]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/netdevice.h
Merge branch 'mlxsw-fixes'
[mirror_ubuntu-bionic-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>
59cc1f61 55#include <linux/hashtable.h>
a59e2ecb 56
115c1d6e 57struct netpoll_info;
313162d0 58struct device;
c1f19b51 59struct phy_device;
704232c2
JB
60/* 802.11 specific */
61struct wireless_dev;
98a18b6f
AA
62/* 802.15.4 specific */
63struct wpan_dev;
03c57747 64struct mpls_dev;
7c46a640
AD
65/* UDP Tunnel offloads */
66struct udp_tunnel_info;
a7862b45 67struct bpf_prog;
1da177e4 68
f629d208
JP
69void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
d07d7507 71
9a1654ba
JP
72/* Backlog congestion levels */
73#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74#define NET_RX_DROP 1 /* packet dropped */
75
572a9d7b
PM
76/*
77 * Transmit return codes: transmit return codes originate from three different
78 * namespaces:
79 *
80 * - qdisc return codes
81 * - driver transmit return codes
82 * - errno values
83 *
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
90 * others are propagated to higher layers.
91 */
92
93/* qdisc ->enqueue() return codes. */
94#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
95#define NET_XMIT_DROP 0x01 /* skb dropped */
96#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 97#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 98
b9df3cb8
GR
99/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 102#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
103#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
104
dc1f8bf6 105/* Driver transmit return codes */
9a1654ba 106#define NETDEV_TX_MASK 0xf0
572a9d7b 107
dc1f8bf6 108enum netdev_tx {
572a9d7b 109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
112};
113typedef enum netdev_tx netdev_tx_t;
114
9a1654ba
JP
115/*
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
118 */
119static inline bool dev_xmit_complete(int rc)
120{
121 /*
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
126 */
127 if (likely(rc < NET_XMIT_MASK))
128 return true;
129
130 return false;
131}
132
1da177e4 133/*
5e82b4b2 134 * Compute the worst-case header length according to the protocols
1da177e4
LT
135 * used.
136 */
fe2918b0 137
c0eb4540
KS
138#if defined(CONFIG_HYPERV_NET)
139# define LL_MAX_HEADER 128
140#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
141# if defined(CONFIG_MAC80211_MESH)
142# define LL_MAX_HEADER 128
143# else
144# define LL_MAX_HEADER 96
145# endif
1da177e4 146#else
8388e3da 147# define LL_MAX_HEADER 32
1da177e4
LT
148#endif
149
d11ead75
BH
150#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
152#define MAX_HEADER LL_MAX_HEADER
153#else
154#define MAX_HEADER (LL_MAX_HEADER + 48)
155#endif
156
157/*
be1f3c2c
BH
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
1da177e4 160 */
fe2918b0 161
d94d9fee 162struct net_device_stats {
3cfde79c
BH
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
1da177e4 172 unsigned long collisions;
1da177e4 173 unsigned long rx_length_errors;
3cfde79c
BH
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
1da177e4
LT
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
1da177e4
LT
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
186};
187
1da177e4
LT
188
189#include <linux/cache.h>
190#include <linux/skbuff.h>
191
adc9300e 192#ifdef CONFIG_RPS
c5905afb
IM
193#include <linux/static_key.h>
194extern struct static_key rps_needed;
adc9300e
ED
195#endif
196
1da177e4
LT
197struct neighbour;
198struct neigh_parms;
199struct sk_buff;
200
f001fde5
JP
201struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
ccffad25
JP
205#define NETDEV_HW_ADDR_T_LAN 1
206#define NETDEV_HW_ADDR_T_SAN 2
207#define NETDEV_HW_ADDR_T_SLAVE 3
208#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 209#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 210 bool global_use;
4cd729b0 211 int sync_cnt;
8f8f103d 212 int refcount;
4543fbef 213 int synced;
f001fde5
JP
214 struct rcu_head rcu_head;
215};
216
31278e71
JP
217struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220};
221
22bedad3
JP
222#define netdev_hw_addr_list_count(l) ((l)->count)
223#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224#define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 226
22bedad3
JP
227#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229#define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 231
22bedad3
JP
232#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 234#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 236
d94d9fee 237struct hh_cache {
f6b72b62 238 u16 hh_len;
5c25f686 239 u16 __pad;
3644f0ce 240 seqlock_t hh_lock;
1da177e4
LT
241
242 /* cached hardware header; allow for machine alignment needs. */
243#define HH_DATA_MOD 16
244#define HH_DATA_OFF(__len) \
5ba0eac6 245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
246#define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
249};
250
5e82b4b2 251/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
252 * Alternative is:
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
255 *
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
258 */
259#define LL_RESERVED_SPACE(dev) \
f5184d26 260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 261#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 263
3b04ddde
SH
264struct header_ops {
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
95c96174 267 const void *saddr, unsigned int len);
3b04ddde 268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
270 void (*cache_update)(struct hh_cache *hh,
271 const struct net_device *dev,
272 const unsigned char *haddr);
2793a23a 273 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
274};
275
1da177e4 276/* These flag bits are private to the generic network queueing
5e82b4b2 277 * layer; they may not be explicitly referenced by any other
1da177e4
LT
278 * code.
279 */
280
d94d9fee 281enum netdev_state_t {
1da177e4
LT
282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
1da177e4 284 __LINK_STATE_NOCARRIER,
b00055aa
SR
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
1da177e4
LT
287};
288
289
290/*
5e82b4b2 291 * This structure holds boot-time configured netdevice settings. They
fe2918b0 292 * are then used in the device probing.
1da177e4
LT
293 */
294struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297};
298#define NETDEV_BOOT_SETUP_MAX 8
299
f629d208 300int __init netdev_boot_setup(char *str);
1da177e4 301
bea3348e
SH
302/*
303 * Structure for NAPI scheduling similar to tasklet but with weighting
304 */
305struct napi_struct {
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 309 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
310 * can remove from the list right before clearing the bit.
311 */
312 struct list_head poll_list;
313
314 unsigned long state;
315 int weight;
404f7c9e 316 unsigned int gro_count;
bea3348e
SH
317 int (*poll)(struct napi_struct *, int);
318#ifdef CONFIG_NETPOLL
319 spinlock_t poll_lock;
320 int poll_owner;
bea3348e 321#endif
5d38a079 322 struct net_device *dev;
d565b0a1 323 struct sk_buff *gro_list;
5d38a079 324 struct sk_buff *skb;
3b47d303 325 struct hrtimer timer;
404f7c9e 326 struct list_head dev_list;
af12fa6e
ET
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
bea3348e
SH
329};
330
d94d9fee 331enum {
bea3348e 332 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 333 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
bea3348e
SH
337};
338
5b252f0c 339enum gro_result {
d1c76af9
HX
340 GRO_MERGED,
341 GRO_MERGED_FREE,
342 GRO_HELD,
343 GRO_NORMAL,
344 GRO_DROP,
345};
5b252f0c 346typedef enum gro_result gro_result_t;
d1c76af9 347
8a4eb573
JP
348/*
349 * enum rx_handler_result - Possible return values for rx_handlers.
350 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
351 * further.
352 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
353 * case skb->dev was changed by rx_handler.
354 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 355 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
356 *
357 * rx_handlers are functions called from inside __netif_receive_skb(), to do
358 * special processing of the skb, prior to delivery to protocol handlers.
359 *
360 * Currently, a net_device can only have a single rx_handler registered. Trying
361 * to register a second rx_handler will return -EBUSY.
362 *
363 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
364 * To unregister a rx_handler on a net_device, use
365 * netdev_rx_handler_unregister().
366 *
367 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
368 * do with the skb.
369 *
5e82b4b2 370 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 371 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 372 * the skb to be delivered in some other way.
8a4eb573
JP
373 *
374 * If the rx_handler changed skb->dev, to divert the skb to another
375 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
376 * new device will be called if it exists.
377 *
5e82b4b2 378 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 379 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 380 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 381 *
5e82b4b2 382 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
383 * delivered, it should return RX_HANDLER_PASS.
384 *
385 * A device without a registered rx_handler will behave as if rx_handler
386 * returned RX_HANDLER_PASS.
387 */
388
389enum rx_handler_result {
390 RX_HANDLER_CONSUMED,
391 RX_HANDLER_ANOTHER,
392 RX_HANDLER_EXACT,
393 RX_HANDLER_PASS,
394};
395typedef enum rx_handler_result rx_handler_result_t;
396typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 397
f629d208 398void __napi_schedule(struct napi_struct *n);
bc9ad166 399void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 400
4d29515f 401static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
402{
403 return test_bit(NAPI_STATE_DISABLE, &n->state);
404}
405
bea3348e 406/**
5e82b4b2
BH
407 * napi_schedule_prep - check if NAPI can be scheduled
408 * @n: NAPI context
bea3348e
SH
409 *
410 * Test if NAPI routine is already running, and if not mark
5e82b4b2 411 * it as running. This is used as a condition variable to
a0a46196
DM
412 * insure only one NAPI poll instance runs. We also make
413 * sure there is no pending NAPI disable.
bea3348e 414 */
4d29515f 415static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 416{
a0a46196
DM
417 return !napi_disable_pending(n) &&
418 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
419}
420
421/**
422 * napi_schedule - schedule NAPI poll
5e82b4b2 423 * @n: NAPI context
bea3348e
SH
424 *
425 * Schedule NAPI poll routine to be called if it is not already
426 * running.
427 */
428static inline void napi_schedule(struct napi_struct *n)
429{
430 if (napi_schedule_prep(n))
431 __napi_schedule(n);
432}
433
bc9ad166
ED
434/**
435 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 436 * @n: NAPI context
bc9ad166
ED
437 *
438 * Variant of napi_schedule(), assuming hard irqs are masked.
439 */
440static inline void napi_schedule_irqoff(struct napi_struct *n)
441{
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
444}
445
bfe13f54 446/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 447static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
448{
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
4d29515f 451 return true;
bfe13f54 452 }
4d29515f 453 return false;
bfe13f54
RD
454}
455
3b47d303
ED
456void __napi_complete(struct napi_struct *n);
457void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
458/**
459 * napi_complete - NAPI processing complete
5e82b4b2 460 * @n: NAPI context
bea3348e
SH
461 *
462 * Mark NAPI processing as complete.
3b47d303 463 * Consider using napi_complete_done() instead.
bea3348e 464 */
3b47d303
ED
465static inline void napi_complete(struct napi_struct *n)
466{
467 return napi_complete_done(n, 0);
468}
bea3348e 469
af12fa6e
ET
470/**
471 * napi_hash_add - add a NAPI to global hashtable
5e82b4b2 472 * @napi: NAPI context
af12fa6e 473 *
5e82b4b2
BH
474 * Generate a new napi_id and store a @napi under it in napi_hash.
475 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
93d05d4a 476 * Note: This is normally automatically done from netif_napi_add(),
5e82b4b2 477 * so might disappear in a future Linux version.
af12fa6e 478 */
f629d208 479void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
480
481/**
482 * napi_hash_del - remove a NAPI from global table
5e82b4b2 483 * @napi: NAPI context
af12fa6e 484 *
5e82b4b2 485 * Warning: caller must observe RCU grace period
34cbe27e
ED
486 * before freeing memory containing @napi, if
487 * this function returns true.
93d05d4a 488 * Note: core networking stack automatically calls it
5e82b4b2 489 * from netif_napi_del().
93d05d4a 490 * Drivers might want to call this helper to combine all
5e82b4b2 491 * the needed RCU grace periods into a single one.
af12fa6e 492 */
34cbe27e 493bool napi_hash_del(struct napi_struct *napi);
af12fa6e 494
bea3348e
SH
495/**
496 * napi_disable - prevent NAPI from scheduling
5e82b4b2 497 * @n: NAPI context
bea3348e
SH
498 *
499 * Stop NAPI from being scheduled on this context.
500 * Waits till any outstanding processing completes.
501 */
3b47d303 502void napi_disable(struct napi_struct *n);
bea3348e
SH
503
504/**
505 * napi_enable - enable NAPI scheduling
5e82b4b2 506 * @n: NAPI context
bea3348e
SH
507 *
508 * Resume NAPI from being scheduled on this context.
509 * Must be paired with napi_disable.
510 */
511static inline void napi_enable(struct napi_struct *n)
512{
513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 514 smp_mb__before_atomic();
bea3348e 515 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 516 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
517}
518
c264c3de
SH
519/**
520 * napi_synchronize - wait until NAPI is not running
5e82b4b2 521 * @n: NAPI context
c264c3de
SH
522 *
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
526 */
527static inline void napi_synchronize(const struct napi_struct *n)
528{
facc432f
AB
529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
531 msleep(1);
532 else
533 barrier();
c264c3de 534}
c264c3de 535
d94d9fee 536enum netdev_queue_state_t {
73466498
TH
537 __QUEUE_STATE_DRV_XOFF,
538 __QUEUE_STATE_STACK_XOFF,
c3f26a26 539 __QUEUE_STATE_FROZEN,
79d16385 540};
8e2f1a63
DB
541
542#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
543#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
544#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
545
546#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 QUEUE_STATE_FROZEN)
549#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
550 QUEUE_STATE_FROZEN)
551
73466498
TH
552/*
553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
554 * netif_tx_* functions below are used to manipulate this flag. The
555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556 * queue independently. The netif_xmit_*stopped functions below are called
557 * to check if the queue has been stopped by the driver or stack (either
558 * of the XOFF bits are set in the state). Drivers should not need to call
559 * netif_xmit*stopped functions, they should only be using netif_tx_*.
560 */
79d16385 561
bb949fbd 562struct netdev_queue {
6a321cb3 563/*
5e82b4b2 564 * read-mostly part
6a321cb3 565 */
bb949fbd 566 struct net_device *dev;
46e5da40 567 struct Qdisc __rcu *qdisc;
b0e1e646 568 struct Qdisc *qdisc_sleeping;
ccf5ff69 569#ifdef CONFIG_SYSFS
1d24eb48
TH
570 struct kobject kobj;
571#endif
f2cd2d3e
ED
572#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 int numa_node;
574#endif
c0ef079c
FW
575 unsigned long tx_maxrate;
576 /*
577 * Number of TX timeouts for this queue
578 * (/sys/class/net/DEV/Q/trans_timeout)
579 */
580 unsigned long trans_timeout;
6a321cb3 581/*
5e82b4b2 582 * write-mostly part
6a321cb3
ED
583 */
584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
585 int xmit_lock_owner;
9d21493b 586 /*
9b36627a 587 * Time (in jiffies) of last Tx
9d21493b
ED
588 */
589 unsigned long trans_start;
ccf5ff69 590
114cf580
TH
591 unsigned long state;
592
593#ifdef CONFIG_BQL
594 struct dql dql;
595#endif
e8a0464c 596} ____cacheline_aligned_in_smp;
bb949fbd 597
f2cd2d3e
ED
598static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599{
600#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 return q->numa_node;
602#else
b236da69 603 return NUMA_NO_NODE;
f2cd2d3e
ED
604#endif
605}
606
607static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608{
609#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 q->numa_node = node;
611#endif
612}
613
df334545 614#ifdef CONFIG_RPS
0a9627f2
TH
615/*
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
618 */
619struct rps_map {
620 unsigned int len;
621 struct rcu_head rcu;
622 u16 cpus[0];
623};
60b778ce 624#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 625
fec5e652 626/*
c445477d
BH
627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628 * tail pointer for that CPU's input queue at the time of last enqueue, and
629 * a hardware filter index.
fec5e652
TH
630 */
631struct rps_dev_flow {
632 u16 cpu;
c445477d 633 u16 filter;
fec5e652
TH
634 unsigned int last_qtail;
635};
c445477d 636#define RPS_NO_FILTER 0xffff
fec5e652
TH
637
638/*
639 * The rps_dev_flow_table structure contains a table of flow mappings.
640 */
641struct rps_dev_flow_table {
642 unsigned int mask;
643 struct rcu_head rcu;
fec5e652
TH
644 struct rps_dev_flow flows[0];
645};
646#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 647 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
648
649/*
650 * The rps_sock_flow_table contains mappings of flows to the last CPU
651 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
652 * Each entry is a 32bit value. Upper part is the high-order bits
653 * of flow hash, lower part is CPU number.
567e4b79 654 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 657 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
658 */
659struct rps_sock_flow_table {
567e4b79 660 u32 mask;
93c1af6c
ED
661
662 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 663};
567e4b79 664#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
665
666#define RPS_NO_CPU 0xffff
667
567e4b79
ED
668extern u32 rps_cpu_mask;
669extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670
fec5e652
TH
671static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
672 u32 hash)
673{
674 if (table && hash) {
567e4b79
ED
675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
fec5e652 677
5e82b4b2 678 /* We only give a hint, preemption can change CPU under us */
567e4b79 679 val |= raw_smp_processor_id();
fec5e652 680
567e4b79
ED
681 if (table->ents[index] != val)
682 table->ents[index] = val;
fec5e652
TH
683 }
684}
685
c445477d 686#ifdef CONFIG_RFS_ACCEL
f629d208
JP
687bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
688 u16 filter_id);
c445477d 689#endif
a953be53 690#endif /* CONFIG_RPS */
c445477d 691
0a9627f2
TH
692/* This structure contains an instance of an RX queue. */
693struct netdev_rx_queue {
a953be53 694#ifdef CONFIG_RPS
6e3f7faf
ED
695 struct rps_map __rcu *rps_map;
696 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 697#endif
6e3f7faf 698 struct kobject kobj;
fe822240 699 struct net_device *dev;
0a9627f2 700} ____cacheline_aligned_in_smp;
a953be53
MD
701
702/*
703 * RX queue sysfs structures and functions.
