]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
net: add netif_is_ovs_master helper with IFF_OPENVSWITCH private flag
[mirror_ubuntu-artful-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
cf85d08f 44#include <net/dsa.h>
7a6b6f51 45#ifdef CONFIG_DCB
2f90b865
AD
46#include <net/dcbnl.h>
47#endif
5bc1421e 48#include <net/netprio_cgroup.h>
a050c33f 49
a59e2ecb 50#include <linux/netdev_features.h>
77162022 51#include <linux/neighbour.h>
607ca46e 52#include <uapi/linux/netdevice.h>
61bd3857 53#include <uapi/linux/if_bonding.h>
a59e2ecb 54
115c1d6e 55struct netpoll_info;
313162d0 56struct device;
c1f19b51 57struct phy_device;
704232c2
JB
58/* 802.11 specific */
59struct wireless_dev;
98a18b6f
AA
60/* 802.15.4 specific */
61struct wpan_dev;
03c57747 62struct mpls_dev;
1da177e4 63
f629d208
JP
64void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
d07d7507 66
9a1654ba
JP
67/* Backlog congestion levels */
68#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69#define NET_RX_DROP 1 /* packet dropped */
70
572a9d7b
PM
71/*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88/* qdisc ->enqueue() return codes. */
89#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
90#define NET_XMIT_DROP 0x01 /* skb dropped */
91#define NET_XMIT_CN 0x02 /* congestion notification */
92#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 94
b9df3cb8
GR
95/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 98#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
99#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
dc1f8bf6 101/* Driver transmit return codes */
9a1654ba 102#define NETDEV_TX_MASK 0xf0
572a9d7b 103
dc1f8bf6 104enum netdev_tx {
572a9d7b 105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
dc1f8bf6
SH
109};
110typedef enum netdev_tx netdev_tx_t;
111
9a1654ba
JP
112/*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116static inline bool dev_xmit_complete(int rc)
117{
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128}
129
1da177e4
LT
130/*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
fe2918b0 134
d11ead75 135#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
136# if defined(CONFIG_MAC80211_MESH)
137# define LL_MAX_HEADER 128
138# else
139# define LL_MAX_HEADER 96
140# endif
1da177e4 141#else
8388e3da 142# define LL_MAX_HEADER 32
1da177e4
LT
143#endif
144
d11ead75
BH
145#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
147#define MAX_HEADER LL_MAX_HEADER
148#else
149#define MAX_HEADER (LL_MAX_HEADER + 48)
150#endif
151
152/*
be1f3c2c
BH
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
1da177e4 155 */
fe2918b0 156
d94d9fee 157struct net_device_stats {
3cfde79c
BH
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
1da177e4 167 unsigned long collisions;
1da177e4 168 unsigned long rx_length_errors;
3cfde79c
BH
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
1da177e4
LT
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
1da177e4
LT
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
181};
182
1da177e4
LT
183
184#include <linux/cache.h>
185#include <linux/skbuff.h>
186
adc9300e 187#ifdef CONFIG_RPS
c5905afb
IM
188#include <linux/static_key.h>
189extern struct static_key rps_needed;
adc9300e
ED
190#endif
191
1da177e4
LT
192struct neighbour;
193struct neigh_parms;
194struct sk_buff;
195
f001fde5
JP
196struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
199 unsigned char type;
ccffad25
JP
200#define NETDEV_HW_ADDR_T_LAN 1
201#define NETDEV_HW_ADDR_T_SAN 2
202#define NETDEV_HW_ADDR_T_SLAVE 3
203#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 204#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 205 bool global_use;
4cd729b0 206 int sync_cnt;
8f8f103d 207 int refcount;
4543fbef 208 int synced;
f001fde5
JP
209 struct rcu_head rcu_head;
210};
211
31278e71
JP
212struct netdev_hw_addr_list {
213 struct list_head list;
214 int count;
215};
216
22bedad3
JP
217#define netdev_hw_addr_list_count(l) ((l)->count)
218#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219#define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 221
22bedad3
JP
222#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224#define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 226
22bedad3
JP
227#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 229#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 231
d94d9fee 232struct hh_cache {
f6b72b62 233 u16 hh_len;
5c25f686 234 u16 __pad;
3644f0ce 235 seqlock_t hh_lock;
1da177e4
LT
236
237 /* cached hardware header; allow for machine alignment needs. */
238#define HH_DATA_MOD 16
239#define HH_DATA_OFF(__len) \
5ba0eac6 240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
241#define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
244};
245
246/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * Alternative is:
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 *
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
253 */
254#define LL_RESERVED_SPACE(dev) \
f5184d26 255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 256#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 258
3b04ddde
SH
259struct header_ops {
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
95c96174 262 const void *saddr, unsigned int len);
3b04ddde 263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268};
269
1da177e4
LT
270/* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
272 * code.
273 */
274
d94d9fee 275enum netdev_state_t {
1da177e4
LT
276 __LINK_STATE_START,
277 __LINK_STATE_PRESENT,
1da177e4 278 __LINK_STATE_NOCARRIER,
b00055aa
SR
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
1da177e4
LT
281};
282
283
284/*
285 * This structure holds at boot time configured netdevice settings. They
fe2918b0 286 * are then used in the device probing.
1da177e4
LT
287 */
288struct netdev_boot_setup {
289 char name[IFNAMSIZ];
290 struct ifmap map;
291};
292#define NETDEV_BOOT_SETUP_MAX 8
293
f629d208 294int __init netdev_boot_setup(char *str);
1da177e4 295
bea3348e
SH
296/*
297 * Structure for NAPI scheduling similar to tasklet but with weighting
298 */
299struct napi_struct {
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
305 */
306 struct list_head poll_list;
307
308 unsigned long state;
309 int weight;
404f7c9e 310 unsigned int gro_count;
bea3348e
SH
311 int (*poll)(struct napi_struct *, int);
312#ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
314 int poll_owner;
bea3348e 315#endif
5d38a079 316 struct net_device *dev;
d565b0a1 317 struct sk_buff *gro_list;
5d38a079 318 struct sk_buff *skb;
3b47d303 319 struct hrtimer timer;
404f7c9e 320 struct list_head dev_list;
af12fa6e
ET
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
bea3348e
SH
323};
324
d94d9fee 325enum {
bea3348e 326 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 327 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
af12fa6e 329 NAPI_STATE_HASHED, /* In NAPI hash */
bea3348e
SH
330};
331
5b252f0c 332enum gro_result {
d1c76af9
HX
333 GRO_MERGED,
334 GRO_MERGED_FREE,
335 GRO_HELD,
336 GRO_NORMAL,
337 GRO_DROP,
338};
5b252f0c 339typedef enum gro_result gro_result_t;
d1c76af9 340
8a4eb573
JP
341/*
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * further.
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 *
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
352 *
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
355 *
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
359 *
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
361 * do with the skb.
362 *
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
366 *
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
370 *
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 373 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
374 *
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
377 *
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
380 */
381
382enum rx_handler_result {
383 RX_HANDLER_CONSUMED,
384 RX_HANDLER_ANOTHER,
385 RX_HANDLER_EXACT,
386 RX_HANDLER_PASS,
387};
388typedef enum rx_handler_result rx_handler_result_t;
389typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 390
f629d208 391void __napi_schedule(struct napi_struct *n);
bc9ad166 392void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 393
4d29515f 394static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
395{
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
397}
398
bea3348e
SH
399/**
400 * napi_schedule_prep - check if napi can be scheduled
401 * @n: napi context
402 *
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
a0a46196
DM
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
bea3348e 407 */
4d29515f 408static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 409{
a0a46196
DM
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
412}
413
414/**
415 * napi_schedule - schedule NAPI poll
416 * @n: napi context
417 *
418 * Schedule NAPI poll routine to be called if it is not already
419 * running.
420 */
421static inline void napi_schedule(struct napi_struct *n)
422{
423 if (napi_schedule_prep(n))
424 __napi_schedule(n);
425}
426
bc9ad166
ED
427/**
428 * napi_schedule_irqoff - schedule NAPI poll
429 * @n: napi context
430 *
431 * Variant of napi_schedule(), assuming hard irqs are masked.
432 */
433static inline void napi_schedule_irqoff(struct napi_struct *n)
434{
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
437}
438
bfe13f54 439/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 440static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
441{
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
4d29515f 444 return true;
bfe13f54 445 }
4d29515f 446 return false;
bfe13f54
RD
447}
448
3b47d303
ED
449void __napi_complete(struct napi_struct *n);
450void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
451/**
452 * napi_complete - NAPI processing complete
453 * @n: napi context
454 *
455 * Mark NAPI processing as complete.
3b47d303 456 * Consider using napi_complete_done() instead.
bea3348e 457 */
3b47d303
ED
458static inline void napi_complete(struct napi_struct *n)
459{
460 return napi_complete_done(n, 0);
461}
bea3348e 462
af12fa6e
ET
463/**
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
466 *
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
469 */
f629d208 470struct napi_struct *napi_by_id(unsigned int napi_id);
af12fa6e
ET
471
472/**
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
475 *
476 * generate a new napi_id and store a @napi under it in napi_hash
477 */
f629d208 478void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
479
480/**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
483 *
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
486 */
f629d208 487void napi_hash_del(struct napi_struct *napi);
af12fa6e 488
bea3348e
SH
489/**
490 * napi_disable - prevent NAPI from scheduling
491 * @n: napi context
492 *
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
495 */
3b47d303 496void napi_disable(struct napi_struct *n);
bea3348e
SH
497
498/**
499 * napi_enable - enable NAPI scheduling
500 * @n: napi context
501 *
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
504 */
505static inline void napi_enable(struct napi_struct *n)
506{
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 508 smp_mb__before_atomic();
bea3348e
SH
509 clear_bit(NAPI_STATE_SCHED, &n->state);
510}
511
c264c3de
SH
512#ifdef CONFIG_SMP
513/**
514 * napi_synchronize - wait until NAPI is not running
515 * @n: napi context
516 *
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
520 */
521static inline void napi_synchronize(const struct napi_struct *n)
522{
523 while (test_bit(NAPI_STATE_SCHED, &n->state))
524 msleep(1);
525}
526#else
527# define napi_synchronize(n) barrier()
528#endif
529
d94d9fee 530enum netdev_queue_state_t {
73466498
TH
531 __QUEUE_STATE_DRV_XOFF,
532 __QUEUE_STATE_STACK_XOFF,
c3f26a26 533 __QUEUE_STATE_FROZEN,
79d16385 534};
8e2f1a63
DB
535
536#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
537#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
538#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
539
540#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
541#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
542 QUEUE_STATE_FROZEN)
543#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
544 QUEUE_STATE_FROZEN)
545
73466498
TH
546/*
547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
548 * netif_tx_* functions below are used to manipulate this flag. The
549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
550 * queue independently. The netif_xmit_*stopped functions below are called
551 * to check if the queue has been stopped by the driver or stack (either
552 * of the XOFF bits are set in the state). Drivers should not need to call
553 * netif_xmit*stopped functions, they should only be using netif_tx_*.
554 */
79d16385 555
bb949fbd 556struct netdev_queue {
6a321cb3
ED
557/*
558 * read mostly part
559 */
bb949fbd 560 struct net_device *dev;
46e5da40 561 struct Qdisc __rcu *qdisc;
b0e1e646 562 struct Qdisc *qdisc_sleeping;
ccf5ff69 563#ifdef CONFIG_SYSFS
1d24eb48
TH
564 struct kobject kobj;
565#endif
f2cd2d3e
ED
566#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
567 int numa_node;
568#endif
6a321cb3
ED
569/*
570 * write mostly part
571 */
572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
573 int xmit_lock_owner;
9d21493b
ED
574 /*
575 * please use this field instead of dev->trans_start
576 */
577 unsigned long trans_start;
ccf5ff69 578
579 /*
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
582 */
583 unsigned long trans_timeout;
114cf580
TH
584
585 unsigned long state;
586
587#ifdef CONFIG_BQL
588 struct dql dql;
589#endif
822b3b2e 590 unsigned long tx_maxrate;
e8a0464c 591} ____cacheline_aligned_in_smp;
bb949fbd 592
f2cd2d3e
ED
593static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
594{
595#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 return q->numa_node;
597#else
b236da69 598 return NUMA_NO_NODE;
f2cd2d3e
ED
599#endif
600}
601
602static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
603{
604#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 q->numa_node = node;
606#endif
607}
608
df334545 609#ifdef CONFIG_RPS
0a9627f2
TH
610/*
611 * This structure holds an RPS map which can be of variable length. The
612 * map is an array of CPUs.
613 */
614struct rps_map {
615 unsigned int len;
616 struct rcu_head rcu;
617 u16 cpus[0];
618};
60b778ce 619#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 620
fec5e652 621/*
c445477d
BH
622 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
623 * tail pointer for that CPU's input queue at the time of last enqueue, and
624 * a hardware filter index.
fec5e652
TH
625 */
626struct rps_dev_flow {
627 u16 cpu;
c445477d 628 u16 filter;
fec5e652
TH
629 unsigned int last_qtail;
630};
c445477d 631#define RPS_NO_FILTER 0xffff
fec5e652
TH
632
633/*
634 * The rps_dev_flow_table structure contains a table of flow mappings.
635 */
636struct rps_dev_flow_table {
637 unsigned int mask;
638 struct rcu_head rcu;
fec5e652
TH
639 struct rps_dev_flow flows[0];
640};
641#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 642 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
643
644/*
645 * The rps_sock_flow_table contains mappings of flows to the last CPU
646 * on which they were processed by the application (set in recvmsg).
567e4b79
ED
647 * Each entry is a 32bit value. Upper part is the high order bits
648 * of flow hash, lower part is cpu number.
649 * rps_cpu_mask is used to partition the space, depending on number of
650 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
651 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
652 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
653 */
654struct rps_sock_flow_table {
567e4b79 655 u32 mask;
93c1af6c
ED
656
657 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 658};
567e4b79 659#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
660
661#define RPS_NO_CPU 0xffff
662
567e4b79
ED
663extern u32 rps_cpu_mask;
664extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665
fec5e652
TH
666static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
667 u32 hash)
668{
669 if (table && hash) {
567e4b79
ED
670 unsigned int index = hash & table->mask;
671 u32 val = hash & ~rps_cpu_mask;
fec5e652
TH
672
673 /* We only give a hint, preemption can change cpu under us */
567e4b79 674 val |= raw_smp_processor_id();
fec5e652 675
567e4b79
ED
676 if (table->ents[index] != val)
677 table->ents[index] = val;
fec5e652
TH
678 }
679}
680
c445477d 681#ifdef CONFIG_RFS_ACCEL
f629d208
JP
682bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
683 u16 filter_id);
c445477d 684#endif
a953be53 685#endif /* CONFIG_RPS */
c445477d 686
0a9627f2
TH
687/* This structure contains an instance of an RX queue. */
688struct netdev_rx_queue {
a953be53 689#ifdef CONFIG_RPS
6e3f7faf
ED
690 struct rps_map __rcu *rps_map;
691 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 692#endif
6e3f7faf 693 struct kobject kobj;
fe822240 694 struct net_device *dev;
0a9627f2 695} ____cacheline_aligned_in_smp;
a953be53
MD
696
697/*
698 * RX queue sysfs structures and functions.
699 */
700struct rx_queue_attribute {
701 struct attribute attr;
702 ssize_t (*show)(struct netdev_rx_queue *queue,
703 struct rx_queue_attribute *attr, char *buf);
704 ssize_t (*store)(struct netdev_rx_queue *queue,
705 struct rx_queue_attribute *attr, const char *buf, size_t len);
706};
d314774c 707
bf264145
TH
708#ifdef CONFIG_XPS
709/*
710 * This structure holds an XPS map which can be of variable length. The
711 * map is an array of queues.
712 */
713struct xps_map {
714 unsigned int len;
715 unsigned int alloc_len;
716 struct rcu_head rcu;
717 u16 queues[0];
718};
60b778ce 719#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
bf264145
TH
720#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
721 / sizeof(u16))
722
723/*
724 * This structure holds all XPS maps for device. Maps are indexed by CPU.
