]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - include/linux/netdevice.h
cxgb4: increase max tx rate limit to 100 Gbps
[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>
e817f856 47#include <net/xdp.h>
a050c33f 48
a59e2ecb 49#include <linux/netdev_features.h>
77162022 50#include <linux/neighbour.h>
607ca46e 51#include <uapi/linux/netdevice.h>
61bd3857 52#include <uapi/linux/if_bonding.h>
e4c6734e 53#include <uapi/linux/pkt_cls.h>
59cc1f61 54#include <linux/hashtable.h>
a59e2ecb 55
115c1d6e 56struct netpoll_info;
313162d0 57struct device;
c1f19b51 58struct phy_device;
2f657a60 59struct dsa_port;
c6e970a0 60
704232c2
JB
61/* 802.11 specific */
62struct wireless_dev;
98a18b6f
AA
63/* 802.15.4 specific */
64struct wpan_dev;
03c57747 65struct mpls_dev;
7c46a640
AD
66/* UDP Tunnel offloads */
67struct udp_tunnel_info;
a7862b45 68struct bpf_prog;
814abfab 69struct xdp_buff;
1da177e4 70
f629d208
JP
71void netdev_set_default_ethtool_ops(struct net_device *dev,
72 const struct ethtool_ops *ops);
d07d7507 73
9a1654ba
JP
74/* Backlog congestion levels */
75#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
76#define NET_RX_DROP 1 /* packet dropped */
77
572a9d7b
PM
78/*
79 * Transmit return codes: transmit return codes originate from three different
80 * namespaces:
81 *
82 * - qdisc return codes
83 * - driver transmit return codes
84 * - errno values
85 *
86 * Drivers are allowed to return any one of those in their hard_start_xmit()
87 * function. Real network devices commonly used with qdiscs should only return
88 * the driver transmit return codes though - when qdiscs are used, the actual
89 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
90 * higher layers. Virtual network devices transmit synchronously; in this case
91 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
92 * others are propagated to higher layers.
93 */
94
95/* qdisc ->enqueue() return codes. */
96#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
97#define NET_XMIT_DROP 0x01 /* skb dropped */
98#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 99#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 100
b9df3cb8
GR
101/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 104#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
105#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
dc1f8bf6 107/* Driver transmit return codes */
9a1654ba 108#define NETDEV_TX_MASK 0xf0
572a9d7b 109
dc1f8bf6 110enum netdev_tx {
572a9d7b 111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
114};
115typedef enum netdev_tx netdev_tx_t;
116
9a1654ba
JP
117/*
118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 */
121static inline bool dev_xmit_complete(int rc)
122{
123 /*
124 * Positive cases with an skb consumed by a driver:
125 * - successful transmission (rc == NETDEV_TX_OK)
126 * - error while transmitting (rc < 0)
127 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 */
129 if (likely(rc < NET_XMIT_MASK))
130 return true;
131
132 return false;
133}
134
1da177e4 135/*
5e82b4b2 136 * Compute the worst-case header length according to the protocols
1da177e4
LT
137 * used.
138 */
fe2918b0 139
c0eb4540
KS
140#if defined(CONFIG_HYPERV_NET)
141# define LL_MAX_HEADER 128
142#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
143# if defined(CONFIG_MAC80211_MESH)
144# define LL_MAX_HEADER 128
145# else
146# define LL_MAX_HEADER 96
147# endif
1da177e4 148#else
8388e3da 149# define LL_MAX_HEADER 32
1da177e4
LT
150#endif
151
d11ead75
BH
152#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
153 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
154#define MAX_HEADER LL_MAX_HEADER
155#else
156#define MAX_HEADER (LL_MAX_HEADER + 48)
157#endif
158
159/*
be1f3c2c
BH
160 * Old network device statistics. Fields are native words
161 * (unsigned long) so they can be read and written atomically.
1da177e4 162 */
fe2918b0 163
d94d9fee 164struct net_device_stats {
3cfde79c
BH
165 unsigned long rx_packets;
166 unsigned long tx_packets;
167 unsigned long rx_bytes;
168 unsigned long tx_bytes;
169 unsigned long rx_errors;
170 unsigned long tx_errors;
171 unsigned long rx_dropped;
172 unsigned long tx_dropped;
173 unsigned long multicast;
1da177e4 174 unsigned long collisions;
1da177e4 175 unsigned long rx_length_errors;
3cfde79c
BH
176 unsigned long rx_over_errors;
177 unsigned long rx_crc_errors;
178 unsigned long rx_frame_errors;
179 unsigned long rx_fifo_errors;
180 unsigned long rx_missed_errors;
1da177e4
LT
181 unsigned long tx_aborted_errors;
182 unsigned long tx_carrier_errors;
183 unsigned long tx_fifo_errors;
184 unsigned long tx_heartbeat_errors;
185 unsigned long tx_window_errors;
1da177e4
LT
186 unsigned long rx_compressed;
187 unsigned long tx_compressed;
188};
189
1da177e4
LT
190
191#include <linux/cache.h>
192#include <linux/skbuff.h>
193
adc9300e 194#ifdef CONFIG_RPS
c5905afb
IM
195#include <linux/static_key.h>
196extern struct static_key rps_needed;
13bfff25 197extern struct static_key rfs_needed;
adc9300e
ED
198#endif
199
1da177e4
LT
200struct neighbour;
201struct neigh_parms;
202struct sk_buff;
203
f001fde5
JP
204struct netdev_hw_addr {
205 struct list_head list;
206 unsigned char addr[MAX_ADDR_LEN];
207 unsigned char type;
ccffad25
JP
208#define NETDEV_HW_ADDR_T_LAN 1
209#define NETDEV_HW_ADDR_T_SAN 2
210#define NETDEV_HW_ADDR_T_SLAVE 3
211#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 212#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 213 bool global_use;
4cd729b0 214 int sync_cnt;
8f8f103d 215 int refcount;
4543fbef 216 int synced;
f001fde5
JP
217 struct rcu_head rcu_head;
218};
219
31278e71
JP
220struct netdev_hw_addr_list {
221 struct list_head list;
222 int count;
223};
224
22bedad3
JP
225#define netdev_hw_addr_list_count(l) ((l)->count)
226#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
227#define netdev_hw_addr_list_for_each(ha, l) \
228 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 229
22bedad3
JP
230#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
231#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
232#define netdev_for_each_uc_addr(ha, dev) \
233 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 234
22bedad3
JP
235#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
236#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 237#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 238 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 239
d94d9fee 240struct hh_cache {
5b3dc2f3 241 unsigned int hh_len;
3644f0ce 242 seqlock_t hh_lock;
1da177e4
LT
243
244 /* cached hardware header; allow for machine alignment needs. */
245#define HH_DATA_MOD 16
246#define HH_DATA_OFF(__len) \
5ba0eac6 247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
248#define HH_DATA_ALIGN(__len) \
249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251};
252
5e82b4b2 253/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
254 * Alternative is:
255 * dev->hard_header_len ? (dev->hard_header_len +
256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 *
258 * We could use other alignment values, but we must maintain the
259 * relationship HH alignment <= LL alignment.
260 */
261#define LL_RESERVED_SPACE(dev) \
f5184d26 262 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 263#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 264 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 265
3b04ddde
SH
266struct header_ops {
267 int (*create) (struct sk_buff *skb, struct net_device *dev,
268 unsigned short type, const void *daddr,
95c96174 269 const void *saddr, unsigned int len);
3b04ddde 270 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
2793a23a 275 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
276};
277
1da177e4 278/* These flag bits are private to the generic network queueing
5e82b4b2 279 * layer; they may not be explicitly referenced by any other
1da177e4
LT
280 * code.
281 */
282
d94d9fee 283enum netdev_state_t {
1da177e4
LT
284 __LINK_STATE_START,
285 __LINK_STATE_PRESENT,
1da177e4 286 __LINK_STATE_NOCARRIER,
b00055aa
SR
287 __LINK_STATE_LINKWATCH_PENDING,
288 __LINK_STATE_DORMANT,
1da177e4
LT
289};
290
291
292/*
5e82b4b2 293 * This structure holds boot-time configured netdevice settings. They
fe2918b0 294 * are then used in the device probing.
1da177e4
LT
295 */
296struct netdev_boot_setup {
297 char name[IFNAMSIZ];
298 struct ifmap map;
299};
300#define NETDEV_BOOT_SETUP_MAX 8
301
f629d208 302int __init netdev_boot_setup(char *str);
1da177e4 303
bea3348e
SH
304/*
305 * Structure for NAPI scheduling similar to tasklet but with weighting
306 */
307struct napi_struct {
308 /* The poll_list must only be managed by the entity which
309 * changes the state of the NAPI_STATE_SCHED bit. This means
310 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 311 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
312 * can remove from the list right before clearing the bit.
313 */
314 struct list_head poll_list;
315
316 unsigned long state;
317 int weight;
404f7c9e 318 unsigned int gro_count;
bea3348e
SH
319 int (*poll)(struct napi_struct *, int);
320#ifdef CONFIG_NETPOLL
bea3348e 321 int poll_owner;
bea3348e 322#endif
5d38a079 323 struct net_device *dev;
d565b0a1 324 struct sk_buff *gro_list;
5d38a079 325 struct sk_buff *skb;
3b47d303 326 struct hrtimer timer;
404f7c9e 327 struct list_head dev_list;
af12fa6e
ET
328 struct hlist_node napi_hash_node;
329 unsigned int napi_id;
bea3348e
SH
330};
331
d94d9fee 332enum {
bea3348e 333 NAPI_STATE_SCHED, /* Poll is scheduled */
39e6c820 334 NAPI_STATE_MISSED, /* reschedule a napi */
a0a46196 335 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 336 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
337 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
338 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
339 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
340};
341
342enum {
39e6c820
ED
343 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
344 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
345 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
346 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
347 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
348 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
349 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
350};
351
5b252f0c 352enum gro_result {
d1c76af9
HX
353 GRO_MERGED,
354 GRO_MERGED_FREE,
355 GRO_HELD,
356 GRO_NORMAL,
357 GRO_DROP,
25393d3f 358 GRO_CONSUMED,
d1c76af9 359};
5b252f0c 360typedef enum gro_result gro_result_t;
d1c76af9 361
8a4eb573
JP
362/*
363 * enum rx_handler_result - Possible return values for rx_handlers.
364 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365 * further.
366 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
367 * case skb->dev was changed by rx_handler.
368 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 369 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
370 *
371 * rx_handlers are functions called from inside __netif_receive_skb(), to do
372 * special processing of the skb, prior to delivery to protocol handlers.
373 *
374 * Currently, a net_device can only have a single rx_handler registered. Trying
375 * to register a second rx_handler will return -EBUSY.
376 *
377 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
378 * To unregister a rx_handler on a net_device, use
379 * netdev_rx_handler_unregister().
380 *
381 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
382 * do with the skb.
383 *
5e82b4b2 384 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 385 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 386 * the skb to be delivered in some other way.
8a4eb573
JP
387 *
388 * If the rx_handler changed skb->dev, to divert the skb to another
389 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
390 * new device will be called if it exists.
391 *
5e82b4b2 392 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 393 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 394 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 395 *
5e82b4b2 396 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
397 * delivered, it should return RX_HANDLER_PASS.
398 *
399 * A device without a registered rx_handler will behave as if rx_handler
400 * returned RX_HANDLER_PASS.
401 */
402
403enum rx_handler_result {
404 RX_HANDLER_CONSUMED,
405 RX_HANDLER_ANOTHER,
406 RX_HANDLER_EXACT,
407 RX_HANDLER_PASS,
408};
409typedef enum rx_handler_result rx_handler_result_t;
410typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 411
f629d208 412void __napi_schedule(struct napi_struct *n);
bc9ad166 413void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 414
4d29515f 415static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
416{
417 return test_bit(NAPI_STATE_DISABLE, &n->state);
418}
419
39e6c820 420bool napi_schedule_prep(struct napi_struct *n);
bea3348e
SH
421
422/**
423 * napi_schedule - schedule NAPI poll
5e82b4b2 424 * @n: NAPI context
bea3348e
SH
425 *
426 * Schedule NAPI poll routine to be called if it is not already
427 * running.
428 */
429static inline void napi_schedule(struct napi_struct *n)
430{
431 if (napi_schedule_prep(n))
432 __napi_schedule(n);
433}
434
bc9ad166
ED
435/**
436 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 437 * @n: NAPI context
bc9ad166
ED
438 *
439 * Variant of napi_schedule(), assuming hard irqs are masked.
440 */
441static inline void napi_schedule_irqoff(struct napi_struct *n)
442{
443 if (napi_schedule_prep(n))
444 __napi_schedule_irqoff(n);
445}
446
bfe13f54 447/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 448static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
449{
450 if (napi_schedule_prep(napi)) {
451 __napi_schedule(napi);
4d29515f 452 return true;
bfe13f54 453 }
4d29515f 454 return false;
bfe13f54
RD
455}
456
364b6055 457bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
458/**
459 * napi_complete - NAPI processing complete
5e82b4b2 460 * @n: NAPI context
bea3348e
SH
461 *
462 * Mark NAPI processing as complete.
3b47d303 463 * Consider using napi_complete_done() instead.
364b6055 464 * Return false if device should avoid rearming interrupts.
bea3348e 465 */
364b6055 466static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
467{
468 return napi_complete_done(n, 0);
469}
bea3348e 470
af12fa6e
ET
471/**
472 * napi_hash_del - remove a NAPI from global table
5e82b4b2 473 * @napi: NAPI context
af12fa6e 474 *
5e82b4b2 475 * Warning: caller must observe RCU grace period
34cbe27e
ED
476 * before freeing memory containing @napi, if
477 * this function returns true.
93d05d4a 478 * Note: core networking stack automatically calls it
5e82b4b2 479 * from netif_napi_del().
93d05d4a 480 * Drivers might want to call this helper to combine all
5e82b4b2 481 * the needed RCU grace periods into a single one.
af12fa6e 482 */
34cbe27e 483bool napi_hash_del(struct napi_struct *napi);
af12fa6e 484
bea3348e
SH
485/**
486 * napi_disable - prevent NAPI from scheduling
5e82b4b2 487 * @n: NAPI context
bea3348e
SH
488 *
489 * Stop NAPI from being scheduled on this context.
490 * Waits till any outstanding processing completes.
491 */
3b47d303 492void napi_disable(struct napi_struct *n);
bea3348e
SH
493
494/**
495 * napi_enable - enable NAPI scheduling
5e82b4b2 496 * @n: NAPI context
bea3348e
SH
497 *
498 * Resume NAPI from being scheduled on this context.
499 * Must be paired with napi_disable.
500 */
501static inline void napi_enable(struct napi_struct *n)
502{
503 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 504 smp_mb__before_atomic();
bea3348e 505 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 506 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
507}
508
c264c3de
SH
509/**
510 * napi_synchronize - wait until NAPI is not running
5e82b4b2 511 * @n: NAPI context
c264c3de
SH
512 *
513 * Wait until NAPI is done being scheduled on this context.
514 * Waits till any outstanding processing completes but
515 * does not disable future activations.
516 */
517static inline void napi_synchronize(const struct napi_struct *n)
518{
facc432f
AB
519 if (IS_ENABLED(CONFIG_SMP))
520 while (test_bit(NAPI_STATE_SCHED, &n->state))
521 msleep(1);
522 else
523 barrier();
c264c3de 524}
c264c3de 525
d94d9fee 526enum netdev_queue_state_t {
73466498
TH
527 __QUEUE_STATE_DRV_XOFF,
528 __QUEUE_STATE_STACK_XOFF,
c3f26a26 529 __QUEUE_STATE_FROZEN,
79d16385 530};
8e2f1a63
DB
531
532#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
533#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
534#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
535
536#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
537#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 QUEUE_STATE_FROZEN)
539#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
540 QUEUE_STATE_FROZEN)
541
73466498
TH
542/*
543 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
544 * netif_tx_* functions below are used to manipulate this flag. The
545 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
546 * queue independently. The netif_xmit_*stopped functions below are called
547 * to check if the queue has been stopped by the driver or stack (either
548 * of the XOFF bits are set in the state). Drivers should not need to call
549 * netif_xmit*stopped functions, they should only be using netif_tx_*.
550 */
79d16385 551
bb949fbd 552struct netdev_queue {
6a321cb3 553/*
5e82b4b2 554 * read-mostly part
6a321cb3 555 */
bb949fbd 556 struct net_device *dev;
46e5da40 557 struct Qdisc __rcu *qdisc;
b0e1e646 558 struct Qdisc *qdisc_sleeping;
ccf5ff69 559#ifdef CONFIG_SYSFS
1d24eb48
TH
560 struct kobject kobj;
561#endif
f2cd2d3e
ED
562#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
563 int numa_node;
564#endif
c0ef079c
FW
565 unsigned long tx_maxrate;
566 /*
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
569 */
570 unsigned long trans_timeout;
6a321cb3 571/*
5e82b4b2 572 * write-mostly part
6a321cb3
ED
573 */
574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
575 int xmit_lock_owner;
9d21493b 576 /*
9b36627a 577 * Time (in jiffies) of last Tx
9d21493b
ED
578 */
579 unsigned long trans_start;
ccf5ff69 580
114cf580
TH
581 unsigned long state;
582
583#ifdef CONFIG_BQL
584 struct dql dql;
585#endif
e8a0464c 586} ____cacheline_aligned_in_smp;
bb949fbd 587
79134e6c
ED
588extern int sysctl_fb_tunnels_only_for_init_net;
589
590static inline bool net_has_fallback_tunnels(const struct net *net)
591{
592 return net == &init_net || !sysctl_fb_tunnels_only_for_init_net;
593}
594
f2cd2d3e
ED
595static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
596{
597#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
598 return q->numa_node;
599#else
b236da69 600 return NUMA_NO_NODE;
f2cd2d3e
ED
601#endif
602}
603
604static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
605{
606#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
607 q->numa_node = node;
608#endif
609}
610
df334545 611#ifdef CONFIG_RPS
0a9627f2
TH
612/*
613 * This structure holds an RPS map which can be of variable length. The
614 * map is an array of CPUs.
615 */
616struct rps_map {
617 unsigned int len;
618 struct rcu_head rcu;
619 u16 cpus[0];
620};
60b778ce 621#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 622
fec5e652 623/*
c445477d
BH
624 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
625 * tail pointer for that CPU's input queue at the time of last enqueue, and
626 * a hardware filter index.
fec5e652
TH
627 */
628struct rps_dev_flow {
629 u16 cpu;
c445477d 630 u16 filter;
fec5e652
TH
631 unsigned int last_qtail;
632};
c445477d 633#define RPS_NO_FILTER 0xffff
fec5e652
TH
634
635/*
636 * The rps_dev_flow_table structure contains a table of flow mappings.
