]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/linux/netdevice.h
net: move napi_hash[] into read mostly section
[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 */
d64b5e85
ED
329 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
330 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
bea3348e
SH
331};
332
5b252f0c 333enum gro_result {
d1c76af9
HX
334 GRO_MERGED,
335 GRO_MERGED_FREE,
336 GRO_HELD,
337 GRO_NORMAL,
338 GRO_DROP,
339};
5b252f0c 340typedef enum gro_result gro_result_t;
d1c76af9 341
8a4eb573
JP
342/*
343 * enum rx_handler_result - Possible return values for rx_handlers.
344 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
345 * further.
346 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
347 * case skb->dev was changed by rx_handler.
348 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
349 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
350 *
351 * rx_handlers are functions called from inside __netif_receive_skb(), to do
352 * special processing of the skb, prior to delivery to protocol handlers.
353 *
354 * Currently, a net_device can only have a single rx_handler registered. Trying
355 * to register a second rx_handler will return -EBUSY.
356 *
357 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
358 * To unregister a rx_handler on a net_device, use
359 * netdev_rx_handler_unregister().
360 *
361 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
362 * do with the skb.
363 *
364 * If the rx_handler consumed to skb in some way, it should return
365 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
366 * the skb to be delivered in some other ways.
367 *
368 * If the rx_handler changed skb->dev, to divert the skb to another
369 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
370 * new device will be called if it exists.
371 *
372 * If the rx_handler consider the skb should be ignored, it should return
373 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 374 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
375 *
376 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
377 * delivered, it should return RX_HANDLER_PASS.
378 *
379 * A device without a registered rx_handler will behave as if rx_handler
380 * returned RX_HANDLER_PASS.
381 */
382
383enum rx_handler_result {
384 RX_HANDLER_CONSUMED,
385 RX_HANDLER_ANOTHER,
386 RX_HANDLER_EXACT,
387 RX_HANDLER_PASS,
388};
389typedef enum rx_handler_result rx_handler_result_t;
390typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 391
f629d208 392void __napi_schedule(struct napi_struct *n);
bc9ad166 393void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 394
4d29515f 395static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
396{
397 return test_bit(NAPI_STATE_DISABLE, &n->state);
398}
399
bea3348e
SH
400/**
401 * napi_schedule_prep - check if napi can be scheduled
402 * @n: napi context
403 *
404 * Test if NAPI routine is already running, and if not mark
405 * it as running. This is used as a condition variable
a0a46196
DM
406 * insure only one NAPI poll instance runs. We also make
407 * sure there is no pending NAPI disable.
bea3348e 408 */
4d29515f 409static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 410{
a0a46196
DM
411 return !napi_disable_pending(n) &&
412 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
413}
414
415/**
416 * napi_schedule - schedule NAPI poll
417 * @n: napi context
418 *
419 * Schedule NAPI poll routine to be called if it is not already
420 * running.
421 */
422static inline void napi_schedule(struct napi_struct *n)
423{
424 if (napi_schedule_prep(n))
425 __napi_schedule(n);
426}
427
bc9ad166
ED
428/**
429 * napi_schedule_irqoff - schedule NAPI poll
430 * @n: napi context
431 *
432 * Variant of napi_schedule(), assuming hard irqs are masked.
433 */
434static inline void napi_schedule_irqoff(struct napi_struct *n)
435{
436 if (napi_schedule_prep(n))
437 __napi_schedule_irqoff(n);
438}
439
bfe13f54 440/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 441static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
442{
443 if (napi_schedule_prep(napi)) {
444 __napi_schedule(napi);
4d29515f 445 return true;
bfe13f54 446 }
4d29515f 447 return false;
bfe13f54
RD
448}
449
3b47d303
ED
450void __napi_complete(struct napi_struct *n);
451void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
452/**
453 * napi_complete - NAPI processing complete
454 * @n: napi context
455 *
456 * Mark NAPI processing as complete.
3b47d303 457 * Consider using napi_complete_done() instead.
bea3348e 458 */
3b47d303
ED
459static inline void napi_complete(struct napi_struct *n)
460{
461 return napi_complete_done(n, 0);
462}
bea3348e 463
af12fa6e
ET
464/**
465 * napi_hash_add - add a NAPI to global hashtable
466 * @napi: napi context
467 *
468 * generate a new napi_id and store a @napi under it in napi_hash
469 */
f629d208 470void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
471
472/**
473 * napi_hash_del - remove a NAPI from global table
474 * @napi: napi context
475 *
476 * Warning: caller must observe rcu grace period
477 * before freeing memory containing @napi
478 */
f629d208 479void napi_hash_del(struct napi_struct *napi);
af12fa6e 480
bea3348e
SH
481/**
482 * napi_disable - prevent NAPI from scheduling
483 * @n: napi context
484 *
485 * Stop NAPI from being scheduled on this context.
486 * Waits till any outstanding processing completes.
487 */
3b47d303 488void napi_disable(struct napi_struct *n);
bea3348e
SH
489
490/**
491 * napi_enable - enable NAPI scheduling
492 * @n: napi context
493 *
494 * Resume NAPI from being scheduled on this context.
495 * Must be paired with napi_disable.
496 */
497static inline void napi_enable(struct napi_struct *n)
498{
499 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 500 smp_mb__before_atomic();
bea3348e 501 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 502 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
503}
504
c264c3de
SH
505#ifdef CONFIG_SMP
506/**
507 * napi_synchronize - wait until NAPI is not running
508 * @n: napi context
509 *
510 * Wait until NAPI is done being scheduled on this context.
511 * Waits till any outstanding processing completes but
512 * does not disable future activations.
513 */
514static inline void napi_synchronize(const struct napi_struct *n)
515{
516 while (test_bit(NAPI_STATE_SCHED, &n->state))
517 msleep(1);
518}
519#else
520# define napi_synchronize(n) barrier()
521#endif
522
d94d9fee 523enum netdev_queue_state_t {
73466498
TH
524 __QUEUE_STATE_DRV_XOFF,
525 __QUEUE_STATE_STACK_XOFF,
c3f26a26 526 __QUEUE_STATE_FROZEN,
79d16385 527};
8e2f1a63
DB
528
529#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
530#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
531#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
532
533#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
534#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
535 QUEUE_STATE_FROZEN)
536#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
537 QUEUE_STATE_FROZEN)
538
73466498
TH
539/*
540 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
541 * netif_tx_* functions below are used to manipulate this flag. The
542 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
543 * queue independently. The netif_xmit_*stopped functions below are called
544 * to check if the queue has been stopped by the driver or stack (either
545 * of the XOFF bits are set in the state). Drivers should not need to call
546 * netif_xmit*stopped functions, they should only be using netif_tx_*.
547 */
79d16385 548
bb949fbd 549struct netdev_queue {
6a321cb3
ED
550/*
551 * read mostly part
552 */
bb949fbd 553 struct net_device *dev;
46e5da40 554 struct Qdisc __rcu *qdisc;
b0e1e646 555 struct Qdisc *qdisc_sleeping;
ccf5ff69 556#ifdef CONFIG_SYSFS
1d24eb48
TH
557 struct kobject kobj;
558#endif
f2cd2d3e
ED
559#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
560 int numa_node;
561#endif
6a321cb3
ED
562/*
563 * write mostly part
564 */
565 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
566 int xmit_lock_owner;
9d21493b
ED
567 /*
568 * please use this field instead of dev->trans_start
569 */
570 unsigned long trans_start;
ccf5ff69 571
572 /*
573 * Number of TX timeouts for this queue
574 * (/sys/class/net/DEV/Q/trans_timeout)
575 */
576 unsigned long trans_timeout;
114cf580
TH
577
578 unsigned long state;
579
580#ifdef CONFIG_BQL
581 struct dql dql;
582#endif
822b3b2e 583 unsigned long tx_maxrate;
e8a0464c 584} ____cacheline_aligned_in_smp;
bb949fbd 585
f2cd2d3e
ED
586static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
587{
588#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
589 return q->numa_node;
590#else
b236da69 591 return NUMA_NO_NODE;
f2cd2d3e
ED
592#endif
593}
594
595static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
596{
597#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
598 q->numa_node = node;
599#endif
600}
601
df334545 602#ifdef CONFIG_RPS
0a9627f2
TH
603/*
604 * This structure holds an RPS map which can be of variable length. The
605 * map is an array of CPUs.
606 */
607struct rps_map {
608 unsigned int len;
609 struct rcu_head rcu;
610 u16 cpus[0];
611};
60b778ce 612#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 613
fec5e652 614/*
c445477d
BH
615 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
616 * tail pointer for that CPU's input queue at the time of last enqueue, and
617 * a hardware filter index.
fec5e652
TH
618 */
619struct rps_dev_flow {
620 u16 cpu;
c445477d 621 u16 filter;
fec5e652
TH
622 unsigned int last_qtail;
623};
c445477d 624#define RPS_NO_FILTER 0xffff
fec5e652
TH
625
626/*
627 * The rps_dev_flow_table structure contains a table of flow mappings.
628 */
629struct rps_dev_flow_table {
630 unsigned int mask;
631 struct rcu_head rcu;
fec5e652
TH
632 struct rps_dev_flow flows[0];
633};
634#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 635 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
636
637/*
638 * The rps_sock_flow_table contains mappings of flows to the last CPU
639 * on which they were processed by the application (set in recvmsg).
567e4b79
ED
640 * Each entry is a 32bit value. Upper part is the high order bits
641 * of flow hash, lower part is cpu number.
642 * rps_cpu_mask is used to partition the space, depending on number of
643 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
644 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
645 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
646 */
647struct rps_sock_flow_table {
567e4b79 648 u32 mask;
93c1af6c
ED
649
650 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 651};
567e4b79 652#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
653
654#define RPS_NO_CPU 0xffff
655
567e4b79
ED
656extern u32 rps_cpu_mask;
657extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
658
fec5e652
TH
659static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
660 u32 hash)
661{
662 if (table && hash) {
567e4b79
ED
663 unsigned int index = hash & table->mask;
664 u32 val = hash & ~rps_cpu_mask;
fec5e652
TH
665
666 /* We only give a hint, preemption can change cpu under us */
567e4b79 667 val |= raw_smp_processor_id();
fec5e652 668
567e4b79
ED
669 if (table->ents[index] != val)
670 table->ents[index] = val;
fec5e652
TH
671 }
672}
673
c445477d 674#ifdef CONFIG_RFS_ACCEL
f629d208
JP
675bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
676 u16 filter_id);
c445477d 677#endif
a953be53 678#endif /* CONFIG_RPS */
c445477d 679
0a9627f2
TH
680/* This structure contains an instance of an RX queue. */
681struct netdev_rx_queue {
a953be53 682#ifdef CONFIG_RPS
6e3f7faf
ED
683 struct rps_map __rcu *rps_map;
684 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 685#endif
6e3f7faf 686 struct kobject kobj;
fe822240 687 struct net_device *dev;
0a9627f2 688} ____cacheline_aligned_in_smp;
a953be53
MD
689
690/*
691 * RX queue sysfs structures and functions.
692 */
693struct rx_queue_attribute {
694 struct attribute attr;
695 ssize_t (*show)(struct netdev_rx_queue *queue,
696 struct rx_queue_attribute *attr, char *buf);
697 ssize_t (*store)(struct netdev_rx_queue *queue,
698 struct rx_queue_attribute *attr, const char *buf, size_t len);
699};
d314774c 700
bf264145
TH
701#ifdef CONFIG_XPS
702/*
703 * This structure holds an XPS map which can be of variable length. The
704 * map is an array of queues.
705 */
706struct xps_map {
707 unsigned int len;
708 unsigned int alloc_len;
709 struct rcu_head rcu;
710 u16 queues[0];
711};
60b778ce 712#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
713#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
714 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
715
716/*
717 * This structure holds all XPS maps for device. Maps are indexed by CPU.
718 */
719struct xps_dev_maps {
720 struct rcu_head rcu;
a4177869 721 struct xps_map __rcu *cpu_map[0];
bf264145
TH
722};
723#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
724 (nr_cpu_ids * sizeof(struct xps_map *)))
725#endif /* CONFIG_XPS */
726
4f57c087
JF
727#define TC_MAX_QUEUE 16
728#define TC_BITMASK 15
729/* HW offloaded queuing disciplines txq count and offset maps */
730struct netdev_tc_txq {
731 u16 count;
732 u16 offset;
733};
734
68bad94e
NP
735#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
736/*
737 * This structure is to hold information about the device
738 * configured to run FCoE protocol stack.
739 */
740struct netdev_fcoe_hbainfo {
741 char manufacturer[64];
742 char serial_number[64];
743 char hardware_version[64];
744 char driver_version[64];
745 char optionrom_version[64];
746 char firmware_version[64];
747 char model[256];
748 char model_description[256];
749};
750#endif
751
02637fce 752#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 753
02637fce
JP
754/* This structure holds a unique identifier to identify some
755 * physical item (port for example) used by a netdevice.
66b52b0d 756 */
02637fce
JP
757struct netdev_phys_item_id {
758 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
759 unsigned char id_len;
760};
761
d754f98b
SF
762static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
763 struct netdev_phys_item_id *b)
764{
765 return a->id_len == b->id_len &&
766 memcmp(a->id, b->id, a->id_len) == 0;
767}
768
99932d4f
DB
769typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 struct sk_buff *skb);
771
d314774c
SH
772/*
773 * This structure defines the management hooks for network devices.
00829823
SH
774 * The following hooks can be defined; unless noted otherwise, they are
775 * optional and can be filled with a null pointer.
d314774c
SH
776 *
777 * int (*ndo_init)(struct net_device *dev);
778 * This function is called once when network device is registered.
779 * The network device can use this to any late stage initializaton
780 * or semantic validattion. It can fail with an error code which will
781 * be propogated back to register_netdev
782 *
783 * void (*ndo_uninit)(struct net_device *dev);
784 * This function is called when device is unregistered or when registration
785 * fails. It is not called if init fails.
786 *
787 * int (*ndo_open)(struct net_device *dev);
788 * This function is called when network device transistions to the up
789 * state.
790 *
791 * int (*ndo_stop)(struct net_device *dev);
792 * This function is called when network device transistions to the down
793 * state.
794 *
dc1f8bf6
SH
795 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
796 * struct net_device *dev);
00829823 797 * Called when a packet needs to be transmitted.
e79d8429
RR
798 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
799 * the queue before that can happen; it's for obsolete devices and weird
800 * corner cases, but the stack really does a non-trivial amount
801 * of useless work if you return NETDEV_TX_BUSY.
dc1f8bf6 802 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
803 * Required can not be NULL.
804 *
f663dd9a 805 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 806 * void *accel_priv, select_queue_fallback_t fallback);
00829823
SH
807 * Called to decide which queue to when device supports multiple
808 * transmit queues.
