]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/netdevice.h
net: check UDP tunnel RX port offload feature before calling tunnel ndo ndo
[mirror_ubuntu-bionic-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
7a6b6f51 44#ifdef CONFIG_DCB
2f90b865
AD
45#include <net/dcbnl.h>
46#endif
5bc1421e 47#include <net/netprio_cgroup.h>
a050c33f 48
a59e2ecb 49#include <linux/netdev_features.h>
77162022 50#include <linux/neighbour.h>
607ca46e 51#include <uapi/linux/netdevice.h>
61bd3857 52#include <uapi/linux/if_bonding.h>
e4c6734e 53#include <uapi/linux/pkt_cls.h>
59cc1f61 54#include <linux/hashtable.h>
a59e2ecb 55
115c1d6e 56struct netpoll_info;
313162d0 57struct device;
c1f19b51 58struct phy_device;
c6e970a0
AL
59struct dsa_switch_tree;
60
704232c2
JB
61/* 802.11 specific */
62struct wireless_dev;
98a18b6f
AA
63/* 802.15.4 specific */
64struct wpan_dev;
03c57747 65struct mpls_dev;
7c46a640
AD
66/* UDP Tunnel offloads */
67struct udp_tunnel_info;
a7862b45 68struct bpf_prog;
814abfab 69struct xdp_buff;
1da177e4 70
f629d208
JP
71void netdev_set_default_ethtool_ops(struct net_device *dev,
72 const struct ethtool_ops *ops);
d07d7507 73
9a1654ba
JP
74/* Backlog congestion levels */
75#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
76#define NET_RX_DROP 1 /* packet dropped */
77
572a9d7b
PM
78/*
79 * Transmit return codes: transmit return codes originate from three different
80 * namespaces:
81 *
82 * - qdisc return codes
83 * - driver transmit return codes
84 * - errno values
85 *
86 * Drivers are allowed to return any one of those in their hard_start_xmit()
87 * function. Real network devices commonly used with qdiscs should only return
88 * the driver transmit return codes though - when qdiscs are used, the actual
89 * transmission happens asynchronously, so the value is not propagated to
5e82b4b2
BH
90 * higher layers. Virtual network devices transmit synchronously; in this case
91 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
572a9d7b
PM
92 * others are propagated to higher layers.
93 */
94
95/* qdisc ->enqueue() return codes. */
96#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
97#define NET_XMIT_DROP 0x01 /* skb dropped */
98#define NET_XMIT_CN 0x02 /* congestion notification */
9a1654ba 99#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 100
b9df3cb8
GR
101/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 104#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
105#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
dc1f8bf6 107/* Driver transmit return codes */
9a1654ba 108#define NETDEV_TX_MASK 0xf0
572a9d7b 109
dc1f8bf6 110enum netdev_tx {
572a9d7b 111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
dc1f8bf6
SH
114};
115typedef enum netdev_tx netdev_tx_t;
116
9a1654ba
JP
117/*
118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 */
121static inline bool dev_xmit_complete(int rc)
122{
123 /*
124 * Positive cases with an skb consumed by a driver:
125 * - successful transmission (rc == NETDEV_TX_OK)
126 * - error while transmitting (rc < 0)
127 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 */
129 if (likely(rc < NET_XMIT_MASK))
130 return true;
131
132 return false;
133}
134
1da177e4 135/*
5e82b4b2 136 * Compute the worst-case header length according to the protocols
1da177e4
LT
137 * used.
138 */
fe2918b0 139
c0eb4540
KS
140#if defined(CONFIG_HYPERV_NET)
141# define LL_MAX_HEADER 128
142#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
143# if defined(CONFIG_MAC80211_MESH)
144# define LL_MAX_HEADER 128
145# else
146# define LL_MAX_HEADER 96
147# endif
1da177e4 148#else
8388e3da 149# define LL_MAX_HEADER 32
1da177e4
LT
150#endif
151
d11ead75
BH
152#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
153 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
154#define MAX_HEADER LL_MAX_HEADER
155#else
156#define MAX_HEADER (LL_MAX_HEADER + 48)
157#endif
158
159/*
be1f3c2c
BH
160 * Old network device statistics. Fields are native words
161 * (unsigned long) so they can be read and written atomically.
1da177e4 162 */
fe2918b0 163
d94d9fee 164struct net_device_stats {
3cfde79c
BH
165 unsigned long rx_packets;
166 unsigned long tx_packets;
167 unsigned long rx_bytes;
168 unsigned long tx_bytes;
169 unsigned long rx_errors;
170 unsigned long tx_errors;
171 unsigned long rx_dropped;
172 unsigned long tx_dropped;
173 unsigned long multicast;
1da177e4 174 unsigned long collisions;
1da177e4 175 unsigned long rx_length_errors;
3cfde79c
BH
176 unsigned long rx_over_errors;
177 unsigned long rx_crc_errors;
178 unsigned long rx_frame_errors;
179 unsigned long rx_fifo_errors;
180 unsigned long rx_missed_errors;
1da177e4
LT
181 unsigned long tx_aborted_errors;
182 unsigned long tx_carrier_errors;
183 unsigned long tx_fifo_errors;
184 unsigned long tx_heartbeat_errors;
185 unsigned long tx_window_errors;
1da177e4
LT
186 unsigned long rx_compressed;
187 unsigned long tx_compressed;
188};
189
1da177e4
LT
190
191#include <linux/cache.h>
192#include <linux/skbuff.h>
193
adc9300e 194#ifdef CONFIG_RPS
c5905afb
IM
195#include <linux/static_key.h>
196extern struct static_key rps_needed;
13bfff25 197extern struct static_key rfs_needed;
adc9300e
ED
198#endif
199
1da177e4
LT
200struct neighbour;
201struct neigh_parms;
202struct sk_buff;
203
f001fde5
JP
204struct netdev_hw_addr {
205 struct list_head list;
206 unsigned char addr[MAX_ADDR_LEN];
207 unsigned char type;
ccffad25
JP
208#define NETDEV_HW_ADDR_T_LAN 1
209#define NETDEV_HW_ADDR_T_SAN 2
210#define NETDEV_HW_ADDR_T_SLAVE 3
211#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 212#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 213 bool global_use;
4cd729b0 214 int sync_cnt;
8f8f103d 215 int refcount;
4543fbef 216 int synced;
f001fde5
JP
217 struct rcu_head rcu_head;
218};
219
31278e71
JP
220struct netdev_hw_addr_list {
221 struct list_head list;
222 int count;
223};
224
22bedad3
JP
225#define netdev_hw_addr_list_count(l) ((l)->count)
226#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
227#define netdev_hw_addr_list_for_each(ha, l) \
228 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 229
22bedad3
JP
230#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
231#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
232#define netdev_for_each_uc_addr(ha, dev) \
233 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 234
22bedad3
JP
235#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
236#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 237#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 238 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 239
d94d9fee 240struct hh_cache {
5b3dc2f3 241 unsigned int hh_len;
3644f0ce 242 seqlock_t hh_lock;
1da177e4
LT
243
244 /* cached hardware header; allow for machine alignment needs. */
245#define HH_DATA_MOD 16
246#define HH_DATA_OFF(__len) \
5ba0eac6 247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
248#define HH_DATA_ALIGN(__len) \
249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251};
252
5e82b4b2 253/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
1da177e4
LT
254 * Alternative is:
255 * dev->hard_header_len ? (dev->hard_header_len +
256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 *
258 * We could use other alignment values, but we must maintain the
259 * relationship HH alignment <= LL alignment.
260 */
261#define LL_RESERVED_SPACE(dev) \
f5184d26 262 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 263#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 264 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 265
3b04ddde
SH
266struct header_ops {
267 int (*create) (struct sk_buff *skb, struct net_device *dev,
268 unsigned short type, const void *daddr,
95c96174 269 const void *saddr, unsigned int len);
3b04ddde 270 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
2793a23a 275 bool (*validate)(const char *ll_header, unsigned int len);
3b04ddde
SH
276};
277
1da177e4 278/* These flag bits are private to the generic network queueing
5e82b4b2 279 * layer; they may not be explicitly referenced by any other
1da177e4
LT
280 * code.
281 */
282
d94d9fee 283enum netdev_state_t {
1da177e4
LT
284 __LINK_STATE_START,
285 __LINK_STATE_PRESENT,
1da177e4 286 __LINK_STATE_NOCARRIER,
b00055aa
SR
287 __LINK_STATE_LINKWATCH_PENDING,
288 __LINK_STATE_DORMANT,
1da177e4
LT
289};
290
291
292/*
5e82b4b2 293 * This structure holds boot-time configured netdevice settings. They
fe2918b0 294 * are then used in the device probing.
1da177e4
LT
295 */
296struct netdev_boot_setup {
297 char name[IFNAMSIZ];
298 struct ifmap map;
299};
300#define NETDEV_BOOT_SETUP_MAX 8
301
f629d208 302int __init netdev_boot_setup(char *str);
1da177e4 303
bea3348e
SH
304/*
305 * Structure for NAPI scheduling similar to tasklet but with weighting
306 */
307struct napi_struct {
308 /* The poll_list must only be managed by the entity which
309 * changes the state of the NAPI_STATE_SCHED bit. This means
310 * whoever atomically sets that bit can add this napi_struct
5e82b4b2 311 * to the per-CPU poll_list, and whoever clears that bit
bea3348e
SH
312 * can remove from the list right before clearing the bit.
313 */
314 struct list_head poll_list;
315
316 unsigned long state;
317 int weight;
404f7c9e 318 unsigned int gro_count;
bea3348e
SH
319 int (*poll)(struct napi_struct *, int);
320#ifdef CONFIG_NETPOLL
bea3348e 321 int poll_owner;
bea3348e 322#endif
5d38a079 323 struct net_device *dev;
d565b0a1 324 struct sk_buff *gro_list;
5d38a079 325 struct sk_buff *skb;
3b47d303 326 struct hrtimer timer;
404f7c9e 327 struct list_head dev_list;
af12fa6e
ET
328 struct hlist_node napi_hash_node;
329 unsigned int napi_id;
bea3348e
SH
330};
331
d94d9fee 332enum {
bea3348e 333 NAPI_STATE_SCHED, /* Poll is scheduled */
39e6c820 334 NAPI_STATE_MISSED, /* reschedule a napi */
a0a46196 335 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 336 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
d64b5e85
ED
337 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
338 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
217f6974
ED
339 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
340};
341
342enum {
39e6c820
ED
343 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
344 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
345 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
346 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
347 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
348 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
349 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
bea3348e
SH
350};
351
5b252f0c 352enum gro_result {
d1c76af9
HX
353 GRO_MERGED,
354 GRO_MERGED_FREE,
355 GRO_HELD,
356 GRO_NORMAL,
357 GRO_DROP,
25393d3f 358 GRO_CONSUMED,
d1c76af9 359};
5b252f0c 360typedef enum gro_result gro_result_t;
d1c76af9 361
8a4eb573
JP
362/*
363 * enum rx_handler_result - Possible return values for rx_handlers.
364 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365 * further.
366 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
367 * case skb->dev was changed by rx_handler.
368 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
5e82b4b2 369 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
8a4eb573
JP
370 *
371 * rx_handlers are functions called from inside __netif_receive_skb(), to do
372 * special processing of the skb, prior to delivery to protocol handlers.
373 *
374 * Currently, a net_device can only have a single rx_handler registered. Trying
375 * to register a second rx_handler will return -EBUSY.
376 *
377 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
378 * To unregister a rx_handler on a net_device, use
379 * netdev_rx_handler_unregister().
380 *
381 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
382 * do with the skb.
383 *
5e82b4b2 384 * If the rx_handler consumed the skb in some way, it should return
8a4eb573 385 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
5e82b4b2 386 * the skb to be delivered in some other way.
8a4eb573
JP
387 *
388 * If the rx_handler changed skb->dev, to divert the skb to another
389 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
390 * new device will be called if it exists.
391 *
5e82b4b2 392 * If the rx_handler decides the skb should be ignored, it should return
8a4eb573 393 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 394 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573 395 *
5e82b4b2 396 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
8a4eb573
JP
397 * delivered, it should return RX_HANDLER_PASS.
398 *
399 * A device without a registered rx_handler will behave as if rx_handler
400 * returned RX_HANDLER_PASS.
401 */
402
403enum rx_handler_result {
404 RX_HANDLER_CONSUMED,
405 RX_HANDLER_ANOTHER,
406 RX_HANDLER_EXACT,
407 RX_HANDLER_PASS,
408};
409typedef enum rx_handler_result rx_handler_result_t;
410typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 411
f629d208 412void __napi_schedule(struct napi_struct *n);
bc9ad166 413void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 414
4d29515f 415static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
416{
417 return test_bit(NAPI_STATE_DISABLE, &n->state);
418}
419
39e6c820 420bool napi_schedule_prep(struct napi_struct *n);
bea3348e
SH
421
422/**
423 * napi_schedule - schedule NAPI poll
5e82b4b2 424 * @n: NAPI context
bea3348e
SH
425 *
426 * Schedule NAPI poll routine to be called if it is not already
427 * running.
428 */
429static inline void napi_schedule(struct napi_struct *n)
430{
431 if (napi_schedule_prep(n))
432 __napi_schedule(n);
433}
434
bc9ad166
ED
435/**
436 * napi_schedule_irqoff - schedule NAPI poll
5e82b4b2 437 * @n: NAPI context
bc9ad166
ED
438 *
439 * Variant of napi_schedule(), assuming hard irqs are masked.
440 */
441static inline void napi_schedule_irqoff(struct napi_struct *n)
442{
443 if (napi_schedule_prep(n))
444 __napi_schedule_irqoff(n);
445}
446
bfe13f54 447/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 448static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
449{
450 if (napi_schedule_prep(napi)) {
451 __napi_schedule(napi);
4d29515f 452 return true;
bfe13f54 453 }
4d29515f 454 return false;
bfe13f54
RD
455}
456
364b6055 457bool napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
458/**
459 * napi_complete - NAPI processing complete
5e82b4b2 460 * @n: NAPI context
bea3348e
SH
461 *
462 * Mark NAPI processing as complete.
3b47d303 463 * Consider using napi_complete_done() instead.
364b6055 464 * Return false if device should avoid rearming interrupts.
bea3348e 465 */
364b6055 466static inline bool napi_complete(struct napi_struct *n)
3b47d303
ED
467{
468 return napi_complete_done(n, 0);
469}
bea3348e 470
af12fa6e
ET
471/**
472 * napi_hash_del - remove a NAPI from global table
5e82b4b2 473 * @napi: NAPI context
af12fa6e 474 *
5e82b4b2 475 * Warning: caller must observe RCU grace period
34cbe27e
ED
476 * before freeing memory containing @napi, if
477 * this function returns true.
93d05d4a 478 * Note: core networking stack automatically calls it
5e82b4b2 479 * from netif_napi_del().
93d05d4a 480 * Drivers might want to call this helper to combine all
5e82b4b2 481 * the needed RCU grace periods into a single one.
af12fa6e 482 */
34cbe27e 483bool napi_hash_del(struct napi_struct *napi);
af12fa6e 484
bea3348e
SH
485/**
486 * napi_disable - prevent NAPI from scheduling
5e82b4b2 487 * @n: NAPI context
bea3348e
SH
488 *
489 * Stop NAPI from being scheduled on this context.
490 * Waits till any outstanding processing completes.
491 */
3b47d303 492void napi_disable(struct napi_struct *n);
bea3348e
SH
493
494/**
495 * napi_enable - enable NAPI scheduling
5e82b4b2 496 * @n: NAPI context
bea3348e
SH
497 *
498 * Resume NAPI from being scheduled on this context.
499 * Must be paired with napi_disable.
500 */
501static inline void napi_enable(struct napi_struct *n)
502{
503 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 504 smp_mb__before_atomic();
bea3348e 505 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 506 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
507}
508
c264c3de
SH
509/**
510 * napi_synchronize - wait until NAPI is not running
5e82b4b2 511 * @n: NAPI context
c264c3de
SH
512 *
513 * Wait until NAPI is done being scheduled on this context.
514 * Waits till any outstanding processing completes but
515 * does not disable future activations.
516 */
517static inline void napi_synchronize(const struct napi_struct *n)
518{
facc432f
AB
519 if (IS_ENABLED(CONFIG_SMP))
520 while (test_bit(NAPI_STATE_SCHED, &n->state))
521 msleep(1);
522 else
523 barrier();
c264c3de 524}
c264c3de 525
d94d9fee 526enum netdev_queue_state_t {
73466498
TH
527 __QUEUE_STATE_DRV_XOFF,
528 __QUEUE_STATE_STACK_XOFF,
c3f26a26 529 __QUEUE_STATE_FROZEN,
79d16385 530};
8e2f1a63
DB
531
532#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
533#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
534#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
535
536#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
537#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 QUEUE_STATE_FROZEN)
539#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
540 QUEUE_STATE_FROZEN)
541
73466498
TH
542/*
543 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
544 * netif_tx_* functions below are used to manipulate this flag. The
545 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
546 * queue independently. The netif_xmit_*stopped functions below are called
547 * to check if the queue has been stopped by the driver or stack (either
548 * of the XOFF bits are set in the state). Drivers should not need to call
549 * netif_xmit*stopped functions, they should only be using netif_tx_*.
550 */
79d16385 551
bb949fbd 552struct netdev_queue {
6a321cb3 553/*
5e82b4b2 554 * read-mostly part
6a321cb3 555 */
bb949fbd 556 struct net_device *dev;
46e5da40 557 struct Qdisc __rcu *qdisc;
b0e1e646 558 struct Qdisc *qdisc_sleeping;
ccf5ff69 559#ifdef CONFIG_SYSFS
1d24eb48
TH
560 struct kobject kobj;
561#endif
f2cd2d3e
ED
562#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
563 int numa_node;
564#endif
c0ef079c
FW
565 unsigned long tx_maxrate;
566 /*
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
569 */
570 unsigned long trans_timeout;
6a321cb3 571/*
5e82b4b2 572 * write-mostly part
6a321cb3
ED
573 */
574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
575 int xmit_lock_owner;
9d21493b 576 /*
9b36627a 577 * Time (in jiffies) of last Tx
9d21493b
ED
578 */
579 unsigned long trans_start;
ccf5ff69 580
114cf580
TH
581 unsigned long state;
582
583#ifdef CONFIG_BQL
584 struct dql dql;
585#endif
e8a0464c 586} ____cacheline_aligned_in_smp;
bb949fbd 587
f2cd2d3e
ED
588static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
589{
590#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
591 return q->numa_node;
592#else
b236da69 593 return NUMA_NO_NODE;
f2cd2d3e
ED
594#endif
595}
596
597static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 q->numa_node = node;
601#endif
602}
603
df334545 604#ifdef CONFIG_RPS
0a9627f2
TH
605/*
606 * This structure holds an RPS map which can be of variable length. The
607 * map is an array of CPUs.
