net: sched: change return value of ndo_setup_tc for driver supporting mqprio only
[linux-block.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;
814abfab 69struct xdp_buff;
1da177e4 70
f629d208
JP
71void netdev_set_default_ethtool_ops(struct net_device *dev,
72 const struct ethtool_ops *ops);
d07d7507 73
9a1654ba
JP
74/* Backlog congestion levels */
75#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
76#define NET_RX_DROP 1 /* packet dropped */
77
572a9d7b
PM
78/*
79 * Transmit return codes: transmit return codes originate from three different
80 * namespaces:
81 *
82 * - qdisc return codes
83 * - driver transmit return codes
84 * - errno values
85 *
86 * Drivers are allowed to return any one of those in their hard_start_xmit()
87 * function. Real network devices commonly used with qdiscs should only return
88 * the driver transmit return codes though - when qdiscs are used, the actual
89 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
90 * higher layers. Virtual network devices transmit synchronously; in this case
91 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
92 * others are propagated to higher layers.
93 */
94
95/* qdisc ->enqueue() return codes. */
96#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
97#define NET_XMIT_DROP 0x01 /* skb dropped */
98#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 99#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 100
b9df3cb8
GR
101/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 104#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
105#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
dc1f8bf6 107/* Driver transmit return codes */
9a1654ba 108#define NETDEV_TX_MASK 0xf0
572a9d7b 109
dc1f8bf6 110enum netdev_tx {
572a9d7b 111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
114};
115typedef enum netdev_tx netdev_tx_t;
116
9a1654ba
JP
117/*
118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 */
121static inline bool dev_xmit_complete(int rc)
122{
123 /*
124 * Positive cases with an skb consumed by a driver:
125 * - successful transmission (rc == NETDEV_TX_OK)
126 * - error while transmitting (rc < 0)
127 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 */
129 if (likely(rc < NET_XMIT_MASK))
130 return true;
131
132 return false;
133}
134
1da177e4 135/*
5e82b4b2 136 * Compute the worst-case header length according to the protocols
1da177e4
LT
137 * used.
138 */
fe2918b0 139
c0eb4540
KS
140#if defined(CONFIG_HYPERV_NET)
141# define LL_MAX_HEADER 128
142#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
143# if defined(CONFIG_MAC80211_MESH)
144# define LL_MAX_HEADER 128
145# else
146# define LL_MAX_HEADER 96
147# endif
1da177e4 148#else
8388e3da 149# define LL_MAX_HEADER 32
1da177e4
LT
150#endif
151
d11ead75
BH
152#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
153 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
154#define MAX_HEADER LL_MAX_HEADER
155#else
156#define MAX_HEADER (LL_MAX_HEADER + 48)
157#endif
158
159/*
be1f3c2c
BH
160 * Old network device statistics. Fields are native words
161 * (unsigned long) so they can be read and written atomically.
1da177e4 162 */
fe2918b0 163
d94d9fee 164struct net_device_stats {
3cfde79c
BH
165 unsigned long rx_packets;
166 unsigned long tx_packets;
167 unsigned long rx_bytes;
168 unsigned long tx_bytes;
169 unsigned long rx_errors;
170 unsigned long tx_errors;
171 unsigned long rx_dropped;
172 unsigned long tx_dropped;
173 unsigned long multicast;
1da177e4 174 unsigned long collisions;
1da177e4 175 unsigned long rx_length_errors;
3cfde79c
BH
176 unsigned long rx_over_errors;
177 unsigned long rx_crc_errors;
178 unsigned long rx_frame_errors;
179 unsigned long rx_fifo_errors;
180 unsigned long rx_missed_errors;
1da177e4
LT
181 unsigned long tx_aborted_errors;
182 unsigned long tx_carrier_errors;
183 unsigned long tx_fifo_errors;
184 unsigned long tx_heartbeat_errors;
185 unsigned long tx_window_errors;
1da177e4
LT
186 unsigned long rx_compressed;
187 unsigned long tx_compressed;
188};
189
1da177e4
LT
190
191#include <linux/cache.h>
192#include <linux/skbuff.h>
193
adc9300e 194#ifdef CONFIG_RPS
c5905afb
IM
195#include <linux/static_key.h>
196extern struct static_key rps_needed;
13bfff25 197extern struct static_key rfs_needed;
adc9300e
ED
198#endif
199
1da177e4
LT
200struct neighbour;
201struct neigh_parms;
202struct sk_buff;
203
f001fde5
JP
204struct netdev_hw_addr {
205 struct list_head list;
206 unsigned char addr[MAX_ADDR_LEN];
207 unsigned char type;
ccffad25
JP
208#define NETDEV_HW_ADDR_T_LAN 1
209#define NETDEV_HW_ADDR_T_SAN 2
210#define NETDEV_HW_ADDR_T_SLAVE 3
211#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 212#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 213 bool global_use;
4cd729b0 214 int sync_cnt;
8f8f103d 215 int refcount;
4543fbef 216 int synced;
f001fde5
JP
217 struct rcu_head rcu_head;
218};
219
31278e71
JP
220struct netdev_hw_addr_list {
221 struct list_head list;
222 int count;
223};
224
22bedad3
JP
225#define netdev_hw_addr_list_count(l) ((l)->count)
226#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
227#define netdev_hw_addr_list_for_each(ha, l) \
228 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 229
22bedad3
JP
230#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
231#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
232#define netdev_for_each_uc_addr(ha, dev) \
233 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 234
22bedad3
JP
235#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
236#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 237#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 238 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 239
d94d9fee 240struct hh_cache {
5b3dc2f3 241 unsigned int hh_len;
3644f0ce 242 seqlock_t hh_lock;
1da177e4
LT
243
244 /* cached hardware header; allow for machine alignment needs. */
245#define HH_DATA_MOD 16
246#define HH_DATA_OFF(__len) \
5ba0eac6 247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
248#define HH_DATA_ALIGN(__len) \
249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251};
252
5e82b4b2 253/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
254 * Alternative is:
255 * dev->hard_header_len ? (dev->hard_header_len +
256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 *
258 * We could use other alignment values, but we must maintain the
259 * relationship HH alignment <= LL alignment.
260 */
261#define LL_RESERVED_SPACE(dev) \
f5184d26 262 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 263#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 264 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 265
3b04ddde
SH
266struct header_ops {
267 int (*create) (struct sk_buff *skb, struct net_device *dev,
268 unsigned short type, const void *daddr,
95c96174 269 const void *saddr, unsigned int len);
3b04ddde 270 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
2793a23a 275 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
276};
277
1da177e4 278/* These flag bits are private to the generic network queueing
5e82b4b2 279 * layer; they may not be explicitly referenced by any other
1da177e4
LT
280 * code.
281 */
282
d94d9fee 283enum netdev_state_t {
1da177e4
LT
284 __LINK_STATE_START,
285 __LINK_STATE_PRESENT,
1da177e4 286 __LINK_STATE_NOCARRIER,
b00055aa
SR
287 __LINK_STATE_LINKWATCH_PENDING,
288 __LINK_STATE_DORMANT,
1da177e4
LT
289};
290
291
292/*
5e82b4b2 293 * This structure holds boot-time configured netdevice settings. They
fe2918b0 294 * are then used in the device probing.
1da177e4
LT
295 */
296struct netdev_boot_setup {
297 char name[IFNAMSIZ];
298 struct ifmap map;
299};
300#define NETDEV_BOOT_SETUP_MAX 8
301
f629d208 302int __init netdev_boot_setup(char *str);
1da177e4 303
bea3348e
SH
304/*
305 * Structure for NAPI scheduling similar to tasklet but with weighting
306 */
307struct napi_struct {
308 /* The poll_list must only be managed by the entity which
309 * changes the state of the NAPI_STATE_SCHED bit. This means
310 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 311 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
312 * can remove from the list right before clearing the bit.
313 */
314 struct list_head poll_list;
315
316 unsigned long state;
317 int weight;
404f7c9e 318 unsigned int gro_count;
bea3348e
SH
319 int (*poll)(struct napi_struct *, int);
320#ifdef CONFIG_NETPOLL
bea3348e 321 int poll_owner;
bea3348e 322#endif
5d38a079 323 struct net_device *dev;
d565b0a1 324 struct sk_buff *gro_list;
5d38a079 325 struct sk_buff *skb;
3b47d303 326 struct hrtimer timer;
404f7c9e 327 struct list_head dev_list;
af12fa6e
ET
328 struct hlist_node napi_hash_node;
329 unsigned int napi_id;
bea3348e
SH
330};
331
d94d9fee 332enum {
bea3348e 333 NAPI_STATE_SCHED, /* Poll is scheduled */
39e6c820 334 NAPI_STATE_MISSED, /* reschedule a napi */
a0a46196 335 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 336 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
337 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
338 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
339 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
340};
341
342enum {
39e6c820
ED
343 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
344 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
345 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
346 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
347 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
348 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
349 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
350};
351
5b252f0c 352enum gro_result {
d1c76af9
HX
353 GRO_MERGED,
354 GRO_MERGED_FREE,
355 GRO_HELD,
356 GRO_NORMAL,
357 GRO_DROP,
25393d3f 358 GRO_CONSUMED,
d1c76af9 359};
5b252f0c 360typedef enum gro_result gro_result_t;
d1c76af9 361
8a4eb573
JP
362/*
363 * enum rx_handler_result - Possible return values for rx_handlers.
364 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365 * further.
366 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
367 * case skb->dev was changed by rx_handler.
368 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 369 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
370 *
371 * rx_handlers are functions called from inside __netif_receive_skb(), to do
372 * special processing of the skb, prior to delivery to protocol handlers.
373 *
374 * Currently, a net_device can only have a single rx_handler registered. Trying
375 * to register a second rx_handler will return -EBUSY.
376 *
377 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
378 * To unregister a rx_handler on a net_device, use
379 * netdev_rx_handler_unregister().
380 *
381 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
382 * do with the skb.
383 *
5e82b4b2 384 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 385 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 386 * the skb to be delivered in some other way.
8a4eb573
JP
387 *
388 * If the rx_handler changed skb->dev, to divert the skb to another
389 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
390 * new device will be called if it exists.
391 *
5e82b4b2 392 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 393 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 394 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 395 *
5e82b4b2 396 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
397 * delivered, it should return RX_HANDLER_PASS.
398 *
399 * A device without a registered rx_handler will behave as if rx_handler
400 * returned RX_HANDLER_PASS.
401 */
402
403enum rx_handler_result {
404 RX_HANDLER_CONSUMED,
405 RX_HANDLER_ANOTHER,
406 RX_HANDLER_EXACT,
407 RX_HANDLER_PASS,
408};
409typedef enum rx_handler_result rx_handler_result_t;
410typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 411
f629d208 412void __napi_schedule(struct napi_struct *n);
bc9ad166 413void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 414
4d29515f 415static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
416{
417 return test_bit(NAPI_STATE_DISABLE, &n->state);
418}
419
39e6c820 420bool napi_schedule_prep(struct napi_struct *n);
bea3348e
SH
421
422/**
423 * napi_schedule - schedule NAPI poll
5e82b4b2 424 * @n: NAPI context
bea3348e
SH
425 *
426 * Schedule NAPI poll routine to be called if it is not already
427 * running.
428 */
429static inline void napi_schedule(struct napi_struct *n)
430{
431 if (napi_schedule_prep(n))
432 __napi_schedule(n);
433}
434
bc9ad166
ED
435/**
436 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 437 * @n: NAPI context
bc9ad166
ED
438 *
439 * Variant of napi_schedule(), assuming hard irqs are masked.
440 */
441static inline void napi_schedule_irqoff(struct napi_struct *n)
442{
443 if (napi_schedule_prep(n))
444 __napi_schedule_irqoff(n);
445}
446
bfe13f54 447/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 448static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
449{
450 if (napi_schedule_prep(napi)) {
451 __napi_schedule(napi);
4d29515f 452 return true;
bfe13f54 453 }
4d29515f 454 return false;
bfe13f54
RD
455}
456
364b6055 457bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
458/**
459 * napi_complete - NAPI processing complete
5e82b4b2 460 * @n: NAPI context
bea3348e
SH
461 *
462 * Mark NAPI processing as complete.
3b47d303 463 * Consider using napi_complete_done() instead.
364b6055 464 * Return false if device should avoid rearming interrupts.
bea3348e 465 */
364b6055 466static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
467{
468 return napi_complete_done(n, 0);
469}
bea3348e 470
af12fa6e
ET
471/**
472 * napi_hash_del - remove a NAPI from global table
5e82b4b2 473 * @napi: NAPI context
af12fa6e 474 *
5e82b4b2 475 * Warning: caller must observe RCU grace period
34cbe27e
ED
476 * before freeing memory containing @napi, if
477 * this function returns true.
93d05d4a 478 * Note: core networking stack automatically calls it
5e82b4b2 479 * from netif_napi_del().
93d05d4a 480 * Drivers might want to call this helper to combine all
5e82b4b2 481 * the needed RCU grace periods into a single one.
af12fa6e 482 */
34cbe27e 483bool napi_hash_del(struct napi_struct *napi);
af12fa6e 484
bea3348e
SH
485/**
486 * napi_disable - prevent NAPI from scheduling
5e82b4b2 487 * @n: NAPI context
bea3348e
SH
488 *
489 * Stop NAPI from being scheduled on this context.
490 * Waits till any outstanding processing completes.
491 */
3b47d303 492void napi_disable(struct napi_struct *n);
bea3348e
SH
493
494/**
495 * napi_enable - enable NAPI scheduling
5e82b4b2 496 * @n: NAPI context
bea3348e
SH
497 *
498 * Resume NAPI from being scheduled on this context.
499 * Must be paired with napi_disable.
500 */
501static inline void napi_enable(struct napi_struct *n)
502{
503 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 504 smp_mb__before_atomic();
bea3348e 505 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 506 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
507}
508
c264c3de
SH
509/**
510 * napi_synchronize - wait until NAPI is not running
5e82b4b2 511 * @n: NAPI context
c264c3de
SH
512 *
513 * Wait until NAPI is done being scheduled on this context.
514 * Waits till any outstanding processing completes but
515 * does not disable future activations.
516 */
517static inline void napi_synchronize(const struct napi_struct *n)
518{
facc432f
AB
519 if (IS_ENABLED(CONFIG_SMP))
520 while (test_bit(NAPI_STATE_SCHED, &n->state))
521 msleep(1);
522 else
523 barrier();
c264c3de 524}
c264c3de 525
d94d9fee 526enum netdev_queue_state_t {
73466498
TH
527 __QUEUE_STATE_DRV_XOFF,
528 __QUEUE_STATE_STACK_XOFF,
c3f26a26 529 __QUEUE_STATE_FROZEN,
79d16385 530};
8e2f1a63
DB
531
532#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
533#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
534#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
535
536#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
537#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 QUEUE_STATE_FROZEN)
539#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
540 QUEUE_STATE_FROZEN)
541
73466498
TH
542/*
543 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
544 * netif_tx_* functions below are used to manipulate this flag. The
545 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
546 * queue independently. The netif_xmit_*stopped functions below are called
547 * to check if the queue has been stopped by the driver or stack (either
548 * of the XOFF bits are set in the state). Drivers should not need to call
549 * netif_xmit*stopped functions, they should only be using netif_tx_*.
550 */
79d16385 551
bb949fbd 552struct netdev_queue {
6a321cb3 553/*
5e82b4b2 554 * read-mostly part
6a321cb3 555 */
bb949fbd 556 struct net_device *dev;
46e5da40 557 struct Qdisc __rcu *qdisc;
b0e1e646 558 struct Qdisc *qdisc_sleeping;
ccf5ff69 559#ifdef CONFIG_SYSFS
1d24eb48
TH
560 struct kobject kobj;
561#endif
f2cd2d3e
ED
562#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
563 int numa_node;
564#endif
c0ef079c
FW
565 unsigned long tx_maxrate;
566 /*
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
569 */
570 unsigned long trans_timeout;
6a321cb3 571/*
5e82b4b2 572 * write-mostly part
6a321cb3
ED
573 */
574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
575 int xmit_lock_owner;
9d21493b 576 /*
9b36627a 577 * Time (in jiffies) of last Tx
9d21493b
ED
578 */
579 unsigned long trans_start;
ccf5ff69 580
114cf580
TH
581 unsigned long state;
582
583#ifdef CONFIG_BQL
584 struct dql dql;
585#endif
e8a0464c 586} ____cacheline_aligned_in_smp;
bb949fbd 587
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 */
2572ac53 777enum tc_setup_type {
16e5cc64 778 TC_SETUP_MQPRIO,
a1b7c5fd 779 TC_SETUP_CLSU32,
5b33f488 780 TC_SETUP_CLSFLOWER,
ade9b658 781 TC_SETUP_CLSMATCHALL,
332ae8e2 782 TC_SETUP_CLSBPF,
16e5cc64
JF
783};
784
a1b7c5fd
JF
785struct tc_cls_u32_offload;
786
16e5cc64 787struct tc_to_netdev {
16e5cc64 788 union {
a1b7c5fd 789 struct tc_cls_u32_offload *cls_u32;
5b33f488 790 struct tc_cls_flower_offload *cls_flower;
b87f7936 791 struct tc_cls_matchall_offload *cls_mall;
332ae8e2 792 struct tc_cls_bpf_offload *cls_bpf;
56f36acd 793 struct tc_mqprio_qopt *mqprio;
16e5cc64
JF
794 };
795};
796
a7862b45
BB
797/* These structures hold the attributes of xdp state that are being passed
798 * to the netdevice through the xdp op.
