]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
Merge branch 'mlx4-drivers-version-update'
[mirror_ubuntu-artful-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
7a6b6f51 44#ifdef CONFIG_DCB
2f90b865
AD
45#include <net/dcbnl.h>
46#endif
5bc1421e 47#include <net/netprio_cgroup.h>
a050c33f 48
a59e2ecb 49#include <linux/netdev_features.h>
77162022 50#include <linux/neighbour.h>
607ca46e 51#include <uapi/linux/netdevice.h>
61bd3857 52#include <uapi/linux/if_bonding.h>
e4c6734e 53#include <uapi/linux/pkt_cls.h>
59cc1f61 54#include <linux/hashtable.h>
a59e2ecb 55
115c1d6e 56struct netpoll_info;
313162d0 57struct device;
c1f19b51 58struct phy_device;
c6e970a0
AL
59struct dsa_switch_tree;
60
704232c2
JB
61/* 802.11 specific */
62struct wireless_dev;
98a18b6f
AA
63/* 802.15.4 specific */
64struct wpan_dev;
03c57747 65struct mpls_dev;
7c46a640
AD
66/* UDP Tunnel offloads */
67struct udp_tunnel_info;
a7862b45 68struct bpf_prog;
1da177e4 69
f629d208
JP
70void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
d07d7507 72
9a1654ba
JP
73/* Backlog congestion levels */
74#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75#define NET_RX_DROP 1 /* packet dropped */
76
572a9d7b
PM
77/*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
91 * others are propagated to higher layers.
92 */
93
94/* qdisc ->enqueue() return codes. */
95#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
96#define NET_XMIT_DROP 0x01 /* skb dropped */
97#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 98#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 99
b9df3cb8
GR
100/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 103#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
104#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105
dc1f8bf6 106/* Driver transmit return codes */
9a1654ba 107#define NETDEV_TX_MASK 0xf0
572a9d7b 108
dc1f8bf6 109enum netdev_tx {
572a9d7b 110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
113};
114typedef enum netdev_tx netdev_tx_t;
115
9a1654ba
JP
116/*
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 */
120static inline bool dev_xmit_complete(int rc)
121{
122 /*
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 */
128 if (likely(rc < NET_XMIT_MASK))
129 return true;
130
131 return false;
132}
133
1da177e4 134/*
5e82b4b2 135 * Compute the worst-case header length according to the protocols
1da177e4
LT
136 * used.
137 */
fe2918b0 138
c0eb4540
KS
139#if defined(CONFIG_HYPERV_NET)
140# define LL_MAX_HEADER 128
141#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
142# if defined(CONFIG_MAC80211_MESH)
143# define LL_MAX_HEADER 128
144# else
145# define LL_MAX_HEADER 96
146# endif
1da177e4 147#else
8388e3da 148# define LL_MAX_HEADER 32
1da177e4
LT
149#endif
150
d11ead75
BH
151#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
153#define MAX_HEADER LL_MAX_HEADER
154#else
155#define MAX_HEADER (LL_MAX_HEADER + 48)
156#endif
157
158/*
be1f3c2c
BH
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
1da177e4 161 */
fe2918b0 162
d94d9fee 163struct net_device_stats {
3cfde79c
BH
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
1da177e4 173 unsigned long collisions;
1da177e4 174 unsigned long rx_length_errors;
3cfde79c
BH
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
1da177e4
LT
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
1da177e4
LT
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187};
188
1da177e4
LT
189
190#include <linux/cache.h>
191#include <linux/skbuff.h>
192
adc9300e 193#ifdef CONFIG_RPS
c5905afb
IM
194#include <linux/static_key.h>
195extern struct static_key rps_needed;
13bfff25 196extern struct static_key rfs_needed;
adc9300e
ED
197#endif
198
1da177e4
LT
199struct neighbour;
200struct neigh_parms;
201struct sk_buff;
202
f001fde5
JP
203struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
206 unsigned char type;
ccffad25
JP
207#define NETDEV_HW_ADDR_T_LAN 1
208#define NETDEV_HW_ADDR_T_SAN 2
209#define NETDEV_HW_ADDR_T_SLAVE 3
210#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 211#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 212 bool global_use;
4cd729b0 213 int sync_cnt;
8f8f103d 214 int refcount;
4543fbef 215 int synced;
f001fde5
JP
216 struct rcu_head rcu_head;
217};
218
31278e71
JP
219struct netdev_hw_addr_list {
220 struct list_head list;
221 int count;
222};
223
22bedad3
JP
224#define netdev_hw_addr_list_count(l) ((l)->count)
225#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226#define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 228
22bedad3
JP
229#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231#define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 233
22bedad3
JP
234#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 236#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 238
d94d9fee 239struct hh_cache {
5b3dc2f3 240 unsigned int hh_len;
3644f0ce 241 seqlock_t hh_lock;
1da177e4
LT
242
243 /* cached hardware header; allow for machine alignment needs. */
244#define HH_DATA_MOD 16
245#define HH_DATA_OFF(__len) \
5ba0eac6 246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
247#define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250};
251
5e82b4b2 252/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260#define LL_RESERVED_SPACE(dev) \
f5184d26 261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 262#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 264
3b04ddde
SH
265struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
95c96174 268 const void *saddr, unsigned int len);
3b04ddde 269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
2793a23a 274 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
275};
276
1da177e4 277/* These flag bits are private to the generic network queueing
5e82b4b2 278 * layer; they may not be explicitly referenced by any other
1da177e4
LT
279 * code.
280 */
281
d94d9fee 282enum netdev_state_t {
1da177e4
LT
283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
1da177e4 285 __LINK_STATE_NOCARRIER,
b00055aa
SR
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
1da177e4
LT
288};
289
290
291/*
5e82b4b2 292 * This structure holds boot-time configured netdevice settings. They
fe2918b0 293 * are then used in the device probing.
1da177e4
LT
294 */
295struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298};
299#define NETDEV_BOOT_SETUP_MAX 8
300
f629d208 301int __init netdev_boot_setup(char *str);
1da177e4 302
bea3348e
SH
303/*
304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 310 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
404f7c9e 317 unsigned int gro_count;
bea3348e
SH
318 int (*poll)(struct napi_struct *, int);
319#ifdef CONFIG_NETPOLL
bea3348e 320 int poll_owner;
bea3348e 321#endif
5d38a079 322 struct net_device *dev;
d565b0a1 323 struct sk_buff *gro_list;
5d38a079 324 struct sk_buff *skb;
3b47d303 325 struct hrtimer timer;
404f7c9e 326 struct list_head dev_list;
af12fa6e
ET
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
bea3348e
SH
329};
330
d94d9fee 331enum {
bea3348e 332 NAPI_STATE_SCHED, /* Poll is scheduled */
39e6c820 333 NAPI_STATE_MISSED, /* reschedule a napi */
a0a46196 334 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 335 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
336 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
337 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
338 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
339};
340
341enum {
39e6c820
ED
342 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
343 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
344 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
345 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
346 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
347 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
348 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
349};
350
5b252f0c 351enum gro_result {
d1c76af9
HX
352 GRO_MERGED,
353 GRO_MERGED_FREE,
354 GRO_HELD,
355 GRO_NORMAL,
356 GRO_DROP,
25393d3f 357 GRO_CONSUMED,
d1c76af9 358};
5b252f0c 359typedef enum gro_result gro_result_t;
d1c76af9 360
8a4eb573
JP
361/*
362 * enum rx_handler_result - Possible return values for rx_handlers.
363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
364 * further.
365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366 * case skb->dev was changed by rx_handler.
367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
369 *
370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
371 * special processing of the skb, prior to delivery to protocol handlers.
372 *
373 * Currently, a net_device can only have a single rx_handler registered. Trying
374 * to register a second rx_handler will return -EBUSY.
375 *
376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377 * To unregister a rx_handler on a net_device, use
378 * netdev_rx_handler_unregister().
379 *
380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
381 * do with the skb.
382 *
5e82b4b2 383 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 385 * the skb to be delivered in some other way.
8a4eb573
JP
386 *
387 * If the rx_handler changed skb->dev, to divert the skb to another
388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389 * new device will be called if it exists.
390 *
5e82b4b2 391 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 393 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 394 *
5e82b4b2 395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
396 * delivered, it should return RX_HANDLER_PASS.
397 *
398 * A device without a registered rx_handler will behave as if rx_handler
399 * returned RX_HANDLER_PASS.
400 */
401
402enum rx_handler_result {
403 RX_HANDLER_CONSUMED,
404 RX_HANDLER_ANOTHER,
405 RX_HANDLER_EXACT,
406 RX_HANDLER_PASS,
407};
408typedef enum rx_handler_result rx_handler_result_t;
409typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 410
f629d208 411void __napi_schedule(struct napi_struct *n);
bc9ad166 412void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 413
4d29515f 414static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
415{
416 return test_bit(NAPI_STATE_DISABLE, &n->state);
417}
418
39e6c820 419bool napi_schedule_prep(struct napi_struct *n);
bea3348e
SH
420
421/**
422 * napi_schedule - schedule NAPI poll
5e82b4b2 423 * @n: NAPI context
bea3348e
SH
424 *
425 * Schedule NAPI poll routine to be called if it is not already
426 * running.
427 */
428static inline void napi_schedule(struct napi_struct *n)
429{
430 if (napi_schedule_prep(n))
431 __napi_schedule(n);
432}
433
bc9ad166
ED
434/**
435 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 436 * @n: NAPI context
bc9ad166
ED
437 *
438 * Variant of napi_schedule(), assuming hard irqs are masked.
439 */
440static inline void napi_schedule_irqoff(struct napi_struct *n)
441{
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
444}
445
bfe13f54 446/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 447static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
448{
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
4d29515f 451 return true;
bfe13f54 452 }
4d29515f 453 return false;
bfe13f54
RD
454}
455
364b6055 456bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
457/**
458 * napi_complete - NAPI processing complete
5e82b4b2 459 * @n: NAPI context
bea3348e
SH
460 *
461 * Mark NAPI processing as complete.
3b47d303 462 * Consider using napi_complete_done() instead.
364b6055 463 * Return false if device should avoid rearming interrupts.
bea3348e 464 */
364b6055 465static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
466{
467 return napi_complete_done(n, 0);
468}
bea3348e 469
af12fa6e
ET
470/**
471 * napi_hash_del - remove a NAPI from global table
5e82b4b2 472 * @napi: NAPI context
af12fa6e 473 *
5e82b4b2 474 * Warning: caller must observe RCU grace period
34cbe27e
ED
475 * before freeing memory containing @napi, if
476 * this function returns true.
93d05d4a 477 * Note: core networking stack automatically calls it
5e82b4b2 478 * from netif_napi_del().
93d05d4a 479 * Drivers might want to call this helper to combine all
5e82b4b2 480 * the needed RCU grace periods into a single one.
af12fa6e 481 */
34cbe27e 482bool napi_hash_del(struct napi_struct *napi);
af12fa6e 483
bea3348e
SH
484/**
485 * napi_disable - prevent NAPI from scheduling
5e82b4b2 486 * @n: NAPI context
bea3348e
SH
487 *
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
490 */
3b47d303 491void napi_disable(struct napi_struct *n);
bea3348e
SH
492
493/**
494 * napi_enable - enable NAPI scheduling
5e82b4b2 495 * @n: NAPI context
bea3348e
SH
496 *
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
499 */
500static inline void napi_enable(struct napi_struct *n)
501{
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 503 smp_mb__before_atomic();
bea3348e 504 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 505 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
506}
507
c264c3de
SH
508/**
509 * napi_synchronize - wait until NAPI is not running
5e82b4b2 510 * @n: NAPI context
c264c3de
SH
511 *
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
515 */
516static inline void napi_synchronize(const struct napi_struct *n)
517{
facc432f
AB
518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
520 msleep(1);
521 else
522 barrier();
c264c3de 523}
c264c3de 524
d94d9fee 525enum netdev_queue_state_t {
73466498
TH
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
c3f26a26 528 __QUEUE_STATE_FROZEN,
79d16385 529};
8e2f1a63
DB
530
531#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
534
535#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
537 QUEUE_STATE_FROZEN)
538#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
539 QUEUE_STATE_FROZEN)
540
73466498
TH
541/*
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
549 */
79d16385 550
bb949fbd 551struct netdev_queue {
6a321cb3 552/*
5e82b4b2 553 * read-mostly part
6a321cb3 554 */
bb949fbd 555 struct net_device *dev;
46e5da40 556 struct Qdisc __rcu *qdisc;
b0e1e646 557 struct Qdisc *qdisc_sleeping;
ccf5ff69 558#ifdef CONFIG_SYSFS
1d24eb48
TH
559 struct kobject kobj;
560#endif
f2cd2d3e
ED
561#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 int numa_node;
563#endif
c0ef079c
FW
564 unsigned long tx_maxrate;
565 /*
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
568 */
569 unsigned long trans_timeout;
6a321cb3 570/*
5e82b4b2 571 * write-mostly part
6a321cb3
ED
572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
9d21493b 575 /*
9b36627a 576 * Time (in jiffies) of last Tx
9d21493b
ED
577 */
578 unsigned long trans_start;
ccf5ff69 579
114cf580
TH
580 unsigned long state;
581
582#ifdef CONFIG_BQL
583 struct dql dql;
584#endif
e8a0464c 585} ____cacheline_aligned_in_smp;
bb949fbd 586
f2cd2d3e
ED
587static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
588{
589#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 return q->numa_node;
591#else
b236da69 592 return NUMA_NO_NODE;
f2cd2d3e
ED
593#endif
594}
595
596static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
597{
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 q->numa_node = node;
600#endif
601}
602
df334545 603#ifdef CONFIG_RPS
0a9627f2
TH
604/*
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
607 */
608struct rps_map {
609 unsigned int len;
610 struct rcu_head rcu;
611 u16 cpus[0];
612};
60b778ce 613#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 614
fec5e652 615/*
c445477d
BH
616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
fec5e652
TH
619 */
620struct rps_dev_flow {
621 u16 cpu;
c445477d 622 u16 filter;
fec5e652
TH
623 unsigned int last_qtail;
624};
c445477d 625#define RPS_NO_FILTER 0xffff
fec5e652
TH
626
627/*
628 * The rps_dev_flow_table structure contains a table of flow mappings.
629 */
630struct rps_dev_flow_table {
631 unsigned int mask;
632 struct rcu_head rcu;
fec5e652
TH
633 struct rps_dev_flow flows[0];
634};
635#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 636 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
637
638/*
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
567e4b79 643 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 646 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
647 */
648struct rps_sock_flow_table {
567e4b79 649 u32 mask;
93c1af6c
ED
650
651 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 652};
567e4b79 653#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
654
655#define RPS_NO_CPU 0xffff
656
567e4b79
ED
657extern u32 rps_cpu_mask;
658extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
659
fec5e652
TH
660static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
661 u32 hash)
662{
663 if (table && hash) {
567e4b79
ED
664 unsigned int index = hash & table->mask;
665 u32 val = hash & ~rps_cpu_mask;
fec5e652 666
5e82b4b2 667 /* We only give a hint, preemption can change CPU under us */
567e4b79 668 val |= raw_smp_processor_id();
fec5e652 669
567e4b79
ED
670 if (table->ents[index] != val)
671 table->ents[index] = val;
fec5e652
TH
672 }
673}
674
c445477d 675#ifdef CONFIG_RFS_ACCEL
f629d208
JP
676bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
677 u16 filter_id);
c445477d 678#endif
a953be53 679#endif /* CONFIG_RPS */
c445477d 680
0a9627f2
TH
681/* This structure contains an instance of an RX queue. */
682struct netdev_rx_queue {
a953be53 683#ifdef CONFIG_RPS
6e3f7faf
ED
684 struct rps_map __rcu *rps_map;
685 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 686#endif
6e3f7faf 687 struct kobject kobj;
fe822240 688 struct net_device *dev;
0a9627f2 689} ____cacheline_aligned_in_smp;
a953be53
MD
690
691/*
692 * RX queue sysfs structures and functions.
693 */
694struct rx_queue_attribute {
695 struct attribute attr;
696 ssize_t (*show)(struct netdev_rx_queue *queue,
697 struct rx_queue_attribute *attr, char *buf);
698 ssize_t (*store)(struct netdev_rx_queue *queue,
699 struct rx_queue_attribute *attr, const char *buf, size_t len);
700};
d314774c 701
bf264145
TH
702#ifdef CONFIG_XPS
703/*
704 * This structure holds an XPS map which can be of variable length. The
705 * map is an array of queues.
706 */
707struct xps_map {
708 unsigned int len;
709 unsigned int alloc_len;
710 struct rcu_head rcu;
711 u16 queues[0];
712};
60b778ce 713#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
714#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
715 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
716
717/*
718 * This structure holds all XPS maps for device. Maps are indexed by CPU.
719 */
720struct xps_dev_maps {
721 struct rcu_head rcu;
a4177869 722 struct xps_map __rcu *cpu_map[0];
bf264145 723};
184c449f
AD
724#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
725 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
726#endif /* CONFIG_XPS */
727
4f57c087
JF
728#define TC_MAX_QUEUE 16
729#define TC_BITMASK 15
730/* HW offloaded queuing disciplines txq count and offset maps */
731struct netdev_tc_txq {
732 u16 count;
733 u16 offset;
734};
735
68bad94e
NP
736#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
737/*
738 * This structure is to hold information about the device
739 * configured to run FCoE protocol stack.
