]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/netdevice.h
sctp: start t5 timer only when peer rwnd is 0 and local state is SHUTDOWN_PENDING
[mirror_ubuntu-bionic-kernel.git] / include / linux / netdevice.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
113aa838 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
d7fe0f24 28#include <linux/timer.h>
187f1882 29#include <linux/bug.h>
bea3348e 30#include <linux/delay.h>
60063497 31#include <linux/atomic.h>
53511453 32#include <linux/prefetch.h>
1da177e4
LT
33#include <asm/cache.h>
34#include <asm/byteorder.h>
35
1da177e4 36#include <linux/percpu.h>
4d5b78c0 37#include <linux/rculist.h>
db217334 38#include <linux/dmaengine.h>
bea3348e 39#include <linux/workqueue.h>
114cf580 40#include <linux/dynamic_queue_limits.h>
1da177e4 41
b1b67dd4 42#include <linux/ethtool.h>
a050c33f 43#include <net/net_namespace.h>
cf85d08f 44#include <net/dsa.h>
7a6b6f51 45#ifdef CONFIG_DCB
2f90b865
AD
46#include <net/dcbnl.h>
47#endif
5bc1421e 48#include <net/netprio_cgroup.h>
a050c33f 49
a59e2ecb 50#include <linux/netdev_features.h>
77162022 51#include <linux/neighbour.h>
607ca46e 52#include <uapi/linux/netdevice.h>
61bd3857 53#include <uapi/linux/if_bonding.h>
a59e2ecb 54
115c1d6e 55struct netpoll_info;
313162d0 56struct device;
c1f19b51 57struct phy_device;
704232c2
JB
58/* 802.11 specific */
59struct wireless_dev;
98a18b6f
AA
60/* 802.15.4 specific */
61struct wpan_dev;
03c57747 62struct mpls_dev;
1da177e4 63
f629d208
JP
64void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
d07d7507 66
9a1654ba
JP
67/* Backlog congestion levels */
68#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69#define NET_RX_DROP 1 /* packet dropped */
70
572a9d7b
PM
71/*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88/* qdisc ->enqueue() return codes. */
89#define NET_XMIT_SUCCESS 0x00
9a1654ba
JP
90#define NET_XMIT_DROP 0x01 /* skb dropped */
91#define NET_XMIT_CN 0x02 /* congestion notification */
92#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
1da177e4 94
b9df3cb8
GR
95/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
572a9d7b 98#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
1da177e4
LT
99#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
dc1f8bf6 101/* Driver transmit return codes */
9a1654ba 102#define NETDEV_TX_MASK 0xf0
572a9d7b 103
dc1f8bf6 104enum netdev_tx {
572a9d7b 105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
9a1654ba
JP
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
dc1f8bf6
SH
109};
110typedef enum netdev_tx netdev_tx_t;
111
9a1654ba
JP
112/*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116static inline bool dev_xmit_complete(int rc)
117{
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128}
129
1da177e4
LT
130/*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
fe2918b0 134
d11ead75 135#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
8388e3da
DM
136# if defined(CONFIG_MAC80211_MESH)
137# define LL_MAX_HEADER 128
138# else
139# define LL_MAX_HEADER 96
140# endif
1da177e4 141#else
8388e3da 142# define LL_MAX_HEADER 32
1da177e4
LT
143#endif
144
d11ead75
BH
145#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
1da177e4
LT
147#define MAX_HEADER LL_MAX_HEADER
148#else
149#define MAX_HEADER (LL_MAX_HEADER + 48)
150#endif
151
152/*
be1f3c2c
BH
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
1da177e4 155 */
fe2918b0 156
d94d9fee 157struct net_device_stats {
3cfde79c
BH
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
1da177e4 167 unsigned long collisions;
1da177e4 168 unsigned long rx_length_errors;
3cfde79c
BH
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
1da177e4
LT
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
1da177e4
LT
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
181};
182
1da177e4
LT
183
184#include <linux/cache.h>
185#include <linux/skbuff.h>
186
adc9300e 187#ifdef CONFIG_RPS
c5905afb
IM
188#include <linux/static_key.h>
189extern struct static_key rps_needed;
adc9300e
ED
190#endif
191
1da177e4
LT
192struct neighbour;
193struct neigh_parms;
194struct sk_buff;
195
f001fde5
JP
196struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
199 unsigned char type;
ccffad25
JP
200#define NETDEV_HW_ADDR_T_LAN 1
201#define NETDEV_HW_ADDR_T_SAN 2
202#define NETDEV_HW_ADDR_T_SLAVE 3
203#define NETDEV_HW_ADDR_T_UNICAST 4
22bedad3 204#define NETDEV_HW_ADDR_T_MULTICAST 5
22bedad3 205 bool global_use;
4cd729b0 206 int sync_cnt;
8f8f103d 207 int refcount;
4543fbef 208 int synced;
f001fde5
JP
209 struct rcu_head rcu_head;
210};
211
31278e71
JP
212struct netdev_hw_addr_list {
213 struct list_head list;
214 int count;
215};
216
22bedad3
JP
217#define netdev_hw_addr_list_count(l) ((l)->count)
218#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219#define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
32e7bfc4 221
22bedad3
JP
222#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224#define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
6683ece3 226
22bedad3
JP
227#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
18e225f2 229#define netdev_for_each_mc_addr(ha, dev) \
22bedad3 230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
6683ece3 231
d94d9fee 232struct hh_cache {
f6b72b62 233 u16 hh_len;
5c25f686 234 u16 __pad;
3644f0ce 235 seqlock_t hh_lock;
1da177e4
LT
236
237 /* cached hardware header; allow for machine alignment needs. */
238#define HH_DATA_MOD 16
239#define HH_DATA_OFF(__len) \
5ba0eac6 240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
1da177e4
LT
241#define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
244};
245
246/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * Alternative is:
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 *
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
253 */
254#define LL_RESERVED_SPACE(dev) \
f5184d26 255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 256#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
f5184d26 257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
1da177e4 258
3b04ddde
SH
259struct header_ops {
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
95c96174 262 const void *saddr, unsigned int len);
3b04ddde 263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
e69dd336 264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
3b04ddde
SH
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268};
269
1da177e4
LT
270/* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
272 * code.
273 */
274
d94d9fee 275enum netdev_state_t {
1da177e4
LT
276 __LINK_STATE_START,
277 __LINK_STATE_PRESENT,
1da177e4 278 __LINK_STATE_NOCARRIER,
b00055aa
SR
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
1da177e4
LT
281};
282
283
284/*
285 * This structure holds at boot time configured netdevice settings. They
fe2918b0 286 * are then used in the device probing.
1da177e4
LT
287 */
288struct netdev_boot_setup {
289 char name[IFNAMSIZ];
290 struct ifmap map;
291};
292#define NETDEV_BOOT_SETUP_MAX 8
293
f629d208 294int __init netdev_boot_setup(char *str);
1da177e4 295
bea3348e
SH
296/*
297 * Structure for NAPI scheduling similar to tasklet but with weighting
298 */
299struct napi_struct {
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
305 */
306 struct list_head poll_list;
307
308 unsigned long state;
309 int weight;
404f7c9e 310 unsigned int gro_count;
bea3348e
SH
311 int (*poll)(struct napi_struct *, int);
312#ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
314 int poll_owner;
bea3348e 315#endif
5d38a079 316 struct net_device *dev;
d565b0a1 317 struct sk_buff *gro_list;
5d38a079 318 struct sk_buff *skb;
3b47d303 319 struct hrtimer timer;
404f7c9e 320 struct list_head dev_list;
af12fa6e
ET
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
bea3348e
SH
323};
324
d94d9fee 325enum {
bea3348e 326 NAPI_STATE_SCHED, /* Poll is scheduled */
a0a46196 327 NAPI_STATE_DISABLE, /* Disable pending */
7b363e44 328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
af12fa6e 329 NAPI_STATE_HASHED, /* In NAPI hash */
bea3348e
SH
330};
331
5b252f0c 332enum gro_result {
d1c76af9
HX
333 GRO_MERGED,
334 GRO_MERGED_FREE,
335 GRO_HELD,
336 GRO_NORMAL,
337 GRO_DROP,
338};
5b252f0c 339typedef enum gro_result gro_result_t;
d1c76af9 340
8a4eb573
JP
341/*
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * further.
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 *
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
352 *
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
355 *
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
359 *
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
361 * do with the skb.
362 *
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
366 *
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
370 *
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
d93cf068 373 * are registered on exact device (ptype->dev == skb->dev).
8a4eb573
JP
374 *
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
377 *
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
380 */
381
382enum rx_handler_result {
383 RX_HANDLER_CONSUMED,
384 RX_HANDLER_ANOTHER,
385 RX_HANDLER_EXACT,
386 RX_HANDLER_PASS,
387};
388typedef enum rx_handler_result rx_handler_result_t;
389typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
ab95bfe0 390
f629d208 391void __napi_schedule(struct napi_struct *n);
bc9ad166 392void __napi_schedule_irqoff(struct napi_struct *n);
bea3348e 393
4d29515f 394static inline bool napi_disable_pending(struct napi_struct *n)
a0a46196
DM
395{
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
397}
398
bea3348e
SH
399/**
400 * napi_schedule_prep - check if napi can be scheduled
401 * @n: napi context
402 *
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
a0a46196
DM
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
bea3348e 407 */
4d29515f 408static inline bool napi_schedule_prep(struct napi_struct *n)
bea3348e 409{
a0a46196
DM
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
bea3348e
SH
412}
413
414/**
415 * napi_schedule - schedule NAPI poll
416 * @n: napi context
417 *
418 * Schedule NAPI poll routine to be called if it is not already
419 * running.
420 */
421static inline void napi_schedule(struct napi_struct *n)
422{
423 if (napi_schedule_prep(n))
424 __napi_schedule(n);
425}
426
bc9ad166
ED
427/**
428 * napi_schedule_irqoff - schedule NAPI poll
429 * @n: napi context
430 *
431 * Variant of napi_schedule(), assuming hard irqs are masked.
432 */
433static inline void napi_schedule_irqoff(struct napi_struct *n)
434{
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
437}
438
bfe13f54 439/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
4d29515f 440static inline bool napi_reschedule(struct napi_struct *napi)
bfe13f54
RD
441{
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
4d29515f 444 return true;
bfe13f54 445 }
4d29515f 446 return false;
bfe13f54
RD
447}
448
3b47d303
ED
449void __napi_complete(struct napi_struct *n);
450void napi_complete_done(struct napi_struct *n, int work_done);
bea3348e
SH
451/**
452 * napi_complete - NAPI processing complete
453 * @n: napi context
454 *
455 * Mark NAPI processing as complete.
3b47d303 456 * Consider using napi_complete_done() instead.
bea3348e 457 */
3b47d303
ED
458static inline void napi_complete(struct napi_struct *n)
459{
460 return napi_complete_done(n, 0);
461}
bea3348e 462
af12fa6e
ET
463/**
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
466 *
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
469 */
f629d208 470struct napi_struct *napi_by_id(unsigned int napi_id);
af12fa6e
ET
471
472/**
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
475 *
476 * generate a new napi_id and store a @napi under it in napi_hash
477 */
f629d208 478void napi_hash_add(struct napi_struct *napi);
af12fa6e
ET
479
480/**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
483 *
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
486 */
f629d208 487void napi_hash_del(struct napi_struct *napi);
af12fa6e 488
bea3348e
SH
489/**
490 * napi_disable - prevent NAPI from scheduling
491 * @n: napi context
492 *
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
495 */
3b47d303 496void napi_disable(struct napi_struct *n);
bea3348e
SH
497
498/**
499 * napi_enable - enable NAPI scheduling
500 * @n: napi context
501 *
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
504 */
505static inline void napi_enable(struct napi_struct *n)
506{
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
4e857c58 508 smp_mb__before_atomic();
bea3348e 509 clear_bit(NAPI_STATE_SCHED, &n->state);
2d8bff12 510 clear_bit(NAPI_STATE_NPSVC, &n->state);
bea3348e
SH
511}
512
c264c3de
SH
513#ifdef CONFIG_SMP
514/**
515 * napi_synchronize - wait until NAPI is not running
516 * @n: napi context
517 *
518 * Wait until NAPI is done being scheduled on this context.
519 * Waits till any outstanding processing completes but
520 * does not disable future activations.
521 */
522static inline void napi_synchronize(const struct napi_struct *n)
523{
524 while (test_bit(NAPI_STATE_SCHED, &n->state))
525 msleep(1);
526}
527#else
528# define napi_synchronize(n) barrier()
529#endif
530
d94d9fee 531enum netdev_queue_state_t {
73466498
TH
532 __QUEUE_STATE_DRV_XOFF,
533 __QUEUE_STATE_STACK_XOFF,
c3f26a26 534 __QUEUE_STATE_FROZEN,
79d16385 535};
8e2f1a63
DB
536
537#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
538#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
539#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
540
541#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
542#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
543 QUEUE_STATE_FROZEN)
544#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
545 QUEUE_STATE_FROZEN)
546
73466498
TH
547/*
548 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
549 * netif_tx_* functions below are used to manipulate this flag. The
550 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
551 * queue independently. The netif_xmit_*stopped functions below are called
552 * to check if the queue has been stopped by the driver or stack (either
553 * of the XOFF bits are set in the state). Drivers should not need to call
554 * netif_xmit*stopped functions, they should only be using netif_tx_*.
555 */
79d16385 556
bb949fbd 557struct netdev_queue {
6a321cb3
ED
558/*
559 * read mostly part
560 */
bb949fbd 561 struct net_device *dev;
46e5da40 562 struct Qdisc __rcu *qdisc;
b0e1e646 563 struct Qdisc *qdisc_sleeping;
ccf5ff69 564#ifdef CONFIG_SYSFS
1d24eb48
TH
565 struct kobject kobj;
566#endif
f2cd2d3e
ED
567#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
568 int numa_node;
569#endif
6a321cb3
ED
570/*
571 * write mostly part
572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
9d21493b
ED
575 /*
576 * please use this field instead of dev->trans_start
577 */
578 unsigned long trans_start;
ccf5ff69 579
580 /*
581 * Number of TX timeouts for this queue
582 * (/sys/class/net/DEV/Q/trans_timeout)
583 */
584 unsigned long trans_timeout;
114cf580
TH
585
586 unsigned long state;
587
588#ifdef CONFIG_BQL
589 struct dql dql;
590#endif
822b3b2e 591 unsigned long tx_maxrate;
e8a0464c 592} ____cacheline_aligned_in_smp;
bb949fbd 593
f2cd2d3e
ED
594static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
595{
596#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
597 return q->numa_node;
598#else
b236da69 599 return NUMA_NO_NODE;
f2cd2d3e
ED
600#endif
601}
602
603static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
604{
605#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 q->numa_node = node;
607#endif
608}
609
df334545 610#ifdef CONFIG_RPS
0a9627f2
TH
611/*
612 * This structure holds an RPS map which can be of variable length. The
613 * map is an array of CPUs.
614 */
615struct rps_map {
616 unsigned int len;
617 struct rcu_head rcu;
618 u16 cpus[0];
619};
60b778ce 620#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
0a9627f2 621
fec5e652 622/*
c445477d
BH
623 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
624 * tail pointer for that CPU's input queue at the time of last enqueue, and
625 * a hardware filter index.
fec5e652
TH
626 */
627struct rps_dev_flow {
628 u16 cpu;
c445477d 629 u16 filter;
fec5e652
TH
630 unsigned int last_qtail;
631};
c445477d 632#define RPS_NO_FILTER 0xffff
fec5e652
TH
633
634/*
635 * The rps_dev_flow_table structure contains a table of flow mappings.
636 */
637struct rps_dev_flow_table {
638 unsigned int mask;
639 struct rcu_head rcu;
fec5e652
TH
640 struct rps_dev_flow flows[0];
641};
642#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
60b778ce 643 ((_num) * sizeof(struct rps_dev_flow)))
fec5e652
TH
644
645/*
646 * The rps_sock_flow_table contains mappings of flows to the last CPU
647 * on which they were processed by the application (set in recvmsg).
567e4b79
ED
648 * Each entry is a 32bit value. Upper part is the high order bits
649 * of flow hash, lower part is cpu number.
650 * rps_cpu_mask is used to partition the space, depending on number of
651 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
652 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
653 * meaning we use 32-6=26 bits for the hash.
fec5e652
TH
654 */
655struct rps_sock_flow_table {
567e4b79 656 u32 mask;
93c1af6c
ED
657
658 u32 ents[0] ____cacheline_aligned_in_smp;
fec5e652 659};
567e4b79 660#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
fec5e652
TH
661
662#define RPS_NO_CPU 0xffff
663
567e4b79
ED
664extern u32 rps_cpu_mask;
665extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
666
fec5e652
TH
667static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
668 u32 hash)
669{
670 if (table && hash) {
567e4b79
ED
671 unsigned int index = hash & table->mask;
672 u32 val = hash & ~rps_cpu_mask;
fec5e652
TH
673
674 /* We only give a hint, preemption can change cpu under us */
567e4b79 675 val |= raw_smp_processor_id();
fec5e652 676
567e4b79
ED
677 if (table->ents[index] != val)
678 table->ents[index] = val;
fec5e652
TH
679 }
680}
681
c445477d 682#ifdef CONFIG_RFS_ACCEL
f629d208
JP
683bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
684 u16 filter_id);
c445477d 685#endif
a953be53 686#endif /* CONFIG_RPS */
c445477d 687
0a9627f2
TH
688/* This structure contains an instance of an RX queue. */
689struct netdev_rx_queue {
a953be53 690#ifdef CONFIG_RPS
6e3f7faf
ED
691 struct rps_map __rcu *rps_map;
692 struct rps_dev_flow_table __rcu *rps_flow_table;
a953be53 693#endif
6e3f7faf 694 struct kobject kobj;
fe822240 695 struct net_device *dev;
0a9627f2 696} ____cacheline_aligned_in_smp;
a953be53
MD
697
698/*
699 * RX queue sysfs structures and functions.
