]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
xfrm: Add mode handlers for IPsec on layer 2
[mirror_ubuntu-artful-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
7a6b6f51 44#ifdef CONFIG_DCB
2f90b865
AD
45#include <net/dcbnl.h>
46#endif
5bc1421e 47#include <net/netprio_cgroup.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;
c6e970a0
AL
59struct dsa_switch_tree;
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;
1da177e4 69
f629d208
JP
70void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
d07d7507 72
9a1654ba
JP
73/* Backlog congestion levels */
74#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75#define NET_RX_DROP 1 /* packet dropped */
76
572a9d7b
PM
77/*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
91 * others are propagated to higher layers.
92 */
93
94/* qdisc ->enqueue() return codes. */
95#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
96#define NET_XMIT_DROP 0x01 /* skb dropped */
97#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 98#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 99
b9df3cb8
GR
100/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 103#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
104#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105
dc1f8bf6 106/* Driver transmit return codes */
9a1654ba 107#define NETDEV_TX_MASK 0xf0
572a9d7b 108
dc1f8bf6 109enum netdev_tx {
572a9d7b 110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
113};
114typedef enum netdev_tx netdev_tx_t;
115
9a1654ba
JP
116/*
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 */
120static inline bool dev_xmit_complete(int rc)
121{
122 /*
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 */
128 if (likely(rc < NET_XMIT_MASK))
129 return true;
130
131 return false;
132}
133
1da177e4 134/*
5e82b4b2 135 * Compute the worst-case header length according to the protocols
1da177e4
LT
136 * used.
137 */
fe2918b0 138
c0eb4540
KS
139#if defined(CONFIG_HYPERV_NET)
140# define LL_MAX_HEADER 128
141#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
142# if defined(CONFIG_MAC80211_MESH)
143# define LL_MAX_HEADER 128
144# else
145# define LL_MAX_HEADER 96
146# endif
1da177e4 147#else
8388e3da 148# define LL_MAX_HEADER 32
1da177e4
LT
149#endif
150
d11ead75
BH
151#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
153#define MAX_HEADER LL_MAX_HEADER
154#else
155#define MAX_HEADER (LL_MAX_HEADER + 48)
156#endif
157
158/*
be1f3c2c
BH
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
1da177e4 161 */
fe2918b0 162
d94d9fee 163struct net_device_stats {
3cfde79c
BH
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
1da177e4 173 unsigned long collisions;
1da177e4 174 unsigned long rx_length_errors;
3cfde79c
BH
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
1da177e4
LT
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
1da177e4
LT
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187};
188
1da177e4
LT
189
190#include <linux/cache.h>
191#include <linux/skbuff.h>
192
adc9300e 193#ifdef CONFIG_RPS
c5905afb
IM
194#include <linux/static_key.h>
195extern struct static_key rps_needed;
13bfff25 196extern struct static_key rfs_needed;
adc9300e
ED
197#endif
198
1da177e4
LT
199struct neighbour;
200struct neigh_parms;
201struct sk_buff;
202
f001fde5
JP
203struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
206 unsigned char type;
ccffad25
JP
207#define NETDEV_HW_ADDR_T_LAN 1
208#define NETDEV_HW_ADDR_T_SAN 2
209#define NETDEV_HW_ADDR_T_SLAVE 3
210#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 211#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 212 bool global_use;
4cd729b0 213 int sync_cnt;
8f8f103d 214 int refcount;
4543fbef 215 int synced;
f001fde5
JP
216 struct rcu_head rcu_head;
217};
218
31278e71
JP
219struct netdev_hw_addr_list {
220 struct list_head list;
221 int count;
222};
223
22bedad3
JP
224#define netdev_hw_addr_list_count(l) ((l)->count)
225#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226#define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 228
22bedad3
JP
229#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231#define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 233
22bedad3
JP
234#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 236#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 238
d94d9fee 239struct hh_cache {
f6b72b62 240 u16 hh_len;
5c25f686 241 u16 __pad;
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
f2cd2d3e
ED
588static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
589{
590#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
591 return q->numa_node;
592#else
b236da69 593 return NUMA_NO_NODE;
f2cd2d3e
ED
594#endif
595}
596
597static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 q->numa_node = node;
601#endif
602}
603
df334545 604#ifdef CONFIG_RPS
0a9627f2
TH
605/*
606 * This structure holds an RPS map which can be of variable length. The
607 * map is an array of CPUs.
608 */
609struct rps_map {
610 unsigned int len;
611 struct rcu_head rcu;
612 u16 cpus[0];
613};
60b778ce 614#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 615
fec5e652 616/*
c445477d
BH
617 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
618 * tail pointer for that CPU's input queue at the time of last enqueue, and
619 * a hardware filter index.
fec5e652
TH
620 */
621struct rps_dev_flow {
622 u16 cpu;
c445477d 623 u16 filter;
fec5e652
TH
624 unsigned int last_qtail;
625};
c445477d 626#define RPS_NO_FILTER 0xffff
fec5e652
TH
627
628/*
629 * The rps_dev_flow_table structure contains a table of flow mappings.
630 */
631struct rps_dev_flow_table {
632 unsigned int mask;
633 struct rcu_head rcu;
fec5e652
TH
634 struct rps_dev_flow flows[0];
635};
636#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 637 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
638
639/*
640 * The rps_sock_flow_table contains mappings of flows to the last CPU
641 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
642 * Each entry is a 32bit value. Upper part is the high-order bits
643 * of flow hash, lower part is CPU number.
567e4b79 644 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
645 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
646 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 647 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
648 */
649struct rps_sock_flow_table {
567e4b79 650 u32 mask;
93c1af6c
ED
651
652 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 653};
567e4b79 654#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
655
656#define RPS_NO_CPU 0xffff
657
567e4b79
ED
658extern u32 rps_cpu_mask;
659extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
660
fec5e652
TH
661static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
662 u32 hash)
663{
664 if (table && hash) {
567e4b79
ED
665 unsigned int index = hash & table->mask;
666 u32 val = hash & ~rps_cpu_mask;
fec5e652 667
5e82b4b2 668 /* We only give a hint, preemption can change CPU under us */
567e4b79 669 val |= raw_smp_processor_id();
fec5e652 670
567e4b79
ED
671 if (table->ents[index] != val)
672 table->ents[index] = val;
fec5e652
TH
673 }
674}
675
c445477d 676#ifdef CONFIG_RFS_ACCEL
f629d208
JP
677bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
678 u16 filter_id);
c445477d 679#endif
a953be53 680#endif /* CONFIG_RPS */
c445477d 681
0a9627f2
TH
682/* This structure contains an instance of an RX queue. */
683struct netdev_rx_queue {
a953be53 684#ifdef CONFIG_RPS
6e3f7faf
ED
685 struct rps_map __rcu *rps_map;
686 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 687#endif
6e3f7faf 688 struct kobject kobj;
fe822240 689 struct net_device *dev;
0a9627f2 690} ____cacheline_aligned_in_smp;
a953be53
MD
691
692/*
693 * RX queue sysfs structures and functions.
694 */
695struct rx_queue_attribute {
696 struct attribute attr;
697 ssize_t (*show)(struct netdev_rx_queue *queue,
698 struct rx_queue_attribute *attr, char *buf);
699 ssize_t (*store)(struct netdev_rx_queue *queue,
700 struct rx_queue_attribute *attr, const char *buf, size_t len);
701};
d314774c 702
bf264145
TH
703#ifdef CONFIG_XPS
704/*
705 * This structure holds an XPS map which can be of variable length. The
706 * map is an array of queues.
707 */
708struct xps_map {
709 unsigned int len;
710 unsigned int alloc_len;
711 struct rcu_head rcu;
712 u16 queues[0];
713};
60b778ce 714#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
715#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
716 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
717
718/*
719 * This structure holds all XPS maps for device. Maps are indexed by CPU.
720 */
721struct xps_dev_maps {
722 struct rcu_head rcu;
a4177869 723 struct xps_map __rcu *cpu_map[0];
bf264145 724};
184c449f
AD
725#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
726 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
727#endif /* CONFIG_XPS */
728
4f57c087
JF
729#define TC_MAX_QUEUE 16
730#define TC_BITMASK 15
731/* HW offloaded queuing disciplines txq count and offset maps */
732struct netdev_tc_txq {
733 u16 count;
734 u16 offset;
735};
736
68bad94e
NP
737#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
738/*
739 * This structure is to hold information about the device
740 * configured to run FCoE protocol stack.
741 */
742struct netdev_fcoe_hbainfo {
743 char manufacturer[64];
744 char serial_number[64];
745 char hardware_version[64];
746 char driver_version[64];
747 char optionrom_version[64];
748 char firmware_version[64];
749 char model[256];
750 char model_description[256];
751};
752#endif
753
02637fce 754#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 755
02637fce
JP
756/* This structure holds a unique identifier to identify some
757 * physical item (port for example) used by a netdevice.
66b52b0d 758 */
02637fce
JP
759struct netdev_phys_item_id {
760 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
761 unsigned char id_len;
762};
763
d754f98b
SF
764static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
765 struct netdev_phys_item_id *b)
766{
767 return a->id_len == b->id_len &&
768 memcmp(a->id, b->id, a->id_len) == 0;
769}
770
99932d4f
DB
771typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
772 struct sk_buff *skb);
773
a1b7c5fd 774/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
775 * that are being passed to the netdevice through the setup_tc op.
776 */
777enum {
778 TC_SETUP_MQPRIO,
a1b7c5fd 779 TC_SETUP_CLSU32,
5b33f488 780 TC_SETUP_CLSFLOWER,
b87f7936 781 TC_SETUP_MATCHALL,
332ae8e2 782 TC_SETUP_CLSBPF,
16e5cc64
JF
783};
784
a1b7c5fd
JF
785struct tc_cls_u32_offload;
786
16e5cc64
JF
787struct tc_to_netdev {
788 unsigned int type;
789 union {
a1b7c5fd 790 struct tc_cls_u32_offload *cls_u32;
5b33f488 791 struct tc_cls_flower_offload *cls_flower;
b87f7936 792 struct tc_cls_matchall_offload *cls_mall;
332ae8e2 793 struct tc_cls_bpf_offload *cls_bpf;
56f36acd 794 struct tc_mqprio_qopt *mqprio;
16e5cc64 795 };
7091d8c7 796 bool egress_dev;
16e5cc64
JF
797};
798
a7862b45
BB
799/* These structures hold the attributes of xdp state that are being passed
800 * to the netdevice through the xdp op.
801 */
802enum xdp_netdev_command {
803 /* Set or clear a bpf program used in the earliest stages of packet
804 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
805 * is responsible for calling bpf_prog_put on any old progs that are
806 * stored. In case of error, the callee need not release the new prog
807 * reference, but on success it takes ownership and must bpf_prog_put
808 * when it is no longer used.
809 */
810 XDP_SETUP_PROG,
811 /* Check if a bpf program is set on the device. The callee should
812 * return true if a program is currently attached and running.
813 */
814 XDP_QUERY_PROG,
815};
816
817struct netdev_xdp {
818 enum xdp_netdev_command command;
819 union {
820 /* XDP_SETUP_PROG */
821 struct bpf_prog *prog;
822 /* XDP_QUERY_PROG */
823 bool prog_attached;
824 };
825};
16e5cc64 826
d314774c
SH
827/*
828 * This structure defines the management hooks for network devices.
00829823
SH
829 * The following hooks can be defined; unless noted otherwise, they are
830 * optional and can be filled with a null pointer.
d314774c
SH
831 *
832 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
833 * This function is called once when a network device is registered.
834 * The network device can use this for any late stage initialization
835 * or semantic validation. It can fail with an error code which will
836 * be propagated back to register_netdev.
d314774c
SH
837 *
838 * void (*ndo_uninit)(struct net_device *dev);
839 * This function is called when device is unregistered or when registration
840 * fails. It is not called if init fails.
841 *
842 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 843 * This function is called when a network device transitions to the up
d314774c
SH
844 * state.
845 *
846 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 847 * This function is called when a network device transitions to the down
d314774c
SH
848 * state.
849 *
dc1f8bf6
SH
850 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
851 * struct net_device *dev);
00829823 852 * Called when a packet needs to be transmitted.
e79d8429
RR
853 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
854 * the queue before that can happen; it's for obsolete devices and weird
855 * corner cases, but the stack really does a non-trivial amount
856 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 857 * Required; cannot be NULL.
00829823 858 *
1a2a1444
DM
859 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
860 * struct net_device *dev
861 * netdev_features_t features);
862 * Called by core transmit path to determine if device is capable of
863 * performing offload operations on a given packet. This is to give
864 * the device an opportunity to implement any restrictions that cannot
865 * be otherwise expressed by feature flags. The check is called with
866 * the set of features that the stack has calculated and it returns
867 * those the driver believes to be appropriate.
cdba756f 868 *
f663dd9a 869 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 870 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 871 * Called to decide which queue to use when device supports multiple
00829823
SH
872 * transmit queues.
873 *
d314774c
SH
874 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
875 * This function is called to allow device receiver to make
5e82b4b2 876 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
877 *
878 * void (*ndo_set_rx_mode)(struct net_device *dev);
879 * This function is called device changes address list filtering.
01789349 880 * If driver handles unicast address filtering, it should set
5e82b4b2 881 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
882 *
883 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
884 * This function is called when the Media Access Control address
37b607c5 885 * needs to be changed. If this interface is not defined, the
5e82b4b2 886 * MAC address can not be changed.
d314774c
SH
887 *
888 * int (*ndo_validate_addr)(struct net_device *dev);
889 * Test if Media Access Control address is valid for the device.
890 *
891 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
892 * Called when a user requests an ioctl which can't be handled by
893 * the generic interface code. If not defined ioctls return
d314774c
SH
894 * not supported error code.
895 *
896 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
897 * Used to set network devices bus interface parameters. This interface
5e82b4b2 898 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
899 * interface (PCI) for low level management.
900 *
901 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
902 * Called when a user wants to change the Maximum Transfer Unit
903 * of a device. If not defined, any request to change MTU will
904 * will return an error.
905 *
00829823 906 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 907 * Callback used when the transmitter has not made any progress
d314774c
SH
908 * for dev->watchdog ticks.
909 *
bc1f4470 910 * void (*ndo_get_stats64)(struct net_device *dev,
911 * struct rtnl_link_stats64 *storage);
d308e38f 912 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 913 * Called when a user wants to get the network device usage
be1f3c2c 914 * statistics. Drivers must do one of the following:
3cfde79c
BH
915 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
916 * rtnl_link_stats64 structure passed by the caller.
82695d9b 917 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
918 * (which should normally be dev->stats) and return a pointer to
919 * it. The structure may be changed asynchronously only if each
920 * field is written atomically.
921 * 3. Update dev->stats asynchronously and atomically, and define
922 * neither operation.
d314774c 923 *
3df5b3c6 924 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
925 * Return true if this device supports offload stats of this attr_id.
926 *
927 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
928 * void *attr_data)
929 * Get statistics for offload operations by attr_id. Write it into the
930 * attr_data pointer.
931 *
5d632cb7 932 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 933 * If device supports VLAN filtering this function is called when a
80d5c368 934 * VLAN id is registered.
d314774c 935 *
5d632cb7 936 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 937 * If device supports VLAN filtering this function is called when a
80d5c368 938 * VLAN id is unregistered.
d314774c
SH
939 *
940 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
941 *
942 * SR-IOV management functions.
943 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
944 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
945 * u8 qos, __be16 proto);
ed616689
SC
946 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
947 * int max_tx_rate);
5f8444a3 948 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 949 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
950 * int (*ndo_get_vf_config)(struct net_device *dev,
951 * int vf, struct ifla_vf_info *ivf);
1d8faf48 952 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
953 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
954 * struct nlattr *port[]);
01a3d796
VZ
955 *
956 * Enable or disable the VF ability to query its RSS Redirection Table and
957 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 958 * with PF and querying it may introduce a theoretical security risk.
