lower part is CPU number.
567e4b79 652 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
653 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
654 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 655 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
656 */
657struct rps_sock_flow_table {
567e4b79 658 u32 mask;
93c1af6c
ED
659
660 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 661};
567e4b79 662#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
663
664#define RPS_NO_CPU 0xffff
665
567e4b79
ED
666extern u32 rps_cpu_mask;
667extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
668
fec5e652
TH
669static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
670 u32 hash)
671{
672 if (table && hash) {
567e4b79
ED
673 unsigned int index = hash & table->mask;
674 u32 val = hash & ~rps_cpu_mask;
fec5e652 675
5e82b4b2 676 /* We only give a hint, preemption can change CPU under us */
567e4b79 677 val |= raw_smp_processor_id();
fec5e652 678
567e4b79
ED
679 if (table->ents[index] != val)
680 table->ents[index] = val;
fec5e652
TH
681 }
682}
683
c445477d 684#ifdef CONFIG_RFS_ACCEL
f629d208
JP
685bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
686 u16 filter_id);
c445477d 687#endif
a953be53 688#endif /* CONFIG_RPS */
c445477d 689
0a9627f2
TH
690/* This structure contains an instance of an RX queue. */
691struct netdev_rx_queue {
a953be53 692#ifdef CONFIG_RPS
6e3f7faf
ED
693 struct rps_map __rcu *rps_map;
694 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 695#endif
6e3f7faf 696 struct kobject kobj;
fe822240 697 struct net_device *dev;
0a9627f2 698} ____cacheline_aligned_in_smp;
a953be53
MD
699
700/*
701 * RX queue sysfs structures and functions.
702 */
703struct rx_queue_attribute {
704 struct attribute attr;
705 ssize_t (*show)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, char *buf);
707 ssize_t (*store)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, const char *buf, size_t len);
709};
d314774c 710
bf264145
TH
711#ifdef CONFIG_XPS
712/*
713 * This structure holds an XPS map which can be of variable length. The
714 * map is an array of queues.
715 */
716struct xps_map {
717 unsigned int len;
718 unsigned int alloc_len;
719 struct rcu_head rcu;
720 u16 queues[0];
721};
60b778ce 722#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
723#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
724 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
725
726/*
727 * This structure holds all XPS maps for device. Maps are indexed by CPU.
728 */
729struct xps_dev_maps {
730 struct rcu_head rcu;
a4177869 731 struct xps_map __rcu *cpu_map[0];
bf264145
TH
732};
733#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
734 (nr_cpu_ids * sizeof(struct xps_map *)))
735#endif /* CONFIG_XPS */
736
4f57c087
JF
737#define TC_MAX_QUEUE 16
738#define TC_BITMASK 15
739/* HW offloaded queuing disciplines txq count and offset maps */
740struct netdev_tc_txq {
741 u16 count;
742 u16 offset;
743};
744
68bad94e
NP
745#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
746/*
747 * This structure is to hold information about the device
748 * configured to run FCoE protocol stack.
749 */
750struct netdev_fcoe_hbainfo {
751 char manufacturer[64];
752 char serial_number[64];
753 char hardware_version[64];
754 char driver_version[64];
755 char optionrom_version[64];
756 char firmware_version[64];
757 char model[256];
758 char model_description[256];
759};
760#endif
761
02637fce 762#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 763
02637fce
JP
764/* This structure holds a unique identifier to identify some
765 * physical item (port for example) used by a netdevice.
66b52b0d 766 */
02637fce
JP
767struct netdev_phys_item_id {
768 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
769 unsigned char id_len;
770};
771
d754f98b
SF
772static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
773 struct netdev_phys_item_id *b)
774{
775 return a->id_len == b->id_len &&
776 memcmp(a->id, b->id, a->id_len) == 0;
777}
778
99932d4f
DB
779typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
780 struct sk_buff *skb);
781
a1b7c5fd 782/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
783 * that are being passed to the netdevice through the setup_tc op.
784 */
785enum {
786 TC_SETUP_MQPRIO,
a1b7c5fd 787 TC_SETUP_CLSU32,
5b33f488 788 TC_SETUP_CLSFLOWER,
16e5cc64
JF
789};
790
a1b7c5fd
JF
791struct tc_cls_u32_offload;
792
16e5cc64
JF
793struct tc_to_netdev {
794 unsigned int type;
795 union {
796 u8 tc;
a1b7c5fd 797 struct tc_cls_u32_offload *cls_u32;
5b33f488 798 struct tc_cls_flower_offload *cls_flower;
16e5cc64
JF
799 };
800};
801
802
d314774c
SH
803/*
804 * This structure defines the management hooks for network devices.
00829823
SH
805 * The following hooks can be defined; unless noted otherwise, they are
806 * optional and can be filled with a null pointer.
d314774c
SH
807 *
808 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
809 * This function is called once when a network device is registered.
810 * The network device can use this for any late stage initialization
811 * or semantic validation. It can fail with an error code which will
812 * be propagated back to register_netdev.
d314774c
SH
813 *
814 * void (*ndo_uninit)(struct net_device *dev);
815 * This function is called when device is unregistered or when registration
816 * fails. It is not called if init fails.
817 *
818 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 819 * This function is called when a network device transitions to the up
d314774c
SH
820 * state.
821 *
822 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 823 * This function is called when a network device transitions to the down
d314774c
SH
824 * state.
825 *
dc1f8bf6
SH
826 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
827 * struct net_device *dev);
00829823 828 * Called when a packet needs to be transmitted.
e79d8429
RR
829 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
830 * the queue before that can happen; it's for obsolete devices and weird
831 * corner cases, but the stack really does a non-trivial amount
832 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 833 * Required; cannot be NULL.
00829823 834 *
cdba756f
ED
835 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
836 * netdev_features_t features);
837 * Adjusts the requested feature flags according to device-specific
838 * constraints, and returns the resulting flags. Must not modify
839 * the device state.
840 *
f663dd9a 841 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 842 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 843 * Called to decide which queue to use when device supports multiple
00829823
SH
844 * transmit queues.
845 *
d314774c
SH
846 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
847 * This function is called to allow device receiver to make
5e82b4b2 848 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
849 *
850 * void (*ndo_set_rx_mode)(struct net_device *dev);
851 * This function is called device changes address list filtering.
01789349 852 * If driver handles unicast address filtering, it should set
5e82b4b2 853 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
854 *
855 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
856 * This function is called when the Media Access Control address
37b607c5 857 * needs to be changed. If this interface is not defined, the
5e82b4b2 858 * MAC address can not be changed.
d314774c
SH
859 *
860 * int (*ndo_validate_addr)(struct net_device *dev);
861 * Test if Media Access Control address is valid for the device.
862 *
863 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
864 * Called when a user requests an ioctl which can't be handled by
865 * the generic interface code. If not defined ioctls return
d314774c
SH
866 * not supported error code.
867 *
868 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
869 * Used to set network devices bus interface parameters. This interface
5e82b4b2 870 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
871 * interface (PCI) for low level management.
872 *
873 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
874 * Called when a user wants to change the Maximum Transfer Unit
875 * of a device. If not defined, any request to change MTU will
876 * will return an error.
877 *
00829823 878 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 879 * Callback used when the transmitter has not made any progress
d314774c
SH
880 * for dev->watchdog ticks.
881 *
3cfde79c 882 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 883 * struct rtnl_link_stats64 *storage);
d308e38f 884 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 885 * Called when a user wants to get the network device usage
be1f3c2c 886 * statistics. Drivers must do one of the following:
3cfde79c
BH
887 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
888 * rtnl_link_stats64 structure passed by the caller.
82695d9b 889 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
890 * (which should normally be dev->stats) and return a pointer to
891 * it. The structure may be changed asynchronously only if each
892 * field is written atomically.
893 * 3. Update dev->stats asynchronously and atomically, and define
894 * neither operation.
d314774c 895 *
5d632cb7 896 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 897 * If device supports VLAN filtering this function is called when a
80d5c368 898 * VLAN id is registered.
d314774c 899 *
5d632cb7 900 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 901 * If device supports VLAN filtering this function is called when a
80d5c368 902 * VLAN id is unregistered.
d314774c
SH
903 *
904 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
905 *
906 * SR-IOV management functions.
907 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
908 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
909 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
910 * int max_tx_rate);
5f8444a3 911 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 912 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
913 * int (*ndo_get_vf_config)(struct net_device *dev,
914 * int vf, struct ifla_vf_info *ivf);
1d8faf48 915 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
916 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
917 * struct nlattr *port[]);
01a3d796
VZ
918 *
919 * Enable or disable the VF ability to query its RSS Redirection Table and
920 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 921 * with PF and querying it may introduce a theoretical security risk.
01a3d796 922 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 923 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
924 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
925 * Called to setup 'tc' number of traffic classes in the net device. This
926 * is always called from the stack with the rtnl lock held and netif tx
927 * queues stopped. This allows the netdevice to perform queue management
928 * safely.
c445477d 929 *
e9bce845
YZ
930 * Fiber Channel over Ethernet (FCoE) offload functions.
931 * int (*ndo_fcoe_enable)(struct net_device *dev);
932 * Called when the FCoE protocol stack wants to start using LLD for FCoE
933 * so the underlying device can perform whatever needed configuration or
934 * initialization to support acceleration of FCoE traffic.
935 *
936 * int (*ndo_fcoe_disable)(struct net_device *dev);
937 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
938 * so the underlying device can perform whatever needed clean-ups to
939 * stop supporting acceleration of FCoE traffic.
940 *
941 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
942 * struct scatterlist *sgl, unsigned int sgc);
943 * Called when the FCoE Initiator wants to initialize an I/O that
944 * is a possible candidate for Direct Data Placement (DDP). The LLD can
945 * perform necessary setup and returns 1 to indicate the device is set up
946 * successfully to perform DDP on this I/O, otherwise this returns 0.
947 *
948 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
949 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
950 * indicated by the FC exchange id 'xid', so the underlying device can
951 * clean up and reuse resources for later DDP requests.
952 *
953 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
954 * struct scatterlist *sgl, unsigned int sgc);
955 * Called when the FCoE Target wants to initialize an I/O that
956 * is a possible candidate for Direct Data Placement (DDP). The LLD can
957 * perform necessary setup and returns 1 to indicate the device is set up
958 * successfully to perform DDP on this I/O, otherwise this returns 0.
959 *
68bad94e
NP
960 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
961 * struct netdev_fcoe_hbainfo *hbainfo);
962 * Called when the FCoE Protocol stack wants information on the underlying
963 * device. This information is utilized by the FCoE protocol stack to
964 * register attributes with Fiber Channel management service as per the
965 * FC-GS Fabric Device Management Information(FDMI) specification.
966 *
e9bce845
YZ
967 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
968 * Called when the underlying device wants to override default World Wide
969 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
970 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
971 * protocol stack to use.
972 *
c445477d
BH
973 * RFS acceleration.
974 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
975 * u16 rxq_index, u32 flow_id);
976 * Set hardware filter for RFS. rxq_index is the target queue index;
977 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
978 * Return the filter ID on success, or a negative error code.
fbaec0ea 979 *
8b98a70c 980 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
981 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
982 * Called to make another netdev an underling.
983 *
984 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
985 * Called to release previously enslaved netdev.
5455c699
MM
986 *
987 * Feature/offload setting functions.
c8f44aff 988 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
989 * Called to update device configuration to new features. Passed
990 * feature set might be less than what was returned by ndo_fix_features()).
991 * Must return >0 or -errno if it changed dev->features itself.
992 *
edc7d573 993 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
994 * struct net_device *dev,
f6f6424b 995 * const unsigned char *addr, u16 vid, u16 flags)
77162022 996 * Adds an FDB entry to dev for addr.
1690be63
VY
997 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
998 * struct net_device *dev,
f6f6424b 999 * const unsigned char *addr, u16 vid)
77162022
JF
1000 * Deletes the FDB entry from dev coresponding to addr.
1001 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
1002 * struct net_device *dev, struct net_device *filter_dev,
1003 * int idx)
77162022
JF
1004 * Used to add FDB entries to dump requests. Implementers should add
1005 * entries to skb and update idx with the number of entries.
e5a55a89 1006 *
ad41faa8
ND
1007 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1008 * u16 flags)
e5a55a89 1009 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1010 * struct net_device *dev, u32 filter_mask,
1011 * int nlflags)
ad41faa8
ND
1012 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1013 * u16 flags);
4bf84c35
JP
1014 *
1015 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1016 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1017 * which do not represent real hardware may define this to allow their
1018 * userspace components to manage their virtual carrier state. Devices
1019 * that determine carrier state from physical hardware properties (eg
1020 * network cables) or protocol-dependent mechanisms (eg
1021 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1022 *
1023 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1024 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1025 * Called to get ID of physical port of this device. If driver does
1026 * not implement this, it is assumed that the hw is not able to have
1027 * multiple net devices on single physical port.
53cf5275 1028 *
7c46a640
AD
1029 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1030 * struct udp_tunnel_info *ti);
1031 * Called by UDP tunnel to notify a driver about the UDP port and socket
1032 * address family that a UDP tunnel is listnening to. It is called only
1033 * when a new port starts listening. The operation is protected by the
1034 * RTNL.
1035 *
1036 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1037 * struct udp_tunnel_info *ti);
1038 * Called by UDP tunnel to notify the driver about a UDP port and socket
1039 * address family that the UDP tunnel is not listening to anymore. The
1040 * operation is protected by the RTNL.
1041 *
a6cc0cfa
JF
1042 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1043 * struct net_device *dev)
1044 * Called by upper layer devices to accelerate switching or other
1045 * station functionality into hardware. 'pdev is the lowerdev
1046 * to use for the offload and 'dev' is the net device that will
1047 * back the offload. Returns a pointer to the private structure
1048 * the upper layer will maintain.
1049 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1050 * Called by upper layer device to delete the station created
1051 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1052 * the station and priv is the structure returned by the add
1053 * operation.
1054 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1055 * struct net_device *dev,
1056 * void *priv);
1057 * Callback to use for xmit over the accelerated station. This
1058 * is used in place of ndo_start_xmit on accelerated net
1059 * devices.
5e82b4b2
BH
1060 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1061 * struct net_device *dev
1062 * netdev_features_t features);
04ffcb25 1063 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1064 * performing offload operations on a given packet. This is to give
1065 * the device an opportunity to implement any restrictions that cannot
1066 * be otherwise expressed by feature flags. The check is called with
1067 * the set of features that the stack has calculated and it returns
1068 * those the driver believes to be appropriate.
822b3b2e
JF
1069 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1070 * int queue_index, u32 maxrate);
1071 * Called when a user wants to set a max-rate limitation of specific
1072 * TX queue.
a54acb3a
ND
1073 * int (*ndo_get_iflink)(const struct net_device *dev);
1074 * Called to get the iflink value of this device.
d746d707 1075 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1076 * bool proto_down);
d746d707
AK
1077 * This function is used to pass protocol port error state information
1078 * to the switch driver. The switch driver can react to the proto_down
1079 * by doing a phys down on the associated switch port.
fc4099f1
PS
1080 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1081 * This function is used to get egress tunnel information for given skb.