637 */
638struct rps_dev_flow_table {
639 unsigned int mask;
640 struct rcu_head rcu;
fec5e652
TH
641 struct rps_dev_flow flows[0];
642};
643#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 644 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
645
646/*
647 * The rps_sock_flow_table contains mappings of flows to the last CPU
648 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
649 * Each entry is a 32bit value. Upper part is the high-order bits
650 * of flow hash, lower part is CPU number.
567e4b79 651 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
652 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
653 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 654 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
655 */
656struct rps_sock_flow_table {
567e4b79 657 u32 mask;
93c1af6c
ED
658
659 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 660};
567e4b79 661#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
662
663#define RPS_NO_CPU 0xffff
664
567e4b79
ED
665extern u32 rps_cpu_mask;
666extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
667
fec5e652
TH
668static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
669 u32 hash)
670{
671 if (table && hash) {
567e4b79
ED
672 unsigned int index = hash & table->mask;
673 u32 val = hash & ~rps_cpu_mask;
fec5e652 674
5e82b4b2 675 /* We only give a hint, preemption can change CPU under us */
567e4b79 676 val |= raw_smp_processor_id();
fec5e652 677
567e4b79
ED
678 if (table->ents[index] != val)
679 table->ents[index] = val;
fec5e652
TH
680 }
681}
682
c445477d 683#ifdef CONFIG_RFS_ACCEL
f629d208
JP
684bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
685 u16 filter_id);
c445477d 686#endif
a953be53 687#endif /* CONFIG_RPS */
c445477d 688
0a9627f2
TH
689/* This structure contains an instance of an RX queue. */
690struct netdev_rx_queue {
a953be53 691#ifdef CONFIG_RPS
6e3f7faf
ED
692 struct rps_map __rcu *rps_map;
693 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 694#endif
6e3f7faf 695 struct kobject kobj;
fe822240 696 struct net_device *dev;
e817f856 697 struct xdp_rxq_info xdp_rxq;
0a9627f2 698} ____cacheline_aligned_in_smp;
a953be53
MD
699
700/*
701 * RX queue sysfs structures and functions.
702 */
703struct rx_queue_attribute {
704 struct attribute attr;
718ad681 705 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
a953be53 706 ssize_t (*store)(struct netdev_rx_queue *queue,
718ad681 707 const char *buf, size_t len);
a953be53 708};
d314774c 709
bf264145
TH
710#ifdef CONFIG_XPS
711/*
712 * This structure holds an XPS map which can be of variable length. The
713 * map is an array of queues.
714 */
715struct xps_map {
716 unsigned int len;
717 unsigned int alloc_len;
718 struct rcu_head rcu;
719 u16 queues[0];
720};
60b778ce 721#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
722#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
723 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
724
725/*
726 * This structure holds all XPS maps for device. Maps are indexed by CPU.
727 */
728struct xps_dev_maps {
729 struct rcu_head rcu;
a4177869 730 struct xps_map __rcu *cpu_map[0];
bf264145 731};
184c449f
AD
732#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
733 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
734#endif /* CONFIG_XPS */
735
4f57c087
JF
736#define TC_MAX_QUEUE 16
737#define TC_BITMASK 15
738/* HW offloaded queuing disciplines txq count and offset maps */
739struct netdev_tc_txq {
740 u16 count;
741 u16 offset;
742};
743
68bad94e
NP
744#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
745/*
746 * This structure is to hold information about the device
747 * configured to run FCoE protocol stack.
748 */
749struct netdev_fcoe_hbainfo {
750 char manufacturer[64];
751 char serial_number[64];
752 char hardware_version[64];
753 char driver_version[64];
754 char optionrom_version[64];
755 char firmware_version[64];
756 char model[256];
757 char model_description[256];
758};
759#endif
760
02637fce 761#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 762
02637fce
JP
763/* This structure holds a unique identifier to identify some
764 * physical item (port for example) used by a netdevice.
66b52b0d 765 */
02637fce
JP
766struct netdev_phys_item_id {
767 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
768 unsigned char id_len;
769};
770
d754f98b
SF
771static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
772 struct netdev_phys_item_id *b)
773{
774 return a->id_len == b->id_len &&
775 memcmp(a->id, b->id, a->id_len) == 0;
776}
777
99932d4f
DB
778typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
779 struct sk_buff *skb);
780
2572ac53 781enum tc_setup_type {
575ed7d3 782 TC_SETUP_QDISC_MQPRIO,
a1b7c5fd 783 TC_SETUP_CLSU32,
5b33f488 784 TC_SETUP_CLSFLOWER,
ade9b658 785 TC_SETUP_CLSMATCHALL,
332ae8e2 786 TC_SETUP_CLSBPF,
8c4083b3 787 TC_SETUP_BLOCK,
8521db4c 788 TC_SETUP_QDISC_CBS,
602f3baf 789 TC_SETUP_QDISC_RED,
7fdb61b4 790 TC_SETUP_QDISC_PRIO,
16e5cc64
JF
791};
792
f4e63525
JK
793/* These structures hold the attributes of bpf state that are being passed
794 * to the netdevice through the bpf op.
a7862b45 795 */
f4e63525 796enum bpf_netdev_command {
a7862b45
BB
797 /* Set or clear a bpf program used in the earliest stages of packet
798 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
799 * is responsible for calling bpf_prog_put on any old progs that are
800 * stored. In case of error, the callee need not release the new prog
801 * reference, but on success it takes ownership and must bpf_prog_put
802 * when it is no longer used.
803 */
804 XDP_SETUP_PROG,
ee5d032f 805 XDP_SETUP_PROG_HW,
a7862b45 806 /* Check if a bpf program is set on the device. The callee should
ce158e58
JK
807 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
808 * is equivalent to XDP_ATTACHED_DRV.
a7862b45
BB
809 */
810 XDP_QUERY_PROG,
ab3f0063
JK
811 /* BPF program for offload callbacks, invoked at program load time. */
812 BPF_OFFLOAD_VERIFIER_PREP,
813 BPF_OFFLOAD_TRANSLATE,
814 BPF_OFFLOAD_DESTROY,
a3884572
JK
815 BPF_OFFLOAD_MAP_ALLOC,
816 BPF_OFFLOAD_MAP_FREE,
a7862b45
BB
817};
818
cae1927c 819struct bpf_prog_offload_ops;
ddf9f970
JK
820struct netlink_ext_ack;
821
f4e63525
JK
822struct netdev_bpf {
823 enum bpf_netdev_command command;
a7862b45
BB
824 union {
825 /* XDP_SETUP_PROG */
ddf9f970 826 struct {
32d60277 827 u32 flags;
ddf9f970
JK
828 struct bpf_prog *prog;
829 struct netlink_ext_ack *extack;
830 };
a7862b45 831 /* XDP_QUERY_PROG */
58038695 832 struct {
ce158e58 833 u8 prog_attached;
58038695 834 u32 prog_id;
92f0292b
JK
835 /* flags with which program was installed */
836 u32 prog_flags;
58038695 837 };
ab3f0063
JK
838 /* BPF_OFFLOAD_VERIFIER_PREP */
839 struct {
840 struct bpf_prog *prog;
cae1927c 841 const struct bpf_prog_offload_ops *ops; /* callee set */
ab3f0063
JK
842 } verifier;
843 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
844 struct {
845 struct bpf_prog *prog;
846 } offload;
a3884572
JK
847 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
848 struct {
849 struct bpf_offloaded_map *offmap;
850 };
a7862b45
BB
851 };
852};
16e5cc64 853
d77e38e6
SK
854#ifdef CONFIG_XFRM_OFFLOAD
855struct xfrmdev_ops {
856 int (*xdo_dev_state_add) (struct xfrm_state *x);
857 void (*xdo_dev_state_delete) (struct xfrm_state *x);
858 void (*xdo_dev_state_free) (struct xfrm_state *x);
859 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
860 struct xfrm_state *x);
50bd870a 861 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
d77e38e6
SK
862};
863#endif
864
6c557001
FW
865struct dev_ifalias {
866 struct rcu_head rcuhead;
867 char ifalias[];
868};
869
d314774c
SH
870/*
871 * This structure defines the management hooks for network devices.
00829823
SH
872 * The following hooks can be defined; unless noted otherwise, they are
873 * optional and can be filled with a null pointer.
d314774c
SH
874 *
875 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
876 * This function is called once when a network device is registered.
877 * The network device can use this for any late stage initialization
878 * or semantic validation. It can fail with an error code which will
879 * be propagated back to register_netdev.
d314774c
SH
880 *
881 * void (*ndo_uninit)(struct net_device *dev);
882 * This function is called when device is unregistered or when registration
883 * fails. It is not called if init fails.
884 *
885 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 886 * This function is called when a network device transitions to the up
d314774c
SH
887 * state.
888 *
889 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 890 * This function is called when a network device transitions to the down
d314774c
SH
891 * state.
892 *
dc1f8bf6
SH
893 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
894 * struct net_device *dev);
00829823 895 * Called when a packet needs to be transmitted.
e79d8429
RR
896 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
897 * the queue before that can happen; it's for obsolete devices and weird
898 * corner cases, but the stack really does a non-trivial amount
899 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 900 * Required; cannot be NULL.
00829823 901 *
1a2a1444
DM
902 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
903 * struct net_device *dev
904 * netdev_features_t features);
905 * Called by core transmit path to determine if device is capable of
906 * performing offload operations on a given packet. This is to give
907 * the device an opportunity to implement any restrictions that cannot
908 * be otherwise expressed by feature flags. The check is called with
909 * the set of features that the stack has calculated and it returns
910 * those the driver believes to be appropriate.
cdba756f 911 *
f663dd9a 912 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 913 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 914 * Called to decide which queue to use when device supports multiple
00829823
SH
915 * transmit queues.
916 *
d314774c
SH
917 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
918 * This function is called to allow device receiver to make
5e82b4b2 919 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
920 *
921 * void (*ndo_set_rx_mode)(struct net_device *dev);
922 * This function is called device changes address list filtering.
01789349 923 * If driver handles unicast address filtering, it should set
5e82b4b2 924 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
925 *
926 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
927 * This function is called when the Media Access Control address
37b607c5 928 * needs to be changed. If this interface is not defined, the
5e82b4b2 929 * MAC address can not be changed.
d314774c
SH
930 *
931 * int (*ndo_validate_addr)(struct net_device *dev);
932 * Test if Media Access Control address is valid for the device.
933 *
934 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
935 * Called when a user requests an ioctl which can't be handled by
936 * the generic interface code. If not defined ioctls return
d314774c
SH
937 * not supported error code.
938 *
939 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
940 * Used to set network devices bus interface parameters. This interface
5e82b4b2 941 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
942 * interface (PCI) for low level management.
943 *
944 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
945 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 946 * of a device.
d314774c 947 *
00829823 948 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 949 * Callback used when the transmitter has not made any progress
d314774c
SH
950 * for dev->watchdog ticks.
951 *
bc1f4470 952 * void (*ndo_get_stats64)(struct net_device *dev,
953 * struct rtnl_link_stats64 *storage);
d308e38f 954 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 955 * Called when a user wants to get the network device usage
be1f3c2c 956 * statistics. Drivers must do one of the following:
3cfde79c
BH
957 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
958 * rtnl_link_stats64 structure passed by the caller.
82695d9b 959 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
960 * (which should normally be dev->stats) and return a pointer to
961 * it. The structure may be changed asynchronously only if each
962 * field is written atomically.
963 * 3. Update dev->stats asynchronously and atomically, and define
964 * neither operation.
d314774c 965 *
3df5b3c6 966 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
967 * Return true if this device supports offload stats of this attr_id.
968 *
969 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
970 * void *attr_data)
971 * Get statistics for offload operations by attr_id. Write it into the
972 * attr_data pointer.
973 *
5d632cb7 974 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 975 * If device supports VLAN filtering this function is called when a
80d5c368 976 * VLAN id is registered.
d314774c 977 *
5d632cb7 978 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 979 * If device supports VLAN filtering this function is called when a
80d5c368 980 * VLAN id is unregistered.
d314774c
SH
981 *
982 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
983 *
984 * SR-IOV management functions.
985 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
986 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
987 * u8 qos, __be16 proto);
ed616689
SC
988 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
989 * int max_tx_rate);
5f8444a3 990 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 991 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
992 * int (*ndo_get_vf_config)(struct net_device *dev,
993 * int vf, struct ifla_vf_info *ivf);
1d8faf48 994 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
995 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
996 * struct nlattr *port[]);
01a3d796
VZ
997 *
998 * Enable or disable the VF ability to query its RSS Redirection Table and
999 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 1000 * with PF and querying it may introduce a theoretical security risk.
01a3d796 1001 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 1002 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
2572ac53 1003 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
de4784ca 1004 * void *type_data);
6a4bc2b4
FF
1005 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1006 * This is always called from the stack with the rtnl lock held and netif
1007 * tx queues stopped. This allows the netdevice to perform queue
1008 * management safely.
c445477d 1009 *
e9bce845
YZ
1010 * Fiber Channel over Ethernet (FCoE) offload functions.
1011 * int (*ndo_fcoe_enable)(struct net_device *dev);
1012 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1013 * so the underlying device can perform whatever needed configuration or
1014 * initialization to support acceleration of FCoE traffic.
1015 *
1016 * int (*ndo_fcoe_disable)(struct net_device *dev);
1017 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1018 * so the underlying device can perform whatever needed clean-ups to
1019 * stop supporting acceleration of FCoE traffic.
1020 *
1021 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1022 * struct scatterlist *sgl, unsigned int sgc);
1023 * Called when the FCoE Initiator wants to initialize an I/O that
1024 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1025 * perform necessary setup and returns 1 to indicate the device is set up
1026 * successfully to perform DDP on this I/O, otherwise this returns 0.
1027 *
1028 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1029 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1030 * indicated by the FC exchange id 'xid', so the underlying device can
1031 * clean up and reuse resources for later DDP requests.
1032 *
1033 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1034 * struct scatterlist *sgl, unsigned int sgc);
1035 * Called when the FCoE Target wants to initialize an I/O that
1036 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1037 * perform necessary setup and returns 1 to indicate the device is set up
1038 * successfully to perform DDP on this I/O, otherwise this returns 0.
1039 *
68bad94e
NP
1040 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1041 * struct netdev_fcoe_hbainfo *hbainfo);
1042 * Called when the FCoE Protocol stack wants information on the underlying
1043 * device. This information is utilized by the FCoE protocol stack to
1044 * register attributes with Fiber Channel management service as per the
1045 * FC-GS Fabric Device Management Information(FDMI) specification.
1046 *
e9bce845
YZ
1047 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1048 * Called when the underlying device wants to override default World Wide
1049 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1050 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1051 * protocol stack to use.
1052 *
c445477d
BH
1053 * RFS acceleration.
1054 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1055 * u16 rxq_index, u32 flow_id);
1056 * Set hardware filter for RFS. rxq_index is the target queue index;
1057 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1058 * Return the filter ID on success, or a negative error code.
fbaec0ea 1059 *
8b98a70c 1060 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1061 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1062 * Called to make another netdev an underling.
1063 *
1064 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1065 * Called to release previously enslaved netdev.
5455c699
MM
1066 *
1067 * Feature/offload setting functions.
1a2a1444
DM
1068 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1069 * netdev_features_t features);
1070 * Adjusts the requested feature flags according to device-specific
1071 * constraints, and returns the resulting flags. Must not modify
1072 * the device state.
1073 *
c8f44aff 1074 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1075 * Called to update device configuration to new features. Passed
1076 * feature set might be less than what was returned by ndo_fix_features()).
1077 * Must return >0 or -errno if it changed dev->features itself.
1078 *
edc7d573 1079 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1080 * struct net_device *dev,
f6f6424b 1081 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1082 * Adds an FDB entry to dev for addr.
1690be63
VY
1083 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1084 * struct net_device *dev,
f6f6424b 1085 * const unsigned char *addr, u16 vid)
77162022
JF
1086 * Deletes the FDB entry from dev coresponding to addr.
1087 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1088 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1089 * int *idx)
77162022
JF
1090 * Used to add FDB entries to dump requests. Implementers should add
1091 * entries to skb and update idx with the number of entries.
e5a55a89 1092 *
ad41faa8
ND
1093 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1094 * u16 flags)
e5a55a89 1095 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1096 * struct net_device *dev, u32 filter_mask,
1097 * int nlflags)
ad41faa8
ND
1098 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1099 * u16 flags);
4bf84c35
JP
1100 *
1101 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1102 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1103 * which do not represent real hardware may define this to allow their
1104 * userspace components to manage their virtual carrier state. Devices
1105 * that determine carrier state from physical hardware properties (eg
1106 * network cables) or protocol-dependent mechanisms (eg
1107 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1108 *
1109 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1110 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1111 * Called to get ID of physical port of this device. If driver does
1112 * not implement this, it is assumed that the hw is not able to have
1113 * multiple net devices on single physical port.
53cf5275 1114 *
7c46a640
AD
1115 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1116 * struct udp_tunnel_info *ti);
1117 * Called by UDP tunnel to notify a driver about the UDP port and socket
1118 * address family that a UDP tunnel is listnening to. It is called only
1119 * when a new port starts listening. The operation is protected by the
1120 * RTNL.
1121 *
1122 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1123 * struct udp_tunnel_info *ti);
1124 * Called by UDP tunnel to notify the driver about a UDP port and socket
1125 * address family that the UDP tunnel is not listening to anymore. The
1126 * operation is protected by the RTNL.
1127 *
a6cc0cfa
JF
1128 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1129 * struct net_device *dev)
1130 * Called by upper layer devices to accelerate switching or other
1131 * station functionality into hardware. 'pdev is the lowerdev
1132 * to use for the offload and 'dev' is the net device that will
1133 * back the offload. Returns a pointer to the private structure
1134 * the upper layer will maintain.
1135 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1136 * Called by upper layer device to delete the station created
1137 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1138 * the station and priv is the structure returned by the add
1139 * operation.
822b3b2e
JF
1140 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1141 * int queue_index, u32 maxrate);
1142 * Called when a user wants to set a max-rate limitation of specific
1143 * TX queue.
a54acb3a
ND
1144 * int (*ndo_get_iflink)(const struct net_device *dev);
1145 * Called to get the iflink value of this device.
d746d707 1146 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1147 * bool proto_down);
d746d707
AK
1148 * This function is used to pass protocol port error state information
1149 * to the switch driver. The switch driver can react to the proto_down
1150 * by doing a phys down on the associated switch port.
fc4099f1
PS
1151 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1152 * This function is used to get egress tunnel information for given skb.