01a3d796 959 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 960 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
6a4bc2b4
FF
961 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
962 * __be16 protocol, struct tc_to_netdev *tc);
963 * Called to setup any 'tc' scheduler, classifier or action on @dev.
964 * This is always called from the stack with the rtnl lock held and netif
965 * tx queues stopped. This allows the netdevice to perform queue
966 * management safely.
c445477d 967 *
e9bce845
YZ
968 * Fiber Channel over Ethernet (FCoE) offload functions.
969 * int (*ndo_fcoe_enable)(struct net_device *dev);
970 * Called when the FCoE protocol stack wants to start using LLD for FCoE
971 * so the underlying device can perform whatever needed configuration or
972 * initialization to support acceleration of FCoE traffic.
973 *
974 * int (*ndo_fcoe_disable)(struct net_device *dev);
975 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
976 * so the underlying device can perform whatever needed clean-ups to
977 * stop supporting acceleration of FCoE traffic.
978 *
979 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
980 * struct scatterlist *sgl, unsigned int sgc);
981 * Called when the FCoE Initiator wants to initialize an I/O that
982 * is a possible candidate for Direct Data Placement (DDP). The LLD can
983 * perform necessary setup and returns 1 to indicate the device is set up
984 * successfully to perform DDP on this I/O, otherwise this returns 0.
985 *
986 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
987 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
988 * indicated by the FC exchange id 'xid', so the underlying device can
989 * clean up and reuse resources for later DDP requests.
990 *
991 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
992 * struct scatterlist *sgl, unsigned int sgc);
993 * Called when the FCoE Target wants to initialize an I/O that
994 * is a possible candidate for Direct Data Placement (DDP). The LLD can
995 * perform necessary setup and returns 1 to indicate the device is set up
996 * successfully to perform DDP on this I/O, otherwise this returns 0.
997 *
68bad94e
NP
998 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
999 * struct netdev_fcoe_hbainfo *hbainfo);
1000 * Called when the FCoE Protocol stack wants information on the underlying
1001 * device. This information is utilized by the FCoE protocol stack to
1002 * register attributes with Fiber Channel management service as per the
1003 * FC-GS Fabric Device Management Information(FDMI) specification.
1004 *
e9bce845
YZ
1005 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1006 * Called when the underlying device wants to override default World Wide
1007 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1008 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1009 * protocol stack to use.
1010 *
c445477d
BH
1011 * RFS acceleration.
1012 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1013 * u16 rxq_index, u32 flow_id);
1014 * Set hardware filter for RFS. rxq_index is the target queue index;
1015 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1016 * Return the filter ID on success, or a negative error code.
fbaec0ea 1017 *
8b98a70c 1018 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1019 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1020 * Called to make another netdev an underling.
1021 *
1022 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1023 * Called to release previously enslaved netdev.
5455c699
MM
1024 *
1025 * Feature/offload setting functions.
1a2a1444
DM
1026 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1027 * netdev_features_t features);
1028 * Adjusts the requested feature flags according to device-specific
1029 * constraints, and returns the resulting flags. Must not modify
1030 * the device state.
1031 *
c8f44aff 1032 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1033 * Called to update device configuration to new features. Passed
1034 * feature set might be less than what was returned by ndo_fix_features()).
1035 * Must return >0 or -errno if it changed dev->features itself.
1036 *
edc7d573 1037 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1038 * struct net_device *dev,
f6f6424b 1039 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1040 * Adds an FDB entry to dev for addr.
1690be63
VY
1041 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1042 * struct net_device *dev,
f6f6424b 1043 * const unsigned char *addr, u16 vid)
77162022
JF
1044 * Deletes the FDB entry from dev coresponding to addr.
1045 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1046 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1047 * int *idx)
77162022
JF
1048 * Used to add FDB entries to dump requests. Implementers should add
1049 * entries to skb and update idx with the number of entries.
e5a55a89 1050 *
ad41faa8
ND
1051 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1052 * u16 flags)
e5a55a89 1053 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1054 * struct net_device *dev, u32 filter_mask,
1055 * int nlflags)
ad41faa8
ND
1056 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1057 * u16 flags);
4bf84c35
JP
1058 *
1059 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1060 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1061 * which do not represent real hardware may define this to allow their
1062 * userspace components to manage their virtual carrier state. Devices
1063 * that determine carrier state from physical hardware properties (eg
1064 * network cables) or protocol-dependent mechanisms (eg
1065 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1066 *
1067 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1068 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1069 * Called to get ID of physical port of this device. If driver does
1070 * not implement this, it is assumed that the hw is not able to have
1071 * multiple net devices on single physical port.
53cf5275 1072 *
7c46a640
AD
1073 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1074 * struct udp_tunnel_info *ti);
1075 * Called by UDP tunnel to notify a driver about the UDP port and socket
1076 * address family that a UDP tunnel is listnening to. It is called only
1077 * when a new port starts listening. The operation is protected by the
1078 * RTNL.
1079 *
1080 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1081 * struct udp_tunnel_info *ti);
1082 * Called by UDP tunnel to notify the driver about a UDP port and socket
1083 * address family that the UDP tunnel is not listening to anymore. The
1084 * operation is protected by the RTNL.
1085 *
a6cc0cfa
JF
1086 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1087 * struct net_device *dev)
1088 * Called by upper layer devices to accelerate switching or other
1089 * station functionality into hardware. 'pdev is the lowerdev
1090 * to use for the offload and 'dev' is the net device that will
1091 * back the offload. Returns a pointer to the private structure
1092 * the upper layer will maintain.
1093 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1094 * Called by upper layer device to delete the station created
1095 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1096 * the station and priv is the structure returned by the add
1097 * operation.
1098 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1099 * struct net_device *dev,
1100 * void *priv);
1101 * Callback to use for xmit over the accelerated station. This
1102 * is used in place of ndo_start_xmit on accelerated net
1103 * devices.
822b3b2e
JF
1104 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1105 * int queue_index, u32 maxrate);
1106 * Called when a user wants to set a max-rate limitation of specific
1107 * TX queue.
a54acb3a
ND
1108 * int (*ndo_get_iflink)(const struct net_device *dev);
1109 * Called to get the iflink value of this device.
d746d707 1110 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1111 * bool proto_down);
d746d707
AK
1112 * This function is used to pass protocol port error state information
1113 * to the switch driver. The switch driver can react to the proto_down
1114 * by doing a phys down on the associated switch port.
fc4099f1
PS
1115 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1116 * This function is used to get egress tunnel information for given skb.
1117 * This is useful for retrieving outer tunnel header parameters while
1118 * sampling packet.
871b642a
PA
1119 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1120 * This function is used to specify the headroom that the skb must
1121 * consider when allocation skb during packet reception. Setting
1122 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1123 * forward. Setting a negative value resets the rx headroom to the
871b642a 1124 * default value.
a7862b45
BB
1125 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1126 * This function is used to set or query state related to XDP on the
1127 * netdevice. See definition of enum xdp_netdev_command for details.
d746d707 1128 *
d314774c
SH
1129 */
1130struct net_device_ops {
1131 int (*ndo_init)(struct net_device *dev);
1132 void (*ndo_uninit)(struct net_device *dev);
1133 int (*ndo_open)(struct net_device *dev);
1134 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1135 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1136 struct net_device *dev);
1137 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1138 struct net_device *dev,
1139 netdev_features_t features);
00829823 1140 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1141 struct sk_buff *skb,
99932d4f
DB
1142 void *accel_priv,
1143 select_queue_fallback_t fallback);
d314774c
SH
1144 void (*ndo_change_rx_flags)(struct net_device *dev,
1145 int flags);
d314774c 1146 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1147 int (*ndo_set_mac_address)(struct net_device *dev,
1148 void *addr);
d314774c 1149 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1150 int (*ndo_do_ioctl)(struct net_device *dev,
1151 struct ifreq *ifr, int cmd);
d314774c
SH
1152 int (*ndo_set_config)(struct net_device *dev,
1153 struct ifmap *map);
00829823
SH
1154 int (*ndo_change_mtu)(struct net_device *dev,
1155 int new_mtu);
1156 int (*ndo_neigh_setup)(struct net_device *dev,
1157 struct neigh_parms *);
d314774c
SH
1158 void (*ndo_tx_timeout) (struct net_device *dev);
1159
bc1f4470 1160 void (*ndo_get_stats64)(struct net_device *dev,
1161 struct rtnl_link_stats64 *storage);
3df5b3c6 1162 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1163 int (*ndo_get_offload_stats)(int attr_id,
1164 const struct net_device *dev,
1165 void *attr_data);
d314774c
SH
1166 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1167
8e586137 1168 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1169 __be16 proto, u16 vid);
8e586137 1170 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1171 __be16 proto, u16 vid);
d314774c 1172#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1173 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1174 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1175 struct netpoll_info *info);
0e34e931 1176 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1177#endif
95c26df8
WM
1178 int (*ndo_set_vf_mac)(struct net_device *dev,
1179 int queue, u8 *mac);
1180 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1181 int queue, u16 vlan,
1182 u8 qos, __be16 proto);
ed616689
SC
1183 int (*ndo_set_vf_rate)(struct net_device *dev,
1184 int vf, int min_tx_rate,
1185 int max_tx_rate);
5f8444a3
GR
1186 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1187 int vf, bool setting);
dd461d6a
HS
1188 int (*ndo_set_vf_trust)(struct net_device *dev,
1189 int vf, bool setting);
95c26df8
WM
1190 int (*ndo_get_vf_config)(struct net_device *dev,
1191 int vf,
1192 struct ifla_vf_info *ivf);
1d8faf48
RE
1193 int (*ndo_set_vf_link_state)(struct net_device *dev,
1194 int vf, int link_state);
3b766cd8
EBE
1195 int (*ndo_get_vf_stats)(struct net_device *dev,
1196 int vf,
1197 struct ifla_vf_stats
1198 *vf_stats);
57b61080
SF
1199 int (*ndo_set_vf_port)(struct net_device *dev,
1200 int vf,
1201 struct nlattr *port[]);
1202 int (*ndo_get_vf_port)(struct net_device *dev,
1203 int vf, struct sk_buff *skb);
cc8e27cc
EC
1204 int (*ndo_set_vf_guid)(struct net_device *dev,
1205 int vf, u64 guid,
1206 int guid_type);
01a3d796
VZ
1207 int (*ndo_set_vf_rss_query_en)(
1208 struct net_device *dev,
1209 int vf, bool setting);
16e5cc64
JF
1210 int (*ndo_setup_tc)(struct net_device *dev,
1211 u32 handle,
1212 __be16 protocol,
1213 struct tc_to_netdev *tc);
d11ead75 1214#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1215 int (*ndo_fcoe_enable)(struct net_device *dev);
1216 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1217 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1218 u16 xid,
1219 struct scatterlist *sgl,
1220 unsigned int sgc);
1221 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1222 u16 xid);
6247e086
YZ
1223 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1224 u16 xid,
1225 struct scatterlist *sgl,
1226 unsigned int sgc);
68bad94e
NP
1227 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1228 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1229#endif
1230
d11ead75 1231#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1232#define NETDEV_FCOE_WWNN 0
1233#define NETDEV_FCOE_WWPN 1
1234 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1235 u64 *wwn, int type);
4d288d57 1236#endif
3c9c36bc 1237
c445477d
BH
1238#ifdef CONFIG_RFS_ACCEL
1239 int (*ndo_rx_flow_steer)(struct net_device *dev,
1240 const struct sk_buff *skb,
1241 u16 rxq_index,
1242 u32 flow_id);
1243#endif
fbaec0ea
JP
1244 int (*ndo_add_slave)(struct net_device *dev,
1245 struct net_device *slave_dev);
1246 int (*ndo_del_slave)(struct net_device *dev,
1247 struct net_device *slave_dev);
c8f44aff
MM
1248 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1249 netdev_features_t features);
5455c699 1250 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1251 netdev_features_t features);
503eebc2
JP
1252 int (*ndo_neigh_construct)(struct net_device *dev,
1253 struct neighbour *n);
1254 void (*ndo_neigh_destroy)(struct net_device *dev,
1255 struct neighbour *n);
77162022
JF
1256
1257 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1258 struct nlattr *tb[],
77162022 1259 struct net_device *dev,
6b6e2725 1260 const unsigned char *addr,
f6f6424b 1261 u16 vid,
77162022
JF
1262 u16 flags);
1263 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1264 struct nlattr *tb[],
77162022 1265 struct net_device *dev,
f6f6424b
JP
1266 const unsigned char *addr,
1267 u16 vid);
77162022
JF
1268 int (*ndo_fdb_dump)(struct sk_buff *skb,
1269 struct netlink_callback *cb,
1270 struct net_device *dev,
5d5eacb3 1271 struct net_device *filter_dev,
d297653d 1272 int *idx);
e5a55a89
JF
1273
1274 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1275 struct nlmsghdr *nlh,
1276 u16 flags);
e5a55a89
JF
1277 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1278 u32 pid, u32 seq,
6cbdceeb 1279 struct net_device *dev,
46c264da
ND
1280 u32 filter_mask,
1281 int nlflags);
407af329 1282 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1283 struct nlmsghdr *nlh,
1284 u16 flags);
4bf84c35
JP
1285 int (*ndo_change_carrier)(struct net_device *dev,
1286 bool new_carrier);
66b52b0d 1287 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1288 struct netdev_phys_item_id *ppid);
db24a904
DA
1289 int (*ndo_get_phys_port_name)(struct net_device *dev,
1290 char *name, size_t len);
7c46a640
AD
1291 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1292 struct udp_tunnel_info *ti);
1293 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1294 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1295 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1296 struct net_device *dev);
1297 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1298 void *priv);
1299
1300 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1301 struct net_device *dev,
1302 void *priv);
25175ba5 1303 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1304 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1305 int queue_index,
1306 u32 maxrate);
a54acb3a 1307 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1308 int (*ndo_change_proto_down)(struct net_device *dev,
1309 bool proto_down);
fc4099f1
PS
1310 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1311 struct sk_buff *skb);
871b642a
PA
1312 void (*ndo_set_rx_headroom)(struct net_device *dev,
1313 int needed_headroom);
a7862b45
BB
1314 int (*ndo_xdp)(struct net_device *dev,
1315 struct netdev_xdp *xdp);
d314774c
SH
1316};
1317
7aa98047
LR
1318/**
1319 * enum net_device_priv_flags - &struct net_device priv_flags
1320 *
1321 * These are the &struct net_device, they are only set internally
1322 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1323 * userspace; this means that the order of these flags can change
7aa98047
LR
1324 * during any kernel release.
1325 *
1326 * You should have a pretty good reason to be extending these flags.