809 *
d314774c
SH
810 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
811 * This function is called to allow device receiver to make
812 * changes to configuration when multicast or promiscious is enabled.
813 *
814 * void (*ndo_set_rx_mode)(struct net_device *dev);
815 * This function is called device changes address list filtering.
01789349
JP
816 * If driver handles unicast address filtering, it should set
817 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
818 *
819 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
820 * This function is called when the Media Access Control address
37b607c5 821 * needs to be changed. If this interface is not defined, the
d314774c
SH
822 * mac address can not be changed.
823 *
824 * int (*ndo_validate_addr)(struct net_device *dev);
825 * Test if Media Access Control address is valid for the device.
826 *
827 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
828 * Called when a user request an ioctl which can't be handled by
829 * the generic interface code. If not defined ioctl's return
830 * not supported error code.
831 *
832 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
833 * Used to set network devices bus interface parameters. This interface
834 * is retained for legacy reason, new devices should use the bus
835 * interface (PCI) for low level management.
836 *
837 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
838 * Called when a user wants to change the Maximum Transfer Unit
839 * of a device. If not defined, any request to change MTU will
840 * will return an error.
841 *
00829823 842 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
843 * Callback uses when the transmitter has not made any progress
844 * for dev->watchdog ticks.
845 *
3cfde79c 846 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 847 * struct rtnl_link_stats64 *storage);
d308e38f 848 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 849 * Called when a user wants to get the network device usage
be1f3c2c 850 * statistics. Drivers must do one of the following:
3cfde79c
BH
851 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
852 * rtnl_link_stats64 structure passed by the caller.
82695d9b 853 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
854 * (which should normally be dev->stats) and return a pointer to
855 * it. The structure may be changed asynchronously only if each
856 * field is written atomically.
857 * 3. Update dev->stats asynchronously and atomically, and define
858 * neither operation.
d314774c 859 *
5d632cb7 860 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
861 * If device support VLAN filtering this function is called when a
862 * VLAN id is registered.
d314774c 863 *
5d632cb7 864 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
865 * If device support VLAN filtering this function is called when a
866 * VLAN id is unregistered.
d314774c
SH
867 *
868 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
869 *
870 * SR-IOV management functions.
871 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
872 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
873 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
874 * int max_tx_rate);
5f8444a3 875 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 876 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
877 * int (*ndo_get_vf_config)(struct net_device *dev,
878 * int vf, struct ifla_vf_info *ivf);
1d8faf48 879 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
880 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
881 * struct nlattr *port[]);
01a3d796
VZ
882 *
883 * Enable or disable the VF ability to query its RSS Redirection Table and
884 * Hash Key. This is needed since on some devices VF share this information
885 * with PF and querying it may adduce a theoretical security risk.
886 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 887 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
888 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
889 * Called to setup 'tc' number of traffic classes in the net device. This
890 * is always called from the stack with the rtnl lock held and netif tx
891 * queues stopped. This allows the netdevice to perform queue management
892 * safely.
c445477d 893 *
e9bce845
YZ
894 * Fiber Channel over Ethernet (FCoE) offload functions.
895 * int (*ndo_fcoe_enable)(struct net_device *dev);
896 * Called when the FCoE protocol stack wants to start using LLD for FCoE
897 * so the underlying device can perform whatever needed configuration or
898 * initialization to support acceleration of FCoE traffic.
899 *
900 * int (*ndo_fcoe_disable)(struct net_device *dev);
901 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
902 * so the underlying device can perform whatever needed clean-ups to
903 * stop supporting acceleration of FCoE traffic.
904 *
905 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
906 * struct scatterlist *sgl, unsigned int sgc);
907 * Called when the FCoE Initiator wants to initialize an I/O that
908 * is a possible candidate for Direct Data Placement (DDP). The LLD can
909 * perform necessary setup and returns 1 to indicate the device is set up
910 * successfully to perform DDP on this I/O, otherwise this returns 0.
911 *
912 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
913 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
914 * indicated by the FC exchange id 'xid', so the underlying device can
915 * clean up and reuse resources for later DDP requests.
916 *
917 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
918 * struct scatterlist *sgl, unsigned int sgc);
919 * Called when the FCoE Target wants to initialize an I/O that
920 * is a possible candidate for Direct Data Placement (DDP). The LLD can
921 * perform necessary setup and returns 1 to indicate the device is set up
922 * successfully to perform DDP on this I/O, otherwise this returns 0.
923 *
68bad94e
NP
924 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
925 * struct netdev_fcoe_hbainfo *hbainfo);
926 * Called when the FCoE Protocol stack wants information on the underlying
927 * device. This information is utilized by the FCoE protocol stack to
928 * register attributes with Fiber Channel management service as per the
929 * FC-GS Fabric Device Management Information(FDMI) specification.
930 *
e9bce845
YZ
931 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
932 * Called when the underlying device wants to override default World Wide
933 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
934 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
935 * protocol stack to use.
936 *
c445477d
BH
937 * RFS acceleration.
938 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
939 * u16 rxq_index, u32 flow_id);
940 * Set hardware filter for RFS. rxq_index is the target queue index;
941 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
942 * Return the filter ID on success, or a negative error code.
fbaec0ea 943 *
8b98a70c 944 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
945 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
946 * Called to make another netdev an underling.
947 *
948 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
949 * Called to release previously enslaved netdev.
5455c699
MM
950 *
951 * Feature/offload setting functions.
c8f44aff
MM
952 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
953 * netdev_features_t features);
5455c699
MM
954 * Adjusts the requested feature flags according to device-specific
955 * constraints, and returns the resulting flags. Must not modify
956 * the device state.
957 *
c8f44aff 958 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
959 * Called to update device configuration to new features. Passed
960 * feature set might be less than what was returned by ndo_fix_features()).
961 * Must return >0 or -errno if it changed dev->features itself.
962 *
edc7d573 963 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
964 * struct net_device *dev,
f6f6424b 965 * const unsigned char *addr, u16 vid, u16 flags)
77162022 966 * Adds an FDB entry to dev for addr.
1690be63
VY
967 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
968 * struct net_device *dev,
f6f6424b 969 * const unsigned char *addr, u16 vid)
77162022
JF
970 * Deletes the FDB entry from dev coresponding to addr.
971 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
972 * struct net_device *dev, struct net_device *filter_dev,
973 * int idx)
77162022
JF
974 * Used to add FDB entries to dump requests. Implementers should add
975 * entries to skb and update idx with the number of entries.
e5a55a89 976 *
ad41faa8
ND
977 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
978 * u16 flags)
e5a55a89 979 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
980 * struct net_device *dev, u32 filter_mask,
981 * int nlflags)
ad41faa8
ND
982 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
983 * u16 flags);
4bf84c35
JP
984 *
985 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
986 * Called to change device carrier. Soft-devices (like dummy, team, etc)
987 * which do not represent real hardware may define this to allow their
988 * userspace components to manage their virtual carrier state. Devices
989 * that determine carrier state from physical hardware properties (eg
990 * network cables) or protocol-dependent mechanisms (eg
991 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
992 *
993 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 994 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
995 * Called to get ID of physical port of this device. If driver does
996 * not implement this, it is assumed that the hw is not able to have
997 * multiple net devices on single physical port.
53cf5275
JG
998 *
999 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 1000 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1001 * Called by vxlan to notiy a driver about the UDP port and socket
1002 * address family that vxlan is listnening to. It is called only when
1003 * a new port starts listening. The operation is protected by the
1004 * vxlan_net->sock_lock.
1005 *
1006 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 1007 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1008 * Called by vxlan to notify the driver about a UDP port and socket
1009 * address family that vxlan is not listening to anymore. The operation
1010 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
1011 *
1012 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1013 * struct net_device *dev)
1014 * Called by upper layer devices to accelerate switching or other
1015 * station functionality into hardware. 'pdev is the lowerdev
1016 * to use for the offload and 'dev' is the net device that will
1017 * back the offload. Returns a pointer to the private structure
1018 * the upper layer will maintain.
1019 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1020 * Called by upper layer device to delete the station created
1021 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1022 * the station and priv is the structure returned by the add
1023 * operation.
1024 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1025 * struct net_device *dev,
1026 * void *priv);
1027 * Callback to use for xmit over the accelerated station. This
1028 * is used in place of ndo_start_xmit on accelerated net
1029 * devices.
5f35227e
JG
1030 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1031 * struct net_device *dev
1032 * netdev_features_t features);
04ffcb25 1033 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1034 * performing offload operations on a given packet. This is to give
1035 * the device an opportunity to implement any restrictions that cannot
1036 * be otherwise expressed by feature flags. The check is called with
1037 * the set of features that the stack has calculated and it returns
1038 * those the driver believes to be appropriate.
822b3b2e
JF
1039 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1040 * int queue_index, u32 maxrate);
1041 * Called when a user wants to set a max-rate limitation of specific
1042 * TX queue.
a54acb3a
ND
1043 * int (*ndo_get_iflink)(const struct net_device *dev);
1044 * Called to get the iflink value of this device.
d746d707
AK
1045 * void (*ndo_change_proto_down)(struct net_device *dev,
1046 * bool proto_down);
1047 * This function is used to pass protocol port error state information
1048 * to the switch driver. The switch driver can react to the proto_down
1049 * by doing a phys down on the associated switch port.
fc4099f1
PS
1050 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1051 * This function is used to get egress tunnel information for given skb.
1052 * This is useful for retrieving outer tunnel header parameters while
1053 * sampling packet.
d746d707 1054 *
d314774c
SH
1055 */
1056struct net_device_ops {
1057 int (*ndo_init)(struct net_device *dev);
1058 void (*ndo_uninit)(struct net_device *dev);
1059 int (*ndo_open)(struct net_device *dev);
1060 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 1061 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
1062 struct net_device *dev);
1063 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1064 struct sk_buff *skb,
99932d4f
DB
1065 void *accel_priv,
1066 select_queue_fallback_t fallback);
d314774c
SH
1067 void (*ndo_change_rx_flags)(struct net_device *dev,
1068 int flags);
d314774c 1069 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1070 int (*ndo_set_mac_address)(struct net_device *dev,
1071 void *addr);
d314774c 1072 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1073 int (*ndo_do_ioctl)(struct net_device *dev,
1074 struct ifreq *ifr, int cmd);
d314774c
SH
1075 int (*ndo_set_config)(struct net_device *dev,
1076 struct ifmap *map);
00829823
SH
1077 int (*ndo_change_mtu)(struct net_device *dev,
1078 int new_mtu);
1079 int (*ndo_neigh_setup)(struct net_device *dev,
1080 struct neigh_parms *);
d314774c
SH
1081 void (*ndo_tx_timeout) (struct net_device *dev);
1082
28172739
ED
1083 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1084 struct rtnl_link_stats64 *storage);
d314774c
SH
1085 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1086
8e586137 1087 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1088 __be16 proto, u16 vid);
8e586137 1089 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1090 __be16 proto, u16 vid);
d314774c 1091#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1092 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1093 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1094 struct netpoll_info *info);
0e34e931 1095 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1096#endif
e0d1095a 1097#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1098 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1099#endif
95c26df8
WM
1100 int (*ndo_set_vf_mac)(struct net_device *dev,
1101 int queue, u8 *mac);
1102 int (*ndo_set_vf_vlan)(struct net_device *dev,
1103 int queue, u16 vlan, u8 qos);
ed616689
SC
1104 int (*ndo_set_vf_rate)(struct net_device *dev,
1105 int vf, int min_tx_rate,
1106 int max_tx_rate);
5f8444a3
GR
1107 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1108 int vf, bool setting);
dd461d6a
HS
1109 int (*ndo_set_vf_trust)(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);
fc4099f1
PS
1229 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1230 struct sk_buff *skb);
d314774c
SH
1231};
1232
7aa98047
LR
1233/**
1234 * enum net_device_priv_flags - &struct net_device priv_flags
1235 *
1236 * These are the &struct net_device, they are only set internally
1237 * by drivers and used in the kernel. These flags are invisible to
1238 * userspace, this means that the order of these flags can change
1239 * during any kernel release.
1240 *
1241 * You should have a pretty good reason to be extending these flags.
1242 *
1243 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1244 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1245 * @IFF_BONDING: bonding master or slave
7aa98047 1246 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1247 * @IFF_WAN_HDLC: WAN HDLC device
1248 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1249 * release skb->dst
1250 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1251 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1252 * @IFF_MACVLAN_PORT: device used as macvlan port
1253 * @IFF_BRIDGE_PORT: device used as bridge port
1254 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1255 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1256 * @IFF_UNICAST_FLT: Supports unicast filtering
1257 * @IFF_TEAM_PORT: device used as team port
1258 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1259 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1260 * change when it's running
1261 * @IFF_MACVLAN: Macvlan device
007979ea 1262 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1263 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1264 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1265 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
7aa98047
LR
1266 */
1267enum netdev_priv_flags {
1268 IFF_802_1Q_VLAN = 1<<0,
1269 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1270 IFF_BONDING = 1<<2,
1271 IFF_ISATAP = 1<<3,
1272 IFF_WAN_HDLC = 1<<4,
1273 IFF_XMIT_DST_RELEASE = 1<<5,
1274 IFF_DONT_BRIDGE = 1<<6,
1275 IFF_DISABLE_NETPOLL = 1<<7,
1276 IFF_MACVLAN_PORT = 1<<8,
1277 IFF_BRIDGE_PORT = 1<<9,
1278 IFF_OVS_DATAPATH = 1<<10,
1279 IFF_TX_SKB_SHARING = 1<<11,
1280 IFF_UNICAST_FLT = 1<<12,
1281 IFF_TEAM_PORT = 1<<13,
1282 IFF_SUPP_NOFCS = 1<<14,
1283 IFF_LIVE_ADDR_CHANGE = 1<<15,
1284 IFF_MACVLAN = 1<<16,
1285 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1286 IFF_IPVLAN_MASTER = 1<<18,
1287 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1288 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1289 IFF_NO_QUEUE = 1<<21,
1290 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1291 IFF_L3MDEV_SLAVE = 1<<23,
7aa98047
LR
1292};
1293
1294#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1295#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1296#define IFF_BONDING IFF_BONDING
7aa98047 1297#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1298#define IFF_WAN_HDLC IFF_WAN_HDLC
1299#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1300#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1301#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1302#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1303#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1304#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1305#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1306#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1307#define IFF_TEAM_PORT IFF_TEAM_PORT
1308#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1309#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1310#define IFF_MACVLAN IFF_MACVLAN
02875878 1311#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1312#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1313#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1314#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1315#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1316#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1317#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
7aa98047 1318
536721b1
KK
1319/**
1320 * struct net_device - The DEVICE structure.
