]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
udp: fix poll()
[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 826 /* XDP_QUERY_PROG */
58038695
MKL
827 struct {
828 bool prog_attached;
829 u32 prog_id;
830 };
a7862b45
BB
831 };
832};
16e5cc64 833
d77e38e6
SK
834#ifdef CONFIG_XFRM_OFFLOAD
835struct xfrmdev_ops {
836 int (*xdo_dev_state_add) (struct xfrm_state *x);
837 void (*xdo_dev_state_delete) (struct xfrm_state *x);
838 void (*xdo_dev_state_free) (struct xfrm_state *x);
839 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
840 struct xfrm_state *x);
841};
842#endif
843
d314774c
SH
844/*
845 * This structure defines the management hooks for network devices.
00829823
SH
846 * The following hooks can be defined; unless noted otherwise, they are
847 * optional and can be filled with a null pointer.
d314774c
SH
848 *
849 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
850 * This function is called once when a network device is registered.
851 * The network device can use this for any late stage initialization
852 * or semantic validation. It can fail with an error code which will
853 * be propagated back to register_netdev.
d314774c
SH
854 *
855 * void (*ndo_uninit)(struct net_device *dev);
856 * This function is called when device is unregistered or when registration
857 * fails. It is not called if init fails.
858 *
859 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 860 * This function is called when a network device transitions to the up
d314774c
SH
861 * state.
862 *
863 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 864 * This function is called when a network device transitions to the down
d314774c
SH
865 * state.
866 *
dc1f8bf6
SH
867 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
868 * struct net_device *dev);
00829823 869 * Called when a packet needs to be transmitted.
e79d8429
RR
870 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
871 * the queue before that can happen; it's for obsolete devices and weird
872 * corner cases, but the stack really does a non-trivial amount
873 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 874 * Required; cannot be NULL.
00829823 875 *
1a2a1444
DM
876 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
877 * struct net_device *dev
878 * netdev_features_t features);
879 * Called by core transmit path to determine if device is capable of
880 * performing offload operations on a given packet. This is to give
881 * the device an opportunity to implement any restrictions that cannot
882 * be otherwise expressed by feature flags. The check is called with
883 * the set of features that the stack has calculated and it returns
884 * those the driver believes to be appropriate.
cdba756f 885 *
f663dd9a 886 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 887 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 888 * Called to decide which queue to use when device supports multiple
00829823
SH
889 * transmit queues.
890 *
d314774c
SH
891 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
892 * This function is called to allow device receiver to make
5e82b4b2 893 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
894 *
895 * void (*ndo_set_rx_mode)(struct net_device *dev);
896 * This function is called device changes address list filtering.
01789349 897 * If driver handles unicast address filtering, it should set
5e82b4b2 898 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
899 *
900 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
901 * This function is called when the Media Access Control address
37b607c5 902 * needs to be changed. If this interface is not defined, the
5e82b4b2 903 * MAC address can not be changed.
d314774c
SH
904 *
905 * int (*ndo_validate_addr)(struct net_device *dev);
906 * Test if Media Access Control address is valid for the device.
907 *
908 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
909 * Called when a user requests an ioctl which can't be handled by
910 * the generic interface code. If not defined ioctls return
d314774c
SH
911 * not supported error code.
912 *
913 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
914 * Used to set network devices bus interface parameters. This interface
5e82b4b2 915 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
916 * interface (PCI) for low level management.
917 *
918 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
919 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 920 * of a device.
d314774c 921 *
00829823 922 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 923 * Callback used when the transmitter has not made any progress
d314774c
SH
924 * for dev->watchdog ticks.
925 *
bc1f4470 926 * void (*ndo_get_stats64)(struct net_device *dev,
927 * struct rtnl_link_stats64 *storage);
d308e38f 928 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 929 * Called when a user wants to get the network device usage
be1f3c2c 930 * statistics. Drivers must do one of the following:
3cfde79c
BH
931 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
932 * rtnl_link_stats64 structure passed by the caller.
82695d9b 933 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
934 * (which should normally be dev->stats) and return a pointer to
935 * it. The structure may be changed asynchronously only if each
936 * field is written atomically.
937 * 3. Update dev->stats asynchronously and atomically, and define
938 * neither operation.
d314774c 939 *
3df5b3c6 940 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
941 * Return true if this device supports offload stats of this attr_id.
942 *
943 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
944 * void *attr_data)
945 * Get statistics for offload operations by attr_id. Write it into the
946 * attr_data pointer.
947 *
5d632cb7 948 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 949 * If device supports VLAN filtering this function is called when a
80d5c368 950 * VLAN id is registered.
d314774c 951 *
5d632cb7 952 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 953 * If device supports VLAN filtering this function is called when a
80d5c368 954 * VLAN id is unregistered.
d314774c
SH
955 *
956 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
957 *
958 * SR-IOV management functions.
959 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
960 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
961 * u8 qos, __be16 proto);
ed616689
SC
962 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
963 * int max_tx_rate);
5f8444a3 964 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 965 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
966 * int (*ndo_get_vf_config)(struct net_device *dev,
967 * int vf, struct ifla_vf_info *ivf);
1d8faf48 968 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
969 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
970 * struct nlattr *port[]);
01a3d796
VZ
971 *
972 * Enable or disable the VF ability to query its RSS Redirection Table and
973 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 974 * with PF and querying it may introduce a theoretical security risk.
01a3d796 975 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 976 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
a5fcf8a6 977 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle, u32 chain_index,
6a4bc2b4
FF
978 * __be16 protocol, struct tc_to_netdev *tc);
979 * Called to setup any 'tc' scheduler, classifier or action on @dev.
980 * This is always called from the stack with the rtnl lock held and netif
981 * tx queues stopped. This allows the netdevice to perform queue
982 * management safely.
c445477d 983 *
e9bce845
YZ
984 * Fiber Channel over Ethernet (FCoE) offload functions.
985 * int (*ndo_fcoe_enable)(struct net_device *dev);
986 * Called when the FCoE protocol stack wants to start using LLD for FCoE
987 * so the underlying device can perform whatever needed configuration or
988 * initialization to support acceleration of FCoE traffic.
989 *
990 * int (*ndo_fcoe_disable)(struct net_device *dev);
991 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
992 * so the underlying device can perform whatever needed clean-ups to
993 * stop supporting acceleration of FCoE traffic.
994 *
995 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
996 * struct scatterlist *sgl, unsigned int sgc);
997 * Called when the FCoE Initiator wants to initialize an I/O that
998 * is a possible candidate for Direct Data Placement (DDP). The LLD can
999 * perform necessary setup and returns 1 to indicate the device is set up
1000 * successfully to perform DDP on this I/O, otherwise this returns 0.
1001 *
1002 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1003 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1004 * indicated by the FC exchange id 'xid', so the underlying device can
1005 * clean up and reuse resources for later DDP requests.
1006 *
1007 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1008 * struct scatterlist *sgl, unsigned int sgc);
1009 * Called when the FCoE Target wants to initialize an I/O that
1010 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1011 * perform necessary setup and returns 1 to indicate the device is set up
1012 * successfully to perform DDP on this I/O, otherwise this returns 0.
1013 *
68bad94e
NP
1014 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1015 * struct netdev_fcoe_hbainfo *hbainfo);
1016 * Called when the FCoE Protocol stack wants information on the underlying
1017 * device. This information is utilized by the FCoE protocol stack to
1018 * register attributes with Fiber Channel management service as per the
1019 * FC-GS Fabric Device Management Information(FDMI) specification.
1020 *
e9bce845
YZ
1021 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1022 * Called when the underlying device wants to override default World Wide
1023 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1024 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1025 * protocol stack to use.
1026 *
c445477d
BH
1027 * RFS acceleration.
1028 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1029 * u16 rxq_index, u32 flow_id);
1030 * Set hardware filter for RFS. rxq_index is the target queue index;
1031 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1032 * Return the filter ID on success, or a negative error code.
fbaec0ea 1033 *
8b98a70c 1034 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1035 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1036 * Called to make another netdev an underling.
1037 *
1038 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1039 * Called to release previously enslaved netdev.
5455c699
MM
1040 *
1041 * Feature/offload setting functions.
1a2a1444
DM
1042 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1043 * netdev_features_t features);
1044 * Adjusts the requested feature flags according to device-specific
1045 * constraints, and returns the resulting flags. Must not modify
1046 * the device state.
1047 *
c8f44aff 1048 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1049 * Called to update device configuration to new features. Passed
1050 * feature set might be less than what was returned by ndo_fix_features()).
1051 * Must return >0 or -errno if it changed dev->features itself.
1052 *
edc7d573 1053 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1054 * struct net_device *dev,
f6f6424b 1055 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1056 * Adds an FDB entry to dev for addr.
1690be63
VY
1057 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1058 * struct net_device *dev,
f6f6424b 1059 * const unsigned char *addr, u16 vid)
77162022
JF
1060 * Deletes the FDB entry from dev coresponding to addr.
1061 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1062 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1063 * int *idx)
77162022
JF
1064 * Used to add FDB entries to dump requests. Implementers should add
1065 * entries to skb and update idx with the number of entries.
e5a55a89 1066 *
ad41faa8
ND
1067 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1068 * u16 flags)
e5a55a89 1069 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1070 * struct net_device *dev, u32 filter_mask,
1071 * int nlflags)
ad41faa8
ND
1072 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1073 * u16 flags);
4bf84c35
JP
1074 *
1075 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1076 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1077 * which do not represent real hardware may define this to allow their
1078 * userspace components to manage their virtual carrier state. Devices
1079 * that determine carrier state from physical hardware properties (eg
1080 * network cables) or protocol-dependent mechanisms (eg
1081 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1082 *
1083 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1084 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1085 * Called to get ID of physical port of this device. If driver does
1086 * not implement this, it is assumed that the hw is not able to have
1087 * multiple net devices on single physical port.
53cf5275 1088 *
7c46a640
AD
1089 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1090 * struct udp_tunnel_info *ti);
1091 * Called by UDP tunnel to notify a driver about the UDP port and socket
1092 * address family that a UDP tunnel is listnening to. It is called only
1093 * when a new port starts listening. The operation is protected by the
1094 * RTNL.
1095 *
1096 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1097 * struct udp_tunnel_info *ti);
1098 * Called by UDP tunnel to notify the driver about a UDP port and socket
1099 * address family that the UDP tunnel is not listening to anymore. The
1100 * operation is protected by the RTNL.
1101 *
a6cc0cfa
JF
1102 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1103 * struct net_device *dev)
1104 * Called by upper layer devices to accelerate switching or other
1105 * station functionality into hardware. 'pdev is the lowerdev
1106 * to use for the offload and 'dev' is the net device that will
1107 * back the offload. Returns a pointer to the private structure
1108 * the upper layer will maintain.
1109 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1110 * Called by upper layer device to delete the station created
1111 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1112 * the station and priv is the structure returned by the add
1113 * operation.
1114 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1115 * struct net_device *dev,
1116 * void *priv);
1117 * Callback to use for xmit over the accelerated station. This
1118 * is used in place of ndo_start_xmit on accelerated net
1119 * devices.
822b3b2e
JF
1120 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1121 * int queue_index, u32 maxrate);
1122 * Called when a user wants to set a max-rate limitation of specific
1123 * TX queue.
a54acb3a
ND
1124 * int (*ndo_get_iflink)(const struct net_device *dev);
1125 * Called to get the iflink value of this device.
d746d707 1126 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1127 * bool proto_down);
d746d707
AK
1128 * This function is used to pass protocol port error state information
1129 * to the switch driver. The switch driver can react to the proto_down
1130 * by doing a phys down on the associated switch port.
fc4099f1
PS
1131 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1132 * This function is used to get egress tunnel information for given skb.
1133 * This is useful for retrieving outer tunnel header parameters while
1134 * sampling packet.
871b642a
PA
1135 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1136 * This function is used to specify the headroom that the skb must
1137 * consider when allocation skb during packet reception. Setting
1138 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1139 * forward. Setting a negative value resets the rx headroom to the
871b642a 1140 * default value.
a7862b45
BB
1141 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1142 * This function is used to set or query state related to XDP on the
1143 * netdevice. See definition of enum xdp_netdev_command for details.