1327 *
1328 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1329 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1330 * @IFF_BONDING: bonding master or slave
7aa98047 1331 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1332 * @IFF_WAN_HDLC: WAN HDLC device
1333 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1334 * release skb->dst
1335 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1336 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1337 * @IFF_MACVLAN_PORT: device used as macvlan port
1338 * @IFF_BRIDGE_PORT: device used as bridge port
1339 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1340 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1341 * @IFF_UNICAST_FLT: Supports unicast filtering
1342 * @IFF_TEAM_PORT: device used as team port
1343 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1344 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1345 * change when it's running
1346 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1347 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1348 * underlying stacked devices
1349 * @IFF_IPVLAN_MASTER: IPvlan master device
1350 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1351 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1352 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1353 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1354 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1355 * @IFF_TEAM: device is a team device
d4ab4286 1356 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1357 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1358 * entity (i.e. the master device for bridged veth)
3c175784 1359 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1360 */
1361enum netdev_priv_flags {
1362 IFF_802_1Q_VLAN = 1<<0,
1363 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1364 IFF_BONDING = 1<<2,
1365 IFF_ISATAP = 1<<3,
1366 IFF_WAN_HDLC = 1<<4,
1367 IFF_XMIT_DST_RELEASE = 1<<5,
1368 IFF_DONT_BRIDGE = 1<<6,
1369 IFF_DISABLE_NETPOLL = 1<<7,
1370 IFF_MACVLAN_PORT = 1<<8,
1371 IFF_BRIDGE_PORT = 1<<9,
1372 IFF_OVS_DATAPATH = 1<<10,
1373 IFF_TX_SKB_SHARING = 1<<11,
1374 IFF_UNICAST_FLT = 1<<12,
1375 IFF_TEAM_PORT = 1<<13,
1376 IFF_SUPP_NOFCS = 1<<14,
1377 IFF_LIVE_ADDR_CHANGE = 1<<15,
1378 IFF_MACVLAN = 1<<16,
1379 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1380 IFF_IPVLAN_MASTER = 1<<18,
1381 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1382 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1383 IFF_NO_QUEUE = 1<<21,
1384 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1385 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1386 IFF_TEAM = 1<<24,
d4ab4286 1387 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1388 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1389 IFF_MACSEC = 1<<27,
7aa98047
LR
1390};
1391
1392#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1393#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1394#define IFF_BONDING IFF_BONDING
7aa98047 1395#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1396#define IFF_WAN_HDLC IFF_WAN_HDLC
1397#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1398#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1399#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1400#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1401#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1402#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1403#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1404#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1405#define IFF_TEAM_PORT IFF_TEAM_PORT
1406#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1407#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1408#define IFF_MACVLAN IFF_MACVLAN
02875878 1409#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1410#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1411#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1412#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1413#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1414#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1415#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1416#define IFF_TEAM IFF_TEAM
d4ab4286 1417#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1418#define IFF_MACSEC IFF_MACSEC
7aa98047 1419
536721b1
KK
1420/**
1421 * struct net_device - The DEVICE structure.
1422 * Actually, this whole structure is a big mistake. It mixes I/O
1423 * data with strictly "high-level" data, and it has to know about
1424 * almost every data structure used in the INET module.
1425 *
1426 * @name: This is the first field of the "visible" part of this structure
1427 * (i.e. as seen by users in the "Space.c" file). It is the name
1428 * of the interface.
1429 *
1430 * @name_hlist: Device name hash chain, please keep it close to name[]
1431 * @ifalias: SNMP alias
1432 * @mem_end: Shared memory end
1433 * @mem_start: Shared memory start
1434 * @base_addr: Device I/O address
1435 * @irq: Device IRQ number
1436 *
14ffbbb8
TG
1437 * @carrier_changes: Stats to monitor carrier on<->off transitions
1438 *
536721b1
KK
1439 * @state: Generic network queuing layer state, see netdev_state_t
1440 * @dev_list: The global list of network devices
5e82b4b2
BH
1441 * @napi_list: List entry used for polling NAPI devices
1442 * @unreg_list: List entry when we are unregistering the
1443 * device; see the function unregister_netdev
1444 * @close_list: List entry used when we are closing the device
62d885fe
BP
1445 * @ptype_all: Device-specific packet handlers for all protocols
1446 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1447 *
1448 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1449 * @features: Currently active device features
1450 * @hw_features: User-changeable features
1451 *
1452 * @wanted_features: User-requested features
1453 * @vlan_features: Mask of features inheritable by VLAN devices
1454 *
1455 * @hw_enc_features: Mask of features inherited by encapsulating devices
1456 * This field indicates what encapsulation
1457 * offloads the hardware is capable of doing,
1458 * and drivers will need to set them appropriately.
1459 *
1460 * @mpls_features: Mask of features inheritable by MPLS
1461 *
1462 * @ifindex: interface index
5e82b4b2 1463 * @group: The group the device belongs to
536721b1
KK
1464 *
1465 * @stats: Statistics struct, which was left as a legacy, use
1466 * rtnl_link_stats64 instead
1467 *
1468 * @rx_dropped: Dropped packets by core network,
1469 * do not use this in drivers
1470 * @tx_dropped: Dropped packets by core network,
1471 * do not use this in drivers
6e7333d3
JW
1472 * @rx_nohandler: nohandler dropped packets by core network on
1473 * inactive devices, do not use this in drivers
536721b1 1474 *
536721b1
KK
1475 * @wireless_handlers: List of functions to handle Wireless Extensions,
1476 * instead of ioctl,
1477 * see <net/iw_handler.h> for details.
1478 * @wireless_data: Instance data managed by the core of wireless extensions
1479 *
1480 * @netdev_ops: Includes several pointers to callbacks,
1481 * if one wants to override the ndo_*() functions
1482 * @ethtool_ops: Management operations
f997c55c
AA
1483 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1484 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1485 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1486 * of Layer 2 headers.
1487 *
1488 * @flags: Interface flags (a la BSD)
1489 * @priv_flags: Like 'flags' but invisible to userspace,
1490 * see if.h for the definitions
1491 * @gflags: Global flags ( kept as legacy )
1492 * @padded: How much padding added by alloc_netdev()
1493 * @operstate: RFC2863 operstate
1494 * @link_mode: Mapping policy to operstate
1495 * @if_port: Selectable AUI, TP, ...
1496 * @dma: DMA channel
1497 * @mtu: Interface MTU value
61e84623
JW
1498 * @min_mtu: Interface Minimum MTU value
1499 * @max_mtu: Interface Maximum MTU value
536721b1 1500 * @type: Interface hardware type
2793a23a 1501 * @hard_header_len: Maximum hardware header length.
217e6fa2 1502 * @min_header_len: Minimum hardware header length
536721b1
KK
1503 *
1504 * @needed_headroom: Extra headroom the hardware may need, but not in all
1505 * cases can this be guaranteed
1506 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1507 * cases can this be guaranteed. Some cases also use
1508 * LL_MAX_HEADER instead to allocate the skb
1509 *
1510 * interface address info:
1511 *
1512 * @perm_addr: Permanent hw address
1513 * @addr_assign_type: Hw address assignment type
1514 * @addr_len: Hardware address length
8626a0c8 1515 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1516 * @dev_id: Used to differentiate devices that share
1517 * the same link layer address
1518 * @dev_port: Used to differentiate devices that share
1519 * the same function
1520 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1521 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1522 * has been enabled due to the need to listen to
1523 * additional unicast addresses in a device that
1524 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1525 * @uc: unicast mac addresses
1526 * @mc: multicast mac addresses
1527 * @dev_addrs: list of device hw addresses
1528 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1529 * @promiscuity: Number of times the NIC is told to work in
1530 * promiscuous mode; if it becomes 0 the NIC will
1531 * exit promiscuous mode
536721b1
KK
1532 * @allmulti: Counter, enables or disables allmulticast mode
1533 *
1534 * @vlan_info: VLAN info
1535 * @dsa_ptr: dsa specific data
1536 * @tipc_ptr: TIPC specific data
1537 * @atalk_ptr: AppleTalk link
1538 * @ip_ptr: IPv4 specific data
1539 * @dn_ptr: DECnet specific data
1540 * @ip6_ptr: IPv6 specific data
1541 * @ax25_ptr: AX.25 specific data
1542 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1543 *
536721b1
KK
1544 * @dev_addr: Hw address (before bcast,
1545 * because most packets are unicast)
1546 *
1547 * @_rx: Array of RX queues
1548 * @num_rx_queues: Number of RX queues
1549 * allocated at register_netdev() time
1550 * @real_num_rx_queues: Number of RX queues currently active in device
1551 *
1552 * @rx_handler: handler for received packets
1553 * @rx_handler_data: XXX: need comments on this one
1554 * @ingress_queue: XXX: need comments on this one
1555 * @broadcast: hw bcast address
1556 *
14ffbbb8
TG
1557 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1558 * indexed by RX queue number. Assigned by driver.
1559 * This must only be set if the ndo_rx_flow_steer
1560 * operation is defined
1561 * @index_hlist: Device index hash chain
1562 *
536721b1
KK
1563 * @_tx: Array of TX queues
1564 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1565 * @real_num_tx_queues: Number of TX queues currently active in device
1566 * @qdisc: Root qdisc from userspace point of view
1567 * @tx_queue_len: Max frames per queue allowed
1568 * @tx_global_lock: XXX: need comments on this one
1569 *
1570 * @xps_maps: XXX: need comments on this one
1571 *
536721b1 1572 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1573 * the watchdog (see dev_watchdog())
536721b1
KK
1574 * @watchdog_timer: List of timers
1575 *
1576 * @pcpu_refcnt: Number of references to this device
1577 * @todo_list: Delayed register/unregister
536721b1
KK
1578 * @link_watch_list: XXX: need comments on this one
1579 *
1580 * @reg_state: Register/unregister state machine
1581 * @dismantle: Device is going to be freed
1582 * @rtnl_link_state: This enum represents the phases of creating
1583 * a new link
1584 *
1585 * @destructor: Called from unregister,
1586 * can be used to call free_netdev
1587 * @npinfo: XXX: need comments on this one
1588 * @nd_net: Network namespace this network device is inside
1589 *
1590 * @ml_priv: Mid-layer private
1591 * @lstats: Loopback statistics
1592 * @tstats: Tunnel statistics
1593 * @dstats: Dummy statistics
1594 * @vstats: Virtual ethernet statistics
1595 *
1596 * @garp_port: GARP
1597 * @mrp_port: MRP
1598 *
1599 * @dev: Class/net/name entry
1600 * @sysfs_groups: Space for optional device, statistics and wireless
1601 * sysfs groups
1602 *
1603 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1604 * @rtnl_link_ops: Rtnl_link_ops
1605 *
1606 * @gso_max_size: Maximum size of generic segmentation offload
1607 * @gso_max_segs: Maximum number of segments that can be passed to the
1608 * NIC for GSO
1609 *
1610 * @dcbnl_ops: Data Center Bridging netlink ops
1611 * @num_tc: Number of traffic classes in the net device
1612 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1613 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1614 *
1615 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1616 *
1617 * @priomap: XXX: need comments on this one
1618 * @phydev: Physical device may attach itself
1619 * for hardware timestamping
1620 *
123b3652
ED
1621 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1622 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1623 *
d746d707
AK
1624 * @proto_down: protocol port state information can be sent to the
1625 * switch driver and used to set the phys state of the
1626 * switch port.
1627 *
1da177e4
LT
1628 * FIXME: cleanup struct net_device such that network protocol info
1629 * moves out.
1630 */
1631
d94d9fee 1632struct net_device {
1da177e4 1633 char name[IFNAMSIZ];
9356b8fc 1634 struct hlist_node name_hlist;
0b815a1a 1635 char *ifalias;
1da177e4
LT
1636 /*
1637 * I/O specific fields
1638 * FIXME: Merge these and struct ifmap into one
1639 */
536721b1
KK
1640 unsigned long mem_end;
1641 unsigned long mem_start;
1642 unsigned long base_addr;
1643 int irq;
1da177e4 1644
14ffbbb8
TG
1645 atomic_t carrier_changes;
1646
1da177e4 1647 /*
536721b1
KK
1648 * Some hardware also needs these fields (state,dev_list,
1649 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1650 * part of the usual set specified in Space.c.
1651 */
1652
1da177e4
LT
1653 unsigned long state;
1654
7562f876 1655 struct list_head dev_list;
bea3348e 1656 struct list_head napi_list;
44a0873d 1657 struct list_head unreg_list;
5cde2829 1658 struct list_head close_list;
7866a621
SN
1659 struct list_head ptype_all;
1660 struct list_head ptype_specific;
2f268f12 1661
2f268f12
VF
1662 struct {
1663 struct list_head upper;
1664 struct list_head lower;
1665 } adj_list;
1666
c8f44aff 1667 netdev_features_t features;
c8f44aff 1668 netdev_features_t hw_features;
c8f44aff 1669 netdev_features_t wanted_features;
c8f44aff 1670 netdev_features_t vlan_features;
6a674e9c 1671 netdev_features_t hw_enc_features;
0d89d203 1672 netdev_features_t mpls_features;
802ab55a 1673 netdev_features_t gso_partial_features;
04ed3e74 1674
1da177e4 1675 int ifindex;
7a66bbc9 1676 int group;
1da177e4 1677
c45d286e 1678 struct net_device_stats stats;
015f0688 1679
015f0688
ED
1680 atomic_long_t rx_dropped;
1681 atomic_long_t tx_dropped;
6e7333d3 1682 atomic_long_t rx_nohandler;
1da177e4 1683
b86e0280 1684#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1685 const struct iw_handler_def *wireless_handlers;
1686 struct iw_public_data *wireless_data;
b86e0280 1687#endif
d314774c 1688 const struct net_device_ops *netdev_ops;
76fd8593 1689 const struct ethtool_ops *ethtool_ops;
4170604f 1690#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1691 const struct switchdev_ops *switchdev_ops;
4170604f 1692#endif
1b69c6d0
DA
1693#ifdef CONFIG_NET_L3_MASTER_DEV
1694 const struct l3mdev_ops *l3mdev_ops;
1695#endif
f997c55c
AA
1696#if IS_ENABLED(CONFIG_IPV6)
1697 const struct ndisc_ops *ndisc_ops;
1698#endif
1da177e4 1699
3b04ddde
SH
1700 const struct header_ops *header_ops;
1701
536721b1
KK
1702 unsigned int flags;
1703 unsigned int priv_flags;
1704
1da177e4 1705 unsigned short gflags;
536721b1 1706 unsigned short padded;
1da177e4 1707
536721b1
KK
1708 unsigned char operstate;
1709 unsigned char link_mode;
b00055aa 1710
536721b1
KK
1711 unsigned char if_port;
1712 unsigned char dma;
bdc220da 1713
536721b1 1714 unsigned int mtu;
61e84623
JW
1715 unsigned int min_mtu;
1716 unsigned int max_mtu;
536721b1
KK
1717 unsigned short type;
1718 unsigned short hard_header_len;
217e6fa2 1719 unsigned short min_header_len;
1da177e4 1720
f5184d26
JB
1721 unsigned short needed_headroom;
1722 unsigned short needed_tailroom;
1723
1da177e4 1724 /* Interface address info. */
536721b1
KK
1725 unsigned char perm_addr[MAX_ADDR_LEN];
1726 unsigned char addr_assign_type;
1727 unsigned char addr_len;
a0a9663d 1728 unsigned short neigh_priv_len;
536721b1
KK
1729 unsigned short dev_id;
1730 unsigned short dev_port;
ccffad25 1731 spinlock_t addr_list_lock;
14ffbbb8
TG
1732 unsigned char name_assign_type;
1733 bool uc_promisc;
536721b1
KK
1734 struct netdev_hw_addr_list uc;
1735 struct netdev_hw_addr_list mc;