1321 * Actually, this whole structure is a big mistake. It mixes I/O
1322 * data with strictly "high-level" data, and it has to know about
1323 * almost every data structure used in the INET module.
1324 *
1325 * @name: This is the first field of the "visible" part of this structure
1326 * (i.e. as seen by users in the "Space.c" file). It is the name
1327 * of the interface.
1328 *
1329 * @name_hlist: Device name hash chain, please keep it close to name[]
1330 * @ifalias: SNMP alias
1331 * @mem_end: Shared memory end
1332 * @mem_start: Shared memory start
1333 * @base_addr: Device I/O address
1334 * @irq: Device IRQ number
1335 *
14ffbbb8
TG
1336 * @carrier_changes: Stats to monitor carrier on<->off transitions
1337 *
536721b1
KK
1338 * @state: Generic network queuing layer state, see netdev_state_t
1339 * @dev_list: The global list of network devices
1340 * @napi_list: List entry, that is used for polling napi devices
1341 * @unreg_list: List entry, that is used, when we are unregistering the
1342 * device, see the function unregister_netdev
1343 * @close_list: List entry, that is used, when we are closing the device
1344 *
1345 * @adj_list: Directly linked devices, like slaves for bonding
1346 * @all_adj_list: All linked devices, *including* neighbours
1347 * @features: Currently active device features
1348 * @hw_features: User-changeable features
1349 *
1350 * @wanted_features: User-requested features
1351 * @vlan_features: Mask of features inheritable by VLAN devices
1352 *
1353 * @hw_enc_features: Mask of features inherited by encapsulating devices
1354 * This field indicates what encapsulation
1355 * offloads the hardware is capable of doing,
1356 * and drivers will need to set them appropriately.
1357 *
1358 * @mpls_features: Mask of features inheritable by MPLS
1359 *
1360 * @ifindex: interface index
388069d3 1361 * @group: The group, that the device belongs to
536721b1
KK
1362 *
1363 * @stats: Statistics struct, which was left as a legacy, use
1364 * rtnl_link_stats64 instead
1365 *
1366 * @rx_dropped: Dropped packets by core network,
1367 * do not use this in drivers
1368 * @tx_dropped: Dropped packets by core network,
1369 * do not use this in drivers
1370 *
536721b1
KK
1371 * @wireless_handlers: List of functions to handle Wireless Extensions,
1372 * instead of ioctl,
1373 * see <net/iw_handler.h> for details.
1374 * @wireless_data: Instance data managed by the core of wireless extensions
1375 *
1376 * @netdev_ops: Includes several pointers to callbacks,
1377 * if one wants to override the ndo_*() functions
1378 * @ethtool_ops: Management operations
d476059e 1379 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1380 * of Layer 2 headers.
1381 *
1382 * @flags: Interface flags (a la BSD)
1383 * @priv_flags: Like 'flags' but invisible to userspace,
1384 * see if.h for the definitions
1385 * @gflags: Global flags ( kept as legacy )
1386 * @padded: How much padding added by alloc_netdev()
1387 * @operstate: RFC2863 operstate
1388 * @link_mode: Mapping policy to operstate
1389 * @if_port: Selectable AUI, TP, ...
1390 * @dma: DMA channel
1391 * @mtu: Interface MTU value
1392 * @type: Interface hardware type
1393 * @hard_header_len: Hardware header length
1394 *
1395 * @needed_headroom: Extra headroom the hardware may need, but not in all
1396 * cases can this be guaranteed
1397 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1398 * cases can this be guaranteed. Some cases also use
1399 * LL_MAX_HEADER instead to allocate the skb
1400 *
1401 * interface address info:
1402 *
1403 * @perm_addr: Permanent hw address
1404 * @addr_assign_type: Hw address assignment type
1405 * @addr_len: Hardware address length
1406 * @neigh_priv_len; Used in neigh_alloc(),
1407 * initialized only in atm/clip.c
1408 * @dev_id: Used to differentiate devices that share
1409 * the same link layer address
1410 * @dev_port: Used to differentiate devices that share
1411 * the same function
1412 * @addr_list_lock: XXX: need comments on this one
536721b1
KK
1413 * @uc_promisc: Counter, that indicates, that promiscuous mode
1414 * has been enabled due to the need to listen to
1415 * additional unicast addresses in a device that
1416 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1417 * @uc: unicast mac addresses
1418 * @mc: multicast mac addresses
1419 * @dev_addrs: list of device hw addresses
1420 * @queues_kset: Group of all Kobjects in the Tx and RX queues
536721b1
KK
1421 * @promiscuity: Number of times, the NIC is told to work in
1422 * Promiscuous mode, if it becomes 0 the NIC will
1423 * exit from working in Promiscuous mode
1424 * @allmulti: Counter, enables or disables allmulticast mode
1425 *
1426 * @vlan_info: VLAN info
1427 * @dsa_ptr: dsa specific data
1428 * @tipc_ptr: TIPC specific data
1429 * @atalk_ptr: AppleTalk link
1430 * @ip_ptr: IPv4 specific data
1431 * @dn_ptr: DECnet specific data
1432 * @ip6_ptr: IPv6 specific data
1433 * @ax25_ptr: AX.25 specific data
1434 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1435 *
1436 * @last_rx: Time of last Rx
1437 * @dev_addr: Hw address (before bcast,
1438 * because most packets are unicast)
1439 *
1440 * @_rx: Array of RX queues
1441 * @num_rx_queues: Number of RX queues
1442 * allocated at register_netdev() time
1443 * @real_num_rx_queues: Number of RX queues currently active in device
1444 *
1445 * @rx_handler: handler for received packets
1446 * @rx_handler_data: XXX: need comments on this one
1447 * @ingress_queue: XXX: need comments on this one
1448 * @broadcast: hw bcast address
1449 *
14ffbbb8
TG
1450 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1451 * indexed by RX queue number. Assigned by driver.
1452 * This must only be set if the ndo_rx_flow_steer
1453 * operation is defined
1454 * @index_hlist: Device index hash chain
1455 *
536721b1
KK
1456 * @_tx: Array of TX queues
1457 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1458 * @real_num_tx_queues: Number of TX queues currently active in device
1459 * @qdisc: Root qdisc from userspace point of view
1460 * @tx_queue_len: Max frames per queue allowed
1461 * @tx_global_lock: XXX: need comments on this one
1462 *
1463 * @xps_maps: XXX: need comments on this one
1464 *
0c4f691f
SF
1465 * @offload_fwd_mark: Offload device fwding mark
1466 *
536721b1
KK
1467 * @trans_start: Time (in jiffies) of last Tx
1468 * @watchdog_timeo: Represents the timeout that is used by
1469 * the watchdog ( see dev_watchdog() )
1470 * @watchdog_timer: List of timers
1471 *
1472 * @pcpu_refcnt: Number of references to this device
1473 * @todo_list: Delayed register/unregister
536721b1
KK
1474 * @link_watch_list: XXX: need comments on this one
1475 *
1476 * @reg_state: Register/unregister state machine
1477 * @dismantle: Device is going to be freed
1478 * @rtnl_link_state: This enum represents the phases of creating
1479 * a new link
1480 *
1481 * @destructor: Called from unregister,
1482 * can be used to call free_netdev
1483 * @npinfo: XXX: need comments on this one
1484 * @nd_net: Network namespace this network device is inside
1485 *
1486 * @ml_priv: Mid-layer private
1487 * @lstats: Loopback statistics
1488 * @tstats: Tunnel statistics
1489 * @dstats: Dummy statistics
1490 * @vstats: Virtual ethernet statistics
1491 *
1492 * @garp_port: GARP
1493 * @mrp_port: MRP
1494 *
1495 * @dev: Class/net/name entry
1496 * @sysfs_groups: Space for optional device, statistics and wireless
1497 * sysfs groups
1498 *
1499 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1500 * @rtnl_link_ops: Rtnl_link_ops
1501 *
1502 * @gso_max_size: Maximum size of generic segmentation offload
1503 * @gso_max_segs: Maximum number of segments that can be passed to the
1504 * NIC for GSO
fcbeb976
ED
1505 * @gso_min_segs: Minimum number of segments that can be passed to the
1506 * NIC for GSO
536721b1
KK
1507 *
1508 * @dcbnl_ops: Data Center Bridging netlink ops
1509 * @num_tc: Number of traffic classes in the net device
1510 * @tc_to_txq: XXX: need comments on this one
1511 * @prio_tc_map XXX: need comments on this one
1512 *
1513 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1514 *
1515 * @priomap: XXX: need comments on this one
1516 * @phydev: Physical device may attach itself
1517 * for hardware timestamping
1518 *
1519 * @qdisc_tx_busylock: XXX: need comments on this one
1520 *
d746d707
AK
1521 * @proto_down: protocol port state information can be sent to the
1522 * switch driver and used to set the phys state of the
1523 * switch port.
1524 *
1da177e4
LT
1525 * FIXME: cleanup struct net_device such that network protocol info
1526 * moves out.
1527 */
1528
d94d9fee 1529struct net_device {
1da177e4 1530 char name[IFNAMSIZ];
9356b8fc 1531 struct hlist_node name_hlist;
0b815a1a 1532 char *ifalias;
1da177e4
LT
1533 /*
1534 * I/O specific fields
1535 * FIXME: Merge these and struct ifmap into one
1536 */
536721b1
KK
1537 unsigned long mem_end;
1538 unsigned long mem_start;
1539 unsigned long base_addr;
1540 int irq;
1da177e4 1541
14ffbbb8
TG
1542 atomic_t carrier_changes;
1543
1da177e4 1544 /*
536721b1
KK
1545 * Some hardware also needs these fields (state,dev_list,
1546 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1547 * part of the usual set specified in Space.c.
1548 */
1549
1da177e4
LT
1550 unsigned long state;
1551
7562f876 1552 struct list_head dev_list;
bea3348e 1553 struct list_head napi_list;
44a0873d 1554 struct list_head unreg_list;
5cde2829 1555 struct list_head close_list;
7866a621
SN
1556 struct list_head ptype_all;
1557 struct list_head ptype_specific;
2f268f12 1558
2f268f12
VF
1559 struct {
1560 struct list_head upper;
1561 struct list_head lower;
1562 } adj_list;
1563
2f268f12
VF
1564 struct {
1565 struct list_head upper;
1566 struct list_head lower;
1567 } all_adj_list;
4c3d5e7b 1568
c8f44aff 1569 netdev_features_t features;
c8f44aff 1570 netdev_features_t hw_features;
c8f44aff 1571 netdev_features_t wanted_features;
c8f44aff 1572 netdev_features_t vlan_features;
6a674e9c 1573 netdev_features_t hw_enc_features;
0d89d203 1574 netdev_features_t mpls_features;
04ed3e74 1575
1da177e4 1576 int ifindex;
7a66bbc9 1577 int group;
1da177e4 1578
c45d286e 1579 struct net_device_stats stats;
015f0688 1580
015f0688
ED
1581 atomic_long_t rx_dropped;
1582 atomic_long_t tx_dropped;
1da177e4 1583
b86e0280 1584#ifdef CONFIG_WIRELESS_EXT
1da177e4 1585 const struct iw_handler_def * wireless_handlers;
1da177e4 1586 struct iw_public_data * wireless_data;
b86e0280 1587#endif
d314774c 1588 const struct net_device_ops *netdev_ops;
76fd8593 1589 const struct ethtool_ops *ethtool_ops;
4170604f 1590#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1591 const struct switchdev_ops *switchdev_ops;
4170604f 1592#endif
1b69c6d0
DA
1593#ifdef CONFIG_NET_L3_MASTER_DEV
1594 const struct l3mdev_ops *l3mdev_ops;
1595#endif
1da177e4 1596
3b04ddde
SH
1597 const struct header_ops *header_ops;
1598
536721b1
KK
1599 unsigned int flags;
1600 unsigned int priv_flags;
1601
1da177e4 1602 unsigned short gflags;
536721b1 1603 unsigned short padded;
1da177e4 1604
536721b1
KK
1605 unsigned char operstate;
1606 unsigned char link_mode;
b00055aa 1607
536721b1
KK
1608 unsigned char if_port;
1609 unsigned char dma;
bdc220da 1610
536721b1
KK
1611 unsigned int mtu;
1612 unsigned short type;
1613 unsigned short hard_header_len;
1da177e4 1614
f5184d26
JB
1615 unsigned short needed_headroom;
1616 unsigned short needed_tailroom;
1617
1da177e4 1618 /* Interface address info. */
536721b1
KK
1619 unsigned char perm_addr[MAX_ADDR_LEN];
1620 unsigned char addr_assign_type;
1621 unsigned char addr_len;
a0a9663d 1622 unsigned short neigh_priv_len;
536721b1
KK
1623 unsigned short dev_id;
1624 unsigned short dev_port;
ccffad25 1625 spinlock_t addr_list_lock;
14ffbbb8
TG
1626 unsigned char name_assign_type;
1627 bool uc_promisc;
536721b1
KK
1628 struct netdev_hw_addr_list uc;
1629 struct netdev_hw_addr_list mc;
1630 struct netdev_hw_addr_list dev_addrs;
1631
4c3d5e7b
ED
1632#ifdef CONFIG_SYSFS
1633 struct kset *queues_kset;
1634#endif
9d45abe1
WC
1635 unsigned int promiscuity;
1636 unsigned int allmulti;
1da177e4 1637
1da177e4
LT
1638
1639 /* Protocol specific pointers */
65ac6a5f 1640
d11ead75 1641#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1642 struct vlan_info __rcu *vlan_info;
65ac6a5f 1643#endif
34a430d7 1644#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1645 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1646#endif
1647#if IS_ENABLED(CONFIG_TIPC)
536721b1 1648 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1649#endif
536721b1
KK
1650 void *atalk_ptr;
1651 struct in_device __rcu *ip_ptr;
1652 struct dn_dev __rcu *dn_ptr;
1653 struct inet6_dev __rcu *ip6_ptr;
1654 void *ax25_ptr;
1655 struct wireless_dev *ieee80211_ptr;
98a18b6f 1656 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1657#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1658 struct mpls_dev __rcu *mpls_ptr;
1659#endif
1da177e4 1660
9356b8fc 1661/*
cd13539b 1662 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1663 */
536721b1 1664 unsigned long last_rx;
4dc89133 1665
9356b8fc 1666 /* Interface address info used in eth_type_trans() */
536721b1 1667 unsigned char *dev_addr;
f001fde5 1668
0a9627f2 1669
a953be53 1670#ifdef CONFIG_SYSFS
0a9627f2
TH
1671 struct netdev_rx_queue *_rx;
1672
0a9627f2 1673 unsigned int num_rx_queues;
62fe0b40 1674 unsigned int real_num_rx_queues;
c445477d 1675
df334545 1676#endif
0a9627f2 1677
3b47d303 1678 unsigned long gro_flush_timeout;
61391cde 1679 rx_handler_func_t __rcu *rx_handler;
1680 void __rcu *rx_handler_data;
e8a0464c 1681
4cda01e8 1682#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1683 struct tcf_proto __rcu *ingress_cl_list;
1684#endif
24824a09 1685 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1686#ifdef CONFIG_NETFILTER_INGRESS
1687 struct list_head nf_hooks_ingress;
1688#endif
d2788d34 1689
536721b1 1690 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1691#ifdef CONFIG_RFS_ACCEL
1692 struct cpu_rmap *rx_cpu_rmap;
1693#endif
1694 struct hlist_node index_hlist;
cd13539b
ED
1695
1696/*
1697 * Cache lines mostly used on transmit path
1698 */
e8a0464c
DM
1699 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1700 unsigned int num_tx_queues;
fd2ea0a7 1701 unsigned int real_num_tx_queues;
af356afa 1702 struct Qdisc *qdisc;
536721b1 1703 unsigned long tx_queue_len;
c3f26a26 1704 spinlock_t tx_global_lock;
14ffbbb8 1705 int watchdog_timeo;
cd13539b 1706
bf264145 1707#ifdef CONFIG_XPS
a4177869 1708 struct xps_dev_maps __rcu *xps_maps;
bf264145 1709#endif
1d24eb48 1710
0c4f691f
SF
1711#ifdef CONFIG_NET_SWITCHDEV
1712 u32 offload_fwd_mark;
1713#endif
1714
9356b8fc 1715 /* These may be needed for future network-power-down code. */
9d21493b
ED
1716
1717 /*
1718 * trans_start here is expensive for high speed devices on SMP,
1719 * please use netdev_queue->trans_start instead.
