netlink: add NULL-friendly helper for setting extended ACK message
[linux-2.6-block.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
7a6b6f51 44#ifdef CONFIG_DCB
2f90b865
AD
45#include <net/dcbnl.h>
46#endif
5bc1421e 47#include <net/netprio_cgroup.h>
a050c33f 48
a59e2ecb 49#include <linux/netdev_features.h>
77162022 50#include <linux/neighbour.h>
607ca46e 51#include <uapi/linux/netdevice.h>
61bd3857 52#include <uapi/linux/if_bonding.h>
e4c6734e 53#include <uapi/linux/pkt_cls.h>
59cc1f61 54#include <linux/hashtable.h>
a59e2ecb 55
115c1d6e 56struct netpoll_info;
313162d0 57struct device;
c1f19b51 58struct phy_device;
c6e970a0
AL
59struct dsa_switch_tree;
60
704232c2
JB
61/* 802.11 specific */
62struct wireless_dev;
98a18b6f
AA
63/* 802.15.4 specific */
64struct wpan_dev;
03c57747 65struct mpls_dev;
7c46a640
AD
66/* UDP Tunnel offloads */
67struct udp_tunnel_info;
a7862b45 68struct bpf_prog;
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
816struct netdev_xdp {
817 enum xdp_netdev_command command;
818 union {
819 /* XDP_SETUP_PROG */
820 struct bpf_prog *prog;
821 /* XDP_QUERY_PROG */
822 bool prog_attached;
823 };
824};
16e5cc64 825
d77e38e6
SK
826#ifdef CONFIG_XFRM_OFFLOAD
827struct xfrmdev_ops {
828 int (*xdo_dev_state_add) (struct xfrm_state *x);
829 void (*xdo_dev_state_delete) (struct xfrm_state *x);
830 void (*xdo_dev_state_free) (struct xfrm_state *x);
831 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
832 struct xfrm_state *x);
833};
834#endif
835
d314774c
SH
836/*
837 * This structure defines the management hooks for network devices.
00829823
SH
838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
d314774c
SH
840 *
841 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
d314774c
SH
846 *
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
850 *
851 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 852 * This function is called when a network device transitions to the up
d314774c
SH
853 * state.
854 *
855 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 856 * This function is called when a network device transitions to the down
d314774c
SH
857 * state.
858 *
dc1f8bf6
SH
859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
00829823 861 * Called when a packet needs to be transmitted.
e79d8429
RR
862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 866 * Required; cannot be NULL.
00829823 867 *
1a2a1444
DM
868 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
869 * struct net_device *dev
870 * netdev_features_t features);
871 * Called by core transmit path to determine if device is capable of
872 * performing offload operations on a given packet. This is to give
873 * the device an opportunity to implement any restrictions that cannot
874 * be otherwise expressed by feature flags. The check is called with
875 * the set of features that the stack has calculated and it returns
876 * those the driver believes to be appropriate.
cdba756f 877 *
f663dd9a 878 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 879 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 880 * Called to decide which queue to use when device supports multiple
00829823
SH
881 * transmit queues.
882 *
d314774c
SH
883 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
884 * This function is called to allow device receiver to make
5e82b4b2 885 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
886 *
887 * void (*ndo_set_rx_mode)(struct net_device *dev);
888 * This function is called device changes address list filtering.
01789349 889 * If driver handles unicast address filtering, it should set
5e82b4b2 890 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
891 *
892 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
893 * This function is called when the Media Access Control address
37b607c5 894 * needs to be changed. If this interface is not defined, the
5e82b4b2 895 * MAC address can not be changed.
d314774c
SH
896 *
897 * int (*ndo_validate_addr)(struct net_device *dev);
898 * Test if Media Access Control address is valid for the device.
899 *
900 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
901 * Called when a user requests an ioctl which can't be handled by
902 * the generic interface code. If not defined ioctls return
d314774c
SH
903 * not supported error code.
904 *
905 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
906 * Used to set network devices bus interface parameters. This interface
5e82b4b2 907 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
908 * interface (PCI) for low level management.
909 *
910 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
911 * Called when a user wants to change the Maximum Transfer Unit
912 * of a device. If not defined, any request to change MTU will
913 * will return an error.
914 *
00829823 915 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 916 * Callback used when the transmitter has not made any progress
d314774c
SH
917 * for dev->watchdog ticks.
918 *
bc1f4470 919 * void (*ndo_get_stats64)(struct net_device *dev,
920 * struct rtnl_link_stats64 *storage);
d308e38f 921 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 922 * Called when a user wants to get the network device usage
be1f3c2c 923 * statistics. Drivers must do one of the following:
3cfde79c
BH
924 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
925 * rtnl_link_stats64 structure passed by the caller.
82695d9b 926 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
927 * (which should normally be dev->stats) and return a pointer to
928 * it. The structure may be changed asynchronously only if each
929 * field is written atomically.
930 * 3. Update dev->stats asynchronously and atomically, and define
931 * neither operation.
d314774c 932 *
3df5b3c6 933 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
934 * Return true if this device supports offload stats of this attr_id.
935 *
936 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
937 * void *attr_data)
938 * Get statistics for offload operations by attr_id. Write it into the
939 * attr_data pointer.
940 *
5d632cb7 941 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 942 * If device supports VLAN filtering this function is called when a
80d5c368 943 * VLAN id is registered.
d314774c 944 *
5d632cb7 945 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 946 * If device supports VLAN filtering this function is called when a
80d5c368 947 * VLAN id is unregistered.
d314774c
SH
948 *
949 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
950 *
951 * SR-IOV management functions.
952 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
953 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
954 * u8 qos, __be16 proto);
ed616689
SC
955 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
956 * int max_tx_rate);
5f8444a3 957 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 958 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
959 * int (*ndo_get_vf_config)(struct net_device *dev,
960 * int vf, struct ifla_vf_info *ivf);
1d8faf48 961 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
962 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
963 * struct nlattr *port[]);
01a3d796
VZ
964 *
965 * Enable or disable the VF ability to query its RSS Redirection Table and
966 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 967 * with PF and querying it may introduce a theoretical security risk.
01a3d796 968 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 969 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
6a4bc2b4
FF
970 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
971 * __be16 protocol, struct tc_to_netdev *tc);
972 * Called to setup any 'tc' scheduler, classifier or action on @dev.
973 * This is always called from the stack with the rtnl lock held and netif
974 * tx queues stopped. This allows the netdevice to perform queue
975 * management safely.
c445477d 976 *
e9bce845
YZ
977 * Fiber Channel over Ethernet (FCoE) offload functions.
978 * int (*ndo_fcoe_enable)(struct net_device *dev);
979 * Called when the FCoE protocol stack wants to start using LLD for FCoE
980 * so the underlying device can perform whatever needed configuration or
981 * initialization to support acceleration of FCoE traffic.
982 *
983 * int (*ndo_fcoe_disable)(struct net_device *dev);
984 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
985 * so the underlying device can perform whatever needed clean-ups to
986 * stop supporting acceleration of FCoE traffic.
987 *
988 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
989 * struct scatterlist *sgl, unsigned int sgc);
990 * Called when the FCoE Initiator wants to initialize an I/O that
991 * is a possible candidate for Direct Data Placement (DDP). The LLD can
992 * perform necessary setup and returns 1 to indicate the device is set up
993 * successfully to perform DDP on this I/O, otherwise this returns 0.
994 *
995 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
996 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
997 * indicated by the FC exchange id 'xid', so the underlying device can
998 * clean up and reuse resources for later DDP requests.
999 *
1000 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1001 * struct scatterlist *sgl, unsigned int sgc);
1002 * Called when the FCoE Target wants to initialize an I/O that
1003 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1004 * perform necessary setup and returns 1 to indicate the device is set up
1005 * successfully to perform DDP on this I/O, otherwise this returns 0.
1006 *
68bad94e
NP
1007 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1008 * struct netdev_fcoe_hbainfo *hbainfo);
1009 * Called when the FCoE Protocol stack wants information on the underlying
1010 * device. This information is utilized by the FCoE protocol stack to
1011 * register attributes with Fiber Channel management service as per the
1012 * FC-GS Fabric Device Management Information(FDMI) specification.
1013 *
e9bce845
YZ
1014 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1015 * Called when the underlying device wants to override default World Wide
1016 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1017 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1018 * protocol stack to use.
1019 *
c445477d
BH
1020 * RFS acceleration.
1021 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1022 * u16 rxq_index, u32 flow_id);
1023 * Set hardware filter for RFS. rxq_index is the target queue index;
1024 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1025 * Return the filter ID on success, or a negative error code.
fbaec0ea 1026 *
8b98a70c 1027 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1028 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1029 * Called to make another netdev an underling.
1030 *
1031 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1032 * Called to release previously enslaved netdev.
5455c699
MM
1033 *
1034 * Feature/offload setting functions.
1a2a1444
DM
1035 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1036 * netdev_features_t features);
1037 * Adjusts the requested feature flags according to device-specific
1038 * constraints, and returns the resulting flags. Must not modify
1039 * the device state.
1040 *
c8f44aff 1041 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1042 * Called to update device configuration to new features. Passed
1043 * feature set might be less than what was returned by ndo_fix_features()).
1044 * Must return >0 or -errno if it changed dev->features itself.
1045 *
edc7d573 1046 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1047 * struct net_device *dev,
f6f6424b 1048 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1049 * Adds an FDB entry to dev for addr.
1690be63
VY
1050 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1051 * struct net_device *dev,
f6f6424b 1052 * const unsigned char *addr, u16 vid)
77162022
JF
1053 * Deletes the FDB entry from dev coresponding to addr.
1054 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1055 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1056 * int *idx)
77162022
JF
1057 * Used to add FDB entries to dump requests. Implementers should add
1058 * entries to skb and update idx with the number of entries.
e5a55a89 1059 *
ad41faa8
ND
1060 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1061 * u16 flags)
e5a55a89 1062 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1063 * struct net_device *dev, u32 filter_mask,
1064 * int nlflags)
ad41faa8
ND
1065 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1066 * u16 flags);
4bf84c35
JP
1067 *
1068 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1069 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1070 * which do not represent real hardware may define this to allow their
1071 * userspace components to manage their virtual carrier state. Devices
1072 * that determine carrier state from physical hardware properties (eg
1073 * network cables) or protocol-dependent mechanisms (eg
1074 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1075 *
1076 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1077 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1078 * Called to get ID of physical port of this device. If driver does
1079 * not implement this, it is assumed that the hw is not able to have
1080 * multiple net devices on single physical port.
53cf5275 1081 *
7c46a640
AD
1082 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1083 * struct udp_tunnel_info *ti);
1084 * Called by UDP tunnel to notify a driver about the UDP port and socket
1085 * address family that a UDP tunnel is listnening to. It is called only
1086 * when a new port starts listening. The operation is protected by the
1087 * RTNL.
1088 *
1089 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1090 * struct udp_tunnel_info *ti);
1091 * Called by UDP tunnel to notify the driver about a UDP port and socket
1092 * address family that the UDP tunnel is not listening to anymore. The
1093 * operation is protected by the RTNL.
1094 *
a6cc0cfa
JF
1095 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1096 * struct net_device *dev)
1097 * Called by upper layer devices to accelerate switching or other
1098 * station functionality into hardware. 'pdev is the lowerdev
1099 * to use for the offload and 'dev' is the net device that will
1100 * back the offload. Returns a pointer to the private structure
1101 * the upper layer will maintain.
1102 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1103 * Called by upper layer device to delete the station created
1104 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1105 * the station and priv is the structure returned by the add
1106 * operation.
1107 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1108 * struct net_device *dev,
1109 * void *priv);
1110 * Callback to use for xmit over the accelerated station. This
1111 * is used in place of ndo_start_xmit on accelerated net
1112 * devices.
822b3b2e
JF
1113 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1114 * int queue_index, u32 maxrate);
1115 * Called when a user wants to set a max-rate limitation of specific
1116 * TX queue.
a54acb3a
ND
1117 * int (*ndo_get_iflink)(const struct net_device *dev);
1118 * Called to get the iflink value of this device.
d746d707 1119 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1120 * bool proto_down);
d746d707
AK
1121 * This function is used to pass protocol port error state information
1122 * to the switch driver. The switch driver can react to the proto_down
1123 * by doing a phys down on the associated switch port.
fc4099f1
PS
1124 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1125 * This function is used to get egress tunnel information for given skb.
1126 * This is useful for retrieving outer tunnel header parameters while
1127 * sampling packet.
871b642a
PA
1128 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1129 * This function is used to specify the headroom that the skb must
1130 * consider when allocation skb during packet reception. Setting
1131 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1132 * forward. Setting a negative value resets the rx headroom to the
871b642a 1133 * default value.
a7862b45
BB
1134 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1135 * This function is used to set or query state related to XDP on the
1136 * netdevice. See definition of enum xdp_netdev_command for details.