d746d707 1144 *
d314774c
SH
1145 */
1146struct net_device_ops {
1147 int (*ndo_init)(struct net_device *dev);
1148 void (*ndo_uninit)(struct net_device *dev);
1149 int (*ndo_open)(struct net_device *dev);
1150 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1151 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1152 struct net_device *dev);
1153 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1154 struct net_device *dev,
1155 netdev_features_t features);
00829823 1156 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1157 struct sk_buff *skb,
99932d4f
DB
1158 void *accel_priv,
1159 select_queue_fallback_t fallback);
d314774c
SH
1160 void (*ndo_change_rx_flags)(struct net_device *dev,
1161 int flags);
d314774c 1162 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1163 int (*ndo_set_mac_address)(struct net_device *dev,
1164 void *addr);
d314774c 1165 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1166 int (*ndo_do_ioctl)(struct net_device *dev,
1167 struct ifreq *ifr, int cmd);
d314774c
SH
1168 int (*ndo_set_config)(struct net_device *dev,
1169 struct ifmap *map);
00829823
SH
1170 int (*ndo_change_mtu)(struct net_device *dev,
1171 int new_mtu);
1172 int (*ndo_neigh_setup)(struct net_device *dev,
1173 struct neigh_parms *);
d314774c
SH
1174 void (*ndo_tx_timeout) (struct net_device *dev);
1175
bc1f4470 1176 void (*ndo_get_stats64)(struct net_device *dev,
1177 struct rtnl_link_stats64 *storage);
3df5b3c6 1178 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1179 int (*ndo_get_offload_stats)(int attr_id,
1180 const struct net_device *dev,
1181 void *attr_data);
d314774c
SH
1182 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1183
8e586137 1184 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1185 __be16 proto, u16 vid);
8e586137 1186 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1187 __be16 proto, u16 vid);
d314774c 1188#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1189 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1190 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1191 struct netpoll_info *info);
0e34e931 1192 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1193#endif
95c26df8
WM
1194 int (*ndo_set_vf_mac)(struct net_device *dev,
1195 int queue, u8 *mac);
1196 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1197 int queue, u16 vlan,
1198 u8 qos, __be16 proto);
ed616689
SC
1199 int (*ndo_set_vf_rate)(struct net_device *dev,
1200 int vf, int min_tx_rate,
1201 int max_tx_rate);
5f8444a3
GR
1202 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1203 int vf, bool setting);
dd461d6a
HS
1204 int (*ndo_set_vf_trust)(struct net_device *dev,
1205 int vf, bool setting);
95c26df8
WM
1206 int (*ndo_get_vf_config)(struct net_device *dev,
1207 int vf,
1208 struct ifla_vf_info *ivf);
1d8faf48
RE
1209 int (*ndo_set_vf_link_state)(struct net_device *dev,
1210 int vf, int link_state);
3b766cd8
EBE
1211 int (*ndo_get_vf_stats)(struct net_device *dev,
1212 int vf,
1213 struct ifla_vf_stats
1214 *vf_stats);
57b61080
SF
1215 int (*ndo_set_vf_port)(struct net_device *dev,
1216 int vf,
1217 struct nlattr *port[]);
1218 int (*ndo_get_vf_port)(struct net_device *dev,
1219 int vf, struct sk_buff *skb);
cc8e27cc
EC
1220 int (*ndo_set_vf_guid)(struct net_device *dev,
1221 int vf, u64 guid,
1222 int guid_type);
01a3d796
VZ
1223 int (*ndo_set_vf_rss_query_en)(
1224 struct net_device *dev,
1225 int vf, bool setting);
16e5cc64 1226 int (*ndo_setup_tc)(struct net_device *dev,
a5fcf8a6 1227 u32 handle, u32 chain_index,
16e5cc64
JF
1228 __be16 protocol,
1229 struct tc_to_netdev *tc);
d11ead75 1230#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1231 int (*ndo_fcoe_enable)(struct net_device *dev);
1232 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1233 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1234 u16 xid,
1235 struct scatterlist *sgl,
1236 unsigned int sgc);
1237 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1238 u16 xid);
6247e086
YZ
1239 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1240 u16 xid,
1241 struct scatterlist *sgl,
1242 unsigned int sgc);
68bad94e
NP
1243 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1244 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1245#endif
1246
d11ead75 1247#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1248#define NETDEV_FCOE_WWNN 0
1249#define NETDEV_FCOE_WWPN 1
1250 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1251 u64 *wwn, int type);
4d288d57 1252#endif
3c9c36bc 1253
c445477d
BH
1254#ifdef CONFIG_RFS_ACCEL
1255 int (*ndo_rx_flow_steer)(struct net_device *dev,
1256 const struct sk_buff *skb,
1257 u16 rxq_index,
1258 u32 flow_id);
1259#endif
fbaec0ea
JP
1260 int (*ndo_add_slave)(struct net_device *dev,
1261 struct net_device *slave_dev);
1262 int (*ndo_del_slave)(struct net_device *dev,
1263 struct net_device *slave_dev);
c8f44aff
MM
1264 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1265 netdev_features_t features);
5455c699 1266 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1267 netdev_features_t features);
503eebc2
JP
1268 int (*ndo_neigh_construct)(struct net_device *dev,
1269 struct neighbour *n);
1270 void (*ndo_neigh_destroy)(struct net_device *dev,
1271 struct neighbour *n);
77162022
JF
1272
1273 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1274 struct nlattr *tb[],
77162022 1275 struct net_device *dev,
6b6e2725 1276 const unsigned char *addr,
f6f6424b 1277 u16 vid,
77162022
JF
1278 u16 flags);
1279 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1280 struct nlattr *tb[],
77162022 1281 struct net_device *dev,
f6f6424b
JP
1282 const unsigned char *addr,
1283 u16 vid);
77162022
JF
1284 int (*ndo_fdb_dump)(struct sk_buff *skb,
1285 struct netlink_callback *cb,
1286 struct net_device *dev,
5d5eacb3 1287 struct net_device *filter_dev,
d297653d 1288 int *idx);
e5a55a89
JF
1289
1290 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1291 struct nlmsghdr *nlh,
1292 u16 flags);
e5a55a89
JF
1293 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1294 u32 pid, u32 seq,
6cbdceeb 1295 struct net_device *dev,
46c264da
ND
1296 u32 filter_mask,
1297 int nlflags);
407af329 1298 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1299 struct nlmsghdr *nlh,
1300 u16 flags);
4bf84c35
JP
1301 int (*ndo_change_carrier)(struct net_device *dev,
1302 bool new_carrier);
66b52b0d 1303 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1304 struct netdev_phys_item_id *ppid);
db24a904
DA
1305 int (*ndo_get_phys_port_name)(struct net_device *dev,
1306 char *name, size_t len);
7c46a640
AD
1307 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1308 struct udp_tunnel_info *ti);
1309 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1310 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1311 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1312 struct net_device *dev);
1313 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1314 void *priv);
1315
1316 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1317 struct net_device *dev,
1318 void *priv);
25175ba5 1319 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1320 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1321 int queue_index,
1322 u32 maxrate);
a54acb3a 1323 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1324 int (*ndo_change_proto_down)(struct net_device *dev,
1325 bool proto_down);
fc4099f1
PS
1326 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1327 struct sk_buff *skb);
871b642a
PA
1328 void (*ndo_set_rx_headroom)(struct net_device *dev,
1329 int needed_headroom);
a7862b45
BB
1330 int (*ndo_xdp)(struct net_device *dev,
1331 struct netdev_xdp *xdp);
d314774c
SH
1332};
1333
7aa98047
LR
1334/**
1335 * enum net_device_priv_flags - &struct net_device priv_flags
1336 *
1337 * These are the &struct net_device, they are only set internally
1338 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1339 * userspace; this means that the order of these flags can change
7aa98047
LR
1340 * during any kernel release.
1341 *
1342 * You should have a pretty good reason to be extending these flags.
1343 *
1344 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1345 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1346 * @IFF_BONDING: bonding master or slave
7aa98047 1347 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1348 * @IFF_WAN_HDLC: WAN HDLC device
1349 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1350 * release skb->dst
1351 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1352 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1353 * @IFF_MACVLAN_PORT: device used as macvlan port
1354 * @IFF_BRIDGE_PORT: device used as bridge port
1355 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1356 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1357 * @IFF_UNICAST_FLT: Supports unicast filtering
1358 * @IFF_TEAM_PORT: device used as team port
1359 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1360 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1361 * change when it's running
1362 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1363 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1364 * underlying stacked devices
1365 * @IFF_IPVLAN_MASTER: IPvlan master device
1366 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1367 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1368 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1369 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1370 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1371 * @IFF_TEAM: device is a team device
d4ab4286 1372 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1373 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1374 * entity (i.e. the master device for bridged veth)
3c175784 1375 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1376 */
1377enum netdev_priv_flags {
1378 IFF_802_1Q_VLAN = 1<<0,
1379 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1380 IFF_BONDING = 1<<2,
1381 IFF_ISATAP = 1<<3,
1382 IFF_WAN_HDLC = 1<<4,
1383 IFF_XMIT_DST_RELEASE = 1<<5,
1384 IFF_DONT_BRIDGE = 1<<6,
1385 IFF_DISABLE_NETPOLL = 1<<7,
1386 IFF_MACVLAN_PORT = 1<<8,
1387 IFF_BRIDGE_PORT = 1<<9,
1388 IFF_OVS_DATAPATH = 1<<10,
1389 IFF_TX_SKB_SHARING = 1<<11,
1390 IFF_UNICAST_FLT = 1<<12,
1391 IFF_TEAM_PORT = 1<<13,
1392 IFF_SUPP_NOFCS = 1<<14,
1393 IFF_LIVE_ADDR_CHANGE = 1<<15,
1394 IFF_MACVLAN = 1<<16,
1395 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1396 IFF_IPVLAN_MASTER = 1<<18,
1397 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1398 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1399 IFF_NO_QUEUE = 1<<21,
1400 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1401 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1402 IFF_TEAM = 1<<24,
d4ab4286 1403 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1404 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1405 IFF_MACSEC = 1<<27,
7aa98047
LR
1406};
1407
1408#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1409#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1410#define IFF_BONDING IFF_BONDING
7aa98047 1411#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1412#define IFF_WAN_HDLC IFF_WAN_HDLC
1413#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1414#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1415#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1416#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1417#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1418#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1419#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1420#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1421#define IFF_TEAM_PORT IFF_TEAM_PORT
1422#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1423#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1424#define IFF_MACVLAN IFF_MACVLAN
02875878 1425#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1426#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1427#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1428#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1429#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1430#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1431#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1432#define IFF_TEAM IFF_TEAM
d4ab4286 1433#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1434#define IFF_MACSEC IFF_MACSEC
7aa98047 1435
536721b1
KK
1436/**
1437 * struct net_device - The DEVICE structure.
1438 * Actually, this whole structure is a big mistake. It mixes I/O
1439 * data with strictly "high-level" data, and it has to know about
1440 * almost every data structure used in the INET module.
1441 *
1442 * @name: This is the first field of the "visible" part of this structure
1443 * (i.e. as seen by users in the "Space.c" file). It is the name
1444 * of the interface.
1445 *
1446 * @name_hlist: Device name hash chain, please keep it close to name[]
1447 * @ifalias: SNMP alias
1448 * @mem_end: Shared memory end
1449 * @mem_start: Shared memory start
1450 * @base_addr: Device I/O address
1451 * @irq: Device IRQ number
1452 *
14ffbbb8
TG
1453 * @carrier_changes: Stats to monitor carrier on<->off transitions
1454 *
536721b1
KK
1455 * @state: Generic network queuing layer state, see netdev_state_t
1456 * @dev_list: The global list of network devices
5e82b4b2
BH
1457 * @napi_list: List entry used for polling NAPI devices
1458 * @unreg_list: List entry when we are unregistering the
1459 * device; see the function unregister_netdev
1460 * @close_list: List entry used when we are closing the device
62d885fe
BP
1461 * @ptype_all: Device-specific packet handlers for all protocols
1462 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1463 *
1464 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1465 * @features: Currently active device features
1466 * @hw_features: User-changeable features
1467 *
1468 * @wanted_features: User-requested features
1469 * @vlan_features: Mask of features inheritable by VLAN devices
1470 *
1471 * @hw_enc_features: Mask of features inherited by encapsulating devices
1472 * This field indicates what encapsulation
1473 * offloads the hardware is capable of doing,
1474 * and drivers will need to set them appropriately.
1475 *
1476 * @mpls_features: Mask of features inheritable by MPLS
1477 *
1478 * @ifindex: interface index
5e82b4b2 1479 * @group: The group the device belongs to
536721b1
KK
1480 *
1481 * @stats: Statistics struct, which was left as a legacy, use
1482 * rtnl_link_stats64 instead
1483 *
1484 * @rx_dropped: Dropped packets by core network,
1485 * do not use this in drivers
1486 * @tx_dropped: Dropped packets by core network,
1487 * do not use this in drivers
6e7333d3
JW
1488 * @rx_nohandler: nohandler dropped packets by core network on
1489 * inactive devices, do not use this in drivers
536721b1 1490 *
536721b1
KK
1491 * @wireless_handlers: List of functions to handle Wireless Extensions,
1492 * instead of ioctl,
1493 * see <net/iw_handler.h> for details.
1494 * @wireless_data: Instance data managed by the core of wireless extensions
1495 *
1496 * @netdev_ops: Includes several pointers to callbacks,
1497 * if one wants to override the ndo_*() functions
1498 * @ethtool_ops: Management operations
f997c55c
AA
1499 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1500 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1501 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1502 * of Layer 2 headers.
1503 *
1504 * @flags: Interface flags (a la BSD)
1505 * @priv_flags: Like 'flags' but invisible to userspace,
1506 * see if.h for the definitions
1507 * @gflags: Global flags ( kept as legacy )
1508 * @padded: How much padding added by alloc_netdev()
1509 * @operstate: RFC2863 operstate
1510 * @link_mode: Mapping policy to operstate
1511 * @if_port: Selectable AUI, TP, ...
1512 * @dma: DMA channel
1513 * @mtu: Interface MTU value
61e84623
JW
1514 * @min_mtu: Interface Minimum MTU value
1515 * @max_mtu: Interface Maximum MTU value
536721b1 1516 * @type: Interface hardware type
2793a23a 1517 * @hard_header_len: Maximum hardware header length.
217e6fa2 1518 * @min_header_len: Minimum hardware header length
536721b1
KK
1519 *
1520 * @needed_headroom: Extra headroom the hardware may need, but not in all
1521 * cases can this be guaranteed
1522 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1523 * cases can this be guaranteed. Some cases also use
1524 * LL_MAX_HEADER instead to allocate the skb
1525 *
1526 * interface address info:
1527 *
1528 * @perm_addr: Permanent hw address
1529 * @addr_assign_type: Hw address assignment type
1530 * @addr_len: Hardware address length
8626a0c8 1531 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1532 * @dev_id: Used to differentiate devices that share
1533 * the same link layer address
1534 * @dev_port: Used to differentiate devices that share
1535 * the same function
1536 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1537 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1538 * has been enabled due to the need to listen to
1539 * additional unicast addresses in a device that
1540 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1541 * @uc: unicast mac addresses
1542 * @mc: multicast mac addresses
1543 * @dev_addrs: list of device hw addresses
1544 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1545 * @promiscuity: Number of times the NIC is told to work in
1546 * promiscuous mode; if it becomes 0 the NIC will
1547 * exit promiscuous mode
536721b1
KK
1548 * @allmulti: Counter, enables or disables allmulticast mode
1549 *
1550 * @vlan_info: VLAN info
1551 * @dsa_ptr: dsa specific data
1552 * @tipc_ptr: TIPC specific data
1553 * @atalk_ptr: AppleTalk link
1554 * @ip_ptr: IPv4 specific data
1555 * @dn_ptr: DECnet specific data
1556 * @ip6_ptr: IPv6 specific data
1557 * @ax25_ptr: AX.25 specific data
1558 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1559 *
536721b1
KK
1560 * @dev_addr: Hw address (before bcast,
1561 * because most packets are unicast)
1562 *
1563 * @_rx: Array of RX queues
1564 * @num_rx_queues: Number of RX queues
1565 * allocated at register_netdev() time
1566 * @real_num_rx_queues: Number of RX queues currently active in device
1567 *
1568 * @rx_handler: handler for received packets
1569 * @rx_handler_data: XXX: need comments on this one
1570 * @ingress_queue: XXX: need comments on this one
1571 * @broadcast: hw bcast address
1572 *
14ffbbb8
TG
1573 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1574 * indexed by RX queue number. Assigned by driver.
1575 * This must only be set if the ndo_rx_flow_steer
1576 * operation is defined
1577 * @index_hlist: Device index hash chain
1578 *
536721b1
KK
1579 * @_tx: Array of TX queues
1580 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1581 * @real_num_tx_queues: Number of TX queues currently active in device
1582 * @qdisc: Root qdisc from userspace point of view
1583 * @tx_queue_len: Max frames per queue allowed
1584 * @tx_global_lock: XXX: need comments on this one
1585 *
1586 * @xps_maps: XXX: need comments on this one
1587 *
536721b1 1588 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1589 * the watchdog (see dev_watchdog())
536721b1
KK
1590 * @watchdog_timer: List of timers
1591 *
1592 * @pcpu_refcnt: Number of references to this device
1593 * @todo_list: Delayed register/unregister
536721b1
KK
1594 * @link_watch_list: XXX: need comments on this one
1595 *
1596 * @reg_state: Register/unregister state machine
1597 * @dismantle: Device is going to be freed
1598 * @rtnl_link_state: This enum represents the phases of creating
1599 * a new link
1600 *
cf124db5
DM
1601 * @needs_free_netdev: Should unregister perform free_netdev?