1720 */
536721b1 1721 unsigned long trans_start;
9356b8fc 1722
9356b8fc
ED
1723 struct timer_list watchdog_timer;
1724
29b4433d 1725 int __percpu *pcpu_refcnt;
1da177e4 1726 struct list_head todo_list;
1da177e4 1727
e014debe 1728 struct list_head link_watch_list;
572a103d 1729
1da177e4 1730 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1731 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1732 NETREG_UNREGISTERING, /* called unregister_netdevice */
1733 NETREG_UNREGISTERED, /* completed unregister todo */
1734 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1735 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1736 } reg_state:8;
1737
536721b1 1738 bool dismantle;
a2835763
PM
1739
1740 enum {
1741 RTNL_LINK_INITIALIZED,
1742 RTNL_LINK_INITIALIZING,
1743 } rtnl_link_state:16;
1da177e4 1744
d314774c 1745 void (*destructor)(struct net_device *dev);
1da177e4 1746
1da177e4 1747#ifdef CONFIG_NETPOLL
5fbee843 1748 struct netpoll_info __rcu *npinfo;
1da177e4 1749#endif
eae792b7 1750
0c5c9fb5 1751 possible_net_t nd_net;
4a1c5371 1752
4951704b 1753 /* mid-layer private */
a7855c78 1754 union {
536721b1
KK
1755 void *ml_priv;
1756 struct pcpu_lstats __percpu *lstats;
8f84985f 1757 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1758 struct pcpu_dstats __percpu *dstats;
1759 struct pcpu_vstats __percpu *vstats;
a7855c78 1760 };
536721b1 1761
3cc77ec7 1762 struct garp_port __rcu *garp_port;
febf018d 1763 struct mrp_port __rcu *mrp_port;
1da177e4 1764
536721b1 1765 struct device dev;
0c509a6c 1766 const struct attribute_group *sysfs_groups[4];
a953be53 1767 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1768
38f7b870 1769 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1770
82cc1a7a
PWJ
1771 /* for setting kernel sock attribute on TCP connection setup */
1772#define GSO_MAX_SIZE 65536
1773 unsigned int gso_max_size;
30b678d8
BH
1774#define GSO_MAX_SEGS 65535
1775 u16 gso_max_segs;
fcbeb976 1776 u16 gso_min_segs;
7a6b6f51 1777#ifdef CONFIG_DCB
32953543 1778 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1779#endif
4f57c087
JF
1780 u8 num_tc;
1781 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1782 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1783
d11ead75 1784#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1785 unsigned int fcoe_ddp_xid;
5bc1421e 1786#endif
86f8515f 1787#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1788 struct netprio_map __rcu *priomap;
4d288d57 1789#endif
c1f19b51 1790 struct phy_device *phydev;
23d3b8bf 1791 struct lock_class_key *qdisc_tx_busylock;
d746d707 1792 bool proto_down;
1da177e4 1793};
43cb76d9 1794#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1795
1796#define NETDEV_ALIGN 32
1da177e4 1797
4f57c087
JF
1798static inline
1799int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1800{
1801 return dev->prio_tc_map[prio & TC_BITMASK];
1802}
1803
1804static inline
1805int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1806{
1807 if (tc >= dev->num_tc)
1808 return -EINVAL;
1809
1810 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1811 return 0;
1812}
1813
1814static inline
1815void netdev_reset_tc(struct net_device *dev)
1816{
1817 dev->num_tc = 0;
1818 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1819 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1820}
1821
1822static inline
1823int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1824{
1825 if (tc >= dev->num_tc)
1826 return -EINVAL;
1827
1828 dev->tc_to_txq[tc].count = count;
1829 dev->tc_to_txq[tc].offset = offset;
1830 return 0;
1831}
1832
1833static inline
1834int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1835{
1836 if (num_tc > TC_MAX_QUEUE)
1837 return -EINVAL;
1838
1839 dev->num_tc = num_tc;
1840 return 0;
1841}
1842
1843static inline
1844int netdev_get_num_tc(struct net_device *dev)
1845{
1846 return dev->num_tc;
1847}
1848
e8a0464c
DM
1849static inline
1850struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1851 unsigned int index)
1852{
1853 return &dev->_tx[index];
1854}
1855
10c51b56
DB
1856static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1857 const struct sk_buff *skb)
1858{
1859 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1860}
1861
e8a0464c
DM
1862static inline void netdev_for_each_tx_queue(struct net_device *dev,
1863 void (*f)(struct net_device *,
1864 struct netdev_queue *,
1865 void *),
1866 void *arg)
1867{
1868 unsigned int i;
1869
1870 for (i = 0; i < dev->num_tx_queues; i++)
1871 f(dev, &dev->_tx[i], arg);
1872}
1873
f629d208 1874struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1875 struct sk_buff *skb,
1876 void *accel_priv);
8c4c49df 1877
c346dca1
YH
1878/*
1879 * Net namespace inlines
1880 */
1881static inline
1882struct net *dev_net(const struct net_device *dev)
1883{
c2d9ba9b 1884 return read_pnet(&dev->nd_net);
c346dca1
YH
1885}
1886
1887static inline
f5aa23fd 1888void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 1889{
0c5c9fb5 1890 write_pnet(&dev->nd_net, net);
c346dca1
YH
1891}
1892
3e8a72d1 1893static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 1894{
3fc88677 1895#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
1896 if (dev->dsa_ptr != NULL)
1897 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 1898#endif
5aed85ce 1899 return false;
396138f0
LB
1900}
1901
bea3348e
SH
1902/**
1903 * netdev_priv - access network device private data
1904 * @dev: network device
1905 *
1906 * Get network device private data
1907 */
6472ce60 1908static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1909{
1ce8e7b5 1910 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1911}
1912
1da177e4
LT
1913/* Set the sysfs physical device reference for the network logical device
1914 * if set prior to registration will cause a symlink during initialization.
1915 */
43cb76d9 1916#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1917
384912ed 1918/* Set the sysfs device type for the network logical device to allow
3f79410c 1919 * fine-grained identification of different network device types. For
384912ed
MH
1920 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1921 */
1922#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1923
82dc3c63
ED
1924/* Default NAPI poll() weight
1925 * Device drivers are strongly advised to not use bigger value
1926 */
1927#define NAPI_POLL_WEIGHT 64
1928
3b582cc1
SH
1929/**
1930 * netif_napi_add - initialize a napi context
1931 * @dev: network device
1932 * @napi: napi context
1933 * @poll: polling function
1934 * @weight: default weight
1935 *
1936 * netif_napi_add() must be used to initialize a napi context prior to calling
1937 * *any* of the other napi related functions.
1938 */
d565b0a1
HX
1939void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1940 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1941
d64b5e85
ED
1942/**
1943 * netif_tx_napi_add - initialize a napi context
1944 * @dev: network device
1945 * @napi: napi context
1946 * @poll: polling function
1947 * @weight: default weight
1948 *
1949 * This variant of netif_napi_add() should be used from drivers using NAPI
1950 * to exclusively poll a TX queue.
1951 * This will avoid we add it into napi_hash[], thus polluting this hash table.
1952 */
1953static inline void netif_tx_napi_add(struct net_device *dev,
1954 struct napi_struct *napi,
1955 int (*poll)(struct napi_struct *, int),
1956 int weight)
1957{
1958 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
1959 netif_napi_add(dev, napi, poll, weight);
1960}
1961
d8156534
AD
1962/**
1963 * netif_napi_del - remove a napi context
1964 * @napi: napi context
1965 *
1966 * netif_napi_del() removes a napi context from the network device napi list
1967 */
d565b0a1
HX
1968void netif_napi_del(struct napi_struct *napi);
1969
1970struct napi_gro_cb {
78a478d0
HX
1971 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1972 void *frag0;
1973
7489594c
HX
1974 /* Length of frag0. */
1975 unsigned int frag0_len;
1976
86911732
HX
1977 /* This indicates where we are processing relative to skb->data. */
1978 int data_offset;
1979
d565b0a1 1980 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
1981 u16 flush;
1982
1983 /* Save the IP ID here and check when we get to the transport layer */
1984 u16 flush_id;
d565b0a1
HX
1985
1986 /* Number of segments aggregated. */
2e71a6f8
ED
1987 u16 count;
1988
15e2396d
TH
1989 /* Start offset for remote checksum offload */
1990 u16 gro_remcsum_start;
1991
2e71a6f8
ED
1992 /* jiffies when first packet was created/queued */
1993 unsigned long age;
86347245 1994
afe93325 1995 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
1996 u16 proto;
1997
baa32ff4
TH
1998 /* This is non-zero if the packet may be of the same flow. */
1999 u8 same_flow:1;
2000
b582ef09 2001 /* Used in udp_gro_receive */
573e8fca
TH
2002 u8 udp_mark:1;
2003
2004 /* GRO checksum is valid */
2005 u8 csum_valid:1;
2006
662880f4
TH
2007 /* Number of checksums via CHECKSUM_UNNECESSARY */
2008 u8 csum_cnt:3;
c3c7c254 2009
baa32ff4
TH
2010 /* Free the skb? */
2011 u8 free:2;
2012#define NAPI_GRO_FREE 1
2013#define NAPI_GRO_FREE_STOLEN_HEAD 2
2014
efc98d08
TH
2015 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2016 u8 is_ipv6:1;
2017
baa32ff4
TH
2018 /* 7 bit hole */
2019
bf5a755f
JC
2020 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2021 __wsum csum;
2022
c3c7c254
ED
2023 /* used in skb_gro_receive() slow path */
2024 struct sk_buff *last;
d565b0a1
HX
2025};
2026
2027#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2028
1da177e4 2029struct packet_type {
f2ccd8fa
DM
2030 __be16 type; /* This is really htons(ether_type). */
2031 struct net_device *dev; /* NULL is wildcarded here */
2032 int (*func) (struct sk_buff *,
2033 struct net_device *,
2034 struct packet_type *,
2035 struct net_device *);
c0de08d0
EL
2036 bool (*id_match)(struct packet_type *ptype,
2037 struct sock *sk);
1da177e4
LT
2038 void *af_packet_priv;
2039 struct list_head list;
2040};
2041
f191a1d1 2042struct offload_callbacks {
576a30eb 2043 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2044 netdev_features_t features);
d565b0a1 2045 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2046 struct sk_buff *skb);
299603e8 2047 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2048};
2049
2050struct packet_offload {
2051 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2052 u16 priority;
f191a1d1
VY
2053 struct offload_callbacks callbacks;
2054 struct list_head list;
1da177e4
LT
2055};
2056
a2b12f3c
TH
2057struct udp_offload;
2058
2059struct udp_offload_callbacks {
2060 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2061 struct sk_buff *skb,
2062 struct udp_offload *uoff);
2063 int (*gro_complete)(struct sk_buff *skb,
2064 int nhoff,
2065 struct udp_offload *uoff);
2066};
2067
b582ef09
OG
2068struct udp_offload {
2069 __be16 port;
afe93325 2070 u8 ipproto;
a2b12f3c 2071 struct udp_offload_callbacks callbacks;
b582ef09
OG
2072};
2073
8f84985f
LR
2074/* often modified stats are per cpu, other are shared (netdev->stats) */
2075struct pcpu_sw_netstats {
2076 u64 rx_packets;
2077 u64 rx_bytes;
2078 u64 tx_packets;
2079 u64 tx_bytes;
2080 struct u64_stats_sync syncp;
2081};
2082
aabc92bb
PNA
2083#define __netdev_alloc_pcpu_stats(type, gfp) \
2084({ \
2085 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2086 if (pcpu_stats) { \
2087 int __cpu; \
2088 for_each_possible_cpu(__cpu) { \
2089 typeof(type) *stat; \
2090 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2091 u64_stats_init(&stat->syncp); \
2092 } \
2093 } \
2094 pcpu_stats; \
1c213bd2
WC
2095})
2096
aabc92bb
PNA
2097#define netdev_alloc_pcpu_stats(type) \
2098 __netdev_alloc_pcpu_stats(type, GFP_KERNEL);
2099
1da177e4
LT
2100#include <linux/notifier.h>
2101
dcfe1421
AW
2102/* netdevice notifier chain. Please remember to update the rtnetlink
2103 * notification exclusion list in rtnetlink_event() when adding new
2104 * types.