d746d707 1137 *
d314774c
SH
1138 */
1139struct net_device_ops {
1140 int (*ndo_init)(struct net_device *dev);
1141 void (*ndo_uninit)(struct net_device *dev);
1142 int (*ndo_open)(struct net_device *dev);
1143 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1144 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1145 struct net_device *dev);
1146 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1147 struct net_device *dev,
1148 netdev_features_t features);
00829823 1149 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1150 struct sk_buff *skb,
99932d4f
DB
1151 void *accel_priv,
1152 select_queue_fallback_t fallback);
d314774c
SH
1153 void (*ndo_change_rx_flags)(struct net_device *dev,
1154 int flags);
d314774c 1155 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1156 int (*ndo_set_mac_address)(struct net_device *dev,
1157 void *addr);
d314774c 1158 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1159 int (*ndo_do_ioctl)(struct net_device *dev,
1160 struct ifreq *ifr, int cmd);
d314774c
SH
1161 int (*ndo_set_config)(struct net_device *dev,
1162 struct ifmap *map);
00829823
SH
1163 int (*ndo_change_mtu)(struct net_device *dev,
1164 int new_mtu);
1165 int (*ndo_neigh_setup)(struct net_device *dev,
1166 struct neigh_parms *);
d314774c
SH
1167 void (*ndo_tx_timeout) (struct net_device *dev);
1168
bc1f4470 1169 void (*ndo_get_stats64)(struct net_device *dev,
1170 struct rtnl_link_stats64 *storage);
3df5b3c6 1171 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1172 int (*ndo_get_offload_stats)(int attr_id,
1173 const struct net_device *dev,
1174 void *attr_data);
d314774c
SH
1175 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1176
8e586137 1177 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1178 __be16 proto, u16 vid);
8e586137 1179 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1180 __be16 proto, u16 vid);
d314774c 1181#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1182 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1183 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1184 struct netpoll_info *info);
0e34e931 1185 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1186#endif
95c26df8
WM
1187 int (*ndo_set_vf_mac)(struct net_device *dev,
1188 int queue, u8 *mac);
1189 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1190 int queue, u16 vlan,
1191 u8 qos, __be16 proto);
ed616689
SC
1192 int (*ndo_set_vf_rate)(struct net_device *dev,
1193 int vf, int min_tx_rate,
1194 int max_tx_rate);
5f8444a3
GR
1195 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1196 int vf, bool setting);
dd461d6a
HS
1197 int (*ndo_set_vf_trust)(struct net_device *dev,
1198 int vf, bool setting);
95c26df8
WM
1199 int (*ndo_get_vf_config)(struct net_device *dev,
1200 int vf,
1201 struct ifla_vf_info *ivf);
1d8faf48
RE
1202 int (*ndo_set_vf_link_state)(struct net_device *dev,
1203 int vf, int link_state);
3b766cd8
EBE
1204 int (*ndo_get_vf_stats)(struct net_device *dev,
1205 int vf,
1206 struct ifla_vf_stats
1207 *vf_stats);
57b61080
SF
1208 int (*ndo_set_vf_port)(struct net_device *dev,
1209 int vf,
1210 struct nlattr *port[]);
1211 int (*ndo_get_vf_port)(struct net_device *dev,
1212 int vf, struct sk_buff *skb);
cc8e27cc
EC
1213 int (*ndo_set_vf_guid)(struct net_device *dev,
1214 int vf, u64 guid,
1215 int guid_type);
01a3d796
VZ
1216 int (*ndo_set_vf_rss_query_en)(
1217 struct net_device *dev,
1218 int vf, bool setting);
16e5cc64
JF
1219 int (*ndo_setup_tc)(struct net_device *dev,
1220 u32 handle,
1221 __be16 protocol,
1222 struct tc_to_netdev *tc);
d11ead75 1223#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1224 int (*ndo_fcoe_enable)(struct net_device *dev);
1225 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1226 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1227 u16 xid,
1228 struct scatterlist *sgl,
1229 unsigned int sgc);
1230 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1231 u16 xid);
6247e086
YZ
1232 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1233 u16 xid,
1234 struct scatterlist *sgl,
1235 unsigned int sgc);
68bad94e
NP
1236 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1237 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1238#endif
1239
d11ead75 1240#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1241#define NETDEV_FCOE_WWNN 0
1242#define NETDEV_FCOE_WWPN 1
1243 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1244 u64 *wwn, int type);
4d288d57 1245#endif
3c9c36bc 1246
c445477d
BH
1247#ifdef CONFIG_RFS_ACCEL
1248 int (*ndo_rx_flow_steer)(struct net_device *dev,
1249 const struct sk_buff *skb,
1250 u16 rxq_index,
1251 u32 flow_id);
1252#endif
fbaec0ea
JP
1253 int (*ndo_add_slave)(struct net_device *dev,
1254 struct net_device *slave_dev);
1255 int (*ndo_del_slave)(struct net_device *dev,
1256 struct net_device *slave_dev);
c8f44aff
MM
1257 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1258 netdev_features_t features);
5455c699 1259 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1260 netdev_features_t features);
503eebc2
JP
1261 int (*ndo_neigh_construct)(struct net_device *dev,
1262 struct neighbour *n);
1263 void (*ndo_neigh_destroy)(struct net_device *dev,
1264 struct neighbour *n);
77162022
JF
1265
1266 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1267 struct nlattr *tb[],
77162022 1268 struct net_device *dev,
6b6e2725 1269 const unsigned char *addr,
f6f6424b 1270 u16 vid,
77162022
JF
1271 u16 flags);
1272 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1273 struct nlattr *tb[],
77162022 1274 struct net_device *dev,
f6f6424b
JP
1275 const unsigned char *addr,
1276 u16 vid);
77162022
JF
1277 int (*ndo_fdb_dump)(struct sk_buff *skb,
1278 struct netlink_callback *cb,
1279 struct net_device *dev,
5d5eacb3 1280 struct net_device *filter_dev,
d297653d 1281 int *idx);
e5a55a89
JF
1282
1283 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1284 struct nlmsghdr *nlh,
1285 u16 flags);
e5a55a89
JF
1286 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1287 u32 pid, u32 seq,
6cbdceeb 1288 struct net_device *dev,
46c264da
ND
1289 u32 filter_mask,
1290 int nlflags);
407af329 1291 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1292 struct nlmsghdr *nlh,
1293 u16 flags);
4bf84c35
JP
1294 int (*ndo_change_carrier)(struct net_device *dev,
1295 bool new_carrier);
66b52b0d 1296 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1297 struct netdev_phys_item_id *ppid);
db24a904
DA
1298 int (*ndo_get_phys_port_name)(struct net_device *dev,
1299 char *name, size_t len);
7c46a640
AD
1300 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1301 struct udp_tunnel_info *ti);
1302 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1303 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1304 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1305 struct net_device *dev);
1306 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1307 void *priv);
1308
1309 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1310 struct net_device *dev,
1311 void *priv);
25175ba5 1312 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1313 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1314 int queue_index,
1315 u32 maxrate);
a54acb3a 1316 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1317 int (*ndo_change_proto_down)(struct net_device *dev,
1318 bool proto_down);
fc4099f1
PS
1319 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1320 struct sk_buff *skb);
871b642a
PA
1321 void (*ndo_set_rx_headroom)(struct net_device *dev,
1322 int needed_headroom);
a7862b45
BB
1323 int (*ndo_xdp)(struct net_device *dev,
1324 struct netdev_xdp *xdp);
d314774c
SH
1325};
1326
7aa98047
LR
1327/**
1328 * enum net_device_priv_flags - &struct net_device priv_flags
1329 *
1330 * These are the &struct net_device, they are only set internally
1331 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1332 * userspace; this means that the order of these flags can change
7aa98047
LR
1333 * during any kernel release.
1334 *
1335 * You should have a pretty good reason to be extending these flags.
1336 *
1337 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1338 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1339 * @IFF_BONDING: bonding master or slave
7aa98047 1340 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1341 * @IFF_WAN_HDLC: WAN HDLC device
1342 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1343 * release skb->dst
1344 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1345 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1346 * @IFF_MACVLAN_PORT: device used as macvlan port
1347 * @IFF_BRIDGE_PORT: device used as bridge port
1348 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1349 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1350 * @IFF_UNICAST_FLT: Supports unicast filtering
1351 * @IFF_TEAM_PORT: device used as team port
1352 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1353 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1354 * change when it's running
1355 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1356 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1357 * underlying stacked devices
1358 * @IFF_IPVLAN_MASTER: IPvlan master device
1359 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1360 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1361 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1362 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1363 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1364 * @IFF_TEAM: device is a team device
d4ab4286 1365 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1366 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1367 * entity (i.e. the master device for bridged veth)
3c175784 1368 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1369 */
1370enum netdev_priv_flags {
1371 IFF_802_1Q_VLAN = 1<<0,
1372 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1373 IFF_BONDING = 1<<2,
1374 IFF_ISATAP = 1<<3,
1375 IFF_WAN_HDLC = 1<<4,
1376 IFF_XMIT_DST_RELEASE = 1<<5,
1377 IFF_DONT_BRIDGE = 1<<6,
1378 IFF_DISABLE_NETPOLL = 1<<7,
1379 IFF_MACVLAN_PORT = 1<<8,
1380 IFF_BRIDGE_PORT = 1<<9,
1381 IFF_OVS_DATAPATH = 1<<10,
1382 IFF_TX_SKB_SHARING = 1<<11,
1383 IFF_UNICAST_FLT = 1<<12,
1384 IFF_TEAM_PORT = 1<<13,
1385 IFF_SUPP_NOFCS = 1<<14,
1386 IFF_LIVE_ADDR_CHANGE = 1<<15,
1387 IFF_MACVLAN = 1<<16,
1388 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1389 IFF_IPVLAN_MASTER = 1<<18,
1390 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1391 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1392 IFF_NO_QUEUE = 1<<21,
1393 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1394 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1395 IFF_TEAM = 1<<24,
d4ab4286 1396 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1397 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1398 IFF_MACSEC = 1<<27,
7aa98047
LR
1399};
1400
1401#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1402#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1403#define IFF_BONDING IFF_BONDING
7aa98047 1404#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1405#define IFF_WAN_HDLC IFF_WAN_HDLC
1406#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1407#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1408#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1409#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1410#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1411#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1412#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1413#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1414#define IFF_TEAM_PORT IFF_TEAM_PORT
1415#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1416#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1417#define IFF_MACVLAN IFF_MACVLAN
02875878 1418#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1419#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1420#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1421#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1422#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1423#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1424#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1425#define IFF_TEAM IFF_TEAM
d4ab4286 1426#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1427#define IFF_MACSEC IFF_MACSEC
7aa98047 1428
536721b1
KK
1429/**
1430 * struct net_device - The DEVICE structure.
1431 * Actually, this whole structure is a big mistake. It mixes I/O
1432 * data with strictly "high-level" data, and it has to know about
1433 * almost every data structure used in the INET module.
1434 *
1435 * @name: This is the first field of the "visible" part of this structure
1436 * (i.e. as seen by users in the "Space.c" file). It is the name
1437 * of the interface.
1438 *
1439 * @name_hlist: Device name hash chain, please keep it close to name[]
1440 * @ifalias: SNMP alias
1441 * @mem_end: Shared memory end
1442 * @mem_start: Shared memory start
1443 * @base_addr: Device I/O address
1444 * @irq: Device IRQ number
1445 *
14ffbbb8
TG
1446 * @carrier_changes: Stats to monitor carrier on<->off transitions
1447 *
536721b1
KK
1448 * @state: Generic network queuing layer state, see netdev_state_t
1449 * @dev_list: The global list of network devices
5e82b4b2
BH
1450 * @napi_list: List entry used for polling NAPI devices
1451 * @unreg_list: List entry when we are unregistering the
1452 * device; see the function unregister_netdev
1453 * @close_list: List entry used when we are closing the device
62d885fe
BP
1454 * @ptype_all: Device-specific packet handlers for all protocols
1455 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1456 *
1457 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1458 * @features: Currently active device features
1459 * @hw_features: User-changeable features
1460 *
1461 * @wanted_features: User-requested features
1462 * @vlan_features: Mask of features inheritable by VLAN devices
1463 *
1464 * @hw_enc_features: Mask of features inherited by encapsulating devices
1465 * This field indicates what encapsulation
1466 * offloads the hardware is capable of doing,
1467 * and drivers will need to set them appropriately.
1468 *
1469 * @mpls_features: Mask of features inheritable by MPLS
1470 *
1471 * @ifindex: interface index
5e82b4b2 1472 * @group: The group the device belongs to
536721b1
KK
1473 *
1474 * @stats: Statistics struct, which was left as a legacy, use
1475 * rtnl_link_stats64 instead
1476 *
1477 * @rx_dropped: Dropped packets by core network,
1478 * do not use this in drivers
1479 * @tx_dropped: Dropped packets by core network,
1480 * do not use this in drivers
6e7333d3
JW
1481 * @rx_nohandler: nohandler dropped packets by core network on
1482 * inactive devices, do not use this in drivers
536721b1 1483 *
536721b1
KK
1484 * @wireless_handlers: List of functions to handle Wireless Extensions,
1485 * instead of ioctl,
1486 * see <net/iw_handler.h> for details.
1487 * @wireless_data: Instance data managed by the core of wireless extensions
1488 *
1489 * @netdev_ops: Includes several pointers to callbacks,
1490 * if one wants to override the ndo_*() functions
1491 * @ethtool_ops: Management operations
f997c55c
AA
1492 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1493 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1494 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1495 * of Layer 2 headers.
1496 *
1497 * @flags: Interface flags (a la BSD)
1498 * @priv_flags: Like 'flags' but invisible to userspace,
1499 * see if.h for the definitions
1500 * @gflags: Global flags ( kept as legacy )
1501 * @padded: How much padding added by alloc_netdev()
1502 * @operstate: RFC2863 operstate
1503 * @link_mode: Mapping policy to operstate
1504 * @if_port: Selectable AUI, TP, ...
1505 * @dma: DMA channel
1506 * @mtu: Interface MTU value
61e84623
JW
1507 * @min_mtu: Interface Minimum MTU value
1508 * @max_mtu: Interface Maximum MTU value
536721b1 1509 * @type: Interface hardware type
2793a23a 1510 * @hard_header_len: Maximum hardware header length.
217e6fa2 1511 * @min_header_len: Minimum hardware header length
536721b1
KK
1512 *
1513 * @needed_headroom: Extra headroom the hardware may need, but not in all
1514 * cases can this be guaranteed
1515 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1516 * cases can this be guaranteed. Some cases also use
1517 * LL_MAX_HEADER instead to allocate the skb
1518 *
1519 * interface address info:
1520 *
1521 * @perm_addr: Permanent hw address
1522 * @addr_assign_type: Hw address assignment type
1523 * @addr_len: Hardware address length
8626a0c8 1524 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1525 * @dev_id: Used to differentiate devices that share
1526 * the same link layer address
1527 * @dev_port: Used to differentiate devices that share
1528 * the same function
1529 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1530 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1531 * has been enabled due to the need to listen to
1532 * additional unicast addresses in a device that
1533 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1534 * @uc: unicast mac addresses
1535 * @mc: multicast mac addresses
1536 * @dev_addrs: list of device hw addresses
1537 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1538 * @promiscuity: Number of times the NIC is told to work in
1539 * promiscuous mode; if it becomes 0 the NIC will
1540 * exit promiscuous mode
536721b1
KK
1541 * @allmulti: Counter, enables or disables allmulticast mode
1542 *
1543 * @vlan_info: VLAN info
1544 * @dsa_ptr: dsa specific data
1545 * @tipc_ptr: TIPC specific data
1546 * @atalk_ptr: AppleTalk link
1547 * @ip_ptr: IPv4 specific data
1548 * @dn_ptr: DECnet specific data
1549 * @ip6_ptr: IPv6 specific data
1550 * @ax25_ptr: AX.25 specific data
1551 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1552 *
536721b1
KK
1553 * @dev_addr: Hw address (before bcast,
1554 * because most packets are unicast)
1555 *
1556 * @_rx: Array of RX queues
1557 * @num_rx_queues: Number of RX queues
1558 * allocated at register_netdev() time
1559 * @real_num_rx_queues: Number of RX queues currently active in device
1560 *
1561 * @rx_handler: handler for received packets
1562 * @rx_handler_data: XXX: need comments on this one
1563 * @ingress_queue: XXX: need comments on this one
1564 * @broadcast: hw bcast address
1565 *
14ffbbb8
TG
1566 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1567 * indexed by RX queue number. Assigned by driver.