700 */
701struct rx_queue_attribute {
702 struct attribute attr;
703 ssize_t (*show)(struct netdev_rx_queue *queue,
704 struct rx_queue_attribute *attr, char *buf);
705 ssize_t (*store)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, const char *buf, size_t len);
707};
d314774c 708
bf264145
TH
709#ifdef CONFIG_XPS
710/*
711 * This structure holds an XPS map which can be of variable length. The
712 * map is an array of queues.
713 */
714struct xps_map {
715 unsigned int len;
716 unsigned int alloc_len;
717 struct rcu_head rcu;
718 u16 queues[0];
719};
60b778ce 720#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
c59f419b
HD
721#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
722 - sizeof(struct xps_map)) / sizeof(u16))
bf264145
TH
723
724/*
725 * This structure holds all XPS maps for device. Maps are indexed by CPU.
726 */
727struct xps_dev_maps {
728 struct rcu_head rcu;
a4177869 729 struct xps_map __rcu *cpu_map[0];
bf264145
TH
730};
731#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
732 (nr_cpu_ids * sizeof(struct xps_map *)))
733#endif /* CONFIG_XPS */
734
4f57c087
JF
735#define TC_MAX_QUEUE 16
736#define TC_BITMASK 15
737/* HW offloaded queuing disciplines txq count and offset maps */
738struct netdev_tc_txq {
739 u16 count;
740 u16 offset;
741};
742
68bad94e
NP
743#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
744/*
745 * This structure is to hold information about the device
746 * configured to run FCoE protocol stack.
747 */
748struct netdev_fcoe_hbainfo {
749 char manufacturer[64];
750 char serial_number[64];
751 char hardware_version[64];
752 char driver_version[64];
753 char optionrom_version[64];
754 char firmware_version[64];
755 char model[256];
756 char model_description[256];
757};
758#endif
759
02637fce 760#define MAX_PHYS_ITEM_ID_LEN 32
66b52b0d 761
02637fce
JP
762/* This structure holds a unique identifier to identify some
763 * physical item (port for example) used by a netdevice.
66b52b0d 764 */
02637fce
JP
765struct netdev_phys_item_id {
766 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
66b52b0d
JP
767 unsigned char id_len;
768};
769
d754f98b
SF
770static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
771 struct netdev_phys_item_id *b)
772{
773 return a->id_len == b->id_len &&
774 memcmp(a->id, b->id, a->id_len) == 0;
775}
776
99932d4f
DB
777typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
778 struct sk_buff *skb);
779
d314774c
SH
780/*
781 * This structure defines the management hooks for network devices.
00829823
SH
782 * The following hooks can be defined; unless noted otherwise, they are
783 * optional and can be filled with a null pointer.
d314774c
SH
784 *
785 * int (*ndo_init)(struct net_device *dev);
786 * This function is called once when network device is registered.
787 * The network device can use this to any late stage initializaton
788 * or semantic validattion. It can fail with an error code which will
789 * be propogated back to register_netdev
790 *
791 * void (*ndo_uninit)(struct net_device *dev);
792 * This function is called when device is unregistered or when registration
793 * fails. It is not called if init fails.
794 *
795 * int (*ndo_open)(struct net_device *dev);
796 * This function is called when network device transistions to the up
797 * state.
798 *
799 * int (*ndo_stop)(struct net_device *dev);
800 * This function is called when network device transistions to the down
801 * state.
802 *
dc1f8bf6
SH
803 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
804 * struct net_device *dev);
00829823 805 * Called when a packet needs to be transmitted.
e79d8429
RR
806 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
807 * the queue before that can happen; it's for obsolete devices and weird
808 * corner cases, but the stack really does a non-trivial amount
809 * of useless work if you return NETDEV_TX_BUSY.
dc1f8bf6 810 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
00829823
SH
811 * Required can not be NULL.
812 *
f663dd9a 813 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
99932d4f 814 * void *accel_priv, select_queue_fallback_t fallback);
00829823
SH
815 * Called to decide which queue to when device supports multiple
816 * transmit queues.
817 *
d314774c
SH
818 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
819 * This function is called to allow device receiver to make
820 * changes to configuration when multicast or promiscious is enabled.
821 *
822 * void (*ndo_set_rx_mode)(struct net_device *dev);
823 * This function is called device changes address list filtering.
01789349
JP
824 * If driver handles unicast address filtering, it should set
825 * IFF_UNICAST_FLT to its priv_flags.
d314774c
SH
826 *
827 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
828 * This function is called when the Media Access Control address
37b607c5 829 * needs to be changed. If this interface is not defined, the
d314774c
SH
830 * mac address can not be changed.
831 *
832 * int (*ndo_validate_addr)(struct net_device *dev);
833 * Test if Media Access Control address is valid for the device.
834 *
835 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
836 * Called when a user request an ioctl which can't be handled by
837 * the generic interface code. If not defined ioctl's return
838 * not supported error code.
839 *
840 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
841 * Used to set network devices bus interface parameters. This interface
842 * is retained for legacy reason, new devices should use the bus
843 * interface (PCI) for low level management.
844 *
845 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
846 * Called when a user wants to change the Maximum Transfer Unit
847 * of a device. If not defined, any request to change MTU will
848 * will return an error.
849 *
00829823 850 * void (*ndo_tx_timeout)(struct net_device *dev);
d314774c
SH
851 * Callback uses when the transmitter has not made any progress
852 * for dev->watchdog ticks.
853 *
3cfde79c 854 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
28172739 855 * struct rtnl_link_stats64 *storage);
d308e38f 856 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
d314774c 857 * Called when a user wants to get the network device usage
be1f3c2c 858 * statistics. Drivers must do one of the following:
3cfde79c
BH
859 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
860 * rtnl_link_stats64 structure passed by the caller.
82695d9b 861 * 2. Define @ndo_get_stats to update a net_device_stats structure
be1f3c2c
BH
862 * (which should normally be dev->stats) and return a pointer to
863 * it. The structure may be changed asynchronously only if each
864 * field is written atomically.
865 * 3. Update dev->stats asynchronously and atomically, and define
866 * neither operation.
d314774c 867 *
5d632cb7 868 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
869 * If device support VLAN filtering this function is called when a
870 * VLAN id is registered.
d314774c 871 *
5d632cb7 872 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
80d5c368
PM
873 * If device support VLAN filtering this function is called when a
874 * VLAN id is unregistered.
d314774c
SH
875 *
876 * void (*ndo_poll_controller)(struct net_device *dev);
95c26df8
WM
877 *
878 * SR-IOV management functions.
879 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
880 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
ed616689
SC
881 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
882 * int max_tx_rate);
5f8444a3 883 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
dd461d6a 884 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
95c26df8
WM
885 * int (*ndo_get_vf_config)(struct net_device *dev,
886 * int vf, struct ifla_vf_info *ivf);
1d8faf48 887 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
57b61080
SF
888 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
889 * struct nlattr *port[]);
01a3d796
VZ
890 *
891 * Enable or disable the VF ability to query its RSS Redirection Table and
892 * Hash Key. This is needed since on some devices VF share this information
893 * with PF and querying it may adduce a theoretical security risk.
894 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
57b61080 895 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
4f57c087
JF
896 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
897 * Called to setup 'tc' number of traffic classes in the net device. This
898 * is always called from the stack with the rtnl lock held and netif tx
899 * queues stopped. This allows the netdevice to perform queue management
900 * safely.
c445477d 901 *
e9bce845
YZ
902 * Fiber Channel over Ethernet (FCoE) offload functions.
903 * int (*ndo_fcoe_enable)(struct net_device *dev);
904 * Called when the FCoE protocol stack wants to start using LLD for FCoE
905 * so the underlying device can perform whatever needed configuration or
906 * initialization to support acceleration of FCoE traffic.
907 *
908 * int (*ndo_fcoe_disable)(struct net_device *dev);
909 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
910 * so the underlying device can perform whatever needed clean-ups to
911 * stop supporting acceleration of FCoE traffic.
912 *
913 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
914 * struct scatterlist *sgl, unsigned int sgc);
915 * Called when the FCoE Initiator wants to initialize an I/O that
916 * is a possible candidate for Direct Data Placement (DDP). The LLD can
917 * perform necessary setup and returns 1 to indicate the device is set up
918 * successfully to perform DDP on this I/O, otherwise this returns 0.
919 *
920 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
921 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
922 * indicated by the FC exchange id 'xid', so the underlying device can
923 * clean up and reuse resources for later DDP requests.
924 *
925 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
926 * struct scatterlist *sgl, unsigned int sgc);
927 * Called when the FCoE Target wants to initialize an I/O that
928 * is a possible candidate for Direct Data Placement (DDP). The LLD can
929 * perform necessary setup and returns 1 to indicate the device is set up
930 * successfully to perform DDP on this I/O, otherwise this returns 0.
931 *
68bad94e
NP
932 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
933 * struct netdev_fcoe_hbainfo *hbainfo);
934 * Called when the FCoE Protocol stack wants information on the underlying
935 * device. This information is utilized by the FCoE protocol stack to
936 * register attributes with Fiber Channel management service as per the
937 * FC-GS Fabric Device Management Information(FDMI) specification.
938 *
e9bce845
YZ
939 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
940 * Called when the underlying device wants to override default World Wide
941 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
942 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
943 * protocol stack to use.
944 *
c445477d
BH
945 * RFS acceleration.
946 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
947 * u16 rxq_index, u32 flow_id);
948 * Set hardware filter for RFS. rxq_index is the target queue index;
949 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
950 * Return the filter ID on success, or a negative error code.
fbaec0ea 951 *
8b98a70c 952 * Slave management functions (for bridge, bonding, etc).
fbaec0ea
JP
953 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
954 * Called to make another netdev an underling.
955 *
956 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
957 * Called to release previously enslaved netdev.
5455c699
MM
958 *
959 * Feature/offload setting functions.
c8f44aff
MM
960 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
961 * netdev_features_t features);
5455c699
MM
962 * Adjusts the requested feature flags according to device-specific
963 * constraints, and returns the resulting flags. Must not modify
964 * the device state.
965 *
c8f44aff 966 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
5455c699
MM
967 * Called to update device configuration to new features. Passed
968 * feature set might be less than what was returned by ndo_fix_features()).
969 * Must return >0 or -errno if it changed dev->features itself.
970 *
edc7d573 971 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
972 * struct net_device *dev,
f6f6424b 973 * const unsigned char *addr, u16 vid, u16 flags)
77162022 974 * Adds an FDB entry to dev for addr.
1690be63
VY
975 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
976 * struct net_device *dev,
f6f6424b 977 * const unsigned char *addr, u16 vid)
77162022
JF
978 * Deletes the FDB entry from dev coresponding to addr.
979 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS
980 * struct net_device *dev, struct net_device *filter_dev,
981 * int idx)
77162022
JF
982 * Used to add FDB entries to dump requests. Implementers should add
983 * entries to skb and update idx with the number of entries.
e5a55a89 984 *
ad41faa8
ND
985 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
986 * u16 flags)
e5a55a89 987 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
46c264da
ND
988 * struct net_device *dev, u32 filter_mask,
989 * int nlflags)
ad41faa8
ND
990 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
991 * u16 flags);
4bf84c35
JP
992 *
993 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
994 * Called to change device carrier. Soft-devices (like dummy, team, etc)
995 * which do not represent real hardware may define this to allow their
996 * userspace components to manage their virtual carrier state. Devices
997 * that determine carrier state from physical hardware properties (eg
998 * network cables) or protocol-dependent mechanisms (eg
999 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
66b52b0d
JP
1000 *
1001 * int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1002 * struct netdev_phys_item_id *ppid);
66b52b0d
JP
1003 * Called to get ID of physical port of this device. If driver does
1004 * not implement this, it is assumed that the hw is not able to have
1005 * multiple net devices on single physical port.
53cf5275
JG
1006 *
1007 * void (*ndo_add_vxlan_port)(struct net_device *dev,
35e42379 1008 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1009 * Called by vxlan to notiy a driver about the UDP port and socket
1010 * address family that vxlan is listnening to. It is called only when
1011 * a new port starts listening. The operation is protected by the
1012 * vxlan_net->sock_lock.
1013 *
1014 * void (*ndo_del_vxlan_port)(struct net_device *dev,
35e42379 1015 * sa_family_t sa_family, __be16 port);
53cf5275
JG
1016 * Called by vxlan to notify the driver about a UDP port and socket
1017 * address family that vxlan is not listening to anymore. The operation
1018 * is protected by the vxlan_net->sock_lock.
a6cc0cfa
JF
1019 *
1020 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1021 * struct net_device *dev)
1022 * Called by upper layer devices to accelerate switching or other
1023 * station functionality into hardware. 'pdev is the lowerdev
1024 * to use for the offload and 'dev' is the net device that will
1025 * back the offload. Returns a pointer to the private structure
1026 * the upper layer will maintain.
1027 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1028 * Called by upper layer device to delete the station created
1029 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1030 * the station and priv is the structure returned by the add
1031 * operation.
1032 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1033 * struct net_device *dev,
1034 * void *priv);
1035 * Callback to use for xmit over the accelerated station. This
1036 * is used in place of ndo_start_xmit on accelerated net
1037 * devices.
5f35227e
JG
1038 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1039 * struct net_device *dev
1040 * netdev_features_t features);
04ffcb25 1041 * Called by core transmit path to determine if device is capable of
5f35227e
JG
1042 * performing offload operations on a given packet. This is to give
1043 * the device an opportunity to implement any restrictions that cannot
1044 * be otherwise expressed by feature flags. The check is called with
1045 * the set of features that the stack has calculated and it returns
1046 * those the driver believes to be appropriate.
822b3b2e
JF
1047 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1048 * int queue_index, u32 maxrate);
1049 * Called when a user wants to set a max-rate limitation of specific
1050 * TX queue.
a54acb3a
ND
1051 * int (*ndo_get_iflink)(const struct net_device *dev);
1052 * Called to get the iflink value of this device.
d746d707
AK
1053 * void (*ndo_change_proto_down)(struct net_device *dev,
1054 * bool proto_down);
1055 * This function is used to pass protocol port error state information
1056 * to the switch driver. The switch driver can react to the proto_down
1057 * by doing a phys down on the associated switch port.
fc4099f1
PS
1058 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1059 * This function is used to get egress tunnel information for given skb.
1060 * This is useful for retrieving outer tunnel header parameters while
1061 * sampling packet.