704 */
705struct rx_queue_attribute {
706 struct attribute attr;
707 ssize_t (*show)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, char *buf);
709 ssize_t (*store)(struct netdev_rx_queue *queue,
710 struct rx_queue_attribute *attr, const char *buf, size_t len);
711};
d314774c 712
bf264145
TH
713#ifdef CONFIG_XPS
714/*
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
717 */
718struct xps_map {
719 unsigned int len;
720 unsigned int alloc_len;
721 struct rcu_head rcu;
722 u16 queues[0];
723};
60b778ce 724#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
725#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
727
728/*
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 */
731struct xps_dev_maps {
732 struct rcu_head rcu;
a4177869 733 struct xps_map __rcu *cpu_map[0];
bf264145
TH
734};
735#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
736 (nr_cpu_ids * sizeof(struct xps_map *)))
737#endif /* CONFIG_XPS */
738
4f57c087
JF
739#define TC_MAX_QUEUE 16
740#define TC_BITMASK 15
741/* HW offloaded queuing disciplines txq count and offset maps */
742struct netdev_tc_txq {
743 u16 count;
744 u16 offset;
745};
746
68bad94e
NP
747#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748/*
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
751 */
752struct netdev_fcoe_hbainfo {
753 char manufacturer[64];
754 char serial_number[64];
755 char hardware_version[64];
756 char driver_version[64];
757 char optionrom_version[64];
758 char firmware_version[64];
759 char model[256];
760 char model_description[256];
761};
762#endif
763
02637fce 764#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 765
02637fce
JP
766/* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
66b52b0d 768 */
02637fce
JP
769struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
771 unsigned char id_len;
772};
773
d754f98b
SF
774static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
776{
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
779}
780
99932d4f
DB
781typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
783
a1b7c5fd 784/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
785 * that are being passed to the netdevice through the setup_tc op.
786 */
787enum {
788 TC_SETUP_MQPRIO,
a1b7c5fd 789 TC_SETUP_CLSU32,
5b33f488 790 TC_SETUP_CLSFLOWER,
b87f7936 791 TC_SETUP_MATCHALL,
332ae8e2 792 TC_SETUP_CLSBPF,
16e5cc64
JF
793};
794
a1b7c5fd
JF
795struct tc_cls_u32_offload;
796
16e5cc64
JF
797struct tc_to_netdev {
798 unsigned int type;
799 union {
800 u8 tc;
a1b7c5fd 801 struct tc_cls_u32_offload *cls_u32;
5b33f488 802 struct tc_cls_flower_offload *cls_flower;
b87f7936 803 struct tc_cls_matchall_offload *cls_mall;
332ae8e2 804 struct tc_cls_bpf_offload *cls_bpf;
16e5cc64
JF
805 };
806};
807
a7862b45
BB
808/* These structures hold the attributes of xdp state that are being passed
809 * to the netdevice through the xdp op.
810 */
811enum xdp_netdev_command {
812 /* Set or clear a bpf program used in the earliest stages of packet
813 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
814 * is responsible for calling bpf_prog_put on any old progs that are
815 * stored. In case of error, the callee need not release the new prog
816 * reference, but on success it takes ownership and must bpf_prog_put
817 * when it is no longer used.
818 */
819 XDP_SETUP_PROG,
820 /* Check if a bpf program is set on the device. The callee should
821 * return true if a program is currently attached and running.
822 */
823 XDP_QUERY_PROG,
824};
825
826struct netdev_xdp {
827 enum xdp_netdev_command command;
828 union {
829 /* XDP_SETUP_PROG */
830 struct bpf_prog *prog;
831 /* XDP_QUERY_PROG */
832 bool prog_attached;
833 };
834};
16e5cc64 835
d314774c
SH
836/*
837 * This structure defines the management hooks for network devices.
00829823
SH
838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
d314774c
SH
840 *
841 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
d314774c
SH
846 *
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
850 *
851 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 852 * This function is called when a network device transitions to the up
d314774c
SH
853 * state.
854 *
855 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 856 * This function is called when a network device transitions to the down
d314774c
SH
857 * state.
858 *
dc1f8bf6
SH
859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
00829823 861 * Called when a packet needs to be transmitted.
e79d8429
RR
862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 866 * Required; cannot be NULL.
00829823 867 *
cdba756f
ED
868 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
869 * netdev_features_t features);
870 * Adjusts the requested feature flags according to device-specific
871 * constraints, and returns the resulting flags. Must not modify
872 * the device state.
873 *
f663dd9a 874 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 875 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 876 * Called to decide which queue to use when device supports multiple
00829823
SH
877 * transmit queues.
878 *
d314774c
SH
879 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
880 * This function is called to allow device receiver to make
5e82b4b2 881 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
882 *
883 * void (*ndo_set_rx_mode)(struct net_device *dev);
884 * This function is called device changes address list filtering.
01789349 885 * If driver handles unicast address filtering, it should set
5e82b4b2 886 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
887 *
888 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
889 * This function is called when the Media Access Control address
37b607c5 890 * needs to be changed. If this interface is not defined, the
5e82b4b2 891 * MAC address can not be changed.
d314774c
SH
892 *
893 * int (*ndo_validate_addr)(struct net_device *dev);
894 * Test if Media Access Control address is valid for the device.
895 *
896 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
897 * Called when a user requests an ioctl which can't be handled by
898 * the generic interface code. If not defined ioctls return
d314774c
SH
899 * not supported error code.
900 *
901 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
902 * Used to set network devices bus interface parameters. This interface
5e82b4b2 903 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
904 * interface (PCI) for low level management.
905 *
906 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
907 * Called when a user wants to change the Maximum Transfer Unit
908 * of a device. If not defined, any request to change MTU will
909 * will return an error.
910 *
00829823 911 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 912 * Callback used when the transmitter has not made any progress
d314774c
SH
913 * for dev->watchdog ticks.
914 *
3cfde79c 915 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 916 * struct rtnl_link_stats64 *storage);
d308e38f 917 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 918 * Called when a user wants to get the network device usage
be1f3c2c 919 * statistics. Drivers must do one of the following:
3cfde79c
BH
920 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
921 * rtnl_link_stats64 structure passed by the caller.
82695d9b 922 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
923 * (which should normally be dev->stats) and return a pointer to
924 * it. The structure may be changed asynchronously only if each
925 * field is written atomically.
926 * 3. Update dev->stats asynchronously and atomically, and define
927 * neither operation.
d314774c 928 *
2c9d85d4
NF
929 * bool (*ndo_has_offload_stats)(int attr_id)
930 * Return true if this device supports offload stats of this attr_id.
931 *
932 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
933 * void *attr_data)
934 * Get statistics for offload operations by attr_id. Write it into the
935 * attr_data pointer.
936 *
5d632cb7 937 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 938 * If device supports VLAN filtering this function is called when a
80d5c368 939 * VLAN id is registered.
d314774c 940 *
5d632cb7 941 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 942 * If device supports VLAN filtering this function is called when a
80d5c368 943 * VLAN id is unregistered.
d314774c
SH
944 *
945 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
946 *
947 * SR-IOV management functions.
948 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
949 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
950 * u8 qos, __be16 proto);
ed616689
SC
951 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
952 * int max_tx_rate);
5f8444a3 953 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 954 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
955 * int (*ndo_get_vf_config)(struct net_device *dev,
956 * int vf, struct ifla_vf_info *ivf);
1d8faf48 957 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
958 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
959 * struct nlattr *port[]);
01a3d796
VZ
960 *
961 * Enable or disable the VF ability to query its RSS Redirection Table and
962 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 963 * with PF and querying it may introduce a theoretical security risk.
01a3d796 964 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 965 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
966 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
967 * Called to setup 'tc' number of traffic classes in the net device. This
968 * is always called from the stack with the rtnl lock held and netif tx
969 * queues stopped. This allows the netdevice to perform queue management
970 * safely.
c445477d 971 *
e9bce845
YZ
972 * Fiber Channel over Ethernet (FCoE) offload functions.
973 * int (*ndo_fcoe_enable)(struct net_device *dev);
974 * Called when the FCoE protocol stack wants to start using LLD for FCoE
975 * so the underlying device can perform whatever needed configuration or
976 * initialization to support acceleration of FCoE traffic.
977 *
978 * int (*ndo_fcoe_disable)(struct net_device *dev);
979 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
980 * so the underlying device can perform whatever needed clean-ups to
981 * stop supporting acceleration of FCoE traffic.
982 *
983 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
984 * struct scatterlist *sgl, unsigned int sgc);
985 * Called when the FCoE Initiator wants to initialize an I/O that
986 * is a possible candidate for Direct Data Placement (DDP). The LLD can
987 * perform necessary setup and returns 1 to indicate the device is set up
988 * successfully to perform DDP on this I/O, otherwise this returns 0.
989 *
990 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
991 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
992 * indicated by the FC exchange id 'xid', so the underlying device can
993 * clean up and reuse resources for later DDP requests.
994 *
995 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
996 * struct scatterlist *sgl, unsigned int sgc);
997 * Called when the FCoE Target wants to initialize an I/O that
998 * is a possible candidate for Direct Data Placement (DDP). The LLD can
999 * perform necessary setup and returns 1 to indicate the device is set up
1000 * successfully to perform DDP on this I/O, otherwise this returns 0.
1001 *
68bad94e
NP
1002 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1003 * struct netdev_fcoe_hbainfo *hbainfo);
1004 * Called when the FCoE Protocol stack wants information on the underlying
1005 * device. This information is utilized by the FCoE protocol stack to
1006 * register attributes with Fiber Channel management service as per the
1007 * FC-GS Fabric Device Management Information(FDMI) specification.
1008 *
e9bce845
YZ
1009 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1010 * Called when the underlying device wants to override default World Wide
1011 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1012 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1013 * protocol stack to use.
1014 *
c445477d
BH
1015 * RFS acceleration.
1016 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1017 * u16 rxq_index, u32 flow_id);
1018 * Set hardware filter for RFS. rxq_index is the target queue index;
1019 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1020 * Return the filter ID on success, or a negative error code.
fbaec0ea 1021 *
8b98a70c 1022 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1023 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1024 * Called to make another netdev an underling.
1025 *
1026 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1027 * Called to release previously enslaved netdev.
5455c699
MM
1028 *
1029 * Feature/offload setting functions.
c8f44aff 1030 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1031 * Called to update device configuration to new features. Passed
1032 * feature set might be less than what was returned by ndo_fix_features()).
1033 * Must return >0 or -errno if it changed dev->features itself.
1034 *
edc7d573 1035 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1036 * struct net_device *dev,
f6f6424b 1037 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1038 * Adds an FDB entry to dev for addr.
1690be63
VY
1039 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1040 * struct net_device *dev,
f6f6424b 1041 * const unsigned char *addr, u16 vid)
77162022
JF
1042 * Deletes the FDB entry from dev coresponding to addr.
1043 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1044 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1045 * int *idx)
77162022
JF
1046 * Used to add FDB entries to dump requests. Implementers should add
1047 * entries to skb and update idx with the number of entries.
e5a55a89 1048 *
ad41faa8
ND
1049 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1050 * u16 flags)
e5a55a89 1051 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1052 * struct net_device *dev, u32 filter_mask,
1053 * int nlflags)
ad41faa8
ND
1054 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1055 * u16 flags);
4bf84c35
JP
1056 *
1057 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1058 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1059 * which do not represent real hardware may define this to allow their
1060 * userspace components to manage their virtual carrier state. Devices
1061 * that determine carrier state from physical hardware properties (eg
1062 * network cables) or protocol-dependent mechanisms (eg
1063 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1064 *
1065 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1066 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1067 * Called to get ID of physical port of this device. If driver does
1068 * not implement this, it is assumed that the hw is not able to have
1069 * multiple net devices on single physical port.
53cf5275 1070 *
7c46a640
AD
1071 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1072 * struct udp_tunnel_info *ti);
1073 * Called by UDP tunnel to notify a driver about the UDP port and socket
1074 * address family that a UDP tunnel is listnening to. It is called only
1075 * when a new port starts listening. The operation is protected by the
1076 * RTNL.
1077 *
1078 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1079 * struct udp_tunnel_info *ti);
1080 * Called by UDP tunnel to notify the driver about a UDP port and socket
1081 * address family that the UDP tunnel is not listening to anymore. The
1082 * operation is protected by the RTNL.
1083 *
a6cc0cfa
JF
1084 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1085 * struct net_device *dev)
1086 * Called by upper layer devices to accelerate switching or other
1087 * station functionality into hardware. 'pdev is the lowerdev
1088 * to use for the offload and 'dev' is the net device that will
1089 * back the offload. Returns a pointer to the private structure
1090 * the upper layer will maintain.
1091 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1092 * Called by upper layer device to delete the station created
1093 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1094 * the station and priv is the structure returned by the add
1095 * operation.
1096 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1097 * struct net_device *dev,
1098 * void *priv);
1099 * Callback to use for xmit over the accelerated station. This
1100 * is used in place of ndo_start_xmit on accelerated net
1101 * devices.
5e82b4b2
BH
1102 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1103 * struct net_device *dev
1104 * netdev_features_t features);
04ffcb25 1105 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1106 * performing offload operations on a given packet. This is to give
1107 * the device an opportunity to implement any restrictions that cannot
1108 * be otherwise expressed by feature flags. The check is called with
1109 * the set of features that the stack has calculated and it returns
1110 * those the driver believes to be appropriate.
822b3b2e
JF
1111 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1112 * int queue_index, u32 maxrate);
1113 * Called when a user wants to set a max-rate limitation of specific
1114 * TX queue.
a54acb3a
ND
1115 * int (*ndo_get_iflink)(const struct net_device *dev);
1116 * Called to get the iflink value of this device.
d746d707 1117 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1118 * bool proto_down);
d746d707
AK
1119 * This function is used to pass protocol port error state information
1120 * to the switch driver. The switch driver can react to the proto_down
1121 * by doing a phys down on the associated switch port.
fc4099f1
PS
1122 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1123 * This function is used to get egress tunnel information for given skb.
1124 * This is useful for retrieving outer tunnel header parameters while
1125 * sampling packet.
871b642a
PA
1126 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1127 * This function is used to specify the headroom that the skb must
1128 * consider when allocation skb during packet reception. Setting
1129 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1130 * forward. Setting a negative value resets the rx headroom to the
871b642a 1131 * default value.
a7862b45
BB
1132 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1133 * This function is used to set or query state related to XDP on the
1134 * netdevice. See definition of enum xdp_netdev_command for details.