1568 * This must only be set if the ndo_rx_flow_steer
1569 * operation is defined
1570 * @index_hlist: Device index hash chain
1571 *
536721b1
KK
1572 * @_tx: Array of TX queues
1573 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1574 * @real_num_tx_queues: Number of TX queues currently active in device
1575 * @qdisc: Root qdisc from userspace point of view
1576 * @tx_queue_len: Max frames per queue allowed
1577 * @tx_global_lock: XXX: need comments on this one
1578 *
1579 * @xps_maps: XXX: need comments on this one
1580 *
536721b1 1581 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1582 * the watchdog (see dev_watchdog())
536721b1
KK
1583 * @watchdog_timer: List of timers
1584 *
1585 * @pcpu_refcnt: Number of references to this device
1586 * @todo_list: Delayed register/unregister
536721b1
KK
1587 * @link_watch_list: XXX: need comments on this one
1588 *
1589 * @reg_state: Register/unregister state machine
1590 * @dismantle: Device is going to be freed
1591 * @rtnl_link_state: This enum represents the phases of creating
1592 * a new link
1593 *
1594 * @destructor: Called from unregister,
1595 * can be used to call free_netdev
1596 * @npinfo: XXX: need comments on this one
1597 * @nd_net: Network namespace this network device is inside
1598 *
1599 * @ml_priv: Mid-layer private
1600 * @lstats: Loopback statistics
1601 * @tstats: Tunnel statistics
1602 * @dstats: Dummy statistics
1603 * @vstats: Virtual ethernet statistics
1604 *
1605 * @garp_port: GARP
1606 * @mrp_port: MRP
1607 *
1608 * @dev: Class/net/name entry
1609 * @sysfs_groups: Space for optional device, statistics and wireless
1610 * sysfs groups
1611 *
1612 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1613 * @rtnl_link_ops: Rtnl_link_ops
1614 *
1615 * @gso_max_size: Maximum size of generic segmentation offload
1616 * @gso_max_segs: Maximum number of segments that can be passed to the
1617 * NIC for GSO
1618 *
1619 * @dcbnl_ops: Data Center Bridging netlink ops
1620 * @num_tc: Number of traffic classes in the net device
1621 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1622 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1623 *
1624 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1625 *
1626 * @priomap: XXX: need comments on this one
1627 * @phydev: Physical device may attach itself
1628 * for hardware timestamping
1629 *
123b3652
ED
1630 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1631 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1632 *
d746d707
AK
1633 * @proto_down: protocol port state information can be sent to the
1634 * switch driver and used to set the phys state of the
1635 * switch port.
1636 *
1da177e4
LT
1637 * FIXME: cleanup struct net_device such that network protocol info
1638 * moves out.
1639 */
1640
d94d9fee 1641struct net_device {
1da177e4 1642 char name[IFNAMSIZ];
9356b8fc 1643 struct hlist_node name_hlist;
0b815a1a 1644 char *ifalias;
1da177e4
LT
1645 /*
1646 * I/O specific fields
1647 * FIXME: Merge these and struct ifmap into one
1648 */
536721b1
KK
1649 unsigned long mem_end;
1650 unsigned long mem_start;
1651 unsigned long base_addr;
1652 int irq;
1da177e4 1653
14ffbbb8
TG
1654 atomic_t carrier_changes;
1655
1da177e4 1656 /*
536721b1
KK
1657 * Some hardware also needs these fields (state,dev_list,
1658 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1659 * part of the usual set specified in Space.c.
1660 */
1661
1da177e4
LT
1662 unsigned long state;
1663
7562f876 1664 struct list_head dev_list;
bea3348e 1665 struct list_head napi_list;
44a0873d 1666 struct list_head unreg_list;
5cde2829 1667 struct list_head close_list;
7866a621
SN
1668 struct list_head ptype_all;
1669 struct list_head ptype_specific;
2f268f12 1670
2f268f12
VF
1671 struct {
1672 struct list_head upper;
1673 struct list_head lower;
1674 } adj_list;
1675
c8f44aff 1676 netdev_features_t features;
c8f44aff 1677 netdev_features_t hw_features;
c8f44aff 1678 netdev_features_t wanted_features;
c8f44aff 1679 netdev_features_t vlan_features;
6a674e9c 1680 netdev_features_t hw_enc_features;
0d89d203 1681 netdev_features_t mpls_features;
802ab55a 1682 netdev_features_t gso_partial_features;
04ed3e74 1683
1da177e4 1684 int ifindex;
7a66bbc9 1685 int group;
1da177e4 1686
c45d286e 1687 struct net_device_stats stats;
015f0688 1688
015f0688
ED
1689 atomic_long_t rx_dropped;
1690 atomic_long_t tx_dropped;
6e7333d3 1691 atomic_long_t rx_nohandler;
1da177e4 1692
b86e0280 1693#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1694 const struct iw_handler_def *wireless_handlers;
1695 struct iw_public_data *wireless_data;
b86e0280 1696#endif
d314774c 1697 const struct net_device_ops *netdev_ops;
76fd8593 1698 const struct ethtool_ops *ethtool_ops;
4170604f 1699#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1700 const struct switchdev_ops *switchdev_ops;
4170604f 1701#endif
1b69c6d0
DA
1702#ifdef CONFIG_NET_L3_MASTER_DEV
1703 const struct l3mdev_ops *l3mdev_ops;
1704#endif
f997c55c
AA
1705#if IS_ENABLED(CONFIG_IPV6)
1706 const struct ndisc_ops *ndisc_ops;
1707#endif
1da177e4 1708
d77e38e6
SK
1709#ifdef CONFIG_XFRM
1710 const struct xfrmdev_ops *xfrmdev_ops;
1711#endif
1712
3b04ddde
SH
1713 const struct header_ops *header_ops;
1714
536721b1
KK
1715 unsigned int flags;
1716 unsigned int priv_flags;
1717
1da177e4 1718 unsigned short gflags;
536721b1 1719 unsigned short padded;
1da177e4 1720
536721b1
KK
1721 unsigned char operstate;
1722 unsigned char link_mode;
b00055aa 1723
536721b1
KK
1724 unsigned char if_port;
1725 unsigned char dma;
bdc220da 1726
536721b1 1727 unsigned int mtu;
61e84623
JW
1728 unsigned int min_mtu;
1729 unsigned int max_mtu;
536721b1
KK
1730 unsigned short type;
1731 unsigned short hard_header_len;
d92be7a4 1732 unsigned char min_header_len;
1da177e4 1733
f5184d26
JB
1734 unsigned short needed_headroom;
1735 unsigned short needed_tailroom;
1736
1da177e4 1737 /* Interface address info. */
536721b1
KK
1738 unsigned char perm_addr[MAX_ADDR_LEN];
1739 unsigned char addr_assign_type;
1740 unsigned char addr_len;
a0a9663d 1741 unsigned short neigh_priv_len;
536721b1
KK
1742 unsigned short dev_id;
1743 unsigned short dev_port;
ccffad25 1744 spinlock_t addr_list_lock;
14ffbbb8
TG
1745 unsigned char name_assign_type;
1746 bool uc_promisc;
536721b1
KK
1747 struct netdev_hw_addr_list uc;
1748 struct netdev_hw_addr_list mc;
1749 struct netdev_hw_addr_list dev_addrs;
1750
4c3d5e7b
ED
1751#ifdef CONFIG_SYSFS
1752 struct kset *queues_kset;
1753#endif
9d45abe1
WC
1754 unsigned int promiscuity;
1755 unsigned int allmulti;
1da177e4 1756
1da177e4 1757
5e82b4b2 1758 /* Protocol-specific pointers */
65ac6a5f 1759
d11ead75 1760#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1761 struct vlan_info __rcu *vlan_info;
65ac6a5f 1762#endif
34a430d7 1763#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1764 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1765#endif
1766#if IS_ENABLED(CONFIG_TIPC)
536721b1 1767 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1768#endif
536721b1
KK
1769 void *atalk_ptr;
1770 struct in_device __rcu *ip_ptr;
1771 struct dn_dev __rcu *dn_ptr;
1772 struct inet6_dev __rcu *ip6_ptr;
1773 void *ax25_ptr;
1774 struct wireless_dev *ieee80211_ptr;
98a18b6f 1775 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1776#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1777 struct mpls_dev __rcu *mpls_ptr;
1778#endif
1da177e4 1779
9356b8fc 1780/*
cd13539b 1781 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1782 */
9356b8fc 1783 /* Interface address info used in eth_type_trans() */
536721b1 1784 unsigned char *dev_addr;
f001fde5 1785
a953be53 1786#ifdef CONFIG_SYSFS
0a9627f2
TH
1787 struct netdev_rx_queue *_rx;
1788
0a9627f2 1789 unsigned int num_rx_queues;
62fe0b40 1790 unsigned int real_num_rx_queues;
df334545 1791#endif
0a9627f2 1792
7acedaf5 1793 struct bpf_prog __rcu *xdp_prog;
3b47d303 1794 unsigned long gro_flush_timeout;
61391cde 1795 rx_handler_func_t __rcu *rx_handler;
1796 void __rcu *rx_handler_data;
e8a0464c 1797
4cda01e8 1798#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1799 struct tcf_proto __rcu *ingress_cl_list;
1800#endif
24824a09 1801 struct netdev_queue __rcu *ingress_queue;
e687ad60 1802#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1803 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1804#endif
d2788d34 1805
536721b1 1806 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1807#ifdef CONFIG_RFS_ACCEL
1808 struct cpu_rmap *rx_cpu_rmap;
1809#endif
1810 struct hlist_node index_hlist;
cd13539b
ED
1811
1812/*
1813 * Cache lines mostly used on transmit path
1814 */
e8a0464c
DM
1815 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1816 unsigned int num_tx_queues;
fd2ea0a7 1817 unsigned int real_num_tx_queues;
af356afa 1818 struct Qdisc *qdisc;
59cc1f61
JK
1819#ifdef CONFIG_NET_SCHED
1820 DECLARE_HASHTABLE (qdisc_hash, 4);
1821#endif
536721b1 1822 unsigned long tx_queue_len;
c3f26a26 1823 spinlock_t tx_global_lock;
14ffbbb8 1824 int watchdog_timeo;
cd13539b 1825
bf264145 1826#ifdef CONFIG_XPS
a4177869 1827 struct xps_dev_maps __rcu *xps_maps;
bf264145 1828#endif
1f211a1b
DB
1829#ifdef CONFIG_NET_CLS_ACT
1830 struct tcf_proto __rcu *egress_cl_list;
1831#endif
0c4f691f 1832
9356b8fc 1833 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1834 struct timer_list watchdog_timer;
1835
29b4433d 1836 int __percpu *pcpu_refcnt;
1da177e4 1837 struct list_head todo_list;
1da177e4 1838
e014debe 1839 struct list_head link_watch_list;
572a103d 1840
1da177e4 1841 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1842 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1843 NETREG_UNREGISTERING, /* called unregister_netdevice */
1844 NETREG_UNREGISTERED, /* completed unregister todo */
1845 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1846 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1847 } reg_state:8;
1848
536721b1 1849 bool dismantle;
a2835763
PM
1850
1851 enum {
1852 RTNL_LINK_INITIALIZED,
1853 RTNL_LINK_INITIALIZING,
1854 } rtnl_link_state:16;
1da177e4 1855
d314774c 1856 void (*destructor)(struct net_device *dev);
1da177e4 1857
1da177e4 1858#ifdef CONFIG_NETPOLL
5fbee843 1859 struct netpoll_info __rcu *npinfo;
1da177e4 1860#endif
eae792b7 1861
0c5c9fb5 1862 possible_net_t nd_net;
4a1c5371 1863
4951704b 1864 /* mid-layer private */
a7855c78 1865 union {
536721b1
KK
1866 void *ml_priv;
1867 struct pcpu_lstats __percpu *lstats;
8f84985f 1868 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1869 struct pcpu_dstats __percpu *dstats;
1870 struct pcpu_vstats __percpu *vstats;
a7855c78 1871 };
536721b1 1872
fb585b44 1873#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1874 struct garp_port __rcu *garp_port;
fb585b44
TK
1875#endif
1876#if IS_ENABLED(CONFIG_MRP)
febf018d 1877 struct mrp_port __rcu *mrp_port;
fb585b44 1878#endif
1da177e4 1879
5e82b4b2 1880 struct device dev;
0c509a6c 1881 const struct attribute_group *sysfs_groups[4];
a953be53 1882 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1883
38f7b870 1884 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1885
82cc1a7a
PWJ
1886 /* for setting kernel sock attribute on TCP connection setup */
1887#define GSO_MAX_SIZE 65536
1888 unsigned int gso_max_size;
30b678d8
BH
1889#define GSO_MAX_SEGS 65535
1890 u16 gso_max_segs;
743b03a8 1891
7a6b6f51 1892#ifdef CONFIG_DCB
32953543 1893 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1894#endif
5e82b4b2
BH
1895 u8 num_tc;
1896 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1897 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1898
d11ead75 1899#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1900 unsigned int fcoe_ddp_xid;
5bc1421e 1901#endif
86f8515f 1902#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1903 struct netprio_map __rcu *priomap;
4d288d57 1904#endif
5e82b4b2
BH
1905 struct phy_device *phydev;
1906 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1907 struct lock_class_key *qdisc_running_key;
5e82b4b2 1908 bool proto_down;
1da177e4 1909};
43cb76d9 1910#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1911
b5cdae32
DM
1912static inline bool netif_elide_gro(const struct net_device *dev)
1913{
1914 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1915 return true;
1916 return false;
1917}
1918
1da177e4 1919#define NETDEV_ALIGN 32
1da177e4 1920
4f57c087
JF
1921static inline
1922int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1923{
1924 return dev->prio_tc_map[prio & TC_BITMASK];
1925}
1926
1927static inline
1928int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1929{
1930 if (tc >= dev->num_tc)
1931 return -EINVAL;
1932
1933 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1934 return 0;
1935}
1936
8d059b0f 1937int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1938void netdev_reset_tc(struct net_device *dev);
1939int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1940int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1941
1942static inline
1943int netdev_get_num_tc(struct net_device *dev)
1944{
1945 return dev->num_tc;
1946}
1947
e8a0464c
DM
1948static inline
1949struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1950 unsigned int index)
1951{
1952 return &dev->_tx[index];
1953}
1954
10c51b56
DB
1955static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1956 const struct sk_buff *skb)
1957{
1958 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1959}
1960
e8a0464c
DM
1961static inline void netdev_for_each_tx_queue(struct net_device *dev,
1962 void (*f)(struct net_device *,
1963 struct netdev_queue *,
1964 void *),
1965 void *arg)
1966{
1967 unsigned int i;
1968
1969 for (i = 0; i < dev->num_tx_queues; i++)
1970 f(dev, &dev->_tx[i], arg);
1971}
1972
d3fff6c4
ED
1973#define netdev_lockdep_set_classes(dev) \
1974{ \
1975 static struct lock_class_key qdisc_tx_busylock_key; \
1976 static struct lock_class_key qdisc_running_key; \
1977 static struct lock_class_key qdisc_xmit_lock_key; \
1978 static struct lock_class_key dev_addr_list_lock_key; \
1979 unsigned int i; \
1980 \
1981 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1982 (dev)->qdisc_running_key = &qdisc_running_key; \
1983 lockdep_set_class(&(dev)->addr_list_lock, \
1984 &dev_addr_list_lock_key); \
1985 for (i = 0; i < (dev)->num_tx_queues; i++) \
1986 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1987 &qdisc_xmit_lock_key); \
1988}
1989
f629d208 1990struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1991 struct sk_buff *skb,
1992 void *accel_priv);
8c4c49df 1993
871b642a
PA
1994/* returns the headroom that the master device needs to take in account
1995 * when forwarding to this dev
1996 */
1997static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1998{
1999 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2000}
2001
2002static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2003{
2004 if (dev->netdev_ops->ndo_set_rx_headroom)
2005 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2006}
2007
2008/* set the device rx headroom to the dev's default */
2009static inline void netdev_reset_rx_headroom(struct net_device *dev)
2010{
2011 netdev_set_rx_headroom(dev, -1);
2012}
2013
c346dca1
YH
2014/*
2015 * Net namespace inlines
2016 */
2017static inline
2018struct net *dev_net(const struct net_device *dev)
2019{
c2d9ba9b 2020 return read_pnet(&dev->nd_net);
c346dca1
YH
2021}
2022
2023static inline
f5aa23fd 2024void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2025{
0c5c9fb5 2026 write_pnet(&dev->nd_net, net);
c346dca1
YH
2027}
2028
bea3348e
SH
2029/**
2030 * netdev_priv - access network device private data
2031 * @dev: network device
2032 *
2033 * Get network device private data
2034 */
6472ce60 2035static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2036{
1ce8e7b5 2037 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2038}
2039
1da177e4
LT
2040/* Set the sysfs physical device reference for the network logical device
2041 * if set prior to registration will cause a symlink during initialization.