740 */
741struct netdev_fcoe_hbainfo {
742 char manufacturer[64];
743 char serial_number[64];
744 char hardware_version[64];
745 char driver_version[64];
746 char optionrom_version[64];
747 char firmware_version[64];
748 char model[256];
749 char model_description[256];
750};
751#endif
752
02637fce 753#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 754
02637fce
JP
755/* This structure holds a unique identifier to identify some
756 * physical item (port for example) used by a netdevice.
66b52b0d 757 */
02637fce
JP
758struct netdev_phys_item_id {
759 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
760 unsigned char id_len;
761};
762
d754f98b
SF
763static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
764 struct netdev_phys_item_id *b)
765{
766 return a->id_len == b->id_len &&
767 memcmp(a->id, b->id, a->id_len) == 0;
768}
769
99932d4f
DB
770typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
771 struct sk_buff *skb);
772
a1b7c5fd 773/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
774 * that are being passed to the netdevice through the setup_tc op.
775 */
776enum {
777 TC_SETUP_MQPRIO,
a1b7c5fd 778 TC_SETUP_CLSU32,
5b33f488 779 TC_SETUP_CLSFLOWER,
b87f7936 780 TC_SETUP_MATCHALL,
332ae8e2 781 TC_SETUP_CLSBPF,
16e5cc64
JF
782};
783
a1b7c5fd
JF
784struct tc_cls_u32_offload;
785
16e5cc64
JF
786struct tc_to_netdev {
787 unsigned int type;
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 794 };
7091d8c7 795 bool egress_dev;
16e5cc64
JF
796};
797
a7862b45
BB
798/* These structures hold the attributes of xdp state that are being passed
799 * to the netdevice through the xdp op.
800 */
801enum xdp_netdev_command {
802 /* Set or clear a bpf program used in the earliest stages of packet
803 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
804 * is responsible for calling bpf_prog_put on any old progs that are
805 * stored. In case of error, the callee need not release the new prog
806 * reference, but on success it takes ownership and must bpf_prog_put
807 * when it is no longer used.
808 */
809 XDP_SETUP_PROG,
810 /* Check if a bpf program is set on the device. The callee should
811 * return true if a program is currently attached and running.
812 */
813 XDP_QUERY_PROG,
814};
815
ddf9f970
JK
816struct netlink_ext_ack;
817
a7862b45
BB
818struct netdev_xdp {
819 enum xdp_netdev_command command;
820 union {
821 /* XDP_SETUP_PROG */
ddf9f970
JK
822 struct {
823 struct bpf_prog *prog;
824 struct netlink_ext_ack *extack;
825 };
a7862b45
BB
826 /* XDP_QUERY_PROG */
827 bool prog_attached;
828 };
829};
16e5cc64 830
d77e38e6
SK
831#ifdef CONFIG_XFRM_OFFLOAD
832struct xfrmdev_ops {
833 int (*xdo_dev_state_add) (struct xfrm_state *x);
834 void (*xdo_dev_state_delete) (struct xfrm_state *x);
835 void (*xdo_dev_state_free) (struct xfrm_state *x);
836 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
837 struct xfrm_state *x);
838};
839#endif
840
d314774c
SH
841/*
842 * This structure defines the management hooks for network devices.
00829823
SH
843 * The following hooks can be defined; unless noted otherwise, they are
844 * optional and can be filled with a null pointer.
d314774c
SH
845 *
846 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
847 * This function is called once when a network device is registered.
848 * The network device can use this for any late stage initialization
849 * or semantic validation. It can fail with an error code which will
850 * be propagated back to register_netdev.
d314774c
SH
851 *
852 * void (*ndo_uninit)(struct net_device *dev);
853 * This function is called when device is unregistered or when registration
854 * fails. It is not called if init fails.
855 *
856 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 857 * This function is called when a network device transitions to the up
d314774c
SH
858 * state.
859 *
860 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 861 * This function is called when a network device transitions to the down
d314774c
SH
862 * state.
863 *
dc1f8bf6
SH
864 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
865 * struct net_device *dev);
00829823 866 * Called when a packet needs to be transmitted.
e79d8429
RR
867 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
868 * the queue before that can happen; it's for obsolete devices and weird
869 * corner cases, but the stack really does a non-trivial amount
870 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 871 * Required; cannot be NULL.
00829823 872 *
1a2a1444
DM
873 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
874 * struct net_device *dev
875 * netdev_features_t features);
876 * Called by core transmit path to determine if device is capable of
877 * performing offload operations on a given packet. This is to give
878 * the device an opportunity to implement any restrictions that cannot
879 * be otherwise expressed by feature flags. The check is called with
880 * the set of features that the stack has calculated and it returns
881 * those the driver believes to be appropriate.
cdba756f 882 *
f663dd9a 883 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 884 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 885 * Called to decide which queue to use when device supports multiple
00829823
SH
886 * transmit queues.
887 *
d314774c
SH
888 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
889 * This function is called to allow device receiver to make
5e82b4b2 890 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
891 *
892 * void (*ndo_set_rx_mode)(struct net_device *dev);
893 * This function is called device changes address list filtering.
01789349 894 * If driver handles unicast address filtering, it should set
5e82b4b2 895 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
896 *
897 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
898 * This function is called when the Media Access Control address
37b607c5 899 * needs to be changed. If this interface is not defined, the
5e82b4b2 900 * MAC address can not be changed.
d314774c
SH
901 *
902 * int (*ndo_validate_addr)(struct net_device *dev);
903 * Test if Media Access Control address is valid for the device.
904 *
905 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
906 * Called when a user requests an ioctl which can't be handled by
907 * the generic interface code. If not defined ioctls return
d314774c
SH
908 * not supported error code.
909 *
910 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
911 * Used to set network devices bus interface parameters. This interface
5e82b4b2 912 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
913 * interface (PCI) for low level management.
914 *
915 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
916 * Called when a user wants to change the Maximum Transfer Unit
917 * of a device. If not defined, any request to change MTU will
918 * will return an error.
919 *
00829823 920 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 921 * Callback used when the transmitter has not made any progress
d314774c
SH
922 * for dev->watchdog ticks.
923 *
bc1f4470 924 * void (*ndo_get_stats64)(struct net_device *dev,
925 * struct rtnl_link_stats64 *storage);
d308e38f 926 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 927 * Called when a user wants to get the network device usage
be1f3c2c 928 * statistics. Drivers must do one of the following:
3cfde79c
BH
929 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
930 * rtnl_link_stats64 structure passed by the caller.
82695d9b 931 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
932 * (which should normally be dev->stats) and return a pointer to
933 * it. The structure may be changed asynchronously only if each
934 * field is written atomically.
935 * 3. Update dev->stats asynchronously and atomically, and define
936 * neither operation.
d314774c 937 *
3df5b3c6 938 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
939 * Return true if this device supports offload stats of this attr_id.
940 *
941 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
942 * void *attr_data)
943 * Get statistics for offload operations by attr_id. Write it into the
944 * attr_data pointer.
945 *
5d632cb7 946 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 947 * If device supports VLAN filtering this function is called when a
80d5c368 948 * VLAN id is registered.
d314774c 949 *
5d632cb7 950 * int (*ndo_vlan_rx_kill_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 unregistered.
d314774c
SH
953 *
954 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
955 *
956 * SR-IOV management functions.
957 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
958 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
959 * u8 qos, __be16 proto);
ed616689
SC
960 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
961 * int max_tx_rate);
5f8444a3 962 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 963 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
964 * int (*ndo_get_vf_config)(struct net_device *dev,
965 * int vf, struct ifla_vf_info *ivf);
1d8faf48 966 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
967 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
968 * struct nlattr *port[]);
01a3d796
VZ
969 *
970 * Enable or disable the VF ability to query its RSS Redirection Table and
971 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 972 * with PF and querying it may introduce a theoretical security risk.
01a3d796 973 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 974 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
6a4bc2b4
FF
975 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
976 * __be16 protocol, struct tc_to_netdev *tc);
977 * Called to setup any 'tc' scheduler, classifier or action on @dev.
978 * This is always called from the stack with the rtnl lock held and netif
979 * tx queues stopped. This allows the netdevice to perform queue
980 * management safely.
c445477d 981 *
e9bce845
YZ
982 * Fiber Channel over Ethernet (FCoE) offload functions.
983 * int (*ndo_fcoe_enable)(struct net_device *dev);
984 * Called when the FCoE protocol stack wants to start using LLD for FCoE
985 * so the underlying device can perform whatever needed configuration or
986 * initialization to support acceleration of FCoE traffic.
987 *
988 * int (*ndo_fcoe_disable)(struct net_device *dev);
989 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
990 * so the underlying device can perform whatever needed clean-ups to
991 * stop supporting acceleration of FCoE traffic.
992 *
993 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
994 * struct scatterlist *sgl, unsigned int sgc);
995 * Called when the FCoE Initiator wants to initialize an I/O that
996 * is a possible candidate for Direct Data Placement (DDP). The LLD can
997 * perform necessary setup and returns 1 to indicate the device is set up
998 * successfully to perform DDP on this I/O, otherwise this returns 0.
999 *
1000 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1001 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1002 * indicated by the FC exchange id 'xid', so the underlying device can
1003 * clean up and reuse resources for later DDP requests.
1004 *
1005 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1006 * struct scatterlist *sgl, unsigned int sgc);
1007 * Called when the FCoE Target wants to initialize an I/O that
1008 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1009 * perform necessary setup and returns 1 to indicate the device is set up
1010 * successfully to perform DDP on this I/O, otherwise this returns 0.
1011 *
68bad94e
NP
1012 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1013 * struct netdev_fcoe_hbainfo *hbainfo);
1014 * Called when the FCoE Protocol stack wants information on the underlying
1015 * device. This information is utilized by the FCoE protocol stack to
1016 * register attributes with Fiber Channel management service as per the
1017 * FC-GS Fabric Device Management Information(FDMI) specification.
1018 *
e9bce845
YZ
1019 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1020 * Called when the underlying device wants to override default World Wide
1021 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1022 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1023 * protocol stack to use.
1024 *
c445477d
BH
1025 * RFS acceleration.
1026 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1027 * u16 rxq_index, u32 flow_id);
1028 * Set hardware filter for RFS. rxq_index is the target queue index;
1029 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1030 * Return the filter ID on success, or a negative error code.
fbaec0ea 1031 *
8b98a70c 1032 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1033 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1034 * Called to make another netdev an underling.
1035 *
1036 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1037 * Called to release previously enslaved netdev.
5455c699
MM
1038 *
1039 * Feature/offload setting functions.
1a2a1444
DM
1040 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1041 * netdev_features_t features);
1042 * Adjusts the requested feature flags according to device-specific
1043 * constraints, and returns the resulting flags. Must not modify
1044 * the device state.
1045 *
c8f44aff 1046 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1047 * Called to update device configuration to new features. Passed
1048 * feature set might be less than what was returned by ndo_fix_features()).
1049 * Must return >0 or -errno if it changed dev->features itself.
1050 *
edc7d573 1051 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1052 * struct net_device *dev,
f6f6424b 1053 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1054 * Adds an FDB entry to dev for addr.
1690be63
VY
1055 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1056 * struct net_device *dev,
f6f6424b 1057 * const unsigned char *addr, u16 vid)
77162022
JF
1058 * Deletes the FDB entry from dev coresponding to addr.
1059 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1060 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1061 * int *idx)
77162022
JF
1062 * Used to add FDB entries to dump requests. Implementers should add
1063 * entries to skb and update idx with the number of entries.
e5a55a89 1064 *
ad41faa8
ND
1065 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1066 * u16 flags)
e5a55a89 1067 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1068 * struct net_device *dev, u32 filter_mask,
1069 * int nlflags)
ad41faa8
ND
1070 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1071 * u16 flags);
4bf84c35
JP
1072 *
1073 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1074 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1075 * which do not represent real hardware may define this to allow their
1076 * userspace components to manage their virtual carrier state. Devices
1077 * that determine carrier state from physical hardware properties (eg
1078 * network cables) or protocol-dependent mechanisms (eg
1079 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1080 *
1081 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1082 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1083 * Called to get ID of physical port of this device. If driver does
1084 * not implement this, it is assumed that the hw is not able to have
1085 * multiple net devices on single physical port.
53cf5275 1086 *
7c46a640
AD
1087 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1088 * struct udp_tunnel_info *ti);
1089 * Called by UDP tunnel to notify a driver about the UDP port and socket
1090 * address family that a UDP tunnel is listnening to. It is called only
1091 * when a new port starts listening. The operation is protected by the
1092 * RTNL.
1093 *
1094 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1095 * struct udp_tunnel_info *ti);
1096 * Called by UDP tunnel to notify the driver about a UDP port and socket
1097 * address family that the UDP tunnel is not listening to anymore. The
1098 * operation is protected by the RTNL.
1099 *
a6cc0cfa
JF
1100 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1101 * struct net_device *dev)
1102 * Called by upper layer devices to accelerate switching or other
1103 * station functionality into hardware. 'pdev is the lowerdev
1104 * to use for the offload and 'dev' is the net device that will
1105 * back the offload. Returns a pointer to the private structure
1106 * the upper layer will maintain.
1107 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1108 * Called by upper layer device to delete the station created
1109 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1110 * the station and priv is the structure returned by the add
1111 * operation.
1112 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1113 * struct net_device *dev,
1114 * void *priv);
1115 * Callback to use for xmit over the accelerated station. This
1116 * is used in place of ndo_start_xmit on accelerated net
1117 * devices.
822b3b2e
JF
1118 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1119 * int queue_index, u32 maxrate);
1120 * Called when a user wants to set a max-rate limitation of specific
1121 * TX queue.
a54acb3a
ND
1122 * int (*ndo_get_iflink)(const struct net_device *dev);
1123 * Called to get the iflink value of this device.
d746d707 1124 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1125 * bool proto_down);
d746d707
AK
1126 * This function is used to pass protocol port error state information
1127 * to the switch driver. The switch driver can react to the proto_down
1128 * by doing a phys down on the associated switch port.
fc4099f1
PS
1129 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1130 * This function is used to get egress tunnel information for given skb.
1131 * This is useful for retrieving outer tunnel header parameters while
1132 * sampling packet.
871b642a
PA
1133 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1134 * This function is used to specify the headroom that the skb must
1135 * consider when allocation skb during packet reception. Setting
1136 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1137 * forward. Setting a negative value resets the rx headroom to the
871b642a 1138 * default value.
a7862b45
BB
1139 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1140 * This function is used to set or query state related to XDP on the
1141 * netdevice. See definition of enum xdp_netdev_command for details.
