]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - include/linux/netdevice.h
xfrm: Separate ESP handling from segmentation for GRO packets.
[mirror_ubuntu-hirsute-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
bea3348e 38#include <linux/workqueue.h>
114cf580 39#include <linux/dynamic_queue_limits.h>
1da177e4 40
b1b67dd4 41#include <linux/ethtool.h>
a050c33f 42#include <net/net_namespace.h>
7a6b6f51 43#ifdef CONFIG_DCB
2f90b865
AD
44#include <net/dcbnl.h>
45#endif
5bc1421e 46#include <net/netprio_cgroup.h>
a050c33f 47
a59e2ecb 48#include <linux/netdev_features.h>
77162022 49#include <linux/neighbour.h>
607ca46e 50#include <uapi/linux/netdevice.h>
61bd3857 51#include <uapi/linux/if_bonding.h>
e4c6734e 52#include <uapi/linux/pkt_cls.h>
59cc1f61 53#include <linux/hashtable.h>
a59e2ecb 54
115c1d6e 55struct netpoll_info;
313162d0 56struct device;
c1f19b51 57struct phy_device;
2f657a60 58struct dsa_port;
c6e970a0 59
704232c2
JB
60/* 802.11 specific */
61struct wireless_dev;
98a18b6f
AA
62/* 802.15.4 specific */
63struct wpan_dev;
03c57747 64struct mpls_dev;
7c46a640
AD
65/* UDP Tunnel offloads */
66struct udp_tunnel_info;
a7862b45 67struct bpf_prog;
814abfab 68struct xdp_buff;
1da177e4 69
f629d208
JP
70void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
d07d7507 72
9a1654ba
JP
73/* Backlog congestion levels */
74#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75#define NET_RX_DROP 1 /* packet dropped */
76
572a9d7b
PM
77/*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
91 * others are propagated to higher layers.
92 */
93
94/* qdisc ->enqueue() return codes. */
95#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
96#define NET_XMIT_DROP 0x01 /* skb dropped */
97#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 98#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 99
b9df3cb8
GR
100/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 103#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
104#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105
dc1f8bf6 106/* Driver transmit return codes */
9a1654ba 107#define NETDEV_TX_MASK 0xf0
572a9d7b 108
dc1f8bf6 109enum netdev_tx {
572a9d7b 110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
113};
114typedef enum netdev_tx netdev_tx_t;
115
9a1654ba
JP
116/*
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 */
120static inline bool dev_xmit_complete(int rc)
121{
122 /*
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 */
128 if (likely(rc < NET_XMIT_MASK))
129 return true;
130
131 return false;
132}
133
1da177e4 134/*
5e82b4b2 135 * Compute the worst-case header length according to the protocols
1da177e4
LT
136 * used.
137 */
fe2918b0 138
c0eb4540
KS
139#if defined(CONFIG_HYPERV_NET)
140# define LL_MAX_HEADER 128
141#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
142# if defined(CONFIG_MAC80211_MESH)
143# define LL_MAX_HEADER 128
144# else
145# define LL_MAX_HEADER 96
146# endif
1da177e4 147#else
8388e3da 148# define LL_MAX_HEADER 32
1da177e4
LT
149#endif
150
d11ead75
BH
151#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
153#define MAX_HEADER LL_MAX_HEADER
154#else
155#define MAX_HEADER (LL_MAX_HEADER + 48)
156#endif
157
158/*
be1f3c2c
BH
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
1da177e4 161 */
fe2918b0 162
d94d9fee 163struct net_device_stats {
3cfde79c
BH
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
1da177e4 173 unsigned long collisions;
1da177e4 174 unsigned long rx_length_errors;
3cfde79c
BH
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
1da177e4
LT
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
1da177e4
LT
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187};
188
1da177e4
LT
189
190#include <linux/cache.h>
191#include <linux/skbuff.h>
192
adc9300e 193#ifdef CONFIG_RPS
c5905afb
IM
194#include <linux/static_key.h>
195extern struct static_key rps_needed;
13bfff25 196extern struct static_key rfs_needed;
adc9300e
ED
197#endif
198
1da177e4
LT
199struct neighbour;
200struct neigh_parms;
201struct sk_buff;
202
f001fde5
JP
203struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
206 unsigned char type;
ccffad25
JP
207#define NETDEV_HW_ADDR_T_LAN 1
208#define NETDEV_HW_ADDR_T_SAN 2
209#define NETDEV_HW_ADDR_T_SLAVE 3
210#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 211#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 212 bool global_use;
4cd729b0 213 int sync_cnt;
8f8f103d 214 int refcount;
4543fbef 215 int synced;
f001fde5
JP
216 struct rcu_head rcu_head;
217};
218
31278e71
JP
219struct netdev_hw_addr_list {
220 struct list_head list;
221 int count;
222};
223
22bedad3
JP
224#define netdev_hw_addr_list_count(l) ((l)->count)
225#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226#define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 228
22bedad3
JP
229#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231#define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 233
22bedad3
JP
234#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 236#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 238
d94d9fee 239struct hh_cache {
5b3dc2f3 240 unsigned int hh_len;
3644f0ce 241 seqlock_t hh_lock;
1da177e4
LT
242
243 /* cached hardware header; allow for machine alignment needs. */
244#define HH_DATA_MOD 16
245#define HH_DATA_OFF(__len) \
5ba0eac6 246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
247#define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250};
251
5e82b4b2 252/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260#define LL_RESERVED_SPACE(dev) \
f5184d26 261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 262#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 264
3b04ddde
SH
265struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
95c96174 268 const void *saddr, unsigned int len);
3b04ddde 269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
2793a23a 274 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
275};
276
1da177e4 277/* These flag bits are private to the generic network queueing
5e82b4b2 278 * layer; they may not be explicitly referenced by any other
1da177e4
LT
279 * code.
280 */
281
d94d9fee 282enum netdev_state_t {
1da177e4
LT
283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
1da177e4 285 __LINK_STATE_NOCARRIER,
b00055aa
SR
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
1da177e4
LT
288};
289
290
291/*
5e82b4b2 292 * This structure holds boot-time configured netdevice settings. They
fe2918b0 293 * are then used in the device probing.
1da177e4
LT
294 */
295struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298};
299#define NETDEV_BOOT_SETUP_MAX 8
300
f629d208 301int __init netdev_boot_setup(char *str);
1da177e4 302
bea3348e
SH
303/*
304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 310 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
404f7c9e 317 unsigned int gro_count;
bea3348e
SH
318 int (*poll)(struct napi_struct *, int);
319#ifdef CONFIG_NETPOLL
bea3348e 320 int poll_owner;
bea3348e 321#endif
5d38a079 322 struct net_device *dev;
d565b0a1 323 struct sk_buff *gro_list;
5d38a079 324 struct sk_buff *skb;
3b47d303 325 struct hrtimer timer;
404f7c9e 326 struct list_head dev_list;
af12fa6e
ET
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
bea3348e
SH
329};
330
d94d9fee 331enum {
bea3348e 332 NAPI_STATE_SCHED, /* Poll is scheduled */
39e6c820 333 NAPI_STATE_MISSED, /* reschedule a napi */
a0a46196 334 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 335 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
336 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
337 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
338 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
339};
340
341enum {
39e6c820
ED
342 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
343 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
344 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
345 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
346 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
347 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
348 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
349};
350
5b252f0c 351enum gro_result {
d1c76af9
HX
352 GRO_MERGED,
353 GRO_MERGED_FREE,
354 GRO_HELD,
355 GRO_NORMAL,
356 GRO_DROP,
25393d3f 357 GRO_CONSUMED,
d1c76af9 358};
5b252f0c 359typedef enum gro_result gro_result_t;
d1c76af9 360
8a4eb573
JP
361/*
362 * enum rx_handler_result - Possible return values for rx_handlers.
363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
364 * further.
365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366 * case skb->dev was changed by rx_handler.
367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
369 *
370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
371 * special processing of the skb, prior to delivery to protocol handlers.
372 *
373 * Currently, a net_device can only have a single rx_handler registered. Trying
374 * to register a second rx_handler will return -EBUSY.
375 *
376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377 * To unregister a rx_handler on a net_device, use
378 * netdev_rx_handler_unregister().
379 *
380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
381 * do with the skb.
382 *
5e82b4b2 383 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 385 * the skb to be delivered in some other way.
8a4eb573
JP
386 *
387 * If the rx_handler changed skb->dev, to divert the skb to another
388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389 * new device will be called if it exists.
390 *
5e82b4b2 391 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 393 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 394 *
5e82b4b2 395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
396 * delivered, it should return RX_HANDLER_PASS.
397 *
398 * A device without a registered rx_handler will behave as if rx_handler
399 * returned RX_HANDLER_PASS.
400 */
401
402enum rx_handler_result {
403 RX_HANDLER_CONSUMED,
404 RX_HANDLER_ANOTHER,
405 RX_HANDLER_EXACT,
406 RX_HANDLER_PASS,
407};
408typedef enum rx_handler_result rx_handler_result_t;
409typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 410
f629d208 411void __napi_schedule(struct napi_struct *n);
bc9ad166 412void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 413
4d29515f 414static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
415{
416 return test_bit(NAPI_STATE_DISABLE, &n->state);
417}
418
39e6c820 419bool napi_schedule_prep(struct napi_struct *n);
bea3348e
SH
420
421/**
422 * napi_schedule - schedule NAPI poll
5e82b4b2 423 * @n: NAPI context
bea3348e
SH
424 *
425 * Schedule NAPI poll routine to be called if it is not already
426 * running.
427 */
428static inline void napi_schedule(struct napi_struct *n)
429{
430 if (napi_schedule_prep(n))
431 __napi_schedule(n);
432}
433
bc9ad166
ED
434/**
435 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 436 * @n: NAPI context
bc9ad166
ED
437 *
438 * Variant of napi_schedule(), assuming hard irqs are masked.
439 */
440static inline void napi_schedule_irqoff(struct napi_struct *n)
441{
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
444}
445
bfe13f54 446/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 447static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
448{
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
4d29515f 451 return true;
bfe13f54 452 }
4d29515f 453 return false;
bfe13f54
RD
454}
455
364b6055 456bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
457/**
458 * napi_complete - NAPI processing complete
5e82b4b2 459 * @n: NAPI context
bea3348e
SH
460 *
461 * Mark NAPI processing as complete.
3b47d303 462 * Consider using napi_complete_done() instead.
364b6055 463 * Return false if device should avoid rearming interrupts.
bea3348e 464 */
364b6055 465static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
466{
467 return napi_complete_done(n, 0);
468}
bea3348e 469
af12fa6e
ET
470/**
471 * napi_hash_del - remove a NAPI from global table
5e82b4b2 472 * @napi: NAPI context
af12fa6e 473 *
5e82b4b2 474 * Warning: caller must observe RCU grace period
34cbe27e
ED
475 * before freeing memory containing @napi, if
476 * this function returns true.
93d05d4a 477 * Note: core networking stack automatically calls it
5e82b4b2 478 * from netif_napi_del().
93d05d4a 479 * Drivers might want to call this helper to combine all
5e82b4b2 480 * the needed RCU grace periods into a single one.
af12fa6e 481 */
34cbe27e 482bool napi_hash_del(struct napi_struct *napi);
af12fa6e 483
bea3348e
SH
484/**
485 * napi_disable - prevent NAPI from scheduling
5e82b4b2 486 * @n: NAPI context
bea3348e
SH
487 *
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
490 */
3b47d303 491void napi_disable(struct napi_struct *n);
bea3348e
SH
492
493/**
494 * napi_enable - enable NAPI scheduling
5e82b4b2 495 * @n: NAPI context
bea3348e
SH
496 *
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
499 */
500static inline void napi_enable(struct napi_struct *n)
501{
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 503 smp_mb__before_atomic();
bea3348e 504 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 505 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
506}
507
c264c3de
SH
508/**
509 * napi_synchronize - wait until NAPI is not running
5e82b4b2 510 * @n: NAPI context
c264c3de
SH
511 *
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
515 */
516static inline void napi_synchronize(const struct napi_struct *n)
517{
facc432f
AB
518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
520 msleep(1);
521 else
522 barrier();
c264c3de 523}
c264c3de 524
d94d9fee 525enum netdev_queue_state_t {
73466498
TH
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
c3f26a26 528 __QUEUE_STATE_FROZEN,
79d16385 529};
8e2f1a63
DB
530
531#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
534
535#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
537 QUEUE_STATE_FROZEN)
538#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
539 QUEUE_STATE_FROZEN)
540
73466498
TH
541/*
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
549 */
79d16385 550
bb949fbd 551struct netdev_queue {
6a321cb3 552/*
5e82b4b2 553 * read-mostly part
6a321cb3 554 */
bb949fbd 555 struct net_device *dev;
46e5da40 556 struct Qdisc __rcu *qdisc;
b0e1e646 557 struct Qdisc *qdisc_sleeping;
ccf5ff69 558#ifdef CONFIG_SYSFS
1d24eb48
TH
559 struct kobject kobj;
560#endif
f2cd2d3e
ED
561#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 int numa_node;
563#endif
c0ef079c
FW
564 unsigned long tx_maxrate;
565 /*
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
568 */
569 unsigned long trans_timeout;
6a321cb3 570/*
5e82b4b2 571 * write-mostly part
6a321cb3
ED
572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
9d21493b 575 /*
9b36627a 576 * Time (in jiffies) of last Tx
9d21493b
ED
577 */
578 unsigned long trans_start;
ccf5ff69 579
114cf580
TH
580 unsigned long state;
581
582#ifdef CONFIG_BQL
583 struct dql dql;
584#endif
e8a0464c 585} ____cacheline_aligned_in_smp;
bb949fbd 586
f2cd2d3e
ED
587static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
588{
589#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 return q->numa_node;
591#else
b236da69 592 return NUMA_NO_NODE;
f2cd2d3e
ED
593#endif
594}
595
596static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
597{
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 q->numa_node = node;
600#endif
601}
602
df334545 603#ifdef CONFIG_RPS
0a9627f2
TH
604/*
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
607 */
608struct rps_map {
609 unsigned int len;
610 struct rcu_head rcu;
611 u16 cpus[0];
612};
60b778ce 613#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 614
fec5e652 615/*
c445477d
BH
616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
fec5e652
TH
619 */
620struct rps_dev_flow {
621 u16 cpu;
c445477d 622 u16 filter;
fec5e652
TH
623 unsigned int last_qtail;
624};
c445477d 625#define RPS_NO_FILTER 0xffff
fec5e652
TH
626
627/*
628 * The rps_dev_flow_table structure contains a table of flow mappings.
629 */
630struct rps_dev_flow_table {
631 unsigned int mask;
632 struct rcu_head rcu;
fec5e652
TH
633 struct rps_dev_flow flows[0];
634};
635#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 636 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
637
638/*
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
567e4b79 643 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 646 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
647 */
648struct rps_sock_flow_table {
567e4b79 649 u32 mask;
93c1af6c
ED
650
651 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 652};
567e4b79 653#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
654
655#define RPS_NO_CPU 0xffff
656
567e4b79
ED
657extern u32 rps_cpu_mask;
658extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
659
fec5e652
TH
660static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
661 u32 hash)
662{
663 if (table && hash) {
567e4b79
ED
664 unsigned int index = hash & table->mask;
665 u32 val = hash & ~rps_cpu_mask;
fec5e652 666
5e82b4b2 667 /* We only give a hint, preemption can change CPU under us */
567e4b79 668 val |= raw_smp_processor_id();
fec5e652 669
567e4b79
ED
670 if (table->ents[index] != val)
671 table->ents[index] = val;
fec5e652
TH
672 }
673}
674
c445477d 675#ifdef CONFIG_RFS_ACCEL
f629d208
JP
676bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
677 u16 filter_id);
c445477d 678#endif
a953be53 679#endif /* CONFIG_RPS */
c445477d 680
0a9627f2
TH
681/* This structure contains an instance of an RX queue. */
682struct netdev_rx_queue {
a953be53 683#ifdef CONFIG_RPS
6e3f7faf
ED
684 struct rps_map __rcu *rps_map;
685 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 686#endif
6e3f7faf 687 struct kobject kobj;
fe822240 688 struct net_device *dev;
0a9627f2 689} ____cacheline_aligned_in_smp;
a953be53
MD
690
691/*
692 * RX queue sysfs structures and functions.
693 */
694struct rx_queue_attribute {
695 struct attribute attr;
718ad681 696 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
a953be53 697 ssize_t (*store)(struct netdev_rx_queue *queue,
718ad681 698 const char *buf, size_t len);
a953be53 699};
d314774c 700
bf264145
TH
701#ifdef CONFIG_XPS
702/*
703 * This structure holds an XPS map which can be of variable length. The
704 * map is an array of queues.
705 */
706struct xps_map {
707 unsigned int len;
708 unsigned int alloc_len;
709 struct rcu_head rcu;
710 u16 queues[0];
711};
60b778ce 712#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
713#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
714 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
715
716/*
717 * This structure holds all XPS maps for device. Maps are indexed by CPU.
718 */
719struct xps_dev_maps {
720 struct rcu_head rcu;
a4177869 721 struct xps_map __rcu *cpu_map[0];
bf264145 722};
184c449f
AD
723#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
724 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
725#endif /* CONFIG_XPS */
726
4f57c087
JF
727#define TC_MAX_QUEUE 16
728#define TC_BITMASK 15
729/* HW offloaded queuing disciplines txq count and offset maps */
730struct netdev_tc_txq {
731 u16 count;
732 u16 offset;
733};
734
68bad94e
NP
735#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
736/*
737 * This structure is to hold information about the device
738 * configured to run FCoE protocol stack.