725 */
726struct xps_dev_maps {
727 struct rcu_head rcu;
a4177869 728 struct xps_map __rcu *cpu_map[0];
bf264145
TH
729};
730#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
731 (nr_cpu_ids * sizeof(struct xps_map *)))
732#endif /* CONFIG_XPS */
733
4f57c087
JF
734#define TC_MAX_QUEUE 16
735#define TC_BITMASK 15
736/* HW offloaded queuing disciplines txq count and offset maps */
737struct netdev_tc_txq {
738 u16 count;
739 u16 offset;
740};
741
68bad94e
NP
742#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
743/*
744 * This structure is to hold information about the device
745 * configured to run FCoE protocol stack.
746 */
747struct netdev_fcoe_hbainfo {
748 char manufacturer[64];
749 char serial_number[64];
750 char hardware_version[64];
751 char driver_version[64];
752 char optionrom_version[64];
753 char firmware_version[64];
754 char model[256];
755 char model_description[256];
756};
757#endif
758
02637fce 759#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 760
02637fce
JP
761/* This structure holds a unique identifier to identify some
762 * physical item (port for example) used by a netdevice.
66b52b0d 763 */
02637fce
JP
764struct netdev_phys_item_id {
765 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
766 unsigned char id_len;
767};
768
d754f98b
SF
769static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
770 struct netdev_phys_item_id *b)
771{
772 return a->id_len == b->id_len &&
773 memcmp(a->id, b->id, a->id_len) == 0;
774}
775
99932d4f
DB
776typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
777 struct sk_buff *skb);
778
d314774c
SH
779/*
780 * This structure defines the management hooks for network devices.
00829823
SH
781 * The following hooks can be defined; unless noted otherwise, they are
782 * optional and can be filled with a null pointer.
d314774c
SH
783 *
784 * int (*ndo_init)(struct net_device *dev);
785 * This function is called once when network device is registered.
786 * The network device can use this to any late stage initializaton
787 * or semantic validattion. It can fail with an error code which will
788 * be propogated back to register_netdev
789 *
790 * void (*ndo_uninit)(struct net_device *dev);
791 * This function is called when device is unregistered or when registration
792 * fails. It is not called if init fails.
793 *
794 * int (*ndo_open)(struct net_device *dev);
795 * This function is called when network device transistions to the up
796 * state.
797 *
798 * int (*ndo_stop)(struct net_device *dev);
799 * This function is called when network device transistions to the down
800 * state.
801 *
dc1f8bf6
SH
802 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
803 * struct net_device *dev);
00829823 804 * Called when a packet needs to be transmitted.
e79d8429
RR
805 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
806 * the queue before that can happen; it's for obsolete devices and weird
807 * corner cases, but the stack really does a non-trivial amount
808 * of useless work if you return NETDEV_TX_BUSY.
dc1f8bf6 809 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
810 * Required can not be NULL.
811 *
f663dd9a 812 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 813 * void *accel_priv, select_queue_fallback_t fallback);
00829823
SH
814 * Called to decide which queue to when device supports multiple
815 * transmit queues.
816 *
d314774c
SH
817 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
818 * This function is called to allow device receiver to make
819 * changes to configuration when multicast or promiscious is enabled.
820 *
821 * void (*ndo_set_rx_mode)(struct net_device *dev);
822 * This function is called device changes address list filtering.
01789349
JP
823 * If driver handles unicast address filtering, it should set
824 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
825 *
826 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
827 * This function is called when the Media Access Control address
37b607c5 828 * needs to be changed. If this interface is not defined, the
d314774c
SH
829 * mac address can not be changed.
830 *
831 * int (*ndo_validate_addr)(struct net_device *dev);
832 * Test if Media Access Control address is valid for the device.
833 *
834 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
835 * Called when a user request an ioctl which can't be handled by
836 * the generic interface code. If not defined ioctl's return
837 * not supported error code.
838 *
839 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
840 * Used to set network devices bus interface parameters. This interface
841 * is retained for legacy reason, new devices should use the bus
842 * interface (PCI) for low level management.
843 *
844 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
845 * Called when a user wants to change the Maximum Transfer Unit
846 * of a device. If not defined, any request to change MTU will
847 * will return an error.
848 *
00829823 849 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
850 * Callback uses when the transmitter has not made any progress
851 * for dev->watchdog ticks.
852 *
3cfde79c 853 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 854 * struct rtnl_link_stats64 *storage);
d308e38f 855 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 856 * Called when a user wants to get the network device usage
be1f3c2c 857 * statistics. Drivers must do one of the following:
3cfde79c
BH
858 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
859 * rtnl_link_stats64 structure passed by the caller.
82695d9b 860 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
861 * (which should normally be dev->stats) and return a pointer to
862 * it. The structure may be changed asynchronously only if each
863 * field is written atomically.
864 * 3. Update dev->stats asynchronously and atomically, and define
865 * neither operation.
d314774c 866 *
5d632cb7 867 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
868 * If device support VLAN filtering this function is called when a
869 * VLAN id is registered.
d314774c 870 *
5d632cb7 871 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
872 * If device support VLAN filtering this function is called when a
873 * VLAN id is unregistered.
d314774c
SH
874 *
875 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
876 *
877 * SR-IOV management functions.
878 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
879 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
880 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
881 * int max_tx_rate);
5f8444a3 882 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
883 * int (*ndo_get_vf_config)(struct net_device *dev,
884 * int vf, struct ifla_vf_info *ivf);
1d8faf48 885 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
886 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
887 * struct nlattr *port[]);
01a3d796
VZ
888 *
889 * Enable or disable the VF ability to query its RSS Redirection Table and
890 * Hash Key. This is needed since on some devices VF share this information
891 * with PF and querying it may adduce a theoretical security risk.
892 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 893 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
894 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
895 * Called to setup 'tc' number of traffic classes in the net device. This
896 * is always called from the stack with the rtnl lock held and netif tx
897 * queues stopped. This allows the netdevice to perform queue management
898 * safely.
c445477d 899 *
e9bce845
YZ
900 * Fiber Channel over Ethernet (FCoE) offload functions.
901 * int (*ndo_fcoe_enable)(struct net_device *dev);
902 * Called when the FCoE protocol stack wants to start using LLD for FCoE
903 * so the underlying device can perform whatever needed configuration or
904 * initialization to support acceleration of FCoE traffic.
905 *
906 * int (*ndo_fcoe_disable)(struct net_device *dev);
907 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
908 * so the underlying device can perform whatever needed clean-ups to
909 * stop supporting acceleration of FCoE traffic.
910 *
911 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
912 * struct scatterlist *sgl, unsigned int sgc);
913 * Called when the FCoE Initiator wants to initialize an I/O that
914 * is a possible candidate for Direct Data Placement (DDP). The LLD can
915 * perform necessary setup and returns 1 to indicate the device is set up
916 * successfully to perform DDP on this I/O, otherwise this returns 0.
917 *
918 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
919 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
920 * indicated by the FC exchange id 'xid', so the underlying device can
921 * clean up and reuse resources for later DDP requests.
922 *
923 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
924 * struct scatterlist *sgl, unsigned int sgc);
925 * Called when the FCoE Target wants to initialize an I/O that
926 * is a possible candidate for Direct Data Placement (DDP). The LLD can
927 * perform necessary setup and returns 1 to indicate the device is set up
928 * successfully to perform DDP on this I/O, otherwise this returns 0.
929 *
68bad94e
NP
930 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
931 * struct netdev_fcoe_hbainfo *hbainfo);
932 * Called when the FCoE Protocol stack wants information on the underlying
933 * device. This information is utilized by the FCoE protocol stack to
934 * register attributes with Fiber Channel management service as per the
935 * FC-GS Fabric Device Management Information(FDMI) specification.
936 *
e9bce845
YZ
937 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
938 * Called when the underlying device wants to override default World Wide
939 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
940 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
941 * protocol stack to use.
942 *
c445477d
BH
943 * RFS acceleration.
944 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
945 * u16 rxq_index, u32 flow_id);
946 * Set hardware filter for RFS. rxq_index is the target queue index;
947 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
948 * Return the filter ID on success, or a negative error code.
fbaec0ea 949 *
8b98a70c 950 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
951 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
952 * Called to make another netdev an underling.
953 *
954 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
955 * Called to release previously enslaved netdev.
5455c699
MM
956 *
957 * Feature/offload setting functions.
c8f44aff
MM
958 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
959 * netdev_features_t features);
5455c699
MM
960 * Adjusts the requested feature flags according to device-specific
961 * constraints, and returns the resulting flags. Must not modify
962 * the device state.
963 *
c8f44aff 964 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
965 * Called to update device configuration to new features. Passed
966 * feature set might be less than what was returned by ndo_fix_features()).
967 * Must return >0 or -errno if it changed dev->features itself.
968 *
edc7d573 969 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
970 * struct net_device *dev,
f6f6424b 971 * const unsigned char *addr, u16 vid, u16 flags)
77162022 972 * Adds an FDB entry to dev for addr.
1690be63
VY
973 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
974 * struct net_device *dev,
f6f6424b 975 * const unsigned char *addr, u16 vid)
77162022
JF
976 * Deletes the FDB entry from dev coresponding to addr.
977 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
978 * struct net_device *dev, struct net_device *filter_dev,
979 * int idx)
77162022
JF
980 * Used to add FDB entries to dump requests. Implementers should add
981 * entries to skb and update idx with the number of entries.
e5a55a89 982 *
ad41faa8
ND
983 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
984 * u16 flags)
e5a55a89 985 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
986 * struct net_device *dev, u32 filter_mask,
987 * int nlflags)
ad41faa8
ND
988 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
989 * u16 flags);
4bf84c35
JP
990 *
991 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
992 * Called to change device carrier. Soft-devices (like dummy, team, etc)
993 * which do not represent real hardware may define this to allow their
994 * userspace components to manage their virtual carrier state. Devices
995 * that determine carrier state from physical hardware properties (eg
996 * network cables) or protocol-dependent mechanisms (eg
997 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
998 *
999 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1000 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1001 * Called to get ID of physical port of this device. If driver does
1002 * not implement this, it is assumed that the hw is not able to have
1003 * multiple net devices on single physical port.
53cf5275
JG
1004 *
1005 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 1006 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1007 * Called by vxlan to notiy a driver about the UDP port and socket
1008 * address family that vxlan is listnening to. It is called only when
1009 * a new port starts listening. The operation is protected by the
1010 * vxlan_net->sock_lock.
1011 *
1012 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 1013 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1014 * Called by vxlan to notify the driver about a UDP port and socket
1015 * address family that vxlan is not listening to anymore. The operation
1016 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
1017 *
1018 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1019 * struct net_device *dev)
1020 * Called by upper layer devices to accelerate switching or other
1021 * station functionality into hardware. 'pdev is the lowerdev
1022 * to use for the offload and 'dev' is the net device that will
1023 * back the offload. Returns a pointer to the private structure
1024 * the upper layer will maintain.
1025 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1026 * Called by upper layer device to delete the station created
1027 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1028 * the station and priv is the structure returned by the add
1029 * operation.
1030 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1031 * struct net_device *dev,
1032 * void *priv);
1033 * Callback to use for xmit over the accelerated station. This
1034 * is used in place of ndo_start_xmit on accelerated net
1035 * devices.
5f35227e
JG
1036 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1037 * struct net_device *dev
1038 * netdev_features_t features);
04ffcb25 1039 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1040 * performing offload operations on a given packet. This is to give
1041 * the device an opportunity to implement any restrictions that cannot
1042 * be otherwise expressed by feature flags. The check is called with
1043 * the set of features that the stack has calculated and it returns
1044 * those the driver believes to be appropriate.
822b3b2e
JF
1045 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1046 * int queue_index, u32 maxrate);
1047 * Called when a user wants to set a max-rate limitation of specific
1048 * TX queue.
a54acb3a
ND
1049 * int (*ndo_get_iflink)(const struct net_device *dev);
1050 * Called to get the iflink value of this device.
d746d707
AK
1051 * void (*ndo_change_proto_down)(struct net_device *dev,
1052 * bool proto_down);
1053 * This function is used to pass protocol port error state information
1054 * to the switch driver. The switch driver can react to the proto_down
1055 * by doing a phys down on the associated switch port.
1056 *
d314774c
SH
1057 */
1058struct net_device_ops {
1059 int (*ndo_init)(struct net_device *dev);
1060 void (*ndo_uninit)(struct net_device *dev);
1061 int (*ndo_open)(struct net_device *dev);
1062 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 1063 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
1064 struct net_device *dev);
1065 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1066 struct sk_buff *skb,
99932d4f
DB
1067 void *accel_priv,
1068 select_queue_fallback_t fallback);
d314774c
SH
1069 void (*ndo_change_rx_flags)(struct net_device *dev,
1070 int flags);
d314774c 1071 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1072 int (*ndo_set_mac_address)(struct net_device *dev,
1073 void *addr);
d314774c 1074 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1075 int (*ndo_do_ioctl)(struct net_device *dev,
1076 struct ifreq *ifr, int cmd);
d314774c
SH
1077 int (*ndo_set_config)(struct net_device *dev,
1078 struct ifmap *map);
00829823
SH
1079 int (*ndo_change_mtu)(struct net_device *dev,
1080 int new_mtu);
1081 int (*ndo_neigh_setup)(struct net_device *dev,
1082 struct neigh_parms *);
d314774c
SH
1083 void (*ndo_tx_timeout) (struct net_device *dev);
1084
28172739
ED
1085 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1086 struct rtnl_link_stats64 *storage);
d314774c
SH
1087 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1088
8e586137 1089 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1090 __be16 proto, u16 vid);
8e586137 1091 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1092 __be16 proto, u16 vid);
d314774c 1093#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1094 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1095 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1096 struct netpoll_info *info);
0e34e931 1097 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1098#endif
e0d1095a 1099#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1100 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1101#endif
95c26df8
WM
1102 int (*ndo_set_vf_mac)(struct net_device *dev,
1103 int queue, u8 *mac);
1104 int (*ndo_set_vf_vlan)(struct net_device *dev,
1105 int queue, u16 vlan, u8 qos);
ed616689
SC
1106 int (*ndo_set_vf_rate)(struct net_device *dev,
1107 int vf, int min_tx_rate,
1108 int max_tx_rate);
5f8444a3
GR
1109 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1110 int vf, bool setting);
95c26df8
WM
1111 int (*ndo_get_vf_config)(struct net_device *dev,
1112 int vf,
1113 struct ifla_vf_info *ivf);
1d8faf48
RE
1114 int (*ndo_set_vf_link_state)(struct net_device *dev,
1115 int vf, int link_state);
3b766cd8
EBE
1116 int (*ndo_get_vf_stats)(struct net_device *dev,
1117 int vf,
1118 struct ifla_vf_stats
1119 *vf_stats);
57b61080
SF
1120 int (*ndo_set_vf_port)(struct net_device *dev,
1121 int vf,
1122 struct nlattr *port[]);
1123 int (*ndo_get_vf_port)(struct net_device *dev,
1124 int vf, struct sk_buff *skb);
01a3d796
VZ
1125 int (*ndo_set_vf_rss_query_en)(
1126 struct net_device *dev,
1127 int vf, bool setting);
4f57c087 1128 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
d11ead75 1129#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1130 int (*ndo_fcoe_enable)(struct net_device *dev);
1131 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1132 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1133 u16 xid,
1134 struct scatterlist *sgl,
1135 unsigned int sgc);
1136 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1137 u16 xid);
6247e086
YZ
1138 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1139 u16 xid,
1140 struct scatterlist *sgl,
1141 unsigned int sgc);
68bad94e
NP
1142 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1143 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1144#endif
1145
d11ead75 1146#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1147#define NETDEV_FCOE_WWNN 0
1148#define NETDEV_FCOE_WWPN 1
1149 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1150 u64 *wwn, int type);
4d288d57 1151#endif
3c9c36bc 1152
c445477d
BH
1153#ifdef CONFIG_RFS_ACCEL
1154 int (*ndo_rx_flow_steer)(struct net_device *dev,
1155 const struct sk_buff *skb,
1156 u16 rxq_index,
1157 u32 flow_id);
1158#endif
fbaec0ea
JP
1159 int (*ndo_add_slave)(struct net_device *dev,
1160 struct net_device *slave_dev);
1161 int (*ndo_del_slave)(struct net_device *dev,
1162 struct net_device *slave_dev);
c8f44aff
MM
1163 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1164 netdev_features_t features);
5455c699 1165 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1166 netdev_features_t features);
da6a8fa0 1167 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1168 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1169
1170 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1171 struct nlattr *tb[],
77162022 1172 struct net_device *dev,
6b6e2725 1173 const unsigned char *addr,
f6f6424b 1174 u16 vid,
77162022
JF
1175 u16 flags);
1176 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1177 struct nlattr *tb[],
77162022 1178 struct net_device *dev,
f6f6424b
JP
1179 const unsigned char *addr,
1180 u16 vid);
77162022
JF
1181 int (*ndo_fdb_dump)(struct sk_buff *skb,
1182 struct netlink_callback *cb,
1183 struct net_device *dev,
5d5eacb3 1184 struct net_device *filter_dev,
77162022 1185 int idx);
e5a55a89
JF
1186
1187 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1188 struct nlmsghdr *nlh,
1189 u16 flags);
e5a55a89
JF
1190 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1191 u32 pid, u32 seq,
6cbdceeb 1192 struct net_device *dev,
46c264da
ND
1193 u32 filter_mask,
1194 int nlflags);
407af329 1195 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1196 struct nlmsghdr *nlh,
1197 u16 flags);
4bf84c35
JP
1198 int (*ndo_change_carrier)(struct net_device *dev,
1199 bool new_carrier);
66b52b0d 1200 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1201 struct netdev_phys_item_id *ppid);
db24a904
DA
1202 int (*ndo_get_phys_port_name)(struct net_device *dev,
1203 char *name, size_t len);
53cf5275
JG
1204 void (*ndo_add_vxlan_port)(struct net_device *dev,
1205 sa_family_t sa_family,
35e42379 1206 __be16 port);
53cf5275
JG
1207 void (*ndo_del_vxlan_port)(struct net_device *dev,
1208 sa_family_t sa_family,
35e42379 1209 __be16 port);
a6cc0cfa
JF
1210
1211 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1212 struct net_device *dev);
1213 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1214 void *priv);
1215
1216 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1217 struct net_device *dev,
1218 void *priv);
25175ba5 1219 int (*ndo_get_lock_subclass)(struct net_device *dev);
5f35227e
JG
1220 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1221 struct net_device *dev,
1222 netdev_features_t features);
822b3b2e
JF
1223 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1224 int queue_index,
1225 u32 maxrate);
a54acb3a 1226 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1227 int (*ndo_change_proto_down)(struct net_device *dev,
1228 bool proto_down);
d314774c
SH
1229};
1230
7aa98047
LR
1231/**
1232 * enum net_device_priv_flags - &struct net_device priv_flags
1233 *
1234 * These are the &struct net_device, they are only set internally
1235 * by drivers and used in the kernel. These flags are invisible to
1236 * userspace, this means that the order of these flags can change
1237 * during any kernel release.