799 */
800enum xdp_netdev_command {
801 /* Set or clear a bpf program used in the earliest stages of packet
802 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
803 * is responsible for calling bpf_prog_put on any old progs that are
804 * stored. In case of error, the callee need not release the new prog
805 * reference, but on success it takes ownership and must bpf_prog_put
806 * when it is no longer used.
807 */
808 XDP_SETUP_PROG,
ee5d032f 809 XDP_SETUP_PROG_HW,
a7862b45 810 /* Check if a bpf program is set on the device. The callee should
ce158e58
JK
811 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
812 * is equivalent to XDP_ATTACHED_DRV.
a7862b45
BB
813 */
814 XDP_QUERY_PROG,
815};
816
ddf9f970
JK
817struct netlink_ext_ack;
818
a7862b45
BB
819struct netdev_xdp {
820 enum xdp_netdev_command command;
821 union {
822 /* XDP_SETUP_PROG */
ddf9f970 823 struct {
32d60277 824 u32 flags;
ddf9f970
JK
825 struct bpf_prog *prog;
826 struct netlink_ext_ack *extack;
827 };
a7862b45 828 /* XDP_QUERY_PROG */
58038695 829 struct {
ce158e58 830 u8 prog_attached;
58038695
MKL
831 u32 prog_id;
832 };
a7862b45
BB
833 };
834};
16e5cc64 835
d77e38e6
SK
836#ifdef CONFIG_XFRM_OFFLOAD
837struct xfrmdev_ops {
838 int (*xdo_dev_state_add) (struct xfrm_state *x);
839 void (*xdo_dev_state_delete) (struct xfrm_state *x);
840 void (*xdo_dev_state_free) (struct xfrm_state *x);
841 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
842 struct xfrm_state *x);
843};
844#endif
845
d314774c
SH
846/*
847 * This structure defines the management hooks for network devices.
00829823
SH
848 * The following hooks can be defined; unless noted otherwise, they are
849 * optional and can be filled with a null pointer.
d314774c
SH
850 *
851 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
852 * This function is called once when a network device is registered.
853 * The network device can use this for any late stage initialization
854 * or semantic validation. It can fail with an error code which will
855 * be propagated back to register_netdev.
d314774c
SH
856 *
857 * void (*ndo_uninit)(struct net_device *dev);
858 * This function is called when device is unregistered or when registration
859 * fails. It is not called if init fails.
860 *
861 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 862 * This function is called when a network device transitions to the up
d314774c
SH
863 * state.
864 *
865 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 866 * This function is called when a network device transitions to the down
d314774c
SH
867 * state.
868 *
dc1f8bf6
SH
869 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
870 * struct net_device *dev);
00829823 871 * Called when a packet needs to be transmitted.
e79d8429
RR
872 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
873 * the queue before that can happen; it's for obsolete devices and weird
874 * corner cases, but the stack really does a non-trivial amount
875 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 876 * Required; cannot be NULL.
00829823 877 *
1a2a1444
DM
878 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
879 * struct net_device *dev
880 * netdev_features_t features);
881 * Called by core transmit path to determine if device is capable of
882 * performing offload operations on a given packet. This is to give
883 * the device an opportunity to implement any restrictions that cannot
884 * be otherwise expressed by feature flags. The check is called with
885 * the set of features that the stack has calculated and it returns
886 * those the driver believes to be appropriate.
cdba756f 887 *
f663dd9a 888 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 889 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 890 * Called to decide which queue to use when device supports multiple
00829823
SH
891 * transmit queues.
892 *
d314774c
SH
893 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
894 * This function is called to allow device receiver to make
5e82b4b2 895 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
896 *
897 * void (*ndo_set_rx_mode)(struct net_device *dev);
898 * This function is called device changes address list filtering.
01789349 899 * If driver handles unicast address filtering, it should set
5e82b4b2 900 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
901 *
902 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
903 * This function is called when the Media Access Control address
37b607c5 904 * needs to be changed. If this interface is not defined, the
5e82b4b2 905 * MAC address can not be changed.
d314774c
SH
906 *
907 * int (*ndo_validate_addr)(struct net_device *dev);
908 * Test if Media Access Control address is valid for the device.
909 *
910 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
911 * Called when a user requests an ioctl which can't be handled by
912 * the generic interface code. If not defined ioctls return
d314774c
SH
913 * not supported error code.
914 *
915 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
916 * Used to set network devices bus interface parameters. This interface
5e82b4b2 917 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
918 * interface (PCI) for low level management.
919 *
920 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
921 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 922 * of a device.
d314774c 923 *
00829823 924 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 925 * Callback used when the transmitter has not made any progress
d314774c
SH
926 * for dev->watchdog ticks.
927 *
bc1f4470 928 * void (*ndo_get_stats64)(struct net_device *dev,
929 * struct rtnl_link_stats64 *storage);
d308e38f 930 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 931 * Called when a user wants to get the network device usage
be1f3c2c 932 * statistics. Drivers must do one of the following:
3cfde79c
BH
933 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
934 * rtnl_link_stats64 structure passed by the caller.
82695d9b 935 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
936 * (which should normally be dev->stats) and return a pointer to
937 * it. The structure may be changed asynchronously only if each
938 * field is written atomically.
939 * 3. Update dev->stats asynchronously and atomically, and define
940 * neither operation.
d314774c 941 *
3df5b3c6 942 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
943 * Return true if this device supports offload stats of this attr_id.
944 *
945 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
946 * void *attr_data)
947 * Get statistics for offload operations by attr_id. Write it into the
948 * attr_data pointer.
949 *
5d632cb7 950 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 951 * If device supports VLAN filtering this function is called when a
80d5c368 952 * VLAN id is registered.
d314774c 953 *
5d632cb7 954 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 955 * If device supports VLAN filtering this function is called when a
80d5c368 956 * VLAN id is unregistered.
d314774c
SH
957 *
958 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
959 *
960 * SR-IOV management functions.
961 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
962 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
963 * u8 qos, __be16 proto);
ed616689
SC
964 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
965 * int max_tx_rate);
5f8444a3 966 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 967 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
968 * int (*ndo_get_vf_config)(struct net_device *dev,
969 * int vf, struct ifla_vf_info *ivf);
1d8faf48 970 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
971 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
972 * struct nlattr *port[]);
01a3d796
VZ
973 *
974 * Enable or disable the VF ability to query its RSS Redirection Table and
975 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 976 * with PF and querying it may introduce a theoretical security risk.
01a3d796 977 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 978 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
2572ac53 979 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
2572ac53 980 * struct tc_to_netdev *tc);
6a4bc2b4
FF
981 * Called to setup any 'tc' scheduler, classifier or action on @dev.
982 * This is always called from the stack with the rtnl lock held and netif
983 * tx queues stopped. This allows the netdevice to perform queue
984 * management safely.
c445477d 985 *
e9bce845
YZ
986 * Fiber Channel over Ethernet (FCoE) offload functions.
987 * int (*ndo_fcoe_enable)(struct net_device *dev);
988 * Called when the FCoE protocol stack wants to start using LLD for FCoE
989 * so the underlying device can perform whatever needed configuration or
990 * initialization to support acceleration of FCoE traffic.
991 *
992 * int (*ndo_fcoe_disable)(struct net_device *dev);
993 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
994 * so the underlying device can perform whatever needed clean-ups to
995 * stop supporting acceleration of FCoE traffic.
996 *
997 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
998 * struct scatterlist *sgl, unsigned int sgc);
999 * Called when the FCoE Initiator wants to initialize an I/O that
1000 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1001 * perform necessary setup and returns 1 to indicate the device is set up
1002 * successfully to perform DDP on this I/O, otherwise this returns 0.
1003 *
1004 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1005 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1006 * indicated by the FC exchange id 'xid', so the underlying device can
1007 * clean up and reuse resources for later DDP requests.
1008 *
1009 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1010 * struct scatterlist *sgl, unsigned int sgc);
1011 * Called when the FCoE Target wants to initialize an I/O that
1012 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1013 * perform necessary setup and returns 1 to indicate the device is set up
1014 * successfully to perform DDP on this I/O, otherwise this returns 0.
1015 *
68bad94e
NP
1016 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1017 * struct netdev_fcoe_hbainfo *hbainfo);
1018 * Called when the FCoE Protocol stack wants information on the underlying
1019 * device. This information is utilized by the FCoE protocol stack to
1020 * register attributes with Fiber Channel management service as per the
1021 * FC-GS Fabric Device Management Information(FDMI) specification.
1022 *
e9bce845
YZ
1023 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1024 * Called when the underlying device wants to override default World Wide
1025 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1026 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1027 * protocol stack to use.
1028 *
c445477d
BH
1029 * RFS acceleration.
1030 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1031 * u16 rxq_index, u32 flow_id);
1032 * Set hardware filter for RFS. rxq_index is the target queue index;
1033 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1034 * Return the filter ID on success, or a negative error code.
fbaec0ea 1035 *
8b98a70c 1036 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1037 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1038 * Called to make another netdev an underling.
1039 *
1040 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1041 * Called to release previously enslaved netdev.
5455c699
MM
1042 *
1043 * Feature/offload setting functions.
1a2a1444
DM
1044 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1045 * netdev_features_t features);
1046 * Adjusts the requested feature flags according to device-specific
1047 * constraints, and returns the resulting flags. Must not modify
1048 * the device state.
1049 *
c8f44aff 1050 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1051 * Called to update device configuration to new features. Passed
1052 * feature set might be less than what was returned by ndo_fix_features()).
1053 * Must return >0 or -errno if it changed dev->features itself.
1054 *
edc7d573 1055 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1056 * struct net_device *dev,
f6f6424b 1057 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1058 * Adds an FDB entry to dev for addr.
1690be63
VY
1059 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1060 * struct net_device *dev,
f6f6424b 1061 * const unsigned char *addr, u16 vid)
77162022
JF
1062 * Deletes the FDB entry from dev coresponding to addr.
1063 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1064 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1065 * int *idx)
77162022
JF
1066 * Used to add FDB entries to dump requests. Implementers should add
1067 * entries to skb and update idx with the number of entries.
e5a55a89 1068 *
ad41faa8
ND
1069 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1070 * u16 flags)
e5a55a89 1071 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1072 * struct net_device *dev, u32 filter_mask,
1073 * int nlflags)
ad41faa8
ND
1074 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1075 * u16 flags);
4bf84c35
JP
1076 *
1077 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1078 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1079 * which do not represent real hardware may define this to allow their
1080 * userspace components to manage their virtual carrier state. Devices
1081 * that determine carrier state from physical hardware properties (eg
1082 * network cables) or protocol-dependent mechanisms (eg
1083 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1084 *
1085 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1086 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1087 * Called to get ID of physical port of this device. If driver does
1088 * not implement this, it is assumed that the hw is not able to have
1089 * multiple net devices on single physical port.
53cf5275 1090 *
7c46a640
AD
1091 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1092 * struct udp_tunnel_info *ti);
1093 * Called by UDP tunnel to notify a driver about the UDP port and socket
1094 * address family that a UDP tunnel is listnening to. It is called only
1095 * when a new port starts listening. The operation is protected by the
1096 * RTNL.
1097 *
1098 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1099 * struct udp_tunnel_info *ti);
1100 * Called by UDP tunnel to notify the driver about a UDP port and socket
1101 * address family that the UDP tunnel is not listening to anymore. The
1102 * operation is protected by the RTNL.
1103 *
a6cc0cfa
JF
1104 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1105 * struct net_device *dev)
1106 * Called by upper layer devices to accelerate switching or other
1107 * station functionality into hardware. 'pdev is the lowerdev
1108 * to use for the offload and 'dev' is the net device that will
1109 * back the offload. Returns a pointer to the private structure
1110 * the upper layer will maintain.
1111 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1112 * Called by upper layer device to delete the station created
1113 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1114 * the station and priv is the structure returned by the add
1115 * operation.
822b3b2e
JF
1116 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1117 * int queue_index, u32 maxrate);
1118 * Called when a user wants to set a max-rate limitation of specific
1119 * TX queue.
a54acb3a
ND
1120 * int (*ndo_get_iflink)(const struct net_device *dev);
1121 * Called to get the iflink value of this device.
d746d707 1122 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1123 * bool proto_down);
d746d707
AK
1124 * This function is used to pass protocol port error state information
1125 * to the switch driver. The switch driver can react to the proto_down
1126 * by doing a phys down on the associated switch port.
fc4099f1
PS
1127 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1128 * This function is used to get egress tunnel information for given skb.
1129 * This is useful for retrieving outer tunnel header parameters while
1130 * sampling packet.
871b642a
PA
1131 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1132 * This function is used to specify the headroom that the skb must
1133 * consider when allocation skb during packet reception. Setting
1134 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1135 * forward. Setting a negative value resets the rx headroom to the
871b642a 1136 * default value.
a7862b45
BB
1137 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1138 * This function is used to set or query state related to XDP on the
1139 * netdevice. See definition of enum xdp_netdev_command for details.
814abfab
JF
1140 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1141 * This function is used to submit a XDP packet for transmit on a
1142 * netdevice.
11393cc9
JF
1143 * void (*ndo_xdp_flush)(struct net_device *dev);
1144 * This function is used to inform the driver to flush a paticular
1145 * xpd tx queue. Must be called on same CPU as xdp_xmit.