d746d707 1142 *
d314774c
SH
1143 */
1144struct net_device_ops {
1145 int (*ndo_init)(struct net_device *dev);
1146 void (*ndo_uninit)(struct net_device *dev);
1147 int (*ndo_open)(struct net_device *dev);
1148 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1149 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1150 struct net_device *dev);
1151 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1152 struct net_device *dev,
1153 netdev_features_t features);
00829823 1154 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1155 struct sk_buff *skb,
99932d4f
DB
1156 void *accel_priv,
1157 select_queue_fallback_t fallback);
d314774c
SH
1158 void (*ndo_change_rx_flags)(struct net_device *dev,
1159 int flags);
d314774c 1160 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1161 int (*ndo_set_mac_address)(struct net_device *dev,
1162 void *addr);
d314774c 1163 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1164 int (*ndo_do_ioctl)(struct net_device *dev,
1165 struct ifreq *ifr, int cmd);
d314774c
SH
1166 int (*ndo_set_config)(struct net_device *dev,
1167 struct ifmap *map);
00829823
SH
1168 int (*ndo_change_mtu)(struct net_device *dev,
1169 int new_mtu);
1170 int (*ndo_neigh_setup)(struct net_device *dev,
1171 struct neigh_parms *);
d314774c
SH
1172 void (*ndo_tx_timeout) (struct net_device *dev);
1173
bc1f4470 1174 void (*ndo_get_stats64)(struct net_device *dev,
1175 struct rtnl_link_stats64 *storage);
3df5b3c6 1176 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1177 int (*ndo_get_offload_stats)(int attr_id,
1178 const struct net_device *dev,
1179 void *attr_data);
d314774c
SH
1180 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1181
8e586137 1182 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1183 __be16 proto, u16 vid);
8e586137 1184 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1185 __be16 proto, u16 vid);
d314774c 1186#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1187 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1188 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1189 struct netpoll_info *info);
0e34e931 1190 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1191#endif
95c26df8
WM
1192 int (*ndo_set_vf_mac)(struct net_device *dev,
1193 int queue, u8 *mac);
1194 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1195 int queue, u16 vlan,
1196 u8 qos, __be16 proto);
ed616689
SC
1197 int (*ndo_set_vf_rate)(struct net_device *dev,
1198 int vf, int min_tx_rate,
1199 int max_tx_rate);
5f8444a3
GR
1200 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1201 int vf, bool setting);
dd461d6a
HS
1202 int (*ndo_set_vf_trust)(struct net_device *dev,
1203 int vf, bool setting);
95c26df8
WM
1204 int (*ndo_get_vf_config)(struct net_device *dev,
1205 int vf,
1206 struct ifla_vf_info *ivf);
1d8faf48
RE
1207 int (*ndo_set_vf_link_state)(struct net_device *dev,
1208 int vf, int link_state);
3b766cd8
EBE
1209 int (*ndo_get_vf_stats)(struct net_device *dev,
1210 int vf,
1211 struct ifla_vf_stats
1212 *vf_stats);
57b61080
SF
1213 int (*ndo_set_vf_port)(struct net_device *dev,
1214 int vf,
1215 struct nlattr *port[]);
1216 int (*ndo_get_vf_port)(struct net_device *dev,
1217 int vf, struct sk_buff *skb);
cc8e27cc
EC
1218 int (*ndo_set_vf_guid)(struct net_device *dev,
1219 int vf, u64 guid,
1220 int guid_type);
01a3d796
VZ
1221 int (*ndo_set_vf_rss_query_en)(
1222 struct net_device *dev,
1223 int vf, bool setting);
16e5cc64
JF
1224 int (*ndo_setup_tc)(struct net_device *dev,
1225 u32 handle,
1226 __be16 protocol,
1227 struct tc_to_netdev *tc);
d11ead75 1228#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1229 int (*ndo_fcoe_enable)(struct net_device *dev);
1230 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1231 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1232 u16 xid,
1233 struct scatterlist *sgl,
1234 unsigned int sgc);
1235 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1236 u16 xid);
6247e086
YZ
1237 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1238 u16 xid,
1239 struct scatterlist *sgl,
1240 unsigned int sgc);
68bad94e
NP
1241 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1242 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1243#endif
1244
d11ead75 1245#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1246#define NETDEV_FCOE_WWNN 0
1247#define NETDEV_FCOE_WWPN 1
1248 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1249 u64 *wwn, int type);
4d288d57 1250#endif
3c9c36bc 1251
c445477d
BH
1252#ifdef CONFIG_RFS_ACCEL
1253 int (*ndo_rx_flow_steer)(struct net_device *dev,
1254 const struct sk_buff *skb,
1255 u16 rxq_index,
1256 u32 flow_id);
1257#endif
fbaec0ea
JP
1258 int (*ndo_add_slave)(struct net_device *dev,
1259 struct net_device *slave_dev);
1260 int (*ndo_del_slave)(struct net_device *dev,
1261 struct net_device *slave_dev);
c8f44aff
MM
1262 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1263 netdev_features_t features);
5455c699 1264 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1265 netdev_features_t features);
503eebc2
JP
1266 int (*ndo_neigh_construct)(struct net_device *dev,
1267 struct neighbour *n);
1268 void (*ndo_neigh_destroy)(struct net_device *dev,
1269 struct neighbour *n);
77162022
JF
1270
1271 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1272 struct nlattr *tb[],
77162022 1273 struct net_device *dev,
6b6e2725 1274 const unsigned char *addr,
f6f6424b 1275 u16 vid,
77162022
JF
1276 u16 flags);
1277 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1278 struct nlattr *tb[],
77162022 1279 struct net_device *dev,
f6f6424b
JP
1280 const unsigned char *addr,
1281 u16 vid);
77162022
JF
1282 int (*ndo_fdb_dump)(struct sk_buff *skb,
1283 struct netlink_callback *cb,
1284 struct net_device *dev,
5d5eacb3 1285 struct net_device *filter_dev,
d297653d 1286 int *idx);
e5a55a89
JF
1287
1288 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1289 struct nlmsghdr *nlh,
1290 u16 flags);
e5a55a89
JF
1291 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1292 u32 pid, u32 seq,
6cbdceeb 1293 struct net_device *dev,
46c264da
ND
1294 u32 filter_mask,
1295 int nlflags);
407af329 1296 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1297 struct nlmsghdr *nlh,
1298 u16 flags);
4bf84c35
JP
1299 int (*ndo_change_carrier)(struct net_device *dev,
1300 bool new_carrier);
66b52b0d 1301 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1302 struct netdev_phys_item_id *ppid);
db24a904
DA
1303 int (*ndo_get_phys_port_name)(struct net_device *dev,
1304 char *name, size_t len);
7c46a640
AD
1305 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1306 struct udp_tunnel_info *ti);
1307 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1308 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1309 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1310 struct net_device *dev);
1311 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1312 void *priv);
1313
1314 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1315 struct net_device *dev,
1316 void *priv);
25175ba5 1317 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1318 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1319 int queue_index,
1320 u32 maxrate);
a54acb3a 1321 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1322 int (*ndo_change_proto_down)(struct net_device *dev,
1323 bool proto_down);
fc4099f1
PS
1324 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1325 struct sk_buff *skb);
871b642a
PA
1326 void (*ndo_set_rx_headroom)(struct net_device *dev,
1327 int needed_headroom);
a7862b45
BB
1328 int (*ndo_xdp)(struct net_device *dev,
1329 struct netdev_xdp *xdp);
d314774c
SH
1330};
1331
7aa98047
LR
1332/**
1333 * enum net_device_priv_flags - &struct net_device priv_flags
1334 *
1335 * These are the &struct net_device, they are only set internally
1336 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1337 * userspace; this means that the order of these flags can change
7aa98047
LR
1338 * during any kernel release.
1339 *
1340 * You should have a pretty good reason to be extending these flags.
1341 *
1342 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1343 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1344 * @IFF_BONDING: bonding master or slave
7aa98047 1345 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1346 * @IFF_WAN_HDLC: WAN HDLC device
1347 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1348 * release skb->dst
1349 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1350 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1351 * @IFF_MACVLAN_PORT: device used as macvlan port
1352 * @IFF_BRIDGE_PORT: device used as bridge port
1353 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1354 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1355 * @IFF_UNICAST_FLT: Supports unicast filtering
1356 * @IFF_TEAM_PORT: device used as team port
1357 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1358 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1359 * change when it's running
1360 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1361 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1362 * underlying stacked devices
1363 * @IFF_IPVLAN_MASTER: IPvlan master device
1364 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1365 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1366 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1367 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1368 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1369 * @IFF_TEAM: device is a team device
d4ab4286 1370 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1371 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1372 * entity (i.e. the master device for bridged veth)
3c175784 1373 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1374 */
1375enum netdev_priv_flags {
1376 IFF_802_1Q_VLAN = 1<<0,
1377 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1378 IFF_BONDING = 1<<2,
1379 IFF_ISATAP = 1<<3,
1380 IFF_WAN_HDLC = 1<<4,
1381 IFF_XMIT_DST_RELEASE = 1<<5,
1382 IFF_DONT_BRIDGE = 1<<6,
1383 IFF_DISABLE_NETPOLL = 1<<7,
1384 IFF_MACVLAN_PORT = 1<<8,
1385 IFF_BRIDGE_PORT = 1<<9,
1386 IFF_OVS_DATAPATH = 1<<10,
1387 IFF_TX_SKB_SHARING = 1<<11,
1388 IFF_UNICAST_FLT = 1<<12,
1389 IFF_TEAM_PORT = 1<<13,
1390 IFF_SUPP_NOFCS = 1<<14,
1391 IFF_LIVE_ADDR_CHANGE = 1<<15,
1392 IFF_MACVLAN = 1<<16,
1393 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1394 IFF_IPVLAN_MASTER = 1<<18,
1395 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1396 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1397 IFF_NO_QUEUE = 1<<21,
1398 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1399 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1400 IFF_TEAM = 1<<24,
d4ab4286 1401 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1402 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1403 IFF_MACSEC = 1<<27,
7aa98047
LR
1404};
1405
1406#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1407#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1408#define IFF_BONDING IFF_BONDING
7aa98047 1409#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1410#define IFF_WAN_HDLC IFF_WAN_HDLC
1411#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1412#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1413#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1414#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1415#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1416#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1417#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1418#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1419#define IFF_TEAM_PORT IFF_TEAM_PORT
1420#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1421#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1422#define IFF_MACVLAN IFF_MACVLAN
02875878 1423#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1424#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1425#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1426#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1427#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1428#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1429#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1430#define IFF_TEAM IFF_TEAM
d4ab4286 1431#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1432#define IFF_MACSEC IFF_MACSEC
7aa98047 1433
536721b1
KK
1434/**
1435 * struct net_device - The DEVICE structure.
1436 * Actually, this whole structure is a big mistake. It mixes I/O
1437 * data with strictly "high-level" data, and it has to know about
1438 * almost every data structure used in the INET module.
1439 *
1440 * @name: This is the first field of the "visible" part of this structure
1441 * (i.e. as seen by users in the "Space.c" file). It is the name
1442 * of the interface.
1443 *
1444 * @name_hlist: Device name hash chain, please keep it close to name[]
1445 * @ifalias: SNMP alias
1446 * @mem_end: Shared memory end
1447 * @mem_start: Shared memory start
1448 * @base_addr: Device I/O address
1449 * @irq: Device IRQ number
1450 *
14ffbbb8
TG
1451 * @carrier_changes: Stats to monitor carrier on<->off transitions
1452 *
536721b1
KK
1453 * @state: Generic network queuing layer state, see netdev_state_t
1454 * @dev_list: The global list of network devices
5e82b4b2
BH
1455 * @napi_list: List entry used for polling NAPI devices
1456 * @unreg_list: List entry when we are unregistering the
1457 * device; see the function unregister_netdev
1458 * @close_list: List entry used when we are closing the device
62d885fe
BP
1459 * @ptype_all: Device-specific packet handlers for all protocols
1460 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1461 *
1462 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1463 * @features: Currently active device features
1464 * @hw_features: User-changeable features
1465 *
1466 * @wanted_features: User-requested features
1467 * @vlan_features: Mask of features inheritable by VLAN devices
1468 *
1469 * @hw_enc_features: Mask of features inherited by encapsulating devices
1470 * This field indicates what encapsulation
1471 * offloads the hardware is capable of doing,
1472 * and drivers will need to set them appropriately.
1473 *
1474 * @mpls_features: Mask of features inheritable by MPLS
1475 *
1476 * @ifindex: interface index
5e82b4b2 1477 * @group: The group the device belongs to
536721b1
KK
1478 *
1479 * @stats: Statistics struct, which was left as a legacy, use
1480 * rtnl_link_stats64 instead
1481 *
1482 * @rx_dropped: Dropped packets by core network,
1483 * do not use this in drivers
1484 * @tx_dropped: Dropped packets by core network,
1485 * do not use this in drivers
6e7333d3
JW
1486 * @rx_nohandler: nohandler dropped packets by core network on
1487 * inactive devices, do not use this in drivers
536721b1 1488 *
536721b1
KK
1489 * @wireless_handlers: List of functions to handle Wireless Extensions,
1490 * instead of ioctl,
1491 * see <net/iw_handler.h> for details.
1492 * @wireless_data: Instance data managed by the core of wireless extensions
1493 *
1494 * @netdev_ops: Includes several pointers to callbacks,
1495 * if one wants to override the ndo_*() functions
1496 * @ethtool_ops: Management operations
f997c55c
AA
1497 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1498 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1499 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1500 * of Layer 2 headers.
1501 *
1502 * @flags: Interface flags (a la BSD)
1503 * @priv_flags: Like 'flags' but invisible to userspace,
1504 * see if.h for the definitions
1505 * @gflags: Global flags ( kept as legacy )
1506 * @padded: How much padding added by alloc_netdev()
1507 * @operstate: RFC2863 operstate
1508 * @link_mode: Mapping policy to operstate
1509 * @if_port: Selectable AUI, TP, ...
1510 * @dma: DMA channel
1511 * @mtu: Interface MTU value
61e84623
JW
1512 * @min_mtu: Interface Minimum MTU value
1513 * @max_mtu: Interface Maximum MTU value
536721b1 1514 * @type: Interface hardware type
2793a23a 1515 * @hard_header_len: Maximum hardware header length.
217e6fa2 1516 * @min_header_len: Minimum hardware header length
536721b1
KK
1517 *
1518 * @needed_headroom: Extra headroom the hardware may need, but not in all
1519 * cases can this be guaranteed
1520 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1521 * cases can this be guaranteed. Some cases also use
1522 * LL_MAX_HEADER instead to allocate the skb
1523 *
1524 * interface address info:
1525 *
1526 * @perm_addr: Permanent hw address
1527 * @addr_assign_type: Hw address assignment type
1528 * @addr_len: Hardware address length
8626a0c8 1529 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1530 * @dev_id: Used to differentiate devices that share
1531 * the same link layer address
1532 * @dev_port: Used to differentiate devices that share
1533 * the same function
1534 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1535 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1536 * has been enabled due to the need to listen to
1537 * additional unicast addresses in a device that
1538 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1539 * @uc: unicast mac addresses
1540 * @mc: multicast mac addresses
1541 * @dev_addrs: list of device hw addresses
1542 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1543 * @promiscuity: Number of times the NIC is told to work in
1544 * promiscuous mode; if it becomes 0 the NIC will
1545 * exit promiscuous mode
536721b1
KK
1546 * @allmulti: Counter, enables or disables allmulticast mode
1547 *
1548 * @vlan_info: VLAN info
1549 * @dsa_ptr: dsa specific data
1550 * @tipc_ptr: TIPC specific data
1551 * @atalk_ptr: AppleTalk link
1552 * @ip_ptr: IPv4 specific data
1553 * @dn_ptr: DECnet specific data
1554 * @ip6_ptr: IPv6 specific data
1555 * @ax25_ptr: AX.25 specific data
1556 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1557 *
536721b1
KK
1558 * @dev_addr: Hw address (before bcast,
1559 * because most packets are unicast)
1560 *
1561 * @_rx: Array of RX queues
1562 * @num_rx_queues: Number of RX queues
1563 * allocated at register_netdev() time
1564 * @real_num_rx_queues: Number of RX queues currently active in device
1565 *
1566 * @rx_handler: handler for received packets
1567 * @rx_handler_data: XXX: need comments on this one
1568 * @ingress_queue: XXX: need comments on this one
1569 * @broadcast: hw bcast address
1570 *
14ffbbb8
TG
1571 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1572 * indexed by RX queue number. Assigned by driver.
1573 * This must only be set if the ndo_rx_flow_steer
1574 * operation is defined
1575 * @index_hlist: Device index hash chain
1576 *
536721b1
KK
1577 * @_tx: Array of TX queues
1578 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1579 * @real_num_tx_queues: Number of TX queues currently active in device
1580 * @qdisc: Root qdisc from userspace point of view
1581 * @tx_queue_len: Max frames per queue allowed
1582 * @tx_global_lock: XXX: need comments on this one
1583 *
1584 * @xps_maps: XXX: need comments on this one
1585 *
536721b1 1586 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1587 * the watchdog (see dev_watchdog())
536721b1
KK
1588 * @watchdog_timer: List of timers
1589 *
1590 * @pcpu_refcnt: Number of references to this device
1591 * @todo_list: Delayed register/unregister
536721b1
KK
1592 * @link_watch_list: XXX: need comments on this one
1593 *
1594 * @reg_state: Register/unregister state machine
1595 * @dismantle: Device is going to be freed
1596 * @rtnl_link_state: This enum represents the phases of creating
1597 * a new link
1598 *
1599 * @destructor: Called from unregister,
1600 * can be used to call free_netdev
1601 * @npinfo: XXX: need comments on this one
1602 * @nd_net: Network namespace this network device is inside
1603 *
1604 * @ml_priv: Mid-layer private
1605 * @lstats: Loopback statistics
1606 * @tstats: Tunnel statistics
1607 * @dstats: Dummy statistics
1608 * @vstats: Virtual ethernet statistics
1609 *
1610 * @garp_port: GARP
1611 * @mrp_port: MRP
1612 *
1613 * @dev: Class/net/name entry
1614 * @sysfs_groups: Space for optional device, statistics and wireless
1615 * sysfs groups
1616 *
1617 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1618 * @rtnl_link_ops: Rtnl_link_ops
1619 *
1620 * @gso_max_size: Maximum size of generic segmentation offload
1621 * @gso_max_segs: Maximum number of segments that can be passed to the
1622 * NIC for GSO
1623 *
1624 * @dcbnl_ops: Data Center Bridging netlink ops
1625 * @num_tc: Number of traffic classes in the net device
1626 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1627 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1628 *
1629 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1630 *
1631 * @priomap: XXX: need comments on this one
1632 * @phydev: Physical device may attach itself
1633 * for hardware timestamping
1634 *
123b3652
ED
1635 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1636 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1637 *
d746d707
AK
1638 * @proto_down: protocol port state information can be sent to the
1639 * switch driver and used to set the phys state of the
1640 * switch port.
1641 *
1da177e4
LT
1642 * FIXME: cleanup struct net_device such that network protocol info
1643 * moves out.
1644 */
1645
d94d9fee 1646struct net_device {
1da177e4 1647 char name[IFNAMSIZ];
9356b8fc 1648 struct hlist_node name_hlist;
0b815a1a 1649 char *ifalias;
1da177e4
LT
1650 /*
1651 * I/O specific fields
1652 * FIXME: Merge these and struct ifmap into one
1653 */
536721b1
KK
1654 unsigned long mem_end;
1655 unsigned long mem_start;
1656 unsigned long base_addr;
1657 int irq;
1da177e4 1658
14ffbbb8
TG
1659 atomic_t carrier_changes;
1660
1da177e4 1661 /*
536721b1
KK
1662 * Some hardware also needs these fields (state,dev_list,
1663 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1664 * part of the usual set specified in Space.c.