1238 *
1239 * You should have a pretty good reason to be extending these flags.
1240 *
1241 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1242 * @IFF_EBRIDGE: Ethernet bridging device
1243 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1244 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1245 * @IFF_MASTER_ALB: bonding master, balance-alb
1246 * @IFF_BONDING: bonding master or slave
1247 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1248 * @IFF_ISATAP: ISATAP interface (RFC4214)
1249 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1250 * @IFF_WAN_HDLC: WAN HDLC device
1251 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1252 * release skb->dst
1253 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1254 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1255 * @IFF_MACVLAN_PORT: device used as macvlan port
1256 * @IFF_BRIDGE_PORT: device used as bridge port
1257 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1258 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1259 * @IFF_UNICAST_FLT: Supports unicast filtering
1260 * @IFF_TEAM_PORT: device used as team port
1261 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1262 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1263 * change when it's running
1264 * @IFF_MACVLAN: Macvlan device
808d28c4 1265 * @IFF_VRF_MASTER: device is a VRF master
fa8187c9 1266 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1267 * @IFF_OPENVSWITCH: device is a Open vSwitch master
7aa98047
LR
1268 */
1269enum netdev_priv_flags {
1270 IFF_802_1Q_VLAN = 1<<0,
1271 IFF_EBRIDGE = 1<<1,
1272 IFF_SLAVE_INACTIVE = 1<<2,
1273 IFF_MASTER_8023AD = 1<<3,
1274 IFF_MASTER_ALB = 1<<4,
1275 IFF_BONDING = 1<<5,
1276 IFF_SLAVE_NEEDARP = 1<<6,
1277 IFF_ISATAP = 1<<7,
1278 IFF_MASTER_ARPMON = 1<<8,
1279 IFF_WAN_HDLC = 1<<9,
1280 IFF_XMIT_DST_RELEASE = 1<<10,
1281 IFF_DONT_BRIDGE = 1<<11,
1282 IFF_DISABLE_NETPOLL = 1<<12,
1283 IFF_MACVLAN_PORT = 1<<13,
1284 IFF_BRIDGE_PORT = 1<<14,
1285 IFF_OVS_DATAPATH = 1<<15,
1286 IFF_TX_SKB_SHARING = 1<<16,
1287 IFF_UNICAST_FLT = 1<<17,
1288 IFF_TEAM_PORT = 1<<18,
1289 IFF_SUPP_NOFCS = 1<<19,
1290 IFF_LIVE_ADDR_CHANGE = 1<<20,
1291 IFF_MACVLAN = 1<<21,
02875878 1292 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
2ad7bf36
MB
1293 IFF_IPVLAN_MASTER = 1<<23,
1294 IFF_IPVLAN_SLAVE = 1<<24,
4e3c8992 1295 IFF_VRF_MASTER = 1<<25,
fa8187c9 1296 IFF_NO_QUEUE = 1<<26,
35d4e172 1297 IFF_OPENVSWITCH = 1<<27,
7aa98047
LR
1298};
1299
1300#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1301#define IFF_EBRIDGE IFF_EBRIDGE
1302#define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1303#define IFF_MASTER_8023AD IFF_MASTER_8023AD
1304#define IFF_MASTER_ALB IFF_MASTER_ALB
1305#define IFF_BONDING IFF_BONDING
1306#define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1307#define IFF_ISATAP IFF_ISATAP
1308#define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1309#define IFF_WAN_HDLC IFF_WAN_HDLC
1310#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1311#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1312#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1313#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1314#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1315#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1316#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1317#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1318#define IFF_TEAM_PORT IFF_TEAM_PORT
1319#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1320#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1321#define IFF_MACVLAN IFF_MACVLAN
02875878 1322#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1323#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1324#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
4e3c8992 1325#define IFF_VRF_MASTER IFF_VRF_MASTER
fa8187c9 1326#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1327#define IFF_OPENVSWITCH IFF_OPENVSWITCH
7aa98047 1328
536721b1
KK
1329/**
1330 * struct net_device - The DEVICE structure.
1331 * Actually, this whole structure is a big mistake. It mixes I/O
1332 * data with strictly "high-level" data, and it has to know about
1333 * almost every data structure used in the INET module.
1334 *
1335 * @name: This is the first field of the "visible" part of this structure
1336 * (i.e. as seen by users in the "Space.c" file). It is the name
1337 * of the interface.
1338 *
1339 * @name_hlist: Device name hash chain, please keep it close to name[]
1340 * @ifalias: SNMP alias
1341 * @mem_end: Shared memory end
1342 * @mem_start: Shared memory start
1343 * @base_addr: Device I/O address
1344 * @irq: Device IRQ number
1345 *
14ffbbb8
TG
1346 * @carrier_changes: Stats to monitor carrier on<->off transitions
1347 *
536721b1
KK
1348 * @state: Generic network queuing layer state, see netdev_state_t
1349 * @dev_list: The global list of network devices
1350 * @napi_list: List entry, that is used for polling napi devices
1351 * @unreg_list: List entry, that is used, when we are unregistering the
1352 * device, see the function unregister_netdev
1353 * @close_list: List entry, that is used, when we are closing the device
1354 *
1355 * @adj_list: Directly linked devices, like slaves for bonding
1356 * @all_adj_list: All linked devices, *including* neighbours
1357 * @features: Currently active device features
1358 * @hw_features: User-changeable features
1359 *
1360 * @wanted_features: User-requested features
1361 * @vlan_features: Mask of features inheritable by VLAN devices
1362 *
1363 * @hw_enc_features: Mask of features inherited by encapsulating devices
1364 * This field indicates what encapsulation
1365 * offloads the hardware is capable of doing,
1366 * and drivers will need to set them appropriately.
1367 *
1368 * @mpls_features: Mask of features inheritable by MPLS
1369 *
1370 * @ifindex: interface index
388069d3 1371 * @group: The group, that the device belongs to
536721b1
KK
1372 *
1373 * @stats: Statistics struct, which was left as a legacy, use
1374 * rtnl_link_stats64 instead
1375 *
1376 * @rx_dropped: Dropped packets by core network,
1377 * do not use this in drivers
1378 * @tx_dropped: Dropped packets by core network,
1379 * do not use this in drivers
1380 *
536721b1
KK
1381 * @wireless_handlers: List of functions to handle Wireless Extensions,
1382 * instead of ioctl,
1383 * see <net/iw_handler.h> for details.
1384 * @wireless_data: Instance data managed by the core of wireless extensions
1385 *
1386 * @netdev_ops: Includes several pointers to callbacks,
1387 * if one wants to override the ndo_*() functions
1388 * @ethtool_ops: Management operations
d476059e 1389 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1390 * of Layer 2 headers.
1391 *
1392 * @flags: Interface flags (a la BSD)
1393 * @priv_flags: Like 'flags' but invisible to userspace,
1394 * see if.h for the definitions
1395 * @gflags: Global flags ( kept as legacy )
1396 * @padded: How much padding added by alloc_netdev()
1397 * @operstate: RFC2863 operstate
1398 * @link_mode: Mapping policy to operstate
1399 * @if_port: Selectable AUI, TP, ...
1400 * @dma: DMA channel
1401 * @mtu: Interface MTU value
1402 * @type: Interface hardware type
1403 * @hard_header_len: Hardware header length
1404 *
1405 * @needed_headroom: Extra headroom the hardware may need, but not in all
1406 * cases can this be guaranteed
1407 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1408 * cases can this be guaranteed. Some cases also use
1409 * LL_MAX_HEADER instead to allocate the skb
1410 *
1411 * interface address info:
1412 *
1413 * @perm_addr: Permanent hw address
1414 * @addr_assign_type: Hw address assignment type
1415 * @addr_len: Hardware address length
1416 * @neigh_priv_len; Used in neigh_alloc(),
1417 * initialized only in atm/clip.c
1418 * @dev_id: Used to differentiate devices that share
1419 * the same link layer address
1420 * @dev_port: Used to differentiate devices that share
1421 * the same function
1422 * @addr_list_lock: XXX: need comments on this one
536721b1
KK
1423 * @uc_promisc: Counter, that indicates, that promiscuous mode
1424 * has been enabled due to the need to listen to
1425 * additional unicast addresses in a device that
1426 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1427 * @uc: unicast mac addresses
1428 * @mc: multicast mac addresses
1429 * @dev_addrs: list of device hw addresses
1430 * @queues_kset: Group of all Kobjects in the Tx and RX queues
536721b1
KK
1431 * @promiscuity: Number of times, the NIC is told to work in
1432 * Promiscuous mode, if it becomes 0 the NIC will
1433 * exit from working in Promiscuous mode
1434 * @allmulti: Counter, enables or disables allmulticast mode
1435 *
1436 * @vlan_info: VLAN info
1437 * @dsa_ptr: dsa specific data
1438 * @tipc_ptr: TIPC specific data
1439 * @atalk_ptr: AppleTalk link
1440 * @ip_ptr: IPv4 specific data
1441 * @dn_ptr: DECnet specific data
1442 * @ip6_ptr: IPv6 specific data
1443 * @ax25_ptr: AX.25 specific data
4e3c8992 1444 * @vrf_ptr: VRF specific data
536721b1
KK
1445 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1446 *
1447 * @last_rx: Time of last Rx
1448 * @dev_addr: Hw address (before bcast,
1449 * because most packets are unicast)
1450 *
1451 * @_rx: Array of RX queues
1452 * @num_rx_queues: Number of RX queues
1453 * allocated at register_netdev() time
1454 * @real_num_rx_queues: Number of RX queues currently active in device
1455 *
1456 * @rx_handler: handler for received packets
1457 * @rx_handler_data: XXX: need comments on this one
1458 * @ingress_queue: XXX: need comments on this one
1459 * @broadcast: hw bcast address
1460 *
14ffbbb8
TG
1461 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1462 * indexed by RX queue number. Assigned by driver.
1463 * This must only be set if the ndo_rx_flow_steer
1464 * operation is defined
1465 * @index_hlist: Device index hash chain
1466 *
536721b1
KK
1467 * @_tx: Array of TX queues
1468 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1469 * @real_num_tx_queues: Number of TX queues currently active in device
1470 * @qdisc: Root qdisc from userspace point of view
1471 * @tx_queue_len: Max frames per queue allowed
1472 * @tx_global_lock: XXX: need comments on this one
1473 *
1474 * @xps_maps: XXX: need comments on this one
1475 *
0c4f691f
SF
1476 * @offload_fwd_mark: Offload device fwding mark
1477 *
536721b1
KK
1478 * @trans_start: Time (in jiffies) of last Tx
1479 * @watchdog_timeo: Represents the timeout that is used by
1480 * the watchdog ( see dev_watchdog() )
1481 * @watchdog_timer: List of timers
1482 *
1483 * @pcpu_refcnt: Number of references to this device
1484 * @todo_list: Delayed register/unregister
536721b1
KK
1485 * @link_watch_list: XXX: need comments on this one
1486 *
1487 * @reg_state: Register/unregister state machine
1488 * @dismantle: Device is going to be freed
1489 * @rtnl_link_state: This enum represents the phases of creating
1490 * a new link
1491 *
1492 * @destructor: Called from unregister,
1493 * can be used to call free_netdev
1494 * @npinfo: XXX: need comments on this one
1495 * @nd_net: Network namespace this network device is inside
1496 *
1497 * @ml_priv: Mid-layer private
1498 * @lstats: Loopback statistics
1499 * @tstats: Tunnel statistics
1500 * @dstats: Dummy statistics
1501 * @vstats: Virtual ethernet statistics
1502 *
1503 * @garp_port: GARP
1504 * @mrp_port: MRP
1505 *
1506 * @dev: Class/net/name entry
1507 * @sysfs_groups: Space for optional device, statistics and wireless
1508 * sysfs groups
1509 *
1510 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1511 * @rtnl_link_ops: Rtnl_link_ops
1512 *
1513 * @gso_max_size: Maximum size of generic segmentation offload
1514 * @gso_max_segs: Maximum number of segments that can be passed to the
1515 * NIC for GSO
fcbeb976
ED
1516 * @gso_min_segs: Minimum number of segments that can be passed to the
1517 * NIC for GSO
536721b1
KK
1518 *
1519 * @dcbnl_ops: Data Center Bridging netlink ops
1520 * @num_tc: Number of traffic classes in the net device
1521 * @tc_to_txq: XXX: need comments on this one
1522 * @prio_tc_map XXX: need comments on this one
1523 *
1524 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1525 *
1526 * @priomap: XXX: need comments on this one
1527 * @phydev: Physical device may attach itself
1528 * for hardware timestamping
1529 *
1530 * @qdisc_tx_busylock: XXX: need comments on this one
1531 *
d746d707
AK
1532 * @proto_down: protocol port state information can be sent to the
1533 * switch driver and used to set the phys state of the
1534 * switch port.