1736 struct netdev_hw_addr_list dev_addrs;
1737
4c3d5e7b
ED
1738#ifdef CONFIG_SYSFS
1739 struct kset *queues_kset;
1740#endif
9d45abe1
WC
1741 unsigned int promiscuity;
1742 unsigned int allmulti;
1da177e4 1743
1da177e4 1744
5e82b4b2 1745 /* Protocol-specific pointers */
65ac6a5f 1746
d11ead75 1747#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1748 struct vlan_info __rcu *vlan_info;
65ac6a5f 1749#endif
34a430d7 1750#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1751 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1752#endif
1753#if IS_ENABLED(CONFIG_TIPC)
536721b1 1754 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1755#endif
536721b1
KK
1756 void *atalk_ptr;
1757 struct in_device __rcu *ip_ptr;
1758 struct dn_dev __rcu *dn_ptr;
1759 struct inet6_dev __rcu *ip6_ptr;
1760 void *ax25_ptr;
1761 struct wireless_dev *ieee80211_ptr;
98a18b6f 1762 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1763#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1764 struct mpls_dev __rcu *mpls_ptr;
1765#endif
1da177e4 1766
9356b8fc 1767/*
cd13539b 1768 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1769 */
9356b8fc 1770 /* Interface address info used in eth_type_trans() */
536721b1 1771 unsigned char *dev_addr;
f001fde5 1772
a953be53 1773#ifdef CONFIG_SYSFS
0a9627f2
TH
1774 struct netdev_rx_queue *_rx;
1775
0a9627f2 1776 unsigned int num_rx_queues;
62fe0b40 1777 unsigned int real_num_rx_queues;
df334545 1778#endif
0a9627f2 1779
3b47d303 1780 unsigned long gro_flush_timeout;
61391cde 1781 rx_handler_func_t __rcu *rx_handler;
1782 void __rcu *rx_handler_data;
e8a0464c 1783
4cda01e8 1784#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1785 struct tcf_proto __rcu *ingress_cl_list;
1786#endif
24824a09 1787 struct netdev_queue __rcu *ingress_queue;
e687ad60 1788#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1789 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1790#endif
d2788d34 1791
536721b1 1792 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1793#ifdef CONFIG_RFS_ACCEL
1794 struct cpu_rmap *rx_cpu_rmap;
1795#endif
1796 struct hlist_node index_hlist;
cd13539b
ED
1797
1798/*
1799 * Cache lines mostly used on transmit path
1800 */
e8a0464c
DM
1801 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1802 unsigned int num_tx_queues;
fd2ea0a7 1803 unsigned int real_num_tx_queues;
af356afa 1804 struct Qdisc *qdisc;
59cc1f61
JK
1805#ifdef CONFIG_NET_SCHED
1806 DECLARE_HASHTABLE (qdisc_hash, 4);
1807#endif
536721b1 1808 unsigned long tx_queue_len;
c3f26a26 1809 spinlock_t tx_global_lock;
14ffbbb8 1810 int watchdog_timeo;
cd13539b 1811
bf264145 1812#ifdef CONFIG_XPS
a4177869 1813 struct xps_dev_maps __rcu *xps_maps;
bf264145 1814#endif
1f211a1b
DB
1815#ifdef CONFIG_NET_CLS_ACT
1816 struct tcf_proto __rcu *egress_cl_list;
1817#endif
0c4f691f 1818
9356b8fc 1819 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1820 struct timer_list watchdog_timer;
1821
29b4433d 1822 int __percpu *pcpu_refcnt;
1da177e4 1823 struct list_head todo_list;
1da177e4 1824
e014debe 1825 struct list_head link_watch_list;
572a103d 1826
1da177e4 1827 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1828 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1829 NETREG_UNREGISTERING, /* called unregister_netdevice */
1830 NETREG_UNREGISTERED, /* completed unregister todo */
1831 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1832 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1833 } reg_state:8;
1834
536721b1 1835 bool dismantle;
a2835763
PM
1836
1837 enum {
1838 RTNL_LINK_INITIALIZED,
1839 RTNL_LINK_INITIALIZING,
1840 } rtnl_link_state:16;
1da177e4 1841
d314774c 1842 void (*destructor)(struct net_device *dev);
1da177e4 1843
1da177e4 1844#ifdef CONFIG_NETPOLL
5fbee843 1845 struct netpoll_info __rcu *npinfo;
1da177e4 1846#endif
eae792b7 1847
0c5c9fb5 1848 possible_net_t nd_net;
4a1c5371 1849
4951704b 1850 /* mid-layer private */
a7855c78 1851 union {
536721b1
KK
1852 void *ml_priv;
1853 struct pcpu_lstats __percpu *lstats;
8f84985f 1854 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1855 struct pcpu_dstats __percpu *dstats;
1856 struct pcpu_vstats __percpu *vstats;
a7855c78 1857 };
536721b1 1858
fb585b44 1859#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1860 struct garp_port __rcu *garp_port;
fb585b44
TK
1861#endif
1862#if IS_ENABLED(CONFIG_MRP)
febf018d 1863 struct mrp_port __rcu *mrp_port;
fb585b44 1864#endif
1da177e4 1865
5e82b4b2 1866 struct device dev;
0c509a6c 1867 const struct attribute_group *sysfs_groups[4];
a953be53 1868 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1869
38f7b870 1870 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1871
82cc1a7a
PWJ
1872 /* for setting kernel sock attribute on TCP connection setup */
1873#define GSO_MAX_SIZE 65536
1874 unsigned int gso_max_size;
30b678d8
BH
1875#define GSO_MAX_SEGS 65535
1876 u16 gso_max_segs;
743b03a8 1877
7a6b6f51 1878#ifdef CONFIG_DCB
32953543 1879 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1880#endif
5e82b4b2
BH
1881 u8 num_tc;
1882 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1883 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1884
d11ead75 1885#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1886 unsigned int fcoe_ddp_xid;
5bc1421e 1887#endif
86f8515f 1888#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1889 struct netprio_map __rcu *priomap;
4d288d57 1890#endif
5e82b4b2
BH
1891 struct phy_device *phydev;
1892 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1893 struct lock_class_key *qdisc_running_key;
5e82b4b2 1894 bool proto_down;
1da177e4 1895};
43cb76d9 1896#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1897
1898#define NETDEV_ALIGN 32
1da177e4 1899
4f57c087
JF
1900static inline
1901int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1902{
1903 return dev->prio_tc_map[prio & TC_BITMASK];
1904}
1905
1906static inline
1907int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1908{
1909 if (tc >= dev->num_tc)
1910 return -EINVAL;
1911
1912 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1913 return 0;
1914}
1915
8d059b0f 1916int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1917void netdev_reset_tc(struct net_device *dev);
1918int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1919int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1920
1921static inline
1922int netdev_get_num_tc(struct net_device *dev)
1923{
1924 return dev->num_tc;
1925}
1926
e8a0464c
DM
1927static inline
1928struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1929 unsigned int index)
1930{
1931 return &dev->_tx[index];
1932}
1933
10c51b56
DB
1934static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1935 const struct sk_buff *skb)
1936{
1937 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1938}
1939
e8a0464c
DM
1940static inline void netdev_for_each_tx_queue(struct net_device *dev,
1941 void (*f)(struct net_device *,
1942 struct netdev_queue *,
1943 void *),
1944 void *arg)
1945{
1946 unsigned int i;
1947
1948 for (i = 0; i < dev->num_tx_queues; i++)
1949 f(dev, &dev->_tx[i], arg);
1950}
1951
d3fff6c4
ED
1952#define netdev_lockdep_set_classes(dev) \
1953{ \
1954 static struct lock_class_key qdisc_tx_busylock_key; \
1955 static struct lock_class_key qdisc_running_key; \
1956 static struct lock_class_key qdisc_xmit_lock_key; \
1957 static struct lock_class_key dev_addr_list_lock_key; \
1958 unsigned int i; \
1959 \
1960 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1961 (dev)->qdisc_running_key = &qdisc_running_key; \
1962 lockdep_set_class(&(dev)->addr_list_lock, \
1963 &dev_addr_list_lock_key); \
1964 for (i = 0; i < (dev)->num_tx_queues; i++) \
1965 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1966 &qdisc_xmit_lock_key); \
1967}
1968
f629d208 1969struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1970 struct sk_buff *skb,
1971 void *accel_priv);
8c4c49df 1972
871b642a
PA
1973/* returns the headroom that the master device needs to take in account
1974 * when forwarding to this dev
1975 */
1976static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1977{
1978 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1979}
1980
1981static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1982{
1983 if (dev->netdev_ops->ndo_set_rx_headroom)
1984 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1985}
1986
1987/* set the device rx headroom to the dev's default */
1988static inline void netdev_reset_rx_headroom(struct net_device *dev)
1989{
1990 netdev_set_rx_headroom(dev, -1);
1991}
1992
c346dca1
YH
1993/*
1994 * Net namespace inlines
1995 */
1996static inline
1997struct net *dev_net(const struct net_device *dev)
1998{
c2d9ba9b 1999 return read_pnet(&dev->nd_net);
c346dca1
YH
2000}
2001
2002static inline
f5aa23fd 2003void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2004{
0c5c9fb5 2005 write_pnet(&dev->nd_net, net);
c346dca1
YH
2006}
2007
bea3348e
SH
2008/**
2009 * netdev_priv - access network device private data
2010 * @dev: network device
2011 *
2012 * Get network device private data
2013 */
6472ce60 2014static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2015{
1ce8e7b5 2016 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2017}
2018
1da177e4
LT
2019/* Set the sysfs physical device reference for the network logical device
2020 * if set prior to registration will cause a symlink during initialization.
2021 */
43cb76d9 2022#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2023
384912ed 2024/* Set the sysfs device type for the network logical device to allow
3f79410c 2025 * fine-grained identification of different network device types. For
5e82b4b2 2026 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2027 */
2028#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2029
82dc3c63
ED
2030/* Default NAPI poll() weight
2031 * Device drivers are strongly advised to not use bigger value
2032 */
2033#define NAPI_POLL_WEIGHT 64
2034
3b582cc1 2035/**
5e82b4b2 2036 * netif_napi_add - initialize a NAPI context
3b582cc1 2037 * @dev: network device
5e82b4b2 2038 * @napi: NAPI context
3b582cc1
SH
2039 * @poll: polling function
2040 * @weight: default weight
2041 *
5e82b4b2
BH
2042 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2043 * *any* of the other NAPI-related functions.
3b582cc1 2044 */
d565b0a1
HX
2045void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2046 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2047
d64b5e85 2048/**
5e82b4b2 2049 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2050 * @dev: network device
5e82b4b2 2051 * @napi: NAPI context
d64b5e85
ED
2052 * @poll: polling function
2053 * @weight: default weight
2054 *
2055 * This variant of netif_napi_add() should be used from drivers using NAPI
2056 * to exclusively poll a TX queue.
2057 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2058 */
2059static inline void netif_tx_napi_add(struct net_device *dev,
2060 struct napi_struct *napi,
2061 int (*poll)(struct napi_struct *, int),
2062 int weight)
2063{
2064 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2065 netif_napi_add(dev, napi, poll, weight);
2066}
2067
d8156534 2068/**
5e82b4b2
BH
2069 * netif_napi_del - remove a NAPI context
2070 * @napi: NAPI context
d8156534 2071 *
5e82b4b2 2072 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2073 */
d565b0a1
HX
2074void netif_napi_del(struct napi_struct *napi);
2075
2076struct napi_gro_cb {
78a478d0 2077 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2078 void *frag0;
78a478d0 2079
7489594c
HX
2080 /* Length of frag0. */
2081 unsigned int frag0_len;
2082
86911732 2083 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2084 int data_offset;
86911732 2085
d565b0a1 2086 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2087 u16 flush;
2088
2089 /* Save the IP ID here and check when we get to the transport layer */
2090 u16 flush_id;
d565b0a1
HX
2091
2092 /* Number of segments aggregated. */
2e71a6f8
ED
2093 u16 count;
2094
15e2396d
TH
2095 /* Start offset for remote checksum offload */
2096 u16 gro_remcsum_start;
2097
2e71a6f8
ED
2098 /* jiffies when first packet was created/queued */
2099 unsigned long age;
86347245 2100
afe93325 2101 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2102 u16 proto;
2103
baa32ff4
TH
2104 /* This is non-zero if the packet may be of the same flow. */
2105 u8 same_flow:1;
2106
fac8e0f5
JG
2107 /* Used in tunnel GRO receive */
2108 u8 encap_mark:1;
573e8fca
TH
2109
2110 /* GRO checksum is valid */
2111 u8 csum_valid:1;
2112
662880f4
TH
2113 /* Number of checksums via CHECKSUM_UNNECESSARY */
2114 u8 csum_cnt:3;
c3c7c254 2115
baa32ff4
TH
2116 /* Free the skb? */
2117 u8 free:2;
2118#define NAPI_GRO_FREE 1
2119#define NAPI_GRO_FREE_STOLEN_HEAD 2
2120
efc98d08
TH
2121 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2122 u8 is_ipv6:1;
2123
a0ca153f
AD
2124 /* Used in GRE, set in fou/gue_gro_receive */
2125 u8 is_fou:1;
2126
1530545e
AD
2127 /* Used to determine if flush_id can be ignored */
2128 u8 is_atomic:1;
2129
fcd91dd4
SD
2130 /* Number of gro_receive callbacks this packet already went through */
2131 u8 recursion_counter:4;
2132
2133 /* 1 bit hole */
baa32ff4 2134
bf5a755f
JC
2135 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2136 __wsum csum;
2137
c3c7c254
ED
2138 /* used in skb_gro_receive() slow path */
2139 struct sk_buff *last;
d565b0a1
HX
2140};
2141
2142#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2143
fcd91dd4
SD
2144#define GRO_RECURSION_LIMIT 15
2145static inline int gro_recursion_inc_test(struct sk_buff *skb)
2146{
2147 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2148}
2149
2150typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2151static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2152 struct sk_buff **head,
2153 struct sk_buff *skb)
2154{
2155 if (unlikely(gro_recursion_inc_test(skb))) {
2156 NAPI_GRO_CB(skb)->flush |= 1;
2157 return NULL;
2158 }
2159
2160 return cb(head, skb);
2161}
2162
2163typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2164 struct sk_buff *);
2165static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2166 struct sock *sk,
2167 struct sk_buff **head,
2168 struct sk_buff *skb)
2169{
2170 if (unlikely(gro_recursion_inc_test(skb))) {
2171 NAPI_GRO_CB(skb)->flush |= 1;
2172 return NULL;
2173 }
2174
2175 return cb(sk, head, skb);
2176}
2177
1da177e4 2178struct packet_type {
f2ccd8fa
DM
2179 __be16 type; /* This is really htons(ether_type). */
2180 struct net_device *dev; /* NULL is wildcarded here */
2181 int (*func) (struct sk_buff *,
2182 struct net_device *,
2183 struct packet_type *,
2184 struct net_device *);
c0de08d0
EL
2185 bool (*id_match)(struct packet_type *ptype,
2186 struct sock *sk);
1da177e4
LT
2187 void *af_packet_priv;
2188 struct list_head list;
2189};
2190
f191a1d1 2191struct offload_callbacks {
576a30eb 2192 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2193 netdev_features_t features);
d565b0a1 2194 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2195 struct sk_buff *skb);
299603e8 2196 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2197};
2198
2199struct packet_offload {
2200 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2201 u16 priority;
f191a1d1
VY
2202 struct offload_callbacks callbacks;
2203 struct list_head list;
1da177e4
LT
2204};
2205
5e82b4b2 2206/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2207struct pcpu_sw_netstats {
2208 u64 rx_packets;
2209 u64 rx_bytes;
2210 u64 tx_packets;
2211 u64 tx_bytes;
2212 struct u64_stats_sync syncp;
2213};
2214
aabc92bb
PNA
2215#define __netdev_alloc_pcpu_stats(type, gfp) \
2216({ \
2217 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2218 if (pcpu_stats) { \
2219 int __cpu; \
2220 for_each_possible_cpu(__cpu) { \
2221 typeof(type) *stat; \
2222 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2223 u64_stats_init(&stat->syncp); \
2224 } \
2225 } \
2226 pcpu_stats; \
1c213bd2
WC
2227})
2228
aabc92bb 2229#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2230 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2231
764f5e54
JP
2232enum netdev_lag_tx_type {
2233 NETDEV_LAG_TX_TYPE_UNKNOWN,
2234 NETDEV_LAG_TX_TYPE_RANDOM,
2235 NETDEV_LAG_TX_TYPE_BROADCAST,
2236 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2237 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2238 NETDEV_LAG_TX_TYPE_HASH,
2239};
2240
2241struct netdev_lag_upper_info {
2242 enum netdev_lag_tx_type tx_type;
2243};
2244
fb1b2e3c
JP
2245struct netdev_lag_lower_state_info {
2246 u8 link_up : 1,
2247 tx_enabled : 1;
2248};
2249
1da177e4
LT
2250#include <linux/notifier.h>
2251
dcfe1421
AW
2252/* netdevice notifier chain. Please remember to update the rtnetlink
2253 * notification exclusion list in rtnetlink_event() when adding new
2254 * types.