d746d707 1135 *
d314774c
SH
1136 */
1137struct net_device_ops {
1138 int (*ndo_init)(struct net_device *dev);
1139 void (*ndo_uninit)(struct net_device *dev);
1140 int (*ndo_open)(struct net_device *dev);
1141 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1142 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1143 struct net_device *dev);
1144 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1145 struct net_device *dev,
1146 netdev_features_t features);
00829823 1147 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1148 struct sk_buff *skb,
99932d4f
DB
1149 void *accel_priv,
1150 select_queue_fallback_t fallback);
d314774c
SH
1151 void (*ndo_change_rx_flags)(struct net_device *dev,
1152 int flags);
d314774c 1153 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1154 int (*ndo_set_mac_address)(struct net_device *dev,
1155 void *addr);
d314774c 1156 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1157 int (*ndo_do_ioctl)(struct net_device *dev,
1158 struct ifreq *ifr, int cmd);
d314774c
SH
1159 int (*ndo_set_config)(struct net_device *dev,
1160 struct ifmap *map);
00829823
SH
1161 int (*ndo_change_mtu)(struct net_device *dev,
1162 int new_mtu);
1163 int (*ndo_neigh_setup)(struct net_device *dev,
1164 struct neigh_parms *);
d314774c
SH
1165 void (*ndo_tx_timeout) (struct net_device *dev);
1166
28172739
ED
1167 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1168 struct rtnl_link_stats64 *storage);
2c9d85d4
NF
1169 bool (*ndo_has_offload_stats)(int attr_id);
1170 int (*ndo_get_offload_stats)(int attr_id,
1171 const struct net_device *dev,
1172 void *attr_data);
d314774c
SH
1173 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1174
8e586137 1175 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1176 __be16 proto, u16 vid);
8e586137 1177 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1178 __be16 proto, u16 vid);
d314774c 1179#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1180 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1181 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1182 struct netpoll_info *info);
0e34e931 1183 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1184#endif
e0d1095a 1185#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1186 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1187#endif
95c26df8
WM
1188 int (*ndo_set_vf_mac)(struct net_device *dev,
1189 int queue, u8 *mac);
1190 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1191 int queue, u16 vlan,
1192 u8 qos, __be16 proto);
ed616689
SC
1193 int (*ndo_set_vf_rate)(struct net_device *dev,
1194 int vf, int min_tx_rate,
1195 int max_tx_rate);
5f8444a3
GR
1196 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1197 int vf, bool setting);
dd461d6a
HS
1198 int (*ndo_set_vf_trust)(struct net_device *dev,
1199 int vf, bool setting);
95c26df8
WM
1200 int (*ndo_get_vf_config)(struct net_device *dev,
1201 int vf,
1202 struct ifla_vf_info *ivf);
1d8faf48
RE
1203 int (*ndo_set_vf_link_state)(struct net_device *dev,
1204 int vf, int link_state);
3b766cd8
EBE
1205 int (*ndo_get_vf_stats)(struct net_device *dev,
1206 int vf,
1207 struct ifla_vf_stats
1208 *vf_stats);
57b61080
SF
1209 int (*ndo_set_vf_port)(struct net_device *dev,
1210 int vf,
1211 struct nlattr *port[]);
1212 int (*ndo_get_vf_port)(struct net_device *dev,
1213 int vf, struct sk_buff *skb);
cc8e27cc
EC
1214 int (*ndo_set_vf_guid)(struct net_device *dev,
1215 int vf, u64 guid,
1216 int guid_type);
01a3d796
VZ
1217 int (*ndo_set_vf_rss_query_en)(
1218 struct net_device *dev,
1219 int vf, bool setting);
16e5cc64
JF
1220 int (*ndo_setup_tc)(struct net_device *dev,
1221 u32 handle,
1222 __be16 protocol,
1223 struct tc_to_netdev *tc);
d11ead75 1224#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1225 int (*ndo_fcoe_enable)(struct net_device *dev);
1226 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1227 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1228 u16 xid,
1229 struct scatterlist *sgl,
1230 unsigned int sgc);
1231 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1232 u16 xid);
6247e086
YZ
1233 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1234 u16 xid,
1235 struct scatterlist *sgl,
1236 unsigned int sgc);
68bad94e
NP
1237 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1238 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1239#endif
1240
d11ead75 1241#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1242#define NETDEV_FCOE_WWNN 0
1243#define NETDEV_FCOE_WWPN 1
1244 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1245 u64 *wwn, int type);
4d288d57 1246#endif
3c9c36bc 1247
c445477d
BH
1248#ifdef CONFIG_RFS_ACCEL
1249 int (*ndo_rx_flow_steer)(struct net_device *dev,
1250 const struct sk_buff *skb,
1251 u16 rxq_index,
1252 u32 flow_id);
1253#endif
fbaec0ea
JP
1254 int (*ndo_add_slave)(struct net_device *dev,
1255 struct net_device *slave_dev);
1256 int (*ndo_del_slave)(struct net_device *dev,
1257 struct net_device *slave_dev);
c8f44aff
MM
1258 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1259 netdev_features_t features);
5455c699 1260 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1261 netdev_features_t features);
503eebc2
JP
1262 int (*ndo_neigh_construct)(struct net_device *dev,
1263 struct neighbour *n);
1264 void (*ndo_neigh_destroy)(struct net_device *dev,
1265 struct neighbour *n);
77162022
JF
1266
1267 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1268 struct nlattr *tb[],
77162022 1269 struct net_device *dev,
6b6e2725 1270 const unsigned char *addr,
f6f6424b 1271 u16 vid,
77162022
JF
1272 u16 flags);
1273 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1274 struct nlattr *tb[],
77162022 1275 struct net_device *dev,
f6f6424b
JP
1276 const unsigned char *addr,
1277 u16 vid);
77162022
JF
1278 int (*ndo_fdb_dump)(struct sk_buff *skb,
1279 struct netlink_callback *cb,
1280 struct net_device *dev,
5d5eacb3 1281 struct net_device *filter_dev,
d297653d 1282 int *idx);
e5a55a89
JF
1283
1284 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1285 struct nlmsghdr *nlh,
1286 u16 flags);
e5a55a89
JF
1287 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1288 u32 pid, u32 seq,
6cbdceeb 1289 struct net_device *dev,
46c264da
ND
1290 u32 filter_mask,
1291 int nlflags);
407af329 1292 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1293 struct nlmsghdr *nlh,
1294 u16 flags);
4bf84c35
JP
1295 int (*ndo_change_carrier)(struct net_device *dev,
1296 bool new_carrier);
66b52b0d 1297 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1298 struct netdev_phys_item_id *ppid);
db24a904
DA
1299 int (*ndo_get_phys_port_name)(struct net_device *dev,
1300 char *name, size_t len);
7c46a640
AD
1301 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1302 struct udp_tunnel_info *ti);
1303 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1304 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1305 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1306 struct net_device *dev);
1307 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1308 void *priv);
1309
1310 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1311 struct net_device *dev,
1312 void *priv);
25175ba5 1313 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1314 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1315 int queue_index,
1316 u32 maxrate);
a54acb3a 1317 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1318 int (*ndo_change_proto_down)(struct net_device *dev,
1319 bool proto_down);
fc4099f1
PS
1320 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1321 struct sk_buff *skb);
871b642a
PA
1322 void (*ndo_set_rx_headroom)(struct net_device *dev,
1323 int needed_headroom);
a7862b45
BB
1324 int (*ndo_xdp)(struct net_device *dev,
1325 struct netdev_xdp *xdp);
d314774c
SH
1326};
1327
7aa98047
LR
1328/**
1329 * enum net_device_priv_flags - &struct net_device priv_flags
1330 *
1331 * These are the &struct net_device, they are only set internally
1332 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1333 * userspace; this means that the order of these flags can change
7aa98047
LR
1334 * during any kernel release.
1335 *
1336 * You should have a pretty good reason to be extending these flags.
1337 *
1338 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1339 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1340 * @IFF_BONDING: bonding master or slave
7aa98047 1341 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1342 * @IFF_WAN_HDLC: WAN HDLC device
1343 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1344 * release skb->dst
1345 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1346 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1347 * @IFF_MACVLAN_PORT: device used as macvlan port
1348 * @IFF_BRIDGE_PORT: device used as bridge port
1349 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1350 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1351 * @IFF_UNICAST_FLT: Supports unicast filtering
1352 * @IFF_TEAM_PORT: device used as team port
1353 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1354 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1355 * change when it's running
1356 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1357 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1358 * underlying stacked devices
1359 * @IFF_IPVLAN_MASTER: IPvlan master device
1360 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1361 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1362 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1363 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1364 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1365 * @IFF_TEAM: device is a team device
d4ab4286 1366 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1367 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1368 * entity (i.e. the master device for bridged veth)
3c175784 1369 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1370 */
1371enum netdev_priv_flags {
1372 IFF_802_1Q_VLAN = 1<<0,
1373 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1374 IFF_BONDING = 1<<2,
1375 IFF_ISATAP = 1<<3,
1376 IFF_WAN_HDLC = 1<<4,
1377 IFF_XMIT_DST_RELEASE = 1<<5,
1378 IFF_DONT_BRIDGE = 1<<6,
1379 IFF_DISABLE_NETPOLL = 1<<7,
1380 IFF_MACVLAN_PORT = 1<<8,
1381 IFF_BRIDGE_PORT = 1<<9,
1382 IFF_OVS_DATAPATH = 1<<10,
1383 IFF_TX_SKB_SHARING = 1<<11,
1384 IFF_UNICAST_FLT = 1<<12,
1385 IFF_TEAM_PORT = 1<<13,
1386 IFF_SUPP_NOFCS = 1<<14,
1387 IFF_LIVE_ADDR_CHANGE = 1<<15,
1388 IFF_MACVLAN = 1<<16,
1389 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1390 IFF_IPVLAN_MASTER = 1<<18,
1391 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1392 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1393 IFF_NO_QUEUE = 1<<21,
1394 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1395 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1396 IFF_TEAM = 1<<24,
d4ab4286 1397 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1398 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1399 IFF_MACSEC = 1<<27,
7aa98047
LR
1400};
1401
1402#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1403#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1404#define IFF_BONDING IFF_BONDING
7aa98047 1405#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1406#define IFF_WAN_HDLC IFF_WAN_HDLC
1407#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1408#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1409#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1410#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1411#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1412#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1413#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1414#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1415#define IFF_TEAM_PORT IFF_TEAM_PORT
1416#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1417#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1418#define IFF_MACVLAN IFF_MACVLAN
02875878 1419#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1420#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1421#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1422#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1423#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1424#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1425#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1426#define IFF_TEAM IFF_TEAM
d4ab4286 1427#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1428#define IFF_MACSEC IFF_MACSEC
7aa98047 1429
536721b1
KK
1430/**
1431 * struct net_device - The DEVICE structure.
1432 * Actually, this whole structure is a big mistake. It mixes I/O
1433 * data with strictly "high-level" data, and it has to know about
1434 * almost every data structure used in the INET module.
1435 *
1436 * @name: This is the first field of the "visible" part of this structure
1437 * (i.e. as seen by users in the "Space.c" file). It is the name
1438 * of the interface.
1439 *
1440 * @name_hlist: Device name hash chain, please keep it close to name[]
1441 * @ifalias: SNMP alias
1442 * @mem_end: Shared memory end
1443 * @mem_start: Shared memory start
1444 * @base_addr: Device I/O address
1445 * @irq: Device IRQ number
1446 *
14ffbbb8
TG
1447 * @carrier_changes: Stats to monitor carrier on<->off transitions
1448 *
536721b1
KK
1449 * @state: Generic network queuing layer state, see netdev_state_t
1450 * @dev_list: The global list of network devices
5e82b4b2
BH
1451 * @napi_list: List entry used for polling NAPI devices
1452 * @unreg_list: List entry when we are unregistering the
1453 * device; see the function unregister_netdev
1454 * @close_list: List entry used when we are closing the device
62d885fe
BP
1455 * @ptype_all: Device-specific packet handlers for all protocols
1456 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1457 *
1458 * @adj_list: Directly linked devices, like slaves for bonding
1459 * @all_adj_list: All linked devices, *including* neighbours
1460 * @features: Currently active device features
1461 * @hw_features: User-changeable features
1462 *
1463 * @wanted_features: User-requested features
1464 * @vlan_features: Mask of features inheritable by VLAN devices
1465 *
1466 * @hw_enc_features: Mask of features inherited by encapsulating devices
1467 * This field indicates what encapsulation
1468 * offloads the hardware is capable of doing,
1469 * and drivers will need to set them appropriately.
1470 *
1471 * @mpls_features: Mask of features inheritable by MPLS
1472 *
1473 * @ifindex: interface index
5e82b4b2 1474 * @group: The group the device belongs to
536721b1
KK
1475 *
1476 * @stats: Statistics struct, which was left as a legacy, use
1477 * rtnl_link_stats64 instead
1478 *
1479 * @rx_dropped: Dropped packets by core network,
1480 * do not use this in drivers
1481 * @tx_dropped: Dropped packets by core network,
1482 * do not use this in drivers
6e7333d3
JW
1483 * @rx_nohandler: nohandler dropped packets by core network on
1484 * inactive devices, do not use this in drivers
536721b1 1485 *
536721b1
KK
1486 * @wireless_handlers: List of functions to handle Wireless Extensions,
1487 * instead of ioctl,
1488 * see <net/iw_handler.h> for details.
1489 * @wireless_data: Instance data managed by the core of wireless extensions
1490 *
1491 * @netdev_ops: Includes several pointers to callbacks,
1492 * if one wants to override the ndo_*() functions
1493 * @ethtool_ops: Management operations
f997c55c
AA
1494 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1495 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1496 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1497 * of Layer 2 headers.
1498 *
1499 * @flags: Interface flags (a la BSD)
1500 * @priv_flags: Like 'flags' but invisible to userspace,
1501 * see if.h for the definitions
1502 * @gflags: Global flags ( kept as legacy )
1503 * @padded: How much padding added by alloc_netdev()
1504 * @operstate: RFC2863 operstate
1505 * @link_mode: Mapping policy to operstate
1506 * @if_port: Selectable AUI, TP, ...
1507 * @dma: DMA channel
1508 * @mtu: Interface MTU value
1509 * @type: Interface hardware type
2793a23a 1510 * @hard_header_len: Maximum hardware header length.
536721b1
KK
1511 *
1512 * @needed_headroom: Extra headroom the hardware may need, but not in all
1513 * cases can this be guaranteed
1514 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1515 * cases can this be guaranteed. Some cases also use
1516 * LL_MAX_HEADER instead to allocate the skb
1517 *
1518 * interface address info:
1519 *
1520 * @perm_addr: Permanent hw address
1521 * @addr_assign_type: Hw address assignment type
1522 * @addr_len: Hardware address length
8626a0c8 1523 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1524 * @dev_id: Used to differentiate devices that share
1525 * the same link layer address
1526 * @dev_port: Used to differentiate devices that share
1527 * the same function
1528 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1529 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1530 * has been enabled due to the need to listen to
1531 * additional unicast addresses in a device that
1532 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1533 * @uc: unicast mac addresses
1534 * @mc: multicast mac addresses
1535 * @dev_addrs: list of device hw addresses
1536 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1537 * @promiscuity: Number of times the NIC is told to work in
1538 * promiscuous mode; if it becomes 0 the NIC will
1539 * exit promiscuous mode
536721b1
KK
1540 * @allmulti: Counter, enables or disables allmulticast mode
1541 *
1542 * @vlan_info: VLAN info
1543 * @dsa_ptr: dsa specific data
1544 * @tipc_ptr: TIPC specific data
1545 * @atalk_ptr: AppleTalk link
1546 * @ip_ptr: IPv4 specific data
1547 * @dn_ptr: DECnet specific data
1548 * @ip6_ptr: IPv6 specific data
1549 * @ax25_ptr: AX.25 specific data
1550 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1551 *
1552 * @last_rx: Time of last Rx
1553 * @dev_addr: Hw address (before bcast,
1554 * because most packets are unicast)
1555 *
1556 * @_rx: Array of RX queues
1557 * @num_rx_queues: Number of RX queues
1558 * allocated at register_netdev() time
1559 * @real_num_rx_queues: Number of RX queues currently active in device
1560 *
1561 * @rx_handler: handler for received packets
1562 * @rx_handler_data: XXX: need comments on this one
1563 * @ingress_queue: XXX: need comments on this one
1564 * @broadcast: hw bcast address
1565 *
14ffbbb8
TG
1566 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1567 * indexed by RX queue number. Assigned by driver.
1568 * This must only be set if the ndo_rx_flow_steer
1569 * operation is defined
1570 * @index_hlist: Device index hash chain
1571 *
536721b1
KK
1572 * @_tx: Array of TX queues
1573 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1574 * @real_num_tx_queues: Number of TX queues currently active in device
1575 * @qdisc: Root qdisc from userspace point of view
1576 * @tx_queue_len: Max frames per queue allowed
1577 * @tx_global_lock: XXX: need comments on this one
1578 *
1579 * @xps_maps: XXX: need comments on this one
1580 *
536721b1 1581 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1582 * the watchdog (see dev_watchdog())
536721b1
KK
1583 * @watchdog_timer: List of timers
1584 *
1585 * @pcpu_refcnt: Number of references to this device
1586 * @todo_list: Delayed register/unregister
536721b1
KK
1587 * @link_watch_list: XXX: need comments on this one
1588 *
1589 * @reg_state: Register/unregister state machine
1590 * @dismantle: Device is going to be freed
1591 * @rtnl_link_state: This enum represents the phases of creating
1592 * a new link
1593 *
1594 * @destructor: Called from unregister,
1595 * can be used to call free_netdev
1596 * @npinfo: XXX: need comments on this one
1597 * @nd_net: Network namespace this network device is inside
1598 *
1599 * @ml_priv: Mid-layer private
1600 * @lstats: Loopback statistics
1601 * @tstats: Tunnel statistics
1602 * @dstats: Dummy statistics
1603 * @vstats: Virtual ethernet statistics
1604 *
1605 * @garp_port: GARP
1606 * @mrp_port: MRP
1607 *
1608 * @dev: Class/net/name entry
1609 * @sysfs_groups: Space for optional device, statistics and wireless
1610 * sysfs groups
1611 *
1612 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1613 * @rtnl_link_ops: Rtnl_link_ops
1614 *
1615 * @gso_max_size: Maximum size of generic segmentation offload
1616 * @gso_max_segs: Maximum number of segments that can be passed to the
1617 * NIC for GSO
1618 *
1619 * @dcbnl_ops: Data Center Bridging netlink ops
1620 * @num_tc: Number of traffic classes in the net device
1621 * @tc_to_txq: XXX: need comments on this one
1622 * @prio_tc_map XXX: need comments on this one
1623 *
1624 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1625 *
1626 * @priomap: XXX: need comments on this one
1627 * @phydev: Physical device may attach itself
1628 * for hardware timestamping
1629 *
123b3652
ED
1630 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1631 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1632 *
d746d707
AK
1633 * @proto_down: protocol port state information can be sent to the
1634 * switch driver and used to set the phys state of the
1635 * switch port.
1636 *
1da177e4
LT
1637 * FIXME: cleanup struct net_device such that network protocol info
1638 * moves out.