1535 *
1da177e4
LT
1536 * FIXME: cleanup struct net_device such that network protocol info
1537 * moves out.
1538 */
1539
d94d9fee 1540struct net_device {
1da177e4 1541 char name[IFNAMSIZ];
9356b8fc 1542 struct hlist_node name_hlist;
0b815a1a 1543 char *ifalias;
1da177e4
LT
1544 /*
1545 * I/O specific fields
1546 * FIXME: Merge these and struct ifmap into one
1547 */
536721b1
KK
1548 unsigned long mem_end;
1549 unsigned long mem_start;
1550 unsigned long base_addr;
1551 int irq;
1da177e4 1552
14ffbbb8
TG
1553 atomic_t carrier_changes;
1554
1da177e4 1555 /*
536721b1
KK
1556 * Some hardware also needs these fields (state,dev_list,
1557 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1558 * part of the usual set specified in Space.c.
1559 */
1560
1da177e4
LT
1561 unsigned long state;
1562
7562f876 1563 struct list_head dev_list;
bea3348e 1564 struct list_head napi_list;
44a0873d 1565 struct list_head unreg_list;
5cde2829 1566 struct list_head close_list;
7866a621
SN
1567 struct list_head ptype_all;
1568 struct list_head ptype_specific;
2f268f12 1569
2f268f12
VF
1570 struct {
1571 struct list_head upper;
1572 struct list_head lower;
1573 } adj_list;
1574
2f268f12
VF
1575 struct {
1576 struct list_head upper;
1577 struct list_head lower;
1578 } all_adj_list;
4c3d5e7b 1579
c8f44aff 1580 netdev_features_t features;
c8f44aff 1581 netdev_features_t hw_features;
c8f44aff 1582 netdev_features_t wanted_features;
c8f44aff 1583 netdev_features_t vlan_features;
6a674e9c 1584 netdev_features_t hw_enc_features;
0d89d203 1585 netdev_features_t mpls_features;
04ed3e74 1586
1da177e4 1587 int ifindex;
7a66bbc9 1588 int group;
1da177e4 1589
c45d286e 1590 struct net_device_stats stats;
015f0688 1591
015f0688
ED
1592 atomic_long_t rx_dropped;
1593 atomic_long_t tx_dropped;
1da177e4 1594
b86e0280 1595#ifdef CONFIG_WIRELESS_EXT
1da177e4 1596 const struct iw_handler_def * wireless_handlers;
1da177e4 1597 struct iw_public_data * wireless_data;
b86e0280 1598#endif
d314774c 1599 const struct net_device_ops *netdev_ops;
76fd8593 1600 const struct ethtool_ops *ethtool_ops;
4170604f 1601#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1602 const struct switchdev_ops *switchdev_ops;
4170604f 1603#endif
1da177e4 1604
3b04ddde
SH
1605 const struct header_ops *header_ops;
1606
536721b1
KK
1607 unsigned int flags;
1608 unsigned int priv_flags;
1609
1da177e4 1610 unsigned short gflags;
536721b1 1611 unsigned short padded;
1da177e4 1612
536721b1
KK
1613 unsigned char operstate;
1614 unsigned char link_mode;
b00055aa 1615
536721b1
KK
1616 unsigned char if_port;
1617 unsigned char dma;
bdc220da 1618
536721b1
KK
1619 unsigned int mtu;
1620 unsigned short type;
1621 unsigned short hard_header_len;
1da177e4 1622
f5184d26
JB
1623 unsigned short needed_headroom;
1624 unsigned short needed_tailroom;
1625
1da177e4 1626 /* Interface address info. */
536721b1
KK
1627 unsigned char perm_addr[MAX_ADDR_LEN];
1628 unsigned char addr_assign_type;
1629 unsigned char addr_len;
a0a9663d 1630 unsigned short neigh_priv_len;
536721b1
KK
1631 unsigned short dev_id;
1632 unsigned short dev_port;
ccffad25 1633 spinlock_t addr_list_lock;
14ffbbb8
TG
1634 unsigned char name_assign_type;
1635 bool uc_promisc;
536721b1
KK
1636 struct netdev_hw_addr_list uc;
1637 struct netdev_hw_addr_list mc;
1638 struct netdev_hw_addr_list dev_addrs;
1639
4c3d5e7b
ED
1640#ifdef CONFIG_SYSFS
1641 struct kset *queues_kset;
1642#endif
9d45abe1
WC
1643 unsigned int promiscuity;
1644 unsigned int allmulti;
1da177e4 1645
1da177e4
LT
1646
1647 /* Protocol specific pointers */
65ac6a5f 1648
d11ead75 1649#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1650 struct vlan_info __rcu *vlan_info;
65ac6a5f 1651#endif
34a430d7 1652#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1653 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1654#endif
1655#if IS_ENABLED(CONFIG_TIPC)
536721b1 1656 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1657#endif
536721b1
KK
1658 void *atalk_ptr;
1659 struct in_device __rcu *ip_ptr;
1660 struct dn_dev __rcu *dn_ptr;
1661 struct inet6_dev __rcu *ip6_ptr;
1662 void *ax25_ptr;
4e3c8992 1663 struct net_vrf_dev __rcu *vrf_ptr;
536721b1 1664 struct wireless_dev *ieee80211_ptr;
98a18b6f 1665 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1666#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1667 struct mpls_dev __rcu *mpls_ptr;
1668#endif
1da177e4 1669
9356b8fc 1670/*
cd13539b 1671 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1672 */
536721b1 1673 unsigned long last_rx;
4dc89133 1674
9356b8fc 1675 /* Interface address info used in eth_type_trans() */
536721b1 1676 unsigned char *dev_addr;
f001fde5 1677
0a9627f2 1678
a953be53 1679#ifdef CONFIG_SYSFS
0a9627f2
TH
1680 struct netdev_rx_queue *_rx;
1681
0a9627f2 1682 unsigned int num_rx_queues;
62fe0b40 1683 unsigned int real_num_rx_queues;
c445477d 1684
df334545 1685#endif
0a9627f2 1686
3b47d303 1687 unsigned long gro_flush_timeout;
61391cde 1688 rx_handler_func_t __rcu *rx_handler;
1689 void __rcu *rx_handler_data;
e8a0464c 1690
4cda01e8 1691#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1692 struct tcf_proto __rcu *ingress_cl_list;
1693#endif
24824a09 1694 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1695#ifdef CONFIG_NETFILTER_INGRESS
1696 struct list_head nf_hooks_ingress;
1697#endif
d2788d34 1698
536721b1 1699 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1700#ifdef CONFIG_RFS_ACCEL
1701 struct cpu_rmap *rx_cpu_rmap;
1702#endif
1703 struct hlist_node index_hlist;
cd13539b
ED
1704
1705/*
1706 * Cache lines mostly used on transmit path
1707 */
e8a0464c
DM
1708 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1709 unsigned int num_tx_queues;
fd2ea0a7 1710 unsigned int real_num_tx_queues;
af356afa 1711 struct Qdisc *qdisc;
536721b1 1712 unsigned long tx_queue_len;
c3f26a26 1713 spinlock_t tx_global_lock;
14ffbbb8 1714 int watchdog_timeo;
cd13539b 1715
bf264145 1716#ifdef CONFIG_XPS
a4177869 1717 struct xps_dev_maps __rcu *xps_maps;
bf264145 1718#endif
1d24eb48 1719
0c4f691f
SF
1720#ifdef CONFIG_NET_SWITCHDEV
1721 u32 offload_fwd_mark;
1722#endif
1723
9356b8fc 1724 /* These may be needed for future network-power-down code. */
9d21493b
ED
1725
1726 /*
1727 * trans_start here is expensive for high speed devices on SMP,
1728 * please use netdev_queue->trans_start instead.
1729 */
536721b1 1730 unsigned long trans_start;
9356b8fc 1731
9356b8fc
ED
1732 struct timer_list watchdog_timer;
1733
29b4433d 1734 int __percpu *pcpu_refcnt;
1da177e4 1735 struct list_head todo_list;
1da177e4 1736
e014debe 1737 struct list_head link_watch_list;
572a103d 1738
1da177e4 1739 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1740 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1741 NETREG_UNREGISTERING, /* called unregister_netdevice */
1742 NETREG_UNREGISTERED, /* completed unregister todo */
1743 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1744 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1745 } reg_state:8;
1746
536721b1 1747 bool dismantle;
a2835763
PM
1748
1749 enum {
1750 RTNL_LINK_INITIALIZED,
1751 RTNL_LINK_INITIALIZING,
1752 } rtnl_link_state:16;
1da177e4 1753
d314774c 1754 void (*destructor)(struct net_device *dev);
1da177e4 1755
1da177e4 1756#ifdef CONFIG_NETPOLL
5fbee843 1757 struct netpoll_info __rcu *npinfo;
1da177e4 1758#endif
eae792b7 1759
0c5c9fb5 1760 possible_net_t nd_net;
4a1c5371 1761
4951704b 1762 /* mid-layer private */
a7855c78 1763 union {
536721b1
KK
1764 void *ml_priv;
1765 struct pcpu_lstats __percpu *lstats;
8f84985f 1766 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1767 struct pcpu_dstats __percpu *dstats;
1768 struct pcpu_vstats __percpu *vstats;
a7855c78 1769 };
536721b1 1770
3cc77ec7 1771 struct garp_port __rcu *garp_port;
febf018d 1772 struct mrp_port __rcu *mrp_port;
1da177e4 1773
536721b1 1774 struct device dev;
0c509a6c 1775 const struct attribute_group *sysfs_groups[4];
a953be53 1776 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1777
38f7b870 1778 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1779
82cc1a7a
PWJ
1780 /* for setting kernel sock attribute on TCP connection setup */
1781#define GSO_MAX_SIZE 65536
1782 unsigned int gso_max_size;
30b678d8
BH
1783#define GSO_MAX_SEGS 65535
1784 u16 gso_max_segs;
fcbeb976 1785 u16 gso_min_segs;
7a6b6f51 1786#ifdef CONFIG_DCB
32953543 1787 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1788#endif
4f57c087
JF
1789 u8 num_tc;
1790 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1791 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1792
d11ead75 1793#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1794 unsigned int fcoe_ddp_xid;
5bc1421e 1795#endif
86f8515f 1796#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1797 struct netprio_map __rcu *priomap;
4d288d57 1798#endif
c1f19b51 1799 struct phy_device *phydev;
23d3b8bf 1800 struct lock_class_key *qdisc_tx_busylock;
d746d707 1801 bool proto_down;
1da177e4 1802};
43cb76d9 1803#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1804
1805#define NETDEV_ALIGN 32
1da177e4 1806
4f57c087
JF
1807static inline
1808int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1809{
1810 return dev->prio_tc_map[prio & TC_BITMASK];
1811}
1812
1813static inline
1814int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1815{
1816 if (tc >= dev->num_tc)
1817 return -EINVAL;
1818
1819 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1820 return 0;
1821}
1822
1823static inline
1824void netdev_reset_tc(struct net_device *dev)
1825{
1826 dev->num_tc = 0;
1827 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1828 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1829}
1830
1831static inline
1832int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1833{
1834 if (tc >= dev->num_tc)
1835 return -EINVAL;
1836
1837 dev->tc_to_txq[tc].count = count;
1838 dev->tc_to_txq[tc].offset = offset;
1839 return 0;
1840}
1841
1842static inline
1843int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1844{
1845 if (num_tc > TC_MAX_QUEUE)
1846 return -EINVAL;
1847
1848 dev->num_tc = num_tc;
1849 return 0;
1850}
1851
1852static inline
1853int netdev_get_num_tc(struct net_device *dev)
1854{
1855 return dev->num_tc;
1856}
1857
e8a0464c
DM
1858static inline
1859struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1860 unsigned int index)
1861{
1862 return &dev->_tx[index];
1863}
1864
10c51b56
DB
1865static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1866 const struct sk_buff *skb)
1867{
1868 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1869}
1870
e8a0464c
DM
1871static inline void netdev_for_each_tx_queue(struct net_device *dev,
1872 void (*f)(struct net_device *,
1873 struct netdev_queue *,
1874 void *),
1875 void *arg)
1876{
1877 unsigned int i;
1878
1879 for (i = 0; i < dev->num_tx_queues; i++)
1880 f(dev, &dev->_tx[i], arg);
1881}
1882
f629d208 1883struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1884 struct sk_buff *skb,
1885 void *accel_priv);
8c4c49df 1886
c346dca1
YH
1887/*
1888 * Net namespace inlines
1889 */
1890static inline
1891struct net *dev_net(const struct net_device *dev)
1892{
c2d9ba9b 1893 return read_pnet(&dev->nd_net);
c346dca1
YH
1894}
1895
1896static inline
f5aa23fd 1897void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 1898{
0c5c9fb5 1899 write_pnet(&dev->nd_net, net);
c346dca1
YH
1900}
1901
3e8a72d1 1902static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 1903{
3fc88677 1904#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
1905 if (dev->dsa_ptr != NULL)
1906 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 1907#endif
5aed85ce 1908 return false;
396138f0
LB
1909}
1910
bea3348e
SH
1911/**
1912 * netdev_priv - access network device private data
1913 * @dev: network device
1914 *
1915 * Get network device private data
1916 */
6472ce60 1917static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1918{
1ce8e7b5 1919 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1920}
1921
1da177e4
LT
1922/* Set the sysfs physical device reference for the network logical device
1923 * if set prior to registration will cause a symlink during initialization.
1924 */
43cb76d9 1925#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1926
384912ed 1927/* Set the sysfs device type for the network logical device to allow
3f79410c 1928 * fine-grained identification of different network device types. For
384912ed
MH
1929 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1930 */
1931#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1932
82dc3c63
ED
1933/* Default NAPI poll() weight
1934 * Device drivers are strongly advised to not use bigger value
1935 */
1936#define NAPI_POLL_WEIGHT 64
1937
3b582cc1
SH
1938/**
1939 * netif_napi_add - initialize a napi context
1940 * @dev: network device
1941 * @napi: napi context
1942 * @poll: polling function
1943 * @weight: default weight
1944 *
1945 * netif_napi_add() must be used to initialize a napi context prior to calling
1946 * *any* of the other napi related functions.