739 */
740struct netdev_fcoe_hbainfo {
741 char manufacturer[64];
742 char serial_number[64];
743 char hardware_version[64];
744 char driver_version[64];
745 char optionrom_version[64];
746 char firmware_version[64];
747 char model[256];
748 char model_description[256];
749};
750#endif
751
02637fce 752#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 753
02637fce
JP
754/* This structure holds a unique identifier to identify some
755 * physical item (port for example) used by a netdevice.
66b52b0d 756 */
02637fce
JP
757struct netdev_phys_item_id {
758 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
759 unsigned char id_len;
760};
761
d754f98b
SF
762static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
763 struct netdev_phys_item_id *b)
764{
765 return a->id_len == b->id_len &&
766 memcmp(a->id, b->id, a->id_len) == 0;
767}
768
99932d4f
DB
769typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 struct sk_buff *skb);
771
2572ac53 772enum tc_setup_type {
575ed7d3 773 TC_SETUP_QDISC_MQPRIO,
a1b7c5fd 774 TC_SETUP_CLSU32,
5b33f488 775 TC_SETUP_CLSFLOWER,
ade9b658 776 TC_SETUP_CLSMATCHALL,
332ae8e2 777 TC_SETUP_CLSBPF,
8c4083b3 778 TC_SETUP_BLOCK,
8521db4c 779 TC_SETUP_QDISC_CBS,
602f3baf 780 TC_SETUP_QDISC_RED,
16e5cc64
JF
781};
782
f4e63525
JK
783/* These structures hold the attributes of bpf state that are being passed
784 * to the netdevice through the bpf op.
a7862b45 785 */
f4e63525 786enum bpf_netdev_command {
a7862b45
BB
787 /* Set or clear a bpf program used in the earliest stages of packet
788 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
789 * is responsible for calling bpf_prog_put on any old progs that are
790 * stored. In case of error, the callee need not release the new prog
791 * reference, but on success it takes ownership and must bpf_prog_put
792 * when it is no longer used.
793 */
794 XDP_SETUP_PROG,
ee5d032f 795 XDP_SETUP_PROG_HW,
a7862b45 796 /* Check if a bpf program is set on the device. The callee should
ce158e58
JK
797 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
798 * is equivalent to XDP_ATTACHED_DRV.
a7862b45
BB
799 */
800 XDP_QUERY_PROG,
ab3f0063
JK
801 /* BPF program for offload callbacks, invoked at program load time. */
802 BPF_OFFLOAD_VERIFIER_PREP,
803 BPF_OFFLOAD_TRANSLATE,
804 BPF_OFFLOAD_DESTROY,
a7862b45
BB
805};
806
ab3f0063 807struct bpf_ext_analyzer_ops;
ddf9f970
JK
808struct netlink_ext_ack;
809
f4e63525
JK
810struct netdev_bpf {
811 enum bpf_netdev_command command;
a7862b45
BB
812 union {
813 /* XDP_SETUP_PROG */
ddf9f970 814 struct {
32d60277 815 u32 flags;
ddf9f970
JK
816 struct bpf_prog *prog;
817 struct netlink_ext_ack *extack;
818 };
a7862b45 819 /* XDP_QUERY_PROG */
58038695 820 struct {
ce158e58 821 u8 prog_attached;
58038695 822 u32 prog_id;
92f0292b
JK
823 /* flags with which program was installed */
824 u32 prog_flags;
58038695 825 };
ab3f0063
JK
826 /* BPF_OFFLOAD_VERIFIER_PREP */
827 struct {
828 struct bpf_prog *prog;
829 const struct bpf_ext_analyzer_ops *ops; /* callee set */
830 } verifier;
831 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
832 struct {
833 struct bpf_prog *prog;
834 } offload;
a7862b45
BB
835 };
836};
16e5cc64 837
d77e38e6
SK
838#ifdef CONFIG_XFRM_OFFLOAD
839struct xfrmdev_ops {
840 int (*xdo_dev_state_add) (struct xfrm_state *x);
841 void (*xdo_dev_state_delete) (struct xfrm_state *x);
842 void (*xdo_dev_state_free) (struct xfrm_state *x);
843 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
844 struct xfrm_state *x);
845};
846#endif
847
6c557001
FW
848struct dev_ifalias {
849 struct rcu_head rcuhead;
850 char ifalias[];
851};
852
d314774c
SH
853/*
854 * This structure defines the management hooks for network devices.
00829823
SH
855 * The following hooks can be defined; unless noted otherwise, they are
856 * optional and can be filled with a null pointer.
d314774c
SH
857 *
858 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
859 * This function is called once when a network device is registered.
860 * The network device can use this for any late stage initialization
861 * or semantic validation. It can fail with an error code which will
862 * be propagated back to register_netdev.
d314774c
SH
863 *
864 * void (*ndo_uninit)(struct net_device *dev);
865 * This function is called when device is unregistered or when registration
866 * fails. It is not called if init fails.
867 *
868 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 869 * This function is called when a network device transitions to the up
d314774c
SH
870 * state.
871 *
872 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 873 * This function is called when a network device transitions to the down
d314774c
SH
874 * state.
875 *
dc1f8bf6
SH
876 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
877 * struct net_device *dev);
00829823 878 * Called when a packet needs to be transmitted.
e79d8429
RR
879 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
880 * the queue before that can happen; it's for obsolete devices and weird
881 * corner cases, but the stack really does a non-trivial amount
882 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 883 * Required; cannot be NULL.
00829823 884 *
1a2a1444
DM
885 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
886 * struct net_device *dev
887 * netdev_features_t features);
888 * Called by core transmit path to determine if device is capable of
889 * performing offload operations on a given packet. This is to give
890 * the device an opportunity to implement any restrictions that cannot
891 * be otherwise expressed by feature flags. The check is called with
892 * the set of features that the stack has calculated and it returns
893 * those the driver believes to be appropriate.
cdba756f 894 *
f663dd9a 895 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 896 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 897 * Called to decide which queue to use when device supports multiple
00829823
SH
898 * transmit queues.
899 *
d314774c
SH
900 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
901 * This function is called to allow device receiver to make
5e82b4b2 902 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
903 *
904 * void (*ndo_set_rx_mode)(struct net_device *dev);
905 * This function is called device changes address list filtering.
01789349 906 * If driver handles unicast address filtering, it should set
5e82b4b2 907 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
908 *
909 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
910 * This function is called when the Media Access Control address
37b607c5 911 * needs to be changed. If this interface is not defined, the
5e82b4b2 912 * MAC address can not be changed.
d314774c
SH
913 *
914 * int (*ndo_validate_addr)(struct net_device *dev);
915 * Test if Media Access Control address is valid for the device.
916 *
917 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
918 * Called when a user requests an ioctl which can't be handled by
919 * the generic interface code. If not defined ioctls return
d314774c
SH
920 * not supported error code.
921 *
922 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
923 * Used to set network devices bus interface parameters. This interface
5e82b4b2 924 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
925 * interface (PCI) for low level management.
926 *
927 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
928 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 929 * of a device.
d314774c 930 *
00829823 931 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 932 * Callback used when the transmitter has not made any progress
d314774c
SH
933 * for dev->watchdog ticks.
934 *
bc1f4470 935 * void (*ndo_get_stats64)(struct net_device *dev,
936 * struct rtnl_link_stats64 *storage);
d308e38f 937 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 938 * Called when a user wants to get the network device usage
be1f3c2c 939 * statistics. Drivers must do one of the following:
3cfde79c
BH
940 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
941 * rtnl_link_stats64 structure passed by the caller.
82695d9b 942 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
943 * (which should normally be dev->stats) and return a pointer to
944 * it. The structure may be changed asynchronously only if each
945 * field is written atomically.
946 * 3. Update dev->stats asynchronously and atomically, and define
947 * neither operation.
d314774c 948 *
3df5b3c6 949 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
950 * Return true if this device supports offload stats of this attr_id.
951 *
952 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
953 * void *attr_data)
954 * Get statistics for offload operations by attr_id. Write it into the
955 * attr_data pointer.
956 *
5d632cb7 957 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 958 * If device supports VLAN filtering this function is called when a
80d5c368 959 * VLAN id is registered.
d314774c 960 *
5d632cb7 961 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 962 * If device supports VLAN filtering this function is called when a
80d5c368 963 * VLAN id is unregistered.
d314774c
SH
964 *
965 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
966 *
967 * SR-IOV management functions.
968 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
969 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
970 * u8 qos, __be16 proto);
ed616689
SC
971 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
972 * int max_tx_rate);
5f8444a3 973 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 974 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
975 * int (*ndo_get_vf_config)(struct net_device *dev,
976 * int vf, struct ifla_vf_info *ivf);
1d8faf48 977 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
978 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
979 * struct nlattr *port[]);
01a3d796
VZ
980 *
981 * Enable or disable the VF ability to query its RSS Redirection Table and
982 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 983 * with PF and querying it may introduce a theoretical security risk.
01a3d796 984 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 985 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
2572ac53 986 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
de4784ca 987 * void *type_data);
6a4bc2b4
FF
988 * Called to setup any 'tc' scheduler, classifier or action on @dev.
989 * This is always called from the stack with the rtnl lock held and netif
990 * tx queues stopped. This allows the netdevice to perform queue
991 * management safely.
c445477d 992 *
e9bce845
YZ
993 * Fiber Channel over Ethernet (FCoE) offload functions.
994 * int (*ndo_fcoe_enable)(struct net_device *dev);
995 * Called when the FCoE protocol stack wants to start using LLD for FCoE
996 * so the underlying device can perform whatever needed configuration or
997 * initialization to support acceleration of FCoE traffic.
998 *
999 * int (*ndo_fcoe_disable)(struct net_device *dev);
1000 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1001 * so the underlying device can perform whatever needed clean-ups to
1002 * stop supporting acceleration of FCoE traffic.
1003 *
1004 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1005 * struct scatterlist *sgl, unsigned int sgc);
1006 * Called when the FCoE Initiator wants to initialize an I/O that
1007 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1008 * perform necessary setup and returns 1 to indicate the device is set up
1009 * successfully to perform DDP on this I/O, otherwise this returns 0.
1010 *
1011 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1012 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1013 * indicated by the FC exchange id 'xid', so the underlying device can
1014 * clean up and reuse resources for later DDP requests.
1015 *
1016 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1017 * struct scatterlist *sgl, unsigned int sgc);
1018 * Called when the FCoE Target wants to initialize an I/O that
1019 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1020 * perform necessary setup and returns 1 to indicate the device is set up
1021 * successfully to perform DDP on this I/O, otherwise this returns 0.
1022 *
68bad94e
NP
1023 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1024 * struct netdev_fcoe_hbainfo *hbainfo);
1025 * Called when the FCoE Protocol stack wants information on the underlying
1026 * device. This information is utilized by the FCoE protocol stack to
1027 * register attributes with Fiber Channel management service as per the
1028 * FC-GS Fabric Device Management Information(FDMI) specification.
1029 *
e9bce845
YZ
1030 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1031 * Called when the underlying device wants to override default World Wide
1032 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1033 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1034 * protocol stack to use.
1035 *
c445477d
BH
1036 * RFS acceleration.
1037 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1038 * u16 rxq_index, u32 flow_id);
1039 * Set hardware filter for RFS. rxq_index is the target queue index;
1040 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1041 * Return the filter ID on success, or a negative error code.
fbaec0ea 1042 *
8b98a70c 1043 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1044 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1045 * Called to make another netdev an underling.
1046 *
1047 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1048 * Called to release previously enslaved netdev.
5455c699
MM
1049 *
1050 * Feature/offload setting functions.
1a2a1444
DM
1051 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1052 * netdev_features_t features);
1053 * Adjusts the requested feature flags according to device-specific
1054 * constraints, and returns the resulting flags. Must not modify
1055 * the device state.
1056 *
c8f44aff 1057 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1058 * Called to update device configuration to new features. Passed
1059 * feature set might be less than what was returned by ndo_fix_features()).
1060 * Must return >0 or -errno if it changed dev->features itself.
1061 *
edc7d573 1062 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1063 * struct net_device *dev,
f6f6424b 1064 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1065 * Adds an FDB entry to dev for addr.
1690be63
VY
1066 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1067 * struct net_device *dev,
f6f6424b 1068 * const unsigned char *addr, u16 vid)
77162022
JF
1069 * Deletes the FDB entry from dev coresponding to addr.
1070 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1071 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1072 * int *idx)
77162022
JF
1073 * Used to add FDB entries to dump requests. Implementers should add
1074 * entries to skb and update idx with the number of entries.
e5a55a89 1075 *
ad41faa8
ND
1076 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1077 * u16 flags)
e5a55a89 1078 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1079 * struct net_device *dev, u32 filter_mask,
1080 * int nlflags)
ad41faa8
ND
1081 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1082 * u16 flags);
4bf84c35
JP
1083 *
1084 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1085 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1086 * which do not represent real hardware may define this to allow their
1087 * userspace components to manage their virtual carrier state. Devices
1088 * that determine carrier state from physical hardware properties (eg
1089 * network cables) or protocol-dependent mechanisms (eg
1090 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1091 *
1092 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1093 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1094 * Called to get ID of physical port of this device. If driver does
1095 * not implement this, it is assumed that the hw is not able to have
1096 * multiple net devices on single physical port.
53cf5275 1097 *
7c46a640
AD
1098 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1099 * struct udp_tunnel_info *ti);
1100 * Called by UDP tunnel to notify a driver about the UDP port and socket
1101 * address family that a UDP tunnel is listnening to. It is called only
1102 * when a new port starts listening. The operation is protected by the
1103 * RTNL.
1104 *
1105 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1106 * struct udp_tunnel_info *ti);
1107 * Called by UDP tunnel to notify the driver about a UDP port and socket
1108 * address family that the UDP tunnel is not listening to anymore. The
1109 * operation is protected by the RTNL.
1110 *
a6cc0cfa
JF
1111 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1112 * struct net_device *dev)
1113 * Called by upper layer devices to accelerate switching or other
1114 * station functionality into hardware. 'pdev is the lowerdev
1115 * to use for the offload and 'dev' is the net device that will
1116 * back the offload. Returns a pointer to the private structure
1117 * the upper layer will maintain.
1118 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1119 * Called by upper layer device to delete the station created
1120 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1121 * the station and priv is the structure returned by the add
1122 * operation.
822b3b2e
JF
1123 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1124 * int queue_index, u32 maxrate);
1125 * Called when a user wants to set a max-rate limitation of specific
1126 * TX queue.
a54acb3a
ND
1127 * int (*ndo_get_iflink)(const struct net_device *dev);
1128 * Called to get the iflink value of this device.
d746d707 1129 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1130 * bool proto_down);
d746d707
AK
1131 * This function is used to pass protocol port error state information
1132 * to the switch driver. The switch driver can react to the proto_down
1133 * by doing a phys down on the associated switch port.
fc4099f1
PS
1134 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1135 * This function is used to get egress tunnel information for given skb.
1136 * This is useful for retrieving outer tunnel header parameters while
1137 * sampling packet.
871b642a
PA
1138 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1139 * This function is used to specify the headroom that the skb must
1140 * consider when allocation skb during packet reception. Setting
1141 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1142 * forward. Setting a negative value resets the rx headroom to the
871b642a 1143 * default value.
f4e63525 1144 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
a7862b45 1145 * This function is used to set or query state related to XDP on the
f4e63525
JK
1146 * netdevice and manage BPF offload. See definition of
1147 * enum bpf_netdev_command for details.
814abfab
JF
1148 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1149 * This function is used to submit a XDP packet for transmit on a
1150 * netdevice.