1602 * @priv_destructor: Called from unregister
536721b1
KK
1603 * @npinfo: XXX: need comments on this one
1604 * @nd_net: Network namespace this network device is inside
1605 *
1606 * @ml_priv: Mid-layer private
1607 * @lstats: Loopback statistics
1608 * @tstats: Tunnel statistics
1609 * @dstats: Dummy statistics
1610 * @vstats: Virtual ethernet statistics
1611 *
1612 * @garp_port: GARP
1613 * @mrp_port: MRP
1614 *
1615 * @dev: Class/net/name entry
1616 * @sysfs_groups: Space for optional device, statistics and wireless
1617 * sysfs groups
1618 *
1619 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1620 * @rtnl_link_ops: Rtnl_link_ops
1621 *
1622 * @gso_max_size: Maximum size of generic segmentation offload
1623 * @gso_max_segs: Maximum number of segments that can be passed to the
1624 * NIC for GSO
1625 *
1626 * @dcbnl_ops: Data Center Bridging netlink ops
1627 * @num_tc: Number of traffic classes in the net device
1628 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1629 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1630 *
1631 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1632 *
1633 * @priomap: XXX: need comments on this one
1634 * @phydev: Physical device may attach itself
1635 * for hardware timestamping
1636 *
123b3652
ED
1637 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1638 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1639 *
d746d707
AK
1640 * @proto_down: protocol port state information can be sent to the
1641 * switch driver and used to set the phys state of the
1642 * switch port.
1643 *
1da177e4
LT
1644 * FIXME: cleanup struct net_device such that network protocol info
1645 * moves out.
1646 */
1647
d94d9fee 1648struct net_device {
1da177e4 1649 char name[IFNAMSIZ];
9356b8fc 1650 struct hlist_node name_hlist;
0b815a1a 1651 char *ifalias;
1da177e4
LT
1652 /*
1653 * I/O specific fields
1654 * FIXME: Merge these and struct ifmap into one
1655 */
536721b1
KK
1656 unsigned long mem_end;
1657 unsigned long mem_start;
1658 unsigned long base_addr;
1659 int irq;
1da177e4 1660
14ffbbb8
TG
1661 atomic_t carrier_changes;
1662
1da177e4 1663 /*
536721b1
KK
1664 * Some hardware also needs these fields (state,dev_list,
1665 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1666 * part of the usual set specified in Space.c.
1667 */
1668
1da177e4
LT
1669 unsigned long state;
1670
7562f876 1671 struct list_head dev_list;
bea3348e 1672 struct list_head napi_list;
44a0873d 1673 struct list_head unreg_list;
5cde2829 1674 struct list_head close_list;
7866a621
SN
1675 struct list_head ptype_all;
1676 struct list_head ptype_specific;
2f268f12 1677
2f268f12
VF
1678 struct {
1679 struct list_head upper;
1680 struct list_head lower;
1681 } adj_list;
1682
c8f44aff 1683 netdev_features_t features;
c8f44aff 1684 netdev_features_t hw_features;
c8f44aff 1685 netdev_features_t wanted_features;
c8f44aff 1686 netdev_features_t vlan_features;
6a674e9c 1687 netdev_features_t hw_enc_features;
0d89d203 1688 netdev_features_t mpls_features;
802ab55a 1689 netdev_features_t gso_partial_features;
04ed3e74 1690
1da177e4 1691 int ifindex;
7a66bbc9 1692 int group;
1da177e4 1693
c45d286e 1694 struct net_device_stats stats;
015f0688 1695
015f0688
ED
1696 atomic_long_t rx_dropped;
1697 atomic_long_t tx_dropped;
6e7333d3 1698 atomic_long_t rx_nohandler;
1da177e4 1699
b86e0280 1700#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1701 const struct iw_handler_def *wireless_handlers;
1702 struct iw_public_data *wireless_data;
b86e0280 1703#endif
d314774c 1704 const struct net_device_ops *netdev_ops;
76fd8593 1705 const struct ethtool_ops *ethtool_ops;
4170604f 1706#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1707 const struct switchdev_ops *switchdev_ops;
4170604f 1708#endif
1b69c6d0
DA
1709#ifdef CONFIG_NET_L3_MASTER_DEV
1710 const struct l3mdev_ops *l3mdev_ops;
1711#endif
f997c55c
AA
1712#if IS_ENABLED(CONFIG_IPV6)
1713 const struct ndisc_ops *ndisc_ops;
1714#endif
1da177e4 1715
d77e38e6
SK
1716#ifdef CONFIG_XFRM
1717 const struct xfrmdev_ops *xfrmdev_ops;
1718#endif
1719
3b04ddde
SH
1720 const struct header_ops *header_ops;
1721
536721b1
KK
1722 unsigned int flags;
1723 unsigned int priv_flags;
1724
1da177e4 1725 unsigned short gflags;
536721b1 1726 unsigned short padded;
1da177e4 1727
536721b1
KK
1728 unsigned char operstate;
1729 unsigned char link_mode;
b00055aa 1730
536721b1
KK
1731 unsigned char if_port;
1732 unsigned char dma;
bdc220da 1733
536721b1 1734 unsigned int mtu;
61e84623
JW
1735 unsigned int min_mtu;
1736 unsigned int max_mtu;
536721b1
KK
1737 unsigned short type;
1738 unsigned short hard_header_len;
d92be7a4 1739 unsigned char min_header_len;
1da177e4 1740
f5184d26
JB
1741 unsigned short needed_headroom;
1742 unsigned short needed_tailroom;
1743
1da177e4 1744 /* Interface address info. */
536721b1
KK
1745 unsigned char perm_addr[MAX_ADDR_LEN];
1746 unsigned char addr_assign_type;
1747 unsigned char addr_len;
a0a9663d 1748 unsigned short neigh_priv_len;
536721b1
KK
1749 unsigned short dev_id;
1750 unsigned short dev_port;
ccffad25 1751 spinlock_t addr_list_lock;
14ffbbb8
TG
1752 unsigned char name_assign_type;
1753 bool uc_promisc;
536721b1
KK
1754 struct netdev_hw_addr_list uc;
1755 struct netdev_hw_addr_list mc;
1756 struct netdev_hw_addr_list dev_addrs;
1757
4c3d5e7b
ED
1758#ifdef CONFIG_SYSFS
1759 struct kset *queues_kset;
1760#endif
9d45abe1
WC
1761 unsigned int promiscuity;
1762 unsigned int allmulti;
1da177e4 1763
1da177e4 1764
5e82b4b2 1765 /* Protocol-specific pointers */
65ac6a5f 1766
d11ead75 1767#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1768 struct vlan_info __rcu *vlan_info;
65ac6a5f 1769#endif
34a430d7 1770#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1771 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1772#endif
1773#if IS_ENABLED(CONFIG_TIPC)
536721b1 1774 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1775#endif
536721b1
KK
1776 void *atalk_ptr;
1777 struct in_device __rcu *ip_ptr;
1778 struct dn_dev __rcu *dn_ptr;
1779 struct inet6_dev __rcu *ip6_ptr;
1780 void *ax25_ptr;
1781 struct wireless_dev *ieee80211_ptr;
98a18b6f 1782 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1783#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1784 struct mpls_dev __rcu *mpls_ptr;
1785#endif
1da177e4 1786
9356b8fc 1787/*
cd13539b 1788 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1789 */
9356b8fc 1790 /* Interface address info used in eth_type_trans() */
536721b1 1791 unsigned char *dev_addr;
f001fde5 1792
a953be53 1793#ifdef CONFIG_SYSFS
0a9627f2
TH
1794 struct netdev_rx_queue *_rx;
1795
0a9627f2 1796 unsigned int num_rx_queues;
62fe0b40 1797 unsigned int real_num_rx_queues;
df334545 1798#endif
0a9627f2 1799
7acedaf5 1800 struct bpf_prog __rcu *xdp_prog;
3b47d303 1801 unsigned long gro_flush_timeout;
61391cde 1802 rx_handler_func_t __rcu *rx_handler;
1803 void __rcu *rx_handler_data;
e8a0464c 1804
4cda01e8 1805#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1806 struct tcf_proto __rcu *ingress_cl_list;
1807#endif
24824a09 1808 struct netdev_queue __rcu *ingress_queue;
e687ad60 1809#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1810 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1811#endif
d2788d34 1812
536721b1 1813 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1814#ifdef CONFIG_RFS_ACCEL
1815 struct cpu_rmap *rx_cpu_rmap;
1816#endif
1817 struct hlist_node index_hlist;
cd13539b
ED
1818
1819/*
1820 * Cache lines mostly used on transmit path
1821 */
e8a0464c
DM
1822 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1823 unsigned int num_tx_queues;
fd2ea0a7 1824 unsigned int real_num_tx_queues;
af356afa 1825 struct Qdisc *qdisc;
59cc1f61
JK
1826#ifdef CONFIG_NET_SCHED
1827 DECLARE_HASHTABLE (qdisc_hash, 4);
1828#endif
0cd29503 1829 unsigned int tx_queue_len;
c3f26a26 1830 spinlock_t tx_global_lock;
14ffbbb8 1831 int watchdog_timeo;
cd13539b 1832
bf264145 1833#ifdef CONFIG_XPS
a4177869 1834 struct xps_dev_maps __rcu *xps_maps;
bf264145 1835#endif
1f211a1b
DB
1836#ifdef CONFIG_NET_CLS_ACT
1837 struct tcf_proto __rcu *egress_cl_list;
1838#endif
0c4f691f 1839
9356b8fc 1840 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1841 struct timer_list watchdog_timer;
1842
29b4433d 1843 int __percpu *pcpu_refcnt;
1da177e4 1844 struct list_head todo_list;
1da177e4 1845
e014debe 1846 struct list_head link_watch_list;
572a103d 1847
1da177e4 1848 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1849 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1850 NETREG_UNREGISTERING, /* called unregister_netdevice */
1851 NETREG_UNREGISTERED, /* completed unregister todo */
1852 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1853 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1854 } reg_state:8;
1855
536721b1 1856 bool dismantle;
a2835763
PM
1857
1858 enum {
1859 RTNL_LINK_INITIALIZED,
1860 RTNL_LINK_INITIALIZING,
1861 } rtnl_link_state:16;
1da177e4 1862
cf124db5
DM
1863 bool needs_free_netdev;
1864 void (*priv_destructor)(struct net_device *dev);
1da177e4 1865
1da177e4 1866#ifdef CONFIG_NETPOLL
5fbee843 1867 struct netpoll_info __rcu *npinfo;
1da177e4 1868#endif
eae792b7 1869
0c5c9fb5 1870 possible_net_t nd_net;
4a1c5371 1871
4951704b 1872 /* mid-layer private */
a7855c78 1873 union {
536721b1
KK
1874 void *ml_priv;
1875 struct pcpu_lstats __percpu *lstats;
8f84985f 1876 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1877 struct pcpu_dstats __percpu *dstats;
1878 struct pcpu_vstats __percpu *vstats;
a7855c78 1879 };
536721b1 1880
fb585b44 1881#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1882 struct garp_port __rcu *garp_port;
fb585b44
TK
1883#endif
1884#if IS_ENABLED(CONFIG_MRP)
febf018d 1885 struct mrp_port __rcu *mrp_port;
fb585b44 1886#endif
1da177e4 1887
5e82b4b2 1888 struct device dev;
0c509a6c 1889 const struct attribute_group *sysfs_groups[4];
a953be53 1890 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1891
38f7b870 1892 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1893
82cc1a7a
PWJ
1894 /* for setting kernel sock attribute on TCP connection setup */
1895#define GSO_MAX_SIZE 65536
1896 unsigned int gso_max_size;
30b678d8
BH
1897#define GSO_MAX_SEGS 65535
1898 u16 gso_max_segs;
743b03a8 1899
7a6b6f51 1900#ifdef CONFIG_DCB
32953543 1901 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1902#endif
5e82b4b2
BH
1903 u8 num_tc;
1904 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1905 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1906
d11ead75 1907#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1908 unsigned int fcoe_ddp_xid;
5bc1421e 1909#endif
86f8515f 1910#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1911 struct netprio_map __rcu *priomap;
4d288d57 1912#endif
5e82b4b2
BH
1913 struct phy_device *phydev;
1914 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1915 struct lock_class_key *qdisc_running_key;
5e82b4b2 1916 bool proto_down;
1da177e4 1917};
43cb76d9 1918#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1919
b5cdae32
DM
1920static inline bool netif_elide_gro(const struct net_device *dev)
1921{
1922 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1923 return true;
1924 return false;
1925}
1926
1da177e4 1927#define NETDEV_ALIGN 32
1da177e4 1928
4f57c087
JF
1929static inline
1930int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1931{
1932 return dev->prio_tc_map[prio & TC_BITMASK];
1933}
1934
1935static inline
1936int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1937{
1938 if (tc >= dev->num_tc)
1939 return -EINVAL;
1940
1941 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1942 return 0;
1943}
1944
8d059b0f 1945int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1946void netdev_reset_tc(struct net_device *dev);
1947int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1948int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1949
1950static inline
1951int netdev_get_num_tc(struct net_device *dev)
1952{
1953 return dev->num_tc;
1954}
1955
e8a0464c
DM
1956static inline
1957struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1958 unsigned int index)
1959{
1960 return &dev->_tx[index];
1961}
1962
10c51b56
DB
1963static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1964 const struct sk_buff *skb)
1965{
1966 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1967}
1968
e8a0464c
DM
1969static inline void netdev_for_each_tx_queue(struct net_device *dev,
1970 void (*f)(struct net_device *,
1971 struct netdev_queue *,
1972 void *),
1973 void *arg)
1974{
1975 unsigned int i;
1976
1977 for (i = 0; i < dev->num_tx_queues; i++)
1978 f(dev, &dev->_tx[i], arg);
1979}
1980
d3fff6c4
ED
1981#define netdev_lockdep_set_classes(dev) \
1982{ \
1983 static struct lock_class_key qdisc_tx_busylock_key; \
1984 static struct lock_class_key qdisc_running_key; \
1985 static struct lock_class_key qdisc_xmit_lock_key; \
1986 static struct lock_class_key dev_addr_list_lock_key; \
1987 unsigned int i; \
1988 \
1989 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1990 (dev)->qdisc_running_key = &qdisc_running_key; \
1991 lockdep_set_class(&(dev)->addr_list_lock, \
1992 &dev_addr_list_lock_key); \
1993 for (i = 0; i < (dev)->num_tx_queues; i++) \
1994 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1995 &qdisc_xmit_lock_key); \
1996}
1997
f629d208 1998struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1999 struct sk_buff *skb,
2000 void *accel_priv);
8c4c49df 2001
871b642a
PA
2002/* returns the headroom that the master device needs to take in account
2003 * when forwarding to this dev
2004 */
2005static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2006{
2007 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2008}
2009
2010static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2011{
2012 if (dev->netdev_ops->ndo_set_rx_headroom)
2013 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2014}
2015
2016/* set the device rx headroom to the dev's default */
2017static inline void netdev_reset_rx_headroom(struct net_device *dev)
2018{
2019 netdev_set_rx_headroom(dev, -1);
2020}
2021
c346dca1
YH
2022/*
2023 * Net namespace inlines
2024 */
2025static inline
2026struct net *dev_net(const struct net_device *dev)
2027{
c2d9ba9b 2028 return read_pnet(&dev->nd_net);
c346dca1
YH
2029}
2030
2031static inline
f5aa23fd 2032void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2033{
0c5c9fb5 2034 write_pnet(&dev->nd_net, net);
c346dca1
YH
2035}
2036
bea3348e
SH
2037/**
2038 * netdev_priv - access network device private data
2039 * @dev: network device
2040 *
2041 * Get network device private data
2042 */
6472ce60 2043static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2044{
1ce8e7b5 2045 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2046}
2047
1da177e4
LT
2048/* Set the sysfs physical device reference for the network logical device
2049 * if set prior to registration will cause a symlink during initialization.