1665 */
1666
1da177e4
LT
1667 unsigned long state;
1668
7562f876 1669 struct list_head dev_list;
bea3348e 1670 struct list_head napi_list;
44a0873d 1671 struct list_head unreg_list;
5cde2829 1672 struct list_head close_list;
7866a621
SN
1673 struct list_head ptype_all;
1674 struct list_head ptype_specific;
2f268f12 1675
2f268f12
VF
1676 struct {
1677 struct list_head upper;
1678 struct list_head lower;
1679 } adj_list;
1680
c8f44aff 1681 netdev_features_t features;
c8f44aff 1682 netdev_features_t hw_features;
c8f44aff 1683 netdev_features_t wanted_features;
c8f44aff 1684 netdev_features_t vlan_features;
6a674e9c 1685 netdev_features_t hw_enc_features;
0d89d203 1686 netdev_features_t mpls_features;
802ab55a 1687 netdev_features_t gso_partial_features;
04ed3e74 1688
1da177e4 1689 int ifindex;
7a66bbc9 1690 int group;
1da177e4 1691
c45d286e 1692 struct net_device_stats stats;
015f0688 1693
015f0688
ED
1694 atomic_long_t rx_dropped;
1695 atomic_long_t tx_dropped;
6e7333d3 1696 atomic_long_t rx_nohandler;
1da177e4 1697
b86e0280 1698#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1699 const struct iw_handler_def *wireless_handlers;
1700 struct iw_public_data *wireless_data;
b86e0280 1701#endif
d314774c 1702 const struct net_device_ops *netdev_ops;
76fd8593 1703 const struct ethtool_ops *ethtool_ops;
4170604f 1704#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1705 const struct switchdev_ops *switchdev_ops;
4170604f 1706#endif
1b69c6d0
DA
1707#ifdef CONFIG_NET_L3_MASTER_DEV
1708 const struct l3mdev_ops *l3mdev_ops;
1709#endif
f997c55c
AA
1710#if IS_ENABLED(CONFIG_IPV6)
1711 const struct ndisc_ops *ndisc_ops;
1712#endif
1da177e4 1713
d77e38e6
SK
1714#ifdef CONFIG_XFRM
1715 const struct xfrmdev_ops *xfrmdev_ops;
1716#endif
1717
3b04ddde
SH
1718 const struct header_ops *header_ops;
1719
536721b1
KK
1720 unsigned int flags;
1721 unsigned int priv_flags;
1722
1da177e4 1723 unsigned short gflags;
536721b1 1724 unsigned short padded;
1da177e4 1725
536721b1
KK
1726 unsigned char operstate;
1727 unsigned char link_mode;
b00055aa 1728
536721b1
KK
1729 unsigned char if_port;
1730 unsigned char dma;
bdc220da 1731
536721b1 1732 unsigned int mtu;
61e84623
JW
1733 unsigned int min_mtu;
1734 unsigned int max_mtu;
536721b1
KK
1735 unsigned short type;
1736 unsigned short hard_header_len;
d92be7a4 1737 unsigned char min_header_len;
1da177e4 1738
f5184d26
JB
1739 unsigned short needed_headroom;
1740 unsigned short needed_tailroom;
1741
1da177e4 1742 /* Interface address info. */
536721b1
KK
1743 unsigned char perm_addr[MAX_ADDR_LEN];
1744 unsigned char addr_assign_type;
1745 unsigned char addr_len;
a0a9663d 1746 unsigned short neigh_priv_len;
536721b1
KK
1747 unsigned short dev_id;
1748 unsigned short dev_port;
ccffad25 1749 spinlock_t addr_list_lock;
14ffbbb8
TG
1750 unsigned char name_assign_type;
1751 bool uc_promisc;
536721b1
KK
1752 struct netdev_hw_addr_list uc;
1753 struct netdev_hw_addr_list mc;
1754 struct netdev_hw_addr_list dev_addrs;
1755
4c3d5e7b
ED
1756#ifdef CONFIG_SYSFS
1757 struct kset *queues_kset;
1758#endif
9d45abe1
WC
1759 unsigned int promiscuity;
1760 unsigned int allmulti;
1da177e4 1761
1da177e4 1762
5e82b4b2 1763 /* Protocol-specific pointers */
65ac6a5f 1764
d11ead75 1765#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1766 struct vlan_info __rcu *vlan_info;
65ac6a5f 1767#endif
34a430d7 1768#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1769 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1770#endif
1771#if IS_ENABLED(CONFIG_TIPC)
536721b1 1772 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1773#endif
536721b1
KK
1774 void *atalk_ptr;
1775 struct in_device __rcu *ip_ptr;
1776 struct dn_dev __rcu *dn_ptr;
1777 struct inet6_dev __rcu *ip6_ptr;
1778 void *ax25_ptr;
1779 struct wireless_dev *ieee80211_ptr;
98a18b6f 1780 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1781#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1782 struct mpls_dev __rcu *mpls_ptr;
1783#endif
1da177e4 1784
9356b8fc 1785/*
cd13539b 1786 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1787 */
9356b8fc 1788 /* Interface address info used in eth_type_trans() */
536721b1 1789 unsigned char *dev_addr;
f001fde5 1790
a953be53 1791#ifdef CONFIG_SYSFS
0a9627f2
TH
1792 struct netdev_rx_queue *_rx;
1793
0a9627f2 1794 unsigned int num_rx_queues;
62fe0b40 1795 unsigned int real_num_rx_queues;
df334545 1796#endif
0a9627f2 1797
7acedaf5 1798 struct bpf_prog __rcu *xdp_prog;
3b47d303 1799 unsigned long gro_flush_timeout;
61391cde 1800 rx_handler_func_t __rcu *rx_handler;
1801 void __rcu *rx_handler_data;
e8a0464c 1802
4cda01e8 1803#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1804 struct tcf_proto __rcu *ingress_cl_list;
1805#endif
24824a09 1806 struct netdev_queue __rcu *ingress_queue;
e687ad60 1807#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1808 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1809#endif
d2788d34 1810
536721b1 1811 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1812#ifdef CONFIG_RFS_ACCEL
1813 struct cpu_rmap *rx_cpu_rmap;
1814#endif
1815 struct hlist_node index_hlist;
cd13539b
ED
1816
1817/*
1818 * Cache lines mostly used on transmit path
1819 */
e8a0464c
DM
1820 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1821 unsigned int num_tx_queues;
fd2ea0a7 1822 unsigned int real_num_tx_queues;
af356afa 1823 struct Qdisc *qdisc;
59cc1f61
JK
1824#ifdef CONFIG_NET_SCHED
1825 DECLARE_HASHTABLE (qdisc_hash, 4);
1826#endif
0cd29503 1827 unsigned int tx_queue_len;
c3f26a26 1828 spinlock_t tx_global_lock;
14ffbbb8 1829 int watchdog_timeo;
cd13539b 1830
bf264145 1831#ifdef CONFIG_XPS
a4177869 1832 struct xps_dev_maps __rcu *xps_maps;
bf264145 1833#endif
1f211a1b
DB
1834#ifdef CONFIG_NET_CLS_ACT
1835 struct tcf_proto __rcu *egress_cl_list;
1836#endif
0c4f691f 1837
9356b8fc 1838 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1839 struct timer_list watchdog_timer;
1840
29b4433d 1841 int __percpu *pcpu_refcnt;
1da177e4 1842 struct list_head todo_list;
1da177e4 1843
e014debe 1844 struct list_head link_watch_list;
572a103d 1845
1da177e4 1846 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1847 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1848 NETREG_UNREGISTERING, /* called unregister_netdevice */
1849 NETREG_UNREGISTERED, /* completed unregister todo */
1850 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1851 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1852 } reg_state:8;
1853
536721b1 1854 bool dismantle;
a2835763
PM
1855
1856 enum {
1857 RTNL_LINK_INITIALIZED,
1858 RTNL_LINK_INITIALIZING,
1859 } rtnl_link_state:16;
1da177e4 1860
d314774c 1861 void (*destructor)(struct net_device *dev);
1da177e4 1862
1da177e4 1863#ifdef CONFIG_NETPOLL
5fbee843 1864 struct netpoll_info __rcu *npinfo;
1da177e4 1865#endif
eae792b7 1866
0c5c9fb5 1867 possible_net_t nd_net;
4a1c5371 1868
4951704b 1869 /* mid-layer private */
a7855c78 1870 union {
536721b1
KK
1871 void *ml_priv;
1872 struct pcpu_lstats __percpu *lstats;
8f84985f 1873 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1874 struct pcpu_dstats __percpu *dstats;
1875 struct pcpu_vstats __percpu *vstats;
a7855c78 1876 };
536721b1 1877
fb585b44 1878#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1879 struct garp_port __rcu *garp_port;
fb585b44
TK
1880#endif
1881#if IS_ENABLED(CONFIG_MRP)
febf018d 1882 struct mrp_port __rcu *mrp_port;
fb585b44 1883#endif
1da177e4 1884
5e82b4b2 1885 struct device dev;
0c509a6c 1886 const struct attribute_group *sysfs_groups[4];
a953be53 1887 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1888
38f7b870 1889 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1890
82cc1a7a
PWJ
1891 /* for setting kernel sock attribute on TCP connection setup */
1892#define GSO_MAX_SIZE 65536
1893 unsigned int gso_max_size;
30b678d8
BH
1894#define GSO_MAX_SEGS 65535
1895 u16 gso_max_segs;
743b03a8 1896
7a6b6f51 1897#ifdef CONFIG_DCB
32953543 1898 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1899#endif
5e82b4b2
BH
1900 u8 num_tc;
1901 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1902 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1903
d11ead75 1904#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1905 unsigned int fcoe_ddp_xid;
5bc1421e 1906#endif
86f8515f 1907#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1908 struct netprio_map __rcu *priomap;
4d288d57 1909#endif
5e82b4b2
BH
1910 struct phy_device *phydev;
1911 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1912 struct lock_class_key *qdisc_running_key;
5e82b4b2 1913 bool proto_down;
1da177e4 1914};
43cb76d9 1915#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1916
b5cdae32
DM
1917static inline bool netif_elide_gro(const struct net_device *dev)
1918{
1919 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1920 return true;
1921 return false;
1922}
1923
1da177e4 1924#define NETDEV_ALIGN 32
1da177e4 1925
4f57c087
JF
1926static inline
1927int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1928{
1929 return dev->prio_tc_map[prio & TC_BITMASK];
1930}
1931
1932static inline
1933int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1934{
1935 if (tc >= dev->num_tc)
1936 return -EINVAL;
1937
1938 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1939 return 0;
1940}
1941
8d059b0f 1942int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1943void netdev_reset_tc(struct net_device *dev);
1944int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1945int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1946
1947static inline
1948int netdev_get_num_tc(struct net_device *dev)
1949{
1950 return dev->num_tc;
1951}
1952
e8a0464c
DM
1953static inline
1954struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1955 unsigned int index)
1956{
1957 return &dev->_tx[index];
1958}
1959
10c51b56
DB
1960static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1961 const struct sk_buff *skb)
1962{
1963 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1964}
1965
e8a0464c
DM
1966static inline void netdev_for_each_tx_queue(struct net_device *dev,
1967 void (*f)(struct net_device *,
1968 struct netdev_queue *,
1969 void *),
1970 void *arg)
1971{
1972 unsigned int i;
1973
1974 for (i = 0; i < dev->num_tx_queues; i++)
1975 f(dev, &dev->_tx[i], arg);
1976}
1977
d3fff6c4
ED
1978#define netdev_lockdep_set_classes(dev) \
1979{ \
1980 static struct lock_class_key qdisc_tx_busylock_key; \
1981 static struct lock_class_key qdisc_running_key; \
1982 static struct lock_class_key qdisc_xmit_lock_key; \
1983 static struct lock_class_key dev_addr_list_lock_key; \
1984 unsigned int i; \
1985 \
1986 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1987 (dev)->qdisc_running_key = &qdisc_running_key; \
1988 lockdep_set_class(&(dev)->addr_list_lock, \
1989 &dev_addr_list_lock_key); \
1990 for (i = 0; i < (dev)->num_tx_queues; i++) \
1991 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1992 &qdisc_xmit_lock_key); \
1993}
1994
f629d208 1995struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1996 struct sk_buff *skb,
1997 void *accel_priv);
8c4c49df 1998
871b642a
PA
1999/* returns the headroom that the master device needs to take in account
2000 * when forwarding to this dev
2001 */
2002static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2003{
2004 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2005}
2006
2007static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2008{
2009 if (dev->netdev_ops->ndo_set_rx_headroom)
2010 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2011}
2012
2013/* set the device rx headroom to the dev's default */
2014static inline void netdev_reset_rx_headroom(struct net_device *dev)
2015{
2016 netdev_set_rx_headroom(dev, -1);
2017}
2018
c346dca1
YH
2019/*
2020 * Net namespace inlines
2021 */
2022static inline
2023struct net *dev_net(const struct net_device *dev)
2024{
c2d9ba9b 2025 return read_pnet(&dev->nd_net);
c346dca1
YH
2026}
2027
2028static inline
f5aa23fd 2029void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2030{
0c5c9fb5 2031 write_pnet(&dev->nd_net, net);
c346dca1
YH
2032}
2033
bea3348e
SH
2034/**
2035 * netdev_priv - access network device private data
2036 * @dev: network device
2037 *
2038 * Get network device private data
2039 */
6472ce60 2040static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2041{
1ce8e7b5 2042 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2043}
2044
1da177e4
LT
2045/* Set the sysfs physical device reference for the network logical device
2046 * if set prior to registration will cause a symlink during initialization.
2047 */
43cb76d9 2048#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2049
384912ed 2050/* Set the sysfs device type for the network logical device to allow
3f79410c 2051 * fine-grained identification of different network device types. For
5e82b4b2 2052 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2053 */
2054#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2055
82dc3c63
ED
2056/* Default NAPI poll() weight
2057 * Device drivers are strongly advised to not use bigger value
2058 */
2059#define NAPI_POLL_WEIGHT 64
2060
3b582cc1 2061/**
5e82b4b2 2062 * netif_napi_add - initialize a NAPI context
3b582cc1 2063 * @dev: network device
5e82b4b2 2064 * @napi: NAPI context
3b582cc1
SH
2065 * @poll: polling function
2066 * @weight: default weight
2067 *
5e82b4b2
BH
2068 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2069 * *any* of the other NAPI-related functions.
3b582cc1 2070 */
d565b0a1
HX
2071void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2072 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2073
d64b5e85 2074/**
5e82b4b2 2075 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2076 * @dev: network device
5e82b4b2 2077 * @napi: NAPI context
d64b5e85
ED
2078 * @poll: polling function
2079 * @weight: default weight
2080 *
2081 * This variant of netif_napi_add() should be used from drivers using NAPI
2082 * to exclusively poll a TX queue.