608 */
609struct rps_map {
610 unsigned int len;
611 struct rcu_head rcu;
612 u16 cpus[0];
613};
60b778ce 614#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 615
fec5e652 616/*
c445477d
BH
617 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
618 * tail pointer for that CPU's input queue at the time of last enqueue, and
619 * a hardware filter index.
fec5e652
TH
620 */
621struct rps_dev_flow {
622 u16 cpu;
c445477d 623 u16 filter;
fec5e652
TH
624 unsigned int last_qtail;
625};
c445477d 626#define RPS_NO_FILTER 0xffff
fec5e652
TH
627
628/*
629 * The rps_dev_flow_table structure contains a table of flow mappings.
630 */
631struct rps_dev_flow_table {
632 unsigned int mask;
633 struct rcu_head rcu;
fec5e652
TH
634 struct rps_dev_flow flows[0];
635};
636#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 637 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
638
639/*
640 * The rps_sock_flow_table contains mappings of flows to the last CPU
641 * on which they were processed by the application (set in recvmsg).
5e82b4b2
BH
642 * Each entry is a 32bit value. Upper part is the high-order bits
643 * of flow hash, lower part is CPU number.
567e4b79 644 * rps_cpu_mask is used to partition the space, depending on number of
5e82b4b2
BH
645 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
646 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
567e4b79 647 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
648 */
649struct rps_sock_flow_table {
567e4b79 650 u32 mask;
93c1af6c
ED
651
652 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 653};
567e4b79 654#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
655
656#define RPS_NO_CPU 0xffff
657
567e4b79
ED
658extern u32 rps_cpu_mask;
659extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
660
fec5e652
TH
661static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
662 u32 hash)
663{
664 if (table && hash) {
567e4b79
ED
665 unsigned int index = hash & table->mask;
666 u32 val = hash & ~rps_cpu_mask;
fec5e652 667
5e82b4b2 668 /* We only give a hint, preemption can change CPU under us */
567e4b79 669 val |= raw_smp_processor_id();
fec5e652 670
567e4b79
ED
671 if (table->ents[index] != val)
672 table->ents[index] = val;
fec5e652
TH
673 }
674}
675
c445477d 676#ifdef CONFIG_RFS_ACCEL
f629d208
JP
677bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
678 u16 filter_id);
c445477d 679#endif
a953be53 680#endif /* CONFIG_RPS */
c445477d 681
0a9627f2
TH
682/* This structure contains an instance of an RX queue. */
683struct netdev_rx_queue {
a953be53 684#ifdef CONFIG_RPS
6e3f7faf
ED
685 struct rps_map __rcu *rps_map;
686 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 687#endif
6e3f7faf 688 struct kobject kobj;
fe822240 689 struct net_device *dev;
0a9627f2 690} ____cacheline_aligned_in_smp;
a953be53
MD
691
692/*
693 * RX queue sysfs structures and functions.
694 */
695struct rx_queue_attribute {
696 struct attribute attr;
697 ssize_t (*show)(struct netdev_rx_queue *queue,
698 struct rx_queue_attribute *attr, char *buf);
699 ssize_t (*store)(struct netdev_rx_queue *queue,
700 struct rx_queue_attribute *attr, const char *buf, size_t len);
701};
d314774c 702
bf264145
TH
703#ifdef CONFIG_XPS
704/*
705 * This structure holds an XPS map which can be of variable length. The
706 * map is an array of queues.
707 */
708struct xps_map {
709 unsigned int len;
710 unsigned int alloc_len;
711 struct rcu_head rcu;
712 u16 queues[0];
713};
60b778ce 714#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
715#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
716 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
717
718/*
719 * This structure holds all XPS maps for device. Maps are indexed by CPU.
720 */
721struct xps_dev_maps {
722 struct rcu_head rcu;
a4177869 723 struct xps_map __rcu *cpu_map[0];
bf264145 724};
184c449f
AD
725#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
726 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
bf264145
TH
727#endif /* CONFIG_XPS */
728
4f57c087
JF
729#define TC_MAX_QUEUE 16
730#define TC_BITMASK 15
731/* HW offloaded queuing disciplines txq count and offset maps */
732struct netdev_tc_txq {
733 u16 count;
734 u16 offset;
735};
736
68bad94e
NP
737#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
738/*
739 * This structure is to hold information about the device
740 * configured to run FCoE protocol stack.
741 */
742struct netdev_fcoe_hbainfo {
743 char manufacturer[64];
744 char serial_number[64];
745 char hardware_version[64];
746 char driver_version[64];
747 char optionrom_version[64];
748 char firmware_version[64];
749 char model[256];
750 char model_description[256];
751};
752#endif
753
02637fce 754#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 755
02637fce
JP
756/* This structure holds a unique identifier to identify some
757 * physical item (port for example) used by a netdevice.
66b52b0d 758 */
02637fce
JP
759struct netdev_phys_item_id {
760 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
761 unsigned char id_len;
762};
763
d754f98b
SF
764static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
765 struct netdev_phys_item_id *b)
766{
767 return a->id_len == b->id_len &&
768 memcmp(a->id, b->id, a->id_len) == 0;
769}
770
99932d4f
DB
771typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
772 struct sk_buff *skb);
773
a1b7c5fd 774/* These structures hold the attributes of qdisc and classifiers
16e5cc64
JF
775 * that are being passed to the netdevice through the setup_tc op.
776 */
777enum {
778 TC_SETUP_MQPRIO,
a1b7c5fd 779 TC_SETUP_CLSU32,
5b33f488 780 TC_SETUP_CLSFLOWER,
b87f7936 781 TC_SETUP_MATCHALL,
332ae8e2 782 TC_SETUP_CLSBPF,
16e5cc64
JF
783};
784
a1b7c5fd
JF
785struct tc_cls_u32_offload;
786
16e5cc64
JF
787struct tc_to_netdev {
788 unsigned int type;
789 union {
a1b7c5fd 790 struct tc_cls_u32_offload *cls_u32;
5b33f488 791 struct tc_cls_flower_offload *cls_flower;
b87f7936 792 struct tc_cls_matchall_offload *cls_mall;
332ae8e2 793 struct tc_cls_bpf_offload *cls_bpf;
56f36acd 794 struct tc_mqprio_qopt *mqprio;
16e5cc64 795 };
7091d8c7 796 bool egress_dev;
16e5cc64
JF
797};
798
a7862b45
BB
799/* These structures hold the attributes of xdp state that are being passed
800 * to the netdevice through the xdp op.
801 */
802enum xdp_netdev_command {
803 /* Set or clear a bpf program used in the earliest stages of packet
804 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
805 * is responsible for calling bpf_prog_put on any old progs that are
806 * stored. In case of error, the callee need not release the new prog
807 * reference, but on success it takes ownership and must bpf_prog_put
808 * when it is no longer used.
809 */
810 XDP_SETUP_PROG,
ee5d032f 811 XDP_SETUP_PROG_HW,
a7862b45 812 /* Check if a bpf program is set on the device. The callee should
ce158e58
JK
813 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
814 * is equivalent to XDP_ATTACHED_DRV.
a7862b45
BB
815 */
816 XDP_QUERY_PROG,
817};
818
ddf9f970
JK
819struct netlink_ext_ack;
820
a7862b45
BB
821struct netdev_xdp {
822 enum xdp_netdev_command command;
823 union {
824 /* XDP_SETUP_PROG */
ddf9f970 825 struct {
32d60277 826 u32 flags;
ddf9f970
JK
827 struct bpf_prog *prog;
828 struct netlink_ext_ack *extack;
829 };
a7862b45 830 /* XDP_QUERY_PROG */
58038695 831 struct {
ce158e58 832 u8 prog_attached;
58038695
MKL
833 u32 prog_id;
834 };
a7862b45
BB
835 };
836};
16e5cc64 837
d77e38e6
SK
838#ifdef CONFIG_XFRM_OFFLOAD
839struct xfrmdev_ops {
840 int (*xdo_dev_state_add) (struct xfrm_state *x);
841 void (*xdo_dev_state_delete) (struct xfrm_state *x);
842 void (*xdo_dev_state_free) (struct xfrm_state *x);
843 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
844 struct xfrm_state *x);
845};
846#endif
847
d314774c
SH
848/*
849 * This structure defines the management hooks for network devices.
00829823
SH
850 * The following hooks can be defined; unless noted otherwise, they are
851 * optional and can be filled with a null pointer.
d314774c
SH
852 *
853 * int (*ndo_init)(struct net_device *dev);
5e82b4b2
BH
854 * This function is called once when a network device is registered.
855 * The network device can use this for any late stage initialization
856 * or semantic validation. It can fail with an error code which will
857 * be propagated back to register_netdev.
d314774c
SH
858 *
859 * void (*ndo_uninit)(struct net_device *dev);
860 * This function is called when device is unregistered or when registration
861 * fails. It is not called if init fails.
862 *
863 * int (*ndo_open)(struct net_device *dev);
5e82b4b2 864 * This function is called when a network device transitions to the up
d314774c
SH
865 * state.
866 *
867 * int (*ndo_stop)(struct net_device *dev);
5e82b4b2 868 * This function is called when a network device transitions to the down
d314774c
SH
869 * state.
870 *
dc1f8bf6
SH
871 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
872 * struct net_device *dev);
00829823 873 * Called when a packet needs to be transmitted.
e79d8429
RR
874 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
875 * the queue before that can happen; it's for obsolete devices and weird
876 * corner cases, but the stack really does a non-trivial amount
877 * of useless work if you return NETDEV_TX_BUSY.
5e82b4b2 878 * Required; cannot be NULL.
00829823 879 *
1a2a1444
DM
880 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
881 * struct net_device *dev
882 * netdev_features_t features);
883 * Called by core transmit path to determine if device is capable of
884 * performing offload operations on a given packet. This is to give
885 * the device an opportunity to implement any restrictions that cannot
886 * be otherwise expressed by feature flags. The check is called with
887 * the set of features that the stack has calculated and it returns
888 * those the driver believes to be appropriate.
cdba756f 889 *
f663dd9a 890 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 891 * void *accel_priv, select_queue_fallback_t fallback);
5e82b4b2 892 * Called to decide which queue to use when device supports multiple
00829823
SH
893 * transmit queues.
894 *
d314774c
SH
895 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
896 * This function is called to allow device receiver to make
5e82b4b2 897 * changes to configuration when multicast or promiscuous is enabled.
d314774c
SH
898 *
899 * void (*ndo_set_rx_mode)(struct net_device *dev);
900 * This function is called device changes address list filtering.
01789349 901 * If driver handles unicast address filtering, it should set
5e82b4b2 902 * IFF_UNICAST_FLT in its priv_flags.
d314774c
SH
903 *
904 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
905 * This function is called when the Media Access Control address
37b607c5 906 * needs to be changed. If this interface is not defined, the
5e82b4b2 907 * MAC address can not be changed.
d314774c
SH
908 *
909 * int (*ndo_validate_addr)(struct net_device *dev);
910 * Test if Media Access Control address is valid for the device.
911 *
912 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
5e82b4b2
BH
913 * Called when a user requests an ioctl which can't be handled by
914 * the generic interface code. If not defined ioctls return
d314774c
SH
915 * not supported error code.
916 *
917 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
918 * Used to set network devices bus interface parameters. This interface
5e82b4b2 919 * is retained for legacy reasons; new devices should use the bus
d314774c
SH
920 * interface (PCI) for low level management.
921 *
922 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
923 * Called when a user wants to change the Maximum Transfer Unit
db46a0e1 924 * of a device.
d314774c 925 *
00829823 926 * void (*ndo_tx_timeout)(struct net_device *dev);
5e82b4b2 927 * Callback used when the transmitter has not made any progress
d314774c
SH
928 * for dev->watchdog ticks.
929 *
bc1f4470 930 * void (*ndo_get_stats64)(struct net_device *dev,
931 * struct rtnl_link_stats64 *storage);
d308e38f 932 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 933 * Called when a user wants to get the network device usage
be1f3c2c 934 * statistics. Drivers must do one of the following:
3cfde79c
BH
935 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
936 * rtnl_link_stats64 structure passed by the caller.
82695d9b 937 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
938 * (which should normally be dev->stats) and return a pointer to
939 * it. The structure may be changed asynchronously only if each
940 * field is written atomically.
941 * 3. Update dev->stats asynchronously and atomically, and define
942 * neither operation.
d314774c 943 *
3df5b3c6 944 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
2c9d85d4
NF
945 * Return true if this device supports offload stats of this attr_id.
946 *
947 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
948 * void *attr_data)
949 * Get statistics for offload operations by attr_id. Write it into the
950 * attr_data pointer.
951 *
5d632cb7 952 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 953 * If device supports VLAN filtering this function is called when a
80d5c368 954 * VLAN id is registered.
d314774c 955 *
5d632cb7 956 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
5e82b4b2 957 * If device supports VLAN filtering this function is called when a
80d5c368 958 * VLAN id is unregistered.
d314774c
SH
959 *
960 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
961 *
962 * SR-IOV management functions.
963 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
79aab093
MS
964 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
965 * u8 qos, __be16 proto);
ed616689
SC
966 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
967 * int max_tx_rate);
5f8444a3 968 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 969 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
970 * int (*ndo_get_vf_config)(struct net_device *dev,
971 * int vf, struct ifla_vf_info *ivf);
1d8faf48 972 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
973 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
974 * struct nlattr *port[]);
01a3d796
VZ
975 *
976 * Enable or disable the VF ability to query its RSS Redirection Table and
977 * Hash Key. This is needed since on some devices VF share this information
5e82b4b2 978 * with PF and querying it may introduce a theoretical security risk.
01a3d796 979 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 980 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
a5fcf8a6 981 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle, u32 chain_index,
6a4bc2b4
FF
982 * __be16 protocol, struct tc_to_netdev *tc);
983 * Called to setup any 'tc' scheduler, classifier or action on @dev.
984 * This is always called from the stack with the rtnl lock held and netif
985 * tx queues stopped. This allows the netdevice to perform queue
986 * management safely.
c445477d 987 *
e9bce845
YZ
988 * Fiber Channel over Ethernet (FCoE) offload functions.
989 * int (*ndo_fcoe_enable)(struct net_device *dev);
990 * Called when the FCoE protocol stack wants to start using LLD for FCoE
991 * so the underlying device can perform whatever needed configuration or
992 * initialization to support acceleration of FCoE traffic.
993 *
994 * int (*ndo_fcoe_disable)(struct net_device *dev);
995 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
996 * so the underlying device can perform whatever needed clean-ups to
997 * stop supporting acceleration of FCoE traffic.
998 *
999 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1000 * struct scatterlist *sgl, unsigned int sgc);
1001 * Called when the FCoE Initiator wants to initialize an I/O that
1002 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1003 * perform necessary setup and returns 1 to indicate the device is set up
1004 * successfully to perform DDP on this I/O, otherwise this returns 0.
1005 *
1006 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1007 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1008 * indicated by the FC exchange id 'xid', so the underlying device can
1009 * clean up and reuse resources for later DDP requests.
1010 *
1011 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1012 * struct scatterlist *sgl, unsigned int sgc);
1013 * Called when the FCoE Target wants to initialize an I/O that
1014 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1015 * perform necessary setup and returns 1 to indicate the device is set up
1016 * successfully to perform DDP on this I/O, otherwise this returns 0.
1017 *
68bad94e
NP
1018 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1019 * struct netdev_fcoe_hbainfo *hbainfo);
1020 * Called when the FCoE Protocol stack wants information on the underlying
1021 * device. This information is utilized by the FCoE protocol stack to
1022 * register attributes with Fiber Channel management service as per the
1023 * FC-GS Fabric Device Management Information(FDMI) specification.
1024 *
e9bce845
YZ
1025 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1026 * Called when the underlying device wants to override default World Wide
1027 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1028 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1029 * protocol stack to use.
1030 *
c445477d
BH
1031 * RFS acceleration.
1032 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1033 * u16 rxq_index, u32 flow_id);
1034 * Set hardware filter for RFS. rxq_index is the target queue index;
1035 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1036 * Return the filter ID on success, or a negative error code.
fbaec0ea 1037 *
8b98a70c 1038 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
1039 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1040 * Called to make another netdev an underling.
1041 *
1042 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1043 * Called to release previously enslaved netdev.
5455c699
MM
1044 *
1045 * Feature/offload setting functions.
1a2a1444
DM
1046 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1047 * netdev_features_t features);
1048 * Adjusts the requested feature flags according to device-specific
1049 * constraints, and returns the resulting flags. Must not modify
1050 * the device state.
1051 *
c8f44aff 1052 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
1053 * Called to update device configuration to new features. Passed
1054 * feature set might be less than what was returned by ndo_fix_features()).
1055 * Must return >0 or -errno if it changed dev->features itself.
1056 *
edc7d573 1057 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1058 * struct net_device *dev,
f6f6424b 1059 * const unsigned char *addr, u16 vid, u16 flags)
77162022 1060 * Adds an FDB entry to dev for addr.
1690be63
VY
1061 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1062 * struct net_device *dev,
f6f6424b 1063 * const unsigned char *addr, u16 vid)
77162022
JF
1064 * Deletes the FDB entry from dev coresponding to addr.
1065 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1066 * struct net_device *dev, struct net_device *filter_dev,
d297653d 1067 * int *idx)
77162022
JF
1068 * Used to add FDB entries to dump requests. Implementers should add
1069 * entries to skb and update idx with the number of entries.
e5a55a89 1070 *
ad41faa8
ND
1071 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1072 * u16 flags)
e5a55a89 1073 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
1074 * struct net_device *dev, u32 filter_mask,
1075 * int nlflags)
ad41faa8
ND
1076 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1077 * u16 flags);
4bf84c35
JP
1078 *
1079 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1080 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1081 * which do not represent real hardware may define this to allow their
1082 * userspace components to manage their virtual carrier state. Devices
1083 * that determine carrier state from physical hardware properties (eg
1084 * network cables) or protocol-dependent mechanisms (eg
1085 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1086 *
1087 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1088 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1089 * Called to get ID of physical port of this device. If driver does
1090 * not implement this, it is assumed that the hw is not able to have
1091 * multiple net devices on single physical port.
53cf5275 1092 *
7c46a640
AD
1093 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1094 * struct udp_tunnel_info *ti);
1095 * Called by UDP tunnel to notify a driver about the UDP port and socket
1096 * address family that a UDP tunnel is listnening to. It is called only
1097 * when a new port starts listening. The operation is protected by the
1098 * RTNL.
1099 *
1100 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1101 * struct udp_tunnel_info *ti);
1102 * Called by UDP tunnel to notify the driver about a UDP port and socket
1103 * address family that the UDP tunnel is not listening to anymore. The
1104 * operation is protected by the RTNL.
1105 *
a6cc0cfa
JF
1106 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1107 * struct net_device *dev)
1108 * Called by upper layer devices to accelerate switching or other
1109 * station functionality into hardware. 'pdev is the lowerdev
1110 * to use for the offload and 'dev' is the net device that will
1111 * back the offload. Returns a pointer to the private structure
1112 * the upper layer will maintain.
