selftests/net: in timestamping, strncpy needs to preserve null byte
[linux-block.git] / include / linux / netdevice.h
CommitLineData
2874c5fd 1/* SPDX-License-Identifier: GPL-2.0-or-later */
1da177e4
LT
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
02c30a84 11 * Authors: Ross Biro
1da177e4
LT
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
1da177e4
LT
19 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
d7fe0f24 24#include <linux/timer.h>
187f1882 25#include <linux/bug.h>
bea3348e 26#include <linux/delay.h>
60063497 27#include <linux/atomic.h>
53511453 28#include <linux/prefetch.h>
1da177e4
LT
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31
1da177e4 32#include <linux/percpu.h>
4d5b78c0 33#include <linux/rculist.h>
bea3348e 34#include <linux/workqueue.h>
114cf580 35#include <linux/dynamic_queue_limits.h>
1da177e4 36
b1b67dd4 37#include <linux/ethtool.h>
a050c33f 38#include <net/net_namespace.h>
7a6b6f51 39#ifdef CONFIG_DCB
2f90b865
AD
40#include <net/dcbnl.h>
41#endif
5bc1421e 42#include <net/netprio_cgroup.h>
e817f856 43#include <net/xdp.h>
a050c33f 44
a59e2ecb 45#include <linux/netdev_features.h>
77162022 46#include <linux/neighbour.h>
607ca46e 47#include <uapi/linux/netdevice.h>
61bd3857 48#include <uapi/linux/if_bonding.h>
e4c6734e 49#include <uapi/linux/pkt_cls.h>
59cc1f61 50#include <linux/hashtable.h>
a59e2ecb 51
115c1d6e 52struct netpoll_info;
313162d0 53struct device;
c1f19b51 54struct phy_device;
2f657a60 55struct dsa_port;
607259a6 56struct ip_tunnel_parm;
30e9bb84
AT
57struct macsec_context;
58struct macsec_ops;
c6e970a0 59
e679c9c1 60struct sfp_bus;
704232c2
JB
61/* 802.11 specific */
62struct wireless_dev;
98a18b6f
AA
63/* 802.15.4 specific */
64struct wpan_dev;
03c57747 65struct mpls_dev;
7c46a640
AD
66/* UDP Tunnel offloads */
67struct udp_tunnel_info;
a7862b45 68struct bpf_prog;
814abfab 69struct xdp_buff;
1da177e4 70
f629d208
JP
71void netdev_set_default_ethtool_ops(struct net_device *dev,
72 const struct ethtool_ops *ops);
d07d7507 73
9a1654ba
JP
74/* Backlog congestion levels */
75#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
76#define NET_RX_DROP 1 /* packet dropped */
77
7151affe
TY
78#define MAX_NEST_DEV 8
79
572a9d7b
PM
80/*
81 * Transmit return codes: transmit return codes originate from three different
82 * namespaces:
83 *
84 * - qdisc return codes
85 * - driver transmit return codes
86 * - errno values
87 *
88 * Drivers are allowed to return any one of those in their hard_start_xmit()
89 * function. Real network devices commonly used with qdiscs should only return
90 * the driver transmit return codes though - when qdiscs are used, the actual
91 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
92 * higher layers. Virtual network devices transmit synchronously; in this case
93 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
94 * others are propagated to higher layers.
95 */
96
97/* qdisc ->enqueue() return codes. */
98#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
99#define NET_XMIT_DROP 0x01 /* skb dropped */
100#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 101#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 102
b9df3cb8
GR
103/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
104 * indicates that the device will soon be dropping packets, or already drops
105 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 106#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
107#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
108
dc1f8bf6 109/* Driver transmit return codes */
9a1654ba 110#define NETDEV_TX_MASK 0xf0
572a9d7b 111
dc1f8bf6 112enum netdev_tx {
572a9d7b 113 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
114 NETDEV_TX_OK = 0x00, /* driver took care of packet */
115 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
116};
117typedef enum netdev_tx netdev_tx_t;
118
9a1654ba
JP
119/*
120 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
121 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
122 */
123static inline bool dev_xmit_complete(int rc)
124{
125 /*
126 * Positive cases with an skb consumed by a driver:
127 * - successful transmission (rc == NETDEV_TX_OK)
128 * - error while transmitting (rc < 0)
129 * - error while queueing to a different device (rc & NET_XMIT_MASK)
130 */
131 if (likely(rc < NET_XMIT_MASK))
132 return true;
133
134 return false;
135}
136
1da177e4 137/*
5e82b4b2 138 * Compute the worst-case header length according to the protocols
1da177e4
LT
139 * used.
140 */
fe2918b0 141
c0eb4540
KS
142#if defined(CONFIG_HYPERV_NET)
143# define LL_MAX_HEADER 128
144#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
145# if defined(CONFIG_MAC80211_MESH)
146# define LL_MAX_HEADER 128
147# else
148# define LL_MAX_HEADER 96
149# endif
1da177e4 150#else
8388e3da 151# define LL_MAX_HEADER 32
1da177e4
LT
152#endif
153
d11ead75
BH
154#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
155 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
156#define MAX_HEADER LL_MAX_HEADER
157#else
158#define MAX_HEADER (LL_MAX_HEADER + 48)
159#endif
160
161/*
be1f3c2c
BH
162 * Old network device statistics. Fields are native words
163 * (unsigned long) so they can be read and written atomically.
1da177e4 164 */
fe2918b0 165
d94d9fee 166struct net_device_stats {
3cfde79c
BH
167 unsigned long rx_packets;
168 unsigned long tx_packets;
169 unsigned long rx_bytes;
170 unsigned long tx_bytes;
171 unsigned long rx_errors;
172 unsigned long tx_errors;
173 unsigned long rx_dropped;
174 unsigned long tx_dropped;
175 unsigned long multicast;
1da177e4 176 unsigned long collisions;
1da177e4 177 unsigned long rx_length_errors;
3cfde79c
BH
178 unsigned long rx_over_errors;
179 unsigned long rx_crc_errors;
180 unsigned long rx_frame_errors;
181 unsigned long rx_fifo_errors;
182 unsigned long rx_missed_errors;
1da177e4
LT
183 unsigned long tx_aborted_errors;
184 unsigned long tx_carrier_errors;
185 unsigned long tx_fifo_errors;
186 unsigned long tx_heartbeat_errors;
187 unsigned long tx_window_errors;
1da177e4
LT
188 unsigned long rx_compressed;
189 unsigned long tx_compressed;
190};
191
1da177e4
LT
192
193#include <linux/cache.h>
194#include <linux/skbuff.h>
195
adc9300e 196#ifdef CONFIG_RPS
c5905afb 197#include <linux/static_key.h>
dc05360f
ED
198extern struct static_key_false rps_needed;
199extern struct static_key_false rfs_needed;
adc9300e
ED
200#endif
201
1da177e4
LT
202struct neighbour;
203struct neigh_parms;
204struct sk_buff;
205
f001fde5
JP
206struct netdev_hw_addr {
207 struct list_head list;
208 unsigned char addr[MAX_ADDR_LEN];
209 unsigned char type;
ccffad25
JP
210#define NETDEV_HW_ADDR_T_LAN 1
211#define NETDEV_HW_ADDR_T_SAN 2
212#define NETDEV_HW_ADDR_T_SLAVE 3
213#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 214#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 215 bool global_use;
4cd729b0 216 int sync_cnt;
8f8f103d 217 int refcount;
4543fbef 218 int synced;
f001fde5
JP
219 struct rcu_head rcu_head;
220};
221
31278e71
JP
222struct netdev_hw_addr_list {
223 struct list_head list;
224 int count;
225};
226
22bedad3
JP
227#define netdev_hw_addr_list_count(l) ((l)->count)
228#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
229#define netdev_hw_addr_list_for_each(ha, l) \
230 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 231
22bedad3
JP
232#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
233#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
234#define netdev_for_each_uc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 236
22bedad3
JP
237#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
238#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 239#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 240 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 241
d94d9fee 242struct hh_cache {
5b3dc2f3 243 unsigned int hh_len;
3644f0ce 244 seqlock_t hh_lock;
1da177e4
LT
245
246 /* cached hardware header; allow for machine alignment needs. */
247#define HH_DATA_MOD 16
248#define HH_DATA_OFF(__len) \
5ba0eac6 249 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
250#define HH_DATA_ALIGN(__len) \
251 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
252 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
253};
254
5e82b4b2 255/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
256 * Alternative is:
257 * dev->hard_header_len ? (dev->hard_header_len +
258 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
259 *
260 * We could use other alignment values, but we must maintain the
261 * relationship HH alignment <= LL alignment.
262 */
263#define LL_RESERVED_SPACE(dev) \
f5184d26 264 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 265#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 266 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 267
3b04ddde
SH
268struct header_ops {
269 int (*create) (struct sk_buff *skb, struct net_device *dev,
270 unsigned short type, const void *daddr,
95c96174 271 const void *saddr, unsigned int len);
3b04ddde 272 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 273 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
274 void (*cache_update)(struct hh_cache *hh,
275 const struct net_device *dev,
276 const unsigned char *haddr);
2793a23a 277 bool (*validate)(const char *ll_header, unsigned int len);
e78b2915 278 __be16 (*parse_protocol)(const struct sk_buff *skb);
3b04ddde
SH
279};
280
1da177e4 281/* These flag bits are private to the generic network queueing
5e82b4b2 282 * layer; they may not be explicitly referenced by any other
1da177e4
LT
283 * code.
284 */
285
d94d9fee 286enum netdev_state_t {
1da177e4
LT
287 __LINK_STATE_START,
288 __LINK_STATE_PRESENT,
1da177e4 289 __LINK_STATE_NOCARRIER,
b00055aa
SR
290 __LINK_STATE_LINKWATCH_PENDING,
291 __LINK_STATE_DORMANT,
eec517cd 292 __LINK_STATE_TESTING,
1da177e4
LT
293};
294
295
296/*
5e82b4b2 297 * This structure holds boot-time configured netdevice settings. They
fe2918b0 298 * are then used in the device probing.
1da177e4
LT
299 */
300struct netdev_boot_setup {
301 char name[IFNAMSIZ];
302 struct ifmap map;
303};
304#define NETDEV_BOOT_SETUP_MAX 8
305
f629d208 306int __init netdev_boot_setup(char *str);
1da177e4 307
6312fe77
LR
308struct gro_list {
309 struct list_head list;
310 int count;
311};
312
bea3348e 313/*
d9f37d01
LR
314 * size of gro hash buckets, must less than bit number of
315 * napi_struct::gro_bitmask
bea3348e 316 */
07d78363 317#define GRO_HASH_BUCKETS 8
d9f37d01
LR
318
319/*
320 * Structure for NAPI scheduling similar to tasklet but with weighting
321 */
bea3348e
SH
322struct napi_struct {
323 /* The poll_list must only be managed by the entity which
324 * changes the state of the NAPI_STATE_SCHED bit. This means
325 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 326 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
327 * can remove from the list right before clearing the bit.
328 */
329 struct list_head poll_list;
330
331 unsigned long state;
332 int weight;
6f8b12d6 333 int defer_hard_irqs_count;
d9f37d01 334 unsigned long gro_bitmask;
bea3348e
SH
335 int (*poll)(struct napi_struct *, int);
336#ifdef CONFIG_NETPOLL
bea3348e 337 int poll_owner;
bea3348e 338#endif
5d38a079 339 struct net_device *dev;
6312fe77 340 struct gro_list gro_hash[GRO_HASH_BUCKETS];
5d38a079 341 struct sk_buff *skb;
323ebb61
EC
342 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
343 int rx_count; /* length of rx_list */
3b47d303 344 struct hrtimer timer;
404f7c9e 345 struct list_head dev_list;
af12fa6e
ET
346 struct hlist_node napi_hash_node;
347 unsigned int napi_id;
bea3348e
SH
348};
349
d94d9fee 350enum {
bea3348e 351 NAPI_STATE_SCHED, /* Poll is scheduled */
39e6c820 352 NAPI_STATE_MISSED, /* reschedule a napi */
a0a46196 353 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 354 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
355 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
356 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
357 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
358};
359
360enum {
39e6c820
ED
361 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
362 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
363 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
364 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
365 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
366 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
367 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
368};
369
5b252f0c 370enum gro_result {
d1c76af9
HX
371 GRO_MERGED,
372 GRO_MERGED_FREE,
373 GRO_HELD,
374 GRO_NORMAL,
375 GRO_DROP,
25393d3f 376 GRO_CONSUMED,
d1c76af9 377};
5b252f0c 378typedef enum gro_result gro_result_t;
d1c76af9 379
8a4eb573
JP
380/*
381 * enum rx_handler_result - Possible return values for rx_handlers.
382 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
383 * further.
384 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
385 * case skb->dev was changed by rx_handler.
386 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 387 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
388 *
389 * rx_handlers are functions called from inside __netif_receive_skb(), to do
390 * special processing of the skb, prior to delivery to protocol handlers.
391 *
392 * Currently, a net_device can only have a single rx_handler registered. Trying
393 * to register a second rx_handler will return -EBUSY.
394 *
395 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
396 * To unregister a rx_handler on a net_device, use
397 * netdev_rx_handler_unregister().
398 *
399 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
400 * do with the skb.
401 *
5e82b4b2 402 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 403 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 404 * the skb to be delivered in some other way.
8a4eb573
JP
405 *
406 * If the rx_handler changed skb->dev, to divert the skb to another
407 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
408 * new device will be called if it exists.
409 *
5e82b4b2 410 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 411 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 412 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 413 *
5e82b4b2 414 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
415 * delivered, it should return RX_HANDLER_PASS.
416 *
417 * A device without a registered rx_handler will behave as if rx_handler
418 * returned RX_HANDLER_PASS.
419 */
420
421enum rx_handler_result {
422 RX_HANDLER_CONSUMED,
423 RX_HANDLER_ANOTHER,
424 RX_HANDLER_EXACT,
425 RX_HANDLER_PASS,
426};
427typedef enum rx_handler_result rx_handler_result_t;
428typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 429
f629d208 430void __napi_schedule(struct napi_struct *n);
bc9ad166 431void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 432
4d29515f 433static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
434{
435 return test_bit(NAPI_STATE_DISABLE, &n->state);
436}
437
39e6c820 438bool napi_schedule_prep(struct napi_struct *n);
bea3348e
SH
439
440/**
441 * napi_schedule - schedule NAPI poll
5e82b4b2 442 * @n: NAPI context
bea3348e
SH
443 *
444 * Schedule NAPI poll routine to be called if it is not already
445 * running.
446 */
447static inline void napi_schedule(struct napi_struct *n)
448{
449 if (napi_schedule_prep(n))
450 __napi_schedule(n);
451}
452
bc9ad166
ED
453/**
454 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 455 * @n: NAPI context
bc9ad166
ED
456 *
457 * Variant of napi_schedule(), assuming hard irqs are masked.
458 */
459static inline void napi_schedule_irqoff(struct napi_struct *n)
460{
461 if (napi_schedule_prep(n))
462 __napi_schedule_irqoff(n);
463}
464
bfe13f54 465/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 466static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
467{
468 if (napi_schedule_prep(napi)) {
469 __napi_schedule(napi);
4d29515f 470 return true;
bfe13f54 471 }
4d29515f 472 return false;
bfe13f54
RD
473}
474
364b6055 475bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
476/**
477 * napi_complete - NAPI processing complete
5e82b4b2 478 * @n: NAPI context
bea3348e
SH
479 *
480 * Mark NAPI processing as complete.
3b47d303 481 * Consider using napi_complete_done() instead.
364b6055 482 * Return false if device should avoid rearming interrupts.
bea3348e 483 */
364b6055 484static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
485{
486 return napi_complete_done(n, 0);
487}
bea3348e 488
af12fa6e
ET
489/**
490 * napi_hash_del - remove a NAPI from global table
5e82b4b2 491 * @napi: NAPI context
af12fa6e 492 *
5e82b4b2 493 * Warning: caller must observe RCU grace period
34cbe27e
ED
494 * before freeing memory containing @napi, if
495 * this function returns true.
93d05d4a 496 * Note: core networking stack automatically calls it
5e82b4b2 497 * from netif_napi_del().
93d05d4a 498 * Drivers might want to call this helper to combine all
5e82b4b2 499 * the needed RCU grace periods into a single one.
af12fa6e 500 */
34cbe27e 501bool napi_hash_del(struct napi_struct *napi);
af12fa6e 502
bea3348e
SH
503/**
504 * napi_disable - prevent NAPI from scheduling
5e82b4b2 505 * @n: NAPI context
bea3348e
SH
506 *
507 * Stop NAPI from being scheduled on this context.
508 * Waits till any outstanding processing completes.
509 */
3b47d303 510void napi_disable(struct napi_struct *n);
bea3348e
SH
511
512/**
513 * napi_enable - enable NAPI scheduling
5e82b4b2 514 * @n: NAPI context
bea3348e
SH
515 *
516 * Resume NAPI from being scheduled on this context.
517 * Must be paired with napi_disable.
518 */
519static inline void napi_enable(struct napi_struct *n)
520{
521 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 522 smp_mb__before_atomic();
bea3348e 523 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 524 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
525}
526
c264c3de
SH
527/**
528 * napi_synchronize - wait until NAPI is not running
5e82b4b2 529 * @n: NAPI context
c264c3de
SH
530 *
531 * Wait until NAPI is done being scheduled on this context.
532 * Waits till any outstanding processing completes but
533 * does not disable future activations.
534 */
535static inline void napi_synchronize(const struct napi_struct *n)
536{
facc432f
AB
537 if (IS_ENABLED(CONFIG_SMP))
538 while (test_bit(NAPI_STATE_SCHED, &n->state))
539 msleep(1);
540 else
541 barrier();
c264c3de 542}
c264c3de 543
6c5c9581
MK
544/**
545 * napi_if_scheduled_mark_missed - if napi is running, set the
546 * NAPIF_STATE_MISSED
547 * @n: NAPI context
548 *
549 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
550 * NAPI is scheduled.
551 **/
552static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
553{
554 unsigned long val, new;
555
556 do {
557 val = READ_ONCE(n->state);
558 if (val & NAPIF_STATE_DISABLE)
559 return true;
560
561 if (!(val & NAPIF_STATE_SCHED))
562 return false;
563
564 new = val | NAPIF_STATE_MISSED;
565 } while (cmpxchg(&n->state, val, new) != val);
566
567 return true;
568}
569
d94d9fee 570enum netdev_queue_state_t {
73466498
TH
571 __QUEUE_STATE_DRV_XOFF,
572 __QUEUE_STATE_STACK_XOFF,
c3f26a26 573 __QUEUE_STATE_FROZEN,
79d16385 574};
8e2f1a63
DB
575
576#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
577#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
578#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
579
580#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
581#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
582 QUEUE_STATE_FROZEN)
583#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
584 QUEUE_STATE_FROZEN)
585
73466498
TH
586/*
587 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
588 * netif_tx_* functions below are used to manipulate this flag. The
589 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
590 * queue independently. The netif_xmit_*stopped functions below are called
591 * to check if the queue has been stopped by the driver or stack (either
592 * of the XOFF bits are set in the state). Drivers should not need to call
593 * netif_xmit*stopped functions, they should only be using netif_tx_*.
594 */
79d16385 595
bb949fbd 596struct netdev_queue {
6a321cb3 597/*
5e82b4b2 598 * read-mostly part
6a321cb3 599 */
bb949fbd 600 struct net_device *dev;
46e5da40 601 struct Qdisc __rcu *qdisc;
b0e1e646 602 struct Qdisc *qdisc_sleeping;
ccf5ff69 603#ifdef CONFIG_SYSFS
1d24eb48
TH
604 struct kobject kobj;
605#endif
f2cd2d3e
ED
606#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
607 int numa_node;
608#endif
c0ef079c
FW
609 unsigned long tx_maxrate;
610 /*
611 * Number of TX timeouts for this queue
612 * (/sys/class/net/DEV/Q/trans_timeout)
613 */
614 unsigned long trans_timeout;
ffcfe25b
AD
615
616 /* Subordinate device that the queue has been assigned to */
617 struct net_device *sb_dev;
661b8d1b
MK
618#ifdef CONFIG_XDP_SOCKETS
619 struct xdp_umem *umem;
620#endif
6a321cb3 621/*
5e82b4b2 622 * write-mostly part
6a321cb3
ED
623 */
624 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
625 int xmit_lock_owner;
9d21493b 626 /*
9b36627a 627 * Time (in jiffies) of last Tx
9d21493b
ED
628 */
629 unsigned long trans_start;
ccf5ff69 630
114cf580
TH
631 unsigned long state;
632
633#ifdef CONFIG_BQL
634 struct dql dql;
635#endif
e8a0464c 636} ____cacheline_aligned_in_smp;
bb949fbd 637
79134e6c 638extern int sysctl_fb_tunnels_only_for_init_net;
856c395c 639extern int sysctl_devconf_inherit_init_net;
79134e6c
ED
640
641static inline bool net_has_fallback_tunnels(const struct net *net)
642{
be9fc097
AB
643 return net == &init_net ||
644 !IS_ENABLED(CONFIG_SYSCTL) ||
645 !sysctl_fb_tunnels_only_for_init_net;
79134e6c
ED
646}
647
f2cd2d3e
ED
648static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
649{
650#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
651 return q->numa_node;
652#else
b236da69 653 return NUMA_NO_NODE;
f2cd2d3e
ED
654#endif
655}
656
657static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
658{
659#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
660 q->numa_node = node;
661#endif
662}
663
df334545 664#ifdef CONFIG_RPS
0a9627f2
TH
665/*
666 * This structure holds an RPS map which can be of variable length. The
667 * map is an array of CPUs.