11393cc9 1151 * void (*ndo_xdp_flush)(struct net_device *dev);
f16ded59
RR
1152 * This function is used to inform the driver to flush a particular
1153 * xdp tx queue. Must be called on same CPU as xdp_xmit.
d314774c
SH
1154 */
1155struct net_device_ops {
1156 int (*ndo_init)(struct net_device *dev);
1157 void (*ndo_uninit)(struct net_device *dev);
1158 int (*ndo_open)(struct net_device *dev);
1159 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1160 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1161 struct net_device *dev);
1162 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1163 struct net_device *dev,
1164 netdev_features_t features);
00829823 1165 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1166 struct sk_buff *skb,
99932d4f
DB
1167 void *accel_priv,
1168 select_queue_fallback_t fallback);
d314774c
SH
1169 void (*ndo_change_rx_flags)(struct net_device *dev,
1170 int flags);
d314774c 1171 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1172 int (*ndo_set_mac_address)(struct net_device *dev,
1173 void *addr);
d314774c 1174 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1175 int (*ndo_do_ioctl)(struct net_device *dev,
1176 struct ifreq *ifr, int cmd);
d314774c
SH
1177 int (*ndo_set_config)(struct net_device *dev,
1178 struct ifmap *map);
00829823
SH
1179 int (*ndo_change_mtu)(struct net_device *dev,
1180 int new_mtu);
1181 int (*ndo_neigh_setup)(struct net_device *dev,
1182 struct neigh_parms *);
d314774c
SH
1183 void (*ndo_tx_timeout) (struct net_device *dev);
1184
bc1f4470 1185 void (*ndo_get_stats64)(struct net_device *dev,
1186 struct rtnl_link_stats64 *storage);
3df5b3c6 1187 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1188 int (*ndo_get_offload_stats)(int attr_id,
1189 const struct net_device *dev,
1190 void *attr_data);
d314774c
SH
1191 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1192
8e586137 1193 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1194 __be16 proto, u16 vid);
8e586137 1195 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1196 __be16 proto, u16 vid);
d314774c 1197#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1198 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1199 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1200 struct netpoll_info *info);
0e34e931 1201 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1202#endif
95c26df8
WM
1203 int (*ndo_set_vf_mac)(struct net_device *dev,
1204 int queue, u8 *mac);
1205 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1206 int queue, u16 vlan,
1207 u8 qos, __be16 proto);
ed616689
SC
1208 int (*ndo_set_vf_rate)(struct net_device *dev,
1209 int vf, int min_tx_rate,
1210 int max_tx_rate);
5f8444a3
GR
1211 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1212 int vf, bool setting);
dd461d6a
HS
1213 int (*ndo_set_vf_trust)(struct net_device *dev,
1214 int vf, bool setting);
95c26df8
WM
1215 int (*ndo_get_vf_config)(struct net_device *dev,
1216 int vf,
1217 struct ifla_vf_info *ivf);
1d8faf48
RE
1218 int (*ndo_set_vf_link_state)(struct net_device *dev,
1219 int vf, int link_state);
3b766cd8
EBE
1220 int (*ndo_get_vf_stats)(struct net_device *dev,
1221 int vf,
1222 struct ifla_vf_stats
1223 *vf_stats);
57b61080
SF
1224 int (*ndo_set_vf_port)(struct net_device *dev,
1225 int vf,
1226 struct nlattr *port[]);
1227 int (*ndo_get_vf_port)(struct net_device *dev,
1228 int vf, struct sk_buff *skb);
cc8e27cc
EC
1229 int (*ndo_set_vf_guid)(struct net_device *dev,
1230 int vf, u64 guid,
1231 int guid_type);
01a3d796
VZ
1232 int (*ndo_set_vf_rss_query_en)(
1233 struct net_device *dev,
1234 int vf, bool setting);
16e5cc64 1235 int (*ndo_setup_tc)(struct net_device *dev,
2572ac53 1236 enum tc_setup_type type,
de4784ca 1237 void *type_data);
d11ead75 1238#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1239 int (*ndo_fcoe_enable)(struct net_device *dev);
1240 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1241 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1242 u16 xid,
1243 struct scatterlist *sgl,
1244 unsigned int sgc);
1245 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1246 u16 xid);
6247e086
YZ
1247 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1248 u16 xid,
1249 struct scatterlist *sgl,
1250 unsigned int sgc);
68bad94e
NP
1251 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1252 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1253#endif
1254
d11ead75 1255#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1256#define NETDEV_FCOE_WWNN 0
1257#define NETDEV_FCOE_WWPN 1
1258 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1259 u64 *wwn, int type);
4d288d57 1260#endif
3c9c36bc 1261
c445477d
BH
1262#ifdef CONFIG_RFS_ACCEL
1263 int (*ndo_rx_flow_steer)(struct net_device *dev,
1264 const struct sk_buff *skb,
1265 u16 rxq_index,
1266 u32 flow_id);
1267#endif
fbaec0ea 1268 int (*ndo_add_slave)(struct net_device *dev,
33eaf2a6
DA
1269 struct net_device *slave_dev,
1270 struct netlink_ext_ack *extack);
fbaec0ea
JP
1271 int (*ndo_del_slave)(struct net_device *dev,
1272 struct net_device *slave_dev);
c8f44aff
MM
1273 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1274 netdev_features_t features);
5455c699 1275 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1276 netdev_features_t features);
503eebc2
JP
1277 int (*ndo_neigh_construct)(struct net_device *dev,
1278 struct neighbour *n);
1279 void (*ndo_neigh_destroy)(struct net_device *dev,
1280 struct neighbour *n);
77162022
JF
1281
1282 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1283 struct nlattr *tb[],
77162022 1284 struct net_device *dev,
6b6e2725 1285 const unsigned char *addr,
f6f6424b 1286 u16 vid,
77162022
JF
1287 u16 flags);
1288 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1289 struct nlattr *tb[],
77162022 1290 struct net_device *dev,
f6f6424b
JP
1291 const unsigned char *addr,
1292 u16 vid);
77162022
JF
1293 int (*ndo_fdb_dump)(struct sk_buff *skb,
1294 struct netlink_callback *cb,
1295 struct net_device *dev,
5d5eacb3 1296 struct net_device *filter_dev,
d297653d 1297 int *idx);
e5a55a89
JF
1298
1299 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1300 struct nlmsghdr *nlh,
1301 u16 flags);
e5a55a89
JF
1302 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1303 u32 pid, u32 seq,
6cbdceeb 1304 struct net_device *dev,
46c264da
ND
1305 u32 filter_mask,
1306 int nlflags);
407af329 1307 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1308 struct nlmsghdr *nlh,
1309 u16 flags);
4bf84c35
JP
1310 int (*ndo_change_carrier)(struct net_device *dev,
1311 bool new_carrier);
66b52b0d 1312 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1313 struct netdev_phys_item_id *ppid);
db24a904
DA
1314 int (*ndo_get_phys_port_name)(struct net_device *dev,
1315 char *name, size_t len);
7c46a640
AD
1316 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1317 struct udp_tunnel_info *ti);
1318 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1319 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1320 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1321 struct net_device *dev);
1322 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1323 void *priv);
1324
25175ba5 1325 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1326 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1327 int queue_index,
1328 u32 maxrate);
a54acb3a 1329 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1330 int (*ndo_change_proto_down)(struct net_device *dev,
1331 bool proto_down);
fc4099f1
PS
1332 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1333 struct sk_buff *skb);
871b642a
PA
1334 void (*ndo_set_rx_headroom)(struct net_device *dev,
1335 int needed_headroom);
f4e63525
JK
1336 int (*ndo_bpf)(struct net_device *dev,
1337 struct netdev_bpf *bpf);
814abfab
JF
1338 int (*ndo_xdp_xmit)(struct net_device *dev,
1339 struct xdp_buff *xdp);
11393cc9 1340 void (*ndo_xdp_flush)(struct net_device *dev);
d314774c
SH
1341};
1342
7aa98047
LR
1343/**
1344 * enum net_device_priv_flags - &struct net_device priv_flags
1345 *
1346 * These are the &struct net_device, they are only set internally
1347 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1348 * userspace; this means that the order of these flags can change
7aa98047
LR
1349 * during any kernel release.
1350 *
1351 * You should have a pretty good reason to be extending these flags.
1352 *
1353 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1354 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1355 * @IFF_BONDING: bonding master or slave
7aa98047 1356 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1357 * @IFF_WAN_HDLC: WAN HDLC device
1358 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1359 * release skb->dst
1360 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1361 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1362 * @IFF_MACVLAN_PORT: device used as macvlan port
1363 * @IFF_BRIDGE_PORT: device used as bridge port
1364 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1365 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1366 * @IFF_UNICAST_FLT: Supports unicast filtering
1367 * @IFF_TEAM_PORT: device used as team port
1368 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1369 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1370 * change when it's running
1371 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1372 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1373 * underlying stacked devices
1374 * @IFF_IPVLAN_MASTER: IPvlan master device
1375 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1376 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1377 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1378 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1379 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1380 * @IFF_TEAM: device is a team device
d4ab4286 1381 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1382 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1383 * entity (i.e. the master device for bridged veth)
3c175784 1384 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1385 */
1386enum netdev_priv_flags {
1387 IFF_802_1Q_VLAN = 1<<0,
1388 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1389 IFF_BONDING = 1<<2,
1390 IFF_ISATAP = 1<<3,
1391 IFF_WAN_HDLC = 1<<4,
1392 IFF_XMIT_DST_RELEASE = 1<<5,
1393 IFF_DONT_BRIDGE = 1<<6,
1394 IFF_DISABLE_NETPOLL = 1<<7,
1395 IFF_MACVLAN_PORT = 1<<8,
1396 IFF_BRIDGE_PORT = 1<<9,
1397 IFF_OVS_DATAPATH = 1<<10,
1398 IFF_TX_SKB_SHARING = 1<<11,
1399 IFF_UNICAST_FLT = 1<<12,
1400 IFF_TEAM_PORT = 1<<13,
1401 IFF_SUPP_NOFCS = 1<<14,
1402 IFF_LIVE_ADDR_CHANGE = 1<<15,
1403 IFF_MACVLAN = 1<<16,
1404 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1405 IFF_IPVLAN_MASTER = 1<<18,
1406 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1407 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1408 IFF_NO_QUEUE = 1<<21,
1409 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1410 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1411 IFF_TEAM = 1<<24,
d4ab4286 1412 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1413 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1414 IFF_MACSEC = 1<<27,
7aa98047
LR
1415};
1416
1417#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1418#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1419#define IFF_BONDING IFF_BONDING
7aa98047 1420#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1421#define IFF_WAN_HDLC IFF_WAN_HDLC
1422#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1423#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1424#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1425#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1426#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1427#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1428#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1429#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1430#define IFF_TEAM_PORT IFF_TEAM_PORT
1431#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1432#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1433#define IFF_MACVLAN IFF_MACVLAN
02875878 1434#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1435#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1436#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1437#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1438#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1439#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1440#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1441#define IFF_TEAM IFF_TEAM
d4ab4286 1442#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1443#define IFF_MACSEC IFF_MACSEC
7aa98047 1444
536721b1
KK
1445/**
1446 * struct net_device - The DEVICE structure.
d651983d
MCC
1447 *
1448 * Actually, this whole structure is a big mistake. It mixes I/O
1449 * data with strictly "high-level" data, and it has to know about
1450 * almost every data structure used in the INET module.
536721b1
KK
1451 *
1452 * @name: This is the first field of the "visible" part of this structure
1453 * (i.e. as seen by users in the "Space.c" file). It is the name
d651983d 1454 * of the interface.
536721b1
KK
1455 *
1456 * @name_hlist: Device name hash chain, please keep it close to name[]
1457 * @ifalias: SNMP alias
1458 * @mem_end: Shared memory end
1459 * @mem_start: Shared memory start
1460 * @base_addr: Device I/O address
1461 * @irq: Device IRQ number
1462 *
14ffbbb8
TG
1463 * @carrier_changes: Stats to monitor carrier on<->off transitions
1464 *
536721b1
KK
1465 * @state: Generic network queuing layer state, see netdev_state_t
1466 * @dev_list: The global list of network devices
5e82b4b2
BH
1467 * @napi_list: List entry used for polling NAPI devices
1468 * @unreg_list: List entry when we are unregistering the
1469 * device; see the function unregister_netdev
1470 * @close_list: List entry used when we are closing the device
62d885fe
BP
1471 * @ptype_all: Device-specific packet handlers for all protocols
1472 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1473 *
1474 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1475 * @features: Currently active device features
1476 * @hw_features: User-changeable features
1477 *
1478 * @wanted_features: User-requested features
1479 * @vlan_features: Mask of features inheritable by VLAN devices
1480 *
1481 * @hw_enc_features: Mask of features inherited by encapsulating devices
1482 * This field indicates what encapsulation
1483 * offloads the hardware is capable of doing,
1484 * and drivers will need to set them appropriately.
1485 *
1486 * @mpls_features: Mask of features inheritable by MPLS
1487 *
1488 * @ifindex: interface index
5e82b4b2 1489 * @group: The group the device belongs to
536721b1
KK
1490 *
1491 * @stats: Statistics struct, which was left as a legacy, use
1492 * rtnl_link_stats64 instead
1493 *
1494 * @rx_dropped: Dropped packets by core network,
1495 * do not use this in drivers
1496 * @tx_dropped: Dropped packets by core network,
1497 * do not use this in drivers
6e7333d3
JW
1498 * @rx_nohandler: nohandler dropped packets by core network on
1499 * inactive devices, do not use this in drivers
536721b1 1500 *
536721b1
KK
1501 * @wireless_handlers: List of functions to handle Wireless Extensions,
1502 * instead of ioctl,
1503 * see <net/iw_handler.h> for details.
1504 * @wireless_data: Instance data managed by the core of wireless extensions
1505 *
1506 * @netdev_ops: Includes several pointers to callbacks,
1507 * if one wants to override the ndo_*() functions
1508 * @ethtool_ops: Management operations
f997c55c
AA
1509 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1510 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1511 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1512 * of Layer 2 headers.
1513 *
1514 * @flags: Interface flags (a la BSD)
1515 * @priv_flags: Like 'flags' but invisible to userspace,
1516 * see if.h for the definitions
1517 * @gflags: Global flags ( kept as legacy )
1518 * @padded: How much padding added by alloc_netdev()
1519 * @operstate: RFC2863 operstate
1520 * @link_mode: Mapping policy to operstate
1521 * @if_port: Selectable AUI, TP, ...
1522 * @dma: DMA channel
1523 * @mtu: Interface MTU value
61e84623
JW
1524 * @min_mtu: Interface Minimum MTU value
1525 * @max_mtu: Interface Maximum MTU value
536721b1 1526 * @type: Interface hardware type
2793a23a 1527 * @hard_header_len: Maximum hardware header length.
217e6fa2 1528 * @min_header_len: Minimum hardware header length
536721b1
KK
1529 *
1530 * @needed_headroom: Extra headroom the hardware may need, but not in all
1531 * cases can this be guaranteed
1532 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1533 * cases can this be guaranteed. Some cases also use
1534 * LL_MAX_HEADER instead to allocate the skb
1535 *
1536 * interface address info:
1537 *
1538 * @perm_addr: Permanent hw address
1539 * @addr_assign_type: Hw address assignment type
1540 * @addr_len: Hardware address length
8626a0c8 1541 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1542 * @dev_id: Used to differentiate devices that share
1543 * the same link layer address
1544 * @dev_port: Used to differentiate devices that share
1545 * the same function
1546 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1547 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1548 * has been enabled due to the need to listen to
1549 * additional unicast addresses in a device that
1550 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1551 * @uc: unicast mac addresses
1552 * @mc: multicast mac addresses
1553 * @dev_addrs: list of device hw addresses
1554 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1555 * @promiscuity: Number of times the NIC is told to work in
1556 * promiscuous mode; if it becomes 0 the NIC will
1557 * exit promiscuous mode
536721b1
KK
1558 * @allmulti: Counter, enables or disables allmulticast mode
1559 *
1560 * @vlan_info: VLAN info
1561 * @dsa_ptr: dsa specific data
1562 * @tipc_ptr: TIPC specific data
1563 * @atalk_ptr: AppleTalk link
1564 * @ip_ptr: IPv4 specific data
1565 * @dn_ptr: DECnet specific data
1566 * @ip6_ptr: IPv6 specific data
1567 * @ax25_ptr: AX.25 specific data
1568 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1569 *
536721b1
KK
1570 * @dev_addr: Hw address (before bcast,
1571 * because most packets are unicast)
1572 *
1573 * @_rx: Array of RX queues
1574 * @num_rx_queues: Number of RX queues
1575 * allocated at register_netdev() time
1576 * @real_num_rx_queues: Number of RX queues currently active in device
1577 *
1578 * @rx_handler: handler for received packets
1579 * @rx_handler_data: XXX: need comments on this one
46209401
JP
1580 * @miniq_ingress: ingress/clsact qdisc specific data for
1581 * ingress processing
536721b1
KK
1582 * @ingress_queue: XXX: need comments on this one
1583 * @broadcast: hw bcast address
1584 *
14ffbbb8
TG
1585 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1586 * indexed by RX queue number. Assigned by driver.
1587 * This must only be set if the ndo_rx_flow_steer
1588 * operation is defined
1589 * @index_hlist: Device index hash chain
1590 *
536721b1
KK
1591 * @_tx: Array of TX queues
1592 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1593 * @real_num_tx_queues: Number of TX queues currently active in device
1594 * @qdisc: Root qdisc from userspace point of view
1595 * @tx_queue_len: Max frames per queue allowed
1596 * @tx_global_lock: XXX: need comments on this one
1597 *
1598 * @xps_maps: XXX: need comments on this one
46209401
JP
1599 * @miniq_egress: clsact qdisc specific data for
1600 * egress processing
536721b1 1601 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1602 * the watchdog (see dev_watchdog())
536721b1
KK
1603 * @watchdog_timer: List of timers
1604 *
1605 * @pcpu_refcnt: Number of references to this device
1606 * @todo_list: Delayed register/unregister
536721b1
KK
1607 * @link_watch_list: XXX: need comments on this one
1608 *
1609 * @reg_state: Register/unregister state machine
1610 * @dismantle: Device is going to be freed
1611 * @rtnl_link_state: This enum represents the phases of creating
1612 * a new link
1613 *
cf124db5
DM
1614 * @needs_free_netdev: Should unregister perform free_netdev?
1615 * @priv_destructor: Called from unregister
536721b1
KK
1616 * @npinfo: XXX: need comments on this one
1617 * @nd_net: Network namespace this network device is inside
1618 *
1619 * @ml_priv: Mid-layer private
1620 * @lstats: Loopback statistics
1621 * @tstats: Tunnel statistics
1622 * @dstats: Dummy statistics
1623 * @vstats: Virtual ethernet statistics
1624 *
1625 * @garp_port: GARP
1626 * @mrp_port: MRP
1627 *
1628 * @dev: Class/net/name entry
1629 * @sysfs_groups: Space for optional device, statistics and wireless
1630 * sysfs groups
1631 *
1632 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1633 * @rtnl_link_ops: Rtnl_link_ops
1634 *
1635 * @gso_max_size: Maximum size of generic segmentation offload
1636 * @gso_max_segs: Maximum number of segments that can be passed to the
1637 * NIC for GSO
1638 *
1639 * @dcbnl_ops: Data Center Bridging netlink ops
1640 * @num_tc: Number of traffic classes in the net device
1641 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1642 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1643 *
1644 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1645 *
1646 * @priomap: XXX: need comments on this one
1647 * @phydev: Physical device may attach itself
1648 * for hardware timestamping
1649 *
123b3652
ED
1650 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1651 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1652 *
d746d707
AK
1653 * @proto_down: protocol port state information can be sent to the
1654 * switch driver and used to set the phys state of the
1655 * switch port.
1656 *
1da177e4
LT
1657 * FIXME: cleanup struct net_device such that network protocol info
1658 * moves out.