1153 * This is useful for retrieving outer tunnel header parameters while
1154 * sampling packet.
871b642a
PA
1155 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1156 * This function is used to specify the headroom that the skb must
1157 * consider when allocation skb during packet reception. Setting
1158 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1159 * forward. Setting a negative value resets the rx headroom to the
871b642a 1160 * default value.
f4e63525 1161 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
a7862b45 1162 * This function is used to set or query state related to XDP on the
f4e63525
JK
1163 * netdevice and manage BPF offload. See definition of
1164 * enum bpf_netdev_command for details.
814abfab
JF
1165 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1166 * This function is used to submit a XDP packet for transmit on a
1167 * netdevice.
11393cc9 1168 * void (*ndo_xdp_flush)(struct net_device *dev);
f16ded59
RR
1169 * This function is used to inform the driver to flush a particular
1170 * xdp tx queue. Must be called on same CPU as xdp_xmit.
d314774c
SH
1171 */
1172struct net_device_ops {
1173 int (*ndo_init)(struct net_device *dev);
1174 void (*ndo_uninit)(struct net_device *dev);
1175 int (*ndo_open)(struct net_device *dev);
1176 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1177 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1178 struct net_device *dev);
1179 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1180 struct net_device *dev,
1181 netdev_features_t features);
00829823 1182 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1183 struct sk_buff *skb,
99932d4f
DB
1184 void *accel_priv,
1185 select_queue_fallback_t fallback);
d314774c
SH
1186 void (*ndo_change_rx_flags)(struct net_device *dev,
1187 int flags);
d314774c 1188 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1189 int (*ndo_set_mac_address)(struct net_device *dev,
1190 void *addr);
d314774c 1191 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1192 int (*ndo_do_ioctl)(struct net_device *dev,
1193 struct ifreq *ifr, int cmd);
d314774c
SH
1194 int (*ndo_set_config)(struct net_device *dev,
1195 struct ifmap *map);
00829823
SH
1196 int (*ndo_change_mtu)(struct net_device *dev,
1197 int new_mtu);
1198 int (*ndo_neigh_setup)(struct net_device *dev,
1199 struct neigh_parms *);
d314774c
SH
1200 void (*ndo_tx_timeout) (struct net_device *dev);
1201
bc1f4470 1202 void (*ndo_get_stats64)(struct net_device *dev,
1203 struct rtnl_link_stats64 *storage);
3df5b3c6 1204 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1205 int (*ndo_get_offload_stats)(int attr_id,
1206 const struct net_device *dev,
1207 void *attr_data);
d314774c
SH
1208 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1209
8e586137 1210 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1211 __be16 proto, u16 vid);
8e586137 1212 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1213 __be16 proto, u16 vid);
d314774c 1214#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1215 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1216 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1217 struct netpoll_info *info);
0e34e931 1218 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1219#endif
95c26df8
WM
1220 int (*ndo_set_vf_mac)(struct net_device *dev,
1221 int queue, u8 *mac);
1222 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1223 int queue, u16 vlan,
1224 u8 qos, __be16 proto);
ed616689
SC
1225 int (*ndo_set_vf_rate)(struct net_device *dev,
1226 int vf, int min_tx_rate,
1227 int max_tx_rate);
5f8444a3
GR
1228 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1229 int vf, bool setting);
dd461d6a
HS
1230 int (*ndo_set_vf_trust)(struct net_device *dev,
1231 int vf, bool setting);
95c26df8
WM
1232 int (*ndo_get_vf_config)(struct net_device *dev,
1233 int vf,
1234 struct ifla_vf_info *ivf);
1d8faf48
RE
1235 int (*ndo_set_vf_link_state)(struct net_device *dev,
1236 int vf, int link_state);
3b766cd8
EBE
1237 int (*ndo_get_vf_stats)(struct net_device *dev,
1238 int vf,
1239 struct ifla_vf_stats
1240 *vf_stats);
57b61080
SF
1241 int (*ndo_set_vf_port)(struct net_device *dev,
1242 int vf,
1243 struct nlattr *port[]);
1244 int (*ndo_get_vf_port)(struct net_device *dev,
1245 int vf, struct sk_buff *skb);
cc8e27cc
EC
1246 int (*ndo_set_vf_guid)(struct net_device *dev,
1247 int vf, u64 guid,
1248 int guid_type);
01a3d796
VZ
1249 int (*ndo_set_vf_rss_query_en)(
1250 struct net_device *dev,
1251 int vf, bool setting);
16e5cc64 1252 int (*ndo_setup_tc)(struct net_device *dev,
2572ac53 1253 enum tc_setup_type type,
de4784ca 1254 void *type_data);
d11ead75 1255#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1256 int (*ndo_fcoe_enable)(struct net_device *dev);
1257 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1258 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1259 u16 xid,
1260 struct scatterlist *sgl,
1261 unsigned int sgc);
1262 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1263 u16 xid);
6247e086
YZ
1264 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1265 u16 xid,
1266 struct scatterlist *sgl,
1267 unsigned int sgc);
68bad94e
NP
1268 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1269 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1270#endif
1271
d11ead75 1272#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1273#define NETDEV_FCOE_WWNN 0
1274#define NETDEV_FCOE_WWPN 1
1275 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1276 u64 *wwn, int type);
4d288d57 1277#endif
3c9c36bc 1278
c445477d
BH
1279#ifdef CONFIG_RFS_ACCEL
1280 int (*ndo_rx_flow_steer)(struct net_device *dev,
1281 const struct sk_buff *skb,
1282 u16 rxq_index,
1283 u32 flow_id);
1284#endif
fbaec0ea 1285 int (*ndo_add_slave)(struct net_device *dev,
33eaf2a6
DA
1286 struct net_device *slave_dev,
1287 struct netlink_ext_ack *extack);
fbaec0ea
JP
1288 int (*ndo_del_slave)(struct net_device *dev,
1289 struct net_device *slave_dev);
c8f44aff
MM
1290 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1291 netdev_features_t features);
5455c699 1292 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1293 netdev_features_t features);
503eebc2
JP
1294 int (*ndo_neigh_construct)(struct net_device *dev,
1295 struct neighbour *n);
1296 void (*ndo_neigh_destroy)(struct net_device *dev,
1297 struct neighbour *n);
77162022
JF
1298
1299 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1300 struct nlattr *tb[],
77162022 1301 struct net_device *dev,
6b6e2725 1302 const unsigned char *addr,
f6f6424b 1303 u16 vid,
77162022
JF
1304 u16 flags);
1305 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1306 struct nlattr *tb[],
77162022 1307 struct net_device *dev,
f6f6424b
JP
1308 const unsigned char *addr,
1309 u16 vid);
77162022
JF
1310 int (*ndo_fdb_dump)(struct sk_buff *skb,
1311 struct netlink_callback *cb,
1312 struct net_device *dev,
5d5eacb3 1313 struct net_device *filter_dev,
d297653d 1314 int *idx);
e5a55a89
JF
1315
1316 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1317 struct nlmsghdr *nlh,
1318 u16 flags);
e5a55a89
JF
1319 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1320 u32 pid, u32 seq,
6cbdceeb 1321 struct net_device *dev,
46c264da
ND
1322 u32 filter_mask,
1323 int nlflags);
407af329 1324 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1325 struct nlmsghdr *nlh,
1326 u16 flags);
4bf84c35
JP
1327 int (*ndo_change_carrier)(struct net_device *dev,
1328 bool new_carrier);
66b52b0d 1329 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1330 struct netdev_phys_item_id *ppid);
db24a904
DA
1331 int (*ndo_get_phys_port_name)(struct net_device *dev,
1332 char *name, size_t len);
7c46a640
AD
1333 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1334 struct udp_tunnel_info *ti);
1335 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1336 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1337 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1338 struct net_device *dev);
1339 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1340 void *priv);
1341
25175ba5 1342 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1343 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1344 int queue_index,
1345 u32 maxrate);
a54acb3a 1346 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1347 int (*ndo_change_proto_down)(struct net_device *dev,
1348 bool proto_down);
fc4099f1
PS
1349 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1350 struct sk_buff *skb);
871b642a
PA
1351 void (*ndo_set_rx_headroom)(struct net_device *dev,
1352 int needed_headroom);
f4e63525
JK
1353 int (*ndo_bpf)(struct net_device *dev,
1354 struct netdev_bpf *bpf);
814abfab
JF
1355 int (*ndo_xdp_xmit)(struct net_device *dev,
1356 struct xdp_buff *xdp);
11393cc9 1357 void (*ndo_xdp_flush)(struct net_device *dev);
d314774c
SH
1358};
1359
7aa98047
LR
1360/**
1361 * enum net_device_priv_flags - &struct net_device priv_flags
1362 *
1363 * These are the &struct net_device, they are only set internally
1364 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1365 * userspace; this means that the order of these flags can change
7aa98047
LR
1366 * during any kernel release.
1367 *
1368 * You should have a pretty good reason to be extending these flags.
1369 *
1370 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1371 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1372 * @IFF_BONDING: bonding master or slave
7aa98047 1373 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1374 * @IFF_WAN_HDLC: WAN HDLC device
1375 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1376 * release skb->dst
1377 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1378 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1379 * @IFF_MACVLAN_PORT: device used as macvlan port
1380 * @IFF_BRIDGE_PORT: device used as bridge port
1381 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1382 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1383 * @IFF_UNICAST_FLT: Supports unicast filtering
1384 * @IFF_TEAM_PORT: device used as team port
1385 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1386 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1387 * change when it's running
1388 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1389 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1390 * underlying stacked devices
007979ea 1391 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1392 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1393 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1394 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1395 * @IFF_TEAM: device is a team device
d4ab4286 1396 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1397 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1398 * entity (i.e. the master device for bridged veth)
3c175784 1399 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1400 */
1401enum netdev_priv_flags {
1402 IFF_802_1Q_VLAN = 1<<0,
1403 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1404 IFF_BONDING = 1<<2,
1405 IFF_ISATAP = 1<<3,
1406 IFF_WAN_HDLC = 1<<4,
1407 IFF_XMIT_DST_RELEASE = 1<<5,
1408 IFF_DONT_BRIDGE = 1<<6,
1409 IFF_DISABLE_NETPOLL = 1<<7,
1410 IFF_MACVLAN_PORT = 1<<8,
1411 IFF_BRIDGE_PORT = 1<<9,
1412 IFF_OVS_DATAPATH = 1<<10,
1413 IFF_TX_SKB_SHARING = 1<<11,
1414 IFF_UNICAST_FLT = 1<<12,
1415 IFF_TEAM_PORT = 1<<13,
1416 IFF_SUPP_NOFCS = 1<<14,
1417 IFF_LIVE_ADDR_CHANGE = 1<<15,
1418 IFF_MACVLAN = 1<<16,
1419 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1ec54cb4
PA
1420 IFF_L3MDEV_MASTER = 1<<18,
1421 IFF_NO_QUEUE = 1<<19,
1422 IFF_OPENVSWITCH = 1<<20,
1423 IFF_L3MDEV_SLAVE = 1<<21,
1424 IFF_TEAM = 1<<22,
1425 IFF_RXFH_CONFIGURED = 1<<23,
1426 IFF_PHONY_HEADROOM = 1<<24,
1427 IFF_MACSEC = 1<<25,
7aa98047
LR
1428};
1429
1430#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1431#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1432#define IFF_BONDING IFF_BONDING
7aa98047 1433#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1434#define IFF_WAN_HDLC IFF_WAN_HDLC
1435#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1436#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1437#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1438#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1439#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1440#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1441#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1442#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1443#define IFF_TEAM_PORT IFF_TEAM_PORT
1444#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1445#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1446#define IFF_MACVLAN IFF_MACVLAN
02875878 1447#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
007979ea 1448#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1449#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1450#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1451#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1452#define IFF_TEAM IFF_TEAM
d4ab4286 1453#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1454#define IFF_MACSEC IFF_MACSEC
7aa98047 1455
536721b1
KK
1456/**
1457 * struct net_device - The DEVICE structure.
d651983d
MCC
1458 *
1459 * Actually, this whole structure is a big mistake. It mixes I/O
1460 * data with strictly "high-level" data, and it has to know about
1461 * almost every data structure used in the INET module.
536721b1
KK
1462 *
1463 * @name: This is the first field of the "visible" part of this structure
1464 * (i.e. as seen by users in the "Space.c" file). It is the name
d651983d 1465 * of the interface.
536721b1
KK
1466 *
1467 * @name_hlist: Device name hash chain, please keep it close to name[]
1468 * @ifalias: SNMP alias
1469 * @mem_end: Shared memory end
1470 * @mem_start: Shared memory start
1471 * @base_addr: Device I/O address
1472 * @irq: Device IRQ number
1473 *
1474 * @state: Generic network queuing layer state, see netdev_state_t
1475 * @dev_list: The global list of network devices
5e82b4b2
BH
1476 * @napi_list: List entry used for polling NAPI devices
1477 * @unreg_list: List entry when we are unregistering the
1478 * device; see the function unregister_netdev
1479 * @close_list: List entry used when we are closing the device
62d885fe
BP
1480 * @ptype_all: Device-specific packet handlers for all protocols
1481 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1482 *
1483 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1484 * @features: Currently active device features
1485 * @hw_features: User-changeable features
1486 *
1487 * @wanted_features: User-requested features
1488 * @vlan_features: Mask of features inheritable by VLAN devices
1489 *
1490 * @hw_enc_features: Mask of features inherited by encapsulating devices
1491 * This field indicates what encapsulation
1492 * offloads the hardware is capable of doing,
1493 * and drivers will need to set them appropriately.
1494 *
1495 * @mpls_features: Mask of features inheritable by MPLS
1496 *
1497 * @ifindex: interface index
5e82b4b2 1498 * @group: The group the device belongs to
536721b1
KK
1499 *
1500 * @stats: Statistics struct, which was left as a legacy, use
1501 * rtnl_link_stats64 instead
1502 *
1503 * @rx_dropped: Dropped packets by core network,
1504 * do not use this in drivers
1505 * @tx_dropped: Dropped packets by core network,
1506 * do not use this in drivers
6e7333d3
JW
1507 * @rx_nohandler: nohandler dropped packets by core network on
1508 * inactive devices, do not use this in drivers
9e55e5d3
FF
1509 * @carrier_up_count: Number of times the carrier has been up
1510 * @carrier_down_count: Number of times the carrier has been down
536721b1 1511 *
536721b1
KK
1512 * @wireless_handlers: List of functions to handle Wireless Extensions,
1513 * instead of ioctl,
1514 * see <net/iw_handler.h> for details.
1515 * @wireless_data: Instance data managed by the core of wireless extensions
1516 *
1517 * @netdev_ops: Includes several pointers to callbacks,
1518 * if one wants to override the ndo_*() functions
1519 * @ethtool_ops: Management operations
f997c55c
AA
1520 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1521 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1522 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1523 * of Layer 2 headers.
1524 *
1525 * @flags: Interface flags (a la BSD)
1526 * @priv_flags: Like 'flags' but invisible to userspace,
1527 * see if.h for the definitions
1528 * @gflags: Global flags ( kept as legacy )
1529 * @padded: How much padding added by alloc_netdev()
1530 * @operstate: RFC2863 operstate
1531 * @link_mode: Mapping policy to operstate
1532 * @if_port: Selectable AUI, TP, ...
1533 * @dma: DMA channel
1534 * @mtu: Interface MTU value
61e84623
JW
1535 * @min_mtu: Interface Minimum MTU value
1536 * @max_mtu: Interface Maximum MTU value
536721b1 1537 * @type: Interface hardware type
2793a23a 1538 * @hard_header_len: Maximum hardware header length.
217e6fa2 1539 * @min_header_len: Minimum hardware header length
536721b1
KK
1540 *
1541 * @needed_headroom: Extra headroom the hardware may need, but not in all
1542 * cases can this be guaranteed
1543 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1544 * cases can this be guaranteed. Some cases also use
1545 * LL_MAX_HEADER instead to allocate the skb
1546 *
1547 * interface address info:
1548 *
1549 * @perm_addr: Permanent hw address
1550 * @addr_assign_type: Hw address assignment type
1551 * @addr_len: Hardware address length
8626a0c8 1552 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1553 * @dev_id: Used to differentiate devices that share
1554 * the same link layer address
1555 * @dev_port: Used to differentiate devices that share
1556 * the same function
1557 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1558 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1559 * has been enabled due to the need to listen to
1560 * additional unicast addresses in a device that
1561 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1562 * @uc: unicast mac addresses
1563 * @mc: multicast mac addresses
1564 * @dev_addrs: list of device hw addresses
1565 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1566 * @promiscuity: Number of times the NIC is told to work in
1567 * promiscuous mode; if it becomes 0 the NIC will
1568 * exit promiscuous mode
536721b1
KK
1569 * @allmulti: Counter, enables or disables allmulticast mode
1570 *
1571 * @vlan_info: VLAN info
1572 * @dsa_ptr: dsa specific data
1573 * @tipc_ptr: TIPC specific data
1574 * @atalk_ptr: AppleTalk link
1575 * @ip_ptr: IPv4 specific data
1576 * @dn_ptr: DECnet specific data
1577 * @ip6_ptr: IPv6 specific data
1578 * @ax25_ptr: AX.25 specific data
1579 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1580 *
536721b1
KK
1581 * @dev_addr: Hw address (before bcast,
1582 * because most packets are unicast)
1583 *
1584 * @_rx: Array of RX queues
1585 * @num_rx_queues: Number of RX queues
1586 * allocated at register_netdev() time
1587 * @real_num_rx_queues: Number of RX queues currently active in device
1588 *
1589 * @rx_handler: handler for received packets
1590 * @rx_handler_data: XXX: need comments on this one
46209401
JP
1591 * @miniq_ingress: ingress/clsact qdisc specific data for
1592 * ingress processing
536721b1
KK
1593 * @ingress_queue: XXX: need comments on this one
1594 * @broadcast: hw bcast address
1595 *
14ffbbb8
TG
1596 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1597 * indexed by RX queue number. Assigned by driver.
