]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - include/linux/netdevice.h
xdp: pass XDP flags into install handlers
[mirror_ubuntu-hirsute-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
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 822 struct {
32d60277 823 u32 flags;
ddf9f970
JK
824 struct bpf_prog *prog;
825 struct netlink_ext_ack *extack;
826 };
a7862b45 827 /* XDP_QUERY_PROG */
58038695
MKL
828 struct {
829 bool prog_attached;
830 u32 prog_id;
831 };
a7862b45
BB
832 };
833};
16e5cc64 834
d77e38e6
SK
835#ifdef CONFIG_XFRM_OFFLOAD
836struct xfrmdev_ops {
837 int (*xdo_dev_state_add) (struct xfrm_state *x);
838 void (*xdo_dev_state_delete) (struct xfrm_state *x);
839 void (*xdo_dev_state_free) (struct xfrm_state *x);
840 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
841 struct xfrm_state *x);
842};
843#endif
844
d314774c
SH
845/*
846 * This structure defines the management hooks for network devices.
00829823
SH
847 * The following hooks can be defined; unless noted otherwise, they are
848 * optional and can be filled with a null pointer.
d314774c
SH
849 *
850 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
851 * This function is called once when a network device is registered.
852 * The network device can use this for any late stage initialization
853 * or semantic validation. It can fail with an error code which will
854 * be propagated back to register_netdev.
d314774c
SH
855 *
856 * void (*ndo_uninit)(struct net_device *dev);
857 * This function is called when device is unregistered or when registration
858 * fails. It is not called if init fails.
859 *
860 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 861 * This function is called when a network device transitions to the up
d314774c
SH
862 * state.
863 *
864 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 865 * This function is called when a network device transitions to the down
d314774c
SH
866 * state.
867 *
dc1f8bf6
SH
868 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
869 * struct net_device *dev);
00829823 870 * Called when a packet needs to be transmitted.
e79d8429
RR
871 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
872 * the queue before that can happen; it's for obsolete devices and weird
873 * corner cases, but the stack really does a non-trivial amount
874 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 875 * Required; cannot be NULL.
00829823 876 *
1a2a1444
DM
877 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
878 * struct net_device *dev
879 * netdev_features_t features);
880 * Called by core transmit path to determine if device is capable of
881 * performing offload operations on a given packet. This is to give
882 * the device an opportunity to implement any restrictions that cannot
883 * be otherwise expressed by feature flags. The check is called with
884 * the set of features that the stack has calculated and it returns
885 * those the driver believes to be appropriate.
cdba756f 886 *
f663dd9a 887 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 888 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 889 * Called to decide which queue to use when device supports multiple
00829823
SH
890 * transmit queues.
891 *
d314774c
SH
892 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
893 * This function is called to allow device receiver to make
5e82b4b2 894 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
895 *
896 * void (*ndo_set_rx_mode)(struct net_device *dev);
897 * This function is called device changes address list filtering.
01789349 898 * If driver handles unicast address filtering, it should set
5e82b4b2 899 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
900 *
901 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
902 * This function is called when the Media Access Control address
37b607c5 903 * needs to be changed. If this interface is not defined, the
5e82b4b2 904 * MAC address can not be changed.
d314774c
SH
905 *
906 * int (*ndo_validate_addr)(struct net_device *dev);
907 * Test if Media Access Control address is valid for the device.
908 *
909 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
910 * Called when a user requests an ioctl which can't be handled by
911 * the generic interface code. If not defined ioctls return
d314774c
SH
912 * not supported error code.
913 *
914 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
915 * Used to set network devices bus interface parameters. This interface
5e82b4b2 916 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
917 * interface (PCI) for low level management.
918 *
919 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
920 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 921 * of a device.
d314774c 922 *
00829823 923 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 924 * Callback used when the transmitter has not made any progress
d314774c
SH
925 * for dev->watchdog ticks.
926 *
bc1f4470 927 * void (*ndo_get_stats64)(struct net_device *dev,
928 * struct rtnl_link_stats64 *storage);
d308e38f 929 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 930 * Called when a user wants to get the network device usage
be1f3c2c 931 * statistics. Drivers must do one of the following:
3cfde79c
BH
932 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
933 * rtnl_link_stats64 structure passed by the caller.
82695d9b 934 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
935 * (which should normally be dev->stats) and return a pointer to
936 * it. The structure may be changed asynchronously only if each
937 * field is written atomically.
938 * 3. Update dev->stats asynchronously and atomically, and define
939 * neither operation.
d314774c 940 *
3df5b3c6 941 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
942 * Return true if this device supports offload stats of this attr_id.
943 *
944 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
945 * void *attr_data)
946 * Get statistics for offload operations by attr_id. Write it into the
947 * attr_data pointer.
948 *
5d632cb7 949 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 950 * If device supports VLAN filtering this function is called when a
80d5c368 951 * VLAN id is registered.
d314774c 952 *
5d632cb7 953 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 954 * If device supports VLAN filtering this function is called when a
80d5c368 955 * VLAN id is unregistered.
d314774c
SH
956 *
957 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
958 *
959 * SR-IOV management functions.
960 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
961 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
962 * u8 qos, __be16 proto);
ed616689
SC
963 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
964 * int max_tx_rate);
5f8444a3 965 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 966 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
967 * int (*ndo_get_vf_config)(struct net_device *dev,
968 * int vf, struct ifla_vf_info *ivf);
1d8faf48 969 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
970 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
971 * struct nlattr *port[]);
01a3d796
VZ
972 *
973 * Enable or disable the VF ability to query its RSS Redirection Table and
974 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 975 * with PF and querying it may introduce a theoretical security risk.
01a3d796 976 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 977 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
a5fcf8a6 978 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle, u32 chain_index,
6a4bc2b4
FF
979 * __be16 protocol, struct tc_to_netdev *tc);
980 * Called to setup any 'tc' scheduler, classifier or action on @dev.
981 * This is always called from the stack with the rtnl lock held and netif
982 * tx queues stopped. This allows the netdevice to perform queue
983 * management safely.
c445477d 984 *
e9bce845
YZ
985 * Fiber Channel over Ethernet (FCoE) offload functions.
986 * int (*ndo_fcoe_enable)(struct net_device *dev);
987 * Called when the FCoE protocol stack wants to start using LLD for FCoE
988 * so the underlying device can perform whatever needed configuration or
989 * initialization to support acceleration of FCoE traffic.
990 *
991 * int (*ndo_fcoe_disable)(struct net_device *dev);
992 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
993 * so the underlying device can perform whatever needed clean-ups to
994 * stop supporting acceleration of FCoE traffic.
995 *
996 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
997 * struct scatterlist *sgl, unsigned int sgc);
998 * Called when the FCoE Initiator wants to initialize an I/O that
999 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1000 * perform necessary setup and returns 1 to indicate the device is set up
1001 * successfully to perform DDP on this I/O, otherwise this returns 0.
1002 *
1003 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1004 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1005 * indicated by the FC exchange id 'xid', so the underlying device can
1006 * clean up and reuse resources for later DDP requests.
1007 *
1008 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1009 * struct scatterlist *sgl, unsigned int sgc);
1010 * Called when the FCoE Target wants to initialize an I/O that
1011 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1012 * perform necessary setup and returns 1 to indicate the device is set up
1013 * successfully to perform DDP on this I/O, otherwise this returns 0.
1014 *
68bad94e
NP
1015 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1016 * struct netdev_fcoe_hbainfo *hbainfo);
1017 * Called when the FCoE Protocol stack wants information on the underlying
1018 * device. This information is utilized by the FCoE protocol stack to
1019 * register attributes with Fiber Channel management service as per the
1020 * FC-GS Fabric Device Management Information(FDMI) specification.
1021 *
e9bce845
YZ
1022 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1023 * Called when the underlying device wants to override default World Wide
1024 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1025 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1026 * protocol stack to use.
1027 *
c445477d
BH
1028 * RFS acceleration.
1029 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1030 * u16 rxq_index, u32 flow_id);
1031 * Set hardware filter for RFS. rxq_index is the target queue index;
1032 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1033 * Return the filter ID on success, or a negative error code.
fbaec0ea 1034 *
8b98a70c 1035 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1036 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1037 * Called to make another netdev an underling.
1038 *
1039 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1040 * Called to release previously enslaved netdev.
5455c699
MM
1041 *
1042 * Feature/offload setting functions.
1a2a1444
DM
1043 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1044 * netdev_features_t features);
1045 * Adjusts the requested feature flags according to device-specific
1046 * constraints, and returns the resulting flags. Must not modify
1047 * the device state.
1048 *
c8f44aff 1049 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1050 * Called to update device configuration to new features. Passed
1051 * feature set might be less than what was returned by ndo_fix_features()).
1052 * Must return >0 or -errno if it changed dev->features itself.
1053 *
edc7d573 1054 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1055 * struct net_device *dev,
f6f6424b 1056 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1057 * Adds an FDB entry to dev for addr.
1690be63
VY
1058 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1059 * struct net_device *dev,
f6f6424b 1060 * const unsigned char *addr, u16 vid)
77162022
JF
1061 * Deletes the FDB entry from dev coresponding to addr.
1062 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1063 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1064 * int *idx)
77162022
JF
1065 * Used to add FDB entries to dump requests. Implementers should add
1066 * entries to skb and update idx with the number of entries.
e5a55a89 1067 *
ad41faa8
ND
1068 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1069 * u16 flags)
e5a55a89 1070 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1071 * struct net_device *dev, u32 filter_mask,
1072 * int nlflags)
ad41faa8
ND
1073 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1074 * u16 flags);
4bf84c35
JP
1075 *
1076 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1077 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1078 * which do not represent real hardware may define this to allow their
1079 * userspace components to manage their virtual carrier state. Devices
1080 * that determine carrier state from physical hardware properties (eg
1081 * network cables) or protocol-dependent mechanisms (eg
1082 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1083 *
1084 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1085 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1086 * Called to get ID of physical port of this device. If driver does
1087 * not implement this, it is assumed that the hw is not able to have
1088 * multiple net devices on single physical port.
53cf5275 1089 *
7c46a640
AD
1090 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1091 * struct udp_tunnel_info *ti);
1092 * Called by UDP tunnel to notify a driver about the UDP port and socket
1093 * address family that a UDP tunnel is listnening to. It is called only
1094 * when a new port starts listening. The operation is protected by the
1095 * RTNL.
1096 *
1097 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1098 * struct udp_tunnel_info *ti);
1099 * Called by UDP tunnel to notify the driver about a UDP port and socket
1100 * address family that the UDP tunnel is not listening to anymore. The
1101 * operation is protected by the RTNL.
1102 *
a6cc0cfa
JF
1103 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1104 * struct net_device *dev)
1105 * Called by upper layer devices to accelerate switching or other
1106 * station functionality into hardware. 'pdev is the lowerdev
1107 * to use for the offload and 'dev' is the net device that will
1108 * back the offload. Returns a pointer to the private structure
1109 * the upper layer will maintain.
1110 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1111 * Called by upper layer device to delete the station created
1112 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1113 * the station and priv is the structure returned by the add
1114 * operation.
1115 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1116 * struct net_device *dev,
1117 * void *priv);
1118 * Callback to use for xmit over the accelerated station. This
1119 * is used in place of ndo_start_xmit on accelerated net
1120 * devices.
822b3b2e
JF
1121 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1122 * int queue_index, u32 maxrate);
1123 * Called when a user wants to set a max-rate limitation of specific
1124 * TX queue.
a54acb3a
ND
1125 * int (*ndo_get_iflink)(const struct net_device *dev);
1126 * Called to get the iflink value of this device.
d746d707 1127 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1128 * bool proto_down);
d746d707
AK
1129 * This function is used to pass protocol port error state information
1130 * to the switch driver. The switch driver can react to the proto_down
1131 * by doing a phys down on the associated switch port.
fc4099f1
PS
1132 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1133 * This function is used to get egress tunnel information for given skb.
1134 * This is useful for retrieving outer tunnel header parameters while
1135 * sampling packet.
871b642a
PA
1136 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1137 * This function is used to specify the headroom that the skb must
1138 * consider when allocation skb during packet reception. Setting
1139 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1140 * forward. Setting a negative value resets the rx headroom to the
871b642a 1141 * default value.
a7862b45
BB
1142 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1143 * This function is used to set or query state related to XDP on the
1144 * netdevice. See definition of enum xdp_netdev_command for details.