1659 */
1660
d94d9fee 1661struct net_device {
1da177e4 1662 char name[IFNAMSIZ];
9356b8fc 1663 struct hlist_node name_hlist;
6c557001 1664 struct dev_ifalias __rcu *ifalias;
1da177e4
LT
1665 /*
1666 * I/O specific fields
1667 * FIXME: Merge these and struct ifmap into one
1668 */
536721b1
KK
1669 unsigned long mem_end;
1670 unsigned long mem_start;
1671 unsigned long base_addr;
1672 int irq;
1da177e4 1673
14ffbbb8
TG
1674 atomic_t carrier_changes;
1675
1da177e4 1676 /*
536721b1
KK
1677 * Some hardware also needs these fields (state,dev_list,
1678 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1679 * part of the usual set specified in Space.c.
1680 */
1681
1da177e4
LT
1682 unsigned long state;
1683
7562f876 1684 struct list_head dev_list;
bea3348e 1685 struct list_head napi_list;
44a0873d 1686 struct list_head unreg_list;
5cde2829 1687 struct list_head close_list;
7866a621
SN
1688 struct list_head ptype_all;
1689 struct list_head ptype_specific;
2f268f12 1690
2f268f12
VF
1691 struct {
1692 struct list_head upper;
1693 struct list_head lower;
1694 } adj_list;
1695
c8f44aff 1696 netdev_features_t features;
c8f44aff 1697 netdev_features_t hw_features;
c8f44aff 1698 netdev_features_t wanted_features;
c8f44aff 1699 netdev_features_t vlan_features;
6a674e9c 1700 netdev_features_t hw_enc_features;
0d89d203 1701 netdev_features_t mpls_features;
802ab55a 1702 netdev_features_t gso_partial_features;
04ed3e74 1703
1da177e4 1704 int ifindex;
7a66bbc9 1705 int group;
1da177e4 1706
c45d286e 1707 struct net_device_stats stats;
015f0688 1708
015f0688
ED
1709 atomic_long_t rx_dropped;
1710 atomic_long_t tx_dropped;
6e7333d3 1711 atomic_long_t rx_nohandler;
1da177e4 1712
b86e0280 1713#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1714 const struct iw_handler_def *wireless_handlers;
1715 struct iw_public_data *wireless_data;
b86e0280 1716#endif
d314774c 1717 const struct net_device_ops *netdev_ops;
76fd8593 1718 const struct ethtool_ops *ethtool_ops;
4170604f 1719#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1720 const struct switchdev_ops *switchdev_ops;
4170604f 1721#endif
1b69c6d0
DA
1722#ifdef CONFIG_NET_L3_MASTER_DEV
1723 const struct l3mdev_ops *l3mdev_ops;
1724#endif
f997c55c
AA
1725#if IS_ENABLED(CONFIG_IPV6)
1726 const struct ndisc_ops *ndisc_ops;
1727#endif
1da177e4 1728
d77e38e6
SK
1729#ifdef CONFIG_XFRM
1730 const struct xfrmdev_ops *xfrmdev_ops;
1731#endif
1732
3b04ddde
SH
1733 const struct header_ops *header_ops;
1734
536721b1
KK
1735 unsigned int flags;
1736 unsigned int priv_flags;
1737
1da177e4 1738 unsigned short gflags;
536721b1 1739 unsigned short padded;
1da177e4 1740
536721b1
KK
1741 unsigned char operstate;
1742 unsigned char link_mode;
b00055aa 1743
536721b1
KK
1744 unsigned char if_port;
1745 unsigned char dma;
bdc220da 1746
536721b1 1747 unsigned int mtu;
61e84623
JW
1748 unsigned int min_mtu;
1749 unsigned int max_mtu;
536721b1
KK
1750 unsigned short type;
1751 unsigned short hard_header_len;
d92be7a4 1752 unsigned char min_header_len;
1da177e4 1753
f5184d26
JB
1754 unsigned short needed_headroom;
1755 unsigned short needed_tailroom;
1756
1da177e4 1757 /* Interface address info. */
536721b1
KK
1758 unsigned char perm_addr[MAX_ADDR_LEN];
1759 unsigned char addr_assign_type;
1760 unsigned char addr_len;
a0a9663d 1761 unsigned short neigh_priv_len;
536721b1
KK
1762 unsigned short dev_id;
1763 unsigned short dev_port;
ccffad25 1764 spinlock_t addr_list_lock;
14ffbbb8
TG
1765 unsigned char name_assign_type;
1766 bool uc_promisc;
536721b1
KK
1767 struct netdev_hw_addr_list uc;
1768 struct netdev_hw_addr_list mc;
1769 struct netdev_hw_addr_list dev_addrs;
1770
4c3d5e7b
ED
1771#ifdef CONFIG_SYSFS
1772 struct kset *queues_kset;
1773#endif
9d45abe1
WC
1774 unsigned int promiscuity;
1775 unsigned int allmulti;
1da177e4 1776
1da177e4 1777
5e82b4b2 1778 /* Protocol-specific pointers */
65ac6a5f 1779
d11ead75 1780#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1781 struct vlan_info __rcu *vlan_info;
65ac6a5f 1782#endif
34a430d7 1783#if IS_ENABLED(CONFIG_NET_DSA)
2f657a60 1784 struct dsa_port *dsa_ptr;
37cb0620
YX
1785#endif
1786#if IS_ENABLED(CONFIG_TIPC)
536721b1 1787 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1788#endif
536721b1
KK
1789 void *atalk_ptr;
1790 struct in_device __rcu *ip_ptr;
1791 struct dn_dev __rcu *dn_ptr;
1792 struct inet6_dev __rcu *ip6_ptr;
1793 void *ax25_ptr;
1794 struct wireless_dev *ieee80211_ptr;
98a18b6f 1795 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1796#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1797 struct mpls_dev __rcu *mpls_ptr;
1798#endif
1da177e4 1799
9356b8fc 1800/*
cd13539b 1801 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1802 */
9356b8fc 1803 /* Interface address info used in eth_type_trans() */
536721b1 1804 unsigned char *dev_addr;
f001fde5 1805
a953be53 1806#ifdef CONFIG_SYSFS
0a9627f2
TH
1807 struct netdev_rx_queue *_rx;
1808
0a9627f2 1809 unsigned int num_rx_queues;
62fe0b40 1810 unsigned int real_num_rx_queues;
df334545 1811#endif
0a9627f2 1812
7acedaf5 1813 struct bpf_prog __rcu *xdp_prog;
3b47d303 1814 unsigned long gro_flush_timeout;
61391cde 1815 rx_handler_func_t __rcu *rx_handler;
1816 void __rcu *rx_handler_data;
e8a0464c 1817
4cda01e8 1818#ifdef CONFIG_NET_CLS_ACT
46209401 1819 struct mini_Qdisc __rcu *miniq_ingress;
d2788d34 1820#endif
24824a09 1821 struct netdev_queue __rcu *ingress_queue;
e687ad60 1822#ifdef CONFIG_NETFILTER_INGRESS
960632ec 1823 struct nf_hook_entries __rcu *nf_hooks_ingress;
e687ad60 1824#endif
d2788d34 1825
536721b1 1826 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1827#ifdef CONFIG_RFS_ACCEL
1828 struct cpu_rmap *rx_cpu_rmap;
1829#endif
1830 struct hlist_node index_hlist;
cd13539b
ED
1831
1832/*
1833 * Cache lines mostly used on transmit path
1834 */
e8a0464c
DM
1835 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1836 unsigned int num_tx_queues;
fd2ea0a7 1837 unsigned int real_num_tx_queues;
af356afa 1838 struct Qdisc *qdisc;
59cc1f61
JK
1839#ifdef CONFIG_NET_SCHED
1840 DECLARE_HASHTABLE (qdisc_hash, 4);
1841#endif
0cd29503 1842 unsigned int tx_queue_len;
c3f26a26 1843 spinlock_t tx_global_lock;
14ffbbb8 1844 int watchdog_timeo;
cd13539b 1845
bf264145 1846#ifdef CONFIG_XPS
a4177869 1847 struct xps_dev_maps __rcu *xps_maps;
bf264145 1848#endif
1f211a1b 1849#ifdef CONFIG_NET_CLS_ACT
46209401 1850 struct mini_Qdisc __rcu *miniq_egress;
1f211a1b 1851#endif
0c4f691f 1852
9356b8fc 1853 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1854 struct timer_list watchdog_timer;
1855
29b4433d 1856 int __percpu *pcpu_refcnt;
1da177e4 1857 struct list_head todo_list;
1da177e4 1858
e014debe 1859 struct list_head link_watch_list;
572a103d 1860
1da177e4 1861 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1862 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1863 NETREG_UNREGISTERING, /* called unregister_netdevice */
1864 NETREG_UNREGISTERED, /* completed unregister todo */
1865 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1866 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1867 } reg_state:8;
1868
536721b1 1869 bool dismantle;
a2835763
PM
1870
1871 enum {
1872 RTNL_LINK_INITIALIZED,
1873 RTNL_LINK_INITIALIZING,
1874 } rtnl_link_state:16;
1da177e4 1875
cf124db5
DM
1876 bool needs_free_netdev;
1877 void (*priv_destructor)(struct net_device *dev);
1da177e4 1878
1da177e4 1879#ifdef CONFIG_NETPOLL
5fbee843 1880 struct netpoll_info __rcu *npinfo;
1da177e4 1881#endif
eae792b7 1882
0c5c9fb5 1883 possible_net_t nd_net;
4a1c5371 1884
4951704b 1885 /* mid-layer private */
a7855c78 1886 union {
536721b1
KK
1887 void *ml_priv;
1888 struct pcpu_lstats __percpu *lstats;
8f84985f 1889 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1890 struct pcpu_dstats __percpu *dstats;
1891 struct pcpu_vstats __percpu *vstats;
a7855c78 1892 };
536721b1 1893
fb585b44 1894#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1895 struct garp_port __rcu *garp_port;
fb585b44
TK
1896#endif
1897#if IS_ENABLED(CONFIG_MRP)
febf018d 1898 struct mrp_port __rcu *mrp_port;
fb585b44 1899#endif
1da177e4 1900
5e82b4b2 1901 struct device dev;
0c509a6c 1902 const struct attribute_group *sysfs_groups[4];
a953be53 1903 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1904
38f7b870 1905 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1906
82cc1a7a
PWJ
1907 /* for setting kernel sock attribute on TCP connection setup */
1908#define GSO_MAX_SIZE 65536
1909 unsigned int gso_max_size;
30b678d8
BH
1910#define GSO_MAX_SEGS 65535
1911 u16 gso_max_segs;
743b03a8 1912
7a6b6f51 1913#ifdef CONFIG_DCB
32953543 1914 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1915#endif
5e82b4b2
BH
1916 u8 num_tc;
1917 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1918 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1919
d11ead75 1920#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1921 unsigned int fcoe_ddp_xid;
5bc1421e 1922#endif
86f8515f 1923#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1924 struct netprio_map __rcu *priomap;
4d288d57 1925#endif
5e82b4b2
BH
1926 struct phy_device *phydev;
1927 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1928 struct lock_class_key *qdisc_running_key;
5e82b4b2 1929 bool proto_down;
1da177e4 1930};
43cb76d9 1931#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1932
b5cdae32
DM
1933static inline bool netif_elide_gro(const struct net_device *dev)
1934{
1935 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1936 return true;
1937 return false;
1938}
1939
1da177e4 1940#define NETDEV_ALIGN 32
1da177e4 1941
4f57c087
JF
1942static inline
1943int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1944{
1945 return dev->prio_tc_map[prio & TC_BITMASK];
1946}
1947
1948static inline
1949int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1950{
1951 if (tc >= dev->num_tc)
1952 return -EINVAL;
1953
1954 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1955 return 0;
1956}
1957
8d059b0f 1958int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1959void netdev_reset_tc(struct net_device *dev);
1960int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1961int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1962
1963static inline
1964int netdev_get_num_tc(struct net_device *dev)
1965{
1966 return dev->num_tc;
1967}
1968
e8a0464c
DM
1969static inline
1970struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1971 unsigned int index)
1972{
1973 return &dev->_tx[index];
1974}
1975
10c51b56
DB
1976static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1977 const struct sk_buff *skb)
1978{
1979 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1980}
1981
e8a0464c
DM
1982static inline void netdev_for_each_tx_queue(struct net_device *dev,
1983 void (*f)(struct net_device *,
1984 struct netdev_queue *,
1985 void *),
1986 void *arg)
1987{
1988 unsigned int i;
1989
1990 for (i = 0; i < dev->num_tx_queues; i++)
1991 f(dev, &dev->_tx[i], arg);
1992}
1993
d3fff6c4
ED
1994#define netdev_lockdep_set_classes(dev) \
1995{ \
1996 static struct lock_class_key qdisc_tx_busylock_key; \
1997 static struct lock_class_key qdisc_running_key; \
1998 static struct lock_class_key qdisc_xmit_lock_key; \
1999 static struct lock_class_key dev_addr_list_lock_key; \
2000 unsigned int i; \
2001 \
2002 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2003 (dev)->qdisc_running_key = &qdisc_running_key; \
2004 lockdep_set_class(&(dev)->addr_list_lock, \
2005 &dev_addr_list_lock_key); \
2006 for (i = 0; i < (dev)->num_tx_queues; i++) \
2007 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2008 &qdisc_xmit_lock_key); \
2009}
2010
f629d208 2011struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
2012 struct sk_buff *skb,
2013 void *accel_priv);
8c4c49df 2014
871b642a
PA
2015/* returns the headroom that the master device needs to take in account
2016 * when forwarding to this dev
2017 */
2018static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2019{
2020 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2021}
2022
2023static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2024{
2025 if (dev->netdev_ops->ndo_set_rx_headroom)
2026 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2027}
2028
2029/* set the device rx headroom to the dev's default */
2030static inline void netdev_reset_rx_headroom(struct net_device *dev)
2031{
2032 netdev_set_rx_headroom(dev, -1);
2033}
2034
c346dca1
YH
2035/*
2036 * Net namespace inlines
2037 */
2038static inline
2039struct net *dev_net(const struct net_device *dev)
2040{
c2d9ba9b 2041 return read_pnet(&dev->nd_net);
c346dca1
YH
2042}
2043
2044static inline
f5aa23fd 2045void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2046{
0c5c9fb5 2047 write_pnet(&dev->nd_net, net);
c346dca1
YH
2048}
2049
bea3348e
SH
2050/**
2051 * netdev_priv - access network device private data
2052 * @dev: network device
2053 *
2054 * Get network device private data
2055 */
6472ce60 2056static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2057{
1ce8e7b5 2058 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2059}
2060
1da177e4
LT
2061/* Set the sysfs physical device reference for the network logical device
2062 * if set prior to registration will cause a symlink during initialization.
2063 */
43cb76d9 2064#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2065
384912ed 2066/* Set the sysfs device type for the network logical device to allow
3f79410c 2067 * fine-grained identification of different network device types. For
5e82b4b2 2068 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2069 */
2070#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2071
82dc3c63
ED
2072/* Default NAPI poll() weight
2073 * Device drivers are strongly advised to not use bigger value
2074 */
2075#define NAPI_POLL_WEIGHT 64
2076
3b582cc1 2077/**
5e82b4b2 2078 * netif_napi_add - initialize a NAPI context
3b582cc1 2079 * @dev: network device
5e82b4b2 2080 * @napi: NAPI context
3b582cc1
SH
2081 * @poll: polling function
2082 * @weight: default weight
2083 *
5e82b4b2
BH
2084 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2085 * *any* of the other NAPI-related functions.
3b582cc1 2086 */
d565b0a1
HX
2087void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2088 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2089
d64b5e85 2090/**
5e82b4b2 2091 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2092 * @dev: network device
5e82b4b2 2093 * @napi: NAPI context
d64b5e85
ED
2094 * @poll: polling function
2095 * @weight: default weight
2096 *
2097 * This variant of netif_napi_add() should be used from drivers using NAPI
2098 * to exclusively poll a TX queue.