1598 * This must only be set if the ndo_rx_flow_steer
1599 * operation is defined
1600 * @index_hlist: Device index hash chain
1601 *
536721b1
KK
1602 * @_tx: Array of TX queues
1603 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1604 * @real_num_tx_queues: Number of TX queues currently active in device
1605 * @qdisc: Root qdisc from userspace point of view
1606 * @tx_queue_len: Max frames per queue allowed
1607 * @tx_global_lock: XXX: need comments on this one
1608 *
1609 * @xps_maps: XXX: need comments on this one
46209401
JP
1610 * @miniq_egress: clsact qdisc specific data for
1611 * egress processing
536721b1 1612 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1613 * the watchdog (see dev_watchdog())
536721b1
KK
1614 * @watchdog_timer: List of timers
1615 *
1616 * @pcpu_refcnt: Number of references to this device
1617 * @todo_list: Delayed register/unregister
536721b1
KK
1618 * @link_watch_list: XXX: need comments on this one
1619 *
1620 * @reg_state: Register/unregister state machine
1621 * @dismantle: Device is going to be freed
1622 * @rtnl_link_state: This enum represents the phases of creating
1623 * a new link
1624 *
cf124db5
DM
1625 * @needs_free_netdev: Should unregister perform free_netdev?
1626 * @priv_destructor: Called from unregister
536721b1
KK
1627 * @npinfo: XXX: need comments on this one
1628 * @nd_net: Network namespace this network device is inside
1629 *
1630 * @ml_priv: Mid-layer private
1631 * @lstats: Loopback statistics
1632 * @tstats: Tunnel statistics
1633 * @dstats: Dummy statistics
1634 * @vstats: Virtual ethernet statistics
1635 *
1636 * @garp_port: GARP
1637 * @mrp_port: MRP
1638 *
1639 * @dev: Class/net/name entry
1640 * @sysfs_groups: Space for optional device, statistics and wireless
1641 * sysfs groups
1642 *
1643 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1644 * @rtnl_link_ops: Rtnl_link_ops
1645 *
1646 * @gso_max_size: Maximum size of generic segmentation offload
1647 * @gso_max_segs: Maximum number of segments that can be passed to the
1648 * NIC for GSO
1649 *
1650 * @dcbnl_ops: Data Center Bridging netlink ops
1651 * @num_tc: Number of traffic classes in the net device
1652 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1653 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1654 *
1655 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1656 *
1657 * @priomap: XXX: need comments on this one
1658 * @phydev: Physical device may attach itself
1659 * for hardware timestamping
1660 *
123b3652
ED
1661 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1662 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1663 *
d746d707
AK
1664 * @proto_down: protocol port state information can be sent to the
1665 * switch driver and used to set the phys state of the
1666 * switch port.
1667 *
1da177e4
LT
1668 * FIXME: cleanup struct net_device such that network protocol info
1669 * moves out.
1670 */
1671
d94d9fee 1672struct net_device {
1da177e4 1673 char name[IFNAMSIZ];
9356b8fc 1674 struct hlist_node name_hlist;
6c557001 1675 struct dev_ifalias __rcu *ifalias;
1da177e4
LT
1676 /*
1677 * I/O specific fields
1678 * FIXME: Merge these and struct ifmap into one
1679 */
536721b1
KK
1680 unsigned long mem_end;
1681 unsigned long mem_start;
1682 unsigned long base_addr;
1683 int irq;
1da177e4
LT
1684
1685 /*
536721b1
KK
1686 * Some hardware also needs these fields (state,dev_list,
1687 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1688 * part of the usual set specified in Space.c.
1689 */
1690
1da177e4
LT
1691 unsigned long state;
1692
7562f876 1693 struct list_head dev_list;
bea3348e 1694 struct list_head napi_list;
44a0873d 1695 struct list_head unreg_list;
5cde2829 1696 struct list_head close_list;
7866a621
SN
1697 struct list_head ptype_all;
1698 struct list_head ptype_specific;
2f268f12 1699
2f268f12
VF
1700 struct {
1701 struct list_head upper;
1702 struct list_head lower;
1703 } adj_list;
1704
c8f44aff 1705 netdev_features_t features;
c8f44aff 1706 netdev_features_t hw_features;
c8f44aff 1707 netdev_features_t wanted_features;
c8f44aff 1708 netdev_features_t vlan_features;
6a674e9c 1709 netdev_features_t hw_enc_features;
0d89d203 1710 netdev_features_t mpls_features;
802ab55a 1711 netdev_features_t gso_partial_features;
04ed3e74 1712
1da177e4 1713 int ifindex;
7a66bbc9 1714 int group;
1da177e4 1715
c45d286e 1716 struct net_device_stats stats;
015f0688 1717
015f0688
ED
1718 atomic_long_t rx_dropped;
1719 atomic_long_t tx_dropped;
6e7333d3 1720 atomic_long_t rx_nohandler;
1da177e4 1721
b2d3bcfa
DD
1722 /* Stats to monitor link on/off, flapping */
1723 atomic_t carrier_up_count;
1724 atomic_t carrier_down_count;
1725
b86e0280 1726#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1727 const struct iw_handler_def *wireless_handlers;
1728 struct iw_public_data *wireless_data;
b86e0280 1729#endif
d314774c 1730 const struct net_device_ops *netdev_ops;
76fd8593 1731 const struct ethtool_ops *ethtool_ops;
4170604f 1732#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1733 const struct switchdev_ops *switchdev_ops;
4170604f 1734#endif
1b69c6d0
DA
1735#ifdef CONFIG_NET_L3_MASTER_DEV
1736 const struct l3mdev_ops *l3mdev_ops;
1737#endif
f997c55c
AA
1738#if IS_ENABLED(CONFIG_IPV6)
1739 const struct ndisc_ops *ndisc_ops;
1740#endif
1da177e4 1741
9cb0d21d 1742#ifdef CONFIG_XFRM_OFFLOAD
d77e38e6
SK
1743 const struct xfrmdev_ops *xfrmdev_ops;
1744#endif
1745
3b04ddde
SH
1746 const struct header_ops *header_ops;
1747
536721b1
KK
1748 unsigned int flags;
1749 unsigned int priv_flags;
1750
1da177e4 1751 unsigned short gflags;
536721b1 1752 unsigned short padded;
1da177e4 1753
536721b1
KK
1754 unsigned char operstate;
1755 unsigned char link_mode;
b00055aa 1756
536721b1
KK
1757 unsigned char if_port;
1758 unsigned char dma;
bdc220da 1759
536721b1 1760 unsigned int mtu;
61e84623
JW
1761 unsigned int min_mtu;
1762 unsigned int max_mtu;
536721b1
KK
1763 unsigned short type;
1764 unsigned short hard_header_len;
d92be7a4 1765 unsigned char min_header_len;
1da177e4 1766
f5184d26
JB
1767 unsigned short needed_headroom;
1768 unsigned short needed_tailroom;
1769
1da177e4 1770 /* Interface address info. */
536721b1
KK
1771 unsigned char perm_addr[MAX_ADDR_LEN];
1772 unsigned char addr_assign_type;
1773 unsigned char addr_len;
a0a9663d 1774 unsigned short neigh_priv_len;
536721b1
KK
1775 unsigned short dev_id;
1776 unsigned short dev_port;
ccffad25 1777 spinlock_t addr_list_lock;
14ffbbb8
TG
1778 unsigned char name_assign_type;
1779 bool uc_promisc;
536721b1
KK
1780 struct netdev_hw_addr_list uc;
1781 struct netdev_hw_addr_list mc;
1782 struct netdev_hw_addr_list dev_addrs;
1783
4c3d5e7b
ED
1784#ifdef CONFIG_SYSFS
1785 struct kset *queues_kset;
1786#endif
9d45abe1
WC
1787 unsigned int promiscuity;
1788 unsigned int allmulti;
1da177e4 1789
1da177e4 1790
5e82b4b2 1791 /* Protocol-specific pointers */
65ac6a5f 1792
d11ead75 1793#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1794 struct vlan_info __rcu *vlan_info;
65ac6a5f 1795#endif
34a430d7 1796#if IS_ENABLED(CONFIG_NET_DSA)
2f657a60 1797 struct dsa_port *dsa_ptr;
37cb0620
YX
1798#endif
1799#if IS_ENABLED(CONFIG_TIPC)
536721b1 1800 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1801#endif
89e58148 1802#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
536721b1 1803 void *atalk_ptr;
89e58148 1804#endif
536721b1 1805 struct in_device __rcu *ip_ptr;
330c7272 1806#if IS_ENABLED(CONFIG_DECNET)
536721b1 1807 struct dn_dev __rcu *dn_ptr;
330c7272 1808#endif
536721b1 1809 struct inet6_dev __rcu *ip6_ptr;
19ff13f2 1810#if IS_ENABLED(CONFIG_AX25)
536721b1 1811 void *ax25_ptr;
19ff13f2 1812#endif
536721b1 1813 struct wireless_dev *ieee80211_ptr;
98a18b6f 1814 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1815#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1816 struct mpls_dev __rcu *mpls_ptr;
1817#endif
1da177e4 1818
9356b8fc 1819/*
cd13539b 1820 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1821 */
9356b8fc 1822 /* Interface address info used in eth_type_trans() */
536721b1 1823 unsigned char *dev_addr;
f001fde5 1824
0a9627f2 1825 struct netdev_rx_queue *_rx;
0a9627f2 1826 unsigned int num_rx_queues;
62fe0b40 1827 unsigned int real_num_rx_queues;
0a9627f2 1828
7acedaf5 1829 struct bpf_prog __rcu *xdp_prog;
3b47d303 1830 unsigned long gro_flush_timeout;
61391cde 1831 rx_handler_func_t __rcu *rx_handler;
1832 void __rcu *rx_handler_data;
e8a0464c 1833
4cda01e8 1834#ifdef CONFIG_NET_CLS_ACT
46209401 1835 struct mini_Qdisc __rcu *miniq_ingress;
d2788d34 1836#endif
24824a09 1837 struct netdev_queue __rcu *ingress_queue;
e687ad60 1838#ifdef CONFIG_NETFILTER_INGRESS
960632ec 1839 struct nf_hook_entries __rcu *nf_hooks_ingress;
e687ad60 1840#endif
d2788d34 1841
536721b1 1842 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1843#ifdef CONFIG_RFS_ACCEL
1844 struct cpu_rmap *rx_cpu_rmap;
1845#endif
1846 struct hlist_node index_hlist;
cd13539b
ED
1847
1848/*
1849 * Cache lines mostly used on transmit path
1850 */
e8a0464c
DM
1851 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1852 unsigned int num_tx_queues;
fd2ea0a7 1853 unsigned int real_num_tx_queues;
af356afa 1854 struct Qdisc *qdisc;
59cc1f61
JK
1855#ifdef CONFIG_NET_SCHED
1856 DECLARE_HASHTABLE (qdisc_hash, 4);
1857#endif
0cd29503 1858 unsigned int tx_queue_len;
c3f26a26 1859 spinlock_t tx_global_lock;
14ffbbb8 1860 int watchdog_timeo;
cd13539b 1861
bf264145 1862#ifdef CONFIG_XPS
a4177869 1863 struct xps_dev_maps __rcu *xps_maps;
bf264145 1864#endif
1f211a1b 1865#ifdef CONFIG_NET_CLS_ACT
46209401 1866 struct mini_Qdisc __rcu *miniq_egress;
1f211a1b 1867#endif
0c4f691f 1868
9356b8fc 1869 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1870 struct timer_list watchdog_timer;
1871
29b4433d 1872 int __percpu *pcpu_refcnt;
1da177e4 1873 struct list_head todo_list;
1da177e4 1874
e014debe 1875 struct list_head link_watch_list;
572a103d 1876
1da177e4 1877 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1878 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1879 NETREG_UNREGISTERING, /* called unregister_netdevice */
1880 NETREG_UNREGISTERED, /* completed unregister todo */
1881 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1882 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1883 } reg_state:8;
1884
536721b1 1885 bool dismantle;
a2835763
PM
1886
1887 enum {
1888 RTNL_LINK_INITIALIZED,
1889 RTNL_LINK_INITIALIZING,
1890 } rtnl_link_state:16;
1da177e4 1891
cf124db5
DM
1892 bool needs_free_netdev;
1893 void (*priv_destructor)(struct net_device *dev);
1da177e4 1894
1da177e4 1895#ifdef CONFIG_NETPOLL
5fbee843 1896 struct netpoll_info __rcu *npinfo;
1da177e4 1897#endif
eae792b7 1898
0c5c9fb5 1899 possible_net_t nd_net;
4a1c5371 1900
4951704b 1901 /* mid-layer private */
a7855c78 1902 union {
536721b1
KK
1903 void *ml_priv;
1904 struct pcpu_lstats __percpu *lstats;
8f84985f 1905 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1906 struct pcpu_dstats __percpu *dstats;
1907 struct pcpu_vstats __percpu *vstats;
a7855c78 1908 };
536721b1 1909
fb585b44 1910#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1911 struct garp_port __rcu *garp_port;
fb585b44
TK
1912#endif
1913#if IS_ENABLED(CONFIG_MRP)
febf018d 1914 struct mrp_port __rcu *mrp_port;
fb585b44 1915#endif
1da177e4 1916
5e82b4b2 1917 struct device dev;
0c509a6c 1918 const struct attribute_group *sysfs_groups[4];
a953be53 1919 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1920
38f7b870 1921 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1922
82cc1a7a
PWJ
1923 /* for setting kernel sock attribute on TCP connection setup */
1924#define GSO_MAX_SIZE 65536
1925 unsigned int gso_max_size;
30b678d8
BH
1926#define GSO_MAX_SEGS 65535
1927 u16 gso_max_segs;
743b03a8 1928
7a6b6f51 1929#ifdef CONFIG_DCB
32953543 1930 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1931#endif
5e82b4b2
BH
1932 u8 num_tc;
1933 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1934 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1935
d11ead75 1936#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1937 unsigned int fcoe_ddp_xid;
5bc1421e 1938#endif
86f8515f 1939#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1940 struct netprio_map __rcu *priomap;
4d288d57 1941#endif
5e82b4b2
BH
1942 struct phy_device *phydev;
1943 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1944 struct lock_class_key *qdisc_running_key;
5e82b4b2 1945 bool proto_down;
1da177e4 1946};
43cb76d9 1947#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1948
b5cdae32
DM
1949static inline bool netif_elide_gro(const struct net_device *dev)
1950{
1951 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1952 return true;
1953 return false;
1954}
1955
1da177e4 1956#define NETDEV_ALIGN 32
1da177e4 1957
4f57c087
JF
1958static inline
1959int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1960{
1961 return dev->prio_tc_map[prio & TC_BITMASK];
1962}
1963
1964static inline
1965int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1966{
1967 if (tc >= dev->num_tc)
1968 return -EINVAL;
1969
1970 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1971 return 0;
1972}
1973
8d059b0f 1974int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1975void netdev_reset_tc(struct net_device *dev);
1976int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1977int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1978
1979static inline
1980int netdev_get_num_tc(struct net_device *dev)
1981{
1982 return dev->num_tc;
1983}
1984
e8a0464c
DM
1985static inline
1986struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1987 unsigned int index)
1988{
1989 return &dev->_tx[index];
1990}
1991
10c51b56
DB
1992static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1993 const struct sk_buff *skb)
1994{
1995 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1996}
1997
e8a0464c
DM
1998static inline void netdev_for_each_tx_queue(struct net_device *dev,
1999 void (*f)(struct net_device *,
2000 struct netdev_queue *,
2001 void *),
2002 void *arg)
2003{
2004 unsigned int i;
2005
2006 for (i = 0; i < dev->num_tx_queues; i++)
2007 f(dev, &dev->_tx[i], arg);
2008}
2009
d3fff6c4
ED
2010#define netdev_lockdep_set_classes(dev) \
2011{ \
2012 static struct lock_class_key qdisc_tx_busylock_key; \
2013 static struct lock_class_key qdisc_running_key; \
2014 static struct lock_class_key qdisc_xmit_lock_key; \
2015 static struct lock_class_key dev_addr_list_lock_key; \
2016 unsigned int i; \
2017 \
2018 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2019 (dev)->qdisc_running_key = &qdisc_running_key; \
2020 lockdep_set_class(&(dev)->addr_list_lock, \
2021 &dev_addr_list_lock_key); \
2022 for (i = 0; i < (dev)->num_tx_queues; i++) \
2023 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2024 &qdisc_xmit_lock_key); \
2025}
2026
f629d208 2027struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
2028 struct sk_buff *skb,
2029 void *accel_priv);
8c4c49df 2030
871b642a
PA
2031/* returns the headroom that the master device needs to take in account
2032 * when forwarding to this dev
2033 */
2034static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2035{
2036 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2037}
2038
2039static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2040{
2041 if (dev->netdev_ops->ndo_set_rx_headroom)
2042 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2043}
2044
2045/* set the device rx headroom to the dev's default */
2046static inline void netdev_reset_rx_headroom(struct net_device *dev)
2047{
2048 netdev_set_rx_headroom(dev, -1);
2049}
2050
c346dca1
YH
2051/*
2052 * Net namespace inlines
2053 */
2054static inline
2055struct net *dev_net(const struct net_device *dev)
2056{
c2d9ba9b 2057 return read_pnet(&dev->nd_net);
c346dca1
YH
2058}
2059
2060static inline
f5aa23fd 2061void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2062{
0c5c9fb5 2063 write_pnet(&dev->nd_net, net);
c346dca1
YH
2064}
2065
bea3348e
SH
2066/**
2067 * netdev_priv - access network device private data
2068 * @dev: network device
2069 *
2070 * Get network device private data
2071 */
6472ce60 2072static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2073{
1ce8e7b5 2074 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2075}
2076
1da177e4
LT
2077/* Set the sysfs physical device reference for the network logical device
2078 * if set prior to registration will cause a symlink during initialization.
2079 */
43cb76d9 2080#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2081
384912ed 2082/* Set the sysfs device type for the network logical device to allow
3f79410c 2083 * fine-grained identification of different network device types. For
5e82b4b2 2084 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2085 */
2086#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2087
82dc3c63
ED
2088/* Default NAPI poll() weight
2089 * Device drivers are strongly advised to not use bigger value
2090 */
2091#define NAPI_POLL_WEIGHT 64
2092
3b582cc1 2093/**
5e82b4b2 2094 * netif_napi_add - initialize a NAPI context
3b582cc1 2095 * @dev: network device
5e82b4b2 2096 * @napi: NAPI context
3b582cc1
SH
2097 * @poll: polling function
2098 * @weight: default weight
2099 *
5e82b4b2
BH
2100 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2101 * *any* of the other NAPI-related functions.