668 */
669struct rps_map {
670 unsigned int len;
671 struct rcu_head rcu;
bb4cf02d 672 u16 cpus[];
0a9627f2 673};
60b778ce 674#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 675
fec5e652 676/*
c445477d
BH
677 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
678 * tail pointer for that CPU's input queue at the time of last enqueue, and
679 * a hardware filter index.
fec5e652
TH
680 */
681struct rps_dev_flow {
682 u16 cpu;
c445477d 683 u16 filter;
fec5e652
TH
684 unsigned int last_qtail;
685};
c445477d 686#define RPS_NO_FILTER 0xffff
fec5e652
TH
687
688/*
689 * The rps_dev_flow_table structure contains a table of flow mappings.
690 */
691struct rps_dev_flow_table {
692 unsigned int mask;
693 struct rcu_head rcu;
bb4cf02d 694 struct rps_dev_flow flows[];
fec5e652
TH
695};
696#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 697 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
698
699/*
700 * The rps_sock_flow_table contains mappings of flows to the last CPU
701 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
702 * Each entry is a 32bit value. Upper part is the high-order bits
703 * of flow hash, lower part is CPU number.
567e4b79 704 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
705 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
706 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 707 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
708 */
709struct rps_sock_flow_table {
567e4b79 710 u32 mask;
93c1af6c 711
bb4cf02d 712 u32 ents[] ____cacheline_aligned_in_smp;
fec5e652 713};
567e4b79 714#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
715
716#define RPS_NO_CPU 0xffff
717
567e4b79
ED
718extern u32 rps_cpu_mask;
719extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
720
fec5e652
TH
721static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
722 u32 hash)
723{
724 if (table && hash) {
567e4b79
ED
725 unsigned int index = hash & table->mask;
726 u32 val = hash & ~rps_cpu_mask;
fec5e652 727
5e82b4b2 728 /* We only give a hint, preemption can change CPU under us */
567e4b79 729 val |= raw_smp_processor_id();
fec5e652 730
567e4b79
ED
731 if (table->ents[index] != val)
732 table->ents[index] = val;
fec5e652
TH
733 }
734}
735
c445477d 736#ifdef CONFIG_RFS_ACCEL
f629d208
JP
737bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
738 u16 filter_id);
c445477d 739#endif
a953be53 740#endif /* CONFIG_RPS */
c445477d 741
0a9627f2
TH
742/* This structure contains an instance of an RX queue. */
743struct netdev_rx_queue {
a953be53 744#ifdef CONFIG_RPS
6e3f7faf
ED
745 struct rps_map __rcu *rps_map;
746 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 747#endif
6e3f7faf 748 struct kobject kobj;
fe822240 749 struct net_device *dev;
e817f856 750 struct xdp_rxq_info xdp_rxq;
661b8d1b
MK
751#ifdef CONFIG_XDP_SOCKETS
752 struct xdp_umem *umem;
753#endif
0a9627f2 754} ____cacheline_aligned_in_smp;
a953be53
MD
755
756/*
757 * RX queue sysfs structures and functions.
758 */
759struct rx_queue_attribute {
760 struct attribute attr;
718ad681 761 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
a953be53 762 ssize_t (*store)(struct netdev_rx_queue *queue,
718ad681 763 const char *buf, size_t len);
a953be53 764};
d314774c 765
bf264145
TH
766#ifdef CONFIG_XPS
767/*
768 * This structure holds an XPS map which can be of variable length. The
769 * map is an array of queues.
770 */
771struct xps_map {
772 unsigned int len;
773 unsigned int alloc_len;
774 struct rcu_head rcu;
bb4cf02d 775 u16 queues[];
bf264145 776};
60b778ce 777#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
778#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
779 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
780
781/*
782 * This structure holds all XPS maps for device. Maps are indexed by CPU.
783 */
784struct xps_dev_maps {
785 struct rcu_head rcu;
bb4cf02d 786 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
bf264145 787};
80d19669
AN
788
789#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
184c449f 790 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
80d19669
AN
791
792#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
793 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
794
bf264145
TH
795#endif /* CONFIG_XPS */
796
4f57c087
JF
797#define TC_MAX_QUEUE 16
798#define TC_BITMASK 15
799/* HW offloaded queuing disciplines txq count and offset maps */
800struct netdev_tc_txq {
801 u16 count;
802 u16 offset;
803};
804
68bad94e
NP
805#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
806/*
807 * This structure is to hold information about the device
808 * configured to run FCoE protocol stack.
809 */
810struct netdev_fcoe_hbainfo {
811 char manufacturer[64];
812 char serial_number[64];
813 char hardware_version[64];
814 char driver_version[64];
815 char optionrom_version[64];
816 char firmware_version[64];
817 char model[256];
818 char model_description[256];
819};
820#endif
821
02637fce 822#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 823
02637fce
JP
824/* This structure holds a unique identifier to identify some
825 * physical item (port for example) used by a netdevice.
66b52b0d 826 */
02637fce
JP
827struct netdev_phys_item_id {
828 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
829 unsigned char id_len;
830};
831
d754f98b
SF
832static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
833 struct netdev_phys_item_id *b)
834{
835 return a->id_len == b->id_len &&
836 memcmp(a->id, b->id, a->id_len) == 0;
837}
838
99932d4f 839typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
8ec56fc3
AD
840 struct sk_buff *skb,
841 struct net_device *sb_dev);
99932d4f 842
2572ac53 843enum tc_setup_type {
575ed7d3 844 TC_SETUP_QDISC_MQPRIO,
a1b7c5fd 845 TC_SETUP_CLSU32,
5b33f488 846 TC_SETUP_CLSFLOWER,
ade9b658 847 TC_SETUP_CLSMATCHALL,
332ae8e2 848 TC_SETUP_CLSBPF,
8c4083b3 849 TC_SETUP_BLOCK,
8521db4c 850 TC_SETUP_QDISC_CBS,
602f3baf 851 TC_SETUP_QDISC_RED,
7fdb61b4 852 TC_SETUP_QDISC_PRIO,
f971b132 853 TC_SETUP_QDISC_MQ,
25db26a9 854 TC_SETUP_QDISC_ETF,
98b0e5f6 855 TC_SETUP_ROOT_QDISC,
890d8d23 856 TC_SETUP_QDISC_GRED,
9c66d156 857 TC_SETUP_QDISC_TAPRIO,
c29f74e0 858 TC_SETUP_FT,
d35eb52b 859 TC_SETUP_QDISC_ETS,
ef6aadcc 860 TC_SETUP_QDISC_TBF,
aaca9408 861 TC_SETUP_QDISC_FIFO,
16e5cc64
JF
862};
863
f4e63525
JK
864/* These structures hold the attributes of bpf state that are being passed
865 * to the netdevice through the bpf op.
a7862b45 866 */
f4e63525 867enum bpf_netdev_command {
a7862b45
BB
868 /* Set or clear a bpf program used in the earliest stages of packet
869 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
870 * is responsible for calling bpf_prog_put on any old progs that are
871 * stored. In case of error, the callee need not release the new prog
872 * reference, but on success it takes ownership and must bpf_prog_put
873 * when it is no longer used.
874 */
875 XDP_SETUP_PROG,
ee5d032f 876 XDP_SETUP_PROG_HW,
a7862b45 877 XDP_QUERY_PROG,
a25717d2 878 XDP_QUERY_PROG_HW,
ab3f0063 879 /* BPF program for offload callbacks, invoked at program load time. */
a3884572
JK
880 BPF_OFFLOAD_MAP_ALLOC,
881 BPF_OFFLOAD_MAP_FREE,
74515c57 882 XDP_SETUP_XSK_UMEM,
a7862b45
BB
883};
884
cae1927c 885struct bpf_prog_offload_ops;
ddf9f970 886struct netlink_ext_ack;
74515c57 887struct xdp_umem;
75ccae62 888struct xdp_dev_bulk_queue;
ddf9f970 889
f4e63525
JK
890struct netdev_bpf {
891 enum bpf_netdev_command command;
a7862b45
BB
892 union {
893 /* XDP_SETUP_PROG */
ddf9f970 894 struct {
32d60277 895 u32 flags;
ddf9f970
JK
896 struct bpf_prog *prog;
897 struct netlink_ext_ack *extack;
898 };
a25717d2 899 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
58038695 900 struct {
58038695 901 u32 prog_id;
92f0292b
JK
902 /* flags with which program was installed */
903 u32 prog_flags;
58038695 904 };
a3884572
JK
905 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
906 struct {
907 struct bpf_offloaded_map *offmap;
908 };
f8ebfaf6 909 /* XDP_SETUP_XSK_UMEM */
74515c57 910 struct {
f8ebfaf6
JS
911 struct xdp_umem *umem;
912 u16 queue_id;
74515c57 913 } xsk;
a7862b45
BB
914 };
915};
16e5cc64 916
9116e5e2
MK
917/* Flags for ndo_xsk_wakeup. */
918#define XDP_WAKEUP_RX (1 << 0)
919#define XDP_WAKEUP_TX (1 << 1)
920
d77e38e6
SK
921#ifdef CONFIG_XFRM_OFFLOAD
922struct xfrmdev_ops {
923 int (*xdo_dev_state_add) (struct xfrm_state *x);
924 void (*xdo_dev_state_delete) (struct xfrm_state *x);
925 void (*xdo_dev_state_free) (struct xfrm_state *x);
926 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
927 struct xfrm_state *x);
50bd870a 928 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
d77e38e6
SK
929};
930#endif
931
6c557001
FW
932struct dev_ifalias {
933 struct rcu_head rcuhead;
934 char ifalias[];
935};
936
b473b0d2 937struct devlink;
da68b4ad 938struct tlsdev_ops;
b473b0d2 939
ff927412
JP
940struct netdev_name_node {
941 struct hlist_node hlist;
36fbf1e5 942 struct list_head list;
ff927412
JP
943 struct net_device *dev;
944 const char *name;
945};
946
36fbf1e5
JP
947int netdev_name_node_alt_create(struct net_device *dev, const char *name);
948int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
949
93642e14
JP
950struct netdev_net_notifier {
951 struct list_head list;
952 struct notifier_block *nb;
953};
954
d314774c
SH
955/*
956 * This structure defines the management hooks for network devices.
00829823
SH
957 * The following hooks can be defined; unless noted otherwise, they are
958 * optional and can be filled with a null pointer.
d314774c
SH
959 *
960 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
961 * This function is called once when a network device is registered.
962 * The network device can use this for any late stage initialization
963 * or semantic validation. It can fail with an error code which will
964 * be propagated back to register_netdev.
d314774c
SH
965 *
966 * void (*ndo_uninit)(struct net_device *dev);
967 * This function is called when device is unregistered or when registration
968 * fails. It is not called if init fails.
969 *
970 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 971 * This function is called when a network device transitions to the up
d314774c
SH
972 * state.
973 *
974 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 975 * This function is called when a network device transitions to the down
d314774c
SH
976 * state.
977 *
dc1f8bf6
SH
978 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
979 * struct net_device *dev);
00829823 980 * Called when a packet needs to be transmitted.
e79d8429
RR
981 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
982 * the queue before that can happen; it's for obsolete devices and weird
983 * corner cases, but the stack really does a non-trivial amount
984 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 985 * Required; cannot be NULL.
00829823 986 *
1a2a1444
DM
987 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
988 * struct net_device *dev
989 * netdev_features_t features);
990 * Called by core transmit path to determine if device is capable of
991 * performing offload operations on a given packet. This is to give
992 * the device an opportunity to implement any restrictions that cannot
993 * be otherwise expressed by feature flags. The check is called with
994 * the set of features that the stack has calculated and it returns
995 * those the driver believes to be appropriate.
cdba756f 996 *
f663dd9a 997 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
a350ecce 998 * struct net_device *sb_dev);
5e82b4b2 999 * Called to decide which queue to use when device supports multiple
00829823
SH
1000 * transmit queues.
1001 *
d314774c
SH
1002 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1003 * This function is called to allow device receiver to make
5e82b4b2 1004 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
1005 *
1006 * void (*ndo_set_rx_mode)(struct net_device *dev);
1007 * This function is called device changes address list filtering.
01789349 1008 * If driver handles unicast address filtering, it should set
5e82b4b2 1009 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
1010 *
1011 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1012 * This function is called when the Media Access Control address
37b607c5 1013 * needs to be changed. If this interface is not defined, the
5e82b4b2 1014 * MAC address can not be changed.
d314774c
SH
1015 *
1016 * int (*ndo_validate_addr)(struct net_device *dev);
1017 * Test if Media Access Control address is valid for the device.
1018 *
1019 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
1020 * Called when a user requests an ioctl which can't be handled by
1021 * the generic interface code. If not defined ioctls return
d314774c
SH
1022 * not supported error code.
1023 *
1024 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1025 * Used to set network devices bus interface parameters. This interface
5e82b4b2 1026 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
1027 * interface (PCI) for low level management.
1028 *
1029 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1030 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 1031 * of a device.
d314774c 1032 *
0290bd29 1033 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
5e82b4b2 1034 * Callback used when the transmitter has not made any progress
d314774c
SH
1035 * for dev->watchdog ticks.
1036 *
bc1f4470 1037 * void (*ndo_get_stats64)(struct net_device *dev,
1038 * struct rtnl_link_stats64 *storage);
d308e38f 1039 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 1040 * Called when a user wants to get the network device usage
be1f3c2c 1041 * statistics. Drivers must do one of the following:
3cfde79c
BH
1042 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1043 * rtnl_link_stats64 structure passed by the caller.
82695d9b 1044 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
1045 * (which should normally be dev->stats) and return a pointer to
1046 * it. The structure may be changed asynchronously only if each
1047 * field is written atomically.
1048 * 3. Update dev->stats asynchronously and atomically, and define
1049 * neither operation.
d314774c 1050 *
3df5b3c6 1051 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
1052 * Return true if this device supports offload stats of this attr_id.
1053 *
1054 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1055 * void *attr_data)
1056 * Get statistics for offload operations by attr_id. Write it into the
1057 * attr_data pointer.
1058 *
5d632cb7 1059 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 1060 * If device supports VLAN filtering this function is called when a
80d5c368 1061 * VLAN id is registered.
d314774c 1062 *
5d632cb7 1063 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 1064 * If device supports VLAN filtering this function is called when a
80d5c368 1065 * VLAN id is unregistered.
d314774c
SH
1066 *
1067 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
1068 *
1069 * SR-IOV management functions.
1070 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
1071 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1072 * u8 qos, __be16 proto);
ed616689
SC
1073 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1074 * int max_tx_rate);
5f8444a3 1075 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 1076 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
1077 * int (*ndo_get_vf_config)(struct net_device *dev,
1078 * int vf, struct ifla_vf_info *ivf);
1d8faf48 1079 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
1080 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1081 * struct nlattr *port[]);
01a3d796
VZ
1082 *
1083 * Enable or disable the VF ability to query its RSS Redirection Table and
1084 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 1085 * with PF and querying it may introduce a theoretical security risk.
01a3d796 1086 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 1087 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
2572ac53 1088 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
de4784ca 1089 * void *type_data);
6a4bc2b4
FF
1090 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1091 * This is always called from the stack with the rtnl lock held and netif
1092 * tx queues stopped. This allows the netdevice to perform queue
1093 * management safely.
c445477d 1094 *
e9bce845
YZ
1095 * Fiber Channel over Ethernet (FCoE) offload functions.
1096 * int (*ndo_fcoe_enable)(struct net_device *dev);
1097 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1098 * so the underlying device can perform whatever needed configuration or
1099 * initialization to support acceleration of FCoE traffic.
1100 *
1101 * int (*ndo_fcoe_disable)(struct net_device *dev);
1102 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1103 * so the underlying device can perform whatever needed clean-ups to
1104 * stop supporting acceleration of FCoE traffic.
1105 *
1106 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1107 * struct scatterlist *sgl, unsigned int sgc);
1108 * Called when the FCoE Initiator wants to initialize an I/O that
1109 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1110 * perform necessary setup and returns 1 to indicate the device is set up
1111 * successfully to perform DDP on this I/O, otherwise this returns 0.
1112 *
1113 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1114 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1115 * indicated by the FC exchange id 'xid', so the underlying device can
1116 * clean up and reuse resources for later DDP requests.
1117 *
1118 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1119 * struct scatterlist *sgl, unsigned int sgc);
1120 * Called when the FCoE Target wants to initialize an I/O that
1121 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1122 * perform necessary setup and returns 1 to indicate the device is set up
1123 * successfully to perform DDP on this I/O, otherwise this returns 0.
1124 *
68bad94e
NP
1125 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1126 * struct netdev_fcoe_hbainfo *hbainfo);
1127 * Called when the FCoE Protocol stack wants information on the underlying
1128 * device. This information is utilized by the FCoE protocol stack to
1129 * register attributes with Fiber Channel management service as per the
1130 * FC-GS Fabric Device Management Information(FDMI) specification.
1131 *
e9bce845
YZ
1132 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1133 * Called when the underlying device wants to override default World Wide
1134 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1135 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1136 * protocol stack to use.
1137 *
c445477d
BH
1138 * RFS acceleration.
1139 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1140 * u16 rxq_index, u32 flow_id);
1141 * Set hardware filter for RFS. rxq_index is the target queue index;
1142 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1143 * Return the filter ID on success, or a negative error code.
fbaec0ea 1144 *
8b98a70c 1145 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1146 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1147 * Called to make another netdev an underling.
1148 *
1149 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1150 * Called to release previously enslaved netdev.
5455c699 1151 *
cff9f12b
MG
1152 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1153 * struct sk_buff *skb,
1154 * bool all_slaves);
1155 * Get the xmit slave of master device. If all_slaves is true, function
1156 * assume all the slaves can transmit.
1157 *
5455c699 1158 * Feature/offload setting functions.
1a2a1444
DM
1159 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1160 * netdev_features_t features);
1161 * Adjusts the requested feature flags according to device-specific
1162 * constraints, and returns the resulting flags. Must not modify
1163 * the device state.
1164 *
c8f44aff 1165 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1166 * Called to update device configuration to new features. Passed
1167 * feature set might be less than what was returned by ndo_fix_features()).
1168 * Must return >0 or -errno if it changed dev->features itself.
1169 *
edc7d573 1170 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1171 * struct net_device *dev,
87b0984e
PM
1172 * const unsigned char *addr, u16 vid, u16 flags,
1173 * struct netlink_ext_ack *extack);
77162022 1174 * Adds an FDB entry to dev for addr.
1690be63
VY
1175 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1176 * struct net_device *dev,
f6f6424b 1177 * const unsigned char *addr, u16 vid)
77162022
JF
1178 * Deletes the FDB entry from dev coresponding to addr.
1179 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1180 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1181 * int *idx)
77162022
JF
1182 * Used to add FDB entries to dump requests. Implementers should add
1183 * entries to skb and update idx with the number of entries.
e5a55a89 1184 *
ad41faa8 1185 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
2fd527b7 1186 * u16 flags, struct netlink_ext_ack *extack)
e5a55a89 1187 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1188 * struct net_device *dev, u32 filter_mask,
1189 * int nlflags)
ad41faa8
ND
1190 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1191 * u16 flags);
4bf84c35
JP
1192 *
1193 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1194 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1195 * which do not represent real hardware may define this to allow their
1196 * userspace components to manage their virtual carrier state. Devices
1197 * that determine carrier state from physical hardware properties (eg
1198 * network cables) or protocol-dependent mechanisms (eg
1199 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1200 *
1201 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1202 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1203 * Called to get ID of physical port of this device. If driver does
1204 * not implement this, it is assumed that the hw is not able to have
1205 * multiple net devices on single physical port.
53cf5275 1206 *
d6abc596
FF
1207 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1208 * struct netdev_phys_item_id *ppid)
1209 * Called to get the parent ID of the physical port of this device.
1210 *
7c46a640
AD
1211 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1212 * struct udp_tunnel_info *ti);
1213 * Called by UDP tunnel to notify a driver about the UDP port and socket
1214 * address family that a UDP tunnel is listnening to. It is called only
1215 * when a new port starts listening. The operation is protected by the
1216 * RTNL.
1217 *
1218 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1219 * struct udp_tunnel_info *ti);
1220 * Called by UDP tunnel to notify the driver about a UDP port and socket
1221 * address family that the UDP tunnel is not listening to anymore. The
1222 * operation is protected by the RTNL.
1223 *
a6cc0cfa
JF
1224 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1225 * struct net_device *dev)
1226 * Called by upper layer devices to accelerate switching or other
1227 * station functionality into hardware. 'pdev is the lowerdev
1228 * to use for the offload and 'dev' is the net device that will
1229 * back the offload. Returns a pointer to the private structure
1230 * the upper layer will maintain.
1231 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1232 * Called by upper layer device to delete the station created
1233 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1234 * the station and priv is the structure returned by the add
1235 * operation.
822b3b2e
JF
1236 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1237 * int queue_index, u32 maxrate);
1238 * Called when a user wants to set a max-rate limitation of specific
1239 * TX queue.
a54acb3a
ND
1240 * int (*ndo_get_iflink)(const struct net_device *dev);
1241 * Called to get the iflink value of this device.
d746d707 1242 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1243 * bool proto_down);
d746d707
AK
1244 * This function is used to pass protocol port error state information
1245 * to the switch driver. The switch driver can react to the proto_down
1246 * by doing a phys down on the associated switch port.
fc4099f1
PS
1247 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1248 * This function is used to get egress tunnel information for given skb.
1249 * This is useful for retrieving outer tunnel header parameters while
1250 * sampling packet.
871b642a
PA
1251 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1252 * This function is used to specify the headroom that the skb must
1253 * consider when allocation skb during packet reception. Setting
1254 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1255 * forward. Setting a negative value resets the rx headroom to the
871b642a 1256 * default value.
f4e63525 1257 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
a7862b45 1258 * This function is used to set or query state related to XDP on the
f4e63525
JK
1259 * netdevice and manage BPF offload. See definition of
1260 * enum bpf_netdev_command for details.
42b33468
JDB
1261 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1262 * u32 flags);
735fc405
JDB
1263 * This function is used to submit @n XDP packets for transmit on a
1264 * netdevice. Returns number of frames successfully transmitted, frames
1265 * that got dropped are freed/returned via xdp_return_frame().