d746d707 1145 *
d314774c
SH
1146 */
1147struct net_device_ops {
1148 int (*ndo_init)(struct net_device *dev);
1149 void (*ndo_uninit)(struct net_device *dev);
1150 int (*ndo_open)(struct net_device *dev);
1151 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1152 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1153 struct net_device *dev);
1154 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1155 struct net_device *dev,
1156 netdev_features_t features);
00829823 1157 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1158 struct sk_buff *skb,
99932d4f
DB
1159 void *accel_priv,
1160 select_queue_fallback_t fallback);
d314774c
SH
1161 void (*ndo_change_rx_flags)(struct net_device *dev,
1162 int flags);
d314774c 1163 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1164 int (*ndo_set_mac_address)(struct net_device *dev,
1165 void *addr);
d314774c 1166 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1167 int (*ndo_do_ioctl)(struct net_device *dev,
1168 struct ifreq *ifr, int cmd);
d314774c
SH
1169 int (*ndo_set_config)(struct net_device *dev,
1170 struct ifmap *map);
00829823
SH
1171 int (*ndo_change_mtu)(struct net_device *dev,
1172 int new_mtu);
1173 int (*ndo_neigh_setup)(struct net_device *dev,
1174 struct neigh_parms *);
d314774c
SH
1175 void (*ndo_tx_timeout) (struct net_device *dev);
1176
bc1f4470 1177 void (*ndo_get_stats64)(struct net_device *dev,
1178 struct rtnl_link_stats64 *storage);
3df5b3c6 1179 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1180 int (*ndo_get_offload_stats)(int attr_id,
1181 const struct net_device *dev,
1182 void *attr_data);
d314774c
SH
1183 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1184
8e586137 1185 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1186 __be16 proto, u16 vid);
8e586137 1187 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1188 __be16 proto, u16 vid);
d314774c 1189#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1190 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1191 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1192 struct netpoll_info *info);
0e34e931 1193 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1194#endif
95c26df8
WM
1195 int (*ndo_set_vf_mac)(struct net_device *dev,
1196 int queue, u8 *mac);
1197 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1198 int queue, u16 vlan,
1199 u8 qos, __be16 proto);
ed616689
SC
1200 int (*ndo_set_vf_rate)(struct net_device *dev,
1201 int vf, int min_tx_rate,
1202 int max_tx_rate);
5f8444a3
GR
1203 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1204 int vf, bool setting);
dd461d6a
HS
1205 int (*ndo_set_vf_trust)(struct net_device *dev,
1206 int vf, bool setting);
95c26df8
WM
1207 int (*ndo_get_vf_config)(struct net_device *dev,
1208 int vf,
1209 struct ifla_vf_info *ivf);
1d8faf48
RE
1210 int (*ndo_set_vf_link_state)(struct net_device *dev,
1211 int vf, int link_state);
3b766cd8
EBE
1212 int (*ndo_get_vf_stats)(struct net_device *dev,
1213 int vf,
1214 struct ifla_vf_stats
1215 *vf_stats);
57b61080
SF
1216 int (*ndo_set_vf_port)(struct net_device *dev,
1217 int vf,
1218 struct nlattr *port[]);
1219 int (*ndo_get_vf_port)(struct net_device *dev,
1220 int vf, struct sk_buff *skb);
cc8e27cc
EC
1221 int (*ndo_set_vf_guid)(struct net_device *dev,
1222 int vf, u64 guid,
1223 int guid_type);
01a3d796
VZ
1224 int (*ndo_set_vf_rss_query_en)(
1225 struct net_device *dev,
1226 int vf, bool setting);
16e5cc64 1227 int (*ndo_setup_tc)(struct net_device *dev,
a5fcf8a6 1228 u32 handle, u32 chain_index,
16e5cc64
JF
1229 __be16 protocol,
1230 struct tc_to_netdev *tc);
d11ead75 1231#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1232 int (*ndo_fcoe_enable)(struct net_device *dev);
1233 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1234 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1235 u16 xid,
1236 struct scatterlist *sgl,
1237 unsigned int sgc);
1238 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1239 u16 xid);
6247e086
YZ
1240 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1241 u16 xid,
1242 struct scatterlist *sgl,
1243 unsigned int sgc);
68bad94e
NP
1244 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1245 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1246#endif
1247
d11ead75 1248#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1249#define NETDEV_FCOE_WWNN 0
1250#define NETDEV_FCOE_WWPN 1
1251 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1252 u64 *wwn, int type);
4d288d57 1253#endif
3c9c36bc 1254
c445477d
BH
1255#ifdef CONFIG_RFS_ACCEL
1256 int (*ndo_rx_flow_steer)(struct net_device *dev,
1257 const struct sk_buff *skb,
1258 u16 rxq_index,
1259 u32 flow_id);
1260#endif
fbaec0ea
JP
1261 int (*ndo_add_slave)(struct net_device *dev,
1262 struct net_device *slave_dev);
1263 int (*ndo_del_slave)(struct net_device *dev,
1264 struct net_device *slave_dev);
c8f44aff
MM
1265 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1266 netdev_features_t features);
5455c699 1267 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1268 netdev_features_t features);
503eebc2
JP
1269 int (*ndo_neigh_construct)(struct net_device *dev,
1270 struct neighbour *n);
1271 void (*ndo_neigh_destroy)(struct net_device *dev,
1272 struct neighbour *n);
77162022
JF
1273
1274 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1275 struct nlattr *tb[],
77162022 1276 struct net_device *dev,
6b6e2725 1277 const unsigned char *addr,
f6f6424b 1278 u16 vid,
77162022
JF
1279 u16 flags);
1280 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1281 struct nlattr *tb[],
77162022 1282 struct net_device *dev,
f6f6424b
JP
1283 const unsigned char *addr,
1284 u16 vid);
77162022
JF
1285 int (*ndo_fdb_dump)(struct sk_buff *skb,
1286 struct netlink_callback *cb,
1287 struct net_device *dev,
5d5eacb3 1288 struct net_device *filter_dev,
d297653d 1289 int *idx);
e5a55a89
JF
1290
1291 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1292 struct nlmsghdr *nlh,
1293 u16 flags);
e5a55a89
JF
1294 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1295 u32 pid, u32 seq,
6cbdceeb 1296 struct net_device *dev,
46c264da
ND
1297 u32 filter_mask,
1298 int nlflags);
407af329 1299 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1300 struct nlmsghdr *nlh,
1301 u16 flags);
4bf84c35
JP
1302 int (*ndo_change_carrier)(struct net_device *dev,
1303 bool new_carrier);
66b52b0d 1304 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1305 struct netdev_phys_item_id *ppid);
db24a904
DA
1306 int (*ndo_get_phys_port_name)(struct net_device *dev,
1307 char *name, size_t len);
7c46a640
AD
1308 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1309 struct udp_tunnel_info *ti);
1310 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1311 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1312 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1313 struct net_device *dev);
1314 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1315 void *priv);
1316
1317 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1318 struct net_device *dev,
1319 void *priv);
25175ba5 1320 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1321 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1322 int queue_index,
1323 u32 maxrate);
a54acb3a 1324 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1325 int (*ndo_change_proto_down)(struct net_device *dev,
1326 bool proto_down);
fc4099f1
PS
1327 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1328 struct sk_buff *skb);
871b642a
PA
1329 void (*ndo_set_rx_headroom)(struct net_device *dev,
1330 int needed_headroom);
a7862b45
BB
1331 int (*ndo_xdp)(struct net_device *dev,
1332 struct netdev_xdp *xdp);
d314774c
SH
1333};
1334
7aa98047
LR
1335/**
1336 * enum net_device_priv_flags - &struct net_device priv_flags
1337 *
1338 * These are the &struct net_device, they are only set internally
1339 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1340 * userspace; this means that the order of these flags can change
7aa98047
LR
1341 * during any kernel release.
1342 *
1343 * You should have a pretty good reason to be extending these flags.
1344 *
1345 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1346 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1347 * @IFF_BONDING: bonding master or slave
7aa98047 1348 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1349 * @IFF_WAN_HDLC: WAN HDLC device
1350 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1351 * release skb->dst
1352 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1353 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1354 * @IFF_MACVLAN_PORT: device used as macvlan port
1355 * @IFF_BRIDGE_PORT: device used as bridge port
1356 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1357 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1358 * @IFF_UNICAST_FLT: Supports unicast filtering
1359 * @IFF_TEAM_PORT: device used as team port
1360 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1361 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1362 * change when it's running
1363 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1364 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1365 * underlying stacked devices
1366 * @IFF_IPVLAN_MASTER: IPvlan master device
1367 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1368 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1369 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1370 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1371 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1372 * @IFF_TEAM: device is a team device
d4ab4286 1373 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1374 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1375 * entity (i.e. the master device for bridged veth)
3c175784 1376 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1377 */
1378enum netdev_priv_flags {
1379 IFF_802_1Q_VLAN = 1<<0,
1380 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1381 IFF_BONDING = 1<<2,
1382 IFF_ISATAP = 1<<3,
1383 IFF_WAN_HDLC = 1<<4,
1384 IFF_XMIT_DST_RELEASE = 1<<5,
1385 IFF_DONT_BRIDGE = 1<<6,
1386 IFF_DISABLE_NETPOLL = 1<<7,
1387 IFF_MACVLAN_PORT = 1<<8,
1388 IFF_BRIDGE_PORT = 1<<9,
1389 IFF_OVS_DATAPATH = 1<<10,
1390 IFF_TX_SKB_SHARING = 1<<11,
1391 IFF_UNICAST_FLT = 1<<12,
1392 IFF_TEAM_PORT = 1<<13,
1393 IFF_SUPP_NOFCS = 1<<14,
1394 IFF_LIVE_ADDR_CHANGE = 1<<15,
1395 IFF_MACVLAN = 1<<16,
1396 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1397 IFF_IPVLAN_MASTER = 1<<18,
1398 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1399 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1400 IFF_NO_QUEUE = 1<<21,
1401 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1402 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1403 IFF_TEAM = 1<<24,
d4ab4286 1404 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1405 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1406 IFF_MACSEC = 1<<27,
7aa98047
LR
1407};
1408
1409#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1410#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1411#define IFF_BONDING IFF_BONDING
7aa98047 1412#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1413#define IFF_WAN_HDLC IFF_WAN_HDLC
1414#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1415#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1416#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1417#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1418#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1419#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1420#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1421#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1422#define IFF_TEAM_PORT IFF_TEAM_PORT
1423#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1424#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1425#define IFF_MACVLAN IFF_MACVLAN
02875878 1426#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1427#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1428#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1429#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1430#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1431#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1432#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1433#define IFF_TEAM IFF_TEAM
d4ab4286 1434#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1435#define IFF_MACSEC IFF_MACSEC
7aa98047 1436
536721b1
KK
1437/**
1438 * struct net_device - The DEVICE structure.
1439 * Actually, this whole structure is a big mistake. It mixes I/O
1440 * data with strictly "high-level" data, and it has to know about
1441 * almost every data structure used in the INET module.
1442 *
1443 * @name: This is the first field of the "visible" part of this structure
1444 * (i.e. as seen by users in the "Space.c" file). It is the name
1445 * of the interface.
1446 *
1447 * @name_hlist: Device name hash chain, please keep it close to name[]
1448 * @ifalias: SNMP alias
1449 * @mem_end: Shared memory end
1450 * @mem_start: Shared memory start
1451 * @base_addr: Device I/O address
1452 * @irq: Device IRQ number
1453 *
14ffbbb8
TG
1454 * @carrier_changes: Stats to monitor carrier on<->off transitions
1455 *
536721b1
KK
1456 * @state: Generic network queuing layer state, see netdev_state_t
1457 * @dev_list: The global list of network devices
5e82b4b2
BH
1458 * @napi_list: List entry used for polling NAPI devices
1459 * @unreg_list: List entry when we are unregistering the
1460 * device; see the function unregister_netdev
1461 * @close_list: List entry used when we are closing the device
62d885fe
BP
1462 * @ptype_all: Device-specific packet handlers for all protocols
1463 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1464 *
1465 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1466 * @features: Currently active device features
1467 * @hw_features: User-changeable features
1468 *
1469 * @wanted_features: User-requested features
1470 * @vlan_features: Mask of features inheritable by VLAN devices
1471 *
1472 * @hw_enc_features: Mask of features inherited by encapsulating devices
1473 * This field indicates what encapsulation
1474 * offloads the hardware is capable of doing,
1475 * and drivers will need to set them appropriately.
1476 *
1477 * @mpls_features: Mask of features inheritable by MPLS
1478 *
1479 * @ifindex: interface index
5e82b4b2 1480 * @group: The group the device belongs to
536721b1
KK
1481 *
1482 * @stats: Statistics struct, which was left as a legacy, use
1483 * rtnl_link_stats64 instead
1484 *
1485 * @rx_dropped: Dropped packets by core network,
1486 * do not use this in drivers
1487 * @tx_dropped: Dropped packets by core network,
1488 * do not use this in drivers
6e7333d3
JW
1489 * @rx_nohandler: nohandler dropped packets by core network on
1490 * inactive devices, do not use this in drivers
536721b1 1491 *
536721b1
KK
1492 * @wireless_handlers: List of functions to handle Wireless Extensions,
1493 * instead of ioctl,
1494 * see <net/iw_handler.h> for details.
1495 * @wireless_data: Instance data managed by the core of wireless extensions
1496 *
1497 * @netdev_ops: Includes several pointers to callbacks,
1498 * if one wants to override the ndo_*() functions
1499 * @ethtool_ops: Management operations
f997c55c
AA
1500 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1501 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1502 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1503 * of Layer 2 headers.
1504 *
1505 * @flags: Interface flags (a la BSD)
1506 * @priv_flags: Like 'flags' but invisible to userspace,
1507 * see if.h for the definitions
1508 * @gflags: Global flags ( kept as legacy )
1509 * @padded: How much padding added by alloc_netdev()
1510 * @operstate: RFC2863 operstate
1511 * @link_mode: Mapping policy to operstate
1512 * @if_port: Selectable AUI, TP, ...
1513 * @dma: DMA channel
1514 * @mtu: Interface MTU value
61e84623
JW
1515 * @min_mtu: Interface Minimum MTU value
1516 * @max_mtu: Interface Maximum MTU value
536721b1 1517 * @type: Interface hardware type
2793a23a 1518 * @hard_header_len: Maximum hardware header length.
217e6fa2 1519 * @min_header_len: Minimum hardware header length
536721b1
KK
1520 *
1521 * @needed_headroom: Extra headroom the hardware may need, but not in all
1522 * cases can this be guaranteed
1523 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1524 * cases can this be guaranteed. Some cases also use
1525 * LL_MAX_HEADER instead to allocate the skb
1526 *
1527 * interface address info:
1528 *
1529 * @perm_addr: Permanent hw address
1530 * @addr_assign_type: Hw address assignment type
1531 * @addr_len: Hardware address length
8626a0c8 1532 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1533 * @dev_id: Used to differentiate devices that share
1534 * the same link layer address
1535 * @dev_port: Used to differentiate devices that share
1536 * the same function
1537 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1538 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1539 * has been enabled due to the need to listen to
1540 * additional unicast addresses in a device that
1541 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1542 * @uc: unicast mac addresses
1543 * @mc: multicast mac addresses
1544 * @dev_addrs: list of device hw addresses
1545 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1546 * @promiscuity: Number of times the NIC is told to work in
1547 * promiscuous mode; if it becomes 0 the NIC will
1548 * exit promiscuous mode
536721b1
KK
1549 * @allmulti: Counter, enables or disables allmulticast mode
1550 *
1551 * @vlan_info: VLAN info
1552 * @dsa_ptr: dsa specific data
1553 * @tipc_ptr: TIPC specific data
1554 * @atalk_ptr: AppleTalk link
1555 * @ip_ptr: IPv4 specific data
1556 * @dn_ptr: DECnet specific data
1557 * @ip6_ptr: IPv6 specific data
1558 * @ax25_ptr: AX.25 specific data
1559 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1560 *
536721b1
KK
1561 * @dev_addr: Hw address (before bcast,
1562 * because most packets are unicast)
1563 *
1564 * @_rx: Array of RX queues
1565 * @num_rx_queues: Number of RX queues
1566 * allocated at register_netdev() time
1567 * @real_num_rx_queues: Number of RX queues currently active in device
1568 *
1569 * @rx_handler: handler for received packets
1570 * @rx_handler_data: XXX: need comments on this one
1571 * @ingress_queue: XXX: need comments on this one
1572 * @broadcast: hw bcast address
1573 *
14ffbbb8
TG
1574 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1575 * indexed by RX queue number. Assigned by driver.
1576 * This must only be set if the ndo_rx_flow_steer
1577 * operation is defined
1578 * @index_hlist: Device index hash chain
1579 *
536721b1
KK
1580 * @_tx: Array of TX queues
1581 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1582 * @real_num_tx_queues: Number of TX queues currently active in device
1583 * @qdisc: Root qdisc from userspace point of view
1584 * @tx_queue_len: Max frames per queue allowed
1585 * @tx_global_lock: XXX: need comments on this one
1586 *
1587 * @xps_maps: XXX: need comments on this one
1588 *
536721b1 1589 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1590 * the watchdog (see dev_watchdog())
536721b1
KK
1591 * @watchdog_timer: List of timers
1592 *
1593 * @pcpu_refcnt: Number of references to this device
1594 * @todo_list: Delayed register/unregister
536721b1
KK
1595 * @link_watch_list: XXX: need comments on this one
1596 *
1597 * @reg_state: Register/unregister state machine
1598 * @dismantle: Device is going to be freed
1599 * @rtnl_link_state: This enum represents the phases of creating
1600 * a new link
1601 *
cf124db5
DM
1602 * @needs_free_netdev: Should unregister perform free_netdev?