d746d707 1062 *
d314774c
SH
1063 */
1064struct net_device_ops {
1065 int (*ndo_init)(struct net_device *dev);
1066 void (*ndo_uninit)(struct net_device *dev);
1067 int (*ndo_open)(struct net_device *dev);
1068 int (*ndo_stop)(struct net_device *dev);
dc1f8bf6 1069 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
00829823
SH
1070 struct net_device *dev);
1071 u16 (*ndo_select_queue)(struct net_device *dev,
f663dd9a 1072 struct sk_buff *skb,
99932d4f
DB
1073 void *accel_priv,
1074 select_queue_fallback_t fallback);
d314774c
SH
1075 void (*ndo_change_rx_flags)(struct net_device *dev,
1076 int flags);
d314774c 1077 void (*ndo_set_rx_mode)(struct net_device *dev);
d314774c
SH
1078 int (*ndo_set_mac_address)(struct net_device *dev,
1079 void *addr);
d314774c 1080 int (*ndo_validate_addr)(struct net_device *dev);
d314774c
SH
1081 int (*ndo_do_ioctl)(struct net_device *dev,
1082 struct ifreq *ifr, int cmd);
d314774c
SH
1083 int (*ndo_set_config)(struct net_device *dev,
1084 struct ifmap *map);
00829823
SH
1085 int (*ndo_change_mtu)(struct net_device *dev,
1086 int new_mtu);
1087 int (*ndo_neigh_setup)(struct net_device *dev,
1088 struct neigh_parms *);
d314774c
SH
1089 void (*ndo_tx_timeout) (struct net_device *dev);
1090
28172739
ED
1091 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1092 struct rtnl_link_stats64 *storage);
d314774c
SH
1093 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1094
8e586137 1095 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
80d5c368 1096 __be16 proto, u16 vid);
8e586137 1097 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
80d5c368 1098 __be16 proto, u16 vid);
d314774c 1099#ifdef CONFIG_NET_POLL_CONTROLLER
d314774c 1100 void (*ndo_poll_controller)(struct net_device *dev);
4247e161 1101 int (*ndo_netpoll_setup)(struct net_device *dev,
a8779ec1 1102 struct netpoll_info *info);
0e34e931 1103 void (*ndo_netpoll_cleanup)(struct net_device *dev);
06021292 1104#endif
e0d1095a 1105#ifdef CONFIG_NET_RX_BUSY_POLL
8b80cda5 1106 int (*ndo_busy_poll)(struct napi_struct *dev);
d314774c 1107#endif
95c26df8
WM
1108 int (*ndo_set_vf_mac)(struct net_device *dev,
1109 int queue, u8 *mac);
1110 int (*ndo_set_vf_vlan)(struct net_device *dev,
1111 int queue, u16 vlan, u8 qos);
ed616689
SC
1112 int (*ndo_set_vf_rate)(struct net_device *dev,
1113 int vf, int min_tx_rate,
1114 int max_tx_rate);
5f8444a3
GR
1115 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1116 int vf, bool setting);
dd461d6a
HS
1117 int (*ndo_set_vf_trust)(struct net_device *dev,
1118 int vf, bool setting);
95c26df8
WM
1119 int (*ndo_get_vf_config)(struct net_device *dev,
1120 int vf,
1121 struct ifla_vf_info *ivf);
1d8faf48
RE
1122 int (*ndo_set_vf_link_state)(struct net_device *dev,
1123 int vf, int link_state);
3b766cd8
EBE
1124 int (*ndo_get_vf_stats)(struct net_device *dev,
1125 int vf,
1126 struct ifla_vf_stats
1127 *vf_stats);
57b61080
SF
1128 int (*ndo_set_vf_port)(struct net_device *dev,
1129 int vf,
1130 struct nlattr *port[]);
1131 int (*ndo_get_vf_port)(struct net_device *dev,
1132 int vf, struct sk_buff *skb);
01a3d796
VZ
1133 int (*ndo_set_vf_rss_query_en)(
1134 struct net_device *dev,
1135 int vf, bool setting);
4f57c087 1136 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
d11ead75 1137#if IS_ENABLED(CONFIG_FCOE)
cb454399
YZ
1138 int (*ndo_fcoe_enable)(struct net_device *dev);
1139 int (*ndo_fcoe_disable)(struct net_device *dev);
4d288d57
YZ
1140 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1141 u16 xid,
1142 struct scatterlist *sgl,
1143 unsigned int sgc);
1144 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1145 u16 xid);
6247e086
YZ
1146 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1147 u16 xid,
1148 struct scatterlist *sgl,
1149 unsigned int sgc);
68bad94e
NP
1150 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1151 struct netdev_fcoe_hbainfo *hbainfo);
3c9c36bc
BPG
1152#endif
1153
d11ead75 1154#if IS_ENABLED(CONFIG_LIBFCOE)
df5c7945
YZ
1155#define NETDEV_FCOE_WWNN 0
1156#define NETDEV_FCOE_WWPN 1
1157 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1158 u64 *wwn, int type);
4d288d57 1159#endif
3c9c36bc 1160
c445477d
BH
1161#ifdef CONFIG_RFS_ACCEL
1162 int (*ndo_rx_flow_steer)(struct net_device *dev,
1163 const struct sk_buff *skb,
1164 u16 rxq_index,
1165 u32 flow_id);
1166#endif
fbaec0ea
JP
1167 int (*ndo_add_slave)(struct net_device *dev,
1168 struct net_device *slave_dev);
1169 int (*ndo_del_slave)(struct net_device *dev,
1170 struct net_device *slave_dev);
c8f44aff
MM
1171 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1172 netdev_features_t features);
5455c699 1173 int (*ndo_set_features)(struct net_device *dev,
c8f44aff 1174 netdev_features_t features);
da6a8fa0 1175 int (*ndo_neigh_construct)(struct neighbour *n);
447f2191 1176 void (*ndo_neigh_destroy)(struct neighbour *n);
77162022
JF
1177
1178 int (*ndo_fdb_add)(struct ndmsg *ndm,
edc7d573 1179 struct nlattr *tb[],
77162022 1180 struct net_device *dev,
6b6e2725 1181 const unsigned char *addr,
f6f6424b 1182 u16 vid,
77162022
JF
1183 u16 flags);
1184 int (*ndo_fdb_del)(struct ndmsg *ndm,
1690be63 1185 struct nlattr *tb[],
77162022 1186 struct net_device *dev,
f6f6424b
JP
1187 const unsigned char *addr,
1188 u16 vid);
77162022
JF
1189 int (*ndo_fdb_dump)(struct sk_buff *skb,
1190 struct netlink_callback *cb,
1191 struct net_device *dev,
5d5eacb3 1192 struct net_device *filter_dev,
77162022 1193 int idx);
e5a55a89
JF
1194
1195 int (*ndo_bridge_setlink)(struct net_device *dev,
add511b3
RP
1196 struct nlmsghdr *nlh,
1197 u16 flags);
e5a55a89
JF
1198 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1199 u32 pid, u32 seq,
6cbdceeb 1200 struct net_device *dev,
46c264da
ND
1201 u32 filter_mask,
1202 int nlflags);
407af329 1203 int (*ndo_bridge_dellink)(struct net_device *dev,
add511b3
RP
1204 struct nlmsghdr *nlh,
1205 u16 flags);
4bf84c35
JP
1206 int (*ndo_change_carrier)(struct net_device *dev,
1207 bool new_carrier);
66b52b0d 1208 int (*ndo_get_phys_port_id)(struct net_device *dev,
02637fce 1209 struct netdev_phys_item_id *ppid);
db24a904
DA
1210 int (*ndo_get_phys_port_name)(struct net_device *dev,
1211 char *name, size_t len);
53cf5275
JG
1212 void (*ndo_add_vxlan_port)(struct net_device *dev,
1213 sa_family_t sa_family,
35e42379 1214 __be16 port);
53cf5275
JG
1215 void (*ndo_del_vxlan_port)(struct net_device *dev,
1216 sa_family_t sa_family,
35e42379 1217 __be16 port);
a6cc0cfa
JF
1218
1219 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1220 struct net_device *dev);
1221 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1222 void *priv);
1223
1224 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1225 struct net_device *dev,
1226 void *priv);
25175ba5 1227 int (*ndo_get_lock_subclass)(struct net_device *dev);
5f35227e
JG
1228 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1229 struct net_device *dev,
1230 netdev_features_t features);
822b3b2e
JF
1231 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1232 int queue_index,
1233 u32 maxrate);
a54acb3a 1234 int (*ndo_get_iflink)(const struct net_device *dev);
d746d707
AK
1235 int (*ndo_change_proto_down)(struct net_device *dev,
1236 bool proto_down);
fc4099f1
PS
1237 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1238 struct sk_buff *skb);
d314774c
SH
1239};
1240
7aa98047
LR
1241/**
1242 * enum net_device_priv_flags - &struct net_device priv_flags
1243 *
1244 * These are the &struct net_device, they are only set internally
1245 * by drivers and used in the kernel. These flags are invisible to
1246 * userspace, this means that the order of these flags can change
1247 * during any kernel release.
1248 *
1249 * You should have a pretty good reason to be extending these flags.
1250 *
1251 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1252 * @IFF_EBRIDGE: Ethernet bridging device
7aa98047 1253 * @IFF_BONDING: bonding master or slave
7aa98047 1254 * @IFF_ISATAP: ISATAP interface (RFC4214)
7aa98047
LR
1255 * @IFF_WAN_HDLC: WAN HDLC device
1256 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1257 * release skb->dst
1258 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1259 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1260 * @IFF_MACVLAN_PORT: device used as macvlan port
1261 * @IFF_BRIDGE_PORT: device used as bridge port
1262 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1263 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1264 * @IFF_UNICAST_FLT: Supports unicast filtering
1265 * @IFF_TEAM_PORT: device used as team port
1266 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1267 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1268 * change when it's running
1269 * @IFF_MACVLAN: Macvlan device
007979ea 1270 * @IFF_L3MDEV_MASTER: device is an L3 master device
fa8187c9 1271 * @IFF_NO_QUEUE: device can run without qdisc attached
35d4e172 1272 * @IFF_OPENVSWITCH: device is a Open vSwitch master
fee6d4c7 1273 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
7aa98047
LR
1274 */
1275enum netdev_priv_flags {
1276 IFF_802_1Q_VLAN = 1<<0,
1277 IFF_EBRIDGE = 1<<1,
0dc1549b
JP
1278 IFF_BONDING = 1<<2,
1279 IFF_ISATAP = 1<<3,
1280 IFF_WAN_HDLC = 1<<4,
1281 IFF_XMIT_DST_RELEASE = 1<<5,
1282 IFF_DONT_BRIDGE = 1<<6,
1283 IFF_DISABLE_NETPOLL = 1<<7,
1284 IFF_MACVLAN_PORT = 1<<8,
1285 IFF_BRIDGE_PORT = 1<<9,
1286 IFF_OVS_DATAPATH = 1<<10,
1287 IFF_TX_SKB_SHARING = 1<<11,
1288 IFF_UNICAST_FLT = 1<<12,
1289 IFF_TEAM_PORT = 1<<13,
1290 IFF_SUPP_NOFCS = 1<<14,
1291 IFF_LIVE_ADDR_CHANGE = 1<<15,
1292 IFF_MACVLAN = 1<<16,
1293 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1294 IFF_IPVLAN_MASTER = 1<<18,
1295 IFF_IPVLAN_SLAVE = 1<<19,
007979ea 1296 IFF_L3MDEV_MASTER = 1<<20,
0dc1549b
JP
1297 IFF_NO_QUEUE = 1<<21,
1298 IFF_OPENVSWITCH = 1<<22,
fee6d4c7 1299 IFF_L3MDEV_SLAVE = 1<<23,
7aa98047
LR
1300};
1301
1302#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1303#define IFF_EBRIDGE IFF_EBRIDGE
7aa98047 1304#define IFF_BONDING IFF_BONDING
7aa98047 1305#define IFF_ISATAP IFF_ISATAP
7aa98047
LR
1306#define IFF_WAN_HDLC IFF_WAN_HDLC
1307#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1308#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1309#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1310#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1311#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1312#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1313#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1314#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1315#define IFF_TEAM_PORT IFF_TEAM_PORT
1316#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1317#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1318#define IFF_MACVLAN IFF_MACVLAN
02875878 1319#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
2ad7bf36
MB
1320#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1321#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
007979ea 1322#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
fa8187c9 1323#define IFF_NO_QUEUE IFF_NO_QUEUE
35d4e172 1324#define IFF_OPENVSWITCH IFF_OPENVSWITCH
8f25348b 1325#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
7aa98047 1326
536721b1
KK
1327/**
1328 * struct net_device - The DEVICE structure.
1329 * Actually, this whole structure is a big mistake. It mixes I/O
1330 * data with strictly "high-level" data, and it has to know about
1331 * almost every data structure used in the INET module.
1332 *
1333 * @name: This is the first field of the "visible" part of this structure
1334 * (i.e. as seen by users in the "Space.c" file). It is the name
1335 * of the interface.
1336 *
1337 * @name_hlist: Device name hash chain, please keep it close to name[]
1338 * @ifalias: SNMP alias
1339 * @mem_end: Shared memory end
1340 * @mem_start: Shared memory start
1341 * @base_addr: Device I/O address
1342 * @irq: Device IRQ number
1343 *
14ffbbb8
TG
1344 * @carrier_changes: Stats to monitor carrier on<->off transitions
1345 *
536721b1
KK
1346 * @state: Generic network queuing layer state, see netdev_state_t
1347 * @dev_list: The global list of network devices
1348 * @napi_list: List entry, that is used for polling napi devices
1349 * @unreg_list: List entry, that is used, when we are unregistering the
1350 * device, see the function unregister_netdev
1351 * @close_list: List entry, that is used, when we are closing the device
1352 *
1353 * @adj_list: Directly linked devices, like slaves for bonding
1354 * @all_adj_list: All linked devices, *including* neighbours
1355 * @features: Currently active device features
1356 * @hw_features: User-changeable features
1357 *
1358 * @wanted_features: User-requested features
1359 * @vlan_features: Mask of features inheritable by VLAN devices
1360 *
1361 * @hw_enc_features: Mask of features inherited by encapsulating devices
1362 * This field indicates what encapsulation
1363 * offloads the hardware is capable of doing,
1364 * and drivers will need to set them appropriately.
1365 *
1366 * @mpls_features: Mask of features inheritable by MPLS
1367 *
1368 * @ifindex: interface index
388069d3 1369 * @group: The group, that the device belongs to
536721b1
KK
1370 *
1371 * @stats: Statistics struct, which was left as a legacy, use
1372 * rtnl_link_stats64 instead
1373 *
1374 * @rx_dropped: Dropped packets by core network,
1375 * do not use this in drivers
1376 * @tx_dropped: Dropped packets by core network,
1377 * do not use this in drivers
1378 *
536721b1
KK
1379 * @wireless_handlers: List of functions to handle Wireless Extensions,
1380 * instead of ioctl,
1381 * see <net/iw_handler.h> for details.
1382 * @wireless_data: Instance data managed by the core of wireless extensions
1383 *
1384 * @netdev_ops: Includes several pointers to callbacks,
1385 * if one wants to override the ndo_*() functions
1386 * @ethtool_ops: Management operations
d476059e 1387 * @header_ops: Includes callbacks for creating,parsing,caching,etc
536721b1
KK
1388 * of Layer 2 headers.
1389 *
1390 * @flags: Interface flags (a la BSD)
1391 * @priv_flags: Like 'flags' but invisible to userspace,
1392 * see if.h for the definitions
1393 * @gflags: Global flags ( kept as legacy )
1394 * @padded: How much padding added by alloc_netdev()
1395 * @operstate: RFC2863 operstate
1396 * @link_mode: Mapping policy to operstate
1397 * @if_port: Selectable AUI, TP, ...
1398 * @dma: DMA channel
1399 * @mtu: Interface MTU value
1400 * @type: Interface hardware type
880621c2
MB
1401 * @hard_header_len: Hardware header length, which means that this is the
1402 * minimum size of a packet.
536721b1
KK
1403 *
1404 * @needed_headroom: Extra headroom the hardware may need, but not in all
1405 * cases can this be guaranteed
1406 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1407 * cases can this be guaranteed. Some cases also use
1408 * LL_MAX_HEADER instead to allocate the skb
1409 *
1410 * interface address info:
1411 *
1412 * @perm_addr: Permanent hw address
1413 * @addr_assign_type: Hw address assignment type
1414 * @addr_len: Hardware address length
1415 * @neigh_priv_len; Used in neigh_alloc(),
1416 * initialized only in atm/clip.c
1417 * @dev_id: Used to differentiate devices that share
1418 * the same link layer address
1419 * @dev_port: Used to differentiate devices that share
1420 * the same function
1421 * @addr_list_lock: XXX: need comments on this one
536721b1
KK
1422 * @uc_promisc: Counter, that indicates, that promiscuous mode
1423 * has been enabled due to the need to listen to
1424 * additional unicast addresses in a device that
1425 * does not implement ndo_set_rx_mode()
14ffbbb8
TG
1426 * @uc: unicast mac addresses
1427 * @mc: multicast mac addresses
1428 * @dev_addrs: list of device hw addresses
1429 * @queues_kset: Group of all Kobjects in the Tx and RX queues
536721b1
KK
1430 * @promiscuity: Number of times, the NIC is told to work in
1431 * Promiscuous mode, if it becomes 0 the NIC will
1432 * exit from working in Promiscuous mode
1433 * @allmulti: Counter, enables or disables allmulticast mode
1434 *
1435 * @vlan_info: VLAN info
1436 * @dsa_ptr: dsa specific data
1437 * @tipc_ptr: TIPC specific data
1438 * @atalk_ptr: AppleTalk link
1439 * @ip_ptr: IPv4 specific data
1440 * @dn_ptr: DECnet specific data
1441 * @ip6_ptr: IPv6 specific data
1442 * @ax25_ptr: AX.25 specific data
1443 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1444 *
1445 * @last_rx: Time of last Rx
1446 * @dev_addr: Hw address (before bcast,
1447 * because most packets are unicast)
1448 *
1449 * @_rx: Array of RX queues
1450 * @num_rx_queues: Number of RX queues
1451 * allocated at register_netdev() time
1452 * @real_num_rx_queues: Number of RX queues currently active in device
1453 *
1454 * @rx_handler: handler for received packets
1455 * @rx_handler_data: XXX: need comments on this one
1456 * @ingress_queue: XXX: need comments on this one
1457 * @broadcast: hw bcast address
1458 *
14ffbbb8
TG
1459 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1460 * indexed by RX queue number. Assigned by driver.
1461 * This must only be set if the ndo_rx_flow_steer
1462 * operation is defined
1463 * @index_hlist: Device index hash chain
1464 *
536721b1
KK
1465 * @_tx: Array of TX queues
1466 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1467 * @real_num_tx_queues: Number of TX queues currently active in device
1468 * @qdisc: Root qdisc from userspace point of view
1469 * @tx_queue_len: Max frames per queue allowed
1470 * @tx_global_lock: XXX: need comments on this one
1471 *
1472 * @xps_maps: XXX: need comments on this one
1473 *
0c4f691f
SF
1474 * @offload_fwd_mark: Offload device fwding mark
1475 *
536721b1
KK
1476 * @trans_start: Time (in jiffies) of last Tx
1477 * @watchdog_timeo: Represents the timeout that is used by
1478 * the watchdog ( see dev_watchdog() )
1479 * @watchdog_timer: List of timers
1480 *
1481 * @pcpu_refcnt: Number of references to this device
1482 * @todo_list: Delayed register/unregister
536721b1
KK
1483 * @link_watch_list: XXX: need comments on this one
1484 *
1485 * @reg_state: Register/unregister state machine
1486 * @dismantle: Device is going to be freed
1487 * @rtnl_link_state: This enum represents the phases of creating
1488 * a new link
1489 *
1490 * @destructor: Called from unregister,
1491 * can be used to call free_netdev
1492 * @npinfo: XXX: need comments on this one
1493 * @nd_net: Network namespace this network device is inside
1494 *
1495 * @ml_priv: Mid-layer private
1496 * @lstats: Loopback statistics
1497 * @tstats: Tunnel statistics
1498 * @dstats: Dummy statistics
1499 * @vstats: Virtual ethernet statistics
1500 *
1501 * @garp_port: GARP
1502 * @mrp_port: MRP
1503 *
1504 * @dev: Class/net/name entry
1505 * @sysfs_groups: Space for optional device, statistics and wireless
1506 * sysfs groups
1507 *
1508 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1509 * @rtnl_link_ops: Rtnl_link_ops
1510 *
1511 * @gso_max_size: Maximum size of generic segmentation offload
1512 * @gso_max_segs: Maximum number of segments that can be passed to the
1513 * NIC for GSO
fcbeb976
ED
1514 * @gso_min_segs: Minimum number of segments that can be passed to the
1515 * NIC for GSO
536721b1
KK
1516 *
1517 * @dcbnl_ops: Data Center Bridging netlink ops
1518 * @num_tc: Number of traffic classes in the net device
1519 * @tc_to_txq: XXX: need comments on this one
1520 * @prio_tc_map XXX: need comments on this one
1521 *
1522 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1523 *
1524 * @priomap: XXX: need comments on this one
1525 * @phydev: Physical device may attach itself
1526 * for hardware timestamping
1527 *
1528 * @qdisc_tx_busylock: XXX: need comments on this one
1529 *
d746d707
AK
1530 * @proto_down: protocol port state information can be sent to the
1531 * switch driver and used to set the phys state of the
1532 * switch port.
1533 *
1da177e4
LT
1534 * FIXME: cleanup struct net_device such that network protocol info
1535 * moves out.