2042 */
43cb76d9 2043#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2044
384912ed 2045/* Set the sysfs device type for the network logical device to allow
3f79410c 2046 * fine-grained identification of different network device types. For
5e82b4b2 2047 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2048 */
2049#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2050
82dc3c63
ED
2051/* Default NAPI poll() weight
2052 * Device drivers are strongly advised to not use bigger value
2053 */
2054#define NAPI_POLL_WEIGHT 64
2055
3b582cc1 2056/**
5e82b4b2 2057 * netif_napi_add - initialize a NAPI context
3b582cc1 2058 * @dev: network device
5e82b4b2 2059 * @napi: NAPI context
3b582cc1
SH
2060 * @poll: polling function
2061 * @weight: default weight
2062 *
5e82b4b2
BH
2063 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2064 * *any* of the other NAPI-related functions.
3b582cc1 2065 */
d565b0a1
HX
2066void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2067 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2068
d64b5e85 2069/**
5e82b4b2 2070 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2071 * @dev: network device
5e82b4b2 2072 * @napi: NAPI context
d64b5e85
ED
2073 * @poll: polling function
2074 * @weight: default weight
2075 *
2076 * This variant of netif_napi_add() should be used from drivers using NAPI
2077 * to exclusively poll a TX queue.
2078 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2079 */
2080static inline void netif_tx_napi_add(struct net_device *dev,
2081 struct napi_struct *napi,
2082 int (*poll)(struct napi_struct *, int),
2083 int weight)
2084{
2085 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2086 netif_napi_add(dev, napi, poll, weight);
2087}
2088
d8156534 2089/**
5e82b4b2
BH
2090 * netif_napi_del - remove a NAPI context
2091 * @napi: NAPI context
d8156534 2092 *
5e82b4b2 2093 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2094 */
d565b0a1
HX
2095void netif_napi_del(struct napi_struct *napi);
2096
2097struct napi_gro_cb {
78a478d0 2098 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2099 void *frag0;
78a478d0 2100
7489594c
HX
2101 /* Length of frag0. */
2102 unsigned int frag0_len;
2103
86911732 2104 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2105 int data_offset;
86911732 2106
d565b0a1 2107 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2108 u16 flush;
2109
2110 /* Save the IP ID here and check when we get to the transport layer */
2111 u16 flush_id;
d565b0a1
HX
2112
2113 /* Number of segments aggregated. */
2e71a6f8
ED
2114 u16 count;
2115
15e2396d
TH
2116 /* Start offset for remote checksum offload */
2117 u16 gro_remcsum_start;
2118
2e71a6f8
ED
2119 /* jiffies when first packet was created/queued */
2120 unsigned long age;
86347245 2121
afe93325 2122 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2123 u16 proto;
2124
baa32ff4
TH
2125 /* This is non-zero if the packet may be of the same flow. */
2126 u8 same_flow:1;
2127
fac8e0f5
JG
2128 /* Used in tunnel GRO receive */
2129 u8 encap_mark:1;
573e8fca
TH
2130
2131 /* GRO checksum is valid */
2132 u8 csum_valid:1;
2133
662880f4
TH
2134 /* Number of checksums via CHECKSUM_UNNECESSARY */
2135 u8 csum_cnt:3;
c3c7c254 2136
baa32ff4
TH
2137 /* Free the skb? */
2138 u8 free:2;
2139#define NAPI_GRO_FREE 1
2140#define NAPI_GRO_FREE_STOLEN_HEAD 2
2141
efc98d08
TH
2142 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2143 u8 is_ipv6:1;
2144
a0ca153f
AD
2145 /* Used in GRE, set in fou/gue_gro_receive */
2146 u8 is_fou:1;
2147
1530545e
AD
2148 /* Used to determine if flush_id can be ignored */
2149 u8 is_atomic:1;
2150
fcd91dd4
SD
2151 /* Number of gro_receive callbacks this packet already went through */
2152 u8 recursion_counter:4;
2153
2154 /* 1 bit hole */
baa32ff4 2155
bf5a755f
JC
2156 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2157 __wsum csum;
2158
c3c7c254
ED
2159 /* used in skb_gro_receive() slow path */
2160 struct sk_buff *last;
d565b0a1
HX
2161};
2162
2163#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2164
fcd91dd4
SD
2165#define GRO_RECURSION_LIMIT 15
2166static inline int gro_recursion_inc_test(struct sk_buff *skb)
2167{
2168 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2169}
2170
2171typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2172static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2173 struct sk_buff **head,
2174 struct sk_buff *skb)
2175{
2176 if (unlikely(gro_recursion_inc_test(skb))) {
2177 NAPI_GRO_CB(skb)->flush |= 1;
2178 return NULL;
2179 }
2180
2181 return cb(head, skb);
2182}
2183
2184typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2185 struct sk_buff *);
2186static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2187 struct sock *sk,
2188 struct sk_buff **head,
2189 struct sk_buff *skb)
2190{
2191 if (unlikely(gro_recursion_inc_test(skb))) {
2192 NAPI_GRO_CB(skb)->flush |= 1;
2193 return NULL;
2194 }
2195
2196 return cb(sk, head, skb);
2197}
2198
1da177e4 2199struct packet_type {
f2ccd8fa
DM
2200 __be16 type; /* This is really htons(ether_type). */
2201 struct net_device *dev; /* NULL is wildcarded here */
2202 int (*func) (struct sk_buff *,
2203 struct net_device *,
2204 struct packet_type *,
2205 struct net_device *);
c0de08d0
EL
2206 bool (*id_match)(struct packet_type *ptype,
2207 struct sock *sk);
1da177e4
LT
2208 void *af_packet_priv;
2209 struct list_head list;
2210};
2211
f191a1d1 2212struct offload_callbacks {
576a30eb 2213 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2214 netdev_features_t features);
d565b0a1 2215 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2216 struct sk_buff *skb);
299603e8 2217 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2218};
2219
2220struct packet_offload {
2221 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2222 u16 priority;
f191a1d1
VY
2223 struct offload_callbacks callbacks;
2224 struct list_head list;
1da177e4
LT
2225};
2226
5e82b4b2 2227/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2228struct pcpu_sw_netstats {
2229 u64 rx_packets;
2230 u64 rx_bytes;
2231 u64 tx_packets;
2232 u64 tx_bytes;
2233 struct u64_stats_sync syncp;
2234};
2235
aabc92bb
PNA
2236#define __netdev_alloc_pcpu_stats(type, gfp) \
2237({ \
2238 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2239 if (pcpu_stats) { \
2240 int __cpu; \
2241 for_each_possible_cpu(__cpu) { \
2242 typeof(type) *stat; \
2243 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2244 u64_stats_init(&stat->syncp); \
2245 } \
2246 } \
2247 pcpu_stats; \
1c213bd2
WC
2248})
2249
aabc92bb 2250#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2251 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2252
764f5e54
JP
2253enum netdev_lag_tx_type {
2254 NETDEV_LAG_TX_TYPE_UNKNOWN,
2255 NETDEV_LAG_TX_TYPE_RANDOM,
2256 NETDEV_LAG_TX_TYPE_BROADCAST,
2257 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2258 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2259 NETDEV_LAG_TX_TYPE_HASH,
2260};
2261
2262struct netdev_lag_upper_info {
2263 enum netdev_lag_tx_type tx_type;
2264};
2265
fb1b2e3c
JP
2266struct netdev_lag_lower_state_info {
2267 u8 link_up : 1,
2268 tx_enabled : 1;
2269};
2270
1da177e4
LT
2271#include <linux/notifier.h>
2272
dcfe1421
AW
2273/* netdevice notifier chain. Please remember to update the rtnetlink
2274 * notification exclusion list in rtnetlink_event() when adding new
2275 * types.
2276 */
2277#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2278#define NETDEV_DOWN 0x0002
2279#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2280 detected a hardware crash and restarted
2281 - we can use this eg to kick tcp sessions
2282 once done */
2283#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2284#define NETDEV_REGISTER 0x0005
2285#define NETDEV_UNREGISTER 0x0006
1d486bfb 2286#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2287#define NETDEV_CHANGEADDR 0x0008
2288#define NETDEV_GOING_DOWN 0x0009
2289#define NETDEV_CHANGENAME 0x000A
2290#define NETDEV_FEAT_CHANGE 0x000B
2291#define NETDEV_BONDING_FAILOVER 0x000C
2292#define NETDEV_PRE_UP 0x000D
2293#define NETDEV_PRE_TYPE_CHANGE 0x000E
2294#define NETDEV_POST_TYPE_CHANGE 0x000F
2295#define NETDEV_POST_INIT 0x0010
0115e8e3 2296#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2297#define NETDEV_RELEASE 0x0012
2298#define NETDEV_NOTIFY_PEERS 0x0013
2299#define NETDEV_JOIN 0x0014
42e52bf9 2300#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2301#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2302#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2303#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2304#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2305#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2306#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2307#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2308#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2309
f629d208
JP
2310int register_netdevice_notifier(struct notifier_block *nb);
2311int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2312
2313struct netdev_notifier_info {
2314 struct net_device *dev;
2315};
2316
be9efd36
JP
2317struct netdev_notifier_change_info {
2318 struct netdev_notifier_info info; /* must be first */
2319 unsigned int flags_changed;
2320};
2321
0e4ead9d
JP
2322struct netdev_notifier_changeupper_info {
2323 struct netdev_notifier_info info; /* must be first */
2324 struct net_device *upper_dev; /* new upper dev */
2325 bool master; /* is upper dev master */
5e82b4b2 2326 bool linking; /* is the notification for link or unlink */
29bf24af 2327 void *upper_info; /* upper dev info */
0e4ead9d
JP
2328};
2329
04d48266
JP
2330struct netdev_notifier_changelowerstate_info {
2331 struct netdev_notifier_info info; /* must be first */
2332 void *lower_state_info; /* is lower dev state */
2333};
2334
75538c2b
CW
2335static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2336 struct net_device *dev)
2337{
2338 info->dev = dev;
2339}
2340
351638e7
JP
2341static inline struct net_device *
2342netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2343{
2344 return info->dev;
2345}
2346
f629d208 2347int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2348
2349
1da177e4
LT
2350extern rwlock_t dev_base_lock; /* Device list lock */
2351
881d966b
EB
2352#define for_each_netdev(net, d) \
2353 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2354#define for_each_netdev_reverse(net, d) \
2355 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2356#define for_each_netdev_rcu(net, d) \
2357 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2358#define for_each_netdev_safe(net, d, n) \
2359 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2360#define for_each_netdev_continue(net, d) \
2361 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2362#define for_each_netdev_continue_rcu(net, d) \
2363 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2364#define for_each_netdev_in_bond_rcu(bond, slave) \
2365 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2366 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2367#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2368
a050c33f
DL
2369static inline struct net_device *next_net_device(struct net_device *dev)
2370{
2371 struct list_head *lh;
2372 struct net *net;
2373
c346dca1 2374 net = dev_net(dev);
a050c33f
DL
2375 lh = dev->dev_list.next;
2376 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2377}
2378
ce81b76a
ED
2379static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2380{
2381 struct list_head *lh;
2382 struct net *net;
2383
2384 net = dev_net(dev);
ccf43438 2385 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2386 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2387}
2388
a050c33f
DL
2389static inline struct net_device *first_net_device(struct net *net)
2390{
2391 return list_empty(&net->dev_base_head) ? NULL :
2392 net_device_entry(net->dev_base_head.next);
2393}
7562f876 2394
ccf43438
ED
2395static inline struct net_device *first_net_device_rcu(struct net *net)
2396{
2397 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2398
2399 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2400}
2401
f629d208
JP
2402int netdev_boot_setup_check(struct net_device *dev);
2403unsigned long netdev_boot_base(const char *prefix, int unit);
2404struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2405 const char *hwaddr);
2406struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2407struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2408void dev_add_pack(struct packet_type *pt);
2409void dev_remove_pack(struct packet_type *pt);
2410void __dev_remove_pack(struct packet_type *pt);
2411void dev_add_offload(struct packet_offload *po);
2412void dev_remove_offload(struct packet_offload *po);
f629d208 2413
a54acb3a 2414int dev_get_iflink(const struct net_device *dev);
fc4099f1 2415int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2416struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2417 unsigned short mask);
f629d208
JP
2418struct net_device *dev_get_by_name(struct net *net, const char *name);
2419struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2420struct net_device *__dev_get_by_name(struct net *net, const char *name);
2421int dev_alloc_name(struct net_device *dev, const char *name);
2422int dev_open(struct net_device *dev);
2423int dev_close(struct net_device *dev);
99c4a26a 2424int dev_close_many(struct list_head *head, bool unlink);
f629d208 2425void dev_disable_lro(struct net_device *dev);
0c4b51f0 2426int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2427int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2428int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2429int register_netdevice(struct net_device *dev);
2430void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2431void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2432static inline void unregister_netdevice(struct net_device *dev)
2433{
2434 unregister_netdevice_queue(dev, NULL);
2435}
2436
f629d208
JP
2437int netdev_refcnt_read(const struct net_device *dev);
2438void free_netdev(struct net_device *dev);
74d332c1 2439void netdev_freemem(struct net_device *dev);
f629d208
JP
2440void synchronize_net(void);
2441int init_dummy_netdev(struct net_device *dev);
937f1ba5 2442
f60e5990 2443DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2444#define XMIT_RECURSION_LIMIT 10
2445
f60e5990 2446static inline int dev_recursion_level(void)
2447{
2448 return this_cpu_read(xmit_recursion);
2449}
2450
f629d208
JP
2451struct net_device *dev_get_by_index(struct net *net, int ifindex);
2452struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2453struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2454int netdev_get_name(struct net *net, char *name, int ifindex);
2455int dev_restart(struct net_device *dev);
f629d208 2456int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2457
2458static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2459{
2460 return NAPI_GRO_CB(skb)->data_offset;
2461}
2462
2463static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2464{
2465 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2466}
2467
2468static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2469{
2470 NAPI_GRO_CB(skb)->data_offset += len;
2471}
2472
a5b1cf28
HX
2473static inline void *skb_gro_header_fast(struct sk_buff *skb,
2474 unsigned int offset)
86911732 2475{
a5b1cf28
HX
2476 return NAPI_GRO_CB(skb)->frag0 + offset;
2477}
78a478d0 2478
a5b1cf28
HX
2479static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2480{
2481 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2482}
78a478d0 2483
57ea52a8
HX
2484static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2485{
2486 NAPI_GRO_CB(skb)->frag0 = NULL;
2487 NAPI_GRO_CB(skb)->frag0_len = 0;
2488}
2489
a5b1cf28
HX
2490static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2491 unsigned int offset)
2492{
17dd759c
HX
2493 if (!pskb_may_pull(skb, hlen))
2494 return NULL;
2495
57ea52a8 2496 skb_gro_frag0_invalidate(skb);
17dd759c 2497 return skb->data + offset;
86911732 2498}
1da177e4 2499
36e7b1b8
HX
2500static inline void *skb_gro_network_header(struct sk_buff *skb)
2501{
78d3fd0b
HX
2502 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2503 skb_network_offset(skb);
36e7b1b8
HX
2504}
2505
bf5a755f
JC
2506static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2507 const void *start, unsigned int len)
2508{
573e8fca 2509 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2510 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2511 csum_partial(start, len, 0));
2512}
2513
573e8fca
TH
2514/* GRO checksum functions. These are logical equivalents of the normal
2515 * checksum functions (in skbuff.h) except that they operate on the GRO
2516 * offsets and fields in sk_buff.