1603 * @priv_destructor: Called from unregister
536721b1
KK
1604 * @npinfo: XXX: need comments on this one
1605 * @nd_net: Network namespace this network device is inside
1606 *
1607 * @ml_priv: Mid-layer private
1608 * @lstats: Loopback statistics
1609 * @tstats: Tunnel statistics
1610 * @dstats: Dummy statistics
1611 * @vstats: Virtual ethernet statistics
1612 *
1613 * @garp_port: GARP
1614 * @mrp_port: MRP
1615 *
1616 * @dev: Class/net/name entry
1617 * @sysfs_groups: Space for optional device, statistics and wireless
1618 * sysfs groups
1619 *
1620 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1621 * @rtnl_link_ops: Rtnl_link_ops
1622 *
1623 * @gso_max_size: Maximum size of generic segmentation offload
1624 * @gso_max_segs: Maximum number of segments that can be passed to the
1625 * NIC for GSO
1626 *
1627 * @dcbnl_ops: Data Center Bridging netlink ops
1628 * @num_tc: Number of traffic classes in the net device
1629 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1630 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1631 *
1632 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1633 *
1634 * @priomap: XXX: need comments on this one
1635 * @phydev: Physical device may attach itself
1636 * for hardware timestamping
1637 *
123b3652
ED
1638 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1639 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1640 *
d746d707
AK
1641 * @proto_down: protocol port state information can be sent to the
1642 * switch driver and used to set the phys state of the
1643 * switch port.
1644 *
1da177e4
LT
1645 * FIXME: cleanup struct net_device such that network protocol info
1646 * moves out.
1647 */
1648
d94d9fee 1649struct net_device {
1da177e4 1650 char name[IFNAMSIZ];
9356b8fc 1651 struct hlist_node name_hlist;
0b815a1a 1652 char *ifalias;
1da177e4
LT
1653 /*
1654 * I/O specific fields
1655 * FIXME: Merge these and struct ifmap into one
1656 */
536721b1
KK
1657 unsigned long mem_end;
1658 unsigned long mem_start;
1659 unsigned long base_addr;
1660 int irq;
1da177e4 1661
14ffbbb8
TG
1662 atomic_t carrier_changes;
1663
1da177e4 1664 /*
536721b1
KK
1665 * Some hardware also needs these fields (state,dev_list,
1666 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1667 * part of the usual set specified in Space.c.
1668 */
1669
1da177e4
LT
1670 unsigned long state;
1671
7562f876 1672 struct list_head dev_list;
bea3348e 1673 struct list_head napi_list;
44a0873d 1674 struct list_head unreg_list;
5cde2829 1675 struct list_head close_list;
7866a621
SN
1676 struct list_head ptype_all;
1677 struct list_head ptype_specific;
2f268f12 1678
2f268f12
VF
1679 struct {
1680 struct list_head upper;
1681 struct list_head lower;
1682 } adj_list;
1683
c8f44aff 1684 netdev_features_t features;
c8f44aff 1685 netdev_features_t hw_features;
c8f44aff 1686 netdev_features_t wanted_features;
c8f44aff 1687 netdev_features_t vlan_features;
6a674e9c 1688 netdev_features_t hw_enc_features;
0d89d203 1689 netdev_features_t mpls_features;
802ab55a 1690 netdev_features_t gso_partial_features;
04ed3e74 1691
1da177e4 1692 int ifindex;
7a66bbc9 1693 int group;
1da177e4 1694
c45d286e 1695 struct net_device_stats stats;
015f0688 1696
015f0688
ED
1697 atomic_long_t rx_dropped;
1698 atomic_long_t tx_dropped;
6e7333d3 1699 atomic_long_t rx_nohandler;
1da177e4 1700
b86e0280 1701#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1702 const struct iw_handler_def *wireless_handlers;
1703 struct iw_public_data *wireless_data;
b86e0280 1704#endif
d314774c 1705 const struct net_device_ops *netdev_ops;
76fd8593 1706 const struct ethtool_ops *ethtool_ops;
4170604f 1707#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1708 const struct switchdev_ops *switchdev_ops;
4170604f 1709#endif
1b69c6d0
DA
1710#ifdef CONFIG_NET_L3_MASTER_DEV
1711 const struct l3mdev_ops *l3mdev_ops;
1712#endif
f997c55c
AA
1713#if IS_ENABLED(CONFIG_IPV6)
1714 const struct ndisc_ops *ndisc_ops;
1715#endif
1da177e4 1716
d77e38e6
SK
1717#ifdef CONFIG_XFRM
1718 const struct xfrmdev_ops *xfrmdev_ops;
1719#endif
1720
3b04ddde
SH
1721 const struct header_ops *header_ops;
1722
536721b1
KK
1723 unsigned int flags;
1724 unsigned int priv_flags;
1725
1da177e4 1726 unsigned short gflags;
536721b1 1727 unsigned short padded;
1da177e4 1728
536721b1
KK
1729 unsigned char operstate;
1730 unsigned char link_mode;
b00055aa 1731
536721b1
KK
1732 unsigned char if_port;
1733 unsigned char dma;
bdc220da 1734
536721b1 1735 unsigned int mtu;
61e84623
JW
1736 unsigned int min_mtu;
1737 unsigned int max_mtu;
536721b1
KK
1738 unsigned short type;
1739 unsigned short hard_header_len;
d92be7a4 1740 unsigned char min_header_len;
1da177e4 1741
f5184d26
JB
1742 unsigned short needed_headroom;
1743 unsigned short needed_tailroom;
1744
1da177e4 1745 /* Interface address info. */
536721b1
KK
1746 unsigned char perm_addr[MAX_ADDR_LEN];
1747 unsigned char addr_assign_type;
1748 unsigned char addr_len;
a0a9663d 1749 unsigned short neigh_priv_len;
536721b1
KK
1750 unsigned short dev_id;
1751 unsigned short dev_port;
ccffad25 1752 spinlock_t addr_list_lock;
14ffbbb8
TG
1753 unsigned char name_assign_type;
1754 bool uc_promisc;
536721b1
KK
1755 struct netdev_hw_addr_list uc;
1756 struct netdev_hw_addr_list mc;
1757 struct netdev_hw_addr_list dev_addrs;
1758
4c3d5e7b
ED
1759#ifdef CONFIG_SYSFS
1760 struct kset *queues_kset;
1761#endif
9d45abe1
WC
1762 unsigned int promiscuity;
1763 unsigned int allmulti;
1da177e4 1764
1da177e4 1765
5e82b4b2 1766 /* Protocol-specific pointers */
65ac6a5f 1767
d11ead75 1768#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1769 struct vlan_info __rcu *vlan_info;
65ac6a5f 1770#endif
34a430d7 1771#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1772 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1773#endif
1774#if IS_ENABLED(CONFIG_TIPC)
536721b1 1775 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1776#endif
536721b1
KK
1777 void *atalk_ptr;
1778 struct in_device __rcu *ip_ptr;
1779 struct dn_dev __rcu *dn_ptr;
1780 struct inet6_dev __rcu *ip6_ptr;
1781 void *ax25_ptr;
1782 struct wireless_dev *ieee80211_ptr;
98a18b6f 1783 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1784#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1785 struct mpls_dev __rcu *mpls_ptr;
1786#endif
1da177e4 1787
9356b8fc 1788/*
cd13539b 1789 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1790 */
9356b8fc 1791 /* Interface address info used in eth_type_trans() */
536721b1 1792 unsigned char *dev_addr;
f001fde5 1793
a953be53 1794#ifdef CONFIG_SYSFS
0a9627f2
TH
1795 struct netdev_rx_queue *_rx;
1796
0a9627f2 1797 unsigned int num_rx_queues;
62fe0b40 1798 unsigned int real_num_rx_queues;
df334545 1799#endif
0a9627f2 1800
7acedaf5 1801 struct bpf_prog __rcu *xdp_prog;
3b47d303 1802 unsigned long gro_flush_timeout;
61391cde 1803 rx_handler_func_t __rcu *rx_handler;
1804 void __rcu *rx_handler_data;
e8a0464c 1805
4cda01e8 1806#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1807 struct tcf_proto __rcu *ingress_cl_list;
1808#endif
24824a09 1809 struct netdev_queue __rcu *ingress_queue;
e687ad60 1810#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1811 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1812#endif
d2788d34 1813
536721b1 1814 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1815#ifdef CONFIG_RFS_ACCEL
1816 struct cpu_rmap *rx_cpu_rmap;
1817#endif
1818 struct hlist_node index_hlist;
cd13539b
ED
1819
1820/*
1821 * Cache lines mostly used on transmit path
1822 */
e8a0464c
DM
1823 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1824 unsigned int num_tx_queues;
fd2ea0a7 1825 unsigned int real_num_tx_queues;
af356afa 1826 struct Qdisc *qdisc;
59cc1f61
JK
1827#ifdef CONFIG_NET_SCHED
1828 DECLARE_HASHTABLE (qdisc_hash, 4);
1829#endif
0cd29503 1830 unsigned int tx_queue_len;
c3f26a26 1831 spinlock_t tx_global_lock;
14ffbbb8 1832 int watchdog_timeo;
cd13539b 1833
bf264145 1834#ifdef CONFIG_XPS
a4177869 1835 struct xps_dev_maps __rcu *xps_maps;
bf264145 1836#endif
1f211a1b
DB
1837#ifdef CONFIG_NET_CLS_ACT
1838 struct tcf_proto __rcu *egress_cl_list;
1839#endif
0c4f691f 1840
9356b8fc 1841 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1842 struct timer_list watchdog_timer;
1843
29b4433d 1844 int __percpu *pcpu_refcnt;
1da177e4 1845 struct list_head todo_list;
1da177e4 1846
e014debe 1847 struct list_head link_watch_list;
572a103d 1848
1da177e4 1849 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1850 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1851 NETREG_UNREGISTERING, /* called unregister_netdevice */
1852 NETREG_UNREGISTERED, /* completed unregister todo */
1853 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1854 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1855 } reg_state:8;
1856
536721b1 1857 bool dismantle;
a2835763
PM
1858
1859 enum {
1860 RTNL_LINK_INITIALIZED,
1861 RTNL_LINK_INITIALIZING,
1862 } rtnl_link_state:16;
1da177e4 1863
cf124db5
DM
1864 bool needs_free_netdev;
1865 void (*priv_destructor)(struct net_device *dev);
1da177e4 1866
1da177e4 1867#ifdef CONFIG_NETPOLL
5fbee843 1868 struct netpoll_info __rcu *npinfo;
1da177e4 1869#endif
eae792b7 1870
0c5c9fb5 1871 possible_net_t nd_net;
4a1c5371 1872
4951704b 1873 /* mid-layer private */
a7855c78 1874 union {
536721b1
KK
1875 void *ml_priv;
1876 struct pcpu_lstats __percpu *lstats;
8f84985f 1877 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1878 struct pcpu_dstats __percpu *dstats;
1879 struct pcpu_vstats __percpu *vstats;
a7855c78 1880 };
536721b1 1881
fb585b44 1882#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1883 struct garp_port __rcu *garp_port;
fb585b44
TK
1884#endif
1885#if IS_ENABLED(CONFIG_MRP)
febf018d 1886 struct mrp_port __rcu *mrp_port;
fb585b44 1887#endif
1da177e4 1888
5e82b4b2 1889 struct device dev;
0c509a6c 1890 const struct attribute_group *sysfs_groups[4];
a953be53 1891 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1892
38f7b870 1893 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1894
82cc1a7a
PWJ
1895 /* for setting kernel sock attribute on TCP connection setup */
1896#define GSO_MAX_SIZE 65536
1897 unsigned int gso_max_size;
30b678d8
BH
1898#define GSO_MAX_SEGS 65535
1899 u16 gso_max_segs;
743b03a8 1900
7a6b6f51 1901#ifdef CONFIG_DCB
32953543 1902 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1903#endif
5e82b4b2
BH
1904 u8 num_tc;
1905 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1906 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1907
d11ead75 1908#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1909 unsigned int fcoe_ddp_xid;
5bc1421e 1910#endif
86f8515f 1911#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1912 struct netprio_map __rcu *priomap;
4d288d57 1913#endif
5e82b4b2
BH
1914 struct phy_device *phydev;
1915 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1916 struct lock_class_key *qdisc_running_key;
5e82b4b2 1917 bool proto_down;
1da177e4 1918};
43cb76d9 1919#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1920
b5cdae32
DM
1921static inline bool netif_elide_gro(const struct net_device *dev)
1922{
1923 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1924 return true;
1925 return false;
1926}
1927
1da177e4 1928#define NETDEV_ALIGN 32
1da177e4 1929
4f57c087
JF
1930static inline
1931int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1932{
1933 return dev->prio_tc_map[prio & TC_BITMASK];
1934}
1935
1936static inline
1937int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1938{
1939 if (tc >= dev->num_tc)
1940 return -EINVAL;
1941
1942 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1943 return 0;
1944}
1945
8d059b0f 1946int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1947void netdev_reset_tc(struct net_device *dev);
1948int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1949int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1950
1951static inline
1952int netdev_get_num_tc(struct net_device *dev)
1953{
1954 return dev->num_tc;
1955}
1956
e8a0464c
DM
1957static inline
1958struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1959 unsigned int index)
1960{
1961 return &dev->_tx[index];
1962}
1963
10c51b56
DB
1964static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1965 const struct sk_buff *skb)
1966{
1967 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1968}
1969
e8a0464c
DM
1970static inline void netdev_for_each_tx_queue(struct net_device *dev,
1971 void (*f)(struct net_device *,
1972 struct netdev_queue *,
1973 void *),
1974 void *arg)
1975{
1976 unsigned int i;
1977
1978 for (i = 0; i < dev->num_tx_queues; i++)
1979 f(dev, &dev->_tx[i], arg);
1980}
1981
d3fff6c4
ED
1982#define netdev_lockdep_set_classes(dev) \
1983{ \
1984 static struct lock_class_key qdisc_tx_busylock_key; \
1985 static struct lock_class_key qdisc_running_key; \
1986 static struct lock_class_key qdisc_xmit_lock_key; \
1987 static struct lock_class_key dev_addr_list_lock_key; \
1988 unsigned int i; \
1989 \
1990 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1991 (dev)->qdisc_running_key = &qdisc_running_key; \
1992 lockdep_set_class(&(dev)->addr_list_lock, \
1993 &dev_addr_list_lock_key); \
1994 for (i = 0; i < (dev)->num_tx_queues; i++) \
1995 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1996 &qdisc_xmit_lock_key); \
1997}
1998
f629d208 1999struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
2000 struct sk_buff *skb,
2001 void *accel_priv);
8c4c49df 2002
871b642a
PA
2003/* returns the headroom that the master device needs to take in account
2004 * when forwarding to this dev
2005 */
2006static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2007{
2008 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2009}
2010
2011static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2012{
2013 if (dev->netdev_ops->ndo_set_rx_headroom)
2014 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2015}
2016
2017/* set the device rx headroom to the dev's default */
2018static inline void netdev_reset_rx_headroom(struct net_device *dev)
2019{
2020 netdev_set_rx_headroom(dev, -1);
2021}
2022
c346dca1
YH
2023/*
2024 * Net namespace inlines
2025 */
2026static inline
2027struct net *dev_net(const struct net_device *dev)
2028{
c2d9ba9b 2029 return read_pnet(&dev->nd_net);
c346dca1
YH
2030}
2031
2032static inline
f5aa23fd 2033void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2034{
0c5c9fb5 2035 write_pnet(&dev->nd_net, net);
c346dca1
YH
2036}
2037
bea3348e
SH
2038/**
2039 * netdev_priv - access network device private data
2040 * @dev: network device
2041 *
2042 * Get network device private data
2043 */
6472ce60 2044static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2045{
1ce8e7b5 2046 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2047}
2048
1da177e4
LT
2049/* Set the sysfs physical device reference for the network logical device
2050 * if set prior to registration will cause a symlink during initialization.