1536 */
1537
d94d9fee 1538struct net_device {
1da177e4 1539 char name[IFNAMSIZ];
9356b8fc 1540 struct hlist_node name_hlist;
0b815a1a 1541 char *ifalias;
1da177e4
LT
1542 /*
1543 * I/O specific fields
1544 * FIXME: Merge these and struct ifmap into one
1545 */
536721b1
KK
1546 unsigned long mem_end;
1547 unsigned long mem_start;
1548 unsigned long base_addr;
1549 int irq;
1da177e4 1550
14ffbbb8
TG
1551 atomic_t carrier_changes;
1552
1da177e4 1553 /*
536721b1
KK
1554 * Some hardware also needs these fields (state,dev_list,
1555 * napi_list,unreg_list,close_list) but they are not
1da177e4
LT
1556 * part of the usual set specified in Space.c.
1557 */
1558
1da177e4
LT
1559 unsigned long state;
1560
7562f876 1561 struct list_head dev_list;
bea3348e 1562 struct list_head napi_list;
44a0873d 1563 struct list_head unreg_list;
5cde2829 1564 struct list_head close_list;
7866a621
SN
1565 struct list_head ptype_all;
1566 struct list_head ptype_specific;
2f268f12 1567
2f268f12
VF
1568 struct {
1569 struct list_head upper;
1570 struct list_head lower;
1571 } adj_list;
1572
2f268f12
VF
1573 struct {
1574 struct list_head upper;
1575 struct list_head lower;
1576 } all_adj_list;
4c3d5e7b 1577
c8f44aff 1578 netdev_features_t features;
c8f44aff 1579 netdev_features_t hw_features;
c8f44aff 1580 netdev_features_t wanted_features;
c8f44aff 1581 netdev_features_t vlan_features;
6a674e9c 1582 netdev_features_t hw_enc_features;
0d89d203 1583 netdev_features_t mpls_features;
04ed3e74 1584
1da177e4 1585 int ifindex;
7a66bbc9 1586 int group;
1da177e4 1587
c45d286e 1588 struct net_device_stats stats;
015f0688 1589
015f0688
ED
1590 atomic_long_t rx_dropped;
1591 atomic_long_t tx_dropped;
1da177e4 1592
b86e0280 1593#ifdef CONFIG_WIRELESS_EXT
1da177e4 1594 const struct iw_handler_def * wireless_handlers;
1da177e4 1595 struct iw_public_data * wireless_data;
b86e0280 1596#endif
d314774c 1597 const struct net_device_ops *netdev_ops;
76fd8593 1598 const struct ethtool_ops *ethtool_ops;
4170604f 1599#ifdef CONFIG_NET_SWITCHDEV
9d47c0a2 1600 const struct switchdev_ops *switchdev_ops;
4170604f 1601#endif
1b69c6d0
DA
1602#ifdef CONFIG_NET_L3_MASTER_DEV
1603 const struct l3mdev_ops *l3mdev_ops;
1604#endif
1da177e4 1605
3b04ddde
SH
1606 const struct header_ops *header_ops;
1607
536721b1
KK
1608 unsigned int flags;
1609 unsigned int priv_flags;
1610
1da177e4 1611 unsigned short gflags;
536721b1 1612 unsigned short padded;
1da177e4 1613
536721b1
KK
1614 unsigned char operstate;
1615 unsigned char link_mode;
b00055aa 1616
536721b1
KK
1617 unsigned char if_port;
1618 unsigned char dma;
bdc220da 1619
536721b1
KK
1620 unsigned int mtu;
1621 unsigned short type;
1622 unsigned short hard_header_len;
1da177e4 1623
f5184d26
JB
1624 unsigned short needed_headroom;
1625 unsigned short needed_tailroom;
1626
1da177e4 1627 /* Interface address info. */
536721b1
KK
1628 unsigned char perm_addr[MAX_ADDR_LEN];
1629 unsigned char addr_assign_type;
1630 unsigned char addr_len;
a0a9663d 1631 unsigned short neigh_priv_len;
536721b1
KK
1632 unsigned short dev_id;
1633 unsigned short dev_port;
ccffad25 1634 spinlock_t addr_list_lock;
14ffbbb8
TG
1635 unsigned char name_assign_type;
1636 bool uc_promisc;
536721b1
KK
1637 struct netdev_hw_addr_list uc;
1638 struct netdev_hw_addr_list mc;
1639 struct netdev_hw_addr_list dev_addrs;
1640
4c3d5e7b
ED
1641#ifdef CONFIG_SYSFS
1642 struct kset *queues_kset;
1643#endif
9d45abe1
WC
1644 unsigned int promiscuity;
1645 unsigned int allmulti;
1da177e4 1646
1da177e4
LT
1647
1648 /* Protocol specific pointers */
65ac6a5f 1649
d11ead75 1650#if IS_ENABLED(CONFIG_VLAN_8021Q)
536721b1 1651 struct vlan_info __rcu *vlan_info;
65ac6a5f 1652#endif
34a430d7 1653#if IS_ENABLED(CONFIG_NET_DSA)
536721b1 1654 struct dsa_switch_tree *dsa_ptr;
37cb0620
YX
1655#endif
1656#if IS_ENABLED(CONFIG_TIPC)
536721b1 1657 struct tipc_bearer __rcu *tipc_ptr;
91da11f8 1658#endif
536721b1
KK
1659 void *atalk_ptr;
1660 struct in_device __rcu *ip_ptr;
1661 struct dn_dev __rcu *dn_ptr;
1662 struct inet6_dev __rcu *ip6_ptr;
1663 void *ax25_ptr;
1664 struct wireless_dev *ieee80211_ptr;
98a18b6f 1665 struct wpan_dev *ieee802154_ptr;
03c57747
RS
1666#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1667 struct mpls_dev __rcu *mpls_ptr;
1668#endif
1da177e4 1669
9356b8fc 1670/*
cd13539b 1671 * Cache lines mostly used on receive path (including eth_type_trans())
9356b8fc 1672 */
536721b1 1673 unsigned long last_rx;
4dc89133 1674
9356b8fc 1675 /* Interface address info used in eth_type_trans() */
536721b1 1676 unsigned char *dev_addr;
f001fde5 1677
0a9627f2 1678
a953be53 1679#ifdef CONFIG_SYSFS
0a9627f2
TH
1680 struct netdev_rx_queue *_rx;
1681
0a9627f2 1682 unsigned int num_rx_queues;
62fe0b40 1683 unsigned int real_num_rx_queues;
c445477d 1684
df334545 1685#endif
0a9627f2 1686
3b47d303 1687 unsigned long gro_flush_timeout;
61391cde 1688 rx_handler_func_t __rcu *rx_handler;
1689 void __rcu *rx_handler_data;
e8a0464c 1690
4cda01e8 1691#ifdef CONFIG_NET_CLS_ACT
d2788d34
DB
1692 struct tcf_proto __rcu *ingress_cl_list;
1693#endif
24824a09 1694 struct netdev_queue __rcu *ingress_queue;
e687ad60
PN
1695#ifdef CONFIG_NETFILTER_INGRESS
1696 struct list_head nf_hooks_ingress;
1697#endif
d2788d34 1698
536721b1 1699 unsigned char broadcast[MAX_ADDR_LEN];
14ffbbb8
TG
1700#ifdef CONFIG_RFS_ACCEL
1701 struct cpu_rmap *rx_cpu_rmap;
1702#endif
1703 struct hlist_node index_hlist;
cd13539b
ED
1704
1705/*
1706 * Cache lines mostly used on transmit path
1707 */
e8a0464c
DM
1708 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1709 unsigned int num_tx_queues;
fd2ea0a7 1710 unsigned int real_num_tx_queues;
af356afa 1711 struct Qdisc *qdisc;
536721b1 1712 unsigned long tx_queue_len;
c3f26a26 1713 spinlock_t tx_global_lock;
14ffbbb8 1714 int watchdog_timeo;
cd13539b 1715
bf264145 1716#ifdef CONFIG_XPS
a4177869 1717 struct xps_dev_maps __rcu *xps_maps;
bf264145 1718#endif
1d24eb48 1719
0c4f691f
SF
1720#ifdef CONFIG_NET_SWITCHDEV
1721 u32 offload_fwd_mark;
1722#endif
1723
9356b8fc 1724 /* These may be needed for future network-power-down code. */
9d21493b
ED
1725
1726 /*
1727 * trans_start here is expensive for high speed devices on SMP,
1728 * please use netdev_queue->trans_start instead.
1729 */
536721b1 1730 unsigned long trans_start;
9356b8fc 1731
9356b8fc
ED
1732 struct timer_list watchdog_timer;
1733
29b4433d 1734 int __percpu *pcpu_refcnt;
1da177e4 1735 struct list_head todo_list;
1da177e4 1736
e014debe 1737 struct list_head link_watch_list;
572a103d 1738
1da177e4 1739 enum { NETREG_UNINITIALIZED=0,
b17a7c17 1740 NETREG_REGISTERED, /* completed register_netdevice */
1da177e4
LT
1741 NETREG_UNREGISTERING, /* called unregister_netdevice */
1742 NETREG_UNREGISTERED, /* completed unregister todo */
1743 NETREG_RELEASED, /* called free_netdev */
937f1ba5 1744 NETREG_DUMMY, /* dummy device for NAPI poll */
449f4544
ED
1745 } reg_state:8;
1746
536721b1 1747 bool dismantle;
a2835763
PM
1748
1749 enum {
1750 RTNL_LINK_INITIALIZED,
1751 RTNL_LINK_INITIALIZING,
1752 } rtnl_link_state:16;
1da177e4 1753
d314774c 1754 void (*destructor)(struct net_device *dev);
1da177e4 1755
1da177e4 1756#ifdef CONFIG_NETPOLL
5fbee843 1757 struct netpoll_info __rcu *npinfo;
1da177e4 1758#endif
eae792b7 1759
0c5c9fb5 1760 possible_net_t nd_net;
4a1c5371 1761
4951704b 1762 /* mid-layer private */
a7855c78 1763 union {
536721b1
KK
1764 void *ml_priv;
1765 struct pcpu_lstats __percpu *lstats;
8f84985f 1766 struct pcpu_sw_netstats __percpu *tstats;
536721b1
KK
1767 struct pcpu_dstats __percpu *dstats;
1768 struct pcpu_vstats __percpu *vstats;
a7855c78 1769 };
536721b1 1770
3cc77ec7 1771 struct garp_port __rcu *garp_port;
febf018d 1772 struct mrp_port __rcu *mrp_port;
1da177e4 1773
536721b1 1774 struct device dev;
0c509a6c 1775 const struct attribute_group *sysfs_groups[4];
a953be53 1776 const struct attribute_group *sysfs_rx_queue_group;
38f7b870 1777
38f7b870 1778 const struct rtnl_link_ops *rtnl_link_ops;
f25f4e44 1779
82cc1a7a
PWJ
1780 /* for setting kernel sock attribute on TCP connection setup */
1781#define GSO_MAX_SIZE 65536
1782 unsigned int gso_max_size;
30b678d8
BH
1783#define GSO_MAX_SEGS 65535
1784 u16 gso_max_segs;
fcbeb976 1785 u16 gso_min_segs;
7a6b6f51 1786#ifdef CONFIG_DCB
32953543 1787 const struct dcbnl_rtnl_ops *dcbnl_ops;
2f90b865 1788#endif
4f57c087
JF
1789 u8 num_tc;
1790 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1791 u8 prio_tc_map[TC_BITMASK + 1];
2f90b865 1792
d11ead75 1793#if IS_ENABLED(CONFIG_FCOE)
4d288d57 1794 unsigned int fcoe_ddp_xid;
5bc1421e 1795#endif
86f8515f 1796#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
5bc1421e 1797 struct netprio_map __rcu *priomap;
4d288d57 1798#endif
c1f19b51 1799 struct phy_device *phydev;
23d3b8bf 1800 struct lock_class_key *qdisc_tx_busylock;
d746d707 1801 bool proto_down;
1da177e4 1802};
43cb76d9 1803#define to_net_dev(d) container_of(d, struct net_device, dev)
1da177e4
LT
1804
1805#define NETDEV_ALIGN 32
1da177e4 1806
4f57c087
JF
1807static inline
1808int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1809{
1810 return dev->prio_tc_map[prio & TC_BITMASK];
1811}
1812
1813static inline
1814int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1815{
1816 if (tc >= dev->num_tc)
1817 return -EINVAL;
1818
1819 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1820 return 0;
1821}
1822
1823static inline
1824void netdev_reset_tc(struct net_device *dev)
1825{
1826 dev->num_tc = 0;
1827 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1828 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1829}
1830
1831static inline
1832int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1833{
1834 if (tc >= dev->num_tc)
1835 return -EINVAL;
1836
1837 dev->tc_to_txq[tc].count = count;
1838 dev->tc_to_txq[tc].offset = offset;
1839 return 0;
1840}
1841
1842static inline
1843int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1844{
1845 if (num_tc > TC_MAX_QUEUE)
1846 return -EINVAL;
1847
1848 dev->num_tc = num_tc;
1849 return 0;
1850}
1851
1852static inline
1853int netdev_get_num_tc(struct net_device *dev)
1854{
1855 return dev->num_tc;
1856}
1857
e8a0464c
DM
1858static inline
1859struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1860 unsigned int index)
1861{
1862 return &dev->_tx[index];
1863}
1864
10c51b56
DB
1865static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1866 const struct sk_buff *skb)
1867{
1868 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1869}
1870
e8a0464c
DM
1871static inline void netdev_for_each_tx_queue(struct net_device *dev,
1872 void (*f)(struct net_device *,
1873 struct netdev_queue *,
1874 void *),
1875 void *arg)
1876{
1877 unsigned int i;
1878
1879 for (i = 0; i < dev->num_tx_queues; i++)
1880 f(dev, &dev->_tx[i], arg);
1881}
1882
f629d208 1883struct netdev_queue *netdev_pick_tx(struct net_device *dev,
f663dd9a
JW
1884 struct sk_buff *skb,
1885 void *accel_priv);
8c4c49df 1886
c346dca1
YH
1887/*
1888 * Net namespace inlines
1889 */
1890static inline
1891struct net *dev_net(const struct net_device *dev)
1892{
c2d9ba9b 1893 return read_pnet(&dev->nd_net);
c346dca1
YH
1894}
1895
1896static inline
f5aa23fd 1897void dev_net_set(struct net_device *dev, struct net *net)
c346dca1 1898{
0c5c9fb5 1899 write_pnet(&dev->nd_net, net);
c346dca1
YH
1900}
1901
3e8a72d1 1902static inline bool netdev_uses_dsa(struct net_device *dev)
cf85d08f 1903{
3fc88677 1904#if IS_ENABLED(CONFIG_NET_DSA)
5aed85ce
FF
1905 if (dev->dsa_ptr != NULL)
1906 return dsa_uses_tagged_protocol(dev->dsa_ptr);
396138f0 1907#endif
5aed85ce 1908 return false;
396138f0
LB
1909}
1910
bea3348e
SH
1911/**
1912 * netdev_priv - access network device private data
1913 * @dev: network device
1914 *
1915 * Get network device private data
1916 */
6472ce60 1917static inline void *netdev_priv(const struct net_device *dev)
1da177e4 1918{
1ce8e7b5 1919 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1da177e4
LT
1920}
1921
1da177e4
LT
1922/* Set the sysfs physical device reference for the network logical device
1923 * if set prior to registration will cause a symlink during initialization.
1924 */
43cb76d9 1925#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1da177e4 1926
384912ed 1927/* Set the sysfs device type for the network logical device to allow
3f79410c 1928 * fine-grained identification of different network device types. For
384912ed
MH
1929 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1930 */
1931#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1932
82dc3c63
ED
1933/* Default NAPI poll() weight
1934 * Device drivers are strongly advised to not use bigger value
1935 */
1936#define NAPI_POLL_WEIGHT 64
1937
3b582cc1
SH
1938/**
1939 * netif_napi_add - initialize a napi context
1940 * @dev: network device
1941 * @napi: napi context
1942 * @poll: polling function
1943 * @weight: default weight
1944 *
1945 * netif_napi_add() must be used to initialize a napi context prior to calling
1946 * *any* of the other napi related functions.