2099 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2100 */
2101static inline void netif_tx_napi_add(struct net_device *dev,
2102 struct napi_struct *napi,
2103 int (*poll)(struct napi_struct *, int),
2104 int weight)
2105{
2106 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2107 netif_napi_add(dev, napi, poll, weight);
2108}
2109
d8156534 2110/**
5e82b4b2
BH
2111 * netif_napi_del - remove a NAPI context
2112 * @napi: NAPI context
d8156534 2113 *
5e82b4b2 2114 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2115 */
d565b0a1
HX
2116void netif_napi_del(struct napi_struct *napi);
2117
2118struct napi_gro_cb {
78a478d0 2119 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2120 void *frag0;
78a478d0 2121
7489594c
HX
2122 /* Length of frag0. */
2123 unsigned int frag0_len;
2124
86911732 2125 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2126 int data_offset;
86911732 2127
d565b0a1 2128 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2129 u16 flush;
2130
2131 /* Save the IP ID here and check when we get to the transport layer */
2132 u16 flush_id;
d565b0a1
HX
2133
2134 /* Number of segments aggregated. */
2e71a6f8
ED
2135 u16 count;
2136
15e2396d
TH
2137 /* Start offset for remote checksum offload */
2138 u16 gro_remcsum_start;
2139
2e71a6f8
ED
2140 /* jiffies when first packet was created/queued */
2141 unsigned long age;
86347245 2142
afe93325 2143 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2144 u16 proto;
2145
baa32ff4
TH
2146 /* This is non-zero if the packet may be of the same flow. */
2147 u8 same_flow:1;
2148
fac8e0f5
JG
2149 /* Used in tunnel GRO receive */
2150 u8 encap_mark:1;
573e8fca
TH
2151
2152 /* GRO checksum is valid */
2153 u8 csum_valid:1;
2154
662880f4
TH
2155 /* Number of checksums via CHECKSUM_UNNECESSARY */
2156 u8 csum_cnt:3;
c3c7c254 2157
baa32ff4
TH
2158 /* Free the skb? */
2159 u8 free:2;
2160#define NAPI_GRO_FREE 1
2161#define NAPI_GRO_FREE_STOLEN_HEAD 2
2162
efc98d08
TH
2163 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2164 u8 is_ipv6:1;
2165
a0ca153f
AD
2166 /* Used in GRE, set in fou/gue_gro_receive */
2167 u8 is_fou:1;
2168
1530545e
AD
2169 /* Used to determine if flush_id can be ignored */
2170 u8 is_atomic:1;
2171
fcd91dd4
SD
2172 /* Number of gro_receive callbacks this packet already went through */
2173 u8 recursion_counter:4;
2174
2175 /* 1 bit hole */
baa32ff4 2176
bf5a755f
JC
2177 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2178 __wsum csum;
2179
c3c7c254
ED
2180 /* used in skb_gro_receive() slow path */
2181 struct sk_buff *last;
d565b0a1
HX
2182};
2183
2184#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2185
fcd91dd4
SD
2186#define GRO_RECURSION_LIMIT 15
2187static inline int gro_recursion_inc_test(struct sk_buff *skb)
2188{
2189 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2190}
2191
2192typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2193static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2194 struct sk_buff **head,
2195 struct sk_buff *skb)
2196{
2197 if (unlikely(gro_recursion_inc_test(skb))) {
2198 NAPI_GRO_CB(skb)->flush |= 1;
2199 return NULL;
2200 }
2201
2202 return cb(head, skb);
2203}
2204
2205typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2206 struct sk_buff *);
2207static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2208 struct sock *sk,
2209 struct sk_buff **head,
2210 struct sk_buff *skb)
2211{
2212 if (unlikely(gro_recursion_inc_test(skb))) {
2213 NAPI_GRO_CB(skb)->flush |= 1;
2214 return NULL;
2215 }
2216
2217 return cb(sk, head, skb);
2218}
2219
1da177e4 2220struct packet_type {
f2ccd8fa
DM
2221 __be16 type; /* This is really htons(ether_type). */
2222 struct net_device *dev; /* NULL is wildcarded here */
2223 int (*func) (struct sk_buff *,
2224 struct net_device *,
2225 struct packet_type *,
2226 struct net_device *);
c0de08d0
EL
2227 bool (*id_match)(struct packet_type *ptype,
2228 struct sock *sk);
1da177e4
LT
2229 void *af_packet_priv;
2230 struct list_head list;
2231};
2232
f191a1d1 2233struct offload_callbacks {
576a30eb 2234 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2235 netdev_features_t features);
d565b0a1 2236 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2237 struct sk_buff *skb);
299603e8 2238 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2239};
2240
2241struct packet_offload {
2242 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2243 u16 priority;
f191a1d1
VY
2244 struct offload_callbacks callbacks;
2245 struct list_head list;
1da177e4
LT
2246};
2247
5e82b4b2 2248/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2249struct pcpu_sw_netstats {
2250 u64 rx_packets;
2251 u64 rx_bytes;
2252 u64 tx_packets;
2253 u64 tx_bytes;
2254 struct u64_stats_sync syncp;
2255};
2256
aabc92bb
PNA
2257#define __netdev_alloc_pcpu_stats(type, gfp) \
2258({ \
2259 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2260 if (pcpu_stats) { \
2261 int __cpu; \
2262 for_each_possible_cpu(__cpu) { \
2263 typeof(type) *stat; \
2264 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2265 u64_stats_init(&stat->syncp); \
2266 } \
2267 } \
2268 pcpu_stats; \
1c213bd2
WC
2269})
2270
aabc92bb 2271#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2272 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2273
764f5e54
JP
2274enum netdev_lag_tx_type {
2275 NETDEV_LAG_TX_TYPE_UNKNOWN,
2276 NETDEV_LAG_TX_TYPE_RANDOM,
2277 NETDEV_LAG_TX_TYPE_BROADCAST,
2278 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2279 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2280 NETDEV_LAG_TX_TYPE_HASH,
2281};
2282
2283struct netdev_lag_upper_info {
2284 enum netdev_lag_tx_type tx_type;
2285};
2286
fb1b2e3c
JP
2287struct netdev_lag_lower_state_info {
2288 u8 link_up : 1,
2289 tx_enabled : 1;
2290};
2291
1da177e4
LT
2292#include <linux/notifier.h>
2293
dcfe1421
AW
2294/* netdevice notifier chain. Please remember to update the rtnetlink
2295 * notification exclusion list in rtnetlink_event() when adding new
2296 * types.
2297 */
2298#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2299#define NETDEV_DOWN 0x0002
2300#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2301 detected a hardware crash and restarted
2302 - we can use this eg to kick tcp sessions
2303 once done */
2304#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2305#define NETDEV_REGISTER 0x0005
2306#define NETDEV_UNREGISTER 0x0006
1d486bfb 2307#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2308#define NETDEV_CHANGEADDR 0x0008
2309#define NETDEV_GOING_DOWN 0x0009
2310#define NETDEV_CHANGENAME 0x000A
2311#define NETDEV_FEAT_CHANGE 0x000B
2312#define NETDEV_BONDING_FAILOVER 0x000C
2313#define NETDEV_PRE_UP 0x000D
2314#define NETDEV_PRE_TYPE_CHANGE 0x000E
2315#define NETDEV_POST_TYPE_CHANGE 0x000F
2316#define NETDEV_POST_INIT 0x0010
0115e8e3 2317#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2318#define NETDEV_RELEASE 0x0012
2319#define NETDEV_NOTIFY_PEERS 0x0013
2320#define NETDEV_JOIN 0x0014
42e52bf9 2321#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2322#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2323#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2324#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2325#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2326#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2327#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2328#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
296d8ee3 2329#define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
08294a26 2330#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2331
f629d208
JP
2332int register_netdevice_notifier(struct notifier_block *nb);
2333int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2334
2335struct netdev_notifier_info {
51d0c047
DA
2336 struct net_device *dev;
2337 struct netlink_ext_ack *extack;
351638e7
JP
2338};
2339
be9efd36
JP
2340struct netdev_notifier_change_info {
2341 struct netdev_notifier_info info; /* must be first */
2342 unsigned int flags_changed;
2343};
2344
0e4ead9d
JP
2345struct netdev_notifier_changeupper_info {
2346 struct netdev_notifier_info info; /* must be first */
2347 struct net_device *upper_dev; /* new upper dev */
2348 bool master; /* is upper dev master */
5e82b4b2 2349 bool linking; /* is the notification for link or unlink */
29bf24af 2350 void *upper_info; /* upper dev info */
0e4ead9d
JP
2351};
2352
04d48266
JP
2353struct netdev_notifier_changelowerstate_info {
2354 struct netdev_notifier_info info; /* must be first */
2355 void *lower_state_info; /* is lower dev state */
2356};
2357
75538c2b
CW
2358static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2359 struct net_device *dev)
2360{
2361 info->dev = dev;
51d0c047 2362 info->extack = NULL;
75538c2b
CW
2363}
2364
351638e7
JP
2365static inline struct net_device *
2366netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2367{
2368 return info->dev;
2369}
2370
51d0c047
DA
2371static inline struct netlink_ext_ack *
2372netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2373{
2374 return info->extack;
2375}
2376
f629d208 2377int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2378
2379
1da177e4
LT
2380extern rwlock_t dev_base_lock; /* Device list lock */
2381
881d966b
EB
2382#define for_each_netdev(net, d) \
2383 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2384#define for_each_netdev_reverse(net, d) \
2385 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2386#define for_each_netdev_rcu(net, d) \
2387 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2388#define for_each_netdev_safe(net, d, n) \
2389 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2390#define for_each_netdev_continue(net, d) \
2391 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2392#define for_each_netdev_continue_rcu(net, d) \
2393 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2394#define for_each_netdev_in_bond_rcu(bond, slave) \
2395 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2396 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2397#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2398
a050c33f
DL
2399static inline struct net_device *next_net_device(struct net_device *dev)
2400{
2401 struct list_head *lh;
2402 struct net *net;
2403
c346dca1 2404 net = dev_net(dev);
a050c33f
DL
2405 lh = dev->dev_list.next;
2406 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2407}
2408
ce81b76a
ED
2409static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2410{
2411 struct list_head *lh;
2412 struct net *net;
2413
2414 net = dev_net(dev);
ccf43438 2415 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2416 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2417}
2418
a050c33f
DL
2419static inline struct net_device *first_net_device(struct net *net)
2420{
2421 return list_empty(&net->dev_base_head) ? NULL :
2422 net_device_entry(net->dev_base_head.next);
2423}
7562f876 2424
ccf43438
ED
2425static inline struct net_device *first_net_device_rcu(struct net *net)
2426{
2427 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2428
2429 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2430}
2431
f629d208
JP
2432int netdev_boot_setup_check(struct net_device *dev);
2433unsigned long netdev_boot_base(const char *prefix, int unit);
2434struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2435 const char *hwaddr);
2436struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2437struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2438void dev_add_pack(struct packet_type *pt);
2439void dev_remove_pack(struct packet_type *pt);
2440void __dev_remove_pack(struct packet_type *pt);
2441void dev_add_offload(struct packet_offload *po);
2442void dev_remove_offload(struct packet_offload *po);
f629d208 2443
a54acb3a 2444int dev_get_iflink(const struct net_device *dev);
fc4099f1 2445int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2446struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2447 unsigned short mask);
f629d208
JP
2448struct net_device *dev_get_by_name(struct net *net, const char *name);
2449struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2450struct net_device *__dev_get_by_name(struct net *net, const char *name);
2451int dev_alloc_name(struct net_device *dev, const char *name);
2452int dev_open(struct net_device *dev);
7051b88a 2453void dev_close(struct net_device *dev);
2454void dev_close_many(struct list_head *head, bool unlink);
f629d208 2455void dev_disable_lro(struct net_device *dev);
0c4b51f0 2456int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2457int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2458int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2459int register_netdevice(struct net_device *dev);
2460void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2461void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2462static inline void unregister_netdevice(struct net_device *dev)
2463{
2464 unregister_netdevice_queue(dev, NULL);
2465}
2466
f629d208
JP
2467int netdev_refcnt_read(const struct net_device *dev);
2468void free_netdev(struct net_device *dev);
74d332c1 2469void netdev_freemem(struct net_device *dev);
f629d208
JP
2470void synchronize_net(void);
2471int init_dummy_netdev(struct net_device *dev);
937f1ba5 2472
f60e5990 2473DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2474#define XMIT_RECURSION_LIMIT 10
2475
f60e5990 2476static inline int dev_recursion_level(void)
2477{
2478 return this_cpu_read(xmit_recursion);
2479}
2480
f629d208
JP
2481struct net_device *dev_get_by_index(struct net *net, int ifindex);
2482struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2483struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2484struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2485int netdev_get_name(struct net *net, char *name, int ifindex);
2486int dev_restart(struct net_device *dev);
f629d208 2487int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2488
2489static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2490{
2491 return NAPI_GRO_CB(skb)->data_offset;
2492}
2493
2494static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2495{
2496 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2497}
2498
2499static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2500{
2501 NAPI_GRO_CB(skb)->data_offset += len;
2502}
2503
a5b1cf28
HX
2504static inline void *skb_gro_header_fast(struct sk_buff *skb,
2505 unsigned int offset)
86911732 2506{
a5b1cf28
HX
2507 return NAPI_GRO_CB(skb)->frag0 + offset;
2508}
78a478d0 2509
a5b1cf28
HX
2510static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2511{
2512 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2513}
78a478d0 2514
57ea52a8
HX
2515static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2516{
2517 NAPI_GRO_CB(skb)->frag0 = NULL;
2518 NAPI_GRO_CB(skb)->frag0_len = 0;
2519}
2520
a5b1cf28
HX
2521static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2522 unsigned int offset)
2523{
17dd759c
HX
2524 if (!pskb_may_pull(skb, hlen))
2525 return NULL;
2526
57ea52a8 2527 skb_gro_frag0_invalidate(skb);
17dd759c 2528 return skb->data + offset;
86911732 2529}
1da177e4 2530
36e7b1b8
HX
2531static inline void *skb_gro_network_header(struct sk_buff *skb)
2532{
78d3fd0b
HX
2533 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2534 skb_network_offset(skb);
36e7b1b8
HX
2535}
2536
bf5a755f
JC
2537static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2538 const void *start, unsigned int len)
2539{
573e8fca 2540 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2541 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2542 csum_partial(start, len, 0));
2543}
2544
573e8fca
TH
2545/* GRO checksum functions. These are logical equivalents of the normal
2546 * checksum functions (in skbuff.h) except that they operate on the GRO
2547 * offsets and fields in sk_buff.
2548 */
2549
2550__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2551
15e2396d
TH
2552static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2553{
b7fe10e5 2554 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2555}
2556
573e8fca
TH
2557static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2558 bool zero_okay,
2559 __sum16 check)
2560{
6edec0e6
TH
2561 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2562 skb_checksum_start_offset(skb) <
2563 skb_gro_offset(skb)) &&
15e2396d 2564 !skb_at_gro_remcsum_start(skb) &&
662880f4 2565 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2566 (!zero_okay || check));
2567}
2568
2569static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2570 __wsum psum)
2571{
2572 if (NAPI_GRO_CB(skb)->csum_valid &&
2573 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2574 return 0;
2575
2576 NAPI_GRO_CB(skb)->csum = psum;
2577
2578 return __skb_gro_checksum_complete(skb);
2579}
2580
573e8fca
TH
2581static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2582{
662880f4
TH
2583 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2584 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2585 NAPI_GRO_CB(skb)->csum_cnt--;
2586 } else {
2587 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2588 * verified a new top level checksum or an encapsulated one
2589 * during GRO. This saves work if we fallback to normal path.
2590 */
2591 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2592 }
2593}
2594
2595#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2596 compute_pseudo) \
2597({ \
2598 __sum16 __ret = 0; \
2599 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2600 __ret = __skb_gro_checksum_validate_complete(skb, \
2601 compute_pseudo(skb, proto)); \
219f1d79 2602 if (!__ret) \
573e8fca
TH
2603 skb_gro_incr_csum_unnecessary(skb); \
2604 __ret; \
2605})
2606
2607#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2608 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2609
2610#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2611 compute_pseudo) \
2612 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2613
2614#define skb_gro_checksum_simple_validate(skb) \
2615 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2616
d96535a1
TH
2617static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2618{
2619 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2620 !NAPI_GRO_CB(skb)->csum_valid);
2621}
2622
2623static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2624 __sum16 check, __wsum pseudo)
2625{
2626 NAPI_GRO_CB(skb)->csum = ~pseudo;
2627 NAPI_GRO_CB(skb)->csum_valid = 1;
2628}
2629
2630#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2631do { \
2632 if (__skb_gro_checksum_convert_check(skb)) \
2633 __skb_gro_checksum_convert(skb, check, \
2634 compute_pseudo(skb, proto)); \
2635} while (0)
2636
26c4f7da
TH
2637struct gro_remcsum {
2638 int offset;
2639 __wsum delta;
2640};
2641
2642static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2643{
846cd667 2644 grc->offset = 0;
26c4f7da
TH
2645 grc->delta = 0;
2646}
2647
b7fe10e5
TH
2648static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2649 unsigned int off, size_t hdrlen,
2650 int start, int offset,
2651 struct gro_remcsum *grc,
2652 bool nopartial)
dcdc8994
TH
2653{
2654 __wsum delta;
b7fe10e5 2655 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2656
2657 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2658
15e2396d 2659 if (!nopartial) {
b7fe10e5
TH
2660 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2661 return ptr;
2662 }
2663
2664 ptr = skb_gro_header_fast(skb, off);
2665 if (skb_gro_header_hard(skb, off + plen)) {
2666 ptr = skb_gro_header_slow(skb, off + plen, off);
2667 if (!ptr)
2668 return NULL;
15e2396d
TH
2669 }
2670
b7fe10e5
TH
2671 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2672 start, offset);
dcdc8994
TH
2673
2674 /* Adjust skb->csum since we changed the packet */
dcdc8994 2675 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2676
b7fe10e5 2677 grc->offset = off + hdrlen + offset;
26c4f7da 2678 grc->delta = delta;
b7fe10e5
TH
2679
2680 return ptr;
dcdc8994
TH
2681}
2682
26c4f7da
TH
2683static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2684 struct gro_remcsum *grc)
2685{
b7fe10e5
TH
2686 void *ptr;
2687 size_t plen = grc->offset + sizeof(u16);
2688
26c4f7da
TH
2689 if (!grc->delta)
2690 return;
2691
b7fe10e5
TH
2692 ptr = skb_gro_header_fast(skb, grc->offset);
2693 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2694 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2695 if (!ptr)
2696 return;
2697 }
2698
2699 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2700}
dcdc8994 2701
25393d3f
SK
2702#ifdef CONFIG_XFRM_OFFLOAD
2703static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2704{
2705 if (PTR_ERR(pp) != -EINPROGRESS)
2706 NAPI_GRO_CB(skb)->flush |= flush;
2707}
2708#else
5f114163
SK
2709static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2710{
2711 NAPI_GRO_CB(skb)->flush |= flush;
2712}
25393d3f 2713#endif
5f114163 2714
0c4e8581
SH
2715static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2716 unsigned short type,
3b04ddde 2717 const void *daddr, const void *saddr,
95c96174 2718 unsigned int len)
0c4e8581 2719{
f1ecfd5d 2720 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2721 return 0;
3b04ddde
SH
2722
2723 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2724}
2725
b95cce35
SH
2726static inline int dev_parse_header(const struct sk_buff *skb,
2727 unsigned char *haddr)
2728{
2729 const struct net_device *dev = skb->dev;
2730
1b83336b 2731 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2732 return 0;
3b04ddde 2733 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2734}
2735
2793a23a
WB
2736/* ll_header must have at least hard_header_len allocated */
2737static inline bool dev_validate_header(const struct net_device *dev,
2738 char *ll_header, int len)
2739{
2740 if (likely(len >= dev->hard_header_len))
2741 return true;
217e6fa2
WB
2742 if (len < dev->min_header_len)
2743 return false;
2793a23a
WB
2744
2745 if (capable(CAP_SYS_RAWIO)) {
2746 memset(ll_header + len, 0, dev->hard_header_len - len);
2747 return true;
2748 }
2749
2750 if (dev->header_ops && dev->header_ops->validate)
2751 return dev->header_ops->validate(ll_header, len);
2752
2753 return false;
2754}
2755
1da177e4 2756typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2757int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2758static inline int unregister_gifconf(unsigned int family)
2759{
2760 return register_gifconf(family, NULL);
2761}
2762
99bbc707 2763#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2764#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2765struct sd_flow_limit {
2766 u64 count;
2767 unsigned int num_buckets;
2768 unsigned int history_head;
2769 u16 history[FLOW_LIMIT_HISTORY];
2770 u8 buckets[];
2771};
2772
2773extern int netdev_flow_limit_table_len;
2774#endif /* CONFIG_NET_FLOW_LIMIT */
2775
1da177e4 2776/*
5e82b4b2 2777 * Incoming packets are placed on per-CPU queues
1da177e4 2778 */
d94d9fee 2779struct softnet_data {
1da177e4 2780 struct list_head poll_list;
6e7676c1 2781 struct sk_buff_head process_queue;
1da177e4 2782
dee42870 2783 /* stats */
cd7b5396
DM
2784 unsigned int processed;
2785 unsigned int time_squeeze;
cd7b5396 2786 unsigned int received_rps;
fd793d89 2787#ifdef CONFIG_RPS
88751275 2788 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2789#endif
2790#ifdef CONFIG_NET_FLOW_LIMIT
2791 struct sd_flow_limit __rcu *flow_limit;
2792#endif
2793 struct Qdisc *output_queue;
2794 struct Qdisc **output_queue_tailp;
2795 struct sk_buff *completion_queue;
88751275 2796
4cdb1e2e 2797#ifdef CONFIG_RPS
501e7ef5
ED
2798 /* input_queue_head should be written by cpu owning this struct,
2799 * and only read by other cpus. Worth using a cache line.