2255 */
2256#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2257#define NETDEV_DOWN 0x0002
2258#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2259 detected a hardware crash and restarted
2260 - we can use this eg to kick tcp sessions
2261 once done */
2262#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2263#define NETDEV_REGISTER 0x0005
2264#define NETDEV_UNREGISTER 0x0006
1d486bfb 2265#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2266#define NETDEV_CHANGEADDR 0x0008
2267#define NETDEV_GOING_DOWN 0x0009
2268#define NETDEV_CHANGENAME 0x000A
2269#define NETDEV_FEAT_CHANGE 0x000B
2270#define NETDEV_BONDING_FAILOVER 0x000C
2271#define NETDEV_PRE_UP 0x000D
2272#define NETDEV_PRE_TYPE_CHANGE 0x000E
2273#define NETDEV_POST_TYPE_CHANGE 0x000F
2274#define NETDEV_POST_INIT 0x0010
0115e8e3 2275#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2276#define NETDEV_RELEASE 0x0012
2277#define NETDEV_NOTIFY_PEERS 0x0013
2278#define NETDEV_JOIN 0x0014
42e52bf9 2279#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2280#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2281#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2282#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2283#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2284#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2285#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2286#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2287#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2288
f629d208
JP
2289int register_netdevice_notifier(struct notifier_block *nb);
2290int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2291
2292struct netdev_notifier_info {
2293 struct net_device *dev;
2294};
2295
be9efd36
JP
2296struct netdev_notifier_change_info {
2297 struct netdev_notifier_info info; /* must be first */
2298 unsigned int flags_changed;
2299};
2300
0e4ead9d
JP
2301struct netdev_notifier_changeupper_info {
2302 struct netdev_notifier_info info; /* must be first */
2303 struct net_device *upper_dev; /* new upper dev */
2304 bool master; /* is upper dev master */
5e82b4b2 2305 bool linking; /* is the notification for link or unlink */
29bf24af 2306 void *upper_info; /* upper dev info */
0e4ead9d
JP
2307};
2308
04d48266
JP
2309struct netdev_notifier_changelowerstate_info {
2310 struct netdev_notifier_info info; /* must be first */
2311 void *lower_state_info; /* is lower dev state */
2312};
2313
75538c2b
CW
2314static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2315 struct net_device *dev)
2316{
2317 info->dev = dev;
2318}
2319
351638e7
JP
2320static inline struct net_device *
2321netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2322{
2323 return info->dev;
2324}
2325
f629d208 2326int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2327
2328
1da177e4
LT
2329extern rwlock_t dev_base_lock; /* Device list lock */
2330
881d966b
EB
2331#define for_each_netdev(net, d) \
2332 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2333#define for_each_netdev_reverse(net, d) \
2334 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2335#define for_each_netdev_rcu(net, d) \
2336 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2337#define for_each_netdev_safe(net, d, n) \
2338 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2339#define for_each_netdev_continue(net, d) \
2340 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2341#define for_each_netdev_continue_rcu(net, d) \
2342 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2343#define for_each_netdev_in_bond_rcu(bond, slave) \
2344 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2345 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2346#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2347
a050c33f
DL
2348static inline struct net_device *next_net_device(struct net_device *dev)
2349{
2350 struct list_head *lh;
2351 struct net *net;
2352
c346dca1 2353 net = dev_net(dev);
a050c33f
DL
2354 lh = dev->dev_list.next;
2355 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2356}
2357
ce81b76a
ED
2358static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2359{
2360 struct list_head *lh;
2361 struct net *net;
2362
2363 net = dev_net(dev);
ccf43438 2364 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2365 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2366}
2367
a050c33f
DL
2368static inline struct net_device *first_net_device(struct net *net)
2369{
2370 return list_empty(&net->dev_base_head) ? NULL :
2371 net_device_entry(net->dev_base_head.next);
2372}
7562f876 2373
ccf43438
ED
2374static inline struct net_device *first_net_device_rcu(struct net *net)
2375{
2376 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2377
2378 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2379}
2380
f629d208
JP
2381int netdev_boot_setup_check(struct net_device *dev);
2382unsigned long netdev_boot_base(const char *prefix, int unit);
2383struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2384 const char *hwaddr);
2385struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2386struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2387void dev_add_pack(struct packet_type *pt);
2388void dev_remove_pack(struct packet_type *pt);
2389void __dev_remove_pack(struct packet_type *pt);
2390void dev_add_offload(struct packet_offload *po);
2391void dev_remove_offload(struct packet_offload *po);
f629d208 2392
a54acb3a 2393int dev_get_iflink(const struct net_device *dev);
fc4099f1 2394int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2395struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2396 unsigned short mask);
f629d208
JP
2397struct net_device *dev_get_by_name(struct net *net, const char *name);
2398struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2399struct net_device *__dev_get_by_name(struct net *net, const char *name);
2400int dev_alloc_name(struct net_device *dev, const char *name);
2401int dev_open(struct net_device *dev);
2402int dev_close(struct net_device *dev);
99c4a26a 2403int dev_close_many(struct list_head *head, bool unlink);
f629d208 2404void dev_disable_lro(struct net_device *dev);
0c4b51f0 2405int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2406int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2407int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2408int register_netdevice(struct net_device *dev);
2409void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2410void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2411static inline void unregister_netdevice(struct net_device *dev)
2412{
2413 unregister_netdevice_queue(dev, NULL);
2414}
2415
f629d208
JP
2416int netdev_refcnt_read(const struct net_device *dev);
2417void free_netdev(struct net_device *dev);
74d332c1 2418void netdev_freemem(struct net_device *dev);
f629d208
JP
2419void synchronize_net(void);
2420int init_dummy_netdev(struct net_device *dev);
937f1ba5 2421
f60e5990 2422DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2423#define XMIT_RECURSION_LIMIT 10
2424
f60e5990 2425static inline int dev_recursion_level(void)
2426{
2427 return this_cpu_read(xmit_recursion);
2428}
2429
f629d208
JP
2430struct net_device *dev_get_by_index(struct net *net, int ifindex);
2431struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2432struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2433int netdev_get_name(struct net *net, char *name, int ifindex);
2434int dev_restart(struct net_device *dev);
f629d208 2435int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2436
2437static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2438{
2439 return NAPI_GRO_CB(skb)->data_offset;
2440}
2441
2442static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2443{
2444 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2445}
2446
2447static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2448{
2449 NAPI_GRO_CB(skb)->data_offset += len;
2450}
2451
a5b1cf28
HX
2452static inline void *skb_gro_header_fast(struct sk_buff *skb,
2453 unsigned int offset)
86911732 2454{
a5b1cf28
HX
2455 return NAPI_GRO_CB(skb)->frag0 + offset;
2456}
78a478d0 2457
a5b1cf28
HX
2458static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2459{
2460 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2461}
78a478d0 2462
57ea52a8
HX
2463static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2464{
2465 NAPI_GRO_CB(skb)->frag0 = NULL;
2466 NAPI_GRO_CB(skb)->frag0_len = 0;
2467}
2468
a5b1cf28
HX
2469static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2470 unsigned int offset)
2471{
17dd759c
HX
2472 if (!pskb_may_pull(skb, hlen))
2473 return NULL;
2474
57ea52a8 2475 skb_gro_frag0_invalidate(skb);
17dd759c 2476 return skb->data + offset;
86911732 2477}
1da177e4 2478
36e7b1b8
HX
2479static inline void *skb_gro_network_header(struct sk_buff *skb)
2480{
78d3fd0b
HX
2481 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2482 skb_network_offset(skb);
36e7b1b8
HX
2483}
2484
bf5a755f
JC
2485static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2486 const void *start, unsigned int len)
2487{
573e8fca 2488 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2489 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2490 csum_partial(start, len, 0));
2491}
2492
573e8fca
TH
2493/* GRO checksum functions. These are logical equivalents of the normal
2494 * checksum functions (in skbuff.h) except that they operate on the GRO
2495 * offsets and fields in sk_buff.
2496 */
2497
2498__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2499
15e2396d
TH
2500static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2501{
b7fe10e5 2502 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2503}
2504
573e8fca
TH
2505static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2506 bool zero_okay,
2507 __sum16 check)
2508{
6edec0e6
TH
2509 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2510 skb_checksum_start_offset(skb) <
2511 skb_gro_offset(skb)) &&
15e2396d 2512 !skb_at_gro_remcsum_start(skb) &&
662880f4 2513 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2514 (!zero_okay || check));
2515}
2516
2517static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2518 __wsum psum)
2519{
2520 if (NAPI_GRO_CB(skb)->csum_valid &&
2521 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2522 return 0;
2523
2524 NAPI_GRO_CB(skb)->csum = psum;
2525
2526 return __skb_gro_checksum_complete(skb);
2527}
2528
573e8fca
TH
2529static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2530{
662880f4
TH
2531 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2532 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2533 NAPI_GRO_CB(skb)->csum_cnt--;
2534 } else {
2535 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2536 * verified a new top level checksum or an encapsulated one
2537 * during GRO. This saves work if we fallback to normal path.
2538 */
2539 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2540 }
2541}
2542
2543#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2544 compute_pseudo) \
2545({ \
2546 __sum16 __ret = 0; \
2547 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2548 __ret = __skb_gro_checksum_validate_complete(skb, \
2549 compute_pseudo(skb, proto)); \
5a212329
TH
2550 if (__ret) \
2551 __skb_mark_checksum_bad(skb); \
2552 else \
573e8fca
TH
2553 skb_gro_incr_csum_unnecessary(skb); \
2554 __ret; \
2555})
2556
2557#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2558 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2559
2560#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2561 compute_pseudo) \
2562 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2563
2564#define skb_gro_checksum_simple_validate(skb) \
2565 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2566
d96535a1
TH
2567static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2568{
2569 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2570 !NAPI_GRO_CB(skb)->csum_valid);
2571}
2572
2573static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2574 __sum16 check, __wsum pseudo)
2575{
2576 NAPI_GRO_CB(skb)->csum = ~pseudo;
2577 NAPI_GRO_CB(skb)->csum_valid = 1;
2578}
2579
2580#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2581do { \
2582 if (__skb_gro_checksum_convert_check(skb)) \
2583 __skb_gro_checksum_convert(skb, check, \
2584 compute_pseudo(skb, proto)); \
2585} while (0)
2586
26c4f7da
TH
2587struct gro_remcsum {
2588 int offset;
2589 __wsum delta;
2590};
2591
2592static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2593{
846cd667 2594 grc->offset = 0;
26c4f7da
TH
2595 grc->delta = 0;
2596}
2597
b7fe10e5
TH
2598static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2599 unsigned int off, size_t hdrlen,
2600 int start, int offset,
2601 struct gro_remcsum *grc,
2602 bool nopartial)
dcdc8994
TH
2603{
2604 __wsum delta;
b7fe10e5 2605 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2606
2607 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2608
15e2396d 2609 if (!nopartial) {
b7fe10e5
TH
2610 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2611 return ptr;
2612 }
2613
2614 ptr = skb_gro_header_fast(skb, off);
2615 if (skb_gro_header_hard(skb, off + plen)) {
2616 ptr = skb_gro_header_slow(skb, off + plen, off);
2617 if (!ptr)
2618 return NULL;
15e2396d
TH
2619 }
2620
b7fe10e5
TH
2621 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2622 start, offset);
dcdc8994
TH
2623
2624 /* Adjust skb->csum since we changed the packet */
dcdc8994 2625 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2626
b7fe10e5 2627 grc->offset = off + hdrlen + offset;
26c4f7da 2628 grc->delta = delta;
b7fe10e5
TH
2629
2630 return ptr;
dcdc8994
TH
2631}
2632
26c4f7da
TH
2633static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2634 struct gro_remcsum *grc)
2635{
b7fe10e5
TH
2636 void *ptr;
2637 size_t plen = grc->offset + sizeof(u16);
2638
26c4f7da
TH
2639 if (!grc->delta)
2640 return;
2641
b7fe10e5
TH
2642 ptr = skb_gro_header_fast(skb, grc->offset);
2643 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2644 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2645 if (!ptr)
2646 return;
2647 }
2648
2649 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2650}
dcdc8994 2651
25393d3f
SK
2652#ifdef CONFIG_XFRM_OFFLOAD
2653static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2654{
2655 if (PTR_ERR(pp) != -EINPROGRESS)
2656 NAPI_GRO_CB(skb)->flush |= flush;
2657}
2658#else
5f114163
SK
2659static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2660{
2661 NAPI_GRO_CB(skb)->flush |= flush;
2662}
25393d3f 2663#endif
5f114163 2664
0c4e8581
SH
2665static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2666 unsigned short type,
3b04ddde 2667 const void *daddr, const void *saddr,
95c96174 2668 unsigned int len)
0c4e8581 2669{
f1ecfd5d 2670 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2671 return 0;
3b04ddde
SH
2672
2673 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2674}
2675
b95cce35
SH
2676static inline int dev_parse_header(const struct sk_buff *skb,
2677 unsigned char *haddr)
2678{
2679 const struct net_device *dev = skb->dev;
2680
1b83336b 2681 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2682 return 0;
3b04ddde 2683 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2684}
2685
2793a23a
WB
2686/* ll_header must have at least hard_header_len allocated */
2687static inline bool dev_validate_header(const struct net_device *dev,
2688 char *ll_header, int len)
2689{
2690 if (likely(len >= dev->hard_header_len))
2691 return true;
217e6fa2
WB
2692 if (len < dev->min_header_len)
2693 return false;
2793a23a
WB
2694
2695 if (capable(CAP_SYS_RAWIO)) {
2696 memset(ll_header + len, 0, dev->hard_header_len - len);
2697 return true;
2698 }
2699
2700 if (dev->header_ops && dev->header_ops->validate)
2701 return dev->header_ops->validate(ll_header, len);
2702
2703 return false;
2704}
2705
1da177e4 2706typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2707int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2708static inline int unregister_gifconf(unsigned int family)
2709{
2710 return register_gifconf(family, NULL);
2711}
2712
99bbc707 2713#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2714#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2715struct sd_flow_limit {
2716 u64 count;
2717 unsigned int num_buckets;
2718 unsigned int history_head;
2719 u16 history[FLOW_LIMIT_HISTORY];
2720 u8 buckets[];
2721};
2722
2723extern int netdev_flow_limit_table_len;
2724#endif /* CONFIG_NET_FLOW_LIMIT */
2725
1da177e4 2726/*
5e82b4b2 2727 * Incoming packets are placed on per-CPU queues
1da177e4 2728 */
d94d9fee 2729struct softnet_data {
1da177e4 2730 struct list_head poll_list;
6e7676c1 2731 struct sk_buff_head process_queue;
1da177e4 2732
dee42870 2733 /* stats */
cd7b5396
DM
2734 unsigned int processed;
2735 unsigned int time_squeeze;
cd7b5396 2736 unsigned int received_rps;
fd793d89 2737#ifdef CONFIG_RPS
88751275 2738 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2739#endif
2740#ifdef CONFIG_NET_FLOW_LIMIT
2741 struct sd_flow_limit __rcu *flow_limit;
2742#endif
2743 struct Qdisc *output_queue;
2744 struct Qdisc **output_queue_tailp;
2745 struct sk_buff *completion_queue;
88751275 2746
4cdb1e2e 2747#ifdef CONFIG_RPS
501e7ef5
ED
2748 /* input_queue_head should be written by cpu owning this struct,
2749 * and only read by other cpus. Worth using a cache line.
2750 */
2751 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2752
2753 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2754 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2755 struct softnet_data *rps_ipi_next;
2756 unsigned int cpu;
76cc8b13 2757 unsigned int input_queue_tail;
1e94d72f 2758#endif
95c96174 2759 unsigned int dropped;
0a9627f2 2760 struct sk_buff_head input_pkt_queue;
bea3348e 2761 struct napi_struct backlog;
99bbc707 2762
1da177e4
LT
2763};
2764
76cc8b13 2765static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2766{
2767#ifdef CONFIG_RPS
76cc8b13
TH
2768 sd->input_queue_head++;
2769#endif
2770}
2771
2772static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2773 unsigned int *qtail)
2774{
2775#ifdef CONFIG_RPS
2776 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2777#endif
2778}
2779
0a9627f2 2780DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2781
f629d208 2782void __netif_schedule(struct Qdisc *q);
46e5da40 2783void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2784
fd2ea0a7
DM
2785static inline void netif_tx_schedule_all(struct net_device *dev)
2786{
2787 unsigned int i;
2788
2789 for (i = 0; i < dev->num_tx_queues; i++)
2790 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2791}
2792
f9a7cbbf 2793static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2794{
73466498 2795 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2796}
2797
bea3348e
SH
2798/**
2799 * netif_start_queue - allow transmit
2800 * @dev: network device
2801 *
2802 * Allow upper layers to call the device hard_start_xmit routine.