1113 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1114 * Called by upper layer device to delete the station created
1115 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1116 * the station and priv is the structure returned by the add
1117 * operation.
822b3b2e
JF
1118 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1119 * int queue_index, u32 maxrate);
1120 * Called when a user wants to set a max-rate limitation of specific
1121 * TX queue.
a54acb3a
ND
1122 * int (*ndo_get_iflink)(const struct net_device *dev);
1123 * Called to get the iflink value of this device.
d746d707 1124 * void (*ndo_change_proto_down)(struct net_device *dev,
5e82b4b2 1125 * bool proto_down);
d746d707
AK
1126 * This function is used to pass protocol port error state information
1127 * to the switch driver. The switch driver can react to the proto_down
1128 * by doing a phys down on the associated switch port.
fc4099f1
PS
1129 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1130 * This function is used to get egress tunnel information for given skb.
1131 * This is useful for retrieving outer tunnel header parameters while
1132 * sampling packet.
871b642a
PA
1133 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1134 * This function is used to specify the headroom that the skb must
1135 * consider when allocation skb during packet reception. Setting
1136 * appropriate rx headroom value allows avoiding skb head copy on
5e82b4b2 1137 * forward. Setting a negative value resets the rx headroom to the
871b642a 1138 * default value.
a7862b45
BB
1139 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1140 * This function is used to set or query state related to XDP on the
1141 * netdevice. See definition of enum xdp_netdev_command for details.
814abfab
JF
1142 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1143 * This function is used to submit a XDP packet for transmit on a
1144 * netdevice.
11393cc9
JF
1145 * void (*ndo_xdp_flush)(struct net_device *dev);
1146 * This function is used to inform the driver to flush a paticular
1147 * xpd tx queue. Must be called on same CPU as xdp_xmit.
d314774c
SH
1148 */
1149struct net_device_ops {
1150 int (*ndo_init)(struct net_device *dev);
1151 void (*ndo_uninit)(struct net_device *dev);
1152 int (*ndo_open)(struct net_device *dev);
1153 int (*ndo_stop)(struct net_device *dev);
cdba756f
ED
1154 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1155 struct net_device *dev);
1156 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1157 struct net_device *dev,
1158 netdev_features_t features);
00829823 1159 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1160 struct sk_buff *skb,
99932d4f
DB
1161 void *accel_priv,
1162 select_queue_fallback_t fallback);
d314774c
SH
1163 void (*ndo_change_rx_flags)(struct net_device *dev,
1164 int flags);
d314774c 1165 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1166 int (*ndo_set_mac_address)(struct net_device *dev,
1167 void *addr);
d314774c 1168 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1169 int (*ndo_do_ioctl)(struct net_device *dev,
1170 struct ifreq *ifr, int cmd);
d314774c
SH
1171 int (*ndo_set_config)(struct net_device *dev,
1172 struct ifmap *map);
00829823
SH
1173 int (*ndo_change_mtu)(struct net_device *dev,
1174 int new_mtu);
1175 int (*ndo_neigh_setup)(struct net_device *dev,
1176 struct neigh_parms *);
d314774c
SH
1177 void (*ndo_tx_timeout) (struct net_device *dev);
1178
bc1f4470 1179 void (*ndo_get_stats64)(struct net_device *dev,
1180 struct rtnl_link_stats64 *storage);
3df5b3c6 1181 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
2c9d85d4
NF
1182 int (*ndo_get_offload_stats)(int attr_id,
1183 const struct net_device *dev,
1184 void *attr_data);
d314774c
SH
1185 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1186
8e586137 1187 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1188 __be16 proto, u16 vid);
8e586137 1189 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1190 __be16 proto, u16 vid);
d314774c 1191#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1192 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1193 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1194 struct netpoll_info *info);
0e34e931 1195 void (*ndo_netpoll_cleanup)(struct net_device *dev);
d314774c 1196#endif
95c26df8
WM
1197 int (*ndo_set_vf_mac)(struct net_device *dev,
1198 int queue, u8 *mac);
1199 int (*ndo_set_vf_vlan)(struct net_device *dev,
79aab093
MS
1200 int queue, u16 vlan,
1201 u8 qos, __be16 proto);
ed616689
SC
1202 int (*ndo_set_vf_rate)(struct net_device *dev,
1203 int vf, int min_tx_rate,
1204 int max_tx_rate);
5f8444a3
GR
1205 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1206 int vf, bool setting);
dd461d6a
HS
1207 int (*ndo_set_vf_trust)(struct net_device *dev,
1208 int vf, bool setting);
95c26df8
WM
1209 int (*ndo_get_vf_config)(struct net_device *dev,
1210 int vf,
1211 struct ifla_vf_info *ivf);
1d8faf48
RE
1212 int (*ndo_set_vf_link_state)(struct net_device *dev,
1213 int vf, int link_state);
3b766cd8
EBE
1214 int (*ndo_get_vf_stats)(struct net_device *dev,
1215 int vf,
1216 struct ifla_vf_stats
1217 *vf_stats);
57b61080
SF
1218 int (*ndo_set_vf_port)(struct net_device *dev,
1219 int vf,
1220 struct nlattr *port[]);
1221 int (*ndo_get_vf_port)(struct net_device *dev,
1222 int vf, struct sk_buff *skb);
cc8e27cc
EC
1223 int (*ndo_set_vf_guid)(struct net_device *dev,
1224 int vf, u64 guid,
1225 int guid_type);
01a3d796
VZ
1226 int (*ndo_set_vf_rss_query_en)(
1227 struct net_device *dev,
1228 int vf, bool setting);
16e5cc64 1229 int (*ndo_setup_tc)(struct net_device *dev,
a5fcf8a6 1230 u32 handle, u32 chain_index,
16e5cc64
JF
1231 __be16 protocol,
1232 struct tc_to_netdev *tc);
d11ead75 1233#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1234 int (*ndo_fcoe_enable)(struct net_device *dev);
1235 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1236 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1237 u16 xid,
1238 struct scatterlist *sgl,
1239 unsigned int sgc);
1240 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1241 u16 xid);
6247e086
YZ
1242 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1243 u16 xid,
1244 struct scatterlist *sgl,
1245 unsigned int sgc);
68bad94e
NP
1246 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1247 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1248#endif
1249
d11ead75 1250#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1251#define NETDEV_FCOE_WWNN 0
1252#define NETDEV_FCOE_WWPN 1
1253 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1254 u64 *wwn, int type);
4d288d57 1255#endif
3c9c36bc 1256
c445477d
BH
1257#ifdef CONFIG_RFS_ACCEL
1258 int (*ndo_rx_flow_steer)(struct net_device *dev,
1259 const struct sk_buff *skb,
1260 u16 rxq_index,
1261 u32 flow_id);
1262#endif
fbaec0ea
JP
1263 int (*ndo_add_slave)(struct net_device *dev,
1264 struct net_device *slave_dev);
1265 int (*ndo_del_slave)(struct net_device *dev,
1266 struct net_device *slave_dev);
c8f44aff
MM
1267 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1268 netdev_features_t features);
5455c699 1269 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1270 netdev_features_t features);
503eebc2
JP
1271 int (*ndo_neigh_construct)(struct net_device *dev,
1272 struct neighbour *n);
1273 void (*ndo_neigh_destroy)(struct net_device *dev,
1274 struct neighbour *n);
77162022
JF
1275
1276 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1277 struct nlattr *tb[],
77162022 1278 struct net_device *dev,
6b6e2725 1279 const unsigned char *addr,
f6f6424b 1280 u16 vid,
77162022
JF
1281 u16 flags);
1282 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1283 struct nlattr *tb[],
77162022 1284 struct net_device *dev,
f6f6424b
JP
1285 const unsigned char *addr,
1286 u16 vid);
77162022
JF
1287 int (*ndo_fdb_dump)(struct sk_buff *skb,
1288 struct netlink_callback *cb,
1289 struct net_device *dev,
5d5eacb3 1290 struct net_device *filter_dev,
d297653d 1291 int *idx);
e5a55a89
JF
1292
1293 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1294 struct nlmsghdr *nlh,
1295 u16 flags);
e5a55a89
JF
1296 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1297 u32 pid, u32 seq,
6cbdceeb 1298 struct net_device *dev,
46c264da
ND
1299 u32 filter_mask,
1300 int nlflags);
407af329 1301 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1302 struct nlmsghdr *nlh,
1303 u16 flags);
4bf84c35
JP
1304 int (*ndo_change_carrier)(struct net_device *dev,
1305 bool new_carrier);
66b52b0d 1306 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1307 struct netdev_phys_item_id *ppid);
db24a904
DA
1308 int (*ndo_get_phys_port_name)(struct net_device *dev,
1309 char *name, size_t len);
7c46a640
AD
1310 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1311 struct udp_tunnel_info *ti);
1312 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1313 struct udp_tunnel_info *ti);
a6cc0cfa
JF
1314 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1315 struct net_device *dev);
1316 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1317 void *priv);
1318
25175ba5 1319 int (*ndo_get_lock_subclass)(struct net_device *dev);
822b3b2e
JF
1320 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1321 int queue_index,
1322 u32 maxrate);
a54acb3a 1323 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1324 int (*ndo_change_proto_down)(struct net_device *dev,
1325 bool proto_down);
fc4099f1
PS
1326 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1327 struct sk_buff *skb);
871b642a
PA
1328 void (*ndo_set_rx_headroom)(struct net_device *dev,
1329 int needed_headroom);
a7862b45
BB
1330 int (*ndo_xdp)(struct net_device *dev,
1331 struct netdev_xdp *xdp);
814abfab
JF
1332 int (*ndo_xdp_xmit)(struct net_device *dev,
1333 struct xdp_buff *xdp);
11393cc9 1334 void (*ndo_xdp_flush)(struct net_device *dev);
d314774c
SH
1335};
1336
7aa98047
LR
1337/**
1338 * enum net_device_priv_flags - &struct net_device priv_flags
1339 *
1340 * These are the &struct net_device, they are only set internally
1341 * by drivers and used in the kernel. These flags are invisible to
5e82b4b2 1342 * userspace; this means that the order of these flags can change
7aa98047
LR
1343 * during any kernel release.
1344 *
1345 * You should have a pretty good reason to be extending these flags.
1346 *
1347 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1348 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1349 * @IFF_BONDING: bonding master or slave
7aa98047 1350 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1351 * @IFF_WAN_HDLC: WAN HDLC device
1352 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1353 * release skb->dst
1354 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1355 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1356 * @IFF_MACVLAN_PORT: device used as macvlan port
1357 * @IFF_BRIDGE_PORT: device used as bridge port
1358 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1359 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1360 * @IFF_UNICAST_FLT: Supports unicast filtering
1361 * @IFF_TEAM_PORT: device used as team port
1362 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1363 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1364 * change when it's running
1365 * @IFF_MACVLAN: Macvlan device
6d0e24cd
LB
1366 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1367 * underlying stacked devices
1368 * @IFF_IPVLAN_MASTER: IPvlan master device
1369 * @IFF_IPVLAN_SLAVE: IPvlan slave device
007979ea 1370 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1371 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1372 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1373 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
c981e421 1374 * @IFF_TEAM: device is a team device
d4ab4286 1375 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
871b642a
PA
1376 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1377 * entity (i.e. the master device for bridged veth)
3c175784 1378 * @IFF_MACSEC: device is a MACsec device
7aa98047
LR
1379 */
1380enum netdev_priv_flags {
1381 IFF_802_1Q_VLAN = 1<<0,
1382 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1383 IFF_BONDING = 1<<2,
1384 IFF_ISATAP = 1<<3,
1385 IFF_WAN_HDLC = 1<<4,
1386 IFF_XMIT_DST_RELEASE = 1<<5,
1387 IFF_DONT_BRIDGE = 1<<6,
1388 IFF_DISABLE_NETPOLL = 1<<7,
1389 IFF_MACVLAN_PORT = 1<<8,
1390 IFF_BRIDGE_PORT = 1<<9,
1391 IFF_OVS_DATAPATH = 1<<10,
1392 IFF_TX_SKB_SHARING = 1<<11,
1393 IFF_UNICAST_FLT = 1<<12,
1394 IFF_TEAM_PORT = 1<<13,
1395 IFF_SUPP_NOFCS = 1<<14,
1396 IFF_LIVE_ADDR_CHANGE = 1<<15,
1397 IFF_MACVLAN = 1<<16,
1398 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1399 IFF_IPVLAN_MASTER = 1<<18,
1400 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1401 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1402 IFF_NO_QUEUE = 1<<21,
1403 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1404 IFF_L3MDEV_SLAVE = 1<<23,
c981e421 1405 IFF_TEAM = 1<<24,
d4ab4286 1406 IFF_RXFH_CONFIGURED = 1<<25,
871b642a 1407 IFF_PHONY_HEADROOM = 1<<26,
3c175784 1408 IFF_MACSEC = 1<<27,
7aa98047
LR
1409};
1410
1411#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1412#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1413#define IFF_BONDING IFF_BONDING
7aa98047 1414#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1415#define IFF_WAN_HDLC IFF_WAN_HDLC
1416#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1417#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1418#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1419#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1420#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1421#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1422#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1423#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1424#define IFF_TEAM_PORT IFF_TEAM_PORT
1425#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1426#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1427#define IFF_MACVLAN IFF_MACVLAN
02875878 1428#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1429#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1430#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1431#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1432#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1433#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1434#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
c981e421 1435#define IFF_TEAM IFF_TEAM
d4ab4286 1436#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
3c175784 1437#define IFF_MACSEC IFF_MACSEC
7aa98047 1438
536721b1
KK
1439/**
1440 * struct net_device - The DEVICE structure.
d651983d
MCC
1441 *
1442 * Actually, this whole structure is a big mistake. It mixes I/O
1443 * data with strictly "high-level" data, and it has to know about
1444 * almost every data structure used in the INET module.
536721b1
KK
1445 *
1446 * @name: This is the first field of the "visible" part of this structure
1447 * (i.e. as seen by users in the "Space.c" file). It is the name
d651983d 1448 * of the interface.
536721b1
KK
1449 *
1450 * @name_hlist: Device name hash chain, please keep it close to name[]
1451 * @ifalias: SNMP alias
1452 * @mem_end: Shared memory end
1453 * @mem_start: Shared memory start
1454 * @base_addr: Device I/O address
1455 * @irq: Device IRQ number
1456 *
14ffbbb8
TG
1457 * @carrier_changes: Stats to monitor carrier on<->off transitions
1458 *
536721b1
KK
1459 * @state: Generic network queuing layer state, see netdev_state_t
1460 * @dev_list: The global list of network devices
5e82b4b2
BH
1461 * @napi_list: List entry used for polling NAPI devices
1462 * @unreg_list: List entry when we are unregistering the
1463 * device; see the function unregister_netdev
1464 * @close_list: List entry used when we are closing the device
62d885fe
BP
1465 * @ptype_all: Device-specific packet handlers for all protocols
1466 * @ptype_specific: Device-specific, protocol-specific packet handlers
536721b1
KK
1467 *
1468 * @adj_list: Directly linked devices, like slaves for bonding
536721b1
KK
1469 * @features: Currently active device features
1470 * @hw_features: User-changeable features
1471 *
1472 * @wanted_features: User-requested features
1473 * @vlan_features: Mask of features inheritable by VLAN devices
1474 *
1475 * @hw_enc_features: Mask of features inherited by encapsulating devices
1476 * This field indicates what encapsulation
1477 * offloads the hardware is capable of doing,
1478 * and drivers will need to set them appropriately.
1479 *
1480 * @mpls_features: Mask of features inheritable by MPLS
1481 *
1482 * @ifindex: interface index
5e82b4b2 1483 * @group: The group the device belongs to
536721b1
KK
1484 *
1485 * @stats: Statistics struct, which was left as a legacy, use
1486 * rtnl_link_stats64 instead
1487 *
1488 * @rx_dropped: Dropped packets by core network,
1489 * do not use this in drivers
1490 * @tx_dropped: Dropped packets by core network,
1491 * do not use this in drivers
6e7333d3
JW
1492 * @rx_nohandler: nohandler dropped packets by core network on
1493 * inactive devices, do not use this in drivers
536721b1 1494 *
536721b1
KK
1495 * @wireless_handlers: List of functions to handle Wireless Extensions,
1496 * instead of ioctl,
1497 * see <net/iw_handler.h> for details.
1498 * @wireless_data: Instance data managed by the core of wireless extensions
1499 *
1500 * @netdev_ops: Includes several pointers to callbacks,
1501 * if one wants to override the ndo_*() functions
1502 * @ethtool_ops: Management operations
f997c55c
AA
1503 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1504 * discovery handling. Necessary for e.g. 6LoWPAN.
d476059e 1505 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1506 * of Layer 2 headers.
1507 *
1508 * @flags: Interface flags (a la BSD)
1509 * @priv_flags: Like 'flags' but invisible to userspace,
1510 * see if.h for the definitions
1511 * @gflags: Global flags ( kept as legacy )
1512 * @padded: How much padding added by alloc_netdev()
1513 * @operstate: RFC2863 operstate
1514 * @link_mode: Mapping policy to operstate
1515 * @if_port: Selectable AUI, TP, ...
1516 * @dma: DMA channel
1517 * @mtu: Interface MTU value
61e84623
JW
1518 * @min_mtu: Interface Minimum MTU value
1519 * @max_mtu: Interface Maximum MTU value
536721b1 1520 * @type: Interface hardware type
2793a23a 1521 * @hard_header_len: Maximum hardware header length.
217e6fa2 1522 * @min_header_len: Minimum hardware header length
536721b1
KK
1523 *
1524 * @needed_headroom: Extra headroom the hardware may need, but not in all
1525 * cases can this be guaranteed
1526 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1527 * cases can this be guaranteed. Some cases also use
1528 * LL_MAX_HEADER instead to allocate the skb
1529 *
1530 * interface address info:
1531 *
1532 * @perm_addr: Permanent hw address
1533 * @addr_assign_type: Hw address assignment type
1534 * @addr_len: Hardware address length
8626a0c8 1535 * @neigh_priv_len: Used in neigh_alloc()
536721b1
KK
1536 * @dev_id: Used to differentiate devices that share
1537 * the same link layer address
1538 * @dev_port: Used to differentiate devices that share
1539 * the same function
1540 * @addr_list_lock: XXX: need comments on this one
5e82b4b2 1541 * @uc_promisc: Counter that indicates promiscuous mode
536721b1
KK
1542 * has been enabled due to the need to listen to
1543 * additional unicast addresses in a device that
1544 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1545 * @uc: unicast mac addresses
1546 * @mc: multicast mac addresses
1547 * @dev_addrs: list of device hw addresses
1548 * @queues_kset: Group of all Kobjects in the Tx and RX queues
5e82b4b2
BH
1549 * @promiscuity: Number of times the NIC is told to work in
1550 * promiscuous mode; if it becomes 0 the NIC will
1551 * exit promiscuous mode
536721b1
KK
1552 * @allmulti: Counter, enables or disables allmulticast mode
1553 *
1554 * @vlan_info: VLAN info
1555 * @dsa_ptr: dsa specific data
1556 * @tipc_ptr: TIPC specific data
1557 * @atalk_ptr: AppleTalk link
1558 * @ip_ptr: IPv4 specific data
1559 * @dn_ptr: DECnet specific data
1560 * @ip6_ptr: IPv6 specific data
1561 * @ax25_ptr: AX.25 specific data
1562 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1563 *
536721b1
KK
1564 * @dev_addr: Hw address (before bcast,
1565 * because most packets are unicast)
1566 *
1567 * @_rx: Array of RX queues
1568 * @num_rx_queues: Number of RX queues
1569 * allocated at register_netdev() time
1570 * @real_num_rx_queues: Number of RX queues currently active in device
1571 *
1572 * @rx_handler: handler for received packets
1573 * @rx_handler_data: XXX: need comments on this one
1574 * @ingress_queue: XXX: need comments on this one
1575 * @broadcast: hw bcast address
1576 *
14ffbbb8
TG
1577 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1578 * indexed by RX queue number. Assigned by driver.