d314774c
SH
1146 */
1147struct net_device_ops {
1148 int (*ndo_init)(struct net_device *dev);
1149 void (*ndo_uninit)(struct net_device *dev);
1150 int (*ndo_open)(struct net_device *dev);
1151 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1152 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1153 struct net_device *dev);
1154 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1155 struct net_device *dev,
1156 netdev_features_t features);
00829823 1157 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1158 struct sk_buff *skb,
99932d4f
DB
1159 void *accel_priv,
1160 select_queue_fallback_t fallback);
d314774c
SH
1161 void (*ndo_change_rx_flags)(struct net_device *dev,
1162 int flags);
d314774c 1163 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1164 int (*ndo_set_mac_address)(struct net_device *dev,
1165 void *addr);
d314774c 1166 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1167 int (*ndo_do_ioctl)(struct net_device *dev,
1168 struct ifreq *ifr, int cmd);
d314774c
SH
1169 int (*ndo_set_config)(struct net_device *dev,
1170 struct ifmap *map);
00829823
SH
1171 int (*ndo_change_mtu)(struct net_device *dev,
1172 int new_mtu);
1173 int (*ndo_neigh_setup)(struct net_device *dev,
1174 struct neigh_parms *);
d314774c
SH
1175 void (*ndo_tx_timeout) (struct net_device *dev);
1176
bc1f4470 1177 void (*ndo_get_stats64)(struct net_device *dev,
1178 struct rtnl_link_stats64 *storage);
3df5b3c6 1179 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1180 int (*ndo_get_offload_stats)(int attr_id,
1181 const struct net_device *dev,
1182 void *attr_data);
d314774c
SH
1183 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1184
8e586137 1185 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1186 __be16 proto, u16 vid);
8e586137 1187 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1188 __be16 proto, u16 vid);
d314774c 1189#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1190 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1191 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1192 struct netpoll_info *info);
0e34e931 1193 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1194#endif
95c26df8
WM
1195 int (*ndo_set_vf_mac)(struct net_device *dev,
1196 int queue, u8 *mac);
1197 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1198 int queue, u16 vlan,
1199 u8 qos, __be16 proto);
ed616689
SC
1200 int (*ndo_set_vf_rate)(struct net_device *dev,
1201 int vf, int min_tx_rate,
1202 int max_tx_rate);
5f8444a3
GR
1203 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1204 int vf, bool setting);
dd461d6a
HS
1205 int (*ndo_set_vf_trust)(struct net_device *dev,
1206 int vf, bool setting);
95c26df8
WM
1207 int (*ndo_get_vf_config)(struct net_device *dev,
1208 int vf,
1209 struct ifla_vf_info *ivf);
1d8faf48
RE
1210 int (*ndo_set_vf_link_state)(struct net_device *dev,
1211 int vf, int link_state);
3b766cd8
EBE
1212 int (*ndo_get_vf_stats)(struct net_device *dev,
1213 int vf,
1214 struct ifla_vf_stats
1215 *vf_stats);
57b61080
SF
1216 int (*ndo_set_vf_port)(struct net_device *dev,
1217 int vf,
1218 struct nlattr *port[]);
1219 int (*ndo_get_vf_port)(struct net_device *dev,
1220 int vf, struct sk_buff *skb);
cc8e27cc
EC
1221 int (*ndo_set_vf_guid)(struct net_device *dev,
1222 int vf, u64 guid,
1223 int guid_type);
01a3d796
VZ
1224 int (*ndo_set_vf_rss_query_en)(
1225 struct net_device *dev,
1226 int vf, bool setting);
16e5cc64 1227 int (*ndo_setup_tc)(struct net_device *dev,
2572ac53 1228 enum tc_setup_type type,
16e5cc64 1229 struct tc_to_netdev *tc);
d11ead75 1230#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1231 int (*ndo_fcoe_enable)(struct net_device *dev);
1232 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1233 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1234 u16 xid,
1235 struct scatterlist *sgl,
1236 unsigned int sgc);
1237 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1238 u16 xid);
6247e086
YZ
1239 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1240 u16 xid,
1241 struct scatterlist *sgl,
1242 unsigned int sgc);
68bad94e
NP
1243 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1244 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1245#endif
1246
d11ead75 1247#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1248#define NETDEV_FCOE_WWNN 0
1249#define NETDEV_FCOE_WWPN 1
1250 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1251 u64 *wwn, int type);
4d288d57 1252#endif
3c9c36bc 1253
c445477d
BH
1254#ifdef CONFIG_RFS_ACCEL
1255 int (*ndo_rx_flow_steer)(struct net_device *dev,
1256 const struct sk_buff *skb,
1257 u16 rxq_index,
1258 u32 flow_id);
1259#endif
fbaec0ea
JP
1260 int (*ndo_add_slave)(struct net_device *dev,
1261 struct net_device *slave_dev);
1262 int (*ndo_del_slave)(struct net_device *dev,
1263 struct net_device *slave_dev);
c8f44aff
MM
1264 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1265 netdev_features_t features);
5455c699 1266 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1267 netdev_features_t features);
503eebc2
JP
1268 int (*ndo_neigh_construct)(struct net_device *dev,
1269 struct neighbour *n);
1270 void (*ndo_neigh_destroy)(struct net_device *dev,
1271 struct neighbour *n);
77162022
JF
1272
1273 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1274 struct nlattr *tb[],
77162022 1275 struct net_device *dev,
6b6e2725 1276 const unsigned char *addr,
f6f6424b 1277 u16 vid,
77162022
JF
1278 u16 flags);
1279 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1280 struct nlattr *tb[],
77162022 1281 struct net_device *dev,
f6f6424b
JP
1282 const unsigned char *addr,
1283 u16 vid);
77162022
JF
1284 int (*ndo_fdb_dump)(struct sk_buff *skb,
1285 struct netlink_callback *cb,
1286 struct net_device *dev,
5d5eacb3 1287 struct net_device *filter_dev,
d297653d 1288 int *idx);
e5a55a89
JF
1289
1290 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1291 struct nlmsghdr *nlh,
1292 u16 flags);
e5a55a89
JF
1293 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1294 u32 pid, u32 seq,
6cbdceeb 1295 struct net_device *dev,
46c264da
ND
1296 u32 filter_mask,
1297 int nlflags);
407af329 1298 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1299 struct nlmsghdr *nlh,
1300 u16 flags);
4bf84c35
JP
1301 int (*ndo_change_carrier)(struct net_device *dev,
1302 bool new_carrier);
66b52b0d 1303 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1304 struct netdev_phys_item_id *ppid);
db24a904
DA
1305 int (*ndo_get_phys_port_name)(struct net_device *dev,
1306 char *name, size_t len);
7c46a640
AD
1307 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1308 struct udp_tunnel_info *ti);
1309 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1310 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1311 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1312 struct net_device *dev);
1313 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1314 void *priv);
1315
25175ba5 1316 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1317 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1318 int queue_index,
1319 u32 maxrate);
a54acb3a 1320 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1321 int (*ndo_change_proto_down)(struct net_device *dev,
1322 bool proto_down);
fc4099f1
PS
1323 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1324 struct sk_buff *skb);
871b642a
PA
1325 void (*ndo_set_rx_headroom)(struct net_device *dev,
1326 int needed_headroom);
a7862b45
BB
1327 int (*ndo_xdp)(struct net_device *dev,
1328 struct netdev_xdp *xdp);
814abfab
JF
1329 int (*ndo_xdp_xmit)(struct net_device *dev,
1330 struct xdp_buff *xdp);
11393cc9 1331 void (*ndo_xdp_flush)(struct net_device *dev);
d314774c
SH
1332};
1333
7aa98047
LR
1334/**
1335 * enum net_device_priv_flags - &struct net_device priv_flags
1336 *
1337 * These are the &struct net_device, they are only set internally
1338 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1339 * userspace; this means that the order of these flags can change
7aa98047
LR
1340 * during any kernel release.
1341 *
1342 * You should have a pretty good reason to be extending these flags.
1343 *
1344 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1345 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1346 * @IFF_BONDING: bonding master or slave
7aa98047 1347 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1348 * @IFF_WAN_HDLC: WAN HDLC device
1349 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1350 * release skb->dst
1351 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1352 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1353 * @IFF_MACVLAN_PORT: device used as macvlan port
1354 * @IFF_BRIDGE_PORT: device used as bridge port
1355 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1356 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1357 * @IFF_UNICAST_FLT: Supports unicast filtering
1358 * @IFF_TEAM_PORT: device used as team port
1359 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1360 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1361 * change when it's running
1362 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1363 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1364 * underlying stacked devices
1365 * @IFF_IPVLAN_MASTER: IPvlan master device
1366 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1367 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1368 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1369 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1370 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1371 * @IFF_TEAM: device is a team device
d4ab4286 1372 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1373 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1374 * entity (i.e. the master device for bridged veth)
3c175784 1375 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1376 */
1377enum netdev_priv_flags {
1378 IFF_802_1Q_VLAN = 1<<0,
1379 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1380 IFF_BONDING = 1<<2,
1381 IFF_ISATAP = 1<<3,
1382 IFF_WAN_HDLC = 1<<4,
1383 IFF_XMIT_DST_RELEASE = 1<<5,
1384 IFF_DONT_BRIDGE = 1<<6,
1385 IFF_DISABLE_NETPOLL = 1<<7,
1386 IFF_MACVLAN_PORT = 1<<8,
1387 IFF_BRIDGE_PORT = 1<<9,
1388 IFF_OVS_DATAPATH = 1<<10,
1389 IFF_TX_SKB_SHARING = 1<<11,
1390 IFF_UNICAST_FLT = 1<<12,
1391 IFF_TEAM_PORT = 1<<13,
1392 IFF_SUPP_NOFCS = 1<<14,
1393 IFF_LIVE_ADDR_CHANGE = 1<<15,
1394 IFF_MACVLAN = 1<<16,
1395 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1396 IFF_IPVLAN_MASTER = 1<<18,
1397 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1398 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1399 IFF_NO_QUEUE = 1<<21,
1400 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1401 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1402 IFF_TEAM = 1<<24,
d4ab4286 1403 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1404 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1405 IFF_MACSEC = 1<<27,
7aa98047
LR
1406};
1407
1408#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1409#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1410#define IFF_BONDING IFF_BONDING
7aa98047 1411#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1412#define IFF_WAN_HDLC IFF_WAN_HDLC
1413#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1414#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1415#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1416#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1417#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1418#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1419#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1420#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1421#define IFF_TEAM_PORT IFF_TEAM_PORT
1422#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1423#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1424#define IFF_MACVLAN IFF_MACVLAN
02875878 1425#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1426#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1427#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1428#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1429#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1430#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1431#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1432#define IFF_TEAM IFF_TEAM
d4ab4286 1433#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1434#define IFF_MACSEC IFF_MACSEC
7aa98047 1435
536721b1
KK
1436/**
1437 * struct net_device - The DEVICE structure.
d651983d
MCC
1438 *
1439 * Actually, this whole structure is a big mistake. It mixes I/O
1440 * data with strictly "high-level" data, and it has to know about
1441 * almost every data structure used in the INET module.
536721b1
KK
1442 *
1443 * @name: This is the first field of the "visible" part of this structure
1444 * (i.e. as seen by users in the "Space.c" file). It is the name
d651983d 1445 * of the interface.
536721b1
KK
1446 *
1447 * @name_hlist: Device name hash chain, please keep it close to name[]
1448 * @ifalias: SNMP alias
1449 * @mem_end: Shared memory end
1450 * @mem_start: Shared memory start
1451 * @base_addr: Device I/O address
1452 * @irq: Device IRQ number
1453 *
14ffbbb8
TG
1454 * @carrier_changes: Stats to monitor carrier on<->off transitions
1455 *
536721b1
KK
1456 * @state: Generic network queuing layer state, see netdev_state_t
1457 * @dev_list: The global list of network devices
5e82b4b2
BH
1458 * @napi_list: List entry used for polling NAPI devices
1459 * @unreg_list: List entry when we are unregistering the
1460 * device; see the function unregister_netdev
1461 * @close_list: List entry used when we are closing the device
62d885fe
BP
1462 * @ptype_all: Device-specific packet handlers for all protocols
1463 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1464 *
1465 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1466 * @features: Currently active device features
1467 * @hw_features: User-changeable features
1468 *
1469 * @wanted_features: User-requested features
1470 * @vlan_features: Mask of features inheritable by VLAN devices
1471 *
1472 * @hw_enc_features: Mask of features inherited by encapsulating devices
1473 * This field indicates what encapsulation
1474 * offloads the hardware is capable of doing,
1475 * and drivers will need to set them appropriately.
1476 *
1477 * @mpls_features: Mask of features inheritable by MPLS
1478 *
1479 * @ifindex: interface index
5e82b4b2 1480 * @group: The group the device belongs to
536721b1
KK
1481 *
1482 * @stats: Statistics struct, which was left as a legacy, use
1483 * rtnl_link_stats64 instead
1484 *
1485 * @rx_dropped: Dropped packets by core network,
1486 * do not use this in drivers
1487 * @tx_dropped: Dropped packets by core network,
1488 * do not use this in drivers
6e7333d3
JW
1489 * @rx_nohandler: nohandler dropped packets by core network on
1490 * inactive devices, do not use this in drivers
536721b1 1491 *
536721b1
KK
1492 * @wireless_handlers: List of functions to handle Wireless Extensions,
1493 * instead of ioctl,
1494 * see <net/iw_handler.h> for details.
1495 * @wireless_data: Instance data managed by the core of wireless extensions
1496 *
1497 * @netdev_ops: Includes several pointers to callbacks,
1498 * if one wants to override the ndo_*() functions
1499 * @ethtool_ops: Management operations
f997c55c
AA
1500 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1501 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1502 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1503 * of Layer 2 headers.
1504 *
1505 * @flags: Interface flags (a la BSD)
1506 * @priv_flags: Like 'flags' but invisible to userspace,
1507 * see if.h for the definitions
1508 * @gflags: Global flags ( kept as legacy )
1509 * @padded: How much padding added by alloc_netdev()
1510 * @operstate: RFC2863 operstate
1511 * @link_mode: Mapping policy to operstate
1512 * @if_port: Selectable AUI, TP, ...
1513 * @dma: DMA channel
1514 * @mtu: Interface MTU value
61e84623
JW
1515 * @min_mtu: Interface Minimum MTU value
1516 * @max_mtu: Interface Maximum MTU value
536721b1 1517 * @type: Interface hardware type
2793a23a 1518 * @hard_header_len: Maximum hardware header length.
217e6fa2 1519 * @min_header_len: Minimum hardware header length
536721b1
KK
1520 *
1521 * @needed_headroom: Extra headroom the hardware may need, but not in all
1522 * cases can this be guaranteed
1523 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1524 * cases can this be guaranteed. Some cases also use
1525 * LL_MAX_HEADER instead to allocate the skb
1526 *
1527 * interface address info:
1528 *
1529 * @perm_addr: Permanent hw address
1530 * @addr_assign_type: Hw address assignment type
1531 * @addr_len: Hardware address length
8626a0c8 1532 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1533 * @dev_id: Used to differentiate devices that share
1534 * the same link layer address
1535 * @dev_port: Used to differentiate devices that share
1536 * the same function
1537 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1538 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1539 * has been enabled due to the need to listen to
1540 * additional unicast addresses in a device that
1541 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1542 * @uc: unicast mac addresses
1543 * @mc: multicast mac addresses
1544 * @dev_addrs: list of device hw addresses
1545 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1546 * @promiscuity: Number of times the NIC is told to work in
1547 * promiscuous mode; if it becomes 0 the NIC will
1548 * exit promiscuous mode
536721b1
KK
1549 * @allmulti: Counter, enables or disables allmulticast mode
1550 *
1551 * @vlan_info: VLAN info
1552 * @dsa_ptr: dsa specific data
1553 * @tipc_ptr: TIPC specific data
1554 * @atalk_ptr: AppleTalk link
1555 * @ip_ptr: IPv4 specific data
1556 * @dn_ptr: DECnet specific data
1557 * @ip6_ptr: IPv6 specific data
1558 * @ax25_ptr: AX.25 specific data
1559 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1560 *
536721b1
KK
1561 * @dev_addr: Hw address (before bcast,
1562 * because most packets are unicast)
1563 *
1564 * @_rx: Array of RX queues
1565 * @num_rx_queues: Number of RX queues
1566 * allocated at register_netdev() time
1567 * @real_num_rx_queues: Number of RX queues currently active in device
1568 *
1569 * @rx_handler: handler for received packets
1570 * @rx_handler_data: XXX: need comments on this one
1571 * @ingress_queue: XXX: need comments on this one
1572 * @broadcast: hw bcast address
1573 *
14ffbbb8
TG
1574 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1575 * indexed by RX queue number. Assigned by driver.
1576 * This must only be set if the ndo_rx_flow_steer
1577 * operation is defined
1578 * @index_hlist: Device index hash chain
1579 *
536721b1
KK
1580 * @_tx: Array of TX queues
1581 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1582 * @real_num_tx_queues: Number of TX queues currently active in device
1583 * @qdisc: Root qdisc from userspace point of view
1584 * @tx_queue_len: Max frames per queue allowed
1585 * @tx_global_lock: XXX: need comments on this one
1586 *
1587 * @xps_maps: XXX: need comments on this one
1588 *
536721b1 1589 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1590 * the watchdog (see dev_watchdog())
536721b1
KK
1591 * @watchdog_timer: List of timers
1592 *
1593 * @pcpu_refcnt: Number of references to this device
1594 * @todo_list: Delayed register/unregister
536721b1
KK
1595 * @link_watch_list: XXX: need comments on this one
1596 *
1597 * @reg_state: Register/unregister state machine
1598 * @dismantle: Device is going to be freed
1599 * @rtnl_link_state: This enum represents the phases of creating
1600 * a new link
1601 *
cf124db5
DM
1602 * @needs_free_netdev: Should unregister perform free_netdev?
1603 * @priv_destructor: Called from unregister
536721b1
KK
1604 * @npinfo: XXX: need comments on this one
1605 * @nd_net: Network namespace this network device is inside
1606 *
1607 * @ml_priv: Mid-layer private
1608 * @lstats: Loopback statistics
1609 * @tstats: Tunnel statistics
1610 * @dstats: Dummy statistics
1611 * @vstats: Virtual ethernet statistics
1612 *
1613 * @garp_port: GARP
1614 * @mrp_port: MRP
1615 *
1616 * @dev: Class/net/name entry
1617 * @sysfs_groups: Space for optional device, statistics and wireless
1618 * sysfs groups
1619 *
1620 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1621 * @rtnl_link_ops: Rtnl_link_ops
1622 *
1623 * @gso_max_size: Maximum size of generic segmentation offload
1624 * @gso_max_segs: Maximum number of segments that can be passed to the
1625 * NIC for GSO
1626 *
1627 * @dcbnl_ops: Data Center Bridging netlink ops
1628 * @num_tc: Number of traffic classes in the net device
1629 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1630 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1631 *
1632 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1633 *
1634 * @priomap: XXX: need comments on this one
1635 * @phydev: Physical device may attach itself
1636 * for hardware timestamping
1637 *
123b3652
ED
1638 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1639 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1640 *
d746d707
AK
1641 * @proto_down: protocol port state information can be sent to the
1642 * switch driver and used to set the phys state of the
1643 * switch port.
1644 *
1da177e4
LT
1645 * FIXME: cleanup struct net_device such that network protocol info
1646 * moves out.
1647 */
1648
d94d9fee 1649struct net_device {
1da177e4 1650 char name[IFNAMSIZ];
9356b8fc 1651 struct hlist_node name_hlist;
0b815a1a 1652 char *ifalias;
1da177e4
LT
1653 /*
1654 * I/O specific fields
1655 * FIXME: Merge these and struct ifmap into one
1656 */
536721b1
KK
1657 unsigned long mem_end;
1658 unsigned long mem_start;
1659 unsigned long base_addr;
1660 int irq;
1da177e4 1661
14ffbbb8
TG
1662 atomic_t carrier_changes;
1663
1da177e4 1664 /*
536721b1
KK
1665 * Some hardware also needs these fields (state,dev_list,
1666 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1667 * part of the usual set specified in Space.c.