2051 */
43cb76d9 2052#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2053
384912ed 2054/* Set the sysfs device type for the network logical device to allow
3f79410c 2055 * fine-grained identification of different network device types. For
5e82b4b2 2056 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2057 */
2058#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2059
82dc3c63
ED
2060/* Default NAPI poll() weight
2061 * Device drivers are strongly advised to not use bigger value
2062 */
2063#define NAPI_POLL_WEIGHT 64
2064
3b582cc1 2065/**
5e82b4b2 2066 * netif_napi_add - initialize a NAPI context
3b582cc1 2067 * @dev: network device
5e82b4b2 2068 * @napi: NAPI context
3b582cc1
SH
2069 * @poll: polling function
2070 * @weight: default weight
2071 *
5e82b4b2
BH
2072 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2073 * *any* of the other NAPI-related functions.
3b582cc1 2074 */
d565b0a1
HX
2075void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2076 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2077
d64b5e85 2078/**
5e82b4b2 2079 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2080 * @dev: network device
5e82b4b2 2081 * @napi: NAPI context
d64b5e85
ED
2082 * @poll: polling function
2083 * @weight: default weight
2084 *
2085 * This variant of netif_napi_add() should be used from drivers using NAPI
2086 * to exclusively poll a TX queue.
2087 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2088 */
2089static inline void netif_tx_napi_add(struct net_device *dev,
2090 struct napi_struct *napi,
2091 int (*poll)(struct napi_struct *, int),
2092 int weight)
2093{
2094 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2095 netif_napi_add(dev, napi, poll, weight);
2096}
2097
d8156534 2098/**
5e82b4b2
BH
2099 * netif_napi_del - remove a NAPI context
2100 * @napi: NAPI context
d8156534 2101 *
5e82b4b2 2102 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2103 */
d565b0a1
HX
2104void netif_napi_del(struct napi_struct *napi);
2105
2106struct napi_gro_cb {
78a478d0 2107 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2108 void *frag0;
78a478d0 2109
7489594c
HX
2110 /* Length of frag0. */
2111 unsigned int frag0_len;
2112
86911732 2113 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2114 int data_offset;
86911732 2115
d565b0a1 2116 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2117 u16 flush;
2118
2119 /* Save the IP ID here and check when we get to the transport layer */
2120 u16 flush_id;
d565b0a1
HX
2121
2122 /* Number of segments aggregated. */
2e71a6f8
ED
2123 u16 count;
2124
15e2396d
TH
2125 /* Start offset for remote checksum offload */
2126 u16 gro_remcsum_start;
2127
2e71a6f8
ED
2128 /* jiffies when first packet was created/queued */
2129 unsigned long age;
86347245 2130
afe93325 2131 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2132 u16 proto;
2133
baa32ff4
TH
2134 /* This is non-zero if the packet may be of the same flow. */
2135 u8 same_flow:1;
2136
fac8e0f5
JG
2137 /* Used in tunnel GRO receive */
2138 u8 encap_mark:1;
573e8fca
TH
2139
2140 /* GRO checksum is valid */
2141 u8 csum_valid:1;
2142
662880f4
TH
2143 /* Number of checksums via CHECKSUM_UNNECESSARY */
2144 u8 csum_cnt:3;
c3c7c254 2145
baa32ff4
TH
2146 /* Free the skb? */
2147 u8 free:2;
2148#define NAPI_GRO_FREE 1
2149#define NAPI_GRO_FREE_STOLEN_HEAD 2
2150
efc98d08
TH
2151 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2152 u8 is_ipv6:1;
2153
a0ca153f
AD
2154 /* Used in GRE, set in fou/gue_gro_receive */
2155 u8 is_fou:1;
2156
1530545e
AD
2157 /* Used to determine if flush_id can be ignored */
2158 u8 is_atomic:1;
2159
fcd91dd4
SD
2160 /* Number of gro_receive callbacks this packet already went through */
2161 u8 recursion_counter:4;
2162
2163 /* 1 bit hole */
baa32ff4 2164
bf5a755f
JC
2165 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2166 __wsum csum;
2167
c3c7c254
ED
2168 /* used in skb_gro_receive() slow path */
2169 struct sk_buff *last;
d565b0a1
HX
2170};
2171
2172#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2173
fcd91dd4
SD
2174#define GRO_RECURSION_LIMIT 15
2175static inline int gro_recursion_inc_test(struct sk_buff *skb)
2176{
2177 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2178}
2179
2180typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2181static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2182 struct sk_buff **head,
2183 struct sk_buff *skb)
2184{
2185 if (unlikely(gro_recursion_inc_test(skb))) {
2186 NAPI_GRO_CB(skb)->flush |= 1;
2187 return NULL;
2188 }
2189
2190 return cb(head, skb);
2191}
2192
2193typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2194 struct sk_buff *);
2195static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2196 struct sock *sk,
2197 struct sk_buff **head,
2198 struct sk_buff *skb)
2199{
2200 if (unlikely(gro_recursion_inc_test(skb))) {
2201 NAPI_GRO_CB(skb)->flush |= 1;
2202 return NULL;
2203 }
2204
2205 return cb(sk, head, skb);
2206}
2207
1da177e4 2208struct packet_type {
f2ccd8fa
DM
2209 __be16 type; /* This is really htons(ether_type). */
2210 struct net_device *dev; /* NULL is wildcarded here */
2211 int (*func) (struct sk_buff *,
2212 struct net_device *,
2213 struct packet_type *,
2214 struct net_device *);
c0de08d0
EL
2215 bool (*id_match)(struct packet_type *ptype,
2216 struct sock *sk);
1da177e4
LT
2217 void *af_packet_priv;
2218 struct list_head list;
2219};
2220
f191a1d1 2221struct offload_callbacks {
576a30eb 2222 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2223 netdev_features_t features);
d565b0a1 2224 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2225 struct sk_buff *skb);
299603e8 2226 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2227};
2228
2229struct packet_offload {
2230 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2231 u16 priority;
f191a1d1
VY
2232 struct offload_callbacks callbacks;
2233 struct list_head list;
1da177e4
LT
2234};
2235
5e82b4b2 2236/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2237struct pcpu_sw_netstats {
2238 u64 rx_packets;
2239 u64 rx_bytes;
2240 u64 tx_packets;
2241 u64 tx_bytes;
2242 struct u64_stats_sync syncp;
2243};
2244
aabc92bb
PNA
2245#define __netdev_alloc_pcpu_stats(type, gfp) \
2246({ \
2247 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2248 if (pcpu_stats) { \
2249 int __cpu; \
2250 for_each_possible_cpu(__cpu) { \
2251 typeof(type) *stat; \
2252 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2253 u64_stats_init(&stat->syncp); \
2254 } \
2255 } \
2256 pcpu_stats; \
1c213bd2
WC
2257})
2258
aabc92bb 2259#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2260 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2261
764f5e54
JP
2262enum netdev_lag_tx_type {
2263 NETDEV_LAG_TX_TYPE_UNKNOWN,
2264 NETDEV_LAG_TX_TYPE_RANDOM,
2265 NETDEV_LAG_TX_TYPE_BROADCAST,
2266 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2267 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2268 NETDEV_LAG_TX_TYPE_HASH,
2269};
2270
2271struct netdev_lag_upper_info {
2272 enum netdev_lag_tx_type tx_type;
2273};
2274
fb1b2e3c
JP
2275struct netdev_lag_lower_state_info {
2276 u8 link_up : 1,
2277 tx_enabled : 1;
2278};
2279
1da177e4
LT
2280#include <linux/notifier.h>
2281
dcfe1421
AW
2282/* netdevice notifier chain. Please remember to update the rtnetlink
2283 * notification exclusion list in rtnetlink_event() when adding new
2284 * types.
2285 */
2286#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2287#define NETDEV_DOWN 0x0002
2288#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2289 detected a hardware crash and restarted
2290 - we can use this eg to kick tcp sessions
2291 once done */
2292#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2293#define NETDEV_REGISTER 0x0005
2294#define NETDEV_UNREGISTER 0x0006
1d486bfb 2295#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2296#define NETDEV_CHANGEADDR 0x0008
2297#define NETDEV_GOING_DOWN 0x0009
2298#define NETDEV_CHANGENAME 0x000A
2299#define NETDEV_FEAT_CHANGE 0x000B
2300#define NETDEV_BONDING_FAILOVER 0x000C
2301#define NETDEV_PRE_UP 0x000D
2302#define NETDEV_PRE_TYPE_CHANGE 0x000E
2303#define NETDEV_POST_TYPE_CHANGE 0x000F
2304#define NETDEV_POST_INIT 0x0010
0115e8e3 2305#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2306#define NETDEV_RELEASE 0x0012
2307#define NETDEV_NOTIFY_PEERS 0x0013
2308#define NETDEV_JOIN 0x0014
42e52bf9 2309#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2310#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2311#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2312#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2313#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2314#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2315#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2316#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2317#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2318
f629d208
JP
2319int register_netdevice_notifier(struct notifier_block *nb);
2320int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2321
2322struct netdev_notifier_info {
2323 struct net_device *dev;
2324};
2325
be9efd36
JP
2326struct netdev_notifier_change_info {
2327 struct netdev_notifier_info info; /* must be first */
2328 unsigned int flags_changed;
2329};
2330
0e4ead9d
JP
2331struct netdev_notifier_changeupper_info {
2332 struct netdev_notifier_info info; /* must be first */
2333 struct net_device *upper_dev; /* new upper dev */
2334 bool master; /* is upper dev master */
5e82b4b2 2335 bool linking; /* is the notification for link or unlink */
29bf24af 2336 void *upper_info; /* upper dev info */
0e4ead9d
JP
2337};
2338
04d48266
JP
2339struct netdev_notifier_changelowerstate_info {
2340 struct netdev_notifier_info info; /* must be first */
2341 void *lower_state_info; /* is lower dev state */
2342};
2343
75538c2b
CW
2344static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2345 struct net_device *dev)
2346{
2347 info->dev = dev;
2348}
2349
351638e7
JP
2350static inline struct net_device *
2351netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2352{
2353 return info->dev;
2354}
2355
f629d208 2356int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2357
2358
1da177e4
LT
2359extern rwlock_t dev_base_lock; /* Device list lock */
2360
881d966b
EB
2361#define for_each_netdev(net, d) \
2362 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2363#define for_each_netdev_reverse(net, d) \
2364 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2365#define for_each_netdev_rcu(net, d) \
2366 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2367#define for_each_netdev_safe(net, d, n) \
2368 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2369#define for_each_netdev_continue(net, d) \
2370 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2371#define for_each_netdev_continue_rcu(net, d) \
2372 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2373#define for_each_netdev_in_bond_rcu(bond, slave) \
2374 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2375 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2376#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2377
a050c33f
DL
2378static inline struct net_device *next_net_device(struct net_device *dev)
2379{
2380 struct list_head *lh;
2381 struct net *net;
2382
c346dca1 2383 net = dev_net(dev);
a050c33f
DL
2384 lh = dev->dev_list.next;
2385 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2386}
2387
ce81b76a
ED
2388static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2389{
2390 struct list_head *lh;
2391 struct net *net;
2392
2393 net = dev_net(dev);
ccf43438 2394 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2395 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2396}
2397
a050c33f
DL
2398static inline struct net_device *first_net_device(struct net *net)
2399{
2400 return list_empty(&net->dev_base_head) ? NULL :
2401 net_device_entry(net->dev_base_head.next);
2402}
7562f876 2403
ccf43438
ED
2404static inline struct net_device *first_net_device_rcu(struct net *net)
2405{
2406 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2407
2408 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2409}
2410
f629d208
JP
2411int netdev_boot_setup_check(struct net_device *dev);
2412unsigned long netdev_boot_base(const char *prefix, int unit);
2413struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2414 const char *hwaddr);
2415struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2416struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2417void dev_add_pack(struct packet_type *pt);
2418void dev_remove_pack(struct packet_type *pt);
2419void __dev_remove_pack(struct packet_type *pt);
2420void dev_add_offload(struct packet_offload *po);
2421void dev_remove_offload(struct packet_offload *po);
f629d208 2422
a54acb3a 2423int dev_get_iflink(const struct net_device *dev);
fc4099f1 2424int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2425struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2426 unsigned short mask);
f629d208
JP
2427struct net_device *dev_get_by_name(struct net *net, const char *name);
2428struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2429struct net_device *__dev_get_by_name(struct net *net, const char *name);
2430int dev_alloc_name(struct net_device *dev, const char *name);
2431int dev_open(struct net_device *dev);
2432int dev_close(struct net_device *dev);
99c4a26a 2433int dev_close_many(struct list_head *head, bool unlink);
f629d208 2434void dev_disable_lro(struct net_device *dev);
0c4b51f0 2435int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2436int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2437int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2438int register_netdevice(struct net_device *dev);
2439void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2440void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2441static inline void unregister_netdevice(struct net_device *dev)
2442{
2443 unregister_netdevice_queue(dev, NULL);
2444}
2445
f629d208
JP
2446int netdev_refcnt_read(const struct net_device *dev);
2447void free_netdev(struct net_device *dev);
74d332c1 2448void netdev_freemem(struct net_device *dev);
f629d208
JP
2449void synchronize_net(void);
2450int init_dummy_netdev(struct net_device *dev);
937f1ba5 2451
f60e5990 2452DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2453#define XMIT_RECURSION_LIMIT 10
2454
f60e5990 2455static inline int dev_recursion_level(void)
2456{
2457 return this_cpu_read(xmit_recursion);
2458}
2459
f629d208
JP
2460struct net_device *dev_get_by_index(struct net *net, int ifindex);
2461struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2462struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2463struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2464int netdev_get_name(struct net *net, char *name, int ifindex);
2465int dev_restart(struct net_device *dev);
f629d208 2466int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2467
2468static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2469{
2470 return NAPI_GRO_CB(skb)->data_offset;
2471}
2472
2473static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2474{
2475 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2476}
2477
2478static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2479{
2480 NAPI_GRO_CB(skb)->data_offset += len;
2481}
2482
a5b1cf28
HX
2483static inline void *skb_gro_header_fast(struct sk_buff *skb,
2484 unsigned int offset)
86911732 2485{
a5b1cf28
HX
2486 return NAPI_GRO_CB(skb)->frag0 + offset;
2487}
78a478d0 2488
a5b1cf28
HX
2489static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2490{
2491 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2492}
78a478d0 2493
57ea52a8
HX
2494static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2495{
2496 NAPI_GRO_CB(skb)->frag0 = NULL;
2497 NAPI_GRO_CB(skb)->frag0_len = 0;
2498}
2499
a5b1cf28
HX
2500static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2501 unsigned int offset)
2502{
17dd759c
HX
2503 if (!pskb_may_pull(skb, hlen))
2504 return NULL;
2505
57ea52a8 2506 skb_gro_frag0_invalidate(skb);
17dd759c 2507 return skb->data + offset;
86911732 2508}
1da177e4 2509
36e7b1b8
HX
2510static inline void *skb_gro_network_header(struct sk_buff *skb)
2511{
78d3fd0b
HX
2512 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2513 skb_network_offset(skb);
36e7b1b8
HX
2514}
2515
bf5a755f
JC
2516static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2517 const void *start, unsigned int len)
2518{
573e8fca 2519 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2520 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2521 csum_partial(start, len, 0));
2522}
2523
573e8fca
TH
2524/* GRO checksum functions. These are logical equivalents of the normal
2525 * checksum functions (in skbuff.h) except that they operate on the GRO
2526 * offsets and fields in sk_buff.