1639 */
1640
d94d9fee 1641struct net_device {
1da177e4 1642 char name[IFNAMSIZ];
9356b8fc 1643 struct hlist_node name_hlist;
0b815a1a 1644 char *ifalias;
1da177e4
LT
1645 /*
1646 * I/O specific fields
1647 * FIXME: Merge these and struct ifmap into one
1648 */
536721b1
KK
1649 unsigned long mem_end;
1650 unsigned long mem_start;
1651 unsigned long base_addr;
1652 int irq;
1da177e4 1653
14ffbbb8
TG
1654 atomic_t carrier_changes;
1655
1da177e4 1656 /*
536721b1
KK
1657 * Some hardware also needs these fields (state,dev_list,
1658 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1659 * part of the usual set specified in Space.c.
1660 */
1661
1da177e4
LT
1662 unsigned long state;
1663
7562f876 1664 struct list_head dev_list;
bea3348e 1665 struct list_head napi_list;
44a0873d 1666 struct list_head unreg_list;
5cde2829 1667 struct list_head close_list;
7866a621
SN
1668 struct list_head ptype_all;
1669 struct list_head ptype_specific;
2f268f12 1670
2f268f12
VF
1671 struct {
1672 struct list_head upper;
1673 struct list_head lower;
1674 } adj_list;
1675
2f268f12
VF
1676 struct {
1677 struct list_head upper;
1678 struct list_head lower;
1679 } all_adj_list;
4c3d5e7b 1680
c8f44aff 1681 netdev_features_t features;
c8f44aff 1682 netdev_features_t hw_features;
c8f44aff 1683 netdev_features_t wanted_features;
c8f44aff 1684 netdev_features_t vlan_features;
6a674e9c 1685 netdev_features_t hw_enc_features;
0d89d203 1686 netdev_features_t mpls_features;
802ab55a 1687 netdev_features_t gso_partial_features;
04ed3e74 1688
1da177e4 1689 int ifindex;
7a66bbc9 1690 int group;
1da177e4 1691
c45d286e 1692 struct net_device_stats stats;
015f0688 1693
015f0688
ED
1694 atomic_long_t rx_dropped;
1695 atomic_long_t tx_dropped;
6e7333d3 1696 atomic_long_t rx_nohandler;
1da177e4 1697
b86e0280 1698#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1699 const struct iw_handler_def *wireless_handlers;
1700 struct iw_public_data *wireless_data;
b86e0280 1701#endif
d314774c 1702 const struct net_device_ops *netdev_ops;
76fd8593 1703 const struct ethtool_ops *ethtool_ops;
4170604f 1704#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1705 const struct switchdev_ops *switchdev_ops;
4170604f 1706#endif
1b69c6d0
DA
1707#ifdef CONFIG_NET_L3_MASTER_DEV
1708 const struct l3mdev_ops *l3mdev_ops;
1709#endif
f997c55c
AA
1710#if IS_ENABLED(CONFIG_IPV6)
1711 const struct ndisc_ops *ndisc_ops;
1712#endif
1da177e4 1713
3b04ddde
SH
1714 const struct header_ops *header_ops;
1715
536721b1
KK
1716 unsigned int flags;
1717 unsigned int priv_flags;
1718
1da177e4 1719 unsigned short gflags;
536721b1 1720 unsigned short padded;
1da177e4 1721
536721b1
KK
1722 unsigned char operstate;
1723 unsigned char link_mode;
b00055aa 1724
536721b1
KK
1725 unsigned char if_port;
1726 unsigned char dma;
bdc220da 1727
536721b1
KK
1728 unsigned int mtu;
1729 unsigned short type;
1730 unsigned short hard_header_len;
1da177e4 1731
f5184d26
JB
1732 unsigned short needed_headroom;
1733 unsigned short needed_tailroom;
1734
1da177e4 1735 /* Interface address info. */
536721b1
KK
1736 unsigned char perm_addr[MAX_ADDR_LEN];
1737 unsigned char addr_assign_type;
1738 unsigned char addr_len;
a0a9663d 1739 unsigned short neigh_priv_len;
536721b1
KK
1740 unsigned short dev_id;
1741 unsigned short dev_port;
ccffad25 1742 spinlock_t addr_list_lock;
14ffbbb8
TG
1743 unsigned char name_assign_type;
1744 bool uc_promisc;
536721b1
KK
1745 struct netdev_hw_addr_list uc;
1746 struct netdev_hw_addr_list mc;
1747 struct netdev_hw_addr_list dev_addrs;
1748
4c3d5e7b
ED
1749#ifdef CONFIG_SYSFS
1750 struct kset *queues_kset;
1751#endif
9d45abe1
WC
1752 unsigned int promiscuity;
1753 unsigned int allmulti;
1da177e4 1754
1da177e4 1755
5e82b4b2 1756 /* Protocol-specific pointers */
65ac6a5f 1757
d11ead75 1758#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1759 struct vlan_info __rcu *vlan_info;
65ac6a5f 1760#endif
34a430d7 1761#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1762 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1763#endif
1764#if IS_ENABLED(CONFIG_TIPC)
536721b1 1765 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1766#endif
536721b1
KK
1767 void *atalk_ptr;
1768 struct in_device __rcu *ip_ptr;
1769 struct dn_dev __rcu *dn_ptr;
1770 struct inet6_dev __rcu *ip6_ptr;
1771 void *ax25_ptr;
1772 struct wireless_dev *ieee80211_ptr;
98a18b6f 1773 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1774#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1775 struct mpls_dev __rcu *mpls_ptr;
1776#endif
1da177e4 1777
9356b8fc 1778/*
cd13539b 1779 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1780 */
536721b1 1781 unsigned long last_rx;
4dc89133 1782
9356b8fc 1783 /* Interface address info used in eth_type_trans() */
536721b1 1784 unsigned char *dev_addr;
f001fde5 1785
a953be53 1786#ifdef CONFIG_SYSFS
0a9627f2
TH
1787 struct netdev_rx_queue *_rx;
1788
0a9627f2 1789 unsigned int num_rx_queues;
62fe0b40 1790 unsigned int real_num_rx_queues;
df334545 1791#endif
0a9627f2 1792
3b47d303 1793 unsigned long gro_flush_timeout;
61391cde 1794 rx_handler_func_t __rcu *rx_handler;
1795 void __rcu *rx_handler_data;
e8a0464c 1796
4cda01e8 1797#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1798 struct tcf_proto __rcu *ingress_cl_list;
1799#endif
24824a09 1800 struct netdev_queue __rcu *ingress_queue;
e687ad60 1801#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1802 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1803#endif
d2788d34 1804
536721b1 1805 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1806#ifdef CONFIG_RFS_ACCEL
1807 struct cpu_rmap *rx_cpu_rmap;
1808#endif
1809 struct hlist_node index_hlist;
cd13539b
ED
1810
1811/*
1812 * Cache lines mostly used on transmit path
1813 */
e8a0464c
DM
1814 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1815 unsigned int num_tx_queues;
fd2ea0a7 1816 unsigned int real_num_tx_queues;
af356afa 1817 struct Qdisc *qdisc;
59cc1f61
JK
1818#ifdef CONFIG_NET_SCHED
1819 DECLARE_HASHTABLE (qdisc_hash, 4);
1820#endif
536721b1 1821 unsigned long tx_queue_len;
c3f26a26 1822 spinlock_t tx_global_lock;
14ffbbb8 1823 int watchdog_timeo;
cd13539b 1824
bf264145 1825#ifdef CONFIG_XPS
a4177869 1826 struct xps_dev_maps __rcu *xps_maps;
bf264145 1827#endif
1f211a1b
DB
1828#ifdef CONFIG_NET_CLS_ACT
1829 struct tcf_proto __rcu *egress_cl_list;
1830#endif
0c4f691f 1831
9356b8fc 1832 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1833 struct timer_list watchdog_timer;
1834
29b4433d 1835 int __percpu *pcpu_refcnt;
1da177e4 1836 struct list_head todo_list;
1da177e4 1837
e014debe 1838 struct list_head link_watch_list;
572a103d 1839
1da177e4 1840 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1841 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1842 NETREG_UNREGISTERING, /* called unregister_netdevice */
1843 NETREG_UNREGISTERED, /* completed unregister todo */
1844 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1845 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1846 } reg_state:8;
1847
536721b1 1848 bool dismantle;
a2835763
PM
1849
1850 enum {
1851 RTNL_LINK_INITIALIZED,
1852 RTNL_LINK_INITIALIZING,
1853 } rtnl_link_state:16;
1da177e4 1854
d314774c 1855 void (*destructor)(struct net_device *dev);
1da177e4 1856
1da177e4 1857#ifdef CONFIG_NETPOLL
5fbee843 1858 struct netpoll_info __rcu *npinfo;
1da177e4 1859#endif
eae792b7 1860
0c5c9fb5 1861 possible_net_t nd_net;
4a1c5371 1862
4951704b 1863 /* mid-layer private */
a7855c78 1864 union {
536721b1
KK
1865 void *ml_priv;
1866 struct pcpu_lstats __percpu *lstats;
8f84985f 1867 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1868 struct pcpu_dstats __percpu *dstats;
1869 struct pcpu_vstats __percpu *vstats;
a7855c78 1870 };
536721b1 1871
3cc77ec7 1872 struct garp_port __rcu *garp_port;
febf018d 1873 struct mrp_port __rcu *mrp_port;
1da177e4 1874
5e82b4b2 1875 struct device dev;
0c509a6c 1876 const struct attribute_group *sysfs_groups[4];
a953be53 1877 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1878
38f7b870 1879 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1880
82cc1a7a
PWJ
1881 /* for setting kernel sock attribute on TCP connection setup */
1882#define GSO_MAX_SIZE 65536
1883 unsigned int gso_max_size;
30b678d8
BH
1884#define GSO_MAX_SEGS 65535
1885 u16 gso_max_segs;
743b03a8 1886
7a6b6f51 1887#ifdef CONFIG_DCB
32953543 1888 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1889#endif
5e82b4b2
BH
1890 u8 num_tc;
1891 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1892 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1893
d11ead75 1894#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1895 unsigned int fcoe_ddp_xid;
5bc1421e 1896#endif
86f8515f 1897#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1898 struct netprio_map __rcu *priomap;
4d288d57 1899#endif
5e82b4b2
BH
1900 struct phy_device *phydev;
1901 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1902 struct lock_class_key *qdisc_running_key;
5e82b4b2 1903 bool proto_down;
1da177e4 1904};
43cb76d9 1905#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1906
1907#define NETDEV_ALIGN 32
1da177e4 1908
4f57c087
JF
1909static inline
1910int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1911{
1912 return dev->prio_tc_map[prio & TC_BITMASK];
1913}
1914
1915static inline
1916int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1917{
1918 if (tc >= dev->num_tc)
1919 return -EINVAL;
1920
1921 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1922 return 0;
1923}
1924
1925static inline
1926void netdev_reset_tc(struct net_device *dev)
1927{
1928 dev->num_tc = 0;
1929 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1930 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1931}
1932
1933static inline
1934int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1935{
1936 if (tc >= dev->num_tc)
1937 return -EINVAL;
1938
1939 dev->tc_to_txq[tc].count = count;
1940 dev->tc_to_txq[tc].offset = offset;
1941 return 0;
1942}
1943
1944static inline
1945int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1946{
1947 if (num_tc > TC_MAX_QUEUE)
1948 return -EINVAL;
1949
1950 dev->num_tc = num_tc;
1951 return 0;
1952}
1953
1954static inline
1955int netdev_get_num_tc(struct net_device *dev)
1956{
1957 return dev->num_tc;
1958}
1959
e8a0464c
DM
1960static inline
1961struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1962 unsigned int index)
1963{
1964 return &dev->_tx[index];
1965}
1966
10c51b56
DB
1967static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1968 const struct sk_buff *skb)
1969{
1970 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1971}
1972
e8a0464c
DM
1973static inline void netdev_for_each_tx_queue(struct net_device *dev,
1974 void (*f)(struct net_device *,
1975 struct netdev_queue *,
1976 void *),
1977 void *arg)
1978{
1979 unsigned int i;
1980
1981 for (i = 0; i < dev->num_tx_queues; i++)
1982 f(dev, &dev->_tx[i], arg);
1983}
1984
d3fff6c4
ED
1985#define netdev_lockdep_set_classes(dev) \
1986{ \
1987 static struct lock_class_key qdisc_tx_busylock_key; \
1988 static struct lock_class_key qdisc_running_key; \
1989 static struct lock_class_key qdisc_xmit_lock_key; \
1990 static struct lock_class_key dev_addr_list_lock_key; \
1991 unsigned int i; \
1992 \
1993 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1994 (dev)->qdisc_running_key = &qdisc_running_key; \
1995 lockdep_set_class(&(dev)->addr_list_lock, \
1996 &dev_addr_list_lock_key); \
1997 for (i = 0; i < (dev)->num_tx_queues; i++) \
1998 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1999 &qdisc_xmit_lock_key); \
2000}
2001
f629d208 2002struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
2003 struct sk_buff *skb,
2004 void *accel_priv);
8c4c49df 2005
871b642a
PA
2006/* returns the headroom that the master device needs to take in account
2007 * when forwarding to this dev
2008 */
2009static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2010{
2011 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2012}
2013
2014static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2015{
2016 if (dev->netdev_ops->ndo_set_rx_headroom)
2017 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2018}
2019
2020/* set the device rx headroom to the dev's default */
2021static inline void netdev_reset_rx_headroom(struct net_device *dev)
2022{
2023 netdev_set_rx_headroom(dev, -1);
2024}
2025
c346dca1
YH
2026/*
2027 * Net namespace inlines
2028 */
2029static inline
2030struct net *dev_net(const struct net_device *dev)
2031{
c2d9ba9b 2032 return read_pnet(&dev->nd_net);
c346dca1
YH
2033}
2034
2035static inline
f5aa23fd 2036void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2037{
0c5c9fb5 2038 write_pnet(&dev->nd_net, net);
c346dca1
YH
2039}
2040
3e8a72d1 2041static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 2042{
3fc88677 2043#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
2044 if (dev->dsa_ptr != NULL)
2045 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 2046#endif
5aed85ce 2047 return false;
396138f0
LB
2048}
2049
bea3348e
SH
2050/**
2051 * netdev_priv - access network device private data
2052 * @dev: network device
2053 *
2054 * Get network device private data
2055 */
6472ce60 2056static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2057{
1ce8e7b5 2058 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2059}
2060
1da177e4
LT
2061/* Set the sysfs physical device reference for the network logical device
2062 * if set prior to registration will cause a symlink during initialization.
2063 */
43cb76d9 2064#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2065
384912ed 2066/* Set the sysfs device type for the network logical device to allow
3f79410c 2067 * fine-grained identification of different network device types. For
5e82b4b2 2068 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2069 */
2070#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2071
82dc3c63
ED
2072/* Default NAPI poll() weight
2073 * Device drivers are strongly advised to not use bigger value
2074 */
2075#define NAPI_POLL_WEIGHT 64
2076
3b582cc1 2077/**
5e82b4b2 2078 * netif_napi_add - initialize a NAPI context
3b582cc1 2079 * @dev: network device
5e82b4b2 2080 * @napi: NAPI context
3b582cc1
SH
2081 * @poll: polling function
2082 * @weight: default weight
2083 *
5e82b4b2
BH
2084 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2085 * *any* of the other NAPI-related functions.
3b582cc1 2086 */
d565b0a1
HX
2087void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2088 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2089
d64b5e85 2090/**
5e82b4b2 2091 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2092 * @dev: network device
5e82b4b2 2093 * @napi: NAPI context
d64b5e85
ED
2094 * @poll: polling function
2095 * @weight: default weight
2096 *
2097 * This variant of netif_napi_add() should be used from drivers using NAPI
2098 * to exclusively poll a TX queue.