1266 * Returns negative number, means general error invoking ndo, meaning
1267 * no frames were xmit'ed and core-caller will free all frames.
9116e5e2
MK
1268 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1269 * This function is used to wake up the softirq, ksoftirqd or kthread
1270 * responsible for sending and/or receiving packets on a specific
1271 * queue id bound to an AF_XDP socket. The flags field specifies if
1272 * only RX, only Tx, or both should be woken up using the flags
1273 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
5dc37bb9
JP
1274 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1275 * Get devlink port instance associated with a given netdev.
b473b0d2
JK
1276 * Called with a reference on the netdevice and devlink locks only,
1277 * rtnl_lock is not held.
607259a6
CH
1278 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1279 * int cmd);
1280 * Add, change, delete or get information on an IPv4 tunnel.
d314774c
SH
1281 */
1282struct net_device_ops {
1283 int (*ndo_init)(struct net_device *dev);
1284 void (*ndo_uninit)(struct net_device *dev);
1285 int (*ndo_open)(struct net_device *dev);
1286 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1287 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1288 struct net_device *dev);
1289 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1290 struct net_device *dev,
1291 netdev_features_t features);
00829823 1292 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1293 struct sk_buff *skb,
a350ecce 1294 struct net_device *sb_dev);
d314774c
SH
1295 void (*ndo_change_rx_flags)(struct net_device *dev,
1296 int flags);
d314774c 1297 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1298 int (*ndo_set_mac_address)(struct net_device *dev,
1299 void *addr);
d314774c 1300 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1301 int (*ndo_do_ioctl)(struct net_device *dev,
1302 struct ifreq *ifr, int cmd);
d314774c
SH
1303 int (*ndo_set_config)(struct net_device *dev,
1304 struct ifmap *map);
00829823
SH
1305 int (*ndo_change_mtu)(struct net_device *dev,
1306 int new_mtu);
1307 int (*ndo_neigh_setup)(struct net_device *dev,
1308 struct neigh_parms *);
0290bd29
MT
1309 void (*ndo_tx_timeout) (struct net_device *dev,
1310 unsigned int txqueue);
d314774c 1311
bc1f4470 1312 void (*ndo_get_stats64)(struct net_device *dev,
1313 struct rtnl_link_stats64 *storage);
3df5b3c6 1314 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1315 int (*ndo_get_offload_stats)(int attr_id,
1316 const struct net_device *dev,
1317 void *attr_data);
d314774c
SH
1318 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1319
8e586137 1320 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1321 __be16 proto, u16 vid);
8e586137 1322 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1323 __be16 proto, u16 vid);
d314774c 1324#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1325 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1326 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1327 struct netpoll_info *info);
0e34e931 1328 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1329#endif
95c26df8
WM
1330 int (*ndo_set_vf_mac)(struct net_device *dev,
1331 int queue, u8 *mac);
1332 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1333 int queue, u16 vlan,
1334 u8 qos, __be16 proto);
ed616689
SC
1335 int (*ndo_set_vf_rate)(struct net_device *dev,
1336 int vf, int min_tx_rate,
1337 int max_tx_rate);
5f8444a3
GR
1338 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1339 int vf, bool setting);
dd461d6a
HS
1340 int (*ndo_set_vf_trust)(struct net_device *dev,
1341 int vf, bool setting);
95c26df8
WM
1342 int (*ndo_get_vf_config)(struct net_device *dev,
1343 int vf,
1344 struct ifla_vf_info *ivf);
1d8faf48
RE
1345 int (*ndo_set_vf_link_state)(struct net_device *dev,
1346 int vf, int link_state);
3b766cd8
EBE
1347 int (*ndo_get_vf_stats)(struct net_device *dev,
1348 int vf,
1349 struct ifla_vf_stats
1350 *vf_stats);
57b61080
SF
1351 int (*ndo_set_vf_port)(struct net_device *dev,
1352 int vf,
1353 struct nlattr *port[]);
1354 int (*ndo_get_vf_port)(struct net_device *dev,
1355 int vf, struct sk_buff *skb);
30aad417
DG
1356 int (*ndo_get_vf_guid)(struct net_device *dev,
1357 int vf,
1358 struct ifla_vf_guid *node_guid,
1359 struct ifla_vf_guid *port_guid);
cc8e27cc
EC
1360 int (*ndo_set_vf_guid)(struct net_device *dev,
1361 int vf, u64 guid,
1362 int guid_type);
01a3d796
VZ
1363 int (*ndo_set_vf_rss_query_en)(
1364 struct net_device *dev,
1365 int vf, bool setting);
16e5cc64 1366 int (*ndo_setup_tc)(struct net_device *dev,
2572ac53 1367 enum tc_setup_type type,
de4784ca 1368 void *type_data);
d11ead75 1369#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1370 int (*ndo_fcoe_enable)(struct net_device *dev);
1371 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1372 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1373 u16 xid,
1374 struct scatterlist *sgl,
1375 unsigned int sgc);
1376 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1377 u16 xid);
6247e086
YZ
1378 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1379 u16 xid,
1380 struct scatterlist *sgl,
1381 unsigned int sgc);
68bad94e
NP
1382 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1383 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1384#endif
1385
d11ead75 1386#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1387#define NETDEV_FCOE_WWNN 0
1388#define NETDEV_FCOE_WWPN 1
1389 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1390 u64 *wwn, int type);
4d288d57 1391#endif
3c9c36bc 1392
c445477d
BH
1393#ifdef CONFIG_RFS_ACCEL
1394 int (*ndo_rx_flow_steer)(struct net_device *dev,
1395 const struct sk_buff *skb,
1396 u16 rxq_index,
1397 u32 flow_id);
1398#endif
fbaec0ea 1399 int (*ndo_add_slave)(struct net_device *dev,
33eaf2a6
DA
1400 struct net_device *slave_dev,
1401 struct netlink_ext_ack *extack);
fbaec0ea
JP
1402 int (*ndo_del_slave)(struct net_device *dev,
1403 struct net_device *slave_dev);
cff9f12b
MG
1404 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1405 struct sk_buff *skb,
1406 bool all_slaves);
c8f44aff
MM
1407 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1408 netdev_features_t features);
5455c699 1409 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1410 netdev_features_t features);
503eebc2
JP
1411 int (*ndo_neigh_construct)(struct net_device *dev,
1412 struct neighbour *n);
1413 void (*ndo_neigh_destroy)(struct net_device *dev,
1414 struct neighbour *n);
77162022
JF
1415
1416 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1417 struct nlattr *tb[],
77162022 1418 struct net_device *dev,
6b6e2725 1419 const unsigned char *addr,
f6f6424b 1420 u16 vid,
87b0984e
PM
1421 u16 flags,
1422 struct netlink_ext_ack *extack);
77162022 1423 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1424 struct nlattr *tb[],
77162022 1425 struct net_device *dev,
f6f6424b
JP
1426 const unsigned char *addr,
1427 u16 vid);
77162022
JF
1428 int (*ndo_fdb_dump)(struct sk_buff *skb,
1429 struct netlink_callback *cb,
1430 struct net_device *dev,
5d5eacb3 1431 struct net_device *filter_dev,
d297653d 1432 int *idx);
5b2f94b2
RP
1433 int (*ndo_fdb_get)(struct sk_buff *skb,
1434 struct nlattr *tb[],
1435 struct net_device *dev,
1436 const unsigned char *addr,
1437 u16 vid, u32 portid, u32 seq,
1438 struct netlink_ext_ack *extack);
e5a55a89 1439 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3 1440 struct nlmsghdr *nlh,
2fd527b7
PM
1441 u16 flags,
1442 struct netlink_ext_ack *extack);
e5a55a89
JF
1443 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1444 u32 pid, u32 seq,
6cbdceeb 1445 struct net_device *dev,
46c264da
ND
1446 u32 filter_mask,
1447 int nlflags);
407af329 1448 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1449 struct nlmsghdr *nlh,
1450 u16 flags);
4bf84c35
JP
1451 int (*ndo_change_carrier)(struct net_device *dev,
1452 bool new_carrier);
66b52b0d 1453 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1454 struct netdev_phys_item_id *ppid);
d6abc596
FF
1455 int (*ndo_get_port_parent_id)(struct net_device *dev,
1456 struct netdev_phys_item_id *ppid);
db24a904
DA
1457 int (*ndo_get_phys_port_name)(struct net_device *dev,
1458 char *name, size_t len);
7c46a640
AD
1459 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1460 struct udp_tunnel_info *ti);
1461 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1462 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1463 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1464 struct net_device *dev);
1465 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1466 void *priv);
1467
822b3b2e
JF
1468 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1469 int queue_index,
1470 u32 maxrate);
a54acb3a 1471 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1472 int (*ndo_change_proto_down)(struct net_device *dev,
1473 bool proto_down);
fc4099f1
PS
1474 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1475 struct sk_buff *skb);
871b642a
PA
1476 void (*ndo_set_rx_headroom)(struct net_device *dev,
1477 int needed_headroom);
f4e63525
JK
1478 int (*ndo_bpf)(struct net_device *dev,
1479 struct netdev_bpf *bpf);
735fc405 1480 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
42b33468
JDB
1481 struct xdp_frame **xdp,
1482 u32 flags);
9116e5e2
MK
1483 int (*ndo_xsk_wakeup)(struct net_device *dev,
1484 u32 queue_id, u32 flags);
5dc37bb9 1485 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
607259a6
CH
1486 int (*ndo_tunnel_ctl)(struct net_device *dev,
1487 struct ip_tunnel_parm *p, int cmd);
d314774c
SH
1488};
1489
7aa98047
LR
1490/**
1491 * enum net_device_priv_flags - &struct net_device priv_flags
1492 *
1493 * These are the &struct net_device, they are only set internally
1494 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1495 * userspace; this means that the order of these flags can change
7aa98047
LR
1496 * during any kernel release.
1497 *
1498 * You should have a pretty good reason to be extending these flags.
1499 *
1500 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1501 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1502 * @IFF_BONDING: bonding master or slave
7aa98047 1503 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1504 * @IFF_WAN_HDLC: WAN HDLC device
1505 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1506 * release skb->dst
1507 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1508 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1509 * @IFF_MACVLAN_PORT: device used as macvlan port
1510 * @IFF_BRIDGE_PORT: device used as bridge port
1511 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1512 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1513 * @IFF_UNICAST_FLT: Supports unicast filtering
1514 * @IFF_TEAM_PORT: device used as team port
1515 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1516 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1517 * change when it's running
1518 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1519 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1520 * underlying stacked devices
007979ea 1521 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1522 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1523 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1524 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1525 * @IFF_TEAM: device is a team device
d4ab4286 1526 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1527 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1528 * entity (i.e. the master device for bridged veth)
3c175784 1529 * @IFF_MACSEC: device is a MACsec device
f5426250 1530 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
30c8bd5a
SS
1531 * @IFF_FAILOVER: device is a failover master device
1532 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
d5256083 1533 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
8065a779 1534 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
7aa98047
LR
1535 */
1536enum netdev_priv_flags {
1537 IFF_802_1Q_VLAN = 1<<0,
1538 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1539 IFF_BONDING = 1<<2,
1540 IFF_ISATAP = 1<<3,
1541 IFF_WAN_HDLC = 1<<4,
1542 IFF_XMIT_DST_RELEASE = 1<<5,
1543 IFF_DONT_BRIDGE = 1<<6,
1544 IFF_DISABLE_NETPOLL = 1<<7,
1545 IFF_MACVLAN_PORT = 1<<8,
1546 IFF_BRIDGE_PORT = 1<<9,
1547 IFF_OVS_DATAPATH = 1<<10,
1548 IFF_TX_SKB_SHARING = 1<<11,
1549 IFF_UNICAST_FLT = 1<<12,
1550 IFF_TEAM_PORT = 1<<13,
1551 IFF_SUPP_NOFCS = 1<<14,
1552 IFF_LIVE_ADDR_CHANGE = 1<<15,
1553 IFF_MACVLAN = 1<<16,
1554 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1ec54cb4
PA
1555 IFF_L3MDEV_MASTER = 1<<18,
1556 IFF_NO_QUEUE = 1<<19,
1557 IFF_OPENVSWITCH = 1<<20,
1558 IFF_L3MDEV_SLAVE = 1<<21,
1559 IFF_TEAM = 1<<22,
1560 IFF_RXFH_CONFIGURED = 1<<23,
1561 IFF_PHONY_HEADROOM = 1<<24,
1562 IFF_MACSEC = 1<<25,
f5426250 1563 IFF_NO_RX_HANDLER = 1<<26,
30c8bd5a
SS
1564 IFF_FAILOVER = 1<<27,
1565 IFF_FAILOVER_SLAVE = 1<<28,
d5256083 1566 IFF_L3MDEV_RX_HANDLER = 1<<29,
8065a779 1567 IFF_LIVE_RENAME_OK = 1<<30,
7aa98047
LR
1568};
1569
1570#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1571#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1572#define IFF_BONDING IFF_BONDING
7aa98047 1573#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1574#define IFF_WAN_HDLC IFF_WAN_HDLC
1575#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1576#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1577#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1578#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1579#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1580#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1581#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1582#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1583#define IFF_TEAM_PORT IFF_TEAM_PORT
1584#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1585#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1586#define IFF_MACVLAN IFF_MACVLAN
02875878 1587#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
007979ea 1588#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1589#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1590#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1591#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1592#define IFF_TEAM IFF_TEAM
d4ab4286 1593#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1594#define IFF_MACSEC IFF_MACSEC
f5426250 1595#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
30c8bd5a
SS
1596#define IFF_FAILOVER IFF_FAILOVER
1597#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
d5256083 1598#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
8065a779 1599#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
7aa98047 1600
536721b1
KK
1601/**
1602 * struct net_device - The DEVICE structure.
d651983d
MCC
1603 *
1604 * Actually, this whole structure is a big mistake. It mixes I/O
1605 * data with strictly "high-level" data, and it has to know about
1606 * almost every data structure used in the INET module.
536721b1
KK
1607 *
1608 * @name: This is the first field of the "visible" part of this structure
1609 * (i.e. as seen by users in the "Space.c" file). It is the name
d651983d 1610 * of the interface.
536721b1 1611 *
ff927412 1612 * @name_node: Name hashlist node
536721b1
KK
1613 * @ifalias: SNMP alias
1614 * @mem_end: Shared memory end
1615 * @mem_start: Shared memory start
1616 * @base_addr: Device I/O address
1617 * @irq: Device IRQ number
1618 *
1619 * @state: Generic network queuing layer state, see netdev_state_t
1620 * @dev_list: The global list of network devices
5e82b4b2
BH
1621 * @napi_list: List entry used for polling NAPI devices
1622 * @unreg_list: List entry when we are unregistering the
1623 * device; see the function unregister_netdev
1624 * @close_list: List entry used when we are closing the device
62d885fe
BP
1625 * @ptype_all: Device-specific packet handlers for all protocols
1626 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1627 *
1628 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1629 * @features: Currently active device features
1630 * @hw_features: User-changeable features
1631 *
1632 * @wanted_features: User-requested features
1633 * @vlan_features: Mask of features inheritable by VLAN devices
1634 *
1635 * @hw_enc_features: Mask of features inherited by encapsulating devices
1636 * This field indicates what encapsulation
1637 * offloads the hardware is capable of doing,
1638 * and drivers will need to set them appropriately.
1639 *
1640 * @mpls_features: Mask of features inheritable by MPLS
a1fa83bd 1641 * @gso_partial_features: value(s) from NETIF_F_GSO\*
536721b1
KK
1642 *
1643 * @ifindex: interface index
5e82b4b2 1644 * @group: The group the device belongs to
536721b1
KK
1645 *
1646 * @stats: Statistics struct, which was left as a legacy, use
1647 * rtnl_link_stats64 instead
1648 *
1649 * @rx_dropped: Dropped packets by core network,
1650 * do not use this in drivers
1651 * @tx_dropped: Dropped packets by core network,
1652 * do not use this in drivers
6e7333d3
JW
1653 * @rx_nohandler: nohandler dropped packets by core network on
1654 * inactive devices, do not use this in drivers
9e55e5d3
FF
1655 * @carrier_up_count: Number of times the carrier has been up
1656 * @carrier_down_count: Number of times the carrier has been down
536721b1 1657 *
536721b1
KK
1658 * @wireless_handlers: List of functions to handle Wireless Extensions,
1659 * instead of ioctl,
1660 * see <net/iw_handler.h> for details.
1661 * @wireless_data: Instance data managed by the core of wireless extensions
1662 *
1663 * @netdev_ops: Includes several pointers to callbacks,
1664 * if one wants to override the ndo_*() functions
1665 * @ethtool_ops: Management operations
a1fa83bd 1666 * @l3mdev_ops: Layer 3 master device operations
f997c55c
AA
1667 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1668 * discovery handling. Necessary for e.g. 6LoWPAN.
a1fa83bd
RD
1669 * @xfrmdev_ops: Transformation offload operations
1670 * @tlsdev_ops: Transport Layer Security offload operations
d476059e 1671 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1672 * of Layer 2 headers.
1673 *
1674 * @flags: Interface flags (a la BSD)
1675 * @priv_flags: Like 'flags' but invisible to userspace,
1676 * see if.h for the definitions
1677 * @gflags: Global flags ( kept as legacy )
1678 * @padded: How much padding added by alloc_netdev()
1679 * @operstate: RFC2863 operstate
1680 * @link_mode: Mapping policy to operstate
1681 * @if_port: Selectable AUI, TP, ...
1682 * @dma: DMA channel
1683 * @mtu: Interface MTU value
61e84623
JW
1684 * @min_mtu: Interface Minimum MTU value
1685 * @max_mtu: Interface Maximum MTU value
536721b1 1686 * @type: Interface hardware type
2793a23a 1687 * @hard_header_len: Maximum hardware header length.
217e6fa2 1688 * @min_header_len: Minimum hardware header length
536721b1
KK
1689 *
1690 * @needed_headroom: Extra headroom the hardware may need, but not in all
1691 * cases can this be guaranteed
1692 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1693 * cases can this be guaranteed. Some cases also use
1694 * LL_MAX_HEADER instead to allocate the skb
1695 *
1696 * interface address info:
1697 *
1698 * @perm_addr: Permanent hw address
1699 * @addr_assign_type: Hw address assignment type
1700 * @addr_len: Hardware address length
5343da4c
TY
1701 * @upper_level: Maximum depth level of upper devices.
1702 * @lower_level: Maximum depth level of lower devices.
8626a0c8 1703 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1704 * @dev_id: Used to differentiate devices that share
1705 * the same link layer address
1706 * @dev_port: Used to differentiate devices that share
1707 * the same function
1708 * @addr_list_lock: XXX: need comments on this one
a1fa83bd 1709 * @name_assign_type: network interface name assignment type
5e82b4b2 1710 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1711 * has been enabled due to the need to listen to
1712 * additional unicast addresses in a device that
1713 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1714 * @uc: unicast mac addresses
1715 * @mc: multicast mac addresses
1716 * @dev_addrs: list of device hw addresses
1717 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1718 * @promiscuity: Number of times the NIC is told to work in
1719 * promiscuous mode; if it becomes 0 the NIC will
1720 * exit promiscuous mode
536721b1
KK
1721 * @allmulti: Counter, enables or disables allmulticast mode
1722 *
1723 * @vlan_info: VLAN info
1724 * @dsa_ptr: dsa specific data
1725 * @tipc_ptr: TIPC specific data
1726 * @atalk_ptr: AppleTalk link
1727 * @ip_ptr: IPv4 specific data
1728 * @dn_ptr: DECnet specific data
1729 * @ip6_ptr: IPv6 specific data
1730 * @ax25_ptr: AX.25 specific data
1731 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
a1fa83bd
RD
1732 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1733 * device struct
1734 * @mpls_ptr: mpls_dev struct pointer
536721b1 1735 *
536721b1
KK
1736 * @dev_addr: Hw address (before bcast,
1737 * because most packets are unicast)
1738 *
1739 * @_rx: Array of RX queues
1740 * @num_rx_queues: Number of RX queues
1741 * allocated at register_netdev() time
1742 * @real_num_rx_queues: Number of RX queues currently active in device
a1fa83bd
RD
1743 * @xdp_prog: XDP sockets filter program pointer
1744 * @gro_flush_timeout: timeout for GRO layer in NAPI
536721b1
KK
1745 *
1746 * @rx_handler: handler for received packets
1747 * @rx_handler_data: XXX: need comments on this one
46209401
JP
1748 * @miniq_ingress: ingress/clsact qdisc specific data for
1749 * ingress processing
536721b1 1750 * @ingress_queue: XXX: need comments on this one
2f5e70c8 1751 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
536721b1
KK
1752 * @broadcast: hw bcast address
1753 *
14ffbbb8
TG
1754 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1755 * indexed by RX queue number. Assigned by driver.
1756 * This must only be set if the ndo_rx_flow_steer
1757 * operation is defined
1758 * @index_hlist: Device index hash chain
1759 *
536721b1
KK
1760 * @_tx: Array of TX queues
1761 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1762 * @real_num_tx_queues: Number of TX queues currently active in device
1763 * @qdisc: Root qdisc from userspace point of view
1764 * @tx_queue_len: Max frames per queue allowed
1765 * @tx_global_lock: XXX: need comments on this one
a1fa83bd
RD
1766 * @xdp_bulkq: XDP device bulk queue
1767 * @xps_cpus_map: all CPUs map for XPS device
1768 * @xps_rxqs_map: all RXQs map for XPS device
536721b1
KK
1769 *
1770 * @xps_maps: XXX: need comments on this one
46209401
JP
1771 * @miniq_egress: clsact qdisc specific data for
1772 * egress processing
a1fa83bd 1773 * @qdisc_hash: qdisc hash table
536721b1 1774 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1775 * the watchdog (see dev_watchdog())
536721b1
KK
1776 * @watchdog_timer: List of timers
1777 *
1778 * @pcpu_refcnt: Number of references to this device
1779 * @todo_list: Delayed register/unregister
536721b1
KK
1780 * @link_watch_list: XXX: need comments on this one
1781 *
1782 * @reg_state: Register/unregister state machine
1783 * @dismantle: Device is going to be freed
1784 * @rtnl_link_state: This enum represents the phases of creating
1785 * a new link
1786 *
cf124db5
DM
1787 * @needs_free_netdev: Should unregister perform free_netdev?