1947 */
d565b0a1
HX
1948void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1949 int (*poll)(struct napi_struct *, int), int weight);
bea3348e 1950
d8156534
AD
1951/**
1952 * netif_napi_del - remove a napi context
1953 * @napi: napi context
1954 *
1955 * netif_napi_del() removes a napi context from the network device napi list
1956 */
d565b0a1
HX
1957void netif_napi_del(struct napi_struct *napi);
1958
1959struct napi_gro_cb {
78a478d0
HX
1960 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1961 void *frag0;
1962
7489594c
HX
1963 /* Length of frag0. */
1964 unsigned int frag0_len;
1965
86911732
HX
1966 /* This indicates where we are processing relative to skb->data. */
1967 int data_offset;
1968
d565b0a1 1969 /* This is non-zero if the packet cannot be merged with the new skb. */
bf5a755f
JC
1970 u16 flush;
1971
1972 /* Save the IP ID here and check when we get to the transport layer */
1973 u16 flush_id;
d565b0a1
HX
1974
1975 /* Number of segments aggregated. */
2e71a6f8
ED
1976 u16 count;
1977
15e2396d
TH
1978 /* Start offset for remote checksum offload */
1979 u16 gro_remcsum_start;
1980
2e71a6f8
ED
1981 /* jiffies when first packet was created/queued */
1982 unsigned long age;
86347245 1983
afe93325 1984 /* Used in ipv6_gro_receive() and foo-over-udp */
b582ef09
OG
1985 u16 proto;
1986
baa32ff4
TH
1987 /* This is non-zero if the packet may be of the same flow. */
1988 u8 same_flow:1;
1989
b582ef09 1990 /* Used in udp_gro_receive */
573e8fca
TH
1991 u8 udp_mark:1;
1992
1993 /* GRO checksum is valid */
1994 u8 csum_valid:1;
1995
662880f4
TH
1996 /* Number of checksums via CHECKSUM_UNNECESSARY */
1997 u8 csum_cnt:3;
c3c7c254 1998
baa32ff4
TH
1999 /* Free the skb? */
2000 u8 free:2;
2001#define NAPI_GRO_FREE 1
2002#define NAPI_GRO_FREE_STOLEN_HEAD 2
2003
efc98d08
TH
2004 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2005 u8 is_ipv6:1;
2006
baa32ff4
TH
2007 /* 7 bit hole */
2008
bf5a755f
JC
2009 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2010 __wsum csum;
2011
c3c7c254
ED
2012 /* used in skb_gro_receive() slow path */
2013 struct sk_buff *last;
d565b0a1
HX
2014};
2015
2016#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
d8156534 2017
1da177e4 2018struct packet_type {
f2ccd8fa
DM
2019 __be16 type; /* This is really htons(ether_type). */
2020 struct net_device *dev; /* NULL is wildcarded here */
2021 int (*func) (struct sk_buff *,
2022 struct net_device *,
2023 struct packet_type *,
2024 struct net_device *);
c0de08d0
EL
2025 bool (*id_match)(struct packet_type *ptype,
2026 struct sock *sk);
1da177e4
LT
2027 void *af_packet_priv;
2028 struct list_head list;
2029};
2030
f191a1d1 2031struct offload_callbacks {
576a30eb 2032 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
c8f44aff 2033 netdev_features_t features);
d565b0a1 2034 struct sk_buff **(*gro_receive)(struct sk_buff **head,
a2b12f3c 2035 struct sk_buff *skb);
299603e8 2036 int (*gro_complete)(struct sk_buff *skb, int nhoff);
f191a1d1
VY
2037};
2038
2039struct packet_offload {
2040 __be16 type; /* This is really htons(ether_type). */
bdef7de4 2041 u16 priority;
f191a1d1
VY
2042 struct offload_callbacks callbacks;
2043 struct list_head list;
1da177e4
LT
2044};
2045
a2b12f3c
TH
2046struct udp_offload;
2047
2048struct udp_offload_callbacks {
2049 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2050 struct sk_buff *skb,
2051 struct udp_offload *uoff);
2052 int (*gro_complete)(struct sk_buff *skb,
2053 int nhoff,
2054 struct udp_offload *uoff);
2055};
2056
b582ef09
OG
2057struct udp_offload {
2058 __be16 port;
afe93325 2059 u8 ipproto;
a2b12f3c 2060 struct udp_offload_callbacks callbacks;
b582ef09
OG
2061};
2062
8f84985f
LR
2063/* often modified stats are per cpu, other are shared (netdev->stats) */
2064struct pcpu_sw_netstats {
2065 u64 rx_packets;
2066 u64 rx_bytes;
2067 u64 tx_packets;
2068 u64 tx_bytes;
2069 struct u64_stats_sync syncp;
2070};
2071
aabc92bb
PNA
2072#define __netdev_alloc_pcpu_stats(type, gfp) \
2073({ \
2074 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2075 if (pcpu_stats) { \
2076 int __cpu; \
2077 for_each_possible_cpu(__cpu) { \
2078 typeof(type) *stat; \
2079 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2080 u64_stats_init(&stat->syncp); \
2081 } \
2082 } \
2083 pcpu_stats; \
1c213bd2
WC
2084})
2085
aabc92bb
PNA
2086#define netdev_alloc_pcpu_stats(type) \
2087 __netdev_alloc_pcpu_stats(type, GFP_KERNEL);
2088
1da177e4
LT
2089#include <linux/notifier.h>
2090
dcfe1421
AW
2091/* netdevice notifier chain. Please remember to update the rtnetlink
2092 * notification exclusion list in rtnetlink_event() when adding new
2093 * types.
2094 */
2095#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2096#define NETDEV_DOWN 0x0002
2097#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2098 detected a hardware crash and restarted
2099 - we can use this eg to kick tcp sessions
2100 once done */
2101#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2102#define NETDEV_REGISTER 0x0005
2103#define NETDEV_UNREGISTER 0x0006
1d486bfb 2104#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
dcfe1421
AW
2105#define NETDEV_CHANGEADDR 0x0008
2106#define NETDEV_GOING_DOWN 0x0009
2107#define NETDEV_CHANGENAME 0x000A
2108#define NETDEV_FEAT_CHANGE 0x000B
2109#define NETDEV_BONDING_FAILOVER 0x000C
2110#define NETDEV_PRE_UP 0x000D
2111#define NETDEV_PRE_TYPE_CHANGE 0x000E
2112#define NETDEV_POST_TYPE_CHANGE 0x000F
2113#define NETDEV_POST_INIT 0x0010
0115e8e3 2114#define NETDEV_UNREGISTER_FINAL 0x0011
dcfe1421
AW
2115#define NETDEV_RELEASE 0x0012
2116#define NETDEV_NOTIFY_PEERS 0x0013
2117#define NETDEV_JOIN 0x0014
42e52bf9 2118#define NETDEV_CHANGEUPPER 0x0015
4aa5dee4 2119#define NETDEV_RESEND_IGMP 0x0016
1d486bfb 2120#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
d4261e56 2121#define NETDEV_CHANGEINFODATA 0x0018
61bd3857 2122#define NETDEV_BONDING_INFO 0x0019
573c7ba0 2123#define NETDEV_PRECHANGEUPPER 0x001A
dcfe1421 2124
f629d208
JP
2125int register_netdevice_notifier(struct notifier_block *nb);
2126int unregister_netdevice_notifier(struct notifier_block *nb);
351638e7
JP
2127
2128struct netdev_notifier_info {
2129 struct net_device *dev;
2130};
2131
be9efd36
JP
2132struct netdev_notifier_change_info {
2133 struct netdev_notifier_info info; /* must be first */
2134 unsigned int flags_changed;
2135};
2136
0e4ead9d
JP
2137struct netdev_notifier_changeupper_info {
2138 struct netdev_notifier_info info; /* must be first */
2139 struct net_device *upper_dev; /* new upper dev */
2140 bool master; /* is upper dev master */
2141 bool linking; /* is the nofication for link or unlink */
2142};
2143
75538c2b
CW
2144static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2145 struct net_device *dev)
2146{
2147 info->dev = dev;
2148}
2149
351638e7
JP
2150static inline struct net_device *
2151netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2152{
2153 return info->dev;
2154}
2155
f629d208 2156int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
dcfe1421
AW
2157
2158
1da177e4
LT
2159extern rwlock_t dev_base_lock; /* Device list lock */
2160
881d966b
EB
2161#define for_each_netdev(net, d) \
2162 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
dcbccbd4
EB
2163#define for_each_netdev_reverse(net, d) \
2164 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
c6d14c84
ED
2165#define for_each_netdev_rcu(net, d) \
2166 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
881d966b
EB
2167#define for_each_netdev_safe(net, d, n) \
2168 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2169#define for_each_netdev_continue(net, d) \
2170 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
254245d2 2171#define for_each_netdev_continue_rcu(net, d) \
2172 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
8a7fbfab 2173#define for_each_netdev_in_bond_rcu(bond, slave) \
2174 for_each_netdev_rcu(&init_net, slave) \
4ccce02e 2175 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
881d966b 2176#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
7562f876 2177
a050c33f
DL
2178static inline struct net_device *next_net_device(struct net_device *dev)
2179{
2180 struct list_head *lh;
2181 struct net *net;
2182
c346dca1 2183 net = dev_net(dev);
a050c33f
DL
2184 lh = dev->dev_list.next;
2185 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2186}
2187
ce81b76a
ED
2188static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2189{
2190 struct list_head *lh;
2191 struct net *net;
2192
2193 net = dev_net(dev);
ccf43438 2194 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
ce81b76a
ED
2195 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2196}
2197
a050c33f
DL
2198static inline struct net_device *first_net_device(struct net *net)
2199{
2200 return list_empty(&net->dev_base_head) ? NULL :
2201 net_device_entry(net->dev_base_head.next);
2202}
7562f876 2203
ccf43438
ED
2204static inline struct net_device *first_net_device_rcu(struct net *net)
2205{
2206 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2207
2208 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2209}
2210
f629d208
JP
2211int netdev_boot_setup_check(struct net_device *dev);
2212unsigned long netdev_boot_base(const char *prefix, int unit);
2213struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2214 const char *hwaddr);
2215struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2216struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2217void dev_add_pack(struct packet_type *pt);
2218void dev_remove_pack(struct packet_type *pt);
2219void __dev_remove_pack(struct packet_type *pt);
2220void dev_add_offload(struct packet_offload *po);
2221void dev_remove_offload(struct packet_offload *po);
f629d208 2222
a54acb3a 2223int dev_get_iflink(const struct net_device *dev);
fc4099f1 2224int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
6c555490
WC
2225struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2226 unsigned short mask);
f629d208
JP
2227struct net_device *dev_get_by_name(struct net *net, const char *name);
2228struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2229struct net_device *__dev_get_by_name(struct net *net, const char *name);
2230int dev_alloc_name(struct net_device *dev, const char *name);
2231int dev_open(struct net_device *dev);
2232int dev_close(struct net_device *dev);
99c4a26a 2233int dev_close_many(struct list_head *head, bool unlink);
f629d208 2234void dev_disable_lro(struct net_device *dev);
0c4b51f0 2235int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2b4aa3ce 2236int dev_queue_xmit(struct sk_buff *skb);
f663dd9a 2237int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
f629d208
JP
2238int register_netdevice(struct net_device *dev);
2239void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2240void unregister_netdevice_many(struct list_head *head);
44a0873d
ED
2241static inline void unregister_netdevice(struct net_device *dev)
2242{
2243 unregister_netdevice_queue(dev, NULL);
2244}
2245
f629d208
JP
2246int netdev_refcnt_read(const struct net_device *dev);
2247void free_netdev(struct net_device *dev);
74d332c1 2248void netdev_freemem(struct net_device *dev);
f629d208
JP
2249void synchronize_net(void);
2250int init_dummy_netdev(struct net_device *dev);
937f1ba5 2251
f60e5990 2252DECLARE_PER_CPU(int, xmit_recursion);
2253static inline int dev_recursion_level(void)
2254{
2255 return this_cpu_read(xmit_recursion);
2256}
2257
f629d208
JP
2258struct net_device *dev_get_by_index(struct net *net, int ifindex);
2259struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2260struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2261int netdev_get_name(struct net *net, char *name, int ifindex);
2262int dev_restart(struct net_device *dev);
f629d208 2263int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
86911732
HX
2264
2265static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2266{
2267 return NAPI_GRO_CB(skb)->data_offset;
2268}
2269
2270static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2271{
2272 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2273}
2274
2275static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2276{
2277 NAPI_GRO_CB(skb)->data_offset += len;
2278}
2279
a5b1cf28
HX
2280static inline void *skb_gro_header_fast(struct sk_buff *skb,
2281 unsigned int offset)
86911732 2282{
a5b1cf28
HX
2283 return NAPI_GRO_CB(skb)->frag0 + offset;
2284}
78a478d0 2285
a5b1cf28
HX
2286static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2287{
2288 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2289}
78a478d0 2290
a5b1cf28
HX
2291static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2292 unsigned int offset)
2293{
17dd759c
HX
2294 if (!pskb_may_pull(skb, hlen))
2295 return NULL;
2296
a5b1cf28
HX
2297 NAPI_GRO_CB(skb)->frag0 = NULL;
2298 NAPI_GRO_CB(skb)->frag0_len = 0;
17dd759c 2299 return skb->data + offset;
86911732 2300}
1da177e4 2301
36e7b1b8
HX
2302static inline void *skb_gro_network_header(struct sk_buff *skb)
2303{
78d3fd0b
HX
2304 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2305 skb_network_offset(skb);
36e7b1b8
HX
2306}
2307
bf5a755f
JC
2308static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2309 const void *start, unsigned int len)
2310{
573e8fca 2311 if (NAPI_GRO_CB(skb)->csum_valid)
bf5a755f
JC
2312 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2313 csum_partial(start, len, 0));
2314}
2315
573e8fca
TH
2316/* GRO checksum functions. These are logical equivalents of the normal
2317 * checksum functions (in skbuff.h) except that they operate on the GRO
2318 * offsets and fields in sk_buff.
2319 */
2320
2321__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2322
15e2396d
TH
2323static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2324{
b7fe10e5 2325 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
15e2396d
TH
2326}
2327
573e8fca
TH
2328static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2329 bool zero_okay,
2330 __sum16 check)
2331{
6edec0e6
TH
2332 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2333 skb_checksum_start_offset(skb) <
2334 skb_gro_offset(skb)) &&
15e2396d 2335 !skb_at_gro_remcsum_start(skb) &&
662880f4 2336 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
573e8fca
TH
2337 (!zero_okay || check));
2338}
2339
2340static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2341 __wsum psum)
2342{
2343 if (NAPI_GRO_CB(skb)->csum_valid &&
2344 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2345 return 0;
2346
2347 NAPI_GRO_CB(skb)->csum = psum;
2348
2349 return __skb_gro_checksum_complete(skb);
2350}
2351
573e8fca
TH
2352static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2353{
662880f4
TH
2354 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2355 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2356 NAPI_GRO_CB(skb)->csum_cnt--;
2357 } else {
2358 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2359 * verified a new top level checksum or an encapsulated one
2360 * during GRO. This saves work if we fallback to normal path.
2361 */
2362 __skb_incr_checksum_unnecessary(skb);
573e8fca
TH
2363 }
2364}
2365
2366#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2367 compute_pseudo) \
2368({ \
2369 __sum16 __ret = 0; \
2370 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2371 __ret = __skb_gro_checksum_validate_complete(skb, \
2372 compute_pseudo(skb, proto)); \
5a212329
TH
2373 if (__ret) \
2374 __skb_mark_checksum_bad(skb); \
2375 else \
573e8fca
TH
2376 skb_gro_incr_csum_unnecessary(skb); \
2377 __ret; \
2378})
2379
2380#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2381 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2382
2383#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2384 compute_pseudo) \
2385 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2386
2387#define skb_gro_checksum_simple_validate(skb) \
2388 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2389
d96535a1
TH
2390static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2391{
2392 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2393 !NAPI_GRO_CB(skb)->csum_valid);
2394}
2395
2396static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2397 __sum16 check, __wsum pseudo)
2398{
2399 NAPI_GRO_CB(skb)->csum = ~pseudo;
2400 NAPI_GRO_CB(skb)->csum_valid = 1;
2401}
2402
2403#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2404do { \
2405 if (__skb_gro_checksum_convert_check(skb)) \
2406 __skb_gro_checksum_convert(skb, check, \
2407 compute_pseudo(skb, proto)); \
2408} while (0)
2409
26c4f7da
TH
2410struct gro_remcsum {
2411 int offset;
2412 __wsum delta;
2413};
2414
2415static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2416{
846cd667 2417 grc->offset = 0;
26c4f7da
TH
2418 grc->delta = 0;
2419}
2420
b7fe10e5
TH
2421static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2422 unsigned int off, size_t hdrlen,
2423 int start, int offset,
2424 struct gro_remcsum *grc,
2425 bool nopartial)
dcdc8994
TH
2426{
2427 __wsum delta;
b7fe10e5 2428 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
dcdc8994
TH
2429
2430 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2431
15e2396d 2432 if (!nopartial) {
b7fe10e5
TH
2433 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2434 return ptr;
2435 }
2436
2437 ptr = skb_gro_header_fast(skb, off);
2438 if (skb_gro_header_hard(skb, off + plen)) {
2439 ptr = skb_gro_header_slow(skb, off + plen, off);
2440 if (!ptr)
2441 return NULL;
15e2396d
TH
2442 }
2443
b7fe10e5
TH
2444 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2445 start, offset);
dcdc8994
TH
2446
2447 /* Adjust skb->csum since we changed the packet */
dcdc8994 2448 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
26c4f7da 2449
b7fe10e5 2450 grc->offset = off + hdrlen + offset;
26c4f7da 2451 grc->delta = delta;
b7fe10e5
TH
2452
2453 return ptr;
dcdc8994
TH
2454}
2455
26c4f7da
TH
2456static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2457 struct gro_remcsum *grc)
2458{
b7fe10e5
TH
2459 void *ptr;
2460 size_t plen = grc->offset + sizeof(u16);
2461
26c4f7da
TH
2462 if (!grc->delta)
2463 return;
2464
b7fe10e5
TH
2465 ptr = skb_gro_header_fast(skb, grc->offset);
2466 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2467 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2468 if (!ptr)
2469 return;
2470 }
2471
2472 remcsum_unadjust((__sum16 *)ptr, grc->delta);
26c4f7da 2473}
dcdc8994 2474
0c4e8581
SH
2475static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2476 unsigned short type,
3b04ddde 2477 const void *daddr, const void *saddr,
95c96174 2478 unsigned int len)
0c4e8581 2479{
f1ecfd5d 2480 if (!dev->header_ops || !dev->header_ops->create)
0c4e8581 2481 return 0;
3b04ddde
SH
2482
2483 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
0c4e8581
SH
2484}
2485
b95cce35
SH
2486static inline int dev_parse_header(const struct sk_buff *skb,
2487 unsigned char *haddr)
2488{
2489 const struct net_device *dev = skb->dev;
2490
1b83336b 2491 if (!dev->header_ops || !dev->header_ops->parse)
b95cce35 2492 return 0;
3b04ddde 2493 return dev->header_ops->parse(skb, haddr);
b95cce35
SH
2494}
2495
1da177e4 2496typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
f629d208 2497int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
1da177e4
LT
2498static inline int unregister_gifconf(unsigned int family)
2499{
2500 return register_gifconf(family, NULL);
2501}
2502
99bbc707 2503#ifdef CONFIG_NET_FLOW_LIMIT
5f121b9a 2504#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
99bbc707
WB
2505struct sd_flow_limit {
2506 u64 count;
2507 unsigned int num_buckets;
2508 unsigned int history_head;
2509 u16 history[FLOW_LIMIT_HISTORY];
2510 u8 buckets[];
2511};
2512
2513extern int netdev_flow_limit_table_len;
2514#endif /* CONFIG_NET_FLOW_LIMIT */
2515
1da177e4 2516/*
88751275 2517 * Incoming packets are placed on per-cpu queues
1da177e4 2518 */
d94d9fee 2519struct softnet_data {
1da177e4 2520 struct list_head poll_list;
6e7676c1 2521 struct sk_buff_head process_queue;
1da177e4 2522
dee42870 2523 /* stats */
cd7b5396
DM
2524 unsigned int processed;
2525 unsigned int time_squeeze;
2526 unsigned int cpu_collision;
2527 unsigned int received_rps;
fd793d89 2528#ifdef CONFIG_RPS
88751275 2529 struct softnet_data *rps_ipi_list;
4cdb1e2e
ED
2530#endif
2531#ifdef CONFIG_NET_FLOW_LIMIT
2532 struct sd_flow_limit __rcu *flow_limit;
2533#endif
2534 struct Qdisc *output_queue;
2535 struct Qdisc **output_queue_tailp;
2536 struct sk_buff *completion_queue;
88751275 2537
4cdb1e2e 2538#ifdef CONFIG_RPS
88751275 2539 /* Elements below can be accessed between CPUs for RPS */
0a9627f2 2540 struct call_single_data csd ____cacheline_aligned_in_smp;
88751275
ED
2541 struct softnet_data *rps_ipi_next;
2542 unsigned int cpu;
fec5e652 2543 unsigned int input_queue_head;
76cc8b13 2544 unsigned int input_queue_tail;
1e94d72f 2545#endif
95c96174 2546 unsigned int dropped;
0a9627f2 2547 struct sk_buff_head input_pkt_queue;
bea3348e 2548 struct napi_struct backlog;
99bbc707 2549
1da177e4
LT
2550};
2551
76cc8b13 2552static inline void input_queue_head_incr(struct softnet_data *sd)
fec5e652
TH
2553{
2554#ifdef CONFIG_RPS
76cc8b13
TH
2555 sd->input_queue_head++;
2556#endif
2557}
2558
2559static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2560 unsigned int *qtail)
2561{
2562#ifdef CONFIG_RPS
2563 *qtail = ++sd->input_queue_tail;
fec5e652
TH
2564#endif
2565}
2566
0a9627f2 2567DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1da177e4 2568
f629d208 2569void __netif_schedule(struct Qdisc *q);
46e5da40 2570void netif_schedule_queue(struct netdev_queue *txq);
86d804e1 2571
fd2ea0a7
DM
2572static inline void netif_tx_schedule_all(struct net_device *dev)
2573{
2574 unsigned int i;
2575
2576 for (i = 0; i < dev->num_tx_queues; i++)
2577 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2578}
2579
d29f749e
DJ
2580static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2581{
73466498 2582 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2583}
2584
bea3348e
SH
2585/**
2586 * netif_start_queue - allow transmit
2587 * @dev: network device
2588 *
2589 * Allow upper layers to call the device hard_start_xmit routine.