2105 */
2106#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2107#define NETDEV_DOWN 0x0002
2108#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2109 detected a hardware crash and restarted
2110 - we can use this eg to kick tcp sessions
2111 once done */
2112#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2113#define NETDEV_REGISTER 0x0005
2114#define NETDEV_UNREGISTER 0x0006
1d486bfb 2115#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2116#define NETDEV_CHANGEADDR 0x0008
2117#define NETDEV_GOING_DOWN 0x0009
2118#define NETDEV_CHANGENAME 0x000A
2119#define NETDEV_FEAT_CHANGE 0x000B
2120#define NETDEV_BONDING_FAILOVER 0x000C
2121#define NETDEV_PRE_UP 0x000D
2122#define NETDEV_PRE_TYPE_CHANGE 0x000E
2123#define NETDEV_POST_TYPE_CHANGE 0x000F
2124#define NETDEV_POST_INIT 0x0010
0115e8e3 2125#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2126#define NETDEV_RELEASE 0x0012
2127#define NETDEV_NOTIFY_PEERS 0x0013
2128#define NETDEV_JOIN 0x0014
42e52bf9 2129#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2130#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2131#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2132#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2133#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2134#define NETDEV_PRECHANGEUPPER 0x001A
dcfe1421 2135
f629d208
JP
2136int register_netdevice_notifier(struct notifier_block *nb);
2137int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2138
2139struct netdev_notifier_info {
2140 struct net_device *dev;
2141};
2142
be9efd36
JP
2143struct netdev_notifier_change_info {
2144 struct netdev_notifier_info info; /* must be first */
2145 unsigned int flags_changed;
2146};
2147
0e4ead9d
JP
2148struct netdev_notifier_changeupper_info {
2149 struct netdev_notifier_info info; /* must be first */
2150 struct net_device *upper_dev; /* new upper dev */
2151 bool master; /* is upper dev master */
2152 bool linking; /* is the nofication for link or unlink */
2153};
2154
75538c2b
CW
2155static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2156 struct net_device *dev)
2157{
2158 info->dev = dev;
2159}
2160
351638e7
JP
2161static inline struct net_device *
2162netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2163{
2164 return info->dev;
2165}
2166
f629d208 2167int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2168
2169
1da177e4
LT
2170extern rwlock_t dev_base_lock; /* Device list lock */
2171
881d966b
EB
2172#define for_each_netdev(net, d) \
2173 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2174#define for_each_netdev_reverse(net, d) \
2175 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2176#define for_each_netdev_rcu(net, d) \
2177 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2178#define for_each_netdev_safe(net, d, n) \
2179 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2180#define for_each_netdev_continue(net, d) \
2181 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2182#define for_each_netdev_continue_rcu(net, d) \
2183 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2184#define for_each_netdev_in_bond_rcu(bond, slave) \
2185 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2186 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2187#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2188
a050c33f
DL
2189static inline struct net_device *next_net_device(struct net_device *dev)
2190{
2191 struct list_head *lh;
2192 struct net *net;
2193
c346dca1 2194 net = dev_net(dev);
a050c33f
DL
2195 lh = dev->dev_list.next;
2196 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2197}
2198
ce81b76a
ED
2199static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2200{
2201 struct list_head *lh;
2202 struct net *net;
2203
2204 net = dev_net(dev);
ccf43438 2205 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2206 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2207}
2208
a050c33f
DL
2209static inline struct net_device *first_net_device(struct net *net)
2210{
2211 return list_empty(&net->dev_base_head) ? NULL :
2212 net_device_entry(net->dev_base_head.next);
2213}
7562f876 2214
ccf43438
ED
2215static inline struct net_device *first_net_device_rcu(struct net *net)
2216{
2217 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2218
2219 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2220}
2221
f629d208
JP
2222int netdev_boot_setup_check(struct net_device *dev);
2223unsigned long netdev_boot_base(const char *prefix, int unit);
2224struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2225 const char *hwaddr);
2226struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2227struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2228void dev_add_pack(struct packet_type *pt);
2229void dev_remove_pack(struct packet_type *pt);
2230void __dev_remove_pack(struct packet_type *pt);
2231void dev_add_offload(struct packet_offload *po);
2232void dev_remove_offload(struct packet_offload *po);
f629d208 2233
a54acb3a 2234int dev_get_iflink(const struct net_device *dev);
fc4099f1 2235int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2236struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2237 unsigned short mask);
f629d208
JP
2238struct net_device *dev_get_by_name(struct net *net, const char *name);
2239struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2240struct net_device *__dev_get_by_name(struct net *net, const char *name);
2241int dev_alloc_name(struct net_device *dev, const char *name);
2242int dev_open(struct net_device *dev);
2243int dev_close(struct net_device *dev);
99c4a26a 2244int dev_close_many(struct list_head *head, bool unlink);
f629d208 2245void dev_disable_lro(struct net_device *dev);
0c4b51f0 2246int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2247int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2248int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2249int register_netdevice(struct net_device *dev);
2250void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2251void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2252static inline void unregister_netdevice(struct net_device *dev)
2253{
2254 unregister_netdevice_queue(dev, NULL);
2255}
2256
f629d208
JP
2257int netdev_refcnt_read(const struct net_device *dev);
2258void free_netdev(struct net_device *dev);
74d332c1 2259void netdev_freemem(struct net_device *dev);
f629d208
JP
2260void synchronize_net(void);
2261int init_dummy_netdev(struct net_device *dev);
937f1ba5 2262
f60e5990 2263DECLARE_PER_CPU(int, xmit_recursion);
2264static inline int dev_recursion_level(void)
2265{
2266 return this_cpu_read(xmit_recursion);
2267}
2268
f629d208
JP
2269struct net_device *dev_get_by_index(struct net *net, int ifindex);
2270struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2271struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2272int netdev_get_name(struct net *net, char *name, int ifindex);
2273int dev_restart(struct net_device *dev);
f629d208 2274int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2275
2276static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2277{
2278 return NAPI_GRO_CB(skb)->data_offset;
2279}
2280
2281static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2282{
2283 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2284}
2285
2286static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2287{
2288 NAPI_GRO_CB(skb)->data_offset += len;
2289}
2290
a5b1cf28
HX
2291static inline void *skb_gro_header_fast(struct sk_buff *skb,
2292 unsigned int offset)
86911732 2293{
a5b1cf28
HX
2294 return NAPI_GRO_CB(skb)->frag0 + offset;
2295}
78a478d0 2296
a5b1cf28
HX
2297static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2298{
2299 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2300}
78a478d0 2301
a5b1cf28
HX
2302static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2303 unsigned int offset)
2304{
17dd759c
HX
2305 if (!pskb_may_pull(skb, hlen))
2306 return NULL;
2307
a5b1cf28
HX
2308 NAPI_GRO_CB(skb)->frag0 = NULL;
2309 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2310 return skb->data + offset;
86911732 2311}
1da177e4 2312
36e7b1b8
HX
2313static inline void *skb_gro_network_header(struct sk_buff *skb)
2314{
78d3fd0b
HX
2315 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2316 skb_network_offset(skb);
36e7b1b8
HX
2317}
2318
bf5a755f
JC
2319static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2320 const void *start, unsigned int len)
2321{
573e8fca 2322 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2323 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2324 csum_partial(start, len, 0));
2325}
2326
573e8fca
TH
2327/* GRO checksum functions. These are logical equivalents of the normal
2328 * checksum functions (in skbuff.h) except that they operate on the GRO
2329 * offsets and fields in sk_buff.
2330 */
2331
2332__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2333
15e2396d
TH
2334static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2335{
b7fe10e5 2336 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2337}
2338
573e8fca
TH
2339static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2340 bool zero_okay,
2341 __sum16 check)
2342{
6edec0e6
TH
2343 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2344 skb_checksum_start_offset(skb) <
2345 skb_gro_offset(skb)) &&
15e2396d 2346 !skb_at_gro_remcsum_start(skb) &&
662880f4 2347 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2348 (!zero_okay || check));
2349}
2350
2351static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2352 __wsum psum)
2353{
2354 if (NAPI_GRO_CB(skb)->csum_valid &&
2355 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2356 return 0;
2357
2358 NAPI_GRO_CB(skb)->csum = psum;
2359
2360 return __skb_gro_checksum_complete(skb);
2361}
2362
573e8fca
TH
2363static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2364{
662880f4
TH
2365 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2366 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2367 NAPI_GRO_CB(skb)->csum_cnt--;
2368 } else {
2369 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2370 * verified a new top level checksum or an encapsulated one
2371 * during GRO. This saves work if we fallback to normal path.
2372 */
2373 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2374 }
2375}
2376
2377#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2378 compute_pseudo) \
2379({ \
2380 __sum16 __ret = 0; \
2381 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2382 __ret = __skb_gro_checksum_validate_complete(skb, \
2383 compute_pseudo(skb, proto)); \
5a212329
TH
2384 if (__ret) \
2385 __skb_mark_checksum_bad(skb); \
2386 else \
573e8fca
TH
2387 skb_gro_incr_csum_unnecessary(skb); \
2388 __ret; \
2389})
2390
2391#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2392 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2393
2394#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2395 compute_pseudo) \
2396 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2397
2398#define skb_gro_checksum_simple_validate(skb) \
2399 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2400
d96535a1
TH
2401static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2402{
2403 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2404 !NAPI_GRO_CB(skb)->csum_valid);
2405}
2406
2407static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2408 __sum16 check, __wsum pseudo)
2409{
2410 NAPI_GRO_CB(skb)->csum = ~pseudo;
2411 NAPI_GRO_CB(skb)->csum_valid = 1;
2412}
2413
2414#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2415do { \
2416 if (__skb_gro_checksum_convert_check(skb)) \
2417 __skb_gro_checksum_convert(skb, check, \
2418 compute_pseudo(skb, proto)); \
2419} while (0)
2420
26c4f7da
TH
2421struct gro_remcsum {
2422 int offset;
2423 __wsum delta;
2424};
2425
2426static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2427{
846cd667 2428 grc->offset = 0;
26c4f7da
TH
2429 grc->delta = 0;
2430}
2431
b7fe10e5
TH
2432static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2433 unsigned int off, size_t hdrlen,
2434 int start, int offset,
2435 struct gro_remcsum *grc,
2436 bool nopartial)
dcdc8994
TH
2437{
2438 __wsum delta;
b7fe10e5 2439 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2440
2441 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2442
15e2396d 2443 if (!nopartial) {
b7fe10e5
TH
2444 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2445 return ptr;
2446 }
2447
2448 ptr = skb_gro_header_fast(skb, off);
2449 if (skb_gro_header_hard(skb, off + plen)) {
2450 ptr = skb_gro_header_slow(skb, off + plen, off);
2451 if (!ptr)
2452 return NULL;
15e2396d
TH
2453 }
2454
b7fe10e5
TH
2455 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2456 start, offset);
dcdc8994
TH
2457
2458 /* Adjust skb->csum since we changed the packet */
dcdc8994 2459 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2460
b7fe10e5 2461 grc->offset = off + hdrlen + offset;
26c4f7da 2462 grc->delta = delta;
b7fe10e5
TH
2463
2464 return ptr;
dcdc8994
TH
2465}
2466
26c4f7da
TH
2467static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2468 struct gro_remcsum *grc)
2469{
b7fe10e5
TH
2470 void *ptr;
2471 size_t plen = grc->offset + sizeof(u16);
2472
26c4f7da
TH
2473 if (!grc->delta)
2474 return;
2475
b7fe10e5
TH
2476 ptr = skb_gro_header_fast(skb, grc->offset);
2477 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2478 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2479 if (!ptr)
2480 return;
2481 }
2482
2483 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2484}
dcdc8994 2485
0c4e8581
SH
2486static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2487 unsigned short type,
3b04ddde 2488 const void *daddr, const void *saddr,
95c96174 2489 unsigned int len)
0c4e8581 2490{
f1ecfd5d 2491 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2492 return 0;
3b04ddde
SH
2493
2494 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2495}
2496
b95cce35
SH
2497static inline int dev_parse_header(const struct sk_buff *skb,
2498 unsigned char *haddr)
2499{
2500 const struct net_device *dev = skb->dev;
2501
1b83336b 2502 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2503 return 0;
3b04ddde 2504 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2505}
2506
1da177e4 2507typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2508int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2509static inline int unregister_gifconf(unsigned int family)
2510{
2511 return register_gifconf(family, NULL);
2512}
2513
99bbc707 2514#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2515#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2516struct sd_flow_limit {
2517 u64 count;
2518 unsigned int num_buckets;
2519 unsigned int history_head;
2520 u16 history[FLOW_LIMIT_HISTORY];
2521 u8 buckets[];
2522};
2523
2524extern int netdev_flow_limit_table_len;
2525#endif /* CONFIG_NET_FLOW_LIMIT */
2526
1da177e4 2527/*
88751275 2528 * Incoming packets are placed on per-cpu queues
1da177e4 2529 */
d94d9fee 2530struct softnet_data {
1da177e4 2531 struct list_head poll_list;
6e7676c1 2532 struct sk_buff_head process_queue;
1da177e4 2533
dee42870 2534 /* stats */
cd7b5396
DM
2535 unsigned int processed;
2536 unsigned int time_squeeze;
2537 unsigned int cpu_collision;
2538 unsigned int received_rps;
fd793d89 2539#ifdef CONFIG_RPS
88751275 2540 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2541#endif
2542#ifdef CONFIG_NET_FLOW_LIMIT
2543 struct sd_flow_limit __rcu *flow_limit;
2544#endif
2545 struct Qdisc *output_queue;
2546 struct Qdisc **output_queue_tailp;
2547 struct sk_buff *completion_queue;
88751275 2548
4cdb1e2e 2549#ifdef CONFIG_RPS
88751275 2550 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 2551 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2552 struct softnet_data *rps_ipi_next;
2553 unsigned int cpu;
fec5e652 2554 unsigned int input_queue_head;
76cc8b13 2555 unsigned int input_queue_tail;
1e94d72f 2556#endif
95c96174 2557 unsigned int dropped;
0a9627f2 2558 struct sk_buff_head input_pkt_queue;
bea3348e 2559 struct napi_struct backlog;
99bbc707 2560
1da177e4
LT
2561};
2562
76cc8b13 2563static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2564{
2565#ifdef CONFIG_RPS
76cc8b13
TH
2566 sd->input_queue_head++;
2567#endif
2568}
2569
2570static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2571 unsigned int *qtail)
2572{
2573#ifdef CONFIG_RPS
2574 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2575#endif
2576}
2577
0a9627f2 2578DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2579
f629d208 2580void __netif_schedule(struct Qdisc *q);
46e5da40 2581void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2582
fd2ea0a7
DM
2583static inline void netif_tx_schedule_all(struct net_device *dev)
2584{
2585 unsigned int i;
2586
2587 for (i = 0; i < dev->num_tx_queues; i++)
2588 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2589}
2590
d29f749e
DJ
2591static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2592{
73466498 2593 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2594}
2595
bea3348e
SH
2596/**
2597 * netif_start_queue - allow transmit
2598 * @dev: network device
2599 *
2600 * Allow upper layers to call the device hard_start_xmit routine.
2601 */
1da177e4
LT
2602static inline void netif_start_queue(struct net_device *dev)
2603{
e8a0464c 2604 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2605}
2606
fd2ea0a7
DM
2607static inline void netif_tx_start_all_queues(struct net_device *dev)
2608{
2609 unsigned int i;
2610
2611 for (i = 0; i < dev->num_tx_queues; i++) {
2612 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2613 netif_tx_start_queue(txq);
2614 }
2615}
2616
46e5da40 2617void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2618
d29f749e
DJ
2619/**
2620 * netif_wake_queue - restart transmit
2621 * @dev: network device
2622 *
2623 * Allow upper layers to call the device hard_start_xmit routine.