2803 */
1da177e4
LT
2804static inline void netif_start_queue(struct net_device *dev)
2805{
e8a0464c 2806 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2807}
2808
fd2ea0a7
DM
2809static inline void netif_tx_start_all_queues(struct net_device *dev)
2810{
2811 unsigned int i;
2812
2813 for (i = 0; i < dev->num_tx_queues; i++) {
2814 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2815 netif_tx_start_queue(txq);
2816 }
2817}
2818
46e5da40 2819void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2820
d29f749e
DJ
2821/**
2822 * netif_wake_queue - restart transmit
2823 * @dev: network device
2824 *
2825 * Allow upper layers to call the device hard_start_xmit routine.
2826 * Used for flow control when transmit resources are available.
2827 */
79d16385
DM
2828static inline void netif_wake_queue(struct net_device *dev)
2829{
e8a0464c 2830 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2831}
2832
fd2ea0a7
DM
2833static inline void netif_tx_wake_all_queues(struct net_device *dev)
2834{
2835 unsigned int i;
2836
2837 for (i = 0; i < dev->num_tx_queues; i++) {
2838 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2839 netif_tx_wake_queue(txq);
2840 }
2841}
2842
f9a7cbbf 2843static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2844{
73466498 2845 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2846}
2847
bea3348e
SH
2848/**
2849 * netif_stop_queue - stop transmitted packets
2850 * @dev: network device
2851 *
2852 * Stop upper layers calling the device hard_start_xmit routine.
2853 * Used for flow control when transmit resources are unavailable.
2854 */
1da177e4
LT
2855static inline void netif_stop_queue(struct net_device *dev)
2856{
e8a0464c 2857 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2858}
2859
a2029240 2860void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2861
4d29515f 2862static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2863{
73466498 2864 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2865}
2866
bea3348e
SH
2867/**
2868 * netif_queue_stopped - test if transmit queue is flowblocked
2869 * @dev: network device
2870 *
2871 * Test if transmit queue on device is currently unable to send.
2872 */
4d29515f 2873static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2874{
e8a0464c 2875 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2876}
2877
4d29515f 2878static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2879{
73466498
TH
2880 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2881}
2882
8e2f1a63
DB
2883static inline bool
2884netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2885{
2886 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2887}
2888
8e2f1a63
DB
2889static inline bool
2890netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2891{
2892 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2893}
2894
53511453
ED
2895/**
2896 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2897 * @dev_queue: pointer to transmit queue
2898 *
2899 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2900 * to give appropriate hint to the CPU.
53511453
ED
2901 */
2902static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2903{
2904#ifdef CONFIG_BQL
2905 prefetchw(&dev_queue->dql.num_queued);
2906#endif
2907}
2908
2909/**
2910 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2911 * @dev_queue: pointer to transmit queue
2912 *
2913 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2914 * to give appropriate hint to the CPU.
53511453
ED
2915 */
2916static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2917{
2918#ifdef CONFIG_BQL
2919 prefetchw(&dev_queue->dql.limit);
2920#endif
2921}
2922
c5d67bd7
TH
2923static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2924 unsigned int bytes)
2925{
114cf580
TH
2926#ifdef CONFIG_BQL
2927 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2928
2929 if (likely(dql_avail(&dev_queue->dql) >= 0))
2930 return;
2931
2932 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2933
2934 /*
2935 * The XOFF flag must be set before checking the dql_avail below,
2936 * because in netdev_tx_completed_queue we update the dql_completed
2937 * before checking the XOFF flag.
2938 */
2939 smp_mb();
2940
2941 /* check again in case another CPU has just made room avail */
2942 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2943 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2944#endif
c5d67bd7
TH
2945}
2946
0042d0c8
FF
2947/**
2948 * netdev_sent_queue - report the number of bytes queued to hardware
2949 * @dev: network device
2950 * @bytes: number of bytes queued to the hardware device queue
2951 *
2952 * Report the number of bytes queued for sending/completion to the network
2953 * device hardware queue. @bytes should be a good approximation and should
2954 * exactly match netdev_completed_queue() @bytes
2955 */
c5d67bd7
TH
2956static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2957{
2958 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2959}
2960
2961static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2962 unsigned int pkts, unsigned int bytes)
c5d67bd7 2963{
114cf580 2964#ifdef CONFIG_BQL
b37c0fbe
AD
2965 if (unlikely(!bytes))
2966 return;
2967
2968 dql_completed(&dev_queue->dql, bytes);
2969
2970 /*
2971 * Without the memory barrier there is a small possiblity that
2972 * netdev_tx_sent_queue will miss the update and cause the queue to
2973 * be stopped forever
2974 */
2975 smp_mb();
2976
2977 if (dql_avail(&dev_queue->dql) < 0)
2978 return;
2979
2980 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2981 netif_schedule_queue(dev_queue);
114cf580 2982#endif
c5d67bd7
TH
2983}
2984
0042d0c8
FF
2985/**
2986 * netdev_completed_queue - report bytes and packets completed by device
2987 * @dev: network device
2988 * @pkts: actual number of packets sent over the medium
2989 * @bytes: actual number of bytes sent over the medium
2990 *
2991 * Report the number of bytes and packets transmitted by the network device
2992 * hardware queue over the physical medium, @bytes must exactly match the
2993 * @bytes amount passed to netdev_sent_queue()
2994 */
c5d67bd7 2995static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2996 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2997{
2998 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2999}
3000
3001static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3002{
114cf580 3003#ifdef CONFIG_BQL
5c490354 3004 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3005 dql_reset(&q->dql);
3006#endif
c5d67bd7
TH
3007}
3008
0042d0c8
FF
3009/**
3010 * netdev_reset_queue - reset the packets and bytes count of a network device
3011 * @dev_queue: network device
3012 *
3013 * Reset the bytes and packet count of a network device and clear the
3014 * software flow control OFF bit for this network device
3015 */
c5d67bd7
TH
3016static inline void netdev_reset_queue(struct net_device *dev_queue)
3017{
3018 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3019}
3020
b9507bda
DB
3021/**
3022 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3023 * @dev: network device
3024 * @queue_index: given tx queue index
3025 *
3026 * Returns 0 if given tx queue index >= number of device tx queues,
3027 * otherwise returns the originally passed tx queue index.
3028 */
3029static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3030{
3031 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3032 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3033 dev->name, queue_index,
3034 dev->real_num_tx_queues);
3035 return 0;
3036 }
3037
3038 return queue_index;
3039}
3040
bea3348e
SH
3041/**
3042 * netif_running - test if up
3043 * @dev: network device
3044 *
3045 * Test if the device has been brought up.
3046 */
4d29515f 3047static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3048{
3049 return test_bit(__LINK_STATE_START, &dev->state);
3050}
3051
f25f4e44 3052/*
5e82b4b2 3053 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3054 * stop, and a check if it's stopped. All other device management is
3055 * done at the overall netdevice level.
3056 * Also test the device if we're multiqueue.
3057 */
bea3348e
SH
3058
3059/**
3060 * netif_start_subqueue - allow sending packets on subqueue
3061 * @dev: network device
3062 * @queue_index: sub queue index
3063 *
3064 * Start individual transmit queue of a device with multiple transmit queues.
3065 */
f25f4e44
PWJ
3066static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3067{
fd2ea0a7 3068 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3069
3070 netif_tx_start_queue(txq);
f25f4e44
PWJ
3071}
3072
bea3348e
SH
3073/**
3074 * netif_stop_subqueue - stop sending packets on subqueue
3075 * @dev: network device
3076 * @queue_index: sub queue index
3077 *
3078 * Stop individual transmit queue of a device with multiple transmit queues.
3079 */
f25f4e44
PWJ
3080static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3081{
fd2ea0a7 3082 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3083 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3084}
3085
bea3348e
SH
3086/**
3087 * netif_subqueue_stopped - test status of subqueue
3088 * @dev: network device
3089 * @queue_index: sub queue index
3090 *
3091 * Check individual transmit queue of a device with multiple transmit queues.
3092 */
4d29515f
DM
3093static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3094 u16 queue_index)
f25f4e44 3095{
fd2ea0a7 3096 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3097
3098 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3099}
3100
4d29515f
DM
3101static inline bool netif_subqueue_stopped(const struct net_device *dev,
3102 struct sk_buff *skb)
668f895a
PE
3103{
3104 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3105}
bea3348e 3106
738b35cc
FF
3107/**
3108 * netif_wake_subqueue - allow sending packets on subqueue
3109 * @dev: network device
3110 * @queue_index: sub queue index
3111 *
3112 * Resume individual transmit queue of a device with multiple transmit queues.
3113 */
3114static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3115{
3116 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3117
3118 netif_tx_wake_queue(txq);
3119}
f25f4e44 3120
537c00de 3121#ifdef CONFIG_XPS
53af53ae 3122int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3123 u16 index);
537c00de
AD
3124#else
3125static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3126 const struct cpumask *mask,
537c00de
AD
3127 u16 index)
3128{
3129 return 0;
3130}
3131#endif
3132
5605c762
JP
3133u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3134 unsigned int num_tx_queues);
3135
a3d22a68
VZ
3136/*
3137 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3138 * as a distribution range limit for the returned value.
3139 */
3140static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3141 struct sk_buff *skb)
a3d22a68
VZ
3142{
3143 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3144}
3145
bea3348e
SH
3146/**
3147 * netif_is_multiqueue - test if device has multiple transmit queues
3148 * @dev: network device
3149 *
3150 * Check if device has multiple transmit queues
bea3348e 3151 */
4d29515f 3152static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3153{
a02cec21 3154 return dev->num_tx_queues > 1;
f25f4e44 3155}
1da177e4 3156
f629d208 3157int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3158
a953be53 3159#ifdef CONFIG_SYSFS
f629d208 3160int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3161#else
3162static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3163 unsigned int rxq)
3164{
3165 return 0;
3166}
3167#endif
3168
a953be53
MD
3169#ifdef CONFIG_SYSFS
3170static inline unsigned int get_netdev_rx_queue_index(
3171 struct netdev_rx_queue *queue)
3172{
3173 struct net_device *dev = queue->dev;
3174 int index = queue - dev->_rx;
3175
3176 BUG_ON(index >= dev->num_rx_queues);
3177 return index;
3178}
3179#endif
3180
16917b87 3181#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3182int netif_get_num_default_rss_queues(void);
16917b87 3183
e6247027
ED
3184enum skb_free_reason {
3185 SKB_REASON_CONSUMED,
3186 SKB_REASON_DROPPED,
3187};
3188
3189void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3190void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3191
e6247027
ED
3192/*
3193 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3194 * interrupt context or with hardware interrupts being disabled.
3195 * (in_irq() || irqs_disabled())
3196 *
3197 * We provide four helpers that can be used in following contexts :
3198 *
3199 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3200 * replacing kfree_skb(skb)
3201 *
3202 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3203 * Typically used in place of consume_skb(skb) in TX completion path
3204 *
3205 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3206 * replacing kfree_skb(skb)
3207 *
3208 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3209 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3210 */
e6247027
ED
3211static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3212{
3213 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3214}
3215
3216static inline void dev_consume_skb_irq(struct sk_buff *skb)
3217{
3218 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3219}
3220
3221static inline void dev_kfree_skb_any(struct sk_buff *skb)
3222{
3223 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3224}
3225
3226static inline void dev_consume_skb_any(struct sk_buff *skb)
3227{
3228 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3229}
1da177e4 3230
f629d208
JP
3231int netif_rx(struct sk_buff *skb);
3232int netif_rx_ni(struct sk_buff *skb);
04eb4489 3233int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3234gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3235void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3236struct sk_buff *napi_get_frags(struct napi_struct *napi);
3237gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3238struct packet_offload *gro_find_receive_by_type(__be16 type);
3239struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3240
3241static inline void napi_free_frags(struct napi_struct *napi)
3242{
3243 kfree_skb(napi->skb);
3244 napi->skb = NULL;
3245}
3246
24b27fc4 3247bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3248int netdev_rx_handler_register(struct net_device *dev,
3249 rx_handler_func_t *rx_handler,
3250 void *rx_handler_data);
3251void netdev_rx_handler_unregister(struct net_device *dev);
3252
3253bool dev_valid_name(const char *name);
3254int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3255int dev_ethtool(struct net *net, struct ifreq *);
3256unsigned int dev_get_flags(const struct net_device *);
3257int __dev_change_flags(struct net_device *, unsigned int flags);
3258int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3259void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3260 unsigned int gchanges);
f629d208
JP
3261int dev_change_name(struct net_device *, const char *);
3262int dev_set_alias(struct net_device *, const char *, size_t);
3263int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3264int dev_set_mtu(struct net_device *, int);
3265void dev_set_group(struct net_device *, int);
3266int dev_set_mac_address(struct net_device *, struct sockaddr *);
3267int dev_change_carrier(struct net_device *, bool new_carrier);
3268int dev_get_phys_port_id(struct net_device *dev,
02637fce 3269 struct netdev_phys_item_id *ppid);
db24a904
DA
3270int dev_get_phys_port_name(struct net_device *dev,
3271 char *name, size_t len);
d746d707 3272int dev_change_proto_down(struct net_device *dev, bool proto_down);
85de8576 3273int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
55a93b3e 3274struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3275struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3276 struct netdev_queue *txq, int *ret);
a0265d28 3277int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3278int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3279bool is_skb_forwardable(const struct net_device *dev,
3280 const struct sk_buff *skb);
1da177e4 3281
4e3264d2
MKL
3282static __always_inline int ____dev_forward_skb(struct net_device *dev,
3283 struct sk_buff *skb)
3284{
3285 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3286 unlikely(!is_skb_forwardable(dev, skb))) {
3287 atomic_long_inc(&dev->rx_dropped);
3288 kfree_skb(skb);
3289 return NET_RX_DROP;
3290 }
3291
3292 skb_scrub_packet(skb, true);
3293 skb->priority = 0;
3294 return 0;
3295}
3296
74b20582
DA
3297void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3298
20380731 3299extern int netdev_budget;
1da177e4
LT
3300
3301/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3302void netdev_run_todo(void);
1da177e4 3303
bea3348e
SH
3304/**
3305 * dev_put - release reference to device
3306 * @dev: network device
3307 *
9ef4429b 3308 * Release reference to device to allow it to be freed.
bea3348e 3309 */
1da177e4
LT
3310static inline void dev_put(struct net_device *dev)
3311{
933393f5 3312 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3313}
3314
bea3348e
SH
3315/**
3316 * dev_hold - get reference to device
3317 * @dev: network device
3318 *
9ef4429b 3319 * Hold reference to device to keep it from being freed.
bea3348e 3320 */
15333061
SH
3321static inline void dev_hold(struct net_device *dev)
3322{
933393f5 3323 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3324}
1da177e4
LT
3325
3326/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3327 * and _off may be called from IRQ context, but it is caller
3328 * who is responsible for serialization of these calls.
b00055aa
SR
3329 *
3330 * The name carrier is inappropriate, these functions should really be
3331 * called netif_lowerlayer_*() because they represent the state of any
3332 * kind of lower layer not just hardware media.
1da177e4
LT
3333 */
3334
f629d208
JP
3335void linkwatch_init_dev(struct net_device *dev);
3336void linkwatch_fire_event(struct net_device *dev);
3337void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3338
bea3348e
SH
3339/**
3340 * netif_carrier_ok - test if carrier present
3341 * @dev: network device
3342 *
3343 * Check if carrier is present on device
3344 */
4d29515f 3345static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3346{
3347 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3348}
3349
f629d208 3350unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3351
f629d208 3352void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3353
f629d208 3354void netif_carrier_on(struct net_device *dev);
1da177e4 3355
f629d208 3356void netif_carrier_off(struct net_device *dev);
1da177e4 3357
bea3348e
SH
3358/**
3359 * netif_dormant_on - mark device as dormant.