3b582cc1 2102 */
d565b0a1
HX
2103void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2104 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2105
d64b5e85 2106/**
5e82b4b2 2107 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2108 * @dev: network device
5e82b4b2 2109 * @napi: NAPI context
d64b5e85
ED
2110 * @poll: polling function
2111 * @weight: default weight
2112 *
2113 * This variant of netif_napi_add() should be used from drivers using NAPI
2114 * to exclusively poll a TX queue.
2115 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2116 */
2117static inline void netif_tx_napi_add(struct net_device *dev,
2118 struct napi_struct *napi,
2119 int (*poll)(struct napi_struct *, int),
2120 int weight)
2121{
2122 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2123 netif_napi_add(dev, napi, poll, weight);
2124}
2125
d8156534 2126/**
5e82b4b2
BH
2127 * netif_napi_del - remove a NAPI context
2128 * @napi: NAPI context
d8156534 2129 *
5e82b4b2 2130 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2131 */
d565b0a1
HX
2132void netif_napi_del(struct napi_struct *napi);
2133
2134struct napi_gro_cb {
78a478d0 2135 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2136 void *frag0;
78a478d0 2137
7489594c
HX
2138 /* Length of frag0. */
2139 unsigned int frag0_len;
2140
86911732 2141 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2142 int data_offset;
86911732 2143
d565b0a1 2144 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2145 u16 flush;
2146
2147 /* Save the IP ID here and check when we get to the transport layer */
2148 u16 flush_id;
d565b0a1
HX
2149
2150 /* Number of segments aggregated. */
2e71a6f8
ED
2151 u16 count;
2152
15e2396d
TH
2153 /* Start offset for remote checksum offload */
2154 u16 gro_remcsum_start;
2155
2e71a6f8
ED
2156 /* jiffies when first packet was created/queued */
2157 unsigned long age;
86347245 2158
afe93325 2159 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2160 u16 proto;
2161
baa32ff4
TH
2162 /* This is non-zero if the packet may be of the same flow. */
2163 u8 same_flow:1;
2164
fac8e0f5
JG
2165 /* Used in tunnel GRO receive */
2166 u8 encap_mark:1;
573e8fca
TH
2167
2168 /* GRO checksum is valid */
2169 u8 csum_valid:1;
2170
662880f4
TH
2171 /* Number of checksums via CHECKSUM_UNNECESSARY */
2172 u8 csum_cnt:3;
c3c7c254 2173
baa32ff4
TH
2174 /* Free the skb? */
2175 u8 free:2;
2176#define NAPI_GRO_FREE 1
2177#define NAPI_GRO_FREE_STOLEN_HEAD 2
2178
efc98d08
TH
2179 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2180 u8 is_ipv6:1;
2181
a0ca153f
AD
2182 /* Used in GRE, set in fou/gue_gro_receive */
2183 u8 is_fou:1;
2184
1530545e
AD
2185 /* Used to determine if flush_id can be ignored */
2186 u8 is_atomic:1;
2187
fcd91dd4
SD
2188 /* Number of gro_receive callbacks this packet already went through */
2189 u8 recursion_counter:4;
2190
2191 /* 1 bit hole */
baa32ff4 2192
bf5a755f
JC
2193 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2194 __wsum csum;
2195
c3c7c254
ED
2196 /* used in skb_gro_receive() slow path */
2197 struct sk_buff *last;
d565b0a1
HX
2198};
2199
2200#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2201
fcd91dd4
SD
2202#define GRO_RECURSION_LIMIT 15
2203static inline int gro_recursion_inc_test(struct sk_buff *skb)
2204{
2205 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2206}
2207
2208typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2209static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2210 struct sk_buff **head,
2211 struct sk_buff *skb)
2212{
2213 if (unlikely(gro_recursion_inc_test(skb))) {
2214 NAPI_GRO_CB(skb)->flush |= 1;
2215 return NULL;
2216 }
2217
2218 return cb(head, skb);
2219}
2220
2221typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2222 struct sk_buff *);
2223static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2224 struct sock *sk,
2225 struct sk_buff **head,
2226 struct sk_buff *skb)
2227{
2228 if (unlikely(gro_recursion_inc_test(skb))) {
2229 NAPI_GRO_CB(skb)->flush |= 1;
2230 return NULL;
2231 }
2232
2233 return cb(sk, head, skb);
2234}
2235
1da177e4 2236struct packet_type {
f2ccd8fa
DM
2237 __be16 type; /* This is really htons(ether_type). */
2238 struct net_device *dev; /* NULL is wildcarded here */
2239 int (*func) (struct sk_buff *,
2240 struct net_device *,
2241 struct packet_type *,
2242 struct net_device *);
c0de08d0
EL
2243 bool (*id_match)(struct packet_type *ptype,
2244 struct sock *sk);
1da177e4
LT
2245 void *af_packet_priv;
2246 struct list_head list;
2247};
2248
f191a1d1 2249struct offload_callbacks {
576a30eb 2250 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2251 netdev_features_t features);
d565b0a1 2252 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2253 struct sk_buff *skb);
299603e8 2254 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2255};
2256
2257struct packet_offload {
2258 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2259 u16 priority;
f191a1d1
VY
2260 struct offload_callbacks callbacks;
2261 struct list_head list;
1da177e4
LT
2262};
2263
5e82b4b2 2264/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2265struct pcpu_sw_netstats {
2266 u64 rx_packets;
2267 u64 rx_bytes;
2268 u64 tx_packets;
2269 u64 tx_bytes;
2270 struct u64_stats_sync syncp;
2271};
2272
aabc92bb
PNA
2273#define __netdev_alloc_pcpu_stats(type, gfp) \
2274({ \
2275 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2276 if (pcpu_stats) { \
2277 int __cpu; \
2278 for_each_possible_cpu(__cpu) { \
2279 typeof(type) *stat; \
2280 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2281 u64_stats_init(&stat->syncp); \
2282 } \
2283 } \
2284 pcpu_stats; \
1c213bd2
WC
2285})
2286
aabc92bb 2287#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2288 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2289
764f5e54
JP
2290enum netdev_lag_tx_type {
2291 NETDEV_LAG_TX_TYPE_UNKNOWN,
2292 NETDEV_LAG_TX_TYPE_RANDOM,
2293 NETDEV_LAG_TX_TYPE_BROADCAST,
2294 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2295 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2296 NETDEV_LAG_TX_TYPE_HASH,
2297};
2298
2299struct netdev_lag_upper_info {
2300 enum netdev_lag_tx_type tx_type;
2301};
2302
fb1b2e3c
JP
2303struct netdev_lag_lower_state_info {
2304 u8 link_up : 1,
2305 tx_enabled : 1;
2306};
2307
1da177e4
LT
2308#include <linux/notifier.h>
2309
dcfe1421
AW
2310/* netdevice notifier chain. Please remember to update the rtnetlink
2311 * notification exclusion list in rtnetlink_event() when adding new
2312 * types.
2313 */
2314#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2315#define NETDEV_DOWN 0x0002
2316#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2317 detected a hardware crash and restarted
2318 - we can use this eg to kick tcp sessions
2319 once done */
2320#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2321#define NETDEV_REGISTER 0x0005
2322#define NETDEV_UNREGISTER 0x0006
1d486bfb 2323#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2324#define NETDEV_CHANGEADDR 0x0008
2325#define NETDEV_GOING_DOWN 0x0009
2326#define NETDEV_CHANGENAME 0x000A
2327#define NETDEV_FEAT_CHANGE 0x000B
2328#define NETDEV_BONDING_FAILOVER 0x000C
2329#define NETDEV_PRE_UP 0x000D
2330#define NETDEV_PRE_TYPE_CHANGE 0x000E
2331#define NETDEV_POST_TYPE_CHANGE 0x000F
2332#define NETDEV_POST_INIT 0x0010
0115e8e3 2333#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2334#define NETDEV_RELEASE 0x0012
2335#define NETDEV_NOTIFY_PEERS 0x0013
2336#define NETDEV_JOIN 0x0014
42e52bf9 2337#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2338#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2339#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2340#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2341#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2342#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2343#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2344#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
296d8ee3 2345#define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
08294a26 2346#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2347
f629d208
JP
2348int register_netdevice_notifier(struct notifier_block *nb);
2349int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2350
2351struct netdev_notifier_info {
51d0c047
DA
2352 struct net_device *dev;
2353 struct netlink_ext_ack *extack;
351638e7
JP
2354};
2355
be9efd36
JP
2356struct netdev_notifier_change_info {
2357 struct netdev_notifier_info info; /* must be first */
2358 unsigned int flags_changed;
2359};
2360
0e4ead9d
JP
2361struct netdev_notifier_changeupper_info {
2362 struct netdev_notifier_info info; /* must be first */
2363 struct net_device *upper_dev; /* new upper dev */
2364 bool master; /* is upper dev master */
5e82b4b2 2365 bool linking; /* is the notification for link or unlink */
29bf24af 2366 void *upper_info; /* upper dev info */
0e4ead9d
JP
2367};
2368
04d48266
JP
2369struct netdev_notifier_changelowerstate_info {
2370 struct netdev_notifier_info info; /* must be first */
2371 void *lower_state_info; /* is lower dev state */
2372};
2373
75538c2b
CW
2374static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2375 struct net_device *dev)
2376{
2377 info->dev = dev;
51d0c047 2378 info->extack = NULL;
75538c2b
CW
2379}
2380
351638e7
JP
2381static inline struct net_device *
2382netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2383{
2384 return info->dev;
2385}
2386
51d0c047
DA
2387static inline struct netlink_ext_ack *
2388netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2389{
2390 return info->extack;
2391}
2392
f629d208 2393int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2394
2395
1da177e4
LT
2396extern rwlock_t dev_base_lock; /* Device list lock */
2397
881d966b
EB
2398#define for_each_netdev(net, d) \
2399 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2400#define for_each_netdev_reverse(net, d) \
2401 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2402#define for_each_netdev_rcu(net, d) \
2403 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2404#define for_each_netdev_safe(net, d, n) \
2405 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2406#define for_each_netdev_continue(net, d) \
2407 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2408#define for_each_netdev_continue_rcu(net, d) \
2409 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2410#define for_each_netdev_in_bond_rcu(bond, slave) \
2411 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2412 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2413#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2414
a050c33f
DL
2415static inline struct net_device *next_net_device(struct net_device *dev)
2416{
2417 struct list_head *lh;
2418 struct net *net;
2419
c346dca1 2420 net = dev_net(dev);
a050c33f
DL
2421 lh = dev->dev_list.next;
2422 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2423}
2424
ce81b76a
ED
2425static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2426{
2427 struct list_head *lh;
2428 struct net *net;
2429
2430 net = dev_net(dev);
ccf43438 2431 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2432 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2433}
2434
a050c33f
DL
2435static inline struct net_device *first_net_device(struct net *net)
2436{
2437 return list_empty(&net->dev_base_head) ? NULL :
2438 net_device_entry(net->dev_base_head.next);
2439}
7562f876 2440
ccf43438
ED
2441static inline struct net_device *first_net_device_rcu(struct net *net)
2442{
2443 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2444
2445 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2446}
2447
f629d208
JP
2448int netdev_boot_setup_check(struct net_device *dev);
2449unsigned long netdev_boot_base(const char *prefix, int unit);
2450struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2451 const char *hwaddr);
2452struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2453struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2454void dev_add_pack(struct packet_type *pt);
2455void dev_remove_pack(struct packet_type *pt);
2456void __dev_remove_pack(struct packet_type *pt);
2457void dev_add_offload(struct packet_offload *po);
2458void dev_remove_offload(struct packet_offload *po);
f629d208 2459
a54acb3a 2460int dev_get_iflink(const struct net_device *dev);
fc4099f1 2461int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2462struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2463 unsigned short mask);
f629d208
JP
2464struct net_device *dev_get_by_name(struct net *net, const char *name);
2465struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2466struct net_device *__dev_get_by_name(struct net *net, const char *name);
2467int dev_alloc_name(struct net_device *dev, const char *name);
2468int dev_open(struct net_device *dev);
7051b88a 2469void dev_close(struct net_device *dev);
2470void dev_close_many(struct list_head *head, bool unlink);
f629d208 2471void dev_disable_lro(struct net_device *dev);
0c4b51f0 2472int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2473int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2474int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2475int register_netdevice(struct net_device *dev);
2476void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2477void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2478static inline void unregister_netdevice(struct net_device *dev)
2479{
2480 unregister_netdevice_queue(dev, NULL);
2481}
2482
f629d208
JP
2483int netdev_refcnt_read(const struct net_device *dev);
2484void free_netdev(struct net_device *dev);
74d332c1 2485void netdev_freemem(struct net_device *dev);
f629d208
JP
2486void synchronize_net(void);
2487int init_dummy_netdev(struct net_device *dev);
937f1ba5 2488
f60e5990 2489DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2490#define XMIT_RECURSION_LIMIT 10
2491
f60e5990 2492static inline int dev_recursion_level(void)
2493{
2494 return this_cpu_read(xmit_recursion);
2495}
2496
f629d208
JP
2497struct net_device *dev_get_by_index(struct net *net, int ifindex);
2498struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2499struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2500struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2501int netdev_get_name(struct net *net, char *name, int ifindex);
2502int dev_restart(struct net_device *dev);
f629d208 2503int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2504
2505static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2506{
2507 return NAPI_GRO_CB(skb)->data_offset;
2508}
2509
2510static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2511{
2512 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2513}
2514
2515static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2516{
2517 NAPI_GRO_CB(skb)->data_offset += len;
2518}
2519
a5b1cf28
HX
2520static inline void *skb_gro_header_fast(struct sk_buff *skb,
2521 unsigned int offset)
86911732 2522{
a5b1cf28
HX
2523 return NAPI_GRO_CB(skb)->frag0 + offset;
2524}
78a478d0 2525
a5b1cf28
HX
2526static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2527{
2528 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2529}
78a478d0 2530
57ea52a8
HX
2531static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2532{
2533 NAPI_GRO_CB(skb)->frag0 = NULL;
2534 NAPI_GRO_CB(skb)->frag0_len = 0;
2535}
2536
a5b1cf28
HX
2537static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2538 unsigned int offset)
2539{
17dd759c
HX
2540 if (!pskb_may_pull(skb, hlen))
2541 return NULL;
2542
57ea52a8 2543 skb_gro_frag0_invalidate(skb);
17dd759c 2544 return skb->data + offset;
86911732 2545}
1da177e4 2546
36e7b1b8
HX
2547static inline void *skb_gro_network_header(struct sk_buff *skb)
2548{
78d3fd0b
HX
2549 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2550 skb_network_offset(skb);
36e7b1b8
HX
2551}
2552
bf5a755f
JC
2553static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2554 const void *start, unsigned int len)
2555{
573e8fca 2556 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2557 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2558 csum_partial(start, len, 0));
2559}
2560
573e8fca
TH
2561/* GRO checksum functions. These are logical equivalents of the normal
2562 * checksum functions (in skbuff.h) except that they operate on the GRO
2563 * offsets and fields in sk_buff.
2564 */
2565
2566__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2567
15e2396d
TH
2568static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2569{
b7fe10e5 2570 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2571}
2572
573e8fca
TH
2573static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2574 bool zero_okay,
2575 __sum16 check)
2576{
6edec0e6
TH
2577 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2578 skb_checksum_start_offset(skb) <
2579 skb_gro_offset(skb)) &&
15e2396d 2580 !skb_at_gro_remcsum_start(skb) &&
662880f4 2581 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2582 (!zero_okay || check));
2583}
2584
2585static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2586 __wsum psum)
2587{
2588 if (NAPI_GRO_CB(skb)->csum_valid &&
2589 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2590 return 0;
2591
2592 NAPI_GRO_CB(skb)->csum = psum;
2593
2594 return __skb_gro_checksum_complete(skb);
2595}
2596
573e8fca
TH
2597static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2598{
662880f4
TH
2599 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2600 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2601 NAPI_GRO_CB(skb)->csum_cnt--;
2602 } else {
2603 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2604 * verified a new top level checksum or an encapsulated one
2605 * during GRO. This saves work if we fallback to normal path.