2527 */
2528
2529__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2530
15e2396d
TH
2531static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2532{
b7fe10e5 2533 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2534}
2535
573e8fca
TH
2536static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2537 bool zero_okay,
2538 __sum16 check)
2539{
6edec0e6
TH
2540 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2541 skb_checksum_start_offset(skb) <
2542 skb_gro_offset(skb)) &&
15e2396d 2543 !skb_at_gro_remcsum_start(skb) &&
662880f4 2544 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2545 (!zero_okay || check));
2546}
2547
2548static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2549 __wsum psum)
2550{
2551 if (NAPI_GRO_CB(skb)->csum_valid &&
2552 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2553 return 0;
2554
2555 NAPI_GRO_CB(skb)->csum = psum;
2556
2557 return __skb_gro_checksum_complete(skb);
2558}
2559
573e8fca
TH
2560static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2561{
662880f4
TH
2562 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2563 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2564 NAPI_GRO_CB(skb)->csum_cnt--;
2565 } else {
2566 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2567 * verified a new top level checksum or an encapsulated one
2568 * during GRO. This saves work if we fallback to normal path.
2569 */
2570 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2571 }
2572}
2573
2574#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2575 compute_pseudo) \
2576({ \
2577 __sum16 __ret = 0; \
2578 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2579 __ret = __skb_gro_checksum_validate_complete(skb, \
2580 compute_pseudo(skb, proto)); \
219f1d79 2581 if (!__ret) \
573e8fca
TH
2582 skb_gro_incr_csum_unnecessary(skb); \
2583 __ret; \
2584})
2585
2586#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2587 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2588
2589#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2590 compute_pseudo) \
2591 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2592
2593#define skb_gro_checksum_simple_validate(skb) \
2594 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2595
d96535a1
TH
2596static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2597{
2598 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2599 !NAPI_GRO_CB(skb)->csum_valid);
2600}
2601
2602static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2603 __sum16 check, __wsum pseudo)
2604{
2605 NAPI_GRO_CB(skb)->csum = ~pseudo;
2606 NAPI_GRO_CB(skb)->csum_valid = 1;
2607}
2608
2609#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2610do { \
2611 if (__skb_gro_checksum_convert_check(skb)) \
2612 __skb_gro_checksum_convert(skb, check, \
2613 compute_pseudo(skb, proto)); \
2614} while (0)
2615
26c4f7da
TH
2616struct gro_remcsum {
2617 int offset;
2618 __wsum delta;
2619};
2620
2621static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2622{
846cd667 2623 grc->offset = 0;
26c4f7da
TH
2624 grc->delta = 0;
2625}
2626
b7fe10e5
TH
2627static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2628 unsigned int off, size_t hdrlen,
2629 int start, int offset,
2630 struct gro_remcsum *grc,
2631 bool nopartial)
dcdc8994
TH
2632{
2633 __wsum delta;
b7fe10e5 2634 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2635
2636 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2637
15e2396d 2638 if (!nopartial) {
b7fe10e5
TH
2639 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2640 return ptr;
2641 }
2642
2643 ptr = skb_gro_header_fast(skb, off);
2644 if (skb_gro_header_hard(skb, off + plen)) {
2645 ptr = skb_gro_header_slow(skb, off + plen, off);
2646 if (!ptr)
2647 return NULL;
15e2396d
TH
2648 }
2649
b7fe10e5
TH
2650 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2651 start, offset);
dcdc8994
TH
2652
2653 /* Adjust skb->csum since we changed the packet */
dcdc8994 2654 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2655
b7fe10e5 2656 grc->offset = off + hdrlen + offset;
26c4f7da 2657 grc->delta = delta;
b7fe10e5
TH
2658
2659 return ptr;
dcdc8994
TH
2660}
2661
26c4f7da
TH
2662static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2663 struct gro_remcsum *grc)
2664{
b7fe10e5
TH
2665 void *ptr;
2666 size_t plen = grc->offset + sizeof(u16);
2667
26c4f7da
TH
2668 if (!grc->delta)
2669 return;
2670
b7fe10e5
TH
2671 ptr = skb_gro_header_fast(skb, grc->offset);
2672 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2673 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2674 if (!ptr)
2675 return;
2676 }
2677
2678 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2679}
dcdc8994 2680
25393d3f
SK
2681#ifdef CONFIG_XFRM_OFFLOAD
2682static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2683{
2684 if (PTR_ERR(pp) != -EINPROGRESS)
2685 NAPI_GRO_CB(skb)->flush |= flush;
2686}
2687#else
5f114163
SK
2688static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2689{
2690 NAPI_GRO_CB(skb)->flush |= flush;
2691}
25393d3f 2692#endif
5f114163 2693
0c4e8581
SH
2694static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2695 unsigned short type,
3b04ddde 2696 const void *daddr, const void *saddr,
95c96174 2697 unsigned int len)
0c4e8581 2698{
f1ecfd5d 2699 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2700 return 0;
3b04ddde
SH
2701
2702 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2703}
2704
b95cce35
SH
2705static inline int dev_parse_header(const struct sk_buff *skb,
2706 unsigned char *haddr)
2707{
2708 const struct net_device *dev = skb->dev;
2709
1b83336b 2710 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2711 return 0;
3b04ddde 2712 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2713}
2714
2793a23a
WB
2715/* ll_header must have at least hard_header_len allocated */
2716static inline bool dev_validate_header(const struct net_device *dev,
2717 char *ll_header, int len)
2718{
2719 if (likely(len >= dev->hard_header_len))
2720 return true;
217e6fa2
WB
2721 if (len < dev->min_header_len)
2722 return false;
2793a23a
WB
2723
2724 if (capable(CAP_SYS_RAWIO)) {
2725 memset(ll_header + len, 0, dev->hard_header_len - len);
2726 return true;
2727 }
2728
2729 if (dev->header_ops && dev->header_ops->validate)
2730 return dev->header_ops->validate(ll_header, len);
2731
2732 return false;
2733}
2734
1da177e4 2735typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2736int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2737static inline int unregister_gifconf(unsigned int family)
2738{
2739 return register_gifconf(family, NULL);
2740}
2741
99bbc707 2742#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2743#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2744struct sd_flow_limit {
2745 u64 count;
2746 unsigned int num_buckets;
2747 unsigned int history_head;
2748 u16 history[FLOW_LIMIT_HISTORY];
2749 u8 buckets[];
2750};
2751
2752extern int netdev_flow_limit_table_len;
2753#endif /* CONFIG_NET_FLOW_LIMIT */
2754
1da177e4 2755/*
5e82b4b2 2756 * Incoming packets are placed on per-CPU queues
1da177e4 2757 */
d94d9fee 2758struct softnet_data {
1da177e4 2759 struct list_head poll_list;
6e7676c1 2760 struct sk_buff_head process_queue;
1da177e4 2761
dee42870 2762 /* stats */
cd7b5396
DM
2763 unsigned int processed;
2764 unsigned int time_squeeze;
cd7b5396 2765 unsigned int received_rps;
fd793d89 2766#ifdef CONFIG_RPS
88751275 2767 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2768#endif
2769#ifdef CONFIG_NET_FLOW_LIMIT
2770 struct sd_flow_limit __rcu *flow_limit;
2771#endif
2772 struct Qdisc *output_queue;
2773 struct Qdisc **output_queue_tailp;
2774 struct sk_buff *completion_queue;
88751275 2775
4cdb1e2e 2776#ifdef CONFIG_RPS
501e7ef5
ED
2777 /* input_queue_head should be written by cpu owning this struct,
2778 * and only read by other cpus. Worth using a cache line.
2779 */
2780 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2781
2782 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2783 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2784 struct softnet_data *rps_ipi_next;
2785 unsigned int cpu;
76cc8b13 2786 unsigned int input_queue_tail;
1e94d72f 2787#endif
95c96174 2788 unsigned int dropped;
0a9627f2 2789 struct sk_buff_head input_pkt_queue;
bea3348e 2790 struct napi_struct backlog;
99bbc707 2791
1da177e4
LT
2792};
2793
76cc8b13 2794static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2795{
2796#ifdef CONFIG_RPS
76cc8b13
TH
2797 sd->input_queue_head++;
2798#endif
2799}
2800
2801static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2802 unsigned int *qtail)
2803{
2804#ifdef CONFIG_RPS
2805 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2806#endif
2807}
2808
0a9627f2 2809DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2810
f629d208 2811void __netif_schedule(struct Qdisc *q);
46e5da40 2812void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2813
fd2ea0a7
DM
2814static inline void netif_tx_schedule_all(struct net_device *dev)
2815{
2816 unsigned int i;
2817
2818 for (i = 0; i < dev->num_tx_queues; i++)
2819 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2820}
2821
f9a7cbbf 2822static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2823{
73466498 2824 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2825}
2826
bea3348e
SH
2827/**
2828 * netif_start_queue - allow transmit
2829 * @dev: network device
2830 *
2831 * Allow upper layers to call the device hard_start_xmit routine.
2832 */
1da177e4
LT
2833static inline void netif_start_queue(struct net_device *dev)
2834{
e8a0464c 2835 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2836}
2837
fd2ea0a7
DM
2838static inline void netif_tx_start_all_queues(struct net_device *dev)
2839{
2840 unsigned int i;
2841
2842 for (i = 0; i < dev->num_tx_queues; i++) {
2843 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2844 netif_tx_start_queue(txq);
2845 }
2846}
2847
46e5da40 2848void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2849
d29f749e
DJ
2850/**
2851 * netif_wake_queue - restart transmit
2852 * @dev: network device
2853 *
2854 * Allow upper layers to call the device hard_start_xmit routine.
2855 * Used for flow control when transmit resources are available.
2856 */
79d16385
DM
2857static inline void netif_wake_queue(struct net_device *dev)
2858{
e8a0464c 2859 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2860}
2861
fd2ea0a7
DM
2862static inline void netif_tx_wake_all_queues(struct net_device *dev)
2863{
2864 unsigned int i;
2865
2866 for (i = 0; i < dev->num_tx_queues; i++) {
2867 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2868 netif_tx_wake_queue(txq);
2869 }
2870}
2871
f9a7cbbf 2872static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2873{
73466498 2874 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2875}
2876
bea3348e
SH
2877/**
2878 * netif_stop_queue - stop transmitted packets
2879 * @dev: network device
2880 *
2881 * Stop upper layers calling the device hard_start_xmit routine.
2882 * Used for flow control when transmit resources are unavailable.
2883 */
1da177e4
LT
2884static inline void netif_stop_queue(struct net_device *dev)
2885{
e8a0464c 2886 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2887}
2888
a2029240 2889void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2890
4d29515f 2891static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2892{
73466498 2893 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2894}
2895
bea3348e
SH
2896/**
2897 * netif_queue_stopped - test if transmit queue is flowblocked
2898 * @dev: network device
2899 *
2900 * Test if transmit queue on device is currently unable to send.
2901 */
4d29515f 2902static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2903{
e8a0464c 2904 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2905}
2906
4d29515f 2907static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2908{
73466498
TH
2909 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2910}
2911
8e2f1a63
DB
2912static inline bool
2913netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2914{
2915 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2916}
2917
8e2f1a63
DB
2918static inline bool
2919netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2920{
2921 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2922}
2923
53511453
ED
2924/**
2925 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2926 * @dev_queue: pointer to transmit queue
2927 *
2928 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2929 * to give appropriate hint to the CPU.
53511453
ED
2930 */
2931static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2932{
2933#ifdef CONFIG_BQL
2934 prefetchw(&dev_queue->dql.num_queued);
2935#endif
2936}
2937
2938/**
2939 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2940 * @dev_queue: pointer to transmit queue
2941 *
2942 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2943 * to give appropriate hint to the CPU.
53511453
ED
2944 */
2945static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2946{
2947#ifdef CONFIG_BQL
2948 prefetchw(&dev_queue->dql.limit);
2949#endif
2950}
2951
c5d67bd7
TH
2952static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2953 unsigned int bytes)
2954{
114cf580
TH
2955#ifdef CONFIG_BQL
2956 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2957
2958 if (likely(dql_avail(&dev_queue->dql) >= 0))
2959 return;
2960
2961 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2962
2963 /*
2964 * The XOFF flag must be set before checking the dql_avail below,
2965 * because in netdev_tx_completed_queue we update the dql_completed
2966 * before checking the XOFF flag.