2800 */
2801 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2802
2803 /* Elements below can be accessed between CPUs for RPS/RFS */
966a9671 2804 call_single_data_t csd ____cacheline_aligned_in_smp;
88751275
ED
2805 struct softnet_data *rps_ipi_next;
2806 unsigned int cpu;
76cc8b13 2807 unsigned int input_queue_tail;
1e94d72f 2808#endif
95c96174 2809 unsigned int dropped;
0a9627f2 2810 struct sk_buff_head input_pkt_queue;
bea3348e 2811 struct napi_struct backlog;
99bbc707 2812
1da177e4
LT
2813};
2814
76cc8b13 2815static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2816{
2817#ifdef CONFIG_RPS
76cc8b13
TH
2818 sd->input_queue_head++;
2819#endif
2820}
2821
2822static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2823 unsigned int *qtail)
2824{
2825#ifdef CONFIG_RPS
2826 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2827#endif
2828}
2829
0a9627f2 2830DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2831
f629d208 2832void __netif_schedule(struct Qdisc *q);
46e5da40 2833void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2834
fd2ea0a7
DM
2835static inline void netif_tx_schedule_all(struct net_device *dev)
2836{
2837 unsigned int i;
2838
2839 for (i = 0; i < dev->num_tx_queues; i++)
2840 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2841}
2842
f9a7cbbf 2843static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2844{
73466498 2845 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2846}
2847
bea3348e
SH
2848/**
2849 * netif_start_queue - allow transmit
2850 * @dev: network device
2851 *
2852 * Allow upper layers to call the device hard_start_xmit routine.
2853 */
1da177e4
LT
2854static inline void netif_start_queue(struct net_device *dev)
2855{
e8a0464c 2856 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2857}
2858
fd2ea0a7
DM
2859static inline void netif_tx_start_all_queues(struct net_device *dev)
2860{
2861 unsigned int i;
2862
2863 for (i = 0; i < dev->num_tx_queues; i++) {
2864 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2865 netif_tx_start_queue(txq);
2866 }
2867}
2868
46e5da40 2869void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2870
d29f749e
DJ
2871/**
2872 * netif_wake_queue - restart transmit
2873 * @dev: network device
2874 *
2875 * Allow upper layers to call the device hard_start_xmit routine.
2876 * Used for flow control when transmit resources are available.
2877 */
79d16385
DM
2878static inline void netif_wake_queue(struct net_device *dev)
2879{
e8a0464c 2880 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2881}
2882
fd2ea0a7
DM
2883static inline void netif_tx_wake_all_queues(struct net_device *dev)
2884{
2885 unsigned int i;
2886
2887 for (i = 0; i < dev->num_tx_queues; i++) {
2888 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2889 netif_tx_wake_queue(txq);
2890 }
2891}
2892
f9a7cbbf 2893static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2894{
73466498 2895 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2896}
2897
bea3348e
SH
2898/**
2899 * netif_stop_queue - stop transmitted packets
2900 * @dev: network device
2901 *
2902 * Stop upper layers calling the device hard_start_xmit routine.
2903 * Used for flow control when transmit resources are unavailable.
2904 */
1da177e4
LT
2905static inline void netif_stop_queue(struct net_device *dev)
2906{
e8a0464c 2907 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2908}
2909
a2029240 2910void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2911
4d29515f 2912static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2913{
73466498 2914 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2915}
2916
bea3348e
SH
2917/**
2918 * netif_queue_stopped - test if transmit queue is flowblocked
2919 * @dev: network device
2920 *
2921 * Test if transmit queue on device is currently unable to send.
2922 */
4d29515f 2923static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2924{
e8a0464c 2925 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2926}
2927
4d29515f 2928static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2929{
73466498
TH
2930 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2931}
2932
8e2f1a63
DB
2933static inline bool
2934netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2935{
2936 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2937}
2938
8e2f1a63
DB
2939static inline bool
2940netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2941{
2942 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2943}
2944
53511453
ED
2945/**
2946 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2947 * @dev_queue: pointer to transmit queue
2948 *
2949 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2950 * to give appropriate hint to the CPU.
53511453
ED
2951 */
2952static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2953{
2954#ifdef CONFIG_BQL
2955 prefetchw(&dev_queue->dql.num_queued);
2956#endif
2957}
2958
2959/**
2960 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2961 * @dev_queue: pointer to transmit queue
2962 *
2963 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2964 * to give appropriate hint to the CPU.
53511453
ED
2965 */
2966static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2967{
2968#ifdef CONFIG_BQL
2969 prefetchw(&dev_queue->dql.limit);
2970#endif
2971}
2972
c5d67bd7
TH
2973static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2974 unsigned int bytes)
2975{
114cf580
TH
2976#ifdef CONFIG_BQL
2977 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2978
2979 if (likely(dql_avail(&dev_queue->dql) >= 0))
2980 return;
2981
2982 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2983
2984 /*
2985 * The XOFF flag must be set before checking the dql_avail below,
2986 * because in netdev_tx_completed_queue we update the dql_completed
2987 * before checking the XOFF flag.
2988 */
2989 smp_mb();
2990
2991 /* check again in case another CPU has just made room avail */
2992 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2993 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2994#endif
c5d67bd7
TH
2995}
2996
0042d0c8
FF
2997/**
2998 * netdev_sent_queue - report the number of bytes queued to hardware
2999 * @dev: network device
3000 * @bytes: number of bytes queued to the hardware device queue
3001 *
3002 * Report the number of bytes queued for sending/completion to the network
3003 * device hardware queue. @bytes should be a good approximation and should
3004 * exactly match netdev_completed_queue() @bytes
3005 */
c5d67bd7
TH
3006static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3007{
3008 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3009}
3010
3011static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 3012 unsigned int pkts, unsigned int bytes)
c5d67bd7 3013{
114cf580 3014#ifdef CONFIG_BQL
b37c0fbe
AD
3015 if (unlikely(!bytes))
3016 return;
3017
3018 dql_completed(&dev_queue->dql, bytes);
3019
3020 /*
3021 * Without the memory barrier there is a small possiblity that
3022 * netdev_tx_sent_queue will miss the update and cause the queue to
3023 * be stopped forever
3024 */
3025 smp_mb();
3026
3027 if (dql_avail(&dev_queue->dql) < 0)
3028 return;
3029
3030 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3031 netif_schedule_queue(dev_queue);
114cf580 3032#endif
c5d67bd7
TH
3033}
3034
0042d0c8
FF
3035/**
3036 * netdev_completed_queue - report bytes and packets completed by device
3037 * @dev: network device
3038 * @pkts: actual number of packets sent over the medium
3039 * @bytes: actual number of bytes sent over the medium
3040 *
3041 * Report the number of bytes and packets transmitted by the network device
3042 * hardware queue over the physical medium, @bytes must exactly match the
3043 * @bytes amount passed to netdev_sent_queue()
3044 */
c5d67bd7 3045static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3046 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3047{
3048 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3049}
3050
3051static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3052{
114cf580 3053#ifdef CONFIG_BQL
5c490354 3054 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3055 dql_reset(&q->dql);
3056#endif
c5d67bd7
TH
3057}
3058
0042d0c8
FF
3059/**
3060 * netdev_reset_queue - reset the packets and bytes count of a network device
3061 * @dev_queue: network device
3062 *
3063 * Reset the bytes and packet count of a network device and clear the
3064 * software flow control OFF bit for this network device
3065 */
c5d67bd7
TH
3066static inline void netdev_reset_queue(struct net_device *dev_queue)
3067{
3068 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3069}
3070
b9507bda
DB
3071/**
3072 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3073 * @dev: network device
3074 * @queue_index: given tx queue index
3075 *
3076 * Returns 0 if given tx queue index >= number of device tx queues,
3077 * otherwise returns the originally passed tx queue index.
3078 */
3079static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3080{
3081 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3082 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3083 dev->name, queue_index,
3084 dev->real_num_tx_queues);
3085 return 0;
3086 }
3087
3088 return queue_index;
3089}
3090
bea3348e
SH
3091/**
3092 * netif_running - test if up
3093 * @dev: network device
3094 *
3095 * Test if the device has been brought up.
3096 */
4d29515f 3097static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3098{
3099 return test_bit(__LINK_STATE_START, &dev->state);
3100}
3101
f25f4e44 3102/*
5e82b4b2 3103 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3104 * stop, and a check if it's stopped. All other device management is
3105 * done at the overall netdevice level.
3106 * Also test the device if we're multiqueue.
3107 */
bea3348e
SH
3108
3109/**
3110 * netif_start_subqueue - allow sending packets on subqueue
3111 * @dev: network device
3112 * @queue_index: sub queue index
3113 *
3114 * Start individual transmit queue of a device with multiple transmit queues.
3115 */
f25f4e44
PWJ
3116static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3117{
fd2ea0a7 3118 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3119
3120 netif_tx_start_queue(txq);
f25f4e44
PWJ
3121}
3122
bea3348e
SH
3123/**
3124 * netif_stop_subqueue - stop sending packets on subqueue
3125 * @dev: network device
3126 * @queue_index: sub queue index
3127 *
3128 * Stop individual transmit queue of a device with multiple transmit queues.
3129 */
f25f4e44
PWJ
3130static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3131{
fd2ea0a7 3132 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3133 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3134}
3135
bea3348e
SH
3136/**
3137 * netif_subqueue_stopped - test status of subqueue
3138 * @dev: network device
3139 * @queue_index: sub queue index
3140 *
3141 * Check individual transmit queue of a device with multiple transmit queues.
3142 */
4d29515f
DM
3143static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3144 u16 queue_index)
f25f4e44 3145{
fd2ea0a7 3146 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3147
3148 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3149}
3150
4d29515f
DM
3151static inline bool netif_subqueue_stopped(const struct net_device *dev,
3152 struct sk_buff *skb)
668f895a
PE
3153{
3154 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3155}
bea3348e 3156
738b35cc
FF
3157/**
3158 * netif_wake_subqueue - allow sending packets on subqueue
3159 * @dev: network device
3160 * @queue_index: sub queue index
3161 *
3162 * Resume individual transmit queue of a device with multiple transmit queues.
3163 */
3164static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3165{
3166 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3167
3168 netif_tx_wake_queue(txq);
3169}
f25f4e44 3170
537c00de 3171#ifdef CONFIG_XPS
53af53ae 3172int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3173 u16 index);
537c00de
AD
3174#else
3175static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3176 const struct cpumask *mask,
537c00de
AD
3177 u16 index)
3178{
3179 return 0;
3180}
3181#endif
3182
5605c762
JP
3183u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3184 unsigned int num_tx_queues);
3185
a3d22a68
VZ
3186/*
3187 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3188 * as a distribution range limit for the returned value.
3189 */
3190static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3191 struct sk_buff *skb)
a3d22a68
VZ
3192{
3193 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3194}
3195
bea3348e
SH
3196/**
3197 * netif_is_multiqueue - test if device has multiple transmit queues
3198 * @dev: network device
3199 *
3200 * Check if device has multiple transmit queues
bea3348e 3201 */
4d29515f 3202static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3203{
a02cec21 3204 return dev->num_tx_queues > 1;
f25f4e44 3205}
1da177e4 3206
f629d208 3207int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3208
a953be53 3209#ifdef CONFIG_SYSFS
f629d208 3210int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3211#else
3212static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3213 unsigned int rxq)
3214{
3215 return 0;
3216}
3217#endif
3218
a953be53
MD
3219#ifdef CONFIG_SYSFS
3220static inline unsigned int get_netdev_rx_queue_index(
3221 struct netdev_rx_queue *queue)
3222{
3223 struct net_device *dev = queue->dev;
3224 int index = queue - dev->_rx;
3225
3226 BUG_ON(index >= dev->num_rx_queues);
3227 return index;
3228}
3229#endif
3230
16917b87 3231#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3232int netif_get_num_default_rss_queues(void);
16917b87 3233
e6247027
ED
3234enum skb_free_reason {
3235 SKB_REASON_CONSUMED,
3236 SKB_REASON_DROPPED,
3237};
3238
3239void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3240void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3241
e6247027
ED
3242/*
3243 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3244 * interrupt context or with hardware interrupts being disabled.
3245 * (in_irq() || irqs_disabled())
3246 *
3247 * We provide four helpers that can be used in following contexts :
3248 *
3249 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3250 * replacing kfree_skb(skb)
3251 *
3252 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3253 * Typically used in place of consume_skb(skb) in TX completion path
3254 *
3255 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3256 * replacing kfree_skb(skb)
3257 *
3258 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3259 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3260 */
e6247027
ED
3261static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3262{
3263 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3264}
3265
3266static inline void dev_consume_skb_irq(struct sk_buff *skb)
3267{
3268 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3269}
3270
3271static inline void dev_kfree_skb_any(struct sk_buff *skb)
3272{
3273 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3274}
3275
3276static inline void dev_consume_skb_any(struct sk_buff *skb)
3277{
3278 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3279}
1da177e4 3280
7c497478
JW
3281void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3282int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
f629d208
JP
3283int netif_rx(struct sk_buff *skb);
3284int netif_rx_ni(struct sk_buff *skb);
04eb4489 3285int netif_receive_skb(struct sk_buff *skb);
1c601d82 3286int netif_receive_skb_core(struct sk_buff *skb);
f629d208
JP
3287gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3288void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3289struct sk_buff *napi_get_frags(struct napi_struct *napi);
3290gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3291struct packet_offload *gro_find_receive_by_type(__be16 type);
3292struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3293
3294static inline void napi_free_frags(struct napi_struct *napi)
3295{
3296 kfree_skb(napi->skb);
3297 napi->skb = NULL;
3298}
3299
24b27fc4 3300bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3301int netdev_rx_handler_register(struct net_device *dev,
3302 rx_handler_func_t *rx_handler,
3303 void *rx_handler_data);
3304void netdev_rx_handler_unregister(struct net_device *dev);
3305
3306bool dev_valid_name(const char *name);
3307int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3308int dev_ethtool(struct net *net, struct ifreq *);
3309unsigned int dev_get_flags(const struct net_device *);
3310int __dev_change_flags(struct net_device *, unsigned int flags);
3311int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3312void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3313 unsigned int gchanges);
f629d208
JP
3314int dev_change_name(struct net_device *, const char *);
3315int dev_set_alias(struct net_device *, const char *, size_t);
6c557001 3316int dev_get_alias(const struct net_device *, char *, size_t);
f629d208 3317int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3318int __dev_set_mtu(struct net_device *, int);
f629d208
JP
3319int dev_set_mtu(struct net_device *, int);
3320void dev_set_group(struct net_device *, int);
3321int dev_set_mac_address(struct net_device *, struct sockaddr *);
3322int dev_change_carrier(struct net_device *, bool new_carrier);
3323int dev_get_phys_port_id(struct net_device *dev,
02637fce 3324 struct netdev_phys_item_id *ppid);
db24a904
DA
3325int dev_get_phys_port_name(struct net_device *dev,
3326 char *name, size_t len);
d746d707 3327int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3328struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3329struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3330 struct netdev_queue *txq, int *ret);
d67b9cd2 3331
f4e63525 3332typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
d67b9cd2
DB
3333int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3334 int fd, u32 flags);
118b4aa2
JK
3335void __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3336 struct netdev_bpf *xdp);
d67b9cd2 3337
a0265d28 3338int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3339int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3340bool is_skb_forwardable(const struct net_device *dev,
3341 const struct sk_buff *skb);
1da177e4 3342
4e3264d2
MKL
3343static __always_inline int ____dev_forward_skb(struct net_device *dev,
3344 struct sk_buff *skb)
3345{
3346 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3347 unlikely(!is_skb_forwardable(dev, skb))) {
3348 atomic_long_inc(&dev->rx_dropped);
3349 kfree_skb(skb);
3350 return NET_RX_DROP;
3351 }
3352
3353 skb_scrub_packet(skb, true);
3354 skb->priority = 0;
3355 return 0;
3356}
3357
74b20582
DA
3358void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3359
20380731 3360extern int netdev_budget;
7acf8a1e 3361extern unsigned int netdev_budget_usecs;
1da177e4
LT
3362
3363/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3364void netdev_run_todo(void);
1da177e4 3365
bea3348e
SH
3366/**
3367 * dev_put - release reference to device
3368 * @dev: network device
3369 *
9ef4429b 3370 * Release reference to device to allow it to be freed.
bea3348e 3371 */
1da177e4
LT
3372static inline void dev_put(struct net_device *dev)
3373{
933393f5 3374 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3375}
3376
bea3348e
SH
3377/**
3378 * dev_hold - get reference to device
3379 * @dev: network device
3380 *
9ef4429b 3381 * Hold reference to device to keep it from being freed.
bea3348e 3382 */
15333061
SH
3383static inline void dev_hold(struct net_device *dev)
3384{
933393f5 3385 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3386}
1da177e4
LT
3387
3388/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3389 * and _off may be called from IRQ context, but it is caller
3390 * who is responsible for serialization of these calls.
b00055aa
SR
3391 *
3392 * The name carrier is inappropriate, these functions should really be
3393 * called netif_lowerlayer_*() because they represent the state of any
3394 * kind of lower layer not just hardware media.