1579 * This must only be set if the ndo_rx_flow_steer
1580 * operation is defined
1581 * @index_hlist: Device index hash chain
1582 *
536721b1
KK
1583 * @_tx: Array of TX queues
1584 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1585 * @real_num_tx_queues: Number of TX queues currently active in device
1586 * @qdisc: Root qdisc from userspace point of view
1587 * @tx_queue_len: Max frames per queue allowed
1588 * @tx_global_lock: XXX: need comments on this one
1589 *
1590 * @xps_maps: XXX: need comments on this one
1591 *
536721b1 1592 * @watchdog_timeo: Represents the timeout that is used by
5e82b4b2 1593 * the watchdog (see dev_watchdog())
536721b1
KK
1594 * @watchdog_timer: List of timers
1595 *
1596 * @pcpu_refcnt: Number of references to this device
1597 * @todo_list: Delayed register/unregister
536721b1
KK
1598 * @link_watch_list: XXX: need comments on this one
1599 *
1600 * @reg_state: Register/unregister state machine
1601 * @dismantle: Device is going to be freed
1602 * @rtnl_link_state: This enum represents the phases of creating
1603 * a new link
1604 *
cf124db5
DM
1605 * @needs_free_netdev: Should unregister perform free_netdev?
1606 * @priv_destructor: Called from unregister
536721b1
KK
1607 * @npinfo: XXX: need comments on this one
1608 * @nd_net: Network namespace this network device is inside
1609 *
1610 * @ml_priv: Mid-layer private
1611 * @lstats: Loopback statistics
1612 * @tstats: Tunnel statistics
1613 * @dstats: Dummy statistics
1614 * @vstats: Virtual ethernet statistics
1615 *
1616 * @garp_port: GARP
1617 * @mrp_port: MRP
1618 *
1619 * @dev: Class/net/name entry
1620 * @sysfs_groups: Space for optional device, statistics and wireless
1621 * sysfs groups
1622 *
1623 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1624 * @rtnl_link_ops: Rtnl_link_ops
1625 *
1626 * @gso_max_size: Maximum size of generic segmentation offload
1627 * @gso_max_segs: Maximum number of segments that can be passed to the
1628 * NIC for GSO
1629 *
1630 * @dcbnl_ops: Data Center Bridging netlink ops
1631 * @num_tc: Number of traffic classes in the net device
1632 * @tc_to_txq: XXX: need comments on this one
920c1cd3 1633 * @prio_tc_map: XXX: need comments on this one
536721b1
KK
1634 *
1635 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1636 *
1637 * @priomap: XXX: need comments on this one
1638 * @phydev: Physical device may attach itself
1639 * for hardware timestamping
1640 *
123b3652
ED
1641 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1642 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
536721b1 1643 *
d746d707
AK
1644 * @proto_down: protocol port state information can be sent to the
1645 * switch driver and used to set the phys state of the
1646 * switch port.
1647 *
1da177e4
LT
1648 * FIXME: cleanup struct net_device such that network protocol info
1649 * moves out.
1650 */
1651
d94d9fee 1652struct net_device {
1da177e4 1653 char name[IFNAMSIZ];
9356b8fc 1654 struct hlist_node name_hlist;
0b815a1a 1655 char *ifalias;
1da177e4
LT
1656 /*
1657 * I/O specific fields
1658 * FIXME: Merge these and struct ifmap into one
1659 */
536721b1
KK
1660 unsigned long mem_end;
1661 unsigned long mem_start;
1662 unsigned long base_addr;
1663 int irq;
1da177e4 1664
14ffbbb8
TG
1665 atomic_t carrier_changes;
1666
1da177e4 1667 /*
536721b1
KK
1668 * Some hardware also needs these fields (state,dev_list,
1669 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1670 * part of the usual set specified in Space.c.
1671 */
1672
1da177e4
LT
1673 unsigned long state;
1674
7562f876 1675 struct list_head dev_list;
bea3348e 1676 struct list_head napi_list;
44a0873d 1677 struct list_head unreg_list;
5cde2829 1678 struct list_head close_list;
7866a621
SN
1679 struct list_head ptype_all;
1680 struct list_head ptype_specific;
2f268f12 1681
2f268f12
VF
1682 struct {
1683 struct list_head upper;
1684 struct list_head lower;
1685 } adj_list;
1686
c8f44aff 1687 netdev_features_t features;
c8f44aff 1688 netdev_features_t hw_features;
c8f44aff 1689 netdev_features_t wanted_features;
c8f44aff 1690 netdev_features_t vlan_features;
6a674e9c 1691 netdev_features_t hw_enc_features;
0d89d203 1692 netdev_features_t mpls_features;
802ab55a 1693 netdev_features_t gso_partial_features;
04ed3e74 1694
1da177e4 1695 int ifindex;
7a66bbc9 1696 int group;
1da177e4 1697
c45d286e 1698 struct net_device_stats stats;
015f0688 1699
015f0688
ED
1700 atomic_long_t rx_dropped;
1701 atomic_long_t tx_dropped;
6e7333d3 1702 atomic_long_t rx_nohandler;
1da177e4 1703
b86e0280 1704#ifdef CONFIG_WIRELESS_EXT
5e82b4b2
BH
1705 const struct iw_handler_def *wireless_handlers;
1706 struct iw_public_data *wireless_data;
b86e0280 1707#endif
d314774c 1708 const struct net_device_ops *netdev_ops;
76fd8593 1709 const struct ethtool_ops *ethtool_ops;
4170604f 1710#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1711 const struct switchdev_ops *switchdev_ops;
4170604f 1712#endif
1b69c6d0
DA
1713#ifdef CONFIG_NET_L3_MASTER_DEV
1714 const struct l3mdev_ops *l3mdev_ops;
1715#endif
f997c55c
AA
1716#if IS_ENABLED(CONFIG_IPV6)
1717 const struct ndisc_ops *ndisc_ops;
1718#endif
1da177e4 1719
d77e38e6
SK
1720#ifdef CONFIG_XFRM
1721 const struct xfrmdev_ops *xfrmdev_ops;
1722#endif
1723
3b04ddde
SH
1724 const struct header_ops *header_ops;
1725
536721b1
KK
1726 unsigned int flags;
1727 unsigned int priv_flags;
1728
1da177e4 1729 unsigned short gflags;
536721b1 1730 unsigned short padded;
1da177e4 1731
536721b1
KK
1732 unsigned char operstate;
1733 unsigned char link_mode;
b00055aa 1734
536721b1
KK
1735 unsigned char if_port;
1736 unsigned char dma;
bdc220da 1737
536721b1 1738 unsigned int mtu;
61e84623
JW
1739 unsigned int min_mtu;
1740 unsigned int max_mtu;
536721b1
KK
1741 unsigned short type;
1742 unsigned short hard_header_len;
d92be7a4 1743 unsigned char min_header_len;
1da177e4 1744
f5184d26
JB
1745 unsigned short needed_headroom;
1746 unsigned short needed_tailroom;
1747
1da177e4 1748 /* Interface address info. */
536721b1
KK
1749 unsigned char perm_addr[MAX_ADDR_LEN];
1750 unsigned char addr_assign_type;
1751 unsigned char addr_len;
a0a9663d 1752 unsigned short neigh_priv_len;
536721b1
KK
1753 unsigned short dev_id;
1754 unsigned short dev_port;
ccffad25 1755 spinlock_t addr_list_lock;
14ffbbb8
TG
1756 unsigned char name_assign_type;
1757 bool uc_promisc;
536721b1
KK
1758 struct netdev_hw_addr_list uc;
1759 struct netdev_hw_addr_list mc;
1760 struct netdev_hw_addr_list dev_addrs;
1761
4c3d5e7b
ED
1762#ifdef CONFIG_SYSFS
1763 struct kset *queues_kset;
1764#endif
9d45abe1
WC
1765 unsigned int promiscuity;
1766 unsigned int allmulti;
1da177e4 1767
1da177e4 1768
5e82b4b2 1769 /* Protocol-specific pointers */
65ac6a5f 1770
d11ead75 1771#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1772 struct vlan_info __rcu *vlan_info;
65ac6a5f 1773#endif
34a430d7 1774#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1775 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1776#endif
1777#if IS_ENABLED(CONFIG_TIPC)
536721b1 1778 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1779#endif
536721b1
KK
1780 void *atalk_ptr;
1781 struct in_device __rcu *ip_ptr;
1782 struct dn_dev __rcu *dn_ptr;
1783 struct inet6_dev __rcu *ip6_ptr;
1784 void *ax25_ptr;
1785 struct wireless_dev *ieee80211_ptr;
98a18b6f 1786 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1787#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1788 struct mpls_dev __rcu *mpls_ptr;
1789#endif
1da177e4 1790
9356b8fc 1791/*
cd13539b 1792 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1793 */
9356b8fc 1794 /* Interface address info used in eth_type_trans() */
536721b1 1795 unsigned char *dev_addr;
f001fde5 1796
a953be53 1797#ifdef CONFIG_SYSFS
0a9627f2
TH
1798 struct netdev_rx_queue *_rx;
1799
0a9627f2 1800 unsigned int num_rx_queues;
62fe0b40 1801 unsigned int real_num_rx_queues;
df334545 1802#endif
0a9627f2 1803
7acedaf5 1804 struct bpf_prog __rcu *xdp_prog;
3b47d303 1805 unsigned long gro_flush_timeout;
61391cde 1806 rx_handler_func_t __rcu *rx_handler;
1807 void __rcu *rx_handler_data;
e8a0464c 1808
4cda01e8 1809#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1810 struct tcf_proto __rcu *ingress_cl_list;
1811#endif
24824a09 1812 struct netdev_queue __rcu *ingress_queue;
e687ad60 1813#ifdef CONFIG_NETFILTER_INGRESS
e3b37f11 1814 struct nf_hook_entry __rcu *nf_hooks_ingress;
e687ad60 1815#endif
d2788d34 1816
536721b1 1817 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1818#ifdef CONFIG_RFS_ACCEL
1819 struct cpu_rmap *rx_cpu_rmap;
1820#endif
1821 struct hlist_node index_hlist;
cd13539b
ED
1822
1823/*
1824 * Cache lines mostly used on transmit path
1825 */
e8a0464c
DM
1826 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1827 unsigned int num_tx_queues;
fd2ea0a7 1828 unsigned int real_num_tx_queues;
af356afa 1829 struct Qdisc *qdisc;
59cc1f61
JK
1830#ifdef CONFIG_NET_SCHED
1831 DECLARE_HASHTABLE (qdisc_hash, 4);
1832#endif
0cd29503 1833 unsigned int tx_queue_len;
c3f26a26 1834 spinlock_t tx_global_lock;
14ffbbb8 1835 int watchdog_timeo;
cd13539b 1836
bf264145 1837#ifdef CONFIG_XPS
a4177869 1838 struct xps_dev_maps __rcu *xps_maps;
bf264145 1839#endif
1f211a1b
DB
1840#ifdef CONFIG_NET_CLS_ACT
1841 struct tcf_proto __rcu *egress_cl_list;
1842#endif
0c4f691f 1843
9356b8fc 1844 /* These may be needed for future network-power-down code. */
9356b8fc
ED
1845 struct timer_list watchdog_timer;
1846
29b4433d 1847 int __percpu *pcpu_refcnt;
1da177e4 1848 struct list_head todo_list;
1da177e4 1849
e014debe 1850 struct list_head link_watch_list;
572a103d 1851
1da177e4 1852 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1853 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1854 NETREG_UNREGISTERING, /* called unregister_netdevice */
1855 NETREG_UNREGISTERED, /* completed unregister todo */
1856 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1857 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1858 } reg_state:8;
1859
536721b1 1860 bool dismantle;
a2835763
PM
1861
1862 enum {
1863 RTNL_LINK_INITIALIZED,
1864 RTNL_LINK_INITIALIZING,
1865 } rtnl_link_state:16;
1da177e4 1866
cf124db5
DM
1867 bool needs_free_netdev;
1868 void (*priv_destructor)(struct net_device *dev);
1da177e4 1869
1da177e4 1870#ifdef CONFIG_NETPOLL
5fbee843 1871 struct netpoll_info __rcu *npinfo;
1da177e4 1872#endif
eae792b7 1873
0c5c9fb5 1874 possible_net_t nd_net;
4a1c5371 1875
4951704b 1876 /* mid-layer private */
a7855c78 1877 union {
536721b1
KK
1878 void *ml_priv;
1879 struct pcpu_lstats __percpu *lstats;
8f84985f 1880 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1881 struct pcpu_dstats __percpu *dstats;
1882 struct pcpu_vstats __percpu *vstats;
a7855c78 1883 };
536721b1 1884
fb585b44 1885#if IS_ENABLED(CONFIG_GARP)
3cc77ec7 1886 struct garp_port __rcu *garp_port;
fb585b44
TK
1887#endif
1888#if IS_ENABLED(CONFIG_MRP)
febf018d 1889 struct mrp_port __rcu *mrp_port;
fb585b44 1890#endif
1da177e4 1891
5e82b4b2 1892 struct device dev;
0c509a6c 1893 const struct attribute_group *sysfs_groups[4];
a953be53 1894 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1895
38f7b870 1896 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1897
82cc1a7a
PWJ
1898 /* for setting kernel sock attribute on TCP connection setup */
1899#define GSO_MAX_SIZE 65536
1900 unsigned int gso_max_size;
30b678d8
BH
1901#define GSO_MAX_SEGS 65535
1902 u16 gso_max_segs;
743b03a8 1903
7a6b6f51 1904#ifdef CONFIG_DCB
32953543 1905 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1906#endif
5e82b4b2
BH
1907 u8 num_tc;
1908 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1909 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1910
d11ead75 1911#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1912 unsigned int fcoe_ddp_xid;
5bc1421e 1913#endif
86f8515f 1914#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1915 struct netprio_map __rcu *priomap;
4d288d57 1916#endif
5e82b4b2
BH
1917 struct phy_device *phydev;
1918 struct lock_class_key *qdisc_tx_busylock;
f9eb8aea 1919 struct lock_class_key *qdisc_running_key;
5e82b4b2 1920 bool proto_down;
1da177e4 1921};
43cb76d9 1922#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4 1923
b5cdae32
DM
1924static inline bool netif_elide_gro(const struct net_device *dev)
1925{
1926 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1927 return true;
1928 return false;
1929}
1930
1da177e4 1931#define NETDEV_ALIGN 32
1da177e4 1932
4f57c087
JF
1933static inline
1934int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1935{
1936 return dev->prio_tc_map[prio & TC_BITMASK];
1937}
1938
1939static inline
1940int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1941{
1942 if (tc >= dev->num_tc)
1943 return -EINVAL;
1944
1945 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1946 return 0;
1947}
1948
8d059b0f 1949int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
9cf1f6a8
AD
1950void netdev_reset_tc(struct net_device *dev);
1951int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1952int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
4f57c087
JF
1953
1954static inline
1955int netdev_get_num_tc(struct net_device *dev)
1956{
1957 return dev->num_tc;
1958}
1959
e8a0464c
DM
1960static inline
1961struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1962 unsigned int index)
1963{
1964 return &dev->_tx[index];
1965}
1966
10c51b56
DB
1967static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1968 const struct sk_buff *skb)
1969{
1970 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1971}
1972
e8a0464c
DM
1973static inline void netdev_for_each_tx_queue(struct net_device *dev,
1974 void (*f)(struct net_device *,
1975 struct netdev_queue *,
1976 void *),
1977 void *arg)
1978{
1979 unsigned int i;
1980
1981 for (i = 0; i < dev->num_tx_queues; i++)
1982 f(dev, &dev->_tx[i], arg);
1983}
1984
d3fff6c4
ED
1985#define netdev_lockdep_set_classes(dev) \
1986{ \
1987 static struct lock_class_key qdisc_tx_busylock_key; \
1988 static struct lock_class_key qdisc_running_key; \
1989 static struct lock_class_key qdisc_xmit_lock_key; \
1990 static struct lock_class_key dev_addr_list_lock_key; \
1991 unsigned int i; \
1992 \
1993 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1994 (dev)->qdisc_running_key = &qdisc_running_key; \
1995 lockdep_set_class(&(dev)->addr_list_lock, \
1996 &dev_addr_list_lock_key); \
1997 for (i = 0; i < (dev)->num_tx_queues; i++) \
1998 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1999 &qdisc_xmit_lock_key); \
2000}
2001
f629d208 2002struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
2003 struct sk_buff *skb,
2004 void *accel_priv);
8c4c49df 2005
871b642a
PA
2006/* returns the headroom that the master device needs to take in account
2007 * when forwarding to this dev
2008 */
2009static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2010{
2011 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2012}
2013
2014static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2015{
2016 if (dev->netdev_ops->ndo_set_rx_headroom)
2017 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2018}
2019
2020/* set the device rx headroom to the dev's default */
2021static inline void netdev_reset_rx_headroom(struct net_device *dev)
2022{
2023 netdev_set_rx_headroom(dev, -1);
2024}
2025
c346dca1
YH
2026/*
2027 * Net namespace inlines
2028 */
2029static inline
2030struct net *dev_net(const struct net_device *dev)
2031{
c2d9ba9b 2032 return read_pnet(&dev->nd_net);
c346dca1
YH
2033}
2034
2035static inline
f5aa23fd 2036void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 2037{
0c5c9fb5 2038 write_pnet(&dev->nd_net, net);
c346dca1
YH
2039}
2040
bea3348e
SH
2041/**
2042 * netdev_priv - access network device private data
2043 * @dev: network device
2044 *
2045 * Get network device private data
2046 */
6472ce60 2047static inline void *netdev_priv(const struct net_device *dev)
1da177e4 2048{
1ce8e7b5 2049 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
2050}
2051
1da177e4
LT
2052/* Set the sysfs physical device reference for the network logical device
2053 * if set prior to registration will cause a symlink during initialization.