1668 */
1669
1da177e4
LT
1670 unsigned long state;
1671
7562f876 1672 struct list_head dev_list;
bea3348e 1673 struct list_head napi_list;
44a0873d 1674 struct list_head unreg_list;
5cde2829 1675 struct list_head close_list;
7866a621
SN
1676 struct list_head ptype_all;
1677 struct list_head ptype_specific;
2f268f12 1678
2f268f12
VF
1679 struct {
1680 struct list_head upper;
1681 struct list_head lower;
1682 } adj_list;
1683
c8f44aff 1684 netdev_features_t features;
c8f44aff 1685 netdev_features_t hw_features;
c8f44aff 1686 netdev_features_t wanted_features;
c8f44aff 1687 netdev_features_t vlan_features;
6a674e9c 1688 netdev_features_t hw_enc_features;
0d89d203 1689 netdev_features_t mpls_features;
802ab55a 1690 netdev_features_t gso_partial_features;
04ed3e74 1691
1da177e4 1692 int ifindex;
7a66bbc9 1693 int group;
1da177e4 1694
c45d286e 1695 struct net_device_stats stats;
015f0688 1696
015f0688
ED
1697 atomic_long_t rx_dropped;
1698 atomic_long_t tx_dropped;
6e7333d3 1699 atomic_long_t rx_nohandler;
1da177e4 1700
b86e0280 1701#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1702 const struct iw_handler_def *wireless_handlers;
1703 struct iw_public_data *wireless_data;
b86e0280 1704#endif
d314774c 1705 const struct net_device_ops *netdev_ops;
76fd8593 1706 const struct ethtool_ops *ethtool_ops;
4170604f 1707#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1708 const struct switchdev_ops *switchdev_ops;
4170604f 1709#endif
1b69c6d0
DA
1710#ifdef CONFIG_NET_L3_MASTER_DEV
1711 const struct l3mdev_ops *l3mdev_ops;
1712#endif
f997c55c
AA
1713#if IS_ENABLED(CONFIG_IPV6)
1714 const struct ndisc_ops *ndisc_ops;
1715#endif
1da177e4 1716
d77e38e6
SK
1717#ifdef CONFIG_XFRM
1718 const struct xfrmdev_ops *xfrmdev_ops;
1719#endif
1720
3b04ddde
SH
1721 const struct header_ops *header_ops;
1722
536721b1
KK
1723 unsigned int flags;
1724 unsigned int priv_flags;
1725
1da177e4 1726 unsigned short gflags;
536721b1 1727 unsigned short padded;
1da177e4 1728
536721b1
KK
1729 unsigned char operstate;
1730 unsigned char link_mode;
b00055aa 1731
536721b1
KK
1732 unsigned char if_port;
1733 unsigned char dma;
bdc220da 1734
536721b1 1735 unsigned int mtu;
61e84623
JW
1736 unsigned int min_mtu;
1737 unsigned int max_mtu;
536721b1
KK
1738 unsigned short type;
1739 unsigned short hard_header_len;
d92be7a4 1740 unsigned char min_header_len;
1da177e4 1741
f5184d26
JB
1742 unsigned short needed_headroom;
1743 unsigned short needed_tailroom;
1744
1da177e4 1745 /* Interface address info. */
536721b1
KK
1746 unsigned char perm_addr[MAX_ADDR_LEN];
1747 unsigned char addr_assign_type;
1748 unsigned char addr_len;
a0a9663d 1749 unsigned short neigh_priv_len;
536721b1
KK
1750 unsigned short dev_id;
1751 unsigned short dev_port;
ccffad25 1752 spinlock_t addr_list_lock;
14ffbbb8
TG
1753 unsigned char name_assign_type;
1754 bool uc_promisc;
536721b1
KK
1755 struct netdev_hw_addr_list uc;
1756 struct netdev_hw_addr_list mc;
1757 struct netdev_hw_addr_list dev_addrs;
1758
4c3d5e7b
ED
1759#ifdef CONFIG_SYSFS
1760 struct kset *queues_kset;
1761#endif
9d45abe1
WC
1762 unsigned int promiscuity;
1763 unsigned int allmulti;
1da177e4 1764
1da177e4 1765
5e82b4b2 1766 /* Protocol-specific pointers */
65ac6a5f 1767
d11ead75 1768#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1769 struct vlan_info __rcu *vlan_info;
65ac6a5f 1770#endif
34a430d7 1771#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1772 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1773#endif
1774#if IS_ENABLED(CONFIG_TIPC)
536721b1 1775 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1776#endif
536721b1
KK
1777 void *atalk_ptr;
1778 struct in_device __rcu *ip_ptr;
1779 struct dn_dev __rcu *dn_ptr;
1780 struct inet6_dev __rcu *ip6_ptr;
1781 void *ax25_ptr;
1782 struct wireless_dev *ieee80211_ptr;
98a18b6f 1783 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1784#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1785 struct mpls_dev __rcu *mpls_ptr;
1786#endif
1da177e4 1787
9356b8fc 1788/*
cd13539b 1789 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1790 */
9356b8fc 1791 /* Interface address info used in eth_type_trans() */
536721b1 1792 unsigned char *dev_addr;
f001fde5 1793
a953be53 1794#ifdef CONFIG_SYSFS
0a9627f2
TH
1795 struct netdev_rx_queue *_rx;
1796
0a9627f2 1797 unsigned int num_rx_queues;
62fe0b40 1798 unsigned int real_num_rx_queues;
df334545 1799#endif
0a9627f2 1800
7acedaf5 1801 struct bpf_prog __rcu *xdp_prog;
3b47d303 1802 unsigned long gro_flush_timeout;
61391cde 1803 rx_handler_func_t __rcu *rx_handler;
1804 void __rcu *rx_handler_data;
e8a0464c 1805
4cda01e8 1806#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1807 struct tcf_proto __rcu *ingress_cl_list;
1808#endif
24824a09 1809 struct netdev_queue __rcu *ingress_queue;
e687ad60 1810#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1811 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1812#endif
d2788d34 1813
536721b1 1814 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1815#ifdef CONFIG_RFS_ACCEL
1816 struct cpu_rmap *rx_cpu_rmap;
1817#endif
1818 struct hlist_node index_hlist;
cd13539b
ED
1819
1820/*
1821 * Cache lines mostly used on transmit path
1822 */
e8a0464c
DM
1823 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1824 unsigned int num_tx_queues;
fd2ea0a7 1825 unsigned int real_num_tx_queues;
af356afa 1826 struct Qdisc *qdisc;
59cc1f61
JK
1827#ifdef CONFIG_NET_SCHED
1828 DECLARE_HASHTABLE (qdisc_hash, 4);
1829#endif
0cd29503 1830 unsigned int tx_queue_len;
c3f26a26 1831 spinlock_t tx_global_lock;
14ffbbb8 1832 int watchdog_timeo;
cd13539b 1833
bf264145 1834#ifdef CONFIG_XPS
a4177869 1835 struct xps_dev_maps __rcu *xps_maps;
bf264145 1836#endif
1f211a1b
DB
1837#ifdef CONFIG_NET_CLS_ACT
1838 struct tcf_proto __rcu *egress_cl_list;
1839#endif
0c4f691f 1840
9356b8fc 1841 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1842 struct timer_list watchdog_timer;
1843
29b4433d 1844 int __percpu *pcpu_refcnt;
1da177e4 1845 struct list_head todo_list;
1da177e4 1846
e014debe 1847 struct list_head link_watch_list;
572a103d 1848
1da177e4 1849 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1850 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1851 NETREG_UNREGISTERING, /* called unregister_netdevice */
1852 NETREG_UNREGISTERED, /* completed unregister todo */
1853 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1854 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1855 } reg_state:8;
1856
536721b1 1857 bool dismantle;
a2835763
PM
1858
1859 enum {
1860 RTNL_LINK_INITIALIZED,
1861 RTNL_LINK_INITIALIZING,
1862 } rtnl_link_state:16;
1da177e4 1863
cf124db5
DM
1864 bool needs_free_netdev;
1865 void (*priv_destructor)(struct net_device *dev);
1da177e4 1866
1da177e4 1867#ifdef CONFIG_NETPOLL
5fbee843 1868 struct netpoll_info __rcu *npinfo;
1da177e4 1869#endif
eae792b7 1870
0c5c9fb5 1871 possible_net_t nd_net;
4a1c5371 1872
4951704b 1873 /* mid-layer private */
a7855c78 1874 union {
536721b1
KK
1875 void *ml_priv;
1876 struct pcpu_lstats __percpu *lstats;
8f84985f 1877 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1878 struct pcpu_dstats __percpu *dstats;
1879 struct pcpu_vstats __percpu *vstats;
a7855c78 1880 };
536721b1 1881
fb585b44 1882#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1883 struct garp_port __rcu *garp_port;
fb585b44
TK
1884#endif
1885#if IS_ENABLED(CONFIG_MRP)
febf018d 1886 struct mrp_port __rcu *mrp_port;
fb585b44 1887#endif
1da177e4 1888
5e82b4b2 1889 struct device dev;
0c509a6c 1890 const struct attribute_group *sysfs_groups[4];
a953be53 1891 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1892
38f7b870 1893 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1894
82cc1a7a
PWJ
1895 /* for setting kernel sock attribute on TCP connection setup */
1896#define GSO_MAX_SIZE 65536
1897 unsigned int gso_max_size;
30b678d8
BH
1898#define GSO_MAX_SEGS 65535
1899 u16 gso_max_segs;
743b03a8 1900
7a6b6f51 1901#ifdef CONFIG_DCB
32953543 1902 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1903#endif
5e82b4b2
BH
1904 u8 num_tc;
1905 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1906 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1907
d11ead75 1908#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1909 unsigned int fcoe_ddp_xid;
5bc1421e 1910#endif
86f8515f 1911#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1912 struct netprio_map __rcu *priomap;
4d288d57 1913#endif
5e82b4b2
BH
1914 struct phy_device *phydev;
1915 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1916 struct lock_class_key *qdisc_running_key;
5e82b4b2 1917 bool proto_down;
1da177e4 1918};
43cb76d9 1919#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1920
b5cdae32
DM
1921static inline bool netif_elide_gro(const struct net_device *dev)
1922{
1923 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1924 return true;
1925 return false;
1926}
1927
1da177e4 1928#define NETDEV_ALIGN 32
1da177e4 1929
4f57c087
JF
1930static inline
1931int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1932{
1933 return dev->prio_tc_map[prio & TC_BITMASK];
1934}
1935
1936static inline
1937int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1938{
1939 if (tc >= dev->num_tc)
1940 return -EINVAL;
1941
1942 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1943 return 0;
1944}
1945
8d059b0f 1946int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1947void netdev_reset_tc(struct net_device *dev);
1948int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1949int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1950
1951static inline
1952int netdev_get_num_tc(struct net_device *dev)
1953{
1954 return dev->num_tc;
1955}
1956
e8a0464c
DM
1957static inline
1958struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1959 unsigned int index)
1960{
1961 return &dev->_tx[index];
1962}
1963
10c51b56
DB
1964static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1965 const struct sk_buff *skb)
1966{
1967 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1968}
1969
e8a0464c
DM
1970static inline void netdev_for_each_tx_queue(struct net_device *dev,
1971 void (*f)(struct net_device *,
1972 struct netdev_queue *,
1973 void *),
1974 void *arg)
1975{
1976 unsigned int i;
1977
1978 for (i = 0; i < dev->num_tx_queues; i++)
1979 f(dev, &dev->_tx[i], arg);
1980}
1981
d3fff6c4
ED
1982#define netdev_lockdep_set_classes(dev) \
1983{ \
1984 static struct lock_class_key qdisc_tx_busylock_key; \
1985 static struct lock_class_key qdisc_running_key; \
1986 static struct lock_class_key qdisc_xmit_lock_key; \
1987 static struct lock_class_key dev_addr_list_lock_key; \
1988 unsigned int i; \
1989 \
1990 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1991 (dev)->qdisc_running_key = &qdisc_running_key; \
1992 lockdep_set_class(&(dev)->addr_list_lock, \
1993 &dev_addr_list_lock_key); \
1994 for (i = 0; i < (dev)->num_tx_queues; i++) \
1995 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1996 &qdisc_xmit_lock_key); \
1997}
1998
f629d208 1999struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
2000 struct sk_buff *skb,
2001 void *accel_priv);
8c4c49df 2002
871b642a
PA
2003/* returns the headroom that the master device needs to take in account
2004 * when forwarding to this dev
2005 */
2006static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2007{
2008 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2009}
2010
2011static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2012{
2013 if (dev->netdev_ops->ndo_set_rx_headroom)
2014 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2015}
2016
2017/* set the device rx headroom to the dev's default */
2018static inline void netdev_reset_rx_headroom(struct net_device *dev)
2019{
2020 netdev_set_rx_headroom(dev, -1);
2021}
2022
c346dca1
YH
2023/*
2024 * Net namespace inlines
2025 */
2026static inline
2027struct net *dev_net(const struct net_device *dev)
2028{
c2d9ba9b 2029 return read_pnet(&dev->nd_net);
c346dca1
YH
2030}
2031
2032static inline
f5aa23fd 2033void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2034{
0c5c9fb5 2035 write_pnet(&dev->nd_net, net);
c346dca1
YH
2036}
2037
bea3348e
SH
2038/**
2039 * netdev_priv - access network device private data
2040 * @dev: network device
2041 *
2042 * Get network device private data
2043 */
6472ce60 2044static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2045{
1ce8e7b5 2046 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2047}
2048
1da177e4
LT
2049/* Set the sysfs physical device reference for the network logical device
2050 * if set prior to registration will cause a symlink during initialization.
2051 */
43cb76d9 2052#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2053
384912ed 2054/* Set the sysfs device type for the network logical device to allow
3f79410c 2055 * fine-grained identification of different network device types. For
5e82b4b2 2056 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2057 */
2058#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2059
82dc3c63
ED
2060/* Default NAPI poll() weight
2061 * Device drivers are strongly advised to not use bigger value
2062 */
2063#define NAPI_POLL_WEIGHT 64
2064
3b582cc1 2065/**
5e82b4b2 2066 * netif_napi_add - initialize a NAPI context
3b582cc1 2067 * @dev: network device
5e82b4b2 2068 * @napi: NAPI context
3b582cc1
SH
2069 * @poll: polling function
2070 * @weight: default weight
2071 *
5e82b4b2
BH
2072 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2073 * *any* of the other NAPI-related functions.
3b582cc1 2074 */
d565b0a1
HX
2075void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2076 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2077
d64b5e85 2078/**
5e82b4b2 2079 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2080 * @dev: network device
5e82b4b2 2081 * @napi: NAPI context
d64b5e85
ED
2082 * @poll: polling function
2083 * @weight: default weight
2084 *
2085 * This variant of netif_napi_add() should be used from drivers using NAPI
2086 * to exclusively poll a TX queue.