3360 * @dev: network device
3361 *
3362 * Mark device as dormant (as per RFC2863).
3363 *
3364 * The dormant state indicates that the relevant interface is not
3365 * actually in a condition to pass packets (i.e., it is not 'up') but is
3366 * in a "pending" state, waiting for some external event. For "on-
3367 * demand" interfaces, this new state identifies the situation where the
3368 * interface is waiting for events to place it in the up state.
bea3348e 3369 */
b00055aa
SR
3370static inline void netif_dormant_on(struct net_device *dev)
3371{
3372 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3373 linkwatch_fire_event(dev);
3374}
3375
bea3348e
SH
3376/**
3377 * netif_dormant_off - set device as not dormant.
3378 * @dev: network device
3379 *
3380 * Device is not in dormant state.
3381 */
b00055aa
SR
3382static inline void netif_dormant_off(struct net_device *dev)
3383{
3384 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3385 linkwatch_fire_event(dev);
3386}
3387
bea3348e
SH
3388/**
3389 * netif_dormant - test if carrier present
3390 * @dev: network device
3391 *
3392 * Check if carrier is present on device
3393 */
4d29515f 3394static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3395{
3396 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3397}
3398
3399
bea3348e
SH
3400/**
3401 * netif_oper_up - test if device is operational
3402 * @dev: network device
3403 *
3404 * Check if carrier is operational
3405 */
4d29515f 3406static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3407{
b00055aa
SR
3408 return (dev->operstate == IF_OPER_UP ||
3409 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3410}
3411
bea3348e
SH
3412/**
3413 * netif_device_present - is device available or removed
3414 * @dev: network device
3415 *
3416 * Check if device has not been removed from system.
3417 */
4d29515f 3418static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3419{
3420 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3421}
3422
f629d208 3423void netif_device_detach(struct net_device *dev);
1da177e4 3424
f629d208 3425void netif_device_attach(struct net_device *dev);
1da177e4
LT
3426
3427/*
3428 * Network interface message level settings
3429 */
1da177e4
LT
3430
3431enum {
3432 NETIF_MSG_DRV = 0x0001,
3433 NETIF_MSG_PROBE = 0x0002,
3434 NETIF_MSG_LINK = 0x0004,
3435 NETIF_MSG_TIMER = 0x0008,
3436 NETIF_MSG_IFDOWN = 0x0010,
3437 NETIF_MSG_IFUP = 0x0020,
3438 NETIF_MSG_RX_ERR = 0x0040,
3439 NETIF_MSG_TX_ERR = 0x0080,
3440 NETIF_MSG_TX_QUEUED = 0x0100,
3441 NETIF_MSG_INTR = 0x0200,
3442 NETIF_MSG_TX_DONE = 0x0400,
3443 NETIF_MSG_RX_STATUS = 0x0800,
3444 NETIF_MSG_PKTDATA = 0x1000,
3445 NETIF_MSG_HW = 0x2000,
3446 NETIF_MSG_WOL = 0x4000,
3447};
3448
3449#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3450#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3451#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3452#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3453#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3454#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3455#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3456#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3457#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3458#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3459#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3460#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3461#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3462#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3463#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3464
3465static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3466{
3467 /* use default */
3468 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3469 return default_msg_enable_bits;
3470 if (debug_value == 0) /* no output */
3471 return 0;
3472 /* set low N bits */
3473 return (1 << debug_value) - 1;
3474}
3475
c773e847 3476static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3477{
c773e847
DM
3478 spin_lock(&txq->_xmit_lock);
3479 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3480}
3481
5a717f4f
MT
3482static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3483{
3484 __acquire(&txq->_xmit_lock);
3485 return true;
3486}
3487
3488static inline void __netif_tx_release(struct netdev_queue *txq)
3489{
3490 __release(&txq->_xmit_lock);
3491}
3492
fd2ea0a7
DM
3493static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3494{
3495 spin_lock_bh(&txq->_xmit_lock);
3496 txq->xmit_lock_owner = smp_processor_id();
3497}
3498
4d29515f 3499static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3500{
4d29515f 3501 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3502 if (likely(ok))
3503 txq->xmit_lock_owner = smp_processor_id();
3504 return ok;
3505}
3506
3507static inline void __netif_tx_unlock(struct netdev_queue *txq)
3508{
3509 txq->xmit_lock_owner = -1;
3510 spin_unlock(&txq->_xmit_lock);
3511}
3512
3513static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3514{
3515 txq->xmit_lock_owner = -1;
3516 spin_unlock_bh(&txq->_xmit_lock);
3517}
3518
08baf561
ED
3519static inline void txq_trans_update(struct netdev_queue *txq)
3520{
3521 if (txq->xmit_lock_owner != -1)
3522 txq->trans_start = jiffies;
3523}
3524
ba162f8e
FW
3525/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3526static inline void netif_trans_update(struct net_device *dev)
3527{
9b36627a
FW
3528 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3529
3530 if (txq->trans_start != jiffies)
3531 txq->trans_start = jiffies;
ba162f8e
FW
3532}
3533
d29f749e
DJ
3534/**
3535 * netif_tx_lock - grab network device transmit lock
3536 * @dev: network device
d29f749e
DJ
3537 *
3538 * Get network device transmit lock
3539 */
22dd7495
JHS
3540static inline void netif_tx_lock(struct net_device *dev)
3541{
e8a0464c 3542 unsigned int i;
c3f26a26 3543 int cpu;
c773e847 3544
c3f26a26
DM
3545 spin_lock(&dev->tx_global_lock);
3546 cpu = smp_processor_id();
e8a0464c
DM
3547 for (i = 0; i < dev->num_tx_queues; i++) {
3548 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3549
3550 /* We are the only thread of execution doing a
3551 * freeze, but we have to grab the _xmit_lock in
3552 * order to synchronize with threads which are in
3553 * the ->hard_start_xmit() handler and already
3554 * checked the frozen bit.
3555 */
e8a0464c 3556 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3557 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3558 __netif_tx_unlock(txq);
e8a0464c 3559 }
932ff279
HX
3560}
3561
3562static inline void netif_tx_lock_bh(struct net_device *dev)
3563{
e8a0464c
DM
3564 local_bh_disable();
3565 netif_tx_lock(dev);
932ff279
HX
3566}
3567
932ff279
HX
3568static inline void netif_tx_unlock(struct net_device *dev)
3569{
e8a0464c
DM
3570 unsigned int i;
3571
3572 for (i = 0; i < dev->num_tx_queues; i++) {
3573 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3574
c3f26a26
DM
3575 /* No need to grab the _xmit_lock here. If the
3576 * queue is not stopped for another reason, we
3577 * force a schedule.
3578 */
3579 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3580 netif_schedule_queue(txq);
c3f26a26
DM
3581 }
3582 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3583}
3584
3585static inline void netif_tx_unlock_bh(struct net_device *dev)
3586{
e8a0464c
DM
3587 netif_tx_unlock(dev);
3588 local_bh_enable();
932ff279
HX
3589}
3590
c773e847 3591#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3592 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3593 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3594 } else { \
3595 __netif_tx_acquire(txq); \
22dd7495
JHS
3596 } \
3597}
3598
5efeac44
EB
3599#define HARD_TX_TRYLOCK(dev, txq) \
3600 (((dev->features & NETIF_F_LLTX) == 0) ? \
3601 __netif_tx_trylock(txq) : \
5a717f4f 3602 __netif_tx_acquire(txq))
5efeac44 3603
c773e847 3604#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3605 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3606 __netif_tx_unlock(txq); \
5a717f4f
MT
3607 } else { \
3608 __netif_tx_release(txq); \
22dd7495
JHS
3609 } \
3610}
3611
1da177e4
LT
3612static inline void netif_tx_disable(struct net_device *dev)
3613{
fd2ea0a7 3614 unsigned int i;
c3f26a26 3615 int cpu;
fd2ea0a7 3616
c3f26a26
DM
3617 local_bh_disable();
3618 cpu = smp_processor_id();
fd2ea0a7
DM
3619 for (i = 0; i < dev->num_tx_queues; i++) {
3620 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3621
3622 __netif_tx_lock(txq, cpu);
fd2ea0a7 3623 netif_tx_stop_queue(txq);
c3f26a26 3624 __netif_tx_unlock(txq);
fd2ea0a7 3625 }
c3f26a26 3626 local_bh_enable();
1da177e4
LT
3627}
3628
e308a5d8
DM
3629static inline void netif_addr_lock(struct net_device *dev)
3630{
3631 spin_lock(&dev->addr_list_lock);
3632}
3633
2429f7ac
JP
3634static inline void netif_addr_lock_nested(struct net_device *dev)
3635{
25175ba5
VY
3636 int subclass = SINGLE_DEPTH_NESTING;
3637
3638 if (dev->netdev_ops->ndo_get_lock_subclass)
3639 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3640
3641 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3642}
3643
e308a5d8
DM
3644static inline void netif_addr_lock_bh(struct net_device *dev)
3645{
3646 spin_lock_bh(&dev->addr_list_lock);
3647}
3648
3649static inline void netif_addr_unlock(struct net_device *dev)
3650{
3651 spin_unlock(&dev->addr_list_lock);
3652}
3653
3654static inline void netif_addr_unlock_bh(struct net_device *dev)
3655{
3656 spin_unlock_bh(&dev->addr_list_lock);
3657}
3658
f001fde5 3659/*
31278e71 3660 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3661 * rcu_read_lock held.
3662 */
3663#define for_each_dev_addr(dev, ha) \
31278e71 3664 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3665
1da177e4
LT
3666/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3667
f629d208 3668void ether_setup(struct net_device *dev);
1da177e4
LT
3669
3670/* Support for loadable net-drivers */
f629d208 3671struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3672 unsigned char name_assign_type,
f629d208
JP
3673 void (*setup)(struct net_device *),
3674 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3675#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3676 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3677
c835a677
TG
3678#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3679 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3680 count)
36909ea4 3681
f629d208
JP
3682int register_netdev(struct net_device *dev);
3683void unregister_netdev(struct net_device *dev);
f001fde5 3684
22bedad3 3685/* General hardware address lists handling functions */
f629d208
JP
3686int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3687 struct netdev_hw_addr_list *from_list, int addr_len);
3688void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3689 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3690int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3691 struct net_device *dev,
3692 int (*sync)(struct net_device *, const unsigned char *),
3693 int (*unsync)(struct net_device *,
3694 const unsigned char *));
3695void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3696 struct net_device *dev,
3697 int (*unsync)(struct net_device *,
3698 const unsigned char *));
f629d208 3699void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3700
f001fde5 3701/* Functions used for device addresses handling */
f629d208
JP
3702int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3703 unsigned char addr_type);
3704int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3705 unsigned char addr_type);
f629d208
JP
3706void dev_addr_flush(struct net_device *dev);
3707int dev_addr_init(struct net_device *dev);
a748ee24
JP
3708
3709/* Functions used for unicast addresses handling */
f629d208
JP
3710int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3711int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3712int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3713int dev_uc_sync(struct net_device *to, struct net_device *from);
3714int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3715void dev_uc_unsync(struct net_device *to, struct net_device *from);
3716void dev_uc_flush(struct net_device *dev);
3717void dev_uc_init(struct net_device *dev);
f001fde5 3718
670e5b8e
AD
3719/**
3720 * __dev_uc_sync - Synchonize device's unicast list
3721 * @dev: device to sync
3722 * @sync: function to call if address should be added
3723 * @unsync: function to call if address should be removed
3724 *
3725 * Add newly added addresses to the interface, and release
3726 * addresses that have been deleted.
5e82b4b2 3727 */
670e5b8e
AD
3728static inline int __dev_uc_sync(struct net_device *dev,
3729 int (*sync)(struct net_device *,
3730 const unsigned char *),
3731 int (*unsync)(struct net_device *,
3732 const unsigned char *))
3733{
3734 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3735}
3736
3737/**
e793c0f7 3738 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3739 * @dev: device to sync
3740 * @unsync: function to call if address should be removed
3741 *
3742 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3743 */
670e5b8e
AD
3744static inline void __dev_uc_unsync(struct net_device *dev,
3745 int (*unsync)(struct net_device *,
3746 const unsigned char *))
3747{
3748 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3749}
3750
22bedad3 3751/* Functions used for multicast addresses handling */
f629d208
JP
3752int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3753int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3754int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3755int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3756int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3757int dev_mc_sync(struct net_device *to, struct net_device *from);
3758int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3759void dev_mc_unsync(struct net_device *to, struct net_device *from);
3760void dev_mc_flush(struct net_device *dev);
3761void dev_mc_init(struct net_device *dev);
f001fde5 3762
670e5b8e
AD
3763/**
3764 * __dev_mc_sync - Synchonize device's multicast list
3765 * @dev: device to sync
3766 * @sync: function to call if address should be added
3767 * @unsync: function to call if address should be removed
3768 *
3769 * Add newly added addresses to the interface, and release
3770 * addresses that have been deleted.