1788 * @priv_destructor: Called from unregister
536721b1
KK
1789 * @npinfo: XXX: need comments on this one
1790 * @nd_net: Network namespace this network device is inside
1791 *
1792 * @ml_priv: Mid-layer private
1793 * @lstats: Loopback statistics
1794 * @tstats: Tunnel statistics
1795 * @dstats: Dummy statistics
1796 * @vstats: Virtual ethernet statistics
1797 *
1798 * @garp_port: GARP
1799 * @mrp_port: MRP
1800 *
1801 * @dev: Class/net/name entry
1802 * @sysfs_groups: Space for optional device, statistics and wireless
1803 * sysfs groups
1804 *
1805 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1806 * @rtnl_link_ops: Rtnl_link_ops
1807 *
1808 * @gso_max_size: Maximum size of generic segmentation offload
1809 * @gso_max_segs: Maximum number of segments that can be passed to the
1810 * NIC for GSO
1811 *
1812 * @dcbnl_ops: Data Center Bridging netlink ops
1813 * @num_tc: Number of traffic classes in the net device
1814 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1815 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1816 *
1817 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1818 *
1819 * @priomap: XXX: need comments on this one
1820 * @phydev: Physical device may attach itself
1821 * for hardware timestamping
e679c9c1 1822 * @sfp_bus: attached &struct sfp_bus structure.
1a33e10e 1823 *
ab92d68f
TY
1824 * @addr_list_lock_key: lockdep class annotating
1825 * net_device->addr_list_lock spinlock
1a33e10e
CW
1826 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1827 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1828 *
d746d707
AK
1829 * @proto_down: protocol port state information can be sent to the
1830 * switch driver and used to set the phys state of the
1831 * switch port.
1832 *
61941143
HK
1833 * @wol_enabled: Wake-on-LAN is enabled
1834 *
93642e14
JP
1835 * @net_notifier_list: List of per-net netdev notifier block
1836 * that follow this device when it is moved
1837 * to another network namespace.
1838 *
30e9bb84
AT
1839 * @macsec_ops: MACsec offloading ops
1840 *
1da177e4
LT
1841 * FIXME: cleanup struct net_device such that network protocol info
1842 * moves out.
1843 */
1844
d94d9fee 1845struct net_device {
1da177e4 1846 char name[IFNAMSIZ];
ff927412 1847 struct netdev_name_node *name_node;
6c557001 1848 struct dev_ifalias __rcu *ifalias;
1da177e4
LT
1849 /*
1850 * I/O specific fields
1851 * FIXME: Merge these and struct ifmap into one
1852 */
536721b1
KK
1853 unsigned long mem_end;
1854 unsigned long mem_start;
1855 unsigned long base_addr;
1856 int irq;
1da177e4
LT
1857
1858 /*
536721b1
KK
1859 * Some hardware also needs these fields (state,dev_list,
1860 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1861 * part of the usual set specified in Space.c.
1862 */
1863
1da177e4
LT
1864 unsigned long state;
1865
7562f876 1866 struct list_head dev_list;
bea3348e 1867 struct list_head napi_list;
44a0873d 1868 struct list_head unreg_list;
5cde2829 1869 struct list_head close_list;
7866a621
SN
1870 struct list_head ptype_all;
1871 struct list_head ptype_specific;
2f268f12 1872
2f268f12
VF
1873 struct {
1874 struct list_head upper;
1875 struct list_head lower;
1876 } adj_list;
1877
c8f44aff 1878 netdev_features_t features;
c8f44aff 1879 netdev_features_t hw_features;
c8f44aff 1880 netdev_features_t wanted_features;
c8f44aff 1881 netdev_features_t vlan_features;
6a674e9c 1882 netdev_features_t hw_enc_features;
0d89d203 1883 netdev_features_t mpls_features;
802ab55a 1884 netdev_features_t gso_partial_features;
04ed3e74 1885
1da177e4 1886 int ifindex;
7a66bbc9 1887 int group;
1da177e4 1888
c45d286e 1889 struct net_device_stats stats;
015f0688 1890
015f0688
ED
1891 atomic_long_t rx_dropped;
1892 atomic_long_t tx_dropped;
6e7333d3 1893 atomic_long_t rx_nohandler;
1da177e4 1894
b2d3bcfa
DD
1895 /* Stats to monitor link on/off, flapping */
1896 atomic_t carrier_up_count;
1897 atomic_t carrier_down_count;
1898
b86e0280 1899#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1900 const struct iw_handler_def *wireless_handlers;
1901 struct iw_public_data *wireless_data;
b86e0280 1902#endif
d314774c 1903 const struct net_device_ops *netdev_ops;
76fd8593 1904 const struct ethtool_ops *ethtool_ops;
1b69c6d0
DA
1905#ifdef CONFIG_NET_L3_MASTER_DEV
1906 const struct l3mdev_ops *l3mdev_ops;
1907#endif
f997c55c
AA
1908#if IS_ENABLED(CONFIG_IPV6)
1909 const struct ndisc_ops *ndisc_ops;
1910#endif
1da177e4 1911
9cb0d21d 1912#ifdef CONFIG_XFRM_OFFLOAD
d77e38e6
SK
1913 const struct xfrmdev_ops *xfrmdev_ops;
1914#endif
1915
a5c37c63
IL
1916#if IS_ENABLED(CONFIG_TLS_DEVICE)
1917 const struct tlsdev_ops *tlsdev_ops;
1918#endif
1919
3b04ddde
SH
1920 const struct header_ops *header_ops;
1921
536721b1
KK
1922 unsigned int flags;
1923 unsigned int priv_flags;
1924
1da177e4 1925 unsigned short gflags;
536721b1 1926 unsigned short padded;
1da177e4 1927
536721b1
KK
1928 unsigned char operstate;
1929 unsigned char link_mode;
b00055aa 1930
536721b1
KK
1931 unsigned char if_port;
1932 unsigned char dma;
bdc220da 1933
501a90c9
ED
1934 /* Note : dev->mtu is often read without holding a lock.
1935 * Writers usually hold RTNL.
1936 * It is recommended to use READ_ONCE() to annotate the reads,
1937 * and to use WRITE_ONCE() to annotate the writes.
1938 */
536721b1 1939 unsigned int mtu;
61e84623
JW
1940 unsigned int min_mtu;
1941 unsigned int max_mtu;
536721b1
KK
1942 unsigned short type;
1943 unsigned short hard_header_len;
d92be7a4 1944 unsigned char min_header_len;
1da177e4 1945
f5184d26
JB
1946 unsigned short needed_headroom;
1947 unsigned short needed_tailroom;
1948
1da177e4 1949 /* Interface address info. */
536721b1
KK
1950 unsigned char perm_addr[MAX_ADDR_LEN];
1951 unsigned char addr_assign_type;
1952 unsigned char addr_len;
5343da4c
TY
1953 unsigned char upper_level;
1954 unsigned char lower_level;
a0a9663d 1955 unsigned short neigh_priv_len;
536721b1
KK
1956 unsigned short dev_id;
1957 unsigned short dev_port;
ccffad25 1958 spinlock_t addr_list_lock;
14ffbbb8
TG
1959 unsigned char name_assign_type;
1960 bool uc_promisc;
536721b1
KK
1961 struct netdev_hw_addr_list uc;
1962 struct netdev_hw_addr_list mc;
1963 struct netdev_hw_addr_list dev_addrs;
1964
4c3d5e7b
ED
1965#ifdef CONFIG_SYSFS
1966 struct kset *queues_kset;
1967#endif
9d45abe1
WC
1968 unsigned int promiscuity;
1969 unsigned int allmulti;
1da177e4 1970
1da177e4 1971
5e82b4b2 1972 /* Protocol-specific pointers */
65ac6a5f 1973
d11ead75 1974#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1975 struct vlan_info __rcu *vlan_info;
65ac6a5f 1976#endif
34a430d7 1977#if IS_ENABLED(CONFIG_NET_DSA)
2f657a60 1978 struct dsa_port *dsa_ptr;
37cb0620
YX
1979#endif
1980#if IS_ENABLED(CONFIG_TIPC)
536721b1 1981 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1982#endif
89e58148 1983#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
536721b1 1984 void *atalk_ptr;
89e58148 1985#endif
536721b1 1986 struct in_device __rcu *ip_ptr;
330c7272 1987#if IS_ENABLED(CONFIG_DECNET)
536721b1 1988 struct dn_dev __rcu *dn_ptr;
330c7272 1989#endif
536721b1 1990 struct inet6_dev __rcu *ip6_ptr;
19ff13f2 1991#if IS_ENABLED(CONFIG_AX25)
536721b1 1992 void *ax25_ptr;
19ff13f2 1993#endif
536721b1 1994 struct wireless_dev *ieee80211_ptr;
98a18b6f 1995 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1996#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1997 struct mpls_dev __rcu *mpls_ptr;
1998#endif
1da177e4 1999
9356b8fc 2000/*
cd13539b 2001 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 2002 */
9356b8fc 2003 /* Interface address info used in eth_type_trans() */
536721b1 2004 unsigned char *dev_addr;
f001fde5 2005
0a9627f2 2006 struct netdev_rx_queue *_rx;
0a9627f2 2007 unsigned int num_rx_queues;
62fe0b40 2008 unsigned int real_num_rx_queues;
0a9627f2 2009
7acedaf5 2010 struct bpf_prog __rcu *xdp_prog;
3b47d303 2011 unsigned long gro_flush_timeout;
6f8b12d6 2012 int napi_defer_hard_irqs;
61391cde 2013 rx_handler_func_t __rcu *rx_handler;
2014 void __rcu *rx_handler_data;
e8a0464c 2015
4cda01e8 2016#ifdef CONFIG_NET_CLS_ACT
46209401 2017 struct mini_Qdisc __rcu *miniq_ingress;
d2788d34 2018#endif
24824a09 2019 struct netdev_queue __rcu *ingress_queue;
e687ad60 2020#ifdef CONFIG_NETFILTER_INGRESS
960632ec 2021 struct nf_hook_entries __rcu *nf_hooks_ingress;
e687ad60 2022#endif
d2788d34 2023
536721b1 2024 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
2025#ifdef CONFIG_RFS_ACCEL
2026 struct cpu_rmap *rx_cpu_rmap;
2027#endif
2028 struct hlist_node index_hlist;
cd13539b
ED
2029
2030/*
2031 * Cache lines mostly used on transmit path
2032 */
e8a0464c
DM
2033 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2034 unsigned int num_tx_queues;
fd2ea0a7 2035 unsigned int real_num_tx_queues;
af356afa 2036 struct Qdisc *qdisc;
0cd29503 2037 unsigned int tx_queue_len;
c3f26a26 2038 spinlock_t tx_global_lock;
75ccae62
THJ
2039
2040 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
cd13539b 2041
bf264145 2042#ifdef CONFIG_XPS
80d19669
AN
2043 struct xps_dev_maps __rcu *xps_cpus_map;
2044 struct xps_dev_maps __rcu *xps_rxqs_map;
bf264145 2045#endif
1f211a1b 2046#ifdef CONFIG_NET_CLS_ACT
46209401 2047 struct mini_Qdisc __rcu *miniq_egress;
1f211a1b 2048#endif
0c4f691f 2049
75ccae62
THJ
2050#ifdef CONFIG_NET_SCHED
2051 DECLARE_HASHTABLE (qdisc_hash, 4);
2052#endif
9356b8fc 2053 /* These may be needed for future network-power-down code. */
9356b8fc 2054 struct timer_list watchdog_timer;
75ccae62 2055 int watchdog_timeo;
9356b8fc 2056
1da177e4 2057 struct list_head todo_list;
75ccae62 2058 int __percpu *pcpu_refcnt;
1da177e4 2059
e014debe 2060 struct list_head link_watch_list;
572a103d 2061
1da177e4 2062 enum { NETREG_UNINITIALIZED=0,
b17a7c17 2063 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
2064 NETREG_UNREGISTERING, /* called unregister_netdevice */
2065 NETREG_UNREGISTERED, /* completed unregister todo */
2066 NETREG_RELEASED, /* called free_netdev */
937f1ba5 2067 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
2068 } reg_state:8;
2069
536721b1 2070 bool dismantle;
a2835763
PM
2071
2072 enum {
2073 RTNL_LINK_INITIALIZED,
2074 RTNL_LINK_INITIALIZING,
2075 } rtnl_link_state:16;
1da177e4 2076
cf124db5
DM
2077 bool needs_free_netdev;
2078 void (*priv_destructor)(struct net_device *dev);
1da177e4 2079
1da177e4 2080#ifdef CONFIG_NETPOLL
5fbee843 2081 struct netpoll_info __rcu *npinfo;
1da177e4 2082#endif
eae792b7 2083
0c5c9fb5 2084 possible_net_t nd_net;
4a1c5371 2085
4951704b 2086 /* mid-layer private */
a7855c78 2087 union {
536721b1
KK
2088 void *ml_priv;
2089 struct pcpu_lstats __percpu *lstats;
8f84985f 2090 struct pcpu_sw_netstats __percpu *tstats;
536721b1 2091 struct pcpu_dstats __percpu *dstats;
a7855c78 2092 };
536721b1 2093
fb585b44 2094#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 2095 struct garp_port __rcu *garp_port;
fb585b44
TK
2096#endif
2097#if IS_ENABLED(CONFIG_MRP)
febf018d 2098 struct mrp_port __rcu *mrp_port;
fb585b44 2099#endif
1da177e4 2100
5e82b4b2 2101 struct device dev;
0c509a6c 2102 const struct attribute_group *sysfs_groups[4];
a953be53 2103 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 2104
38f7b870 2105 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 2106
82cc1a7a
PWJ
2107 /* for setting kernel sock attribute on TCP connection setup */
2108#define GSO_MAX_SIZE 65536
2109 unsigned int gso_max_size;
30b678d8
BH
2110#define GSO_MAX_SEGS 65535
2111 u16 gso_max_segs;
743b03a8 2112
7a6b6f51 2113#ifdef CONFIG_DCB
32953543 2114 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 2115#endif
ffcfe25b 2116 s16 num_tc;
5e82b4b2
BH
2117 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2118 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 2119
d11ead75 2120#if IS_ENABLED(CONFIG_FCOE)
4d288d57 2121 unsigned int fcoe_ddp_xid;
5bc1421e 2122#endif
86f8515f 2123#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 2124 struct netprio_map __rcu *priomap;
4d288d57 2125#endif
5e82b4b2 2126 struct phy_device *phydev;
e679c9c1 2127 struct sfp_bus *sfp_bus;
ab92d68f 2128 struct lock_class_key addr_list_lock_key;
1a33e10e
CW
2129 struct lock_class_key *qdisc_tx_busylock;
2130 struct lock_class_key *qdisc_running_key;
5e82b4b2 2131 bool proto_down;
61941143 2132 unsigned wol_enabled:1;
93642e14
JP
2133
2134 struct list_head net_notifier_list;
30e9bb84
AT
2135
2136#if IS_ENABLED(CONFIG_MACSEC)
2137 /* MACsec management functions */
2138 const struct macsec_ops *macsec_ops;
2139#endif
1da177e4 2140};
43cb76d9 2141#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 2142
b5cdae32
DM
2143static inline bool netif_elide_gro(const struct net_device *dev)
2144{
2145 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2146 return true;
2147 return false;
2148}
2149
1da177e4 2150#define NETDEV_ALIGN 32
1da177e4 2151
4f57c087
JF
2152static inline
2153int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2154{
2155 return dev->prio_tc_map[prio & TC_BITMASK];
2156}
2157
2158static inline
2159int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2160{
2161 if (tc >= dev->num_tc)
2162 return -EINVAL;
2163
2164 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2165 return 0;
2166}
2167
8d059b0f 2168int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
2169void netdev_reset_tc(struct net_device *dev);
2170int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2171int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
2172
2173static inline
2174int netdev_get_num_tc(struct net_device *dev)
2175{
2176 return dev->num_tc;
2177}
2178
ffcfe25b
AD
2179void netdev_unbind_sb_channel(struct net_device *dev,
2180 struct net_device *sb_dev);
2181int netdev_bind_sb_channel_queue(struct net_device *dev,
2182 struct net_device *sb_dev,
2183 u8 tc, u16 count, u16 offset);
2184int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2185static inline int netdev_get_sb_channel(struct net_device *dev)
2186{
2187 return max_t(int, -dev->num_tc, 0);
2188}
2189
e8a0464c
DM
2190static inline
2191struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2192 unsigned int index)
2193{
2194 return &dev->_tx[index];
2195}
2196
10c51b56
DB
2197static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2198 const struct sk_buff *skb)
2199{
2200 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2201}
2202
e8a0464c
DM
2203static inline void netdev_for_each_tx_queue(struct net_device *dev,
2204 void (*f)(struct net_device *,
2205 struct netdev_queue *,
2206 void *),
2207 void *arg)
2208{
2209 unsigned int i;
2210
2211 for (i = 0; i < dev->num_tx_queues; i++)
2212 f(dev, &dev->_tx[i], arg);
2213}
2214
1a33e10e
CW
2215#define netdev_lockdep_set_classes(dev) \
2216{ \
2217 static struct lock_class_key qdisc_tx_busylock_key; \
2218 static struct lock_class_key qdisc_running_key; \
2219 static struct lock_class_key qdisc_xmit_lock_key; \
2220 unsigned int i; \
2221 \
2222 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2223 (dev)->qdisc_running_key = &qdisc_running_key; \
2224 for (i = 0; i < (dev)->num_tx_queues; i++) \
2225 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2226 &qdisc_xmit_lock_key); \
2227}
2228
b71b5837
PA
2229u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2230 struct net_device *sb_dev);
4bd97d51
PA
2231struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2232 struct sk_buff *skb,
2233 struct net_device *sb_dev);
8c4c49df 2234
871b642a
PA
2235/* returns the headroom that the master device needs to take in account
2236 * when forwarding to this dev
2237 */
2238static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2239{
2240 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2241}
2242
2243static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2244{
2245 if (dev->netdev_ops->ndo_set_rx_headroom)
2246 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2247}
2248
2249/* set the device rx headroom to the dev's default */
2250static inline void netdev_reset_rx_headroom(struct net_device *dev)
2251{
2252 netdev_set_rx_headroom(dev, -1);
2253}
2254
c346dca1
YH
2255/*
2256 * Net namespace inlines
2257 */
2258static inline
2259struct net *dev_net(const struct net_device *dev)
2260{
c2d9ba9b 2261 return read_pnet(&dev->nd_net);
c346dca1
YH
2262}
2263
2264static inline
f5aa23fd 2265void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2266{
0c5c9fb5 2267 write_pnet(&dev->nd_net, net);
c346dca1
YH
2268}
2269
bea3348e
SH
2270/**
2271 * netdev_priv - access network device private data
2272 * @dev: network device
2273 *
2274 * Get network device private data
2275 */
6472ce60 2276static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2277{
1ce8e7b5 2278 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2279}
2280
1da177e4
LT
2281/* Set the sysfs physical device reference for the network logical device
2282 * if set prior to registration will cause a symlink during initialization.
2283 */
43cb76d9 2284#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2285
384912ed 2286/* Set the sysfs device type for the network logical device to allow
3f79410c 2287 * fine-grained identification of different network device types. For
5e82b4b2 2288 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2289 */
2290#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2291
82dc3c63
ED
2292/* Default NAPI poll() weight
2293 * Device drivers are strongly advised to not use bigger value
2294 */
2295#define NAPI_POLL_WEIGHT 64
2296
3b582cc1 2297/**
5e82b4b2 2298 * netif_napi_add - initialize a NAPI context
3b582cc1 2299 * @dev: network device
5e82b4b2 2300 * @napi: NAPI context
3b582cc1
SH
2301 * @poll: polling function
2302 * @weight: default weight
2303 *
5e82b4b2
BH
2304 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2305 * *any* of the other NAPI-related functions.
3b582cc1 2306 */
d565b0a1
HX
2307void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2308 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2309
d64b5e85 2310/**
5e82b4b2 2311 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2312 * @dev: network device
5e82b4b2 2313 * @napi: NAPI context
d64b5e85
ED
2314 * @poll: polling function
2315 * @weight: default weight
2316 *
2317 * This variant of netif_napi_add() should be used from drivers using NAPI
2318 * to exclusively poll a TX queue.