2087 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2088 */
2089static inline void netif_tx_napi_add(struct net_device *dev,
2090 struct napi_struct *napi,
2091 int (*poll)(struct napi_struct *, int),
2092 int weight)
2093{
2094 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2095 netif_napi_add(dev, napi, poll, weight);
2096}
2097
d8156534 2098/**
5e82b4b2
BH
2099 * netif_napi_del - remove a NAPI context
2100 * @napi: NAPI context
d8156534 2101 *
5e82b4b2 2102 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2103 */
d565b0a1
HX
2104void netif_napi_del(struct napi_struct *napi);
2105
2106struct napi_gro_cb {
78a478d0 2107 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2108 void *frag0;
78a478d0 2109
7489594c
HX
2110 /* Length of frag0. */
2111 unsigned int frag0_len;
2112
86911732 2113 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2114 int data_offset;
86911732 2115
d565b0a1 2116 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2117 u16 flush;
2118
2119 /* Save the IP ID here and check when we get to the transport layer */
2120 u16 flush_id;
d565b0a1
HX
2121
2122 /* Number of segments aggregated. */
2e71a6f8
ED
2123 u16 count;
2124
15e2396d
TH
2125 /* Start offset for remote checksum offload */
2126 u16 gro_remcsum_start;
2127
2e71a6f8
ED
2128 /* jiffies when first packet was created/queued */
2129 unsigned long age;
86347245 2130
afe93325 2131 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2132 u16 proto;
2133
baa32ff4
TH
2134 /* This is non-zero if the packet may be of the same flow. */
2135 u8 same_flow:1;
2136
fac8e0f5
JG
2137 /* Used in tunnel GRO receive */
2138 u8 encap_mark:1;
573e8fca
TH
2139
2140 /* GRO checksum is valid */
2141 u8 csum_valid:1;
2142
662880f4
TH
2143 /* Number of checksums via CHECKSUM_UNNECESSARY */
2144 u8 csum_cnt:3;
c3c7c254 2145
baa32ff4
TH
2146 /* Free the skb? */
2147 u8 free:2;
2148#define NAPI_GRO_FREE 1
2149#define NAPI_GRO_FREE_STOLEN_HEAD 2
2150
efc98d08
TH
2151 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2152 u8 is_ipv6:1;
2153
a0ca153f
AD
2154 /* Used in GRE, set in fou/gue_gro_receive */
2155 u8 is_fou:1;
2156
1530545e
AD
2157 /* Used to determine if flush_id can be ignored */
2158 u8 is_atomic:1;
2159
fcd91dd4
SD
2160 /* Number of gro_receive callbacks this packet already went through */
2161 u8 recursion_counter:4;
2162
2163 /* 1 bit hole */
baa32ff4 2164
bf5a755f
JC
2165 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2166 __wsum csum;
2167
c3c7c254
ED
2168 /* used in skb_gro_receive() slow path */
2169 struct sk_buff *last;
d565b0a1
HX
2170};
2171
2172#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2173
fcd91dd4
SD
2174#define GRO_RECURSION_LIMIT 15
2175static inline int gro_recursion_inc_test(struct sk_buff *skb)
2176{
2177 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2178}
2179
2180typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2181static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2182 struct sk_buff **head,
2183 struct sk_buff *skb)
2184{
2185 if (unlikely(gro_recursion_inc_test(skb))) {
2186 NAPI_GRO_CB(skb)->flush |= 1;
2187 return NULL;
2188 }
2189
2190 return cb(head, skb);
2191}
2192
2193typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2194 struct sk_buff *);
2195static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2196 struct sock *sk,
2197 struct sk_buff **head,
2198 struct sk_buff *skb)
2199{
2200 if (unlikely(gro_recursion_inc_test(skb))) {
2201 NAPI_GRO_CB(skb)->flush |= 1;
2202 return NULL;
2203 }
2204
2205 return cb(sk, head, skb);
2206}
2207
1da177e4 2208struct packet_type {
f2ccd8fa
DM
2209 __be16 type; /* This is really htons(ether_type). */
2210 struct net_device *dev; /* NULL is wildcarded here */
2211 int (*func) (struct sk_buff *,
2212 struct net_device *,
2213 struct packet_type *,
2214 struct net_device *);
c0de08d0
EL
2215 bool (*id_match)(struct packet_type *ptype,
2216 struct sock *sk);
1da177e4
LT
2217 void *af_packet_priv;
2218 struct list_head list;
2219};
2220
f191a1d1 2221struct offload_callbacks {
576a30eb 2222 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2223 netdev_features_t features);
d565b0a1 2224 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2225 struct sk_buff *skb);
299603e8 2226 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2227};
2228
2229struct packet_offload {
2230 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2231 u16 priority;
f191a1d1
VY
2232 struct offload_callbacks callbacks;
2233 struct list_head list;
1da177e4
LT
2234};
2235
5e82b4b2 2236/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2237struct pcpu_sw_netstats {
2238 u64 rx_packets;
2239 u64 rx_bytes;
2240 u64 tx_packets;
2241 u64 tx_bytes;
2242 struct u64_stats_sync syncp;
2243};
2244
aabc92bb
PNA
2245#define __netdev_alloc_pcpu_stats(type, gfp) \
2246({ \
2247 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2248 if (pcpu_stats) { \
2249 int __cpu; \
2250 for_each_possible_cpu(__cpu) { \
2251 typeof(type) *stat; \
2252 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2253 u64_stats_init(&stat->syncp); \
2254 } \
2255 } \
2256 pcpu_stats; \
1c213bd2
WC
2257})
2258
aabc92bb 2259#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2260 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2261
764f5e54
JP
2262enum netdev_lag_tx_type {
2263 NETDEV_LAG_TX_TYPE_UNKNOWN,
2264 NETDEV_LAG_TX_TYPE_RANDOM,
2265 NETDEV_LAG_TX_TYPE_BROADCAST,
2266 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2267 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2268 NETDEV_LAG_TX_TYPE_HASH,
2269};
2270
2271struct netdev_lag_upper_info {
2272 enum netdev_lag_tx_type tx_type;
2273};
2274
fb1b2e3c
JP
2275struct netdev_lag_lower_state_info {
2276 u8 link_up : 1,
2277 tx_enabled : 1;
2278};
2279
1da177e4
LT
2280#include <linux/notifier.h>
2281
dcfe1421
AW
2282/* netdevice notifier chain. Please remember to update the rtnetlink
2283 * notification exclusion list in rtnetlink_event() when adding new
2284 * types.
2285 */
2286#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2287#define NETDEV_DOWN 0x0002
2288#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2289 detected a hardware crash and restarted
2290 - we can use this eg to kick tcp sessions
2291 once done */
2292#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2293#define NETDEV_REGISTER 0x0005
2294#define NETDEV_UNREGISTER 0x0006
1d486bfb 2295#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2296#define NETDEV_CHANGEADDR 0x0008
2297#define NETDEV_GOING_DOWN 0x0009
2298#define NETDEV_CHANGENAME 0x000A
2299#define NETDEV_FEAT_CHANGE 0x000B
2300#define NETDEV_BONDING_FAILOVER 0x000C
2301#define NETDEV_PRE_UP 0x000D
2302#define NETDEV_PRE_TYPE_CHANGE 0x000E
2303#define NETDEV_POST_TYPE_CHANGE 0x000F
2304#define NETDEV_POST_INIT 0x0010
0115e8e3 2305#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2306#define NETDEV_RELEASE 0x0012
2307#define NETDEV_NOTIFY_PEERS 0x0013
2308#define NETDEV_JOIN 0x0014
42e52bf9 2309#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2310#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2311#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2312#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2313#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2314#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2315#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2316#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
296d8ee3 2317#define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
08294a26 2318#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2319
f629d208
JP
2320int register_netdevice_notifier(struct notifier_block *nb);
2321int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2322
2323struct netdev_notifier_info {
2324 struct net_device *dev;
2325};
2326
be9efd36
JP
2327struct netdev_notifier_change_info {
2328 struct netdev_notifier_info info; /* must be first */
2329 unsigned int flags_changed;
2330};
2331
0e4ead9d
JP
2332struct netdev_notifier_changeupper_info {
2333 struct netdev_notifier_info info; /* must be first */
2334 struct net_device *upper_dev; /* new upper dev */
2335 bool master; /* is upper dev master */
5e82b4b2 2336 bool linking; /* is the notification for link or unlink */
29bf24af 2337 void *upper_info; /* upper dev info */
0e4ead9d
JP
2338};
2339
04d48266
JP
2340struct netdev_notifier_changelowerstate_info {
2341 struct netdev_notifier_info info; /* must be first */
2342 void *lower_state_info; /* is lower dev state */
2343};
2344
75538c2b
CW
2345static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2346 struct net_device *dev)
2347{
2348 info->dev = dev;
2349}
2350
351638e7
JP
2351static inline struct net_device *
2352netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2353{
2354 return info->dev;
2355}
2356
f629d208 2357int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2358
2359
1da177e4
LT
2360extern rwlock_t dev_base_lock; /* Device list lock */
2361
881d966b
EB
2362#define for_each_netdev(net, d) \
2363 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2364#define for_each_netdev_reverse(net, d) \
2365 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2366#define for_each_netdev_rcu(net, d) \
2367 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2368#define for_each_netdev_safe(net, d, n) \
2369 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2370#define for_each_netdev_continue(net, d) \
2371 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2372#define for_each_netdev_continue_rcu(net, d) \
2373 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2374#define for_each_netdev_in_bond_rcu(bond, slave) \
2375 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2376 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2377#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2378
a050c33f
DL
2379static inline struct net_device *next_net_device(struct net_device *dev)
2380{
2381 struct list_head *lh;
2382 struct net *net;
2383
c346dca1 2384 net = dev_net(dev);
a050c33f
DL
2385 lh = dev->dev_list.next;
2386 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2387}
2388
ce81b76a
ED
2389static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2390{
2391 struct list_head *lh;
2392 struct net *net;
2393
2394 net = dev_net(dev);
ccf43438 2395 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2396 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2397}
2398
a050c33f
DL
2399static inline struct net_device *first_net_device(struct net *net)
2400{
2401 return list_empty(&net->dev_base_head) ? NULL :
2402 net_device_entry(net->dev_base_head.next);
2403}
7562f876 2404
ccf43438
ED
2405static inline struct net_device *first_net_device_rcu(struct net *net)
2406{
2407 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2408
2409 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2410}
2411
f629d208
JP
2412int netdev_boot_setup_check(struct net_device *dev);
2413unsigned long netdev_boot_base(const char *prefix, int unit);
2414struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2415 const char *hwaddr);
2416struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2417struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2418void dev_add_pack(struct packet_type *pt);
2419void dev_remove_pack(struct packet_type *pt);
2420void __dev_remove_pack(struct packet_type *pt);
2421void dev_add_offload(struct packet_offload *po);
2422void dev_remove_offload(struct packet_offload *po);
f629d208 2423
a54acb3a 2424int dev_get_iflink(const struct net_device *dev);
fc4099f1 2425int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2426struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2427 unsigned short mask);
f629d208
JP
2428struct net_device *dev_get_by_name(struct net *net, const char *name);
2429struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2430struct net_device *__dev_get_by_name(struct net *net, const char *name);
2431int dev_alloc_name(struct net_device *dev, const char *name);
2432int dev_open(struct net_device *dev);
7051b88a 2433void dev_close(struct net_device *dev);
2434void dev_close_many(struct list_head *head, bool unlink);
f629d208 2435void dev_disable_lro(struct net_device *dev);
0c4b51f0 2436int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2437int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2438int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2439int register_netdevice(struct net_device *dev);
2440void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2441void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2442static inline void unregister_netdevice(struct net_device *dev)
2443{
2444 unregister_netdevice_queue(dev, NULL);
2445}
2446
f629d208
JP
2447int netdev_refcnt_read(const struct net_device *dev);
2448void free_netdev(struct net_device *dev);
74d332c1 2449void netdev_freemem(struct net_device *dev);
f629d208
JP
2450void synchronize_net(void);
2451int init_dummy_netdev(struct net_device *dev);
937f1ba5 2452
f60e5990 2453DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2454#define XMIT_RECURSION_LIMIT 10
2455
f60e5990 2456static inline int dev_recursion_level(void)
2457{
2458 return this_cpu_read(xmit_recursion);
2459}
2460
f629d208
JP
2461struct net_device *dev_get_by_index(struct net *net, int ifindex);
2462struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2463struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2464struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2465int netdev_get_name(struct net *net, char *name, int ifindex);
2466int dev_restart(struct net_device *dev);
f629d208 2467int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2468
2469static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2470{
2471 return NAPI_GRO_CB(skb)->data_offset;
2472}
2473
2474static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2475{
2476 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2477}
2478
2479static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2480{
2481 NAPI_GRO_CB(skb)->data_offset += len;
2482}
2483
a5b1cf28
HX
2484static inline void *skb_gro_header_fast(struct sk_buff *skb,
2485 unsigned int offset)
86911732 2486{
a5b1cf28
HX
2487 return NAPI_GRO_CB(skb)->frag0 + offset;
2488}
78a478d0 2489
a5b1cf28
HX
2490static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2491{
2492 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2493}
78a478d0 2494
57ea52a8
HX
2495static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2496{
2497 NAPI_GRO_CB(skb)->frag0 = NULL;
2498 NAPI_GRO_CB(skb)->frag0_len = 0;
2499}
2500
a5b1cf28
HX
2501static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2502 unsigned int offset)
2503{
17dd759c
HX
2504 if (!pskb_may_pull(skb, hlen))
2505 return NULL;
2506
57ea52a8 2507 skb_gro_frag0_invalidate(skb);
17dd759c 2508 return skb->data + offset;
86911732 2509}
1da177e4 2510
36e7b1b8
HX
2511static inline void *skb_gro_network_header(struct sk_buff *skb)
2512{
78d3fd0b
HX
2513 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2514 skb_network_offset(skb);
36e7b1b8
HX
2515}
2516
bf5a755f
JC
2517static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2518 const void *start, unsigned int len)
2519{
573e8fca 2520 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2521 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2522 csum_partial(start, len, 0));
2523}
2524
573e8fca
TH
2525/* GRO checksum functions. These are logical equivalents of the normal
2526 * checksum functions (in skbuff.h) except that they operate on the GRO
2527 * offsets and fields in sk_buff.
2528 */
2529
2530__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2531
15e2396d
TH
2532static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2533{
b7fe10e5 2534 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2535}
2536
573e8fca
TH
2537static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2538 bool zero_okay,
2539 __sum16 check)
2540{
6edec0e6
TH
2541 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2542 skb_checksum_start_offset(skb) <
2543 skb_gro_offset(skb)) &&
15e2396d 2544 !skb_at_gro_remcsum_start(skb) &&
662880f4 2545 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2546 (!zero_okay || check));
2547}
2548
2549static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2550 __wsum psum)
2551{
2552 if (NAPI_GRO_CB(skb)->csum_valid &&
2553 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2554 return 0;
2555
2556 NAPI_GRO_CB(skb)->csum = psum;
2557
2558 return __skb_gro_checksum_complete(skb);
2559}
2560
573e8fca
TH
2561static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2562{
662880f4
TH
2563 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2564 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2565 NAPI_GRO_CB(skb)->csum_cnt--;
2566 } else {
2567 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2568 * verified a new top level checksum or an encapsulated one
2569 * during GRO. This saves work if we fallback to normal path.
2570 */
2571 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2572 }
2573}
2574
2575#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2576 compute_pseudo) \
2577({ \
2578 __sum16 __ret = 0; \
2579 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2580 __ret = __skb_gro_checksum_validate_complete(skb, \
2581 compute_pseudo(skb, proto)); \
219f1d79 2582 if (!__ret) \
573e8fca
TH
2583 skb_gro_incr_csum_unnecessary(skb); \
2584 __ret; \
2585})
2586
2587#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2588 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2589
2590#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2591 compute_pseudo) \
2592 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2593
2594#define skb_gro_checksum_simple_validate(skb) \
2595 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2596
d96535a1
TH
2597static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2598{
2599 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2600 !NAPI_GRO_CB(skb)->csum_valid);
2601}
2602
2603static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2604 __sum16 check, __wsum pseudo)
2605{
2606 NAPI_GRO_CB(skb)->csum = ~pseudo;
2607 NAPI_GRO_CB(skb)->csum_valid = 1;
2608}
2609
2610#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2611do { \
2612 if (__skb_gro_checksum_convert_check(skb)) \
2613 __skb_gro_checksum_convert(skb, check, \
2614 compute_pseudo(skb, proto)); \
2615} while (0)
2616
26c4f7da
TH
2617struct gro_remcsum {
2618 int offset;
2619 __wsum delta;
2620};
2621
2622static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2623{
846cd667 2624 grc->offset = 0;
26c4f7da
TH
2625 grc->delta = 0;
2626}
2627
b7fe10e5
TH
2628static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2629 unsigned int off, size_t hdrlen,
2630 int start, int offset,
2631 struct gro_remcsum *grc,
2632 bool nopartial)
dcdc8994
TH
2633{
2634 __wsum delta;
b7fe10e5 2635 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2636
2637 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2638
15e2396d 2639 if (!nopartial) {
b7fe10e5
TH
2640 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2641 return ptr;
2642 }
2643
2644 ptr = skb_gro_header_fast(skb, off);
2645 if (skb_gro_header_hard(skb, off + plen)) {
2646 ptr = skb_gro_header_slow(skb, off + plen, off);
2647 if (!ptr)
2648 return NULL;
15e2396d
TH
2649 }
2650
b7fe10e5
TH
2651 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2652 start, offset);
dcdc8994
TH
2653
2654 /* Adjust skb->csum since we changed the packet */
dcdc8994 2655 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2656
b7fe10e5 2657 grc->offset = off + hdrlen + offset;
26c4f7da 2658 grc->delta = delta;
b7fe10e5
TH
2659
2660 return ptr;
dcdc8994
TH
2661}
2662
26c4f7da
TH
2663static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2664 struct gro_remcsum *grc)
2665{
b7fe10e5
TH
2666 void *ptr;
2667 size_t plen = grc->offset + sizeof(u16);
2668
26c4f7da
TH
2669 if (!grc->delta)
2670 return;
2671
b7fe10e5
TH
2672 ptr = skb_gro_header_fast(skb, grc->offset);
2673 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2674 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2675 if (!ptr)
2676 return;
2677 }
2678
2679 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2680}
dcdc8994 2681
25393d3f
SK
2682#ifdef CONFIG_XFRM_OFFLOAD
2683static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2684{
2685 if (PTR_ERR(pp) != -EINPROGRESS)
2686 NAPI_GRO_CB(skb)->flush |= flush;
2687}
2688#else
5f114163
SK
2689static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2690{
2691 NAPI_GRO_CB(skb)->flush |= flush;
2692}
25393d3f 2693#endif
5f114163 2694
0c4e8581
SH
2695static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2696 unsigned short type,
3b04ddde 2697 const void *daddr, const void *saddr,
95c96174 2698 unsigned int len)
0c4e8581 2699{
f1ecfd5d 2700 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2701 return 0;
3b04ddde
SH
2702
2703 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2704}
2705
b95cce35
SH
2706static inline int dev_parse_header(const struct sk_buff *skb,
2707 unsigned char *haddr)
2708{
2709 const struct net_device *dev = skb->dev;
2710
1b83336b 2711 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2712 return 0;
3b04ddde 2713 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2714}
2715
2793a23a
WB
2716/* ll_header must have at least hard_header_len allocated */
2717static inline bool dev_validate_header(const struct net_device *dev,
2718 char *ll_header, int len)
2719{
2720 if (likely(len >= dev->hard_header_len))
2721 return true;
217e6fa2
WB
2722 if (len < dev->min_header_len)
2723 return false;
2793a23a
WB
2724
2725 if (capable(CAP_SYS_RAWIO)) {
2726 memset(ll_header + len, 0, dev->hard_header_len - len);
2727 return true;
2728 }
2729
2730 if (dev->header_ops && dev->header_ops->validate)
2731 return dev->header_ops->validate(ll_header, len);
2732
2733 return false;
2734}
2735
1da177e4 2736typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2737int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2738static inline int unregister_gifconf(unsigned int family)
2739{
2740 return register_gifconf(family, NULL);
2741}
2742
99bbc707 2743#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2744#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2745struct sd_flow_limit {
2746 u64 count;
2747 unsigned int num_buckets;
2748 unsigned int history_head;
2749 u16 history[FLOW_LIMIT_HISTORY];
2750 u8 buckets[];
2751};
2752
2753extern int netdev_flow_limit_table_len;
2754#endif /* CONFIG_NET_FLOW_LIMIT */
2755
1da177e4 2756/*
5e82b4b2 2757 * Incoming packets are placed on per-CPU queues
1da177e4 2758 */
d94d9fee 2759struct softnet_data {
1da177e4 2760 struct list_head poll_list;
6e7676c1 2761 struct sk_buff_head process_queue;
1da177e4 2762
dee42870 2763 /* stats */
cd7b5396
DM
2764 unsigned int processed;
2765 unsigned int time_squeeze;
cd7b5396 2766 unsigned int received_rps;
fd793d89 2767#ifdef CONFIG_RPS
88751275 2768 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2769#endif
2770#ifdef CONFIG_NET_FLOW_LIMIT
2771 struct sd_flow_limit __rcu *flow_limit;
2772#endif
2773 struct Qdisc *output_queue;
2774 struct Qdisc **output_queue_tailp;
2775 struct sk_buff *completion_queue;
88751275 2776
4cdb1e2e 2777#ifdef CONFIG_RPS
501e7ef5
ED
2778 /* input_queue_head should be written by cpu owning this struct,
2779 * and only read by other cpus. Worth using a cache line.