2050 */
43cb76d9 2051#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2052
384912ed 2053/* Set the sysfs device type for the network logical device to allow
3f79410c 2054 * fine-grained identification of different network device types. For
5e82b4b2 2055 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2056 */
2057#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2058
82dc3c63
ED
2059/* Default NAPI poll() weight
2060 * Device drivers are strongly advised to not use bigger value
2061 */
2062#define NAPI_POLL_WEIGHT 64
2063
3b582cc1 2064/**
5e82b4b2 2065 * netif_napi_add - initialize a NAPI context
3b582cc1 2066 * @dev: network device
5e82b4b2 2067 * @napi: NAPI context
3b582cc1
SH
2068 * @poll: polling function
2069 * @weight: default weight
2070 *
5e82b4b2
BH
2071 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2072 * *any* of the other NAPI-related functions.
3b582cc1 2073 */
d565b0a1
HX
2074void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2075 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2076
d64b5e85 2077/**
5e82b4b2 2078 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2079 * @dev: network device
5e82b4b2 2080 * @napi: NAPI context
d64b5e85
ED
2081 * @poll: polling function
2082 * @weight: default weight
2083 *
2084 * This variant of netif_napi_add() should be used from drivers using NAPI
2085 * to exclusively poll a TX queue.
2086 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2087 */
2088static inline void netif_tx_napi_add(struct net_device *dev,
2089 struct napi_struct *napi,
2090 int (*poll)(struct napi_struct *, int),
2091 int weight)
2092{
2093 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2094 netif_napi_add(dev, napi, poll, weight);
2095}
2096
d8156534 2097/**
5e82b4b2
BH
2098 * netif_napi_del - remove a NAPI context
2099 * @napi: NAPI context
d8156534 2100 *
5e82b4b2 2101 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2102 */
d565b0a1
HX
2103void netif_napi_del(struct napi_struct *napi);
2104
2105struct napi_gro_cb {
78a478d0 2106 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2107 void *frag0;
78a478d0 2108
7489594c
HX
2109 /* Length of frag0. */
2110 unsigned int frag0_len;
2111
86911732 2112 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2113 int data_offset;
86911732 2114
d565b0a1 2115 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2116 u16 flush;
2117
2118 /* Save the IP ID here and check when we get to the transport layer */
2119 u16 flush_id;
d565b0a1
HX
2120
2121 /* Number of segments aggregated. */
2e71a6f8
ED
2122 u16 count;
2123
15e2396d
TH
2124 /* Start offset for remote checksum offload */
2125 u16 gro_remcsum_start;
2126
2e71a6f8
ED
2127 /* jiffies when first packet was created/queued */
2128 unsigned long age;
86347245 2129
afe93325 2130 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2131 u16 proto;
2132
baa32ff4
TH
2133 /* This is non-zero if the packet may be of the same flow. */
2134 u8 same_flow:1;
2135
fac8e0f5
JG
2136 /* Used in tunnel GRO receive */
2137 u8 encap_mark:1;
573e8fca
TH
2138
2139 /* GRO checksum is valid */
2140 u8 csum_valid:1;
2141
662880f4
TH
2142 /* Number of checksums via CHECKSUM_UNNECESSARY */
2143 u8 csum_cnt:3;
c3c7c254 2144
baa32ff4
TH
2145 /* Free the skb? */
2146 u8 free:2;
2147#define NAPI_GRO_FREE 1
2148#define NAPI_GRO_FREE_STOLEN_HEAD 2
2149
efc98d08
TH
2150 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2151 u8 is_ipv6:1;
2152
a0ca153f
AD
2153 /* Used in GRE, set in fou/gue_gro_receive */
2154 u8 is_fou:1;
2155
1530545e
AD
2156 /* Used to determine if flush_id can be ignored */
2157 u8 is_atomic:1;
2158
fcd91dd4
SD
2159 /* Number of gro_receive callbacks this packet already went through */
2160 u8 recursion_counter:4;
2161
2162 /* 1 bit hole */
baa32ff4 2163
bf5a755f
JC
2164 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2165 __wsum csum;
2166
c3c7c254
ED
2167 /* used in skb_gro_receive() slow path */
2168 struct sk_buff *last;
d565b0a1
HX
2169};
2170
2171#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2172
fcd91dd4
SD
2173#define GRO_RECURSION_LIMIT 15
2174static inline int gro_recursion_inc_test(struct sk_buff *skb)
2175{
2176 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2177}
2178
2179typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2180static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2181 struct sk_buff **head,
2182 struct sk_buff *skb)
2183{
2184 if (unlikely(gro_recursion_inc_test(skb))) {
2185 NAPI_GRO_CB(skb)->flush |= 1;
2186 return NULL;
2187 }
2188
2189 return cb(head, skb);
2190}
2191
2192typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2193 struct sk_buff *);
2194static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2195 struct sock *sk,
2196 struct sk_buff **head,
2197 struct sk_buff *skb)
2198{
2199 if (unlikely(gro_recursion_inc_test(skb))) {
2200 NAPI_GRO_CB(skb)->flush |= 1;
2201 return NULL;
2202 }
2203
2204 return cb(sk, head, skb);
2205}
2206
1da177e4 2207struct packet_type {
f2ccd8fa
DM
2208 __be16 type; /* This is really htons(ether_type). */
2209 struct net_device *dev; /* NULL is wildcarded here */
2210 int (*func) (struct sk_buff *,
2211 struct net_device *,
2212 struct packet_type *,
2213 struct net_device *);
c0de08d0
EL
2214 bool (*id_match)(struct packet_type *ptype,
2215 struct sock *sk);
1da177e4
LT
2216 void *af_packet_priv;
2217 struct list_head list;
2218};
2219
f191a1d1 2220struct offload_callbacks {
576a30eb 2221 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2222 netdev_features_t features);
d565b0a1 2223 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2224 struct sk_buff *skb);
299603e8 2225 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2226};
2227
2228struct packet_offload {
2229 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2230 u16 priority;
f191a1d1
VY
2231 struct offload_callbacks callbacks;
2232 struct list_head list;
1da177e4
LT
2233};
2234
5e82b4b2 2235/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2236struct pcpu_sw_netstats {
2237 u64 rx_packets;
2238 u64 rx_bytes;
2239 u64 tx_packets;
2240 u64 tx_bytes;
2241 struct u64_stats_sync syncp;
2242};
2243
aabc92bb
PNA
2244#define __netdev_alloc_pcpu_stats(type, gfp) \
2245({ \
2246 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2247 if (pcpu_stats) { \
2248 int __cpu; \
2249 for_each_possible_cpu(__cpu) { \
2250 typeof(type) *stat; \
2251 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2252 u64_stats_init(&stat->syncp); \
2253 } \
2254 } \
2255 pcpu_stats; \
1c213bd2
WC
2256})
2257
aabc92bb 2258#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2259 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2260
764f5e54
JP
2261enum netdev_lag_tx_type {
2262 NETDEV_LAG_TX_TYPE_UNKNOWN,
2263 NETDEV_LAG_TX_TYPE_RANDOM,
2264 NETDEV_LAG_TX_TYPE_BROADCAST,
2265 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2266 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2267 NETDEV_LAG_TX_TYPE_HASH,
2268};
2269
2270struct netdev_lag_upper_info {
2271 enum netdev_lag_tx_type tx_type;
2272};
2273
fb1b2e3c
JP
2274struct netdev_lag_lower_state_info {
2275 u8 link_up : 1,
2276 tx_enabled : 1;
2277};
2278
1da177e4
LT
2279#include <linux/notifier.h>
2280
dcfe1421
AW
2281/* netdevice notifier chain. Please remember to update the rtnetlink
2282 * notification exclusion list in rtnetlink_event() when adding new
2283 * types.
2284 */
2285#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2286#define NETDEV_DOWN 0x0002
2287#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2288 detected a hardware crash and restarted
2289 - we can use this eg to kick tcp sessions
2290 once done */
2291#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2292#define NETDEV_REGISTER 0x0005
2293#define NETDEV_UNREGISTER 0x0006
1d486bfb 2294#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2295#define NETDEV_CHANGEADDR 0x0008
2296#define NETDEV_GOING_DOWN 0x0009
2297#define NETDEV_CHANGENAME 0x000A
2298#define NETDEV_FEAT_CHANGE 0x000B
2299#define NETDEV_BONDING_FAILOVER 0x000C
2300#define NETDEV_PRE_UP 0x000D
2301#define NETDEV_PRE_TYPE_CHANGE 0x000E
2302#define NETDEV_POST_TYPE_CHANGE 0x000F
2303#define NETDEV_POST_INIT 0x0010
0115e8e3 2304#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2305#define NETDEV_RELEASE 0x0012
2306#define NETDEV_NOTIFY_PEERS 0x0013
2307#define NETDEV_JOIN 0x0014
42e52bf9 2308#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2309#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2310#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2311#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2312#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2313#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2314#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2315#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2316#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2317
f629d208
JP
2318int register_netdevice_notifier(struct notifier_block *nb);
2319int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2320
2321struct netdev_notifier_info {
2322 struct net_device *dev;
2323};
2324
be9efd36
JP
2325struct netdev_notifier_change_info {
2326 struct netdev_notifier_info info; /* must be first */
2327 unsigned int flags_changed;
2328};
2329
0e4ead9d
JP
2330struct netdev_notifier_changeupper_info {
2331 struct netdev_notifier_info info; /* must be first */
2332 struct net_device *upper_dev; /* new upper dev */
2333 bool master; /* is upper dev master */
5e82b4b2 2334 bool linking; /* is the notification for link or unlink */
29bf24af 2335 void *upper_info; /* upper dev info */
0e4ead9d
JP
2336};
2337
04d48266
JP
2338struct netdev_notifier_changelowerstate_info {
2339 struct netdev_notifier_info info; /* must be first */
2340 void *lower_state_info; /* is lower dev state */
2341};
2342
75538c2b
CW
2343static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2344 struct net_device *dev)
2345{
2346 info->dev = dev;
2347}
2348
351638e7
JP
2349static inline struct net_device *
2350netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2351{
2352 return info->dev;
2353}
2354
f629d208 2355int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2356
2357
1da177e4
LT
2358extern rwlock_t dev_base_lock; /* Device list lock */
2359
881d966b
EB
2360#define for_each_netdev(net, d) \
2361 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2362#define for_each_netdev_reverse(net, d) \
2363 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2364#define for_each_netdev_rcu(net, d) \
2365 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2366#define for_each_netdev_safe(net, d, n) \
2367 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2368#define for_each_netdev_continue(net, d) \
2369 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2370#define for_each_netdev_continue_rcu(net, d) \
2371 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2372#define for_each_netdev_in_bond_rcu(bond, slave) \
2373 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2374 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2375#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2376
a050c33f
DL
2377static inline struct net_device *next_net_device(struct net_device *dev)
2378{
2379 struct list_head *lh;
2380 struct net *net;
2381
c346dca1 2382 net = dev_net(dev);
a050c33f
DL
2383 lh = dev->dev_list.next;
2384 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2385}
2386
ce81b76a
ED
2387static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2388{
2389 struct list_head *lh;
2390 struct net *net;
2391
2392 net = dev_net(dev);
ccf43438 2393 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2394 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2395}
2396
a050c33f
DL
2397static inline struct net_device *first_net_device(struct net *net)
2398{
2399 return list_empty(&net->dev_base_head) ? NULL :
2400 net_device_entry(net->dev_base_head.next);
2401}
7562f876 2402
ccf43438
ED
2403static inline struct net_device *first_net_device_rcu(struct net *net)
2404{
2405 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2406
2407 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2408}
2409
f629d208
JP
2410int netdev_boot_setup_check(struct net_device *dev);
2411unsigned long netdev_boot_base(const char *prefix, int unit);
2412struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2413 const char *hwaddr);
2414struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2415struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2416void dev_add_pack(struct packet_type *pt);
2417void dev_remove_pack(struct packet_type *pt);
2418void __dev_remove_pack(struct packet_type *pt);
2419void dev_add_offload(struct packet_offload *po);
2420void dev_remove_offload(struct packet_offload *po);
f629d208 2421
a54acb3a 2422int dev_get_iflink(const struct net_device *dev);
fc4099f1 2423int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2424struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2425 unsigned short mask);
f629d208
JP
2426struct net_device *dev_get_by_name(struct net *net, const char *name);
2427struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2428struct net_device *__dev_get_by_name(struct net *net, const char *name);
2429int dev_alloc_name(struct net_device *dev, const char *name);
2430int dev_open(struct net_device *dev);
2431int dev_close(struct net_device *dev);
99c4a26a 2432int dev_close_many(struct list_head *head, bool unlink);
f629d208 2433void dev_disable_lro(struct net_device *dev);
0c4b51f0 2434int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2435int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2436int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2437int register_netdevice(struct net_device *dev);
2438void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2439void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2440static inline void unregister_netdevice(struct net_device *dev)
2441{
2442 unregister_netdevice_queue(dev, NULL);
2443}
2444
f629d208
JP
2445int netdev_refcnt_read(const struct net_device *dev);
2446void free_netdev(struct net_device *dev);
74d332c1 2447void netdev_freemem(struct net_device *dev);
f629d208
JP
2448void synchronize_net(void);
2449int init_dummy_netdev(struct net_device *dev);
937f1ba5 2450
f60e5990 2451DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2452#define XMIT_RECURSION_LIMIT 10
2453
f60e5990 2454static inline int dev_recursion_level(void)
2455{
2456 return this_cpu_read(xmit_recursion);
2457}
2458
f629d208
JP
2459struct net_device *dev_get_by_index(struct net *net, int ifindex);
2460struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2461struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2462struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2463int netdev_get_name(struct net *net, char *name, int ifindex);
2464int dev_restart(struct net_device *dev);
f629d208 2465int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2466
2467static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2468{
2469 return NAPI_GRO_CB(skb)->data_offset;
2470}
2471
2472static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2473{
2474 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2475}
2476
2477static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2478{
2479 NAPI_GRO_CB(skb)->data_offset += len;
2480}
2481
a5b1cf28
HX
2482static inline void *skb_gro_header_fast(struct sk_buff *skb,
2483 unsigned int offset)
86911732 2484{
a5b1cf28
HX
2485 return NAPI_GRO_CB(skb)->frag0 + offset;
2486}
78a478d0 2487
a5b1cf28
HX
2488static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2489{
2490 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2491}
78a478d0 2492
57ea52a8
HX
2493static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2494{
2495 NAPI_GRO_CB(skb)->frag0 = NULL;
2496 NAPI_GRO_CB(skb)->frag0_len = 0;
2497}
2498
a5b1cf28
HX
2499static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2500 unsigned int offset)
2501{
17dd759c
HX
2502 if (!pskb_may_pull(skb, hlen))
2503 return NULL;
2504
57ea52a8 2505 skb_gro_frag0_invalidate(skb);
17dd759c 2506 return skb->data + offset;
86911732 2507}
1da177e4 2508
36e7b1b8
HX
2509static inline void *skb_gro_network_header(struct sk_buff *skb)
2510{
78d3fd0b
HX
2511 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2512 skb_network_offset(skb);
36e7b1b8
HX
2513}
2514
bf5a755f
JC
2515static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2516 const void *start, unsigned int len)
2517{
573e8fca 2518 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2519 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2520 csum_partial(start, len, 0));
2521}
2522
573e8fca
TH
2523/* GRO checksum functions. These are logical equivalents of the normal
2524 * checksum functions (in skbuff.h) except that they operate on the GRO
2525 * offsets and fields in sk_buff.