2967 */
2968 smp_mb();
2969
2970 /* check again in case another CPU has just made room avail */
2971 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2972 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2973#endif
c5d67bd7
TH
2974}
2975
0042d0c8
FF
2976/**
2977 * netdev_sent_queue - report the number of bytes queued to hardware
2978 * @dev: network device
2979 * @bytes: number of bytes queued to the hardware device queue
2980 *
2981 * Report the number of bytes queued for sending/completion to the network
2982 * device hardware queue. @bytes should be a good approximation and should
2983 * exactly match netdev_completed_queue() @bytes
2984 */
c5d67bd7
TH
2985static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2986{
2987 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2988}
2989
2990static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2991 unsigned int pkts, unsigned int bytes)
c5d67bd7 2992{
114cf580 2993#ifdef CONFIG_BQL
b37c0fbe
AD
2994 if (unlikely(!bytes))
2995 return;
2996
2997 dql_completed(&dev_queue->dql, bytes);
2998
2999 /*
3000 * Without the memory barrier there is a small possiblity that
3001 * netdev_tx_sent_queue will miss the update and cause the queue to
3002 * be stopped forever
3003 */
3004 smp_mb();
3005
3006 if (dql_avail(&dev_queue->dql) < 0)
3007 return;
3008
3009 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3010 netif_schedule_queue(dev_queue);
114cf580 3011#endif
c5d67bd7
TH
3012}
3013
0042d0c8
FF
3014/**
3015 * netdev_completed_queue - report bytes and packets completed by device
3016 * @dev: network device
3017 * @pkts: actual number of packets sent over the medium
3018 * @bytes: actual number of bytes sent over the medium
3019 *
3020 * Report the number of bytes and packets transmitted by the network device
3021 * hardware queue over the physical medium, @bytes must exactly match the
3022 * @bytes amount passed to netdev_sent_queue()
3023 */
c5d67bd7 3024static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3025 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3026{
3027 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3028}
3029
3030static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3031{
114cf580 3032#ifdef CONFIG_BQL
5c490354 3033 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3034 dql_reset(&q->dql);
3035#endif
c5d67bd7
TH
3036}
3037
0042d0c8
FF
3038/**
3039 * netdev_reset_queue - reset the packets and bytes count of a network device
3040 * @dev_queue: network device
3041 *
3042 * Reset the bytes and packet count of a network device and clear the
3043 * software flow control OFF bit for this network device
3044 */
c5d67bd7
TH
3045static inline void netdev_reset_queue(struct net_device *dev_queue)
3046{
3047 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3048}
3049
b9507bda
DB
3050/**
3051 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3052 * @dev: network device
3053 * @queue_index: given tx queue index
3054 *
3055 * Returns 0 if given tx queue index >= number of device tx queues,
3056 * otherwise returns the originally passed tx queue index.
3057 */
3058static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3059{
3060 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3061 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3062 dev->name, queue_index,
3063 dev->real_num_tx_queues);
3064 return 0;
3065 }
3066
3067 return queue_index;
3068}
3069
bea3348e
SH
3070/**
3071 * netif_running - test if up
3072 * @dev: network device
3073 *
3074 * Test if the device has been brought up.
3075 */
4d29515f 3076static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3077{
3078 return test_bit(__LINK_STATE_START, &dev->state);
3079}
3080
f25f4e44 3081/*
5e82b4b2 3082 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3083 * stop, and a check if it's stopped. All other device management is
3084 * done at the overall netdevice level.
3085 * Also test the device if we're multiqueue.
3086 */
bea3348e
SH
3087
3088/**
3089 * netif_start_subqueue - allow sending packets on subqueue
3090 * @dev: network device
3091 * @queue_index: sub queue index
3092 *
3093 * Start individual transmit queue of a device with multiple transmit queues.
3094 */
f25f4e44
PWJ
3095static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3096{
fd2ea0a7 3097 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3098
3099 netif_tx_start_queue(txq);
f25f4e44
PWJ
3100}
3101
bea3348e
SH
3102/**
3103 * netif_stop_subqueue - stop sending packets on subqueue
3104 * @dev: network device
3105 * @queue_index: sub queue index
3106 *
3107 * Stop individual transmit queue of a device with multiple transmit queues.
3108 */
f25f4e44
PWJ
3109static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3110{
fd2ea0a7 3111 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3112 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3113}
3114
bea3348e
SH
3115/**
3116 * netif_subqueue_stopped - test status of subqueue
3117 * @dev: network device
3118 * @queue_index: sub queue index
3119 *
3120 * Check individual transmit queue of a device with multiple transmit queues.
3121 */
4d29515f
DM
3122static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3123 u16 queue_index)
f25f4e44 3124{
fd2ea0a7 3125 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3126
3127 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3128}
3129
4d29515f
DM
3130static inline bool netif_subqueue_stopped(const struct net_device *dev,
3131 struct sk_buff *skb)
668f895a
PE
3132{
3133 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3134}
bea3348e 3135
738b35cc
FF
3136/**
3137 * netif_wake_subqueue - allow sending packets on subqueue
3138 * @dev: network device
3139 * @queue_index: sub queue index
3140 *
3141 * Resume individual transmit queue of a device with multiple transmit queues.
3142 */
3143static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3144{
3145 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3146
3147 netif_tx_wake_queue(txq);
3148}
f25f4e44 3149
537c00de 3150#ifdef CONFIG_XPS
53af53ae 3151int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3152 u16 index);
537c00de
AD
3153#else
3154static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3155 const struct cpumask *mask,
537c00de
AD
3156 u16 index)
3157{
3158 return 0;
3159}
3160#endif
3161
5605c762
JP
3162u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3163 unsigned int num_tx_queues);
3164
a3d22a68
VZ
3165/*
3166 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3167 * as a distribution range limit for the returned value.
3168 */
3169static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3170 struct sk_buff *skb)
a3d22a68
VZ
3171{
3172 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3173}
3174
bea3348e
SH
3175/**
3176 * netif_is_multiqueue - test if device has multiple transmit queues
3177 * @dev: network device
3178 *
3179 * Check if device has multiple transmit queues
bea3348e 3180 */
4d29515f 3181static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3182{
a02cec21 3183 return dev->num_tx_queues > 1;
f25f4e44 3184}
1da177e4 3185
f629d208 3186int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3187
a953be53 3188#ifdef CONFIG_SYSFS
f629d208 3189int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3190#else
3191static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3192 unsigned int rxq)
3193{
3194 return 0;
3195}
3196#endif
3197
a953be53
MD
3198#ifdef CONFIG_SYSFS
3199static inline unsigned int get_netdev_rx_queue_index(
3200 struct netdev_rx_queue *queue)
3201{
3202 struct net_device *dev = queue->dev;
3203 int index = queue - dev->_rx;
3204
3205 BUG_ON(index >= dev->num_rx_queues);
3206 return index;
3207}
3208#endif
3209
16917b87 3210#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3211int netif_get_num_default_rss_queues(void);
16917b87 3212
e6247027
ED
3213enum skb_free_reason {
3214 SKB_REASON_CONSUMED,
3215 SKB_REASON_DROPPED,
3216};
3217
3218void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3219void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3220
e6247027
ED
3221/*
3222 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3223 * interrupt context or with hardware interrupts being disabled.
3224 * (in_irq() || irqs_disabled())
3225 *
3226 * We provide four helpers that can be used in following contexts :
3227 *
3228 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3229 * replacing kfree_skb(skb)
3230 *
3231 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3232 * Typically used in place of consume_skb(skb) in TX completion path
3233 *
3234 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3235 * replacing kfree_skb(skb)
3236 *
3237 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3238 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3239 */
e6247027
ED
3240static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3241{
3242 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3243}
3244
3245static inline void dev_consume_skb_irq(struct sk_buff *skb)
3246{
3247 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3248}
3249
3250static inline void dev_kfree_skb_any(struct sk_buff *skb)
3251{
3252 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3253}
3254
3255static inline void dev_consume_skb_any(struct sk_buff *skb)
3256{
3257 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3258}
1da177e4 3259
f629d208
JP
3260int netif_rx(struct sk_buff *skb);
3261int netif_rx_ni(struct sk_buff *skb);
04eb4489 3262int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3263gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3264void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3265struct sk_buff *napi_get_frags(struct napi_struct *napi);
3266gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3267struct packet_offload *gro_find_receive_by_type(__be16 type);
3268struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3269
3270static inline void napi_free_frags(struct napi_struct *napi)
3271{
3272 kfree_skb(napi->skb);
3273 napi->skb = NULL;
3274}
3275
24b27fc4 3276bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3277int netdev_rx_handler_register(struct net_device *dev,
3278 rx_handler_func_t *rx_handler,
3279 void *rx_handler_data);
3280void netdev_rx_handler_unregister(struct net_device *dev);
3281
3282bool dev_valid_name(const char *name);
3283int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3284int dev_ethtool(struct net *net, struct ifreq *);
3285unsigned int dev_get_flags(const struct net_device *);
3286int __dev_change_flags(struct net_device *, unsigned int flags);
3287int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3288void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3289 unsigned int gchanges);
f629d208
JP
3290int dev_change_name(struct net_device *, const char *);
3291int dev_set_alias(struct net_device *, const char *, size_t);
3292int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3293int dev_set_mtu(struct net_device *, int);
3294void dev_set_group(struct net_device *, int);
3295int dev_set_mac_address(struct net_device *, struct sockaddr *);
3296int dev_change_carrier(struct net_device *, bool new_carrier);
3297int dev_get_phys_port_id(struct net_device *dev,
02637fce 3298 struct netdev_phys_item_id *ppid);
db24a904
DA
3299int dev_get_phys_port_name(struct net_device *dev,
3300 char *name, size_t len);
d746d707 3301int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3302struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3303struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3304 struct netdev_queue *txq, int *ret);
d67b9cd2
DB
3305
3306typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3307int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3308 int fd, u32 flags);
58038695 3309bool __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
d67b9cd2 3310
a0265d28 3311int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3312int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3313bool is_skb_forwardable(const struct net_device *dev,
3314 const struct sk_buff *skb);
1da177e4 3315
4e3264d2
MKL
3316static __always_inline int ____dev_forward_skb(struct net_device *dev,
3317 struct sk_buff *skb)
3318{
3319 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3320 unlikely(!is_skb_forwardable(dev, skb))) {
3321 atomic_long_inc(&dev->rx_dropped);
3322 kfree_skb(skb);
3323 return NET_RX_DROP;
3324 }
3325
3326 skb_scrub_packet(skb, true);
3327 skb->priority = 0;
3328 return 0;
3329}
3330
74b20582
DA
3331void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3332
20380731 3333extern int netdev_budget;
7acf8a1e 3334extern unsigned int netdev_budget_usecs;
1da177e4
LT
3335
3336/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3337void netdev_run_todo(void);
1da177e4 3338
bea3348e
SH
3339/**
3340 * dev_put - release reference to device
3341 * @dev: network device
3342 *
9ef4429b 3343 * Release reference to device to allow it to be freed.
bea3348e 3344 */
1da177e4
LT
3345static inline void dev_put(struct net_device *dev)
3346{
933393f5 3347 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3348}
3349
bea3348e
SH
3350/**
3351 * dev_hold - get reference to device
3352 * @dev: network device
3353 *
9ef4429b 3354 * Hold reference to device to keep it from being freed.
bea3348e 3355 */
15333061
SH
3356static inline void dev_hold(struct net_device *dev)
3357{
933393f5 3358 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3359}
1da177e4
LT
3360
3361/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3362 * and _off may be called from IRQ context, but it is caller
3363 * who is responsible for serialization of these calls.
b00055aa
SR
3364 *
3365 * The name carrier is inappropriate, these functions should really be
3366 * called netif_lowerlayer_*() because they represent the state of any
3367 * kind of lower layer not just hardware media.
1da177e4
LT
3368 */
3369
f629d208
JP
3370void linkwatch_init_dev(struct net_device *dev);
3371void linkwatch_fire_event(struct net_device *dev);
3372void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3373
bea3348e
SH
3374/**
3375 * netif_carrier_ok - test if carrier present
3376 * @dev: network device
3377 *
3378 * Check if carrier is present on device
3379 */
4d29515f 3380static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3381{
3382 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3383}
3384
f629d208 3385unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3386
f629d208 3387void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3388
f629d208 3389void netif_carrier_on(struct net_device *dev);
1da177e4 3390
f629d208 3391void netif_carrier_off(struct net_device *dev);
1da177e4 3392
bea3348e
SH
3393/**
3394 * netif_dormant_on - mark device as dormant.
3395 * @dev: network device
3396 *
3397 * Mark device as dormant (as per RFC2863).
3398 *
3399 * The dormant state indicates that the relevant interface is not
3400 * actually in a condition to pass packets (i.e., it is not 'up') but is
3401 * in a "pending" state, waiting for some external event. For "on-
3402 * demand" interfaces, this new state identifies the situation where the
3403 * interface is waiting for events to place it in the up state.
bea3348e 3404 */
b00055aa
SR
3405static inline void netif_dormant_on(struct net_device *dev)
3406{
3407 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3408 linkwatch_fire_event(dev);
3409}
3410
bea3348e
SH
3411/**
3412 * netif_dormant_off - set device as not dormant.
3413 * @dev: network device
3414 *
3415 * Device is not in dormant state.
3416 */
b00055aa
SR
3417static inline void netif_dormant_off(struct net_device *dev)
3418{
3419 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3420 linkwatch_fire_event(dev);
3421}
3422
bea3348e 3423/**
8ecbc40a 3424 * netif_dormant - test if device is dormant
bea3348e
SH
3425 * @dev: network device
3426 *
8ecbc40a 3427 * Check if device is dormant.
bea3348e 3428 */
4d29515f 3429static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3430{
3431 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3432}
3433
3434
bea3348e
SH
3435/**
3436 * netif_oper_up - test if device is operational
3437 * @dev: network device
3438 *
3439 * Check if carrier is operational
3440 */
4d29515f 3441static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3442{
b00055aa
SR
3443 return (dev->operstate == IF_OPER_UP ||
3444 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3445}
3446
bea3348e
SH
3447/**
3448 * netif_device_present - is device available or removed
3449 * @dev: network device
3450 *
3451 * Check if device has not been removed from system.