2054 */
43cb76d9 2055#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 2056
384912ed 2057/* Set the sysfs device type for the network logical device to allow
3f79410c 2058 * fine-grained identification of different network device types. For
5e82b4b2 2059 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
384912ed
MH
2060 */
2061#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2062
82dc3c63
ED
2063/* Default NAPI poll() weight
2064 * Device drivers are strongly advised to not use bigger value
2065 */
2066#define NAPI_POLL_WEIGHT 64
2067
3b582cc1 2068/**
5e82b4b2 2069 * netif_napi_add - initialize a NAPI context
3b582cc1 2070 * @dev: network device
5e82b4b2 2071 * @napi: NAPI context
3b582cc1
SH
2072 * @poll: polling function
2073 * @weight: default weight
2074 *
5e82b4b2
BH
2075 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2076 * *any* of the other NAPI-related functions.
3b582cc1 2077 */
d565b0a1
HX
2078void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2079 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 2080
d64b5e85 2081/**
5e82b4b2 2082 * netif_tx_napi_add - initialize a NAPI context
d64b5e85 2083 * @dev: network device
5e82b4b2 2084 * @napi: NAPI context
d64b5e85
ED
2085 * @poll: polling function
2086 * @weight: default weight
2087 *
2088 * This variant of netif_napi_add() should be used from drivers using NAPI
2089 * to exclusively poll a TX queue.
2090 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2091 */
2092static inline void netif_tx_napi_add(struct net_device *dev,
2093 struct napi_struct *napi,
2094 int (*poll)(struct napi_struct *, int),
2095 int weight)
2096{
2097 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2098 netif_napi_add(dev, napi, poll, weight);
2099}
2100
d8156534 2101/**
5e82b4b2
BH
2102 * netif_napi_del - remove a NAPI context
2103 * @napi: NAPI context
d8156534 2104 *
5e82b4b2 2105 * netif_napi_del() removes a NAPI context from the network device NAPI list
d8156534 2106 */
d565b0a1
HX
2107void netif_napi_del(struct napi_struct *napi);
2108
2109struct napi_gro_cb {
78a478d0 2110 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
5e82b4b2 2111 void *frag0;
78a478d0 2112
7489594c
HX
2113 /* Length of frag0. */
2114 unsigned int frag0_len;
2115
86911732 2116 /* This indicates where we are processing relative to skb->data. */
5e82b4b2 2117 int data_offset;
86911732 2118
d565b0a1 2119 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
2120 u16 flush;
2121
2122 /* Save the IP ID here and check when we get to the transport layer */
2123 u16 flush_id;
d565b0a1
HX
2124
2125 /* Number of segments aggregated. */
2e71a6f8
ED
2126 u16 count;
2127
15e2396d
TH
2128 /* Start offset for remote checksum offload */
2129 u16 gro_remcsum_start;
2130
2e71a6f8
ED
2131 /* jiffies when first packet was created/queued */
2132 unsigned long age;
86347245 2133
afe93325 2134 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
2135 u16 proto;
2136
baa32ff4
TH
2137 /* This is non-zero if the packet may be of the same flow. */
2138 u8 same_flow:1;
2139
fac8e0f5
JG
2140 /* Used in tunnel GRO receive */
2141 u8 encap_mark:1;
573e8fca
TH
2142
2143 /* GRO checksum is valid */
2144 u8 csum_valid:1;
2145
662880f4
TH
2146 /* Number of checksums via CHECKSUM_UNNECESSARY */
2147 u8 csum_cnt:3;
c3c7c254 2148
baa32ff4
TH
2149 /* Free the skb? */
2150 u8 free:2;
2151#define NAPI_GRO_FREE 1
2152#define NAPI_GRO_FREE_STOLEN_HEAD 2
2153
efc98d08
TH
2154 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2155 u8 is_ipv6:1;
2156
a0ca153f
AD
2157 /* Used in GRE, set in fou/gue_gro_receive */
2158 u8 is_fou:1;
2159
1530545e
AD
2160 /* Used to determine if flush_id can be ignored */
2161 u8 is_atomic:1;
2162
fcd91dd4
SD
2163 /* Number of gro_receive callbacks this packet already went through */
2164 u8 recursion_counter:4;
2165
2166 /* 1 bit hole */
baa32ff4 2167
bf5a755f
JC
2168 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2169 __wsum csum;
2170
c3c7c254
ED
2171 /* used in skb_gro_receive() slow path */
2172 struct sk_buff *last;
d565b0a1
HX
2173};
2174
2175#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2176
fcd91dd4
SD
2177#define GRO_RECURSION_LIMIT 15
2178static inline int gro_recursion_inc_test(struct sk_buff *skb)
2179{
2180 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2181}
2182
2183typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2184static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2185 struct sk_buff **head,
2186 struct sk_buff *skb)
2187{
2188 if (unlikely(gro_recursion_inc_test(skb))) {
2189 NAPI_GRO_CB(skb)->flush |= 1;
2190 return NULL;
2191 }
2192
2193 return cb(head, skb);
2194}
2195
2196typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2197 struct sk_buff *);
2198static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2199 struct sock *sk,
2200 struct sk_buff **head,
2201 struct sk_buff *skb)
2202{
2203 if (unlikely(gro_recursion_inc_test(skb))) {
2204 NAPI_GRO_CB(skb)->flush |= 1;
2205 return NULL;
2206 }
2207
2208 return cb(sk, head, skb);
2209}
2210
1da177e4 2211struct packet_type {
f2ccd8fa
DM
2212 __be16 type; /* This is really htons(ether_type). */
2213 struct net_device *dev; /* NULL is wildcarded here */
2214 int (*func) (struct sk_buff *,
2215 struct net_device *,
2216 struct packet_type *,
2217 struct net_device *);
c0de08d0
EL
2218 bool (*id_match)(struct packet_type *ptype,
2219 struct sock *sk);
1da177e4
LT
2220 void *af_packet_priv;
2221 struct list_head list;
2222};
2223
f191a1d1 2224struct offload_callbacks {
576a30eb 2225 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2226 netdev_features_t features);
d565b0a1 2227 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2228 struct sk_buff *skb);
299603e8 2229 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2230};
2231
2232struct packet_offload {
2233 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2234 u16 priority;
f191a1d1
VY
2235 struct offload_callbacks callbacks;
2236 struct list_head list;
1da177e4
LT
2237};
2238
5e82b4b2 2239/* often modified stats are per-CPU, other are shared (netdev->stats) */
8f84985f
LR
2240struct pcpu_sw_netstats {
2241 u64 rx_packets;
2242 u64 rx_bytes;
2243 u64 tx_packets;
2244 u64 tx_bytes;
2245 struct u64_stats_sync syncp;
2246};
2247
aabc92bb
PNA
2248#define __netdev_alloc_pcpu_stats(type, gfp) \
2249({ \
2250 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2251 if (pcpu_stats) { \
2252 int __cpu; \
2253 for_each_possible_cpu(__cpu) { \
2254 typeof(type) *stat; \
2255 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2256 u64_stats_init(&stat->syncp); \
2257 } \
2258 } \
2259 pcpu_stats; \
1c213bd2
WC
2260})
2261
aabc92bb 2262#define netdev_alloc_pcpu_stats(type) \
326fcfa5 2263 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
aabc92bb 2264
764f5e54
JP
2265enum netdev_lag_tx_type {
2266 NETDEV_LAG_TX_TYPE_UNKNOWN,
2267 NETDEV_LAG_TX_TYPE_RANDOM,
2268 NETDEV_LAG_TX_TYPE_BROADCAST,
2269 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2270 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2271 NETDEV_LAG_TX_TYPE_HASH,
2272};
2273
2274struct netdev_lag_upper_info {
2275 enum netdev_lag_tx_type tx_type;
2276};
2277
fb1b2e3c
JP
2278struct netdev_lag_lower_state_info {
2279 u8 link_up : 1,
2280 tx_enabled : 1;
2281};
2282
1da177e4
LT
2283#include <linux/notifier.h>
2284
dcfe1421
AW
2285/* netdevice notifier chain. Please remember to update the rtnetlink
2286 * notification exclusion list in rtnetlink_event() when adding new
2287 * types.
2288 */
2289#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2290#define NETDEV_DOWN 0x0002
2291#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2292 detected a hardware crash and restarted
2293 - we can use this eg to kick tcp sessions
2294 once done */
2295#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2296#define NETDEV_REGISTER 0x0005
2297#define NETDEV_UNREGISTER 0x0006
1d486bfb 2298#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2299#define NETDEV_CHANGEADDR 0x0008
2300#define NETDEV_GOING_DOWN 0x0009
2301#define NETDEV_CHANGENAME 0x000A
2302#define NETDEV_FEAT_CHANGE 0x000B
2303#define NETDEV_BONDING_FAILOVER 0x000C
2304#define NETDEV_PRE_UP 0x000D
2305#define NETDEV_PRE_TYPE_CHANGE 0x000E
2306#define NETDEV_POST_TYPE_CHANGE 0x000F
2307#define NETDEV_POST_INIT 0x0010
0115e8e3 2308#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2309#define NETDEV_RELEASE 0x0012
2310#define NETDEV_NOTIFY_PEERS 0x0013
2311#define NETDEV_JOIN 0x0014
42e52bf9 2312#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2313#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2314#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2315#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2316#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2317#define NETDEV_PRECHANGEUPPER 0x001A
04d48266 2318#define NETDEV_CHANGELOWERSTATE 0x001B
7c46a640 2319#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
08294a26 2320#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
dcfe1421 2321
f629d208
JP
2322int register_netdevice_notifier(struct notifier_block *nb);
2323int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2324
2325struct netdev_notifier_info {
2326 struct net_device *dev;
2327};
2328
be9efd36
JP
2329struct netdev_notifier_change_info {
2330 struct netdev_notifier_info info; /* must be first */
2331 unsigned int flags_changed;
2332};
2333
0e4ead9d
JP
2334struct netdev_notifier_changeupper_info {
2335 struct netdev_notifier_info info; /* must be first */
2336 struct net_device *upper_dev; /* new upper dev */
2337 bool master; /* is upper dev master */
5e82b4b2 2338 bool linking; /* is the notification for link or unlink */
29bf24af 2339 void *upper_info; /* upper dev info */
0e4ead9d
JP
2340};
2341
04d48266
JP
2342struct netdev_notifier_changelowerstate_info {
2343 struct netdev_notifier_info info; /* must be first */
2344 void *lower_state_info; /* is lower dev state */
2345};
2346
75538c2b
CW
2347static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2348 struct net_device *dev)
2349{
2350 info->dev = dev;
2351}
2352
351638e7
JP
2353static inline struct net_device *
2354netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2355{
2356 return info->dev;
2357}
2358
f629d208 2359int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2360
2361
1da177e4
LT
2362extern rwlock_t dev_base_lock; /* Device list lock */
2363
881d966b
EB
2364#define for_each_netdev(net, d) \
2365 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2366#define for_each_netdev_reverse(net, d) \
2367 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2368#define for_each_netdev_rcu(net, d) \
2369 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2370#define for_each_netdev_safe(net, d, n) \
2371 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2372#define for_each_netdev_continue(net, d) \
2373 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2374#define for_each_netdev_continue_rcu(net, d) \
2375 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2376#define for_each_netdev_in_bond_rcu(bond, slave) \
2377 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2378 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2379#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2380
a050c33f
DL
2381static inline struct net_device *next_net_device(struct net_device *dev)
2382{
2383 struct list_head *lh;
2384 struct net *net;
2385
c346dca1 2386 net = dev_net(dev);
a050c33f
DL
2387 lh = dev->dev_list.next;
2388 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2389}
2390
ce81b76a
ED
2391static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2392{
2393 struct list_head *lh;
2394 struct net *net;
2395
2396 net = dev_net(dev);
ccf43438 2397 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2398 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2399}
2400
a050c33f
DL
2401static inline struct net_device *first_net_device(struct net *net)
2402{
2403 return list_empty(&net->dev_base_head) ? NULL :
2404 net_device_entry(net->dev_base_head.next);
2405}
7562f876 2406
ccf43438
ED
2407static inline struct net_device *first_net_device_rcu(struct net *net)
2408{
2409 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2410
2411 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2412}
2413
f629d208
JP
2414int netdev_boot_setup_check(struct net_device *dev);
2415unsigned long netdev_boot_base(const char *prefix, int unit);
2416struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2417 const char *hwaddr);
2418struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2419struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2420void dev_add_pack(struct packet_type *pt);
2421void dev_remove_pack(struct packet_type *pt);
2422void __dev_remove_pack(struct packet_type *pt);
2423void dev_add_offload(struct packet_offload *po);
2424void dev_remove_offload(struct packet_offload *po);
f629d208 2425
a54acb3a 2426int dev_get_iflink(const struct net_device *dev);
fc4099f1 2427int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2428struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2429 unsigned short mask);
f629d208
JP
2430struct net_device *dev_get_by_name(struct net *net, const char *name);
2431struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2432struct net_device *__dev_get_by_name(struct net *net, const char *name);
2433int dev_alloc_name(struct net_device *dev, const char *name);
2434int dev_open(struct net_device *dev);
7051b88a 2435void dev_close(struct net_device *dev);
2436void dev_close_many(struct list_head *head, bool unlink);
f629d208 2437void dev_disable_lro(struct net_device *dev);
0c4b51f0 2438int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2439int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2440int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2441int register_netdevice(struct net_device *dev);
2442void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2443void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2444static inline void unregister_netdevice(struct net_device *dev)
2445{
2446 unregister_netdevice_queue(dev, NULL);
2447}
2448
f629d208
JP
2449int netdev_refcnt_read(const struct net_device *dev);
2450void free_netdev(struct net_device *dev);
74d332c1 2451void netdev_freemem(struct net_device *dev);
f629d208
JP
2452void synchronize_net(void);
2453int init_dummy_netdev(struct net_device *dev);
937f1ba5 2454
f60e5990 2455DECLARE_PER_CPU(int, xmit_recursion);
a70b506e
DB
2456#define XMIT_RECURSION_LIMIT 10
2457
f60e5990 2458static inline int dev_recursion_level(void)
2459{
2460 return this_cpu_read(xmit_recursion);
2461}
2462
f629d208
JP
2463struct net_device *dev_get_by_index(struct net *net, int ifindex);
2464struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2465struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
90b602f8 2466struct net_device *dev_get_by_napi_id(unsigned int napi_id);
f629d208
JP
2467int netdev_get_name(struct net *net, char *name, int ifindex);
2468int dev_restart(struct net_device *dev);
f629d208 2469int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2470
2471static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2472{
2473 return NAPI_GRO_CB(skb)->data_offset;
2474}
2475
2476static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2477{
2478 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2479}
2480
2481static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2482{
2483 NAPI_GRO_CB(skb)->data_offset += len;
2484}
2485
a5b1cf28
HX
2486static inline void *skb_gro_header_fast(struct sk_buff *skb,
2487 unsigned int offset)
86911732 2488{
a5b1cf28
HX
2489 return NAPI_GRO_CB(skb)->frag0 + offset;
2490}
78a478d0 2491
a5b1cf28
HX
2492static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2493{
2494 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2495}
78a478d0 2496
57ea52a8
HX
2497static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2498{
2499 NAPI_GRO_CB(skb)->frag0 = NULL;
2500 NAPI_GRO_CB(skb)->frag0_len = 0;
2501}
2502
a5b1cf28
HX
2503static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2504 unsigned int offset)
2505{
17dd759c
HX
2506 if (!pskb_may_pull(skb, hlen))
2507 return NULL;
2508
57ea52a8 2509 skb_gro_frag0_invalidate(skb);
17dd759c 2510 return skb->data + offset;
86911732 2511}
1da177e4 2512
36e7b1b8
HX
2513static inline void *skb_gro_network_header(struct sk_buff *skb)
2514{
78d3fd0b
HX
2515 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2516 skb_network_offset(skb);
36e7b1b8
HX
2517}
2518
bf5a755f
JC
2519static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2520 const void *start, unsigned int len)
2521{
573e8fca 2522 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2523 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2524 csum_partial(start, len, 0));
2525}
2526
573e8fca
TH
2527/* GRO checksum functions. These are logical equivalents of the normal
2528 * checksum functions (in skbuff.h) except that they operate on the GRO
2529 * offsets and fields in sk_buff.
2530 */
2531
2532__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2533
15e2396d
TH
2534static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2535{
b7fe10e5 2536 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2537}
2538
573e8fca
TH
2539static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2540 bool zero_okay,
2541 __sum16 check)
2542{
6edec0e6
TH
2543 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2544 skb_checksum_start_offset(skb) <
2545 skb_gro_offset(skb)) &&
15e2396d 2546 !skb_at_gro_remcsum_start(skb) &&
662880f4 2547 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2548 (!zero_okay || check));
2549}
2550
2551static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2552 __wsum psum)
2553{
2554 if (NAPI_GRO_CB(skb)->csum_valid &&
2555 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2556 return 0;
2557
2558 NAPI_GRO_CB(skb)->csum = psum;
2559
2560 return __skb_gro_checksum_complete(skb);
2561}
2562
573e8fca
TH
2563static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2564{
662880f4
TH
2565 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2566 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2567 NAPI_GRO_CB(skb)->csum_cnt--;
2568 } else {
2569 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2570 * verified a new top level checksum or an encapsulated one
2571 * during GRO. This saves work if we fallback to normal path.