2319 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2320 */
2321static inline void netif_tx_napi_add(struct net_device *dev,
2322 struct napi_struct *napi,
2323 int (*poll)(struct napi_struct *, int),
2324 int weight)
2325{
2326 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2327 netif_napi_add(dev, napi, poll, weight);
2328}
2329
d8156534 2330/**
5e82b4b2
BH
2331 * netif_napi_del - remove a NAPI context
2332 * @napi: NAPI context
d8156534 2333 *
5e82b4b2 2334 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2335 */
d565b0a1
HX
2336void netif_napi_del(struct napi_struct *napi);
2337
2338struct napi_gro_cb {
78a478d0 2339 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2340 void *frag0;
78a478d0 2341
7489594c
HX
2342 /* Length of frag0. */
2343 unsigned int frag0_len;
2344
86911732 2345 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2346 int data_offset;
86911732 2347
d565b0a1 2348 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2349 u16 flush;
2350
2351 /* Save the IP ID here and check when we get to the transport layer */
2352 u16 flush_id;
d565b0a1
HX
2353
2354 /* Number of segments aggregated. */
2e71a6f8
ED
2355 u16 count;
2356
15e2396d
TH
2357 /* Start offset for remote checksum offload */
2358 u16 gro_remcsum_start;
2359
2e71a6f8
ED
2360 /* jiffies when first packet was created/queued */
2361 unsigned long age;
86347245 2362
afe93325 2363 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2364 u16 proto;
2365
baa32ff4
TH
2366 /* This is non-zero if the packet may be of the same flow. */
2367 u8 same_flow:1;
2368
fac8e0f5
JG
2369 /* Used in tunnel GRO receive */
2370 u8 encap_mark:1;
573e8fca
TH
2371
2372 /* GRO checksum is valid */
2373 u8 csum_valid:1;
2374
662880f4
TH
2375 /* Number of checksums via CHECKSUM_UNNECESSARY */
2376 u8 csum_cnt:3;
c3c7c254 2377
baa32ff4
TH
2378 /* Free the skb? */
2379 u8 free:2;
2380#define NAPI_GRO_FREE 1
2381#define NAPI_GRO_FREE_STOLEN_HEAD 2
2382
efc98d08
TH
2383 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2384 u8 is_ipv6:1;
2385
a0ca153f
AD
2386 /* Used in GRE, set in fou/gue_gro_receive */
2387 u8 is_fou:1;
2388
1530545e
AD
2389 /* Used to determine if flush_id can be ignored */
2390 u8 is_atomic:1;
2391
fcd91dd4
SD
2392 /* Number of gro_receive callbacks this packet already went through */
2393 u8 recursion_counter:4;
2394
3a1296a3
SK
2395 /* GRO is done by frag_list pointer chaining. */
2396 u8 is_flist:1;
baa32ff4 2397
bf5a755f
JC
2398 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2399 __wsum csum;
2400
c3c7c254
ED
2401 /* used in skb_gro_receive() slow path */
2402 struct sk_buff *last;
d565b0a1
HX
2403};
2404
2405#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2406
fcd91dd4
SD
2407#define GRO_RECURSION_LIMIT 15
2408static inline int gro_recursion_inc_test(struct sk_buff *skb)
2409{
2410 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2411}
2412
d4546c25
DM
2413typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2414static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2415 struct list_head *head,
2416 struct sk_buff *skb)
fcd91dd4
SD
2417{
2418 if (unlikely(gro_recursion_inc_test(skb))) {
2419 NAPI_GRO_CB(skb)->flush |= 1;
2420 return NULL;
2421 }
2422
2423 return cb(head, skb);
2424}
2425
d4546c25
DM
2426typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2427 struct sk_buff *);
2428static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2429 struct sock *sk,
2430 struct list_head *head,
2431 struct sk_buff *skb)
fcd91dd4
SD
2432{
2433 if (unlikely(gro_recursion_inc_test(skb))) {
2434 NAPI_GRO_CB(skb)->flush |= 1;
2435 return NULL;
2436 }
2437
2438 return cb(sk, head, skb);
2439}
2440
1da177e4 2441struct packet_type {
f2ccd8fa 2442 __be16 type; /* This is really htons(ether_type). */
fa788d98 2443 bool ignore_outgoing;
f2ccd8fa
DM
2444 struct net_device *dev; /* NULL is wildcarded here */
2445 int (*func) (struct sk_buff *,
2446 struct net_device *,
2447 struct packet_type *,
2448 struct net_device *);
17266ee9
EC
2449 void (*list_func) (struct list_head *,
2450 struct packet_type *,
2451 struct net_device *);
c0de08d0
EL
2452 bool (*id_match)(struct packet_type *ptype,
2453 struct sock *sk);
1da177e4
LT
2454 void *af_packet_priv;
2455 struct list_head list;
2456};
2457
f191a1d1 2458struct offload_callbacks {
576a30eb 2459 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2460 netdev_features_t features);
d4546c25
DM
2461 struct sk_buff *(*gro_receive)(struct list_head *head,
2462 struct sk_buff *skb);
299603e8 2463 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2464};
2465
2466struct packet_offload {
2467 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2468 u16 priority;
f191a1d1
VY
2469 struct offload_callbacks callbacks;
2470 struct list_head list;
1da177e4
LT
2471};
2472
5e82b4b2 2473/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2474struct pcpu_sw_netstats {
2475 u64 rx_packets;
2476 u64 rx_bytes;
2477 u64 tx_packets;
2478 u64 tx_bytes;
2479 struct u64_stats_sync syncp;
9a5ee462 2480} __aligned(4 * sizeof(u64));
52bb6677
LR
2481
2482struct pcpu_lstats {
fd2f4737
ED
2483 u64_stats_t packets;
2484 u64_stats_t bytes;
52bb6677 2485 struct u64_stats_sync syncp;
9a5ee462 2486} __aligned(2 * sizeof(u64));
8f84985f 2487
de7d5084
ED
2488void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2489
dd5382a0
ED
2490static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2491{
2492 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2493
2494 u64_stats_update_begin(&lstats->syncp);
fd2f4737
ED
2495 u64_stats_add(&lstats->bytes, len);
2496 u64_stats_inc(&lstats->packets);
dd5382a0
ED
2497 u64_stats_update_end(&lstats->syncp);
2498}
2499
aabc92bb
PNA
2500#define __netdev_alloc_pcpu_stats(type, gfp) \
2501({ \
2502 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2503 if (pcpu_stats) { \
2504 int __cpu; \
2505 for_each_possible_cpu(__cpu) { \
2506 typeof(type) *stat; \
2507 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2508 u64_stats_init(&stat->syncp); \
2509 } \
2510 } \
2511 pcpu_stats; \
1c213bd2
WC
2512})
2513
aabc92bb 2514#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2515 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2516
764f5e54
JP
2517enum netdev_lag_tx_type {
2518 NETDEV_LAG_TX_TYPE_UNKNOWN,
2519 NETDEV_LAG_TX_TYPE_RANDOM,
2520 NETDEV_LAG_TX_TYPE_BROADCAST,
2521 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2522 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2523 NETDEV_LAG_TX_TYPE_HASH,
2524};
2525
f44aa9ef
JH
2526enum netdev_lag_hash {
2527 NETDEV_LAG_HASH_NONE,
2528 NETDEV_LAG_HASH_L2,
2529 NETDEV_LAG_HASH_L34,
2530 NETDEV_LAG_HASH_L23,
2531 NETDEV_LAG_HASH_E23,
2532 NETDEV_LAG_HASH_E34,
2533 NETDEV_LAG_HASH_UNKNOWN,
2534};
2535
764f5e54
JP
2536struct netdev_lag_upper_info {
2537 enum netdev_lag_tx_type tx_type;
f44aa9ef 2538 enum netdev_lag_hash hash_type;
764f5e54
JP
2539};
2540
fb1b2e3c
JP
2541struct netdev_lag_lower_state_info {
2542 u8 link_up : 1,
2543 tx_enabled : 1;
2544};
2545
1da177e4
LT
2546#include <linux/notifier.h>
2547
ede2762d
KT
2548/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2549 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2550 * adding new types.
dcfe1421 2551 */
ede2762d
KT
2552enum netdev_cmd {
2553 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2554 NETDEV_DOWN,
2555 NETDEV_REBOOT, /* Tell a protocol stack a network interface
dcfe1421
AW
2556 detected a hardware crash and restarted
2557 - we can use this eg to kick tcp sessions
2558 once done */
ede2762d
KT
2559 NETDEV_CHANGE, /* Notify device state change */
2560 NETDEV_REGISTER,
2561 NETDEV_UNREGISTER,
2562 NETDEV_CHANGEMTU, /* notify after mtu change happened */
1570415f
PM
2563 NETDEV_CHANGEADDR, /* notify after the address change */
2564 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
ede2762d
KT
2565 NETDEV_GOING_DOWN,
2566 NETDEV_CHANGENAME,
2567 NETDEV_FEAT_CHANGE,
2568 NETDEV_BONDING_FAILOVER,
2569 NETDEV_PRE_UP,
2570 NETDEV_PRE_TYPE_CHANGE,
2571 NETDEV_POST_TYPE_CHANGE,
2572 NETDEV_POST_INIT,
ede2762d
KT
2573 NETDEV_RELEASE,
2574 NETDEV_NOTIFY_PEERS,
2575 NETDEV_JOIN,
2576 NETDEV_CHANGEUPPER,
2577 NETDEV_RESEND_IGMP,
2578 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2579 NETDEV_CHANGEINFODATA,
2580 NETDEV_BONDING_INFO,
2581 NETDEV_PRECHANGEUPPER,
2582 NETDEV_CHANGELOWERSTATE,
2583 NETDEV_UDP_TUNNEL_PUSH_INFO,
2584 NETDEV_UDP_TUNNEL_DROP_INFO,
2585 NETDEV_CHANGE_TX_QUEUE_LEN,
9daae9bd
GP
2586 NETDEV_CVLAN_FILTER_PUSH_INFO,
2587 NETDEV_CVLAN_FILTER_DROP_INFO,
2588 NETDEV_SVLAN_FILTER_PUSH_INFO,
2589 NETDEV_SVLAN_FILTER_DROP_INFO,
ede2762d
KT
2590};
2591const char *netdev_cmd_to_name(enum netdev_cmd cmd);
dcfe1421 2592
f629d208
JP
2593int register_netdevice_notifier(struct notifier_block *nb);
2594int unregister_netdevice_notifier(struct notifier_block *nb);
a30c7b42
JP
2595int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2596int unregister_netdevice_notifier_net(struct net *net,
2597 struct notifier_block *nb);
93642e14
JP
2598int register_netdevice_notifier_dev_net(struct net_device *dev,
2599 struct notifier_block *nb,
2600 struct netdev_net_notifier *nn);
2601int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2602 struct notifier_block *nb,
2603 struct netdev_net_notifier *nn);
351638e7
JP
2604
2605struct netdev_notifier_info {
51d0c047
DA
2606 struct net_device *dev;
2607 struct netlink_ext_ack *extack;
351638e7
JP
2608};
2609
af7d6cce
SD
2610struct netdev_notifier_info_ext {
2611 struct netdev_notifier_info info; /* must be first */
2612 union {
2613 u32 mtu;
2614 } ext;
2615};
2616
be9efd36
JP
2617struct netdev_notifier_change_info {
2618 struct netdev_notifier_info info; /* must be first */
2619 unsigned int flags_changed;
2620};
2621
0e4ead9d
JP
2622struct netdev_notifier_changeupper_info {
2623 struct netdev_notifier_info info; /* must be first */
2624 struct net_device *upper_dev; /* new upper dev */
2625 bool master; /* is upper dev master */
5e82b4b2 2626 bool linking; /* is the notification for link or unlink */
29bf24af 2627 void *upper_info; /* upper dev info */
0e4ead9d
JP
2628};
2629
04d48266
JP
2630struct netdev_notifier_changelowerstate_info {
2631 struct netdev_notifier_info info; /* must be first */
2632 void *lower_state_info; /* is lower dev state */
2633};
2634
1570415f
PM
2635struct netdev_notifier_pre_changeaddr_info {
2636 struct netdev_notifier_info info; /* must be first */
2637 const unsigned char *dev_addr;
2638};
2639
75538c2b
CW
2640static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2641 struct net_device *dev)
2642{
2643 info->dev = dev;
51d0c047 2644 info->extack = NULL;
75538c2b
CW
2645}
2646
351638e7
JP
2647static inline struct net_device *
2648netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2649{
2650 return info->dev;
2651}
2652
51d0c047
DA
2653static inline struct netlink_ext_ack *
2654netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2655{
2656 return info->extack;
2657}
2658
f629d208 2659int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2660
2661
1da177e4
LT
2662extern rwlock_t dev_base_lock; /* Device list lock */
2663
881d966b
EB
2664#define for_each_netdev(net, d) \
2665 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2666#define for_each_netdev_reverse(net, d) \
2667 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2668#define for_each_netdev_rcu(net, d) \
2669 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2670#define for_each_netdev_safe(net, d, n) \
2671 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2672#define for_each_netdev_continue(net, d) \
2673 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
afa0df59
JP
2674#define for_each_netdev_continue_reverse(net, d) \
2675 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2676 dev_list)
254245d2 2677#define for_each_netdev_continue_rcu(net, d) \
2678 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2679#define for_each_netdev_in_bond_rcu(bond, slave) \
2680 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2681 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2682#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2683
a050c33f
DL
2684static inline struct net_device *next_net_device(struct net_device *dev)
2685{
2686 struct list_head *lh;
2687 struct net *net;
2688
c346dca1 2689 net = dev_net(dev);
a050c33f
DL
2690 lh = dev->dev_list.next;
2691 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2692}
2693
ce81b76a
ED
2694static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2695{
2696 struct list_head *lh;
2697 struct net *net;
2698
2699 net = dev_net(dev);
ccf43438 2700 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2701 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2702}
2703
a050c33f
DL
2704static inline struct net_device *first_net_device(struct net *net)
2705{
2706 return list_empty(&net->dev_base_head) ? NULL :
2707 net_device_entry(net->dev_base_head.next);
2708}
7562f876 2709
ccf43438
ED
2710static inline struct net_device *first_net_device_rcu(struct net *net)
2711{
2712 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2713
2714 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2715}
2716
f629d208
JP
2717int netdev_boot_setup_check(struct net_device *dev);
2718unsigned long netdev_boot_base(const char *prefix, int unit);
2719struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2720 const char *hwaddr);
2721struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2722struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2723void dev_add_pack(struct packet_type *pt);
2724void dev_remove_pack(struct packet_type *pt);
2725void __dev_remove_pack(struct packet_type *pt);
2726void dev_add_offload(struct packet_offload *po);
2727void dev_remove_offload(struct packet_offload *po);
f629d208 2728
a54acb3a 2729int dev_get_iflink(const struct net_device *dev);
fc4099f1 2730int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2731struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2732 unsigned short mask);
f629d208
JP
2733struct net_device *dev_get_by_name(struct net *net, const char *name);
2734struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2735struct net_device *__dev_get_by_name(struct net *net, const char *name);
2736int dev_alloc_name(struct net_device *dev, const char *name);
00f54e68 2737int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
7051b88a 2738void dev_close(struct net_device *dev);
2739void dev_close_many(struct list_head *head, bool unlink);
f629d208 2740void dev_disable_lro(struct net_device *dev);
0c4b51f0 2741int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
a4ea8a3d 2742u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
a350ecce 2743 struct net_device *sb_dev);
a4ea8a3d 2744u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
a350ecce 2745 struct net_device *sb_dev);
2b4aa3ce 2746int dev_queue_xmit(struct sk_buff *skb);
eadec877 2747int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
865b03f2 2748int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
f629d208
JP
2749int register_netdevice(struct net_device *dev);
2750void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2751void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2752static inline void unregister_netdevice(struct net_device *dev)
2753{
2754 unregister_netdevice_queue(dev, NULL);
2755}
2756
f629d208
JP
2757int netdev_refcnt_read(const struct net_device *dev);
2758void free_netdev(struct net_device *dev);
74d332c1 2759void netdev_freemem(struct net_device *dev);
f629d208
JP
2760void synchronize_net(void);
2761int init_dummy_netdev(struct net_device *dev);
937f1ba5 2762
cff9f12b
MG
2763struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2764 struct sk_buff *skb,
2765 bool all_slaves);
f629d208
JP
2766struct net_device *dev_get_by_index(struct net *net, int ifindex);
2767struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2768struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2769struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2770int netdev_get_name(struct net *net, char *name, int ifindex);
2771int dev_restart(struct net_device *dev);
d4546c25 2772int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
3a1296a3 2773int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
86911732
HX
2774
2775static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2776{
2777 return NAPI_GRO_CB(skb)->data_offset;
2778}
2779
2780static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2781{
2782 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2783}
2784
2785static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2786{
2787 NAPI_GRO_CB(skb)->data_offset += len;
2788}
2789
a5b1cf28
HX
2790static inline void *skb_gro_header_fast(struct sk_buff *skb,
2791 unsigned int offset)
86911732 2792{
a5b1cf28
HX
2793 return NAPI_GRO_CB(skb)->frag0 + offset;
2794}
78a478d0 2795
a5b1cf28
HX
2796static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2797{
2798 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2799}
78a478d0 2800
57ea52a8
HX
2801static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2802{
2803 NAPI_GRO_CB(skb)->frag0 = NULL;
2804 NAPI_GRO_CB(skb)->frag0_len = 0;
2805}
2806
a5b1cf28
HX
2807static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2808 unsigned int offset)
2809{
17dd759c
HX
2810 if (!pskb_may_pull(skb, hlen))
2811 return NULL;
2812
57ea52a8 2813 skb_gro_frag0_invalidate(skb);
17dd759c 2814 return skb->data + offset;
86911732 2815}
1da177e4 2816
36e7b1b8
HX
2817static inline void *skb_gro_network_header(struct sk_buff *skb)
2818{
78d3fd0b
HX
2819 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2820 skb_network_offset(skb);
36e7b1b8
HX
2821}
2822
bf5a755f
JC
2823static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2824 const void *start, unsigned int len)
2825{
573e8fca 2826 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2827 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2828 csum_partial(start, len, 0));
2829}
2830
573e8fca
TH
2831/* GRO checksum functions. These are logical equivalents of the normal
2832 * checksum functions (in skbuff.h) except that they operate on the GRO
2833 * offsets and fields in sk_buff.
2834 */
2835
2836__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2837
15e2396d
TH
2838static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2839{
b7fe10e5 2840 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2841}
2842
573e8fca
TH
2843static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2844 bool zero_okay,
2845 __sum16 check)
2846{
6edec0e6
TH
2847 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2848 skb_checksum_start_offset(skb) <
2849 skb_gro_offset(skb)) &&
15e2396d 2850 !skb_at_gro_remcsum_start(skb) &&
662880f4 2851 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2852 (!zero_okay || check));
2853}
2854
2855static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2856 __wsum psum)
2857{
2858 if (NAPI_GRO_CB(skb)->csum_valid &&
2859 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2860 return 0;
2861
2862 NAPI_GRO_CB(skb)->csum = psum;
2863
2864 return __skb_gro_checksum_complete(skb);
2865}
2866
573e8fca
TH
2867static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2868{
662880f4
TH
2869 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2870 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2871 NAPI_GRO_CB(skb)->csum_cnt--;
2872 } else {
2873 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2874 * verified a new top level checksum or an encapsulated one
2875 * during GRO. This saves work if we fallback to normal path.