2099 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2100 */
2101static inline void netif_tx_napi_add(struct net_device *dev,
2102 struct napi_struct *napi,
2103 int (*poll)(struct napi_struct *, int),
2104 int weight)
2105{
2106 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2107 netif_napi_add(dev, napi, poll, weight);
2108}
2109
d8156534 2110/**
5e82b4b2
BH
2111 * netif_napi_del - remove a NAPI context
2112 * @napi: NAPI context
d8156534 2113 *
5e82b4b2 2114 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2115 */
d565b0a1
HX
2116void netif_napi_del(struct napi_struct *napi);
2117
2118struct napi_gro_cb {
78a478d0 2119 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2120 void *frag0;
78a478d0 2121
7489594c
HX
2122 /* Length of frag0. */
2123 unsigned int frag0_len;
2124
86911732 2125 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2126 int data_offset;
86911732 2127
d565b0a1 2128 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2129 u16 flush;
2130
2131 /* Save the IP ID here and check when we get to the transport layer */
2132 u16 flush_id;
d565b0a1
HX
2133
2134 /* Number of segments aggregated. */
2e71a6f8
ED
2135 u16 count;
2136
15e2396d
TH
2137 /* Start offset for remote checksum offload */
2138 u16 gro_remcsum_start;
2139
2e71a6f8
ED
2140 /* jiffies when first packet was created/queued */
2141 unsigned long age;
86347245 2142
afe93325 2143 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2144 u16 proto;
2145
baa32ff4
TH
2146 /* This is non-zero if the packet may be of the same flow. */
2147 u8 same_flow:1;
2148
fac8e0f5
JG
2149 /* Used in tunnel GRO receive */
2150 u8 encap_mark:1;
573e8fca
TH
2151
2152 /* GRO checksum is valid */
2153 u8 csum_valid:1;
2154
662880f4
TH
2155 /* Number of checksums via CHECKSUM_UNNECESSARY */
2156 u8 csum_cnt:3;
c3c7c254 2157
baa32ff4
TH
2158 /* Free the skb? */
2159 u8 free:2;
2160#define NAPI_GRO_FREE 1
2161#define NAPI_GRO_FREE_STOLEN_HEAD 2
2162
efc98d08
TH
2163 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2164 u8 is_ipv6:1;
2165
a0ca153f
AD
2166 /* Used in GRE, set in fou/gue_gro_receive */
2167 u8 is_fou:1;
2168
1530545e
AD
2169 /* Used to determine if flush_id can be ignored */
2170 u8 is_atomic:1;
2171
fcd91dd4
SD
2172 /* Number of gro_receive callbacks this packet already went through */
2173 u8 recursion_counter:4;
2174
2175 /* 1 bit hole */
baa32ff4 2176
bf5a755f
JC
2177 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2178 __wsum csum;
2179
c3c7c254
ED
2180 /* used in skb_gro_receive() slow path */
2181 struct sk_buff *last;
d565b0a1
HX
2182};
2183
2184#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2185
fcd91dd4
SD
2186#define GRO_RECURSION_LIMIT 15
2187static inline int gro_recursion_inc_test(struct sk_buff *skb)
2188{
2189 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2190}
2191
2192typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2193static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2194 struct sk_buff **head,
2195 struct sk_buff *skb)
2196{
2197 if (unlikely(gro_recursion_inc_test(skb))) {
2198 NAPI_GRO_CB(skb)->flush |= 1;
2199 return NULL;
2200 }
2201
2202 return cb(head, skb);
2203}
2204
2205typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2206 struct sk_buff *);
2207static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2208 struct sock *sk,
2209 struct sk_buff **head,
2210 struct sk_buff *skb)
2211{
2212 if (unlikely(gro_recursion_inc_test(skb))) {
2213 NAPI_GRO_CB(skb)->flush |= 1;
2214 return NULL;
2215 }
2216
2217 return cb(sk, head, skb);
2218}
2219
1da177e4 2220struct packet_type {
f2ccd8fa
DM
2221 __be16 type; /* This is really htons(ether_type). */
2222 struct net_device *dev; /* NULL is wildcarded here */
2223 int (*func) (struct sk_buff *,
2224 struct net_device *,
2225 struct packet_type *,
2226 struct net_device *);
c0de08d0
EL
2227 bool (*id_match)(struct packet_type *ptype,
2228 struct sock *sk);
1da177e4
LT
2229 void *af_packet_priv;
2230 struct list_head list;
2231};
2232
f191a1d1 2233struct offload_callbacks {
576a30eb 2234 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2235 netdev_features_t features);
d565b0a1 2236 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2237 struct sk_buff *skb);
299603e8 2238 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2239};
2240
2241struct packet_offload {
2242 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2243 u16 priority;
f191a1d1
VY
2244 struct offload_callbacks callbacks;
2245 struct list_head list;
1da177e4
LT
2246};
2247
5e82b4b2 2248/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2249struct pcpu_sw_netstats {
2250 u64 rx_packets;
2251 u64 rx_bytes;
2252 u64 tx_packets;
2253 u64 tx_bytes;
2254 struct u64_stats_sync syncp;
2255};
2256
aabc92bb
PNA
2257#define __netdev_alloc_pcpu_stats(type, gfp) \
2258({ \
2259 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2260 if (pcpu_stats) { \
2261 int __cpu; \
2262 for_each_possible_cpu(__cpu) { \
2263 typeof(type) *stat; \
2264 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2265 u64_stats_init(&stat->syncp); \
2266 } \
2267 } \
2268 pcpu_stats; \
1c213bd2
WC
2269})
2270
aabc92bb 2271#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2272 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2273
764f5e54
JP
2274enum netdev_lag_tx_type {
2275 NETDEV_LAG_TX_TYPE_UNKNOWN,
2276 NETDEV_LAG_TX_TYPE_RANDOM,
2277 NETDEV_LAG_TX_TYPE_BROADCAST,
2278 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2279 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2280 NETDEV_LAG_TX_TYPE_HASH,
2281};
2282
2283struct netdev_lag_upper_info {
2284 enum netdev_lag_tx_type tx_type;
2285};
2286
fb1b2e3c
JP
2287struct netdev_lag_lower_state_info {
2288 u8 link_up : 1,
2289 tx_enabled : 1;
2290};
2291
1da177e4
LT
2292#include <linux/notifier.h>
2293
dcfe1421
AW
2294/* netdevice notifier chain. Please remember to update the rtnetlink
2295 * notification exclusion list in rtnetlink_event() when adding new
2296 * types.
2297 */
2298#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2299#define NETDEV_DOWN 0x0002
2300#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2301 detected a hardware crash and restarted
2302 - we can use this eg to kick tcp sessions
2303 once done */
2304#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2305#define NETDEV_REGISTER 0x0005
2306#define NETDEV_UNREGISTER 0x0006
1d486bfb 2307#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2308#define NETDEV_CHANGEADDR 0x0008
2309#define NETDEV_GOING_DOWN 0x0009
2310#define NETDEV_CHANGENAME 0x000A
2311#define NETDEV_FEAT_CHANGE 0x000B
2312#define NETDEV_BONDING_FAILOVER 0x000C
2313#define NETDEV_PRE_UP 0x000D
2314#define NETDEV_PRE_TYPE_CHANGE 0x000E
2315#define NETDEV_POST_TYPE_CHANGE 0x000F
2316#define NETDEV_POST_INIT 0x0010
0115e8e3 2317#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2318#define NETDEV_RELEASE 0x0012
2319#define NETDEV_NOTIFY_PEERS 0x0013
2320#define NETDEV_JOIN 0x0014
42e52bf9 2321#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2322#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2323#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2324#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2325#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2326#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2327#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2328#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2329#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2330
f629d208
JP
2331int register_netdevice_notifier(struct notifier_block *nb);
2332int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2333
2334struct netdev_notifier_info {
2335 struct net_device *dev;
2336};
2337
be9efd36
JP
2338struct netdev_notifier_change_info {
2339 struct netdev_notifier_info info; /* must be first */
2340 unsigned int flags_changed;
2341};
2342
0e4ead9d
JP
2343struct netdev_notifier_changeupper_info {
2344 struct netdev_notifier_info info; /* must be first */
2345 struct net_device *upper_dev; /* new upper dev */
2346 bool master; /* is upper dev master */
5e82b4b2 2347 bool linking; /* is the notification for link or unlink */
29bf24af 2348 void *upper_info; /* upper dev info */
0e4ead9d
JP
2349};
2350
04d48266
JP
2351struct netdev_notifier_changelowerstate_info {
2352 struct netdev_notifier_info info; /* must be first */
2353 void *lower_state_info; /* is lower dev state */
2354};
2355
75538c2b
CW
2356static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2357 struct net_device *dev)
2358{
2359 info->dev = dev;
2360}
2361
351638e7
JP
2362static inline struct net_device *
2363netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2364{
2365 return info->dev;
2366}
2367
f629d208 2368int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2369
2370
1da177e4
LT
2371extern rwlock_t dev_base_lock; /* Device list lock */
2372
881d966b
EB
2373#define for_each_netdev(net, d) \
2374 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2375#define for_each_netdev_reverse(net, d) \
2376 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2377#define for_each_netdev_rcu(net, d) \
2378 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2379#define for_each_netdev_safe(net, d, n) \
2380 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2381#define for_each_netdev_continue(net, d) \
2382 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2383#define for_each_netdev_continue_rcu(net, d) \
2384 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2385#define for_each_netdev_in_bond_rcu(bond, slave) \
2386 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2387 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2388#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2389
a050c33f
DL
2390static inline struct net_device *next_net_device(struct net_device *dev)
2391{
2392 struct list_head *lh;
2393 struct net *net;
2394
c346dca1 2395 net = dev_net(dev);
a050c33f
DL
2396 lh = dev->dev_list.next;
2397 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2398}
2399
ce81b76a
ED
2400static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2401{
2402 struct list_head *lh;
2403 struct net *net;
2404
2405 net = dev_net(dev);
ccf43438 2406 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2407 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2408}
2409
a050c33f
DL
2410static inline struct net_device *first_net_device(struct net *net)
2411{
2412 return list_empty(&net->dev_base_head) ? NULL :
2413 net_device_entry(net->dev_base_head.next);
2414}
7562f876 2415
ccf43438
ED
2416static inline struct net_device *first_net_device_rcu(struct net *net)
2417{
2418 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2419
2420 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2421}
2422
f629d208
JP
2423int netdev_boot_setup_check(struct net_device *dev);
2424unsigned long netdev_boot_base(const char *prefix, int unit);
2425struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2426 const char *hwaddr);
2427struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2428struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2429void dev_add_pack(struct packet_type *pt);
2430void dev_remove_pack(struct packet_type *pt);
2431void __dev_remove_pack(struct packet_type *pt);
2432void dev_add_offload(struct packet_offload *po);
2433void dev_remove_offload(struct packet_offload *po);
f629d208 2434
a54acb3a 2435int dev_get_iflink(const struct net_device *dev);
fc4099f1 2436int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2437struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2438 unsigned short mask);
f629d208
JP
2439struct net_device *dev_get_by_name(struct net *net, const char *name);
2440struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2441struct net_device *__dev_get_by_name(struct net *net, const char *name);
2442int dev_alloc_name(struct net_device *dev, const char *name);
2443int dev_open(struct net_device *dev);
2444int dev_close(struct net_device *dev);
99c4a26a 2445int dev_close_many(struct list_head *head, bool unlink);
f629d208 2446void dev_disable_lro(struct net_device *dev);
0c4b51f0 2447int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2448int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2449int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2450int register_netdevice(struct net_device *dev);
2451void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2452void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2453static inline void unregister_netdevice(struct net_device *dev)
2454{
2455 unregister_netdevice_queue(dev, NULL);
2456}
2457
f629d208
JP
2458int netdev_refcnt_read(const struct net_device *dev);
2459void free_netdev(struct net_device *dev);
74d332c1 2460void netdev_freemem(struct net_device *dev);
f629d208
JP
2461void synchronize_net(void);
2462int init_dummy_netdev(struct net_device *dev);
937f1ba5 2463
f60e5990 2464DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2465#define XMIT_RECURSION_LIMIT 10
2466
f60e5990 2467static inline int dev_recursion_level(void)
2468{
2469 return this_cpu_read(xmit_recursion);
2470}
2471
f629d208
JP
2472struct net_device *dev_get_by_index(struct net *net, int ifindex);
2473struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2474struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2475int netdev_get_name(struct net *net, char *name, int ifindex);
2476int dev_restart(struct net_device *dev);
f629d208 2477int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2478
2479static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2480{
2481 return NAPI_GRO_CB(skb)->data_offset;
2482}
2483
2484static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2485{
2486 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2487}
2488
2489static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2490{
2491 NAPI_GRO_CB(skb)->data_offset += len;
2492}
2493
a5b1cf28
HX
2494static inline void *skb_gro_header_fast(struct sk_buff *skb,
2495 unsigned int offset)
86911732 2496{
a5b1cf28
HX
2497 return NAPI_GRO_CB(skb)->frag0 + offset;
2498}
78a478d0 2499
a5b1cf28
HX
2500static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2501{
2502 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2503}
78a478d0 2504
a5b1cf28
HX
2505static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2506 unsigned int offset)
2507{
17dd759c
HX
2508 if (!pskb_may_pull(skb, hlen))
2509 return NULL;
2510
a5b1cf28
HX
2511 NAPI_GRO_CB(skb)->frag0 = NULL;
2512 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2513 return skb->data + offset;
86911732 2514}
1da177e4 2515
36e7b1b8
HX
2516static inline void *skb_gro_network_header(struct sk_buff *skb)
2517{
78d3fd0b
HX
2518 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2519 skb_network_offset(skb);
36e7b1b8
HX
2520}
2521
bf5a755f
JC
2522static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2523 const void *start, unsigned int len)
2524{
573e8fca 2525 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2526 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2527 csum_partial(start, len, 0));
2528}
2529
573e8fca
TH
2530/* GRO checksum functions. These are logical equivalents of the normal
2531 * checksum functions (in skbuff.h) except that they operate on the GRO
2532 * offsets and fields in sk_buff.
2533 */
2534
2535__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2536
15e2396d
TH
2537static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2538{
b7fe10e5 2539 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2540}
2541
573e8fca
TH
2542static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2543 bool zero_okay,
2544 __sum16 check)
2545{
6edec0e6
TH
2546 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2547 skb_checksum_start_offset(skb) <
2548 skb_gro_offset(skb)) &&
15e2396d 2549 !skb_at_gro_remcsum_start(skb) &&
662880f4 2550 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2551 (!zero_okay || check));
2552}
2553
2554static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2555 __wsum psum)
2556{
2557 if (NAPI_GRO_CB(skb)->csum_valid &&
2558 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2559 return 0;
2560
2561 NAPI_GRO_CB(skb)->csum = psum;
2562
2563 return __skb_gro_checksum_complete(skb);
2564}
2565
573e8fca
TH
2566static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2567{
662880f4
TH
2568 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2569 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2570 NAPI_GRO_CB(skb)->csum_cnt--;
2571 } else {
2572 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2573 * verified a new top level checksum or an encapsulated one
2574 * during GRO. This saves work if we fallback to normal path.