2780 */
2781 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2782
2783 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2784 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2785 struct softnet_data *rps_ipi_next;
2786 unsigned int cpu;
76cc8b13 2787 unsigned int input_queue_tail;
1e94d72f 2788#endif
95c96174 2789 unsigned int dropped;
0a9627f2 2790 struct sk_buff_head input_pkt_queue;
bea3348e 2791 struct napi_struct backlog;
99bbc707 2792
1da177e4
LT
2793};
2794
76cc8b13 2795static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2796{
2797#ifdef CONFIG_RPS
76cc8b13
TH
2798 sd->input_queue_head++;
2799#endif
2800}
2801
2802static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2803 unsigned int *qtail)
2804{
2805#ifdef CONFIG_RPS
2806 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2807#endif
2808}
2809
0a9627f2 2810DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2811
f629d208 2812void __netif_schedule(struct Qdisc *q);
46e5da40 2813void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2814
fd2ea0a7
DM
2815static inline void netif_tx_schedule_all(struct net_device *dev)
2816{
2817 unsigned int i;
2818
2819 for (i = 0; i < dev->num_tx_queues; i++)
2820 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2821}
2822
f9a7cbbf 2823static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2824{
73466498 2825 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2826}
2827
bea3348e
SH
2828/**
2829 * netif_start_queue - allow transmit
2830 * @dev: network device
2831 *
2832 * Allow upper layers to call the device hard_start_xmit routine.
2833 */
1da177e4
LT
2834static inline void netif_start_queue(struct net_device *dev)
2835{
e8a0464c 2836 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2837}
2838
fd2ea0a7
DM
2839static inline void netif_tx_start_all_queues(struct net_device *dev)
2840{
2841 unsigned int i;
2842
2843 for (i = 0; i < dev->num_tx_queues; i++) {
2844 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2845 netif_tx_start_queue(txq);
2846 }
2847}
2848
46e5da40 2849void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2850
d29f749e
DJ
2851/**
2852 * netif_wake_queue - restart transmit
2853 * @dev: network device
2854 *
2855 * Allow upper layers to call the device hard_start_xmit routine.
2856 * Used for flow control when transmit resources are available.
2857 */
79d16385
DM
2858static inline void netif_wake_queue(struct net_device *dev)
2859{
e8a0464c 2860 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2861}
2862
fd2ea0a7
DM
2863static inline void netif_tx_wake_all_queues(struct net_device *dev)
2864{
2865 unsigned int i;
2866
2867 for (i = 0; i < dev->num_tx_queues; i++) {
2868 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2869 netif_tx_wake_queue(txq);
2870 }
2871}
2872
f9a7cbbf 2873static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2874{
73466498 2875 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2876}
2877
bea3348e
SH
2878/**
2879 * netif_stop_queue - stop transmitted packets
2880 * @dev: network device
2881 *
2882 * Stop upper layers calling the device hard_start_xmit routine.
2883 * Used for flow control when transmit resources are unavailable.
2884 */
1da177e4
LT
2885static inline void netif_stop_queue(struct net_device *dev)
2886{
e8a0464c 2887 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2888}
2889
a2029240 2890void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2891
4d29515f 2892static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2893{
73466498 2894 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2895}
2896
bea3348e
SH
2897/**
2898 * netif_queue_stopped - test if transmit queue is flowblocked
2899 * @dev: network device
2900 *
2901 * Test if transmit queue on device is currently unable to send.
2902 */
4d29515f 2903static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2904{
e8a0464c 2905 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2906}
2907
4d29515f 2908static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2909{
73466498
TH
2910 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2911}
2912
8e2f1a63
DB
2913static inline bool
2914netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2915{
2916 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2917}
2918
8e2f1a63
DB
2919static inline bool
2920netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2921{
2922 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2923}
2924
53511453
ED
2925/**
2926 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2927 * @dev_queue: pointer to transmit queue
2928 *
2929 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2930 * to give appropriate hint to the CPU.
53511453
ED
2931 */
2932static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2933{
2934#ifdef CONFIG_BQL
2935 prefetchw(&dev_queue->dql.num_queued);
2936#endif
2937}
2938
2939/**
2940 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2941 * @dev_queue: pointer to transmit queue
2942 *
2943 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2944 * to give appropriate hint to the CPU.
53511453
ED
2945 */
2946static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2947{
2948#ifdef CONFIG_BQL
2949 prefetchw(&dev_queue->dql.limit);
2950#endif
2951}
2952
c5d67bd7
TH
2953static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2954 unsigned int bytes)
2955{
114cf580
TH
2956#ifdef CONFIG_BQL
2957 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2958
2959 if (likely(dql_avail(&dev_queue->dql) >= 0))
2960 return;
2961
2962 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2963
2964 /*
2965 * The XOFF flag must be set before checking the dql_avail below,
2966 * because in netdev_tx_completed_queue we update the dql_completed
2967 * before checking the XOFF flag.
2968 */
2969 smp_mb();
2970
2971 /* check again in case another CPU has just made room avail */
2972 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2973 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2974#endif
c5d67bd7
TH
2975}
2976
0042d0c8
FF
2977/**
2978 * netdev_sent_queue - report the number of bytes queued to hardware
2979 * @dev: network device
2980 * @bytes: number of bytes queued to the hardware device queue
2981 *
2982 * Report the number of bytes queued for sending/completion to the network
2983 * device hardware queue. @bytes should be a good approximation and should
2984 * exactly match netdev_completed_queue() @bytes
2985 */
c5d67bd7
TH
2986static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2987{
2988 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2989}
2990
2991static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2992 unsigned int pkts, unsigned int bytes)
c5d67bd7 2993{
114cf580 2994#ifdef CONFIG_BQL
b37c0fbe
AD
2995 if (unlikely(!bytes))
2996 return;
2997
2998 dql_completed(&dev_queue->dql, bytes);
2999
3000 /*
3001 * Without the memory barrier there is a small possiblity that
3002 * netdev_tx_sent_queue will miss the update and cause the queue to
3003 * be stopped forever
3004 */
3005 smp_mb();
3006
3007 if (dql_avail(&dev_queue->dql) < 0)
3008 return;
3009
3010 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3011 netif_schedule_queue(dev_queue);
114cf580 3012#endif
c5d67bd7
TH
3013}
3014
0042d0c8
FF
3015/**
3016 * netdev_completed_queue - report bytes and packets completed by device
3017 * @dev: network device
3018 * @pkts: actual number of packets sent over the medium
3019 * @bytes: actual number of bytes sent over the medium
3020 *
3021 * Report the number of bytes and packets transmitted by the network device
3022 * hardware queue over the physical medium, @bytes must exactly match the
3023 * @bytes amount passed to netdev_sent_queue()
3024 */
c5d67bd7 3025static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3026 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3027{
3028 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3029}
3030
3031static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3032{
114cf580 3033#ifdef CONFIG_BQL
5c490354 3034 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3035 dql_reset(&q->dql);
3036#endif
c5d67bd7
TH
3037}
3038
0042d0c8
FF
3039/**
3040 * netdev_reset_queue - reset the packets and bytes count of a network device
3041 * @dev_queue: network device
3042 *
3043 * Reset the bytes and packet count of a network device and clear the
3044 * software flow control OFF bit for this network device
3045 */
c5d67bd7
TH
3046static inline void netdev_reset_queue(struct net_device *dev_queue)
3047{
3048 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3049}
3050
b9507bda
DB
3051/**
3052 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3053 * @dev: network device
3054 * @queue_index: given tx queue index
3055 *
3056 * Returns 0 if given tx queue index >= number of device tx queues,
3057 * otherwise returns the originally passed tx queue index.
3058 */
3059static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3060{
3061 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3062 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3063 dev->name, queue_index,
3064 dev->real_num_tx_queues);
3065 return 0;
3066 }
3067
3068 return queue_index;
3069}
3070
bea3348e
SH
3071/**
3072 * netif_running - test if up
3073 * @dev: network device
3074 *
3075 * Test if the device has been brought up.
3076 */
4d29515f 3077static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3078{
3079 return test_bit(__LINK_STATE_START, &dev->state);
3080}
3081
f25f4e44 3082/*
5e82b4b2 3083 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3084 * stop, and a check if it's stopped. All other device management is
3085 * done at the overall netdevice level.
3086 * Also test the device if we're multiqueue.
3087 */
bea3348e
SH
3088
3089/**
3090 * netif_start_subqueue - allow sending packets on subqueue
3091 * @dev: network device
3092 * @queue_index: sub queue index
3093 *
3094 * Start individual transmit queue of a device with multiple transmit queues.
3095 */
f25f4e44
PWJ
3096static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3097{
fd2ea0a7 3098 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3099
3100 netif_tx_start_queue(txq);
f25f4e44
PWJ
3101}
3102
bea3348e
SH
3103/**
3104 * netif_stop_subqueue - stop sending packets on subqueue
3105 * @dev: network device
3106 * @queue_index: sub queue index
3107 *
3108 * Stop individual transmit queue of a device with multiple transmit queues.
3109 */
f25f4e44
PWJ
3110static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3111{
fd2ea0a7 3112 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3113 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3114}
3115
bea3348e
SH
3116/**
3117 * netif_subqueue_stopped - test status of subqueue
3118 * @dev: network device
3119 * @queue_index: sub queue index
3120 *
3121 * Check individual transmit queue of a device with multiple transmit queues.
3122 */
4d29515f
DM
3123static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3124 u16 queue_index)
f25f4e44 3125{
fd2ea0a7 3126 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3127
3128 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3129}
3130
4d29515f
DM
3131static inline bool netif_subqueue_stopped(const struct net_device *dev,
3132 struct sk_buff *skb)
668f895a
PE
3133{
3134 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3135}
bea3348e 3136
738b35cc
FF
3137/**
3138 * netif_wake_subqueue - allow sending packets on subqueue
3139 * @dev: network device
3140 * @queue_index: sub queue index
3141 *
3142 * Resume individual transmit queue of a device with multiple transmit queues.
3143 */
3144static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3145{
3146 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3147
3148 netif_tx_wake_queue(txq);
3149}
f25f4e44 3150
537c00de 3151#ifdef CONFIG_XPS
53af53ae 3152int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3153 u16 index);
537c00de
AD
3154#else
3155static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3156 const struct cpumask *mask,
537c00de
AD
3157 u16 index)
3158{
3159 return 0;
3160}
3161#endif
3162
5605c762
JP
3163u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3164 unsigned int num_tx_queues);
3165
a3d22a68
VZ
3166/*
3167 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3168 * as a distribution range limit for the returned value.
3169 */
3170static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3171 struct sk_buff *skb)
a3d22a68
VZ
3172{
3173 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3174}
3175
bea3348e
SH
3176/**
3177 * netif_is_multiqueue - test if device has multiple transmit queues
3178 * @dev: network device
3179 *
3180 * Check if device has multiple transmit queues
bea3348e 3181 */
4d29515f 3182static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3183{
a02cec21 3184 return dev->num_tx_queues > 1;
f25f4e44 3185}
1da177e4 3186
f629d208 3187int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3188
a953be53 3189#ifdef CONFIG_SYSFS
f629d208 3190int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3191#else
3192static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3193 unsigned int rxq)
3194{
3195 return 0;
3196}
3197#endif
3198
a953be53
MD
3199#ifdef CONFIG_SYSFS
3200static inline unsigned int get_netdev_rx_queue_index(
3201 struct netdev_rx_queue *queue)
3202{
3203 struct net_device *dev = queue->dev;
3204 int index = queue - dev->_rx;
3205
3206 BUG_ON(index >= dev->num_rx_queues);
3207 return index;
3208}
3209#endif
3210
16917b87 3211#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3212int netif_get_num_default_rss_queues(void);
16917b87 3213
e6247027
ED
3214enum skb_free_reason {
3215 SKB_REASON_CONSUMED,
3216 SKB_REASON_DROPPED,
3217};
3218
3219void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3220void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3221
e6247027
ED
3222/*
3223 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3224 * interrupt context or with hardware interrupts being disabled.
3225 * (in_irq() || irqs_disabled())
3226 *
3227 * We provide four helpers that can be used in following contexts :
3228 *
3229 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3230 * replacing kfree_skb(skb)
3231 *
3232 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3233 * Typically used in place of consume_skb(skb) in TX completion path
3234 *
3235 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3236 * replacing kfree_skb(skb)
3237 *
3238 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3239 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3240 */
e6247027
ED
3241static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3242{
3243 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3244}
3245
3246static inline void dev_consume_skb_irq(struct sk_buff *skb)
3247{
3248 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3249}
3250
3251static inline void dev_kfree_skb_any(struct sk_buff *skb)
3252{
3253 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3254}
3255
3256static inline void dev_consume_skb_any(struct sk_buff *skb)
3257{
3258 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3259}
1da177e4 3260
f629d208
JP
3261int netif_rx(struct sk_buff *skb);
3262int netif_rx_ni(struct sk_buff *skb);
04eb4489 3263int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3264gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3265void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3266struct sk_buff *napi_get_frags(struct napi_struct *napi);
3267gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3268struct packet_offload *gro_find_receive_by_type(__be16 type);
3269struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3270
3271static inline void napi_free_frags(struct napi_struct *napi)
3272{
3273 kfree_skb(napi->skb);
3274 napi->skb = NULL;
3275}
3276
24b27fc4 3277bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3278int netdev_rx_handler_register(struct net_device *dev,
3279 rx_handler_func_t *rx_handler,
3280 void *rx_handler_data);
3281void netdev_rx_handler_unregister(struct net_device *dev);
3282
3283bool dev_valid_name(const char *name);
3284int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3285int dev_ethtool(struct net *net, struct ifreq *);
3286unsigned int dev_get_flags(const struct net_device *);
3287int __dev_change_flags(struct net_device *, unsigned int flags);
3288int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3289void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3290 unsigned int gchanges);
f629d208
JP
3291int dev_change_name(struct net_device *, const char *);
3292int dev_set_alias(struct net_device *, const char *, size_t);
3293int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3294int __dev_set_mtu(struct net_device *, int);
f629d208
JP
3295int dev_set_mtu(struct net_device *, int);
3296void dev_set_group(struct net_device *, int);
3297int dev_set_mac_address(struct net_device *, struct sockaddr *);
3298int dev_change_carrier(struct net_device *, bool new_carrier);
3299int dev_get_phys_port_id(struct net_device *dev,
02637fce 3300 struct netdev_phys_item_id *ppid);
db24a904
DA
3301int dev_get_phys_port_name(struct net_device *dev,
3302 char *name, size_t len);
d746d707 3303int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3304struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3305struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3306 struct netdev_queue *txq, int *ret);
d67b9cd2
DB
3307
3308typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3309int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3310 int fd, u32 flags);
ce158e58 3311u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
d67b9cd2 3312
a0265d28 3313int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3314int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3315bool is_skb_forwardable(const struct net_device *dev,
3316 const struct sk_buff *skb);
1da177e4 3317
4e3264d2
MKL
3318static __always_inline int ____dev_forward_skb(struct net_device *dev,
3319 struct sk_buff *skb)
3320{
3321 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3322 unlikely(!is_skb_forwardable(dev, skb))) {
3323 atomic_long_inc(&dev->rx_dropped);
3324 kfree_skb(skb);
3325 return NET_RX_DROP;
3326 }
3327
3328 skb_scrub_packet(skb, true);
3329 skb->priority = 0;
3330 return 0;
3331}
3332
74b20582
DA
3333void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3334
20380731 3335extern int netdev_budget;
7acf8a1e 3336extern unsigned int netdev_budget_usecs;
1da177e4
LT
3337
3338/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3339void netdev_run_todo(void);
1da177e4 3340
bea3348e
SH
3341/**
3342 * dev_put - release reference to device
3343 * @dev: network device
3344 *
9ef4429b 3345 * Release reference to device to allow it to be freed.
bea3348e 3346 */
1da177e4
LT
3347static inline void dev_put(struct net_device *dev)
3348{
933393f5 3349 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3350}
3351
bea3348e
SH
3352/**
3353 * dev_hold - get reference to device
3354 * @dev: network device
3355 *
9ef4429b 3356 * Hold reference to device to keep it from being freed.
bea3348e 3357 */
15333061
SH
3358static inline void dev_hold(struct net_device *dev)
3359{
933393f5 3360 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3361}
1da177e4
LT
3362
3363/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3364 * and _off may be called from IRQ context, but it is caller
3365 * who is responsible for serialization of these calls.
b00055aa
SR
3366 *
3367 * The name carrier is inappropriate, these functions should really be
3368 * called netif_lowerlayer_*() because they represent the state of any
3369 * kind of lower layer not just hardware media.