2590 */
1da177e4
LT
2591static inline void netif_start_queue(struct net_device *dev)
2592{
e8a0464c 2593 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2594}
2595
fd2ea0a7
DM
2596static inline void netif_tx_start_all_queues(struct net_device *dev)
2597{
2598 unsigned int i;
2599
2600 for (i = 0; i < dev->num_tx_queues; i++) {
2601 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2602 netif_tx_start_queue(txq);
2603 }
2604}
2605
46e5da40 2606void netif_tx_wake_queue(struct netdev_queue *dev_queue);
79d16385 2607
d29f749e
DJ
2608/**
2609 * netif_wake_queue - restart transmit
2610 * @dev: network device
2611 *
2612 * Allow upper layers to call the device hard_start_xmit routine.
2613 * Used for flow control when transmit resources are available.
2614 */
79d16385
DM
2615static inline void netif_wake_queue(struct net_device *dev)
2616{
e8a0464c 2617 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2618}
2619
fd2ea0a7
DM
2620static inline void netif_tx_wake_all_queues(struct net_device *dev)
2621{
2622 unsigned int i;
2623
2624 for (i = 0; i < dev->num_tx_queues; i++) {
2625 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2626 netif_tx_wake_queue(txq);
2627 }
2628}
2629
d29f749e
DJ
2630static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2631{
73466498 2632 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2633}
2634
bea3348e
SH
2635/**
2636 * netif_stop_queue - stop transmitted packets
2637 * @dev: network device
2638 *
2639 * Stop upper layers calling the device hard_start_xmit routine.
2640 * Used for flow control when transmit resources are unavailable.
2641 */
1da177e4
LT
2642static inline void netif_stop_queue(struct net_device *dev)
2643{
e8a0464c 2644 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2645}
2646
a2029240 2647void netif_tx_stop_all_queues(struct net_device *dev);
fd2ea0a7 2648
4d29515f 2649static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
d29f749e 2650{
73466498 2651 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
d29f749e
DJ
2652}
2653
bea3348e
SH
2654/**
2655 * netif_queue_stopped - test if transmit queue is flowblocked
2656 * @dev: network device
2657 *
2658 * Test if transmit queue on device is currently unable to send.
2659 */
4d29515f 2660static inline bool netif_queue_stopped(const struct net_device *dev)
1da177e4 2661{
e8a0464c 2662 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1da177e4
LT
2663}
2664
4d29515f 2665static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
c3f26a26 2666{
73466498
TH
2667 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2668}
2669
8e2f1a63
DB
2670static inline bool
2671netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
73466498
TH
2672{
2673 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2674}
2675
8e2f1a63
DB
2676static inline bool
2677netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2678{
2679 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2680}
2681
53511453
ED
2682/**
2683 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2684 * @dev_queue: pointer to transmit queue
2685 *
2686 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2687 * to give appropriate hint to the cpu.
2688 */
2689static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2690{
2691#ifdef CONFIG_BQL
2692 prefetchw(&dev_queue->dql.num_queued);
2693#endif
2694}
2695
2696/**
2697 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2698 * @dev_queue: pointer to transmit queue
2699 *
2700 * BQL enabled drivers might use this helper in their TX completion path,
2701 * to give appropriate hint to the cpu.
2702 */
2703static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2704{
2705#ifdef CONFIG_BQL
2706 prefetchw(&dev_queue->dql.limit);
2707#endif
2708}
2709
c5d67bd7
TH
2710static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2711 unsigned int bytes)
2712{
114cf580
TH
2713#ifdef CONFIG_BQL
2714 dql_queued(&dev_queue->dql, bytes);
b37c0fbe
AD
2715
2716 if (likely(dql_avail(&dev_queue->dql) >= 0))
2717 return;
2718
2719 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2720
2721 /*
2722 * The XOFF flag must be set before checking the dql_avail below,
2723 * because in netdev_tx_completed_queue we update the dql_completed
2724 * before checking the XOFF flag.
2725 */
2726 smp_mb();
2727
2728 /* check again in case another CPU has just made room avail */
2729 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2730 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
114cf580 2731#endif
c5d67bd7
TH
2732}
2733
0042d0c8
FF
2734/**
2735 * netdev_sent_queue - report the number of bytes queued to hardware
2736 * @dev: network device
2737 * @bytes: number of bytes queued to the hardware device queue
2738 *
2739 * Report the number of bytes queued for sending/completion to the network
2740 * device hardware queue. @bytes should be a good approximation and should
2741 * exactly match netdev_completed_queue() @bytes
2742 */
c5d67bd7
TH
2743static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2744{
2745 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2746}
2747
2748static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
95c96174 2749 unsigned int pkts, unsigned int bytes)
c5d67bd7 2750{
114cf580 2751#ifdef CONFIG_BQL
b37c0fbe
AD
2752 if (unlikely(!bytes))
2753 return;
2754
2755 dql_completed(&dev_queue->dql, bytes);
2756
2757 /*
2758 * Without the memory barrier there is a small possiblity that
2759 * netdev_tx_sent_queue will miss the update and cause the queue to
2760 * be stopped forever
2761 */
2762 smp_mb();
2763
2764 if (dql_avail(&dev_queue->dql) < 0)
2765 return;
2766
2767 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2768 netif_schedule_queue(dev_queue);
114cf580 2769#endif
c5d67bd7
TH
2770}
2771
0042d0c8
FF
2772/**
2773 * netdev_completed_queue - report bytes and packets completed by device
2774 * @dev: network device
2775 * @pkts: actual number of packets sent over the medium
2776 * @bytes: actual number of bytes sent over the medium
2777 *
2778 * Report the number of bytes and packets transmitted by the network device
2779 * hardware queue over the physical medium, @bytes must exactly match the
2780 * @bytes amount passed to netdev_sent_queue()
2781 */
c5d67bd7 2782static inline void netdev_completed_queue(struct net_device *dev,
95c96174 2783 unsigned int pkts, unsigned int bytes)
c5d67bd7
TH
2784{
2785 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2786}
2787
2788static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2789{
114cf580 2790#ifdef CONFIG_BQL
5c490354 2791 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
114cf580
TH
2792 dql_reset(&q->dql);
2793#endif
c5d67bd7
TH
2794}
2795
0042d0c8
FF
2796/**
2797 * netdev_reset_queue - reset the packets and bytes count of a network device
2798 * @dev_queue: network device
2799 *
2800 * Reset the bytes and packet count of a network device and clear the
2801 * software flow control OFF bit for this network device
2802 */
c5d67bd7
TH
2803static inline void netdev_reset_queue(struct net_device *dev_queue)
2804{
2805 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
c3f26a26
DM
2806}
2807
b9507bda
DB
2808/**
2809 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2810 * @dev: network device
2811 * @queue_index: given tx queue index
2812 *
2813 * Returns 0 if given tx queue index >= number of device tx queues,
2814 * otherwise returns the originally passed tx queue index.
2815 */
2816static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2817{
2818 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2819 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2820 dev->name, queue_index,
2821 dev->real_num_tx_queues);
2822 return 0;
2823 }
2824
2825 return queue_index;
2826}
2827
bea3348e
SH
2828/**
2829 * netif_running - test if up
2830 * @dev: network device
2831 *
2832 * Test if the device has been brought up.
2833 */
4d29515f 2834static inline bool netif_running(const struct net_device *dev)
1da177e4
LT
2835{
2836 return test_bit(__LINK_STATE_START, &dev->state);
2837}
2838
f25f4e44
PWJ
2839/*
2840 * Routines to manage the subqueues on a device. We only need start
2841 * stop, and a check if it's stopped. All other device management is
2842 * done at the overall netdevice level.
2843 * Also test the device if we're multiqueue.
2844 */
bea3348e
SH
2845
2846/**
2847 * netif_start_subqueue - allow sending packets on subqueue
2848 * @dev: network device
2849 * @queue_index: sub queue index
2850 *
2851 * Start individual transmit queue of a device with multiple transmit queues.
2852 */
f25f4e44
PWJ
2853static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2854{
fd2ea0a7 2855 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2856
2857 netif_tx_start_queue(txq);
f25f4e44
PWJ
2858}
2859
bea3348e
SH
2860/**
2861 * netif_stop_subqueue - stop sending packets on subqueue
2862 * @dev: network device
2863 * @queue_index: sub queue index
2864 *
2865 * Stop individual transmit queue of a device with multiple transmit queues.
2866 */
f25f4e44
PWJ
2867static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2868{
fd2ea0a7 2869 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f 2870 netif_tx_stop_queue(txq);
f25f4e44
PWJ
2871}
2872
bea3348e
SH
2873/**
2874 * netif_subqueue_stopped - test status of subqueue
2875 * @dev: network device
2876 * @queue_index: sub queue index
2877 *
2878 * Check individual transmit queue of a device with multiple transmit queues.
2879 */
4d29515f
DM
2880static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2881 u16 queue_index)
f25f4e44 2882{
fd2ea0a7 2883 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
7b3d3e4f
KK
2884
2885 return netif_tx_queue_stopped(txq);
f25f4e44
PWJ
2886}
2887
4d29515f
DM
2888static inline bool netif_subqueue_stopped(const struct net_device *dev,
2889 struct sk_buff *skb)
668f895a
PE
2890{
2891 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2892}
bea3348e 2893
46e5da40 2894void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
f25f4e44 2895
537c00de 2896#ifdef CONFIG_XPS
53af53ae 2897int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
f629d208 2898 u16 index);
537c00de
AD
2899#else
2900static inline int netif_set_xps_queue(struct net_device *dev,
3573540c 2901 const struct cpumask *mask,
537c00de
AD
2902 u16 index)
2903{
2904 return 0;
2905}
2906#endif
2907
5605c762
JP
2908u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2909 unsigned int num_tx_queues);
2910
a3d22a68
VZ
2911/*
2912 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2913 * as a distribution range limit for the returned value.
2914 */
2915static inline u16 skb_tx_hash(const struct net_device *dev,
0e001614 2916 struct sk_buff *skb)
a3d22a68
VZ
2917{
2918 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2919}
2920
bea3348e
SH
2921/**
2922 * netif_is_multiqueue - test if device has multiple transmit queues
2923 * @dev: network device
2924 *
2925 * Check if device has multiple transmit queues
bea3348e 2926 */
4d29515f 2927static inline bool netif_is_multiqueue(const struct net_device *dev)
f25f4e44 2928{
a02cec21 2929 return dev->num_tx_queues > 1;
f25f4e44 2930}
1da177e4 2931
f629d208 2932int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
f0796d5c 2933
a953be53 2934#ifdef CONFIG_SYSFS
f629d208 2935int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
62fe0b40
BH
2936#else
2937static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2938 unsigned int rxq)
2939{
2940 return 0;
2941}
2942#endif
2943
a953be53
MD
2944#ifdef CONFIG_SYSFS
2945static inline unsigned int get_netdev_rx_queue_index(
2946 struct netdev_rx_queue *queue)
2947{
2948 struct net_device *dev = queue->dev;
2949 int index = queue - dev->_rx;
2950
2951 BUG_ON(index >= dev->num_rx_queues);
2952 return index;
2953}
2954#endif
2955
16917b87 2956#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
f629d208 2957int netif_get_num_default_rss_queues(void);
16917b87 2958
e6247027
ED
2959enum skb_free_reason {
2960 SKB_REASON_CONSUMED,
2961 SKB_REASON_DROPPED,
2962};
2963
2964void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2965void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
1da177e4 2966
e6247027
ED
2967/*
2968 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2969 * interrupt context or with hardware interrupts being disabled.
2970 * (in_irq() || irqs_disabled())
2971 *
2972 * We provide four helpers that can be used in following contexts :
2973 *
2974 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2975 * replacing kfree_skb(skb)
2976 *
2977 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2978 * Typically used in place of consume_skb(skb) in TX completion path
2979 *
2980 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2981 * replacing kfree_skb(skb)
2982 *
2983 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2984 * and consumed a packet. Used in place of consume_skb(skb)
1da177e4 2985 */
e6247027
ED
2986static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2987{
2988 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2989}
2990
2991static inline void dev_consume_skb_irq(struct sk_buff *skb)
2992{
2993 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2994}
2995
2996static inline void dev_kfree_skb_any(struct sk_buff *skb)
2997{
2998 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2999}
3000
3001static inline void dev_consume_skb_any(struct sk_buff *skb)
3002{
3003 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3004}
1da177e4 3005
f629d208
JP
3006int netif_rx(struct sk_buff *skb);
3007int netif_rx_ni(struct sk_buff *skb);
04eb4489 3008int netif_receive_skb(struct sk_buff *skb);
f629d208
JP
3009gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3010void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3011struct sk_buff *napi_get_frags(struct napi_struct *napi);
3012gro_result_t napi_gro_frags(struct napi_struct *napi);
bf5a755f
JC
3013struct packet_offload *gro_find_receive_by_type(__be16 type);
3014struct packet_offload *gro_find_complete_by_type(__be16 type);
76620aaf
HX
3015
3016static inline void napi_free_frags(struct napi_struct *napi)
3017{
3018 kfree_skb(napi->skb);
3019 napi->skb = NULL;
3020}
3021
f629d208
JP
3022int netdev_rx_handler_register(struct net_device *dev,
3023 rx_handler_func_t *rx_handler,
3024 void *rx_handler_data);
3025void netdev_rx_handler_unregister(struct net_device *dev);
3026
3027bool dev_valid_name(const char *name);
3028int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3029int dev_ethtool(struct net *net, struct ifreq *);
3030unsigned int dev_get_flags(const struct net_device *);
3031int __dev_change_flags(struct net_device *, unsigned int flags);
3032int dev_change_flags(struct net_device *, unsigned int);
cb178190
DM
3033void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3034 unsigned int gchanges);
f629d208
JP
3035int dev_change_name(struct net_device *, const char *);
3036int dev_set_alias(struct net_device *, const char *, size_t);
3037int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3038int dev_set_mtu(struct net_device *, int);
3039void dev_set_group(struct net_device *, int);
3040int dev_set_mac_address(struct net_device *, struct sockaddr *);
3041int dev_change_carrier(struct net_device *, bool new_carrier);
3042int dev_get_phys_port_id(struct net_device *dev,
02637fce 3043 struct netdev_phys_item_id *ppid);
db24a904
DA
3044int dev_get_phys_port_name(struct net_device *dev,
3045 char *name, size_t len);
d746d707 3046int dev_change_proto_down(struct net_device *dev, bool proto_down);
55a93b3e 3047struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
ce93718f
DM
3048struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3049 struct netdev_queue *txq, int *ret);
a0265d28 3050int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
f629d208 3051int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
1ee481fb 3052bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
1da177e4 3053
20380731 3054extern int netdev_budget;
1da177e4
LT
3055
3056/* Called by rtnetlink.c:rtnl_unlock() */
f629d208 3057void netdev_run_todo(void);
1da177e4 3058
bea3348e
SH
3059/**
3060 * dev_put - release reference to device
3061 * @dev: network device
3062 *
9ef4429b 3063 * Release reference to device to allow it to be freed.
bea3348e 3064 */
1da177e4
LT
3065static inline void dev_put(struct net_device *dev)
3066{
933393f5 3067 this_cpu_dec(*dev->pcpu_refcnt);
1da177e4
LT
3068}
3069
bea3348e
SH
3070/**
3071 * dev_hold - get reference to device
3072 * @dev: network device
3073 *
9ef4429b 3074 * Hold reference to device to keep it from being freed.
bea3348e 3075 */
15333061
SH
3076static inline void dev_hold(struct net_device *dev)
3077{
933393f5 3078 this_cpu_inc(*dev->pcpu_refcnt);
15333061 3079}
1da177e4
LT
3080
3081/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3082 * and _off may be called from IRQ context, but it is caller
3083 * who is responsible for serialization of these calls.
b00055aa
SR
3084 *
3085 * The name carrier is inappropriate, these functions should really be
3086 * called netif_lowerlayer_*() because they represent the state of any
3087 * kind of lower layer not just hardware media.