1947 */
d565b0a1
HX
1948void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1949 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1950
d8156534
AD
1951/**
1952 * netif_napi_del - remove a napi context
1953 * @napi: napi context
1954 *
1955 * netif_napi_del() removes a napi context from the network device napi list
1956 */
d565b0a1
HX
1957void netif_napi_del(struct napi_struct *napi);
1958
1959struct napi_gro_cb {
78a478d0
HX
1960 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1961 void *frag0;
1962
7489594c
HX
1963 /* Length of frag0. */
1964 unsigned int frag0_len;
1965
86911732
HX
1966 /* This indicates where we are processing relative to skb->data. */
1967 int data_offset;
1968
d565b0a1 1969 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
1970 u16 flush;
1971
1972 /* Save the IP ID here and check when we get to the transport layer */
1973 u16 flush_id;
d565b0a1
HX
1974
1975 /* Number of segments aggregated. */
2e71a6f8
ED
1976 u16 count;
1977
15e2396d
TH
1978 /* Start offset for remote checksum offload */
1979 u16 gro_remcsum_start;
1980
2e71a6f8
ED
1981 /* jiffies when first packet was created/queued */
1982 unsigned long age;
86347245 1983
afe93325 1984 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
1985 u16 proto;
1986
baa32ff4
TH
1987 /* This is non-zero if the packet may be of the same flow. */
1988 u8 same_flow:1;
1989
b582ef09 1990 /* Used in udp_gro_receive */
573e8fca
TH
1991 u8 udp_mark:1;
1992
1993 /* GRO checksum is valid */
1994 u8 csum_valid:1;
1995
662880f4
TH
1996 /* Number of checksums via CHECKSUM_UNNECESSARY */
1997 u8 csum_cnt:3;
c3c7c254 1998
baa32ff4
TH
1999 /* Free the skb? */
2000 u8 free:2;
2001#define NAPI_GRO_FREE 1
2002#define NAPI_GRO_FREE_STOLEN_HEAD 2
2003
efc98d08
TH
2004 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2005 u8 is_ipv6:1;
2006
baa32ff4
TH
2007 /* 7 bit hole */
2008
bf5a755f
JC
2009 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2010 __wsum csum;
2011
c3c7c254
ED
2012 /* used in skb_gro_receive() slow path */
2013 struct sk_buff *last;
d565b0a1
HX
2014};
2015
2016#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2017
1da177e4 2018struct packet_type {
f2ccd8fa
DM
2019 __be16 type; /* This is really htons(ether_type). */
2020 struct net_device *dev; /* NULL is wildcarded here */
2021 int (*func) (struct sk_buff *,
2022 struct net_device *,
2023 struct packet_type *,
2024 struct net_device *);
c0de08d0
EL
2025 bool (*id_match)(struct packet_type *ptype,
2026 struct sock *sk);
1da177e4
LT
2027 void *af_packet_priv;
2028 struct list_head list;
2029};
2030
f191a1d1 2031struct offload_callbacks {
576a30eb 2032 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2033 netdev_features_t features);
d565b0a1 2034 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2035 struct sk_buff *skb);
299603e8 2036 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2037};
2038
2039struct packet_offload {
2040 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2041 u16 priority;
f191a1d1
VY
2042 struct offload_callbacks callbacks;
2043 struct list_head list;
1da177e4
LT
2044};
2045
a2b12f3c
TH
2046struct udp_offload;
2047
2048struct udp_offload_callbacks {
2049 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2050 struct sk_buff *skb,
2051 struct udp_offload *uoff);
2052 int (*gro_complete)(struct sk_buff *skb,
2053 int nhoff,
2054 struct udp_offload *uoff);
2055};
2056
b582ef09
OG
2057struct udp_offload {
2058 __be16 port;
afe93325 2059 u8 ipproto;
a2b12f3c 2060 struct udp_offload_callbacks callbacks;
b582ef09
OG
2061};
2062
8f84985f
LR
2063/* often modified stats are per cpu, other are shared (netdev->stats) */
2064struct pcpu_sw_netstats {
2065 u64 rx_packets;
2066 u64 rx_bytes;
2067 u64 tx_packets;
2068 u64 tx_bytes;
2069 struct u64_stats_sync syncp;
2070};
2071
1c213bd2
WC
2072#define netdev_alloc_pcpu_stats(type) \
2073({ \
693350c2 2074 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1c213bd2 2075 if (pcpu_stats) { \
ec65aafb
JB
2076 int __cpu; \
2077 for_each_possible_cpu(__cpu) { \
1c213bd2 2078 typeof(type) *stat; \
ec65aafb 2079 stat = per_cpu_ptr(pcpu_stats, __cpu); \
1c213bd2
WC
2080 u64_stats_init(&stat->syncp); \
2081 } \
2082 } \
2083 pcpu_stats; \
2084})
2085
1da177e4
LT
2086#include <linux/notifier.h>
2087
dcfe1421
AW
2088/* netdevice notifier chain. Please remember to update the rtnetlink
2089 * notification exclusion list in rtnetlink_event() when adding new
2090 * types.
2091 */
2092#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2093#define NETDEV_DOWN 0x0002
2094#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2095 detected a hardware crash and restarted
2096 - we can use this eg to kick tcp sessions
2097 once done */
2098#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2099#define NETDEV_REGISTER 0x0005
2100#define NETDEV_UNREGISTER 0x0006
1d486bfb 2101#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2102#define NETDEV_CHANGEADDR 0x0008
2103#define NETDEV_GOING_DOWN 0x0009
2104#define NETDEV_CHANGENAME 0x000A
2105#define NETDEV_FEAT_CHANGE 0x000B
2106#define NETDEV_BONDING_FAILOVER 0x000C
2107#define NETDEV_PRE_UP 0x000D
2108#define NETDEV_PRE_TYPE_CHANGE 0x000E
2109#define NETDEV_POST_TYPE_CHANGE 0x000F
2110#define NETDEV_POST_INIT 0x0010
0115e8e3 2111#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2112#define NETDEV_RELEASE 0x0012
2113#define NETDEV_NOTIFY_PEERS 0x0013
2114#define NETDEV_JOIN 0x0014
42e52bf9 2115#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2116#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2117#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2118#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2119#define NETDEV_BONDING_INFO 0x0019
dcfe1421 2120
f629d208
JP
2121int register_netdevice_notifier(struct notifier_block *nb);
2122int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2123
2124struct netdev_notifier_info {
2125 struct net_device *dev;
2126};
2127
be9efd36
JP
2128struct netdev_notifier_change_info {
2129 struct netdev_notifier_info info; /* must be first */
2130 unsigned int flags_changed;
2131};
2132
0e4ead9d
JP
2133struct netdev_notifier_changeupper_info {
2134 struct netdev_notifier_info info; /* must be first */
2135 struct net_device *upper_dev; /* new upper dev */
2136 bool master; /* is upper dev master */
2137 bool linking; /* is the nofication for link or unlink */
2138};
2139
75538c2b
CW
2140static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2141 struct net_device *dev)
2142{
2143 info->dev = dev;
2144}
2145
351638e7
JP
2146static inline struct net_device *
2147netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2148{
2149 return info->dev;
2150}
2151
f629d208 2152int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2153
2154
1da177e4
LT
2155extern rwlock_t dev_base_lock; /* Device list lock */
2156
881d966b
EB
2157#define for_each_netdev(net, d) \
2158 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2159#define for_each_netdev_reverse(net, d) \
2160 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2161#define for_each_netdev_rcu(net, d) \
2162 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2163#define for_each_netdev_safe(net, d, n) \
2164 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2165#define for_each_netdev_continue(net, d) \
2166 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2167#define for_each_netdev_continue_rcu(net, d) \
2168 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2169#define for_each_netdev_in_bond_rcu(bond, slave) \
2170 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2171 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2172#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2173
a050c33f
DL
2174static inline struct net_device *next_net_device(struct net_device *dev)
2175{
2176 struct list_head *lh;
2177 struct net *net;
2178
c346dca1 2179 net = dev_net(dev);
a050c33f
DL
2180 lh = dev->dev_list.next;
2181 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2182}
2183
ce81b76a
ED
2184static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2185{
2186 struct list_head *lh;
2187 struct net *net;
2188
2189 net = dev_net(dev);
ccf43438 2190 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2191 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2192}
2193
a050c33f
DL
2194static inline struct net_device *first_net_device(struct net *net)
2195{
2196 return list_empty(&net->dev_base_head) ? NULL :
2197 net_device_entry(net->dev_base_head.next);
2198}
7562f876 2199
ccf43438
ED
2200static inline struct net_device *first_net_device_rcu(struct net *net)
2201{
2202 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2203
2204 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2205}
2206
f629d208
JP
2207int netdev_boot_setup_check(struct net_device *dev);
2208unsigned long netdev_boot_base(const char *prefix, int unit);
2209struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2210 const char *hwaddr);
2211struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2212struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2213void dev_add_pack(struct packet_type *pt);
2214void dev_remove_pack(struct packet_type *pt);
2215void __dev_remove_pack(struct packet_type *pt);
2216void dev_add_offload(struct packet_offload *po);
2217void dev_remove_offload(struct packet_offload *po);
f629d208 2218
a54acb3a 2219int dev_get_iflink(const struct net_device *dev);
6c555490
WC
2220struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2221 unsigned short mask);
f629d208
JP
2222struct net_device *dev_get_by_name(struct net *net, const char *name);
2223struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2224struct net_device *__dev_get_by_name(struct net *net, const char *name);
2225int dev_alloc_name(struct net_device *dev, const char *name);
2226int dev_open(struct net_device *dev);
2227int dev_close(struct net_device *dev);
99c4a26a 2228int dev_close_many(struct list_head *head, bool unlink);
f629d208 2229void dev_disable_lro(struct net_device *dev);
7026b1dd
DM
2230int dev_loopback_xmit(struct sock *sk, struct sk_buff *newskb);
2231int dev_queue_xmit_sk(struct sock *sk, struct sk_buff *skb);
2232static inline int dev_queue_xmit(struct sk_buff *skb)
2233{
2234 return dev_queue_xmit_sk(skb->sk, skb);
2235}
f663dd9a 2236int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2237int register_netdevice(struct net_device *dev);
2238void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2239void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2240static inline void unregister_netdevice(struct net_device *dev)
2241{
2242 unregister_netdevice_queue(dev, NULL);
2243}
2244
f629d208
JP
2245int netdev_refcnt_read(const struct net_device *dev);
2246void free_netdev(struct net_device *dev);
74d332c1 2247void netdev_freemem(struct net_device *dev);
f629d208
JP
2248void synchronize_net(void);
2249int init_dummy_netdev(struct net_device *dev);
937f1ba5 2250
f60e5990 2251DECLARE_PER_CPU(int, xmit_recursion);
2252static inline int dev_recursion_level(void)
2253{
2254 return this_cpu_read(xmit_recursion);
2255}
2256
f629d208
JP
2257struct net_device *dev_get_by_index(struct net *net, int ifindex);
2258struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2259struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2260int netdev_get_name(struct net *net, char *name, int ifindex);
2261int dev_restart(struct net_device *dev);
f629d208 2262int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2263
2264static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2265{
2266 return NAPI_GRO_CB(skb)->data_offset;
2267}
2268
2269static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2270{
2271 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2272}
2273
2274static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2275{
2276 NAPI_GRO_CB(skb)->data_offset += len;
2277}
2278
a5b1cf28
HX
2279static inline void *skb_gro_header_fast(struct sk_buff *skb,
2280 unsigned int offset)
86911732 2281{
a5b1cf28
HX
2282 return NAPI_GRO_CB(skb)->frag0 + offset;
2283}
78a478d0 2284
a5b1cf28
HX
2285static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2286{
2287 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2288}
78a478d0 2289
a5b1cf28
HX
2290static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2291 unsigned int offset)
2292{
17dd759c
HX
2293 if (!pskb_may_pull(skb, hlen))
2294 return NULL;
2295
a5b1cf28
HX
2296 NAPI_GRO_CB(skb)->frag0 = NULL;
2297 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2298 return skb->data + offset;
86911732 2299}
1da177e4 2300
36e7b1b8
HX
2301static inline void *skb_gro_network_header(struct sk_buff *skb)
2302{
78d3fd0b
HX
2303 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2304 skb_network_offset(skb);
36e7b1b8
HX
2305}
2306
bf5a755f
JC
2307static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2308 const void *start, unsigned int len)
2309{
573e8fca 2310 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2311 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2312 csum_partial(start, len, 0));
2313}
2314
573e8fca
TH
2315/* GRO checksum functions. These are logical equivalents of the normal
2316 * checksum functions (in skbuff.h) except that they operate on the GRO
2317 * offsets and fields in sk_buff.
2318 */
2319
2320__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2321
15e2396d
TH
2322static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2323{
b7fe10e5 2324 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2325}
2326
573e8fca
TH
2327static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2328 bool zero_okay,
2329 __sum16 check)
2330{
6edec0e6
TH
2331 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2332 skb_checksum_start_offset(skb) <
2333 skb_gro_offset(skb)) &&
15e2396d 2334 !skb_at_gro_remcsum_start(skb) &&
662880f4 2335 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2336 (!zero_okay || check));
2337}
2338
2339static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2340 __wsum psum)
2341{
2342 if (NAPI_GRO_CB(skb)->csum_valid &&
2343 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2344 return 0;
2345
2346 NAPI_GRO_CB(skb)->csum = psum;
2347
2348 return __skb_gro_checksum_complete(skb);
2349}
2350
573e8fca
TH
2351static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2352{
662880f4
TH
2353 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2354 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2355 NAPI_GRO_CB(skb)->csum_cnt--;
2356 } else {
2357 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2358 * verified a new top level checksum or an encapsulated one
2359 * during GRO. This saves work if we fallback to normal path.
2360 */
2361 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2362 }
2363}
2364
2365#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2366 compute_pseudo) \
2367({ \
2368 __sum16 __ret = 0; \
2369 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2370 __ret = __skb_gro_checksum_validate_complete(skb, \
2371 compute_pseudo(skb, proto)); \
5a212329
TH
2372 if (__ret) \
2373 __skb_mark_checksum_bad(skb); \
2374 else \
573e8fca
TH
2375 skb_gro_incr_csum_unnecessary(skb); \
2376 __ret; \
2377})
2378
2379#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2380 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2381
2382#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2383 compute_pseudo) \
2384 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2385
2386#define skb_gro_checksum_simple_validate(skb) \
2387 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2388
d96535a1
TH
2389static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2390{
2391 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2392 !NAPI_GRO_CB(skb)->csum_valid);
2393}
2394
2395static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2396 __sum16 check, __wsum pseudo)
2397{
2398 NAPI_GRO_CB(skb)->csum = ~pseudo;
2399 NAPI_GRO_CB(skb)->csum_valid = 1;
2400}
2401
2402#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2403do { \
2404 if (__skb_gro_checksum_convert_check(skb)) \
2405 __skb_gro_checksum_convert(skb, check, \
2406 compute_pseudo(skb, proto)); \
2407} while (0)
2408
26c4f7da
TH
2409struct gro_remcsum {
2410 int offset;
2411 __wsum delta;
2412};
2413
2414static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2415{
846cd667 2416 grc->offset = 0;
26c4f7da
TH
2417 grc->delta = 0;
2418}
2419
b7fe10e5
TH
2420static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2421 unsigned int off, size_t hdrlen,
2422 int start, int offset,
2423 struct gro_remcsum *grc,
2424 bool nopartial)
dcdc8994
TH
2425{
2426 __wsum delta;
b7fe10e5 2427 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2428
2429 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2430
15e2396d 2431 if (!nopartial) {
b7fe10e5
TH
2432 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2433 return ptr;
2434 }
2435
2436 ptr = skb_gro_header_fast(skb, off);
2437 if (skb_gro_header_hard(skb, off + plen)) {
2438 ptr = skb_gro_header_slow(skb, off + plen, off);
2439 if (!ptr)
2440 return NULL;
15e2396d
TH
2441 }
2442
b7fe10e5
TH
2443 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2444 start, offset);
dcdc8994
TH
2445
2446 /* Adjust skb->csum since we changed the packet */
dcdc8994 2447 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2448
b7fe10e5 2449 grc->offset = off + hdrlen + offset;
26c4f7da 2450 grc->delta = delta;
b7fe10e5
TH
2451
2452 return ptr;
dcdc8994
TH
2453}
2454
26c4f7da
TH
2455static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2456 struct gro_remcsum *grc)
2457{
b7fe10e5
TH
2458 void *ptr;
2459 size_t plen = grc->offset + sizeof(u16);
2460
26c4f7da
TH
2461 if (!grc->delta)
2462 return;
2463
b7fe10e5
TH
2464 ptr = skb_gro_header_fast(skb, grc->offset);
2465 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2466 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2467 if (!ptr)
2468 return;
2469 }
2470
2471 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2472}
dcdc8994 2473
0c4e8581
SH
2474static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2475 unsigned short type,
3b04ddde 2476 const void *daddr, const void *saddr,
95c96174 2477 unsigned int len)
0c4e8581 2478{
f1ecfd5d 2479 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2480 return 0;
3b04ddde
SH
2481
2482 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2483}
2484
b95cce35
SH
2485static inline int dev_parse_header(const struct sk_buff *skb,
2486 unsigned char *haddr)
2487{
2488 const struct net_device *dev = skb->dev;
2489
1b83336b 2490 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2491 return 0;
3b04ddde 2492 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2493}
2494
1da177e4 2495typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2496int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2497static inline int unregister_gifconf(unsigned int family)
2498{
2499 return register_gifconf(family, NULL);
2500}
2501
99bbc707 2502#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2503#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2504struct sd_flow_limit {
2505 u64 count;
2506 unsigned int num_buckets;
2507 unsigned int history_head;
2508 u16 history[FLOW_LIMIT_HISTORY];
2509 u8 buckets[];
2510};
2511
2512extern int netdev_flow_limit_table_len;
2513#endif /* CONFIG_NET_FLOW_LIMIT */
2514
1da177e4 2515/*
88751275 2516 * Incoming packets are placed on per-cpu queues
1da177e4 2517 */
d94d9fee 2518struct softnet_data {
1da177e4 2519 struct list_head poll_list;
6e7676c1 2520 struct sk_buff_head process_queue;
1da177e4 2521
dee42870 2522 /* stats */
cd7b5396
DM
2523 unsigned int processed;
2524 unsigned int time_squeeze;
2525 unsigned int cpu_collision;
2526 unsigned int received_rps;
fd793d89 2527#ifdef CONFIG_RPS
88751275 2528 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2529#endif
2530#ifdef CONFIG_NET_FLOW_LIMIT
2531 struct sd_flow_limit __rcu *flow_limit;
2532#endif
2533 struct Qdisc *output_queue;
2534 struct Qdisc **output_queue_tailp;
2535 struct sk_buff *completion_queue;
88751275 2536
4cdb1e2e 2537#ifdef CONFIG_RPS
88751275 2538 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 2539 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2540 struct softnet_data *rps_ipi_next;
2541 unsigned int cpu;
fec5e652 2542 unsigned int input_queue_head;
76cc8b13 2543 unsigned int input_queue_tail;
1e94d72f 2544#endif
95c96174 2545 unsigned int dropped;
0a9627f2 2546 struct sk_buff_head input_pkt_queue;
bea3348e 2547 struct napi_struct backlog;
99bbc707 2548
1da177e4
LT
2549};
2550
76cc8b13 2551static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2552{
2553#ifdef CONFIG_RPS
76cc8b13
TH
2554 sd->input_queue_head++;
2555#endif
2556}
2557
2558static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2559 unsigned int *qtail)
2560{
2561#ifdef CONFIG_RPS
2562 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2563#endif
2564}
2565
0a9627f2 2566DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2567
f629d208 2568void __netif_schedule(struct Qdisc *q);
46e5da40 2569void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2570
fd2ea0a7
DM
2571static inline void netif_tx_schedule_all(struct net_device *dev)
2572{
2573 unsigned int i;
2574
2575 for (i = 0; i < dev->num_tx_queues; i++)
2576 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2577}
2578
d29f749e
DJ
2579static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2580{
73466498 2581 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2582}
2583
bea3348e
SH
2584/**
2585 * netif_start_queue - allow transmit
2586 * @dev: network device
2587 *
2588 * Allow upper layers to call the device hard_start_xmit routine.