2083 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2084 */
2085static inline void netif_tx_napi_add(struct net_device *dev,
2086 struct napi_struct *napi,
2087 int (*poll)(struct napi_struct *, int),
2088 int weight)
2089{
2090 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2091 netif_napi_add(dev, napi, poll, weight);
2092}
2093
d8156534 2094/**
5e82b4b2
BH
2095 * netif_napi_del - remove a NAPI context
2096 * @napi: NAPI context
d8156534 2097 *
5e82b4b2 2098 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2099 */
d565b0a1
HX
2100void netif_napi_del(struct napi_struct *napi);
2101
2102struct napi_gro_cb {
78a478d0 2103 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2104 void *frag0;
78a478d0 2105
7489594c
HX
2106 /* Length of frag0. */
2107 unsigned int frag0_len;
2108
86911732 2109 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2110 int data_offset;
86911732 2111
d565b0a1 2112 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2113 u16 flush;
2114
2115 /* Save the IP ID here and check when we get to the transport layer */
2116 u16 flush_id;
d565b0a1
HX
2117
2118 /* Number of segments aggregated. */
2e71a6f8
ED
2119 u16 count;
2120
15e2396d
TH
2121 /* Start offset for remote checksum offload */
2122 u16 gro_remcsum_start;
2123
2e71a6f8
ED
2124 /* jiffies when first packet was created/queued */
2125 unsigned long age;
86347245 2126
afe93325 2127 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2128 u16 proto;
2129
baa32ff4
TH
2130 /* This is non-zero if the packet may be of the same flow. */
2131 u8 same_flow:1;
2132
fac8e0f5
JG
2133 /* Used in tunnel GRO receive */
2134 u8 encap_mark:1;
573e8fca
TH
2135
2136 /* GRO checksum is valid */
2137 u8 csum_valid:1;
2138
662880f4
TH
2139 /* Number of checksums via CHECKSUM_UNNECESSARY */
2140 u8 csum_cnt:3;
c3c7c254 2141
baa32ff4
TH
2142 /* Free the skb? */
2143 u8 free:2;
2144#define NAPI_GRO_FREE 1
2145#define NAPI_GRO_FREE_STOLEN_HEAD 2
2146
efc98d08
TH
2147 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2148 u8 is_ipv6:1;
2149
a0ca153f
AD
2150 /* Used in GRE, set in fou/gue_gro_receive */
2151 u8 is_fou:1;
2152
1530545e
AD
2153 /* Used to determine if flush_id can be ignored */
2154 u8 is_atomic:1;
2155
fcd91dd4
SD
2156 /* Number of gro_receive callbacks this packet already went through */
2157 u8 recursion_counter:4;
2158
2159 /* 1 bit hole */
baa32ff4 2160
bf5a755f
JC
2161 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2162 __wsum csum;
2163
c3c7c254
ED
2164 /* used in skb_gro_receive() slow path */
2165 struct sk_buff *last;
d565b0a1
HX
2166};
2167
2168#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2169
fcd91dd4
SD
2170#define GRO_RECURSION_LIMIT 15
2171static inline int gro_recursion_inc_test(struct sk_buff *skb)
2172{
2173 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2174}
2175
2176typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2177static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2178 struct sk_buff **head,
2179 struct sk_buff *skb)
2180{
2181 if (unlikely(gro_recursion_inc_test(skb))) {
2182 NAPI_GRO_CB(skb)->flush |= 1;
2183 return NULL;
2184 }
2185
2186 return cb(head, skb);
2187}
2188
2189typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2190 struct sk_buff *);
2191static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2192 struct sock *sk,
2193 struct sk_buff **head,
2194 struct sk_buff *skb)
2195{
2196 if (unlikely(gro_recursion_inc_test(skb))) {
2197 NAPI_GRO_CB(skb)->flush |= 1;
2198 return NULL;
2199 }
2200
2201 return cb(sk, head, skb);
2202}
2203
1da177e4 2204struct packet_type {
f2ccd8fa
DM
2205 __be16 type; /* This is really htons(ether_type). */
2206 struct net_device *dev; /* NULL is wildcarded here */
2207 int (*func) (struct sk_buff *,
2208 struct net_device *,
2209 struct packet_type *,
2210 struct net_device *);
c0de08d0
EL
2211 bool (*id_match)(struct packet_type *ptype,
2212 struct sock *sk);
1da177e4
LT
2213 void *af_packet_priv;
2214 struct list_head list;
2215};
2216
f191a1d1 2217struct offload_callbacks {
576a30eb 2218 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2219 netdev_features_t features);
d565b0a1 2220 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2221 struct sk_buff *skb);
299603e8 2222 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2223};
2224
2225struct packet_offload {
2226 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2227 u16 priority;
f191a1d1
VY
2228 struct offload_callbacks callbacks;
2229 struct list_head list;
1da177e4
LT
2230};
2231
5e82b4b2 2232/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2233struct pcpu_sw_netstats {
2234 u64 rx_packets;
2235 u64 rx_bytes;
2236 u64 tx_packets;
2237 u64 tx_bytes;
2238 struct u64_stats_sync syncp;
2239};
2240
aabc92bb
PNA
2241#define __netdev_alloc_pcpu_stats(type, gfp) \
2242({ \
2243 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2244 if (pcpu_stats) { \
2245 int __cpu; \
2246 for_each_possible_cpu(__cpu) { \
2247 typeof(type) *stat; \
2248 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2249 u64_stats_init(&stat->syncp); \
2250 } \
2251 } \
2252 pcpu_stats; \
1c213bd2
WC
2253})
2254
aabc92bb 2255#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2256 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2257
764f5e54
JP
2258enum netdev_lag_tx_type {
2259 NETDEV_LAG_TX_TYPE_UNKNOWN,
2260 NETDEV_LAG_TX_TYPE_RANDOM,
2261 NETDEV_LAG_TX_TYPE_BROADCAST,
2262 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2263 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2264 NETDEV_LAG_TX_TYPE_HASH,
2265};
2266
2267struct netdev_lag_upper_info {
2268 enum netdev_lag_tx_type tx_type;
2269};
2270
fb1b2e3c
JP
2271struct netdev_lag_lower_state_info {
2272 u8 link_up : 1,
2273 tx_enabled : 1;
2274};
2275
1da177e4
LT
2276#include <linux/notifier.h>
2277
dcfe1421
AW
2278/* netdevice notifier chain. Please remember to update the rtnetlink
2279 * notification exclusion list in rtnetlink_event() when adding new
2280 * types.
2281 */
2282#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2283#define NETDEV_DOWN 0x0002
2284#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2285 detected a hardware crash and restarted
2286 - we can use this eg to kick tcp sessions
2287 once done */
2288#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2289#define NETDEV_REGISTER 0x0005
2290#define NETDEV_UNREGISTER 0x0006
1d486bfb 2291#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2292#define NETDEV_CHANGEADDR 0x0008
2293#define NETDEV_GOING_DOWN 0x0009
2294#define NETDEV_CHANGENAME 0x000A
2295#define NETDEV_FEAT_CHANGE 0x000B
2296#define NETDEV_BONDING_FAILOVER 0x000C
2297#define NETDEV_PRE_UP 0x000D
2298#define NETDEV_PRE_TYPE_CHANGE 0x000E
2299#define NETDEV_POST_TYPE_CHANGE 0x000F
2300#define NETDEV_POST_INIT 0x0010
0115e8e3 2301#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2302#define NETDEV_RELEASE 0x0012
2303#define NETDEV_NOTIFY_PEERS 0x0013
2304#define NETDEV_JOIN 0x0014
42e52bf9 2305#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2306#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2307#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2308#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2309#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2310#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2311#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2312#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2313#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2314
f629d208
JP
2315int register_netdevice_notifier(struct notifier_block *nb);
2316int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2317
2318struct netdev_notifier_info {
2319 struct net_device *dev;
2320};
2321
be9efd36
JP
2322struct netdev_notifier_change_info {
2323 struct netdev_notifier_info info; /* must be first */
2324 unsigned int flags_changed;
2325};
2326
0e4ead9d
JP
2327struct netdev_notifier_changeupper_info {
2328 struct netdev_notifier_info info; /* must be first */
2329 struct net_device *upper_dev; /* new upper dev */
2330 bool master; /* is upper dev master */
5e82b4b2 2331 bool linking; /* is the notification for link or unlink */
29bf24af 2332 void *upper_info; /* upper dev info */
0e4ead9d
JP
2333};
2334
04d48266
JP
2335struct netdev_notifier_changelowerstate_info {
2336 struct netdev_notifier_info info; /* must be first */
2337 void *lower_state_info; /* is lower dev state */
2338};
2339
75538c2b
CW
2340static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2341 struct net_device *dev)
2342{
2343 info->dev = dev;
2344}
2345
351638e7
JP
2346static inline struct net_device *
2347netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2348{
2349 return info->dev;
2350}
2351
f629d208 2352int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2353
2354
1da177e4
LT
2355extern rwlock_t dev_base_lock; /* Device list lock */
2356
881d966b
EB
2357#define for_each_netdev(net, d) \
2358 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2359#define for_each_netdev_reverse(net, d) \
2360 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2361#define for_each_netdev_rcu(net, d) \
2362 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2363#define for_each_netdev_safe(net, d, n) \
2364 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2365#define for_each_netdev_continue(net, d) \
2366 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2367#define for_each_netdev_continue_rcu(net, d) \
2368 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2369#define for_each_netdev_in_bond_rcu(bond, slave) \
2370 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2371 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2372#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2373
a050c33f
DL
2374static inline struct net_device *next_net_device(struct net_device *dev)
2375{
2376 struct list_head *lh;
2377 struct net *net;
2378
c346dca1 2379 net = dev_net(dev);
a050c33f
DL
2380 lh = dev->dev_list.next;
2381 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2382}
2383
ce81b76a
ED
2384static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2385{
2386 struct list_head *lh;
2387 struct net *net;
2388
2389 net = dev_net(dev);
ccf43438 2390 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2391 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2392}
2393
a050c33f
DL
2394static inline struct net_device *first_net_device(struct net *net)
2395{
2396 return list_empty(&net->dev_base_head) ? NULL :
2397 net_device_entry(net->dev_base_head.next);
2398}
7562f876 2399
ccf43438
ED
2400static inline struct net_device *first_net_device_rcu(struct net *net)
2401{
2402 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2403
2404 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2405}
2406
f629d208
JP
2407int netdev_boot_setup_check(struct net_device *dev);
2408unsigned long netdev_boot_base(const char *prefix, int unit);
2409struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2410 const char *hwaddr);
2411struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2412struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2413void dev_add_pack(struct packet_type *pt);
2414void dev_remove_pack(struct packet_type *pt);
2415void __dev_remove_pack(struct packet_type *pt);
2416void dev_add_offload(struct packet_offload *po);
2417void dev_remove_offload(struct packet_offload *po);
f629d208 2418
a54acb3a 2419int dev_get_iflink(const struct net_device *dev);
fc4099f1 2420int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2421struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2422 unsigned short mask);
f629d208
JP
2423struct net_device *dev_get_by_name(struct net *net, const char *name);
2424struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2425struct net_device *__dev_get_by_name(struct net *net, const char *name);
2426int dev_alloc_name(struct net_device *dev, const char *name);
2427int dev_open(struct net_device *dev);
2428int dev_close(struct net_device *dev);
99c4a26a 2429int dev_close_many(struct list_head *head, bool unlink);
f629d208 2430void dev_disable_lro(struct net_device *dev);
0c4b51f0 2431int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2432int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2433int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2434int register_netdevice(struct net_device *dev);
2435void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2436void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2437static inline void unregister_netdevice(struct net_device *dev)
2438{
2439 unregister_netdevice_queue(dev, NULL);
2440}
2441
f629d208
JP
2442int netdev_refcnt_read(const struct net_device *dev);
2443void free_netdev(struct net_device *dev);
74d332c1 2444void netdev_freemem(struct net_device *dev);
f629d208
JP
2445void synchronize_net(void);
2446int init_dummy_netdev(struct net_device *dev);
937f1ba5 2447
f60e5990 2448DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2449#define XMIT_RECURSION_LIMIT 10
2450
f60e5990 2451static inline int dev_recursion_level(void)
2452{
2453 return this_cpu_read(xmit_recursion);
2454}
2455
f629d208
JP
2456struct net_device *dev_get_by_index(struct net *net, int ifindex);
2457struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2458struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2459struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2460int netdev_get_name(struct net *net, char *name, int ifindex);
2461int dev_restart(struct net_device *dev);
f629d208 2462int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2463
2464static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2465{
2466 return NAPI_GRO_CB(skb)->data_offset;
2467}
2468
2469static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2470{
2471 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2472}
2473
2474static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2475{
2476 NAPI_GRO_CB(skb)->data_offset += len;
2477}
2478
a5b1cf28
HX
2479static inline void *skb_gro_header_fast(struct sk_buff *skb,
2480 unsigned int offset)
86911732 2481{
a5b1cf28
HX
2482 return NAPI_GRO_CB(skb)->frag0 + offset;
2483}
78a478d0 2484
a5b1cf28
HX
2485static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2486{
2487 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2488}
78a478d0 2489
57ea52a8
HX
2490static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2491{
2492 NAPI_GRO_CB(skb)->frag0 = NULL;
2493 NAPI_GRO_CB(skb)->frag0_len = 0;
2494}
2495
a5b1cf28
HX
2496static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2497 unsigned int offset)
2498{
17dd759c
HX
2499 if (!pskb_may_pull(skb, hlen))
2500 return NULL;
2501
57ea52a8 2502 skb_gro_frag0_invalidate(skb);
17dd759c 2503 return skb->data + offset;
86911732 2504}
1da177e4 2505
36e7b1b8
HX
2506static inline void *skb_gro_network_header(struct sk_buff *skb)
2507{
78d3fd0b
HX
2508 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2509 skb_network_offset(skb);
36e7b1b8
HX
2510}
2511
bf5a755f
JC
2512static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2513 const void *start, unsigned int len)
2514{
573e8fca 2515 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2516 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2517 csum_partial(start, len, 0));
2518}
2519
573e8fca
TH
2520/* GRO checksum functions. These are logical equivalents of the normal
2521 * checksum functions (in skbuff.h) except that they operate on the GRO
2522 * offsets and fields in sk_buff.
2523 */
2524
2525__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2526
15e2396d
TH
2527static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2528{
b7fe10e5 2529 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2530}
2531
573e8fca
TH
2532static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2533 bool zero_okay,
2534 __sum16 check)
2535{
6edec0e6
TH
2536 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2537 skb_checksum_start_offset(skb) <
2538 skb_gro_offset(skb)) &&
15e2396d 2539 !skb_at_gro_remcsum_start(skb) &&
662880f4 2540 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2541 (!zero_okay || check));
2542}
2543
2544static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2545 __wsum psum)
2546{
2547 if (NAPI_GRO_CB(skb)->csum_valid &&
2548 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2549 return 0;
2550
2551 NAPI_GRO_CB(skb)->csum = psum;
2552
2553 return __skb_gro_checksum_complete(skb);
2554}
2555
573e8fca
TH
2556static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2557{
662880f4
TH
2558 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2559 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2560 NAPI_GRO_CB(skb)->csum_cnt--;
2561 } else {
2562 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2563 * verified a new top level checksum or an encapsulated one
2564 * during GRO. This saves work if we fallback to normal path.
2565 */
2566 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2567 }
2568}
2569
2570#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2571 compute_pseudo) \
2572({ \
2573 __sum16 __ret = 0; \
2574 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2575 __ret = __skb_gro_checksum_validate_complete(skb, \
2576 compute_pseudo(skb, proto)); \
219f1d79 2577 if (!__ret) \
573e8fca
TH
2578 skb_gro_incr_csum_unnecessary(skb); \
2579 __ret; \
2580})
2581
2582#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2583 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2584
2585#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2586 compute_pseudo) \
2587 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2588
2589#define skb_gro_checksum_simple_validate(skb) \
2590 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2591
d96535a1
TH
2592static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2593{
2594 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2595 !NAPI_GRO_CB(skb)->csum_valid);
2596}
2597
2598static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2599 __sum16 check, __wsum pseudo)
2600{
2601 NAPI_GRO_CB(skb)->csum = ~pseudo;
2602 NAPI_GRO_CB(skb)->csum_valid = 1;
2603}
2604
2605#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2606do { \
2607 if (__skb_gro_checksum_convert_check(skb)) \
2608 __skb_gro_checksum_convert(skb, check, \
2609 compute_pseudo(skb, proto)); \
2610} while (0)
2611
26c4f7da
TH
2612struct gro_remcsum {
2613 int offset;
2614 __wsum delta;
2615};
2616
2617static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2618{
846cd667 2619 grc->offset = 0;
26c4f7da
TH
2620 grc->delta = 0;
2621}
2622
b7fe10e5
TH
2623static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2624 unsigned int off, size_t hdrlen,
2625 int start, int offset,
2626 struct gro_remcsum *grc,
2627 bool nopartial)
dcdc8994
TH
2628{
2629 __wsum delta;
b7fe10e5 2630 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2631
2632 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2633
15e2396d 2634 if (!nopartial) {
b7fe10e5
TH
2635 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2636 return ptr;
2637 }
2638
2639 ptr = skb_gro_header_fast(skb, off);
2640 if (skb_gro_header_hard(skb, off + plen)) {
2641 ptr = skb_gro_header_slow(skb, off + plen, off);
2642 if (!ptr)
2643 return NULL;
15e2396d
TH
2644 }
2645
b7fe10e5
TH
2646 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2647 start, offset);
dcdc8994
TH
2648
2649 /* Adjust skb->csum since we changed the packet */
dcdc8994 2650 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2651
b7fe10e5 2652 grc->offset = off + hdrlen + offset;
26c4f7da 2653 grc->delta = delta;
b7fe10e5
TH
2654
2655 return ptr;
dcdc8994
TH
2656}
2657
26c4f7da
TH
2658static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2659 struct gro_remcsum *grc)
2660{
b7fe10e5
TH
2661 void *ptr;
2662 size_t plen = grc->offset + sizeof(u16);
2663
26c4f7da
TH
2664 if (!grc->delta)
2665 return;
2666
b7fe10e5
TH
2667 ptr = skb_gro_header_fast(skb, grc->offset);
2668 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2669 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2670 if (!ptr)
2671 return;
2672 }
2673
2674 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2675}
dcdc8994 2676
25393d3f
SK
2677#ifdef CONFIG_XFRM_OFFLOAD
2678static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2679{
2680 if (PTR_ERR(pp) != -EINPROGRESS)
2681 NAPI_GRO_CB(skb)->flush |= flush;
2682}
2683#else
5f114163
SK
2684static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2685{
2686 NAPI_GRO_CB(skb)->flush |= flush;
2687}
25393d3f 2688#endif
5f114163 2689
0c4e8581
SH
2690static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2691 unsigned short type,
3b04ddde 2692 const void *daddr, const void *saddr,
95c96174 2693 unsigned int len)
0c4e8581 2694{
f1ecfd5d 2695 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2696 return 0;
3b04ddde
SH
2697
2698 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2699}
2700
b95cce35
SH
2701static inline int dev_parse_header(const struct sk_buff *skb,
2702 unsigned char *haddr)
2703{
2704 const struct net_device *dev = skb->dev;
2705
1b83336b 2706 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2707 return 0;
3b04ddde 2708 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2709}
2710
2793a23a
WB
2711/* ll_header must have at least hard_header_len allocated */
2712static inline bool dev_validate_header(const struct net_device *dev,
2713 char *ll_header, int len)
2714{
2715 if (likely(len >= dev->hard_header_len))
2716 return true;
217e6fa2
WB
2717 if (len < dev->min_header_len)
2718 return false;
2793a23a
WB
2719
2720 if (capable(CAP_SYS_RAWIO)) {
2721 memset(ll_header + len, 0, dev->hard_header_len - len);
2722 return true;
2723 }
2724
2725 if (dev->header_ops && dev->header_ops->validate)
2726 return dev->header_ops->validate(ll_header, len);
2727
2728 return false;
2729}
2730
1da177e4 2731typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2732int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2733static inline int unregister_gifconf(unsigned int family)
2734{
2735 return register_gifconf(family, NULL);
2736}
2737
99bbc707 2738#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2739#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2740struct sd_flow_limit {
2741 u64 count;
2742 unsigned int num_buckets;
2743 unsigned int history_head;
2744 u16 history[FLOW_LIMIT_HISTORY];
2745 u8 buckets[];
2746};
2747
2748extern int netdev_flow_limit_table_len;
2749#endif /* CONFIG_NET_FLOW_LIMIT */
2750
1da177e4 2751/*
5e82b4b2 2752 * Incoming packets are placed on per-CPU queues
1da177e4 2753 */
d94d9fee 2754struct softnet_data {
1da177e4 2755 struct list_head poll_list;
6e7676c1 2756 struct sk_buff_head process_queue;
1da177e4 2757
dee42870 2758 /* stats */
cd7b5396
DM
2759 unsigned int processed;
2760 unsigned int time_squeeze;
cd7b5396 2761 unsigned int received_rps;
fd793d89 2762#ifdef CONFIG_RPS
88751275 2763 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2764#endif
2765#ifdef CONFIG_NET_FLOW_LIMIT
2766 struct sd_flow_limit __rcu *flow_limit;
2767#endif
2768 struct Qdisc *output_queue;
2769 struct Qdisc **output_queue_tailp;
2770 struct sk_buff *completion_queue;
88751275 2771
4cdb1e2e 2772#ifdef CONFIG_RPS
501e7ef5
ED
2773 /* input_queue_head should be written by cpu owning this struct,
2774 * and only read by other cpus. Worth using a cache line.