2575 */
2576 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2577 }
2578}
2579
2580#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2581 compute_pseudo) \
2582({ \
2583 __sum16 __ret = 0; \
2584 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2585 __ret = __skb_gro_checksum_validate_complete(skb, \
2586 compute_pseudo(skb, proto)); \
5a212329
TH
2587 if (__ret) \
2588 __skb_mark_checksum_bad(skb); \
2589 else \
573e8fca
TH
2590 skb_gro_incr_csum_unnecessary(skb); \
2591 __ret; \
2592})
2593
2594#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2595 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2596
2597#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2598 compute_pseudo) \
2599 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2600
2601#define skb_gro_checksum_simple_validate(skb) \
2602 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2603
d96535a1
TH
2604static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2605{
2606 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2607 !NAPI_GRO_CB(skb)->csum_valid);
2608}
2609
2610static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2611 __sum16 check, __wsum pseudo)
2612{
2613 NAPI_GRO_CB(skb)->csum = ~pseudo;
2614 NAPI_GRO_CB(skb)->csum_valid = 1;
2615}
2616
2617#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2618do { \
2619 if (__skb_gro_checksum_convert_check(skb)) \
2620 __skb_gro_checksum_convert(skb, check, \
2621 compute_pseudo(skb, proto)); \
2622} while (0)
2623
26c4f7da
TH
2624struct gro_remcsum {
2625 int offset;
2626 __wsum delta;
2627};
2628
2629static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2630{
846cd667 2631 grc->offset = 0;
26c4f7da
TH
2632 grc->delta = 0;
2633}
2634
b7fe10e5
TH
2635static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2636 unsigned int off, size_t hdrlen,
2637 int start, int offset,
2638 struct gro_remcsum *grc,
2639 bool nopartial)
dcdc8994
TH
2640{
2641 __wsum delta;
b7fe10e5 2642 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2643
2644 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2645
15e2396d 2646 if (!nopartial) {
b7fe10e5
TH
2647 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2648 return ptr;
2649 }
2650
2651 ptr = skb_gro_header_fast(skb, off);
2652 if (skb_gro_header_hard(skb, off + plen)) {
2653 ptr = skb_gro_header_slow(skb, off + plen, off);
2654 if (!ptr)
2655 return NULL;
15e2396d
TH
2656 }
2657
b7fe10e5
TH
2658 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2659 start, offset);
dcdc8994
TH
2660
2661 /* Adjust skb->csum since we changed the packet */
dcdc8994 2662 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2663
b7fe10e5 2664 grc->offset = off + hdrlen + offset;
26c4f7da 2665 grc->delta = delta;
b7fe10e5
TH
2666
2667 return ptr;
dcdc8994
TH
2668}
2669
26c4f7da
TH
2670static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2671 struct gro_remcsum *grc)
2672{
b7fe10e5
TH
2673 void *ptr;
2674 size_t plen = grc->offset + sizeof(u16);
2675
26c4f7da
TH
2676 if (!grc->delta)
2677 return;
2678
b7fe10e5
TH
2679 ptr = skb_gro_header_fast(skb, grc->offset);
2680 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2681 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2682 if (!ptr)
2683 return;
2684 }
2685
2686 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2687}
dcdc8994 2688
6ae23ad3
TH
2689struct skb_csum_offl_spec {
2690 __u16 ipv4_okay:1,
2691 ipv6_okay:1,
2692 encap_okay:1,
2693 ip_options_okay:1,
2694 ext_hdrs_okay:1,
2695 tcp_okay:1,
2696 udp_okay:1,
2697 sctp_okay:1,
2698 vlan_okay:1,
2699 no_encapped_ipv6:1,
2700 no_not_encapped:1;
2701};
2702
2703bool __skb_csum_offload_chk(struct sk_buff *skb,
2704 const struct skb_csum_offl_spec *spec,
2705 bool *csum_encapped,
2706 bool csum_help);
2707
2708static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2709 const struct skb_csum_offl_spec *spec,
2710 bool *csum_encapped,
2711 bool csum_help)
2712{
2713 if (skb->ip_summed != CHECKSUM_PARTIAL)
2714 return false;
2715
2716 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2717}
2718
2719static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2720 const struct skb_csum_offl_spec *spec)
2721{
2722 bool csum_encapped;
2723
2724 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2725}
2726
2727static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2728{
2729 static const struct skb_csum_offl_spec csum_offl_spec = {
2730 .ipv4_okay = 1,
2731 .ip_options_okay = 1,
2732 .ipv6_okay = 1,
2733 .vlan_okay = 1,
2734 .tcp_okay = 1,
2735 .udp_okay = 1,
2736 };
2737
2738 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2739}
2740
2741static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2742{
2743 static const struct skb_csum_offl_spec csum_offl_spec = {
2744 .ipv4_okay = 1,
2745 .ip_options_okay = 1,
2746 .tcp_okay = 1,
2747 .udp_okay = 1,
2748 .vlan_okay = 1,
2749 };
2750
2751 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2752}
2753
0c4e8581
SH
2754static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2755 unsigned short type,
3b04ddde 2756 const void *daddr, const void *saddr,
95c96174 2757 unsigned int len)
0c4e8581 2758{
f1ecfd5d 2759 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2760 return 0;
3b04ddde
SH
2761
2762 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2763}
2764
b95cce35
SH
2765static inline int dev_parse_header(const struct sk_buff *skb,
2766 unsigned char *haddr)
2767{
2768 const struct net_device *dev = skb->dev;
2769
1b83336b 2770 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2771 return 0;
3b04ddde 2772 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2773}
2774
2793a23a
WB
2775/* ll_header must have at least hard_header_len allocated */
2776static inline bool dev_validate_header(const struct net_device *dev,
2777 char *ll_header, int len)
2778{
2779 if (likely(len >= dev->hard_header_len))
2780 return true;
2781
2782 if (capable(CAP_SYS_RAWIO)) {
2783 memset(ll_header + len, 0, dev->hard_header_len - len);
2784 return true;
2785 }
2786
2787 if (dev->header_ops && dev->header_ops->validate)
2788 return dev->header_ops->validate(ll_header, len);
2789
2790 return false;
2791}
2792
1da177e4 2793typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2794int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2795static inline int unregister_gifconf(unsigned int family)
2796{
2797 return register_gifconf(family, NULL);
2798}
2799
99bbc707 2800#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2801#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2802struct sd_flow_limit {
2803 u64 count;
2804 unsigned int num_buckets;
2805 unsigned int history_head;
2806 u16 history[FLOW_LIMIT_HISTORY];
2807 u8 buckets[];
2808};
2809
2810extern int netdev_flow_limit_table_len;
2811#endif /* CONFIG_NET_FLOW_LIMIT */
2812
1da177e4 2813/*
5e82b4b2 2814 * Incoming packets are placed on per-CPU queues
1da177e4 2815 */
d94d9fee 2816struct softnet_data {
1da177e4 2817 struct list_head poll_list;
6e7676c1 2818 struct sk_buff_head process_queue;
1da177e4 2819
dee42870 2820 /* stats */
cd7b5396
DM
2821 unsigned int processed;
2822 unsigned int time_squeeze;
cd7b5396 2823 unsigned int received_rps;
fd793d89 2824#ifdef CONFIG_RPS
88751275 2825 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2826#endif
2827#ifdef CONFIG_NET_FLOW_LIMIT
2828 struct sd_flow_limit __rcu *flow_limit;
2829#endif
2830 struct Qdisc *output_queue;
2831 struct Qdisc **output_queue_tailp;
2832 struct sk_buff *completion_queue;
88751275 2833
4cdb1e2e 2834#ifdef CONFIG_RPS
501e7ef5
ED
2835 /* input_queue_head should be written by cpu owning this struct,
2836 * and only read by other cpus. Worth using a cache line.
2837 */
2838 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2839
2840 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2841 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2842 struct softnet_data *rps_ipi_next;
2843 unsigned int cpu;
76cc8b13 2844 unsigned int input_queue_tail;
1e94d72f 2845#endif
95c96174 2846 unsigned int dropped;
0a9627f2 2847 struct sk_buff_head input_pkt_queue;
bea3348e 2848 struct napi_struct backlog;
99bbc707 2849
1da177e4
LT
2850};
2851
76cc8b13 2852static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2853{
2854#ifdef CONFIG_RPS
76cc8b13
TH
2855 sd->input_queue_head++;
2856#endif
2857}
2858
2859static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2860 unsigned int *qtail)
2861{
2862#ifdef CONFIG_RPS
2863 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2864#endif
2865}
2866
0a9627f2 2867DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2868
f629d208 2869void __netif_schedule(struct Qdisc *q);
46e5da40 2870void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2871
fd2ea0a7
DM
2872static inline void netif_tx_schedule_all(struct net_device *dev)
2873{
2874 unsigned int i;
2875
2876 for (i = 0; i < dev->num_tx_queues; i++)
2877 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2878}
2879
f9a7cbbf 2880static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2881{
73466498 2882 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2883}
2884
bea3348e
SH
2885/**
2886 * netif_start_queue - allow transmit
2887 * @dev: network device
2888 *
2889 * Allow upper layers to call the device hard_start_xmit routine.
2890 */
1da177e4
LT
2891static inline void netif_start_queue(struct net_device *dev)
2892{
e8a0464c 2893 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2894}
2895
fd2ea0a7
DM
2896static inline void netif_tx_start_all_queues(struct net_device *dev)
2897{
2898 unsigned int i;
2899
2900 for (i = 0; i < dev->num_tx_queues; i++) {
2901 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2902 netif_tx_start_queue(txq);
2903 }
2904}
2905
46e5da40 2906void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2907
d29f749e
DJ
2908/**
2909 * netif_wake_queue - restart transmit
2910 * @dev: network device
2911 *
2912 * Allow upper layers to call the device hard_start_xmit routine.
2913 * Used for flow control when transmit resources are available.
2914 */
79d16385
DM
2915static inline void netif_wake_queue(struct net_device *dev)
2916{
e8a0464c 2917 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2918}
2919
fd2ea0a7
DM
2920static inline void netif_tx_wake_all_queues(struct net_device *dev)
2921{
2922 unsigned int i;
2923
2924 for (i = 0; i < dev->num_tx_queues; i++) {
2925 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2926 netif_tx_wake_queue(txq);
2927 }
2928}
2929
f9a7cbbf 2930static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2931{
73466498 2932 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2933}
2934
bea3348e
SH
2935/**
2936 * netif_stop_queue - stop transmitted packets
2937 * @dev: network device
2938 *
2939 * Stop upper layers calling the device hard_start_xmit routine.
2940 * Used for flow control when transmit resources are unavailable.
2941 */
1da177e4
LT
2942static inline void netif_stop_queue(struct net_device *dev)
2943{
e8a0464c 2944 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2945}
2946
a2029240 2947void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2948
4d29515f 2949static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2950{
73466498 2951 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2952}
2953
bea3348e
SH
2954/**
2955 * netif_queue_stopped - test if transmit queue is flowblocked
2956 * @dev: network device
2957 *
2958 * Test if transmit queue on device is currently unable to send.
2959 */
4d29515f 2960static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2961{
e8a0464c 2962 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2963}
2964
4d29515f 2965static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2966{
73466498
TH
2967 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2968}
2969
8e2f1a63
DB
2970static inline bool
2971netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2972{
2973 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2974}
2975
8e2f1a63
DB
2976static inline bool
2977netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2978{
2979 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2980}
2981
53511453
ED
2982/**
2983 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2984 * @dev_queue: pointer to transmit queue
2985 *
2986 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2987 * to give appropriate hint to the CPU.
53511453
ED
2988 */
2989static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2990{
2991#ifdef CONFIG_BQL
2992 prefetchw(&dev_queue->dql.num_queued);
2993#endif
2994}
2995
2996/**
2997 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2998 * @dev_queue: pointer to transmit queue
2999 *
3000 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 3001 * to give appropriate hint to the CPU.
53511453
ED
3002 */
3003static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3004{
3005#ifdef CONFIG_BQL
3006 prefetchw(&dev_queue->dql.limit);
3007#endif
3008}
3009
c5d67bd7
TH
3010static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3011 unsigned int bytes)
3012{
114cf580
TH
3013#ifdef CONFIG_BQL
3014 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
3015
3016 if (likely(dql_avail(&dev_queue->dql) >= 0))
3017 return;
3018
3019 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3020
3021 /*
3022 * The XOFF flag must be set before checking the dql_avail below,
3023 * because in netdev_tx_completed_queue we update the dql_completed
3024 * before checking the XOFF flag.
3025 */
3026 smp_mb();
3027
3028 /* check again in case another CPU has just made room avail */
3029 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3030 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 3031#endif
c5d67bd7
TH
3032}
3033
0042d0c8
FF
3034/**
3035 * netdev_sent_queue - report the number of bytes queued to hardware
3036 * @dev: network device
3037 * @bytes: number of bytes queued to the hardware device queue
3038 *
3039 * Report the number of bytes queued for sending/completion to the network
3040 * device hardware queue. @bytes should be a good approximation and should
3041 * exactly match netdev_completed_queue() @bytes
3042 */
c5d67bd7
TH
3043static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3044{
3045 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3046}
3047
3048static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 3049 unsigned int pkts, unsigned int bytes)
c5d67bd7 3050{
114cf580 3051#ifdef CONFIG_BQL
b37c0fbe
AD
3052 if (unlikely(!bytes))
3053 return;
3054
3055 dql_completed(&dev_queue->dql, bytes);
3056
3057 /*
3058 * Without the memory barrier there is a small possiblity that
3059 * netdev_tx_sent_queue will miss the update and cause the queue to
3060 * be stopped forever
3061 */
3062 smp_mb();
3063
3064 if (dql_avail(&dev_queue->dql) < 0)
3065 return;
3066
3067 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3068 netif_schedule_queue(dev_queue);
114cf580 3069#endif
c5d67bd7
TH
3070}
3071
0042d0c8
FF
3072/**
3073 * netdev_completed_queue - report bytes and packets completed by device
3074 * @dev: network device
3075 * @pkts: actual number of packets sent over the medium
3076 * @bytes: actual number of bytes sent over the medium
3077 *
3078 * Report the number of bytes and packets transmitted by the network device
3079 * hardware queue over the physical medium, @bytes must exactly match the
3080 * @bytes amount passed to netdev_sent_queue()
3081 */
c5d67bd7 3082static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3083 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3084{
3085 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3086}
3087
3088static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3089{
114cf580 3090#ifdef CONFIG_BQL
5c490354 3091 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3092 dql_reset(&q->dql);
3093#endif
c5d67bd7
TH
3094}
3095
0042d0c8
FF
3096/**
3097 * netdev_reset_queue - reset the packets and bytes count of a network device
3098 * @dev_queue: network device
3099 *
3100 * Reset the bytes and packet count of a network device and clear the
3101 * software flow control OFF bit for this network device
3102 */
c5d67bd7
TH
3103static inline void netdev_reset_queue(struct net_device *dev_queue)
3104{
3105 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3106}
3107
b9507bda
DB
3108/**
3109 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3110 * @dev: network device
3111 * @queue_index: given tx queue index
3112 *
3113 * Returns 0 if given tx queue index >= number of device tx queues,
3114 * otherwise returns the originally passed tx queue index.
3115 */
3116static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3117{
3118 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3119 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3120 dev->name, queue_index,
3121 dev->real_num_tx_queues);
3122 return 0;
3123 }
3124
3125 return queue_index;
3126}
3127
bea3348e
SH
3128/**
3129 * netif_running - test if up
3130 * @dev: network device
3131 *
3132 * Test if the device has been brought up.
3133 */
4d29515f 3134static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3135{
3136 return test_bit(__LINK_STATE_START, &dev->state);
3137}
3138
f25f4e44 3139/*
5e82b4b2 3140 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3141 * stop, and a check if it's stopped. All other device management is
3142 * done at the overall netdevice level.
3143 * Also test the device if we're multiqueue.
3144 */
bea3348e
SH
3145
3146/**
3147 * netif_start_subqueue - allow sending packets on subqueue
3148 * @dev: network device
3149 * @queue_index: sub queue index
3150 *
3151 * Start individual transmit queue of a device with multiple transmit queues.
3152 */
f25f4e44
PWJ
3153static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3154{
fd2ea0a7 3155 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3156
3157 netif_tx_start_queue(txq);
f25f4e44
PWJ
3158}
3159
bea3348e
SH
3160/**
3161 * netif_stop_subqueue - stop sending packets on subqueue
3162 * @dev: network device
3163 * @queue_index: sub queue index
3164 *
3165 * Stop individual transmit queue of a device with multiple transmit queues.
3166 */
f25f4e44
PWJ
3167static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3168{
fd2ea0a7 3169 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3170 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3171}
3172
bea3348e
SH
3173/**
3174 * netif_subqueue_stopped - test status of subqueue
3175 * @dev: network device
3176 * @queue_index: sub queue index
3177 *
3178 * Check individual transmit queue of a device with multiple transmit queues.
3179 */
4d29515f
DM
3180static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3181 u16 queue_index)
f25f4e44 3182{
fd2ea0a7 3183 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3184
3185 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3186}
3187
4d29515f
DM
3188static inline bool netif_subqueue_stopped(const struct net_device *dev,
3189 struct sk_buff *skb)
668f895a
PE
3190{
3191 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3192}
bea3348e 3193
46e5da40 3194void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 3195
537c00de 3196#ifdef CONFIG_XPS
53af53ae 3197int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3198 u16 index);
537c00de
AD
3199#else
3200static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3201 const struct cpumask *mask,
537c00de
AD
3202 u16 index)
3203{
3204 return 0;
3205}
3206#endif
3207
5605c762
JP
3208u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3209 unsigned int num_tx_queues);
3210
a3d22a68
VZ
3211/*
3212 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3213 * as a distribution range limit for the returned value.
3214 */
3215static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3216 struct sk_buff *skb)
a3d22a68
VZ
3217{
3218 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3219}
3220
bea3348e
SH
3221/**
3222 * netif_is_multiqueue - test if device has multiple transmit queues
3223 * @dev: network device
3224 *
3225 * Check if device has multiple transmit queues
bea3348e 3226 */
4d29515f 3227static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3228{
a02cec21 3229 return dev->num_tx_queues > 1;
f25f4e44 3230}
1da177e4 3231
f629d208 3232int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3233
a953be53 3234#ifdef CONFIG_SYSFS
f629d208 3235int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3236#else
3237static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3238 unsigned int rxq)
3239{
3240 return 0;
3241}
3242#endif
3243
a953be53
MD
3244#ifdef CONFIG_SYSFS
3245static inline unsigned int get_netdev_rx_queue_index(
3246 struct netdev_rx_queue *queue)
3247{
3248 struct net_device *dev = queue->dev;
3249 int index = queue - dev->_rx;
3250
3251 BUG_ON(index >= dev->num_rx_queues);
3252 return index;
3253}
3254#endif
3255
16917b87 3256#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3257int netif_get_num_default_rss_queues(void);
16917b87 3258
e6247027
ED
3259enum skb_free_reason {
3260 SKB_REASON_CONSUMED,
3261 SKB_REASON_DROPPED,
3262};
3263
3264void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3265void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3266
e6247027
ED
3267/*
3268 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3269 * interrupt context or with hardware interrupts being disabled.