1082 * This is useful for retrieving outer tunnel header parameters while
1083 * sampling packet.
871b642a
PA
1084 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1085 * This function is used to specify the headroom that the skb must
1086 * consider when allocation skb during packet reception. Setting
1087 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1088 * forward. Setting a negative value resets the rx headroom to the
871b642a 1089 * default value.
d746d707 1090 *
d314774c
SH
1091 */
1092struct net_device_ops {
1093 int (*ndo_init)(struct net_device *dev);
1094 void (*ndo_uninit)(struct net_device *dev);
1095 int (*ndo_open)(struct net_device *dev);
1096 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1097 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1098 struct net_device *dev);
1099 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1100 struct net_device *dev,
1101 netdev_features_t features);
00829823 1102 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1103 struct sk_buff *skb,
99932d4f
DB
1104 void *accel_priv,
1105 select_queue_fallback_t fallback);
d314774c
SH
1106 void (*ndo_change_rx_flags)(struct net_device *dev,
1107 int flags);
d314774c 1108 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1109 int (*ndo_set_mac_address)(struct net_device *dev,
1110 void *addr);
d314774c 1111 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1112 int (*ndo_do_ioctl)(struct net_device *dev,
1113 struct ifreq *ifr, int cmd);
d314774c
SH
1114 int (*ndo_set_config)(struct net_device *dev,
1115 struct ifmap *map);
00829823
SH
1116 int (*ndo_change_mtu)(struct net_device *dev,
1117 int new_mtu);
1118 int (*ndo_neigh_setup)(struct net_device *dev,
1119 struct neigh_parms *);
d314774c
SH
1120 void (*ndo_tx_timeout) (struct net_device *dev);
1121
28172739
ED
1122 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1123 struct rtnl_link_stats64 *storage);
d314774c
SH
1124 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1125
8e586137 1126 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1127 __be16 proto, u16 vid);
8e586137 1128 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1129 __be16 proto, u16 vid);
d314774c 1130#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1131 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1132 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1133 struct netpoll_info *info);
0e34e931 1134 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1135#endif
e0d1095a 1136#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1137 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1138#endif
95c26df8
WM
1139 int (*ndo_set_vf_mac)(struct net_device *dev,
1140 int queue, u8 *mac);
1141 int (*ndo_set_vf_vlan)(struct net_device *dev,
1142 int queue, u16 vlan, u8 qos);
ed616689
SC
1143 int (*ndo_set_vf_rate)(struct net_device *dev,
1144 int vf, int min_tx_rate,
1145 int max_tx_rate);
5f8444a3
GR
1146 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1147 int vf, bool setting);
dd461d6a
HS
1148 int (*ndo_set_vf_trust)(struct net_device *dev,
1149 int vf, bool setting);
95c26df8
WM
1150 int (*ndo_get_vf_config)(struct net_device *dev,
1151 int vf,
1152 struct ifla_vf_info *ivf);
1d8faf48
RE
1153 int (*ndo_set_vf_link_state)(struct net_device *dev,
1154 int vf, int link_state);
3b766cd8
EBE
1155 int (*ndo_get_vf_stats)(struct net_device *dev,
1156 int vf,
1157 struct ifla_vf_stats
1158 *vf_stats);
57b61080
SF
1159 int (*ndo_set_vf_port)(struct net_device *dev,
1160 int vf,
1161 struct nlattr *port[]);
1162 int (*ndo_get_vf_port)(struct net_device *dev,
1163 int vf, struct sk_buff *skb);
cc8e27cc
EC
1164 int (*ndo_set_vf_guid)(struct net_device *dev,
1165 int vf, u64 guid,
1166 int guid_type);
01a3d796
VZ
1167 int (*ndo_set_vf_rss_query_en)(
1168 struct net_device *dev,
1169 int vf, bool setting);
16e5cc64
JF
1170 int (*ndo_setup_tc)(struct net_device *dev,
1171 u32 handle,
1172 __be16 protocol,
1173 struct tc_to_netdev *tc);
d11ead75 1174#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1175 int (*ndo_fcoe_enable)(struct net_device *dev);
1176 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1177 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1178 u16 xid,
1179 struct scatterlist *sgl,
1180 unsigned int sgc);
1181 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1182 u16 xid);
6247e086
YZ
1183 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1184 u16 xid,
1185 struct scatterlist *sgl,
1186 unsigned int sgc);
68bad94e
NP
1187 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1188 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1189#endif
1190
d11ead75 1191#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1192#define NETDEV_FCOE_WWNN 0
1193#define NETDEV_FCOE_WWPN 1
1194 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1195 u64 *wwn, int type);
4d288d57 1196#endif
3c9c36bc 1197
c445477d
BH
1198#ifdef CONFIG_RFS_ACCEL
1199 int (*ndo_rx_flow_steer)(struct net_device *dev,
1200 const struct sk_buff *skb,
1201 u16 rxq_index,
1202 u32 flow_id);
1203#endif
fbaec0ea
JP
1204 int (*ndo_add_slave)(struct net_device *dev,
1205 struct net_device *slave_dev);
1206 int (*ndo_del_slave)(struct net_device *dev,
1207 struct net_device *slave_dev);
c8f44aff
MM
1208 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1209 netdev_features_t features);
5455c699 1210 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1211 netdev_features_t features);
da6a8fa0 1212 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1213 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1214
1215 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1216 struct nlattr *tb[],
77162022 1217 struct net_device *dev,
6b6e2725 1218 const unsigned char *addr,
f6f6424b 1219 u16 vid,
77162022
JF
1220 u16 flags);
1221 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1222 struct nlattr *tb[],
77162022 1223 struct net_device *dev,
f6f6424b
JP
1224 const unsigned char *addr,
1225 u16 vid);
77162022
JF
1226 int (*ndo_fdb_dump)(struct sk_buff *skb,
1227 struct netlink_callback *cb,
1228 struct net_device *dev,
5d5eacb3 1229 struct net_device *filter_dev,
77162022 1230 int idx);
e5a55a89
JF
1231
1232 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1233 struct nlmsghdr *nlh,
1234 u16 flags);
e5a55a89
JF
1235 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1236 u32 pid, u32 seq,
6cbdceeb 1237 struct net_device *dev,
46c264da
ND
1238 u32 filter_mask,
1239 int nlflags);
407af329 1240 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1241 struct nlmsghdr *nlh,
1242 u16 flags);
4bf84c35
JP
1243 int (*ndo_change_carrier)(struct net_device *dev,
1244 bool new_carrier);
66b52b0d 1245 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1246 struct netdev_phys_item_id *ppid);
db24a904
DA
1247 int (*ndo_get_phys_port_name)(struct net_device *dev,
1248 char *name, size_t len);
7c46a640
AD
1249 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1250 struct udp_tunnel_info *ti);
1251 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1252 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1253 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1254 struct net_device *dev);
1255 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1256 void *priv);
1257
1258 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1259 struct net_device *dev,
1260 void *priv);
25175ba5 1261 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1262 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1263 int queue_index,
1264 u32 maxrate);
a54acb3a 1265 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1266 int (*ndo_change_proto_down)(struct net_device *dev,
1267 bool proto_down);
fc4099f1
PS
1268 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1269 struct sk_buff *skb);
871b642a
PA
1270 void (*ndo_set_rx_headroom)(struct net_device *dev,
1271 int needed_headroom);
d314774c
SH
1272};
1273
7aa98047
LR
1274/**
1275 * enum net_device_priv_flags - &struct net_device priv_flags
1276 *
1277 * These are the &struct net_device, they are only set internally
1278 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1279 * userspace; this means that the order of these flags can change
7aa98047
LR
1280 * during any kernel release.
1281 *
1282 * You should have a pretty good reason to be extending these flags.
1283 *
1284 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1285 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1286 * @IFF_BONDING: bonding master or slave
7aa98047 1287 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1288 * @IFF_WAN_HDLC: WAN HDLC device
1289 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1290 * release skb->dst
1291 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1292 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1293 * @IFF_MACVLAN_PORT: device used as macvlan port
1294 * @IFF_BRIDGE_PORT: device used as bridge port
1295 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1296 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1297 * @IFF_UNICAST_FLT: Supports unicast filtering
1298 * @IFF_TEAM_PORT: device used as team port
1299 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1300 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1301 * change when it's running
1302 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1303 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1304 * underlying stacked devices
1305 * @IFF_IPVLAN_MASTER: IPvlan master device
1306 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1307 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1308 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1309 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1310 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1311 * @IFF_TEAM: device is a team device
d4ab4286 1312 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1313 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1314 * entity (i.e. the master device for bridged veth)
3c175784 1315 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1316 */
1317enum netdev_priv_flags {
1318 IFF_802_1Q_VLAN = 1<<0,
1319 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1320 IFF_BONDING = 1<<2,
1321 IFF_ISATAP = 1<<3,
1322 IFF_WAN_HDLC = 1<<4,
1323 IFF_XMIT_DST_RELEASE = 1<<5,
1324 IFF_DONT_BRIDGE = 1<<6,
1325 IFF_DISABLE_NETPOLL = 1<<7,
1326 IFF_MACVLAN_PORT = 1<<8,
1327 IFF_BRIDGE_PORT = 1<<9,
1328 IFF_OVS_DATAPATH = 1<<10,
1329 IFF_TX_SKB_SHARING = 1<<11,
1330 IFF_UNICAST_FLT = 1<<12,
1331 IFF_TEAM_PORT = 1<<13,
1332 IFF_SUPP_NOFCS = 1<<14,
1333 IFF_LIVE_ADDR_CHANGE = 1<<15,
1334 IFF_MACVLAN = 1<<16,
1335 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1336 IFF_IPVLAN_MASTER = 1<<18,
1337 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1338 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1339 IFF_NO_QUEUE = 1<<21,
1340 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1341 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1342 IFF_TEAM = 1<<24,
d4ab4286 1343 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1344 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1345 IFF_MACSEC = 1<<27,
7aa98047
LR
1346};
1347
1348#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1349#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1350#define IFF_BONDING IFF_BONDING
7aa98047 1351#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1352#define IFF_WAN_HDLC IFF_WAN_HDLC
1353#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1354#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1355#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1356#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1357#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1358#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1359#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1360#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1361#define IFF_TEAM_PORT IFF_TEAM_PORT
1362#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1363#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1364#define IFF_MACVLAN IFF_MACVLAN
02875878 1365#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1366#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1367#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1368#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1369#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1370#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1371#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1372#define IFF_TEAM IFF_TEAM
d4ab4286 1373#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1374#define IFF_MACSEC IFF_MACSEC
7aa98047 1375
536721b1
KK
1376/**
1377 * struct net_device - The DEVICE structure.
1378 * Actually, this whole structure is a big mistake. It mixes I/O
1379 * data with strictly "high-level" data, and it has to know about
1380 * almost every data structure used in the INET module.
1381 *
1382 * @name: This is the first field of the "visible" part of this structure
1383 * (i.e. as seen by users in the "Space.c" file). It is the name
1384 * of the interface.
1385 *
1386 * @name_hlist: Device name hash chain, please keep it close to name[]
1387 * @ifalias: SNMP alias
1388 * @mem_end: Shared memory end
1389 * @mem_start: Shared memory start
1390 * @base_addr: Device I/O address
1391 * @irq: Device IRQ number
1392 *
14ffbbb8
TG
1393 * @carrier_changes: Stats to monitor carrier on<->off transitions
1394 *
536721b1
KK
1395 * @state: Generic network queuing layer state, see netdev_state_t
1396 * @dev_list: The global list of network devices
5e82b4b2
BH
1397 * @napi_list: List entry used for polling NAPI devices
1398 * @unreg_list: List entry when we are unregistering the
1399 * device; see the function unregister_netdev
1400 * @close_list: List entry used when we are closing the device
62d885fe
BP
1401 * @ptype_all: Device-specific packet handlers for all protocols
1402 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1403 *
1404 * @adj_list: Directly linked devices, like slaves for bonding
1405 * @all_adj_list: All linked devices, *including* neighbours
1406 * @features: Currently active device features
1407 * @hw_features: User-changeable features
1408 *
1409 * @wanted_features: User-requested features
1410 * @vlan_features: Mask of features inheritable by VLAN devices
1411 *
1412 * @hw_enc_features: Mask of features inherited by encapsulating devices
1413 * This field indicates what encapsulation
1414 * offloads the hardware is capable of doing,
1415 * and drivers will need to set them appropriately.
1416 *
1417 * @mpls_features: Mask of features inheritable by MPLS
1418 *
1419 * @ifindex: interface index
5e82b4b2 1420 * @group: The group the device belongs to
536721b1
KK
1421 *
1422 * @stats: Statistics struct, which was left as a legacy, use
1423 * rtnl_link_stats64 instead
1424 *
1425 * @rx_dropped: Dropped packets by core network,
1426 * do not use this in drivers
1427 * @tx_dropped: Dropped packets by core network,
1428 * do not use this in drivers
6e7333d3
JW
1429 * @rx_nohandler: nohandler dropped packets by core network on
1430 * inactive devices, do not use this in drivers
536721b1 1431 *
536721b1
KK
1432 * @wireless_handlers: List of functions to handle Wireless Extensions,
1433 * instead of ioctl,
1434 * see <net/iw_handler.h> for details.
1435 * @wireless_data: Instance data managed by the core of wireless extensions
1436 *
1437 * @netdev_ops: Includes several pointers to callbacks,
1438 * if one wants to override the ndo_*() functions
1439 * @ethtool_ops: Management operations
f997c55c
AA
1440 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1441 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1442 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1443 * of Layer 2 headers.
1444 *
1445 * @flags: Interface flags (a la BSD)
1446 * @priv_flags: Like 'flags' but invisible to userspace,
1447 * see if.h for the definitions
1448 * @gflags: Global flags ( kept as legacy )
1449 * @padded: How much padding added by alloc_netdev()
1450 * @operstate: RFC2863 operstate
1451 * @link_mode: Mapping policy to operstate
1452 * @if_port: Selectable AUI, TP, ...
1453 * @dma: DMA channel
1454 * @mtu: Interface MTU value
1455 * @type: Interface hardware type
2793a23a 1456 * @hard_header_len: Maximum hardware header length.