2526 */
2527
2528__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2529
15e2396d
TH
2530static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2531{
b7fe10e5 2532 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2533}
2534
573e8fca
TH
2535static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2536 bool zero_okay,
2537 __sum16 check)
2538{
6edec0e6
TH
2539 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2540 skb_checksum_start_offset(skb) <
2541 skb_gro_offset(skb)) &&
15e2396d 2542 !skb_at_gro_remcsum_start(skb) &&
662880f4 2543 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2544 (!zero_okay || check));
2545}
2546
2547static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2548 __wsum psum)
2549{
2550 if (NAPI_GRO_CB(skb)->csum_valid &&
2551 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2552 return 0;
2553
2554 NAPI_GRO_CB(skb)->csum = psum;
2555
2556 return __skb_gro_checksum_complete(skb);
2557}
2558
573e8fca
TH
2559static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2560{
662880f4
TH
2561 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2562 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2563 NAPI_GRO_CB(skb)->csum_cnt--;
2564 } else {
2565 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2566 * verified a new top level checksum or an encapsulated one
2567 * during GRO. This saves work if we fallback to normal path.
2568 */
2569 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2570 }
2571}
2572
2573#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2574 compute_pseudo) \
2575({ \
2576 __sum16 __ret = 0; \
2577 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2578 __ret = __skb_gro_checksum_validate_complete(skb, \
2579 compute_pseudo(skb, proto)); \
219f1d79 2580 if (!__ret) \
573e8fca
TH
2581 skb_gro_incr_csum_unnecessary(skb); \
2582 __ret; \
2583})
2584
2585#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2586 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2587
2588#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2589 compute_pseudo) \
2590 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2591
2592#define skb_gro_checksum_simple_validate(skb) \
2593 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2594
d96535a1
TH
2595static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2596{
2597 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2598 !NAPI_GRO_CB(skb)->csum_valid);
2599}
2600
2601static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2602 __sum16 check, __wsum pseudo)
2603{
2604 NAPI_GRO_CB(skb)->csum = ~pseudo;
2605 NAPI_GRO_CB(skb)->csum_valid = 1;
2606}
2607
2608#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2609do { \
2610 if (__skb_gro_checksum_convert_check(skb)) \
2611 __skb_gro_checksum_convert(skb, check, \
2612 compute_pseudo(skb, proto)); \
2613} while (0)
2614
26c4f7da
TH
2615struct gro_remcsum {
2616 int offset;
2617 __wsum delta;
2618};
2619
2620static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2621{
846cd667 2622 grc->offset = 0;
26c4f7da
TH
2623 grc->delta = 0;
2624}
2625
b7fe10e5
TH
2626static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2627 unsigned int off, size_t hdrlen,
2628 int start, int offset,
2629 struct gro_remcsum *grc,
2630 bool nopartial)
dcdc8994
TH
2631{
2632 __wsum delta;
b7fe10e5 2633 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2634
2635 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2636
15e2396d 2637 if (!nopartial) {
b7fe10e5
TH
2638 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2639 return ptr;
2640 }
2641
2642 ptr = skb_gro_header_fast(skb, off);
2643 if (skb_gro_header_hard(skb, off + plen)) {
2644 ptr = skb_gro_header_slow(skb, off + plen, off);
2645 if (!ptr)
2646 return NULL;
15e2396d
TH
2647 }
2648
b7fe10e5
TH
2649 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2650 start, offset);
dcdc8994
TH
2651
2652 /* Adjust skb->csum since we changed the packet */
dcdc8994 2653 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2654
b7fe10e5 2655 grc->offset = off + hdrlen + offset;
26c4f7da 2656 grc->delta = delta;
b7fe10e5
TH
2657
2658 return ptr;
dcdc8994
TH
2659}
2660
26c4f7da
TH
2661static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2662 struct gro_remcsum *grc)
2663{
b7fe10e5
TH
2664 void *ptr;
2665 size_t plen = grc->offset + sizeof(u16);
2666
26c4f7da
TH
2667 if (!grc->delta)
2668 return;
2669
b7fe10e5
TH
2670 ptr = skb_gro_header_fast(skb, grc->offset);
2671 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2672 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2673 if (!ptr)
2674 return;
2675 }
2676
2677 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2678}
dcdc8994 2679
25393d3f
SK
2680#ifdef CONFIG_XFRM_OFFLOAD
2681static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2682{
2683 if (PTR_ERR(pp) != -EINPROGRESS)
2684 NAPI_GRO_CB(skb)->flush |= flush;
2685}
2686#else
5f114163
SK
2687static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2688{
2689 NAPI_GRO_CB(skb)->flush |= flush;
2690}
25393d3f 2691#endif
5f114163 2692
0c4e8581
SH
2693static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2694 unsigned short type,
3b04ddde 2695 const void *daddr, const void *saddr,
95c96174 2696 unsigned int len)
0c4e8581 2697{
f1ecfd5d 2698 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2699 return 0;
3b04ddde
SH
2700
2701 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2702}
2703
b95cce35
SH
2704static inline int dev_parse_header(const struct sk_buff *skb,
2705 unsigned char *haddr)
2706{
2707 const struct net_device *dev = skb->dev;
2708
1b83336b 2709 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2710 return 0;
3b04ddde 2711 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2712}
2713
2793a23a
WB
2714/* ll_header must have at least hard_header_len allocated */
2715static inline bool dev_validate_header(const struct net_device *dev,
2716 char *ll_header, int len)
2717{
2718 if (likely(len >= dev->hard_header_len))
2719 return true;
217e6fa2
WB
2720 if (len < dev->min_header_len)
2721 return false;
2793a23a
WB
2722
2723 if (capable(CAP_SYS_RAWIO)) {
2724 memset(ll_header + len, 0, dev->hard_header_len - len);
2725 return true;
2726 }
2727
2728 if (dev->header_ops && dev->header_ops->validate)
2729 return dev->header_ops->validate(ll_header, len);
2730
2731 return false;
2732}
2733
1da177e4 2734typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2735int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2736static inline int unregister_gifconf(unsigned int family)
2737{
2738 return register_gifconf(family, NULL);
2739}
2740
99bbc707 2741#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2742#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2743struct sd_flow_limit {
2744 u64 count;
2745 unsigned int num_buckets;
2746 unsigned int history_head;
2747 u16 history[FLOW_LIMIT_HISTORY];
2748 u8 buckets[];
2749};
2750
2751extern int netdev_flow_limit_table_len;
2752#endif /* CONFIG_NET_FLOW_LIMIT */
2753
1da177e4 2754/*
5e82b4b2 2755 * Incoming packets are placed on per-CPU queues
1da177e4 2756 */
d94d9fee 2757struct softnet_data {
1da177e4 2758 struct list_head poll_list;
6e7676c1 2759 struct sk_buff_head process_queue;
1da177e4 2760
dee42870 2761 /* stats */
cd7b5396
DM
2762 unsigned int processed;
2763 unsigned int time_squeeze;
cd7b5396 2764 unsigned int received_rps;
fd793d89 2765#ifdef CONFIG_RPS
88751275 2766 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2767#endif
2768#ifdef CONFIG_NET_FLOW_LIMIT
2769 struct sd_flow_limit __rcu *flow_limit;
2770#endif
2771 struct Qdisc *output_queue;
2772 struct Qdisc **output_queue_tailp;
2773 struct sk_buff *completion_queue;
88751275 2774
4cdb1e2e 2775#ifdef CONFIG_RPS
501e7ef5
ED
2776 /* input_queue_head should be written by cpu owning this struct,
2777 * and only read by other cpus. Worth using a cache line.
2778 */
2779 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2780
2781 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2782 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2783 struct softnet_data *rps_ipi_next;
2784 unsigned int cpu;
76cc8b13 2785 unsigned int input_queue_tail;
1e94d72f 2786#endif
95c96174 2787 unsigned int dropped;
0a9627f2 2788 struct sk_buff_head input_pkt_queue;
bea3348e 2789 struct napi_struct backlog;
99bbc707 2790
1da177e4
LT
2791};
2792
76cc8b13 2793static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2794{
2795#ifdef CONFIG_RPS
76cc8b13
TH
2796 sd->input_queue_head++;
2797#endif
2798}
2799
2800static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2801 unsigned int *qtail)
2802{
2803#ifdef CONFIG_RPS
2804 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2805#endif
2806}
2807
0a9627f2 2808DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2809
f629d208 2810void __netif_schedule(struct Qdisc *q);
46e5da40 2811void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2812
fd2ea0a7
DM
2813static inline void netif_tx_schedule_all(struct net_device *dev)
2814{
2815 unsigned int i;
2816
2817 for (i = 0; i < dev->num_tx_queues; i++)
2818 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2819}
2820
f9a7cbbf 2821static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2822{
73466498 2823 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2824}
2825
bea3348e
SH
2826/**
2827 * netif_start_queue - allow transmit
2828 * @dev: network device
2829 *
2830 * Allow upper layers to call the device hard_start_xmit routine.
2831 */
1da177e4
LT
2832static inline void netif_start_queue(struct net_device *dev)
2833{
e8a0464c 2834 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2835}
2836
fd2ea0a7
DM
2837static inline void netif_tx_start_all_queues(struct net_device *dev)
2838{
2839 unsigned int i;
2840
2841 for (i = 0; i < dev->num_tx_queues; i++) {
2842 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2843 netif_tx_start_queue(txq);
2844 }
2845}
2846
46e5da40 2847void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2848
d29f749e
DJ
2849/**
2850 * netif_wake_queue - restart transmit
2851 * @dev: network device
2852 *
2853 * Allow upper layers to call the device hard_start_xmit routine.
2854 * Used for flow control when transmit resources are available.
2855 */
79d16385
DM
2856static inline void netif_wake_queue(struct net_device *dev)
2857{
e8a0464c 2858 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2859}
2860
fd2ea0a7
DM
2861static inline void netif_tx_wake_all_queues(struct net_device *dev)
2862{
2863 unsigned int i;
2864
2865 for (i = 0; i < dev->num_tx_queues; i++) {
2866 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2867 netif_tx_wake_queue(txq);
2868 }
2869}
2870
f9a7cbbf 2871static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2872{
73466498 2873 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2874}
2875
bea3348e
SH
2876/**
2877 * netif_stop_queue - stop transmitted packets
2878 * @dev: network device
2879 *
2880 * Stop upper layers calling the device hard_start_xmit routine.
2881 * Used for flow control when transmit resources are unavailable.
2882 */
1da177e4
LT
2883static inline void netif_stop_queue(struct net_device *dev)
2884{
e8a0464c 2885 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2886}
2887
a2029240 2888void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2889
4d29515f 2890static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2891{
73466498 2892 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2893}
2894
bea3348e
SH
2895/**
2896 * netif_queue_stopped - test if transmit queue is flowblocked
2897 * @dev: network device
2898 *
2899 * Test if transmit queue on device is currently unable to send.
2900 */
4d29515f 2901static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2902{
e8a0464c 2903 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2904}
2905
4d29515f 2906static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2907{
73466498
TH
2908 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2909}
2910
8e2f1a63
DB
2911static inline bool
2912netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2913{
2914 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2915}
2916
8e2f1a63
DB
2917static inline bool
2918netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2919{
2920 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2921}
2922
53511453
ED
2923/**
2924 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2925 * @dev_queue: pointer to transmit queue
2926 *
2927 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2928 * to give appropriate hint to the CPU.
53511453
ED
2929 */
2930static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2931{
2932#ifdef CONFIG_BQL
2933 prefetchw(&dev_queue->dql.num_queued);
2934#endif
2935}
2936
2937/**
2938 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2939 * @dev_queue: pointer to transmit queue
2940 *
2941 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2942 * to give appropriate hint to the CPU.
53511453
ED
2943 */
2944static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2945{
2946#ifdef CONFIG_BQL
2947 prefetchw(&dev_queue->dql.limit);
2948#endif
2949}
2950
c5d67bd7
TH
2951static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2952 unsigned int bytes)
2953{
114cf580
TH
2954#ifdef CONFIG_BQL
2955 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2956
2957 if (likely(dql_avail(&dev_queue->dql) >= 0))
2958 return;
2959
2960 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2961
2962 /*
2963 * The XOFF flag must be set before checking the dql_avail below,
2964 * because in netdev_tx_completed_queue we update the dql_completed
2965 * before checking the XOFF flag.