1da177e4
LT
3395 */
3396
f629d208
JP
3397void linkwatch_init_dev(struct net_device *dev);
3398void linkwatch_fire_event(struct net_device *dev);
3399void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3400
bea3348e
SH
3401/**
3402 * netif_carrier_ok - test if carrier present
3403 * @dev: network device
3404 *
3405 * Check if carrier is present on device
3406 */
4d29515f 3407static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3408{
3409 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3410}
3411
f629d208 3412unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3413
f629d208 3414void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3415
f629d208 3416void netif_carrier_on(struct net_device *dev);
1da177e4 3417
f629d208 3418void netif_carrier_off(struct net_device *dev);
1da177e4 3419
bea3348e
SH
3420/**
3421 * netif_dormant_on - mark device as dormant.
3422 * @dev: network device
3423 *
3424 * Mark device as dormant (as per RFC2863).
3425 *
3426 * The dormant state indicates that the relevant interface is not
3427 * actually in a condition to pass packets (i.e., it is not 'up') but is
3428 * in a "pending" state, waiting for some external event. For "on-
3429 * demand" interfaces, this new state identifies the situation where the
3430 * interface is waiting for events to place it in the up state.
bea3348e 3431 */
b00055aa
SR
3432static inline void netif_dormant_on(struct net_device *dev)
3433{
3434 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3435 linkwatch_fire_event(dev);
3436}
3437
bea3348e
SH
3438/**
3439 * netif_dormant_off - set device as not dormant.
3440 * @dev: network device
3441 *
3442 * Device is not in dormant state.
3443 */
b00055aa
SR
3444static inline void netif_dormant_off(struct net_device *dev)
3445{
3446 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3447 linkwatch_fire_event(dev);
3448}
3449
bea3348e 3450/**
8ecbc40a 3451 * netif_dormant - test if device is dormant
bea3348e
SH
3452 * @dev: network device
3453 *
8ecbc40a 3454 * Check if device is dormant.
bea3348e 3455 */
4d29515f 3456static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3457{
3458 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3459}
3460
3461
bea3348e
SH
3462/**
3463 * netif_oper_up - test if device is operational
3464 * @dev: network device
3465 *
3466 * Check if carrier is operational
3467 */
4d29515f 3468static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3469{
b00055aa
SR
3470 return (dev->operstate == IF_OPER_UP ||
3471 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3472}
3473
bea3348e
SH
3474/**
3475 * netif_device_present - is device available or removed
3476 * @dev: network device
3477 *
3478 * Check if device has not been removed from system.
3479 */
4d29515f 3480static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3481{
3482 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3483}
3484
f629d208 3485void netif_device_detach(struct net_device *dev);
1da177e4 3486
f629d208 3487void netif_device_attach(struct net_device *dev);
1da177e4
LT
3488
3489/*
3490 * Network interface message level settings
3491 */
1da177e4
LT
3492
3493enum {
3494 NETIF_MSG_DRV = 0x0001,
3495 NETIF_MSG_PROBE = 0x0002,
3496 NETIF_MSG_LINK = 0x0004,
3497 NETIF_MSG_TIMER = 0x0008,
3498 NETIF_MSG_IFDOWN = 0x0010,
3499 NETIF_MSG_IFUP = 0x0020,
3500 NETIF_MSG_RX_ERR = 0x0040,
3501 NETIF_MSG_TX_ERR = 0x0080,
3502 NETIF_MSG_TX_QUEUED = 0x0100,
3503 NETIF_MSG_INTR = 0x0200,
3504 NETIF_MSG_TX_DONE = 0x0400,
3505 NETIF_MSG_RX_STATUS = 0x0800,
3506 NETIF_MSG_PKTDATA = 0x1000,
3507 NETIF_MSG_HW = 0x2000,
3508 NETIF_MSG_WOL = 0x4000,
3509};
3510
3511#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3512#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3513#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3514#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3515#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3516#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3517#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3518#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3519#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3520#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3521#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3522#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3523#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3524#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3525#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3526
3527static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3528{
3529 /* use default */
3530 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3531 return default_msg_enable_bits;
3532 if (debug_value == 0) /* no output */
3533 return 0;
3534 /* set low N bits */
3535 return (1 << debug_value) - 1;
3536}
3537
c773e847 3538static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3539{
c773e847
DM
3540 spin_lock(&txq->_xmit_lock);
3541 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3542}
3543
5a717f4f
MT
3544static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3545{
3546 __acquire(&txq->_xmit_lock);
3547 return true;
3548}
3549
3550static inline void __netif_tx_release(struct netdev_queue *txq)
3551{
3552 __release(&txq->_xmit_lock);
3553}
3554
fd2ea0a7
DM
3555static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3556{
3557 spin_lock_bh(&txq->_xmit_lock);
3558 txq->xmit_lock_owner = smp_processor_id();
3559}
3560
4d29515f 3561static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3562{
4d29515f 3563 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3564 if (likely(ok))
3565 txq->xmit_lock_owner = smp_processor_id();
3566 return ok;
3567}
3568
3569static inline void __netif_tx_unlock(struct netdev_queue *txq)
3570{
3571 txq->xmit_lock_owner = -1;
3572 spin_unlock(&txq->_xmit_lock);
3573}
3574
3575static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3576{
3577 txq->xmit_lock_owner = -1;
3578 spin_unlock_bh(&txq->_xmit_lock);
3579}
3580
08baf561
ED
3581static inline void txq_trans_update(struct netdev_queue *txq)
3582{
3583 if (txq->xmit_lock_owner != -1)
3584 txq->trans_start = jiffies;
3585}
3586
ba162f8e
FW
3587/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3588static inline void netif_trans_update(struct net_device *dev)
3589{
9b36627a
FW
3590 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3591
3592 if (txq->trans_start != jiffies)
3593 txq->trans_start = jiffies;
ba162f8e
FW
3594}
3595
d29f749e
DJ
3596/**
3597 * netif_tx_lock - grab network device transmit lock
3598 * @dev: network device
d29f749e
DJ
3599 *
3600 * Get network device transmit lock
3601 */
22dd7495
JHS
3602static inline void netif_tx_lock(struct net_device *dev)
3603{
e8a0464c 3604 unsigned int i;
c3f26a26 3605 int cpu;
c773e847 3606
c3f26a26
DM
3607 spin_lock(&dev->tx_global_lock);
3608 cpu = smp_processor_id();
e8a0464c
DM
3609 for (i = 0; i < dev->num_tx_queues; i++) {
3610 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3611
3612 /* We are the only thread of execution doing a
3613 * freeze, but we have to grab the _xmit_lock in
3614 * order to synchronize with threads which are in
3615 * the ->hard_start_xmit() handler and already
3616 * checked the frozen bit.
3617 */
e8a0464c 3618 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3619 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3620 __netif_tx_unlock(txq);
e8a0464c 3621 }
932ff279
HX
3622}
3623
3624static inline void netif_tx_lock_bh(struct net_device *dev)
3625{
e8a0464c
DM
3626 local_bh_disable();
3627 netif_tx_lock(dev);
932ff279
HX
3628}
3629
932ff279
HX
3630static inline void netif_tx_unlock(struct net_device *dev)
3631{
e8a0464c
DM
3632 unsigned int i;
3633
3634 for (i = 0; i < dev->num_tx_queues; i++) {
3635 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3636
c3f26a26
DM
3637 /* No need to grab the _xmit_lock here. If the
3638 * queue is not stopped for another reason, we
3639 * force a schedule.
3640 */
3641 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3642 netif_schedule_queue(txq);
c3f26a26
DM
3643 }
3644 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3645}
3646
3647static inline void netif_tx_unlock_bh(struct net_device *dev)
3648{
e8a0464c
DM
3649 netif_tx_unlock(dev);
3650 local_bh_enable();
932ff279
HX
3651}
3652
c773e847 3653#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3654 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3655 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3656 } else { \
3657 __netif_tx_acquire(txq); \
22dd7495
JHS
3658 } \
3659}
3660
5efeac44
EB
3661#define HARD_TX_TRYLOCK(dev, txq) \
3662 (((dev->features & NETIF_F_LLTX) == 0) ? \
3663 __netif_tx_trylock(txq) : \
5a717f4f 3664 __netif_tx_acquire(txq))
5efeac44 3665
c773e847 3666#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3667 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3668 __netif_tx_unlock(txq); \
5a717f4f
MT
3669 } else { \
3670 __netif_tx_release(txq); \
22dd7495
JHS
3671 } \
3672}
3673
1da177e4
LT
3674static inline void netif_tx_disable(struct net_device *dev)
3675{
fd2ea0a7 3676 unsigned int i;
c3f26a26 3677 int cpu;
fd2ea0a7 3678
c3f26a26
DM
3679 local_bh_disable();
3680 cpu = smp_processor_id();
fd2ea0a7
DM
3681 for (i = 0; i < dev->num_tx_queues; i++) {
3682 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3683
3684 __netif_tx_lock(txq, cpu);
fd2ea0a7 3685 netif_tx_stop_queue(txq);
c3f26a26 3686 __netif_tx_unlock(txq);
fd2ea0a7 3687 }
c3f26a26 3688 local_bh_enable();
1da177e4
LT
3689}
3690
e308a5d8
DM
3691static inline void netif_addr_lock(struct net_device *dev)
3692{
3693 spin_lock(&dev->addr_list_lock);
3694}
3695
2429f7ac
JP
3696static inline void netif_addr_lock_nested(struct net_device *dev)
3697{
25175ba5
VY
3698 int subclass = SINGLE_DEPTH_NESTING;
3699
3700 if (dev->netdev_ops->ndo_get_lock_subclass)
3701 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3702
3703 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3704}
3705
e308a5d8
DM
3706static inline void netif_addr_lock_bh(struct net_device *dev)
3707{
3708 spin_lock_bh(&dev->addr_list_lock);
3709}
3710
3711static inline void netif_addr_unlock(struct net_device *dev)
3712{
3713 spin_unlock(&dev->addr_list_lock);
3714}
3715
3716static inline void netif_addr_unlock_bh(struct net_device *dev)
3717{
3718 spin_unlock_bh(&dev->addr_list_lock);
3719}
3720
f001fde5 3721/*
31278e71 3722 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3723 * rcu_read_lock held.
3724 */
3725#define for_each_dev_addr(dev, ha) \
31278e71 3726 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3727
1da177e4
LT
3728/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3729
f629d208 3730void ether_setup(struct net_device *dev);
1da177e4
LT
3731
3732/* Support for loadable net-drivers */
f629d208 3733struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3734 unsigned char name_assign_type,
f629d208
JP
3735 void (*setup)(struct net_device *),
3736 unsigned int txqs, unsigned int rxqs);
0ad646c8
CW
3737int dev_get_valid_name(struct net *net, struct net_device *dev,
3738 const char *name);
3739
c835a677
TG
3740#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3741 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3742
c835a677
TG
3743#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3744 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3745 count)
36909ea4 3746
f629d208
JP
3747int register_netdev(struct net_device *dev);
3748void unregister_netdev(struct net_device *dev);
f001fde5 3749
22bedad3 3750/* General hardware address lists handling functions */
f629d208
JP
3751int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3752 struct netdev_hw_addr_list *from_list, int addr_len);
3753void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3754 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3755int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3756 struct net_device *dev,
3757 int (*sync)(struct net_device *, const unsigned char *),
3758 int (*unsync)(struct net_device *,
3759 const unsigned char *));
3760void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3761 struct net_device *dev,
3762 int (*unsync)(struct net_device *,
3763 const unsigned char *));
f629d208 3764void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3765
f001fde5 3766/* Functions used for device addresses handling */
f629d208
JP
3767int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3768 unsigned char addr_type);
3769int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3770 unsigned char addr_type);
f629d208
JP
3771void dev_addr_flush(struct net_device *dev);
3772int dev_addr_init(struct net_device *dev);
a748ee24
JP
3773
3774/* Functions used for unicast addresses handling */
f629d208
JP
3775int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3776int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3777int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3778int dev_uc_sync(struct net_device *to, struct net_device *from);
3779int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3780void dev_uc_unsync(struct net_device *to, struct net_device *from);
3781void dev_uc_flush(struct net_device *dev);
3782void dev_uc_init(struct net_device *dev);
f001fde5 3783
670e5b8e
AD
3784/**
3785 * __dev_uc_sync - Synchonize device's unicast list
3786 * @dev: device to sync
3787 * @sync: function to call if address should be added
3788 * @unsync: function to call if address should be removed
3789 *
3790 * Add newly added addresses to the interface, and release
3791 * addresses that have been deleted.
5e82b4b2 3792 */
670e5b8e
AD
3793static inline int __dev_uc_sync(struct net_device *dev,
3794 int (*sync)(struct net_device *,
3795 const unsigned char *),
3796 int (*unsync)(struct net_device *,
3797 const unsigned char *))
3798{
3799 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3800}
3801
3802/**
e793c0f7 3803 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3804 * @dev: device to sync
3805 * @unsync: function to call if address should be removed
3806 *
3807 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3808 */
670e5b8e
AD
3809static inline void __dev_uc_unsync(struct net_device *dev,
3810 int (*unsync)(struct net_device *,
3811 const unsigned char *))
3812{
3813 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3814}
3815
22bedad3 3816/* Functions used for multicast addresses handling */
f629d208
JP
3817int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3818int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3819int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3820int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3821int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3822int dev_mc_sync(struct net_device *to, struct net_device *from);
3823int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3824void dev_mc_unsync(struct net_device *to, struct net_device *from);
3825void dev_mc_flush(struct net_device *dev);
3826void dev_mc_init(struct net_device *dev);
f001fde5 3827
670e5b8e
AD
3828/**
3829 * __dev_mc_sync - Synchonize device's multicast list
3830 * @dev: device to sync
3831 * @sync: function to call if address should be added
3832 * @unsync: function to call if address should be removed
3833 *
3834 * Add newly added addresses to the interface, and release
3835 * addresses that have been deleted.