2624 * Used for flow control when transmit resources are available.
2625 */
79d16385
DM
2626static inline void netif_wake_queue(struct net_device *dev)
2627{
e8a0464c 2628 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2629}
2630
fd2ea0a7
DM
2631static inline void netif_tx_wake_all_queues(struct net_device *dev)
2632{
2633 unsigned int i;
2634
2635 for (i = 0; i < dev->num_tx_queues; i++) {
2636 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2637 netif_tx_wake_queue(txq);
2638 }
2639}
2640
d29f749e
DJ
2641static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2642{
73466498 2643 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2644}
2645
bea3348e
SH
2646/**
2647 * netif_stop_queue - stop transmitted packets
2648 * @dev: network device
2649 *
2650 * Stop upper layers calling the device hard_start_xmit routine.
2651 * Used for flow control when transmit resources are unavailable.
2652 */
1da177e4
LT
2653static inline void netif_stop_queue(struct net_device *dev)
2654{
e8a0464c 2655 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2656}
2657
a2029240 2658void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2659
4d29515f 2660static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2661{
73466498 2662 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2663}
2664
bea3348e
SH
2665/**
2666 * netif_queue_stopped - test if transmit queue is flowblocked
2667 * @dev: network device
2668 *
2669 * Test if transmit queue on device is currently unable to send.
2670 */
4d29515f 2671static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2672{
e8a0464c 2673 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2674}
2675
4d29515f 2676static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2677{
73466498
TH
2678 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2679}
2680
8e2f1a63
DB
2681static inline bool
2682netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2683{
2684 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2685}
2686
8e2f1a63
DB
2687static inline bool
2688netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2689{
2690 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2691}
2692
53511453
ED
2693/**
2694 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2695 * @dev_queue: pointer to transmit queue
2696 *
2697 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2698 * to give appropriate hint to the cpu.
2699 */
2700static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2701{
2702#ifdef CONFIG_BQL
2703 prefetchw(&dev_queue->dql.num_queued);
2704#endif
2705}
2706
2707/**
2708 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2709 * @dev_queue: pointer to transmit queue
2710 *
2711 * BQL enabled drivers might use this helper in their TX completion path,
2712 * to give appropriate hint to the cpu.
2713 */
2714static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2715{
2716#ifdef CONFIG_BQL
2717 prefetchw(&dev_queue->dql.limit);
2718#endif
2719}
2720
c5d67bd7
TH
2721static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2722 unsigned int bytes)
2723{
114cf580
TH
2724#ifdef CONFIG_BQL
2725 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2726
2727 if (likely(dql_avail(&dev_queue->dql) >= 0))
2728 return;
2729
2730 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2731
2732 /*
2733 * The XOFF flag must be set before checking the dql_avail below,
2734 * because in netdev_tx_completed_queue we update the dql_completed
2735 * before checking the XOFF flag.
2736 */
2737 smp_mb();
2738
2739 /* check again in case another CPU has just made room avail */
2740 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2741 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2742#endif
c5d67bd7
TH
2743}
2744
0042d0c8
FF
2745/**
2746 * netdev_sent_queue - report the number of bytes queued to hardware
2747 * @dev: network device
2748 * @bytes: number of bytes queued to the hardware device queue
2749 *
2750 * Report the number of bytes queued for sending/completion to the network
2751 * device hardware queue. @bytes should be a good approximation and should
2752 * exactly match netdev_completed_queue() @bytes
2753 */
c5d67bd7
TH
2754static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2755{
2756 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2757}
2758
2759static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2760 unsigned int pkts, unsigned int bytes)
c5d67bd7 2761{
114cf580 2762#ifdef CONFIG_BQL
b37c0fbe
AD
2763 if (unlikely(!bytes))
2764 return;
2765
2766 dql_completed(&dev_queue->dql, bytes);
2767
2768 /*
2769 * Without the memory barrier there is a small possiblity that
2770 * netdev_tx_sent_queue will miss the update and cause the queue to
2771 * be stopped forever
2772 */
2773 smp_mb();
2774
2775 if (dql_avail(&dev_queue->dql) < 0)
2776 return;
2777
2778 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2779 netif_schedule_queue(dev_queue);
114cf580 2780#endif
c5d67bd7
TH
2781}
2782
0042d0c8
FF
2783/**
2784 * netdev_completed_queue - report bytes and packets completed by device
2785 * @dev: network device
2786 * @pkts: actual number of packets sent over the medium
2787 * @bytes: actual number of bytes sent over the medium
2788 *
2789 * Report the number of bytes and packets transmitted by the network device
2790 * hardware queue over the physical medium, @bytes must exactly match the
2791 * @bytes amount passed to netdev_sent_queue()
2792 */
c5d67bd7 2793static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2794 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2795{
2796 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2797}
2798
2799static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2800{
114cf580 2801#ifdef CONFIG_BQL
5c490354 2802 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2803 dql_reset(&q->dql);
2804#endif
c5d67bd7
TH
2805}
2806
0042d0c8
FF
2807/**
2808 * netdev_reset_queue - reset the packets and bytes count of a network device
2809 * @dev_queue: network device
2810 *
2811 * Reset the bytes and packet count of a network device and clear the
2812 * software flow control OFF bit for this network device
2813 */
c5d67bd7
TH
2814static inline void netdev_reset_queue(struct net_device *dev_queue)
2815{
2816 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2817}
2818
b9507bda
DB
2819/**
2820 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2821 * @dev: network device
2822 * @queue_index: given tx queue index
2823 *
2824 * Returns 0 if given tx queue index >= number of device tx queues,
2825 * otherwise returns the originally passed tx queue index.
2826 */
2827static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2828{
2829 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2830 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2831 dev->name, queue_index,
2832 dev->real_num_tx_queues);
2833 return 0;
2834 }
2835
2836 return queue_index;
2837}
2838
bea3348e
SH
2839/**
2840 * netif_running - test if up
2841 * @dev: network device
2842 *
2843 * Test if the device has been brought up.
2844 */
4d29515f 2845static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2846{
2847 return test_bit(__LINK_STATE_START, &dev->state);
2848}
2849
f25f4e44
PWJ
2850/*
2851 * Routines to manage the subqueues on a device. We only need start
2852 * stop, and a check if it's stopped. All other device management is
2853 * done at the overall netdevice level.
2854 * Also test the device if we're multiqueue.
2855 */
bea3348e
SH
2856
2857/**
2858 * netif_start_subqueue - allow sending packets on subqueue
2859 * @dev: network device
2860 * @queue_index: sub queue index
2861 *
2862 * Start individual transmit queue of a device with multiple transmit queues.
2863 */
f25f4e44
PWJ
2864static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2865{
fd2ea0a7 2866 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2867
2868 netif_tx_start_queue(txq);
f25f4e44
PWJ
2869}
2870
bea3348e
SH
2871/**
2872 * netif_stop_subqueue - stop sending packets on subqueue
2873 * @dev: network device
2874 * @queue_index: sub queue index
2875 *
2876 * Stop individual transmit queue of a device with multiple transmit queues.
2877 */
f25f4e44
PWJ
2878static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2879{
fd2ea0a7 2880 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 2881 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2882}
2883
bea3348e
SH
2884/**
2885 * netif_subqueue_stopped - test status of subqueue
2886 * @dev: network device
2887 * @queue_index: sub queue index
2888 *
2889 * Check individual transmit queue of a device with multiple transmit queues.
2890 */
4d29515f
DM
2891static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2892 u16 queue_index)
f25f4e44 2893{
fd2ea0a7 2894 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2895
2896 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2897}
2898
4d29515f
DM
2899static inline bool netif_subqueue_stopped(const struct net_device *dev,
2900 struct sk_buff *skb)
668f895a
PE
2901{
2902 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2903}
bea3348e 2904
46e5da40 2905void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 2906
537c00de 2907#ifdef CONFIG_XPS
53af53ae 2908int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2909 u16 index);
537c00de
AD
2910#else
2911static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2912 const struct cpumask *mask,
537c00de
AD
2913 u16 index)
2914{
2915 return 0;
2916}
2917#endif
2918
5605c762
JP
2919u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2920 unsigned int num_tx_queues);
2921
a3d22a68
VZ
2922/*
2923 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2924 * as a distribution range limit for the returned value.
2925 */
2926static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 2927 struct sk_buff *skb)
a3d22a68
VZ
2928{
2929 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2930}
2931
bea3348e
SH
2932/**
2933 * netif_is_multiqueue - test if device has multiple transmit queues
2934 * @dev: network device
2935 *
2936 * Check if device has multiple transmit queues
bea3348e 2937 */
4d29515f 2938static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2939{
a02cec21 2940 return dev->num_tx_queues > 1;
f25f4e44 2941}
1da177e4 2942
f629d208 2943int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2944
a953be53 2945#ifdef CONFIG_SYSFS
f629d208 2946int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
2947#else
2948static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2949 unsigned int rxq)
2950{
2951 return 0;
2952}
2953#endif
2954
a953be53
MD
2955#ifdef CONFIG_SYSFS
2956static inline unsigned int get_netdev_rx_queue_index(
2957 struct netdev_rx_queue *queue)
2958{
2959 struct net_device *dev = queue->dev;
2960 int index = queue - dev->_rx;
2961
2962 BUG_ON(index >= dev->num_rx_queues);
2963 return index;
2964}
2965#endif
2966
16917b87 2967#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 2968int netif_get_num_default_rss_queues(void);
16917b87 2969
e6247027
ED
2970enum skb_free_reason {
2971 SKB_REASON_CONSUMED,
2972 SKB_REASON_DROPPED,
2973};
2974
2975void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2976void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 2977
e6247027
ED
2978/*
2979 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2980 * interrupt context or with hardware interrupts being disabled.
2981 * (in_irq() || irqs_disabled())
2982 *
2983 * We provide four helpers that can be used in following contexts :
2984 *
2985 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2986 * replacing kfree_skb(skb)
2987 *
2988 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2989 * Typically used in place of consume_skb(skb) in TX completion path
2990 *
2991 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2992 * replacing kfree_skb(skb)
2993 *
2994 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2995 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 2996 */
e6247027
ED
2997static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2998{
2999 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3000}
3001
3002static inline void dev_consume_skb_irq(struct sk_buff *skb)
3003{
3004 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3005}
3006
3007static inline void dev_kfree_skb_any(struct sk_buff *skb)
3008{
3009 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3010}
3011
3012static inline void dev_consume_skb_any(struct sk_buff *skb)
3013{
3014 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3015}
1da177e4 3016
f629d208
JP
3017int netif_rx(struct sk_buff *skb);
3018int netif_rx_ni(struct sk_buff *skb);
04eb4489 3019int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3020gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3021void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3022struct sk_buff *napi_get_frags(struct napi_struct *napi);
3023gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3024struct packet_offload *gro_find_receive_by_type(__be16 type);
3025struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3026
3027static inline void napi_free_frags(struct napi_struct *napi)
3028{
3029 kfree_skb(napi->skb);
3030 napi->skb = NULL;
3031}
3032
f629d208
JP
3033int netdev_rx_handler_register(struct net_device *dev,
3034 rx_handler_func_t *rx_handler,
3035 void *rx_handler_data);
3036void netdev_rx_handler_unregister(struct net_device *dev);
3037
3038bool dev_valid_name(const char *name);
3039int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3040int dev_ethtool(struct net *net, struct ifreq *);
3041unsigned int dev_get_flags(const struct net_device *);
3042int __dev_change_flags(struct net_device *, unsigned int flags);
3043int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3044void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3045 unsigned int gchanges);
f629d208
JP
3046int dev_change_name(struct net_device *, const char *);
3047int dev_set_alias(struct net_device *, const char *, size_t);
3048int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3049int dev_set_mtu(struct net_device *, int);
3050void dev_set_group(struct net_device *, int);
3051int dev_set_mac_address(struct net_device *, struct sockaddr *);
3052int dev_change_carrier(struct net_device *, bool new_carrier);
3053int dev_get_phys_port_id(struct net_device *dev,
02637fce 3054 struct netdev_phys_item_id *ppid);
db24a904
DA
3055int dev_get_phys_port_name(struct net_device *dev,
3056 char *name, size_t len);
d746d707 3057int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3058struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3059struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3060 struct netdev_queue *txq, int *ret);
a0265d28 3061int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3062int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1ee481fb 3063bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
1da177e4 3064
20380731 3065extern int netdev_budget;
1da177e4
LT
3066
3067/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3068void netdev_run_todo(void);
1da177e4 3069
bea3348e
SH
3070/**
3071 * dev_put - release reference to device
3072 * @dev: network device
3073 *
9ef4429b 3074 * Release reference to device to allow it to be freed.
bea3348e 3075 */
1da177e4
LT
3076static inline void dev_put(struct net_device *dev)
3077{
933393f5 3078 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3079}
3080
bea3348e
SH
3081/**
3082 * dev_hold - get reference to device
3083 * @dev: network device
3084 *
9ef4429b 3085 * Hold reference to device to keep it from being freed.
bea3348e 3086 */
15333061
SH
3087static inline void dev_hold(struct net_device *dev)
3088{
933393f5 3089 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3090}
1da177e4
LT
3091
3092/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3093 * and _off may be called from IRQ context, but it is caller
3094 * who is responsible for serialization of these calls.
b00055aa
SR
3095 *
3096 * The name carrier is inappropriate, these functions should really be
3097 * called netif_lowerlayer_*() because they represent the state of any
3098 * kind of lower layer not just hardware media.
1da177e4
LT
3099 */
3100
f629d208
JP
3101void linkwatch_init_dev(struct net_device *dev);
3102void linkwatch_fire_event(struct net_device *dev);
3103void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3104
bea3348e
SH
3105/**
3106 * netif_carrier_ok - test if carrier present
3107 * @dev: network device
3108 *
3109 * Check if carrier is present on device
3110 */
4d29515f 3111static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3112{
3113 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3114}
3115
f629d208 3116unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3117
f629d208 3118void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3119
f629d208 3120void netif_carrier_on(struct net_device *dev);
1da177e4 3121
f629d208 3122void netif_carrier_off(struct net_device *dev);
1da177e4 3123
bea3348e
SH
3124/**
3125 * netif_dormant_on - mark device as dormant.
3126 * @dev: network device
3127 *
3128 * Mark device as dormant (as per RFC2863).
3129 *
3130 * The dormant state indicates that the relevant interface is not
3131 * actually in a condition to pass packets (i.e., it is not 'up') but is
3132 * in a "pending" state, waiting for some external event. For "on-
3133 * demand" interfaces, this new state identifies the situation where the
3134 * interface is waiting for events to place it in the up state.