536721b1
KK
1457 *
1458 * @needed_headroom: Extra headroom the hardware may need, but not in all
1459 * cases can this be guaranteed
1460 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1461 * cases can this be guaranteed. Some cases also use
1462 * LL_MAX_HEADER instead to allocate the skb
1463 *
1464 * interface address info:
1465 *
1466 * @perm_addr: Permanent hw address
1467 * @addr_assign_type: Hw address assignment type
1468 * @addr_len: Hardware address length
8626a0c8 1469 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1470 * @dev_id: Used to differentiate devices that share
1471 * the same link layer address
1472 * @dev_port: Used to differentiate devices that share
1473 * the same function
1474 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1475 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1476 * has been enabled due to the need to listen to
1477 * additional unicast addresses in a device that
1478 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1479 * @uc: unicast mac addresses
1480 * @mc: multicast mac addresses
1481 * @dev_addrs: list of device hw addresses
1482 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1483 * @promiscuity: Number of times the NIC is told to work in
1484 * promiscuous mode; if it becomes 0 the NIC will
1485 * exit promiscuous mode
536721b1
KK
1486 * @allmulti: Counter, enables or disables allmulticast mode
1487 *
1488 * @vlan_info: VLAN info
1489 * @dsa_ptr: dsa specific data
1490 * @tipc_ptr: TIPC specific data
1491 * @atalk_ptr: AppleTalk link
1492 * @ip_ptr: IPv4 specific data
1493 * @dn_ptr: DECnet specific data
1494 * @ip6_ptr: IPv6 specific data
1495 * @ax25_ptr: AX.25 specific data
1496 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1497 *
1498 * @last_rx: Time of last Rx
1499 * @dev_addr: Hw address (before bcast,
1500 * because most packets are unicast)
1501 *
1502 * @_rx: Array of RX queues
1503 * @num_rx_queues: Number of RX queues
1504 * allocated at register_netdev() time
1505 * @real_num_rx_queues: Number of RX queues currently active in device
1506 *
1507 * @rx_handler: handler for received packets
1508 * @rx_handler_data: XXX: need comments on this one
1509 * @ingress_queue: XXX: need comments on this one
1510 * @broadcast: hw bcast address
1511 *
14ffbbb8
TG
1512 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1513 * indexed by RX queue number. Assigned by driver.
1514 * This must only be set if the ndo_rx_flow_steer
1515 * operation is defined
1516 * @index_hlist: Device index hash chain
1517 *
536721b1
KK
1518 * @_tx: Array of TX queues
1519 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1520 * @real_num_tx_queues: Number of TX queues currently active in device
1521 * @qdisc: Root qdisc from userspace point of view
1522 * @tx_queue_len: Max frames per queue allowed
1523 * @tx_global_lock: XXX: need comments on this one
1524 *
1525 * @xps_maps: XXX: need comments on this one
1526 *
0c4f691f
SF
1527 * @offload_fwd_mark: Offload device fwding mark
1528 *
536721b1 1529 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1530 * the watchdog (see dev_watchdog())
536721b1
KK
1531 * @watchdog_timer: List of timers
1532 *
1533 * @pcpu_refcnt: Number of references to this device
1534 * @todo_list: Delayed register/unregister
536721b1
KK
1535 * @link_watch_list: XXX: need comments on this one
1536 *
1537 * @reg_state: Register/unregister state machine
1538 * @dismantle: Device is going to be freed
1539 * @rtnl_link_state: This enum represents the phases of creating
1540 * a new link
1541 *
1542 * @destructor: Called from unregister,
1543 * can be used to call free_netdev
1544 * @npinfo: XXX: need comments on this one
1545 * @nd_net: Network namespace this network device is inside
1546 *
1547 * @ml_priv: Mid-layer private
1548 * @lstats: Loopback statistics
1549 * @tstats: Tunnel statistics
1550 * @dstats: Dummy statistics
1551 * @vstats: Virtual ethernet statistics
1552 *
1553 * @garp_port: GARP
1554 * @mrp_port: MRP
1555 *
1556 * @dev: Class/net/name entry
1557 * @sysfs_groups: Space for optional device, statistics and wireless
1558 * sysfs groups
1559 *
1560 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1561 * @rtnl_link_ops: Rtnl_link_ops
1562 *
1563 * @gso_max_size: Maximum size of generic segmentation offload
1564 * @gso_max_segs: Maximum number of segments that can be passed to the
1565 * NIC for GSO
1566 *
1567 * @dcbnl_ops: Data Center Bridging netlink ops
1568 * @num_tc: Number of traffic classes in the net device
1569 * @tc_to_txq: XXX: need comments on this one
1570 * @prio_tc_map XXX: need comments on this one
1571 *
1572 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1573 *
1574 * @priomap: XXX: need comments on this one
1575 * @phydev: Physical device may attach itself
1576 * for hardware timestamping
1577 *
123b3652
ED
1578 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1579 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1580 *
d746d707
AK
1581 * @proto_down: protocol port state information can be sent to the
1582 * switch driver and used to set the phys state of the
1583 * switch port.
1584 *
1da177e4
LT
1585 * FIXME: cleanup struct net_device such that network protocol info
1586 * moves out.
1587 */
1588
d94d9fee 1589struct net_device {
1da177e4 1590 char name[IFNAMSIZ];
9356b8fc 1591 struct hlist_node name_hlist;
0b815a1a 1592 char *ifalias;
1da177e4
LT
1593 /*
1594 * I/O specific fields
1595 * FIXME: Merge these and struct ifmap into one
1596 */
536721b1
KK
1597 unsigned long mem_end;
1598 unsigned long mem_start;
1599 unsigned long base_addr;
1600 int irq;
1da177e4 1601
14ffbbb8
TG
1602 atomic_t carrier_changes;
1603
1da177e4 1604 /*
536721b1
KK
1605 * Some hardware also needs these fields (state,dev_list,
1606 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1607 * part of the usual set specified in Space.c.
1608 */
1609
1da177e4
LT
1610 unsigned long state;
1611
7562f876 1612 struct list_head dev_list;
bea3348e 1613 struct list_head napi_list;
44a0873d 1614 struct list_head unreg_list;
5cde2829 1615 struct list_head close_list;
7866a621
SN
1616 struct list_head ptype_all;
1617 struct list_head ptype_specific;
2f268f12 1618
2f268f12
VF
1619 struct {
1620 struct list_head upper;
1621 struct list_head lower;
1622 } adj_list;
1623
2f268f12
VF
1624 struct {
1625 struct list_head upper;
1626 struct list_head lower;
1627 } all_adj_list;
4c3d5e7b 1628
c8f44aff 1629 netdev_features_t features;
c8f44aff 1630 netdev_features_t hw_features;
c8f44aff 1631 netdev_features_t wanted_features;
c8f44aff 1632 netdev_features_t vlan_features;
6a674e9c 1633 netdev_features_t hw_enc_features;
0d89d203 1634 netdev_features_t mpls_features;
802ab55a 1635 netdev_features_t gso_partial_features;
04ed3e74 1636
1da177e4 1637 int ifindex;
7a66bbc9 1638 int group;
1da177e4 1639
c45d286e 1640 struct net_device_stats stats;
015f0688 1641
015f0688
ED
1642 atomic_long_t rx_dropped;
1643 atomic_long_t tx_dropped;
6e7333d3 1644 atomic_long_t rx_nohandler;
1da177e4 1645
b86e0280 1646#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1647 const struct iw_handler_def *wireless_handlers;
1648 struct iw_public_data *wireless_data;
b86e0280 1649#endif
d314774c 1650 const struct net_device_ops *netdev_ops;
76fd8593 1651 const struct ethtool_ops *ethtool_ops;
4170604f 1652#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1653 const struct switchdev_ops *switchdev_ops;
4170604f 1654#endif
1b69c6d0
DA
1655#ifdef CONFIG_NET_L3_MASTER_DEV
1656 const struct l3mdev_ops *l3mdev_ops;
1657#endif
f997c55c
AA
1658#if IS_ENABLED(CONFIG_IPV6)
1659 const struct ndisc_ops *ndisc_ops;
1660#endif
1da177e4 1661
3b04ddde
SH
1662 const struct header_ops *header_ops;
1663
536721b1
KK
1664 unsigned int flags;
1665 unsigned int priv_flags;
1666
1da177e4 1667 unsigned short gflags;
536721b1 1668 unsigned short padded;
1da177e4 1669
536721b1
KK
1670 unsigned char operstate;
1671 unsigned char link_mode;
b00055aa 1672
536721b1
KK
1673 unsigned char if_port;
1674 unsigned char dma;
bdc220da 1675
536721b1
KK
1676 unsigned int mtu;
1677 unsigned short type;
1678 unsigned short hard_header_len;
1da177e4 1679
f5184d26
JB
1680 unsigned short needed_headroom;
1681 unsigned short needed_tailroom;
1682
1da177e4 1683 /* Interface address info. */
536721b1
KK
1684 unsigned char perm_addr[MAX_ADDR_LEN];
1685 unsigned char addr_assign_type;
1686 unsigned char addr_len;
a0a9663d 1687 unsigned short neigh_priv_len;
536721b1
KK
1688 unsigned short dev_id;
1689 unsigned short dev_port;
ccffad25 1690 spinlock_t addr_list_lock;
14ffbbb8
TG
1691 unsigned char name_assign_type;
1692 bool uc_promisc;
536721b1
KK
1693 struct netdev_hw_addr_list uc;
1694 struct netdev_hw_addr_list mc;
1695 struct netdev_hw_addr_list dev_addrs;
1696
4c3d5e7b
ED
1697#ifdef CONFIG_SYSFS
1698 struct kset *queues_kset;
1699#endif
9d45abe1
WC
1700 unsigned int promiscuity;
1701 unsigned int allmulti;
1da177e4 1702
1da177e4 1703
5e82b4b2 1704 /* Protocol-specific pointers */
65ac6a5f 1705
d11ead75 1706#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1707 struct vlan_info __rcu *vlan_info;
65ac6a5f 1708#endif
34a430d7 1709#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1710 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1711#endif
1712#if IS_ENABLED(CONFIG_TIPC)
536721b1 1713 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1714#endif
536721b1
KK
1715 void *atalk_ptr;
1716 struct in_device __rcu *ip_ptr;
1717 struct dn_dev __rcu *dn_ptr;
1718 struct inet6_dev __rcu *ip6_ptr;
1719 void *ax25_ptr;
1720 struct wireless_dev *ieee80211_ptr;
98a18b6f 1721 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1722#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1723 struct mpls_dev __rcu *mpls_ptr;
1724#endif
1da177e4 1725
9356b8fc 1726/*
cd13539b 1727 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1728 */
536721b1 1729 unsigned long last_rx;
4dc89133 1730
9356b8fc 1731 /* Interface address info used in eth_type_trans() */
536721b1 1732 unsigned char *dev_addr;
f001fde5 1733
a953be53 1734#ifdef CONFIG_SYSFS
0a9627f2
TH
1735 struct netdev_rx_queue *_rx;
1736
0a9627f2 1737 unsigned int num_rx_queues;
62fe0b40 1738 unsigned int real_num_rx_queues;
df334545 1739#endif
0a9627f2 1740
3b47d303 1741 unsigned long gro_flush_timeout;
61391cde 1742 rx_handler_func_t __rcu *rx_handler;
1743 void __rcu *rx_handler_data;
e8a0464c 1744
4cda01e8 1745#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1746 struct tcf_proto __rcu *ingress_cl_list;
1747#endif
24824a09 1748 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1749#ifdef CONFIG_NETFILTER_INGRESS
1750 struct list_head nf_hooks_ingress;
1751#endif
d2788d34 1752
536721b1 1753 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1754#ifdef CONFIG_RFS_ACCEL
1755 struct cpu_rmap *rx_cpu_rmap;
1756#endif
1757 struct hlist_node index_hlist;
cd13539b
ED
1758
1759/*
1760 * Cache lines mostly used on transmit path
1761 */
e8a0464c
DM
1762 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1763 unsigned int num_tx_queues;
fd2ea0a7 1764 unsigned int real_num_tx_queues;
af356afa 1765 struct Qdisc *qdisc;
536721b1 1766 unsigned long tx_queue_len;
c3f26a26 1767 spinlock_t tx_global_lock;
14ffbbb8 1768 int watchdog_timeo;
cd13539b 1769
bf264145 1770#ifdef CONFIG_XPS
a4177869 1771 struct xps_dev_maps __rcu *xps_maps;
bf264145 1772#endif
1f211a1b
DB
1773#ifdef CONFIG_NET_CLS_ACT
1774 struct tcf_proto __rcu *egress_cl_list;
1775#endif
0c4f691f
SF
1776#ifdef CONFIG_NET_SWITCHDEV
1777 u32 offload_fwd_mark;
1778#endif
1779
9356b8fc 1780 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1781 struct timer_list watchdog_timer;
1782
29b4433d 1783 int __percpu *pcpu_refcnt;
1da177e4 1784 struct list_head todo_list;
1da177e4 1785
e014debe 1786 struct list_head link_watch_list;
572a103d 1787
1da177e4 1788 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1789 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1790 NETREG_UNREGISTERING, /* called unregister_netdevice */
1791 NETREG_UNREGISTERED, /* completed unregister todo */
1792 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1793 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1794 } reg_state:8;
1795
536721b1 1796 bool dismantle;
a2835763
PM
1797
1798 enum {
1799 RTNL_LINK_INITIALIZED,
1800 RTNL_LINK_INITIALIZING,
1801 } rtnl_link_state:16;
1da177e4 1802
d314774c 1803 void (*destructor)(struct net_device *dev);
1da177e4 1804
1da177e4 1805#ifdef CONFIG_NETPOLL
5fbee843 1806 struct netpoll_info __rcu *npinfo;
1da177e4 1807#endif
eae792b7 1808
0c5c9fb5 1809 possible_net_t nd_net;
4a1c5371 1810
4951704b 1811 /* mid-layer private */
a7855c78 1812 union {
536721b1
KK
1813 void *ml_priv;
1814 struct pcpu_lstats __percpu *lstats;
8f84985f 1815 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1816 struct pcpu_dstats __percpu *dstats;
1817 struct pcpu_vstats __percpu *vstats;
a7855c78 1818 };
536721b1 1819
3cc77ec7 1820 struct garp_port __rcu *garp_port;
febf018d 1821 struct mrp_port __rcu *mrp_port;
1da177e4 1822
5e82b4b2 1823 struct device dev;
0c509a6c 1824 const struct attribute_group *sysfs_groups[4];
a953be53 1825 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1826
38f7b870 1827 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1828
82cc1a7a
PWJ
1829 /* for setting kernel sock attribute on TCP connection setup */
1830#define GSO_MAX_SIZE 65536
1831 unsigned int gso_max_size;
30b678d8
BH
1832#define GSO_MAX_SEGS 65535
1833 u16 gso_max_segs;
743b03a8 1834
7a6b6f51 1835#ifdef CONFIG_DCB
32953543 1836 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1837#endif
5e82b4b2
BH
1838 u8 num_tc;
1839 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1840 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1841
d11ead75 1842#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1843 unsigned int fcoe_ddp_xid;
5bc1421e 1844#endif
86f8515f 1845#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1846 struct netprio_map __rcu *priomap;
4d288d57 1847#endif
5e82b4b2
BH
1848 struct phy_device *phydev;
1849 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1850 struct lock_class_key *qdisc_running_key;
5e82b4b2 1851 bool proto_down;
1da177e4 1852};
43cb76d9 1853#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1854
1855#define NETDEV_ALIGN 32
1da177e4 1856
4f57c087
JF
1857static inline
1858int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1859{
1860 return dev->prio_tc_map[prio & TC_BITMASK];
1861}
1862
1863static inline
1864int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1865{
1866 if (tc >= dev->num_tc)
1867 return -EINVAL;
1868
1869 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1870 return 0;
1871}
1872
1873static inline
1874void netdev_reset_tc(struct net_device *dev)
1875{
1876 dev->num_tc = 0;
1877 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1878 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1879}
1880
1881static inline
1882int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1883{
1884 if (tc >= dev->num_tc)
1885 return -EINVAL;
1886
1887 dev->tc_to_txq[tc].count = count;
1888 dev->tc_to_txq[tc].offset = offset;
1889 return 0;
1890}
1891
1892static inline
1893int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1894{
1895 if (num_tc > TC_MAX_QUEUE)
1896 return -EINVAL;
1897
1898 dev->num_tc = num_tc;
1899 return 0;
1900}
1901
1902static inline
1903int netdev_get_num_tc(struct net_device *dev)
1904{
1905 return dev->num_tc;
1906}
1907
e8a0464c
DM
1908static inline
1909struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1910 unsigned int index)
1911{
1912 return &dev->_tx[index];
1913}
1914
10c51b56
DB
1915static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1916 const struct sk_buff *skb)
1917{
1918 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1919}
1920
e8a0464c
DM
1921static inline void netdev_for_each_tx_queue(struct net_device *dev,
1922 void (*f)(struct net_device *,
1923 struct netdev_queue *,
1924 void *),
1925 void *arg)
1926{
1927 unsigned int i;
1928
1929 for (i = 0; i < dev->num_tx_queues; i++)
1930 f(dev, &dev->_tx[i], arg);
1931}
1932
d3fff6c4
ED
1933#define netdev_lockdep_set_classes(dev) \
1934{ \
1935 static struct lock_class_key qdisc_tx_busylock_key; \
1936 static struct lock_class_key qdisc_running_key; \
1937 static struct lock_class_key qdisc_xmit_lock_key; \
1938 static struct lock_class_key dev_addr_list_lock_key; \
1939 unsigned int i; \
1940 \
1941 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1942 (dev)->qdisc_running_key = &qdisc_running_key; \
1943 lockdep_set_class(&(dev)->addr_list_lock, \
1944 &dev_addr_list_lock_key); \
1945 for (i = 0; i < (dev)->num_tx_queues; i++) \
1946 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1947 &qdisc_xmit_lock_key); \
1948}
1949
f629d208 1950struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1951 struct sk_buff *skb,
1952 void *accel_priv);
8c4c49df 1953
871b642a
PA
1954/* returns the headroom that the master device needs to take in account
1955 * when forwarding to this dev
1956 */
1957static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1958{
1959 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1960}
1961
1962static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1963{
1964 if (dev->netdev_ops->ndo_set_rx_headroom)
1965 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1966}
1967
1968/* set the device rx headroom to the dev's default */
1969static inline void netdev_reset_rx_headroom(struct net_device *dev)
1970{
1971 netdev_set_rx_headroom(dev, -1);
1972}
1973
c346dca1
YH
1974/*
1975 * Net namespace inlines
1976 */
1977static inline
1978struct net *dev_net(const struct net_device *dev)
1979{
c2d9ba9b 1980 return read_pnet(&dev->nd_net);
c346dca1
YH
1981}
1982
1983static inline
f5aa23fd 1984void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 1985{
0c5c9fb5 1986 write_pnet(&dev->nd_net, net);
c346dca1
YH
1987}
1988
3e8a72d1 1989static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 1990{
3fc88677 1991#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
1992 if (dev->dsa_ptr != NULL)
1993 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 1994#endif
5aed85ce 1995 return false;
396138f0
LB
1996}
1997
bea3348e
SH
1998/**
1999 * netdev_priv - access network device private data
2000 * @dev: network device
2001 *
2002 * Get network device private data
2003 */
6472ce60 2004static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2005{
1ce8e7b5 2006 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2007}
2008
1da177e4
LT
2009/* Set the sysfs physical device reference for the network logical device
2010 * if set prior to registration will cause a symlink during initialization.