3270 * (in_irq() || irqs_disabled())
3271 *
3272 * We provide four helpers that can be used in following contexts :
3273 *
3274 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3275 * replacing kfree_skb(skb)
3276 *
3277 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3278 * Typically used in place of consume_skb(skb) in TX completion path
3279 *
3280 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3281 * replacing kfree_skb(skb)
3282 *
3283 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3284 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3285 */
e6247027
ED
3286static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3287{
3288 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3289}
3290
3291static inline void dev_consume_skb_irq(struct sk_buff *skb)
3292{
3293 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3294}
3295
3296static inline void dev_kfree_skb_any(struct sk_buff *skb)
3297{
3298 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3299}
3300
3301static inline void dev_consume_skb_any(struct sk_buff *skb)
3302{
3303 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3304}
1da177e4 3305
f629d208
JP
3306int netif_rx(struct sk_buff *skb);
3307int netif_rx_ni(struct sk_buff *skb);
04eb4489 3308int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3309gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3310void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3311struct sk_buff *napi_get_frags(struct napi_struct *napi);
3312gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3313struct packet_offload *gro_find_receive_by_type(__be16 type);
3314struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3315
3316static inline void napi_free_frags(struct napi_struct *napi)
3317{
3318 kfree_skb(napi->skb);
3319 napi->skb = NULL;
3320}
3321
24b27fc4 3322bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3323int netdev_rx_handler_register(struct net_device *dev,
3324 rx_handler_func_t *rx_handler,
3325 void *rx_handler_data);
3326void netdev_rx_handler_unregister(struct net_device *dev);
3327
3328bool dev_valid_name(const char *name);
3329int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3330int dev_ethtool(struct net *net, struct ifreq *);
3331unsigned int dev_get_flags(const struct net_device *);
3332int __dev_change_flags(struct net_device *, unsigned int flags);
3333int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3334void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3335 unsigned int gchanges);
f629d208
JP
3336int dev_change_name(struct net_device *, const char *);
3337int dev_set_alias(struct net_device *, const char *, size_t);
3338int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3339int dev_set_mtu(struct net_device *, int);
3340void dev_set_group(struct net_device *, int);
3341int dev_set_mac_address(struct net_device *, struct sockaddr *);
3342int dev_change_carrier(struct net_device *, bool new_carrier);
3343int dev_get_phys_port_id(struct net_device *dev,
02637fce 3344 struct netdev_phys_item_id *ppid);
db24a904
DA
3345int dev_get_phys_port_name(struct net_device *dev,
3346 char *name, size_t len);
d746d707 3347int dev_change_proto_down(struct net_device *dev, bool proto_down);
a7862b45 3348int dev_change_xdp_fd(struct net_device *dev, int fd);
55a93b3e 3349struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3350struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3351 struct netdev_queue *txq, int *ret);
a0265d28 3352int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3353int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3354bool is_skb_forwardable(const struct net_device *dev,
3355 const struct sk_buff *skb);
1da177e4 3356
74b20582
DA
3357void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3358
20380731 3359extern int netdev_budget;
1da177e4
LT
3360
3361/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3362void netdev_run_todo(void);
1da177e4 3363
bea3348e
SH
3364/**
3365 * dev_put - release reference to device
3366 * @dev: network device
3367 *
9ef4429b 3368 * Release reference to device to allow it to be freed.
bea3348e 3369 */
1da177e4
LT
3370static inline void dev_put(struct net_device *dev)
3371{
933393f5 3372 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3373}
3374
bea3348e
SH
3375/**
3376 * dev_hold - get reference to device
3377 * @dev: network device
3378 *
9ef4429b 3379 * Hold reference to device to keep it from being freed.
bea3348e 3380 */
15333061
SH
3381static inline void dev_hold(struct net_device *dev)
3382{
933393f5 3383 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3384}
1da177e4
LT
3385
3386/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3387 * and _off may be called from IRQ context, but it is caller
3388 * who is responsible for serialization of these calls.
b00055aa
SR
3389 *
3390 * The name carrier is inappropriate, these functions should really be
3391 * called netif_lowerlayer_*() because they represent the state of any
3392 * kind of lower layer not just hardware media.
1da177e4
LT
3393 */
3394
f629d208
JP
3395void linkwatch_init_dev(struct net_device *dev);
3396void linkwatch_fire_event(struct net_device *dev);
3397void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3398
bea3348e
SH
3399/**
3400 * netif_carrier_ok - test if carrier present
3401 * @dev: network device
3402 *
3403 * Check if carrier is present on device
3404 */
4d29515f 3405static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3406{
3407 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3408}
3409
f629d208 3410unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3411
f629d208 3412void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3413
f629d208 3414void netif_carrier_on(struct net_device *dev);
1da177e4 3415
f629d208 3416void netif_carrier_off(struct net_device *dev);
1da177e4 3417
bea3348e
SH
3418/**
3419 * netif_dormant_on - mark device as dormant.
3420 * @dev: network device
3421 *
3422 * Mark device as dormant (as per RFC2863).
3423 *
3424 * The dormant state indicates that the relevant interface is not
3425 * actually in a condition to pass packets (i.e., it is not 'up') but is
3426 * in a "pending" state, waiting for some external event. For "on-
3427 * demand" interfaces, this new state identifies the situation where the
3428 * interface is waiting for events to place it in the up state.
bea3348e 3429 */
b00055aa
SR
3430static inline void netif_dormant_on(struct net_device *dev)
3431{
3432 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3433 linkwatch_fire_event(dev);
3434}
3435
bea3348e
SH
3436/**
3437 * netif_dormant_off - set device as not dormant.
3438 * @dev: network device
3439 *
3440 * Device is not in dormant state.
3441 */
b00055aa
SR
3442static inline void netif_dormant_off(struct net_device *dev)
3443{
3444 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3445 linkwatch_fire_event(dev);
3446}
3447
bea3348e
SH
3448/**
3449 * netif_dormant - test if carrier present
3450 * @dev: network device
3451 *
3452 * Check if carrier is present on device
3453 */
4d29515f 3454static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3455{
3456 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3457}
3458
3459
bea3348e
SH
3460/**
3461 * netif_oper_up - test if device is operational
3462 * @dev: network device
3463 *
3464 * Check if carrier is operational
3465 */
4d29515f 3466static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3467{
b00055aa
SR
3468 return (dev->operstate == IF_OPER_UP ||
3469 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3470}
3471
bea3348e
SH
3472/**
3473 * netif_device_present - is device available or removed
3474 * @dev: network device
3475 *
3476 * Check if device has not been removed from system.
3477 */
4d29515f 3478static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3479{
3480 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3481}
3482
f629d208 3483void netif_device_detach(struct net_device *dev);
1da177e4 3484
f629d208 3485void netif_device_attach(struct net_device *dev);
1da177e4
LT
3486
3487/*
3488 * Network interface message level settings
3489 */
1da177e4
LT
3490
3491enum {
3492 NETIF_MSG_DRV = 0x0001,
3493 NETIF_MSG_PROBE = 0x0002,
3494 NETIF_MSG_LINK = 0x0004,
3495 NETIF_MSG_TIMER = 0x0008,
3496 NETIF_MSG_IFDOWN = 0x0010,
3497 NETIF_MSG_IFUP = 0x0020,
3498 NETIF_MSG_RX_ERR = 0x0040,
3499 NETIF_MSG_TX_ERR = 0x0080,
3500 NETIF_MSG_TX_QUEUED = 0x0100,
3501 NETIF_MSG_INTR = 0x0200,
3502 NETIF_MSG_TX_DONE = 0x0400,
3503 NETIF_MSG_RX_STATUS = 0x0800,
3504 NETIF_MSG_PKTDATA = 0x1000,
3505 NETIF_MSG_HW = 0x2000,
3506 NETIF_MSG_WOL = 0x4000,
3507};
3508
3509#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3510#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3511#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3512#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3513#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3514#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3515#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3516#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3517#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3518#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3519#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3520#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3521#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3522#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3523#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3524
3525static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3526{
3527 /* use default */
3528 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3529 return default_msg_enable_bits;
3530 if (debug_value == 0) /* no output */
3531 return 0;
3532 /* set low N bits */
3533 return (1 << debug_value) - 1;
3534}
3535
c773e847 3536static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3537{
c773e847
DM
3538 spin_lock(&txq->_xmit_lock);
3539 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3540}
3541
fd2ea0a7
DM
3542static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3543{
3544 spin_lock_bh(&txq->_xmit_lock);
3545 txq->xmit_lock_owner = smp_processor_id();
3546}
3547
4d29515f 3548static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3549{
4d29515f 3550 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3551 if (likely(ok))
3552 txq->xmit_lock_owner = smp_processor_id();
3553 return ok;
3554}
3555
3556static inline void __netif_tx_unlock(struct netdev_queue *txq)
3557{
3558 txq->xmit_lock_owner = -1;
3559 spin_unlock(&txq->_xmit_lock);
3560}
3561
3562static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3563{
3564 txq->xmit_lock_owner = -1;
3565 spin_unlock_bh(&txq->_xmit_lock);
3566}
3567
08baf561
ED
3568static inline void txq_trans_update(struct netdev_queue *txq)
3569{
3570 if (txq->xmit_lock_owner != -1)
3571 txq->trans_start = jiffies;
3572}
3573
ba162f8e
FW
3574/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3575static inline void netif_trans_update(struct net_device *dev)
3576{
9b36627a
FW
3577 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3578
3579 if (txq->trans_start != jiffies)
3580 txq->trans_start = jiffies;
ba162f8e
FW
3581}
3582
d29f749e
DJ
3583/**
3584 * netif_tx_lock - grab network device transmit lock
3585 * @dev: network device
d29f749e
DJ
3586 *
3587 * Get network device transmit lock
3588 */
22dd7495
JHS
3589static inline void netif_tx_lock(struct net_device *dev)
3590{
e8a0464c 3591 unsigned int i;
c3f26a26 3592 int cpu;
c773e847 3593
c3f26a26
DM
3594 spin_lock(&dev->tx_global_lock);
3595 cpu = smp_processor_id();
e8a0464c
DM
3596 for (i = 0; i < dev->num_tx_queues; i++) {
3597 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3598
3599 /* We are the only thread of execution doing a
3600 * freeze, but we have to grab the _xmit_lock in
3601 * order to synchronize with threads which are in
3602 * the ->hard_start_xmit() handler and already
3603 * checked the frozen bit.
3604 */
e8a0464c 3605 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3606 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3607 __netif_tx_unlock(txq);
e8a0464c 3608 }
932ff279
HX
3609}
3610
3611static inline void netif_tx_lock_bh(struct net_device *dev)
3612{
e8a0464c
DM
3613 local_bh_disable();
3614 netif_tx_lock(dev);
932ff279
HX
3615}
3616
932ff279
HX
3617static inline void netif_tx_unlock(struct net_device *dev)
3618{
e8a0464c
DM
3619 unsigned int i;
3620
3621 for (i = 0; i < dev->num_tx_queues; i++) {
3622 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3623
c3f26a26
DM
3624 /* No need to grab the _xmit_lock here. If the
3625 * queue is not stopped for another reason, we
3626 * force a schedule.
3627 */
3628 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3629 netif_schedule_queue(txq);
c3f26a26
DM
3630 }
3631 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3632}
3633
3634static inline void netif_tx_unlock_bh(struct net_device *dev)
3635{
e8a0464c
DM
3636 netif_tx_unlock(dev);
3637 local_bh_enable();
932ff279
HX
3638}
3639
c773e847 3640#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3641 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3642 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3643 } \
3644}
3645
5efeac44
EB
3646#define HARD_TX_TRYLOCK(dev, txq) \
3647 (((dev->features & NETIF_F_LLTX) == 0) ? \
3648 __netif_tx_trylock(txq) : \
3649 true )
3650
c773e847 3651#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3652 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3653 __netif_tx_unlock(txq); \
22dd7495
JHS
3654 } \
3655}
3656
1da177e4
LT
3657static inline void netif_tx_disable(struct net_device *dev)
3658{
fd2ea0a7 3659 unsigned int i;
c3f26a26 3660 int cpu;
fd2ea0a7 3661
c3f26a26
DM
3662 local_bh_disable();
3663 cpu = smp_processor_id();
fd2ea0a7
DM
3664 for (i = 0; i < dev->num_tx_queues; i++) {
3665 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3666
3667 __netif_tx_lock(txq, cpu);
fd2ea0a7 3668 netif_tx_stop_queue(txq);
c3f26a26 3669 __netif_tx_unlock(txq);
fd2ea0a7 3670 }
c3f26a26 3671 local_bh_enable();
1da177e4
LT
3672}
3673
e308a5d8
DM
3674static inline void netif_addr_lock(struct net_device *dev)
3675{
3676 spin_lock(&dev->addr_list_lock);
3677}
3678
2429f7ac
JP
3679static inline void netif_addr_lock_nested(struct net_device *dev)
3680{
25175ba5
VY
3681 int subclass = SINGLE_DEPTH_NESTING;
3682
3683 if (dev->netdev_ops->ndo_get_lock_subclass)
3684 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3685
3686 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3687}
3688
e308a5d8
DM
3689static inline void netif_addr_lock_bh(struct net_device *dev)
3690{
3691 spin_lock_bh(&dev->addr_list_lock);
3692}
3693
3694static inline void netif_addr_unlock(struct net_device *dev)
3695{
3696 spin_unlock(&dev->addr_list_lock);
3697}
3698
3699static inline void netif_addr_unlock_bh(struct net_device *dev)
3700{
3701 spin_unlock_bh(&dev->addr_list_lock);
3702}
3703
f001fde5 3704/*
31278e71 3705 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3706 * rcu_read_lock held.
3707 */
3708#define for_each_dev_addr(dev, ha) \
31278e71 3709 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3710
1da177e4
LT
3711/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3712
f629d208 3713void ether_setup(struct net_device *dev);
1da177e4
LT
3714
3715/* Support for loadable net-drivers */
f629d208 3716struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3717 unsigned char name_assign_type,
f629d208
JP
3718 void (*setup)(struct net_device *),
3719 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3720#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3721 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3722
c835a677
TG
3723#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3724 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3725 count)
36909ea4 3726
f629d208
JP
3727int register_netdev(struct net_device *dev);
3728void unregister_netdev(struct net_device *dev);
f001fde5 3729
22bedad3 3730/* General hardware address lists handling functions */
f629d208
JP
3731int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3732 struct netdev_hw_addr_list *from_list, int addr_len);
3733void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3734 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3735int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3736 struct net_device *dev,
3737 int (*sync)(struct net_device *, const unsigned char *),
3738 int (*unsync)(struct net_device *,
3739 const unsigned char *));
3740void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3741 struct net_device *dev,
3742 int (*unsync)(struct net_device *,
3743 const unsigned char *));
f629d208 3744void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3745
f001fde5 3746/* Functions used for device addresses handling */
f629d208
JP
3747int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3748 unsigned char addr_type);
3749int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3750 unsigned char addr_type);
f629d208
JP
3751void dev_addr_flush(struct net_device *dev);
3752int dev_addr_init(struct net_device *dev);
a748ee24
JP
3753
3754/* Functions used for unicast addresses handling */
f629d208
JP
3755int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3756int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3757int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3758int dev_uc_sync(struct net_device *to, struct net_device *from);
3759int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3760void dev_uc_unsync(struct net_device *to, struct net_device *from);
3761void dev_uc_flush(struct net_device *dev);
3762void dev_uc_init(struct net_device *dev);
f001fde5 3763
670e5b8e
AD
3764/**
3765 * __dev_uc_sync - Synchonize device's unicast list
3766 * @dev: device to sync
3767 * @sync: function to call if address should be added
3768 * @unsync: function to call if address should be removed
3769 *
3770 * Add newly added addresses to the interface, and release
3771 * addresses that have been deleted.
5e82b4b2 3772 */
670e5b8e
AD
3773static inline int __dev_uc_sync(struct net_device *dev,
3774 int (*sync)(struct net_device *,
3775 const unsigned char *),
3776 int (*unsync)(struct net_device *,
3777 const unsigned char *))
3778{
3779 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3780}
3781
3782/**
e793c0f7 3783 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3784 * @dev: device to sync
3785 * @unsync: function to call if address should be removed
3786 *
3787 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3788 */
670e5b8e
AD
3789static inline void __dev_uc_unsync(struct net_device *dev,
3790 int (*unsync)(struct net_device *,
3791 const unsigned char *))
3792{
3793 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3794}
3795
22bedad3 3796/* Functions used for multicast addresses handling */
f629d208
JP
3797int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3798int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3799int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3800int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3801int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3802int dev_mc_sync(struct net_device *to, struct net_device *from);
3803int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3804void dev_mc_unsync(struct net_device *to, struct net_device *from);
3805void dev_mc_flush(struct net_device *dev);
3806void dev_mc_init(struct net_device *dev);
f001fde5 3807
670e5b8e
AD
3808/**
3809 * __dev_mc_sync - Synchonize device's multicast list
3810 * @dev: device to sync
3811 * @sync: function to call if address should be added
3812 * @unsync: function to call if address should be removed
3813 *
3814 * Add newly added addresses to the interface, and release
3815 * addresses that have been deleted.