2876 */
2877 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2878 }
2879}
2880
2881#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2882 compute_pseudo) \
2883({ \
2884 __sum16 __ret = 0; \
2885 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2886 __ret = __skb_gro_checksum_validate_complete(skb, \
2887 compute_pseudo(skb, proto)); \
219f1d79 2888 if (!__ret) \
573e8fca
TH
2889 skb_gro_incr_csum_unnecessary(skb); \
2890 __ret; \
2891})
2892
2893#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2894 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2895
2896#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2897 compute_pseudo) \
2898 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2899
2900#define skb_gro_checksum_simple_validate(skb) \
2901 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2902
d96535a1
TH
2903static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2904{
2905 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2906 !NAPI_GRO_CB(skb)->csum_valid);
2907}
2908
2909static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
b39c78b2 2910 __wsum pseudo)
d96535a1
TH
2911{
2912 NAPI_GRO_CB(skb)->csum = ~pseudo;
2913 NAPI_GRO_CB(skb)->csum_valid = 1;
2914}
2915
b39c78b2 2916#define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
2917do { \
2918 if (__skb_gro_checksum_convert_check(skb)) \
b39c78b2 2919 __skb_gro_checksum_convert(skb, \
d96535a1
TH
2920 compute_pseudo(skb, proto)); \
2921} while (0)
2922
26c4f7da
TH
2923struct gro_remcsum {
2924 int offset;
2925 __wsum delta;
2926};
2927
2928static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2929{
846cd667 2930 grc->offset = 0;
26c4f7da
TH
2931 grc->delta = 0;
2932}
2933
b7fe10e5
TH
2934static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2935 unsigned int off, size_t hdrlen,
2936 int start, int offset,
2937 struct gro_remcsum *grc,
2938 bool nopartial)
dcdc8994
TH
2939{
2940 __wsum delta;
b7fe10e5 2941 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2942
2943 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2944
15e2396d 2945 if (!nopartial) {
b7fe10e5
TH
2946 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2947 return ptr;
2948 }
2949
2950 ptr = skb_gro_header_fast(skb, off);
2951 if (skb_gro_header_hard(skb, off + plen)) {
2952 ptr = skb_gro_header_slow(skb, off + plen, off);
2953 if (!ptr)
2954 return NULL;
15e2396d
TH
2955 }
2956
b7fe10e5
TH
2957 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2958 start, offset);
dcdc8994
TH
2959
2960 /* Adjust skb->csum since we changed the packet */
dcdc8994 2961 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2962
b7fe10e5 2963 grc->offset = off + hdrlen + offset;
26c4f7da 2964 grc->delta = delta;
b7fe10e5
TH
2965
2966 return ptr;
dcdc8994
TH
2967}
2968
26c4f7da
TH
2969static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2970 struct gro_remcsum *grc)
2971{
b7fe10e5
TH
2972 void *ptr;
2973 size_t plen = grc->offset + sizeof(u16);
2974
26c4f7da
TH
2975 if (!grc->delta)
2976 return;
2977
b7fe10e5
TH
2978 ptr = skb_gro_header_fast(skb, grc->offset);
2979 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2980 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2981 if (!ptr)
2982 return;
2983 }
2984
2985 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2986}
dcdc8994 2987
25393d3f 2988#ifdef CONFIG_XFRM_OFFLOAD
d4546c25 2989static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
25393d3f
SK
2990{
2991 if (PTR_ERR(pp) != -EINPROGRESS)
2992 NAPI_GRO_CB(skb)->flush |= flush;
2993}
603d4cf8 2994static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
5cd3da4b 2995 struct sk_buff *pp,
603d4cf8
SD
2996 int flush,
2997 struct gro_remcsum *grc)
2998{
2999 if (PTR_ERR(pp) != -EINPROGRESS) {
3000 NAPI_GRO_CB(skb)->flush |= flush;
3001 skb_gro_remcsum_cleanup(skb, grc);
3002 skb->remcsum_offload = 0;
3003 }
3004}
25393d3f 3005#else
d4546c25 3006static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
5f114163
SK
3007{
3008 NAPI_GRO_CB(skb)->flush |= flush;
3009}
603d4cf8 3010static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
5cd3da4b 3011 struct sk_buff *pp,
603d4cf8
SD
3012 int flush,
3013 struct gro_remcsum *grc)
3014{
3015 NAPI_GRO_CB(skb)->flush |= flush;
3016 skb_gro_remcsum_cleanup(skb, grc);
3017 skb->remcsum_offload = 0;
3018}
25393d3f 3019#endif
5f114163 3020
0c4e8581
SH
3021static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3022 unsigned short type,
3b04ddde 3023 const void *daddr, const void *saddr,
95c96174 3024 unsigned int len)
0c4e8581 3025{
f1ecfd5d 3026 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 3027 return 0;
3b04ddde
SH
3028
3029 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
3030}
3031
b95cce35
SH
3032static inline int dev_parse_header(const struct sk_buff *skb,
3033 unsigned char *haddr)
3034{
3035 const struct net_device *dev = skb->dev;
3036
1b83336b 3037 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 3038 return 0;
3b04ddde 3039 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
3040}
3041
e78b2915
MM
3042static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3043{
3044 const struct net_device *dev = skb->dev;
3045
3046 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3047 return 0;
3048 return dev->header_ops->parse_protocol(skb);
3049}
3050
2793a23a
WB
3051/* ll_header must have at least hard_header_len allocated */
3052static inline bool dev_validate_header(const struct net_device *dev,
3053 char *ll_header, int len)
3054{
3055 if (likely(len >= dev->hard_header_len))
3056 return true;
217e6fa2
WB
3057 if (len < dev->min_header_len)
3058 return false;
2793a23a
WB
3059
3060 if (capable(CAP_SYS_RAWIO)) {
3061 memset(ll_header + len, 0, dev->hard_header_len - len);
3062 return true;
3063 }
3064
3065 if (dev->header_ops && dev->header_ops->validate)
3066 return dev->header_ops->validate(ll_header, len);
3067
3068 return false;
3069}
3070
36fd633e
AV
3071typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3072 int len, int size);
f629d208 3073int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
3074static inline int unregister_gifconf(unsigned int family)
3075{
3076 return register_gifconf(family, NULL);
3077}
3078
99bbc707 3079#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 3080#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
3081struct sd_flow_limit {
3082 u64 count;
3083 unsigned int num_buckets;
3084 unsigned int history_head;
3085 u16 history[FLOW_LIMIT_HISTORY];
3086 u8 buckets[];
3087};
3088
3089extern int netdev_flow_limit_table_len;
3090#endif /* CONFIG_NET_FLOW_LIMIT */
3091
1da177e4 3092/*
5e82b4b2 3093 * Incoming packets are placed on per-CPU queues
1da177e4 3094 */
d94d9fee 3095struct softnet_data {
1da177e4 3096 struct list_head poll_list;
6e7676c1 3097 struct sk_buff_head process_queue;
1da177e4 3098
dee42870 3099 /* stats */
cd7b5396
DM
3100 unsigned int processed;
3101 unsigned int time_squeeze;
cd7b5396 3102 unsigned int received_rps;
fd793d89 3103#ifdef CONFIG_RPS
88751275 3104 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
3105#endif
3106#ifdef CONFIG_NET_FLOW_LIMIT
3107 struct sd_flow_limit __rcu *flow_limit;
3108#endif
3109 struct Qdisc *output_queue;
3110 struct Qdisc **output_queue_tailp;
3111 struct sk_buff *completion_queue;
f53c7239
SK
3112#ifdef CONFIG_XFRM_OFFLOAD
3113 struct sk_buff_head xfrm_backlog;
3114#endif
97cdcf37
FW
3115 /* written and read only by owning cpu: */
3116 struct {
3117 u16 recursion;
3118 u8 more;
3119 } xmit;
4cdb1e2e 3120#ifdef CONFIG_RPS
501e7ef5
ED
3121 /* input_queue_head should be written by cpu owning this struct,
3122 * and only read by other cpus. Worth using a cache line.
3123 */
3124 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3125
3126 /* Elements below can be accessed between CPUs for RPS/RFS */
966a9671 3127 call_single_data_t csd ____cacheline_aligned_in_smp;
88751275
ED
3128 struct softnet_data *rps_ipi_next;
3129 unsigned int cpu;
76cc8b13 3130 unsigned int input_queue_tail;
1e94d72f 3131#endif
95c96174 3132 unsigned int dropped;
0a9627f2 3133 struct sk_buff_head input_pkt_queue;
bea3348e 3134 struct napi_struct backlog;
99bbc707 3135
1da177e4
LT
3136};
3137
76cc8b13 3138static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
3139{
3140#ifdef CONFIG_RPS
76cc8b13
TH
3141 sd->input_queue_head++;
3142#endif
3143}
3144
3145static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3146 unsigned int *qtail)
3147{
3148#ifdef CONFIG_RPS
3149 *qtail = ++sd->input_queue_tail;
fec5e652
TH
3150#endif
3151}
3152
0a9627f2 3153DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 3154
97cdcf37
FW
3155static inline int dev_recursion_level(void)
3156{
28b05b92 3157 return this_cpu_read(softnet_data.xmit.recursion);
97cdcf37
FW
3158}
3159
3160#define XMIT_RECURSION_LIMIT 10
3161static inline bool dev_xmit_recursion(void)
3162{
3163 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3164 XMIT_RECURSION_LIMIT);
3165}
3166
3167static inline void dev_xmit_recursion_inc(void)
3168{
3169 __this_cpu_inc(softnet_data.xmit.recursion);
3170}
3171
3172static inline void dev_xmit_recursion_dec(void)
3173{
3174 __this_cpu_dec(softnet_data.xmit.recursion);
3175}
3176
f629d208 3177void __netif_schedule(struct Qdisc *q);
46e5da40 3178void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 3179
fd2ea0a7
DM
3180static inline void netif_tx_schedule_all(struct net_device *dev)
3181{
3182 unsigned int i;
3183
3184 for (i = 0; i < dev->num_tx_queues; i++)
3185 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3186}
3187
f9a7cbbf 3188static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 3189{
73466498 3190 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
3191}
3192
bea3348e
SH
3193/**
3194 * netif_start_queue - allow transmit
3195 * @dev: network device
3196 *
3197 * Allow upper layers to call the device hard_start_xmit routine.
3198 */
1da177e4
LT
3199static inline void netif_start_queue(struct net_device *dev)
3200{
e8a0464c 3201 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3202}
3203
fd2ea0a7
DM
3204static inline void netif_tx_start_all_queues(struct net_device *dev)
3205{
3206 unsigned int i;
3207
3208 for (i = 0; i < dev->num_tx_queues; i++) {
3209 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3210 netif_tx_start_queue(txq);
3211 }
3212}
3213
46e5da40 3214void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 3215
d29f749e
DJ
3216/**
3217 * netif_wake_queue - restart transmit
3218 * @dev: network device
3219 *
3220 * Allow upper layers to call the device hard_start_xmit routine.
3221 * Used for flow control when transmit resources are available.
3222 */
79d16385
DM
3223static inline void netif_wake_queue(struct net_device *dev)
3224{
e8a0464c 3225 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3226}
3227
fd2ea0a7
DM
3228static inline void netif_tx_wake_all_queues(struct net_device *dev)
3229{
3230 unsigned int i;
3231
3232 for (i = 0; i < dev->num_tx_queues; i++) {
3233 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3234 netif_tx_wake_queue(txq);
3235 }
3236}
3237
f9a7cbbf 3238static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 3239{
73466498 3240 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
3241}
3242
bea3348e
SH
3243/**
3244 * netif_stop_queue - stop transmitted packets
3245 * @dev: network device
3246 *
3247 * Stop upper layers calling the device hard_start_xmit routine.
3248 * Used for flow control when transmit resources are unavailable.
3249 */
1da177e4
LT
3250static inline void netif_stop_queue(struct net_device *dev)
3251{
e8a0464c 3252 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3253}
3254
a2029240 3255void netif_tx_stop_all_queues(struct net_device *dev);
ab92d68f 3256void netdev_update_lockdep_key(struct net_device *dev);
fd2ea0a7 3257
4d29515f 3258static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 3259{
73466498 3260 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
3261}
3262
bea3348e
SH
3263/**
3264 * netif_queue_stopped - test if transmit queue is flowblocked
3265 * @dev: network device
3266 *
3267 * Test if transmit queue on device is currently unable to send.
3268 */
4d29515f 3269static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 3270{
e8a0464c 3271 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
3272}
3273
4d29515f 3274static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 3275{
73466498
TH
3276 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3277}
3278
8e2f1a63
DB
3279static inline bool
3280netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
3281{
3282 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3283}
3284
8e2f1a63
DB
3285static inline bool
3286netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3287{
3288 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3289}
3290
53511453
ED
3291/**
3292 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3293 * @dev_queue: pointer to transmit queue
3294 *
3295 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 3296 * to give appropriate hint to the CPU.
53511453
ED
3297 */
3298static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3299{
3300#ifdef CONFIG_BQL
3301 prefetchw(&dev_queue->dql.num_queued);
3302#endif
3303}
3304
3305/**
3306 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3307 * @dev_queue: pointer to transmit queue
3308 *
3309 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 3310 * to give appropriate hint to the CPU.
53511453
ED
3311 */
3312static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3313{
3314#ifdef CONFIG_BQL
3315 prefetchw(&dev_queue->dql.limit);
3316#endif
3317}
3318
c5d67bd7
TH
3319static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3320 unsigned int bytes)
3321{
114cf580
TH
3322#ifdef CONFIG_BQL
3323 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
3324
3325 if (likely(dql_avail(&dev_queue->dql) >= 0))
3326 return;
3327
3328 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3329
3330 /*
3331 * The XOFF flag must be set before checking the dql_avail below,
3332 * because in netdev_tx_completed_queue we update the dql_completed
3333 * before checking the XOFF flag.
3334 */
3335 smp_mb();
3336
3337 /* check again in case another CPU has just made room avail */
3338 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3339 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 3340#endif
c5d67bd7
TH
3341}
3342
3e59020a
ED
3343/* Variant of netdev_tx_sent_queue() for drivers that are aware
3344 * that they should not test BQL status themselves.
3345 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3346 * skb of a batch.
3347 * Returns true if the doorbell must be used to kick the NIC.
3348 */
3349static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3350 unsigned int bytes,
3351 bool xmit_more)
3352{
3353 if (xmit_more) {
3354#ifdef CONFIG_BQL
3355 dql_queued(&dev_queue->dql, bytes);
3356#endif
3357 return netif_tx_queue_stopped(dev_queue);
3358 }
3359 netdev_tx_sent_queue(dev_queue, bytes);
3360 return true;
3361}
3362
0042d0c8
FF
3363/**
3364 * netdev_sent_queue - report the number of bytes queued to hardware
3365 * @dev: network device
3366 * @bytes: number of bytes queued to the hardware device queue
3367 *
3368 * Report the number of bytes queued for sending/completion to the network
3369 * device hardware queue. @bytes should be a good approximation and should
3370 * exactly match netdev_completed_queue() @bytes
3371 */
c5d67bd7
TH
3372static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3373{
3374 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3375}
3376
620344c4
HK
3377static inline bool __netdev_sent_queue(struct net_device *dev,
3378 unsigned int bytes,
3379 bool xmit_more)
3380{
3381 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3382 xmit_more);
3383}
3384
c5d67bd7 3385static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 3386 unsigned int pkts, unsigned int bytes)
c5d67bd7 3387{
114cf580 3388#ifdef CONFIG_BQL
b37c0fbe
AD
3389 if (unlikely(!bytes))
3390 return;
3391
3392 dql_completed(&dev_queue->dql, bytes);
3393
3394 /*
3395 * Without the memory barrier there is a small possiblity that
3396 * netdev_tx_sent_queue will miss the update and cause the queue to
3397 * be stopped forever
3398 */
3399 smp_mb();
3400
f3acd33d 3401 if (unlikely(dql_avail(&dev_queue->dql) < 0))
b37c0fbe
AD
3402 return;
3403
3404 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3405 netif_schedule_queue(dev_queue);
114cf580 3406#endif
c5d67bd7
TH
3407}
3408
0042d0c8
FF
3409/**
3410 * netdev_completed_queue - report bytes and packets completed by device
3411 * @dev: network device
3412 * @pkts: actual number of packets sent over the medium
3413 * @bytes: actual number of bytes sent over the medium
3414 *
3415 * Report the number of bytes and packets transmitted by the network device
3416 * hardware queue over the physical medium, @bytes must exactly match the
3417 * @bytes amount passed to netdev_sent_queue()
3418 */
c5d67bd7 3419static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3420 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3421{
3422 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3423}
3424
3425static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3426{
114cf580 3427#ifdef CONFIG_BQL
5c490354 3428 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3429 dql_reset(&q->dql);
3430#endif
c5d67bd7
TH
3431}
3432
0042d0c8
FF
3433/**
3434 * netdev_reset_queue - reset the packets and bytes count of a network device
3435 * @dev_queue: network device
3436 *
3437 * Reset the bytes and packet count of a network device and clear the
3438 * software flow control OFF bit for this network device
3439 */
c5d67bd7
TH
3440static inline void netdev_reset_queue(struct net_device *dev_queue)
3441{
3442 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3443}
3444
b9507bda
DB
3445/**
3446 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3447 * @dev: network device
3448 * @queue_index: given tx queue index
3449 *
3450 * Returns 0 if given tx queue index >= number of device tx queues,
3451 * otherwise returns the originally passed tx queue index.
3452 */
3453static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3454{
3455 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3456 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3457 dev->name, queue_index,
3458 dev->real_num_tx_queues);
3459 return 0;
3460 }
3461
3462 return queue_index;
3463}
3464
bea3348e
SH
3465/**
3466 * netif_running - test if up
3467 * @dev: network device
3468 *
3469 * Test if the device has been brought up.
3470 */
4d29515f 3471static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3472{
3473 return test_bit(__LINK_STATE_START, &dev->state);
3474}
3475
f25f4e44 3476/*
5e82b4b2 3477 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3478 * stop, and a check if it's stopped. All other device management is
3479 * done at the overall netdevice level.
3480 * Also test the device if we're multiqueue.
3481 */
bea3348e
SH
3482
3483/**
3484 * netif_start_subqueue - allow sending packets on subqueue
3485 * @dev: network device
3486 * @queue_index: sub queue index
3487 *
3488 * Start individual transmit queue of a device with multiple transmit queues.
3489 */
f25f4e44
PWJ
3490static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3491{
fd2ea0a7 3492 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3493
3494 netif_tx_start_queue(txq);
f25f4e44
PWJ
3495}
3496
bea3348e
SH
3497/**
3498 * netif_stop_subqueue - stop sending packets on subqueue
3499 * @dev: network device
3500 * @queue_index: sub queue index
3501 *
3502 * Stop individual transmit queue of a device with multiple transmit queues.
3503 */
f25f4e44
PWJ
3504static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3505{
fd2ea0a7 3506 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3507 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3508}
3509
bea3348e
SH
3510/**
3511 * netif_subqueue_stopped - test status of subqueue
3512 * @dev: network device
3513 * @queue_index: sub queue index
3514 *
3515 * Check individual transmit queue of a device with multiple transmit queues.
3516 */
4d29515f
DM
3517static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3518 u16 queue_index)
f25f4e44 3519{
fd2ea0a7 3520 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3521
3522 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3523}
3524
4d29515f
DM
3525static inline bool netif_subqueue_stopped(const struct net_device *dev,
3526 struct sk_buff *skb)
668f895a
PE
3527{
3528 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3529}
bea3348e 3530
738b35cc
FF
3531/**
3532 * netif_wake_subqueue - allow sending packets on subqueue
3533 * @dev: network device
3534 * @queue_index: sub queue index
3535 *
3536 * Resume individual transmit queue of a device with multiple transmit queues.
3537 */
3538static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3539{
3540 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3541
3542 netif_tx_wake_queue(txq);
3543}
f25f4e44 3544
537c00de 3545#ifdef CONFIG_XPS
53af53ae 3546int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3547 u16 index);
80d19669
AN
3548int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3549 u16 index, bool is_rxqs_map);
3550
3551/**
3552 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3553 * @j: CPU/Rx queue index
3554 * @mask: bitmask of all cpus/rx queues
3555 * @nr_bits: number of bits in the bitmask
3556 *
3557 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3558 */
3559static inline bool netif_attr_test_mask(unsigned long j,
3560 const unsigned long *mask,
3561 unsigned int nr_bits)
3562{
3563 cpu_max_bits_warn(j, nr_bits);
3564 return test_bit(j, mask);
3565}
3566
3567/**
3568 * netif_attr_test_online - Test for online CPU/Rx queue
3569 * @j: CPU/Rx queue index
3570 * @online_mask: bitmask for CPUs/Rx queues that are online
3571 * @nr_bits: number of bits in the bitmask
3572 *
3573 * Returns true if a CPU/Rx queue is online.
3574 */
3575static inline bool netif_attr_test_online(unsigned long j,
3576 const unsigned long *online_mask,
3577 unsigned int nr_bits)
3578{
3579 cpu_max_bits_warn(j, nr_bits);
3580
3581 if (online_mask)
3582 return test_bit(j, online_mask);
3583
3584 return (j < nr_bits);
3585}
3586
3587/**
3588 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3589 * @n: CPU/Rx queue index
3590 * @srcp: the cpumask/Rx queue mask pointer
3591 * @nr_bits: number of bits in the bitmask
3592 *
3593 * Returns >= nr_bits if no further CPUs/Rx queues set.
3594 */
3595static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3596 unsigned int nr_bits)
3597{
3598 /* -1 is a legal arg here. */
3599 if (n != -1)
3600 cpu_max_bits_warn(n, nr_bits);
3601
3602 if (srcp)
3603 return find_next_bit(srcp, nr_bits, n + 1);
3604
3605 return n + 1;
3606}
3607
3608/**
a1fa83bd 3609 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
80d19669
AN
3610 * @n: CPU/Rx queue index
3611 * @src1p: the first CPUs/Rx queues mask pointer
3612 * @src2p: the second CPUs/Rx queues mask pointer
3613 * @nr_bits: number of bits in the bitmask
3614 *
3615 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3616 */
3617static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3618 const unsigned long *src2p,
3619 unsigned int nr_bits)
3620{
3621 /* -1 is a legal arg here. */
3622 if (n != -1)
3623 cpu_max_bits_warn(n, nr_bits);
3624
3625 if (src1p && src2p)
3626 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3627 else if (src1p)
3628 return find_next_bit(src1p, nr_bits, n + 1);
3629 else if (src2p)
3630 return find_next_bit(src2p, nr_bits, n + 1);
3631
3632 return n + 1;
3633}
537c00de
AD
3634#else
3635static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3636 const struct cpumask *mask,
537c00de
AD
3637 u16 index)
3638{
3639 return 0;
3640}
c9fbb2d2
KK
3641
3642static inline int __netif_set_xps_queue(struct net_device *dev,
3643 const unsigned long *mask,
3644 u16 index, bool is_rxqs_map)
3645{
3646 return 0;
3647}
537c00de
AD
3648#endif
3649
bea3348e
SH
3650/**
3651 * netif_is_multiqueue - test if device has multiple transmit queues
3652 * @dev: network device
3653 *
3654 * Check if device has multiple transmit queues
bea3348e 3655 */
4d29515f 3656static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3657{
a02cec21 3658 return dev->num_tx_queues > 1;
f25f4e44 3659}
1da177e4 3660
f629d208 3661int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3662
a953be53 3663#ifdef CONFIG_SYSFS
f629d208 3664int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3665#else
3666static inline int netif_set_real_num_rx_queues(struct net_device *dev,
c29c2ebd 3667 unsigned int rxqs)
62fe0b40 3668{
c29c2ebd 3669 dev->real_num_rx_queues = rxqs;
62fe0b40
BH
3670 return 0;
3671}
3672#endif
3673
65073a67
DB
3674static inline struct netdev_rx_queue *
3675__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3676{
3677 return dev->_rx + rxq;
3678}
3679
a953be53
MD
3680#ifdef CONFIG_SYSFS
3681static inline unsigned int get_netdev_rx_queue_index(
3682 struct netdev_rx_queue *queue)
3683{
3684 struct net_device *dev = queue->dev;
3685 int index = queue - dev->_rx;
3686
3687 BUG_ON(index >= dev->num_rx_queues);
3688 return index;
3689}
3690#endif
3691
16917b87 3692#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3693int netif_get_num_default_rss_queues(void);
16917b87 3694
e6247027
ED
3695enum skb_free_reason {
3696 SKB_REASON_CONSUMED,
3697 SKB_REASON_DROPPED,
3698};
3699
3700void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3701void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3702
e6247027
ED
3703/*
3704 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3705 * interrupt context or with hardware interrupts being disabled.