2011 */
43cb76d9 2012#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2013
384912ed 2014/* Set the sysfs device type for the network logical device to allow
3f79410c 2015 * fine-grained identification of different network device types. For
5e82b4b2 2016 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2017 */
2018#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2019
82dc3c63
ED
2020/* Default NAPI poll() weight
2021 * Device drivers are strongly advised to not use bigger value
2022 */
2023#define NAPI_POLL_WEIGHT 64
2024
3b582cc1 2025/**
5e82b4b2 2026 * netif_napi_add - initialize a NAPI context
3b582cc1 2027 * @dev: network device
5e82b4b2 2028 * @napi: NAPI context
3b582cc1
SH
2029 * @poll: polling function
2030 * @weight: default weight
2031 *
5e82b4b2
BH
2032 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2033 * *any* of the other NAPI-related functions.
3b582cc1 2034 */
d565b0a1
HX
2035void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2036 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2037
d64b5e85 2038/**
5e82b4b2 2039 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2040 * @dev: network device
5e82b4b2 2041 * @napi: NAPI context
d64b5e85
ED
2042 * @poll: polling function
2043 * @weight: default weight
2044 *
2045 * This variant of netif_napi_add() should be used from drivers using NAPI
2046 * to exclusively poll a TX queue.
2047 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2048 */
2049static inline void netif_tx_napi_add(struct net_device *dev,
2050 struct napi_struct *napi,
2051 int (*poll)(struct napi_struct *, int),
2052 int weight)
2053{
2054 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2055 netif_napi_add(dev, napi, poll, weight);
2056}
2057
d8156534 2058/**
5e82b4b2
BH
2059 * netif_napi_del - remove a NAPI context
2060 * @napi: NAPI context
d8156534 2061 *
5e82b4b2 2062 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2063 */
d565b0a1
HX
2064void netif_napi_del(struct napi_struct *napi);
2065
2066struct napi_gro_cb {
78a478d0 2067 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2068 void *frag0;
78a478d0 2069
7489594c
HX
2070 /* Length of frag0. */
2071 unsigned int frag0_len;
2072
86911732 2073 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2074 int data_offset;
86911732 2075
d565b0a1 2076 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2077 u16 flush;
2078
2079 /* Save the IP ID here and check when we get to the transport layer */
2080 u16 flush_id;
d565b0a1
HX
2081
2082 /* Number of segments aggregated. */
2e71a6f8
ED
2083 u16 count;
2084
15e2396d
TH
2085 /* Start offset for remote checksum offload */
2086 u16 gro_remcsum_start;
2087
2e71a6f8
ED
2088 /* jiffies when first packet was created/queued */
2089 unsigned long age;
86347245 2090
afe93325 2091 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2092 u16 proto;
2093
baa32ff4
TH
2094 /* This is non-zero if the packet may be of the same flow. */
2095 u8 same_flow:1;
2096
fac8e0f5
JG
2097 /* Used in tunnel GRO receive */
2098 u8 encap_mark:1;
573e8fca
TH
2099
2100 /* GRO checksum is valid */
2101 u8 csum_valid:1;
2102
662880f4
TH
2103 /* Number of checksums via CHECKSUM_UNNECESSARY */
2104 u8 csum_cnt:3;
c3c7c254 2105
baa32ff4
TH
2106 /* Free the skb? */
2107 u8 free:2;
2108#define NAPI_GRO_FREE 1
2109#define NAPI_GRO_FREE_STOLEN_HEAD 2
2110
efc98d08
TH
2111 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2112 u8 is_ipv6:1;
2113
a0ca153f
AD
2114 /* Used in GRE, set in fou/gue_gro_receive */
2115 u8 is_fou:1;
2116
1530545e
AD
2117 /* Used to determine if flush_id can be ignored */
2118 u8 is_atomic:1;
2119
2120 /* 5 bit hole */
baa32ff4 2121
bf5a755f
JC
2122 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2123 __wsum csum;
2124
c3c7c254
ED
2125 /* used in skb_gro_receive() slow path */
2126 struct sk_buff *last;
d565b0a1
HX
2127};
2128
2129#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2130
1da177e4 2131struct packet_type {
f2ccd8fa
DM
2132 __be16 type; /* This is really htons(ether_type). */
2133 struct net_device *dev; /* NULL is wildcarded here */
2134 int (*func) (struct sk_buff *,
2135 struct net_device *,
2136 struct packet_type *,
2137 struct net_device *);
c0de08d0
EL
2138 bool (*id_match)(struct packet_type *ptype,
2139 struct sock *sk);
1da177e4
LT
2140 void *af_packet_priv;
2141 struct list_head list;
2142};
2143
f191a1d1 2144struct offload_callbacks {
576a30eb 2145 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2146 netdev_features_t features);
d565b0a1 2147 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2148 struct sk_buff *skb);
299603e8 2149 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2150};
2151
2152struct packet_offload {
2153 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2154 u16 priority;
f191a1d1
VY
2155 struct offload_callbacks callbacks;
2156 struct list_head list;
1da177e4
LT
2157};
2158
5e82b4b2 2159/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2160struct pcpu_sw_netstats {
2161 u64 rx_packets;
2162 u64 rx_bytes;
2163 u64 tx_packets;
2164 u64 tx_bytes;
2165 struct u64_stats_sync syncp;
2166};
2167
aabc92bb
PNA
2168#define __netdev_alloc_pcpu_stats(type, gfp) \
2169({ \
2170 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2171 if (pcpu_stats) { \
2172 int __cpu; \
2173 for_each_possible_cpu(__cpu) { \
2174 typeof(type) *stat; \
2175 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2176 u64_stats_init(&stat->syncp); \
2177 } \
2178 } \
2179 pcpu_stats; \
1c213bd2
WC
2180})
2181
aabc92bb 2182#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2183 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2184
764f5e54
JP
2185enum netdev_lag_tx_type {
2186 NETDEV_LAG_TX_TYPE_UNKNOWN,
2187 NETDEV_LAG_TX_TYPE_RANDOM,
2188 NETDEV_LAG_TX_TYPE_BROADCAST,
2189 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2190 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2191 NETDEV_LAG_TX_TYPE_HASH,
2192};
2193
2194struct netdev_lag_upper_info {
2195 enum netdev_lag_tx_type tx_type;
2196};
2197
fb1b2e3c
JP
2198struct netdev_lag_lower_state_info {
2199 u8 link_up : 1,
2200 tx_enabled : 1;
2201};
2202
1da177e4
LT
2203#include <linux/notifier.h>
2204
dcfe1421
AW
2205/* netdevice notifier chain. Please remember to update the rtnetlink
2206 * notification exclusion list in rtnetlink_event() when adding new
2207 * types.