2517 */
2518
2519__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2520
15e2396d
TH
2521static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2522{
b7fe10e5 2523 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2524}
2525
573e8fca
TH
2526static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2527 bool zero_okay,
2528 __sum16 check)
2529{
6edec0e6
TH
2530 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2531 skb_checksum_start_offset(skb) <
2532 skb_gro_offset(skb)) &&
15e2396d 2533 !skb_at_gro_remcsum_start(skb) &&
662880f4 2534 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2535 (!zero_okay || check));
2536}
2537
2538static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2539 __wsum psum)
2540{
2541 if (NAPI_GRO_CB(skb)->csum_valid &&
2542 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2543 return 0;
2544
2545 NAPI_GRO_CB(skb)->csum = psum;
2546
2547 return __skb_gro_checksum_complete(skb);
2548}
2549
573e8fca
TH
2550static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2551{
662880f4
TH
2552 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2553 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2554 NAPI_GRO_CB(skb)->csum_cnt--;
2555 } else {
2556 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2557 * verified a new top level checksum or an encapsulated one
2558 * during GRO. This saves work if we fallback to normal path.
2559 */
2560 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2561 }
2562}
2563
2564#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2565 compute_pseudo) \
2566({ \
2567 __sum16 __ret = 0; \
2568 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2569 __ret = __skb_gro_checksum_validate_complete(skb, \
2570 compute_pseudo(skb, proto)); \
5a212329
TH
2571 if (__ret) \
2572 __skb_mark_checksum_bad(skb); \
2573 else \
573e8fca
TH
2574 skb_gro_incr_csum_unnecessary(skb); \
2575 __ret; \
2576})
2577
2578#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2579 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2580
2581#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2582 compute_pseudo) \
2583 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2584
2585#define skb_gro_checksum_simple_validate(skb) \
2586 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2587
d96535a1
TH
2588static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2589{
2590 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2591 !NAPI_GRO_CB(skb)->csum_valid);
2592}
2593
2594static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2595 __sum16 check, __wsum pseudo)
2596{
2597 NAPI_GRO_CB(skb)->csum = ~pseudo;
2598 NAPI_GRO_CB(skb)->csum_valid = 1;
2599}
2600
2601#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2602do { \
2603 if (__skb_gro_checksum_convert_check(skb)) \
2604 __skb_gro_checksum_convert(skb, check, \
2605 compute_pseudo(skb, proto)); \
2606} while (0)
2607
26c4f7da
TH
2608struct gro_remcsum {
2609 int offset;
2610 __wsum delta;
2611};
2612
2613static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2614{
846cd667 2615 grc->offset = 0;
26c4f7da
TH
2616 grc->delta = 0;
2617}
2618
b7fe10e5
TH
2619static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2620 unsigned int off, size_t hdrlen,
2621 int start, int offset,
2622 struct gro_remcsum *grc,
2623 bool nopartial)
dcdc8994
TH
2624{
2625 __wsum delta;
b7fe10e5 2626 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2627
2628 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2629
15e2396d 2630 if (!nopartial) {
b7fe10e5
TH
2631 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2632 return ptr;
2633 }
2634
2635 ptr = skb_gro_header_fast(skb, off);
2636 if (skb_gro_header_hard(skb, off + plen)) {
2637 ptr = skb_gro_header_slow(skb, off + plen, off);
2638 if (!ptr)
2639 return NULL;
15e2396d
TH
2640 }
2641
b7fe10e5
TH
2642 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2643 start, offset);
dcdc8994
TH
2644
2645 /* Adjust skb->csum since we changed the packet */
dcdc8994 2646 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2647
b7fe10e5 2648 grc->offset = off + hdrlen + offset;
26c4f7da 2649 grc->delta = delta;
b7fe10e5
TH
2650
2651 return ptr;
dcdc8994
TH
2652}
2653
26c4f7da
TH
2654static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2655 struct gro_remcsum *grc)
2656{
b7fe10e5
TH
2657 void *ptr;
2658 size_t plen = grc->offset + sizeof(u16);
2659
26c4f7da
TH
2660 if (!grc->delta)
2661 return;
2662
b7fe10e5
TH
2663 ptr = skb_gro_header_fast(skb, grc->offset);
2664 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2665 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2666 if (!ptr)
2667 return;
2668 }
2669
2670 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2671}
dcdc8994 2672
25393d3f
SK
2673#ifdef CONFIG_XFRM_OFFLOAD
2674static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2675{
2676 if (PTR_ERR(pp) != -EINPROGRESS)
2677 NAPI_GRO_CB(skb)->flush |= flush;
2678}
2679#else
5f114163
SK
2680static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2681{
2682 NAPI_GRO_CB(skb)->flush |= flush;
2683}
25393d3f 2684#endif
5f114163 2685
0c4e8581
SH
2686static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2687 unsigned short type,
3b04ddde 2688 const void *daddr, const void *saddr,
95c96174 2689 unsigned int len)
0c4e8581 2690{
f1ecfd5d 2691 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2692 return 0;
3b04ddde
SH
2693
2694 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2695}
2696
b95cce35
SH
2697static inline int dev_parse_header(const struct sk_buff *skb,
2698 unsigned char *haddr)
2699{
2700 const struct net_device *dev = skb->dev;
2701
1b83336b 2702 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2703 return 0;
3b04ddde 2704 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2705}
2706
2793a23a
WB
2707/* ll_header must have at least hard_header_len allocated */
2708static inline bool dev_validate_header(const struct net_device *dev,
2709 char *ll_header, int len)
2710{
2711 if (likely(len >= dev->hard_header_len))
2712 return true;
217e6fa2
WB
2713 if (len < dev->min_header_len)
2714 return false;
2793a23a
WB
2715
2716 if (capable(CAP_SYS_RAWIO)) {
2717 memset(ll_header + len, 0, dev->hard_header_len - len);
2718 return true;
2719 }
2720
2721 if (dev->header_ops && dev->header_ops->validate)
2722 return dev->header_ops->validate(ll_header, len);
2723
2724 return false;
2725}
2726
1da177e4 2727typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2728int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2729static inline int unregister_gifconf(unsigned int family)
2730{
2731 return register_gifconf(family, NULL);
2732}
2733
99bbc707 2734#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2735#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2736struct sd_flow_limit {
2737 u64 count;
2738 unsigned int num_buckets;
2739 unsigned int history_head;
2740 u16 history[FLOW_LIMIT_HISTORY];
2741 u8 buckets[];
2742};
2743
2744extern int netdev_flow_limit_table_len;
2745#endif /* CONFIG_NET_FLOW_LIMIT */
2746
1da177e4 2747/*
5e82b4b2 2748 * Incoming packets are placed on per-CPU queues
1da177e4 2749 */
d94d9fee 2750struct softnet_data {
1da177e4 2751 struct list_head poll_list;
6e7676c1 2752 struct sk_buff_head process_queue;
1da177e4 2753
dee42870 2754 /* stats */
cd7b5396
DM
2755 unsigned int processed;
2756 unsigned int time_squeeze;
cd7b5396 2757 unsigned int received_rps;
fd793d89 2758#ifdef CONFIG_RPS
88751275 2759 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2760#endif
2761#ifdef CONFIG_NET_FLOW_LIMIT
2762 struct sd_flow_limit __rcu *flow_limit;
2763#endif
2764 struct Qdisc *output_queue;
2765 struct Qdisc **output_queue_tailp;
2766 struct sk_buff *completion_queue;
88751275 2767
4cdb1e2e 2768#ifdef CONFIG_RPS
501e7ef5
ED
2769 /* input_queue_head should be written by cpu owning this struct,
2770 * and only read by other cpus. Worth using a cache line.
2771 */
2772 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2773
2774 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2775 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2776 struct softnet_data *rps_ipi_next;
2777 unsigned int cpu;
76cc8b13 2778 unsigned int input_queue_tail;
1e94d72f 2779#endif
95c96174 2780 unsigned int dropped;
0a9627f2 2781 struct sk_buff_head input_pkt_queue;
bea3348e 2782 struct napi_struct backlog;
99bbc707 2783
1da177e4
LT
2784};
2785
76cc8b13 2786static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2787{
2788#ifdef CONFIG_RPS
76cc8b13
TH
2789 sd->input_queue_head++;
2790#endif
2791}
2792
2793static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2794 unsigned int *qtail)
2795{
2796#ifdef CONFIG_RPS
2797 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2798#endif
2799}
2800
0a9627f2 2801DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2802
f629d208 2803void __netif_schedule(struct Qdisc *q);
46e5da40 2804void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2805
fd2ea0a7
DM
2806static inline void netif_tx_schedule_all(struct net_device *dev)
2807{
2808 unsigned int i;
2809
2810 for (i = 0; i < dev->num_tx_queues; i++)
2811 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2812}
2813
f9a7cbbf 2814static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2815{
73466498 2816 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2817}
2818
bea3348e
SH
2819/**
2820 * netif_start_queue - allow transmit
2821 * @dev: network device
2822 *
2823 * Allow upper layers to call the device hard_start_xmit routine.
2824 */
1da177e4
LT
2825static inline void netif_start_queue(struct net_device *dev)
2826{
e8a0464c 2827 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2828}
2829
fd2ea0a7
DM
2830static inline void netif_tx_start_all_queues(struct net_device *dev)
2831{
2832 unsigned int i;
2833
2834 for (i = 0; i < dev->num_tx_queues; i++) {
2835 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2836 netif_tx_start_queue(txq);
2837 }
2838}
2839
46e5da40 2840void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2841
d29f749e
DJ
2842/**
2843 * netif_wake_queue - restart transmit
2844 * @dev: network device
2845 *
2846 * Allow upper layers to call the device hard_start_xmit routine.
2847 * Used for flow control when transmit resources are available.
2848 */
79d16385
DM
2849static inline void netif_wake_queue(struct net_device *dev)
2850{
e8a0464c 2851 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2852}
2853
fd2ea0a7
DM
2854static inline void netif_tx_wake_all_queues(struct net_device *dev)
2855{
2856 unsigned int i;
2857
2858 for (i = 0; i < dev->num_tx_queues; i++) {
2859 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2860 netif_tx_wake_queue(txq);
2861 }
2862}
2863
f9a7cbbf 2864static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2865{
73466498 2866 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2867}
2868
bea3348e
SH
2869/**
2870 * netif_stop_queue - stop transmitted packets
2871 * @dev: network device
2872 *
2873 * Stop upper layers calling the device hard_start_xmit routine.
2874 * Used for flow control when transmit resources are unavailable.
2875 */
1da177e4
LT
2876static inline void netif_stop_queue(struct net_device *dev)
2877{
e8a0464c 2878 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2879}
2880
a2029240 2881void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2882
4d29515f 2883static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2884{
73466498 2885 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2886}
2887
bea3348e
SH
2888/**
2889 * netif_queue_stopped - test if transmit queue is flowblocked
2890 * @dev: network device
2891 *
2892 * Test if transmit queue on device is currently unable to send.
2893 */
4d29515f 2894static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2895{
e8a0464c 2896 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2897}
2898
4d29515f 2899static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2900{
73466498
TH
2901 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2902}
2903
8e2f1a63
DB
2904static inline bool
2905netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2906{
2907 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2908}
2909
8e2f1a63
DB
2910static inline bool
2911netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2912{
2913 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2914}
2915
53511453
ED
2916/**
2917 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2918 * @dev_queue: pointer to transmit queue
2919 *
2920 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2921 * to give appropriate hint to the CPU.
53511453
ED
2922 */
2923static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2924{
2925#ifdef CONFIG_BQL
2926 prefetchw(&dev_queue->dql.num_queued);
2927#endif
2928}
2929
2930/**
2931 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2932 * @dev_queue: pointer to transmit queue
2933 *
2934 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2935 * to give appropriate hint to the CPU.
53511453
ED
2936 */
2937static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2938{
2939#ifdef CONFIG_BQL
2940 prefetchw(&dev_queue->dql.limit);
2941#endif
2942}
2943
c5d67bd7
TH
2944static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2945 unsigned int bytes)
2946{
114cf580
TH
2947#ifdef CONFIG_BQL
2948 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2949
2950 if (likely(dql_avail(&dev_queue->dql) >= 0))
2951 return;
2952
2953 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2954
2955 /*
2956 * The XOFF flag must be set before checking the dql_avail below,
2957 * because in netdev_tx_completed_queue we update the dql_completed
2958 * before checking the XOFF flag.