5e82b4b2 3771 */
670e5b8e
AD
3772static inline int __dev_mc_sync(struct net_device *dev,
3773 int (*sync)(struct net_device *,
3774 const unsigned char *),
3775 int (*unsync)(struct net_device *,
3776 const unsigned char *))
3777{
3778 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3779}
3780
3781/**
e793c0f7 3782 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3783 * @dev: device to sync
3784 * @unsync: function to call if address should be removed
3785 *
3786 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3787 */
670e5b8e
AD
3788static inline void __dev_mc_unsync(struct net_device *dev,
3789 int (*unsync)(struct net_device *,
3790 const unsigned char *))
3791{
3792 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3793}
3794
4417da66 3795/* Functions used for secondary unicast and multicast support */
f629d208
JP
3796void dev_set_rx_mode(struct net_device *dev);
3797void __dev_set_rx_mode(struct net_device *dev);
3798int dev_set_promiscuity(struct net_device *dev, int inc);
3799int dev_set_allmulti(struct net_device *dev, int inc);
3800void netdev_state_change(struct net_device *dev);
3801void netdev_notify_peers(struct net_device *dev);
3802void netdev_features_change(struct net_device *dev);
1da177e4 3803/* Load a device via the kmod */
f629d208
JP
3804void dev_load(struct net *net, const char *name);
3805struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3806 struct rtnl_link_stats64 *storage);
3807void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3808 const struct net_device_stats *netdev_stats);
eeda3fd6 3809
1da177e4 3810extern int netdev_max_backlog;
3b098e2d 3811extern int netdev_tstamp_prequeue;
1da177e4 3812extern int weight_p;
3d48b53f
MT
3813extern int dev_weight_rx_bias;
3814extern int dev_weight_tx_bias;
3815extern int dev_rx_weight;
3816extern int dev_tx_weight;
9ff162a8 3817
f629d208 3818bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3819struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3820 struct list_head **iter);
f629d208
JP
3821struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3822 struct list_head **iter);
8b5be856 3823
44a40855
VY
3824/* iterate through upper list, must be called under RCU read lock */
3825#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3826 for (iter = &(dev)->adj_list.upper, \
3827 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3828 updev; \
3829 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3830
1a3f060c
DA
3831int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3832 int (*fn)(struct net_device *upper_dev,
3833 void *data),
3834 void *data);
3835
3836bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3837 struct net_device *upper_dev);
3838
f629d208
JP
3839void *netdev_lower_get_next_private(struct net_device *dev,
3840 struct list_head **iter);
3841void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3842 struct list_head **iter);
31088a11
VF
3843
3844#define netdev_for_each_lower_private(dev, priv, iter) \
3845 for (iter = (dev)->adj_list.lower.next, \
3846 priv = netdev_lower_get_next_private(dev, &(iter)); \
3847 priv; \
3848 priv = netdev_lower_get_next_private(dev, &(iter)))
3849
3850#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3851 for (iter = &(dev)->adj_list.lower, \
3852 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3853 priv; \
3854 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3855
4085ebe8
VY
3856void *netdev_lower_get_next(struct net_device *dev,
3857 struct list_head **iter);
7ce856aa 3858
4085ebe8 3859#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3860 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3861 ldev = netdev_lower_get_next(dev, &(iter)); \
3862 ldev; \
3863 ldev = netdev_lower_get_next(dev, &(iter)))
3864
7ce856aa
JP
3865struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3866 struct list_head **iter);
3867struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3868 struct list_head **iter);
3869
1a3f060c
DA
3870int netdev_walk_all_lower_dev(struct net_device *dev,
3871 int (*fn)(struct net_device *lower_dev,
3872 void *data),
3873 void *data);
3874int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3875 int (*fn)(struct net_device *lower_dev,
3876 void *data),
3877 void *data);
3878
f629d208 3879void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3880void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3881struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3882struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3883int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3884int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3885 struct net_device *upper_dev,
29bf24af 3886 void *upper_priv, void *upper_info);
f629d208
JP
3887void netdev_upper_dev_unlink(struct net_device *dev,
3888 struct net_device *upper_dev);
5bb025fa 3889void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3890void *netdev_lower_dev_get_private(struct net_device *dev,
3891 struct net_device *lower_dev);
04d48266
JP
3892void netdev_lower_state_changed(struct net_device *lower_dev,
3893 void *lower_state_info);
960fb622
ED
3894
3895/* RSS keys are 40 or 52 bytes long */
3896#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3897extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3898void netdev_rss_key_fill(void *buffer, size_t len);
3899
952fcfd0 3900int dev_get_nest_level(struct net_device *dev);
f629d208
JP
3901int skb_checksum_help(struct sk_buff *skb);
3902struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3903 netdev_features_t features, bool tx_path);
3904struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3905 netdev_features_t features);
12b0004d 3906
61bd3857
MS
3907struct netdev_bonding_info {
3908 ifslave slave;
3909 ifbond master;
3910};
3911
3912struct netdev_notifier_bonding_info {
3913 struct netdev_notifier_info info; /* must be first */
3914 struct netdev_bonding_info bonding_info;
3915};
3916
3917void netdev_bonding_info_change(struct net_device *dev,
3918 struct netdev_bonding_info *bonding_info);
3919
12b0004d
CW
3920static inline
3921struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3922{
3923 return __skb_gso_segment(skb, features, true);
3924}
53d6471c 3925__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3926
3927static inline bool can_checksum_protocol(netdev_features_t features,
3928 __be16 protocol)
3929{
c8cd0989
TH
3930 if (protocol == htons(ETH_P_FCOE))
3931 return !!(features & NETIF_F_FCOE_CRC);
3932
3933 /* Assume this is an IP checksum (not SCTP CRC) */
3934
3935 if (features & NETIF_F_HW_CSUM) {
3936 /* Can checksum everything */
3937 return true;
3938 }
3939
3940 switch (protocol) {
3941 case htons(ETH_P_IP):
3942 return !!(features & NETIF_F_IP_CSUM);
3943 case htons(ETH_P_IPV6):
3944 return !!(features & NETIF_F_IPV6_CSUM);
3945 default:
3946 return false;
3947 }
ec5f0615 3948}
12b0004d 3949
fb286bb2 3950#ifdef CONFIG_BUG
f629d208 3951void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3952#else
3953static inline void netdev_rx_csum_fault(struct net_device *dev)
3954{
3955}
3956#endif
1da177e4 3957/* rx skb timestamps */
f629d208
JP
3958void net_enable_timestamp(void);
3959void net_disable_timestamp(void);
1da177e4 3960
20380731 3961#ifdef CONFIG_PROC_FS
f629d208 3962int __init dev_proc_init(void);
900ff8c6
CW
3963#else
3964#define dev_proc_init() 0
20380731
ACM
3965#endif
3966
4798248e 3967static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3968 struct sk_buff *skb, struct net_device *dev,
3969 bool more)
4798248e 3970{
fa2dbdc2 3971 skb->xmit_more = more ? 1 : 0;
0b725a2c 3972 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3973}
3974
10b3ad8c 3975static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3976 struct netdev_queue *txq, bool more)
4798248e
DM
3977{
3978 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3979 int rc;
4798248e 3980
fa2dbdc2 3981 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3982 if (rc == NETDEV_TX_OK)
3983 txq_trans_update(txq);
3984
3985 return rc;
4798248e
DM
3986}
3987
42a2d923
LT
3988int netdev_class_create_file_ns(struct class_attribute *class_attr,
3989 const void *ns);
3990void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3991 const void *ns);
58292cbe
TH
3992
3993static inline int netdev_class_create_file(struct class_attribute *class_attr)
3994{
3995 return netdev_class_create_file_ns(class_attr, NULL);
3996}
3997
3998static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3999{
4000 netdev_class_remove_file_ns(class_attr, NULL);
4001}
b8a9787e 4002
04600794
JB
4003extern struct kobj_ns_type_operations net_ns_type_operations;
4004
f629d208 4005const char *netdev_drivername(const struct net_device *dev);
6579e57b 4006
f629d208 4007void linkwatch_run_queue(void);
20380731 4008
da08143b
MK
4009static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4010 netdev_features_t f2)
4011{
c8cd0989
TH
4012 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4013 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4014 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4015 else
b6a0e72a 4016 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4017 }
da08143b 4018
c8cd0989 4019 return f1 & f2;
da08143b
MK
4020}
4021
c8f44aff
MM
4022static inline netdev_features_t netdev_get_wanted_features(
4023 struct net_device *dev)
5455c699
MM
4024{
4025 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4026}
c8f44aff
MM
4027netdev_features_t netdev_increment_features(netdev_features_t all,
4028 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4029
4030/* Allow TSO being used on stacked device :
4031 * Performing the GSO segmentation before last device
4032 * is a performance improvement.
4033 */
4034static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4035 netdev_features_t mask)
4036{
4037 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4038}
4039
6cb6a27c 4040int __netdev_update_features(struct net_device *dev);
5455c699 4041void netdev_update_features(struct net_device *dev);
afe12cc8 4042void netdev_change_features(struct net_device *dev);
7f353bf2 4043
fc4a7489
PM
4044void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4045 struct net_device *dev);
4046
e38f3025
TM
4047netdev_features_t passthru_features_check(struct sk_buff *skb,
4048 struct net_device *dev,
4049 netdev_features_t features);
c1e756bf 4050netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4051
4d29515f 4052static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4053{
7b748340 4054 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4055
4056 /* check flags correspondence */
4057 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4058 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4059 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4060 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4061 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4062 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4063 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4064 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4065 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4066 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4067 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4068 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4069 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4070 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4071 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4072 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4073 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0345e186 4074
d6b4991a 4075 return (features & feature) == feature;
576a30eb
HX
4076}
4077
4d29515f 4078static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4079{
278b2513 4080 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4081 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4082}
4083
8b86a61d 4084static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4085 netdev_features_t features)
7967168c 4086{
fc741216 4087 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4088 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4089 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4090}
4091
82cc1a7a
PWJ
4092static inline void netif_set_gso_max_size(struct net_device *dev,
4093 unsigned int size)
4094{
4095 dev->gso_max_size = size;
4096}
4097
7a7ffbab
WCC
4098static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4099 int pulled_hlen, u16 mac_offset,
4100 int mac_len)
4101{
4102 skb->protocol = protocol;
4103 skb->encapsulation = 1;
4104 skb_push(skb, pulled_hlen);
4105 skb_reset_transport_header(skb);
4106 skb->mac_header = mac_offset;
4107 skb->network_header = skb->mac_header + mac_len;
4108 skb->mac_len = mac_len;
4109}
4110
3c175784
SD
4111static inline bool netif_is_macsec(const struct net_device *dev)
4112{
4113 return dev->priv_flags & IFF_MACSEC;
4114}
4115
b618aaa9 4116static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4117{
4118 return dev->priv_flags & IFF_MACVLAN;
4119}
4120
b618aaa9 4121static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4122{
4123 return dev->priv_flags & IFF_MACVLAN_PORT;
4124}
4125
b618aaa9 4126static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4127{
4128 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4129}
4130
b618aaa9 4131static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4132{
4133 return dev->priv_flags & IFF_IPVLAN_MASTER;
4134}
4135
b618aaa9 4136static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4137{
4138 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4139}
4140
b618aaa9 4141static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4142{
4143 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4144}
4145
3bdc0eba
BG
4146static inline bool netif_supports_nofcs(struct net_device *dev)
4147{
4148 return dev->priv_flags & IFF_SUPP_NOFCS;
4149}
4150
007979ea 4151static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4152{
007979ea 4153 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4154}
4155
fee6d4c7
DA
4156static inline bool netif_is_l3_slave(const struct net_device *dev)
4157{
4158 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4159}
4160
0894ae3f
JP
4161static inline bool netif_is_bridge_master(const struct net_device *dev)
4162{
4163 return dev->priv_flags & IFF_EBRIDGE;
4164}
4165
28f9ee22
VY
4166static inline bool netif_is_bridge_port(const struct net_device *dev)
4167{
4168 return dev->priv_flags & IFF_BRIDGE_PORT;
4169}
4170
35d4e172
JP
4171static inline bool netif_is_ovs_master(const struct net_device *dev)
4172{
4173 return dev->priv_flags & IFF_OPENVSWITCH;
4174}
4175
b618aaa9 4176static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4177{
4178 return dev->priv_flags & IFF_TEAM;
4179}
4180
b618aaa9 4181static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4182{
4183 return dev->priv_flags & IFF_TEAM_PORT;
4184}
4185
b618aaa9 4186static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4187{
4188 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4189}
4190
b618aaa9 4191static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4192{
4193 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4194}
4195
d4ab4286
KJ
4196static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4197{
4198 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4199}
4200
02875878
ED
4201/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4202static inline void netif_keep_dst(struct net_device *dev)
4203{
4204 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4205}
4206
18d3df3e
PA
4207/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4208static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4209{
4210 /* TODO: reserve and use an additional IFF bit, if we get more users */
4211 return dev->priv_flags & IFF_MACSEC;
4212}
4213
505d4f73 4214extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4215
571ba423
JP
4216/* Logging, debugging and troubleshooting/diagnostic helpers. */
4217
4218/* netdev_printk helpers, similar to dev_printk */
4219
4220static inline const char *netdev_name(const struct net_device *dev)
4221{
c6f854d5
VF
4222 if (!dev->name[0] || strchr(dev->name, '%'))
4223 return "(unnamed net_device)";
571ba423
JP
4224 return dev->name;
4225}
4226
ccc7f496
VF
4227static inline const char *netdev_reg_state(const struct net_device *dev)
4228{
4229 switch (dev->reg_state) {
4230 case NETREG_UNINITIALIZED: return " (uninitialized)";
4231 case NETREG_REGISTERED: return "";
4232 case NETREG_UNREGISTERING: return " (unregistering)";
4233 case NETREG_UNREGISTERED: return " (unregistered)";
4234 case NETREG_RELEASED: return " (released)";
4235 case NETREG_DUMMY: return " (dummy)";
4236 }
4237
4238 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4239 return " (unknown)";
4240}
4241
f629d208 4242__printf(3, 4)
6ea754eb
JP
4243void netdev_printk(const char *level, const struct net_device *dev,
4244 const char *format, ...);
f629d208 4245__printf(2, 3)
6ea754eb 4246void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4247__printf(2, 3)
6ea754eb 4248void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4249__printf(2, 3)
6ea754eb 4250void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4251__printf(2, 3)
6ea754eb 4252void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4253__printf(2, 3)
6ea754eb 4254void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4255__printf(2, 3)
6ea754eb 4256void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4257__printf(2, 3)
6ea754eb 4258void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4259
8909c9ad
VK
4260#define MODULE_ALIAS_NETDEV(device) \
4261 MODULE_ALIAS("netdev-" device)
4262
b558c96f 4263#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4264#define netdev_dbg(__dev, format, args...) \
4265do { \
ffa10cb4 4266 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4267} while (0)
b558c96f
JC
4268#elif defined(DEBUG)
4269#define netdev_dbg(__dev, format, args...) \
4270 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4271#else
4272#define netdev_dbg(__dev, format, args...) \
4273({ \
4274 if (0) \
4275 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4276})
4277#endif
4278
4279#if defined(VERBOSE_DEBUG)
4280#define netdev_vdbg netdev_dbg
4281#else
4282
4283#define netdev_vdbg(dev, format, args...) \
4284({ \
4285 if (0) \
4286 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4287 0; \
4288})
4289#endif
4290
4291/*
4292 * netdev_WARN() acts like dev_printk(), but with the key difference
4293 * of using a WARN/WARN_ON to get the message out, including the
4294 * file/line information and a backtrace.
4295 */
4296#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4297 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4298 netdev_reg_state(dev), ##args)
571ba423 4299
b3d95c5c
JP
4300/* netif printk helpers, similar to netdev_printk */
4301
4302#define netif_printk(priv, type, level, dev, fmt, args...) \
4303do { \
4304 if (netif_msg_##type(priv)) \
4305 netdev_printk(level, (dev), fmt, ##args); \
4306} while (0)
4307
f45f4321
JP
4308#define netif_level(level, priv, type, dev, fmt, args...) \
4309do { \
4310 if (netif_msg_##type(priv)) \
4311 netdev_##level(dev, fmt, ##args); \
4312} while (0)
4313
b3d95c5c 4314#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4315 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4316#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4317 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4318#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4319 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4320#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4321 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4322#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4323 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4324#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4325 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4326#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4327 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4328
0053ea9c 4329#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4330#define netif_dbg(priv, type, netdev, format, args...) \
4331do { \
4332 if (netif_msg_##type(priv)) \
b5fb0a03 4333 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4334} while (0)
0053ea9c
JP
4335#elif defined(DEBUG)
4336#define netif_dbg(priv, type, dev, format, args...) \
4337 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4338#else
4339#define netif_dbg(priv, type, dev, format, args...) \
4340({ \
4341 if (0) \
4342 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4343 0; \
4344})
4345#endif
4346
f617f276
EC
4347/* if @cond then downgrade to debug, else print at @level */
4348#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4349 do { \
4350 if (cond) \
4351 netif_dbg(priv, type, netdev, fmt, ##args); \
4352 else \
4353 netif_ ## level(priv, type, netdev, fmt, ##args); \
4354 } while (0)
4355
b3d95c5c 4356#if defined(VERBOSE_DEBUG)
bcfcc450 4357#define netif_vdbg netif_dbg
b3d95c5c
JP
4358#else
4359#define netif_vdbg(priv, type, dev, format, args...) \
4360({ \
4361 if (0) \
a4ed89cb 4362 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4363 0; \
4364})
4365#endif
571ba423 4366
900ff8c6
CW
4367/*
4368 * The list of packet types we will receive (as opposed to discard)
4369 * and the routines to invoke.
4370 *
4371 * Why 16. Because with 16 the only overlap we get on a hash of the
4372 * low nibble of the protocol value is RARP/SNAP/X.25.
4373 *
4374 * NOTE: That is no longer true with the addition of VLAN tags. Not
4375 * sure which should go first, but I bet it won't make much
4376 * difference if we are running VLANs. The good news is that
4377 * this protocol won't be in the list unless compiled in, so
4378 * the average user (w/out VLANs) will not be adversely affected.
4379 * --BLG
4380 *
4381 * 0800 IP
4382 * 8100 802.1Q VLAN
4383 * 0001 802.3
4384 * 0002 AX.25
4385 * 0004 802.2
4386 * 8035 RARP
4387 * 0005 SNAP
4388 * 0805 X.25
4389 * 0806 ARP
4390 * 8137 IPX
4391 * 0009 Localtalk
4392 * 86DD IPv6
4393 */
4394#define PTYPE_HASH_SIZE (16)
4395#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4396
385a154c 4397#endif /* _LINUX_NETDEVICE_H */