2966 */
2967 smp_mb();
2968
2969 /* check again in case another CPU has just made room avail */
2970 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2971 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2972#endif
c5d67bd7
TH
2973}
2974
0042d0c8
FF
2975/**
2976 * netdev_sent_queue - report the number of bytes queued to hardware
2977 * @dev: network device
2978 * @bytes: number of bytes queued to the hardware device queue
2979 *
2980 * Report the number of bytes queued for sending/completion to the network
2981 * device hardware queue. @bytes should be a good approximation and should
2982 * exactly match netdev_completed_queue() @bytes
2983 */
c5d67bd7
TH
2984static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2985{
2986 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2987}
2988
2989static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2990 unsigned int pkts, unsigned int bytes)
c5d67bd7 2991{
114cf580 2992#ifdef CONFIG_BQL
b37c0fbe
AD
2993 if (unlikely(!bytes))
2994 return;
2995
2996 dql_completed(&dev_queue->dql, bytes);
2997
2998 /*
2999 * Without the memory barrier there is a small possiblity that
3000 * netdev_tx_sent_queue will miss the update and cause the queue to
3001 * be stopped forever
3002 */
3003 smp_mb();
3004
3005 if (dql_avail(&dev_queue->dql) < 0)
3006 return;
3007
3008 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3009 netif_schedule_queue(dev_queue);
114cf580 3010#endif
c5d67bd7
TH
3011}
3012
0042d0c8
FF
3013/**
3014 * netdev_completed_queue - report bytes and packets completed by device
3015 * @dev: network device
3016 * @pkts: actual number of packets sent over the medium
3017 * @bytes: actual number of bytes sent over the medium
3018 *
3019 * Report the number of bytes and packets transmitted by the network device
3020 * hardware queue over the physical medium, @bytes must exactly match the
3021 * @bytes amount passed to netdev_sent_queue()
3022 */
c5d67bd7 3023static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3024 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3025{
3026 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3027}
3028
3029static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3030{
114cf580 3031#ifdef CONFIG_BQL
5c490354 3032 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3033 dql_reset(&q->dql);
3034#endif
c5d67bd7
TH
3035}
3036
0042d0c8
FF
3037/**
3038 * netdev_reset_queue - reset the packets and bytes count of a network device
3039 * @dev_queue: network device
3040 *
3041 * Reset the bytes and packet count of a network device and clear the
3042 * software flow control OFF bit for this network device
3043 */
c5d67bd7
TH
3044static inline void netdev_reset_queue(struct net_device *dev_queue)
3045{
3046 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3047}
3048
b9507bda
DB
3049/**
3050 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3051 * @dev: network device
3052 * @queue_index: given tx queue index
3053 *
3054 * Returns 0 if given tx queue index >= number of device tx queues,
3055 * otherwise returns the originally passed tx queue index.
3056 */
3057static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3058{
3059 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3060 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3061 dev->name, queue_index,
3062 dev->real_num_tx_queues);
3063 return 0;
3064 }
3065
3066 return queue_index;
3067}
3068
bea3348e
SH
3069/**
3070 * netif_running - test if up
3071 * @dev: network device
3072 *
3073 * Test if the device has been brought up.
3074 */
4d29515f 3075static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3076{
3077 return test_bit(__LINK_STATE_START, &dev->state);
3078}
3079
f25f4e44 3080/*
5e82b4b2 3081 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3082 * stop, and a check if it's stopped. All other device management is
3083 * done at the overall netdevice level.
3084 * Also test the device if we're multiqueue.
3085 */
bea3348e
SH
3086
3087/**
3088 * netif_start_subqueue - allow sending packets on subqueue
3089 * @dev: network device
3090 * @queue_index: sub queue index
3091 *
3092 * Start individual transmit queue of a device with multiple transmit queues.
3093 */
f25f4e44
PWJ
3094static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3095{
fd2ea0a7 3096 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3097
3098 netif_tx_start_queue(txq);
f25f4e44
PWJ
3099}
3100
bea3348e
SH
3101/**
3102 * netif_stop_subqueue - stop sending packets on subqueue
3103 * @dev: network device
3104 * @queue_index: sub queue index
3105 *
3106 * Stop individual transmit queue of a device with multiple transmit queues.
3107 */
f25f4e44
PWJ
3108static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3109{
fd2ea0a7 3110 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3111 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3112}
3113
bea3348e
SH
3114/**
3115 * netif_subqueue_stopped - test status of subqueue
3116 * @dev: network device
3117 * @queue_index: sub queue index
3118 *
3119 * Check individual transmit queue of a device with multiple transmit queues.
3120 */
4d29515f
DM
3121static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3122 u16 queue_index)
f25f4e44 3123{
fd2ea0a7 3124 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3125
3126 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3127}
3128
4d29515f
DM
3129static inline bool netif_subqueue_stopped(const struct net_device *dev,
3130 struct sk_buff *skb)
668f895a
PE
3131{
3132 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3133}
bea3348e 3134
738b35cc
FF
3135/**
3136 * netif_wake_subqueue - allow sending packets on subqueue
3137 * @dev: network device
3138 * @queue_index: sub queue index
3139 *
3140 * Resume individual transmit queue of a device with multiple transmit queues.
3141 */
3142static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3143{
3144 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3145
3146 netif_tx_wake_queue(txq);
3147}
f25f4e44 3148
537c00de 3149#ifdef CONFIG_XPS
53af53ae 3150int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3151 u16 index);
537c00de
AD
3152#else
3153static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3154 const struct cpumask *mask,
537c00de
AD
3155 u16 index)
3156{
3157 return 0;
3158}
3159#endif
3160
5605c762
JP
3161u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3162 unsigned int num_tx_queues);
3163
a3d22a68
VZ
3164/*
3165 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3166 * as a distribution range limit for the returned value.
3167 */
3168static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3169 struct sk_buff *skb)
a3d22a68
VZ
3170{
3171 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3172}
3173
bea3348e
SH
3174/**
3175 * netif_is_multiqueue - test if device has multiple transmit queues
3176 * @dev: network device
3177 *
3178 * Check if device has multiple transmit queues
bea3348e 3179 */
4d29515f 3180static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3181{
a02cec21 3182 return dev->num_tx_queues > 1;
f25f4e44 3183}
1da177e4 3184
f629d208 3185int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3186
a953be53 3187#ifdef CONFIG_SYSFS
f629d208 3188int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3189#else
3190static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3191 unsigned int rxq)
3192{
3193 return 0;
3194}
3195#endif
3196
a953be53
MD
3197#ifdef CONFIG_SYSFS
3198static inline unsigned int get_netdev_rx_queue_index(
3199 struct netdev_rx_queue *queue)
3200{
3201 struct net_device *dev = queue->dev;
3202 int index = queue - dev->_rx;
3203
3204 BUG_ON(index >= dev->num_rx_queues);
3205 return index;
3206}
3207#endif
3208
16917b87 3209#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3210int netif_get_num_default_rss_queues(void);
16917b87 3211
e6247027
ED
3212enum skb_free_reason {
3213 SKB_REASON_CONSUMED,
3214 SKB_REASON_DROPPED,
3215};
3216
3217void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3218void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3219
e6247027
ED
3220/*
3221 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3222 * interrupt context or with hardware interrupts being disabled.
3223 * (in_irq() || irqs_disabled())
3224 *
3225 * We provide four helpers that can be used in following contexts :
3226 *
3227 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3228 * replacing kfree_skb(skb)
3229 *
3230 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3231 * Typically used in place of consume_skb(skb) in TX completion path
3232 *
3233 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3234 * replacing kfree_skb(skb)
3235 *
3236 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3237 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3238 */
e6247027
ED
3239static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3240{
3241 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3242}
3243
3244static inline void dev_consume_skb_irq(struct sk_buff *skb)
3245{
3246 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3247}
3248
3249static inline void dev_kfree_skb_any(struct sk_buff *skb)
3250{
3251 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3252}
3253
3254static inline void dev_consume_skb_any(struct sk_buff *skb)
3255{
3256 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3257}
1da177e4 3258
f629d208
JP
3259int netif_rx(struct sk_buff *skb);
3260int netif_rx_ni(struct sk_buff *skb);
04eb4489 3261int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3262gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3263void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3264struct sk_buff *napi_get_frags(struct napi_struct *napi);
3265gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3266struct packet_offload *gro_find_receive_by_type(__be16 type);
3267struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3268
3269static inline void napi_free_frags(struct napi_struct *napi)
3270{
3271 kfree_skb(napi->skb);
3272 napi->skb = NULL;
3273}
3274
24b27fc4 3275bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3276int netdev_rx_handler_register(struct net_device *dev,
3277 rx_handler_func_t *rx_handler,
3278 void *rx_handler_data);
3279void netdev_rx_handler_unregister(struct net_device *dev);
3280
3281bool dev_valid_name(const char *name);
3282int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3283int dev_ethtool(struct net *net, struct ifreq *);
3284unsigned int dev_get_flags(const struct net_device *);
3285int __dev_change_flags(struct net_device *, unsigned int flags);
3286int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3287void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3288 unsigned int gchanges);
f629d208
JP
3289int dev_change_name(struct net_device *, const char *);
3290int dev_set_alias(struct net_device *, const char *, size_t);
3291int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3292int dev_set_mtu(struct net_device *, int);
3293void dev_set_group(struct net_device *, int);
3294int dev_set_mac_address(struct net_device *, struct sockaddr *);
3295int dev_change_carrier(struct net_device *, bool new_carrier);
3296int dev_get_phys_port_id(struct net_device *dev,
02637fce 3297 struct netdev_phys_item_id *ppid);
db24a904
DA
3298int dev_get_phys_port_name(struct net_device *dev,
3299 char *name, size_t len);
d746d707 3300int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3301struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3302struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3303 struct netdev_queue *txq, int *ret);
d67b9cd2
DB
3304
3305typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3306int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3307 int fd, u32 flags);
58038695 3308bool __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
d67b9cd2 3309
a0265d28 3310int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3311int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3312bool is_skb_forwardable(const struct net_device *dev,
3313 const struct sk_buff *skb);
1da177e4 3314
4e3264d2
MKL
3315static __always_inline int ____dev_forward_skb(struct net_device *dev,
3316 struct sk_buff *skb)
3317{
3318 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3319 unlikely(!is_skb_forwardable(dev, skb))) {
3320 atomic_long_inc(&dev->rx_dropped);
3321 kfree_skb(skb);
3322 return NET_RX_DROP;
3323 }
3324
3325 skb_scrub_packet(skb, true);
3326 skb->priority = 0;
3327 return 0;
3328}
3329
74b20582
DA
3330void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3331
20380731 3332extern int netdev_budget;
7acf8a1e 3333extern unsigned int netdev_budget_usecs;
1da177e4
LT
3334
3335/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3336void netdev_run_todo(void);
1da177e4 3337
bea3348e
SH
3338/**
3339 * dev_put - release reference to device
3340 * @dev: network device
3341 *
9ef4429b 3342 * Release reference to device to allow it to be freed.
bea3348e 3343 */
1da177e4
LT
3344static inline void dev_put(struct net_device *dev)
3345{
933393f5 3346 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3347}
3348
bea3348e
SH
3349/**
3350 * dev_hold - get reference to device
3351 * @dev: network device
3352 *
9ef4429b 3353 * Hold reference to device to keep it from being freed.
bea3348e 3354 */
15333061
SH
3355static inline void dev_hold(struct net_device *dev)
3356{
933393f5 3357 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3358}
1da177e4
LT
3359
3360/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3361 * and _off may be called from IRQ context, but it is caller
3362 * who is responsible for serialization of these calls.
b00055aa
SR
3363 *
3364 * The name carrier is inappropriate, these functions should really be
3365 * called netif_lowerlayer_*() because they represent the state of any
3366 * kind of lower layer not just hardware media.
1da177e4
LT
3367 */
3368
f629d208
JP
3369void linkwatch_init_dev(struct net_device *dev);
3370void linkwatch_fire_event(struct net_device *dev);
3371void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3372
bea3348e
SH
3373/**
3374 * netif_carrier_ok - test if carrier present
3375 * @dev: network device
3376 *
3377 * Check if carrier is present on device
3378 */
4d29515f 3379static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3380{
3381 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3382}
3383
f629d208 3384unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3385
f629d208 3386void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3387
f629d208 3388void netif_carrier_on(struct net_device *dev);
1da177e4 3389
f629d208 3390void netif_carrier_off(struct net_device *dev);
1da177e4 3391
bea3348e
SH
3392/**
3393 * netif_dormant_on - mark device as dormant.
3394 * @dev: network device
3395 *
3396 * Mark device as dormant (as per RFC2863).
3397 *
3398 * The dormant state indicates that the relevant interface is not
3399 * actually in a condition to pass packets (i.e., it is not 'up') but is
3400 * in a "pending" state, waiting for some external event. For "on-
3401 * demand" interfaces, this new state identifies the situation where the
3402 * interface is waiting for events to place it in the up state.
bea3348e 3403 */
b00055aa
SR
3404static inline void netif_dormant_on(struct net_device *dev)
3405{
3406 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3407 linkwatch_fire_event(dev);
3408}
3409
bea3348e
SH
3410/**
3411 * netif_dormant_off - set device as not dormant.
3412 * @dev: network device
3413 *
3414 * Device is not in dormant state.
3415 */
b00055aa
SR
3416static inline void netif_dormant_off(struct net_device *dev)
3417{
3418 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3419 linkwatch_fire_event(dev);
3420}
3421
bea3348e 3422/**
8ecbc40a 3423 * netif_dormant - test if device is dormant
bea3348e
SH
3424 * @dev: network device
3425 *
8ecbc40a 3426 * Check if device is dormant.
bea3348e 3427 */
4d29515f 3428static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3429{
3430 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3431}
3432
3433
bea3348e
SH
3434/**
3435 * netif_oper_up - test if device is operational
3436 * @dev: network device
3437 *
3438 * Check if carrier is operational
3439 */
4d29515f 3440static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3441{
b00055aa
SR
3442 return (dev->operstate == IF_OPER_UP ||
3443 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3444}
3445
bea3348e
SH
3446/**
3447 * netif_device_present - is device available or removed
3448 * @dev: network device
3449 *
3450 * Check if device has not been removed from system.