2959 */
2960 smp_mb();
2961
2962 /* check again in case another CPU has just made room avail */
2963 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2964 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2965#endif
c5d67bd7
TH
2966}
2967
0042d0c8
FF
2968/**
2969 * netdev_sent_queue - report the number of bytes queued to hardware
2970 * @dev: network device
2971 * @bytes: number of bytes queued to the hardware device queue
2972 *
2973 * Report the number of bytes queued for sending/completion to the network
2974 * device hardware queue. @bytes should be a good approximation and should
2975 * exactly match netdev_completed_queue() @bytes
2976 */
c5d67bd7
TH
2977static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2978{
2979 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2980}
2981
2982static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2983 unsigned int pkts, unsigned int bytes)
c5d67bd7 2984{
114cf580 2985#ifdef CONFIG_BQL
b37c0fbe
AD
2986 if (unlikely(!bytes))
2987 return;
2988
2989 dql_completed(&dev_queue->dql, bytes);
2990
2991 /*
2992 * Without the memory barrier there is a small possiblity that
2993 * netdev_tx_sent_queue will miss the update and cause the queue to
2994 * be stopped forever
2995 */
2996 smp_mb();
2997
2998 if (dql_avail(&dev_queue->dql) < 0)
2999 return;
3000
3001 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3002 netif_schedule_queue(dev_queue);
114cf580 3003#endif
c5d67bd7
TH
3004}
3005
0042d0c8
FF
3006/**
3007 * netdev_completed_queue - report bytes and packets completed by device
3008 * @dev: network device
3009 * @pkts: actual number of packets sent over the medium
3010 * @bytes: actual number of bytes sent over the medium
3011 *
3012 * Report the number of bytes and packets transmitted by the network device
3013 * hardware queue over the physical medium, @bytes must exactly match the
3014 * @bytes amount passed to netdev_sent_queue()
3015 */
c5d67bd7 3016static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3017 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3018{
3019 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3020}
3021
3022static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3023{
114cf580 3024#ifdef CONFIG_BQL
5c490354 3025 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3026 dql_reset(&q->dql);
3027#endif
c5d67bd7
TH
3028}
3029
0042d0c8
FF
3030/**
3031 * netdev_reset_queue - reset the packets and bytes count of a network device
3032 * @dev_queue: network device
3033 *
3034 * Reset the bytes and packet count of a network device and clear the
3035 * software flow control OFF bit for this network device
3036 */
c5d67bd7
TH
3037static inline void netdev_reset_queue(struct net_device *dev_queue)
3038{
3039 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3040}
3041
b9507bda
DB
3042/**
3043 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3044 * @dev: network device
3045 * @queue_index: given tx queue index
3046 *
3047 * Returns 0 if given tx queue index >= number of device tx queues,
3048 * otherwise returns the originally passed tx queue index.
3049 */
3050static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3051{
3052 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3053 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3054 dev->name, queue_index,
3055 dev->real_num_tx_queues);
3056 return 0;
3057 }
3058
3059 return queue_index;
3060}
3061
bea3348e
SH
3062/**
3063 * netif_running - test if up
3064 * @dev: network device
3065 *
3066 * Test if the device has been brought up.
3067 */
4d29515f 3068static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3069{
3070 return test_bit(__LINK_STATE_START, &dev->state);
3071}
3072
f25f4e44 3073/*
5e82b4b2 3074 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3075 * stop, and a check if it's stopped. All other device management is
3076 * done at the overall netdevice level.
3077 * Also test the device if we're multiqueue.
3078 */
bea3348e
SH
3079
3080/**
3081 * netif_start_subqueue - allow sending packets on subqueue
3082 * @dev: network device
3083 * @queue_index: sub queue index
3084 *
3085 * Start individual transmit queue of a device with multiple transmit queues.
3086 */
f25f4e44
PWJ
3087static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3088{
fd2ea0a7 3089 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3090
3091 netif_tx_start_queue(txq);
f25f4e44
PWJ
3092}
3093
bea3348e
SH
3094/**
3095 * netif_stop_subqueue - stop sending packets on subqueue
3096 * @dev: network device
3097 * @queue_index: sub queue index
3098 *
3099 * Stop individual transmit queue of a device with multiple transmit queues.
3100 */
f25f4e44
PWJ
3101static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3102{
fd2ea0a7 3103 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3104 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3105}
3106
bea3348e
SH
3107/**
3108 * netif_subqueue_stopped - test status of subqueue
3109 * @dev: network device
3110 * @queue_index: sub queue index
3111 *
3112 * Check individual transmit queue of a device with multiple transmit queues.
3113 */
4d29515f
DM
3114static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3115 u16 queue_index)
f25f4e44 3116{
fd2ea0a7 3117 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3118
3119 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3120}
3121
4d29515f
DM
3122static inline bool netif_subqueue_stopped(const struct net_device *dev,
3123 struct sk_buff *skb)
668f895a
PE
3124{
3125 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3126}
bea3348e 3127
738b35cc
FF
3128/**
3129 * netif_wake_subqueue - allow sending packets on subqueue
3130 * @dev: network device
3131 * @queue_index: sub queue index
3132 *
3133 * Resume individual transmit queue of a device with multiple transmit queues.
3134 */
3135static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3136{
3137 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3138
3139 netif_tx_wake_queue(txq);
3140}
f25f4e44 3141
537c00de 3142#ifdef CONFIG_XPS
53af53ae 3143int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3144 u16 index);
537c00de
AD
3145#else
3146static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3147 const struct cpumask *mask,
537c00de
AD
3148 u16 index)
3149{
3150 return 0;
3151}
3152#endif
3153
5605c762
JP
3154u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3155 unsigned int num_tx_queues);
3156
a3d22a68
VZ
3157/*
3158 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3159 * as a distribution range limit for the returned value.
3160 */
3161static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3162 struct sk_buff *skb)
a3d22a68
VZ
3163{
3164 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3165}
3166
bea3348e
SH
3167/**
3168 * netif_is_multiqueue - test if device has multiple transmit queues
3169 * @dev: network device
3170 *
3171 * Check if device has multiple transmit queues
bea3348e 3172 */
4d29515f 3173static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3174{
a02cec21 3175 return dev->num_tx_queues > 1;
f25f4e44 3176}
1da177e4 3177
f629d208 3178int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3179
a953be53 3180#ifdef CONFIG_SYSFS
f629d208 3181int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3182#else
3183static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3184 unsigned int rxq)
3185{
3186 return 0;
3187}
3188#endif
3189
a953be53
MD
3190#ifdef CONFIG_SYSFS
3191static inline unsigned int get_netdev_rx_queue_index(
3192 struct netdev_rx_queue *queue)
3193{
3194 struct net_device *dev = queue->dev;
3195 int index = queue - dev->_rx;
3196
3197 BUG_ON(index >= dev->num_rx_queues);
3198 return index;
3199}
3200#endif
3201
16917b87 3202#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3203int netif_get_num_default_rss_queues(void);
16917b87 3204
e6247027
ED
3205enum skb_free_reason {
3206 SKB_REASON_CONSUMED,
3207 SKB_REASON_DROPPED,
3208};
3209
3210void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3211void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3212
e6247027
ED
3213/*
3214 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3215 * interrupt context or with hardware interrupts being disabled.
3216 * (in_irq() || irqs_disabled())
3217 *
3218 * We provide four helpers that can be used in following contexts :
3219 *
3220 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3221 * replacing kfree_skb(skb)
3222 *
3223 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3224 * Typically used in place of consume_skb(skb) in TX completion path
3225 *
3226 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3227 * replacing kfree_skb(skb)
3228 *
3229 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3230 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3231 */
e6247027
ED
3232static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3233{
3234 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3235}
3236
3237static inline void dev_consume_skb_irq(struct sk_buff *skb)
3238{
3239 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3240}
3241
3242static inline void dev_kfree_skb_any(struct sk_buff *skb)
3243{
3244 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3245}
3246
3247static inline void dev_consume_skb_any(struct sk_buff *skb)
3248{
3249 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3250}
1da177e4 3251
f629d208
JP
3252int netif_rx(struct sk_buff *skb);
3253int netif_rx_ni(struct sk_buff *skb);
04eb4489 3254int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3255gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3256void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3257struct sk_buff *napi_get_frags(struct napi_struct *napi);
3258gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3259struct packet_offload *gro_find_receive_by_type(__be16 type);
3260struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3261
3262static inline void napi_free_frags(struct napi_struct *napi)
3263{
3264 kfree_skb(napi->skb);
3265 napi->skb = NULL;
3266}
3267
24b27fc4 3268bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3269int netdev_rx_handler_register(struct net_device *dev,
3270 rx_handler_func_t *rx_handler,
3271 void *rx_handler_data);
3272void netdev_rx_handler_unregister(struct net_device *dev);
3273
3274bool dev_valid_name(const char *name);
3275int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3276int dev_ethtool(struct net *net, struct ifreq *);
3277unsigned int dev_get_flags(const struct net_device *);
3278int __dev_change_flags(struct net_device *, unsigned int flags);
3279int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3280void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3281 unsigned int gchanges);
f629d208
JP
3282int dev_change_name(struct net_device *, const char *);
3283int dev_set_alias(struct net_device *, const char *, size_t);
3284int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3285int dev_set_mtu(struct net_device *, int);
3286void dev_set_group(struct net_device *, int);
3287int dev_set_mac_address(struct net_device *, struct sockaddr *);
3288int dev_change_carrier(struct net_device *, bool new_carrier);
3289int dev_get_phys_port_id(struct net_device *dev,
02637fce 3290 struct netdev_phys_item_id *ppid);
db24a904
DA
3291int dev_get_phys_port_name(struct net_device *dev,
3292 char *name, size_t len);
d746d707 3293int dev_change_proto_down(struct net_device *dev, bool proto_down);
85de8576 3294int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
55a93b3e 3295struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3296struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3297 struct netdev_queue *txq, int *ret);
a0265d28 3298int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3299int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3300bool is_skb_forwardable(const struct net_device *dev,
3301 const struct sk_buff *skb);
1da177e4 3302
4e3264d2
MKL
3303static __always_inline int ____dev_forward_skb(struct net_device *dev,
3304 struct sk_buff *skb)
3305{
3306 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3307 unlikely(!is_skb_forwardable(dev, skb))) {
3308 atomic_long_inc(&dev->rx_dropped);
3309 kfree_skb(skb);
3310 return NET_RX_DROP;
3311 }
3312
3313 skb_scrub_packet(skb, true);
3314 skb->priority = 0;
3315 return 0;
3316}
3317
74b20582
DA
3318void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3319
20380731 3320extern int netdev_budget;
7acf8a1e 3321extern unsigned int netdev_budget_usecs;
1da177e4
LT
3322
3323/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3324void netdev_run_todo(void);
1da177e4 3325
bea3348e
SH
3326/**
3327 * dev_put - release reference to device
3328 * @dev: network device
3329 *
9ef4429b 3330 * Release reference to device to allow it to be freed.
bea3348e 3331 */
1da177e4
LT
3332static inline void dev_put(struct net_device *dev)
3333{
933393f5 3334 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3335}
3336
bea3348e
SH
3337/**
3338 * dev_hold - get reference to device
3339 * @dev: network device
3340 *
9ef4429b 3341 * Hold reference to device to keep it from being freed.
bea3348e 3342 */
15333061
SH
3343static inline void dev_hold(struct net_device *dev)
3344{
933393f5 3345 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3346}
1da177e4
LT
3347
3348/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3349 * and _off may be called from IRQ context, but it is caller
3350 * who is responsible for serialization of these calls.
b00055aa
SR
3351 *
3352 * The name carrier is inappropriate, these functions should really be
3353 * called netif_lowerlayer_*() because they represent the state of any
3354 * kind of lower layer not just hardware media.
1da177e4
LT
3355 */
3356
f629d208
JP
3357void linkwatch_init_dev(struct net_device *dev);
3358void linkwatch_fire_event(struct net_device *dev);
3359void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3360
bea3348e
SH
3361/**
3362 * netif_carrier_ok - test if carrier present
3363 * @dev: network device
3364 *
3365 * Check if carrier is present on device
3366 */
4d29515f 3367static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3368{
3369 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3370}
3371
f629d208 3372unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3373
f629d208 3374void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3375
f629d208 3376void netif_carrier_on(struct net_device *dev);
1da177e4 3377
f629d208 3378void netif_carrier_off(struct net_device *dev);
1da177e4 3379
bea3348e
SH
3380/**
3381 * netif_dormant_on - mark device as dormant.
3382 * @dev: network device
3383 *
3384 * Mark device as dormant (as per RFC2863).
3385 *
3386 * The dormant state indicates that the relevant interface is not
3387 * actually in a condition to pass packets (i.e., it is not 'up') but is
3388 * in a "pending" state, waiting for some external event. For "on-
3389 * demand" interfaces, this new state identifies the situation where the
3390 * interface is waiting for events to place it in the up state.
bea3348e 3391 */
b00055aa
SR
3392static inline void netif_dormant_on(struct net_device *dev)
3393{
3394 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3395 linkwatch_fire_event(dev);
3396}
3397
bea3348e
SH
3398/**
3399 * netif_dormant_off - set device as not dormant.
3400 * @dev: network device
3401 *
3402 * Device is not in dormant state.
3403 */
b00055aa
SR
3404static inline void netif_dormant_off(struct net_device *dev)
3405{
3406 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3407 linkwatch_fire_event(dev);
3408}
3409
bea3348e 3410/**
8ecbc40a 3411 * netif_dormant - test if device is dormant
bea3348e
SH
3412 * @dev: network device
3413 *
8ecbc40a 3414 * Check if device is dormant.
bea3348e 3415 */
4d29515f 3416static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3417{
3418 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3419}
3420
3421
bea3348e
SH
3422/**
3423 * netif_oper_up - test if device is operational
3424 * @dev: network device
3425 *
3426 * Check if carrier is operational
3427 */
4d29515f 3428static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3429{
b00055aa
SR
3430 return (dev->operstate == IF_OPER_UP ||
3431 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3432}
3433
bea3348e
SH
3434/**
3435 * netif_device_present - is device available or removed
3436 * @dev: network device
3437 *
3438 * Check if device has not been removed from system.