2572 */
2573 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2574 }
2575}
2576
2577#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2578 compute_pseudo) \
2579({ \
2580 __sum16 __ret = 0; \
2581 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2582 __ret = __skb_gro_checksum_validate_complete(skb, \
2583 compute_pseudo(skb, proto)); \
219f1d79 2584 if (!__ret) \
573e8fca
TH
2585 skb_gro_incr_csum_unnecessary(skb); \
2586 __ret; \
2587})
2588
2589#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2590 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2591
2592#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2593 compute_pseudo) \
2594 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2595
2596#define skb_gro_checksum_simple_validate(skb) \
2597 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2598
d96535a1
TH
2599static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2600{
2601 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2602 !NAPI_GRO_CB(skb)->csum_valid);
2603}
2604
2605static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2606 __sum16 check, __wsum pseudo)
2607{
2608 NAPI_GRO_CB(skb)->csum = ~pseudo;
2609 NAPI_GRO_CB(skb)->csum_valid = 1;
2610}
2611
2612#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2613do { \
2614 if (__skb_gro_checksum_convert_check(skb)) \
2615 __skb_gro_checksum_convert(skb, check, \
2616 compute_pseudo(skb, proto)); \
2617} while (0)
2618
26c4f7da
TH
2619struct gro_remcsum {
2620 int offset;
2621 __wsum delta;
2622};
2623
2624static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2625{
846cd667 2626 grc->offset = 0;
26c4f7da
TH
2627 grc->delta = 0;
2628}
2629
b7fe10e5
TH
2630static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2631 unsigned int off, size_t hdrlen,
2632 int start, int offset,
2633 struct gro_remcsum *grc,
2634 bool nopartial)
dcdc8994
TH
2635{
2636 __wsum delta;
b7fe10e5 2637 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2638
2639 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2640
15e2396d 2641 if (!nopartial) {
b7fe10e5
TH
2642 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2643 return ptr;
2644 }
2645
2646 ptr = skb_gro_header_fast(skb, off);
2647 if (skb_gro_header_hard(skb, off + plen)) {
2648 ptr = skb_gro_header_slow(skb, off + plen, off);
2649 if (!ptr)
2650 return NULL;
15e2396d
TH
2651 }
2652
b7fe10e5
TH
2653 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2654 start, offset);
dcdc8994
TH
2655
2656 /* Adjust skb->csum since we changed the packet */
dcdc8994 2657 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2658
b7fe10e5 2659 grc->offset = off + hdrlen + offset;
26c4f7da 2660 grc->delta = delta;
b7fe10e5
TH
2661
2662 return ptr;
dcdc8994
TH
2663}
2664
26c4f7da
TH
2665static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2666 struct gro_remcsum *grc)
2667{
b7fe10e5
TH
2668 void *ptr;
2669 size_t plen = grc->offset + sizeof(u16);
2670
26c4f7da
TH
2671 if (!grc->delta)
2672 return;
2673
b7fe10e5
TH
2674 ptr = skb_gro_header_fast(skb, grc->offset);
2675 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2676 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2677 if (!ptr)
2678 return;
2679 }
2680
2681 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2682}
dcdc8994 2683
25393d3f
SK
2684#ifdef CONFIG_XFRM_OFFLOAD
2685static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2686{
2687 if (PTR_ERR(pp) != -EINPROGRESS)
2688 NAPI_GRO_CB(skb)->flush |= flush;
2689}
2690#else
5f114163
SK
2691static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2692{
2693 NAPI_GRO_CB(skb)->flush |= flush;
2694}
25393d3f 2695#endif
5f114163 2696
0c4e8581
SH
2697static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2698 unsigned short type,
3b04ddde 2699 const void *daddr, const void *saddr,
95c96174 2700 unsigned int len)
0c4e8581 2701{
f1ecfd5d 2702 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2703 return 0;
3b04ddde
SH
2704
2705 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2706}
2707
b95cce35
SH
2708static inline int dev_parse_header(const struct sk_buff *skb,
2709 unsigned char *haddr)
2710{
2711 const struct net_device *dev = skb->dev;
2712
1b83336b 2713 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2714 return 0;
3b04ddde 2715 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2716}
2717
2793a23a
WB
2718/* ll_header must have at least hard_header_len allocated */
2719static inline bool dev_validate_header(const struct net_device *dev,
2720 char *ll_header, int len)
2721{
2722 if (likely(len >= dev->hard_header_len))
2723 return true;
217e6fa2
WB
2724 if (len < dev->min_header_len)
2725 return false;
2793a23a
WB
2726
2727 if (capable(CAP_SYS_RAWIO)) {
2728 memset(ll_header + len, 0, dev->hard_header_len - len);
2729 return true;
2730 }
2731
2732 if (dev->header_ops && dev->header_ops->validate)
2733 return dev->header_ops->validate(ll_header, len);
2734
2735 return false;
2736}
2737
1da177e4 2738typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2739int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2740static inline int unregister_gifconf(unsigned int family)
2741{
2742 return register_gifconf(family, NULL);
2743}
2744
99bbc707 2745#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2746#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2747struct sd_flow_limit {
2748 u64 count;
2749 unsigned int num_buckets;
2750 unsigned int history_head;
2751 u16 history[FLOW_LIMIT_HISTORY];
2752 u8 buckets[];
2753};
2754
2755extern int netdev_flow_limit_table_len;
2756#endif /* CONFIG_NET_FLOW_LIMIT */
2757
1da177e4 2758/*
5e82b4b2 2759 * Incoming packets are placed on per-CPU queues
1da177e4 2760 */
d94d9fee 2761struct softnet_data {
1da177e4 2762 struct list_head poll_list;
6e7676c1 2763 struct sk_buff_head process_queue;
1da177e4 2764
dee42870 2765 /* stats */
cd7b5396
DM
2766 unsigned int processed;
2767 unsigned int time_squeeze;
cd7b5396 2768 unsigned int received_rps;
fd793d89 2769#ifdef CONFIG_RPS
88751275 2770 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2771#endif
2772#ifdef CONFIG_NET_FLOW_LIMIT
2773 struct sd_flow_limit __rcu *flow_limit;
2774#endif
2775 struct Qdisc *output_queue;
2776 struct Qdisc **output_queue_tailp;
2777 struct sk_buff *completion_queue;
88751275 2778
4cdb1e2e 2779#ifdef CONFIG_RPS
501e7ef5
ED
2780 /* input_queue_head should be written by cpu owning this struct,
2781 * and only read by other cpus. Worth using a cache line.
2782 */
2783 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2784
2785 /* Elements below can be accessed between CPUs for RPS/RFS */
0a9627f2 2786 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2787 struct softnet_data *rps_ipi_next;
2788 unsigned int cpu;
76cc8b13 2789 unsigned int input_queue_tail;
1e94d72f 2790#endif
95c96174 2791 unsigned int dropped;
0a9627f2 2792 struct sk_buff_head input_pkt_queue;
bea3348e 2793 struct napi_struct backlog;
99bbc707 2794
1da177e4
LT
2795};
2796
76cc8b13 2797static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2798{
2799#ifdef CONFIG_RPS
76cc8b13
TH
2800 sd->input_queue_head++;
2801#endif
2802}
2803
2804static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2805 unsigned int *qtail)
2806{
2807#ifdef CONFIG_RPS
2808 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2809#endif
2810}
2811
0a9627f2 2812DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2813
f629d208 2814void __netif_schedule(struct Qdisc *q);
46e5da40 2815void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2816
fd2ea0a7
DM
2817static inline void netif_tx_schedule_all(struct net_device *dev)
2818{
2819 unsigned int i;
2820
2821 for (i = 0; i < dev->num_tx_queues; i++)
2822 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2823}
2824
f9a7cbbf 2825static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
d29f749e 2826{
73466498 2827 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2828}
2829
bea3348e
SH
2830/**
2831 * netif_start_queue - allow transmit
2832 * @dev: network device
2833 *
2834 * Allow upper layers to call the device hard_start_xmit routine.
2835 */
1da177e4
LT
2836static inline void netif_start_queue(struct net_device *dev)
2837{
e8a0464c 2838 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2839}
2840
fd2ea0a7
DM
2841static inline void netif_tx_start_all_queues(struct net_device *dev)
2842{
2843 unsigned int i;
2844
2845 for (i = 0; i < dev->num_tx_queues; i++) {
2846 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2847 netif_tx_start_queue(txq);
2848 }
2849}
2850
46e5da40 2851void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2852
d29f749e
DJ
2853/**
2854 * netif_wake_queue - restart transmit
2855 * @dev: network device
2856 *
2857 * Allow upper layers to call the device hard_start_xmit routine.
2858 * Used for flow control when transmit resources are available.
2859 */
79d16385
DM
2860static inline void netif_wake_queue(struct net_device *dev)
2861{
e8a0464c 2862 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2863}
2864
fd2ea0a7
DM
2865static inline void netif_tx_wake_all_queues(struct net_device *dev)
2866{
2867 unsigned int i;
2868
2869 for (i = 0; i < dev->num_tx_queues; i++) {
2870 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2871 netif_tx_wake_queue(txq);
2872 }
2873}
2874
f9a7cbbf 2875static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
d29f749e 2876{
73466498 2877 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2878}
2879
bea3348e
SH
2880/**
2881 * netif_stop_queue - stop transmitted packets
2882 * @dev: network device
2883 *
2884 * Stop upper layers calling the device hard_start_xmit routine.
2885 * Used for flow control when transmit resources are unavailable.
2886 */
1da177e4
LT
2887static inline void netif_stop_queue(struct net_device *dev)
2888{
e8a0464c 2889 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2890}
2891
a2029240 2892void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2893
4d29515f 2894static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2895{
73466498 2896 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2897}
2898
bea3348e
SH
2899/**
2900 * netif_queue_stopped - test if transmit queue is flowblocked
2901 * @dev: network device
2902 *
2903 * Test if transmit queue on device is currently unable to send.
2904 */
4d29515f 2905static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2906{
e8a0464c 2907 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2908}
2909
4d29515f 2910static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2911{
73466498
TH
2912 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2913}
2914
8e2f1a63
DB
2915static inline bool
2916netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2917{
2918 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2919}
2920
8e2f1a63
DB
2921static inline bool
2922netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2923{
2924 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2925}
2926
53511453
ED
2927/**
2928 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2929 * @dev_queue: pointer to transmit queue
2930 *
2931 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
5e82b4b2 2932 * to give appropriate hint to the CPU.
53511453
ED
2933 */
2934static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2935{
2936#ifdef CONFIG_BQL
2937 prefetchw(&dev_queue->dql.num_queued);
2938#endif
2939}
2940
2941/**
2942 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2943 * @dev_queue: pointer to transmit queue
2944 *
2945 * BQL enabled drivers might use this helper in their TX completion path,
5e82b4b2 2946 * to give appropriate hint to the CPU.
53511453
ED
2947 */
2948static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2949{
2950#ifdef CONFIG_BQL
2951 prefetchw(&dev_queue->dql.limit);
2952#endif
2953}
2954
c5d67bd7
TH
2955static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2956 unsigned int bytes)
2957{
114cf580
TH
2958#ifdef CONFIG_BQL
2959 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2960
2961 if (likely(dql_avail(&dev_queue->dql) >= 0))
2962 return;
2963
2964 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2965
2966 /*
2967 * The XOFF flag must be set before checking the dql_avail below,
2968 * because in netdev_tx_completed_queue we update the dql_completed
2969 * before checking the XOFF flag.
2970 */
2971 smp_mb();
2972
2973 /* check again in case another CPU has just made room avail */
2974 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2975 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2976#endif
c5d67bd7
TH
2977}
2978
0042d0c8
FF
2979/**
2980 * netdev_sent_queue - report the number of bytes queued to hardware
2981 * @dev: network device
2982 * @bytes: number of bytes queued to the hardware device queue
2983 *
2984 * Report the number of bytes queued for sending/completion to the network
2985 * device hardware queue. @bytes should be a good approximation and should
2986 * exactly match netdev_completed_queue() @bytes
2987 */
c5d67bd7
TH
2988static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2989{
2990 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2991}
2992
2993static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2994 unsigned int pkts, unsigned int bytes)
c5d67bd7 2995{
114cf580 2996#ifdef CONFIG_BQL
b37c0fbe
AD
2997 if (unlikely(!bytes))
2998 return;
2999
3000 dql_completed(&dev_queue->dql, bytes);
3001
3002 /*
3003 * Without the memory barrier there is a small possiblity that
3004 * netdev_tx_sent_queue will miss the update and cause the queue to
3005 * be stopped forever
3006 */
3007 smp_mb();
3008
3009 if (dql_avail(&dev_queue->dql) < 0)
3010 return;
3011
3012 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3013 netif_schedule_queue(dev_queue);
114cf580 3014#endif
c5d67bd7
TH
3015}
3016
0042d0c8
FF
3017/**
3018 * netdev_completed_queue - report bytes and packets completed by device
3019 * @dev: network device
3020 * @pkts: actual number of packets sent over the medium
3021 * @bytes: actual number of bytes sent over the medium
3022 *
3023 * Report the number of bytes and packets transmitted by the network device
3024 * hardware queue over the physical medium, @bytes must exactly match the
3025 * @bytes amount passed to netdev_sent_queue()
3026 */
c5d67bd7 3027static inline void netdev_completed_queue(struct net_device *dev,
95c96174 3028 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
3029{
3030 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3031}
3032
3033static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3034{
114cf580 3035#ifdef CONFIG_BQL
5c490354 3036 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
3037 dql_reset(&q->dql);
3038#endif
c5d67bd7
TH
3039}
3040
0042d0c8
FF
3041/**
3042 * netdev_reset_queue - reset the packets and bytes count of a network device
3043 * @dev_queue: network device
3044 *
3045 * Reset the bytes and packet count of a network device and clear the
3046 * software flow control OFF bit for this network device
3047 */
c5d67bd7
TH
3048static inline void netdev_reset_queue(struct net_device *dev_queue)
3049{
3050 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
3051}
3052
b9507bda
DB
3053/**
3054 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3055 * @dev: network device
3056 * @queue_index: given tx queue index
3057 *
3058 * Returns 0 if given tx queue index >= number of device tx queues,
3059 * otherwise returns the originally passed tx queue index.
3060 */
3061static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3062{
3063 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3064 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3065 dev->name, queue_index,
3066 dev->real_num_tx_queues);
3067 return 0;
3068 }
3069
3070 return queue_index;
3071}
3072
bea3348e
SH
3073/**
3074 * netif_running - test if up
3075 * @dev: network device
3076 *
3077 * Test if the device has been brought up.
3078 */
4d29515f 3079static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
3080{
3081 return test_bit(__LINK_STATE_START, &dev->state);
3082}
3083
f25f4e44 3084/*
5e82b4b2 3085 * Routines to manage the subqueues on a device. We only need start,
f25f4e44
PWJ
3086 * stop, and a check if it's stopped. All other device management is
3087 * done at the overall netdevice level.
3088 * Also test the device if we're multiqueue.
3089 */
bea3348e
SH
3090
3091/**
3092 * netif_start_subqueue - allow sending packets on subqueue
3093 * @dev: network device
3094 * @queue_index: sub queue index
3095 *
3096 * Start individual transmit queue of a device with multiple transmit queues.
3097 */
f25f4e44
PWJ
3098static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3099{
fd2ea0a7 3100 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3101
3102 netif_tx_start_queue(txq);
f25f4e44
PWJ
3103}
3104
bea3348e
SH
3105/**
3106 * netif_stop_subqueue - stop sending packets on subqueue
3107 * @dev: network device
3108 * @queue_index: sub queue index
3109 *
3110 * Stop individual transmit queue of a device with multiple transmit queues.
3111 */
f25f4e44
PWJ
3112static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3113{
fd2ea0a7 3114 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 3115 netif_tx_stop_queue(txq);
f25f4e44
PWJ
3116}
3117
bea3348e
SH
3118/**
3119 * netif_subqueue_stopped - test status of subqueue
3120 * @dev: network device
3121 * @queue_index: sub queue index
3122 *
3123 * Check individual transmit queue of a device with multiple transmit queues.
3124 */
4d29515f
DM
3125static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3126 u16 queue_index)
f25f4e44 3127{
fd2ea0a7 3128 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
3129
3130 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
3131}
3132
4d29515f
DM
3133static inline bool netif_subqueue_stopped(const struct net_device *dev,
3134 struct sk_buff *skb)
668f895a
PE
3135{
3136 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3137}
bea3348e 3138
738b35cc
FF
3139/**
3140 * netif_wake_subqueue - allow sending packets on subqueue
3141 * @dev: network device
3142 * @queue_index: sub queue index
3143 *
3144 * Resume individual transmit queue of a device with multiple transmit queues.
3145 */
3146static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3147{
3148 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3149
3150 netif_tx_wake_queue(txq);
3151}
f25f4e44 3152
537c00de 3153#ifdef CONFIG_XPS
53af53ae 3154int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 3155 u16 index);
537c00de
AD
3156#else
3157static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 3158 const struct cpumask *mask,
537c00de
AD
3159 u16 index)
3160{
3161 return 0;
3162}
3163#endif
3164
5605c762
JP
3165u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3166 unsigned int num_tx_queues);
3167
a3d22a68
VZ
3168/*
3169 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3170 * as a distribution range limit for the returned value.
3171 */
3172static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 3173 struct sk_buff *skb)
a3d22a68
VZ
3174{
3175 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3176}
3177
bea3348e
SH
3178/**
3179 * netif_is_multiqueue - test if device has multiple transmit queues
3180 * @dev: network device
3181 *
3182 * Check if device has multiple transmit queues
bea3348e 3183 */
4d29515f 3184static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 3185{
a02cec21 3186 return dev->num_tx_queues > 1;
f25f4e44 3187}
1da177e4 3188
f629d208 3189int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 3190
a953be53 3191#ifdef CONFIG_SYSFS
f629d208 3192int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
3193#else
3194static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3195 unsigned int rxq)
3196{
3197 return 0;
3198}
3199#endif
3200
a953be53
MD
3201#ifdef CONFIG_SYSFS
3202static inline unsigned int get_netdev_rx_queue_index(
3203 struct netdev_rx_queue *queue)
3204{
3205 struct net_device *dev = queue->dev;
3206 int index = queue - dev->_rx;
3207
3208 BUG_ON(index >= dev->num_rx_queues);
3209 return index;
3210}
3211#endif
3212
16917b87 3213#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 3214int netif_get_num_default_rss_queues(void);
16917b87 3215
e6247027
ED
3216enum skb_free_reason {
3217 SKB_REASON_CONSUMED,
3218 SKB_REASON_DROPPED,
3219};
3220
3221void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3222void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 3223
e6247027
ED
3224/*
3225 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3226 * interrupt context or with hardware interrupts being disabled.