3452 */
4d29515f 3453static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3454{
3455 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3456}
3457
f629d208 3458void netif_device_detach(struct net_device *dev);
1da177e4 3459
f629d208 3460void netif_device_attach(struct net_device *dev);
1da177e4
LT
3461
3462/*
3463 * Network interface message level settings
3464 */
1da177e4
LT
3465
3466enum {
3467 NETIF_MSG_DRV = 0x0001,
3468 NETIF_MSG_PROBE = 0x0002,
3469 NETIF_MSG_LINK = 0x0004,
3470 NETIF_MSG_TIMER = 0x0008,
3471 NETIF_MSG_IFDOWN = 0x0010,
3472 NETIF_MSG_IFUP = 0x0020,
3473 NETIF_MSG_RX_ERR = 0x0040,
3474 NETIF_MSG_TX_ERR = 0x0080,
3475 NETIF_MSG_TX_QUEUED = 0x0100,
3476 NETIF_MSG_INTR = 0x0200,
3477 NETIF_MSG_TX_DONE = 0x0400,
3478 NETIF_MSG_RX_STATUS = 0x0800,
3479 NETIF_MSG_PKTDATA = 0x1000,
3480 NETIF_MSG_HW = 0x2000,
3481 NETIF_MSG_WOL = 0x4000,
3482};
3483
3484#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3485#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3486#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3487#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3488#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3489#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3490#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3491#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3492#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3493#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3494#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3495#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3496#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3497#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3498#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3499
3500static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3501{
3502 /* use default */
3503 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3504 return default_msg_enable_bits;
3505 if (debug_value == 0) /* no output */
3506 return 0;
3507 /* set low N bits */
3508 return (1 << debug_value) - 1;
3509}
3510
c773e847 3511static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3512{
c773e847
DM
3513 spin_lock(&txq->_xmit_lock);
3514 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3515}
3516
5a717f4f
MT
3517static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3518{
3519 __acquire(&txq->_xmit_lock);
3520 return true;
3521}
3522
3523static inline void __netif_tx_release(struct netdev_queue *txq)
3524{
3525 __release(&txq->_xmit_lock);
3526}
3527
fd2ea0a7
DM
3528static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3529{
3530 spin_lock_bh(&txq->_xmit_lock);
3531 txq->xmit_lock_owner = smp_processor_id();
3532}
3533
4d29515f 3534static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3535{
4d29515f 3536 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3537 if (likely(ok))
3538 txq->xmit_lock_owner = smp_processor_id();
3539 return ok;
3540}
3541
3542static inline void __netif_tx_unlock(struct netdev_queue *txq)
3543{
3544 txq->xmit_lock_owner = -1;
3545 spin_unlock(&txq->_xmit_lock);
3546}
3547
3548static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3549{
3550 txq->xmit_lock_owner = -1;
3551 spin_unlock_bh(&txq->_xmit_lock);
3552}
3553
08baf561
ED
3554static inline void txq_trans_update(struct netdev_queue *txq)
3555{
3556 if (txq->xmit_lock_owner != -1)
3557 txq->trans_start = jiffies;
3558}
3559
ba162f8e
FW
3560/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3561static inline void netif_trans_update(struct net_device *dev)
3562{
9b36627a
FW
3563 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3564
3565 if (txq->trans_start != jiffies)
3566 txq->trans_start = jiffies;
ba162f8e
FW
3567}
3568
d29f749e
DJ
3569/**
3570 * netif_tx_lock - grab network device transmit lock
3571 * @dev: network device
d29f749e
DJ
3572 *
3573 * Get network device transmit lock
3574 */
22dd7495
JHS
3575static inline void netif_tx_lock(struct net_device *dev)
3576{
e8a0464c 3577 unsigned int i;
c3f26a26 3578 int cpu;
c773e847 3579
c3f26a26
DM
3580 spin_lock(&dev->tx_global_lock);
3581 cpu = smp_processor_id();
e8a0464c
DM
3582 for (i = 0; i < dev->num_tx_queues; i++) {
3583 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3584
3585 /* We are the only thread of execution doing a
3586 * freeze, but we have to grab the _xmit_lock in
3587 * order to synchronize with threads which are in
3588 * the ->hard_start_xmit() handler and already
3589 * checked the frozen bit.
3590 */
e8a0464c 3591 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3592 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3593 __netif_tx_unlock(txq);
e8a0464c 3594 }
932ff279
HX
3595}
3596
3597static inline void netif_tx_lock_bh(struct net_device *dev)
3598{
e8a0464c
DM
3599 local_bh_disable();
3600 netif_tx_lock(dev);
932ff279
HX
3601}
3602
932ff279
HX
3603static inline void netif_tx_unlock(struct net_device *dev)
3604{
e8a0464c
DM
3605 unsigned int i;
3606
3607 for (i = 0; i < dev->num_tx_queues; i++) {
3608 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3609
c3f26a26
DM
3610 /* No need to grab the _xmit_lock here. If the
3611 * queue is not stopped for another reason, we
3612 * force a schedule.
3613 */
3614 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3615 netif_schedule_queue(txq);
c3f26a26
DM
3616 }
3617 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3618}
3619
3620static inline void netif_tx_unlock_bh(struct net_device *dev)
3621{
e8a0464c
DM
3622 netif_tx_unlock(dev);
3623 local_bh_enable();
932ff279
HX
3624}
3625
c773e847 3626#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3627 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3628 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3629 } else { \
3630 __netif_tx_acquire(txq); \
22dd7495
JHS
3631 } \
3632}
3633
5efeac44
EB
3634#define HARD_TX_TRYLOCK(dev, txq) \
3635 (((dev->features & NETIF_F_LLTX) == 0) ? \
3636 __netif_tx_trylock(txq) : \
5a717f4f 3637 __netif_tx_acquire(txq))
5efeac44 3638
c773e847 3639#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3640 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3641 __netif_tx_unlock(txq); \
5a717f4f
MT
3642 } else { \
3643 __netif_tx_release(txq); \
22dd7495
JHS
3644 } \
3645}
3646
1da177e4
LT
3647static inline void netif_tx_disable(struct net_device *dev)
3648{
fd2ea0a7 3649 unsigned int i;
c3f26a26 3650 int cpu;
fd2ea0a7 3651
c3f26a26
DM
3652 local_bh_disable();
3653 cpu = smp_processor_id();
fd2ea0a7
DM
3654 for (i = 0; i < dev->num_tx_queues; i++) {
3655 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3656
3657 __netif_tx_lock(txq, cpu);
fd2ea0a7 3658 netif_tx_stop_queue(txq);
c3f26a26 3659 __netif_tx_unlock(txq);
fd2ea0a7 3660 }
c3f26a26 3661 local_bh_enable();
1da177e4
LT
3662}
3663
e308a5d8
DM
3664static inline void netif_addr_lock(struct net_device *dev)
3665{
3666 spin_lock(&dev->addr_list_lock);
3667}
3668
2429f7ac
JP
3669static inline void netif_addr_lock_nested(struct net_device *dev)
3670{
25175ba5
VY
3671 int subclass = SINGLE_DEPTH_NESTING;
3672
3673 if (dev->netdev_ops->ndo_get_lock_subclass)
3674 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3675
3676 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3677}
3678
e308a5d8
DM
3679static inline void netif_addr_lock_bh(struct net_device *dev)
3680{
3681 spin_lock_bh(&dev->addr_list_lock);
3682}
3683
3684static inline void netif_addr_unlock(struct net_device *dev)
3685{
3686 spin_unlock(&dev->addr_list_lock);
3687}
3688
3689static inline void netif_addr_unlock_bh(struct net_device *dev)
3690{
3691 spin_unlock_bh(&dev->addr_list_lock);
3692}
3693
f001fde5 3694/*
31278e71 3695 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3696 * rcu_read_lock held.
3697 */
3698#define for_each_dev_addr(dev, ha) \
31278e71 3699 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3700
1da177e4
LT
3701/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3702
f629d208 3703void ether_setup(struct net_device *dev);
1da177e4
LT
3704
3705/* Support for loadable net-drivers */
f629d208 3706struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3707 unsigned char name_assign_type,
f629d208
JP
3708 void (*setup)(struct net_device *),
3709 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3710#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3711 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3712
c835a677
TG
3713#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3714 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3715 count)
36909ea4 3716
f629d208
JP
3717int register_netdev(struct net_device *dev);
3718void unregister_netdev(struct net_device *dev);
f001fde5 3719
22bedad3 3720/* General hardware address lists handling functions */
f629d208
JP
3721int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3722 struct netdev_hw_addr_list *from_list, int addr_len);
3723void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3724 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3725int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3726 struct net_device *dev,
3727 int (*sync)(struct net_device *, const unsigned char *),
3728 int (*unsync)(struct net_device *,
3729 const unsigned char *));
3730void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3731 struct net_device *dev,
3732 int (*unsync)(struct net_device *,
3733 const unsigned char *));
f629d208 3734void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3735
f001fde5 3736/* Functions used for device addresses handling */
f629d208
JP
3737int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3738 unsigned char addr_type);
3739int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3740 unsigned char addr_type);
f629d208
JP
3741void dev_addr_flush(struct net_device *dev);
3742int dev_addr_init(struct net_device *dev);
a748ee24
JP
3743
3744/* Functions used for unicast addresses handling */
f629d208
JP
3745int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3746int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3747int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3748int dev_uc_sync(struct net_device *to, struct net_device *from);
3749int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3750void dev_uc_unsync(struct net_device *to, struct net_device *from);
3751void dev_uc_flush(struct net_device *dev);
3752void dev_uc_init(struct net_device *dev);
f001fde5 3753
670e5b8e
AD
3754/**
3755 * __dev_uc_sync - Synchonize device's unicast list
3756 * @dev: device to sync
3757 * @sync: function to call if address should be added
3758 * @unsync: function to call if address should be removed
3759 *
3760 * Add newly added addresses to the interface, and release
3761 * addresses that have been deleted.
5e82b4b2 3762 */
670e5b8e
AD
3763static inline int __dev_uc_sync(struct net_device *dev,
3764 int (*sync)(struct net_device *,
3765 const unsigned char *),
3766 int (*unsync)(struct net_device *,
3767 const unsigned char *))
3768{
3769 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3770}
3771
3772/**
e793c0f7 3773 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3774 * @dev: device to sync
3775 * @unsync: function to call if address should be removed
3776 *
3777 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3778 */
670e5b8e
AD
3779static inline void __dev_uc_unsync(struct net_device *dev,
3780 int (*unsync)(struct net_device *,
3781 const unsigned char *))
3782{
3783 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3784}
3785
22bedad3 3786/* Functions used for multicast addresses handling */
f629d208
JP
3787int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3788int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3789int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3790int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3791int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3792int dev_mc_sync(struct net_device *to, struct net_device *from);
3793int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3794void dev_mc_unsync(struct net_device *to, struct net_device *from);
3795void dev_mc_flush(struct net_device *dev);
3796void dev_mc_init(struct net_device *dev);
f001fde5 3797
670e5b8e
AD
3798/**
3799 * __dev_mc_sync - Synchonize device's multicast list
3800 * @dev: device to sync
3801 * @sync: function to call if address should be added
3802 * @unsync: function to call if address should be removed
3803 *
3804 * Add newly added addresses to the interface, and release
3805 * addresses that have been deleted.
5e82b4b2 3806 */
670e5b8e
AD
3807static inline int __dev_mc_sync(struct net_device *dev,
3808 int (*sync)(struct net_device *,
3809 const unsigned char *),
3810 int (*unsync)(struct net_device *,
3811 const unsigned char *))
3812{
3813 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3814}
3815
3816/**
e793c0f7 3817 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3818 * @dev: device to sync
3819 * @unsync: function to call if address should be removed
3820 *
3821 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3822 */
670e5b8e
AD
3823static inline void __dev_mc_unsync(struct net_device *dev,
3824 int (*unsync)(struct net_device *,
3825 const unsigned char *))
3826{
3827 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3828}
3829
4417da66 3830/* Functions used for secondary unicast and multicast support */
f629d208
JP
3831void dev_set_rx_mode(struct net_device *dev);
3832void __dev_set_rx_mode(struct net_device *dev);
3833int dev_set_promiscuity(struct net_device *dev, int inc);
3834int dev_set_allmulti(struct net_device *dev, int inc);
3835void netdev_state_change(struct net_device *dev);
3836void netdev_notify_peers(struct net_device *dev);
3837void netdev_features_change(struct net_device *dev);
1da177e4 3838/* Load a device via the kmod */
f629d208
JP
3839void dev_load(struct net *net, const char *name);
3840struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3841 struct rtnl_link_stats64 *storage);
3842void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3843 const struct net_device_stats *netdev_stats);
eeda3fd6 3844
1da177e4 3845extern int netdev_max_backlog;
3b098e2d 3846extern int netdev_tstamp_prequeue;
1da177e4 3847extern int weight_p;
3d48b53f
MT
3848extern int dev_weight_rx_bias;
3849extern int dev_weight_tx_bias;
3850extern int dev_rx_weight;
3851extern int dev_tx_weight;
9ff162a8 3852
f629d208 3853bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3854struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3855 struct list_head **iter);
f629d208
JP
3856struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3857 struct list_head **iter);
8b5be856 3858
44a40855
VY
3859/* iterate through upper list, must be called under RCU read lock */
3860#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3861 for (iter = &(dev)->adj_list.upper, \
3862 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3863 updev; \
3864 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3865
1a3f060c
DA
3866int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3867 int (*fn)(struct net_device *upper_dev,
3868 void *data),
3869 void *data);
3870
3871bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3872 struct net_device *upper_dev);
3873
f629d208
JP
3874void *netdev_lower_get_next_private(struct net_device *dev,
3875 struct list_head **iter);
3876void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3877 struct list_head **iter);
31088a11
VF
3878
3879#define netdev_for_each_lower_private(dev, priv, iter) \
3880 for (iter = (dev)->adj_list.lower.next, \
3881 priv = netdev_lower_get_next_private(dev, &(iter)); \
3882 priv; \
3883 priv = netdev_lower_get_next_private(dev, &(iter)))
3884
3885#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3886 for (iter = &(dev)->adj_list.lower, \
3887 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3888 priv; \
3889 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3890
4085ebe8
VY
3891void *netdev_lower_get_next(struct net_device *dev,
3892 struct list_head **iter);
7ce856aa 3893
4085ebe8 3894#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3895 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3896 ldev = netdev_lower_get_next(dev, &(iter)); \
3897 ldev; \
3898 ldev = netdev_lower_get_next(dev, &(iter)))
3899
7ce856aa
JP
3900struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3901 struct list_head **iter);
3902struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3903 struct list_head **iter);
3904
1a3f060c
DA
3905int netdev_walk_all_lower_dev(struct net_device *dev,
3906 int (*fn)(struct net_device *lower_dev,
3907 void *data),
3908 void *data);
3909int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3910 int (*fn)(struct net_device *lower_dev,
3911 void *data),
3912 void *data);
3913
f629d208 3914void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3915void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3916struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3917struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3918int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3919int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3920 struct net_device *upper_dev,
29bf24af 3921 void *upper_priv, void *upper_info);
f629d208
JP
3922void netdev_upper_dev_unlink(struct net_device *dev,
3923 struct net_device *upper_dev);
5bb025fa 3924void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3925void *netdev_lower_dev_get_private(struct net_device *dev,
3926 struct net_device *lower_dev);
04d48266
JP
3927void netdev_lower_state_changed(struct net_device *lower_dev,
3928 void *lower_state_info);
960fb622
ED
3929
3930/* RSS keys are 40 or 52 bytes long */
3931#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3932extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3933void netdev_rss_key_fill(void *buffer, size_t len);
3934
952fcfd0 3935int dev_get_nest_level(struct net_device *dev);
f629d208 3936int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 3937int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
3938int skb_csum_hwoffload_help(struct sk_buff *skb,
3939 const netdev_features_t features);
3940
f629d208
JP
3941struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3942 netdev_features_t features, bool tx_path);
3943struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3944 netdev_features_t features);
12b0004d 3945
61bd3857
MS
3946struct netdev_bonding_info {
3947 ifslave slave;
3948 ifbond master;
3949};
3950
3951struct netdev_notifier_bonding_info {
3952 struct netdev_notifier_info info; /* must be first */
3953 struct netdev_bonding_info bonding_info;
3954};
3955
3956void netdev_bonding_info_change(struct net_device *dev,
3957 struct netdev_bonding_info *bonding_info);
3958
12b0004d
CW
3959static inline
3960struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3961{
3962 return __skb_gso_segment(skb, features, true);
3963}
53d6471c 3964__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3965
3966static inline bool can_checksum_protocol(netdev_features_t features,
3967 __be16 protocol)
3968{
c8cd0989
TH
3969 if (protocol == htons(ETH_P_FCOE))
3970 return !!(features & NETIF_F_FCOE_CRC);
3971
3972 /* Assume this is an IP checksum (not SCTP CRC) */
3973
3974 if (features & NETIF_F_HW_CSUM) {
3975 /* Can checksum everything */
3976 return true;
3977 }
3978
3979 switch (protocol) {
3980 case htons(ETH_P_IP):
3981 return !!(features & NETIF_F_IP_CSUM);
3982 case htons(ETH_P_IPV6):
3983 return !!(features & NETIF_F_IPV6_CSUM);
3984 default:
3985 return false;
3986 }
ec5f0615 3987}
12b0004d 3988
fb286bb2 3989#ifdef CONFIG_BUG
f629d208 3990void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3991#else
3992static inline void netdev_rx_csum_fault(struct net_device *dev)
3993{
3994}
3995#endif
1da177e4 3996/* rx skb timestamps */
f629d208
JP
3997void net_enable_timestamp(void);
3998void net_disable_timestamp(void);
1da177e4 3999
20380731 4000#ifdef CONFIG_PROC_FS
f629d208 4001int __init dev_proc_init(void);
900ff8c6
CW
4002#else
4003#define dev_proc_init() 0
20380731
ACM
4004#endif
4005
4798248e 4006static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4007 struct sk_buff *skb, struct net_device *dev,
4008 bool more)
4798248e 4009{
fa2dbdc2 4010 skb->xmit_more = more ? 1 : 0;
0b725a2c 4011 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4012}
4013
10b3ad8c 4014static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4015 struct netdev_queue *txq, bool more)
4798248e
DM
4016{
4017 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4018 int rc;
4798248e 4019
fa2dbdc2 4020 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4021 if (rc == NETDEV_TX_OK)
4022 txq_trans_update(txq);
4023
4024 return rc;
4798248e
DM
4025}
4026
42a2d923
LT
4027int netdev_class_create_file_ns(struct class_attribute *class_attr,
4028 const void *ns);
4029void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4030 const void *ns);
58292cbe
TH
4031
4032static inline int netdev_class_create_file(struct class_attribute *class_attr)
4033{
4034 return netdev_class_create_file_ns(class_attr, NULL);
4035}
4036
4037static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4038{
4039 netdev_class_remove_file_ns(class_attr, NULL);
4040}
b8a9787e 4041
04600794
JB
4042extern struct kobj_ns_type_operations net_ns_type_operations;
4043
f629d208 4044const char *netdev_drivername(const struct net_device *dev);
6579e57b 4045
f629d208 4046void linkwatch_run_queue(void);
20380731 4047
da08143b
MK
4048static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4049 netdev_features_t f2)
4050{
c8cd0989
TH
4051 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4052 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4053 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4054 else
b6a0e72a 4055 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4056 }
da08143b 4057
c8cd0989 4058 return f1 & f2;
da08143b
MK
4059}
4060
c8f44aff
MM
4061static inline netdev_features_t netdev_get_wanted_features(
4062 struct net_device *dev)
5455c699
MM
4063{
4064 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4065}
c8f44aff
MM
4066netdev_features_t netdev_increment_features(netdev_features_t all,
4067 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4068
4069/* Allow TSO being used on stacked device :
4070 * Performing the GSO segmentation before last device
4071 * is a performance improvement.