3439 */
4d29515f 3440static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3441{
3442 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3443}
3444
f629d208 3445void netif_device_detach(struct net_device *dev);
1da177e4 3446
f629d208 3447void netif_device_attach(struct net_device *dev);
1da177e4
LT
3448
3449/*
3450 * Network interface message level settings
3451 */
1da177e4
LT
3452
3453enum {
3454 NETIF_MSG_DRV = 0x0001,
3455 NETIF_MSG_PROBE = 0x0002,
3456 NETIF_MSG_LINK = 0x0004,
3457 NETIF_MSG_TIMER = 0x0008,
3458 NETIF_MSG_IFDOWN = 0x0010,
3459 NETIF_MSG_IFUP = 0x0020,
3460 NETIF_MSG_RX_ERR = 0x0040,
3461 NETIF_MSG_TX_ERR = 0x0080,
3462 NETIF_MSG_TX_QUEUED = 0x0100,
3463 NETIF_MSG_INTR = 0x0200,
3464 NETIF_MSG_TX_DONE = 0x0400,
3465 NETIF_MSG_RX_STATUS = 0x0800,
3466 NETIF_MSG_PKTDATA = 0x1000,
3467 NETIF_MSG_HW = 0x2000,
3468 NETIF_MSG_WOL = 0x4000,
3469};
3470
3471#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3472#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3473#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3474#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3475#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3476#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3477#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3478#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3479#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3480#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3481#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3482#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3483#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3484#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3485#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3486
3487static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3488{
3489 /* use default */
3490 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3491 return default_msg_enable_bits;
3492 if (debug_value == 0) /* no output */
3493 return 0;
3494 /* set low N bits */
3495 return (1 << debug_value) - 1;
3496}
3497
c773e847 3498static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3499{
c773e847
DM
3500 spin_lock(&txq->_xmit_lock);
3501 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3502}
3503
5a717f4f
MT
3504static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3505{
3506 __acquire(&txq->_xmit_lock);
3507 return true;
3508}
3509
3510static inline void __netif_tx_release(struct netdev_queue *txq)
3511{
3512 __release(&txq->_xmit_lock);
3513}
3514
fd2ea0a7
DM
3515static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3516{
3517 spin_lock_bh(&txq->_xmit_lock);
3518 txq->xmit_lock_owner = smp_processor_id();
3519}
3520
4d29515f 3521static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3522{
4d29515f 3523 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3524 if (likely(ok))
3525 txq->xmit_lock_owner = smp_processor_id();
3526 return ok;
3527}
3528
3529static inline void __netif_tx_unlock(struct netdev_queue *txq)
3530{
3531 txq->xmit_lock_owner = -1;
3532 spin_unlock(&txq->_xmit_lock);
3533}
3534
3535static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3536{
3537 txq->xmit_lock_owner = -1;
3538 spin_unlock_bh(&txq->_xmit_lock);
3539}
3540
08baf561
ED
3541static inline void txq_trans_update(struct netdev_queue *txq)
3542{
3543 if (txq->xmit_lock_owner != -1)
3544 txq->trans_start = jiffies;
3545}
3546
ba162f8e
FW
3547/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3548static inline void netif_trans_update(struct net_device *dev)
3549{
9b36627a
FW
3550 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3551
3552 if (txq->trans_start != jiffies)
3553 txq->trans_start = jiffies;
ba162f8e
FW
3554}
3555
d29f749e
DJ
3556/**
3557 * netif_tx_lock - grab network device transmit lock
3558 * @dev: network device
d29f749e
DJ
3559 *
3560 * Get network device transmit lock
3561 */
22dd7495
JHS
3562static inline void netif_tx_lock(struct net_device *dev)
3563{
e8a0464c 3564 unsigned int i;
c3f26a26 3565 int cpu;
c773e847 3566
c3f26a26
DM
3567 spin_lock(&dev->tx_global_lock);
3568 cpu = smp_processor_id();
e8a0464c
DM
3569 for (i = 0; i < dev->num_tx_queues; i++) {
3570 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3571
3572 /* We are the only thread of execution doing a
3573 * freeze, but we have to grab the _xmit_lock in
3574 * order to synchronize with threads which are in
3575 * the ->hard_start_xmit() handler and already
3576 * checked the frozen bit.
3577 */
e8a0464c 3578 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3579 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3580 __netif_tx_unlock(txq);
e8a0464c 3581 }
932ff279
HX
3582}
3583
3584static inline void netif_tx_lock_bh(struct net_device *dev)
3585{
e8a0464c
DM
3586 local_bh_disable();
3587 netif_tx_lock(dev);
932ff279
HX
3588}
3589
932ff279
HX
3590static inline void netif_tx_unlock(struct net_device *dev)
3591{
e8a0464c
DM
3592 unsigned int i;
3593
3594 for (i = 0; i < dev->num_tx_queues; i++) {
3595 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3596
c3f26a26
DM
3597 /* No need to grab the _xmit_lock here. If the
3598 * queue is not stopped for another reason, we
3599 * force a schedule.
3600 */
3601 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3602 netif_schedule_queue(txq);
c3f26a26
DM
3603 }
3604 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3605}
3606
3607static inline void netif_tx_unlock_bh(struct net_device *dev)
3608{
e8a0464c
DM
3609 netif_tx_unlock(dev);
3610 local_bh_enable();
932ff279
HX
3611}
3612
c773e847 3613#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3614 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3615 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3616 } else { \
3617 __netif_tx_acquire(txq); \
22dd7495
JHS
3618 } \
3619}
3620
5efeac44
EB
3621#define HARD_TX_TRYLOCK(dev, txq) \
3622 (((dev->features & NETIF_F_LLTX) == 0) ? \
3623 __netif_tx_trylock(txq) : \
5a717f4f 3624 __netif_tx_acquire(txq))
5efeac44 3625
c773e847 3626#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3627 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3628 __netif_tx_unlock(txq); \
5a717f4f
MT
3629 } else { \
3630 __netif_tx_release(txq); \
22dd7495
JHS
3631 } \
3632}
3633
1da177e4
LT
3634static inline void netif_tx_disable(struct net_device *dev)
3635{
fd2ea0a7 3636 unsigned int i;
c3f26a26 3637 int cpu;
fd2ea0a7 3638
c3f26a26
DM
3639 local_bh_disable();
3640 cpu = smp_processor_id();
fd2ea0a7
DM
3641 for (i = 0; i < dev->num_tx_queues; i++) {
3642 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3643
3644 __netif_tx_lock(txq, cpu);
fd2ea0a7 3645 netif_tx_stop_queue(txq);
c3f26a26 3646 __netif_tx_unlock(txq);
fd2ea0a7 3647 }
c3f26a26 3648 local_bh_enable();
1da177e4
LT
3649}
3650
e308a5d8
DM
3651static inline void netif_addr_lock(struct net_device *dev)
3652{
3653 spin_lock(&dev->addr_list_lock);
3654}
3655
2429f7ac
JP
3656static inline void netif_addr_lock_nested(struct net_device *dev)
3657{
25175ba5
VY
3658 int subclass = SINGLE_DEPTH_NESTING;
3659
3660 if (dev->netdev_ops->ndo_get_lock_subclass)
3661 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3662
3663 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3664}
3665
e308a5d8
DM
3666static inline void netif_addr_lock_bh(struct net_device *dev)
3667{
3668 spin_lock_bh(&dev->addr_list_lock);
3669}
3670
3671static inline void netif_addr_unlock(struct net_device *dev)
3672{
3673 spin_unlock(&dev->addr_list_lock);
3674}
3675
3676static inline void netif_addr_unlock_bh(struct net_device *dev)
3677{
3678 spin_unlock_bh(&dev->addr_list_lock);
3679}
3680
f001fde5 3681/*
31278e71 3682 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3683 * rcu_read_lock held.
3684 */
3685#define for_each_dev_addr(dev, ha) \
31278e71 3686 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3687
1da177e4
LT
3688/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3689
f629d208 3690void ether_setup(struct net_device *dev);
1da177e4
LT
3691
3692/* Support for loadable net-drivers */
f629d208 3693struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3694 unsigned char name_assign_type,
f629d208
JP
3695 void (*setup)(struct net_device *),
3696 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3697#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3698 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3699
c835a677
TG
3700#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3701 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3702 count)
36909ea4 3703
f629d208
JP
3704int register_netdev(struct net_device *dev);
3705void unregister_netdev(struct net_device *dev);
f001fde5 3706
22bedad3 3707/* General hardware address lists handling functions */
f629d208
JP
3708int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3709 struct netdev_hw_addr_list *from_list, int addr_len);
3710void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3711 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3712int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3713 struct net_device *dev,
3714 int (*sync)(struct net_device *, const unsigned char *),
3715 int (*unsync)(struct net_device *,
3716 const unsigned char *));
3717void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3718 struct net_device *dev,
3719 int (*unsync)(struct net_device *,
3720 const unsigned char *));
f629d208 3721void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3722
f001fde5 3723/* Functions used for device addresses handling */
f629d208
JP
3724int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3725 unsigned char addr_type);
3726int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3727 unsigned char addr_type);
f629d208
JP
3728void dev_addr_flush(struct net_device *dev);
3729int dev_addr_init(struct net_device *dev);
a748ee24
JP
3730
3731/* Functions used for unicast addresses handling */
f629d208
JP
3732int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3733int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3734int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3735int dev_uc_sync(struct net_device *to, struct net_device *from);
3736int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3737void dev_uc_unsync(struct net_device *to, struct net_device *from);
3738void dev_uc_flush(struct net_device *dev);
3739void dev_uc_init(struct net_device *dev);
f001fde5 3740
670e5b8e
AD
3741/**
3742 * __dev_uc_sync - Synchonize device's unicast list
3743 * @dev: device to sync
3744 * @sync: function to call if address should be added
3745 * @unsync: function to call if address should be removed
3746 *
3747 * Add newly added addresses to the interface, and release
3748 * addresses that have been deleted.
5e82b4b2 3749 */
670e5b8e
AD
3750static inline int __dev_uc_sync(struct net_device *dev,
3751 int (*sync)(struct net_device *,
3752 const unsigned char *),
3753 int (*unsync)(struct net_device *,
3754 const unsigned char *))
3755{
3756 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3757}
3758
3759/**
e793c0f7 3760 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3761 * @dev: device to sync
3762 * @unsync: function to call if address should be removed
3763 *
3764 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3765 */
670e5b8e
AD
3766static inline void __dev_uc_unsync(struct net_device *dev,
3767 int (*unsync)(struct net_device *,
3768 const unsigned char *))
3769{
3770 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3771}
3772
22bedad3 3773/* Functions used for multicast addresses handling */
f629d208
JP
3774int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3775int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3776int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3777int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3778int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3779int dev_mc_sync(struct net_device *to, struct net_device *from);
3780int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3781void dev_mc_unsync(struct net_device *to, struct net_device *from);
3782void dev_mc_flush(struct net_device *dev);
3783void dev_mc_init(struct net_device *dev);
f001fde5 3784
670e5b8e
AD
3785/**
3786 * __dev_mc_sync - Synchonize device's multicast list
3787 * @dev: device to sync
3788 * @sync: function to call if address should be added
3789 * @unsync: function to call if address should be removed
3790 *
3791 * Add newly added addresses to the interface, and release
3792 * addresses that have been deleted.