2208 */
2209#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2210#define NETDEV_DOWN 0x0002
2211#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2212 detected a hardware crash and restarted
2213 - we can use this eg to kick tcp sessions
2214 once done */
2215#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2216#define NETDEV_REGISTER 0x0005
2217#define NETDEV_UNREGISTER 0x0006
1d486bfb 2218#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2219#define NETDEV_CHANGEADDR 0x0008
2220#define NETDEV_GOING_DOWN 0x0009
2221#define NETDEV_CHANGENAME 0x000A
2222#define NETDEV_FEAT_CHANGE 0x000B
2223#define NETDEV_BONDING_FAILOVER 0x000C
2224#define NETDEV_PRE_UP 0x000D
2225#define NETDEV_PRE_TYPE_CHANGE 0x000E
2226#define NETDEV_POST_TYPE_CHANGE 0x000F
2227#define NETDEV_POST_INIT 0x0010
0115e8e3 2228#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2229#define NETDEV_RELEASE 0x0012
2230#define NETDEV_NOTIFY_PEERS 0x0013
2231#define NETDEV_JOIN 0x0014
42e52bf9 2232#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2233#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2234#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2235#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2236#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2237#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2238#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2239#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
dcfe1421 2240
f629d208
JP
2241int register_netdevice_notifier(struct notifier_block *nb);
2242int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2243
2244struct netdev_notifier_info {
2245 struct net_device *dev;
2246};
2247
be9efd36
JP
2248struct netdev_notifier_change_info {
2249 struct netdev_notifier_info info; /* must be first */
2250 unsigned int flags_changed;
2251};
2252
0e4ead9d
JP
2253struct netdev_notifier_changeupper_info {
2254 struct netdev_notifier_info info; /* must be first */
2255 struct net_device *upper_dev; /* new upper dev */
2256 bool master; /* is upper dev master */
5e82b4b2 2257 bool linking; /* is the notification for link or unlink */
29bf24af 2258 void *upper_info; /* upper dev info */
0e4ead9d
JP
2259};
2260
04d48266
JP
2261struct netdev_notifier_changelowerstate_info {
2262 struct netdev_notifier_info info; /* must be first */
2263 void *lower_state_info; /* is lower dev state */
2264};
2265
75538c2b
CW
2266static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2267 struct net_device *dev)
2268{
2269 info->dev = dev;
2270}
2271
351638e7
JP
2272static inline struct net_device *
2273netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2274{
2275 return info->dev;
2276}
2277
f629d208 2278int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2279
2280
1da177e4
LT
2281extern rwlock_t dev_base_lock; /* Device list lock */
2282
881d966b
EB
2283#define for_each_netdev(net, d) \
2284 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2285#define for_each_netdev_reverse(net, d) \
2286 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2287#define for_each_netdev_rcu(net, d) \
2288 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2289#define for_each_netdev_safe(net, d, n) \
2290 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2291#define for_each_netdev_continue(net, d) \
2292 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2293#define for_each_netdev_continue_rcu(net, d) \
2294 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2295#define for_each_netdev_in_bond_rcu(bond, slave) \
2296 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2297 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2298#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2299
a050c33f
DL
2300static inline struct net_device *next_net_device(struct net_device *dev)
2301{
2302 struct list_head *lh;
2303 struct net *net;
2304
c346dca1 2305 net = dev_net(dev);
a050c33f
DL
2306 lh = dev->dev_list.next;
2307 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2308}
2309
ce81b76a
ED
2310static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2311{
2312 struct list_head *lh;
2313 struct net *net;
2314
2315 net = dev_net(dev);
ccf43438 2316 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2317 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2318}
2319
a050c33f
DL
2320static inline struct net_device *first_net_device(struct net *net)
2321{
2322 return list_empty(&net->dev_base_head) ? NULL :
2323 net_device_entry(net->dev_base_head.next);
2324}
7562f876 2325
ccf43438
ED
2326static inline struct net_device *first_net_device_rcu(struct net *net)
2327{
2328 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2329
2330 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2331}
2332
f629d208
JP
2333int netdev_boot_setup_check(struct net_device *dev);
2334unsigned long netdev_boot_base(const char *prefix, int unit);
2335struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2336 const char *hwaddr);
2337struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2338struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2339void dev_add_pack(struct packet_type *pt);
2340void dev_remove_pack(struct packet_type *pt);
2341void __dev_remove_pack(struct packet_type *pt);
2342void dev_add_offload(struct packet_offload *po);
2343void dev_remove_offload(struct packet_offload *po);
f629d208 2344
a54acb3a 2345int dev_get_iflink(const struct net_device *dev);
fc4099f1 2346int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2347struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2348 unsigned short mask);
f629d208
JP
2349struct net_device *dev_get_by_name(struct net *net, const char *name);
2350struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2351struct net_device *__dev_get_by_name(struct net *net, const char *name);
2352int dev_alloc_name(struct net_device *dev, const char *name);
2353int dev_open(struct net_device *dev);
2354int dev_close(struct net_device *dev);
99c4a26a 2355int dev_close_many(struct list_head *head, bool unlink);
f629d208 2356void dev_disable_lro(struct net_device *dev);
0c4b51f0 2357int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2358int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2359int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2360int register_netdevice(struct net_device *dev);
2361void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2362void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2363static inline void unregister_netdevice(struct net_device *dev)
2364{
2365 unregister_netdevice_queue(dev, NULL);
2366}
2367
f629d208
JP
2368int netdev_refcnt_read(const struct net_device *dev);
2369void free_netdev(struct net_device *dev);
74d332c1 2370void netdev_freemem(struct net_device *dev);
f629d208
JP
2371void synchronize_net(void);
2372int init_dummy_netdev(struct net_device *dev);
937f1ba5 2373
f60e5990 2374DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2375#define XMIT_RECURSION_LIMIT 10
2376
f60e5990 2377static inline int dev_recursion_level(void)
2378{
2379 return this_cpu_read(xmit_recursion);
2380}
2381
f629d208
JP
2382struct net_device *dev_get_by_index(struct net *net, int ifindex);
2383struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2384struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2385int netdev_get_name(struct net *net, char *name, int ifindex);
2386int dev_restart(struct net_device *dev);
f629d208 2387int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2388
2389static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2390{
2391 return NAPI_GRO_CB(skb)->data_offset;
2392}
2393
2394static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2395{
2396 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2397}
2398
2399static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2400{
2401 NAPI_GRO_CB(skb)->data_offset += len;
2402}
2403
a5b1cf28
HX
2404static inline void *skb_gro_header_fast(struct sk_buff *skb,
2405 unsigned int offset)
86911732 2406{
a5b1cf28
HX
2407 return NAPI_GRO_CB(skb)->frag0 + offset;
2408}
78a478d0 2409
a5b1cf28
HX
2410static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2411{
2412 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2413}
78a478d0 2414
a5b1cf28
HX
2415static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2416 unsigned int offset)
2417{
17dd759c
HX
2418 if (!pskb_may_pull(skb, hlen))
2419 return NULL;
2420
a5b1cf28
HX
2421 NAPI_GRO_CB(skb)->frag0 = NULL;
2422 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2423 return skb->data + offset;
86911732 2424}
1da177e4 2425
36e7b1b8
HX
2426static inline void *skb_gro_network_header(struct sk_buff *skb)
2427{
78d3fd0b
HX
2428 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2429 skb_network_offset(skb);
36e7b1b8
HX
2430}
2431
bf5a755f
JC
2432static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2433 const void *start, unsigned int len)
2434{
573e8fca 2435 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2436 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2437 csum_partial(start, len, 0));
2438}
2439
573e8fca
TH
2440/* GRO checksum functions. These are logical equivalents of the normal
2441 * checksum functions (in skbuff.h) except that they operate on the GRO
2442 * offsets and fields in sk_buff.
2443 */
2444
2445__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2446
15e2396d
TH
2447static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2448{
b7fe10e5 2449 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2450}
2451
573e8fca
TH
2452static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2453 bool zero_okay,
2454 __sum16 check)
2455{
6edec0e6
TH
2456 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2457 skb_checksum_start_offset(skb) <
2458 skb_gro_offset(skb)) &&
15e2396d 2459 !skb_at_gro_remcsum_start(skb) &&
662880f4 2460 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2461 (!zero_okay || check));
2462}
2463
2464static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2465 __wsum psum)
2466{
2467 if (NAPI_GRO_CB(skb)->csum_valid &&
2468 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2469 return 0;
2470
2471 NAPI_GRO_CB(skb)->csum = psum;
2472
2473 return __skb_gro_checksum_complete(skb);
2474}
2475
573e8fca
TH
2476static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2477{
662880f4
TH
2478 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2479 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2480 NAPI_GRO_CB(skb)->csum_cnt--;
2481 } else {
2482 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2483 * verified a new top level checksum or an encapsulated one
2484 * during GRO. This saves work if we fallback to normal path.
2485 */
2486 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2487 }
2488}
2489
2490#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2491 compute_pseudo) \
2492({ \
2493 __sum16 __ret = 0; \
2494 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2495 __ret = __skb_gro_checksum_validate_complete(skb, \
2496 compute_pseudo(skb, proto)); \
5a212329
TH
2497 if (__ret) \
2498 __skb_mark_checksum_bad(skb); \
2499 else \
573e8fca
TH
2500 skb_gro_incr_csum_unnecessary(skb); \
2501 __ret; \
2502})
2503
2504#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2505 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2506
2507#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2508 compute_pseudo) \
2509 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2510
2511#define skb_gro_checksum_simple_validate(skb) \
2512 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2513
d96535a1
TH
2514static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2515{
2516 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2517 !NAPI_GRO_CB(skb)->csum_valid);
2518}
2519
2520static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2521 __sum16 check, __wsum pseudo)
2522{
2523 NAPI_GRO_CB(skb)->csum = ~pseudo;
2524 NAPI_GRO_CB(skb)->csum_valid = 1;
2525}
2526
2527#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2528do { \
2529 if (__skb_gro_checksum_convert_check(skb)) \
2530 __skb_gro_checksum_convert(skb, check, \
2531 compute_pseudo(skb, proto)); \
2532} while (0)
2533
26c4f7da
TH
2534struct gro_remcsum {
2535 int offset;
2536 __wsum delta;
2537};
2538
2539static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2540{
846cd667 2541 grc->offset = 0;
26c4f7da
TH
2542 grc->delta = 0;
2543}
2544
b7fe10e5
TH
2545static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2546 unsigned int off, size_t hdrlen,
2547 int start, int offset,
2548 struct gro_remcsum *grc,
2549 bool nopartial)
dcdc8994
TH
2550{
2551 __wsum delta;
b7fe10e5 2552 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2553
2554 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2555
15e2396d 2556 if (!nopartial) {
b7fe10e5
TH
2557 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2558 return ptr;
2559 }
2560
2561 ptr = skb_gro_header_fast(skb, off);
2562 if (skb_gro_header_hard(skb, off + plen)) {
2563 ptr = skb_gro_header_slow(skb, off + plen, off);
2564 if (!ptr)
2565 return NULL;
15e2396d
TH
2566 }
2567
b7fe10e5
TH
2568 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2569 start, offset);
dcdc8994
TH
2570
2571 /* Adjust skb->csum since we changed the packet */
dcdc8994 2572 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2573
b7fe10e5 2574 grc->offset = off + hdrlen + offset;
26c4f7da 2575 grc->delta = delta;
b7fe10e5
TH
2576
2577 return ptr;
dcdc8994
TH
2578}
2579
26c4f7da
TH
2580static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2581 struct gro_remcsum *grc)
2582{
b7fe10e5
TH
2583 void *ptr;
2584 size_t plen = grc->offset + sizeof(u16);
2585
26c4f7da
TH
2586 if (!grc->delta)
2587 return;
2588
b7fe10e5
TH
2589 ptr = skb_gro_header_fast(skb, grc->offset);
2590 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2591 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2592 if (!ptr)
2593 return;
2594 }
2595
2596 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2597}
dcdc8994 2598
6ae23ad3
TH
2599struct skb_csum_offl_spec {
2600 __u16 ipv4_okay:1,
2601 ipv6_okay:1,
2602 encap_okay:1,
2603 ip_options_okay:1,
2604 ext_hdrs_okay:1,
2605 tcp_okay:1,
2606 udp_okay:1,
2607 sctp_okay:1,
2608 vlan_okay:1,
2609 no_encapped_ipv6:1,
2610 no_not_encapped:1;
2611};
2612
2613bool __skb_csum_offload_chk(struct sk_buff *skb,
2614 const struct skb_csum_offl_spec *spec,
2615 bool *csum_encapped,
2616 bool csum_help);
2617
2618static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2619 const struct skb_csum_offl_spec *spec,
2620 bool *csum_encapped,
2621 bool csum_help)
2622{
2623 if (skb->ip_summed != CHECKSUM_PARTIAL)
2624 return false;
2625
2626 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2627}
2628
2629static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2630 const struct skb_csum_offl_spec *spec)
2631{
2632 bool csum_encapped;
2633
2634 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2635}
2636
2637static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2638{
2639 static const struct skb_csum_offl_spec csum_offl_spec = {
2640 .ipv4_okay = 1,
2641 .ip_options_okay = 1,
2642 .ipv6_okay = 1,
2643 .vlan_okay = 1,
2644 .tcp_okay = 1,
2645 .udp_okay = 1,
2646 };
2647
2648 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2649}
2650
2651static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2652{
2653 static const struct skb_csum_offl_spec csum_offl_spec = {
2654 .ipv4_okay = 1,
2655 .ip_options_okay = 1,
2656 .tcp_okay = 1,
2657 .udp_okay = 1,
2658 .vlan_okay = 1,
2659 };
2660
2661 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2662}
2663
0c4e8581
SH
2664static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2665 unsigned short type,
3b04ddde 2666 const void *daddr, const void *saddr,
95c96174 2667 unsigned int len)
0c4e8581 2668{
f1ecfd5d 2669 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2670 return 0;
3b04ddde
SH
2671
2672 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2673}
2674
b95cce35
SH
2675static inline int dev_parse_header(const struct sk_buff *skb,
2676 unsigned char *haddr)
2677{
2678 const struct net_device *dev = skb->dev;
2679
1b83336b 2680 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2681 return 0;
3b04ddde 2682 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2683}
2684
2793a23a
WB
2685/* ll_header must have at least hard_header_len allocated */
2686static inline bool dev_validate_header(const struct net_device *dev,
2687 char *ll_header, int len)
2688{
2689 if (likely(len >= dev->hard_header_len))
2690 return true;
2691
2692 if (capable(CAP_SYS_RAWIO)) {
2693 memset(ll_header + len, 0, dev->hard_header_len - len);
2694 return true;
2695 }
2696
2697 if (dev->header_ops && dev->header_ops->validate)
2698 return dev->header_ops->validate(ll_header, len);
2699
2700 return false;
2701}
2702
1da177e4 2703typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2704int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2705static inline int unregister_gifconf(unsigned int family)
2706{
2707 return register_gifconf(family, NULL);
2708}
2709
99bbc707 2710#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2711#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2712struct sd_flow_limit {
2713 u64 count;
2714 unsigned int num_buckets;
2715 unsigned int history_head;
2716 u16 history[FLOW_LIMIT_HISTORY];
2717 u8 buckets[];
2718};
2719
2720extern int netdev_flow_limit_table_len;
2721#endif /* CONFIG_NET_FLOW_LIMIT */
2722
1da177e4 2723/*
5e82b4b2 2724 * Incoming packets are placed on per-CPU queues
1da177e4 2725 */
d94d9fee 2726struct softnet_data {
1da177e4 2727 struct list_head poll_list;
6e7676c1 2728 struct sk_buff_head process_queue;
1da177e4 2729
dee42870 2730 /* stats */
cd7b5396
DM
2731 unsigned int processed;
2732 unsigned int time_squeeze;
cd7b5396 2733 unsigned int received_rps;
fd793d89 2734#ifdef CONFIG_RPS
88751275 2735 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2736#endif
2737#ifdef CONFIG_NET_FLOW_LIMIT
2738 struct sd_flow_limit __rcu *flow_limit;
2739#endif
2740 struct Qdisc *output_queue;
2741 struct Qdisc **output_queue_tailp;
2742 struct sk_buff *completion_queue;
88751275 2743
4cdb1e2e 2744#ifdef CONFIG_RPS
501e7ef5
ED
2745 /* input_queue_head should be written by cpu owning this struct,
2746 * and only read by other cpus. Worth using a cache line.
2747 */
2748 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2749
2750 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2751 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2752 struct softnet_data *rps_ipi_next;
2753 unsigned int cpu;
76cc8b13 2754 unsigned int input_queue_tail;
1e94d72f 2755#endif
95c96174 2756 unsigned int dropped;
0a9627f2 2757 struct sk_buff_head input_pkt_queue;
bea3348e 2758 struct napi_struct backlog;
99bbc707 2759
1da177e4
LT
2760};
2761
76cc8b13 2762static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2763{
2764#ifdef CONFIG_RPS
76cc8b13
TH
2765 sd->input_queue_head++;
2766#endif
2767}
2768
2769static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2770 unsigned int *qtail)
2771{
2772#ifdef CONFIG_RPS
2773 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2774#endif
2775}
2776
0a9627f2 2777DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2778
f629d208 2779void __netif_schedule(struct Qdisc *q);
46e5da40 2780void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2781
fd2ea0a7
DM
2782static inline void netif_tx_schedule_all(struct net_device *dev)
2783{
2784 unsigned int i;
2785
2786 for (i = 0; i < dev->num_tx_queues; i++)
2787 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2788}
2789
f9a7cbbf 2790static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2791{
73466498 2792 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2793}
2794
bea3348e
SH
2795/**
2796 * netif_start_queue - allow transmit
2797 * @dev: network device
2798 *
2799 * Allow upper layers to call the device hard_start_xmit routine.
2800 */
1da177e4
LT
2801static inline void netif_start_queue(struct net_device *dev)
2802{
e8a0464c 2803 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2804}
2805
fd2ea0a7
DM
2806static inline void netif_tx_start_all_queues(struct net_device *dev)
2807{
2808 unsigned int i;
2809
2810 for (i = 0; i < dev->num_tx_queues; i++) {
2811 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2812 netif_tx_start_queue(txq);
2813 }
2814}
2815
46e5da40 2816void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2817
d29f749e
DJ
2818/**
2819 * netif_wake_queue - restart transmit
2820 * @dev: network device
2821 *
2822 * Allow upper layers to call the device hard_start_xmit routine.
2823 * Used for flow control when transmit resources are available.