2775 */
2776 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2777
2778 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2779 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2780 struct softnet_data *rps_ipi_next;
2781 unsigned int cpu;
76cc8b13 2782 unsigned int input_queue_tail;
1e94d72f 2783#endif
95c96174 2784 unsigned int dropped;
0a9627f2 2785 struct sk_buff_head input_pkt_queue;
bea3348e 2786 struct napi_struct backlog;
99bbc707 2787
1da177e4
LT
2788};
2789
76cc8b13 2790static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2791{
2792#ifdef CONFIG_RPS
76cc8b13
TH
2793 sd->input_queue_head++;
2794#endif
2795}
2796
2797static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2798 unsigned int *qtail)
2799{
2800#ifdef CONFIG_RPS
2801 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2802#endif
2803}
2804
0a9627f2 2805DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2806
f629d208 2807void __netif_schedule(struct Qdisc *q);
46e5da40 2808void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2809
fd2ea0a7
DM
2810static inline void netif_tx_schedule_all(struct net_device *dev)
2811{
2812 unsigned int i;
2813
2814 for (i = 0; i < dev->num_tx_queues; i++)
2815 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2816}
2817
f9a7cbbf 2818static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2819{
73466498 2820 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2821}
2822
bea3348e
SH
2823/**
2824 * netif_start_queue - allow transmit
2825 * @dev: network device
2826 *
2827 * Allow upper layers to call the device hard_start_xmit routine.
2828 */
1da177e4
LT
2829static inline void netif_start_queue(struct net_device *dev)
2830{
e8a0464c 2831 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2832}
2833
fd2ea0a7
DM
2834static inline void netif_tx_start_all_queues(struct net_device *dev)
2835{
2836 unsigned int i;
2837
2838 for (i = 0; i < dev->num_tx_queues; i++) {
2839 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2840 netif_tx_start_queue(txq);
2841 }
2842}
2843
46e5da40 2844void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2845
d29f749e
DJ
2846/**
2847 * netif_wake_queue - restart transmit
2848 * @dev: network device
2849 *
2850 * Allow upper layers to call the device hard_start_xmit routine.
2851 * Used for flow control when transmit resources are available.
2852 */
79d16385
DM
2853static inline void netif_wake_queue(struct net_device *dev)
2854{
e8a0464c 2855 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2856}
2857
fd2ea0a7
DM
2858static inline void netif_tx_wake_all_queues(struct net_device *dev)
2859{
2860 unsigned int i;
2861
2862 for (i = 0; i < dev->num_tx_queues; i++) {
2863 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2864 netif_tx_wake_queue(txq);
2865 }
2866}
2867
f9a7cbbf 2868static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2869{
73466498 2870 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2871}
2872
bea3348e
SH
2873/**
2874 * netif_stop_queue - stop transmitted packets
2875 * @dev: network device
2876 *
2877 * Stop upper layers calling the device hard_start_xmit routine.
2878 * Used for flow control when transmit resources are unavailable.
2879 */
1da177e4
LT
2880static inline void netif_stop_queue(struct net_device *dev)
2881{
e8a0464c 2882 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2883}
2884
a2029240 2885void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2886
4d29515f 2887static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2888{
73466498 2889 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2890}
2891
bea3348e
SH
2892/**
2893 * netif_queue_stopped - test if transmit queue is flowblocked
2894 * @dev: network device
2895 *
2896 * Test if transmit queue on device is currently unable to send.
2897 */
4d29515f 2898static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2899{
e8a0464c 2900 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2901}
2902
4d29515f 2903static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2904{
73466498
TH
2905 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2906}
2907
8e2f1a63
DB
2908static inline bool
2909netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2910{
2911 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2912}
2913
8e2f1a63
DB
2914static inline bool
2915netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2916{
2917 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2918}
2919
53511453
ED
2920/**
2921 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2922 * @dev_queue: pointer to transmit queue
2923 *
2924 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2925 * to give appropriate hint to the CPU.
53511453
ED
2926 */
2927static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2928{
2929#ifdef CONFIG_BQL
2930 prefetchw(&dev_queue->dql.num_queued);
2931#endif
2932}
2933
2934/**
2935 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2936 * @dev_queue: pointer to transmit queue
2937 *
2938 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2939 * to give appropriate hint to the CPU.
53511453
ED
2940 */
2941static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2942{
2943#ifdef CONFIG_BQL
2944 prefetchw(&dev_queue->dql.limit);
2945#endif
2946}
2947
c5d67bd7
TH
2948static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2949 unsigned int bytes)
2950{
114cf580
TH
2951#ifdef CONFIG_BQL
2952 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2953
2954 if (likely(dql_avail(&dev_queue->dql) >= 0))
2955 return;
2956
2957 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2958
2959 /*
2960 * The XOFF flag must be set before checking the dql_avail below,
2961 * because in netdev_tx_completed_queue we update the dql_completed
2962 * before checking the XOFF flag.
2963 */
2964 smp_mb();
2965
2966 /* check again in case another CPU has just made room avail */
2967 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2968 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2969#endif
c5d67bd7
TH
2970}
2971
0042d0c8
FF
2972/**
2973 * netdev_sent_queue - report the number of bytes queued to hardware
2974 * @dev: network device
2975 * @bytes: number of bytes queued to the hardware device queue
2976 *
2977 * Report the number of bytes queued for sending/completion to the network
2978 * device hardware queue. @bytes should be a good approximation and should
2979 * exactly match netdev_completed_queue() @bytes
2980 */
c5d67bd7
TH
2981static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2982{
2983 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2984}
2985
2986static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2987 unsigned int pkts, unsigned int bytes)
c5d67bd7 2988{
114cf580 2989#ifdef CONFIG_BQL
b37c0fbe
AD
2990 if (unlikely(!bytes))
2991 return;
2992
2993 dql_completed(&dev_queue->dql, bytes);
2994
2995 /*
2996 * Without the memory barrier there is a small possiblity that
2997 * netdev_tx_sent_queue will miss the update and cause the queue to
2998 * be stopped forever
2999 */
3000 smp_mb();
3001
3002 if (dql_avail(&dev_queue->dql) < 0)
3003 return;
3004
3005 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3006 netif_schedule_queue(dev_queue);
114cf580 3007#endif
c5d67bd7
TH
3008}
3009
0042d0c8
FF
3010/**
3011 * netdev_completed_queue - report bytes and packets completed by device
3012 * @dev: network device
3013 * @pkts: actual number of packets sent over the medium
3014 * @bytes: actual number of bytes sent over the medium
3015 *
3016 * Report the number of bytes and packets transmitted by the network device
3017 * hardware queue over the physical medium, @bytes must exactly match the
3018 * @bytes amount passed to netdev_sent_queue()
3019 */
c5d67bd7 3020static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3021 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3022{
3023 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3024}
3025
3026static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3027{
114cf580 3028#ifdef CONFIG_BQL
5c490354 3029 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3030 dql_reset(&q->dql);
3031#endif
c5d67bd7
TH
3032}
3033
0042d0c8
FF
3034/**
3035 * netdev_reset_queue - reset the packets and bytes count of a network device
3036 * @dev_queue: network device
3037 *
3038 * Reset the bytes and packet count of a network device and clear the
3039 * software flow control OFF bit for this network device
3040 */
c5d67bd7
TH
3041static inline void netdev_reset_queue(struct net_device *dev_queue)
3042{
3043 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3044}
3045
b9507bda
DB
3046/**
3047 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3048 * @dev: network device
3049 * @queue_index: given tx queue index
3050 *
3051 * Returns 0 if given tx queue index >= number of device tx queues,
3052 * otherwise returns the originally passed tx queue index.
3053 */
3054static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3055{
3056 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3057 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3058 dev->name, queue_index,
3059 dev->real_num_tx_queues);
3060 return 0;
3061 }
3062
3063 return queue_index;
3064}
3065
bea3348e
SH
3066/**
3067 * netif_running - test if up
3068 * @dev: network device
3069 *
3070 * Test if the device has been brought up.
3071 */
4d29515f 3072static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3073{
3074 return test_bit(__LINK_STATE_START, &dev->state);
3075}
3076
f25f4e44 3077/*
5e82b4b2 3078 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3079 * stop, and a check if it's stopped. All other device management is
3080 * done at the overall netdevice level.
3081 * Also test the device if we're multiqueue.
3082 */
bea3348e
SH
3083
3084/**
3085 * netif_start_subqueue - allow sending packets on subqueue
3086 * @dev: network device
3087 * @queue_index: sub queue index
3088 *
3089 * Start individual transmit queue of a device with multiple transmit queues.
3090 */
f25f4e44
PWJ
3091static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3092{
fd2ea0a7 3093 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3094
3095 netif_tx_start_queue(txq);
f25f4e44
PWJ
3096}
3097
bea3348e
SH
3098/**
3099 * netif_stop_subqueue - stop sending packets on subqueue
3100 * @dev: network device
3101 * @queue_index: sub queue index
3102 *
3103 * Stop individual transmit queue of a device with multiple transmit queues.
3104 */
f25f4e44
PWJ
3105static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3106{
fd2ea0a7 3107 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3108 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3109}
3110
bea3348e
SH
3111/**
3112 * netif_subqueue_stopped - test status of subqueue
3113 * @dev: network device
3114 * @queue_index: sub queue index
3115 *
3116 * Check individual transmit queue of a device with multiple transmit queues.
3117 */
4d29515f
DM
3118static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3119 u16 queue_index)
f25f4e44 3120{
fd2ea0a7 3121 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3122
3123 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3124}
3125
4d29515f
DM
3126static inline bool netif_subqueue_stopped(const struct net_device *dev,
3127 struct sk_buff *skb)
668f895a
PE
3128{
3129 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3130}
bea3348e 3131
738b35cc
FF
3132/**
3133 * netif_wake_subqueue - allow sending packets on subqueue
3134 * @dev: network device
3135 * @queue_index: sub queue index
3136 *
3137 * Resume individual transmit queue of a device with multiple transmit queues.
3138 */
3139static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3140{
3141 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3142
3143 netif_tx_wake_queue(txq);
3144}
f25f4e44 3145
537c00de 3146#ifdef CONFIG_XPS
53af53ae 3147int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3148 u16 index);
537c00de
AD
3149#else
3150static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3151 const struct cpumask *mask,
537c00de
AD
3152 u16 index)
3153{
3154 return 0;
3155}
3156#endif
3157
5605c762
JP
3158u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3159 unsigned int num_tx_queues);
3160
a3d22a68
VZ
3161/*
3162 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3163 * as a distribution range limit for the returned value.
3164 */
3165static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3166 struct sk_buff *skb)
a3d22a68
VZ
3167{
3168 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3169}
3170
bea3348e
SH
3171/**
3172 * netif_is_multiqueue - test if device has multiple transmit queues
3173 * @dev: network device
3174 *
3175 * Check if device has multiple transmit queues
bea3348e 3176 */
4d29515f 3177static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3178{
a02cec21 3179 return dev->num_tx_queues > 1;
f25f4e44 3180}
1da177e4 3181
f629d208 3182int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3183
a953be53 3184#ifdef CONFIG_SYSFS
f629d208 3185int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3186#else
3187static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3188 unsigned int rxq)
3189{
3190 return 0;
3191}
3192#endif
3193
a953be53
MD
3194#ifdef CONFIG_SYSFS
3195static inline unsigned int get_netdev_rx_queue_index(
3196 struct netdev_rx_queue *queue)
3197{
3198 struct net_device *dev = queue->dev;
3199 int index = queue - dev->_rx;
3200
3201 BUG_ON(index >= dev->num_rx_queues);
3202 return index;
3203}
3204#endif
3205
16917b87 3206#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3207int netif_get_num_default_rss_queues(void);
16917b87 3208
e6247027
ED
3209enum skb_free_reason {
3210 SKB_REASON_CONSUMED,
3211 SKB_REASON_DROPPED,
3212};
3213
3214void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3215void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3216
e6247027
ED
3217/*
3218 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3219 * interrupt context or with hardware interrupts being disabled.
3220 * (in_irq() || irqs_disabled())
3221 *
3222 * We provide four helpers that can be used in following contexts :
3223 *
3224 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3225 * replacing kfree_skb(skb)
3226 *
3227 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3228 * Typically used in place of consume_skb(skb) in TX completion path
3229 *
3230 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3231 * replacing kfree_skb(skb)
3232 *
3233 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3234 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3235 */
e6247027
ED
3236static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3237{
3238 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3239}
3240
3241static inline void dev_consume_skb_irq(struct sk_buff *skb)
3242{
3243 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3244}
3245
3246static inline void dev_kfree_skb_any(struct sk_buff *skb)
3247{
3248 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3249}
3250
3251static inline void dev_consume_skb_any(struct sk_buff *skb)
3252{
3253 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3254}
1da177e4 3255
f629d208
JP
3256int netif_rx(struct sk_buff *skb);
3257int netif_rx_ni(struct sk_buff *skb);
04eb4489 3258int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3259gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3260void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3261struct sk_buff *napi_get_frags(struct napi_struct *napi);
3262gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3263struct packet_offload *gro_find_receive_by_type(__be16 type);
3264struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3265
3266static inline void napi_free_frags(struct napi_struct *napi)
3267{
3268 kfree_skb(napi->skb);
3269 napi->skb = NULL;
3270}
3271
24b27fc4 3272bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3273int netdev_rx_handler_register(struct net_device *dev,
3274 rx_handler_func_t *rx_handler,
3275 void *rx_handler_data);
3276void netdev_rx_handler_unregister(struct net_device *dev);
3277
3278bool dev_valid_name(const char *name);
3279int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3280int dev_ethtool(struct net *net, struct ifreq *);
3281unsigned int dev_get_flags(const struct net_device *);
3282int __dev_change_flags(struct net_device *, unsigned int flags);
3283int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3284void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3285 unsigned int gchanges);
f629d208
JP
3286int dev_change_name(struct net_device *, const char *);
3287int dev_set_alias(struct net_device *, const char *, size_t);
3288int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3289int dev_set_mtu(struct net_device *, int);
3290void dev_set_group(struct net_device *, int);
3291int dev_set_mac_address(struct net_device *, struct sockaddr *);
3292int dev_change_carrier(struct net_device *, bool new_carrier);
3293int dev_get_phys_port_id(struct net_device *dev,
02637fce 3294 struct netdev_phys_item_id *ppid);
db24a904
DA
3295int dev_get_phys_port_name(struct net_device *dev,
3296 char *name, size_t len);
d746d707 3297int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3298struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3299struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3300 struct netdev_queue *txq, int *ret);
d67b9cd2
DB
3301
3302typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3303int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3304 int fd, u32 flags);
3305bool __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op);
3306
a0265d28 3307int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3308int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3309bool is_skb_forwardable(const struct net_device *dev,
3310 const struct sk_buff *skb);
1da177e4 3311
4e3264d2
MKL
3312static __always_inline int ____dev_forward_skb(struct net_device *dev,
3313 struct sk_buff *skb)
3314{
3315 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3316 unlikely(!is_skb_forwardable(dev, skb))) {
3317 atomic_long_inc(&dev->rx_dropped);
3318 kfree_skb(skb);
3319 return NET_RX_DROP;
3320 }
3321
3322 skb_scrub_packet(skb, true);
3323 skb->priority = 0;
3324 return 0;
3325}
3326
74b20582
DA
3327void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3328
20380731 3329extern int netdev_budget;
7acf8a1e 3330extern unsigned int netdev_budget_usecs;
1da177e4
LT
3331
3332/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3333void netdev_run_todo(void);
1da177e4 3334
bea3348e
SH
3335/**
3336 * dev_put - release reference to device
3337 * @dev: network device
3338 *
9ef4429b 3339 * Release reference to device to allow it to be freed.
bea3348e 3340 */
1da177e4
LT
3341static inline void dev_put(struct net_device *dev)
3342{
933393f5 3343 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3344}
3345
bea3348e
SH
3346/**
3347 * dev_hold - get reference to device
3348 * @dev: network device
3349 *
9ef4429b 3350 * Hold reference to device to keep it from being freed.
bea3348e 3351 */
15333061
SH
3352static inline void dev_hold(struct net_device *dev)
3353{
933393f5 3354 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3355}
1da177e4
LT
3356
3357/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3358 * and _off may be called from IRQ context, but it is caller
3359 * who is responsible for serialization of these calls.
b00055aa
SR
3360 *
3361 * The name carrier is inappropriate, these functions should really be
3362 * called netif_lowerlayer_*() because they represent the state of any
3363 * kind of lower layer not just hardware media.