3227 * (in_irq() || irqs_disabled())
3228 *
3229 * We provide four helpers that can be used in following contexts :
3230 *
3231 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3232 * replacing kfree_skb(skb)
3233 *
3234 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3235 * Typically used in place of consume_skb(skb) in TX completion path
3236 *
3237 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3238 * replacing kfree_skb(skb)
3239 *
3240 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3241 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 3242 */
e6247027
ED
3243static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3244{
3245 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3246}
3247
3248static inline void dev_consume_skb_irq(struct sk_buff *skb)
3249{
3250 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3251}
3252
3253static inline void dev_kfree_skb_any(struct sk_buff *skb)
3254{
3255 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3256}
3257
3258static inline void dev_consume_skb_any(struct sk_buff *skb)
3259{
3260 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3261}
1da177e4 3262
f629d208
JP
3263int netif_rx(struct sk_buff *skb);
3264int netif_rx_ni(struct sk_buff *skb);
04eb4489 3265int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3266gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3267void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3268struct sk_buff *napi_get_frags(struct napi_struct *napi);
3269gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3270struct packet_offload *gro_find_receive_by_type(__be16 type);
3271struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3272
3273static inline void napi_free_frags(struct napi_struct *napi)
3274{
3275 kfree_skb(napi->skb);
3276 napi->skb = NULL;
3277}
3278
24b27fc4 3279bool netdev_is_rx_handler_busy(struct net_device *dev);
f629d208
JP
3280int netdev_rx_handler_register(struct net_device *dev,
3281 rx_handler_func_t *rx_handler,
3282 void *rx_handler_data);
3283void netdev_rx_handler_unregister(struct net_device *dev);
3284
3285bool dev_valid_name(const char *name);
3286int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3287int dev_ethtool(struct net *net, struct ifreq *);
3288unsigned int dev_get_flags(const struct net_device *);
3289int __dev_change_flags(struct net_device *, unsigned int flags);
3290int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3291void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3292 unsigned int gchanges);
f629d208
JP
3293int dev_change_name(struct net_device *, const char *);
3294int dev_set_alias(struct net_device *, const char *, size_t);
3295int dev_change_net_namespace(struct net_device *, struct net *, const char *);
f51048c3 3296int __dev_set_mtu(struct net_device *, int);
f629d208
JP
3297int dev_set_mtu(struct net_device *, int);
3298void dev_set_group(struct net_device *, int);
3299int dev_set_mac_address(struct net_device *, struct sockaddr *);
3300int dev_change_carrier(struct net_device *, bool new_carrier);
3301int dev_get_phys_port_id(struct net_device *dev,
02637fce 3302 struct netdev_phys_item_id *ppid);
db24a904
DA
3303int dev_get_phys_port_name(struct net_device *dev,
3304 char *name, size_t len);
d746d707 3305int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3306struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3307struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3308 struct netdev_queue *txq, int *ret);
d67b9cd2
DB
3309
3310typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3311int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3312 int fd, u32 flags);
ce158e58 3313u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
d67b9cd2 3314
a0265d28 3315int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3316int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f4b05d27
NA
3317bool is_skb_forwardable(const struct net_device *dev,
3318 const struct sk_buff *skb);
1da177e4 3319
4e3264d2
MKL
3320static __always_inline int ____dev_forward_skb(struct net_device *dev,
3321 struct sk_buff *skb)
3322{
3323 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3324 unlikely(!is_skb_forwardable(dev, skb))) {
3325 atomic_long_inc(&dev->rx_dropped);
3326 kfree_skb(skb);
3327 return NET_RX_DROP;
3328 }
3329
3330 skb_scrub_packet(skb, true);
3331 skb->priority = 0;
3332 return 0;
3333}
3334
74b20582
DA
3335void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3336
20380731 3337extern int netdev_budget;
7acf8a1e 3338extern unsigned int netdev_budget_usecs;
1da177e4
LT
3339
3340/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3341void netdev_run_todo(void);
1da177e4 3342
bea3348e
SH
3343/**
3344 * dev_put - release reference to device
3345 * @dev: network device
3346 *
9ef4429b 3347 * Release reference to device to allow it to be freed.
bea3348e 3348 */
1da177e4
LT
3349static inline void dev_put(struct net_device *dev)
3350{
933393f5 3351 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3352}
3353
bea3348e
SH
3354/**
3355 * dev_hold - get reference to device
3356 * @dev: network device
3357 *
9ef4429b 3358 * Hold reference to device to keep it from being freed.
bea3348e 3359 */
15333061
SH
3360static inline void dev_hold(struct net_device *dev)
3361{
933393f5 3362 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3363}
1da177e4
LT
3364
3365/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3366 * and _off may be called from IRQ context, but it is caller
3367 * who is responsible for serialization of these calls.
b00055aa
SR
3368 *
3369 * The name carrier is inappropriate, these functions should really be
3370 * called netif_lowerlayer_*() because they represent the state of any
3371 * kind of lower layer not just hardware media.
1da177e4
LT
3372 */
3373
f629d208
JP
3374void linkwatch_init_dev(struct net_device *dev);
3375void linkwatch_fire_event(struct net_device *dev);
3376void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3377
bea3348e
SH
3378/**
3379 * netif_carrier_ok - test if carrier present
3380 * @dev: network device
3381 *
3382 * Check if carrier is present on device
3383 */
4d29515f 3384static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3385{
3386 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3387}
3388
f629d208 3389unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3390
f629d208 3391void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3392
f629d208 3393void netif_carrier_on(struct net_device *dev);
1da177e4 3394
f629d208 3395void netif_carrier_off(struct net_device *dev);
1da177e4 3396
bea3348e
SH
3397/**
3398 * netif_dormant_on - mark device as dormant.
3399 * @dev: network device
3400 *
3401 * Mark device as dormant (as per RFC2863).
3402 *
3403 * The dormant state indicates that the relevant interface is not
3404 * actually in a condition to pass packets (i.e., it is not 'up') but is
3405 * in a "pending" state, waiting for some external event. For "on-
3406 * demand" interfaces, this new state identifies the situation where the
3407 * interface is waiting for events to place it in the up state.
bea3348e 3408 */
b00055aa
SR
3409static inline void netif_dormant_on(struct net_device *dev)
3410{
3411 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3412 linkwatch_fire_event(dev);
3413}
3414
bea3348e
SH
3415/**
3416 * netif_dormant_off - set device as not dormant.
3417 * @dev: network device
3418 *
3419 * Device is not in dormant state.
3420 */
b00055aa
SR
3421static inline void netif_dormant_off(struct net_device *dev)
3422{
3423 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3424 linkwatch_fire_event(dev);
3425}
3426
bea3348e 3427/**
8ecbc40a 3428 * netif_dormant - test if device is dormant
bea3348e
SH
3429 * @dev: network device
3430 *
8ecbc40a 3431 * Check if device is dormant.
bea3348e 3432 */
4d29515f 3433static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3434{
3435 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3436}
3437
3438
bea3348e
SH
3439/**
3440 * netif_oper_up - test if device is operational
3441 * @dev: network device
3442 *
3443 * Check if carrier is operational
3444 */
4d29515f 3445static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3446{
b00055aa
SR
3447 return (dev->operstate == IF_OPER_UP ||
3448 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3449}
3450
bea3348e
SH
3451/**
3452 * netif_device_present - is device available or removed
3453 * @dev: network device
3454 *
3455 * Check if device has not been removed from system.
3456 */
4d29515f 3457static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3458{
3459 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3460}
3461
f629d208 3462void netif_device_detach(struct net_device *dev);
1da177e4 3463
f629d208 3464void netif_device_attach(struct net_device *dev);
1da177e4
LT
3465
3466/*
3467 * Network interface message level settings
3468 */
1da177e4
LT
3469
3470enum {
3471 NETIF_MSG_DRV = 0x0001,
3472 NETIF_MSG_PROBE = 0x0002,
3473 NETIF_MSG_LINK = 0x0004,
3474 NETIF_MSG_TIMER = 0x0008,
3475 NETIF_MSG_IFDOWN = 0x0010,
3476 NETIF_MSG_IFUP = 0x0020,
3477 NETIF_MSG_RX_ERR = 0x0040,
3478 NETIF_MSG_TX_ERR = 0x0080,
3479 NETIF_MSG_TX_QUEUED = 0x0100,
3480 NETIF_MSG_INTR = 0x0200,
3481 NETIF_MSG_TX_DONE = 0x0400,
3482 NETIF_MSG_RX_STATUS = 0x0800,
3483 NETIF_MSG_PKTDATA = 0x1000,
3484 NETIF_MSG_HW = 0x2000,
3485 NETIF_MSG_WOL = 0x4000,
3486};
3487
3488#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3489#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3490#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3491#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3492#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3493#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3494#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3495#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3496#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3497#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3498#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3499#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3500#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3501#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3502#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3503
3504static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3505{
3506 /* use default */
3507 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3508 return default_msg_enable_bits;
3509 if (debug_value == 0) /* no output */
3510 return 0;
3511 /* set low N bits */
3512 return (1 << debug_value) - 1;
3513}
3514
c773e847 3515static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3516{
c773e847
DM
3517 spin_lock(&txq->_xmit_lock);
3518 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3519}
3520
5a717f4f
MT
3521static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3522{
3523 __acquire(&txq->_xmit_lock);
3524 return true;
3525}
3526
3527static inline void __netif_tx_release(struct netdev_queue *txq)
3528{
3529 __release(&txq->_xmit_lock);
3530}
3531
fd2ea0a7
DM
3532static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3533{
3534 spin_lock_bh(&txq->_xmit_lock);
3535 txq->xmit_lock_owner = smp_processor_id();
3536}
3537
4d29515f 3538static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3539{
4d29515f 3540 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3541 if (likely(ok))
3542 txq->xmit_lock_owner = smp_processor_id();
3543 return ok;
3544}
3545
3546static inline void __netif_tx_unlock(struct netdev_queue *txq)
3547{
3548 txq->xmit_lock_owner = -1;
3549 spin_unlock(&txq->_xmit_lock);
3550}
3551
3552static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3553{
3554 txq->xmit_lock_owner = -1;
3555 spin_unlock_bh(&txq->_xmit_lock);
3556}
3557
08baf561
ED
3558static inline void txq_trans_update(struct netdev_queue *txq)
3559{
3560 if (txq->xmit_lock_owner != -1)
3561 txq->trans_start = jiffies;
3562}
3563
ba162f8e
FW
3564/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3565static inline void netif_trans_update(struct net_device *dev)
3566{
9b36627a
FW
3567 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3568
3569 if (txq->trans_start != jiffies)
3570 txq->trans_start = jiffies;
ba162f8e
FW
3571}
3572
d29f749e
DJ
3573/**
3574 * netif_tx_lock - grab network device transmit lock
3575 * @dev: network device
d29f749e
DJ
3576 *
3577 * Get network device transmit lock
3578 */
22dd7495
JHS
3579static inline void netif_tx_lock(struct net_device *dev)
3580{
e8a0464c 3581 unsigned int i;
c3f26a26 3582 int cpu;
c773e847 3583
c3f26a26
DM
3584 spin_lock(&dev->tx_global_lock);
3585 cpu = smp_processor_id();
e8a0464c
DM
3586 for (i = 0; i < dev->num_tx_queues; i++) {
3587 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3588
3589 /* We are the only thread of execution doing a
3590 * freeze, but we have to grab the _xmit_lock in
3591 * order to synchronize with threads which are in
3592 * the ->hard_start_xmit() handler and already
3593 * checked the frozen bit.
3594 */
e8a0464c 3595 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3596 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3597 __netif_tx_unlock(txq);
e8a0464c 3598 }
932ff279
HX
3599}
3600
3601static inline void netif_tx_lock_bh(struct net_device *dev)
3602{
e8a0464c
DM
3603 local_bh_disable();
3604 netif_tx_lock(dev);
932ff279
HX
3605}
3606
932ff279
HX
3607static inline void netif_tx_unlock(struct net_device *dev)
3608{
e8a0464c
DM
3609 unsigned int i;
3610
3611 for (i = 0; i < dev->num_tx_queues; i++) {
3612 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3613
c3f26a26
DM
3614 /* No need to grab the _xmit_lock here. If the
3615 * queue is not stopped for another reason, we
3616 * force a schedule.
3617 */
3618 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3619 netif_schedule_queue(txq);
c3f26a26
DM
3620 }
3621 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3622}
3623
3624static inline void netif_tx_unlock_bh(struct net_device *dev)
3625{
e8a0464c
DM
3626 netif_tx_unlock(dev);
3627 local_bh_enable();
932ff279
HX
3628}
3629
c773e847 3630#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3631 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3632 __netif_tx_lock(txq, cpu); \
5a717f4f
MT
3633 } else { \
3634 __netif_tx_acquire(txq); \
22dd7495
JHS
3635 } \
3636}
3637
5efeac44
EB
3638#define HARD_TX_TRYLOCK(dev, txq) \
3639 (((dev->features & NETIF_F_LLTX) == 0) ? \
3640 __netif_tx_trylock(txq) : \
5a717f4f 3641 __netif_tx_acquire(txq))
5efeac44 3642
c773e847 3643#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3644 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3645 __netif_tx_unlock(txq); \
5a717f4f
MT
3646 } else { \
3647 __netif_tx_release(txq); \
22dd7495
JHS
3648 } \
3649}
3650
1da177e4
LT
3651static inline void netif_tx_disable(struct net_device *dev)
3652{
fd2ea0a7 3653 unsigned int i;
c3f26a26 3654 int cpu;
fd2ea0a7 3655
c3f26a26
DM
3656 local_bh_disable();
3657 cpu = smp_processor_id();
fd2ea0a7
DM
3658 for (i = 0; i < dev->num_tx_queues; i++) {
3659 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3660
3661 __netif_tx_lock(txq, cpu);
fd2ea0a7 3662 netif_tx_stop_queue(txq);
c3f26a26 3663 __netif_tx_unlock(txq);
fd2ea0a7 3664 }
c3f26a26 3665 local_bh_enable();
1da177e4
LT
3666}
3667
e308a5d8
DM
3668static inline void netif_addr_lock(struct net_device *dev)
3669{
3670 spin_lock(&dev->addr_list_lock);
3671}
3672
2429f7ac
JP
3673static inline void netif_addr_lock_nested(struct net_device *dev)
3674{
25175ba5
VY
3675 int subclass = SINGLE_DEPTH_NESTING;
3676
3677 if (dev->netdev_ops->ndo_get_lock_subclass)
3678 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3679
3680 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3681}
3682
e308a5d8
DM
3683static inline void netif_addr_lock_bh(struct net_device *dev)
3684{
3685 spin_lock_bh(&dev->addr_list_lock);
3686}
3687
3688static inline void netif_addr_unlock(struct net_device *dev)
3689{
3690 spin_unlock(&dev->addr_list_lock);
3691}
3692
3693static inline void netif_addr_unlock_bh(struct net_device *dev)
3694{
3695 spin_unlock_bh(&dev->addr_list_lock);
3696}
3697
f001fde5 3698/*
31278e71 3699 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3700 * rcu_read_lock held.
3701 */
3702#define for_each_dev_addr(dev, ha) \
31278e71 3703 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3704
1da177e4
LT
3705/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3706
f629d208 3707void ether_setup(struct net_device *dev);
1da177e4
LT
3708
3709/* Support for loadable net-drivers */
f629d208 3710struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3711 unsigned char name_assign_type,
f629d208
JP
3712 void (*setup)(struct net_device *),
3713 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3714#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3715 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3716
c835a677
TG
3717#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3718 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3719 count)
36909ea4 3720
f629d208
JP
3721int register_netdev(struct net_device *dev);
3722void unregister_netdev(struct net_device *dev);
f001fde5 3723
22bedad3 3724/* General hardware address lists handling functions */
f629d208
JP
3725int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3726 struct netdev_hw_addr_list *from_list, int addr_len);
3727void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3728 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3729int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3730 struct net_device *dev,
3731 int (*sync)(struct net_device *, const unsigned char *),
3732 int (*unsync)(struct net_device *,
3733 const unsigned char *));
3734void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3735 struct net_device *dev,
3736 int (*unsync)(struct net_device *,
3737 const unsigned char *));
f629d208 3738void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3739
f001fde5 3740/* Functions used for device addresses handling */
f629d208
JP
3741int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3742 unsigned char addr_type);
3743int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3744 unsigned char addr_type);
f629d208
JP
3745void dev_addr_flush(struct net_device *dev);
3746int dev_addr_init(struct net_device *dev);
a748ee24
JP
3747
3748/* Functions used for unicast addresses handling */
f629d208
JP
3749int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3750int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3751int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3752int dev_uc_sync(struct net_device *to, struct net_device *from);
3753int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3754void dev_uc_unsync(struct net_device *to, struct net_device *from);
3755void dev_uc_flush(struct net_device *dev);
3756void dev_uc_init(struct net_device *dev);
f001fde5 3757
670e5b8e
AD
3758/**
3759 * __dev_uc_sync - Synchonize device's unicast list
3760 * @dev: device to sync
3761 * @sync: function to call if address should be added
3762 * @unsync: function to call if address should be removed
3763 *
3764 * Add newly added addresses to the interface, and release
3765 * addresses that have been deleted.
5e82b4b2 3766 */
670e5b8e
AD
3767static inline int __dev_uc_sync(struct net_device *dev,
3768 int (*sync)(struct net_device *,
3769 const unsigned char *),
3770 int (*unsync)(struct net_device *,
3771 const unsigned char *))
3772{
3773 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3774}
3775
3776/**
e793c0f7 3777 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3778 * @dev: device to sync
3779 * @unsync: function to call if address should be removed
3780 *
3781 * Remove all addresses that were added to the device by dev_uc_sync().
5e82b4b2 3782 */
670e5b8e
AD
3783static inline void __dev_uc_unsync(struct net_device *dev,
3784 int (*unsync)(struct net_device *,
3785 const unsigned char *))
3786{
3787 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3788}
3789
22bedad3 3790/* Functions used for multicast addresses handling */
f629d208
JP
3791int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3792int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3793int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3794int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3795int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3796int dev_mc_sync(struct net_device *to, struct net_device *from);
3797int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3798void dev_mc_unsync(struct net_device *to, struct net_device *from);
3799void dev_mc_flush(struct net_device *dev);
3800void dev_mc_init(struct net_device *dev);
f001fde5 3801
670e5b8e
AD
3802/**
3803 * __dev_mc_sync - Synchonize device's multicast list
3804 * @dev: device to sync
3805 * @sync: function to call if address should be added
3806 * @unsync: function to call if address should be removed
3807 *
3808 * Add newly added addresses to the interface, and release
3809 * addresses that have been deleted.
5e82b4b2 3810 */
670e5b8e
AD
3811static inline int __dev_mc_sync(struct net_device *dev,
3812 int (*sync)(struct net_device *,
3813 const unsigned char *),
3814 int (*unsync)(struct net_device *,
3815 const unsigned char *))
3816{
3817 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3818}
3819
3820/**
e793c0f7 3821 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3822 * @dev: device to sync
3823 * @unsync: function to call if address should be removed
3824 *
3825 * Remove all addresses that were added to the device by dev_mc_sync().