2824 */
79d16385
DM
2825static inline void netif_wake_queue(struct net_device *dev)
2826{
e8a0464c 2827 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2828}
2829
fd2ea0a7
DM
2830static inline void netif_tx_wake_all_queues(struct net_device *dev)
2831{
2832 unsigned int i;
2833
2834 for (i = 0; i < dev->num_tx_queues; i++) {
2835 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2836 netif_tx_wake_queue(txq);
2837 }
2838}
2839
f9a7cbbf 2840static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2841{
73466498 2842 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2843}
2844
bea3348e
SH
2845/**
2846 * netif_stop_queue - stop transmitted packets
2847 * @dev: network device
2848 *
2849 * Stop upper layers calling the device hard_start_xmit routine.
2850 * Used for flow control when transmit resources are unavailable.
2851 */
1da177e4
LT
2852static inline void netif_stop_queue(struct net_device *dev)
2853{
e8a0464c 2854 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2855}
2856
a2029240 2857void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2858
4d29515f 2859static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2860{
73466498 2861 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2862}
2863
bea3348e
SH
2864/**
2865 * netif_queue_stopped - test if transmit queue is flowblocked
2866 * @dev: network device
2867 *
2868 * Test if transmit queue on device is currently unable to send.
2869 */
4d29515f 2870static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2871{
e8a0464c 2872 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2873}
2874
4d29515f 2875static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2876{
73466498
TH
2877 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2878}
2879
8e2f1a63
DB
2880static inline bool
2881netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2882{
2883 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2884}
2885
8e2f1a63
DB
2886static inline bool
2887netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2888{
2889 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2890}
2891
53511453
ED
2892/**
2893 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2894 * @dev_queue: pointer to transmit queue
2895 *
2896 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2897 * to give appropriate hint to the CPU.
53511453
ED
2898 */
2899static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2900{
2901#ifdef CONFIG_BQL
2902 prefetchw(&dev_queue->dql.num_queued);
2903#endif
2904}
2905
2906/**
2907 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2908 * @dev_queue: pointer to transmit queue
2909 *
2910 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2911 * to give appropriate hint to the CPU.
53511453
ED
2912 */
2913static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2914{
2915#ifdef CONFIG_BQL
2916 prefetchw(&dev_queue->dql.limit);
2917#endif
2918}
2919
c5d67bd7
TH
2920static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2921 unsigned int bytes)
2922{
114cf580
TH
2923#ifdef CONFIG_BQL
2924 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2925
2926 if (likely(dql_avail(&dev_queue->dql) >= 0))
2927 return;
2928
2929 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2930
2931 /*
2932 * The XOFF flag must be set before checking the dql_avail below,
2933 * because in netdev_tx_completed_queue we update the dql_completed
2934 * before checking the XOFF flag.
2935 */
2936 smp_mb();
2937
2938 /* check again in case another CPU has just made room avail */
2939 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2940 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2941#endif
c5d67bd7
TH
2942}
2943
0042d0c8
FF
2944/**
2945 * netdev_sent_queue - report the number of bytes queued to hardware
2946 * @dev: network device
2947 * @bytes: number of bytes queued to the hardware device queue
2948 *
2949 * Report the number of bytes queued for sending/completion to the network
2950 * device hardware queue. @bytes should be a good approximation and should
2951 * exactly match netdev_completed_queue() @bytes
2952 */
c5d67bd7
TH
2953static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2954{
2955 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2956}
2957
2958static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2959 unsigned int pkts, unsigned int bytes)
c5d67bd7 2960{
114cf580 2961#ifdef CONFIG_BQL
b37c0fbe
AD
2962 if (unlikely(!bytes))
2963 return;
2964
2965 dql_completed(&dev_queue->dql, bytes);
2966
2967 /*
2968 * Without the memory barrier there is a small possiblity that
2969 * netdev_tx_sent_queue will miss the update and cause the queue to
2970 * be stopped forever
2971 */
2972 smp_mb();
2973
2974 if (dql_avail(&dev_queue->dql) < 0)
2975 return;
2976
2977 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2978 netif_schedule_queue(dev_queue);
114cf580 2979#endif
c5d67bd7
TH
2980}
2981
0042d0c8
FF
2982/**
2983 * netdev_completed_queue - report bytes and packets completed by device
2984 * @dev: network device
2985 * @pkts: actual number of packets sent over the medium
2986 * @bytes: actual number of bytes sent over the medium
2987 *
2988 * Report the number of bytes and packets transmitted by the network device
2989 * hardware queue over the physical medium, @bytes must exactly match the
2990 * @bytes amount passed to netdev_sent_queue()
2991 */
c5d67bd7 2992static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2993 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2994{
2995 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2996}
2997
2998static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2999{
114cf580 3000#ifdef CONFIG_BQL
5c490354 3001 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3002 dql_reset(&q->dql);
3003#endif
c5d67bd7
TH
3004}
3005
0042d0c8
FF
3006/**
3007 * netdev_reset_queue - reset the packets and bytes count of a network device
3008 * @dev_queue: network device
3009 *
3010 * Reset the bytes and packet count of a network device and clear the
3011 * software flow control OFF bit for this network device
3012 */
c5d67bd7
TH
3013static inline void netdev_reset_queue(struct net_device *dev_queue)
3014{
3015 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3016}
3017
b9507bda
DB
3018/**
3019 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3020 * @dev: network device
3021 * @queue_index: given tx queue index
3022 *
3023 * Returns 0 if given tx queue index >= number of device tx queues,
3024 * otherwise returns the originally passed tx queue index.
3025 */
3026static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3027{
3028 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3029 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3030 dev->name, queue_index,
3031 dev->real_num_tx_queues);
3032 return 0;
3033 }
3034
3035 return queue_index;
3036}
3037
bea3348e
SH
3038/**
3039 * netif_running - test if up
3040 * @dev: network device
3041 *
3042 * Test if the device has been brought up.
3043 */
4d29515f 3044static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3045{
3046 return test_bit(__LINK_STATE_START, &dev->state);
3047}
3048
f25f4e44 3049/*
5e82b4b2 3050 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3051 * stop, and a check if it's stopped. All other device management is
3052 * done at the overall netdevice level.
3053 * Also test the device if we're multiqueue.
3054 */
bea3348e
SH
3055
3056/**
3057 * netif_start_subqueue - allow sending packets on subqueue
3058 * @dev: network device
3059 * @queue_index: sub queue index
3060 *
3061 * Start individual transmit queue of a device with multiple transmit queues.
3062 */
f25f4e44
PWJ
3063static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3064{
fd2ea0a7 3065 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3066
3067 netif_tx_start_queue(txq);
f25f4e44
PWJ
3068}
3069
bea3348e
SH
3070/**
3071 * netif_stop_subqueue - stop sending packets on subqueue
3072 * @dev: network device
3073 * @queue_index: sub queue index
3074 *
3075 * Stop individual transmit queue of a device with multiple transmit queues.
3076 */
f25f4e44
PWJ
3077static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3078{
fd2ea0a7 3079 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3080 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3081}
3082
bea3348e
SH
3083/**
3084 * netif_subqueue_stopped - test status of subqueue
3085 * @dev: network device
3086 * @queue_index: sub queue index
3087 *
3088 * Check individual transmit queue of a device with multiple transmit queues.
3089 */
4d29515f
DM
3090static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3091 u16 queue_index)
f25f4e44 3092{
fd2ea0a7 3093 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3094
3095 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3096}
3097
4d29515f
DM
3098static inline bool netif_subqueue_stopped(const struct net_device *dev,
3099 struct sk_buff *skb)
668f895a
PE
3100{
3101 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3102}
bea3348e 3103
46e5da40 3104void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 3105
537c00de 3106#ifdef CONFIG_XPS
53af53ae 3107int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3108 u16 index);
537c00de
AD
3109#else
3110static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3111 const struct cpumask *mask,
537c00de
AD
3112 u16 index)
3113{
3114 return 0;
3115}
3116#endif
3117
5605c762
JP
3118u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3119 unsigned int num_tx_queues);
3120
a3d22a68
VZ
3121/*
3122 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3123 * as a distribution range limit for the returned value.
3124 */
3125static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3126 struct sk_buff *skb)
a3d22a68
VZ
3127{
3128 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3129}
3130
bea3348e
SH
3131/**
3132 * netif_is_multiqueue - test if device has multiple transmit queues
3133 * @dev: network device
3134 *
3135 * Check if device has multiple transmit queues
bea3348e 3136 */
4d29515f 3137static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3138{
a02cec21 3139 return dev->num_tx_queues > 1;
f25f4e44 3140}
1da177e4 3141
f629d208 3142int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3143
a953be53 3144#ifdef CONFIG_SYSFS
f629d208 3145int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3146#else
3147static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3148 unsigned int rxq)
3149{
3150 return 0;
3151}
3152#endif
3153
a953be53
MD
3154#ifdef CONFIG_SYSFS
3155static inline unsigned int get_netdev_rx_queue_index(
3156 struct netdev_rx_queue *queue)
3157{
3158 struct net_device *dev = queue->dev;
3159 int index = queue - dev->_rx;
3160
3161 BUG_ON(index >= dev->num_rx_queues);
3162 return index;
3163}
3164#endif
3165
16917b87 3166#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3167int netif_get_num_default_rss_queues(void);
16917b87 3168
e6247027
ED
3169enum skb_free_reason {
3170 SKB_REASON_CONSUMED,
3171 SKB_REASON_DROPPED,
3172};
3173
3174void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3175void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3176
e6247027
ED
3177/*
3178 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3179 * interrupt context or with hardware interrupts being disabled.
3180 * (in_irq() || irqs_disabled())
3181 *
3182 * We provide four helpers that can be used in following contexts :
3183 *
3184 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3185 * replacing kfree_skb(skb)
3186 *
3187 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3188 * Typically used in place of consume_skb(skb) in TX completion path
3189 *
3190 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3191 * replacing kfree_skb(skb)
3192 *
3193 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3194 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3195 */
e6247027
ED
3196static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3197{
3198 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3199}
3200
3201static inline void dev_consume_skb_irq(struct sk_buff *skb)
3202{
3203 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3204}
3205
3206static inline void dev_kfree_skb_any(struct sk_buff *skb)
3207{
3208 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3209}
3210
3211static inline void dev_consume_skb_any(struct sk_buff *skb)
3212{
3213 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3214}
1da177e4 3215
f629d208
JP
3216int netif_rx(struct sk_buff *skb);
3217int netif_rx_ni(struct sk_buff *skb);
04eb4489 3218int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3219gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3220void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3221struct sk_buff *napi_get_frags(struct napi_struct *napi);
3222gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3223struct packet_offload *gro_find_receive_by_type(__be16 type);
3224struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3225
3226static inline void napi_free_frags(struct napi_struct *napi)
3227{
3228 kfree_skb(napi->skb);
3229 napi->skb = NULL;
3230}
3231
f629d208
JP
3232int netdev_rx_handler_register(struct net_device *dev,
3233 rx_handler_func_t *rx_handler,
3234 void *rx_handler_data);
3235void netdev_rx_handler_unregister(struct net_device *dev);
3236
3237bool dev_valid_name(const char *name);
3238int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3239int dev_ethtool(struct net *net, struct ifreq *);
3240unsigned int dev_get_flags(const struct net_device *);
3241int __dev_change_flags(struct net_device *, unsigned int flags);
3242int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3243void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3244 unsigned int gchanges);
f629d208
JP
3245int dev_change_name(struct net_device *, const char *);
3246int dev_set_alias(struct net_device *, const char *, size_t);
3247int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3248int dev_set_mtu(struct net_device *, int);
3249void dev_set_group(struct net_device *, int);
3250int dev_set_mac_address(struct net_device *, struct sockaddr *);
3251int dev_change_carrier(struct net_device *, bool new_carrier);
3252int dev_get_phys_port_id(struct net_device *dev,
02637fce 3253 struct netdev_phys_item_id *ppid);
db24a904
DA
3254int dev_get_phys_port_name(struct net_device *dev,
3255 char *name, size_t len);
d746d707 3256int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3257struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3258struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3259 struct netdev_queue *txq, int *ret);
a0265d28 3260int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3261int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3262bool is_skb_forwardable(const struct net_device *dev,
3263 const struct sk_buff *skb);
1da177e4 3264
74b20582
DA
3265void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3266
20380731 3267extern int netdev_budget;
1da177e4
LT
3268
3269/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3270void netdev_run_todo(void);
1da177e4 3271
bea3348e
SH
3272/**
3273 * dev_put - release reference to device
3274 * @dev: network device
3275 *
9ef4429b 3276 * Release reference to device to allow it to be freed.
bea3348e 3277 */
1da177e4
LT
3278static inline void dev_put(struct net_device *dev)
3279{
933393f5 3280 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3281}
3282
bea3348e
SH
3283/**
3284 * dev_hold - get reference to device
3285 * @dev: network device
3286 *
9ef4429b 3287 * Hold reference to device to keep it from being freed.
bea3348e 3288 */
15333061
SH
3289static inline void dev_hold(struct net_device *dev)
3290{
933393f5 3291 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3292}
1da177e4
LT
3293
3294/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3295 * and _off may be called from IRQ context, but it is caller
3296 * who is responsible for serialization of these calls.
b00055aa
SR
3297 *
3298 * The name carrier is inappropriate, these functions should really be
3299 * called netif_lowerlayer_*() because they represent the state of any
3300 * kind of lower layer not just hardware media.
1da177e4
LT
3301 */
3302
f629d208
JP
3303void linkwatch_init_dev(struct net_device *dev);
3304void linkwatch_fire_event(struct net_device *dev);
3305void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3306
bea3348e
SH
3307/**
3308 * netif_carrier_ok - test if carrier present
3309 * @dev: network device
3310 *
3311 * Check if carrier is present on device
3312 */
4d29515f 3313static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3314{
3315 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3316}
3317
f629d208 3318unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3319
f629d208 3320void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3321
f629d208 3322void netif_carrier_on(struct net_device *dev);
1da177e4 3323
f629d208 3324void netif_carrier_off(struct net_device *dev);
1da177e4 3325
bea3348e
SH
3326/**
3327 * netif_dormant_on - mark device as dormant.
3328 * @dev: network device
3329 *
3330 * Mark device as dormant (as per RFC2863).
3331 *
3332 * The dormant state indicates that the relevant interface is not
3333 * actually in a condition to pass packets (i.e., it is not 'up') but is
3334 * in a "pending" state, waiting for some external event. For "on-
3335 * demand" interfaces, this new state identifies the situation where the
3336 * interface is waiting for events to place it in the up state.
bea3348e 3337 */
b00055aa
SR
3338static inline void netif_dormant_on(struct net_device *dev)
3339{
3340 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3341 linkwatch_fire_event(dev);
3342}
3343
bea3348e
SH
3344/**
3345 * netif_dormant_off - set device as not dormant.
3346 * @dev: network device
3347 *
3348 * Device is not in dormant state.