1da177e4
LT
3364 */
3365
f629d208
JP
3366void linkwatch_init_dev(struct net_device *dev);
3367void linkwatch_fire_event(struct net_device *dev);
3368void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3369
bea3348e
SH
3370/**
3371 * netif_carrier_ok - test if carrier present
3372 * @dev: network device
3373 *
3374 * Check if carrier is present on device
3375 */
4d29515f 3376static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3377{
3378 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3379}
3380
f629d208 3381unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3382
f629d208 3383void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3384
f629d208 3385void netif_carrier_on(struct net_device *dev);
1da177e4 3386
f629d208 3387void netif_carrier_off(struct net_device *dev);
1da177e4 3388
bea3348e
SH
3389/**
3390 * netif_dormant_on - mark device as dormant.
3391 * @dev: network device
3392 *
3393 * Mark device as dormant (as per RFC2863).
3394 *
3395 * The dormant state indicates that the relevant interface is not
3396 * actually in a condition to pass packets (i.e., it is not 'up') but is
3397 * in a "pending" state, waiting for some external event. For "on-
3398 * demand" interfaces, this new state identifies the situation where the
3399 * interface is waiting for events to place it in the up state.
bea3348e 3400 */
b00055aa
SR
3401static inline void netif_dormant_on(struct net_device *dev)
3402{
3403 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3404 linkwatch_fire_event(dev);
3405}
3406
bea3348e
SH
3407/**
3408 * netif_dormant_off - set device as not dormant.
3409 * @dev: network device
3410 *
3411 * Device is not in dormant state.
3412 */
b00055aa
SR
3413static inline void netif_dormant_off(struct net_device *dev)
3414{
3415 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3416 linkwatch_fire_event(dev);
3417}
3418
bea3348e 3419/**
8ecbc40a 3420 * netif_dormant - test if device is dormant
bea3348e
SH
3421 * @dev: network device
3422 *
8ecbc40a 3423 * Check if device is dormant.
bea3348e 3424 */
4d29515f 3425static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3426{
3427 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3428}
3429
3430
bea3348e
SH
3431/**
3432 * netif_oper_up - test if device is operational
3433 * @dev: network device
3434 *
3435 * Check if carrier is operational
3436 */
4d29515f 3437static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3438{
b00055aa
SR
3439 return (dev->operstate == IF_OPER_UP ||
3440 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3441}
3442
bea3348e
SH
3443/**
3444 * netif_device_present - is device available or removed
3445 * @dev: network device
3446 *
3447 * Check if device has not been removed from system.
3448 */
4d29515f 3449static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3450{
3451 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3452}
3453
f629d208 3454void netif_device_detach(struct net_device *dev);
1da177e4 3455
f629d208 3456void netif_device_attach(struct net_device *dev);
1da177e4
LT
3457
3458/*
3459 * Network interface message level settings
3460 */
1da177e4
LT
3461
3462enum {
3463 NETIF_MSG_DRV = 0x0001,
3464 NETIF_MSG_PROBE = 0x0002,
3465 NETIF_MSG_LINK = 0x0004,
3466 NETIF_MSG_TIMER = 0x0008,
3467 NETIF_MSG_IFDOWN = 0x0010,
3468 NETIF_MSG_IFUP = 0x0020,
3469 NETIF_MSG_RX_ERR = 0x0040,
3470 NETIF_MSG_TX_ERR = 0x0080,
3471 NETIF_MSG_TX_QUEUED = 0x0100,
3472 NETIF_MSG_INTR = 0x0200,
3473 NETIF_MSG_TX_DONE = 0x0400,
3474 NETIF_MSG_RX_STATUS = 0x0800,
3475 NETIF_MSG_PKTDATA = 0x1000,
3476 NETIF_MSG_HW = 0x2000,
3477 NETIF_MSG_WOL = 0x4000,
3478};
3479
3480#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3481#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3482#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3483#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3484#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3485#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3486#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3487#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3488#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3489#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3490#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3491#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3492#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3493#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3494#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3495
3496static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3497{
3498 /* use default */
3499 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3500 return default_msg_enable_bits;
3501 if (debug_value == 0) /* no output */
3502 return 0;
3503 /* set low N bits */
3504 return (1 << debug_value) - 1;
3505}
3506
c773e847 3507static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3508{
c773e847
DM
3509 spin_lock(&txq->_xmit_lock);
3510 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3511}
3512
5a717f4f
MT
3513static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3514{
3515 __acquire(&txq->_xmit_lock);
3516 return true;
3517}
3518
3519static inline void __netif_tx_release(struct netdev_queue *txq)
3520{
3521 __release(&txq->_xmit_lock);
3522}
3523
fd2ea0a7
DM
3524static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3525{
3526 spin_lock_bh(&txq->_xmit_lock);
3527 txq->xmit_lock_owner = smp_processor_id();
3528}
3529
4d29515f 3530static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3531{
4d29515f 3532 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3533 if (likely(ok))
3534 txq->xmit_lock_owner = smp_processor_id();
3535 return ok;
3536}
3537
3538static inline void __netif_tx_unlock(struct netdev_queue *txq)
3539{
3540 txq->xmit_lock_owner = -1;
3541 spin_unlock(&txq->_xmit_lock);
3542}
3543
3544static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3545{
3546 txq->xmit_lock_owner = -1;
3547 spin_unlock_bh(&txq->_xmit_lock);
3548}
3549
08baf561
ED
3550static inline void txq_trans_update(struct netdev_queue *txq)
3551{
3552 if (txq->xmit_lock_owner != -1)
3553 txq->trans_start = jiffies;
3554}
3555
ba162f8e
FW
3556/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3557static inline void netif_trans_update(struct net_device *dev)
3558{
9b36627a
FW
3559 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3560
3561 if (txq->trans_start != jiffies)
3562 txq->trans_start = jiffies;
ba162f8e
FW
3563}
3564
d29f749e
DJ
3565/**
3566 * netif_tx_lock - grab network device transmit lock
3567 * @dev: network device
d29f749e
DJ
3568 *
3569 * Get network device transmit lock
3570 */
22dd7495
JHS
3571static inline void netif_tx_lock(struct net_device *dev)
3572{
e8a0464c 3573 unsigned int i;
c3f26a26 3574 int cpu;
c773e847 3575
c3f26a26
DM
3576 spin_lock(&dev->tx_global_lock);
3577 cpu = smp_processor_id();
e8a0464c
DM
3578 for (i = 0; i < dev->num_tx_queues; i++) {
3579 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3580
3581 /* We are the only thread of execution doing a
3582 * freeze, but we have to grab the _xmit_lock in
3583 * order to synchronize with threads which are in
3584 * the ->hard_start_xmit() handler and already
3585 * checked the frozen bit.
3586 */
e8a0464c 3587 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3588 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3589 __netif_tx_unlock(txq);
e8a0464c 3590 }
932ff279
HX
3591}
3592
3593static inline void netif_tx_lock_bh(struct net_device *dev)
3594{
e8a0464c
DM
3595 local_bh_disable();
3596 netif_tx_lock(dev);
932ff279
HX
3597}
3598
932ff279
HX
3599static inline void netif_tx_unlock(struct net_device *dev)
3600{
e8a0464c
DM
3601 unsigned int i;
3602
3603 for (i = 0; i < dev->num_tx_queues; i++) {
3604 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3605
c3f26a26
DM
3606 /* No need to grab the _xmit_lock here. If the
3607 * queue is not stopped for another reason, we
3608 * force a schedule.
3609 */
3610 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3611 netif_schedule_queue(txq);
c3f26a26
DM
3612 }
3613 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3614}
3615
3616static inline void netif_tx_unlock_bh(struct net_device *dev)
3617{
e8a0464c
DM
3618 netif_tx_unlock(dev);
3619 local_bh_enable();
932ff279
HX
3620}
3621
c773e847 3622#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3623 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3624 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3625 } else { \
3626 __netif_tx_acquire(txq); \
22dd7495
JHS
3627 } \
3628}
3629
5efeac44
EB
3630#define HARD_TX_TRYLOCK(dev, txq) \
3631 (((dev->features & NETIF_F_LLTX) == 0) ? \
3632 __netif_tx_trylock(txq) : \
5a717f4f 3633 __netif_tx_acquire(txq))
5efeac44 3634
c773e847 3635#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3636 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3637 __netif_tx_unlock(txq); \
5a717f4f
MT
3638 } else { \
3639 __netif_tx_release(txq); \
22dd7495
JHS
3640 } \
3641}
3642
1da177e4
LT
3643static inline void netif_tx_disable(struct net_device *dev)
3644{
fd2ea0a7 3645 unsigned int i;
c3f26a26 3646 int cpu;
fd2ea0a7 3647
c3f26a26
DM
3648 local_bh_disable();
3649 cpu = smp_processor_id();
fd2ea0a7
DM
3650 for (i = 0; i < dev->num_tx_queues; i++) {
3651 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3652
3653 __netif_tx_lock(txq, cpu);
fd2ea0a7 3654 netif_tx_stop_queue(txq);
c3f26a26 3655 __netif_tx_unlock(txq);
fd2ea0a7 3656 }
c3f26a26 3657 local_bh_enable();
1da177e4
LT
3658}
3659
e308a5d8
DM
3660static inline void netif_addr_lock(struct net_device *dev)
3661{
3662 spin_lock(&dev->addr_list_lock);
3663}
3664
2429f7ac
JP
3665static inline void netif_addr_lock_nested(struct net_device *dev)
3666{
25175ba5
VY
3667 int subclass = SINGLE_DEPTH_NESTING;
3668
3669 if (dev->netdev_ops->ndo_get_lock_subclass)
3670 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3671
3672 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3673}
3674
e308a5d8
DM
3675static inline void netif_addr_lock_bh(struct net_device *dev)
3676{
3677 spin_lock_bh(&dev->addr_list_lock);
3678}
3679
3680static inline void netif_addr_unlock(struct net_device *dev)
3681{
3682 spin_unlock(&dev->addr_list_lock);
3683}
3684
3685static inline void netif_addr_unlock_bh(struct net_device *dev)
3686{
3687 spin_unlock_bh(&dev->addr_list_lock);
3688}
3689
f001fde5 3690/*
31278e71 3691 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3692 * rcu_read_lock held.
3693 */
3694#define for_each_dev_addr(dev, ha) \
31278e71 3695 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3696
1da177e4
LT
3697/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3698
f629d208 3699void ether_setup(struct net_device *dev);
1da177e4
LT
3700
3701/* Support for loadable net-drivers */
f629d208 3702struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3703 unsigned char name_assign_type,
f629d208
JP
3704 void (*setup)(struct net_device *),
3705 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3706#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3707 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3708
c835a677
TG
3709#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3710 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3711 count)
36909ea4 3712
f629d208
JP
3713int register_netdev(struct net_device *dev);
3714void unregister_netdev(struct net_device *dev);
f001fde5 3715
22bedad3 3716/* General hardware address lists handling functions */
f629d208
JP
3717int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3718 struct netdev_hw_addr_list *from_list, int addr_len);
3719void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3720 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3721int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3722 struct net_device *dev,
3723 int (*sync)(struct net_device *, const unsigned char *),
3724 int (*unsync)(struct net_device *,
3725 const unsigned char *));
3726void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3727 struct net_device *dev,
3728 int (*unsync)(struct net_device *,
3729 const unsigned char *));
f629d208 3730void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3731
f001fde5 3732/* Functions used for device addresses handling */
f629d208
JP
3733int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3734 unsigned char addr_type);
3735int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3736 unsigned char addr_type);
f629d208
JP
3737void dev_addr_flush(struct net_device *dev);
3738int dev_addr_init(struct net_device *dev);
a748ee24
JP
3739
3740/* Functions used for unicast addresses handling */
f629d208
JP
3741int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3742int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3743int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3744int dev_uc_sync(struct net_device *to, struct net_device *from);
3745int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3746void dev_uc_unsync(struct net_device *to, struct net_device *from);
3747void dev_uc_flush(struct net_device *dev);
3748void dev_uc_init(struct net_device *dev);
f001fde5 3749
670e5b8e
AD
3750/**
3751 * __dev_uc_sync - Synchonize device's unicast list
3752 * @dev: device to sync
3753 * @sync: function to call if address should be added
3754 * @unsync: function to call if address should be removed
3755 *
3756 * Add newly added addresses to the interface, and release
3757 * addresses that have been deleted.
5e82b4b2 3758 */
670e5b8e
AD
3759static inline int __dev_uc_sync(struct net_device *dev,
3760 int (*sync)(struct net_device *,
3761 const unsigned char *),
3762 int (*unsync)(struct net_device *,
3763 const unsigned char *))
3764{
3765 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3766}
3767
3768/**
e793c0f7 3769 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3770 * @dev: device to sync
3771 * @unsync: function to call if address should be removed
3772 *
3773 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3774 */
670e5b8e
AD
3775static inline void __dev_uc_unsync(struct net_device *dev,
3776 int (*unsync)(struct net_device *,
3777 const unsigned char *))
3778{
3779 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3780}
3781
22bedad3 3782/* Functions used for multicast addresses handling */
f629d208
JP
3783int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3784int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3785int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3786int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3787int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3788int dev_mc_sync(struct net_device *to, struct net_device *from);
3789int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3790void dev_mc_unsync(struct net_device *to, struct net_device *from);
3791void dev_mc_flush(struct net_device *dev);
3792void dev_mc_init(struct net_device *dev);
f001fde5 3793
670e5b8e
AD
3794/**
3795 * __dev_mc_sync - Synchonize device's multicast list
3796 * @dev: device to sync
3797 * @sync: function to call if address should be added
3798 * @unsync: function to call if address should be removed
3799 *
3800 * Add newly added addresses to the interface, and release
3801 * addresses that have been deleted.