5e82b4b2 3816 */
670e5b8e
AD
3817static inline int __dev_mc_sync(struct net_device *dev,
3818 int (*sync)(struct net_device *,
3819 const unsigned char *),
3820 int (*unsync)(struct net_device *,
3821 const unsigned char *))
3822{
3823 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3824}
3825
3826/**
e793c0f7 3827 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3828 * @dev: device to sync
3829 * @unsync: function to call if address should be removed
3830 *
3831 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3832 */
670e5b8e
AD
3833static inline void __dev_mc_unsync(struct net_device *dev,
3834 int (*unsync)(struct net_device *,
3835 const unsigned char *))
3836{
3837 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3838}
3839
4417da66 3840/* Functions used for secondary unicast and multicast support */
f629d208
JP
3841void dev_set_rx_mode(struct net_device *dev);
3842void __dev_set_rx_mode(struct net_device *dev);
3843int dev_set_promiscuity(struct net_device *dev, int inc);
3844int dev_set_allmulti(struct net_device *dev, int inc);
3845void netdev_state_change(struct net_device *dev);
3846void netdev_notify_peers(struct net_device *dev);
3847void netdev_features_change(struct net_device *dev);
1da177e4 3848/* Load a device via the kmod */
f629d208
JP
3849void dev_load(struct net *net, const char *name);
3850struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3851 struct rtnl_link_stats64 *storage);
3852void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3853 const struct net_device_stats *netdev_stats);
eeda3fd6 3854
1da177e4 3855extern int netdev_max_backlog;
3b098e2d 3856extern int netdev_tstamp_prequeue;
1da177e4 3857extern int weight_p;
9ff162a8 3858
f629d208 3859bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3860struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3861 struct list_head **iter);
f629d208
JP
3862struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3863 struct list_head **iter);
8b5be856 3864
44a40855
VY
3865/* iterate through upper list, must be called under RCU read lock */
3866#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3867 for (iter = &(dev)->adj_list.upper, \
3868 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3869 updev; \
3870 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3871
8b5be856 3872/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3873#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3874 for (iter = &(dev)->all_adj_list.upper, \
3875 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3876 updev; \
3877 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3878
f629d208
JP
3879void *netdev_lower_get_next_private(struct net_device *dev,
3880 struct list_head **iter);
3881void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3882 struct list_head **iter);
31088a11
VF
3883
3884#define netdev_for_each_lower_private(dev, priv, iter) \
3885 for (iter = (dev)->adj_list.lower.next, \
3886 priv = netdev_lower_get_next_private(dev, &(iter)); \
3887 priv; \
3888 priv = netdev_lower_get_next_private(dev, &(iter)))
3889
3890#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3891 for (iter = &(dev)->adj_list.lower, \
3892 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3893 priv; \
3894 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3895
4085ebe8
VY
3896void *netdev_lower_get_next(struct net_device *dev,
3897 struct list_head **iter);
7ce856aa 3898
4085ebe8 3899#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3900 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3901 ldev = netdev_lower_get_next(dev, &(iter)); \
3902 ldev; \
3903 ldev = netdev_lower_get_next(dev, &(iter)))
3904
7ce856aa
JP
3905struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3906 struct list_head **iter);
3907struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3908 struct list_head **iter);
3909
3910#define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3911 for (iter = (dev)->all_adj_list.lower.next, \
3912 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3913 ldev; \
3914 ldev = netdev_all_lower_get_next(dev, &(iter)))
3915
3916#define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
e4961b07 3917 for (iter = &(dev)->all_adj_list.lower, \
7ce856aa
JP
3918 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3919 ldev; \
3920 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3921
f629d208 3922void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3923void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3924struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3925struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3926int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3927int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3928 struct net_device *upper_dev,
29bf24af 3929 void *upper_priv, void *upper_info);
f629d208
JP
3930void netdev_upper_dev_unlink(struct net_device *dev,
3931 struct net_device *upper_dev);
5bb025fa 3932void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3933void *netdev_lower_dev_get_private(struct net_device *dev,
3934 struct net_device *lower_dev);
04d48266
JP
3935void netdev_lower_state_changed(struct net_device *lower_dev,
3936 void *lower_state_info);
18bfb924
JP
3937int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3938 struct neighbour *n);
3939void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3940 struct neighbour *n);
960fb622
ED
3941
3942/* RSS keys are 40 or 52 bytes long */
3943#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3944extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3945void netdev_rss_key_fill(void *buffer, size_t len);
3946
952fcfd0 3947int dev_get_nest_level(struct net_device *dev);
f629d208
JP
3948int skb_checksum_help(struct sk_buff *skb);
3949struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3950 netdev_features_t features, bool tx_path);
3951struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3952 netdev_features_t features);
12b0004d 3953
61bd3857
MS
3954struct netdev_bonding_info {
3955 ifslave slave;
3956 ifbond master;
3957};
3958
3959struct netdev_notifier_bonding_info {
3960 struct netdev_notifier_info info; /* must be first */
3961 struct netdev_bonding_info bonding_info;
3962};
3963
3964void netdev_bonding_info_change(struct net_device *dev,
3965 struct netdev_bonding_info *bonding_info);
3966
12b0004d
CW
3967static inline
3968struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3969{
3970 return __skb_gso_segment(skb, features, true);
3971}
53d6471c 3972__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3973
3974static inline bool can_checksum_protocol(netdev_features_t features,
3975 __be16 protocol)
3976{
c8cd0989
TH
3977 if (protocol == htons(ETH_P_FCOE))
3978 return !!(features & NETIF_F_FCOE_CRC);
3979
3980 /* Assume this is an IP checksum (not SCTP CRC) */
3981
3982 if (features & NETIF_F_HW_CSUM) {
3983 /* Can checksum everything */
3984 return true;
3985 }
3986
3987 switch (protocol) {
3988 case htons(ETH_P_IP):
3989 return !!(features & NETIF_F_IP_CSUM);
3990 case htons(ETH_P_IPV6):
3991 return !!(features & NETIF_F_IPV6_CSUM);
3992 default:
3993 return false;
3994 }
ec5f0615 3995}
12b0004d 3996
6ae23ad3
TH
3997/* Map an ethertype into IP protocol if possible */
3998static inline int eproto_to_ipproto(int eproto)
3999{
4000 switch (eproto) {
4001 case htons(ETH_P_IP):
4002 return IPPROTO_IP;
4003 case htons(ETH_P_IPV6):
4004 return IPPROTO_IPV6;
4005 default:
4006 return -1;
4007 }
4008}
4009
fb286bb2 4010#ifdef CONFIG_BUG
f629d208 4011void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
4012#else
4013static inline void netdev_rx_csum_fault(struct net_device *dev)
4014{
4015}
4016#endif
1da177e4 4017/* rx skb timestamps */
f629d208
JP
4018void net_enable_timestamp(void);
4019void net_disable_timestamp(void);
1da177e4 4020
20380731 4021#ifdef CONFIG_PROC_FS
f629d208 4022int __init dev_proc_init(void);
900ff8c6
CW
4023#else
4024#define dev_proc_init() 0
20380731
ACM
4025#endif
4026
4798248e 4027static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4028 struct sk_buff *skb, struct net_device *dev,
4029 bool more)
4798248e 4030{
fa2dbdc2 4031 skb->xmit_more = more ? 1 : 0;
0b725a2c 4032 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4033}
4034
10b3ad8c 4035static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4036 struct netdev_queue *txq, bool more)
4798248e
DM
4037{
4038 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4039 int rc;
4798248e 4040
fa2dbdc2 4041 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4042 if (rc == NETDEV_TX_OK)
4043 txq_trans_update(txq);
4044
4045 return rc;
4798248e
DM
4046}
4047
42a2d923
LT
4048int netdev_class_create_file_ns(struct class_attribute *class_attr,
4049 const void *ns);
4050void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4051 const void *ns);
58292cbe
TH
4052
4053static inline int netdev_class_create_file(struct class_attribute *class_attr)
4054{
4055 return netdev_class_create_file_ns(class_attr, NULL);
4056}
4057
4058static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4059{
4060 netdev_class_remove_file_ns(class_attr, NULL);
4061}
b8a9787e 4062
04600794
JB
4063extern struct kobj_ns_type_operations net_ns_type_operations;
4064
f629d208 4065const char *netdev_drivername(const struct net_device *dev);
6579e57b 4066
f629d208 4067void linkwatch_run_queue(void);
20380731 4068
da08143b
MK
4069static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4070 netdev_features_t f2)
4071{
c8cd0989
TH
4072 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4073 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4074 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4075 else
b6a0e72a 4076 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4077 }
da08143b 4078
c8cd0989 4079 return f1 & f2;
da08143b
MK
4080}
4081
c8f44aff
MM
4082static inline netdev_features_t netdev_get_wanted_features(
4083 struct net_device *dev)
5455c699
MM
4084{
4085 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4086}
c8f44aff
MM
4087netdev_features_t netdev_increment_features(netdev_features_t all,
4088 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4089
4090/* Allow TSO being used on stacked device :
4091 * Performing the GSO segmentation before last device
4092 * is a performance improvement.
4093 */
4094static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4095 netdev_features_t mask)
4096{
4097 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4098}
4099
6cb6a27c 4100int __netdev_update_features(struct net_device *dev);
5455c699 4101void netdev_update_features(struct net_device *dev);
afe12cc8 4102void netdev_change_features(struct net_device *dev);
7f353bf2 4103
fc4a7489
PM
4104void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4105 struct net_device *dev);
4106
e38f3025
TM
4107netdev_features_t passthru_features_check(struct sk_buff *skb,
4108 struct net_device *dev,
4109 netdev_features_t features);
c1e756bf 4110netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4111
4d29515f 4112static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4113{
7b748340 4114 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4115
4116 /* check flags correspondence */
4117 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4118 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4119 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4120 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4121 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4122 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4123 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4124 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4125 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4126 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4127 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4128 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4129 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4130 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4131 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4132 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
0345e186 4133
d6b4991a 4134 return (features & feature) == feature;
576a30eb
HX
4135}
4136
4d29515f 4137static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4138{
278b2513 4139 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4140 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4141}
4142
8b86a61d 4143static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4144 netdev_features_t features)
7967168c 4145{
fc741216 4146 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4147 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4148 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4149}
4150
82cc1a7a
PWJ
4151static inline void netif_set_gso_max_size(struct net_device *dev,
4152 unsigned int size)
4153{
4154 dev->gso_max_size = size;
4155}
4156
7a7ffbab
WCC
4157static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4158 int pulled_hlen, u16 mac_offset,
4159 int mac_len)
4160{
4161 skb->protocol = protocol;
4162 skb->encapsulation = 1;
4163 skb_push(skb, pulled_hlen);
4164 skb_reset_transport_header(skb);
4165 skb->mac_header = mac_offset;
4166 skb->network_header = skb->mac_header + mac_len;
4167 skb->mac_len = mac_len;
4168}
4169
3c175784
SD
4170static inline bool netif_is_macsec(const struct net_device *dev)
4171{
4172 return dev->priv_flags & IFF_MACSEC;
4173}
4174
b618aaa9 4175static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4176{
4177 return dev->priv_flags & IFF_MACVLAN;
4178}
4179
b618aaa9 4180static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4181{
4182 return dev->priv_flags & IFF_MACVLAN_PORT;
4183}
4184
b618aaa9 4185static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4186{
4187 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4188}
4189
b618aaa9 4190static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4191{
4192 return dev->priv_flags & IFF_IPVLAN_MASTER;
4193}
4194
b618aaa9 4195static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4196{
4197 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4198}
4199
b618aaa9 4200static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4201{
4202 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4203}
4204
3bdc0eba
BG
4205static inline bool netif_supports_nofcs(struct net_device *dev)
4206{
4207 return dev->priv_flags & IFF_SUPP_NOFCS;
4208}
4209
007979ea 4210static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4211{
007979ea 4212 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4213}
4214
fee6d4c7
DA
4215static inline bool netif_is_l3_slave(const struct net_device *dev)
4216{
4217 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4218}
4219
0894ae3f
JP
4220static inline bool netif_is_bridge_master(const struct net_device *dev)
4221{
4222 return dev->priv_flags & IFF_EBRIDGE;
4223}
4224
28f9ee22
VY
4225static inline bool netif_is_bridge_port(const struct net_device *dev)
4226{
4227 return dev->priv_flags & IFF_BRIDGE_PORT;
4228}
4229
35d4e172
JP
4230static inline bool netif_is_ovs_master(const struct net_device *dev)
4231{
4232 return dev->priv_flags & IFF_OPENVSWITCH;
4233}
4234
b618aaa9 4235static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4236{
4237 return dev->priv_flags & IFF_TEAM;
4238}
4239
b618aaa9 4240static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4241{
4242 return dev->priv_flags & IFF_TEAM_PORT;
4243}
4244
b618aaa9 4245static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4246{
4247 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4248}
4249
b618aaa9 4250static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4251{
4252 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4253}
4254
d4ab4286
KJ
4255static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4256{
4257 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4258}
4259
02875878
ED
4260/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4261static inline void netif_keep_dst(struct net_device *dev)
4262{
4263 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4264}
4265
18d3df3e
PA
4266/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4267static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4268{
4269 /* TODO: reserve and use an additional IFF bit, if we get more users */
4270 return dev->priv_flags & IFF_MACSEC;
4271}
4272
505d4f73 4273extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4274
571ba423
JP
4275/* Logging, debugging and troubleshooting/diagnostic helpers. */
4276
4277/* netdev_printk helpers, similar to dev_printk */
4278
4279static inline const char *netdev_name(const struct net_device *dev)
4280{
c6f854d5
VF
4281 if (!dev->name[0] || strchr(dev->name, '%'))
4282 return "(unnamed net_device)";
571ba423
JP
4283 return dev->name;
4284}
4285
ccc7f496
VF
4286static inline const char *netdev_reg_state(const struct net_device *dev)
4287{
4288 switch (dev->reg_state) {
4289 case NETREG_UNINITIALIZED: return " (uninitialized)";
4290 case NETREG_REGISTERED: return "";
4291 case NETREG_UNREGISTERING: return " (unregistering)";
4292 case NETREG_UNREGISTERED: return " (unregistered)";
4293 case NETREG_RELEASED: return " (released)";
4294 case NETREG_DUMMY: return " (dummy)";
4295 }
4296
4297 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4298 return " (unknown)";
4299}
4300
f629d208 4301__printf(3, 4)
6ea754eb
JP
4302void netdev_printk(const char *level, const struct net_device *dev,
4303 const char *format, ...);
f629d208 4304__printf(2, 3)
6ea754eb 4305void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4306__printf(2, 3)
6ea754eb 4307void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4308__printf(2, 3)
6ea754eb 4309void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4310__printf(2, 3)
6ea754eb 4311void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4312__printf(2, 3)
6ea754eb 4313void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4314__printf(2, 3)
6ea754eb 4315void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4316__printf(2, 3)
6ea754eb 4317void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4318
8909c9ad
VK
4319#define MODULE_ALIAS_NETDEV(device) \
4320 MODULE_ALIAS("netdev-" device)
4321
b558c96f 4322#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4323#define netdev_dbg(__dev, format, args...) \
4324do { \
ffa10cb4 4325 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4326} while (0)
b558c96f
JC
4327#elif defined(DEBUG)
4328#define netdev_dbg(__dev, format, args...) \
4329 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4330#else
4331#define netdev_dbg(__dev, format, args...) \
4332({ \
4333 if (0) \
4334 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4335})
4336#endif
4337
4338#if defined(VERBOSE_DEBUG)
4339#define netdev_vdbg netdev_dbg
4340#else
4341
4342#define netdev_vdbg(dev, format, args...) \
4343({ \
4344 if (0) \
4345 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4346 0; \
4347})
4348#endif
4349
4350/*
4351 * netdev_WARN() acts like dev_printk(), but with the key difference
4352 * of using a WARN/WARN_ON to get the message out, including the
4353 * file/line information and a backtrace.
4354 */
4355#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4356 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4357 netdev_reg_state(dev), ##args)
571ba423 4358
b3d95c5c
JP
4359/* netif printk helpers, similar to netdev_printk */
4360
4361#define netif_printk(priv, type, level, dev, fmt, args...) \
4362do { \
4363 if (netif_msg_##type(priv)) \
4364 netdev_printk(level, (dev), fmt, ##args); \
4365} while (0)
4366
f45f4321
JP
4367#define netif_level(level, priv, type, dev, fmt, args...) \
4368do { \
4369 if (netif_msg_##type(priv)) \
4370 netdev_##level(dev, fmt, ##args); \
4371} while (0)
4372
b3d95c5c 4373#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4374 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4375#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4376 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4377#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4378 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4379#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4380 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4381#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4382 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4383#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4384 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4385#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4386 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4387
0053ea9c 4388#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4389#define netif_dbg(priv, type, netdev, format, args...) \
4390do { \
4391 if (netif_msg_##type(priv)) \
b5fb0a03 4392 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4393} while (0)
0053ea9c
JP
4394#elif defined(DEBUG)
4395#define netif_dbg(priv, type, dev, format, args...) \
4396 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4397#else
4398#define netif_dbg(priv, type, dev, format, args...) \
4399({ \
4400 if (0) \
4401 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4402 0; \
4403})
4404#endif
4405
4406#if defined(VERBOSE_DEBUG)
bcfcc450 4407#define netif_vdbg netif_dbg
b3d95c5c
JP
4408#else
4409#define netif_vdbg(priv, type, dev, format, args...) \
4410({ \
4411 if (0) \
a4ed89cb 4412 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4413 0; \
4414})
4415#endif
571ba423 4416
900ff8c6
CW
4417/*
4418 * The list of packet types we will receive (as opposed to discard)
4419 * and the routines to invoke.
4420 *
4421 * Why 16. Because with 16 the only overlap we get on a hash of the
4422 * low nibble of the protocol value is RARP/SNAP/X.25.
4423 *
4424 * NOTE: That is no longer true with the addition of VLAN tags. Not
4425 * sure which should go first, but I bet it won't make much
4426 * difference if we are running VLANs. The good news is that
4427 * this protocol won't be in the list unless compiled in, so
4428 * the average user (w/out VLANs) will not be adversely affected.
4429 * --BLG
4430 *
4431 * 0800 IP
4432 * 8100 802.1Q VLAN
4433 * 0001 802.3
4434 * 0002 AX.25
4435 * 0004 802.2
4436 * 8035 RARP
4437 * 0005 SNAP
4438 * 0805 X.25
4439 * 0806 ARP
4440 * 8137 IPX
4441 * 0009 Localtalk
4442 * 86DD IPv6
4443 */
4444#define PTYPE_HASH_SIZE (16)
4445#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4446
385a154c 4447#endif /* _LINUX_NETDEVICE_H */