3451 */
4d29515f 3452static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3453{
3454 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3455}
3456
f629d208 3457void netif_device_detach(struct net_device *dev);
1da177e4 3458
f629d208 3459void netif_device_attach(struct net_device *dev);
1da177e4
LT
3460
3461/*
3462 * Network interface message level settings
3463 */
1da177e4
LT
3464
3465enum {
3466 NETIF_MSG_DRV = 0x0001,
3467 NETIF_MSG_PROBE = 0x0002,
3468 NETIF_MSG_LINK = 0x0004,
3469 NETIF_MSG_TIMER = 0x0008,
3470 NETIF_MSG_IFDOWN = 0x0010,
3471 NETIF_MSG_IFUP = 0x0020,
3472 NETIF_MSG_RX_ERR = 0x0040,
3473 NETIF_MSG_TX_ERR = 0x0080,
3474 NETIF_MSG_TX_QUEUED = 0x0100,
3475 NETIF_MSG_INTR = 0x0200,
3476 NETIF_MSG_TX_DONE = 0x0400,
3477 NETIF_MSG_RX_STATUS = 0x0800,
3478 NETIF_MSG_PKTDATA = 0x1000,
3479 NETIF_MSG_HW = 0x2000,
3480 NETIF_MSG_WOL = 0x4000,
3481};
3482
3483#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3484#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3485#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3486#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3487#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3488#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3489#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3490#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3491#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3492#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3493#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3494#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3495#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3496#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3497#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3498
3499static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3500{
3501 /* use default */
3502 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3503 return default_msg_enable_bits;
3504 if (debug_value == 0) /* no output */
3505 return 0;
3506 /* set low N bits */
3507 return (1 << debug_value) - 1;
3508}
3509
c773e847 3510static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3511{
c773e847
DM
3512 spin_lock(&txq->_xmit_lock);
3513 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3514}
3515
5a717f4f
MT
3516static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3517{
3518 __acquire(&txq->_xmit_lock);
3519 return true;
3520}
3521
3522static inline void __netif_tx_release(struct netdev_queue *txq)
3523{
3524 __release(&txq->_xmit_lock);
3525}
3526
fd2ea0a7
DM
3527static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3528{
3529 spin_lock_bh(&txq->_xmit_lock);
3530 txq->xmit_lock_owner = smp_processor_id();
3531}
3532
4d29515f 3533static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3534{
4d29515f 3535 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3536 if (likely(ok))
3537 txq->xmit_lock_owner = smp_processor_id();
3538 return ok;
3539}
3540
3541static inline void __netif_tx_unlock(struct netdev_queue *txq)
3542{
3543 txq->xmit_lock_owner = -1;
3544 spin_unlock(&txq->_xmit_lock);
3545}
3546
3547static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3548{
3549 txq->xmit_lock_owner = -1;
3550 spin_unlock_bh(&txq->_xmit_lock);
3551}
3552
08baf561
ED
3553static inline void txq_trans_update(struct netdev_queue *txq)
3554{
3555 if (txq->xmit_lock_owner != -1)
3556 txq->trans_start = jiffies;
3557}
3558
ba162f8e
FW
3559/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3560static inline void netif_trans_update(struct net_device *dev)
3561{
9b36627a
FW
3562 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3563
3564 if (txq->trans_start != jiffies)
3565 txq->trans_start = jiffies;
ba162f8e
FW
3566}
3567
d29f749e
DJ
3568/**
3569 * netif_tx_lock - grab network device transmit lock
3570 * @dev: network device
d29f749e
DJ
3571 *
3572 * Get network device transmit lock
3573 */
22dd7495
JHS
3574static inline void netif_tx_lock(struct net_device *dev)
3575{
e8a0464c 3576 unsigned int i;
c3f26a26 3577 int cpu;
c773e847 3578
c3f26a26
DM
3579 spin_lock(&dev->tx_global_lock);
3580 cpu = smp_processor_id();
e8a0464c
DM
3581 for (i = 0; i < dev->num_tx_queues; i++) {
3582 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3583
3584 /* We are the only thread of execution doing a
3585 * freeze, but we have to grab the _xmit_lock in
3586 * order to synchronize with threads which are in
3587 * the ->hard_start_xmit() handler and already
3588 * checked the frozen bit.
3589 */
e8a0464c 3590 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3591 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3592 __netif_tx_unlock(txq);
e8a0464c 3593 }
932ff279
HX
3594}
3595
3596static inline void netif_tx_lock_bh(struct net_device *dev)
3597{
e8a0464c
DM
3598 local_bh_disable();
3599 netif_tx_lock(dev);
932ff279
HX
3600}
3601
932ff279
HX
3602static inline void netif_tx_unlock(struct net_device *dev)
3603{
e8a0464c
DM
3604 unsigned int i;
3605
3606 for (i = 0; i < dev->num_tx_queues; i++) {
3607 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3608
c3f26a26
DM
3609 /* No need to grab the _xmit_lock here. If the
3610 * queue is not stopped for another reason, we
3611 * force a schedule.
3612 */
3613 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3614 netif_schedule_queue(txq);
c3f26a26
DM
3615 }
3616 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3617}
3618
3619static inline void netif_tx_unlock_bh(struct net_device *dev)
3620{
e8a0464c
DM
3621 netif_tx_unlock(dev);
3622 local_bh_enable();
932ff279
HX
3623}
3624
c773e847 3625#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3626 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3627 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3628 } else { \
3629 __netif_tx_acquire(txq); \
22dd7495
JHS
3630 } \
3631}
3632
5efeac44
EB
3633#define HARD_TX_TRYLOCK(dev, txq) \
3634 (((dev->features & NETIF_F_LLTX) == 0) ? \
3635 __netif_tx_trylock(txq) : \
5a717f4f 3636 __netif_tx_acquire(txq))
5efeac44 3637
c773e847 3638#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3639 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3640 __netif_tx_unlock(txq); \
5a717f4f
MT
3641 } else { \
3642 __netif_tx_release(txq); \
22dd7495
JHS
3643 } \
3644}
3645
1da177e4
LT
3646static inline void netif_tx_disable(struct net_device *dev)
3647{
fd2ea0a7 3648 unsigned int i;
c3f26a26 3649 int cpu;
fd2ea0a7 3650
c3f26a26
DM
3651 local_bh_disable();
3652 cpu = smp_processor_id();
fd2ea0a7
DM
3653 for (i = 0; i < dev->num_tx_queues; i++) {
3654 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3655
3656 __netif_tx_lock(txq, cpu);
fd2ea0a7 3657 netif_tx_stop_queue(txq);
c3f26a26 3658 __netif_tx_unlock(txq);
fd2ea0a7 3659 }
c3f26a26 3660 local_bh_enable();
1da177e4
LT
3661}
3662
e308a5d8
DM
3663static inline void netif_addr_lock(struct net_device *dev)
3664{
3665 spin_lock(&dev->addr_list_lock);
3666}
3667
2429f7ac
JP
3668static inline void netif_addr_lock_nested(struct net_device *dev)
3669{
25175ba5
VY
3670 int subclass = SINGLE_DEPTH_NESTING;
3671
3672 if (dev->netdev_ops->ndo_get_lock_subclass)
3673 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3674
3675 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3676}
3677
e308a5d8
DM
3678static inline void netif_addr_lock_bh(struct net_device *dev)
3679{
3680 spin_lock_bh(&dev->addr_list_lock);
3681}
3682
3683static inline void netif_addr_unlock(struct net_device *dev)
3684{
3685 spin_unlock(&dev->addr_list_lock);
3686}
3687
3688static inline void netif_addr_unlock_bh(struct net_device *dev)
3689{
3690 spin_unlock_bh(&dev->addr_list_lock);
3691}
3692
f001fde5 3693/*
31278e71 3694 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3695 * rcu_read_lock held.
3696 */
3697#define for_each_dev_addr(dev, ha) \
31278e71 3698 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3699
1da177e4
LT
3700/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3701
f629d208 3702void ether_setup(struct net_device *dev);
1da177e4
LT
3703
3704/* Support for loadable net-drivers */
f629d208 3705struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3706 unsigned char name_assign_type,
f629d208
JP
3707 void (*setup)(struct net_device *),
3708 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3709#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3710 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3711
c835a677
TG
3712#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3713 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3714 count)
36909ea4 3715
f629d208
JP
3716int register_netdev(struct net_device *dev);
3717void unregister_netdev(struct net_device *dev);
f001fde5 3718
22bedad3 3719/* General hardware address lists handling functions */
f629d208
JP
3720int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3721 struct netdev_hw_addr_list *from_list, int addr_len);
3722void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3723 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3724int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3725 struct net_device *dev,
3726 int (*sync)(struct net_device *, const unsigned char *),
3727 int (*unsync)(struct net_device *,
3728 const unsigned char *));
3729void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3730 struct net_device *dev,
3731 int (*unsync)(struct net_device *,
3732 const unsigned char *));
f629d208 3733void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3734
f001fde5 3735/* Functions used for device addresses handling */
f629d208
JP
3736int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3737 unsigned char addr_type);
3738int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3739 unsigned char addr_type);
f629d208
JP
3740void dev_addr_flush(struct net_device *dev);
3741int dev_addr_init(struct net_device *dev);
a748ee24
JP
3742
3743/* Functions used for unicast addresses handling */
f629d208
JP
3744int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3745int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3746int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3747int dev_uc_sync(struct net_device *to, struct net_device *from);
3748int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3749void dev_uc_unsync(struct net_device *to, struct net_device *from);
3750void dev_uc_flush(struct net_device *dev);
3751void dev_uc_init(struct net_device *dev);
f001fde5 3752
670e5b8e
AD
3753/**
3754 * __dev_uc_sync - Synchonize device's unicast list
3755 * @dev: device to sync
3756 * @sync: function to call if address should be added
3757 * @unsync: function to call if address should be removed
3758 *
3759 * Add newly added addresses to the interface, and release
3760 * addresses that have been deleted.
5e82b4b2 3761 */
670e5b8e
AD
3762static inline int __dev_uc_sync(struct net_device *dev,
3763 int (*sync)(struct net_device *,
3764 const unsigned char *),
3765 int (*unsync)(struct net_device *,
3766 const unsigned char *))
3767{
3768 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3769}
3770
3771/**
e793c0f7 3772 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3773 * @dev: device to sync
3774 * @unsync: function to call if address should be removed
3775 *
3776 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3777 */
670e5b8e
AD
3778static inline void __dev_uc_unsync(struct net_device *dev,
3779 int (*unsync)(struct net_device *,
3780 const unsigned char *))
3781{
3782 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3783}
3784
22bedad3 3785/* Functions used for multicast addresses handling */
f629d208
JP
3786int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3787int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3788int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3789int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3790int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3791int dev_mc_sync(struct net_device *to, struct net_device *from);
3792int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3793void dev_mc_unsync(struct net_device *to, struct net_device *from);
3794void dev_mc_flush(struct net_device *dev);
3795void dev_mc_init(struct net_device *dev);
f001fde5 3796
670e5b8e
AD
3797/**
3798 * __dev_mc_sync - Synchonize device's multicast list
3799 * @dev: device to sync
3800 * @sync: function to call if address should be added
3801 * @unsync: function to call if address should be removed
3802 *
3803 * Add newly added addresses to the interface, and release
3804 * addresses that have been deleted.
5e82b4b2 3805 */
670e5b8e
AD
3806static inline int __dev_mc_sync(struct net_device *dev,
3807 int (*sync)(struct net_device *,
3808 const unsigned char *),
3809 int (*unsync)(struct net_device *,
3810 const unsigned char *))
3811{
3812 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3813}
3814
3815/**
e793c0f7 3816 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3817 * @dev: device to sync
3818 * @unsync: function to call if address should be removed
3819 *
3820 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3821 */
670e5b8e
AD
3822static inline void __dev_mc_unsync(struct net_device *dev,
3823 int (*unsync)(struct net_device *,
3824 const unsigned char *))
3825{
3826 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3827}
3828
4417da66 3829/* Functions used for secondary unicast and multicast support */
f629d208
JP
3830void dev_set_rx_mode(struct net_device *dev);
3831void __dev_set_rx_mode(struct net_device *dev);
3832int dev_set_promiscuity(struct net_device *dev, int inc);
3833int dev_set_allmulti(struct net_device *dev, int inc);
3834void netdev_state_change(struct net_device *dev);
3835void netdev_notify_peers(struct net_device *dev);
3836void netdev_features_change(struct net_device *dev);
1da177e4 3837/* Load a device via the kmod */
f629d208
JP
3838void dev_load(struct net *net, const char *name);
3839struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3840 struct rtnl_link_stats64 *storage);
3841void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3842 const struct net_device_stats *netdev_stats);
eeda3fd6 3843
1da177e4 3844extern int netdev_max_backlog;
3b098e2d 3845extern int netdev_tstamp_prequeue;
1da177e4 3846extern int weight_p;
3d48b53f
MT
3847extern int dev_weight_rx_bias;
3848extern int dev_weight_tx_bias;
3849extern int dev_rx_weight;
3850extern int dev_tx_weight;
9ff162a8 3851
f629d208 3852bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3853struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3854 struct list_head **iter);
f629d208
JP
3855struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3856 struct list_head **iter);
8b5be856 3857
44a40855
VY
3858/* iterate through upper list, must be called under RCU read lock */
3859#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3860 for (iter = &(dev)->adj_list.upper, \
3861 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3862 updev; \
3863 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3864
1a3f060c
DA
3865int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3866 int (*fn)(struct net_device *upper_dev,
3867 void *data),
3868 void *data);
3869
3870bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3871 struct net_device *upper_dev);
3872
f629d208
JP
3873void *netdev_lower_get_next_private(struct net_device *dev,
3874 struct list_head **iter);
3875void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3876 struct list_head **iter);
31088a11
VF
3877
3878#define netdev_for_each_lower_private(dev, priv, iter) \
3879 for (iter = (dev)->adj_list.lower.next, \
3880 priv = netdev_lower_get_next_private(dev, &(iter)); \
3881 priv; \
3882 priv = netdev_lower_get_next_private(dev, &(iter)))
3883
3884#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3885 for (iter = &(dev)->adj_list.lower, \
3886 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3887 priv; \
3888 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3889
4085ebe8
VY
3890void *netdev_lower_get_next(struct net_device *dev,
3891 struct list_head **iter);
7ce856aa 3892
4085ebe8 3893#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3894 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3895 ldev = netdev_lower_get_next(dev, &(iter)); \
3896 ldev; \
3897 ldev = netdev_lower_get_next(dev, &(iter)))
3898
7ce856aa
JP
3899struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3900 struct list_head **iter);
3901struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3902 struct list_head **iter);
3903
1a3f060c
DA
3904int netdev_walk_all_lower_dev(struct net_device *dev,
3905 int (*fn)(struct net_device *lower_dev,
3906 void *data),
3907 void *data);
3908int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3909 int (*fn)(struct net_device *lower_dev,
3910 void *data),
3911 void *data);
3912
f629d208 3913void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3914void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3915struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3916struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3917int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3918int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3919 struct net_device *upper_dev,
29bf24af 3920 void *upper_priv, void *upper_info);
f629d208
JP
3921void netdev_upper_dev_unlink(struct net_device *dev,
3922 struct net_device *upper_dev);
5bb025fa 3923void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3924void *netdev_lower_dev_get_private(struct net_device *dev,
3925 struct net_device *lower_dev);
04d48266
JP
3926void netdev_lower_state_changed(struct net_device *lower_dev,
3927 void *lower_state_info);
960fb622
ED
3928
3929/* RSS keys are 40 or 52 bytes long */
3930#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3931extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3932void netdev_rss_key_fill(void *buffer, size_t len);
3933
952fcfd0 3934int dev_get_nest_level(struct net_device *dev);
f629d208 3935int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 3936int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
3937int skb_csum_hwoffload_help(struct sk_buff *skb,
3938 const netdev_features_t features);
3939
f629d208
JP
3940struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3941 netdev_features_t features, bool tx_path);
3942struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3943 netdev_features_t features);
12b0004d 3944
61bd3857
MS
3945struct netdev_bonding_info {
3946 ifslave slave;
3947 ifbond master;
3948};
3949
3950struct netdev_notifier_bonding_info {
3951 struct netdev_notifier_info info; /* must be first */
3952 struct netdev_bonding_info bonding_info;
3953};
3954
3955void netdev_bonding_info_change(struct net_device *dev,
3956 struct netdev_bonding_info *bonding_info);
3957
12b0004d
CW
3958static inline
3959struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3960{
3961 return __skb_gso_segment(skb, features, true);
3962}
53d6471c 3963__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3964
3965static inline bool can_checksum_protocol(netdev_features_t features,
3966 __be16 protocol)
3967{
c8cd0989
TH
3968 if (protocol == htons(ETH_P_FCOE))
3969 return !!(features & NETIF_F_FCOE_CRC);
3970
3971 /* Assume this is an IP checksum (not SCTP CRC) */
3972
3973 if (features & NETIF_F_HW_CSUM) {
3974 /* Can checksum everything */
3975 return true;
3976 }
3977
3978 switch (protocol) {
3979 case htons(ETH_P_IP):
3980 return !!(features & NETIF_F_IP_CSUM);
3981 case htons(ETH_P_IPV6):
3982 return !!(features & NETIF_F_IPV6_CSUM);
3983 default:
3984 return false;
3985 }
ec5f0615 3986}
12b0004d 3987
fb286bb2 3988#ifdef CONFIG_BUG
f629d208 3989void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3990#else
3991static inline void netdev_rx_csum_fault(struct net_device *dev)
3992{
3993}
3994#endif
1da177e4 3995/* rx skb timestamps */
f629d208
JP
3996void net_enable_timestamp(void);
3997void net_disable_timestamp(void);
1da177e4 3998
20380731 3999#ifdef CONFIG_PROC_FS
f629d208 4000int __init dev_proc_init(void);
900ff8c6
CW
4001#else
4002#define dev_proc_init() 0
20380731
ACM
4003#endif
4004
4798248e 4005static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4006 struct sk_buff *skb, struct net_device *dev,
4007 bool more)
4798248e 4008{
fa2dbdc2 4009 skb->xmit_more = more ? 1 : 0;
0b725a2c 4010 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4011}
4012
10b3ad8c 4013static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4014 struct netdev_queue *txq, bool more)
4798248e
DM
4015{
4016 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4017 int rc;
4798248e 4018
fa2dbdc2 4019 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4020 if (rc == NETDEV_TX_OK)
4021 txq_trans_update(txq);
4022
4023 return rc;
4798248e
DM
4024}
4025
42a2d923
LT
4026int netdev_class_create_file_ns(struct class_attribute *class_attr,
4027 const void *ns);
4028void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4029 const void *ns);
58292cbe
TH
4030
4031static inline int netdev_class_create_file(struct class_attribute *class_attr)
4032{
4033 return netdev_class_create_file_ns(class_attr, NULL);
4034}
4035
4036static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4037{
4038 netdev_class_remove_file_ns(class_attr, NULL);
4039}
b8a9787e 4040
04600794
JB
4041extern struct kobj_ns_type_operations net_ns_type_operations;
4042
f629d208 4043const char *netdev_drivername(const struct net_device *dev);
6579e57b 4044
f629d208 4045void linkwatch_run_queue(void);
20380731 4046
da08143b
MK
4047static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4048 netdev_features_t f2)
4049{
c8cd0989
TH
4050 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4051 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4052 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4053 else
b6a0e72a 4054 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4055 }
da08143b 4056
c8cd0989 4057 return f1 & f2;
da08143b
MK
4058}
4059
c8f44aff
MM
4060static inline netdev_features_t netdev_get_wanted_features(
4061 struct net_device *dev)
5455c699
MM
4062{
4063 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4064}
c8f44aff
MM
4065netdev_features_t netdev_increment_features(netdev_features_t all,
4066 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4067
4068/* Allow TSO being used on stacked device :
4069 * Performing the GSO segmentation before last device
4070 * is a performance improvement.