3706 * (in_irq() || irqs_disabled())
3707 *
3708 * We provide four helpers that can be used in following contexts :
3709 *
3710 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3711 * replacing kfree_skb(skb)
3712 *
3713 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3714 * Typically used in place of consume_skb(skb) in TX completion path
3715 *
3716 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3717 * replacing kfree_skb(skb)
3718 *
3719 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3720 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3721 */
e6247027
ED
3722static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3723{
3724 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3725}
3726
3727static inline void dev_consume_skb_irq(struct sk_buff *skb)
3728{
3729 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3730}
3731
3732static inline void dev_kfree_skb_any(struct sk_buff *skb)
3733{
3734 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3735}
3736
3737static inline void dev_consume_skb_any(struct sk_buff *skb)
3738{
3739 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3740}
1da177e4 3741
7c497478
JW
3742void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3743int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
f629d208
JP
3744int netif_rx(struct sk_buff *skb);
3745int netif_rx_ni(struct sk_buff *skb);
04eb4489 3746int netif_receive_skb(struct sk_buff *skb);
1c601d82 3747int netif_receive_skb_core(struct sk_buff *skb);
f6ad8c1b 3748void netif_receive_skb_list(struct list_head *head);
f629d208
JP
3749gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3750void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3751struct sk_buff *napi_get_frags(struct napi_struct *napi);
3752gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3753struct packet_offload *gro_find_receive_by_type(__be16 type);
3754struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3755
3756static inline void napi_free_frags(struct napi_struct *napi)
3757{
3758 kfree_skb(napi->skb);
3759 napi->skb = NULL;
3760}
3761
24b27fc4 3762bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3763int netdev_rx_handler_register(struct net_device *dev,
3764 rx_handler_func_t *rx_handler,
3765 void *rx_handler_data);
3766void netdev_rx_handler_unregister(struct net_device *dev);
3767
3768bool dev_valid_name(const char *name);
44c02a2c
AV
3769int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3770 bool *need_copyout);
36fd633e 3771int dev_ifconf(struct net *net, struct ifconf *, int);
f629d208
JP
3772int dev_ethtool(struct net *net, struct ifreq *);
3773unsigned int dev_get_flags(const struct net_device *);
6d040321
PM
3774int __dev_change_flags(struct net_device *dev, unsigned int flags,
3775 struct netlink_ext_ack *extack);
567c5e13
PM
3776int dev_change_flags(struct net_device *dev, unsigned int flags,
3777 struct netlink_ext_ack *extack);
cb178190
DM
3778void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3779 unsigned int gchanges);
f629d208
JP
3780int dev_change_name(struct net_device *, const char *);
3781int dev_set_alias(struct net_device *, const char *, size_t);
6c557001 3782int dev_get_alias(const struct net_device *, char *, size_t);
f629d208 3783int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3784int __dev_set_mtu(struct net_device *, int);
d836f5c6
ED
3785int dev_validate_mtu(struct net_device *dev, int mtu,
3786 struct netlink_ext_ack *extack);
7a4c53be
SH
3787int dev_set_mtu_ext(struct net_device *dev, int mtu,
3788 struct netlink_ext_ack *extack);
f629d208 3789int dev_set_mtu(struct net_device *, int);
6a643ddb 3790int dev_change_tx_queue_len(struct net_device *, unsigned long);
f629d208 3791void dev_set_group(struct net_device *, int);
d59cdf94
PM
3792int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3793 struct netlink_ext_ack *extack);
3a37a963
PM
3794int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3795 struct netlink_ext_ack *extack);
f629d208
JP
3796int dev_change_carrier(struct net_device *, bool new_carrier);
3797int dev_get_phys_port_id(struct net_device *dev,
02637fce 3798 struct netdev_phys_item_id *ppid);
db24a904
DA
3799int dev_get_phys_port_name(struct net_device *dev,
3800 char *name, size_t len);
d6abc596
FF
3801int dev_get_port_parent_id(struct net_device *dev,
3802 struct netdev_phys_item_id *ppid, bool recurse);
3803bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
d746d707 3804int dev_change_proto_down(struct net_device *dev, bool proto_down);
b5899679 3805int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
f53c7239 3806struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
ce93718f
DM
3807struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3808 struct netdev_queue *txq, int *ret);
d67b9cd2 3809
f4e63525 3810typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
d67b9cd2 3811int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
92234c8f 3812 int fd, int expected_fd, u32 flags);
a25717d2
JK
3813u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3814 enum bpf_netdev_command cmd);
84c6b868 3815int xdp_umem_query(struct net_device *dev, u16 queue_id);
d67b9cd2 3816
a0265d28 3817int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3818int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3819bool is_skb_forwardable(const struct net_device *dev,
3820 const struct sk_buff *skb);
1da177e4 3821
4e3264d2
MKL
3822static __always_inline int ____dev_forward_skb(struct net_device *dev,
3823 struct sk_buff *skb)
3824{
3825 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3826 unlikely(!is_skb_forwardable(dev, skb))) {
3827 atomic_long_inc(&dev->rx_dropped);
3828 kfree_skb(skb);
3829 return NET_RX_DROP;
3830 }
3831
3832 skb_scrub_packet(skb, true);
3833 skb->priority = 0;
3834 return 0;
3835}
3836
9f9a742d 3837bool dev_nit_active(struct net_device *dev);
74b20582
DA
3838void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3839
20380731 3840extern int netdev_budget;
7acf8a1e 3841extern unsigned int netdev_budget_usecs;
1da177e4
LT
3842
3843/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3844void netdev_run_todo(void);
1da177e4 3845
bea3348e
SH
3846/**
3847 * dev_put - release reference to device
3848 * @dev: network device
3849 *
9ef4429b 3850 * Release reference to device to allow it to be freed.
bea3348e 3851 */
1da177e4
LT
3852static inline void dev_put(struct net_device *dev)
3853{
933393f5 3854 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3855}
3856
bea3348e
SH
3857/**
3858 * dev_hold - get reference to device
3859 * @dev: network device
3860 *
9ef4429b 3861 * Hold reference to device to keep it from being freed.
bea3348e 3862 */
15333061
SH
3863static inline void dev_hold(struct net_device *dev)
3864{
933393f5 3865 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3866}
1da177e4
LT
3867
3868/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3869 * and _off may be called from IRQ context, but it is caller
3870 * who is responsible for serialization of these calls.
b00055aa
SR
3871 *
3872 * The name carrier is inappropriate, these functions should really be
3873 * called netif_lowerlayer_*() because they represent the state of any
3874 * kind of lower layer not just hardware media.
1da177e4
LT
3875 */
3876
f629d208
JP
3877void linkwatch_init_dev(struct net_device *dev);
3878void linkwatch_fire_event(struct net_device *dev);
3879void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3880
bea3348e
SH
3881/**
3882 * netif_carrier_ok - test if carrier present
3883 * @dev: network device
3884 *
3885 * Check if carrier is present on device
3886 */
4d29515f 3887static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3888{
3889 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3890}
3891
f629d208 3892unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3893
f629d208 3894void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3895
f629d208 3896void netif_carrier_on(struct net_device *dev);
1da177e4 3897
f629d208 3898void netif_carrier_off(struct net_device *dev);
1da177e4 3899
bea3348e
SH
3900/**
3901 * netif_dormant_on - mark device as dormant.
3902 * @dev: network device
3903 *
3904 * Mark device as dormant (as per RFC2863).
3905 *
3906 * The dormant state indicates that the relevant interface is not
3907 * actually in a condition to pass packets (i.e., it is not 'up') but is
3908 * in a "pending" state, waiting for some external event. For "on-
3909 * demand" interfaces, this new state identifies the situation where the
3910 * interface is waiting for events to place it in the up state.
bea3348e 3911 */
b00055aa
SR
3912static inline void netif_dormant_on(struct net_device *dev)
3913{
3914 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3915 linkwatch_fire_event(dev);
3916}
3917
bea3348e
SH
3918/**
3919 * netif_dormant_off - set device as not dormant.
3920 * @dev: network device
3921 *
3922 * Device is not in dormant state.
3923 */
b00055aa
SR
3924static inline void netif_dormant_off(struct net_device *dev)
3925{
3926 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3927 linkwatch_fire_event(dev);
3928}
3929
bea3348e 3930/**
8ecbc40a 3931 * netif_dormant - test if device is dormant
bea3348e
SH
3932 * @dev: network device
3933 *
8ecbc40a 3934 * Check if device is dormant.
bea3348e 3935 */
4d29515f 3936static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3937{
3938 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3939}
3940
3941
eec517cd
AL
3942/**
3943 * netif_testing_on - mark device as under test.
3944 * @dev: network device
3945 *
3946 * Mark device as under test (as per RFC2863).
3947 *
3948 * The testing state indicates that some test(s) must be performed on
3949 * the interface. After completion, of the test, the interface state
3950 * will change to up, dormant, or down, as appropriate.
3951 */
3952static inline void netif_testing_on(struct net_device *dev)
3953{
3954 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
3955 linkwatch_fire_event(dev);
3956}
3957
3958/**
3959 * netif_testing_off - set device as not under test.
3960 * @dev: network device
3961 *
3962 * Device is not in testing state.
3963 */
3964static inline void netif_testing_off(struct net_device *dev)
3965{
3966 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
3967 linkwatch_fire_event(dev);
3968}
3969
3970/**
3971 * netif_testing - test if device is under test
3972 * @dev: network device
3973 *
3974 * Check if device is under test
3975 */
3976static inline bool netif_testing(const struct net_device *dev)
3977{
3978 return test_bit(__LINK_STATE_TESTING, &dev->state);
3979}
3980
3981
bea3348e
SH
3982/**
3983 * netif_oper_up - test if device is operational
3984 * @dev: network device
3985 *
3986 * Check if carrier is operational
3987 */
4d29515f 3988static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3989{
b00055aa
SR
3990 return (dev->operstate == IF_OPER_UP ||
3991 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3992}
3993
bea3348e
SH
3994/**
3995 * netif_device_present - is device available or removed
3996 * @dev: network device
3997 *
3998 * Check if device has not been removed from system.
3999 */
4d29515f 4000static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
4001{
4002 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4003}
4004
f629d208 4005void netif_device_detach(struct net_device *dev);
1da177e4 4006
f629d208 4007void netif_device_attach(struct net_device *dev);
1da177e4
LT
4008
4009/*
4010 * Network interface message level settings
4011 */
1da177e4
LT
4012
4013enum {
6a94b8cc
MK
4014 NETIF_MSG_DRV_BIT,
4015 NETIF_MSG_PROBE_BIT,
4016 NETIF_MSG_LINK_BIT,
4017 NETIF_MSG_TIMER_BIT,
4018 NETIF_MSG_IFDOWN_BIT,
4019 NETIF_MSG_IFUP_BIT,
4020 NETIF_MSG_RX_ERR_BIT,
4021 NETIF_MSG_TX_ERR_BIT,
4022 NETIF_MSG_TX_QUEUED_BIT,
4023 NETIF_MSG_INTR_BIT,
4024 NETIF_MSG_TX_DONE_BIT,
4025 NETIF_MSG_RX_STATUS_BIT,
4026 NETIF_MSG_PKTDATA_BIT,
4027 NETIF_MSG_HW_BIT,
4028 NETIF_MSG_WOL_BIT,
4029
4030 /* When you add a new bit above, update netif_msg_class_names array
4031 * in net/ethtool/common.c
4032 */
4033 NETIF_MSG_CLASS_COUNT,
1da177e4 4034};
6a94b8cc
MK
4035/* Both ethtool_ops interface and internal driver implementation use u32 */
4036static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4037
4038#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4039#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4040
4041#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4042#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4043#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4044#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4045#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4046#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4047#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4048#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4049#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4050#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4051#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4052#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4053#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4054#define NETIF_MSG_HW __NETIF_MSG(HW)
4055#define NETIF_MSG_WOL __NETIF_MSG(WOL)
1da177e4
LT
4056
4057#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4058#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4059#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4060#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4061#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4062#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4063#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4064#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4065#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4066#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4067#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4068#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4069#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4070#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4071#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4072
4073static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4074{
4075 /* use default */
4076 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4077 return default_msg_enable_bits;
4078 if (debug_value == 0) /* no output */
4079 return 0;
4080 /* set low N bits */
f4d7b3e2 4081 return (1U << debug_value) - 1;
1da177e4
LT
4082}
4083
c773e847 4084static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 4085{
c773e847
DM
4086 spin_lock(&txq->_xmit_lock);
4087 txq->xmit_lock_owner = cpu;
22dd7495
JHS
4088}
4089
5a717f4f
MT
4090static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4091{
4092 __acquire(&txq->_xmit_lock);
4093 return true;
4094}
4095
4096static inline void __netif_tx_release(struct netdev_queue *txq)
4097{
4098 __release(&txq->_xmit_lock);
4099}
4100
fd2ea0a7
DM
4101static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4102{
4103 spin_lock_bh(&txq->_xmit_lock);
4104 txq->xmit_lock_owner = smp_processor_id();
4105}
4106
4d29515f 4107static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 4108{
4d29515f 4109 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
4110 if (likely(ok))
4111 txq->xmit_lock_owner = smp_processor_id();
4112 return ok;
4113}
4114
4115static inline void __netif_tx_unlock(struct netdev_queue *txq)
4116{
4117 txq->xmit_lock_owner = -1;
4118 spin_unlock(&txq->_xmit_lock);
4119}
4120
4121static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4122{
4123 txq->xmit_lock_owner = -1;
4124 spin_unlock_bh(&txq->_xmit_lock);
4125}
4126
08baf561
ED
4127static inline void txq_trans_update(struct netdev_queue *txq)
4128{
4129 if (txq->xmit_lock_owner != -1)
4130 txq->trans_start = jiffies;
4131}
4132
ba162f8e
FW
4133/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4134static inline void netif_trans_update(struct net_device *dev)
4135{
9b36627a
FW
4136 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4137
4138 if (txq->trans_start != jiffies)
4139 txq->trans_start = jiffies;
ba162f8e
FW
4140}
4141
d29f749e
DJ
4142/**
4143 * netif_tx_lock - grab network device transmit lock
4144 * @dev: network device
d29f749e
DJ
4145 *
4146 * Get network device transmit lock
4147 */
22dd7495
JHS
4148static inline void netif_tx_lock(struct net_device *dev)
4149{
e8a0464c 4150 unsigned int i;
c3f26a26 4151 int cpu;
c773e847 4152
c3f26a26
DM
4153 spin_lock(&dev->tx_global_lock);
4154 cpu = smp_processor_id();
e8a0464c
DM
4155 for (i = 0; i < dev->num_tx_queues; i++) {
4156 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
4157
4158 /* We are the only thread of execution doing a
4159 * freeze, but we have to grab the _xmit_lock in
4160 * order to synchronize with threads which are in
4161 * the ->hard_start_xmit() handler and already
4162 * checked the frozen bit.
4163 */
e8a0464c 4164 __netif_tx_lock(txq, cpu);
c3f26a26
DM
4165 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4166 __netif_tx_unlock(txq);
e8a0464c 4167 }
932ff279
HX
4168}
4169
4170static inline void netif_tx_lock_bh(struct net_device *dev)
4171{
e8a0464c
DM
4172 local_bh_disable();
4173 netif_tx_lock(dev);
932ff279
HX
4174}
4175
932ff279
HX
4176static inline void netif_tx_unlock(struct net_device *dev)
4177{
e8a0464c
DM
4178 unsigned int i;
4179
4180 for (i = 0; i < dev->num_tx_queues; i++) {
4181 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 4182
c3f26a26
DM
4183 /* No need to grab the _xmit_lock here. If the
4184 * queue is not stopped for another reason, we
4185 * force a schedule.
4186 */
4187 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 4188 netif_schedule_queue(txq);
c3f26a26
DM
4189 }
4190 spin_unlock(&dev->tx_global_lock);
932ff279
HX
4191}
4192
4193static inline void netif_tx_unlock_bh(struct net_device *dev)
4194{
e8a0464c
DM
4195 netif_tx_unlock(dev);
4196 local_bh_enable();
932ff279
HX
4197}
4198
c773e847 4199#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 4200 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 4201 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
4202 } else { \
4203 __netif_tx_acquire(txq); \
22dd7495
JHS
4204 } \
4205}
4206
5efeac44
EB
4207#define HARD_TX_TRYLOCK(dev, txq) \
4208 (((dev->features & NETIF_F_LLTX) == 0) ? \
4209 __netif_tx_trylock(txq) : \
5a717f4f 4210 __netif_tx_acquire(txq))
5efeac44 4211
c773e847 4212#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 4213 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 4214 __netif_tx_unlock(txq); \
5a717f4f
MT
4215 } else { \
4216 __netif_tx_release(txq); \
22dd7495
JHS
4217 } \
4218}
4219
1da177e4
LT
4220static inline void netif_tx_disable(struct net_device *dev)
4221{
fd2ea0a7 4222 unsigned int i;
c3f26a26 4223 int cpu;
fd2ea0a7 4224
c3f26a26
DM
4225 local_bh_disable();
4226 cpu = smp_processor_id();
fd2ea0a7
DM
4227 for (i = 0; i < dev->num_tx_queues; i++) {
4228 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
4229
4230 __netif_tx_lock(txq, cpu);
fd2ea0a7 4231 netif_tx_stop_queue(txq);
c3f26a26 4232 __netif_tx_unlock(txq);
fd2ea0a7 4233 }
c3f26a26 4234 local_bh_enable();
1da177e4
LT
4235}
4236
e308a5d8
DM
4237static inline void netif_addr_lock(struct net_device *dev)
4238{
4239 spin_lock(&dev->addr_list_lock);
4240}
4241
4242static inline void netif_addr_lock_bh(struct net_device *dev)
4243{
4244 spin_lock_bh(&dev->addr_list_lock);
4245}
4246
4247static inline void netif_addr_unlock(struct net_device *dev)
4248{
4249 spin_unlock(&dev->addr_list_lock);
4250}
4251
4252static inline void netif_addr_unlock_bh(struct net_device *dev)
4253{
4254 spin_unlock_bh(&dev->addr_list_lock);
4255}
4256
f001fde5 4257/*
31278e71 4258 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
4259 * rcu_read_lock held.
4260 */
4261#define for_each_dev_addr(dev, ha) \
31278e71 4262 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 4263
1da177e4
LT
4264/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4265
f629d208 4266void ether_setup(struct net_device *dev);
1da177e4
LT
4267
4268/* Support for loadable net-drivers */
f629d208 4269struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 4270 unsigned char name_assign_type,
f629d208
JP
4271 void (*setup)(struct net_device *),
4272 unsigned int txqs, unsigned int rxqs);
c835a677
TG
4273#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4274 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 4275
c835a677
TG
4276#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4277 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4278 count)
36909ea4 4279
f629d208
JP
4280int register_netdev(struct net_device *dev);
4281void unregister_netdev(struct net_device *dev);
f001fde5 4282
cd16627f
BG
4283int devm_register_netdev(struct device *dev, struct net_device *ndev);
4284
22bedad3 4285/* General hardware address lists handling functions */
f629d208
JP
4286int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4287 struct netdev_hw_addr_list *from_list, int addr_len);
4288void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4289 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
4290int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4291 struct net_device *dev,
4292 int (*sync)(struct net_device *, const unsigned char *),
4293 int (*unsync)(struct net_device *,
4294 const unsigned char *));
e7946760
IK
4295int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4296 struct net_device *dev,
4297 int (*sync)(struct net_device *,
4298 const unsigned char *, int),
4299 int (*unsync)(struct net_device *,
4300 const unsigned char *, int));
4301void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4302 struct net_device *dev,
4303 int (*unsync)(struct net_device *,
4304 const unsigned char *, int));
670e5b8e
AD
4305void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4306 struct net_device *dev,
4307 int (*unsync)(struct net_device *,
4308 const unsigned char *));
f629d208 4309void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 4310
f001fde5 4311/* Functions used for device addresses handling */
f629d208
JP
4312int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4313 unsigned char addr_type);
4314int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4315 unsigned char addr_type);
f629d208
JP
4316void dev_addr_flush(struct net_device *dev);
4317int dev_addr_init(struct net_device *dev);
a748ee24
JP
4318
4319/* Functions used for unicast addresses handling */
f629d208
JP
4320int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4321int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4322int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4323int dev_uc_sync(struct net_device *to, struct net_device *from);
4324int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4325void dev_uc_unsync(struct net_device *to, struct net_device *from);
4326void dev_uc_flush(struct net_device *dev);
4327void dev_uc_init(struct net_device *dev);
f001fde5 4328
670e5b8e
AD
4329/**
4330 * __dev_uc_sync - Synchonize device's unicast list
4331 * @dev: device to sync
4332 * @sync: function to call if address should be added
4333 * @unsync: function to call if address should be removed
4334 *
4335 * Add newly added addresses to the interface, and release
4336 * addresses that have been deleted.
5e82b4b2 4337 */
670e5b8e
AD
4338static inline int __dev_uc_sync(struct net_device *dev,
4339 int (*sync)(struct net_device *,
4340 const unsigned char *),
4341 int (*unsync)(struct net_device *,
4342 const unsigned char *))
4343{
4344 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4345}
4346
4347/**
e793c0f7 4348 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
4349 * @dev: device to sync
4350 * @unsync: function to call if address should be removed
4351 *
4352 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 4353 */
670e5b8e
AD
4354static inline void __dev_uc_unsync(struct net_device *dev,
4355 int (*unsync)(struct net_device *,
4356 const unsigned char *))
4357{
4358 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4359}
4360
22bedad3 4361/* Functions used for multicast addresses handling */
f629d208
JP
4362int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4363int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4364int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4365int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4366int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4367int dev_mc_sync(struct net_device *to, struct net_device *from);
4368int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4369void dev_mc_unsync(struct net_device *to, struct net_device *from);
4370void dev_mc_flush(struct net_device *dev);
4371void dev_mc_init(struct net_device *dev);
f001fde5 4372
670e5b8e
AD
4373/**
4374 * __dev_mc_sync - Synchonize device's multicast list
4375 * @dev: device to sync
4376 * @sync: function to call if address should be added
4377 * @unsync: function to call if address should be removed
4378 *
4379 * Add newly added addresses to the interface, and release
4380 * addresses that have been deleted.