2589 */
1da177e4
LT
2590static inline void netif_start_queue(struct net_device *dev)
2591{
e8a0464c 2592 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2593}
2594
fd2ea0a7
DM
2595static inline void netif_tx_start_all_queues(struct net_device *dev)
2596{
2597 unsigned int i;
2598
2599 for (i = 0; i < dev->num_tx_queues; i++) {
2600 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2601 netif_tx_start_queue(txq);
2602 }
2603}
2604
46e5da40 2605void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2606
d29f749e
DJ
2607/**
2608 * netif_wake_queue - restart transmit
2609 * @dev: network device
2610 *
2611 * Allow upper layers to call the device hard_start_xmit routine.
2612 * Used for flow control when transmit resources are available.
2613 */
79d16385
DM
2614static inline void netif_wake_queue(struct net_device *dev)
2615{
e8a0464c 2616 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2617}
2618
fd2ea0a7
DM
2619static inline void netif_tx_wake_all_queues(struct net_device *dev)
2620{
2621 unsigned int i;
2622
2623 for (i = 0; i < dev->num_tx_queues; i++) {
2624 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2625 netif_tx_wake_queue(txq);
2626 }
2627}
2628
d29f749e
DJ
2629static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2630{
73466498 2631 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2632}
2633
bea3348e
SH
2634/**
2635 * netif_stop_queue - stop transmitted packets
2636 * @dev: network device
2637 *
2638 * Stop upper layers calling the device hard_start_xmit routine.
2639 * Used for flow control when transmit resources are unavailable.
2640 */
1da177e4
LT
2641static inline void netif_stop_queue(struct net_device *dev)
2642{
e8a0464c 2643 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2644}
2645
a2029240 2646void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2647
4d29515f 2648static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2649{
73466498 2650 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2651}
2652
bea3348e
SH
2653/**
2654 * netif_queue_stopped - test if transmit queue is flowblocked
2655 * @dev: network device
2656 *
2657 * Test if transmit queue on device is currently unable to send.
2658 */
4d29515f 2659static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2660{
e8a0464c 2661 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2662}
2663
4d29515f 2664static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2665{
73466498
TH
2666 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2667}
2668
8e2f1a63
DB
2669static inline bool
2670netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2671{
2672 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2673}
2674
8e2f1a63
DB
2675static inline bool
2676netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2677{
2678 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2679}
2680
53511453
ED
2681/**
2682 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2683 * @dev_queue: pointer to transmit queue
2684 *
2685 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2686 * to give appropriate hint to the cpu.
2687 */
2688static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2689{
2690#ifdef CONFIG_BQL
2691 prefetchw(&dev_queue->dql.num_queued);
2692#endif
2693}
2694
2695/**
2696 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2697 * @dev_queue: pointer to transmit queue
2698 *
2699 * BQL enabled drivers might use this helper in their TX completion path,
2700 * to give appropriate hint to the cpu.
2701 */
2702static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2703{
2704#ifdef CONFIG_BQL
2705 prefetchw(&dev_queue->dql.limit);
2706#endif
2707}
2708
c5d67bd7
TH
2709static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2710 unsigned int bytes)
2711{
114cf580
TH
2712#ifdef CONFIG_BQL
2713 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2714
2715 if (likely(dql_avail(&dev_queue->dql) >= 0))
2716 return;
2717
2718 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2719
2720 /*
2721 * The XOFF flag must be set before checking the dql_avail below,
2722 * because in netdev_tx_completed_queue we update the dql_completed
2723 * before checking the XOFF flag.
2724 */
2725 smp_mb();
2726
2727 /* check again in case another CPU has just made room avail */
2728 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2729 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2730#endif
c5d67bd7
TH
2731}
2732
0042d0c8
FF
2733/**
2734 * netdev_sent_queue - report the number of bytes queued to hardware
2735 * @dev: network device
2736 * @bytes: number of bytes queued to the hardware device queue
2737 *
2738 * Report the number of bytes queued for sending/completion to the network
2739 * device hardware queue. @bytes should be a good approximation and should
2740 * exactly match netdev_completed_queue() @bytes
2741 */
c5d67bd7
TH
2742static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2743{
2744 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2745}
2746
2747static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2748 unsigned int pkts, unsigned int bytes)
c5d67bd7 2749{
114cf580 2750#ifdef CONFIG_BQL
b37c0fbe
AD
2751 if (unlikely(!bytes))
2752 return;
2753
2754 dql_completed(&dev_queue->dql, bytes);
2755
2756 /*
2757 * Without the memory barrier there is a small possiblity that
2758 * netdev_tx_sent_queue will miss the update and cause the queue to
2759 * be stopped forever
2760 */
2761 smp_mb();
2762
2763 if (dql_avail(&dev_queue->dql) < 0)
2764 return;
2765
2766 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2767 netif_schedule_queue(dev_queue);
114cf580 2768#endif
c5d67bd7
TH
2769}
2770
0042d0c8
FF
2771/**
2772 * netdev_completed_queue - report bytes and packets completed by device
2773 * @dev: network device
2774 * @pkts: actual number of packets sent over the medium
2775 * @bytes: actual number of bytes sent over the medium
2776 *
2777 * Report the number of bytes and packets transmitted by the network device
2778 * hardware queue over the physical medium, @bytes must exactly match the
2779 * @bytes amount passed to netdev_sent_queue()
2780 */
c5d67bd7 2781static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2782 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2783{
2784 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2785}
2786
2787static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2788{
114cf580 2789#ifdef CONFIG_BQL
5c490354 2790 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2791 dql_reset(&q->dql);
2792#endif
c5d67bd7
TH
2793}
2794
0042d0c8
FF
2795/**
2796 * netdev_reset_queue - reset the packets and bytes count of a network device
2797 * @dev_queue: network device
2798 *
2799 * Reset the bytes and packet count of a network device and clear the
2800 * software flow control OFF bit for this network device
2801 */
c5d67bd7
TH
2802static inline void netdev_reset_queue(struct net_device *dev_queue)
2803{
2804 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2805}
2806
b9507bda
DB
2807/**
2808 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2809 * @dev: network device
2810 * @queue_index: given tx queue index
2811 *
2812 * Returns 0 if given tx queue index >= number of device tx queues,
2813 * otherwise returns the originally passed tx queue index.
2814 */
2815static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2816{
2817 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2818 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2819 dev->name, queue_index,
2820 dev->real_num_tx_queues);
2821 return 0;
2822 }
2823
2824 return queue_index;
2825}
2826
bea3348e
SH
2827/**
2828 * netif_running - test if up
2829 * @dev: network device
2830 *
2831 * Test if the device has been brought up.
2832 */
4d29515f 2833static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2834{
2835 return test_bit(__LINK_STATE_START, &dev->state);
2836}
2837
f25f4e44
PWJ
2838/*
2839 * Routines to manage the subqueues on a device. We only need start
2840 * stop, and a check if it's stopped. All other device management is
2841 * done at the overall netdevice level.
2842 * Also test the device if we're multiqueue.
2843 */
bea3348e
SH
2844
2845/**
2846 * netif_start_subqueue - allow sending packets on subqueue
2847 * @dev: network device
2848 * @queue_index: sub queue index
2849 *
2850 * Start individual transmit queue of a device with multiple transmit queues.
2851 */
f25f4e44
PWJ
2852static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2853{
fd2ea0a7 2854 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2855
2856 netif_tx_start_queue(txq);
f25f4e44
PWJ
2857}
2858
bea3348e
SH
2859/**
2860 * netif_stop_subqueue - stop sending packets on subqueue
2861 * @dev: network device
2862 * @queue_index: sub queue index
2863 *
2864 * Stop individual transmit queue of a device with multiple transmit queues.
2865 */
f25f4e44
PWJ
2866static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2867{
fd2ea0a7 2868 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 2869 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2870}
2871
bea3348e
SH
2872/**
2873 * netif_subqueue_stopped - test status of subqueue
2874 * @dev: network device
2875 * @queue_index: sub queue index
2876 *
2877 * Check individual transmit queue of a device with multiple transmit queues.
2878 */
4d29515f
DM
2879static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2880 u16 queue_index)
f25f4e44 2881{
fd2ea0a7 2882 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2883
2884 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2885}
2886
4d29515f
DM
2887static inline bool netif_subqueue_stopped(const struct net_device *dev,
2888 struct sk_buff *skb)
668f895a
PE
2889{
2890 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2891}
bea3348e 2892
46e5da40 2893void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 2894
537c00de 2895#ifdef CONFIG_XPS
53af53ae 2896int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2897 u16 index);
537c00de
AD
2898#else
2899static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2900 const struct cpumask *mask,
537c00de
AD
2901 u16 index)
2902{
2903 return 0;
2904}
2905#endif
2906
5605c762
JP
2907u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2908 unsigned int num_tx_queues);
2909
a3d22a68
VZ
2910/*
2911 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2912 * as a distribution range limit for the returned value.
2913 */
2914static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 2915 struct sk_buff *skb)
a3d22a68
VZ
2916{
2917 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2918}
2919
bea3348e
SH
2920/**
2921 * netif_is_multiqueue - test if device has multiple transmit queues
2922 * @dev: network device
2923 *
2924 * Check if device has multiple transmit queues
bea3348e 2925 */
4d29515f 2926static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2927{
a02cec21 2928 return dev->num_tx_queues > 1;
f25f4e44 2929}
1da177e4 2930
f629d208 2931int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2932
a953be53 2933#ifdef CONFIG_SYSFS
f629d208 2934int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
2935#else
2936static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2937 unsigned int rxq)
2938{
2939 return 0;
2940}
2941#endif
2942
a953be53
MD
2943#ifdef CONFIG_SYSFS
2944static inline unsigned int get_netdev_rx_queue_index(
2945 struct netdev_rx_queue *queue)
2946{
2947 struct net_device *dev = queue->dev;
2948 int index = queue - dev->_rx;
2949
2950 BUG_ON(index >= dev->num_rx_queues);
2951 return index;
2952}
2953#endif
2954
16917b87 2955#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 2956int netif_get_num_default_rss_queues(void);
16917b87 2957
e6247027
ED
2958enum skb_free_reason {
2959 SKB_REASON_CONSUMED,
2960 SKB_REASON_DROPPED,
2961};
2962
2963void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2964void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 2965
e6247027
ED
2966/*
2967 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2968 * interrupt context or with hardware interrupts being disabled.
2969 * (in_irq() || irqs_disabled())
2970 *
2971 * We provide four helpers that can be used in following contexts :
2972 *
2973 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2974 * replacing kfree_skb(skb)
2975 *
2976 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2977 * Typically used in place of consume_skb(skb) in TX completion path
2978 *
2979 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2980 * replacing kfree_skb(skb)
2981 *
2982 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2983 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 2984 */
e6247027
ED
2985static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2986{
2987 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2988}
2989
2990static inline void dev_consume_skb_irq(struct sk_buff *skb)
2991{
2992 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2993}
2994
2995static inline void dev_kfree_skb_any(struct sk_buff *skb)
2996{
2997 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2998}
2999
3000static inline void dev_consume_skb_any(struct sk_buff *skb)
3001{
3002 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3003}
1da177e4 3004
f629d208
JP
3005int netif_rx(struct sk_buff *skb);
3006int netif_rx_ni(struct sk_buff *skb);
7026b1dd
DM
3007int netif_receive_skb_sk(struct sock *sk, struct sk_buff *skb);
3008static inline int netif_receive_skb(struct sk_buff *skb)
3009{
3010 return netif_receive_skb_sk(skb->sk, skb);
3011}
f629d208
JP
3012gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3013void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3014struct sk_buff *napi_get_frags(struct napi_struct *napi);
3015gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3016struct packet_offload *gro_find_receive_by_type(__be16 type);
3017struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3018
3019static inline void napi_free_frags(struct napi_struct *napi)
3020{
3021 kfree_skb(napi->skb);
3022 napi->skb = NULL;
3023}
3024
f629d208
JP
3025int netdev_rx_handler_register(struct net_device *dev,
3026 rx_handler_func_t *rx_handler,
3027 void *rx_handler_data);
3028void netdev_rx_handler_unregister(struct net_device *dev);
3029
3030bool dev_valid_name(const char *name);
3031int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3032int dev_ethtool(struct net *net, struct ifreq *);
3033unsigned int dev_get_flags(const struct net_device *);
3034int __dev_change_flags(struct net_device *, unsigned int flags);
3035int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3036void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3037 unsigned int gchanges);
f629d208
JP
3038int dev_change_name(struct net_device *, const char *);
3039int dev_set_alias(struct net_device *, const char *, size_t);
3040int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3041int dev_set_mtu(struct net_device *, int);
3042void dev_set_group(struct net_device *, int);
3043int dev_set_mac_address(struct net_device *, struct sockaddr *);
3044int dev_change_carrier(struct net_device *, bool new_carrier);
3045int dev_get_phys_port_id(struct net_device *dev,
02637fce 3046 struct netdev_phys_item_id *ppid);
db24a904
DA
3047int dev_get_phys_port_name(struct net_device *dev,
3048 char *name, size_t len);
d746d707 3049int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3050struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3051struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3052 struct netdev_queue *txq, int *ret);
a0265d28 3053int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3054int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1ee481fb 3055bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
1da177e4 3056
20380731 3057extern int netdev_budget;
1da177e4
LT
3058
3059/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3060void netdev_run_todo(void);
1da177e4 3061
bea3348e
SH
3062/**
3063 * dev_put - release reference to device
3064 * @dev: network device
3065 *
9ef4429b 3066 * Release reference to device to allow it to be freed.
bea3348e 3067 */
1da177e4
LT
3068static inline void dev_put(struct net_device *dev)
3069{
933393f5 3070 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3071}
3072
bea3348e
SH
3073/**
3074 * dev_hold - get reference to device
3075 * @dev: network device
3076 *
9ef4429b 3077 * Hold reference to device to keep it from being freed.
bea3348e 3078 */
15333061
SH
3079static inline void dev_hold(struct net_device *dev)
3080{
933393f5 3081 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3082}
1da177e4
LT
3083
3084/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3085 * and _off may be called from IRQ context, but it is caller
3086 * who is responsible for serialization of these calls.
b00055aa
SR
3087 *
3088 * The name carrier is inappropriate, these functions should really be
3089 * called netif_lowerlayer_*() because they represent the state of any
3090 * kind of lower layer not just hardware media.