1da177e4
LT
3088 */
3089
f629d208
JP
3090void linkwatch_init_dev(struct net_device *dev);
3091void linkwatch_fire_event(struct net_device *dev);
3092void linkwatch_forget_dev(struct net_device *dev);
1da177e4 3093
bea3348e
SH
3094/**
3095 * netif_carrier_ok - test if carrier present
3096 * @dev: network device
3097 *
3098 * Check if carrier is present on device
3099 */
4d29515f 3100static inline bool netif_carrier_ok(const struct net_device *dev)
1da177e4
LT
3101{
3102 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3103}
3104
f629d208 3105unsigned long dev_trans_start(struct net_device *dev);
9d21493b 3106
f629d208 3107void __netdev_watchdog_up(struct net_device *dev);
1da177e4 3108
f629d208 3109void netif_carrier_on(struct net_device *dev);
1da177e4 3110
f629d208 3111void netif_carrier_off(struct net_device *dev);
1da177e4 3112
bea3348e
SH
3113/**
3114 * netif_dormant_on - mark device as dormant.
3115 * @dev: network device
3116 *
3117 * Mark device as dormant (as per RFC2863).
3118 *
3119 * The dormant state indicates that the relevant interface is not
3120 * actually in a condition to pass packets (i.e., it is not 'up') but is
3121 * in a "pending" state, waiting for some external event. For "on-
3122 * demand" interfaces, this new state identifies the situation where the
3123 * interface is waiting for events to place it in the up state.
3124 *
3125 */
b00055aa
SR
3126static inline void netif_dormant_on(struct net_device *dev)
3127{
3128 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3129 linkwatch_fire_event(dev);
3130}
3131
bea3348e
SH
3132/**
3133 * netif_dormant_off - set device as not dormant.
3134 * @dev: network device
3135 *
3136 * Device is not in dormant state.
3137 */
b00055aa
SR
3138static inline void netif_dormant_off(struct net_device *dev)
3139{
3140 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3141 linkwatch_fire_event(dev);
3142}
3143
bea3348e
SH
3144/**
3145 * netif_dormant - test if carrier present
3146 * @dev: network device
3147 *
3148 * Check if carrier is present on device
3149 */
4d29515f 3150static inline bool netif_dormant(const struct net_device *dev)
b00055aa
SR
3151{
3152 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3153}
3154
3155
bea3348e
SH
3156/**
3157 * netif_oper_up - test if device is operational
3158 * @dev: network device
3159 *
3160 * Check if carrier is operational
3161 */
4d29515f 3162static inline bool netif_oper_up(const struct net_device *dev)
d94d9fee 3163{
b00055aa
SR
3164 return (dev->operstate == IF_OPER_UP ||
3165 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3166}
3167
bea3348e
SH
3168/**
3169 * netif_device_present - is device available or removed
3170 * @dev: network device
3171 *
3172 * Check if device has not been removed from system.
3173 */
4d29515f 3174static inline bool netif_device_present(struct net_device *dev)
1da177e4
LT
3175{
3176 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3177}
3178
f629d208 3179void netif_device_detach(struct net_device *dev);
1da177e4 3180
f629d208 3181void netif_device_attach(struct net_device *dev);
1da177e4
LT
3182
3183/*
3184 * Network interface message level settings
3185 */
1da177e4
LT
3186
3187enum {
3188 NETIF_MSG_DRV = 0x0001,
3189 NETIF_MSG_PROBE = 0x0002,
3190 NETIF_MSG_LINK = 0x0004,
3191 NETIF_MSG_TIMER = 0x0008,
3192 NETIF_MSG_IFDOWN = 0x0010,
3193 NETIF_MSG_IFUP = 0x0020,
3194 NETIF_MSG_RX_ERR = 0x0040,
3195 NETIF_MSG_TX_ERR = 0x0080,
3196 NETIF_MSG_TX_QUEUED = 0x0100,
3197 NETIF_MSG_INTR = 0x0200,
3198 NETIF_MSG_TX_DONE = 0x0400,
3199 NETIF_MSG_RX_STATUS = 0x0800,
3200 NETIF_MSG_PKTDATA = 0x1000,
3201 NETIF_MSG_HW = 0x2000,
3202 NETIF_MSG_WOL = 0x4000,
3203};
3204
3205#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3206#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3207#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3208#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3209#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3210#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3211#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3212#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3213#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3214#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3215#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3216#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3217#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3218#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3219#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3220
3221static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3222{
3223 /* use default */
3224 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3225 return default_msg_enable_bits;
3226 if (debug_value == 0) /* no output */
3227 return 0;
3228 /* set low N bits */
3229 return (1 << debug_value) - 1;
3230}
3231
c773e847 3232static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
932ff279 3233{
c773e847
DM
3234 spin_lock(&txq->_xmit_lock);
3235 txq->xmit_lock_owner = cpu;
22dd7495
JHS
3236}
3237
fd2ea0a7
DM
3238static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3239{
3240 spin_lock_bh(&txq->_xmit_lock);
3241 txq->xmit_lock_owner = smp_processor_id();
3242}
3243
4d29515f 3244static inline bool __netif_tx_trylock(struct netdev_queue *txq)
c3f26a26 3245{
4d29515f 3246 bool ok = spin_trylock(&txq->_xmit_lock);
c3f26a26
DM
3247 if (likely(ok))
3248 txq->xmit_lock_owner = smp_processor_id();
3249 return ok;
3250}
3251
3252static inline void __netif_tx_unlock(struct netdev_queue *txq)
3253{
3254 txq->xmit_lock_owner = -1;
3255 spin_unlock(&txq->_xmit_lock);
3256}
3257
3258static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3259{
3260 txq->xmit_lock_owner = -1;
3261 spin_unlock_bh(&txq->_xmit_lock);
3262}
3263
08baf561
ED
3264static inline void txq_trans_update(struct netdev_queue *txq)
3265{
3266 if (txq->xmit_lock_owner != -1)
3267 txq->trans_start = jiffies;
3268}
3269
d29f749e
DJ
3270/**
3271 * netif_tx_lock - grab network device transmit lock
3272 * @dev: network device
d29f749e
DJ
3273 *
3274 * Get network device transmit lock
3275 */
22dd7495
JHS
3276static inline void netif_tx_lock(struct net_device *dev)
3277{
e8a0464c 3278 unsigned int i;
c3f26a26 3279 int cpu;
c773e847 3280
c3f26a26
DM
3281 spin_lock(&dev->tx_global_lock);
3282 cpu = smp_processor_id();
e8a0464c
DM
3283 for (i = 0; i < dev->num_tx_queues; i++) {
3284 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3285
3286 /* We are the only thread of execution doing a
3287 * freeze, but we have to grab the _xmit_lock in
3288 * order to synchronize with threads which are in
3289 * the ->hard_start_xmit() handler and already
3290 * checked the frozen bit.
3291 */
e8a0464c 3292 __netif_tx_lock(txq, cpu);
c3f26a26
DM
3293 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3294 __netif_tx_unlock(txq);
e8a0464c 3295 }
932ff279
HX
3296}
3297
3298static inline void netif_tx_lock_bh(struct net_device *dev)
3299{
e8a0464c
DM
3300 local_bh_disable();
3301 netif_tx_lock(dev);
932ff279
HX
3302}
3303
932ff279
HX
3304static inline void netif_tx_unlock(struct net_device *dev)
3305{
e8a0464c
DM
3306 unsigned int i;
3307
3308 for (i = 0; i < dev->num_tx_queues; i++) {
3309 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c773e847 3310
c3f26a26
DM
3311 /* No need to grab the _xmit_lock here. If the
3312 * queue is not stopped for another reason, we
3313 * force a schedule.
3314 */
3315 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
7b3d3e4f 3316 netif_schedule_queue(txq);
c3f26a26
DM
3317 }
3318 spin_unlock(&dev->tx_global_lock);
932ff279
HX
3319}
3320
3321static inline void netif_tx_unlock_bh(struct net_device *dev)
3322{
e8a0464c
DM
3323 netif_tx_unlock(dev);
3324 local_bh_enable();
932ff279
HX
3325}
3326
c773e847 3327#define HARD_TX_LOCK(dev, txq, cpu) { \
22dd7495 3328 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3329 __netif_tx_lock(txq, cpu); \
22dd7495
JHS
3330 } \
3331}
3332
5efeac44
EB
3333#define HARD_TX_TRYLOCK(dev, txq) \
3334 (((dev->features & NETIF_F_LLTX) == 0) ? \
3335 __netif_tx_trylock(txq) : \
3336 true )
3337
c773e847 3338#define HARD_TX_UNLOCK(dev, txq) { \
22dd7495 3339 if ((dev->features & NETIF_F_LLTX) == 0) { \
c773e847 3340 __netif_tx_unlock(txq); \
22dd7495
JHS
3341 } \
3342}
3343
1da177e4
LT
3344static inline void netif_tx_disable(struct net_device *dev)
3345{
fd2ea0a7 3346 unsigned int i;
c3f26a26 3347 int cpu;
fd2ea0a7 3348
c3f26a26
DM
3349 local_bh_disable();
3350 cpu = smp_processor_id();
fd2ea0a7
DM
3351 for (i = 0; i < dev->num_tx_queues; i++) {
3352 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
c3f26a26
DM
3353
3354 __netif_tx_lock(txq, cpu);
fd2ea0a7 3355 netif_tx_stop_queue(txq);
c3f26a26 3356 __netif_tx_unlock(txq);
fd2ea0a7 3357 }
c3f26a26 3358 local_bh_enable();
1da177e4
LT
3359}
3360
e308a5d8
DM
3361static inline void netif_addr_lock(struct net_device *dev)
3362{
3363 spin_lock(&dev->addr_list_lock);
3364}
3365
2429f7ac
JP
3366static inline void netif_addr_lock_nested(struct net_device *dev)
3367{
25175ba5
VY
3368 int subclass = SINGLE_DEPTH_NESTING;
3369
3370 if (dev->netdev_ops->ndo_get_lock_subclass)
3371 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3372
3373 spin_lock_nested(&dev->addr_list_lock, subclass);
2429f7ac
JP
3374}
3375
e308a5d8
DM
3376static inline void netif_addr_lock_bh(struct net_device *dev)
3377{
3378 spin_lock_bh(&dev->addr_list_lock);
3379}
3380
3381static inline void netif_addr_unlock(struct net_device *dev)
3382{
3383 spin_unlock(&dev->addr_list_lock);
3384}
3385
3386static inline void netif_addr_unlock_bh(struct net_device *dev)
3387{
3388 spin_unlock_bh(&dev->addr_list_lock);
3389}
3390
f001fde5 3391/*
31278e71 3392 * dev_addrs walker. Should be used only for read access. Call with
f001fde5
JP
3393 * rcu_read_lock held.
3394 */
3395#define for_each_dev_addr(dev, ha) \
31278e71 3396 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
f001fde5 3397
1da177e4
LT
3398/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3399
f629d208 3400void ether_setup(struct net_device *dev);
1da177e4
LT
3401
3402/* Support for loadable net-drivers */
f629d208 3403struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
c835a677 3404 unsigned char name_assign_type,
f629d208
JP
3405 void (*setup)(struct net_device *),
3406 unsigned int txqs, unsigned int rxqs);
c835a677
TG
3407#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3408 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
36909ea4 3409
c835a677
TG
3410#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3411 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3412 count)
36909ea4 3413
f629d208
JP
3414int register_netdev(struct net_device *dev);
3415void unregister_netdev(struct net_device *dev);
f001fde5 3416
22bedad3 3417/* General hardware address lists handling functions */
f629d208
JP
3418int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3419 struct netdev_hw_addr_list *from_list, int addr_len);
3420void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3421 struct netdev_hw_addr_list *from_list, int addr_len);
670e5b8e
AD
3422int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3423 struct net_device *dev,
3424 int (*sync)(struct net_device *, const unsigned char *),
3425 int (*unsync)(struct net_device *,
3426 const unsigned char *));
3427void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3428 struct net_device *dev,
3429 int (*unsync)(struct net_device *,
3430 const unsigned char *));
f629d208 3431void __hw_addr_init(struct netdev_hw_addr_list *list);
22bedad3 3432
f001fde5 3433/* Functions used for device addresses handling */
f629d208
JP
3434int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3435 unsigned char addr_type);
3436int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3437 unsigned char addr_type);
f629d208
JP
3438void dev_addr_flush(struct net_device *dev);
3439int dev_addr_init(struct net_device *dev);
a748ee24
JP
3440
3441/* Functions used for unicast addresses handling */
f629d208
JP
3442int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3443int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3444int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3445int dev_uc_sync(struct net_device *to, struct net_device *from);
3446int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3447void dev_uc_unsync(struct net_device *to, struct net_device *from);
3448void dev_uc_flush(struct net_device *dev);
3449void dev_uc_init(struct net_device *dev);
f001fde5 3450
670e5b8e
AD
3451/**
3452 * __dev_uc_sync - Synchonize device's unicast list
3453 * @dev: device to sync
3454 * @sync: function to call if address should be added
3455 * @unsync: function to call if address should be removed
3456 *
3457 * Add newly added addresses to the interface, and release
3458 * addresses that have been deleted.
3459 **/
3460static inline int __dev_uc_sync(struct net_device *dev,
3461 int (*sync)(struct net_device *,
3462 const unsigned char *),
3463 int (*unsync)(struct net_device *,
3464 const unsigned char *))
3465{
3466 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3467}
3468
3469/**
e793c0f7 3470 * __dev_uc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3471 * @dev: device to sync
3472 * @unsync: function to call if address should be removed
3473 *
3474 * Remove all addresses that were added to the device by dev_uc_sync().
3475 **/
3476static inline void __dev_uc_unsync(struct net_device *dev,
3477 int (*unsync)(struct net_device *,
3478 const unsigned char *))
3479{
3480 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3481}
3482
22bedad3 3483/* Functions used for multicast addresses handling */
f629d208
JP
3484int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3485int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3486int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3487int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3488int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3489int dev_mc_sync(struct net_device *to, struct net_device *from);
3490int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3491void dev_mc_unsync(struct net_device *to, struct net_device *from);
3492void dev_mc_flush(struct net_device *dev);
3493void dev_mc_init(struct net_device *dev);
f001fde5 3494
670e5b8e
AD
3495/**
3496 * __dev_mc_sync - Synchonize device's multicast list
3497 * @dev: device to sync
3498 * @sync: function to call if address should be added
3499 * @unsync: function to call if address should be removed
3500 *
3501 * Add newly added addresses to the interface, and release
3502 * addresses that have been deleted.
3503 **/
3504static inline int __dev_mc_sync(struct net_device *dev,
3505 int (*sync)(struct net_device *,
3506 const unsigned char *),
3507 int (*unsync)(struct net_device *,
3508 const unsigned char *))
3509{
3510 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3511}
3512
3513/**
e793c0f7 3514 * __dev_mc_unsync - Remove synchronized addresses from device
670e5b8e
AD
3515 * @dev: device to sync
3516 * @unsync: function to call if address should be removed
3517 *
3518 * Remove all addresses that were added to the device by dev_mc_sync().