5e82b4b2 3836 */
670e5b8e
AD
3837static inline int __dev_mc_sync(struct net_device *dev,
3838 int (*sync)(struct net_device *,
3839 const unsigned char *),
3840 int (*unsync)(struct net_device *,
3841 const unsigned char *))
3842{
3843 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3844}
3845
3846/**
e793c0f7 3847 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3848 * @dev: device to sync
3849 * @unsync: function to call if address should be removed
3850 *
3851 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3852 */
670e5b8e
AD
3853static inline void __dev_mc_unsync(struct net_device *dev,
3854 int (*unsync)(struct net_device *,
3855 const unsigned char *))
3856{
3857 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3858}
3859
4417da66 3860/* Functions used for secondary unicast and multicast support */
f629d208
JP
3861void dev_set_rx_mode(struct net_device *dev);
3862void __dev_set_rx_mode(struct net_device *dev);
3863int dev_set_promiscuity(struct net_device *dev, int inc);
3864int dev_set_allmulti(struct net_device *dev, int inc);
3865void netdev_state_change(struct net_device *dev);
3866void netdev_notify_peers(struct net_device *dev);
3867void netdev_features_change(struct net_device *dev);
1da177e4 3868/* Load a device via the kmod */
f629d208
JP
3869void dev_load(struct net *net, const char *name);
3870struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3871 struct rtnl_link_stats64 *storage);
3872void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3873 const struct net_device_stats *netdev_stats);
eeda3fd6 3874
1da177e4 3875extern int netdev_max_backlog;
3b098e2d 3876extern int netdev_tstamp_prequeue;
1da177e4 3877extern int weight_p;
3d48b53f
MT
3878extern int dev_weight_rx_bias;
3879extern int dev_weight_tx_bias;
3880extern int dev_rx_weight;
3881extern int dev_tx_weight;
9ff162a8 3882
f629d208 3883bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3884struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3885 struct list_head **iter);
f629d208
JP
3886struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3887 struct list_head **iter);
8b5be856 3888
44a40855
VY
3889/* iterate through upper list, must be called under RCU read lock */
3890#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3891 for (iter = &(dev)->adj_list.upper, \
3892 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3893 updev; \
3894 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3895
1a3f060c
DA
3896int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3897 int (*fn)(struct net_device *upper_dev,
3898 void *data),
3899 void *data);
3900
3901bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3902 struct net_device *upper_dev);
3903
25cc72a3
IS
3904bool netdev_has_any_upper_dev(struct net_device *dev);
3905
f629d208
JP
3906void *netdev_lower_get_next_private(struct net_device *dev,
3907 struct list_head **iter);
3908void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3909 struct list_head **iter);
31088a11
VF
3910
3911#define netdev_for_each_lower_private(dev, priv, iter) \
3912 for (iter = (dev)->adj_list.lower.next, \
3913 priv = netdev_lower_get_next_private(dev, &(iter)); \
3914 priv; \
3915 priv = netdev_lower_get_next_private(dev, &(iter)))
3916
3917#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3918 for (iter = &(dev)->adj_list.lower, \
3919 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3920 priv; \
3921 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3922
4085ebe8
VY
3923void *netdev_lower_get_next(struct net_device *dev,
3924 struct list_head **iter);
7ce856aa 3925
4085ebe8 3926#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3927 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3928 ldev = netdev_lower_get_next(dev, &(iter)); \
3929 ldev; \
3930 ldev = netdev_lower_get_next(dev, &(iter)))
3931
7ce856aa
JP
3932struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3933 struct list_head **iter);
3934struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3935 struct list_head **iter);
3936
1a3f060c
DA
3937int netdev_walk_all_lower_dev(struct net_device *dev,
3938 int (*fn)(struct net_device *lower_dev,
3939 void *data),
3940 void *data);
3941int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3942 int (*fn)(struct net_device *lower_dev,
3943 void *data),
3944 void *data);
3945
f629d208 3946void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3947void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3948struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3949struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
42ab19ee
DA
3950int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
3951 struct netlink_ext_ack *extack);
f629d208 3952int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3953 struct net_device *upper_dev,
42ab19ee
DA
3954 void *upper_priv, void *upper_info,
3955 struct netlink_ext_ack *extack);
f629d208
JP
3956void netdev_upper_dev_unlink(struct net_device *dev,
3957 struct net_device *upper_dev);
5bb025fa 3958void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3959void *netdev_lower_dev_get_private(struct net_device *dev,
3960 struct net_device *lower_dev);
04d48266
JP
3961void netdev_lower_state_changed(struct net_device *lower_dev,
3962 void *lower_state_info);
960fb622
ED
3963
3964/* RSS keys are 40 or 52 bytes long */
3965#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3966extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3967void netdev_rss_key_fill(void *buffer, size_t len);
3968
952fcfd0 3969int dev_get_nest_level(struct net_device *dev);
f629d208 3970int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 3971int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
3972int skb_csum_hwoffload_help(struct sk_buff *skb,
3973 const netdev_features_t features);
3974
f629d208
JP
3975struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3976 netdev_features_t features, bool tx_path);
3977struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3978 netdev_features_t features);
12b0004d 3979
61bd3857
MS
3980struct netdev_bonding_info {
3981 ifslave slave;
3982 ifbond master;
3983};
3984
3985struct netdev_notifier_bonding_info {
3986 struct netdev_notifier_info info; /* must be first */
3987 struct netdev_bonding_info bonding_info;
3988};
3989
3990void netdev_bonding_info_change(struct net_device *dev,
3991 struct netdev_bonding_info *bonding_info);
3992
12b0004d
CW
3993static inline
3994struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3995{
3996 return __skb_gso_segment(skb, features, true);
3997}
53d6471c 3998__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3999
4000static inline bool can_checksum_protocol(netdev_features_t features,
4001 __be16 protocol)
4002{
c8cd0989
TH
4003 if (protocol == htons(ETH_P_FCOE))
4004 return !!(features & NETIF_F_FCOE_CRC);
4005
4006 /* Assume this is an IP checksum (not SCTP CRC) */
4007
4008 if (features & NETIF_F_HW_CSUM) {
4009 /* Can checksum everything */
4010 return true;
4011 }
4012
4013 switch (protocol) {
4014 case htons(ETH_P_IP):
4015 return !!(features & NETIF_F_IP_CSUM);
4016 case htons(ETH_P_IPV6):
4017 return !!(features & NETIF_F_IPV6_CSUM);
4018 default:
4019 return false;
4020 }
ec5f0615 4021}
12b0004d 4022
fb286bb2 4023#ifdef CONFIG_BUG
f629d208 4024void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
4025#else
4026static inline void netdev_rx_csum_fault(struct net_device *dev)
4027{
4028}
4029#endif
1da177e4 4030/* rx skb timestamps */
f629d208
JP
4031void net_enable_timestamp(void);
4032void net_disable_timestamp(void);
1da177e4 4033
20380731 4034#ifdef CONFIG_PROC_FS
f629d208 4035int __init dev_proc_init(void);
900ff8c6
CW
4036#else
4037#define dev_proc_init() 0
20380731
ACM
4038#endif
4039
4798248e 4040static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4041 struct sk_buff *skb, struct net_device *dev,
4042 bool more)
4798248e 4043{
fa2dbdc2 4044 skb->xmit_more = more ? 1 : 0;
0b725a2c 4045 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4046}
4047
10b3ad8c 4048static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4049 struct netdev_queue *txq, bool more)
4798248e
DM
4050{
4051 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4052 int rc;
4798248e 4053
fa2dbdc2 4054 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4055 if (rc == NETDEV_TX_OK)
4056 txq_trans_update(txq);
4057
4058 return rc;
4798248e
DM
4059}
4060
b793dc5c 4061int netdev_class_create_file_ns(const struct class_attribute *class_attr,
42a2d923 4062 const void *ns);
b793dc5c 4063void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
42a2d923 4064 const void *ns);
58292cbe 4065
b793dc5c 4066static inline int netdev_class_create_file(const struct class_attribute *class_attr)
58292cbe
TH
4067{
4068 return netdev_class_create_file_ns(class_attr, NULL);
4069}
4070
b793dc5c 4071static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
58292cbe
TH
4072{
4073 netdev_class_remove_file_ns(class_attr, NULL);
4074}
b8a9787e 4075
737aec57 4076extern const struct kobj_ns_type_operations net_ns_type_operations;
04600794 4077
f629d208 4078const char *netdev_drivername(const struct net_device *dev);
6579e57b 4079
f629d208 4080void linkwatch_run_queue(void);
20380731 4081
da08143b
MK
4082static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4083 netdev_features_t f2)
4084{
c8cd0989
TH
4085 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4086 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4087 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4088 else
b6a0e72a 4089 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4090 }
da08143b 4091
c8cd0989 4092 return f1 & f2;
da08143b
MK
4093}
4094
c8f44aff
MM
4095static inline netdev_features_t netdev_get_wanted_features(
4096 struct net_device *dev)
5455c699
MM
4097{
4098 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4099}
c8f44aff
MM
4100netdev_features_t netdev_increment_features(netdev_features_t all,
4101 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4102
4103/* Allow TSO being used on stacked device :
4104 * Performing the GSO segmentation before last device
4105 * is a performance improvement.
4106 */
4107static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4108 netdev_features_t mask)
4109{
4110 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4111}
4112
6cb6a27c 4113int __netdev_update_features(struct net_device *dev);
5455c699 4114void netdev_update_features(struct net_device *dev);
afe12cc8 4115void netdev_change_features(struct net_device *dev);
7f353bf2 4116
fc4a7489
PM
4117void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4118 struct net_device *dev);
4119
e38f3025
TM
4120netdev_features_t passthru_features_check(struct sk_buff *skb,
4121 struct net_device *dev,
4122 netdev_features_t features);
c1e756bf 4123netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4124
4d29515f 4125static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4126{
7b748340 4127 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4128
4129 /* check flags correspondence */
4130 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4131 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4132 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4133 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4134 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4135 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4136 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4137 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4138 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4139 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4140 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4141 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4142 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4143 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4144 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4145 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0c19f846 4146 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
0345e186 4147
d6b4991a 4148 return (features & feature) == feature;
576a30eb
HX
4149}
4150
4d29515f 4151static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4152{
278b2513 4153 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4154 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4155}
4156
8b86a61d 4157static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4158 netdev_features_t features)
7967168c 4159{
fc741216 4160 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4161 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4162 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4163}
4164
82cc1a7a
PWJ
4165static inline void netif_set_gso_max_size(struct net_device *dev,
4166 unsigned int size)
4167{
4168 dev->gso_max_size = size;
4169}
4170
7a7ffbab
WCC
4171static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4172 int pulled_hlen, u16 mac_offset,
4173 int mac_len)
4174{
4175 skb->protocol = protocol;
4176 skb->encapsulation = 1;
4177 skb_push(skb, pulled_hlen);
4178 skb_reset_transport_header(skb);
4179 skb->mac_header = mac_offset;
4180 skb->network_header = skb->mac_header + mac_len;
4181 skb->mac_len = mac_len;
4182}
4183
3c175784
SD
4184static inline bool netif_is_macsec(const struct net_device *dev)
4185{
4186 return dev->priv_flags & IFF_MACSEC;
4187}
4188
b618aaa9 4189static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4190{
4191 return dev->priv_flags & IFF_MACVLAN;
4192}
4193
b618aaa9 4194static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4195{
4196 return dev->priv_flags & IFF_MACVLAN_PORT;
4197}
4198
b618aaa9 4199static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4200{
4201 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4202}
4203
b618aaa9 4204static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4205{
4206 return dev->priv_flags & IFF_IPVLAN_MASTER;
4207}
4208
b618aaa9 4209static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4210{
4211 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4212}
4213
b618aaa9 4214static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4215{
4216 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4217}
4218
3bdc0eba
BG
4219static inline bool netif_supports_nofcs(struct net_device *dev)
4220{
4221 return dev->priv_flags & IFF_SUPP_NOFCS;
4222}
4223
007979ea 4224static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4225{
007979ea 4226 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4227}
4228
fee6d4c7
DA
4229static inline bool netif_is_l3_slave(const struct net_device *dev)
4230{
4231 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4232}
4233
0894ae3f
JP
4234static inline bool netif_is_bridge_master(const struct net_device *dev)
4235{
4236 return dev->priv_flags & IFF_EBRIDGE;
4237}
4238
28f9ee22
VY
4239static inline bool netif_is_bridge_port(const struct net_device *dev)
4240{
4241 return dev->priv_flags & IFF_BRIDGE_PORT;
4242}
4243
35d4e172
JP
4244static inline bool netif_is_ovs_master(const struct net_device *dev)
4245{
4246 return dev->priv_flags & IFF_OPENVSWITCH;
4247}
4248
5be66141
JP
4249static inline bool netif_is_ovs_port(const struct net_device *dev)
4250{
4251 return dev->priv_flags & IFF_OVS_DATAPATH;
4252}
4253
b618aaa9 4254static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4255{
4256 return dev->priv_flags & IFF_TEAM;
4257}
4258
b618aaa9 4259static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4260{
4261 return dev->priv_flags & IFF_TEAM_PORT;
4262}
4263
b618aaa9 4264static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4265{
4266 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4267}
4268
b618aaa9 4269static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4270{
4271 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4272}
4273
d4ab4286
KJ
4274static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4275{
4276 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4277}
4278
02875878
ED
4279/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4280static inline void netif_keep_dst(struct net_device *dev)
4281{
4282 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4283}
4284
18d3df3e
PA
4285/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4286static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4287{
4288 /* TODO: reserve and use an additional IFF bit, if we get more users */
4289 return dev->priv_flags & IFF_MACSEC;
4290}
4291
505d4f73 4292extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4293
571ba423
JP
4294/* Logging, debugging and troubleshooting/diagnostic helpers. */
4295
4296/* netdev_printk helpers, similar to dev_printk */
4297
4298static inline const char *netdev_name(const struct net_device *dev)
4299{
c6f854d5
VF
4300 if (!dev->name[0] || strchr(dev->name, '%'))
4301 return "(unnamed net_device)";
571ba423
JP
4302 return dev->name;
4303}
4304
8397ed36
DA
4305static inline bool netdev_unregistering(const struct net_device *dev)
4306{
4307 return dev->reg_state == NETREG_UNREGISTERING;
4308}
4309
ccc7f496
VF
4310static inline const char *netdev_reg_state(const struct net_device *dev)
4311{
4312 switch (dev->reg_state) {
4313 case NETREG_UNINITIALIZED: return " (uninitialized)";
4314 case NETREG_REGISTERED: return "";
4315 case NETREG_UNREGISTERING: return " (unregistering)";
4316 case NETREG_UNREGISTERED: return " (unregistered)";
4317 case NETREG_RELEASED: return " (released)";
4318 case NETREG_DUMMY: return " (dummy)";
4319 }
4320
4321 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4322 return " (unknown)";
4323}
4324
f629d208 4325__printf(3, 4)
6ea754eb
JP
4326void netdev_printk(const char *level, const struct net_device *dev,
4327 const char *format, ...);
f629d208 4328__printf(2, 3)
6ea754eb 4329void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4330__printf(2, 3)
6ea754eb 4331void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4332__printf(2, 3)
6ea754eb 4333void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4334__printf(2, 3)
6ea754eb 4335void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4336__printf(2, 3)
6ea754eb 4337void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4338__printf(2, 3)
6ea754eb 4339void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4340__printf(2, 3)
6ea754eb 4341void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4342
375ef2b1
GP
4343#define netdev_level_once(level, dev, fmt, ...) \
4344do { \
4345 static bool __print_once __read_mostly; \
4346 \
4347 if (!__print_once) { \
4348 __print_once = true; \
4349 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4350 } \
4351} while (0)
4352
4353#define netdev_emerg_once(dev, fmt, ...) \
4354 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4355#define netdev_alert_once(dev, fmt, ...) \
4356 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4357#define netdev_crit_once(dev, fmt, ...) \
4358 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4359#define netdev_err_once(dev, fmt, ...) \
4360 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4361#define netdev_warn_once(dev, fmt, ...) \
4362 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4363#define netdev_notice_once(dev, fmt, ...) \
4364 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4365#define netdev_info_once(dev, fmt, ...) \
4366 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4367
8909c9ad
VK
4368#define MODULE_ALIAS_NETDEV(device) \
4369 MODULE_ALIAS("netdev-" device)
4370
b558c96f 4371#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4372#define netdev_dbg(__dev, format, args...) \
4373do { \
ffa10cb4 4374 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4375} while (0)
b558c96f
JC
4376#elif defined(DEBUG)
4377#define netdev_dbg(__dev, format, args...) \
4378 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4379#else
4380#define netdev_dbg(__dev, format, args...) \
4381({ \
4382 if (0) \
4383 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4384})
4385#endif
4386
4387#if defined(VERBOSE_DEBUG)
4388#define netdev_vdbg netdev_dbg
4389#else
4390
4391#define netdev_vdbg(dev, format, args...) \
4392({ \
4393 if (0) \
4394 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4395 0; \
4396})
4397#endif
4398
4399/*
4400 * netdev_WARN() acts like dev_printk(), but with the key difference
4401 * of using a WARN/WARN_ON to get the message out, including the
4402 * file/line information and a backtrace.
4403 */
4404#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4405 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4406 netdev_reg_state(dev), ##args)
571ba423 4407
375ef2b1
GP
4408#define netdev_WARN_ONCE(dev, condition, format, arg...) \
4409 WARN_ONCE(1, "netdevice: %s%s\n" format, netdev_name(dev) \
4410 netdev_reg_state(dev), ##args)
4411
b3d95c5c
JP
4412/* netif printk helpers, similar to netdev_printk */
4413
4414#define netif_printk(priv, type, level, dev, fmt, args...) \
4415do { \
4416 if (netif_msg_##type(priv)) \
4417 netdev_printk(level, (dev), fmt, ##args); \
4418} while (0)
4419
f45f4321
JP
4420#define netif_level(level, priv, type, dev, fmt, args...) \
4421do { \
4422 if (netif_msg_##type(priv)) \
4423 netdev_##level(dev, fmt, ##args); \
4424} while (0)
4425
b3d95c5c 4426#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4427 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4428#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4429 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4430#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4431 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4432#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4433 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4434#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4435 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4436#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4437 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4438#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4439 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4440
0053ea9c 4441#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4442#define netif_dbg(priv, type, netdev, format, args...) \
4443do { \
4444 if (netif_msg_##type(priv)) \
b5fb0a03 4445 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4446} while (0)
0053ea9c
JP
4447#elif defined(DEBUG)
4448#define netif_dbg(priv, type, dev, format, args...) \
4449 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4450#else
4451#define netif_dbg(priv, type, dev, format, args...) \
4452({ \
4453 if (0) \
4454 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4455 0; \
4456})
4457#endif
4458
f617f276
EC
4459/* if @cond then downgrade to debug, else print at @level */
4460#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4461 do { \
4462 if (cond) \
4463 netif_dbg(priv, type, netdev, fmt, ##args); \
4464 else \
4465 netif_ ## level(priv, type, netdev, fmt, ##args); \
4466 } while (0)
4467
b3d95c5c 4468#if defined(VERBOSE_DEBUG)
bcfcc450 4469#define netif_vdbg netif_dbg
b3d95c5c
JP
4470#else
4471#define netif_vdbg(priv, type, dev, format, args...) \
4472({ \
4473 if (0) \
a4ed89cb 4474 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4475 0; \
4476})
4477#endif
571ba423 4478
900ff8c6
CW
4479/*
4480 * The list of packet types we will receive (as opposed to discard)
4481 * and the routines to invoke.
4482 *
4483 * Why 16. Because with 16 the only overlap we get on a hash of the
4484 * low nibble of the protocol value is RARP/SNAP/X.25.
4485 *
900ff8c6 4486 * 0800 IP
900ff8c6
CW
4487 * 0001 802.3
4488 * 0002 AX.25
4489 * 0004 802.2
4490 * 8035 RARP
4491 * 0005 SNAP
4492 * 0805 X.25
4493 * 0806 ARP
4494 * 8137 IPX
4495 * 0009 Localtalk
4496 * 86DD IPv6
4497 */
4498#define PTYPE_HASH_SIZE (16)
4499#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4500
385a154c 4501#endif /* _LINUX_NETDEVICE_H */