5e82b4b2 3826 */
670e5b8e
AD
3827static inline void __dev_mc_unsync(struct net_device *dev,
3828 int (*unsync)(struct net_device *,
3829 const unsigned char *))
3830{
3831 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3832}
3833
4417da66 3834/* Functions used for secondary unicast and multicast support */
f629d208
JP
3835void dev_set_rx_mode(struct net_device *dev);
3836void __dev_set_rx_mode(struct net_device *dev);
3837int dev_set_promiscuity(struct net_device *dev, int inc);
3838int dev_set_allmulti(struct net_device *dev, int inc);
3839void netdev_state_change(struct net_device *dev);
3840void netdev_notify_peers(struct net_device *dev);
3841void netdev_features_change(struct net_device *dev);
1da177e4 3842/* Load a device via the kmod */
f629d208
JP
3843void dev_load(struct net *net, const char *name);
3844struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3845 struct rtnl_link_stats64 *storage);
3846void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3847 const struct net_device_stats *netdev_stats);
eeda3fd6 3848
1da177e4 3849extern int netdev_max_backlog;
3b098e2d 3850extern int netdev_tstamp_prequeue;
1da177e4 3851extern int weight_p;
3d48b53f
MT
3852extern int dev_weight_rx_bias;
3853extern int dev_weight_tx_bias;
3854extern int dev_rx_weight;
3855extern int dev_tx_weight;
9ff162a8 3856
f629d208 3857bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3858struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3859 struct list_head **iter);
f629d208
JP
3860struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3861 struct list_head **iter);
8b5be856 3862
44a40855
VY
3863/* iterate through upper list, must be called under RCU read lock */
3864#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3865 for (iter = &(dev)->adj_list.upper, \
3866 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3867 updev; \
3868 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3869
1a3f060c
DA
3870int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3871 int (*fn)(struct net_device *upper_dev,
3872 void *data),
3873 void *data);
3874
3875bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3876 struct net_device *upper_dev);
3877
f629d208
JP
3878void *netdev_lower_get_next_private(struct net_device *dev,
3879 struct list_head **iter);
3880void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3881 struct list_head **iter);
31088a11
VF
3882
3883#define netdev_for_each_lower_private(dev, priv, iter) \
3884 for (iter = (dev)->adj_list.lower.next, \
3885 priv = netdev_lower_get_next_private(dev, &(iter)); \
3886 priv; \
3887 priv = netdev_lower_get_next_private(dev, &(iter)))
3888
3889#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3890 for (iter = &(dev)->adj_list.lower, \
3891 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3892 priv; \
3893 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3894
4085ebe8
VY
3895void *netdev_lower_get_next(struct net_device *dev,
3896 struct list_head **iter);
7ce856aa 3897
4085ebe8 3898#define netdev_for_each_lower_dev(dev, ldev, iter) \
cfdd28be 3899 for (iter = (dev)->adj_list.lower.next, \
4085ebe8
VY
3900 ldev = netdev_lower_get_next(dev, &(iter)); \
3901 ldev; \
3902 ldev = netdev_lower_get_next(dev, &(iter)))
3903
7ce856aa
JP
3904struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3905 struct list_head **iter);
3906struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3907 struct list_head **iter);
3908
1a3f060c
DA
3909int netdev_walk_all_lower_dev(struct net_device *dev,
3910 int (*fn)(struct net_device *lower_dev,
3911 void *data),
3912 void *data);
3913int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3914 int (*fn)(struct net_device *lower_dev,
3915 void *data),
3916 void *data);
3917
f629d208 3918void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3919void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3920struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3921struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3922int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3923int netdev_master_upper_dev_link(struct net_device *dev,
6dffb044 3924 struct net_device *upper_dev,
29bf24af 3925 void *upper_priv, void *upper_info);
f629d208
JP
3926void netdev_upper_dev_unlink(struct net_device *dev,
3927 struct net_device *upper_dev);
5bb025fa 3928void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3929void *netdev_lower_dev_get_private(struct net_device *dev,
3930 struct net_device *lower_dev);
04d48266
JP
3931void netdev_lower_state_changed(struct net_device *lower_dev,
3932 void *lower_state_info);
960fb622
ED
3933
3934/* RSS keys are 40 or 52 bytes long */
3935#define NETDEV_RSS_KEY_LEN 52
ba905f5e 3936extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
960fb622
ED
3937void netdev_rss_key_fill(void *buffer, size_t len);
3938
952fcfd0 3939int dev_get_nest_level(struct net_device *dev);
f629d208 3940int skb_checksum_help(struct sk_buff *skb);
b72b5bf6 3941int skb_crc32c_csum_help(struct sk_buff *skb);
43c26a1a
DC
3942int skb_csum_hwoffload_help(struct sk_buff *skb,
3943 const netdev_features_t features);
3944
f629d208
JP
3945struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3946 netdev_features_t features, bool tx_path);
3947struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3948 netdev_features_t features);
12b0004d 3949
61bd3857
MS
3950struct netdev_bonding_info {
3951 ifslave slave;
3952 ifbond master;
3953};
3954
3955struct netdev_notifier_bonding_info {
3956 struct netdev_notifier_info info; /* must be first */
3957 struct netdev_bonding_info bonding_info;
3958};
3959
3960void netdev_bonding_info_change(struct net_device *dev,
3961 struct netdev_bonding_info *bonding_info);
3962
12b0004d
CW
3963static inline
3964struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3965{
3966 return __skb_gso_segment(skb, features, true);
3967}
53d6471c 3968__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3969
3970static inline bool can_checksum_protocol(netdev_features_t features,
3971 __be16 protocol)
3972{
c8cd0989
TH
3973 if (protocol == htons(ETH_P_FCOE))
3974 return !!(features & NETIF_F_FCOE_CRC);
3975
3976 /* Assume this is an IP checksum (not SCTP CRC) */
3977
3978 if (features & NETIF_F_HW_CSUM) {
3979 /* Can checksum everything */
3980 return true;
3981 }
3982
3983 switch (protocol) {
3984 case htons(ETH_P_IP):
3985 return !!(features & NETIF_F_IP_CSUM);
3986 case htons(ETH_P_IPV6):
3987 return !!(features & NETIF_F_IPV6_CSUM);
3988 default:
3989 return false;
3990 }
ec5f0615 3991}
12b0004d 3992
fb286bb2 3993#ifdef CONFIG_BUG
f629d208 3994void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3995#else
3996static inline void netdev_rx_csum_fault(struct net_device *dev)
3997{
3998}
3999#endif
1da177e4 4000/* rx skb timestamps */
f629d208
JP
4001void net_enable_timestamp(void);
4002void net_disable_timestamp(void);
1da177e4 4003
20380731 4004#ifdef CONFIG_PROC_FS
f629d208 4005int __init dev_proc_init(void);
900ff8c6
CW
4006#else
4007#define dev_proc_init() 0
20380731
ACM
4008#endif
4009
4798248e 4010static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
4011 struct sk_buff *skb, struct net_device *dev,
4012 bool more)
4798248e 4013{
fa2dbdc2 4014 skb->xmit_more = more ? 1 : 0;
0b725a2c 4015 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
4016}
4017
10b3ad8c 4018static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 4019 struct netdev_queue *txq, bool more)
4798248e
DM
4020{
4021 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 4022 int rc;
4798248e 4023
fa2dbdc2 4024 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
4025 if (rc == NETDEV_TX_OK)
4026 txq_trans_update(txq);
4027
4028 return rc;
4798248e
DM
4029}
4030
42a2d923
LT
4031int netdev_class_create_file_ns(struct class_attribute *class_attr,
4032 const void *ns);
4033void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4034 const void *ns);
58292cbe
TH
4035
4036static inline int netdev_class_create_file(struct class_attribute *class_attr)
4037{
4038 return netdev_class_create_file_ns(class_attr, NULL);
4039}
4040
4041static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4042{
4043 netdev_class_remove_file_ns(class_attr, NULL);
4044}
b8a9787e 4045
04600794
JB
4046extern struct kobj_ns_type_operations net_ns_type_operations;
4047
f629d208 4048const char *netdev_drivername(const struct net_device *dev);
6579e57b 4049
f629d208 4050void linkwatch_run_queue(void);
20380731 4051
da08143b
MK
4052static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4053 netdev_features_t f2)
4054{
c8cd0989
TH
4055 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4056 if (f1 & NETIF_F_HW_CSUM)
b6a0e72a 4057 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4058 else
b6a0e72a 4059 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
c8cd0989 4060 }
da08143b 4061
c8cd0989 4062 return f1 & f2;
da08143b
MK
4063}
4064
c8f44aff
MM
4065static inline netdev_features_t netdev_get_wanted_features(
4066 struct net_device *dev)
5455c699
MM
4067{
4068 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4069}
c8f44aff
MM
4070netdev_features_t netdev_increment_features(netdev_features_t all,
4071 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
4072
4073/* Allow TSO being used on stacked device :
4074 * Performing the GSO segmentation before last device
4075 * is a performance improvement.
4076 */
4077static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4078 netdev_features_t mask)
4079{
4080 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4081}
4082
6cb6a27c 4083int __netdev_update_features(struct net_device *dev);
5455c699 4084void netdev_update_features(struct net_device *dev);
afe12cc8 4085void netdev_change_features(struct net_device *dev);
7f353bf2 4086
fc4a7489
PM
4087void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4088 struct net_device *dev);
4089
e38f3025
TM
4090netdev_features_t passthru_features_check(struct sk_buff *skb,
4091 struct net_device *dev,
4092 netdev_features_t features);
c1e756bf 4093netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 4094
4d29515f 4095static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 4096{
7b748340 4097 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
4098
4099 /* check flags correspondence */
4100 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
0345e186
MM
4101 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4102 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
cbc53e08 4103 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
0345e186
MM
4104 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4105 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4106 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4107 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
7e13318d
TH
4108 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4109 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4b28252c
TH
4110 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4111 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
802ab55a 4112 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
e585f236 4113 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
90017acc 4114 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
c7ef8f0c 4115 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
0345e186 4116
d6b4991a 4117 return (features & feature) == feature;
576a30eb
HX
4118}
4119
4d29515f 4120static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 4121{
278b2513 4122 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 4123 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
4124}
4125
8b86a61d 4126static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 4127 netdev_features_t features)
7967168c 4128{
fc741216 4129 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
4130 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4131 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
4132}
4133
82cc1a7a
PWJ
4134static inline void netif_set_gso_max_size(struct net_device *dev,
4135 unsigned int size)
4136{
4137 dev->gso_max_size = size;
4138}
4139
7a7ffbab
WCC
4140static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4141 int pulled_hlen, u16 mac_offset,
4142 int mac_len)
4143{
4144 skb->protocol = protocol;
4145 skb->encapsulation = 1;
4146 skb_push(skb, pulled_hlen);
4147 skb_reset_transport_header(skb);
4148 skb->mac_header = mac_offset;
4149 skb->network_header = skb->mac_header + mac_len;
4150 skb->mac_len = mac_len;
4151}
4152
3c175784
SD
4153static inline bool netif_is_macsec(const struct net_device *dev)
4154{
4155 return dev->priv_flags & IFF_MACSEC;
4156}
4157
b618aaa9 4158static inline bool netif_is_macvlan(const struct net_device *dev)
a6cc0cfa
JF
4159{
4160 return dev->priv_flags & IFF_MACVLAN;
4161}
4162
b618aaa9 4163static inline bool netif_is_macvlan_port(const struct net_device *dev)
2f33e7d5
MB
4164{
4165 return dev->priv_flags & IFF_MACVLAN_PORT;
4166}
4167
b618aaa9 4168static inline bool netif_is_ipvlan(const struct net_device *dev)
5933fea7
MB
4169{
4170 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4171}
4172
b618aaa9 4173static inline bool netif_is_ipvlan_port(const struct net_device *dev)
5933fea7
MB
4174{
4175 return dev->priv_flags & IFF_IPVLAN_MASTER;
4176}
4177
b618aaa9 4178static inline bool netif_is_bond_master(const struct net_device *dev)
8a7fbfab 4179{
4180 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4181}
4182
b618aaa9 4183static inline bool netif_is_bond_slave(const struct net_device *dev)
1765a575
JP
4184{
4185 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4186}
4187
3bdc0eba
BG
4188static inline bool netif_supports_nofcs(struct net_device *dev)
4189{
4190 return dev->priv_flags & IFF_SUPP_NOFCS;
4191}
4192
007979ea 4193static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 4194{
007979ea 4195 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
4196}
4197
fee6d4c7
DA
4198static inline bool netif_is_l3_slave(const struct net_device *dev)
4199{
4200 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4201}
4202
0894ae3f
JP
4203static inline bool netif_is_bridge_master(const struct net_device *dev)
4204{
4205 return dev->priv_flags & IFF_EBRIDGE;
4206}
4207
28f9ee22
VY
4208static inline bool netif_is_bridge_port(const struct net_device *dev)
4209{
4210 return dev->priv_flags & IFF_BRIDGE_PORT;
4211}
4212
35d4e172
JP
4213static inline bool netif_is_ovs_master(const struct net_device *dev)
4214{
4215 return dev->priv_flags & IFF_OPENVSWITCH;
4216}
4217
5be66141
JP
4218static inline bool netif_is_ovs_port(const struct net_device *dev)
4219{
4220 return dev->priv_flags & IFF_OVS_DATAPATH;
4221}
4222
b618aaa9 4223static inline bool netif_is_team_master(const struct net_device *dev)
c981e421
JP
4224{
4225 return dev->priv_flags & IFF_TEAM;
4226}
4227
b618aaa9 4228static inline bool netif_is_team_port(const struct net_device *dev)
f7f019ee
JP
4229{
4230 return dev->priv_flags & IFF_TEAM_PORT;
4231}
4232
b618aaa9 4233static inline bool netif_is_lag_master(const struct net_device *dev)
7be61833
JP
4234{
4235 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4236}
4237
b618aaa9 4238static inline bool netif_is_lag_port(const struct net_device *dev)
e0ba1414
JP
4239{
4240 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4241}
4242
d4ab4286
KJ
4243static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4244{
4245 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4246}
4247
02875878
ED
4248/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4249static inline void netif_keep_dst(struct net_device *dev)
4250{
4251 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4252}
4253
18d3df3e
PA
4254/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4255static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4256{
4257 /* TODO: reserve and use an additional IFF bit, if we get more users */
4258 return dev->priv_flags & IFF_MACSEC;
4259}
4260
505d4f73 4261extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 4262
571ba423
JP
4263/* Logging, debugging and troubleshooting/diagnostic helpers. */
4264
4265/* netdev_printk helpers, similar to dev_printk */
4266
4267static inline const char *netdev_name(const struct net_device *dev)
4268{
c6f854d5
VF
4269 if (!dev->name[0] || strchr(dev->name, '%'))
4270 return "(unnamed net_device)";
571ba423
JP
4271 return dev->name;
4272}
4273
8397ed36
DA
4274static inline bool netdev_unregistering(const struct net_device *dev)
4275{
4276 return dev->reg_state == NETREG_UNREGISTERING;
4277}
4278
ccc7f496
VF
4279static inline const char *netdev_reg_state(const struct net_device *dev)
4280{
4281 switch (dev->reg_state) {
4282 case NETREG_UNINITIALIZED: return " (uninitialized)";
4283 case NETREG_REGISTERED: return "";
4284 case NETREG_UNREGISTERING: return " (unregistering)";
4285 case NETREG_UNREGISTERED: return " (unregistered)";
4286 case NETREG_RELEASED: return " (released)";
4287 case NETREG_DUMMY: return " (dummy)";
4288 }
4289
4290 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4291 return " (unknown)";
4292}
4293
f629d208 4294__printf(3, 4)
6ea754eb
JP
4295void netdev_printk(const char *level, const struct net_device *dev,
4296 const char *format, ...);
f629d208 4297__printf(2, 3)
6ea754eb 4298void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 4299__printf(2, 3)
6ea754eb 4300void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 4301__printf(2, 3)
6ea754eb 4302void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 4303__printf(2, 3)
6ea754eb 4304void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 4305__printf(2, 3)
6ea754eb 4306void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 4307__printf(2, 3)
6ea754eb 4308void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 4309__printf(2, 3)
6ea754eb 4310void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 4311
8909c9ad
VK
4312#define MODULE_ALIAS_NETDEV(device) \
4313 MODULE_ALIAS("netdev-" device)
4314
b558c96f 4315#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
4316#define netdev_dbg(__dev, format, args...) \
4317do { \
ffa10cb4 4318 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 4319} while (0)
b558c96f
JC
4320#elif defined(DEBUG)
4321#define netdev_dbg(__dev, format, args...) \
4322 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
4323#else
4324#define netdev_dbg(__dev, format, args...) \
4325({ \
4326 if (0) \
4327 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
4328})
4329#endif
4330
4331#if defined(VERBOSE_DEBUG)
4332#define netdev_vdbg netdev_dbg
4333#else
4334
4335#define netdev_vdbg(dev, format, args...) \
4336({ \
4337 if (0) \
4338 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4339 0; \
4340})
4341#endif
4342
4343/*
4344 * netdev_WARN() acts like dev_printk(), but with the key difference
4345 * of using a WARN/WARN_ON to get the message out, including the
4346 * file/line information and a backtrace.
4347 */
4348#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
4349 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4350 netdev_reg_state(dev), ##args)
571ba423 4351
b3d95c5c
JP
4352/* netif printk helpers, similar to netdev_printk */
4353
4354#define netif_printk(priv, type, level, dev, fmt, args...) \
4355do { \
4356 if (netif_msg_##type(priv)) \
4357 netdev_printk(level, (dev), fmt, ##args); \
4358} while (0)
4359
f45f4321
JP
4360#define netif_level(level, priv, type, dev, fmt, args...) \
4361do { \
4362 if (netif_msg_##type(priv)) \
4363 netdev_##level(dev, fmt, ##args); \
4364} while (0)
4365
b3d95c5c 4366#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 4367 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 4368#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 4369 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 4370#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 4371 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 4372#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 4373 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 4374#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4375 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4376#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4377 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4378#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4379 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4380
0053ea9c 4381#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4382#define netif_dbg(priv, type, netdev, format, args...) \
4383do { \
4384 if (netif_msg_##type(priv)) \
b5fb0a03 4385 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4386} while (0)
0053ea9c
JP
4387#elif defined(DEBUG)
4388#define netif_dbg(priv, type, dev, format, args...) \
4389 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4390#else
4391#define netif_dbg(priv, type, dev, format, args...) \
4392({ \
4393 if (0) \
4394 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4395 0; \
4396})
4397#endif
4398
f617f276
EC
4399/* if @cond then downgrade to debug, else print at @level */
4400#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4401 do { \
4402 if (cond) \
4403 netif_dbg(priv, type, netdev, fmt, ##args); \
4404 else \
4405 netif_ ## level(priv, type, netdev, fmt, ##args); \
4406 } while (0)
4407
b3d95c5c 4408#if defined(VERBOSE_DEBUG)
bcfcc450 4409#define netif_vdbg netif_dbg
b3d95c5c
JP
4410#else
4411#define netif_vdbg(priv, type, dev, format, args...) \
4412({ \
4413 if (0) \
a4ed89cb 4414 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4415 0; \
4416})
4417#endif
571ba423 4418
900ff8c6
CW
4419/*
4420 * The list of packet types we will receive (as opposed to discard)
4421 * and the routines to invoke.
4422 *
4423 * Why 16. Because with 16 the only overlap we get on a hash of the
4424 * low nibble of the protocol value is RARP/SNAP/X.25.
4425 *
4426 * NOTE: That is no longer true with the addition of VLAN tags. Not
4427 * sure which should go first, but I bet it won't make much
4428 * difference if we are running VLANs. The good news is that
4429 * this protocol won't be in the list unless compiled in, so
4430 * the average user (w/out VLANs) will not be adversely affected.
4431 * --BLG
4432 *
4433 * 0800 IP
4434 * 8100 802.1Q VLAN
4435 * 0001 802.3
4436 * 0002 AX.25
4437 * 0004 802.2
4438 * 8035 RARP
4439 * 0005 SNAP
4440 * 0805 X.25
4441 * 0806 ARP
4442 * 8137 IPX
4443 * 0009 Localtalk
4444 * 86DD IPv6
4445 */
4446#define PTYPE_HASH_SIZE (16)
4447#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4448
385a154c 4449#endif /* _LINUX_NETDEVICE_H */