2606 */
2607 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2608 }
2609}
2610
2611#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2612 compute_pseudo) \
2613({ \
2614 __sum16 __ret = 0; \
2615 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2616 __ret = __skb_gro_checksum_validate_complete(skb, \
2617 compute_pseudo(skb, proto)); \
219f1d79 2618 if (!__ret) \
573e8fca
TH
2619 skb_gro_incr_csum_unnecessary(skb); \
2620 __ret; \
2621})
2622
2623#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2624 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2625
2626#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2627 compute_pseudo) \
2628 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2629
2630#define skb_gro_checksum_simple_validate(skb) \
2631 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2632
d96535a1
TH
2633static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2634{
2635 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2636 !NAPI_GRO_CB(skb)->csum_valid);
2637}
2638
2639static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2640 __sum16 check, __wsum pseudo)
2641{
2642 NAPI_GRO_CB(skb)->csum = ~pseudo;
2643 NAPI_GRO_CB(skb)->csum_valid = 1;
2644}
2645
2646#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2647do { \
2648 if (__skb_gro_checksum_convert_check(skb)) \
2649 __skb_gro_checksum_convert(skb, check, \
2650 compute_pseudo(skb, proto)); \
2651} while (0)
2652
26c4f7da
TH
2653struct gro_remcsum {
2654 int offset;
2655 __wsum delta;
2656};
2657
2658static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2659{
846cd667 2660 grc->offset = 0;
26c4f7da
TH
2661 grc->delta = 0;
2662}
2663
b7fe10e5
TH
2664static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2665 unsigned int off, size_t hdrlen,
2666 int start, int offset,
2667 struct gro_remcsum *grc,
2668 bool nopartial)
dcdc8994
TH
2669{
2670 __wsum delta;
b7fe10e5 2671 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2672
2673 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2674
15e2396d 2675 if (!nopartial) {
b7fe10e5
TH
2676 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2677 return ptr;
2678 }
2679
2680 ptr = skb_gro_header_fast(skb, off);
2681 if (skb_gro_header_hard(skb, off + plen)) {
2682 ptr = skb_gro_header_slow(skb, off + plen, off);
2683 if (!ptr)
2684 return NULL;
15e2396d
TH
2685 }
2686
b7fe10e5
TH
2687 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2688 start, offset);
dcdc8994
TH
2689
2690 /* Adjust skb->csum since we changed the packet */
dcdc8994 2691 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2692
b7fe10e5 2693 grc->offset = off + hdrlen + offset;
26c4f7da 2694 grc->delta = delta;
b7fe10e5
TH
2695
2696 return ptr;
dcdc8994
TH
2697}
2698
26c4f7da
TH
2699static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2700 struct gro_remcsum *grc)
2701{
b7fe10e5
TH
2702 void *ptr;
2703 size_t plen = grc->offset + sizeof(u16);
2704
26c4f7da
TH
2705 if (!grc->delta)
2706 return;
2707
b7fe10e5
TH
2708 ptr = skb_gro_header_fast(skb, grc->offset);
2709 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2710 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2711 if (!ptr)
2712 return;
2713 }
2714
2715 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2716}
dcdc8994 2717
25393d3f
SK
2718#ifdef CONFIG_XFRM_OFFLOAD
2719static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2720{
2721 if (PTR_ERR(pp) != -EINPROGRESS)
2722 NAPI_GRO_CB(skb)->flush |= flush;
2723}
2724#else
5f114163
SK
2725static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2726{
2727 NAPI_GRO_CB(skb)->flush |= flush;
2728}
25393d3f 2729#endif
5f114163 2730
0c4e8581
SH
2731static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2732 unsigned short type,
3b04ddde 2733 const void *daddr, const void *saddr,
95c96174 2734 unsigned int len)
0c4e8581 2735{
f1ecfd5d 2736 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2737 return 0;
3b04ddde
SH
2738
2739 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2740}
2741
b95cce35
SH
2742static inline int dev_parse_header(const struct sk_buff *skb,
2743 unsigned char *haddr)
2744{
2745 const struct net_device *dev = skb->dev;
2746
1b83336b 2747 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2748 return 0;
3b04ddde 2749 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2750}
2751
2793a23a
WB
2752/* ll_header must have at least hard_header_len allocated */
2753static inline bool dev_validate_header(const struct net_device *dev,
2754 char *ll_header, int len)
2755{
2756 if (likely(len >= dev->hard_header_len))
2757 return true;
217e6fa2
WB
2758 if (len < dev->min_header_len)
2759 return false;
2793a23a
WB
2760
2761 if (capable(CAP_SYS_RAWIO)) {
2762 memset(ll_header + len, 0, dev->hard_header_len - len);
2763 return true;
2764 }
2765
2766 if (dev->header_ops && dev->header_ops->validate)
2767 return dev->header_ops->validate(ll_header, len);
2768
2769 return false;
2770}
2771
36fd633e
AV
2772typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2773 int len, int size);
f629d208 2774int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2775static inline int unregister_gifconf(unsigned int family)
2776{
2777 return register_gifconf(family, NULL);
2778}
2779
99bbc707 2780#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2781#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2782struct sd_flow_limit {
2783 u64 count;
2784 unsigned int num_buckets;
2785 unsigned int history_head;
2786 u16 history[FLOW_LIMIT_HISTORY];
2787 u8 buckets[];
2788};
2789
2790extern int netdev_flow_limit_table_len;
2791#endif /* CONFIG_NET_FLOW_LIMIT */
2792
1da177e4 2793/*
5e82b4b2 2794 * Incoming packets are placed on per-CPU queues
1da177e4 2795 */
d94d9fee 2796struct softnet_data {
1da177e4 2797 struct list_head poll_list;
6e7676c1 2798 struct sk_buff_head process_queue;
1da177e4 2799
dee42870 2800 /* stats */
cd7b5396
DM
2801 unsigned int processed;
2802 unsigned int time_squeeze;
cd7b5396 2803 unsigned int received_rps;
fd793d89 2804#ifdef CONFIG_RPS
88751275 2805 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2806#endif
2807#ifdef CONFIG_NET_FLOW_LIMIT
2808 struct sd_flow_limit __rcu *flow_limit;
2809#endif
2810 struct Qdisc *output_queue;
2811 struct Qdisc **output_queue_tailp;
2812 struct sk_buff *completion_queue;
f53c7239
SK
2813#ifdef CONFIG_XFRM_OFFLOAD
2814 struct sk_buff_head xfrm_backlog;
2815#endif
4cdb1e2e 2816#ifdef CONFIG_RPS
501e7ef5
ED
2817 /* input_queue_head should be written by cpu owning this struct,
2818 * and only read by other cpus. Worth using a cache line.
2819 */
2820 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2821
2822 /* Elements below can be accessed between CPUs for RPS/RFS */
966a9671 2823 call_single_data_t csd ____cacheline_aligned_in_smp;
88751275
ED
2824 struct softnet_data *rps_ipi_next;
2825 unsigned int cpu;
76cc8b13 2826 unsigned int input_queue_tail;
1e94d72f 2827#endif
95c96174 2828 unsigned int dropped;
0a9627f2 2829 struct sk_buff_head input_pkt_queue;
bea3348e 2830 struct napi_struct backlog;
99bbc707 2831
1da177e4
LT
2832};
2833
76cc8b13 2834static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2835{
2836#ifdef CONFIG_RPS
76cc8b13
TH
2837 sd->input_queue_head++;
2838#endif
2839}
2840
2841static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2842 unsigned int *qtail)
2843{
2844#ifdef CONFIG_RPS
2845 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2846#endif
2847}
2848
0a9627f2 2849DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2850
f629d208 2851void __netif_schedule(struct Qdisc *q);
46e5da40 2852void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2853
fd2ea0a7
DM
2854static inline void netif_tx_schedule_all(struct net_device *dev)
2855{
2856 unsigned int i;
2857
2858 for (i = 0; i < dev->num_tx_queues; i++)
2859 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2860}
2861
f9a7cbbf 2862static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2863{
73466498 2864 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2865}
2866
bea3348e
SH
2867/**
2868 * netif_start_queue - allow transmit
2869 * @dev: network device
2870 *
2871 * Allow upper layers to call the device hard_start_xmit routine.
2872 */
1da177e4
LT
2873static inline void netif_start_queue(struct net_device *dev)
2874{
e8a0464c 2875 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2876}
2877
fd2ea0a7
DM
2878static inline void netif_tx_start_all_queues(struct net_device *dev)
2879{
2880 unsigned int i;
2881
2882 for (i = 0; i < dev->num_tx_queues; i++) {
2883 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2884 netif_tx_start_queue(txq);
2885 }
2886}
2887
46e5da40 2888void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2889
d29f749e
DJ
2890/**
2891 * netif_wake_queue - restart transmit
2892 * @dev: network device
2893 *
2894 * Allow upper layers to call the device hard_start_xmit routine.
2895 * Used for flow control when transmit resources are available.
2896 */
79d16385
DM
2897static inline void netif_wake_queue(struct net_device *dev)
2898{
e8a0464c 2899 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2900}
2901
fd2ea0a7
DM
2902static inline void netif_tx_wake_all_queues(struct net_device *dev)
2903{
2904 unsigned int i;
2905
2906 for (i = 0; i < dev->num_tx_queues; i++) {
2907 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2908 netif_tx_wake_queue(txq);
2909 }
2910}
2911
f9a7cbbf 2912static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2913{
73466498 2914 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2915}
2916
bea3348e
SH
2917/**
2918 * netif_stop_queue - stop transmitted packets
2919 * @dev: network device
2920 *
2921 * Stop upper layers calling the device hard_start_xmit routine.
2922 * Used for flow control when transmit resources are unavailable.
2923 */
1da177e4
LT
2924static inline void netif_stop_queue(struct net_device *dev)
2925{
e8a0464c 2926 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2927}
2928
a2029240 2929void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2930
4d29515f 2931static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2932{
73466498 2933 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2934}
2935
bea3348e
SH
2936/**
2937 * netif_queue_stopped - test if transmit queue is flowblocked
2938 * @dev: network device
2939 *
2940 * Test if transmit queue on device is currently unable to send.
2941 */
4d29515f 2942static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2943{
e8a0464c 2944 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2945}
2946
4d29515f 2947static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2948{
73466498
TH
2949 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2950}
2951
8e2f1a63
DB
2952static inline bool
2953netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2954{
2955 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2956}
2957
8e2f1a63
DB
2958static inline bool
2959netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2960{
2961 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2962}
2963
53511453
ED
2964/**
2965 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2966 * @dev_queue: pointer to transmit queue
2967 *
2968 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2969 * to give appropriate hint to the CPU.
53511453
ED
2970 */
2971static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2972{
2973#ifdef CONFIG_BQL
2974 prefetchw(&dev_queue->dql.num_queued);
2975#endif
2976}
2977
2978/**
2979 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2980 * @dev_queue: pointer to transmit queue
2981 *
2982 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2983 * to give appropriate hint to the CPU.
53511453
ED
2984 */
2985static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2986{
2987#ifdef CONFIG_BQL
2988 prefetchw(&dev_queue->dql.limit);
2989#endif
2990}
2991
c5d67bd7
TH
2992static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2993 unsigned int bytes)
2994{
114cf580
TH
2995#ifdef CONFIG_BQL
2996 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2997
2998 if (likely(dql_avail(&dev_queue->dql) >= 0))
2999 return;
3000
3001 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3002
3003 /*
3004 * The XOFF flag must be set before checking the dql_avail below,
3005 * because in netdev_tx_completed_queue we update the dql_completed
3006 * before checking the XOFF flag.
3007 */
3008 smp_mb();
3009
3010 /* check again in case another CPU has just made room avail */
3011 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3012 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 3013#endif
c5d67bd7
TH
3014}
3015
0042d0c8
FF
3016/**
3017 * netdev_sent_queue - report the number of bytes queued to hardware
3018 * @dev: network device
3019 * @bytes: number of bytes queued to the hardware device queue
3020 *
3021 * Report the number of bytes queued for sending/completion to the network
3022 * device hardware queue. @bytes should be a good approximation and should
3023 * exactly match netdev_completed_queue() @bytes
3024 */
c5d67bd7
TH
3025static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3026{
3027 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3028}
3029
3030static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 3031 unsigned int pkts, unsigned int bytes)
c5d67bd7 3032{
114cf580 3033#ifdef CONFIG_BQL
b37c0fbe
AD
3034 if (unlikely(!bytes))
3035 return;
3036
3037 dql_completed(&dev_queue->dql, bytes);
3038
3039 /*
3040 * Without the memory barrier there is a small possiblity that
3041 * netdev_tx_sent_queue will miss the update and cause the queue to
3042 * be stopped forever
3043 */
3044 smp_mb();
3045
3046 if (dql_avail(&dev_queue->dql) < 0)
3047 return;
3048
3049 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3050 netif_schedule_queue(dev_queue);
114cf580 3051#endif
c5d67bd7
TH
3052}
3053
0042d0c8
FF
3054/**
3055 * netdev_completed_queue - report bytes and packets completed by device
3056 * @dev: network device
3057 * @pkts: actual number of packets sent over the medium
3058 * @bytes: actual number of bytes sent over the medium
3059 *
3060 * Report the number of bytes and packets transmitted by the network device
3061 * hardware queue over the physical medium, @bytes must exactly match the
3062 * @bytes amount passed to netdev_sent_queue()
3063 */
c5d67bd7 3064static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3065 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3066{
3067 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3068}
3069
3070static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3071{
114cf580 3072#ifdef CONFIG_BQL
5c490354 3073 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3074 dql_reset(&q->dql);
3075#endif
c5d67bd7
TH
3076}
3077
0042d0c8
FF
3078/**
3079 * netdev_reset_queue - reset the packets and bytes count of a network device
3080 * @dev_queue: network device
3081 *
3082 * Reset the bytes and packet count of a network device and clear the
3083 * software flow control OFF bit for this network device
3084 */
c5d67bd7
TH
3085static inline void netdev_reset_queue(struct net_device *dev_queue)
3086{
3087 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3088}
3089
b9507bda
DB
3090/**
3091 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3092 * @dev: network device
3093 * @queue_index: given tx queue index
3094 *
3095 * Returns 0 if given tx queue index >= number of device tx queues,
3096 * otherwise returns the originally passed tx queue index.
3097 */
3098static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3099{
3100 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3101 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3102 dev->name, queue_index,
3103 dev->real_num_tx_queues);
3104 return 0;
3105 }
3106
3107 return queue_index;
3108}
3109
bea3348e
SH
3110/**
3111 * netif_running - test if up
3112 * @dev: network device
3113 *
3114 * Test if the device has been brought up.
3115 */
4d29515f 3116static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3117{
3118 return test_bit(__LINK_STATE_START, &dev->state);
3119}
3120
f25f4e44 3121/*
5e82b4b2 3122 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3123 * stop, and a check if it's stopped. All other device management is
3124 * done at the overall netdevice level.
3125 * Also test the device if we're multiqueue.
3126 */
bea3348e
SH
3127
3128/**
3129 * netif_start_subqueue - allow sending packets on subqueue
3130 * @dev: network device
3131 * @queue_index: sub queue index
3132 *
3133 * Start individual transmit queue of a device with multiple transmit queues.
3134 */
f25f4e44
PWJ
3135static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3136{
fd2ea0a7 3137 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3138
3139 netif_tx_start_queue(txq);
f25f4e44
PWJ
3140}
3141
bea3348e
SH
3142/**
3143 * netif_stop_subqueue - stop sending packets on subqueue
3144 * @dev: network device
3145 * @queue_index: sub queue index
3146 *
3147 * Stop individual transmit queue of a device with multiple transmit queues.
3148 */
f25f4e44
PWJ
3149static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3150{
fd2ea0a7 3151 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3152 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3153}
3154
bea3348e
SH
3155/**
3156 * netif_subqueue_stopped - test status of subqueue
3157 * @dev: network device
3158 * @queue_index: sub queue index
3159 *
3160 * Check individual transmit queue of a device with multiple transmit queues.
3161 */
4d29515f
DM
3162static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3163 u16 queue_index)
f25f4e44 3164{
fd2ea0a7 3165 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3166
3167 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3168}
3169
4d29515f
DM
3170static inline bool netif_subqueue_stopped(const struct net_device *dev,
3171 struct sk_buff *skb)
668f895a
PE
3172{
3173 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3174}
bea3348e 3175
738b35cc
FF
3176/**
3177 * netif_wake_subqueue - allow sending packets on subqueue
3178 * @dev: network device
3179 * @queue_index: sub queue index
3180 *
3181 * Resume individual transmit queue of a device with multiple transmit queues.
3182 */
3183static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3184{
3185 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3186
3187 netif_tx_wake_queue(txq);
3188}
f25f4e44 3189
537c00de 3190#ifdef CONFIG_XPS
53af53ae 3191int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3192 u16 index);
537c00de
AD
3193#else
3194static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3195 const struct cpumask *mask,
537c00de
AD
3196 u16 index)
3197{
3198 return 0;
3199}
3200#endif
3201
5605c762
JP
3202u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3203 unsigned int num_tx_queues);
3204
a3d22a68
VZ
3205/*
3206 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3207 * as a distribution range limit for the returned value.
3208 */
3209static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3210 struct sk_buff *skb)
a3d22a68
VZ
3211{
3212 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3213}
3214
bea3348e
SH
3215/**
3216 * netif_is_multiqueue - test if device has multiple transmit queues
3217 * @dev: network device
3218 *
3219 * Check if device has multiple transmit queues
bea3348e 3220 */
4d29515f 3221static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3222{
a02cec21 3223 return dev->num_tx_queues > 1;
f25f4e44 3224}
1da177e4 3225
f629d208 3226int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3227
a953be53 3228#ifdef CONFIG_SYSFS
f629d208 3229int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3230#else
3231static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3232 unsigned int rxq)
3233{
3234 return 0;
3235}
3236#endif
3237
65073a67
DB
3238static inline struct netdev_rx_queue *
3239__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3240{
3241 return dev->_rx + rxq;
3242}
3243
a953be53
MD
3244#ifdef CONFIG_SYSFS
3245static inline unsigned int get_netdev_rx_queue_index(
3246 struct netdev_rx_queue *queue)
3247{
3248 struct net_device *dev = queue->dev;
3249 int index = queue - dev->_rx;
3250
3251 BUG_ON(index >= dev->num_rx_queues);
3252 return index;
3253}
3254#endif
3255
16917b87 3256#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3257int netif_get_num_default_rss_queues(void);
16917b87 3258
e6247027
ED
3259enum skb_free_reason {
3260 SKB_REASON_CONSUMED,
3261 SKB_REASON_DROPPED,
3262};
3263
3264void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3265void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3266
e6247027
ED
3267/*
3268 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3269 * interrupt context or with hardware interrupts being disabled.