3349 */
b00055aa
SR
3350static inline void netif_dormant_off(struct net_device *dev)
3351{
3352 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3353 linkwatch_fire_event(dev);
3354}
3355
bea3348e
SH
3356/**
3357 * netif_dormant - test if carrier present
3358 * @dev: network device
3359 *
3360 * Check if carrier is present on device
3361 */
4d29515f 3362static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3363{
3364 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3365}
3366
3367
bea3348e
SH
3368/**
3369 * netif_oper_up - test if device is operational
3370 * @dev: network device
3371 *
3372 * Check if carrier is operational
3373 */
4d29515f 3374static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3375{
b00055aa
SR
3376 return (dev->operstate == IF_OPER_UP ||
3377 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3378}
3379
bea3348e
SH
3380/**
3381 * netif_device_present - is device available or removed
3382 * @dev: network device
3383 *
3384 * Check if device has not been removed from system.
3385 */
4d29515f 3386static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3387{
3388 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3389}
3390
f629d208 3391void netif_device_detach(struct net_device *dev);
1da177e4 3392
f629d208 3393void netif_device_attach(struct net_device *dev);
1da177e4
LT
3394
3395/*
3396 * Network interface message level settings
3397 */
1da177e4
LT
3398
3399enum {
3400 NETIF_MSG_DRV = 0x0001,
3401 NETIF_MSG_PROBE = 0x0002,
3402 NETIF_MSG_LINK = 0x0004,
3403 NETIF_MSG_TIMER = 0x0008,
3404 NETIF_MSG_IFDOWN = 0x0010,
3405 NETIF_MSG_IFUP = 0x0020,
3406 NETIF_MSG_RX_ERR = 0x0040,
3407 NETIF_MSG_TX_ERR = 0x0080,
3408 NETIF_MSG_TX_QUEUED = 0x0100,
3409 NETIF_MSG_INTR = 0x0200,
3410 NETIF_MSG_TX_DONE = 0x0400,
3411 NETIF_MSG_RX_STATUS = 0x0800,
3412 NETIF_MSG_PKTDATA = 0x1000,
3413 NETIF_MSG_HW = 0x2000,
3414 NETIF_MSG_WOL = 0x4000,
3415};
3416
3417#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3418#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3419#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3420#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3421#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3422#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3423#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3424#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3425#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3426#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3427#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3428#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3429#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3430#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3431#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3432
3433static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3434{
3435 /* use default */
3436 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3437 return default_msg_enable_bits;
3438 if (debug_value == 0) /* no output */
3439 return 0;
3440 /* set low N bits */
3441 return (1 << debug_value) - 1;
3442}
3443
c773e847 3444static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3445{
c773e847
DM
3446 spin_lock(&txq->_xmit_lock);
3447 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3448}
3449
fd2ea0a7
DM
3450static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3451{
3452 spin_lock_bh(&txq->_xmit_lock);
3453 txq->xmit_lock_owner = smp_processor_id();
3454}
3455
4d29515f 3456static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3457{
4d29515f 3458 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3459 if (likely(ok))
3460 txq->xmit_lock_owner = smp_processor_id();
3461 return ok;
3462}
3463
3464static inline void __netif_tx_unlock(struct netdev_queue *txq)
3465{
3466 txq->xmit_lock_owner = -1;
3467 spin_unlock(&txq->_xmit_lock);
3468}
3469
3470static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3471{
3472 txq->xmit_lock_owner = -1;
3473 spin_unlock_bh(&txq->_xmit_lock);
3474}
3475
08baf561
ED
3476static inline void txq_trans_update(struct netdev_queue *txq)
3477{
3478 if (txq->xmit_lock_owner != -1)
3479 txq->trans_start = jiffies;
3480}
3481
ba162f8e
FW
3482/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3483static inline void netif_trans_update(struct net_device *dev)
3484{
9b36627a
FW
3485 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3486
3487 if (txq->trans_start != jiffies)
3488 txq->trans_start = jiffies;
ba162f8e
FW
3489}
3490
d29f749e
DJ
3491/**
3492 * netif_tx_lock - grab network device transmit lock
3493 * @dev: network device
d29f749e
DJ
3494 *
3495 * Get network device transmit lock
3496 */
22dd7495
JHS
3497static inline void netif_tx_lock(struct net_device *dev)
3498{
e8a0464c 3499 unsigned int i;
c3f26a26 3500 int cpu;
c773e847 3501
c3f26a26
DM
3502 spin_lock(&dev->tx_global_lock);
3503 cpu = smp_processor_id();
e8a0464c
DM
3504 for (i = 0; i < dev->num_tx_queues; i++) {
3505 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3506
3507 /* We are the only thread of execution doing a
3508 * freeze, but we have to grab the _xmit_lock in
3509 * order to synchronize with threads which are in
3510 * the ->hard_start_xmit() handler and already
3511 * checked the frozen bit.
3512 */
e8a0464c 3513 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3514 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3515 __netif_tx_unlock(txq);
e8a0464c 3516 }
932ff279
HX
3517}
3518
3519static inline void netif_tx_lock_bh(struct net_device *dev)
3520{
e8a0464c
DM
3521 local_bh_disable();
3522 netif_tx_lock(dev);
932ff279
HX
3523}
3524
932ff279
HX
3525static inline void netif_tx_unlock(struct net_device *dev)
3526{
e8a0464c
DM
3527 unsigned int i;
3528
3529 for (i = 0; i < dev->num_tx_queues; i++) {
3530 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3531
c3f26a26
DM
3532 /* No need to grab the _xmit_lock here. If the
3533 * queue is not stopped for another reason, we
3534 * force a schedule.
3535 */
3536 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3537 netif_schedule_queue(txq);
c3f26a26
DM
3538 }
3539 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3540}
3541
3542static inline void netif_tx_unlock_bh(struct net_device *dev)
3543{
e8a0464c
DM
3544 netif_tx_unlock(dev);
3545 local_bh_enable();
932ff279
HX
3546}
3547
c773e847 3548#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3549 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3550 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3551 } \
3552}
3553
5efeac44
EB
3554#define HARD_TX_TRYLOCK(dev, txq) \
3555 (((dev->features & NETIF_F_LLTX) == 0) ? \
3556 __netif_tx_trylock(txq) : \
3557 true )
3558
c773e847 3559#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3560 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3561 __netif_tx_unlock(txq); \
22dd7495
JHS
3562 } \
3563}
3564
1da177e4
LT
3565static inline void netif_tx_disable(struct net_device *dev)
3566{
fd2ea0a7 3567 unsigned int i;
c3f26a26 3568 int cpu;
fd2ea0a7 3569
c3f26a26
DM
3570 local_bh_disable();
3571 cpu = smp_processor_id();
fd2ea0a7
DM
3572 for (i = 0; i < dev->num_tx_queues; i++) {
3573 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3574
3575 __netif_tx_lock(txq, cpu);
fd2ea0a7 3576 netif_tx_stop_queue(txq);
c3f26a26 3577 __netif_tx_unlock(txq);
fd2ea0a7 3578 }
c3f26a26 3579 local_bh_enable();
1da177e4
LT
3580}
3581
e308a5d8
DM
3582static inline void netif_addr_lock(struct net_device *dev)
3583{
3584 spin_lock(&dev->addr_list_lock);
3585}
3586
2429f7ac
JP
3587static inline void netif_addr_lock_nested(struct net_device *dev)
3588{
25175ba5
VY
3589 int subclass = SINGLE_DEPTH_NESTING;
3590
3591 if (dev->netdev_ops->ndo_get_lock_subclass)
3592 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3593
3594 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3595}
3596
e308a5d8
DM
3597static inline void netif_addr_lock_bh(struct net_device *dev)
3598{
3599 spin_lock_bh(&dev->addr_list_lock);
3600}
3601
3602static inline void netif_addr_unlock(struct net_device *dev)
3603{
3604 spin_unlock(&dev->addr_list_lock);
3605}
3606
3607static inline void netif_addr_unlock_bh(struct net_device *dev)
3608{
3609 spin_unlock_bh(&dev->addr_list_lock);
3610}
3611
f001fde5 3612/*
31278e71 3613 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3614 * rcu_read_lock held.
3615 */
3616#define for_each_dev_addr(dev, ha) \
31278e71 3617 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3618
1da177e4
LT
3619/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3620
f629d208 3621void ether_setup(struct net_device *dev);
1da177e4
LT
3622
3623/* Support for loadable net-drivers */
f629d208 3624struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3625 unsigned char name_assign_type,
f629d208
JP
3626 void (*setup)(struct net_device *),
3627 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3628#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3629 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3630
c835a677
TG
3631#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3632 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3633 count)
36909ea4 3634
f629d208
JP
3635int register_netdev(struct net_device *dev);
3636void unregister_netdev(struct net_device *dev);
f001fde5 3637
22bedad3 3638/* General hardware address lists handling functions */
f629d208
JP
3639int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3640 struct netdev_hw_addr_list *from_list, int addr_len);
3641void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3642 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3643int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3644 struct net_device *dev,
3645 int (*sync)(struct net_device *, const unsigned char *),
3646 int (*unsync)(struct net_device *,
3647 const unsigned char *));
3648void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3649 struct net_device *dev,
3650 int (*unsync)(struct net_device *,
3651 const unsigned char *));
f629d208 3652void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3653
f001fde5 3654/* Functions used for device addresses handling */
f629d208
JP
3655int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3656 unsigned char addr_type);
3657int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3658 unsigned char addr_type);
f629d208
JP
3659void dev_addr_flush(struct net_device *dev);
3660int dev_addr_init(struct net_device *dev);
a748ee24
JP
3661
3662/* Functions used for unicast addresses handling */
f629d208
JP
3663int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3664int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3665int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3666int dev_uc_sync(struct net_device *to, struct net_device *from);
3667int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3668void dev_uc_unsync(struct net_device *to, struct net_device *from);
3669void dev_uc_flush(struct net_device *dev);
3670void dev_uc_init(struct net_device *dev);
f001fde5 3671
670e5b8e
AD
3672/**
3673 * __dev_uc_sync - Synchonize device's unicast list
3674 * @dev: device to sync
3675 * @sync: function to call if address should be added
3676 * @unsync: function to call if address should be removed
3677 *
3678 * Add newly added addresses to the interface, and release
3679 * addresses that have been deleted.
5e82b4b2 3680 */
670e5b8e
AD
3681static inline int __dev_uc_sync(struct net_device *dev,
3682 int (*sync)(struct net_device *,
3683 const unsigned char *),
3684 int (*unsync)(struct net_device *,
3685 const unsigned char *))
3686{
3687 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3688}
3689
3690/**
e793c0f7 3691 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3692 * @dev: device to sync
3693 * @unsync: function to call if address should be removed
3694 *
3695 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3696 */
670e5b8e
AD
3697static inline void __dev_uc_unsync(struct net_device *dev,
3698 int (*unsync)(struct net_device *,
3699 const unsigned char *))
3700{
3701 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3702}
3703
22bedad3 3704/* Functions used for multicast addresses handling */
f629d208
JP
3705int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3706int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3707int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3708int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3709int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3710int dev_mc_sync(struct net_device *to, struct net_device *from);
3711int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3712void dev_mc_unsync(struct net_device *to, struct net_device *from);
3713void dev_mc_flush(struct net_device *dev);
3714void dev_mc_init(struct net_device *dev);
f001fde5 3715
670e5b8e
AD
3716/**
3717 * __dev_mc_sync - Synchonize device's multicast list
3718 * @dev: device to sync
3719 * @sync: function to call if address should be added
3720 * @unsync: function to call if address should be removed
3721 *
3722 * Add newly added addresses to the interface, and release
3723 * addresses that have been deleted.