3135 *
3136 */
b00055aa
SR
3137static inline void netif_dormant_on(struct net_device *dev)
3138{
3139 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3140 linkwatch_fire_event(dev);
3141}
3142
bea3348e
SH
3143/**
3144 * netif_dormant_off - set device as not dormant.
3145 * @dev: network device
3146 *
3147 * Device is not in dormant state.
3148 */
b00055aa
SR
3149static inline void netif_dormant_off(struct net_device *dev)
3150{
3151 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3152 linkwatch_fire_event(dev);
3153}
3154
bea3348e
SH
3155/**
3156 * netif_dormant - test if carrier present
3157 * @dev: network device
3158 *
3159 * Check if carrier is present on device
3160 */
4d29515f 3161static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3162{
3163 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3164}
3165
3166
bea3348e
SH
3167/**
3168 * netif_oper_up - test if device is operational
3169 * @dev: network device
3170 *
3171 * Check if carrier is operational
3172 */
4d29515f 3173static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3174{
b00055aa
SR
3175 return (dev->operstate == IF_OPER_UP ||
3176 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3177}
3178
bea3348e
SH
3179/**
3180 * netif_device_present - is device available or removed
3181 * @dev: network device
3182 *
3183 * Check if device has not been removed from system.
3184 */
4d29515f 3185static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3186{
3187 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3188}
3189
f629d208 3190void netif_device_detach(struct net_device *dev);
1da177e4 3191
f629d208 3192void netif_device_attach(struct net_device *dev);
1da177e4
LT
3193
3194/*
3195 * Network interface message level settings
3196 */
1da177e4
LT
3197
3198enum {
3199 NETIF_MSG_DRV = 0x0001,
3200 NETIF_MSG_PROBE = 0x0002,
3201 NETIF_MSG_LINK = 0x0004,
3202 NETIF_MSG_TIMER = 0x0008,
3203 NETIF_MSG_IFDOWN = 0x0010,
3204 NETIF_MSG_IFUP = 0x0020,
3205 NETIF_MSG_RX_ERR = 0x0040,
3206 NETIF_MSG_TX_ERR = 0x0080,
3207 NETIF_MSG_TX_QUEUED = 0x0100,
3208 NETIF_MSG_INTR = 0x0200,
3209 NETIF_MSG_TX_DONE = 0x0400,
3210 NETIF_MSG_RX_STATUS = 0x0800,
3211 NETIF_MSG_PKTDATA = 0x1000,
3212 NETIF_MSG_HW = 0x2000,
3213 NETIF_MSG_WOL = 0x4000,
3214};
3215
3216#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3217#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3218#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3219#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3220#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3221#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3222#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3223#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3224#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3225#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3226#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3227#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3228#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3229#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3230#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3231
3232static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3233{
3234 /* use default */
3235 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3236 return default_msg_enable_bits;
3237 if (debug_value == 0) /* no output */
3238 return 0;
3239 /* set low N bits */
3240 return (1 << debug_value) - 1;
3241}
3242
c773e847 3243static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3244{
c773e847
DM
3245 spin_lock(&txq->_xmit_lock);
3246 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3247}
3248
fd2ea0a7
DM
3249static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3250{
3251 spin_lock_bh(&txq->_xmit_lock);
3252 txq->xmit_lock_owner = smp_processor_id();
3253}
3254
4d29515f 3255static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3256{
4d29515f 3257 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3258 if (likely(ok))
3259 txq->xmit_lock_owner = smp_processor_id();
3260 return ok;
3261}
3262
3263static inline void __netif_tx_unlock(struct netdev_queue *txq)
3264{
3265 txq->xmit_lock_owner = -1;
3266 spin_unlock(&txq->_xmit_lock);
3267}
3268
3269static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3270{
3271 txq->xmit_lock_owner = -1;
3272 spin_unlock_bh(&txq->_xmit_lock);
3273}
3274
08baf561
ED
3275static inline void txq_trans_update(struct netdev_queue *txq)
3276{
3277 if (txq->xmit_lock_owner != -1)
3278 txq->trans_start = jiffies;
3279}
3280
d29f749e
DJ
3281/**
3282 * netif_tx_lock - grab network device transmit lock
3283 * @dev: network device
d29f749e
DJ
3284 *
3285 * Get network device transmit lock
3286 */
22dd7495
JHS
3287static inline void netif_tx_lock(struct net_device *dev)
3288{
e8a0464c 3289 unsigned int i;
c3f26a26 3290 int cpu;
c773e847 3291
c3f26a26
DM
3292 spin_lock(&dev->tx_global_lock);
3293 cpu = smp_processor_id();
e8a0464c
DM
3294 for (i = 0; i < dev->num_tx_queues; i++) {
3295 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3296
3297 /* We are the only thread of execution doing a
3298 * freeze, but we have to grab the _xmit_lock in
3299 * order to synchronize with threads which are in
3300 * the ->hard_start_xmit() handler and already
3301 * checked the frozen bit.
3302 */
e8a0464c 3303 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3304 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3305 __netif_tx_unlock(txq);
e8a0464c 3306 }
932ff279
HX
3307}
3308
3309static inline void netif_tx_lock_bh(struct net_device *dev)
3310{
e8a0464c
DM
3311 local_bh_disable();
3312 netif_tx_lock(dev);
932ff279
HX
3313}
3314
932ff279
HX
3315static inline void netif_tx_unlock(struct net_device *dev)
3316{
e8a0464c
DM
3317 unsigned int i;
3318
3319 for (i = 0; i < dev->num_tx_queues; i++) {
3320 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3321
c3f26a26
DM
3322 /* No need to grab the _xmit_lock here. If the
3323 * queue is not stopped for another reason, we
3324 * force a schedule.
3325 */
3326 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3327 netif_schedule_queue(txq);
c3f26a26
DM
3328 }
3329 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3330}
3331
3332static inline void netif_tx_unlock_bh(struct net_device *dev)
3333{
e8a0464c
DM
3334 netif_tx_unlock(dev);
3335 local_bh_enable();
932ff279
HX
3336}
3337
c773e847 3338#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3339 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3340 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3341 } \
3342}
3343
5efeac44
EB
3344#define HARD_TX_TRYLOCK(dev, txq) \
3345 (((dev->features & NETIF_F_LLTX) == 0) ? \
3346 __netif_tx_trylock(txq) : \
3347 true )
3348
c773e847 3349#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3350 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3351 __netif_tx_unlock(txq); \
22dd7495
JHS
3352 } \
3353}
3354
1da177e4
LT
3355static inline void netif_tx_disable(struct net_device *dev)
3356{
fd2ea0a7 3357 unsigned int i;
c3f26a26 3358 int cpu;
fd2ea0a7 3359
c3f26a26
DM
3360 local_bh_disable();
3361 cpu = smp_processor_id();
fd2ea0a7
DM
3362 for (i = 0; i < dev->num_tx_queues; i++) {
3363 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3364
3365 __netif_tx_lock(txq, cpu);
fd2ea0a7 3366 netif_tx_stop_queue(txq);
c3f26a26 3367 __netif_tx_unlock(txq);
fd2ea0a7 3368 }
c3f26a26 3369 local_bh_enable();
1da177e4
LT
3370}
3371
e308a5d8
DM
3372static inline void netif_addr_lock(struct net_device *dev)
3373{
3374 spin_lock(&dev->addr_list_lock);
3375}
3376
2429f7ac
JP
3377static inline void netif_addr_lock_nested(struct net_device *dev)
3378{
25175ba5
VY
3379 int subclass = SINGLE_DEPTH_NESTING;
3380
3381 if (dev->netdev_ops->ndo_get_lock_subclass)
3382 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3383
3384 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3385}
3386
e308a5d8
DM
3387static inline void netif_addr_lock_bh(struct net_device *dev)
3388{
3389 spin_lock_bh(&dev->addr_list_lock);
3390}
3391
3392static inline void netif_addr_unlock(struct net_device *dev)
3393{
3394 spin_unlock(&dev->addr_list_lock);
3395}
3396
3397static inline void netif_addr_unlock_bh(struct net_device *dev)
3398{
3399 spin_unlock_bh(&dev->addr_list_lock);
3400}
3401
f001fde5 3402/*
31278e71 3403 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3404 * rcu_read_lock held.
3405 */
3406#define for_each_dev_addr(dev, ha) \
31278e71 3407 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3408
1da177e4
LT
3409/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3410
f629d208 3411void ether_setup(struct net_device *dev);
1da177e4
LT
3412
3413/* Support for loadable net-drivers */
f629d208 3414struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3415 unsigned char name_assign_type,
f629d208
JP
3416 void (*setup)(struct net_device *),
3417 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3418#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3419 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3420
c835a677
TG
3421#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3422 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3423 count)
36909ea4 3424
f629d208
JP
3425int register_netdev(struct net_device *dev);
3426void unregister_netdev(struct net_device *dev);
f001fde5 3427
22bedad3 3428/* General hardware address lists handling functions */
f629d208
JP
3429int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3430 struct netdev_hw_addr_list *from_list, int addr_len);
3431void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3432 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3433int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3434 struct net_device *dev,
3435 int (*sync)(struct net_device *, const unsigned char *),
3436 int (*unsync)(struct net_device *,
3437 const unsigned char *));
3438void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3439 struct net_device *dev,
3440 int (*unsync)(struct net_device *,
3441 const unsigned char *));
f629d208 3442void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3443
f001fde5 3444/* Functions used for device addresses handling */
f629d208
JP
3445int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3446 unsigned char addr_type);
3447int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3448 unsigned char addr_type);
f629d208
JP
3449void dev_addr_flush(struct net_device *dev);
3450int dev_addr_init(struct net_device *dev);
a748ee24
JP
3451
3452/* Functions used for unicast addresses handling */
f629d208
JP
3453int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3454int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3455int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3456int dev_uc_sync(struct net_device *to, struct net_device *from);
3457int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3458void dev_uc_unsync(struct net_device *to, struct net_device *from);
3459void dev_uc_flush(struct net_device *dev);
3460void dev_uc_init(struct net_device *dev);
f001fde5 3461
670e5b8e
AD
3462/**
3463 * __dev_uc_sync - Synchonize device's unicast list
3464 * @dev: device to sync
3465 * @sync: function to call if address should be added
3466 * @unsync: function to call if address should be removed
3467 *
3468 * Add newly added addresses to the interface, and release
3469 * addresses that have been deleted.
3470 **/
3471static inline int __dev_uc_sync(struct net_device *dev,
3472 int (*sync)(struct net_device *,
3473 const unsigned char *),
3474 int (*unsync)(struct net_device *,
3475 const unsigned char *))
3476{
3477 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3478}
3479
3480/**
e793c0f7 3481 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3482 * @dev: device to sync
3483 * @unsync: function to call if address should be removed
3484 *
3485 * Remove all addresses that were added to the device by dev_uc_sync().
3486 **/
3487static inline void __dev_uc_unsync(struct net_device *dev,
3488 int (*unsync)(struct net_device *,
3489 const unsigned char *))
3490{
3491 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3492}
3493
22bedad3 3494/* Functions used for multicast addresses handling */
f629d208
JP
3495int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3496int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3497int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3498int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3499int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3500int dev_mc_sync(struct net_device *to, struct net_device *from);
3501int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3502void dev_mc_unsync(struct net_device *to, struct net_device *from);
3503void dev_mc_flush(struct net_device *dev);
3504void dev_mc_init(struct net_device *dev);
f001fde5 3505
670e5b8e
AD
3506/**
3507 * __dev_mc_sync - Synchonize device's multicast list
3508 * @dev: device to sync
3509 * @sync: function to call if address should be added
3510 * @unsync: function to call if address should be removed
3511 *
3512 * Add newly added addresses to the interface, and release
3513 * addresses that have been deleted.
3514 **/
3515static inline int __dev_mc_sync(struct net_device *dev,
3516 int (*sync)(struct net_device *,
3517 const unsigned char *),
3518 int (*unsync)(struct net_device *,
3519 const unsigned char *))
3520{
3521 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3522}
3523
3524/**
e793c0f7 3525 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3526 * @dev: device to sync
3527 * @unsync: function to call if address should be removed
3528 *
3529 * Remove all addresses that were added to the device by dev_mc_sync().