4072 */
4073static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4074 netdev_features_t mask)
4075{
4076 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4077}
4078
6cb6a27c 4079int __netdev_update_features(struct net_device *dev);
5455c699 4080void netdev_update_features(struct net_device *dev);
afe12cc8 4081void netdev_change_features(struct net_device *dev);
7f353bf2 4082
fc4a7489
PM
4083void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4084 struct net_device *dev);
4085
e38f3025
TM
4086netdev_features_t passthru_features_check(struct sk_buff *skb,
4087 struct net_device *dev,
4088 netdev_features_t features);
c1e756bf 4089netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4090
4d29515f 4091static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4092{
7b748340 4093 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4094
4095 /* check flags correspondence */
4096 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4097 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4098 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4099 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4100 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4101 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4102 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4103 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4104 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4105 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4106 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4107 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4108 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4109 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4110 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4111 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4112 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0345e186 4113
d6b4991a 4114 return (features & feature) == feature;
576a30eb
HX
4115}
4116
4d29515f 4117static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4118{
278b2513 4119 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4120 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4121}
4122
8b86a61d 4123static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4124 netdev_features_t features)
7967168c 4125{
fc741216 4126 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4127 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4128 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4129}
4130
82cc1a7a
PWJ
4131static inline void netif_set_gso_max_size(struct net_device *dev,
4132 unsigned int size)
4133{
4134 dev->gso_max_size = size;
4135}
4136
7a7ffbab
WCC
4137static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4138 int pulled_hlen, u16 mac_offset,
4139 int mac_len)
4140{
4141 skb->protocol = protocol;
4142 skb->encapsulation = 1;
4143 skb_push(skb, pulled_hlen);
4144 skb_reset_transport_header(skb);
4145 skb->mac_header = mac_offset;
4146 skb->network_header = skb->mac_header + mac_len;
4147 skb->mac_len = mac_len;
4148}
4149
3c175784
SD
4150static inline bool netif_is_macsec(const struct net_device *dev)
4151{
4152 return dev->priv_flags & IFF_MACSEC;
4153}
4154
b618aaa9 4155static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4156{
4157 return dev->priv_flags & IFF_MACVLAN;
4158}
4159
b618aaa9 4160static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4161{
4162 return dev->priv_flags & IFF_MACVLAN_PORT;
4163}
4164
b618aaa9 4165static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4166{
4167 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4168}
4169
b618aaa9 4170static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4171{
4172 return dev->priv_flags & IFF_IPVLAN_MASTER;
4173}
4174
b618aaa9 4175static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4176{
4177 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4178}
4179
b618aaa9 4180static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4181{
4182 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4183}
4184
3bdc0eba
BG
4185static inline bool netif_supports_nofcs(struct net_device *dev)
4186{
4187 return dev->priv_flags & IFF_SUPP_NOFCS;
4188}
4189
007979ea 4190static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4191{
007979ea 4192 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4193}
4194
fee6d4c7
DA
4195static inline bool netif_is_l3_slave(const struct net_device *dev)
4196{
4197 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4198}
4199
0894ae3f
JP
4200static inline bool netif_is_bridge_master(const struct net_device *dev)
4201{
4202 return dev->priv_flags & IFF_EBRIDGE;
4203}
4204
28f9ee22
VY
4205static inline bool netif_is_bridge_port(const struct net_device *dev)
4206{
4207 return dev->priv_flags & IFF_BRIDGE_PORT;
4208}
4209
35d4e172
JP
4210static inline bool netif_is_ovs_master(const struct net_device *dev)
4211{
4212 return dev->priv_flags & IFF_OPENVSWITCH;
4213}
4214
5be66141
JP
4215static inline bool netif_is_ovs_port(const struct net_device *dev)
4216{
4217 return dev->priv_flags & IFF_OVS_DATAPATH;
4218}
4219
b618aaa9 4220static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4221{
4222 return dev->priv_flags & IFF_TEAM;
4223}
4224
b618aaa9 4225static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4226{
4227 return dev->priv_flags & IFF_TEAM_PORT;
4228}
4229
b618aaa9 4230static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4231{
4232 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4233}
4234
b618aaa9 4235static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4236{
4237 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4238}
4239
d4ab4286
KJ
4240static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4241{
4242 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4243}
4244
02875878
ED
4245/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4246static inline void netif_keep_dst(struct net_device *dev)
4247{
4248 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4249}
4250
18d3df3e
PA
4251/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4252static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4253{
4254 /* TODO: reserve and use an additional IFF bit, if we get more users */
4255 return dev->priv_flags & IFF_MACSEC;
4256}
4257
505d4f73 4258extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4259
571ba423
JP
4260/* Logging, debugging and troubleshooting/diagnostic helpers. */
4261
4262/* netdev_printk helpers, similar to dev_printk */
4263
4264static inline const char *netdev_name(const struct net_device *dev)
4265{
c6f854d5
VF
4266 if (!dev->name[0] || strchr(dev->name, '%'))
4267 return "(unnamed net_device)";
571ba423
JP
4268 return dev->name;
4269}
4270
8397ed36
DA
4271static inline bool netdev_unregistering(const struct net_device *dev)
4272{
4273 return dev->reg_state == NETREG_UNREGISTERING;
4274}
4275
ccc7f496
VF
4276static inline const char *netdev_reg_state(const struct net_device *dev)
4277{
4278 switch (dev->reg_state) {
4279 case NETREG_UNINITIALIZED: return " (uninitialized)";
4280 case NETREG_REGISTERED: return "";
4281 case NETREG_UNREGISTERING: return " (unregistering)";
4282 case NETREG_UNREGISTERED: return " (unregistered)";
4283 case NETREG_RELEASED: return " (released)";
4284 case NETREG_DUMMY: return " (dummy)";
4285 }
4286
4287 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4288 return " (unknown)";
4289}
4290
f629d208 4291__printf(3, 4)
6ea754eb
JP
4292void netdev_printk(const char *level, const struct net_device *dev,
4293 const char *format, ...);
f629d208 4294__printf(2, 3)
6ea754eb 4295void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4296__printf(2, 3)
6ea754eb 4297void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4298__printf(2, 3)
6ea754eb 4299void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4300__printf(2, 3)
6ea754eb 4301void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4302__printf(2, 3)
6ea754eb 4303void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4304__printf(2, 3)
6ea754eb 4305void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4306__printf(2, 3)
6ea754eb 4307void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4308
8909c9ad
VK
4309#define MODULE_ALIAS_NETDEV(device) \
4310 MODULE_ALIAS("netdev-" device)
4311
b558c96f 4312#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4313#define netdev_dbg(__dev, format, args...) \
4314do { \
ffa10cb4 4315 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4316} while (0)
b558c96f
JC
4317#elif defined(DEBUG)
4318#define netdev_dbg(__dev, format, args...) \
4319 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4320#else
4321#define netdev_dbg(__dev, format, args...) \
4322({ \
4323 if (0) \
4324 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4325})
4326#endif
4327
4328#if defined(VERBOSE_DEBUG)
4329#define netdev_vdbg netdev_dbg
4330#else
4331
4332#define netdev_vdbg(dev, format, args...) \
4333({ \
4334 if (0) \
4335 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4336 0; \
4337})
4338#endif
4339
4340/*
4341 * netdev_WARN() acts like dev_printk(), but with the key difference
4342 * of using a WARN/WARN_ON to get the message out, including the
4343 * file/line information and a backtrace.
4344 */
4345#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4346 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4347 netdev_reg_state(dev), ##args)
571ba423 4348
b3d95c5c
JP
4349/* netif printk helpers, similar to netdev_printk */
4350
4351#define netif_printk(priv, type, level, dev, fmt, args...) \
4352do { \
4353 if (netif_msg_##type(priv)) \
4354 netdev_printk(level, (dev), fmt, ##args); \
4355} while (0)
4356
f45f4321
JP
4357#define netif_level(level, priv, type, dev, fmt, args...) \
4358do { \
4359 if (netif_msg_##type(priv)) \
4360 netdev_##level(dev, fmt, ##args); \
4361} while (0)
4362
b3d95c5c 4363#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4364 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4365#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4366 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4367#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4368 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4369#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4370 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4371#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4372 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4373#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4374 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4375#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4376 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4377
0053ea9c 4378#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4379#define netif_dbg(priv, type, netdev, format, args...) \
4380do { \
4381 if (netif_msg_##type(priv)) \
b5fb0a03 4382 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4383} while (0)
0053ea9c
JP
4384#elif defined(DEBUG)
4385#define netif_dbg(priv, type, dev, format, args...) \
4386 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4387#else
4388#define netif_dbg(priv, type, dev, format, args...) \
4389({ \
4390 if (0) \
4391 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4392 0; \
4393})
4394#endif
4395
f617f276
EC
4396/* if @cond then downgrade to debug, else print at @level */
4397#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4398 do { \
4399 if (cond) \
4400 netif_dbg(priv, type, netdev, fmt, ##args); \
4401 else \
4402 netif_ ## level(priv, type, netdev, fmt, ##args); \
4403 } while (0)
4404
b3d95c5c 4405#if defined(VERBOSE_DEBUG)
bcfcc450 4406#define netif_vdbg netif_dbg
b3d95c5c
JP
4407#else
4408#define netif_vdbg(priv, type, dev, format, args...) \
4409({ \
4410 if (0) \
a4ed89cb 4411 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4412 0; \
4413})
4414#endif
571ba423 4415
900ff8c6
CW
4416/*
4417 * The list of packet types we will receive (as opposed to discard)
4418 * and the routines to invoke.
4419 *
4420 * Why 16. Because with 16 the only overlap we get on a hash of the
4421 * low nibble of the protocol value is RARP/SNAP/X.25.
4422 *
4423 * NOTE: That is no longer true with the addition of VLAN tags. Not
4424 * sure which should go first, but I bet it won't make much
4425 * difference if we are running VLANs. The good news is that
4426 * this protocol won't be in the list unless compiled in, so
4427 * the average user (w/out VLANs) will not be adversely affected.
4428 * --BLG
4429 *
4430 * 0800 IP
4431 * 8100 802.1Q VLAN
4432 * 0001 802.3
4433 * 0002 AX.25
4434 * 0004 802.2
4435 * 8035 RARP
4436 * 0005 SNAP
4437 * 0805 X.25
4438 * 0806 ARP
4439 * 8137 IPX
4440 * 0009 Localtalk
4441 * 86DD IPv6
4442 */
4443#define PTYPE_HASH_SIZE (16)
4444#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4445
385a154c 4446#endif /* _LINUX_NETDEVICE_H */