4071 */
4072static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4073 netdev_features_t mask)
4074{
4075 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4076}
4077
6cb6a27c 4078int __netdev_update_features(struct net_device *dev);
5455c699 4079void netdev_update_features(struct net_device *dev);
afe12cc8 4080void netdev_change_features(struct net_device *dev);
7f353bf2 4081
fc4a7489
PM
4082void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4083 struct net_device *dev);
4084
e38f3025
TM
4085netdev_features_t passthru_features_check(struct sk_buff *skb,
4086 struct net_device *dev,
4087 netdev_features_t features);
c1e756bf 4088netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4089
4d29515f 4090static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4091{
7b748340 4092 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4093
4094 /* check flags correspondence */
4095 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4096 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4097 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4098 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4099 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4100 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4101 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4102 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4103 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4104 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4105 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4106 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4107 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4108 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4109 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4110 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4111 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0345e186 4112
d6b4991a 4113 return (features & feature) == feature;
576a30eb
HX
4114}
4115
4d29515f 4116static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4117{
278b2513 4118 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4119 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4120}
4121
8b86a61d 4122static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4123 netdev_features_t features)
7967168c 4124{
fc741216 4125 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4126 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4127 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4128}
4129
82cc1a7a
PWJ
4130static inline void netif_set_gso_max_size(struct net_device *dev,
4131 unsigned int size)
4132{
4133 dev->gso_max_size = size;
4134}
4135
7a7ffbab
WCC
4136static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4137 int pulled_hlen, u16 mac_offset,
4138 int mac_len)
4139{
4140 skb->protocol = protocol;
4141 skb->encapsulation = 1;
4142 skb_push(skb, pulled_hlen);
4143 skb_reset_transport_header(skb);
4144 skb->mac_header = mac_offset;
4145 skb->network_header = skb->mac_header + mac_len;
4146 skb->mac_len = mac_len;
4147}
4148
3c175784
SD
4149static inline bool netif_is_macsec(const struct net_device *dev)
4150{
4151 return dev->priv_flags & IFF_MACSEC;
4152}
4153
b618aaa9 4154static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4155{
4156 return dev->priv_flags & IFF_MACVLAN;
4157}
4158
b618aaa9 4159static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4160{
4161 return dev->priv_flags & IFF_MACVLAN_PORT;
4162}
4163
b618aaa9 4164static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4165{
4166 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4167}
4168
b618aaa9 4169static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4170{
4171 return dev->priv_flags & IFF_IPVLAN_MASTER;
4172}
4173
b618aaa9 4174static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4175{
4176 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4177}
4178
b618aaa9 4179static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4180{
4181 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4182}
4183
3bdc0eba
BG
4184static inline bool netif_supports_nofcs(struct net_device *dev)
4185{
4186 return dev->priv_flags & IFF_SUPP_NOFCS;
4187}
4188
007979ea 4189static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4190{
007979ea 4191 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4192}
4193
fee6d4c7
DA
4194static inline bool netif_is_l3_slave(const struct net_device *dev)
4195{
4196 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4197}
4198
0894ae3f
JP
4199static inline bool netif_is_bridge_master(const struct net_device *dev)
4200{
4201 return dev->priv_flags & IFF_EBRIDGE;
4202}
4203
28f9ee22
VY
4204static inline bool netif_is_bridge_port(const struct net_device *dev)
4205{
4206 return dev->priv_flags & IFF_BRIDGE_PORT;
4207}
4208
35d4e172
JP
4209static inline bool netif_is_ovs_master(const struct net_device *dev)
4210{
4211 return dev->priv_flags & IFF_OPENVSWITCH;
4212}
4213
5be66141
JP
4214static inline bool netif_is_ovs_port(const struct net_device *dev)
4215{
4216 return dev->priv_flags & IFF_OVS_DATAPATH;
4217}
4218
b618aaa9 4219static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4220{
4221 return dev->priv_flags & IFF_TEAM;
4222}
4223
b618aaa9 4224static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4225{
4226 return dev->priv_flags & IFF_TEAM_PORT;
4227}
4228
b618aaa9 4229static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4230{
4231 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4232}
4233
b618aaa9 4234static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4235{
4236 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4237}
4238
d4ab4286
KJ
4239static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4240{
4241 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4242}
4243
02875878
ED
4244/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4245static inline void netif_keep_dst(struct net_device *dev)
4246{
4247 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4248}
4249
18d3df3e
PA
4250/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4251static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4252{
4253 /* TODO: reserve and use an additional IFF bit, if we get more users */
4254 return dev->priv_flags & IFF_MACSEC;
4255}
4256
505d4f73 4257extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4258
571ba423
JP
4259/* Logging, debugging and troubleshooting/diagnostic helpers. */
4260
4261/* netdev_printk helpers, similar to dev_printk */
4262
4263static inline const char *netdev_name(const struct net_device *dev)
4264{
c6f854d5
VF
4265 if (!dev->name[0] || strchr(dev->name, '%'))
4266 return "(unnamed net_device)";
571ba423
JP
4267 return dev->name;
4268}
4269
8397ed36
DA
4270static inline bool netdev_unregistering(const struct net_device *dev)
4271{
4272 return dev->reg_state == NETREG_UNREGISTERING;
4273}
4274
ccc7f496
VF
4275static inline const char *netdev_reg_state(const struct net_device *dev)
4276{
4277 switch (dev->reg_state) {
4278 case NETREG_UNINITIALIZED: return " (uninitialized)";
4279 case NETREG_REGISTERED: return "";
4280 case NETREG_UNREGISTERING: return " (unregistering)";
4281 case NETREG_UNREGISTERED: return " (unregistered)";
4282 case NETREG_RELEASED: return " (released)";
4283 case NETREG_DUMMY: return " (dummy)";
4284 }
4285
4286 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4287 return " (unknown)";
4288}
4289
f629d208 4290__printf(3, 4)
6ea754eb
JP
4291void netdev_printk(const char *level, const struct net_device *dev,
4292 const char *format, ...);
f629d208 4293__printf(2, 3)
6ea754eb 4294void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4295__printf(2, 3)
6ea754eb 4296void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4297__printf(2, 3)
6ea754eb 4298void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4299__printf(2, 3)
6ea754eb 4300void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4301__printf(2, 3)
6ea754eb 4302void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4303__printf(2, 3)
6ea754eb 4304void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4305__printf(2, 3)
6ea754eb 4306void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4307
8909c9ad
VK
4308#define MODULE_ALIAS_NETDEV(device) \
4309 MODULE_ALIAS("netdev-" device)
4310
b558c96f 4311#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4312#define netdev_dbg(__dev, format, args...) \
4313do { \
ffa10cb4 4314 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4315} while (0)
b558c96f
JC
4316#elif defined(DEBUG)
4317#define netdev_dbg(__dev, format, args...) \
4318 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4319#else
4320#define netdev_dbg(__dev, format, args...) \
4321({ \
4322 if (0) \
4323 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4324})
4325#endif
4326
4327#if defined(VERBOSE_DEBUG)
4328#define netdev_vdbg netdev_dbg
4329#else
4330
4331#define netdev_vdbg(dev, format, args...) \
4332({ \
4333 if (0) \
4334 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4335 0; \
4336})
4337#endif
4338
4339/*
4340 * netdev_WARN() acts like dev_printk(), but with the key difference
4341 * of using a WARN/WARN_ON to get the message out, including the
4342 * file/line information and a backtrace.
4343 */
4344#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4345 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4346 netdev_reg_state(dev), ##args)
571ba423 4347
b3d95c5c
JP
4348/* netif printk helpers, similar to netdev_printk */
4349
4350#define netif_printk(priv, type, level, dev, fmt, args...) \
4351do { \
4352 if (netif_msg_##type(priv)) \
4353 netdev_printk(level, (dev), fmt, ##args); \
4354} while (0)
4355
f45f4321
JP
4356#define netif_level(level, priv, type, dev, fmt, args...) \
4357do { \
4358 if (netif_msg_##type(priv)) \
4359 netdev_##level(dev, fmt, ##args); \
4360} while (0)
4361
b3d95c5c 4362#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4363 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4364#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4365 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4366#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4367 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4368#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4369 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4370#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4371 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4372#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4373 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4374#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4375 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4376
0053ea9c 4377#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4378#define netif_dbg(priv, type, netdev, format, args...) \
4379do { \
4380 if (netif_msg_##type(priv)) \
b5fb0a03 4381 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4382} while (0)
0053ea9c
JP
4383#elif defined(DEBUG)
4384#define netif_dbg(priv, type, dev, format, args...) \
4385 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4386#else
4387#define netif_dbg(priv, type, dev, format, args...) \
4388({ \
4389 if (0) \
4390 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4391 0; \
4392})
4393#endif
4394
f617f276
EC
4395/* if @cond then downgrade to debug, else print at @level */
4396#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4397 do { \
4398 if (cond) \
4399 netif_dbg(priv, type, netdev, fmt, ##args); \
4400 else \
4401 netif_ ## level(priv, type, netdev, fmt, ##args); \
4402 } while (0)
4403
b3d95c5c 4404#if defined(VERBOSE_DEBUG)
bcfcc450 4405#define netif_vdbg netif_dbg
b3d95c5c
JP
4406#else
4407#define netif_vdbg(priv, type, dev, format, args...) \
4408({ \
4409 if (0) \
a4ed89cb 4410 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4411 0; \
4412})
4413#endif
571ba423 4414
900ff8c6
CW
4415/*
4416 * The list of packet types we will receive (as opposed to discard)
4417 * and the routines to invoke.
4418 *
4419 * Why 16. Because with 16 the only overlap we get on a hash of the
4420 * low nibble of the protocol value is RARP/SNAP/X.25.
4421 *
4422 * NOTE: That is no longer true with the addition of VLAN tags. Not
4423 * sure which should go first, but I bet it won't make much
4424 * difference if we are running VLANs. The good news is that
4425 * this protocol won't be in the list unless compiled in, so
4426 * the average user (w/out VLANs) will not be adversely affected.
4427 * --BLG
4428 *
4429 * 0800 IP
4430 * 8100 802.1Q VLAN
4431 * 0001 802.3
4432 * 0002 AX.25
4433 * 0004 802.2
4434 * 8035 RARP
4435 * 0005 SNAP
4436 * 0805 X.25
4437 * 0806 ARP
4438 * 8137 IPX
4439 * 0009 Localtalk
4440 * 86DD IPv6
4441 */
4442#define PTYPE_HASH_SIZE (16)
4443#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4444
385a154c 4445#endif /* _LINUX_NETDEVICE_H */