1da177e4
LT
3091 */
3092
f629d208
JP
3093void linkwatch_init_dev(struct net_device *dev);
3094void linkwatch_fire_event(struct net_device *dev);
3095void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3096
bea3348e
SH
3097/**
3098 * netif_carrier_ok - test if carrier present
3099 * @dev: network device
3100 *
3101 * Check if carrier is present on device
3102 */
4d29515f 3103static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3104{
3105 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3106}
3107
f629d208 3108unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3109
f629d208 3110void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3111
f629d208 3112void netif_carrier_on(struct net_device *dev);
1da177e4 3113
f629d208 3114void netif_carrier_off(struct net_device *dev);
1da177e4 3115
bea3348e
SH
3116/**
3117 * netif_dormant_on - mark device as dormant.
3118 * @dev: network device
3119 *
3120 * Mark device as dormant (as per RFC2863).
3121 *
3122 * The dormant state indicates that the relevant interface is not
3123 * actually in a condition to pass packets (i.e., it is not 'up') but is
3124 * in a "pending" state, waiting for some external event. For "on-
3125 * demand" interfaces, this new state identifies the situation where the
3126 * interface is waiting for events to place it in the up state.
3127 *
3128 */
b00055aa
SR
3129static inline void netif_dormant_on(struct net_device *dev)
3130{
3131 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3132 linkwatch_fire_event(dev);
3133}
3134
bea3348e
SH
3135/**
3136 * netif_dormant_off - set device as not dormant.
3137 * @dev: network device
3138 *
3139 * Device is not in dormant state.
3140 */
b00055aa
SR
3141static inline void netif_dormant_off(struct net_device *dev)
3142{
3143 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3144 linkwatch_fire_event(dev);
3145}
3146
bea3348e
SH
3147/**
3148 * netif_dormant - test if carrier present
3149 * @dev: network device
3150 *
3151 * Check if carrier is present on device
3152 */
4d29515f 3153static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3154{
3155 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3156}
3157
3158
bea3348e
SH
3159/**
3160 * netif_oper_up - test if device is operational
3161 * @dev: network device
3162 *
3163 * Check if carrier is operational
3164 */
4d29515f 3165static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3166{
b00055aa
SR
3167 return (dev->operstate == IF_OPER_UP ||
3168 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3169}
3170
bea3348e
SH
3171/**
3172 * netif_device_present - is device available or removed
3173 * @dev: network device
3174 *
3175 * Check if device has not been removed from system.
3176 */
4d29515f 3177static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3178{
3179 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3180}
3181
f629d208 3182void netif_device_detach(struct net_device *dev);
1da177e4 3183
f629d208 3184void netif_device_attach(struct net_device *dev);
1da177e4
LT
3185
3186/*
3187 * Network interface message level settings
3188 */
1da177e4
LT
3189
3190enum {
3191 NETIF_MSG_DRV = 0x0001,
3192 NETIF_MSG_PROBE = 0x0002,
3193 NETIF_MSG_LINK = 0x0004,
3194 NETIF_MSG_TIMER = 0x0008,
3195 NETIF_MSG_IFDOWN = 0x0010,
3196 NETIF_MSG_IFUP = 0x0020,
3197 NETIF_MSG_RX_ERR = 0x0040,
3198 NETIF_MSG_TX_ERR = 0x0080,
3199 NETIF_MSG_TX_QUEUED = 0x0100,
3200 NETIF_MSG_INTR = 0x0200,
3201 NETIF_MSG_TX_DONE = 0x0400,
3202 NETIF_MSG_RX_STATUS = 0x0800,
3203 NETIF_MSG_PKTDATA = 0x1000,
3204 NETIF_MSG_HW = 0x2000,
3205 NETIF_MSG_WOL = 0x4000,
3206};
3207
3208#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3209#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3210#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3211#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3212#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3213#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3214#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3215#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3216#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3217#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3218#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3219#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3220#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3221#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3222#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3223
3224static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3225{
3226 /* use default */
3227 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3228 return default_msg_enable_bits;
3229 if (debug_value == 0) /* no output */
3230 return 0;
3231 /* set low N bits */
3232 return (1 << debug_value) - 1;
3233}
3234
c773e847 3235static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3236{
c773e847
DM
3237 spin_lock(&txq->_xmit_lock);
3238 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3239}
3240
fd2ea0a7
DM
3241static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3242{
3243 spin_lock_bh(&txq->_xmit_lock);
3244 txq->xmit_lock_owner = smp_processor_id();
3245}
3246
4d29515f 3247static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3248{
4d29515f 3249 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3250 if (likely(ok))
3251 txq->xmit_lock_owner = smp_processor_id();
3252 return ok;
3253}
3254
3255static inline void __netif_tx_unlock(struct netdev_queue *txq)
3256{
3257 txq->xmit_lock_owner = -1;
3258 spin_unlock(&txq->_xmit_lock);
3259}
3260
3261static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3262{
3263 txq->xmit_lock_owner = -1;
3264 spin_unlock_bh(&txq->_xmit_lock);
3265}
3266
08baf561
ED
3267static inline void txq_trans_update(struct netdev_queue *txq)
3268{
3269 if (txq->xmit_lock_owner != -1)
3270 txq->trans_start = jiffies;
3271}
3272
d29f749e
DJ
3273/**
3274 * netif_tx_lock - grab network device transmit lock
3275 * @dev: network device
d29f749e
DJ
3276 *
3277 * Get network device transmit lock
3278 */
22dd7495
JHS
3279static inline void netif_tx_lock(struct net_device *dev)
3280{
e8a0464c 3281 unsigned int i;
c3f26a26 3282 int cpu;
c773e847 3283
c3f26a26
DM
3284 spin_lock(&dev->tx_global_lock);
3285 cpu = smp_processor_id();
e8a0464c
DM
3286 for (i = 0; i < dev->num_tx_queues; i++) {
3287 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3288
3289 /* We are the only thread of execution doing a
3290 * freeze, but we have to grab the _xmit_lock in
3291 * order to synchronize with threads which are in
3292 * the ->hard_start_xmit() handler and already
3293 * checked the frozen bit.
3294 */
e8a0464c 3295 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3296 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3297 __netif_tx_unlock(txq);
e8a0464c 3298 }
932ff279
HX
3299}
3300
3301static inline void netif_tx_lock_bh(struct net_device *dev)
3302{
e8a0464c
DM
3303 local_bh_disable();
3304 netif_tx_lock(dev);
932ff279
HX
3305}
3306
932ff279
HX
3307static inline void netif_tx_unlock(struct net_device *dev)
3308{
e8a0464c
DM
3309 unsigned int i;
3310
3311 for (i = 0; i < dev->num_tx_queues; i++) {
3312 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3313
c3f26a26
DM
3314 /* No need to grab the _xmit_lock here. If the
3315 * queue is not stopped for another reason, we
3316 * force a schedule.
3317 */
3318 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3319 netif_schedule_queue(txq);
c3f26a26
DM
3320 }
3321 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3322}
3323
3324static inline void netif_tx_unlock_bh(struct net_device *dev)
3325{
e8a0464c
DM
3326 netif_tx_unlock(dev);
3327 local_bh_enable();
932ff279
HX
3328}
3329
c773e847 3330#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3331 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3332 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3333 } \
3334}
3335
5efeac44
EB
3336#define HARD_TX_TRYLOCK(dev, txq) \
3337 (((dev->features & NETIF_F_LLTX) == 0) ? \
3338 __netif_tx_trylock(txq) : \
3339 true )
3340
c773e847 3341#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3342 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3343 __netif_tx_unlock(txq); \
22dd7495
JHS
3344 } \
3345}
3346
1da177e4
LT
3347static inline void netif_tx_disable(struct net_device *dev)
3348{
fd2ea0a7 3349 unsigned int i;
c3f26a26 3350 int cpu;
fd2ea0a7 3351
c3f26a26
DM
3352 local_bh_disable();
3353 cpu = smp_processor_id();
fd2ea0a7
DM
3354 for (i = 0; i < dev->num_tx_queues; i++) {
3355 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3356
3357 __netif_tx_lock(txq, cpu);
fd2ea0a7 3358 netif_tx_stop_queue(txq);
c3f26a26 3359 __netif_tx_unlock(txq);
fd2ea0a7 3360 }
c3f26a26 3361 local_bh_enable();
1da177e4
LT
3362}
3363
e308a5d8
DM
3364static inline void netif_addr_lock(struct net_device *dev)
3365{
3366 spin_lock(&dev->addr_list_lock);
3367}
3368
2429f7ac
JP
3369static inline void netif_addr_lock_nested(struct net_device *dev)
3370{
25175ba5
VY
3371 int subclass = SINGLE_DEPTH_NESTING;
3372
3373 if (dev->netdev_ops->ndo_get_lock_subclass)
3374 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3375
3376 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3377}
3378
e308a5d8
DM
3379static inline void netif_addr_lock_bh(struct net_device *dev)
3380{
3381 spin_lock_bh(&dev->addr_list_lock);
3382}
3383
3384static inline void netif_addr_unlock(struct net_device *dev)
3385{
3386 spin_unlock(&dev->addr_list_lock);
3387}
3388
3389static inline void netif_addr_unlock_bh(struct net_device *dev)
3390{
3391 spin_unlock_bh(&dev->addr_list_lock);
3392}
3393
f001fde5 3394/*
31278e71 3395 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3396 * rcu_read_lock held.
3397 */
3398#define for_each_dev_addr(dev, ha) \
31278e71 3399 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3400
1da177e4
LT
3401/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3402
f629d208 3403void ether_setup(struct net_device *dev);
1da177e4
LT
3404
3405/* Support for loadable net-drivers */
f629d208 3406struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3407 unsigned char name_assign_type,
f629d208
JP
3408 void (*setup)(struct net_device *),
3409 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3410#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3411 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3412
c835a677
TG
3413#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3414 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3415 count)
36909ea4 3416
f629d208
JP
3417int register_netdev(struct net_device *dev);
3418void unregister_netdev(struct net_device *dev);
f001fde5 3419
22bedad3 3420/* General hardware address lists handling functions */
f629d208
JP
3421int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3422 struct netdev_hw_addr_list *from_list, int addr_len);
3423void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3424 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3425int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3426 struct net_device *dev,
3427 int (*sync)(struct net_device *, const unsigned char *),
3428 int (*unsync)(struct net_device *,
3429 const unsigned char *));
3430void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3431 struct net_device *dev,
3432 int (*unsync)(struct net_device *,
3433 const unsigned char *));
f629d208 3434void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3435
f001fde5 3436/* Functions used for device addresses handling */
f629d208
JP
3437int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3438 unsigned char addr_type);
3439int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3440 unsigned char addr_type);
f629d208
JP
3441void dev_addr_flush(struct net_device *dev);
3442int dev_addr_init(struct net_device *dev);
a748ee24
JP
3443
3444/* Functions used for unicast addresses handling */
f629d208
JP
3445int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3446int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3447int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3448int dev_uc_sync(struct net_device *to, struct net_device *from);
3449int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3450void dev_uc_unsync(struct net_device *to, struct net_device *from);
3451void dev_uc_flush(struct net_device *dev);
3452void dev_uc_init(struct net_device *dev);
f001fde5 3453
670e5b8e
AD
3454/**
3455 * __dev_uc_sync - Synchonize device's unicast list
3456 * @dev: device to sync
3457 * @sync: function to call if address should be added
3458 * @unsync: function to call if address should be removed
3459 *
3460 * Add newly added addresses to the interface, and release
3461 * addresses that have been deleted.
3462 **/
3463static inline int __dev_uc_sync(struct net_device *dev,
3464 int (*sync)(struct net_device *,
3465 const unsigned char *),
3466 int (*unsync)(struct net_device *,
3467 const unsigned char *))
3468{
3469 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3470}
3471
3472/**
e793c0f7 3473 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3474 * @dev: device to sync
3475 * @unsync: function to call if address should be removed
3476 *
3477 * Remove all addresses that were added to the device by dev_uc_sync().
3478 **/
3479static inline void __dev_uc_unsync(struct net_device *dev,
3480 int (*unsync)(struct net_device *,
3481 const unsigned char *))
3482{
3483 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3484}
3485
22bedad3 3486/* Functions used for multicast addresses handling */
f629d208
JP
3487int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3488int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3489int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3490int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3491int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3492int dev_mc_sync(struct net_device *to, struct net_device *from);
3493int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3494void dev_mc_unsync(struct net_device *to, struct net_device *from);
3495void dev_mc_flush(struct net_device *dev);
3496void dev_mc_init(struct net_device *dev);
f001fde5 3497
670e5b8e
AD
3498/**
3499 * __dev_mc_sync - Synchonize device's multicast list
3500 * @dev: device to sync
3501 * @sync: function to call if address should be added
3502 * @unsync: function to call if address should be removed
3503 *
3504 * Add newly added addresses to the interface, and release
3505 * addresses that have been deleted.
3506 **/
3507static inline int __dev_mc_sync(struct net_device *dev,
3508 int (*sync)(struct net_device *,
3509 const unsigned char *),
3510 int (*unsync)(struct net_device *,
3511 const unsigned char *))
3512{
3513 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3514}
3515
3516/**
e793c0f7 3517 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3518 * @dev: device to sync
3519 * @unsync: function to call if address should be removed
3520 *
3521 * Remove all addresses that were added to the device by dev_mc_sync().