3519 **/
3520static inline void __dev_mc_unsync(struct net_device *dev,
3521 int (*unsync)(struct net_device *,
3522 const unsigned char *))
3523{
3524 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3525}
3526
4417da66 3527/* Functions used for secondary unicast and multicast support */
f629d208
JP
3528void dev_set_rx_mode(struct net_device *dev);
3529void __dev_set_rx_mode(struct net_device *dev);
3530int dev_set_promiscuity(struct net_device *dev, int inc);
3531int dev_set_allmulti(struct net_device *dev, int inc);
3532void netdev_state_change(struct net_device *dev);
3533void netdev_notify_peers(struct net_device *dev);
3534void netdev_features_change(struct net_device *dev);
1da177e4 3535/* Load a device via the kmod */
f629d208
JP
3536void dev_load(struct net *net, const char *name);
3537struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3538 struct rtnl_link_stats64 *storage);
3539void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3540 const struct net_device_stats *netdev_stats);
eeda3fd6 3541
1da177e4 3542extern int netdev_max_backlog;
3b098e2d 3543extern int netdev_tstamp_prequeue;
1da177e4 3544extern int weight_p;
0a14842f 3545extern int bpf_jit_enable;
9ff162a8 3546
f629d208 3547bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
44a40855
VY
3548struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3549 struct list_head **iter);
f629d208
JP
3550struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3551 struct list_head **iter);
8b5be856 3552
44a40855
VY
3553/* iterate through upper list, must be called under RCU read lock */
3554#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3555 for (iter = &(dev)->adj_list.upper, \
3556 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3557 updev; \
3558 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3559
8b5be856 3560/* iterate through upper list, must be called under RCU read lock */
2f268f12
VF
3561#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3562 for (iter = &(dev)->all_adj_list.upper, \
3563 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3564 updev; \
3565 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
8b5be856 3566
f629d208
JP
3567void *netdev_lower_get_next_private(struct net_device *dev,
3568 struct list_head **iter);
3569void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3570 struct list_head **iter);
31088a11
VF
3571
3572#define netdev_for_each_lower_private(dev, priv, iter) \
3573 for (iter = (dev)->adj_list.lower.next, \
3574 priv = netdev_lower_get_next_private(dev, &(iter)); \
3575 priv; \
3576 priv = netdev_lower_get_next_private(dev, &(iter)))
3577
3578#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3579 for (iter = &(dev)->adj_list.lower, \
3580 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3581 priv; \
3582 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3583
4085ebe8
VY
3584void *netdev_lower_get_next(struct net_device *dev,
3585 struct list_head **iter);
3586#define netdev_for_each_lower_dev(dev, ldev, iter) \
3587 for (iter = &(dev)->adj_list.lower, \
3588 ldev = netdev_lower_get_next(dev, &(iter)); \
3589 ldev; \
3590 ldev = netdev_lower_get_next(dev, &(iter)))
3591
f629d208 3592void *netdev_adjacent_get_private(struct list_head *adj_list);
e001bfad 3593void *netdev_lower_get_first_private_rcu(struct net_device *dev);
f629d208
JP
3594struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3595struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3596int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3597int netdev_master_upper_dev_link(struct net_device *dev,
9ff162a8 3598 struct net_device *upper_dev);
f629d208
JP
3599int netdev_master_upper_dev_link_private(struct net_device *dev,
3600 struct net_device *upper_dev,
3601 void *private);
3602void netdev_upper_dev_unlink(struct net_device *dev,
3603 struct net_device *upper_dev);
5bb025fa 3604void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
f629d208
JP
3605void *netdev_lower_dev_get_private(struct net_device *dev,
3606 struct net_device *lower_dev);
960fb622
ED
3607
3608/* RSS keys are 40 or 52 bytes long */
3609#define NETDEV_RSS_KEY_LEN 52
3610extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3611void netdev_rss_key_fill(void *buffer, size_t len);
3612
4085ebe8
VY
3613int dev_get_nest_level(struct net_device *dev,
3614 bool (*type_check)(struct net_device *dev));
f629d208
JP
3615int skb_checksum_help(struct sk_buff *skb);
3616struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3617 netdev_features_t features, bool tx_path);
3618struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3619 netdev_features_t features);
12b0004d 3620
61bd3857
MS
3621struct netdev_bonding_info {
3622 ifslave slave;
3623 ifbond master;
3624};
3625
3626struct netdev_notifier_bonding_info {
3627 struct netdev_notifier_info info; /* must be first */
3628 struct netdev_bonding_info bonding_info;
3629};
3630
3631void netdev_bonding_info_change(struct net_device *dev,
3632 struct netdev_bonding_info *bonding_info);
3633
12b0004d
CW
3634static inline
3635struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3636{
3637 return __skb_gso_segment(skb, features, true);
3638}
53d6471c 3639__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
ec5f0615
PS
3640
3641static inline bool can_checksum_protocol(netdev_features_t features,
3642 __be16 protocol)
3643{
3644 return ((features & NETIF_F_GEN_CSUM) ||
3645 ((features & NETIF_F_V4_CSUM) &&
3646 protocol == htons(ETH_P_IP)) ||
3647 ((features & NETIF_F_V6_CSUM) &&
3648 protocol == htons(ETH_P_IPV6)) ||
3649 ((features & NETIF_F_FCOE_CRC) &&
3650 protocol == htons(ETH_P_FCOE)));
3651}
12b0004d 3652
fb286bb2 3653#ifdef CONFIG_BUG
f629d208 3654void netdev_rx_csum_fault(struct net_device *dev);
fb286bb2
HX
3655#else
3656static inline void netdev_rx_csum_fault(struct net_device *dev)
3657{
3658}
3659#endif
1da177e4 3660/* rx skb timestamps */
f629d208
JP
3661void net_enable_timestamp(void);
3662void net_disable_timestamp(void);
1da177e4 3663
20380731 3664#ifdef CONFIG_PROC_FS
f629d208 3665int __init dev_proc_init(void);
900ff8c6
CW
3666#else
3667#define dev_proc_init() 0
20380731
ACM
3668#endif
3669
4798248e 3670static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
fa2dbdc2
DM
3671 struct sk_buff *skb, struct net_device *dev,
3672 bool more)
4798248e 3673{
fa2dbdc2 3674 skb->xmit_more = more ? 1 : 0;
0b725a2c 3675 return ops->ndo_start_xmit(skb, dev);
4798248e
DM
3676}
3677
10b3ad8c 3678static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
fa2dbdc2 3679 struct netdev_queue *txq, bool more)
4798248e
DM
3680{
3681 const struct net_device_ops *ops = dev->netdev_ops;
10b3ad8c 3682 int rc;
4798248e 3683
fa2dbdc2 3684 rc = __netdev_start_xmit(ops, skb, dev, more);
10b3ad8c
DM
3685 if (rc == NETDEV_TX_OK)
3686 txq_trans_update(txq);
3687
3688 return rc;
4798248e
DM
3689}
3690
42a2d923
LT
3691int netdev_class_create_file_ns(struct class_attribute *class_attr,
3692 const void *ns);
3693void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3694 const void *ns);
58292cbe
TH
3695
3696static inline int netdev_class_create_file(struct class_attribute *class_attr)
3697{
3698 return netdev_class_create_file_ns(class_attr, NULL);
3699}
3700
3701static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3702{
3703 netdev_class_remove_file_ns(class_attr, NULL);
3704}
b8a9787e 3705
04600794
JB
3706extern struct kobj_ns_type_operations net_ns_type_operations;
3707
f629d208 3708const char *netdev_drivername(const struct net_device *dev);
6579e57b 3709
f629d208 3710void linkwatch_run_queue(void);
20380731 3711
da08143b
MK
3712static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3713 netdev_features_t f2)
3714{
3715 if (f1 & NETIF_F_GEN_CSUM)
3716 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3717 if (f2 & NETIF_F_GEN_CSUM)
3718 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3719 f1 &= f2;
3720 if (f1 & NETIF_F_GEN_CSUM)
3721 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3722
3723 return f1;
3724}
3725
c8f44aff
MM
3726static inline netdev_features_t netdev_get_wanted_features(
3727 struct net_device *dev)
5455c699
MM
3728{
3729 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3730}
c8f44aff
MM
3731netdev_features_t netdev_increment_features(netdev_features_t all,
3732 netdev_features_t one, netdev_features_t mask);
b0ce3508
ED
3733
3734/* Allow TSO being used on stacked device :
3735 * Performing the GSO segmentation before last device
3736 * is a performance improvement.
3737 */
3738static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3739 netdev_features_t mask)
3740{
3741 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3742}
3743
6cb6a27c 3744int __netdev_update_features(struct net_device *dev);
5455c699 3745void netdev_update_features(struct net_device *dev);
afe12cc8 3746void netdev_change_features(struct net_device *dev);
7f353bf2 3747
fc4a7489
PM
3748void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3749 struct net_device *dev);
3750
e38f3025
TM
3751netdev_features_t passthru_features_check(struct sk_buff *skb,
3752 struct net_device *dev,
3753 netdev_features_t features);
c1e756bf 3754netdev_features_t netif_skb_features(struct sk_buff *skb);
58e998c6 3755
4d29515f 3756static inline bool net_gso_ok(netdev_features_t features, int gso_type)
576a30eb 3757{
c8f44aff 3758 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
0345e186
MM
3759
3760 /* check flags correspondence */
3761 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3762 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3763 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3764 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3765 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3766 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4b28252c
TH
3767 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3768 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3769 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3770 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3771 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3772 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
e585f236 3773 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
0345e186 3774
d6b4991a 3775 return (features & feature) == feature;
576a30eb
HX
3776}
3777
4d29515f 3778static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
bcd76111 3779{
278b2513 3780 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
21dc3301 3781 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
bcd76111
HX
3782}
3783
8b86a61d 3784static inline bool netif_needs_gso(struct sk_buff *skb,
4d29515f 3785 netdev_features_t features)
7967168c 3786{
fc741216 3787 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
cdbee74c
YZ
3788 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3789 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
7967168c
HX
3790}
3791
82cc1a7a
PWJ
3792static inline void netif_set_gso_max_size(struct net_device *dev,
3793 unsigned int size)
3794{
3795 dev->gso_max_size = size;
3796}
3797
7a7ffbab
WCC
3798static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3799 int pulled_hlen, u16 mac_offset,
3800 int mac_len)
3801{
3802 skb->protocol = protocol;
3803 skb->encapsulation = 1;
3804 skb_push(skb, pulled_hlen);
3805 skb_reset_transport_header(skb);
3806 skb->mac_header = mac_offset;
3807 skb->network_header = skb->mac_header + mac_len;
3808 skb->mac_len = mac_len;
3809}
3810
a6cc0cfa
JF
3811static inline bool netif_is_macvlan(struct net_device *dev)
3812{
3813 return dev->priv_flags & IFF_MACVLAN;
3814}
3815
2f33e7d5
MB
3816static inline bool netif_is_macvlan_port(struct net_device *dev)
3817{
3818 return dev->priv_flags & IFF_MACVLAN_PORT;
3819}
3820
5933fea7
MB
3821static inline bool netif_is_ipvlan(struct net_device *dev)
3822{
3823 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3824}
3825
3826static inline bool netif_is_ipvlan_port(struct net_device *dev)
3827{
3828 return dev->priv_flags & IFF_IPVLAN_MASTER;
3829}
3830
8a7fbfab 3831static inline bool netif_is_bond_master(struct net_device *dev)
3832{
3833 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3834}
3835
4d29515f 3836static inline bool netif_is_bond_slave(struct net_device *dev)
1765a575
JP
3837{
3838 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3839}
3840
3bdc0eba
BG
3841static inline bool netif_supports_nofcs(struct net_device *dev)
3842{
3843 return dev->priv_flags & IFF_SUPP_NOFCS;
3844}
3845
007979ea 3846static inline bool netif_is_l3_master(const struct net_device *dev)
4e3c8992 3847{
007979ea 3848 return dev->priv_flags & IFF_L3MDEV_MASTER;
4e3c8992
DA
3849}
3850
fee6d4c7
DA
3851static inline bool netif_is_l3_slave(const struct net_device *dev)
3852{
3853 return dev->priv_flags & IFF_L3MDEV_SLAVE;
3854}
3855
0894ae3f
JP
3856static inline bool netif_is_bridge_master(const struct net_device *dev)
3857{
3858 return dev->priv_flags & IFF_EBRIDGE;
3859}
3860
28f9ee22
VY
3861static inline bool netif_is_bridge_port(const struct net_device *dev)
3862{
3863 return dev->priv_flags & IFF_BRIDGE_PORT;
3864}
3865
35d4e172
JP
3866static inline bool netif_is_ovs_master(const struct net_device *dev)
3867{
3868 return dev->priv_flags & IFF_OPENVSWITCH;
3869}
3870
02875878
ED
3871/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3872static inline void netif_keep_dst(struct net_device *dev)
3873{
3874 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3875}
3876
505d4f73 3877extern struct pernet_operations __net_initdata loopback_net_ops;
b1b67dd4 3878
571ba423
JP
3879/* Logging, debugging and troubleshooting/diagnostic helpers. */
3880
3881/* netdev_printk helpers, similar to dev_printk */
3882
3883static inline const char *netdev_name(const struct net_device *dev)
3884{
c6f854d5
VF
3885 if (!dev->name[0] || strchr(dev->name, '%'))
3886 return "(unnamed net_device)";
571ba423
JP
3887 return dev->name;
3888}
3889
ccc7f496
VF
3890static inline const char *netdev_reg_state(const struct net_device *dev)
3891{
3892 switch (dev->reg_state) {
3893 case NETREG_UNINITIALIZED: return " (uninitialized)";
3894 case NETREG_REGISTERED: return "";
3895 case NETREG_UNREGISTERING: return " (unregistering)";
3896 case NETREG_UNREGISTERED: return " (unregistered)";
3897 case NETREG_RELEASED: return " (released)";
3898 case NETREG_DUMMY: return " (dummy)";
3899 }
3900
3901 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3902 return " (unknown)";
3903}
3904
f629d208 3905__printf(3, 4)
6ea754eb
JP
3906void netdev_printk(const char *level, const struct net_device *dev,
3907 const char *format, ...);
f629d208 3908__printf(2, 3)
6ea754eb 3909void netdev_emerg(const struct net_device *dev, const char *format, ...);
f629d208 3910__printf(2, 3)
6ea754eb 3911void netdev_alert(const struct net_device *dev, const char *format, ...);
f629d208 3912__printf(2, 3)
6ea754eb 3913void netdev_crit(const struct net_device *dev, const char *format, ...);
f629d208 3914__printf(2, 3)
6ea754eb 3915void netdev_err(const struct net_device *dev, const char *format, ...);
f629d208 3916__printf(2, 3)
6ea754eb 3917void netdev_warn(const struct net_device *dev, const char *format, ...);
f629d208 3918__printf(2, 3)
6ea754eb 3919void netdev_notice(const struct net_device *dev, const char *format, ...);
f629d208 3920__printf(2, 3)
6ea754eb 3921void netdev_info(const struct net_device *dev, const char *format, ...);
571ba423 3922
8909c9ad
VK
3923#define MODULE_ALIAS_NETDEV(device) \
3924 MODULE_ALIAS("netdev-" device)
3925
b558c96f 3926#if defined(CONFIG_DYNAMIC_DEBUG)
571ba423
JP
3927#define netdev_dbg(__dev, format, args...) \
3928do { \
ffa10cb4 3929 dynamic_netdev_dbg(__dev, format, ##args); \
571ba423 3930} while (0)
b558c96f
JC
3931#elif defined(DEBUG)
3932#define netdev_dbg(__dev, format, args...) \
3933 netdev_printk(KERN_DEBUG, __dev, format, ##args)
571ba423
JP
3934#else
3935#define netdev_dbg(__dev, format, args...) \
3936({ \
3937 if (0) \
3938 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
571ba423
JP
3939})
3940#endif
3941
3942#if defined(VERBOSE_DEBUG)
3943#define netdev_vdbg netdev_dbg
3944#else
3945
3946#define netdev_vdbg(dev, format, args...) \
3947({ \
3948 if (0) \
3949 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3950 0; \
3951})
3952#endif
3953
3954/*
3955 * netdev_WARN() acts like dev_printk(), but with the key difference
3956 * of using a WARN/WARN_ON to get the message out, including the
3957 * file/line information and a backtrace.
3958 */
3959#define netdev_WARN(dev, format, args...) \
ccc7f496
VF
3960 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3961 netdev_reg_state(dev), ##args)
571ba423 3962
b3d95c5c
JP
3963/* netif printk helpers, similar to netdev_printk */
3964
3965#define netif_printk(priv, type, level, dev, fmt, args...) \
3966do { \
3967 if (netif_msg_##type(priv)) \
3968 netdev_printk(level, (dev), fmt, ##args); \
3969} while (0)
3970
f45f4321
JP
3971#define netif_level(level, priv, type, dev, fmt, args...) \
3972do { \
3973 if (netif_msg_##type(priv)) \
3974 netdev_##level(dev, fmt, ##args); \
3975} while (0)
3976
b3d95c5c 3977#define netif_emerg(priv, type, dev, fmt, args...) \
f45f4321 3978 netif_level(emerg, priv, type, dev, fmt, ##args)
b3d95c5c 3979#define netif_alert(priv, type, dev, fmt, args...) \
f45f4321 3980 netif_level(alert, priv, type, dev, fmt, ##args)
b3d95c5c 3981#define netif_crit(priv, type, dev, fmt, args...) \
f45f4321 3982 netif_level(crit, priv, type, dev, fmt, ##args)
b3d95c5c 3983#define netif_err(priv, type, dev, fmt, args...) \
f45f4321 3984 netif_level(err, priv, type, dev, fmt, ##args)
b3d95c5c 3985#define netif_warn(priv, type, dev, fmt, args...) \
f45f4321 3986 netif_level(warn, priv, type, dev, fmt, ##args)
b3d95c5c 3987#define netif_notice(priv, type, dev, fmt, args...) \
f45f4321 3988 netif_level(notice, priv, type, dev, fmt, ##args)
b3d95c5c 3989#define netif_info(priv, type, dev, fmt, args...) \
f45f4321 3990 netif_level(info, priv, type, dev, fmt, ##args)
b3d95c5c 3991
0053ea9c 3992#if defined(CONFIG_DYNAMIC_DEBUG)
b3d95c5c
JP
3993#define netif_dbg(priv, type, netdev, format, args...) \
3994do { \
3995 if (netif_msg_##type(priv)) \
b5fb0a03 3996 dynamic_netdev_dbg(netdev, format, ##args); \
b3d95c5c 3997} while (0)
0053ea9c
JP
3998#elif defined(DEBUG)
3999#define netif_dbg(priv, type, dev, format, args...) \
4000 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
b3d95c5c
JP
4001#else
4002#define netif_dbg(priv, type, dev, format, args...) \
4003({ \
4004 if (0) \
4005 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4006 0; \
4007})
4008#endif
4009
4010#if defined(VERBOSE_DEBUG)
bcfcc450 4011#define netif_vdbg netif_dbg
b3d95c5c
JP
4012#else
4013#define netif_vdbg(priv, type, dev, format, args...) \
4014({ \
4015 if (0) \
a4ed89cb 4016 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
b3d95c5c
JP
4017 0; \
4018})
4019#endif
571ba423 4020
900ff8c6
CW
4021/*
4022 * The list of packet types we will receive (as opposed to discard)
4023 * and the routines to invoke.
4024 *
4025 * Why 16. Because with 16 the only overlap we get on a hash of the
4026 * low nibble of the protocol value is RARP/SNAP/X.25.
4027 *
4028 * NOTE: That is no longer true with the addition of VLAN tags. Not
4029 * sure which should go first, but I bet it won't make much
4030 * difference if we are running VLANs. The good news is that
4031 * this protocol won't be in the list unless compiled in, so
4032 * the average user (w/out VLANs) will not be adversely affected.
4033 * --BLG
4034 *
4035 * 0800 IP
4036 * 8100 802.1Q VLAN
4037 * 0001 802.3
4038 * 0002 AX.25
4039 * 0004 802.2
4040 * 8035 RARP
4041 * 0005 SNAP
4042 * 0805 X.25
4043 * 0806 ARP
4044 * 8137 IPX
4045 * 0009 Localtalk
4046 * 86DD IPv6
4047 */
4048#define PTYPE_HASH_SIZE (16)
4049#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4050
385a154c 4051#endif /* _LINUX_NETDEVICE_H */