1da177e4
LT
3370 */
3371
f629d208
JP
3372void linkwatch_init_dev(struct net_device *dev);
3373void linkwatch_fire_event(struct net_device *dev);
3374void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3375
bea3348e
SH
3376/**
3377 * netif_carrier_ok - test if carrier present
3378 * @dev: network device
3379 *
3380 * Check if carrier is present on device
3381 */
4d29515f 3382static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3383{
3384 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3385}
3386
f629d208 3387unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3388
f629d208 3389void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3390
f629d208 3391void netif_carrier_on(struct net_device *dev);
1da177e4 3392
f629d208 3393void netif_carrier_off(struct net_device *dev);
1da177e4 3394
bea3348e
SH
3395/**
3396 * netif_dormant_on - mark device as dormant.
3397 * @dev: network device
3398 *
3399 * Mark device as dormant (as per RFC2863).
3400 *
3401 * The dormant state indicates that the relevant interface is not
3402 * actually in a condition to pass packets (i.e., it is not 'up') but is
3403 * in a "pending" state, waiting for some external event. For "on-
3404 * demand" interfaces, this new state identifies the situation where the
3405 * interface is waiting for events to place it in the up state.
bea3348e 3406 */
b00055aa
SR
3407static inline void netif_dormant_on(struct net_device *dev)
3408{
3409 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3410 linkwatch_fire_event(dev);
3411}
3412
bea3348e
SH
3413/**
3414 * netif_dormant_off - set device as not dormant.
3415 * @dev: network device
3416 *
3417 * Device is not in dormant state.
3418 */
b00055aa
SR
3419static inline void netif_dormant_off(struct net_device *dev)
3420{
3421 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3422 linkwatch_fire_event(dev);
3423}
3424
bea3348e 3425/**
8ecbc40a 3426 * netif_dormant - test if device is dormant
bea3348e
SH
3427 * @dev: network device
3428 *
8ecbc40a 3429 * Check if device is dormant.
bea3348e 3430 */
4d29515f 3431static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3432{
3433 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3434}
3435
3436
bea3348e
SH
3437/**
3438 * netif_oper_up - test if device is operational
3439 * @dev: network device
3440 *
3441 * Check if carrier is operational
3442 */
4d29515f 3443static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3444{
b00055aa
SR
3445 return (dev->operstate == IF_OPER_UP ||
3446 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3447}
3448
bea3348e
SH
3449/**
3450 * netif_device_present - is device available or removed
3451 * @dev: network device
3452 *
3453 * Check if device has not been removed from system.
3454 */
4d29515f 3455static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3456{
3457 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3458}
3459
f629d208 3460void netif_device_detach(struct net_device *dev);
1da177e4 3461
f629d208 3462void netif_device_attach(struct net_device *dev);
1da177e4
LT
3463
3464/*
3465 * Network interface message level settings
3466 */
1da177e4
LT
3467
3468enum {
3469 NETIF_MSG_DRV = 0x0001,
3470 NETIF_MSG_PROBE = 0x0002,
3471 NETIF_MSG_LINK = 0x0004,
3472 NETIF_MSG_TIMER = 0x0008,
3473 NETIF_MSG_IFDOWN = 0x0010,
3474 NETIF_MSG_IFUP = 0x0020,
3475 NETIF_MSG_RX_ERR = 0x0040,
3476 NETIF_MSG_TX_ERR = 0x0080,
3477 NETIF_MSG_TX_QUEUED = 0x0100,
3478 NETIF_MSG_INTR = 0x0200,
3479 NETIF_MSG_TX_DONE = 0x0400,
3480 NETIF_MSG_RX_STATUS = 0x0800,
3481 NETIF_MSG_PKTDATA = 0x1000,
3482 NETIF_MSG_HW = 0x2000,
3483 NETIF_MSG_WOL = 0x4000,
3484};
3485
3486#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3487#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3488#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3489#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3490#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3491#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3492#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3493#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3494#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3495#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3496#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3497#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3498#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3499#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3500#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3501
3502static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3503{
3504 /* use default */
3505 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3506 return default_msg_enable_bits;
3507 if (debug_value == 0) /* no output */
3508 return 0;
3509 /* set low N bits */
3510 return (1 << debug_value) - 1;
3511}
3512
c773e847 3513static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3514{
c773e847
DM
3515 spin_lock(&txq->_xmit_lock);
3516 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3517}
3518
5a717f4f
MT
3519static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3520{
3521 __acquire(&txq->_xmit_lock);
3522 return true;
3523}
3524
3525static inline void __netif_tx_release(struct netdev_queue *txq)
3526{
3527 __release(&txq->_xmit_lock);
3528}
3529
fd2ea0a7
DM
3530static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3531{
3532 spin_lock_bh(&txq->_xmit_lock);
3533 txq->xmit_lock_owner = smp_processor_id();
3534}
3535
4d29515f 3536static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3537{
4d29515f 3538 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3539 if (likely(ok))
3540 txq->xmit_lock_owner = smp_processor_id();
3541 return ok;
3542}
3543
3544static inline void __netif_tx_unlock(struct netdev_queue *txq)
3545{
3546 txq->xmit_lock_owner = -1;
3547 spin_unlock(&txq->_xmit_lock);
3548}
3549
3550static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3551{
3552 txq->xmit_lock_owner = -1;
3553 spin_unlock_bh(&txq->_xmit_lock);
3554}
3555
08baf561
ED
3556static inline void txq_trans_update(struct netdev_queue *txq)
3557{
3558 if (txq->xmit_lock_owner != -1)
3559 txq->trans_start = jiffies;
3560}
3561
ba162f8e
FW
3562/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3563static inline void netif_trans_update(struct net_device *dev)
3564{
9b36627a
FW
3565 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3566
3567 if (txq->trans_start != jiffies)
3568 txq->trans_start = jiffies;
ba162f8e
FW
3569}
3570
d29f749e
DJ
3571/**
3572 * netif_tx_lock - grab network device transmit lock
3573 * @dev: network device
d29f749e
DJ
3574 *
3575 * Get network device transmit lock
3576 */
22dd7495
JHS
3577static inline void netif_tx_lock(struct net_device *dev)
3578{
e8a0464c 3579 unsigned int i;
c3f26a26 3580 int cpu;
c773e847 3581
c3f26a26
DM
3582 spin_lock(&dev->tx_global_lock);
3583 cpu = smp_processor_id();
e8a0464c
DM
3584 for (i = 0; i < dev->num_tx_queues; i++) {
3585 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3586
3587 /* We are the only thread of execution doing a
3588 * freeze, but we have to grab the _xmit_lock in
3589 * order to synchronize with threads which are in
3590 * the ->hard_start_xmit() handler and already
3591 * checked the frozen bit.
3592 */
e8a0464c 3593 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3594 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3595 __netif_tx_unlock(txq);
e8a0464c 3596 }
932ff279
HX
3597}
3598
3599static inline void netif_tx_lock_bh(struct net_device *dev)
3600{
e8a0464c
DM
3601 local_bh_disable();
3602 netif_tx_lock(dev);
932ff279
HX
3603}
3604
932ff279
HX
3605static inline void netif_tx_unlock(struct net_device *dev)
3606{
e8a0464c
DM
3607 unsigned int i;
3608
3609 for (i = 0; i < dev->num_tx_queues; i++) {
3610 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3611
c3f26a26
DM
3612 /* No need to grab the _xmit_lock here. If the
3613 * queue is not stopped for another reason, we
3614 * force a schedule.
3615 */
3616 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3617 netif_schedule_queue(txq);
c3f26a26
DM
3618 }
3619 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3620}
3621
3622static inline void netif_tx_unlock_bh(struct net_device *dev)
3623{
e8a0464c
DM
3624 netif_tx_unlock(dev);
3625 local_bh_enable();
932ff279
HX
3626}
3627
c773e847 3628#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3629 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3630 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3631 } else { \
3632 __netif_tx_acquire(txq); \
22dd7495
JHS
3633 } \
3634}
3635
5efeac44
EB
3636#define HARD_TX_TRYLOCK(dev, txq) \
3637 (((dev->features & NETIF_F_LLTX) == 0) ? \
3638 __netif_tx_trylock(txq) : \
5a717f4f 3639 __netif_tx_acquire(txq))
5efeac44 3640
c773e847 3641#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3642 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3643 __netif_tx_unlock(txq); \
5a717f4f
MT
3644 } else { \
3645 __netif_tx_release(txq); \
22dd7495
JHS
3646 } \
3647}
3648
1da177e4
LT
3649static inline void netif_tx_disable(struct net_device *dev)
3650{
fd2ea0a7 3651 unsigned int i;
c3f26a26 3652 int cpu;
fd2ea0a7 3653
c3f26a26
DM
3654 local_bh_disable();
3655 cpu = smp_processor_id();
fd2ea0a7
DM
3656 for (i = 0; i < dev->num_tx_queues; i++) {
3657 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3658
3659 __netif_tx_lock(txq, cpu);
fd2ea0a7 3660 netif_tx_stop_queue(txq);
c3f26a26 3661 __netif_tx_unlock(txq);
fd2ea0a7 3662 }
c3f26a26 3663 local_bh_enable();
1da177e4
LT
3664}
3665
e308a5d8
DM
3666static inline void netif_addr_lock(struct net_device *dev)
3667{
3668 spin_lock(&dev->addr_list_lock);
3669}
3670
2429f7ac
JP
3671static inline void netif_addr_lock_nested(struct net_device *dev)
3672{
25175ba5
VY
3673 int subclass = SINGLE_DEPTH_NESTING;
3674
3675 if (dev->netdev_ops->ndo_get_lock_subclass)
3676 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3677
3678 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3679}
3680
e308a5d8
DM
3681static inline void netif_addr_lock_bh(struct net_device *dev)
3682{
3683 spin_lock_bh(&dev->addr_list_lock);
3684}
3685
3686static inline void netif_addr_unlock(struct net_device *dev)
3687{
3688 spin_unlock(&dev->addr_list_lock);
3689}
3690
3691static inline void netif_addr_unlock_bh(struct net_device *dev)
3692{
3693 spin_unlock_bh(&dev->addr_list_lock);
3694}
3695
f001fde5 3696/*
31278e71 3697 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3698 * rcu_read_lock held.
3699 */
3700#define for_each_dev_addr(dev, ha) \
31278e71 3701 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3702
1da177e4
LT
3703/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3704
f629d208 3705void ether_setup(struct net_device *dev);
1da177e4
LT
3706
3707/* Support for loadable net-drivers */
f629d208 3708struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3709 unsigned char name_assign_type,
f629d208
JP
3710 void (*setup)(struct net_device *),
3711 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3712#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3713 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3714
c835a677
TG
3715#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3716 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3717 count)
36909ea4 3718
f629d208
JP
3719int register_netdev(struct net_device *dev);
3720void unregister_netdev(struct net_device *dev);
f001fde5 3721
22bedad3 3722/* General hardware address lists handling functions */
f629d208
JP
3723int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3724 struct netdev_hw_addr_list *from_list, int addr_len);
3725void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3726 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3727int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3728 struct net_device *dev,
3729 int (*sync)(struct net_device *, const unsigned char *),
3730 int (*unsync)(struct net_device *,
3731 const unsigned char *));
3732void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3733 struct net_device *dev,
3734 int (*unsync)(struct net_device *,
3735 const unsigned char *));
f629d208 3736void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3737
f001fde5 3738/* Functions used for device addresses handling */
f629d208
JP
3739int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3740 unsigned char addr_type);
3741int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3742 unsigned char addr_type);
f629d208
JP
3743void dev_addr_flush(struct net_device *dev);
3744int dev_addr_init(struct net_device *dev);
a748ee24
JP
3745
3746/* Functions used for unicast addresses handling */
f629d208
JP
3747int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3748int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3749int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3750int dev_uc_sync(struct net_device *to, struct net_device *from);
3751int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3752void dev_uc_unsync(struct net_device *to, struct net_device *from);
3753void dev_uc_flush(struct net_device *dev);
3754void dev_uc_init(struct net_device *dev);
f001fde5 3755
670e5b8e
AD
3756/**
3757 * __dev_uc_sync - Synchonize device's unicast list
3758 * @dev: device to sync
3759 * @sync: function to call if address should be added
3760 * @unsync: function to call if address should be removed
3761 *
3762 * Add newly added addresses to the interface, and release
3763 * addresses that have been deleted.
5e82b4b2 3764 */
670e5b8e
AD
3765static inline int __dev_uc_sync(struct net_device *dev,
3766 int (*sync)(struct net_device *,
3767 const unsigned char *),
3768 int (*unsync)(struct net_device *,
3769 const unsigned char *))
3770{
3771 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3772}
3773
3774/**
e793c0f7 3775 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3776 * @dev: device to sync
3777 * @unsync: function to call if address should be removed
3778 *
3779 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3780 */
670e5b8e
AD
3781static inline void __dev_uc_unsync(struct net_device *dev,
3782 int (*unsync)(struct net_device *,
3783 const unsigned char *))
3784{
3785 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3786}
3787
22bedad3 3788/* Functions used for multicast addresses handling */
f629d208
JP
3789int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3790int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3791int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3792int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3793int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3794int dev_mc_sync(struct net_device *to, struct net_device *from);
3795int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3796void dev_mc_unsync(struct net_device *to, struct net_device *from);
3797void dev_mc_flush(struct net_device *dev);
3798void dev_mc_init(struct net_device *dev);
f001fde5 3799
670e5b8e
AD
3800/**
3801 * __dev_mc_sync - Synchonize device's multicast list
3802 * @dev: device to sync
3803 * @sync: function to call if address should be added
3804 * @unsync: function to call if address should be removed
3805 *
3806 * Add newly added addresses to the interface, and release
3807 * addresses that have been deleted.