5e82b4b2 3724 */
670e5b8e
AD
3725static inline int __dev_mc_sync(struct net_device *dev,
3726 int (*sync)(struct net_device *,
3727 const unsigned char *),
3728 int (*unsync)(struct net_device *,
3729 const unsigned char *))
3730{
3731 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3732}
3733
3734/**
e793c0f7 3735 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3736 * @dev: device to sync
3737 * @unsync: function to call if address should be removed
3738 *
3739 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3740 */
670e5b8e
AD
3741static inline void __dev_mc_unsync(struct net_device *dev,
3742 int (*unsync)(struct net_device *,
3743 const unsigned char *))
3744{
3745 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3746}
3747
4417da66 3748/* Functions used for secondary unicast and multicast support */
f629d208
JP
3749void dev_set_rx_mode(struct net_device *dev);
3750void __dev_set_rx_mode(struct net_device *dev);
3751int dev_set_promiscuity(struct net_device *dev, int inc);
3752int dev_set_allmulti(struct net_device *dev, int inc);
3753void netdev_state_change(struct net_device *dev);
3754void netdev_notify_peers(struct net_device *dev);
3755void netdev_features_change(struct net_device *dev);
1da177e4 3756/* Load a device via the kmod */
f629d208
JP
3757void dev_load(struct net *net, const char *name);
3758struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3759 struct rtnl_link_stats64 *storage);
3760void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3761 const struct net_device_stats *netdev_stats);
eeda3fd6 3762
1da177e4 3763extern int netdev_max_backlog;
3b098e2d 3764extern int netdev_tstamp_prequeue;
1da177e4 3765extern int weight_p;
9ff162a8 3766
f629d208 3767bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3768struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3769 struct list_head **iter);
f629d208
JP
3770struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3771 struct list_head **iter);
8b5be856 3772
44a40855
VY
3773/* iterate through upper list, must be called under RCU read lock */
3774#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3775 for (iter = &(dev)->adj_list.upper, \
3776 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3777 updev; \
3778 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3779
8b5be856 3780/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3781#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3782 for (iter = &(dev)->all_adj_list.upper, \
3783 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3784 updev; \
3785 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3786
f629d208
JP
3787void *netdev_lower_get_next_private(struct net_device *dev,
3788 struct list_head **iter);
3789void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3790 struct list_head **iter);
31088a11
VF
3791
3792#define netdev_for_each_lower_private(dev, priv, iter) \
3793 for (iter = (dev)->adj_list.lower.next, \
3794 priv = netdev_lower_get_next_private(dev, &(iter)); \
3795 priv; \
3796 priv = netdev_lower_get_next_private(dev, &(iter)))
3797
3798#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3799 for (iter = &(dev)->adj_list.lower, \
3800 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3801 priv; \
3802 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3803
4085ebe8
VY
3804void *netdev_lower_get_next(struct net_device *dev,
3805 struct list_head **iter);
3806#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3807 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3808 ldev = netdev_lower_get_next(dev, &(iter)); \
3809 ldev; \
3810 ldev = netdev_lower_get_next(dev, &(iter)))
3811
f629d208 3812void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3813void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3814struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3815struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3816int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3817int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3818 struct net_device *upper_dev,
29bf24af 3819 void *upper_priv, void *upper_info);
f629d208
JP
3820void netdev_upper_dev_unlink(struct net_device *dev,
3821 struct net_device *upper_dev);
5bb025fa 3822void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3823void *netdev_lower_dev_get_private(struct net_device *dev,
3824 struct net_device *lower_dev);
04d48266
JP
3825void netdev_lower_state_changed(struct net_device *lower_dev,
3826 void *lower_state_info);
960fb622
ED
3827
3828/* RSS keys are 40 or 52 bytes long */
3829#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3830extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3831void netdev_rss_key_fill(void *buffer, size_t len);
3832
4085ebe8 3833int dev_get_nest_level(struct net_device *dev,
b618aaa9 3834 bool (*type_check)(const struct net_device *dev));
f629d208
JP
3835int skb_checksum_help(struct sk_buff *skb);
3836struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3837 netdev_features_t features, bool tx_path);
3838struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3839 netdev_features_t features);
12b0004d 3840
61bd3857
MS
3841struct netdev_bonding_info {
3842 ifslave slave;
3843 ifbond master;
3844};
3845
3846struct netdev_notifier_bonding_info {
3847 struct netdev_notifier_info info; /* must be first */
3848 struct netdev_bonding_info bonding_info;
3849};
3850
3851void netdev_bonding_info_change(struct net_device *dev,
3852 struct netdev_bonding_info *bonding_info);
3853
12b0004d
CW
3854static inline
3855struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3856{
3857 return __skb_gso_segment(skb, features, true);
3858}
53d6471c 3859__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3860
3861static inline bool can_checksum_protocol(netdev_features_t features,
3862 __be16 protocol)
3863{
c8cd0989
TH
3864 if (protocol == htons(ETH_P_FCOE))
3865 return !!(features & NETIF_F_FCOE_CRC);
3866
3867 /* Assume this is an IP checksum (not SCTP CRC) */
3868
3869 if (features & NETIF_F_HW_CSUM) {
3870 /* Can checksum everything */
3871 return true;
3872 }
3873
3874 switch (protocol) {
3875 case htons(ETH_P_IP):
3876 return !!(features & NETIF_F_IP_CSUM);
3877 case htons(ETH_P_IPV6):
3878 return !!(features & NETIF_F_IPV6_CSUM);
3879 default:
3880 return false;
3881 }
ec5f0615 3882}
12b0004d 3883
6ae23ad3
TH
3884/* Map an ethertype into IP protocol if possible */
3885static inline int eproto_to_ipproto(int eproto)
3886{
3887 switch (eproto) {
3888 case htons(ETH_P_IP):
3889 return IPPROTO_IP;
3890 case htons(ETH_P_IPV6):
3891 return IPPROTO_IPV6;
3892 default:
3893 return -1;
3894 }
3895}
3896
fb286bb2 3897#ifdef CONFIG_BUG
f629d208 3898void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3899#else
3900static inline void netdev_rx_csum_fault(struct net_device *dev)
3901{
3902}
3903#endif
1da177e4 3904/* rx skb timestamps */
f629d208
JP
3905void net_enable_timestamp(void);
3906void net_disable_timestamp(void);
1da177e4 3907
20380731 3908#ifdef CONFIG_PROC_FS
f629d208 3909int __init dev_proc_init(void);
900ff8c6
CW
3910#else
3911#define dev_proc_init() 0
20380731
ACM
3912#endif
3913
4798248e 3914static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3915 struct sk_buff *skb, struct net_device *dev,
3916 bool more)
4798248e 3917{
fa2dbdc2 3918 skb->xmit_more = more ? 1 : 0;
0b725a2c 3919 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3920}
3921
10b3ad8c 3922static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3923 struct netdev_queue *txq, bool more)
4798248e
DM
3924{
3925 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3926 int rc;
4798248e 3927
fa2dbdc2 3928 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3929 if (rc == NETDEV_TX_OK)
3930 txq_trans_update(txq);
3931
3932 return rc;
4798248e
DM
3933}
3934
42a2d923
LT
3935int netdev_class_create_file_ns(struct class_attribute *class_attr,
3936 const void *ns);
3937void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3938 const void *ns);
58292cbe
TH
3939
3940static inline int netdev_class_create_file(struct class_attribute *class_attr)
3941{
3942 return netdev_class_create_file_ns(class_attr, NULL);
3943}
3944
3945static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3946{
3947 netdev_class_remove_file_ns(class_attr, NULL);
3948}
b8a9787e 3949
04600794
JB
3950extern struct kobj_ns_type_operations net_ns_type_operations;
3951
f629d208 3952const char *netdev_drivername(const struct net_device *dev);
6579e57b 3953
f629d208 3954void linkwatch_run_queue(void);
20380731 3955
da08143b
MK
3956static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3957 netdev_features_t f2)
3958{
c8cd0989
TH
3959 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3960 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 3961 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 3962 else
b6a0e72a 3963 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 3964 }
da08143b 3965
c8cd0989 3966 return f1 & f2;
da08143b
MK
3967}
3968
c8f44aff
MM
3969static inline netdev_features_t netdev_get_wanted_features(
3970 struct net_device *dev)
5455c699
MM
3971{
3972 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3973}
c8f44aff
MM
3974netdev_features_t netdev_increment_features(netdev_features_t all,
3975 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3976
3977/* Allow TSO being used on stacked device :
3978 * Performing the GSO segmentation before last device
3979 * is a performance improvement.
3980 */
3981static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3982 netdev_features_t mask)
3983{
3984 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3985}
3986
6cb6a27c 3987int __netdev_update_features(struct net_device *dev);
5455c699 3988void netdev_update_features(struct net_device *dev);
afe12cc8 3989void netdev_change_features(struct net_device *dev);
7f353bf2 3990
fc4a7489
PM
3991void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3992 struct net_device *dev);
3993
e38f3025
TM
3994netdev_features_t passthru_features_check(struct sk_buff *skb,
3995 struct net_device *dev,
3996 netdev_features_t features);
c1e756bf 3997netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3998
4d29515f 3999static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4000{
7b748340 4001 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4002
4003 /* check flags correspondence */
4004 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4005 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4006 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4007 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4008 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4009 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4010 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4011 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4012 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4013 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4014 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4015 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4016 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4017 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4018 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4019 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
0345e186 4020
d6b4991a 4021 return (features & feature) == feature;
576a30eb
HX
4022}
4023
4d29515f 4024static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4025{
278b2513 4026 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4027 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4028}
4029
8b86a61d 4030static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4031 netdev_features_t features)
7967168c 4032{
fc741216 4033 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4034 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4035 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4036}
4037
82cc1a7a
PWJ
4038static inline void netif_set_gso_max_size(struct net_device *dev,
4039 unsigned int size)
4040{
4041 dev->gso_max_size = size;
4042}
4043
7a7ffbab
WCC
4044static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4045 int pulled_hlen, u16 mac_offset,
4046 int mac_len)
4047{
4048 skb->protocol = protocol;
4049 skb->encapsulation = 1;
4050 skb_push(skb, pulled_hlen);
4051 skb_reset_transport_header(skb);
4052 skb->mac_header = mac_offset;
4053 skb->network_header = skb->mac_header + mac_len;
4054 skb->mac_len = mac_len;
4055}
4056
3c175784
SD
4057static inline bool netif_is_macsec(const struct net_device *dev)
4058{
4059 return dev->priv_flags & IFF_MACSEC;
4060}
4061
b618aaa9 4062static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4063{
4064 return dev->priv_flags & IFF_MACVLAN;
4065}
4066
b618aaa9 4067static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4068{
4069 return dev->priv_flags & IFF_MACVLAN_PORT;
4070}
4071
b618aaa9 4072static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4073{
4074 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4075}
4076
b618aaa9 4077static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4078{
4079 return dev->priv_flags & IFF_IPVLAN_MASTER;
4080}
4081
b618aaa9 4082static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4083{
4084 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4085}
4086
b618aaa9 4087static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4088{
4089 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4090}
4091
3bdc0eba
BG
4092static inline bool netif_supports_nofcs(struct net_device *dev)
4093{
4094 return dev->priv_flags & IFF_SUPP_NOFCS;
4095}
4096
007979ea 4097static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4098{
007979ea 4099 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4100}
4101
fee6d4c7
DA
4102static inline bool netif_is_l3_slave(const struct net_device *dev)
4103{
4104 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4105}
4106
0894ae3f
JP
4107static inline bool netif_is_bridge_master(const struct net_device *dev)
4108{
4109 return dev->priv_flags & IFF_EBRIDGE;
4110}
4111
28f9ee22
VY
4112static inline bool netif_is_bridge_port(const struct net_device *dev)
4113{
4114 return dev->priv_flags & IFF_BRIDGE_PORT;
4115}
4116
35d4e172
JP
4117static inline bool netif_is_ovs_master(const struct net_device *dev)
4118{
4119 return dev->priv_flags & IFF_OPENVSWITCH;
4120}
4121
b618aaa9 4122static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4123{
4124 return dev->priv_flags & IFF_TEAM;
4125}
4126
b618aaa9 4127static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4128{
4129 return dev->priv_flags & IFF_TEAM_PORT;
4130}
4131
b618aaa9 4132static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4133{
4134 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4135}
4136
b618aaa9 4137static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4138{
4139 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4140}
4141
d4ab4286
KJ
4142static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4143{
4144 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4145}
4146
02875878
ED
4147/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4148static inline void netif_keep_dst(struct net_device *dev)
4149{
4150 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4151}
4152
505d4f73 4153extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4154
571ba423
JP
4155/* Logging, debugging and troubleshooting/diagnostic helpers. */
4156
4157/* netdev_printk helpers, similar to dev_printk */
4158
4159static inline const char *netdev_name(const struct net_device *dev)
4160{
c6f854d5
VF
4161 if (!dev->name[0] || strchr(dev->name, '%'))
4162 return "(unnamed net_device)";
571ba423
JP
4163 return dev->name;
4164}
4165
ccc7f496
VF
4166static inline const char *netdev_reg_state(const struct net_device *dev)
4167{
4168 switch (dev->reg_state) {
4169 case NETREG_UNINITIALIZED: return " (uninitialized)";
4170 case NETREG_REGISTERED: return "";
4171 case NETREG_UNREGISTERING: return " (unregistering)";
4172 case NETREG_UNREGISTERED: return " (unregistered)";
4173 case NETREG_RELEASED: return " (released)";
4174 case NETREG_DUMMY: return " (dummy)";
4175 }
4176
4177 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4178 return " (unknown)";
4179}
4180
f629d208 4181__printf(3, 4)
6ea754eb
JP
4182void netdev_printk(const char *level, const struct net_device *dev,
4183 const char *format, ...);
f629d208 4184__printf(2, 3)
6ea754eb 4185void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4186__printf(2, 3)
6ea754eb 4187void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4188__printf(2, 3)
6ea754eb 4189void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4190__printf(2, 3)
6ea754eb 4191void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4192__printf(2, 3)
6ea754eb 4193void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4194__printf(2, 3)
6ea754eb 4195void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4196__printf(2, 3)
6ea754eb 4197void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4198
8909c9ad
VK
4199#define MODULE_ALIAS_NETDEV(device) \
4200 MODULE_ALIAS("netdev-" device)
4201
b558c96f 4202#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4203#define netdev_dbg(__dev, format, args...) \
4204do { \
ffa10cb4 4205 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4206} while (0)
b558c96f
JC
4207#elif defined(DEBUG)
4208#define netdev_dbg(__dev, format, args...) \
4209 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4210#else
4211#define netdev_dbg(__dev, format, args...) \
4212({ \
4213 if (0) \
4214 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4215})
4216#endif
4217
4218#if defined(VERBOSE_DEBUG)
4219#define netdev_vdbg netdev_dbg
4220#else
4221
4222#define netdev_vdbg(dev, format, args...) \
4223({ \
4224 if (0) \
4225 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4226 0; \
4227})
4228#endif
4229
4230/*
4231 * netdev_WARN() acts like dev_printk(), but with the key difference
4232 * of using a WARN/WARN_ON to get the message out, including the
4233 * file/line information and a backtrace.
4234 */
4235#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4236 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4237 netdev_reg_state(dev), ##args)
571ba423 4238
b3d95c5c
JP
4239/* netif printk helpers, similar to netdev_printk */
4240
4241#define netif_printk(priv, type, level, dev, fmt, args...) \
4242do { \
4243 if (netif_msg_##type(priv)) \
4244 netdev_printk(level, (dev), fmt, ##args); \
4245} while (0)
4246
f45f4321
JP
4247#define netif_level(level, priv, type, dev, fmt, args...) \
4248do { \
4249 if (netif_msg_##type(priv)) \
4250 netdev_##level(dev, fmt, ##args); \
4251} while (0)
4252
b3d95c5c 4253#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4254 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4255#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4256 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4257#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4258 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4259#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4260 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4261#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4262 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4263#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4264 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4265#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4266 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4267
0053ea9c 4268#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4269#define netif_dbg(priv, type, netdev, format, args...) \
4270do { \
4271 if (netif_msg_##type(priv)) \
b5fb0a03 4272 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4273} while (0)
0053ea9c
JP
4274#elif defined(DEBUG)
4275#define netif_dbg(priv, type, dev, format, args...) \
4276 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4277#else
4278#define netif_dbg(priv, type, dev, format, args...) \
4279({ \
4280 if (0) \
4281 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4282 0; \
4283})
4284#endif
4285
4286#if defined(VERBOSE_DEBUG)
bcfcc450 4287#define netif_vdbg netif_dbg
b3d95c5c
JP
4288#else
4289#define netif_vdbg(priv, type, dev, format, args...) \
4290({ \
4291 if (0) \
a4ed89cb 4292 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4293 0; \
4294})
4295#endif
571ba423 4296
900ff8c6
CW
4297/*
4298 * The list of packet types we will receive (as opposed to discard)
4299 * and the routines to invoke.
4300 *
4301 * Why 16. Because with 16 the only overlap we get on a hash of the
4302 * low nibble of the protocol value is RARP/SNAP/X.25.
4303 *
4304 * NOTE: That is no longer true with the addition of VLAN tags. Not
4305 * sure which should go first, but I bet it won't make much
4306 * difference if we are running VLANs. The good news is that
4307 * this protocol won't be in the list unless compiled in, so
4308 * the average user (w/out VLANs) will not be adversely affected.
4309 * --BLG
4310 *
4311 * 0800 IP
4312 * 8100 802.1Q VLAN
4313 * 0001 802.3
4314 * 0002 AX.25
4315 * 0004 802.2
4316 * 8035 RARP
4317 * 0005 SNAP
4318 * 0805 X.25
4319 * 0806 ARP
4320 * 8137 IPX
4321 * 0009 Localtalk
4322 * 86DD IPv6
4323 */
4324#define PTYPE_HASH_SIZE (16)
4325#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4326
385a154c 4327#endif /* _LINUX_NETDEVICE_H */