5e82b4b2 3793 */
670e5b8e
AD
3794static inline int __dev_mc_sync(struct net_device *dev,
3795 int (*sync)(struct net_device *,
3796 const unsigned char *),
3797 int (*unsync)(struct net_device *,
3798 const unsigned char *))
3799{
3800 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3801}
3802
3803/**
e793c0f7 3804 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3805 * @dev: device to sync
3806 * @unsync: function to call if address should be removed
3807 *
3808 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3809 */
670e5b8e
AD
3810static inline void __dev_mc_unsync(struct net_device *dev,
3811 int (*unsync)(struct net_device *,
3812 const unsigned char *))
3813{
3814 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3815}
3816
4417da66 3817/* Functions used for secondary unicast and multicast support */
f629d208
JP
3818void dev_set_rx_mode(struct net_device *dev);
3819void __dev_set_rx_mode(struct net_device *dev);
3820int dev_set_promiscuity(struct net_device *dev, int inc);
3821int dev_set_allmulti(struct net_device *dev, int inc);
3822void netdev_state_change(struct net_device *dev);
3823void netdev_notify_peers(struct net_device *dev);
3824void netdev_features_change(struct net_device *dev);
1da177e4 3825/* Load a device via the kmod */
f629d208
JP
3826void dev_load(struct net *net, const char *name);
3827struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3828 struct rtnl_link_stats64 *storage);
3829void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3830 const struct net_device_stats *netdev_stats);
eeda3fd6 3831
1da177e4 3832extern int netdev_max_backlog;
3b098e2d 3833extern int netdev_tstamp_prequeue;
1da177e4 3834extern int weight_p;
3d48b53f
MT
3835extern int dev_weight_rx_bias;
3836extern int dev_weight_tx_bias;
3837extern int dev_rx_weight;
3838extern int dev_tx_weight;
9ff162a8 3839
f629d208 3840bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3841struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3842 struct list_head **iter);
f629d208
JP
3843struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3844 struct list_head **iter);
8b5be856 3845
44a40855
VY
3846/* iterate through upper list, must be called under RCU read lock */
3847#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3848 for (iter = &(dev)->adj_list.upper, \
3849 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3850 updev; \
3851 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3852
1a3f060c
DA
3853int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3854 int (*fn)(struct net_device *upper_dev,
3855 void *data),
3856 void *data);
3857
3858bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3859 struct net_device *upper_dev);
3860
f629d208
JP
3861void *netdev_lower_get_next_private(struct net_device *dev,
3862 struct list_head **iter);
3863void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3864 struct list_head **iter);
31088a11
VF
3865
3866#define netdev_for_each_lower_private(dev, priv, iter) \
3867 for (iter = (dev)->adj_list.lower.next, \
3868 priv = netdev_lower_get_next_private(dev, &(iter)); \
3869 priv; \
3870 priv = netdev_lower_get_next_private(dev, &(iter)))
3871
3872#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3873 for (iter = &(dev)->adj_list.lower, \
3874 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3875 priv; \
3876 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3877
4085ebe8
VY
3878void *netdev_lower_get_next(struct net_device *dev,
3879 struct list_head **iter);
7ce856aa 3880
4085ebe8 3881#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3882 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3883 ldev = netdev_lower_get_next(dev, &(iter)); \
3884 ldev; \
3885 ldev = netdev_lower_get_next(dev, &(iter)))
3886
7ce856aa
JP
3887struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3888 struct list_head **iter);
3889struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3890 struct list_head **iter);
3891
1a3f060c
DA
3892int netdev_walk_all_lower_dev(struct net_device *dev,
3893 int (*fn)(struct net_device *lower_dev,
3894 void *data),
3895 void *data);
3896int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3897 int (*fn)(struct net_device *lower_dev,
3898 void *data),
3899 void *data);
3900
f629d208 3901void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3902void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3903struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3904struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3905int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3906int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3907 struct net_device *upper_dev,
29bf24af 3908 void *upper_priv, void *upper_info);
f629d208
JP
3909void netdev_upper_dev_unlink(struct net_device *dev,
3910 struct net_device *upper_dev);
5bb025fa 3911void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3912void *netdev_lower_dev_get_private(struct net_device *dev,
3913 struct net_device *lower_dev);
04d48266
JP
3914void netdev_lower_state_changed(struct net_device *lower_dev,
3915 void *lower_state_info);
960fb622
ED
3916
3917/* RSS keys are 40 or 52 bytes long */
3918#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3919extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3920void netdev_rss_key_fill(void *buffer, size_t len);
3921
952fcfd0 3922int dev_get_nest_level(struct net_device *dev);
f629d208
JP
3923int skb_checksum_help(struct sk_buff *skb);
3924struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3925 netdev_features_t features, bool tx_path);
3926struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3927 netdev_features_t features);
12b0004d 3928
61bd3857
MS
3929struct netdev_bonding_info {
3930 ifslave slave;
3931 ifbond master;
3932};
3933
3934struct netdev_notifier_bonding_info {
3935 struct netdev_notifier_info info; /* must be first */
3936 struct netdev_bonding_info bonding_info;
3937};
3938
3939void netdev_bonding_info_change(struct net_device *dev,
3940 struct netdev_bonding_info *bonding_info);
3941
12b0004d
CW
3942static inline
3943struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3944{
3945 return __skb_gso_segment(skb, features, true);
3946}
53d6471c 3947__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3948
3949static inline bool can_checksum_protocol(netdev_features_t features,
3950 __be16 protocol)
3951{
c8cd0989
TH
3952 if (protocol == htons(ETH_P_FCOE))
3953 return !!(features & NETIF_F_FCOE_CRC);
3954
3955 /* Assume this is an IP checksum (not SCTP CRC) */
3956
3957 if (features & NETIF_F_HW_CSUM) {
3958 /* Can checksum everything */
3959 return true;
3960 }
3961
3962 switch (protocol) {
3963 case htons(ETH_P_IP):
3964 return !!(features & NETIF_F_IP_CSUM);
3965 case htons(ETH_P_IPV6):
3966 return !!(features & NETIF_F_IPV6_CSUM);
3967 default:
3968 return false;
3969 }
ec5f0615 3970}
12b0004d 3971
fb286bb2 3972#ifdef CONFIG_BUG
f629d208 3973void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3974#else
3975static inline void netdev_rx_csum_fault(struct net_device *dev)
3976{
3977}
3978#endif
1da177e4 3979/* rx skb timestamps */
f629d208
JP
3980void net_enable_timestamp(void);
3981void net_disable_timestamp(void);
1da177e4 3982
20380731 3983#ifdef CONFIG_PROC_FS
f629d208 3984int __init dev_proc_init(void);
900ff8c6
CW
3985#else
3986#define dev_proc_init() 0
20380731
ACM
3987#endif
3988
4798248e 3989static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3990 struct sk_buff *skb, struct net_device *dev,
3991 bool more)
4798248e 3992{
fa2dbdc2 3993 skb->xmit_more = more ? 1 : 0;
0b725a2c 3994 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3995}
3996
10b3ad8c 3997static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3998 struct netdev_queue *txq, bool more)
4798248e
DM
3999{
4000 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4001 int rc;
4798248e 4002
fa2dbdc2 4003 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4004 if (rc == NETDEV_TX_OK)
4005 txq_trans_update(txq);
4006
4007 return rc;
4798248e
DM
4008}
4009
42a2d923
LT
4010int netdev_class_create_file_ns(struct class_attribute *class_attr,
4011 const void *ns);
4012void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4013 const void *ns);
58292cbe
TH
4014
4015static inline int netdev_class_create_file(struct class_attribute *class_attr)
4016{
4017 return netdev_class_create_file_ns(class_attr, NULL);
4018}
4019
4020static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4021{
4022 netdev_class_remove_file_ns(class_attr, NULL);
4023}
b8a9787e 4024
04600794
JB
4025extern struct kobj_ns_type_operations net_ns_type_operations;
4026
f629d208 4027const char *netdev_drivername(const struct net_device *dev);
6579e57b 4028
f629d208 4029void linkwatch_run_queue(void);
20380731 4030
da08143b
MK
4031static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4032 netdev_features_t f2)
4033{
c8cd0989
TH
4034 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4035 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4036 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4037 else
b6a0e72a 4038 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4039 }
da08143b 4040
c8cd0989 4041 return f1 & f2;
da08143b
MK
4042}
4043
c8f44aff
MM
4044static inline netdev_features_t netdev_get_wanted_features(
4045 struct net_device *dev)
5455c699
MM
4046{
4047 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4048}
c8f44aff
MM
4049netdev_features_t netdev_increment_features(netdev_features_t all,
4050 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4051
4052/* Allow TSO being used on stacked device :
4053 * Performing the GSO segmentation before last device
4054 * is a performance improvement.
4055 */
4056static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4057 netdev_features_t mask)
4058{
4059 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4060}
4061
6cb6a27c 4062int __netdev_update_features(struct net_device *dev);
5455c699 4063void netdev_update_features(struct net_device *dev);
afe12cc8 4064void netdev_change_features(struct net_device *dev);
7f353bf2 4065
fc4a7489
PM
4066void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4067 struct net_device *dev);
4068
e38f3025
TM
4069netdev_features_t passthru_features_check(struct sk_buff *skb,
4070 struct net_device *dev,
4071 netdev_features_t features);
c1e756bf 4072netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4073
4d29515f 4074static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4075{
7b748340 4076 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4077
4078 /* check flags correspondence */
4079 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4080 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4081 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4082 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4083 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4084 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4085 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4086 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4087 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4088 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4089 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4090 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4091 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4092 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4093 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4094 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4095 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0345e186 4096
d6b4991a 4097 return (features & feature) == feature;
576a30eb
HX
4098}
4099
4d29515f 4100static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4101{
278b2513 4102 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4103 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4104}
4105
8b86a61d 4106static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4107 netdev_features_t features)
7967168c 4108{
fc741216 4109 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4110 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4111 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4112}
4113
82cc1a7a
PWJ
4114static inline void netif_set_gso_max_size(struct net_device *dev,
4115 unsigned int size)
4116{
4117 dev->gso_max_size = size;
4118}
4119
7a7ffbab
WCC
4120static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4121 int pulled_hlen, u16 mac_offset,
4122 int mac_len)
4123{
4124 skb->protocol = protocol;
4125 skb->encapsulation = 1;
4126 skb_push(skb, pulled_hlen);
4127 skb_reset_transport_header(skb);
4128 skb->mac_header = mac_offset;
4129 skb->network_header = skb->mac_header + mac_len;
4130 skb->mac_len = mac_len;
4131}
4132
3c175784
SD
4133static inline bool netif_is_macsec(const struct net_device *dev)
4134{
4135 return dev->priv_flags & IFF_MACSEC;
4136}
4137
b618aaa9 4138static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4139{
4140 return dev->priv_flags & IFF_MACVLAN;
4141}
4142
b618aaa9 4143static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4144{
4145 return dev->priv_flags & IFF_MACVLAN_PORT;
4146}
4147
b618aaa9 4148static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4149{
4150 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4151}
4152
b618aaa9 4153static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4154{
4155 return dev->priv_flags & IFF_IPVLAN_MASTER;
4156}
4157
b618aaa9 4158static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4159{
4160 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4161}
4162
b618aaa9 4163static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4164{
4165 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4166}
4167
3bdc0eba
BG
4168static inline bool netif_supports_nofcs(struct net_device *dev)
4169{
4170 return dev->priv_flags & IFF_SUPP_NOFCS;
4171}
4172
007979ea 4173static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4174{
007979ea 4175 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4176}
4177
fee6d4c7
DA
4178static inline bool netif_is_l3_slave(const struct net_device *dev)
4179{
4180 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4181}
4182
0894ae3f
JP
4183static inline bool netif_is_bridge_master(const struct net_device *dev)
4184{
4185 return dev->priv_flags & IFF_EBRIDGE;
4186}
4187
28f9ee22
VY
4188static inline bool netif_is_bridge_port(const struct net_device *dev)
4189{
4190 return dev->priv_flags & IFF_BRIDGE_PORT;
4191}
4192
35d4e172
JP
4193static inline bool netif_is_ovs_master(const struct net_device *dev)
4194{
4195 return dev->priv_flags & IFF_OPENVSWITCH;
4196}
4197
5be66141
JP
4198static inline bool netif_is_ovs_port(const struct net_device *dev)
4199{
4200 return dev->priv_flags & IFF_OVS_DATAPATH;
4201}
4202
b618aaa9 4203static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4204{
4205 return dev->priv_flags & IFF_TEAM;
4206}
4207
b618aaa9 4208static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4209{
4210 return dev->priv_flags & IFF_TEAM_PORT;
4211}
4212
b618aaa9 4213static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4214{
4215 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4216}
4217
b618aaa9 4218static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4219{
4220 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4221}
4222
d4ab4286
KJ
4223static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4224{
4225 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4226}
4227
02875878
ED
4228/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4229static inline void netif_keep_dst(struct net_device *dev)
4230{
4231 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4232}
4233
18d3df3e
PA
4234/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4235static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4236{
4237 /* TODO: reserve and use an additional IFF bit, if we get more users */
4238 return dev->priv_flags & IFF_MACSEC;
4239}
4240
505d4f73 4241extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4242
571ba423
JP
4243/* Logging, debugging and troubleshooting/diagnostic helpers. */
4244
4245/* netdev_printk helpers, similar to dev_printk */
4246
4247static inline const char *netdev_name(const struct net_device *dev)
4248{
c6f854d5
VF
4249 if (!dev->name[0] || strchr(dev->name, '%'))
4250 return "(unnamed net_device)";
571ba423
JP
4251 return dev->name;
4252}
4253
ccc7f496
VF
4254static inline const char *netdev_reg_state(const struct net_device *dev)
4255{
4256 switch (dev->reg_state) {
4257 case NETREG_UNINITIALIZED: return " (uninitialized)";
4258 case NETREG_REGISTERED: return "";
4259 case NETREG_UNREGISTERING: return " (unregistering)";
4260 case NETREG_UNREGISTERED: return " (unregistered)";
4261 case NETREG_RELEASED: return " (released)";
4262 case NETREG_DUMMY: return " (dummy)";
4263 }
4264
4265 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4266 return " (unknown)";
4267}
4268
f629d208 4269__printf(3, 4)
6ea754eb
JP
4270void netdev_printk(const char *level, const struct net_device *dev,
4271 const char *format, ...);
f629d208 4272__printf(2, 3)
6ea754eb 4273void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4274__printf(2, 3)
6ea754eb 4275void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4276__printf(2, 3)
6ea754eb 4277void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4278__printf(2, 3)
6ea754eb 4279void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4280__printf(2, 3)
6ea754eb 4281void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4282__printf(2, 3)
6ea754eb 4283void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4284__printf(2, 3)
6ea754eb 4285void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4286
8909c9ad
VK
4287#define MODULE_ALIAS_NETDEV(device) \
4288 MODULE_ALIAS("netdev-" device)
4289
b558c96f 4290#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4291#define netdev_dbg(__dev, format, args...) \
4292do { \
ffa10cb4 4293 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4294} while (0)
b558c96f
JC
4295#elif defined(DEBUG)
4296#define netdev_dbg(__dev, format, args...) \
4297 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4298#else
4299#define netdev_dbg(__dev, format, args...) \
4300({ \
4301 if (0) \
4302 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4303})
4304#endif
4305
4306#if defined(VERBOSE_DEBUG)
4307#define netdev_vdbg netdev_dbg
4308#else
4309
4310#define netdev_vdbg(dev, format, args...) \
4311({ \
4312 if (0) \
4313 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4314 0; \
4315})
4316#endif
4317
4318/*
4319 * netdev_WARN() acts like dev_printk(), but with the key difference
4320 * of using a WARN/WARN_ON to get the message out, including the
4321 * file/line information and a backtrace.
4322 */
4323#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4324 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4325 netdev_reg_state(dev), ##args)
571ba423 4326
b3d95c5c
JP
4327/* netif printk helpers, similar to netdev_printk */
4328
4329#define netif_printk(priv, type, level, dev, fmt, args...) \
4330do { \
4331 if (netif_msg_##type(priv)) \
4332 netdev_printk(level, (dev), fmt, ##args); \
4333} while (0)
4334
f45f4321
JP
4335#define netif_level(level, priv, type, dev, fmt, args...) \
4336do { \
4337 if (netif_msg_##type(priv)) \
4338 netdev_##level(dev, fmt, ##args); \
4339} while (0)
4340
b3d95c5c 4341#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4342 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4343#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4344 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4345#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4346 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4347#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4348 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4349#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4350 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4351#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4352 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4353#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4354 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4355
0053ea9c 4356#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4357#define netif_dbg(priv, type, netdev, format, args...) \
4358do { \
4359 if (netif_msg_##type(priv)) \
b5fb0a03 4360 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4361} while (0)
0053ea9c
JP
4362#elif defined(DEBUG)
4363#define netif_dbg(priv, type, dev, format, args...) \
4364 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4365#else
4366#define netif_dbg(priv, type, dev, format, args...) \
4367({ \
4368 if (0) \
4369 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4370 0; \
4371})
4372#endif
4373
f617f276
EC
4374/* if @cond then downgrade to debug, else print at @level */
4375#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4376 do { \
4377 if (cond) \
4378 netif_dbg(priv, type, netdev, fmt, ##args); \
4379 else \
4380 netif_ ## level(priv, type, netdev, fmt, ##args); \
4381 } while (0)
4382
b3d95c5c 4383#if defined(VERBOSE_DEBUG)
bcfcc450 4384#define netif_vdbg netif_dbg
b3d95c5c
JP
4385#else
4386#define netif_vdbg(priv, type, dev, format, args...) \
4387({ \
4388 if (0) \
a4ed89cb 4389 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4390 0; \
4391})
4392#endif
571ba423 4393
900ff8c6
CW
4394/*
4395 * The list of packet types we will receive (as opposed to discard)
4396 * and the routines to invoke.
4397 *
4398 * Why 16. Because with 16 the only overlap we get on a hash of the
4399 * low nibble of the protocol value is RARP/SNAP/X.25.
4400 *
4401 * NOTE: That is no longer true with the addition of VLAN tags. Not
4402 * sure which should go first, but I bet it won't make much
4403 * difference if we are running VLANs. The good news is that
4404 * this protocol won't be in the list unless compiled in, so
4405 * the average user (w/out VLANs) will not be adversely affected.
4406 * --BLG
4407 *
4408 * 0800 IP
4409 * 8100 802.1Q VLAN
4410 * 0001 802.3
4411 * 0002 AX.25
4412 * 0004 802.2
4413 * 8035 RARP
4414 * 0005 SNAP
4415 * 0805 X.25
4416 * 0806 ARP
4417 * 8137 IPX
4418 * 0009 Localtalk
4419 * 86DD IPv6
4420 */
4421#define PTYPE_HASH_SIZE (16)
4422#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4423
385a154c 4424#endif /* _LINUX_NETDEVICE_H */