3522 **/
3523static inline void __dev_mc_unsync(struct net_device *dev,
3524 int (*unsync)(struct net_device *,
3525 const unsigned char *))
3526{
3527 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3528}
3529
4417da66 3530/* Functions used for secondary unicast and multicast support */
f629d208
JP
3531void dev_set_rx_mode(struct net_device *dev);
3532void __dev_set_rx_mode(struct net_device *dev);
3533int dev_set_promiscuity(struct net_device *dev, int inc);
3534int dev_set_allmulti(struct net_device *dev, int inc);
3535void netdev_state_change(struct net_device *dev);
3536void netdev_notify_peers(struct net_device *dev);
3537void netdev_features_change(struct net_device *dev);
1da177e4 3538/* Load a device via the kmod */
f629d208
JP
3539void dev_load(struct net *net, const char *name);
3540struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3541 struct rtnl_link_stats64 *storage);
3542void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3543 const struct net_device_stats *netdev_stats);
eeda3fd6 3544
1da177e4 3545extern int netdev_max_backlog;
3b098e2d 3546extern int netdev_tstamp_prequeue;
1da177e4 3547extern int weight_p;
0a14842f 3548extern int bpf_jit_enable;
9ff162a8 3549
f629d208 3550bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3551struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3552 struct list_head **iter);
f629d208
JP
3553struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3554 struct list_head **iter);
8b5be856 3555
44a40855
VY
3556/* iterate through upper list, must be called under RCU read lock */
3557#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3558 for (iter = &(dev)->adj_list.upper, \
3559 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3560 updev; \
3561 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3562
8b5be856 3563/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3564#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3565 for (iter = &(dev)->all_adj_list.upper, \
3566 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3567 updev; \
3568 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3569
f629d208
JP
3570void *netdev_lower_get_next_private(struct net_device *dev,
3571 struct list_head **iter);
3572void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3573 struct list_head **iter);
31088a11
VF
3574
3575#define netdev_for_each_lower_private(dev, priv, iter) \
3576 for (iter = (dev)->adj_list.lower.next, \
3577 priv = netdev_lower_get_next_private(dev, &(iter)); \
3578 priv; \
3579 priv = netdev_lower_get_next_private(dev, &(iter)))
3580
3581#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3582 for (iter = &(dev)->adj_list.lower, \
3583 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3584 priv; \
3585 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3586
4085ebe8
VY
3587void *netdev_lower_get_next(struct net_device *dev,
3588 struct list_head **iter);
3589#define netdev_for_each_lower_dev(dev, ldev, iter) \
3590 for (iter = &(dev)->adj_list.lower, \
3591 ldev = netdev_lower_get_next(dev, &(iter)); \
3592 ldev; \
3593 ldev = netdev_lower_get_next(dev, &(iter)))
3594
f629d208 3595void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3596void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3597struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3598struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3599int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3600int netdev_master_upper_dev_link(struct net_device *dev,
9ff162a8 3601 struct net_device *upper_dev);
f629d208
JP
3602int netdev_master_upper_dev_link_private(struct net_device *dev,
3603 struct net_device *upper_dev,
3604 void *private);
3605void netdev_upper_dev_unlink(struct net_device *dev,
3606 struct net_device *upper_dev);
5bb025fa 3607void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3608void *netdev_lower_dev_get_private(struct net_device *dev,
3609 struct net_device *lower_dev);
960fb622
ED
3610
3611/* RSS keys are 40 or 52 bytes long */
3612#define NETDEV_RSS_KEY_LEN 52
3613extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3614void netdev_rss_key_fill(void *buffer, size_t len);
3615
4085ebe8
VY
3616int dev_get_nest_level(struct net_device *dev,
3617 bool (*type_check)(struct net_device *dev));
f629d208
JP
3618int skb_checksum_help(struct sk_buff *skb);
3619struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3620 netdev_features_t features, bool tx_path);
3621struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3622 netdev_features_t features);
12b0004d 3623
61bd3857
MS
3624struct netdev_bonding_info {
3625 ifslave slave;
3626 ifbond master;
3627};
3628
3629struct netdev_notifier_bonding_info {
3630 struct netdev_notifier_info info; /* must be first */
3631 struct netdev_bonding_info bonding_info;
3632};
3633
3634void netdev_bonding_info_change(struct net_device *dev,
3635 struct netdev_bonding_info *bonding_info);
3636
12b0004d
CW
3637static inline
3638struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3639{
3640 return __skb_gso_segment(skb, features, true);
3641}
53d6471c 3642__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3643
3644static inline bool can_checksum_protocol(netdev_features_t features,
3645 __be16 protocol)
3646{
3647 return ((features & NETIF_F_GEN_CSUM) ||
3648 ((features & NETIF_F_V4_CSUM) &&
3649 protocol == htons(ETH_P_IP)) ||
3650 ((features & NETIF_F_V6_CSUM) &&
3651 protocol == htons(ETH_P_IPV6)) ||
3652 ((features & NETIF_F_FCOE_CRC) &&
3653 protocol == htons(ETH_P_FCOE)));
3654}
12b0004d 3655
fb286bb2 3656#ifdef CONFIG_BUG
f629d208 3657void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3658#else
3659static inline void netdev_rx_csum_fault(struct net_device *dev)
3660{
3661}
3662#endif
1da177e4 3663/* rx skb timestamps */
f629d208
JP
3664void net_enable_timestamp(void);
3665void net_disable_timestamp(void);
1da177e4 3666
20380731 3667#ifdef CONFIG_PROC_FS
f629d208 3668int __init dev_proc_init(void);
900ff8c6
CW
3669#else
3670#define dev_proc_init() 0
20380731
ACM
3671#endif
3672
4798248e 3673static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3674 struct sk_buff *skb, struct net_device *dev,
3675 bool more)
4798248e 3676{
fa2dbdc2 3677 skb->xmit_more = more ? 1 : 0;
0b725a2c 3678 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3679}
3680
10b3ad8c 3681static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3682 struct netdev_queue *txq, bool more)
4798248e
DM
3683{
3684 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3685 int rc;
4798248e 3686
fa2dbdc2 3687 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3688 if (rc == NETDEV_TX_OK)
3689 txq_trans_update(txq);
3690
3691 return rc;
4798248e
DM
3692}
3693
42a2d923
LT
3694int netdev_class_create_file_ns(struct class_attribute *class_attr,
3695 const void *ns);
3696void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3697 const void *ns);
58292cbe
TH
3698
3699static inline int netdev_class_create_file(struct class_attribute *class_attr)
3700{
3701 return netdev_class_create_file_ns(class_attr, NULL);
3702}
3703
3704static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3705{
3706 netdev_class_remove_file_ns(class_attr, NULL);
3707}
b8a9787e 3708
04600794
JB
3709extern struct kobj_ns_type_operations net_ns_type_operations;
3710
f629d208 3711const char *netdev_drivername(const struct net_device *dev);
6579e57b 3712
f629d208 3713void linkwatch_run_queue(void);
20380731 3714
da08143b
MK
3715static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3716 netdev_features_t f2)
3717{
3718 if (f1 & NETIF_F_GEN_CSUM)
3719 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3720 if (f2 & NETIF_F_GEN_CSUM)
3721 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3722 f1 &= f2;
3723 if (f1 & NETIF_F_GEN_CSUM)
3724 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3725
3726 return f1;
3727}
3728
c8f44aff
MM
3729static inline netdev_features_t netdev_get_wanted_features(
3730 struct net_device *dev)
5455c699
MM
3731{
3732 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3733}
c8f44aff
MM
3734netdev_features_t netdev_increment_features(netdev_features_t all,
3735 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3736
3737/* Allow TSO being used on stacked device :
3738 * Performing the GSO segmentation before last device
3739 * is a performance improvement.
3740 */
3741static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3742 netdev_features_t mask)
3743{
3744 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3745}
3746
6cb6a27c 3747int __netdev_update_features(struct net_device *dev);
5455c699 3748void netdev_update_features(struct net_device *dev);
afe12cc8 3749void netdev_change_features(struct net_device *dev);
7f353bf2 3750
fc4a7489
PM
3751void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3752 struct net_device *dev);
3753
e38f3025
TM
3754netdev_features_t passthru_features_check(struct sk_buff *skb,
3755 struct net_device *dev,
3756 netdev_features_t features);
c1e756bf 3757netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3758
4d29515f 3759static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3760{
c8f44aff 3761 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3762
3763 /* check flags correspondence */
3764 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3765 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3766 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3767 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3768 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3769 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
3770 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3771 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3772 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3773 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3774 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3775 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
e585f236 3776 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
0345e186 3777
d6b4991a 3778 return (features & feature) == feature;
576a30eb
HX
3779}
3780
4d29515f 3781static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3782{
278b2513 3783 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3784 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3785}
3786
8b86a61d 3787static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 3788 netdev_features_t features)
7967168c 3789{
fc741216 3790 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3791 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3792 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3793}
3794
82cc1a7a
PWJ
3795static inline void netif_set_gso_max_size(struct net_device *dev,
3796 unsigned int size)
3797{
3798 dev->gso_max_size = size;
3799}
3800
7a7ffbab
WCC
3801static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3802 int pulled_hlen, u16 mac_offset,
3803 int mac_len)
3804{
3805 skb->protocol = protocol;
3806 skb->encapsulation = 1;
3807 skb_push(skb, pulled_hlen);
3808 skb_reset_transport_header(skb);
3809 skb->mac_header = mac_offset;
3810 skb->network_header = skb->mac_header + mac_len;
3811 skb->mac_len = mac_len;
3812}
3813
a6cc0cfa
JF
3814static inline bool netif_is_macvlan(struct net_device *dev)
3815{
3816 return dev->priv_flags & IFF_MACVLAN;
3817}
3818
2f33e7d5
MB
3819static inline bool netif_is_macvlan_port(struct net_device *dev)
3820{
3821 return dev->priv_flags & IFF_MACVLAN_PORT;
3822}
3823
5933fea7
MB
3824static inline bool netif_is_ipvlan(struct net_device *dev)
3825{
3826 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3827}
3828
3829static inline bool netif_is_ipvlan_port(struct net_device *dev)
3830{
3831 return dev->priv_flags & IFF_IPVLAN_MASTER;
3832}
3833
8a7fbfab 3834static inline bool netif_is_bond_master(struct net_device *dev)
3835{
3836 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3837}
3838
4d29515f 3839static inline bool netif_is_bond_slave(struct net_device *dev)
1765a575
JP
3840{
3841 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3842}
3843
3bdc0eba
BG
3844static inline bool netif_supports_nofcs(struct net_device *dev)
3845{
3846 return dev->priv_flags & IFF_SUPP_NOFCS;
3847}
3848
4e3c8992
DA
3849static inline bool netif_is_vrf(const struct net_device *dev)
3850{
3851 return dev->priv_flags & IFF_VRF_MASTER;
3852}
3853
0894ae3f
JP
3854static inline bool netif_is_bridge_master(const struct net_device *dev)
3855{
3856 return dev->priv_flags & IFF_EBRIDGE;
3857}
3858
35d4e172
JP
3859static inline bool netif_is_ovs_master(const struct net_device *dev)
3860{
3861 return dev->priv_flags & IFF_OPENVSWITCH;
3862}
3863
4e3c8992
DA
3864static inline bool netif_index_is_vrf(struct net *net, int ifindex)
3865{
4e3c8992
DA
3866 bool rc = false;
3867
2f52bdcf
DA
3868#if IS_ENABLED(CONFIG_NET_VRF)
3869 struct net_device *dev;
3870
3871 if (ifindex == 0)
3872 return false;
3873
3874 rcu_read_lock();
3875
3876 dev = dev_get_by_index_rcu(net, ifindex);
4e3c8992
DA
3877 if (dev)
3878 rc = netif_is_vrf(dev);
3879
2f52bdcf
DA
3880 rcu_read_unlock();
3881#endif
4e3c8992
DA
3882 return rc;
3883}
3884
02875878
ED
3885/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3886static inline void netif_keep_dst(struct net_device *dev)
3887{
3888 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3889}
3890
505d4f73 3891extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 3892
571ba423
JP
3893/* Logging, debugging and troubleshooting/diagnostic helpers. */
3894
3895/* netdev_printk helpers, similar to dev_printk */
3896
3897static inline const char *netdev_name(const struct net_device *dev)
3898{
c6f854d5
VF
3899 if (!dev->name[0] || strchr(dev->name, '%'))
3900 return "(unnamed net_device)";
571ba423
JP
3901 return dev->name;
3902}
3903
ccc7f496
VF
3904static inline const char *netdev_reg_state(const struct net_device *dev)
3905{
3906 switch (dev->reg_state) {
3907 case NETREG_UNINITIALIZED: return " (uninitialized)";
3908 case NETREG_REGISTERED: return "";
3909 case NETREG_UNREGISTERING: return " (unregistering)";
3910 case NETREG_UNREGISTERED: return " (unregistered)";
3911 case NETREG_RELEASED: return " (released)";
3912 case NETREG_DUMMY: return " (dummy)";
3913 }
3914
3915 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3916 return " (unknown)";
3917}
3918
f629d208 3919__printf(3, 4)
6ea754eb
JP
3920void netdev_printk(const char *level, const struct net_device *dev,
3921 const char *format, ...);
f629d208 3922__printf(2, 3)
6ea754eb 3923void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 3924__printf(2, 3)
6ea754eb 3925void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 3926__printf(2, 3)
6ea754eb 3927void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 3928__printf(2, 3)
6ea754eb 3929void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 3930__printf(2, 3)
6ea754eb 3931void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 3932__printf(2, 3)
6ea754eb 3933void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 3934__printf(2, 3)
6ea754eb 3935void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 3936
8909c9ad
VK
3937#define MODULE_ALIAS_NETDEV(device) \
3938 MODULE_ALIAS("netdev-" device)
3939
b558c96f 3940#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
3941#define netdev_dbg(__dev, format, args...) \
3942do { \
ffa10cb4 3943 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 3944} while (0)
b558c96f
JC
3945#elif defined(DEBUG)
3946#define netdev_dbg(__dev, format, args...) \
3947 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
3948#else
3949#define netdev_dbg(__dev, format, args...) \
3950({ \
3951 if (0) \
3952 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
3953})
3954#endif
3955
3956#if defined(VERBOSE_DEBUG)
3957#define netdev_vdbg netdev_dbg
3958#else
3959
3960#define netdev_vdbg(dev, format, args...) \
3961({ \
3962 if (0) \
3963 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3964 0; \
3965})
3966#endif
3967
3968/*
3969 * netdev_WARN() acts like dev_printk(), but with the key difference
3970 * of using a WARN/WARN_ON to get the message out, including the
3971 * file/line information and a backtrace.
3972 */
3973#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
3974 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3975 netdev_reg_state(dev), ##args)
571ba423 3976
b3d95c5c
JP
3977/* netif printk helpers, similar to netdev_printk */
3978
3979#define netif_printk(priv, type, level, dev, fmt, args...) \
3980do { \
3981 if (netif_msg_##type(priv)) \
3982 netdev_printk(level, (dev), fmt, ##args); \
3983} while (0)
3984
f45f4321
JP
3985#define netif_level(level, priv, type, dev, fmt, args...) \
3986do { \
3987 if (netif_msg_##type(priv)) \
3988 netdev_##level(dev, fmt, ##args); \
3989} while (0)
3990
b3d95c5c 3991#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 3992 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 3993#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 3994 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 3995#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 3996 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 3997#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 3998 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 3999#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 4000 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 4001#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 4002 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4003#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4004 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4005
0053ea9c 4006#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4007#define netif_dbg(priv, type, netdev, format, args...) \
4008do { \
4009 if (netif_msg_##type(priv)) \
b5fb0a03 4010 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4011} while (0)
0053ea9c
JP
4012#elif defined(DEBUG)
4013#define netif_dbg(priv, type, dev, format, args...) \
4014 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4015#else
4016#define netif_dbg(priv, type, dev, format, args...) \
4017({ \
4018 if (0) \
4019 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4020 0; \
4021})
4022#endif
4023
4024#if defined(VERBOSE_DEBUG)
bcfcc450 4025#define netif_vdbg netif_dbg
b3d95c5c
JP
4026#else
4027#define netif_vdbg(priv, type, dev, format, args...) \
4028({ \
4029 if (0) \
a4ed89cb 4030 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4031 0; \
4032})
4033#endif
571ba423 4034
900ff8c6
CW
4035/*
4036 * The list of packet types we will receive (as opposed to discard)
4037 * and the routines to invoke.
4038 *
4039 * Why 16. Because with 16 the only overlap we get on a hash of the
4040 * low nibble of the protocol value is RARP/SNAP/X.25.
4041 *
4042 * NOTE: That is no longer true with the addition of VLAN tags. Not
4043 * sure which should go first, but I bet it won't make much
4044 * difference if we are running VLANs. The good news is that
4045 * this protocol won't be in the list unless compiled in, so
4046 * the average user (w/out VLANs) will not be adversely affected.
4047 * --BLG
4048 *
4049 * 0800 IP
4050 * 8100 802.1Q VLAN
4051 * 0001 802.3
4052 * 0002 AX.25
4053 * 0004 802.2
4054 * 8035 RARP
4055 * 0005 SNAP
4056 * 0805 X.25
4057 * 0806 ARP
4058 * 8137 IPX
4059 * 0009 Localtalk
4060 * 86DD IPv6
4061 */
4062#define PTYPE_HASH_SIZE (16)
4063#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4064
385a154c 4065#endif /* _LINUX_NETDEVICE_H */