5e82b4b2 3808 */
670e5b8e
AD
3809static inline int __dev_mc_sync(struct net_device *dev,
3810 int (*sync)(struct net_device *,
3811 const unsigned char *),
3812 int (*unsync)(struct net_device *,
3813 const unsigned char *))
3814{
3815 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3816}
3817
3818/**
e793c0f7 3819 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3820 * @dev: device to sync
3821 * @unsync: function to call if address should be removed
3822 *
3823 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3824 */
670e5b8e
AD
3825static inline void __dev_mc_unsync(struct net_device *dev,
3826 int (*unsync)(struct net_device *,
3827 const unsigned char *))
3828{
3829 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3830}
3831
4417da66 3832/* Functions used for secondary unicast and multicast support */
f629d208
JP
3833void dev_set_rx_mode(struct net_device *dev);
3834void __dev_set_rx_mode(struct net_device *dev);
3835int dev_set_promiscuity(struct net_device *dev, int inc);
3836int dev_set_allmulti(struct net_device *dev, int inc);
3837void netdev_state_change(struct net_device *dev);
3838void netdev_notify_peers(struct net_device *dev);
3839void netdev_features_change(struct net_device *dev);
1da177e4 3840/* Load a device via the kmod */
f629d208
JP
3841void dev_load(struct net *net, const char *name);
3842struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3843 struct rtnl_link_stats64 *storage);
3844void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3845 const struct net_device_stats *netdev_stats);
eeda3fd6 3846
1da177e4 3847extern int netdev_max_backlog;
3b098e2d 3848extern int netdev_tstamp_prequeue;
1da177e4 3849extern int weight_p;
3d48b53f
MT
3850extern int dev_weight_rx_bias;
3851extern int dev_weight_tx_bias;
3852extern int dev_rx_weight;
3853extern int dev_tx_weight;
9ff162a8 3854
f629d208 3855bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3856struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3857 struct list_head **iter);
f629d208
JP
3858struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3859 struct list_head **iter);
8b5be856 3860
44a40855
VY
3861/* iterate through upper list, must be called under RCU read lock */
3862#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3863 for (iter = &(dev)->adj_list.upper, \
3864 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3865 updev; \
3866 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3867
1a3f060c
DA
3868int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3869 int (*fn)(struct net_device *upper_dev,
3870 void *data),
3871 void *data);
3872
3873bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3874 struct net_device *upper_dev);
3875
f629d208
JP
3876void *netdev_lower_get_next_private(struct net_device *dev,
3877 struct list_head **iter);
3878void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3879 struct list_head **iter);
31088a11
VF
3880
3881#define netdev_for_each_lower_private(dev, priv, iter) \
3882 for (iter = (dev)->adj_list.lower.next, \
3883 priv = netdev_lower_get_next_private(dev, &(iter)); \
3884 priv; \
3885 priv = netdev_lower_get_next_private(dev, &(iter)))
3886
3887#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3888 for (iter = &(dev)->adj_list.lower, \
3889 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3890 priv; \
3891 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3892
4085ebe8
VY
3893void *netdev_lower_get_next(struct net_device *dev,
3894 struct list_head **iter);
7ce856aa 3895
4085ebe8 3896#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3897 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3898 ldev = netdev_lower_get_next(dev, &(iter)); \
3899 ldev; \
3900 ldev = netdev_lower_get_next(dev, &(iter)))
3901
7ce856aa
JP
3902struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3903 struct list_head **iter);
3904struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3905 struct list_head **iter);
3906
1a3f060c
DA
3907int netdev_walk_all_lower_dev(struct net_device *dev,
3908 int (*fn)(struct net_device *lower_dev,
3909 void *data),
3910 void *data);
3911int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3912 int (*fn)(struct net_device *lower_dev,
3913 void *data),
3914 void *data);
3915
f629d208 3916void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3917void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3918struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3919struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3920int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3921int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3922 struct net_device *upper_dev,
29bf24af 3923 void *upper_priv, void *upper_info);
f629d208
JP
3924void netdev_upper_dev_unlink(struct net_device *dev,
3925 struct net_device *upper_dev);
5bb025fa 3926void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3927void *netdev_lower_dev_get_private(struct net_device *dev,
3928 struct net_device *lower_dev);
04d48266
JP
3929void netdev_lower_state_changed(struct net_device *lower_dev,
3930 void *lower_state_info);
960fb622
ED
3931
3932/* RSS keys are 40 or 52 bytes long */
3933#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3934extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3935void netdev_rss_key_fill(void *buffer, size_t len);
3936
952fcfd0 3937int dev_get_nest_level(struct net_device *dev);
f629d208 3938int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 3939int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
3940int skb_csum_hwoffload_help(struct sk_buff *skb,
3941 const netdev_features_t features);
3942
f629d208
JP
3943struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3944 netdev_features_t features, bool tx_path);
3945struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3946 netdev_features_t features);
12b0004d 3947
61bd3857
MS
3948struct netdev_bonding_info {
3949 ifslave slave;
3950 ifbond master;
3951};
3952
3953struct netdev_notifier_bonding_info {
3954 struct netdev_notifier_info info; /* must be first */
3955 struct netdev_bonding_info bonding_info;
3956};
3957
3958void netdev_bonding_info_change(struct net_device *dev,
3959 struct netdev_bonding_info *bonding_info);
3960
12b0004d
CW
3961static inline
3962struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3963{
3964 return __skb_gso_segment(skb, features, true);
3965}
53d6471c 3966__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3967
3968static inline bool can_checksum_protocol(netdev_features_t features,
3969 __be16 protocol)
3970{
c8cd0989
TH
3971 if (protocol == htons(ETH_P_FCOE))
3972 return !!(features & NETIF_F_FCOE_CRC);
3973
3974 /* Assume this is an IP checksum (not SCTP CRC) */
3975
3976 if (features & NETIF_F_HW_CSUM) {
3977 /* Can checksum everything */
3978 return true;
3979 }
3980
3981 switch (protocol) {
3982 case htons(ETH_P_IP):
3983 return !!(features & NETIF_F_IP_CSUM);
3984 case htons(ETH_P_IPV6):
3985 return !!(features & NETIF_F_IPV6_CSUM);
3986 default:
3987 return false;
3988 }
ec5f0615 3989}
12b0004d 3990
fb286bb2 3991#ifdef CONFIG_BUG
f629d208 3992void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3993#else
3994static inline void netdev_rx_csum_fault(struct net_device *dev)
3995{
3996}
3997#endif
1da177e4 3998/* rx skb timestamps */
f629d208
JP
3999void net_enable_timestamp(void);
4000void net_disable_timestamp(void);
1da177e4 4001
20380731 4002#ifdef CONFIG_PROC_FS
f629d208 4003int __init dev_proc_init(void);
900ff8c6
CW
4004#else
4005#define dev_proc_init() 0
20380731
ACM
4006#endif
4007
4798248e 4008static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4009 struct sk_buff *skb, struct net_device *dev,
4010 bool more)
4798248e 4011{
fa2dbdc2 4012 skb->xmit_more = more ? 1 : 0;
0b725a2c 4013 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4014}
4015
10b3ad8c 4016static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4017 struct netdev_queue *txq, bool more)
4798248e
DM
4018{
4019 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4020 int rc;
4798248e 4021
fa2dbdc2 4022 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4023 if (rc == NETDEV_TX_OK)
4024 txq_trans_update(txq);
4025
4026 return rc;
4798248e
DM
4027}
4028
42a2d923
LT
4029int netdev_class_create_file_ns(struct class_attribute *class_attr,
4030 const void *ns);
4031void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4032 const void *ns);
58292cbe
TH
4033
4034static inline int netdev_class_create_file(struct class_attribute *class_attr)
4035{
4036 return netdev_class_create_file_ns(class_attr, NULL);
4037}
4038
4039static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4040{
4041 netdev_class_remove_file_ns(class_attr, NULL);
4042}
b8a9787e 4043
04600794
JB
4044extern struct kobj_ns_type_operations net_ns_type_operations;
4045
f629d208 4046const char *netdev_drivername(const struct net_device *dev);
6579e57b 4047
f629d208 4048void linkwatch_run_queue(void);
20380731 4049
da08143b
MK
4050static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4051 netdev_features_t f2)
4052{
c8cd0989
TH
4053 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4054 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4055 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4056 else
b6a0e72a 4057 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4058 }
da08143b 4059
c8cd0989 4060 return f1 & f2;
da08143b
MK
4061}
4062
c8f44aff
MM
4063static inline netdev_features_t netdev_get_wanted_features(
4064 struct net_device *dev)
5455c699
MM
4065{
4066 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4067}
c8f44aff
MM
4068netdev_features_t netdev_increment_features(netdev_features_t all,
4069 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4070
4071/* Allow TSO being used on stacked device :
4072 * Performing the GSO segmentation before last device
4073 * is a performance improvement.
4074 */
4075static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4076 netdev_features_t mask)
4077{
4078 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4079}
4080
6cb6a27c 4081int __netdev_update_features(struct net_device *dev);
5455c699 4082void netdev_update_features(struct net_device *dev);
afe12cc8 4083void netdev_change_features(struct net_device *dev);
7f353bf2 4084
fc4a7489
PM
4085void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4086 struct net_device *dev);
4087
e38f3025
TM
4088netdev_features_t passthru_features_check(struct sk_buff *skb,
4089 struct net_device *dev,
4090 netdev_features_t features);
c1e756bf 4091netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4092
4d29515f 4093static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4094{
7b748340 4095 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4096
4097 /* check flags correspondence */
4098 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4099 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4100 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4101 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4102 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4103 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4104 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4105 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4106 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4107 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4108 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4109 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4110 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4111 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4112 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4113 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0345e186 4114
d6b4991a 4115 return (features & feature) == feature;
576a30eb
HX
4116}
4117
4d29515f 4118static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4119{
278b2513 4120 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4121 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4122}
4123
8b86a61d 4124static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4125 netdev_features_t features)
7967168c 4126{
fc741216 4127 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4128 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4129 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4130}
4131
82cc1a7a
PWJ
4132static inline void netif_set_gso_max_size(struct net_device *dev,
4133 unsigned int size)
4134{
4135 dev->gso_max_size = size;
4136}
4137
7a7ffbab
WCC
4138static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4139 int pulled_hlen, u16 mac_offset,
4140 int mac_len)
4141{
4142 skb->protocol = protocol;
4143 skb->encapsulation = 1;
4144 skb_push(skb, pulled_hlen);
4145 skb_reset_transport_header(skb);
4146 skb->mac_header = mac_offset;
4147 skb->network_header = skb->mac_header + mac_len;
4148 skb->mac_len = mac_len;
4149}
4150
3c175784
SD
4151static inline bool netif_is_macsec(const struct net_device *dev)
4152{
4153 return dev->priv_flags & IFF_MACSEC;
4154}
4155
b618aaa9 4156static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4157{
4158 return dev->priv_flags & IFF_MACVLAN;
4159}
4160
b618aaa9 4161static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4162{
4163 return dev->priv_flags & IFF_MACVLAN_PORT;
4164}
4165
b618aaa9 4166static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4167{
4168 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4169}
4170
b618aaa9 4171static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4172{
4173 return dev->priv_flags & IFF_IPVLAN_MASTER;
4174}
4175
b618aaa9 4176static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4177{
4178 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4179}
4180
b618aaa9 4181static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4182{
4183 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4184}
4185
3bdc0eba
BG
4186static inline bool netif_supports_nofcs(struct net_device *dev)
4187{
4188 return dev->priv_flags & IFF_SUPP_NOFCS;
4189}
4190
007979ea 4191static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4192{
007979ea 4193 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4194}
4195
fee6d4c7
DA
4196static inline bool netif_is_l3_slave(const struct net_device *dev)
4197{
4198 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4199}
4200
0894ae3f
JP
4201static inline bool netif_is_bridge_master(const struct net_device *dev)
4202{
4203 return dev->priv_flags & IFF_EBRIDGE;
4204}
4205
28f9ee22
VY
4206static inline bool netif_is_bridge_port(const struct net_device *dev)
4207{
4208 return dev->priv_flags & IFF_BRIDGE_PORT;
4209}
4210
35d4e172
JP
4211static inline bool netif_is_ovs_master(const struct net_device *dev)
4212{
4213 return dev->priv_flags & IFF_OPENVSWITCH;
4214}
4215
5be66141
JP
4216static inline bool netif_is_ovs_port(const struct net_device *dev)
4217{
4218 return dev->priv_flags & IFF_OVS_DATAPATH;
4219}
4220
b618aaa9 4221static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4222{
4223 return dev->priv_flags & IFF_TEAM;
4224}
4225
b618aaa9 4226static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4227{
4228 return dev->priv_flags & IFF_TEAM_PORT;
4229}
4230
b618aaa9 4231static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4232{
4233 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4234}
4235
b618aaa9 4236static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4237{
4238 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4239}
4240
d4ab4286
KJ
4241static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4242{
4243 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4244}
4245
02875878
ED
4246/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4247static inline void netif_keep_dst(struct net_device *dev)
4248{
4249 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4250}
4251
18d3df3e
PA
4252/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4253static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4254{
4255 /* TODO: reserve and use an additional IFF bit, if we get more users */
4256 return dev->priv_flags & IFF_MACSEC;
4257}
4258
505d4f73 4259extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4260
571ba423
JP
4261/* Logging, debugging and troubleshooting/diagnostic helpers. */
4262
4263/* netdev_printk helpers, similar to dev_printk */
4264
4265static inline const char *netdev_name(const struct net_device *dev)
4266{
c6f854d5
VF
4267 if (!dev->name[0] || strchr(dev->name, '%'))
4268 return "(unnamed net_device)";
571ba423
JP
4269 return dev->name;
4270}
4271
8397ed36
DA
4272static inline bool netdev_unregistering(const struct net_device *dev)
4273{
4274 return dev->reg_state == NETREG_UNREGISTERING;
4275}
4276
ccc7f496
VF
4277static inline const char *netdev_reg_state(const struct net_device *dev)
4278{
4279 switch (dev->reg_state) {
4280 case NETREG_UNINITIALIZED: return " (uninitialized)";
4281 case NETREG_REGISTERED: return "";
4282 case NETREG_UNREGISTERING: return " (unregistering)";
4283 case NETREG_UNREGISTERED: return " (unregistered)";
4284 case NETREG_RELEASED: return " (released)";
4285 case NETREG_DUMMY: return " (dummy)";
4286 }
4287
4288 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4289 return " (unknown)";
4290}
4291
f629d208 4292__printf(3, 4)
6ea754eb
JP
4293void netdev_printk(const char *level, const struct net_device *dev,
4294 const char *format, ...);
f629d208 4295__printf(2, 3)
6ea754eb 4296void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4297__printf(2, 3)
6ea754eb 4298void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4299__printf(2, 3)
6ea754eb 4300void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4301__printf(2, 3)
6ea754eb 4302void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4303__printf(2, 3)
6ea754eb 4304void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4305__printf(2, 3)
6ea754eb 4306void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4307__printf(2, 3)
6ea754eb 4308void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4309
8909c9ad
VK
4310#define MODULE_ALIAS_NETDEV(device) \
4311 MODULE_ALIAS("netdev-" device)
4312
b558c96f 4313#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4314#define netdev_dbg(__dev, format, args...) \
4315do { \
ffa10cb4 4316 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4317} while (0)
b558c96f
JC
4318#elif defined(DEBUG)
4319#define netdev_dbg(__dev, format, args...) \
4320 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4321#else
4322#define netdev_dbg(__dev, format, args...) \
4323({ \
4324 if (0) \
4325 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4326})
4327#endif
4328
4329#if defined(VERBOSE_DEBUG)
4330#define netdev_vdbg netdev_dbg
4331#else
4332
4333#define netdev_vdbg(dev, format, args...) \
4334({ \
4335 if (0) \
4336 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4337 0; \
4338})
4339#endif
4340
4341/*
4342 * netdev_WARN() acts like dev_printk(), but with the key difference
4343 * of using a WARN/WARN_ON to get the message out, including the
4344 * file/line information and a backtrace.
4345 */
4346#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4347 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4348 netdev_reg_state(dev), ##args)
571ba423 4349
b3d95c5c
JP
4350/* netif printk helpers, similar to netdev_printk */
4351
4352#define netif_printk(priv, type, level, dev, fmt, args...) \
4353do { \
4354 if (netif_msg_##type(priv)) \
4355 netdev_printk(level, (dev), fmt, ##args); \
4356} while (0)
4357
f45f4321
JP
4358#define netif_level(level, priv, type, dev, fmt, args...) \
4359do { \
4360 if (netif_msg_##type(priv)) \
4361 netdev_##level(dev, fmt, ##args); \
4362} while (0)
4363
b3d95c5c 4364#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4365 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4366#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4367 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4368#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4369 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4370#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4371 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4372#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4373 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4374#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4375 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4376#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4377 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4378
0053ea9c 4379#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4380#define netif_dbg(priv, type, netdev, format, args...) \
4381do { \
4382 if (netif_msg_##type(priv)) \
b5fb0a03 4383 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4384} while (0)
0053ea9c
JP
4385#elif defined(DEBUG)
4386#define netif_dbg(priv, type, dev, format, args...) \
4387 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4388#else
4389#define netif_dbg(priv, type, dev, format, args...) \
4390({ \
4391 if (0) \
4392 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4393 0; \
4394})
4395#endif
4396
f617f276
EC
4397/* if @cond then downgrade to debug, else print at @level */
4398#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4399 do { \
4400 if (cond) \
4401 netif_dbg(priv, type, netdev, fmt, ##args); \
4402 else \
4403 netif_ ## level(priv, type, netdev, fmt, ##args); \
4404 } while (0)
4405
b3d95c5c 4406#if defined(VERBOSE_DEBUG)
bcfcc450 4407#define netif_vdbg netif_dbg
b3d95c5c
JP
4408#else
4409#define netif_vdbg(priv, type, dev, format, args...) \
4410({ \
4411 if (0) \
a4ed89cb 4412 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4413 0; \
4414})
4415#endif
571ba423 4416
900ff8c6
CW
4417/*
4418 * The list of packet types we will receive (as opposed to discard)
4419 * and the routines to invoke.
4420 *
4421 * Why 16. Because with 16 the only overlap we get on a hash of the
4422 * low nibble of the protocol value is RARP/SNAP/X.25.
4423 *
4424 * NOTE: That is no longer true with the addition of VLAN tags. Not
4425 * sure which should go first, but I bet it won't make much
4426 * difference if we are running VLANs. The good news is that
4427 * this protocol won't be in the list unless compiled in, so
4428 * the average user (w/out VLANs) will not be adversely affected.
4429 * --BLG
4430 *
4431 * 0800 IP
4432 * 8100 802.1Q VLAN
4433 * 0001 802.3
4434 * 0002 AX.25
4435 * 0004 802.2
4436 * 8035 RARP
4437 * 0005 SNAP
4438 * 0805 X.25
4439 * 0806 ARP
4440 * 8137 IPX
4441 * 0009 Localtalk
4442 * 86DD IPv6
4443 */
4444#define PTYPE_HASH_SIZE (16)
4445#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4446
385a154c 4447#endif /* _LINUX_NETDEVICE_H */