3530 **/
3531static inline void __dev_mc_unsync(struct net_device *dev,
3532 int (*unsync)(struct net_device *,
3533 const unsigned char *))
3534{
3535 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3536}
3537
4417da66 3538/* Functions used for secondary unicast and multicast support */
f629d208
JP
3539void dev_set_rx_mode(struct net_device *dev);
3540void __dev_set_rx_mode(struct net_device *dev);
3541int dev_set_promiscuity(struct net_device *dev, int inc);
3542int dev_set_allmulti(struct net_device *dev, int inc);
3543void netdev_state_change(struct net_device *dev);
3544void netdev_notify_peers(struct net_device *dev);
3545void netdev_features_change(struct net_device *dev);
1da177e4 3546/* Load a device via the kmod */
f629d208
JP
3547void dev_load(struct net *net, const char *name);
3548struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3549 struct rtnl_link_stats64 *storage);
3550void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3551 const struct net_device_stats *netdev_stats);
eeda3fd6 3552
1da177e4 3553extern int netdev_max_backlog;
3b098e2d 3554extern int netdev_tstamp_prequeue;
1da177e4 3555extern int weight_p;
0a14842f 3556extern int bpf_jit_enable;
9ff162a8 3557
f629d208 3558bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3559struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3560 struct list_head **iter);
f629d208
JP
3561struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3562 struct list_head **iter);
8b5be856 3563
44a40855
VY
3564/* iterate through upper list, must be called under RCU read lock */
3565#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3566 for (iter = &(dev)->adj_list.upper, \
3567 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3568 updev; \
3569 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3570
8b5be856 3571/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3572#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3573 for (iter = &(dev)->all_adj_list.upper, \
3574 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3575 updev; \
3576 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3577
f629d208
JP
3578void *netdev_lower_get_next_private(struct net_device *dev,
3579 struct list_head **iter);
3580void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3581 struct list_head **iter);
31088a11
VF
3582
3583#define netdev_for_each_lower_private(dev, priv, iter) \
3584 for (iter = (dev)->adj_list.lower.next, \
3585 priv = netdev_lower_get_next_private(dev, &(iter)); \
3586 priv; \
3587 priv = netdev_lower_get_next_private(dev, &(iter)))
3588
3589#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3590 for (iter = &(dev)->adj_list.lower, \
3591 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3592 priv; \
3593 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3594
4085ebe8
VY
3595void *netdev_lower_get_next(struct net_device *dev,
3596 struct list_head **iter);
3597#define netdev_for_each_lower_dev(dev, ldev, iter) \
3598 for (iter = &(dev)->adj_list.lower, \
3599 ldev = netdev_lower_get_next(dev, &(iter)); \
3600 ldev; \
3601 ldev = netdev_lower_get_next(dev, &(iter)))
3602
f629d208 3603void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3604void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3605struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3606struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3607int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3608int netdev_master_upper_dev_link(struct net_device *dev,
9ff162a8 3609 struct net_device *upper_dev);
f629d208
JP
3610int netdev_master_upper_dev_link_private(struct net_device *dev,
3611 struct net_device *upper_dev,
3612 void *private);
3613void netdev_upper_dev_unlink(struct net_device *dev,
3614 struct net_device *upper_dev);
5bb025fa 3615void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3616void *netdev_lower_dev_get_private(struct net_device *dev,
3617 struct net_device *lower_dev);
960fb622
ED
3618
3619/* RSS keys are 40 or 52 bytes long */
3620#define NETDEV_RSS_KEY_LEN 52
3621extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3622void netdev_rss_key_fill(void *buffer, size_t len);
3623
4085ebe8
VY
3624int dev_get_nest_level(struct net_device *dev,
3625 bool (*type_check)(struct net_device *dev));
f629d208
JP
3626int skb_checksum_help(struct sk_buff *skb);
3627struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3628 netdev_features_t features, bool tx_path);
3629struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3630 netdev_features_t features);
12b0004d 3631
61bd3857
MS
3632struct netdev_bonding_info {
3633 ifslave slave;
3634 ifbond master;
3635};
3636
3637struct netdev_notifier_bonding_info {
3638 struct netdev_notifier_info info; /* must be first */
3639 struct netdev_bonding_info bonding_info;
3640};
3641
3642void netdev_bonding_info_change(struct net_device *dev,
3643 struct netdev_bonding_info *bonding_info);
3644
12b0004d
CW
3645static inline
3646struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3647{
3648 return __skb_gso_segment(skb, features, true);
3649}
53d6471c 3650__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3651
3652static inline bool can_checksum_protocol(netdev_features_t features,
3653 __be16 protocol)
3654{
3655 return ((features & NETIF_F_GEN_CSUM) ||
3656 ((features & NETIF_F_V4_CSUM) &&
3657 protocol == htons(ETH_P_IP)) ||
3658 ((features & NETIF_F_V6_CSUM) &&
3659 protocol == htons(ETH_P_IPV6)) ||
3660 ((features & NETIF_F_FCOE_CRC) &&
3661 protocol == htons(ETH_P_FCOE)));
3662}
12b0004d 3663
fb286bb2 3664#ifdef CONFIG_BUG
f629d208 3665void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3666#else
3667static inline void netdev_rx_csum_fault(struct net_device *dev)
3668{
3669}
3670#endif
1da177e4 3671/* rx skb timestamps */
f629d208
JP
3672void net_enable_timestamp(void);
3673void net_disable_timestamp(void);
1da177e4 3674
20380731 3675#ifdef CONFIG_PROC_FS
f629d208 3676int __init dev_proc_init(void);
900ff8c6
CW
3677#else
3678#define dev_proc_init() 0
20380731
ACM
3679#endif
3680
4798248e 3681static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3682 struct sk_buff *skb, struct net_device *dev,
3683 bool more)
4798248e 3684{
fa2dbdc2 3685 skb->xmit_more = more ? 1 : 0;
0b725a2c 3686 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3687}
3688
10b3ad8c 3689static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3690 struct netdev_queue *txq, bool more)
4798248e
DM
3691{
3692 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3693 int rc;
4798248e 3694
fa2dbdc2 3695 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3696 if (rc == NETDEV_TX_OK)
3697 txq_trans_update(txq);
3698
3699 return rc;
4798248e
DM
3700}
3701
42a2d923
LT
3702int netdev_class_create_file_ns(struct class_attribute *class_attr,
3703 const void *ns);
3704void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3705 const void *ns);
58292cbe
TH
3706
3707static inline int netdev_class_create_file(struct class_attribute *class_attr)
3708{
3709 return netdev_class_create_file_ns(class_attr, NULL);
3710}
3711
3712static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3713{
3714 netdev_class_remove_file_ns(class_attr, NULL);
3715}
b8a9787e 3716
04600794
JB
3717extern struct kobj_ns_type_operations net_ns_type_operations;
3718
f629d208 3719const char *netdev_drivername(const struct net_device *dev);
6579e57b 3720
f629d208 3721void linkwatch_run_queue(void);
20380731 3722
da08143b
MK
3723static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3724 netdev_features_t f2)
3725{
3726 if (f1 & NETIF_F_GEN_CSUM)
3727 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3728 if (f2 & NETIF_F_GEN_CSUM)
3729 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3730 f1 &= f2;
3731 if (f1 & NETIF_F_GEN_CSUM)
3732 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3733
3734 return f1;
3735}
3736
c8f44aff
MM
3737static inline netdev_features_t netdev_get_wanted_features(
3738 struct net_device *dev)
5455c699
MM
3739{
3740 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3741}
c8f44aff
MM
3742netdev_features_t netdev_increment_features(netdev_features_t all,
3743 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3744
3745/* Allow TSO being used on stacked device :
3746 * Performing the GSO segmentation before last device
3747 * is a performance improvement.
3748 */
3749static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3750 netdev_features_t mask)
3751{
3752 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3753}
3754
6cb6a27c 3755int __netdev_update_features(struct net_device *dev);
5455c699 3756void netdev_update_features(struct net_device *dev);
afe12cc8 3757void netdev_change_features(struct net_device *dev);
7f353bf2 3758
fc4a7489
PM
3759void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3760 struct net_device *dev);
3761
e38f3025
TM
3762netdev_features_t passthru_features_check(struct sk_buff *skb,
3763 struct net_device *dev,
3764 netdev_features_t features);
c1e756bf 3765netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3766
4d29515f 3767static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3768{
c8f44aff 3769 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3770
3771 /* check flags correspondence */
3772 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3773 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3774 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3775 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3776 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3777 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
3778 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3779 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3780 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3781 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3782 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3783 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
e585f236 3784 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
0345e186 3785
d6b4991a 3786 return (features & feature) == feature;
576a30eb
HX
3787}
3788
4d29515f 3789static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3790{
278b2513 3791 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3792 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3793}
3794
8b86a61d 3795static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 3796 netdev_features_t features)
7967168c 3797{
fc741216 3798 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3799 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3800 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3801}
3802
82cc1a7a
PWJ
3803static inline void netif_set_gso_max_size(struct net_device *dev,
3804 unsigned int size)
3805{
3806 dev->gso_max_size = size;
3807}
3808
7a7ffbab
WCC
3809static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3810 int pulled_hlen, u16 mac_offset,
3811 int mac_len)
3812{
3813 skb->protocol = protocol;
3814 skb->encapsulation = 1;
3815 skb_push(skb, pulled_hlen);
3816 skb_reset_transport_header(skb);
3817 skb->mac_header = mac_offset;
3818 skb->network_header = skb->mac_header + mac_len;
3819 skb->mac_len = mac_len;
3820}
3821
a6cc0cfa
JF
3822static inline bool netif_is_macvlan(struct net_device *dev)
3823{
3824 return dev->priv_flags & IFF_MACVLAN;
3825}
3826
2f33e7d5
MB
3827static inline bool netif_is_macvlan_port(struct net_device *dev)
3828{
3829 return dev->priv_flags & IFF_MACVLAN_PORT;
3830}
3831
5933fea7
MB
3832static inline bool netif_is_ipvlan(struct net_device *dev)
3833{
3834 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3835}
3836
3837static inline bool netif_is_ipvlan_port(struct net_device *dev)
3838{
3839 return dev->priv_flags & IFF_IPVLAN_MASTER;
3840}
3841
8a7fbfab 3842static inline bool netif_is_bond_master(struct net_device *dev)
3843{
3844 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3845}
3846
4d29515f 3847static inline bool netif_is_bond_slave(struct net_device *dev)
1765a575
JP
3848{
3849 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3850}
3851
3bdc0eba
BG
3852static inline bool netif_supports_nofcs(struct net_device *dev)
3853{
3854 return dev->priv_flags & IFF_SUPP_NOFCS;
3855}
3856
007979ea 3857static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 3858{
007979ea 3859 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
3860}
3861
fee6d4c7
DA
3862static inline bool netif_is_l3_slave(const struct net_device *dev)
3863{
3864 return dev->priv_flags & IFF_L3MDEV_SLAVE;
3865}
3866
0894ae3f
JP
3867static inline bool netif_is_bridge_master(const struct net_device *dev)
3868{
3869 return dev->priv_flags & IFF_EBRIDGE;
3870}
3871
28f9ee22
VY
3872static inline bool netif_is_bridge_port(const struct net_device *dev)
3873{
3874 return dev->priv_flags & IFF_BRIDGE_PORT;
3875}
3876
35d4e172
JP
3877static inline bool netif_is_ovs_master(const struct net_device *dev)
3878{
3879 return dev->priv_flags & IFF_OPENVSWITCH;
3880}
3881
02875878
ED
3882/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3883static inline void netif_keep_dst(struct net_device *dev)
3884{
3885 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3886}
3887
505d4f73 3888extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 3889
571ba423
JP
3890/* Logging, debugging and troubleshooting/diagnostic helpers. */
3891
3892/* netdev_printk helpers, similar to dev_printk */
3893
3894static inline const char *netdev_name(const struct net_device *dev)
3895{
c6f854d5
VF
3896 if (!dev->name[0] || strchr(dev->name, '%'))
3897 return "(unnamed net_device)";
571ba423
JP
3898 return dev->name;
3899}
3900
ccc7f496
VF
3901static inline const char *netdev_reg_state(const struct net_device *dev)
3902{
3903 switch (dev->reg_state) {
3904 case NETREG_UNINITIALIZED: return " (uninitialized)";
3905 case NETREG_REGISTERED: return "";
3906 case NETREG_UNREGISTERING: return " (unregistering)";
3907 case NETREG_UNREGISTERED: return " (unregistered)";
3908 case NETREG_RELEASED: return " (released)";
3909 case NETREG_DUMMY: return " (dummy)";
3910 }
3911
3912 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3913 return " (unknown)";
3914}
3915
f629d208 3916__printf(3, 4)
6ea754eb
JP
3917void netdev_printk(const char *level, const struct net_device *dev,
3918 const char *format, ...);
f629d208 3919__printf(2, 3)
6ea754eb 3920void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 3921__printf(2, 3)
6ea754eb 3922void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 3923__printf(2, 3)
6ea754eb 3924void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 3925__printf(2, 3)
6ea754eb 3926void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 3927__printf(2, 3)
6ea754eb 3928void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 3929__printf(2, 3)
6ea754eb 3930void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 3931__printf(2, 3)
6ea754eb 3932void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 3933
8909c9ad
VK
3934#define MODULE_ALIAS_NETDEV(device) \
3935 MODULE_ALIAS("netdev-" device)
3936
b558c96f 3937#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
3938#define netdev_dbg(__dev, format, args...) \
3939do { \
ffa10cb4 3940 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 3941} while (0)
b558c96f
JC
3942#elif defined(DEBUG)
3943#define netdev_dbg(__dev, format, args...) \
3944 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
3945#else
3946#define netdev_dbg(__dev, format, args...) \
3947({ \
3948 if (0) \
3949 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
3950})
3951#endif
3952
3953#if defined(VERBOSE_DEBUG)
3954#define netdev_vdbg netdev_dbg
3955#else
3956
3957#define netdev_vdbg(dev, format, args...) \
3958({ \
3959 if (0) \
3960 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3961 0; \
3962})
3963#endif
3964
3965/*
3966 * netdev_WARN() acts like dev_printk(), but with the key difference
3967 * of using a WARN/WARN_ON to get the message out, including the
3968 * file/line information and a backtrace.
3969 */
3970#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
3971 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3972 netdev_reg_state(dev), ##args)
571ba423 3973
b3d95c5c
JP
3974/* netif printk helpers, similar to netdev_printk */
3975
3976#define netif_printk(priv, type, level, dev, fmt, args...) \
3977do { \
3978 if (netif_msg_##type(priv)) \
3979 netdev_printk(level, (dev), fmt, ##args); \
3980} while (0)
3981
f45f4321
JP
3982#define netif_level(level, priv, type, dev, fmt, args...) \
3983do { \
3984 if (netif_msg_##type(priv)) \
3985 netdev_##level(dev, fmt, ##args); \
3986} while (0)
3987
b3d95c5c 3988#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 3989 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 3990#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 3991 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 3992#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 3993 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 3994#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 3995 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 3996#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 3997 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 3998#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 3999 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 4000#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 4001 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 4002
0053ea9c 4003#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
4004#define netif_dbg(priv, type, netdev, format, args...) \
4005do { \
4006 if (netif_msg_##type(priv)) \
b5fb0a03 4007 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 4008} while (0)
0053ea9c
JP
4009#elif defined(DEBUG)
4010#define netif_dbg(priv, type, dev, format, args...) \
4011 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4012#else
4013#define netif_dbg(priv, type, dev, format, args...) \
4014({ \
4015 if (0) \
4016 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4017 0; \
4018})
4019#endif
4020
4021#if defined(VERBOSE_DEBUG)
bcfcc450 4022#define netif_vdbg netif_dbg
b3d95c5c
JP
4023#else
4024#define netif_vdbg(priv, type, dev, format, args...) \
4025({ \
4026 if (0) \
a4ed89cb 4027 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4028 0; \
4029})
4030#endif
571ba423 4031
900ff8c6
CW
4032/*
4033 * The list of packet types we will receive (as opposed to discard)
4034 * and the routines to invoke.
4035 *
4036 * Why 16. Because with 16 the only overlap we get on a hash of the
4037 * low nibble of the protocol value is RARP/SNAP/X.25.
4038 *
4039 * NOTE: That is no longer true with the addition of VLAN tags. Not
4040 * sure which should go first, but I bet it won't make much
4041 * difference if we are running VLANs. The good news is that
4042 * this protocol won't be in the list unless compiled in, so
4043 * the average user (w/out VLANs) will not be adversely affected.
4044 * --BLG
4045 *
4046 * 0800 IP
4047 * 8100 802.1Q VLAN
4048 * 0001 802.3
4049 * 0002 AX.25
4050 * 0004 802.2
4051 * 8035 RARP
4052 * 0005 SNAP
4053 * 0805 X.25
4054 * 0806 ARP
4055 * 8137 IPX
4056 * 0009 Localtalk
4057 * 86DD IPv6
4058 */
4059#define PTYPE_HASH_SIZE (16)
4060#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4061
385a154c 4062#endif /* _LINUX_NETDEVICE_H */