3270 * (in_irq() || irqs_disabled())
3271 *
3272 * We provide four helpers that can be used in following contexts :
3273 *
3274 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3275 * replacing kfree_skb(skb)
3276 *
3277 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3278 * Typically used in place of consume_skb(skb) in TX completion path
3279 *
3280 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3281 * replacing kfree_skb(skb)
3282 *
3283 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3284 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3285 */
e6247027
ED
3286static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3287{
3288 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3289}
3290
3291static inline void dev_consume_skb_irq(struct sk_buff *skb)
3292{
3293 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3294}
3295
3296static inline void dev_kfree_skb_any(struct sk_buff *skb)
3297{
3298 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3299}
3300
3301static inline void dev_consume_skb_any(struct sk_buff *skb)
3302{
3303 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3304}
1da177e4 3305
7c497478
JW
3306void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3307int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
f629d208
JP
3308int netif_rx(struct sk_buff *skb);
3309int netif_rx_ni(struct sk_buff *skb);
04eb4489 3310int netif_receive_skb(struct sk_buff *skb);
1c601d82 3311int netif_receive_skb_core(struct sk_buff *skb);
f629d208
JP
3312gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3313void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3314struct sk_buff *napi_get_frags(struct napi_struct *napi);
3315gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3316struct packet_offload *gro_find_receive_by_type(__be16 type);
3317struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3318
3319static inline void napi_free_frags(struct napi_struct *napi)
3320{
3321 kfree_skb(napi->skb);
3322 napi->skb = NULL;
3323}
3324
24b27fc4 3325bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3326int netdev_rx_handler_register(struct net_device *dev,
3327 rx_handler_func_t *rx_handler,
3328 void *rx_handler_data);
3329void netdev_rx_handler_unregister(struct net_device *dev);
3330
3331bool dev_valid_name(const char *name);
44c02a2c
AV
3332int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3333 bool *need_copyout);
36fd633e 3334int dev_ifconf(struct net *net, struct ifconf *, int);
f629d208
JP
3335int dev_ethtool(struct net *net, struct ifreq *);
3336unsigned int dev_get_flags(const struct net_device *);
3337int __dev_change_flags(struct net_device *, unsigned int flags);
3338int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3339void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3340 unsigned int gchanges);
f629d208
JP
3341int dev_change_name(struct net_device *, const char *);
3342int dev_set_alias(struct net_device *, const char *, size_t);
6c557001 3343int dev_get_alias(const struct net_device *, char *, size_t);
f629d208 3344int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3345int __dev_set_mtu(struct net_device *, int);
f629d208 3346int dev_set_mtu(struct net_device *, int);
6a643ddb 3347int dev_change_tx_queue_len(struct net_device *, unsigned long);
f629d208
JP
3348void dev_set_group(struct net_device *, int);
3349int dev_set_mac_address(struct net_device *, struct sockaddr *);
3350int dev_change_carrier(struct net_device *, bool new_carrier);
3351int dev_get_phys_port_id(struct net_device *dev,
02637fce 3352 struct netdev_phys_item_id *ppid);
db24a904
DA
3353int dev_get_phys_port_name(struct net_device *dev,
3354 char *name, size_t len);
d746d707 3355int dev_change_proto_down(struct net_device *dev, bool proto_down);
f53c7239 3356struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
ce93718f
DM
3357struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3358 struct netdev_queue *txq, int *ret);
d67b9cd2 3359
f4e63525 3360typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
d67b9cd2
DB
3361int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3362 int fd, u32 flags);
118b4aa2
JK
3363void __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3364 struct netdev_bpf *xdp);
d67b9cd2 3365
a0265d28 3366int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3367int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3368bool is_skb_forwardable(const struct net_device *dev,
3369 const struct sk_buff *skb);
1da177e4 3370
4e3264d2
MKL
3371static __always_inline int ____dev_forward_skb(struct net_device *dev,
3372 struct sk_buff *skb)
3373{
3374 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3375 unlikely(!is_skb_forwardable(dev, skb))) {
3376 atomic_long_inc(&dev->rx_dropped);
3377 kfree_skb(skb);
3378 return NET_RX_DROP;
3379 }
3380
3381 skb_scrub_packet(skb, true);
3382 skb->priority = 0;
3383 return 0;
3384}
3385
74b20582
DA
3386void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3387
20380731 3388extern int netdev_budget;
7acf8a1e 3389extern unsigned int netdev_budget_usecs;
1da177e4
LT
3390
3391/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3392void netdev_run_todo(void);
1da177e4 3393
bea3348e
SH
3394/**
3395 * dev_put - release reference to device
3396 * @dev: network device
3397 *
9ef4429b 3398 * Release reference to device to allow it to be freed.
bea3348e 3399 */
1da177e4
LT
3400static inline void dev_put(struct net_device *dev)
3401{
933393f5 3402 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3403}
3404
bea3348e
SH
3405/**
3406 * dev_hold - get reference to device
3407 * @dev: network device
3408 *
9ef4429b 3409 * Hold reference to device to keep it from being freed.
bea3348e 3410 */
15333061
SH
3411static inline void dev_hold(struct net_device *dev)
3412{
933393f5 3413 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3414}
1da177e4
LT
3415
3416/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3417 * and _off may be called from IRQ context, but it is caller
3418 * who is responsible for serialization of these calls.
b00055aa
SR
3419 *
3420 * The name carrier is inappropriate, these functions should really be
3421 * called netif_lowerlayer_*() because they represent the state of any
3422 * kind of lower layer not just hardware media.
1da177e4
LT
3423 */
3424
f629d208
JP
3425void linkwatch_init_dev(struct net_device *dev);
3426void linkwatch_fire_event(struct net_device *dev);
3427void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3428
bea3348e
SH
3429/**
3430 * netif_carrier_ok - test if carrier present
3431 * @dev: network device
3432 *
3433 * Check if carrier is present on device
3434 */
4d29515f 3435static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3436{
3437 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3438}
3439
f629d208 3440unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3441
f629d208 3442void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3443
f629d208 3444void netif_carrier_on(struct net_device *dev);
1da177e4 3445
f629d208 3446void netif_carrier_off(struct net_device *dev);
1da177e4 3447
bea3348e
SH
3448/**
3449 * netif_dormant_on - mark device as dormant.
3450 * @dev: network device
3451 *
3452 * Mark device as dormant (as per RFC2863).
3453 *
3454 * The dormant state indicates that the relevant interface is not
3455 * actually in a condition to pass packets (i.e., it is not 'up') but is
3456 * in a "pending" state, waiting for some external event. For "on-
3457 * demand" interfaces, this new state identifies the situation where the
3458 * interface is waiting for events to place it in the up state.
bea3348e 3459 */
b00055aa
SR
3460static inline void netif_dormant_on(struct net_device *dev)
3461{
3462 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3463 linkwatch_fire_event(dev);
3464}
3465
bea3348e
SH
3466/**
3467 * netif_dormant_off - set device as not dormant.
3468 * @dev: network device
3469 *
3470 * Device is not in dormant state.
3471 */
b00055aa
SR
3472static inline void netif_dormant_off(struct net_device *dev)
3473{
3474 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3475 linkwatch_fire_event(dev);
3476}
3477
bea3348e 3478/**
8ecbc40a 3479 * netif_dormant - test if device is dormant
bea3348e
SH
3480 * @dev: network device
3481 *
8ecbc40a 3482 * Check if device is dormant.
bea3348e 3483 */
4d29515f 3484static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3485{
3486 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3487}
3488
3489
bea3348e
SH
3490/**
3491 * netif_oper_up - test if device is operational
3492 * @dev: network device
3493 *
3494 * Check if carrier is operational
3495 */
4d29515f 3496static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3497{
b00055aa
SR
3498 return (dev->operstate == IF_OPER_UP ||
3499 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3500}
3501
bea3348e
SH
3502/**
3503 * netif_device_present - is device available or removed
3504 * @dev: network device
3505 *
3506 * Check if device has not been removed from system.
3507 */
4d29515f 3508static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3509{
3510 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3511}
3512
f629d208 3513void netif_device_detach(struct net_device *dev);
1da177e4 3514
f629d208 3515void netif_device_attach(struct net_device *dev);
1da177e4
LT
3516
3517/*
3518 * Network interface message level settings
3519 */
1da177e4
LT
3520
3521enum {
3522 NETIF_MSG_DRV = 0x0001,
3523 NETIF_MSG_PROBE = 0x0002,
3524 NETIF_MSG_LINK = 0x0004,
3525 NETIF_MSG_TIMER = 0x0008,
3526 NETIF_MSG_IFDOWN = 0x0010,
3527 NETIF_MSG_IFUP = 0x0020,
3528 NETIF_MSG_RX_ERR = 0x0040,
3529 NETIF_MSG_TX_ERR = 0x0080,
3530 NETIF_MSG_TX_QUEUED = 0x0100,
3531 NETIF_MSG_INTR = 0x0200,
3532 NETIF_MSG_TX_DONE = 0x0400,
3533 NETIF_MSG_RX_STATUS = 0x0800,
3534 NETIF_MSG_PKTDATA = 0x1000,
3535 NETIF_MSG_HW = 0x2000,
3536 NETIF_MSG_WOL = 0x4000,
3537};
3538
3539#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3540#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3541#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3542#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3543#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3544#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3545#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3546#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3547#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3548#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3549#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3550#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3551#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3552#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3553#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3554
3555static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3556{
3557 /* use default */
3558 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3559 return default_msg_enable_bits;
3560 if (debug_value == 0) /* no output */
3561 return 0;
3562 /* set low N bits */
3563 return (1 << debug_value) - 1;
3564}
3565
c773e847 3566static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3567{
c773e847
DM
3568 spin_lock(&txq->_xmit_lock);
3569 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3570}
3571
5a717f4f
MT
3572static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3573{
3574 __acquire(&txq->_xmit_lock);
3575 return true;
3576}
3577
3578static inline void __netif_tx_release(struct netdev_queue *txq)
3579{
3580 __release(&txq->_xmit_lock);
3581}
3582
fd2ea0a7
DM
3583static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3584{
3585 spin_lock_bh(&txq->_xmit_lock);
3586 txq->xmit_lock_owner = smp_processor_id();
3587}
3588
4d29515f 3589static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3590{
4d29515f 3591 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3592 if (likely(ok))
3593 txq->xmit_lock_owner = smp_processor_id();
3594 return ok;
3595}
3596
3597static inline void __netif_tx_unlock(struct netdev_queue *txq)
3598{
3599 txq->xmit_lock_owner = -1;
3600 spin_unlock(&txq->_xmit_lock);
3601}
3602
3603static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3604{
3605 txq->xmit_lock_owner = -1;
3606 spin_unlock_bh(&txq->_xmit_lock);
3607}
3608
08baf561
ED
3609static inline void txq_trans_update(struct netdev_queue *txq)
3610{
3611 if (txq->xmit_lock_owner != -1)
3612 txq->trans_start = jiffies;
3613}
3614
ba162f8e
FW
3615/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3616static inline void netif_trans_update(struct net_device *dev)
3617{
9b36627a
FW
3618 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3619
3620 if (txq->trans_start != jiffies)
3621 txq->trans_start = jiffies;
ba162f8e
FW
3622}
3623
d29f749e
DJ
3624/**
3625 * netif_tx_lock - grab network device transmit lock
3626 * @dev: network device
d29f749e
DJ
3627 *
3628 * Get network device transmit lock
3629 */
22dd7495
JHS
3630static inline void netif_tx_lock(struct net_device *dev)
3631{
e8a0464c 3632 unsigned int i;
c3f26a26 3633 int cpu;
c773e847 3634
c3f26a26
DM
3635 spin_lock(&dev->tx_global_lock);
3636 cpu = smp_processor_id();
e8a0464c
DM
3637 for (i = 0; i < dev->num_tx_queues; i++) {
3638 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3639
3640 /* We are the only thread of execution doing a
3641 * freeze, but we have to grab the _xmit_lock in
3642 * order to synchronize with threads which are in
3643 * the ->hard_start_xmit() handler and already
3644 * checked the frozen bit.
3645 */
e8a0464c 3646 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3647 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3648 __netif_tx_unlock(txq);
e8a0464c 3649 }
932ff279
HX
3650}
3651
3652static inline void netif_tx_lock_bh(struct net_device *dev)
3653{
e8a0464c
DM
3654 local_bh_disable();
3655 netif_tx_lock(dev);
932ff279
HX
3656}
3657
932ff279
HX
3658static inline void netif_tx_unlock(struct net_device *dev)
3659{
e8a0464c
DM
3660 unsigned int i;
3661
3662 for (i = 0; i < dev->num_tx_queues; i++) {
3663 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3664
c3f26a26
DM
3665 /* No need to grab the _xmit_lock here. If the
3666 * queue is not stopped for another reason, we
3667 * force a schedule.
3668 */
3669 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3670 netif_schedule_queue(txq);
c3f26a26
DM
3671 }
3672 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3673}
3674
3675static inline void netif_tx_unlock_bh(struct net_device *dev)
3676{
e8a0464c
DM
3677 netif_tx_unlock(dev);
3678 local_bh_enable();
932ff279
HX
3679}
3680
c773e847 3681#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3682 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3683 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3684 } else { \
3685 __netif_tx_acquire(txq); \
22dd7495
JHS
3686 } \
3687}
3688
5efeac44
EB
3689#define HARD_TX_TRYLOCK(dev, txq) \
3690 (((dev->features & NETIF_F_LLTX) == 0) ? \
3691 __netif_tx_trylock(txq) : \
5a717f4f 3692 __netif_tx_acquire(txq))
5efeac44 3693
c773e847 3694#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3695 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3696 __netif_tx_unlock(txq); \
5a717f4f
MT
3697 } else { \
3698 __netif_tx_release(txq); \
22dd7495
JHS
3699 } \
3700}
3701
1da177e4
LT
3702static inline void netif_tx_disable(struct net_device *dev)
3703{
fd2ea0a7 3704 unsigned int i;
c3f26a26 3705 int cpu;
fd2ea0a7 3706
c3f26a26
DM
3707 local_bh_disable();
3708 cpu = smp_processor_id();
fd2ea0a7
DM
3709 for (i = 0; i < dev->num_tx_queues; i++) {
3710 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3711
3712 __netif_tx_lock(txq, cpu);
fd2ea0a7 3713 netif_tx_stop_queue(txq);
c3f26a26 3714 __netif_tx_unlock(txq);
fd2ea0a7 3715 }
c3f26a26 3716 local_bh_enable();
1da177e4
LT
3717}
3718
e308a5d8
DM
3719static inline void netif_addr_lock(struct net_device *dev)
3720{
3721 spin_lock(&dev->addr_list_lock);
3722}
3723
2429f7ac
JP
3724static inline void netif_addr_lock_nested(struct net_device *dev)
3725{
25175ba5
VY
3726 int subclass = SINGLE_DEPTH_NESTING;
3727
3728 if (dev->netdev_ops->ndo_get_lock_subclass)
3729 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3730
3731 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3732}
3733
e308a5d8
DM
3734static inline void netif_addr_lock_bh(struct net_device *dev)
3735{
3736 spin_lock_bh(&dev->addr_list_lock);
3737}
3738
3739static inline void netif_addr_unlock(struct net_device *dev)
3740{
3741 spin_unlock(&dev->addr_list_lock);
3742}
3743
3744static inline void netif_addr_unlock_bh(struct net_device *dev)
3745{
3746 spin_unlock_bh(&dev->addr_list_lock);
3747}
3748
f001fde5 3749/*
31278e71 3750 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3751 * rcu_read_lock held.
3752 */
3753#define for_each_dev_addr(dev, ha) \
31278e71 3754 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3755
1da177e4
LT
3756/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3757
f629d208 3758void ether_setup(struct net_device *dev);
1da177e4
LT
3759
3760/* Support for loadable net-drivers */
f629d208 3761struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3762 unsigned char name_assign_type,
f629d208
JP
3763 void (*setup)(struct net_device *),
3764 unsigned int txqs, unsigned int rxqs);
0ad646c8
CW
3765int dev_get_valid_name(struct net *net, struct net_device *dev,
3766 const char *name);
3767
c835a677
TG
3768#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3769 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3770
c835a677
TG
3771#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3772 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3773 count)
36909ea4 3774
f629d208
JP
3775int register_netdev(struct net_device *dev);
3776void unregister_netdev(struct net_device *dev);
f001fde5 3777
22bedad3 3778/* General hardware address lists handling functions */
f629d208
JP
3779int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3780 struct netdev_hw_addr_list *from_list, int addr_len);
3781void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3782 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3783int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3784 struct net_device *dev,
3785 int (*sync)(struct net_device *, const unsigned char *),
3786 int (*unsync)(struct net_device *,
3787 const unsigned char *));
3788void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3789 struct net_device *dev,
3790 int (*unsync)(struct net_device *,
3791 const unsigned char *));
f629d208 3792void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3793
f001fde5 3794/* Functions used for device addresses handling */
f629d208
JP
3795int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3796 unsigned char addr_type);
3797int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3798 unsigned char addr_type);
f629d208
JP
3799void dev_addr_flush(struct net_device *dev);
3800int dev_addr_init(struct net_device *dev);
a748ee24
JP
3801
3802/* Functions used for unicast addresses handling */
f629d208
JP
3803int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3804int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3805int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3806int dev_uc_sync(struct net_device *to, struct net_device *from);
3807int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3808void dev_uc_unsync(struct net_device *to, struct net_device *from);
3809void dev_uc_flush(struct net_device *dev);
3810void dev_uc_init(struct net_device *dev);
f001fde5 3811
670e5b8e
AD
3812/**
3813 * __dev_uc_sync - Synchonize device's unicast list
3814 * @dev: device to sync
3815 * @sync: function to call if address should be added
3816 * @unsync: function to call if address should be removed
3817 *
3818 * Add newly added addresses to the interface, and release
3819 * addresses that have been deleted.
5e82b4b2 3820 */
670e5b8e
AD
3821static inline int __dev_uc_sync(struct net_device *dev,
3822 int (*sync)(struct net_device *,
3823 const unsigned char *),
3824 int (*unsync)(struct net_device *,
3825 const unsigned char *))
3826{
3827 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3828}
3829
3830/**
e793c0f7 3831 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3832 * @dev: device to sync
3833 * @unsync: function to call if address should be removed
3834 *
3835 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3836 */
670e5b8e
AD
3837static inline void __dev_uc_unsync(struct net_device *dev,
3838 int (*unsync)(struct net_device *,
3839 const unsigned char *))
3840{
3841 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3842}
3843
22bedad3 3844/* Functions used for multicast addresses handling */
f629d208
JP
3845int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3846int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3847int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3848int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3849int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3850int dev_mc_sync(struct net_device *to, struct net_device *from);
3851int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3852void dev_mc_unsync(struct net_device *to, struct net_device *from);
3853void dev_mc_flush(struct net_device *dev);
3854void dev_mc_init(struct net_device *dev);
f001fde5 3855
670e5b8e
AD
3856/**
3857 * __dev_mc_sync - Synchonize device's multicast list
3858 * @dev: device to sync
3859 * @sync: function to call if address should be added
3860 * @unsync: function to call if address should be removed
3861 *
3862 * Add newly added addresses to the interface, and release
3863 * addresses that have been deleted.