5e82b4b2 3802 */
670e5b8e
AD
3803static inline int __dev_mc_sync(struct net_device *dev,
3804 int (*sync)(struct net_device *,
3805 const unsigned char *),
3806 int (*unsync)(struct net_device *,
3807 const unsigned char *))
3808{
3809 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3810}
3811
3812/**
e793c0f7 3813 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3814 * @dev: device to sync
3815 * @unsync: function to call if address should be removed
3816 *
3817 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3818 */
670e5b8e
AD
3819static inline void __dev_mc_unsync(struct net_device *dev,
3820 int (*unsync)(struct net_device *,
3821 const unsigned char *))
3822{
3823 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3824}
3825
4417da66 3826/* Functions used for secondary unicast and multicast support */
f629d208
JP
3827void dev_set_rx_mode(struct net_device *dev);
3828void __dev_set_rx_mode(struct net_device *dev);
3829int dev_set_promiscuity(struct net_device *dev, int inc);
3830int dev_set_allmulti(struct net_device *dev, int inc);
3831void netdev_state_change(struct net_device *dev);
3832void netdev_notify_peers(struct net_device *dev);
3833void netdev_features_change(struct net_device *dev);
1da177e4 3834/* Load a device via the kmod */
f629d208
JP
3835void dev_load(struct net *net, const char *name);
3836struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3837 struct rtnl_link_stats64 *storage);
3838void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3839 const struct net_device_stats *netdev_stats);
eeda3fd6 3840
1da177e4 3841extern int netdev_max_backlog;
3b098e2d 3842extern int netdev_tstamp_prequeue;
1da177e4 3843extern int weight_p;
3d48b53f
MT
3844extern int dev_weight_rx_bias;
3845extern int dev_weight_tx_bias;
3846extern int dev_rx_weight;
3847extern int dev_tx_weight;
9ff162a8 3848
f629d208 3849bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3850struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3851 struct list_head **iter);
f629d208
JP
3852struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3853 struct list_head **iter);
8b5be856 3854
44a40855
VY
3855/* iterate through upper list, must be called under RCU read lock */
3856#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3857 for (iter = &(dev)->adj_list.upper, \
3858 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3859 updev; \
3860 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3861
1a3f060c
DA
3862int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3863 int (*fn)(struct net_device *upper_dev,
3864 void *data),
3865 void *data);
3866
3867bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3868 struct net_device *upper_dev);
3869
f629d208
JP
3870void *netdev_lower_get_next_private(struct net_device *dev,
3871 struct list_head **iter);
3872void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3873 struct list_head **iter);
31088a11
VF
3874
3875#define netdev_for_each_lower_private(dev, priv, iter) \
3876 for (iter = (dev)->adj_list.lower.next, \
3877 priv = netdev_lower_get_next_private(dev, &(iter)); \
3878 priv; \
3879 priv = netdev_lower_get_next_private(dev, &(iter)))
3880
3881#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3882 for (iter = &(dev)->adj_list.lower, \
3883 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3884 priv; \
3885 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3886
4085ebe8
VY
3887void *netdev_lower_get_next(struct net_device *dev,
3888 struct list_head **iter);
7ce856aa 3889
4085ebe8 3890#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3891 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3892 ldev = netdev_lower_get_next(dev, &(iter)); \
3893 ldev; \
3894 ldev = netdev_lower_get_next(dev, &(iter)))
3895
7ce856aa
JP
3896struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3897 struct list_head **iter);
3898struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3899 struct list_head **iter);
3900
1a3f060c
DA
3901int netdev_walk_all_lower_dev(struct net_device *dev,
3902 int (*fn)(struct net_device *lower_dev,
3903 void *data),
3904 void *data);
3905int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3906 int (*fn)(struct net_device *lower_dev,
3907 void *data),
3908 void *data);
3909
f629d208 3910void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3911void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3912struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3913struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3914int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3915int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3916 struct net_device *upper_dev,
29bf24af 3917 void *upper_priv, void *upper_info);
f629d208
JP
3918void netdev_upper_dev_unlink(struct net_device *dev,
3919 struct net_device *upper_dev);
5bb025fa 3920void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3921void *netdev_lower_dev_get_private(struct net_device *dev,
3922 struct net_device *lower_dev);
04d48266
JP
3923void netdev_lower_state_changed(struct net_device *lower_dev,
3924 void *lower_state_info);
960fb622
ED
3925
3926/* RSS keys are 40 or 52 bytes long */
3927#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3928extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3929void netdev_rss_key_fill(void *buffer, size_t len);
3930
952fcfd0 3931int dev_get_nest_level(struct net_device *dev);
f629d208 3932int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 3933int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
3934int skb_csum_hwoffload_help(struct sk_buff *skb,
3935 const netdev_features_t features);
3936
f629d208
JP
3937struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3938 netdev_features_t features, bool tx_path);
3939struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3940 netdev_features_t features);
12b0004d 3941
61bd3857
MS
3942struct netdev_bonding_info {
3943 ifslave slave;
3944 ifbond master;
3945};
3946
3947struct netdev_notifier_bonding_info {
3948 struct netdev_notifier_info info; /* must be first */
3949 struct netdev_bonding_info bonding_info;
3950};
3951
3952void netdev_bonding_info_change(struct net_device *dev,
3953 struct netdev_bonding_info *bonding_info);
3954
12b0004d
CW
3955static inline
3956struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3957{
3958 return __skb_gso_segment(skb, features, true);
3959}
53d6471c 3960__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3961
3962static inline bool can_checksum_protocol(netdev_features_t features,
3963 __be16 protocol)
3964{
c8cd0989
TH
3965 if (protocol == htons(ETH_P_FCOE))
3966 return !!(features & NETIF_F_FCOE_CRC);
3967
3968 /* Assume this is an IP checksum (not SCTP CRC) */
3969
3970 if (features & NETIF_F_HW_CSUM) {
3971 /* Can checksum everything */
3972 return true;
3973 }
3974
3975 switch (protocol) {
3976 case htons(ETH_P_IP):
3977 return !!(features & NETIF_F_IP_CSUM);
3978 case htons(ETH_P_IPV6):
3979 return !!(features & NETIF_F_IPV6_CSUM);
3980 default:
3981 return false;
3982 }
ec5f0615 3983}
12b0004d 3984
fb286bb2 3985#ifdef CONFIG_BUG
f629d208 3986void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3987#else
3988static inline void netdev_rx_csum_fault(struct net_device *dev)
3989{
3990}
3991#endif
1da177e4 3992/* rx skb timestamps */
f629d208
JP
3993void net_enable_timestamp(void);
3994void net_disable_timestamp(void);
1da177e4 3995
20380731 3996#ifdef CONFIG_PROC_FS
f629d208 3997int __init dev_proc_init(void);
900ff8c6
CW
3998#else
3999#define dev_proc_init() 0
20380731
ACM
4000#endif
4001
4798248e 4002static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4003 struct sk_buff *skb, struct net_device *dev,
4004 bool more)
4798248e 4005{
fa2dbdc2 4006 skb->xmit_more = more ? 1 : 0;
0b725a2c 4007 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4008}
4009
10b3ad8c 4010static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4011 struct netdev_queue *txq, bool more)
4798248e
DM
4012{
4013 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4014 int rc;
4798248e 4015
fa2dbdc2 4016 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4017 if (rc == NETDEV_TX_OK)
4018 txq_trans_update(txq);
4019
4020 return rc;
4798248e
DM
4021}
4022
42a2d923
LT
4023int netdev_class_create_file_ns(struct class_attribute *class_attr,
4024 const void *ns);
4025void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4026 const void *ns);
58292cbe
TH
4027
4028static inline int netdev_class_create_file(struct class_attribute *class_attr)
4029{
4030 return netdev_class_create_file_ns(class_attr, NULL);
4031}
4032
4033static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4034{
4035 netdev_class_remove_file_ns(class_attr, NULL);
4036}
b8a9787e 4037
04600794
JB
4038extern struct kobj_ns_type_operations net_ns_type_operations;
4039
f629d208 4040const char *netdev_drivername(const struct net_device *dev);
6579e57b 4041
f629d208 4042void linkwatch_run_queue(void);
20380731 4043
da08143b
MK
4044static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4045 netdev_features_t f2)
4046{
c8cd0989
TH
4047 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4048 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4049 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4050 else
b6a0e72a 4051 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4052 }
da08143b 4053
c8cd0989 4054 return f1 & f2;
da08143b
MK
4055}
4056
c8f44aff
MM
4057static inline netdev_features_t netdev_get_wanted_features(
4058 struct net_device *dev)
5455c699
MM
4059{
4060 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4061}
c8f44aff
MM
4062netdev_features_t netdev_increment_features(netdev_features_t all,
4063 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4064
4065/* Allow TSO being used on stacked device :
4066 * Performing the GSO segmentation before last device
4067 * is a performance improvement.
4068 */
4069static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4070 netdev_features_t mask)
4071{
4072 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4073}
4074
6cb6a27c 4075int __netdev_update_features(struct net_device *dev);
5455c699 4076void netdev_update_features(struct net_device *dev);
afe12cc8 4077void netdev_change_features(struct net_device *dev);
7f353bf2 4078
fc4a7489
PM
4079void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4080 struct net_device *dev);
4081
e38f3025
TM
4082netdev_features_t passthru_features_check(struct sk_buff *skb,
4083 struct net_device *dev,
4084 netdev_features_t features);
c1e756bf 4085netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4086
4d29515f 4087static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4088{
7b748340 4089 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4090
4091 /* check flags correspondence */
4092 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4093 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4094 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4095 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4096 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4097 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4098 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4099 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4100 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4101 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4102 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4103 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4104 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4105 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4106 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4107 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4108 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0345e186 4109
d6b4991a 4110 return (features & feature) == feature;
576a30eb
HX
4111}
4112
4d29515f 4113static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4114{
278b2513 4115 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4116 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4117}
4118
8b86a61d 4119static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4120 netdev_features_t features)
7967168c 4121{
fc741216 4122 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4123 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4124 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4125}
4126
82cc1a7a
PWJ
4127static inline void netif_set_gso_max_size(struct net_device *dev,
4128 unsigned int size)
4129{
4130 dev->gso_max_size = size;
4131}
4132
7a7ffbab
WCC
4133static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4134 int pulled_hlen, u16 mac_offset,
4135 int mac_len)
4136{
4137 skb->protocol = protocol;
4138 skb->encapsulation = 1;
4139 skb_push(skb, pulled_hlen);
4140 skb_reset_transport_header(skb);
4141 skb->mac_header = mac_offset;
4142 skb->network_header = skb->mac_header + mac_len;
4143 skb->mac_len = mac_len;
4144}
4145
3c175784
SD
4146static inline bool netif_is_macsec(const struct net_device *dev)
4147{
4148 return dev->priv_flags & IFF_MACSEC;
4149}
4150
b618aaa9 4151static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4152{
4153 return dev->priv_flags & IFF_MACVLAN;
4154}
4155
b618aaa9 4156static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4157{
4158 return dev->priv_flags & IFF_MACVLAN_PORT;
4159}
4160
b618aaa9 4161static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4162{
4163 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4164}
4165
b618aaa9 4166static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4167{
4168 return dev->priv_flags & IFF_IPVLAN_MASTER;
4169}
4170
b618aaa9 4171static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4172{
4173 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4174}
4175
b618aaa9 4176static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4177{
4178 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4179}
4180
3bdc0eba
BG
4181static inline bool netif_supports_nofcs(struct net_device *dev)
4182{
4183 return dev->priv_flags & IFF_SUPP_NOFCS;
4184}
4185
007979ea 4186static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4187{
007979ea 4188 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4189}
4190
fee6d4c7
DA
4191static inline bool netif_is_l3_slave(const struct net_device *dev)
4192{
4193 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4194}
4195
0894ae3f
JP
4196static inline bool netif_is_bridge_master(const struct net_device *dev)
4197{
4198 return dev->priv_flags & IFF_EBRIDGE;
4199}
4200
28f9ee22
VY
4201static inline bool netif_is_bridge_port(const struct net_device *dev)
4202{
4203 return dev->priv_flags & IFF_BRIDGE_PORT;
4204}
4205
35d4e172
JP
4206static inline bool netif_is_ovs_master(const struct net_device *dev)
4207{
4208 return dev->priv_flags & IFF_OPENVSWITCH;
4209}
4210
5be66141
JP
4211static inline bool netif_is_ovs_port(const struct net_device *dev)
4212{
4213 return dev->priv_flags & IFF_OVS_DATAPATH;
4214}
4215
b618aaa9 4216static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4217{
4218 return dev->priv_flags & IFF_TEAM;
4219}
4220
b618aaa9 4221static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4222{
4223 return dev->priv_flags & IFF_TEAM_PORT;
4224}
4225
b618aaa9 4226static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4227{
4228 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4229}
4230
b618aaa9 4231static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4232{
4233 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4234}
4235
d4ab4286
KJ
4236static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4237{
4238 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4239}
4240
02875878
ED
4241/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4242static inline void netif_keep_dst(struct net_device *dev)
4243{
4244 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4245}
4246
18d3df3e
PA
4247/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4248static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4249{
4250 /* TODO: reserve and use an additional IFF bit, if we get more users */
4251 return dev->priv_flags & IFF_MACSEC;
4252}
4253
505d4f73 4254extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4255
571ba423
JP
4256/* Logging, debugging and troubleshooting/diagnostic helpers. */
4257
4258/* netdev_printk helpers, similar to dev_printk */
4259
4260static inline const char *netdev_name(const struct net_device *dev)
4261{
c6f854d5
VF
4262 if (!dev->name[0] || strchr(dev->name, '%'))
4263 return "(unnamed net_device)";
571ba423
JP
4264 return dev->name;
4265}
4266
ccc7f496
VF
4267static inline const char *netdev_reg_state(const struct net_device *dev)
4268{
4269 switch (dev->reg_state) {
4270 case NETREG_UNINITIALIZED: return " (uninitialized)";
4271 case NETREG_REGISTERED: return "";
4272 case NETREG_UNREGISTERING: return " (unregistering)";
4273 case NETREG_UNREGISTERED: return " (unregistered)";
4274 case NETREG_RELEASED: return " (released)";
4275 case NETREG_DUMMY: return " (dummy)";
4276 }
4277
4278 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4279 return " (unknown)";
4280}
4281
f629d208 4282__printf(3, 4)
6ea754eb
JP
4283void netdev_printk(const char *level, const struct net_device *dev,
4284 const char *format, ...);
f629d208 4285__printf(2, 3)
6ea754eb 4286void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4287__printf(2, 3)
6ea754eb 4288void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4289__printf(2, 3)
6ea754eb 4290void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4291__printf(2, 3)
6ea754eb 4292void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4293__printf(2, 3)
6ea754eb 4294void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4295__printf(2, 3)
6ea754eb 4296void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4297__printf(2, 3)
6ea754eb 4298void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4299
8909c9ad
VK
4300#define MODULE_ALIAS_NETDEV(device) \
4301 MODULE_ALIAS("netdev-" device)
4302
b558c96f 4303#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4304#define netdev_dbg(__dev, format, args...) \
4305do { \
ffa10cb4 4306 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4307} while (0)
b558c96f
JC
4308#elif defined(DEBUG)
4309#define netdev_dbg(__dev, format, args...) \
4310 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4311#else
4312#define netdev_dbg(__dev, format, args...) \
4313({ \
4314 if (0) \
4315 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4316})
4317#endif
4318
4319#if defined(VERBOSE_DEBUG)
4320#define netdev_vdbg netdev_dbg
4321#else
4322
4323#define netdev_vdbg(dev, format, args...) \
4324({ \
4325 if (0) \
4326 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4327 0; \
4328})
4329#endif
4330
4331/*
4332 * netdev_WARN() acts like dev_printk(), but with the key difference
4333 * of using a WARN/WARN_ON to get the message out, including the
4334 * file/line information and a backtrace.
4335 */
4336#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4337 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4338 netdev_reg_state(dev), ##args)
571ba423 4339
b3d95c5c
JP
4340/* netif printk helpers, similar to netdev_printk */
4341
4342#define netif_printk(priv, type, level, dev, fmt, args...) \
4343do { \
4344 if (netif_msg_##type(priv)) \
4345 netdev_printk(level, (dev), fmt, ##args); \
4346} while (0)
4347
f45f4321
JP
4348#define netif_level(level, priv, type, dev, fmt, args...) \
4349do { \
4350 if (netif_msg_##type(priv)) \
4351 netdev_##level(dev, fmt, ##args); \
4352} while (0)
4353
b3d95c5c 4354#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4355 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4356#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4357 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4358#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4359 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4360#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4361 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4362#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4363 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4364#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4365 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4366#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4367 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4368
0053ea9c 4369#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4370#define netif_dbg(priv, type, netdev, format, args...) \
4371do { \
4372 if (netif_msg_##type(priv)) \
b5fb0a03 4373 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4374} while (0)
0053ea9c
JP
4375#elif defined(DEBUG)
4376#define netif_dbg(priv, type, dev, format, args...) \
4377 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4378#else
4379#define netif_dbg(priv, type, dev, format, args...) \
4380({ \
4381 if (0) \
4382 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4383 0; \
4384})
4385#endif
4386
f617f276
EC
4387/* if @cond then downgrade to debug, else print at @level */
4388#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4389 do { \
4390 if (cond) \
4391 netif_dbg(priv, type, netdev, fmt, ##args); \
4392 else \
4393 netif_ ## level(priv, type, netdev, fmt, ##args); \
4394 } while (0)
4395
b3d95c5c 4396#if defined(VERBOSE_DEBUG)
bcfcc450 4397#define netif_vdbg netif_dbg
b3d95c5c
JP
4398#else
4399#define netif_vdbg(priv, type, dev, format, args...) \
4400({ \
4401 if (0) \
a4ed89cb 4402 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4403 0; \
4404})
4405#endif
571ba423 4406
900ff8c6
CW
4407/*
4408 * The list of packet types we will receive (as opposed to discard)
4409 * and the routines to invoke.
4410 *
4411 * Why 16. Because with 16 the only overlap we get on a hash of the
4412 * low nibble of the protocol value is RARP/SNAP/X.25.
4413 *
4414 * NOTE: That is no longer true with the addition of VLAN tags. Not
4415 * sure which should go first, but I bet it won't make much
4416 * difference if we are running VLANs. The good news is that
4417 * this protocol won't be in the list unless compiled in, so
4418 * the average user (w/out VLANs) will not be adversely affected.
4419 * --BLG
4420 *
4421 * 0800 IP
4422 * 8100 802.1Q VLAN
4423 * 0001 802.3
4424 * 0002 AX.25
4425 * 0004 802.2
4426 * 8035 RARP
4427 * 0005 SNAP
4428 * 0805 X.25
4429 * 0806 ARP
4430 * 8137 IPX
4431 * 0009 Localtalk
4432 * 86DD IPv6
4433 */
4434#define PTYPE_HASH_SIZE (16)
4435#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4436
385a154c 4437#endif /* _LINUX_NETDEVICE_H */