5e82b4b2 4381 */
670e5b8e
AD
4382static inline int __dev_mc_sync(struct net_device *dev,
4383 int (*sync)(struct net_device *,
4384 const unsigned char *),
4385 int (*unsync)(struct net_device *,
4386 const unsigned char *))
4387{
4388 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4389}
4390
4391/**
e793c0f7 4392 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
4393 * @dev: device to sync
4394 * @unsync: function to call if address should be removed
4395 *
4396 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 4397 */
670e5b8e
AD
4398static inline void __dev_mc_unsync(struct net_device *dev,
4399 int (*unsync)(struct net_device *,
4400 const unsigned char *))
4401{
4402 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4403}
4404
4417da66 4405/* Functions used for secondary unicast and multicast support */
f629d208
JP
4406void dev_set_rx_mode(struct net_device *dev);
4407void __dev_set_rx_mode(struct net_device *dev);
4408int dev_set_promiscuity(struct net_device *dev, int inc);
4409int dev_set_allmulti(struct net_device *dev, int inc);
4410void netdev_state_change(struct net_device *dev);
4411void netdev_notify_peers(struct net_device *dev);
4412void netdev_features_change(struct net_device *dev);
1da177e4 4413/* Load a device via the kmod */
f629d208
JP
4414void dev_load(struct net *net, const char *name);
4415struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4416 struct rtnl_link_stats64 *storage);
4417void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4418 const struct net_device_stats *netdev_stats);
eeda3fd6 4419
1da177e4 4420extern int netdev_max_backlog;
3b098e2d 4421extern int netdev_tstamp_prequeue;
1da177e4 4422extern int weight_p;
3d48b53f
MT
4423extern int dev_weight_rx_bias;
4424extern int dev_weight_tx_bias;
4425extern int dev_rx_weight;
4426extern int dev_tx_weight;
323ebb61 4427extern int gro_normal_batch;
9ff162a8 4428
f629d208 4429bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
4430struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4431 struct list_head **iter);
f629d208
JP
4432struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4433 struct list_head **iter);
8b5be856 4434
44a40855
VY
4435/* iterate through upper list, must be called under RCU read lock */
4436#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4437 for (iter = &(dev)->adj_list.upper, \
4438 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4439 updev; \
4440 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4441
1a3f060c
DA
4442int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4443 int (*fn)(struct net_device *upper_dev,
4444 void *data),
4445 void *data);
4446
4447bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4448 struct net_device *upper_dev);
4449
25cc72a3
IS
4450bool netdev_has_any_upper_dev(struct net_device *dev);
4451
f629d208
JP
4452void *netdev_lower_get_next_private(struct net_device *dev,
4453 struct list_head **iter);
4454void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4455 struct list_head **iter);
31088a11
VF
4456
4457#define netdev_for_each_lower_private(dev, priv, iter) \
4458 for (iter = (dev)->adj_list.lower.next, \
4459 priv = netdev_lower_get_next_private(dev, &(iter)); \
4460 priv; \
4461 priv = netdev_lower_get_next_private(dev, &(iter)))
4462
4463#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4464 for (iter = &(dev)->adj_list.lower, \
4465 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4466 priv; \
4467 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4468
4085ebe8
VY
4469void *netdev_lower_get_next(struct net_device *dev,
4470 struct list_head **iter);
7ce856aa 4471
4085ebe8 4472#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 4473 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
4474 ldev = netdev_lower_get_next(dev, &(iter)); \
4475 ldev; \
4476 ldev = netdev_lower_get_next(dev, &(iter)))
4477
7151affe 4478struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
7ce856aa 4479 struct list_head **iter);
1a3f060c
DA
4480int netdev_walk_all_lower_dev(struct net_device *dev,
4481 int (*fn)(struct net_device *lower_dev,
4482 void *data),
4483 void *data);
4484int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4485 int (*fn)(struct net_device *lower_dev,
4486 void *data),
4487 void *data);
4488
f629d208 4489void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 4490void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
4491struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4492struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
42ab19ee
DA
4493int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4494 struct netlink_ext_ack *extack);
f629d208 4495int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 4496 struct net_device *upper_dev,
42ab19ee
DA
4497 void *upper_priv, void *upper_info,
4498 struct netlink_ext_ack *extack);
f629d208
JP
4499void netdev_upper_dev_unlink(struct net_device *dev,
4500 struct net_device *upper_dev);
32b6d34f
TY
4501int netdev_adjacent_change_prepare(struct net_device *old_dev,
4502 struct net_device *new_dev,
4503 struct net_device *dev,
4504 struct netlink_ext_ack *extack);
4505void netdev_adjacent_change_commit(struct net_device *old_dev,
4506 struct net_device *new_dev,
4507 struct net_device *dev);
4508void netdev_adjacent_change_abort(struct net_device *old_dev,
4509 struct net_device *new_dev,
4510 struct net_device *dev);
5bb025fa 4511void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
4512void *netdev_lower_dev_get_private(struct net_device *dev,
4513 struct net_device *lower_dev);
04d48266
JP
4514void netdev_lower_state_changed(struct net_device *lower_dev,
4515 void *lower_state_info);
960fb622
ED
4516
4517/* RSS keys are 40 or 52 bytes long */
4518#define NETDEV_RSS_KEY_LEN 52
ba905f5e 4519extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
4520void netdev_rss_key_fill(void *buffer, size_t len);
4521
f629d208 4522int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 4523int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
4524int skb_csum_hwoffload_help(struct sk_buff *skb,
4525 const netdev_features_t features);
4526
f629d208
JP
4527struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4528 netdev_features_t features, bool tx_path);
4529struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4530 netdev_features_t features);
12b0004d 4531
61bd3857
MS
4532struct netdev_bonding_info {
4533 ifslave slave;
4534 ifbond master;
4535};
4536
4537struct netdev_notifier_bonding_info {
4538 struct netdev_notifier_info info; /* must be first */
4539 struct netdev_bonding_info bonding_info;
4540};
4541
4542void netdev_bonding_info_change(struct net_device *dev,
4543 struct netdev_bonding_info *bonding_info);
4544
6b08d6c1
MK
4545#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4546void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4547#else
4548static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4549 const void *data)
4550{
4551}
4552#endif
4553
12b0004d
CW
4554static inline
4555struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4556{
4557 return __skb_gso_segment(skb, features, true);
4558}
53d6471c 4559__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
4560
4561static inline bool can_checksum_protocol(netdev_features_t features,
4562 __be16 protocol)
4563{
c8cd0989
TH
4564 if (protocol == htons(ETH_P_FCOE))
4565 return !!(features & NETIF_F_FCOE_CRC);
4566
4567 /* Assume this is an IP checksum (not SCTP CRC) */
4568
4569 if (features & NETIF_F_HW_CSUM) {
4570 /* Can checksum everything */
4571 return true;
4572 }
4573
4574 switch (protocol) {
4575 case htons(ETH_P_IP):
4576 return !!(features & NETIF_F_IP_CSUM);
4577 case htons(ETH_P_IPV6):
4578 return !!(features & NETIF_F_IPV6_CSUM);
4579 default:
4580 return false;
4581 }
ec5f0615 4582}
12b0004d 4583
fb286bb2 4584#ifdef CONFIG_BUG
7fe50ac8 4585void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
fb286bb2 4586#else
7fe50ac8
CW
4587static inline void netdev_rx_csum_fault(struct net_device *dev,
4588 struct sk_buff *skb)
fb286bb2
HX
4589{
4590}
4591#endif
1da177e4 4592/* rx skb timestamps */
f629d208
JP
4593void net_enable_timestamp(void);
4594void net_disable_timestamp(void);
1da177e4 4595
20380731 4596#ifdef CONFIG_PROC_FS
f629d208 4597int __init dev_proc_init(void);
900ff8c6
CW
4598#else
4599#define dev_proc_init() 0
20380731
ACM
4600#endif
4601
4798248e 4602static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4603 struct sk_buff *skb, struct net_device *dev,
4604 bool more)
4798248e 4605{
6b16f9ee 4606 __this_cpu_write(softnet_data.xmit.more, more);
0b725a2c 4607 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4608}
4609
97cdcf37
FW
4610static inline bool netdev_xmit_more(void)
4611{
4612 return __this_cpu_read(softnet_data.xmit.more);
4613}
4614
10b3ad8c 4615static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4616 struct netdev_queue *txq, bool more)
4798248e
DM
4617{
4618 const struct net_device_ops *ops = dev->netdev_ops;
2183435c 4619 netdev_tx_t rc;
4798248e 4620
fa2dbdc2 4621 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4622 if (rc == NETDEV_TX_OK)
4623 txq_trans_update(txq);
4624
4625 return rc;
4798248e
DM
4626}
4627
b793dc5c 4628int netdev_class_create_file_ns(const struct class_attribute *class_attr,
42a2d923 4629 const void *ns);
b793dc5c 4630void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
42a2d923 4631 const void *ns);
58292cbe 4632
b793dc5c 4633static inline int netdev_class_create_file(const struct class_attribute *class_attr)
58292cbe
TH
4634{
4635 return netdev_class_create_file_ns(class_attr, NULL);
4636}
4637
b793dc5c 4638static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
58292cbe
TH
4639{
4640 netdev_class_remove_file_ns(class_attr, NULL);
4641}
b8a9787e 4642
737aec57 4643extern const struct kobj_ns_type_operations net_ns_type_operations;
04600794 4644
f629d208 4645const char *netdev_drivername(const struct net_device *dev);
6579e57b 4646
f629d208 4647void linkwatch_run_queue(void);
20380731 4648
da08143b
MK
4649static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4650 netdev_features_t f2)
4651{
c8cd0989
TH
4652 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4653 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4654 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4655 else
b6a0e72a 4656 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4657 }
da08143b 4658
c8cd0989 4659 return f1 & f2;
da08143b
MK
4660}
4661
c8f44aff
MM
4662static inline netdev_features_t netdev_get_wanted_features(
4663 struct net_device *dev)
5455c699
MM
4664{
4665 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4666}
c8f44aff
MM
4667netdev_features_t netdev_increment_features(netdev_features_t all,
4668 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4669
4670/* Allow TSO being used on stacked device :
4671 * Performing the GSO segmentation before last device
4672 * is a performance improvement.
4673 */
4674static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4675 netdev_features_t mask)
4676{
4677 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4678}
4679
6cb6a27c 4680int __netdev_update_features(struct net_device *dev);
5455c699 4681void netdev_update_features(struct net_device *dev);
afe12cc8 4682void netdev_change_features(struct net_device *dev);
7f353bf2 4683
fc4a7489
PM
4684void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4685 struct net_device *dev);
4686
e38f3025
TM
4687netdev_features_t passthru_features_check(struct sk_buff *skb,
4688 struct net_device *dev,
4689 netdev_features_t features);
c1e756bf 4690netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4691
4d29515f 4692static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4693{
7b748340 4694 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4695
4696 /* check flags correspondence */
4697 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4698 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4699 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4700 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4701 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4702 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4703 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4704 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4705 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4706 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4707 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4708 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4709 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4710 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4711 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4712 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0c19f846 4713 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
83aa025f 4714 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
3b335832 4715 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
0345e186 4716
d6b4991a 4717 return (features & feature) == feature;
576a30eb
HX
4718}
4719
4d29515f 4720static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4721{
278b2513 4722 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4723 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4724}
4725
8b86a61d 4726static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4727 netdev_features_t features)
7967168c 4728{
fc741216 4729 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4730 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4731 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4732}
4733
82cc1a7a
PWJ
4734static inline void netif_set_gso_max_size(struct net_device *dev,
4735 unsigned int size)
4736{
4737 dev->gso_max_size = size;
4738}
4739
7a7ffbab
WCC
4740static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4741 int pulled_hlen, u16 mac_offset,
4742 int mac_len)
4743{
4744 skb->protocol = protocol;
4745 skb->encapsulation = 1;
4746 skb_push(skb, pulled_hlen);
4747 skb_reset_transport_header(skb);
4748 skb->mac_header = mac_offset;
4749 skb->network_header = skb->mac_header + mac_len;
4750 skb->mac_len = mac_len;
4751}
4752
3c175784
SD
4753static inline bool netif_is_macsec(const struct net_device *dev)
4754{
4755 return dev->priv_flags & IFF_MACSEC;
4756}
4757
b618aaa9 4758static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4759{
4760 return dev->priv_flags & IFF_MACVLAN;
4761}
4762
b618aaa9 4763static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4764{
4765 return dev->priv_flags & IFF_MACVLAN_PORT;
4766}
4767
b618aaa9 4768static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4769{
4770 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4771}
4772
b618aaa9 4773static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4774{
4775 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4776}
4777
3bdc0eba
BG
4778static inline bool netif_supports_nofcs(struct net_device *dev)
4779{
4780 return dev->priv_flags & IFF_SUPP_NOFCS;
4781}
4782
d5256083
DB
4783static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4784{
4785 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4786}
4787
007979ea 4788static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4789{
007979ea 4790 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4791}
4792
fee6d4c7
DA
4793static inline bool netif_is_l3_slave(const struct net_device *dev)
4794{
4795 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4796}
4797
0894ae3f
JP
4798static inline bool netif_is_bridge_master(const struct net_device *dev)
4799{
4800 return dev->priv_flags & IFF_EBRIDGE;
4801}
4802
28f9ee22
VY
4803static inline bool netif_is_bridge_port(const struct net_device *dev)
4804{
4805 return dev->priv_flags & IFF_BRIDGE_PORT;
4806}
4807
35d4e172
JP
4808static inline bool netif_is_ovs_master(const struct net_device *dev)
4809{
4810 return dev->priv_flags & IFF_OPENVSWITCH;
4811}
4812
5be66141
JP
4813static inline bool netif_is_ovs_port(const struct net_device *dev)
4814{
4815 return dev->priv_flags & IFF_OVS_DATAPATH;
4816}
4817
b618aaa9 4818static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4819{
4820 return dev->priv_flags & IFF_TEAM;
4821}
4822
b618aaa9 4823static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4824{
4825 return dev->priv_flags & IFF_TEAM_PORT;
4826}
4827
b618aaa9 4828static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4829{
4830 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4831}
4832
b618aaa9 4833static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4834{
4835 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4836}
4837
d4ab4286
KJ
4838static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4839{
4840 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4841}
4842
30c8bd5a
SS
4843static inline bool netif_is_failover(const struct net_device *dev)
4844{
4845 return dev->priv_flags & IFF_FAILOVER;
4846}
4847
4848static inline bool netif_is_failover_slave(const struct net_device *dev)
4849{
4850 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4851}
4852
02875878
ED
4853/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4854static inline void netif_keep_dst(struct net_device *dev)
4855{
4856 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4857}
4858
18d3df3e
PA
4859/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4860static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4861{
4862 /* TODO: reserve and use an additional IFF bit, if we get more users */
4863 return dev->priv_flags & IFF_MACSEC;
4864}
4865
505d4f73 4866extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4867
571ba423
JP
4868/* Logging, debugging and troubleshooting/diagnostic helpers. */
4869
4870/* netdev_printk helpers, similar to dev_printk */
4871
4872static inline const char *netdev_name(const struct net_device *dev)
4873{
c6f854d5
VF
4874 if (!dev->name[0] || strchr(dev->name, '%'))
4875 return "(unnamed net_device)";
571ba423
JP
4876 return dev->name;
4877}
4878
8397ed36
DA
4879static inline bool netdev_unregistering(const struct net_device *dev)
4880{
4881 return dev->reg_state == NETREG_UNREGISTERING;
4882}
4883
ccc7f496
VF
4884static inline const char *netdev_reg_state(const struct net_device *dev)
4885{
4886 switch (dev->reg_state) {
4887 case NETREG_UNINITIALIZED: return " (uninitialized)";
4888 case NETREG_REGISTERED: return "";
4889 case NETREG_UNREGISTERING: return " (unregistering)";
4890 case NETREG_UNREGISTERED: return " (unregistered)";
4891 case NETREG_RELEASED: return " (released)";
4892 case NETREG_DUMMY: return " (dummy)";
4893 }
4894
4895 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4896 return " (unknown)";
4897}
4898
ce3fdb69 4899__printf(3, 4) __cold
6ea754eb
JP
4900void netdev_printk(const char *level, const struct net_device *dev,
4901 const char *format, ...);
ce3fdb69 4902__printf(2, 3) __cold
6ea754eb 4903void netdev_emerg(const struct net_device *dev, const char *format, ...);
ce3fdb69 4904__printf(2, 3) __cold
6ea754eb 4905void netdev_alert(const struct net_device *dev, const char *format, ...);
ce3fdb69 4906__printf(2, 3) __cold
6ea754eb 4907void netdev_crit(const struct net_device *dev, const char *format, ...);
ce3fdb69 4908__printf(2, 3) __cold
6ea754eb 4909void netdev_err(const struct net_device *dev, const char *format, ...);
ce3fdb69 4910__printf(2, 3) __cold
6ea754eb 4911void netdev_warn(const struct net_device *dev, const char *format, ...);
ce3fdb69 4912__printf(2, 3) __cold
6ea754eb 4913void netdev_notice(const struct net_device *dev, const char *format, ...);
ce3fdb69 4914__printf(2, 3) __cold
6ea754eb 4915void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4916
375ef2b1
GP
4917#define netdev_level_once(level, dev, fmt, ...) \
4918do { \
4919 static bool __print_once __read_mostly; \
4920 \
4921 if (!__print_once) { \
4922 __print_once = true; \
4923 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4924 } \
4925} while (0)
4926
4927#define netdev_emerg_once(dev, fmt, ...) \
4928 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4929#define netdev_alert_once(dev, fmt, ...) \
4930 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4931#define netdev_crit_once(dev, fmt, ...) \
4932 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4933#define netdev_err_once(dev, fmt, ...) \
4934 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4935#define netdev_warn_once(dev, fmt, ...) \
4936 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4937#define netdev_notice_once(dev, fmt, ...) \
4938 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4939#define netdev_info_once(dev, fmt, ...) \
4940 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4941
8909c9ad
VK
4942#define MODULE_ALIAS_NETDEV(device) \
4943 MODULE_ALIAS("netdev-" device)
4944
b558c96f 4945#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4946#define netdev_dbg(__dev, format, args...) \
4947do { \
ffa10cb4 4948 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4949} while (0)
b558c96f
JC
4950#elif defined(DEBUG)
4951#define netdev_dbg(__dev, format, args...) \
4952 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4953#else
4954#define netdev_dbg(__dev, format, args...) \
4955({ \
4956 if (0) \
4957 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4958})
4959#endif
4960
4961#if defined(VERBOSE_DEBUG)
4962#define netdev_vdbg netdev_dbg
4963#else
4964
4965#define netdev_vdbg(dev, format, args...) \
4966({ \
4967 if (0) \
4968 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4969 0; \
4970})
4971#endif
4972
4973/*
4974 * netdev_WARN() acts like dev_printk(), but with the key difference
4975 * of using a WARN/WARN_ON to get the message out, including the
4976 * file/line information and a backtrace.
4977 */
4978#define netdev_WARN(dev, format, args...) \
e1cfe3d0 4979 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
ccc7f496 4980 netdev_reg_state(dev), ##args)
571ba423 4981
72dd831e 4982#define netdev_WARN_ONCE(dev, format, args...) \
e1cfe3d0 4983 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
375ef2b1
GP
4984 netdev_reg_state(dev), ##args)
4985
b3d95c5c
JP
4986/* netif printk helpers, similar to netdev_printk */
4987
4988#define netif_printk(priv, type, level, dev, fmt, args...) \
4989do { \
4990 if (netif_msg_##type(priv)) \
4991 netdev_printk(level, (dev), fmt, ##args); \
4992} while (0)
4993
f45f4321
JP
4994#define netif_level(level, priv, type, dev, fmt, args...) \
4995do { \
4996 if (netif_msg_##type(priv)) \
4997 netdev_##level(dev, fmt, ##args); \
4998} while (0)
4999
b3d95c5c 5000#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 5001 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 5002#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 5003 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 5004#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 5005 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 5006#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 5007 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 5008#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 5009 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 5010#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 5011 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 5012#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 5013 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 5014
0053ea9c 5015#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
5016#define netif_dbg(priv, type, netdev, format, args...) \
5017do { \
5018 if (netif_msg_##type(priv)) \
b5fb0a03 5019 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 5020} while (0)
0053ea9c
JP
5021#elif defined(DEBUG)
5022#define netif_dbg(priv, type, dev, format, args...) \
5023 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
5024#else
5025#define netif_dbg(priv, type, dev, format, args...) \
5026({ \
5027 if (0) \
5028 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5029 0; \
5030})
5031#endif
5032
f617f276
EC
5033/* if @cond then downgrade to debug, else print at @level */
5034#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5035 do { \
5036 if (cond) \
5037 netif_dbg(priv, type, netdev, fmt, ##args); \
5038 else \
5039 netif_ ## level(priv, type, netdev, fmt, ##args); \
5040 } while (0)
5041
b3d95c5c 5042#if defined(VERBOSE_DEBUG)
bcfcc450 5043#define netif_vdbg netif_dbg
b3d95c5c
JP
5044#else
5045#define netif_vdbg(priv, type, dev, format, args...) \
5046({ \
5047 if (0) \
a4ed89cb 5048 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
5049 0; \
5050})
5051#endif
571ba423 5052
900ff8c6
CW
5053/*
5054 * The list of packet types we will receive (as opposed to discard)
5055 * and the routines to invoke.
5056 *
5057 * Why 16. Because with 16 the only overlap we get on a hash of the
5058 * low nibble of the protocol value is RARP/SNAP/X.25.
5059 *
900ff8c6 5060 * 0800 IP
900ff8c6
CW
5061 * 0001 802.3
5062 * 0002 AX.25
5063 * 0004 802.2
5064 * 8035 RARP
5065 * 0005 SNAP
5066 * 0805 X.25
5067 * 0806 ARP
5068 * 8137 IPX
5069 * 0009 Localtalk
5070 * 86DD IPv6
5071 */
5072#define PTYPE_HASH_SIZE (16)
5073#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5074
4de83b88
MB
5075extern struct net_device *blackhole_netdev;
5076
385a154c 5077#endif /* _LINUX_NETDEVICE_H */