5e82b4b2 3864 */
670e5b8e
AD
3865static inline int __dev_mc_sync(struct net_device *dev,
3866 int (*sync)(struct net_device *,
3867 const unsigned char *),
3868 int (*unsync)(struct net_device *,
3869 const unsigned char *))
3870{
3871 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3872}
3873
3874/**
e793c0f7 3875 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3876 * @dev: device to sync
3877 * @unsync: function to call if address should be removed
3878 *
3879 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3880 */
670e5b8e
AD
3881static inline void __dev_mc_unsync(struct net_device *dev,
3882 int (*unsync)(struct net_device *,
3883 const unsigned char *))
3884{
3885 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3886}
3887
4417da66 3888/* Functions used for secondary unicast and multicast support */
f629d208
JP
3889void dev_set_rx_mode(struct net_device *dev);
3890void __dev_set_rx_mode(struct net_device *dev);
3891int dev_set_promiscuity(struct net_device *dev, int inc);
3892int dev_set_allmulti(struct net_device *dev, int inc);
3893void netdev_state_change(struct net_device *dev);
3894void netdev_notify_peers(struct net_device *dev);
3895void netdev_features_change(struct net_device *dev);
1da177e4 3896/* Load a device via the kmod */
f629d208
JP
3897void dev_load(struct net *net, const char *name);
3898struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3899 struct rtnl_link_stats64 *storage);
3900void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3901 const struct net_device_stats *netdev_stats);
eeda3fd6 3902
1da177e4 3903extern int netdev_max_backlog;
3b098e2d 3904extern int netdev_tstamp_prequeue;
1da177e4 3905extern int weight_p;
3d48b53f
MT
3906extern int dev_weight_rx_bias;
3907extern int dev_weight_tx_bias;
3908extern int dev_rx_weight;
3909extern int dev_tx_weight;
9ff162a8 3910
f629d208 3911bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3912struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3913 struct list_head **iter);
f629d208
JP
3914struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3915 struct list_head **iter);
8b5be856 3916
44a40855
VY
3917/* iterate through upper list, must be called under RCU read lock */
3918#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3919 for (iter = &(dev)->adj_list.upper, \
3920 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3921 updev; \
3922 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3923
1a3f060c
DA
3924int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3925 int (*fn)(struct net_device *upper_dev,
3926 void *data),
3927 void *data);
3928
3929bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3930 struct net_device *upper_dev);
3931
25cc72a3
IS
3932bool netdev_has_any_upper_dev(struct net_device *dev);
3933
f629d208
JP
3934void *netdev_lower_get_next_private(struct net_device *dev,
3935 struct list_head **iter);
3936void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3937 struct list_head **iter);
31088a11
VF
3938
3939#define netdev_for_each_lower_private(dev, priv, iter) \
3940 for (iter = (dev)->adj_list.lower.next, \
3941 priv = netdev_lower_get_next_private(dev, &(iter)); \
3942 priv; \
3943 priv = netdev_lower_get_next_private(dev, &(iter)))
3944
3945#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3946 for (iter = &(dev)->adj_list.lower, \
3947 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3948 priv; \
3949 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3950
4085ebe8
VY
3951void *netdev_lower_get_next(struct net_device *dev,
3952 struct list_head **iter);
7ce856aa 3953
4085ebe8 3954#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3955 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3956 ldev = netdev_lower_get_next(dev, &(iter)); \
3957 ldev; \
3958 ldev = netdev_lower_get_next(dev, &(iter)))
3959
7ce856aa
JP
3960struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3961 struct list_head **iter);
3962struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3963 struct list_head **iter);
3964
1a3f060c
DA
3965int netdev_walk_all_lower_dev(struct net_device *dev,
3966 int (*fn)(struct net_device *lower_dev,
3967 void *data),
3968 void *data);
3969int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3970 int (*fn)(struct net_device *lower_dev,
3971 void *data),
3972 void *data);
3973
f629d208 3974void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3975void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3976struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3977struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
42ab19ee
DA
3978int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
3979 struct netlink_ext_ack *extack);
f629d208 3980int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3981 struct net_device *upper_dev,
42ab19ee
DA
3982 void *upper_priv, void *upper_info,
3983 struct netlink_ext_ack *extack);
f629d208
JP
3984void netdev_upper_dev_unlink(struct net_device *dev,
3985 struct net_device *upper_dev);
5bb025fa 3986void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3987void *netdev_lower_dev_get_private(struct net_device *dev,
3988 struct net_device *lower_dev);
04d48266
JP
3989void netdev_lower_state_changed(struct net_device *lower_dev,
3990 void *lower_state_info);
960fb622
ED
3991
3992/* RSS keys are 40 or 52 bytes long */
3993#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3994extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3995void netdev_rss_key_fill(void *buffer, size_t len);
3996
952fcfd0 3997int dev_get_nest_level(struct net_device *dev);
f629d208 3998int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 3999int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
4000int skb_csum_hwoffload_help(struct sk_buff *skb,
4001 const netdev_features_t features);
4002
f629d208
JP
4003struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4004 netdev_features_t features, bool tx_path);
4005struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4006 netdev_features_t features);
12b0004d 4007
61bd3857
MS
4008struct netdev_bonding_info {
4009 ifslave slave;
4010 ifbond master;
4011};
4012
4013struct netdev_notifier_bonding_info {
4014 struct netdev_notifier_info info; /* must be first */
4015 struct netdev_bonding_info bonding_info;
4016};
4017
4018void netdev_bonding_info_change(struct net_device *dev,
4019 struct netdev_bonding_info *bonding_info);
4020
12b0004d
CW
4021static inline
4022struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4023{
4024 return __skb_gso_segment(skb, features, true);
4025}
53d6471c 4026__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
4027
4028static inline bool can_checksum_protocol(netdev_features_t features,
4029 __be16 protocol)
4030{
c8cd0989
TH
4031 if (protocol == htons(ETH_P_FCOE))
4032 return !!(features & NETIF_F_FCOE_CRC);
4033
4034 /* Assume this is an IP checksum (not SCTP CRC) */
4035
4036 if (features & NETIF_F_HW_CSUM) {
4037 /* Can checksum everything */
4038 return true;
4039 }
4040
4041 switch (protocol) {
4042 case htons(ETH_P_IP):
4043 return !!(features & NETIF_F_IP_CSUM);
4044 case htons(ETH_P_IPV6):
4045 return !!(features & NETIF_F_IPV6_CSUM);
4046 default:
4047 return false;
4048 }
ec5f0615 4049}
12b0004d 4050
fb286bb2 4051#ifdef CONFIG_BUG
f629d208 4052void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
4053#else
4054static inline void netdev_rx_csum_fault(struct net_device *dev)
4055{
4056}
4057#endif
1da177e4 4058/* rx skb timestamps */
f629d208
JP
4059void net_enable_timestamp(void);
4060void net_disable_timestamp(void);
1da177e4 4061
20380731 4062#ifdef CONFIG_PROC_FS
f629d208 4063int __init dev_proc_init(void);
900ff8c6
CW
4064#else
4065#define dev_proc_init() 0
20380731
ACM
4066#endif
4067
4798248e 4068static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4069 struct sk_buff *skb, struct net_device *dev,
4070 bool more)
4798248e 4071{
fa2dbdc2 4072 skb->xmit_more = more ? 1 : 0;
0b725a2c 4073 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4074}
4075
10b3ad8c 4076static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4077 struct netdev_queue *txq, bool more)
4798248e
DM
4078{
4079 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4080 int rc;
4798248e 4081
fa2dbdc2 4082 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4083 if (rc == NETDEV_TX_OK)
4084 txq_trans_update(txq);
4085
4086 return rc;
4798248e
DM
4087}
4088
b793dc5c 4089int netdev_class_create_file_ns(const struct class_attribute *class_attr,
42a2d923 4090 const void *ns);
b793dc5c 4091void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
42a2d923 4092 const void *ns);
58292cbe 4093
b793dc5c 4094static inline int netdev_class_create_file(const struct class_attribute *class_attr)
58292cbe
TH
4095{
4096 return netdev_class_create_file_ns(class_attr, NULL);
4097}
4098
b793dc5c 4099static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
58292cbe
TH
4100{
4101 netdev_class_remove_file_ns(class_attr, NULL);
4102}
b8a9787e 4103
737aec57 4104extern const struct kobj_ns_type_operations net_ns_type_operations;
04600794 4105
f629d208 4106const char *netdev_drivername(const struct net_device *dev);
6579e57b 4107
f629d208 4108void linkwatch_run_queue(void);
20380731 4109
da08143b
MK
4110static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4111 netdev_features_t f2)
4112{
c8cd0989
TH
4113 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4114 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4115 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4116 else
b6a0e72a 4117 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4118 }
da08143b 4119
c8cd0989 4120 return f1 & f2;
da08143b
MK
4121}
4122
c8f44aff
MM
4123static inline netdev_features_t netdev_get_wanted_features(
4124 struct net_device *dev)
5455c699
MM
4125{
4126 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4127}
c8f44aff
MM
4128netdev_features_t netdev_increment_features(netdev_features_t all,
4129 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4130
4131/* Allow TSO being used on stacked device :
4132 * Performing the GSO segmentation before last device
4133 * is a performance improvement.
4134 */
4135static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4136 netdev_features_t mask)
4137{
4138 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4139}
4140
6cb6a27c 4141int __netdev_update_features(struct net_device *dev);
5455c699 4142void netdev_update_features(struct net_device *dev);
afe12cc8 4143void netdev_change_features(struct net_device *dev);
7f353bf2 4144
fc4a7489
PM
4145void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4146 struct net_device *dev);
4147
e38f3025
TM
4148netdev_features_t passthru_features_check(struct sk_buff *skb,
4149 struct net_device *dev,
4150 netdev_features_t features);
c1e756bf 4151netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4152
4d29515f 4153static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4154{
7b748340 4155 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4156
4157 /* check flags correspondence */
4158 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4159 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4160 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4161 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4162 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4163 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4164 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4165 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4166 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4167 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4168 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4169 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4170 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4171 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4172 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4173 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0c19f846 4174 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
0345e186 4175
d6b4991a 4176 return (features & feature) == feature;
576a30eb
HX
4177}
4178
4d29515f 4179static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4180{
278b2513 4181 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4182 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4183}
4184
8b86a61d 4185static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4186 netdev_features_t features)
7967168c 4187{
fc741216 4188 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4189 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4190 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4191}
4192
82cc1a7a
PWJ
4193static inline void netif_set_gso_max_size(struct net_device *dev,
4194 unsigned int size)
4195{
4196 dev->gso_max_size = size;
4197}
4198
7a7ffbab
WCC
4199static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4200 int pulled_hlen, u16 mac_offset,
4201 int mac_len)
4202{
4203 skb->protocol = protocol;
4204 skb->encapsulation = 1;
4205 skb_push(skb, pulled_hlen);
4206 skb_reset_transport_header(skb);
4207 skb->mac_header = mac_offset;
4208 skb->network_header = skb->mac_header + mac_len;
4209 skb->mac_len = mac_len;
4210}
4211
3c175784
SD
4212static inline bool netif_is_macsec(const struct net_device *dev)
4213{
4214 return dev->priv_flags & IFF_MACSEC;
4215}
4216
b618aaa9 4217static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4218{
4219 return dev->priv_flags & IFF_MACVLAN;
4220}
4221
b618aaa9 4222static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4223{
4224 return dev->priv_flags & IFF_MACVLAN_PORT;
4225}
4226
b618aaa9 4227static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4228{
4229 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4230}
4231
b618aaa9 4232static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4233{
4234 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4235}
4236
3bdc0eba
BG
4237static inline bool netif_supports_nofcs(struct net_device *dev)
4238{
4239 return dev->priv_flags & IFF_SUPP_NOFCS;
4240}
4241
007979ea 4242static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4243{
007979ea 4244 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4245}
4246
fee6d4c7
DA
4247static inline bool netif_is_l3_slave(const struct net_device *dev)
4248{
4249 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4250}
4251
0894ae3f
JP
4252static inline bool netif_is_bridge_master(const struct net_device *dev)
4253{
4254 return dev->priv_flags & IFF_EBRIDGE;
4255}
4256
28f9ee22
VY
4257static inline bool netif_is_bridge_port(const struct net_device *dev)
4258{
4259 return dev->priv_flags & IFF_BRIDGE_PORT;
4260}
4261
35d4e172
JP
4262static inline bool netif_is_ovs_master(const struct net_device *dev)
4263{
4264 return dev->priv_flags & IFF_OPENVSWITCH;
4265}
4266
5be66141
JP
4267static inline bool netif_is_ovs_port(const struct net_device *dev)
4268{
4269 return dev->priv_flags & IFF_OVS_DATAPATH;
4270}
4271
b618aaa9 4272static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4273{
4274 return dev->priv_flags & IFF_TEAM;
4275}
4276
b618aaa9 4277static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4278{
4279 return dev->priv_flags & IFF_TEAM_PORT;
4280}
4281
b618aaa9 4282static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4283{
4284 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4285}
4286
b618aaa9 4287static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4288{
4289 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4290}
4291
d4ab4286
KJ
4292static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4293{
4294 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4295}
4296
02875878
ED
4297/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4298static inline void netif_keep_dst(struct net_device *dev)
4299{
4300 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4301}
4302
18d3df3e
PA
4303/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4304static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4305{
4306 /* TODO: reserve and use an additional IFF bit, if we get more users */
4307 return dev->priv_flags & IFF_MACSEC;
4308}
4309
505d4f73 4310extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4311
571ba423
JP
4312/* Logging, debugging and troubleshooting/diagnostic helpers. */
4313
4314/* netdev_printk helpers, similar to dev_printk */
4315
4316static inline const char *netdev_name(const struct net_device *dev)
4317{
c6f854d5
VF
4318 if (!dev->name[0] || strchr(dev->name, '%'))
4319 return "(unnamed net_device)";
571ba423
JP
4320 return dev->name;
4321}
4322
8397ed36
DA
4323static inline bool netdev_unregistering(const struct net_device *dev)
4324{
4325 return dev->reg_state == NETREG_UNREGISTERING;
4326}
4327
ccc7f496
VF
4328static inline const char *netdev_reg_state(const struct net_device *dev)
4329{
4330 switch (dev->reg_state) {
4331 case NETREG_UNINITIALIZED: return " (uninitialized)";
4332 case NETREG_REGISTERED: return "";
4333 case NETREG_UNREGISTERING: return " (unregistering)";
4334 case NETREG_UNREGISTERED: return " (unregistered)";
4335 case NETREG_RELEASED: return " (released)";
4336 case NETREG_DUMMY: return " (dummy)";
4337 }
4338
4339 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4340 return " (unknown)";
4341}
4342
f629d208 4343__printf(3, 4)
6ea754eb
JP
4344void netdev_printk(const char *level, const struct net_device *dev,
4345 const char *format, ...);
f629d208 4346__printf(2, 3)
6ea754eb 4347void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4348__printf(2, 3)
6ea754eb 4349void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4350__printf(2, 3)
6ea754eb 4351void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4352__printf(2, 3)
6ea754eb 4353void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4354__printf(2, 3)
6ea754eb 4355void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4356__printf(2, 3)
6ea754eb 4357void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4358__printf(2, 3)
6ea754eb 4359void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4360
375ef2b1
GP
4361#define netdev_level_once(level, dev, fmt, ...) \
4362do { \
4363 static bool __print_once __read_mostly; \
4364 \
4365 if (!__print_once) { \
4366 __print_once = true; \
4367 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4368 } \
4369} while (0)
4370
4371#define netdev_emerg_once(dev, fmt, ...) \
4372 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4373#define netdev_alert_once(dev, fmt, ...) \
4374 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4375#define netdev_crit_once(dev, fmt, ...) \
4376 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4377#define netdev_err_once(dev, fmt, ...) \
4378 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4379#define netdev_warn_once(dev, fmt, ...) \
4380 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4381#define netdev_notice_once(dev, fmt, ...) \
4382 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4383#define netdev_info_once(dev, fmt, ...) \
4384 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4385
8909c9ad
VK
4386#define MODULE_ALIAS_NETDEV(device) \
4387 MODULE_ALIAS("netdev-" device)
4388
b558c96f 4389#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4390#define netdev_dbg(__dev, format, args...) \
4391do { \
ffa10cb4 4392 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4393} while (0)
b558c96f
JC
4394#elif defined(DEBUG)
4395#define netdev_dbg(__dev, format, args...) \
4396 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4397#else
4398#define netdev_dbg(__dev, format, args...) \
4399({ \
4400 if (0) \
4401 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4402})
4403#endif
4404
4405#if defined(VERBOSE_DEBUG)
4406#define netdev_vdbg netdev_dbg
4407#else
4408
4409#define netdev_vdbg(dev, format, args...) \
4410({ \
4411 if (0) \
4412 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4413 0; \
4414})
4415#endif
4416
4417/*
4418 * netdev_WARN() acts like dev_printk(), but with the key difference
4419 * of using a WARN/WARN_ON to get the message out, including the
4420 * file/line information and a backtrace.
4421 */
4422#define netdev_WARN(dev, format, args...) \
e1cfe3d0 4423 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
ccc7f496 4424 netdev_reg_state(dev), ##args)
571ba423 4425
72dd831e 4426#define netdev_WARN_ONCE(dev, format, args...) \
e1cfe3d0 4427 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
375ef2b1
GP
4428 netdev_reg_state(dev), ##args)
4429
b3d95c5c
JP
4430/* netif printk helpers, similar to netdev_printk */
4431
4432#define netif_printk(priv, type, level, dev, fmt, args...) \
4433do { \
4434 if (netif_msg_##type(priv)) \
4435 netdev_printk(level, (dev), fmt, ##args); \
4436} while (0)
4437
f45f4321
JP
4438#define netif_level(level, priv, type, dev, fmt, args...) \
4439do { \
4440 if (netif_msg_##type(priv)) \
4441 netdev_##level(dev, fmt, ##args); \
4442} while (0)
4443
b3d95c5c 4444#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4445 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4446#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4447 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4448#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4449 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4450#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4451 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4452#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4453 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4454#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4455 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4456#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4457 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4458
0053ea9c 4459#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4460#define netif_dbg(priv, type, netdev, format, args...) \
4461do { \
4462 if (netif_msg_##type(priv)) \
b5fb0a03 4463 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4464} while (0)
0053ea9c
JP
4465#elif defined(DEBUG)
4466#define netif_dbg(priv, type, dev, format, args...) \
4467 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4468#else
4469#define netif_dbg(priv, type, dev, format, args...) \
4470({ \
4471 if (0) \
4472 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4473 0; \
4474})
4475#endif
4476
f617f276
EC
4477/* if @cond then downgrade to debug, else print at @level */
4478#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4479 do { \
4480 if (cond) \
4481 netif_dbg(priv, type, netdev, fmt, ##args); \
4482 else \
4483 netif_ ## level(priv, type, netdev, fmt, ##args); \
4484 } while (0)
4485
b3d95c5c 4486#if defined(VERBOSE_DEBUG)
bcfcc450 4487#define netif_vdbg netif_dbg
b3d95c5c
JP
4488#else
4489#define netif_vdbg(priv, type, dev, format, args...) \
4490({ \
4491 if (0) \
a4ed89cb 4492 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4493 0; \
4494})
4495#endif
571ba423 4496
900ff8c6
CW
4497/*
4498 * The list of packet types we will receive (as opposed to discard)
4499 * and the routines to invoke.
4500 *
4501 * Why 16. Because with 16 the only overlap we get on a hash of the
4502 * low nibble of the protocol value is RARP/SNAP/X.25.
4503 *
900ff8c6 4504 * 0800 IP
900ff8c6
CW
4505 * 0001 802.3
4506 * 0002 AX.25
4507 * 0004 802.2
4508 * 8035 RARP
4509 * 0005 SNAP
4510 * 0805 X.25
4511 * 0806 ARP
4512 * 8137 IPX
4513 * 0009 Localtalk
4514 * 86DD IPv6
4515 */
4516#define PTYPE_HASH_SIZE (16)
4517#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4518
385a154c 4519#endif /* _LINUX_NETDEVICE_H */